WIP dump/load heap

Not needed because a) parallel writes to the same object are forbidden
b) subsequent writes don't depend on previous ones anyway.

.gitea/workflows/buildenv.Dockerfile

@@ -20,7 +20,7 @@ RUN echo 'deb http://deb.debian.org/debian bullseye-backports main' >> /etc/apt/
 
 RUN apt-get update
 RUN apt-get -y install etcd qemu-system-x86 qemu-block-extra qemu-utils fio libasan5 \
-    liburing1 liburing-dev libgoogle-perftools-dev devscripts libjerasure-dev cmake libibverbs-dev libisal-dev
+    libgoogle-perftools-dev devscripts libjerasure-dev cmake libibverbs-dev libisal-dev
 RUN apt-get -y build-dep fio qemu=`dpkg -s qemu-system-x86|grep ^Version:|awk '{print $2}'`
 RUN apt-get update && apt-get -y install jq lp-solve sudo nfs-common fdisk parted
 RUN apt-get --download-only source fio qemu=`dpkg -s qemu-system-x86|grep ^Version:|awk '{print $2}'`

.gitea/workflows/test.yml

@@ -144,6 +144,24 @@ jobs:
           echo ""
         done
 
+  test_change_pg_count_online:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: /root/vitastor/tests/test_change_pg_count_online.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
   test_change_pg_size:
     runs-on: ubuntu-latest
     needs: build
@@ -792,6 +810,24 @@ jobs:
           echo ""
         done
 
+  test_reweight_half:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: /root/vitastor/tests/test_reweight_half.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
   test_heal_csum_32k_dmj:
     runs-on: ubuntu-latest
     needs: build

.gitmodules

@@ -4,3 +4,6 @@
 [submodule "json11"]
 	path = json11
 	url = ../json11.git
+[submodule "emhash"]
+	path = emhash
+	url = ../emhash.git

CMakeLists.txt

@@ -2,6 +2,19 @@ cmake_minimum_required(VERSION 2.8.12)
 
 project(vitastor)
 
-set(VITASTOR_VERSION "2.2.2")
+set(VITASTOR_VERSION "2.3.0")
 
+include(CTest)
+
+add_custom_target(build_tests)
+add_custom_target(test
+	COMMAND
+	echo leak:tcmalloc > ${CMAKE_CURRENT_BINARY_DIR}/lsan-suppress.txt &&
+	env LSAN_OPTIONS=suppressions=${CMAKE_CURRENT_BINARY_DIR}/lsan-suppress.txt ${CMAKE_CTEST_COMMAND}
+)
+# make -j16 -C ../../build test_heap && ../../build/src/test/test_heap
+# make -j16 -C ../../build test_heap && rm -f $(find ../../build -name '*.gcda') && ctest -V -T test -T coverage -R heap --test-dir ../../build && (cd ../../build; gcovr -f ../src --html --html-nested -o coverage/index.html; cd ../src/test)
+# make -j16 -C ../../build test_blockstore && rm -f $(find ../../build -name '*.gcda') && ctest -V -T test -T coverage -R blockstore --test-dir ../../build && (cd ../../build; gcovr -f ../src --html --html-nested -o coverage/index.html; cd ../src/test)
+# kcov --include-path=../../../src ../../kcov ./test_blockstore
+add_dependencies(test build_tests)
 add_subdirectory(src)

README-ru.md

@@ -19,7 +19,7 @@ Vitastor нацелен в первую очередь на SSD и SSD+HDD кл
 TCP и RDMA и на хорошем железе может достигать задержки 4 КБ чтения и записи на уровне ~0.1 мс,
 что примерно в 10 раз быстрее, чем Ceph и другие популярные программные СХД.
 
-Vitastor поддерживает QEMU-драйвер, протоколы NBD и NFS, драйверы OpenStack, OpenNebula, Proxmox, Kubernetes.
+Vitastor поддерживает QEMU-драйвер, протоколы UBLK, NBD и NFS, драйверы OpenStack, OpenNebula, Proxmox, Kubernetes.
 Другие драйверы могут также быть легко реализованы.
 
 Подробности смотрите в документации по ссылкам. Можете начать отсюда: [Быстрый старт](docs/intro/quickstart.ru.md).
@@ -64,8 +64,9 @@ Vitastor поддерживает QEMU-драйвер, протоколы NBD и
   - [vitastor-cli](docs/usage/cli.ru.md) (консольный интерфейс)
   - [vitastor-disk](docs/usage/disk.ru.md) (управление дисками)
   - [fio](docs/usage/fio.ru.md) для тестов производительности
-  - [NBD](docs/usage/nbd.ru.md) для монтирования ядром
-  - [QEMU и qemu-img](docs/usage/qemu.ru.md)
+  - [UBLK](docs/usage/ublk.ru.md) для монтирования ядром
+  - [NBD](docs/usage/nbd.ru.md) - старый интерфейс для монтирования ядром
+  - [QEMU, qemu-img и VDUSE](docs/usage/qemu.ru.md)
   - [NFS](docs/usage/nfs.ru.md) кластерная файловая система и псевдо-ФС прокси
   - [Администрирование](docs/usage/admin.ru.md)
 - Производительность

README.md

@@ -19,7 +19,7 @@ supports TCP and RDMA and may achieve 4 KB read and write latency as low as ~0.1
 with proper hardware which is ~10 times faster than other popular SDS's like Ceph
 or internal systems of public clouds.
 
-Vitastor supports QEMU, NBD, NFS protocols, OpenStack, OpenNebula, Proxmox, Kubernetes drivers.
+Vitastor supports QEMU, UBLK, NBD, NFS protocols, OpenStack, OpenNebula, Proxmox, Kubernetes drivers.
 More drivers may be created easily.
 
 Read more details in the documentation. You can start from here: [Quick Start](docs/intro/quickstart.en.md).
@@ -64,8 +64,9 @@ Read more details in the documentation. You can start from here: [Quick Start](d
   - [vitastor-cli](docs/usage/cli.en.md) (command-line interface)
   - [vitastor-disk](docs/usage/disk.en.md) (disk management tool)
   - [fio](docs/usage/fio.en.md) for benchmarks
-  - [NBD](docs/usage/nbd.en.md) for kernel mounts
-  - [QEMU and qemu-img](docs/usage/qemu.en.md)
+  - [UBLK](docs/usage/ublk.en.md) for kernel mounts
+  - [NBD](docs/usage/nbd.en.md) - old interface for kernel mounts
+  - [QEMU, qemu-img and VDUSE](docs/usage/qemu.en.md)
   - [NFS](docs/usage/nfs.en.md) clustered file system and pseudo-FS proxy
   - [Administration](docs/usage/admin.en.md)
 - Performance

csi/Dockerfile

@@ -36,7 +36,7 @@ RUN (echo deb http://vitastor.io/debian bookworm main > /etc/apt/sources.list.d/
     ((echo 'Package: *'; echo 'Pin: origin "vitastor.io"'; echo 'Pin-Priority: 1000') > /etc/apt/preferences.d/vitastor.pref) && \
     wget -q -O /etc/apt/trusted.gpg.d/vitastor.gpg https://vitastor.io/debian/pubkey.gpg && \
     apt-get update && \
-    apt-get install -y vitastor-client && \
+    apt-get install -y vitastor-client ibverbs-providers && \
     wget https://vitastor.io/archive/qemu/qemu-bookworm-9.2.2%2Bds-1%2Bvitastor4/qemu-utils_9.2.2%2Bds-1%2Bvitastor4_amd64.deb && \
     wget https://vitastor.io/archive/qemu/qemu-bookworm-9.2.2%2Bds-1%2Bvitastor4/qemu-block-extra_9.2.2%2Bds-1%2Bvitastor4_amd64.deb && \
     dpkg -x qemu-utils*.deb tmp1 && \

csi/Makefile

@@ -1,4 +1,4 @@
-VITASTOR_VERSION ?= v2.2.2
+VITASTOR_VERSION ?= v2.3.0
 
 all: build push
 

csi/deploy/004-csi-nodeplugin.yaml

@@ -49,7 +49,7 @@ spec:
             capabilities:
               add: ["SYS_ADMIN"]
             allowPrivilegeEscalation: true
-          image: vitalif/vitastor-csi:v2.2.2
+          image: vitalif/vitastor-csi:v2.3.0
           args:
             - "--node=$(NODE_ID)"
             - "--endpoint=$(CSI_ENDPOINT)"

csi/deploy/007-csi-provisioner.yaml

@@ -121,7 +121,7 @@ spec:
             privileged: true
             capabilities:
               add: ["SYS_ADMIN"]
-          image: vitalif/vitastor-csi:v2.2.2
+          image: vitalif/vitastor-csi:v2.3.0
           args:
             - "--node=$(NODE_ID)"
             - "--endpoint=$(CSI_ENDPOINT)"

csi/src/config.go

@@ -5,7 +5,7 @@ package vitastor
 
 const (
     vitastorCSIDriverName    = "csi.vitastor.io"
-    vitastorCSIDriverVersion = "2.2.2"
+    vitastorCSIDriverVersion = "2.3.0"
 )
 
 // Config struct fills the parameters of request or user input

debian/build-vitastor-bookworm.sh

@@ -1,7 +1,4 @@
 #!/bin/bash
 
-cat < vitastor.Dockerfile > ../Dockerfile
-cd ..
-mkdir -p packages
-sudo podman build --build-arg DISTRO=debian --build-arg REL=bookworm -v `pwd`/packages:/root/packages -f Dockerfile .
-rm Dockerfile
+docker build --build-arg DISTRO=debian --build-arg REL=bookworm -t vitastor-buildenv:bookworm -f vitastor-buildenv.Dockerfile .
+docker run -i --rm -e REL=bookworm -v `dirname $0`/../:/root/vitastor vitastor-buildenv:bookworm /root/vitastor/debian/vitastor-build.sh

debian/build-vitastor-bullseye.sh

@@ -1,7 +1,4 @@
 #!/bin/bash
 
-cat < vitastor.Dockerfile > ../Dockerfile
-cd ..
-mkdir -p packages
-sudo podman build --build-arg DISTRO=debian --build-arg REL=bullseye -v `pwd`/packages:/root/packages -f Dockerfile .
-rm Dockerfile
+docker build --build-arg DISTRO=debian --build-arg REL=bullseye -t vitastor-buildenv:bullseye -f vitastor-buildenv.Dockerfile .
+docker run -i --rm -e REL=bullseye -v `dirname $0`/../:/root/vitastor vitastor-buildenv:bullseye /root/vitastor/debian/vitastor-build.sh

debian/build-vitastor-buster.sh

@@ -1,7 +1,4 @@
 #!/bin/bash
 
-cat < vitastor.Dockerfile > ../Dockerfile
-cd ..
-mkdir -p packages
-sudo podman build --build-arg DISTRO=debian --build-arg REL=buster -v `pwd`/packages:/root/packages -f Dockerfile .
-rm Dockerfile
+docker build --build-arg DISTRO=debian --build-arg REL=buster -t vitastor-buildenv:buster -f vitastor-buildenv.Dockerfile .
+docker run -i --rm -e REL=buster -v `dirname $0`/../:/root/vitastor vitastor-buildenv:buster /root/vitastor/debian/vitastor-build.sh

debian/build-vitastor-trixie.sh

@@ -0,0 +1,4 @@
+#!/bin/bash
+
+docker build --build-arg DISTRO=debian --build-arg REL=trixie -t vitastor-buildenv:trixie -f vitastor-buildenv.Dockerfile .
+docker run -i --rm -e REL=trixie -v `dirname $0`/../:/root/vitastor vitastor-buildenv:trixie /root/vitastor/debian/vitastor-build.sh

debian/build-vitastor-ubuntu-jammy.sh

@@ -1,7 +1,5 @@
 #!/bin/bash
+# Ubuntu 22.04 Jammy Jellyfish
 
-cat < vitastor.Dockerfile > ../Dockerfile
-cd ..
-mkdir -p packages
-sudo podman build --build-arg DISTRO=ubuntu --build-arg REL=jammy -v `pwd`/packages:/root/packages -f Dockerfile .
-rm Dockerfile
+docker build --build-arg DISTRO=ubuntu --build-arg REL=jammy -t vitastor-buildenv:jammy -f vitastor-buildenv.Dockerfile .
+docker run -i --rm -e REL=jammy -v `dirname $0`/../:/root/vitastor vitastor-buildenv:jammy /root/vitastor/debian/vitastor-build.sh

debian/build-vitastor-ubuntu-noble.sh

@@ -0,0 +1,5 @@
+#!/bin/bash
+# 24.04 Noble Numbat
+
+docker build --build-arg DISTRO=ubuntu --build-arg REL=noble -t vitastor-buildenv:noble -f vitastor-buildenv.Dockerfile .
+docker run -i --rm -e REL=noble -v `dirname $0`/../:/root/vitastor vitastor-buildenv:noble /root/vitastor/debian/vitastor-build.sh

debian/changelog

@@ -1,4 +1,4 @@
-vitastor (2.2.2-1) unstable; urgency=medium
+vitastor (2.3.0-1) unstable; urgency=medium
 
   * Bugfixes
 

debian/control

@@ -2,9 +2,9 @@ Source: vitastor
 Section: admin
 Priority: optional
 Maintainer: Vitaliy Filippov <vitalif@yourcmc.ru>
-Build-Depends: debhelper, liburing-dev (>= 0.6), g++ (>= 8), libstdc++6 (>= 8),
+Build-Depends: debhelper, g++ (>= 8), libstdc++6 (>= 8),
   linux-libc-dev, libgoogle-perftools-dev, libjerasure-dev, libgf-complete-dev,
-  libibverbs-dev, libisal-dev, cmake, pkg-config, libnl-3-dev, libnl-genl-3-dev,
+  libibverbs-dev, librdmacm-dev, libisal-dev, cmake, pkg-config, libnl-3-dev, libnl-genl-3-dev,
   node-bindings <!nocheck>, node-gyp, node-nan
 Standards-Version: 4.5.0
 Homepage: https://vitastor.io/

debian/patched-qemu.Dockerfile

@@ -26,7 +26,7 @@ RUN if [ "$REL" = "buster" -o "$REL" = "bullseye" -o "$REL" = "bookworm" ]; then
     echo 'APT::Install-Suggests false;' >> /etc/apt/apt.conf
 
 RUN apt-get update
-RUN DEBIAN_FRONTEND=noninteractive TZ=Europe/Moscow apt-get -y install fio liburing-dev libgoogle-perftools-dev devscripts
+RUN DEBIAN_FRONTEND=noninteractive TZ=Europe/Moscow apt-get -y install fio libgoogle-perftools-dev devscripts
 RUN DEBIAN_FRONTEND=noninteractive TZ=Europe/Moscow apt-get -y build-dep qemu
 # To build a custom version
 #RUN cp /root/packages/qemu-orig/* /root

debian/vitastor-build.sh

@@ -0,0 +1,60 @@
+#!/bin/bash
+# To be ran inside buildenv docker
+
+set -e -x
+
+[ -e /usr/lib/x86_64-linux-gnu/pkgconfig/libisal.pc ] || cp /root/vitastor/debian/libisal.pc /usr/lib/x86_64-linux-gnu/pkgconfig
+
+mkdir -p /root/fio-build/
+cd /root/fio-build/
+rm -rf /root/fio-build/*
+dpkg-source -x /root/fio*.dsc
+
+FULLVER=`head -n1 /root/vitastor/debian/changelog | perl -pe 's/^.*\((.*?)\).*$/$1/'`
+VER=${FULLVER%%-*}
+rm -rf /root/vitastor-$VER
+mkdir /root/vitastor-$VER
+cd /root/vitastor
+cp -a $(ls | grep -v packages) /root/vitastor-$VER
+
+rm -rf /root/vitastor/packages/vitastor-$REL
+mkdir -p /root/vitastor/packages/vitastor-$REL
+mv /root/vitastor-$VER /root/vitastor/packages/vitastor-$REL/
+
+cd /root/vitastor/packages/vitastor-$REL/vitastor-$VER
+
+rm -rf fio
+ln -s /root/fio-build/fio-*/ ./fio
+FIO=`head -n1 fio/debian/changelog | perl -pe 's/^.*\((.*?)\).*$/$1/'`
+ls /usr/include/linux/raw.h || cp ./debian/raw.h /usr/include/linux/raw.h
+sh copy-fio-includes.sh
+rm fio
+mkdir -p a b debian/patches
+mv fio-copy b/fio
+diff -NaurpbB a b > debian/patches/fio-headers.patch || true
+echo fio-headers.patch >> debian/patches/series
+rm -rf a b
+
+echo "dep:fio=$FIO" > debian/fio_version
+
+cd /root/vitastor/packages/vitastor-$REL/vitastor-$VER
+mkdir mon/node_modules
+cd mon/node_modules
+curl -s https://git.yourcmc.ru/vitalif/antietcd/archive/master.tar.gz | tar -zx
+curl -s https://git.yourcmc.ru/vitalif/tinyraft/archive/master.tar.gz | tar -zx
+
+cd /root/vitastor/packages/vitastor-$REL
+tar --sort=name --mtime='2020-01-01' --owner=0 --group=0 --exclude=debian -cJf vitastor_$VER.orig.tar.xz vitastor-$VER
+cd vitastor-$VER
+DEBEMAIL="Vitaliy Filippov <vitalif@yourcmc.ru>" dch -D $REL -v "$FULLVER""$REL" "Rebuild for $REL"
+DEB_BUILD_OPTIONS=nocheck dpkg-buildpackage --jobs=auto -sa
+rm -rf /root/vitastor/packages/vitastor-$REL/vitastor-*/
+
+# Why does ubuntu rename debug packages to *.ddeb?
+cd /root/vitastor/packages/vitastor-$REL
+if ls *.ddeb >/dev/null; then
+    perl -i -pe 's/\.ddeb/.deb/' *.buildinfo *.changes
+    for i in *.ddeb; do
+        mv $i ${i%%.ddeb}.deb
+    done
+fi

debian/vitastor-buildenv.Dockerfile

@@ -0,0 +1,31 @@
+# Build environment for building Vitastor packages for Debian inside a container
+# cd ..
+# docker build --build-arg DISTRO=debian --build-arg REL=bullseye -f debian/vitastor.Dockerfile -t vitastor-buildenv:bullseye .
+# docker run --rm -e REL=bullseye -v ./:/root/vitastor /root/vitastor/debian/vitastor-build.sh
+
+ARG DISTRO=debian
+ARG REL=
+FROM $DISTRO:$REL
+ARG DISTRO=debian
+ARG REL=
+
+WORKDIR /root
+
+RUN set -e -x; \
+    if [ "$REL" = "buster" ]; then \
+        perl -i -pe 's/deb.debian.org/archive.debian.org/' /etc/apt/sources.list; \
+        apt-get update; \
+        apt-get -y install wget; \
+        wget https://vitastor.io/debian/pubkey.gpg -O /etc/apt/trusted.gpg.d/vitastor.gpg; \
+        echo "deb https://vitastor.io/debian $REL main" >> /etc/apt/sources.list; \
+    fi; \
+    grep '^deb ' /etc/apt/sources.list | perl -pe 's/^deb/deb-src/' >> /etc/apt/sources.list; \
+    perl -i -pe 's/Types: deb$/Types: deb deb-src/' /etc/apt/sources.list.d/*.sources || true; \
+    echo 'APT::Install-Recommends false;' >> /etc/apt/apt.conf; \
+    echo 'APT::Install-Suggests false;' >> /etc/apt/apt.conf
+
+RUN apt-get update && \
+    apt-get -y install fio libgoogle-perftools-dev devscripts libjerasure-dev cmake \
+        libibverbs-dev librdmacm-dev libisal-dev libnl-3-dev libnl-genl-3-dev curl nodejs npm node-nan node-bindings && \
+    apt-get -y build-dep fio && \
+    apt-get --download-only source fio

debian/vitastor-client.install

@@ -2,6 +2,7 @@ usr/bin/vita
 usr/bin/vitastor-cli
 usr/bin/vitastor-rm
 usr/bin/vitastor-nbd
+usr/bin/vitastor-ublk
 usr/bin/vitastor-nfs
 usr/bin/vitastor-kv
 usr/bin/vitastor-kv-stress

debian/vitastor.Dockerfile

@@ -1,65 +0,0 @@
-# Build Vitastor packages for Debian inside a container
-# cd ..; podman build --build-arg DISTRO=debian --build-arg REL=bullseye -v `pwd`/packages:/root/packages -f debian/vitastor.Dockerfile .
-
-ARG DISTRO=debian
-ARG REL=
-FROM $DISTRO:$REL
-ARG DISTRO=debian
-ARG REL=
-
-WORKDIR /root
-
-RUN set -e -x; \
-    if [ "$REL" = "buster" ]; then \
-        apt-get update; \
-        apt-get -y install wget; \
-        wget https://vitastor.io/debian/pubkey.gpg -O /etc/apt/trusted.gpg.d/vitastor.gpg; \
-        echo "deb https://vitastor.io/debian $REL main" >> /etc/apt/sources.list; \
-    fi; \
-    grep '^deb ' /etc/apt/sources.list | perl -pe 's/^deb/deb-src/' >> /etc/apt/sources.list; \
-    perl -i -pe 's/Types: deb$/Types: deb deb-src/' /etc/apt/sources.list.d/debian.sources || true; \
-    echo 'APT::Install-Recommends false;' >> /etc/apt/apt.conf; \
-    echo 'APT::Install-Suggests false;' >> /etc/apt/apt.conf
-
-RUN apt-get update && \
-    apt-get -y install fio liburing-dev libgoogle-perftools-dev devscripts libjerasure-dev cmake \
-        libibverbs-dev librdmacm-dev libisal-dev libnl-3-dev libnl-genl-3-dev curl nodejs npm node-nan node-bindings && \
-    apt-get -y build-dep fio && \
-    apt-get --download-only source fio
-
-ADD . /root/vitastor
-RUN set -e -x; \
-    [ -e /usr/lib/x86_64-linux-gnu/pkgconfig/libisal.pc ] || cp /root/vitastor/debian/libisal.pc /usr/lib/x86_64-linux-gnu/pkgconfig; \
-    mkdir -p /root/fio-build/; \
-    cd /root/fio-build/; \
-    rm -rf /root/fio-build/*; \
-    dpkg-source -x /root/fio*.dsc; \
-    mkdir -p /root/packages/vitastor-$REL; \
-    rm -rf /root/packages/vitastor-$REL/*; \
-    cd /root/packages/vitastor-$REL; \
-    FULLVER=$(head -n1 /root/vitastor/debian/changelog | perl -pe 's/^.*\((.*?)\).*$/$1/'); \
-    VER=${FULLVER%%-*}; \
-    cp -r /root/vitastor vitastor-$VER; \
-    cd vitastor-$VER; \
-    ln -s /root/fio-build/fio-*/ ./fio; \
-    FIO=$(head -n1 fio/debian/changelog | perl -pe 's/^.*\((.*?)\).*$/$1/'); \
-    ls /usr/include/linux/raw.h || cp ./debian/raw.h /usr/include/linux/raw.h; \
-    sh copy-fio-includes.sh; \
-    rm fio; \
-    mkdir -p a b debian/patches; \
-    mv fio-copy b/fio; \
-    diff -NaurpbB a b > debian/patches/fio-headers.patch || true; \
-    echo fio-headers.patch >> debian/patches/series; \
-    rm -rf a b; \
-    echo "dep:fio=$FIO" > debian/fio_version; \
-    cd /root/packages/vitastor-$REL/vitastor-$VER; \
-    mkdir mon/node_modules; \
-    cd mon/node_modules; \
-    curl -s https://git.yourcmc.ru/vitalif/antietcd/archive/master.tar.gz | tar -zx; \
-    curl -s https://git.yourcmc.ru/vitalif/tinyraft/archive/master.tar.gz | tar -zx; \
-    cd /root/packages/vitastor-$REL; \
-    tar --sort=name --mtime='2020-01-01' --owner=0 --group=0 --exclude=debian -cJf vitastor_$VER.orig.tar.xz vitastor-$VER; \
-    cd vitastor-$VER; \
-    DEBFULLNAME="Vitaliy Filippov <vitalif@yourcmc.ru>" dch -D $REL -v "$FULLVER""$REL" "Rebuild for $REL"; \
-    DEB_BUILD_OPTIONS=nocheck dpkg-buildpackage --jobs=auto -sa; \
-    rm -rf /root/packages/vitastor-$REL/vitastor-*/

docker/Dockerfile

@@ -3,7 +3,7 @@
 FROM debian:bookworm
 
 ADD etc/apt /etc/apt/
-RUN apt-get update && apt-get -y install vitastor udev systemd qemu-system-x86 qemu-system-common qemu-block-extra qemu-utils jq nfs-common && apt-get clean
+RUN apt-get update && apt-get -y install vitastor ibverbs-providers udev systemd qemu-system-x86 qemu-system-common qemu-block-extra qemu-utils jq nfs-common && apt-get clean
 ADD sleep.sh /usr/bin/
 ADD install.sh /usr/bin/
 ADD scripts /opt/scripts/

docker/Makefile

@@ -1,4 +1,4 @@
-VITASTOR_VERSION ?= v2.2.2
+VITASTOR_VERSION ?= v2.3.0
 
 all: build push
 

docker/etc/apt/preferences

@@ -0,0 +1,3 @@
+Package: *
+Pin: release n=bookworm-backports
+Pin-Priority: 500

docker/etc/vitastor/docker.conf

@@ -4,7 +4,7 @@
 #
 
 # Desired Vitastor version
-VITASTOR_VERSION=v2.2.2
+VITASTOR_VERSION=v2.3.0
 
 # Additional arguments for all containers
 # For example, you may want to specify a custom logging driver here

docs/config/client.en.md

@@ -25,6 +25,9 @@ affect their interaction with the cluster.
 - [nbd_max_part](#nbd_max_part)
 - [osd_nearfull_ratio](#osd_nearfull_ratio)
 - [hostname](#hostname)
+- [ublk_queue_depth](#ublk_queue_depth)
+- [ublk_max_io_size](#ublk_max_io_size)
+- [qemu_file_mirror_path](#qemu_file_mirror_path)
 
 ## client_iothread_count
 
@@ -225,3 +228,28 @@ without destroying and recreating OSDs.
 Clients use host name to find their distance to OSDs when [localized reads](pool.en.md#local_reads)
 are enabled. By default, standard [gethostname](https://man7.org/linux/man-pages/man2/gethostname.2.html)
 function is used to determine host name, but you can also override it with this parameter.
+
+## ublk_queue_depth
+
+- Type: integer
+- Default: 256
+
+Default queue depth for [Vitastor ublk servers](../usage/ublk.en.md).
+
+## ublk_max_io_size
+
+- Type: integer
+
+Default maximum I/O size for Vitastor [ublk servers](../usage/ublk.en.md).
+The largest of 1 MB and pool block size multiplied by EC data chunk count is used if not specified.
+
+## qemu_file_mirror_path
+
+- Type: string
+
+When set to an FS directory path (for example, `/mnt/vitastor/`), `qemu-img info` and similar
+QAPI commands return the name of the image inside this directory instead of normal
+`vitastor://?image=abc` URI as `filename`.
+
+This allows to then mount this path using [vitastor-nfs](../usage/nfs.en.md) and trick
+third-party systems like Veeam which rely on `filename` in the image info but don't support Vitastor.

docs/config/client.ru.md

@@ -25,6 +25,9 @@
 - [nbd_max_part](#nbd_max_part)
 - [osd_nearfull_ratio](#osd_nearfull_ratio)
 - [hostname](#hostname)
+- [ublk_queue_depth](#ublk_queue_depth)
+- [ublk_max_io_size](#ublk_max_io_size)
+- [qemu_file_mirror_path](#qemu_file_mirror_path)
 
 ## client_iothread_count
 
@@ -230,3 +233,30 @@ RDMA и хотите повысить пиковую производитель
 [локальные чтения](pool.ru.md#local_reads). По умолчанию для определения имени
 хоста используется стандартная функция [gethostname](https://man7.org/linux/man-pages/man2/gethostname.2.html),
 но вы также можете задать имя хоста вручную данным параметром.
+
+## ublk_queue_depth
+
+- Тип: целое число
+- Значение по умолчанию: 256
+
+Глубина очереди по умолчанию для [ublk-серверов Vitastor](../usage/ublk.ru.md).
+
+## ublk_max_io_size
+
+- Тип: целое число
+
+Максимальный размер запроса ввода-вывода для [ublk-серверов Vitastor](../usage/ublk.ru.md).
+Если не задан, используется максимум из 1 МБ и размера блока пула, умноженного на число частей
+данных EC-пула.
+
+## qemu_file_mirror_path
+
+- Тип: строка
+
+Если установить эту опцию равной пути к каталогу в ФС, команда `qemu-img info` и подобные
+команды QAPI будут возвращать в поле `filename` имя образа внутри заданного каталога вместо
+обычного адреса типа `vitastor://?image=abc`.
+
+Это позволяет смонтировать этот путь с помощью [vitastor-nfs](../usage/nfs.ru.md) и обмануть
+сторонние системы типа Veeam, которые полагаются на поле `filename` в информации об образе QEMU,
+но не поддерживают Vitastor.

docs/config/network.en.md

@@ -74,7 +74,7 @@ Consider `use_rdmacm` for such networks.
 ## use_rdmacm
 
 - Type: boolean
-- Default: true
+- Default: false
 
 Use an alternative implementation of RDMA through RDMA-CM (Connection
 Manager). Works with all RDMA networks: Infiniband, iWARP and

docs/config/network.ru.md

@@ -74,7 +74,7 @@ RDMA-устройства, но они не имеют соединения с
 ## use_rdmacm
 
 - Тип: булево (да/нет)
-- Значение по умолчанию: true
+- Значение по умолчанию: false
 
 Использовать альтернативную реализацию RDMA на основе RDMA-CM (Connection
 Manager). Работает со всеми типами RDMA-сетей: Infiniband, iWARP и

docs/config/osd.en.md

@@ -491,7 +491,7 @@ Can be used to slow down scrubbing if it affects user load too much.
 ## scrub_list_limit
 
 - Type: integer
-- Default: 1000
+- Default: 262144
 - Can be changed online: yes
 
 Number of objects to list in one listing operation during scrub.

docs/config/osd.ru.md

@@ -514,7 +514,7 @@ fsync небезопасным даже с режимом "directsync".
 ## scrub_list_limit
 
 - Тип: целое число
-- Значение по умолчанию: 1000
+- Значение по умолчанию: 262144
 - Можно менять на лету: да
 
 Размер загружаемых за одну операцию списков объектов в процессе фоновой

docs/config/src/client.yml

@@ -283,3 +283,36 @@
     [локальные чтения](pool.ru.md#local_reads). По умолчанию для определения имени
     хоста используется стандартная функция [gethostname](https://man7.org/linux/man-pages/man2/gethostname.2.html),
     но вы также можете задать имя хоста вручную данным параметром.
+- name: ublk_queue_depth
+  type: int
+  default: 256
+  online: false
+  info: Default queue depth for [Vitastor ublk servers](../usage/ublk.en.md).
+  info_ru: Глубина очереди по умолчанию для [ublk-серверов Vitastor](../usage/ublk.ru.md).
+- name: ublk_max_io_size
+  type: int
+  online: false
+  info: |
+    Default maximum I/O size for Vitastor [ublk servers](../usage/ublk.en.md).
+    The largest of 1 MB and pool block size multiplied by EC data chunk count is used if not specified.
+  info_ru: |
+    Максимальный размер запроса ввода-вывода для [ublk-серверов Vitastor](../usage/ublk.ru.md).
+    Если не задан, используется максимум из 1 МБ и размера блока пула, умноженного на число частей
+    данных EC-пула.
+- name: qemu_file_mirror_path
+  type: string
+  info: |
+    When set to an FS directory path (for example, `/mnt/vitastor/`), `qemu-img info` and similar
+    QAPI commands return the name of the image inside this directory instead of normal
+    `vitastor://?image=abc` URI as `filename`.
+
+    This allows to then mount this path using [vitastor-nfs](../usage/nfs.en.md) and trick
+    third-party systems like Veeam which rely on `filename` in the image info but don't support Vitastor.
+  info_ru: |
+    Если установить эту опцию равной пути к каталогу в ФС, команда `qemu-img info` и подобные
+    команды QAPI будут возвращать в поле `filename` имя образа внутри заданного каталога вместо
+    обычного адреса типа `vitastor://?image=abc`.
+
+    Это позволяет смонтировать этот путь с помощью [vitastor-nfs](../usage/nfs.ru.md) и обмануть
+    сторонние системы типа Veeam, которые полагаются на поле `filename` в информации об образе QEMU,
+    но не поддерживают Vitastor.

docs/config/src/included.en.md

@@ -24,6 +24,8 @@
 
 {{../../installation/kubernetes.en.md}}
 
+{{../../installation/s3.en.md}}
+
 {{../../installation/source.en.md}}
 
 {{../../config.en.md|indent=1}}
@@ -54,6 +56,8 @@
 
 {{../../usage/fio.en.md}}
 
+{{../../usage/ublk.en.md}}
+
 {{../../usage/nbd.en.md}}
 
 {{../../usage/qemu.en.md}}

docs/config/src/included.ru.md

@@ -26,6 +26,8 @@
 
 {{../../installation/source.ru.md}}
 
+{{../../installation/s3.ru.md}}
+
 {{../../config.ru.md|indent=1}}
 
 {{../../config/common.ru.md|indent=2}}
@@ -54,6 +56,8 @@
 
 {{../../usage/fio.ru.md}}
 
+{{../../usage/ublk.ru.md}}
+
 {{../../usage/nbd.ru.md}}
 
 {{../../usage/qemu.ru.md}}

docs/config/src/network.yml

@@ -51,7 +51,7 @@
     Рассмотрите включение `use_rdmacm` для таких сетей.
 - name: use_rdmacm
   type: bool
-  default: true
+  default: false
   info: |
     Use an alternative implementation of RDMA through RDMA-CM (Connection
     Manager). Works with all RDMA networks: Infiniband, iWARP and

docs/config/src/osd.yml

@@ -566,7 +566,7 @@
     сильно влияет на пользовательскую нагрузку.
 - name: scrub_list_limit
   type: int
-  default: 1000
+  default: 262144
   online: true
   info: |
     Number of objects to list in one listing operation during scrub.

docs/installation/docker.en.md

@@ -26,9 +26,9 @@ at Vitastor Kubernetes operator: https://github.com/Antilles7227/vitastor-operat
 The instruction is very simple.
 
 1. Download a Docker image of the desired version: \
-   `docker pull vitastor:v2.2.2`
+   `docker pull vitalif/vitastor:v2.3.0`
 2. Install scripts to the host system: \
-   `docker run --rm -it -v /etc:/host-etc -v /usr/bin:/host-bin vitastor:v2.2.2 install.sh`
+   `docker run --rm -it -v /etc:/host-etc -v /usr/bin:/host-bin vitalif/vitastor:v2.3.0 install.sh`
 3. Reload udev rules: \
    `udevadm control --reload-rules`
 

docs/installation/docker.ru.md

@@ -25,9 +25,9 @@ Vitastor можно установить в Docker/Podman. При этом etcd,
 Инструкция по установке максимально простая.
 
 1. Скачайте Docker-образ желаемой версии: \
-   `docker pull vitastor:v2.2.2`
+   `docker pull vitalif/vitastor:v2.3.0`
 2. Установите скрипты в хост-систему командой: \
-   `docker run --rm -it -v /etc:/host-etc -v /usr/bin:/host-bin vitastor:v2.2.2 install.sh`
+   `docker run --rm -it -v /etc:/host-etc -v /usr/bin:/host-bin vitalif/vitastor:v2.3.0 install.sh`
 3. Перезагрузите правила udev: \
    `udevadm control --reload-rules`
 

docs/installation/packages.en.md

@@ -11,12 +11,20 @@
 - Trust Vitastor package signing key:
   `wget https://vitastor.io/debian/pubkey.gpg -O /etc/apt/trusted.gpg.d/vitastor.gpg`
 - Add Vitastor package repository to your /etc/apt/sources.list:
-  - Debian 12 (Bookworm/Sid): `deb https://vitastor.io/debian bookworm main`
+  - Debian 13 (Trixie/Sid): `deb https://vitastor.io/debian trixie main`
+  - Debian 12 (Bookworm): `deb https://vitastor.io/debian bookworm main`
   - Debian 11 (Bullseye): `deb https://vitastor.io/debian bullseye main`
   - Debian 10 (Buster): `deb https://vitastor.io/debian buster main`
   - Ubuntu 22.04 (Jammy): `deb https://vitastor.io/debian jammy main`
+  - Ubuntu 24.04 (Noble): `deb https://vitastor.io/debian noble main`
   - Add `-oldstable` to bookworm/bullseye/buster in this line to install the last
     stable version from 0.9.x branch instead of 1.x
+- To always prefer vitastor-patched QEMU and Libvirt versions, add the following to `/etc/apt/preferences`:
+  ```
+  Package: *
+  Pin: origin "vitastor.io"
+  Pin-Priority: 501
+  ```
 - Install packages: `apt update; apt install vitastor lp-solve etcd linux-image-amd64 qemu-system-x86`
 
 ## CentOS
@@ -42,7 +50,6 @@
   recommended because io_uring is a relatively new technology and there is
   at least one bug which reproduces with io_uring and HP SmartArray
   controllers in 5.4
-- liburing 0.4 or newer
 - lp_solve
 - etcd 3.4.15 or newer. Earlier versions won't work because of various bugs,
   for example [#12402](https://github.com/etcd-io/etcd/pull/12402).

docs/installation/packages.ru.md

@@ -11,12 +11,20 @@
 - Добавьте ключ репозитория Vitastor:
   `wget https://vitastor.io/debian/pubkey.gpg -O /etc/apt/trusted.gpg.d/vitastor.gpg`
 - Добавьте репозиторий Vitastor в /etc/apt/sources.list:
-  - Debian 12 (Bookworm/Sid): `deb https://vitastor.io/debian bookworm main`
+  - Debian 13 (Trixie/Sid): `deb https://vitastor.io/debian trixie main`
+  - Debian 12 (Bookworm): `deb https://vitastor.io/debian bookworm main`
   - Debian 11 (Bullseye): `deb https://vitastor.io/debian bullseye main`
   - Debian 10 (Buster): `deb https://vitastor.io/debian buster main`
   - Ubuntu 22.04 (Jammy): `deb https://vitastor.io/debian jammy main`
+  - Ubuntu 24.04 (Noble): `deb https://vitastor.io/debian noble main`
   - Добавьте `-oldstable` к слову bookworm/bullseye/buster в этой строке, чтобы
     установить последнюю стабильную версию из ветки 0.9.x вместо 1.x
+- Чтобы всегда предпочитались версии пакетов QEMU и Libvirt с патчами Vitastor, добавьте в `/etc/apt/preferences`:
+  ```
+  Package: *
+  Pin: origin "vitastor.io"
+  Pin-Priority: 501
+  ```
 - Установите пакеты: `apt update; apt install vitastor lp-solve etcd linux-image-amd64 qemu-system-x86`
 
 ## CentOS
@@ -41,7 +49,6 @@
 - Ядро Linux 5.4 или новее, для поддержки io_uring. Рекомендуется даже 5.8,
   так как io_uring - относительно новый интерфейс и в версиях до 5.8 встречались
   некоторые баги, например, зависание с io_uring и контроллером HP SmartArray
-- liburing 0.4 или новее
 - lp_solve
 - etcd 3.4.15 или новее. Более старые версии не будут работать из-за разных багов,
   например, [#12402](https://github.com/etcd-io/etcd/pull/12402).

docs/installation/proxmox.en.md

@@ -9,7 +9,7 @@
 To enable Vitastor support in Proxmox Virtual Environment (6.4-8.x are supported):
 
 - Add the corresponding Vitastor Debian repository into sources.list on Proxmox hosts:
-  bookworm for 8.1+, pve8.0 for 8.0, bullseye for 7.4, pve7.3 for 7.3, pve7.2 for 7.2, pve7.1 for 7.1, buster for 6.4
+  trixie for 9.0+, bookworm for 8.1+, pve8.0 for 8.0, bullseye for 7.4, pve7.3 for 7.3, pve7.2 for 7.2, pve7.1 for 7.1, buster for 6.4
 - Install vitastor-client, pve-qemu-kvm, pve-storage-vitastor (* or see note) packages from Vitastor repository
 - Define storage in `/etc/pve/storage.cfg` (see below)
 - Block network access from VMs to Vitastor network (to OSDs and etcd),

docs/installation/proxmox.ru.md

@@ -9,7 +9,7 @@
 Чтобы подключить Vitastor к Proxmox Virtual Environment (поддерживаются версии 6.4-8.x):
 
 - Добавьте соответствующий Debian-репозиторий Vitastor в sources.list на хостах Proxmox:
-  bookworm для 8.1+, pve8.0 для 8.0, bullseye для 7.4, pve7.3 для 7.3, pve7.2 для 7.2, pve7.1 для 7.1, buster для 6.4
+  trixie для 9.0+, bookworm для 8.1+, pve8.0 для 8.0, bullseye для 7.4, pve7.3 для 7.3, pve7.2 для 7.2, pve7.1 для 7.1, buster для 6.4
 - Установите пакеты vitastor-client, pve-qemu-kvm, pve-storage-vitastor (* или см. сноску) из репозитория Vitastor
 - Определите тип хранилища в `/etc/pve/storage.cfg` (см. ниже)
 - Обязательно заблокируйте доступ от виртуальных машин к сети Vitastor (OSD и etcd), т.к. Vitastor (пока) не поддерживает аутентификацию

docs/installation/source.en.md

@@ -15,7 +15,7 @@
 - gcc and g++ 8 or newer, clang 10 or newer, or other compiler with C++11 plus
   designated initializers support from C++20
 - CMake
-- liburing, jerasure headers and libraries
+- jerasure headers and libraries
 - ISA-L, libibverbs and librdmacm headers and libraries (optional)
 - tcmalloc (google-perftools-dev)
 

docs/installation/source.ru.md

@@ -15,7 +15,7 @@
 - gcc и g++ >= 8, либо clang >= 10, либо другой компилятор с поддержкой C++11 плюс
   назначенных инициализаторов (designated initializers) из C++20
 - CMake
-- Заголовки и библиотеки liburing, jerasure
+- Заголовки и библиотеки jerasure
 - Опционально - заголовки и библиотеки ISA-L, libibverbs, librdmacm
 - tcmalloc (google-perftools-dev)
 

docs/intro/features.en.md

@@ -52,7 +52,7 @@
 - Generic user-space client library
 - [Native QEMU driver](../usage/qemu.en.md)
 - [Loadable fio engine for benchmarks](../usage/fio.en.md)
-- [NBD proxy for kernel mounts](../usage/nbd.en.md)
+- [UBLK](../usage/ublk.en.md) and [NBD](../usage/nbd.en.md) servers for kernel mounts
 - [Simplified NFS proxy for file-based image access emulation (suitable for VMWare)](../usage/nfs.en.md#pseudo-fs)
 
 ## Roadmap

docs/intro/features.ru.md

@@ -54,7 +54,7 @@
 - Общая пользовательская клиентская библиотека для работы с кластером
 - [Драйвер диска для QEMU](../usage/qemu.ru.md)
 - [Драйвер диска для утилиты тестирования производительности fio](../usage/fio.ru.md)
-- [NBD-прокси для монтирования образов ядром](../usage/nbd.ru.md) ("блочное устройство в режиме пользователя")
+- [UBLK](../usage/ublk.ru.md) и [NBD](../usage/nbd.ru.md) серверы для монтирования образов ядром ("блочное устройство в режиме пользователя")
 - [Упрощённая NFS-прокси для эмуляции файлового доступа к образам (подходит для VMWare)](../usage/nfs.ru.md#псевдо-фс)
 
 ## Планы развития

docs/usage/disk.en.md

@@ -73,6 +73,8 @@ Options (automatic mode):
 --max_other 10%
   Use disks for OSD data even if they already have non-Vitastor partitions,
   but only if these take up no more than this percent of disk space.
+--dry-run
+  Check and print new OSD count for each disk but do not actually create them.
 ```
 
 Options (single-device mode):

docs/usage/disk.ru.md

@@ -74,6 +74,8 @@ vitastor-disk - инструмент командной строки для уп
 --max_other 10%
   Использовать диски под данные OSD, даже если на них уже есть не-Vitastor-овые
   разделы, но только в случае, если они занимают не более данного процента диска.
+--dry-run
+  Проверить и вывести число новых OSD для каждого диска, но не создавать их.
 ```
 
 Опции для режима одного OSD:

docs/usage/nfs.en.md

@@ -89,6 +89,8 @@ POSIX features currently not implemented in VitastorFS:
   instead of actually allocated space
 - Access times (`atime`) are not tracked (like `-o noatime`)
 - Modification time (`mtime`) is updated lazily every second (like `-o lazytime`)
+- Permission enforcement is disabled by default (and Linux NFS client doesn't
+  enforce them too). Use `--enforce 1` to enable it.
 
