Series comparison

-[PULL 0/5] Block patches
+[PULL 0/8] Block patches
-The following changes since commit 711c0418c8c1ce3a24346f058b001c4c5a2f0f81:
+The following changes since commit 1be5a765c08cee3a9587c8a8d3fc2ea247b13f9c:
-  Merge remote-tracking branch 'remotes/philmd/tags/mips-20210702' into staging (2021-07-04 14:04:12 +0100)
+  Merge tag 'for-upstream' of https://gitlab.com/bonzini/qemu into staging (2022-04-19 18:22:16 -0700)
 are available in the Git repository at:
-  https://gitlab.com/stefanha/qemu.git tags/block-pull-request
+  https://gitlab.com/hreitz/qemu.git tags/pull-block-2022-04-20
-for you to fetch changes up to 9f460c64e13897117f35ffb61f6f5e0102cabc70:
+for you to fetch changes up to 0423f75351ab83b844a31349218b0eadd830e07a:
-  block/io: Merge discard request alignments (2021-07-06 14:28:55 +0100)
+  qcow2: Add errp to rebuild_refcount_structure() (2022-04-20 12:09:17 +0200)
 ----------------------------------------------------------------
-Pull request
+Block patches:
 - Some changes for qcow2's refcount repair algorithm to make it work for
   qcow2 images stored on block devices
 - Skip test cases that require zstd when support for it is missing
 - Some refactoring in the iotests' meson.build
 ----------------------------------------------------------------
+Hanna Reitz (6):
+  iotests.py: Add supports_qcow2_zstd_compression()
+  iotests/065: Check for zstd support
+  iotests/303: Check for zstd support
+  qcow2: Improve refcount structure rebuilding
+  iotests/108: Test new refcount rebuild algorithm
+  qcow2: Add errp to rebuild_refcount_structure()
-Akihiko Odaki (3):
+Thomas Huth (2):
-  block/file-posix: Optimize for macOS
+  tests/qemu-iotests/meson.build: Improve the indentation
-  block: Add backend_defaults property
+  tests/qemu-iotests: Move the bash and sanitizer checks to meson.build
   block/io: Merge discard request alignments
-Stefan Hajnoczi (2):
+ block/qcow2-refcount.c         | 353 +++++++++++++++++++++++----------
-  util/async: add a human-readable name to BHs for debugging
+ tests/check-block.sh           |  26 ---
-  util/async: print leaked BH name when AioContext finalizes
+ tests/qemu-iotests/065         |  24 ++-
+ tests/qemu-iotests/108         | 259 +++++++++++++++++++++++-
- include/block/aio.h            | 31 ++++++++++++++++++++++---
+ tests/qemu-iotests/108.out     |  81 ++++++++
- include/hw/block/block.h       |  3 +++
+ tests/qemu-iotests/303         |   4 +-
- include/qemu/main-loop.h       |  4 +++-
+ tests/qemu-iotests/iotests.py  |  20 ++
- block/file-posix.c             | 27 ++++++++++++++++++++--
+ tests/qemu-iotests/meson.build |  73 ++++---
- block/io.c                     |  2 ++
+files changed, 673 insertions(+), 167 deletions(-)
  hw/block/block.c               | 42 ++++++++++++++++++++++++++++++----
  tests/unit/ptimer-test-stubs.c |  2 +-
  util/async.c                   | 25 ++++++++++++++++----
  util/main-loop.c               |  4 ++--
  tests/qemu-iotests/172.out     | 38 ++++++++++++++++++++++++++++++
 files changed, 161 insertions(+), 17 deletions(-)
 --
-.31.1
+.35.1

-[PULL 2/5] util/async: print leaked BH name when AioContext finalizes
+[PULL 1/8] tests/qemu-iotests/meson.build: Improve the indentation
-BHs must be deleted before the AioContext is finalized. If not, it's a
+From: Thomas Huth <thuth@redhat.com>
 bug and probably indicates that some part of the program still expects
 the BH to run in the future. That can lead to memory leaks, inconsistent
 state, or just hangs.
-Unfortunately the assert(flags & BH_DELETED) call in aio_ctx_finalize()
+By using subdir_done(), we can get rid of one level of indentation
-is difficult to debug because the assertion failure contains no
+in this file. This will make it easier to add more conditions to
-information about the BH!
+skip the iotests in future patches.
-Use the QEMUBH name field added in the previous patch to show a useful
+Reviewed-by: Hanna Reitz <hreitz@redhat.com>
-error when a leaked BH is detected.
+Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
 Signed-off-by: Thomas Huth <thuth@redhat.com>
 Message-Id: <20220223093840.2515281-3-thuth@redhat.com>
 Signed-off-by: Hanna Reitz <hreitz@redhat.com>
 ---
  tests/qemu-iotests/meson.build | 61 ++++++++++++++++++----------------
 file changed, 32 insertions(+), 29 deletions(-)
-Suggested-by: Eric Ernst <eric.g.ernst@gmail.com>
+diff --git a/tests/qemu-iotests/meson.build b/tests/qemu-iotests/meson.build
-Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
+index XXXXXXX..XXXXXXX 100644
-Message-Id: <20210414200247.917496-3-stefanha@redhat.com>
+--- a/tests/qemu-iotests/meson.build
----
++++ b/tests/qemu-iotests/meson.build
- util/async.c | 16 ++++++++++++++--
+@@ -XXX,XX +XXX,XX @@
-file changed, 14 insertions(+), 2 deletions(-)
+-if have_tools and targetos != 'windows' and not get_option('gprof')
 -  qemu_iotests_binaries = [qemu_img, qemu_io, qemu_nbd, qsd]
 -  qemu_iotests_env = {'PYTHON': python.full_path()}
 -  qemu_iotests_formats = {
 -    'qcow2': 'quick',
 -    'raw': 'slow',
 -    'qed': 'thorough',
 -    'vmdk': 'thorough',
 -    'vpc': 'thorough'
 -  }
 -
 -  foreach k, v : emulators
 -    if k.startswith('qemu-system-')
 -      qemu_iotests_binaries += v
 -    endif
 -  endforeach
 -  foreach format, speed: qemu_iotests_formats
 -    if speed == 'quick'
 -      suites = 'block'
 -    else
 -      suites = ['block-' + speed, speed]
 -    endif
 -    test('qemu-iotests ' + format, sh, args: [files('../check-block.sh'), format],
 -         depends: qemu_iotests_binaries, env: qemu_iotests_env,
 -         protocol: 'tap',
 -         suite: suites,
 -         timeout: 0,
 -         is_parallel: false)
 -  endforeach
 +if not have_tools or targetos == 'windows' or get_option('gprof')
 +  subdir_done()
  endif
 +
 +qemu_iotests_binaries = [qemu_img, qemu_io, qemu_nbd, qsd]
 +qemu_iotests_env = {'PYTHON': python.full_path()}
 +qemu_iotests_formats = {
 +  'qcow2': 'quick',
 +  'raw': 'slow',
 +  'qed': 'thorough',
 +  'vmdk': 'thorough',
 +  'vpc': 'thorough'
 +}
 +
 +foreach k, v : emulators
 +  if k.startswith('qemu-system-')
 +    qemu_iotests_binaries += v
 +  endif
 +endforeach
 +
 +foreach format, speed: qemu_iotests_formats
 +  if speed == 'quick'
 +    suites = 'block'
 +  else
 +    suites = ['block-' + speed, speed]
 +  endif
 +  test('qemu-iotests ' + format, sh, args: [files('../check-block.sh'), format],
 +       depends: qemu_iotests_binaries, env: qemu_iotests_env,
 +       protocol: 'tap',
 +       suite: suites,
 +       timeout: 0,
 +       is_parallel: false)
 +endforeach
 --
 .35.1
-diff --git a/util/async.c b/util/async.c
-index XXXXXXX..XXXXXXX 100644
---- a/util/async.c
-+++ b/util/async.c
-@@ -XXX,XX +XXX,XX @@ aio_ctx_finalize(GSource     *source)
-     assert(QSIMPLEQ_EMPTY(&ctx->bh_slice_list));
-     while ((bh = aio_bh_dequeue(&ctx->bh_list, &flags))) {
--        /* qemu_bh_delete() must have been called on BHs in this AioContext */
--        assert(flags & BH_DELETED);
-+        /*
-+         * qemu_bh_delete() must have been called on BHs in this AioContext. In
-+         * many cases memory leaks, hangs, or inconsistent state occur when a
-+         * BH is leaked because something still expects it to run.
-+         *
-+         * If you hit this, fix the lifecycle of the BH so that
-+         * qemu_bh_delete() and any associated cleanup is called before the
-+         * AioContext is finalized.
-+         */
-+        if (unlikely(!(flags & BH_DELETED))) {
-+            fprintf(stderr, "%s: BH '%s' leaked, aborting...\n",
-+                    __func__, bh->name);
-+            abort();
-+        }
-         g_free(bh);
-     }
---
-.31.1

-New patch
+[PULL 2/8] tests/qemu-iotests: Move the bash and sanitizer checks to meson.build
+From: Thomas Huth <thuth@redhat.com>
+We want to get rid of check-block.sh in the long run, so let's move
+the checks for the bash version and sanitizers from check-block.sh
+into the meson.build file instead.
+Signed-off-by: Thomas Huth <thuth@redhat.com>
+Message-Id: <20220223093840.2515281-4-thuth@redhat.com>
+Signed-off-by: Hanna Reitz <hreitz@redhat.com>
+---
+ tests/check-block.sh           | 26 --------------------------
+ tests/qemu-iotests/meson.build | 14 ++++++++++++++
+files changed, 14 insertions(+), 26 deletions(-)
+diff --git a/tests/check-block.sh b/tests/check-block.sh
+index XXXXXXX..XXXXXXX 100755
+--- a/tests/check-block.sh
++++ b/tests/check-block.sh
+@@ -XXX,XX +XXX,XX @@ skip() {
+     exit 0
+ }
+-# Disable tests with any sanitizer except for specific ones
+-SANITIZE_FLAGS=$( grep "CFLAGS.*-fsanitize" config-host.mak 2>/dev/null )
+-ALLOWED_SANITIZE_FLAGS="safe-stack cfi-icall"
+-#Remove all occurrencies of allowed Sanitize flags
+-for j in ${ALLOWED_SANITIZE_FLAGS}; do
+-    TMP_FLAGS=${SANITIZE_FLAGS}
+-    SANITIZE_FLAGS=""
+-    for i in ${TMP_FLAGS}; do
+-        if ! echo ${i} | grep -q "${j}" 2>/dev/null; then
+-            SANITIZE_FLAGS="${SANITIZE_FLAGS} ${i}"
+-        fi
+-    done
+-done
+-if echo ${SANITIZE_FLAGS} | grep -q "\-fsanitize" 2>/dev/null; then
+-    # Have a sanitize flag that is not allowed, stop
+-    skip "Sanitizers are enabled ==> Not running the qemu-iotests."
+-fi
+-
+ if [ -z "$(find . -name 'qemu-system-*' -print)" ]; then
+     skip "No qemu-system binary available ==> Not running the qemu-iotests."
+ fi
+-if ! command -v bash >/dev/null 2>&1 ; then
+-    skip "bash not available ==> Not running the qemu-iotests."
+-fi
+-
+-if LANG=C bash --version | grep -q 'GNU bash, version [123]' ; then
+-    skip "bash version too old ==> Not running the qemu-iotests."
+-fi
+-
+ cd tests/qemu-iotests
+ # QEMU_CHECK_BLOCK_AUTO is used to disable some unstable sub-tests
+diff --git a/tests/qemu-iotests/meson.build b/tests/qemu-iotests/meson.build
+index XXXXXXX..XXXXXXX 100644
+--- a/tests/qemu-iotests/meson.build
++++ b/tests/qemu-iotests/meson.build
+@@ -XXX,XX +XXX,XX @@ if not have_tools or targetos == 'windows' or get_option('gprof')
+   subdir_done()
+ endif
++foreach cflag: config_host['QEMU_CFLAGS'].split()
++  if cflag.startswith('-fsanitize') and \
++     not cflag.contains('safe-stack') and not cflag.contains('cfi-icall')
++    message('Sanitizers are enabled ==> Disabled the qemu-iotests.')
++    subdir_done()
++  endif
++endforeach
++
++bash = find_program('bash', required: false, version: '>= 4.0')
++if not bash.found()
++  message('bash >= v4.0 not available ==> Disabled the qemu-iotests.')
++  subdir_done()
++endif
++
+ qemu_iotests_binaries = [qemu_img, qemu_io, qemu_nbd, qsd]
+ qemu_iotests_env = {'PYTHON': python.full_path()}
+ qemu_iotests_formats = {
+--
+.35.1

-New patch
+[PULL 3/8] iotests.py: Add supports_qcow2_zstd_compression()
+Suggested-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
+Signed-off-by: Hanna Reitz <hreitz@redhat.com>
+Message-Id: <20220323105522.53660-2-hreitz@redhat.com>
+Reviewed-by: Vladimir Sementsov-Ogievskiy <v.sementsov-og@mail.ru>
+---
+ tests/qemu-iotests/iotests.py | 20 ++++++++++++++++++++
+file changed, 20 insertions(+)
+diff --git a/tests/qemu-iotests/iotests.py b/tests/qemu-iotests/iotests.py
+index XXXXXXX..XXXXXXX 100644
+--- a/tests/qemu-iotests/iotests.py
++++ b/tests/qemu-iotests/iotests.py
+@@ -XXX,XX +XXX,XX @@ def verify_working_luks():
+     if not working:
+         notrun(reason)
++def supports_qcow2_zstd_compression() -> bool:
++    img_file = f'{test_dir}/qcow2-zstd-test.qcow2'
++    res = qemu_img('create', '-f', 'qcow2', '-o', 'compression_type=zstd',
++                   img_file, '0',
++                   check=False)
++    try:
++        os.remove(img_file)
++    except OSError:
++        pass
++
++    if res.returncode == 1 and \
++            "'compression-type' does not accept value 'zstd'" in res.stdout:
++        return False
++    else:
++        return True
++
++def verify_qcow2_zstd_compression():
++    if not supports_qcow2_zstd_compression():
++        notrun('zstd compression not supported')
++
+ def qemu_pipe(*args: str) -> str:
+     """
+     Run qemu with an option to print something and exit (e.g. a help option).
+--
+.35.1

-New patch
+[PULL 4/8] iotests/065: Check for zstd support
+Some test cases run in iotest 065 want to run with zstd compression just
+for added coverage.  Run them with zlib if there is no zstd support
+compiled in.
+Reported-by: Thomas Huth <thuth@redhat.com>
+Fixes: 12a936171d71f839dc907ff ("iotest 065: explicit compression type")
+Signed-off-by: Hanna Reitz <hreitz@redhat.com>
+Message-Id: <20220323105522.53660-3-hreitz@redhat.com>
+Reviewed-by: Vladimir Sementsov-Ogievskiy <v.sementsov-og@mail.ru>
+---
+ tests/qemu-iotests/065 | 24 ++++++++++++++++++------
+file changed, 18 insertions(+), 6 deletions(-)
+diff --git a/tests/qemu-iotests/065 b/tests/qemu-iotests/065
+index XXXXXXX..XXXXXXX 100755
+--- a/tests/qemu-iotests/065
++++ b/tests/qemu-iotests/065
+@@ -XXX,XX +XXX,XX @@ import os
+ import re
+ import json
+ import iotests
+-from iotests import qemu_img, qemu_img_info
++from iotests import qemu_img, qemu_img_info, supports_qcow2_zstd_compression
+ import unittest
+ test_img = os.path.join(iotests.test_dir, 'test.img')
+@@ -XXX,XX +XXX,XX @@ class TestQCow2(TestQemuImgInfo):
+ class TestQCow3NotLazy(TestQemuImgInfo):
+     '''Testing a qcow2 version 3 image with lazy refcounts disabled'''
+-    img_options = 'compat=1.1,lazy_refcounts=off,compression_type=zstd'
++    if supports_qcow2_zstd_compression():
++        compression_type = 'zstd'
++    else:
++        compression_type = 'zlib'
++
++    img_options = 'compat=1.1,lazy_refcounts=off'
++    img_options += f',compression_type={compression_type}'
+     json_compare = { 'compat': '1.1', 'lazy-refcounts': False,
+                      'refcount-bits': 16, 'corrupt': False,
+-                     'compression-type': 'zstd', 'extended-l2': False }
+-    human_compare = [ 'compat: 1.1', 'compression type: zstd',
++                     'compression-type': compression_type, 'extended-l2': False }
++    human_compare = [ 'compat: 1.1', f'compression type: {compression_type}',
+                       'lazy refcounts: false', 'refcount bits: 16',
+                       'corrupt: false', 'extended l2: false' ]
+@@ -XXX,XX +XXX,XX @@ class TestQCow3NotLazyQMP(TestQMP):
+ class TestQCow3LazyQMP(TestQMP):
+     '''Testing a qcow2 version 3 image with lazy refcounts enabled, opening
+        with lazy refcounts disabled'''
+-    img_options = 'compat=1.1,lazy_refcounts=on,compression_type=zstd'
++    if supports_qcow2_zstd_compression():
++        compression_type = 'zstd'
++    else:
++        compression_type = 'zlib'
++
++    img_options = 'compat=1.1,lazy_refcounts=on'
++    img_options += f',compression_type={compression_type}'
+     qemu_options = 'lazy-refcounts=off'
+     compare = { 'compat': '1.1', 'lazy-refcounts': True,
+                 'refcount-bits': 16, 'corrupt': False,
+-                'compression-type': 'zstd', 'extended-l2': False }
++                'compression-type': compression_type, 'extended-l2': False }
+ TestImageInfoSpecific = None
+ TestQemuImgInfo = None
+--
+.35.1

