| Message ID | 1488817322-11397-6-git-send-email-kwolf@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 03/06/2017 10:21 AM, Kevin Wolf wrote: > For blockdev-snapshot, external_snapshot_prepare() accepts an arbitrary > node reference at first and only checks later whether it already has a > backing file. Between those places, other errors can occur. > > Therefore checking in external_snapshot_abort() whether state->new_bs > has a backing file is not sufficient to tell whether bdrv_append() was > already completed or not. Trying to undo the bdrv_append() when it > wasn't even executed is wrong. > > Introduce a new boolean flag in the state to fix this. > > Signed-off-by: Kevin Wolf <kwolf@redhat.com> > --- > blockdev.c | 4 +++- > 1 file changed, 3 insertions(+), 1 deletion(-) > Reviewed-by: Eric Blake <eblake@redhat.com> By the way, how are you finding all these spots? Is it existing qemu-io tests that are failing? And if so, would mentioning which test exposed the problem being fixed be worth adding in the commit messages? If not, are there some qemu-io tests to be added?
Am 06.03.2017 um 21:23 hat Eric Blake geschrieben: > On 03/06/2017 10:21 AM, Kevin Wolf wrote: > > For blockdev-snapshot, external_snapshot_prepare() accepts an arbitrary > > node reference at first and only checks later whether it already has a > > backing file. Between those places, other errors can occur. > > > > Therefore checking in external_snapshot_abort() whether state->new_bs > > has a backing file is not sufficient to tell whether bdrv_append() was > > already completed or not. Trying to undo the bdrv_append() when it > > wasn't even executed is wrong. > > > > Introduce a new boolean flag in the state to fix this. > > > > Signed-off-by: Kevin Wolf <kwolf@redhat.com> > > --- > > blockdev.c | 4 +++- > > 1 file changed, 3 insertions(+), 1 deletion(-) > > > > Reviewed-by: Eric Blake <eblake@redhat.com> > > By the way, how are you finding all these spots? Is it existing qemu-io > tests that are failing? And if so, would mentioning which test exposed > the problem being fixed be worth adding in the commit messages? If not, > are there some qemu-io tests to be added? Most of the fixes here are based on qemu-iotests failures that only appeared after fixing the error path in change_parent_backing_link(), which is now later in this series. So there is no commit at which qemu-iotests cases are actually failing, even though they helped me spot some problems. Kevin
diff --git a/blockdev.c b/blockdev.c index 8eb4e84..af67ce4 100644 --- a/blockdev.c +++ b/blockdev.c @@ -1614,6 +1614,7 @@ typedef struct ExternalSnapshotState { BlockDriverState *old_bs; BlockDriverState *new_bs; AioContext *aio_context; + bool overlay_appended; } ExternalSnapshotState; static void external_snapshot_prepare(BlkActionState *common, @@ -1780,6 +1781,7 @@ static void external_snapshot_prepare(BlkActionState *common, error_propagate(errp, local_err); return; } + state->overlay_appended = true; } static void external_snapshot_commit(BlkActionState *common) @@ -1803,7 +1805,7 @@ static void external_snapshot_abort(BlkActionState *common) ExternalSnapshotState *state = DO_UPCAST(ExternalSnapshotState, common, common); if (state->new_bs) { - if (state->new_bs->backing) { + if (state->overlay_appended) { bdrv_replace_in_backing_chain(state->new_bs, state->old_bs); } }
For blockdev-snapshot, external_snapshot_prepare() accepts an arbitrary node reference at first and only checks later whether it already has a backing file. Between those places, other errors can occur. Therefore checking in external_snapshot_abort() whether state->new_bs has a backing file is not sufficient to tell whether bdrv_append() was already completed or not. Trying to undo the bdrv_append() when it wasn't even executed is wrong. Introduce a new boolean flag in the state to fix this. Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- blockdev.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)