 Other notable missing features which should be addressed in the future:
 - Inode ID reuse. Currently inode IDs always grow, the limit is 2^48 inodes, so
@@ -258,4 +260,5 @@ Options:
 | `--nfspath <PATH>` | set NFS export path to \<PATH> (default is /)            |
 | `--pidfile <FILE>` | write process ID to the specified file                   |
 | `--logfile <FILE>` | log to the specified file                                |
+| `--enforce 1`      | enforce permissions at the server side (no by default)   |
 | `--foreground 1`   | stay in foreground, do not daemonize                     |

docs/usage/nfs.ru.md

@@ -91,6 +91,8 @@ JSON-формате :-). Для инспекции содержимого БД
   stat(2), так что `du` всегда показывает сумму размеров файлов, а не фактически занятое место
 - Времена доступа (`atime`) не отслеживаются (как будто ФС смонтирована с `-o noatime`)
 - Времена модификации (`mtime`) отслеживаются асинхронно (как будто ФС смонтирована с `-o lazytime`)
+- Привилегии доступа по умолчанию не проверяются сервером (клиент NFS Linux их также не проверяет).
+  Чтобы включить проверки, используйте опцию `--enforce 1`.
 
 Другие недостающие функции, которые нужно добавить в будущем:
 - Переиспользование номеров инодов. В текущей реализации номера инодов всё время
@@ -270,4 +272,5 @@ VitastorFS из GPUDirect.
 | `--nfspath <PATH>` | установить путь NFS-экспорта в \<PATH> (по умолчанию /) |
 | `--pidfile <FILE>` | записать ID процесса в заданный файл                    |
 | `--logfile <FILE>` | записывать логи в заданный файл                         |
+| `--enforce 1`      | проверять права доступа на стороне сервера (по умолчанию нет) |
 | `--foreground 1`   | не уходить в фон после запуска                          |

docs/usage/qemu.en.md

@@ -130,23 +130,16 @@ Linux kernel, starting with version 5.15, supports a new interface for attaching
 to the host - VDUSE (vDPA Device in Userspace). QEMU, starting with 7.2, has support for
 exporting QEMU block devices over this protocol using qemu-storage-daemon.
 
-VDUSE is currently the best interface to attach Vitastor disks as kernel devices because:
-- It avoids data copies and thus achieves much better performance than [NBD](nbd.en.md)
-- It doesn't have NBD timeout problem - the device doesn't die if an operation executes for too long
+VDUSE advantages:
+
+- VDUSE copies memory 1 time instead of 2, and is thus faster than [NBD](nbd.en.md) for linear read/write.
+- It doesn't have NBD timeout problem - the device doesn't die if an operation executes for too long.
 - It doesn't have hung device problem - if the userspace process dies it can be restarted (!)
-  and block device will continue operation
-- It doesn't seem to have the device number limit
+  and block device will continue operation (UBLK can do it too).
+- It doesn't seem to have the device number limit (UBLK also doesn't).
 
-Example performance comparison:
-
-|                      | direct fio  | NBD         | VDUSE       |
-|----------------------|-------------|-------------|-------------|
-| linear write         | 3.85 GB/s   | 1.12 GB/s   | 3.85 GB/s   |
-| 4k random write Q128 | 240000 iops | 120000 iops | 178000 iops |
-| 4k random write Q1   | 9500 iops   | 7620 iops   | 7640 iops   |
-| linear read          | 4.3 GB/s    | 1.8 GB/s    | 2.85 GB/s   |
-| 4k random read Q128  | 287000 iops | 140000 iops | 189000 iops |
-| 4k random read Q1    | 9600 iops   | 7640 iops   | 7780 iops   |
+At the same time, VDUSE may be slower or faster than [UBLK](ublk.en.md) for linear read/write,
+and iops-wise it's sometimes even slower than NBD. See performance comparison examples at the page [UBLK](ublk.en.md).
 
 To try VDUSE you need at least Linux 5.15, built with VDUSE support
 (CONFIG_VDPA=m, CONFIG_VDPA_USER=m, CONFIG_VIRTIO_VDPA=m).
@@ -193,3 +186,12 @@ To remove the device:
 vdpa dev del test1
 kill <qemu-storage-daemon_process_PID>
 ```
+
+## Veeam
+
+Vitastor QEMU driver has a feature that allows to trick third-party systems like Veeam not able to parse qemu-img
+vitastor URIs: [qemu_file_mirror_path](../config/client.en.md#qemu_file_mirror_path).
+
+To make such systems work, you should set this option to an FS directory path (for example, `/mnt/vitastor/`) and
+mount this directory using [`vitastor-nfs mount --block`](../usage/nfs.en.md). It will make them access
+your images using files and, hopefully, succeed in doing their normal job :).

docs/usage/qemu.ru.md

@@ -132,24 +132,16 @@ qemu-system-x86_64 -enable-kvm -m 2048 -M accel=kvm,memory-backend=mem \
 к системе - VDUSE (vDPA Device in Userspace), а в QEMU, начиная с версии 7.2, есть поддержка
 экспорта блочных устройств QEMU по этому протоколу через qemu-storage-daemon.
 
-VDUSE - на данный момент лучший интерфейс для подключения дисков Vitastor в виде блочных
-устройств на уровне ядра, ибо:
-- VDUSE не копирует данные и поэтому достигает значительно лучшей производительности, чем [NBD](nbd.ru.md)
-- Также оно не имеет проблемы NBD-таймаута - устройство не умирает, если операция выполняется слишком долго
-- Также оно не имеет проблемы подвисающих устройств - если процесс-обработчик умирает, его можно
-  перезапустить (!) и блочное устройство продолжит работать
-- По-видимому, у него нет предела числа подключаемых в систему устройств
+Преимущества VDUSE:
 
-Пример сравнения производительности:
+- VDUSE копирует данные 1 раз, а не 2, и поэтому он быстрее, чем [NBD](nbd.ru.md) при линейном доступе.
+- VDUSE не имеет проблемы NBD-таймаута - устройство не умирает, если операция выполняется слишком долго.
+- VDUSE не имеет проблемы подвисающих устройств - если процесс-обработчик умирает, его можно
+  перезапустить (!) и блочное устройство продолжит работать (в UBLK это тоже поддерживается).
+- По-видимому, у него нет предела числа подключаемых в систему устройств (в UBLK лимита тоже нет).
 
-|                          | Прямой fio  | NBD         | VDUSE       |
-|--------------------------|-------------|-------------|-------------|
-| линейная запись          | 3.85 GB/s   | 1.12 GB/s   | 3.85 GB/s   |
-| 4k случайная запись Q128 | 240000 iops | 120000 iops | 178000 iops |
-| 4k случайная запись Q1   | 9500 iops   | 7620 iops   | 7640 iops   |
-| линейное чтение          | 4.3 GB/s    | 1.8 GB/s    | 2.85 GB/s   |
-| 4k случайное чтение Q128 | 287000 iops | 140000 iops | 189000 iops |
-| 4k случайное чтение Q1   | 9600 iops   | 7640 iops   | 7780 iops   |
+Однако, при линейном доступе VDUSE может быть медленнее UBLK (а может быть и быстрее), а по iops
+VDUSE иногда даже медленнее NBD. Пример сравнения производительности смотрите на странице [UBLK](ublk.ru.md).
 
 Чтобы попробовать VDUSE, вам нужно ядро Linux как минимум версии 5.15, собранное с поддержкой
 VDUSE (CONFIG_VDPA=m, CONFIG_VDPA_USER=m, CONFIG_VIRTIO_VDPA=m).
@@ -196,3 +188,12 @@ vdpa dev add name test1 mgmtdev vduse
 vdpa dev del test1
 kill <PID_процесса_qemu-storage-daemon>
 ```
+
+## Veeam
+
+Драйвер Vitastor QEMU имеет функцию, которая позволяет обманывать сторонние системы типа Veeam, которые
+не могут сами по себе разобрать адреса дисков в vitastor: [qemu_file_mirror_path](../config/client.ru.md#qemu_file_mirror_path).
+
+Чтобы заставить такие системы работать, вам нужно установить эту опцию равной пути к некоторому каталогу
+в ФС (например, `/mnt/vitastor/`) и примонтировать этот каталог с помощью [`vitastor-nfs mount --block`](../usage/nfs.ru.md).
+Они начнут обращаться к образам как к файлам и, вероятно, смогут заработать корректно :).

docs/usage/ublk.en.md

@@ -0,0 +1,116 @@
+[Documentation](../../README.md#documentation) → Usage → UBLK
+
+-----
+
+[Читать на русском](ublk.ru.md)
+
+# UBLK
+
+[ublk](https://docs.kernel.org/block/ublk.html) is a new io_uring-based Linux interface
+for user-space block device drivers, available since Linux 6.0.
+
+It's not zero-copy, but it's still a fast implementation, outperforming both [NBD](nbd.en.md)
+and [VDUSE](qemu.en.md#vduse) iops-wise and may or may not outperform VDUSE in linear I/O MB/s.
+ublk also allows to recover devices even if the server (vitastor-ublk process) dies.
+
+## Example performance comparison
+
+TCP (100G), 3 hosts each with 6 NVMe OSDs, 3 replicas, single client
+
+|                      | direct fio  | NBD         | VDUSE      | UBLK        |
+|----------------------|-------------|-------------|------------|-------------|
+| linear write         | 3807 MB/s   | 1832 MB/s   | 3226 MB/s  | 3027 MB/s   |
+| linear read          | 3067 MB/s   | 1885 MB/s   | 1800 MB/s  | 2076 MB/s   |
+| 4k random write Q128 | 128624 iops | 91060 iops  | 94621 iops | 149450 iops |
+| 4k random read Q128  | 117769 iops | 153408 iops | 93157 iops | 171987 iops |
+| 4k random write Q1   | 8090 iops   | 6442 iops   | 6316 iops  | 7272 iops   |
+| 4k random read Q1    | 9474 iops   | 7200 iops   | 6840 iops  | 8038 iops   |
+
+RDMA (100G), 3 hosts each with 6 NVMe OSDs, 3 replicas, single client
+
+|                      | direct fio  | NBD         | VDUSE       | UBLK        |
+|----------------------|-------------|-------------|-------------|-------------|
+| linear write         | 6998 MB/s   | 1878 MB/s   | 4249 MB/s   | 3140 MB/s   |
+| linear read          | 8628 MB/s   | 3389 MB/s   | 5062 MB/s   | 3674 MB/s   |
+| 4k random write Q128 | 222541 iops | 181589 iops | 138281 iops | 218222 iops |
+| 4k random read Q128  | 412647 iops | 239987 iops | 151663 iops | 269583 iops |
+| 4k random write Q1   | 11601 iops  | 8592 iops   | 9111 iops   | 10000 iops  |
+| 4k random read Q1    | 10102 iops  | 7788 iops   | 8111 iops   | 8965 iops   |
+
+## Commands
+
+vitastor-ublk supports the following commands:
+
+- [map](#map)
+- [unmap](#unmap)
+- [ls](#ls)
+
+## map
+
+To create a local block device for a Vitastor image run:
+
+```
+vitastor-ublk map [/dev/ublkbN] --image testimg
+```
+
+It will output a block device name like /dev/ublkb0 which you can then use as a normal disk.
+
+You can also use `--pool <POOL> --inode <INODE> --size <SIZE>` instead of `--image <IMAGE>` if you want.
+
+vitastor-ublk supports all usual Vitastor configuration options like `--config_path <path_to_config>` plus ublk-specific:
+
+* `--recover` \
+  Recover a mapped device if the previous ublk server is dead.
+* `--queue_depth 256` \
+  Maximum queue size for the device.
+* `--max_io_size 1M` \
+  Maximum single I/O size for the device. Default: `max(1 MB, pool block size * EC part count)`.
+* `--readonly` \
+  Make the device read-only.
+* `--hdd` \
+  Mark the device as rotational.
+* `--logfile /path/to/log/file.txt` \
+  Write log messages to the specified file instead of dropping them (in background mode)
+  or printing them to the standard output (in foreground mode).
+* `--dev_num N` \
+  Use the specified device /dev/ublkbN instead of automatic selection (alternative syntax
+  to /dev/ublkbN positional parameter).
+* `--foreground 1` \
+  Stay in foreground, do not daemonize.
+
+Note that `ublk_queue_depth` and `ublk_max_io_size` may also be specified
+in `/etc/vitastor/vitastor.conf` or in other configuration file specified with `--config_path`.
+
+## unmap
+
+To unmap the device run:
+
+```
+vitastor-ublk unmap /dev/ublkb0
+```
+
+## ls
+
+```
+vitastor-ublk ls [--json]
+```
+
+List mapped images.
+
+Example output (normal format):
+
+```
+/dev/ublkb0
+image: bench
+pid: 584536
+
+/dev/ublkb1
+image: bench1
+pid: 584546
+```
+
+Example output (JSON format):
+
+```
+{"/dev/ublkb0": {"image": "bench", "pid": 584536}, "/dev/ublkb1": {"image": "bench1", "pid": 584546}}
+```

docs/usage/ublk.ru.md

@@ -0,0 +1,121 @@
+[Документация](../../README-ru.md#документация) → Использование → UBLK
+
+-----
+
+[Read in English](ublk.en.md)
+
+# UBLK
+
+[ublk](https://docs.kernel.org/block/ublk.html) - это новый Linux-интерфейс на основе io_uring
+для реализации блочных устройств в пространстве пользователя, доступный, начиная с Linux 6.0.
+
+ublk тоже копирует память (т.е. не является zero-copy), но по IOPS всё равно обгоняет и
+[NBD](nbd.ru.md), и [VDUSE](qemu.ru.md#vduse), и иногда может даже обгонять VDUSE по
+скорости линейного доступа. Также ublk позволяет оживлять устройства, у которых умер
+сервер (процесс-обработчик vitastor-ublk).
+
+## Пример сравнения производительности
+
+TCP (100G), 3 сервера с 6 NVMe OSD каждый, 3 реплики, один клиент
+
+|                          | Прямой fio  | NBD         | VDUSE      | UBLK        |
+|--------------------------|-------------|-------------|------------|-------------|
+| линейная запись          | 3807 MB/s   | 1832 MB/s   | 3226 MB/s  | 3027 MB/s   |
+| линейное чтение          | 3067 MB/s   | 1885 MB/s   | 1800 MB/s  | 2076 MB/s   |
+| 4k случайная запись Q128 | 128624 iops | 91060 iops  | 94621 iops | 149450 iops |
+| 4k случайное чтение Q128 | 117769 iops | 153408 iops | 93157 iops | 171987 iops |
+| 4k случайная запись Q1   | 8090 iops   | 6442 iops   | 6316 iops  | 7272 iops   |
+| 4k случайное чтение Q1   | 9474 iops   | 7200 iops   | 6840 iops  | 8038 iops   |
+
+RDMA (100G), 3 сервера с 6 NVMe OSD каждый, 3 реплики, один клиент
+
+|                          | Прямой fio  | NBD         | VDUSE       | UBLK        |
+|--------------------------|-------------|-------------|-------------|-------------|
+| линейная запись          | 6998 MB/s   | 1878 MB/s   | 4249 MB/s   | 3140 MB/s   |
+| линейное чтение          | 8628 MB/s   | 3389 MB/s   | 5062 MB/s   | 3674 MB/s   |
+| 4k случайная запись Q128 | 222541 iops | 181589 iops | 138281 iops | 218222 iops |
+| 4k случайное чтение Q128 | 412647 iops | 239987 iops | 151663 iops | 269583 iops |
+| 4k случайная запись Q1   | 11601 iops  | 8592 iops   | 9111 iops   | 10000 iops  |
+| 4k случайное чтение Q1   | 10102 iops  | 7788 iops   | 8111 iops   | 8965 iops   |
+
+## Команды
+
+vitastor-ublk поддерживает следующие команды:
+
+- [map](#map)
+- [unmap](#unmap)
+- [ls](#ls)
+
+## map
+
+Чтобы создать локальное блочное устройство для образа, выполните команду:
+
+```
+vitastor-ublk map [/dev/ublkbN] --image testimg
+```
+
+Команда напечатает название блочного устройства вида /dev/ublkb0, которое потом можно
+будет использовать как обычный диск.
+
+Для обращения по номеру инода, аналогично другим командам, можно использовать опции
+`--pool <POOL> --inode <INODE> --size <SIZE>` вместо `--image testimg`.
+
+vitastor-ublk поддерживает все обычные опции Vitastor, например, `--config_path <path_to_config>`,
+плюс специфичные для ublk:
+
+* `--recover` \
+  Восстановить ранее подключённое устройство, у которого умер обработчик.
+* `--queue_depth 256` \
+  Максимальная глубина очереди устройства.
+* `--max_io_size 1M` \
+  Максимальный размер запроса ввода-вывода для устройства. По умолчанию: `max(1 MB, блок данных пула * число частей данных EC)`.
+* `--readonly` \
+  Подключить устройство в режиме только для чтения.
+* `--hdd` \
+  Пометить устройство как вращающийся жёсткий диск (флаг rotational).
+* `--logfile /path/to/log/file.txt` \
+  Писать сообщения о процессе работы в заданный файл, вместо пропуска их
+  при фоновом режиме запуска или печати на стандартный вывод при запуске
+  в консоли с `--foreground 1`.
+* `--dev_num N` \
+  Использовать заданное устройство `/dev/ublkbN` вместо автоматического подбора.
+* `--foreground 1` \
+  Не уводить процесс в фоновый режим.
+
+Обратите внимание, что опции `ublk_queue_depth` и `ublk_max_io_size` можно
+также задавать в `/etc/vitastor/vitastor.conf` или в другом файле конфигурации,
+заданном опцией `--config_path`.
+
+## unmap
+
+Для отключения устройства выполните:
+
+```
+vitastor-ublk unmap /dev/ublkb0
+```
+
+## ls
+
+```
+vitastor-ublk ls [--json]
+```
+
+Вывести подключённые устройства.
+
+Пример вывода в обычном формате:
+
+```
+/dev/ublkb0
+image: bench
+pid: 584536
+
+/dev/ublkb1
+image: bench1
+pid: 584546
+```
+
+Пример вывода в JSON-формате:
+
+```
+{"/dev/ublkb0": {"image": "bench", "pid": 584536}, "/dev/ublkb1": {"image": "bench1", "pid": 584546}}
+```

mon/mon.js

@@ -96,6 +96,7 @@ class Mon
                     }
                     else
                     {
+                        res.setHeader('Content-Type', 'text/plain; version=0.0.4; charset=utf-8');
                         res.write(export_prometheus_metrics(this.state));
                     }
                 }

mon/osd_tree.js

@@ -15,7 +15,7 @@ function get_osd_tree(global_config, state)
         const stat = state.osd.stats[osd_num];
         const osd_cfg = state.config.osd[osd_num];
         let reweight = osd_cfg == null ? 1 : Number(osd_cfg.reweight);
-        if (isNaN(reweight) || reweight < 0 || reweight > 0)
+        if (isNaN(reweight) || reweight < 0 || reweight > 1)
             reweight = 1;
         if (stat && stat.size && reweight && (state.osd.state[osd_num] || Number(stat.time) >= down_time ||
             osd_cfg && osd_cfg.noout))
@@ -179,7 +179,7 @@ function filter_osds_by_block_layout(orig_tree, osd_stats, block_size, bitmap_gr
         if (orig_tree[osd].level === 'osd')
         {
             const osd_stat = osd_stats[osd];
-            if (osd_stat && (osd_stat.bs_block_size && osd_stat.bs_block_size != block_size ||
+            if (osd_stat && (osd_stat.data_block_size && osd_stat.data_block_size != block_size ||
                 osd_stat.bitmap_granularity && osd_stat.bitmap_granularity != bitmap_granularity ||
                 osd_stat.immediate_commit == 'small' && immediate_commit == 'all' ||
                 osd_stat.immediate_commit == 'none' && immediate_commit != 'none'))

mon/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vitastor-mon",
-  "version": "2.2.2",
+  "version": "2.3.0",
   "description": "Vitastor SDS monitor service",
   "main": "mon-main.js",
   "scripts": {
@@ -9,7 +9,7 @@
   "author": "Vitaliy Filippov",
   "license": "UNLICENSED",
   "dependencies": {
-    "antietcd": "^1.1.2",
+    "antietcd": "^1.1.3",
     "sprintf-js": "^1.1.2",
     "ws": "^7.2.5"
   },

node-binding/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vitastor",
-  "version": "2.2.2",
+  "version": "2.3.0",
   "description": "Low-level native bindings to Vitastor client library",
   "main": "index.js",
   "keywords": [

patches/VitastorPlugin.pm

@@ -261,7 +261,7 @@ sub free_image
     my ($vtype, $name, $vmid, undef, undef, undef) = $class->parse_volname($volname);
     $class->deactivate_volume($storeid, $scfg, $volname);
     my $full_list = run_cli($scfg, [ 'ls', '-l' ]);
-    my $list = _process_list($scfg, $storeid, $full_list);
+    my $list = _process_list($scfg, $storeid, $full_list, 0);
     # Remove image and all its snapshots
     my $rm_names = {
         map { ($prefix.$_->{name} => 1) }
@@ -269,6 +269,10 @@ sub free_image
         @$list
     };
     my $children = [ grep { $_->{parent_name} && $rm_names->{$_->{parent_name}} } @$full_list ];
+    $children = [ grep { 
+        substr($_->{name}, 0, length($prefix.$name)) ne $prefix.$name &&
+        substr($_->{name}, 0, length($prefix.$name)+1) ne $prefix.$name.'@'
+    } @$children ];
     die "Image has children: ".join(', ', map {
         substr($_->{name}, 0, length $prefix) eq $prefix
             ? substr($_->name, length $prefix)
@@ -288,14 +292,15 @@ sub free_image
 
 sub _process_list
 {
-    my ($scfg, $storeid, $result) = @_;
+    my ($scfg, $storeid, $result, $skip_snapshot) = @_;
+    $skip_snapshot = 1 if !defined $skip_snapshot;
     my $prefix = defined $scfg->{vitastor_prefix} ? $scfg->{vitastor_prefix} : 'pve/';
     my $list = [];
     foreach my $el (@$result)
     {
         next if !$el->{name} || length($prefix) && substr($el->{name}, 0, length $prefix) ne $prefix;
         my $name = substr($el->{name}, length $prefix);
-        next if $name =~ /@/;
+        next if $skip_snapshot && $name =~ /@/;
         my ($owner) = $name =~ /^(?:vm|base)-(\d+)-/s;
         next if !defined $owner;
         my $parent = !defined $el->{parent_name}
@@ -494,4 +499,55 @@ sub rename_volume
     return "${storeid}:${base_name}${target_volname}";
 }
 
+sub _monkey_patch_qemu_blockdev_options
+{
+    my ($cfg, $volid, $machine_version, $options) = @_;
+    my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid);
+
+    my $scfg = PVE::Storage::storage_config($cfg, $storeid);
+
+    my $plugin = PVE::Storage::Plugin->lookup($scfg->{type});
+
+    my ($vtype) = $plugin->parse_volname($volname);
+    die "cannot use volume of type '$vtype' as a QEMU blockdevice\n"
+        if $vtype ne 'images' && $vtype ne 'iso' && $vtype ne 'import';
+
+    return $plugin->qemu_blockdev_options($scfg, $storeid, $volname, $machine_version, $options);
+}
+
+sub qemu_blockdev_options
+{
+    my ($class, $scfg, $storeid, $volname, $machine_version, $options) = @_;
+    my $prefix = defined $scfg->{vitastor_prefix} ? $scfg->{vitastor_prefix} : 'pve/';
+    my ($vtype, $name, $vmid) = $class->parse_volname($volname);
+    $name .= '@'.$options->{'snapshot-name'} if $options->{'snapshot-name'};
+    if ($scfg->{vitastor_nbd})
+    {
+        my $mapped = run_cli($scfg, [ 'ls' ], binary => '/usr/bin/vitastor-nbd');
+        my ($kerneldev) = grep { $mapped->{$_}->{image} eq $prefix.$name } keys %$mapped;
+        die "Image not mapped via NBD" if !$kerneldev;
+        return { driver => 'host_device', filename => $kerneldev };
+    }
+    my $blockdev = {
+        driver => 'vitastor',
+        image => $prefix.$name,
+    };
+    if ($scfg->{vitastor_config_path})
+    {
+        $blockdev->{'config-path'} = $scfg->{vitastor_config_path};
+    }
+    if ($scfg->{vitastor_etcd_address})
+    {
+        # FIXME This is the only exception: etcd_address -> etcd_host for qemu
+        $blockdev->{'etcd-host'} = $scfg->{vitastor_etcd_address};
+    }
+    if ($scfg->{vitastor_etcd_prefix})
+    {
+        $blockdev->{'etcd-prefix'} = $scfg->{vitastor_etcd_prefix};
+    }
+    return $blockdev;
+}
+
+*PVE::Storage::qemu_blockdev_options = *_monkey_patch_qemu_blockdev_options;
+
 1;

patches/cinder-vitastor.py

@@ -50,7 +50,7 @@ from cinder.volume import configuration
 from cinder.volume import driver
 from cinder.volume import volume_utils
 
-VITASTOR_VERSION = '2.2.2'
+VITASTOR_VERSION = '2.3.0'
 
 LOG = logging.getLogger(__name__)
 