-[PULL 4/5] block: Add backend_defaults property
+[PULL 5/8] iotests/303: Check for zstd support
-From: Akihiko Odaki <akihiko.odaki@gmail.com>
+runs two test cases, one of which requires zstd support.
 Unfortunately, given that this is not a unittest-style test, we cannot
 easily skip that single case, and instead can only skip the whole test.
-backend_defaults property allow users to control if default block
+(Alternatively, we could split this test into a zlib and a zstd part,
-properties should be decided with backend information.
+but that seems excessive, given that this test is not in auto and thus
 likely only run by developers who have zstd support compiled in.)
-If it is off, any backend information will be discarded, which is
+Fixes: 677e0bae686e7c670a71d1f ("iotest 303: explicit compression type")
-suitable if you plan to perform live migration to a different disk backend.
+Signed-off-by: Hanna Reitz <hreitz@redhat.com>
 Reviewed-by: Vladimir Sementsov-Ogievskiy <v.sementsov-og@mail.ru>
 Message-Id: <20220323105522.53660-4-hreitz@redhat.com>
 ---
  tests/qemu-iotests/303 | 4 +++-
 file changed, 3 insertions(+), 1 deletion(-)
-If it is on, a block device may utilize backend information more
+diff --git a/tests/qemu-iotests/303 b/tests/qemu-iotests/303
-aggressively.
+index XXXXXXX..XXXXXXX 100755
+--- a/tests/qemu-iotests/303
-By default, it is auto, which uses backend information for block
++++ b/tests/qemu-iotests/303
 sizes and ignores the others, which is consistent with the older
 versions.
 Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
 Message-id: 20210705130458.97642-2-akihiko.odaki@gmail.com
 Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
 ---
  include/hw/block/block.h   |  3 +++
  hw/block/block.c           | 42 ++++++++++++++++++++++++++++++++++----
  tests/qemu-iotests/172.out | 38 ++++++++++++++++++++++++++++++++++
 files changed, 79 insertions(+), 4 deletions(-)
 diff --git a/include/hw/block/block.h b/include/hw/block/block.h
 index XXXXXXX..XXXXXXX 100644
 --- a/include/hw/block/block.h
 +++ b/include/hw/block/block.h
 @@ -XXX,XX +XXX,XX @@
- typedef struct BlockConf {
+ import iotests
-     BlockBackend *blk;
+ import subprocess
-+    OnOffAuto backend_defaults;
+-from iotests import qemu_img_create, qemu_io, file_path, log, filter_qemu_io
-     uint32_t physical_block_size;
++from iotests import qemu_img_create, qemu_io, file_path, log, filter_qemu_io, \
-     uint32_t logical_block_size;
++        verify_qcow2_zstd_compression
-     uint32_t min_io_size;
-@@ -XXX,XX +XXX,XX @@ static inline unsigned int get_physical_block_exp(BlockConf *conf)
+ iotests.script_initialize(supported_fmts=['qcow2'],
- }
+                           unsupported_imgopts=['refcount_bits', 'compat'])
++verify_qcow2_zstd_compression()
- #define DEFINE_BLOCK_PROPERTIES_BASE(_state, _conf)                     \
-+    DEFINE_PROP_ON_OFF_AUTO("backend_defaults", _state,                 \
+ disk = file_path('disk')
-+                            _conf.backend_defaults, ON_OFF_AUTO_AUTO),  \
+ chunk = 1024 * 1024
      DEFINE_PROP_BLOCKSIZE("logical_block_size", _state,                 \
                            _conf.logical_block_size),                    \
      DEFINE_PROP_BLOCKSIZE("physical_block_size", _state,                \
 diff --git a/hw/block/block.c b/hw/block/block.c
 index XXXXXXX..XXXXXXX 100644
 --- a/hw/block/block.c
 +++ b/hw/block/block.c
@@ -XXX,XX +XXX,XX @@ bool blkconf_blocksizes(BlockConf *conf, Error **errp)
  {
      BlockBackend *blk = conf->blk;
      BlockSizes blocksizes;
 -    int backend_ret;
 +    BlockDriverState *bs;
 +    bool use_blocksizes;
 +    bool use_bs;
 +
 +    switch (conf->backend_defaults) {
 +    case ON_OFF_AUTO_AUTO:
 +        use_blocksizes = !blk_probe_blocksizes(blk, &blocksizes);
 +        use_bs = false;
 +        break;
 +
 +    case ON_OFF_AUTO_ON:
 +        use_blocksizes = !blk_probe_blocksizes(blk, &blocksizes);
 +        bs = blk_bs(blk);
 +        use_bs = bs;
 +        break;
 +
 +    case ON_OFF_AUTO_OFF:
 +        use_blocksizes = false;
 +        use_bs = false;
 +        break;
 +
 +    default:
 +        abort();
 +    }
 -    backend_ret = blk_probe_blocksizes(blk, &blocksizes);
      /* fill in detected values if they are not defined via qemu command line */
      if (!conf->physical_block_size) {
 -        if (!backend_ret) {
 +        if (use_blocksizes) {
             conf->physical_block_size = blocksizes.phys;
          } else {
              conf->physical_block_size = BDRV_SECTOR_SIZE;
          }
      }
      if (!conf->logical_block_size) {
 -        if (!backend_ret) {
 +        if (use_blocksizes) {
              conf->logical_block_size = blocksizes.log;
          } else {
              conf->logical_block_size = BDRV_SECTOR_SIZE;
          }
      }
 +    if (use_bs) {
 +        if (!conf->opt_io_size) {
 +            conf->opt_io_size = bs->bl.opt_transfer;
 +        }
 +        if (conf->discard_granularity == -1) {
 +            if (bs->bl.pdiscard_alignment) {
 +                conf->discard_granularity = bs->bl.pdiscard_alignment;
 +            } else if (bs->bl.request_alignment != 1) {
 +                conf->discard_granularity = bs->bl.request_alignment;
 +            }
 +        }
 +    }
      if (conf->logical_block_size > conf->physical_block_size) {
          error_setg(errp,
 diff --git a/tests/qemu-iotests/172.out b/tests/qemu-iotests/172.out
 index XXXXXXX..XXXXXXX 100644
 --- a/tests/qemu-iotests/172.out
 +++ b/tests/qemu-iotests/172.out
@@ -XXX,XX +XXX,XX @@ Testing:
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = "floppy0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fda TEST_DIR/t.qcow2
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = "floppy0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fdb TEST_DIR/t.qcow2
                dev: floppy, id ""
                  unit = 1 (0x1)
                  drive = "floppy1"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fdb TEST_DIR/t.qcow2
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = "floppy0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fda TEST_DIR/t.qcow2 -fdb TEST_DIR/t.qcow2.2
                dev: floppy, id ""
                  unit = 1 (0x1)
                  drive = "floppy1"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fda TEST_DIR/t.qcow2 -fdb TEST_DIR/t.qcow2.2
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = "floppy0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fdb
                dev: floppy, id ""
                  unit = 1 (0x1)
                  drive = "floppy1"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fdb
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = "floppy0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=floppy,file=TEST_DIR/t.qcow2
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = "floppy0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=floppy,file=TEST_DIR/t.qcow2,index=1
                dev: floppy, id ""
                  unit = 1 (0x1)
                  drive = "floppy1"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=floppy,file=TEST_DIR/t.qcow2,index=1
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = "floppy0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=floppy,file=TEST_DIR/t.qcow2 -drive if=floppy,file=TEST_DIR/t
                dev: floppy, id ""
                  unit = 1 (0x1)
                  drive = "floppy1"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=floppy,file=TEST_DIR/t.qcow2 -drive if=floppy,file=TEST_DIR/t
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = "floppy0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=none,file=TEST_DIR/t.qcow2 -device floppy,drive=none0
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = "none0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=none,file=TEST_DIR/t.qcow2 -device floppy,drive=none0,unit=1
                dev: floppy, id ""
                  unit = 1 (0x1)
                  drive = "none0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=none,file=TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.qco
                dev: floppy, id ""
                  unit = 1 (0x1)
                  drive = "none1"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=none,file=TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.qco
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = "none0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fda TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.qcow2.2 -device fl
                dev: floppy, id ""
                  unit = 1 (0x1)
                  drive = "none0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fda TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.qcow2.2 -device fl
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = "floppy0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fda TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.qcow2.2 -device fl
                dev: floppy, id ""
                  unit = 1 (0x1)
                  drive = "none0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fda TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.qcow2.2 -device fl
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = "floppy0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fdb TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.qcow2.2 -device fl
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = "none0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fdb TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.qcow2.2 -device fl
                dev: floppy, id ""
                  unit = 1 (0x1)
                  drive = "floppy1"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fdb TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.qcow2.2 -device fl
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = "none0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fdb TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.qcow2.2 -device fl
                dev: floppy, id ""
                  unit = 1 (0x1)
                  drive = "floppy1"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=floppy,file=TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.q
                dev: floppy, id ""
                  unit = 1 (0x1)
                  drive = "none0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=floppy,file=TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.q
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = "floppy0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=floppy,file=TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.q
                dev: floppy, id ""
                  unit = 1 (0x1)
                  drive = "none0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=floppy,file=TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.q
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = "floppy0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=none,file=TEST_DIR/t.qcow2 -global floppy.drive=none0 -device
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = "none0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -device floppy
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = ""
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -device floppy,drive-type=120
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = ""
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -device floppy,drive-type=144
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = ""
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -device floppy,drive-type=288
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = ""
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=none,file=TEST_DIR/t.qcow2 -device floppy,drive=none0,drive-t
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = "none0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=none,file=TEST_DIR/t.qcow2 -device floppy,drive=none0,drive-t
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = "none0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=none,file=TEST_DIR/t.qcow2 -device floppy,drive=none0,logical
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = "none0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=none,file=TEST_DIR/t.qcow2 -device floppy,drive=none0,physica
                dev: floppy, id ""
                  unit = 0 (0x0)
                  drive = "none0"
 +                backend_defaults = "auto"
                  logical_block_size = 512 (512 B)
                  physical_block_size = 512 (512 B)
                  min_io_size = 0 (0 B)
 --
-.31.1
+.35.1

