qcow2: persistent dirty bitmaps
v22 cover qcow2: persistent dirty bitmaps
Hi all!
There is a new update of qcow2-bitmap series - v22.
web: https://src.openvz.org/users/vsementsov/repos/qemu/browse?at=qcow2-bitmap-v22
git: https://src.openvz.org/scm/~vsementsov/qemu.git (tag qcow2-bitmap-v22)
v22:
Rebase on master, so changes, mostly related to new dirty bitmaps mutex:
10: - asserts now in bdrv_{re,}set_dirty_bitmap_locked functions.
- also add assert into bdrv_undo_clear_dirty_bitmap (the only change, not related to rebase)
- add mutex lock into bdrv_dirty_bitmap_set_readonly (as it changes bitmap list, so the lock should be taken)
- return instead of go-to in qmp_block_dirty_bitmap_clear
- in dirty-bitmaps.h, move bdrv_dirty_bitmap_set_readonly before block "Functions that require manual locking", move
bdrv_dirty_bitmap_readonly and bdrv_has_readonly_bitmaps into this block
15: - add mutex lock/unlock into bdrv_dirty_bitmap_set_autoload
- in dirty-bitmaps.h, move bdrv_dirty_bitmap_set_autoload before block "Functions that require manual locking", move bdrv_dirty_bitmap_get_autoload into this block
17: - add mutex lock/unlock into bdrv_dirty_bitmap_set_persistance
- in dirty-bitmaps.h, move bdrv_dirty_bitmap_set_persistance before block "Functions that require manual locking", move bdrv_dirty_bitmap_get_persistance and
bdrv_has_changed_persistent_bitmaps into this block
18: in dirty-bitmaps.h, move bdrv_dirty_bitmap_next into block
"Functions that require manual locking". (do not remove r-b, as it is
just one empty line removed before function declaration)
23: return instead of go-to in qmp_block_dirty_bitmap_add
24: return instead of go-to in qmp_block_dirty_bitmap_add
25: - return instead of go-to
- remove aio_context_acquire/release calls
- no aio_context parameter for block_dirty_bitmap_lookup
- in dirty-bitmaps.h, move bdrv_dirty_bitmap_sha256 into block
"Functions that require manual locking".
29: - return instead of go-to in qmp_block_dirty_bitmap_remove
r-b's are dropped from 10,15,17,25.
v21:
09,10: improve comment, add r-b's by Max and John
10: improve comment,k
11,12: add r-b by John
13: prepend local_err with additional info (Max), add r-b by John 14: add r-b's by Max and John
20,30: add r-b by Max
v20:
handle reopening images ro and rw.
On reopening ro: store bitmaps (storing sets 'IN_USE'=0 in the image) and mark them readonly (set readonly flag in BlockDirtyBitmap)
After reopening rw: mark bitmaps IN_USE in the image and unset readonly flag in BlockDirtyBitmap
09: new
10: improve comment
add parameter 'value' to bdrv_dirty_bitmap_set_readonly
11: use new parameter of bdrv_dirty_bitmap_set_readonly 12-14, 20: new
v19:
rebased on master
05: move 'sign-off' over 'reviewed-by's
08: error_report -> error_setg in qcow2_truncate (because of rebase) 09: return EPERM in bdrv_aligned_pwritev and bdrv_co_pdiscard if there
are readonly bitmaps. EPERM is chosen because it is already used for readonly image in bdrv_co_pdiscard.
Also handle readonly bitmap in block_dirty_bitmap_clear_prepare and qmp_block_dirty_bitmap_clear
Max's r-b is not added
10: fix grammar in comment
add Max's r-b
12, 13, 15, 21: add Max's r-b
24: fix grammar in comment
25: fix grammar and wording in comment
also, I see contextual changes in inactiavate mechanism. Hope, they do not affect these series.
v18:
rebased on master (sorry for v17)
08: contextual: qcow2_do_open is changed instead of qcow2_open
rename s/need_update_header/update_header/ in qcow2_do_open, to not do it in 10
save r-b's by Max and John
09: new patch
10: load_bitmap_data: do not clear bitmap parameter - it should be already cleared
(it actually created before single load_bitmap_data() call)
if some bitmaps are loaded, but we can't write the image (it is readonly
or inactive), so we can't mark bitmaps "in use" in the image, mark
corresponding BdrvDirtyBitmap read-only.
change error_setg to error_setg_errno for "Can't update bitmap directory"
no needs to rename s/need_update_header/update_header/ here, as it done in 08
13: function bdrv_has_persistent_bitmaps becomes bdrv_has_changed_persistent_bitmaps,
to handle readonly field.
14: declaration moved to the bottom of .h, save r-b's
15: firstly check bdrv_has_changed_persistent_bitmaps and only then fail on !can_write, and then QSIMPLEQ_INIT(&drop_tables)
skip readonly bitmaps in saving loop
18: remove '#optional', 2.9 -> 2.10, save r-b's
19: remove '#optional', 2.9 -> 2.10, save r-b's
20: 2.9 -> 2.10, save r-b's
21: add check of read-only image open, drop r-b's
24: add comment to qapi/block-core.json, that block-dirty-bitmap-add removes bitmap
from storage. r-b's by Max and John saved
v17:
08: add r-b's by Max and John
09: clear unknown autoclear features from BDRVQcow2State before calling
qcow2_load_autoloading_dirty_bitmaps(), and also do not extra update header if it is updated by qcow2_load_autoloading_dirty_bitmaps().
11: new patch, splitted out from 16
12: rewrite commit message
14: changing bdrv_close moved to separate new patch 11
s/1/1ULL/ ; s/%u/%llu/ for two errors
16: s/No/Not/
add Max's r-b
24: new patch
v16:
mostly by Kevin's comments:
07: just moved here, as preparation for merging refcounts to 08
08: move "qcow2-bitmap: refcounts" staff to this patch, to not break qcow2-img check
drop r-b's
move necessary supporting static functions to this patch too (to not break compilation) fprintf -> error_report
other small changes with error messages and comments
code style
for qcow2-bitmap.c:
'include "exec/log.h"' was dropped
s/1/(1U << 0) for BME_FLAG_IN_USE
add BME_TABLE_ENTRY_FLAG_ALL_ONES to replace magic 1
don't check 'cluster_size <= 0' in check_table_entry
old "[PATCH v15 08/25] block: introduce auto-loading bitmaps" was dropped
09: was "[PATCH v15 09/25] qcow2: add .bdrv_load_autoloading_dirty_bitmaps"
drop r-b's
some staff was moved to 08
update_header_sync - error on bdrv_flush fail
rename disk_sectors_in_bitmap_cluster to sectors_covered_by_bitmap_cluster
and adjust comment.
so, variable for storing this function result: s/dsc/sbc/
s/1/BME_TABLE_ENTRY_FLAG_ALL_ONES/
also, as Qcow2BitmapTable already introduced in 08,
s/table_offset/table.offset/ and s/table_size/table.size, etc..
update_ext_header_and_dir_in_place: add comments, add additional
update_header_sync, to reduce indeterminacy in case of error.
call qcow2_load_autoloading_dirty_bitmaps directly from qcow2_open
no .bdrv_load_autoloading_dirty_bitmaps
11: drop bdrv_store_persistent_dirty_bitmaps at all.
drop r-b's
13: rename patch, rewrite commit msg
drop r-b's
move bdrv_release_named_dirty_bitmaps in bdrv_close() after drv->bdrv_close() Qcow2BitmapTable is already introduced, so no
s/table_offset/table.offset/ and s/table_size/table.size, etc.. here
old 25 with check_constraints_on_bitmap() improvements merged here (Eric)
like in 09, s/dsc/sbc/ and s/disk_sectors_in_bitmap_cluster/sectors_covered_by_bitmap_cluster/ no .bdrv_store_persistent_dirty_bitmaps
call qcow2_store_persistent_dirty_bitmaps directly from qcow2_inactivate
15: corresponding part of 25 merged here. Add John's r-b, as 25 was also reviewed by John. 16: add John's r-b
v15:
13,14: add John's r-b
15: qcow2_can_store_new_dirty_bitmap:
- check max dirty bitmaps and bitmap directory overhead
- switch to error_prepend
rm Max's r-b
not add John's r-b
17-24: add John's r-b
25: changed because 15 changed,
not add John's r-b
v14:
07: use '|=' to update need_update_header
add John's r-b
add Max's r-b
09: remove unused bitmap_table_to_cpu()
left Max's r-b, hope it's ok
add John's r-b
10: remove extra new line
add John's r-b
11: add John's r-b
12: add John's r-b
13: small fixes by John's review:
- remove weird g_free of NULL pointer from
if (tb == NULL) {
g_free(tb);
return -EINVAL;
}
- remove extra comment "/* errp is already set */"
- s/"Too large bitmap directory"/"Bitmap directory is too large"/
left Max's r-b, hope you don't mind
22: add Max's r-b
23: add Max's r-b
24: add Max's r-b
25: new patch to improve error message on check_constraints_on_bitmap fail
v13: Just a fix for style checker.
13: line over 80
14: line over 80
22: s/if () \n{/if () {/
v12:
07: do not update header in qcow2_read_extensions, instead do it in the
end of qcow2_open, where it is updated also to clear unknown autoclear features.
Wrong bdrv_is_root_node used instead of bdrv_is_read_only is fixed automatically.
08: add Max's r-b
09: contextual: s/raw_bsd/raw-format/
qcow2-bitmap.c: copyright s/2016/2017/
fix compilation: move update_header_sync() to this patch
add Max's r-b
13: update_header_sync() is moved to 09
add Max's r-b
15: add Max's r-b
16: As qmp-commands.txt is removed from master, remove it here too.
Sentence "For now only Qcow2 disks support persistent bitmaps.
Default is false." moved to qapi/block-core.json.
Hope that is OK, Max's r-b is not dropped.
17: qmp-commands.txt deleted, r-b is not dropped too.
18: s/is failed/has failed/, add Max's r-b
19: copyright: s/2016/2017/
21: drop "res->corruptions > 0 => ret = -EINVAL", add Max's r-b
22: actually, patch is replaced with a new one, however some parts are the same.
instead of changing bdrv_release_dirty_bitmap behaviour, create new bdrv_remove_persistent_dirty_bitmap
23: improve comment, about not-exists is not an error
24: new patch
Old 24 (qcow2-bitmap: cache bitmap list in BDRVQcow2State) will be sent separately, to be applied after these series.
Also: about bdrv_dirty_bitmap_set_persistance: I've decided not change its behaviour for now and just call bdrv_remove_persistent_dirty_bitmap from qmp_block_dirty_bitmap_remove. Reasons:
1. Do not change reviewed part.
2. I'm not sure, that this complication of BdrvDirtyBitmap.persistent semantics is good (current semantics are just: .persistent means that bitmap should be saved on disk close). .persistent actually is not very related to "is there stored version of the bitmap in the image".
3. It may be done later.
4. It is not actually needed for now, as the only usage is dropping bitmap in bdrv_remove_persistent_dirty_bitmap. May be in future it will be needed, when we will have qmp interfaces for finer control of persistent dirty bitmaps.
v11:
Fix automatic build test fail.
18: wording - "ASCII representation of SHA256 ..."
24: fix redeclaration error. (This is still RFC, may be not needed patch,
see description in v10 below).
v10:
07: rm John's r-b
not add Max's r-b
fix subject s/dirty bitmaps/bitmaps
improve WARNING message
remove header ext if autoclear bit is unset, through qcow2_updata_header() - r-b blocking change
08: move bdrv_load_autoloading_dirty_bitmaps under asserts block (Max)
fix typo in commit message
move bdrv_load_autoloading_dirty_bitmaps after asserts
John's r-b
09: rewrite check_dir_entry
remove extra feature cluster_size=0 from check_table_entry
which clears autoclear bit before trying to update bitmap directory bitmap_list_store - do not free old clusters if in_place=true
changes in comments, fix indents
add functions
bitmap_list_count
update_ext_header_and_dir_in_place (unset autoclear bit before trying to
modify bitmap directory)
(for usage in qcow2_load_autoloading_dirty_bitmaps)
11: add node name to error report
Max's r-b
13: add type Qcow2BitmapTable
move from bm.table_* to bm.table.*
rewrite check_constraints_on_bitmap
flush(bs->file) instead of flush(bs) in update_ext_header_and_dir assert(DIV_ROUND_UP(bm_size, dsc) == tb_size);
and assert(write_size <= s->cluster_size); in store_bitmap_data
fix: s/tb_size/tb_size * sizeof(tb[0]) in store_bitmap
in qcow2_store_persistent_dirty_bitmaps:
fail if !can_write
fix counting of new_nb_bitmaps and new_dir_size
free old clusters _after_ successful directory update (aggregate old bitmap
tables into drop_tables)
2 14: rename to can_store_new_dirty_bitmap
Max's r-b
15: rename to can_store_new_dirty_bitmap
rm extra !!
16: Max's r-b
rename to can_store_new_dirty_bitmap
indenting
Since s/2.8/2.9
17: Max's r-b
Since s/2.8/2.9
18: hashing can fail in comment for qapi
Since s/2.8/2.9
remove dead comment
19: indentation
Max's r-b
20: Max's r-b
21: fix error handling
return 0 if nb_bitmaps == 0
new patches:
22-23: remove bitmap from file on bdrv_release_dirty_bitmap
24: experimental, RFC: cache bitmap list to bs, to not reload it from file.
I'm not sure that is needed now, but if you want I can merge it to earlier patches.
v9:
rebase on master!
01-04,06,07: John's r-b
03: rewording in comment ("The concurrent resetting ..."), John's r-b 07: reword error messages
commit message: dirty bitmap -> bitmap
09: fix uninitialized ret in bitmap_list_store() (fix compilation warning) 18: fix compilation of tests/test-hbitmap
also, change constants names for qcow2: DIRTY_BITMAP -> BITMAP (as it is not dirty bitmaps extension, but just bitmaps extension), QCOW -> QCOW2 (bitmaps are only for qcow2)
v8:
Patches 01-06 are mostly unchanged, except for spelling and wording. 02 - add Eric's r-b
03,04 - add Max's r-b
07-09 is refactored "bitmap-read" part of the feature. Main changes:
- introduce bitmap list - normal list, which represents bitmap directory. Function bitmap_list_load loads bitmap directory, checks everything and creates normal list.
- introduce .bdrv_load_autoloading_dirty_bitmaps : instead of loading bitmaps in qcow2_open, lets load them only in bdrv_open, to avoid reloading in bdrv_snapshot_goto
- do not delete bitmaps after read, but mark them IN_USE
10 has contextual changes and rewording of comment. I've added Max's r-b, hope it's ok.
11: add error_report("Persistent bitmaps are lost") in case of failed bitmap store
12: add Max's r-b
13 is refactored "bitmap-store" part of the feature. see 07-09 description
- for now I just free old clusters and allocate new. This will be improved with a separate patch.
patch about "delete bitmaps on truncate" is removed. This case will be handled later. IN_USE, autoclear, check-constraints things are merged into other patches.
14-15 are mew patches, to early check possibility of creating persistent bitmap with
specified name and granularity in specified BDS
16-17: spelling, rewording, indenting, tiny code simplifying, check can_store (by 14-15),
s/if (autoload && !persistent)/if (has_autoload && !persistent)/, rebase (deleted qmp-commands.hx)
18: alternative to md5 in query-block:
- separated qmp command -x-debug-block-dirty-bitmap-sha256
- as adviced by Daniel P. Berrange in my parallel thread:
- sha256 instead of md5
- use qcrypto_hash_* instead of GChecksum
- fix bug =) (size was wrong in hbitmap_md5)
19: s/3999/3fff/, use x-debug-block-dirty-bitmap-sha256
20: new patch to rename and publish inc_refcounts
21: some fixes and refactoring mostyly by Max's comments.
Max, Eric, great tanks for your review!
I hope, I've covered most of your comments by this update.
TODO:
- handle reopening image RO->RW and incoming migration, set IN_USE for existing loaded bitmaps in these cases.
- reuse old, already allocated data clusters for bitmaps storing
- truncate bitmaps in the image on truncate
v7:
https://src.openvz.org/users/vsementsov/repos/qemu/browse?at=refs%2Ftags%2Fqcow2-bitmap-v7
based on block-next (https://github.com/XanClic/qemu/commits/block-next)
- a lot of refactoring and reordering of patches.
- dead code removed (bdrv_dirty_bitmap_load, etc.)
- do not maintain extra data for now
- do not store dirty bitmap directory in memory (as we use it seldom, we can just reread if needed)
By Kevin's review:
01 - commit message changed: fix->improvement (as it was not a bug) 03 - r-b
04 - r-b
05 - add 21 patch to fix spec, also, removed all (I hope) mentions of
"Bitmap Header", switch to one unified name for it - "Bitmap Directory Entry", to avoid misunderstanding with Qcow2 header. (also, add patch 22, to fix it in spec)
v6.06 - improve for_each_dir_entry loop, reorder patches, other small fixes v6.07 ~> v7.09 - dead code removed, I've moved to one function
.bdrv_store_persistent_bitmaps and have wrapper and callback in one patch (with also some other staff. If it not ok, I can split them)
v6.08 - about keeping bitmap directory instead of bitmap list: no I don't keep
it at all.
v6.16 - old bdrv_ bitmap-storing related functions are removed. The new one is
bdrv_store_persistent_bitmaps.
v6:
https://src.openvz.org/users/vsementsov/repos/qemu/browse?at=refs%2Ftags%2Fqcow2-bitmap-v6
based on block-next (https://github.com/XanClic/qemu/commits/block-next)
There are a lot of changes, reorderings and additions in comparement with v5. One principal thing: now bitmaps are removed from image after loading instead of marking them in_use. It is simpler and we do not need to store superfluous data. Also, we are no more interested in command line interface to dirty bitmaps. So it is dropped. If someone needs it I can add it later.
Vladimir Sementsov-Ogievskiy (30):
specs/qcow2: fix bitmap granularity qemu-specific note specs/qcow2: do not use wording 'bitmap header'
hbitmap: improve dirty iter
tests: add hbitmap iter test
block: fix bdrv_dirty_bitmap_granularity signature block/dirty-bitmap: add deserialize_ones func
qcow2-refcount: rename inc_refcounts() and make it public
qcow2: add bitmaps extension
block/dirty-bitmap: fix comment for BlockDirtyBitmap.disabled field block/dirty-bitmap: add readonly field to BdrvDirtyBitmap
qcow2: autoloading dirty bitmaps
block: refactor bdrv_reopen_commit
block: new bdrv_reopen_bitmaps_rw interface
qcow2: support .bdrv_reopen_bitmaps_rw block/dirty-bitmap: add autoload field to BdrvDirtyBitmap
block: bdrv_close: release bitmaps after drv->bdrv_close
block: introduce persistent dirty bitmaps block/dirty-bitmap: add bdrv_dirty_bitmap_next()
qcow2: add persistent dirty bitmaps support
qcow2: store bitmaps on reopening image as read-only
block: add bdrv_can_store_new_dirty_bitmap
qcow2: add .bdrv_can_store_new_dirty_bitmap
qmp: add persistent flag to block-dirty-bitmap-add
qmp: add autoload parameter to block-dirty-bitmap-add
qmp: add x-debug-block-dirty-bitmap-sha256
iotests: test qcow2 persistent dirty bitmap block/dirty-bitmap: add bdrv_remove_persistent_dirty_bitmap
qcow2: add .bdrv_remove_persistent_dirty_bitmap
qmp: block-dirty-bitmap-remove: remove persistent
block: release persistent bitmaps on inactivate
block.c | 65 +- block/Makefile.objs | 2 +- block/dirty-bitmap.c | 154 ++++- block/io.c | 8 + block/qcow2-bitmap.c | 1481 ++++++++++++++++++++++++++++++++++++++++++ block/qcow2-refcount.c | 59 +- block/qcow2.c | 155 ++++- block/qcow2.h | 43 ++ blockdev.c | 73 ++- docs/interop/qcow2.txt | 8 +- include/block/block.h | 3 + include/block/block_int.h | 14 + include/block/dirty-bitmap.h | 22 +- include/qemu/hbitmap.h | 49 +- qapi/block-core.json | 42 +- tests/Makefile.include | 2 +- tests/qemu-iotests/165 | 105 +++ tests/qemu-iotests/165.out | 5 + tests/qemu-iotests/group | 1 + tests/test-hbitmap.c | 19 + util/hbitmap.c | 51 +- 21 files changed, 2280 insertions(+), 81 deletions(-) create mode 100644 block/qcow2-bitmap.c create mode 100755 tests/qemu-iotests/165 create mode 100644 tests/qemu-iotests/165.out
Signature · 1 line
2.11.1
PATCH v22 1/30 specs/qcow2: fix bitmap granularity qemu-specific note
--- docs/interop/qcow2.txt | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)
docs/interop/qcow2.txt +1 -2
@@ -472,8 +472,7 @@ Structure of a bitmap directory entry: 17: granularity_bits Granularity bits. Valid values: 0 - 63. - Note: Qemu currently doesn't support granularity_bits - greater than 31. + Note: Qemu currently supports only values 9 - 31. Granularity is calculated as granularity = 1 << granularity_bits
Signature · 1 line
2.11.1
PATCH v22 2/30 specs/qcow2: do not use wording 'bitmap header'
In reply to Vladimir Sementsov-OgievskiyA bitmap directory entry is sometimes called a 'bitmap header'. This patch leaves only one name - 'bitmap directory entry'. The name 'bitmap header' creates misunderstandings with 'qcow2 header' and 'qcow2 bitmap header extension' (which is extension of qcow2 header)
--- docs/interop/qcow2.txt | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-)
docs/interop/qcow2.txt +2 -3
@@ -201,7 +201,7 @@ The fields of the bitmaps extension are: 8 - 15: bitmap_directory_size Size of the bitmap directory in bytes. It is the cumulative - size of all (nb_bitmaps) bitmap headers. + size of all (nb_bitmaps) bitmap directory entries. 16 - 23: bitmap_directory_offset Offset into the image file at which the bitmap directory
@@ -426,8 +426,7 @@ Each bitmap saved in the image is described in a bitmap directory entry. The bitmap directory is a contiguous area in the image file, whose starting offset and length are given by the header extension fields bitmap_directory_offset and bitmap_directory_size. The entries of the bitmap directory have variable -length, depending on the lengths of the bitmap name and extra data. These -entries are also called bitmap headers. +length, depending on the lengths of the bitmap name and extra data. Structure of a bitmap directory entry:
Signature · 1 line
2.11.1
PATCH v22 3/30 hbitmap: improve dirty iter
In reply to Vladimir Sementsov-OgievskiyMake dirty iter resistant to resetting bits in corresponding HBitmap.
--- include/qemu/hbitmap.h | 26 ++++---------------------- util/hbitmap.c | 23 ++++++++++++++++++++++- 2 files changed, 26 insertions(+), 23 deletions(-)
include/qemu/hbitmap.h +4 -22
@@ -256,10 +256,9 @@ void hbitmap_free(HBitmap *hb);* the lowest-numbered bit that is set in @hb, starting at @first. * * Concurrent setting of bits is acceptable, and will at worst cause the -* iteration to miss some of those bits. Resetting bits before the current -* position of the iterator is also okay. However, concurrent resetting of -* bits can lead to unexpected behavior if the iterator has not yet reached -* those bits. +* iteration to miss some of those bits. +* +* The concurrent resetting of bits is OK. */ void hbitmap_iter_init (HBitmapIter *hbi,const HBitmap *hb,uint64_t first);
@@ -298,24 +297,7 @@ void hbitmap_free_meta(HBitmap *hb);* Return the next bit that is set in @hbi's associated HBitmap, * or -1 if all remaining bits are zero. */ -static inline int64_t hbitmap_iter_next (HBitmapIter *hbi) -{ -unsigned long cur = hbi->cur[HBITMAP_LEVELS -1 ]; -int64_t item; - -if (cur ==0 ) { - cur =hbitmap_iter_skip_words (hbi); -if (cur ==0 ) { -return -1 ; - } - } - -/* The next call will resume work from the next bit. */ - hbi->cur[HBITMAP_LEVELS -1 ] = cur & (cur -1 ); - item = ((uint64_t )hbi->pos <<BITS_PER_LEVEL ) +ctzl (cur); - -return item << hbi->granularity; -} +int64_t hbitmap_iter_next (HBitmapIter *hbi);/** * hbitmap_iter_next_word:
util/hbitmap.c +22 -1
@@ -106,8 +106,9 @@ unsigned long hbitmap_iter_skip_words(HBitmapIter *hbi)unsigned long cur;do { - cur = hbi->cur[--i]; + i--; pos >>=BITS_PER_LEVEL ; + cur = hbi->cur[i] & hb->levels[i][pos]; }while (cur ==0 );/* Check for end of iteration. We always use fewer than BITS_PER_LONG
@@ -139,6 +140,26 @@ unsigned long hbitmap_iter_skip_words(HBitmapIter *hbi)return cur; } +int64_t hbitmap_iter_next (HBitmapIter *hbi) +{ +unsigned long cur = hbi->cur[HBITMAP_LEVELS -1 ] & + hbi->hb->levels[HBITMAP_LEVELS -1 ][hbi->pos]; +int64_t item; + +if (cur ==0 ) { + cur =hbitmap_iter_skip_words (hbi); +if (cur ==0 ) { +return -1 ; + } + } + +/* The next call will resume work from the next bit. */ + hbi->cur[HBITMAP_LEVELS -1 ] = cur & (cur -1 ); + item = ((uint64_t )hbi->pos <<BITS_PER_LEVEL ) +ctzl (cur); + +return item << hbi->granularity; +} +void hbitmap_iter_init (HBitmapIter *hbi,const HBitmap *hb,uint64_t first) {unsigned i, bit;
Signature · 1 line
2.11.1
PATCH v22 4/30 tests: add hbitmap iter test
In reply to Vladimir Sementsov-OgievskiyTest that hbitmap iter is resistant to bitmap resetting.
--- tests/test-hbitmap.c | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+)
tests/test-hbitmap.c +19 -0
@@ -909,6 +909,22 @@ static void hbitmap_test_add(const char *testpath, hbitmap_test_teardown); } +static void test_hbitmap_iter_and_reset (TestHBitmapData *data, +const void *unused) +{ +HBitmapIter hbi; + +hbitmap_test_init (data,L1 *2 ,0 ); +hbitmap_set (data->hb,0 , data->size); + +hbitmap_iter_init (&hbi, data->hb,BITS_PER_LONG -1 ); + +hbitmap_iter_next (&hbi); + +hbitmap_reset_all (data->hb); +hbitmap_iter_next (&hbi); +} +int main (int argc,char **argv) {g_test_init (&argc, &argv,NULL );
@@ -966,6 +982,9 @@ int main(int argc, char **argv) test_hbitmap_serialize_part);hbitmap_test_add ("/hbitmap/serialize/zeroes" , test_hbitmap_serialize_zeroes); + +hbitmap_test_add ("/hbitmap/iter/iter_and_reset" , + test_hbitmap_iter_and_reset);g_test_run ();return 0 ;
Signature · 1 line
2.11.1
PATCH v22 5/30 block: fix bdrv_dirty_bitmap_granularity signature
In reply to Vladimir Sementsov-OgievskiyMake getter signature const-correct. This allows other functions with const dirty bitmap parameter use bdrv_dirty_bitmap_granularity().
--- block/dirty-bitmap.c | 2 +- include/block/dirty-bitmap.h | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
block/dirty-bitmap.c +1 -1
@@ -455,7 +455,7 @@ uint32_t bdrv_get_default_bitmap_granularity(BlockDriverState *bs)return granularity; } -uint32_t bdrv_dirty_bitmap_granularity (BdrvDirtyBitmap *bitmap) +uint32_t bdrv_dirty_bitmap_granularity (const BdrvDirtyBitmap *bitmap) {return BDRV_SECTOR_SIZE <<hbitmap_granularity (bitmap->bitmap); }
include/block/dirty-bitmap.h +1 -1
@@ -29,7 +29,7 @@ void bdrv_disable_dirty_bitmap(BdrvDirtyBitmap *bitmap);void bdrv_enable_dirty_bitmap (BdrvDirtyBitmap *bitmap);BlockDirtyInfoList *bdrv_query_dirty_bitmaps (BlockDriverState *bs);uint32_t bdrv_get_default_bitmap_granularity (BlockDriverState *bs); -uint32_t bdrv_dirty_bitmap_granularity (BdrvDirtyBitmap *bitmap); +uint32_t bdrv_dirty_bitmap_granularity (const BdrvDirtyBitmap *bitmap);uint32_t bdrv_dirty_bitmap_meta_granularity (BdrvDirtyBitmap *bitmap);bool bdrv_dirty_bitmap_enabled (BdrvDirtyBitmap *bitmap);bool bdrv_dirty_bitmap_frozen (BdrvDirtyBitmap *bitmap);
Signature · 1 line
2.11.1
PATCH v22 6/30 block/dirty-bitmap: add deserialize_ones func
In reply to Vladimir Sementsov-OgievskiyAdd bdrv_dirty_bitmap_deserialize_ones() function, which is needed for qcow2 bitmap loading, to handle unallocated bitmap parts, marked as all-ones.
