bump version to 9.2.0-5

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>

Makefile

@@ -17,10 +17,14 @@ all: $(DEBS)
 
 .PHONY: submodule
 submodule:
-	test -f "$(SRCDIR)/configure" || git submodule update --init --recursive
+ifeq ($(shell test -f "$(SRCDIR)/configure" && echo 1 || echo 0), 0)
+	git submodule update --init --recursive
+	cd $(SRCDIR); meson subprojects download
+endif
 
 PC_BIOS_FW_PURGE_LIST_IN = \
 	hppa-firmware.img \
+	hppa-firmware64.img \
 	openbios-ppc \
 	openbios-sparc32 \
 	openbios-sparc64 \
@@ -28,7 +32,8 @@ PC_BIOS_FW_PURGE_LIST_IN = \
 	s390-ccw.img \
 	s390-netboot.img \
 	u-boot.e500 \
-	.*\.dtb \
+	.*[a-zA-Z0-9]\.dtb \
+	.*[a-zA-Z0-9]\.dts \
 	qemu_vga.ndrv \
 	slof.bin \
 	opensbi-riscv.*-generic-fw_dynamic.bin \
@@ -36,16 +41,14 @@ PC_BIOS_FW_PURGE_LIST_IN = \
 BLOB_PURGE_SED_CMDS = $(foreach FILE,$(PC_BIOS_FW_PURGE_LIST_IN),-e "/$(FILE)/d")
 BLOB_PURGE_FILTER = $(foreach FILE,$(PC_BIOS_FW_PURGE_LIST_IN),-e "$(FILE)")
 
-$(BUILDDIR): keycodemapdb | submodule
+$(BUILDDIR): submodule
 	# check if qemu/ was used for a build
 	# if so, please run 'make distclean' in the submodule and try again
 	test ! -f $(SRCDIR)/build/config.status
 	rm -rf $@.tmp $@
 	cp -a $(SRCDIR) $@.tmp
 	cp -a debian $@.tmp/debian
-	rm -rf $@.tmp/ui/keycodemapdb
 	rm -rf $@.tmp/roms/edk2 # packaged separately
-	cp -a keycodemapdb $@.tmp/ui/
 	find $@.tmp/pc-bios -type f | grep $(BLOB_PURGE_FILTER) | xargs rm -f
 	sed -i $(BLOB_PURGE_SED_CMDS) $@.tmp/pc-bios/meson.build
 	echo "git clone git://git.proxmox.com/git/pve-qemu.git\\ngit checkout $(GITVERSION)" > $@.tmp/debian/SOURCE
@@ -55,7 +58,7 @@ $(BUILDDIR): keycodemapdb | submodule
 deb kvm: $(DEBS)
 $(DEB_DBG): $(DEB)
 $(DEB): $(BUILDDIR)
-	cd $(BUILDDIR); dpkg-buildpackage -b -us -uc -j32
+	cd $(BUILDDIR); dpkg-buildpackage -b -us -uc
 	lintian $(DEBS)
 
 sbuild: $(DSC)
@@ -73,17 +76,6 @@ dsc:
 $(DSC): $(ORIG_SRC_TAR) $(BUILDDIR)
 	cd $(BUILDDIR); dpkg-buildpackage -S -us -uc -d
 
-.PHONY: update
-update:
-	cd $(SRCDIR) && git submodule deinit ui/keycodemapdb || true
-	rm -rf $(SRCDIR)/ui/keycodemapdb
-	mkdir $(SRCDIR)/ui/keycodemapdb
-	cd $(SRCDIR) && git submodule update --init ui/keycodemapdb
-	rm -rf keycodemapdb
-	mkdir keycodemapdb
-	cp -R $(SRCDIR)/ui/keycodemapdb/* keycodemapdb/
-	git add keycodemapdb
-
 .PHONY: upload
 upload: UPLOAD_DIST ?= $(DEB_DISTRIBUTION)
 upload: $(DEBS)

debian/changelog

@@ -1,15 +1,310 @@
-pve-qemu-kvm (8.0.2-3+vitastor2) bookworm; urgency=medium
+pve-qemu-kvm (9.2.0-5) bookworm; urgency=medium
 
-  * Improve performance by adding io_uring support
-  * Fix qemu-img deadlocks after iothread fixes
+  * pve backup: backup-access api: simplify bitmap logic
 
- -- Vitaliy Filippov <vitalif@yourcmc.ru>  Tue, 19 Jul 2023 02:07:02 +0300
+ -- Proxmox Support Team <support@proxmox.com>  Fri, 04 Apr 2025 16:15:58 +0200
 
-pve-qemu-kvm (8.0.2-3+vitastor1) bookworm; urgency=medium
+pve-qemu-kvm (9.2.0-4) bookworm; urgency=medium
 
-  * Add Vitastor support
+  * various async snapshot improvements, inclduing using a dedicated IO thread
+    for the state file when doing a live snapshot. That should reduce load on
+    the main thread and for it to get stuck on IO, i.e. same benefits as using
+    a dedicated IO thread for regular drives. This is particularly interesting
+    when the VM state storage is a network storage like NFS. It should also
+    address #6262.
 
- -- Vitaliy Filippov <vitalif@yourcmc.ru>  Sat, 24 Jun 2023 00:50:32 +0300
+  * pve backup: implement basic features and API in preperation for external
+    backup provider storage plugins.
+
+ -- Proxmox Support Team <support@proxmox.com>  Thu, 03 Apr 2025 17:00:34 +0200
+
+pve-qemu-kvm (9.2.0-3) bookworm; urgency=medium
+
+  * revert changes to the High Precision Event Timer (HPET) to fix performance
+    regression
+
+ -- Proxmox Support Team <support@proxmox.com>  Wed, 26 Mar 2025 09:56:01 +0100
+
+pve-qemu-kvm (9.2.0-2) bookworm; urgency=medium
+
+  * fix assertion failure when migrating a VM with multiple disks on a
+    replicated ZFS.
+
+ -- Proxmox Support Team <support@proxmox.com>  Mon, 24 Feb 2025 17:33:34 +0100
+
+pve-qemu-kvm (9.2.0-1) bookworm; urgency=medium
+
+  * update submodule and patches to QEMU 9.2.0
+
+ -- Proxmox Support Team <support@proxmox.com>  Tue, 04 Feb 2025 08:49:20 +0100
+
+pve-qemu-kvm (9.1.2-3) bookworm; urgency=medium
+
+  * async snapshot: explicitly specify raw format when loading the VM state
+    file
+
+  * vma create: rework CLI parameters for passing disk to a more structured
+    style and use that to allow explicitly specifying the format
+
+ -- Proxmox Support Team <support@proxmox.com>  Fri, 24 Jan 2025 16:12:34 +0100
+
+pve-qemu-kvm (9.1.2-2) bookworm; urgency=medium
+
+  * adapt machine version deprecation for Proxmox VE release and support
+    cycle.
+
+ -- Proxmox Support Team <support@proxmox.com>  Fri, 17 Jan 2025 16:34:06 +0100
+
+pve-qemu-kvm (9.1.2-1) bookworm; urgency=medium
+
+  * update submodule and patches to QEMU 9.1.2
+
+  * improve error handling and edge cases with fleecing backups.
+
+ -- Proxmox Support Team <support@proxmox.com>  Wed, 11 Dec 2024 16:47:21 +0100
+
+pve-qemu-kvm (9.0.2-4) bookworm; urgency=medium
+
+  * async snapshot: ensure any dynamic vCPU-throttling applied for
+    auto-converge gets always disabled again after finishing the snapshot.
+
+ -- Proxmox Support Team <support@proxmox.com>  Sun, 10 Nov 2024 11:23:09 +0100
+
+pve-qemu-kvm (9.0.2-3) bookworm; urgency=medium
+
+  * pick up fix for VirtIO PCI regressions
+
+  * pick up stable fixes for 9.0, including fixes for VirtIO-net, ARM and
+    x86(_64) emulation, CVEs to harden NBD server against malicious clients,
+    as well as a few others (VNC, physmem, Intel IOMMU, ...).
+
+ -- Proxmox Support Team <support@proxmox.com>  Fri, 06 Sep 2024 16:21:42 +0200
+
+pve-qemu-kvm (9.0.2-2) bookworm; urgency=medium
+
+  * actually update submodule to QEMU 9.0.2. The previous release was still
+    based on 9.0.0 by mistake.
+
+ -- Proxmox Support Team <support@proxmox.com>  Wed, 07 Aug 2024 10:16:01 +0200
+
+pve-qemu-kvm (9.0.2-1) bookworm; urgency=medium
+
+  * update submodule and patches to QEMU 9.0.2. While our version had most
+    stable fixes included already, there are new fixes for VirtIO and VGA
+    display screen blanking (#4786)
+
+  * backport fix for a regression with the LSI-53c895a controller and one for
+    the boot order getting ignored for USB storage
+
+ -- Proxmox Support Team <support@proxmox.com>  Mon, 29 Jul 2024 18:59:40 +0200
+
+pve-qemu-kvm (9.0.0-6) bookworm; urgency=medium
+
+  * fix a regression in the zeroinit block driver that prevented importing and
+    cloning disks to RBD storages which are not using the krbd setting
+
+ -- Proxmox Support Team <support@proxmox.com>  Mon, 08 Jul 2024 16:11:15 +0200
+
+pve-qemu-kvm (9.0.0-5) bookworm; urgency=medium
+
+  * backport fix for CVE-2024-4467 to prevent malicious qcow2 image files from
+    already causing bad effects if being queried via 'qemu-img info'. For
+    Proxmox VE, this is an additional safe guard, as currently it directly
+    creates and manages the qcow2 images used by VMs and does not allow
+    unprivileged users to import them
+
+  * fix #4726: code cleanup: avoid superfluous check in vma backup code
+
+ -- Proxmox Support Team <support@proxmox.com>  Wed, 03 Jul 2024 13:13:35 +0200
+
+pve-qemu-kvm (9.0.0-4) bookworm; urgency=medium
+
+  * fix crash after saving a snapshot without including VM state when a VirtIO
+    block device with iothread is configured.
+
+  * fix edge case in error handling when opening a block device from PBS fails
+
+  * minor code cleanup in backup code
+
+ -- Proxmox Support Team <support@proxmox.com>  Mon, 01 Jul 2024 11:26:11 +0200
+
+pve-qemu-kvm (9.0.0-3) bookworm; urgency=medium
+
+  * fix crash when doing resize after hotplugging a disk using io_uring
+
+  * fix some minor issues in software CPU emulation (i.e. non-KVM) for ARM and
+    x86(_64)
+
+ -- Proxmox Support Team <support@proxmox.com>  Wed, 29 May 2024 15:55:44 +0200
+
+pve-qemu-kvm (9.0.0-2) bookworm; urgency=medium
+
+  * fix #5409: backup: fix copy-before-write timeout
+
+  * backup: improve error when copy-before-write fails for fleecing
+
+  * fix forwards and backwards migration with VirtIO-GPU display
+
+  * fix a regression in pflash device introduced in 8.2
+
+  * revert a commit for VirtIO PCI devices that turned out to cause more
+    potential security issues than what it fixed
+
+  * move compatibility flags for a new VirtIO-net feature to the correct
+    machine type. The feature was introduced in QEMU 8.2, but the
+    compatibility flags got added to machine version 8.0 instead of 8.1. This
+    breaks backwards migration with machine version 8.1 from a 8.2/9.0 binary
+    to an 8.1 binary, in cases where the guest kernel enables the feature
+    (e.g. Ubuntu 23.10).
+    While that breaks migration with machine version 8.1 from an unpatched to
+    a patched binary, Proxmox VE only ever had 8.2 on the test repository and
+    9.0 not yet in any public repository.
+
+ -- Proxmox Support Team <support@proxmox.com>  Fri, 17 May 2024 17:04:52 +0200
+
+pve-qemu-kvm (9.0.0-1) bookworm; urgency=medium
+
+  * update submodule and patches to QEMU 9.0.0
+
+ -- Proxmox Support Team <support@proxmox.com>  Mon, 29 Apr 2024 10:51:37 +0200
+
+pve-qemu-kvm (8.2.2-1) bookworm; urgency=medium
+
+  * update submodule and patches to QEMU 8.2.2
+
+ -- Proxmox Support Team <support@proxmox.com>  Sat, 27 Apr 2024 12:44:30 +0200
+
+pve-qemu-kvm (8.1.5-5) bookworm; urgency=medium
+
+  * implement support for backup fleecing
+
+ -- Proxmox Support Team <support@proxmox.com>  Thu, 11 Apr 2024 17:46:48 +0200
+
+pve-qemu-kvm (8.1.5-4) bookworm; urgency=medium
+
+  * fix live-import for certain kinds of VMDK images that rely on padding
+
+  * backup: avoid bubbling up first error if it's an ECANCELED one, as those
+    are often a result of cancling the job due to running into an actual
+    issue.
+
+  * backup: factor out & clean up gathering device info into helper
+
+ -- Proxmox Support Team <support@proxmox.com>  Tue, 12 Mar 2024 14:08:40 +0100
+
+pve-qemu-kvm (8.1.5-3) bookworm; urgency=medium
+
+  * backport fix for potential deadlock during QMP stop command if the VM has
+    disks attached through VirtIO-Block and IO-Thread enabled
+
+  * fix #4507: add patch to automatically increase NOFILE soft limit
+
+ -- Proxmox Support Team <support@proxmox.com>  Wed, 21 Feb 2024 20:11:23 +0100
+
+pve-qemu-kvm (8.1.5-2) bookworm; urgency=medium
+
+  * work around for a situation where guest IO might get stuck, if the VM is
+    configure  with iothread and VirtIO block/SCSI
+
+ -- Proxmox Support Team <support@proxmox.com>  Fri, 02 Feb 2024 19:41:27 +0100
+
+pve-qemu-kvm (8.1.5-1) bookworm; urgency=medium
+
+  * update to 8.1.5 stable release, including more relevant fixes like:
+    - virtio-net: correctly copy vnet header when flushing TX
+    - hw/pflash: implement update buffer for block writes
+    - Fixes to i386 emulation and ARM emulation.
+
+ -- Proxmox Support Team <support@proxmox.com>  Fri, 02 Feb 2024 19:08:13 +0100
+
+pve-qemu-kvm (8.1.2-6) bookworm; urgency=medium
+
+  * revert attempted fix to avoid rare issue with stuck guest IO when using
+    iothread, because it caused a much more common issue with iothreads
+    consuming too much CPU
+
+ -- Proxmox Support Team <support@proxmox.com>  Fri, 15 Dec 2023 14:22:06 +0100
+
+pve-qemu-kvm (8.1.2-5) bookworm; urgency=medium
+
+  * backport workaround for stuck guest IO with iothread and VirtIO block/SCSI
+    in some rare edge cases
+
+  * backport fix for potential deadlock when issuing the "resize" QMP command
+    for a disk that is using iothread
+
+ -- Proxmox Support Team <support@proxmox.com>  Mon, 11 Dec 2023 16:58:27 +0100
+
+pve-qemu-kvm (8.1.2-4) bookworm; urgency=medium
+
+  * fix vnc clipboard in the host to guest direction
+
+ -- Proxmox Support Team <support@proxmox.com>  Wed, 22 Nov 2023 14:28:21 +0100
+
+pve-qemu-kvm (8.1.2-3) bookworm; urgency=medium
+
+  * fix #5054: backport fix for software reset with SATA, avoiding breakage
+    with, e.g., some FreeBSD VMs
+
+ -- Proxmox Support Team <support@proxmox.com>  Mon, 20 Nov 2023 10:24:50 +0100
+
+pve-qemu-kvm (8.1.2-2) bookworm; urgency=medium
+
+  * revert "x86: acpi: workaround Windows not handling name references in
+    Package properly" as that seems to have broken networking (and possibly
+    other things) one some localized variants of Windows (e.g., the German
+    versions).
+
+ -- Proxmox Support Team <support@proxmox.com>  Fri, 17 Nov 2023 11:55:23 +0100
+
+pve-qemu-kvm (8.1.2-1) bookworm; urgency=medium
+
+  * update submodule and patches to QEMU 8.1.2
+
+  * use QEMU's keycode-map-db again instead of our static copy from QEMU 6.0
+
+  * disable graph locking, newly introduced in the 8.1 release, as it has
+    still various deadlock issuess, e.g., during canceling backup jobs.
+
+ -- Proxmox Support Team <support@proxmox.com>  Tue, 24 Oct 2023 13:42:45 +0200
+
+pve-qemu-kvm (8.0.2-7) bookworm; urgency=medium
+
+  * fix #2874: SATA: avoid unsolicited write to sector 0 during reset
+
+ -- Proxmox Support Team <support@proxmox.com>  Wed, 04 Oct 2023 08:33:35 +0200
+
+pve-qemu-kvm (8.0.2-6) bookworm; urgency=medium
+
+  * fix #1534: vma: add extract-filter for disk images allowing users to pass
+    a comma separated list of the disks they want to extract from an archive.
+
+  * backup: create jobs in a drained section to avoid subtle bugs where
+    something interferes with the block-copy-state bitmap on initialization
+
+  * backup: drop experimental, and since a while also fully broken, directory
+    backup format (BACKUP_FORMAT_DIR). This format was never exposed via the
+    Proxmox VE API, but only available via QMP, as its broken since QEMU 8 and
+    we got zero reports about that, it's safe to assume that there are no
+    public users, so just remove it completely.
+
+ -- Proxmox Support Team <support@proxmox.com>  Wed, 06 Sep 2023 17:03:59 +0200
+
+pve-qemu-kvm (8.0.2-5) bookworm; urgency=medium
+
+  * improve memory footprint after backup by not keeping as much memory
+    resident.
+
+  * fix file descriptor leak for vhost (used by default by vNICs).
+
+ -- Proxmox Support Team <support@proxmox.com>  Wed, 16 Aug 2023 11:52:24 +0200
+
+pve-qemu-kvm (8.0.2-4) bookworm; urgency=medium
+
+  * fix resume for snapshot and hibernate in combination with iothread and
+    dirty bitmap
+
+ -- Proxmox Support Team <support@proxmox.com>  Fri, 28 Jul 2023 12:58:22 +0200
 
 pve-qemu-kvm (8.0.2-3) bookworm; urgency=medium
 

debian/control

@@ -37,6 +37,7 @@ Build-Depends: debhelper-compat (= 13),
                python3-minimal,
                python3-sphinx,
                python3-sphinx-rtd-theme,
+               python3-venv,
                quilt,
                uuid-dev,
                xfslibs-dev,
@@ -58,7 +59,6 @@ Depends: ceph-common (>= 0.48),
          libspice-server1 (>= 0.14.0~),
          libusb-1.0-0 (>= 1.0.17-1),
          libusbredirparser1 (>= 0.6-2),
-         vitastor-client (>= 0.9.4),
          libuuid1,
          ${misc:Depends},
          ${shlibs:Depends},
@@ -80,6 +80,7 @@ Replaces: pve-kvm,
           qemu-system-arm,
           qemu-system-x86,
           qemu-utils,
+Breaks: qemu-server (<= 8.0.6)
 Description: Full virtualization on x86 hardware
  Using KVM, one can run multiple virtual PCs, each running unmodified Linux or
  Windows images. Each virtual machine has private virtualized hardware: a

debian/patches/bitmap-mirror/0001-drive-mirror-add-support-for-sync-bitmap-mode-never.patch

@@ -27,18 +27,18 @@ Signed-off-by: Ma Haocong <mahaocong@didichuxing.com>
 Signed-off-by: John Snow <jsnow@redhat.com>
 Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
 Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
-[FE: rebased for 8.0]
+[FE: rebased for 9.1.2]
 Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
 ---
- block/mirror.c                         | 98 +++++++++++++++++++++-----
+ block/mirror.c                         | 99 ++++++++++++++++++++------
  blockdev.c                             | 38 +++++++++-
  include/block/block_int-global-state.h |  4 +-
- qapi/block-core.json                   | 29 ++++++--
+ qapi/block-core.json                   | 25 ++++++-
  tests/unit/test-block-iothread.c       |  4 +-
- 5 files changed, 144 insertions(+), 29 deletions(-)
+ 5 files changed, 142 insertions(+), 28 deletions(-)
 
 diff --git a/block/mirror.c b/block/mirror.c
-index 663e2b7002..9099c75992 100644
+index 2afe700b4d..c3d4be9b15 100644
 --- a/block/mirror.c
 +++ b/block/mirror.c
 @@ -51,7 +51,7 @@ typedef struct MirrorBlockJob {
@@ -50,7 +50,7 @@ index 663e2b7002..9099c75992 100644
      BlockMirrorBackingMode backing_mode;
      /* Whether the target image requires explicit zero-initialization */
      bool zero_target;
-@@ -65,6 +65,8 @@ typedef struct MirrorBlockJob {
+@@ -73,6 +73,8 @@ typedef struct MirrorBlockJob {
      size_t buf_size;
      int64_t bdev_length;
      unsigned long *cow_bitmap;
@@ -59,9 +59,9 @@ index 663e2b7002..9099c75992 100644
      BdrvDirtyBitmap *dirty_bitmap;
      BdrvDirtyBitmapIter *dbi;
      uint8_t *buf;
-@@ -703,7 +705,8 @@ static int mirror_exit_common(Job *job)
-     bdrv_child_refresh_perms(mirror_top_bs, mirror_top_bs->backing,
+@@ -723,7 +725,8 @@ static int mirror_exit_common(Job *job)
                               &error_abort);
+ 
      if (!abort && s->backing_mode == MIRROR_SOURCE_BACKING_CHAIN) {
 -        BlockDriverState *backing = s->is_none_mode ? src : s->base;
 +        BlockDriverState *backing;
@@ -69,7 +69,7 @@ index 663e2b7002..9099c75992 100644
          BlockDriverState *unfiltered_target = bdrv_skip_filters(target_bs);
  
          if (bdrv_cow_bs(unfiltered_target) != backing) {
-@@ -801,6 +804,16 @@ static void mirror_abort(Job *job)
+@@ -824,6 +827,16 @@ static void mirror_abort(Job *job)
      assert(ret == 0);
  }
  
@@ -86,7 +86,7 @@ index 663e2b7002..9099c75992 100644
  static void coroutine_fn mirror_throttle(MirrorBlockJob *s)
  {
      int64_t now = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
-@@ -987,7 +1000,8 @@ static int coroutine_fn mirror_run(Job *job, Error **errp)
+@@ -1020,7 +1033,8 @@ static int coroutine_fn mirror_run(Job *job, Error **errp)
      mirror_free_init(s);
  
      s->last_pause_ns = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
@@ -96,7 +96,7 @@ index 663e2b7002..9099c75992 100644
          ret = mirror_dirty_init(s);
          if (ret < 0 || job_is_cancelled(&s->common.job)) {
              goto immediate_exit;
-@@ -1240,6 +1254,7 @@ static const BlockJobDriver mirror_job_driver = {
+@@ -1309,6 +1323,7 @@ static const BlockJobDriver mirror_job_driver = {
          .run                    = mirror_run,
          .prepare                = mirror_prepare,
          .abort                  = mirror_abort,
@@ -104,7 +104,7 @@ index 663e2b7002..9099c75992 100644
          .pause                  = mirror_pause,
          .complete               = mirror_complete,
          .cancel                 = mirror_cancel,
-@@ -1256,6 +1271,7 @@ static const BlockJobDriver commit_active_job_driver = {
+@@ -1327,6 +1342,7 @@ static const BlockJobDriver commit_active_job_driver = {
          .run                    = mirror_run,
          .prepare                = mirror_prepare,
          .abort                  = mirror_abort,
@@ -112,7 +112,7 @@ index 663e2b7002..9099c75992 100644
          .pause                  = mirror_pause,
          .complete               = mirror_complete,
          .cancel                 = commit_active_cancel,
-@@ -1647,7 +1663,10 @@ static BlockJob *mirror_start_job(
+@@ -1719,7 +1735,10 @@ static BlockJob *mirror_start_job(
                               BlockCompletionFunc *cb,
                               void *opaque,
                               const BlockJobDriver *driver,
@@ -123,10 +123,10 @@ index 663e2b7002..9099c75992 100644
 +                             BlockDriverState *base,
                               bool auto_complete, const char *filter_node_name,
                               bool is_mirror, MirrorCopyMode copy_mode,
-                              Error **errp)
-@@ -1659,10 +1678,39 @@ static BlockJob *mirror_start_job(
-     uint64_t target_perms, target_shared_perms;
-     int ret;
+                              bool base_ro,
+@@ -1734,10 +1753,39 @@ static BlockJob *mirror_start_job(
+ 
+     GLOBAL_STATE_CODE();
  
 -    if (granularity == 0) {
 -        granularity = bdrv_get_default_bitmap_granularity(target);
@@ -166,7 +166,7 @@ index 663e2b7002..9099c75992 100644
      assert(is_power_of_2(granularity));
  
      if (buf_size < 0) {
-@@ -1793,7 +1841,9 @@ static BlockJob *mirror_start_job(
+@@ -1878,7 +1926,9 @@ static BlockJob *mirror_start_job(
      s->replaces = g_strdup(replaces);
      s->on_source_error = on_source_error;
      s->on_target_error = on_target_error;
@@ -176,10 +176,10 @@ index 663e2b7002..9099c75992 100644
 +    s->bitmap_mode = bitmap_mode;
      s->backing_mode = backing_mode;
      s->zero_target = zero_target;
-     s->copy_mode = copy_mode;
-@@ -1814,6 +1864,18 @@ static BlockJob *mirror_start_job(
-         bdrv_disable_dirty_bitmap(s->dirty_bitmap);
-     }
+     qatomic_set(&s->copy_mode, copy_mode);
+@@ -1904,6 +1954,18 @@ static BlockJob *mirror_start_job(
+      */
+     bdrv_disable_dirty_bitmap(s->dirty_bitmap);
  
 +    if (s->sync_bitmap) {
 +        bdrv_dirty_bitmap_set_busy(s->sync_bitmap, true);
@@ -193,10 +193,10 @@ index 663e2b7002..9099c75992 100644
 +        }
 +    }
 +
+     bdrv_graph_wrlock();
      ret = block_job_add_bdrv(&s->common, "source", bs, 0,
                               BLK_PERM_WRITE_UNCHANGED | BLK_PERM_WRITE |
-                              BLK_PERM_CONSISTENT_READ,
-@@ -1891,6 +1953,9 @@ fail:
+@@ -1986,6 +2048,9 @@ fail:
          if (s->dirty_bitmap) {
              bdrv_release_dirty_bitmap(s->dirty_bitmap);
          }
@@ -206,7 +206,7 @@ index 663e2b7002..9099c75992 100644
          job_early_fail(&s->common.job);
      }
  
-@@ -1908,31 +1973,25 @@ void mirror_start(const char *job_id, BlockDriverState *bs,
+@@ -2008,35 +2073,28 @@ void mirror_start(const char *job_id, BlockDriverState *bs,
                    BlockDriverState *target, const char *replaces,
                    int creation_flags, int64_t speed,
                    uint32_t granularity, int64_t buf_size,
@@ -231,19 +231,23 @@ index 663e2b7002..9099c75992 100644
 -                   MirrorSyncMode_str(mode));
 -        return;
 -    }
+-
+     bdrv_graph_rdlock_main_loop();
 -    is_none_mode = mode == MIRROR_SYNC_MODE_NONE;
      base = mode == MIRROR_SYNC_MODE_TOP ? bdrv_backing_chain_next(bs) : NULL;
+     bdrv_graph_rdunlock_main_loop();
+ 
      mirror_start_job(job_id, bs, creation_flags, target, replaces,
                       speed, granularity, buf_size, backing_mode, zero_target,
                       on_source_error, on_target_error, unmap, NULL, NULL,
 -                     &mirror_job_driver, is_none_mode, base, false,
--                     filter_node_name, true, copy_mode, errp);
+-                     filter_node_name, true, copy_mode, false, errp);
 +                     &mirror_job_driver, mode, bitmap, bitmap_mode, base,
-+                     false, filter_node_name, true, copy_mode, errp);
++                     false, filter_node_name, true, copy_mode, false, errp);
  }
  
  BlockJob *commit_active_start(const char *job_id, BlockDriverState *bs,
-@@ -1959,7 +2018,8 @@ BlockJob *commit_active_start(const char *job_id, BlockDriverState *bs,
+@@ -2063,7 +2121,8 @@ BlockJob *commit_active_start(const char *job_id, BlockDriverState *bs,
                       job_id, bs, creation_flags, base, NULL, speed, 0, 0,
                       MIRROR_LEAVE_BACKING_CHAIN, false,
                       on_error, on_error, true, cb, opaque,
@@ -251,13 +255,13 @@ index 663e2b7002..9099c75992 100644
 +                     &commit_active_job_driver, MIRROR_SYNC_MODE_FULL,
 +                     NULL, 0, base, auto_complete,
                       filter_node_name, false, MIRROR_COPY_MODE_BACKGROUND,
-                      errp);
+                      base_read_only, errp);
      if (!job) {
 diff --git a/blockdev.c b/blockdev.c
-index e464daea58..50e4a9c682 100644
+index 6740663fda..38fa63155c 100644
 --- a/blockdev.c
 +++ b/blockdev.c
-@@ -2942,6 +2942,9 @@ static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
+@@ -2781,6 +2781,9 @@ static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
                                     BlockDriverState *target,
                                     const char *replaces,
                                     enum MirrorSyncMode sync,
@@ -267,15 +271,15 @@ index e464daea58..50e4a9c682 100644
                                     BlockMirrorBackingMode backing_mode,
                                     bool zero_target,
                                     bool has_speed, int64_t speed,
-@@ -2960,6 +2963,7 @@ static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
+@@ -2799,6 +2802,7 @@ static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
  {
      BlockDriverState *unfiltered_bs;
      int job_flags = JOB_DEFAULT;
 +    BdrvDirtyBitmap *bitmap = NULL;
  
-     if (!has_speed) {
-         speed = 0;
-@@ -3011,6 +3015,29 @@ static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
+     GLOBAL_STATE_CODE();
+     GRAPH_RDLOCK_GUARD_MAINLOOP();
+@@ -2853,6 +2857,29 @@ static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
          sync = MIRROR_SYNC_MODE_FULL;
      }
  
@@ -305,7 +309,7 @@ index e464daea58..50e4a9c682 100644
      if (!replaces) {
          /* We want to mirror from @bs, but keep implicit filters on top */
          unfiltered_bs = bdrv_skip_implicit_filters(bs);
-@@ -3056,8 +3083,8 @@ static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
+@@ -2894,8 +2921,8 @@ static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
       * and will allow to check whether the node still exist at mirror completion
       */
      mirror_start(job_id, bs, target,
@@ -316,7 +320,7 @@ index e464daea58..50e4a9c682 100644
                   on_source_error, on_target_error, unmap, filter_node_name,
                   copy_mode, errp);
  }
-@@ -3202,6 +3229,8 @@ void qmp_drive_mirror(DriveMirror *arg, Error **errp)
+@@ -3039,6 +3066,8 @@ void qmp_drive_mirror(DriveMirror *arg, Error **errp)
  
      blockdev_mirror_common(arg->job_id, bs, target_bs,
                             arg->replaces, arg->sync,
@@ -325,7 +329,7 @@ index e464daea58..50e4a9c682 100644
                             backing_mode, zero_target,
                             arg->has_speed, arg->speed,
                             arg->has_granularity, arg->granularity,
-@@ -3223,6 +3252,8 @@ void qmp_blockdev_mirror(const char *job_id,
+@@ -3058,6 +3087,8 @@ void qmp_blockdev_mirror(const char *job_id,
                           const char *device, const char *target,
                           const char *replaces,
                           MirrorSyncMode sync,
@@ -334,7 +338,7 @@ index e464daea58..50e4a9c682 100644
                           bool has_speed, int64_t speed,
                           bool has_granularity, uint32_t granularity,
                           bool has_buf_size, int64_t buf_size,
-@@ -3271,7 +3302,8 @@ void qmp_blockdev_mirror(const char *job_id,
+@@ -3098,7 +3129,8 @@ void qmp_blockdev_mirror(const char *job_id,
      }
  
      blockdev_mirror_common(job_id, bs, target_bs,
@@ -345,10 +349,10 @@ index e464daea58..50e4a9c682 100644
                             has_granularity, granularity,
                             has_buf_size, buf_size,
 diff --git a/include/block/block_int-global-state.h b/include/block/block_int-global-state.h
-index 902406eb99..d559be928c 100644
+index eb2d92a226..f0c642b194 100644
 --- a/include/block/block_int-global-state.h
 +++ b/include/block/block_int-global-state.h
-@@ -152,7 +152,9 @@ void mirror_start(const char *job_id, BlockDriverState *bs,
+@@ -158,7 +158,9 @@ void mirror_start(const char *job_id, BlockDriverState *bs,
                    BlockDriverState *target, const char *replaces,
                    int creation_flags, int64_t speed,
                    uint32_t granularity, int64_t buf_size,
@@ -360,31 +364,26 @@ index 902406eb99..d559be928c 100644
                    BlockdevOnError on_source_error,
                    BlockdevOnError on_target_error,
 diff --git a/qapi/block-core.json b/qapi/block-core.json
-index c05ad0c07e..3c945c1f93 100644
+index fd3bcc1c17..48ba32049f 100644
 --- a/qapi/block-core.json
 +++ b/qapi/block-core.json
-@@ -2095,10 +2095,19 @@
- #        (all the disk, only the sectors allocated in the topmost image, or
- #        only new I/O).
+@@ -2178,6 +2178,15 @@
+ #     destination (all the disk, only the sectors allocated in the
+ #     topmost image, or only new I/O).
  #
-+# @bitmap: The name of a bitmap to use for sync=bitmap mode. This argument must
-+#          be present for bitmap mode and absent otherwise. The bitmap's
-+#          granularity is used instead of @granularity (since 4.1).
++# @bitmap: The name of a bitmap to use for sync=bitmap mode. This
++#     argument must be present for bitmap mode and absent otherwise.
++#     The bitmap's granularity is used instead of @granularity (Since
++#     4.1).
 +#
-+# @bitmap-mode: Specifies the type of data the bitmap should contain after
-+#               the operation concludes. Must be present if sync is "bitmap".
-+#               Must NOT be present otherwise. (Since 4.1)
++# @bitmap-mode: Specifies the type of data the bitmap should contain
++#     after the operation concludes. Must be present if sync is
++#     "bitmap". Must NOT be present otherwise. (Since 4.1)
 +#
- # @granularity: granularity of the dirty bitmap, default is 64K
- #               if the image format doesn't have clusters, 4K if the clusters
- #               are smaller than that, else the cluster size.  Must be a
--#               power of 2 between 512 and 64M (since 1.4).
-+#               power of 2 between 512 and 64M. Must not be specified if
-+#               @bitmap is present (since 1.4).
- #
- # @buf-size: maximum amount of data in flight from source to
- #            target (since 1.4).
-@@ -2138,7 +2147,9 @@
+ # @granularity: granularity of the dirty bitmap, default is 64K if the
+ #     image format doesn't have clusters, 4K if the clusters are
+ #     smaller than that, else the cluster size.  Must be a power of 2
+@@ -2220,7 +2229,9 @@
  { 'struct': 'DriveMirror',
    'data': { '*job-id': 'str', 'device': 'str', 'target': 'str',
              '*format': 'str', '*node-name': 'str', '*replaces': 'str',
@@ -395,28 +394,23 @@ index c05ad0c07e..3c945c1f93 100644
              '*speed': 'int', '*granularity': 'uint32',
              '*buf-size': 'int', '*on-source-error': 'BlockdevOnError',
              '*on-target-error': 'BlockdevOnError',
-@@ -2417,10 +2428,19 @@
- #        (all the disk, only the sectors allocated in the topmost image, or
- #        only new I/O).
+@@ -2499,6 +2510,15 @@
+ #     destination (all the disk, only the sectors allocated in the
+ #     topmost image, or only new I/O).
  #
-+# @bitmap: The name of a bitmap to use for sync=bitmap mode. This argument must
-+#          be present for bitmap mode and absent otherwise. The bitmap's
-+#          granularity is used instead of @granularity (since 4.1).
++# @bitmap: The name of a bitmap to use for sync=bitmap mode. This
++#     argument must be present for bitmap mode and absent otherwise.
++#     The bitmap's granularity is used instead of @granularity (since
++#     4.1).
 +#
-+# @bitmap-mode: Specifies the type of data the bitmap should contain after
-+#               the operation concludes. Must be present if sync is "bitmap".
-+#               Must NOT be present otherwise. (Since 4.1)
++# @bitmap-mode: Specifies the type of data the bitmap should contain
++#     after the operation concludes. Must be present if sync is
++#     "bitmap". Must NOT be present otherwise. (Since 4.1)
 +#
- # @granularity: granularity of the dirty bitmap, default is 64K
- #               if the image format doesn't have clusters, 4K if the clusters
- #               are smaller than that, else the cluster size.  Must be a
--#               power of 2 between 512 and 64M
-+#               power of 2 between 512 and 64M . Must not be specified if
-+#               @bitmap is present.
- #
- # @buf-size: maximum amount of data in flight from source to
- #            target
-@@ -2470,7 +2490,8 @@
+ # @granularity: granularity of the dirty bitmap, default is 64K if the
+ #     image format doesn't have clusters, 4K if the clusters are
+ #     smaller than that, else the cluster size.  Must be a power of 2
+@@ -2547,7 +2567,8 @@
  { 'command': 'blockdev-mirror',
    'data': { '*job-id': 'str', 'device': 'str', 'target': 'str',
              '*replaces': 'str',
@@ -427,10 +421,10 @@ index c05ad0c07e..3c945c1f93 100644
              '*buf-size': 'int', '*on-source-error': 'BlockdevOnError',
              '*on-target-error': 'BlockdevOnError',
 diff --git a/tests/unit/test-block-iothread.c b/tests/unit/test-block-iothread.c
-index 3a5e1eb2c4..c1ecc49073 100644
+index 20ed54f570..4f50a99334 100644
 --- a/tests/unit/test-block-iothread.c
 +++ b/tests/unit/test-block-iothread.c
-@@ -757,8 +757,8 @@ static void test_propagate_mirror(void)
+@@ -755,8 +755,8 @@ static void test_propagate_mirror(void)
  
      /* Start a mirror job */
      mirror_start("job0", src, target, NULL, JOB_DEFAULT, 0, 0, 0,
@@ -440,4 +434,4 @@ index 3a5e1eb2c4..c1ecc49073 100644
 +                 false, BLOCKDEV_ON_ERROR_REPORT, BLOCKDEV_ON_ERROR_REPORT,
                   false, "filter_node", MIRROR_COPY_MODE_BACKGROUND,
                   &error_abort);
-     WITH_JOB_LOCK_GUARD() {
+ 

debian/patches/bitmap-mirror/0002-drive-mirror-add-support-for-conditional-and-always-.patch

@@ -24,10 +24,10 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
  1 file changed, 18 insertions(+), 6 deletions(-)
 
 diff --git a/block/mirror.c b/block/mirror.c
-index 9099c75992..e2ff42067b 100644
+index c3d4be9b15..7b6f7c0068 100644
 --- a/block/mirror.c
 +++ b/block/mirror.c
-@@ -680,8 +680,6 @@ static int mirror_exit_common(Job *job)
+@@ -694,8 +694,6 @@ static int mirror_exit_common(Job *job)
          bdrv_unfreeze_backing_chain(mirror_top_bs, target_bs);
      }
  
@@ -36,9 +36,9 @@ index 9099c75992..e2ff42067b 100644
      /* Make sure that the source BDS doesn't go away during bdrv_replace_node,
       * before we can call bdrv_drained_end */
      bdrv_ref(src);
-@@ -782,6 +780,18 @@ static int mirror_exit_common(Job *job)
-     block_job_remove_all_bdrv(bjob);
-     bdrv_replace_node(mirror_top_bs, mirror_top_bs->backing->bs, &error_abort);
+@@ -805,6 +803,18 @@ static int mirror_exit_common(Job *job)
+     bdrv_drained_end(target_bs);
+     bdrv_unref(target_bs);
  
 +    if (s->sync_bitmap) {
 +        if (s->bitmap_mode == BITMAP_SYNC_MODE_ALWAYS ||
@@ -55,7 +55,7 @@ index 9099c75992..e2ff42067b 100644
      bs_opaque->job = NULL;
  
      bdrv_drained_end(src);
-@@ -1688,10 +1698,6 @@ static BlockJob *mirror_start_job(
+@@ -1763,10 +1773,6 @@ static BlockJob *mirror_start_job(
                         " sync mode",
                         MirrorSyncMode_str(sync_mode));
              return NULL;
@@ -66,7 +66,7 @@ index 9099c75992..e2ff42067b 100644
          }
      } else if (bitmap) {
          error_setg(errp,
-@@ -1708,6 +1714,12 @@ static BlockJob *mirror_start_job(
+@@ -1783,6 +1789,12 @@ static BlockJob *mirror_start_job(
              return NULL;
          }
          granularity = bdrv_dirty_bitmap_granularity(bitmap);

debian/patches/bitmap-mirror/0003-mirror-add-check-for-bitmap-mode-without-bitmap.patch

@@ -16,10 +16,10 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
  1 file changed, 3 insertions(+)
 
 diff --git a/blockdev.c b/blockdev.c
-index 50e4a9c682..e6b2b1e338 100644
+index 38fa63155c..204cf6fad1 100644
 --- a/blockdev.c
 +++ b/blockdev.c
-@@ -3036,6 +3036,9 @@ static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
+@@ -2878,6 +2878,9 @@ static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
          if (bdrv_dirty_bitmap_check(bitmap, BDRV_BITMAP_ALLOW_RO, errp)) {
              return;
          }

debian/patches/bitmap-mirror/0004-mirror-switch-to-bdrv_dirty_bitmap_merge_internal.patch

@@ -16,10 +16,10 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
  1 file changed, 4 insertions(+), 7 deletions(-)
 
 diff --git a/block/mirror.c b/block/mirror.c
-index e2ff42067b..f42953837b 100644
+index 7b6f7c0068..2b1c07095d 100644
 --- a/block/mirror.c
 +++ b/block/mirror.c
-@@ -786,8 +786,8 @@ static int mirror_exit_common(Job *job)
+@@ -809,8 +809,8 @@ static int mirror_exit_common(Job *job)
               job->ret == 0 && ret == 0)) {
              /* Success; synchronize copy back to sync. */
              bdrv_clear_dirty_bitmap(s->sync_bitmap, NULL);
@@ -30,7 +30,7 @@ index e2ff42067b..f42953837b 100644
          }
      }
      bdrv_release_dirty_bitmap(s->dirty_bitmap);
-@@ -1881,11 +1881,8 @@ static BlockJob *mirror_start_job(
+@@ -1971,11 +1971,8 @@ static BlockJob *mirror_start_job(
      }
  
      if (s->sync_mode == MIRROR_SYNC_MODE_BITMAP) {
@@ -43,4 +43,4 @@ index e2ff42067b..f42953837b 100644
 +                                         NULL, true);
      }
  
-     ret = block_job_add_bdrv(&s->common, "source", bs, 0,
+     bdrv_graph_wrlock();

debian/patches/bitmap-mirror/0006-mirror-move-some-checks-to-qmp.patch

@@ -12,7 +12,7 @@ uniform w.r.t. backup block jobs.
 
 Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
 Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
-[FE: rebase for 8.0]
+[FE: rebase for 8.2.2]
 Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
 ---
  block/mirror.c             | 28 +++------------
@@ -21,12 +21,12 @@ Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
  3 files changed, 70 insertions(+), 59 deletions(-)
 
 diff --git a/block/mirror.c b/block/mirror.c
-index f42953837b..8f79efaa87 100644
+index 2b1c07095d..f5787b380c 100644
 --- a/block/mirror.c
 +++ b/block/mirror.c
-@@ -1688,31 +1688,13 @@ static BlockJob *mirror_start_job(
-     uint64_t target_perms, target_shared_perms;
-     int ret;
+@@ -1763,31 +1763,13 @@ static BlockJob *mirror_start_job(
+ 
+     GLOBAL_STATE_CODE();
  
 -    if (sync_mode == MIRROR_SYNC_MODE_INCREMENTAL) {
 -        error_setg(errp, "Sync mode '%s' not supported",
@@ -62,10 +62,10 @@ index f42953837b..8f79efaa87 100644
  
          if (bitmap_mode != BITMAP_SYNC_MODE_NEVER) {
 diff --git a/blockdev.c b/blockdev.c
-index e6b2b1e338..bdae211a54 100644
+index 204cf6fad1..79d47b1920 100644
 --- a/blockdev.c
 +++ b/blockdev.c
-@@ -3015,7 +3015,36 @@ static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
+@@ -2857,7 +2857,36 @@ static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
          sync = MIRROR_SYNC_MODE_FULL;
      }
  

debian/patches/extra/0001-monitor-qmp-fix-race-with-clients-disconnecting-earl.patch

@@ -48,7 +48,7 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
  6 files changed, 59 insertions(+), 5 deletions(-)
 
 diff --git a/include/monitor/monitor.h b/include/monitor/monitor.h
-index 033390f699..ad35d4fea8 100644
+index c3740ec616..7f38ce6b8b 100644
 --- a/include/monitor/monitor.h
 +++ b/include/monitor/monitor.h
 @@ -16,6 +16,7 @@ extern QemuOptsList qemu_mon_opts;
@@ -60,10 +60,10 @@ index 033390f699..ad35d4fea8 100644
  void monitor_init_globals(void);
  void monitor_init_globals_core(void);
 diff --git a/monitor/monitor-internal.h b/monitor/monitor-internal.h
-index 53e3808054..a19f8cbc2b 100644
+index cb628f681d..93dbd62fc2 100644
 --- a/monitor/monitor-internal.h
 +++ b/monitor/monitor-internal.h
-@@ -152,6 +152,13 @@ typedef struct {
+@@ -151,6 +151,13 @@ typedef struct {
      QemuMutex qmp_queue_lock;
      /* Input queue that holds all the parsed QMP requests */
      GQueue *qmp_requests;
@@ -78,10 +78,10 @@ index 53e3808054..a19f8cbc2b 100644
  
  /**
 diff --git a/monitor/monitor.c b/monitor/monitor.c
-index 8dc96f6af9..f3c38cb714 100644
+index 56786c0ccc..30071d0c8a 100644
 --- a/monitor/monitor.c
 +++ b/monitor/monitor.c
-@@ -135,6 +135,21 @@ bool monitor_cur_is_qmp(void)
+@@ -116,6 +116,21 @@ bool monitor_cur_is_qmp(void)
      return cur_mon && monitor_is_qmp(cur_mon);
  }
  
@@ -104,10 +104,10 @@ index 8dc96f6af9..f3c38cb714 100644
   * Is @mon is using readline?
   * Note: not all HMP monitors use readline, e.g., gdbserver has a
 diff --git a/monitor/qmp.c b/monitor/qmp.c
-index 092c527b6f..6b8cfcf6d8 100644
+index 5e538f34c0..eb181d5979 100644
 --- a/monitor/qmp.c
 +++ b/monitor/qmp.c
-@@ -141,6 +141,8 @@ static void monitor_qmp_dispatch(MonitorQMP *mon, QObject *req)
+@@ -165,6 +165,8 @@ static void monitor_qmp_dispatch(MonitorQMP *mon, QObject *req)
      QDict *rsp;
      QDict *error;
  
@@ -116,7 +116,7 @@ index 092c527b6f..6b8cfcf6d8 100644
      rsp = qmp_dispatch(mon->commands, req, qmp_oob_enabled(mon),
                         &mon->common);
  
-@@ -156,7 +158,17 @@ static void monitor_qmp_dispatch(MonitorQMP *mon, QObject *req)
+@@ -180,7 +182,17 @@ static void monitor_qmp_dispatch(MonitorQMP *mon, QObject *req)
          }
      }
  
@@ -135,7 +135,7 @@ index 092c527b6f..6b8cfcf6d8 100644
      qobject_unref(rsp);
  }
  
-@@ -444,6 +456,7 @@ static void monitor_qmp_event(void *opaque, QEMUChrEvent event)
+@@ -461,6 +473,7 @@ static void monitor_qmp_event(void *opaque, QEMUChrEvent event)
  
      switch (event) {
      case CHR_EVENT_OPENED:
@@ -144,7 +144,7 @@ index 092c527b6f..6b8cfcf6d8 100644
          monitor_qmp_caps_reset(mon);
          data = qmp_greeting(mon);
 diff --git a/qapi/qmp-dispatch.c b/qapi/qmp-dispatch.c
-index 0990873ec8..e605003771 100644
+index 176b549473..790bb7d1da 100644
 --- a/qapi/qmp-dispatch.c
 +++ b/qapi/qmp-dispatch.c
 @@ -117,16 +117,28 @@ typedef struct QmpDispatchBH {
@@ -180,16 +180,16 @@ index 0990873ec8..e605003771 100644
      aio_co_wake(data->co);
  }
  
-@@ -231,6 +243,7 @@ QDict *qmp_dispatch(const QmpCommandList *cmds, QObject *request,
+@@ -253,6 +265,7 @@ QDict *coroutine_mixed_fn qmp_dispatch(const QmpCommandList *cmds, QObject *requ
              .ret        = &ret,
              .errp       = &err,
              .co         = qemu_coroutine_self(),
 +            .conn_nr    = monitor_get_connection_nr(cur_mon),
          };
-         aio_bh_schedule_oneshot(qemu_get_aio_context(), do_qmp_dispatch_bh,
+         aio_bh_schedule_oneshot(iohandler_get_aio_context(), do_qmp_dispatch_bh,
                                  &data);
 diff --git a/stubs/monitor-core.c b/stubs/monitor-core.c
-index afa477aae6..d3ff124bf3 100644
+index 1894cdfe1f..d74d0459f0 100644
 --- a/stubs/monitor-core.c
 +++ b/stubs/monitor-core.c
 @@ -12,6 +12,11 @@ Monitor *monitor_set_cur(Coroutine *co, Monitor *mon)
@@ -201,6 +201,6 @@ index afa477aae6..d3ff124bf3 100644
 +    return -1;
 +}
 +
- void monitor_init_qmp(Chardev *chr, bool pretty, Error **errp)
+ void qapi_event_emit(QAPIEvent event, QDict *qdict)
  {
  }

debian/patches/extra/0002-ide-avoid-potential-deadlock-when-draining-during-tr.patch

@@ -55,10 +55,10 @@ Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
  1 file changed, 6 insertions(+), 6 deletions(-)
 
 diff --git a/hw/ide/core.c b/hw/ide/core.c
-index 45d14a25e9..08e1f0c3d7 100644
+index 08d9218455..20d8c0cf66 100644
 --- a/hw/ide/core.c
 +++ b/hw/ide/core.c
-@@ -444,7 +444,7 @@ static void ide_trim_bh_cb(void *opaque)
+@@ -456,7 +456,7 @@ static void ide_trim_bh_cb(void *opaque)
      iocb->bh = NULL;
      qemu_aio_unref(iocb);
  
@@ -67,7 +67,7 @@ index 45d14a25e9..08e1f0c3d7 100644
      blk_dec_in_flight(blk);
  }
  
-@@ -504,6 +504,8 @@ static void ide_issue_trim_cb(void *opaque, int ret)
+@@ -516,6 +516,8 @@ static void ide_issue_trim_cb(void *opaque, int ret)
  done:
      iocb->aiocb = NULL;
      if (iocb->bh) {
@@ -76,8 +76,8 @@ index 45d14a25e9..08e1f0c3d7 100644
          replay_bh_schedule_event(iocb->bh);
      }
  }
-@@ -515,9 +517,6 @@ BlockAIOCB *ide_issue_trim(
-     IDEState *s = opaque;
+@@ -528,9 +530,6 @@ BlockAIOCB *ide_issue_trim(
+     IDEDevice *dev = s->unit ? s->bus->slave : s->bus->master;
      TrimAIOCB *iocb;
  
 -    /* Paired with a decrement in ide_trim_bh_cb() */
@@ -85,8 +85,8 @@ index 45d14a25e9..08e1f0c3d7 100644
 -
      iocb = blk_aio_get(&trim_aiocb_info, s->blk, cb, cb_opaque);
      iocb->s = s;
-     iocb->bh = qemu_bh_new(ide_trim_bh_cb, iocb);
-@@ -740,8 +739,9 @@ void ide_cancel_dma_sync(IDEState *s)
+     iocb->bh = qemu_bh_new_guarded(ide_trim_bh_cb, iocb,
+@@ -754,8 +753,9 @@ void ide_cancel_dma_sync(IDEState *s)
       */
      if (s->bus->dma->aiocb) {
          trace_ide_cancel_dma_sync_remaining();

debian/patches/extra/0002-scsi-megasas-Internal-cdbs-have-16-byte-length.patch

@@ -1,69 +0,0 @@
-From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
-From: Guenter Roeck <linux@roeck-us.net>
-Date: Tue, 28 Feb 2023 09:11:29 -0800
-Subject: [PATCH] scsi: megasas: Internal cdbs have 16-byte length
-MIME-Version: 1.0
-Content-Type: text/plain; charset=UTF-8
-Content-Transfer-Encoding: 8bit
-
-Host drivers do not necessarily set cdb_len in megasas io commands.
-With commits 6d1511cea0 ("scsi: Reject commands if the CDB length
-exceeds buf_len") and fe9d8927e2 ("scsi: Add buf_len parameter to
-scsi_req_new()"), this results in failures to boot Linux from affected
-SCSI drives because cdb_len is set to 0 by the host driver.
-Set the cdb length to its actual size to solve the problem.
-
-Signed-off-by: Guenter Roeck <linux@roeck-us.net>
-Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
-(picked-up from https://lists.nongnu.org/archive/html/qemu-devel/2023-02/msg08653.html)
-Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
----
- hw/scsi/megasas.c | 14 ++------------
- 1 file changed, 2 insertions(+), 12 deletions(-)
-
-diff --git a/hw/scsi/megasas.c b/hw/scsi/megasas.c
-index 9cbbb16121..d624866bb6 100644
---- a/hw/scsi/megasas.c
-+++ b/hw/scsi/megasas.c
-@@ -1780,7 +1780,7 @@ static int megasas_handle_io(MegasasState *s, MegasasCmd *cmd, int frame_cmd)
-     uint8_t cdb[16];
-     int len;
-     struct SCSIDevice *sdev = NULL;
--    int target_id, lun_id, cdb_len;
-+    int target_id, lun_id;
- 
-     lba_count = le32_to_cpu(cmd->frame->io.header.data_len);
-     lba_start_lo = le32_to_cpu(cmd->frame->io.lba_lo);
-@@ -1789,7 +1789,6 @@ static int megasas_handle_io(MegasasState *s, MegasasCmd *cmd, int frame_cmd)
- 
-     target_id = cmd->frame->header.target_id;
-     lun_id = cmd->frame->header.lun_id;
--    cdb_len = cmd->frame->header.cdb_len;
- 
-     if (target_id < MFI_MAX_LD && lun_id == 0) {
-         sdev = scsi_device_find(&s->bus, 0, target_id, lun_id);
-@@ -1804,15 +1803,6 @@ static int megasas_handle_io(MegasasState *s, MegasasCmd *cmd, int frame_cmd)
-         return MFI_STAT_DEVICE_NOT_FOUND;
-     }
- 
--    if (cdb_len > 16) {
--        trace_megasas_scsi_invalid_cdb_len(
--            mfi_frame_desc(frame_cmd), 1, target_id, lun_id, cdb_len);
--        megasas_write_sense(cmd, SENSE_CODE(INVALID_OPCODE));
--        cmd->frame->header.scsi_status = CHECK_CONDITION;
--        s->event_count++;
--        return MFI_STAT_SCSI_DONE_WITH_ERROR;
--    }
--
-     cmd->iov_size = lba_count * sdev->blocksize;
-     if (megasas_map_sgl(s, cmd, &cmd->frame->io.sgl)) {
-         megasas_write_sense(cmd, SENSE_CODE(TARGET_FAILURE));
-@@ -1823,7 +1813,7 @@ static int megasas_handle_io(MegasasState *s, MegasasCmd *cmd, int frame_cmd)
- 
-     megasas_encode_lba(cdb, lba_start, lba_count, is_write);
-     cmd->req = scsi_req_new(sdev, cmd->index,
--                            lun_id, cdb, cdb_len, cmd);
-+                            lun_id, cdb, sizeof(cdb), cmd);
-     if (!cmd->req) {
-         trace_megasas_scsi_req_alloc_failed(
-             mfi_frame_desc(frame_cmd), target_id, lun_id);

debian/patches/extra/0003-tcg-Reset-free_temps-before-tcg_optimize.patch

@@ -0,0 +1,82 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Richard Henderson <richard.henderson@linaro.org>
+Date: Sat, 7 Dec 2024 18:14:45 +0000
+Subject: [PATCH] tcg: Reset free_temps before tcg_optimize
+MIME-Version: 1.0
+Content-Type: text/plain; charset=UTF-8
+Content-Transfer-Encoding: 8bit
+
+When allocating new temps during tcg_optmize, do not re-use
+any EBB temps that were used within the TB.  We do not have
+any idea what span of the TB in which the temp was live.
+
+Introduce tcg_temp_ebb_reset_freed and use before tcg_optimize,
+as well as replacing the equivalent in plugin_gen_inject and
+tcg_func_start.
+
+Cc: qemu-stable@nongnu.org
+Fixes: fb04ab7ddd8 ("tcg/optimize: Lower TCG_COND_TST{EQ,NE} if unsupported")
+Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2711
+Reported-by: wannacu <wannacu2049@gmail.com>
+Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
+Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
+Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
+(cherry picked from commit 04e006ab36a8565b92d4e21dd346367fbade7d74)
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ accel/tcg/plugin-gen.c          | 2 +-
+ include/tcg/tcg-temp-internal.h | 6 ++++++
+ tcg/tcg.c                       | 5 ++++-
+ 3 files changed, 11 insertions(+), 2 deletions(-)
+
+diff --git a/accel/tcg/plugin-gen.c b/accel/tcg/plugin-gen.c
+index 0f47bfbb48..1ef075552c 100644
+--- a/accel/tcg/plugin-gen.c
++++ b/accel/tcg/plugin-gen.c
+@@ -275,7 +275,7 @@ static void plugin_gen_inject(struct qemu_plugin_tb *plugin_tb)
+      * that might be live within the existing opcode stream.
+      * The simplest solution is to release them all and create new.
+      */
+-    memset(tcg_ctx->free_temps, 0, sizeof(tcg_ctx->free_temps));
++    tcg_temp_ebb_reset_freed(tcg_ctx);
+ 
+     QTAILQ_FOREACH_SAFE(op, &tcg_ctx->ops, link, next) {
+         switch (op->opc) {
+diff --git a/include/tcg/tcg-temp-internal.h b/include/tcg/tcg-temp-internal.h
+index 44192c55a9..98f91e68b7 100644
+--- a/include/tcg/tcg-temp-internal.h
++++ b/include/tcg/tcg-temp-internal.h
+@@ -42,4 +42,10 @@ TCGv_i64 tcg_temp_ebb_new_i64(void);
+ TCGv_ptr tcg_temp_ebb_new_ptr(void);
+ TCGv_i128 tcg_temp_ebb_new_i128(void);
+ 
++/* Forget all freed EBB temps, so that new allocations produce new temps. */
++static inline void tcg_temp_ebb_reset_freed(TCGContext *s)
++{
++    memset(s->free_temps, 0, sizeof(s->free_temps));
++}
++
+ #endif /* TCG_TEMP_FREE_H */
+diff --git a/tcg/tcg.c b/tcg/tcg.c
+index 0babae1b88..4578b185be 100644
+--- a/tcg/tcg.c
++++ b/tcg/tcg.c
+@@ -1489,7 +1489,7 @@ void tcg_func_start(TCGContext *s)
+     s->nb_temps = s->nb_globals;
+ 
+     /* No temps have been previously allocated for size or locality.  */
+-    memset(s->free_temps, 0, sizeof(s->free_temps));
++    tcg_temp_ebb_reset_freed(s);
+ 
+     /* No constant temps have been previously allocated. */
+     for (int i = 0; i < TCG_TYPE_COUNT; ++i) {
+@@ -6120,6 +6120,9 @@ int tcg_gen_code(TCGContext *s, TranslationBlock *tb, uint64_t pc_start)
+     }
+ #endif
+ 
++    /* Do not reuse any EBB that may be allocated within the TB. */
++    tcg_temp_ebb_reset_freed(s);
++
+     tcg_optimize(s);
+ 
+     reachable_code_pass(s);

debian/patches/extra/0004-target-i386-Reset-TSCs-of-parked-vCPUs-too-on-VM-res.patch

