... and implement it under POSIX. When a ThreadContext is provided,
create new threads via the context such that these new threads obtain a
properly configured CPU affinity.
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Message-Id: <20221014134720.168738-6-david@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Let's make it easier to pin threads created via a ThreadContext to
all host CPUs currently belonging to a given set of host NUMA nodes --
which is the common case.
"node-affinity" is simply a shortcut for setting "cpu-affinity" manually
to the list of host CPUs belonging to the set of host nodes. This property
can only be written.
A simple QEMU example to set the CPU affinity to host node 1 on a system
with two nodes, 24 CPUs each, whereby odd-numbered host CPUs belong to
host node 1:
qemu-system-x86_64 -S \
-object thread-context,id=tc1,node-affinity=1
And we can query the cpu-affinity via HMP/QMP:
(qemu) qom-get tc1 cpu-affinity
[
1,
3,
5,
7,
9,
11,
13,
15,
17,
19,
21,
23,
25,
27,
29,
31,
33,
35,
37,
39,
41,
43,
45,
47
]
We cannot query the node-affinity:
(qemu) qom-get tc1 node-affinity
Error: Insufficient permission to perform this operation
But note that due to dynamic library loading this example will not work
before we actually make use of thread_context_create_thread() in QEMU
code, because the type will otherwise not get registered. We'll wire
this up next to make it work.
Note that if the host CPUs for a host node change due do CPU hot(un)plug
CPU onlining/offlining (i.e., lscpu output changes) after the ThreadContext
was started, the CPU affinity will not get updated.
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Acked-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20221014134720.168738-5-david@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Setting the CPU affinity of QEMU threads is a bit problematic, because
QEMU doesn't always have permissions to set the CPU affinity itself,
for example, with seccomp after initialized by QEMU:
-sandbox enable=on,resourcecontrol=deny
General information about CPU affinities can be found in the man page of
taskset:
CPU affinity is a scheduler property that "bonds" a process to a given
set of CPUs on the system. The Linux scheduler will honor the given CPU
affinity and the process will not run on any other CPUs.
While upper layers are already aware of how to handle CPU affinities for
long-lived threads like iothreads or vcpu threads, especially short-lived
threads, as used for memory-backend preallocation, are more involved to
handle. These threads are created on demand and upper layers are not even
able to identify and configure them.
Introduce the concept of a ThreadContext, that is essentially a thread
used for creating new threads. All threads created via that context
thread inherit the configured CPU affinity. Consequently, it's
sufficient to create a ThreadContext and configure it once, and have all
threads created via that ThreadContext inherit the same CPU affinity.
The CPU affinity of a ThreadContext can be configured two ways:
(1) Obtaining the thread id via the "thread-id" property and setting the
CPU affinity manually (e.g., via taskset).
(2) Setting the "cpu-affinity" property and letting QEMU try set the
CPU affinity itself. This will fail if QEMU doesn't have permissions
to do so anymore after seccomp was initialized.
A simple QEMU example to set the CPU affinity to host CPU 0,1,6,7 would be:
qemu-system-x86_64 -S \
-object thread-context,id=tc1,cpu-affinity=0-1,cpu-affinity=6-7
And we can query it via HMP/QMP:
(qemu) qom-get tc1 cpu-affinity
[
0,
1,
6,
7
]
But note that due to dynamic library loading this example will not work
before we actually make use of thread_context_create_thread() in QEMU
code, because the type will otherwise not get registered. We'll wire
this up next to make it work.
In general, the interface behaves like pthread_setaffinity_np(): host
CPU numbers that are currently not available are ignored; only host CPU
numbers that are impossible with the current kernel will fail. If the
list of host CPU numbers does not include a single CPU that is
available, setting the CPU affinity will fail.
A ThreadContext can be reused, simply by reconfiguring the CPU affinity.
Note that the CPU affinity of previously created threads will not get
adjusted.
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Acked-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20221014134720.168738-4-david@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Usually, we let upper layers handle CPU pinning, because
pthread_setaffinity_np() (-> sched_setaffinity()) is blocked via
seccomp when starting QEMU with
-sandbox enable=on,resourcecontrol=deny
However, we want to configure and observe the CPU affinity of threads
from QEMU directly in some cases when the sandbox option is either not
enabled or not active yet.
So let's add a way to configure CPU pinning via
qemu_thread_set_affinity() and obtain CPU affinity via
qemu_thread_get_affinity() and implement them under POSIX using
pthread_setaffinity_np() + pthread_getaffinity_np().
