bitmaps persistent and migration fixes
v2 cover bitmaps persistent and migration fixes
v2:
01: new, proposed by Max
02: fix leaving cat processes
Vladimir Sementsov-Ogievskiy (2):
qcow2: try load bitmaps only once
iotests: fix 169
block/qcow2.h | 1 + block/qcow2.c | 16 ++++++++-------- tests/qemu-iotests/169 | 48 +++++++++++++++++++++++++++--------------------- 3 files changed, 36 insertions(+), 29 deletions(-)
Signature · 1 line
2.11.1
PATCH v2 1/2 qcow2: try load bitmaps only once
Checking reopen by existence of some bitmaps is wrong, as it may be some other bitmaps, or on the other hand, user may remove bitmaps. This criteria is bad. To simplify things and make behavior more predictable let's just add a flag to remember, that we've already tried to load bitmaps on open and do not want do it again.
--- block/qcow2.h | 1 + block/qcow2.c | 16 ++++++++-------- 2 files changed, 9 insertions(+), 8 deletions(-)
block/qcow2.h +1 -0
@@ -298,6 +298,7 @@ typedef struct BDRVQcow2State {uint32_t nb_bitmaps;uint64_t bitmap_directory_size;uint64_t bitmap_directory_offset; +bool dirty_bitmaps_loaded;int flags;int qcow_version;
block/qcow2.c +8 -8
@@ -1142,6 +1142,7 @@ static int coroutine_fn qcow2_do_open(BlockDriverState *bs, QDict *options,uint64_t ext_end;uint64_t l1_vm_state_index;bool update_header = false; +bool header_updated = false; ret =bdrv_pread (bs->file,0 , &header,sizeof (header));if (ret <0 ) {
@@ -1480,10 +1481,9 @@ static int coroutine_fn qcow2_do_open(BlockDriverState *bs, QDict *options, s->autoclear_features &=QCOW2_AUTOCLEAR_MASK ; } -if (bdrv_dirty_bitmap_next (bs,NULL )) { -/* It's some kind of reopen with already existing dirty bitmaps. There -* are no known cases where we need loading bitmaps in such situation, -* so it's safer don't load them. +if (s->dirty_bitmaps_loaded) { +/* It's some kind of reopen. There are no known cases where we need +* loading bitmaps in such situation, so it's safer don't load them. * * Moreover, if we have some readonly bitmaps and we are reopening for * rw we should reopen bitmaps correspondingly.
@@ -1491,13 +1491,13 @@ static int coroutine_fn qcow2_do_open(BlockDriverState *bs, QDict *options,if (bdrv_has_readonly_bitmaps (bs) && !bdrv_is_read_only (bs) && !(bdrv_get_flags (bs) &BDRV_O_INACTIVE )) { -bool header_updated = false;qcow2_reopen_bitmaps_rw_hint (bs, &header_updated, &local_err); - update_header = update_header && !header_updated; } - }else if (qcow2_load_dirty_bitmaps (bs, &local_err)) { - update_header = false; + }else { + header_updated =qcow2_load_dirty_bitmaps (bs, &local_err); + s->dirty_bitmaps_loaded = true; } + update_header = update_header && !header_updated;if (local_err !=NULL ) {error_propagate (errp, local_err); ret = -EINVAL ;
Signature · 1 line
2.11.1
Re: [Qemu-devel] [PATCH v2 1/2] qcow2: try load bitmaps only once
Quoted 5 lines · Vladimir Sementsov-Ogievskiy
Checking reopen by existence of some bitmaps is wrong, as it may be some other bitmaps, or on the other hand, user may remove bitmaps. This criteria is bad. To simplify things and make behavior more predictable let's just add a flag to remember, that we've already tried to load bitmaps on open and do not want do it again.
Wording suggestion:
Checking whether we are reopening an image based on the existence of some bitmaps is wrong, as the set of bitmaps may have changed (for example, the user may have added or removed bitmaps in the meantime). To simplify things and make behavior more predictable, let's just add a flag to remember whether we've already tried to load bitmaps, so that we only attempt it on the initial open.