@@ -0,0 +1,149 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: "Maciej S. Szmigiero" <maciej.szmigiero@oracle.com>
+Date: Thu, 12 Dec 2024 15:51:15 +0100
+Subject: [PATCH] target/i386: Reset TSCs of parked vCPUs too on VM reset
+
+Since commit 5286c3662294 ("target/i386: properly reset TSC on reset")
+QEMU writes the special value of "1" to each online vCPU TSC on VM reset
+to reset it.
+
+However parked vCPUs don't get that handling and due to that their TSCs
+get desynchronized when the VM gets reset.
+This in turn causes KVM to turn off PVCLOCK_TSC_STABLE_BIT in its exported
+PV clock.
+Note that KVM has no understanding of vCPU being currently parked.
+
+Without PVCLOCK_TSC_STABLE_BIT the sched clock is marked unstable in
+the guest's kvm_sched_clock_init().
+This causes a performance regressions to show in some tests.
+
+Fix this issue by writing the special value of "1" also to TSCs of parked
+vCPUs on VM reset.
+
+Reproducing the issue:
+1) Boot a VM with "-smp 2,maxcpus=3" or similar
+
+2) device_add host-x86_64-cpu,id=vcpu,node-id=0,socket-id=0,core-id=2,thread-id=0
+
+3) Wait a few seconds
+
+4) device_del vcpu
+
+5) Inside the VM run:
+# echo "t" >/proc/sysrq-trigger; dmesg | grep sched_clock_stable
+Observe the sched_clock_stable() value is 1.
+
+6) Reboot the VM
+
+7) Once the VM boots once again run inside it:
+# echo "t" >/proc/sysrq-trigger; dmesg | grep sched_clock_stable
+Observe the sched_clock_stable() value is now 0.
+
+Fixes: 5286c3662294 ("target/i386: properly reset TSC on reset")
+Signed-off-by: Maciej S. Szmigiero <maciej.szmigiero@oracle.com>
+Link: https://lore.kernel.org/r/5a605a88e9a231386dc803c60f5fed9b48108139.1734014926.git.maciej.szmigiero@oracle.com
+Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
+(cherry picked from commit 3f2a05b31ee9ce2ddb6c75a9bc3f5e7f7af9a76f)
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ accel/kvm/kvm-all.c                | 11 +++++++++++
+ configs/targets/i386-softmmu.mak   |  1 +
+ configs/targets/x86_64-softmmu.mak |  1 +
+ include/sysemu/kvm.h               |  8 ++++++++
+ target/i386/kvm/kvm.c              | 15 +++++++++++++++
+ 5 files changed, 36 insertions(+)
+
+diff --git a/accel/kvm/kvm-all.c b/accel/kvm/kvm-all.c
+index 801cff16a5..dec1d1c16a 100644
+--- a/accel/kvm/kvm-all.c
++++ b/accel/kvm/kvm-all.c
+@@ -437,6 +437,16 @@ int kvm_unpark_vcpu(KVMState *s, unsigned long vcpu_id)
+     return kvm_fd;
+ }
+ 
++static void kvm_reset_parked_vcpus(void *param)
++{
++    KVMState *s = param;
++    struct KVMParkedVcpu *cpu;
++
++    QLIST_FOREACH(cpu, &s->kvm_parked_vcpus, node) {
++        kvm_arch_reset_parked_vcpu(cpu->vcpu_id, cpu->kvm_fd);
++    }
++}
++
+ int kvm_create_vcpu(CPUState *cpu)
+ {
+     unsigned long vcpu_id = kvm_arch_vcpu_id(cpu);
+@@ -2728,6 +2738,7 @@ static int kvm_init(MachineState *ms)
+     }
+ 
+     qemu_register_reset(kvm_unpoison_all, NULL);
++    qemu_register_reset(kvm_reset_parked_vcpus, s);
+ 
+     if (s->kernel_irqchip_allowed) {
+         kvm_irqchip_create(s);
+diff --git a/configs/targets/i386-softmmu.mak b/configs/targets/i386-softmmu.mak
+index 2ac69d5ba3..2eb0e86250 100644
+--- a/configs/targets/i386-softmmu.mak
++++ b/configs/targets/i386-softmmu.mak
+@@ -1,4 +1,5 @@
+ TARGET_ARCH=i386
+ TARGET_SUPPORTS_MTTCG=y
+ TARGET_KVM_HAVE_GUEST_DEBUG=y
++TARGET_KVM_HAVE_RESET_PARKED_VCPU=y
+ TARGET_XML_FILES= gdb-xml/i386-32bit.xml
+diff --git a/configs/targets/x86_64-softmmu.mak b/configs/targets/x86_64-softmmu.mak
+index e12ac3dc59..920e9a4200 100644
+--- a/configs/targets/x86_64-softmmu.mak
++++ b/configs/targets/x86_64-softmmu.mak
+@@ -2,4 +2,5 @@ TARGET_ARCH=x86_64
+ TARGET_BASE_ARCH=i386
+ TARGET_SUPPORTS_MTTCG=y
+ TARGET_KVM_HAVE_GUEST_DEBUG=y
++TARGET_KVM_HAVE_RESET_PARKED_VCPU=y
+ TARGET_XML_FILES= gdb-xml/i386-64bit.xml
+diff --git a/include/sysemu/kvm.h b/include/sysemu/kvm.h
+index c3a60b2890..ab17c09a55 100644
+--- a/include/sysemu/kvm.h
++++ b/include/sysemu/kvm.h
+@@ -377,6 +377,14 @@ int kvm_arch_init(MachineState *ms, KVMState *s);
+ int kvm_arch_init_vcpu(CPUState *cpu);
+ int kvm_arch_destroy_vcpu(CPUState *cpu);
+ 
++#ifdef TARGET_KVM_HAVE_RESET_PARKED_VCPU
++void kvm_arch_reset_parked_vcpu(unsigned long vcpu_id, int kvm_fd);
++#else
++static inline void kvm_arch_reset_parked_vcpu(unsigned long vcpu_id, int kvm_fd)
++{
++}
++#endif
++
+ bool kvm_vcpu_id_is_valid(int vcpu_id);
+ 
+ /* Returns VCPU ID to be used on KVM_CREATE_VCPU ioctl() */
+diff --git a/target/i386/kvm/kvm.c b/target/i386/kvm/kvm.c
+index 8e17942c3b..2ff618fbf1 100644
+--- a/target/i386/kvm/kvm.c
++++ b/target/i386/kvm/kvm.c
+@@ -2415,6 +2415,21 @@ void kvm_arch_after_reset_vcpu(X86CPU *cpu)
+     }
+ }
+ 
++void kvm_arch_reset_parked_vcpu(unsigned long vcpu_id, int kvm_fd)
++{
++    g_autofree struct kvm_msrs *msrs = NULL;
++
++    msrs = g_malloc0(sizeof(*msrs) + sizeof(msrs->entries[0]));
++    msrs->entries[0].index = MSR_IA32_TSC;
++    msrs->entries[0].data = 1; /* match the value in x86_cpu_reset() */
++    msrs->nmsrs++;
++
++    if (ioctl(kvm_fd, KVM_SET_MSRS, msrs) != 1) {
++        warn_report("parked vCPU %lu TSC reset failed: %d",
++                    vcpu_id, errno);
++    }
++}
++
+ void kvm_arch_do_init_vcpu(X86CPU *cpu)
+ {
+     CPUX86State *env = &cpu->env;

debian/patches/extra/0004-ui-return-NULL-when-getting-cursor-without-a-console.patch

@@ -1,36 +0,0 @@
-From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
-From: =?UTF-8?q?Marc-Andr=C3=A9=20Lureau?= <marcandre.lureau@redhat.com>
-Date: Fri, 28 Apr 2023 19:48:06 +0400
-Subject: [PATCH] ui: return NULL when getting cursor without a console
-MIME-Version: 1.0
-Content-Type: text/plain; charset=UTF-8
-Content-Transfer-Encoding: 8bit
-
-VNC may try to get the current cursor even when there are no consoles
-and crashes. Simple reproducer is qemu with -nodefaults.
-
-Fixes: (again)
-https://gitlab.com/qemu-project/qemu/-/issues/1548
-
-Fixes: commit 385ac97f8 ("ui: keep current cursor with QemuConsole")
-Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
-Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
-(picked up from https://lists.nongnu.org/archive/html/qemu-devel/2023-04/msg05598.html)
-Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
----
- ui/console.c | 2 +-
- 1 file changed, 1 insertion(+), 1 deletion(-)
-
-diff --git a/ui/console.c b/ui/console.c
-index e173731e20..7461446e71 100644
---- a/ui/console.c
-+++ b/ui/console.c
-@@ -2306,7 +2306,7 @@ QEMUCursor *qemu_console_get_cursor(QemuConsole *con)
-     if (con == NULL) {
-         con = active_console;
-     }
--    return con->cursor;
-+    return con ? con->cursor : NULL;
- }
- 
- bool qemu_console_is_visible(QemuConsole *con)

debian/patches/extra/0005-i386-cpu-Mark-avx10_version-filtered-when-prefix-is-.patch

@@ -0,0 +1,41 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Zhao Liu <zhao1.liu@intel.com>
+Date: Wed, 6 Nov 2024 11:07:18 +0800
+Subject: [PATCH] i386/cpu: Mark avx10_version filtered when prefix is NULL
+
+In x86_cpu_filter_features(), if host doesn't support AVX10, the
+configured avx10_version should be marked as filtered regardless of
+whether prefix is NULL or not.
+
+Check prefix before warn_report() instead of checking for
+have_filtered_features.
+
+Cc: qemu-stable@nongnu.org
+Fixes: commit bccfb846fd52 ("target/i386: add AVX10 feature and AVX10 version property")
+Signed-off-by: Zhao Liu <zhao1.liu@intel.com>
+Reviewed-by: Tao Su <tao1.su@linux.intel.com>
+Link: https://lore.kernel.org/r/20241106030728.553238-2-zhao1.liu@intel.com
+Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
+(cherry picked from commit cf4c263551886964c5d58bd7b675b13fd497b402)
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ target/i386/cpu.c | 6 ++++--
+ 1 file changed, 4 insertions(+), 2 deletions(-)
+
+diff --git a/target/i386/cpu.c b/target/i386/cpu.c
+index 3725dbbc4b..1981aeaba5 100644
+--- a/target/i386/cpu.c
++++ b/target/i386/cpu.c
+@@ -7718,8 +7718,10 @@ static bool x86_cpu_filter_features(X86CPU *cpu, bool verbose)
+             env->avx10_version = version;
+             have_filtered_features = true;
+         }
+-    } else if (env->avx10_version && prefix) {
+-        warn_report("%s: avx10.%d.", prefix, env->avx10_version);
++    } else if (env->avx10_version) {
++        if (prefix) {
++            warn_report("%s: avx10.%d.", prefix, env->avx10_version);
++        }
+         have_filtered_features = true;
+     }
+ 

debian/patches/extra/0005-memory-prevent-dma-reentracy-issues.patch

@@ -1,130 +0,0 @@
-From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
-From: Alexander Bulekov <alxndr@bu.edu>
-Date: Thu, 27 Apr 2023 17:10:06 -0400
-Subject: [PATCH] memory: prevent dma-reentracy issues
-
-Add a flag to the DeviceState, when a device is engaged in PIO/MMIO/DMA.
-This flag is set/checked prior to calling a device's MemoryRegion
-handlers, and set when device code initiates DMA.  The purpose of this
-flag is to prevent two types of DMA-based reentrancy issues:
-
-1.) mmio -> dma -> mmio case
-2.) bh -> dma write -> mmio case
-
-These issues have led to problems such as stack-exhaustion and
-use-after-frees.
-
-Summary of the problem from Peter Maydell:
-https://lore.kernel.org/qemu-devel/CAFEAcA_23vc7hE3iaM-JVA6W38LK4hJoWae5KcknhPRD5fPBZA@mail.gmail.com
-
-Resolves: https://gitlab.com/qemu-project/qemu/-/issues/62
-Resolves: https://gitlab.com/qemu-project/qemu/-/issues/540
-Resolves: https://gitlab.com/qemu-project/qemu/-/issues/541
-Resolves: https://gitlab.com/qemu-project/qemu/-/issues/556
-Resolves: https://gitlab.com/qemu-project/qemu/-/issues/557
-Resolves: https://gitlab.com/qemu-project/qemu/-/issues/827
-Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1282
-Resolves: CVE-2023-0330
-
-Signed-off-by: Alexander Bulekov <alxndr@bu.edu>
-Reviewed-by: Thomas Huth <thuth@redhat.com>
-Message-Id: <20230427211013.2994127-2-alxndr@bu.edu>
-[thuth: Replace warn_report() with warn_report_once()]
-Signed-off-by: Thomas Huth <thuth@redhat.com>
-(cherry-picked from commit a2e1753b8054344f32cf94f31c6399a58794a380)
-Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
----
- include/exec/memory.h  |  5 +++++
- include/hw/qdev-core.h |  7 +++++++
- softmmu/memory.c       | 16 ++++++++++++++++
- 3 files changed, 28 insertions(+)
-
-diff --git a/include/exec/memory.h b/include/exec/memory.h
-index 15ade918ba..e45ce6061f 100644
---- a/include/exec/memory.h
-+++ b/include/exec/memory.h
-@@ -767,6 +767,8 @@ struct MemoryRegion {
-     bool is_iommu;
-     RAMBlock *ram_block;
-     Object *owner;
-+    /* owner as TYPE_DEVICE. Used for re-entrancy checks in MR access hotpath */
-+    DeviceState *dev;
- 
-     const MemoryRegionOps *ops;
-     void *opaque;
-@@ -791,6 +793,9 @@ struct MemoryRegion {
-     unsigned ioeventfd_nb;
-     MemoryRegionIoeventfd *ioeventfds;
-     RamDiscardManager *rdm; /* Only for RAM */
-+
-+    /* For devices designed to perform re-entrant IO into their own IO MRs */
-+    bool disable_reentrancy_guard;
- };
- 
- struct IOMMUMemoryRegion {
-diff --git a/include/hw/qdev-core.h b/include/hw/qdev-core.h
-index bd50ad5ee1..7623703943 100644
---- a/include/hw/qdev-core.h
-+++ b/include/hw/qdev-core.h
-@@ -162,6 +162,10 @@ struct NamedClockList {
-     QLIST_ENTRY(NamedClockList) node;
- };
- 
-+typedef struct {
-+    bool engaged_in_io;
-+} MemReentrancyGuard;
-+
- /**
-  * DeviceState:
-  * @realized: Indicates whether the device has been fully constructed.
-@@ -194,6 +198,9 @@ struct DeviceState {
-     int alias_required_for_version;
-     ResettableState reset;
-     GSList *unplug_blockers;
-+
-+    /* Is the device currently in mmio/pio/dma? Used to prevent re-entrancy */
-+    MemReentrancyGuard mem_reentrancy_guard;
- };
- 
- struct DeviceListener {
-diff --git a/softmmu/memory.c b/softmmu/memory.c
-index b1a6cae6f5..b7b3386e9d 100644
---- a/softmmu/memory.c
-+++ b/softmmu/memory.c
-@@ -542,6 +542,18 @@ static MemTxResult access_with_adjusted_size(hwaddr addr,
-         access_size_max = 4;
-     }
- 
-+    /* Do not allow more than one simultaneous access to a device's IO Regions */
-+    if (mr->dev && !mr->disable_reentrancy_guard &&
-+        !mr->ram_device && !mr->ram && !mr->rom_device && !mr->readonly) {
-+        if (mr->dev->mem_reentrancy_guard.engaged_in_io) {
-+            warn_report_once("Blocked re-entrant IO on MemoryRegion: "
-+                             "%s at addr: 0x%" HWADDR_PRIX,
-+                             memory_region_name(mr), addr);
-+            return MEMTX_ACCESS_ERROR;
-+        }
-+        mr->dev->mem_reentrancy_guard.engaged_in_io = true;
-+    }
-+
-     /* FIXME: support unaligned access? */
-     access_size = MAX(MIN(size, access_size_max), access_size_min);
-     access_mask = MAKE_64BIT_MASK(0, access_size * 8);
-@@ -556,6 +568,9 @@ static MemTxResult access_with_adjusted_size(hwaddr addr,
-                         access_mask, attrs);
-         }
-     }
-+    if (mr->dev) {
-+        mr->dev->mem_reentrancy_guard.engaged_in_io = false;
-+    }
-     return r;
- }
- 
-@@ -1170,6 +1185,7 @@ static void memory_region_do_init(MemoryRegion *mr,
-     }
-     mr->name = g_strdup(name);
-     mr->owner = owner;
-+    mr->dev = (DeviceState *) object_dynamic_cast(mr->owner, TYPE_DEVICE);
-     mr->ram_block = NULL;
- 
-     if (name) {

debian/patches/extra/0006-lsi53c895a-disable-reentrancy-detection-for-script-R.patch

@@ -1,39 +0,0 @@
-From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
-From: Alexander Bulekov <alxndr@bu.edu>
-Date: Thu, 27 Apr 2023 17:10:10 -0400
-Subject: [PATCH] lsi53c895a: disable reentrancy detection for script RAM
-
-As the code is designed to use the memory APIs to access the script ram,
-disable reentrancy checks for the pseudo-RAM ram_io MemoryRegion.
-
-In the future, ram_io may be converted from an IO to a proper RAM MemoryRegion.
-
-Reported-by: Fiona Ebner <f.ebner@proxmox.com>
-Signed-off-by: Alexander Bulekov <alxndr@bu.edu>
-Reviewed-by: Thomas Huth <thuth@redhat.com>
-Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
-Message-Id: <20230427211013.2994127-6-alxndr@bu.edu>
-Signed-off-by: Thomas Huth <thuth@redhat.com>
-(cherry-picked from commit bfd6e7ae6a72b84e2eb9574f56e6ec037f05182c)
-Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
----
- hw/scsi/lsi53c895a.c | 6 ++++++
- 1 file changed, 6 insertions(+)
-
-diff --git a/hw/scsi/lsi53c895a.c b/hw/scsi/lsi53c895a.c
-index bbf32d3f73..17af67935f 100644
---- a/hw/scsi/lsi53c895a.c
-+++ b/hw/scsi/lsi53c895a.c
-@@ -2313,6 +2313,12 @@ static void lsi_scsi_realize(PCIDevice *dev, Error **errp)
-     memory_region_init_io(&s->io_io, OBJECT(s), &lsi_io_ops, s,
-                           "lsi-io", 256);
- 
-+    /*
-+     * Since we use the address-space API to interact with ram_io, disable the
-+     * re-entrancy guard.
-+     */
-+    s->ram_io.disable_reentrancy_guard = true;
-+
-     address_space_init(&s->pci_io_as, pci_address_space_io(dev), "lsi-pci-io");
-     qdev_init_gpio_out(d, &s->ext_irq, 1);
- 

debian/patches/extra/0006-net-Fix-announce_self.patch

@@ -0,0 +1,67 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Laurent Vivier <lvivier@redhat.com>
+Date: Fri, 17 Jan 2025 12:17:08 +0100
+Subject: [PATCH] net: Fix announce_self
+
+b9ad513e1876 ("net: Remove receive_raw()") adds an iovec entry
+in qemu_deliver_packet_iov() to add the virtio-net header
+in the data when QEMU_NET_PACKET_FLAG_RAW is set but forgets
+to increase the number of iovec entries in the array, so
+receive_iov() will only send the first entry (the virtio-net
+entry, full of 0) and no data. The packet will be discarded.
+
+The only user of QEMU_NET_PACKET_FLAG_RAW is announce_self.
+
+We can see the problem with tcpdump:
+
+- QEMU parameters:
+
+  .. -monitor stdio \
+     -netdev bridge,id=netdev0,br=virbr0 \
+     -device virtio-net,mac=9a:2b:2c:2d:2e:2f,netdev=netdev0 \
+
+- HMP command:
+
+  (qemu) announce_self
+
+- TCP dump:
+
+  $ sudo tcpdump -nxi virbr0
+
+  without the fix:
+
+    <nothing>
+
+  with the fix:
+
+   ARP, Reverse Request who-is 9a:2b:2c:2d:2e:2f tell 9a:2b:2c:2d:2e:2f, length 46
+        0x0000:  0001 0800 0604 0003 9a2b 2c2d 2e2f 0000
+        0x0010:  0000 9a2b 2c2d 2e2f 0000 0000 0000 0000
+        0x0020:  0000 0000 0000 0000 0000 0000 0000
+
+Reported-by: Xiaohui Li <xiaohli@redhat.com>
+Bug: https://issues.redhat.com/browse/RHEL-73891
+Fixes: b9ad513e1876 ("net: Remove receive_raw()")
+Cc: akihiko.odaki@daynix.com
+Signed-off-by: Laurent Vivier <lvivier@redhat.com>
+Reviewed-by: Akihiko Odaki <akihiko.odaki@daynix.com>
+Acked-by: Jason Wang <jasowang@redhat.com>
+Reviewed-by: Michael Tokarev <mjt@tls.msk.ru>
+(picked from https://lore.kernel.org/qemu-devel/20250117111709.970789-2-lvivier@redhat.com/)
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ net/net.c | 1 +
+ 1 file changed, 1 insertion(+)
+
+diff --git a/net/net.c b/net/net.c
+index 7ef6885876..fefa701bb2 100644
+--- a/net/net.c
++++ b/net/net.c
+@@ -822,6 +822,7 @@ static ssize_t qemu_deliver_packet_iov(NetClientState *sender,
+         iov_copy[0].iov_len =  nc->vnet_hdr_len;
+         memcpy(&iov_copy[1], iov, iovcnt * sizeof(*iov));
+         iov = iov_copy;
++        iovcnt++;
+     }
+ 
+     if (nc->info->receive_iov) {

debian/patches/extra/0007-bcm2835_property-disable-reentrancy-detection-for-io.patch

@@ -1,37 +0,0 @@
-From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
-From: Alexander Bulekov <alxndr@bu.edu>
-Date: Thu, 27 Apr 2023 17:10:11 -0400
-Subject: [PATCH] bcm2835_property: disable reentrancy detection for iomem
-
-As the code is designed for re-entrant calls from bcm2835_property to
-bcm2835_mbox and back into bcm2835_property, mark iomem as
-reentrancy-safe.
-
-Signed-off-by: Alexander Bulekov <alxndr@bu.edu>
-Reviewed-by: Thomas Huth <thuth@redhat.com>
-Message-Id: <20230427211013.2994127-7-alxndr@bu.edu>
-Signed-off-by: Thomas Huth <thuth@redhat.com>
-(cherry-picked from commit 985c4a4e547afb9573b6bd6843d20eb2c3d1d1cd)
-Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
----
- hw/misc/bcm2835_property.c | 7 +++++++
- 1 file changed, 7 insertions(+)
-
-diff --git a/hw/misc/bcm2835_property.c b/hw/misc/bcm2835_property.c
-index 890ae7bae5..de056ea2df 100644
---- a/hw/misc/bcm2835_property.c
-+++ b/hw/misc/bcm2835_property.c
-@@ -382,6 +382,13 @@ static void bcm2835_property_init(Object *obj)
- 
-     memory_region_init_io(&s->iomem, OBJECT(s), &bcm2835_property_ops, s,
-                           TYPE_BCM2835_PROPERTY, 0x10);
-+
-+    /*
-+     * bcm2835_property_ops call into bcm2835_mbox, which in-turn reads from
-+     * iomem. As such, mark iomem as re-entracy safe.
-+     */
-+    s->iomem.disable_reentrancy_guard = true;
-+
-     sysbus_init_mmio(SYS_BUS_DEVICE(s), &s->iomem);
-     sysbus_init_irq(SYS_BUS_DEVICE(s), &s->mbox_irq);
- }

debian/patches/extra/0007-net-dump-Correctly-compute-Ethernet-packet-offset.patch

@@ -0,0 +1,67 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Laurent Vivier <lvivier@redhat.com>
+Date: Fri, 17 Jan 2025 12:17:09 +0100
+Subject: [PATCH] net/dump: Correctly compute Ethernet packet offset
+
+When a packet is sent with QEMU_NET_PACKET_FLAG_RAW by QEMU it
+never includes virtio-net header even if qemu_get_vnet_hdr_len()
+is not 0, and filter-dump is not managing this case.
+
+The only user of QEMU_NET_PACKET_FLAG_RAW is announce_self,
+we can show the problem using it and tcpddump:
+
+- QEMU parameters:
+
+  .. -monitor stdio \
+     -netdev bridge,id=netdev0,br=virbr0 \
+     -device virtio-net,mac=9a:2b:2c:2d:2e:2f,netdev=netdev0 \
+     -object filter-dump,netdev=netdev0,file=log.pcap,id=pcap0
+
+- HMP command:
+
+  (qemu) announce_self
+
+- TCP dump:
+
+  $ tcpdump -nxr log.pcap
+
+  without the fix:
+
+    08:00:06:04:00:03 > 2e:2f:80:35:00:01, ethertype Unknown (0x9a2b), length 50:
+         0x0000:  2c2d 2e2f 0000 0000 9a2b 2c2d 2e2f 0000
+         0x0010:  0000 0000 0000 0000 0000 0000 0000 0000
+         0x0020:  0000 0000
+
+  with the fix:
+
+    ARP, Reverse Request who-is 9a:2b:2c:2d:2e:2f tell 9a:2b:2c:2d:2e:2f, length 46
+         0x0000:  0001 0800 0604 0003 9a2b 2c2d 2e2f 0000
+         0x0010:  0000 9a2b 2c2d 2e2f 0000 0000 0000 0000
+         0x0020:  0000 0000 0000 0000 0000 0000 0000
+
+Fixes: 481c52320a26 ("net: Strip virtio-net header when dumping")
+Cc: akihiko.odaki@daynix.com
+Signed-off-by: Laurent Vivier <lvivier@redhat.com>
+Reviewed-by: Akihiko Odaki <akihiko.odaki@daynix.com>
+Acked-by: Jason Wang <jasowang@redhat.com>
+Reviewed-by: Michael Tokarev <mjt@tls.msk.ru>
+(picked from https://lore.kernel.org/qemu-devel/20250117111709.970789-3-lvivier@redhat.com/)
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ net/dump.c | 3 ++-
+ 1 file changed, 2 insertions(+), 1 deletion(-)
+
+diff --git a/net/dump.c b/net/dump.c
+index 956e34a123..42ab8d7716 100644
+--- a/net/dump.c
++++ b/net/dump.c
+@@ -155,7 +155,8 @@ static ssize_t filter_dump_receive_iov(NetFilterState *nf, NetClientState *sndr,
+ {
+     NetFilterDumpState *nfds = FILTER_DUMP(nf);
+ 
+-    dump_receive_iov(&nfds->ds, iov, iovcnt, qemu_get_vnet_hdr_len(nf->netdev));
++    dump_receive_iov(&nfds->ds, iov, iovcnt, flags & QEMU_NET_PACKET_FLAG_RAW ?
++                     0 : qemu_get_vnet_hdr_len(nf->netdev));
+     return 0;
+ }
+ 

debian/patches/extra/0008-pci-acpi-Windows-PCI-Label-Id-bug-workaround.patch

@@ -0,0 +1,96 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Igor Mammedov <imammedo@redhat.com>
+Date: Wed, 15 Jan 2025 13:53:41 +0100
+Subject: [PATCH] pci: acpi: Windows 'PCI Label Id' bug workaround
+
+Current versions of Windows call _DSM(func=7) regardless
+of whether it is supported or not. It leads to NICs having bogus
+'PCI Label Id = 0', where none should be set at all.
+
+Also presence of 'PCI Label Id' triggers another Windows bug
+on localized versions that leads to hangs. The later bug is fixed
+in latest updates for 'Windows Server' but not in consumer
+versions of Windows (and there is no plans to fix it
+as far as I'm aware).
+
+Given it's easy, implement Microsoft suggested workaround
+(return invalid Package) so that affected Windows versions
+could boot on QEMU.
+This would effectvely remove bogus 'PCI Label Id's on NICs,
+but MS teem confirmed that flipping 'PCI Label Id' should not
+change 'Network Connection' ennumeration, so it should be safe
+for QEMU to change _DSM without any compat code.
+
+Smoke tested with WinXP and WS2022
+Resolves: https://gitlab.com/qemu-project/qemu/-/issues/774
+Signed-off-by: Igor Mammedov <imammedo@redhat.com>
+Message-Id: <20250115125342.3883374-3-imammedo@redhat.com>
+Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
+Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
+(cherry picked from commit 0b053391985abcc40b16ac8fc4a7f6588d1d95c1)
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ hw/i386/acpi-build.c | 33 +++++++++++++++++++++++----------
+ 1 file changed, 23 insertions(+), 10 deletions(-)
+
+diff --git a/hw/i386/acpi-build.c b/hw/i386/acpi-build.c
+index 9fcc2897b8..f7b961e04c 100644
+--- a/hw/i386/acpi-build.c
++++ b/hw/i386/acpi-build.c
+@@ -654,6 +654,7 @@ static Aml *aml_pci_pdsm(void)
+     Aml *acpi_index = aml_local(2);
+     Aml *zero = aml_int(0);
+     Aml *one = aml_int(1);
++    Aml *not_supp = aml_int(0xFFFFFFFF);
+     Aml *func = aml_arg(2);
+     Aml *params = aml_arg(4);
+     Aml *bnum = aml_derefof(aml_index(params, aml_int(0)));
+@@ -678,7 +679,7 @@ static Aml *aml_pci_pdsm(void)
+          */
+         ifctx1 = aml_if(aml_lnot(
+                      aml_or(aml_equal(acpi_index, zero),
+-                            aml_equal(acpi_index, aml_int(0xFFFFFFFF)), NULL)
++                            aml_equal(acpi_index, not_supp), NULL)
+                  ));
+         {
+             /* have supported functions */
+@@ -704,18 +705,30 @@ static Aml *aml_pci_pdsm(void)
+     {
+        Aml *pkg = aml_package(2);
+ 
+-       aml_append(pkg, zero);
+-       /*
+-        * optional, if not impl. should return null string
+-        */
+-       aml_append(pkg, aml_string("%s", ""));
+-       aml_append(ifctx, aml_store(pkg, ret));
+-
+        aml_append(ifctx, aml_store(aml_call2("AIDX", bnum, sunum), acpi_index));
++       aml_append(ifctx, aml_store(pkg, ret));
+        /*
+-        * update acpi-index to actual value
++        * Windows calls func=7 without checking if it's available,
++        * as workaround Microsoft has suggested to return invalid for func7
++        * Package, so return 2 elements package but only initialize elements
++        * when acpi_index is supported and leave them uninitialized, which
++        * leads elements to being Uninitialized ObjectType and should trip
++        * Windows into discarding result as an unexpected and prevent setting
++        * bogus 'PCI Label' on the device.
+         */
+-       aml_append(ifctx, aml_store(acpi_index, aml_index(ret, zero)));
++       ifctx1 = aml_if(aml_lnot(aml_lor(
++                    aml_equal(acpi_index, zero), aml_equal(acpi_index, not_supp)
++                )));
++       {
++           aml_append(ifctx1, aml_store(acpi_index, aml_index(ret, zero)));
++           /*
++            * optional, if not impl. should return null string
++            */
++           aml_append(ifctx1, aml_store(aml_string("%s", ""),
++                                        aml_index(ret, one)));
++       }
++       aml_append(ifctx, ifctx1);
++
+        aml_append(ifctx, aml_return(ret));
+     }
+ 

debian/patches/extra/0008-raven-disable-reentrancy-detection-for-iomem.patch

@@ -1,35 +0,0 @@
-From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
-From: Alexander Bulekov <alxndr@bu.edu>
-Date: Thu, 27 Apr 2023 17:10:12 -0400
-Subject: [PATCH] raven: disable reentrancy detection for iomem
-
-As the code is designed for re-entrant calls from raven_io_ops to
-pci-conf, mark raven_io_ops as reentrancy-safe.
-
-Signed-off-by: Alexander Bulekov <alxndr@bu.edu>
-Message-Id: <20230427211013.2994127-8-alxndr@bu.edu>
-Signed-off-by: Thomas Huth <thuth@redhat.com>
-(cherry-picked from commit 6dad5a6810d9c60ca320d01276f6133bbcfa1fc7)
-Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
----
- hw/pci-host/raven.c | 7 +++++++
- 1 file changed, 7 insertions(+)
-
-diff --git a/hw/pci-host/raven.c b/hw/pci-host/raven.c
-index 072ffe3c5e..9a11ac4b2b 100644
---- a/hw/pci-host/raven.c
-+++ b/hw/pci-host/raven.c
-@@ -294,6 +294,13 @@ static void raven_pcihost_initfn(Object *obj)
-     memory_region_init(&s->pci_memory, obj, "pci-memory", 0x3f000000);
-     address_space_init(&s->pci_io_as, &s->pci_io, "raven-io");
- 
-+    /*
-+     * Raven's raven_io_ops use the address-space API to access pci-conf-idx
-+     * (which is also owned by the raven device). As such, mark the
-+     * pci_io_non_contiguous as re-entrancy safe.
-+     */
-+    s->pci_io_non_contiguous.disable_reentrancy_guard = true;
-+
-     /* CPU address space */
-     memory_region_add_subregion(address_space_mem, PCI_IO_BASE_ADDR,
-                                 &s->pci_io);

debian/patches/extra/0009-apic-disable-reentrancy-detection-for-apic-msi.patch

@@ -1,36 +0,0 @@
-From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
-From: Alexander Bulekov <alxndr@bu.edu>
-Date: Thu, 27 Apr 2023 17:10:13 -0400
-Subject: [PATCH] apic: disable reentrancy detection for apic-msi
-
-As the code is designed for re-entrant calls to apic-msi, mark apic-msi
-as reentrancy-safe.
-
-Signed-off-by: Alexander Bulekov <alxndr@bu.edu>
-Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
-Message-Id: <20230427211013.2994127-9-alxndr@bu.edu>
-Signed-off-by: Thomas Huth <thuth@redhat.com>
-(cherry-picked from commit 50795ee051a342c681a9b45671c552fbd6274db8)
-Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
----
- hw/intc/apic.c | 7 +++++++
- 1 file changed, 7 insertions(+)
-
-diff --git a/hw/intc/apic.c b/hw/intc/apic.c
-index 20b5a94073..ac3d47d231 100644
---- a/hw/intc/apic.c
-+++ b/hw/intc/apic.c
-@@ -885,6 +885,13 @@ static void apic_realize(DeviceState *dev, Error **errp)
-     memory_region_init_io(&s->io_memory, OBJECT(s), &apic_io_ops, s, "apic-msi",
-                           APIC_SPACE_SIZE);
- 
-+    /*
-+     * apic-msi's apic_mem_write can call into ioapic_eoi_broadcast, which can
-+     * write back to apic-msi. As such mark the apic-msi region re-entrancy
-+     * safe.
-+     */
-+    s->io_memory.disable_reentrancy_guard = true;
-+
-     s->timer = timer_new_ns(QEMU_CLOCK_VIRTUAL, apic_timer, s);
-     local_apics[s->id] = s;
- 

debian/patches/extra/0009-hw-usb-hcd-xhci-pci-Use-modulo-to-select-MSI-vector-.patch

@@ -0,0 +1,53 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Phil Dennis-Jordan <phil@philjordan.eu>
+Date: Fri, 13 Dec 2024 17:06:14 +0100
+Subject: [PATCH] hw/usb/hcd-xhci-pci: Use modulo to select MSI vector as per
+ spec
+MIME-Version: 1.0
+Content-Type: text/plain; charset=UTF-8
+Content-Transfer-Encoding: 8bit
+
+QEMU would crash with a failed assertion if the XHCI controller
+attempted to raise the interrupt on an interrupter corresponding
+to a MSI vector with a higher index than the highest configured
+for the device by the guest driver.
+
+This behaviour is correct on the MSI/PCI side: per PCI 3.0 spec,
+devices must ensure they do not send MSI notifications for
+vectors beyond the range of those allocated by the system/driver
+software. Unlike MSI-X, there is no generic way for handling
+aliasing in the case of fewer allocated vectors than requested,
+so the specifics are up to device implementors. (Section
+6.8.3.4. "Sending Messages")
+
+It turns out the XHCI spec (Implementation Note in section 4.17,
+"Interrupters") requires that the host controller signal the MSI
+vector with the number computed by taking the interrupter number
+modulo the number of enabled MSI vectors.
+
+This change introduces that modulo calculation, fixing the
+failed assertion. This makes the device work correctly in MSI mode
+with macOS's XHCI driver, which only allocates a single vector.
+
+Signed-off-by: Phil Dennis-Jordan <phil@philjordan.eu>
+Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
+Message-ID: <20250112210056.16658-2-phil@philjordan.eu>
+Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
+(cherry picked from commit bb5b7fced6b5d3334ab20702fc846e47bb1fb731)
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ hw/usb/hcd-xhci-pci.c | 1 +
+ 1 file changed, 1 insertion(+)
+
+diff --git a/hw/usb/hcd-xhci-pci.c b/hw/usb/hcd-xhci-pci.c
+index a039f5778a..516e6909d2 100644
+--- a/hw/usb/hcd-xhci-pci.c
++++ b/hw/usb/hcd-xhci-pci.c
+@@ -74,6 +74,7 @@ static bool xhci_pci_intr_raise(XHCIState *xhci, int n, bool level)
+     }
+ 
+     if (msi_enabled(pci_dev) && level) {
++        n %= msi_nr_vectors_allocated(pci_dev);
+         msi_notify(pci_dev, n);
+         return true;
+     }

debian/patches/extra/0010-pci-ensure-valid-link-status-bits-for-downstream-por.patch

@@ -0,0 +1,63 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Sebastian Ott <sebott@redhat.com>
+Date: Tue, 3 Dec 2024 13:19:28 +0100
+Subject: [PATCH] pci: ensure valid link status bits for downstream ports
+
+PCI hotplug for downstream endpoints on arm fails because Linux'
+PCIe hotplug driver doesn't like the QEMU provided LNKSTA:
+
+  pcieport 0000:08:01.0: pciehp: Slot(2): Card present
+  pcieport 0000:08:01.0: pciehp: Slot(2): Link Up
+  pcieport 0000:08:01.0: pciehp: Slot(2): Cannot train link: status 0x2000
+
+There's 2 cases where LNKSTA isn't setup properly:
+* the downstream device has no express capability
+* max link width of the bridge is 0
+
+Move the sanity checks added via 88c869198aa63
+("pci: Sanity test minimum downstream LNKSTA") outside of the
+branch to make sure downstream ports always have a valid LNKSTA.
+
+Signed-off-by: Sebastian Ott <sebott@redhat.com>
+Tested-by: Zhenyu Zhang <zhenyzha@redhat.com>
+Message-Id: <20241203121928.14861-1-sebott@redhat.com>
+Reviewed-by: Alex Williamson <alex.williamson@redhat.com>
+Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
+Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
+(cherry picked from commit 694632fd44987cc4618612a38ad151047524a590)
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ hw/pci/pcie.c | 12 ++++++++----
+ 1 file changed, 8 insertions(+), 4 deletions(-)
+
+diff --git a/hw/pci/pcie.c b/hw/pci/pcie.c
+index 0b455c8654..1b12db6fa2 100644
+--- a/hw/pci/pcie.c
++++ b/hw/pci/pcie.c
+@@ -1113,18 +1113,22 @@ void pcie_sync_bridge_lnk(PCIDevice *bridge_dev)
+         if ((lnksta & PCI_EXP_LNKSTA_NLW) > (lnkcap & PCI_EXP_LNKCAP_MLW)) {
+             lnksta &= ~PCI_EXP_LNKSTA_NLW;
+             lnksta |= lnkcap & PCI_EXP_LNKCAP_MLW;
+-        } else if (!(lnksta & PCI_EXP_LNKSTA_NLW)) {
+-            lnksta |= QEMU_PCI_EXP_LNKSTA_NLW(QEMU_PCI_EXP_LNK_X1);
+         }
+ 
+         if ((lnksta & PCI_EXP_LNKSTA_CLS) > (lnkcap & PCI_EXP_LNKCAP_SLS)) {
+             lnksta &= ~PCI_EXP_LNKSTA_CLS;
+             lnksta |= lnkcap & PCI_EXP_LNKCAP_SLS;
+-        } else if (!(lnksta & PCI_EXP_LNKSTA_CLS)) {
+-            lnksta |= QEMU_PCI_EXP_LNKSTA_CLS(QEMU_PCI_EXP_LNK_2_5GT);
+         }
+     }
+ 
++    if (!(lnksta & PCI_EXP_LNKSTA_NLW)) {
++        lnksta |= QEMU_PCI_EXP_LNKSTA_NLW(QEMU_PCI_EXP_LNK_X1);
++    }
++
++    if (!(lnksta & PCI_EXP_LNKSTA_CLS)) {
++        lnksta |= QEMU_PCI_EXP_LNKSTA_CLS(QEMU_PCI_EXP_LNK_2_5GT);
++    }
++
+     pci_word_test_and_clear_mask(exp_cap + PCI_EXP_LNKSTA,
+                                  PCI_EXP_LNKSTA_CLS | PCI_EXP_LNKSTA_NLW);
+     pci_word_test_and_set_mask(exp_cap + PCI_EXP_LNKSTA, lnksta &

debian/patches/extra/0011-pci-msix-Fix-msix-pba-read-vector-poll-end-calculati.patch

@@ -0,0 +1,36 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Nicholas Piggin <npiggin@gmail.com>
+Date: Thu, 12 Dec 2024 22:04:02 +1000
+Subject: [PATCH] pci/msix: Fix msix pba read vector poll end calculation
+MIME-Version: 1.0
+Content-Type: text/plain; charset=UTF-8
+Content-Transfer-Encoding: 8bit
+
+The end vector calculation has a bug that results in polling fewer
+than required vectors when reading at a non-zero offset in PBA memory.
+
+Fixes: bbef882cc193 ("msi: add API to get notified about pending bit poll")
+Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
+Message-Id: <20241212120402.1475053-1-npiggin@gmail.com>
+Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
+Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
+Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
+(cherry picked from commit 42e2a7a0ab23784e44fcb18369e06067abc89305)
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ hw/pci/msix.c | 2 +-
+ 1 file changed, 1 insertion(+), 1 deletion(-)
+
+diff --git a/hw/pci/msix.c b/hw/pci/msix.c
+index 487e49834e..cc6e79ec67 100644
+--- a/hw/pci/msix.c
++++ b/hw/pci/msix.c
+@@ -250,7 +250,7 @@ static uint64_t msix_pba_mmio_read(void *opaque, hwaddr addr,
+     PCIDevice *dev = opaque;
+     if (dev->msix_vector_poll_notifier) {
+         unsigned vector_start = addr * 8;
+-        unsigned vector_end = MIN(addr + size * 8, dev->msix_entries_nr);
++        unsigned vector_end = MIN((addr + size) * 8, dev->msix_entries_nr);
+         dev->msix_vector_poll_notifier(dev, vector_start, vector_end);
+     }
+ 

debian/patches/pve/0001-PVE-Config-block-file-change-locking-default-to-off.patch

@@ -14,10 +14,10 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
  1 file changed, 2 insertions(+), 2 deletions(-)
 
 diff --git a/block/file-posix.c b/block/file-posix.c
-index c2dee3f056..9681bd0434 100644
+index 90fa54352c..e2ea071315 100644
 --- a/block/file-posix.c
 +++ b/block/file-posix.c
-@@ -553,7 +553,7 @@ static QemuOptsList raw_runtime_opts = {
+@@ -564,7 +564,7 @@ static QemuOptsList raw_runtime_opts = {
          {
              .name = "locking",
              .type = QEMU_OPT_STRING,
@@ -26,7 +26,7 @@ index c2dee3f056..9681bd0434 100644
          },
          {
              .name = "pr-manager",
-@@ -653,7 +653,7 @@ static int raw_open_common(BlockDriverState *bs, QDict *options,
+@@ -664,7 +664,7 @@ static int raw_open_common(BlockDriverState *bs, QDict *options,
          s->use_lock = false;
          break;
      case ON_OFF_AUTO_AUTO:

debian/patches/pve/0002-PVE-Config-Adjust-network-script-path-to-etc-kvm.patch

@@ -9,12 +9,12 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
  1 file changed, 2 insertions(+), 2 deletions(-)
 
 diff --git a/include/net/net.h b/include/net/net.h
-index 1448d00afb..d1601d32c1 100644
+index cdd5b109b0..653a37e9d1 100644
 --- a/include/net/net.h
 +++ b/include/net/net.h
-@@ -258,8 +258,8 @@ void netdev_add(QemuOpts *opts, Error **errp);
+@@ -305,8 +305,8 @@ void netdev_add(QemuOpts *opts, Error **errp);
+ 
  int net_hub_id_for_client(NetClientState *nc, int *id);
- NetClientState *net_hub_port_find(int hub_id);
  
 -#define DEFAULT_NETWORK_SCRIPT CONFIG_SYSCONFDIR "/qemu-ifup"
 -#define DEFAULT_NETWORK_DOWN_SCRIPT CONFIG_SYSCONFDIR "/qemu-ifdown"

debian/patches/pve/0003-PVE-Config-set-the-CPU-model-to-kvm64-32-instead-of-.patch

@@ -10,10 +10,10 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
  1 file changed, 2 insertions(+), 2 deletions(-)
 
 diff --git a/target/i386/cpu.h b/target/i386/cpu.h
-index d243e290d3..3489b05ec4 100644
+index 4c239a6970..be09263fb0 100644
 --- a/target/i386/cpu.h
 +++ b/target/i386/cpu.h
-@@ -2203,9 +2203,9 @@ uint64_t cpu_get_tsc(CPUX86State *env);
+@@ -2475,9 +2475,9 @@ uint64_t cpu_get_tsc(CPUX86State *env);
  #define CPU_RESOLVING_TYPE TYPE_X86_CPU
  
  #ifdef TARGET_X86_64

debian/patches/pve/0004-PVE-Config-ui-spice-default-to-pve-certificates.patch

@@ -9,10 +9,10 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
  1 file changed, 9 insertions(+), 6 deletions(-)
 
 diff --git a/ui/spice-core.c b/ui/spice-core.c
-index 52a59386d7..b20c25aee0 100644
+index bd9dbe03f1..a7ecaad9c7 100644
 --- a/ui/spice-core.c
 +++ b/ui/spice-core.c
-@@ -691,32 +691,35 @@ static void qemu_spice_init(void)
+@@ -690,32 +690,35 @@ static void qemu_spice_init(void)
  
      if (tls_port) {
          x509_dir = qemu_opt_get(opts, "x509-dir");

debian/patches/pve/0005-PVE-Config-glusterfs-no-default-logfile-if-daemonize.patch

@@ -9,10 +9,10 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
  1 file changed, 11 insertions(+), 4 deletions(-)
 
 diff --git a/block/gluster.c b/block/gluster.c
-index 185a83e5e5..f11a40aa9e 100644
+index e9c038042b..c8457a5014 100644
 --- a/block/gluster.c
 +++ b/block/gluster.c
-@@ -43,7 +43,7 @@
+@@ -42,7 +42,7 @@
  #define GLUSTER_DEBUG_DEFAULT       4
  #define GLUSTER_DEBUG_MAX           9
  #define GLUSTER_OPT_LOGFILE         "logfile"
@@ -21,15 +21,15 @@ index 185a83e5e5..f11a40aa9e 100644
  /*
   * Several versions of GlusterFS (3.12? -> 6.0.1) fail when the transfer size
   * is greater or equal to 1024 MiB, so we are limiting the transfer size to 512
-@@ -425,6 +425,7 @@ static struct glfs *qemu_gluster_glfs_init(BlockdevOptionsGluster *gconf,
+@@ -421,6 +421,7 @@ static struct glfs *qemu_gluster_glfs_init(BlockdevOptionsGluster *gconf,
      int old_errno;
      SocketAddressList *server;
-     unsigned long long port;
+     uint64_t port;
 +    const char *logfile;
  
      glfs = glfs_find_preopened(gconf->volume);
      if (glfs) {
-@@ -467,9 +468,15 @@ static struct glfs *qemu_gluster_glfs_init(BlockdevOptionsGluster *gconf,
+@@ -463,9 +464,15 @@ static struct glfs *qemu_gluster_glfs_init(BlockdevOptionsGluster *gconf,
          }
      }
  

debian/patches/pve/0006-PVE-Config-rbd-block-rbd-disable-rbd_cache_writethro.patch

@@ -18,7 +18,7 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
  1 file changed, 2 insertions(+)
 
 diff --git a/block/rbd.c b/block/rbd.c
-index 978671411e..a4749f3b1b 100644
+index 04ed0e242e..728bce3b1e 100644
 --- a/block/rbd.c
 +++ b/block/rbd.c
 @@ -963,6 +963,8 @@ static int qemu_rbd_connect(rados_t *cluster, rados_ioctx_t *io_ctx,

debian/patches/pve/0007-PVE-Up-glusterfs-allow-partial-reads.patch

@@ -16,10 +16,10 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
  1 file changed, 9 insertions(+), 1 deletion(-)
 
 diff --git a/block/gluster.c b/block/gluster.c
-index f11a40aa9e..6756e6b886 100644
+index c8457a5014..c3a9555591 100644
 --- a/block/gluster.c
 +++ b/block/gluster.c
-@@ -58,6 +58,7 @@ typedef struct GlusterAIOCB {
+@@ -57,6 +57,7 @@ typedef struct GlusterAIOCB {
      int ret;
      Coroutine *coroutine;
      AioContext *aio_context;
@@ -27,7 +27,7 @@ index f11a40aa9e..6756e6b886 100644
  } GlusterAIOCB;
  
  typedef struct BDRVGlusterState {
-@@ -753,8 +754,10 @@ static void gluster_finish_aiocb(struct glfs_fd *fd, ssize_t ret,
+@@ -746,8 +747,10 @@ static void gluster_finish_aiocb(struct glfs_fd *fd, ssize_t ret,
          acb->ret = 0; /* Success */
      } else if (ret < 0) {
          acb->ret = -errno; /* Read/Write failed */
@@ -39,7 +39,7 @@ index f11a40aa9e..6756e6b886 100644
      }
  
      aio_co_schedule(acb->aio_context, acb->coroutine);
-@@ -1021,6 +1024,7 @@ static coroutine_fn int qemu_gluster_co_pwrite_zeroes(BlockDriverState *bs,
+@@ -1018,6 +1021,7 @@ static coroutine_fn int qemu_gluster_co_pwrite_zeroes(BlockDriverState *bs,
      acb.ret = 0;
      acb.coroutine = qemu_coroutine_self();
      acb.aio_context = bdrv_get_aio_context(bs);
@@ -47,7 +47,7 @@ index f11a40aa9e..6756e6b886 100644
  
      ret = glfs_zerofill_async(s->fd, offset, bytes, gluster_finish_aiocb, &acb);
      if (ret < 0) {
-@@ -1201,9 +1205,11 @@ static coroutine_fn int qemu_gluster_co_rw(BlockDriverState *bs,
+@@ -1198,9 +1202,11 @@ static coroutine_fn int qemu_gluster_co_rw(BlockDriverState *bs,
      acb.aio_context = bdrv_get_aio_context(bs);
  
      if (write) {
@@ -59,7 +59,7 @@ index f11a40aa9e..6756e6b886 100644
          ret = glfs_preadv_async(s->fd, qiov->iov, qiov->niov, offset, 0,
                                  gluster_finish_aiocb, &acb);
      }
-@@ -1266,6 +1272,7 @@ static coroutine_fn int qemu_gluster_co_flush_to_disk(BlockDriverState *bs)
+@@ -1263,6 +1269,7 @@ static coroutine_fn int qemu_gluster_co_flush_to_disk(BlockDriverState *bs)
      acb.ret = 0;
      acb.coroutine = qemu_coroutine_self();
      acb.aio_context = bdrv_get_aio_context(bs);
@@ -67,7 +67,7 @@ index f11a40aa9e..6756e6b886 100644
  
      ret = glfs_fsync_async(s->fd, gluster_finish_aiocb, &acb);
      if (ret < 0) {
-@@ -1314,6 +1321,7 @@ static coroutine_fn int qemu_gluster_co_pdiscard(BlockDriverState *bs,
+@@ -1311,6 +1318,7 @@ static coroutine_fn int qemu_gluster_co_pdiscard(BlockDriverState *bs,
      acb.ret = 0;
      acb.coroutine = qemu_coroutine_self();
      acb.aio_context = bdrv_get_aio_context(bs);

debian/patches/pve/0008-PVE-Up-qemu-img-return-success-on-info-without-snaps.patch

@@ -9,10 +9,10 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
  1 file changed, 2 insertions(+), 1 deletion(-)
 
 diff --git a/qemu-img.c b/qemu-img.c
-index 9aeac69fa6..0919fac1f1 100644
+index 7668f86769..2575e97b43 100644
 --- a/qemu-img.c
 +++ b/qemu-img.c
-@@ -3059,7 +3059,8 @@ static int img_info(int argc, char **argv)
+@@ -3075,7 +3075,8 @@ static int img_info(int argc, char **argv)
      list = collect_image_info_list(image_opts, filename, fmt, chain,
                                     force_share);
      if (!list) {

debian/patches/pve/0009-PVE-Up-qemu-img-dd-add-osize-and-read-from-to-stdin-.patch

@@ -38,10 +38,10 @@ Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
  2 files changed, 133 insertions(+), 73 deletions(-)
 
 diff --git a/qemu-img-cmds.hx b/qemu-img-cmds.hx
-index 1b1dab5b17..d1616c045a 100644
+index c9dd70a892..048788b23d 100644
 --- a/qemu-img-cmds.hx
 +++ b/qemu-img-cmds.hx
-@@ -58,9 +58,9 @@ SRST
+@@ -60,9 +60,9 @@ SRST
  ERST
  
  DEF("dd", img_dd,
@@ -54,10 +54,10 @@ index 1b1dab5b17..d1616c045a 100644
  
  DEF("info", img_info,
 diff --git a/qemu-img.c b/qemu-img.c
-index 0919fac1f1..c584de648c 100644
+index 2575e97b43..8ec68b346f 100644
 --- a/qemu-img.c
 +++ b/qemu-img.c
-@@ -4885,10 +4885,12 @@ static int img_bitmap(int argc, char **argv)
+@@ -4993,10 +4993,12 @@ static int img_bitmap(int argc, char **argv)
  #define C_IF      04
  #define C_OF      010
  #define C_SKIP    020
@@ -70,7 +70,7 @@ index 0919fac1f1..c584de648c 100644
  };
  
  struct DdIo {
-@@ -4964,6 +4966,19 @@ static int img_dd_skip(const char *arg,
+@@ -5072,6 +5074,19 @@ static int img_dd_skip(const char *arg,
      return 0;
  }
  
@@ -90,7 +90,7 @@ index 0919fac1f1..c584de648c 100644
  static int img_dd(int argc, char **argv)
  {
      int ret = 0;
-@@ -5004,6 +5019,7 @@ static int img_dd(int argc, char **argv)
+@@ -5112,6 +5127,7 @@ static int img_dd(int argc, char **argv)
          { "if", img_dd_if, C_IF },
          { "of", img_dd_of, C_OF },
          { "skip", img_dd_skip, C_SKIP },
@@ -98,7 +98,7 @@ index 0919fac1f1..c584de648c 100644
          { NULL, NULL, 0 }
      };
      const struct option long_options[] = {
-@@ -5079,91 +5095,112 @@ static int img_dd(int argc, char **argv)
+@@ -5187,91 +5203,112 @@ static int img_dd(int argc, char **argv)
          arg = NULL;
      }
  
@@ -275,7 +275,7 @@ index 0919fac1f1..c584de648c 100644
      }
  
      if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz ||
-@@ -5180,20 +5217,43 @@ static int img_dd(int argc, char **argv)
+@@ -5288,20 +5325,43 @@ static int img_dd(int argc, char **argv)
      in.buf = g_new(uint8_t, in.bsz);
  
      for (out_pos = 0; in_pos < size; ) {

debian/patches/pve/0010-PVE-Up-qemu-img-dd-add-isize-parameter.patch

@@ -16,10 +16,10 @@ Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
  1 file changed, 25 insertions(+), 3 deletions(-)
 
 diff --git a/qemu-img.c b/qemu-img.c
-index c584de648c..a57ceeddfe 100644
+index 8ec68b346f..b98184bba1 100644
 --- a/qemu-img.c
 +++ b/qemu-img.c
-@@ -4886,11 +4886,13 @@ static int img_bitmap(int argc, char **argv)
+@@ -4994,11 +4994,13 @@ static int img_bitmap(int argc, char **argv)
  #define C_OF      010
  #define C_SKIP    020
  #define C_OSIZE   040
@@ -33,7 +33,7 @@ index c584de648c..a57ceeddfe 100644
  };
  
  struct DdIo {
-@@ -4979,6 +4981,19 @@ static int img_dd_osize(const char *arg,
+@@ -5087,6 +5089,19 @@ static int img_dd_osize(const char *arg,
      return 0;
  }
  
@@ -53,7 +53,7 @@ index c584de648c..a57ceeddfe 100644
  static int img_dd(int argc, char **argv)
  {
      int ret = 0;
-@@ -4993,12 +5008,14 @@ static int img_dd(int argc, char **argv)
+@@ -5101,12 +5116,14 @@ static int img_dd(int argc, char **argv)
      int c, i;
      const char *out_fmt = "raw";
      const char *fmt = NULL;
@@ -69,7 +69,7 @@ index c584de648c..a57ceeddfe 100644
      };
      struct DdIo in = {
          .bsz = 512, /* Block size is by default 512 bytes */
-@@ -5020,6 +5037,7 @@ static int img_dd(int argc, char **argv)
+@@ -5128,6 +5145,7 @@ static int img_dd(int argc, char **argv)
          { "of", img_dd_of, C_OF },
          { "skip", img_dd_skip, C_SKIP },
          { "osize", img_dd_osize, C_OSIZE },
@@ -77,7 +77,7 @@ index c584de648c..a57ceeddfe 100644
          { NULL, NULL, 0 }
      };
      const struct option long_options[] = {
-@@ -5216,9 +5234,10 @@ static int img_dd(int argc, char **argv)
+@@ -5324,9 +5342,10 @@ static int img_dd(int argc, char **argv)
  
      in.buf = g_new(uint8_t, in.bsz);
  
@@ -90,7 +90,7 @@ index c584de648c..a57ceeddfe 100644
          if (blk1) {
              in_ret = blk_pread(blk1, in_pos, bytes, in.buf, 0);
              if (in_ret == 0) {
-@@ -5227,6 +5246,9 @@ static int img_dd(int argc, char **argv)
+@@ -5335,6 +5354,9 @@ static int img_dd(int argc, char **argv)
          } else {
              in_ret = read(STDIN_FILENO, in.buf, bytes);
              if (in_ret == 0) {

debian/patches/pve/0011-PVE-Up-qemu-img-dd-add-n-skip_create.patch

@@ -13,10 +13,10 @@ Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
  3 files changed, 26 insertions(+), 12 deletions(-)
 
 diff --git a/docs/tools/qemu-img.rst b/docs/tools/qemu-img.rst
-index 15aeddc6d8..5e713e231d 100644
+index 3653adb963..d83e8fb3c0 100644
 --- a/docs/tools/qemu-img.rst
 +++ b/docs/tools/qemu-img.rst
-@@ -208,6 +208,10 @@ Parameters to convert subcommand:
+@@ -212,6 +212,10 @@ Parameters to convert subcommand:
  
  Parameters to dd subcommand:
  
@@ -27,7 +27,7 @@ index 15aeddc6d8..5e713e231d 100644
  .. program:: qemu-img-dd
  
  .. option:: bs=BLOCK_SIZE
-@@ -488,7 +492,7 @@ Command description:
+@@ -492,7 +496,7 @@ Command description:
    it doesn't need to be specified separately in this case.
  
