Am 21.05.25 um 18:05 schrieb Kevin Wolf: > Am 20.05.2025 um 12:29 hat Fiona Ebner geschrieben: >> This is in preparation to mark bdrv_drained_begin() as GRAPH_UNLOCKED. >> >> Note that even if bdrv_drained_begin() would already be marked as > > "if ... were already marked"
Ack. ---snip 8<--- >> diff --git a/block.c b/block.c >> index 01144c895e..7148618504 100644 >> --- a/block.c >> +++ b/block.c >> @@ -106,9 +106,9 @@ static void bdrv_reopen_abort(BDRVReopenState >> *reopen_state); >> >> static bool bdrv_backing_overridden(BlockDriverState *bs); >> >> -static bool bdrv_change_aio_context(BlockDriverState *bs, AioContext *ctx, >> - GHashTable *visited, Transaction *tran, >> - Error **errp); >> +static bool GRAPH_RDLOCK >> +bdrv_change_aio_context(BlockDriverState *bs, AioContext *ctx, >> + GHashTable *visited, Transaction *tran, Error >> **errp); > > For static functions, we should have rhe GRAPH_RDLOCK annotation both > here and in the actual definition, too. Will fix in v3! >> /* If non-zero, use only whitelisted block drivers */ >> static int use_bdrv_whitelist; >> @@ -3040,8 +3040,10 @@ static void GRAPH_WRLOCK >> bdrv_attach_child_common_abort(void *opaque) >> >> /* No need to visit `child`, because it has been detached already */ >> visited = g_hash_table_new(NULL, NULL); >> + bdrv_drain_all_begin(); >> ret = s->child->klass->change_aio_ctx(s->child, s->old_parent_ctx, >> visited, tran, &error_abort); >> + bdrv_drain_all_end(); >> g_hash_table_destroy(visited); >> >> /* transaction is supposed to always succeed */ >> @@ -3122,9 +3124,11 @@ bdrv_attach_child_common(BlockDriverState *child_bs, >> bool ret_child; >> >> g_hash_table_add(visited, new_child); >> + bdrv_drain_all_begin(); >> ret_child = child_class->change_aio_ctx(new_child, child_ctx, >> visited, aio_ctx_tran, >> NULL); >> + bdrv_drain_all_end(); >> if (ret_child == true) { >> error_free(local_err); >> ret = 0; > > Should we document in the header file that BdrvChildClass.change_aio_ctx > is called with the node drained? > > We could add assertions to bdrv_child/parent_change_aio_context or at > least comments to this effect. (Assertions might be over the top because > it's easy to verify that both are only called from > bdrv_change_aio_context().) Sounds good. Would the following be okay? For bdrv_parent_change_aio_context() and for change_aio_ctx() the same (except the name of @child is @c for the former): > /* > * Changes the AioContext of for @child to @ctx and recursively for the > * associated block nodes and all their children and parents. Returns true if > * the change is possible and the transaction @tran can be continued. Returns > * false and sets @errp if not and the transaction must be aborted. > * > * @visited will accumulate all visited BdrvChild objects. The caller is > * responsible for freeing the list afterwards. > * > * Must be called with the affected block nodes drained. > */ and for bdrv_child_change_aio_context() slightly different: > /* > * Changes the AioContext of for @c->bs to @ctx and recursively for all its > * children and parents. Returns true if the change is possible and the > * transaction @tran can be continued. Returns false and sets @errp if not and > * the transaction must be aborted. > * > * @visited will accumulate all visited BdrvChild objects. The caller is > * responsible for freeing the list afterwards. > * > * Must be called with the affected block nodes drained. > */ ---snip 8<--- >> @@ -7720,7 +7717,11 @@ int bdrv_try_change_aio_context(BlockDriverState *bs, >> AioContext *ctx, >> if (ignore_child) { >> g_hash_table_add(visited, ignore_child); >> } >> + bdrv_drain_all_begin(); >> + bdrv_graph_rdlock_main_loop(); >> ret = bdrv_change_aio_context(bs, ctx, visited, tran, errp); >> + bdrv_graph_rdunlock_main_loop(); >> + bdrv_drain_all_end(); >> g_hash_table_destroy(visited); > > I think you're ending the drained section too early here. Previously, > the nodes were kept drained until after tran_abort/commit(), and I think > that's important (tran_commit() is the thing that actually switches the > AioContext). Right, I'll change this in v3. Best Regards, Fiona