Quoted 8 lines
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>--- block/qcow2.h | 1 + block/qcow2.c | 16 ++++++++-------- 2 files changed, 9 insertions(+), 8 deletions(-)
block/qcow2.c static int coroutine_fn qcow2_do_open(BlockDriverState *bs, QDict *options, · line 1486
Quoted 10 lines
block/qcow2.c +3
-4@@ -1480,10 +1481,9 @@ static int coroutine_fn qcow2_do_open(BlockDriverState *bs, QDict *options, s->autoclear_features &=QCOW2_AUTOCLEAR_MASK ; } -if (bdrv_dirty_bitmap_next (bs,NULL )) { -/* It's some kind of reopen with already existing dirty bitmaps. There -* are no known cases where we need loading bitmaps in such situation, -* so it's safer don't load them. +if (s->dirty_bitmaps_loaded) { +/* It's some kind of reopen. There are no known cases where we need +* loading bitmaps in such situation, so it's safer don't load them.
Pre-existing wording, but sounds better as:
There are no known cases where we need to reload bitmaps in such a situation, so it's safer to skip them.
Quoted 8 lines
block/qcow2.c +0
-1* * Moreover, if we have some readonly bitmaps and we are reopening for * rw we should reopen bitmaps correspondingly. @@ -1491,13 +1491,13 @@ static int coroutine_fn qcow2_do_open(BlockDriverState *bs, QDict *options,if (bdrv_has_readonly_bitmaps (bs) && !bdrv_is_read_only (bs) && !(bdrv_get_flags (bs) &BDRV_O_INACTIVE )) { -bool header_updated = false;
block/qcow2.c static int coroutine_fn qcow2_do_open(BlockDriverState *bs, QDict *options, · line 1500
Quoted 10 lines
block/qcow2.c +4
-3qcow2_reopen_bitmaps_rw_hint (bs, &header_updated, &local_err); - update_header = update_header && !header_updated; } - }else if (qcow2_load_dirty_bitmaps (bs, &local_err)) { - update_header = false; + }else { + header_updated =qcow2_load_dirty_bitmaps (bs, &local_err); + s->dirty_bitmaps_loaded = true; } + update_header = update_header && !header_updated;
Could write this as 'update_header &= !header_updated;'
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
Re: [Qemu-devel] [PATCH v2 1/2] qcow2: try load bitmaps only once
Quoted 6 lines · Eric Blake
Quoted 5 lines · Vladimir Sementsov-Ogievskiy
Checking reopen by existence of some bitmaps is wrong, as it may be some other bitmaps, or on the other hand, user may remove bitmaps. This criteria is bad. To simplify things and make behavior more predictable let's just add a flag to remember, that we've already tried to load bitmaps on open and do not want do it again.
Quoted 8 lines
Wording suggestion:
Checking whether we are reopening an image based on the existence of some bitmaps is wrong, as the set of bitmaps may have changed (for example, the user may have added or removed bitmaps in the meantime). To simplify things and make behavior more predictable, let's just add a flag to remember whether we've already tried to load bitmaps, so that we only attempt it on the initial open.
You've dropped an idea supposed by Max, about migration related bitmaps,
I said "some other bitmaps", because now I doubt is it possible: we have migration related bitmaps on source, not on target..
Anyway, I'm ok with all your suggestions, thank you. I can respin, or they may be squashed-in inflight.
Quoted 43 lines
Quoted 8 lines
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>--- block/qcow2.h | 1 + block/qcow2.c | 16 ++++++++-------- 2 files changed, 9 insertions(+), 8 deletions(-)block/qcow2.c static int coroutine_fn qcow2_do_open(BlockDriverState *bs, QDict *options, · line 1486
Quoted 10 lines
block/qcow2.c +3
-4@@ -1480,10 +1481,9 @@ static int coroutine_fn qcow2_do_open(BlockDriverState *bs, QDict *options, s->autoclear_features &=QCOW2_AUTOCLEAR_MASK ; } -if (bdrv_dirty_bitmap_next (bs,NULL )) { -/* It's some kind of reopen with already existing dirty bitmaps. There -* are no known cases where we need loading bitmaps in such situation, -* so it's safer don't load them. +if (s->dirty_bitmaps_loaded) { +/* It's some kind of reopen. There are no known cases where we need +* loading bitmaps in such situation, so it's safer don't load them. Pre-existing wording, but sounds better as:
There are no known cases where we need to reload bitmaps in such a situation, so it's safer to skip them.