  
@@ -36,7 +36,7 @@ index 15aeddc6d8..5e713e231d 100644
  
    dd copies from *INPUT* file to *OUTPUT* file converting it from
    *FMT* format to *OUTPUT_FMT* format.
-@@ -499,6 +503,11 @@ Command description:
+@@ -503,6 +507,11 @@ Command description:
  
    The size syntax is similar to :manpage:`dd(1)`'s size syntax.
  
@@ -49,10 +49,10 @@ index 15aeddc6d8..5e713e231d 100644
  
    Give information about the disk image *FILENAME*. Use it in
 diff --git a/qemu-img-cmds.hx b/qemu-img-cmds.hx
-index d1616c045a..b5b0bb4467 100644
+index 048788b23d..0b29a67a06 100644
 --- a/qemu-img-cmds.hx
 +++ b/qemu-img-cmds.hx
-@@ -58,9 +58,9 @@ SRST
+@@ -60,9 +60,9 @@ SRST
  ERST
  
  DEF("dd", img_dd,
@@ -65,10 +65,10 @@ index d1616c045a..b5b0bb4467 100644
  
  DEF("info", img_info,
 diff --git a/qemu-img.c b/qemu-img.c
-index a57ceeddfe..06d814e39c 100644
+index b98184bba1..6fc8384f64 100644
 --- a/qemu-img.c
 +++ b/qemu-img.c
-@@ -5010,7 +5010,7 @@ static int img_dd(int argc, char **argv)
+@@ -5118,7 +5118,7 @@ static int img_dd(int argc, char **argv)
      const char *fmt = NULL;
      int64_t size = 0, readsize = 0;
      int64_t out_pos, in_pos;
@@ -77,7 +77,7 @@ index a57ceeddfe..06d814e39c 100644
      struct DdInfo dd = {
          .flags = 0,
          .count = 0,
-@@ -5048,7 +5048,7 @@ static int img_dd(int argc, char **argv)
+@@ -5156,7 +5156,7 @@ static int img_dd(int argc, char **argv)
          { 0, 0, 0, 0 }
      };
  
@@ -86,7 +86,7 @@ index a57ceeddfe..06d814e39c 100644
          if (c == EOF) {
              break;
          }
-@@ -5068,6 +5068,9 @@ static int img_dd(int argc, char **argv)
+@@ -5176,6 +5176,9 @@ static int img_dd(int argc, char **argv)
          case 'h':
              help();
              break;
@@ -96,7 +96,7 @@ index a57ceeddfe..06d814e39c 100644
          case 'U':
              force_share = true;
              break;
-@@ -5198,13 +5201,15 @@ static int img_dd(int argc, char **argv)
+@@ -5306,13 +5309,15 @@ static int img_dd(int argc, char **argv)
                                  size - in.bsz * in.offset, &error_abort);
          }
  

debian/patches/pve/0012-qemu-img-dd-add-l-option-for-loading-a-snapshot.patch

@@ -12,10 +12,10 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
  3 files changed, 36 insertions(+), 7 deletions(-)
 
 diff --git a/docs/tools/qemu-img.rst b/docs/tools/qemu-img.rst
-index 5e713e231d..9390d5e5cf 100644
+index d83e8fb3c0..61c6b21859 100644
 --- a/docs/tools/qemu-img.rst
 +++ b/docs/tools/qemu-img.rst
-@@ -492,10 +492,10 @@ Command description:
+@@ -496,10 +496,10 @@ Command description:
    it doesn't need to be specified separately in this case.
  
  
@@ -30,10 +30,10 @@ index 5e713e231d..9390d5e5cf 100644
    The data is by default read and written using blocks of 512 bytes but can be
    modified by specifying *BLOCK_SIZE*. If count=\ *BLOCKS* is specified
 diff --git a/qemu-img-cmds.hx b/qemu-img-cmds.hx
-index b5b0bb4467..36f97e1f19 100644
+index 0b29a67a06..758f397232 100644
 --- a/qemu-img-cmds.hx
 +++ b/qemu-img-cmds.hx
-@@ -58,9 +58,9 @@ SRST
+@@ -60,9 +60,9 @@ SRST
  ERST
  
  DEF("dd", img_dd,
@@ -46,10 +46,10 @@ index b5b0bb4467..36f97e1f19 100644
  
  DEF("info", img_info,
 diff --git a/qemu-img.c b/qemu-img.c
-index 06d814e39c..e2c06c496d 100644
+index 6fc8384f64..a6c88e0860 100644
 --- a/qemu-img.c
 +++ b/qemu-img.c
-@@ -5002,6 +5002,7 @@ static int img_dd(int argc, char **argv)
+@@ -5110,6 +5110,7 @@ static int img_dd(int argc, char **argv)
      BlockDriver *drv = NULL, *proto_drv = NULL;
      BlockBackend *blk1 = NULL, *blk2 = NULL;
      QemuOpts *opts = NULL;
@@ -57,7 +57,7 @@ index 06d814e39c..e2c06c496d 100644
      QemuOptsList *create_opts = NULL;
      Error *local_err = NULL;
      bool image_opts = false;
-@@ -5011,6 +5012,7 @@ static int img_dd(int argc, char **argv)
+@@ -5119,6 +5120,7 @@ static int img_dd(int argc, char **argv)
      int64_t size = 0, readsize = 0;
      int64_t out_pos, in_pos;
      bool force_share = false, skip_create = false;
@@ -65,7 +65,7 @@ index 06d814e39c..e2c06c496d 100644
      struct DdInfo dd = {
          .flags = 0,
          .count = 0,
-@@ -5048,7 +5050,7 @@ static int img_dd(int argc, char **argv)
+@@ -5156,7 +5158,7 @@ static int img_dd(int argc, char **argv)
          { 0, 0, 0, 0 }
      };
  
@@ -74,7 +74,7 @@ index 06d814e39c..e2c06c496d 100644
          if (c == EOF) {
              break;
          }
-@@ -5071,6 +5073,19 @@ static int img_dd(int argc, char **argv)
+@@ -5179,6 +5181,19 @@ static int img_dd(int argc, char **argv)
          case 'n':
              skip_create = true;
              break;
@@ -94,7 +94,7 @@ index 06d814e39c..e2c06c496d 100644
          case 'U':
              force_share = true;
              break;
-@@ -5130,11 +5145,24 @@ static int img_dd(int argc, char **argv)
+@@ -5238,11 +5253,24 @@ static int img_dd(int argc, char **argv)
      if (dd.flags & C_IF) {
          blk1 = img_open(image_opts, in.filename, fmt, 0, false, false,
                          force_share);
@@ -120,7 +120,7 @@ index 06d814e39c..e2c06c496d 100644
      }
  
      if (dd.flags & C_OSIZE) {
-@@ -5289,6 +5317,7 @@ static int img_dd(int argc, char **argv)
+@@ -5397,6 +5425,7 @@ static int img_dd(int argc, char **argv)
  out:
      g_free(arg);
      qemu_opts_del(opts);

debian/patches/pve/0013-PVE-virtio-balloon-improve-query-balloon.patch

@@ -18,10 +18,10 @@ Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
  4 files changed, 82 insertions(+), 4 deletions(-)
 
 diff --git a/hw/core/machine-hmp-cmds.c b/hw/core/machine-hmp-cmds.c
-index c3e55ef9e9..0e32e6201f 100644
+index 8701f00cc7..3b4c5ef403 100644
 --- a/hw/core/machine-hmp-cmds.c
 +++ b/hw/core/machine-hmp-cmds.c
-@@ -169,7 +169,35 @@ void hmp_info_balloon(Monitor *mon, const QDict *qdict)
+@@ -179,7 +179,35 @@ void hmp_info_balloon(Monitor *mon, const QDict *qdict)
          return;
      }
  
@@ -59,10 +59,10 @@ index c3e55ef9e9..0e32e6201f 100644
      qapi_free_BalloonInfo(info);
  }
 diff --git a/hw/virtio/virtio-balloon.c b/hw/virtio/virtio-balloon.c
-index 746f07c4d2..a41854b902 100644
+index afd2ad6dd6..c724218c17 100644
 --- a/hw/virtio/virtio-balloon.c
 +++ b/hw/virtio/virtio-balloon.c
-@@ -804,8 +804,37 @@ static uint64_t virtio_balloon_get_features(VirtIODevice *vdev, uint64_t f,
+@@ -795,8 +795,37 @@ static uint64_t virtio_balloon_get_features(VirtIODevice *vdev, uint64_t f,
  static void virtio_balloon_stat(void *opaque, BalloonInfo *info)
  {
      VirtIOBalloon *dev = opaque;
@@ -103,12 +103,12 @@ index 746f07c4d2..a41854b902 100644
  
  static void virtio_balloon_to_target(void *opaque, ram_addr_t target)
 diff --git a/qapi/machine.json b/qapi/machine.json
-index 604b686e59..15f5f86683 100644
+index a6b8795b09..9f7ed0eaa0 100644
 --- a/qapi/machine.json
 +++ b/qapi/machine.json
-@@ -1056,9 +1056,29 @@
- # @actual: the logical size of the VM in bytes
- #          Formula used: logical_vm_size = vm_ram_size - balloon_size
+@@ -1163,9 +1163,29 @@
+ # @actual: the logical size of the VM in bytes Formula used:
+ #     logical_vm_size = vm_ram_size - balloon_size
  #
 +# @last_update: time when stats got updated from guest
 +#
@@ -138,10 +138,10 @@ index 604b686e59..15f5f86683 100644
  ##
  # @query-balloon:
 diff --git a/qapi/pragma.json b/qapi/pragma.json
-index 7f810b0e97..325e684411 100644
+index 023a2ef7bc..6aaa9cb975 100644
 --- a/qapi/pragma.json
 +++ b/qapi/pragma.json
-@@ -35,6 +35,7 @@
+@@ -81,6 +81,7 @@
      'member-name-exceptions': [     # visible in:
          'ACPISlotType',             # query-acpi-ospm-status
          'AcpiTableOptions',         # -acpitable

debian/patches/pve/0014-PVE-qapi-modify-query-machines.patch

@@ -13,10 +13,10 @@ Signed-off-by: Dietmar Maurer <dietmar@proxmox.com>
  2 files changed, 9 insertions(+), 1 deletion(-)
 
 diff --git a/hw/core/machine-qmp-cmds.c b/hw/core/machine-qmp-cmds.c
-index b98ff15089..24595f618c 100644
+index 130217da8f..52a6d74820 100644
 --- a/hw/core/machine-qmp-cmds.c
 +++ b/hw/core/machine-qmp-cmds.c
-@@ -103,6 +103,12 @@ MachineInfoList *qmp_query_machines(Error **errp)
+@@ -90,6 +90,12 @@ MachineInfoList *qmp_query_machines(bool has_compat_props, bool compat_props,
          info->numa_mem_supported = mc->numa_mem_supported;
          info->deprecated = !!mc->deprecation_reason;
          info->acpi = !!object_class_property_find(OBJECT_CLASS(mc), "acpi");
@@ -30,19 +30,19 @@ index b98ff15089..24595f618c 100644
              info->default_cpu_type = g_strdup(mc->default_cpu_type);
          }
 diff --git a/qapi/machine.json b/qapi/machine.json
-index 15f5f86683..c904280085 100644
+index 9f7ed0eaa0..16366b774a 100644
 --- a/qapi/machine.json
 +++ b/qapi/machine.json
-@@ -138,6 +138,8 @@
+@@ -167,6 +167,8 @@
  #
  # @is-default: whether the machine is default
  #
 +# @is-current: whether this machine is currently used
 +#
  # @cpu-max: maximum number of CPUs supported by the machine type
- #           (since 1.5)
+ #     (since 1.5)
  #
-@@ -161,7 +163,7 @@
+@@ -199,7 +201,7 @@
  ##
  { 'struct': 'MachineInfo',
    'data': { 'name': 'str', '*alias': 'str',
@@ -50,4 +50,4 @@ index 15f5f86683..c904280085 100644
 +            '*is-default': 'bool', '*is-current': 'bool', 'cpu-max': 'int',
              'hotpluggable-cpus': 'bool',  'numa-mem-supported': 'bool',
              'deprecated': 'bool', '*default-cpu-type': 'str',
-             '*default-ram-id': 'str', 'acpi': 'bool' } }
+             '*default-ram-id': 'str', 'acpi': 'bool',

debian/patches/pve/0015-PVE-qapi-modify-spice-query.patch

@@ -14,10 +14,10 @@ Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
  2 files changed, 7 insertions(+)
 
 diff --git a/qapi/ui.json b/qapi/ui.json
-index 98322342f7..316d4dc933 100644
+index 460a26b981..42b911bda3 100644
 --- a/qapi/ui.json
 +++ b/qapi/ui.json
-@@ -310,11 +310,14 @@
+@@ -312,11 +312,14 @@
  #
  # @channels: a list of @SpiceChannel for each active spice channel
  #
@@ -33,7 +33,7 @@ index 98322342f7..316d4dc933 100644
    'if': 'CONFIG_SPICE' }
  
 diff --git a/ui/spice-core.c b/ui/spice-core.c
-index b20c25aee0..26baeb7846 100644
+index a7ecaad9c7..fecf002d50 100644
 --- a/ui/spice-core.c
 +++ b/ui/spice-core.c
 @@ -548,6 +548,10 @@ static SpiceInfo *qmp_query_spice_real(Error **errp)

debian/patches/pve/0016-PVE-add-IOChannel-implementation-for-savevm-async.patch

@@ -14,20 +14,21 @@ Additionally, allows tracking the current position from the outside
 (intended to be used for progress tracking).
 
 Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
 ---
- migration/channel-savevm-async.c | 183 +++++++++++++++++++++++++++++++
+ migration/channel-savevm-async.c | 184 +++++++++++++++++++++++++++++++
  migration/channel-savevm-async.h |  51 +++++++++
  migration/meson.build            |   1 +
- 3 files changed, 235 insertions(+)
+ 3 files changed, 236 insertions(+)
  create mode 100644 migration/channel-savevm-async.c
  create mode 100644 migration/channel-savevm-async.h
 
 diff --git a/migration/channel-savevm-async.c b/migration/channel-savevm-async.c
 new file mode 100644
-index 0000000000..aab081ce07
+index 0000000000..081a192f49
 --- /dev/null
 +++ b/migration/channel-savevm-async.c
-@@ -0,0 +1,183 @@
+@@ -0,0 +1,184 @@
 +/*
 + * QIO Channel implementation to be used by savevm-async QMP calls
 + */
@@ -174,8 +175,9 @@ index 0000000000..aab081ce07
 +
 +static void
 +qio_channel_savevm_async_set_aio_fd_handler(QIOChannel *ioc,
-+                                            AioContext *ctx,
++                                            AioContext *read_ctx,
 +                                            IOHandler *io_read,
++                                            AioContext *write_ctx,
 +                                            IOHandler *io_write,
 +                                            void *opaque)
 +{
@@ -269,14 +271,14 @@ index 0000000000..17ae2cb261
 +
 +#endif /* QIO_CHANNEL_SAVEVM_ASYNC_H */
 diff --git a/migration/meson.build b/migration/meson.build
-index 0d1bb9f96e..8a142fc7a9 100644
+index d53cf3417a..b00d58064d 100644
 --- a/migration/meson.build
 +++ b/migration/meson.build
-@@ -13,6 +13,7 @@ softmmu_ss.add(files(
+@@ -13,6 +13,7 @@ system_ss.add(files(
    'block-dirty-bitmap.c',
    'channel.c',
    'channel-block.c',
 +  'channel-savevm-async.c',
-   'colo-failover.c',
-   'colo.c',
+   'cpu-throttle.c',
+   'dirtyrate.c',
    'exec.c',

debian/patches/pve/0017-PVE-add-savevm-async-for-background-state-snapshots.patch

@@ -27,7 +27,10 @@ Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
 [FE: further improve aborting
      adapt to removal of QEMUFileOps
      improve condition for entering final stage
-     adapt to QAPI and other changes for 8.0]
+     adapt to QAPI and other changes for 8.2
+     make sure to not call vm_start() from coroutine
+     stop CPU throttling after finishing
+     force raw format when loading state as suggested by Friedrich Weber]
 Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
 ---
  hmp-commands-info.hx         |  13 +
@@ -35,20 +38,20 @@ Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
  include/migration/snapshot.h |   2 +
  include/monitor/hmp.h        |   3 +
  migration/meson.build        |   1 +
- migration/savevm-async.c     | 533 +++++++++++++++++++++++++++++++++++
+ migration/savevm-async.c     | 571 +++++++++++++++++++++++++++++++++++
  monitor/hmp-cmds.c           |  38 +++
  qapi/migration.json          |  34 +++
- qapi/misc.json               |  16 ++
+ qapi/misc.json               |  18 ++
  qemu-options.hx              |  12 +
- softmmu/vl.c                 |  10 +
- 11 files changed, 679 insertions(+)
+ system/vl.c                  |  10 +
+ 11 files changed, 719 insertions(+)
  create mode 100644 migration/savevm-async.c
 
 diff --git a/hmp-commands-info.hx b/hmp-commands-info.hx
-index 47d63d26db..a166bff3d5 100644
+index c59cd6637b..d1a7b99add 100644
 --- a/hmp-commands-info.hx
 +++ b/hmp-commands-info.hx
-@@ -540,6 +540,19 @@ SRST
+@@ -512,6 +512,19 @@ SRST
      Show current migration parameters.
  ERST
  
@@ -69,10 +72,10 @@ index 47d63d26db..a166bff3d5 100644
          .name       = "balloon",
          .args_type  = "",
 diff --git a/hmp-commands.hx b/hmp-commands.hx
-index bb85ee1d26..d9f9f42d11 100644
+index 06746f0afc..0c7c6f2c16 100644
 --- a/hmp-commands.hx
 +++ b/hmp-commands.hx
-@@ -1846,3 +1846,20 @@ SRST
+@@ -1859,3 +1859,20 @@ SRST
    List event channels in the guest
  ERST
  #endif
@@ -94,18 +97,18 @@ index bb85ee1d26..d9f9f42d11 100644
 +        .coroutine  = true,
 +    },
 diff --git a/include/migration/snapshot.h b/include/migration/snapshot.h
-index e72083b117..c846d37806 100644
+index 9e4dcaaa75..2581730d74 100644
 --- a/include/migration/snapshot.h
 +++ b/include/migration/snapshot.h
-@@ -61,4 +61,6 @@ bool delete_snapshot(const char *name,
-                     bool has_devices, strList *devices,
-                     Error **errp);
+@@ -68,4 +68,6 @@ bool delete_snapshot(const char *name,
+  */
+ void load_snapshot_resume(RunState state);
  
 +int load_snapshot_from_blockdev(const char *filename, Error **errp);
 +
  #endif
 diff --git a/include/monitor/hmp.h b/include/monitor/hmp.h
-index fdb69b7f9c..fdf6b45fb8 100644
+index ae116d9804..2596cc2426 100644
 --- a/include/monitor/hmp.h
 +++ b/include/monitor/hmp.h
 @@ -28,6 +28,7 @@ void hmp_info_status(Monitor *mon, const QDict *qdict);
@@ -116,7 +119,7 @@ index fdb69b7f9c..fdf6b45fb8 100644
  void hmp_info_migrate(Monitor *mon, const QDict *qdict);
  void hmp_info_migrate_capabilities(Monitor *mon, const QDict *qdict);
  void hmp_info_migrate_parameters(Monitor *mon, const QDict *qdict);
-@@ -94,6 +95,8 @@ void hmp_closefd(Monitor *mon, const QDict *qdict);
+@@ -92,6 +93,8 @@ void hmp_closefd(Monitor *mon, const QDict *qdict);
  void hmp_mouse_move(Monitor *mon, const QDict *qdict);
  void hmp_mouse_button(Monitor *mon, const QDict *qdict);
  void hmp_mouse_set(Monitor *mon, const QDict *qdict);
@@ -126,11 +129,11 @@ index fdb69b7f9c..fdf6b45fb8 100644
  void coroutine_fn hmp_screendump(Monitor *mon, const QDict *qdict);
  void hmp_chardev_add(Monitor *mon, const QDict *qdict);
 diff --git a/migration/meson.build b/migration/meson.build
-index 8a142fc7a9..a7824b5266 100644
+index b00d58064d..075b013971 100644
 --- a/migration/meson.build
 +++ b/migration/meson.build
-@@ -25,6 +25,7 @@ softmmu_ss.add(files(
-   'multifd-zlib.c',
+@@ -29,6 +29,7 @@ system_ss.add(files(
+   'options.c',
    'postcopy-ram.c',
    'savevm.c',
 +  'savevm-async.c',
@@ -139,18 +142,21 @@ index 8a142fc7a9..a7824b5266 100644
    'threadinfo.c',
 diff --git a/migration/savevm-async.c b/migration/savevm-async.c
 new file mode 100644
-index 0000000000..ac1fac6378
+index 0000000000..ee8ef316d0
 --- /dev/null
 +++ b/migration/savevm-async.c
-@@ -0,0 +1,533 @@
+@@ -0,0 +1,571 @@
 +#include "qemu/osdep.h"
 +#include "migration/channel-savevm-async.h"
 +#include "migration/migration.h"
++#include "migration/migration-stats.h"
++#include "migration/options.h"
 +#include "migration/savevm.h"
 +#include "migration/snapshot.h"
 +#include "migration/global_state.h"
 +#include "migration/ram.h"
 +#include "migration/qemu-file.h"
++#include "sysemu/cpu-throttle.h"
 +#include "sysemu/sysemu.h"
 +#include "sysemu/runstate.h"
 +#include "block/block.h"
@@ -162,6 +168,7 @@ index 0000000000..ac1fac6378
 +#include "qapi/qapi-commands-misc.h"
 +#include "qapi/qapi-commands-block.h"
 +#include "qemu/cutils.h"
++#include "qemu/error-report.h"
 +#include "qemu/timer.h"
 +#include "qemu/main-loop.h"
 +#include "qemu/rcu.h"
@@ -255,6 +262,7 @@ index 0000000000..ac1fac6378
 +    }
 +
 +    if (snap_state.target) {
++        BlockDriverState *target_bs = blk_bs(snap_state.target);
 +        if (!savevm_aborted()) {
 +            /* try to truncate, but ignore errors (will fail on block devices).
 +            * note1: bdrv_read() need whole blocks, so we need to round up
@@ -263,7 +271,9 @@ index 0000000000..ac1fac6378
 +            size_t size = QEMU_ALIGN_UP(snap_state.bs_pos, BDRV_SECTOR_SIZE*2);
 +            blk_truncate(snap_state.target, size, false, PREALLOC_MODE_OFF, 0, NULL);
 +        }
-+        blk_op_unblock_all(snap_state.target, snap_state.blocker);
++        if (target_bs) {
++            bdrv_op_unblock_all(target_bs, snap_state.blocker);
++        }
 +        error_free(snap_state.blocker);
 +        snap_state.blocker = NULL;
 +        blk_unref(snap_state.target);
@@ -275,7 +285,7 @@ index 0000000000..ac1fac6378
 +    return ret;
 +}
 +
-+static void save_snapshot_error(const char *fmt, ...)
++static void G_GNUC_PRINTF(1, 2) save_snapshot_error(const char *fmt, ...)
 +{
 +    va_list ap;
 +    char *msg;
@@ -287,7 +297,7 @@ index 0000000000..ac1fac6378
 +    DPRINTF("save_snapshot_error: %s\n", msg);
 +
 +    if (!snap_state.error) {
-+        error_set(&snap_state.error, ERROR_CLASS_GENERIC_ERROR, "%s", msg);
++        error_setg(&snap_state.error, "%s", msg);
 +    }
 +
 +    g_free (msg);
@@ -298,7 +308,6 @@ index 0000000000..ac1fac6378
 +static void process_savevm_finalize(void *opaque)
 +{
 +    int ret;
-+    AioContext *iohandler_ctx = iohandler_get_aio_context();
 +    MigrationState *ms = migrate_get_current();
 +
 +    bool aborted = savevm_aborted();
@@ -315,9 +324,7 @@ index 0000000000..ac1fac6378
 +     * so move it back. It can stay in the main context and live out its live
 +     * there, since we're done with it after this method ends anyway.
 +     */
-+    aio_context_acquire(iohandler_ctx);
 +    blk_set_aio_context(snap_state.target, qemu_get_aio_context(), NULL);
-+    aio_context_release(iohandler_ctx);
 +
 +    ret = vm_stop_force_state(RUN_STATE_FINISH_MIGRATE);
 +    if (ret < 0) {
@@ -342,6 +349,12 @@ index 0000000000..ac1fac6378
 +        ret || aborted ? MIGRATION_STATUS_FAILED : MIGRATION_STATUS_COMPLETED);
 +    ms->to_dst_file = NULL;
 +
++    /*
++     * Same as in migration_iteration_finish(): saving RAM might've turned on CPU throttling for
++     * auto-converge, make sure to disable it.
++     */
++    cpu_throttle_stop();
++
 +    qemu_savevm_state_cleanup();
 +
 +    ret = save_snapshot_cleanup();
@@ -394,12 +407,12 @@ index 0000000000..ac1fac6378
 +         * lock. Similar to what is done in migration.c, call the exact variant
 +         * only once pend_precopy in the estimate is below the threshold.
 +         */
-+        qemu_mutex_unlock_iothread();
++        bql_unlock();
 +        qemu_savevm_state_pending_estimate(&pend_precopy, &pend_postcopy);
 +        if (pend_precopy <= threshold) {
 +            qemu_savevm_state_pending_exact(&pend_precopy, &pend_postcopy);
 +        }
-+        qemu_mutex_lock_iothread();
++        bql_lock();
 +        pending_size = pend_precopy + pend_postcopy;
 +
 +        /*
@@ -420,11 +433,7 @@ index 0000000000..ac1fac6378
 +            DPRINTF("savevm iterate pending size %lu ret %d\n", pending_size, ret);
 +        } else {
 +            qemu_system_wakeup_request(QEMU_WAKEUP_REASON_OTHER, NULL);
-+            ret = global_state_store();
-+            if (ret) {
-+                save_snapshot_error("global_state_store error %d", ret);
-+                break;
-+            }
++            global_state_store();
 +
 +            DPRINTF("savevm iterate complete\n");
 +            break;
@@ -443,21 +452,25 @@ index 0000000000..ac1fac6378
 +     * so move there now and after every flush.
 +     */
 +    aio_co_reschedule_self(qemu_get_aio_context());
-+    for (bs = bdrv_first(&it); bs; bs = bdrv_next(&it)) {
++    bdrv_graph_co_rdlock();
++    bs = bdrv_first(&it);
++    bdrv_graph_co_rdunlock();
++    while (bs) {
 +        /* target has BDRV_O_NO_FLUSH, no sense calling bdrv_flush on it */
-+        if (bs == blk_bs(snap_state.target)) {
-+            continue;
-+        }
-+
-+        AioContext *bs_ctx = bdrv_get_aio_context(bs);
-+        if (bs_ctx != qemu_get_aio_context()) {
-+            DPRINTF("savevm: async flushing drive %s\n", bs->filename);
-+            aio_co_reschedule_self(bs_ctx);
-+            bdrv_graph_co_rdlock();
-+            bdrv_flush(bs);
-+            bdrv_graph_co_rdunlock();
-+            aio_co_reschedule_self(qemu_get_aio_context());
++        if (bs != blk_bs(snap_state.target)) {
++            AioContext *bs_ctx = bdrv_get_aio_context(bs);
++            if (bs_ctx != qemu_get_aio_context()) {
++                DPRINTF("savevm: async flushing drive %s\n", bs->filename);
++                aio_co_reschedule_self(bs_ctx);
++                bdrv_graph_co_rdlock();
++                bdrv_flush(bs);
++                bdrv_graph_co_rdunlock();
++                aio_co_reschedule_self(qemu_get_aio_context());
++            }
 +        }
++        bdrv_graph_co_rdlock();
++        bs = bdrv_next(&it);
++        bdrv_graph_co_rdunlock();
 +    }
 +
 +    DPRINTF("timing: async flushing took %ld ms\n",
@@ -471,23 +484,18 @@ index 0000000000..ac1fac6378
 +    Error *local_err = NULL;
 +    MigrationState *ms = migrate_get_current();
 +    AioContext *iohandler_ctx = iohandler_get_aio_context();
++    BlockDriverState *target_bs = NULL;
++    int ret = 0;
 +
 +    int bdrv_oflags = BDRV_O_RDWR | BDRV_O_RESIZE | BDRV_O_NO_FLUSH;
 +
 +    if (snap_state.state != SAVE_STATE_DONE) {
-+        error_set(errp, ERROR_CLASS_GENERIC_ERROR,
-+                  "VM snapshot already started\n");
++        error_setg(errp, "VM snapshot already started\n");
 +        return;
 +    }
 +
-+    if (migration_is_running(ms->state)) {
-+        error_set(errp, ERROR_CLASS_GENERIC_ERROR, QERR_MIGRATION_ACTIVE);
-+        return;
-+    }
-+
-+    if (migrate_use_block()) {
-+        error_set(errp, ERROR_CLASS_GENERIC_ERROR,
-+                  "Block migration and snapshots are incompatible");
++    if (migration_is_running()) {
++        error_setg(errp, "There's a migration process in progress");
 +        return;
 +    }
 +
@@ -519,7 +527,12 @@ index 0000000000..ac1fac6378
 +    qdict_put_str(options, "driver", "raw");
 +    snap_state.target = blk_new_open(statefile, NULL, options, bdrv_oflags, &local_err);
 +    if (!snap_state.target) {
-+        error_set(errp, ERROR_CLASS_GENERIC_ERROR, "failed to open '%s'", statefile);
++        error_setg(errp, "failed to open '%s'", statefile);
++        goto restart;
++    }
++    target_bs = blk_bs(snap_state.target);
++    if (!target_bs) {
++        error_setg(errp, "failed to open '%s' - no block driver state", statefile);
 +        goto restart;
 +    }
 +
@@ -528,7 +541,7 @@ index 0000000000..ac1fac6378
 +    snap_state.file = qemu_file_new_output(ioc);
 +
 +    if (!snap_state.file) {
-+        error_set(errp, ERROR_CLASS_GENERIC_ERROR, "failed to open '%s'", statefile);
++        error_setg(errp, "failed to open '%s'", statefile);
 +        goto restart;
 +    }
 +
@@ -537,27 +550,36 @@ index 0000000000..ac1fac6378
 +     * State is cleared in process_savevm_co, but has to be initialized
 +     * here (blocking main thread, from QMP) to avoid race conditions.
 +     */
-+    migrate_init(ms);
-+    memset(&ram_counters, 0, sizeof(ram_counters));
-+    memset(&compression_counters, 0, sizeof(compression_counters));
++    if (migrate_init(ms, errp)) {
++        return;
++    }
++    memset(&mig_stats, 0, sizeof(mig_stats));
 +    ms->to_dst_file = snap_state.file;
 +
 +    error_setg(&snap_state.blocker, "block device is in use by savevm");
-+    blk_op_block_all(snap_state.target, snap_state.blocker);
++    bdrv_op_block_all(target_bs, snap_state.blocker);
 +
 +    snap_state.state = SAVE_STATE_ACTIVE;
 +    snap_state.finalize_bh = qemu_bh_new(process_savevm_finalize, &snap_state);
-+    snap_state.co = qemu_coroutine_create(&process_savevm_co, NULL);
-+    qemu_mutex_unlock_iothread();
 +    qemu_savevm_state_header(snap_state.file);
-+    qemu_savevm_state_setup(snap_state.file);
-+    qemu_mutex_lock_iothread();
++    ret = qemu_savevm_state_setup(snap_state.file, &local_err);
++    if (ret != 0) {
++        error_setg_errno(errp, -ret, "savevm state setup failed: %s",
++                         local_err ? error_get_pretty(local_err) : "unknown error");
++        return;
++    }
 +
 +    /* Async processing from here on out happens in iohandler context, so let
 +     * the target bdrv have its home there.
 +     */
-+    blk_set_aio_context(snap_state.target, iohandler_ctx, &local_err);
++    ret = blk_set_aio_context(snap_state.target, iohandler_ctx, &local_err);
++    if (ret != 0) {
++        warn_report("failed to set iohandler context for VM state target: %s %s",
++                    local_err ? error_get_pretty(local_err) : "unknown error",
++                    strerror(-ret));
++    }
 +
++    snap_state.co = qemu_coroutine_create(&process_savevm_co, NULL);
 +    aio_co_schedule(iohandler_ctx, snap_state.co);
 +
 +    return;
@@ -572,29 +594,10 @@ index 0000000000..ac1fac6378
 +    }
 +}
 +
-+void coroutine_fn qmp_savevm_end(Error **errp)
++static void coroutine_fn wait_for_close_co(void *opaque)
 +{
 +    int64_t timeout;
 +
-+    if (snap_state.state == SAVE_STATE_DONE) {
-+        error_set(errp, ERROR_CLASS_GENERIC_ERROR,
-+                  "VM snapshot not started\n");
-+        return;
-+    }
-+
-+    if (snap_state.state == SAVE_STATE_ACTIVE) {
-+        snap_state.state = SAVE_STATE_CANCELLED;
-+        goto wait_for_close;
-+    }
-+
-+    if (snap_state.saved_vm_running) {
-+        vm_start();
-+        snap_state.saved_vm_running = false;
-+    }
-+
-+    snap_state.state = SAVE_STATE_DONE;
-+
-+wait_for_close:
 +    if (!snap_state.target) {
 +        DPRINTF("savevm-end: no target file open\n");
 +        return;
@@ -622,25 +625,61 @@ index 0000000000..ac1fac6378
 +    DPRINTF("savevm-end: cleanup done\n");
 +}
 +
++void qmp_savevm_end(Error **errp)
++{
++    if (snap_state.state == SAVE_STATE_DONE) {
++        error_setg(errp, "VM snapshot not started\n");
++        return;
++    }
++
++    Coroutine *wait_for_close = qemu_coroutine_create(wait_for_close_co, NULL);
++
++    if (snap_state.state == SAVE_STATE_ACTIVE) {
++        snap_state.state = SAVE_STATE_CANCELLED;
++        qemu_coroutine_enter(wait_for_close);
++        return;
++    }
++
++    if (snap_state.saved_vm_running) {
++        vm_start();
++        snap_state.saved_vm_running = false;
++    }
++
++    snap_state.state = SAVE_STATE_DONE;
++
++    qemu_coroutine_enter(wait_for_close);
++}
++
 +int load_snapshot_from_blockdev(const char *filename, Error **errp)
 +{
 +    BlockBackend *be;
++    BlockDriverState *bs = NULL;
 +    Error *local_err = NULL;
 +    Error *blocker = NULL;
++    QDict *options;
 +
 +    QEMUFile *f;
 +    size_t bs_pos = 0;
 +    int ret = -EINVAL;
 +
-+    be = blk_new_open(filename, NULL, NULL, 0, &local_err);
++    options = qdict_new();
++    qdict_put_str(options, "driver", "raw");
++
++    be = blk_new_open(filename, NULL, options, 0, &local_err);
 +
 +    if (!be) {
 +        error_setg(errp, "Could not open VM state file");
 +        goto the_end;
 +    }
 +
++    bs = blk_bs(be);
++    if (!bs) {
++        error_setg(errp, "Could not open VM state file - missing block driver state");
++        goto the_end;
++    }
++
 +    error_setg(&blocker, "block device is in use by load state");
-+    blk_op_block_all(be, blocker);
++    bdrv_op_block_all(bs, blocker);
 +
 +    /* restore the VM state */
 +    f = qemu_file_new_input(QIO_CHANNEL(qio_channel_savevm_async_new(be, &bs_pos)));
@@ -670,28 +709,30 @@ index 0000000000..ac1fac6378
 +
 + the_end:
 +    if (be) {
-+        blk_op_unblock_all(be, blocker);
++        if (bs) {
++            bdrv_op_unblock_all(bs, blocker);
++        }
 +        error_free(blocker);
 +        blk_unref(be);
 +    }
 +    return ret;
 +}
 diff --git a/monitor/hmp-cmds.c b/monitor/hmp-cmds.c
-index 6c559b48c8..91be698308 100644
+index f601d06ab8..874084565f 100644
 --- a/monitor/hmp-cmds.c
 +++ b/monitor/hmp-cmds.c
-@@ -22,6 +22,7 @@
- #include "monitor/monitor-internal.h"
+@@ -24,6 +24,7 @@
  #include "qapi/error.h"
  #include "qapi/qapi-commands-control.h"
+ #include "qapi/qapi-commands-machine.h"
 +#include "qapi/qapi-commands-migration.h"
  #include "qapi/qapi-commands-misc.h"
  #include "qapi/qmp/qdict.h"
- #include "qapi/qmp/qerror.h"
-@@ -443,3 +444,40 @@ void hmp_info_mtree(Monitor *mon, const QDict *qdict)
- 
-     mtree_info(flatview, dispatch_tree, owner, disabled);
+ #include "qemu/cutils.h"
+@@ -434,3 +435,40 @@ void hmp_dumpdtb(Monitor *mon, const QDict *qdict)
+     monitor_printf(mon, "dtb dumped to %s", filename);
  }
+ #endif
 +
 +void hmp_savevm_start(Monitor *mon, const QDict *qdict)
 +{
@@ -730,12 +771,12 @@ index 6c559b48c8..91be698308 100644
 +    }
 +}
 diff --git a/qapi/migration.json b/qapi/migration.json
-index c84fa10e86..1702b92553 100644
+index a605dc26db..927b1e1c7d 100644
 --- a/qapi/migration.json
 +++ b/qapi/migration.json
-@@ -261,6 +261,40 @@
-            '*compression': 'CompressionStats',
-            '*socket-address': ['SocketAddress'] } }
+@@ -276,6 +276,40 @@
+            '*dirty-limit-throttle-time-per-round': 'uint64',
+            '*dirty-limit-ring-full-time': 'uint64'} }
  
 +##
 +# @SaveVMInfo:
@@ -775,10 +816,10 @@ index c84fa10e86..1702b92553 100644
  # @query-migrate:
  #
 diff --git a/qapi/misc.json b/qapi/misc.json
-index 6ddd16ea28..e5681ae8a2 100644
+index 559b66f201..7959e89c1e 100644
 --- a/qapi/misc.json
 +++ b/qapi/misc.json
-@@ -469,6 +469,22 @@
+@@ -454,6 +454,24 @@
  ##
  { 'command': 'query-fdsets', 'returns': ['FdsetInfo'] }
  
@@ -787,6 +828,8 @@ index 6ddd16ea28..e5681ae8a2 100644
 +#
 +# Prepare for snapshot and halt VM. Save VM state to statefile.
 +#
++# @statefile: target file that state should be written to.
++#
 +##
 +{ 'command': 'savevm-start', 'data': { '*statefile': 'str' } }
 +
@@ -796,16 +839,16 @@ index 6ddd16ea28..e5681ae8a2 100644
 +# Resume VM after a snapshot.
 +#
 +##
-+{ 'command': 'savevm-end', 'coroutine': true }
++{ 'command': 'savevm-end' }
 +
  ##
  # @CommandLineParameterType:
  #
 diff --git a/qemu-options.hx b/qemu-options.hx
-index fdddfab6ff..fdd551c2bb 100644
+index dacc9790a4..c05f411599 100644
 --- a/qemu-options.hx
 +++ b/qemu-options.hx
-@@ -4398,6 +4398,18 @@ SRST
+@@ -4764,6 +4764,18 @@ SRST
      Start right away with a saved state (``loadvm`` in monitor)
  ERST
  
@@ -824,10 +867,10 @@ index fdddfab6ff..fdd551c2bb 100644
  #ifndef _WIN32
  DEF("daemonize", 0, QEMU_OPTION_daemonize, \
      "-daemonize      daemonize QEMU after initializing\n", QEMU_ARCH_ALL)
-diff --git a/softmmu/vl.c b/softmmu/vl.c
-index ea20b23e4c..0eabc71b68 100644
---- a/softmmu/vl.c
-+++ b/softmmu/vl.c
+diff --git a/system/vl.c b/system/vl.c
+index 2f855d83fb..39d451bb41 100644
+--- a/system/vl.c
++++ b/system/vl.c
 @@ -164,6 +164,7 @@ static const char *accelerators;
  static bool have_custom_ram_size;
  static const char *ram_memdev_id;
@@ -836,10 +879,10 @@ index ea20b23e4c..0eabc71b68 100644
  static QTAILQ_HEAD(, ObjectOption) object_opts = QTAILQ_HEAD_INITIALIZER(object_opts);
  static QTAILQ_HEAD(, DeviceOption) device_opts = QTAILQ_HEAD_INITIALIZER(device_opts);
  static int display_remote;
-@@ -2612,6 +2613,12 @@ void qmp_x_exit_preconfig(Error **errp)
- 
-     if (loadvm) {
+@@ -2725,6 +2726,12 @@ void qmp_x_exit_preconfig(Error **errp)
+         RunState state = autostart ? RUN_STATE_RUNNING : runstate_get();
          load_snapshot(loadvm, NULL, false, NULL, &error_fatal);
+         load_snapshot_resume(state);
 +    } else if (loadstate) {
 +        Error *local_err = NULL;
 +        if (load_snapshot_from_blockdev(loadstate, &local_err) < 0) {
@@ -849,7 +892,7 @@ index ea20b23e4c..0eabc71b68 100644
      }
      if (replay_mode != REPLAY_MODE_NONE) {
          replay_vmstate_init();
-@@ -3159,6 +3166,9 @@ void qemu_init(int argc, char **argv)
+@@ -3262,6 +3269,9 @@ void qemu_init(int argc, char **argv)
              case QEMU_OPTION_loadvm:
                  loadvm = optarg;
                  break;

debian/patches/pve/0018-PVE-add-optional-buffer-size-to-QEMUFile.patch

@@ -13,18 +13,18 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
 [FE: adapt to removal of QEMUFileOps]
 Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
 ---
- migration/qemu-file.c    | 49 +++++++++++++++++++++++++++-------------
+ migration/qemu-file.c    | 48 +++++++++++++++++++++++++++-------------
  migration/qemu-file.h    |  2 ++
- migration/savevm-async.c |  5 ++--
- 3 files changed, 38 insertions(+), 18 deletions(-)
+ migration/savevm-async.c |  5 +++--
+ 3 files changed, 38 insertions(+), 17 deletions(-)
 
 diff --git a/migration/qemu-file.c b/migration/qemu-file.c
-index 102ab3b439..5ced17aba4 100644
+index b6d2f588bd..754dc0b3f7 100644
 --- a/migration/qemu-file.c
 +++ b/migration/qemu-file.c
-@@ -31,8 +31,8 @@
- #include "trace.h"
- #include "qapi/error.h"
+@@ -34,8 +34,8 @@
+ #include "rdma.h"
+ #include "io/channel-file.h"
  
 -#define IO_BUF_SIZE 32768
 -#define MAX_IOV_SIZE MIN_CONST(IOV_MAX, 64)
@@ -32,8 +32,8 @@ index 102ab3b439..5ced17aba4 100644
 +#define MAX_IOV_SIZE MIN_CONST(IOV_MAX, 256)
  
  struct QEMUFile {
-     const QEMUFileHooks *hooks;
-@@ -55,7 +55,8 @@ struct QEMUFile {
+     QIOChannel *ioc;
+@@ -43,7 +43,8 @@ struct QEMUFile {
  
      int buf_index;
      int buf_size; /* 0 when writing */
@@ -43,8 +43,8 @@ index 102ab3b439..5ced17aba4 100644
  
      DECLARE_BITMAP(may_free, MAX_IOV_SIZE);
      struct iovec iov[MAX_IOV_SIZE];
-@@ -127,7 +128,9 @@ bool qemu_file_mode_is_not_valid(const char *mode)
-     return false;
+@@ -100,7 +101,9 @@ int qemu_file_shutdown(QEMUFile *f)
+     return 0;
  }
  
 -static QEMUFile *qemu_file_new_impl(QIOChannel *ioc, bool is_writable)
@@ -54,7 +54,7 @@ index 102ab3b439..5ced17aba4 100644
  {
      QEMUFile *f;
  
-@@ -136,6 +139,8 @@ static QEMUFile *qemu_file_new_impl(QIOChannel *ioc, bool is_writable)
+@@ -109,6 +112,8 @@ static QEMUFile *qemu_file_new_impl(QIOChannel *ioc, bool is_writable)
      object_ref(ioc);
      f->ioc = ioc;
      f->is_writable = is_writable;
@@ -63,7 +63,7 @@ index 102ab3b439..5ced17aba4 100644
  
      return f;
  }
-@@ -146,17 +151,27 @@ static QEMUFile *qemu_file_new_impl(QIOChannel *ioc, bool is_writable)
+@@ -119,17 +124,27 @@ static QEMUFile *qemu_file_new_impl(QIOChannel *ioc, bool is_writable)
   */
  QEMUFile *qemu_file_get_return_path(QEMUFile *f)
  {
@@ -93,8 +93,8 @@ index 102ab3b439..5ced17aba4 100644
 +    return qemu_file_new_impl(ioc, false, buffer_size);
  }
  
- void qemu_file_set_hooks(QEMUFile *f, const QEMUFileHooks *hooks)
-@@ -414,7 +429,7 @@ static ssize_t qemu_fill_buffer(QEMUFile *f)
+ /*
+@@ -327,7 +342,7 @@ static ssize_t coroutine_mixed_fn qemu_fill_buffer(QEMUFile *f)
      do {
          len = qio_channel_read(f->ioc,
                                 (char *)f->buf + pending,
@@ -103,16 +103,17 @@ index 102ab3b439..5ced17aba4 100644
                                 &local_error);
          if (len == QIO_CHANNEL_ERR_BLOCK) {
              if (qemu_in_coroutine()) {
-@@ -464,6 +479,8 @@ int qemu_fclose(QEMUFile *f)
+@@ -367,6 +382,9 @@ int qemu_fclose(QEMUFile *f)
+         ret = ret2;
      }
      g_clear_pointer(&f->ioc, object_unref);
- 
++
 +    free(f->buf);
 +
-     /* If any error was spotted before closing, we should report it
-      * instead of the close() return value.
-      */
-@@ -518,7 +535,7 @@ static void add_buf_to_iovec(QEMUFile *f, size_t len)
+     error_free(f->last_error_obj);
+     g_free(f);
+     trace_qemu_file_fclose();
+@@ -415,7 +433,7 @@ static void add_buf_to_iovec(QEMUFile *f, size_t len)
  {
      if (!add_to_iovec(f, f->buf + f->buf_index, len, false)) {
          f->buf_index += len;
@@ -121,7 +122,7 @@ index 102ab3b439..5ced17aba4 100644
              qemu_fflush(f);
          }
      }
-@@ -544,7 +561,7 @@ void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, size_t size)
+@@ -440,7 +458,7 @@ void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, size_t size)
      }
  
      while (size > 0) {
@@ -130,7 +131,7 @@ index 102ab3b439..5ced17aba4 100644
          if (l > size) {
              l = size;
          }
-@@ -591,8 +608,8 @@ size_t qemu_peek_buffer(QEMUFile *f, uint8_t **buf, size_t size, size_t offset)
+@@ -586,8 +604,8 @@ size_t coroutine_mixed_fn qemu_peek_buffer(QEMUFile *f, uint8_t **buf, size_t si
      size_t index;
  
      assert(!qemu_file_is_writable(f));
@@ -141,7 +142,7 @@ index 102ab3b439..5ced17aba4 100644
  
      /* The 1st byte to read from */
      index = f->buf_index + offset;
-@@ -642,7 +659,7 @@ size_t qemu_get_buffer(QEMUFile *f, uint8_t *buf, size_t size)
+@@ -637,7 +655,7 @@ size_t coroutine_mixed_fn qemu_get_buffer(QEMUFile *f, uint8_t *buf, size_t size
          size_t res;
          uint8_t *src;
  
@@ -150,16 +151,16 @@ index 102ab3b439..5ced17aba4 100644
          if (res == 0) {
              return done;
          }
-@@ -676,7 +693,7 @@ size_t qemu_get_buffer(QEMUFile *f, uint8_t *buf, size_t size)
+@@ -671,7 +689,7 @@ size_t coroutine_mixed_fn qemu_get_buffer(QEMUFile *f, uint8_t *buf, size_t size
   */
- size_t qemu_get_buffer_in_place(QEMUFile *f, uint8_t **buf, size_t size)
+ size_t coroutine_mixed_fn qemu_get_buffer_in_place(QEMUFile *f, uint8_t **buf, size_t size)
  {
 -    if (size < IO_BUF_SIZE) {
 +    if (size < f->buf_allocated_size) {
          size_t res;
          uint8_t *src = NULL;
  
-@@ -701,7 +718,7 @@ int qemu_peek_byte(QEMUFile *f, int offset)
+@@ -696,7 +714,7 @@ int coroutine_mixed_fn qemu_peek_byte(QEMUFile *f, int offset)
      int index = f->buf_index + offset;
  
      assert(!qemu_file_is_writable(f));
@@ -168,34 +169,25 @@ index 102ab3b439..5ced17aba4 100644
  
      if (index >= f->buf_size) {
          qemu_fill_buffer(f);
-@@ -853,7 +870,7 @@ static int qemu_compress_data(z_stream *stream, uint8_t *dest, size_t dest_len,
- ssize_t qemu_put_compression_data(QEMUFile *f, z_stream *stream,
-                                   const uint8_t *p, size_t size)
- {
--    ssize_t blen = IO_BUF_SIZE - f->buf_index - sizeof(int32_t);
-+    ssize_t blen = f->buf_allocated_size - f->buf_index - sizeof(int32_t);
- 
-     if (blen < compressBound(size)) {
-         return -1;
 diff --git a/migration/qemu-file.h b/migration/qemu-file.h
-index 9d0155a2a1..cc06240e8d 100644
+index 11c2120edd..edf3c5d147 100644
 --- a/migration/qemu-file.h
 +++ b/migration/qemu-file.h
-@@ -63,7 +63,9 @@ typedef struct QEMUFileHooks {
- } QEMUFileHooks;
+@@ -30,7 +30,9 @@
+ #include "io/channel.h"
  
  QEMUFile *qemu_file_new_input(QIOChannel *ioc);
 +QEMUFile *qemu_file_new_input_sized(QIOChannel *ioc, size_t buffer_size);
  QEMUFile *qemu_file_new_output(QIOChannel *ioc);
 +QEMUFile *qemu_file_new_output_sized(QIOChannel *ioc, size_t buffer_size);
- void qemu_file_set_hooks(QEMUFile *f, const QEMUFileHooks *hooks);
  int qemu_fclose(QEMUFile *f);
  
+ /*
 diff --git a/migration/savevm-async.c b/migration/savevm-async.c
-index ac1fac6378..ea3b2f36a6 100644
+index ee8ef316d0..1e79fce9ba 100644
 --- a/migration/savevm-async.c
 +++ b/migration/savevm-async.c
-@@ -380,7 +380,7 @@ void qmp_savevm_start(const char *statefile, Error **errp)
+@@ -390,7 +390,7 @@ void qmp_savevm_start(const char *statefile, Error **errp)
  
      QIOChannel *ioc = QIO_CHANNEL(qio_channel_savevm_async_new(snap_state.target,
                                                                 &snap_state.bs_pos));
@@ -203,9 +195,9 @@ index ac1fac6378..ea3b2f36a6 100644
 +    snap_state.file = qemu_file_new_output_sized(ioc, 4 * 1024 * 1024);
  
      if (!snap_state.file) {
-         error_set(errp, ERROR_CLASS_GENERIC_ERROR, "failed to open '%s'", statefile);
-@@ -498,7 +498,8 @@ int load_snapshot_from_blockdev(const char *filename, Error **errp)
-     blk_op_block_all(be, blocker);
+         error_setg(errp, "failed to open '%s'", statefile);
+@@ -534,7 +534,8 @@ int load_snapshot_from_blockdev(const char *filename, Error **errp)
+     bdrv_op_block_all(bs, blocker);
  
      /* restore the VM state */
 -    f = qemu_file_new_input(QIO_CHANNEL(qio_channel_savevm_async_new(be, &bs_pos)));

debian/patches/pve/0019-PVE-block-add-the-zeroinit-block-driver-filter.patch

@@ -4,32 +4,34 @@ Date: Mon, 6 Apr 2020 12:16:47 +0200
 Subject: [PATCH] PVE: block: add the zeroinit block driver filter
 
 Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
-[FE: adapt to changed function signatures]
+[FE: adapt to changed function signatures
+     adhere to block graph lock requirements
+     use dedicated function to open file child]
 Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
 ---
  block/meson.build |   1 +
- block/zeroinit.c  | 200 ++++++++++++++++++++++++++++++++++++++++++++++
- 2 files changed, 201 insertions(+)
+ block/zeroinit.c  | 207 ++++++++++++++++++++++++++++++++++++++++++++++
+ 2 files changed, 208 insertions(+)
  create mode 100644 block/zeroinit.c
 
 diff --git a/block/meson.build b/block/meson.build
-index 382bec0e7d..253fe49fa2 100644
+index f1262ec2ba..6a60b5d6b9 100644
 --- a/block/meson.build
 +++ b/block/meson.build
-@@ -44,6 +44,7 @@ block_ss.add(files(
-   'vmdk.c',
-   'vpc.c',
+@@ -39,6 +39,7 @@ block_ss.add(files(
+   'throttle.c',
+   'throttle-groups.c',
    'write-threshold.c',
 +  'zeroinit.c',
- ), zstd, zlib, gnutls)
+ ), zstd, zlib)
  
- softmmu_ss.add(when: 'CONFIG_TCG', if_true: files('blkreplay.c'))
+ system_ss.add(when: 'CONFIG_TCG', if_true: files('blkreplay.c'))
 diff --git a/block/zeroinit.c b/block/zeroinit.c
 new file mode 100644
-index 0000000000..1257342724
+index 0000000000..2b2b194ccf
 --- /dev/null
 +++ b/block/zeroinit.c
-@@ -0,0 +1,200 @@
+@@ -0,0 +1,207 @@
 +/*
 + * Filter to fake a zero-initialized block device.
 + *
@@ -44,6 +46,7 @@ index 0000000000..1257342724
 +#include "qapi/error.h"
 +#include "block/block_int.h"
 +#include "block/block-io.h"
++#include "block/graph-lock.h"
 +#include "qapi/qmp/qdict.h"
 +#include "qapi/qmp/qstring.h"
 +#include "qemu/cutils.h"
@@ -109,12 +112,9 @@ index 0000000000..1257342724
 +    }
 +
 +    /* Open the raw file */
-+    bs->file = bdrv_open_child(qemu_opt_get(opts, "x-next"), options, "next",
-+                               bs, &child_of_bds,
-+                               BDRV_CHILD_FILTERED | BDRV_CHILD_PRIMARY,
-+                               false, &local_err);
-+    if (local_err) {
-+        ret = -EINVAL;
++    ret = bdrv_open_file_child(qemu_opt_get(opts, "x-next"), options, "next",
++                               bs, &local_err);
++    if (ret < 0) {
 +        error_propagate(errp, local_err);
 +        goto fail;
 +    }
@@ -125,7 +125,9 @@ index 0000000000..1257342724
 +    ret = 0;
 +fail:
 +    if (ret < 0) {
++        bdrv_graph_wrlock();
 +        bdrv_unref_child(bs, bs->file);
++        bdrv_graph_wrunlock();
 +    }
 +    qemu_opts_del(opts);
 +    return ret;
@@ -137,19 +139,22 @@ index 0000000000..1257342724
 +    (void)s;
 +}
 +
-+static coroutine_fn int64_t zeroinit_co_getlength(BlockDriverState *bs)
++static coroutine_fn int64_t GRAPH_RDLOCK
++zeroinit_co_getlength(BlockDriverState *bs)
 +{
 +    return bdrv_co_getlength(bs->file->bs);
 +}
 +
-+static int coroutine_fn zeroinit_co_preadv(BlockDriverState *bs,
-+    int64_t offset, int64_t bytes, QEMUIOVector *qiov, BdrvRequestFlags flags)
++static int coroutine_fn GRAPH_RDLOCK
++zeroinit_co_preadv(BlockDriverState *bs, int64_t offset, int64_t bytes,
++                   QEMUIOVector *qiov, BdrvRequestFlags flags)
 +{
 +    return bdrv_co_preadv(bs->file, offset, bytes, qiov, flags);
 +}
 +
-+static int coroutine_fn zeroinit_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset,
-+                                                 int64_t bytes, BdrvRequestFlags flags)
++static int coroutine_fn GRAPH_RDLOCK
++zeroinit_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset, int64_t bytes,
++                          BdrvRequestFlags flags)
 +{
 +    BDRVZeroinitState *s = bs->opaque;
 +    if (offset >= s->extents)
@@ -157,8 +162,9 @@ index 0000000000..1257342724
 +    return bdrv_pwrite_zeroes(bs->file, offset, bytes, flags);
 +}
 +
-+static int coroutine_fn zeroinit_co_pwritev(BlockDriverState *bs,
-+    int64_t offset, int64_t bytes, QEMUIOVector *qiov, BdrvRequestFlags flags)
++static int coroutine_fn GRAPH_RDLOCK
++zeroinit_co_pwritev(BlockDriverState *bs, int64_t offset, int64_t bytes,
++                    QEMUIOVector *qiov, BdrvRequestFlags flags)
 +{
 +    BDRVZeroinitState *s = bs->opaque;
 +    int64_t extents = offset + bytes;
@@ -167,32 +173,35 @@ index 0000000000..1257342724
 +    return bdrv_co_pwritev(bs->file, offset, bytes, qiov, flags);
 +}
 +
-+static coroutine_fn int zeroinit_co_flush(BlockDriverState *bs)
++static coroutine_fn int GRAPH_RDLOCK
++zeroinit_co_flush(BlockDriverState *bs)
 +{
 +    return bdrv_co_flush(bs->file->bs);
 +}
 +
-+static int zeroinit_has_zero_init(BlockDriverState *bs)
++static int GRAPH_RDLOCK
++zeroinit_has_zero_init(BlockDriverState *bs)
 +{
 +    BDRVZeroinitState *s = bs->opaque;
 +    return s->has_zero_init;
 +}
 +
-+static int coroutine_fn zeroinit_co_pdiscard(BlockDriverState *bs,
-+                                             int64_t offset, int64_t bytes)
++static int coroutine_fn GRAPH_RDLOCK
++zeroinit_co_pdiscard(BlockDriverState *bs, int64_t offset, int64_t bytes)
 +{
 +    return bdrv_co_pdiscard(bs->file, offset, bytes);
 +}
 +
-+static int zeroinit_co_truncate(BlockDriverState *bs, int64_t offset,
-+                                _Bool exact, PreallocMode prealloc,
-+                                BdrvRequestFlags req_flags, Error **errp)
++static int GRAPH_RDLOCK
++zeroinit_co_truncate(BlockDriverState *bs, int64_t offset, _Bool exact,
++                     PreallocMode prealloc, BdrvRequestFlags req_flags,
++                     Error **errp)
 +{
 +    return bdrv_co_truncate(bs->file, offset, exact, prealloc, req_flags, errp);
 +}
 +
-+static coroutine_fn int zeroinit_co_get_info(BlockDriverState *bs,
-+                                             BlockDriverInfo *bdi)
++static coroutine_fn int GRAPH_RDLOCK
++zeroinit_co_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
 +{
 +    return bdrv_co_get_info(bs->file->bs, bdi);
 +}
@@ -203,7 +212,7 @@ index 0000000000..1257342724
 +    .instance_size                    = sizeof(BDRVZeroinitState),
 +
 +    .bdrv_parse_filename              = zeroinit_parse_filename,
-+    .bdrv_file_open                   = zeroinit_open,
++    .bdrv_open                        = zeroinit_open,
 +    .bdrv_close                       = zeroinit_close,
 +    .bdrv_co_getlength                = zeroinit_co_getlength,
 +    .bdrv_child_perm                  = bdrv_default_perms,

debian/patches/pve/0020-PVE-Add-dummy-id-command-line-parameter.patch

@@ -10,14 +10,14 @@ Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
 Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
 ---
  qemu-options.hx | 3 +++
- softmmu/vl.c    | 8 ++++++++
+ system/vl.c     | 8 ++++++++
  2 files changed, 11 insertions(+)
 
 diff --git a/qemu-options.hx b/qemu-options.hx
-index fdd551c2bb..4eb43b7bc5 100644
+index c05f411599..0732077a0e 100644
 --- a/qemu-options.hx
 +++ b/qemu-options.hx
-@@ -1162,6 +1162,9 @@ legacy PC, they are not recommended for modern configurations.
+@@ -1239,6 +1239,9 @@ legacy PC, they are not recommended for modern configurations.
  