--- block/dirty-bitmap.c | 7 +++++++ include/block/dirty-bitmap.h | 3 +++ include/qemu/hbitmap.h | 15 +++++++++++++++ util/hbitmap.c | 17 +++++++++++++++++ 4 files changed, 42 insertions(+)
block/dirty-bitmap.c +7 -0
@@ -586,6 +586,13 @@ void bdrv_dirty_bitmap_deserialize_zeroes(BdrvDirtyBitmap *bitmap,hbitmap_deserialize_zeroes (bitmap->bitmap, start, count, finish); } +void bdrv_dirty_bitmap_deserialize_ones (BdrvDirtyBitmap *bitmap, +uint64_t start,uint64_t count, +bool finish) +{ +hbitmap_deserialize_ones (bitmap->bitmap, start, count, finish); +} +void bdrv_dirty_bitmap_deserialize_finish (BdrvDirtyBitmap *bitmap) {hbitmap_deserialize_finish (bitmap->bitmap);
include/block/dirty-bitmap.h +3 -0
@@ -66,6 +66,9 @@ void bdrv_dirty_bitmap_deserialize_part(BdrvDirtyBitmap *bitmap,void bdrv_dirty_bitmap_deserialize_zeroes (BdrvDirtyBitmap *bitmap,uint64_t start,uint64_t count,bool finish); +void bdrv_dirty_bitmap_deserialize_ones (BdrvDirtyBitmap *bitmap, +uint64_t start,uint64_t count, +bool finish);void bdrv_dirty_bitmap_deserialize_finish (BdrvDirtyBitmap *bitmap);/* Functions that require manual locking. */
include/qemu/hbitmap.h +15 -0
@@ -229,6 +229,21 @@ void hbitmap_deserialize_zeroes(HBitmap *hb, uint64_t start, uint64_t count, bool finish);/** +* hbitmap_deserialize_ones +* @hb: HBitmap to operate on. +* @start: First bit to restore. +* @count: Number of bits to restore. +* @finish: Whether to call hbitmap_deserialize_finish automatically. +* +* Fills the bitmap with ones. +* +* If @finish is false, caller must call hbitmap_serialize_finish before using +* the bitmap. +*/ +void hbitmap_deserialize_ones (HBitmap *hb,uint64_t start,uint64_t count, +bool finish); + +/** * hbitmap_deserialize_finish * @hb: HBitmap to operate on. *
util/hbitmap.c +17 -0
@@ -551,6 +551,23 @@ void hbitmap_deserialize_zeroes(HBitmap *hb, uint64_t start, uint64_t count, } } +void hbitmap_deserialize_ones (HBitmap *hb,uint64_t start,uint64_t count, +bool finish) +{ +uint64_t el_count; +unsigned long *first; + +if (!count) { +return ; + } +serialization_chunk (hb, start, count, &first, &el_count); + +memset (first,0xff , el_count *sizeof (unsigned long )); +if (finish) { +hbitmap_deserialize_finish (hb); + } +} +void hbitmap_deserialize_finish (HBitmap *bitmap) {int64_t i, size, prev_size;
Signature · 1 line
2.11.1
Re: [Qemu-devel] [PATCH v22 06/30] block/dirty-bitmap: add deserialize_ones func
Quoted 19 lines · Vladimir Sementsov-Ogievskiy
Add
bdrv_dirty_bitmap_deserialize_ones()function, which is needed for qcow2 bitmap loading, to handle unallocated bitmap parts, marked as all-ones.Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>Reviewed-by Kevin Wolf <[email protected]>Reviewed-by John Snow <[email protected]>--- block/dirty-bitmap.c | 7 +++++++ include/block/dirty-bitmap.h | 3 +++ include/qemu/hbitmap.h | 15 +++++++++++++++ util/hbitmap.c | 17 +++++++++++++++++ 4 files changed, 42 insertions(+)block/dirty-bitmap.c +0
-0
block/dirty-bitmap.c void bdrv_dirty_bitmap_deserialize_zeroes(BdrvDirtyBitmap *bitmap, · line 594
Quoted 9 lines
block/dirty-bitmap.c +6
-0@@ -586,6 +586,13 @@ void bdrv_dirty_bitmap_deserialize_zeroes(BdrvDirtyBitmap *bitmap,hbitmap_deserialize_zeroes (bitmap->bitmap, start, count, finish); } +void bdrv_dirty_bitmap_deserialize_ones (BdrvDirtyBitmap *bitmap, +uint64_t start,uint64_t count, +bool finish) +{ +hbitmap_deserialize_ones (bitmap->bitmap, start, count, finish); +}
What unit is 'count' in?
/me reads the code, finding serialization_chunk()
Oh, it's in terms of the underlying bitmap->bitmap->granularity (currently the number of sectors). Eww - I'd rather we name the parameter 'bytes' rather than 'count', and then scale it appropriately when passing to the underlying bitmap->bitmap, rather than making the user guess what scale to use.
But as you are just copying from the existing deserialize_zeroes right above, I'd rather keep the semantics the same between the two, so my series will take care of fixing that up, and your series can go in as-is for this patch.
block/dirty-bitmap.c void hbitmap_deserialize_zeroes(HBitmap *hb, uint64_t start, uint64_t count, · line 235
Quoted 9 lines
block/dirty-bitmap.c +5
-0+++ b/include/qemu/hbitmap.h@@ -229,6 +229,21 @@ void hbitmap_deserialize_zeroes(HBitmap *hb, uint64_t start, uint64_t count, bool finish);/** +* hbitmap_deserialize_ones +* @hb: HBitmap to operate on. +* @start: First bit to restore. +* @count: Number of bits to restore.
This part is accurate (the dirty-bitmap is using an underlying bitmap with "one bit per sector" before my series, afterwards it will be "one bit per byte", remembering that hbitmap really stores only one bit per granularity multiple of the underlying unit), if incomplete (the code asserts that things are aligned, but doesn't document that the caller must pass in aligned values); but again, that's matching the pre-existing deserialize_zeroes code.
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
Re: [Qemu-devel] [PATCH v22 06/30] block/dirty-bitmap: add deserialize_ones func
Quoted 11 lines · Eric Blake
Quoted 9 lines · Vladimir Sementsov-Ogievskiy
Add
bdrv_dirty_bitmap_deserialize_ones()function, which is needed for qcow2 bitmap loading, to handle unallocated bitmap parts, marked as all-ones.+ *
hbitmap_deserialize_ones
+ * @hb: HBitmap to operate on.
+ * @start: First bit to restore.
+ * @count: Number of bits to restore.
Quoted 7 lines
This part is accurate (the dirty-bitmap is using an underlying bitmap with "one bit per sector" before my series, afterwards it will be "one bit per byte", remembering that hbitmap really stores only one bit per granularity multiple of the underlying unit), if incomplete (the code asserts that things are aligned, but doesn't document that the caller must pass in aligned values); but again, that's matching the pre-existing
deserialize_zeroescode.
Okay, I looked again; the documentation for hbitmap_serialization_granularity() has a blanket statement that all other hbitmap_serialization_* functions taking a start and count must be aligned. Indirect, but at least documented, so I retract my statement about the docs being incomplete.
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
Re: [Qemu-devel] [PATCH v22 06/30] block/dirty-bitmap: add deserialize_ones func
Quoted 20 lines · Eric Blake
Quoted 11 lines · Eric Blake
Quoted 9 lines · Vladimir Sementsov-Ogievskiy
Add
bdrv_dirty_bitmap_deserialize_ones()function, which is needed for qcow2 bitmap loading, to handle unallocated bitmap parts, marked as all-ones.+ *
hbitmap_deserialize_ones
+ * @hb: HBitmap to operate on.
+ * @start: First bit to restore.
+ * @count: Number of bits to restore.Quoted 7 lines
This part is accurate (the dirty-bitmap is using an underlying bitmap with "one bit per sector" before my series, afterwards it will be "one bit per byte", remembering that hbitmap really stores only one bit per granularity multiple of the underlying unit), if incomplete (the code asserts that things are aligned, but doesn't document that the caller must pass in aligned values); but again, that's matching the pre-existing
deserialize_zeroescode.
Quoted 6 lines
Okay, I looked again; the documentation for
hbitmap_serialization_granularity()has a blanket statement that all other hbitmap_serialization_* functions taking a start and count must be aligned. Indirect, but at least documented, so I retract my statement about the docs being incomplete.
If the docs are confusing, please send patches to amend them; the bitmap code is definitely very confusing at times and I appreciate any and all insight to make the names, variable and documentation read better to be more intuitive.
Thanks!
PATCH v22 7/30 qcow2-refcount: rename inc_refcounts() and make it public
In reply to Vladimir Sementsov-OgievskiyThis is needed for the following patch, which will introduce refcounts checking for qcow2 bitmaps.
--- block/qcow2-refcount.c | 53 ++++++++++++++++++++++++++------------------------ block/qcow2.h | 4 ++++ 2 files changed, 32 insertions(+), 25 deletions(-)
block/qcow2-refcount.c +28 -25
@@ -1323,11 +1323,10 @@ static int realloc_refcount_array(BDRVQcow2State *s, void **array,* * Modifies the number of errors in res. */ -static int inc_refcounts (BlockDriverState *bs, -BdrvCheckResult *res, -void **refcount_table, -int64_t *refcount_table_size, -int64_t offset,int64_t size) +int qcow2_inc_refcounts_imrt (BlockDriverState *bs,BdrvCheckResult *res, +void **refcount_table, +int64_t *refcount_table_size, +int64_t offset,int64_t size) {BDRVQcow2State *s = bs->opaque;uint64_t start, last, cluster_offset, k, refcount;
@@ -1420,8 +1419,9 @@ static int check_refcounts_l2(BlockDriverState *bs, BdrvCheckResult *res, nb_csectors = ((l2_entry >> s->csize_shift) & s->csize_mask) +1 ; l2_entry &= s->cluster_offset_mask; - ret =inc_refcounts (bs, res, refcount_table, refcount_table_size, - l2_entry & ~511 , nb_csectors *512 ); + ret =qcow2_inc_refcounts_imrt (bs, res, + refcount_table, refcount_table_size, + l2_entry & ~511 , nb_csectors *512 );if (ret <0 ) { goto fail; }
@@ -1454,8 +1454,9 @@ static int check_refcounts_l2(BlockDriverState *bs, BdrvCheckResult *res, }/* Mark cluster as used */ - ret =inc_refcounts (bs, res, refcount_table, refcount_table_size, - offset, s->cluster_size); + ret =qcow2_inc_refcounts_imrt (bs, res, + refcount_table, refcount_table_size, + offset, s->cluster_size);if (ret <0 ) { goto fail; }
@@ -1508,8 +1509,8 @@ static int check_refcounts_l1(BlockDriverState *bs, l1_size2 = l1_size *sizeof (uint64_t );/* Mark L1 table as used */ - ret =inc_refcounts (bs, res, refcount_table, refcount_table_size, - l1_table_offset, l1_size2); + ret =qcow2_inc_refcounts_imrt (bs, res, refcount_table, refcount_table_size, + l1_table_offset, l1_size2);if (ret <0 ) { goto fail; }
@@ -1538,8 +1539,9 @@ static int check_refcounts_l1(BlockDriverState *bs,if (l2_offset) {/* Mark L2 table as used */ l2_offset &=L1E_OFFSET_MASK ; - ret =inc_refcounts (bs, res, refcount_table, refcount_table_size, - l2_offset, s->cluster_size); + ret =qcow2_inc_refcounts_imrt (bs, res, + refcount_table, refcount_table_size, + l2_offset, s->cluster_size);if (ret <0 ) { goto fail; }
@@ -1757,14 +1759,15 @@ static int check_refblocks(BlockDriverState *bs, BdrvCheckResult *res, } res->corruptions_fixed++; - ret =inc_refcounts (bs, res, refcount_table, nb_clusters, - offset, s->cluster_size); + ret =qcow2_inc_refcounts_imrt (bs, res, + refcount_table, nb_clusters, + offset, s->cluster_size);if (ret <0 ) {return ret; }/* No need to check whether the refcount is now greater than 1: * This area was just allocated and zeroed, so it can only be -* exactly 1 after inc_refcounts() */ +* exactly 1 after qcow2_inc_refcounts_imrt() */ continue ; resize_fail:
@@ -1779,8 +1782,8 @@ resize_fail: }if (offset !=0 ) { - ret =inc_refcounts (bs, res, refcount_table, nb_clusters, - offset, s->cluster_size); + ret =qcow2_inc_refcounts_imrt (bs, res, refcount_table, nb_clusters, + offset, s->cluster_size);if (ret <0 ) {return ret; }
@@ -1820,8 +1823,8 @@ static int calculate_refcounts(BlockDriverState *bs, BdrvCheckResult *res, }/* header */ - ret =inc_refcounts (bs, res, refcount_table, nb_clusters, -0 , s->cluster_size); + ret =qcow2_inc_refcounts_imrt (bs, res, refcount_table, nb_clusters, +0 , s->cluster_size);if (ret <0 ) {return ret; }
@@ -1842,16 +1845,16 @@ static int calculate_refcounts(BlockDriverState *bs, BdrvCheckResult *res,return ret; } } - ret =inc_refcounts (bs, res, refcount_table, nb_clusters, - s->snapshots_offset, s->snapshots_size); + ret =qcow2_inc_refcounts_imrt (bs, res, refcount_table, nb_clusters, + s->snapshots_offset, s->snapshots_size);if (ret <0 ) {return ret; }/* refcount data */ - ret =inc_refcounts (bs, res, refcount_table, nb_clusters, - s->refcount_table_offset, - s->refcount_table_size *sizeof (uint64_t )); + ret =qcow2_inc_refcounts_imrt (bs, res, refcount_table, nb_clusters, + s->refcount_table_offset, + s->refcount_table_size *sizeof (uint64_t ));if (ret <0 ) {return ret; }
block/qcow2.h +4 -0
@@ -527,6 +527,10 @@ int qcow2_check_metadata_overlap(BlockDriverState *bs, int ign, int64_t offset, int64_t size);int qcow2_pre_write_overlap_check (BlockDriverState *bs,int ign,int64_t offset,int64_t size); +int qcow2_inc_refcounts_imrt (BlockDriverState *bs,BdrvCheckResult *res, +void **refcount_table, +int64_t *refcount_table_size, +int64_t offset,int64_t size); int qcow2_change_refcount_order(BlockDriverState *bs, int refcount_order, BlockDriverAmendStatusCB *status_cb,
Signature · 1 line
2.11.1
PATCH v22 8/30 qcow2: add bitmaps extension
In reply to Vladimir Sementsov-OgievskiyAdd bitmap extension as specified in docs/specs/qcow2.txt. For now, just mirror extension header into Qcow2 state and check constraints. Also, calculate refcounts for qcow2 bitmaps, to not break qemu-img check.
For now, disable image resize if it has bitmaps. It will be fixed later.
--- block/Makefile.objs | 2 +- block/qcow2-bitmap.c | 439 +++++++++++++++++++++++++++++++++++++++++++++++++ block/qcow2-refcount.c | 6 + block/qcow2.c | 124 +++++++++++++- block/qcow2.h | 27 +++ 5 files changed, 592 insertions(+), 6 deletions(-) create mode 100644 block/qcow2-bitmap.c
block/Makefile.objs +1 -1
@@ -1,5 +1,5 @@ block-obj-y += raw-format.o qcow.o vdi.o vmdk.o cloop.o bochs.o vpc.o vvfat.o dmg.o -block-obj-y += qcow2.o qcow2-refcount.o qcow2-cluster.o qcow2-snapshot.o qcow2-cache.o +block-obj-y += qcow2.o qcow2-refcount.o qcow2-cluster.o qcow2-snapshot.o qcow2-cache.o qcow2-bitmap.o block-obj-y += qed.o qed-gencb.o qed-l2-cache.o qed-table.o qed-cluster.o block-obj-y += qed-check.o block-obj-y += vhdx.o vhdx-endian.o vhdx-log.o
block/qcow2-bitmap.c new file +439 -0
439 lines not on this page · raw message
block/qcow2-refcount.c +6 -0
@@ -1859,6 +1859,12 @@ static int calculate_refcounts(BlockDriverState *bs, BdrvCheckResult *res,return ret; } +/* bitmaps */ + ret =qcow2_check_bitmaps_refcounts (bs, res, refcount_table, nb_clusters); +if (ret <0 ) { +return ret; + } +return check_refblocks (bs, res, fix, rebuild, refcount_table, nb_clusters); }
block/qcow2.c +119 -5
183 lines not on this page · raw message
block/qcow2.h +27 -0
@@ -52,6 +52,10 @@* space for snapshot names and IDs */ #define QCOW_MAX_SNAPSHOTS_SIZE (1024 * QCOW_MAX_SNAPSHOTS) +/* Bitmap header extension constraints */ +#define QCOW2_MAX_BITMAPS 65535 +#define QCOW2_MAX_BITMAP_DIRECTORY_SIZE (1024 * QCOW2_MAX_BITMAPS) +/* indicate that the refcount of the referenced cluster is exactly one. */ #define QCOW_OFLAG_COPIED (1ULL << 63)/* indicate that the cluster is compressed (they never have the copied flag) */
@@ -195,6 +199,14 @@ enum { QCOW2_COMPAT_FEAT_MASK = QCOW2_COMPAT_LAZY_REFCOUNTS, }; +/* Autoclear feature bits */ +enum { +QCOW2_AUTOCLEAR_BITMAPS_BITNR =0 , +QCOW2_AUTOCLEAR_BITMAPS =1 <<QCOW2_AUTOCLEAR_BITMAPS_BITNR , + +QCOW2_AUTOCLEAR_MASK =QCOW2_AUTOCLEAR_BITMAPS , +}; +enum qcow2_discard_type {QCOW2_DISCARD_NEVER =0 ,QCOW2_DISCARD_ALWAYS ,
@@ -222,6 +234,13 @@ typedef uint64_t Qcow2GetRefcountFunc(const void *refcount_array,typedef void Qcow2SetRefcountFunc (void *refcount_array,uint64_t index,uint64_t value); +typedef struct Qcow2BitmapHeaderExt { +uint32_t nb_bitmaps; +uint32_t reserved32; +uint64_t bitmap_directory_size; +uint64_t bitmap_directory_offset; +} QEMU_PACKED Qcow2BitmapHeaderExt; + typedef struct BDRVQcow2State { int cluster_bits; int cluster_size;
@@ -264,6 +283,10 @@ typedef struct BDRVQcow2State {unsigned int nb_snapshots;QCowSnapshot *snapshots; +uint32_t nb_bitmaps; +uint64_t bitmap_directory_size; +uint64_t bitmap_directory_offset; +int flags;int qcow_version;bool use_lazy_refcounts;
@@ -602,4 +625,8 @@ int qcow2_cache_get_empty(BlockDriverState *bs, Qcow2Cache *c, uint64_t offset, void **table);void qcow2_cache_put (BlockDriverState *bs,Qcow2Cache *c,void **table); +/* qcow2-bitmap.c functions */ +int qcow2_check_bitmaps_refcounts (BlockDriverState *bs,BdrvCheckResult *res, +void **refcount_table, +int64_t *refcount_table_size);#endif
Signature · 1 line
2.11.1
PATCH v22 9/30 block/dirty-bitmap: fix comment for BlockDirtyBitmap.disabled field
In reply to Vladimir Sementsov-Ogievskiy--- block/dirty-bitmap.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
block/dirty-bitmap.c +2 -1
@@ -43,7 +43,8 @@ struct BdrvDirtyBitmap {BdrvDirtyBitmap *successor;/* Anonymous child; implies frozen status */ char *name;/* Optional non-empty unique ID */ int64_t size;/* Size of the bitmap (Number of sectors) */ -bool disabled;/* Bitmap is read-only */ +bool disabled;/* Bitmap is disabled. It ignores all writes to +the device */ int active_iterators;/* How many iterators are active */ QLIST_ENTRY (BdrvDirtyBitmap) list; };
Signature · 1 line
2.11.1
PATCH v22 10/30 block/dirty-bitmap: add readonly field to BdrvDirtyBitmap
In reply to Vladimir Sementsov-OgievskiyIt will be needed in following commits for persistent bitmaps. If bitmap is loaded from read-only storage (and we can't mark it "in use" in this storage) corresponding BdrvDirtyBitmap should be read-only.
--- block/dirty-bitmap.c | 36 ++++++++++++++++++++++++++++++++++++ block/io.c | 8 ++++++++ blockdev.c | 6 ++++++ include/block/dirty-bitmap.h | 4 ++++ 4 files changed, 54 insertions(+)
block/dirty-bitmap.c +36 -0
@@ -46,6 +46,12 @@ struct BdrvDirtyBitmap {bool disabled;/* Bitmap is disabled. It ignores all writes to the device */ int active_iterators;/* How many iterators are active */ +bool readonly;/* Bitmap is read-only. This field also +prevents the respective image from being +modified (i.e. blocks writes and discards). +Such operations must fail and both the image +and this bitmap must remain unchanged while +this flag is set. */ QLIST_ENTRY (BdrvDirtyBitmap) list; };
@@ -505,6 +511,7 @@ void bdrv_set_dirty_bitmap_locked(BdrvDirtyBitmap *bitmap, int64_t cur_sector, int64_t nr_sectors) {assert (bdrv_dirty_bitmap_enabled (bitmap )); +assert (!bdrv_dirty_bitmap_readonly (bitmap));hbitmap_set (bitmap->bitmap, cur_sector, nr_sectors); }
@@ -521,6 +528,7 @@ void bdrv_reset_dirty_bitmap_locked(BdrvDirtyBitmap *bitmap, int64_t cur_sector, int64_t nr_sectors) {assert (bdrv_dirty_bitmap_enabled (bitmap )); +assert (!bdrv_dirty_bitmap_readonly (bitmap));hbitmap_reset (bitmap->bitmap, cur_sector, nr_sectors); }
@@ -535,6 +543,7 @@ void bdrv_reset_dirty_bitmap(BdrvDirtyBitmap *bitmap,void bdrv_clear_dirty_bitmap (BdrvDirtyBitmap *bitmap,HBitmap **out) {assert (bdrv_dirty_bitmap_enabled (bitmap)); +assert (!bdrv_dirty_bitmap_readonly (bitmap));bdrv_dirty_bitmap_lock (bitmap);if (!out) {hbitmap_reset_all (bitmap->bitmap);
@@ -551,6 +560,7 @@ void bdrv_undo_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *in) {HBitmap *tmp = bitmap->bitmap;assert (bdrv_dirty_bitmap_enabled (bitmap)); +assert (!bdrv_dirty_bitmap_readonly (bitmap)); bitmap->bitmap = in;hbitmap_free (tmp); }
@@ -613,6 +623,7 @@ void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector,if (!bdrv_dirty_bitmap_enabled (bitmap)) {continue ; } +assert (!bdrv_dirty_bitmap_readonly (bitmap));hbitmap_set (bitmap->bitmap, cur_sector, nr_sectors); }bdrv_dirty_bitmaps_unlock (bs);
@@ -635,3 +646,28 @@ int64_t bdrv_get_meta_dirty_count(BdrvDirtyBitmap *bitmap) {return hbitmap_count (bitmap->meta); } + +bool bdrv_dirty_bitmap_readonly (const BdrvDirtyBitmap *bitmap) +{ +return bitmap->readonly; +} + +/* Called with BQL taken. */ +void bdrv_dirty_bitmap_set_readonly (BdrvDirtyBitmap *bitmap,bool value) +{ +qemu_mutex_lock (bitmap->mutex); + bitmap->readonly = value; +qemu_mutex_unlock (bitmap->mutex); +} + +bool bdrv_has_readonly_bitmaps (BlockDriverState *bs) +{ +BdrvDirtyBitmap *bm; + QLIST_FOREACH(bm, &bs->dirty_bitmaps, list) { +if (bm->readonly) { +return true; + } + } + +return false; +}
block/io.c +8 -0
@@ -1343,6 +1343,10 @@ static int coroutine_fn bdrv_aligned_pwritev(BdrvChild *child,uint64_t bytes_remaining = bytes;int max_transfer; +if (bdrv_has_readonly_bitmaps (bs)) { +return -EPERM ; + } +assert (is_power_of_2 (align));assert ((offset & (align -1 )) ==0 );assert ((bytes & (align -1 )) ==0 );
@@ -2435,6 +2439,10 @@ int coroutine_fn bdrv_co_pdiscard(BlockDriverState *bs, int64_t offset,return -ENOMEDIUM ; } +if (bdrv_has_readonly_bitmaps (bs)) { +return -EPERM ; + } + ret =bdrv_check_byte_request (bs, offset, count);if (ret <0 ) {return ret;
blockdev.c +6 -0
@@ -2023,6 +2023,9 @@ static void block_dirty_bitmap_clear_prepare(BlkActionState *common, }else if (!bdrv_dirty_bitmap_enabled (state->bitmap)) {error_setg (errp,"Cannot clear a disabled bitmap" );return ; + }else if (bdrv_dirty_bitmap_readonly (state->bitmap)) { +error_setg (errp,"Cannot clear a readonly bitmap" ); +return ; }bdrv_clear_dirty_bitmap (state->bitmap, &state->backup);
@@ -2791,6 +2794,9 @@ void qmp_block_dirty_bitmap_clear(const char *node, const char *name,"Bitmap '%s' is currently disabled and cannot be cleared" , name);return ; + }else if (bdrv_dirty_bitmap_readonly (bitmap)) { +error_setg (errp,"Bitmap '%s' is readonly and cannot be cleared" , name); +return ; }bdrv_clear_dirty_bitmap (bitmap,NULL );
include/block/dirty-bitmap.h +4 -0
@@ -71,6 +71,8 @@ void bdrv_dirty_bitmap_deserialize_ones(BdrvDirtyBitmap *bitmap, bool finish);void bdrv_dirty_bitmap_deserialize_finish (BdrvDirtyBitmap *bitmap); +void bdrv_dirty_bitmap_set_readonly (BdrvDirtyBitmap *bitmap,bool value); +/* Functions that require manual locking. */ void bdrv_dirty_bitmap_lock (BdrvDirtyBitmap *bitmap);void bdrv_dirty_bitmap_unlock (BdrvDirtyBitmap *bitmap);
@@ -85,5 +87,7 @@ void bdrv_set_dirty_iter(BdrvDirtyBitmapIter *hbi, int64_t sector_num);int64_t bdrv_get_dirty_count (BdrvDirtyBitmap *bitmap);int64_t bdrv_get_meta_dirty_count (BdrvDirtyBitmap *bitmap);void bdrv_dirty_bitmap_truncate (BlockDriverState *bs); +bool bdrv_dirty_bitmap_readonly (const BdrvDirtyBitmap *bitmap); +bool bdrv_has_readonly_bitmaps (BlockDriverState *bs);#endif
Signature · 1 line
2.11.1
Re: [Qemu-devel] [PATCH v22 10/30] block/dirty-bitmap: add readonly field to BdrvDirtyBitmap
Quoted 17 lines · Vladimir Sementsov-Ogievskiy
It will be needed in following commits for persistent bitmaps. If bitmap is loaded from read-only storage (and we can't mark it "in use" in this storage) corresponding BdrvDirtyBitmap should be read-only.
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>--- block/dirty-bitmap.c | 36 ++++++++++++++++++++++++++++++++++++ block/io.c | 8 ++++++++ blockdev.c | 6 ++++++ include/block/dirty-bitmap.h | 4 ++++ 4 files changed, 54 insertions(+)block/dirty-bitmap.c +0
-0
[...]
Quoted 7 lines
block/dirty-bitmap.c +3
-0@@ -635,3 +646,28 @@ int64_t bdrv_get_meta_dirty_count(BdrvDirtyBitmap *bitmap) {return hbitmap_count (bitmap->meta); } + +bool bdrv_dirty_bitmap_readonly (const BdrvDirtyBitmap *bitmap) +{
block/dirty-bitmap.c int64_t bdrv_get_meta_dirty_count(BdrvDirtyBitmap *bitmap) · line 661
Quoted 10 lines
block/dirty-bitmap.c +10
-0+return bitmap->readonly; +} + +/* Called with BQL taken. */ +void bdrv_dirty_bitmap_set_readonly (BdrvDirtyBitmap *bitmap,bool value) +{ +qemu_mutex_lock (bitmap->mutex); + bitmap->readonly = value; +qemu_mutex_unlock (bitmap->mutex); +}
I find it a bit strange to acquire the mutex if this function is called with the BQL taken (which I agree it should be) -- but Paolo has given his OK and I don't think it does harm to acquire the mutex even under BQL.
Max
Quoted 12 lines
block/dirty-bitmap.c +12
-0+ +bool bdrv_has_readonly_bitmaps (BlockDriverState *bs) +{ +BdrvDirtyBitmap *bm; + QLIST_FOREACH(bm, &bs->dirty_bitmaps, list) { +if (bm->readonly) { +return true; + } + } + +return false; +}
PATCH v22 11/30 qcow2: autoloading dirty bitmaps
In reply to Vladimir Sementsov-OgievskiyAuto loading bitmaps are bitmaps in Qcow2, with the AUTO flag set. They are loaded when the image is opened and become BdrvDirtyBitmaps for the corresponding drive.
Extra data in bitmaps is not supported for now.