Implementation under Windows is possible using SetProcessAffinityMask()
+ GetProcessAffinityMask(), however, that is left as future work.
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Message-Id: <20221014134720.168738-3-david@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Let's
* give the function a "qemu_*" style name
* make sure the parameters in the implementation match the prototype
* rename smp_cpus to max_threads, which makes the semantics of that
parameter clearer
... and add a function documentation.
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Message-Id: <20221014134720.168738-2-david@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
According to msdn documentation and Linux man pages, send() should try
to send as much as possible in blocking mode, while recv() may return
earlier with a smaller available amount, we should try to continue
send/recv from there.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20221006113657.2656108-3-marcandre.lureau@redhat.com>
With a pipe or other reasons, read/write may return less than the
requested bytes. This happens with the test-io-channel-command test on
Windows. glib spawn code uses a binary pipe of 4096 bytes, and the first
read returns that much (although more are requested), for some unclear
reason...
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20221006113657.2656108-2-marcandre.lureau@redhat.com>
As described in:
https://learn.microsoft.com/en-us/visualstudio/debugger/how-to-set-a-thread-name-in-native-code?view=vs-2022
SetThreadDescription() is available since Windows 10, version 1607 and
in some versions only by "Run Time Dynamic Linking". Its declaration is
not yet in mingw, so we lookup the function the same way glib does.
Tested with Visual Studio Community 2022 debugger.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Acked-by: Richard Henderson <richard.henderson@linaro.org>
Callers of coroutine_fn must be coroutine_fn themselves, or the call
must be within "if (qemu_in_coroutine())". Apply coroutine_fn to
functions where this holds.
Reviewed-by: Alberto Faria <afaria@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20220922084924.201610-23-pbonzini@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
qemu_coroutine_get_aio_context inspects a coroutine, but it does
not have to be called from the coroutine itself (or from any
coroutine).
Reviewed-by: Alberto Faria <afaria@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20220922084924.201610-6-pbonzini@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
qemu_socketpair() will create a pair of connected sockets
with FD_CLOEXEC set
Signed-off-by: Guoyi Tu <tugy@chinatelecom.cn>
Reviewed-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <17fa1eff729eeabd9a001f4639abccb127ceec81.1661240709.git.tugy@chinatelecom.cn>
The zpcii-disable machine property can be used to force-disable the use
of zPCI interpretation facilities for a VM. By default, this setting
will be off for machine 7.2 and newer.
Signed-off-by: Matthew Rosato <mjrosato@linux.ibm.com>
Message-Id: <20220902172737.170349-9-mjrosato@linux.ibm.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
[thuth: Fix contextual conflict in ccw_machine_7_1_instance_options()]
Signed-off-by: Thomas Huth <thuth@redhat.com>
Version: GnuPG v1
iQEcBAABAgAGBQJjEaMLAAoJEO8Ells5jWIRoRwIAJpwefLgH/+lkd1mtWqxBhuS
KLa0bkcS6nIGnjQzNX/XWipu/5tMbBLzbaKw0myodvoK6Yx0MFog1cWf6gLHuvWH
Jy3ONUrF9umHYuOa9sJJtXv/aP7neNJSB3RW67BaiLCLkaetDj9lLciA/KKMvb/I
JNFtuLVTPibZ5iVTjvifFWmJD/Yk0P8mlrH5yfrA3B2EaaWf1es0GWobGIwwLu9s
ZSqjhMDAhfOW2E1sBh7jFRh4lJX1t1jRhyIGx2bOXevPx2hFHq6FSq+yuJ9OsZvO
wC8mC4DD+fovypDWbv3WLslIejM0+THD8KuBQnZtKX5Mbhc+0cELpIFLUdH95TM=
=eMUT
-----END PGP SIGNATURE-----
Merge tag 'net-pull-request' of https://github.com/jasowang/qemu into staging
# -----BEGIN PGP SIGNATURE-----
# Version: GnuPG v1
#
# iQEcBAABAgAGBQJjEaMLAAoJEO8Ells5jWIRoRwIAJpwefLgH/+lkd1mtWqxBhuS
# KLa0bkcS6nIGnjQzNX/XWipu/5tMbBLzbaKw0myodvoK6Yx0MFog1cWf6gLHuvWH
# Jy3ONUrF9umHYuOa9sJJtXv/aP7neNJSB3RW67BaiLCLkaetDj9lLciA/KKMvb/I
# JNFtuLVTPibZ5iVTjvifFWmJD/Yk0P8mlrH5yfrA3B2EaaWf1es0GWobGIwwLu9s
# ZSqjhMDAhfOW2E1sBh7jFRh4lJX1t1jRhyIGx2bOXevPx2hFHq6FSq+yuJ9OsZvO
# wC8mC4DD+fovypDWbv3WLslIejM0+THD8KuBQnZtKX5Mbhc+0cELpIFLUdH95TM=
# =eMUT
# -----END PGP SIGNATURE-----
# gpg: Signature made Fri 02 Sep 2022 02:30:35 EDT
# gpg: using RSA key EF04965B398D6211
# gpg: Good signature from "Jason Wang (Jason Wang on RedHat) <jasowang@redhat.com>" [full]
# Primary key fingerprint: 215D 46F4 8246 689E C77F 3562 EF04 965B 398D 6211
* tag 'net-pull-request' of https://github.com/jasowang/qemu: (21 commits)
net: tulip: Restrict DMA engine to memories
net/colo.c: Fix the pointer issue reported by Coverity.