  ERST
  
@@ -27,11 +27,11 @@ index fdd551c2bb..4eb43b7bc5 100644
  DEF("fda", HAS_ARG, QEMU_OPTION_fda,
      "-fda/-fdb file  use 'file' as floppy disk 0/1 image\n", QEMU_ARCH_ALL)
  DEF("fdb", HAS_ARG, QEMU_OPTION_fdb, "", QEMU_ARCH_ALL)
-diff --git a/softmmu/vl.c b/softmmu/vl.c
-index 0eabc71b68..323f6a23d4 100644
---- a/softmmu/vl.c
-+++ b/softmmu/vl.c
-@@ -2648,6 +2648,7 @@ void qemu_init(int argc, char **argv)
+diff --git a/system/vl.c b/system/vl.c
+index 39d451bb41..e7cae51f13 100644
+--- a/system/vl.c
++++ b/system/vl.c
+@@ -2762,6 +2762,7 @@ void qemu_init(int argc, char **argv)
      MachineClass *machine_class;
      bool userconfig = true;
      FILE *vmstate_dump_file = NULL;
@@ -39,7 +39,7 @@ index 0eabc71b68..323f6a23d4 100644
  
      qemu_add_opts(&qemu_drive_opts);
      qemu_add_drive_opts(&qemu_legacy_drive_opts);
-@@ -3271,6 +3272,13 @@ void qemu_init(int argc, char **argv)
+@@ -3374,6 +3375,13 @@ void qemu_init(int argc, char **argv)
                  machine_parse_property_opt(qemu_find_opts("smp-opts"),
                                             "smp", optarg);
                  break;
@@ -50,6 +50,6 @@ index 0eabc71b68..323f6a23d4 100644
 +                    exit(1);
 +                }
 +                break;
+ #ifdef CONFIG_VNC
              case QEMU_OPTION_vnc:
                  vnc_parse(optarg);
-                 break;

debian/patches/pve/0021-PVE-Config-Revert-target-i386-disable-LINT0-after-re.patch

@@ -11,10 +11,10 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
  1 file changed, 9 insertions(+)
 
 diff --git a/hw/intc/apic_common.c b/hw/intc/apic_common.c
-index 4a34f03047..59b917e50c 100644
+index 62f3bbf203..89e0c7d995 100644
 --- a/hw/intc/apic_common.c
 +++ b/hw/intc/apic_common.c
-@@ -252,6 +252,15 @@ static void apic_reset_common(DeviceState *dev)
+@@ -263,6 +263,15 @@ static void apic_reset_common(DeviceState *dev)
      info->vapic_base_update(s);
  
      apic_init_reset(dev);

debian/patches/pve/0022-PVE-Up-Config-file-posix-make-locking-optiono-on-cre.patch

@@ -9,14 +9,14 @@ Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
 Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
 ---
  block/file-posix.c   | 59 ++++++++++++++++++++++++++++++--------------
- qapi/block-core.json |  3 ++-
- 2 files changed, 42 insertions(+), 20 deletions(-)
+ qapi/block-core.json |  7 +++++-
+ 2 files changed, 46 insertions(+), 20 deletions(-)
 
 diff --git a/block/file-posix.c b/block/file-posix.c
-index 9681bd0434..044890822d 100644
+index e2ea071315..4c3dc56c8e 100644
 --- a/block/file-posix.c
 +++ b/block/file-posix.c
-@@ -2483,6 +2483,7 @@ raw_co_create(BlockdevCreateOptions *options, Error **errp)
+@@ -2884,6 +2884,7 @@ raw_co_create(BlockdevCreateOptions *options, Error **errp)
      int fd;
      uint64_t perm, shared;
      int result = 0;
@@ -24,7 +24,7 @@ index 9681bd0434..044890822d 100644
  
      /* Validate options and set default values */
      assert(options->driver == BLOCKDEV_DRIVER_FILE);
-@@ -2523,19 +2524,22 @@ raw_co_create(BlockdevCreateOptions *options, Error **errp)
+@@ -2924,19 +2925,22 @@ raw_co_create(BlockdevCreateOptions *options, Error **errp)
      perm = BLK_PERM_WRITE | BLK_PERM_RESIZE;
      shared = BLK_PERM_ALL & ~BLK_PERM_RESIZE;
  
@@ -59,7 +59,7 @@ index 9681bd0434..044890822d 100644
      }
  
      /* Clear the file by truncating it to 0 */
-@@ -2589,13 +2593,15 @@ raw_co_create(BlockdevCreateOptions *options, Error **errp)
+@@ -2990,13 +2994,15 @@ raw_co_create(BlockdevCreateOptions *options, Error **errp)
      }
  
  out_unlock:
@@ -82,7 +82,7 @@ index 9681bd0434..044890822d 100644
      }
  
  out_close:
-@@ -2619,6 +2625,7 @@ raw_co_create_opts(BlockDriver *drv, const char *filename,
+@@ -3020,6 +3026,7 @@ raw_co_create_opts(BlockDriver *drv, const char *filename,
      PreallocMode prealloc;
      char *buf = NULL;
      Error *local_err = NULL;
@@ -90,7 +90,7 @@ index 9681bd0434..044890822d 100644
  
      /* Skip file: protocol prefix */
      strstart(filename, "file:", &filename);
-@@ -2641,6 +2648,18 @@ raw_co_create_opts(BlockDriver *drv, const char *filename,
+@@ -3042,6 +3049,18 @@ raw_co_create_opts(BlockDriver *drv, const char *filename,
          return -EINVAL;
      }
  
@@ -109,7 +109,7 @@ index 9681bd0434..044890822d 100644
      options = (BlockdevCreateOptions) {
          .driver     = BLOCKDEV_DRIVER_FILE,
          .u.file     = {
-@@ -2652,6 +2671,8 @@ raw_co_create_opts(BlockDriver *drv, const char *filename,
+@@ -3053,6 +3072,8 @@ raw_co_create_opts(BlockDriver *drv, const char *filename,
              .nocow              = nocow,
              .has_extent_size_hint = has_extent_size_hint,
              .extent_size_hint   = extent_size_hint,
@@ -119,10 +119,21 @@ index 9681bd0434..044890822d 100644
      };
      return raw_co_create(&options, errp);
 diff --git a/qapi/block-core.json b/qapi/block-core.json
-index 3c945c1f93..542add004b 100644
+index 48ba32049f..321d1fd0e1 100644
 --- a/qapi/block-core.json
 +++ b/qapi/block-core.json
-@@ -4740,7 +4740,8 @@
+@@ -4974,6 +4974,10 @@
+ # @extent-size-hint: Extent size hint to add to the image file; 0 for
+ #     not adding an extent size hint (default: 1 MB, since 5.1)
+ #
++# @locking: whether to enable file locking.  If set to 'auto', only
++#     enable when Open File Descriptor (OFD) locking API is available
++#     (default: auto).
++#
+ # Since: 2.12
+ ##
+ { 'struct': 'BlockdevCreateOptionsFile',
+@@ -4981,7 +4985,8 @@
              'size':                 'size',
              '*preallocation':       'PreallocMode',
              '*nocow':               'bool',

debian/patches/pve/0023-PVE-monitor-disable-oob-capability.patch

@@ -18,10 +18,10 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
  1 file changed, 1 insertion(+), 2 deletions(-)
 
 diff --git a/monitor/qmp.c b/monitor/qmp.c
-index 6b8cfcf6d8..3ec67e32d3 100644
+index eb181d5979..20fc0d20a6 100644
 --- a/monitor/qmp.c
 +++ b/monitor/qmp.c
-@@ -519,8 +519,7 @@ void monitor_init_qmp(Chardev *chr, bool pretty, Error **errp)
+@@ -534,8 +534,7 @@ void monitor_init_qmp(Chardev *chr, bool pretty, Error **errp)
      qemu_chr_fe_set_echo(&mon->common.chr, true);
  
      /* Note: we run QMP monitor in I/O thread when @chr supports that */

debian/patches/pve/0024-PVE-Compat-4.0-used-balloon-qemu-4-0-config-size-fal.patch

@@ -26,10 +26,10 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
  1 file changed, 2 insertions(+), 1 deletion(-)
 
 diff --git a/hw/core/machine.c b/hw/core/machine.c
-index 2f6ccf5623..a5927e92f1 100644
+index f29fe95964..2c327fc36a 100644
 --- a/hw/core/machine.c
 +++ b/hw/core/machine.c
-@@ -142,7 +142,8 @@ GlobalProperty hw_compat_4_0[] = {
+@@ -180,7 +180,8 @@ GlobalProperty hw_compat_4_0[] = {
      { "virtio-vga",     "edid", "false" },
      { "virtio-gpu-device", "edid", "false" },
      { "virtio-device", "use-started", "false" },

debian/patches/pve/0025-PVE-Allow-version-code-in-machine-type.patch

@@ -16,15 +16,15 @@ Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
 ---
  hw/core/machine-qmp-cmds.c |  5 +++++
  include/hw/boards.h        |  2 ++
- qapi/machine.json          |  4 +++-
- softmmu/vl.c               | 25 +++++++++++++++++++++++++
- 4 files changed, 35 insertions(+), 1 deletion(-)
+ qapi/machine.json          |  3 +++
+ system/vl.c                | 24 ++++++++++++++++++++++++
+ 4 files changed, 34 insertions(+)
 
 diff --git a/hw/core/machine-qmp-cmds.c b/hw/core/machine-qmp-cmds.c
-index 24595f618c..ee9cb0cd04 100644
+index 52a6d74820..362128842d 100644
 --- a/hw/core/machine-qmp-cmds.c
 +++ b/hw/core/machine-qmp-cmds.c
-@@ -107,6 +107,11 @@ MachineInfoList *qmp_query_machines(Error **errp)
+@@ -94,6 +94,11 @@ MachineInfoList *qmp_query_machines(bool has_compat_props, bool compat_props,
          if (strcmp(mc->name, MACHINE_GET_CLASS(current_machine)->name) == 0) {
              info->has_is_current = true;
              info->is_current = true;
@@ -37,65 +37,64 @@ index 24595f618c..ee9cb0cd04 100644
  
          if (mc->default_cpu_type) {
 diff --git a/include/hw/boards.h b/include/hw/boards.h
-index 6fbbfd56c8..61a526e97d 100644
+index 36fbb9b59d..d1741ea121 100644
 --- a/include/hw/boards.h
 +++ b/include/hw/boards.h
-@@ -232,6 +232,8 @@ struct MachineClass {
+@@ -268,6 +268,8 @@ struct MachineClass {
      const char *desc;
      const char *deprecation_reason;
  
 +    const char *pve_version;
 +
      void (*init)(MachineState *state);
-     void (*reset)(MachineState *state, ShutdownCause reason);
+     void (*reset)(MachineState *state, ResetType type);
      void (*wakeup)(MachineState *state);
 diff --git a/qapi/machine.json b/qapi/machine.json
-index c904280085..47f3facdb2 100644
+index 16366b774a..12cfd3f260 100644
 --- a/qapi/machine.json
 +++ b/qapi/machine.json
-@@ -159,6 +159,8 @@
+@@ -189,6 +189,8 @@
  #
  # @acpi: machine type supports ACPI (since 8.0)
  #
 +# @pve-version: custom PVE version suffix specified as 'machine+pveN'
 +#
- # Since: 1.2
- ##
- { 'struct': 'MachineInfo',
-@@ -166,7 +168,7 @@
-             '*is-default': 'bool', '*is-current': 'bool', 'cpu-max': 'int',
+ # @compat-props: The machine type's compatibility properties.  Only
+ #     present when query-machines argument @compat-props is true.
+ #     (since 9.1)
+@@ -205,6 +207,7 @@
              'hotpluggable-cpus': 'bool',  'numa-mem-supported': 'bool',
              'deprecated': 'bool', '*default-cpu-type': 'str',
--            '*default-ram-id': 'str', 'acpi': 'bool' } }
-+            '*default-ram-id': 'str', 'acpi': 'bool', '*pve-version': 'str' } }
+             '*default-ram-id': 'str', 'acpi': 'bool',
++            '*pve-version': 'str',
+             '*compat-props': { 'type': ['CompatProperty'],
+                                'features': ['unstable'] } } }
  
- ##
- # @query-machines:
-diff --git a/softmmu/vl.c b/softmmu/vl.c
-index 323f6a23d4..25abdc9da7 100644
---- a/softmmu/vl.c
-+++ b/softmmu/vl.c
-@@ -1578,6 +1578,7 @@ static const QEMUOption *lookup_opt(int argc, char **argv,
- static MachineClass *select_machine(QDict *qdict, Error **errp)
+diff --git a/system/vl.c b/system/vl.c
+index e7cae51f13..3f4916ac5a 100644
+--- a/system/vl.c
++++ b/system/vl.c
+@@ -1675,6 +1675,7 @@ static MachineClass *select_machine(QDict *qdict, Error **errp)
  {
-     const char *optarg = qdict_get_try_str(qdict, "type");
+     ERRP_GUARD();
+     const char *machine_type = qdict_get_try_str(qdict, "type");
 +    const char *pvever = qdict_get_try_str(qdict, "pvever");
-     GSList *machines = object_class_get_list(TYPE_MACHINE, false);
-     MachineClass *machine_class;
-     Error *local_err = NULL;
-@@ -1595,6 +1596,11 @@ static MachineClass *select_machine(QDict *qdict, Error **errp)
-         }
-     }
+     g_autoptr(GSList) machines = object_class_get_list(TYPE_MACHINE, false);
+     MachineClass *machine_class = NULL;
  
-+    if (machine_class) {
+@@ -1694,7 +1695,11 @@ static MachineClass *select_machine(QDict *qdict, Error **errp)
+     if (!machine_class) {
+         error_append_hint(errp,
+                           "Use -machine help to list supported machines\n");
++    } else {
 +        machine_class->pve_version = g_strdup(pvever);
 +        qdict_del(qdict, "pvever");
-+    }
+     }
 +
-     g_slist_free(machines);
-     if (local_err) {
-         error_append_hint(&local_err, "Use -machine help to list supported machines\n");
-@@ -3213,12 +3219,31 @@ void qemu_init(int argc, char **argv)
+     return machine_class;
+ }
+ 
+@@ -3316,12 +3321,31 @@ void qemu_init(int argc, char **argv)
              case QEMU_OPTION_machine:
                  {
                      bool help;

debian/patches/pve/0026-block-backup-move-bcs-bitmap-initialization-to-job-c.patch

@@ -25,7 +25,7 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
  1 file changed, 4 insertions(+), 4 deletions(-)
 
 diff --git a/block/backup.c b/block/backup.c
-index db3791f4d1..39410dcf8d 100644
+index a1292c01ec..2e38b30d67 100644
 --- a/block/backup.c
 +++ b/block/backup.c
 @@ -237,8 +237,8 @@ static void backup_init_bcs_bitmap(BackupBlockJob *job)
@@ -48,9 +48,9 @@ index db3791f4d1..39410dcf8d 100644
      if (s->sync_mode == MIRROR_SYNC_MODE_TOP) {
          int64_t offset = 0;
          int64_t count;
-@@ -495,6 +493,8 @@ BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs,
-     block_job_add_bdrv(&job->common, "target", target, 0, BLK_PERM_ALL,
+@@ -502,6 +500,8 @@ BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs,
                         &error_abort);
+     bdrv_graph_wrunlock();
  
 +    backup_init_bcs_bitmap(job);
 +

debian/patches/pve/0027-PVE-Backup-add-vma-backup-format-code.patch

@@ -10,39 +10,40 @@ skip reads.
 
 Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
 [FE: improvements during create
-     allow partial restore]
+     allow partial restore
+     allow specifying disk formats for create operation]
 Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
 ---
  block/meson.build |   2 +
  meson.build       |   5 +
- vma-reader.c      | 867 +++++++++++++++++++++++++++++++++++++++++++++
- vma-writer.c      | 793 +++++++++++++++++++++++++++++++++++++++++
- vma.c             | 878 ++++++++++++++++++++++++++++++++++++++++++++++
+ vma-reader.c      | 867 ++++++++++++++++++++++++++++++++++++++++++
+ vma-writer.c      | 816 ++++++++++++++++++++++++++++++++++++++++
+ vma.c             | 941 ++++++++++++++++++++++++++++++++++++++++++++++
  vma.h             | 150 ++++++++
- 6 files changed, 2695 insertions(+)
+ 6 files changed, 2781 insertions(+)
  create mode 100644 vma-reader.c
  create mode 100644 vma-writer.c
  create mode 100644 vma.c
  create mode 100644 vma.h
 
 diff --git a/block/meson.build b/block/meson.build
-index 253fe49fa2..744b698a82 100644
+index 6a60b5d6b9..652c8cbdb7 100644
 --- a/block/meson.build
 +++ b/block/meson.build
-@@ -47,6 +47,8 @@ block_ss.add(files(
+@@ -42,6 +42,8 @@ block_ss.add(files(
    'zeroinit.c',
- ), zstd, zlib, gnutls)
+ ), zstd, zlib)
  
 +block_ss.add(files('../vma-writer.c'), libuuid)
 +
- softmmu_ss.add(when: 'CONFIG_TCG', if_true: files('blkreplay.c'))
- softmmu_ss.add(files('block-ram-registrar.c'))
+ system_ss.add(when: 'CONFIG_TCG', if_true: files('blkreplay.c'))
+ system_ss.add(files('block-ram-registrar.c'))
  
 diff --git a/meson.build b/meson.build
-index 30447cfaef..38a4e2bcef 100644
+index 147097c652..b9b673c271 100644
 --- a/meson.build
 +++ b/meson.build
-@@ -1527,6 +1527,8 @@ keyutils = dependency('libkeyutils', required: false,
+@@ -2129,6 +2129,8 @@ endif
  
  has_gettid = cc.has_function('gettid')
  
@@ -51,19 +52,19 @@ index 30447cfaef..38a4e2bcef 100644
  # libselinux
  selinux = dependency('libselinux',
                       required: get_option('selinux'),
-@@ -3650,6 +3652,9 @@ if have_tools
-                dependencies: [blockdev, qemuutil, gnutls, selinux],
+@@ -4344,6 +4346,9 @@ if have_tools
+                dependencies: [blockdev, qemuutil, selinux],
                 install: true)
  
 +  vma = executable('vma', files('vma.c', 'vma-reader.c') + genh,
-+                   dependencies: [authz, block, crypto, io, qom], install: true)
++                   dependencies: [authz, block, crypto, io, qemuutil, qom], install: true)
 +
    subdir('storage-daemon')
-   subdir('contrib/rdmacm-mux')
-   subdir('contrib/elf2dmp')
+ 
+   foreach exe: [ 'qemu-img', 'qemu-io', 'qemu-nbd', 'qemu-storage-daemon']
 diff --git a/vma-reader.c b/vma-reader.c
 new file mode 100644
-index 0000000000..81a891c6b1
+index 0000000000..bb65ad313c
 --- /dev/null
 +++ b/vma-reader.c
 @@ -0,0 +1,867 @@
@@ -88,6 +89,7 @@ index 0000000000..81a891c6b1
 +#include "qemu/ratelimit.h"
 +#include "vma.h"
 +#include "block/block.h"
++#include "block/graph-lock.h"
 +#include "sysemu/block-backend.h"
 +
 +static unsigned char zero_vma_block[VMA_BLOCK_SIZE];
@@ -380,7 +382,6 @@ index 0000000000..81a891c6b1
 +    }
 +
 +
-+    int count = 0;
 +    for (i = 1; i < 256; i++) {
 +        VmaDeviceInfoHeader *dih = &h->dev_info[i];
 +        uint32_t devname_ptr = GUINT32_FROM_BE(dih->devname_ptr);
@@ -388,7 +389,6 @@ index 0000000000..81a891c6b1
 +        const char *devname =  get_header_str(vmar, devname_ptr);
 +
 +        if (size && devname) {
-+            count++;
 +            vmar->devinfo[i].size = size;
 +            vmar->devinfo[i].devname = devname;
 +
@@ -600,8 +600,10 @@ index 0000000000..81a891c6b1
 +    } else {
 +        int res = blk_pwrite(target, sector_num * BDRV_SECTOR_SIZE, nb_sectors * BDRV_SECTOR_SIZE, buf, 0);
 +        if (res < 0) {
++            bdrv_graph_rdlock_main_loop();
 +            error_setg(errp, "blk_pwrite to %s failed (%d)",
 +                       bdrv_get_device_name(blk_bs(target)), res);
++            bdrv_graph_rdunlock_main_loop();
 +            return -1;
 +        }
 +    }
@@ -881,8 +883,7 @@ index 0000000000..81a891c6b1
 +
 +            int64_t cluster_num, end;
 +
-+            end = (vmar->devinfo[i].size + VMA_CLUSTER_SIZE - 1) /
-+                VMA_CLUSTER_SIZE;
++            end = DIV_ROUND_UP(vmar->devinfo[i].size, VMA_CLUSTER_SIZE);
 +
 +            for (cluster_num = 0; cluster_num < end; cluster_num++) {
 +                if (!vma_reader_get_bitmap(rstate, cluster_num)) {
@@ -936,10 +937,10 @@ index 0000000000..81a891c6b1
 +
 diff --git a/vma-writer.c b/vma-writer.c
 new file mode 100644
-index 0000000000..ac7da237d0
+index 0000000000..3f489092df
 --- /dev/null
 +++ b/vma-writer.c
-@@ -0,0 +1,793 @@
+@@ -0,0 +1,816 @@
 +/*
 + * VMA: Virtual Machine Archive
 + *
@@ -955,6 +956,8 @@ index 0000000000..ac7da237d0
 +
 +#include "qemu/osdep.h"
 +#include <glib.h>
++#include <linux/magic.h>
++#include <sys/vfs.h>
 +#include <uuid/uuid.h>
 +
 +#include "vma.h"
@@ -963,6 +966,7 @@ index 0000000000..ac7da237d0
 +#include "qemu/main-loop.h"
 +#include "qemu/coroutine.h"
 +#include "qemu/cutils.h"
++#include "qemu/error-report.h"
 +#include "qemu/memalign.h"
 +
 +#define DEBUG_VMA 0
@@ -1130,8 +1134,7 @@ index 0000000000..ac7da237d0
 +    vmaw->stream_info[n].devname = g_strdup(devname);
 +    vmaw->stream_info[n].size = size;
 +
-+    vmaw->stream_info[n].cluster_count = (size + VMA_CLUSTER_SIZE - 1) /
-+        VMA_CLUSTER_SIZE;
++    vmaw->stream_info[n].cluster_count = DIV_ROUND_UP(size, VMA_CLUSTER_SIZE);
 +
 +    vmaw->stream_count = n;
 +
@@ -1146,10 +1149,10 @@ index 0000000000..ac7da237d0
 +{
 +    assert(qemu_in_coroutine());
 +    AioContext *ctx = qemu_get_current_aio_context();
-+    aio_set_fd_handler(ctx, fd, false, NULL, (IOHandler *)qemu_coroutine_enter,
-+                       NULL, NULL, qemu_coroutine_self());
++    aio_set_fd_handler(ctx, fd, NULL, (IOHandler *)qemu_coroutine_enter, NULL,
++                       NULL, qemu_coroutine_self());
 +    qemu_coroutine_yield();
-+    aio_set_fd_handler(ctx, fd, false, NULL, NULL, NULL, NULL, NULL);
++    aio_set_fd_handler(ctx, fd, NULL, NULL, NULL, NULL, NULL);
 +}
 +
 +static ssize_t coroutine_fn
@@ -1198,6 +1201,23 @@ index 0000000000..ac7da237d0
 +    return (done == bytes) ? bytes : -1;
 +}
 +
++static bool is_path_tmpfs(const char *path) {
++    struct statfs fs;
++    int ret;
++
++    do {
++        ret = statfs(path, &fs);
++    } while (ret != 0 && errno == EINTR);
++
++    if (ret != 0) {
++        warn_report("statfs call for %s failed, assuming not tmpfs - %s\n",
++                    path, strerror(errno));
++        return false;
++    }
++
++    return fs.f_type == TMPFS_MAGIC;
++}
++
 +VmaWriter *vma_writer_create(const char *filename, uuid_t uuid, Error **errp)
 +{
 +    const char *p;
@@ -1247,8 +1267,13 @@ index 0000000000..ac7da237d0
 +            }
 +            /* try to use O_NONBLOCK */
 +            fcntl(vmaw->fd, F_SETFL, fcntl(vmaw->fd, F_GETFL)|O_NONBLOCK);
-+        } else  {
-+            oflags = O_NONBLOCK|O_DIRECT|O_WRONLY|O_EXCL;
++        } else {
++            gchar *dirname = g_path_get_dirname(filename);
++            oflags = O_NONBLOCK|O_WRONLY|O_EXCL;
++            if (!is_path_tmpfs(dirname)) {
++                oflags |= O_DIRECT;
++            }
++            g_free(dirname);
 +            vmaw->fd = qemu_create(filename, oflags, 0644, errp);
 +        }
 +
@@ -1489,17 +1514,16 @@ index 0000000000..ac7da237d0
 +    int i;
 +
 +    g_assert(vmaw != NULL);
++    g_assert(status != NULL);
 +
-+    if (status) {
-+        status->status = vmaw->status;
-+        g_strlcpy(status->errmsg, vmaw->errmsg, sizeof(status->errmsg));
-+        for (i = 0; i <= 255; i++) {
-+            status->stream_info[i] = vmaw->stream_info[i];
-+        }
-+
-+        uuid_unparse_lower(vmaw->uuid, status->uuid_str);
++    status->status = vmaw->status;
++    g_strlcpy(status->errmsg, vmaw->errmsg, sizeof(status->errmsg));
++    for (i = 0; i <= 255; i++) {
++        status->stream_info[i] = vmaw->stream_info[i];
 +    }
 +
++    uuid_unparse_lower(vmaw->uuid, status->uuid_str);
++
 +    status->closed = vmaw->closed;
 +
 +    return vmaw->status;
@@ -1735,10 +1759,10 @@ index 0000000000..ac7da237d0
 +}
 diff --git a/vma.c b/vma.c
 new file mode 100644
-index 0000000000..304f02bc84
+index 0000000000..8d4b4be414
 --- /dev/null
 +++ b/vma.c
-@@ -0,0 +1,878 @@
+@@ -0,0 +1,941 @@
 +/*
 + * VMA: Virtual Machine Archive
 + *
@@ -1770,9 +1794,9 @@ index 0000000000..304f02bc84
 +        "usage: vma command [command options]\n"
 +        "\n"
 +        "vma list <filename>\n"
-+        "vma config <filename> [-c config]\n"
-+        "vma create <filename> [-c config] pathname ...\n"
-+        "vma extract <filename> [-r <fifo>] <targetdir>\n"
++        "vma config <filename> [-c <config>]\n"
++        "vma create <filename> [-c <config>] [-d format=<format>:<device name>=<path> [-d ...]]\n"
++        "vma extract <filename> [-d <drive-list>] [-r <fifo>] <targetdir>\n"
 +        "vma verify <filename> [-v]\n"
 +        ;
 +
@@ -1917,9 +1941,10 @@ index 0000000000..304f02bc84
 +    const char *filename;
 +    const char *dirname;
 +    const char *readmap = NULL;
++    gchar **drive_list = NULL;
 +
 +    for (;;) {
-+        c = getopt(argc, argv, "hvr:");
++        c = getopt(argc, argv, "hvd:r:");
 +        if (c == -1) {
 +            break;
 +        }
@@ -1928,6 +1953,9 @@ index 0000000000..304f02bc84
 +        case 'h':
 +            help();
 +            break;
++        case 'd':
++            drive_list = g_strsplit(optarg, ",", 254);
++            break;
 +        case 'r':
 +            readmap = optarg;
 +            break;
@@ -2047,29 +2075,29 @@ index 0000000000..304f02bc84
 +                        inbuf);
 +            }
 +
-+            RestoreMap *map = g_new0(RestoreMap, 1);
-+            map->devname = g_strdup(devname);
-+            map->path = g_strdup(path);
-+            map->format = format;
-+            map->throttling_bps = bps_value;
-+            map->throttling_group = group;
-+            map->cache = cache;
-+            map->write_zero = write_zero;
-+            map->skip = skip;
++            RestoreMap *restore_map = g_new0(RestoreMap, 1);
++            restore_map->devname = g_strdup(devname);
++            restore_map->path = g_strdup(path);
++            restore_map->format = format;
++            restore_map->throttling_bps = bps_value;
++            restore_map->throttling_group = group;
++            restore_map->cache = cache;
++            restore_map->write_zero = write_zero;
++            restore_map->skip = skip;
 +
-+            g_hash_table_insert(devmap, map->devname, map);
++            g_hash_table_insert(devmap, restore_map->devname, restore_map);
 +
 +        };
 +    }
 +
 +    int i;
 +    int vmstate_fd = -1;
-+    guint8 vmstate_stream = 0;
++    bool drive_rename_bitmap[255];
++    memset(drive_rename_bitmap, 0, sizeof(drive_rename_bitmap));
 +
 +    for (i = 1; i < 255; i++) {
 +        VmaDeviceInfo *di = vma_reader_get_device_info(vmar, i);
 +        if (di && (strcmp(di->devname, "vmstate") == 0)) {
-+            vmstate_stream = i;
 +            char *statefn = g_strdup_printf("%s/vmstate.bin", dirname);
 +            vmstate_fd = open(statefn, O_WRONLY|O_CREAT|O_EXCL, 0644);
 +            if (vmstate_fd < 0) {
@@ -2089,7 +2117,21 @@ index 0000000000..304f02bc84
 +
 +            BlockBackend *blk = NULL;
 +
-+            if (readmap) {
++            if (drive_list) {
++                skip = true;
++                int j;
++                for (j = 0; drive_list[j]; j++) {
++                    if (strcmp(drive_list[j], di->devname) == 0) {
++                        skip = false;
++                        drive_rename_bitmap[i] = true;
++                        break;
++                    }
++                }
++            } else {
++                drive_rename_bitmap[i] = true;
++            }
++
++            if (!skip && readmap) {
 +                RestoreMap *map;
 +                map = (RestoreMap *)g_hash_table_lookup(devmap, di->devname);
 +                if (map == NULL) {
@@ -2102,7 +2144,7 @@ index 0000000000..304f02bc84
 +                cache = map->cache;
 +                write_zero = map->write_zero;
 +                skip = map->skip;
-+            } else {
++            } else if (!skip) {
 +                devfn = g_strdup_printf("%s/tmp-disk-%s.raw",
 +                                        dirname, di->devname);
 +                printf("DEVINFO %s %zd\n", devfn, di->size);
@@ -2183,6 +2225,10 @@ index 0000000000..304f02bc84
 +        }
 +    }
 +
++    if (drive_list) {
++        g_strfreev(drive_list);
++    }
++
 +    if (vma_reader_restore(vmar, vmstate_fd, verbose, &errp) < 0) {
 +        g_error("restore failed - %s", error_get_pretty(errp));
 +    }
@@ -2190,7 +2236,7 @@ index 0000000000..304f02bc84
 +    if (!readmap) {
 +        for (i = 1; i < 255; i++) {
 +            VmaDeviceInfo *di = vma_reader_get_device_info(vmar, i);
-+            if (di && (i != vmstate_stream)) {
++            if (di && drive_rename_bitmap[i]) {
 +                char *tmpfn = g_strdup_printf("%s/tmp-disk-%s.raw",
 +                                              dirname, di->devname);
 +                char *fn = g_strdup_printf("%s/disk-%s.raw",
@@ -2314,13 +2360,13 @@ index 0000000000..304f02bc84
 +        ret = blk_co_preadv(job->target, start * VMA_CLUSTER_SIZE,
 +                            readlen, &qiov, 0);
 +        if (ret < 0) {
-+            vma_writer_set_error(job->vmaw, "read error", -1);
++            vma_writer_set_error(job->vmaw, "read error");
 +            goto out;
 +        }
 +
 +        size_t zb = 0;
 +        if (vma_writer_write(job->vmaw, job->dev_id, start, buf, &zb) < 0) {
-+            vma_writer_set_error(job->vmaw, "backup_dump_cb vma_writer_write failed", -1);
++            vma_writer_set_error(job->vmaw, "backup_dump_cb vma_writer_write failed");
 +            goto out;
 +        }
 +    }
@@ -2338,14 +2384,16 @@ index 0000000000..304f02bc84
 +
 +static int create_archive(int argc, char **argv)
 +{
-+    int i, c;
++    int c;
 +    int verbose = 0;
++    bool expect_format = true;
 +    const char *archivename;
 +    GList *backup_coroutines = NULL;
 +    GList *config_files = NULL;
++    GList *disk_infos = NULL;
 +
 +    for (;;) {
-+        c = getopt(argc, argv, "hvc:");
++        c = getopt(argc, argv, "hvc:d:");
 +        if (c == -1) {
 +            break;
 +        }
@@ -2357,6 +2405,9 @@ index 0000000000..304f02bc84
 +        case 'c':
 +            config_files = g_list_append(config_files, optarg);
 +            break;
++        case 'd':
++            disk_infos = g_list_append(disk_infos, optarg);
++            break;
 +        case 'v':
 +            verbose = 1;
 +            break;
@@ -2402,16 +2453,48 @@ index 0000000000..304f02bc84
 +        l = g_list_next(l);
 +    }
 +
-+    int devcount = 0;
++    /*
++     * Don't allow mixing new and old way to specifiy disks.
++     * TODO PVE 9 drop old way and always require format.
++     */
++    if (optind < argc && g_list_first(disk_infos)) {
++        unlink(archivename);
++        g_error("Unexpected extra argument - specify all devices via '-d'");
++    }
++
 +    while (optind < argc) {
-+        const char *path = argv[optind++];
++        expect_format = false;
++        disk_infos = g_list_append(disk_infos, argv[optind++]);
++    }
++
++    int devcount = 0;
++    GList *disk_l = disk_infos;
++    while (disk_l && disk_l->data) {
++        char *disk_info = disk_l->data;
++        const char *path = NULL;
 +        char *devname = NULL;
-+        path = extract_devname(path, &devname, devcount++);
++        char *format = NULL;
++        QDict *options = qdict_new();
++
++        if (try_parse_option(&disk_info, "format", &format, disk_info)) {
++            qdict_put_str(options, "driver", format);
++        } else {
++            if (expect_format) {
++                unlink(archivename);
++                g_error("No format specified for device: '%s'", disk_info);
++            } else {
++                g_warning("Specifying a device without a format is deprecated"
++                          " - use '-d format=<format>:%s'",
++                          disk_info);
++            }
++        }
++
++        path = extract_devname(disk_info, &devname, devcount++);
 +
 +        Error *errp = NULL;
 +        BlockBackend *target;
 +
-+        target = blk_new_open(path, NULL, NULL, 0, &errp);
++        target = blk_new_open(path, NULL, options, 0, &errp);
 +        if (!target) {
 +            unlink(archivename);
 +            g_error("bdrv_open '%s' failed - %s", path, error_get_pretty(errp));
@@ -2433,6 +2516,8 @@ index 0000000000..304f02bc84
 +        // Don't enter coroutine yet, because it might write the header before
 +        // all streams can be registered.
 +        backup_coroutines = g_list_append(backup_coroutines, co);
++
++        disk_l = g_list_next(disk_l);
 +    }
 +
 +    VmaStatus vmastat;
@@ -2496,6 +2581,7 @@ index 0000000000..304f02bc84
 +    vma_writer_get_status(vmaw, &vmastat);
 +
 +    if (verbose) {
++        int i;
 +        for (i = 0; i < 256; i++) {
 +            VmaStreamInfo *si = &vmastat.stream_info[i];
 +            if (si->size) {
@@ -2513,6 +2599,7 @@ index 0000000000..304f02bc84
 +
 +    g_list_free(backup_coroutines);
 +    g_list_free(config_files);
++    g_list_free(disk_infos);
 +    vma_writer_destroy(vmaw);
 +    return 0;
 +}
@@ -2619,7 +2706,7 @@ index 0000000000..304f02bc84
 +}
 diff --git a/vma.h b/vma.h
 new file mode 100644
-index 0000000000..1b62859165
+index 0000000000..86d2873aa5
 --- /dev/null
 +++ b/vma.h
 @@ -0,0 +1,150 @@
@@ -2757,7 +2844,7 @@ index 0000000000..1b62859165
 +int coroutine_fn vma_writer_flush_output(VmaWriter *vmaw);
 +
 +int vma_writer_get_status(VmaWriter *vmaw, VmaStatus *status);
-+void vma_writer_set_error(VmaWriter *vmaw, const char *fmt, ...);
++void vma_writer_set_error(VmaWriter *vmaw, const char *fmt, ...) G_GNUC_PRINTF(2, 3);
 +
 +
 +VmaReader *vma_reader_create(const char *filename, Error **errp);

debian/patches/pve/0028-PVE-Backup-add-backup-dump-block-driver.patch

@@ -12,20 +12,20 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
 [FE: adapt to coroutine changes]
 Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
 ---
- block/backup-dump.c              | 168 +++++++++++++++++++++++++++++++
+ block/backup-dump.c              | 172 +++++++++++++++++++++++++++++++
  block/backup.c                   |  30 ++----
  block/meson.build                |   1 +
  include/block/block_int-common.h |  35 +++++++
  job.c                            |   3 +-
- 5 files changed, 214 insertions(+), 23 deletions(-)
+ 5 files changed, 218 insertions(+), 23 deletions(-)
  create mode 100644 block/backup-dump.c
 
 diff --git a/block/backup-dump.c b/block/backup-dump.c
 new file mode 100644
-index 0000000000..232a094426
+index 0000000000..e46abf1070
 --- /dev/null
 +++ b/block/backup-dump.c
-@@ -0,0 +1,168 @@
+@@ -0,0 +1,172 @@
 +/*
 + * BlockDriver to send backup data stream to a callback function
 + *
@@ -37,6 +37,8 @@ index 0000000000..232a094426
 + */
 +
 +#include "qemu/osdep.h"
++
++#include "qapi/qmp/qdict.h"
 +#include "qom/object_interfaces.h"
 +#include "block/block_int.h"
 +
@@ -169,7 +171,7 @@ index 0000000000..232a094426
 +block_init(bdrv_backup_dump_init);
 +
 +
-+BlockDriverState *bdrv_backup_dump_create(
++BlockDriverState *coroutine_fn bdrv_co_backup_dump_create(
 +    int dump_cb_block_size,
 +    uint64_t byte_size,
 +    BackupDumpFunc *dump_cb,
@@ -177,9 +179,11 @@ index 0000000000..232a094426
 +    Error **errp)
 +{
 +    BDRVBackupDumpState *state;
-+    BlockDriverState *bs = bdrv_new_open_driver(
-+        &bdrv_backup_dump_drive, NULL, BDRV_O_RDWR, errp);
 +
++    QDict *options = qdict_new();
++    qdict_put_str(options, "driver", "backup-dump-drive");
++
++    BlockDriverState *bs = bdrv_co_open(NULL, NULL, options, BDRV_O_RDWR, errp);
 +    if (!bs) {
 +        return NULL;
 +    }
@@ -195,7 +199,7 @@ index 0000000000..232a094426
 +    return bs;
 +}
 diff --git a/block/backup.c b/block/backup.c
-index 39410dcf8d..af87fa6aa9 100644
+index 2e38b30d67..fe69723ada 100644
 --- a/block/backup.c
 +++ b/block/backup.c
 @@ -29,28 +29,6 @@
@@ -227,7 +231,7 @@ index 39410dcf8d..af87fa6aa9 100644
  static const BlockJobDriver backup_job_driver;
  
  static void backup_cleanup_sync_bitmap(BackupBlockJob *job, int ret)
-@@ -457,6 +435,14 @@ BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs,
+@@ -462,6 +440,14 @@ BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs,
      }
  
      cluster_size = block_copy_cluster_size(bcs);
@@ -243,7 +247,7 @@ index 39410dcf8d..af87fa6aa9 100644
      if (perf->max_chunk && perf->max_chunk < cluster_size) {
          error_setg(errp, "Required max-chunk (%" PRIi64 ") is less than backup "
 diff --git a/block/meson.build b/block/meson.build
-index 744b698a82..f580f95395 100644
+index 652c8cbdb7..e1cf5a2e65 100644
 --- a/block/meson.build
 +++ b/block/meson.build
 @@ -4,6 +4,7 @@ block_ss.add(files(
@@ -251,11 +255,11 @@ index 744b698a82..f580f95395 100644
    'amend.c',
    'backup.c',
 +  'backup-dump.c',
-   'copy-before-write.c',
    'blkdebug.c',
    'blklogwrites.c',
+   'blkverify.c',
 diff --git a/include/block/block_int-common.h b/include/block/block_int-common.h
-index f01bb8b617..d7ffd1826e 100644
+index ebb4e56a50..e717a74e5f 100644
 --- a/include/block/block_int-common.h
 +++ b/include/block/block_int-common.h
 @@ -26,6 +26,7 @@
@@ -272,7 +276,7 @@ index f01bb8b617..d7ffd1826e 100644
  
 +typedef int BackupDumpFunc(void *opaque, uint64_t offset, uint64_t bytes, const void *buf);
 +
-+BlockDriverState *bdrv_backup_dump_create(
++BlockDriverState *coroutine_fn bdrv_co_backup_dump_create(
 +    int dump_cb_block_size,
 +    uint64_t byte_size,
 +    BackupDumpFunc *dump_cb,
@@ -308,10 +312,10 @@ index f01bb8b617..d7ffd1826e 100644
      BDRV_TRACKED_READ,
      BDRV_TRACKED_WRITE,
 diff --git a/job.c b/job.c
-index 72d57f0934..93e22d180b 100644
+index 660ce22c56..baf54c8d60 100644
 --- a/job.c
 +++ b/job.c
-@@ -330,7 +330,8 @@ static bool job_started_locked(Job *job)
+@@ -331,7 +331,8 @@ static bool job_started_locked(Job *job)
  }
  
  /* Called with job_mutex held. */

debian/patches/pve/0029-PVE-Add-sequential-job-transaction-support.patch

@@ -11,10 +11,10 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
  2 files changed, 46 insertions(+)
 
 diff --git a/include/qemu/job.h b/include/qemu/job.h
-index e502787dd8..963cf2bef5 100644
+index 2b873f2576..528cd6acb9 100644
 --- a/include/qemu/job.h
 +++ b/include/qemu/job.h
-@@ -381,6 +381,18 @@ void job_unlock(void);
+@@ -362,6 +362,18 @@ void job_unlock(void);
   */
  JobTxn *job_txn_new(void);
  
@@ -34,10 +34,10 @@ index e502787dd8..963cf2bef5 100644
   * Release a reference that was previously acquired with job_txn_add_job or
   * job_txn_new. If it's the last reference to the object, it will be freed.
 diff --git a/job.c b/job.c
-index 93e22d180b..2b31f1e14f 100644
+index baf54c8d60..3ac5e5cde2 100644
 --- a/job.c
 +++ b/job.c
-@@ -93,6 +93,8 @@ struct JobTxn {
+@@ -94,6 +94,8 @@ struct JobTxn {
  
      /* Reference count */
      int refcnt;
@@ -46,7 +46,7 @@ index 93e22d180b..2b31f1e14f 100644
  };
  
  void job_lock(void)
-@@ -118,6 +120,25 @@ JobTxn *job_txn_new(void)
+@@ -119,6 +121,25 @@ JobTxn *job_txn_new(void)
      return txn;
  }
  
@@ -72,7 +72,7 @@ index 93e22d180b..2b31f1e14f 100644
  /* Called with job_mutex held. */
  static void job_txn_ref_locked(JobTxn *txn)
  {
-@@ -1057,6 +1078,12 @@ static void job_completed_txn_success_locked(Job *job)
+@@ -1042,6 +1063,12 @@ static void job_completed_txn_success_locked(Job *job)
       */
      QLIST_FOREACH(other_job, &txn->jobs, txn_list) {
          if (!job_is_completed_locked(other_job)) {
@@ -85,7 +85,7 @@ index 93e22d180b..2b31f1e14f 100644
              return;
          }
          assert(other_job->ret == 0);
-@@ -1268,6 +1295,13 @@ int job_finish_sync_locked(Job *job,
+@@ -1253,6 +1280,13 @@ int job_finish_sync_locked(Job *job,
          return -EBUSY;
      }
  

debian/patches/pve/0030-PVE-Backup-Proxmox-backup-patches-for-QEMU.patch

@@ -79,34 +79,36 @@ Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
      adapt for new job lock mechanism replacing AioContext locks
      adapt to QAPI changes
      improve canceling
-     allow passing max-workers setting]
+     allow passing max-workers setting
+     use malloc_trim after backup
+     create jobs in a drained section]
 Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
 ---
  block/meson.build              |    5 +
- block/monitor/block-hmp-cmds.c |   40 ++
+ block/monitor/block-hmp-cmds.c |   39 ++
  blockdev.c                     |    1 +
  hmp-commands-info.hx           |   14 +
- hmp-commands.hx                |   31 +
+ hmp-commands.hx                |   29 +
  include/monitor/hmp.h          |    3 +
  meson.build                    |    1 +
  monitor/hmp-cmds.c             |   72 +++
  proxmox-backup-client.c        |  146 +++++
  proxmox-backup-client.h        |   60 ++
- pve-backup.c                   | 1097 ++++++++++++++++++++++++++++++++
- qapi/block-core.json           |  226 +++++++
- qapi/common.json               |   13 +
- qapi/machine.json              |   15 +-
- 14 files changed, 1711 insertions(+), 13 deletions(-)
+ pve-backup.c                   | 1090 ++++++++++++++++++++++++++++++++
+ qapi/block-core.json           |  233 +++++++
+ qapi/common.json               |   14 +
+ qapi/machine.json              |   16 +-
+ 14 files changed, 1709 insertions(+), 14 deletions(-)
  create mode 100644 proxmox-backup-client.c
  create mode 100644 proxmox-backup-client.h
  create mode 100644 pve-backup.c
 
 diff --git a/block/meson.build b/block/meson.build
-index f580f95395..5bcebb934b 100644
+index e1cf5a2e65..2367e1ac1b 100644
 --- a/block/meson.build
 +++ b/block/meson.build
-@@ -49,6 +49,11 @@ block_ss.add(files(
- ), zstd, zlib, gnutls)
+@@ -44,6 +44,11 @@ block_ss.add(files(
+ ), zstd, zlib)
  
  block_ss.add(files('../vma-writer.c'), libuuid)
 +block_ss.add(files(
@@ -115,13 +117,13 @@ index f580f95395..5bcebb934b 100644
 +), libproxmox_backup_qemu)
 +
  
- softmmu_ss.add(when: 'CONFIG_TCG', if_true: files('blkreplay.c'))
- softmmu_ss.add(files('block-ram-registrar.c'))
+ system_ss.add(when: 'CONFIG_TCG', if_true: files('blkreplay.c'))
+ system_ss.add(files('block-ram-registrar.c'))
 diff --git a/block/monitor/block-hmp-cmds.c b/block/monitor/block-hmp-cmds.c
-index ca2599de44..636509b83e 100644
+index bdf2eb50b6..439a7a14c8 100644
 --- a/block/monitor/block-hmp-cmds.c
 +++ b/block/monitor/block-hmp-cmds.c
-@@ -1029,3 +1029,43 @@ void hmp_change_medium(Monitor *mon, const char *device, const char *target,
+@@ -1009,3 +1009,42 @@ void hmp_change_medium(Monitor *mon, const char *device, const char *target,
      qmp_blockdev_change_medium(device, NULL, target, arg, true, force,
                                 !!read_only, read_only_mode, errp);
  }
@@ -139,7 +141,6 @@ index ca2599de44..636509b83e 100644
 +{
 +    Error *error = NULL;
 +
-+    int dir = qdict_get_try_bool(qdict, "directory", 0);
 +    const char *backup_file = qdict_get_str(qdict, "backupfile");
 +    const char *devlist = qdict_get_try_str(qdict, "devlist");
 +    int64_t speed = qdict_get_try_int(qdict, "speed", 0);
@@ -157,7 +158,7 @@ index ca2599de44..636509b83e 100644
 +        false, false, // PBS use-dirty-bitmap
 +        false, false, // PBS compress
 +        false, false, // PBS encrypt
-+        true, dir ? BACKUP_FORMAT_DIR : BACKUP_FORMAT_VMA,
++        true, BACKUP_FORMAT_VMA,
 +        NULL, NULL,
 +        devlist, qdict_haskey(qdict, "speed"), speed,
 +        false, 0, // BackupPerf max-workers
@@ -166,7 +167,7 @@ index ca2599de44..636509b83e 100644
 +    hmp_handle_error(mon, error);
 +}
 diff --git a/blockdev.c b/blockdev.c
-index bdae211a54..315a27fc09 100644
+index 79d47b1920..3f67eb413d 100644
 --- a/blockdev.c
 +++ b/blockdev.c
 @@ -37,6 +37,7 @@
@@ -178,10 +179,10 @@ index bdae211a54..315a27fc09 100644
  #include "monitor/monitor.h"
  #include "qemu/error-report.h"
 diff --git a/hmp-commands-info.hx b/hmp-commands-info.hx
-index a166bff3d5..4b75966c2e 100644
+index d1a7b99add..af588145ff 100644
 --- a/hmp-commands-info.hx
 +++ b/hmp-commands-info.hx
-@@ -486,6 +486,20 @@ SRST
+@@ -458,6 +458,20 @@ SRST
      Show the current VM UUID.
  ERST
  
@@ -203,10 +204,10 @@ index a166bff3d5..4b75966c2e 100644
      {
          .name       = "usernet",
 diff --git a/hmp-commands.hx b/hmp-commands.hx
-index d9f9f42d11..775518fb09 100644
+index 0c7c6f2c16..bf8315f226 100644
 --- a/hmp-commands.hx
 +++ b/hmp-commands.hx
-@@ -101,6 +101,37 @@ ERST
+@@ -101,6 +101,35 @@ ERST
  SRST
  ``block_stream``
    Copy data from a backing file into a block device.
@@ -214,11 +215,9 @@ index d9f9f42d11..775518fb09 100644
 +
 +   {
 +        .name       = "backup",
-+        .args_type  = "directory:-d,backupfile:s,speed:o?,devlist:s?",
-+        .params     = "[-d] backupfile [speed [devlist]]",
-+        .help       = "create a VM Backup."
-+		    "\n\t\t\t Use -d to dump data into a directory instead"
-+		    "\n\t\t\t of using VMA format.",
++        .args_type  = "backupfile:s,speed:o?,devlist:s?",
++        .params     = "backupfile [speed [devlist]]",
++        .help       = "create a VM backup (VMA format).",
 +        .cmd = hmp_backup,
 +        .coroutine  = true,
 +    },
@@ -245,7 +244,7 @@ index d9f9f42d11..775518fb09 100644
  