patches/libvirt-11.5-vitastor.diff

@@ -0,0 +1,637 @@
+diff --git a/include/libvirt/libvirt-storage.h b/include/libvirt/libvirt-storage.h
+index aaad4a3da1..5f5daa8341 100644
+--- a/include/libvirt/libvirt-storage.h
++++ b/include/libvirt/libvirt-storage.h
+@@ -326,6 +326,7 @@ typedef enum {
+     VIR_CONNECT_LIST_STORAGE_POOLS_ZFS           = 1 << 17, /* (Since: 1.2.8) */
+     VIR_CONNECT_LIST_STORAGE_POOLS_VSTORAGE      = 1 << 18, /* (Since: 3.1.0) */
+     VIR_CONNECT_LIST_STORAGE_POOLS_ISCSI_DIRECT  = 1 << 19, /* (Since: 5.6.0) */
++    VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR      = 1 << 20, /* (Since: 5.0.0) */
+ } virConnectListAllStoragePoolsFlags;
+ 
+ int                     virConnectListAllStoragePools(virConnectPtr conn,
+diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c
+index 1e24e41a48..ce359a4cf8 100644
+--- a/src/conf/domain_conf.c
++++ b/src/conf/domain_conf.c
+@@ -7435,7 +7435,8 @@ virDomainDiskSourceNetworkParse(xmlNodePtr node,
+     src->configFile = virXPathString("string(./config/@file)", ctxt);
+ 
+     if (src->protocol == VIR_STORAGE_NET_PROTOCOL_HTTP ||
+-        src->protocol == VIR_STORAGE_NET_PROTOCOL_HTTPS)
++        src->protocol == VIR_STORAGE_NET_PROTOCOL_HTTPS ||
++        src->protocol == VIR_STORAGE_NET_PROTOCOL_VITASTOR)
+         src->query = virXMLPropString(node, "query");
+ 
+     if (virDomainStorageNetworkParseHosts(node, ctxt, &src->hosts, &src->nhosts) < 0)
+@@ -31871,6 +31872,7 @@ virDomainStorageSourceTranslateSourcePool(virStorageSource *src,
+ 
+     case VIR_STORAGE_POOL_MPATH:
+     case VIR_STORAGE_POOL_RBD:
++    case VIR_STORAGE_POOL_VITASTOR:
+     case VIR_STORAGE_POOL_SHEEPDOG:
+     case VIR_STORAGE_POOL_GLUSTER:
+     case VIR_STORAGE_POOL_LAST:
+diff --git a/src/conf/domain_validate.c b/src/conf/domain_validate.c
+index b28af7fa56..d1aae6e43e 100644
+--- a/src/conf/domain_validate.c
++++ b/src/conf/domain_validate.c
+@@ -504,6 +504,7 @@ virDomainDiskDefValidateSourceChainOne(const virStorageSource *src)
+         case VIR_STORAGE_NET_PROTOCOL_RBD:
+             break;
+ 
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+         case VIR_STORAGE_NET_PROTOCOL_NBD:
+         case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+         case VIR_STORAGE_NET_PROTOCOL_GLUSTER:
+@@ -576,7 +577,7 @@ virDomainDiskDefValidateSourceChainOne(const virStorageSource *src)
+         }
+     }
+ 
+-    /* internal snapshots and config files are currently supported only with rbd: */
++    /* internal snapshots are currently supported only with rbd: */
+     if (virStorageSourceGetActualType(src) != VIR_STORAGE_TYPE_NETWORK &&
+         src->protocol != VIR_STORAGE_NET_PROTOCOL_RBD) {
+         if (src->snapshot) {
+@@ -584,10 +585,14 @@ virDomainDiskDefValidateSourceChainOne(const virStorageSource *src)
+                            _("<snapshot> element is currently supported only with 'rbd' disks"));
+             return -1;
+         }
+-
++    }
++    /* config files are currently supported only with rbd and vitastor: */
++    if (virStorageSourceGetActualType(src) != VIR_STORAGE_TYPE_NETWORK &&
++        src->protocol != VIR_STORAGE_NET_PROTOCOL_RBD &&
++        src->protocol != VIR_STORAGE_NET_PROTOCOL_VITASTOR) {
+         if (src->configFile) {
+             virReportError(VIR_ERR_XML_ERROR, "%s",
+-                           _("<config> element is currently supported only with 'rbd' disks"));
++                           _("<config> element is currently supported only with 'rbd' and 'vitastor' disks"));
+             return -1;
+         }
+     }
+diff --git a/src/conf/schemas/domaincommon.rng b/src/conf/schemas/domaincommon.rng
+index 183dd5db5e..dcc0d1a778 100644
+--- a/src/conf/schemas/domaincommon.rng
++++ b/src/conf/schemas/domaincommon.rng
+@@ -2066,6 +2066,35 @@
+     </element>
+   </define>
+ 
++  <define name="diskSourceNetworkProtocolVitastor">
++    <element name="source">
++      <interleave>
++        <attribute name="protocol">
++          <value>vitastor</value>
++        </attribute>
++        <ref name="diskSourceCommon"/>
++        <optional>
++          <attribute name="name"/>
++        </optional>
++        <optional>
++          <attribute name="query"/>
++        </optional>
++        <zeroOrMore>
++          <ref name="diskSourceNetworkHost"/>
++        </zeroOrMore>
++        <optional>
++          <element name="config">
++            <attribute name="file">
++              <ref name="absFilePath"/>
++            </attribute>
++            <empty/>
++          </element>
++        </optional>
++        <empty/>
++      </interleave>
++    </element>
++  </define>
++
+   <define name="diskSourceNetworkProtocolISCSI">
+     <element name="source">
+       <attribute name="protocol">
+@@ -2416,6 +2445,7 @@
+       <ref name="diskSourceNetworkProtocolSimple"/>
+       <ref name="diskSourceNetworkProtocolVxHS"/>
+       <ref name="diskSourceNetworkProtocolNFS"/>
++      <ref name="diskSourceNetworkProtocolVitastor"/>
+     </choice>
+   </define>
+ 
+diff --git a/src/conf/storage_conf.c b/src/conf/storage_conf.c
+index 1dc9365bf2..a8a736be81 100644
+--- a/src/conf/storage_conf.c
++++ b/src/conf/storage_conf.c
+@@ -56,7 +56,7 @@ VIR_ENUM_IMPL(virStoragePool,
+               "logical", "disk", "iscsi",
+               "iscsi-direct", "scsi", "mpath",
+               "rbd", "sheepdog", "gluster",
+-              "zfs", "vstorage",
++              "zfs", "vstorage", "vitastor",
+ );
+ 
+ VIR_ENUM_IMPL(virStoragePoolFormatFileSystem,
+@@ -242,6 +242,18 @@ static virStoragePoolTypeInfo poolTypeInfo[] = {
+           .formatToString = virStorageFileFormatTypeToString,
+       }
+     },
++    {.poolType = VIR_STORAGE_POOL_VITASTOR,
++     .poolOptions = {
++         .flags = (VIR_STORAGE_POOL_SOURCE_HOST |
++                   VIR_STORAGE_POOL_SOURCE_NETWORK |
++                   VIR_STORAGE_POOL_SOURCE_NAME),
++      },
++      .volOptions = {
++          .defaultFormat = VIR_STORAGE_FILE_RAW,
++          .formatFromString = virStorageVolumeFormatFromString,
++          .formatToString = virStorageFileFormatTypeToString,
++      }
++    },
+     {.poolType = VIR_STORAGE_POOL_SHEEPDOG,
+      .poolOptions = {
+          .flags = (VIR_STORAGE_POOL_SOURCE_HOST |
+@@ -538,6 +550,11 @@ virStoragePoolDefParseSource(xmlXPathContextPtr ctxt,
+                        _("element 'name' is mandatory for RBD pool"));
+         return -1;
+     }
++    if (pool_type == VIR_STORAGE_POOL_VITASTOR && source->name == NULL) {
++        virReportError(VIR_ERR_XML_ERROR, "%s",
++                       _("element 'name' is mandatory for Vitastor pool"));
++        return -1;
++    }
+ 
+     if (options->formatFromString) {
+         g_autofree char *format = NULL;
+@@ -1127,6 +1144,7 @@ virStoragePoolDefFormatBuf(virBuffer *buf,
+     /* RBD, Sheepdog, Gluster and Iscsi-direct devices are not local block devs nor
+      * files, so they don't have a target */
+     if (def->type != VIR_STORAGE_POOL_RBD &&
++        def->type != VIR_STORAGE_POOL_VITASTOR &&
+         def->type != VIR_STORAGE_POOL_SHEEPDOG &&
+         def->type != VIR_STORAGE_POOL_GLUSTER &&
+         def->type != VIR_STORAGE_POOL_ISCSI_DIRECT) {
+diff --git a/src/conf/storage_conf.h b/src/conf/storage_conf.h
+index fc67957cfe..720c07ef74 100644
+--- a/src/conf/storage_conf.h
++++ b/src/conf/storage_conf.h
+@@ -103,6 +103,7 @@ typedef enum {
+     VIR_STORAGE_POOL_GLUSTER,  /* Gluster device */
+     VIR_STORAGE_POOL_ZFS,      /* ZFS */
+     VIR_STORAGE_POOL_VSTORAGE, /* Virtuozzo Storage */
++    VIR_STORAGE_POOL_VITASTOR, /* Vitastor */
+ 
+     VIR_STORAGE_POOL_LAST,
+ } virStoragePoolType;
+@@ -454,6 +455,7 @@ VIR_ENUM_DECL(virStoragePartedFs);
+                  VIR_CONNECT_LIST_STORAGE_POOLS_SCSI     | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_MPATH    | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_RBD      | \
++                 VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_SHEEPDOG | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_GLUSTER  | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_ZFS      | \
+diff --git a/src/conf/storage_source_conf.c b/src/conf/storage_source_conf.c
+index 8a063be244..dd9c7f11a2 100644
+--- a/src/conf/storage_source_conf.c
++++ b/src/conf/storage_source_conf.c
+@@ -89,6 +89,7 @@ VIR_ENUM_IMPL(virStorageNetProtocol,
+               "ssh",
+               "vxhs",
+               "nfs",
++              "vitastor",
+ );
+ 
+ 
+@@ -1314,6 +1315,7 @@ virStorageSourceNetworkDefaultPort(virStorageNetProtocol protocol)
+         case VIR_STORAGE_NET_PROTOCOL_GLUSTER:
+             return 24007;
+ 
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+         case VIR_STORAGE_NET_PROTOCOL_RBD:
+             /* we don't provide a default for RBD */
+             return 0;
+diff --git a/src/conf/storage_source_conf.h b/src/conf/storage_source_conf.h
+index ebddf28cd6..873a2be65c 100644
+--- a/src/conf/storage_source_conf.h
++++ b/src/conf/storage_source_conf.h
+@@ -130,6 +130,7 @@ typedef enum {
+     VIR_STORAGE_NET_PROTOCOL_SSH,
+     VIR_STORAGE_NET_PROTOCOL_VXHS,
+     VIR_STORAGE_NET_PROTOCOL_NFS,
++    VIR_STORAGE_NET_PROTOCOL_VITASTOR,
+ 
+     VIR_STORAGE_NET_PROTOCOL_LAST
+ } virStorageNetProtocol;
+diff --git a/src/conf/virstorageobj.c b/src/conf/virstorageobj.c
+index 59fa5da372..4739167f5f 100644
+--- a/src/conf/virstorageobj.c
++++ b/src/conf/virstorageobj.c
+@@ -1438,6 +1438,7 @@ virStoragePoolObjSourceFindDuplicateCb(const void *payload,
+             return 1;
+         break;
+ 
++    case VIR_STORAGE_POOL_VITASTOR:
+     case VIR_STORAGE_POOL_ISCSI_DIRECT:
+     case VIR_STORAGE_POOL_RBD:
+     case VIR_STORAGE_POOL_LAST:
+@@ -1921,6 +1922,8 @@ virStoragePoolObjMatch(virStoragePoolObj *obj,
+                (obj->def->type == VIR_STORAGE_POOL_MPATH))   ||
+               (MATCH(VIR_CONNECT_LIST_STORAGE_POOLS_RBD) &&
+                (obj->def->type == VIR_STORAGE_POOL_RBD))     ||
++              (MATCH(VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR) &&
++               (obj->def->type == VIR_STORAGE_POOL_VITASTOR)) ||
+               (MATCH(VIR_CONNECT_LIST_STORAGE_POOLS_SHEEPDOG) &&
+                (obj->def->type == VIR_STORAGE_POOL_SHEEPDOG)) ||
+               (MATCH(VIR_CONNECT_LIST_STORAGE_POOLS_GLUSTER) &&
+diff --git a/src/libvirt-storage.c b/src/libvirt-storage.c
+index db7660aac4..561df34709 100644
+--- a/src/libvirt-storage.c
++++ b/src/libvirt-storage.c
+@@ -94,6 +94,7 @@ virStoragePoolGetConnect(virStoragePoolPtr pool)
+  * VIR_CONNECT_LIST_STORAGE_POOLS_SCSI
+  * VIR_CONNECT_LIST_STORAGE_POOLS_MPATH
+  * VIR_CONNECT_LIST_STORAGE_POOLS_RBD
++ * VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR
+  * VIR_CONNECT_LIST_STORAGE_POOLS_SHEEPDOG
+  * VIR_CONNECT_LIST_STORAGE_POOLS_GLUSTER
+  * VIR_CONNECT_LIST_STORAGE_POOLS_ZFS
+diff --git a/src/libxl/libxl_conf.c b/src/libxl/libxl_conf.c
+index bdd30dd65a..5353e00b4a 100644
+--- a/src/libxl/libxl_conf.c
++++ b/src/libxl/libxl_conf.c
+@@ -1081,6 +1081,7 @@ libxlMakeNetworkDiskSrcStr(virStorageSource *src,
+     case VIR_STORAGE_NET_PROTOCOL_SSH:
+     case VIR_STORAGE_NET_PROTOCOL_VXHS:
+     case VIR_STORAGE_NET_PROTOCOL_NFS:
++    case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+     case VIR_STORAGE_NET_PROTOCOL_LAST:
+     case VIR_STORAGE_NET_PROTOCOL_NONE:
+         virReportError(VIR_ERR_NO_SUPPORT,
+diff --git a/src/libxl/xen_xl.c b/src/libxl/xen_xl.c
+index ec8de30c01..61eab9606d 100644
+--- a/src/libxl/xen_xl.c
++++ b/src/libxl/xen_xl.c
+@@ -1461,6 +1461,7 @@ xenFormatXLDiskSrcNet(virStorageSource *src)
+     case VIR_STORAGE_NET_PROTOCOL_SSH:
+     case VIR_STORAGE_NET_PROTOCOL_VXHS:
+     case VIR_STORAGE_NET_PROTOCOL_NFS:
++    case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+     case VIR_STORAGE_NET_PROTOCOL_LAST:
+     case VIR_STORAGE_NET_PROTOCOL_NONE:
+         virReportError(VIR_ERR_NO_SUPPORT,
+diff --git a/src/qemu/qemu_block.c b/src/qemu/qemu_block.c
+index 32568d4ae6..e625fa0720 100644
+--- a/src/qemu/qemu_block.c
++++ b/src/qemu/qemu_block.c
+@@ -731,6 +731,38 @@ qemuBlockStorageSourceGetRBDProps(virStorageSource *src,
+ }
+ 
+ 
++static virJSONValue *
++qemuBlockStorageSourceGetVitastorProps(virStorageSource *src)
++{
++    virJSONValue *ret = NULL;
++    virStorageNetHostDef *host;
++    size_t i;
++    g_auto(virBuffer) buf = VIR_BUFFER_INITIALIZER;
++    g_autofree char *etcd = NULL;
++
++    for (i = 0; i < src->nhosts; i++) {
++        host = src->hosts + i;
++        if ((virStorageNetHostTransport)host->transport != VIR_STORAGE_NET_HOST_TRANS_TCP) {
++            return NULL;
++        }
++        virBufferAsprintf(&buf, i > 0 ? ",%s:%u" : "%s:%u", host->name, host->port);
++    }
++    if (src->nhosts > 0) {
++        etcd = virBufferContentAndReset(&buf);
++    }
++
++    if (virJSONValueObjectAdd(&ret,
++                              "S:etcd-host", etcd,
++                              "S:etcd-prefix", src->query,
++                              "S:config-path", src->configFile,
++                              "s:image", src->path,
++                              NULL) < 0)
++        return NULL;
++
++    return ret;
++}
++
++
+ static virJSONValue *
+ qemuBlockStorageSourceGetSshProps(virStorageSource *src)
+ {
+@@ -1082,6 +1114,12 @@ qemuBlockStorageSourceGetBackendProps(virStorageSource *src,
+                 return NULL;
+             break;
+ 
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
++            driver = "vitastor";
++            if (!(fileprops = qemuBlockStorageSourceGetVitastorProps(src)))
++                return NULL;
++            break;
++
+         case VIR_STORAGE_NET_PROTOCOL_SSH:
+             driver = "ssh";
+             if (!(fileprops = qemuBlockStorageSourceGetSshProps(src)))
+@@ -1985,6 +2023,7 @@ qemuBlockGetBackingStoreString(virStorageSource *src,
+ 
+             case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+             case VIR_STORAGE_NET_PROTOCOL_RBD:
++            case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+             case VIR_STORAGE_NET_PROTOCOL_VXHS:
+             case VIR_STORAGE_NET_PROTOCOL_NFS:
+             case VIR_STORAGE_NET_PROTOCOL_SSH:
+@@ -2365,6 +2404,12 @@ qemuBlockStorageSourceCreateGetStorageProps(virStorageSource *src,
+                 return -1;
+             break;
+ 
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
++            driver = "vitastor";
++            if (!(location = qemuBlockStorageSourceGetVitastorProps(src)))
++                return -1;
++            break;
++
+         case VIR_STORAGE_NET_PROTOCOL_SSH:
+             if (srcPriv->nbdkitProcess) {
+                 /* disk creation not yet supported with nbdkit, and even if it
+diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c
+index 0d2548d8d4..91121d6e1f 100644
+--- a/src/qemu/qemu_domain.c
++++ b/src/qemu/qemu_domain.c
+@@ -4526,7 +4526,8 @@ qemuDomainValidateStorageSource(virStorageSource *src,
+     if (src->query &&
+         (actualType != VIR_STORAGE_TYPE_NETWORK ||
+          (src->protocol != VIR_STORAGE_NET_PROTOCOL_HTTPS &&
+-          src->protocol != VIR_STORAGE_NET_PROTOCOL_HTTP))) {
++          src->protocol != VIR_STORAGE_NET_PROTOCOL_HTTP &&
++          src->protocol != VIR_STORAGE_NET_PROTOCOL_VITASTOR))) {
+         virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
+                        _("query is supported only with HTTP(S) protocols"));
+         return -1;
+@@ -8954,6 +8955,7 @@ qemuDomainPrepareStorageSourceTLS(virStorageSource *src,
+         break;
+ 
+     case VIR_STORAGE_NET_PROTOCOL_RBD:
++    case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+     case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+     case VIR_STORAGE_NET_PROTOCOL_GLUSTER:
+     case VIR_STORAGE_NET_PROTOCOL_ISCSI:
+diff --git a/src/qemu/qemu_snapshot.c b/src/qemu/qemu_snapshot.c
+index 8128154749..afb339b9b0 100644
+--- a/src/qemu/qemu_snapshot.c
++++ b/src/qemu/qemu_snapshot.c
+@@ -662,6 +662,7 @@ qemuSnapshotPrepareDiskExternalInactive(virDomainSnapshotDiskDef *snapdisk,
+         case VIR_STORAGE_NET_PROTOCOL_NONE:
+         case VIR_STORAGE_NET_PROTOCOL_NBD:
+         case VIR_STORAGE_NET_PROTOCOL_RBD:
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+         case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+         case VIR_STORAGE_NET_PROTOCOL_GLUSTER:
+         case VIR_STORAGE_NET_PROTOCOL_ISCSI:
+@@ -887,6 +888,7 @@ qemuSnapshotPrepareDiskInternal(virDomainDiskDef *disk,
+         case VIR_STORAGE_NET_PROTOCOL_NONE:
+         case VIR_STORAGE_NET_PROTOCOL_NBD:
+         case VIR_STORAGE_NET_PROTOCOL_RBD:
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+         case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+         case VIR_STORAGE_NET_PROTOCOL_GLUSTER:
+         case VIR_STORAGE_NET_PROTOCOL_ISCSI:
+diff --git a/src/storage/storage_driver.c b/src/storage/storage_driver.c
+index e19e032427..59f91f4710 100644
+--- a/src/storage/storage_driver.c
++++ b/src/storage/storage_driver.c
+@@ -1626,6 +1626,7 @@ storageVolLookupByPathCallback(virStoragePoolObj *obj,
+ 
+         case VIR_STORAGE_POOL_GLUSTER:
+         case VIR_STORAGE_POOL_RBD:
++        case VIR_STORAGE_POOL_VITASTOR:
+         case VIR_STORAGE_POOL_SHEEPDOG:
+         case VIR_STORAGE_POOL_ZFS:
+         case VIR_STORAGE_POOL_LAST:
+diff --git a/src/storage_file/storage_source_backingstore.c b/src/storage_file/storage_source_backingstore.c
+index 80681924ea..8a3ade9ec0 100644
+--- a/src/storage_file/storage_source_backingstore.c
++++ b/src/storage_file/storage_source_backingstore.c
+@@ -287,6 +287,75 @@ virStorageSourceParseRBDColonString(const char *rbdstr,
+ }
+ 
+ 
++static int
++virStorageSourceParseVitastorColonString(const char *colonstr,
++                                         virStorageSource *src)
++{
++    char *p, *e, *next;
++    g_autofree char *options = NULL;
++
++    /* optionally skip the "vitastor:" prefix if provided */
++    if (STRPREFIX(colonstr, "vitastor:"))
++        colonstr += strlen("vitastor:");
++
++    options = g_strdup(colonstr);
++
++    p = options;
++    while (*p) {
++        /* find : delimiter or end of string */
++        for (e = p; *e && *e != ':'; ++e) {
++            if (*e == '\\') {
++                e++;
++                if (*e == '\0')
++                    break;
++            }
++        }
++        if (*e == '\0') {
++            next = e;    /* last kv pair */
++        } else {
++            next = e + 1;
++            *e = '\0';
++        }
++
++        if (STRPREFIX(p, "image=")) {
++            src->path = g_strdup(p + strlen("image="));
++        } else if (STRPREFIX(p, "etcd-prefix=")) {
++            src->query = g_strdup(p + strlen("etcd-prefix="));
++        } else if (STRPREFIX(p, "config-path=")) {
++            src->configFile = g_strdup(p + strlen("config-path="));
++        } else if (STRPREFIX(p, "etcd-host=")) {
++            char *h, *sep;
++
++            h = p + strlen("etcd-host=");
++            while (h < e) {
++                for (sep = h; sep < e; ++sep) {
++                    if (*sep == '\\' && (sep[1] == ',' ||
++                                         sep[1] == ';' ||
++                                         sep[1] == ' ')) {
++                        *sep = '\0';
++                        sep += 2;
++                        break;
++                    }
++                }
++
++                if (virStorageSourceRBDAddHost(src, h) < 0)
++                    return -1;
++
++                h = sep;
++            }
++        }
++
++        p = next;
++    }
++
++    if (!src->path) {
++        return -1;
++    }
++
++    return 0;
++}
++
++
+ static int
+ virStorageSourceParseNBDColonString(const char *nbdstr,
+                                     virStorageSource *src)
+@@ -399,6 +468,11 @@ virStorageSourceParseBackingColon(virStorageSource *src,
+             return -1;
+         break;
+ 
++    case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
++        if (virStorageSourceParseVitastorColonString(path, src) < 0)
++            return -1;
++        break;
++
+     case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+     case VIR_STORAGE_NET_PROTOCOL_LAST:
+     case VIR_STORAGE_NET_PROTOCOL_NONE:
+@@ -975,6 +1049,54 @@ virStorageSourceParseBackingJSONRBD(virStorageSource *src,
+     return 0;
+ }
+ 
++static int
++virStorageSourceParseBackingJSONVitastor(virStorageSource *src,
++                                         virJSONValue *json,
++                                         const char *jsonstr G_GNUC_UNUSED,
++                                         int opaque G_GNUC_UNUSED)
++{
++    const char *filename;
++    const char *image = virJSONValueObjectGetString(json, "image");
++    const char *conf = virJSONValueObjectGetString(json, "config-path");
++    const char *etcd_prefix = virJSONValueObjectGetString(json, "etcd-prefix");
++    virJSONValue *servers = virJSONValueObjectGetArray(json, "server");
++    size_t nservers;
++    size_t i;
++
++    src->type = VIR_STORAGE_TYPE_NETWORK;
++    src->protocol = VIR_STORAGE_NET_PROTOCOL_VITASTOR;
++
++    /* legacy syntax passed via 'filename' option */
++    if ((filename = virJSONValueObjectGetString(json, "filename")))
++        return virStorageSourceParseVitastorColonString(filename, src);
++
++    if (!image) {
++        virReportError(VIR_ERR_INVALID_ARG, "%s",
++                       _("missing image name in Vitastor backing volume "
++                         "JSON specification"));
++        return -1;
++    }
++
++    src->path = g_strdup(image);
++    src->configFile = g_strdup(conf);
++    src->query = g_strdup(etcd_prefix);
++
++    if (servers) {
++        nservers = virJSONValueArraySize(servers);
++
++        src->hosts = g_new0(virStorageNetHostDef, nservers);
++        src->nhosts = nservers;
++
++        for (i = 0; i < nservers; i++) {
++            if (virStorageSourceParseBackingJSONInetSocketAddress(src->hosts + i,
++                                                                  virJSONValueArrayGet(servers, i)) < 0)
++                return -1;
++        }
++    }
++
++    return 0;
++}
++
+ static int
+ virStorageSourceParseBackingJSONRaw(virStorageSource *src,
+                                     virJSONValue *json,
+@@ -1152,6 +1274,7 @@ static const struct virStorageSourceJSONDriverParser jsonParsers[] = {
+     {"sheepdog", false, virStorageSourceParseBackingJSONSheepdog, 0},
+     {"ssh", false, virStorageSourceParseBackingJSONSSH, 0},
+     {"rbd", false, virStorageSourceParseBackingJSONRBD, 0},
++    {"vitastor", false, virStorageSourceParseBackingJSONVitastor, 0},
+     {"raw", true, virStorageSourceParseBackingJSONRaw, 0},
+     {"nfs", false, virStorageSourceParseBackingJSONNFS, 0},
+     {"vxhs", false, virStorageSourceParseBackingJSONVxHS, 0},
+diff --git a/src/test/test_driver.c b/src/test/test_driver.c
+index 25335d9002..cf54069fbe 100644
+--- a/src/test/test_driver.c
++++ b/src/test/test_driver.c
+@@ -7340,6 +7340,7 @@ testStorageVolumeTypeForPool(int pooltype)
+     case VIR_STORAGE_POOL_ISCSI_DIRECT:
+     case VIR_STORAGE_POOL_GLUSTER:
+     case VIR_STORAGE_POOL_RBD:
++    case VIR_STORAGE_POOL_VITASTOR:
+         return VIR_STORAGE_VOL_NETWORK;
+     case VIR_STORAGE_POOL_LOGICAL:
+     case VIR_STORAGE_POOL_DISK:
+diff --git a/tests/storagepoolcapsschemadata/poolcaps-fs.xml b/tests/storagepoolcapsschemadata/poolcaps-fs.xml
+index eee75af746..8bd0a57bdd 100644
+--- a/tests/storagepoolcapsschemadata/poolcaps-fs.xml
++++ b/tests/storagepoolcapsschemadata/poolcaps-fs.xml
+@@ -204,4 +204,11 @@
+       </enum>
+     </volOptions>
+   </pool>
++  <pool type='vitastor' supported='no'>
++    <volOptions>
++      <defaultFormat type='raw'/>
++      <enum name='targetFormatType'>
++      </enum>
++    </volOptions>
++  </pool>
+ </storagepoolCapabilities>
+diff --git a/tests/storagepoolcapsschemadata/poolcaps-full.xml b/tests/storagepoolcapsschemadata/poolcaps-full.xml
+index 805950a937..852df0de16 100644
+--- a/tests/storagepoolcapsschemadata/poolcaps-full.xml
++++ b/tests/storagepoolcapsschemadata/poolcaps-full.xml
+@@ -204,4 +204,11 @@
+       </enum>
+     </volOptions>
+   </pool>
++  <pool type='vitastor' supported='yes'>
++    <volOptions>
++      <defaultFormat type='raw'/>
++      <enum name='targetFormatType'>
++      </enum>
++    </volOptions>
++  </pool>
+ </storagepoolCapabilities>
+diff --git a/tests/storagepoolxml2argvtest.c b/tests/storagepoolxml2argvtest.c
+index d5c2531ab8..b19308ac38 100644
+--- a/tests/storagepoolxml2argvtest.c
++++ b/tests/storagepoolxml2argvtest.c
+@@ -57,6 +57,7 @@ testCompareXMLToArgvFiles(bool shouldFail,
+     case VIR_STORAGE_POOL_GLUSTER:
+     case VIR_STORAGE_POOL_ZFS:
+     case VIR_STORAGE_POOL_VSTORAGE:
++    case VIR_STORAGE_POOL_VITASTOR:
+     case VIR_STORAGE_POOL_LAST:
+     default:
+         VIR_TEST_DEBUG("pool type '%s' has no xml2argv test", defTypeStr);
+diff --git a/tools/virsh-pool.c b/tools/virsh-pool.c
+index 2010ef1356..072e2ff9e8 100644
+--- a/tools/virsh-pool.c
++++ b/tools/virsh-pool.c
+@@ -1187,6 +1187,9 @@ cmdPoolList(vshControl *ctl, const vshCmd *cmd G_GNUC_UNUSED)
+             case VIR_STORAGE_POOL_VSTORAGE:
+                 flags |= VIR_CONNECT_LIST_STORAGE_POOLS_VSTORAGE;
+                 break;
++            case VIR_STORAGE_POOL_VITASTOR:
++                flags |= VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR;
++                break;
+             case VIR_STORAGE_POOL_LAST:
+                 break;
+             }

patches/pve-qemu-10.0-vitastor.patch

@@ -0,0 +1,172 @@
+Index: pve-qemu-kvm-10.0.2/block/meson.build
+===================================================================
+--- pve-qemu-kvm-10.0.2.orig/block/meson.build
++++ pve-qemu-kvm-10.0.2/block/meson.build
+@@ -126,6 +126,7 @@ foreach m : [
+   [libnfs, 'nfs', files('nfs.c')],
+   [libssh, 'ssh', files('ssh.c')],
+   [rbd, 'rbd', files('rbd.c')],
++  [vitastor, 'vitastor', files('vitastor.c')],
+ ]
+   if m[0].found()
+     module_ss = ss.source_set()
+Index: pve-qemu-kvm-10.0.2/meson.build
+===================================================================
+--- pve-qemu-kvm-10.0.2.orig/meson.build
++++ pve-qemu-kvm-10.0.2/meson.build
+@@ -1622,6 +1622,26 @@ if not get_option('rbd').auto() or have_
+   endif
+ endif
+ 
++vitastor = not_found
++if not get_option('vitastor').auto() or have_block
++  libvitastor_client = cc.find_library('vitastor_client', has_headers: ['vitastor_c.h'],
++    required: get_option('vitastor'))
++  if libvitastor_client.found()
++    if cc.links('''
++      #include <vitastor_c.h>
++      int main(void) {
++        vitastor_c_create_qemu(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
++        return 0;
++      }''', dependencies: libvitastor_client)
++      vitastor = declare_dependency(dependencies: libvitastor_client)
++    elif get_option('vitastor').enabled()
++      error('could not link libvitastor_client')
++    else
++      warning('could not link libvitastor_client, disabling')
++    endif
++  endif
++endif
++
+ glusterfs = not_found
+ glusterfs_ftruncate_has_stat = false
+ glusterfs_iocb_has_stat = false
+@@ -2514,6 +2534,7 @@ endif
+ config_host_data.set('CONFIG_OPENGL', opengl.found())
+ config_host_data.set('CONFIG_PLUGIN', get_option('plugins'))
+ config_host_data.set('CONFIG_RBD', rbd.found())
++config_host_data.set('CONFIG_VITASTOR', vitastor.found())
+ config_host_data.set('CONFIG_RDMA', rdma.found())
+ config_host_data.set('CONFIG_RELOCATABLE', get_option('relocatable'))
+ config_host_data.set('CONFIG_SAFESTACK', get_option('safe_stack'))
+@@ -4812,6 +4833,7 @@ summary_info += {'fdt support':       fd
+ summary_info += {'libcap-ng support': libcap_ng}
+ summary_info += {'bpf support':       libbpf}
+ summary_info += {'rbd support':       rbd}
++summary_info += {'vitastor support':  vitastor}
+ summary_info += {'smartcard support': cacard}
+ summary_info += {'U2F support':       u2f}
+ summary_info += {'libusb':            libusb}
+Index: pve-qemu-kvm-10.0.2/meson_options.txt
+===================================================================
+--- pve-qemu-kvm-10.0.2.orig/meson_options.txt
++++ pve-qemu-kvm-10.0.2/meson_options.txt
+@@ -202,6 +202,8 @@ option('pvg', type: 'feature', value: 'a
+        description: 'macOS paravirtualized graphics support')
+ option('rbd', type : 'feature', value : 'auto',
+        description: 'Ceph block device driver')
++option('vitastor', type : 'feature', value : 'auto',
++       description: 'Vitastor block device driver')
+ option('opengl', type : 'feature', value : 'auto',
+        description: 'OpenGL support')
+ option('rdma', type : 'feature', value : 'auto',
+Index: pve-qemu-kvm-10.0.2/qapi/block-core.json
+===================================================================
+--- pve-qemu-kvm-10.0.2.orig/qapi/block-core.json
++++ pve-qemu-kvm-10.0.2/qapi/block-core.json
+@@ -3599,7 +3599,7 @@
+             'raw', 'rbd',
+             { 'name': 'replication', 'if': 'CONFIG_REPLICATION' },
+             'pbs',
+-            'ssh', 'throttle', 'vdi', 'vhdx',
++            'ssh', 'throttle', 'vdi', 'vhdx', 'vitastor',
+             { 'name': 'virtio-blk-vfio-pci', 'if': 'CONFIG_BLKIO' },
+             { 'name': 'virtio-blk-vhost-user', 'if': 'CONFIG_BLKIO' },
+             { 'name': 'virtio-blk-vhost-vdpa', 'if': 'CONFIG_BLKIO' },
+@@ -4725,6 +4725,28 @@
+             '*server': ['InetSocketAddressBase'] } }
+ 
+ ##
++# @BlockdevOptionsVitastor:
++#
++# Driver specific block device options for vitastor
++#
++# @image:       Image name
++# @inode:       Inode number
++# @pool:        Pool ID
++# @size:        Desired image size in bytes
++# @config-path: Path to Vitastor configuration
++# @etcd-host:   etcd connection address(es)
++# @etcd-prefix: etcd key/value prefix
++##
++{ 'struct': 'BlockdevOptionsVitastor',
++  'data': { '*inode': 'uint64',
++            '*pool': 'uint64',
++            '*size': 'uint64',
++            '*image': 'str',
++            '*config-path': 'str',
++            '*etcd-host': 'str',
++            '*etcd-prefix': 'str' } }
++
++##
+ # @ReplicationMode:
+ #
+ # An enumeration of replication modes.
+@@ -5194,6 +5216,7 @@
+       'throttle':   'BlockdevOptionsThrottle',
+       'vdi':        'BlockdevOptionsGenericFormat',
+       'vhdx':       'BlockdevOptionsGenericFormat',
++      'vitastor':   'BlockdevOptionsVitastor',
+       'virtio-blk-vfio-pci':
+                     { 'type': 'BlockdevOptionsVirtioBlkVfioPci',
+                       'if': 'CONFIG_BLKIO' },
+@@ -5674,6 +5697,20 @@
+             '*encrypt' :        'RbdEncryptionCreateOptions' } }
+ 
+ ##
++# @BlockdevCreateOptionsVitastor:
++#
++# Driver specific image creation options for Vitastor.
++#
++# @location: Where to store the new image file.  This location cannot
++#     point to a snapshot.
++#
++# @size: Size of the virtual disk in bytes
++##
++{ 'struct': 'BlockdevCreateOptionsVitastor',
++  'data': { 'location':         'BlockdevOptionsVitastor',
++            'size':             'size' } }
++
++##
+ # @BlockdevVmdkSubformat:
+ #
+ # Subformat options for VMDK images
+@@ -5895,6 +5932,7 @@
+       'ssh':            'BlockdevCreateOptionsSsh',
+       'vdi':            'BlockdevCreateOptionsVdi',
+       'vhdx':           'BlockdevCreateOptionsVhdx',
++      'vitastor':       'BlockdevCreateOptionsVitastor',
+       'vmdk':           'BlockdevCreateOptionsVmdk',
+       'vpc':            'BlockdevCreateOptionsVpc'
+   } }
+Index: pve-qemu-kvm-10.0.2/scripts/meson-buildoptions.sh
+===================================================================
+--- pve-qemu-kvm-10.0.2.orig/scripts/meson-buildoptions.sh
++++ pve-qemu-kvm-10.0.2/scripts/meson-buildoptions.sh
+@@ -175,6 +175,7 @@ meson_options_help() {
+   printf "%s\n" '  qga-vss         build QGA VSS support (broken with MinGW)'
+   printf "%s\n" '  qpl             Query Processing Library support'
+   printf "%s\n" '  rbd             Ceph block device driver'
++  printf "%s\n" '  vitastor        Vitastor block device driver'
+   printf "%s\n" '  rdma            Enable RDMA-based migration'
+   printf "%s\n" '  replication     replication support'
+   printf "%s\n" '  rust            Rust support'
+@@ -458,6 +459,8 @@ _meson_option_parse() {
+     --disable-qpl) printf "%s" -Dqpl=disabled ;;
+     --enable-rbd) printf "%s" -Drbd=enabled ;;
+     --disable-rbd) printf "%s" -Drbd=disabled ;;
++    --enable-vitastor) printf "%s" -Dvitastor=enabled ;;
++    --disable-vitastor) printf "%s" -Dvitastor=disabled ;;
+     --enable-rdma) printf "%s" -Drdma=enabled ;;
+     --disable-rdma) printf "%s" -Drdma=disabled ;;
+     --enable-relocatable) printf "%s" -Drelocatable=true ;;

patches/qemu-10.0-vitastor.patch

@@ -0,0 +1,172 @@
+diff --git a/block/meson.build b/block/meson.build
+index 34b1b2a306..24ca0f1e52 100644
+--- a/block/meson.build
++++ b/block/meson.build
+@@ -114,6 +114,7 @@ foreach m : [
+   [libnfs, 'nfs', files('nfs.c')],
+   [libssh, 'ssh', files('ssh.c')],
+   [rbd, 'rbd', files('rbd.c')],
++  [vitastor, 'vitastor', files('vitastor.c')],
+ ]
+   if m[0].found()
+     module_ss = ss.source_set()
+diff --git a/meson.build b/meson.build
+index 41f68d3806..29eaed9ba4 100644
+--- a/meson.build
++++ b/meson.build
+@@ -1622,6 +1622,26 @@ if not get_option('rbd').auto() or have_block
+   endif
+ endif
+ 
++vitastor = not_found
++if not get_option('vitastor').auto() or have_block
++  libvitastor_client = cc.find_library('vitastor_client', has_headers: ['vitastor_c.h'],
++    required: get_option('vitastor'))
++  if libvitastor_client.found()
++    if cc.links('''
++      #include <vitastor_c.h>
++      int main(void) {
++        vitastor_c_create_qemu(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
++        return 0;
++      }''', dependencies: libvitastor_client)
++      vitastor = declare_dependency(dependencies: libvitastor_client)
++    elif get_option('vitastor').enabled()
++      error('could not link libvitastor_client')
++    else
++      warning('could not link libvitastor_client, disabling')
++    endif
++  endif
++endif
++
+ glusterfs = not_found
+ glusterfs_ftruncate_has_stat = false
+ glusterfs_iocb_has_stat = false
+@@ -2506,6 +2526,7 @@ endif
+ config_host_data.set('CONFIG_OPENGL', opengl.found())
+ config_host_data.set('CONFIG_PLUGIN', get_option('plugins'))
+ config_host_data.set('CONFIG_RBD', rbd.found())
++config_host_data.set('CONFIG_VITASTOR', vitastor.found())
+ config_host_data.set('CONFIG_RDMA', rdma.found())
+ config_host_data.set('CONFIG_RELOCATABLE', get_option('relocatable'))
+ config_host_data.set('CONFIG_SAFESTACK', get_option('safe_stack'))
+@@ -4813,6 +4834,7 @@ summary_info += {'fdt support':       fdt_opt == 'internal' ? 'internal' : fdt}
+ summary_info += {'libcap-ng support': libcap_ng}
+ summary_info += {'bpf support':       libbpf}
+ summary_info += {'rbd support':       rbd}
++summary_info += {'vitastor support':  vitastor}
+ summary_info += {'smartcard support': cacard}
+ summary_info += {'U2F support':       u2f}
+ summary_info += {'libusb':            libusb}
+diff --git a/meson_options.txt b/meson_options.txt
+index 59d973bca0..a3e7123980 100644
+--- a/meson_options.txt
++++ b/meson_options.txt
+@@ -202,6 +202,8 @@ option('pvg', type: 'feature', value: 'auto',
+        description: 'macOS paravirtualized graphics support')
+ option('rbd', type : 'feature', value : 'auto',
+        description: 'Ceph block device driver')
++option('vitastor', type : 'feature', value : 'auto',
++       description: 'Vitastor block device driver')
+ option('opengl', type : 'feature', value : 'auto',
+        description: 'OpenGL support')
+ option('rdma', type : 'feature', value : 'auto',
+diff --git a/qapi/block-core.json b/qapi/block-core.json
+index b1937780e1..a511193620 100644
+--- a/qapi/block-core.json
++++ b/qapi/block-core.json
+@@ -3216,7 +3216,7 @@
+             'parallels', 'preallocate', 'qcow', 'qcow2', 'qed', 'quorum',
+             'raw', 'rbd',
+             { 'name': 'replication', 'if': 'CONFIG_REPLICATION' },
+-            'ssh', 'throttle', 'vdi', 'vhdx',
++            'ssh', 'throttle', 'vdi', 'vhdx', 'vitastor',
+             { 'name': 'virtio-blk-vfio-pci', 'if': 'CONFIG_BLKIO' },
+             { 'name': 'virtio-blk-vhost-user', 'if': 'CONFIG_BLKIO' },
+             { 'name': 'virtio-blk-vhost-vdpa', 'if': 'CONFIG_BLKIO' },
+@@ -4299,6 +4299,28 @@
+             '*key-secret': 'str',
+             '*server': ['InetSocketAddressBase'] } }
+ 
++##
++# @BlockdevOptionsVitastor:
++#
++# Driver specific block device options for vitastor
++#
++# @image:       Image name
++# @inode:       Inode number
++# @pool:        Pool ID
++# @size:        Desired image size in bytes
++# @config-path: Path to Vitastor configuration
++# @etcd-host:   etcd connection address(es)
++# @etcd-prefix: etcd key/value prefix
++##
++{ 'struct': 'BlockdevOptionsVitastor',
++  'data': { '*inode': 'uint64',
++            '*pool': 'uint64',
++            '*size': 'uint64',
++            '*image': 'str',
++            '*config-path': 'str',
++            '*etcd-host': 'str',
++            '*etcd-prefix': 'str' } }
++
+ ##
+ # @ReplicationMode:
+ #
+@@ -4767,6 +4789,7 @@
+       'throttle':   'BlockdevOptionsThrottle',
+       'vdi':        'BlockdevOptionsGenericFormat',
+       'vhdx':       'BlockdevOptionsGenericFormat',
++      'vitastor':   'BlockdevOptionsVitastor',
+       'virtio-blk-vfio-pci':
+                     { 'type': 'BlockdevOptionsVirtioBlkVfioPci',
+                       'if': 'CONFIG_BLKIO' },
+@@ -5240,6 +5263,20 @@
+             '*cluster-size' :   'size',
+             '*encrypt' :        'RbdEncryptionCreateOptions' } }
+ 
++##
++# @BlockdevCreateOptionsVitastor:
++#
++# Driver specific image creation options for Vitastor.
++#
++# @location: Where to store the new image file.  This location cannot
++#     point to a snapshot.
++#
++# @size: Size of the virtual disk in bytes
++##
++{ 'struct': 'BlockdevCreateOptionsVitastor',
++  'data': { 'location':         'BlockdevOptionsVitastor',
++            'size':             'size' } }
++
+ ##
+ # @BlockdevVmdkSubformat:
+ #
+@@ -5462,6 +5499,7 @@
+       'ssh':            'BlockdevCreateOptionsSsh',
+       'vdi':            'BlockdevCreateOptionsVdi',
+       'vhdx':           'BlockdevCreateOptionsVhdx',
++      'vitastor':       'BlockdevCreateOptionsVitastor',
+       'vmdk':           'BlockdevCreateOptionsVmdk',
+       'vpc':            'BlockdevCreateOptionsVpc'
+   } }
+diff --git a/scripts/meson-buildoptions.sh b/scripts/meson-buildoptions.sh
+index 3e8e00852b..45aff3b6a9 100644
+--- a/scripts/meson-buildoptions.sh
++++ b/scripts/meson-buildoptions.sh
+@@ -175,6 +175,7 @@ meson_options_help() {
+   printf "%s\n" '  qga-vss         build QGA VSS support (broken with MinGW)'
+   printf "%s\n" '  qpl             Query Processing Library support'
+   printf "%s\n" '  rbd             Ceph block device driver'
++  printf "%s\n" '  vitastor        Vitastor block device driver'
+   printf "%s\n" '  rdma            Enable RDMA-based migration'
+   printf "%s\n" '  replication     replication support'
+   printf "%s\n" '  rust            Rust support'
+@@ -458,6 +459,8 @@ _meson_option_parse() {
+     --disable-qpl) printf "%s" -Dqpl=disabled ;;
+     --enable-rbd) printf "%s" -Drbd=enabled ;;
+     --disable-rbd) printf "%s" -Drbd=disabled ;;
++    --enable-vitastor) printf "%s" -Dvitastor=enabled ;;
++    --disable-vitastor) printf "%s" -Dvitastor=disabled ;;
+     --enable-rdma) printf "%s" -Drdma=enabled ;;
+     --disable-rdma) printf "%s" -Drdma=disabled ;;
+     --enable-relocatable) printf "%s" -Drelocatable=true ;;

rpm/build-tarball.sh

@@ -7,22 +7,24 @@ set -e
 VITASTOR=$(dirname $0)
 VITASTOR=$(realpath "$VITASTOR/..")
 
-EL=$(rpm --eval '%dist')
-if [ "$EL" = ".el8" ]; then
+REL=$(rpm --eval '%dist')
+REL=${REL##.}
+if [ "$REL" = "el8" ]; then
     # CentOS 8
     . /opt/rh/gcc-toolset-9/enable
-elif [ "$EL" = ".el7" ]; then
+elif [ "$REL" = "el7" ]; then
     # CentOS 7
     . /opt/rh/devtoolset-9/enable
 fi
 cd ~/rpmbuild/SPECS
 rpmbuild -bp fio.spec
 cd $VITASTOR
-VER=$(grep ^Version: rpm/vitastor-el7.spec | awk '{print $2}')
+VER=$(grep ^Version: rpm/vitastor-$REL.spec | awk '{print $2}')
+rm -rf fio
 ln -s ~/rpmbuild/BUILD/fio*/ fio
 sh copy-fio-includes.sh
 rm fio
 mv fio-copy fio
 FIO=`rpm -qi fio | perl -e 'while(<>) { /^Epoch[\s:]+(\S+)/ && print "$1:"; /^Version[\s:]+(\S+)/ && print $1; /^Release[\s:]+(\S+)/ && print "-$1"; }'`
-perl -i -pe 's/(Requires:\s*fio)([^\n]+)?/$1 = '$FIO'/' $VITASTOR/rpm/vitastor-el$EL.spec
-tar --transform "s#^#vitastor-$VER/#" --exclude 'rpm/*.rpm' -czf $VITASTOR/../vitastor-$VER$(rpm --eval '%dist').tar.gz *
+perl -i -pe 's/(Requires:\s*fio)([^\n]+)?/$1 = '$FIO'/' $VITASTOR/rpm/vitastor-$REL.spec
+tar --transform "s#^#vitastor-$VER/#" --exclude 'rpm/*.rpm' -czf $VITASTOR/../vitastor-$VER.$REL.tar.gz $(ls | grep -v packages)

rpm/vitastor-build.sh

@@ -0,0 +1,16 @@
+#!/bin/bash
+
+set -e -x
+REL=$(rpm --eval '%dist')
+REL=${REL##.}
+cd /root/vitastor/rpm
+./build-tarball.sh
+VER=$(grep ^Version: vitastor-$REL.spec | awk '{print $2}')
+cp /root/vitastor-$VER.$REL.tar.gz ~/rpmbuild/SOURCES
+cp vitastor-$REL.spec ~/rpmbuild/SPECS/vitastor.spec
+cd ~/rpmbuild/SPECS/
+rpmbuild -ba vitastor.spec
+mkdir -p /root/vitastor/packages/vitastor-$REL
+rm -rf /root/vitastor/packages/vitastor-$REL/*
+cp ~/rpmbuild/RPMS/*/*vitastor* /root/vitastor/packages/vitastor-$REL/
+cp ~/rpmbuild/SRPMS/vitastor* /root/vitastor/packages/vitastor-$REL/

rpm/vitastor-el7.Dockerfile

@@ -1,5 +1,8 @@
 # Build packages for CentOS 7 inside a container
-# cd ..; podman build -t vitastor-el7 -v `pwd`/packages:/root/packages -f rpm/vitastor-el7.Dockerfile .
+# cd ..
+# docker build -t vitastor-buildenv:el7 -f rpm/vitastor-el7.Dockerfile .
+# docker run -i --rm -v ./:/root/vitastor vitastor-buildenv:el7 /root/vitastor/rpm/vitastor-build.sh
+
 # localedef -i ru_RU -f UTF-8 ru_RU.UTF-8
 
 FROM centos:7
@@ -7,7 +10,9 @@ FROM centos:7
 WORKDIR /root
 
 RUN rm -f /etc/yum.repos.d/CentOS-Media.repo
+RUN sed -i 's/^mirrorlist=/#mirrorlist=/; s!#baseurl=http://mirror.centos.org/centos/\$releasever!baseurl=http://vault.centos.org/7.9.2009!' /etc/yum.repos.d/*.repo
 RUN yum -y --enablerepo=extras install centos-release-scl epel-release yum-utils rpm-build
+RUN perl -i -pe 's!mirrorlist=!#mirrorlist=!s; s!#\s*baseurl=http://mirror.centos.org!baseurl=http://vault.centos.org!' /etc/yum.repos.d/CentOS-SCLo-scl*.repo
 RUN yum -y install https://vitastor.io/rpms/centos/7/vitastor-release-1.0-1.el7.noarch.rpm
 RUN yum -y install devtoolset-9-gcc-c++ devtoolset-9-libatomic-devel gcc make cmake gperftools-devel \
     fio rh-nodejs12 jerasure-devel libisa-l-devel gf-complete-devel rdma-core-devel libnl3-devel
@@ -16,32 +21,3 @@ RUN rpm --nomd5 -i fio*.src.rpm
 RUN rm -f /etc/yum.repos.d/CentOS-Media.repo
 RUN cd ~/rpmbuild/SPECS && yum-builddep -y fio.spec
 RUN yum -y install cmake3
-
-ADD https://vitastor.io/rpms/liburing-el7/liburing-0.7-2.el7.src.rpm /root
-
-RUN set -e; \
-    rpm -i liburing*.src.rpm; \
-    cd ~/rpmbuild/SPECS/; \
-    . /opt/rh/devtoolset-9/enable; \
-    rpmbuild -ba liburing.spec; \
-    mkdir -p /root/packages/liburing-el7; \
-    rm -rf /root/packages/liburing-el7/*; \
-    cp ~/rpmbuild/RPMS/*/liburing* /root/packages/liburing-el7/; \
-    cp ~/rpmbuild/SRPMS/liburing* /root/packages/liburing-el7/
-
-RUN rpm -i `ls /root/packages/liburing-el7/liburing-*.x86_64.rpm | grep -v debug`
-
-ADD . /root/vitastor
-
-RUN set -e; \
-    cd /root/vitastor/rpm; \
-    sh build-tarball.sh; \
-    VER=$(grep ^Version: vitastor-el7.spec | awk '{print $2}'); \
-    cp /root/vitastor-$VER.el7.tar.gz ~/rpmbuild/SOURCES; \
-    cp vitastor-el7.spec ~/rpmbuild/SPECS/vitastor.spec; \
-    cd ~/rpmbuild/SPECS/; \
-    rpmbuild -ba vitastor.spec; \
-    mkdir -p /root/packages/vitastor-el7; \
-    rm -rf /root/packages/vitastor-el7/*; \
-    cp ~/rpmbuild/RPMS/*/*vitastor* /root/packages/vitastor-el7/; \
-    cp ~/rpmbuild/SRPMS/vitastor* /root/packages/vitastor-el7/

rpm/vitastor-el7.spec

@@ -1,13 +1,12 @@
 Name:           vitastor
-Version:        2.2.2
+Version:        2.3.0
 Release:        1%{?dist}
 Summary:        Vitastor, a fast software-defined clustered block storage
 
 License:        Vitastor Network Public License 1.1
 URL:            https://vitastor.io/
-Source0:        vitastor-2.2.2.el7.tar.gz
+Source0:        vitastor-2.3.0.el7.tar.gz
 
-BuildRequires:  liburing-devel >= 0.6
 BuildRequires:  gperftools-devel
 BuildRequires:  devtoolset-9-gcc-c++
 BuildRequires:  rh-nodejs12
@@ -35,8 +34,6 @@ size with configurable redundancy (replication or erasure codes/XOR).
 Summary:        Vitastor - OSD
 Requires:       libJerasure2
 Requires:       libisa-l
-Requires:       liburing >= 0.6
-Requires:       liburing < 2
 Requires:       vitastor-client = %{version}-%{release}
 Requires:       util-linux
 Requires:       parted
@@ -60,8 +57,6 @@ scheduling cluster-level operations.
 