vdpa: Delete CVQ migration blocker
vdpa: Add virtio-net mac address via CVQ at start
vhost_net: add NetClientState->load() callback
vdpa: extract vhost_vdpa_net_cvq_add from vhost_vdpa_net_handle_ctrl_avail
vdpa: Move command buffers map to start of net device
vdpa: add net_vhost_vdpa_cvq_info NetClientInfo
vhost_net: Add NetClientInfo stop callback
vhost_net: Add NetClientInfo start callback
vhost: Do not depend on !NULL VirtQueueElement on vhost_svq_flush
vhost: Delete useless read memory barrier
vhost: use SVQ element ndescs instead of opaque data for desc validation
vhost: stop transfer elem ownership in vhost_handle_guest_kick
vdpa: Use ring hwaddr at vhost_vdpa_svq_unmap_ring
vhost: Always store new kick fd on vhost_svq_set_svq_kick_fd
vdpa: Make SVQ vring unmapping return void
vdpa: Remove SVQ vring from iova_tree at shutdown
util: accept iova_tree_remove_parameter by value
vdpa: do not save failed dma maps in SVQ iova tree
...
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Support for the unix socket has existed both in BSD and Linux for the
longest time, but not on Windows. Since Windows 10 build 17063 [1],
the native support for the unix socket has come to Windows. Starting
this build, two Win32 processes can use the AF_UNIX address family
over Winsock API to communicate with each other.
[1] https://devblogs.microsoft.com/commandline/af_unix-comes-to-windows/
Signed-off-by: Xuzhou Cheng <xuzhou.cheng@windriver.com>
Signed-off-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20220802075200.907360-3-bmeng.cn@gmail.com>
It's convenient to call iova_tree_remove from a map returned from
iova_tree_find or iova_tree_find_iova. With the current code this is not
possible, since we will free it, and then we will try to search for it
again.
Fix it making accepting the map by value, forcing a copy of the
argument. Not applying a fixes tag, since there is no use like that at
the moment.
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
The last user of this function has just been removed, so we can
drop this function now, too.
Message-Id: <20220810125720.3849835-4-thuth@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Commit 06680b15b4 moved qemu_*_exec_dir() to cutils but forgot
to move the macOS dyld(3) include, resulting in the following
error (when building with Homebrew GCC on macOS Monterey 12.4):
[313/1197] Compiling C object libqemuutil.a.p/util_cutils.c.o
FAILED: libqemuutil.a.p/util_cutils.c.o
../../util/cutils.c:1039:13: error: implicit declaration of function '_NSGetExecutablePath' [-Werror=implicit-function-declaration]
1039 | if (_NSGetExecutablePath(fpath, &len) == 0) {
| ^~~~~~~~~~~~~~~~~~~~
../../util/cutils.c:1039:13: error: nested extern declaration of '_NSGetExecutablePath' [-Werror=nested-externs]
Fix by moving the include line to cutils.
Fixes: 06680b15b4 ("include: move qemu_*_exec_dir() to cutils")
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20220809222046.30812-1-f4bug@amsat.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
close() is a *nix function. It works on any file descriptor, and
sockets in *nix are an example of a file descriptor.
closesocket() is a Windows-specific function, which works only
specifically with sockets. Sockets on Windows do not use *nix-style
file descriptors, and socket() returns a handle to a kernel object
instead, so it must be closed with closesocket().