      {
 diff --git a/include/monitor/hmp.h b/include/monitor/hmp.h
-index fdf6b45fb8..e01b2201d8 100644
+index 2596cc2426..9dda91d65a 100644
 --- a/include/monitor/hmp.h
 +++ b/include/monitor/hmp.h
 @@ -32,6 +32,7 @@ void hmp_info_savevm(Monitor *mon, const QDict *qdict);
@@ -256,7 +255,7 @@ index fdf6b45fb8..e01b2201d8 100644
  void hmp_info_cpus(Monitor *mon, const QDict *qdict);
  void hmp_info_vnc(Monitor *mon, const QDict *qdict);
  void hmp_info_spice(Monitor *mon, const QDict *qdict);
-@@ -84,6 +85,8 @@ void hmp_change_vnc(Monitor *mon, const char *device, const char *target,
+@@ -82,6 +83,8 @@ void hmp_change_vnc(Monitor *mon, const char *device, const char *target,
  void hmp_change_medium(Monitor *mon, const char *device, const char *target,
                         const char *arg, const char *read_only, bool force,
                         Error **errp);
@@ -266,10 +265,10 @@ index fdf6b45fb8..e01b2201d8 100644
  void hmp_device_add(Monitor *mon, const QDict *qdict);
  void hmp_device_del(Monitor *mon, const QDict *qdict);
 diff --git a/meson.build b/meson.build
-index 38a4e2bcef..443b3238f9 100644
+index b9b673c271..f6fb9b4fd8 100644
 --- a/meson.build
 +++ b/meson.build
-@@ -1528,6 +1528,7 @@ keyutils = dependency('libkeyutils', required: false,
+@@ -2130,6 +2130,7 @@ endif
  has_gettid = cc.has_function('gettid')
  
  libuuid = cc.find_library('uuid', required: true)
@@ -278,18 +277,18 @@ index 38a4e2bcef..443b3238f9 100644
  # libselinux
  selinux = dependency('libselinux',
 diff --git a/monitor/hmp-cmds.c b/monitor/hmp-cmds.c
-index 91be698308..5b9c231a4c 100644
+index 874084565f..bedeb81f8c 100644
 --- a/monitor/hmp-cmds.c
 +++ b/monitor/hmp-cmds.c
-@@ -21,6 +21,7 @@
+@@ -22,6 +22,7 @@
  #include "qemu/help_option.h"
  #include "monitor/monitor-internal.h"
  #include "qapi/error.h"
 +#include "qapi/qapi-commands-block-core.h"
  #include "qapi/qapi-commands-control.h"
+ #include "qapi/qapi-commands-machine.h"
  #include "qapi/qapi-commands-migration.h"
- #include "qapi/qapi-commands-misc.h"
-@@ -144,6 +145,77 @@ void hmp_sync_profile(Monitor *mon, const QDict *qdict)
+@@ -119,6 +120,77 @@ void hmp_sync_profile(Monitor *mon, const QDict *qdict)
      }
  }
  
@@ -587,10 +586,10 @@ index 0000000000..8cbf645b2c
 +#endif /* PROXMOX_BACKUP_CLIENT_H */
 diff --git a/pve-backup.c b/pve-backup.c
 new file mode 100644
-index 0000000000..dd72ee0ed6
+index 0000000000..fea0152de0
 --- /dev/null
 +++ b/pve-backup.c
-@@ -0,0 +1,1097 @@
+@@ -0,0 +1,1090 @@
 +#include "proxmox-backup-client.h"
 +#include "vma.h"
 +
@@ -601,10 +600,15 @@ index 0000000000..dd72ee0ed6
 +#include "block/block_int-global-state.h"
 +#include "block/blockjob.h"
 +#include "block/dirty-bitmap.h"
++#include "block/graph-lock.h"
 +#include "qapi/qapi-commands-block.h"
 +#include "qapi/qmp/qerror.h"
 +#include "qemu/cutils.h"
 +
++#if defined(CONFIG_MALLOC_TRIM)
++#include <malloc.h>
++#endif
++
 +#include <proxmox-backup-qemu.h>
 +
 +/* PVE backup state and related function */
@@ -622,7 +626,6 @@ index 0000000000..dd72ee0ed6
 + * ---end-bad-example--
 + *
 + * ==> Always use CoMutext inside coroutines.
-+ * ==> Never acquire/release AioContext withing coroutines (because that use QemuRecMutex)
 + *
 + */
 +
@@ -675,7 +678,6 @@ index 0000000000..dd72ee0ed6
 +    uint64_t block_size;
 +    uint8_t dev_id;
 +    int completed_ret; // INT_MAX if not completed
-+    char targetfile[PATH_MAX];
 +    BdrvDirtyBitmap *bitmap;
 +    BlockDriverState *target;
 +    BlockJob *job;
@@ -697,7 +699,7 @@ index 0000000000..dd72ee0ed6
 +    return error_or_canceled;
 +}
 +
-+static void pvebackup_add_transfered_bytes(size_t transferred, size_t zero_bytes, size_t reused)
++static void pvebackup_add_transferred_bytes(size_t transferred, size_t zero_bytes, size_t reused)
 +{
 +    qemu_mutex_lock(&backup_state.stat.lock);
 +    backup_state.stat.zero_bytes += zero_bytes;
@@ -758,7 +760,7 @@ index 0000000000..dd72ee0ed6
 +    }
 +
 +    qemu_co_mutex_unlock(&backup_state.dump_callback_mutex);
-+    pvebackup_add_transfered_bytes(size, is_zero_block ? size : 0, reused);
++    pvebackup_add_transferred_bytes(size, is_zero_block ? size : 0, reused);
 +
 +    return size;
 +}
@@ -816,11 +818,11 @@ index 0000000000..dd72ee0ed6
 +        } else {
 +            if (remaining >= VMA_CLUSTER_SIZE) {
 +                assert(ret == VMA_CLUSTER_SIZE);
-+                pvebackup_add_transfered_bytes(VMA_CLUSTER_SIZE, zero_bytes, 0);
++                pvebackup_add_transferred_bytes(VMA_CLUSTER_SIZE, zero_bytes, 0);
 +                remaining -= VMA_CLUSTER_SIZE;
 +            } else {
 +                assert(ret == remaining);
-+                pvebackup_add_transfered_bytes(remaining, zero_bytes, 0);
++                pvebackup_add_transferred_bytes(remaining, zero_bytes, 0);
 +                remaining = 0;
 +            }
 +        }
@@ -869,6 +871,14 @@ index 0000000000..dd72ee0ed6
 +    backup_state.stat.end_time = time(NULL);
 +    backup_state.stat.finishing = false;
 +    qemu_mutex_unlock(&backup_state.stat.lock);
++
++#if defined(CONFIG_MALLOC_TRIM)
++    /*
++     * Try to reclaim memory for buffers (and, in case of PBS, Rust futures), etc.
++     * Won't happen by default if there is fragmentation.
++     */
++    malloc_trim(4 * 1024 * 1024);
++#endif
 +}
 +
 +static void coroutine_fn pvebackup_co_complete_stream(void *opaque)
@@ -888,7 +898,12 @@ index 0000000000..dd72ee0ed6
 +
 +    qemu_co_mutex_lock(&backup_state.backup_mutex);
 +
-+    if (ret < 0) {
++    /*
++     * All jobs in the transaction will be canceled when one receives an error.
++     * The first error wins, so only set it for ECANCELED if it was the last
++     * job. This allows more interesting errors from other jobs to win.
++     */
++    if (ret < 0 && (ret != -ECANCELED || !g_list_nth(backup_state.di_list, 1))) {
 +        Error *local_err = NULL;
 +        error_setg(&local_err, "job failed with err %d - %s", ret, strerror(-ret));
 +        pvebackup_propagate_error(local_err);
@@ -912,13 +927,6 @@ index 0000000000..dd72ee0ed6
 +        }
 +    }
 +
-+    if (di->job) {
-+        WITH_JOB_LOCK_GUARD() {
-+            job_unref_locked(&di->job->job);
-+            di->job = NULL;
-+        }
-+    }
-+
 +    // remove self from job list
 +    backup_state.di_list = g_list_remove(backup_state.di_list, di);
 +
@@ -938,6 +946,16 @@ index 0000000000..dd72ee0ed6
 +    di->completed_ret = ret;
 +
 +    /*
++     * Needs to happen outside of coroutine, because it takes the graph write lock.
++     */
++    if (di->job) {
++        WITH_JOB_LOCK_GUARD() {
++            job_unref_locked(&di->job->job);
++            di->job = NULL;
++        }
++    }
++
++    /*
 +     * Schedule stream cleanup in async coroutine. close_image and finish might
 +     * take a while, so we can't block on them here. This way it also doesn't
 +     * matter if we're already running in a coroutine or not.
@@ -1020,7 +1038,6 @@ index 0000000000..dd72ee0ed6
 +    const char *file,
 +    const char *name,
 +    BackupFormat format,
-+    const char *backup_dir,
 +    VmaWriter *vmaw,
 +    ProxmoxBackupHandle *pbs,
 +    Error **errp)
@@ -1046,13 +1063,6 @@ index 0000000000..dd72ee0ed6
 +    } else if (format == BACKUP_FORMAT_PBS) {
 +        if (proxmox_backup_co_add_config(pbs, name, (unsigned char *)cdata, clen, errp) < 0)
 +            goto err;
-+    } else if (format == BACKUP_FORMAT_DIR) {
-+        char config_path[PATH_MAX];
-+        snprintf(config_path, PATH_MAX, "%s/%s", backup_dir, name);
-+        if (!g_file_set_contents(config_path, cdata, clen, &err)) {
-+            error_setg(errp, "unable to write config file '%s'", config_path);
-+            goto err;
-+        }
 +    }
 +
 + out:
@@ -1066,8 +1076,7 @@ index 0000000000..dd72ee0ed6
 +}
 +
 +/*
-+ * backup_job_create can *not* be run from a coroutine (and requires an
-+ * acquired AioContext), so this can't either.
++ * backup_job_create can *not* be run from a coroutine, so this can't either.
 + * The caller is responsible that backup_mutex is held nonetheless.
 + */
 +static void create_backup_jobs_bh(void *opaque) {
@@ -1100,8 +1109,7 @@ index 0000000000..dd72ee0ed6
 +            sync_mode = MIRROR_SYNC_MODE_BITMAP;
 +            bitmap_mode = BITMAP_SYNC_MODE_ON_SUCCESS;
 +        }
-+        AioContext *aio_context = bdrv_get_aio_context(di->bs);
-+        aio_context_acquire(aio_context);
++        bdrv_drained_begin(di->bs);
 +
 +        BlockJob *job = backup_job_create(
 +            NULL, di->bs, di->target, backup_state.speed, sync_mode, di->bitmap,
@@ -1109,7 +1117,7 @@ index 0000000000..dd72ee0ed6
 +            BLOCKDEV_ON_ERROR_REPORT, JOB_DEFAULT, pvebackup_complete_cb, di, backup_state.txn,
 +            &local_err);
 +
-+        aio_context_release(aio_context);
++        bdrv_drained_end(di->bs);
 +
 +        di->job = job;
 +        if (job) {
@@ -1161,6 +1169,66 @@ index 0000000000..dd72ee0ed6
 +    aio_co_enter(data->ctx, data->co);
 +}
 +
++/*
++ * Returns a list of device infos, which needs to be freed by the caller. In
++ * case of an error, errp will be set, but the returned value might still be a
++ * list.
++ */
++static GList coroutine_fn GRAPH_RDLOCK *get_device_info(
++    const char *devlist,
++    Error **errp)
++{
++    gchar **devs = NULL;
++    GList *di_list = NULL;
++
++    if (devlist) {
++        devs = g_strsplit_set(devlist, ",;:", -1);
++
++        gchar **d = devs;
++        while (d && *d) {
++            BlockBackend *blk = blk_by_name(*d);
++            if (!blk) {
++                error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
++                          "Device '%s' not found", *d);
++                goto err;
++            }
++            BlockDriverState *bs = blk_bs(blk);
++            if (!bdrv_co_is_inserted(bs)) {
++                error_setg(errp, "Device '%s' has no medium", *d);
++                goto err;
++            }
++            PVEBackupDevInfo *di = g_new0(PVEBackupDevInfo, 1);
++            di->bs = bs;
++            di_list = g_list_append(di_list, di);
++            d++;
++        }
++    } else {
++        BdrvNextIterator it;
++
++        for (BlockDriverState *bs = bdrv_first(&it); bs; bs = bdrv_next(&it)) {
++            if (!bdrv_co_is_inserted(bs) || bdrv_is_read_only(bs)) {
++                continue;
++            }
++
++            PVEBackupDevInfo *di = g_new0(PVEBackupDevInfo, 1);
++            di->bs = bs;
++            di_list = g_list_append(di_list, di);
++        }
++    }
++
++    if (!di_list) {
++        error_set(errp, ERROR_CLASS_GENERIC_ERROR, "empty device list");
++        goto err;
++    }
++
++err:
++    if (devs) {
++        g_strfreev(devs);
++    }
++
++    return di_list;
++}
++
 +UuidInfo coroutine_fn *qmp_backup(
 +    const char *backup_file,
 +    const char *password,
@@ -1186,14 +1254,10 @@ index 0000000000..dd72ee0ed6
 +
 +    qemu_co_mutex_lock(&backup_state.backup_mutex);
 +
-+    BlockBackend *blk;
-+    BlockDriverState *bs = NULL;
-+    const char *backup_dir = NULL;
 +    Error *local_err = NULL;
 +    uuid_t uuid;
 +    VmaWriter *vmaw = NULL;
 +    ProxmoxBackupHandle *pbs = NULL;
-+    gchar **devs = NULL;
 +    GList *di_list = NULL;
 +    GList *l;
 +    UuidInfo *uuid_info;
@@ -1211,48 +1275,14 @@ index 0000000000..dd72ee0ed6
 +    /* Todo: try to auto-detect format based on file name */
 +    format = has_format ? format : BACKUP_FORMAT_VMA;
 +
-+    if (devlist) {
-+        devs = g_strsplit_set(devlist, ",;:", -1);
-+
-+        gchar **d = devs;
-+        while (d && *d) {
-+            blk = blk_by_name(*d);
-+            if (blk) {
-+                bs = blk_bs(blk);
-+                if (!bdrv_co_is_inserted(bs)) {
-+                    error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, *d);
-+                    goto err;
-+                }
-+                PVEBackupDevInfo *di = g_new0(PVEBackupDevInfo, 1);
-+                di->bs = bs;
-+                di_list = g_list_append(di_list, di);
-+            } else {
-+                error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
-+                          "Device '%s' not found", *d);
-+                goto err;
-+            }
-+            d++;
-+        }
-+
-+    } else {
-+        BdrvNextIterator it;
-+
-+        bs = NULL;
-+        for (bs = bdrv_first(&it); bs; bs = bdrv_next(&it)) {
-+            if (!bdrv_co_is_inserted(bs) || bdrv_is_read_only(bs)) {
-+                continue;
-+            }
-+
-+            PVEBackupDevInfo *di = g_new0(PVEBackupDevInfo, 1);
-+            di->bs = bs;
-+            di_list = g_list_append(di_list, di);
-+        }
-+    }
-+
-+    if (!di_list) {
-+        error_set(errp, ERROR_CLASS_GENERIC_ERROR, "empty device list");
++    bdrv_graph_co_rdlock();
++    di_list = get_device_info(devlist, &local_err);
++    bdrv_graph_co_rdunlock();
++    if (local_err) {
++        error_propagate(errp, local_err);
 +        goto err;
 +    }
++    assert(di_list);
 +
 +    size_t total = 0;
 +
@@ -1260,7 +1290,11 @@ index 0000000000..dd72ee0ed6
 +    while (l) {
 +        PVEBackupDevInfo *di = (PVEBackupDevInfo *)l->data;
 +        l = g_list_next(l);
-+        if (bdrv_op_is_blocked(di->bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
++
++        bdrv_graph_co_rdlock();
++        bool blocked = bdrv_op_is_blocked(di->bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp);
++        bdrv_graph_co_rdunlock();
++        if (blocked) {
 +            goto err;
 +        }
 +
@@ -1344,7 +1378,9 @@ index 0000000000..dd72ee0ed6
 +
 +            di->block_size = dump_cb_block_size;
 +
++            bdrv_graph_co_rdlock();
 +            const char *devname = bdrv_get_device_name(di->bs);
++            bdrv_graph_co_rdunlock();
 +            PBSBitmapAction action = PBS_BITMAP_ACTION_NOT_USED;
 +            size_t dirty = di->size;
 +
@@ -1388,7 +1424,7 @@ index 0000000000..dd72ee0ed6
 +                goto err_mutex;
 +            }
 +
-+            if (!(di->target = bdrv_backup_dump_create(dump_cb_block_size, di->size, pvebackup_co_dump_pbs_cb, di, errp))) {
++            if (!(di->target = bdrv_co_backup_dump_create(dump_cb_block_size, di->size, pvebackup_co_dump_pbs_cb, di, errp))) {
 +                goto err_mutex;
 +            }
 +
@@ -1404,9 +1440,7 @@ index 0000000000..dd72ee0ed6
 +    } else if (format == BACKUP_FORMAT_VMA) {
 +        vmaw = vma_writer_create(backup_file, uuid, &local_err);
 +        if (!vmaw) {
-+            if (local_err) {
-+                error_propagate(errp, local_err);
-+            }
++            error_propagate(errp, local_err);
 +            goto err_mutex;
 +        }
 +
@@ -1416,11 +1450,13 @@ index 0000000000..dd72ee0ed6
 +            PVEBackupDevInfo *di = (PVEBackupDevInfo *)l->data;
 +            l = g_list_next(l);
 +
-+            if (!(di->target = bdrv_backup_dump_create(VMA_CLUSTER_SIZE, di->size, pvebackup_co_dump_vma_cb, di, errp))) {
++            if (!(di->target = bdrv_co_backup_dump_create(VMA_CLUSTER_SIZE, di->size, pvebackup_co_dump_vma_cb, di, errp))) {
 +                goto err_mutex;
 +            }
 +
++            bdrv_graph_co_rdlock();
 +            const char *devname = bdrv_get_device_name(di->bs);
++            bdrv_graph_co_rdunlock();
 +            di->dev_id = vma_writer_register_stream(vmaw, devname, di->size);
 +            if (di->dev_id <= 0) {
 +                error_set(errp, ERROR_CLASS_GENERIC_ERROR,
@@ -1428,54 +1464,21 @@ index 0000000000..dd72ee0ed6
 +                goto err_mutex;
 +            }
 +        }
-+    } else if (format == BACKUP_FORMAT_DIR) {
-+        if (mkdir(backup_file, 0640) != 0) {
-+            error_setg_errno(errp, errno, "can't create directory '%s'\n",
-+                             backup_file);
-+            goto err_mutex;
-+        }
-+        backup_dir = backup_file;
-+
-+        l = di_list;
-+        while (l) {
-+            PVEBackupDevInfo *di = (PVEBackupDevInfo *)l->data;
-+            l = g_list_next(l);
-+
-+            const char *devname = bdrv_get_device_name(di->bs);
-+            snprintf(di->targetfile, PATH_MAX, "%s/%s.raw", backup_dir, devname);
-+
-+            int flags = BDRV_O_RDWR;
-+            bdrv_img_create(di->targetfile, "raw", NULL, NULL, NULL,
-+                            di->size, flags, false, &local_err);
-+            if (local_err) {
-+                error_propagate(errp, local_err);
-+                goto err_mutex;
-+            }
-+
-+            di->target = bdrv_co_open(di->targetfile, NULL, NULL, flags, &local_err);
-+            if (!di->target) {
-+                error_propagate(errp, local_err);
-+                goto err_mutex;
-+            }
-+        }
 +    } else {
 +        error_set(errp, ERROR_CLASS_GENERIC_ERROR, "unknown backup format");
 +        goto err_mutex;
 +    }
 +
-+
 +    /* add configuration file to archive */
 +    if (config_file) {
-+        if (pvebackup_co_add_config(config_file, config_name, format, backup_dir,
-+                                    vmaw, pbs, errp) != 0) {
++        if (pvebackup_co_add_config(config_file, config_name, format, vmaw, pbs, errp) != 0) {
 +            goto err_mutex;
 +        }
 +    }
 +
 +    /* add firewall file to archive */
 +    if (firewall_file) {
-+        if (pvebackup_co_add_config(firewall_file, firewall_name, format, backup_dir,
-+                                    vmaw, pbs, errp) != 0) {
++        if (pvebackup_co_add_config(firewall_file, firewall_name, format, vmaw, pbs, errp) != 0) {
 +            goto err_mutex;
 +        }
 +    }
@@ -1565,18 +1568,11 @@ index 0000000000..dd72ee0ed6
 +            bdrv_co_unref(di->target);
 +        }
 +
-+        if (di->targetfile[0]) {
-+            unlink(di->targetfile);
-+        }
 +        g_free(di);
 +    }
 +    g_list_free(di_list);
 +    backup_state.di_list = NULL;
 +
-+    if (devs) {
-+        g_strfreev(devs);
-+    }
-+
 +    if (vmaw) {
 +        Error *err = NULL;
 +        vma_writer_close(vmaw, &err);
@@ -1588,10 +1584,6 @@ index 0000000000..dd72ee0ed6
 +        backup_state.pbs = NULL;
 +    }
 +
-+    if (backup_dir) {
-+        rmdir(backup_dir);
-+    }
-+
 +    qemu_co_mutex_unlock(&backup_state.backup_mutex);
 +    return NULL;
 +}
@@ -1689,10 +1681,10 @@ index 0000000000..dd72ee0ed6
 +    return ret;
 +}
 diff --git a/qapi/block-core.json b/qapi/block-core.json
-index 542add004b..4ec70acf95 100644
+index 321d1fd0e1..68caf30084 100644
 --- a/qapi/block-core.json
 +++ b/qapi/block-core.json
-@@ -835,6 +835,232 @@
+@@ -851,6 +851,239 @@
  { 'command': 'query-block', 'returns': ['BlockInfo'],
    'allow-preconfig': true }
  
@@ -1742,9 +1734,12 @@ index 542add004b..4ec70acf95 100644
 +# An enumeration of supported backup formats.
 +#
 +# @vma: Proxmox vma backup format
++#
++# @pbs: Proxmox backup server format
++#
 +##
 +{ 'enum': 'BackupFormat',
-+  'data': [ 'vma', 'dir', 'pbs' ] }
++  'data': [ 'vma', 'pbs' ] }
 +
 +##
 +# @backup:
@@ -1758,6 +1753,9 @@ index 542add004b..4ec70acf95 100644
 +# @config-file: a configuration file to include into
 +#               the backup archive.
 +#
++# @firewall-file: a firewall configuration file to include into the backup
++#     archive.
++#
 +# @speed: the maximum speed, in bytes per second
 +#
 +# @devlist: list of block device names (separated by ',', ';'
@@ -1825,9 +1823,7 @@ index 542add004b..4ec70acf95 100644
 +#
 +# Cancel the current executing backup process.
 +#
-+# Returns: nothing on success
-+#
-+# Notes: This command succeeds even if there is no backup process running.
++# .. note:: This command succeeds even if there is no backup process running.
 +#
 +##
 +{ 'command': 'backup-cancel', 'coroutine': true }
@@ -1850,6 +1846,9 @@ index 542add004b..4ec70acf95 100644
 +#
 +# @pbs-library-version: Running version of libproxmox-backup-qemu0 library.
 +#
++# @backup-max-workers: Whether the 'max-workers' @BackupPerf setting is
++#     supported or not.
++#
 +##
 +{ 'struct': 'ProxmoxSupportStatus',
 +  'data': { 'pbs-dirty-bitmap': 'bool',
@@ -1926,10 +1925,10 @@ index 542add004b..4ec70acf95 100644
  # @BlockDeviceTimedStats:
  #
 diff --git a/qapi/common.json b/qapi/common.json
-index 356db3f670..aae8a3b682 100644
+index 6ffc7a3789..9c6c671ece 100644
 --- a/qapi/common.json
 +++ b/qapi/common.json
-@@ -206,3 +206,16 @@
+@@ -212,3 +212,17 @@
  ##
  { 'struct': 'HumanReadableText',
    'data': { 'human-readable-text': 'str' } }
@@ -1943,11 +1942,12 @@ index 356db3f670..aae8a3b682 100644
 +#
 +# Since: 0.14.0
 +#
-+# Notes: If no UUID was specified for the guest, a null UUID is returned.
++# .. note:: If no UUID was specified for the guest, a null UUID is
++#     returned.
 +##
 +{ 'struct': 'UuidInfo', 'data': {'UUID': 'str'} }
 diff --git a/qapi/machine.json b/qapi/machine.json
-index 47f3facdb2..46760978ae 100644
+index 12cfd3f260..a8abdb42a3 100644
 --- a/qapi/machine.json
 +++ b/qapi/machine.json
 @@ -4,6 +4,8 @@
@@ -1959,7 +1959,7 @@ index 47f3facdb2..46760978ae 100644
  ##
  # = Machines
  ##
-@@ -228,19 +230,6 @@
+@@ -302,20 +304,6 @@
  ##
  { 'command': 'query-target', 'returns': 'TargetInfo' }
  
@@ -1972,7 +1972,8 @@ index 47f3facdb2..46760978ae 100644
 -#
 -# Since: 0.14
 -#
--# Notes: If no UUID was specified for the guest, a null UUID is returned.
+-# .. note:: If no UUID was specified for the guest, the nil UUID (all
+-#    zeroes) is returned.
 -##
 -{ 'struct': 'UuidInfo', 'data': {'UUID': 'str'} }
 -

debian/patches/pve/0031-PVE-Backup-pbs-restore-new-command-to-restore-from-p.patch

@@ -14,20 +14,20 @@ Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
  create mode 100644 pbs-restore.c
 
 diff --git a/meson.build b/meson.build
-index 443b3238f9..32ab849ce6 100644
+index f6fb9b4fd8..f666d0f028 100644
 --- a/meson.build
 +++ b/meson.build
-@@ -3656,6 +3656,10 @@ if have_tools
+@@ -4350,6 +4350,10 @@ if have_tools
    vma = executable('vma', files('vma.c', 'vma-reader.c') + genh,
-                    dependencies: [authz, block, crypto, io, qom], install: true)
+                    dependencies: [authz, block, crypto, io, qemuutil, qom], install: true)
  
 +  pbs_restore = executable('pbs-restore', files('pbs-restore.c') + genh,
-+                  dependencies: [authz, block, crypto, io, qom,
++                  dependencies: [authz, block, crypto, io, qemuutil, qom,
 +                    libproxmox_backup_qemu], install: true)
 +
    subdir('storage-daemon')
-   subdir('contrib/rdmacm-mux')
-   subdir('contrib/elf2dmp')
+ 
+   foreach exe: [ 'qemu-img', 'qemu-io', 'qemu-nbd', 'qemu-storage-daemon']
 diff --git a/pbs-restore.c b/pbs-restore.c
 new file mode 100644
 index 0000000000..f03d9bab8d

debian/patches/pve/0032-PVE-Add-PBS-block-driver-to-map-backup-archives-into.patch

@@ -14,35 +14,33 @@ Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
      getlength is now a coroutine function]
 Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
 ---
- block/meson.build    |   3 +
- block/pbs.c          | 305 +++++++++++++++++++++++++++++++++++++++++++
- configure            |   9 ++
+ block/meson.build    |   2 +
+ block/pbs.c          | 306 +++++++++++++++++++++++++++++++++++++++++++
  meson.build          |   2 +-
- qapi/block-core.json |  13 ++
+ qapi/block-core.json |  29 ++++
  qapi/pragma.json     |   1 +
- 6 files changed, 332 insertions(+), 1 deletion(-)
+ 5 files changed, 339 insertions(+), 1 deletion(-)
  create mode 100644 block/pbs.c
 
 diff --git a/block/meson.build b/block/meson.build
-index 5bcebb934b..eece0d5743 100644
+index 2367e1ac1b..e178047ec9 100644
 --- a/block/meson.build
 +++ b/block/meson.build
-@@ -54,6 +54,9 @@ block_ss.add(files(
+@@ -49,6 +49,8 @@ block_ss.add(files(
    '../pve-backup.c',
  ), libproxmox_backup_qemu)
  
-+block_ss.add(when: 'CONFIG_PBS_BDRV', if_true: files('pbs.c'))
-+block_ss.add(when: 'CONFIG_PBS_BDRV', if_true: libproxmox_backup_qemu)
++block_ss.add(files('pbs.c'), libproxmox_backup_qemu)
 +
  
- softmmu_ss.add(when: 'CONFIG_TCG', if_true: files('blkreplay.c'))
- softmmu_ss.add(files('block-ram-registrar.c'))
+ system_ss.add(when: 'CONFIG_TCG', if_true: files('blkreplay.c'))
+ system_ss.add(files('block-ram-registrar.c'))
 diff --git a/block/pbs.c b/block/pbs.c
 new file mode 100644
-index 0000000000..a2211e0f3b
+index 0000000000..2d5e28ce8f
 --- /dev/null
 +++ b/block/pbs.c
-@@ -0,0 +1,305 @@
+@@ -0,0 +1,306 @@
 +/*
 + * Proxmox Backup Server read-only block driver
 + */
@@ -70,7 +68,7 @@ index 0000000000..a2211e0f3b
 +
 +typedef struct {
 +    ProxmoxRestoreHandle *conn;
-+    char aid;
++    uint8_t aid;
 +    int64_t length;
 +
 +    char *repository;
@@ -203,12 +201,18 @@ index 0000000000..a2211e0f3b
 +    }
 +
 +    /* acquire handle and length */
-+    s->aid = proxmox_restore_open_image(s->conn, s->archive, &pbs_error);
-+    if (s->aid < 0) {
++    ret = proxmox_restore_open_image(s->conn, s->archive, &pbs_error);
++    if (ret < 0) {
 +        if (pbs_error && errp) error_setg(errp, "PBS open_image failed: %s", pbs_error);
 +        if (pbs_error) proxmox_backup_free_error(pbs_error);
 +        return -ENODEV;
 +    }
++    if (ret > UINT8_MAX) {
++        error_setg(errp, "PBS open_image returned an ID larger than %u", UINT8_MAX);
++        return -ENODEV;
++    }
++    s->aid = ret;
++
 +    s->length = proxmox_restore_get_image_length(s->conn, s->aid, &pbs_error);
 +    if (s->length < 0) {
 +        if (pbs_error && errp) error_setg(errp, "PBS get_image_length failed: %s", pbs_error);
@@ -219,12 +223,6 @@ index 0000000000..a2211e0f3b
 +    return 0;
 +}
 +
-+static int pbs_file_open(BlockDriverState *bs, QDict *options, int flags,
-+                         Error **errp)
-+{
-+    return pbs_open(bs, options, flags, errp);
-+}
-+
 +static void pbs_close(BlockDriverState *bs) {
 +    BDRVPBSState *s = bs->opaque;
 +    g_free(s->repository);
@@ -234,7 +232,8 @@ index 0000000000..a2211e0f3b
 +    proxmox_restore_disconnect(s->conn);
 +}
 +
-+static coroutine_fn int64_t pbs_co_getlength(BlockDriverState *bs)
++static coroutine_fn int64_t GRAPH_RDLOCK
++pbs_co_getlength(BlockDriverState *bs)
 +{
 +    BDRVPBSState *s = bs->opaque;
 +    return s->length;
@@ -251,9 +250,9 @@ index 0000000000..a2211e0f3b
 +    aio_co_schedule(rcb->ctx, rcb->co);
 +}
 +
-+static coroutine_fn int pbs_co_preadv(BlockDriverState *bs,
-+                                      int64_t offset, int64_t bytes,
-+                                      QEMUIOVector *qiov, BdrvRequestFlags flags)
++static coroutine_fn int GRAPH_RDLOCK
++pbs_co_preadv(BlockDriverState *bs, int64_t offset, int64_t bytes,
++              QEMUIOVector *qiov, BdrvRequestFlags flags)
 +{
 +    BDRVPBSState *s = bs->opaque;
 +    int ret;
@@ -298,16 +297,17 @@ index 0000000000..a2211e0f3b
 +    return 0;
 +}
 +
-+static coroutine_fn int pbs_co_pwritev(BlockDriverState *bs,
-+                                       int64_t offset, int64_t bytes,
-+                                       QEMUIOVector *qiov, BdrvRequestFlags flags)
++static coroutine_fn int GRAPH_RDLOCK
++pbs_co_pwritev(BlockDriverState *bs, int64_t offset, int64_t bytes,
++               QEMUIOVector *qiov, BdrvRequestFlags flags)
 +{
 +    fprintf(stderr, "pbs-bdrv: cannot write to backup file, make sure "
 +           "any attached disk devices are set to read-only!\n");
 +    return -EPERM;
 +}
 +
-+static void pbs_refresh_filename(BlockDriverState *bs)
++static void GRAPH_RDLOCK
++pbs_refresh_filename(BlockDriverState *bs)
 +{
 +    BDRVPBSState *s = bs->opaque;
 +    if (s->namespace) {
@@ -330,7 +330,6 @@ index 0000000000..a2211e0f3b
 +
 +    .bdrv_parse_filename    = pbs_parse_filename,
 +
-+    .bdrv_file_open         = pbs_file_open,
 +    .bdrv_open              = pbs_open,
 +    .bdrv_close             = pbs_close,
 +    .bdrv_co_getlength      = pbs_co_getlength,
@@ -348,65 +347,24 @@ index 0000000000..a2211e0f3b
 +}
 +
 +block_init(bdrv_pbs_init);
-diff --git a/configure b/configure
-index a62a3e6be9..1ac0feb46b 100755
---- a/configure
-+++ b/configure
-@@ -288,6 +288,7 @@ linux_user=""
- bsd_user=""
- pie=""
- coroutine=""
-+pbs_bdrv="yes"
- plugins="$default_feature"
- meson=""
- ninja=""
-@@ -873,6 +874,10 @@ for opt do
-   ;;
-   --with-coroutine=*) coroutine="$optarg"
-   ;;
-+  --disable-pbs-bdrv) pbs_bdrv="no"
-+  ;;
-+  --enable-pbs-bdrv) pbs_bdrv="yes"
-+  ;;
-   --with-git=*) git="$optarg"
-   ;;
-   --with-git-submodules=*)
-@@ -1049,6 +1054,7 @@ cat << EOF
-   debug-info      debugging information
-   safe-stack      SafeStack Stack Smash Protection. Depends on
-                   clang/llvm and requires coroutine backend ucontext.
-+  pbs-bdrv        Proxmox backup server read-only block driver support
- 
- NOTE: The object files are built at the place where configure is launched
- EOF
-@@ -2386,6 +2392,9 @@ echo "TARGET_DIRS=$target_list" >> $config_host_mak
- if test "$modules" = "yes"; then
-   echo "CONFIG_MODULES=y" >> $config_host_mak
- fi
-+if test "$pbs_bdrv" = "yes" ; then
-+  echo "CONFIG_PBS_BDRV=y" >> $config_host_mak
-+fi
- 
- # XXX: suppress that
- if [ "$bsd" = "yes" ] ; then
 diff --git a/meson.build b/meson.build
-index 32ab849ce6..69afe3441b 100644
+index f666d0f028..4c85736ec3 100644
 --- a/meson.build
 +++ b/meson.build
-@@ -4041,7 +4041,7 @@ summary_info += {'bzip2 support':     libbzip2}
- summary_info += {'lzfse support':     liblzfse}
- summary_info += {'zstd support':      zstd}
+@@ -4815,7 +4815,7 @@ summary_info += {'Query Processing Library support': qpl}
+ summary_info += {'UADK Library support': uadk}
+ summary_info += {'qatzip support':    qatzip}
  summary_info += {'NUMA host support': numa}
 -summary_info += {'capstone':          capstone}
 +summary_info += {'PBS bdrv support':  config_host.has_key('CONFIG_PBS_BDRV')}
  summary_info += {'libpmem support':   libpmem}
  summary_info += {'libdaxctl support': libdaxctl}
- summary_info += {'libudev':           libudev}
+ summary_info += {'libcbor support':   libcbor}
 diff --git a/qapi/block-core.json b/qapi/block-core.json
-index 4ec70acf95..47118bf83e 100644
+index 68caf30084..d45e8975a7 100644
 --- a/qapi/block-core.json
 +++ b/qapi/block-core.json
-@@ -3301,6 +3301,7 @@
+@@ -3466,6 +3466,7 @@
              'parallels', 'preallocate', 'qcow', 'qcow2', 'qed', 'quorum',
              'raw', 'rbd',
              { 'name': 'replication', 'if': 'CONFIG_REPLICATION' },
@@ -414,7 +372,7 @@ index 4ec70acf95..47118bf83e 100644
              'ssh', 'throttle', 'vdi', 'vhdx',
              { 'name': 'virtio-blk-vfio-pci', 'if': 'CONFIG_BLKIO' },
              { 'name': 'virtio-blk-vhost-user', 'if': 'CONFIG_BLKIO' },
-@@ -3377,6 +3378,17 @@
+@@ -3552,6 +3553,33 @@
  { 'struct': 'BlockdevOptionsNull',
    'data': { '*size': 'int', '*latency-ns': 'uint64', '*read-zeroes': 'bool' } }
  
@@ -423,6 +381,22 @@ index 4ec70acf95..47118bf83e 100644
 +#
 +# Driver specific block device options for the PBS backend.
 +#
++# @repository: Proxmox Backup Server repository.
++#
++# @snapshot: backup snapshots ID.
++#
++# @archive: archive name.
++#
++# @keyfile: keyfile to use for encryption.
++#
++# @password: password to use for connection.
++#
++# @fingerprint: backup server fingerprint.
++#
++# @key_password: password to unlock key.
++#
++# @namespace: namespace where backup snapshot lives.
++#
 +##
 +{ 'struct': 'BlockdevOptionsPbs',
 +  'data': { 'repository': 'str', 'snapshot': 'str', 'archive': 'str',
@@ -432,7 +406,7 @@ index 4ec70acf95..47118bf83e 100644
  ##
  # @BlockdevOptionsNVMe:
  #
-@@ -4750,6 +4762,7 @@
+@@ -4993,6 +5021,7 @@
        'nfs':        'BlockdevOptionsNfs',
        'null-aio':   'BlockdevOptionsNull',
        'null-co':    'BlockdevOptionsNull',
@@ -441,10 +415,10 @@ index 4ec70acf95..47118bf83e 100644
        'nvme-io_uring': { 'type': 'BlockdevOptionsNvmeIoUring',
                           'if': 'CONFIG_BLKIO' },
 diff --git a/qapi/pragma.json b/qapi/pragma.json
-index 325e684411..b6079f6a0e 100644
+index 6aaa9cb975..e9c595c4ba 100644
 --- a/qapi/pragma.json
 +++ b/qapi/pragma.json
-@@ -45,6 +45,7 @@
+@@ -91,6 +91,7 @@
          'BlockInfo',                # query-block
          'BlockdevAioOptions',       # blockdev-add, -blockdev
          'BlockdevDriver',           # blockdev-add, query-blockstats, ...

debian/patches/pve/0033-PVE-redirect-stderr-to-journal-when-daemonized.patch

@@ -9,15 +9,15 @@ fitting.
 Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
 Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
 ---
- meson.build | 2 ++
+ meson.build | 3 ++-
  os-posix.c  | 7 +++++--
- 2 files changed, 7 insertions(+), 2 deletions(-)
+ 2 files changed, 7 insertions(+), 3 deletions(-)
 
 diff --git a/meson.build b/meson.build
-index 69afe3441b..b2e9b2aec7 100644
+index 4c85736ec3..57f666d722 100644
 --- a/meson.build
 +++ b/meson.build
-@@ -1528,6 +1528,7 @@ keyutils = dependency('libkeyutils', required: false,
+@@ -2130,6 +2130,7 @@ endif
  has_gettid = cc.has_function('gettid')
  
  libuuid = cc.find_library('uuid', required: true)
@@ -25,28 +25,29 @@ index 69afe3441b..b2e9b2aec7 100644
  libproxmox_backup_qemu = cc.find_library('proxmox_backup_qemu', required: true)
  
  # libselinux
-@@ -3144,6 +3145,7 @@ if have_block
-   # os-posix.c contains POSIX-specific functions used by qemu-storage-daemon,
-   # os-win32.c does not
-   blockdev_ss.add(when: 'CONFIG_POSIX', if_true: files('os-posix.c'))
-+  blockdev_ss.add(when: 'CONFIG_POSIX', if_true: libsystemd)
-   softmmu_ss.add(when: 'CONFIG_WIN32', if_true: [files('os-win32.c')])
+@@ -3744,7 +3745,7 @@ if have_block
+   if host_os == 'windows'
+     system_ss.add(files('os-win32.c'))
+   else
+-    blockdev_ss.add(files('os-posix.c'))
++    blockdev_ss.add(files('os-posix.c'), libsystemd)
+   endif
  endif
  
 diff --git a/os-posix.c b/os-posix.c
-index 90ea71725f..33745a8c22 100644
+index 43f9a43f3f..a47e46d1c2 100644
 --- a/os-posix.c
 +++ b/os-posix.c
-@@ -28,6 +28,8 @@
+@@ -29,6 +29,8 @@
  #include <pwd.h>
  #include <grp.h>
  #include <libgen.h>
 +#include <systemd/sd-journal.h>
 +#include <syslog.h>
  
- /* Needed early for CONFIG_BSD etc. */
- #include "net/slirp.h"
-@@ -301,9 +303,10 @@ void os_setup_post(void)
+ #include "qemu/error-report.h"
+ #include "qemu/log.h"
+@@ -306,9 +308,10 @@ void os_setup_post(void)
  
          dup2(fd, 0);
          dup2(fd, 1);

debian/patches/pve/0034-PVE-Migrate-dirty-bitmap-state-via-savevm.patch

@@ -26,47 +26,52 @@ Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
  create mode 100644 migration/pbs-state.c
 
 diff --git a/include/migration/misc.h b/include/migration/misc.h
-index 8b49841016..78f63ca400 100644
+index 804eb23c06..c75b146ae6 100644
 --- a/include/migration/misc.h
 +++ b/include/migration/misc.h
-@@ -77,4 +77,7 @@ bool migration_in_bg_snapshot(void);
- /* migration/block-dirty-bitmap.c */
- void dirty_bitmap_mig_init(void);
+@@ -106,4 +106,7 @@ bool migration_incoming_postcopy_advised(void);
+ /* True if background snapshot is active */
+ bool migration_in_bg_snapshot(void);
  
 +/* migration/pbs-state.c */
 +void pbs_state_mig_init(void);
 +
  #endif
 diff --git a/migration/meson.build b/migration/meson.build
-index a7824b5266..de6a271b58 100644
+index 075b013971..eca57cb2a3 100644
 --- a/migration/meson.build
 +++ b/migration/meson.build
-@@ -6,8 +6,10 @@ migration_files = files(
-   'vmstate.c',
+@@ -8,6 +8,7 @@ migration_files = files(
    'qemu-file.c',
    'yank_functions.c',
-+  'pbs-state.c',
  )
- softmmu_ss.add(migration_files)
-+softmmu_ss.add(libproxmox_backup_qemu)
++system_ss.add(libproxmox_backup_qemu)
  
- softmmu_ss.add(files(
+ system_ss.add(files(
    'block-dirty-bitmap.c',
+@@ -27,6 +28,7 @@ system_ss.add(files(
+   'multifd-zlib.c',
+   'multifd-zero-page.c',
+   'options.c',
++  'pbs-state.c',
+   'postcopy-ram.c',
+   'savevm.c',
+   'savevm-async.c',
 diff --git a/migration/migration.c b/migration/migration.c
-index 99f86bd6c2..db229e72c9 100644
+index 8c5bd0a75c..491d9aa017 100644
 --- a/migration/migration.c
 +++ b/migration/migration.c
-@@ -245,6 +245,7 @@ void migration_object_init(void)
-     blk_mig_init();
-     ram_mig_init();
-     dirty_bitmap_mig_init();
+@@ -266,6 +266,7 @@ void migration_object_init(void)
+ 
+     /* Initialize cpu throttle timers */
+     cpu_throttle_init();
 +    pbs_state_mig_init();
  }
  
- void migration_cancel(const Error *error)
+ typedef struct {
 diff --git a/migration/pbs-state.c b/migration/pbs-state.c
 new file mode 100644
-index 0000000000..887e998b9e
+index 0000000000..a97187e4d7
 --- /dev/null
 +++ b/migration/pbs-state.c
 @@ -0,0 +1,104 @@
@@ -115,7 +120,7 @@ index 0000000000..887e998b9e
 +}
 +
 +/* serialize PBS state and send to target via f, called on source */
-+static int pbs_state_save_setup(QEMUFile *f, void *opaque)
++static int pbs_state_save_setup(QEMUFile *f, void *opaque, Error **errp)
 +{
 +    size_t buf_size;
 +    uint8_t *buf = proxmox_export_state(&buf_size);
@@ -175,10 +180,10 @@ index 0000000000..887e998b9e
 +                         NULL);
 +}
 diff --git a/pve-backup.c b/pve-backup.c
-index dd72ee0ed6..cb5312fff3 100644
+index fea0152de0..faa6a9b93c 100644
 --- a/pve-backup.c
 +++ b/pve-backup.c
-@@ -1090,6 +1090,7 @@ ProxmoxSupportStatus *qmp_query_proxmox_support(Error **errp)
+@@ -1083,6 +1083,7 @@ ProxmoxSupportStatus *qmp_query_proxmox_support(Error **errp)
      ret->pbs_library_version = g_strdup(proxmox_backup_qemu_version());
      ret->pbs_dirty_bitmap = true;
      ret->pbs_dirty_bitmap_savevm = true;
@@ -187,10 +192,10 @@ index dd72ee0ed6..cb5312fff3 100644
      ret->pbs_masterkey = true;
      ret->backup_max_workers = true;
 diff --git a/qapi/block-core.json b/qapi/block-core.json
-index 47118bf83e..809f3c61bc 100644
+index d45e8975a7..9795247c1f 100644
 --- a/qapi/block-core.json
 +++ b/qapi/block-core.json
-@@ -984,6 +984,11 @@
+@@ -1004,6 +1004,11 @@
  # @pbs-dirty-bitmap-savevm: True if 'dirty-bitmaps' migration capability can
  #                           safely be set for savevm-async.
  #
@@ -202,7 +207,7 @@ index 47118bf83e..809f3c61bc 100644
  # @pbs-masterkey: True if the QMP backup call supports the 'master_keyfile'
  #                 parameter.
  #
-@@ -994,6 +999,7 @@
+@@ -1017,6 +1022,7 @@
    'data': { 'pbs-dirty-bitmap': 'bool',
              'query-bitmap-info': 'bool',
              'pbs-dirty-bitmap-savevm': 'bool',

debian/patches/pve/0035-migration-block-dirty-bitmap-migrate-other-bitmaps-e.patch

@@ -15,18 +15,22 @@ transferred.
 Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
 Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
 ---
- migration/block-dirty-bitmap.c | 2 +-
- 1 file changed, 1 insertion(+), 1 deletion(-)
+ migration/block-dirty-bitmap.c | 6 +++++-
+ 1 file changed, 5 insertions(+), 1 deletion(-)
 
 diff --git a/migration/block-dirty-bitmap.c b/migration/block-dirty-bitmap.c
-index fe73aa94b1..a6440929fa 100644
+index a7d55048c2..44078ea670 100644
 --- a/migration/block-dirty-bitmap.c
 +++ b/migration/block-dirty-bitmap.c
-@@ -539,7 +539,7 @@ static int add_bitmaps_to_list(DBMSaveState *s, BlockDriverState *bs,
+@@ -539,7 +539,11 @@ static int add_bitmaps_to_list(DBMSaveState *s, BlockDriverState *bs,
+         }
  
-         if (bdrv_dirty_bitmap_check(bitmap, BDRV_BITMAP_DEFAULT, &local_err)) {
-             error_report_err(local_err);
+         if (bdrv_dirty_bitmap_check(bitmap, BDRV_BITMAP_DEFAULT, errp)) {
 -            return -1;
++            if (errp != NULL) {
++                error_report_err(*errp);
++                *errp = NULL;
++            }
 +            continue;
          }
  

debian/patches/pve/0036-PVE-fall-back-to-open-iscsi-initiatorname.patch

@@ -21,7 +21,7 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
  1 file changed, 30 insertions(+)
 
 diff --git a/block/iscsi.c b/block/iscsi.c
-index 9fc0bed90b..1d40933165 100644
+index 979bf90cb7..961714a4be 100644
 --- a/block/iscsi.c
 +++ b/block/iscsi.c
 @@ -1392,12 +1392,42 @@ static char *get_initiator_name(QemuOpts *opts)

debian/patches/pve/0037-PVE-block-stream-increase-chunk-size.patch

@@ -11,7 +11,7 @@ Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
  1 file changed, 1 insertion(+), 1 deletion(-)
 
 diff --git a/block/stream.c b/block/stream.c
-index 7f9e1ecdbb..6a29d80398 100644
+index 9076203193..1d1c65f061 100644
 --- a/block/stream.c
 +++ b/block/stream.c
 @@ -27,7 +27,7 @@ enum {

debian/patches/pve/0038-block-add-alloc-track-driver.patch

@@ -19,27 +19,33 @@ well.
 This only worked if the target supports backing images, so up until now
 only for qcow2, with alloc-track any driver for the target can be used.
 
-If 'auto-remove' is set, alloc-track will automatically detach itself
-once the backing image is removed. It will be replaced by 'file'.
+Replacing the node cannot be done in the
+track_co_change_backing_file() callback, because replacing a node
+cannot happen in a coroutine and requires the block graph lock
+exclusively. Could either become a special option for the stream job,
+or maybe the upcoming blockdev-replace QMP command can be used in the
+future.
 
 Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
 Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
 [FE: adapt to changed function signatures
      make error return value consistent with QEMU
-     avoid premature break during read]
+     avoid premature break during read
+     adhere to block graph lock requirements]
 Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
 ---
- block/alloc-track.c | 352 ++++++++++++++++++++++++++++++++++++++++++++
+ block/alloc-track.c | 366 ++++++++++++++++++++++++++++++++++++++++++++
  block/meson.build   |   1 +
- 2 files changed, 353 insertions(+)
+ block/stream.c      |  34 ++++
+ 3 files changed, 401 insertions(+)
  create mode 100644 block/alloc-track.c
 
 diff --git a/block/alloc-track.c b/block/alloc-track.c
 new file mode 100644
-index 0000000000..b75d7c6460
+index 0000000000..b4a9851144
 --- /dev/null
 +++ b/block/alloc-track.c
-@@ -0,0 +1,352 @@
+@@ -0,0 +1,366 @@
 +/*
 + * Node to allow backing images to be applied to any node. Assumes a blank
 + * image to begin with, only new writes are tracked as allocated, thus this
@@ -56,9 +62,11 @@ index 0000000000..b75d7c6460
 +#include "qapi/error.h"
 +#include "block/block_int.h"
 +#include "block/dirty-bitmap.h"
++#include "block/graph-lock.h"
 +#include "qapi/qmp/qdict.h"
 +#include "qapi/qmp/qstring.h"
 +#include "qemu/cutils.h"
++#include "qemu/error-report.h"
 +#include "qemu/option.h"
 +#include "qemu/module.h"
 +#include "sysemu/block-backend.h"
@@ -67,12 +75,12 @@ index 0000000000..b75d7c6460
 +
 +typedef enum DropState {
 +    DropNone,
-+    DropRequested,
 +    DropInProgress,
 +} DropState;
 +
 +typedef struct {
 +    BdrvDirtyBitmap *bitmap;
++    uint64_t granularity;
 +    DropState drop_state;
 +    bool auto_remove;
 +} BDRVAllocTrackState;
@@ -91,26 +99,29 @@ index 0000000000..b75d7c6460
 +    },
 +};
 +
-+static void track_refresh_limits(BlockDriverState *bs, Error **errp)
++static void GRAPH_RDLOCK
++track_refresh_limits(BlockDriverState *bs, Error **errp)
 +{
-+    BlockDriverInfo bdi;
++    BDRVAllocTrackState *s = bs->opaque;
 +
 +    if (!bs->file) {
 +        return;
 +    }
 +
-+    /* always use alignment from underlying write device so RMW cycle for
-+     * bdrv_pwritev reads data from our backing via track_co_preadv (no partial
-+     * cluster allocation in 'file') */
-+    bdrv_get_info(bs->file->bs, &bdi);
++    /*
++     * Always use alignment from underlying write device so RMW cycle for
++     * bdrv_pwritev reads data from our backing via track_co_preadv. Also use at
++     * least the bitmap granularity.
++     */
 +    bs->bl.request_alignment = MAX(bs->file->bs->bl.request_alignment,
-+                                   MAX(bdi.cluster_size, BDRV_SECTOR_SIZE));
++                                   s->granularity);
 +}
 +
 +static int track_open(BlockDriverState *bs, QDict *options, int flags,
 +                      Error **errp)
 +{
 +    BDRVAllocTrackState *s = bs->opaque;
++    BdrvChild *file = NULL;
 +    QemuOpts *opts;
 +    Error *local_err = NULL;
 +    int ret = 0;
@@ -126,18 +137,45 @@ index 0000000000..b75d7c6460
 +    s->auto_remove = qemu_opt_get_bool(opts, TRACK_OPT_AUTO_REMOVE, false);
 +
 +    /* open the target (write) node, backing will be attached by block layer */
-+    bs->file = bdrv_open_child(NULL, options, "file", bs, &child_of_bds,
-+                               BDRV_CHILD_DATA | BDRV_CHILD_METADATA, false,
-+                               &local_err);
++    file = bdrv_open_child(NULL, options, "file", bs, &child_of_bds,
++                           BDRV_CHILD_DATA | BDRV_CHILD_METADATA, false,
++                           &local_err);
++    bdrv_graph_wrlock();
++    bs->file = file;
++    bdrv_graph_wrunlock();
 +    if (local_err) {
 +        ret = -EINVAL;
 +        error_propagate(errp, local_err);
 +        goto fail;
 +    }
 +
++    bdrv_graph_rdlock_main_loop();
++    BlockDriverInfo bdi = {0};
++    ret = bdrv_get_info(bs->file->bs, &bdi);
++    if (ret < 0) {
++        /*
++         * Not a hard failure. Worst that can happen is partial cluster
++         * allocation in the write target. However, the driver here returns its
++         * allocation status based on the dirty bitmap, so any other data that
++         * maps to such a cluster will still be copied later by a stream job (or
++         * during writes to that cluster).
++         */
++        warn_report("alloc-track: unable to query cluster size for write target: %s",
++                    strerror(ret));
++    }
++    ret = 0;
++    /*
++     * Always consider alignment from underlying write device so RMW cycle for
++     * bdrv_pwritev reads data from our backing via track_co_preadv. Also try to
++     * avoid partial cluster allocation in the write target by considering the
++     * cluster size.
++     */
++    s->granularity = MAX(bs->file->bs->bl.request_alignment,
++                         MAX(bdi.cluster_size, BDRV_SECTOR_SIZE));
 +    track_refresh_limits(bs, errp);
-+    uint64_t gran = bs->bl.request_alignment;
-+    s->bitmap = bdrv_create_dirty_bitmap(bs->file->bs, gran, NULL, &local_err);
++    s->bitmap = bdrv_create_dirty_bitmap(bs->file->bs, s->granularity, NULL,
++                                         &local_err);
++    bdrv_graph_rdunlock_main_loop();
 +    if (local_err) {
 +        ret = -EIO;
 +        error_propagate(errp, local_err);
@@ -148,7 +186,9 @@ index 0000000000..b75d7c6460
 +
 +fail:
 +    if (ret < 0) {
++        bdrv_graph_wrlock();
 +        bdrv_unref_child(bs, bs->file);
++        bdrv_graph_wrunlock();
 +        if (s->bitmap) {
 +            bdrv_release_dirty_bitmap(s->bitmap);
 +        }
@@ -165,13 +205,15 @@ index 0000000000..b75d7c6460
 +    }
 +}
 +
-+static coroutine_fn int64_t track_co_getlength(BlockDriverState *bs)
++static coroutine_fn int64_t GRAPH_RDLOCK
++track_co_getlength(BlockDriverState *bs)
 +{
 +    return bdrv_co_getlength(bs->file->bs);
 +}
 +
-+static int coroutine_fn track_co_preadv(BlockDriverState *bs,
-+    int64_t offset, int64_t bytes, QEMUIOVector *qiov, BdrvRequestFlags flags)
++static int coroutine_fn GRAPH_RDLOCK
++track_co_preadv(BlockDriverState *bs, int64_t offset, int64_t bytes,
++                QEMUIOVector *qiov, BdrvRequestFlags flags)
 +{
 +    BDRVAllocTrackState *s = bs->opaque;
 +    QEMUIOVector local_qiov;
@@ -229,36 +271,39 @@ index 0000000000..b75d7c6460
 +    return ret;
 +}
 +
-+static int coroutine_fn track_co_pwritev(BlockDriverState *bs,
-+    int64_t offset, int64_t bytes, QEMUIOVector *qiov, BdrvRequestFlags flags)
++static int coroutine_fn GRAPH_RDLOCK
++track_co_pwritev(BlockDriverState *bs, int64_t offset, int64_t bytes,
++                 QEMUIOVector *qiov, BdrvRequestFlags flags)
 +{
 +    return bdrv_co_pwritev(bs->file, offset, bytes, qiov, flags);
 +}
 +
-+static int coroutine_fn track_co_pwrite_zeroes(BlockDriverState *bs,
-+    int64_t offset, int64_t bytes, BdrvRequestFlags flags)
++static int coroutine_fn GRAPH_RDLOCK
++track_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset, int64_t bytes,
++                       BdrvRequestFlags flags)
 +{
 +    return bdrv_co_pwrite_zeroes(bs->file, offset, bytes, flags);
 +}
 +
-+static int coroutine_fn track_co_pdiscard(BlockDriverState *bs,
-+    int64_t offset, int64_t bytes)
++static int coroutine_fn GRAPH_RDLOCK
++track_co_pdiscard(BlockDriverState *bs, int64_t offset, int64_t bytes)
 +{
 +    return bdrv_co_pdiscard(bs->file, offset, bytes);
 +}
 +
-+static coroutine_fn int track_co_flush(BlockDriverState *bs)
++static coroutine_fn int GRAPH_RDLOCK
++track_co_flush(BlockDriverState *bs)
 +{
 +    return bdrv_co_flush(bs->file->bs);
 +}
 +
-+static int coroutine_fn track_co_block_status(BlockDriverState *bs,
-+                                              bool want_zero,
-+                                              int64_t offset,
-+                                              int64_t bytes,
-+                                              int64_t *pnum,
-+                                              int64_t *map,
-+                                              BlockDriverState **file)
++static int coroutine_fn GRAPH_RDLOCK
++track_co_block_status(BlockDriverState *bs, bool want_zero,
++                                            int64_t offset,
++                                            int64_t bytes,
++                                            int64_t *pnum,
++                                            int64_t *map,
++                                            BlockDriverState **file)
 +{
 +    BDRVAllocTrackState *s = bs->opaque;
 +
@@ -284,10 +329,10 @@ index 0000000000..b75d7c6460
 +    return 0;
 +}
 +
-+static void track_child_perm(BlockDriverState *bs, BdrvChild *c,
-+                             BdrvChildRole role, BlockReopenQueue *reopen_queue,
-+                             uint64_t perm, uint64_t shared,
-+                             uint64_t *nperm, uint64_t *nshared)
++static void GRAPH_RDLOCK
++track_child_perm(BlockDriverState *bs, BdrvChild *c, BdrvChildRole role,
++                 BlockReopenQueue *reopen_queue, uint64_t perm, uint64_t shared,
++                 uint64_t *nperm, uint64_t *nshared)
 +{
 +    BDRVAllocTrackState *s = bs->opaque;
 +
@@ -310,53 +355,28 @@ index 0000000000..b75d7c6460
 +    }
 +}
 +
-+static void track_drop(void *opaque)
++static int coroutine_fn GRAPH_RDLOCK
++track_co_change_backing_file(BlockDriverState *bs, const char *backing_file,
++                             const char *backing_fmt)
 +{
-+    BlockDriverState *bs = (BlockDriverState*)opaque;
-+    BlockDriverState *file = bs->file->bs;
-+    BDRVAllocTrackState *s = bs->opaque;
-+
-+    assert(file);
-+
-+    /* we rely on the fact that we're not used anywhere else, so let's wait
-+     * until we're only used once - in the drive connected to the guest (and one
-+     * ref is held by bdrv_ref in track_change_backing_file) */
-+    if (bs->refcnt > 2) {
-+        aio_bh_schedule_oneshot(qemu_get_aio_context(), track_drop, opaque);
-+        return;
-+    }
-+    AioContext *aio_context = bdrv_get_aio_context(bs);
-+    aio_context_acquire(aio_context);
-+
-+    bdrv_drained_begin(bs);
-+
-+    /* now that we're drained, we can safely set 'DropInProgress' */
-+    s->drop_state = DropInProgress;
-+    bdrv_child_refresh_perms(bs, bs->file, &error_abort);
-+
-+    bdrv_replace_node(bs, file, &error_abort);
-+    bdrv_set_backing_hd(bs, NULL, &error_abort);
-+    bdrv_drained_end(bs);
-+    bdrv_unref(bs);
-+    aio_context_release(aio_context);
-+}
-+
-+static int track_change_backing_file(BlockDriverState *bs,
-+                                     const char *backing_file,
-+                                     const char *backing_fmt)
-+{
-+    BDRVAllocTrackState *s = bs->opaque;
-+    if (s->auto_remove && s->drop_state == DropNone &&
-+        backing_file == NULL && backing_fmt == NULL)
-+    {
-+        /* backing file has been disconnected, there's no longer any use for
-+         * this node, so let's remove ourselves from the block graph - we need
-+         * to schedule this for later however, since when this function is
-+         * called, the blockjob modifying us is probably not done yet and has a
-+         * blocker on 'bs' */
-+        s->drop_state = DropRequested;
-+        bdrv_ref(bs);
-+        aio_bh_schedule_oneshot(qemu_get_aio_context(), track_drop, (void*)bs);
++    /*
++     * Note that the actual backing file graph change is already done in the
++     * stream job itself with bdrv_set_backing_hd_drained(), so no need to
++     * actually do anything here. But still needs to be implemented, to make
++     * our caller (i.e. bdrv_co_change_backing_file() do the right thing).
++     *
++     * FIXME
++     * We'd like to auto-remove ourselves from the block graph, but it cannot
++     * be done from a coroutine. Currently done in the stream job, where it
++     * kinda fits better, but in the long-term, a special parameter would be
++     * nice (or done via qemu-server via upcoming blockdev-replace QMP command).
++     */
++    if (backing_file == NULL) {
++        BDRVAllocTrackState *s = bs->opaque;
++        bdrv_drained_begin(bs);
++        s->drop_state = DropInProgress;
++        bdrv_child_refresh_perms(bs, bs->file, &error_abort);
++        bdrv_drained_end(bs);
 +    }
 +
 +    return 0;
@@ -366,7 +386,7 @@ index 0000000000..b75d7c6460
 +    .format_name                      = "alloc-track",
 +    .instance_size                    = sizeof(BDRVAllocTrackState),
 +
-+    .bdrv_file_open                   = track_open,
++    .bdrv_open                        = track_open,
 +    .bdrv_close                       = track_close,
 +    .bdrv_co_getlength                = track_co_getlength,
 +    .bdrv_child_perm                  = track_child_perm,
@@ -383,7 +403,7 @@ index 0000000000..b75d7c6460
 +    .supports_backing                 = true,
 +
 +    .bdrv_co_block_status             = track_co_block_status,
-+    .bdrv_change_backing_file         = track_change_backing_file,
++    .bdrv_co_change_backing_file      = track_co_change_backing_file,
 +};
 +
 +static void bdrv_alloc_track_init(void)
@@ -393,7 +413,7 @@ index 0000000000..b75d7c6460
 +
 +block_init(bdrv_alloc_track_init);
 diff --git a/block/meson.build b/block/meson.build
-index eece0d5743..8a68162cc0 100644
+index e178047ec9..7ef7250d31 100644
 --- a/block/meson.build
 +++ b/block/meson.build
 @@ -2,6 +2,7 @@ block_ss.add(genh)
@@ -404,3 +424,48 @@ index eece0d5743..8a68162cc0 100644
    'amend.c',
    'backup.c',
    'backup-dump.c',
+diff --git a/block/stream.c b/block/stream.c
+index 1d1c65f061..d499c8883f 100644
+--- a/block/stream.c
++++ b/block/stream.c
+@@ -120,6 +120,40 @@ static int stream_prepare(Job *job)
+             ret = -EPERM;
+             goto out;
+         }
++
++        /*
++         * This cannot be done in the co_change_backing_file callback, because
++         * bdrv_replace_node() cannot be done in a coroutine. The latter also
++         * requires the graph lock exclusively. Only required for the
++         * alloc-track driver.
++         *
++         * The long-term plan is to either have an explicit parameter for the
++         * stream job or use the upcoming blockdev-replace QMP command.
++         */
++        if (base_id == NULL && strcmp(unfiltered_bs->drv->format_name, "alloc-track") == 0) {
++            BlockDriverState *file_bs;
++
++            bdrv_graph_rdlock_main_loop();
++            file_bs = unfiltered_bs->file->bs;
++            bdrv_graph_rdunlock_main_loop();
++
++            bdrv_ref(unfiltered_bs); // unrefed by bdrv_replace_node()
++            bdrv_drained_begin(file_bs);
++            bdrv_graph_wrlock();
++
++            bdrv_replace_node(unfiltered_bs, file_bs, &local_err);
++
++            bdrv_graph_wrunlock();
++            bdrv_drained_end(file_bs);
++            bdrv_unref(unfiltered_bs);
++
++            if (local_err) {
++                error_prepend(&local_err, "failed to replace alloc-track node: ");
++                error_report_err(local_err);
++                ret = -EPERM;
++                goto out;
++            }
++        }
+     }
+ 
+ out:

debian/patches/pve/0038-block-io-accept-NULL-qiov-in-bdrv_pad_request.patch

@@ -1,33 +0,0 @@
-From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
-From: Stefan Reiter <s.reiter@proxmox.com>
-Date: Tue, 2 Mar 2021 16:11:54 +0100
-Subject: [PATCH] block/io: accept NULL qiov in bdrv_pad_request
-
-Some operations, e.g. block-stream, perform reads while discarding the
-results (only copy-on-read matters). In this case they will pass NULL as
-the target QEMUIOVector, which will however trip bdrv_pad_request, since
-it wants to extend its passed vector.
-
-Simply check for NULL and do nothing, there's no reason to pad the
-target if it will be discarded anyway.
-
-Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
----
- block/io.c | 4 ++++
- 1 file changed, 4 insertions(+)
-
-diff --git a/block/io.c b/block/io.c
-index 2e267a85ab..449a44bf20 100644
---- a/block/io.c
-+++ b/block/io.c
-@@ -1576,6 +1576,10 @@ static int bdrv_pad_request(BlockDriverState *bs,
- {
-     int ret;
- 
-+    if (!qiov) {
-+        return 0;
-+    }
-+
-     bdrv_check_qiov_request(*offset, *bytes, *qiov, *qiov_offset, &error_abort);
- 
-     if (!bdrv_init_padding(bs, *offset, *bytes, pad)) {

debian/patches/pve/0039-Revert-block-rbd-workaround-for-ceph-issue-53784.patch

@@ -7,15 +7,16 @@ This reverts commit fc176116cdea816ceb8dd969080b2b95f58edbc0 in
 preparation to revert 0347a8fd4c3faaedf119be04c197804be40a384b.
 
 Signed-off-by: Fabian Ebner <f.ebner@proxmox.com>
+Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
 ---
  block/rbd.c | 42 ++----------------------------------------
  1 file changed, 2 insertions(+), 40 deletions(-)
 
 diff --git a/block/rbd.c b/block/rbd.c
-index a4749f3b1b..53e0396b51 100644
+index 728bce3b1e..6c9a8e0add 100644
 --- a/block/rbd.c
 +++ b/block/rbd.c
-@@ -1511,7 +1511,6 @@ static int coroutine_fn qemu_rbd_co_block_status(BlockDriverState *bs,
+@@ -1515,7 +1515,6 @@ static int coroutine_fn qemu_rbd_co_block_status(BlockDriverState *bs,
      int status, r;
      RBDDiffIterateReq req = { .offs = offset };
      uint64_t features, flags;
@@ -23,7 +24,7 @@ index a4749f3b1b..53e0396b51 100644
  
      assert(offset + bytes <= s->image_size);
  
-@@ -1539,43 +1538,7 @@ static int coroutine_fn qemu_rbd_co_block_status(BlockDriverState *bs,
+@@ -1543,43 +1542,7 @@ static int coroutine_fn qemu_rbd_co_block_status(BlockDriverState *bs,
          return status;
      }
  
@@ -68,7 +69,7 @@ index a4749f3b1b..53e0396b51 100644
                            qemu_rbd_diff_iterate_cb, &req);
      if (r < 0 && r != QEMU_RBD_EXIT_DIFF_ITERATE2) {
          return status;
-@@ -1594,8 +1557,7 @@ static int coroutine_fn qemu_rbd_co_block_status(BlockDriverState *bs,
+@@ -1598,8 +1561,7 @@ static int coroutine_fn qemu_rbd_co_block_status(BlockDriverState *bs,
          status = BDRV_BLOCK_ZERO | BDRV_BLOCK_OFFSET_VALID;
      }
  

debian/patches/pve/0040-Revert-block-rbd-fix-handling-of-holes-in-.bdrv_co_b.patch

@@ -8,15 +8,16 @@ This reverts commit 9e302f64bb407a9bb097b626da97228c2654cfee in
 preparation to revert 0347a8fd4c3faaedf119be04c197804be40a384b.
 
 Signed-off-by: Fabian Ebner <f.ebner@proxmox.com>
+Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
 ---
  block/rbd.c | 10 +++++-----
  1 file changed, 5 insertions(+), 5 deletions(-)
 
 diff --git a/block/rbd.c b/block/rbd.c
-index 53e0396b51..0913a0af39 100644
+index 6c9a8e0add..6f5fe90f3a 100644
 --- a/block/rbd.c
 +++ b/block/rbd.c
-@@ -1470,11 +1470,11 @@ static int qemu_rbd_diff_iterate_cb(uint64_t offs, size_t len,
+@@ -1474,11 +1474,11 @@ static int qemu_rbd_diff_iterate_cb(uint64_t offs, size_t len,
      RBDDiffIterateReq *req = opaque;
  
      assert(req->offs + req->bytes <= offs);

debian/patches/pve/0041-Revert-block-rbd-implement-bdrv_co_block_status.patch

@@ -18,12 +18,13 @@ Upstream bug report:
 https://gitlab.com/qemu-project/qemu/-/issues/1026
 
 Signed-off-by: Fabian Ebner <f.ebner@proxmox.com>
+Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
 ---
  block/rbd.c | 112 ----------------------------------------------------
  1 file changed, 112 deletions(-)
 
 diff --git a/block/rbd.c b/block/rbd.c
-index 0913a0af39..1dab254517 100644
+index 6f5fe90f3a..24e820d056 100644
 --- a/block/rbd.c
 +++ b/block/rbd.c
 @@ -108,12 +108,6 @@ typedef struct RBDTask {
@@ -39,7 +40,7 @@ index 0913a0af39..1dab254517 100644
  static int qemu_rbd_connect(rados_t *cluster, rados_ioctx_t *io_ctx,
                              BlockdevOptionsRbd *opts, bool cache,
                              const char *keypairs, const char *secretid,
-@@ -1456,111 +1450,6 @@ static ImageInfoSpecific *qemu_rbd_get_specific_info(BlockDriverState *bs,
+@@ -1460,111 +1454,6 @@ static ImageInfoSpecific *qemu_rbd_get_specific_info(BlockDriverState *bs,
      return spec_info;
  }
  
@@ -151,7 +152,7 @@ index 0913a0af39..1dab254517 100644
  static int64_t coroutine_fn qemu_rbd_co_getlength(BlockDriverState *bs)
  {
      BDRVRBDState *s = bs->opaque;
-@@ -1796,7 +1685,6 @@ static BlockDriver bdrv_rbd = {
+@@ -1801,7 +1690,6 @@ static BlockDriver bdrv_rbd = {
  #ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
      .bdrv_co_pwrite_zeroes  = qemu_rbd_co_pwrite_zeroes,
  #endif

debian/patches/pve/0042-alloc-track-error-out-when-auto-remove-is-not-set.patch

@@ -0,0 +1,43 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Tue, 26 Mar 2024 14:57:51 +0100
+Subject: [PATCH] alloc-track: error out when auto-remove is not set
+
+Since replacing the node now happens in the stream job, where the
+option cannot be read from (it's internal to the driver), it will
+always be treated as on.
+
+qemu-server will always set it, make sure to have other users notice
+the change (should they even exist). The option can be fully dropped
+in the future while adding a version guard in qemu-server.
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ block/alloc-track.c | 7 +++++--
+ 1 file changed, 5 insertions(+), 2 deletions(-)
+
+diff --git a/block/alloc-track.c b/block/alloc-track.c
+index b4a9851144..fc7d58a5d0 100644
+--- a/block/alloc-track.c
++++ b/block/alloc-track.c
+@@ -34,7 +34,6 @@ typedef struct {
+     BdrvDirtyBitmap *bitmap;
+     uint64_t granularity;
+     DropState drop_state;
+-    bool auto_remove;
+ } BDRVAllocTrackState;
+ 
+ static QemuOptsList runtime_opts = {
+@@ -86,7 +85,11 @@ static int track_open(BlockDriverState *bs, QDict *options, int flags,
+         goto fail;
+     }
+ 
+-    s->auto_remove = qemu_opt_get_bool(opts, TRACK_OPT_AUTO_REMOVE, false);
++    if (!qemu_opt_get_bool(opts, TRACK_OPT_AUTO_REMOVE, false)) {
++        error_setg(errp, "alloc-track: requires auto-remove option to be set to on");
++        ret = -EINVAL;
++        goto fail;
++    }
+ 
+     /* open the target (write) node, backing will be attached by block layer */
+     file = bdrv_open_child(NULL, options, "file", bs, &child_of_bds,

debian/patches/pve/0043-alloc-track-avoid-seemingly-superfluous-child-permis.patch

@@ -0,0 +1,84 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Wed, 27 Mar 2024 11:15:39 +0100
+Subject: [PATCH] alloc-track: avoid seemingly superfluous child permission
+ update
+
+Doesn't seem necessary nowadays (maybe after commit "alloc-track: fix
+deadlock during drop" where the dropping is not rescheduled and delayed
+anymore or some upstream change). Should there really be some issue,
+instead of having a drop state, this could also be just based off the
+fact whether there is still a backing child.
+
+Dumping the cumulative (shared) permissions for the BDS with a debug
+print yields the same values after this patch and with QEMU 8.1,
+namely 3 and 5.
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ block/alloc-track.c | 26 --------------------------
+ 1 file changed, 26 deletions(-)
+
+diff --git a/block/alloc-track.c b/block/alloc-track.c
+index fc7d58a5d0..b56425b7f0 100644
+--- a/block/alloc-track.c
++++ b/block/alloc-track.c
+@@ -25,15 +25,9 @@
+ 
+ #define TRACK_OPT_AUTO_REMOVE "auto-remove"
+ 
+-typedef enum DropState {
+-    DropNone,
+-    DropInProgress,
+-} DropState;
+-
+ typedef struct {
+     BdrvDirtyBitmap *bitmap;
+     uint64_t granularity;
+-    DropState drop_state;
+ } BDRVAllocTrackState;
+ 
+ static QemuOptsList runtime_opts = {
+@@ -137,8 +131,6 @@ static int track_open(BlockDriverState *bs, QDict *options, int flags,
+         goto fail;
+     }
+ 
+-    s->drop_state = DropNone;
+-
+ fail:
+     if (ret < 0) {
+         bdrv_graph_wrlock();
+@@ -289,18 +281,8 @@ track_child_perm(BlockDriverState *bs, BdrvChild *c, BdrvChildRole role,
+                  BlockReopenQueue *reopen_queue, uint64_t perm, uint64_t shared,
+                  uint64_t *nperm, uint64_t *nshared)
+ {
+-    BDRVAllocTrackState *s = bs->opaque;
+-
+     *nshared = BLK_PERM_ALL;
+ 
+-    /* in case we're currently dropping ourselves, claim to not use any
+-     * permissions at all - which is fine, since from this point on we will
+-     * never issue a read or write anymore */
+-    if (s->drop_state == DropInProgress) {
+-        *nperm = 0;
+-        return;
+-    }
+-
+     if (role & BDRV_CHILD_DATA) {
+         *nperm = perm & DEFAULT_PERM_PASSTHROUGH;
+     } else {
+@@ -326,14 +308,6 @@ track_co_change_backing_file(BlockDriverState *bs, const char *backing_file,
+      * kinda fits better, but in the long-term, a special parameter would be
+      * nice (or done via qemu-server via upcoming blockdev-replace QMP command).
+      */
+-    if (backing_file == NULL) {
+-        BDRVAllocTrackState *s = bs->opaque;
+-        bdrv_drained_begin(bs);
+-        s->drop_state = DropInProgress;
+-        bdrv_child_refresh_perms(bs, bs->file, &error_abort);
+-        bdrv_drained_end(bs);
+-    }
+-
+     return 0;
+ }
+ 

debian/patches/pve/0043-alloc-track-fix-deadlock-during-drop.patch

@@ -1,153 +0,0 @@
-From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
-From: Fiona Ebner <f.ebner@proxmox.com>
-Date: Thu, 6 Apr 2023 14:59:31 +0200
-Subject: [PATCH] alloc-track: fix deadlock during drop
-
-by replacing the block node directly after changing the backing file
-instead of rescheduling it.
-
-With changes in QEMU 8.0, calling bdrv_get_info (and bdrv_unref)
-during drop can lead to a deadlock when using iothread (only triggered
-with multiple disks, except during debugging where it also triggered
-with one disk sometimes):
-1. job_unref_locked acquires the AioContext and calls job->driver->free
-2. track_drop gets scheduled
-3. bdrv_graph_wrlock is called and polls which leads to track_drop being
-   called
-4. track_drop acquires the AioContext recursively
-5. bdrv_get_info is a wrapped coroutine (since 8.0) and thus polls for
-   bdrv_co_get_info. This releases the AioContext, but only once! The
-   documentation for the AIO_WAIT_WHILE macro states that the
-   AioContext lock needs to be acquired exactly once, but there does
-   not seem to be a way for track_drop to know if it acquired the lock
-   recursively or not (without adding further hacks).
-6. Because the AioContext is still held by the main thread once, it can't
-   be acquired before entering bdrv_co_get_info in co_schedule_bh_cb
-   which happens in the iothread
-
-When doing the operation in change_backing_file, the AioContext has
-already been acquired by the caller, so the issue with the recursive
-lock goes away.
-
-The comment explaining why delaying the replace is necessary is
-> we need to schedule this for later however, since when this function
-> is called, the blockjob modifying us is probably not done yet and
-> has a blocker on 'bs'
-
-However, there is no check for blockers in bdrv_replace_node. It would
-need to be done by us, the caller, with check_to_replace_node.
-Furthermore, the mirror job also does its call to bdrv_replace_node
-while there is an active blocker (inserted by mirror itself) and they
-use a specialized version to check for blockers instead of
-check_to_replace_node there. Alloc-track could also do something
-similar to check for other blockers, but it should be fine to rely on
-Proxmox VE that no other operation with the blockdev is going on.
-
-Mirror also drains the target before replacing the node, but the
-target can have other users. In case of alloc-track the file child
-should not be accessible by anybody else and so there can't be an
-in-flight operation for the file child when alloc-track is drained.
-
-The rescheduling based on refcounting is a hack and it doesn't seem to
-be necessary anymore. It's not clear what the original issue from the
-comment was. Testing with older builds with track_drop done directly
-without rescheduling also didn't lead to any noticable issue for me.
-
-One issue it might have been is the one fixed by b1e1af394d
-("block/stream: Drain subtree around graph change"), where
-block-stream had a use-after-free if the base node changed at an
-inconvenient time (which alloc-track's auto-drop does).
-
-It's also not possible to just not auto-replace the alloc-track. Not
-replacing it at all leads to other operations like block resize
-hanging, and there is no good way to replace it manually via QMP
-(there is x-blockdev-change, but it is experimental and doesn't
-implement the required operation yet). Also, it's just cleaner in
-general to not leave unnecessary block nodes lying around.
-
-Suggested-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
-Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
----
- block/alloc-track.c | 54 ++++++++++++++-------------------------------
- 1 file changed, 16 insertions(+), 38 deletions(-)
-
-diff --git a/block/alloc-track.c b/block/alloc-track.c
-index b75d7c6460..76da140a68 100644
---- a/block/alloc-track.c
-+++ b/block/alloc-track.c
-@@ -25,7 +25,6 @@
- 
- typedef enum DropState {
-     DropNone,
--    DropRequested,
-     DropInProgress,
- } DropState;
- 
-@@ -268,37 +267,6 @@ static void track_child_perm(BlockDriverState *bs, BdrvChild *c,
-     }
- }
- 
--static void track_drop(void *opaque)
--{
--    BlockDriverState *bs = (BlockDriverState*)opaque;
--    BlockDriverState *file = bs->file->bs;
--    BDRVAllocTrackState *s = bs->opaque;
--
--    assert(file);
--
--    /* we rely on the fact that we're not used anywhere else, so let's wait
--     * until we're only used once - in the drive connected to the guest (and one
--     * ref is held by bdrv_ref in track_change_backing_file) */
--    if (bs->refcnt > 2) {
--        aio_bh_schedule_oneshot(qemu_get_aio_context(), track_drop, opaque);
--        return;
--    }
--    AioContext *aio_context = bdrv_get_aio_context(bs);
--    aio_context_acquire(aio_context);
--
--    bdrv_drained_begin(bs);
--
--    /* now that we're drained, we can safely set 'DropInProgress' */
--    s->drop_state = DropInProgress;
--    bdrv_child_refresh_perms(bs, bs->file, &error_abort);
--
--    bdrv_replace_node(bs, file, &error_abort);
--    bdrv_set_backing_hd(bs, NULL, &error_abort);
--    bdrv_drained_end(bs);
--    bdrv_unref(bs);
--    aio_context_release(aio_context);
--}
--
- static int track_change_backing_file(BlockDriverState *bs,
-                                      const char *backing_file,
-                                      const char *backing_fmt)
-@@ -308,13 +276,23 @@ static int track_change_backing_file(BlockDriverState *bs,
-         backing_file == NULL && backing_fmt == NULL)
-     {
-         /* backing file has been disconnected, there's no longer any use for
--         * this node, so let's remove ourselves from the block graph - we need
--         * to schedule this for later however, since when this function is
--         * called, the blockjob modifying us is probably not done yet and has a
--         * blocker on 'bs' */
--        s->drop_state = DropRequested;
-+         * this node, so let's remove ourselves from the block graph */
-+        BlockDriverState *file = bs->file->bs;
-+
-+        /* Just to be sure, because bdrv_replace_node unrefs it */
-         bdrv_ref(bs);
--        aio_bh_schedule_oneshot(qemu_get_aio_context(), track_drop, (void*)bs);
-+        bdrv_drained_begin(bs);
-+
-+        /* now that we're drained, we can safely set 'DropInProgress' */
-+        s->drop_state = DropInProgress;
-+
-+        bdrv_child_refresh_perms(bs, bs->file, &error_abort);
-+
-+        bdrv_replace_node(bs, file, &error_abort);
-+        bdrv_set_backing_hd(bs, NULL, &error_abort);
-+
-+        bdrv_drained_end(bs);
-+        bdrv_unref(bs);
-     }
- 
-     return 0;

debian/patches/pve/0044-PVE-backup-add-fleecing-option.patch

@@ -0,0 +1,337 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Thu, 11 Apr 2024 11:29:28 +0200
+Subject: [PATCH] PVE backup: add fleecing option
+
+When a fleecing option is given, it is expected that each device has
+a corresponding "-fleecing" block device already attached, except for
+EFI disk and TPM state, where fleecing is never used.
+
+The following graph was adapted from [0] which also contains more
+details about fleecing.
+
+[guest]
+   |
+   | root
+   v                 file
+[copy-before-write]<------[snapshot-access]
+   |           |
+   | file      | target
+   v           v
+[source] [fleecing]
+
+For fleecing, a copy-before-write filter is inserted on top of the
+source node, as well as a snapshot-access node pointing to the filter
+node which allows to read the consistent state of the image at the
+time it was inserted. New guest writes are passed through the
+copy-before-write filter which will first copy over old data to the
+fleecing image in case that old data is still needed by the
+snapshot-access node.
+
+The backup process will sequentially read from the snapshot access,
+which has a bitmap and knows whether to read from the original image
+or the fleecing image to get the "snapshot" state, i.e. data from the
+source image at the time when the copy-before-write filter was
+inserted. After reading, the copied sections are discarded from the
+fleecing image to reduce space usage.
+
+All of this can be restricted by an initial dirty bitmap to parts of
+the source image that are required for an incremental backup.
+
+For discard to work, it is necessary that the fleecing image does not
+have a larger cluster size than the backup job granularity. Since
+querying that size does not always work, e.g. for RBD with krbd, the
+cluster size will not be reported, a minimum of 4 MiB is used. A job
+with PBS target already has at least this granularity, so it's just
+relevant for other targets. I.e. edge cases where this minimum is not
+enough should be very rare in practice. If ever necessary in the
+future, can still add a passed-in value for the backup QMP command to
+override.
+
+Additionally, the cbw-timeout and on-cbw-error=break-snapshot options
+are set when installing the copy-before-write filter and
+snapshot-access. When an error or timeout occurs, the problematic (and
+each further) snapshot operation will fail and thus cancel the backup
+instead of breaking the guest write.
+
+Note that job_id cannot be inferred from the snapshot-access bs because
+it has no parent, so just pass the one from the original bs.
+
+[0]: https://www.mail-archive.com/qemu-devel@nongnu.org/msg876056.html
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
+---
+ block/monitor/block-hmp-cmds.c |   1 +
+ pve-backup.c                   | 135 ++++++++++++++++++++++++++++++++-
+ qapi/block-core.json           |  10 ++-
+ 3 files changed, 142 insertions(+), 4 deletions(-)
+
+diff --git a/block/monitor/block-hmp-cmds.c b/block/monitor/block-hmp-cmds.c
+index 439a7a14c8..d0e7771dcc 100644
+--- a/block/monitor/block-hmp-cmds.c
++++ b/block/monitor/block-hmp-cmds.c
+@@ -1044,6 +1044,7 @@ void coroutine_fn hmp_backup(Monitor *mon, const QDict *qdict)
+         NULL, NULL,
+         devlist, qdict_haskey(qdict, "speed"), speed,
+         false, 0, // BackupPerf max-workers
++        false, false, // fleecing
+         &error);
+ 
+     hmp_handle_error(mon, error);
+diff --git a/pve-backup.c b/pve-backup.c
+index faa6a9b93c..4b0820c8a7 100644
+--- a/pve-backup.c
++++ b/pve-backup.c
+@@ -7,9 +7,11 @@
+ #include "sysemu/blockdev.h"
+ #include "block/block_int-global-state.h"
+ #include "block/blockjob.h"
++#include "block/copy-before-write.h"
+ #include "block/dirty-bitmap.h"
+ #include "block/graph-lock.h"
+ #include "qapi/qapi-commands-block.h"
++#include "qapi/qmp/qdict.h"
+ #include "qapi/qmp/qerror.h"
+ #include "qemu/cutils.h"
+ 
+@@ -80,8 +82,15 @@ static void pvebackup_init(void)
+ // initialize PVEBackupState at startup
+ opts_init(pvebackup_init);
+ 
++typedef struct PVEBackupFleecingInfo {
++    BlockDriverState *bs;
++    BlockDriverState *cbw;
++    BlockDriverState *snapshot_access;
++} PVEBackupFleecingInfo;
++
+ typedef struct PVEBackupDevInfo {
+     BlockDriverState *bs;
++    PVEBackupFleecingInfo fleecing;
+     size_t size;
+     uint64_t block_size;
+     uint8_t dev_id;
+@@ -353,6 +362,22 @@ static void pvebackup_complete_cb(void *opaque, int ret)
+     PVEBackupDevInfo *di = opaque;
+     di->completed_ret = ret;
+ 
++    /*
++     * Handle block-graph specific cleanup (for fleecing) outside of the coroutine, because the work
++     * won't be done as a coroutine anyways:
++     * - For snapshot_access, allows doing bdrv_unref() directly. Doing it via bdrv_co_unref() would
++     *   just spawn a BH calling bdrv_unref().
++     * - For cbw, draining would need to spawn a BH.
++     */
++    if (di->fleecing.snapshot_access) {
++        bdrv_unref(di->fleecing.snapshot_access);
++        di->fleecing.snapshot_access = NULL;
++    }
++    if (di->fleecing.cbw) {
++        bdrv_cbw_drop(di->fleecing.cbw);
++        di->fleecing.cbw = NULL;
++    }
++
+     /*
+      * Needs to happen outside of coroutine, because it takes the graph write lock.
+      */
+@@ -519,9 +544,77 @@ static void create_backup_jobs_bh(void *opaque) {
+         }
+         bdrv_drained_begin(di->bs);
+ 
++        BackupPerf perf = (BackupPerf){ .max_workers = backup_state.perf.max_workers };
++
++        BlockDriverState *source_bs = di->bs;
++        bool discard_source = false;
++        bdrv_graph_co_rdlock();
++        const char *job_id = bdrv_get_device_name(di->bs);
++        bdrv_graph_co_rdunlock();
++        if (di->fleecing.bs) {
++            QDict *cbw_opts = qdict_new();
++            qdict_put_str(cbw_opts, "driver", "copy-before-write");
++            qdict_put_str(cbw_opts, "file", bdrv_get_node_name(di->bs));
++            qdict_put_str(cbw_opts, "target", bdrv_get_node_name(di->fleecing.bs));
++
++            if (di->bitmap) {
++                /*
++                 * Only guest writes to parts relevant for the backup need to be intercepted with
++                 * old data being copied to the fleecing image.
++                 */
++                qdict_put_str(cbw_opts, "bitmap.node", bdrv_get_node_name(di->bs));
++                qdict_put_str(cbw_opts, "bitmap.name", bdrv_dirty_bitmap_name(di->bitmap));
++            }
++            /*
++             * Fleecing storage is supposed to be fast and it's better to break backup than guest
++             * writes. Certain guest drivers like VirtIO-win have 60 seconds timeout by default, so
++             * abort a bit before that.
++             */
++            qdict_put_str(cbw_opts, "on-cbw-error", "break-snapshot");
++            qdict_put_int(cbw_opts, "cbw-timeout", 45);
++
++            di->fleecing.cbw = bdrv_insert_node(di->bs, cbw_opts, BDRV_O_RDWR, &local_err);
++
++            if (!di->fleecing.cbw) {
++                error_setg(errp, "appending cbw node for fleecing failed: %s",
++                           local_err ? error_get_pretty(local_err) : "unknown error");
++                break;
++            }
++
++            QDict *snapshot_access_opts = qdict_new();
++            qdict_put_str(snapshot_access_opts, "driver", "snapshot-access");
++            qdict_put_str(snapshot_access_opts, "file", bdrv_get_node_name(di->fleecing.cbw));
++
++            di->fleecing.snapshot_access =
++                bdrv_open(NULL, NULL, snapshot_access_opts, BDRV_O_RDWR | BDRV_O_UNMAP, &local_err);
++            if (!di->fleecing.snapshot_access) {
++                error_setg(errp, "setting up snapshot access for fleecing failed: %s",
++                           local_err ? error_get_pretty(local_err) : "unknown error");
++                break;
++            }
++            source_bs = di->fleecing.snapshot_access;
++            discard_source = true;
++
++            /*
++             * bdrv_get_info() just retuns 0 (= doesn't matter) for RBD when using krbd. But discard
++             * on the fleecing image won't work if the backup job's granularity is less than the RBD
++             * object size (default 4 MiB), so it does matter. Always use at least 4 MiB. With a PBS
++             * target, the backup job granularity would already be at least this much.
++             */
++            perf.min_cluster_size = 4 * 1024 * 1024;
++            /*
++             * For discard to work, cluster size for the backup job must be at least the same as for
++             * the fleecing image.
++             */
++            BlockDriverInfo bdi;
++            if (bdrv_get_info(di->fleecing.bs, &bdi) >= 0) {
++                perf.min_cluster_size = MAX(perf.min_cluster_size, bdi.cluster_size);
++            }
++        }
++
+         BlockJob *job = backup_job_create(
+-            NULL, di->bs, di->target, backup_state.speed, sync_mode, di->bitmap,
+-            bitmap_mode, false, NULL, &backup_state.perf, BLOCKDEV_ON_ERROR_REPORT,
++            job_id, source_bs, di->target, backup_state.speed, sync_mode, di->bitmap,
++            bitmap_mode, false, discard_source, NULL, &perf, BLOCKDEV_ON_ERROR_REPORT,
+             BLOCKDEV_ON_ERROR_REPORT, JOB_DEFAULT, pvebackup_complete_cb, di, backup_state.txn,
+             &local_err);
+ 
+@@ -577,6 +670,14 @@ static void create_backup_jobs_bh(void *opaque) {
+     aio_co_enter(data->ctx, data->co);
+ }
+ 
++/*
++ * EFI disk and TPM state are small and it's just not worth setting up fleecing for them.
++ */
++static bool device_uses_fleecing(const char *device_id)
++{
++    return strncmp(device_id, "drive-efidisk", 13) && strncmp(device_id, "drive-tpmstate", 14);
++}
++
+ /*
+  * Returns a list of device infos, which needs to be freed by the caller. In
+  * case of an error, errp will be set, but the returned value might still be a
+@@ -584,6 +685,7 @@ static void create_backup_jobs_bh(void *opaque) {
+  */
+ static GList coroutine_fn GRAPH_RDLOCK *get_device_info(
+     const char *devlist,
++    bool fleecing,
+     Error **errp)
+ {
+     gchar **devs = NULL;
+@@ -607,6 +709,31 @@ static GList coroutine_fn GRAPH_RDLOCK *get_device_info(
+             }
+             PVEBackupDevInfo *di = g_new0(PVEBackupDevInfo, 1);
+             di->bs = bs;
++
++            if (fleecing && device_uses_fleecing(*d)) {
++                g_autofree gchar *fleecing_devid = g_strconcat(*d, "-fleecing", NULL);
++                BlockBackend *fleecing_blk = blk_by_name(fleecing_devid);
++                if (!fleecing_blk) {
++                    error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
++                              "Device '%s' not found", fleecing_devid);
++                    goto err;
++                }
++                BlockDriverState *fleecing_bs = blk_bs(fleecing_blk);
++                if (!bdrv_co_is_inserted(fleecing_bs)) {
++                    error_setg(errp, "Device '%s' has no medium", fleecing_devid);
++                    goto err;
++                }
++                /*
++                 * Fleecing image needs to be the same size to act as a cbw target.
++                 */
++                if (bs->total_sectors != fleecing_bs->total_sectors) {
++                    error_setg(errp, "Size mismatch for '%s' - sector count %ld != %ld",
++                               fleecing_devid, fleecing_bs->total_sectors, bs->total_sectors);
++                    goto err;
++                }
++                di->fleecing.bs = fleecing_bs;
++            }
++
+             di_list = g_list_append(di_list, di);
+             d++;
+         }
+@@ -656,6 +783,7 @@ UuidInfo coroutine_fn *qmp_backup(
+     const char *devlist,
+     bool has_speed, int64_t speed,
+     bool has_max_workers, int64_t max_workers,
++    bool has_fleecing, bool fleecing,
+     Error **errp)
+ {
+     assert(qemu_in_coroutine());
+@@ -684,7 +812,7 @@ UuidInfo coroutine_fn *qmp_backup(
+     format = has_format ? format : BACKUP_FORMAT_VMA;
+ 
+     bdrv_graph_co_rdlock();
+-    di_list = get_device_info(devlist, &local_err);
++    di_list = get_device_info(devlist, has_fleecing && fleecing, &local_err);
+     bdrv_graph_co_rdunlock();
+     if (local_err) {
+         error_propagate(errp, local_err);
+@@ -1087,5 +1215,6 @@ ProxmoxSupportStatus *qmp_query_proxmox_support(Error **errp)
+     ret->query_bitmap_info = true;
+     ret->pbs_masterkey = true;
+     ret->backup_max_workers = true;
++    ret->backup_fleecing = true;
+     return ret;
+ }
+diff --git a/qapi/block-core.json b/qapi/block-core.json
+index 9795247c1f..c581f1f238 100644
+--- a/qapi/block-core.json
++++ b/qapi/block-core.json
+@@ -948,6 +948,10 @@
+ #
+ # @max-workers: see @BackupPerf for details. Default 16.
+ #
++# @fleecing: perform a backup with fleecing. For each device in @devlist, a
++#            corresponing '-fleecing' device with the same size already needs to
++#            be present.
++#
+ # Returns: the uuid of the backup job
+ #
+ ##
+@@ -968,7 +972,8 @@
+                                     '*firewall-file': 'str',
+                                     '*devlist': 'str',
+                                     '*speed': 'int',
+-                                    '*max-workers': 'int' },
++                                    '*max-workers': 'int',
++                                    '*fleecing': 'bool' },
+   'returns': 'UuidInfo', 'coroutine': true }
+ 
+ ##
+@@ -1014,6 +1019,8 @@
+ #
+ # @pbs-library-version: Running version of libproxmox-backup-qemu0 library.
+ #
++# @backup-fleecing: Whether backup fleecing is supported or not.
++#
+ # @backup-max-workers: Whether the 'max-workers' @BackupPerf setting is
+ #     supported or not.
+ #
+@@ -1025,6 +1032,7 @@
+             'pbs-dirty-bitmap-migration': 'bool',
+             'pbs-masterkey': 'bool',
+             'pbs-library-version': 'str',
++            'backup-fleecing': 'bool',
+             'backup-max-workers': 'bool' } }
+ 
+ ##

debian/patches/pve/0044-migration-for-snapshots-hold-the-BQL-during-setup-ca.patch

@@ -1,190 +0,0 @@
-From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
-From: Fiona Ebner <f.ebner@proxmox.com>
-Date: Fri, 5 May 2023 13:39:53 +0200
-Subject: [PATCH] migration: for snapshots, hold the BQL during setup callbacks
-
-In spirit, this is a partial revert of commit 9b09503752 ("migration:
-run setup callbacks out of big lock"), but only for the snapshot case.
-
-For snapshots, the bdrv_writev_vmstate() function is used during setup
-(in QIOChannelBlock backing the QEMUFile), but not holding the BQL
-while calling it could lead to an assertion failure. To understand
-how, first note the following:
-
-1. Generated coroutine wrappers for block layer functions spawn the
-coroutine and use AIO_WAIT_WHILE()/aio_poll() to wait for it.
-2. If the host OS switches threads at an inconvenient time, it can
-happen that a bottom half scheduled for the main thread's AioContext
-is executed as part of a vCPU thread's aio_poll().
-
-An example leading to the assertion failure is as follows:
-
-main thread:
-1. A snapshot-save QMP command gets issued.
-2. snapshot_save_job_bh() is scheduled.
-
-vCPU thread:
-3. aio_poll() for the main thread's AioContext is called (e.g. when
-the guest writes to a pflash device, as part of blk_pwrite which is a
-generated coroutine wrapper).
-4. snapshot_save_job_bh() is executed as part of aio_poll().
-3. qemu_savevm_state() is called.
-4. qemu_mutex_unlock_iothread() is called. Now
-qemu_get_current_aio_context() returns 0x0.
-5. bdrv_writev_vmstate() is executed during the usual savevm setup.
-But this function is a generated coroutine wrapper, so it uses
-AIO_WAIT_WHILE. There, the assertion
-assert(qemu_get_current_aio_context() == qemu_get_aio_context());
-will fail.
-
-To fix it, ensure that the BQL is held during setup. To avoid changing
-the behavior for migration too, introduce conditionals for the setup
-callbacks that need the BQL and only take the lock if it's not already
-held.
-
-Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
----
- include/migration/register.h   |  2 +-
- migration/block-dirty-bitmap.c | 15 ++++++++++++---
- migration/block.c              | 15 ++++++++++++---
- migration/ram.c                | 16 +++++++++++++---
- migration/savevm.c             |  2 --
- 5 files changed, 38 insertions(+), 12 deletions(-)
-
-diff --git a/include/migration/register.h b/include/migration/register.h
-index a8dfd8fefd..fa9b0b0f10 100644
---- a/include/migration/register.h
-+++ b/include/migration/register.h
-@@ -43,9 +43,9 @@ typedef struct SaveVMHandlers {
-      * by other locks.
-      */
-     int (*save_live_iterate)(QEMUFile *f, void *opaque);
-+    int (*save_setup)(QEMUFile *f, void *opaque);
- 
-     /* This runs outside the iothread lock!  */
--    int (*save_setup)(QEMUFile *f, void *opaque);
-     /* Note for save_live_pending:
-      * must_precopy:
-      * - must be migrated in precopy or in stopped state
-diff --git a/migration/block-dirty-bitmap.c b/migration/block-dirty-bitmap.c
-index a6440929fa..69fab3275c 100644
---- a/migration/block-dirty-bitmap.c
-+++ b/migration/block-dirty-bitmap.c
-@@ -1214,10 +1214,17 @@ static int dirty_bitmap_save_setup(QEMUFile *f, void *opaque)
- {
-     DBMSaveState *s = &((DBMState *)opaque)->save;
-     SaveBitmapState *dbms = NULL;
-+    bool release_lock = false;
- 
--    qemu_mutex_lock_iothread();
-+    /* For snapshots, the BQL is held during setup. */
-+    if (!qemu_mutex_iothread_locked()) {
-+        qemu_mutex_lock_iothread();
-+        release_lock = true;
-+    }
-     if (init_dirty_bitmap_migration(s) < 0) {
--        qemu_mutex_unlock_iothread();
-+        if (release_lock) {
-+            qemu_mutex_unlock_iothread();
-+        }
-         return -1;
-     }
- 
-@@ -1225,7 +1232,9 @@ static int dirty_bitmap_save_setup(QEMUFile *f, void *opaque)
-         send_bitmap_start(f, s, dbms);
-     }
-     qemu_put_bitmap_flags(f, DIRTY_BITMAP_MIG_FLAG_EOS);
--    qemu_mutex_unlock_iothread();
-+    if (release_lock) {
-+        qemu_mutex_unlock_iothread();
-+    }
-     return 0;
- }
- 
-diff --git a/migration/block.c b/migration/block.c
-index b2497bbd32..c9d55be642 100644
---- a/migration/block.c
-+++ b/migration/block.c
-@@ -716,21 +716,30 @@ static void block_migration_cleanup(void *opaque)
- static int block_save_setup(QEMUFile *f, void *opaque)
- {
-     int ret;
-+    bool release_lock = false;
- 
-     trace_migration_block_save("setup", block_mig_state.submitted,
-                                block_mig_state.transferred);
- 
--    qemu_mutex_lock_iothread();
-+    /* For snapshots, the BQL is held during setup. */
-+    if (!qemu_mutex_iothread_locked()) {
-+        qemu_mutex_lock_iothread();
-+        release_lock = true;
-+    }
-     ret = init_blk_migration(f);
-     if (ret < 0) {
--        qemu_mutex_unlock_iothread();
-+        if (release_lock) {
-+            qemu_mutex_unlock_iothread();
-+        }
-         return ret;
-     }
- 
-     /* start track dirty blocks */
-     ret = set_dirty_tracking();
- 
--    qemu_mutex_unlock_iothread();
-+    if (release_lock) {
-+        qemu_mutex_unlock_iothread();
-+    }
- 
-     if (ret) {
-         return ret;
-diff --git a/migration/ram.c b/migration/ram.c
-index 79d881f735..0ecbbc3202 100644
---- a/migration/ram.c
-+++ b/migration/ram.c
-@@ -3117,8 +3117,16 @@ static void migration_bitmap_clear_discarded_pages(RAMState *rs)
- 
- static void ram_init_bitmaps(RAMState *rs)
- {
--    /* For memory_global_dirty_log_start below.  */
--    qemu_mutex_lock_iothread();
-+    bool release_lock = false;
-+
-+    /*
-+     * For memory_global_dirty_log_start below.
-+     * For snapshots, the BQL is held during setup.
-+     */
-+    if (!qemu_mutex_iothread_locked()) {
-+        qemu_mutex_lock_iothread();
-+        release_lock = true;
-+    }
-     qemu_mutex_lock_ramlist();
- 
-     WITH_RCU_READ_LOCK_GUARD() {
-@@ -3130,7 +3138,9 @@ static void ram_init_bitmaps(RAMState *rs)
-         }
-     }
-     qemu_mutex_unlock_ramlist();
--    qemu_mutex_unlock_iothread();
-+    if (release_lock) {
-+        qemu_mutex_unlock_iothread();
-+    }
- 
-     /*
-      * After an eventual first bitmap sync, fixup the initial bitmap
-diff --git a/migration/savevm.c b/migration/savevm.c
-index aa54a67fda..fc6a82a555 100644
---- a/migration/savevm.c
-+++ b/migration/savevm.c
-@@ -1621,10 +1621,8 @@ static int qemu_savevm_state(QEMUFile *f, Error **errp)
-     memset(&compression_counters, 0, sizeof(compression_counters));
-     ms->to_dst_file = f;
- 
--    qemu_mutex_unlock_iothread();
-     qemu_savevm_state_header(f);
-     qemu_savevm_state_setup(f);
--    qemu_mutex_lock_iothread();
- 
-     while (qemu_file_get_error(f) == 0) {
-         if (qemu_savevm_state_iterate(f, false) > 0) {

debian/patches/pve/0045-PVE-backup-improve-error-when-copy-before-write-fail.patch

@@ -0,0 +1,117 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Mon, 29 Apr 2024 14:43:58 +0200
+Subject: [PATCH] PVE backup: improve error when copy-before-write fails for
+ fleecing
+
+With fleecing, failure for copy-before-write does not fail the guest
+write, but only sets the snapshot error that is associated to the
+copy-before-write filter, making further requests to the snapshot
+access fail with EACCES, which then also fails the job. But that error
+code is not the root cause of why the backup failed, so bubble up the
+original snapshot error instead.
+
+Reported-by: Friedrich Weber <f.weber@proxmox.com>
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+Tested-by: Friedrich Weber <f.weber@proxmox.com>
+---
+ block/copy-before-write.c | 18 ++++++++++++------
+ block/copy-before-write.h |  1 +
+ pve-backup.c              |  9 +++++++++
+ 3 files changed, 22 insertions(+), 6 deletions(-)
+
+diff --git a/block/copy-before-write.c b/block/copy-before-write.c
+index 81afeff1c7..fdf9cdc0cd 100644
+--- a/block/copy-before-write.c
++++ b/block/copy-before-write.c
+@@ -27,6 +27,7 @@
+ #include "qapi/qmp/qjson.h"
+ 
+ #include "sysemu/block-backend.h"
++#include "qemu/atomic.h"
+ #include "qemu/cutils.h"
+ #include "qapi/error.h"
+ #include "block/block_int.h"
+@@ -75,7 +76,8 @@ typedef struct BDRVCopyBeforeWriteState {
+      * @snapshot_error is normally zero. But on first copy-before-write failure
+      * when @on_cbw_error == ON_CBW_ERROR_BREAK_SNAPSHOT, @snapshot_error takes
+      * value of this error (<0). After that all in-flight and further
+-     * snapshot-API requests will fail with that error.
++     * snapshot-API requests will fail with that error. To be accessed with
++     * atomics.
+      */
+     int snapshot_error;
+ } BDRVCopyBeforeWriteState;
+@@ -115,7 +117,7 @@ static coroutine_fn int cbw_do_copy_before_write(BlockDriverState *bs,
+         return 0;
+     }
+ 
+-    if (s->snapshot_error) {
++    if (qatomic_read(&s->snapshot_error)) {
+         return 0;
+     }
+ 
+@@ -139,9 +141,7 @@ static coroutine_fn int cbw_do_copy_before_write(BlockDriverState *bs,
+     WITH_QEMU_LOCK_GUARD(&s->lock) {
+         if (ret < 0) {
+             assert(s->on_cbw_error == ON_CBW_ERROR_BREAK_SNAPSHOT);
+-            if (!s->snapshot_error) {
+-                s->snapshot_error = ret;
+-            }
++            qatomic_cmpxchg(&s->snapshot_error, 0, ret);
+         } else {
+             bdrv_set_dirty_bitmap(s->done_bitmap, off, end - off);
+         }
+@@ -215,7 +215,7 @@ cbw_snapshot_read_lock(BlockDriverState *bs, int64_t offset, int64_t bytes,
+ 
+     QEMU_LOCK_GUARD(&s->lock);
+ 
+-    if (s->snapshot_error) {
++    if (qatomic_read(&s->snapshot_error)) {
+         g_free(req);
+         return NULL;
+     }
+@@ -595,6 +595,12 @@ void bdrv_cbw_drop(BlockDriverState *bs)
+     bdrv_unref(bs);
+ }
+ 
++int bdrv_cbw_snapshot_error(BlockDriverState *bs)
++{
++    BDRVCopyBeforeWriteState *s = bs->opaque;
++    return qatomic_read(&s->snapshot_error);
++}
++
+ static void cbw_init(void)
+ {
+     bdrv_register(&bdrv_cbw_filter);
+diff --git a/block/copy-before-write.h b/block/copy-before-write.h
+index 2a5d4ba693..969da3620f 100644
+--- a/block/copy-before-write.h
++++ b/block/copy-before-write.h
+@@ -44,5 +44,6 @@ BlockDriverState *bdrv_cbw_append(BlockDriverState *source,
+                                   BlockCopyState **bcs,
+                                   Error **errp);
+ void bdrv_cbw_drop(BlockDriverState *bs);
++int bdrv_cbw_snapshot_error(BlockDriverState *bs);
+ 
+ #endif /* COPY_BEFORE_WRITE_H */
+diff --git a/pve-backup.c b/pve-backup.c
+index 4b0820c8a7..81697d9bf9 100644
+--- a/pve-backup.c
++++ b/pve-backup.c
+@@ -374,6 +374,15 @@ static void pvebackup_complete_cb(void *opaque, int ret)
+         di->fleecing.snapshot_access = NULL;
+     }
+     if (di->fleecing.cbw) {
++        /*
++         * With fleecing, failure for cbw does not fail the guest write, but only sets the snapshot
++         * error, making further requests to the snapshot fail with EACCES, which then also fail the
++         * job. But that code is not the root cause and just confusing, so update it.
++         */
++        int snapshot_error = bdrv_cbw_snapshot_error(di->fleecing.cbw);
++        if (di->completed_ret == -EACCES && snapshot_error) {
++            di->completed_ret = snapshot_error;
++        }
+         bdrv_cbw_drop(di->fleecing.cbw);
+         di->fleecing.cbw = NULL;
+     }

debian/patches/pve/0045-savevm-async-don-t-hold-BQL-during-setup.patch

@@ -1,29 +0,0 @@
-From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
-From: Fiona Ebner <f.ebner@proxmox.com>
-Date: Fri, 5 May 2023 15:30:16 +0200
-Subject: [PATCH] savevm-async: don't hold BQL during setup
-
-See commit "migration: for snapshots, hold the BQL during setup
-callbacks" for why. This is separate, because a version of that one
-will hopefully land upstream.
-
-Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
----
- migration/savevm-async.c | 2 --
- 1 file changed, 2 deletions(-)
-
-diff --git a/migration/savevm-async.c b/migration/savevm-async.c
-index ea3b2f36a6..dd7744ab66 100644
---- a/migration/savevm-async.c
-+++ b/migration/savevm-async.c
-@@ -403,10 +403,8 @@ void qmp_savevm_start(const char *statefile, Error **errp)
-     snap_state.state = SAVE_STATE_ACTIVE;
-     snap_state.finalize_bh = qemu_bh_new(process_savevm_finalize, &snap_state);
-     snap_state.co = qemu_coroutine_create(&process_savevm_co, NULL);
--    qemu_mutex_unlock_iothread();
-     qemu_savevm_state_header(snap_state.file);
-     qemu_savevm_state_setup(snap_state.file);
--    qemu_mutex_lock_iothread();
- 
-     /* Async processing from here on out happens in iohandler context, so let
-      * the target bdrv have its home there.

debian/patches/pve/0046-PVE-backup-fixup-error-handling-for-fleecing.patch

@@ -0,0 +1,103 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Thu, 7 Nov 2024 17:51:14 +0100
+Subject: [PATCH] PVE backup: fixup error handling for fleecing
+
+The drained section needs to be terminated before breaking out of the
+loop in the error scenarios. Otherwise, guest IO on the drive would
+become stuck.
+
+If the job is created successfully, then the job completion callback
+will clean up the snapshot access block nodes. In case failure
+happened before the job is created, there was no cleanup for the
+snapshot access block nodes yet. Add it.
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ pve-backup.c | 38 +++++++++++++++++++++++++-------------
+ 1 file changed, 25 insertions(+), 13 deletions(-)
+
+diff --git a/pve-backup.c b/pve-backup.c
+index 81697d9bf9..320c660589 100644
+--- a/pve-backup.c
++++ b/pve-backup.c
+@@ -357,22 +357,23 @@ static void coroutine_fn pvebackup_co_complete_stream(void *opaque)
+     qemu_co_mutex_unlock(&backup_state.backup_mutex);
+ }
+ 
+-static void pvebackup_complete_cb(void *opaque, int ret)
++static void cleanup_snapshot_access(PVEBackupDevInfo *di)
+ {
+-    PVEBackupDevInfo *di = opaque;
+-    di->completed_ret = ret;
+-
+-    /*
+-     * Handle block-graph specific cleanup (for fleecing) outside of the coroutine, because the work
+-     * won't be done as a coroutine anyways:
+-     * - For snapshot_access, allows doing bdrv_unref() directly. Doing it via bdrv_co_unref() would
+-     *   just spawn a BH calling bdrv_unref().
+-     * - For cbw, draining would need to spawn a BH.
+-     */
+     if (di->fleecing.snapshot_access) {
+         bdrv_unref(di->fleecing.snapshot_access);
+         di->fleecing.snapshot_access = NULL;
+     }
++    if (di->fleecing.cbw) {
++        bdrv_cbw_drop(di->fleecing.cbw);
++        di->fleecing.cbw = NULL;
++    }
++}
++
++static void pvebackup_complete_cb(void *opaque, int ret)
++{
++    PVEBackupDevInfo *di = opaque;
++    di->completed_ret = ret;
++
+     if (di->fleecing.cbw) {
+         /*
+          * With fleecing, failure for cbw does not fail the guest write, but only sets the snapshot
+@@ -383,10 +384,17 @@ static void pvebackup_complete_cb(void *opaque, int ret)
+         if (di->completed_ret == -EACCES && snapshot_error) {
+             di->completed_ret = snapshot_error;
+         }
+-        bdrv_cbw_drop(di->fleecing.cbw);
+-        di->fleecing.cbw = NULL;
+     }
+ 
++    /*
++     * Handle block-graph specific cleanup (for fleecing) outside of the coroutine, because the work
++     * won't be done as a coroutine anyways:
++     * - For snapshot_access, allows doing bdrv_unref() directly. Doing it via bdrv_co_unref() would
++     *   just spawn a BH calling bdrv_unref().
++     * - For cbw, draining would need to spawn a BH.
++     */
++    cleanup_snapshot_access(di);
++
+     /*
+      * Needs to happen outside of coroutine, because it takes the graph write lock.
+      */
+@@ -587,6 +595,7 @@ static void create_backup_jobs_bh(void *opaque) {
+             if (!di->fleecing.cbw) {
+                 error_setg(errp, "appending cbw node for fleecing failed: %s",
+                            local_err ? error_get_pretty(local_err) : "unknown error");
++                bdrv_drained_end(di->bs);
+                 break;
+             }
+ 
+@@ -599,6 +608,8 @@ static void create_backup_jobs_bh(void *opaque) {
+             if (!di->fleecing.snapshot_access) {
+                 error_setg(errp, "setting up snapshot access for fleecing failed: %s",
+                            local_err ? error_get_pretty(local_err) : "unknown error");
++                cleanup_snapshot_access(di);
++                bdrv_drained_end(di->bs);
+                 break;
+             }
+             source_bs = di->fleecing.snapshot_access;
+@@ -637,6 +648,7 @@ static void create_backup_jobs_bh(void *opaque) {
+         }
+ 
+         if (!job || local_err) {
++            cleanup_snapshot_access(di);
+             error_setg(errp, "backup_job_create failed: %s",
+                        local_err ? error_get_pretty(local_err) : "null");
+             break;

debian/patches/pve/0047-PVE-backup-factor-out-setting-up-snapshot-access-for.patch