-[PULL 3/5] block/file-posix: Optimize for macOS
+[PULL 6/8] qcow2: Improve refcount structure rebuilding
-From: Akihiko Odaki <akihiko.odaki@gmail.com>
+When rebuilding the refcount structures (when qemu-img check -r found
 errors with refcount = 0, but reference count > 0), the new refcount
 table defaults to being put at the image file end[1].  There is no good
 reason for that except that it means we will not have to rewrite any
 refblocks we already wrote to disk.
-This commit introduces "punch hole" operation and optimizes transfer
+Changing the code to rewrite those refblocks is not too difficult,
-block size for macOS.
+though, so let us do that.  That is beneficial for images on block
 devices, where we cannot really write beyond the end of the image file.
-Thanks to Konstantin Nazarov for detailed analysis of a flaw in an
+Use this opportunity to add extensive comments to the code, and refactor
-old version of this change:
+it a bit, getting rid of the backwards-jumping goto.
 https://gist.github.com/akihikodaki/87df4149e7ca87f18dc56807ec5a1bc5#gistcomment-3654667
-Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
+[1] Unless there is something allocated in the area pointed to by the
-Message-id: 20210705130458.97642-1-akihiko.odaki@gmail.com
+    last refblock, so we have to write that refblock.  In that case, we
-Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
+    try to put the reftable in there.
 Buglink: https://bugzilla.redhat.com/show_bug.cgi?id=1519071
 Closes: https://gitlab.com/qemu-project/qemu/-/issues/941
 Reviewed-by: Eric Blake <eblake@redhat.com>
 Signed-off-by: Hanna Reitz <hreitz@redhat.com>
 Message-Id: <20220405134652.19278-2-hreitz@redhat.com>
 ---
- block/file-posix.c | 27 +++++++++++++++++++++++++--
+ block/qcow2-refcount.c | 332 +++++++++++++++++++++++++++++------------
-file changed, 25 insertions(+), 2 deletions(-)
+file changed, 235 insertions(+), 97 deletions(-)
-diff --git a/block/file-posix.c b/block/file-posix.c
+diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
 index XXXXXXX..XXXXXXX 100644
---- a/block/file-posix.c
+--- a/block/qcow2-refcount.c
-+++ b/block/file-posix.c
++++ b/block/qcow2-refcount.c
-@@ -XXX,XX +XXX,XX @@
+@@ -XXX,XX +XXX,XX @@ static int64_t alloc_clusters_imrt(BlockDriverState *bs,
- #if defined(HAVE_HOST_BLOCK_DEVICE)
+ }
- #include <paths.h>
- #include <sys/param.h>
+ /*
-+#include <sys/mount.h>
+- * Creates a new refcount structure based solely on the in-memory information
- #include <IOKit/IOKitLib.h>
+- * given through *refcount_table. All necessary allocations will be reflected
- #include <IOKit/IOBSD.h>
+- * in that array.
- #include <IOKit/storage/IOMediaBSDClient.h>
++ * Helper function for rebuild_refcount_structure().
-@@ -XXX,XX +XXX,XX @@ static void raw_refresh_limits(BlockDriverState *bs, Error **errp)
+  *
-         return;
+- * On success, the old refcount structure is leaked (it will be covered by the
 - * new refcount structure).
 + * Scan the range of clusters [first_cluster, end_cluster) for allocated
 + * clusters and write all corresponding refblocks to disk.  The refblock
 + * and allocation data is taken from the in-memory refcount table
 + * *refcount_table[] (of size *nb_clusters), which is basically one big
 + * (unlimited size) refblock for the whole image.
 + *
 + * For these refblocks, clusters are allocated using said in-memory
 + * refcount table.  Care is taken that these allocations are reflected
 + * in the refblocks written to disk.
 + *
 + * The refblocks' offsets are written into a reftable, which is
 + * *on_disk_reftable_ptr[] (of size *on_disk_reftable_entries_ptr).  If
 + * that reftable is of insufficient size, it will be resized to fit.
 + * This reftable is not written to disk.
 + *
 + * (If *on_disk_reftable_ptr is not NULL, the entries within are assumed
 + * to point to existing valid refblocks that do not need to be allocated
 + * again.)
 + *
 + * Return whether the on-disk reftable array was resized (true/false),
 + * or -errno on error.
   */
 -static int rebuild_refcount_structure(BlockDriverState *bs,
 -                                      BdrvCheckResult *res,
 -                                      void **refcount_table,
 -                                      int64_t *nb_clusters)
 +static int rebuild_refcounts_write_refblocks(
 +        BlockDriverState *bs, void **refcount_table, int64_t *nb_clusters,
 +        int64_t first_cluster, int64_t end_cluster,
 +        uint64_t **on_disk_reftable_ptr, uint32_t *on_disk_reftable_entries_ptr
 +    )
  {
      BDRVQcow2State *s = bs->opaque;
 -    int64_t first_free_cluster = 0, reftable_offset = -1, cluster = 0;
 +    int64_t cluster;
      int64_t refblock_offset, refblock_start, refblock_index;
 -    uint32_t reftable_size = 0;
 -    uint64_t *on_disk_reftable = NULL;
 +    int64_t first_free_cluster = 0;
 +    uint64_t *on_disk_reftable = *on_disk_reftable_ptr;
 +    uint32_t on_disk_reftable_entries = *on_disk_reftable_entries_ptr;
      void *on_disk_refblock;
 -    int ret = 0;
 -    struct {
 -        uint64_t reftable_offset;
 -        uint32_t reftable_clusters;
 -    } QEMU_PACKED reftable_offset_and_clusters;
 -
 -    qcow2_cache_empty(bs, s->refcount_block_cache);
 +    bool reftable_grown = false;
 +    int ret;
 -write_refblocks:
 -    for (; cluster < *nb_clusters; cluster++) {
 +    for (cluster = first_cluster; cluster < end_cluster; cluster++) {
 +        /* Check all clusters to find refblocks that contain non-zero entries */
          if (!s->get_refcount(*refcount_table, cluster)) {
              continue;
          }
 +        /*
 +         * This cluster is allocated, so we need to create a refblock
 +         * for it.  The data we will write to disk is just the
 +         * respective slice from *refcount_table, so it will contain
 +         * accurate refcounts for all clusters belonging to this
 +         * refblock.  After we have written it, we will therefore skip
 +         * all remaining clusters in this refblock.
 +         */
 +
          refblock_index = cluster >> s->refcount_block_bits;
          refblock_start = refblock_index << s->refcount_block_bits;
 -        /* Don't allocate a cluster in a refblock already written to disk */
 -        if (first_free_cluster < refblock_start) {
 -            first_free_cluster = refblock_start;
 -        }
 -        refblock_offset = alloc_clusters_imrt(bs, 1, refcount_table,
 -                                              nb_clusters, &first_free_cluster);
 -        if (refblock_offset < 0) {
 -            fprintf(stderr, "ERROR allocating refblock: %s\n",
 -                    strerror(-refblock_offset));
 -            res->check_errors++;
 -            ret = refblock_offset;
 -            goto fail;
 -        }
 +        if (on_disk_reftable_entries > refblock_index &&
 +            on_disk_reftable[refblock_index])
 +        {
 +            /*
 +             * We can get here after a `goto write_refblocks`: We have a
 +             * reftable from a previous run, and the refblock is already
 +             * allocated.  No need to allocate it again.
 +             */
 +            refblock_offset = on_disk_reftable[refblock_index];
 +        } else {
 +            int64_t refblock_cluster_index;
 -        if (reftable_size <= refblock_index) {
 -            uint32_t old_reftable_size = reftable_size;
 -            uint64_t *new_on_disk_reftable;
 +            /* Don't allocate a cluster in a refblock already written to disk */
 +            if (first_free_cluster < refblock_start) {
 +                first_free_cluster = refblock_start;
 +            }
 +            refblock_offset = alloc_clusters_imrt(bs, 1, refcount_table,
 +                                                  nb_clusters,
 +                                                  &first_free_cluster);
 +            if (refblock_offset < 0) {
 +                fprintf(stderr, "ERROR allocating refblock: %s\n",
 +                        strerror(-refblock_offset));
 +                return refblock_offset;
 +            }
 -            reftable_size = ROUND_UP((refblock_index + 1) * REFTABLE_ENTRY_SIZE,
 -                                     s->cluster_size) / REFTABLE_ENTRY_SIZE;
 -            new_on_disk_reftable = g_try_realloc(on_disk_reftable,
 -                                                 reftable_size *
 -                                                 REFTABLE_ENTRY_SIZE);
 -            if (!new_on_disk_reftable) {
 -                res->check_errors++;
 -                ret = -ENOMEM;
 -                goto fail;
 +            refblock_cluster_index = refblock_offset / s->cluster_size;
 +            if (refblock_cluster_index >= end_cluster) {
 +                /*
 +                 * We must write the refblock that holds this refblock's
 +                 * refcount
 +                 */
 +                end_cluster = refblock_cluster_index + 1;
              }
 -            on_disk_reftable = new_on_disk_reftable;
 -            memset(on_disk_reftable + old_reftable_size, 0,
 -                   (reftable_size - old_reftable_size) * REFTABLE_ENTRY_SIZE);
 +            if (on_disk_reftable_entries <= refblock_index) {
 +                on_disk_reftable_entries =
 +                    ROUND_UP((refblock_index + 1) * REFTABLE_ENTRY_SIZE,
 +                             s->cluster_size) / REFTABLE_ENTRY_SIZE;
 +                on_disk_reftable =
 +                    g_try_realloc(on_disk_reftable,
 +                                  on_disk_reftable_entries *
 +                                  REFTABLE_ENTRY_SIZE);
 +                if (!on_disk_reftable) {
 +                    return -ENOMEM;
 +                }
 -            /* The offset we have for the reftable is now no longer valid;
 -             * this will leak that range, but we can easily fix that by running
 -             * a leak-fixing check after this rebuild operation */
 -            reftable_offset = -1;
 -        } else {
 -            assert(on_disk_reftable);
 -        }
 -        on_disk_reftable[refblock_index] = refblock_offset;
 +                memset(on_disk_reftable + *on_disk_reftable_entries_ptr, 0,
 +                       (on_disk_reftable_entries -
 +                        *on_disk_reftable_entries_ptr) *
 +                       REFTABLE_ENTRY_SIZE);
 -        /* If this is apparently the last refblock (for now), try to squeeze the
 -         * reftable in */
 -        if (refblock_index == (*nb_clusters - 1) >> s->refcount_block_bits &&
 -            reftable_offset < 0)
 -        {
 -            uint64_t reftable_clusters = size_to_clusters(s, reftable_size *
 -                                                          REFTABLE_ENTRY_SIZE);
 -            reftable_offset = alloc_clusters_imrt(bs, reftable_clusters,
 -                                                  refcount_table, nb_clusters,
 -                                                  &first_free_cluster);
 -            if (reftable_offset < 0) {
 -                fprintf(stderr, "ERROR allocating reftable: %s\n",
 -                        strerror(-reftable_offset));
 -                res->check_errors++;
 -                ret = reftable_offset;
 -                goto fail;
 +                *on_disk_reftable_ptr = on_disk_reftable;
 +                *on_disk_reftable_entries_ptr = on_disk_reftable_entries;
 +
 +                reftable_grown = true;
 +            } else {
 +                assert(on_disk_reftable);
              }
 +            on_disk_reftable[refblock_index] = refblock_offset;
          }
 +        /* Refblock is allocated, write it to disk */
 +
          ret = qcow2_pre_write_overlap_check(bs, 0, refblock_offset,
                                              s->cluster_size, false);
          if (ret < 0) {
              fprintf(stderr, "ERROR writing refblock: %s\n", strerror(-ret));
 -            goto fail;
 +            return ret;
          }
 -        /* The size of *refcount_table is always cluster-aligned, therefore the
 -         * write operation will not overflow */
 +        /*
 +         * The refblock is simply a slice of *refcount_table.
 +         * Note that the size of *refcount_table is always aligned to
 +         * whole clusters, so the write operation will not result in
 +         * out-of-bounds accesses.
 +         */
          on_disk_refblock = (void *)((char *) *refcount_table +
                                      refblock_index * s->cluster_size);
@@ -XXX,XX +XXX,XX @@ write_refblocks:
                            s->cluster_size);
          if (ret < 0) {
              fprintf(stderr, "ERROR writing refblock: %s\n", strerror(-ret));
 -            goto fail;
 +            return ret;
          }
 -        /* Go to the end of this refblock */
 +        /* This refblock is done, skip to its end */
          cluster = refblock_start + s->refcount_block_size - 1;
      }
-+#if defined(__APPLE__) && (__MACH__)
+-    if (reftable_offset < 0) {
-+    struct statfs buf;
+-        uint64_t post_refblock_start, reftable_clusters;
-+
++    return reftable_grown;
-+    if (!fstatfs(s->fd, &buf)) {
++}
-+        bs->bl.opt_transfer = buf.f_iosize;
++
-+        bs->bl.pdiscard_alignment = buf.f_bsize;
++/*
 + * Creates a new refcount structure based solely on the in-memory information
 + * given through *refcount_table (this in-memory information is basically just
 + * the concatenation of all refblocks).  All necessary allocations will be
 + * reflected in that array.
 + *
 + * On success, the old refcount structure is leaked (it will be covered by the
 + * new refcount structure).
 + */
 +static int rebuild_refcount_structure(BlockDriverState *bs,
 +                                      BdrvCheckResult *res,
 +                                      void **refcount_table,
 +                                      int64_t *nb_clusters)
 +{
 +    BDRVQcow2State *s = bs->opaque;
 +    int64_t reftable_offset = -1;
 +    int64_t reftable_length = 0;
 +    int64_t reftable_clusters;
 +    int64_t refblock_index;
 +    uint32_t on_disk_reftable_entries = 0;
 +    uint64_t *on_disk_reftable = NULL;
 +    int ret = 0;
 +    int reftable_size_changed = 0;
 +    struct {
 +        uint64_t reftable_offset;
 +        uint32_t reftable_clusters;
 +    } QEMU_PACKED reftable_offset_and_clusters;
 +
 +    qcow2_cache_empty(bs, s->refcount_block_cache);
 +
 +    /*
 +     * For each refblock containing entries, we try to allocate a
 +     * cluster (in the in-memory refcount table) and write its offset
 +     * into on_disk_reftable[].  We then write the whole refblock to
 +     * disk (as a slice of the in-memory refcount table).
 +     * This is done by rebuild_refcounts_write_refblocks().
 +     *
 +     * Once we have scanned all clusters, we try to find space for the
 +     * reftable.  This will dirty the in-memory refcount table (i.e.
 +     * make it differ from the refblocks we have already written), so we
 +     * need to run rebuild_refcounts_write_refblocks() again for the
 +     * range of clusters where the reftable has been allocated.
 +     *
 +     * This second run might make the reftable grow again, in which case
 +     * we will need to allocate another space for it, which is why we
 +     * repeat all this until the reftable stops growing.
 +     *
 +     * (This loop will terminate, because with every cluster the
 +     * reftable grows, it can accomodate a multitude of more refcounts,
 +     * so that at some point this must be able to cover the reftable
 +     * and all refblocks describing it.)
 +     *
 +     * We then convert the reftable to big-endian and write it to disk.
 +     *
 +     * Note that we never free any reftable allocations.  Doing so would
 +     * needlessly complicate the algorithm: The eventual second check
 +     * run we do will clean up all leaks we have caused.
 +     */
 +
 +    reftable_size_changed =
 +        rebuild_refcounts_write_refblocks(bs, refcount_table, nb_clusters,
 +                                          0, *nb_clusters,
 +                                          &on_disk_reftable,
 +                                          &on_disk_reftable_entries);
 +    if (reftable_size_changed < 0) {
 +        res->check_errors++;
 +        ret = reftable_size_changed;
 +        goto fail;
 +    }
-+#endif
++
-+
++    /*
-     if (bs->sg || S_ISBLK(st.st_mode)) {
++     * There was no reftable before, so rebuild_refcounts_write_refblocks()
-         int ret = hdev_get_max_hw_transfer(s->fd, &st);
++     * must have increased its size (from 0 to something).
++     */
-@@ -XXX,XX +XXX,XX @@ out:
++    assert(reftable_size_changed);
 +
 +    do {
 +        int64_t reftable_start_cluster, reftable_end_cluster;
 +        int64_t first_free_cluster = 0;
 +
 +        reftable_length = on_disk_reftable_entries * REFTABLE_ENTRY_SIZE;
 +        reftable_clusters = size_to_clusters(s, reftable_length);
 -        post_refblock_start = ROUND_UP(*nb_clusters, s->refcount_block_size);
 -        reftable_clusters =
 -            size_to_clusters(s, reftable_size * REFTABLE_ENTRY_SIZE);
 -        /* Not pretty but simple */
 -        if (first_free_cluster < post_refblock_start) {
 -            first_free_cluster = post_refblock_start;
 -        }
          reftable_offset = alloc_clusters_imrt(bs, reftable_clusters,
                                                refcount_table, nb_clusters,
                                                &first_free_cluster);
@@ -XXX,XX +XXX,XX @@ write_refblocks:
              goto fail;
          }
 -        goto write_refblocks;
 -    }
 +        /*
 +         * We need to update the affected refblocks, so re-run the
 +         * write_refblocks loop for the reftable's range of clusters.
 +         */
 +        assert(offset_into_cluster(s, reftable_offset) == 0);
 +        reftable_start_cluster = reftable_offset / s->cluster_size;
 +        reftable_end_cluster = reftable_start_cluster + reftable_clusters;
 +        reftable_size_changed =
 +            rebuild_refcounts_write_refblocks(bs, refcount_table, nb_clusters,
 +                                              reftable_start_cluster,
 +                                              reftable_end_cluster,
 +                                              &on_disk_reftable,
 +                                              &on_disk_reftable_entries);
 +        if (reftable_size_changed < 0) {
 +            res->check_errors++;
 +            ret = reftable_size_changed;
 +            goto fail;
 +        }
 +
 +        /*
 +         * If the reftable size has changed, we will need to find a new
 +         * allocation, repeating the loop.
 +         */
 +    } while (reftable_size_changed);
 -    for (refblock_index = 0; refblock_index < reftable_size; refblock_index++) {
 +    /* The above loop must have run at least once */
 +    assert(reftable_offset >= 0);
 +
 +    /*
 +     * All allocations are done, all refblocks are written, convert the
 +     * reftable to big-endian and write it to disk.
 +     */
 +
 +    for (refblock_index = 0; refblock_index < on_disk_reftable_entries;
 +         refblock_index++)
 +    {
          cpu_to_be64s(&on_disk_reftable[refblock_index]);
      }
- }
+-    ret = qcow2_pre_write_overlap_check(bs, 0, reftable_offset,
-+#if defined(CONFIG_FALLOCATE) || defined(BLKZEROOUT) || defined(BLKDISCARD)
+-                                        reftable_size * REFTABLE_ENTRY_SIZE,
- static int translate_err(int err)
++    ret = qcow2_pre_write_overlap_check(bs, 0, reftable_offset, reftable_length,
- {
+                                         false);
-     if (err == -ENODEV || err == -ENOSYS || err == -EOPNOTSUPP ||
+     if (ret < 0) {
-@@ -XXX,XX +XXX,XX @@ static int translate_err(int err)
+         fprintf(stderr, "ERROR writing reftable: %s\n", strerror(-ret));
          goto fail;
      }
-     return err;
- }
+-    assert(reftable_size < INT_MAX / REFTABLE_ENTRY_SIZE);
-+#endif
++    assert(reftable_length < INT_MAX);
+     ret = bdrv_pwrite(bs->file, reftable_offset, on_disk_reftable,
- #ifdef CONFIG_FALLOCATE
+-                      reftable_size * REFTABLE_ENTRY_SIZE);
- static int do_fallocate(int fd, int mode, off_t offset, off_t len)
++                      reftable_length);
-@@ -XXX,XX +XXX,XX @@ static int handle_aiocb_discard(void *opaque)
+     if (ret < 0) {
-             }
+         fprintf(stderr, "ERROR writing reftable: %s\n", strerror(-ret));
-         } while (errno == EINTR);
+         goto fail;
+@@ -XXX,XX +XXX,XX @@ write_refblocks:
--        ret = -errno;
+     /* Enter new reftable into the image header */
-+        ret = translate_err(-errno);
+     reftable_offset_and_clusters.reftable_offset = cpu_to_be64(reftable_offset);
- #endif
+     reftable_offset_and_clusters.reftable_clusters =
-     } else {
+-        cpu_to_be32(size_to_clusters(s, reftable_size * REFTABLE_ENTRY_SIZE));
- #ifdef CONFIG_FALLOCATE_PUNCH_HOLE
++        cpu_to_be32(reftable_clusters);
-         ret = do_fallocate(s->fd, FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
+     ret = bdrv_pwrite_sync(bs->file,
-                            aiocb->aio_offset, aiocb->aio_nbytes);
+                            offsetof(QCowHeader, refcount_table_offset),
-+        ret = translate_err(-errno);
+                            &reftable_offset_and_clusters,
-+#elif defined(__APPLE__) && (__MACH__)
+@@ -XXX,XX +XXX,XX @@ write_refblocks:
-+        fpunchhole_t fpunchhole;
+         goto fail;
 +        fpunchhole.fp_flags = 0;
 +        fpunchhole.reserved = 0;
 +        fpunchhole.fp_offset = aiocb->aio_offset;
 +        fpunchhole.fp_length = aiocb->aio_nbytes;
 +        if (fcntl(s->fd, F_PUNCHHOLE, &fpunchhole) == -1) {
 +            ret = errno == ENODEV ? -ENOTSUP : -errno;
 +        } else {
 +            ret = 0;
 +        }
  #endif
      }
--    ret = translate_err(ret);
+-    for (refblock_index = 0; refblock_index < reftable_size; refblock_index++) {
-     if (ret == -ENOTSUP) {
++    for (refblock_index = 0; refblock_index < on_disk_reftable_entries;
-         s->has_discard = false;
++         refblock_index++)
 +    {
          be64_to_cpus(&on_disk_reftable[refblock_index]);
      }
+     s->refcount_table = on_disk_reftable;
+     s->refcount_table_offset = reftable_offset;
+-    s->refcount_table_size = reftable_size;
++    s->refcount_table_size = on_disk_reftable_entries;
+     update_max_refcount_table_index(s);
+     return 0;
 --
-.31.1
+.35.1