--- block/qcow2-bitmap.c | 389 +++++++++++++++++++++++++++++++++++++++++++++++++++ block/qcow2.c | 17 ++- block/qcow2.h | 2 + 3 files changed, 406 insertions(+), 2 deletions(-)
block/qcow2-bitmap.c +389 -0
@@ -44,6 +44,8 @@/* Bitmap directory entry flags */ #define BME_RESERVED_FLAGS 0xfffffffcU +#define BME_FLAG_IN_USE (1U << 0) +#define BME_FLAG_AUTO (1U << 1)/* bits [1, 8] U [56, 63] are reserved */ #define BME_TABLE_ENTRY_RESERVED_MASK 0xff000000000001feULL
@@ -85,6 +87,23 @@ typedef enum BitmapType { BT_DIRTY_TRACKING_BITMAP = 1 } BitmapType; +static inline bool can_write (BlockDriverState *bs) +{ +return !bdrv_is_read_only (bs) && !(bdrv_get_flags (bs) &BDRV_O_INACTIVE ); +} + +static int update_header_sync (BlockDriverState *bs) +{ +int ret; + + ret =qcow2_update_header (bs); +if (ret <0 ) { +return ret; + } + +return bdrv_flush (bs); +} +static int check_table_entry (uint64_t entry,int cluster_size) {uint64_t offset;
@@ -146,6 +165,120 @@ fail:return ret; } +/* This function returns the number of disk sectors covered by a single qcow2 +* cluster of bitmap data. */ +static uint64_t sectors_covered_by_bitmap_cluster (const BDRVQcow2State *s, +const BdrvDirtyBitmap *bitmap) +{ +uint32_t sector_granularity = +bdrv_dirty_bitmap_granularity (bitmap) >>BDRV_SECTOR_BITS ; + +return (uint64_t )sector_granularity * (s->cluster_size <<3 ); +} + +/* load_bitmap_data +* @bitmap_table entries must satisfy specification constraints. +* @bitmap must be cleared */ +static int load_bitmap_data (BlockDriverState *bs, +const uint64_t *bitmap_table, +uint32_t bitmap_table_size, +BdrvDirtyBitmap *bitmap) +{ +int ret =0 ; +BDRVQcow2State *s = bs->opaque; +uint64_t sector, sbc; +uint64_t bm_size =bdrv_dirty_bitmap_size (bitmap); +uint8_t *buf =NULL ; +uint64_t i, tab_size = +size_to_clusters (s, +bdrv_dirty_bitmap_serialization_size (bitmap,0 , bm_size)); + +if (tab_size != bitmap_table_size || tab_size >BME_MAX_TABLE_SIZE ) { +return -EINVAL ; + } + + buf =g_malloc (s->cluster_size); + sbc =sectors_covered_by_bitmap_cluster (s, bitmap); +for (i =0 , sector =0 ; i < tab_size; ++i, sector += sbc) { +uint64_t count =MIN (bm_size - sector, sbc); +uint64_t entry = bitmap_table[i]; +uint64_t offset = entry &BME_TABLE_ENTRY_OFFSET_MASK ; + +assert (check_table_entry (entry, s->cluster_size) ==0 ); + +if (offset ==0 ) { +if (entry &BME_TABLE_ENTRY_FLAG_ALL_ONES ) { +bdrv_dirty_bitmap_deserialize_ones (bitmap, sector, count, + false); + }else { +/* No need to deserialize zeros because the dirty bitmap is +* already cleared */ + } + }else { + ret =bdrv_pread (bs->file, offset, buf, s->cluster_size); +if (ret <0 ) { + goto finish; + } +bdrv_dirty_bitmap_deserialize_part (bitmap, buf, sector, count, + false); + } + } + ret =0 ; + +bdrv_dirty_bitmap_deserialize_finish (bitmap); + +finish: +g_free (buf); + +return ret; +} + +static BdrvDirtyBitmap *load_bitmap (BlockDriverState *bs, +Qcow2Bitmap *bm,Error **errp) +{ +int ret; +uint64_t *bitmap_table =NULL ; +uint32_t granularity; +BdrvDirtyBitmap *bitmap =NULL ; + +if (bm->flags &BME_FLAG_IN_USE ) { +error_setg (errp,"Bitmap '%s' is in use" , bm->name); + goto fail; + } + + ret =bitmap_table_load (bs, &bm->table, &bitmap_table); +if (ret <0 ) { +error_setg_errno (errp, -ret, +"Could not read bitmap_table table from image for " +"bitmap '%s'" , bm->name); + goto fail; + } + + granularity =1U << bm->granularity_bits; + bitmap =bdrv_create_dirty_bitmap (bs, granularity, bm->name, errp); +if (bitmap ==NULL ) { + goto fail; + } + + ret =load_bitmap_data (bs, bitmap_table, bm->table.size, bitmap); +if (ret <0 ) { +error_setg_errno (errp, -ret,"Could not read bitmap '%s' from image" , + bm->name); + goto fail; + } + +g_free (bitmap_table); +return bitmap; + +fail: +g_free (bitmap_table); +if (bitmap !=NULL ) { +bdrv_release_dirty_bitmap (bs, bitmap); + } + +return NULL ; +} +/* * Bitmap List */
@@ -164,6 +297,15 @@ static inline void bitmap_dir_entry_to_cpu(Qcow2BitmapDirEntry *entry)be32_to_cpus (&entry->extra_data_size); } +static inline void bitmap_dir_entry_to_be (Qcow2BitmapDirEntry *entry) +{ +cpu_to_be64s (&entry->bitmap_table_offset); +cpu_to_be32s (&entry->bitmap_table_size); +cpu_to_be32s (&entry->flags); +cpu_to_be16s (&entry->name_size); +cpu_to_be32s (&entry->extra_data_size); +} +static inline int calc_dir_entry_size (size_t name_size,size_t extra_data_size) { return align_offset(sizeof(Qcow2BitmapDirEntry) +
@@ -224,6 +366,17 @@ static int check_dir_entry(BlockDriverState *bs, Qcow2BitmapDirEntry *entry)return fail ? -EINVAL :0 ; } +static inline void bitmap_directory_to_be (uint8_t *dir,size_t size) +{ +uint8_t *end = dir + size; +while (dir < end) { +Qcow2BitmapDirEntry *e = (Qcow2BitmapDirEntry *)dir; + dir +=dir_entry_size (e); + +bitmap_dir_entry_to_be (e); + } +} +/* * Bitmap List public functions */
@@ -258,6 +411,18 @@ static Qcow2BitmapList *bitmap_list_new(void)return bm_list; } +static uint32_t bitmap_list_count (Qcow2BitmapList *bm_list) +{ +Qcow2Bitmap *bm; +uint32_t nb_bitmaps =0 ; + + QSIMPLEQ_FOREACH(bm, bm_list, entry) { + nb_bitmaps++; + } + +return nb_bitmaps; +} +/* bitmap_list_load * Get bitmap list from qcow2 image. Actually reads bitmap directory, * checks it and convert to bitmap list.
@@ -437,3 +602,227 @@ out:return ret; } + +/* bitmap_list_store +* Store bitmap list to qcow2 image as a bitmap directory. +* Everything is checked. +*/ +static int bitmap_list_store (BlockDriverState *bs,Qcow2BitmapList *bm_list, +uint64_t *offset,uint64_t *size,bool in_place) +{ +int ret; +uint8_t *dir; +int64_t dir_offset =0 ; +uint64_t dir_size =0 ; +Qcow2Bitmap *bm; +Qcow2BitmapDirEntry *e; + + QSIMPLEQ_FOREACH(bm, bm_list, entry) { + dir_size +=calc_dir_entry_size (strlen (bm->name),0 ); + } + +if (dir_size ==0 || dir_size >QCOW2_MAX_BITMAP_DIRECTORY_SIZE ) { +return -EINVAL ; + } + +if (in_place) { +if (*size != dir_size || *offset ==0 ) { +return -EINVAL ; + } + + dir_offset = *offset; + } + + dir =g_try_malloc (dir_size); +if (dir ==NULL ) { +return -ENOMEM ; + } + + e = (Qcow2BitmapDirEntry *)dir; + QSIMPLEQ_FOREACH(bm, bm_list, entry) { + e->bitmap_table_offset = bm->table.offset; + e->bitmap_table_size = bm->table.size; + e->flags = bm->flags; + e->type =BT_DIRTY_TRACKING_BITMAP ; + e->granularity_bits = bm->granularity_bits; + e->name_size =strlen (bm->name); + e->extra_data_size =0 ; +memcpy (e +1 , bm->name, e->name_size); + +if (check_dir_entry (bs, e) <0 ) { + ret = -EINVAL ; + goto fail; + } + + e =next_dir_entry (e); + } + +bitmap_directory_to_be (dir, dir_size); + +if (!in_place) { + dir_offset =qcow2_alloc_clusters (bs, dir_size); +if (dir_offset <0 ) { + ret = dir_offset; + goto fail; + } + } + + ret =qcow2_pre_write_overlap_check (bs,0 , dir_offset, dir_size); +if (ret <0 ) { + goto fail; + } + + ret =bdrv_pwrite (bs->file, dir_offset, dir, dir_size); +if (ret <0 ) { + goto fail; + } + +g_free (dir); + +if (!in_place) { + *size = dir_size; + *offset = dir_offset; + } + +return 0 ; + +fail: +g_free (dir); + +if (!in_place && dir_offset >0 ) { +qcow2_free_clusters (bs, dir_offset, dir_size,QCOW2_DISCARD_OTHER ); + } + +return ret; +} + +/* +* Bitmap List end +*/ + +static int update_ext_header_and_dir_in_place (BlockDriverState *bs, +Qcow2BitmapList *bm_list) +{ +BDRVQcow2State *s = bs->opaque; +int ret; + +if (!(s->autoclear_features &QCOW2_AUTOCLEAR_BITMAPS ) || + bm_list ==NULL ||QSIMPLEQ_EMPTY (bm_list) || +bitmap_list_count (bm_list) != s->nb_bitmaps) + { +return -EINVAL ; + } + + s->autoclear_features &= ~(uint64_t )QCOW2_AUTOCLEAR_BITMAPS ; + ret =update_header_sync (bs); +if (ret <0 ) { +/* Two variants are possible here: +* 1. Autoclear flag is dropped, all bitmaps will be lost. +* 2. Autoclear flag is not dropped, old state is left. +*/ +return ret; + } + +/* autoclear bit is not set, so we can safely update bitmap directory */ + + ret =bitmap_list_store (bs, bm_list, &s->bitmap_directory_offset, + &s->bitmap_directory_size, true); +if (ret <0 ) { +/* autoclear bit is cleared, so all leaked clusters would be removed on +* qemu-img check */ +return ret; + } + + ret =update_header_sync (bs); +if (ret <0 ) { +/* autoclear bit is cleared, so all leaked clusters would be removed on +* qemu-img check */ +return ret; + } + + s->autoclear_features |=QCOW2_AUTOCLEAR_BITMAPS ; +return update_header_sync (bs); +/* If final update_header_sync() fails, two variants are possible: +* 1. Autoclear flag is not set, all bitmaps will be lost. +* 2. Autoclear flag is set, header and directory are successfully updated. +*/ +} + +/* for g_slist_foreach for GSList of BdrvDirtyBitmap* elements */ +static void release_dirty_bitmap_helper (gpointer bitmap, +gpointer bs) +{ +bdrv_release_dirty_bitmap (bs, bitmap); +} + +/* for g_slist_foreach for GSList of BdrvDirtyBitmap* elements */ +static void set_readonly_helper (gpointer bitmap,gpointer value) +{ +bdrv_dirty_bitmap_set_readonly (bitmap, (bool )value); +} + +/* qcow2_load_autoloading_dirty_bitmaps() +* Return value is a hint for caller: true means that the Qcow2 header was +* updated. (false doesn't mean that the header should be updated by the +* caller, it just means that updating was not needed or the image cannot be +* written to). +* On failure the function returns false. +*/ +bool qcow2_load_autoloading_dirty_bitmaps (BlockDriverState *bs,Error **errp) +{ +BDRVQcow2State *s = bs->opaque; +Qcow2BitmapList *bm_list; +Qcow2Bitmap *bm; +GSList *created_dirty_bitmaps =NULL ; +bool header_updated = false; + +if (s->nb_bitmaps ==0 ) { +/* No bitmaps - nothing to do */ +return false; + } + + bm_list =bitmap_list_load (bs, s->bitmap_directory_offset, + s->bitmap_directory_size, errp); +if (bm_list ==NULL ) { +return false; + } + + QSIMPLEQ_FOREACH(bm, bm_list, entry) { +if ((bm->flags &BME_FLAG_AUTO ) && !(bm->flags &BME_FLAG_IN_USE )) { +BdrvDirtyBitmap *bitmap =load_bitmap (bs, bm, errp); +if (bitmap ==NULL ) { + goto fail; + } + bm->flags |=BME_FLAG_IN_USE ; + created_dirty_bitmaps = +g_slist_append (created_dirty_bitmaps, bitmap); + } + } + +if (created_dirty_bitmaps !=NULL ) { +if (can_write (bs)) { +/* in_use flags must be updated */ +int ret =update_ext_header_and_dir_in_place (bs, bm_list); +if (ret <0 ) { +error_setg_errno (errp, -ret,"Can't update bitmap directory" ); + goto fail; + } + header_updated = true; + }else { +g_slist_foreach (created_dirty_bitmaps, set_readonly_helper, + (gpointer )true); + } + } + +g_slist_free (created_dirty_bitmaps); +bitmap_list_free (bm_list); + +return header_updated; + +fail: +g_slist_foreach (created_dirty_bitmaps, release_dirty_bitmap_helper, bs); +g_slist_free (created_dirty_bitmaps); +bitmap_list_free (bm_list); + +return false; +}
block/qcow2.c +15 -2
@@ -1239,9 +1239,22 @@ static int qcow2_do_open(BlockDriverState *bs, QDict *options, int flags,/* Clear unknown autoclear feature bits */ update_header |= s->autoclear_features & ~QCOW2_AUTOCLEAR_MASK ; - -if (update_header && !bs->read_only && !(flags &BDRV_O_INACTIVE )) { + update_header = + update_header && !bs->read_only && !(flags &BDRV_O_INACTIVE ); +if (update_header) { s->autoclear_features &=QCOW2_AUTOCLEAR_MASK ; + } + +if (qcow2_load_autoloading_dirty_bitmaps (bs, &local_err)) { + update_header = false; + } +if (local_err !=NULL ) { +error_propagate (errp, local_err); + ret = -EINVAL ; + goto fail; + } + +if (update_header) { ret =qcow2_update_header (bs);if (ret <0 ) {error_setg_errno (errp, -ret,"Could not update qcow2 header" );
block/qcow2.h +2 -0
@@ -629,4 +629,6 @@ void qcow2_cache_put(BlockDriverState *bs, Qcow2Cache *c, void **table);int qcow2_check_bitmaps_refcounts (BlockDriverState *bs,BdrvCheckResult *res,void **refcount_table,int64_t *refcount_table_size); +bool qcow2_load_autoloading_dirty_bitmaps (BlockDriverState *bs,Error **errp); +#endif
Signature · 1 line
2.11.1
Re: [Qemu-devel] [PATCH v22 11/30] qcow2: autoloading dirty bitmaps
Quoted 18 lines · Vladimir Sementsov-Ogievskiy
Auto loading bitmaps are bitmaps in Qcow2, with the AUTO flag set. They are loaded when the image is opened and become BdrvDirtyBitmaps for the corresponding drive.
Extra data in bitmaps is not supported for now.
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>Reviewed-by Max Reitz <[email protected]>Reviewed-by John Snow <[email protected]>--- block/qcow2-bitmap.c | 389 +++++++++++++++++++++++++++++++++++++++++++++++++++ block/qcow2.c | 17 ++- block/qcow2.h | 2 + 3 files changed, 406 insertions(+), 2 deletions(-)block/qcow2-bitmap.c +3
-0+ +/* load_bitmap_data +* @bitmap_table entries must satisfy specification constraints.
block/qcow2-bitmap.c fail: · line 190
Quoted 10 lines
block/qcow2-bitmap.c +10
-0+* @bitmap must be cleared */ +static int load_bitmap_data (BlockDriverState *bs, +const uint64_t *bitmap_table, +uint32_t bitmap_table_size, +BdrvDirtyBitmap *bitmap) +{ +int ret =0 ; +BDRVQcow2State *s = bs->opaque; +uint64_t sector, sbc; +uint64_t bm_size =bdrv_dirty_bitmap_size (bitmap);
Right now, this is returned as the size in sectors (my series changes it to be the size in bytes, but that's being rebased on top of yours)
block/qcow2-bitmap.c fail: · line 194
Quoted 4 lines
block/qcow2-bitmap.c +4
-0+uint8_t *buf =NULL ; +uint64_t i, tab_size = +size_to_clusters (s, +bdrv_dirty_bitmap_serialization_size (bitmap,0 , bm_size));
But bdrv_dirty_bitmap_serialization_size() takes count according to the underlying hbitmap granularity. Hmm - you found a bug in my series - when I convert dirty-bitmap.c to use byte-based counts everywhere, I forgot to convert bdrv_dirty_bitmap_serialization_size(). Good thing yours is going in first ;)
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
PATCH v22 12/30 block: refactor bdrv_reopen_commit
In reply to Vladimir Sementsov-OgievskiyAdd bs local variable to simplify code.
--- block.c | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-)
block.c +8 -6
@@ -2989,9 +2989,11 @@ error:void bdrv_reopen_commit (BDRVReopenState *reopen_state) {BlockDriver *drv; +BlockDriverState *bs;assert (reopen_state !=NULL ); - drv = reopen_state->bs->drv; + bs = reopen_state->bs; + drv = bs->drv;assert (drv !=NULL );/* If there are any driver level actions to take */
@@ -3000,13 +3002,13 @@ void bdrv_reopen_commit(BDRVReopenState *reopen_state) }/* set BDS specific flags now */ -QDECREF (reopen_state->bs->explicit_options); +QDECREF (bs->explicit_options); - reopen_state->bs->explicit_options = reopen_state->explicit_options; - reopen_state->bs->open_flags = reopen_state->flags; - reopen_state->bs->read_only = !(reopen_state->flags &BDRV_O_RDWR ); + bs->explicit_options = reopen_state->explicit_options; + bs->open_flags = reopen_state->flags; + bs->read_only = !(reopen_state->flags &BDRV_O_RDWR ); -bdrv_refresh_limits (reopen_state->bs,NULL ); +bdrv_refresh_limits (bs,NULL ); }/*
Signature · 1 line
2.11.1
PATCH v22 13/30 block: new bdrv_reopen_bitmaps_rw interface
In reply to Vladimir Sementsov-OgievskiyAdd format driver handler, which should mark loaded read-only bitmaps as 'IN_USE' in the image and unset read_only field in corresponding BdrvDirtyBitmap's.
--- block.c | 19 +++++++++++++++++++ include/block/block_int.h | 7 +++++++ 2 files changed, 26 insertions(+)
block.c +19 -0
@@ -2990,12 +2990,16 @@ void bdrv_reopen_commit(BDRVReopenState *reopen_state) {BlockDriver *drv;BlockDriverState *bs; +bool old_can_write, new_can_write;assert (reopen_state !=NULL ); bs = reopen_state->bs; drv = bs->drv;assert (drv !=NULL ); + old_can_write = + !bdrv_is_read_only (bs) && !(bdrv_get_flags (bs) &BDRV_O_INACTIVE ); +/* If there are any driver level actions to take */ if (drv->bdrv_reopen_commit) { drv->bdrv_reopen_commit (reopen_state);
@@ -3009,6 +3013,21 @@ void bdrv_reopen_commit(BDRVReopenState *reopen_state) bs->read_only = !(reopen_state->flags &BDRV_O_RDWR );bdrv_refresh_limits (bs,NULL ); + + new_can_write = + !bdrv_is_read_only (bs) && !(bdrv_get_flags (bs) &BDRV_O_INACTIVE ); +if (!old_can_write && new_can_write && drv->bdrv_reopen_bitmaps_rw) { +Error *local_err =NULL ; +if (drv->bdrv_reopen_bitmaps_rw (bs, &local_err) <0 ) { +/* This is not fatal, bitmaps just left read-only, so all following +* writes will fail. User can remove read-only bitmaps to unblock +* writes. +*/ +error_reportf_err (local_err, +"%s: Failed to make dirty bitmaps writable: " , +bdrv_get_node_name (bs)); + } + } }/*
include/block/block_int.h +7 -0
@@ -381,6 +381,13 @@ struct BlockDriver { uint64_t parent_perm, uint64_t parent_shared, uint64_t *nperm, uint64_t *nshared); +/** +* Bitmaps should be marked as 'IN_USE' in the image on reopening image +* as rw. This handler should realize it. It also should unset readonly +* field of BlockDirtyBitmap's in case of success. +*/ +int (*bdrv_reopen_bitmaps_rw)(BlockDriverState *bs,Error **errp); +QLIST_ENTRY (BlockDriver) list; };
Signature · 1 line
2.11.1
Re: [Qemu-devel] [PATCH v22 13/30] block: new bdrv_reopen_bitmaps_rw interface
Quoted 6 lines · Vladimir Sementsov-Ogievskiy
Add format driver handler, which should mark loaded read-only bitmaps as '
IN_USE' in the image and unsetread_onlyfield in corresponding BdrvDirtyBitmap's.Signed-off-by Vladimir Sementsov-Ogievskiy<[email protected]>Reviewed-by John Snow<[email protected]>
Forgot to add:
Signature · 2 lines
Best regards, Vladimir
Re: [Qemu-devel] [PATCH v22 13/30] block: new bdrv_reopen_bitmaps_rw interface
[meta-comment]
Quoted 7 lines · Vladimir Sementsov-Ogievskiy
Quoted 6 lines · Vladimir Sementsov-Ogievskiy
Add format driver handler, which should mark loaded read-only bitmaps as '
IN_USE' in the image and unsetread_onlyfield in corresponding BdrvDirtyBitmap's.Signed-off-by Vladimir Sementsov-Ogievskiy<[email protected]>Reviewed-by John Snow<[email protected]>
Your original message had spaces before '<' in the email addresses, but it got lost here...
Quoted 5 lines
Forgot to add:
Reviewed-by Max Reitz<[email protected]>
...this one also lacks the space. I'm not sure if git cares, but it may be worth investigating why your mailer eats the space when you reply manually rather than sending via git; and for consistency, it is worth keeping the space (for example, we like to grep 'git log' for learning how active various contributors are, and having a consistent usage of space before < in an email address can make the task easier).
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
Re: [Qemu-devel] [PATCH v22 13/30] block: new bdrv_reopen_bitmaps_rw interface
It is interesting, but I see this problem only in your answers, in my letters I see this white-space on its place.
Quoted 24 lines · Eric Blake
[meta-comment]
Quoted 7 lines · Vladimir Sementsov-Ogievskiy
Quoted 6 lines · Vladimir Sementsov-Ogievskiy
Add format driver handler, which should mark loaded read-only bitmaps as '
IN_USE' in the image and unsetread_onlyfield in corresponding BdrvDirtyBitmap's.Signed-off-by Vladimir Sementsov-Ogievskiy<[email protected]>Reviewed-by John Snow<[email protected]>Your original message had spaces before '<' in the email addresses, but it got lost here...
Quoted 4 lines
Forgot to add:
Reviewed-by Max Reitz<[email protected]>...this one also lacks the space. I'm not sure if git cares, but it may be worth investigating why your mailer eats the space when you reply manually rather than sending via git; and for consistency, it is worth keeping the space (for example, we like to grep 'git log' for learning how active various contributors are, and having a consistent usage of space before < in an email address can make the task easier).
Signature · 2 lines
Best regards, Vladimir
Re: [Qemu-devel] [PATCH v22 13/30] block: new bdrv_reopen_bitmaps_rw interface
Quoted 1 line · Vladimir Sementsov-Ogievskiy
It is interesting, but I see this problem only in your answers, in my letters I see this white-space on its place.
In outgoing letter I see this white-space, but in letter from mailing-list it is absent.
Quoted 27 lines
Quoted 24 lines · Eric Blake
[meta-comment]
Quoted 7 lines · Vladimir Sementsov-Ogievskiy
Quoted 6 lines · Vladimir Sementsov-Ogievskiy
Add format driver handler, which should mark loaded read-only bitmaps as '
IN_USE' in the image and unsetread_onlyfield in corresponding BdrvDirtyBitmap's.Signed-off-by Vladimir Sementsov-Ogievskiy<[email protected]>Reviewed-by John Snow<[email protected]>Your original message had spaces before '<' in the email addresses, but it got lost here...
Quoted 4 lines
Forgot to add:
Reviewed-by Max Reitz<[email protected]>...this one also lacks the space. I'm not sure if git cares, but it may be worth investigating why your mailer eats the space when you reply manually rather than sending via git; and for consistency, it is worth keeping the space (for example, we like to grep 'git log' for learning how active various contributors are, and having a consistent usage of space before < in an email address can make the task easier).
Signature · 2 lines
Best regards, Vladimir
Re: [Qemu-devel] [PATCH v22 13/30] block: new bdrv_reopen_bitmaps_rw interface
Finally, it looks like thunderbird bug
https://bugzilla.mozilla.org/show_bug.cgi?id=1160880
[sorry for so much offtopic]
Quoted 33 lines · Vladimir Sementsov-Ogievskiy
Quoted 1 line · Vladimir Sementsov-Ogievskiy
It is interesting, but I see this problem only in your answers, in my letters I see this white-space on its place.
In outgoing letter I see this white-space, but in letter from mailing-list it is absent.
Quoted 27 lines
Quoted 24 lines · Eric Blake
[meta-comment]
Quoted 7 lines · Vladimir Sementsov-Ogievskiy
Quoted 6 lines · Vladimir Sementsov-Ogievskiy
Add format driver handler, which should mark loaded read-only bitmaps as '
IN_USE' in the image and unsetread_onlyfield in corresponding BdrvDirtyBitmap's.Signed-off-by Vladimir Sementsov-Ogievskiy<[email protected]>Reviewed-by John Snow<[email protected]>Your original message had spaces before '<' in the email addresses, but it got lost here...
Quoted 4 lines
Forgot to add:
Reviewed-by Max Reitz<[email protected]>...this one also lacks the space. I'm not sure if git cares, but it may be worth investigating why your mailer eats the space when you reply manually rather than sending via git; and for consistency, it is worth keeping the space (for example, we like to grep 'git log' for learning how active various contributors are, and having a consistent usage of space before < in an email address can make the task easier).
Signature · 2 lines
Best regards, Vladimir
Re: [Qemu-devel] [PATCH v22 13/30] block: new bdrv_reopen_bitmaps_rw interface
In reply to Vladimir Sementsov-OgievskiyQuoted 2 lines · Vladimir Sementsov-Ogievskiy
It is interesting, but I see this problem only in your answers, in my letters I see this white-space on its place.
That's the good old Thunderbird "format=flowed" bug.
Vladimir, download
http://people.redhat.com/pbonzini/format-flawed.tar.gz and place it into ~/.thunderbird/<profile name>/extensions. It works around the bug.
Paolo
Quoted 14 lines
Quoted 13 lines · Eric Blake
[meta-comment]
Quoted 7 lines · Vladimir Sementsov-Ogievskiy
Quoted 6 lines · Vladimir Sementsov-Ogievskiy
Add format driver handler, which should mark loaded read-only bitmaps as '
IN_USE' in the image and unsetread_onlyfield in corresponding BdrvDirtyBitmap's.Signed-off-by Vladimir Sementsov-Ogievskiy<[email protected]>Reviewed-by John Snow<[email protected]>Your original message had spaces before '<' in the email addresses, but it got lost here...
Re: [Qemu-devel] [PATCH v22 13/30] block: new bdrv_reopen_bitmaps_rw interface
Quoted 9 lines · Paolo Bonzini
Quoted 2 lines · Vladimir Sementsov-Ogievskiy
It is interesting, but I see this problem only in your answers, in my letters I see this white-space on its place.
That's the good old Thunderbird "format=flowed" bug.
Vladimir, download
http://people.redhat.com/pbonzini/format-flawed.tar.gz and place it into ~/.thunderbird/<profile name>/extensions. It works around the bug.
unfortunately, with this buttons 'reply' and 'reply to all' do nothing..
Quoted 17 lines
Paolo
Quoted 14 lines
Quoted 13 lines · Eric Blake
[meta-comment]
Quoted 7 lines · Vladimir Sementsov-Ogievskiy
Quoted 6 lines · Vladimir Sementsov-Ogievskiy
Add format driver handler, which should mark loaded read-only bitmaps as '
IN_USE' in the image and unsetread_onlyfield in corresponding BdrvDirtyBitmap's.Signed-off-by Vladimir Sementsov-Ogievskiy<[email protected]>Reviewed-by John Snow<[email protected]>Your original message had spaces before '<' in the email addresses, but it got lost here...
Signature · 2 lines
Best regards, Vladimir
PATCH v22 14/30 qcow2: support .bdrv_reopen_bitmaps_rw
In reply to Vladimir Sementsov-OgievskiyRealize bdrv_reopen_bitmaps_rw interface.