@@ -0,0 +1,135 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Thu, 7 Nov 2024 17:51:15 +0100
+Subject: [PATCH] PVE backup: factor out setting up snapshot access for
+ fleecing
+
+Avoids some line bloat in the create_backup_jobs_bh() function and is
+in preparation for setting up the snapshot access independently of
+fleecing, in particular that will be useful for providing access to
+the snapshot via NBD.
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ pve-backup.c | 95 ++++++++++++++++++++++++++++++++--------------------
+ 1 file changed, 58 insertions(+), 37 deletions(-)
+
+diff --git a/pve-backup.c b/pve-backup.c
+index 320c660589..d8d0c04b0f 100644
+--- a/pve-backup.c
++++ b/pve-backup.c
+@@ -525,6 +525,62 @@ static int coroutine_fn pvebackup_co_add_config(
+     goto out;
+ }
+ 
++/*
++ * Setup a snapshot-access block node for a device with associated fleecing image.
++ */
++static int setup_snapshot_access(PVEBackupDevInfo *di, Error **errp)
++{
++    Error *local_err = NULL;
++
++    if (!di->fleecing.bs) {
++        error_setg(errp, "no associated fleecing image");
++        return -1;
++    }
++
++    QDict *cbw_opts = qdict_new();
++    qdict_put_str(cbw_opts, "driver", "copy-before-write");
++    qdict_put_str(cbw_opts, "file", bdrv_get_node_name(di->bs));
++    qdict_put_str(cbw_opts, "target", bdrv_get_node_name(di->fleecing.bs));
++
++    if (di->bitmap) {
++        /*
++         * Only guest writes to parts relevant for the backup need to be intercepted with
++         * old data being copied to the fleecing image.
++         */
++        qdict_put_str(cbw_opts, "bitmap.node", bdrv_get_node_name(di->bs));
++        qdict_put_str(cbw_opts, "bitmap.name", bdrv_dirty_bitmap_name(di->bitmap));
++    }
++    /*
++     * Fleecing storage is supposed to be fast and it's better to break backup than guest
++     * writes. Certain guest drivers like VirtIO-win have 60 seconds timeout by default, so
++     * abort a bit before that.
++     */
++    qdict_put_str(cbw_opts, "on-cbw-error", "break-snapshot");
++    qdict_put_int(cbw_opts, "cbw-timeout", 45);
++
++    di->fleecing.cbw = bdrv_insert_node(di->bs, cbw_opts, BDRV_O_RDWR, &local_err);
++
++    if (!di->fleecing.cbw) {
++        error_setg(errp, "appending cbw node for fleecing failed: %s",
++                   local_err ? error_get_pretty(local_err) : "unknown error");
++        return -1;
++    }
++
++    QDict *snapshot_access_opts = qdict_new();
++    qdict_put_str(snapshot_access_opts, "driver", "snapshot-access");
++    qdict_put_str(snapshot_access_opts, "file", bdrv_get_node_name(di->fleecing.cbw));
++
++    di->fleecing.snapshot_access =
++        bdrv_open(NULL, NULL, snapshot_access_opts, BDRV_O_RDWR | BDRV_O_UNMAP, &local_err);
++    if (!di->fleecing.snapshot_access) {
++        error_setg(errp, "setting up snapshot access for fleecing failed: %s",
++                   local_err ? error_get_pretty(local_err) : "unknown error");
++        return -1;
++    }
++
++    return 0;
++}
++
+ /*
+  * backup_job_create can *not* be run from a coroutine, so this can't either.
+  * The caller is responsible that backup_mutex is held nonetheless.
+@@ -569,49 +625,14 @@ static void create_backup_jobs_bh(void *opaque) {
+         const char *job_id = bdrv_get_device_name(di->bs);
+         bdrv_graph_co_rdunlock();
+         if (di->fleecing.bs) {
+-            QDict *cbw_opts = qdict_new();
+-            qdict_put_str(cbw_opts, "driver", "copy-before-write");
+-            qdict_put_str(cbw_opts, "file", bdrv_get_node_name(di->bs));
+-            qdict_put_str(cbw_opts, "target", bdrv_get_node_name(di->fleecing.bs));
+-
+-            if (di->bitmap) {
+-                /*
+-                 * Only guest writes to parts relevant for the backup need to be intercepted with
+-                 * old data being copied to the fleecing image.
+-                 */
+-                qdict_put_str(cbw_opts, "bitmap.node", bdrv_get_node_name(di->bs));
+-                qdict_put_str(cbw_opts, "bitmap.name", bdrv_dirty_bitmap_name(di->bitmap));
+-            }
+-            /*
+-             * Fleecing storage is supposed to be fast and it's better to break backup than guest
+-             * writes. Certain guest drivers like VirtIO-win have 60 seconds timeout by default, so
+-             * abort a bit before that.
+-             */
+-            qdict_put_str(cbw_opts, "on-cbw-error", "break-snapshot");
+-            qdict_put_int(cbw_opts, "cbw-timeout", 45);
+-
+-            di->fleecing.cbw = bdrv_insert_node(di->bs, cbw_opts, BDRV_O_RDWR, &local_err);
+-
+-            if (!di->fleecing.cbw) {
+-                error_setg(errp, "appending cbw node for fleecing failed: %s",
+-                           local_err ? error_get_pretty(local_err) : "unknown error");
+-                bdrv_drained_end(di->bs);
+-                break;
+-            }
+-
+-            QDict *snapshot_access_opts = qdict_new();
+-            qdict_put_str(snapshot_access_opts, "driver", "snapshot-access");
+-            qdict_put_str(snapshot_access_opts, "file", bdrv_get_node_name(di->fleecing.cbw));
+-
+-            di->fleecing.snapshot_access =
+-                bdrv_open(NULL, NULL, snapshot_access_opts, BDRV_O_RDWR | BDRV_O_UNMAP, &local_err);
+-            if (!di->fleecing.snapshot_access) {
++            if (setup_snapshot_access(di, &local_err) < 0) {
+                 error_setg(errp, "setting up snapshot access for fleecing failed: %s",
+                            local_err ? error_get_pretty(local_err) : "unknown error");
+                 cleanup_snapshot_access(di);
+                 bdrv_drained_end(di->bs);
+                 break;
+             }
++
+             source_bs = di->fleecing.snapshot_access;
+             discard_source = true;
+ 

debian/patches/pve/0048-PVE-backup-save-device-name-in-device-info-structure.patch

@@ -0,0 +1,135 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Thu, 7 Nov 2024 17:51:16 +0100
+Subject: [PATCH] PVE backup: save device name in device info structure
+
+The device name needs to be queried while holding the graph read lock
+and since it doesn't change during the whole operation, just get it
+once during setup and avoid the need to query it again in different
+places.
+
+Also in preparation to use it more often in error messages and for the
+upcoming external backup access API.
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ pve-backup.c | 29 +++++++++++++++--------------
+ 1 file changed, 15 insertions(+), 14 deletions(-)
+
+diff --git a/pve-backup.c b/pve-backup.c
+index d8d0c04b0f..e2110ce0db 100644
+--- a/pve-backup.c
++++ b/pve-backup.c
+@@ -94,6 +94,7 @@ typedef struct PVEBackupDevInfo {
+     size_t size;
+     uint64_t block_size;
+     uint8_t dev_id;
++    char* device_name;
+     int completed_ret; // INT_MAX if not completed
+     BdrvDirtyBitmap *bitmap;
+     BlockDriverState *target;
+@@ -327,6 +328,8 @@ static void coroutine_fn pvebackup_co_complete_stream(void *opaque)
+     }
+ 
+     di->bs = NULL;
++    g_free(di->device_name);
++    di->device_name = NULL;
+ 
+     assert(di->target == NULL);
+ 
+@@ -621,9 +624,6 @@ static void create_backup_jobs_bh(void *opaque) {
+ 
+         BlockDriverState *source_bs = di->bs;
+         bool discard_source = false;
+-        bdrv_graph_co_rdlock();
+-        const char *job_id = bdrv_get_device_name(di->bs);
+-        bdrv_graph_co_rdunlock();
+         if (di->fleecing.bs) {
+             if (setup_snapshot_access(di, &local_err) < 0) {
+                 error_setg(errp, "setting up snapshot access for fleecing failed: %s",
+@@ -654,7 +654,7 @@ static void create_backup_jobs_bh(void *opaque) {
+         }
+ 
+         BlockJob *job = backup_job_create(
+-            job_id, source_bs, di->target, backup_state.speed, sync_mode, di->bitmap,
++            di->device_name, source_bs, di->target, backup_state.speed, sync_mode, di->bitmap,
+             bitmap_mode, false, discard_source, NULL, &perf, BLOCKDEV_ON_ERROR_REPORT,
+             BLOCKDEV_ON_ERROR_REPORT, JOB_DEFAULT, pvebackup_complete_cb, di, backup_state.txn,
+             &local_err);
+@@ -751,6 +751,7 @@ static GList coroutine_fn GRAPH_RDLOCK *get_device_info(
+             }
+             PVEBackupDevInfo *di = g_new0(PVEBackupDevInfo, 1);
+             di->bs = bs;
++            di->device_name = g_strdup(bdrv_get_device_name(bs));
+ 
+             if (fleecing && device_uses_fleecing(*d)) {
+                 g_autofree gchar *fleecing_devid = g_strconcat(*d, "-fleecing", NULL);
+@@ -789,6 +790,7 @@ static GList coroutine_fn GRAPH_RDLOCK *get_device_info(
+ 
+             PVEBackupDevInfo *di = g_new0(PVEBackupDevInfo, 1);
+             di->bs = bs;
++            di->device_name = g_strdup(bdrv_get_device_name(bs));
+             di_list = g_list_append(di_list, di);
+         }
+     }
+@@ -956,9 +958,6 @@ UuidInfo coroutine_fn *qmp_backup(
+ 
+             di->block_size = dump_cb_block_size;
+ 
+-            bdrv_graph_co_rdlock();
+-            const char *devname = bdrv_get_device_name(di->bs);
+-            bdrv_graph_co_rdunlock();
+             PBSBitmapAction action = PBS_BITMAP_ACTION_NOT_USED;
+             size_t dirty = di->size;
+ 
+@@ -973,7 +972,8 @@ UuidInfo coroutine_fn *qmp_backup(
+                     }
+                     action = PBS_BITMAP_ACTION_NEW;
+                 } else {
+-                    expect_only_dirty = proxmox_backup_check_incremental(pbs, devname, di->size) != 0;
++                    expect_only_dirty =
++                        proxmox_backup_check_incremental(pbs, di->device_name, di->size) != 0;
+                 }
+ 
+                 if (expect_only_dirty) {
+@@ -997,7 +997,8 @@ UuidInfo coroutine_fn *qmp_backup(
+                 }
+             }
+ 
+-            int dev_id = proxmox_backup_co_register_image(pbs, devname, di->size, expect_only_dirty, errp);
++            int dev_id = proxmox_backup_co_register_image(pbs, di->device_name, di->size,
++                                                          expect_only_dirty, errp);
+             if (dev_id < 0) {
+                 goto err_mutex;
+             }
+@@ -1009,7 +1010,7 @@ UuidInfo coroutine_fn *qmp_backup(
+             di->dev_id = dev_id;
+ 
+             PBSBitmapInfo *info = g_malloc(sizeof(*info));
+-            info->drive = g_strdup(devname);
++            info->drive = g_strdup(di->device_name);
+             info->action = action;
+             info->size = di->size;
+             info->dirty = dirty;
+@@ -1032,10 +1033,7 @@ UuidInfo coroutine_fn *qmp_backup(
+                 goto err_mutex;
+             }
+ 
+-            bdrv_graph_co_rdlock();
+-            const char *devname = bdrv_get_device_name(di->bs);
+-            bdrv_graph_co_rdunlock();
+-            di->dev_id = vma_writer_register_stream(vmaw, devname, di->size);
++            di->dev_id = vma_writer_register_stream(vmaw, di->device_name, di->size);
+             if (di->dev_id <= 0) {
+                 error_set(errp, ERROR_CLASS_GENERIC_ERROR,
+                           "register_stream failed");
+@@ -1146,6 +1144,9 @@ err:
+             bdrv_co_unref(di->target);
+         }
+ 
++        g_free(di->device_name);
++        di->device_name = NULL;
++
+         g_free(di);
+     }
+     g_list_free(di_list);

debian/patches/pve/0049-PVE-backup-include-device-name-in-error-when-setting.patch

@@ -0,0 +1,25 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Thu, 7 Nov 2024 17:51:17 +0100
+Subject: [PATCH] PVE backup: include device name in error when setting up
+ snapshot access fails
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ pve-backup.c | 3 ++-
+ 1 file changed, 2 insertions(+), 1 deletion(-)
+
+diff --git a/pve-backup.c b/pve-backup.c
+index e2110ce0db..32352fb5ec 100644
+--- a/pve-backup.c
++++ b/pve-backup.c
+@@ -626,7 +626,8 @@ static void create_backup_jobs_bh(void *opaque) {
+         bool discard_source = false;
+         if (di->fleecing.bs) {
+             if (setup_snapshot_access(di, &local_err) < 0) {
+-                error_setg(errp, "setting up snapshot access for fleecing failed: %s",
++                error_setg(errp, "%s - setting up snapshot access for fleecing failed: %s",
++                           di->device_name,
+                            local_err ? error_get_pretty(local_err) : "unknown error");
+                 cleanup_snapshot_access(di);
+                 bdrv_drained_end(di->bs);

debian/patches/pve/0050-adapt-machine-version-deprecation-for-Proxmox-VE.patch

@@ -0,0 +1,137 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Fri, 3 Jan 2025 14:03:12 +0100
+Subject: [PATCH] adapt machine version deprecation for Proxmox VE
+
+In commit a35f8577a0 ("include/hw: add macros for deprecation &
+removal of versioned machines"), a new machine version deprecation and
+removal policy was introduced. After only 3 years a machine version
+will be deprecated while being removed after 6 years.
+
+The deprecation is a bit early considering major PVE releases are
+approximately every 2 years. This means that a deprecation warning can
+already happen for a machine version that was introduced during the
+previous major release. This would scare users for no good reason, so
+avoid deprecating machine versions in PVE too early and define a
+baseline of machine versions that will be supported throughout a
+single major PVE release.
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ include/hw/boards.h | 78 +++++++++++++++++++++++++++++----------------
+ 1 file changed, 51 insertions(+), 27 deletions(-)
+
+diff --git a/include/hw/boards.h b/include/hw/boards.h
+index d1741ea121..3f9befda14 100644
+--- a/include/hw/boards.h
++++ b/include/hw/boards.h
+@@ -631,42 +631,66 @@ struct MachineState {
+ 
+ 
+ /*
+- * How many years/major releases for each phase
+- * of the life cycle. Assumes use of versioning
+- * scheme where major is bumped each year
++ * Baseline of machine versions that are still considered supported throughout
++ * current major Proxmox VE release. Machine versions older than this are
++ * considered to be deprecated in Proxmox VE.
++ *
++ * Machine versions older than 6 years are removed just like in upstream QEMU.
++ * (policy takes effect with QEMU 10.1). Assumes yearly major QEMU release.
++ *
++ * QEMU release cylce N.0 in ~April, N.1 in ~August, N.2 in ~December
++ * Debian/PVE release cylce ~every two years in summer
++ *
++ * PVE - last QEMU - machine versions dropped - baseline
++ *   8         9.2              2.3 and older        2.4
++ *   9        11.2              5.2 and older        6.0
++ *  10        13.2              7.2 and older        8.0
++ */
++#define MACHINE_VER_BASELINE_PVE_MAJOR 2
++#define MACHINE_VER_BASELINE_PVE_MINOR 4
++#define MACHINE_VER_DELETION_MAJOR (QEMU_VERSION_MAJOR - 6)
++#define MACHINE_VER_DELETION_MINOR QEMU_VERSION_MINOR
++
++/*
++ * Proxmox VE needs to support the baseline throughout a major PVE release. So
++ * a QEMU release where the baseline is already deleted cannot be used.
++ * Removal policy after 6 years takes effect with QEMU 10.1.
+  */
+-#define MACHINE_VER_DELETION_MAJOR 6
+-#define MACHINE_VER_DEPRECATION_MAJOR 3
++#if ((QEMU_VERSION_MAJOR > 10) || ((QEMU_VERSION_MAJOR == 10) && (QEMU_VERSION_MINOR >= 1)))
++#if ((MACHINE_VER_BASELINE_PVE_MAJOR < MACHINE_VER_DELETION_MAJOR) || \
++     ((MACHINE_VER_BASELINE_PVE_MAJOR == MACHINE_VER_DELETION_MAJOR) && \
++      (MACHINE_VER_BASELINE_PVE_MINOR < MACHINE_VER_DELETION_MINOR)))
++#error "Baseline machine version needed by Proxmox VE not supported anymore by this QEMU release"
++#endif
++#endif
+ 
+ /*
+  * Expands to a static string containing a deprecation
+  * message for a versioned machine type
+  */
+ #define MACHINE_VER_DEPRECATION_MSG \
+-    "machines more than " stringify(MACHINE_VER_DEPRECATION_MAJOR) \
+-    " years old are subject to deletion after " \
+-    stringify(MACHINE_VER_DELETION_MAJOR) " years"
+-
+-#define _MACHINE_VER_IS_EXPIRED_IMPL(cutoff, major, minor) \
+-    (((QEMU_VERSION_MAJOR - major) > cutoff) || \
+-     (((QEMU_VERSION_MAJOR - major) == cutoff) && \
+-      (QEMU_VERSION_MINOR - minor) >= 0))
+-
+-#define _MACHINE_VER_IS_EXPIRED2(cutoff, major, minor) \
+-    _MACHINE_VER_IS_EXPIRED_IMPL(cutoff, major, minor)
+-#define _MACHINE_VER_IS_EXPIRED3(cutoff, major, minor, micro) \
+-    _MACHINE_VER_IS_EXPIRED_IMPL(cutoff, major, minor)
+-#define _MACHINE_VER_IS_EXPIRED4(cutoff, major, minor, _unused, tag) \
+-    _MACHINE_VER_IS_EXPIRED_IMPL(cutoff, major, minor)
+-#define _MACHINE_VER_IS_EXPIRED5(cutoff, major, minor, micro, _unused, tag)   \
+-    _MACHINE_VER_IS_EXPIRED_IMPL(cutoff, major, minor)
+-
+-#define _MACHINE_IS_EXPIRED(cutoff, ...) \
++    "old machine version is subject to deletion during current major Proxmox VE release"
++
++#define _MACHINE_VER_IS_EXPIRED_IMPL(baseline_major, baseline_minor, major, minor) \
++    ((major < baseline_major) || \
++     ((major == baseline_major) && \
++      (minor < baseline_minor)))
++
++#define _MACHINE_VER_IS_EXPIRED2(baseline_major, baseline_minor, major, minor) \
++    _MACHINE_VER_IS_EXPIRED_IMPL(baseline_major, baseline_minor, major, minor)
++#define _MACHINE_VER_IS_EXPIRED3(baseline_major, baseline_minor, major, minor, micro) \
++    _MACHINE_VER_IS_EXPIRED_IMPL(baseline_major, baseline_minor, major, minor)
++#define _MACHINE_VER_IS_EXPIRED4(baseline_major, baseline_minor, major, minor, _unused, tag) \
++    _MACHINE_VER_IS_EXPIRED_IMPL(baseline_major, baseline_minor, major, minor)
++#define _MACHINE_VER_IS_EXPIRED5(baseline_major, baseline_minor, major, minor, micro, _unused, tag)   \
++    _MACHINE_VER_IS_EXPIRED_IMPL(baseline_major, baseline_minor, major, minor)
++
++#define _MACHINE_IS_EXPIRED(baseline_major, baseline_minor, ...) \
+     _MACHINE_VER_PICK(__VA_ARGS__, \
+                       _MACHINE_VER_IS_EXPIRED5, \
+                       _MACHINE_VER_IS_EXPIRED4, \
+                       _MACHINE_VER_IS_EXPIRED3, \
+-                      _MACHINE_VER_IS_EXPIRED2) (cutoff, __VA_ARGS__)
++                      _MACHINE_VER_IS_EXPIRED2) (baseline_major, baseline_minor, __VA_ARGS__)
+ 
+ /*
+  * Evaluates true when a machine type with (major, minor)
+@@ -675,7 +699,7 @@ struct MachineState {
+  * lifecycle rules
+  */
+ #define MACHINE_VER_IS_DEPRECATED(...) \
+-    _MACHINE_IS_EXPIRED(MACHINE_VER_DEPRECATION_MAJOR, __VA_ARGS__)
++    _MACHINE_IS_EXPIRED(MACHINE_VER_BASELINE_PVE_MAJOR, MACHINE_VER_BASELINE_PVE_MINOR, __VA_ARGS__)
+ 
+ /*
+  * Evaluates true when a machine type with (major, minor)
+@@ -684,7 +708,7 @@ struct MachineState {
+  * lifecycle rules
+  */
+ #define MACHINE_VER_SHOULD_DELETE(...) \
+-    _MACHINE_IS_EXPIRED(MACHINE_VER_DELETION_MAJOR, __VA_ARGS__)
++    _MACHINE_IS_EXPIRED(MACHINE_VER_DELETION_MAJOR, MACHINE_VER_DELETION_MINOR, __VA_ARGS__)
+ 
+ /*
+  * Sets the deprecation reason for a versioned machine based

debian/patches/pve/0051-Revert-hpet-avoid-timer-storms-on-periodic-timers.patch

@@ -0,0 +1,50 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Wed, 19 Mar 2025 17:31:05 +0100
+Subject: [PATCH] Revert "hpet: avoid timer storms on periodic timers"
+
+This reverts commit 7c912ffb59e8137091894d767433e65c3df8b0bf.
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ hw/timer/hpet.c | 13 ++-----------
+ 1 file changed, 2 insertions(+), 11 deletions(-)
+
+diff --git a/hw/timer/hpet.c b/hw/timer/hpet.c
+index 5399f1b2a3..8ccc421cbb 100644
+--- a/hw/timer/hpet.c
++++ b/hw/timer/hpet.c
+@@ -59,7 +59,6 @@ typedef struct HPETTimer {  /* timers */
+     uint8_t wrap_flag;      /* timer pop will indicate wrap for one-shot 32-bit
+                              * mode. Next pop will be actual timer expiration.
+                              */
+-    uint64_t last;          /* last value armed, to avoid timer storms */
+ } HPETTimer;
+ 
+ struct HPETState {
+@@ -267,7 +266,6 @@ static int hpet_post_load(void *opaque, int version_id)
+     for (i = 0; i < s->num_timers; i++) {
+         HPETTimer *t = &s->timer[i];
+         t->cmp64 = hpet_calculate_cmp64(t, s->hpet_counter, t->cmp);
+-        t->last = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) - NANOSECONDS_PER_SECOND;
+     }
+     /* Recalculate the offset between the main counter and guest time */
+     if (!s->hpet_offset_saved) {
+@@ -366,15 +364,8 @@ static const VMStateDescription vmstate_hpet = {
+ 
+ static void hpet_arm(HPETTimer *t, uint64_t tick)
+ {
+-    uint64_t ns = hpet_get_ns(t->state, tick);
+-
+-    /* Clamp period to reasonable min value (1 us) */
+-    if (timer_is_periodic(t) && ns - t->last < 1000) {
+-        ns = t->last + 1000;
+-    }
+-
+-    t->last = ns;
+-    timer_mod(t->qemu_timer, ns);
++    /* FIXME: Clamp period to reasonable min value? */
++    timer_mod(t->qemu_timer, hpet_get_ns(t->state, tick));
+ }
+ 
+ /*

debian/patches/pve/0052-Revert-hpet-store-full-64-bit-target-value-of-the-co.patch

@@ -0,0 +1,203 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Wed, 19 Mar 2025 17:31:08 +0100
+Subject: [PATCH] Revert "hpet: store full 64-bit target value of the counter"
+
+This reverts commit 242d665396407f83a6acbffc804882eeb21cfdad.
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ hw/timer/hpet.c | 111 +++++++++++++++++++++++++++---------------------
+ 1 file changed, 62 insertions(+), 49 deletions(-)
+
+diff --git a/hw/timer/hpet.c b/hw/timer/hpet.c
+index 8ccc421cbb..415a9433f1 100644
+--- a/hw/timer/hpet.c
++++ b/hw/timer/hpet.c
+@@ -54,7 +54,6 @@ typedef struct HPETTimer {  /* timers */
+     uint64_t cmp;           /* comparator */
+     uint64_t fsb;           /* FSB route */
+     /* Hidden register state */
+-    uint64_t cmp64;         /* comparator (extended to counter width) */
+     uint64_t period;        /* Last value written to comparator */
+     uint8_t wrap_flag;      /* timer pop will indicate wrap for one-shot 32-bit
+                              * mode. Next pop will be actual timer expiration.
+@@ -116,6 +115,11 @@ static uint32_t timer_enabled(HPETTimer *t)
+ }
+ 
+ static uint32_t hpet_time_after(uint64_t a, uint64_t b)
++{
++    return ((int32_t)(b - a) < 0);
++}
++
++static uint32_t hpet_time_after64(uint64_t a, uint64_t b)
+ {
+     return ((int64_t)(b - a) < 0);
+ }
+@@ -152,32 +156,27 @@ static uint64_t hpet_get_ticks(HPETState *s)
+     return ns_to_ticks(qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) + s->hpet_offset);
+ }
+ 
+-static uint64_t hpet_get_ns(HPETState *s, uint64_t tick)
+-{
+-    return ticks_to_ns(tick) - s->hpet_offset;
+-}
+-
+ /*
+- * calculate next value of the general counter that matches the
+- * target (either entirely, or the low 32-bit only depending on
+- * the timer mode).
++ * calculate diff between comparator value and current ticks
+  */
+-static uint64_t hpet_calculate_cmp64(HPETTimer *t, uint64_t cur_tick, uint64_t target)
++static inline uint64_t hpet_calculate_diff(HPETTimer *t, uint64_t current)
+ {
++
+     if (t->config & HPET_TN_32BIT) {
+-        uint64_t result = deposit64(cur_tick, 0, 32, target);
+-        if (result < cur_tick) {
+-            result += 0x100000000ULL;
+-        }
+-        return result;
++        uint32_t diff, cmp;
++
++        cmp = (uint32_t)t->cmp;
++        diff = cmp - (uint32_t)current;
++        diff = (int32_t)diff > 0 ? diff : (uint32_t)1;
++        return (uint64_t)diff;
+     } else {
+-        return target;
+-    }
+-}
++        uint64_t diff, cmp;
+ 
+-static uint64_t hpet_next_wrap(uint64_t cur_tick)
+-{
+-    return (cur_tick | 0xffffffffU) + 1;
++        cmp = t->cmp;
++        diff = cmp - current;
++        diff = (int64_t)diff > 0 ? diff : (uint64_t)1;
++        return diff;
++    }
+ }
+ 
+ static void update_irq(struct HPETTimer *timer, int set)
+@@ -261,12 +260,7 @@ static bool hpet_validate_num_timers(void *opaque, int version_id)
+ static int hpet_post_load(void *opaque, int version_id)
+ {
+     HPETState *s = opaque;
+-    int i;
+ 
+-    for (i = 0; i < s->num_timers; i++) {
+-        HPETTimer *t = &s->timer[i];
+-        t->cmp64 = hpet_calculate_cmp64(t, s->hpet_counter, t->cmp);
+-    }
+     /* Recalculate the offset between the main counter and guest time */
+     if (!s->hpet_offset_saved) {
+         s->hpet_offset = ticks_to_ns(s->hpet_counter)
+@@ -362,10 +356,14 @@ static const VMStateDescription vmstate_hpet = {
+     }
+ };
+ 
+-static void hpet_arm(HPETTimer *t, uint64_t tick)
++static void hpet_arm(HPETTimer *t, uint64_t ticks)
+ {
+-    /* FIXME: Clamp period to reasonable min value? */
+-    timer_mod(t->qemu_timer, hpet_get_ns(t->state, tick));
++    if (ticks < ns_to_ticks(INT64_MAX / 2)) {
++        timer_mod(t->qemu_timer,
++                  qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) + ticks_to_ns(ticks));
++    } else {
++        timer_del(t->qemu_timer);
++    }
+ }
+ 
+ /*
+@@ -374,44 +372,54 @@ static void hpet_arm(HPETTimer *t, uint64_t tick)
+ static void hpet_timer(void *opaque)
+ {
+     HPETTimer *t = opaque;
++    uint64_t diff;
++
+     uint64_t period = t->period;
+     uint64_t cur_tick = hpet_get_ticks(t->state);
+ 
+     if (timer_is_periodic(t) && period != 0) {
+-        while (hpet_time_after(cur_tick, t->cmp64)) {
+-            t->cmp64 += period;
+-        }
+         if (t->config & HPET_TN_32BIT) {
+-            t->cmp = (uint32_t)t->cmp64;
++            while (hpet_time_after(cur_tick, t->cmp)) {
++                t->cmp = (uint32_t)(t->cmp + t->period);
++            }
+         } else {
+-            t->cmp = t->cmp64;
++            while (hpet_time_after64(cur_tick, t->cmp)) {
++                t->cmp += period;
++            }
++        }
++        diff = hpet_calculate_diff(t, cur_tick);
++        hpet_arm(t, diff);
++    } else if (t->config & HPET_TN_32BIT && !timer_is_periodic(t)) {
++        if (t->wrap_flag) {
++            diff = hpet_calculate_diff(t, cur_tick);
++            hpet_arm(t, diff);
++            t->wrap_flag = 0;
+         }
+-        hpet_arm(t, t->cmp64);
+-    } else if (t->wrap_flag) {
+-        t->wrap_flag = 0;
+-        hpet_arm(t, t->cmp64);
+     }
+     update_irq(t, 1);
+ }
+ 
+ static void hpet_set_timer(HPETTimer *t)
+ {
++    uint64_t diff;
++    uint32_t wrap_diff;  /* how many ticks until we wrap? */
+     uint64_t cur_tick = hpet_get_ticks(t->state);
+ 
++    /* whenever new timer is being set up, make sure wrap_flag is 0 */
+     t->wrap_flag = 0;
+-    t->cmp64 = hpet_calculate_cmp64(t, cur_tick, t->cmp);
+-    if (t->config & HPET_TN_32BIT) {
+-
+-        /* hpet spec says in one-shot 32-bit mode, generate an interrupt when
+-         * counter wraps in addition to an interrupt with comparator match.
+-         */
+-        if (!timer_is_periodic(t) && t->cmp64 > hpet_next_wrap(cur_tick)) {
++    diff = hpet_calculate_diff(t, cur_tick);
++
++    /* hpet spec says in one-shot 32-bit mode, generate an interrupt when
++     * counter wraps in addition to an interrupt with comparator match.
++     */
++    if (t->config & HPET_TN_32BIT && !timer_is_periodic(t)) {
++        wrap_diff = 0xffffffff - (uint32_t)cur_tick;
++        if (wrap_diff < (uint32_t)diff) {
++            diff = wrap_diff;
+             t->wrap_flag = 1;
+-            hpet_arm(t, hpet_next_wrap(cur_tick));
+-            return;
+         }
+     }
+-    hpet_arm(t, t->cmp64);
++    hpet_arm(t, diff);
+ }
+ 
+ static void hpet_del_timer(HPETTimer *t)
+@@ -542,7 +550,12 @@ static void hpet_ram_write(void *opaque, hwaddr addr,
+                 timer->cmp = deposit64(timer->cmp, shift, len, value);
+             }
+             if (timer_is_periodic(timer)) {
+-                timer->period = deposit64(timer->period, shift, len, value);
++                /*
++                 * FIXME: Clamp period to reasonable min value?
++                 * Clamp period to reasonable max value
++                 */
++                new_val = deposit64(timer->period, shift, len, value);
++                timer->period = MIN(new_val, (timer->config & HPET_TN_32BIT ? ~0u : ~0ull) >> 1);
+             }
+             timer->config &= ~HPET_TN_SETVAL;
+             if (hpet_enabled(s)) {

debian/patches/pve/0053-Revert-hpet-accept-64-bit-reads-and-writes.patch

@@ -0,0 +1,281 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Wed, 19 Mar 2025 17:31:09 +0100
+Subject: [PATCH] Revert "hpet: accept 64-bit reads and writes"
+
+This reverts commit c2366567378dd8fb89329816003801f54e30e6f3.
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ hw/timer/hpet.c       | 137 +++++++++++++++++++++++++++++-------------
+ hw/timer/trace-events |   3 +-
+ 2 files changed, 96 insertions(+), 44 deletions(-)
+
+diff --git a/hw/timer/hpet.c b/hw/timer/hpet.c
+index 415a9433f1..e1ac877759 100644
+--- a/hw/timer/hpet.c
++++ b/hw/timer/hpet.c
+@@ -437,7 +437,6 @@ static uint64_t hpet_ram_read(void *opaque, hwaddr addr,
+                               unsigned size)
+ {
+     HPETState *s = opaque;
+-    int shift = (addr & 4) * 8;
+     uint64_t cur_tick;
+ 
+     trace_hpet_ram_read(addr);
+@@ -452,33 +451,52 @@ static uint64_t hpet_ram_read(void *opaque, hwaddr addr,
+             return 0;
+         }
+ 
+-        switch (addr & 0x18) {
+-        case HPET_TN_CFG: // including interrupt capabilities
+-            return timer->config >> shift;
++        switch ((addr - 0x100) % 0x20) {
++        case HPET_TN_CFG:
++            return timer->config;
++        case HPET_TN_CFG + 4: // Interrupt capabilities
++            return timer->config >> 32;
+         case HPET_TN_CMP: // comparator register
+-            return timer->cmp >> shift;
++            return timer->cmp;
++        case HPET_TN_CMP + 4:
++            return timer->cmp >> 32;
+         case HPET_TN_ROUTE:
+-            return timer->fsb >> shift;
++            return timer->fsb;
++        case HPET_TN_ROUTE + 4:
++            return timer->fsb >> 32;
+         default:
+             trace_hpet_ram_read_invalid();
+             break;
+         }
+     } else {
+-        switch (addr & ~4) {
+-        case HPET_ID: // including HPET_PERIOD
+-            return s->capability >> shift;
++        switch (addr) {
++        case HPET_ID:
++            return s->capability;
++        case HPET_PERIOD:
++            return s->capability >> 32;
+         case HPET_CFG:
+-            return s->config >> shift;
++            return s->config;
++        case HPET_CFG + 4:
++            trace_hpet_invalid_hpet_cfg(4);
++            return 0;
+         case HPET_COUNTER:
+             if (hpet_enabled(s)) {
+                 cur_tick = hpet_get_ticks(s);
+             } else {
+                 cur_tick = s->hpet_counter;
+             }
+-            trace_hpet_ram_read_reading_counter(addr & 4, cur_tick);
+-            return cur_tick >> shift;
++            trace_hpet_ram_read_reading_counter(0, cur_tick);
++            return cur_tick;
++        case HPET_COUNTER + 4:
++            if (hpet_enabled(s)) {
++                cur_tick = hpet_get_ticks(s);
++            } else {
++                cur_tick = s->hpet_counter;
++            }
++            trace_hpet_ram_read_reading_counter(4, cur_tick);
++            return cur_tick >> 32;
+         case HPET_STATUS:
+-            return s->isr >> shift;
++            return s->isr;
+         default:
+             trace_hpet_ram_read_invalid();
+             break;
+@@ -492,11 +510,11 @@ static void hpet_ram_write(void *opaque, hwaddr addr,
+ {
+     int i;
+     HPETState *s = opaque;
+-    int shift = (addr & 4) * 8;
+-    int len = MIN(size * 8, 64 - shift);
+     uint64_t old_val, new_val, cleared;
+ 
+     trace_hpet_ram_write(addr, value);
++    old_val = hpet_ram_read(opaque, addr, 4);
++    new_val = value;
+ 
+     /*address range of all TN regs*/
+     if (addr >= 0x100 && addr <= 0x3ff) {
+@@ -508,12 +526,9 @@ static void hpet_ram_write(void *opaque, hwaddr addr,
+             trace_hpet_timer_id_out_of_range(timer_id);
+             return;
+         }
+-        switch (addr & 0x18) {
++        switch ((addr - 0x100) % 0x20) {
+         case HPET_TN_CFG:
+-            trace_hpet_ram_write_tn_cfg(addr & 4);
+-            old_val = timer->config;
+-            new_val = deposit64(old_val, shift, len, value);
+-            new_val = hpet_fixup_reg(new_val, old_val, HPET_TN_CFG_WRITE_MASK);
++            trace_hpet_ram_write_tn_cfg();
+             if (deactivating_bit(old_val, new_val, HPET_TN_TYPE_LEVEL)) {
+                 /*
+                  * Do this before changing timer->config; otherwise, if
+@@ -521,7 +536,8 @@ static void hpet_ram_write(void *opaque, hwaddr addr,
+                  */
+                 update_irq(timer, 0);
+             }
+-            timer->config = new_val;
++            new_val = hpet_fixup_reg(new_val, old_val, HPET_TN_CFG_WRITE_MASK);
++            timer->config = (timer->config & 0xffffffff00000000ULL) | new_val;
+             if (activating_bit(old_val, new_val, HPET_TN_ENABLE)
+                 && (s->isr & (1 << timer_id))) {
+                 update_irq(timer, 1);
+@@ -534,28 +550,56 @@ static void hpet_ram_write(void *opaque, hwaddr addr,
+                 hpet_set_timer(timer);
+             }
+             break;
++        case HPET_TN_CFG + 4: // Interrupt capabilities
++            trace_hpet_ram_write_invalid_tn_cfg(4);
++            break;
+         case HPET_TN_CMP: // comparator register
++            trace_hpet_ram_write_tn_cmp(0);
+             if (timer->config & HPET_TN_32BIT) {
+-                /* High 32-bits are zero, leave them untouched.  */
+-                if (shift) {
+-                    trace_hpet_ram_write_invalid_tn_cmp();
+-                    break;
++                new_val = (uint32_t)new_val;
++            }
++            if (!timer_is_periodic(timer)
++                || (timer->config & HPET_TN_SETVAL)) {
++                timer->cmp = (timer->cmp & 0xffffffff00000000ULL) | new_val;
++            }
++            if (timer_is_periodic(timer)) {
++                /*
++                 * FIXME: Clamp period to reasonable min value?
++                 * Clamp period to reasonable max value
++                 */
++                if (timer->config & HPET_TN_32BIT) {
++                    new_val = MIN(new_val, ~0u >> 1);
+                 }
+-                len = 64;
+-                value = (uint32_t) value;
++                timer->period =
++                    (timer->period & 0xffffffff00000000ULL) | new_val;
++            }
++            /*
++             * FIXME: on a 64-bit write, HPET_TN_SETVAL should apply to the
++             * high bits part as well.
++             */
++            timer->config &= ~HPET_TN_SETVAL;
++            if (hpet_enabled(s)) {
++                hpet_set_timer(timer);
+             }
+-            trace_hpet_ram_write_tn_cmp(addr & 4);
++            break;
++        case HPET_TN_CMP + 4: // comparator register high order
++            if (timer->config & HPET_TN_32BIT) {
++                trace_hpet_ram_write_invalid_tn_cmp();
++                break;
++            }
++            trace_hpet_ram_write_tn_cmp(4);
+             if (!timer_is_periodic(timer)
+                 || (timer->config & HPET_TN_SETVAL)) {
+-                timer->cmp = deposit64(timer->cmp, shift, len, value);
++                timer->cmp = (timer->cmp & 0xffffffffULL) | new_val << 32;
+             }
+             if (timer_is_periodic(timer)) {
+                 /*
+                  * FIXME: Clamp period to reasonable min value?
+                  * Clamp period to reasonable max value
+                  */
+-                new_val = deposit64(timer->period, shift, len, value);
+-                timer->period = MIN(new_val, (timer->config & HPET_TN_32BIT ? ~0u : ~0ull) >> 1);
++                new_val = MIN(new_val, ~0u >> 1);
++                timer->period =
++                    (timer->period & 0xffffffffULL) | new_val << 32;
+             }
+             timer->config &= ~HPET_TN_SETVAL;
+             if (hpet_enabled(s)) {
+@@ -563,7 +607,10 @@ static void hpet_ram_write(void *opaque, hwaddr addr,
+             }
+             break;
+         case HPET_TN_ROUTE:
+-            timer->fsb = deposit64(timer->fsb, shift, len, value);
++            timer->fsb = (timer->fsb & 0xffffffff00000000ULL) | new_val;
++            break;
++        case HPET_TN_ROUTE + 4:
++            timer->fsb = (new_val << 32) | (timer->fsb & 0xffffffff);
+             break;
+         default:
+             trace_hpet_ram_write_invalid();
+@@ -571,14 +618,12 @@ static void hpet_ram_write(void *opaque, hwaddr addr,
+         }
+         return;
+     } else {
+-        switch (addr & ~4) {
++        switch (addr) {
+         case HPET_ID:
+             return;
+         case HPET_CFG:
+-            old_val = s->config;
+-            new_val = deposit64(old_val, shift, len, value);
+             new_val = hpet_fixup_reg(new_val, old_val, HPET_CFG_WRITE_MASK);
+-            s->config = new_val;
++            s->config = (s->config & 0xffffffff00000000ULL) | new_val;
+             if (activating_bit(old_val, new_val, HPET_CFG_ENABLE)) {
+                 /* Enable main counter and interrupt generation. */
+                 s->hpet_offset =
+@@ -608,8 +653,10 @@ static void hpet_ram_write(void *opaque, hwaddr addr,
+                 qemu_set_irq(s->irqs[RTC_ISA_IRQ], s->rtc_irq_level);
+             }
+             break;
++        case HPET_CFG + 4:
++            trace_hpet_invalid_hpet_cfg(4);
++            break;
+         case HPET_STATUS:
+-            new_val = value << shift;
+             cleared = new_val & s->isr;
+             for (i = 0; i < s->num_timers; i++) {
+                 if (cleared & (1 << i)) {
+@@ -621,7 +668,15 @@ static void hpet_ram_write(void *opaque, hwaddr addr,
+             if (hpet_enabled(s)) {
+                 trace_hpet_ram_write_counter_write_while_enabled();
+             }
+-            s->hpet_counter = deposit64(s->hpet_counter, shift, len, value);
++            s->hpet_counter =
++                (s->hpet_counter & 0xffffffff00000000ULL) | value;
++            trace_hpet_ram_write_counter_written(0, value, s->hpet_counter);
++            break;
++        case HPET_COUNTER + 4:
++            trace_hpet_ram_write_counter_write_while_enabled();
++            s->hpet_counter =
++                (s->hpet_counter & 0xffffffffULL) | (((uint64_t)value) << 32);
++            trace_hpet_ram_write_counter_written(4, value, s->hpet_counter);
+             break;
+         default:
+             trace_hpet_ram_write_invalid();
+@@ -635,11 +690,7 @@ static const MemoryRegionOps hpet_ram_ops = {
+     .write = hpet_ram_write,
+     .valid = {
+         .min_access_size = 4,
+-        .max_access_size = 8,
+-    },
+-    .impl = {
+-        .min_access_size = 4,
+-        .max_access_size = 8,
++        .max_access_size = 4,
+     },
+     .endianness = DEVICE_NATIVE_ENDIAN,
+ };
+diff --git a/hw/timer/trace-events b/hw/timer/trace-events
+index 5cfc369fba..219747df2f 100644
+--- a/hw/timer/trace-events
++++ b/hw/timer/trace-events
+@@ -114,7 +114,8 @@ hpet_ram_read_reading_counter(uint8_t reg_off, uint64_t cur_tick) "reading count
+ hpet_ram_read_invalid(void) "invalid hpet_ram_readl"
+ hpet_ram_write(uint64_t addr, uint64_t value) "enter hpet_ram_writel at 0x%" PRIx64 " = 0x%" PRIx64
+ hpet_ram_write_timer_id(uint64_t timer_id) "hpet_ram_writel timer_id = 0x%" PRIx64
+-hpet_ram_write_tn_cfg(uint8_t reg_off) "hpet_ram_writel HPET_TN_CFG + %" PRIu8
++hpet_ram_write_tn_cfg(void) "hpet_ram_writel HPET_TN_CFG"
++hpet_ram_write_invalid_tn_cfg(uint8_t reg_off) "invalid HPET_TN_CFG + %" PRIu8 " write"
+ hpet_ram_write_tn_cmp(uint8_t reg_off) "hpet_ram_writel HPET_TN_CMP + %" PRIu8
+ hpet_ram_write_invalid_tn_cmp(void) "invalid HPET_TN_CMP + 4 write"
+ hpet_ram_write_invalid(void) "invalid hpet_ram_writel"

debian/patches/pve/0054-Revert-hpet-place-read-only-bits-directly-in-new_val.patch

@@ -0,0 +1,64 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Wed, 19 Mar 2025 17:31:10 +0100
+Subject: [PATCH] Revert "hpet: place read-only bits directly in "new_val""
+
+This reverts commit ba88935b0fac2588b0a739f810b58dfabf7f92c8.
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ hw/timer/hpet.c | 15 ++++++++-------
+ 1 file changed, 8 insertions(+), 7 deletions(-)
+
+diff --git a/hw/timer/hpet.c b/hw/timer/hpet.c
+index e1ac877759..b12bbaf10d 100644
+--- a/hw/timer/hpet.c
++++ b/hw/timer/hpet.c
+@@ -510,7 +510,7 @@ static void hpet_ram_write(void *opaque, hwaddr addr,
+ {
+     int i;
+     HPETState *s = opaque;
+-    uint64_t old_val, new_val, cleared;
++    uint64_t old_val, new_val, val;
+ 
+     trace_hpet_ram_write(addr, value);
+     old_val = hpet_ram_read(opaque, addr, 4);
+@@ -536,12 +536,13 @@ static void hpet_ram_write(void *opaque, hwaddr addr,
+                  */
+                 update_irq(timer, 0);
+             }
+-            new_val = hpet_fixup_reg(new_val, old_val, HPET_TN_CFG_WRITE_MASK);
+-            timer->config = (timer->config & 0xffffffff00000000ULL) | new_val;
++            val = hpet_fixup_reg(new_val, old_val, HPET_TN_CFG_WRITE_MASK);
++            timer->config = (timer->config & 0xffffffff00000000ULL) | val;
+             if (activating_bit(old_val, new_val, HPET_TN_ENABLE)
+                 && (s->isr & (1 << timer_id))) {
+                 update_irq(timer, 1);
+             }
++
+             if (new_val & HPET_TN_32BIT) {
+                 timer->cmp = (uint32_t)timer->cmp;
+                 timer->period = (uint32_t)timer->period;
+@@ -622,8 +623,8 @@ static void hpet_ram_write(void *opaque, hwaddr addr,
+         case HPET_ID:
+             return;
+         case HPET_CFG:
+-            new_val = hpet_fixup_reg(new_val, old_val, HPET_CFG_WRITE_MASK);
+-            s->config = (s->config & 0xffffffff00000000ULL) | new_val;
++            val = hpet_fixup_reg(new_val, old_val, HPET_CFG_WRITE_MASK);
++            s->config = (s->config & 0xffffffff00000000ULL) | val;
+             if (activating_bit(old_val, new_val, HPET_CFG_ENABLE)) {
+                 /* Enable main counter and interrupt generation. */
+                 s->hpet_offset =
+@@ -657,9 +658,9 @@ static void hpet_ram_write(void *opaque, hwaddr addr,
+             trace_hpet_invalid_hpet_cfg(4);
+             break;
+         case HPET_STATUS:
+-            cleared = new_val & s->isr;
++            val = new_val & s->isr;
+             for (i = 0; i < s->num_timers; i++) {
+-                if (cleared & (1 << i)) {
++                if (val & (1 << i)) {
+                     update_irq(&s->timer[i], 0);
+                 }
+             }

debian/patches/pve/0055-Revert-hpet-remove-unnecessary-variable-index.patch

@@ -0,0 +1,68 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Wed, 19 Mar 2025 17:31:11 +0100
+Subject: [PATCH] Revert "hpet: remove unnecessary variable "index""
+
+This reverts commit 5895879aca252f4ebb2d1078eaf836c61ec54e9b.
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ hw/timer/hpet.c | 15 ++++++++-------
+ 1 file changed, 8 insertions(+), 7 deletions(-)
+
+diff --git a/hw/timer/hpet.c b/hw/timer/hpet.c
+index b12bbaf10d..6f83d88516 100644
+--- a/hw/timer/hpet.c
++++ b/hw/timer/hpet.c
+@@ -437,12 +437,12 @@ static uint64_t hpet_ram_read(void *opaque, hwaddr addr,
+                               unsigned size)
+ {
+     HPETState *s = opaque;
+-    uint64_t cur_tick;
++    uint64_t cur_tick, index;
+ 
+     trace_hpet_ram_read(addr);
+-
++    index = addr;
+     /*address range of all TN regs*/
+-    if (addr >= 0x100 && addr <= 0x3ff) {
++    if (index >= 0x100 && index <= 0x3ff) {
+         uint8_t timer_id = (addr - 0x100) / 0x20;
+         HPETTimer *timer = &s->timer[timer_id];
+ 
+@@ -469,7 +469,7 @@ static uint64_t hpet_ram_read(void *opaque, hwaddr addr,
+             break;
+         }
+     } else {
+-        switch (addr) {
++        switch (index) {
+         case HPET_ID:
+             return s->capability;
+         case HPET_PERIOD:
+@@ -510,14 +510,15 @@ static void hpet_ram_write(void *opaque, hwaddr addr,
+ {
+     int i;
+     HPETState *s = opaque;
+-    uint64_t old_val, new_val, val;
++    uint64_t old_val, new_val, val, index;
+ 
+     trace_hpet_ram_write(addr, value);
++    index = addr;
+     old_val = hpet_ram_read(opaque, addr, 4);
+     new_val = value;
+ 
+     /*address range of all TN regs*/
+-    if (addr >= 0x100 && addr <= 0x3ff) {
++    if (index >= 0x100 && index <= 0x3ff) {
+         uint8_t timer_id = (addr - 0x100) / 0x20;
+         HPETTimer *timer = &s->timer[timer_id];
+ 
+@@ -619,7 +620,7 @@ static void hpet_ram_write(void *opaque, hwaddr addr,
+         }
+         return;
+     } else {
+-        switch (addr) {
++        switch (index) {
+         case HPET_ID:
+             return;
+         case HPET_CFG:

debian/patches/pve/0056-Revert-hpet-ignore-high-bits-of-comparator-in-32-bit.patch

@@ -0,0 +1,40 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Wed, 19 Mar 2025 17:31:12 +0100
+Subject: [PATCH] Revert "hpet: ignore high bits of comparator in 32-bit mode"
+
+This reverts commit 9eb7fad3546a89ee7cf0e90f5b1daccf89725cea.
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ hw/timer/hpet.c       | 4 ----
+ hw/timer/trace-events | 1 -
+ 2 files changed, 5 deletions(-)
+
+diff --git a/hw/timer/hpet.c b/hw/timer/hpet.c
+index 6f83d88516..509986c0a9 100644
+--- a/hw/timer/hpet.c
++++ b/hw/timer/hpet.c
+@@ -585,10 +585,6 @@ static void hpet_ram_write(void *opaque, hwaddr addr,
+             }
+             break;
+         case HPET_TN_CMP + 4: // comparator register high order
+-            if (timer->config & HPET_TN_32BIT) {
+-                trace_hpet_ram_write_invalid_tn_cmp();
+-                break;
+-            }
+             trace_hpet_ram_write_tn_cmp(4);
+             if (!timer_is_periodic(timer)
+                 || (timer->config & HPET_TN_SETVAL)) {
+diff --git a/hw/timer/trace-events b/hw/timer/trace-events
+index 219747df2f..b86870fb22 100644
+--- a/hw/timer/trace-events
++++ b/hw/timer/trace-events
+@@ -117,7 +117,6 @@ hpet_ram_write_timer_id(uint64_t timer_id) "hpet_ram_writel timer_id = 0x%" PRIx
+ hpet_ram_write_tn_cfg(void) "hpet_ram_writel HPET_TN_CFG"
+ hpet_ram_write_invalid_tn_cfg(uint8_t reg_off) "invalid HPET_TN_CFG + %" PRIu8 " write"
+ hpet_ram_write_tn_cmp(uint8_t reg_off) "hpet_ram_writel HPET_TN_CMP + %" PRIu8
+-hpet_ram_write_invalid_tn_cmp(void) "invalid HPET_TN_CMP + 4 write"
+ hpet_ram_write_invalid(void) "invalid hpet_ram_writel"
+ hpet_ram_write_counter_write_while_enabled(void) "Writing counter while HPET enabled!"
+ hpet_ram_write_counter_written(uint8_t reg_off, uint64_t value, uint64_t counter) "HPET counter + %" PRIu8 "written. crt = 0x%" PRIx64 " -> 0x%" PRIx64

debian/patches/pve/0057-Revert-hpet-fix-and-cleanup-persistence-of-interrupt.patch

@@ -0,0 +1,120 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Wed, 19 Mar 2025 17:31:13 +0100
+Subject: [PATCH] Revert "hpet: fix and cleanup persistence of interrupt
+ status"
+
+This reverts commit f0ccf770789e48b7a73497b465fdc892d28c1339.
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ hw/timer/hpet.c | 60 ++++++++++++++++---------------------------------
+ 1 file changed, 19 insertions(+), 41 deletions(-)
+
+diff --git a/hw/timer/hpet.c b/hw/timer/hpet.c
+index 509986c0a9..402cc960f0 100644
+--- a/hw/timer/hpet.c
++++ b/hw/timer/hpet.c
+@@ -196,31 +196,21 @@ static void update_irq(struct HPETTimer *timer, int set)
+     }
+     s = timer->state;
+     mask = 1 << timer->tn;
+-
+-    if (set && (timer->config & HPET_TN_TYPE_LEVEL)) {
+-        /*
+-         * If HPET_TN_ENABLE bit is 0, "the timer will still operate and
+-         * generate appropriate status bits, but will not cause an interrupt"
+-         */
+-        s->isr |= mask;
+-    } else {
++    if (!set || !timer_enabled(timer) || !hpet_enabled(timer->state)) {
+         s->isr &= ~mask;
+-    }
+-
+-    if (set && timer_enabled(timer) && hpet_enabled(s)) {
+-        if (timer_fsb_route(timer)) {
+-            address_space_stl_le(&address_space_memory, timer->fsb >> 32,
+-                                 timer->fsb & 0xffffffff, MEMTXATTRS_UNSPECIFIED,
+-                                 NULL);
+-        } else if (timer->config & HPET_TN_TYPE_LEVEL) {
+-            qemu_irq_raise(s->irqs[route]);
+-        } else {
+-            qemu_irq_pulse(s->irqs[route]);
+-        }
+-    } else {
+         if (!timer_fsb_route(timer)) {
+             qemu_irq_lower(s->irqs[route]);
+         }
++    } else if (timer_fsb_route(timer)) {
++        address_space_stl_le(&address_space_memory, timer->fsb >> 32,
++                             timer->fsb & 0xffffffff, MEMTXATTRS_UNSPECIFIED,
++                             NULL);
++    } else if (timer->config & HPET_TN_TYPE_LEVEL) {
++        s->isr |= mask;
++        qemu_irq_raise(s->irqs[route]);
++    } else {
++        s->isr &= ~mask;
++        qemu_irq_pulse(s->irqs[route]);
+     }
+ }
+ 
+@@ -424,13 +414,8 @@ static void hpet_set_timer(HPETTimer *t)
+ 
+ static void hpet_del_timer(HPETTimer *t)
+ {
+-    HPETState *s = t->state;
+     timer_del(t->qemu_timer);
+-
+-    if (s->isr & (1 << t->tn)) {
+-        /* For level-triggered interrupt, this leaves ISR set but lowers irq.  */
+-        update_irq(t, 1);
+-    }
++    update_irq(t, 0);
+ }
+ 
+ static uint64_t hpet_ram_read(void *opaque, hwaddr addr,
+@@ -530,26 +515,20 @@ static void hpet_ram_write(void *opaque, hwaddr addr,
+         switch ((addr - 0x100) % 0x20) {
+         case HPET_TN_CFG:
+             trace_hpet_ram_write_tn_cfg();
+-            if (deactivating_bit(old_val, new_val, HPET_TN_TYPE_LEVEL)) {
+-                /*
+-                 * Do this before changing timer->config; otherwise, if
+-                 * HPET_TN_FSB is set, update_irq will not lower the qemu_irq.
+-                 */
++            if (activating_bit(old_val, new_val, HPET_TN_FSB_ENABLE)) {
+                 update_irq(timer, 0);
+             }
+             val = hpet_fixup_reg(new_val, old_val, HPET_TN_CFG_WRITE_MASK);
+             timer->config = (timer->config & 0xffffffff00000000ULL) | val;
+-            if (activating_bit(old_val, new_val, HPET_TN_ENABLE)
+-                && (s->isr & (1 << timer_id))) {
+-                update_irq(timer, 1);
+-            }
+-
+             if (new_val & HPET_TN_32BIT) {
+                 timer->cmp = (uint32_t)timer->cmp;
+                 timer->period = (uint32_t)timer->period;
+             }
+-            if (hpet_enabled(s)) {
++            if (activating_bit(old_val, new_val, HPET_TN_ENABLE) &&
++                hpet_enabled(s)) {
+                 hpet_set_timer(timer);
++            } else if (deactivating_bit(old_val, new_val, HPET_TN_ENABLE)) {
++                hpet_del_timer(timer);
+             }
+             break;
+         case HPET_TN_CFG + 4: // Interrupt capabilities
+@@ -627,10 +606,9 @@ static void hpet_ram_write(void *opaque, hwaddr addr,
+                 s->hpet_offset =
+                     ticks_to_ns(s->hpet_counter) - qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
+                 for (i = 0; i < s->num_timers; i++) {
+-                    if (timer_enabled(&s->timer[i]) && (s->isr & (1 << i))) {
+-                        update_irq(&s->timer[i], 1);
++                    if ((&s->timer[i])->cmp != ~0ULL) {
++                        hpet_set_timer(&s->timer[i]);
+                     }
+-                    hpet_set_timer(&s->timer[i]);
+                 }
+             } else if (deactivating_bit(old_val, new_val, HPET_CFG_ENABLE)) {
+                 /* Halt main counter and disable interrupt generation. */

debian/patches/pve/0058-savevm-async-improve-setting-state-of-snapshot-opera.patch