-[PULL 5/5] block/io: Merge discard request alignments
+[PULL 7/8] iotests/108: Test new refcount rebuild algorithm
-From: Akihiko Odaki <akihiko.odaki@gmail.com>
+One clear problem with how qcow2's refcount structure rebuild algorithm
 used to be before "qcow2: Improve refcount structure rebuilding" was
 that it is prone to failure for qcow2 images on block devices: There is
 generally unused space after the actual image, and if that exceeds what
 one refblock covers, the old algorithm would invariably write the
 reftable past the block device's end, which cannot work.  The new
 algorithm does not have this problem.
-Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
+Test it with three tests:
-Message-id: 20210705130458.97642-3-akihiko.odaki@gmail.com
+(1) Create an image with more empty space at the end than what one
-Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
+    refblock covers, see whether rebuilding the refcount structures
     results in a change in the image file length.  (It should not.)
 (2) Leave precisely enough space somewhere at the beginning of the image
     for the new reftable (and the refblock for that place), see whether
     the new algorithm puts the reftable there.  (It should.)
 (3) Test the original problem: Create (something like) a block device
     with a fixed size, then create a qcow2 image in there, write some
     data, and then have qemu-img check rebuild the refcount structures.
     Before HEAD^, the reftable would have been written past the image
     file end, i.e. outside of what the block device provides, which
     cannot work.  HEAD^ should have fixed that.
     ("Something like a block device" means a loop device if we can use
     one ("sudo -n losetup" works), or a FUSE block export with
     growable=false otherwise.)
 Reviewed-by: Eric Blake <eblake@redhat.com>
 Signed-off-by: Hanna Reitz <hreitz@redhat.com>
 Message-Id: <20220405134652.19278-3-hreitz@redhat.com>
 ---
- block/io.c | 2 ++
+ tests/qemu-iotests/108     | 259 ++++++++++++++++++++++++++++++++++++-
-file changed, 2 insertions(+)
+ tests/qemu-iotests/108.out |  81 ++++++++++++
 files changed, 339 insertions(+), 1 deletion(-)
-diff --git a/block/io.c b/block/io.c
+diff --git a/tests/qemu-iotests/108 b/tests/qemu-iotests/108
 index XXXXXXX..XXXXXXX 100755
 --- a/tests/qemu-iotests/108
 +++ b/tests/qemu-iotests/108
@@ -XXX,XX +XXX,XX @@ status=1    # failure is the default!
  _cleanup()
  {
 -    _cleanup_test_img
 +    _cleanup_test_img
 +    if [ -f "$TEST_DIR/qsd.pid" ]; then
 +        qsd_pid=$(cat "$TEST_DIR/qsd.pid")
 +        kill -KILL "$qsd_pid"
 +        fusermount -u "$TEST_DIR/fuse-export" &>/dev/null
 +    fi
 +    rm -f "$TEST_DIR/fuse-export"
  }
  trap "_cleanup; exit \$status" 0 1 2 3 15
  # get standard environment, filters and checks
  . ./common.rc
  . ./common.filter
 +. ./common.qemu
  # This tests qcow2-specific low-level functionality
  _supported_fmt qcow2
@@ -XXX,XX +XXX,XX @@ _supported_os Linux
  # files
  _unsupported_imgopts 'refcount_bits=\([^1]\|.\([^6]\|$\)\)' data_file
 +# This test either needs sudo -n losetup or FUSE exports to work
 +if sudo -n losetup &>/dev/null; then
 +    loopdev=true
 +else
 +    loopdev=false
 +
 +    # QSD --export fuse will either yield "Parameter 'id' is missing"
 +    # or "Invalid parameter 'fuse'", depending on whether there is
 +    # FUSE support or not.
 +    error=$($QSD --export fuse 2>&1)
 +    if [[ $error = *"'fuse'"* ]]; then
 +        _notrun 'Passwordless sudo for losetup or FUSE support required, but' \
 +                'neither is available'
 +    fi
 +fi
 +
  echo
  echo '=== Repairing an image without any refcount table ==='
  echo
@@ -XXX,XX +XXX,XX @@ _make_test_img 64M
  poke_file "$TEST_IMG" $((0x10008)) "\xff\xff\xff\xff\xff\xff\x00\x00"
  _check_test_img -r all
 +echo
 +echo '=== Check rebuilt reftable location ==='
 +
 +# In an earlier version of the refcount rebuild algorithm, the
 +# reftable was generally placed at the image end (unless something was
 +# allocated in the area covered by the refblock right before the image
 +# file end, then we would try to place the reftable in that refblock).
 +# This was later changed so the reftable would be placed in the
 +# earliest possible location.  Test this.
 +
 +echo
 +echo '--- Does the image size increase? ---'
 +echo
 +
 +# First test: Just create some image, write some data to it, and
 +# resize it so there is free space at the end of the image (enough
 +# that it spans at least one full refblock, which for cluster_size=512
 +# images, spans 128k).  With the old algorithm, the reftable would
 +# have then been placed at the end of the image file, but with the new
 +# one, it will be put in that free space.
 +# We want to check whether the size of the image file increases due to
 +# rebuilding the refcount structures (it should not).
 +
 +_make_test_img -o 'cluster_size=512' 1M
 +# Write something
 +$QEMU_IO -c 'write 0 64k' "$TEST_IMG" | _filter_qemu_io
 +
 +# Add free space
 +file_len=$(stat -c '%s' "$TEST_IMG")
 +truncate -s $((file_len + 256 * 1024)) "$TEST_IMG"
 +
 +# Corrupt the image by saying the image header was not allocated
 +rt_offset=$(peek_file_be "$TEST_IMG" 48 8)
 +rb_offset=$(peek_file_be "$TEST_IMG" $rt_offset 8)
 +poke_file "$TEST_IMG" $rb_offset "\x00\x00"
 +
 +# Check whether rebuilding the refcount structures increases the image
 +# file size
 +file_len=$(stat -c '%s' "$TEST_IMG")
 +echo
 +# The only leaks there can be are the old refcount structures that are
 +# leaked during rebuilding, no need to clutter the output with them
 +_check_test_img -r all | grep -v '^Repairing cluster.*refcount=1 reference=0'
 +echo
 +post_repair_file_len=$(stat -c '%s' "$TEST_IMG")
 +
 +if [[ $file_len -eq $post_repair_file_len ]]; then
 +    echo 'OK: Image size did not change'
 +else
 +    echo 'ERROR: Image size differs' \
 +        "($file_len before, $post_repair_file_len after)"
 +fi
 +
 +echo
 +echo '--- Will the reftable occupy a hole specifically left for it?  ---'
 +echo
 +
 +# Note: With cluster_size=512, every refblock covers 128k.
 +# The reftable covers 8M per reftable cluster.
 +
 +# Create an image that requires two reftable clusters (just because
 +# this is more interesting than a single-clustered reftable).
 +_make_test_img -o 'cluster_size=512' 9M
 +$QEMU_IO -c 'write 0 8M' "$TEST_IMG" | _filter_qemu_io
 +
 +# Writing 8M will have resized the reftable.  Unfortunately, doing so
 +# will leave holes in the file, so we need to fill them up so we can
 +# be sure the whole file is allocated.  Do that by writing
 +# consecutively smaller chunks starting from 8 MB, until the file
 +# length increases even with a chunk size of 512.  Then we must have
 +# filled all holes.
 +ofs=$((8 * 1024 * 1024))
 +block_len=$((16 * 1024))
 +while [[ $block_len -ge 512 ]]; do
 +    file_len=$(stat -c '%s' "$TEST_IMG")
 +    while [[ $(stat -c '%s' "$TEST_IMG") -eq $file_len ]]; do
 +        # Do not include this in the reference output, it does not
 +        # really matter which qemu-io calls we do here exactly
 +        $QEMU_IO -c "write $ofs $block_len" "$TEST_IMG" >/dev/null
 +        ofs=$((ofs + block_len))
 +    done
 +    block_len=$((block_len / 2))
 +done
 +
 +# Fill up to 9M (do not include this in the reference output either,
 +# $ofs is random for all we know)
 +$QEMU_IO -c "write $ofs $((9 * 1024 * 1024 - ofs))" "$TEST_IMG" >/dev/null
 +
 +# Make space as follows:
 +# - For the first refblock: Right at the beginning of the image (this
 +#   refblock is placed in the first place possible),
 +# - For the reftable somewhere soon afterwards, still near the
 +#   beginning of the image (i.e. covered by the first refblock); the
 +#   reftable too is placed in the first place possible, but only after
 +#   all refblocks have been placed)
 +# No space is needed for the other refblocks, because no refblock is
 +# put before the space it covers.  In this test case, we do not mind
 +# if they are placed at the image file's end.
 +
 +# Before we make that space, we have to find out the host offset of
 +# the area that belonged to the two data clusters at guest offset 4k,
 +# because we expect the reftable to be placed there, and we will have
 +# to verify that it is.
 +
 +l1_offset=$(peek_file_be "$TEST_IMG" 40 8)
 +l2_offset=$(peek_file_be "$TEST_IMG" $l1_offset 8)
 +l2_offset=$((l2_offset & 0x00fffffffffffe00))
 +data_4k_offset=$(peek_file_be "$TEST_IMG" \
 +                 $((l2_offset + 4096 / 512 * 8)) 8)
 +data_4k_offset=$((data_4k_offset & 0x00fffffffffffe00))
 +
 +$QEMU_IO -c "discard 0 512" -c "discard 4k 1k" "$TEST_IMG" | _filter_qemu_io
 +
 +# Corrupt the image by saying the image header was not allocated
 +rt_offset=$(peek_file_be "$TEST_IMG" 48 8)
 +rb_offset=$(peek_file_be "$TEST_IMG" $rt_offset 8)
 +poke_file "$TEST_IMG" $rb_offset "\x00\x00"
 +
 +echo
 +# The only leaks there can be are the old refcount structures that are
 +# leaked during rebuilding, no need to clutter the output with them
 +_check_test_img -r all | grep -v '^Repairing cluster.*refcount=1 reference=0'
 +echo
 +
 +# Check whether the reftable was put where we expected
 +rt_offset=$(peek_file_be "$TEST_IMG" 48 8)
 +if [[ $rt_offset -eq $data_4k_offset ]]; then
 +    echo 'OK: Reftable is where we expect it'
 +else
 +    echo "ERROR: Reftable is at $rt_offset, but was expected at $data_4k_offset"
 +fi
 +
 +echo
 +echo '--- Rebuilding refcount structures on block devices ---'
 +echo
 +
 +# A block device cannot really grow, at least not during qemu-img
 +# check.  As mentioned in the above cases, rebuilding the refcount
 +# structure may lead to new refcount structures being written after
 +# the end of the image, and in the past that happened even if there
 +# was more than sufficient space in the image.  Such post-EOF writes
 +# will not work on block devices, so test that the new algorithm
 +# avoids it.
 +
 +# If we have passwordless sudo and losetup, we can use those to create
 +# a block device.  Otherwise, we can resort to qemu's FUSE export to
 +# create a file that isn't growable, which effectively tests the same
 +# thing.
 +
 +_cleanup_test_img
 +truncate -s $((64 * 1024 * 1024)) "$TEST_IMG"
 +
 +if $loopdev; then
 +    export_mp=$(sudo -n losetup --show -f "$TEST_IMG")
 +    export_mp_driver=host_device
 +    sudo -n chmod go+rw "$export_mp"
 +else
 +    # Create non-growable FUSE export that is a bit like an empty
 +    # block device
 +    export_mp="$TEST_DIR/fuse-export"
 +    export_mp_driver=file
 +    touch "$export_mp"
 +
 +    $QSD \
 +        --blockdev file,node-name=export-node,filename="$TEST_IMG" \
 +        --export fuse,id=fuse-export,node-name=export-node,mountpoint="$export_mp",writable=on,growable=off \
 +        --pidfile "$TEST_DIR/qsd.pid" \
 +        --daemonize
 +fi
 +
 +# Now create a qcow2 image on the device -- unfortunately, qemu-img
 +# create force-creates the file, so we have to resort to the
 +# blockdev-create job.
 +_launch_qemu \
 +    --blockdev $export_mp_driver,node-name=file,filename="$export_mp"
 +
 +_send_qemu_cmd \
 +    $QEMU_HANDLE \
 +    '{ "execute": "qmp_capabilities" }' \
 +    'return'
 +
 +# Small cluster size again, so the image needs multiple refblocks
 +_send_qemu_cmd \
 +    $QEMU_HANDLE \
 +    '{ "execute": "blockdev-create",
 +       "arguments": {
 +           "job-id": "create",
 +           "options": {
 +               "driver": "qcow2",
 +               "file": "file",
 +               "size": '$((64 * 1024 * 1024))',
 +               "cluster-size": 512
 +           } } }' \
 +    '"concluded"'
 +
 +_send_qemu_cmd \
 +    $QEMU_HANDLE \
 +    '{ "execute": "job-dismiss", "arguments": { "id": "create" } }' \
 +    'return'
 +
 +_send_qemu_cmd \
 +    $QEMU_HANDLE \
 +    '{ "execute": "quit" }' \
 +    'return'
 +
 +wait=y _cleanup_qemu
 +echo
 +
 +# Write some data
 +$QEMU_IO -c 'write 0 64k' "$export_mp" | _filter_qemu_io
 +
 +# Corrupt the image by saying the image header was not allocated
 +rt_offset=$(peek_file_be "$export_mp" 48 8)
 +rb_offset=$(peek_file_be "$export_mp" $rt_offset 8)
 +poke_file "$export_mp" $rb_offset "\x00\x00"
 +
 +# Repairing such a simple case should just work
 +# (We used to put the reftable at the end of the image file, which can
 +# never work for non-growable devices.)
 +echo
 +TEST_IMG="$export_mp" _check_test_img -r all \
 +    | grep -v '^Repairing cluster.*refcount=1 reference=0'
 +
 +if $loopdev; then
 +    sudo -n losetup -d "$export_mp"
 +else
 +    qsd_pid=$(cat "$TEST_DIR/qsd.pid")
 +    kill -TERM "$qsd_pid"
 +    # Wait for process to exit (cannot `wait` because the QSD is daemonized)
 +    while [ -f "$TEST_DIR/qsd.pid" ]; do
 +        true
 +    done
 +fi
 +
  # success, all done
  echo '*** done'
  rm -f $seq.full
 diff --git a/tests/qemu-iotests/108.out b/tests/qemu-iotests/108.out
 index XXXXXXX..XXXXXXX 100644
---- a/block/io.c
+--- a/tests/qemu-iotests/108.out
-+++ b/block/io.c
++++ b/tests/qemu-iotests/108.out
-@@ -XXX,XX +XXX,XX @@ void bdrv_parent_drained_begin_single(BdrvChild *c, bool poll)
+@@ -XXX,XX +XXX,XX @@ The following inconsistencies were found and repaired:
+leaked clusters
- static void bdrv_merge_limits(BlockLimits *dst, const BlockLimits *src)
+corruptions
- {
-+    dst->pdiscard_alignment = MAX(dst->pdiscard_alignment,
++Double checking the fixed image now...
-+                                  src->pdiscard_alignment);
++No errors were found on the image.
-     dst->opt_transfer = MAX(dst->opt_transfer, src->opt_transfer);
++
-     dst->max_transfer = MIN_NON_ZERO(dst->max_transfer, src->max_transfer);
++=== Check rebuilt reftable location ===
-     dst->max_hw_transfer = MIN_NON_ZERO(dst->max_hw_transfer,
++
 +--- Does the image size increase? ---
 +
 +Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=1048576
 +wrote 65536/65536 bytes at offset 0
 +64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
 +
 +ERROR cluster 0 refcount=0 reference=1
 +Rebuilding refcount structure
 +The following inconsistencies were found and repaired:
 +
 +    0 leaked clusters
 +    1 corruptions
 +
 +Double checking the fixed image now...
 +No errors were found on the image.
 +
 +OK: Image size did not change
 +
 +--- Will the reftable occupy a hole specifically left for it?  ---
 +
 +Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=9437184
 +wrote 8388608/8388608 bytes at offset 0
 +8 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
 +discard 512/512 bytes at offset 0
 +512 bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
 +discard 1024/1024 bytes at offset 4096
 +1 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
 +
 +ERROR cluster 0 refcount=0 reference=1
 +Rebuilding refcount structure
 +The following inconsistencies were found and repaired:
 +
 +    0 leaked clusters
 +    1 corruptions
 +
 +Double checking the fixed image now...
 +No errors were found on the image.
 +
 +OK: Reftable is where we expect it
 +
 +--- Rebuilding refcount structures on block devices ---
 +
 +{ "execute": "qmp_capabilities" }
 +{"return": {}}
 +{ "execute": "blockdev-create",
 +       "arguments": {
 +           "job-id": "create",
 +           "options": {
 +               "driver": "IMGFMT",
 +               "file": "file",
 +               "size": 67108864,
 +               "cluster-size": 512
 +           } } }
 +{"timestamp": {"seconds":  TIMESTAMP, "microseconds":  TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "created", "id": "create"}}
 +{"timestamp": {"seconds":  TIMESTAMP, "microseconds":  TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "running", "id": "create"}}
 +{"return": {}}
 +{"timestamp": {"seconds":  TIMESTAMP, "microseconds":  TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "waiting", "id": "create"}}
 +{"timestamp": {"seconds":  TIMESTAMP, "microseconds":  TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "pending", "id": "create"}}
 +{"timestamp": {"seconds":  TIMESTAMP, "microseconds":  TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "concluded", "id": "create"}}
 +{ "execute": "job-dismiss", "arguments": { "id": "create" } }
 +{"timestamp": {"seconds":  TIMESTAMP, "microseconds":  TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "null", "id": "create"}}
 +{"return": {}}
 +{ "execute": "quit" }
 +{"return": {}}
 +{"timestamp": {"seconds":  TIMESTAMP, "microseconds":  TIMESTAMP}, "event": "SHUTDOWN", "data": {"guest": false, "reason": "host-qmp-quit"}}
 +
 +wrote 65536/65536 bytes at offset 0
 +64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
 +
 +ERROR cluster 0 refcount=0 reference=1
 +Rebuilding refcount structure
 +The following inconsistencies were found and repaired:
 +
 +    0 leaked clusters
 +    1 corruptions
 +
  Double checking the fixed image now...
  No errors were found on the image.
  *** done
 --
-.31.1
+.35.1

-[PULL 1/5] util/async: add a human-readable name to BHs for debugging
+[PULL 8/8] qcow2: Add errp to rebuild_refcount_structure()
-It can be difficult to debug issues with BHs in production environments.
+Instead of fprint()-ing error messages in rebuild_refcount_structure()
-Although BHs can usually be identified by looking up their ->cb()
+and its rebuild_refcounts_write_refblocks() helper, pass them through an
-function pointer, this requires debug information for the program. It is
+Error object to qcow2_check_refcounts() (which will then print it).
 also not possible to print human-readable diagnostics about BHs because
 they have no identifier.
-This patch adds a name to each BH. The name is not unique per instance
+Suggested-by: Eric Blake <eblake@redhat.com>
-but differentiates between cb() functions, which is usually enough. It's
+Signed-off-by: Hanna Reitz <hreitz@redhat.com>
-done by changing aio_bh_new() and friends to macros that stringify cb.
+Message-Id: <20220405134652.19278-4-hreitz@redhat.com>
 Reviewed-by: Eric Blake <eblake@redhat.com>
 ---
  block/qcow2-refcount.c | 33 +++++++++++++++++++--------------
 file changed, 19 insertions(+), 14 deletions(-)
-The next patch will use the name field when reporting leaked BHs.
+diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
 Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
 Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
 Message-Id: <20210414200247.917496-2-stefanha@redhat.com>
 ---
  include/block/aio.h            | 31 ++++++++++++++++++++++++++++---
  include/qemu/main-loop.h       |  4 +++-
  tests/unit/ptimer-test-stubs.c |  2 +-
  util/async.c                   |  9 +++++++--
  util/main-loop.c               |  4 ++--
 files changed, 41 insertions(+), 9 deletions(-)
 diff --git a/include/block/aio.h b/include/block/aio.h
 index XXXXXXX..XXXXXXX 100644
---- a/include/block/aio.h
+--- a/block/qcow2-refcount.c
-+++ b/include/block/aio.h
++++ b/block/qcow2-refcount.c
-@@ -XXX,XX +XXX,XX @@ void aio_context_acquire(AioContext *ctx);
+@@ -XXX,XX +XXX,XX @@ static int64_t alloc_clusters_imrt(BlockDriverState *bs,
- /* Relinquish ownership of the AioContext. */
+ static int rebuild_refcounts_write_refblocks(
- void aio_context_release(AioContext *ctx);
+         BlockDriverState *bs, void **refcount_table, int64_t *nb_clusters,
+         int64_t first_cluster, int64_t end_cluster,
-+/**
+-        uint64_t **on_disk_reftable_ptr, uint32_t *on_disk_reftable_entries_ptr
-+ * aio_bh_schedule_oneshot_full: Allocate a new bottom half structure that will
++        uint64_t **on_disk_reftable_ptr, uint32_t *on_disk_reftable_entries_ptr,
-+ * run only once and as soon as possible.
++        Error **errp
-+ *
+     )
 + * @name: A human-readable identifier for debugging purposes.
 + */
 +void aio_bh_schedule_oneshot_full(AioContext *ctx, QEMUBHFunc *cb, void *opaque,
 +                                  const char *name);
 +
  /**
   * aio_bh_schedule_oneshot: Allocate a new bottom half structure that will run
   * only once and as soon as possible.
 + *
 + * A convenience wrapper for aio_bh_schedule_oneshot_full() that uses cb as the
 + * name string.
   */
 -void aio_bh_schedule_oneshot(AioContext *ctx, QEMUBHFunc *cb, void *opaque);
 +#define aio_bh_schedule_oneshot(ctx, cb, opaque) \
 +    aio_bh_schedule_oneshot_full((ctx), (cb), (opaque), (stringify(cb)))
  /**
 - * aio_bh_new: Allocate a new bottom half structure.
 + * aio_bh_new_full: Allocate a new bottom half structure.
   *
   * Bottom halves are lightweight callbacks whose invocation is guaranteed
   * to be wait-free, thread-safe and signal-safe.  The #QEMUBH structure
   * is opaque and must be allocated prior to its use.
 + *
 + * @name: A human-readable identifier for debugging purposes.
   */
 -QEMUBH *aio_bh_new(AioContext *ctx, QEMUBHFunc *cb, void *opaque);
 +QEMUBH *aio_bh_new_full(AioContext *ctx, QEMUBHFunc *cb, void *opaque,
 +                        const char *name);
 +
 +/**
 + * aio_bh_new: Allocate a new bottom half structure
 + *
 + * A convenience wrapper for aio_bh_new_full() that uses the cb as the name
 + * string.
 + */
 +#define aio_bh_new(ctx, cb, opaque) \
 +    aio_bh_new_full((ctx), (cb), (opaque), (stringify(cb)))
  /**
   * aio_notify: Force processing of pending events.
 diff --git a/include/qemu/main-loop.h b/include/qemu/main-loop.h
 index XXXXXXX..XXXXXXX 100644
 --- a/include/qemu/main-loop.h
 +++ b/include/qemu/main-loop.h
@@ -XXX,XX +XXX,XX @@ void qemu_cond_timedwait_iothread(QemuCond *cond, int ms);
  void qemu_fd_register(int fd);
 -QEMUBH *qemu_bh_new(QEMUBHFunc *cb, void *opaque);
 +#define qemu_bh_new(cb, opaque) \
 +    qemu_bh_new_full((cb), (opaque), (stringify(cb)))
 +QEMUBH *qemu_bh_new_full(QEMUBHFunc *cb, void *opaque, const char *name);
  void qemu_bh_schedule_idle(QEMUBH *bh);
  enum {
 diff --git a/tests/unit/ptimer-test-stubs.c b/tests/unit/ptimer-test-stubs.c
 index XXXXXXX..XXXXXXX 100644
 --- a/tests/unit/ptimer-test-stubs.c
 +++ b/tests/unit/ptimer-test-stubs.c
@@ -XXX,XX +XXX,XX @@ int64_t qemu_clock_deadline_ns_all(QEMUClockType type, int attr_mask)
      return deadline;
  }
 -QEMUBH *qemu_bh_new(QEMUBHFunc *cb, void *opaque)
 +QEMUBH *qemu_bh_new_full(QEMUBHFunc *cb, void *opaque, const char *name)
  {
-     QEMUBH *bh = g_new(QEMUBH, 1);
+     BDRVQcow2State *s = bs->opaque;
+@@ -XXX,XX +XXX,XX @@ static int rebuild_refcounts_write_refblocks(
-diff --git a/util/async.c b/util/async.c
+                                                   nb_clusters,
-index XXXXXXX..XXXXXXX 100644
+                                                   &first_free_cluster);
---- a/util/async.c
+             if (refblock_offset < 0) {
-+++ b/util/async.c
+-                fprintf(stderr, "ERROR allocating refblock: %s\n",
-@@ -XXX,XX +XXX,XX @@ enum {
+-                        strerror(-refblock_offset));
++                error_setg_errno(errp, -refblock_offset,
- struct QEMUBH {
++                                 "ERROR allocating refblock");
-     AioContext *ctx;
+                 return refblock_offset;
-+    const char *name;
+             }
-     QEMUBHFunc *cb;
-     void *opaque;
+@@ -XXX,XX +XXX,XX @@ static int rebuild_refcounts_write_refblocks(
-     QSLIST_ENTRY(QEMUBH) next;
+                                   on_disk_reftable_entries *
-@@ -XXX,XX +XXX,XX @@ static QEMUBH *aio_bh_dequeue(BHList *head, unsigned *flags)
+                                   REFTABLE_ENTRY_SIZE);
-     return bh;
+                 if (!on_disk_reftable) {
- }
++                    error_setg(errp, "ERROR allocating reftable memory");
+                     return -ENOMEM;
--void aio_bh_schedule_oneshot(AioContext *ctx, QEMUBHFunc *cb, void *opaque)
+                 }
-+void aio_bh_schedule_oneshot_full(AioContext *ctx, QEMUBHFunc *cb,
-+                                  void *opaque, const char *name)
+@@ -XXX,XX +XXX,XX @@ static int rebuild_refcounts_write_refblocks(
          ret = qcow2_pre_write_overlap_check(bs, 0, refblock_offset,
                                              s->cluster_size, false);
          if (ret < 0) {
 -            fprintf(stderr, "ERROR writing refblock: %s\n", strerror(-ret));
 +            error_setg_errno(errp, -ret, "ERROR writing refblock");
              return ret;
          }
@@ -XXX,XX +XXX,XX @@ static int rebuild_refcounts_write_refblocks(
          ret = bdrv_pwrite(bs->file, refblock_offset, on_disk_refblock,
                            s->cluster_size);
          if (ret < 0) {
 -            fprintf(stderr, "ERROR writing refblock: %s\n", strerror(-ret));
 +            error_setg_errno(errp, -ret, "ERROR writing refblock");
              return ret;
          }
@@ -XXX,XX +XXX,XX @@ static int rebuild_refcounts_write_refblocks(
  static int rebuild_refcount_structure(BlockDriverState *bs,
                                        BdrvCheckResult *res,
                                        void **refcount_table,
 -                                      int64_t *nb_clusters)
 +                                      int64_t *nb_clusters,
 +                                      Error **errp)
  {
-     QEMUBH *bh;
+     BDRVQcow2State *s = bs->opaque;
-     bh = g_new(QEMUBH, 1);
+     int64_t reftable_offset = -1;
-@@ -XXX,XX +XXX,XX @@ void aio_bh_schedule_oneshot(AioContext *ctx, QEMUBHFunc *cb, void *opaque)
+@@ -XXX,XX +XXX,XX @@ static int rebuild_refcount_structure(BlockDriverState *bs,
-         .ctx = ctx,
+         rebuild_refcounts_write_refblocks(bs, refcount_table, nb_clusters,
-         .cb = cb,
+, *nb_clusters,
-         .opaque = opaque,
+                                           &on_disk_reftable,
-+        .name = name,
+-                                          &on_disk_reftable_entries);
-     };
++                                          &on_disk_reftable_entries, errp);
-     aio_bh_enqueue(bh, BH_SCHEDULED | BH_ONESHOT);
+     if (reftable_size_changed < 0) {
- }
+         res->check_errors++;
+         ret = reftable_size_changed;
--QEMUBH *aio_bh_new(AioContext *ctx, QEMUBHFunc *cb, void *opaque)
+@@ -XXX,XX +XXX,XX @@ static int rebuild_refcount_structure(BlockDriverState *bs,
-+QEMUBH *aio_bh_new_full(AioContext *ctx, QEMUBHFunc *cb, void *opaque,
+                                               refcount_table, nb_clusters,
-+                        const char *name)
+                                               &first_free_cluster);
- {
+         if (reftable_offset < 0) {
-     QEMUBH *bh;
+-            fprintf(stderr, "ERROR allocating reftable: %s\n",
-     bh = g_new(QEMUBH, 1);
+-                    strerror(-reftable_offset));
-@@ -XXX,XX +XXX,XX @@ QEMUBH *aio_bh_new(AioContext *ctx, QEMUBHFunc *cb, void *opaque)
++            error_setg_errno(errp, -reftable_offset,
-         .ctx = ctx,
++                             "ERROR allocating reftable");
-         .cb = cb,
+             res->check_errors++;
-         .opaque = opaque,
+             ret = reftable_offset;
-+        .name = name,
+             goto fail;
-     };
+@@ -XXX,XX +XXX,XX @@ static int rebuild_refcount_structure(BlockDriverState *bs,
-     return bh;
+                                               reftable_start_cluster,
- }
+                                               reftable_end_cluster,
-diff --git a/util/main-loop.c b/util/main-loop.c
+                                               &on_disk_reftable,
-index XXXXXXX..XXXXXXX 100644
+-                                              &on_disk_reftable_entries);
---- a/util/main-loop.c
++                                              &on_disk_reftable_entries, errp);
-+++ b/util/main-loop.c
+         if (reftable_size_changed < 0) {
-@@ -XXX,XX +XXX,XX @@ void main_loop_wait(int nonblocking)
+             res->check_errors++;
+             ret = reftable_size_changed;
- /* Functions to operate on the main QEMU AioContext.  */
+@@ -XXX,XX +XXX,XX @@ static int rebuild_refcount_structure(BlockDriverState *bs,
+     ret = qcow2_pre_write_overlap_check(bs, 0, reftable_offset, reftable_length,
--QEMUBH *qemu_bh_new(QEMUBHFunc *cb, void *opaque)
+                                         false);
-+QEMUBH *qemu_bh_new_full(QEMUBHFunc *cb, void *opaque, const char *name)
+     if (ret < 0) {
- {
+-        fprintf(stderr, "ERROR writing reftable: %s\n", strerror(-ret));
--    return aio_bh_new(qemu_aio_context, cb, opaque);
++        error_setg_errno(errp, -ret, "ERROR writing reftable");
-+    return aio_bh_new_full(qemu_aio_context, cb, opaque, name);
+         goto fail;
- }
+     }
- /*
+@@ -XXX,XX +XXX,XX @@ static int rebuild_refcount_structure(BlockDriverState *bs,
      ret = bdrv_pwrite(bs->file, reftable_offset, on_disk_reftable,
                        reftable_length);
      if (ret < 0) {
 -        fprintf(stderr, "ERROR writing reftable: %s\n", strerror(-ret));
 +        error_setg_errno(errp, -ret, "ERROR writing reftable");
          goto fail;
      }
@@ -XXX,XX +XXX,XX @@ static int rebuild_refcount_structure(BlockDriverState *bs,
                             &reftable_offset_and_clusters,
                             sizeof(reftable_offset_and_clusters));
      if (ret < 0) {
 -        fprintf(stderr, "ERROR setting reftable: %s\n", strerror(-ret));
 +        error_setg_errno(errp, -ret, "ERROR setting reftable");
          goto fail;
      }
@@ -XXX,XX +XXX,XX @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
      if (rebuild && (fix & BDRV_FIX_ERRORS)) {
          BdrvCheckResult old_res = *res;
          int fresh_leaks = 0;
 +        Error *local_err = NULL;
          fprintf(stderr, "Rebuilding refcount structure\n");
          ret = rebuild_refcount_structure(bs, res, &refcount_table,
 -                                         &nb_clusters);
 +                                         &nb_clusters, &local_err);
          if (ret < 0) {
 +            error_report_err(local_err);
              goto fail;
          }
 --
-.31.1
+.35.1