 %package -n vitastor-client
 Summary:        Vitastor - client
-Requires:       liburing >= 0.6
-Requires:       liburing < 2
 
 
 %description -n vitastor-client
@@ -82,7 +77,7 @@ Vitastor library headers for development.
 Summary:        Vitastor - fio drivers
 Group:          Development/Libraries
 Requires:       vitastor-client = %{version}-%{release}
-Requires:       fio = 3.7-1.el7
+Requires:       fio = 3.7-2.el7
 
 
 %description -n vitastor-fio
@@ -169,6 +164,7 @@ chown vitastor:vitastor /var/lib/vitastor
 
 %files -n vitastor-client
 %_bindir/vitastor-nbd
+%_bindir/vitastor-ublk
 %_bindir/vitastor-nfs
 %_bindir/vitastor-cli
 %_bindir/vitastor-rm

rpm/vitastor-el8.Dockerfile

@@ -1,5 +1,7 @@
 # Build packages for CentOS 8 inside a container
-# cd ..; podman build -t vitastor-el8 -v `pwd`/packages:/root/packages -f rpm/vitastor-el8.Dockerfile .
+# cd ..
+# docker build -t vitastor-buildenv:el8 -f rpm/vitastor-el8.Dockerfile .
+# docker run -i --rm -v ./:/root/vitastor vitastor-buildenv:el8 /root/vitastor/rpm/vitastor-build.sh
 
 FROM centos:8
 
@@ -15,32 +17,3 @@ RUN dnf -y install gcc-toolset-9 gcc-toolset-9-gcc-c++ gperftools-devel \
 RUN dnf download --source fio
 RUN rpm --nomd5 -i fio*.src.rpm
 RUN cd ~/rpmbuild/SPECS && dnf builddep -y --enablerepo=powertools --spec fio.spec
-
-ADD https://vitastor.io/rpms/liburing-el7/liburing-0.7-2.el7.src.rpm /root
-
-RUN set -e; \
-    rpm -i liburing*.src.rpm; \
-    cd ~/rpmbuild/SPECS/; \
-    . /opt/rh/gcc-toolset-9/enable; \
-    rpmbuild -ba liburing.spec; \
-    mkdir -p /root/packages/liburing-el8; \
-    rm -rf /root/packages/liburing-el8/*; \
-    cp ~/rpmbuild/RPMS/*/liburing* /root/packages/liburing-el8/; \
-    cp ~/rpmbuild/SRPMS/liburing* /root/packages/liburing-el8/
-
-RUN rpm -i `ls /root/packages/liburing-el8/liburing-*.x86_64.rpm | grep -v debug`
-
-ADD . /root/vitastor
-
-RUN set -e; \
-    cd /root/vitastor/rpm; \
-    sh build-tarball.sh; \
-    VER=$(grep ^Version: vitastor-el8.spec | awk '{print $2}'); \
-    cp /root/vitastor-$VER.el8.tar.gz ~/rpmbuild/SOURCES; \
-    cp vitastor-el8.spec ~/rpmbuild/SPECS/vitastor.spec; \
-    cd ~/rpmbuild/SPECS/; \
-    rpmbuild -ba vitastor.spec; \
-    mkdir -p /root/packages/vitastor-el8; \
-    rm -rf /root/packages/vitastor-el8/*; \
-    cp ~/rpmbuild/RPMS/*/*vitastor* /root/packages/vitastor-el8/; \
-    cp ~/rpmbuild/SRPMS/vitastor* /root/packages/vitastor-el8/

rpm/vitastor-el8.spec

@@ -1,13 +1,12 @@
 Name:           vitastor
-Version:        2.2.2
+Version:        2.3.0
 Release:        1%{?dist}
 Summary:        Vitastor, a fast software-defined clustered block storage
 
 License:        Vitastor Network Public License 1.1
 URL:            https://vitastor.io/
-Source0:        vitastor-2.2.2.el8.tar.gz
+Source0:        vitastor-2.3.0.el8.tar.gz
 
-BuildRequires:  liburing-devel >= 0.6
 BuildRequires:  gperftools-devel
 BuildRequires:  gcc-toolset-9-gcc-c++
 BuildRequires:  nodejs >= 10
@@ -34,8 +33,6 @@ size with configurable redundancy (replication or erasure codes/XOR).
 Summary:        Vitastor - OSD
 Requires:       libJerasure2
 Requires:       libisa-l
-Requires:       liburing >= 0.6
-Requires:       liburing < 2
 Requires:       vitastor-client = %{version}-%{release}
 Requires:       util-linux
 Requires:       parted
@@ -58,8 +55,6 @@ scheduling cluster-level operations.
 
 %package -n vitastor-client
 Summary:        Vitastor - client
-Requires:       liburing >= 0.6
-Requires:       liburing < 2
 
 
 %description -n vitastor-client
@@ -80,7 +75,7 @@ Vitastor library headers for development.
 Summary:        Vitastor - fio drivers
 Group:          Development/Libraries
 Requires:       vitastor-client = %{version}-%{release}
-Requires:       fio = 3.7-3.el8
+Requires:       fio = 3.19-3.el8
 
 
 %description -n vitastor-fio
@@ -166,6 +161,7 @@ chown vitastor:vitastor /var/lib/vitastor
 
 %files -n vitastor-client
 %_bindir/vitastor-nbd
+%_bindir/vitastor-ublk
 %_bindir/vitastor-nfs
 %_bindir/vitastor-cli
 %_bindir/vitastor-rm

rpm/vitastor-el9.Dockerfile

@@ -1,5 +1,7 @@
 # Build packages for AlmaLinux 9 inside a container
-# cd ..; podman build -t vitastor-el9 -v `pwd`/packages:/root/packages -f rpm/vitastor-el9.Dockerfile .
+# cd ..
+# docker build -t vitastor-buildenv:el9 -f rpm/vitastor-el9.Dockerfile .
+# docker run -i --rm -v ./:/root/vitastor vitastor-buildenv:el9 /root/vitastor/rpm/vitastor-build.sh
 
 FROM almalinux:9
 
@@ -8,22 +10,7 @@ WORKDIR /root
 RUN sed -i 's/enabled=0/enabled=1/' /etc/yum.repos.d/*.repo
 RUN dnf -y install epel-release dnf-plugins-core
 RUN dnf -y install https://vitastor.io/rpms/centos/9/vitastor-release-1.0-1.el9.noarch.rpm
-RUN dnf -y install gcc-c++ gperftools-devel fio nodejs rpm-build jerasure-devel libisa-l-devel gf-complete-devel rdma-core-devel libarchive liburing-devel cmake libnl3-devel
+RUN dnf -y install gcc-c++ gperftools-devel fio nodejs rpm-build jerasure-devel libisa-l-devel gf-complete-devel rdma-core-devel libarchive cmake libnl3-devel
 RUN dnf download --source fio
 RUN rpm --nomd5 -i fio*.src.rpm
 RUN cd ~/rpmbuild/SPECS && dnf builddep -y --spec fio.spec
-
-ADD . /root/vitastor
-
-RUN set -e; \
-    cd /root/vitastor/rpm; \
-    sh build-tarball.sh; \
-    VER=$(grep ^Version: vitastor-el9.spec | awk '{print $2}'); \
-    cp /root/vitastor-$VER.el9.tar.gz ~/rpmbuild/SOURCES; \
-    cp vitastor-el9.spec ~/rpmbuild/SPECS/vitastor.spec; \
-    cd ~/rpmbuild/SPECS/; \
-    rpmbuild -ba vitastor.spec; \
-    mkdir -p /root/packages/vitastor-el9; \
-    rm -rf /root/packages/vitastor-el9/*; \
-    cp ~/rpmbuild/RPMS/*/*vitastor* /root/packages/vitastor-el9/; \
-    cp ~/rpmbuild/SRPMS/vitastor* /root/packages/vitastor-el9/

rpm/vitastor-el9.spec

@@ -1,13 +1,12 @@
 Name:           vitastor
-Version:        2.2.2
+Version:        2.3.0
 Release:        1%{?dist}
 Summary:        Vitastor, a fast software-defined clustered block storage
 
 License:        Vitastor Network Public License 1.1
 URL:            https://vitastor.io/
-Source0:        vitastor-2.2.2.el9.tar.gz
+Source0:        vitastor-2.3.0.el9.tar.gz
 
-BuildRequires:  liburing-devel >= 0.6
 BuildRequires:  gperftools-devel
 BuildRequires:  gcc-c++
 BuildRequires:  nodejs >= 10
@@ -159,6 +158,7 @@ chown vitastor:vitastor /var/lib/vitastor
 
 %files -n vitastor-client
 %_bindir/vitastor-nbd
+%_bindir/vitastor-ublk
 %_bindir/vitastor-nfs
 %_bindir/vitastor-cli
 %_bindir/vitastor-rm

src/CMakeLists.txt

@@ -12,20 +12,30 @@ set(WITH_QEMU false CACHE BOOL "Build QEMU driver inside Vitastor source tree")
 set(WITH_FIO true CACHE BOOL "Build FIO driver")
 set(QEMU_PLUGINDIR qemu CACHE STRING "QEMU plugin directory suffix (qemu-kvm on RHEL)")
 set(WITH_ASAN false CACHE BOOL "Build with AddressSanitizer")
+set(WITH_SYSTEM_LIBURING false CACHE BOOL "Use system liburing")
 if("${CMAKE_INSTALL_PREFIX}" MATCHES "^/usr/local/?$")
 	if(EXISTS "/etc/debian_version")
 		set(CMAKE_INSTALL_LIBDIR "lib/${CMAKE_LIBRARY_ARCHITECTURE}")
 	endif()
 	set(CMAKE_INSTALL_RPATH "${CMAKE_INSTALL_PREFIX}/${CMAKE_INSTALL_LIBDIR}")
 endif()
+set(ENABLE_COVERAGE false CACHE BOOL "Enable code coverage")
 
-add_definitions(-DVITASTOR_VERSION="2.2.2")
-add_definitions(-D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -Wall -Wno-sign-compare -Wno-comment -Wno-parentheses -Wno-pointer-arith -fdiagnostics-color=always -fno-omit-frame-pointer -I ${CMAKE_SOURCE_DIR}/src)
+add_definitions(-DVITASTOR_VERSION="2.3.0")
+add_definitions(-D_GNU_SOURCE -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -Wall -Wno-sign-compare -Wno-comment -Wno-parentheses -Wno-pointer-arith -fdiagnostics-color=always -fno-omit-frame-pointer -fvisibility=hidden -I ${CMAKE_SOURCE_DIR}/src)
 add_link_options(-fno-omit-frame-pointer)
 if (${WITH_ASAN})
 	add_definitions(-fsanitize=address)
 	add_link_options(-fsanitize=address -fno-omit-frame-pointer)
 endif (${WITH_ASAN})
+set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -fvisibility-inlines-hidden")
+set(CMAKE_CXX_FLAGS_MINSIZEREL "${CMAKE_CXX_FLAGS_MINSIZEREL} -fvisibility-inlines-hidden")
+set(CMAKE_CXX_FLAGS_RELWITHDEBINFO "${CMAKE_CXX_FLAGS_RELWITHDEBINFO} -fvisibility-inlines-hidden")
+
+if (${ENABLE_COVERAGE})
+	add_definitions(-coverage)
+	add_link_options(-coverage)
+endif()
 
 set(CMAKE_BUILD_TYPE RelWithDebInfo)
 string(REGEX REPLACE "([\\/\\-]O)[^ \t\r\n]*" "\\13" CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE}")
@@ -49,7 +59,6 @@ endmacro(install_symlink)
 check_include_file("linux/nbd-netlink.h" HAVE_NBD_NETLINK_H)
 
 find_package(PkgConfig)
-pkg_check_modules(LIBURING REQUIRED liburing)
 if (${WITH_QEMU})
 	pkg_check_modules(GLIB REQUIRED glib-2.0)
 endif (${WITH_QEMU})
@@ -66,13 +75,14 @@ if (RDMACM_LIBRARIES)
 	add_definitions(-DWITH_RDMACM)
 endif (RDMACM_LIBRARIES)
 
-add_custom_target(build_tests)
-add_custom_target(test
-	COMMAND
-	echo leak:tcmalloc > ${CMAKE_CURRENT_BINARY_DIR}/lsan-suppress.txt &&
-	env LSAN_OPTIONS=suppressions=${CMAKE_CURRENT_BINARY_DIR}/lsan-suppress.txt ${CMAKE_CTEST_COMMAND}
-)
-add_dependencies(test build_tests)
+if (${WITH_SYSTEM_LIBURING})
+	pkg_check_modules(LIBURING REQUIRED liburing>=2.10)
+	include_directories(${LIBURING_INCLUDE_DIRS})
+else()
+	include_directories(${CMAKE_SOURCE_DIR}/src/liburing/include)
+	add_subdirectory(liburing)
+	set(LIBURING_LIBRARIES uring)
+endif (${WITH_SYSTEM_LIBURING})
 
 include_directories(
 	../
@@ -86,7 +96,6 @@ include_directories(
 	${CMAKE_SOURCE_DIR}/src/test
 	${CMAKE_SOURCE_DIR}/src/util
 	/usr/include/jerasure
-	${LIBURING_INCLUDE_DIRS}
 	${IBVERBS_INCLUDE_DIRS}
 )
 
@@ -101,7 +110,7 @@ add_subdirectory(test)
 
 ### Install
 
-install(TARGETS vitastor-osd vitastor-disk vitastor-nbd vitastor-nfs vitastor-cli vitastor-kv vitastor-kv-stress RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
+install(TARGETS vitastor-osd vitastor-disk vitastor-nbd vitastor-ublk vitastor-nfs vitastor-cli vitastor-kv vitastor-kv-stress RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
 install_symlink(vitastor-disk ${CMAKE_INSTALL_PREFIX}/${CMAKE_INSTALL_BINDIR}/vitastor-dump-journal)
 install_symlink(vitastor-cli ${CMAKE_INSTALL_PREFIX}/${CMAKE_INSTALL_BINDIR}/vitastor-rm)
 install_symlink(vitastor-cli ${CMAKE_INSTALL_PREFIX}/${CMAKE_INSTALL_BINDIR}/vita)

src/blockstore/CMakeLists.txt

@@ -2,14 +2,17 @@ cmake_minimum_required(VERSION 2.8.12)
 
 project(vitastor)
 
-# libvitastor_blk.so
-add_library(vitastor_blk SHARED
-	../util/allocator.cpp blockstore.cpp blockstore_impl.cpp blockstore_disk.cpp blockstore_init.cpp blockstore_open.cpp blockstore_journal.cpp blockstore_read.cpp
-	blockstore_write.cpp blockstore_sync.cpp blockstore_stable.cpp blockstore_rollback.cpp blockstore_flush.cpp ../util/crc32c.c ../util/ringloop.cpp
+# libvitastor_blk.a
+add_library(vitastor_blk STATIC
+	../util/allocator.cpp ../util/crc32c.c ../util/ringloop.cpp
+	multilist.cpp blockstore_heap.cpp blockstore_disk.cpp
+	blockstore.cpp blockstore_impl.cpp blockstore_init.cpp blockstore_open.cpp
+	blockstore_flush.cpp blockstore_read.cpp blockstore_stable.cpp blockstore_sync.cpp blockstore_write.cpp
 )
+target_compile_options(vitastor_blk PUBLIC -fPIC)
 target_link_libraries(vitastor_blk
 	${LIBURING_LIBRARIES}
-	tcmalloc_minimal
+	${ISAL_LIBRARIES}
 	# for timerfd_manager
 	vitastor_common
 )

src/blockstore/blockstore.cpp

@@ -3,7 +3,7 @@
 
 #include "blockstore_impl.h"
 
-blockstore_t::blockstore_t(blockstore_config_t & config, ring_loop_t *ringloop, timerfd_manager_t *tfd)
+blockstore_t::blockstore_t(blockstore_config_t & config, ring_loop_i *ringloop, timerfd_manager_t *tfd)
 {
     impl = new blockstore_impl_t(config, ringloop, tfd);
 }
@@ -48,9 +48,9 @@ int blockstore_t::read_bitmap(object_id oid, uint64_t target_version, void *bitm
     return impl->read_bitmap(oid, target_version, bitmap, result_version);
 }
 
-std::map<uint64_t, uint64_t> & blockstore_t::get_inode_space_stats()
+const std::map<uint64_t, uint64_t> & blockstore_t::get_inode_space_stats()
 {
-    return impl->inode_space_stats;
+    return impl->get_inode_space_stats();
 }
 
 void blockstore_t::dump_diagnostics()
@@ -82,8 +82,3 @@ uint32_t blockstore_t::get_bitmap_granularity()
 {
     return impl->get_bitmap_granularity();
 }
-
-void blockstore_t::set_no_inode_stats(const std::vector<uint64_t> & pool_ids)
-{
-    impl->set_no_inode_stats(pool_ids);
-}

src/blockstore/blockstore.h

@@ -22,17 +22,20 @@
 #define DIRECT_IO_ALIGNMENT 512
 #endif
 
-// Memory allocation alignment (page size is usually optimal)
-#ifndef MEM_ALIGNMENT
-#define MEM_ALIGNMENT 4096
-#endif
-
 // Default block size is 128 KB, current allowed range is 4K - 128M
 #define DEFAULT_DATA_BLOCK_ORDER 17
 #define MIN_DATA_BLOCK_SIZE 4*1024
 #define MAX_DATA_BLOCK_SIZE 128*1024*1024
 #define DEFAULT_BITMAP_GRANULARITY 4096
 
+#define MIN_JOURNAL_SIZE 1024*1024
+
+// "VITAstor"
+#define BLOCKSTORE_META_MAGIC_V1 0x726F747341544956l
+#define BLOCKSTORE_META_FORMAT_V1 1
+#define BLOCKSTORE_META_FORMAT_V2 2
+#define BLOCKSTORE_META_FORMAT_HEAP 3
+
 #define BS_OP_MIN 1
 #define BS_OP_READ 1
 #define BS_OP_WRITE 2
@@ -42,13 +45,18 @@
 #define BS_OP_DELETE 6
 #define BS_OP_LIST 7
 #define BS_OP_ROLLBACK 8
-#define BS_OP_SYNC_STAB_ALL 9
-#define BS_OP_MAX 9
+#define BS_OP_MAX 8
 
 #define BS_OP_PRIVATE_DATA_SIZE 256
 
 /*
 
+All operations may be submitted in any order, because reads only see completed writes,
+syncs only sync completed writes and writes don't depend on each other.
+
+The only restriction is that the external code MUST NOT submit multiple writes for one
+object in parallel. This is a natural restriction because `version` numbers are used though.
+
 Blockstore opcode documentation:
 
 ## BS_OP_READ / BS_OP_WRITE / BS_OP_WRITE_STABLE
@@ -113,14 +121,6 @@ Input:
 Output:
 - retval = 0 or negative error number (-ENOENT if no such version for stabilize)
 
-## BS_OP_SYNC_STAB_ALL
-
-ONLY FOR TESTS! Sync and mark all unstable object versions as stable, at once.
-
-Input: Nothing except opcode
-Output:
-- retval = 0 or negative error number (-EINVAL)
-
 ## BS_OP_LIST
 
 Get a list of all objects in this Blockstore.
@@ -144,10 +144,10 @@ Output:
 
 */
 
-struct blockstore_op_t
+struct __attribute__ ((visibility("default"))) blockstore_op_t
 {
     // operation
-    uint64_t opcode;
+    uint64_t opcode = 0;
     // finish callback
     std::function<void (blockstore_op_t*)> callback;
     union __attribute__((__packed__))
@@ -171,9 +171,9 @@ struct blockstore_op_t
             uint32_t list_stable_limit;
         };
     };
-    void *buf;
-    void *bitmap;
-    int retval;
+    uint8_t *buf = NULL;
+    uint8_t *bitmap = NULL;
+    int retval = 0;
 
     uint8_t private_data[BS_OP_PRIVATE_DATA_SIZE];
 };
@@ -182,11 +182,11 @@ typedef std::map<std::string, std::string> blockstore_config_t;
 
 class blockstore_impl_t;
 
-class blockstore_t
+class __attribute__((visibility("default"))) blockstore_t
 {
     blockstore_impl_t *impl;
 public:
-    blockstore_t(blockstore_config_t & config, ring_loop_t *ringloop, timerfd_manager_t *tfd);
+    blockstore_t(blockstore_config_t & config, ring_loop_i *ringloop, timerfd_manager_t *tfd);
     ~blockstore_t();
 
     // Update configuration
@@ -214,10 +214,7 @@ public:
     int read_bitmap(object_id oid, uint64_t target_version, void *bitmap, uint64_t *result_version = NULL);
 
     // Get per-inode space usage statistics
-    std::map<uint64_t, uint64_t> & get_inode_space_stats();
-
-    // Set per-pool no_inode_stats
-    void set_no_inode_stats(const std::vector<uint64_t> & pool_ids);
+    const std::map<uint64_t, uint64_t> & get_inode_space_stats();
 
     // Print diagnostics to stdout
     void dump_diagnostics();

src/blockstore/blockstore_disk.cpp

@@ -2,11 +2,14 @@
 // License: VNPL-1.1 (see README.md for details)
 
 #include <sys/file.h>
+#include <sys/ioctl.h>
+#include <unistd.h>
 
 #include <stdexcept>
 
-#include "blockstore_impl.h"
+#include "blockstore.h"
 #include "blockstore_disk.h"
+#include "blockstore_heap.h"
 #include "str_util.h"
 #include "allocator.h"
 
@@ -44,8 +47,11 @@ void blockstore_disk_t::parse_config(std::map<std::string, std::string> & config
     disk_alignment = parse_size(config["disk_alignment"]);
     journal_block_size = parse_size(config["journal_block_size"]);
     meta_block_size = parse_size(config["meta_block_size"]);
+    meta_block_target_free_space = parse_size(config["meta_block_target_free_space"]);
     bitmap_granularity = parse_size(config["bitmap_granularity"]);
     meta_format = stoull_full(config["meta_format"]);
+    atomic_write_size = (config.find("atomic_write_size") != config.end()
+        ? parse_size(config["atomic_write_size"]) : 4096);
     if (config.find("data_io") == config.end() &&
         config.find("meta_io") == config.end() &&
         config.find("journal_io") == config.end())
@@ -90,12 +96,28 @@ void blockstore_disk_t::parse_config(std::map<std::string, std::string> & config
     if (!min_discard_size)
         min_discard_size = 1024*1024;
     discard_granularity = parse_size(config["discard_granularity"]);
+    inmemory_meta = config["inmemory_metadata"] != "false" && config["inmemory_metadata"] != "0" &&
+        config["inmemory_metadata"] != "no";
+    inmemory_journal = config["inmemory_journal"] != "false" && config["inmemory_journal"] != "0" &&
+        config["inmemory_journal"] != "no";
+    disable_data_fsync = config["disable_data_fsync"] == "true" || config["disable_data_fsync"] == "1" || config["disable_data_fsync"] == "yes";
+    disable_meta_fsync = config["disable_meta_fsync"] == "true" || config["disable_meta_fsync"] == "1" || config["disable_meta_fsync"] == "yes";
+    disable_journal_fsync = config["disable_journal_fsync"] == "true" || config["disable_journal_fsync"] == "1" || config["disable_journal_fsync"] == "yes";
+    if (mock_mode)
+    {
+        data_device_size = parse_size(config["data_device_size"]);
+        data_device_sect = parse_size(config["data_device_sect"]);
+        meta_device_size = parse_size(config["meta_device_size"]);
+        meta_device_sect = parse_size(config["meta_device_sect"]);
+        journal_device_size = parse_size(config["journal_device_size"]);
+        journal_device_sect = parse_size(config["journal_device_sect"]);
+    }
     // Validate
     if (!data_block_size)
     {
         data_block_size = (1 << DEFAULT_DATA_BLOCK_ORDER);
     }
-    if ((block_order = is_power_of_two(data_block_size)) >= 64 || data_block_size < MIN_DATA_BLOCK_SIZE || data_block_size >= MAX_DATA_BLOCK_SIZE)
+    if (is_power_of_two(data_block_size) >= 64 || data_block_size < MIN_DATA_BLOCK_SIZE || data_block_size >= MAX_DATA_BLOCK_SIZE)
     {
         throw std::runtime_error("Bad block size");
     }
@@ -131,6 +153,14 @@ void blockstore_disk_t::parse_config(std::map<std::string, std::string> & config
     {
         throw std::runtime_error("meta_block_size must not exceed "+std::to_string(MAX_DATA_BLOCK_SIZE));
     }
+    if (!meta_block_target_free_space)
+    {
+        meta_block_target_free_space = 800;
+    }
+    if (meta_block_target_free_space >= meta_block_size)
+    {
+        throw std::runtime_error("meta_block_target_free_space must not exceed "+std::to_string(meta_block_size));
+    }
     if (data_offset % disk_alignment)
     {
         throw std::runtime_error("data_offset must be a multiple of disk_alignment = "+std::to_string(disk_alignment));
@@ -179,17 +209,29 @@ void blockstore_disk_t::parse_config(std::map<std::string, std::string> & config
     {
         throw std::runtime_error("journal_offset must be a multiple of journal_block_size = "+std::to_string(journal_block_size));
     }
+    if (!meta_format)
+    {
+        meta_format = BLOCKSTORE_META_FORMAT_HEAP;
+    }
+    if (meta_device == data_device)
+    {
+        disable_meta_fsync = disable_data_fsync;
+    }
+    if (journal_device == meta_device)
+    {
+        disable_journal_fsync = disable_meta_fsync;
+    }
 }
 
-void blockstore_disk_t::calc_lengths(bool skip_meta_check)
+void blockstore_disk_t::calc_lengths()
 {
     // data
     data_len = data_device_size - data_offset;
-    if (data_fd == meta_fd && data_offset < meta_offset)
+    if (data_device == meta_device && data_offset < meta_offset)
     {
         data_len = meta_offset - data_offset;
     }
-    if (data_fd == journal_fd && data_offset < journal_offset)
+    if (data_device == journal_device && data_offset < journal_offset)
     {
         data_len = data_len < journal_offset-data_offset
             ? data_len : journal_offset-data_offset;
@@ -204,23 +246,23 @@ void blockstore_disk_t::calc_lengths(bool skip_meta_check)
         data_len = cfg_data_size;
     }
     // meta
-    uint64_t meta_area_size = (meta_fd == data_fd ? data_device_size : meta_device_size) - meta_offset;
-    if (meta_fd == data_fd && meta_offset <= data_offset)
+    meta_area_size = (meta_device == data_device ? data_device_size : meta_device_size) - meta_offset;
+    if (meta_device == data_device && meta_offset <= data_offset)
     {
         meta_area_size = data_offset - meta_offset;
     }
-    if (meta_fd == journal_fd && meta_offset <= journal_offset)
+    if (meta_device == journal_device && meta_offset <= journal_offset)
     {
         meta_area_size = meta_area_size < journal_offset-meta_offset
             ? meta_area_size : journal_offset-meta_offset;
     }
     // journal
-    journal_len = (journal_fd == data_fd ? data_device_size : (journal_fd == meta_fd ? meta_device_size : journal_device_size)) - journal_offset;
-    if (journal_fd == data_fd && journal_offset <= data_offset)
+    journal_len = (journal_device == data_device ? data_device_size : (journal_device == meta_device ? meta_device_size : journal_device_size)) - journal_offset;
+    if (journal_device == data_device && journal_offset <= data_offset)
     {
         journal_len = data_offset - journal_offset;
     }
-    if (journal_fd == meta_fd && journal_offset <= meta_offset)
+    if (journal_device == meta_device && journal_offset <= meta_offset)
     {
         journal_len = journal_len < meta_offset-journal_offset
             ? journal_len : meta_offset-journal_offset;
@@ -230,37 +272,37 @@ void blockstore_disk_t::calc_lengths(bool skip_meta_check)
     clean_entry_bitmap_size = data_block_size / bitmap_granularity / 8;
     clean_dyn_size = clean_entry_bitmap_size*2 + (csum_block_size
         ? data_block_size/csum_block_size*(data_csum_type & 0xFF) : 0);
-    clean_entry_size = sizeof(clean_disk_entry) + clean_dyn_size + 4 /*entry_csum*/;
-    meta_len = (1 + (block_count - 1 + meta_block_size / clean_entry_size) / (meta_block_size / clean_entry_size)) * meta_block_size;
-    bool new_doesnt_fit = (!meta_format && !skip_meta_check && meta_area_size < meta_len && !data_csum_type);
-    if (meta_format == BLOCKSTORE_META_FORMAT_V1 || new_doesnt_fit)
+    if (meta_format == BLOCKSTORE_META_FORMAT_HEAP)
     {
-        uint64_t clean_entry_v0_size = sizeof(clean_disk_entry) + 2*clean_entry_bitmap_size;
-        uint64_t meta_v0_len = (1 + (block_count - 1 + meta_block_size / clean_entry_v0_size)
-            / (meta_block_size / clean_entry_v0_size)) * meta_block_size;
-        if (meta_format == BLOCKSTORE_META_FORMAT_V1 || meta_area_size >= meta_v0_len)
-        {
-            // Old metadata fits.
-            if (new_doesnt_fit)
-            {
-                printf("Warning: Using old metadata format without checksums because the new format"
-                    " doesn't fit into provided area (%ju bytes required, %ju bytes available)\n", meta_len, meta_area_size);
-            }
-            clean_entry_size = clean_entry_v0_size;
-            meta_len = meta_v0_len;
-            meta_format = BLOCKSTORE_META_FORMAT_V1;
-        }
-        else
-            meta_format = BLOCKSTORE_META_FORMAT_V2;
+        uint32_t entries_per_block = ((meta_block_size-meta_block_target_free_space) /
+            (sizeof(heap_object_t) + sizeof(heap_write_t) + clean_dyn_size));
+        min_meta_len = (block_count+entries_per_block-1) / entries_per_block * meta_block_size;
+    }
+    else if (meta_format == BLOCKSTORE_META_FORMAT_V1)
+    {
+        clean_entry_size = 24 /*sizeof(clean_disk_entry)*/ + 2*clean_entry_bitmap_size;
+        min_meta_len = (1 + (block_count - 1 + meta_block_size / clean_entry_size)
+            / (meta_block_size / clean_entry_size)) * meta_block_size;
+    }
+    else if (meta_format == BLOCKSTORE_META_FORMAT_V2)
+    {
+        clean_entry_size = 24 /*sizeof(clean_disk_entry)*/ + clean_dyn_size + 4 /*entry_csum*/;
+        min_meta_len = (1 + (block_count - 1 + meta_block_size / clean_entry_size) / (meta_block_size / clean_entry_size)) * meta_block_size;
     }
     else
-        meta_format = BLOCKSTORE_META_FORMAT_V2;
-    if (!skip_meta_check && meta_area_size < meta_len)
     {
-        throw std::runtime_error("Metadata area is too small, need at least "+std::to_string(meta_len)+" bytes, have only "+std::to_string(meta_area_size)+" bytes");
+        throw std::runtime_error("meta_format = "+std::to_string(meta_format)+" is not supported");
+    }
+}
+
+void blockstore_disk_t::check_lengths()
+{
+    if (meta_area_size < min_meta_len)
+    {
+        throw std::runtime_error("Metadata area is too small, need at least "+std::to_string(min_meta_len)+" bytes, have only "+std::to_string(meta_area_size)+" bytes");
     }
     // requested journal size
-    if (!skip_meta_check && cfg_journal_size > journal_len)
+    if (cfg_journal_size > journal_len)
     {
         throw std::runtime_error("Requested journal_size is too large");
     }
@@ -321,12 +363,19 @@ static int bs_openmode(const std::string & mode)
 
 void blockstore_disk_t::open_data()
 {
-    data_fd = open(data_device.c_str(), bs_openmode(data_io) | O_RDWR);
+    if (data_fd >= 0)
+    {
+        throw std::runtime_error("data device is already opened");
+    }
+    data_fd = mock_mode ? MOCK_DATA_FD : open(data_device.c_str(), bs_openmode(data_io) | O_RDWR);
     if (data_fd == -1)
     {
         throw std::runtime_error("Failed to open data device "+data_device+": "+std::string(strerror(errno)));
     }
-    check_size(data_fd, &data_device_size, &data_device_sect, "data device");
+    if (!mock_mode)
+    {
+        check_size(data_fd, &data_device_size, &data_device_sect, "data device");
+    }
     if (disk_alignment % data_device_sect)
     {
         throw std::runtime_error(
@@ -338,7 +387,7 @@ void blockstore_disk_t::open_data()
     {
         throw std::runtime_error("data_offset exceeds device size = "+std::to_string(data_device_size));
     }
-    if (!disable_flock && flock(data_fd, LOCK_EX|LOCK_NB) != 0)
+    if (!mock_mode && !disable_flock && flock(data_fd, LOCK_EX|LOCK_NB) != 0)
     {
         throw std::runtime_error(std::string("Failed to lock data device: ") + strerror(errno));
     }
@@ -346,19 +395,26 @@ void blockstore_disk_t::open_data()
 
 void blockstore_disk_t::open_meta()
 {
+    if (meta_fd >= 0)
+    {
+        throw std::runtime_error("metadata device is already opened");
+    }
     if (meta_device != data_device || meta_io != data_io)
     {
-        meta_fd = open(meta_device.c_str(), bs_openmode(meta_io) | O_RDWR);
+        meta_fd = mock_mode ? MOCK_META_FD : open(meta_device.c_str(), bs_openmode(meta_io) | O_RDWR);
         if (meta_fd == -1)
         {
             throw std::runtime_error("Failed to open metadata device "+meta_device+": "+std::string(strerror(errno)));
         }
-        check_size(meta_fd, &meta_device_size, &meta_device_sect, "metadata device");
+        if (!mock_mode)
+        {
+            check_size(meta_fd, &meta_device_size, &meta_device_sect, "metadata device");
+        }
         if (meta_offset >= meta_device_size)
         {
             throw std::runtime_error("meta_offset exceeds device size = "+std::to_string(meta_device_size));
         }
-        if (!disable_flock && meta_device != data_device && flock(meta_fd, LOCK_EX|LOCK_NB) != 0)
+        if (!mock_mode && !disable_flock && meta_device != data_device && flock(meta_fd, LOCK_EX|LOCK_NB) != 0)
         {
             throw std::runtime_error(std::string("Failed to lock metadata device: ") + strerror(errno));
         }
@@ -384,15 +440,26 @@ void blockstore_disk_t::open_meta()
 
 void blockstore_disk_t::open_journal()
 {
+    if (journal_fd >= 0)
+    {
+        throw std::runtime_error("journal device is already opened");
+    }
     if (journal_device != meta_device || journal_io != meta_io)
     {
-        journal_fd = open(journal_device.c_str(), bs_openmode(journal_io) | O_RDWR);
+        journal_fd = mock_mode ? MOCK_JOURNAL_FD : open(journal_device.c_str(), bs_openmode(journal_io) | O_RDWR);
         if (journal_fd == -1)
         {
             throw std::runtime_error("Failed to open journal device "+journal_device+": "+std::string(strerror(errno)));
         }
-        check_size(journal_fd, &journal_device_size, &journal_device_sect, "journal device");
-        if (!disable_flock && journal_device != meta_device && flock(journal_fd, LOCK_EX|LOCK_NB) != 0)
+        if (!mock_mode)
+        {
+            check_size(journal_fd, &journal_device_size, &journal_device_sect, "journal device");
+        }
+        if (journal_offset >= journal_device_size)
+        {
+            throw std::runtime_error("journal_offset exceeds device size = "+std::to_string(journal_device_size));
+        }
+        if (!mock_mode && !disable_flock && journal_device != meta_device && flock(journal_fd, LOCK_EX|LOCK_NB) != 0)
         {
             throw std::runtime_error(std::string("Failed to lock journal device: ") + strerror(errno));
         }
@@ -418,25 +485,32 @@ void blockstore_disk_t::open_journal()
 
 void blockstore_disk_t::close_all()
 {
-    if (data_fd >= 0)
-        close(data_fd);
-    if (meta_fd >= 0 && meta_fd != data_fd)
-        close(meta_fd);
-    if (journal_fd >= 0 && journal_fd != meta_fd)
-        close(journal_fd);
+    if (!mock_mode)
+    {
+        if (data_fd >= 0)
+            close(data_fd);
+        if (meta_fd >= 0 && meta_fd != data_fd)
+            close(meta_fd);
+        if (journal_fd >= 0 && journal_fd != meta_fd)
+            close(journal_fd);
+    }
     data_fd = meta_fd = journal_fd = -1;
 }
 
 // Sadly DISCARD only works through ioctl(), but it seems to always block the device queue,
 // so it's not a big deal that we can only run it synchronously.
-int blockstore_disk_t::trim_data(allocator_t *alloc)
+int blockstore_disk_t::trim_data(std::function<bool(uint64_t)> is_free)
 {
+    if (mock_mode)
+    {
+        return -EINVAL;
+    }
     int r = 0;
     uint64_t j = 0, i = 0;
     uint64_t discarded = 0;
     for (; i <= block_count; i++)
     {
-        if (i >= block_count || alloc->get(i))
+        if (i >= block_count || is_free(i))
         {
             if (i > j && (i-j)*data_block_size >= min_discard_size)
             {

src/blockstore/blockstore_disk.h

@@ -12,6 +12,10 @@
 // Lower byte of checksum type is its length
 #define BLOCKSTORE_CSUM_CRC32C 0x104
 
+#define MOCK_DATA_FD 1000
+#define MOCK_META_FD 1001
+#define MOCK_JOURNAL_FD 1002
+
 class allocator_t;
 
 struct blockstore_disk_t
@@ -22,11 +26,15 @@ struct blockstore_disk_t
     // Required write alignment and journal/metadata/data areas' location alignment
     uint32_t disk_alignment = 4096;
     // Journal block size - minimum_io_size of the journal device is the best choice
-    uint64_t journal_block_size = 4096;
+    uint32_t journal_block_size = 4096;
     // Metadata block size - minimum_io_size of the metadata device is the best choice
-    uint64_t meta_block_size = 4096;
+    uint32_t meta_block_size = 4096;
+    // Atomic write size of the data block device
+    uint32_t atomic_write_size = 4096;
+    // Target free space in metadata blocks
+    uint32_t meta_block_target_free_space = 800;
     // Sparse write tracking granularity. 4 KB is a good choice. Must be a multiple of disk_alignment
-    uint64_t bitmap_granularity = 4096;
+    uint32_t bitmap_granularity = 4096;
     // Data checksum type, BLOCKSTORE_CSUM_NONE or BLOCKSTORE_CSUM_CRC32C
     uint32_t data_csum_type = BLOCKSTORE_CSUM_NONE;
     // Checksum block size, must be a multiple of bitmap_granularity
@@ -36,27 +44,36 @@ struct blockstore_disk_t
     // I/O modes for data, metadata and journal: direct or "" = O_DIRECT, cached = O_SYNC, directsync = O_DIRECT|O_SYNC
     // O_SYNC without O_DIRECT = use Linux page cache for reads and writes
     std::string data_io, meta_io, journal_io;
+    // It is safe to disable fsync() if drive write cache is writethrough
+    bool disable_data_fsync = false, disable_meta_fsync = false, disable_journal_fsync = false;
+    // Keep journal (buffered data) in memory?
+    bool inmemory_meta = true;
+    // Keep metadata in memory?
+    bool inmemory_journal = true;
     // Data discard granularity and minimum size (for the sake of performance)
     bool discard_on_start = false;
     uint64_t min_discard_size = 1024*1024;
     uint64_t discard_granularity = 0;
 
     int meta_fd = -1, data_fd = -1, journal_fd = -1;
-    uint64_t meta_offset, meta_device_sect, meta_device_size, meta_len, meta_format = 0;
+    uint64_t meta_offset, meta_device_sect, meta_device_size, meta_area_size, min_meta_len, meta_format = 0;
     uint64_t data_offset, data_device_sect, data_device_size, data_len;
     uint64_t journal_offset, journal_device_sect, journal_device_size, journal_len;
 
-    uint32_t block_order = 0;
     uint64_t block_count = 0;
-    uint32_t clean_entry_bitmap_size = 0, clean_entry_size = 0, clean_dyn_size = 0;
+    uint32_t clean_entry_bitmap_size = 0;
+    uint32_t clean_entry_size = 0, clean_dyn_size = 0; // for meta_v1/2
+
+    bool mock_mode = false;
 
     void parse_config(std::map<std::string, std::string> & config);
     void open_data();
     void open_meta();
     void open_journal();
-    void calc_lengths(bool skip_meta_check = false);
+    void calc_lengths();
+    void check_lengths();
     void close_all();
-    int trim_data(allocator_t *alloc);
+    int trim_data(std::function<bool(uint64_t)> is_free);
 
     inline uint64_t dirty_dyn_size(uint64_t offset, uint64_t len)
     {

src/blockstore/blockstore_flush.h

@@ -1,22 +1,20 @@
 // Copyright (c) Vitaliy Filippov, 2019+
 // License: VNPL-1.1 (see README.md for details)
 