In QEMU there is already a logic to handle such platform difference
in os-posix.h and os-win32.h, that:
* closesocket maps to close on POSIX
* closesocket maps to a wrapper that calls the real closesocket()
on Windows
Replace the call to close a socket with closesocket() instead.
Signed-off-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
A recent commit moved some Haiku-specific code parts from oslib-posix.c
to cutils.c, but failed to move the corresponding header #include
statement, too, so "make vm-build-haiku.x86_64" is currently broken.
Fix it by moving the header #include, too.
Fixes: 06680b15b4 ("include: move qemu_*_exec_dir() to cutils")
Message-Id: <20220718172026.139004-1-thuth@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Before this change, the directory of the executable was being added to
resolve modules in the build tree. However, get_relocated_path() can now
resolve them with the new bundle mechanism.
Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Message-Id: <20220624145039.49929-5-akihiko.odaki@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Developers often run QEMU without installing. The bundle mechanism
allows to look up files which should be present in installation even in
such a situation.
It is a general mechanism and can find any files in the installation
tree. The build tree will have a new directory, qemu-bundle, to
represent what files the installation tree would have for reference by
the executables.
Note that it abandons compatibility with Windows older than 8. The
extended support for the prior version, 7 ended more than 2 years ago,
and it is unlikely that someone would like to run the latest QEMU on
such an old system.
Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20220624145039.49929-3-akihiko.odaki@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Add new API, to make a time limited call of the coroutine.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@openvz.org>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
It always returns IOVA_OK so nobody uses it.
Acked-by: Jason Wang <jasowang@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20220427154931.3166388-1-eperezma@redhat.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
It seems that aio_wait_kick always required a memory barrier
or atomic operation in the caller, but nobody actually
took care of doing it.
Let's put the barrier in the function instead, and pair it
with another one in AIO_WAIT_WHILE. Read aio_wait_kick()
comment for further explanation.
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Message-Id: <20220524173054.12651-1-eesposit@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
We have too much logic to simply check that bitmaps are of the same
size. Let's just define that hbitmap_merge() and
bdrv_dirty_bitmap_merge_internal() require their argument bitmaps be of
same size, this simplifies things.
Let's look through the callers:
For backup_init_bcs_bitmap() we already assert that merge can't fail.
In bdrv_reclaim_dirty_bitmap_locked() we gracefully handle the error
that can't happen: successor always has same size as its parent, drop
this logic.
In bdrv_merge_dirty_bitmap() we already has assertion and separate
check. Make the check explicit and improve error message.
Signed-off-by: Vladimir Sementsov-Ogievskiy <v.sementsov-og@mail.ru>
Reviewed-by: Nikita Lapshin <nikita.lapshin@virtuozzo.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20220517111206.23585-4-v.sementsov-og@mail.ru>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
On linux, the AT_HWCAP bit PPC_FEATURE_ICACHE_SNOOP indicates
that we can use a simplified 3 instruction flush sequence.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Message-Id: <20220519141131.29839-1-npiggin@gmail.com>
[rth: update after merging cacheflush.c and cacheinfo.c]
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-Id: <20220621014837.189139-4-richard.henderson@linaro.org>
Merge init_ctr_el0 into arch_cache_info. In flush_idcache_range,
use the pre-computed line sizes from the global variables.
Use CONFIG_DARWIN in preference to __APPLE__.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-Id: <20220621014837.189139-3-richard.henderson@linaro.org>
Combine the two files into cacheflush.c. There's a couple of bits
that would be helpful to share between the two, and combining them
seems better than exporting the bits.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-Id: <20220621014837.189139-2-richard.henderson@linaro.org>
This decreases qemu_clock_deadline_ns_all's share from 23.2% to 13% in a
profile of icount-enabled aarch64-softmmu.
Signed-off-by: Idan Horowitz <idan.horowitz@gmail.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20220114004358.299534-2-idan.horowitz@gmail.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Based on already existing QEMU implementation created a signed
256 bit by 128 bit division needed to implement the vector divide
extended signed quadword instruction from PowerISA 3.1
Signed-off-by: Lucas Mateus Castro (alqotel) <lucas.araujo@eldorado.org.br>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20220525134954.85056-6-lucas.araujo@eldorado.org.br>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Based on already existing QEMU implementation, created an unsigned 256
bit by 128 bit division needed to implement the vector divide extended
unsigned instruction from PowerISA3.1
Signed-off-by: Lucas Mateus Castro (alqotel) <lucas.araujo@eldorado.org.br>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20220525134954.85056-5-lucas.araujo@eldorado.org.br>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Extract the knowledge of IEC and SI prefixes out of size_to_str and
freq_to_str, so that it can be reused when printing statistics.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
vCPU execution should be suspended when new BH is scheduled.