--- block/qcow2-bitmap.c | 61 ++++++++++++++++++++++++++++++++++++++++++++++++++++ block/qcow2.c | 2 ++ block/qcow2.h | 1 + 3 files changed, 64 insertions(+)
block/qcow2-bitmap.c +61 -0
@@ -826,3 +826,64 @@ fail:return false; } + +int qcow2_reopen_bitmaps_rw (BlockDriverState *bs,Error **errp) +{ +BDRVQcow2State *s = bs->opaque; +Qcow2BitmapList *bm_list; +Qcow2Bitmap *bm; +GSList *ro_dirty_bitmaps =NULL ; +int ret =0 ; + +if (s->nb_bitmaps ==0 ) { +/* No bitmaps - nothing to do */ +return 0 ; + } + +if (!can_write (bs)) { +error_setg (errp,"Can't write to the image on reopening bitmaps rw" ); +return -EINVAL ; + } + + bm_list =bitmap_list_load (bs, s->bitmap_directory_offset, + s->bitmap_directory_size, errp); +if (bm_list ==NULL ) { +return -EINVAL ; + } + + QSIMPLEQ_FOREACH(bm, bm_list, entry) { +if (!(bm->flags &BME_FLAG_IN_USE )) { +BdrvDirtyBitmap *bitmap =bdrv_find_dirty_bitmap (bs, bm->name); +if (bitmap ==NULL ) { +continue ; + } + +if (!bdrv_dirty_bitmap_readonly (bitmap)) { +error_setg (errp,"Bitmap %s is not readonly but not marked" +"'IN_USE' in the image. Something went wrong," +"all the bitmaps may be corrupted" , bm->name); + ret = -EINVAL ; + goto out; + } + + bm->flags |=BME_FLAG_IN_USE ; + ro_dirty_bitmaps =g_slist_append (ro_dirty_bitmaps, bitmap); + } + } + +if (ro_dirty_bitmaps !=NULL ) { +/* in_use flags must be updated */ + ret =update_ext_header_and_dir_in_place (bs, bm_list); +if (ret <0 ) { +error_setg_errno (errp, -ret,"Can't update bitmap directory" ); + goto out; + } +g_slist_foreach (ro_dirty_bitmaps, set_readonly_helper, false); + } + +out: +g_slist_free (ro_dirty_bitmaps); +bitmap_list_free (bm_list); + +return ret; +}
block/qcow2.c +2 -0
@@ -3595,6 +3595,8 @@ BlockDriver bdrv_qcow2 = { .bdrv_detach_aio_context = qcow2_detach_aio_context, .bdrv_attach_aio_context = qcow2_attach_aio_context, + + .bdrv_reopen_bitmaps_rw = qcow2_reopen_bitmaps_rw, };static void bdrv_qcow2_init (void )
block/qcow2.h +1 -0
@@ -630,5 +630,6 @@ int qcow2_check_bitmaps_refcounts(BlockDriverState *bs, BdrvCheckResult *res,void **refcount_table, int64_t *refcount_table_size);bool qcow2_load_autoloading_dirty_bitmaps (BlockDriverState *bs,Error **errp); +int qcow2_reopen_bitmaps_rw (BlockDriverState *bs,Error **errp);#endif
Signature · 1 line
2.11.1
PATCH v22 15/30 block/dirty-bitmap: add autoload field to BdrvDirtyBitmap
In reply to Vladimir Sementsov-OgievskiyMirror AUTO flag from Qcow2 bitmap in BdrvDirtyBitmap. This will be needed in future, to save this flag back to Qcow2 for persistent bitmaps.
--- block/dirty-bitmap.c | 18 ++++++++++++++++++ block/qcow2-bitmap.c | 2 ++ include/block/dirty-bitmap.h | 2 ++ 3 files changed, 22 insertions(+)
block/dirty-bitmap.c +18 -0
@@ -52,6 +52,8 @@ struct BdrvDirtyBitmap {Such operations must fail and both the image and this bitmap must remain unchanged while this flag is set. */ +bool autoload;/* For persistent bitmaps: bitmap must be +autoloaded on image opening */ QLIST_ENTRY (BdrvDirtyBitmap) list; };
@@ -100,6 +102,7 @@ void bdrv_dirty_bitmap_make_anon(BdrvDirtyBitmap *bitmap)assert (!bdrv_dirty_bitmap_frozen (bitmap));g_free (bitmap->name); bitmap->name =NULL ; + bitmap->autoload = false; }/* Called with BQL taken. */
@@ -296,6 +299,8 @@ BdrvDirtyBitmap *bdrv_dirty_bitmap_abdicate(BlockDriverState *bs, bitmap->name =NULL ; successor->name = name; bitmap->successor =NULL ; + successor->autoload = bitmap->autoload; + bitmap->autoload = false;bdrv_release_dirty_bitmap (bs, bitmap);return successor;
@@ -671,3 +676,16 @@ bool bdrv_has_readonly_bitmaps(BlockDriverState *bs)return false; } + +/* Called with BQL taken. */ +void bdrv_dirty_bitmap_set_autoload (BdrvDirtyBitmap *bitmap,bool autoload) +{ +qemu_mutex_lock (bitmap->mutex); + bitmap->autoload = autoload; +qemu_mutex_unlock (bitmap->mutex); +} + +bool bdrv_dirty_bitmap_get_autoload (const BdrvDirtyBitmap *bitmap) +{ +return bitmap->autoload; +}
block/qcow2-bitmap.c +2 -0
@@ -793,6 +793,8 @@ bool qcow2_load_autoloading_dirty_bitmaps(BlockDriverState *bs, Error **errp)if (bitmap ==NULL ) { goto fail; } + +bdrv_dirty_bitmap_set_autoload (bitmap, true); bm->flags |=BME_FLAG_IN_USE ; created_dirty_bitmaps =g_slist_append (created_dirty_bitmaps, bitmap);
include/block/dirty-bitmap.h +2 -0
@@ -72,6 +72,7 @@ void bdrv_dirty_bitmap_deserialize_ones(BdrvDirtyBitmap *bitmap,void bdrv_dirty_bitmap_deserialize_finish (BdrvDirtyBitmap *bitmap);void bdrv_dirty_bitmap_set_readonly (BdrvDirtyBitmap *bitmap,bool value); +void bdrv_dirty_bitmap_set_autoload (BdrvDirtyBitmap *bitmap,bool autoload);/* Functions that require manual locking. */ void bdrv_dirty_bitmap_lock (BdrvDirtyBitmap *bitmap);
@@ -89,5 +90,6 @@ int64_t bdrv_get_meta_dirty_count(BdrvDirtyBitmap *bitmap);void bdrv_dirty_bitmap_truncate (BlockDriverState *bs);bool bdrv_dirty_bitmap_readonly (const BdrvDirtyBitmap *bitmap);bool bdrv_has_readonly_bitmaps (BlockDriverState *bs); +bool bdrv_dirty_bitmap_get_autoload (const BdrvDirtyBitmap *bitmap);#endif
Signature · 1 line
2.11.1
PATCH v22 16/30 block: bdrv_close: release bitmaps after drv->bdrv_close
In reply to Vladimir Sementsov-OgievskiyRelease bitmaps after 'if (bs->drv) { ... }' block. This will allow format driver to save persistent bitmaps, which will appear in following commits.
--- block.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
block.c +3 -3
@@ -3061,9 +3061,6 @@ static void bdrv_close(BlockDriverState *bs)bdrv_flush (bs);bdrv_drain (bs);/* in case flush left pending I/O */ -bdrv_release_named_dirty_bitmaps (bs); -assert (QLIST_EMPTY (&bs->dirty_bitmaps)); -if (bs->drv) {BdrvChild *child, *next;
@@ -3102,6 +3099,9 @@ static void bdrv_close(BlockDriverState *bs) bs->full_open_options =NULL ; } +bdrv_release_named_dirty_bitmaps (bs); +assert (QLIST_EMPTY (&bs->dirty_bitmaps)); + QLIST_FOREACH_SAFE(ban, &bs->aio_notifiers, list, ban_next) {g_free (ban); }
Signature · 1 line
2.11.1
PATCH v22 17/30 block: introduce persistent dirty bitmaps
In reply to Vladimir Sementsov-OgievskiyNew field BdrvDirtyBitmap.persistent means, that bitmap should be saved by format driver in .bdrv_close and .bdrv_inactivate. No format driver supports it for now.
--- block/dirty-bitmap.c | 29 +++++++++++++++++++++++++++++ block/qcow2-bitmap.c | 1 + include/block/dirty-bitmap.h | 4 ++++ 3 files changed, 34 insertions(+)
block/dirty-bitmap.c +29 -0
@@ -54,6 +54,7 @@ struct BdrvDirtyBitmap {this flag is set. */ bool autoload;/* For persistent bitmaps: bitmap must be autoloaded on image opening */ +bool persistent;/* bitmap must be saved to owner disk image */ QLIST_ENTRY (BdrvDirtyBitmap) list; };
@@ -102,6 +103,7 @@ void bdrv_dirty_bitmap_make_anon(BdrvDirtyBitmap *bitmap)assert (!bdrv_dirty_bitmap_frozen (bitmap));g_free (bitmap->name); bitmap->name =NULL ; + bitmap->persistent = false; bitmap->autoload = false; }
@@ -299,6 +301,8 @@ BdrvDirtyBitmap *bdrv_dirty_bitmap_abdicate(BlockDriverState *bs, bitmap->name =NULL ; successor->name = name; bitmap->successor =NULL ; + successor->persistent = bitmap->persistent; + bitmap->persistent = false; successor->autoload = bitmap->autoload; bitmap->autoload = false;bdrv_release_dirty_bitmap (bs, bitmap);
@@ -689,3 +693,28 @@ bool bdrv_dirty_bitmap_get_autoload(const BdrvDirtyBitmap *bitmap) {return bitmap->autoload; } + +/* Called with BQL taken. */ +void bdrv_dirty_bitmap_set_persistance (BdrvDirtyBitmap *bitmap,bool persistent) +{ +qemu_mutex_lock (bitmap->mutex); + bitmap->persistent = persistent; +qemu_mutex_unlock (bitmap->mutex); +} + +bool bdrv_dirty_bitmap_get_persistance (BdrvDirtyBitmap *bitmap) +{ +return bitmap->persistent; +} + +bool bdrv_has_changed_persistent_bitmaps (BlockDriverState *bs) +{ +BdrvDirtyBitmap *bm; + QLIST_FOREACH(bm, &bs->dirty_bitmaps, list) { +if (bm->persistent && !bm->readonly) { +return true; + } + } + +return false; +}
block/qcow2-bitmap.c +1 -0
@@ -794,6 +794,7 @@ bool qcow2_load_autoloading_dirty_bitmaps(BlockDriverState *bs, Error **errp) goto fail; } +bdrv_dirty_bitmap_set_persistance (bitmap, true);bdrv_dirty_bitmap_set_autoload (bitmap, true); bm->flags |=BME_FLAG_IN_USE ; created_dirty_bitmaps =
include/block/dirty-bitmap.h +4 -0
@@ -73,6 +73,8 @@ void bdrv_dirty_bitmap_deserialize_finish(BdrvDirtyBitmap *bitmap);void bdrv_dirty_bitmap_set_readonly (BdrvDirtyBitmap *bitmap,bool value);void bdrv_dirty_bitmap_set_autoload (BdrvDirtyBitmap *bitmap,bool autoload); +void bdrv_dirty_bitmap_set_persistance (BdrvDirtyBitmap *bitmap, +bool persistent);/* Functions that require manual locking. */ void bdrv_dirty_bitmap_lock (BdrvDirtyBitmap *bitmap);
@@ -91,5 +93,7 @@ void bdrv_dirty_bitmap_truncate(BlockDriverState *bs);bool bdrv_dirty_bitmap_readonly (const BdrvDirtyBitmap *bitmap);bool bdrv_has_readonly_bitmaps (BlockDriverState *bs);bool bdrv_dirty_bitmap_get_autoload (const BdrvDirtyBitmap *bitmap); +bool bdrv_dirty_bitmap_get_persistance (BdrvDirtyBitmap *bitmap); +bool bdrv_has_changed_persistent_bitmaps (BlockDriverState *bs);#endif
Signature · 1 line
2.11.1
Re: [Qemu-devel] [PATCH v22 17/30] block: introduce persistent dirty bitmaps
Quoted 15 lines · Vladimir Sementsov-Ogievskiy
New field BdrvDirtyBitmap.persistent means, that bitmap should be saved by format driver in .bdrv_close and .bdrv_inactivate. No format driver supports it for now.
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>--- block/dirty-bitmap.c | 29 +++++++++++++++++++++++++++++ block/qcow2-bitmap.c | 1 + include/block/dirty-bitmap.h | 4 ++++ 3 files changed, 34 insertions(+)
include/block/dirty-bitmap.h void bdrv_dirty_bitmap_deserialize_finish(BdrvDirtyBitmap *bitmap); · line 78
Quoted 6 lines
include/block/dirty-bitmap.h +2
-0@@ -73,6 +73,8 @@ void bdrv_dirty_bitmap_deserialize_finish(BdrvDirtyBitmap *bitmap);void bdrv_dirty_bitmap_set_readonly (BdrvDirtyBitmap *bitmap,bool value);void bdrv_dirty_bitmap_set_autoload (BdrvDirtyBitmap *bitmap,bool autoload); +void bdrv_dirty_bitmap_set_persistance (BdrvDirtyBitmap *bitmap, +bool persistent);
Indentation is off.
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
PATCH v22 18/30 block/dirty-bitmap: add bdrv_dirty_bitmap_next()
In reply to Vladimir Sementsov-Ogievskiy--- block/dirty-bitmap.c | 7 +++++++ include/block/dirty-bitmap.h | 2 ++ 2 files changed, 9 insertions(+)
block/dirty-bitmap.c +7 -0
@@ -718,3 +718,10 @@ bool bdrv_has_changed_persistent_bitmaps(BlockDriverState *bs)return false; } + +BdrvDirtyBitmap *bdrv_dirty_bitmap_next (BlockDriverState *bs, +BdrvDirtyBitmap *bitmap) +{ +return bitmap ==NULL ?QLIST_FIRST (&bs->dirty_bitmaps) : +QLIST_NEXT (bitmap, list); +}
include/block/dirty-bitmap.h +2 -0
@@ -95,5 +95,7 @@ bool bdrv_has_readonly_bitmaps(BlockDriverState *bs);bool bdrv_dirty_bitmap_get_autoload (const BdrvDirtyBitmap *bitmap);bool bdrv_dirty_bitmap_get_persistance (BdrvDirtyBitmap *bitmap);bool bdrv_has_changed_persistent_bitmaps (BlockDriverState *bs); +BdrvDirtyBitmap *bdrv_dirty_bitmap_next (BlockDriverState *bs, +BdrvDirtyBitmap *bitmap);#endif
Signature · 1 line
2.11.1
PATCH v22 19/30 qcow2: add persistent dirty bitmaps support
In reply to Vladimir Sementsov-OgievskiyStore persistent dirty bitmaps in qcow2 image.
--- block/qcow2-bitmap.c | 475 +++++++++++++++++++++++++++++++++++++++++++++++++++ block/qcow2.c | 9 + block/qcow2.h | 1 + 3 files changed, 485 insertions(+)
block/qcow2-bitmap.c +475 -0
@@ -27,6 +27,7 @@#include "qemu/osdep.h" #include "qapi/error.h" +#include "qemu/cutils.h" #include "block/block_int.h" #include "block/qcow2.h"
@@ -42,6 +43,10 @@#define BME_MIN_GRANULARITY_BITS 9#define BME_MAX_NAME_SIZE 1023 +#if BME_MAX_TABLE_SIZE *8ULL >INT_MAX +#error In the code bitmap table physical size assumed to fit into int +#endif +/* Bitmap directory entry flags */ #define BME_RESERVED_FLAGS 0xfffffffcU#define BME_FLAG_IN_USE (1U << 0)
@@ -72,6 +77,8 @@ typedef struct Qcow2BitmapTable {uint32_t size;/* number of 64bit entries */ QSIMPLEQ_ENTRY (Qcow2BitmapTable) entry; } Qcow2BitmapTable; +typedef QSIMPLEQ_HEAD(Qcow2BitmapTableList, Qcow2BitmapTable) +Qcow2BitmapTableList ; typedef struct Qcow2Bitmap { Qcow2BitmapTable table;
@@ -79,6 +86,8 @@ typedef struct Qcow2Bitmap {uint8_t granularity_bits;char *name; +BdrvDirtyBitmap *dirty_bitmap; +QSIMPLEQ_ENTRY (Qcow2Bitmap) entry; } Qcow2Bitmap; typedef QSIMPLEQ_HEAD(Qcow2BitmapList, Qcow2Bitmap) Qcow2BitmapList;
@@ -104,6 +113,15 @@ static int update_header_sync(BlockDriverState *bs)return bdrv_flush (bs); } +static inline void bitmap_table_to_be (uint64_t *bitmap_table,size_t size) +{ +size_t i; + +for (i =0 ; i < size; ++i) { +cpu_to_be64s (&bitmap_table[i]); + } +} +static int check_table_entry (uint64_t entry,int cluster_size) {uint64_t offset;
@@ -127,6 +145,70 @@ static int check_table_entry(uint64_t entry, int cluster_size)return 0 ; } +static int check_constraints_on_bitmap (BlockDriverState *bs, +const char *name, +uint32_t granularity, +Error **errp) +{ +BDRVQcow2State *s = bs->opaque; +int granularity_bits =ctz32 (granularity); +int64_t len =bdrv_getlength (bs); + +assert (granularity >0 ); +assert ((granularity & (granularity -1 )) ==0 ); + +if (len <0 ) { +error_setg_errno (errp, -len,"Failed to get size of '%s'" , +bdrv_get_device_or_node_name (bs)); +return len; + } + +if (granularity_bits >BME_MAX_GRANULARITY_BITS ) { +error_setg (errp,"Granularity exceeds maximum (%llu bytes)" , +1ULL <<BME_MAX_GRANULARITY_BITS ); +return -EINVAL ; + } +if (granularity_bits <BME_MIN_GRANULARITY_BITS ) { +error_setg (errp,"Granularity is under minimum (%llu bytes)" , +1ULL <<BME_MIN_GRANULARITY_BITS ); +return -EINVAL ; + } + +if ((len > (uint64_t )BME_MAX_PHYS_SIZE << granularity_bits) || + (len > (uint64_t )BME_MAX_TABLE_SIZE * s->cluster_size << + granularity_bits)) + { +error_setg (errp,"Too much space will be occupied by the bitmap. " +"Use larger granularity" ); +return -EINVAL ; + } + +if (strlen (name) >BME_MAX_NAME_SIZE ) { +error_setg (errp,"Name length exceeds maximum (%u characters)" , +BME_MAX_NAME_SIZE ); +return -EINVAL ; + } + +return 0 ; +} + +static void clear_bitmap_table (BlockDriverState *bs,uint64_t *bitmap_table, +uint32_t bitmap_table_size) +{ +BDRVQcow2State *s = bs->opaque; +int i; + +for (i =0 ; i < bitmap_table_size; ++i) { +uint64_t addr = bitmap_table[i] &BME_TABLE_ENTRY_OFFSET_MASK ; +if (!addr) { +continue ; + } + +qcow2_free_clusters (bs, addr, s->cluster_size,QCOW2_DISCARD_OTHER ); + bitmap_table[i] =0 ; + } +} +static int bitmap_table_load (BlockDriverState *bs,Qcow2BitmapTable *tb,uint64_t **bitmap_table) {
@@ -165,6 +247,28 @@ fail:return ret; } +static int free_bitmap_clusters (BlockDriverState *bs,Qcow2BitmapTable *tb) +{ +int ret; +uint64_t *bitmap_table; + + ret =bitmap_table_load (bs, tb, &bitmap_table); +if (ret <0 ) { +assert (bitmap_table ==NULL ); +return ret; + } + +clear_bitmap_table (bs, bitmap_table, tb->size); +qcow2_free_clusters (bs, tb->offset, tb->size *sizeof (uint64_t ), +QCOW2_DISCARD_OTHER ); +g_free (bitmap_table); + + tb->offset =0 ; + tb->size =0 ; + +return 0 ; +} +/* This function returns the number of disk sectors covered by a single qcow2 * cluster of bitmap data. */ static uint64_t sectors_covered_by_bitmap_cluster(const BDRVQcow2State *s,
@@ -748,6 +852,69 @@ static int update_ext_header_and_dir_in_place(BlockDriverState *bs,*/ } +static int update_ext_header_and_dir (BlockDriverState *bs, +Qcow2BitmapList *bm_list) +{ +BDRVQcow2State *s = bs->opaque; +int ret; +uint64_t new_offset =0 ; +uint64_t new_size =0 ; +uint32_t new_nb_bitmaps =0 ; +uint64_t old_offset = s->bitmap_directory_offset; +uint64_t old_size = s->bitmap_directory_size; +uint32_t old_nb_bitmaps = s->nb_bitmaps; +uint64_t old_autocl = s->autoclear_features; + +if (bm_list !=NULL && !QSIMPLEQ_EMPTY (bm_list)) { + new_nb_bitmaps =bitmap_list_count (bm_list); + +if (new_nb_bitmaps >QCOW2_MAX_BITMAPS ) { +return -EINVAL ; + } + + ret =bitmap_list_store (bs, bm_list, &new_offset, &new_size, false); +if (ret <0 ) { +return ret; + } + + ret =bdrv_flush (bs->file->bs); +if (ret <0 ) { + goto fail; + } + + s->autoclear_features |=QCOW2_AUTOCLEAR_BITMAPS ; + }else { + s->autoclear_features &= ~(uint64_t )QCOW2_AUTOCLEAR_BITMAPS ; + } + + s->bitmap_directory_offset = new_offset; + s->bitmap_directory_size = new_size; + s->nb_bitmaps = new_nb_bitmaps; + + ret =update_header_sync (bs); +if (ret <0 ) { + goto fail; + } + +if (old_size >0 ) { +qcow2_free_clusters (bs, old_offset, old_size,QCOW2_DISCARD_OTHER ); + } + +return 0 ; + +fail: +if (new_offset >0 ) { +qcow2_free_clusters (bs, new_offset, new_size,QCOW2_DISCARD_OTHER ); + } + + s->bitmap_directory_offset = old_offset; + s->bitmap_directory_size = old_size; + s->nb_bitmaps = old_nb_bitmaps; + s->autoclear_features = old_autocl; + +return ret; +} +/* for g_slist_foreach for GSList of BdrvDirtyBitmap* elements */ static void release_dirty_bitmap_helper (gpointer bitmap,gpointer bs)
@@ -890,3 +1057,311 @@ out:return ret; } + +/* store_bitmap_data() +* Store bitmap to image, filling bitmap table accordingly. +*/ +static uint64_t *store_bitmap_data (BlockDriverState *bs, +BdrvDirtyBitmap *bitmap, +uint32_t *bitmap_table_size,Error **errp) +{ +int ret; +BDRVQcow2State *s = bs->opaque; +int64_t sector; +uint64_t sbc; +uint64_t bm_size =bdrv_dirty_bitmap_size (bitmap); +const char *bm_name =bdrv_dirty_bitmap_name (bitmap); +uint8_t *buf =NULL ; +BdrvDirtyBitmapIter *dbi; +uint64_t *tb; +uint64_t tb_size = +size_to_clusters (s, +bdrv_dirty_bitmap_serialization_size (bitmap,0 , bm_size)); + +if (tb_size >BME_MAX_TABLE_SIZE || + tb_size * s->cluster_size >BME_MAX_PHYS_SIZE ) + { +error_setg (errp,"Bitmap '%s' is too big" , bm_name); +return NULL ; + } + + tb =g_try_new0 (uint64_t, tb_size); +if (tb ==NULL ) { +error_setg (errp,"No memory" ); +return NULL ; + } + + dbi =bdrv_dirty_iter_new (bitmap,0 ); + buf =g_malloc (s->cluster_size); + sbc =sectors_covered_by_bitmap_cluster (s, bitmap); +assert (DIV_ROUND_UP (bm_size, sbc) == tb_size); + +while ((sector =bdrv_dirty_iter_next (dbi)) !=-1 ) { +uint64_t cluster = sector / sbc; +uint64_t end, write_size; +int64_t off; + + sector = cluster * sbc; + end =MIN (bm_size, sector + sbc); + write_size = +bdrv_dirty_bitmap_serialization_size (bitmap, sector, end - sector); +assert (write_size <= s->cluster_size); + + off =qcow2_alloc_clusters (bs, s->cluster_size); +if (off <0 ) { +error_setg_errno (errp, -off, +"Failed to allocate clusters for bitmap '%s'" , + bm_name); + goto fail; + } + tb[cluster] = off; + +bdrv_dirty_bitmap_serialize_part (bitmap, buf, sector, end - sector); +if (write_size < s->cluster_size) { +memset (buf + write_size,0 , s->cluster_size - write_size); + } + + ret =qcow2_pre_write_overlap_check (bs,0 , off, s->cluster_size); +if (ret <0 ) { +error_setg_errno (errp, -ret,"Qcow2 overlap check failed" ); + goto fail; + } + + ret =bdrv_pwrite (bs->file, off, buf, s->cluster_size); +if (ret <0 ) { +error_setg_errno (errp, -ret,"Failed to write bitmap '%s' to file" , + bm_name); + goto fail; + } + +if (end >= bm_size) { +break ; + } + +bdrv_set_dirty_iter (dbi, end); + } + + *bitmap_table_size = tb_size; +g_free (buf); +bdrv_dirty_iter_free (dbi); + +return tb; + +fail: +clear_bitmap_table (bs, tb, tb_size); +g_free (buf); +bdrv_dirty_iter_free (dbi); +g_free (tb); + +return NULL ; +} + +/* store_bitmap() +* Store bm->dirty_bitmap to qcow2. +* Set bm->table_offset and bm->table_size accordingly. +*/ +static int store_bitmap (BlockDriverState *bs,Qcow2Bitmap *bm,Error **errp) +{ +int ret; +uint64_t *tb; +int64_t tb_offset; +uint32_t tb_size; +BdrvDirtyBitmap *bitmap = bm->dirty_bitmap; +const char *bm_name; + +assert (bitmap !=NULL ); + + bm_name =bdrv_dirty_bitmap_name (bitmap); + + tb =store_bitmap_data (bs, bitmap, &tb_size, errp); +if (tb ==NULL ) { +return -EINVAL ; + } + +assert (tb_size <=BME_MAX_TABLE_SIZE ); + tb_offset =qcow2_alloc_clusters (bs, tb_size *sizeof (tb[0 ])); +if (tb_offset <0 ) { +error_setg_errno (errp, -tb_offset, +"Failed to allocate clusters for bitmap '%s'" , + bm_name); + goto fail; + } + + ret =qcow2_pre_write_overlap_check (bs,0 , tb_offset, + tb_size *sizeof (tb[0 ])); +if (ret <0 ) { +error_setg_errno (errp, -ret,"Qcow2 overlap check failed" ); + goto fail; + } + +bitmap_table_to_be (tb, tb_size); + ret =bdrv_pwrite (bs->file, tb_offset, tb, tb_size *sizeof (tb[0 ])); +if (ret <0 ) { +error_setg_errno (errp, -ret,"Failed to write bitmap '%s' to file" , + bm_name); + goto fail; + } + +g_free (tb); + + bm->table.offset = tb_offset; + bm->table.size = tb_size; + +return 0 ; + +fail: +clear_bitmap_table (bs, tb, tb_size); + +if (tb_offset >0 ) { +qcow2_free_clusters (bs, tb_offset, tb_size *sizeof (tb[0 ]), +QCOW2_DISCARD_OTHER ); + } + +g_free (tb); + +return ret; +} + +static Qcow2Bitmap *find_bitmap_by_name (Qcow2BitmapList *bm_list, +const char *name) +{ +Qcow2Bitmap *bm; + + QSIMPLEQ_FOREACH(bm, bm_list, entry) { +if (strcmp (name, bm->name) ==0 ) { +return bm; + } + } + +return NULL ; +} + +void qcow2_store_persistent_dirty_bitmaps (BlockDriverState *bs,Error **errp) +{ +BdrvDirtyBitmap *bitmap; +BDRVQcow2State *s = bs->opaque; +uint32_t new_nb_bitmaps = s->nb_bitmaps; +uint64_t new_dir_size = s->bitmap_directory_size; +int ret; +Qcow2BitmapList *bm_list; +Qcow2Bitmap *bm; +Qcow2BitmapTableList drop_tables; +Qcow2BitmapTable *tb, *tb_next; + +if (!bdrv_has_changed_persistent_bitmaps (bs)) { +/* nothing to do */ +return ; + } + +if (!can_write (bs)) { +error_setg (errp,"No write access" ); +return ; + } + +QSIMPLEQ_INIT (&drop_tables); + +if (s->nb_bitmaps ==0 ) { + bm_list =bitmap_list_new (); + }else { + bm_list =bitmap_list_load (bs, s->bitmap_directory_offset, + s->bitmap_directory_size, errp); +if (bm_list ==NULL ) { +return ; + } + } + +/* check constraints and names */ +for (bitmap =bdrv_dirty_bitmap_next (bs,NULL ); bitmap !=NULL ; + bitmap =bdrv_dirty_bitmap_next (bs, bitmap)) + { +const char *name =bdrv_dirty_bitmap_name (bitmap); +uint32_t granularity =bdrv_dirty_bitmap_granularity (bitmap); +Qcow2Bitmap *bm; + +if (!bdrv_dirty_bitmap_get_persistance (bitmap) || +bdrv_dirty_bitmap_readonly (bitmap)) + { +continue ; + } + +if (check_constraints_on_bitmap (bs, name, granularity, errp) <0 ) { +error_prepend (errp,"Bitmap '%s' doesn't satisfy the constraints: " , + name); + goto fail; + } + + bm =find_bitmap_by_name (bm_list, name); +if (bm ==NULL ) { +if (++new_nb_bitmaps >QCOW2_MAX_BITMAPS ) { +error_setg (errp,"Too many persistent bitmaps" ); + goto fail; + } + + new_dir_size +=calc_dir_entry_size (strlen (name),0 ); +if (new_dir_size >QCOW2_MAX_BITMAP_DIRECTORY_SIZE ) { +error_setg (errp,"Bitmap directory is too large" ); + goto fail; + } + + bm =g_new0 (Qcow2Bitmap,1 ); + bm->name =g_strdup (name); +QSIMPLEQ_INSERT_TAIL (bm_list, bm, entry); + }else { +if (!(bm->flags &BME_FLAG_IN_USE )) { +error_setg (errp,"Bitmap '%s' already exists in the image" , + name); + goto fail; + } + tb =g_memdup (&bm->table,sizeof (bm->table)); + bm->table.offset =0 ; + bm->table.size =0 ; +QSIMPLEQ_INSERT_TAIL (&drop_tables, tb, entry); + } + bm->flags =bdrv_dirty_bitmap_get_autoload (bitmap) ?BME_FLAG_AUTO :0 ; + bm->granularity_bits =ctz32 (bdrv_dirty_bitmap_granularity (bitmap)); + bm->dirty_bitmap = bitmap; + } + +/* allocate clusters and store bitmaps */ + QSIMPLEQ_FOREACH(bm, bm_list, entry) { +if (bm->dirty_bitmap ==NULL ) { +continue ; + } + + ret =store_bitmap (bs, bm, errp); +if (ret <0 ) { + goto fail; + } + } + + ret =update_ext_header_and_dir (bs, bm_list); +if (ret <0 ) { +error_setg_errno (errp, -ret,"Failed to update bitmap extension" ); + goto fail; + } + +/* Bitmap directory was successfully updated, so, old data can be dropped. +* TODO it is better to reuse these clusters */ + QSIMPLEQ_FOREACH_SAFE(tb, &drop_tables, entry, tb_next) { +free_bitmap_clusters (bs, tb); +g_free (tb); + } + +bitmap_list_free (bm_list); +return ; + +fail: + QSIMPLEQ_FOREACH(bm, bm_list, entry) { +if (bm->dirty_bitmap ==NULL || bm->table.offset ==0 ) { +continue ; + } + +free_bitmap_clusters (bs, &bm->table); + } + + QSIMPLEQ_FOREACH_SAFE(tb, &drop_tables, entry, tb_next) { +g_free (tb); + } + +bitmap_list_free (bm_list); +}
block/qcow2.c +9 -0
@@ -1824,6 +1824,7 @@ static int qcow2_inactivate(BlockDriverState *bs) {BDRVQcow2State *s = bs->opaque;int ret, result =0 ; +Error *local_err =NULL ; ret =qcow2_cache_flush (bs, s->l2_table_cache);if (ret) {
@@ -1839,6 +1840,14 @@ static int qcow2_inactivate(BlockDriverState *bs) strerror(-ret)); } +qcow2_store_persistent_dirty_bitmaps (bs, &local_err); +if (local_err !=NULL ) { + result = -EINVAL ; +error_report_err (local_err); +error_report ("Persistent bitmaps are lost for node '%s'" , +bdrv_get_device_or_node_name (bs)); + } +if (result ==0 ) {qcow2_mark_clean (bs); }
block/qcow2.h +1 -0
@@ -631,5 +631,6 @@ int qcow2_check_bitmaps_refcounts(BlockDriverState *bs, BdrvCheckResult *res, int64_t *refcount_table_size);bool qcow2_load_autoloading_dirty_bitmaps (BlockDriverState *bs,Error **errp);int qcow2_reopen_bitmaps_rw (BlockDriverState *bs,Error **errp); +void qcow2_store_persistent_dirty_bitmaps (BlockDriverState *bs,Error **errp);#endif
Signature · 1 line
2.11.1
Re: [Qemu-devel] [PATCH v22 19/30] qcow2: add persistent dirty bitmaps support
Quoted 14 lines · Vladimir Sementsov-Ogievskiy
Store persistent dirty bitmaps in qcow2 image.