-#define COPY_BUF_JOURNAL 1
-#define COPY_BUF_DATA 2
-#define COPY_BUF_ZERO 4
-#define COPY_BUF_CSUM_FILL 8
-#define COPY_BUF_COALESCED 16
-#define COPY_BUF_META_BLOCK 32
-#define COPY_BUF_JOURNALED_BIG 64
+#define COPY_BUF_JOURNAL    0x01
+#define COPY_BUF_DATA       0x02
+#define COPY_BUF_ZERO       0x04
+#define COPY_BUF_CSUM_FILL  0x08
+#define COPY_BUF_COALESCED  0x10
+#define COPY_BUF_PADDED     0x20
+#define COPY_BUF_SKIP_CSUM  0x40
 
 struct copy_buffer_t
 {
-    int copy_flags;
-    uint64_t offset, len, disk_offset;
-    uint64_t journal_sector; // only for reads: sector+1 if used and !journal.inmemory, otherwise 0
-    void *buf;
-    uint8_t *csum_buf;
-    int *dyn_data;
+    uint32_t copy_flags;
+    uint64_t offset, len, disk_loc, disk_offset, disk_len;
+    uint8_t *buf;
+    uint64_t wr_lsn;
 };
 
 struct meta_sector_t
@@ -27,13 +25,6 @@ struct meta_sector_t
     int usage_count;
 };
 
-struct flusher_sync_t
-{
-    bool fsync_meta;
-    int ready_count;
-    int state;
-};
-
 struct flusher_meta_write_t
 {
     uint64_t sector, pos;
@@ -49,93 +40,75 @@ class journal_flusher_co
 {
     blockstore_impl_t *bs;
     journal_flusher_t *flusher;
-    int wait_state, wait_count, wait_journal_count;
+    int co_id;
+    int wait_state, wait_count;
     struct io_uring_sqe *sqe;
     struct ring_data_t *data;
 
-    std::list<flusher_sync_t>::iterator cur_sync;
+    std::function<void(ring_data_t*)> simple_callback_r, simple_callback_w;
 
-    obj_ver_id cur;
-    std::map<obj_ver_id, dirty_entry>::iterator dirty_it, dirty_start, dirty_end;
-    std::map<object_id, uint64_t>::iterator repeat_it;
-    std::function<void(ring_data_t*)> simple_callback_r, simple_callback_rj, simple_callback_w;
+    object_id cur_oid;
+    uint64_t copy_id;
+    uint64_t compact_lsn;
+    uint64_t cur_version;
+    heap_object_t *cur_obj;
+    heap_write_t *begin_wr, *end_wr;
+    uint32_t modified_block;
+    bool should_repeat;
 
-    bool try_trim = false;
-    bool skip_copy, has_delete, has_writes;
-    std::vector<copy_buffer_t> v;
-    std::vector<copy_buffer_t>::iterator it;
-    int i;
-    bool fill_incomplete, cleared_incomplete;
-    int read_to_fill_incomplete;
+    std::vector<copy_buffer_t> read_vec;
+    uint32_t overwrite_start, overwrite_end;
+    uint32_t big_start, big_end;
+    int i, res;
+    bool read_to_fill_incomplete;
     int copy_count;
-    uint64_t clean_loc, clean_ver, old_clean_loc, old_clean_ver;
+    uint64_t clean_loc;
     flusher_meta_write_t meta_old, meta_new;
-    bool clean_init_bitmap;
-    uint64_t clean_bitmap_offset, clean_bitmap_len;
-    uint8_t *clean_init_dyn_ptr;
-    uint8_t *new_clean_bitmap;
-
-    uint64_t new_trim_pos;
+    bool do_repeat = false;
 
     friend class journal_flusher_t;
-    void scan_dirty();
-    bool read_dirty(int wait_base);
-    bool modify_meta_do_reads(int wait_base);
-    bool wait_meta_reads(int wait_base);
-    bool modify_meta_read(uint64_t meta_loc, flusher_meta_write_t &wr, int wait_base);
-    bool clear_incomplete_csum_block_bits(int wait_base);
-    void calc_block_checksums(uint32_t *new_data_csums, bool skip_overwrites);
-    void update_metadata_entry();
-    bool write_meta_block(flusher_meta_write_t & meta_block, int wait_base);
-    void update_clean_db();
-    void free_data_blocks();
-    bool fsync_batch(bool fsync_meta, int wait_base);
-    bool trim_journal(int wait_base);
+
+    void iterate_checksum_holes(std::function<void(int & pos, uint32_t hole_start, uint32_t hole_end)> cb);
+    void fill_partial_checksum_blocks();
     void free_buffers();
+    int check_and_punch_checksums();
+    bool calc_block_checksums();
+    bool write_meta_block(int wait_base);
+    bool read_buffered(int wait_base);
+    bool fsync_meta(int wait_base);
+    int fsync_buffer(int wait_base);
+    bool trim_lsn(int wait_base);
 public:
     journal_flusher_co();
+    ~journal_flusher_co();
     bool loop();
 };
 
 // Journal flusher itself
 class journal_flusher_t
 {
-    int trim_wanted = 0;
-    bool dequeuing;
-    int min_flusher_count, max_flusher_count, cur_flusher_count, target_flusher_count;
-    int flusher_start_threshold;
+    int force_start = 0;
+    int min_flusher_count = 0, max_flusher_count = 0, cur_flusher_count = 0, target_flusher_count = 0;
     journal_flusher_co *co;
     blockstore_impl_t *bs;
     friend class journal_flusher_co;
 
-    int journal_trim_counter;
-    bool trimming;
-    void* journal_superblock;
+    int advance_lsn_counter = 0;
+    uint64_t compact_counter = 0;
 
-    int active_flushers;
-    int syncing_flushers;
-    std::list<flusher_sync_t> syncs;
-    std::map<object_id, uint64_t> sync_to_repeat;
-
-    std::map<uint64_t, meta_sector_t> meta_sectors;
-    std::deque<object_id> flush_queue;
-    std::map<object_id, uint64_t> flush_versions; // FIXME: consider unordered_map?
-
-    bool try_find_older(std::map<obj_ver_id, dirty_entry>::iterator & dirty_end, obj_ver_id & cur);
-    bool try_find_other(std::map<obj_ver_id, dirty_entry>::iterator & dirty_end, obj_ver_id & cur);
+    int active_flushers = 0;
+    int wanting_meta_fsync = 0;
+    bool fsyncing_meta = false;
+    int syncing_buffer = 0;
 
 public:
     journal_flusher_t(blockstore_impl_t *bs);
     ~journal_flusher_t();
     void loop();
-    bool is_trim_wanted() { return trim_wanted; }
+    int get_syncing_buffer();
+    uint64_t get_compact_counter();
     bool is_active();
-    void mark_trim_possible();
     void request_trim();
     void release_trim();
-    void enqueue_flush(obj_ver_id oid);
-    void unshift_flush(obj_ver_id oid, bool force);
-    void remove_flush(object_id oid);
     void dump_diagnostics();
-    bool is_mutated(uint64_t clean_loc);
 };

src/blockstore/blockstore_heap.h

@@ -0,0 +1,344 @@
+// Metadata storage version 3 ("heap")
+// Copyright (c) Vitaliy Filippov, 2025+
+// License: VNPL-1.1 (see README.md for details)
+
+#pragma once
+
+#define EMH_EXT
+
+#include <map>
+#include <unordered_map>
+#include <set>
+#include <deque>
+#include <vector>
+
+#include "../client/object_id.h"
+#include "../../emhash/hash_table7.hpp"
+#include "../util/wyhash.h"
+#include "blockstore_disk.h"
+#include "multilist.h"
+
+struct pool_shard_settings_t
+{
+    uint32_t pg_count;
+    uint32_t pg_stripe_size;
+};
+
+#define BS_HEAP_TYPE 7
+#define BS_HEAP_SMALL_WRITE 1
+#define BS_HEAP_BIG_WRITE 2
+#define BS_HEAP_TOMBSTONE 3
+#define BS_HEAP_INTENT_WRITE 4
+#define BS_HEAP_STABLE 8
+
+class blockstore_heap_t;
+
+struct __attribute__((__packed__)) heap_write_t
+{
+    // size should have top bit cleared
+    uint16_t size = 0;
+    int16_t next_pos = 0;
+    uint64_t lsn = 0;
+    uint64_t version = 0;
+    uint32_t offset = 0;
+    uint32_t len = 0;
+    uint64_t location = 0;
+    uint8_t flags = 0; // 1|2|3 = small|big|tombstone, 4|0 = stable|unstable
+
+    // uint8_t[] external_bitmap
+    // uint8_t[] internal_bitmap
+    // uint32_t[] checksums
+
+    heap_write_t *next();
+    inline uint8_t type() const { return (flags & BS_HEAP_TYPE); }
+    uint32_t get_size(blockstore_heap_t *heap);
+    uint32_t get_csum_size(blockstore_heap_t *heap);
+    bool needs_recheck(blockstore_heap_t *heap);
+    bool needs_compact(blockstore_heap_t *heap);
+    bool is_compacted(uint64_t compacted_lsn);
+    bool can_be_collapsed(blockstore_heap_t *heap);
+    bool is_allowed_before_compacted(uint64_t compacted_lsn, bool is_last_entry);
+    uint8_t *get_ext_bitmap(blockstore_heap_t *heap);
+    uint8_t *get_int_bitmap(blockstore_heap_t *heap);
+    uint8_t *get_checksums(blockstore_heap_t *heap);
+    uint32_t *get_checksum(blockstore_heap_t *heap);
+};
+
+struct __attribute__((__packed__)) heap_object_t
+{
+    // size should have top bit cleared
+    uint16_t size = 0;
+    // linked list of write entries...
+    // newest entries are stored first to simplify scanning
+    int16_t write_pos = 0;
+    uint32_t crc32c = 0;
+    uint64_t inode = 0;
+    uint64_t stripe = 0;
+
+    heap_write_t *get_writes();
+    uint32_t calc_crc32c();
+};
+
+struct heap_object_lsn_t
+{
+    object_id oid;
+    uint64_t lsn;
+};
+
+inline bool operator < (const heap_object_lsn_t & a, const heap_object_lsn_t & b)
+{
+    return a.oid < b.oid || a.oid == b.oid && a.lsn < b.lsn;
+}
+
+struct tmp_compact_item_t
+{
+    object_id oid;
+    uint64_t lsn;
+    bool compact;
+};
+
+struct heap_mvcc_copy_id_t
+{
+    object_id oid;
+    uint64_t copy_id;
+};
+
+inline bool operator == (const heap_mvcc_copy_id_t & a, const heap_mvcc_copy_id_t & b)
+{
+    return a.oid.inode == b.oid.inode && a.oid.stripe == b.oid.stripe && a.copy_id == b.copy_id;
+}
+
+namespace std
+{
+    template<> struct hash<heap_mvcc_copy_id_t>
+    {
+        inline size_t operator()(const heap_mvcc_copy_id_t &s) const
+        {
+            size_t seed = std::hash<object_id>()(s.oid);
+            // Copy-pasted from spp::hash_combine()
+            seed ^= (s.copy_id + 0xc6a4a7935bd1e995 + (seed << 6) + (seed >> 2));
+            return seed;
+        }
+    };
+};
+
+struct heap_object_mvcc_t
+{
+    uint32_t readers = 0;
+    heap_object_t *entry_copy = NULL;
+};
+
+struct __attribute__((__packed__)) heap_block_info_t
+{
+    uint32_t used_space = 0;
+    uint32_t free_pos = 0;
+    uint8_t *data = NULL;
+};
+
+struct heap_inflight_lsn_t
+{
+    object_id oid;
+    uint64_t flags;
+};
+
+struct heap_refqi_t
+{
+    uint64_t lsn;
+    uint64_t inode;
+    uint64_t location;
+    uint32_t len;
+    bool is_data;
+};
+
+using i64hash_t = wyhash::hash<uint64_t>;
+using heap_block_index_t = emhash7::HashMap<uint64_t, emhash7::HashMap<inode_t, emhash7::HashMap<uint64_t, uint64_t, i64hash_t>, i64hash_t>, i64hash_t>;
+using heap_mvcc_map_t = emhash7::HashMap<heap_mvcc_copy_id_t, heap_object_mvcc_t>;
+
+class blockstore_heap_t
+{
+    friend class heap_write_t;
+    friend class heap_object_t;
+
+    blockstore_disk_t *dsk = NULL;
+    uint8_t* buffer_area = NULL;
+    bool abort_on_corruption = false;
+    bool abort_on_overlap = true;
+    int log_level = 0;
+
+    const uint32_t meta_block_count = 0;
+    uint32_t target_block_free_space = 800;
+
+    uint64_t next_lsn = 0;
+    emhash7::HashMap<pool_id_t, pool_shard_settings_t> pool_shard_settings;
+    // PG => inode => stripe => block number
+    heap_block_index_t block_index;
+    std::vector<heap_block_info_t> block_info;
+    allocator_t *data_alloc = NULL;
+    multilist_index_t *meta_alloc = NULL;
+    uint32_t meta_alloc_count = 0;
+    uint64_t meta_used_space = 0;
+    multilist_alloc_t *buffer_alloc = NULL;
+    heap_mvcc_map_t object_mvcc;
+    std::unordered_map<uint64_t, uint32_t> mvcc_data_refs;
+    std::unordered_map<uint64_t, uint32_t> mvcc_buffer_refs;
+    std::map<uint64_t, uint64_t> inode_space_stats;
+    uint64_t buffer_area_used_space = 0;
+    uint64_t data_used_space = 0;
+
+    // LSN queue: inflight (writing) -> completed [-> fsynced] -> compactable -> compacted [-> fsynced] -> trimmed and removed
+    std::deque<heap_inflight_lsn_t> inflight_lsn;
+    uint32_t to_compact_count = 0;
+    uint64_t first_inflight_lsn = 0;
+    uint64_t completed_lsn = 0;
+    uint64_t fsynced_lsn = 0;
+    uint64_t compacted_lsn = 0;
+    uint64_t next_compact_lsn = 0;
+    std::deque<heap_refqi_t> overwrite_ref_queue;
+
+    std::vector<tmp_compact_item_t> tmp_compact_queue;
+    std::deque<object_id> recheck_queue;
+    int recheck_in_progress = 0;
+    bool in_recheck = false;
+    std::function<void(bool is_data, uint64_t offset, uint64_t len, uint8_t* buf, std::function<void()>)> recheck_cb;
+    int recheck_queue_depth = 0;
+
+    const uint32_t max_write_entry_size;
+
+    uint64_t get_pg_id(inode_t inode, uint64_t stripe);
+    void defragment_block(uint32_t block_num);
+    uint32_t find_block_run(heap_block_info_t & block, uint32_t space);
+    uint32_t find_block_space(uint32_t block_num, uint32_t space);
+    uint32_t block_has_compactable(uint8_t *data);
+    uint32_t compact_object_to(heap_object_t *obj, uint64_t lsn, uint8_t *new_csums, bool do_free);
+    void copy_full_object(uint8_t *dst, heap_object_t *obj);
+    bool mvcc_save_copy(heap_object_t *obj);
+    bool mvcc_check_tracking(object_id oid);
+    void free_mvcc(heap_mvcc_map_t::iterator mvcc_it);
+    void allocate_block(heap_block_info_t & inf);
+    int allocate_new_object(object_id oid, uint32_t full_object_size, uint32_t *modified_block, heap_object_t **new_obj);
+    int add_object(object_id oid, heap_write_t *wr, uint32_t *modified_block);
+    void mark_overwritten(uint64_t over_lsn, uint64_t inode, heap_write_t *wr, heap_write_t *end_wr, bool tracking_active);
+    int update_object(uint32_t block_num, heap_object_t *obj, heap_write_t *wr, uint32_t *modified_block, uint32_t *moved_from_block);
+    void init_erase(uint32_t block_num, heap_object_t *obj);
+    void erase_object(uint32_t block_num, heap_object_t *obj, uint64_t lsn, bool tracking_active);
+    void reindex_block(uint32_t block_num, heap_object_t *from_obj);
+    void erase_block_index(inode_t inode, uint64_t stripe);
+    void deref_data(uint64_t inode, uint64_t location, bool free_at_0);
+    void deref_buffer(uint64_t inode, uint64_t location, uint32_t len, bool free_at_0);
+    void deref_overwrites(uint64_t lsn);
+    void free_object_space(inode_t inode, heap_write_t *from, heap_write_t *to, int mode = 0);
+    void add_used_space(uint32_t block_num, int32_t used_delta);
+    void push_inflight_lsn(object_id oid, uint64_t lsn, uint64_t flags);
+
+public:
+    blockstore_heap_t(blockstore_disk_t *dsk, uint8_t *buffer_area, int log_level = 0);
+    ~blockstore_heap_t();
+    // set initially compacted lsn - should be done before loading
+    void set_compacted_lsn(uint64_t compacted_lsn);
+    uint64_t get_compacted_lsn();
+    // load data from the disk, returns count of loaded write entries
+    void read_blocks(uint64_t disk_offset, uint64_t size, uint8_t *buf,
+        std::function<void(heap_object_t*)> handle_object, std::function<void(uint32_t, uint32_t, uint8_t*)> handle_block);
+    uint64_t load_blocks(uint64_t disk_offset, uint64_t size, uint8_t *buf);
+    // finish loading
+    void finish_load();
+    // recheck small write data after reading the database from disk
+    bool recheck_small_writes(std::function<void(bool is_data, uint64_t offset, uint64_t len, uint8_t* buf, std::function<void()>)> read_buffer, int queue_depth);
+    // initialize metadata area (fill it with empty data)
+    // returns 0 when done, EAGAIN when the caller has to wait more
+    int initialize();
+    // read from the metadata area
+    // returns 0 when done, EAGAIN when the caller has to wait more
+    int read();
+    // reshard database according to the pool's PG count
+    void reshard(pool_id_t pool, uint32_t pg_count, uint32_t pg_stripe_size);
+    // read an object entry and lock it against removal
+    // in the future, may become asynchronous
+    heap_object_t *lock_and_read_entry(object_id oid, uint64_t & copy_id);
+    // re-read a locked object entry with the given lsn (pointer may be invalidated)
+    heap_object_t *read_locked_entry(object_id oid, uint64_t copy_id);
+    // read an object entry without locking it
+    heap_object_t *read_entry(object_id oid, uint32_t *block_num_ptr, bool for_update = false);
+    // unlock an entry
+    bool unlock_entry(object_id oid, uint64_t copy_id);
+    // set or verify checksums in a write request
+    bool calc_checksums(heap_write_t *wr, uint8_t *data, bool set);
+    // set or verify raw block checksums
+    bool calc_block_checksums(uint32_t *block_csums, uint8_t *data, uint8_t *bitmap, uint32_t start, uint32_t end,
+        bool set, std::function<void(uint32_t, uint32_t, uint32_t)> bad_block_cb);
+    bool calc_block_checksums(uint32_t *block_csums, uint8_t *bitmap,
+        uint32_t start, uint32_t end, std::function<uint8_t*(uint32_t start, uint32_t & len)> next,
+        bool set, std::function<void(uint32_t, uint32_t, uint32_t)> bad_block_cb);
+    // copy an object as is
+    int copy_object(heap_object_t *obj, uint32_t *modified_block);
+    // auto-compacts the object, then adds a write entry to it and to the compaction queue
+    // return 0 if OK, or maybe ENOSPC
+    int post_write(object_id oid, heap_write_t *wr, uint32_t *modified_block, uint32_t *moved_from_block);
+    int post_write(uint32_t & block_num, object_id oid, heap_object_t *obj, heap_write_t *wr, uint32_t *moved_from_block);
+    // stabilize an unstable object version
+    // return 0 if OK, ENOENT if not exists
+    int post_stabilize(object_id oid, uint64_t version, uint32_t *modified_block, uint64_t *new_lsn, uint64_t *new_to_lsn);
+    // rollback an unstable object version
+    // return 0 if OK, ENOENT if not exists, EBUSY if already stable
+    int post_rollback(object_id oid, uint64_t version, uint64_t *new_lsn, uint32_t *modified_block);
+    // forget an object
+    // return error code
+    int post_delete(object_id oid, uint64_t *new_lsn, uint32_t *modified_block);
+    int post_delete(uint32_t block_num, heap_object_t *obj, uint64_t *new_lsn);
+    // get the next object to compact
+    // guaranteed to return objects in min lsn order
+    // returns 0 if OK, ENOENT if nothing to compact
+    int get_next_compact(object_id & oid);
+    // get the range of an object eligible for compaction
+    void get_compact_range(heap_object_t *obj, uint64_t max_lsn, heap_write_t **begin_wr, heap_write_t **end_wr);
+    // mark an object as compacted up to the given lsn
+    int compact_object(object_id oid, uint64_t lsn, uint8_t *new_csums);
+    // retrieve object listing from a PG
+    int list_objects(uint32_t pg_num, object_id min_oid, object_id max_oid,
+        obj_ver_id **result_list, size_t *stable_count, size_t *unstable_count);
+    // set a block number for a new object and returns error status: 0, EAGAIN or ENOSPC
+    int get_block_for_new_object(uint32_t & out_block_num, uint32_t size = 0);
+
+    // inflight write tracking
+    void mark_lsn_completed(uint64_t lsn);
+    void mark_lsn_fsynced(uint64_t lsn);
+    void mark_lsn_compacted(uint64_t lsn, bool allow_undone = false);
+    void mark_object_compacted(heap_object_t *obj, uint64_t max_lsn);
+    void mark_lsn_trimmed(uint64_t lsn);
+    uint64_t get_completed_lsn();
+    uint64_t get_fsynced_lsn();
+
+    // data device block allocator functions
+    uint64_t find_free_data();
+    bool is_data_used(uint64_t location);
+    void use_data(inode_t inode, uint64_t location);
+    void free_data(inode_t inode, uint64_t location);
+
+    // buffer device allocator functions
+    uint64_t find_free_buffer_area(uint64_t size);
+    bool is_buffer_area_free(uint64_t location, uint64_t size);
+    void use_buffer_area(inode_t inode, uint64_t location, uint64_t size);
+    void free_buffer_area(inode_t inode, uint64_t location, uint64_t size);
+    uint64_t get_buffer_area_used_space();
+
+    // get metadata block data buffer and used space
+    uint8_t *get_meta_block(uint32_t block_num);
+    uint32_t get_meta_block_used_space(uint32_t block_num);
+
+    // get space usage statistics
+    uint64_t get_data_used_space();
+    const std::map<uint64_t, uint64_t> & get_inode_space_stats();
+    uint64_t get_meta_total_space();
+    uint64_t get_meta_used_space();
+    uint32_t get_meta_nearfull_blocks();
+    uint32_t get_inflight_queue_size();
+    uint32_t get_compact_queue_size();
+    uint32_t get_to_compact_count();
+
+    // get maximum size for a temporary heap_write_t buffer
+    uint32_t get_max_write_entry_size();
+
+    // only for tests
+    void set_abort_on_corruption(bool fail);
+    void set_abort_on_overlap(bool fail);
+};

src/blockstore/blockstore_impl.cpp

@@ -1,13 +1,17 @@
 // Copyright (c) Vitaliy Filippov, 2019+
 // License: VNPL-1.1 (see README.md for details)
 
-#include "blockstore_impl.h"
+#include <stdexcept>
 
-blockstore_impl_t::blockstore_impl_t(blockstore_config_t & config, ring_loop_t *ringloop, timerfd_manager_t *tfd)
+#include "blockstore_impl.h"
+#include "crc32c.h"
+
+blockstore_impl_t::blockstore_impl_t(blockstore_config_t & config, ring_loop_i *ringloop, timerfd_manager_t *tfd, bool mock_mode)
 {
     assert(sizeof(blockstore_op_private_t) <= BS_OP_PRIVATE_DATA_SIZE);
     this->tfd = tfd;
     this->ringloop = ringloop;
+    dsk.mock_mode = mock_mode;
     ring_consumer.loop = [this]() { loop(); };
     ringloop->register_consumer(&ring_consumer);
     initialized = 0;
@@ -17,31 +21,43 @@ blockstore_impl_t::blockstore_impl_t(blockstore_config_t & config, ring_loop_t *
         dsk.open_data();
         dsk.open_meta();
         dsk.open_journal();
-        calc_lengths();
-        alloc_dyn_data = dsk.clean_dyn_size > sizeof(void*) || dsk.csum_block_size > 0;
+        dsk.calc_lengths();
         zero_object = (uint8_t*)memalign_or_die(MEM_ALIGNMENT, dsk.data_block_size);
-        data_alloc = new allocator_t(dsk.block_count);
     }
     catch (std::exception & e)
     {
         dsk.close_all();
         throw;
     }
+    memset(zero_object, 0, dsk.data_block_size);
+    meta_superblock = (uint8_t*)memalign_or_die(MEM_ALIGNMENT, dsk.meta_block_size);
+    memset(meta_superblock, 0, dsk.meta_block_size);
+}
+
+void blockstore_impl_t::init()
+{
     flusher = new journal_flusher_t(this);
+    if (dsk.inmemory_journal)
+    {
+        buffer_area = (uint8_t*)memalign_or_die(MEM_ALIGNMENT, dsk.journal_len);
+    }
+    heap = new blockstore_heap_t(&dsk, buffer_area, log_level);
 }
 
 blockstore_impl_t::~blockstore_impl_t()
 {
-    delete data_alloc;
-    delete flusher;
+    if (flusher)
+        delete flusher;
+    if (heap)
+        delete heap;
+    if (buffer_area)
+        free(buffer_area);
+    if (meta_superblock)
+        free(meta_superblock);
     if (zero_object)
         free(zero_object);
     ringloop->unregister_consumer(&ring_consumer);
     dsk.close_all();
-    if (metadata_buffer)
-        free(metadata_buffer);
-    if (clean_bitmaps)
-        free(clean_bitmaps);
 }
 
 bool blockstore_impl_t::is_started()
@@ -57,10 +73,9 @@ bool blockstore_impl_t::is_stalled()
 // main event loop - produce requests
 void blockstore_impl_t::loop()
 {
-    // FIXME: initialized == 10 is ugly
     if (initialized != 10)
     {
-        // read metadata, then journal
+        // read metadata
         if (initialized == 0)
         {
             metadata_init_reader = new blockstore_init_meta(this);
@@ -73,69 +88,41 @@ void blockstore_impl_t::loop()
             {
                 delete metadata_init_reader;
                 metadata_init_reader = NULL;
-                journal_init_reader = new blockstore_init_journal(this);
-                initialized = 2;
-            }
-        }
-        if (initialized == 2)
-        {
-            int res = journal_init_reader->loop();
-            if (!res)
-            {
-                delete journal_init_reader;
-                journal_init_reader = NULL;
                 initialized = 3;
-                ringloop->wakeup();
             }
         }
         if (initialized == 3)
         {
             if (!readonly && dsk.discard_on_start)
-                dsk.trim_data(data_alloc);
-            if (journal.flush_journal)
-                initialized = 4;
-            else
-                initialized = 10;
-        }
-        if (initialized == 4)
-        {
-            if (readonly)
             {
-                printf("Can't flush the journal in readonly mode\n");
-                exit(1);
+                dsk.trim_data([this](uint64_t block_num){ return heap->is_data_used(block_num * dsk.data_block_size); });
             }
-            flusher->loop();
-            ringloop->submit();
+            initialized = 10;
         }
     }
     else
     {
         // try to submit ops
         unsigned initial_ring_space = ringloop->space_left();
-        // has_writes == 0 - no writes before the current queue item
-        // has_writes == 1 - some writes in progress
-        // has_writes == 2 - tried to submit some writes, but failed
-        int has_writes = 0, op_idx = 0, new_idx = 0;
+        int op_idx = 0, new_idx = 0;
+        bool has_unfinished_writes = false;
         for (; op_idx < submit_queue.size(); op_idx++, new_idx++)
         {
             auto op = submit_queue[op_idx];
             submit_queue[new_idx] = op;
-            // FIXME: This needs some simplification
-            // Writes should not block reads if the ring is not full and reads don't depend on them
-            // In all other cases we should stop submission
             if (PRIV(op)->wait_for)
             {
                 check_wait(op);
                 if (PRIV(op)->wait_for == WAIT_SQE)
                 {
+                    // ring is full, stop submission
                     break;
                 }
                 else if (PRIV(op)->wait_for)
                 {
-                    if (op->opcode == BS_OP_WRITE || op->opcode == BS_OP_WRITE_STABLE || op->opcode == BS_OP_DELETE)
-                    {
-                        has_writes = 2;
-                    }
+                    has_unfinished_writes = has_unfinished_writes || op->opcode == BS_OP_WRITE ||
+                        op->opcode == BS_OP_WRITE_STABLE || op->opcode == BS_OP_DELETE ||
+                        op->opcode == BS_OP_STABLE || op->opcode == BS_OP_ROLLBACK;
                     continue;
                 }
             }
@@ -148,46 +135,33 @@ void blockstore_impl_t::loop()
             {
                 wr_st = dequeue_read(op);
             }
-            else if (op->opcode == BS_OP_WRITE || op->opcode == BS_OP_WRITE_STABLE)
+            else if (op->opcode == BS_OP_WRITE || op->opcode == BS_OP_WRITE_STABLE || op->opcode == BS_OP_DELETE)
             {
-                if (has_writes == 2)
-                {
-                    // Some writes already could not be submitted
-                    continue;
-                }
                 wr_st = dequeue_write(op);
-                has_writes = wr_st > 0 ? 1 : 2;
-            }
-            else if (op->opcode == BS_OP_DELETE)
-            {
-                if (has_writes == 2)
-                {
-                    // Some writes already could not be submitted
-                    continue;
-                }
-                wr_st = dequeue_del(op);
-                has_writes = wr_st > 0 ? 1 : 2;
+                has_unfinished_writes = has_unfinished_writes || (wr_st != 2);
             }
             else if (op->opcode == BS_OP_SYNC)
             {
-                // sync only completed writes?
-                // wait for the data device fsync to complete, then submit journal writes for big writes
-                // then submit an fsync operation
+                // syncs only completed writes, so doesn't have to be blocked by anything
                 wr_st = continue_sync(op);
             }
-            else if (op->opcode == BS_OP_STABLE)
+            else if (op->opcode == BS_OP_STABLE || op->opcode == BS_OP_ROLLBACK)
             {
                 wr_st = dequeue_stable(op);
-            }
-            else if (op->opcode == BS_OP_ROLLBACK)
-            {
-                wr_st = dequeue_rollback(op);
+                has_unfinished_writes = has_unfinished_writes || (wr_st != 2);
             }
             else if (op->opcode == BS_OP_LIST)
             {
-                // LIST doesn't have to be blocked by previous modifications
-                process_list(op);
-                wr_st = 2;
+                // LIST has to be blocked by previous writes and commits/rollbacks
+                if (!has_unfinished_writes)
+                {
+                    process_list(op);
+                    wr_st = 2;
+                }
+                else
+                {
+                    wr_st = 0;
+                }
             }
             if (wr_st == 2)
             {
@@ -196,16 +170,13 @@ void blockstore_impl_t::loop()
             }
             if (wr_st == 0)
             {
+                PRIV(op)->pending_ops = 0;
                 ringloop->restore(prev_sqe_pos);
                 if (PRIV(op)->wait_for == WAIT_SQE)
                 {
                     // ring is full, stop submission
                     break;
                 }
-                else if (PRIV(op)->wait_for == WAIT_JOURNAL)
-                {
-                    PRIV(op)->wait_detail2 = (unstable_writes.size()+unstable_unsynced);
-                }
             }
         }
         if (op_idx != new_idx)
@@ -225,12 +196,6 @@ void blockstore_impl_t::loop()
         {
             throw std::runtime_error(std::string("io_uring_submit: ") + strerror(-ret));
         }
-        for (auto s: journal.submitting_sectors)
-        {
-            // Mark journal sector writes as submitted
-            journal.sector_info[s].submit_id = 0;
-        }
-        journal.submitting_sectors.clear();
         if ((initial_ring_space - ringloop->space_left()) > 0)
         {
             live = true;
@@ -248,7 +213,7 @@ bool blockstore_impl_t::is_safe_to_stop()
     {
         return false;
     }
-    if (unsynced_big_writes.size() > 0 || unsynced_small_writes.size() > 0)
+    if (unsynced_big_write_count > 0 || unsynced_small_write_count > 0)
     {
         if (!readonly && !stop_sync_submitted)
         {
@@ -272,7 +237,7 @@ void blockstore_impl_t::check_wait(blockstore_op_t *op)
 {
     if (PRIV(op)->wait_for == WAIT_SQE)
     {
-        if (ringloop->sqes_left() < PRIV(op)->wait_detail)
+        if (ringloop->space_left() < PRIV(op)->wait_detail)
         {
             // stop submission if there's still no free space
 #ifdef BLOCKSTORE_DEBUG
@@ -282,40 +247,13 @@ void blockstore_impl_t::check_wait(blockstore_op_t *op)
         }
         PRIV(op)->wait_for = 0;
     }
-    else if (PRIV(op)->wait_for == WAIT_JOURNAL)
+    else if (PRIV(op)->wait_for == WAIT_COMPACTION)
     {
-        if (journal.used_start == PRIV(op)->wait_detail &&
-            (unstable_writes.size()+unstable_unsynced) == PRIV(op)->wait_detail2)
+        if (flusher->get_compact_counter() <= PRIV(op)->wait_detail)
         {
             // do not submit
 #ifdef BLOCKSTORE_DEBUG
-            printf("Still waiting to flush journal offset %08jx\n", PRIV(op)->wait_detail);
-#endif
-            return;
-        }
-        flusher->release_trim();
-        PRIV(op)->wait_for = 0;
-    }
-    else if (PRIV(op)->wait_for == WAIT_JOURNAL_BUFFER)
-    {
-        int next = ((journal.cur_sector + 1) % journal.sector_count);
-        if (journal.sector_info[next].flush_count > 0 ||
-            journal.sector_info[next].dirty)
-        {
-            // do not submit
-#ifdef BLOCKSTORE_DEBUG
-            printf("Still waiting for a journal buffer\n");
-#endif
-            return;
-        }
-        PRIV(op)->wait_for = 0;
-    }
-    else if (PRIV(op)->wait_for == WAIT_FREE)
-    {
-        if (!data_alloc->get_free_count() && big_to_flush > 0)
-        {
-#ifdef BLOCKSTORE_DEBUG
-            printf("Still waiting for free space on the data device\n");
+            printf("Still waiting for more flushes\n");
 #endif
             return;
         }
@@ -343,44 +281,6 @@ void blockstore_impl_t::enqueue_op(blockstore_op_t *op)
         ringloop->set_immediate([op]() { std::function<void (blockstore_op_t*)>(op->callback)(op); });
         return;
     }
-    if (op->opcode == BS_OP_SYNC_STAB_ALL)
-    {
-        std::function<void(blockstore_op_t*)> *old_callback = new std::function<void(blockstore_op_t*)>(op->callback);
-        op->opcode = BS_OP_SYNC;
-        op->callback = [this, old_callback](blockstore_op_t *op)
-        {
-            if (op->retval >= 0 && unstable_writes.size() > 0)
-            {
-                op->opcode = BS_OP_STABLE;
-                op->len = unstable_writes.size();
-                obj_ver_id *vers = new obj_ver_id[op->len];
-                op->buf = vers;
-                int i = 0;
-                for (auto it = unstable_writes.begin(); it != unstable_writes.end(); it++, i++)
-                {
-                    vers[i] = {
-                        .oid = it->first,
-                        .version = it->second,
-                    };
-                }
-                unstable_writes.clear();
-                op->callback = [old_callback](blockstore_op_t *op)
-                {
-                    obj_ver_id *vers = (obj_ver_id*)op->buf;
-                    delete[] vers;
-                    op->buf = NULL;
-                    (*old_callback)(op);
-                    delete old_callback;
-                };
-                this->enqueue_op(op);
-            }
-            else
-            {
-                (*old_callback)(op);
-                delete old_callback;
-            }
-        };
-    }
     if ((op->opcode == BS_OP_WRITE || op->opcode == BS_OP_WRITE_STABLE || op->opcode == BS_OP_DELETE) && !enqueue_write(op))
     {
         ringloop->set_immediate([op]() { std::function<void (blockstore_op_t*)>(op->callback)(op); });
@@ -399,75 +299,11 @@ void blockstore_impl_t::init_op(blockstore_op_t *op)
 {
     // Call constructor without allocating memory. We'll call destructor before returning op back
     new ((void*)op->private_data) blockstore_op_private_t;
-    PRIV(op)->min_flushed_journal_sector = PRIV(op)->max_flushed_journal_sector = 0;
     PRIV(op)->wait_for = 0;
     PRIV(op)->op_state = 0;
     PRIV(op)->pending_ops = 0;
 }
 