@@ -0,0 +1,81 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Mon, 31 Mar 2025 16:55:02 +0200
+Subject: [PATCH] savevm-async: improve setting state of snapshot operation in
+ savevm-end handler
+
+One of the callers of wait_for_close_co() already sets the state to
+SAVE_STATE_DONE before, but that is not fully correct, because at that
+moment, the operation is not fully done. In particular, if closing the
+target later fails, the state would even be set to SAVE_STATE_ERROR
+afterwards. DONE -> ERROR is not a valid state transition. Although,
+it should not matter in practice as long as the relevant QMP commands
+are sequential.
+
+The other caller does not set the state and so there seems to be a
+race that could lead to the state not getting set at all. This is
+because before this commit, the wait_for_close_co() function could
+return early when there is no target file anymore. This can only
+happen when canceling and needs to happen right around the time when
+the snapshot is already finishing and closing the target.
+
+Simply avoid the early return and always set the state within the
+wait_for_close_co() function rather than at the call site.
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ migration/savevm-async.c | 33 +++++++++++++++------------------
+ 1 file changed, 15 insertions(+), 18 deletions(-)
+
+diff --git a/migration/savevm-async.c b/migration/savevm-async.c
+index 1e79fce9ba..e63dc6d8a3 100644
+--- a/migration/savevm-async.c
++++ b/migration/savevm-async.c
+@@ -450,23 +450,22 @@ static void coroutine_fn wait_for_close_co(void *opaque)
+ {
+     int64_t timeout;
+ 
+-    if (!snap_state.target) {
+-        DPRINTF("savevm-end: no target file open\n");
+-        return;
+-    }
+-
+-    /* wait until cleanup is done before returning, this ensures that after this
+-     * call exits the statefile will be closed and can be removed immediately */
+-    DPRINTF("savevm-end: waiting for cleanup\n");
+-    timeout = 30L * 1000 * 1000 * 1000;
+-    qemu_co_sleep_ns_wakeable(&snap_state.target_close_wait,
+-                              QEMU_CLOCK_REALTIME, timeout);
+     if (snap_state.target) {
+-        save_snapshot_error("timeout waiting for target file close in "
+-                            "qmp_savevm_end");
+-        /* we cannot assume the snapshot finished in this case, so leave the
+-         * state alone - caller has to figure something out */
+-        return;
++        /* wait until cleanup is done before returning, this ensures that after this
++         * call exits the statefile will be closed and can be removed immediately */
++        DPRINTF("savevm-end: waiting for cleanup\n");
++        timeout = 30L * 1000 * 1000 * 1000;
++        qemu_co_sleep_ns_wakeable(&snap_state.target_close_wait,
++                                  QEMU_CLOCK_REALTIME, timeout);
++        if (snap_state.target) {
++            save_snapshot_error("timeout waiting for target file close in "
++                                "qmp_savevm_end");
++            /* we cannot assume the snapshot finished in this case, so leave the
++             * state alone - caller has to figure something out */
++            return;
++        }
++    } else {
++        DPRINTF("savevm-end: no target file open\n");
+     }
+ 
+     // File closed and no other error, so ensure next snapshot can be started.
+@@ -497,8 +496,6 @@ void qmp_savevm_end(Error **errp)
+         snap_state.saved_vm_running = false;
+     }
+ 
+-    snap_state.state = SAVE_STATE_DONE;
+-
+     qemu_coroutine_enter(wait_for_close);
+ }
+ 

debian/patches/pve/0059-savevm-async-rename-saved_vm_running-to-vm_needs_sta.patch

@@ -0,0 +1,71 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Mon, 31 Mar 2025 16:55:03 +0200
+Subject: [PATCH] savevm-async: rename saved_vm_running to vm_needs_start
+
+This is what the variable actually expresses. Otherwise, setting it
+to false after starting the VM doesn't make sense.
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+---
+ migration/savevm-async.c | 16 ++++++++--------
+ 1 file changed, 8 insertions(+), 8 deletions(-)
+
+diff --git a/migration/savevm-async.c b/migration/savevm-async.c
+index e63dc6d8a3..1e34c31e8b 100644
+--- a/migration/savevm-async.c
++++ b/migration/savevm-async.c
+@@ -51,7 +51,7 @@ static struct SnapshotState {
+     int state;
+     Error *error;
+     Error *blocker;
+-    int saved_vm_running;
++    int vm_needs_start;
+     QEMUFile *file;
+     int64_t total_time;
+     QEMUBH *finalize_bh;
+@@ -224,9 +224,9 @@ static void process_savevm_finalize(void *opaque)
+         save_snapshot_error("process_savevm_cleanup: invalid state: %d",
+                             snap_state.state);
+     }
+-    if (snap_state.saved_vm_running) {
++    if (snap_state.vm_needs_start) {
+         vm_start();
+-        snap_state.saved_vm_running = false;
++        snap_state.vm_needs_start = false;
+     }
+ 
+     DPRINTF("timing: process_savevm_finalize (full) took %ld ms\n",
+@@ -352,7 +352,7 @@ void qmp_savevm_start(const char *statefile, Error **errp)
+     }
+ 
+     /* initialize snapshot info */
+-    snap_state.saved_vm_running = runstate_is_running();
++    snap_state.vm_needs_start = runstate_is_running();
+     snap_state.bs_pos = 0;
+     snap_state.total_time = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
+     snap_state.blocker = NULL;
+@@ -440,9 +440,9 @@ restart:
+ 
+     save_snapshot_error("setup failed");
+ 
+-    if (snap_state.saved_vm_running) {
++    if (snap_state.vm_needs_start) {
+         vm_start();
+-        snap_state.saved_vm_running = false;
++        snap_state.vm_needs_start = false;
+     }
+ }
+ 
+@@ -491,9 +491,9 @@ void qmp_savevm_end(Error **errp)
+         return;
+     }
+ 
+-    if (snap_state.saved_vm_running) {
++    if (snap_state.vm_needs_start) {
+         vm_start();
+-        snap_state.saved_vm_running = false;
++        snap_state.vm_needs_start = false;
+     }
+ 
+     qemu_coroutine_enter(wait_for_close);

debian/patches/pve/0060-savevm-async-improve-runstate-preservation-cleanup-e.patch

@@ -0,0 +1,120 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Mon, 31 Mar 2025 16:55:04 +0200
+Subject: [PATCH] savevm-async: improve runstate preservation, cleanup error
+ handling
+
+Determine if VM needs to be started after finishing right before
+actually stopping the VM instead of at the beginning.
+
+In qmp_savevm_start(), the only path stopping the VM returns right
+aftwards, so there is no need for the vm_start() handling after
+errors.
+
+Lastly, improve the code style for checking whether migrate_init()
+failed by explicitly comparing against 0.
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+[WB: squashed error handling commits, rename goto branch instead of
+     inlining it]
+Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
+---
+ migration/savevm-async.c | 25 ++++++++++---------------
+ 1 file changed, 10 insertions(+), 15 deletions(-)
+
+diff --git a/migration/savevm-async.c b/migration/savevm-async.c
+index 1e34c31e8b..d8d2c80475 100644
+--- a/migration/savevm-async.c
++++ b/migration/savevm-async.c
+@@ -178,6 +178,7 @@ static void process_savevm_finalize(void *opaque)
+      */
+     blk_set_aio_context(snap_state.target, qemu_get_aio_context(), NULL);
+ 
++    snap_state.vm_needs_start = runstate_is_running();
+     ret = vm_stop_force_state(RUN_STATE_FINISH_MIGRATE);
+     if (ret < 0) {
+         save_snapshot_error("vm_stop_force_state error %d", ret);
+@@ -352,7 +353,6 @@ void qmp_savevm_start(const char *statefile, Error **errp)
+     }
+ 
+     /* initialize snapshot info */
+-    snap_state.vm_needs_start = runstate_is_running();
+     snap_state.bs_pos = 0;
+     snap_state.total_time = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
+     snap_state.blocker = NULL;
+@@ -364,13 +364,14 @@ void qmp_savevm_start(const char *statefile, Error **errp)
+     }
+ 
+     if (!statefile) {
++        snap_state.vm_needs_start = runstate_is_running();
+         vm_stop(RUN_STATE_SAVE_VM);
+         snap_state.state = SAVE_STATE_COMPLETED;
+         return;
+     }
+ 
+     if (qemu_savevm_state_blocked(errp)) {
+-        return;
++        goto fail;
+     }
+ 
+     /* Open the image */
+@@ -380,12 +381,12 @@ void qmp_savevm_start(const char *statefile, Error **errp)
+     snap_state.target = blk_new_open(statefile, NULL, options, bdrv_oflags, &local_err);
+     if (!snap_state.target) {
+         error_setg(errp, "failed to open '%s'", statefile);
+-        goto restart;
++        goto fail;
+     }
+     target_bs = blk_bs(snap_state.target);
+     if (!target_bs) {
+         error_setg(errp, "failed to open '%s' - no block driver state", statefile);
+-        goto restart;
++        goto fail;
+     }
+ 
+     QIOChannel *ioc = QIO_CHANNEL(qio_channel_savevm_async_new(snap_state.target,
+@@ -394,7 +395,7 @@ void qmp_savevm_start(const char *statefile, Error **errp)
+ 
+     if (!snap_state.file) {
+         error_setg(errp, "failed to open '%s'", statefile);
+-        goto restart;
++        goto fail;
+     }
+ 
+     /*
+@@ -402,8 +403,8 @@ void qmp_savevm_start(const char *statefile, Error **errp)
+      * State is cleared in process_savevm_co, but has to be initialized
+      * here (blocking main thread, from QMP) to avoid race conditions.
+      */
+-    if (migrate_init(ms, errp)) {
+-        return;
++    if (migrate_init(ms, errp) != 0) {
++        goto fail;
+     }
+     memset(&mig_stats, 0, sizeof(mig_stats));
+     ms->to_dst_file = snap_state.file;
+@@ -418,7 +419,7 @@ void qmp_savevm_start(const char *statefile, Error **errp)
+     if (ret != 0) {
+         error_setg_errno(errp, -ret, "savevm state setup failed: %s",
+                          local_err ? error_get_pretty(local_err) : "unknown error");
+-        return;
++        goto fail;
+     }
+ 
+     /* Async processing from here on out happens in iohandler context, so let
+@@ -436,14 +437,8 @@ void qmp_savevm_start(const char *statefile, Error **errp)
+ 
+     return;
+ 
+-restart:
+-
++fail:
+     save_snapshot_error("setup failed");
+-
+-    if (snap_state.vm_needs_start) {
+-        vm_start();
+-        snap_state.vm_needs_start = false;
+-    }
+ }
+ 
+ static void coroutine_fn wait_for_close_co(void *opaque)

debian/patches/pve/0061-savevm-async-use-dedicated-iothread-for-state-file.patch

@@ -0,0 +1,185 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Mon, 31 Mar 2025 16:55:06 +0200
+Subject: [PATCH] savevm-async: use dedicated iothread for state file
+
+Having the state file be in the iohandler context means that a
+blk_drain_all() call in the main thread or vCPU thread that happens
+while the snapshot is running will result in a deadlock.
+
+For example, the main thread might be stuck in:
+
+> 0  0x00007300ac9552d6 in __ppoll (fds=0x64bd5a411a50, nfds=2, timeout=<optimized out>, timeout@entry=0x0, sigmask=sigmask@entry=0x0) at ../sysdeps/unix/sysv/linux/ppoll.c:42
+> 1  0x000064bd51af3cad in ppoll (__ss=0x0, __timeout=0x0, __nfds=<optimized out>, __fds=<optimized out>) at /usr/include/x86_64-linux-gnu/bits/poll2.h:64
+> 2  0x000064bd51ad8799 in fdmon_poll_wait (ctx=0x64bd58d968a0, ready_list=0x7ffcfcc15558, timeout=-1) at ../util/fdmon-poll.c:79
+> 3  0x000064bd51ad7c3d in aio_poll (ctx=0x64bd58d968a0, blocking=blocking@entry=true) at ../util/aio-posix.c:671
+> 4  0x000064bd519a0b5d in bdrv_drain_all_begin () at ../block/io.c:531
+> 5  bdrv_drain_all_begin () at ../block/io.c:510
+> 6  0x000064bd519943c4 in blk_drain_all () at ../block/block-backend.c:2085
+> 7  0x000064bd5160fc5a in virtio_scsi_dataplane_stop (vdev=0x64bd5a215190) at ../hw/scsi/virtio-scsi-dataplane.c:213
+> 8  0x000064bd51664e90 in virtio_bus_stop_ioeventfd (bus=0x64bd5a215110) at ../hw/virtio/virtio-bus.c:259
+> 9  0x000064bd5166511b in virtio_bus_stop_ioeventfd (bus=<optimized out>) at ../hw/virtio/virtio-bus.c:251
+> 10 virtio_bus_reset (bus=<optimized out>) at ../hw/virtio/virtio-bus.c:107
+> 11 0x000064bd51667431 in virtio_pci_reset (qdev=<optimized out>) at ../hw/virtio/virtio-pci.c:2296
+...
+> 34 0x000064bd517aa951 in pc_machine_reset (machine=<optimized out>, type=<optimized out>) at ../hw/i386/pc.c:1722
+> 35 0x000064bd516aa4c4 in qemu_system_reset (reason=reason@entry=SHUTDOWN_CAUSE_GUEST_RESET) at ../system/runstate.c:525
+> 36 0x000064bd516aaeb9 in main_loop_should_exit (status=<synthetic pointer>) at ../system/runstate.c:801
+> 37 qemu_main_loop () at ../system/runstate.c:834
+
+which is in block/io.c:
+
+> /* Now poll the in-flight requests */
+> AIO_WAIT_WHILE_UNLOCKED(NULL, bdrv_drain_all_poll());
+
+The working theory is: The deadlock happens because the IO is issued
+from the process_savevm_co() coroutine, which doesn't get scheduled
+again to complete in-flight requests when the main thread is stuck
+there polling. The main thread itself is the one that would need to
+schedule it. In case of a vCPU triggering the VirtIO SCSI dataplane
+stop, which happens during (Linux) boot, the vCPU thread will hold the
+big QEMU lock (BQL) blocking the main thread from making progress
+scheduling the process_savevm_co() coroutine.
+
+This change should also help in general to reduce load on the main
+thread and for it to get stuck on IO, i.e. same benefits as using a
+dedicated IO thread for regular drives. This is particularly
+interesting when the VM state storage is a network storage like NFS.
+
+With some luck, it could also help with bug #6262 [0]. The failure
+there happens while issuing/right after the savevm-start QMP command,
+so the most likely coroutine is the process_savevm_co() that was
+previously scheduled to the iohandler context. Likely someone polls
+the iohandler context and wants to enter the already scheduled
+coroutine leading to the abort():
+> qemu_aio_coroutine_enter: Co-routine was already scheduled in 'aio_co_schedule'
+With a dedicated iothread, there hopefully is no such race.
+
+The comment above querying the pending bytes wrongly talked about the
+"iothread lock", but should've been "iohandler lock". This was even
+renamed to BQL (big QEMU lock) a few releases ago. Even if that was
+not a typo to begin with, there are no AioContext locks anymore.
+
+[0]: https://bugzilla.proxmox.com/show_bug.cgi?id=6262
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+[WB: update to the changed error handling in the previous commit]
+Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
+---
+ migration/savevm-async.c | 42 ++++++++++++++++++++++++++++------------
+ 1 file changed, 30 insertions(+), 12 deletions(-)
+
+diff --git a/migration/savevm-async.c b/migration/savevm-async.c
+index d8d2c80475..11ea4c601d 100644
+--- a/migration/savevm-async.c
++++ b/migration/savevm-async.c
+@@ -25,6 +25,7 @@
+ #include "qemu/main-loop.h"
+ #include "qemu/rcu.h"
+ #include "qemu/yank.h"
++#include "sysemu/iothread.h"
+ 
+ /* #define DEBUG_SAVEVM_STATE */
+ 
+@@ -57,6 +58,7 @@ static struct SnapshotState {
+     QEMUBH *finalize_bh;
+     Coroutine *co;
+     QemuCoSleep target_close_wait;
++    IOThread *iothread;
+ } snap_state;
+ 
+ static bool savevm_aborted(void)
+@@ -256,16 +258,13 @@ static void coroutine_fn process_savevm_co(void *opaque)
+         uint64_t threshold = 400 * 1000;
+ 
+         /*
+-         * pending_{estimate,exact} are expected to be called without iothread
+-         * lock. Similar to what is done in migration.c, call the exact variant
+-         * only once pend_precopy in the estimate is below the threshold.
++         * Similar to what is done in migration.c, call the exact variant only
++         * once pend_precopy in the estimate is below the threshold.
+          */
+-        bql_unlock();
+         qemu_savevm_state_pending_estimate(&pend_precopy, &pend_postcopy);
+         if (pend_precopy <= threshold) {
+             qemu_savevm_state_pending_exact(&pend_precopy, &pend_postcopy);
+         }
+-        bql_lock();
+         pending_size = pend_precopy + pend_postcopy;
+ 
+         /*
+@@ -332,11 +331,17 @@ static void coroutine_fn process_savevm_co(void *opaque)
+     qemu_bh_schedule(snap_state.finalize_bh);
+ }
+ 
++static void savevm_cleanup_iothread(void) {
++    if (snap_state.iothread) {
++        iothread_destroy(snap_state.iothread);
++        snap_state.iothread = NULL;
++    }
++}
++
+ void qmp_savevm_start(const char *statefile, Error **errp)
+ {
+     Error *local_err = NULL;
+     MigrationState *ms = migrate_get_current();
+-    AioContext *iohandler_ctx = iohandler_get_aio_context();
+     BlockDriverState *target_bs = NULL;
+     int ret = 0;
+ 
+@@ -374,6 +379,19 @@ void qmp_savevm_start(const char *statefile, Error **errp)
+         goto fail;
+     }
+ 
++    if (snap_state.iothread) {
++        /* This is not expected, so warn about it, but no point in re-creating a new iothread. */
++        warn_report("iothread for snapshot already exists - re-using");
++    } else {
++        snap_state.iothread =
++            iothread_create("__proxmox_savevm_async_iothread__", &local_err);
++        if (!snap_state.iothread) {
++            error_setg(errp, "creating iothread failed: %s",
++                       local_err ? error_get_pretty(local_err) : "unknown error");
++            goto fail;
++        }
++    }
++
+     /* Open the image */
+     QDict *options = NULL;
+     options = qdict_new();
+@@ -422,22 +440,20 @@ void qmp_savevm_start(const char *statefile, Error **errp)
+         goto fail;
+     }
+ 
+-    /* Async processing from here on out happens in iohandler context, so let
+-     * the target bdrv have its home there.
+-     */
+-    ret = blk_set_aio_context(snap_state.target, iohandler_ctx, &local_err);
++    ret = blk_set_aio_context(snap_state.target, snap_state.iothread->ctx, &local_err);
+     if (ret != 0) {
+-        warn_report("failed to set iohandler context for VM state target: %s %s",
++        warn_report("failed to set iothread context for VM state target: %s %s",
+                     local_err ? error_get_pretty(local_err) : "unknown error",
+                     strerror(-ret));
+     }
+ 
+     snap_state.co = qemu_coroutine_create(&process_savevm_co, NULL);
+-    aio_co_schedule(iohandler_ctx, snap_state.co);
++    aio_co_schedule(snap_state.iothread->ctx, snap_state.co);
+ 
+     return;
+ 
+ fail:
++    savevm_cleanup_iothread();
+     save_snapshot_error("setup failed");
+ }
+ 
+@@ -463,6 +479,8 @@ static void coroutine_fn wait_for_close_co(void *opaque)
+         DPRINTF("savevm-end: no target file open\n");
+     }
+ 
++    savevm_cleanup_iothread();
++
+     // File closed and no other error, so ensure next snapshot can be started.
+     if (snap_state.state != SAVE_STATE_ERROR) {
+         snap_state.state = SAVE_STATE_DONE;

debian/patches/pve/0062-savevm-async-treat-failure-to-set-iothread-context-a.patch

@@ -0,0 +1,33 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Mon, 31 Mar 2025 16:55:07 +0200
+Subject: [PATCH] savevm-async: treat failure to set iothread context as a hard
+ failure
+
+This is not expected to ever fail and there might be assumptions about
+having the expected context down the line.
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+[WB: update to changed error handling]
+Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
+---
+ migration/savevm-async.c | 6 +++---
+ 1 file changed, 3 insertions(+), 3 deletions(-)
+
+diff --git a/migration/savevm-async.c b/migration/savevm-async.c
+index 11ea4c601d..f2b10b5519 100644
+--- a/migration/savevm-async.c
++++ b/migration/savevm-async.c
+@@ -442,9 +442,9 @@ void qmp_savevm_start(const char *statefile, Error **errp)
+ 
+     ret = blk_set_aio_context(snap_state.target, snap_state.iothread->ctx, &local_err);
+     if (ret != 0) {
+-        warn_report("failed to set iothread context for VM state target: %s %s",
+-                    local_err ? error_get_pretty(local_err) : "unknown error",
+-                    strerror(-ret));
++        error_setg_errno(errp, -ret, "failed to set iothread context for VM state target: %s",
++                         local_err ? error_get_pretty(local_err) : "unknown error");
++        goto fail;
+     }
+ 
+     snap_state.co = qemu_coroutine_create(&process_savevm_co, NULL);

debian/patches/pve/0063-PVE-backup-clean-up-directly-in-setup_snapshot_acces.patch

@@ -0,0 +1,41 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Thu, 3 Apr 2025 14:30:41 +0200
+Subject: [PATCH] PVE backup: clean up directly in setup_snapshot_access() when
+ it fails
+
+The only thing that might need to be cleaned up after
+setup_snapshot_access() failed is dropping the cbw filter. Do so in
+the single branch it matters inside setup_snapshot_access() itself.
+This avoids the need that callers of setup_snapshot_access() use
+cleanup_snapshot_access() when the call failed.
+
+Suggested-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+Reviewed-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
+---
+ pve-backup.c | 4 +++-
+ 1 file changed, 3 insertions(+), 1 deletion(-)
+
+diff --git a/pve-backup.c b/pve-backup.c
+index 32352fb5ec..2408f182bc 100644
+--- a/pve-backup.c
++++ b/pve-backup.c
+@@ -576,6 +576,9 @@ static int setup_snapshot_access(PVEBackupDevInfo *di, Error **errp)
+     di->fleecing.snapshot_access =
+         bdrv_open(NULL, NULL, snapshot_access_opts, BDRV_O_RDWR | BDRV_O_UNMAP, &local_err);
+     if (!di->fleecing.snapshot_access) {
++        bdrv_cbw_drop(di->fleecing.cbw);
++        di->fleecing.cbw = NULL;
++
+         error_setg(errp, "setting up snapshot access for fleecing failed: %s",
+                    local_err ? error_get_pretty(local_err) : "unknown error");
+         return -1;
+@@ -629,7 +632,6 @@ static void create_backup_jobs_bh(void *opaque) {
+                 error_setg(errp, "%s - setting up snapshot access for fleecing failed: %s",
+                            di->device_name,
+                            local_err ? error_get_pretty(local_err) : "unknown error");
+-                cleanup_snapshot_access(di);
+                 bdrv_drained_end(di->bs);
+                 break;
+             }

debian/patches/pve/0064-PVE-backup-factor-out-helper-to-clear-backup-state-s.patch

@@ -0,0 +1,59 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Thu, 3 Apr 2025 14:30:42 +0200
+Subject: [PATCH] PVE backup: factor out helper to clear backup state's bitmap
+ list
+
+Suggested-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+Reviewed-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
+---
+ pve-backup.c | 28 ++++++++++++++++++----------
+ 1 file changed, 18 insertions(+), 10 deletions(-)
+
+diff --git a/pve-backup.c b/pve-backup.c
+index 2408f182bc..915649b5f9 100644
+--- a/pve-backup.c
++++ b/pve-backup.c
+@@ -811,6 +811,23 @@ err:
+     return di_list;
+ }
+ 
++/*
++ * To be called with the backup_state.stat mutex held.
++ */
++static void clear_backup_state_bitmap_list(void) {
++
++    if (backup_state.stat.bitmap_list) {
++        GList *bl = backup_state.stat.bitmap_list;
++        while (bl) {
++            g_free(((PBSBitmapInfo *)bl->data)->drive);
++            g_free(bl->data);
++            bl = g_list_next(bl);
++        }
++        g_list_free(backup_state.stat.bitmap_list);
++        backup_state.stat.bitmap_list = NULL;
++    }
++}
++
+ UuidInfo coroutine_fn *qmp_backup(
+     const char *backup_file,
+     const char *password,
+@@ -898,16 +915,7 @@ UuidInfo coroutine_fn *qmp_backup(
+     backup_state.stat.reused = 0;
+ 
+     /* clear previous backup's bitmap_list */
+-    if (backup_state.stat.bitmap_list) {
+-        GList *bl = backup_state.stat.bitmap_list;
+-        while (bl) {
+-            g_free(((PBSBitmapInfo *)bl->data)->drive);
+-            g_free(bl->data);
+-            bl = g_list_next(bl);
+-        }
+-        g_list_free(backup_state.stat.bitmap_list);
+-        backup_state.stat.bitmap_list = NULL;
+-    }
++    clear_backup_state_bitmap_list();
+ 
+     if (format == BACKUP_FORMAT_PBS) {
+         if (!password) {

debian/patches/pve/0065-PVE-backup-factor-out-helper-to-initialize-backup-st.patch

@@ -0,0 +1,95 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Thu, 3 Apr 2025 14:30:43 +0200
+Subject: [PATCH] PVE backup: factor out helper to initialize backup state stat
+ struct
+
+Suggested-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+Reviewed-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
+---
+ pve-backup.c | 62 ++++++++++++++++++++++++++++++++--------------------
+ 1 file changed, 38 insertions(+), 24 deletions(-)
+
+diff --git a/pve-backup.c b/pve-backup.c
+index 915649b5f9..88a981f81c 100644
+--- a/pve-backup.c
++++ b/pve-backup.c
+@@ -828,6 +828,43 @@ static void clear_backup_state_bitmap_list(void) {
+     }
+ }
+ 
++/*
++ * Initializes most of the backup state 'stat' struct. Note that 'reused' and
++ * 'bitmap_list' are not changed by this function and need to be handled by
++ * the caller. In particular, 'reused' needs to be set before calling this
++ * function.
++ *
++ * To be called with the backup_state.stat mutex held.
++ */
++static void initialize_backup_state_stat(
++    const char *backup_file,
++    uuid_t uuid,
++    size_t total)
++{
++    if (backup_state.stat.error) {
++        error_free(backup_state.stat.error);
++        backup_state.stat.error = NULL;
++    }
++
++    backup_state.stat.start_time = time(NULL);
++    backup_state.stat.end_time = 0;
++
++    if (backup_state.stat.backup_file) {
++        g_free(backup_state.stat.backup_file);
++    }
++    backup_state.stat.backup_file = g_strdup(backup_file);
++
++    uuid_copy(backup_state.stat.uuid, uuid);
++    uuid_unparse_lower(uuid, backup_state.stat.uuid_str);
++
++    backup_state.stat.total = total;
++    backup_state.stat.dirty = total - backup_state.stat.reused;
++    backup_state.stat.transferred = 0;
++    backup_state.stat.zero_bytes = 0;
++    backup_state.stat.finishing = false;
++    backup_state.stat.starting = true;
++}
++
+ UuidInfo coroutine_fn *qmp_backup(
+     const char *backup_file,
+     const char *password,
+@@ -1070,32 +1107,9 @@ UuidInfo coroutine_fn *qmp_backup(
+         }
+     }
+     /* initialize global backup_state now */
+-    /* note: 'reused' and 'bitmap_list' are initialized earlier */
+-
+-    if (backup_state.stat.error) {
+-        error_free(backup_state.stat.error);
+-        backup_state.stat.error = NULL;
+-    }
+-
+-    backup_state.stat.start_time = time(NULL);
+-    backup_state.stat.end_time = 0;
+-
+-    if (backup_state.stat.backup_file) {
+-        g_free(backup_state.stat.backup_file);
+-    }
+-    backup_state.stat.backup_file = g_strdup(backup_file);
+-
+-    uuid_copy(backup_state.stat.uuid, uuid);
+-    uuid_unparse_lower(uuid, backup_state.stat.uuid_str);
++    initialize_backup_state_stat(backup_file, uuid, total);
+     char *uuid_str = g_strdup(backup_state.stat.uuid_str);
+ 
+-    backup_state.stat.total = total;
+-    backup_state.stat.dirty = total - backup_state.stat.reused;
+-    backup_state.stat.transferred = 0;
+-    backup_state.stat.zero_bytes = 0;
+-    backup_state.stat.finishing = false;
+-    backup_state.stat.starting = true;
+-
+     qemu_mutex_unlock(&backup_state.stat.lock);
+ 
+     backup_state.speed = (has_speed && speed > 0) ? speed : 0;

debian/patches/pve/0066-PVE-backup-add-target-ID-in-backup-state.patch

@@ -0,0 +1,63 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Thu, 3 Apr 2025 14:30:44 +0200
+Subject: [PATCH] PVE backup: add target ID in backup state
+
+In preparation for allowing multiple backup providers and potentially
+multiple targets for a given provider. Each backup target can then
+have its own dirty bitmap and there can be additional checks that the
+current backup state is actually associated to the expected target.
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+Reviewed-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
+---
+ pve-backup.c | 15 ++++++++++++++-
+ 1 file changed, 14 insertions(+), 1 deletion(-)
+
+diff --git a/pve-backup.c b/pve-backup.c
+index 88a981f81c..8789a0667a 100644
+--- a/pve-backup.c
++++ b/pve-backup.c
+@@ -70,6 +70,7 @@ static struct PVEBackupState {
+     JobTxn *txn;
+     CoMutex backup_mutex;
+     CoMutex dump_callback_mutex;
++    char *target_id;
+ } backup_state;
+ 
+ static void pvebackup_init(void)
+@@ -865,6 +866,16 @@ static void initialize_backup_state_stat(
+     backup_state.stat.starting = true;
+ }
+ 
++/*
++ * To be called with the backup_state mutex held.
++ */
++static void backup_state_set_target_id(const char *target_id) {
++    if (backup_state.target_id) {
++        g_free(backup_state.target_id);
++    }
++    backup_state.target_id = g_strdup(target_id);
++}
++
+ UuidInfo coroutine_fn *qmp_backup(
+     const char *backup_file,
+     const char *password,
+@@ -904,7 +915,7 @@ UuidInfo coroutine_fn *qmp_backup(
+ 
+     if (backup_state.di_list) {
+         error_set(errp, ERROR_CLASS_GENERIC_ERROR,
+-                  "previous backup not finished");
++                  "previous backup for target '%s' not finished", backup_state.target_id);
+         qemu_co_mutex_unlock(&backup_state.backup_mutex);
+         return NULL;
+     }
+@@ -1122,6 +1133,8 @@ UuidInfo coroutine_fn *qmp_backup(
+     backup_state.vmaw = vmaw;
+     backup_state.pbs = pbs;
+ 
++    backup_state_set_target_id("Proxmox");
++
+     backup_state.di_list = di_list;
+ 
+     uuid_info = g_malloc0(sizeof(*uuid_info));

debian/patches/pve/0067-PVE-backup-get-device-info-allow-caller-to-specify-f.patch

@@ -0,0 +1,57 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Thu, 3 Apr 2025 14:30:45 +0200
+Subject: [PATCH] PVE backup: get device info: allow caller to specify filter
+ for which devices use fleecing
+
+For providing snapshot-access to external backup providers, EFI and
+TPM also need an associated fleecing image. The new caller will thus
+need a different filter.
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+Reviewed-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
+---
+ pve-backup.c | 9 +++++----
+ 1 file changed, 5 insertions(+), 4 deletions(-)
+
+diff --git a/pve-backup.c b/pve-backup.c
+index 8789a0667a..755f1abcf1 100644
+--- a/pve-backup.c
++++ b/pve-backup.c
+@@ -719,7 +719,7 @@ static void create_backup_jobs_bh(void *opaque) {
+ /*
+  * EFI disk and TPM state are small and it's just not worth setting up fleecing for them.
+  */
+-static bool device_uses_fleecing(const char *device_id)
++static bool fleecing_no_efi_tpm(const char *device_id)
+ {
+     return strncmp(device_id, "drive-efidisk", 13) && strncmp(device_id, "drive-tpmstate", 14);
+ }
+@@ -731,7 +731,7 @@ static bool device_uses_fleecing(const char *device_id)
+  */
+ static GList coroutine_fn GRAPH_RDLOCK *get_device_info(
+     const char *devlist,
+-    bool fleecing,
++    bool (*device_uses_fleecing)(const char*),
+     Error **errp)
+ {
+     gchar **devs = NULL;
+@@ -757,7 +757,7 @@ static GList coroutine_fn GRAPH_RDLOCK *get_device_info(
+             di->bs = bs;
+             di->device_name = g_strdup(bdrv_get_device_name(bs));
+ 
+-            if (fleecing && device_uses_fleecing(*d)) {
++            if (device_uses_fleecing && device_uses_fleecing(*d)) {
+                 g_autofree gchar *fleecing_devid = g_strconcat(*d, "-fleecing", NULL);
+                 BlockBackend *fleecing_blk = blk_by_name(fleecing_devid);
+                 if (!fleecing_blk) {
+@@ -924,7 +924,8 @@ UuidInfo coroutine_fn *qmp_backup(
+     format = has_format ? format : BACKUP_FORMAT_VMA;
+ 
+     bdrv_graph_co_rdlock();
+-    di_list = get_device_info(devlist, has_fleecing && fleecing, &local_err);
++    di_list = get_device_info(devlist, (has_fleecing && fleecing) ? fleecing_no_efi_tpm : NULL,
++                              &local_err);
+     bdrv_graph_co_rdunlock();
+     if (local_err) {
+         error_propagate(errp, local_err);

debian/patches/pve/0068-PVE-backup-implement-backup-access-setup-and-teardow.patch

@@ -0,0 +1,495 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Thu, 3 Apr 2025 14:30:46 +0200
+Subject: [PATCH] PVE backup: implement backup access setup and teardown API
+ for external providers
+
+For external backup providers, the state of the VM's disk images at
+the time the backup is started is preserved via a snapshot-access
+block node. Old data is moved to the fleecing image when new guest
+writes come in. The snapshot-access block node, as well as the
+associated bitmap in case of incremental backup, will be exported via
+NBD to the external provider. The NBD export will be done by the
+management layer, the missing functionality is setting up and tearing
+down the snapshot-access block nodes, which this patch adds.
+
+It is necessary to also set up fleecing for EFI and TPM disks, so that
+old data can be moved out of the way when a new guest write comes in.
+
+There can only be one regular backup or one active backup access at
+a time, because both require replacing the original block node of the
+drive. Thus the backup state is re-used, and checks are added to
+prohibit regular backup while snapshot access is active and vice
+versa.
+
+The block nodes added by the backup-access-setup QMP call are not
+tracked anywhere else (there is no job they are associated to like for
+regular backup). This requires adding a callback for teardown when
+QEMU exits, i.e. in qemu_cleanup(). Otherwise, there will be an
+assertion failure that the block graph is not empty when QEMU exits
+before the backup-access-teardown QMP command is called.
+
+The code for the qmp_backup_access_setup() was based on the existing
+qmp_backup() routine.
+
+The return value for the setup QMP command contains information about
+the snapshot-access block nodes that can be used by the management
+layer to set up the NBD exports.
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+Reviewed-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
+---
+ pve-backup.c         | 264 ++++++++++++++++++++++++++++++++++++++++++-
+ pve-backup.h         |  16 +++
+ qapi/block-core.json |  49 ++++++++
+ system/runstate.c    |   6 +
+ 4 files changed, 329 insertions(+), 6 deletions(-)
+ create mode 100644 pve-backup.h
+
+diff --git a/pve-backup.c b/pve-backup.c
+index 755f1abcf1..091b5bd231 100644
+--- a/pve-backup.c
++++ b/pve-backup.c
+@@ -1,4 +1,5 @@
+ #include "proxmox-backup-client.h"
++#include "pve-backup.h"
+ #include "vma.h"
+ 
+ #include "qemu/osdep.h"
+@@ -588,6 +589,36 @@ static int setup_snapshot_access(PVEBackupDevInfo *di, Error **errp)
+     return 0;
+ }
+ 
++static void setup_all_snapshot_access_bh(void *opaque)
++{
++    assert(!qemu_in_coroutine());
++
++    CoCtxData *data = (CoCtxData*)opaque;
++    Error **errp = (Error**)data->data;
++
++    Error *local_err = NULL;
++
++    GList *l =  backup_state.di_list;
++    while (l) {
++        PVEBackupDevInfo *di = (PVEBackupDevInfo *)l->data;
++        l = g_list_next(l);
++
++        bdrv_drained_begin(di->bs);
++
++        if (setup_snapshot_access(di, &local_err) < 0) {
++            bdrv_drained_end(di->bs);
++            error_setg(errp, "%s - setting up snapshot access failed: %s", di->device_name,
++                       local_err ? error_get_pretty(local_err) : "unknown error");
++            break;
++        }
++
++        bdrv_drained_end(di->bs);
++    }
++
++    /* return */
++    aio_co_enter(data->ctx, data->co);
++}
++
+ /*
+  * backup_job_create can *not* be run from a coroutine, so this can't either.
+  * The caller is responsible that backup_mutex is held nonetheless.
+@@ -724,6 +755,11 @@ static bool fleecing_no_efi_tpm(const char *device_id)
+     return strncmp(device_id, "drive-efidisk", 13) && strncmp(device_id, "drive-tpmstate", 14);
+ }
+ 
++static bool fleecing_all(const char *device_id)
++{
++    return true;
++}
++
+ /*
+  * Returns a list of device infos, which needs to be freed by the caller. In
+  * case of an error, errp will be set, but the returned value might still be a
+@@ -839,8 +875,9 @@ static void clear_backup_state_bitmap_list(void) {
+  */
+ static void initialize_backup_state_stat(
+     const char *backup_file,
+-    uuid_t uuid,
+-    size_t total)
++    uuid_t *uuid,
++    size_t total,
++    bool starting)
+ {
+     if (backup_state.stat.error) {
+         error_free(backup_state.stat.error);
+@@ -855,15 +892,19 @@ static void initialize_backup_state_stat(
+     }
+     backup_state.stat.backup_file = g_strdup(backup_file);
+ 
+-    uuid_copy(backup_state.stat.uuid, uuid);
+-    uuid_unparse_lower(uuid, backup_state.stat.uuid_str);
++    if (uuid) {
++        uuid_copy(backup_state.stat.uuid, *uuid);
++        uuid_unparse_lower(*uuid, backup_state.stat.uuid_str);
++    } else {
++        backup_state.stat.uuid_str[0] = '\0';
++    }
+ 
+     backup_state.stat.total = total;
+     backup_state.stat.dirty = total - backup_state.stat.reused;
+     backup_state.stat.transferred = 0;
+     backup_state.stat.zero_bytes = 0;
+     backup_state.stat.finishing = false;
+-    backup_state.stat.starting = true;
++    backup_state.stat.starting = starting;
+ }
+ 
+ /*
+@@ -876,6 +917,216 @@ static void backup_state_set_target_id(const char *target_id) {
+     backup_state.target_id = g_strdup(target_id);
+ }
+ 
++BackupAccessInfoList *coroutine_fn qmp_backup_access_setup(
++    const char *target_id,
++    const char *devlist,
++    Error **errp)
++{
++    assert(qemu_in_coroutine());
++
++    qemu_co_mutex_lock(&backup_state.backup_mutex);
++
++    Error *local_err = NULL;
++    GList *di_list = NULL;
++    GList *l;
++
++    if (backup_state.di_list) {
++        error_set(errp, ERROR_CLASS_GENERIC_ERROR,
++                  "previous backup for target '%s' not finished", backup_state.target_id);
++        qemu_co_mutex_unlock(&backup_state.backup_mutex);
++        return NULL;
++    }
++
++    bdrv_graph_co_rdlock();
++    di_list = get_device_info(devlist, fleecing_all, &local_err);
++    bdrv_graph_co_rdunlock();
++    if (local_err) {
++        error_propagate(errp, local_err);
++        goto err;
++    }
++    assert(di_list);
++
++    size_t total = 0;
++
++    l = di_list;
++    while (l) {
++        PVEBackupDevInfo *di = (PVEBackupDevInfo *)l->data;
++        l = g_list_next(l);
++
++        ssize_t size = bdrv_getlength(di->bs);
++        if (size < 0) {
++            error_setg_errno(errp, -size, "bdrv_getlength failed");
++            goto err;
++        }
++        di->size = size;
++        total += size;
++
++        di->completed_ret = INT_MAX;
++    }
++
++    qemu_mutex_lock(&backup_state.stat.lock);
++    backup_state.stat.reused = 0;
++
++    /* clear previous backup's bitmap_list */
++    clear_backup_state_bitmap_list();
++
++    /* starting=false, because there is no associated QEMU job */
++    initialize_backup_state_stat(NULL, NULL, total, false);
++
++    qemu_mutex_unlock(&backup_state.stat.lock);
++
++    backup_state_set_target_id(target_id);
++
++    backup_state.vmaw = NULL;
++    backup_state.pbs = NULL;
++
++    backup_state.di_list = di_list;
++
++    /* Run setup_all_snapshot_access_bh outside of coroutine (in BH) but keep
++    * backup_mutex locked. This is fine, a CoMutex can be held across yield
++    * points, and we'll release it as soon as the BH reschedules us.
++    */
++    CoCtxData waker = {
++        .co = qemu_coroutine_self(),
++        .ctx = qemu_get_current_aio_context(),
++        .data = &local_err,
++    };
++    aio_bh_schedule_oneshot(waker.ctx, setup_all_snapshot_access_bh, &waker);
++    qemu_coroutine_yield();
++
++    if (local_err) {
++        error_propagate(errp, local_err);
++        goto err;
++    }
++
++    qemu_co_mutex_unlock(&backup_state.backup_mutex);
++
++    BackupAccessInfoList *bai_head = NULL, **p_bai_next = &bai_head;
++
++    l = di_list;
++    while (l) {
++        PVEBackupDevInfo *di = (PVEBackupDevInfo *)l->data;
++        l = g_list_next(l);
++
++        BackupAccessInfoList *info = g_malloc0(sizeof(*info));
++        info->value = g_malloc0(sizeof(*info->value));
++        info->value->node_name = g_strdup(bdrv_get_node_name(di->fleecing.snapshot_access));
++        info->value->device = g_strdup(di->device_name);
++        info->value->size = di->size;
++
++        *p_bai_next = info;
++        p_bai_next = &info->next;
++    }
++
++    return bai_head;
++
++err:
++
++    l = di_list;
++    while (l) {
++        PVEBackupDevInfo *di = (PVEBackupDevInfo *)l->data;
++        l = g_list_next(l);
++
++        g_free(di->device_name);
++        di->device_name = NULL;
++
++        g_free(di);
++    }
++    g_list_free(di_list);
++    backup_state.di_list = NULL;
++
++    qemu_co_mutex_unlock(&backup_state.backup_mutex);
++    return NULL;
++}
++
++/*
++ * Caller needs to hold the backup mutex or the BQL.
++ */
++void backup_access_teardown(void)
++{
++    GList *l = backup_state.di_list;
++
++    qemu_mutex_lock(&backup_state.stat.lock);
++    backup_state.stat.finishing = true;
++    qemu_mutex_unlock(&backup_state.stat.lock);
++
++    while (l) {
++        PVEBackupDevInfo *di = (PVEBackupDevInfo *)l->data;
++        l = g_list_next(l);
++
++        if (di->fleecing.snapshot_access) {
++            bdrv_unref(di->fleecing.snapshot_access);
++            di->fleecing.snapshot_access = NULL;
++        }
++        if (di->fleecing.cbw) {
++            bdrv_cbw_drop(di->fleecing.cbw);
++            di->fleecing.cbw = NULL;
++        }
++
++        g_free(di->device_name);
++        di->device_name = NULL;
++
++        g_free(di);
++    }
++    g_list_free(backup_state.di_list);
++    backup_state.di_list = NULL;
++
++    qemu_mutex_lock(&backup_state.stat.lock);
++    backup_state.stat.end_time = time(NULL);
++    backup_state.stat.finishing = false;
++    qemu_mutex_unlock(&backup_state.stat.lock);
++}
++
++// Not done in a coroutine, because bdrv_co_unref() and cbw_drop() would just spawn BHs anyways.
++// Caller needs to hold the backup_state.backup_mutex lock
++static void backup_access_teardown_bh(void *opaque)
++{
++    CoCtxData *data = (CoCtxData*)opaque;
++
++    backup_access_teardown();
++
++    /* return */
++    aio_co_enter(data->ctx, data->co);
++}
++
++void coroutine_fn qmp_backup_access_teardown(const char *target_id, Error **errp)
++{
++    assert(qemu_in_coroutine());
++
++    qemu_co_mutex_lock(&backup_state.backup_mutex);
++
++    if (!backup_state.target_id) { // nothing to do
++        qemu_co_mutex_unlock(&backup_state.backup_mutex);
++        return;
++    }
++
++    /*
++     * Continue with target_id == NULL, used by the callback registered for qemu_cleanup()
++     */
++    if (target_id && strcmp(backup_state.target_id, target_id)) {
++        error_setg(errp, "cannot teardown backup access - got target %s instead of %s",
++                   target_id, backup_state.target_id);
++        qemu_co_mutex_unlock(&backup_state.backup_mutex);
++        return;
++    }
++
++    if (!strcmp(backup_state.target_id, "Proxmox VE")) {
++        error_setg(errp, "cannot teardown backup access for PVE - use backup-cancel instead");
++        qemu_co_mutex_unlock(&backup_state.backup_mutex);
++        return;
++    }
++
++    CoCtxData waker = {
++        .co = qemu_coroutine_self(),
++        .ctx = qemu_get_current_aio_context(),
++    };
++    aio_bh_schedule_oneshot(waker.ctx, backup_access_teardown_bh, &waker);
++    qemu_coroutine_yield();
++
++    qemu_co_mutex_unlock(&backup_state.backup_mutex);
++    return;
++}
++
+ UuidInfo coroutine_fn *qmp_backup(
+     const char *backup_file,
+     const char *password,
+@@ -1119,7 +1370,7 @@ UuidInfo coroutine_fn *qmp_backup(
+         }
+     }
+     /* initialize global backup_state now */
+-    initialize_backup_state_stat(backup_file, uuid, total);
++    initialize_backup_state_stat(backup_file, &uuid, total, true);
+     char *uuid_str = g_strdup(backup_state.stat.uuid_str);
+ 
+     qemu_mutex_unlock(&backup_state.stat.lock);
+@@ -1298,5 +1549,6 @@ ProxmoxSupportStatus *qmp_query_proxmox_support(Error **errp)
+     ret->pbs_masterkey = true;
+     ret->backup_max_workers = true;
+     ret->backup_fleecing = true;
++    ret->backup_access_api = true;
+     return ret;
+ }
+diff --git a/pve-backup.h b/pve-backup.h
+new file mode 100644
+index 0000000000..4033bc848f
+--- /dev/null
++++ b/pve-backup.h
+@@ -0,0 +1,16 @@
++/*
++ * Bacup code used by Proxmox VE
++ *
++ * Copyright (C) Proxmox Server Solutions
++ *
++ * This work is licensed under the terms of the GNU GPL, version 2 or later.
++ * See the COPYING file in the top-level directory.
++ *
++ */
++
++#ifndef PVE_BACKUP_H
++#define PVE_BACKUP_H
++
++void backup_access_teardown(void);
++
++#endif /* PVE_BACKUP_H */
+diff --git a/qapi/block-core.json b/qapi/block-core.json
+index c581f1f238..3f092221ce 100644
+--- a/qapi/block-core.json
++++ b/qapi/block-core.json
+@@ -1019,6 +1019,9 @@
+ #
+ # @pbs-library-version: Running version of libproxmox-backup-qemu0 library.
+ #
++# @backup-access-api: Whether backup access API for external providers is
++#     supported or not.
++#
+ # @backup-fleecing: Whether backup fleecing is supported or not.
+ #
+ # @backup-max-workers: Whether the 'max-workers' @BackupPerf setting is
+@@ -1032,6 +1035,7 @@
+             'pbs-dirty-bitmap-migration': 'bool',
+             'pbs-masterkey': 'bool',
+             'pbs-library-version': 'str',
++            'backup-access-api': 'bool',
+             'backup-fleecing': 'bool',
+             'backup-max-workers': 'bool' } }
+ 
+@@ -1098,6 +1102,51 @@
+ ##
+ { 'command': 'query-pbs-bitmap-info', 'returns': ['PBSBitmapInfo'] }
+ 
++##
++# @BackupAccessInfo:
++#
++# Info associated to a snapshot access for backup.  For more information about
++# the bitmap see @BackupAccessBitmapMode.
++#
++# @node-name: the block node name of the snapshot-access node.
++#
++# @device: the device on top of which the snapshot access was created.
++#
++# @size: the size of the block device in bytes.
++#
++##
++{ 'struct': 'BackupAccessInfo',
++  'data': { 'node-name': 'str', 'device': 'str', 'size': 'size' } }
++
++##
++# @backup-access-setup:
++#
++# Set up snapshot access to VM drives for an external backup provider.  No other
++# backup or backup access can be done before tearing down the backup access.
++#
++# @target-id: the unique ID of the backup target.
++#
++# @devlist: list of block device names (separated by ',', ';' or ':'). By
++#     default the backup includes all writable block devices.
++#
++# Returns: a list of @BackupAccessInfo, one for each device.
++#
++##
++{ 'command': 'backup-access-setup',
++  'data': { 'target-id': 'str', '*devlist': 'str' },
++  'returns': [ 'BackupAccessInfo' ], 'coroutine': true }
++
++##
++# @backup-access-teardown:
++#
++# Tear down previously setup snapshot access for the same target.
++#
++# @target-id: the ID of the backup target.
++#
++##
++{ 'command': 'backup-access-teardown', 'data': { 'target-id': 'str' },
++  'coroutine': true }
++
+ ##
+ # @BlockDeviceTimedStats:
+ #
+diff --git a/system/runstate.c b/system/runstate.c
+index c2c9afa905..6f93d7c2fb 100644
+--- a/system/runstate.c
++++ b/system/runstate.c
+@@ -60,6 +60,7 @@
+ #include "sysemu/sysemu.h"
+ #include "sysemu/tpm.h"
+ #include "trace.h"
++#include "pve-backup.h"
+ 
+ static NotifierList exit_notifiers =
+     NOTIFIER_LIST_INITIALIZER(exit_notifiers);
+@@ -920,6 +921,11 @@ void qemu_cleanup(int status)
+      * requests happening from here on anyway.
+      */
+     bdrv_drain_all_begin();
++    /*
++     * The backup access is set up by a QMP command, but is neither owned by a monitor nor
++     * associated to a BlockBackend. Need to tear it down manually here.
++     */
++    backup_access_teardown();
+     job_cancel_sync_all();
+     bdrv_close_all();
+ 

debian/patches/pve/0069-PVE-backup-factor-out-get_single_device_info-helper.patch

@@ -0,0 +1,122 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Fiona Ebner <f.ebner@proxmox.com>
+Date: Thu, 3 Apr 2025 14:30:47 +0200
+Subject: [PATCH] PVE backup: factor out get_single_device_info() helper
+
+Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
+[WB: free di and di->device_name on error]
+Sigend-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
+---
+ pve-backup.c | 90 +++++++++++++++++++++++++++++++---------------------
+ 1 file changed, 53 insertions(+), 37 deletions(-)
+
+diff --git a/pve-backup.c b/pve-backup.c
+index 091b5bd231..8b7414f057 100644
+--- a/pve-backup.c
++++ b/pve-backup.c
+@@ -760,6 +760,57 @@ static bool fleecing_all(const char *device_id)
+     return true;
+ }
+ 
++static PVEBackupDevInfo coroutine_fn GRAPH_RDLOCK *get_single_device_info(
++    const char *device,
++    bool (*device_uses_fleecing)(const char*),
++    Error **errp)
++{
++    BlockBackend *blk = blk_by_name(device);
++    if (!blk) {
++        error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
++                  "Device '%s' not found", device);
++        return NULL;
++    }
++    BlockDriverState *bs = blk_bs(blk);
++    if (!bdrv_co_is_inserted(bs)) {
++        error_setg(errp, "Device '%s' has no medium", device);
++        return NULL;
++    }
++    PVEBackupDevInfo *di = g_new0(PVEBackupDevInfo, 1);
++    di->bs = bs;
++    di->device_name = g_strdup(bdrv_get_device_name(bs));
++
++    if (device_uses_fleecing && device_uses_fleecing(device)) {
++        g_autofree gchar *fleecing_devid = g_strconcat(device, "-fleecing", NULL);
++        BlockBackend *fleecing_blk = blk_by_name(fleecing_devid);
++        if (!fleecing_blk) {
++            error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
++                      "Device '%s' not found", fleecing_devid);
++            goto fail;
++        }
++        BlockDriverState *fleecing_bs = blk_bs(fleecing_blk);
++        if (!bdrv_co_is_inserted(fleecing_bs)) {
++            error_setg(errp, "Device '%s' has no medium", fleecing_devid);
++            goto fail;
++        }
++        /*
++         * Fleecing image needs to be the same size to act as a cbw target.
++         */
++        if (bs->total_sectors != fleecing_bs->total_sectors) {
++            error_setg(errp, "Size mismatch for '%s' - sector count %ld != %ld",
++                       fleecing_devid, fleecing_bs->total_sectors, bs->total_sectors);
++            goto fail;
++        }
++        di->fleecing.bs = fleecing_bs;
++    }
++
++    return di;
++fail:
++    g_free(di->device_name);
++    g_free(di);
++    return NULL;
++}
++
+ /*
+  * Returns a list of device infos, which needs to be freed by the caller. In
+  * case of an error, errp will be set, but the returned value might still be a
+@@ -778,45 +829,10 @@ static GList coroutine_fn GRAPH_RDLOCK *get_device_info(
+ 
+         gchar **d = devs;
+         while (d && *d) {
+-            BlockBackend *blk = blk_by_name(*d);
+-            if (!blk) {
+-                error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
+-                          "Device '%s' not found", *d);
++            PVEBackupDevInfo *di = get_single_device_info(*d, device_uses_fleecing, errp);
++            if (!di) {
+                 goto err;
+             }
+-            BlockDriverState *bs = blk_bs(blk);
+-            if (!bdrv_co_is_inserted(bs)) {
+-                error_setg(errp, "Device '%s' has no medium", *d);
+-                goto err;
+-            }
+-            PVEBackupDevInfo *di = g_new0(PVEBackupDevInfo, 1);
+-            di->bs = bs;
+-            di->device_name = g_strdup(bdrv_get_device_name(bs));
+-
+-            if (device_uses_fleecing && device_uses_fleecing(*d)) {
+-                g_autofree gchar *fleecing_devid = g_strconcat(*d, "-fleecing", NULL);
+-                BlockBackend *fleecing_blk = blk_by_name(fleecing_devid);
+-                if (!fleecing_blk) {
+-                    error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
+-                              "Device '%s' not found", fleecing_devid);
+-                    goto err;
+-                }
+-                BlockDriverState *fleecing_bs = blk_bs(fleecing_blk);
+-                if (!bdrv_co_is_inserted(fleecing_bs)) {
+-                    error_setg(errp, "Device '%s' has no medium", fleecing_devid);
+-                    goto err;
+-                }
+-                /*
+-                 * Fleecing image needs to be the same size to act as a cbw target.
+-                 */
+-                if (bs->total_sectors != fleecing_bs->total_sectors) {
+-                    error_setg(errp, "Size mismatch for '%s' - sector count %ld != %ld",
+-                               fleecing_devid, fleecing_bs->total_sectors, bs->total_sectors);
+-                    goto err;
+-                }
+-                di->fleecing.bs = fleecing_bs;
+-            }
+-
+             di_list = g_list_append(di_list, di);
+             d++;
+         }