The following changes since commit 711c0418c8c1ce3a24346f058b001c4c5a2f0f81:

Merge remote-tracking branch 'remotes/philmd/tags/mips-20210702' into staging (2021-07-04 14:04:12 +0100)

are available in the Git repository at:

https://gitlab.com/stefanha/qemu.git tags/block-pull-request

for you to fetch changes up to 9f460c64e13897117f35ffb61f6f5e0102cabc70:

block/io: Merge discard request alignments (2021-07-06 14:28:55 +0100)

----------------------------------------------------------------
Pull request

----------------------------------------------------------------

Akihiko Odaki (3):
  block/file-posix: Optimize for macOS
  block: Add backend_defaults property
  block/io: Merge discard request alignments

Stefan Hajnoczi (2):
  util/async: add a human-readable name to BHs for debugging
  util/async: print leaked BH name when AioContext finalizes

-- 
2.31.1

It can be difficult to debug issues with BHs in production environments.
Although BHs can usually be identified by looking up their ->cb()
function pointer, this requires debug information for the program. It is
also not possible to print human-readable diagnostics about BHs because
they have no identifier.

This patch adds a name to each BH. The name is not unique per instance
but differentiates between cb() functions, which is usually enough. It's
done by changing aio_bh_new() and friends to macros that stringify cb.

The next patch will use the name field when reporting leaked BHs.

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210414200247.917496-2-stefanha@redhat.com>
---
 include/block/aio.h            | 31 ++++++++++++++++++++++++++++---
 include/qemu/main-loop.h       |  4 +++-
 tests/unit/ptimer-test-stubs.c |  2 +-
 util/async.c                   |  9 +++++++--
 util/main-loop.c               |  4 ++--
 5 files changed, 41 insertions(+), 9 deletions(-)

diff --git a/include/block/aio.h b/include/block/aio.h
index XXXXXXX..XXXXXXX 100644
--- a/include/block/aio.h
+++ b/include/block/aio.h
@@ -XXX,XX +XXX,XX @@ void aio_context_acquire(AioContext *ctx);
 /* Relinquish ownership of the AioContext. */
 void aio_context_release(AioContext *ctx);
 
+/**
+ * aio_bh_schedule_oneshot_full: Allocate a new bottom half structure that will
+ * run only once and as soon as possible.
+ *
+ * @name: A human-readable identifier for debugging purposes.
+ */
+void aio_bh_schedule_oneshot_full(AioContext *ctx, QEMUBHFunc *cb, void *opaque,
+                                  const char *name);
+
 /**
  * aio_bh_schedule_oneshot: Allocate a new bottom half structure that will run
  * only once and as soon as possible.
+ *
+ * A convenience wrapper for aio_bh_schedule_oneshot_full() that uses cb as the
+ * name string.
  */
-void aio_bh_schedule_oneshot(AioContext *ctx, QEMUBHFunc *cb, void *opaque);
+#define aio_bh_schedule_oneshot(ctx, cb, opaque) \
+    aio_bh_schedule_oneshot_full((ctx), (cb), (opaque), (stringify(cb)))
 
 /**
- * aio_bh_new: Allocate a new bottom half structure.
+ * aio_bh_new_full: Allocate a new bottom half structure.
  *
  * Bottom halves are lightweight callbacks whose invocation is guaranteed
  * to be wait-free, thread-safe and signal-safe.  The #QEMUBH structure
  * is opaque and must be allocated prior to its use.
+ *
+ * @name: A human-readable identifier for debugging purposes.
  */
-QEMUBH *aio_bh_new(AioContext *ctx, QEMUBHFunc *cb, void *opaque);
+QEMUBH *aio_bh_new_full(AioContext *ctx, QEMUBHFunc *cb, void *opaque,
+                        const char *name);
+
+/**
+ * aio_bh_new: Allocate a new bottom half structure
+ *
+ * A convenience wrapper for aio_bh_new_full() that uses the cb as the name
+ * string.
+ */
+#define aio_bh_new(ctx, cb, opaque) \
+    aio_bh_new_full((ctx), (cb), (opaque), (stringify(cb)))
 
 /**
  * aio_notify: Force processing of pending events.
diff --git a/include/qemu/main-loop.h b/include/qemu/main-loop.h
index XXXXXXX..XXXXXXX 100644
--- a/include/qemu/main-loop.h
+++ b/include/qemu/main-loop.h
@@ -XXX,XX +XXX,XX @@ void qemu_cond_timedwait_iothread(QemuCond *cond, int ms);
 
 void qemu_fd_register(int fd);
 
-QEMUBH *qemu_bh_new(QEMUBHFunc *cb, void *opaque);
+#define qemu_bh_new(cb, opaque) \
+    qemu_bh_new_full((cb), (opaque), (stringify(cb)))
+QEMUBH *qemu_bh_new_full(QEMUBHFunc *cb, void *opaque, const char *name);
 void qemu_bh_schedule_idle(QEMUBH *bh);
 
 enum {
diff --git a/tests/unit/ptimer-test-stubs.c b/tests/unit/ptimer-test-stubs.c
index XXXXXXX..XXXXXXX 100644
--- a/tests/unit/ptimer-test-stubs.c
+++ b/tests/unit/ptimer-test-stubs.c
@@ -XXX,XX +XXX,XX @@ int64_t qemu_clock_deadline_ns_all(QEMUClockType type, int attr_mask)
     return deadline;
 }
 
-QEMUBH *qemu_bh_new(QEMUBHFunc *cb, void *opaque)
+QEMUBH *qemu_bh_new_full(QEMUBHFunc *cb, void *opaque, const char *name)
 {
     QEMUBH *bh = g_new(QEMUBH, 1);
 
diff --git a/util/async.c b/util/async.c
index XXXXXXX..XXXXXXX 100644
--- a/util/async.c
+++ b/util/async.c
@@ -XXX,XX +XXX,XX @@ enum {
 
 struct QEMUBH {
     AioContext *ctx;
+    const char *name;
     QEMUBHFunc *cb;
     void *opaque;
     QSLIST_ENTRY(QEMUBH) next;
@@ -XXX,XX +XXX,XX @@ static QEMUBH *aio_bh_dequeue(BHList *head, unsigned *flags)
     return bh;
 }
 
-void aio_bh_schedule_oneshot(AioContext *ctx, QEMUBHFunc *cb, void *opaque)
+void aio_bh_schedule_oneshot_full(AioContext *ctx, QEMUBHFunc *cb,
+                                  void *opaque, const char *name)
 {
     QEMUBH *bh;
     bh = g_new(QEMUBH, 1);
@@ -XXX,XX +XXX,XX @@ void aio_bh_schedule_oneshot(AioContext *ctx, QEMUBHFunc *cb, void *opaque)
         .ctx = ctx,
         .cb = cb,
         .opaque = opaque,
+        .name = name,
     };
     aio_bh_enqueue(bh, BH_SCHEDULED | BH_ONESHOT);
 }
 
-QEMUBH *aio_bh_new(AioContext *ctx, QEMUBHFunc *cb, void *opaque)
+QEMUBH *aio_bh_new_full(AioContext *ctx, QEMUBHFunc *cb, void *opaque,
+                        const char *name)
 {
     QEMUBH *bh;
     bh = g_new(QEMUBH, 1);
@@ -XXX,XX +XXX,XX @@ QEMUBH *aio_bh_new(AioContext *ctx, QEMUBHFunc *cb, void *opaque)
         .ctx = ctx,
         .cb = cb,
         .opaque = opaque,
+        .name = name,
     };
     return bh;
 }
diff --git a/util/main-loop.c b/util/main-loop.c
index XXXXXXX..XXXXXXX 100644
--- a/util/main-loop.c
+++ b/util/main-loop.c
@@ -XXX,XX +XXX,XX @@ void main_loop_wait(int nonblocking)
 
 /* Functions to operate on the main QEMU AioContext.  */
 
-QEMUBH *qemu_bh_new(QEMUBHFunc *cb, void *opaque)
+QEMUBH *qemu_bh_new_full(QEMUBHFunc *cb, void *opaque, const char *name)
 {
-    return aio_bh_new(qemu_aio_context, cb, opaque);
+    return aio_bh_new_full(qemu_aio_context, cb, opaque, name);
 }
 
 /*
-- 
2.31.1

BHs must be deleted before the AioContext is finalized. If not, it's a
bug and probably indicates that some part of the program still expects
the BH to run in the future. That can lead to memory leaks, inconsistent
state, or just hangs.

Unfortunately the assert(flags & BH_DELETED) call in aio_ctx_finalize()
is difficult to debug because the assertion failure contains no
information about the BH!

Use the QEMUBH name field added in the previous patch to show a useful
error when a leaked BH is detected.

Suggested-by: Eric Ernst <eric.g.ernst@gmail.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20210414200247.917496-3-stefanha@redhat.com>
---
 util/async.c | 16 ++++++++++++++--
 1 file changed, 14 insertions(+), 2 deletions(-)

diff --git a/util/async.c b/util/async.c
index XXXXXXX..XXXXXXX 100644
--- a/util/async.c
+++ b/util/async.c
@@ -XXX,XX +XXX,XX @@ aio_ctx_finalize(GSource     *source)
     assert(QSIMPLEQ_EMPTY(&ctx->bh_slice_list));
 
     while ((bh = aio_bh_dequeue(&ctx->bh_list, &flags))) {
-        /* qemu_bh_delete() must have been called on BHs in this AioContext */
-        assert(flags & BH_DELETED);
+        /*
+         * qemu_bh_delete() must have been called on BHs in this AioContext. In
+         * many cases memory leaks, hangs, or inconsistent state occur when a
+         * BH is leaked because something still expects it to run.
+         *
+         * If you hit this, fix the lifecycle of the BH so that
+         * qemu_bh_delete() and any associated cleanup is called before the
+         * AioContext is finalized.
+         */
+        if (unlikely(!(flags & BH_DELETED))) {
+            fprintf(stderr, "%s: BH '%s' leaked, aborting...\n",
+                    __func__, bh->name);
+            abort();
+        }
 
         g_free(bh);
     }
-- 
2.31.1

From: Akihiko Odaki <akihiko.odaki@gmail.com>

This commit introduces "punch hole" operation and optimizes transfer
block size for macOS.

Thanks to Konstantin Nazarov for detailed analysis of a flaw in an
old version of this change:
https://gist.github.com/akihikodaki/87df4149e7ca87f18dc56807ec5a1bc5#gistcomment-3654667

Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Message-id: 20210705130458.97642-1-akihiko.odaki@gmail.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 block/file-posix.c | 27 +++++++++++++++++++++++++--
 1 file changed, 25 insertions(+), 2 deletions(-)

diff --git a/block/file-posix.c b/block/file-posix.c
index XXXXXXX..XXXXXXX 100644
--- a/block/file-posix.c
+++ b/block/file-posix.c
@@ -XXX,XX +XXX,XX @@
 #if defined(HAVE_HOST_BLOCK_DEVICE)
 #include <paths.h>
 #include <sys/param.h>
+#include <sys/mount.h>
 #include <IOKit/IOKitLib.h>
 #include <IOKit/IOBSD.h>
 #include <IOKit/storage/IOMediaBSDClient.h>
@@ -XXX,XX +XXX,XX @@ static void raw_refresh_limits(BlockDriverState *bs, Error **errp)
         return;
     }
 
+#if defined(__APPLE__) && (__MACH__)
+    struct statfs buf;
+
+    if (!fstatfs(s->fd, &buf)) {
+        bs->bl.opt_transfer = buf.f_iosize;
+        bs->bl.pdiscard_alignment = buf.f_bsize;
+    }
+#endif
+
     if (bs->sg || S_ISBLK(st.st_mode)) {
         int ret = hdev_get_max_hw_transfer(s->fd, &st);
 
@@ -XXX,XX +XXX,XX @@ out:
     }
 }
 
+#if defined(CONFIG_FALLOCATE) || defined(BLKZEROOUT) || defined(BLKDISCARD)
 static int translate_err(int err)
 {
     if (err == -ENODEV || err == -ENOSYS || err == -EOPNOTSUPP ||
@@ -XXX,XX +XXX,XX @@ static int translate_err(int err)
     }
     return err;
 }
+#endif
 
 #ifdef CONFIG_FALLOCATE
 static int do_fallocate(int fd, int mode, off_t offset, off_t len)
@@ -XXX,XX +XXX,XX @@ static int handle_aiocb_discard(void *opaque)
             }
         } while (errno == EINTR);
 
-        ret = -errno;
+        ret = translate_err(-errno);
 #endif
     } else {
 #ifdef CONFIG_FALLOCATE_PUNCH_HOLE
         ret = do_fallocate(s->fd, FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
                            aiocb->aio_offset, aiocb->aio_nbytes);
+        ret = translate_err(-errno);
+#elif defined(__APPLE__) && (__MACH__)
+        fpunchhole_t fpunchhole;
+        fpunchhole.fp_flags = 0;
+        fpunchhole.reserved = 0;
+        fpunchhole.fp_offset = aiocb->aio_offset;
+        fpunchhole.fp_length = aiocb->aio_nbytes;
+        if (fcntl(s->fd, F_PUNCHHOLE, &fpunchhole) == -1) {
+            ret = errno == ENODEV ? -ENOTSUP : -errno;
+        } else {
+            ret = 0;
+        }
 #endif
     }
 
-    ret = translate_err(ret);
     if (ret == -ENOTSUP) {
         s->has_discard = false;
     }
-- 
2.31.1

From: Akihiko Odaki <akihiko.odaki@gmail.com>

backend_defaults property allow users to control if default block
properties should be decided with backend information.