-static bool replace_stable(object_id oid, uint64_t version, int search_start, int search_end, obj_ver_id* list)
-{
-    while (search_start < search_end)
-    {
-        int pos = search_start+(search_end-search_start)/2;
-        if (oid < list[pos].oid)
-        {
-            search_end = pos;
-        }
-        else if (list[pos].oid < oid)
-        {
-            search_start = pos+1;
-        }
-        else
-        {
-            list[pos].version = version;
-            return true;
-        }
-    }
-    return false;
-}
-
-blockstore_clean_db_t& blockstore_impl_t::clean_db_shard(object_id oid)
-{
-    uint64_t pg_num = 0;
-    uint64_t pool_id = (oid.inode >> (64-POOL_ID_BITS));
-    auto sh_it = clean_db_settings.find(pool_id);
-    if (sh_it != clean_db_settings.end())
-    {
-        // like map_to_pg()
-        pg_num = (oid.stripe / sh_it->second.pg_stripe_size) % sh_it->second.pg_count + 1;
-    }
-    return clean_db_shards[(pool_id << (64-POOL_ID_BITS)) | pg_num];
-}
-
-void blockstore_impl_t::reshard_clean_db(pool_id_t pool, uint32_t pg_count, uint32_t pg_stripe_size)
-{
-    uint64_t pool_id = (uint64_t)pool;
-    std::map<pool_pg_id_t, blockstore_clean_db_t> new_shards;
-    auto sh_it = clean_db_shards.lower_bound((pool_id << (64-POOL_ID_BITS)));
-    while (sh_it != clean_db_shards.end() &&
-        (sh_it->first >> (64-POOL_ID_BITS)) == pool_id)
-    {
-        for (auto & pair: sh_it->second)
-        {
-            // like map_to_pg()
-            uint64_t pg_num = (pair.first.stripe / pg_stripe_size) % pg_count + 1;
-            uint64_t shard_id = (pool_id << (64-POOL_ID_BITS)) | pg_num;
-            new_shards[shard_id][pair.first] = pair.second;
-        }
-        clean_db_shards.erase(sh_it++);
-    }
-    for (sh_it = new_shards.begin(); sh_it != new_shards.end(); sh_it++)
-    {
-        auto & to = clean_db_shards[sh_it->first];
-        to.swap(sh_it->second);
-    }
-    clean_db_settings[pool_id] = (pool_shard_settings_t){
-        .pg_count = pg_count,
-        .pg_stripe_size = pg_stripe_size,
-    };
-}
-
 void blockstore_impl_t::process_list(blockstore_op_t *op)
 {
     uint32_t list_pg = op->pg_number+1;
@@ -476,7 +312,8 @@ void blockstore_impl_t::process_list(blockstore_op_t *op)
     uint64_t min_inode = op->min_oid.inode;
     uint64_t max_inode = op->max_oid.inode;
     // Check PG
-    if (pg_count != 0 && (pg_stripe_size < MIN_DATA_BLOCK_SIZE || list_pg > pg_count))
+    if (!pg_count || (pg_stripe_size < MIN_DATA_BLOCK_SIZE || list_pg > pg_count) ||
+        !INODE_POOL(min_inode) || INODE_POOL(min_inode) != INODE_POOL(max_inode))
     {
         op->retval = -EINVAL;
         FINISH_OP(op);
@@ -484,250 +321,40 @@ void blockstore_impl_t::process_list(blockstore_op_t *op)
     }
     // Check if the DB needs resharding
     // (we don't know about PGs from the beginning, we only create "shards" here)
-    uint64_t first_shard = 0, last_shard = UINT64_MAX;
-    if (min_inode != 0 &&
-        // Check if min_inode == max_inode == pool_id<<N, i.e. this is a pool listing
-        (min_inode >> (64-POOL_ID_BITS)) == (max_inode >> (64-POOL_ID_BITS)))
+    heap->reshard(INODE_POOL(min_inode), pg_count, pg_stripe_size);
+    obj_ver_id *result = NULL;
+    size_t stable_count = 0, unstable_count = 0;
+    int res = heap->list_objects(list_pg, op->min_oid, op->max_oid, &result, &stable_count, &unstable_count);
+    if (op->list_stable_limit)
     {
-        pool_id_t pool_id = (min_inode >> (64-POOL_ID_BITS));
-        if (pg_count > 1)
+        // Ordered result is expected - used by scrub
+        // We use an unordered map
+        std::sort(result, result + stable_count);
+        if (stable_count > op->list_stable_limit)
         {
-            // Per-pg listing
-            auto sh_it = clean_db_settings.find(pool_id);
-            if (sh_it == clean_db_settings.end() ||
-                sh_it->second.pg_count != pg_count ||
-                sh_it->second.pg_stripe_size != pg_stripe_size)
-            {
-                reshard_clean_db(pool_id, pg_count, pg_stripe_size);
-            }
-            first_shard = last_shard = ((uint64_t)pool_id << (64-POOL_ID_BITS)) | list_pg;
-        }
-        else
-        {
-            // Per-pool listing
-            first_shard = ((uint64_t)pool_id << (64-POOL_ID_BITS));
-            last_shard = ((uint64_t)(pool_id+1) << (64-POOL_ID_BITS)) - 1;
+            memmove(result + op->list_stable_limit, result + stable_count, unstable_count);
+            stable_count = op->list_stable_limit;
         }
     }
-    // Copy clean_db entries
-    int stable_count = 0, stable_alloc = 0;
-    if (min_inode != max_inode)
-    {
-        for (auto shard_it = clean_db_shards.lower_bound(first_shard);
-            shard_it != clean_db_shards.end() && shard_it->first <= last_shard;
-            shard_it++)
-        {
-            auto & clean_db = shard_it->second;
-            stable_alloc += clean_db.size();
-        }
-    }
-    if (op->list_stable_limit > 0)
-    {
-        stable_alloc = op->list_stable_limit;
-        if (stable_alloc > 1024*1024)
-            stable_alloc = 1024*1024;
-    }
-    if (stable_alloc < 32768)
-    {
-        stable_alloc = 32768;
-    }
-    obj_ver_id *stable = (obj_ver_id*)malloc(sizeof(obj_ver_id) * stable_alloc);
-    if (!stable)
-    {
-        op->retval = -ENOMEM;
-        FINISH_OP(op);
-        return;
-    }
-    auto max_oid = op->max_oid;
-    bool limited = false;
-    pool_pg_id_t last_shard_id = 0;
-    for (auto shard_it = clean_db_shards.lower_bound(first_shard);
-        shard_it != clean_db_shards.end() && shard_it->first <= last_shard;
-        shard_it++)
-    {
-        auto & clean_db = shard_it->second;
-        auto clean_it = clean_db.begin(), clean_end = clean_db.end();
-        if (op->min_oid.inode != 0 || op->min_oid.stripe != 0)
-        {
-            clean_it = clean_db.lower_bound(op->min_oid);
-        }
-        if ((max_oid.inode != 0 || max_oid.stripe != 0) && !(max_oid < op->min_oid))
-        {
-            clean_end = clean_db.upper_bound(max_oid);
-        }
-        for (; clean_it != clean_end; clean_it++)
-        {
-            if (stable_count >= stable_alloc)
-            {
-                stable_alloc *= 2;
-                obj_ver_id* nst = (obj_ver_id*)realloc(stable, sizeof(obj_ver_id) * stable_alloc);
-                if (!nst)
-                {
-                    op->retval = -ENOMEM;
-                    FINISH_OP(op);
-                    return;
-                }
-                stable = nst;
-            }
-            stable[stable_count++] = {
-                .oid = clean_it->first,
-                .version = clean_it->second.version,
-            };
-            if (op->list_stable_limit > 0 && stable_count >= op->list_stable_limit)
-            {
-                if (!limited)
-                {
-                    limited = true;
-                    max_oid = stable[stable_count-1].oid;
-                }
-                break;
-            }
-        }
-        if (op->list_stable_limit > 0)
-        {
-            // To maintain the order, we have to include objects in the same range from other shards
-            if (last_shard_id != 0 && last_shard_id != shard_it->first)
-                std::sort(stable, stable+stable_count);
-            if (stable_count > op->list_stable_limit)
-                stable_count = op->list_stable_limit;
-        }
-        last_shard_id = shard_it->first;
-    }
-    if (op->list_stable_limit == 0 && first_shard != last_shard)
-    {
-        // If that's not a per-PG listing, sort clean entries (already sorted if list_stable_limit != 0)
-        std::sort(stable, stable+stable_count);
-    }
-    int clean_stable_count = stable_count;
-    // Copy dirty_db entries (sorted, too)
-    int unstable_count = 0, unstable_alloc = 0;
-    obj_ver_id *unstable = NULL;
-    {
-        auto dirty_it = dirty_db.begin(), dirty_end = dirty_db.end();
-        if (op->min_oid.inode != 0 || op->min_oid.stripe != 0)
-        {
-            dirty_it = dirty_db.lower_bound({
-                .oid = op->min_oid,
-                .version = 0,
-            });
-        }
-        if ((max_oid.inode != 0 || max_oid.stripe != 0) && !(max_oid < op->min_oid))
-        {
-            dirty_end = dirty_db.upper_bound({
-                .oid = max_oid,
-                .version = UINT64_MAX,
-            });
-        }
-        for (; dirty_it != dirty_end; dirty_it++)
-        {
-            if (!pg_count || ((dirty_it->first.oid.stripe / pg_stripe_size) % pg_count + 1) == list_pg) // like map_to_pg()
-            {
-                if (IS_DELETE(dirty_it->second.state))
-                {
-                    // Deletions are always stable, so try to zero out two possible entries
-                    if (!replace_stable(dirty_it->first.oid, 0, 0, clean_stable_count, stable))
-                    {
-                        replace_stable(dirty_it->first.oid, 0, clean_stable_count, stable_count, stable);
-                    }
-                }
-                else if (IS_STABLE(dirty_it->second.state) || (dirty_it->second.state & BS_ST_INSTANT))
-                {
-                    // First try to replace a clean stable version in the first part of the list
-                    if (!replace_stable(dirty_it->first.oid, dirty_it->first.version, 0, clean_stable_count, stable))
-                    {
-                        // Then try to replace the last dirty stable version in the second part of the list
-                        if (stable_count > 0 && stable[stable_count-1].oid == dirty_it->first.oid)
-                        {
-                            stable[stable_count-1].version = dirty_it->first.version;
-                        }
-                        else
-                        {
-                            if (stable_count >= stable_alloc)
-                            {
-                                stable_alloc += 32768;
-                                obj_ver_id *nst = (obj_ver_id*)realloc(stable, sizeof(obj_ver_id) * stable_alloc);
-                                if (!nst)
-                                {
-                                    if (unstable)
-                                        free(unstable);
-                                    op->retval = -ENOMEM;
-                                    FINISH_OP(op);
-                                    return;
-                                }
-                                stable = nst;
-                            }
-                            stable[stable_count++] = dirty_it->first;
-                        }
-                    }
-                    if (op->list_stable_limit > 0 && stable_count >= op->list_stable_limit)
-                    {
-                        // Stop here
-                        break;
-                    }
-                }
-                else
-                {
-                    if (unstable_count >= unstable_alloc)
-                    {
-                        unstable_alloc += 32768;
-                        obj_ver_id *nst = (obj_ver_id*)realloc(unstable, sizeof(obj_ver_id) * unstable_alloc);
-                        if (!nst)
-                        {
-                            if (stable)
-                                free(stable);
-                            op->retval = -ENOMEM;
-                            FINISH_OP(op);
-                            return;
-                        }
-                        unstable = nst;
-                    }
-                    unstable[unstable_count++] = dirty_it->first;
-                }
-            }
-        }
-    }
-    // Remove zeroed out stable entries
-    int j = 0;
-    for (int i = 0; i < stable_count; i++)
-    {
-        if (stable[i].version != 0)
-        {
-            stable[j++] = stable[i];
-        }
-    }
-    stable_count = j;
-    if (stable_count+unstable_count > stable_alloc)
-    {
-        stable_alloc = stable_count+unstable_count;
-        obj_ver_id *nst = (obj_ver_id*)realloc(stable, sizeof(obj_ver_id) * stable_alloc);
-        if (!nst)
-        {
-            if (unstable)
-                free(unstable);
-            op->retval = -ENOMEM;
-            FINISH_OP(op);
-            return;
-        }
-        stable = nst;
-    }
-    // Copy unstable entries
-    for (int i = 0; i < unstable_count; i++)
-    {
-        stable[j++] = unstable[i];
-    }
-    free(unstable);
     op->version = stable_count;
-    op->retval = stable_count+unstable_count;
-    op->buf = stable;
+    op->retval = res == 0 ? stable_count+unstable_count : -res;
+    op->buf = (uint8_t*)result;
     FINISH_OP(op);
 }
 
 void blockstore_impl_t::dump_diagnostics()
 {
-    journal.dump_diagnostics();
     flusher->dump_diagnostics();
 }
 
+void blockstore_meta_header_v3_t::set_crc32c()
+{
+    header_csum = 0;
+    uint32_t calc = crc32c(0, this, version == BLOCKSTORE_META_FORMAT_HEAP
+        ? sizeof(blockstore_meta_header_v3_t) : sizeof(blockstore_meta_header_v2_t));
+    header_csum = calc;
+}
+
 void blockstore_impl_t::disk_error_abort(const char *op, int retval, int expected)
 {
     if (retval == -EAGAIN)
@@ -741,85 +368,7 @@ void blockstore_impl_t::disk_error_abort(const char *op, int retval, int expecte
     exit(1);
 }
 
-void blockstore_impl_t::set_no_inode_stats(const std::vector<uint64_t> & pool_ids)
+uint64_t blockstore_impl_t::get_free_block_count()
 {
-    for (auto & np: no_inode_stats)
-    {
-        np.second = 2;
-    }
-    for (auto pool_id: pool_ids)
-    {
-        if (!no_inode_stats[pool_id])
-            recalc_inode_space_stats(pool_id, false);
-        no_inode_stats[pool_id] = 1;
-    }
-    for (auto np_it = no_inode_stats.begin(); np_it != no_inode_stats.end(); )
-    {
-        if (np_it->second == 2)
-        {
-            recalc_inode_space_stats(np_it->first, true);
-            no_inode_stats.erase(np_it++);
-        }
-        else
-            np_it++;
-    }
-}
-
-void blockstore_impl_t::recalc_inode_space_stats(uint64_t pool_id, bool per_inode)
-{
-    auto sp_begin = inode_space_stats.lower_bound((pool_id << (64-POOL_ID_BITS)));
-    auto sp_end = inode_space_stats.lower_bound(((pool_id+1) << (64-POOL_ID_BITS)));
-    inode_space_stats.erase(sp_begin, sp_end);
-    auto sh_it = clean_db_shards.lower_bound((pool_id << (64-POOL_ID_BITS)));
-    while (sh_it != clean_db_shards.end() &&
-        (sh_it->first >> (64-POOL_ID_BITS)) == pool_id)
-    {
-        for (auto & pair: sh_it->second)
-        {
-            uint64_t space_id = per_inode ? pair.first.inode : (pool_id << (64-POOL_ID_BITS));
-            inode_space_stats[space_id] += dsk.data_block_size;
-        }
-        sh_it++;
-    }
-    object_id last_oid = {};
-    bool last_exists = false;
-    auto dirty_it = dirty_db.lower_bound((obj_ver_id){ .oid = { .inode = (pool_id << (64-POOL_ID_BITS)) } });
-    while (dirty_it != dirty_db.end() && (dirty_it->first.oid.inode >> (64-POOL_ID_BITS)) == pool_id)
-    {
-        if (IS_STABLE(dirty_it->second.state) && (IS_BIG_WRITE(dirty_it->second.state) || IS_DELETE(dirty_it->second.state)))
-        {
-            bool exists = false;
-            if (last_oid == dirty_it->first.oid)
-            {
-                exists = last_exists;
-            }
-            else
-            {
-                auto & clean_db = clean_db_shard(dirty_it->first.oid);
-                auto clean_it = clean_db.find(dirty_it->first.oid);
-                exists = clean_it != clean_db.end();
-            }
-            uint64_t space_id = per_inode ? dirty_it->first.oid.inode : (pool_id << (64-POOL_ID_BITS));
-            if (IS_BIG_WRITE(dirty_it->second.state))
-            {
-                if (!exists)
-                    inode_space_stats[space_id] += dsk.data_block_size;
-                last_exists = true;
-            }
-            else
-            {
-                if (exists)
-                {
-                    auto & sp = inode_space_stats[space_id];
-                    if (sp > dsk.data_block_size)
-                        sp -= dsk.data_block_size;
-                    else
-                        inode_space_stats.erase(space_id);
-                }
-                last_exists = false;
-            }
-            last_oid = dirty_it->first.oid;
-        }
-        dirty_it++;
-    }
+    return dsk.block_count - heap->get_data_used_space()/dsk.data_block_size;
 }

src/blockstore/blockstore_impl.h

@@ -5,6 +5,7 @@
 
 #include "blockstore.h"
 #include "blockstore_disk.h"
+#include "blockstore_heap.h"
 
 #include <sys/types.h>
 #include <sys/ioctl.h>
@@ -19,46 +20,18 @@
 #include <deque>
 #include <new>
 #include <unordered_map>
-
-#include "cpp-btree/btree_map.h"
+#include <unordered_set>
 
 #include "malloc_or_die.h"
-#include "allocator.h"
 
 //#define BLOCKSTORE_DEBUG
 
-// States are not stored on disk. Instead, they're deduced from the journal
-
-#define BS_ST_SMALL_WRITE 0x01
-#define BS_ST_BIG_WRITE 0x02
-#define BS_ST_DELETE 0x03
-
-#define BS_ST_WAIT_DEL 0x10
-#define BS_ST_WAIT_BIG 0x20
-#define BS_ST_IN_FLIGHT 0x30
-#define BS_ST_SUBMITTED 0x40
-#define BS_ST_WRITTEN 0x50
-#define BS_ST_SYNCED 0x60
-#define BS_ST_STABLE 0x70
-
-#define BS_ST_INSTANT 0x100
-
 #define IMMEDIATE_NONE 0
 #define IMMEDIATE_SMALL 1
 #define IMMEDIATE_ALL 2
 
-#define BS_ST_TYPE_MASK 0x0F
-#define BS_ST_WORKFLOW_MASK 0xF0
-#define IS_IN_FLIGHT(st) (((st) & 0xF0) <= BS_ST_SUBMITTED)
-#define IS_STABLE(st) (((st) & 0xF0) == BS_ST_STABLE)
-#define IS_SYNCED(st) (((st) & 0xF0) >= BS_ST_SYNCED)
-#define IS_JOURNAL(st) (((st) & 0x0F) == BS_ST_SMALL_WRITE)
-#define IS_BIG_WRITE(st) (((st) & 0x0F) == BS_ST_BIG_WRITE)
-#define IS_DELETE(st) (((st) & 0x0F) == BS_ST_DELETE)
-#define IS_INSTANT(st) (((st) & BS_ST_TYPE_MASK) == BS_ST_DELETE || ((st) & BS_ST_INSTANT))
-
 #define BS_SUBMIT_CHECK_SQES(n) \
-    if (ringloop->sqes_left() < (n))\
+    if (ringloop->space_left() < (n))\
     {\
         /* Pause until there are more requests available */\
         PRIV(op)->wait_detail = (n);\
@@ -90,13 +63,6 @@
         return 0;\
     }
 
-#include "blockstore_journal.h"
-
-// "VITAstor"
-#define BLOCKSTORE_META_MAGIC_V1 0x726F747341544956l
-#define BLOCKSTORE_META_FORMAT_V1 1
-#define BLOCKSTORE_META_FORMAT_V2 2
-
 // metadata header (superblock)
 struct __attribute__((__packed__)) blockstore_meta_header_v1_t
 {
@@ -121,75 +87,26 @@ struct __attribute__((__packed__)) blockstore_meta_header_v2_t
     uint32_t header_csum;
 };
 
-// 32 bytes = 24 bytes + block bitmap (4 bytes by default) + external attributes (also bitmap, 4 bytes by default)
-// per "clean" entry on disk with fixed metadata tables
-struct __attribute__((__packed__)) clean_disk_entry
+struct __attribute__((__packed__)) blockstore_meta_header_v3_t
 {
-    object_id oid;
+    uint64_t zero;
+    uint64_t magic;
     uint64_t version;
-    uint8_t bitmap[];
-    // Two more fields come after bitmap in metadata version 2:
-    // uint32_t data_csum[];
-    // uint32_t entry_csum;
-};
+    uint32_t meta_block_size;
+    uint32_t data_block_size;
+    uint32_t bitmap_granularity;
+    uint32_t data_csum_type;
+    uint32_t csum_block_size;
+    uint32_t header_csum;
+    uint64_t compacted_lsn;
 
-// 32 = 16 + 16 bytes per "clean" entry in memory (object_id => clean_entry)
-struct __attribute__((__packed__)) clean_entry
-{
-    uint64_t version;
-    uint64_t location;
+    void set_crc32c();
 };
 
-// 64 = 24 + 40 bytes per dirty entry in memory (obj_ver_id => dirty_entry). Plus checksums
-struct __attribute__((__packed__)) dirty_entry
-{
-    uint32_t state;
-    uint32_t flags;    // unneeded, but present for alignment
-    uint64_t location; // location in either journal or data -> in BYTES
-    uint32_t offset;   // data offset within object (stripe)
-    uint32_t len;      // data length
-    uint64_t journal_sector; // journal sector used for this entry
-    void* dyn_data;    // dynamic data: external bitmap and data block checksums. may be a pointer to the in-memory journal
-};
-
-// - Sync must be submitted after previous writes/deletes (not before!)
-// - Reads to the same object must be submitted after previous writes/deletes
-//   are written (not necessarily synced) in their location. This is because we
-//   rely on read-modify-write for erasure coding and we must return new data
-//   to calculate parity for subsequent writes
-// - Writes may be submitted in any order, because they don't overlap. Each write
-//   goes into a new location - either on the journal device or on the data device
-// - Stable (stabilize) must be submitted after sync of that object is completed
-//   It's even OK to return an error to the caller if that object is not synced yet
-// - Journal trim may be processed only after all versions are moved to
-//   the main storage AND after all read operations for older versions complete
-// - If an operation can not be submitted because the ring is full
-//   we should stop submission of other operations. Otherwise some "scatter" reads
-//   may end up blocked for a long time.
-// Otherwise, the submit order is free, that is all operations may be submitted immediately
-// In fact, adding a write operation must immediately result in dirty_db being populated
-
 // Suspend operation until there are more free SQEs
 #define WAIT_SQE 1
-// Suspend operation until there are <wait_detail> bytes of free space in the journal on disk
-#define WAIT_JOURNAL 3
-// Suspend operation until the next journal sector buffer is free
-#define WAIT_JOURNAL_BUFFER 4
-// Suspend operation until there is some free space on the data device
-#define WAIT_FREE 5
-
-struct used_clean_obj_t
-{
-    int refs;
-    bool was_freed; // was freed by a parallel flush?
-    bool was_changed; // was changed by a parallel flush?
-};
-
-// https://github.com/algorithm-ninja/cpp-btree
-// https://github.com/greg7mdp/sparsepp/ was used previously, but it was TERRIBLY slow after resizing
-// with sparsepp, random reads dropped to ~700 iops very fast with just as much as ~32k objects in the DB
-typedef btree::btree_map<object_id, clean_entry> blockstore_clean_db_t;
-typedef std::map<obj_ver_id, dirty_entry> blockstore_dirty_db_t;
+// Suspend until something is compacted
+#define WAIT_COMPACTION 2
 
 #include "blockstore_init.h"
 
@@ -202,58 +119,47 @@ struct blockstore_op_private_t
 {
     // Wait status
     int wait_for;
-    uint64_t wait_detail, wait_detail2;
+    uint64_t wait_detail;
     int pending_ops;
     int op_state;
 
+    // Read, write, sync, stabilize
+    uint64_t lsn;
+
     // Read
-    uint64_t clean_block_used;
     std::vector<copy_buffer_t> read_vec;
 
-    // Sync, write
-    uint64_t min_flushed_journal_sector, max_flushed_journal_sector;
+    // Write
+    uint64_t location;
+    bool is_big;
+
+    // Stabilize, rollback
+    int stab_pos;
+
+    // Stabilize
+    uint64_t to_lsn;
 
     // Write
     struct iovec iov_zerofill[3];
-    // Warning: must not have a default value here because it's written to before calling constructor in blockstore_write.cpp O_o
-    uint64_t real_version;
     timespec tv_begin;
-
-    // Sync
-    std::vector<obj_ver_id> sync_big_writes, sync_small_writes;
 };
 
-typedef uint32_t pool_id_t;
-typedef uint64_t pool_pg_id_t;
-
-#define POOL_ID_BITS 16
-
-struct pool_shard_settings_t
-{
-    uint32_t pg_count;
-    uint32_t pg_stripe_size;
-};
-
-#define STAB_SPLIT_DONE 1
-#define STAB_SPLIT_WAIT 2
-#define STAB_SPLIT_SYNC 3
-#define STAB_SPLIT_TODO 4
-
 class blockstore_impl_t
 {
+public:
     blockstore_disk_t dsk;
 
     /******* OPTIONS *******/
     bool readonly = false;
-    // It is safe to disable fsync() if drive write cache is writethrough
-    bool disable_data_fsync = false, disable_meta_fsync = false, disable_journal_fsync = false;
     // Enable if you want every operation to be executed with an "implicit fsync"
     // Suitable only for server SSDs with capacitors, requires disabled data and journal fsyncs
     int immediate_commit = IMMEDIATE_NONE;
     bool inmemory_meta = false;
+    uint32_t meta_write_recheck_parallelism = 0;
     // Maximum and minimum flusher count
-    unsigned max_flusher_count, min_flusher_count;
-    unsigned journal_trim_interval;
+    unsigned max_flusher_count = 0, min_flusher_count = 0;
+    unsigned journal_trim_interval = 0;
+    unsigned flusher_start_threshold = 0;
     // Maximum queue depth
     unsigned max_write_iodepth = 128;
     // Enable small (journaled) write throttling, useful for the SSD+HDD case
@@ -268,139 +174,89 @@ class blockstore_impl_t
     uint64_t autosync_writes = 128;
     // Log level (0-10)
     int log_level = 0;
+    // Enable correct block checksum validation on objects updated with small writes when checksum block
+    // is larger than bitmap_granularity, at the expense of extra metadata fsyncs during compaction
+    bool perfect_csum_update = false;
     /******* END OF OPTIONS *******/
 
     struct ring_consumer_t ring_consumer;
 
-    std::map<pool_id_t, pool_shard_settings_t> clean_db_settings;
-    std::map<pool_pg_id_t, blockstore_clean_db_t> clean_db_shards;
-    std::map<uint64_t, int> no_inode_stats;
-    uint8_t *clean_bitmaps = NULL;
-    blockstore_dirty_db_t dirty_db;
+    blockstore_heap_t *heap = NULL;
+    uint8_t* meta_superblock = NULL;
+    uint8_t *buffer_area = NULL;
     std::vector<blockstore_op_t*> submit_queue;
-    std::vector<obj_ver_id> unsynced_big_writes, unsynced_small_writes;
-    int unsynced_big_write_count = 0, unstable_unsynced = 0;
+    int unsynced_big_write_count = 0, unsynced_small_write_count = 0, unsynced_meta_write_count = 0;
     int unsynced_queued_ops = 0;
-    allocator_t *data_alloc = NULL;
-    uint64_t used_blocks = 0;
     uint8_t *zero_object = NULL;
 
-    void *metadata_buffer = NULL;
-
-    struct journal_t journal;
     journal_flusher_t *flusher;
-    int big_to_flush = 0;
     int write_iodepth = 0;
-    bool alloc_dyn_data = false;
-
-    // clean data blocks referenced by read operations
-    std::map<uint64_t, used_clean_obj_t> used_clean_objects;
+    int inflight_big = 0;
+    bool fsyncing_data = false;
 
     bool live = false, queue_stall = false;
-    ring_loop_t *ringloop;
-    timerfd_manager_t *tfd;
+    ring_loop_i *ringloop = NULL;
+    timerfd_manager_t *tfd = NULL;
 
-    bool stop_sync_submitted;
+    bool stop_sync_submitted = false;
 
     inline struct io_uring_sqe* get_sqe()
     {
         return ringloop->get_sqe();
     }
 
-    friend class blockstore_init_meta;
-    friend class blockstore_init_journal;
-    friend struct blockstore_journal_check_t;
-    friend class journal_flusher_t;
-    friend class journal_flusher_co;
-
-    void calc_lengths();
     void open_data();
     void open_meta();
     void open_journal();
-    uint8_t* get_clean_entry_bitmap(uint64_t block_loc, int offset);
 
-    blockstore_clean_db_t& clean_db_shard(object_id oid);
-    void reshard_clean_db(pool_id_t pool_id, uint32_t pg_count, uint32_t pg_stripe_size);
-    void recalc_inode_space_stats(uint64_t pool_id, bool per_inode);
-
-    // Journaling
-    void prepare_journal_sector_write(int sector, blockstore_op_t *op);
-    void handle_journal_write(ring_data_t *data, uint64_t flush_id);
     void disk_error_abort(const char *op, int retval, int expected);
 
     // Asynchronous init
     int initialized;
     int metadata_buf_size;
     blockstore_init_meta* metadata_init_reader;
-    blockstore_init_journal* journal_init_reader;
 
+    void init();
     void check_wait(blockstore_op_t *op);
     void init_op(blockstore_op_t *op);
 
     // Read
-    int dequeue_read(blockstore_op_t *read_op);
+    int dequeue_read(blockstore_op_t *op);
+    int fulfill_read(blockstore_op_t *op);
+    uint32_t prepare_read(std::vector<copy_buffer_t> & read_vec, heap_object_t *obj, heap_write_t *wr, uint32_t start, uint32_t end);
+    uint32_t prepare_read_with_bitmaps(std::vector<copy_buffer_t> & read_vec, heap_object_t *obj, heap_write_t *wr, uint32_t start, uint32_t end);
+    uint32_t prepare_read_zero(std::vector<copy_buffer_t> & read_vec, uint32_t start, uint32_t end);
+    uint32_t prepare_read_simple(std::vector<copy_buffer_t> & read_vec, heap_object_t *obj, heap_write_t *wr, uint32_t start, uint32_t end);
+    void prepare_disk_read(std::vector<copy_buffer_t> & read_vec, int pos, heap_object_t *obj, heap_write_t *wr,
+        uint32_t blk_start, uint32_t blk_end, uint32_t start, uint32_t end, uint32_t copy_flags);
     void find_holes(std::vector<copy_buffer_t> & read_vec, uint32_t item_start, uint32_t item_end,
-        std::function<int(int, bool, uint32_t, uint32_t)> callback);
-    int fulfill_read(blockstore_op_t *read_op,
-        uint64_t &fulfilled, uint32_t item_start, uint32_t item_end,
-        uint32_t item_state, uint64_t item_version, uint64_t item_location,
-        uint64_t journal_sector, uint8_t *csum, int *dyn_data);
-    bool fulfill_clean_read(blockstore_op_t *read_op, uint64_t & fulfilled,
-        uint8_t *clean_entry_bitmap, int *dyn_data,
-        uint32_t item_start, uint32_t item_end, uint64_t clean_loc, uint64_t clean_ver);
-    int fill_partial_checksum_blocks(std::vector<copy_buffer_t> & rv, uint64_t & fulfilled,
-        uint8_t *clean_entry_bitmap, int *dyn_data, bool from_journal, uint8_t *read_buf, uint64_t read_offset, uint64_t read_end);
-    int pad_journal_read(std::vector<copy_buffer_t> & rv, copy_buffer_t & cp,
-        uint64_t dirty_offset, uint64_t dirty_end, uint64_t dirty_loc, uint8_t *csum_ptr, int *dyn_data,
-        uint64_t offset, uint64_t submit_len, uint64_t & blk_begin, uint64_t & blk_end, uint8_t* & blk_buf);
-    bool read_range_fulfilled(std::vector<copy_buffer_t> & rv, uint64_t & fulfilled, uint8_t *read_buf,
-        uint8_t *clean_entry_bitmap, uint32_t item_start, uint32_t item_end);
-    bool read_checksum_block(blockstore_op_t *op, int rv_pos, uint64_t &fulfilled, uint64_t clean_loc);
-    uint8_t* read_clean_meta_block(blockstore_op_t *read_op, uint64_t clean_loc, int rv_pos);
-    bool verify_padded_checksums(uint8_t *clean_entry_bitmap, uint8_t *csum_buf, uint32_t offset,
-        iovec *iov, int n_iov, std::function<void(uint32_t, uint32_t, uint32_t)> bad_block_cb);
-    bool verify_journal_checksums(uint8_t *csums, uint32_t offset,
-        iovec *iov, int n_iov, std::function<void(uint32_t, uint32_t, uint32_t)> bad_block_cb);
-    bool verify_clean_padded_checksums(blockstore_op_t *op, uint64_t clean_loc, uint8_t *dyn_data, bool from_journal,
-        iovec *iov, int n_iov, std::function<void(uint32_t, uint32_t, uint32_t)> bad_block_cb);
-    int fulfill_read_push(blockstore_op_t *op, void *buf, uint64_t offset, uint64_t len,
-        uint32_t item_state, uint64_t item_version);
+        std::function<void(int&, uint32_t, uint32_t)> callback);
+    void free_read_buffers(std::vector<copy_buffer_t> & rv);
     void handle_read_event(ring_data_t *data, blockstore_op_t *op);
+    bool verify_read_checksums(blockstore_op_t *op);
 
     // Write
     bool enqueue_write(blockstore_op_t *op);
-    void cancel_all_writes(blockstore_op_t *op, blockstore_dirty_db_t::iterator dirty_it, int retval);
+    void prepare_meta_block_write(blockstore_op_t *op, uint64_t modified_block, io_uring_sqe *sqe = NULL);
     int dequeue_write(blockstore_op_t *op);
-    int dequeue_del(blockstore_op_t *op);
+    int make_big_write(blockstore_op_t *op, uint32_t offset, uint32_t len, uint32_t *modified_block, uint32_t *moved_from_block);
     int continue_write(blockstore_op_t *op);
-    void release_journal_sectors(blockstore_op_t *op);
     void handle_write_event(ring_data_t *data, blockstore_op_t *op);
 
     // Sync
     int continue_sync(blockstore_op_t *op);
-    void ack_sync(blockstore_op_t *op);
+    bool submit_fsyncs(int & wait_count);
+    int do_sync(blockstore_op_t *op, int base_state);
 
     // Stabilize
     int dequeue_stable(blockstore_op_t *op);
-    int continue_stable(blockstore_op_t *op);
-    void mark_stable(obj_ver_id ov, bool forget_dirty = false);
-    void stabilize_object(object_id oid, uint64_t max_ver);
-    blockstore_op_t* selective_sync(blockstore_op_t *op);
-    int split_stab_op(blockstore_op_t *op, std::function<int(obj_ver_id v)> decider);
-
-    // Rollback
-    int dequeue_rollback(blockstore_op_t *op);
-    int continue_rollback(blockstore_op_t *op);
-    void mark_rolled_back(const obj_ver_id & ov);
-    void erase_dirty(blockstore_dirty_db_t::iterator dirty_start, blockstore_dirty_db_t::iterator dirty_end, uint64_t clean_loc);
-    void free_dirty_dyn_data(dirty_entry & e);
 
     // List
     void process_list(blockstore_op_t *op);
 
-public:
+/*public:*/
 
-    blockstore_impl_t(blockstore_config_t & config, ring_loop_t *ringloop, timerfd_manager_t *tfd);
+    blockstore_impl_t(blockstore_config_t & config, ring_loop_i *ringloop, timerfd_manager_t *tfd, bool mock_mode = false);
     ~blockstore_impl_t();
 
     void parse_config(blockstore_config_t & config, bool init);
@@ -426,21 +282,13 @@ public:
     // Simplified synchronous operation: get object bitmap & current version
     int read_bitmap(object_id oid, uint64_t target_version, void *bitmap, uint64_t *result_version = NULL);
 
-    // Unstable writes are added here (map of object_id -> version)
-    std::unordered_map<object_id, uint64_t> unstable_writes;
-
-    // Space usage statistics
-    std::map<uint64_t, uint64_t> inode_space_stats;
-
-    // Set per-pool no_inode_stats
-    void set_no_inode_stats(const std::vector<uint64_t> & pool_ids);
-
     // Print diagnostics to stdout
     void dump_diagnostics();
 
+    const std::map<uint64_t, uint64_t> & get_inode_space_stats() { return heap->get_inode_space_stats(); }
     inline uint32_t get_block_size() { return dsk.data_block_size; }
     inline uint64_t get_block_count() { return dsk.block_count; }
-    inline uint64_t get_free_block_count() { return dsk.block_count - used_blocks; }
+    uint64_t get_free_block_count();
     inline uint32_t get_bitmap_granularity() { return dsk.disk_alignment; }
     inline uint64_t get_journal_size() { return dsk.journal_len; }
 };

src/blockstore/blockstore_init.h

@@ -25,47 +25,10 @@ class blockstore_init_meta
     uint64_t next_offset = 0;
     uint64_t last_read_offset = 0;
     uint64_t entries_loaded = 0;
-    unsigned entries_per_block = 0;
     int i = 0, j = 0;
-    std::vector<uint64_t> entries_to_zero;
     bool handle_meta_block(uint8_t *buf, uint64_t count, uint64_t done_cnt);
     void handle_event(ring_data_t *data, int buf_num);
 public:
     blockstore_init_meta(blockstore_impl_t *bs);
     int loop();
 };
-
-struct bs_init_journal_done
-{
-    void *buf;
-    uint64_t pos, len;
-};
-
-class blockstore_init_journal
-{
-    blockstore_impl_t *bs;
-    int wait_state = 0, wait_count = 0, handle_res = 0;
-    uint64_t entries_loaded = 0;
-    uint32_t crc32_last = 0;
-    bool started = false;
-    uint64_t next_free;
-    std::vector<bs_init_journal_done> done;
-    std::vector<obj_ver_id> double_allocs;
-    std::vector<iovec> small_write_data;
-    uint64_t journal_pos = 0;
-    uint64_t continue_pos = 0;
-    void *init_write_buf = NULL;
-    uint64_t init_write_sector = 0;
-    bool wrapped = false;
-    void *submitted_buf;
-    struct io_uring_sqe *sqe;
-    struct ring_data_t *data;
-    journal_entry_start *je_start;
-    std::function<void(ring_data_t*)> simple_callback;
-    int handle_journal_part(void *buf, uint64_t done_pos, uint64_t len);
-    void handle_event(ring_data_t *data);
-    void erase_dirty_object(blockstore_dirty_db_t::iterator dirty_it);
-public:
-    blockstore_init_journal(blockstore_impl_t* bs);
-    int loop();
-};

src/blockstore/blockstore_journal.cpp

@@ -1,356 +0,0 @@
-// Copyright (c) Vitaliy Filippov, 2019+
-// License: VNPL-1.1 (see README.md for details)
-
-#include "blockstore_impl.h"
-
-blockstore_journal_check_t::blockstore_journal_check_t(blockstore_impl_t *bs)
-{
-    this->bs = bs;
-    sectors_to_write = 0;
-    next_pos = bs->journal.next_free;
-    next_sector = bs->journal.cur_sector;
-    first_sector = -1;
-    next_in_pos = bs->journal.in_sector_pos;
-    right_dir = next_pos >= bs->journal.used_start;
-}
-
-// Check if we can write <required> entries of <size> bytes and <data_after> data bytes after them to the journal
-int blockstore_journal_check_t::check_available(blockstore_op_t *op, int entries_required, int size, int data_after)
-{
-    uint64_t prev_next = next_sector;
-    int required = entries_required;
-    while (1)
-    {
-        int fits = bs->journal.no_same_sector_overwrites && next_pos == bs->journal.next_free && bs->journal.sector_info[next_sector].written
-            ? 0
-            : (bs->journal.block_size - next_in_pos) / size;
-        if (fits > 0)
-        {
-            if (fits > required)
-            {
-                fits = required;
-            }
-            if (first_sector == -1)
-            {
-                first_sector = next_sector;
-            }
-            required -= fits;
-            next_in_pos += fits * size;
-            if (next_sector != prev_next || !sectors_to_write)
-            {
-                // Except the previous call to this function
-                sectors_to_write++;
-            }
-        }
-        else if (bs->journal.sector_info[next_sector].dirty)
-        {
-            if (next_sector != prev_next || !sectors_to_write)
-            {
-                // Except the previous call to this function
-                sectors_to_write++;
-            }
-        }
-        if (required <= 0)
-        {
-            break;
-        }
-        next_pos = next_pos + bs->journal.block_size;
-        if (next_pos >= bs->journal.len)
-        {
-            next_pos = bs->journal.block_size;
-            right_dir = false;
-        }
-        next_in_pos = 0;
-        next_sector = ((next_sector + 1) % bs->journal.sector_count);
-        if (next_sector == first_sector)
-        {
-            // next_sector may wrap when all sectors are flushed and the incoming batch is too big
-            // This is an error condition, we can't wait for anything in this case
-            throw std::runtime_error(
-                "Blockstore journal_sector_buffer_count="+std::to_string(bs->journal.sector_count)+
-                " is too small for a batch of "+std::to_string(entries_required)+" entries of "+std::to_string(size)+" bytes"
-            );
-        }
-        if (bs->journal.sector_info[next_sector].flush_count > 0 ||
-            bs->journal.sector_info[next_sector].dirty)
-        {
-            // No memory buffer available. Wait for it.
-            int used = 0, dirty = 0;
-            for (int i = 0; i < bs->journal.sector_count; i++)
-            {
-                if (bs->journal.sector_info[i].dirty)
-                {
-                    dirty++;
-                    used++;
-                }
-                if (bs->journal.sector_info[i].flush_count > 0)
-                {
-                    used++;
-                }
-            }
-            // In fact, it's even more rare than "ran out of journal space", so print a warning
-            printf(
-                "Ran out of journal sector buffers: %d/%ju buffers used (%d dirty), next buffer (%jd)"
-                " is %s and flushed %ju times. Consider increasing \'journal_sector_buffer_count\'\n",
-                used, bs->journal.sector_count, dirty, next_sector,
-                bs->journal.sector_info[next_sector].dirty ? "dirty" : "not dirty",
-                bs->journal.sector_info[next_sector].flush_count
-            );
-            PRIV(op)->wait_for = WAIT_JOURNAL_BUFFER;
-            return 0;
-        }
-    }
-    if (data_after > 0)
-    {
-        next_pos = next_pos + data_after;
-        if (next_pos >= bs->journal.len)
-        {
-            if (right_dir)
-                next_pos = bs->journal.block_size + data_after;
-            right_dir = false;
-        }
-    }
-    if (!right_dir && next_pos >= bs->journal.used_start-bs->journal.block_size)
-    {
-        // No space in the journal. Wait until used_start changes.
-        if (bs->log_level > 5)
-        {
-            printf(
-                "Ran out of journal space (used_start=%08jx, next_free=%08jx, dirty_start=%08jx)\n",
-                bs->journal.used_start, bs->journal.next_free, bs->journal.dirty_start
-            );
-        }
-        PRIV(op)->wait_for = WAIT_JOURNAL;
-        bs->flusher->request_trim();
-        PRIV(op)->wait_detail = bs->journal.used_start;
-        return 0;
-    }
-    return 1;
-}
-
-journal_entry* prefill_single_journal_entry(journal_t & journal, uint16_t type, uint32_t size)
-{
-    if (!journal.entry_fits(size))
-    {
-        assert(!journal.sector_info[journal.cur_sector].dirty);
-        // Move to the next journal sector
-        if (journal.sector_info[journal.cur_sector].flush_count > 0)
-        {
-            // Also select next sector buffer in memory
-            journal.cur_sector = ((journal.cur_sector + 1) % journal.sector_count);
-            assert(!journal.sector_info[journal.cur_sector].flush_count);
-        }
-        else
-        {
-            journal.dirty_start = journal.next_free;
-        }
-        journal.sector_info[journal.cur_sector].written = false;
-        journal.sector_info[journal.cur_sector].offset = journal.next_free;
-        journal.in_sector_pos = 0;
-        auto next_next_free = (journal.next_free+journal.block_size) < journal.len ? journal.next_free + journal.block_size : journal.block_size;
-        // double check that next_free doesn't cross used_start from the left
-        assert(journal.next_free >= journal.used_start && next_next_free >= journal.next_free || next_next_free < journal.used_start);
-        journal.next_free = next_next_free;
-        memset(journal.inmemory
-            ? (uint8_t*)journal.buffer + journal.sector_info[journal.cur_sector].offset
-            : (uint8_t*)journal.sector_buf + journal.block_size*journal.cur_sector, 0, journal.block_size);
-    }
-    journal_entry *je = (struct journal_entry*)(
-        (journal.inmemory
-            ? (uint8_t*)journal.buffer + journal.sector_info[journal.cur_sector].offset
-            : (uint8_t*)journal.sector_buf + journal.block_size*journal.cur_sector) + journal.in_sector_pos
-    );
-    journal.in_sector_pos += size;
-    je->magic = JOURNAL_MAGIC;
-    je->type = type;
-    je->size = size;
-    je->crc32_prev = journal.crc32_last;
-    journal.sector_info[journal.cur_sector].dirty = true;
-    return je;
-}
-
-void blockstore_impl_t::prepare_journal_sector_write(int cur_sector, blockstore_op_t *op)
-{
-    // Don't submit the same sector twice in the same batch
-    if (!journal.sector_info[cur_sector].submit_id)
-    {
-        io_uring_sqe *sqe = get_sqe();
-        // Caller must ensure availability of an SQE
-        assert(sqe != NULL);
-        ring_data_t *data = ((ring_data_t*)sqe->user_data);
-        journal.sector_info[cur_sector].written = true;
-        journal.sector_info[cur_sector].submit_id = ++journal.submit_id;
-        assert(journal.submit_id != 0); // check overflow
-        journal.submitting_sectors.push_back(cur_sector);
-        journal.sector_info[cur_sector].flush_count++;
-        data->iov = (struct iovec){
-            (journal.inmemory
-                ? (uint8_t*)journal.buffer + journal.sector_info[cur_sector].offset
-                : (uint8_t*)journal.sector_buf + journal.block_size*cur_sector),
-            (size_t)journal.block_size
-        };
-        data->callback = [this, flush_id = journal.submit_id](ring_data_t *data) { handle_journal_write(data, flush_id); };
-        my_uring_prep_writev(
-            sqe, dsk.journal_fd, &data->iov, 1, journal.offset + journal.sector_info[cur_sector].offset
-        );
-    }
-    journal.sector_info[cur_sector].dirty = false;
-    // But always remember that this operation has to wait until this exact journal write is finished
-    journal.flushing_ops.emplace(journal.sector_info[cur_sector].submit_id, (pending_journaling_t){
-        .pending = 1,
-        .sector = cur_sector,
-        .op = op,
-    });
-    auto priv = PRIV(op);
-    priv->pending_ops++;
-    if (!priv->min_flushed_journal_sector)
-        priv->min_flushed_journal_sector = 1+cur_sector;
-    assert(priv->min_flushed_journal_sector <= journal.sector_count);
-    priv->max_flushed_journal_sector = 1+cur_sector;
-}
-
-void blockstore_impl_t::handle_journal_write(ring_data_t *data, uint64_t flush_id)
-{
-    live = true;
-    if (data->res != data->iov.iov_len)
-    {
-        // FIXME: our state becomes corrupted after a write error. maybe do something better than just die
-        disk_error_abort("journal write", data->res, data->iov.iov_len);
-    }
-    auto fl_it = journal.flushing_ops.lower_bound(flush_id);
-    if (fl_it != journal.flushing_ops.end() && fl_it->first == flush_id && fl_it->second.sector >= 0)
-    {
-        journal.sector_info[fl_it->second.sector].flush_count--;
-    }
-    auto is_first = fl_it == journal.flushing_ops.begin();
-    while (fl_it != journal.flushing_ops.end())
-    {
-        bool del = false;
-        if (fl_it->first == flush_id)
-        {
-            fl_it->second.pending = 0;
-            del = is_first;
-        }
-        else
-        {
-            del = !fl_it->second.pending;
-        }
-        if (del)
-        {
-            // Do not complete this operation if previous writes are unfinished
-            // Otherwise also complete following operations waiting for this one
-            auto priv = PRIV(fl_it->second.op);
-            priv->pending_ops--;
-            assert(priv->pending_ops >= 0);
-            if (priv->pending_ops == 0)
-            {
-                release_journal_sectors(fl_it->second.op);
-                priv->op_state++;
-                ringloop->wakeup();
-            }
-            journal.flushing_ops.erase(fl_it++);
-        }
-        else
-        {
-            fl_it++;
-        }
-    }
-}
-
-journal_t::~journal_t()
-{
-    if (sector_buf)
-        free(sector_buf);
-    if (sector_info)
-        free(sector_info);
-    if (buffer)
-        free(buffer);
-    sector_buf = NULL;
-    sector_info = NULL;
-    buffer = NULL;
-}
-
-uint64_t journal_t::get_trim_pos()
-{
-    auto journal_used_it = used_sectors.lower_bound(used_start);
-    if (journal_used_it == used_sectors.end())
-    {
-        // Journal is cleared to its end, restart from the beginning
-        journal_used_it = used_sectors.begin();
-        if (journal_used_it == used_sectors.end())
-        {
-            // Journal is empty
-            return next_free;
-        }
-        else
-        {
-            // next_free does not need updating during trim
-#ifdef BLOCKSTORE_DEBUG
-            printf(
-                "Trimming journal (used_start=%08jx, next_free=%08jx, dirty_start=%08jx, new_start=%08jx, new_refcount=%jd)\n",
-                used_start, next_free, dirty_start,
-                journal_used_it->first, journal_used_it->second
-            );
-#endif
-            return journal_used_it->first;
-        }
-    }
-    else if (journal_used_it->first > used_start)
-    {
-        // Journal is cleared up to <journal_used_it>
-#ifdef BLOCKSTORE_DEBUG
-        printf(
-            "Trimming journal (used_start=%08jx, next_free=%08jx, dirty_start=%08jx, new_start=%08jx, new_refcount=%jd)\n",
-            used_start, next_free, dirty_start,
-            journal_used_it->first, journal_used_it->second
-        );
-#endif
-        return journal_used_it->first;
-    }
-    // Can't trim journal
-    return used_start;
-}
-
-void journal_t::dump_diagnostics()
-{
-    auto journal_used_it = used_sectors.lower_bound(used_start);
-    if (journal_used_it == used_sectors.end())
-    {
-        // Journal is cleared to its end, restart from the beginning
-        journal_used_it = used_sectors.begin();
-    }
-    printf(
-        "Journal: used_start=%08jx next_free=%08jx dirty_start=%08jx trim_to=%08jx trim_to_refs=%jd\n",
-        used_start, next_free, dirty_start,
-        journal_used_it == used_sectors.end() ? 0 : journal_used_it->first,
-        journal_used_it == used_sectors.end() ? 0 : journal_used_it->second
-    );
-}
-
-static uint64_t zero_page[4096];
-
-uint32_t crc32c_pad(uint32_t prev_crc, const void *buf, size_t len, size_t left_pad, size_t right_pad)
-{
-    uint32_t r = prev_crc;
-    while (left_pad >= 4096)
-    {
-        r = crc32c(r, zero_page, 4096);
-        left_pad -= 4096;
-    }
-    if (left_pad > 0)
-        r = crc32c(r, zero_page, left_pad);
-    r = crc32c(r, buf, len);
-    while (right_pad >= 4096)
-    {
-        r = crc32c(r, zero_page, 4096);
-        right_pad -= 4096;
-    }
-    if (left_pad > 0)
-        r = crc32c(r, zero_page, right_pad);
-    return r;
-}
-
-uint32_t crc32c_nopad(uint32_t prev_crc, const void *buf, size_t len, size_t left_pad, size_t right_pad)
-{
-    return crc32c(0, buf, len);
-}

src/blockstore/blockstore_open.cpp

@@ -2,6 +2,7 @@
 // License: VNPL-1.1 (see README.md for details)
 