This is needed to avoid guest timeouts caused by the long cycles
of the execution. In replay mode execution may hang when
vCPU sleeps and block event comes to the queue.
This patch adds notification which wakes up vCPU or interrupts
execution of guest code.
Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgalyuk@ispras.ru>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
--
v2: changed first_cpu to current_cpu (suggested by Richard Henderson)
v4: moved vCPU notification to aio_bh_enqueue (suggested by Paolo Bonzini)
Message-Id: <165364837317.688121.17680519919871405281.stgit@pasha-ThinkPad-X280>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
SHGetFolderPath() is a deprecated API:
https://docs.microsoft.com/en-us/windows/win32/api/shlobj_core/nf-shlobj_core-shgetfolderpatha
It is a wrapper for SHGetKnownFolderPath() and CSIDL_COMMON_PATH is
mapped to FOLDERID_ProgramData:
https://docs.microsoft.com/en-us/windows/win32/shell/csidl
g_get_system_data_dirs() is a suitable replacement, as it will have
FOLDERID_ProgramData in the returned list. However, it follows the XDG
Base Directory Specification, if `XDG_DATA_DIRS` is defined, it will be
returned instead.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Stefan Weil <sw@weilnetz.de>
Message-Id: <20220525144140.591926-3-marcandre.lureau@redhat.com>
The function is required by get_relocated_path() (already in cutils),
and used by qemu-ga and may be generally useful.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20220525144140.591926-2-marcandre.lureau@redhat.com>
Just setting the max threads to 0 is enough to stop all workers.
Message-Id: <20220514065012.1149539-4-pbonzini@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Since commit f9fc8932b1 ("thread-posix: remove the posix semaphore
support", 2022-04-06) QemuSemaphore has its own mutex and condition
variable; this adds unnecessary overhead on I/O with small block sizes.
Check the QTAILQ directly instead of adding the indirection of a
semaphore's count. Using a semaphore has not been necessary since
qemu_cond_timedwait was introduced; the new code has to be careful about
spurious wakeups but it is simpler, for example thread_pool_cancel does
not have to worry about synchronizing the semaphore count with the number
of elements of pool->request_list.
Note that the return value of qemu_cond_timedwait (0 for timeout, 1 for
signal or spurious wakeup) is different from that of qemu_sem_timedwait
(-1 for timeout, 0 for success).
Reported-by: Lukáš Doktor <ldoktor@redhat.com>
Suggested-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
Message-Id: <20220514065012.1149539-3-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The completion bottom half was scheduled within the pool->lock
critical section. That actually results in worse performance,
because the worker thread can run its own small critical section
and go to sleep before the bottom half starts running.
Note that this simple change does not produce an improvement without
changing the thread pool QemuSemaphore to a condition variable.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
Message-Id: <20220514065012.1149539-2-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
qemu_co_queue_restart_all is basically the same as qemu_co_enter_all
but without a QemuLockable argument. That's perfectly fine, but only as
long as the function is marked coroutine_fn. If used outside coroutine
context, qemu_co_queue_wait will attempt to take the lock and that
is just broken: if you are calling qemu_co_queue_restart_all outside
coroutine context, the lock is going to be a QemuMutex which cannot be
taken twice by the same thread.
The patch adds the marker to qemu_co_queue_restart_all and to its sole
non-coroutine_fn caller; it then reimplements the function in terms of
qemu_co_enter_all_impl, to remove duplicated code and to clarify that the
latter also works in coroutine context.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20220427130830.150180-4-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Because qemu_co_queue_restart_all does not release the lock, it should
be used only in coroutine context. Introduce a new function that,
like qemu_co_enter_next, does release the lock, and use it whenever
qemu_co_queue_restart_all was used outside coroutine context.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20220427130830.150180-3-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
qemu_co_queue_next is basically the same as qemu_co_enter_next but
without a QemuLockable argument. That's perfectly fine, but only
as long as the function is marked coroutine_fn. If used outside
coroutine context, qemu_co_queue_wait will attempt to take the lock
and that is just broken: if you are calling qemu_co_queue_next outside
coroutine context, the lock is going to be a QemuMutex which cannot be
taken twice by the same thread.