If it is off, any backend information will be discarded, which is
suitable if you plan to perform live migration to a different disk backend.

If it is on, a block device may utilize backend information more
aggressively.

By default, it is auto, which uses backend information for block
sizes and ignores the others, which is consistent with the older
versions.

Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Message-id: 20210705130458.97642-2-akihiko.odaki@gmail.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 include/hw/block/block.h   |  3 +++
 hw/block/block.c           | 42 ++++++++++++++++++++++++++++++++++----
 tests/qemu-iotests/172.out | 38 ++++++++++++++++++++++++++++++++++
 3 files changed, 79 insertions(+), 4 deletions(-)

diff --git a/include/hw/block/block.h b/include/hw/block/block.h
index XXXXXXX..XXXXXXX 100644
--- a/include/hw/block/block.h
+++ b/include/hw/block/block.h
@@ -XXX,XX +XXX,XX @@
 
 typedef struct BlockConf {
     BlockBackend *blk;
+    OnOffAuto backend_defaults;
     uint32_t physical_block_size;
     uint32_t logical_block_size;
     uint32_t min_io_size;
@@ -XXX,XX +XXX,XX @@ static inline unsigned int get_physical_block_exp(BlockConf *conf)
 }
 
 #define DEFINE_BLOCK_PROPERTIES_BASE(_state, _conf)                     \
+    DEFINE_PROP_ON_OFF_AUTO("backend_defaults", _state,                 \
+                            _conf.backend_defaults, ON_OFF_AUTO_AUTO),  \
     DEFINE_PROP_BLOCKSIZE("logical_block_size", _state,                 \
                           _conf.logical_block_size),                    \
     DEFINE_PROP_BLOCKSIZE("physical_block_size", _state,                \
diff --git a/hw/block/block.c b/hw/block/block.c
index XXXXXXX..XXXXXXX 100644
--- a/hw/block/block.c
+++ b/hw/block/block.c
@@ -XXX,XX +XXX,XX @@ bool blkconf_blocksizes(BlockConf *conf, Error **errp)
 {
     BlockBackend *blk = conf->blk;
     BlockSizes blocksizes;
-    int backend_ret;
+    BlockDriverState *bs;
+    bool use_blocksizes;
+    bool use_bs;
+
+    switch (conf->backend_defaults) {
+    case ON_OFF_AUTO_AUTO:
+        use_blocksizes = !blk_probe_blocksizes(blk, &blocksizes);
+        use_bs = false;
+        break;
+
+    case ON_OFF_AUTO_ON:
+        use_blocksizes = !blk_probe_blocksizes(blk, &blocksizes);
+        bs = blk_bs(blk);
+        use_bs = bs;
+        break;
+
+    case ON_OFF_AUTO_OFF:
+        use_blocksizes = false;
+        use_bs = false;
+        break;
+
+    default:
+        abort();
+    }
 
-    backend_ret = blk_probe_blocksizes(blk, &blocksizes);
     /* fill in detected values if they are not defined via qemu command line */
     if (!conf->physical_block_size) {
-        if (!backend_ret) {
+        if (use_blocksizes) {
            conf->physical_block_size = blocksizes.phys;
         } else {
             conf->physical_block_size = BDRV_SECTOR_SIZE;
         }
     }
     if (!conf->logical_block_size) {
-        if (!backend_ret) {
+        if (use_blocksizes) {
             conf->logical_block_size = blocksizes.log;
         } else {
             conf->logical_block_size = BDRV_SECTOR_SIZE;
         }
     }
+    if (use_bs) {
+        if (!conf->opt_io_size) {
+            conf->opt_io_size = bs->bl.opt_transfer;
+        }
+        if (conf->discard_granularity == -1) {
+            if (bs->bl.pdiscard_alignment) {
+                conf->discard_granularity = bs->bl.pdiscard_alignment;
+            } else if (bs->bl.request_alignment != 1) {
+                conf->discard_granularity = bs->bl.request_alignment;
+            }
+        }
+    }
 
     if (conf->logical_block_size > conf->physical_block_size) {
         error_setg(errp,
diff --git a/tests/qemu-iotests/172.out b/tests/qemu-iotests/172.out
index XXXXXXX..XXXXXXX 100644
--- a/tests/qemu-iotests/172.out
+++ b/tests/qemu-iotests/172.out
@@ -XXX,XX +XXX,XX @@ Testing:
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = "floppy0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fda TEST_DIR/t.qcow2
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = "floppy0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fdb TEST_DIR/t.qcow2
               dev: floppy, id ""
                 unit = 1 (0x1)
                 drive = "floppy1"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fdb TEST_DIR/t.qcow2
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = "floppy0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fda TEST_DIR/t.qcow2 -fdb TEST_DIR/t.qcow2.2
               dev: floppy, id ""
                 unit = 1 (0x1)
                 drive = "floppy1"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fda TEST_DIR/t.qcow2 -fdb TEST_DIR/t.qcow2.2
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = "floppy0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fdb
               dev: floppy, id ""
                 unit = 1 (0x1)
                 drive = "floppy1"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fdb
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = "floppy0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=floppy,file=TEST_DIR/t.qcow2
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = "floppy0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=floppy,file=TEST_DIR/t.qcow2,index=1
               dev: floppy, id ""
                 unit = 1 (0x1)
                 drive = "floppy1"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=floppy,file=TEST_DIR/t.qcow2,index=1
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = "floppy0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=floppy,file=TEST_DIR/t.qcow2 -drive if=floppy,file=TEST_DIR/t
               dev: floppy, id ""
                 unit = 1 (0x1)
                 drive = "floppy1"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=floppy,file=TEST_DIR/t.qcow2 -drive if=floppy,file=TEST_DIR/t
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = "floppy0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=none,file=TEST_DIR/t.qcow2 -device floppy,drive=none0
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = "none0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=none,file=TEST_DIR/t.qcow2 -device floppy,drive=none0,unit=1
               dev: floppy, id ""
                 unit = 1 (0x1)
                 drive = "none0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=none,file=TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.qco
               dev: floppy, id ""
                 unit = 1 (0x1)
                 drive = "none1"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=none,file=TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.qco
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = "none0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fda TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.qcow2.2 -device fl
               dev: floppy, id ""
                 unit = 1 (0x1)
                 drive = "none0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fda TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.qcow2.2 -device fl
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = "floppy0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fda TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.qcow2.2 -device fl
               dev: floppy, id ""
                 unit = 1 (0x1)
                 drive = "none0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fda TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.qcow2.2 -device fl
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = "floppy0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fdb TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.qcow2.2 -device fl
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = "none0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fdb TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.qcow2.2 -device fl
               dev: floppy, id ""
                 unit = 1 (0x1)
                 drive = "floppy1"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fdb TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.qcow2.2 -device fl
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = "none0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -fdb TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.qcow2.2 -device fl
               dev: floppy, id ""
                 unit = 1 (0x1)
                 drive = "floppy1"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=floppy,file=TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.q
               dev: floppy, id ""
                 unit = 1 (0x1)
                 drive = "none0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=floppy,file=TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.q
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = "floppy0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=floppy,file=TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.q
               dev: floppy, id ""
                 unit = 1 (0x1)
                 drive = "none0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=floppy,file=TEST_DIR/t.qcow2 -drive if=none,file=TEST_DIR/t.q
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = "floppy0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=none,file=TEST_DIR/t.qcow2 -global floppy.drive=none0 -device
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = "none0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -device floppy
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = ""
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -device floppy,drive-type=120
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = ""
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -device floppy,drive-type=144
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = ""
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -device floppy,drive-type=288
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = ""
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=none,file=TEST_DIR/t.qcow2 -device floppy,drive=none0,drive-t
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = "none0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=none,file=TEST_DIR/t.qcow2 -device floppy,drive=none0,drive-t
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = "none0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=none,file=TEST_DIR/t.qcow2 -device floppy,drive=none0,logical
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = "none0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
@@ -XXX,XX +XXX,XX @@ Testing: -drive if=none,file=TEST_DIR/t.qcow2 -device floppy,drive=none0,physica
               dev: floppy, id ""
                 unit = 0 (0x0)
                 drive = "none0"
+                backend_defaults = "auto"
                 logical_block_size = 512 (512 B)
                 physical_block_size = 512 (512 B)
                 min_io_size = 0 (0 B)
-- 
2.31.1

The following changes since commit 1be5a765c08cee3a9587c8a8d3fc2ea247b13f9c:

Merge tag 'for-upstream' of https://gitlab.com/bonzini/qemu into staging (2022-04-19 18:22:16 -0700)

are available in the Git repository at:

https://gitlab.com/hreitz/qemu.git tags/pull-block-2022-04-20

for you to fetch changes up to 0423f75351ab83b844a31349218b0eadd830e07a:

qcow2: Add errp to rebuild_refcount_structure() (2022-04-20 12:09:17 +0200)

----------------------------------------------------------------
Block patches:
- Some changes for qcow2's refcount repair algorithm to make it work for
  qcow2 images stored on block devices
- Skip test cases that require zstd when support for it is missing
- Some refactoring in the iotests' meson.build

----------------------------------------------------------------
Hanna Reitz (6):
  iotests.py: Add supports_qcow2_zstd_compression()
  iotests/065: Check for zstd support
  iotests/303: Check for zstd support
  qcow2: Improve refcount structure rebuilding
  iotests/108: Test new refcount rebuild algorithm
  qcow2: Add errp to rebuild_refcount_structure()

Thomas Huth (2):
  tests/qemu-iotests/meson.build: Improve the indentation
  tests/qemu-iotests: Move the bash and sanitizer checks to meson.build

-- 
2.35.1

From: Thomas Huth <thuth@redhat.com>

By using subdir_done(), we can get rid of one level of indentation
in this file. This will make it easier to add more conditions to
skip the iotests in future patches.

Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20220223093840.2515281-3-thuth@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
---
 tests/qemu-iotests/meson.build | 61 ++++++++++++++++++----------------
 1 file changed, 32 insertions(+), 29 deletions(-)

diff --git a/tests/qemu-iotests/meson.build b/tests/qemu-iotests/meson.build
index XXXXXXX..XXXXXXX 100644
--- a/tests/qemu-iotests/meson.build
+++ b/tests/qemu-iotests/meson.build
@@ -XXX,XX +XXX,XX @@
-if have_tools and targetos != 'windows' and not get_option('gprof')
-  qemu_iotests_binaries = [qemu_img, qemu_io, qemu_nbd, qsd]
-  qemu_iotests_env = {'PYTHON': python.full_path()}
-  qemu_iotests_formats = {
-    'qcow2': 'quick',
-    'raw': 'slow',
-    'qed': 'thorough',
-    'vmdk': 'thorough',
-    'vpc': 'thorough'
-  }
-
-  foreach k, v : emulators
-    if k.startswith('qemu-system-')
-      qemu_iotests_binaries += v
-    endif
-  endforeach
-  foreach format, speed: qemu_iotests_formats
-    if speed == 'quick'
-      suites = 'block'
-    else
-      suites = ['block-' + speed, speed]
-    endif
-    test('qemu-iotests ' + format, sh, args: [files('../check-block.sh'), format],
-         depends: qemu_iotests_binaries, env: qemu_iotests_env,
-         protocol: 'tap',
-         suite: suites,
-         timeout: 0,
-         is_parallel: false)
-  endforeach
+if not have_tools or targetos == 'windows' or get_option('gprof')
+  subdir_done()
 endif
+
+qemu_iotests_binaries = [qemu_img, qemu_io, qemu_nbd, qsd]
+qemu_iotests_env = {'PYTHON': python.full_path()}
+qemu_iotests_formats = {
+  'qcow2': 'quick',
+  'raw': 'slow',
+  'qed': 'thorough',
+  'vmdk': 'thorough',
+  'vpc': 'thorough'
+}
+
+foreach k, v : emulators
+  if k.startswith('qemu-system-')
+    qemu_iotests_binaries += v
+  endif
+endforeach
+
+foreach format, speed: qemu_iotests_formats
+  if speed == 'quick'
+    suites = 'block'
+  else
+    suites = ['block-' + speed, speed]
+  endif
+  test('qemu-iotests ' + format, sh, args: [files('../check-block.sh'), format],
+       depends: qemu_iotests_binaries, env: qemu_iotests_env,
+       protocol: 'tap',
+       suite: suites,
+       timeout: 0,
+       is_parallel: false)
+endforeach
-- 
2.35.1

From: Thomas Huth <thuth@redhat.com>

We want to get rid of check-block.sh in the long run, so let's move
the checks for the bash version and sanitizers from check-block.sh
into the meson.build file instead.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20220223093840.2515281-4-thuth@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
---
 tests/check-block.sh           | 26 --------------------------
 tests/qemu-iotests/meson.build | 14 ++++++++++++++
 2 files changed, 14 insertions(+), 26 deletions(-)

diff --git a/tests/check-block.sh b/tests/check-block.sh
index XXXXXXX..XXXXXXX 100755
--- a/tests/check-block.sh
+++ b/tests/check-block.sh
@@ -XXX,XX +XXX,XX @@ skip() {
     exit 0
 }
 
-# Disable tests with any sanitizer except for specific ones
-SANITIZE_FLAGS=$( grep "CFLAGS.*-fsanitize" config-host.mak 2>/dev/null )
-ALLOWED_SANITIZE_FLAGS="safe-stack cfi-icall"
-#Remove all occurrencies of allowed Sanitize flags
-for j in ${ALLOWED_SANITIZE_FLAGS}; do
-    TMP_FLAGS=${SANITIZE_FLAGS}
-    SANITIZE_FLAGS=""
-    for i in ${TMP_FLAGS}; do
-        if ! echo ${i} | grep -q "${j}" 2>/dev/null; then
-            SANITIZE_FLAGS="${SANITIZE_FLAGS} ${i}"
-        fi
-    done
-done
-if echo ${SANITIZE_FLAGS} | grep -q "\-fsanitize" 2>/dev/null; then
-    # Have a sanitize flag that is not allowed, stop
-    skip "Sanitizers are enabled ==> Not running the qemu-iotests."
-fi
-
 if [ -z "$(find . -name 'qemu-system-*' -print)" ]; then
     skip "No qemu-system binary available ==> Not running the qemu-iotests."
 fi
 
-if ! command -v bash >/dev/null 2>&1 ; then
-    skip "bash not available ==> Not running the qemu-iotests."
-fi
-
-if LANG=C bash --version | grep -q 'GNU bash, version [123]' ; then
-    skip "bash version too old ==> Not running the qemu-iotests."
-fi
-
 cd tests/qemu-iotests
 
 # QEMU_CHECK_BLOCK_AUTO is used to disable some unstable sub-tests
diff --git a/tests/qemu-iotests/meson.build b/tests/qemu-iotests/meson.build
index XXXXXXX..XXXXXXX 100644
--- a/tests/qemu-iotests/meson.build
+++ b/tests/qemu-iotests/meson.build
@@ -XXX,XX +XXX,XX @@ if not have_tools or targetos == 'windows' or get_option('gprof')
   subdir_done()
 endif
 
+foreach cflag: config_host['QEMU_CFLAGS'].split()
+  if cflag.startswith('-fsanitize') and \
+     not cflag.contains('safe-stack') and not cflag.contains('cfi-icall')
+    message('Sanitizers are enabled ==> Disabled the qemu-iotests.')
+    subdir_done()
+  endif
+endforeach
+
+bash = find_program('bash', required: false, version: '>= 4.0')
+if not bash.found()
+  message('bash >= v4.0 not available ==> Disabled the qemu-iotests.')
+  subdir_done()
+endif
+
 qemu_iotests_binaries = [qemu_img, qemu_io, qemu_nbd, qsd]
 qemu_iotests_env = {'PYTHON': python.full_path()}
 qemu_iotests_formats = {
-- 
2.35.1

Suggested-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20220323105522.53660-2-hreitz@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <v.sementsov-og@mail.ru>
---
 tests/qemu-iotests/iotests.py | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)

diff --git a/tests/qemu-iotests/iotests.py b/tests/qemu-iotests/iotests.py
index XXXXXXX..XXXXXXX 100644
--- a/tests/qemu-iotests/iotests.py
+++ b/tests/qemu-iotests/iotests.py
@@ -XXX,XX +XXX,XX @@ def verify_working_luks():
     if not working:
         notrun(reason)
 
+def supports_qcow2_zstd_compression() -> bool:
+    img_file = f'{test_dir}/qcow2-zstd-test.qcow2'
+    res = qemu_img('create', '-f', 'qcow2', '-o', 'compression_type=zstd',
+                   img_file, '0',
+                   check=False)
+    try:
+        os.remove(img_file)
+    except OSError:
+        pass
+
+    if res.returncode == 1 and \
+            "'compression-type' does not accept value 'zstd'" in res.stdout:
+        return False
+    else:
+        return True
+
+def verify_qcow2_zstd_compression():
+    if not supports_qcow2_zstd_compression():
+        notrun('zstd compression not supported')
+
 def qemu_pipe(*args: str) -> str:
     """
     Run qemu with an option to print something and exit (e.g. a help option).
-- 
2.35.1

Some test cases run in iotest 065 want to run with zstd compression just
for added coverage.  Run them with zlib if there is no zstd support
compiled in.