 #include <sys/file.h>
+#include <stdexcept>
 #include "blockstore_impl.h"
 
 void blockstore_impl_t::parse_config(blockstore_config_t & config, bool init)
@@ -14,12 +15,14 @@ void blockstore_impl_t::parse_config(blockstore_config_t & config, bool init)
     }
     min_flusher_count = strtoull(config["min_flusher_count"].c_str(), NULL, 10);
     journal_trim_interval = strtoull(config["journal_trim_interval"].c_str(), NULL, 10);
+    flusher_start_threshold = strtoull(config["flusher_start_threshold"].c_str(), NULL, 10);
     max_write_iodepth = strtoull(config["max_write_iodepth"].c_str(), NULL, 10);
     throttle_small_writes = config["throttle_small_writes"] == "true" || config["throttle_small_writes"] == "1" || config["throttle_small_writes"] == "yes";
     throttle_target_iops = strtoull(config["throttle_target_iops"].c_str(), NULL, 10);
     throttle_target_mbs = strtoull(config["throttle_target_mbs"].c_str(), NULL, 10);
     throttle_target_parallelism = strtoull(config["throttle_target_parallelism"].c_str(), NULL, 10);
     throttle_threshold_us = strtoull(config["throttle_threshold_us"].c_str(), NULL, 10);
+    perfect_csum_update = config["perfect_csum_update"] == "true" || config["perfect_csum_update"] == "1" || config["perfect_csum_update"] == "yes";
     if (config["autosync_writes"] != "")
     {
         autosync_writes = strtoull(config["autosync_writes"].c_str(), NULL, 10);
@@ -28,13 +31,17 @@ void blockstore_impl_t::parse_config(blockstore_config_t & config, bool init)
     {
         max_flusher_count = 256;
     }
-    if (!min_flusher_count || journal.flush_journal)
+    if (!min_flusher_count)
     {
         min_flusher_count = 1;
     }
     if (!journal_trim_interval)
     {
-        journal_trim_interval = 512;
+        journal_trim_interval = 1024;
+    }
+    if (!flusher_start_threshold)
+    {
+        flusher_start_threshold = 32;
     }
     if (!max_write_iodepth)
     {
@@ -68,23 +75,6 @@ void blockstore_impl_t::parse_config(blockstore_config_t & config, bool init)
     {
         readonly = true;
     }
-    if (config["disable_data_fsync"] == "true" || config["disable_data_fsync"] == "1" || config["disable_data_fsync"] == "yes")
-    {
-        disable_data_fsync = true;
-    }
-    if (config["disable_meta_fsync"] == "true" || config["disable_meta_fsync"] == "1" || config["disable_meta_fsync"] == "yes")
-    {
-        disable_meta_fsync = true;
-    }
-    if (config["disable_journal_fsync"] == "true" || config["disable_journal_fsync"] == "1" || config["disable_journal_fsync"] == "yes")
-    {
-        disable_journal_fsync = true;
-    }
-    if (config["flush_journal"] == "true" || config["flush_journal"] == "1" || config["flush_journal"] == "yes")
-    {
-        // Only flush journal and exit
-        journal.flush_journal = true;
-    }
     if (config["immediate_commit"] == "all")
     {
         immediate_commit = IMMEDIATE_ALL;
@@ -94,85 +84,23 @@ void blockstore_impl_t::parse_config(blockstore_config_t & config, bool init)
         immediate_commit = IMMEDIATE_SMALL;
     }
     metadata_buf_size = strtoull(config["meta_buf_size"].c_str(), NULL, 10);
-    inmemory_meta = config["inmemory_metadata"] != "false" && config["inmemory_metadata"] != "0" &&
-        config["inmemory_metadata"] != "no";
-    journal.sector_count = strtoull(config["journal_sector_buffer_count"].c_str(), NULL, 10);
-    journal.no_same_sector_overwrites = config["journal_no_same_sector_overwrites"] == "true" ||
-        config["journal_no_same_sector_overwrites"] == "1" || config["journal_no_same_sector_overwrites"] == "yes";
-    journal.inmemory = config["inmemory_journal"] != "false" && config["inmemory_journal"] != "0" &&
-        config["inmemory_journal"] != "no";
+    meta_write_recheck_parallelism = strtoull(config["meta_write_recheck_parallelism"].c_str(), NULL, 10);
     log_level = strtoull(config["log_level"].c_str(), NULL, 10);
     // Validate
-    if (journal.sector_count < 2)
-    {
-        journal.sector_count = 32;
-    }
     if (metadata_buf_size < 65536)
     {
         metadata_buf_size = 4*1024*1024;
     }
-    if (dsk.meta_device == dsk.data_device)
+    if (!meta_write_recheck_parallelism)
     {
-        disable_meta_fsync = disable_data_fsync;
+        meta_write_recheck_parallelism = 16;
     }
-    if (dsk.journal_device == dsk.meta_device)
-    {
-        disable_journal_fsync = disable_meta_fsync;
-    }
-    if (immediate_commit != IMMEDIATE_NONE && !disable_journal_fsync)
+    if (immediate_commit != IMMEDIATE_NONE && !dsk.disable_journal_fsync)
     {
         throw std::runtime_error("immediate_commit requires disable_journal_fsync");
     }
-    if (immediate_commit == IMMEDIATE_ALL && !disable_data_fsync)
+    if (immediate_commit == IMMEDIATE_ALL && !dsk.disable_data_fsync)
     {
         throw std::runtime_error("immediate_commit=all requires disable_journal_fsync and disable_data_fsync");
     }
-    // init some fields
-    journal.block_size = dsk.journal_block_size;
-    journal.next_free = dsk.journal_block_size;
-    journal.used_start = dsk.journal_block_size;
-    // no free space because sector is initially unmapped
-    journal.in_sector_pos = dsk.journal_block_size;
-}
-
-void blockstore_impl_t::calc_lengths()
-{
-    dsk.calc_lengths();
-    journal.len = dsk.journal_len;
-    journal.block_size = dsk.journal_block_size;
-    journal.offset = dsk.journal_offset;
-    if (inmemory_meta)
-    {
-        metadata_buffer = memalign(MEM_ALIGNMENT, dsk.meta_len);
-        if (!metadata_buffer)
-            throw std::runtime_error("Failed to allocate memory for the metadata ("+std::to_string(dsk.meta_len/1024/1024)+" MB)");
-    }
-    else if (dsk.clean_entry_bitmap_size || dsk.data_csum_type)
-    {
-        clean_bitmaps = (uint8_t*)malloc(dsk.block_count * 2 * dsk.clean_entry_bitmap_size);
-        if (!clean_bitmaps)
-        {
-            throw std::runtime_error(
-                "Failed to allocate memory for the metadata sparse write bitmap ("+
-                std::to_string(dsk.block_count * 2 * dsk.clean_entry_bitmap_size / 1024 / 1024)+" MB)"
-            );
-        }
-    }
-    if (journal.inmemory)
-    {
-        journal.buffer = memalign(MEM_ALIGNMENT, journal.len);
-        if (!journal.buffer)
-            throw std::runtime_error("Failed to allocate memory for journal ("+std::to_string(journal.len/1024/1024)+" MB)");
-    }
-    else
-    {
-        journal.sector_buf = (uint8_t*)memalign(MEM_ALIGNMENT, journal.sector_count * dsk.journal_block_size);
-        if (!journal.sector_buf)
-            throw std::bad_alloc();
-    }
-    journal.sector_info = (journal_sector_info_t*)calloc(journal.sector_count, sizeof(journal_sector_info_t));
-    if (!journal.sector_info)
-    {
-        throw std::bad_alloc();
-    }
 }

src/blockstore/blockstore_rollback.cpp

@@ -1,258 +0,0 @@
-// Copyright (c) Vitaliy Filippov, 2019+
-// License: VNPL-1.1 (see README.md for details)
-
-#include "blockstore_impl.h"
-
-int blockstore_impl_t::dequeue_rollback(blockstore_op_t *op)
-{
-    if (PRIV(op)->op_state)
-    {
-        return continue_rollback(op);
-    }
-    int r = split_stab_op(op, [this](obj_ver_id ov)
-    {
-        // Check that there are some versions greater than v->version (which may be zero),
-        // check that they're unstable, synced, and not currently written to
-        auto dirty_it = dirty_db.lower_bound((obj_ver_id){
-            .oid = ov.oid,
-            .version = UINT64_MAX,
-        });
-        if (dirty_it == dirty_db.begin())
-        {
-            // Already rolled back, skip this object version
-            return STAB_SPLIT_DONE;
-        }
-        else
-        {
-            dirty_it--;
-            if (dirty_it->first.oid != ov.oid || dirty_it->first.version < ov.version)
-            {
-                // Already rolled back, skip this object version
-                return STAB_SPLIT_DONE;
-            }
-            while (dirty_it->first.oid == ov.oid && dirty_it->first.version > ov.version)
-            {
-                if (IS_IN_FLIGHT(dirty_it->second.state))
-                {
-                    // Object write is still in progress. Wait until the write request completes
-                    return STAB_SPLIT_WAIT;
-                }
-                else if (!IS_SYNCED(dirty_it->second.state) ||
-                    IS_STABLE(dirty_it->second.state))
-                {
-                    // Sync the object
-                    return STAB_SPLIT_SYNC;
-                }
-                if (dirty_it == dirty_db.begin())
-                {
-                    break;
-                }
-                dirty_it--;
-            }
-            return STAB_SPLIT_TODO;
-        }
-    });
-    if (r != 1)
-    {
-        return r;
-    }
-    // Check journal space
-    blockstore_journal_check_t space_check(this);
-    if (!space_check.check_available(op, op->len, sizeof(journal_entry_rollback), 0))
-    {
-        return 0;
-    }
-    // There is sufficient space. Check SQEs
-    BS_SUBMIT_CHECK_SQES(space_check.sectors_to_write);
-    // Prepare and submit journal entries
-    int s = 0;
-    auto v = (obj_ver_id*)op->buf;
-    for (int i = 0; i < op->len; i++, v++)
-    {
-        if (!journal.entry_fits(sizeof(journal_entry_rollback)) &&
-            journal.sector_info[journal.cur_sector].dirty)
-        {
-            prepare_journal_sector_write(journal.cur_sector, op);
-            s++;
-        }
-        journal_entry_rollback *je = (journal_entry_rollback*)
-            prefill_single_journal_entry(journal, JE_ROLLBACK, sizeof(journal_entry_rollback));
-        je->oid = v->oid;
-        je->version = v->version;
-        je->crc32 = je_crc32((journal_entry*)je);
-        journal.crc32_last = je->crc32;
-    }
-    prepare_journal_sector_write(journal.cur_sector, op);
-    s++;
-    assert(s == space_check.sectors_to_write);
-    PRIV(op)->op_state = 1;
-    return 1;
-}
-
-int blockstore_impl_t::continue_rollback(blockstore_op_t *op)
-{
-    if (PRIV(op)->op_state == 2)
-        goto resume_2;
-    else if (PRIV(op)->op_state == 4)
-        goto resume_4;
-    else
-        return 1;
-resume_2:
-    if (!disable_journal_fsync)
-    {
-        BS_SUBMIT_GET_SQE(sqe, data);
-        my_uring_prep_fsync(sqe, dsk.journal_fd, IORING_FSYNC_DATASYNC);
-        data->iov = { 0 };
-        data->callback = [this, op](ring_data_t *data) { handle_write_event(data, op); };
-        PRIV(op)->min_flushed_journal_sector = PRIV(op)->max_flushed_journal_sector = 0;
-        PRIV(op)->pending_ops = 1;
-        PRIV(op)->op_state = 3;
-        return 1;
-    }
-resume_4:
-    obj_ver_id* v;
-    int i;
-    for (i = 0, v = (obj_ver_id*)op->buf; i < op->len; i++, v++)
-    {
-        mark_rolled_back(*v);
-    }
-    // Acknowledge op
-    op->retval = 0;
-    FINISH_OP(op);
-    return 2;
-}
-
-void blockstore_impl_t::mark_rolled_back(const obj_ver_id & ov)
-{
-    auto it = dirty_db.lower_bound((obj_ver_id){
-        .oid = ov.oid,
-        .version = UINT64_MAX,
-    });
-    if (it != dirty_db.begin())
-    {
-        uint64_t max_unstable = 0;
-        auto rm_start = it;
-        auto rm_end = it;
-        it--;
-        while (1)
-        {
-            if (it->first.oid != ov.oid)
-                break;
-            else if (it->first.version <= ov.version)
-            {
-                if (!IS_STABLE(it->second.state))
-                    max_unstable = it->first.version;
-                break;
-            }
-            else if (IS_IN_FLIGHT(it->second.state) || IS_STABLE(it->second.state))
-                break;
-            // Remove entry
-            rm_start = it;
-            if (it == dirty_db.begin())
-                break;
-            it--;
-        }
-        if (rm_start != rm_end)
-        {
-            erase_dirty(rm_start, rm_end, UINT64_MAX);
-            auto unstab_it = unstable_writes.find(ov.oid);
-            if (unstab_it != unstable_writes.end())
-            {
-                if (max_unstable == 0)
-                    unstable_writes.erase(unstab_it);
-                else
-                    unstab_it->second = max_unstable;
-            }
-        }
-    }
-}
-
-void blockstore_impl_t::erase_dirty(blockstore_dirty_db_t::iterator dirty_start, blockstore_dirty_db_t::iterator dirty_end, uint64_t clean_loc)
-{
-    if (dirty_end == dirty_start)
-    {
-        return;
-    }
-    auto dirty_it = dirty_end;
-    dirty_it--;
-    if (IS_DELETE(dirty_it->second.state))
-    {
-        object_id oid = dirty_it->first.oid;
-#ifdef BLOCKSTORE_DEBUG
-        printf("Unblock writes-after-delete %jx:%jx v%ju\n", oid.inode, oid.stripe, dirty_it->first.version);
-#endif
-        dirty_it = dirty_end;
-        // Unblock operations blocked by delete flushing
-        uint32_t next_state = BS_ST_IN_FLIGHT;
-        while (dirty_it != dirty_db.end() && dirty_it->first.oid == oid)
-        {
-            if ((dirty_it->second.state & BS_ST_WORKFLOW_MASK) == BS_ST_WAIT_DEL)
-            {
-                dirty_it->second.state = (dirty_it->second.state & ~BS_ST_WORKFLOW_MASK) | next_state;
-                if (IS_BIG_WRITE(dirty_it->second.state))
-                {
-                    next_state = BS_ST_WAIT_BIG;
-                }
-            }
-            dirty_it++;
-        }
-        dirty_it = dirty_end;
-        dirty_it--;
-    }
-    while (1)
-    {
-        if ((IS_BIG_WRITE(dirty_it->second.state) || IS_DELETE(dirty_it->second.state)) &&
-            IS_STABLE(dirty_it->second.state))
-        {
-            big_to_flush--;
-        }
-        if (IS_BIG_WRITE(dirty_it->second.state) && dirty_it->second.location != clean_loc &&
-            dirty_it->second.location != UINT64_MAX)
-        {
-#ifdef BLOCKSTORE_DEBUG
-            printf("Free block %ju from %jx:%jx v%ju\n", dirty_it->second.location >> dsk.block_order,
-                dirty_it->first.oid.inode, dirty_it->first.oid.stripe, dirty_it->first.version);
-#endif
-            data_alloc->set(dirty_it->second.location >> dsk.block_order, false);
-        }
-        auto used = --journal.used_sectors.at(dirty_it->second.journal_sector);
-#ifdef BLOCKSTORE_DEBUG
-        printf(
-            "remove usage of journal offset %08jx by %jx:%jx v%ju (%ju refs)\n", dirty_it->second.journal_sector,
-            dirty_it->first.oid.inode, dirty_it->first.oid.stripe, dirty_it->first.version, used
-        );
-#endif
-        if (used == 0)
-        {
-            journal.used_sectors.erase(dirty_it->second.journal_sector);
-            if (dirty_it->second.journal_sector == journal.sector_info[journal.cur_sector].offset)
-            {
-                // Mark current sector as "full" to select the new one
-                journal.in_sector_pos = dsk.journal_block_size;
-            }
-            flusher->mark_trim_possible();
-        }
-        free_dirty_dyn_data(dirty_it->second);
-        if (dirty_it == dirty_start)
-        {
-            break;
-        }
-        dirty_it--;
-    }
-    dirty_db.erase(dirty_start, dirty_end);
-}
-
-void blockstore_impl_t::free_dirty_dyn_data(dirty_entry & e)
-{
-    if (e.dyn_data)
-    {
-        if (alloc_dyn_data &&
-            --*((int*)e.dyn_data) == 0) // refcount
-        {
-            // dyn_data contains the bitmap and checksums
-            // free it if it doesn't refer to the in-memory journal
-            free(e.dyn_data);
-        }
-        e.dyn_data = NULL;
-    }
-}

src/blockstore/blockstore_stable.cpp

@@ -3,559 +3,87 @@
 
 #include "blockstore_impl.h"
 
-// Stabilize small write:
-// 1) Copy data from the journal to the data device
-// 2) Increase version on the metadata device and sync it
-// 3) Advance clean_db entry's version, clear previous journal entries
-//
-// This makes 1 4K small write+sync look like:
-// 512b+4K (journal) + sync + 512b (journal) + sync + 4K (data) [+ sync?] + 512b (metadata) + sync.
-// WA = 2.375. It's not the best, SSD FTL-like redirect-write could probably be lower
-// even with defragmentation. But it's fixed and it's still better than in Ceph. :)
-// except for HDD-only clusters, because each write results in 3 seeks.
-
-// Stabilize big write:
-// 1) Copy metadata from the journal to the metadata device
-// 2) Move dirty_db entry to clean_db and clear previous journal entries
-//
-// This makes 1 128K big write+sync look like:
-// 128K (data) + sync + 512b (journal) + sync + 512b (journal) + sync + 512b (metadata) + sync.
-// WA = 1.012. Very good :)
-
-// Stabilize delete:
-// 1) Remove metadata entry and sync it
-// 2) Remove dirty_db entry and clear previous journal entries
-// We have 2 problems here:
-// - In the cluster environment, we must store the "tombstones" of deleted objects until
-//   all replicas (not just quorum) agrees about their deletion. That is, "stabilize" is
-//   not possible for deletes in degraded placement groups
-// - With simple "fixed" metadata tables we can't just clear the metadata entry of the latest
-//   object version. We must clear all previous entries, too.
-// FIXME Fix both problems - probably, by switching from "fixed" metadata tables to "dynamic"
-
-// AND We must do it in batches, for the sake of reduced fsync call count
-// AND We must know what we stabilize. Basic workflow is like:
-// 1) primary OSD receives sync request
-// 2) it submits syncs to blockstore and peers
-// 3) after everyone acks sync it acks sync to the client
-// 4) after a while it takes his synced object list and sends stabilize requests
-//    to peers and to its own blockstore, thus freeing the old version
-
-struct ver_vector_t
-{
-    obj_ver_id *items = NULL;
-    uint64_t alloc = 0, size = 0;
-};
-
-static void init_versions(ver_vector_t & vec, obj_ver_id *start, obj_ver_id *end, uint64_t len)
-{
-    if (!vec.items)
-    {
-        vec.alloc = len;
-        vec.items = (obj_ver_id*)malloc_or_die(sizeof(obj_ver_id) * vec.alloc);
-        for (auto sv = start; sv < end; sv++)
-        {
-            vec.items[vec.size++] = *sv;
-        }
-    }
-}
-
-static void append_version(ver_vector_t & vec, obj_ver_id ov)
-{
-    if (vec.size >= vec.alloc)
-    {
-        vec.alloc = !vec.alloc ? 4 : vec.alloc*2;
-        vec.items = (obj_ver_id*)realloc_or_die(vec.items, sizeof(obj_ver_id) * vec.alloc);
-    }
-    vec.items[vec.size++] = ov;
-}
-
-static bool check_unsynced(std::vector<obj_ver_id> & check, obj_ver_id ov, std::vector<obj_ver_id> & to, int *count)
-{
-    bool found = false;
-    int j = 0, k = 0;
-    while (j < check.size())
-    {
-        if (check[j] == ov)
-            found = true;
-        if (check[j].oid == ov.oid && check[j].version <= ov.version)
-        {
-            to.push_back(check[j++]);
-            if (count)
-                (*count)--;
-        }
-        else
-            check[k++] = check[j++];
-    }
-    check.resize(k);
-    return found;
-}
-
-blockstore_op_t* blockstore_impl_t::selective_sync(blockstore_op_t *op)
-{
-    unsynced_big_write_count -= unsynced_big_writes.size();
-    unsynced_big_writes.swap(PRIV(op)->sync_big_writes);
-    unsynced_big_write_count += unsynced_big_writes.size();
-    unsynced_small_writes.swap(PRIV(op)->sync_small_writes);
-    // Create a sync operation, insert into the end of the queue
-    // And move ourselves into the end too!
-    // Rather hacky but that's what we need...
-    blockstore_op_t *sync_op = new blockstore_op_t;
-    sync_op->opcode = BS_OP_SYNC;
-    sync_op->buf = NULL;
-    sync_op->callback = [](blockstore_op_t *sync_op)
-    {
-        delete sync_op;
-    };
-    init_op(sync_op);
-    int sync_res = continue_sync(sync_op);
-    if (sync_res != 2)
-    {
-        // Put SYNC into the queue if it's not finished yet
-        submit_queue.push_back(sync_op);
-    }
-    // Restore unsynced_writes
-    unsynced_small_writes.swap(PRIV(op)->sync_small_writes);
-    unsynced_big_write_count -= unsynced_big_writes.size();
-    unsynced_big_writes.swap(PRIV(op)->sync_big_writes);
-    unsynced_big_write_count += unsynced_big_writes.size();
-    if (sync_res == 2)
-    {
-        // Sync is immediately completed
-        return NULL;
-    }
-    return sync_op;
-}
-
-// Returns: 2 = stop processing and dequeue, 0 = stop processing and do not dequeue, 1 = proceed with op itself
-int blockstore_impl_t::split_stab_op(blockstore_op_t *op, std::function<int(obj_ver_id v)> decider)
-{
-    bool add_sync = false;
-    ver_vector_t good_vers, bad_vers;
-    obj_ver_id* v;
-    int i, todo = 0;
-    for (i = 0, v = (obj_ver_id*)op->buf; i < op->len; i++, v++)
-    {
-        int action = decider(*v);
-        if (action < 0)
-        {
-            // Rollback changes
-            for (auto & ov: PRIV(op)->sync_big_writes)
-            {
-                unsynced_big_writes.push_back(ov);
-                unsynced_big_write_count++;
-            }
-            for (auto & ov: PRIV(op)->sync_small_writes)
-            {
-                unsynced_small_writes.push_back(ov);
-            }
-            free(good_vers.items);
-            good_vers.items = NULL;
-            free(bad_vers.items);
-            bad_vers.items = NULL;
-            // Error
-            op->retval = action;
-            FINISH_OP(op);
-            return 2;
-        }
-        else if (action == STAB_SPLIT_DONE)
-        {
-            // Already done
-            init_versions(good_vers, (obj_ver_id*)op->buf, v, op->len);
-        }
-        else if (action == STAB_SPLIT_WAIT)
-        {
-            // Already in progress, we just have to wait until it finishes
-            init_versions(good_vers, (obj_ver_id*)op->buf, v, op->len);
-            append_version(bad_vers, *v);
-        }
-        else if (action == STAB_SPLIT_SYNC)
-        {
-            // Needs a SYNC, we have to send a SYNC if not already in progress
-            //
-            // If the object is not present in unsynced_(big|small)_writes then
-            // it's currently being synced. If it's present then we can initiate
-            // its sync ourselves.
-            init_versions(good_vers, (obj_ver_id*)op->buf, v, op->len);
-            append_version(bad_vers, *v);
-            if (!add_sync)
-            {
-                PRIV(op)->sync_big_writes.clear();
-                PRIV(op)->sync_small_writes.clear();
-                add_sync = true;
-            }
-            check_unsynced(unsynced_small_writes, *v, PRIV(op)->sync_small_writes, NULL);
-            check_unsynced(unsynced_big_writes, *v, PRIV(op)->sync_big_writes, &unsynced_big_write_count);
-        }
-        else /* if (action == STAB_SPLIT_TODO) */
-        {
-            if (good_vers.items)
-            {
-                // If we're selecting versions then append it
-                // Main idea is that 99% of the time all versions passed to BS_OP_STABLE are synced
-                // And we don't want to select/allocate anything in that optimistic case
-                append_version(good_vers, *v);
-            }
-            todo++;
-        }
-    }
-    // In a pessimistic scenario, an operation may be split into 3:
-    // - Stabilize synced entries
-    // - Sync unsynced entries
-    // - Continue for unsynced entries after sync
-    add_sync = add_sync && (PRIV(op)->sync_big_writes.size() || PRIV(op)->sync_small_writes.size());
-    if (!todo && !bad_vers.size)
-    {
-        // Already stable
-        op->retval = 0;
-        FINISH_OP(op);
-        return 2;
-    }
-    op->retval = 0;
-    if (!todo && !add_sync)
-    {
-        // Only wait for inflight writes or current in-progress syncs
-        return 0;
-    }
-    blockstore_op_t *sync_op = NULL, *split_stab_op = NULL;
-    if (add_sync)
-    {
-        // Initiate a selective sync for PRIV(op)->sync_(big|small)_writes
-        sync_op = selective_sync(op);
-    }
-    if (bad_vers.size)
-    {
-        // Split part of the request into a separate operation
-        split_stab_op = new blockstore_op_t;
-        split_stab_op->opcode = op->opcode;
-        split_stab_op->buf = bad_vers.items;
-        split_stab_op->len = bad_vers.size;
-        init_op(split_stab_op);
-        submit_queue.push_back(split_stab_op);
-    }
-    if (sync_op || split_stab_op || good_vers.items)
-    {
-        void *orig_buf = op->buf;
-        if (good_vers.items)
-        {
-            op->buf = good_vers.items;
-            op->len = good_vers.size;
-        }
-        // Make a wrapped callback
-        int *split_op_counter = (int*)malloc_or_die(sizeof(int));
-        *split_op_counter = (sync_op ? 1 : 0) + (split_stab_op ? 1 : 0) + (todo ? 1 : 0);
-        auto cb = [op, good_items = good_vers.items,
-            bad_items = bad_vers.items, split_op_counter,
-            orig_buf, real_cb = op->callback](blockstore_op_t *split_op)
-        {
-            if (split_op->retval != 0)
-                op->retval = split_op->retval;
-            (*split_op_counter)--;
-            assert((*split_op_counter) >= 0);
-            if (op != split_op)
-                delete split_op;
-            if (!*split_op_counter)
-            {
-                free(good_items);
-                free(bad_items);
-                free(split_op_counter);
-                op->buf = orig_buf;
-                real_cb(op);
-            }
-        };
-        if (sync_op)
-        {
-            sync_op->callback = cb;
-        }
-        if (split_stab_op)
-        {
-            split_stab_op->callback = cb;
-        }
-        op->callback = cb;
-    }
-    if (!todo)
-    {
-        // All work is postponed
-        op->callback = NULL;
-        return 2;
-    }
-    return 1;
-}
-
+// Handles both stabilize (commit) and rollback
 int blockstore_impl_t::dequeue_stable(blockstore_op_t *op)
 {
-    if (PRIV(op)->op_state)
-    {
-        return continue_stable(op);
-    }
-    int r = split_stab_op(op, [this](obj_ver_id ov)
-    {
-        auto dirty_it = dirty_db.find(ov);
-        if (dirty_it == dirty_db.end())
-        {
-            auto & clean_db = clean_db_shard(ov.oid);
-            auto clean_it = clean_db.find(ov.oid);
-            if (clean_it == clean_db.end() || clean_it->second.version < ov.version)
-            {
-                // No such object version
-                printf("Error: %jx:%jx v%ju not found while stabilizing\n", ov.oid.inode, ov.oid.stripe, ov.version);
-                return -ENOENT;
-            }
-            else
-            {
-                // Already stable
-                return STAB_SPLIT_DONE;
-            }
-        }
-        else if (IS_STABLE(dirty_it->second.state))
-        {
-            // Already stable
-            return STAB_SPLIT_DONE;
-        }
-        while (true)
-        {
-            if (IS_IN_FLIGHT(dirty_it->second.state))
-            {
-                // Object write is still in progress. Wait until the write request completes
-                return STAB_SPLIT_WAIT;
-            }
-            else if (!IS_SYNCED(dirty_it->second.state))
-            {
-                // Object not synced yet - sync it
-                // In previous versions we returned EBUSY here and required
-                // the caller (OSD) to issue a global sync first. But a global sync
-                // waits for all writes in the queue including inflight writes. And
-                // inflight writes may themselves be blocked by unstable writes being
-                // still present in the journal and not flushed away from it.
-                // So we must sync specific objects here.
-                //
-                // Even more, we have to process "stabilize" request in parts. That is,
-                // we must stabilize all objects which are already synced. Otherwise
-                // they may block objects which are NOT synced yet.
-                return STAB_SPLIT_SYNC;
-            }
-            else if (IS_STABLE(dirty_it->second.state))
-            {
-                break;
-            }
-            // Check previous versions too
-            if (dirty_it == dirty_db.begin())
-            {
-                break;
-            }
-            dirty_it--;
-            if (dirty_it->first.oid != ov.oid)
-            {
-                break;
-            }
-        }
-        return STAB_SPLIT_TODO;
-    });
-    if (r != 1)
-    {
-        return r;
-    }
-    // Check journal space
-    blockstore_journal_check_t space_check(this);
-    if (!space_check.check_available(op, op->len, sizeof(journal_entry_stable), 0))
-    {
-        return 0;
-    }
-    // There is sufficient space. Check SQEs
-    BS_SUBMIT_CHECK_SQES(space_check.sectors_to_write);
-    // Prepare and submit journal entries
-    int s = 0;
-    auto v = (obj_ver_id*)op->buf;
-    for (int i = 0; i < op->len; i++, v++)
-    {
-        if (!journal.entry_fits(sizeof(journal_entry_stable)) &&
-            journal.sector_info[journal.cur_sector].dirty)
-        {
-            prepare_journal_sector_write(journal.cur_sector, op);
-            s++;
-        }
-        journal_entry_stable *je = (journal_entry_stable*)
-            prefill_single_journal_entry(journal, JE_STABLE, sizeof(journal_entry_stable));
-        je->oid = v->oid;
-        je->version = v->version;
-        je->crc32 = je_crc32((journal_entry*)je);
-        journal.crc32_last = je->crc32;
-    }
-    prepare_journal_sector_write(journal.cur_sector, op);
-    s++;
-    assert(s == space_check.sectors_to_write);
-    PRIV(op)->op_state = 1;
-    return 1;
-}
-
-int blockstore_impl_t::continue_stable(blockstore_op_t *op)
-{
-    if (PRIV(op)->op_state == 2)
-        goto resume_2;
-    else if (PRIV(op)->op_state == 4)
-        goto resume_4;
-    else
-        return 1;
-resume_2:
-    if (!disable_journal_fsync)
-    {
-        BS_SUBMIT_GET_SQE(sqe, data);
-        my_uring_prep_fsync(sqe, dsk.journal_fd, IORING_FSYNC_DATASYNC);
-        data->iov = { 0 };
-        data->callback = [this, op](ring_data_t *data) { handle_write_event(data, op); };
-        PRIV(op)->min_flushed_journal_sector = PRIV(op)->max_flushed_journal_sector = 0;
-        PRIV(op)->pending_ops = 1;
-        PRIV(op)->op_state = 3;
-        return 1;
-    }
-resume_4:
-    // Mark dirty_db entries as stable, acknowledge op completion
-    obj_ver_id* v;
-    int i;
-    for (i = 0, v = (obj_ver_id*)op->buf; i < op->len; i++, v++)
-    {
-        // Mark all dirty_db entries up to op->version as stable
-#ifdef BLOCKSTORE_DEBUG
-        printf("Stabilize %jx:%jx v%ju\n", v->oid.inode, v->oid.stripe, v->version);
-#endif
-        mark_stable(*v);
-    }
-    // Acknowledge op
+    obj_ver_id *v = (obj_ver_id*)op->buf;
+    auto priv = PRIV(op);
+    if (priv->op_state == 1)      goto resume_1;
+    else if (priv->op_state == 2) goto resume_2;
+    else if (priv->op_state == 3) goto resume_3;
+    else if (priv->op_state == 4) goto resume_4;
+    assert(!priv->op_state);
+    // Modify in-memory state and assign contiguous LSNs
+    priv->stab_pos = 0;
+    priv->lsn = priv->to_lsn = 0;
     op->retval = 0;
+    while (priv->stab_pos < op->len)
+    {
+        uint32_t modified_block = 0;
+        uint64_t new_lsn = 0;
+        uint64_t new_to_lsn = 0;
+        int res = op->opcode == BS_OP_STABLE
+            ? heap->post_stabilize(v[priv->stab_pos].oid, v[priv->stab_pos].version, &modified_block, &new_lsn, &new_to_lsn)
+            : heap->post_rollback(v[priv->stab_pos].oid, v[priv->stab_pos].version, &new_lsn, &modified_block);
+        if (res != 0)
+        {
+            assert(res == ENOENT || res == EBUSY);
+            op->retval = -res;
+        }
+        if (new_lsn)
+        {
+            assert(priv->lsn == 0 || priv->to_lsn == new_lsn-1);
+            if (!priv->lsn)
+                priv->lsn = new_lsn;
+            priv->to_lsn = op->opcode == BS_OP_STABLE ? new_to_lsn : new_lsn;
+        }
+        priv->stab_pos++;
+    }
+    // Submit metadata writes
+    priv->stab_pos = 0;
+resume_1:
+    priv->op_state = 1;
+    while (priv->stab_pos < op->len)
+    {
+        uint32_t block_num = 0;
+        heap_object_t *obj = heap->read_entry(v[priv->stab_pos].oid, &block_num);
+        if (obj)
+        {
+            io_uring_sqe *sqe = get_sqe();
+            if (!sqe)
+            {
+                if (priv->pending_ops > 0)
+                    return 1;
+                priv->wait_detail = 1;
+                priv->wait_for = WAIT_SQE;
+                return 0;
+            }
+            prepare_meta_block_write(op, block_num, sqe);
+        }
+        priv->stab_pos++;
+    }
+    if (priv->pending_ops > 0)
+    {
+        priv->op_state = 1;
+        return 1;
+    }
+    // Mark writes as completed to allow compaction
+    for (uint64_t lsn = priv->lsn; lsn <= priv->to_lsn; lsn++)
+    {
+        heap->mark_lsn_completed(lsn);
+    }
+    unsynced_meta_write_count++;
+    // Fsync, just because our semantics imply that commit (stabilize) is immediately fsynced
+    priv->op_state = 2;
+resume_2:
+resume_3:
+resume_4:
+    int res = do_sync(op, 2);
+    if (res != 2)
+    {
+        return res;
+    }
+    // Done. Don't touch op->retval - if anything resulted in ENOENT, return it as is
     FINISH_OP(op);
     return 2;
 }
-
-void blockstore_impl_t::mark_stable(obj_ver_id v, bool forget_dirty)
-{
-    auto dirty_it = dirty_db.find(v);
-    if (dirty_it != dirty_db.end())
-    {
-        if (IS_INSTANT(dirty_it->second.state))
-        {
-            // 'Instant' (non-EC) operations may complete and try to become stable out of order. Prevent it.
-            auto back_it = dirty_it;
-            while (back_it != dirty_db.begin())
-            {
-                back_it--;
-                if (back_it->first.oid != v.oid)
-                {
-                    break;
-                }
-                if (!IS_STABLE(back_it->second.state))
-                {
-                    // There are preceding unstable versions, can't flush <v>
-                    return;
-                }
-            }
-            while (true)
-            {
-                dirty_it++;
-                if (dirty_it == dirty_db.end() || dirty_it->first.oid != v.oid ||
-                    !IS_SYNCED(dirty_it->second.state))
-                {
-                    dirty_it--;
-                    break;
-                }
-                v.version = dirty_it->first.version;
-            }
-        }
-        while (1)
-        {
-            bool was_stable = IS_STABLE(dirty_it->second.state);
-            if ((dirty_it->second.state & BS_ST_WORKFLOW_MASK) == BS_ST_SYNCED)
-            {
-                dirty_it->second.state = (dirty_it->second.state & ~BS_ST_WORKFLOW_MASK) | BS_ST_STABLE;
-                // Allocations and deletions are counted when they're stabilized
-                if (IS_BIG_WRITE(dirty_it->second.state))
-                {
-                    int exists = -1;
-                    if (dirty_it != dirty_db.begin())
-                    {
-                        auto prev_it = dirty_it;
-                        prev_it--;
-                        if (prev_it->first.oid == v.oid)
-                        {
-                            exists = IS_DELETE(prev_it->second.state) ? 0 : 1;
-                        }
-                    }
-                    if (exists == -1)
-                    {
-                        auto & clean_db = clean_db_shard(v.oid);
-                        auto clean_it = clean_db.find(v.oid);
-                        exists = clean_it != clean_db.end() ? 1 : 0;
-                    }
-                    if (!exists)
-                    {
-                        uint64_t space_id = dirty_it->first.oid.inode;
-                        if (no_inode_stats[dirty_it->first.oid.inode >> (64-POOL_ID_BITS)])
-                            space_id = space_id & ~(((uint64_t)1 << (64-POOL_ID_BITS)) - 1);
-                        inode_space_stats[space_id] += dsk.data_block_size;
-                        used_blocks++;
-                    }
-                    big_to_flush++;
-                }
-                else if (IS_DELETE(dirty_it->second.state))
-                {
-                    uint64_t space_id = dirty_it->first.oid.inode;
-                    if (no_inode_stats[dirty_it->first.oid.inode >> (64-POOL_ID_BITS)])
-                        space_id = space_id & ~(((uint64_t)1 << (64-POOL_ID_BITS)) - 1);
-                    auto & sp = inode_space_stats[space_id];
-                    if (sp > dsk.data_block_size)
-                        sp -= dsk.data_block_size;
-                    else
-                        inode_space_stats.erase(space_id);
-                    used_blocks--;
-                    big_to_flush++;
-                }
-            }
-            else if (IS_IN_FLIGHT(dirty_it->second.state))
-            {
-                // mark_stable should never be called for in-flight or submitted writes
-                printf(
-                    "BUG: Attempt to mark_stable object %jx:%jx v%ju state of which is %x\n",
-                    dirty_it->first.oid.inode, dirty_it->first.oid.stripe, dirty_it->first.version,
-                    dirty_it->second.state
-                );
-                exit(1);
-            }
-            if (forget_dirty && (IS_BIG_WRITE(dirty_it->second.state) ||
-                IS_DELETE(dirty_it->second.state)))
-            {
-                // Big write overrides all previous dirty entries
-                auto erase_end = dirty_it;
-                while (dirty_it != dirty_db.begin())
-                {
-                    dirty_it--;
-                    if (dirty_it->first.oid != v.oid)
-                    {
-                        dirty_it++;
-                        break;
-                    }
-                }
-                auto & clean_db = clean_db_shard(v.oid);
-                auto clean_it = clean_db.find(v.oid);
-                uint64_t clean_loc = clean_it != clean_db.end()
-                    ? clean_it->second.location : UINT64_MAX;
-                erase_dirty(dirty_it, erase_end, clean_loc);
-                break;
-            }
-            if (was_stable || dirty_it == dirty_db.begin())
-            {
-                break;
-            }
-            dirty_it--;
-            if (dirty_it->first.oid != v.oid)
-            {
-                break;
-            }
-        }
-        flusher->enqueue_flush(v);
-    }
-    auto unstab_it = unstable_writes.find(v.oid);
-    if (unstab_it != unstable_writes.end() &&
-        unstab_it->second <= v.version)
-    {
-        unstable_writes.erase(unstab_it);
-    }
-}

src/blockstore/blockstore_sync.cpp

@@ -3,231 +3,112 @@
 
 #include "blockstore_impl.h"
 