The patch adds the marker and reimplements qemu_co_queue_next in terms of
qemu_co_enter_next_impl, to remove duplicated code and to clarify that the
latter also works in coroutine context.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20220427130830.150180-2-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Commit 4c41c69e changed the way the coroutine pool is sized because for
virtio-blk devices with a large queue size and heavy I/O, it was just
too small and caused coroutines to be deleted and reallocated soon
afterwards. The change made the size dynamic based on the number of
queues and the queue size of virtio-blk devices.
There are two important numbers here: Slightly simplified, when a
coroutine terminates, it is generally stored in the global release pool
up to a certain pool size, and if the pool is full, it is freed.
Conversely, when allocating a new coroutine, the coroutines in the
release pool are reused if the pool already has reached a certain
minimum size (the batch size), otherwise we allocate new coroutines.
The problem after commit 4c41c69e is that it not only increases the
maximum pool size (which is the intended effect), but also the batch
size for reusing coroutines (which is a bug). It means that in cases
with many devices and/or a large queue size (which defaults to the
number of vcpus for virtio-blk-pci), many thousand coroutines could be
sitting in the release pool without being reused.
This is not only a waste of memory and allocations, but it actually
makes the QEMU process likely to hit the vm.max_map_count limit on Linux
because each coroutine requires two mappings (its stack and the guard
page for the stack), causing it to abort() in qemu_alloc_stack() because
when the limit is hit, mprotect() starts to fail with ENOMEM.
In order to fix the problem, change the batch size back to 64 to avoid
uselessly accumulating coroutines in the release pool, but keep the
dynamic maximum pool size so that coroutines aren't freed too early
in heavy I/O scenarios.
Note that this fix doesn't strictly make it impossible to hit the limit,
but this would only happen if most of the coroutines are actually in use
at the same time, not just sitting in a pool. This is the same behaviour
as we already had before commit 4c41c69e. Fully preventing this would
require allowing qemu_coroutine_create() to return an error, but it
doesn't seem to be a scenario that people hit in practice.
Cc: qemu-stable@nongnu.org
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=2079938
Fixes: 4c41c69e05
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20220510151020.105528-3-kwolf@redhat.com>
Tested-by: Hiroki Narukawa <hnarukaw@yahoo-corp.jp>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
It's true that these functions currently affect the batch size in which
coroutines are reused (i.e. moved from the global release pool to the
allocation pool of a specific thread), but this is a bug and will be
fixed in a separate patch.
In fact, the comment in the header file already just promises that it
influences the pool size, so reflect this in the name of the functions.
As a nice side effect, the shorter function name makes some line
wrapping unnecessary.
Cc: qemu-stable@nongnu.org
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20220510151020.105528-2-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The thread pool regulates itself: when idle, it kills threads until
empty, when in demand, it creates new threads until full. This behaviour
doesn't play well with latency sensitive workloads where the price of
creating a new thread is too high. For example, when paired with qemu's
'-mlock', or using safety features like SafeStack, creating a new thread
has been measured take multiple milliseconds.
In order to mitigate this let's introduce a new 'EventLoopBase'
property to set the thread pool size. The threads will be created during
the pool's initialization or upon updating the property's value, remain
available during its lifetime regardless of demand, and destroyed upon
freeing it. A properly characterized workload will then be able to
configure the pool to avoid any latency spikes.
Signed-off-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Acked-by: Markus Armbruster <armbru@redhat.com>
Message-id: 20220425075723.20019-4-nsaenzju@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
'event-loop-base' provides basic property handling for all 'AioContext'
based event loops. So let's define a new 'MainLoopClass' that inherits
from it. This will permit tweaking the main loop's properties through
qapi as well as through the command line using the '-object' keyword[1].
Only one instance of 'MainLoopClass' might be created at any time.
'EventLoopBaseClass' learns a new callback, 'can_be_deleted()' so as to
mark 'MainLoop' as non-deletable.
[1] For example:
-object main-loop,id=main-loop,aio-max-batch=<value>
Signed-off-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Acked-by: Markus Armbruster <armbru@redhat.com>
Message-id: 20220425075723.20019-3-nsaenzju@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
David Hildenbrand