| Message ID | 20200310113831.27293-6-kwolf@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | block: Relax restrictions for blockdev-snapshot | expand |
On Tue, Mar 10, 2020 at 12:38:29 +0100, Kevin Wolf wrote: > external_snapshot_prepare() tries to move the overlay to the AioContext > of the backing file (the snapshotted node). However, it's possible that > this doesn't work, but the backing file can instead be moved to the > overlay's AioContext (e.g. opening the backing chain for a mirror > target). > > bdrv_append() already indirectly uses bdrv_attach_node(), which takes > care to move nodes to make sure they use the same AioContext and which > tries both directions. > > So the problem has a simple fix: Just delete the unnecessary extra > bdrv_try_set_aio_context() call in external_snapshot_prepare() and > instead assert in bdrv_append() that both nodes were indeed moved to the > same AioContext. > > Signed-off-by: Kevin Wolf <kwolf@redhat.com> > --- > block.c | 1 + > blockdev.c | 16 ---------------- > 2 files changed, 1 insertion(+), 16 deletions(-) Tested-by: Peter Krempa <pkrempa@redhat.com> Unfortunately I don't feel confident enough in the intricacies of the aio contexts. Nevertheless Fixing bugs by deleting code is awesome!
diff --git a/block.c b/block.c index 79a5a2770f..8fc7b56937 100644 --- a/block.c +++ b/block.c @@ -4365,6 +4365,7 @@ void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to, bdrv_ref(from); assert(qemu_get_current_aio_context() == qemu_get_aio_context()); + assert(bdrv_get_aio_context(from) == bdrv_get_aio_context(to)); bdrv_drained_begin(from); /* Put all parents into @list and calculate their cumulative permissions */ diff --git a/blockdev.c b/blockdev.c index bba0e9775b..95a0b53d57 100644 --- a/blockdev.c +++ b/blockdev.c @@ -1500,9 +1500,7 @@ static void external_snapshot_prepare(BlkActionState *common, DO_UPCAST(ExternalSnapshotState, common, common); TransactionAction *action = common->action; AioContext *aio_context; - AioContext *old_context; uint64_t perm, shared; - int ret; /* 'blockdev-snapshot' and 'blockdev-snapshot-sync' have similar * purpose but a different set of parameters */ @@ -1638,20 +1636,6 @@ static void external_snapshot_prepare(BlkActionState *common, goto out; } - /* Honor bdrv_try_set_aio_context() context acquisition requirements. */ - old_context = bdrv_get_aio_context(state->new_bs); - aio_context_release(aio_context); - aio_context_acquire(old_context); - - ret = bdrv_try_set_aio_context(state->new_bs, aio_context, errp); - - aio_context_release(old_context); - aio_context_acquire(aio_context); - - if (ret < 0) { - goto out; - } - /* This removes our old bs and adds the new bs. This is an operation that * can fail, so we need to do it in .prepare; undoing it for abort is * always possible. */
external_snapshot_prepare() tries to move the overlay to the AioContext of the backing file (the snapshotted node). However, it's possible that this doesn't work, but the backing file can instead be moved to the overlay's AioContext (e.g. opening the backing chain for a mirror target). bdrv_append() already indirectly uses bdrv_attach_node(), which takes care to move nodes to make sure they use the same AioContext and which tries both directions. So the problem has a simple fix: Just delete the unnecessary extra bdrv_try_set_aio_context() call in external_snapshot_prepare() and instead assert in bdrv_append() that both nodes were indeed moved to the same AioContext. Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block.c | 1 + blockdev.c | 16 ---------------- 2 files changed, 1 insertion(+), 16 deletions(-)