-#define SYNC_HAS_SMALL 1
-#define SYNC_HAS_BIG 2
-#define SYNC_DATA_SYNC_SENT 3
-#define SYNC_DATA_SYNC_DONE 4
-#define SYNC_JOURNAL_WRITE_SENT 5
-#define SYNC_JOURNAL_WRITE_DONE 6
-#define SYNC_JOURNAL_SYNC_SENT 7
-#define SYNC_DONE 8
-
 int blockstore_impl_t::continue_sync(blockstore_op_t *op)
 {
-    if (immediate_commit == IMMEDIATE_ALL)
+    if (!PRIV(op)->op_state)
     {
-        // We can return immediately because sync is only dequeued after all previous writes
         op->retval = 0;
+    }
+    int res = do_sync(op, 0);
+    if (res == 2)
+    {
         FINISH_OP(op);
-        return 2;
     }
-    if (PRIV(op)->op_state == 0)
-    {
-        stop_sync_submitted = false;
-        unsynced_big_write_count -= unsynced_big_writes.size();
-        PRIV(op)->sync_big_writes.swap(unsynced_big_writes);
-        PRIV(op)->sync_small_writes.swap(unsynced_small_writes);
-        unsynced_big_writes.clear();
-        unsynced_small_writes.clear();
-        if (PRIV(op)->sync_big_writes.size() > 0)
-            PRIV(op)->op_state = SYNC_HAS_BIG;
-        else if (PRIV(op)->sync_small_writes.size() > 0)
-            PRIV(op)->op_state = SYNC_HAS_SMALL;
-        else
-            PRIV(op)->op_state = SYNC_DONE;
-    }
-    if (PRIV(op)->op_state == SYNC_HAS_SMALL)
-    {
-        // No big writes, just fsync the journal
-        if (journal.sector_info[journal.cur_sector].dirty)
-        {
-            // Write out the last journal sector if it happens to be dirty
-            BS_SUBMIT_CHECK_SQES(1);
-            prepare_journal_sector_write(journal.cur_sector, op);
-            PRIV(op)->op_state = SYNC_JOURNAL_WRITE_SENT;
-            return 1;
-        }
-        else
-        {
-            PRIV(op)->op_state = SYNC_JOURNAL_WRITE_DONE;
-        }
-    }
-    if (PRIV(op)->op_state == SYNC_HAS_BIG)
-    {
-        // 1st step: fsync data
-        if (!disable_data_fsync)
-        {
-            BS_SUBMIT_GET_SQE(sqe, data);
-            my_uring_prep_fsync(sqe, dsk.data_fd, IORING_FSYNC_DATASYNC);
-            data->iov = { 0 };
-            data->callback = [this, op](ring_data_t *data) { handle_write_event(data, op); };
-            PRIV(op)->min_flushed_journal_sector = PRIV(op)->max_flushed_journal_sector = 0;
-            PRIV(op)->pending_ops = 1;
-            PRIV(op)->op_state = SYNC_DATA_SYNC_SENT;
-            return 1;
-        }
-        else
-        {
-            PRIV(op)->op_state = SYNC_DATA_SYNC_DONE;
-        }
-    }
-    if (PRIV(op)->op_state == SYNC_DATA_SYNC_DONE)
-    {
-        // 2nd step: Data device is synced, prepare & write journal entries
-        // Check space in the journal and journal memory buffers
-        blockstore_journal_check_t space_check(this);
-        if (dsk.csum_block_size)
-        {
-            // More complex check because all journal entries have different lengths
-            int left = PRIV(op)->sync_big_writes.size();
-            for (auto & sbw: PRIV(op)->sync_big_writes)
-            {
-                left--;
-                auto & dirty_entry = dirty_db.at(sbw);
-                uint64_t dyn_size = dsk.dirty_dyn_size(dirty_entry.offset, dirty_entry.len);
-                if (!space_check.check_available(op, 1, sizeof(journal_entry_big_write) + dyn_size, 0))
-                {
-                    return 0;
-                }
-            }
-        }
-        else if (!space_check.check_available(op, PRIV(op)->sync_big_writes.size(),
-            sizeof(journal_entry_big_write) + dsk.clean_entry_bitmap_size, 0))
-        {
-            return 0;
-        }
-        // Check SQEs. Don't bother about merging, submit each journal sector as a separate request
-        BS_SUBMIT_CHECK_SQES(space_check.sectors_to_write);
-        // Prepare and submit journal entries
-        auto it = PRIV(op)->sync_big_writes.begin();
-        int s = 0;
-        while (it != PRIV(op)->sync_big_writes.end())
-        {
-            auto & dirty_entry = dirty_db.at(*it);
-            uint64_t dyn_size = dsk.dirty_dyn_size(dirty_entry.offset, dirty_entry.len);
-            if (!journal.entry_fits(sizeof(journal_entry_big_write) + dyn_size) &&
-                journal.sector_info[journal.cur_sector].dirty)
-            {
-                prepare_journal_sector_write(journal.cur_sector, op);
-                s++;
-            }
-            journal_entry_big_write *je = (journal_entry_big_write*)prefill_single_journal_entry(
-                journal, (dirty_entry.state & BS_ST_INSTANT) ? JE_BIG_WRITE_INSTANT : JE_BIG_WRITE,
-                sizeof(journal_entry_big_write) + dyn_size
-            );
-            auto jsec = dirty_entry.journal_sector = journal.sector_info[journal.cur_sector].offset;
-            assert(journal.next_free >= journal.used_start
-                ? (jsec >= journal.used_start && jsec < journal.next_free)
-                : (jsec >= journal.used_start || jsec < journal.next_free));
-            journal.used_sectors[journal.sector_info[journal.cur_sector].offset]++;
-#ifdef BLOCKSTORE_DEBUG
-            printf(
-                "journal offset %08jx is used by %jx:%jx v%ju (%ju refs)\n",
-                dirty_entry.journal_sector, it->oid.inode, it->oid.stripe, it->version,
-                journal.used_sectors[journal.sector_info[journal.cur_sector].offset]
-            );
-#endif
-            je->oid = it->oid;
-            je->version = it->version;
-            je->offset = dirty_entry.offset;
-            je->len = dirty_entry.len;
-            je->location = dirty_entry.location;
-            memcpy((void*)(je+1), (alloc_dyn_data
-                ? (uint8_t*)dirty_entry.dyn_data+sizeof(int) : (uint8_t*)&dirty_entry.dyn_data), dyn_size);
-            je->crc32 = je_crc32((journal_entry*)je);
-            journal.crc32_last = je->crc32;
-            it++;
-        }
-        prepare_journal_sector_write(journal.cur_sector, op);
-        s++;
-        assert(s == space_check.sectors_to_write);
-        PRIV(op)->op_state = SYNC_JOURNAL_WRITE_SENT;
-        return 1;
-    }
-    if (PRIV(op)->op_state == SYNC_JOURNAL_WRITE_DONE)
-    {
-        if (!disable_journal_fsync)
-        {
-            BS_SUBMIT_GET_SQE(sqe, data);
-            my_uring_prep_fsync(sqe, dsk.journal_fd, IORING_FSYNC_DATASYNC);
-            data->iov = { 0 };
-            data->callback = [this, op](ring_data_t *data) { handle_write_event(data, op); };
-            PRIV(op)->min_flushed_journal_sector = PRIV(op)->max_flushed_journal_sector = 0;
-            PRIV(op)->pending_ops = 1;
-            PRIV(op)->op_state = SYNC_JOURNAL_SYNC_SENT;
-            return 1;
-        }
-        else
-        {
-            PRIV(op)->op_state = SYNC_DONE;
-        }
-    }
-    if (PRIV(op)->op_state == SYNC_DONE)
-    {
-        ack_sync(op);
-        return 2;
-    }
-    return 1;
+    return res;
 }
 
-void blockstore_impl_t::ack_sync(blockstore_op_t *op)
+bool blockstore_impl_t::submit_fsyncs(int & wait_count)
 {
-    // Handle states
-    for (auto it = PRIV(op)->sync_big_writes.begin(); it != PRIV(op)->sync_big_writes.end(); it++)
+    int n = ((unsynced_small_write_count > 0 || unsynced_big_write_count > 0 || unsynced_meta_write_count > 0) && !dsk.disable_meta_fsync) +
+        (unsynced_small_write_count > 0 && !dsk.disable_journal_fsync && dsk.journal_fd != dsk.meta_fd) +
+        (unsynced_big_write_count > 0 && !dsk.disable_data_fsync && dsk.data_fd != dsk.meta_fd && dsk.data_fd != dsk.journal_fd);
+    if (ringloop->space_left() < n)
     {
-#ifdef BLOCKSTORE_DEBUG
-        printf("Ack sync big %jx:%jx v%ju\n", it->oid.inode, it->oid.stripe, it->version);
-#endif
-        auto & unstab = unstable_writes[it->oid];
-        unstab = unstab < it->version ? it->version : unstab;
-        auto dirty_it = dirty_db.find(*it);
-        dirty_it->second.state = ((dirty_it->second.state & ~BS_ST_WORKFLOW_MASK) | BS_ST_SYNCED);
-        if (dirty_it->second.state & BS_ST_INSTANT)
-        {
-            mark_stable(dirty_it->first);
-        }
-        else
-        {
-            unstable_unsynced--;
-            assert(unstable_unsynced >= 0);
-        }
-        dirty_it++;
-        while (dirty_it != dirty_db.end() && dirty_it->first.oid == it->oid)
-        {
-            if ((dirty_it->second.state & BS_ST_WORKFLOW_MASK) == BS_ST_WAIT_BIG)
-            {
-                dirty_it->second.state = (dirty_it->second.state & ~BS_ST_WORKFLOW_MASK) | BS_ST_IN_FLIGHT;
-            }
-            dirty_it++;
-        }
+        return false;
     }
-    for (auto it = PRIV(op)->sync_small_writes.begin(); it != PRIV(op)->sync_small_writes.end(); it++)
+    if (!n)
     {
-#ifdef BLOCKSTORE_DEBUG
-        printf("Ack sync small %jx:%jx v%ju\n", it->oid.inode, it->oid.stripe, it->version);
-#endif
-        auto & unstab = unstable_writes[it->oid];
-        unstab = unstab < it->version ? it->version : unstab;
-        if (dirty_db[*it].state == (BS_ST_DELETE | BS_ST_WRITTEN))
-        {
-            dirty_db[*it].state = (BS_ST_DELETE | BS_ST_SYNCED);
-            // Deletions are treated as immediately stable
-            mark_stable(*it);
-        }
-        else /* (BS_ST_INSTANT?) | BS_ST_SMALL_WRITE | BS_ST_WRITTEN */
-        {
-            dirty_db[*it].state = (dirty_db[*it].state & ~BS_ST_WORKFLOW_MASK) | BS_ST_SYNCED;
-            if (dirty_db[*it].state & BS_ST_INSTANT)
-            {
-                mark_stable(*it);
-            }
-            else
-            {
-                unstable_unsynced--;
-                assert(unstable_unsynced >= 0);
-            }
-        }
+        return true;
     }
-    op->retval = 0;
-    FINISH_OP(op);
+    auto cb = [this, & wait_count](ring_data_t *data)
+    {
+        if (data->res != 0)
+            disk_error_abort("sync meta", data->res, 0);
+        wait_count--;
+        assert(wait_count >= 0);
+        if (!wait_count)
+            ringloop->wakeup();
+    };
+    if ((unsynced_small_write_count > 0 || unsynced_big_write_count > 0 || unsynced_meta_write_count > 0) && !dsk.disable_meta_fsync)
+    {
+        // fsync meta
+        io_uring_sqe *sqe = get_sqe();
+        assert(sqe);
+        ring_data_t *data = ((ring_data_t*)sqe->user_data);
+        io_uring_prep_fsync(sqe, dsk.meta_fd, IORING_FSYNC_DATASYNC);
+        data->iov = { 0 };
+        data->callback = cb;
+        wait_count++;
+    }
+    if (unsynced_small_write_count > 0 && !dsk.disable_journal_fsync && dsk.meta_fd != dsk.journal_fd)
+    {
+        // fsync buffer
+        io_uring_sqe *sqe = get_sqe();
+        assert(sqe);
+        ring_data_t *data = ((ring_data_t*)sqe->user_data);
+        io_uring_prep_fsync(sqe, dsk.journal_fd, IORING_FSYNC_DATASYNC);
+        data->iov = { 0 };
+        data->callback = cb;
+        wait_count++;
+    }
+    if (unsynced_big_write_count > 0 && !dsk.disable_data_fsync && dsk.data_fd != dsk.meta_fd && dsk.data_fd != dsk.journal_fd)
+    {
+        // fsync data
+        io_uring_sqe *sqe = get_sqe();
+        assert(sqe);
+        ring_data_t *data = ((ring_data_t*)sqe->user_data);
+        io_uring_prep_fsync(sqe, dsk.data_fd, IORING_FSYNC_DATASYNC);
+        data->iov = { 0 };
+        data->callback = cb;
+        wait_count++;
+    }
+    unsynced_big_write_count = 0;
+    unsynced_small_write_count = 0;
+    unsynced_meta_write_count = 0;
+    return true;
+}
+
+int blockstore_impl_t::do_sync(blockstore_op_t *op, int base_state)
+{
+    int op_state = PRIV(op)->op_state - base_state;
+    if (op_state == 1) goto resume_1;
+    if (op_state == 2) goto resume_2;
+    assert(!op_state);
+    if (flusher->get_syncing_buffer())
+    {
+        // Wait for flusher-initiated sync
+        return 0;
+    }
+    if (dsk.disable_journal_fsync && dsk.disable_meta_fsync && dsk.disable_data_fsync || !unsynced_big_write_count && !unsynced_small_write_count)
+    {
+        // We can return immediately because sync only syncs previous writes
+        unsynced_big_write_count = unsynced_small_write_count = unsynced_meta_write_count = 0;
+        return 2;
+    }
+    PRIV(op)->lsn = heap->get_completed_lsn();
+    if (!submit_fsyncs(PRIV(op)->pending_ops))
+    {
+        PRIV(op)->wait_detail = 1;
+        PRIV(op)->wait_for = WAIT_SQE;
+        return 0;
+    }
+resume_1:
+    if (PRIV(op)->pending_ops > 0)
+    {
+        PRIV(op)->op_state = base_state+1;
+        return 1;
+    }
+resume_2:
+    heap->mark_lsn_fsynced(PRIV(op)->lsn);
+    return 2;
 }

src/blockstore/fio_engine.cpp

@@ -5,14 +5,14 @@
 //
 // Initialize storage for tests:
 //
-// dd if=/dev/zero of=test_data.bin bs=1024 count=1048576
-// dd if=/dev/zero of=test_meta.bin bs=1024 count=256
-// dd if=/dev/zero of=test_journal.bin bs=1024 count=4096
+// dd if=/dev/zero of=test_data.bin bs=1M count=1024
 //
 // Random write:
 //
-// fio -thread -ioengine=./libfio_blockstore.so -name=test -bs=4k -direct=1 -fsync=16 -iodepth=16 -rw=randwrite \
-//     -bs_config='{"data_device":"./test_data.bin"}' -size=1000M
+// [LD_PRELOAD=libasan.so.8] \
+// fio -name=test -thread -ioengine=../build/src/blockstore/libfio_vitastor_blk.so \
+//     -bs=4k -direct=1 -rw=randwrite -iodepth=16 -size=900M -loops=10 \
+//     -bs_config='{"data_device":"./test_data.bin","meta_offset":0,"journal_offset":16777216,"data_offset":33554432,"disable_data_fsync":true,"meta_format":3,"immediate_commit":"all","log_level":100,"journal_no_same_sector_overwrites":true,"journal_sector_buffer_count":1024}'
 //
 // Linear write:
 //
@@ -38,12 +38,14 @@ struct bs_data
     std::vector<io_u*> completed;
     int op_n = 0, inflight = 0;
     bool last_sync = false;
+    bool trace = false;
 };
 
 struct bs_options
 {
     int __pad;
     char *json_config = NULL;
+    int trace = 0;
 };
 
 static struct fio_option options[] = {
@@ -56,6 +58,16 @@ static struct fio_option options[] = {
         .category = FIO_OPT_C_ENGINE,
         .group  = FIO_OPT_G_FILENAME,
     },
+    {
+        .name   = "bs_trace",
+        .lname  = "trace",
+        .type   = FIO_OPT_BOOL,
+        .off1   = offsetof(struct bs_options, trace),
+        .help   = "Trace operations",
+        .def    = "0",
+        .category = FIO_OPT_C_ENGINE,
+        .group  = FIO_OPT_G_FILENAME,
+    },
     {
         .name = NULL,
     },
@@ -63,6 +75,7 @@ static struct fio_option options[] = {
 
 static int bs_setup(struct thread_data *td)
 {
+    bs_options *o = (bs_options*)td->eo;
     bs_data *bsd;
     //fio_file *f;
     //int r;
@@ -83,6 +96,8 @@ static int bs_setup(struct thread_data *td)
         td->o.open_files++;
     }
 
+    bsd->trace = o->trace ? true : false;
+
     //f = td->files[0];
     //f->real_file_size = size;
     return 0;
@@ -168,7 +183,7 @@ static enum fio_q_status bs_queue(struct thread_data *td, struct io_u *io)
     {
     case DDIR_READ:
         op->opcode = BS_OP_READ;
-        op->buf = io->xfer_buf;
+        op->buf = (uint8_t*)io->xfer_buf;
         op->oid = {
             .inode = 1,
             .stripe = io->offset / bsd->bs->get_block_size(),
@@ -176,21 +191,20 @@ static enum fio_q_status bs_queue(struct thread_data *td, struct io_u *io)
         op->version = UINT64_MAX; // last unstable
         op->offset = io->offset % bsd->bs->get_block_size();
         op->len = io->xfer_buflen;
-        op->callback = [io](blockstore_op_t *op)
+        op->callback = [io, n = bsd->op_n](blockstore_op_t *op)
         {
             io->error = op->retval < 0 ? -op->retval : 0;
             bs_data *bsd = (bs_data*)io->engine_data;
             bsd->inflight--;
             bsd->completed.push_back(io);
-#ifdef BLOCKSTORE_DEBUG
-            printf("--- OP_READ %llx n=%d retval=%d\n", io, n, op->retval);
-#endif
+            if (bsd->trace)
+                printf("--- OP_READ %zx n=%d retval=%d\n", (size_t)op, n, op->retval);
             delete op;
         };
         break;
     case DDIR_WRITE:
-        op->opcode = BS_OP_WRITE;
-        op->buf = io->xfer_buf;
+        op->opcode = BS_OP_WRITE_STABLE;
+        op->buf = (uint8_t*)io->xfer_buf;
         op->oid = {
             .inode = 1,
             .stripe = io->offset / bsd->bs->get_block_size(),
@@ -198,30 +212,28 @@ static enum fio_q_status bs_queue(struct thread_data *td, struct io_u *io)
         op->version = 0; // assign automatically
         op->offset = io->offset % bsd->bs->get_block_size();
         op->len = io->xfer_buflen;
-        op->callback = [io](blockstore_op_t *op)
+        op->callback = [io, n = bsd->op_n](blockstore_op_t *op)
         {
             io->error = op->retval < 0 ? -op->retval : 0;
             bs_data *bsd = (bs_data*)io->engine_data;
             bsd->inflight--;
             bsd->completed.push_back(io);
-#ifdef BLOCKSTORE_DEBUG
-            printf("--- OP_WRITE %llx n=%d retval=%d\n", io, n, op->retval);
-#endif
+            if (bsd->trace)
+                printf("--- OP_WRITE %zx n=%d retval=%d\n", (size_t)op, n, op->retval);
             delete op;
         };
         bsd->last_sync = false;
         break;
     case DDIR_SYNC:
-        op->opcode = BS_OP_SYNC_STAB_ALL;
-        op->callback = [io](blockstore_op_t *op)
+        op->opcode = BS_OP_SYNC;
+        op->callback = [io, n = bsd->op_n](blockstore_op_t *op)
         {
             bs_data *bsd = (bs_data*)io->engine_data;
             io->error = op->retval < 0 ? -op->retval : 0;
             bsd->completed.push_back(io);
             bsd->inflight--;
-#ifdef BLOCKSTORE_DEBUG
-            printf("--- OP_SYNC %llx n=%d retval=%d\n", io, n, op->retval);
-#endif
+            if (bsd->trace)
+                printf("--- OP_SYNC %zx n=%d retval=%d\n", (size_t)op, n, op->retval);
             delete op;
         };
         bsd->last_sync = true;
@@ -232,9 +244,8 @@ static enum fio_q_status bs_queue(struct thread_data *td, struct io_u *io)
         return FIO_Q_COMPLETED;
     }
 
-#ifdef BLOCKSTORE_DEBUG
-    printf("+++ %s %llx n=%d\n", op->opcode == OP_READ ? "OP_READ" : (op->opcode == OP_WRITE ? "OP_WRITE" : "OP_SYNC"), io, n);
-#endif
+    if (bsd->trace)
+        printf("+++ %s %zx n=%d\n", op->opcode == BS_OP_READ ? "OP_READ" : (op->opcode == BS_OP_WRITE_STABLE ? "OP_WRITE" : "OP_SYNC"), (size_t)op, bsd->op_n);
     io->error = 0;
     bsd->inflight++;
     bsd->bs->enqueue_op(op);
@@ -290,7 +301,7 @@ static int bs_invalidate(struct thread_data *td, struct fio_file *f)
     return 0;
 }
 
-struct ioengine_ops ioengine = {
+struct ioengine_ops __attribute__((visibility("default"))) ioengine = {
     .name               = "vitastor_blockstore",
     .version            = FIO_IOOPS_VERSION,
     .flags              = FIO_MEMALIGN | FIO_DISKLESSIO | FIO_NOEXTEND,

src/blockstore/meta_v1.h

@@ -1,12 +1,11 @@
+// Old metadata format on-disk structures
 // Copyright (c) Vitaliy Filippov, 2019+
 // License: VNPL-1.1 (see README.md for details)
 
 #pragma once
 
 #include "crc32c.h"
-#include <set>
 
-#define MIN_JOURNAL_SIZE 4*1024*1024
 #define JOURNAL_MAGIC 0x4A33
 #define JOURNAL_VERSION_V1 1
 #define JOURNAL_VERSION_V2 2
@@ -145,77 +144,14 @@ inline uint32_t je_crc32(journal_entry *je)
     return crc32c(0x48674bc7, ((uint8_t*)je)+4, je->size-4);
 }
 
-struct journal_sector_info_t
+// 32 bytes = 24 bytes + block bitmap (4 bytes by default) + external attributes (also bitmap, 4 bytes by default)
+// per "clean" entry on disk with fixed metadata tables
+struct __attribute__((__packed__)) clean_disk_entry
 {
-    uint64_t offset;
-    uint64_t flush_count;
-    bool written;
-    bool dirty;
-    uint64_t submit_id;
+    object_id oid;
+    uint64_t version;
+    uint8_t bitmap[];
+    // Two more fields come after bitmap in metadata version 2:
+    // uint32_t data_csum[];
+    // uint32_t entry_csum;
 };
-
-struct pending_journaling_t
-{
-    int pending;
-    int sector;
-    blockstore_op_t *op;
-};
-
-struct journal_t
-{
-    int fd;
-    bool inmemory = false;
-    bool flush_journal = false;
-    void *buffer = NULL;
-
-    uint64_t block_size;
-    uint64_t offset, len;
-    // Next free block offset
-    uint64_t next_free = 0;
-    // First occupied block offset
-    uint64_t used_start = 0;
-    // End of the last block not used for writing anymore
-    uint64_t dirty_start = 0;
-    uint32_t crc32_last = 0;
-
-    // Current sector(s) used for writing
-    void *sector_buf = NULL;
-    journal_sector_info_t *sector_info = NULL;
-    uint64_t sector_count;
-    bool no_same_sector_overwrites = false;
-    int cur_sector = 0;
-    int in_sector_pos = 0;
-    std::vector<int> submitting_sectors;
-    std::multimap<uint64_t, pending_journaling_t> flushing_ops;
-    uint64_t submit_id = 0;
-
-    // Used sector map
-    // May use ~ 80 MB per 1 GB of used journal space in the worst case
-    std::map<uint64_t, uint64_t> used_sectors;
-
-    ~journal_t();
-    bool trim();
-    uint64_t get_trim_pos();
-    void dump_diagnostics();
-    inline bool entry_fits(int size)
-    {
-        return !(block_size - in_sector_pos < size ||
-            no_same_sector_overwrites && sector_info[cur_sector].written);
-    }
-};
-
-struct blockstore_journal_check_t
-{
-    blockstore_impl_t *bs;
-    uint64_t next_pos, next_sector, next_in_pos;
-    int sectors_to_write, first_sector;
-    bool right_dir; // writing to the end or the beginning of the ring buffer
-
-    blockstore_journal_check_t(blockstore_impl_t *bs);
-    int check_available(blockstore_op_t *op, int required, int size, int data_after);
-};
-
-journal_entry* prefill_single_journal_entry(journal_t & journal, uint16_t type, uint32_t size);
-
-uint32_t crc32c_pad(uint32_t prev_crc, const void *buf, size_t len, size_t left_pad, size_t right_pad);
-uint32_t crc32c_nopad(uint32_t prev_crc, const void *buf, size_t len, size_t left_pad, size_t right_pad);

src/blockstore/multilist.cpp

@@ -0,0 +1,338 @@
+// Variable-length O(1) disk space allocator
+// Copyright (c) Vitaliy Filippov, 2025+
+// License: VNPL-1.1 (see README.md for details)
+
+#include <assert.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <set>
+#include "multilist.h"
+
+multilist_alloc_t::multilist_alloc_t(uint32_t count, uint32_t maxn):
+    count(count), maxn(maxn)
+{
+    // not-so-memory-efficient: 16 MB memory per 1 GB buffer space, but buffer spaces are small, so OK
+    assert(count > 1 && count < 0x80000000);
+    sizes.resize(count);
+    nexts.resize(count); // nexts[i] = 0 -> area is used; nexts[i] = 1 -> no next; nexts[i] >= 2 -> next item
+    prevs.resize(count);
+    heads.resize(maxn); // heads[i] = 0 -> empty list; heads[i] >= 1 -> list head
+    sizes[0] = count;
+    sizes[count-1] = -count; // end
+    nexts[0] = 1;
+    heads[maxn-1] = 1;
+#ifdef MULTILIST_TRACE
+    print();
+#endif
+}
+
+bool multilist_alloc_t::is_free(uint32_t pos)
+{
+    assert(pos < count);
+    if (sizes[pos] < 0)
+        pos += sizes[pos]+1;
+    while (pos > 0 && !sizes[pos])
+        pos--;
+    return nexts[pos] > 0;
+}
+
+uint32_t multilist_alloc_t::find(uint32_t size)
+{
+    assert(size > 0);
+    assert(size <= maxn);
+    for (uint32_t i = size-1; i < maxn; i++)
+    {
+        if (heads[i])
+        {
+            return heads[i]-1;
+        }
+    }
+    return UINT32_MAX;
+}
+
+void multilist_alloc_t::verify()
+{
+    std::set<uint32_t> reachable;
+    for (int i = 0; i < maxn; i++)
+    {
+        uint32_t cur = heads[i];
+        while (cur)
+        {
+            if (!nexts[cur-1])
+            {
+                fprintf(stderr, "ERROR: item %d from freelist %d is not free\n", cur-1, i);
+                print();
+                abort();
+            }
+            if (nexts[cur-1] >= count+2)
+            {
+                fprintf(stderr, "ERROR: next out of range at %d: %d\n", cur-1, nexts[cur-1]);
+                print();
+                abort();
+            }
+            if (!(i < maxn-1 ? sizes[cur-1] == i+1 : (sizes[cur-1] >= i+1)))
+            {
+                fprintf(stderr, "ERROR: item %d is in wrong freelist: expected size %d, but actual size is %d\n", cur-1, i+1, sizes[cur-1]);
+                print();
+                abort();
+            }
+            if (reachable.find(cur-1) != reachable.end())
+            {
+                fprintf(stderr, "ERROR: doubly-claimed item %d\n", cur-1);
+                print();
+                abort();
+            }
+            reachable.insert(cur-1);
+            cur = nexts[cur-1]-1;
+        }
+    }
+    for (int i = 0; i < count; )
+    {
+        if (sizes[i])
+        {
+            assert(i+sizes[i] <= count);
+            if (sizes[i] > 1 && sizes[i+sizes[i]-1] != -sizes[i])
+            {
+                fprintf(stderr, "ERROR: start/end mismatch at %d: sizes[%d] should be %d, but is %d\n", i, i+sizes[i]-1, -sizes[i], sizes[i+sizes[i]-1]);
+                print();
+                abort();
+            }
+            for (int j = i+1; j < i+sizes[i]-1; j++)
+            {
+                if (sizes[j])
+                {
+                    fprintf(stderr, "ERROR: internal non-zero at %d: %d\n", j, sizes[j]);
+                    print();
+                    abort();
+                }
+            }
+            if (nexts[i] && reachable.find(i) == reachable.end())
+            {
+                fprintf(stderr, "ERROR: %d is unreachable from heads\n", i);
+                print();
+                abort();
+            }
+            if (nexts[i] >= 2)
+            {
+                if (nexts[i] >= 2+count)
+                {
+                    fprintf(stderr, "ERROR: next out of range at %d: %d\n", i, nexts[i]);
+                    print();
+                    abort();
+                }
+                if (prevs[nexts[i]-2] != i+1)
+                {
+                    fprintf(stderr, "ERROR: prev[next] (%d) != this (%d) at %d", prevs[nexts[i]-2], i+1, i);
+                    print();
+                    abort();
+                }
+            }
+            i += (sizes[i] > 1 ? sizes[i] : 1);
+        }
+        else
+            i++;
+    }
+}
+
+void multilist_alloc_t::print()
+{
+    printf("heads:");
+    for (int i = 0; i < maxn; i++)
+        if (heads[i])
+            printf(" %u=%u", i, heads[i]);
+    printf("\n");
+    printf("sizes:");
+    for (int i = 0; i < count; i++)
+        if (sizes[i])
+            printf(" %d=%d", i, sizes[i]);
+    printf("\n");
+    printf("prevs:");
+    for (int i = 0; i < count; i++)
+        if (prevs[i])
+            printf(" %d=%d", i, prevs[i]);
+    printf("\n");
+    printf("nexts:");
+    for (int i = 0; i < count; i++)
+        if (nexts[i])
+            printf(" %d=%d", i, nexts[i]);
+    printf("\n");
+    printf("items:");
+    for (int i = 0; i < count; )
+    {
+        if (sizes[i])
+        {
+            printf(" %u=(s:%d,n:%u,p:%u)", i, sizes[i], nexts[i], prevs[i]);
+            assert(i+sizes[i] <= count);
+            i += (sizes[i] > 1 ? sizes[i] : 1);
+        }
+        else
+            i++;
+    }
+    printf("\n");
+}
+
+void multilist_alloc_t::use(uint32_t pos, uint32_t size)
+{
+    assert(pos+size <= count && size > 0);
+    if (sizes[pos] <= 0)
+    {
+        uint32_t start = pos;
+        if (sizes[start] < 0)
+            start += sizes[start]+1;
+        else
+            while (start > 0 && !sizes[start])
+                start--;
+        assert(sizes[start] >= size);
+        use_full(start);
+        uint32_t full = sizes[start];
+        sizes[pos-1] = -pos+start;
+        sizes[start] = pos-start;
+        free(start);
+        sizes[pos+size-1] = -size;
+        sizes[pos] = size;
+        if (pos+size < start+full)
+        {
+            sizes[start+full-1] = -(start+full-pos-size);
+            sizes[pos+size] = start+full-pos-size;
+            free(pos+size);
+        }
+    }
+    else
+    {
+        assert(sizes[pos] >= size);
+        use_full(pos);
+        if (sizes[pos] > size)
+        {
+            uint32_t full = sizes[pos];
+            sizes[pos+size-1] = -size;
+            sizes[pos] = size;
+            sizes[pos+full-1] = -full+size;
+            sizes[pos+size] = full-size;
+            free(pos+size);
+        }
+    }
+#ifdef MULTILIST_TRACE
+    print();
+#endif
+}
+
+void multilist_alloc_t::use_full(uint32_t pos)
+{
+    uint32_t prevsize = sizes[pos];
+    assert(prevsize);
+    assert(nexts[pos]);
+    uint32_t pi = (prevsize < maxn ? prevsize : maxn)-1;
+    if (heads[pi] == pos+1)
+        heads[pi] = nexts[pos]-1;
+    if (prevs[pos])
+        nexts[prevs[pos]-1] = nexts[pos];
+    if (nexts[pos] >= 2)
+        prevs[nexts[pos]-2] = prevs[pos];
+    prevs[pos] = 0;
+    nexts[pos] = 0;
+}
+
+void multilist_alloc_t::free(uint32_t pos)
+{
+    do_free(pos);
+#ifdef MULTILIST_TRACE
+    print();
+#endif
+}
+
+void multilist_alloc_t::do_free(uint32_t pos)
+{
+    assert(!nexts[pos]);
+    uint32_t size = sizes[pos];
+    assert(size > 0);
+    // merge with previous?
+    if (pos > 0 && nexts[pos+(sizes[pos-1] == 1 ? -1 : sizes[pos-1])] > 0)
+    {
+        assert(sizes[pos-1] < 0 || sizes[pos-1] == 1);
+        uint32_t prevsize = sizes[pos-1] < 0 ? -sizes[pos-1] : 1;
+        use_full(pos-prevsize);
+        sizes[pos] = 0;
+        sizes[pos-1] = 0;
+        size += prevsize;
+        pos -= prevsize;
+        sizes[pos+size-1] = -size;
+        sizes[pos] = size;
+    }
+    // merge with next?
+    if (pos+size < count && nexts[pos+size] >= 1)
+    {
+        uint32_t nextsize = sizes[pos+size];
+        use_full(pos+size);
+        sizes[pos+size] = 0;
+        sizes[pos+size-1] = 0;
+        size += nextsize;
+        sizes[pos+size-1] = -size;
+        sizes[pos] = size;
+    }
+    uint32_t ni = (size < maxn ? size : maxn)-1;
+    nexts[pos] = heads[ni]+1;
+    prevs[pos] = 0;
+    if (heads[ni])
+        prevs[heads[ni]-1] = pos+1;
+    heads[ni] = pos+1;
+}
+
+multilist_index_t::multilist_index_t(uint32_t count, uint32_t max_used, uint32_t init_used):
+    count(count), max_used(max_used)
+{
+    assert(init_used < max_used);
+    nexts.resize(count, UINT32_MAX);
+    prevs.resize(count, UINT32_MAX);
+    heads.resize(max_used, UINT32_MAX);
+    for (size_t i = 0; i < count-1; i++)
+    {
+        nexts[i] = i+1;
+        prevs[i+1] = i;
+    }
+    prevs[0] = UINT32_MAX;
+    nexts[count-1] = UINT32_MAX;
+    heads[init_used] = 0;
+}
+
+uint32_t multilist_index_t::find(uint32_t wanted_used)
+{
+    assert(wanted_used < max_used);
+    return heads[wanted_used];
+}
+
+void multilist_index_t::change(uint32_t pos, uint32_t old_used, uint32_t new_used)
+{
+    if (new_used == old_used)
+        return;
+    assert(old_used < max_used && new_used < max_used);
+    if (prevs[pos] != UINT32_MAX)
+        nexts[prevs[pos]] = nexts[pos];
+    if (nexts[pos] != UINT32_MAX)
+        prevs[nexts[pos]] = prevs[pos];
+    if (heads[old_used] == pos)
+        heads[old_used] = nexts[pos];
+    prevs[pos] = UINT32_MAX;
+    if (heads[new_used] != UINT32_MAX)
+        prevs[heads[new_used]] = pos;
+    nexts[pos] = heads[new_used];
+    heads[new_used] = pos;
+}
+
+void multilist_index_t::print()
+{
+    printf("heads:");
+    for (int i = 0; i < max_used; i++)
+        if (heads[i] != UINT32_MAX)
+            printf(" %u=%u", i, heads[i]);
+    printf("\n");
+    printf("prevs:");
+    for (int i = 0; i < count; i++)
+        if (prevs[i] != UINT32_MAX)
+            printf(" %d=%d", i, prevs[i]);
+    printf("\n");
+    printf("nexts:");
+    for (int i = 0; i < count; i++)
+        if (nexts[i] != UINT32_MAX)
+            printf(" %d=%d", i, nexts[i]);
+    printf("\n");
+}

src/blockstore/multilist.h

@@ -0,0 +1,37 @@
+// Variable-length O(1) disk space allocator
+// Copyright (c) Vitaliy Filippov, 2025+
+// License: VNPL-1.1 (see README.md for details)
+
+#pragma once
+
+#include <stdint.h>
+#include <vector>
+
+struct multilist_alloc_t
+{
+    const uint32_t count, maxn;
+    std::vector<int32_t> sizes;
+    std::vector<uint32_t> nexts, prevs, heads;
+
+    multilist_alloc_t(uint32_t count, uint32_t maxn);
+    bool is_free(uint32_t pos);
+    uint32_t find(uint32_t size);
+    void use_full(uint32_t pos);
+    void use(uint32_t pos, uint32_t size);
+    void do_free(uint32_t pos);
+    void free(uint32_t pos);
+    void verify();
+    void print();
+};
+
+struct multilist_index_t
+{
+    const uint32_t count, max_used;
+    std::vector<uint32_t> nexts, prevs, heads;
+
+    // used should be always < max_used
+    multilist_index_t(uint32_t count, uint32_t max_used, uint32_t init_used);
+    uint32_t find(uint32_t wanted_used);
+    void change(uint32_t pos, uint32_t old_used, uint32_t new_used);
+    void print();
+};