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>Reviewed-by Max Reitz <[email protected]>--- block/qcow2-bitmap.c | 475 +++++++++++++++++++++++++++++++++++++++++++++++++++ block/qcow2.c | 9 + block/qcow2.h | 1 + 3 files changed, 485 insertions(+)block/qcow2-bitmap.c +3
-0+ +/* store_bitmap_data() +* Store bitmap to image, filling bitmap table accordingly.
block/qcow2-bitmap.c out: · line 1072
Quoted 10 lines
block/qcow2-bitmap.c +10
-0+*/ +static uint64_t *store_bitmap_data (BlockDriverState *bs, +BdrvDirtyBitmap *bitmap, +uint32_t *bitmap_table_size,Error **errp) +{ +int ret; +BDRVQcow2State *s = bs->opaque; +int64_t sector; +uint64_t sbc; +uint64_t bm_size =bdrv_dirty_bitmap_size (bitmap);
This grabs the size (currently in sectors, although I plan to fix it to be in bytes)...
block/qcow2-bitmap.c out: · line 1079
Quoted 7 lines
block/qcow2-bitmap.c +7
-0+const char *bm_name =bdrv_dirty_bitmap_name (bitmap); +uint8_t *buf =NULL ; +BdrvDirtyBitmapIter *dbi; +uint64_t *tb; +uint64_t tb_size = +size_to_clusters (s, +bdrv_dirty_bitmap_serialization_size (bitmap,0 , bm_size));
...then finds out how many bytes are required to serialize the entire image, where bm_size should be the same as before...
block/qcow2-bitmap.c out: · line 1107
Quoted 9 lines
block/qcow2-bitmap.c +9
-0+while ((sector =bdrv_dirty_iter_next (dbi)) !=-1 ) { +uint64_t cluster = sector / sbc; +uint64_t end, write_size; +int64_t off; + + sector = cluster * sbc; + end =MIN (bm_size, sector + sbc); + write_size = +bdrv_dirty_bitmap_serialization_size (bitmap, sector, end - sector);
But here, rather than tackling the entire image at once, you are subdividing your queries along arbitrary lines. But nowhere do I see a call to bdrv_dirty_bitmap_serialization_align() to make sure your end-sector value is properly aligned; if it is not aligned, you will trigger an assertion failure here...
block/qcow2-bitmap.c out: · line 1119
Quoted 12 lines
block/qcow2-bitmap.c +12
-0+assert (write_size <= s->cluster_size); + + off =qcow2_alloc_clusters (bs, s->cluster_size); +if (off <0 ) { +error_setg_errno (errp, -off, +"Failed to allocate clusters for bitmap '%s'" , + bm_name); + goto fail; + } + tb[cluster] = off; + +bdrv_dirty_bitmap_serialize_part (bitmap, buf, sector, end - sector);
...and again here, during serialization_chunk().
I don't know if that means you need a v23 series, or if we can just patch it in as a followup (perhaps by having me add the usage during my byte-based dirty-bitmap series). I guess it depends on whether we can come up with any bitmap granularity (between 512 bytes and 2M is all the more we are currently supporting, right?) that differs from the qcow2 cluster granularity in a manner that iteration by qcow2 clusters is no longer guaranteed to be bitmap-granularity aligned, and thus trigger an assertion failure on your code as-is.
I also think it's pretty gross to be calculating the iteration bounds by sectors rather than bytes, when we are really worried about clusters (it's easier to track just bytes/clusters than it is to track bytes/sectors/clusters) - but that one is more along the lines of the sector-to-byte conversions I've been tackling, so I won't insist on you changing it if there is no other reason for a v23.
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
Re: [Qemu-devel] [PATCH v22 19/30] qcow2: add persistent dirty bitmaps support
Quoted 50 lines · Eric Blake
Quoted 14 lines · Vladimir Sementsov-Ogievskiy
Store persistent dirty bitmaps in qcow2 image.
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>Reviewed-by Max Reitz <[email protected]>--- block/qcow2-bitmap.c | 475 +++++++++++++++++++++++++++++++++++++++++++++++++++ block/qcow2.c | 9 + block/qcow2.h | 1 + 3 files changed, 485 insertions(+)block/qcow2-bitmap.c +3
-0+ +/* store_bitmap_data() +* Store bitmap to image, filling bitmap table accordingly. block/qcow2-bitmap.c out: · line 1072
Quoted 10 lines
block/qcow2-bitmap.c +10
-0+*/ +static uint64_t *store_bitmap_data (BlockDriverState *bs, +BdrvDirtyBitmap *bitmap, +uint32_t *bitmap_table_size,Error **errp) +{ +int ret; +BDRVQcow2State *s = bs->opaque; +int64_t sector; +uint64_t sbc; +uint64_t bm_size =bdrv_dirty_bitmap_size (bitmap);This grabs the size (currently in sectors, although I plan to fix it to be in bytes)...
block/qcow2-bitmap.c out: · line 1079
Quoted 7 lines
block/qcow2-bitmap.c +7
-0+const char *bm_name =bdrv_dirty_bitmap_name (bitmap); +uint8_t *buf =NULL ; +BdrvDirtyBitmapIter *dbi; +uint64_t *tb; +uint64_t tb_size = +size_to_clusters (s, +bdrv_dirty_bitmap_serialization_size (bitmap,0 , bm_size));...then finds out how many bytes are required to serialize the entire image, where
bm_sizeshould be the same as before...block/qcow2-bitmap.c out: · line 1107
Quoted 9 lines
block/qcow2-bitmap.c +9
-0+while ((sector =bdrv_dirty_iter_next (dbi)) !=-1 ) { +uint64_t cluster = sector / sbc; +uint64_t end, write_size; +int64_t off; + + sector = cluster * sbc; + end =MIN (bm_size, sector + sbc); + write_size = +bdrv_dirty_bitmap_serialization_size (bitmap, sector, end - sector);
Quoted 5 lines
But here, rather than tackling the entire image at once, you are subdividing your queries along arbitrary lines. But nowhere do I see a call to
bdrv_dirty_bitmap_serialization_align()to make sure your end-sector value is properly aligned; if it is not aligned, you will trigger an assertion failure here...
It's 4:21 am here, so I cannot claim to be right, but if I remember correctly, it will automatically aligned because sbc is the number of bits (and thus sectors) a bitmap cluster covers.
Quoted 25 lines
block/qcow2-bitmap.c out: · line 1119
Quoted 13 lines
block/qcow2-bitmap.c +12
-0+assert (write_size <= s->cluster_size); + + off =qcow2_alloc_clusters (bs, s->cluster_size); +if (off <0 ) { +error_setg_errno (errp, -off, +"Failed to allocate clusters for bitmap '%s'" , + bm_name); + goto fail; + } + tb[cluster] = off; + +bdrv_dirty_bitmap_serialize_part (bitmap, buf, sector, end - sector);...and again here, during
serialization_chunk().I don't know if that means you need a v23 series, or if we can just patch it in as a followup (perhaps by having me add the usage during my byte-based dirty-bitmap series). I guess it depends on whether we can come up with any bitmap granularity (between 512 bytes and 2M is all the more we are currently supporting, right?) that differs from the qcow2 cluster granularity in a manner that iteration by qcow2 clusters is no longer guaranteed to be bitmap-granularity aligned, and thus trigger an assertion failure on your code as-is.
Quoted 6 lines
I also think it's pretty gross to be calculating the iteration bounds by sectors rather than bytes, when we are really worried about clusters (it's easier to track just bytes/clusters than it is to track bytes/sectors/clusters) - but that one is more along the lines of the sector-to-byte conversions I've been tackling, so I won't insist on you changing it if there is no other reason for a v23.
I'm for fixing it later. I would have announced the series "applied" an hour ago if I hadn't had to bisect iotest 055 breakage...
Max
Re: [Qemu-devel] [PATCH v22 19/30] qcow2: add persistent dirty bitmaps support
Quoted 39 lines · Max Reitz
Quoted 32 lines · Eric Blake
Quoted 5 lines · Vladimir Sementsov-Ogievskiy
Store persistent dirty bitmaps in qcow2 image.
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>Reviewed-by Max Reitz <[email protected]>---
This grabs the size (currently in sectors, although I plan to fix it to be in bytes)...
block/qcow2-bitmap.c out: · line 1079
Quoted 7 lines
block/qcow2-bitmap.c +7
-0+const char *bm_name =bdrv_dirty_bitmap_name (bitmap); +uint8_t *buf =NULL ; +BdrvDirtyBitmapIter *dbi; +uint64_t *tb; +uint64_t tb_size = +size_to_clusters (s, +bdrv_dirty_bitmap_serialization_size (bitmap,0 , bm_size));...then finds out how many bytes are required to serialize the entire image, where
bm_sizeshould be the same as before...block/qcow2-bitmap.c out: · line 1107
Quoted 9 lines
block/qcow2-bitmap.c +9
-0+while ((sector =bdrv_dirty_iter_next (dbi)) !=-1 ) { +uint64_t cluster = sector / sbc; +uint64_t end, write_size; +int64_t off; + + sector = cluster * sbc; + end =MIN (bm_size, sector + sbc); + write_size = +bdrv_dirty_bitmap_serialization_size (bitmap, sector, end - sector);Quoted 5 lines
But here, rather than tackling the entire image at once, you are subdividing your queries along arbitrary lines. But nowhere do I see a call to
bdrv_dirty_bitmap_serialization_align()to make sure your end-sector value is properly aligned; if it is not aligned, you will trigger an assertion failure here...
Quoted 3 lines
It's 4:21 am here, so I cannot claim to be right, but if I remember correctly, it will automatically aligned because sbc is the number of bits (and thus sectors) a bitmap cluster covers.
Okay, I re-read the spec. First thing I noticed: we have a potential conflict if an image is allowed to be resized:
"All stored bitmaps are related to the virtual disk stored in the same image, so each bitmap size is equal to the virtual disk size."
If you resize an image, does the bitmap size have to be adjusted as well? What if you create one bitmap, then take an internal snapshot, then resize? Or do we declare that (at least for now) the presence of a bitmap is incompatible with the use of an internal snapshot?
Conversely, we state that:
"Structure of a bitmap directory entry:
...
8 - 11: bitmap_table_size
Number of entries in the bitmap table of the bitmap."
Since a bitmap therefore tracks its own size, I think the earlier statement that all bitmap sizes are equal to the virtual disk size is too strict (there seems to be no technical reason why a bitmap can't have a different size that the image).
But, having read that, you are correct that we are subdividing our bitmaps according to what fits in a qcow2 cluster, and the smallest qcow2 cluster we can come up with is 512-bytes (or 4k bits of the bitmap). Checking hbitmap.c, we are merely asserting that our alignment is always a multiple of 64 << hb->granularity. But since we are partitioning a cluster at a time, our minimum alignment will be 512 << hb->granularity, which is always aligned.
So all the more I need to do is add an assertion.
Quoted 2 lines
I'm for fixing it later. I would have announced the series "applied" an hour ago if I hadn't had to bisect iotest 055 breakage...
I'm working it into my v4 posting of dirty-bitmap cleanups.
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
Re: [Qemu-devel] [PATCH v22 19/30] qcow2: add persistent dirty bitmaps support
Quoted 55 lines · Eric Blake
Quoted 39 lines · Max Reitz
Quoted 32 lines · Eric Blake
Quoted 5 lines · Vladimir Sementsov-Ogievskiy
Store persistent dirty bitmaps in qcow2 image.
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>Reviewed-by Max Reitz <[email protected]>---
This grabs the size (currently in sectors, although I plan to fix it to be in bytes)...
block/qcow2-bitmap.c out: · line 1079
Quoted 7 lines
block/qcow2-bitmap.c +7
-0+const char *bm_name =bdrv_dirty_bitmap_name (bitmap); +uint8_t *buf =NULL ; +BdrvDirtyBitmapIter *dbi; +uint64_t *tb; +uint64_t tb_size = +size_to_clusters (s, +bdrv_dirty_bitmap_serialization_size (bitmap,0 , bm_size));...then finds out how many bytes are required to serialize the entire image, where
bm_sizeshould be the same as before...block/qcow2-bitmap.c out: · line 1107
Quoted 9 lines
block/qcow2-bitmap.c +9
-0+while ((sector =bdrv_dirty_iter_next (dbi)) !=-1 ) { +uint64_t cluster = sector / sbc; +uint64_t end, write_size; +int64_t off; + + sector = cluster * sbc; + end =MIN (bm_size, sector + sbc); + write_size = +bdrv_dirty_bitmap_serialization_size (bitmap, sector, end - sector);Quoted 5 lines
But here, rather than tackling the entire image at once, you are subdividing your queries along arbitrary lines. But nowhere do I see a call to
bdrv_dirty_bitmap_serialization_align()to make sure your end-sector value is properly aligned; if it is not aligned, you will trigger an assertion failure here...Quoted 3 lines
It's 4:21 am here, so I cannot claim to be right, but if I remember correctly, it will automatically aligned because sbc is the number of bits (and thus sectors) a bitmap cluster covers.
Okay, I re-read the spec. First thing I noticed: we have a potential conflict if an image is allowed to be resized:
"All stored bitmaps are related to the virtual disk stored in the same image, so each bitmap size is equal to the virtual disk size."
If you resize an image, does the bitmap size have to be adjusted as well? What if you create one bitmap, then take an internal snapshot, then resize? Or do we declare that (at least for now) the presence of a bitmap is incompatible with the use of an internal snapshot?
Quoted 8 lines
Conversely, we state that:
"Structure of a bitmap directory entry:
...
8 - 11:bitmap_table_size
Number of entries in the bitmap table of the bitmap."
This is the number of bitmaps stored in the qcow2, not the size of one particular bitmap.
Quoted 21 lines
Since a bitmap therefore tracks its own size, I think the earlier statement that all bitmap sizes are equal to the virtual disk size is too strict (there seems to be no technical reason why a bitmap can't have a different size that the image).
But, having read that, you are correct that we are subdividing our bitmaps according to what fits in a qcow2 cluster, and the smallest qcow2 cluster we can come up with is 512-bytes (or 4k bits of the bitmap). Checking
hbitmap.c, we are merely asserting that our alignment is always a multiple of 64 <<hb->granularity. But since we are partitioning a cluster at a time, our minimum alignment will be 512 <<hb->granularity, which is always aligned.So all the more I need to do is add an assertion.
Quoted 2 lines
I'm for fixing it later. I would have announced the series "applied" an hour ago if I hadn't had to bisect iotest 055 breakage...
I'm working it into my v4 posting of dirty-bitmap cleanups.
Re: [Qemu-devel] [PATCH v22 19/30] qcow2: add persistent dirty bitmaps support
Quoted 9 lines · John Snow
Quoted 6 lines
"Structure of a bitmap directory entry:
...
8 - 11:bitmap_table_size
Number of entries in the bitmap table of the bitmap."This is the number of bitmaps stored in the qcow2, not the size of one particular bitmap.
No, I was quoting from "Structure of a bitmap directory entry" (the same struct that also includes a per-bitmap granularity).
However, on re-reading the difference between the bitmap table and the bitmap data, I see that the bitmap_table_size it formally the number of cluster mappings that the bitmap occupies - so while it is not a precise size of the bitmap, it IS an approximation (where scaling has introduced lost precision for all sizes that map within the final cluster). But my point remains - we have some imprecision on whether bitmap_table_size has to be clamped to the same number as you would get if the current virtual disk size is converted into bitmaps, or whether it is valid to have a qcow2 file where a bitmap table contains fewer cluster mappings than would be required to cover the whole virtual file, or conversely contains more cluster mappings than what you get even when rounding the virtual table size up to the bitmap granularity and further scaled by how many bits fit per cluster.
Concretely, if I'm using 512-byte clusters (the qcow2 minimum), and qcow2 forces me to use a bitmap granularity no smaller than 9 (each bit of the bitmap covers 512 bytes), then one cluster of bitmap data covers 2M of virtual size. If I have an image with a virtual size of exactly 4M, and a bitmap covering that image, then my bitmap table _should_ have exactly two mappings (bitmap_table_size should be 2, because it required 2 8-byte entries in the table to give the mapping for the 2 clusters used to contain all the bits of the bitmap; the remaining 496 bytes of the bitmap table should be 0 because there are no further mappings). But note that any size in the range (2M, 4M] as the same bitmap_table_size of 2 for that granularity.
If the bitmap is tied to the image (has the 'auto' and/or 'dirty' bit set), then I agree that the bitmap size should match the virtual image size. But if the bitmap is independent, what technical reason do we have that prevents us from having a bitmap covering 2M or less of data (bitmap_table_size of 1), or more than 4M of data (bitmap_table_size of 3 or more), even though it has no relation to the current virtual image size of 4M?
Meanwhile, if I use a bitmap granularity of 1k instead of 512, in the same image with 512-byte clusters, a virtual size of 2M is indistinguishable from a virtual size of 4M (both bitmaps fit within a single cluster, or bitmap_table_size of 1; although the 2M case only uses 256 of the 512 bytes in the bitmap data cluster).
I'm worried that we are too strict in stating that ALL bitmaps are tied to the current virtual image size, but I'm also worried that our spec might not be strict enough in enforcing that certain bitmaps MUST match the virtual size (if the bitmap is automatically tracking writes to the virtual image); and also worried whether we are setting ourselves up for obscure failures based on the interaction of resize and/or internal snapshots when bitmaps are in play (or whether we are being conservative and forbidding those interactions until we've had time to further prove that we handle their interaction safely).
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
Re: [Qemu-devel] [PATCH v22 19/30] qcow2: add persistent dirty bitmaps support
In reply to Eric BlakeQuoted 40 lines · Eric Blake
Quoted 39 lines · Max Reitz
Quoted 37 lines · Eric Blake
Quoted 5 lines · Vladimir Sementsov-Ogievskiy
Store persistent dirty bitmaps in qcow2 image.
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>Reviewed-by Max Reitz <[email protected]>---
This grabs the size (currently in sectors, although I plan to fix it to be in bytes)...
block/qcow2-bitmap.c out: · line 1079
Quoted 7 lines
block/qcow2-bitmap.c +7
-0+const char *bm_name =bdrv_dirty_bitmap_name (bitmap); +uint8_t *buf =NULL ; +BdrvDirtyBitmapIter *dbi; +uint64_t *tb; +uint64_t tb_size = +size_to_clusters (s, +bdrv_dirty_bitmap_serialization_size (bitmap,0 , bm_size));...then finds out how many bytes are required to serialize the entire image, where
bm_sizeshould be the same as before...block/qcow2-bitmap.c out: · line 1107
Quoted 9 lines
block/qcow2-bitmap.c +9
-0+while ((sector =bdrv_dirty_iter_next (dbi)) !=-1 ) { +uint64_t cluster = sector / sbc; +uint64_t end, write_size; +int64_t off; + + sector = cluster * sbc; + end =MIN (bm_size, sector + sbc); + write_size = +bdrv_dirty_bitmap_serialization_size (bitmap, sector, end - sector);But here, rather than tackling the entire image at once, you are subdividing your queries along arbitrary lines. But nowhere do I see a call to
bdrv_dirty_bitmap_serialization_align()to make sure your end-sector value is properly aligned; if it is not aligned, you will trigger an assertion failure here...
Quoted 6 lines
Quoted 3 lines
It's 4:21 am here, so I cannot claim to be right, but if I remember correctly, it will automatically aligned because sbc is the number of bits (and thus sectors) a bitmap cluster covers.
Okay, I re-read the spec. First thing I noticed: we have a potential conflict if an image is allowed to be resized:
It isn't, though. (At least currently qemu won't allow it.)
Quoted 5 lines
"All stored bitmaps are related to the virtual disk stored in the same image, so each bitmap size is equal to the virtual disk size."
If you resize an image, does the bitmap size have to be adjusted as well?
Sure.
Quoted 2 lines
What if you create one bitmap, then take an internal snapshot,
then resize?
v3 images store the virtual disk size for each snapshot. So resizing an image leaves snapshots unaffected.
Now bitmaps are not (yet) tied to snapshots. The spec thus says that the bitmaps always correspond to the active state (it lists "snapshot switching" as something that should dirty dirty bitmaps).
Therefore, if you then resize (which currently is doubly forbidden, because qemu won't allow you to resize images with bitmaps or snapshots), you would have to resize the bitmap, too.
In the future, we might want to allow binding bitmaps to snapshots. Then the bitmap would not be resized but you couldn't see it anymore unless you go back to the snapshot (which would bring the image to its original size).
Quoted 2 lines
Or do we declare that (at least for now) the presence of a
bitmap is incompatible with the use of an internal snapshot?
Well, that's not what the spec says. It clearly lists "snapshot switching" as something that dirty bitmaps track.
Quoted 8 lines
Conversely, we state that:
"Structure of a bitmap directory entry:
...
8 - 11:bitmap_table_size
Number of entries in the bitmap table of the bitmap."
Quoted 4 lines
Since a bitmap therefore tracks its own size, I think the earlier statement that all bitmap sizes are equal to the virtual disk size is too strict (there seems to be no technical reason why a bitmap can't have a different size that the image).
But a logical. qcow2 is not a container format, it's an image format. I for one am very opposed to storing bitmaps in a qcow2 file which have no immediate connection to that virtual disk.
A reason I can see myself supporting would be that you could tie bitmaps to snapshots and then they can have a size that differs from the one of the active layer.
Max
Quoted 15 lines
But, having read that, you are correct that we are subdividing our bitmaps according to what fits in a qcow2 cluster, and the smallest qcow2 cluster we can come up with is 512-bytes (or 4k bits of the bitmap). Checking
hbitmap.c, we are merely asserting that our alignment is always a multiple of 64 <<hb->granularity. But since we are partitioning a cluster at a time, our minimum alignment will be 512 <<hb->granularity, which is always aligned.So all the more I need to do is add an assertion.
Quoted 2 lines
I'm for fixing it later. I would have announced the series "applied" an hour ago if I hadn't had to bisect iotest 055 breakage...
I'm working it into my v4 posting of dirty-bitmap cleanups.
PATCH v22 20/30 qcow2: store bitmaps on reopening image as read-only
In reply to Vladimir Sementsov-OgievskiyStore bitmaps and mark them read-only on reopening image as read-only.
--- block/qcow2-bitmap.c | 22 ++++++++++++++++++++++ block/qcow2.c | 5 +++++ block/qcow2.h | 1 + 3 files changed, 28 insertions(+)
block/qcow2-bitmap.c +22 -0
@@ -1365,3 +1365,25 @@ fail:bitmap_list_free (bm_list); } + +int qcow2_reopen_bitmaps_ro (BlockDriverState *bs,Error **errp) +{ +BdrvDirtyBitmap *bitmap; +Error *local_err =NULL ; + +qcow2_store_persistent_dirty_bitmaps (bs, &local_err); +if (local_err !=NULL ) { +error_propagate (errp, local_err); +return -EINVAL ; + } + +for (bitmap =bdrv_dirty_bitmap_next (bs,NULL ); bitmap !=NULL ; + bitmap =bdrv_dirty_bitmap_next (bs, bitmap)) + { +if (bdrv_dirty_bitmap_get_persistance (bitmap)) { +bdrv_dirty_bitmap_set_readonly (bitmap, true); + } + } + +return 0 ; +}
block/qcow2.c +5 -0
@@ -1382,6 +1382,11 @@ static int qcow2_reopen_prepare(BDRVReopenState *state,/* We need to write out any unwritten data if we reopen read-only. */ if ((state->flags &BDRV_O_RDWR ) ==0 ) { + ret =qcow2_reopen_bitmaps_ro (state->bs, errp); +if (ret <0 ) { + goto fail; + } + ret =bdrv_flush (state->bs);if (ret <0 ) { goto fail;
block/qcow2.h +1 -0
@@ -632,5 +632,6 @@ int qcow2_check_bitmaps_refcounts(BlockDriverState *bs, BdrvCheckResult *res,bool qcow2_load_autoloading_dirty_bitmaps (BlockDriverState *bs,Error **errp);int qcow2_reopen_bitmaps_rw (BlockDriverState *bs,Error **errp);void qcow2_store_persistent_dirty_bitmaps (BlockDriverState *bs,Error **errp); +int qcow2_reopen_bitmaps_ro (BlockDriverState *bs,Error **errp);#endif
Signature · 1 line
2.11.1
PATCH v22 21/30 block: add bdrv_can_store_new_dirty_bitmap
In reply to Vladimir Sementsov-OgievskiyThis will be needed to check some restrictions before making bitmap persistent in qmp-block-dirty-bitmap-add (this functionality will be added by future patch)
--- block.c | 22 ++++++++++++++++++++++ include/block/block.h | 3 +++ include/block/block_int.h | 4 ++++ 3 files changed, 29 insertions(+)
block.c +22 -0
@@ -4954,3 +4954,25 @@ void bdrv_del_child(BlockDriverState *parent_bs, BdrvChild *child, Error **errp) parent_bs->drv->bdrv_del_child (parent_bs, child, errp); } + +bool bdrv_can_store_new_dirty_bitmap (BlockDriverState *bs,const char *name, +uint32_t granularity,Error **errp) +{ +BlockDriver *drv = bs->drv; + +if (!drv) { +error_setg_errno (errp,ENOMEDIUM , +"Can't store persistent bitmaps to %s" , +bdrv_get_device_or_node_name (bs)); +return false; + } + +if (!drv->bdrv_can_store_new_dirty_bitmap) { +error_setg_errno (errp,ENOTSUP , +"Can't store persistent bitmaps to %s" , +bdrv_get_device_or_node_name (bs)); +return false; + } + +return drv->bdrv_can_store_new_dirty_bitmap (bs, name, granularity, errp); +}
include/block/block.h +3 -0
@@ -630,4 +630,7 @@ void bdrv_add_child(BlockDriverState *parent, BlockDriverState *child, Error **errp);void bdrv_del_child (BlockDriverState *parent,BdrvChild *child,Error **errp); +bool bdrv_can_store_new_dirty_bitmap (BlockDriverState *bs,const char *name, +uint32_t granularity,Error **errp); +#endif
include/block/block_int.h +4 -0
@@ -387,6 +387,10 @@ struct BlockDriver {* field of BlockDirtyBitmap's in case of success. */ int (*bdrv_reopen_bitmaps_rw)(BlockDriverState *bs,Error **errp); +bool (*bdrv_can_store_new_dirty_bitmap)(BlockDriverState *bs, +const char *name, +uint32_t granularity, +Error **errp);QLIST_ENTRY (BlockDriver) list; };
Signature · 1 line
2.11.1
PATCH v22 22/30 qcow2: add .bdrv_can_store_new_dirty_bitmap
In reply to Vladimir Sementsov-OgievskiyRealize .bdrv_can_store_new_dirty_bitmap interface.
