On 2019/12/23 21:40, Vladimir Sementsov-Ogievskiy wrote:
> 23.12.2019 12:06, Eiichi Tsukata wrote:
>> bdrv_open_driver() allocates bs->opaque according to drv->instance_size.
>> There is no need to allocate it and overwrite opaque in
>> bdrv_backup_top_append().
>>
>> Reproducer:
>>
>> $ QTEST_QEMU_BINARY=./x86_64-softmmu/qemu-system-x86_64 valgrind -q
>> --leak-check=full tests/test-replication -p /replication/secondary/start
>> ==29792== 24 bytes in 1 blocks are definitely lost in loss record 52 of
>> 226
>> ==29792== at 0x483AB1A: calloc (vg_replace_malloc.c:762)
>> ==29792== by 0x4B07CE0: g_malloc0 (in
>> /usr/lib64/libglib-2.0.so.0.6000.7)
>> ==29792== by 0x12BAB9: bdrv_open_driver (block.c:1289)
>> ==29792== by 0x12BEA9: bdrv_new_open_driver (block.c:1359)
>> ==29792== by 0x1D15CB: bdrv_backup_top_append (backup-top.c:190)
>> ==29792== by 0x1CC11A: backup_job_create (backup.c:439)
>> ==29792== by 0x1CD542: replication_start (replication.c:544)
>> ==29792== by 0x1401B9: replication_start_all (replication.c:52)
>> ==29792== by 0x128B50: test_secondary_start (test-replication.c:427)
>> ...
>>
>> Fixes: 7df7868b9640 ("block: introduce backup-top filter driver")
>> Signed-off-by: Eiichi Tsukata <de...@etsukata.com>
>> ---
>> block/backup-top.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/block/backup-top.c b/block/backup-top.c
>> index 7cdb1f8eba..617217374d 100644
>> --- a/block/backup-top.c
>> +++ b/block/backup-top.c
>> @@ -196,7 +196,7 @@ BlockDriverState
>> *bdrv_backup_top_append(BlockDriverState *source,
>> }
>>
>> top->total_sectors = source->total_sectors;
>> - top->opaque = state = g_new0(BDRVBackupTopState, 1);
>> + state = top->opaque;
>>
>> bdrv_ref(target);
>> state->target = bdrv_attach_child(top, target, "target", &child_file,
>> errp);
>>
>
> Reviewed-by: Vladimir Sementsov-Ogievskiy <vsement...@virtuozzo.com>
>
> Hmm, it was not my idea, I just copied it from mirror.. And there should be
> the same leak. and
> may be in other places:
>
> # git grep 'opaque =.*g_new'
> block/backup-top.c: top->opaque = state = g_new0(BDRVBackupTopState, 1);
> block/file-posix.c: state->opaque = g_new0(BDRVRawReopenState, 1);
> block/gluster.c: state->opaque = g_new0(BDRVGlusterReopenState, 1);
> block/iscsi.c: bs->opaque = g_new0(struct IscsiLun, 1);
> block/mirror.c: bs_opaque = g_new0(MirrorBDSOpaque, 1);
> block/raw-format.c: reopen_state->opaque = g_new0(BDRVRawState, 1);
> block/sheepdog.c: re_s = state->opaque = g_new0(BDRVSheepdogReopenState,
> 1);
>
>
>
Thanks for reviewing.
As you say, block/mirror.c has similar code. But it does not cause the leak.
The difference is bdrv_mirror_top BlockDriver does not have .instance_size
whereas bdrv_backup_top_filter BlockDriver has .instance_size =
sizeof(BDRVBackupTopState).
So when bdrv_open_driver() is called from mirror.c, g_malloc0(0) is
called allocating nothing.
Eiichi