Reported-by: Thomas Huth <thuth@redhat.com>
Fixes: 12a936171d71f839dc907ff ("iotest 065: explicit compression type")
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20220323105522.53660-3-hreitz@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <v.sementsov-og@mail.ru>
---
 tests/qemu-iotests/065 | 24 ++++++++++++++++++------
 1 file changed, 18 insertions(+), 6 deletions(-)

diff --git a/tests/qemu-iotests/065 b/tests/qemu-iotests/065
index XXXXXXX..XXXXXXX 100755
--- a/tests/qemu-iotests/065
+++ b/tests/qemu-iotests/065
@@ -XXX,XX +XXX,XX @@ import os
 import re
 import json
 import iotests
-from iotests import qemu_img, qemu_img_info
+from iotests import qemu_img, qemu_img_info, supports_qcow2_zstd_compression
 import unittest
 
 test_img = os.path.join(iotests.test_dir, 'test.img')
@@ -XXX,XX +XXX,XX @@ class TestQCow2(TestQemuImgInfo):
 
 class TestQCow3NotLazy(TestQemuImgInfo):
     '''Testing a qcow2 version 3 image with lazy refcounts disabled'''
-    img_options = 'compat=1.1,lazy_refcounts=off,compression_type=zstd'
+    if supports_qcow2_zstd_compression():
+        compression_type = 'zstd'
+    else:
+        compression_type = 'zlib'
+
+    img_options = 'compat=1.1,lazy_refcounts=off'
+    img_options += f',compression_type={compression_type}'
     json_compare = { 'compat': '1.1', 'lazy-refcounts': False,
                      'refcount-bits': 16, 'corrupt': False,
-                     'compression-type': 'zstd', 'extended-l2': False }
-    human_compare = [ 'compat: 1.1', 'compression type: zstd',
+                     'compression-type': compression_type, 'extended-l2': False }
+    human_compare = [ 'compat: 1.1', f'compression type: {compression_type}',
                       'lazy refcounts: false', 'refcount bits: 16',
                       'corrupt: false', 'extended l2: false' ]
 
@@ -XXX,XX +XXX,XX @@ class TestQCow3NotLazyQMP(TestQMP):
 class TestQCow3LazyQMP(TestQMP):
     '''Testing a qcow2 version 3 image with lazy refcounts enabled, opening
        with lazy refcounts disabled'''
-    img_options = 'compat=1.1,lazy_refcounts=on,compression_type=zstd'
+    if supports_qcow2_zstd_compression():
+        compression_type = 'zstd'
+    else:
+        compression_type = 'zlib'
+
+    img_options = 'compat=1.1,lazy_refcounts=on'
+    img_options += f',compression_type={compression_type}'
     qemu_options = 'lazy-refcounts=off'
     compare = { 'compat': '1.1', 'lazy-refcounts': True,
                 'refcount-bits': 16, 'corrupt': False,
-                'compression-type': 'zstd', 'extended-l2': False }
+                'compression-type': compression_type, 'extended-l2': False }
 
 TestImageInfoSpecific = None
 TestQemuImgInfo = None
-- 
2.35.1

303 runs two test cases, one of which requires zstd support.
Unfortunately, given that this is not a unittest-style test, we cannot
easily skip that single case, and instead can only skip the whole test.

(Alternatively, we could split this test into a zlib and a zstd part,
but that seems excessive, given that this test is not in auto and thus
likely only run by developers who have zstd support compiled in.)

Fixes: 677e0bae686e7c670a71d1f ("iotest 303: explicit compression type")
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <v.sementsov-og@mail.ru>
Message-Id: <20220323105522.53660-4-hreitz@redhat.com>
---
 tests/qemu-iotests/303 | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/tests/qemu-iotests/303 b/tests/qemu-iotests/303
index XXXXXXX..XXXXXXX 100755
--- a/tests/qemu-iotests/303
+++ b/tests/qemu-iotests/303
@@ -XXX,XX +XXX,XX @@
 
 import iotests
 import subprocess
-from iotests import qemu_img_create, qemu_io, file_path, log, filter_qemu_io
+from iotests import qemu_img_create, qemu_io, file_path, log, filter_qemu_io, \
+        verify_qcow2_zstd_compression
 
 iotests.script_initialize(supported_fmts=['qcow2'],
                           unsupported_imgopts=['refcount_bits', 'compat'])
+verify_qcow2_zstd_compression()
 
 disk = file_path('disk')
 chunk = 1024 * 1024
-- 
2.35.1

When rebuilding the refcount structures (when qemu-img check -r found
errors with refcount = 0, but reference count > 0), the new refcount
table defaults to being put at the image file end[1].  There is no good
reason for that except that it means we will not have to rewrite any
refblocks we already wrote to disk.

Changing the code to rewrite those refblocks is not too difficult,
though, so let us do that.  That is beneficial for images on block
devices, where we cannot really write beyond the end of the image file.

Use this opportunity to add extensive comments to the code, and refactor
it a bit, getting rid of the backwards-jumping goto.

[1] Unless there is something allocated in the area pointed to by the
    last refblock, so we have to write that refblock.  In that case, we
    try to put the reftable in there.

Buglink: https://bugzilla.redhat.com/show_bug.cgi?id=1519071
Closes: https://gitlab.com/qemu-project/qemu/-/issues/941
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20220405134652.19278-2-hreitz@redhat.com>
---
 block/qcow2-refcount.c | 332 +++++++++++++++++++++++++++++------------
 1 file changed, 235 insertions(+), 97 deletions(-)

diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
index XXXXXXX..XXXXXXX 100644
--- a/block/qcow2-refcount.c
+++ b/block/qcow2-refcount.c
@@ -XXX,XX +XXX,XX @@ static int64_t alloc_clusters_imrt(BlockDriverState *bs,
 }
 
 /*
- * Creates a new refcount structure based solely on the in-memory information
- * given through *refcount_table. All necessary allocations will be reflected
- * in that array.
+ * Helper function for rebuild_refcount_structure().
  *
- * On success, the old refcount structure is leaked (it will be covered by the
- * new refcount structure).
+ * Scan the range of clusters [first_cluster, end_cluster) for allocated
+ * clusters and write all corresponding refblocks to disk.  The refblock
+ * and allocation data is taken from the in-memory refcount table
+ * *refcount_table[] (of size *nb_clusters), which is basically one big
+ * (unlimited size) refblock for the whole image.
+ *
+ * For these refblocks, clusters are allocated using said in-memory
+ * refcount table.  Care is taken that these allocations are reflected
+ * in the refblocks written to disk.
+ *
+ * The refblocks' offsets are written into a reftable, which is
+ * *on_disk_reftable_ptr[] (of size *on_disk_reftable_entries_ptr).  If
+ * that reftable is of insufficient size, it will be resized to fit.
+ * This reftable is not written to disk.
+ *
+ * (If *on_disk_reftable_ptr is not NULL, the entries within are assumed
+ * to point to existing valid refblocks that do not need to be allocated
+ * again.)
+ *
+ * Return whether the on-disk reftable array was resized (true/false),
+ * or -errno on error.
  */
-static int rebuild_refcount_structure(BlockDriverState *bs,
-                                      BdrvCheckResult *res,
-                                      void **refcount_table,
-                                      int64_t *nb_clusters)
+static int rebuild_refcounts_write_refblocks(
+        BlockDriverState *bs, void **refcount_table, int64_t *nb_clusters,
+        int64_t first_cluster, int64_t end_cluster,
+        uint64_t **on_disk_reftable_ptr, uint32_t *on_disk_reftable_entries_ptr
+    )
 {
     BDRVQcow2State *s = bs->opaque;
-    int64_t first_free_cluster = 0, reftable_offset = -1, cluster = 0;
+    int64_t cluster;
     int64_t refblock_offset, refblock_start, refblock_index;
-    uint32_t reftable_size = 0;
-    uint64_t *on_disk_reftable = NULL;
+    int64_t first_free_cluster = 0;
+    uint64_t *on_disk_reftable = *on_disk_reftable_ptr;
+    uint32_t on_disk_reftable_entries = *on_disk_reftable_entries_ptr;
     void *on_disk_refblock;
-    int ret = 0;
-    struct {
-        uint64_t reftable_offset;
-        uint32_t reftable_clusters;
-    } QEMU_PACKED reftable_offset_and_clusters;
-
-    qcow2_cache_empty(bs, s->refcount_block_cache);
+    bool reftable_grown = false;
+    int ret;
 
-write_refblocks:
-    for (; cluster < *nb_clusters; cluster++) {
+    for (cluster = first_cluster; cluster < end_cluster; cluster++) {
+        /* Check all clusters to find refblocks that contain non-zero entries */
         if (!s->get_refcount(*refcount_table, cluster)) {
             continue;
         }
 
+        /*
+         * This cluster is allocated, so we need to create a refblock
+         * for it.  The data we will write to disk is just the
+         * respective slice from *refcount_table, so it will contain
+         * accurate refcounts for all clusters belonging to this
+         * refblock.  After we have written it, we will therefore skip
+         * all remaining clusters in this refblock.
+         */
+
         refblock_index = cluster >> s->refcount_block_bits;
         refblock_start = refblock_index << s->refcount_block_bits;
 
-        /* Don't allocate a cluster in a refblock already written to disk */
-        if (first_free_cluster < refblock_start) {
-            first_free_cluster = refblock_start;
-        }
-        refblock_offset = alloc_clusters_imrt(bs, 1, refcount_table,
-                                              nb_clusters, &first_free_cluster);
-        if (refblock_offset < 0) {
-            fprintf(stderr, "ERROR allocating refblock: %s\n",
-                    strerror(-refblock_offset));
-            res->check_errors++;
-            ret = refblock_offset;
-            goto fail;
-        }
+        if (on_disk_reftable_entries > refblock_index &&
+            on_disk_reftable[refblock_index])
+        {
+            /*
+             * We can get here after a `goto write_refblocks`: We have a
+             * reftable from a previous run, and the refblock is already
+             * allocated.  No need to allocate it again.
+             */
+            refblock_offset = on_disk_reftable[refblock_index];
+        } else {
+            int64_t refblock_cluster_index;
 
-        if (reftable_size <= refblock_index) {
-            uint32_t old_reftable_size = reftable_size;
-            uint64_t *new_on_disk_reftable;
+            /* Don't allocate a cluster in a refblock already written to disk */
+            if (first_free_cluster < refblock_start) {
+                first_free_cluster = refblock_start;
+            }
+            refblock_offset = alloc_clusters_imrt(bs, 1, refcount_table,
+                                                  nb_clusters,
+                                                  &first_free_cluster);
+            if (refblock_offset < 0) {
+                fprintf(stderr, "ERROR allocating refblock: %s\n",
+                        strerror(-refblock_offset));
+                return refblock_offset;
+            }
 
-            reftable_size = ROUND_UP((refblock_index + 1) * REFTABLE_ENTRY_SIZE,
-                                     s->cluster_size) / REFTABLE_ENTRY_SIZE;
-            new_on_disk_reftable = g_try_realloc(on_disk_reftable,
-                                                 reftable_size *
-                                                 REFTABLE_ENTRY_SIZE);
-            if (!new_on_disk_reftable) {
-                res->check_errors++;
-                ret = -ENOMEM;
-                goto fail;
+            refblock_cluster_index = refblock_offset / s->cluster_size;
+            if (refblock_cluster_index >= end_cluster) {
+                /*
+                 * We must write the refblock that holds this refblock's
+                 * refcount
+                 */
+                end_cluster = refblock_cluster_index + 1;
             }
-            on_disk_reftable = new_on_disk_reftable;
 
-            memset(on_disk_reftable + old_reftable_size, 0,
-                   (reftable_size - old_reftable_size) * REFTABLE_ENTRY_SIZE);
+            if (on_disk_reftable_entries <= refblock_index) {
+                on_disk_reftable_entries =
+                    ROUND_UP((refblock_index + 1) * REFTABLE_ENTRY_SIZE,
+                             s->cluster_size) / REFTABLE_ENTRY_SIZE;
+                on_disk_reftable =
+                    g_try_realloc(on_disk_reftable,
+                                  on_disk_reftable_entries *
+                                  REFTABLE_ENTRY_SIZE);
+                if (!on_disk_reftable) {
+                    return -ENOMEM;
+                }
 
-            /* The offset we have for the reftable is now no longer valid;
-             * this will leak that range, but we can easily fix that by running
-             * a leak-fixing check after this rebuild operation */
-            reftable_offset = -1;
-        } else {
-            assert(on_disk_reftable);
-        }
-        on_disk_reftable[refblock_index] = refblock_offset;
+                memset(on_disk_reftable + *on_disk_reftable_entries_ptr, 0,
+                       (on_disk_reftable_entries -
+                        *on_disk_reftable_entries_ptr) *
+                       REFTABLE_ENTRY_SIZE);
 
-        /* If this is apparently the last refblock (for now), try to squeeze the
-         * reftable in */
-        if (refblock_index == (*nb_clusters - 1) >> s->refcount_block_bits &&
-            reftable_offset < 0)
-        {
-            uint64_t reftable_clusters = size_to_clusters(s, reftable_size *
-                                                          REFTABLE_ENTRY_SIZE);
-            reftable_offset = alloc_clusters_imrt(bs, reftable_clusters,
-                                                  refcount_table, nb_clusters,
-                                                  &first_free_cluster);
-            if (reftable_offset < 0) {
-                fprintf(stderr, "ERROR allocating reftable: %s\n",
-                        strerror(-reftable_offset));
-                res->check_errors++;
-                ret = reftable_offset;
-                goto fail;
+                *on_disk_reftable_ptr = on_disk_reftable;
+                *on_disk_reftable_entries_ptr = on_disk_reftable_entries;
+
+                reftable_grown = true;
+            } else {
+                assert(on_disk_reftable);
             }
+            on_disk_reftable[refblock_index] = refblock_offset;
         }
 
+        /* Refblock is allocated, write it to disk */
+
         ret = qcow2_pre_write_overlap_check(bs, 0, refblock_offset,
                                             s->cluster_size, false);
         if (ret < 0) {
             fprintf(stderr, "ERROR writing refblock: %s\n", strerror(-ret));
-            goto fail;
+            return ret;
         }
 
-        /* The size of *refcount_table is always cluster-aligned, therefore the
-         * write operation will not overflow */
+        /*
+         * The refblock is simply a slice of *refcount_table.
+         * Note that the size of *refcount_table is always aligned to
+         * whole clusters, so the write operation will not result in
+         * out-of-bounds accesses.
+         */
         on_disk_refblock = (void *)((char *) *refcount_table +
                                     refblock_index * s->cluster_size);
 
@@ -XXX,XX +XXX,XX @@ write_refblocks:
                           s->cluster_size);
         if (ret < 0) {
             fprintf(stderr, "ERROR writing refblock: %s\n", strerror(-ret));
-            goto fail;
+            return ret;
         }
 
-        /* Go to the end of this refblock */
+        /* This refblock is done, skip to its end */
         cluster = refblock_start + s->refcount_block_size - 1;
     }
 
-    if (reftable_offset < 0) {
-        uint64_t post_refblock_start, reftable_clusters;
+    return reftable_grown;
+}
+
+/*
+ * Creates a new refcount structure based solely on the in-memory information
+ * given through *refcount_table (this in-memory information is basically just
+ * the concatenation of all refblocks).  All necessary allocations will be
+ * reflected in that array.
+ *
+ * On success, the old refcount structure is leaked (it will be covered by the
+ * new refcount structure).
+ */
+static int rebuild_refcount_structure(BlockDriverState *bs,
+                                      BdrvCheckResult *res,
+                                      void **refcount_table,
+                                      int64_t *nb_clusters)
+{
+    BDRVQcow2State *s = bs->opaque;
+    int64_t reftable_offset = -1;
+    int64_t reftable_length = 0;
+    int64_t reftable_clusters;
+    int64_t refblock_index;
+    uint32_t on_disk_reftable_entries = 0;
+    uint64_t *on_disk_reftable = NULL;
+    int ret = 0;
+    int reftable_size_changed = 0;
+    struct {
+        uint64_t reftable_offset;
+        uint32_t reftable_clusters;
+    } QEMU_PACKED reftable_offset_and_clusters;
+
+    qcow2_cache_empty(bs, s->refcount_block_cache);
+
+    /*
+     * For each refblock containing entries, we try to allocate a
+     * cluster (in the in-memory refcount table) and write its offset
+     * into on_disk_reftable[].  We then write the whole refblock to
+     * disk (as a slice of the in-memory refcount table).
+     * This is done by rebuild_refcounts_write_refblocks().
+     *
+     * Once we have scanned all clusters, we try to find space for the
+     * reftable.  This will dirty the in-memory refcount table (i.e.
+     * make it differ from the refblocks we have already written), so we
+     * need to run rebuild_refcounts_write_refblocks() again for the
+     * range of clusters where the reftable has been allocated.
+     *
+     * This second run might make the reftable grow again, in which case
+     * we will need to allocate another space for it, which is why we
+     * repeat all this until the reftable stops growing.
+     *
+     * (This loop will terminate, because with every cluster the
+     * reftable grows, it can accomodate a multitude of more refcounts,
+     * so that at some point this must be able to cover the reftable
+     * and all refblocks describing it.)
+     *
+     * We then convert the reftable to big-endian and write it to disk.
+     *
+     * Note that we never free any reftable allocations.  Doing so would
+     * needlessly complicate the algorithm: The eventual second check
+     * run we do will clean up all leaks we have caused.
+     */
+
+    reftable_size_changed =
+        rebuild_refcounts_write_refblocks(bs, refcount_table, nb_clusters,
+                                          0, *nb_clusters,
+                                          &on_disk_reftable,
+                                          &on_disk_reftable_entries);
+    if (reftable_size_changed < 0) {
+        res->check_errors++;
+        ret = reftable_size_changed;
+        goto fail;
+    }
+
+    /*
+     * There was no reftable before, so rebuild_refcounts_write_refblocks()
+     * must have increased its size (from 0 to something).
+     */
+    assert(reftable_size_changed);
+
+    do {
+        int64_t reftable_start_cluster, reftable_end_cluster;
+        int64_t first_free_cluster = 0;
+
+        reftable_length = on_disk_reftable_entries * REFTABLE_ENTRY_SIZE;
+        reftable_clusters = size_to_clusters(s, reftable_length);
 
-        post_refblock_start = ROUND_UP(*nb_clusters, s->refcount_block_size);
-        reftable_clusters =
-            size_to_clusters(s, reftable_size * REFTABLE_ENTRY_SIZE);
-        /* Not pretty but simple */
-        if (first_free_cluster < post_refblock_start) {
-            first_free_cluster = post_refblock_start;
-        }
         reftable_offset = alloc_clusters_imrt(bs, reftable_clusters,
                                               refcount_table, nb_clusters,
                                               &first_free_cluster);
@@ -XXX,XX +XXX,XX @@ write_refblocks:
             goto fail;
         }
 
-        goto write_refblocks;
-    }
+        /*
+         * We need to update the affected refblocks, so re-run the
+         * write_refblocks loop for the reftable's range of clusters.
+         */
+        assert(offset_into_cluster(s, reftable_offset) == 0);
+        reftable_start_cluster = reftable_offset / s->cluster_size;
+        reftable_end_cluster = reftable_start_cluster + reftable_clusters;
+        reftable_size_changed =
+            rebuild_refcounts_write_refblocks(bs, refcount_table, nb_clusters,
+                                              reftable_start_cluster,
+                                              reftable_end_cluster,
+                                              &on_disk_reftable,
+                                              &on_disk_reftable_entries);
+        if (reftable_size_changed < 0) {
+            res->check_errors++;
+            ret = reftable_size_changed;
+            goto fail;
+        }
+
+        /*
+         * If the reftable size has changed, we will need to find a new
+         * allocation, repeating the loop.
+         */
+    } while (reftable_size_changed);
 
-    for (refblock_index = 0; refblock_index < reftable_size; refblock_index++) {
+    /* The above loop must have run at least once */
+    assert(reftable_offset >= 0);
+
+    /*
+     * All allocations are done, all refblocks are written, convert the
+     * reftable to big-endian and write it to disk.
+     */
+
+    for (refblock_index = 0; refblock_index < on_disk_reftable_entries;
+         refblock_index++)
+    {
         cpu_to_be64s(&on_disk_reftable[refblock_index]);
     }
 
-    ret = qcow2_pre_write_overlap_check(bs, 0, reftable_offset,
-                                        reftable_size * REFTABLE_ENTRY_SIZE,
+    ret = qcow2_pre_write_overlap_check(bs, 0, reftable_offset, reftable_length,
                                         false);
     if (ret < 0) {
         fprintf(stderr, "ERROR writing reftable: %s\n", strerror(-ret));
         goto fail;
     }
 
-    assert(reftable_size < INT_MAX / REFTABLE_ENTRY_SIZE);
+    assert(reftable_length < INT_MAX);
     ret = bdrv_pwrite(bs->file, reftable_offset, on_disk_reftable,
-                      reftable_size * REFTABLE_ENTRY_SIZE);
+                      reftable_length);
     if (ret < 0) {
         fprintf(stderr, "ERROR writing reftable: %s\n", strerror(-ret));
         goto fail;
@@ -XXX,XX +XXX,XX @@ write_refblocks:
     /* Enter new reftable into the image header */
     reftable_offset_and_clusters.reftable_offset = cpu_to_be64(reftable_offset);
     reftable_offset_and_clusters.reftable_clusters =
-        cpu_to_be32(size_to_clusters(s, reftable_size * REFTABLE_ENTRY_SIZE));
+        cpu_to_be32(reftable_clusters);
     ret = bdrv_pwrite_sync(bs->file,
                            offsetof(QCowHeader, refcount_table_offset),
                            &reftable_offset_and_clusters,
@@ -XXX,XX +XXX,XX @@ write_refblocks:
         goto fail;
     }
 
-    for (refblock_index = 0; refblock_index < reftable_size; refblock_index++) {
+    for (refblock_index = 0; refblock_index < on_disk_reftable_entries;
+         refblock_index++)
+    {
         be64_to_cpus(&on_disk_reftable[refblock_index]);
     }
     s->refcount_table = on_disk_reftable;
     s->refcount_table_offset = reftable_offset;
-    s->refcount_table_size = reftable_size;
+    s->refcount_table_size = on_disk_reftable_entries;
     update_max_refcount_table_index(s);
 
     return 0;
-- 
2.35.1

One clear problem with how qcow2's refcount structure rebuild algorithm
used to be before "qcow2: Improve refcount structure rebuilding" was
that it is prone to failure for qcow2 images on block devices: There is
generally unused space after the actual image, and if that exceeds what
one refblock covers, the old algorithm would invariably write the
reftable past the block device's end, which cannot work.  The new
algorithm does not have this problem.