--- block/qcow2-bitmap.c | 51 +++++++++++++++++++++++++++++++++++++++++++++++++++ block/qcow2.c | 1 + block/qcow2.h | 4 ++++ 3 files changed, 56 insertions(+)
block/qcow2-bitmap.c +51 -0
@@ -1387,3 +1387,54 @@ int qcow2_reopen_bitmaps_ro(BlockDriverState *bs, Error **errp)return 0 ; } + +bool qcow2_can_store_new_dirty_bitmap (BlockDriverState *bs, +const char *name, +uint32_t granularity, +Error **errp) +{ +BDRVQcow2State *s = bs->opaque; +bool found; +Qcow2BitmapList *bm_list; + +if (check_constraints_on_bitmap (bs, name, granularity, errp) !=0 ) { + goto fail; + } + +if (s->nb_bitmaps ==0 ) { +return true; + } + +if (s->nb_bitmaps >=QCOW2_MAX_BITMAPS ) { +error_setg (errp, +"Maximum number of persistent bitmaps is already reached" ); + goto fail; + } + +if (s->bitmap_directory_size +calc_dir_entry_size (strlen (name),0 ) > +QCOW2_MAX_BITMAP_DIRECTORY_SIZE ) + { +error_setg (errp,"Not enough space in the bitmap directory" ); + goto fail; + } + + bm_list =bitmap_list_load (bs, s->bitmap_directory_offset, + s->bitmap_directory_size, errp); +if (bm_list ==NULL ) { + goto fail; + } + + found =find_bitmap_by_name (bm_list, name); +bitmap_list_free (bm_list); +if (found) { +error_setg (errp,"Bitmap with the same name is already stored" ); + goto fail; + } + +return true; + +fail: +error_prepend (errp,"Can't make bitmap '%s' persistent in '%s': " , + name,bdrv_get_device_or_node_name (bs)); +return false; +}
block/qcow2.c +1 -0
@@ -3611,6 +3611,7 @@ BlockDriver bdrv_qcow2 = { .bdrv_attach_aio_context = qcow2_attach_aio_context, .bdrv_reopen_bitmaps_rw = qcow2_reopen_bitmaps_rw, + .bdrv_can_store_new_dirty_bitmap = qcow2_can_store_new_dirty_bitmap, };static void bdrv_qcow2_init (void )
block/qcow2.h +4 -0
@@ -633,5 +633,9 @@ bool qcow2_load_autoloading_dirty_bitmaps(BlockDriverState *bs, Error **errp);int qcow2_reopen_bitmaps_rw (BlockDriverState *bs,Error **errp);void qcow2_store_persistent_dirty_bitmaps (BlockDriverState *bs,Error **errp);int qcow2_reopen_bitmaps_ro (BlockDriverState *bs,Error **errp); +bool qcow2_can_store_new_dirty_bitmap (BlockDriverState *bs, +const char *name, +uint32_t granularity, +Error **errp);#endif
Signature · 1 line
2.11.1
PATCH v22 23/30 qmp: add persistent flag to block-dirty-bitmap-add
In reply to Vladimir Sementsov-OgievskiyAdd optional 'persistent' flag to qmp command block-dirty-bitmap-add. Default is false.
--- blockdev.c | 18 +++++++++++++++++- qapi/block-core.json | 8 +++++++- 2 files changed, 24 insertions(+), 2 deletions(-)
blockdev.c +17 -1
@@ -1973,6 +1973,7 @@ static void block_dirty_bitmap_add_prepare(BlkActionState *common,/* AIO context taken and released within qmp_block_dirty_bitmap_add */ qmp_block_dirty_bitmap_add (action->node, action->name, action->has_granularity, action->granularity, + action->has_persistent, action->persistent, &local_err);if (!local_err) {
@@ -2720,9 +2721,11 @@ out:void qmp_block_dirty_bitmap_add (const char *node,const char *name,bool has_granularity,uint32_t granularity, +bool has_persistent,bool persistent,Error **errp) {BlockDriverState *bs; +BdrvDirtyBitmap *bitmap;if (!name || name[0 ] =='\0' ) {error_setg (errp,"Bitmap name cannot be empty" );
@@ -2745,7 +2748,20 @@ void qmp_block_dirty_bitmap_add(const char *node, const char *name, granularity =bdrv_get_default_bitmap_granularity (bs); } -bdrv_create_dirty_bitmap (bs, granularity, name, errp); +if (!has_persistent) { + persistent = false; + } + +if (persistent && + !bdrv_can_store_new_dirty_bitmap (bs, name, granularity, errp)) + { +return ; + } + + bitmap =bdrv_create_dirty_bitmap (bs, granularity, name, errp); +if (bitmap !=NULL ) { +bdrv_dirty_bitmap_set_persistance (bitmap, persistent); + } } void qmp_block_dirty_bitmap_remove(const char *node, const char *name,
qapi/block-core.json +7 -1
@@ -1561,10 +1561,16 @@ # @granularity: the bitmap granularity, default is 64k for # block-dirty-bitmap-add # +# @persistent: the bitmap is persistent, i.e. it will be saved to the +# corresponding block device image file on its close. For now only +# Qcow2 disks support persistent bitmaps. Default is false for +# block-dirty-bitmap-add. (Since: 2.10) +# # Since: 2.4 ## { 'struct': 'BlockDirtyBitmapAdd', - 'data': { 'node': 'str', 'name': 'str', '*granularity': 'uint32' } } + 'data': { 'node': 'str', 'name': 'str', '*granularity': 'uint32', + '*persistent': 'bool' } } ## # @block-dirty-bitmap-add:
Signature · 1 line
2.11.1
Re: [Qemu-devel] [PATCH v22 23/30] qmp: add persistent flag to block-dirty-bitmap-add
QAPI schema review only... I apologize for its lateness.
Quoted 8 lines · Vladimir Sementsov-Ogievskiy
Add optional 'persistent' flag to qmp command block-dirty-bitmap-add. Default is false.
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>Signed-off-by Denis V. Lunev <[email protected]>Reviewed-by Max Reitz <[email protected]>Reviewed-by John Snow <[email protected]>---
[...]
qapi/block-core.json line 1567
Quoted 12 lines
qapi/block-core.json +4
-0@@ -1561,10 +1561,16 @@ # @granularity: the bitmap granularity, default is 64k for # block-dirty-bitmap-add # +# @persistent: the bitmap is persistent, i.e. it will be saved to the +# corresponding block device image file on its close. For now only +# Qcow2 disks support persistent bitmaps. Default is false for +# block-dirty-bitmap-add. (Since: 2.10)
"for block-dirty-bitmap-add" suggests there could be other users, with different (but unspecified) defaults. What about replacing the sentence by "(default: false)"?
Please wrap your comment lines around column 70.
Quoted 10 lines
qapi/block-core.json +3
-1+# # Since: 2.4 ## { 'struct': 'BlockDirtyBitmapAdd', - 'data': { 'node': 'str', 'name': 'str', '*granularity': 'uint32' } } + 'data': { 'node': 'str', 'name': 'str', '*granularity': 'uint32', + '*persistent': 'bool' } } ## # @block-dirty-bitmap-add:
Re: [Qemu-devel] [PATCH v22 23/30] qmp: add persistent flag to block-dirty-bitmap-add
Quoted 13 lines · Markus Armbruster
QAPI schema review only... I apologize for its lateness.
Quoted 8 lines · Vladimir Sementsov-Ogievskiy
Add optional 'persistent' flag to qmp command block-dirty-bitmap-add. Default is false.
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>Signed-off-by Denis V. Lunev <[email protected]>Reviewed-by Max Reitz <[email protected]>Reviewed-by John Snow <[email protected]>---
[...]
Quoted 15 lines
qapi/block-core.json line 1567
Quoted 12 lines
qapi/block-core.json +4
-0@@ -1561,10 +1561,16 @@ # @granularity: the bitmap granularity, default is 64k for # block-dirty-bitmap-add # +# @persistent: the bitmap is persistent, i.e. it will be saved to the +# corresponding block device image file on its close. For now only +# Qcow2 disks support persistent bitmaps. Default is false for +# block-dirty-bitmap-add. (Since: 2.10)"for block-dirty-bitmap-add" suggests there could be other users, with different (but unspecified) defaults. What about replacing the sentence by "(default: false)"?
then, this should be done in other places, @granularity is an example.
Quoted 13 lines
Please wrap your comment lines around column 70.
Quoted 10 lines
qapi/block-core.json +3
-1+# # Since: 2.4 ## { 'struct': 'BlockDirtyBitmapAdd', - 'data': { 'node': 'str', 'name': 'str', '*granularity': 'uint32' } } + 'data': { 'node': 'str', 'name': 'str', '*granularity': 'uint32', + '*persistent': 'bool' } } ## # @block-dirty-bitmap-add:
Signature · 2 lines
Best regards, Vladimir
Re: [Qemu-devel] [PATCH v22 23/30] qmp: add persistent flag to block-dirty-bitmap-add
In reply to Markus ArmbrusterQuoted 29 lines · Markus Armbruster
QAPI schema review only... I apologize for its lateness.
Quoted 8 lines · Vladimir Sementsov-Ogievskiy
Add optional 'persistent' flag to qmp command block-dirty-bitmap-add. Default is false.
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>Signed-off-by Denis V. Lunev <[email protected]>Reviewed-by Max Reitz <[email protected]>Reviewed-by John Snow <[email protected]>---
[...]
qapi/block-core.json line 1567
Quoted 12 lines
qapi/block-core.json +4
-0@@ -1561,10 +1561,16 @@ # @granularity: the bitmap granularity, default is 64k for # block-dirty-bitmap-add # +# @persistent: the bitmap is persistent, i.e. it will be saved to the +# corresponding block device image file on its close. For now only +# Qcow2 disks support persistent bitmaps. Default is false for +# block-dirty-bitmap-add. (Since: 2.10)"for block-dirty-bitmap-add" suggests there could be other users, with different (but unspecified) defaults. What about replacing the sentence by "(default: false)"?
Quoted 1 line
Please wrap your comment lines around column 70.
Why 70, is it written somewhere? There are a lot of lines over 70 characters in this file, so, as series are already in Max's block branch I think it would be better to fix the whole file, if it is really needed.
Quoted 11 lines
Quoted 11 lines
qapi/block-core.json +3
-1+# # Since: 2.4 ## { 'struct': 'BlockDirtyBitmapAdd', - 'data': { 'node': 'str', 'name': 'str', '*granularity': 'uint32' } } + 'data': { 'node': 'str', 'name': 'str', '*granularity': 'uint32', + '*persistent': 'bool' } } ## # @block-dirty-bitmap-add:
Signature · 2 lines
Best regards, Vladimir
Re: [Qemu-devel] [PATCH v22 23/30] qmp: add persistent flag to block-dirty-bitmap-add
Quoted 30 lines · Vladimir Sementsov-Ogievskiy
Quoted 29 lines · Markus Armbruster
QAPI schema review only... I apologize for its lateness.
Quoted 8 lines · Vladimir Sementsov-Ogievskiy
Add optional 'persistent' flag to qmp command block-dirty-bitmap-add. Default is false.
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>Signed-off-by Denis V. Lunev <[email protected]>Reviewed-by Max Reitz <[email protected]>Reviewed-by John Snow <[email protected]>---
[...]
qapi/block-core.json line 1567
Quoted 12 lines
qapi/block-core.json +4
-0@@ -1561,10 +1561,16 @@ # @granularity: the bitmap granularity, default is 64k for # block-dirty-bitmap-add # +# @persistent: the bitmap is persistent, i.e. it will be saved to the +# corresponding block device image file on its close. For now only +# Qcow2 disks support persistent bitmaps. Default is false for +# block-dirty-bitmap-add. (Since: 2.10)"for block-dirty-bitmap-add" suggests there could be other users, with different (but unspecified) defaults. What about replacing the sentence by "(default: false)"?
Quoted 6 lines
Quoted 1 line
Please wrap your comment lines around column 70.
Why 70, is it written somewhere? There are a lot of lines over 70 characters in this file, so, as series are already in Max's block branch I think it would be better to fix the whole file, if it is really needed.
There's no hard rule on comment line length. I routinely advise people to wrap anyway, for legibility. Humans tend to have trouble following long lines with their eyes (I sure do). Typographic manuals suggest to limit columns to roughly 60 characters for exactly that reason[*].
I'm not the maintainer here, so this is really advice, not a demand.
[*] https://en.wikipedia.org/wiki/Column_(typography)#Typographic_style
PATCH v22 24/30 qmp: add autoload parameter to block-dirty-bitmap-add
In reply to Vladimir Sementsov-OgievskiyOptional. Default is false.
--- blockdev.c | 18 ++++++++++++++++-- qapi/block-core.json | 6 +++++- 2 files changed, 21 insertions(+), 3 deletions(-)
blockdev.c +16 -2
@@ -1974,6 +1974,7 @@ static void block_dirty_bitmap_add_prepare(BlkActionState *common,qmp_block_dirty_bitmap_add (action->node, action->name, action->has_granularity, action->granularity, action->has_persistent, action->persistent, + action->has_autoload, action->autoload, &local_err);if (!local_err) {
@@ -2722,6 +2723,7 @@ out:void qmp_block_dirty_bitmap_add (const char *node,const char *name,bool has_granularity,uint32_t granularity,bool has_persistent,bool persistent, +bool has_autoload,bool autoload,Error **errp) {BlockDriverState *bs;
@@ -2751,6 +2753,15 @@ void qmp_block_dirty_bitmap_add(const char *node, const char *name,if (!has_persistent) { persistent = false; } +if (!has_autoload) { + autoload = false; + } + +if (has_autoload && !persistent) { +error_setg (errp,"Autoload flag must be used only for persistent " +"bitmaps" ); +return ; + }if (persistent && !bdrv_can_store_new_dirty_bitmap (bs, name, granularity, errp))
@@ -2759,9 +2770,12 @@ void qmp_block_dirty_bitmap_add(const char *node, const char *name, } bitmap =bdrv_create_dirty_bitmap (bs, granularity, name, errp); -if (bitmap !=NULL ) { -bdrv_dirty_bitmap_set_persistance (bitmap, persistent); +if (bitmap ==NULL ) { +return ; } + +bdrv_dirty_bitmap_set_persistance (bitmap, persistent); +bdrv_dirty_bitmap_set_autoload (bitmap, autoload); } void qmp_block_dirty_bitmap_remove(const char *node, const char *name,
qapi/block-core.json +5 -1
@@ -1566,11 +1566,15 @@ # Qcow2 disks support persistent bitmaps. Default is false for # block-dirty-bitmap-add. (Since: 2.10) # +# @autoload: the bitmap will be automatically loaded when the image it is stored +# in is opened. This flag may only be specified for persistent +# bitmaps. Default is false for block-dirty-bitmap-add. (Since: 2.10) +# # Since: 2.4 ## { 'struct': 'BlockDirtyBitmapAdd', 'data': { 'node': 'str', 'name': 'str', '*granularity': 'uint32', - '*persistent': 'bool' } } + '*persistent': 'bool', '*autoload': 'bool' } } ## # @block-dirty-bitmap-add:
Signature · 1 line
2.11.1
Re: [Qemu-devel] [PATCH v22 24/30] qmp: add autoload parameter to block-dirty-bitmap-add
Quoted 10 lines · Vladimir Sementsov-Ogievskiy
Optional. Default is false.
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>Signed-off-by Denis V. Lunev <[email protected]>Reviewed-by Max Reitz <[email protected]>Reviewed-by John Snow <[email protected]>--- blockdev.c | 18 ++++++++++++++++-- qapi/block-core.json | 6 +++++- 2 files changed, 21 insertions(+), 3 deletions(-)
[...]
qapi/block-core.json line 1572
Quoted 12 lines
qapi/block-core.json +4
-0@@ -1566,11 +1566,15 @@ # Qcow2 disks support persistent bitmaps. Default is false for # block-dirty-bitmap-add. (Since: 2.10) # +# @autoload: the bitmap will be automatically loaded when the image it is stored +# in is opened. This flag may only be specified for persistent +# bitmaps. Default is false for block-dirty-bitmap-add. (Since: 2.10) +#
Comments on PATCH 23 apply.
Quoted 9 lines
qapi/block-core.json +1
-1# Since: 2.4 ## { 'struct': 'BlockDirtyBitmapAdd', 'data': { 'node': 'str', 'name': 'str', '*granularity': 'uint32', - '*persistent': 'bool' } } + '*persistent': 'bool', '*autoload': 'bool' } } ## # @block-dirty-bitmap-add:
PATCH v22 25/30 qmp: add x-debug-block-dirty-bitmap-sha256
In reply to Vladimir Sementsov-Ogievskiy--- block/dirty-bitmap.c | 5 +++++ blockdev.c | 25 +++++++++++++++++++++++++ include/block/dirty-bitmap.h | 1 + include/qemu/hbitmap.h | 8 ++++++++ qapi/block-core.json | 27 +++++++++++++++++++++++++++ tests/Makefile.include | 2 +- util/hbitmap.c | 11 +++++++++++ 7 files changed, 78 insertions(+), 1 deletion(-)
block/dirty-bitmap.c +5 -0
@@ -725,3 +725,8 @@ BdrvDirtyBitmap *bdrv_dirty_bitmap_next(BlockDriverState *bs,return bitmap ==NULL ?QLIST_FIRST (&bs->dirty_bitmaps) :QLIST_NEXT (bitmap, list); } + +char *bdrv_dirty_bitmap_sha256 (const BdrvDirtyBitmap *bitmap,Error **errp) +{ +return hbitmap_sha256 (bitmap->bitmap, errp); +}
blockdev.c +25 -0
@@ -2832,6 +2832,31 @@ void qmp_block_dirty_bitmap_clear(const char *node, const char *name,bdrv_clear_dirty_bitmap (bitmap,NULL ); } +BlockDirtyBitmapSha256 *qmp_x_debug_block_dirty_bitmap_sha256 (const char *node, +const char *name, +Error **errp) +{ +BdrvDirtyBitmap *bitmap; +BlockDriverState *bs; +BlockDirtyBitmapSha256 *ret =NULL ; +char *sha256; + + bitmap =block_dirty_bitmap_lookup (node, name, &bs, errp); +if (!bitmap || !bs) { +return NULL ; + } + + sha256 =bdrv_dirty_bitmap_sha256 (bitmap, errp); +if (sha256 ==NULL ) { +return NULL ; + } + + ret =g_new (BlockDirtyBitmapSha256,1 ); + ret->sha256 = sha256; + +return ret; +} +void hmp_drive_del (Monitor *mon,const QDict *qdict) {const char *id =qdict_get_str (qdict,"id" );
include/block/dirty-bitmap.h +1 -0
@@ -97,5 +97,6 @@ bool bdrv_dirty_bitmap_get_persistance(BdrvDirtyBitmap *bitmap);bool bdrv_has_changed_persistent_bitmaps (BlockDriverState *bs);BdrvDirtyBitmap *bdrv_dirty_bitmap_next (BlockDriverState *bs,BdrvDirtyBitmap *bitmap); +char *bdrv_dirty_bitmap_sha256 (const BdrvDirtyBitmap *bitmap,Error **errp);#endif
include/qemu/hbitmap.h +8 -0
@@ -253,6 +253,14 @@ void hbitmap_deserialize_ones(HBitmap *hb, uint64_t start, uint64_t count,void hbitmap_deserialize_finish (HBitmap *hb);/** +* hbitmap_sha256: +* @bitmap: HBitmap to operate on. +* +* Returns SHA256 hash of the last level. +*/ +char *hbitmap_sha256 (const HBitmap *bitmap,Error **errp); + +/** * hbitmap_free: * @hb: HBitmap to operate on. *
qapi/block-core.json +27 -0
@@ -1644,6 +1644,33 @@ 'data': 'BlockDirtyBitmap' } ## +# @BlockDirtyBitmapSha256: +# +# SHA256 hash of dirty bitmap data +# +# @sha256: ASCII representation of SHA256 bitmap hash +# +# Since: 2.10 +## + { 'struct': 'BlockDirtyBitmapSha256', + 'data': {'sha256': 'str'} } + +## +# @x-debug-block-dirty-bitmap-sha256: +# +# Get bitmap SHA256 +# +# Returns: BlockDirtyBitmapSha256 on success +# If @node is not a valid block device, DeviceNotFound +# If @name is not found or if hashing has failed, GenericError with an +# explanation +# +# Since: 2.10 +## + { 'command': 'x-debug-block-dirty-bitmap-sha256', + 'data': 'BlockDirtyBitmap', 'returns': 'BlockDirtyBitmapSha256' } + +## # @blockdev-mirror: # # Start mirroring a block device's writes to a new destination.
tests/Makefile.include +1 -1
@@ -553,7 +553,7 @@ tests/test-blockjob$(EXESUF): tests/test-blockjob.o $(test-block-obj-y) $(test-u tests/test-blockjob-txn$(EXESUF): tests/test-blockjob-txn.o $(test-block-obj-y) $(test-util-obj-y) tests/test-thread-pool$(EXESUF): tests/test-thread-pool.o $(test-block-obj-y) tests/test-iov$(EXESUF): tests/test-iov.o $(test-util-obj-y) -tests/test-hbitmap$(EXESUF): tests/test-hbitmap.o $(test-util-obj-y) +tests/test-hbitmap$(EXESUF): tests/test-hbitmap.o $(test-util-obj-y) $(test-crypto-obj-y) tests/test-x86-cpuid$(EXESUF): tests/test-x86-cpuid.o tests/test-xbzrle$(EXESUF): tests/test-xbzrle.o migration/xbzrle.o migration/page_cache.o $(test-util-obj-y) tests/test-cutils$(EXESUF): tests/test-cutils.o util/cutils.o
util/hbitmap.c +11 -0
@@ -13,6 +13,7 @@#include "qemu/hbitmap.h" #include "qemu/host-utils.h" #include "trace.h" +#include "crypto/hash.h" /* HBitmaps provides an array of bits. The bits are stored as usual in an * array of unsigned longs, but HBitmap is also optimized to provide fast
@@ -727,3 +728,13 @@ void hbitmap_free_meta(HBitmap *hb)hbitmap_free (hb->meta); hb->meta =NULL ; } + +char *hbitmap_sha256 (const HBitmap *bitmap,Error **errp) +{ +size_t size = bitmap->sizes[HBITMAP_LEVELS -1 ] *sizeof (unsigned long ); +char *data = (char *)bitmap->levels[HBITMAP_LEVELS -1 ]; +char *hash =NULL ; +qcrypto_hash_digest (QCRYPTO_HASH_ALG_SHA256 , data, size, &hash, errp); + +return hash; +}
Signature · 1 line
2.11.1
Re: [Qemu-devel] [PATCH v22 25/30] qmp: add x-debug-block-dirty-bitmap-sha256
Quoted 10 lines · Vladimir Sementsov-Ogievskiy
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>--- block/dirty-bitmap.c | 5 +++++ blockdev.c | 25 +++++++++++++++++++++++++ include/block/dirty-bitmap.h | 1 + include/qemu/hbitmap.h | 8 ++++++++ qapi/block-core.json | 27 +++++++++++++++++++++++++++ tests/Makefile.include | 2 +- util/hbitmap.c | 11 +++++++++++ 7 files changed, 78 insertions(+), 1 deletion(-)
[...]
qapi/block-core.json line 1651
Quoted 13 lines
qapi/block-core.json +5
-0@@ -1644,6 +1644,33 @@ 'data': 'BlockDirtyBitmap' } ## +# @BlockDirtyBitmapSha256: +# +# SHA256 hash of dirty bitmap data +# +# @sha256: ASCII representation of SHA256 bitmap hash
Spell it SHA-256, please. The member name @sha256 can stay.
SHA-256 is 256 binary bits. Please specify how they are represented in ASCII. It better be base64 (RFC 4648), because we use that elsewhere.
Quoted 5 lines
qapi/block-core.json +5
-0+# +# Since: 2.10 +## + { 'struct': 'BlockDirtyBitmapSha256', + 'data': {'sha256': 'str'} }
qapi/block-core.json line 1666
Quoted 10 lines
qapi/block-core.json +10
-0+ +## +# @x-debug-block-dirty-bitmap-sha256: +# +# Get bitmap SHA256 +# +# Returns: BlockDirtyBitmapSha256 on success +# If @node is not a valid block device, DeviceNotFound +# If @name is not found or if hashing has failed, GenericError with an +# explanation
Please wrap your comment lines around column 70.
Quoted 10 lines
qapi/block-core.json +7
-0+# +# Since: 2.10 +## + { 'command': 'x-debug-block-dirty-bitmap-sha256', + 'data': 'BlockDirtyBitmap', 'returns': 'BlockDirtyBitmapSha256' } + +## # @blockdev-mirror: # # Start mirroring a block device's writes to a new destination.
[...]
Re: [Qemu-devel] [PATCH v22 25/30] qmp: add x-debug-block-dirty-bitmap-sha256
Quoted 29 lines · Markus Armbruster
Quoted 10 lines · Vladimir Sementsov-Ogievskiy
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>--- block/dirty-bitmap.c | 5 +++++ blockdev.c | 25 +++++++++++++++++++++++++ include/block/dirty-bitmap.h | 1 + include/qemu/hbitmap.h | 8 ++++++++ qapi/block-core.json | 27 +++++++++++++++++++++++++++ tests/Makefile.include | 2 +- util/hbitmap.c | 11 +++++++++++ 7 files changed, 78 insertions(+), 1 deletion(-)[...]
qapi/block-core.json line 1651
Quoted 13 lines
qapi/block-core.json +5
-0@@ -1644,6 +1644,33 @@ 'data': 'BlockDirtyBitmap' } ## +# @BlockDirtyBitmapSha256: +# +# SHA256 hash of dirty bitmap data +# +# @sha256: ASCII representation of SHA256 bitmap hashSpell it SHA-256, please. The member name @sha256 can stay.
Quoted 2 lines
SHA-256 is 256 binary bits. Please specify how they are represented in ASCII. It better be base64 (RFC 4648), because we use that elsewhere.
It is filled later in this patch using qcrypto_hash_digest, so it is just a hex string representing the hash, not base64. For the latter you can use qcrypto_hash_base64
Regards,
Daniel
Signature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
Re: [Qemu-devel] [PATCH v22 25/30] qmp: add x-debug-block-dirty-bitmap-sha256
Quoted 30 lines · Daniel P. Berrange
Quoted 29 lines · Markus Armbruster
Quoted 10 lines · Vladimir Sementsov-Ogievskiy
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>--- block/dirty-bitmap.c | 5 +++++ blockdev.c | 25 +++++++++++++++++++++++++ include/block/dirty-bitmap.h | 1 + include/qemu/hbitmap.h | 8 ++++++++ qapi/block-core.json | 27 +++++++++++++++++++++++++++ tests/Makefile.include | 2 +- util/hbitmap.c | 11 +++++++++++ 7 files changed, 78 insertions(+), 1 deletion(-)[...]
qapi/block-core.json line 1651
Quoted 13 lines
qapi/block-core.json +5
-0@@ -1644,6 +1644,33 @@ 'data': 'BlockDirtyBitmap' } ## +# @BlockDirtyBitmapSha256: +# +# SHA256 hash of dirty bitmap data +# +# @sha256: ASCII representation of SHA256 bitmap hashSpell it SHA-256, please. The member name @sha256 can stay.
Quoted 6 lines
Quoted 2 lines
SHA-256 is 256 binary bits. Please specify how they are represented in ASCII. It better be base64 (RFC 4648), because we use that elsewhere.
It is filled later in this patch using
qcrypto_hash_digest, so it is just a hex string representing the hash, not base64. For the latter you can useqcrypto_hash_base64
I got two points:
1. Whatever encoding we use, it needs to be documented.
2. The fewer binary -> ASCII encodings we use, the better. We already use base64.
Re: [Qemu-devel] [PATCH v22 25/30] qmp: add x-debug-block-dirty-bitmap-sha256
Quoted 36 lines · Markus Armbruster
Quoted 29 lines · Daniel P. Berrange
Quoted 28 lines · Markus Armbruster
Quoted 10 lines · Vladimir Sementsov-Ogievskiy
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>--- block/dirty-bitmap.c | 5 +++++ blockdev.c | 25 +++++++++++++++++++++++++ include/block/dirty-bitmap.h | 1 + include/qemu/hbitmap.h | 8 ++++++++ qapi/block-core.json | 27 +++++++++++++++++++++++++++ tests/Makefile.include | 2 +- util/hbitmap.c | 11 +++++++++++ 7 files changed, 78 insertions(+), 1 deletion(-)[...]
qapi/block-core.json line 1651
Quoted 13 lines
qapi/block-core.json +5
-0@@ -1644,6 +1644,33 @@ 'data': 'BlockDirtyBitmap' } ## +# @BlockDirtyBitmapSha256: +# +# SHA256 hash of dirty bitmap data +# +# @sha256: ASCII representation of SHA256 bitmap hashSpell it SHA-256, please. The member name @sha256 can stay.
Quoted 5 lines
Quoted 2 lines
SHA-256 is 256 binary bits. Please specify how they are represented in ASCII. It better be base64 (RFC 4648), because we use that elsewhere.
It is filled later in this patch using
qcrypto_hash_digest, so it is just a hex string representing the hash, not base64. For the latter you can useqcrypto_hash_base64
Quoted 6 lines
I got two points:
1. Whatever encoding we use, it needs to be documented.
2. The fewer binary -> ASCII encodings we use, the better. We already use base64.
ASCII format for check sum is more common as it is more readable. It is used in the internet to check downloads, it is used by standard utility sha256sum. So, it may be better for the monitor.
However, if it is needed, I can make a follow-up patch, it is very easy, just s/qcrypto_hash_digest/qcrypto_hash_base64/ in util/hbitmap.c. iotest 165 - the only user of the feature - doesn't need any changes.