Quoted 8 lines
block/qcow2.c +0
-1* * Moreover, if we have some readonly bitmaps and we are reopening for * rw we should reopen bitmaps correspondingly. @@ -1491,13 +1491,13 @@ static int coroutine_fn qcow2_do_open(BlockDriverState *bs, QDict *options,if (bdrv_has_readonly_bitmaps (bs) && !bdrv_is_read_only (bs) && !(bdrv_get_flags (bs) &BDRV_O_INACTIVE )) { -bool header_updated = false;block/qcow2.c static int coroutine_fn qcow2_do_open(BlockDriverState *bs, QDict *options, · line 1500
Quoted 10 lines
block/qcow2.c +4
-3qcow2_reopen_bitmaps_rw_hint (bs, &header_updated, &local_err); - update_header = update_header && !header_updated; } - }else if (qcow2_load_dirty_bitmaps (bs, &local_err)) { - update_header = false; + }else { + header_updated =qcow2_load_dirty_bitmaps (bs, &local_err); + s->dirty_bitmaps_loaded = true; } + update_header = update_header && !header_updated;Could write this as '
update_header&= !header_updated;'
Signature · 2 lines
Best regards, Vladimir
Re: [Qemu-devel] [PATCH v2 1/2] qcow2: try load bitmaps only once
In reply to Vladimir Sementsov-OgievskiyQuoted 11 lines · Vladimir Sementsov-Ogievskiy
Checking reopen by existence of some bitmaps is wrong, as it may be some other bitmaps, or on the other hand, user may remove bitmaps. This criteria is bad. To simplify things and make behavior more predictable let's just add a flag to remember, that we've already tried to load bitmaps on open and do not want do it again.
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>--- block/qcow2.h | 1 + block/qcow2.c | 16 ++++++++-------- 2 files changed, 9 insertions(+), 8 deletions(-)
PATCH v2 2/2 iotests: fix 169
In reply to Vladimir Sementsov-OgievskiyImprove and fix 169:
- use MIGRATION events instead of RESUME
- make a TODO: enable dirty-bitmaps capability for offline case
- recreate vm_b without -incoming near test end
This (likely) fixes racy faults at least of the following types:
- timeout on waiting for RESUME event
- sha256 mismatch on line 136 (142 after this patch)
- fail to self.vm_b.launch() on line 135 (141 now after this patch)
And surely fixes cat processes, left after test finish.
--- tests/qemu-iotests/169 | 48 +++++++++++++++++++++++++++--------------------- 1 file changed, 27 insertions(+), 21 deletions(-)
tests/qemu-iotests/169 +27 -21
@@ -31,6 +31,8 @@ disk_a = os.path.join(iotests.test_dir, 'disk_a') disk_b = os.path.join(iotests.test_dir, 'disk_b') size = '1M' mig_file = os.path.join(iotests.test_dir, 'mig_file') +mig_cmd = 'exec: cat > ' + mig_file +incoming_cmd = 'exec: cat ' + mig_file class TestDirtyBitmapMigration(iotests.QMPTestCase):
@@ -49,7 +51,6 @@ class TestDirtyBitmapMigration(iotests.QMPTestCase): self.vm_a.launch() self.vm_b = iotests.VM(path_suffix='b') - self.vm_b.add_incoming("exec: cat '" + mig_file + "'") def add_bitmap(self, vm, granularity, persistent): params = {'node': 'drive0',
@@ -86,36 +87,30 @@ class TestDirtyBitmapMigration(iotests.QMPTestCase): (0xa0201, 0x1000)) should_migrate = migrate_bitmaps or persistent and shared_storage + mig_caps = [{'capability': 'events', 'state': True}] + if migrate_bitmaps: + mig_caps.append({'capability': 'dirty-bitmaps', 'state': True}) + result = self.vm_a.qmp('migrate-set-capabilities', + capabilities=mig_caps) + self.assert_qmp(result, 'return', {}) + + self.vm_b.add_incoming(incoming_cmd if online else "defer") self.vm_b.add_drive(disk_a if shared_storage else disk_b) if online: os.mkfifo(mig_file) self.vm_b.launch() + result = self.vm_b.qmp('migrate-set-capabilities', + capabilities=mig_caps) + self.assert_qmp(result, 'return', {}) self.add_bitmap(self.vm_a, granularity, persistent) for r in regions: self.vm_a.hmp_qemu_io('drive0', 'write %d %d' % r) sha256 = self.get_bitmap_hash(self.vm_a) - if migrate_bitmaps: - capabilities = [{'capability': 'dirty-bitmaps', 'state': True}] - - result = self.vm_a.qmp('migrate-set-capabilities', - capabilities=capabilities) - self.assert_qmp(result, 'return', {}) - - if online: - result = self.vm_b.qmp('migrate-set-capabilities', - capabilities=capabilities) - self.assert_qmp(result, 'return', {}) - - result = self.vm_a.qmp('migrate-set-capabilities', - capabilities=[{'capability': 'events', - 'state': True}]) - self.assert_qmp(result, 'return', {}) - - result = self.vm_a.qmp('migrate', uri='exec:cat>' + mig_file) + result = self.vm_a.qmp('migrate', uri=mig_cmd) while True: event = self.vm_a.event_wait('MIGRATION') if event['data']['status'] == 'completed':
@@ -124,14 +119,25 @@ class TestDirtyBitmapMigration(iotests.QMPTestCase): if not online: self.vm_a.shutdown() self.vm_b.launch() - # TODO enable bitmap capability for vm_b in this case + result = self.vm_b.qmp('migrate-set-capabilities', + capabilities=mig_caps) + self.assert_qmp(result, 'return', {}) + result = self.vm_b.qmp('migrate-incoming', uri=incoming_cmd) + self.assert_qmp(result, 'return', {}) - self.vm_b.event_wait("RESUME", timeout=10.0) + while True: + event = self.vm_b.event_wait('MIGRATION') + if event['data']['status'] == 'completed': + break self.check_bitmap(self.vm_b, sha256 if should_migrate else False) if should_migrate: self.vm_b.shutdown() + # recreate vm_b, as we don't want -incoming option (this will lead + # to "cat" process left alive after test finish) + self.vm_b = iotests.VM(path_suffix='b') + self.vm_b.add_drive(disk_a if shared_storage else disk_b) self.vm_b.launch() self.check_bitmap(self.vm_b, sha256 if persistent else False)
Signature · 1 line
2.11.1
Re: [Qemu-devel] [PATCH v2 2/2] iotests: fix 169
Quoted 11 lines · Vladimir Sementsov-Ogievskiy
Improve and fix 169:
- use MIGRATION events instead of RESUME
- make a TODO: enable dirty-bitmaps capability for offline case
- recreatevm_bwithout -incoming near test endThis (likely) fixes racy faults at least of the following types:
- timeout on waiting for RESUME event
- sha256 mismatch on line 136 (142 after this patch)
- fail toself.vm_b.launch()on line 135 (141 now after this patch)
Quoted 6 lines
And surely fixes cat processes, left after test finish.
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>--- tests/qemu-iotests/169 | 48 +++++++++++++++++++++++++++--------------------- 1 file changed, 27 insertions(+), 21 deletions(-)
Looks good, makes the test pass reliably (here) and doesn't result in overly loud meowing due to an abundance of orphaned cats. Thanks!
Quoted 5 lines · Vladimir Sementsov-Ogievskiy
v2:
01: new, proposed by Max
02: fix leaving cat processes
Quoted 8 lines
Vladimir Sementsov-Ogievskiy (2):
qcow2: try load bitmaps only once
iotests: fix 169block/qcow2.h | 1 + block/qcow2.c | 16 ++++++++-------- tests/qemu-iotests/169 | 48 +++++++++++++++++++++++++++--------------------- 3 files changed, 36 insertions(+), 29 deletions(-)
Thanks, changed the comment wording in patch 1 according to Eric's suggestion and applied to my block branch:
https://github.com/XanClic/qemu/commits/block
I left the commit wording unchanged because as you said there might be more cases where we have bitmaps at the node although we haven't yet loaded any from the file. I certainly don't want to have to think too much about those cases and that to me is the main use of patch 1.
If there is going to be an rc4 (it appears there is), I will send a pull request with these patches for it. If not, I probably won't. Then I'll just add a qemu-stable CC and get them into 2.13 -- unless you disagree and you think that this series should indeed be in 2.12.
Max