Test it with three tests:
(1) Create an image with more empty space at the end than what one
    refblock covers, see whether rebuilding the refcount structures
    results in a change in the image file length.  (It should not.)

(2) Leave precisely enough space somewhere at the beginning of the image
    for the new reftable (and the refblock for that place), see whether
    the new algorithm puts the reftable there.  (It should.)

(3) Test the original problem: Create (something like) a block device
    with a fixed size, then create a qcow2 image in there, write some
    data, and then have qemu-img check rebuild the refcount structures.
    Before HEAD^, the reftable would have been written past the image
    file end, i.e. outside of what the block device provides, which
    cannot work.  HEAD^ should have fixed that.
    ("Something like a block device" means a loop device if we can use
    one ("sudo -n losetup" works), or a FUSE block export with
    growable=false otherwise.)

Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20220405134652.19278-3-hreitz@redhat.com>
---
 tests/qemu-iotests/108     | 259 ++++++++++++++++++++++++++++++++++++-
 tests/qemu-iotests/108.out |  81 ++++++++++++
 2 files changed, 339 insertions(+), 1 deletion(-)

diff --git a/tests/qemu-iotests/108 b/tests/qemu-iotests/108
index XXXXXXX..XXXXXXX 100755
--- a/tests/qemu-iotests/108
+++ b/tests/qemu-iotests/108
@@ -XXX,XX +XXX,XX @@ status=1	# failure is the default!
 
 _cleanup()
 {
-	_cleanup_test_img
+    _cleanup_test_img
+    if [ -f "$TEST_DIR/qsd.pid" ]; then
+        qsd_pid=$(cat "$TEST_DIR/qsd.pid")
+        kill -KILL "$qsd_pid"
+        fusermount -u "$TEST_DIR/fuse-export" &>/dev/null
+    fi
+    rm -f "$TEST_DIR/fuse-export"
 }
 trap "_cleanup; exit \$status" 0 1 2 3 15
 
 # get standard environment, filters and checks
 . ./common.rc
 . ./common.filter
+. ./common.qemu
 
 # This tests qcow2-specific low-level functionality
 _supported_fmt qcow2
@@ -XXX,XX +XXX,XX @@ _supported_os Linux
 # files
 _unsupported_imgopts 'refcount_bits=$[^1]\|.\([^6]\|$$\)' data_file
 
+# This test either needs sudo -n losetup or FUSE exports to work
+if sudo -n losetup &>/dev/null; then
+    loopdev=true
+else
+    loopdev=false
+
+    # QSD --export fuse will either yield "Parameter 'id' is missing"
+    # or "Invalid parameter 'fuse'", depending on whether there is
+    # FUSE support or not.
+    error=$($QSD --export fuse 2>&1)
+    if [[ $error = *"'fuse'"* ]]; then
+        _notrun 'Passwordless sudo for losetup or FUSE support required, but' \
+                'neither is available'
+    fi
+fi
+
 echo
 echo '=== Repairing an image without any refcount table ==='
 echo
@@ -XXX,XX +XXX,XX @@ _make_test_img 64M
 poke_file "$TEST_IMG" $((0x10008)) "\xff\xff\xff\xff\xff\xff\x00\x00"
 _check_test_img -r all
 
+echo
+echo '=== Check rebuilt reftable location ==='
+
+# In an earlier version of the refcount rebuild algorithm, the
+# reftable was generally placed at the image end (unless something was
+# allocated in the area covered by the refblock right before the image
+# file end, then we would try to place the reftable in that refblock).
+# This was later changed so the reftable would be placed in the
+# earliest possible location.  Test this.
+
+echo
+echo '--- Does the image size increase? ---'
+echo
+
+# First test: Just create some image, write some data to it, and
+# resize it so there is free space at the end of the image (enough
+# that it spans at least one full refblock, which for cluster_size=512
+# images, spans 128k).  With the old algorithm, the reftable would
+# have then been placed at the end of the image file, but with the new
+# one, it will be put in that free space.
+# We want to check whether the size of the image file increases due to
+# rebuilding the refcount structures (it should not).
+
+_make_test_img -o 'cluster_size=512' 1M
+# Write something
+$QEMU_IO -c 'write 0 64k' "$TEST_IMG" | _filter_qemu_io
+
+# Add free space
+file_len=$(stat -c '%s' "$TEST_IMG")
+truncate -s $((file_len + 256 * 1024)) "$TEST_IMG"
+
+# Corrupt the image by saying the image header was not allocated
+rt_offset=$(peek_file_be "$TEST_IMG" 48 8)
+rb_offset=$(peek_file_be "$TEST_IMG" $rt_offset 8)
+poke_file "$TEST_IMG" $rb_offset "\x00\x00"
+
+# Check whether rebuilding the refcount structures increases the image
+# file size
+file_len=$(stat -c '%s' "$TEST_IMG")
+echo
+# The only leaks there can be are the old refcount structures that are
+# leaked during rebuilding, no need to clutter the output with them
+_check_test_img -r all | grep -v '^Repairing cluster.*refcount=1 reference=0'
+echo
+post_repair_file_len=$(stat -c '%s' "$TEST_IMG")
+
+if [[ $file_len -eq $post_repair_file_len ]]; then
+    echo 'OK: Image size did not change'
+else
+    echo 'ERROR: Image size differs' \
+        "($file_len before, $post_repair_file_len after)"
+fi
+
+echo
+echo '--- Will the reftable occupy a hole specifically left for it?  ---'
+echo
+
+# Note: With cluster_size=512, every refblock covers 128k.
+# The reftable covers 8M per reftable cluster.
+
+# Create an image that requires two reftable clusters (just because
+# this is more interesting than a single-clustered reftable).
+_make_test_img -o 'cluster_size=512' 9M
+$QEMU_IO -c 'write 0 8M' "$TEST_IMG" | _filter_qemu_io
+
+# Writing 8M will have resized the reftable.  Unfortunately, doing so
+# will leave holes in the file, so we need to fill them up so we can
+# be sure the whole file is allocated.  Do that by writing
+# consecutively smaller chunks starting from 8 MB, until the file
+# length increases even with a chunk size of 512.  Then we must have
+# filled all holes.
+ofs=$((8 * 1024 * 1024))
+block_len=$((16 * 1024))
+while [[ $block_len -ge 512 ]]; do
+    file_len=$(stat -c '%s' "$TEST_IMG")
+    while [[ $(stat -c '%s' "$TEST_IMG") -eq $file_len ]]; do
+        # Do not include this in the reference output, it does not
+        # really matter which qemu-io calls we do here exactly
+        $QEMU_IO -c "write $ofs $block_len" "$TEST_IMG" >/dev/null
+        ofs=$((ofs + block_len))
+    done
+    block_len=$((block_len / 2))
+done
+
+# Fill up to 9M (do not include this in the reference output either,
+# $ofs is random for all we know)
+$QEMU_IO -c "write $ofs $((9 * 1024 * 1024 - ofs))" "$TEST_IMG" >/dev/null
+
+# Make space as follows:
+# - For the first refblock: Right at the beginning of the image (this
+#   refblock is placed in the first place possible),
+# - For the reftable somewhere soon afterwards, still near the
+#   beginning of the image (i.e. covered by the first refblock); the
+#   reftable too is placed in the first place possible, but only after
+#   all refblocks have been placed)
+# No space is needed for the other refblocks, because no refblock is
+# put before the space it covers.  In this test case, we do not mind
+# if they are placed at the image file's end.
+
+# Before we make that space, we have to find out the host offset of
+# the area that belonged to the two data clusters at guest offset 4k,
+# because we expect the reftable to be placed there, and we will have
+# to verify that it is.
+
+l1_offset=$(peek_file_be "$TEST_IMG" 40 8)
+l2_offset=$(peek_file_be "$TEST_IMG" $l1_offset 8)
+l2_offset=$((l2_offset & 0x00fffffffffffe00))
+data_4k_offset=$(peek_file_be "$TEST_IMG" \
+                 $((l2_offset + 4096 / 512 * 8)) 8)
+data_4k_offset=$((data_4k_offset & 0x00fffffffffffe00))
+
+$QEMU_IO -c "discard 0 512" -c "discard 4k 1k" "$TEST_IMG" | _filter_qemu_io
+
+# Corrupt the image by saying the image header was not allocated
+rt_offset=$(peek_file_be "$TEST_IMG" 48 8)
+rb_offset=$(peek_file_be "$TEST_IMG" $rt_offset 8)
+poke_file "$TEST_IMG" $rb_offset "\x00\x00"
+
+echo
+# The only leaks there can be are the old refcount structures that are
+# leaked during rebuilding, no need to clutter the output with them
+_check_test_img -r all | grep -v '^Repairing cluster.*refcount=1 reference=0'
+echo
+
+# Check whether the reftable was put where we expected
+rt_offset=$(peek_file_be "$TEST_IMG" 48 8)
+if [[ $rt_offset -eq $data_4k_offset ]]; then
+    echo 'OK: Reftable is where we expect it'
+else
+    echo "ERROR: Reftable is at $rt_offset, but was expected at $data_4k_offset"
+fi
+
+echo
+echo '--- Rebuilding refcount structures on block devices ---'
+echo
+
+# A block device cannot really grow, at least not during qemu-img
+# check.  As mentioned in the above cases, rebuilding the refcount
+# structure may lead to new refcount structures being written after
+# the end of the image, and in the past that happened even if there
+# was more than sufficient space in the image.  Such post-EOF writes
+# will not work on block devices, so test that the new algorithm
+# avoids it.
+
+# If we have passwordless sudo and losetup, we can use those to create
+# a block device.  Otherwise, we can resort to qemu's FUSE export to
+# create a file that isn't growable, which effectively tests the same
+# thing.
+
+_cleanup_test_img
+truncate -s $((64 * 1024 * 1024)) "$TEST_IMG"
+
+if $loopdev; then
+    export_mp=$(sudo -n losetup --show -f "$TEST_IMG")
+    export_mp_driver=host_device
+    sudo -n chmod go+rw "$export_mp"
+else
+    # Create non-growable FUSE export that is a bit like an empty
+    # block device
+    export_mp="$TEST_DIR/fuse-export"
+    export_mp_driver=file
+    touch "$export_mp"
+
+    $QSD \
+        --blockdev file,node-name=export-node,filename="$TEST_IMG" \
+        --export fuse,id=fuse-export,node-name=export-node,mountpoint="$export_mp",writable=on,growable=off \
+        --pidfile "$TEST_DIR/qsd.pid" \
+        --daemonize
+fi
+
+# Now create a qcow2 image on the device -- unfortunately, qemu-img
+# create force-creates the file, so we have to resort to the
+# blockdev-create job.
+_launch_qemu \
+    --blockdev $export_mp_driver,node-name=file,filename="$export_mp"
+
+_send_qemu_cmd \
+    $QEMU_HANDLE \
+    '{ "execute": "qmp_capabilities" }' \
+    'return'
+
+# Small cluster size again, so the image needs multiple refblocks
+_send_qemu_cmd \
+    $QEMU_HANDLE \
+    '{ "execute": "blockdev-create",
+       "arguments": {
+           "job-id": "create",
+           "options": {
+               "driver": "qcow2",
+               "file": "file",
+               "size": '$((64 * 1024 * 1024))',
+               "cluster-size": 512
+           } } }' \
+    '"concluded"'
+
+_send_qemu_cmd \
+    $QEMU_HANDLE \
+    '{ "execute": "job-dismiss", "arguments": { "id": "create" } }' \
+    'return'
+
+_send_qemu_cmd \
+    $QEMU_HANDLE \
+    '{ "execute": "quit" }' \
+    'return'
+
+wait=y _cleanup_qemu
+echo
+
+# Write some data
+$QEMU_IO -c 'write 0 64k' "$export_mp" | _filter_qemu_io
+
+# Corrupt the image by saying the image header was not allocated
+rt_offset=$(peek_file_be "$export_mp" 48 8)
+rb_offset=$(peek_file_be "$export_mp" $rt_offset 8)
+poke_file "$export_mp" $rb_offset "\x00\x00"
+
+# Repairing such a simple case should just work
+# (We used to put the reftable at the end of the image file, which can
+# never work for non-growable devices.)
+echo
+TEST_IMG="$export_mp" _check_test_img -r all \
+    | grep -v '^Repairing cluster.*refcount=1 reference=0'
+
+if $loopdev; then
+    sudo -n losetup -d "$export_mp"
+else
+    qsd_pid=$(cat "$TEST_DIR/qsd.pid")
+    kill -TERM "$qsd_pid"
+    # Wait for process to exit (cannot `wait` because the QSD is daemonized)
+    while [ -f "$TEST_DIR/qsd.pid" ]; do
+        true
+    done
+fi
+
 # success, all done
 echo '*** done'
 rm -f $seq.full
diff --git a/tests/qemu-iotests/108.out b/tests/qemu-iotests/108.out
index XXXXXXX..XXXXXXX 100644
--- a/tests/qemu-iotests/108.out
+++ b/tests/qemu-iotests/108.out
@@ -XXX,XX +XXX,XX @@ The following inconsistencies were found and repaired:
     0 leaked clusters
     1 corruptions
 
+Double checking the fixed image now...
+No errors were found on the image.
+
+=== Check rebuilt reftable location ===
+
+--- Does the image size increase? ---
+
+Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=1048576
+wrote 65536/65536 bytes at offset 0
+64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+
+ERROR cluster 0 refcount=0 reference=1
+Rebuilding refcount structure
+The following inconsistencies were found and repaired:
+
+    0 leaked clusters
+    1 corruptions
+
+Double checking the fixed image now...
+No errors were found on the image.
+
+OK: Image size did not change
+
+--- Will the reftable occupy a hole specifically left for it?  ---
+
+Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=9437184
+wrote 8388608/8388608 bytes at offset 0
+8 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+discard 512/512 bytes at offset 0
+512 bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+discard 1024/1024 bytes at offset 4096
+1 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+
+ERROR cluster 0 refcount=0 reference=1
+Rebuilding refcount structure
+The following inconsistencies were found and repaired:
+
+    0 leaked clusters
+    1 corruptions
+
+Double checking the fixed image now...
+No errors were found on the image.
+
+OK: Reftable is where we expect it
+
+--- Rebuilding refcount structures on block devices ---
+
+{ "execute": "qmp_capabilities" }
+{"return": {}}
+{ "execute": "blockdev-create",
+       "arguments": {
+           "job-id": "create",
+           "options": {
+               "driver": "IMGFMT",
+               "file": "file",
+               "size": 67108864,
+               "cluster-size": 512
+           } } }
+{"timestamp": {"seconds":  TIMESTAMP, "microseconds":  TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "created", "id": "create"}}
+{"timestamp": {"seconds":  TIMESTAMP, "microseconds":  TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "running", "id": "create"}}
+{"return": {}}
+{"timestamp": {"seconds":  TIMESTAMP, "microseconds":  TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "waiting", "id": "create"}}
+{"timestamp": {"seconds":  TIMESTAMP, "microseconds":  TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "pending", "id": "create"}}
+{"timestamp": {"seconds":  TIMESTAMP, "microseconds":  TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "concluded", "id": "create"}}
+{ "execute": "job-dismiss", "arguments": { "id": "create" } }
+{"timestamp": {"seconds":  TIMESTAMP, "microseconds":  TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "null", "id": "create"}}
+{"return": {}}
+{ "execute": "quit" }
+{"return": {}}
+{"timestamp": {"seconds":  TIMESTAMP, "microseconds":  TIMESTAMP}, "event": "SHUTDOWN", "data": {"guest": false, "reason": "host-qmp-quit"}}
+
+wrote 65536/65536 bytes at offset 0
+64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+
+ERROR cluster 0 refcount=0 reference=1
+Rebuilding refcount structure
+The following inconsistencies were found and repaired:
+
+    0 leaked clusters
+    1 corruptions
+
 Double checking the fixed image now...
 No errors were found on the image.
 *** done
-- 
2.35.1

Instead of fprint()-ing error messages in rebuild_refcount_structure()
and its rebuild_refcounts_write_refblocks() helper, pass them through an
Error object to qcow2_check_refcounts() (which will then print it).

Suggested-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20220405134652.19278-4-hreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
---
 block/qcow2-refcount.c | 33 +++++++++++++++++++--------------
 1 file changed, 19 insertions(+), 14 deletions(-)