Signature · 2 lines
Best regards, Vladimir
Re: [Qemu-devel] [PATCH v22 25/30] qmp: add x-debug-block-dirty-bitmap-sha256
Quoted 44 lines · Vladimir Sementsov-Ogievskiy
Quoted 35 lines · Markus Armbruster
Quoted 28 lines · Daniel P. Berrange
Quoted 27 lines · Markus Armbruster
Quoted 10 lines · Vladimir Sementsov-Ogievskiy
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>--- block/dirty-bitmap.c | 5 +++++ blockdev.c | 25 +++++++++++++++++++++++++ include/block/dirty-bitmap.h | 1 + include/qemu/hbitmap.h | 8 ++++++++ qapi/block-core.json | 27 +++++++++++++++++++++++++++ tests/Makefile.include | 2 +- util/hbitmap.c | 11 +++++++++++ 7 files changed, 78 insertions(+), 1 deletion(-)[...]
qapi/block-core.json line 1650
Quoted 12 lines
qapi/block-core.json +5
-0@@ -1644,6 +1644,33 @@ 'data': 'BlockDirtyBitmap' } ## +# @BlockDirtyBitmapSha256: +# +# SHA256 hash of dirty bitmap data +# +# @sha256: ASCII representation of SHA256 bitmap hashSpell it SHA-256, please. The member name @sha256 can stay.
Quoted 5 lines
Quoted 2 lines
SHA-256 is 256 binary bits. Please specify how they are represented in ASCII. It better be base64 (RFC 4648), because we use that elsewhere.
It is filled later in this patch using
qcrypto_hash_digest, so it is just a hex string representing the hash, not base64. For the latter you can useqcrypto_hash_base64Quoted 6 lines
I got two points:
1. Whatever encoding we use, it needs to be documented.
2. The fewer binary -> ASCII encodings we use, the better. We already use base64.
Quoted 8 lines
ASCII format for check sum is more common as it is more readable. It is used in the internet to check downloads, it is used by standard utility sha256sum. So, it may be better for the monitor.
However, if it is needed, I can make a follow-up patch, it is very easy, just s/qcrypto_hash_digest/qcrypto_hash_base64/ in
util/hbitmap.c. iotest 165 - the only user of the feature - doesn't need any changes.
If the is a standard way to represent SHA-256 in ASCII, use it.
Whatever you use, document it clearly in the QAPI schema.
Re: [Qemu-devel] [PATCH v22 25/30] qmp: add x-debug-block-dirty-bitmap-sha256
Quoted 50 lines · Markus Armbruster
Quoted 48 lines · Vladimir Sementsov-Ogievskiy
Quoted 35 lines · Markus Armbruster
Quoted 28 lines · Daniel P. Berrange
Quoted 27 lines · Markus Armbruster
Quoted 10 lines · Vladimir Sementsov-Ogievskiy
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>--- block/dirty-bitmap.c | 5 +++++ blockdev.c | 25 +++++++++++++++++++++++++ include/block/dirty-bitmap.h | 1 + include/qemu/hbitmap.h | 8 ++++++++ qapi/block-core.json | 27 +++++++++++++++++++++++++++ tests/Makefile.include | 2 +- util/hbitmap.c | 11 +++++++++++ 7 files changed, 78 insertions(+), 1 deletion(-)[...]
qapi/block-core.json line 1650
Quoted 12 lines
qapi/block-core.json +5
-0@@ -1644,6 +1644,33 @@ 'data': 'BlockDirtyBitmap' } ## +# @BlockDirtyBitmapSha256: +# +# SHA256 hash of dirty bitmap data +# +# @sha256: ASCII representation of SHA256 bitmap hashSpell it SHA-256, please. The member name @sha256 can stay.
Quoted 5 lines
Quoted 2 lines
SHA-256 is 256 binary bits. Please specify how they are represented in ASCII. It better be base64 (RFC 4648), because we use that elsewhere.
It is filled later in this patch using
qcrypto_hash_digest, so it is just a hex string representing the hash, not base64. For the latter you can useqcrypto_hash_base64Quoted 6 lines
I got two points:
1. Whatever encoding we use, it needs to be documented.
2. The fewer binary -> ASCII encodings we use, the better. We already use base64.
ASCII format for check sum is more common as it is more readable. It is used in the internet to check downloads, it is used by standard utility sha256sum. So, it may be better for the monitor.
Quoted 9 lines
Quoted 4 lines
However, if it is needed, I can make a follow-up patch, it is very easy, just s/qcrypto_hash_digest/qcrypto_hash_base64/ in
util/hbitmap.c. iotest 165 - the only user of the feature - doesn't need any changes.If the is a standard way to represent SHA-256 in ASCII, use it.
Whatever you use, document it clearly in the QAPI schema.
I... should we, though? It's a debug interface for testing only, basically. Couldn't think of a better way to test it, and people demanded tests.
How's this for documentation:
"The hash will be in an arbitrary format that changes every time you look away from this specification. Any similarity, real or imagined, to a canonical SHA-256 ASCII string is purely coincidental."
Basically, I would actually rather go out of my way to obfuscate this command, not document it...
Maybe that's wrong-headed of me, but I still maintain that it's not terribly important because I'd rather people never, ever try to use this in production.
Re: [Qemu-devel] [PATCH v22 25/30] qmp: add x-debug-block-dirty-bitmap-sha256
Quoted 71 lines · John Snow
Quoted 50 lines · Markus Armbruster
Quoted 48 lines · Vladimir Sementsov-Ogievskiy
Quoted 35 lines · Markus Armbruster
Quoted 28 lines · Daniel P. Berrange
Quoted 27 lines · Markus Armbruster
Quoted 10 lines · Vladimir Sementsov-Ogievskiy
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>--- block/dirty-bitmap.c | 5 +++++ blockdev.c | 25 +++++++++++++++++++++++++ include/block/dirty-bitmap.h | 1 + include/qemu/hbitmap.h | 8 ++++++++ qapi/block-core.json | 27 +++++++++++++++++++++++++++ tests/Makefile.include | 2 +- util/hbitmap.c | 11 +++++++++++ 7 files changed, 78 insertions(+), 1 deletion(-)[...]
qapi/block-core.json line 1650
Quoted 12 lines
qapi/block-core.json +5
-0@@ -1644,6 +1644,33 @@ 'data': 'BlockDirtyBitmap' } ## +# @BlockDirtyBitmapSha256: +# +# SHA256 hash of dirty bitmap data +# +# @sha256: ASCII representation of SHA256 bitmap hashSpell it SHA-256, please. The member name @sha256 can stay.
Quoted 5 lines
Quoted 2 lines
SHA-256 is 256 binary bits. Please specify how they are represented in ASCII. It better be base64 (RFC 4648), because we use that elsewhere.
It is filled later in this patch using
qcrypto_hash_digest, so it is just a hex string representing the hash, not base64. For the latter you can useqcrypto_hash_base64Quoted 6 lines
I got two points:
1. Whatever encoding we use, it needs to be documented.
2. The fewer binary -> ASCII encodings we use, the better. We already use base64.
ASCII format for check sum is more common as it is more readable. It is used in the internet to check downloads, it is used by standard utility sha256sum. So, it may be better for the monitor.
Quoted 9 lines
Quoted 4 lines
However, if it is needed, I can make a follow-up patch, it is very easy, just s/qcrypto_hash_digest/qcrypto_hash_base64/ in
util/hbitmap.c. iotest 165 - the only user of the feature - doesn't need any changes.If the is a standard way to represent SHA-256 in ASCII, use it.
Whatever you use, document it clearly in the QAPI schema.
I... should we, though? It's a debug interface for testing only, basically. Couldn't think of a better way to test it, and people demanded tests.
How's this for documentation:
"The hash will be in an arbitrary format that changes every time you look away from this specification. Any similarity, real or imagined, to a canonical SHA-256 ASCII string is purely coincidental."
Quoted 6 lines
Basically, I would actually rather go out of my way to obfuscate this command, not document it...
Maybe that's wrong-headed of me, but I still maintain that it's not terribly important because I'd rather people never, ever try to use this in production.
It is wrong-headed of you :)
We mark stuff not covered by QMP's compatibility promise with x- prefixes, not with incomplete documentation, and not by obsfuscating it. People trying to use this in non-production need to know how it works, too.
If you think people need to be scared away some more, feel free to write something like "currently SHA-256 ASCII". If that's not enough of a hint for you, you might add "but you shouldn't rely on that".
Re: [Qemu-devel] [PATCH v22 25/30] qmp: add x-debug-block-dirty-bitmap-sha256
Quoted 73 lines · Markus Armbruster
Quoted 71 lines · John Snow
Quoted 50 lines · Markus Armbruster
Quoted 48 lines · Vladimir Sementsov-Ogievskiy
Quoted 35 lines · Markus Armbruster
Quoted 28 lines · Daniel P. Berrange
Quoted 27 lines · Markus Armbruster
Quoted 10 lines · Vladimir Sementsov-Ogievskiy
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>--- block/dirty-bitmap.c | 5 +++++ blockdev.c | 25 +++++++++++++++++++++++++ include/block/dirty-bitmap.h | 1 + include/qemu/hbitmap.h | 8 ++++++++ qapi/block-core.json | 27 +++++++++++++++++++++++++++ tests/Makefile.include | 2 +- util/hbitmap.c | 11 +++++++++++ 7 files changed, 78 insertions(+), 1 deletion(-)[...]
qapi/block-core.json line 1650
Quoted 12 lines
qapi/block-core.json +5
-0@@ -1644,6 +1644,33 @@ 'data': 'BlockDirtyBitmap' } ## +# @BlockDirtyBitmapSha256: +# +# SHA256 hash of dirty bitmap data +# +# @sha256: ASCII representation of SHA256 bitmap hashSpell it SHA-256, please. The member name @sha256 can stay.
Quoted 5 lines
Quoted 2 lines
SHA-256 is 256 binary bits. Please specify how they are represented in ASCII. It better be base64 (RFC 4648), because we use that elsewhere.
It is filled later in this patch using
qcrypto_hash_digest, so it is just a hex string representing the hash, not base64. For the latter you can useqcrypto_hash_base64Quoted 6 lines
I got two points:
1. Whatever encoding we use, it needs to be documented.
2. The fewer binary -> ASCII encodings we use, the better. We already use base64.
ASCII format for check sum is more common as it is more readable. It is used in the internet to check downloads, it is used by standard utility sha256sum. So, it may be better for the monitor.
Quoted 9 lines
Quoted 4 lines
However, if it is needed, I can make a follow-up patch, it is very easy, just s/qcrypto_hash_digest/qcrypto_hash_base64/ in
util/hbitmap.c. iotest 165 - the only user of the feature - doesn't need any changes.If the is a standard way to represent SHA-256 in ASCII, use it.
Whatever you use, document it clearly in the QAPI schema.
I... should we, though? It's a debug interface for testing only, basically. Couldn't think of a better way to test it, and people demanded tests.
How's this for documentation:
"The hash will be in an arbitrary format that changes every time you look away from this specification. Any similarity, real or imagined, to a canonical SHA-256 ASCII string is purely coincidental."
Quoted 9 lines
Quoted 6 lines
Basically, I would actually rather go out of my way to obfuscate this command, not document it...
Maybe that's wrong-headed of me, but I still maintain that it's not terribly important because I'd rather people never, ever try to use this in production.
It is wrong-headed of you :)
I figured so, but I wanted to be very facetiously clear that we didn't really intend for this to be used outside of our own testing suite. I'd love to not even have it be a real QMP interface available in non-testing binaries at all, but that gets weird quickly.
Sorry for being a wise-ass.
Quoted 9 lines
We mark stuff not covered by QMP's compatibility promise with x- prefixes, not with incomplete documentation, and not by obsfuscating it. People trying to use this in non-production need to know how it works, too.
If you think people need to be scared away some more, feel free to write something like "currently SHA-256 ASCII". If that's not enough of a hint for you, you might add "but you shouldn't rely on that".
I'd still like to decorate it with some little emoji skulls, too. ☠☠💀💀 Make sure people know it's a very bad idea to use it. Maybe QEMU can do some wall text whenever it gets used that says "Someone is doing something very naughty when we asked very nicely for them not to."
Kidding again, of course, but just trying to emphasize that this command literally serves no purpose other than for testing and is not expected to be used apart from that.
The format could be changed if you desired, but as this has been on the list for over a year, I'd have to unfortunately request that you make that change if it was important to you.
--js
Re: [Qemu-devel] [Qemu-block] [PATCH v22 25/30] qmp: add x-debug-block-dirty-bitmap-sha256
In reply to Daniel P. BerrangeQuoted 30 lines · Daniel P. Berrange
Quoted 29 lines · Markus Armbruster
Quoted 10 lines · Vladimir Sementsov-Ogievskiy
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>--- block/dirty-bitmap.c | 5 +++++ blockdev.c | 25 +++++++++++++++++++++++++ include/block/dirty-bitmap.h | 1 + include/qemu/hbitmap.h | 8 ++++++++ qapi/block-core.json | 27 +++++++++++++++++++++++++++ tests/Makefile.include | 2 +- util/hbitmap.c | 11 +++++++++++ 7 files changed, 78 insertions(+), 1 deletion(-)[...]
qapi/block-core.json line 1651
Quoted 13 lines
qapi/block-core.json +5
-0@@ -1644,6 +1644,33 @@ 'data': 'BlockDirtyBitmap' } ## +# @BlockDirtyBitmapSha256: +# +# SHA256 hash of dirty bitmap data +# +# @sha256: ASCII representation of SHA256 bitmap hashSpell it SHA-256, please. The member name @sha256 can stay.
Quoted 6 lines
Quoted 2 lines
SHA-256 is 256 binary bits. Please specify how they are represented in ASCII. It better be base64 (RFC 4648), because we use that elsewhere.
It is filled later in this patch using
qcrypto_hash_digest, so it is just a hex string representing the hash, not base64. For the latter you can useqcrypto_hash_base64
hex (aka base16) is used for displaying UUIDs across QMP. Furthermore, this is an x- command, useful only for debugging, so we can feel free to change the underlying implementation without breaking clients (if we prefer to go with the shorter base64 string instead of the longer hex string, even if we make that switch after a release). Still, it's nice to avoid churn, and Markus does have a point about reusing a common base64 theme for displaying arbitrary binary data, rather than having ad hoc differences.
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
PATCH v22 26/30 iotests: test qcow2 persistent dirty bitmap
In reply to Vladimir Sementsov-Ogievskiy--- tests/qemu-iotests/165 | 105 +++++++++++++++++++++++++++++++++++++++++++++ tests/qemu-iotests/165.out | 5 +++ tests/qemu-iotests/group | 1 + 3 files changed, 111 insertions(+) create mode 100755 tests/qemu-iotests/165 create mode 100644 tests/qemu-iotests/165.out
tests/qemu-iotests/165 new file +105 -0
@@ -0,0 +1,105 @@ +#!/usr/bin/env python +# +# Tests for persistent dirty bitmaps. +# +# Copyright: Vladimir Sementsov-Ogievskiy 2015-2017 +# +# This program is free software; you can redistribute it and/or modify +# it under the terms of the GNU General Public License as published by +# the Free Software Foundation; either version 2 of the License, or +# (at your option) any later version. +# +# This program is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +# GNU General Public License for more details. +# +# You should have received a copy of the GNU General Public License +# along with this program. If not, see <http://www.gnu.org/licenses/>. +# + +import os +import re +import iotests +from iotests import qemu_img + +disk = os.path.join(iotests.test_dir, 'disk') +disk_size = 0x40000000 # 1G + +# regions for qemu_io: (start, count) in bytes +regions1 = ((0, 0x100000), + (0x200000, 0x100000)) + +regions2 = ((0x10000000, 0x20000), + (0x3fff0000, 0x10000)) + +class TestPersistentDirtyBitmap(iotests.QMPTestCase): + + def setUp(self): + qemu_img('create', '-f', iotests.imgfmt, disk, str(disk_size)) + + def tearDown(self): + os.remove(disk) + + def mkVm(self): + return iotests.VM().add_drive(disk) + + def mkVmRo(self): + return iotests.VM().add_drive(disk, opts='readonly=on') + + def getSha256(self): + result = self.vm.qmp('x-debug-block-dirty-bitmap-sha256', + node='drive0', name='bitmap0') + return result['return']['sha256'] + + def checkBitmap(self, sha256): + result = self.vm.qmp('x-debug-block-dirty-bitmap-sha256', + node='drive0', name='bitmap0') + self.assert_qmp(result, 'return/sha256', sha256); + + def writeRegions(self, regions): + for r in regions: + self.vm.hmp_qemu_io('drive0', + 'write %d %d' % r) + + def qmpAddBitmap(self): + self.vm.qmp('block-dirty-bitmap-add', node='drive0', + name='bitmap0', persistent=True, autoload=True) + + def test_persistent(self): + self.vm = self.mkVm() + self.vm.launch() + self.qmpAddBitmap() + + self.writeRegions(regions1) + sha256 = self.getSha256() + + self.vm.shutdown() + + self.vm = self.mkVmRo() + self.vm.launch() + self.vm.shutdown() + + #catch 'Persistent bitmaps are lost' possible error + log = self.vm.get_log() + log = re.sub(r'^\[I \d+\.\d+\] OPENED\n', '', log) + log = re.sub(r'\[I \+\d+\.\d+\] CLOSED\n?$', '', log) + if log: + print log + + self.vm = self.mkVm() + self.vm.launch() + + self.checkBitmap(sha256) + self.writeRegions(regions2) + sha256 = self.getSha256() + + self.vm.shutdown() + self.vm.launch() + + self.checkBitmap(sha256) + + self.vm.shutdown() + +if __name__ == '__main__': + iotests.main(supported_fmts=['qcow2'])
tests/qemu-iotests/165.out new file +5 -0
@@ -0,0 +1,5 @@ +. +---------------------------------------------------------------------- +Ran 1 tests + +OK
tests/qemu-iotests/group +1 -0
@@ -163,6 +163,7 @@ 159 rw auto quick 160 rw auto quick 162 auto quick +165 rw auto quick 170 rw auto quick 171 rw auto quick 172 auto
Signature · 1 line
2.11.1
PATCH v22 27/30 block/dirty-bitmap: add bdrv_remove_persistent_dirty_bitmap
In reply to Vladimir Sementsov-OgievskiyInterface for removing persistent bitmap from its storage.
--- block/dirty-bitmap.c | 18 ++++++++++++++++++ include/block/block_int.h | 3 +++ include/block/dirty-bitmap.h | 3 +++ 3 files changed, 24 insertions(+)
block/dirty-bitmap.c +18 -0
@@ -395,6 +395,7 @@ void bdrv_release_dirty_bitmap(BlockDriverState *bs, BdrvDirtyBitmap *bitmap)/** * Release all named dirty bitmaps attached to a BDS (for use in bdrv_close()). * There must not be any frozen bitmaps attached. +* This function does not remove persistent bitmaps from the storage. * Called with BQL taken. */ void bdrv_release_named_dirty_bitmaps (BlockDriverState *bs)
@@ -402,6 +403,23 @@ void bdrv_release_named_dirty_bitmaps(BlockDriverState *bs)bdrv_do_release_matching_dirty_bitmap (bs,NULL , true); } +/** +* Remove persistent dirty bitmap from the storage if it exists. +* Absence of bitmap is not an error, because we have the following scenario: +* BdrvDirtyBitmap can have .persistent = true but not yet saved and have no +* stored version. For such bitmap bdrv_remove_persistent_dirty_bitmap() should +* not fail. +* This function doesn't release corresponding BdrvDirtyBitmap. +*/ +void bdrv_remove_persistent_dirty_bitmap (BlockDriverState *bs, +const char *name, +Error **errp) +{ +if (bs->drv && bs->drv->bdrv_remove_persistent_dirty_bitmap) { + bs->drv->bdrv_remove_persistent_dirty_bitmap (bs, name, errp); + } +} +/* Called with BQL taken. */ void bdrv_disable_dirty_bitmap (BdrvDirtyBitmap *bitmap) {
include/block/block_int.h +3 -0
@@ -391,6 +391,9 @@ struct BlockDriver {const char *name, uint32_t granularity, Error **errp); +void (*bdrv_remove_persistent_dirty_bitmap)(BlockDriverState *bs, +const char *name, +Error **errp);QLIST_ENTRY (BlockDriver) list; };
include/block/dirty-bitmap.h +3 -0
@@ -25,6 +25,9 @@ BdrvDirtyBitmap *bdrv_find_dirty_bitmap(BlockDriverState *bs,void bdrv_dirty_bitmap_make_anon (BdrvDirtyBitmap *bitmap);void bdrv_release_dirty_bitmap (BlockDriverState *bs,BdrvDirtyBitmap *bitmap);void bdrv_release_named_dirty_bitmaps (BlockDriverState *bs); +void bdrv_remove_persistent_dirty_bitmap (BlockDriverState *bs, +const char *name, +Error **errp);void bdrv_disable_dirty_bitmap (BdrvDirtyBitmap *bitmap);void bdrv_enable_dirty_bitmap (BdrvDirtyBitmap *bitmap);BlockDirtyInfoList *bdrv_query_dirty_bitmaps (BlockDriverState *bs);
Signature · 1 line
2.11.1
PATCH v22 28/30 qcow2: add .bdrv_remove_persistent_dirty_bitmap
In reply to Vladimir Sementsov-OgievskiyRealize .bdrv_remove_persistent_dirty_bitmap interface.
--- block/qcow2-bitmap.c | 41 +++++++++++++++++++++++++++++++++++++++++ block/qcow2.c | 1 + block/qcow2.h | 3 +++ 3 files changed, 45 insertions(+)
block/qcow2-bitmap.c +41 -0
@@ -1236,6 +1236,47 @@ static Qcow2Bitmap *find_bitmap_by_name(Qcow2BitmapList *bm_list,return NULL ; } +void qcow2_remove_persistent_dirty_bitmap (BlockDriverState *bs, +const char *name, +Error **errp) +{ +int ret; +BDRVQcow2State *s = bs->opaque; +Qcow2Bitmap *bm; +Qcow2BitmapList *bm_list; + +if (s->nb_bitmaps ==0 ) { +/* Absence of the bitmap is not an error: see explanation above +* bdrv_remove_persistent_dirty_bitmap() definition. */ +return ; + } + + bm_list =bitmap_list_load (bs, s->bitmap_directory_offset, + s->bitmap_directory_size, errp); +if (bm_list ==NULL ) { +return ; + } + + bm =find_bitmap_by_name (bm_list, name); +if (bm ==NULL ) { + goto fail; + } + +QSIMPLEQ_REMOVE (bm_list, bm, Qcow2Bitmap, entry); + + ret =update_ext_header_and_dir (bs, bm_list); +if (ret <0 ) { +error_setg_errno (errp, -ret,"Failed to update bitmap extension" ); + goto fail; + } + +free_bitmap_clusters (bs, &bm->table); + +fail: +bitmap_free (bm); +bitmap_list_free (bm_list); +} +void qcow2_store_persistent_dirty_bitmaps (BlockDriverState *bs,Error **errp) {BdrvDirtyBitmap *bitmap;
block/qcow2.c +1 -0
@@ -3612,6 +3612,7 @@ BlockDriver bdrv_qcow2 = { .bdrv_reopen_bitmaps_rw = qcow2_reopen_bitmaps_rw, .bdrv_can_store_new_dirty_bitmap = qcow2_can_store_new_dirty_bitmap, + .bdrv_remove_persistent_dirty_bitmap = qcow2_remove_persistent_dirty_bitmap, };static void bdrv_qcow2_init (void )
block/qcow2.h +3 -0
@@ -637,5 +637,8 @@ bool qcow2_can_store_new_dirty_bitmap(BlockDriverState *bs,const char *name, uint32_t granularity, Error **errp); +void qcow2_remove_persistent_dirty_bitmap (BlockDriverState *bs, +const char *name, +Error **errp);#endif
Signature · 1 line
2.11.1
PATCH v22 29/30 qmp: block-dirty-bitmap-remove: remove persistent
In reply to Vladimir Sementsov-OgievskiyRemove persistent bitmap from the storage on block-dirty-bitmap-remove.
--- blockdev.c | 10 ++++++++++ qapi/block-core.json | 3 ++- 2 files changed, 12 insertions(+), 1 deletion(-)
blockdev.c +10 -0
@@ -2783,6 +2783,7 @@ void qmp_block_dirty_bitmap_remove(const char *node, const char *name, {BlockDriverState *bs;BdrvDirtyBitmap *bitmap; +Error *local_err =NULL ; bitmap =block_dirty_bitmap_lookup (node, name, &bs, errp);if (!bitmap || !bs) {
@@ -2795,6 +2796,15 @@ void qmp_block_dirty_bitmap_remove(const char *node, const char *name, name);return ; } + +if (bdrv_dirty_bitmap_get_persistance (bitmap)) { +bdrv_remove_persistent_dirty_bitmap (bs, name, &local_err); +if (local_err !=NULL ) { +error_propagate (errp, local_err); +return ; + } + } +bdrv_dirty_bitmap_make_anon (bitmap);bdrv_release_dirty_bitmap (bs, bitmap); }
qapi/block-core.json +2 -1
@@ -1601,7 +1601,8 @@ # @block-dirty-bitmap-remove: # # Stop write tracking and remove the dirty bitmap that was created -# with block-dirty-bitmap-add. +# with block-dirty-bitmap-add. If the bitmap is persistent, remove it from its +# storage too. # # Returns: nothing on success # If @node is not a valid block device or node, DeviceNotFound
Signature · 1 line
2.11.1
PATCH v22 30/30 block: release persistent bitmaps on inactivate
In reply to Vladimir Sementsov-OgievskiyWe should release them here to reload on invalidate cache.
--- block.c | 4 ++++ block/dirty-bitmap.c | 29 +++++++++++++++++++++++------ include/block/dirty-bitmap.h | 1 + 3 files changed, 28 insertions(+), 6 deletions(-)
block.c +4 -0
@@ -4156,6 +4156,10 @@ static int bdrv_inactivate_recurse(BlockDriverState *bs, } } +/* At this point persistent bitmaps should be already stored by the format +* driver */ +bdrv_release_persistent_dirty_bitmaps (bs); +return 0 ; }
block/dirty-bitmap.c +23 -6
@@ -356,15 +356,20 @@ void bdrv_dirty_bitmap_truncate(BlockDriverState *bs)bdrv_dirty_bitmaps_unlock (bs); } +static bool bdrv_dirty_bitmap_has_name (BdrvDirtyBitmap *bitmap) +{ +return !!bdrv_dirty_bitmap_name (bitmap); +} +/* Called with BQL taken. */ -static void bdrv_do_release_matching_dirty_bitmap (BlockDriverState *bs, -BdrvDirtyBitmap *bitmap, -bool only_named) +static void bdrv_do_release_matching_dirty_bitmap ( +BlockDriverState *bs,BdrvDirtyBitmap *bitmap, +bool (*cond)(BdrvDirtyBitmap *bitmap)) {BdrvDirtyBitmap *bm, *next;bdrv_dirty_bitmaps_lock (bs); QLIST_FOREACH_SAFE(bm, &bs->dirty_bitmaps, list, next) { -if ((!bitmap || bm == bitmap) && (!only_named || bm->name)) { +if ((!bitmap || bm == bitmap) && (!cond ||cond (bm))) {assert (!bm->active_iterators);assert (!bdrv_dirty_bitmap_frozen (bm));assert (!bm->meta);
@@ -389,7 +394,7 @@ out:/* Called with BQL taken. */ void bdrv_release_dirty_bitmap (BlockDriverState *bs,BdrvDirtyBitmap *bitmap) { -bdrv_do_release_matching_dirty_bitmap (bs, bitmap, false); +bdrv_do_release_matching_dirty_bitmap (bs, bitmap,NULL ); }/**
@@ -400,7 +405,19 @@ void bdrv_release_dirty_bitmap(BlockDriverState *bs, BdrvDirtyBitmap *bitmap)*/ void bdrv_release_named_dirty_bitmaps (BlockDriverState *bs) { -bdrv_do_release_matching_dirty_bitmap (bs,NULL , true); +bdrv_do_release_matching_dirty_bitmap (bs,NULL , bdrv_dirty_bitmap_has_name); +} + +/** +* Release all persistent dirty bitmaps attached to a BDS (for use in +* bdrv_inactivate_recurse()). +* There must not be any frozen bitmaps attached. +* This function does not remove persistent bitmaps from the storage. +*/ +void bdrv_release_persistent_dirty_bitmaps (BlockDriverState *bs) +{ +bdrv_do_release_matching_dirty_bitmap (bs,NULL , + bdrv_dirty_bitmap_get_persistance); }/**
include/block/dirty-bitmap.h +1 -0
@@ -25,6 +25,7 @@ BdrvDirtyBitmap *bdrv_find_dirty_bitmap(BlockDriverState *bs,void bdrv_dirty_bitmap_make_anon (BdrvDirtyBitmap *bitmap);void bdrv_release_dirty_bitmap (BlockDriverState *bs,BdrvDirtyBitmap *bitmap);void bdrv_release_named_dirty_bitmaps (BlockDriverState *bs); +void bdrv_release_persistent_dirty_bitmaps (BlockDriverState *bs);void bdrv_remove_persistent_dirty_bitmap (BlockDriverState *bs,const char *name,Error **errp);
Signature · 1 line
2.11.1
Quoted 22 lines · Vladimir Sementsov-Ogievskiy
Rebase on master, so changes, mostly related to new dirty bitmaps mutex:
10: - asserts now in bdrv_{re,}set_dirty_bitmap_locked functions.
- also add assert intobdrv_undo_clear_dirty_bitmap(the only change, not related to rebase)
- add mutex lock intobdrv_dirty_bitmap_set_readonly(as it changes bitmap list, so the lock should be taken)
- return instead of go-to inqmp_block_dirty_bitmap_clear
- indirty-bitmaps.h, movebdrv_dirty_bitmap_set_readonlybefore block "Functions that require manual locking", move
bdrv_dirty_bitmap_readonlyandbdrv_has_readonly_bitmapsinto this block
15: - add mutex lock/unlock intobdrv_dirty_bitmap_set_autoload
- indirty-bitmaps.h, movebdrv_dirty_bitmap_set_autoloadbefore block "Functions that require manual locking", movebdrv_dirty_bitmap_get_autoloadinto this block
17: - add mutex lock/unlock intobdrv_dirty_bitmap_set_persistance
- indirty-bitmaps.h, movebdrv_dirty_bitmap_set_persistancebefore block "Functions that require manual locking", movebdrv_dirty_bitmap_get_persistanceand
bdrv_has_changed_persistent_bitmapsinto this block
18: indirty-bitmaps.h, movebdrv_dirty_bitmap_nextinto block
"Functions that require manual locking". (do not remove r-b, as it is
just one empty line removed before function declaration)
Quoted 8 lines
23: return instead of go-to in
qmp_block_dirty_bitmap_add24: return instead of go-to inqmp_block_dirty_bitmap_add25: - return instead of go-to
- remove aio_context_acquire/release calls
- noaio_contextparameter forblock_dirty_bitmap_lookup
- indirty-bitmaps.h, movebdrv_dirty_bitmap_sha256into block "Functions that require manual locking".
29: - return instead of go-to inqmp_block_dirty_bitmap_remove
All looks good, thanks. I'll rebase my own fixes on top of these patches, no need to have you respin them.
Paolo
Quoted 34 lines · Paolo Bonzini
Quoted 22 lines · Vladimir Sementsov-Ogievskiy
Rebase on master, so changes, mostly related to new dirty bitmaps mutex:
10: - asserts now in bdrv_{re,}set_dirty_bitmap_locked functions.
- also add assert intobdrv_undo_clear_dirty_bitmap(the only change, not related to rebase)
- add mutex lock intobdrv_dirty_bitmap_set_readonly(as it changes bitmap list, so the lock should be taken)
- return instead of go-to inqmp_block_dirty_bitmap_clear
- indirty-bitmaps.h, movebdrv_dirty_bitmap_set_readonlybefore block "Functions that require manual locking", move
bdrv_dirty_bitmap_readonlyandbdrv_has_readonly_bitmapsinto this block
15: - add mutex lock/unlock intobdrv_dirty_bitmap_set_autoload
- indirty-bitmaps.h, movebdrv_dirty_bitmap_set_autoloadbefore block "Functions that require manual locking", movebdrv_dirty_bitmap_get_autoloadinto this block
17: - add mutex lock/unlock intobdrv_dirty_bitmap_set_persistance
- indirty-bitmaps.h, movebdrv_dirty_bitmap_set_persistancebefore block "Functions that require manual locking", movebdrv_dirty_bitmap_get_persistanceand
bdrv_has_changed_persistent_bitmapsinto this block
18: indirty-bitmaps.h, movebdrv_dirty_bitmap_nextinto block
"Functions that require manual locking". (do not remove r-b, as it is
just one empty line removed before function declaration)Quoted 8 lines
23: return instead of go-to in
qmp_block_dirty_bitmap_add24: return instead of go-to inqmp_block_dirty_bitmap_add25: - return instead of go-to
- remove aio_context_acquire/release calls
- noaio_contextparameter forblock_dirty_bitmap_lookup
- indirty-bitmaps.h, movebdrv_dirty_bitmap_sha256into block "Functions that require manual locking".
29: - return instead of go-to inqmp_block_dirty_bitmap_remove
Quoted 5 lines
All looks good, thanks. I'll rebase my own fixes on top of these patches, no need to have you respin them.
Paolo
Thank you! And for thunderbird-work-around too!
Signature · 2 lines
Best regards, Vladimir
Quoted 7 lines · Vladimir Sementsov-Ogievskiy
Hi all!
There is a new update of qcow2-bitmap series - v22.
web: https://src.openvz.org/users/vsementsov/repos/qemu/browse?at=qcow2-bitmap-v22
git: https://src.openvz.org/scm/~vsementsov/qemu.git (tag qcow2-bitmap-v22)
Minor rebase conflicts, I staged a branch here:
https://github.com/qemu/qemu/compare/master...jnsnow:review-vlad-22
As for the conflicts, they are:
08: Conflict when qed-gencb was removed; trivial fix.
10: 'count' was changed to 'bytes' which disrupted patch context.
Thanks,
--js
Quoted 494 lines
v22:
Rebase on master, so changes, mostly related to new dirty bitmaps mutex:
10: - asserts now in bdrv_{re,}set_dirty_bitmap_locked functions.
- also add assert intobdrv_undo_clear_dirty_bitmap(the only change, not related to rebase)
- add mutex lock intobdrv_dirty_bitmap_set_readonly(as it changes bitmap list, so the lock should be taken)
- return instead of go-to inqmp_block_dirty_bitmap_clear
- indirty-bitmaps.h, movebdrv_dirty_bitmap_set_readonlybefore block "Functions that require manual locking", move
bdrv_dirty_bitmap_readonlyandbdrv_has_readonly_bitmapsinto this block
15: - add mutex lock/unlock intobdrv_dirty_bitmap_set_autoload
- indirty-bitmaps.h, movebdrv_dirty_bitmap_set_autoloadbefore block "Functions that require manual locking", movebdrv_dirty_bitmap_get_autoloadinto this block
17: - add mutex lock/unlock intobdrv_dirty_bitmap_set_persistance
- indirty-bitmaps.h, movebdrv_dirty_bitmap_set_persistancebefore block "Functions that require manual locking", movebdrv_dirty_bitmap_get_persistanceand
bdrv_has_changed_persistent_bitmapsinto this block
18: indirty-bitmaps.h, movebdrv_dirty_bitmap_nextinto block
"Functions that require manual locking". (do not remove r-b, as it is
just one empty line removed before function declaration)
23: return instead of go-to inqmp_block_dirty_bitmap_add
24: return instead of go-to inqmp_block_dirty_bitmap_add
25: - return instead of go-to
- remove aio_context_acquire/release calls
- noaio_contextparameter forblock_dirty_bitmap_lookup
- indirty-bitmaps.h, movebdrv_dirty_bitmap_sha256into block
"Functions that require manual locking".
29: - return instead of go-to inqmp_block_dirty_bitmap_remover-b's are dropped from 10,15,17,25.
v21:
09,10: improve comment, add r-b's by Max and John
10: improve comment,k
11,12: add r-b by John
13: prependlocal_errwith additional info (Max), add r-b by John 14: add r-b's by Max and John
20,30: add r-b by Maxv20:
handle reopening images ro and rw.
On reopening ro: store bitmaps (storing sets 'IN_USE'=0 in the image) and mark them readonly (set readonly flag in BlockDirtyBitmap)
After reopening rw: mark bitmaps
IN_USEin the image and unset readonly flag in BlockDirtyBitmap09: new
10: improve comment
add parameter 'value' tobdrv_dirty_bitmap_set_readonly
11: use new parameter ofbdrv_dirty_bitmap_set_readonly12-14, 20: newv19:
rebased on master
05: move 'sign-off' over 'reviewed-by's
08:error_report->error_setginqcow2_truncate(because of rebase) 09: return EPERM inbdrv_aligned_pwritevandbdrv_co_pdiscardif there
are readonly bitmaps. EPERM is chosen because it is already used for readonly image inbdrv_co_pdiscard.
Also handle readonly bitmap inblock_dirty_bitmap_clear_prepareandqmp_block_dirty_bitmap_clear
Max's r-b is not added
10: fix grammar in comment
add Max's r-b
12, 13, 15, 21: add Max's r-b
24: fix grammar in comment
25: fix grammar and wording in comment
also, I see contextual changes in inactiavate mechanism. Hope, they do not affect these series.v18:
rebased on master (sorry for v17)
08: contextual:
qcow2_do_openis changed instead ofqcow2_open
rename s/need_update_header/update_header/ inqcow2_do_open, to not do it in 10
save r-b's by Max and John
09: new patch
10:load_bitmap_data: do not clear bitmap parameter - it should be already cleared
(it actually created before singleload_bitmap_data()call)
if some bitmaps are loaded, but we can't write the image (it is readonly
or inactive), so we can't mark bitmaps "in use" in the image, mark
corresponding BdrvDirtyBitmap read-only.
changeerror_setgtoerror_setg_errnofor "Can't update bitmap directory"
no needs to rename s/need_update_header/update_header/ here, as it done in 08
13: functionbdrv_has_persistent_bitmapsbecomesbdrv_has_changed_persistent_bitmaps,
to handle readonly field.
14: declaration moved to the bottom of .h, save r-b's
15: firstly checkbdrv_has_changed_persistent_bitmapsand only then fail on !can_write, and thenQSIMPLEQ_INIT(&drop_tables)
skip readonly bitmaps in saving loop
18: remove '#optional', 2.9 -> 2.10, save r-b's
19: remove '#optional', 2.9 -> 2.10, save r-b's
20: 2.9 -> 2.10, save r-b's
21: add check of read-only image open, drop r-b's
24: add comment toqapi/block-core.json, that block-dirty-bitmap-add removes bitmap
from storage. r-b's by Max and John savedv17:
08: add r-b's by Max and John
09: clear unknown autoclear features from BDRVQcow2State before calling
qcow2_load_autoloading_dirty_bitmaps(), and also do not extra update header if it is updated byqcow2_load_autoloading_dirty_bitmaps().
11: new patch, splitted out from 16
12: rewrite commit message
14: changingbdrv_closemoved to separate new patch 11s/1/1ULL/ ; s/%u/%llu/ for two errors16: s/No/Not/
add Max's r-b
24: new patchv16:
mostly by Kevin's comments:
07: just moved here, as preparation for merging refcounts to 08
08: move "qcow2-bitmap: refcounts" staff to this patch, to not break qcow2-img check
drop r-b's
move necessary supporting static functions to this patch too (to not break compilation) fprintf ->error_report
other small changes with error messages and comments
code style
forqcow2-bitmap.c:
'include "exec/log.h"' was dropped
s/1/(1U << 0) forBME_FLAG_IN_USE
addBME_TABLE_ENTRY_FLAG_ALL_ONESto replace magic 1
don't check 'cluster_size<= 0' incheck_table_entry
old "[PATCH v15 08/25] block: introduce auto-loading bitmaps" was dropped
09: was "[PATCH v15 09/25] qcow2: add .bdrv_load_autoloading_dirty_bitmaps"
drop r-b's
some staff was moved to 08
update_header_sync- error onbdrv_flushfail
renamedisk_sectors_in_bitmap_clustertosectors_covered_by_bitmap_cluster
and adjust comment.
so, variable for storing this function result: s/dsc/sbc/
s/1/BME_TABLE_ENTRY_FLAG_ALL_ONES/
also, as Qcow2BitmapTable already introduced in 08,
s/table_offset/table.offset/ and s/table_size/table.size, etc..
update_ext_header_and_dir_in_place: add comments, add additional
update_header_sync, to reduce indeterminacy in case of error.
callqcow2_load_autoloading_dirty_bitmapsdirectly fromqcow2_open
no .bdrv_load_autoloading_dirty_bitmaps
11: dropbdrv_store_persistent_dirty_bitmapsat all.
drop r-b's
13: rename patch, rewrite commit msg
drop r-b's
movebdrv_release_named_dirty_bitmapsinbdrv_close()afterdrv->bdrv_close()Qcow2BitmapTable is already introduced, so no
s/table_offset/table.offset/ and s/table_size/table.size, etc.. here
old 25 withcheck_constraints_on_bitmap()improvements merged here (Eric)
like in 09, s/dsc/sbc/ and s/disk_sectors_in_bitmap_cluster/sectors_covered_by_bitmap_cluster/ no .bdrv_store_persistent_dirty_bitmaps
callqcow2_store_persistent_dirty_bitmapsdirectly fromqcow2_inactivate
15: corresponding part of 25 merged here. Add John's r-b, as 25 was also reviewed by John. 16: add John's r-bv15:
13,14: add John's r-b
15:qcow2_can_store_new_dirty_bitmap:
- check max dirty bitmaps and bitmap directory overhead
- switch toerror_prepend
rm Max's r-b
not add John's r-b
17-24: add John's r-b
25: changed because 15 changed,
not add John's r-bv14:
07: use '|=' to update
need_update_header
add John's r-b
add Max's r-b
09: remove unusedbitmap_table_to_cpu()
left Max's r-b, hope it's ok
add John's r-b
10: remove extra new line
add John's r-b
11: add John's r-b
12: add John's r-b
13: small fixes by John's review:
- remove weirdg_freeof NULL pointer from
if (tb == NULL) {
g_free(tb);
return-EINVAL;
}
- remove extra comment "/* errp is already set */"
- s/"Too large bitmap directory"/"Bitmap directory is too large"/
left Max's r-b, hope you don't mind
22: add Max's r-b
23: add Max's r-b
24: add Max's r-b
25: new patch to improve error message oncheck_constraints_on_bitmapfailv13: Just a fix for style checker.
13: line over 80
14: line over 80
22: s/if () \n{/if () {/v12:
07: do not update header inqcow2_read_extensions, instead do it in the
end ofqcow2_open, where it is updated also to clear unknown autoclear features.
Wrongbdrv_is_root_nodeused instead ofbdrv_is_read_onlyis fixed automatically.
08: add Max's r-b
09: contextual: s/raw_bsd/raw-format/
qcow2-bitmap.c: copyright s/2016/2017/
fix compilation: moveupdate_header_sync()to this patch
add Max's r-b
13:update_header_sync()is moved to 09
add Max's r-b
15: add Max's r-b
16: As qmp-commands.txt is removed from master, remove it here too.
Sentence "For now only Qcow2 disks support persistent bitmaps.
Default is false." moved toqapi/block-core.json.
Hope that is OK, Max's r-b is not dropped.
17: qmp-commands.txt deleted, r-b is not dropped too.
18: s/is failed/has failed/, add Max's r-b
19: copyright: s/2016/2017/
21: drop "res->corruptions> 0 => ret =-EINVAL", add Max's r-b
22: actually, patch is replaced with a new one, however some parts are the same.
instead of changingbdrv_release_dirty_bitmapbehaviour, create newbdrv_remove_persistent_dirty_bitmap
23: improve comment, about not-exists is not an error
24: new patchOld 24 (qcow2-bitmap: cache bitmap list in BDRVQcow2State) will be sent separately, to be applied after these series.
Also: about
bdrv_dirty_bitmap_set_persistance: I've decided not change its behaviour for now and just callbdrv_remove_persistent_dirty_bitmapfromqmp_block_dirty_bitmap_remove. Reasons:
1. Do not change reviewed part.
2. I'm not sure, that this complication of BdrvDirtyBitmap.persistent semantics is good (current semantics are just: .persistent means that bitmap should be saved on disk close). .persistent actually is not very related to "is there stored version of the bitmap in the image".
3. It may be done later.
4. It is not actually needed for now, as the only usage is dropping bitmap inbdrv_remove_persistent_dirty_bitmap. May be in future it will be needed, when we will have qmp interfaces for finer control of persistent dirty bitmaps.v11:
Fix automatic build test fail.
18: wording - "ASCII representation of SHA256 ..."
24: fix redeclaration error. (This is still RFC, may be not needed patch,
see description in v10 below).v10:
07: rm John's r-b
not add Max's r-b
fix subject s/dirty bitmaps/bitmaps
improve WARNING message
remove header ext if autoclear bit is unset, throughqcow2_updata_header()- r-b blocking change
08: movebdrv_load_autoloading_dirty_bitmapsunder asserts block (Max)
fix typo in commit message
movebdrv_load_autoloading_dirty_bitmapsafter asserts
John's r-b
09: rewritecheck_dir_entry
remove extra feature cluster_size=0 fromcheck_table_entry
which clears autoclear bit before trying to update bitmap directorybitmap_list_store- do not free old clusters if in_place=true
changes in comments, fix indents
add functions
bitmap_list_count
update_ext_header_and_dir_in_place(unset autoclear bit before trying to
modify bitmap directory)
(for usage inqcow2_load_autoloading_dirty_bitmaps)
11: add node name to error report
Max's r-b
13: add type Qcow2BitmapTable
move from bm.table_* to bm.table.*
rewritecheck_constraints_on_bitmap
flush(bs->file)instead of flush(bs) inupdate_ext_header_and_dirassert(DIV_ROUND_UP(bm_size, dsc) == tb_size);
andassert(write_size <= s->cluster_size); instore_bitmap_data
fix: s/tb_size/tb_size *sizeof(tb[0])instore_bitmap
inqcow2_store_persistent_dirty_bitmaps:
fail if !can_write
fix counting ofnew_nb_bitmapsandnew_dir_size
free old clusters _after_ successful directory update (aggregate old bitmap
tables intodrop_tables)
2 14: rename tocan_store_new_dirty_bitmap
Max's r-b
15: rename tocan_store_new_dirty_bitmap
rm extra !!
16: Max's r-b
rename tocan_store_new_dirty_bitmap
indenting
Since s/2.8/2.9
17: Max's r-bSince s/2.8/2.918: hashing can fail in comment for qapi
Since s/2.8/2.9
remove dead comment
19: indentation
Max's r-b
20: Max's r-b
21: fix error handlingreturn 0 if nb_bitmaps == 0new patches:
22-23: remove bitmap from file onbdrv_release_dirty_bitmap
24: experimental, RFC: cache bitmap list to bs, to not reload it from file.
I'm not sure that is needed now, but if you want I can merge it to earlier patches.v9:
rebase on master!
01-04,06,07: John's r-b
03: rewording in comment ("The concurrent resetting ..."), John's r-b 07: reword error messagescommit message: dirty bitmap -> bitmap09: fix uninitialized ret in
bitmap_list_store()(fix compilation warning) 18: fix compilation of tests/test-hbitmapalso, change constants names for qcow2:
DIRTY_BITMAP-> BITMAP (as it is not dirty bitmaps extension, but just bitmaps extension), QCOW -> QCOW2 (bitmaps are only for qcow2)v8:
Patches 01-06 are mostly unchanged, except for spelling and wording. 02 - add Eric's r-b
03,04 - add Max's r-b07-09 is refactored "bitmap-read" part of the feature. Main changes:
- introduce bitmap list - normal list, which represents bitmap directory. Functionbitmap_list_loadloads bitmap directory, checks everything and creates normal list.
- introduce .bdrv_load_autoloading_dirty_bitmaps : instead of loading bitmaps inqcow2_open, lets load them only inbdrv_open, to avoid reloading inbdrv_snapshot_goto
- do not delete bitmaps after read, but mark themIN_USE10 has contextual changes and rewording of comment. I've added Max's r-b, hope it's ok.
11: add
error_report("Persistent bitmaps are lost")in case of failed bitmap store12: add Max's r-b
13 is refactored "bitmap-store" part of the feature. see 07-09 description
- for now I just free old clusters and allocate new. This will be improved with a separate patch.patch about "delete bitmaps on truncate" is removed. This case will be handled later.
IN_USE, autoclear, check-constraints things are merged into other patches.14-15 are mew patches, to early check possibility of creating persistent bitmap with
specified name and granularity in specified BDS16-17: spelling, rewording, indenting, tiny code simplifying, check
can_store(by 14-15),
s/if (autoload && !persistent)/if (has_autoload&& !persistent)/, rebase (deleted qmp-commands.hx)18: alternative to md5 in query-block:
- separated qmp command -x-debug-block-dirty-bitmap-sha256
- as adviced by Daniel P. Berrange in my parallel thread:
- sha256 instead of md5
- use qcrypto_hash_* instead of GChecksum
- fix bug =) (size was wrong inhbitmap_md5)19: s/3999/3fff/, use x-debug-block-dirty-bitmap-sha256
20: new patch to rename and publish
inc_refcounts21: some fixes and refactoring mostyly by Max's comments.
Max, Eric, great tanks for your review!
I hope, I've covered most of your comments by this update.TODO:
- handle reopening imageRO->RWand incoming migration, setIN_USEfor existing loaded bitmaps in these cases.
- reuse old, already allocated data clusters for bitmaps storing
- truncate bitmaps in the image on truncatev7:
https://src.openvz.org/users/vsementsov/repos/qemu/browse?at=refs%2Ftags%2Fqcow2-bitmap-v7
based on block-next (https://github.com/XanClic/qemu/commits/block-next)- a lot of refactoring and reordering of patches.
- dead code removed (bdrv_dirty_bitmap_load, etc.)
- do not maintain extra data for now
- do not store dirty bitmap directory in memory (as we use it seldom, we can just reread if needed)By Kevin's review:
01 - commit message changed:fix->improvement(as it was not a bug) 03 - r-b
04 - r-b
05 - add 21 patch to fix spec, also, removed all (I hope) mentions of
"Bitmap Header", switch to one unified name for it - "Bitmap Directory Entry", to avoid misunderstanding with Qcow2 header. (also, add patch 22, to fix it in spec)
v6.06 - improvefor_each_dir_entryloop, reorder patches, other small fixes v6.07 ~> v7.09 - dead code removed, I've moved to one function
.bdrv_store_persistent_bitmaps and have wrapper and callback in one patch (with also some other staff. If it not ok, I can split them)
v6.08 - about keeping bitmap directory instead of bitmap list: no I don't keep
it at all.
v6.16 - oldbdrv_bitmap-storing related functions are removed. The new one is
bdrv_store_persistent_bitmaps.v6:
https://src.openvz.org/users/vsementsov/repos/qemu/browse?at=refs%2Ftags%2Fqcow2-bitmap-v6
based on block-next (https://github.com/XanClic/qemu/commits/block-next)There are a lot of changes, reorderings and additions in comparement with v5. One principal thing: now bitmaps are removed from image after loading instead of marking them
in_use. It is simpler and we do not need to store superfluous data. Also, we are no more interested in command line interface to dirty bitmaps. So it is dropped. If someone needs it I can add it later.Vladimir Sementsov-Ogievskiy (30):
specs/qcow2: fix bitmap granularity qemu-specific note specs/qcow2: do not use wording 'bitmap header'
hbitmap: improve dirty iter
tests: add hbitmap iter test
block: fixbdrv_dirty_bitmap_granularitysignatureblock/dirty-bitmap: adddeserialize_onesfunc
qcow2-refcount: renameinc_refcounts()and make it public
qcow2: add bitmaps extension
block/dirty-bitmap: fix comment for BlockDirtyBitmap.disabled fieldblock/dirty-bitmap: add readonly field to BdrvDirtyBitmap
qcow2: autoloading dirty bitmaps
block: refactorbdrv_reopen_commit
block: newbdrv_reopen_bitmaps_rwinterface
qcow2: support .bdrv_reopen_bitmaps_rwblock/dirty-bitmap: add autoload field to BdrvDirtyBitmap
block:bdrv_close: release bitmaps afterdrv->bdrv_close
block: introduce persistent dirty bitmapsblock/dirty-bitmap: addbdrv_dirty_bitmap_next()
qcow2: add persistent dirty bitmaps support
qcow2: store bitmaps on reopening image as read-only
block: addbdrv_can_store_new_dirty_bitmap
qcow2: add .bdrv_can_store_new_dirty_bitmap
qmp: add persistent flag to block-dirty-bitmap-add
qmp: add autoload parameter to block-dirty-bitmap-add
qmp: add x-debug-block-dirty-bitmap-sha256
iotests: test qcow2 persistent dirty bitmapblock/dirty-bitmap: addbdrv_remove_persistent_dirty_bitmap
qcow2: add .bdrv_remove_persistent_dirty_bitmap
qmp: block-dirty-bitmap-remove: remove persistent
block: release persistent bitmaps on inactivateblock.c | 65 +- block/Makefile.objs | 2 +- block/dirty-bitmap.c | 154 ++++- block/io.c | 8 + block/qcow2-bitmap.c | 1481 ++++++++++++++++++++++++++++++++++++++++++ block/qcow2-refcount.c | 59 +- block/qcow2.c | 155 ++++- block/qcow2.h | 43 ++ blockdev.c | 73 ++- docs/interop/qcow2.txt | 8 +- include/block/block.h | 3 + include/block/block_int.h | 14 + include/block/dirty-bitmap.h | 22 +- include/qemu/hbitmap.h | 49 +- qapi/block-core.json | 42 +- tests/Makefile.include | 2 +- tests/qemu-iotests/165 | 105 +++ tests/qemu-iotests/165.out | 5 + tests/qemu-iotests/group | 1 + tests/test-hbitmap.c | 19 + util/hbitmap.c | 51 +- 21 files changed, 2280 insertions(+), 81 deletions(-) create mode 100644 block/qcow2-bitmap.c create mode 100755 tests/qemu-iotests/165 create mode 100644 tests/qemu-iotests/165.out
Quoted 3 lines · Vladimir Sementsov-Ogievskiy
Hi all!
There is a new update of qcow2-bitmap series - v22.
Thank you for all the work.
After resolving some conflicts with the qcow2 LUKS series (and after bisecting some unrelated iotest 055 breakage), I've applied this series to my block branch:
https://github.com/XanClic/qemu/commits/block
(I've also fixed the indentation issue Eric has found in patch 17.)
Max
Revisiting an old series:
Quoted 29 lines · Vladimir Sementsov-Ogievskiy
Hi all!
There is a new update of qcow2-bitmap series - v22.
web: https://src.openvz.org/users/vsementsov/repos/qemu/browse?at=qcow2-bitmap-v22
git: https://src.openvz.org/scm/~vsementsov/qemu.git (tag qcow2-bitmap-v22)v22:
Rebase on master, so changes, mostly related to new dirty bitmaps mutex:
qcow2: add persistent dirty bitmaps support
block.c | 65 +- block/Makefile.objs | 2 +- block/dirty-bitmap.c | 154 ++++- block/io.c | 8 + block/qcow2-bitmap.c | 1481 ++++++++++++++++++++++++++++++++++++++++++ block/qcow2-refcount.c | 59 +- block/qcow2.c | 155 ++++- block/qcow2.h | 43 ++ blockdev.c | 73 ++- docs/interop/qcow2.txt | 8 +- include/block/block.h | 3 + include/block/block_int.h | 14 + include/block/dirty-bitmap.h | 22 +- include/qemu/hbitmap.h | 49 +-
Quoted 8 lines
qapi/block-core.json | 42 +- tests/Makefile.include | 2 +- tests/qemu-iotests/165 | 105 +++ tests/qemu-iotests/165.out | 5 + tests/qemu-iotests/group | 1 + tests/test-hbitmap.c | 19 + util/hbitmap.c | 51 +- 21 files changed, 2280 insertions(+), 81 deletions(-)
Nothing in this series touched qemu-img. At the very minimum, it would be nice if 'qemu-img info' were to display a summary of all bitmaps in the qcow2 file; the QMP type ImageInfoSpecificQCow2 should be updated to mention this information, and qemu-img updated to output it. Being able to create and remove persistent bitmaps from an at-rest qcow2 file without having to fire up qemu to issue QMP commands to do so would also be nice.
It's a bit late for getting this request into 2.12, but had it been available, it would make my work at testing that 'qemu-img commit' doesn't corrupt persistent bitmaps.
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
Quoted 45 lines · Eric Blake
Revisiting an old series:
Quoted 26 lines · Vladimir Sementsov-Ogievskiy
Hi all!
There is a new update of qcow2-bitmap series - v22.
web: https://src.openvz.org/users/vsementsov/repos/qemu/browse?at=qcow2-bitmap-v22
git: https://src.openvz.org/scm/~vsementsov/qemu.git (tag qcow2-bitmap-v22)v22:
Rebase on master, so changes, mostly related to new dirty bitmaps mutex:
qcow2: add persistent dirty bitmaps support
block.c | 65 +- block/Makefile.objs | 2 +- block/dirty-bitmap.c | 154 ++++- block/io.c | 8 + block/qcow2-bitmap.c | 1481 ++++++++++++++++++++++++++++++++++++++++++ block/qcow2-refcount.c | 59 +- block/qcow2.c | 155 ++++- block/qcow2.h | 43 ++ blockdev.c | 73 ++- docs/interop/qcow2.txt | 8 +- include/block/block.h | 3 + include/block/block_int.h | 14 + include/block/dirty-bitmap.h | 22 +- include/qemu/hbitmap.h | 49 +-Quoted 8 lines
qapi/block-core.json | 42 +- tests/Makefile.include | 2 +- tests/qemu-iotests/165 | 105 +++ tests/qemu-iotests/165.out | 5 + tests/qemu-iotests/group | 1 + tests/test-hbitmap.c | 19 + util/hbitmap.c | 51 +- 21 files changed, 2280 insertions(+), 81 deletions(-)Nothing in this series touched qemu-img. At the very minimum, it would be nice if 'qemu-img info' were to display a summary of all bitmaps in the qcow2 file; the QMP type ImageInfoSpecificQCow2 should be updated to mention this information, and qemu-img updated to output it. Being able to create and remove persistent bitmaps from an at-rest qcow2 file without having to fire up qemu to issue QMP commands to do so would also be nice.
Quoted 4 lines
It's a bit late for getting this request into 2.12, but had it been available, it would make my work at testing that 'qemu-img commit' doesn't corrupt persistent bitmaps.
You are right, this is needed, but I can't promise something for the nearest future..
Signature · 2 lines
Best regards, Vladimir
Quoted 9 lines · Vladimir Sementsov-Ogievskiy
Quoted 9 lines
Nothing in this series touched qemu-img. At the very minimum, it would be nice if 'qemu-img info' were to display a summary of all bitmaps in the qcow2 file; the QMP type ImageInfoSpecificQCow2 should be updated to mention this information, and qemu-img updated to output it. Being able to create and remove persistent bitmaps from an at-rest qcow2 file without having to fire up qemu to issue QMP commands to do so would also be nice.
It's a bit late for getting this request into 2.12, but had it been
Make that "too late for 2.11"
Quoted 6 lines
Quoted 3 lines
available, it would make my work at testing that 'qemu-img commit' doesn't corrupt persistent bitmaps.
You are right, this is needed, but I can't promise something for the nearest future..
And as we both agree it is work for 2.12 or later, it is not needed right away; so we have a couple of months.
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org