diff options
Diffstat (limited to 'block.c')
| -rw-r--r-- | block.c | 266 |
1 files changed, 160 insertions, 106 deletions
@@ -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); /* If non-zero, use only whitelisted block drivers */ static int use_bdrv_whitelist; @@ -431,7 +431,7 @@ BlockDriverState *bdrv_new(void) bs->block_status_cache = g_new0(BdrvBlockStatusCache, 1); for (i = 0; i < bdrv_drain_all_count; i++) { - bdrv_drained_begin(bs); + bdrv_do_drained_begin_quiesce(bs, NULL); } QTAILQ_INSERT_TAIL(&all_bdrv_states, bs, bs_list); @@ -1226,9 +1226,10 @@ static int bdrv_child_cb_inactivate(BdrvChild *child) return 0; } -static bool bdrv_child_cb_change_aio_ctx(BdrvChild *child, AioContext *ctx, - GHashTable *visited, Transaction *tran, - Error **errp) +static bool GRAPH_RDLOCK +bdrv_child_cb_change_aio_ctx(BdrvChild *child, AioContext *ctx, + GHashTable *visited, Transaction *tran, + Error **errp) { BlockDriverState *bs = child->opaque; return bdrv_change_aio_context(bs, ctx, visited, tran, errp); @@ -1720,7 +1721,7 @@ bdrv_open_driver(BlockDriverState *bs, BlockDriver *drv, const char *node_name, open_failed: bs->drv = NULL; - bdrv_graph_wrlock(); + bdrv_graph_wrlock_drained(); if (bs->file != NULL) { bdrv_unref_child(bs, bs->file); assert(!bs->file); @@ -3027,7 +3028,8 @@ static void GRAPH_WRLOCK bdrv_attach_child_common_abort(void *opaque) bdrv_replace_child_noperm(s->child, NULL); if (bdrv_get_aio_context(bs) != s->old_child_ctx) { - bdrv_try_change_aio_context(bs, s->old_child_ctx, NULL, &error_abort); + bdrv_try_change_aio_context_locked(bs, s->old_child_ctx, NULL, + &error_abort); } if (bdrv_child_get_parent_aio_context(s->child) != s->old_parent_ctx) { @@ -3069,6 +3071,9 @@ static TransactionActionDrv bdrv_attach_child_common_drv = { * * Both @parent_bs and @child_bs can move to a different AioContext in this * function. + * + * All block nodes must be drained before this function is called until after + * the transaction is finalized. */ static BdrvChild * GRAPH_WRLOCK bdrv_attach_child_common(BlockDriverState *child_bs, @@ -3112,8 +3117,8 @@ bdrv_attach_child_common(BlockDriverState *child_bs, parent_ctx = bdrv_child_get_parent_aio_context(new_child); if (child_ctx != parent_ctx) { Error *local_err = NULL; - int ret = bdrv_try_change_aio_context(child_bs, parent_ctx, NULL, - &local_err); + int ret = bdrv_try_change_aio_context_locked(child_bs, parent_ctx, NULL, + &local_err); if (ret < 0 && child_class->change_aio_ctx) { Transaction *aio_ctx_tran = tran_new(); @@ -3153,7 +3158,7 @@ bdrv_attach_child_common(BlockDriverState *child_bs, * stop new requests from coming in. This is fine, we don't care about the * old requests here, they are not for this child. If another place enters a * drain section for the same parent, but wants it to be fully quiesced, it - * will not run most of the the code in .drained_begin() again (which is not + * will not run most of the code in .drained_begin() again (which is not * a problem, we already did this), but it will still poll until the parent * is fully quiesced, so it will not be negatively affected either. */ @@ -3179,6 +3184,9 @@ bdrv_attach_child_common(BlockDriverState *child_bs, * * After calling this function, the transaction @tran may only be completed * while holding a writer lock for the graph. + * + * All block nodes must be drained before this function is called until after + * the transaction is finalized. */ static BdrvChild * GRAPH_WRLOCK bdrv_attach_child_noperm(BlockDriverState *parent_bs, @@ -3220,6 +3228,8 @@ bdrv_attach_child_noperm(BlockDriverState *parent_bs, * * On failure NULL is returned, errp is set and the reference to * child_bs is also dropped. + * + * All block nodes must be drained. */ BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs, const char *child_name, @@ -3259,6 +3269,8 @@ out: * * On failure NULL is returned, errp is set and the reference to * child_bs is also dropped. + * + * All block nodes must be drained. */ BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs, BlockDriverState *child_bs, @@ -3293,7 +3305,11 @@ out: return ret < 0 ? NULL : child; } -/* Callers must ensure that child->frozen is false. */ +/* + * Callers must ensure that child->frozen is false. + * + * All block nodes must be drained. + */ void bdrv_root_unref_child(BdrvChild *child) { BlockDriverState *child_bs = child->bs; @@ -3314,8 +3330,8 @@ void bdrv_root_unref_child(BdrvChild *child) * When the parent requiring a non-default AioContext is removed, the * node moves back to the main AioContext */ - bdrv_try_change_aio_context(child_bs, qemu_get_aio_context(), NULL, - NULL); + bdrv_try_change_aio_context_locked(child_bs, qemu_get_aio_context(), + NULL, NULL); } bdrv_schedule_unref(child_bs); @@ -3388,7 +3404,11 @@ bdrv_unset_inherits_from(BlockDriverState *root, BdrvChild *child, } } -/* Callers must ensure that child->frozen is false. */ +/* + * Callers must ensure that child->frozen is false. + * + * All block nodes must be drained. + */ void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child) { GLOBAL_STATE_CODE(); @@ -3453,6 +3473,9 @@ static BdrvChildRole bdrv_backing_role(BlockDriverState *bs) * * After calling this function, the transaction @tran may only be completed * while holding a writer lock for the graph. + * + * All block nodes must be drained before this function is called until after + * the transaction is finalized. */ static int GRAPH_WRLOCK bdrv_set_file_or_backing_noperm(BlockDriverState *parent_bs, @@ -3545,12 +3568,10 @@ out: * Both @bs and @backing_hd can move to a different AioContext in this * function. * - * If a backing child is already present (i.e. we're detaching a node), that - * child node must be drained. + * All block nodes must be drained. */ -int bdrv_set_backing_hd_drained(BlockDriverState *bs, - BlockDriverState *backing_hd, - Error **errp) +int bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd, + Error **errp) { int ret; Transaction *tran = tran_new(); @@ -3572,28 +3593,6 @@ out: return ret; } -int bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd, - Error **errp) -{ - BlockDriverState *drain_bs; - int ret; - GLOBAL_STATE_CODE(); - - bdrv_graph_rdlock_main_loop(); - drain_bs = bs->backing ? bs->backing->bs : bs; - bdrv_graph_rdunlock_main_loop(); - - bdrv_ref(drain_bs); - bdrv_drained_begin(drain_bs); - bdrv_graph_wrlock(); - ret = bdrv_set_backing_hd_drained(bs, backing_hd, errp); - bdrv_graph_wrunlock(); - bdrv_drained_end(drain_bs); - bdrv_unref(drain_bs); - - return ret; -} - /* * Opens the backing file for a BlockDriverState if not yet open * @@ -3619,7 +3618,8 @@ int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options, Error *local_err = NULL; GLOBAL_STATE_CODE(); - GRAPH_RDLOCK_GUARD_MAINLOOP(); + + bdrv_graph_rdlock_main_loop(); if (bs->backing != NULL) { goto free_exit; @@ -3700,7 +3700,11 @@ int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options, /* Hook up the backing file link; drop our reference, bs owns the * backing_hd reference now */ + bdrv_graph_rdunlock_main_loop(); + bdrv_graph_wrlock_drained(); ret = bdrv_set_backing_hd(bs, backing_hd, errp); + bdrv_graph_wrunlock(); + bdrv_graph_rdlock_main_loop(); bdrv_unref(backing_hd); if (ret < 0) { @@ -3712,6 +3716,7 @@ int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options, free_exit: g_free(backing_filename); qobject_unref(tmp_parent_options); + bdrv_graph_rdunlock_main_loop(); return ret; } @@ -3761,13 +3766,12 @@ done: return bs; } -static BdrvChild *bdrv_open_child_common(const char *filename, - QDict *options, const char *bdref_key, - BlockDriverState *parent, - const BdrvChildClass *child_class, - BdrvChildRole child_role, - bool allow_none, bool parse_filename, - Error **errp) +static BdrvChild * GRAPH_UNLOCKED +bdrv_open_child_common(const char *filename, QDict *options, + const char *bdref_key, BlockDriverState *parent, + const BdrvChildClass *child_class, + BdrvChildRole child_role, bool allow_none, + bool parse_filename, Error **errp) { BlockDriverState *bs; BdrvChild *child; @@ -3780,7 +3784,7 @@ static BdrvChild *bdrv_open_child_common(const char *filename, return NULL; } - bdrv_graph_wrlock(); + bdrv_graph_wrlock_drained(); child = bdrv_attach_child(parent, bs, bdref_key, child_class, child_role, errp); bdrv_graph_wrunlock(); @@ -4358,9 +4362,7 @@ bdrv_recurse_has_child(BlockDriverState *bs, BlockDriverState *child) * returns a pointer to bs_queue, which is either the newly allocated * bs_queue, or the existing bs_queue being used. * - * bs is drained here and undrained by bdrv_reopen_queue_free(). - * - * To be called with bs->aio_context locked. + * bs must be drained. */ static BlockReopenQueue * GRAPH_RDLOCK bdrv_reopen_queue_child(BlockReopenQueue *bs_queue, BlockDriverState *bs, @@ -4379,12 +4381,7 @@ bdrv_reopen_queue_child(BlockReopenQueue *bs_queue, BlockDriverState *bs, GLOBAL_STATE_CODE(); - /* - * Strictly speaking, draining is illegal under GRAPH_RDLOCK. We know that - * we've been called with bdrv_graph_rdlock_main_loop(), though, so it's ok - * in practice. - */ - bdrv_drained_begin(bs); + assert(bs->quiesce_counter > 0); if (bs_queue == NULL) { bs_queue = g_new0(BlockReopenQueue, 1); @@ -4519,12 +4516,17 @@ bdrv_reopen_queue_child(BlockReopenQueue *bs_queue, BlockDriverState *bs, return bs_queue; } -/* To be called with bs->aio_context locked */ BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue, BlockDriverState *bs, QDict *options, bool keep_old_opts) { GLOBAL_STATE_CODE(); + + if (bs_queue == NULL) { + /* Paired with bdrv_drain_all_end() in bdrv_reopen_queue_free(). */ + bdrv_drain_all_begin(); + } + GRAPH_RDLOCK_GUARD_MAINLOOP(); return bdrv_reopen_queue_child(bs_queue, bs, options, NULL, 0, false, @@ -4537,12 +4539,14 @@ void bdrv_reopen_queue_free(BlockReopenQueue *bs_queue) if (bs_queue) { BlockReopenQueueEntry *bs_entry, *next; QTAILQ_FOREACH_SAFE(bs_entry, bs_queue, entry, next) { - bdrv_drained_end(bs_entry->state.bs); qobject_unref(bs_entry->state.explicit_options); qobject_unref(bs_entry->state.options); g_free(bs_entry); } g_free(bs_queue); + + /* Paired with bdrv_drain_all_begin() in bdrv_reopen_queue(). */ + bdrv_drain_all_end(); } } @@ -4709,6 +4713,9 @@ int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only, * Return 0 on success, otherwise return < 0 and set @errp. * * @reopen_state->bs can move to a different AioContext in this function. + * + * All block nodes must be drained before this function is called until after + * the transaction is finalized. */ static int GRAPH_UNLOCKED bdrv_reopen_parse_file_or_backing(BDRVReopenState *reopen_state, @@ -4802,7 +4809,7 @@ bdrv_reopen_parse_file_or_backing(BDRVReopenState *reopen_state, if (old_child_bs) { bdrv_ref(old_child_bs); - bdrv_drained_begin(old_child_bs); + assert(old_child_bs->quiesce_counter > 0); } bdrv_graph_rdunlock_main_loop(); @@ -4814,7 +4821,6 @@ bdrv_reopen_parse_file_or_backing(BDRVReopenState *reopen_state, bdrv_graph_wrunlock(); if (old_child_bs) { - bdrv_drained_end(old_child_bs); bdrv_unref(old_child_bs); } @@ -4843,6 +4849,9 @@ out_rdlock: * * After calling this function, the transaction @change_child_tran may only be * completed while holding a writer lock for the graph. + * + * All block nodes must be drained before this function is called until after + * the transaction is finalized. */ static int GRAPH_UNLOCKED bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue *queue, @@ -5136,7 +5145,7 @@ static void GRAPH_UNLOCKED bdrv_reopen_abort(BDRVReopenState *reopen_state) } -static void bdrv_close(BlockDriverState *bs) +static void GRAPH_UNLOCKED bdrv_close(BlockDriverState *bs) { BdrvAioNotifier *ban, *ban_next; BdrvChild *child, *next; @@ -5156,7 +5165,7 @@ static void bdrv_close(BlockDriverState *bs) bs->drv = NULL; } - bdrv_graph_wrlock(); + bdrv_graph_wrlock_drained(); QLIST_FOREACH_SAFE(child, &bs->children, next, next) { bdrv_unref_child(bs, child); } @@ -5489,10 +5498,7 @@ int bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top, assert(!bs_new->backing); bdrv_graph_rdunlock_main_loop(); - bdrv_drained_begin(bs_top); - bdrv_drained_begin(bs_new); - - bdrv_graph_wrlock(); + bdrv_graph_wrlock_drained(); child = bdrv_attach_child_noperm(bs_new, bs_top, "backing", &child_of_bds, bdrv_backing_role(bs_new), @@ -5514,9 +5520,6 @@ out: bdrv_refresh_limits(bs_top, NULL, NULL); bdrv_graph_wrunlock(); - bdrv_drained_end(bs_top); - bdrv_drained_end(bs_new); - return ret; } @@ -6989,6 +6992,8 @@ bdrv_inactivate_recurse(BlockDriverState *bs, bool top_level) GLOBAL_STATE_CODE(); + assert(bs->quiesce_counter > 0); + if (!bs->drv) { return -ENOMEDIUM; } @@ -7032,9 +7037,7 @@ bdrv_inactivate_recurse(BlockDriverState *bs, bool top_level) return -EPERM; } - bdrv_drained_begin(bs); bs->open_flags |= BDRV_O_INACTIVE; - bdrv_drained_end(bs); /* * Update permissions, they may differ for inactive nodes. @@ -7054,12 +7057,12 @@ bdrv_inactivate_recurse(BlockDriverState *bs, bool top_level) return 0; } +/* All block nodes must be drained. */ int bdrv_inactivate(BlockDriverState *bs, Error **errp) { int ret; GLOBAL_STATE_CODE(); - GRAPH_RDLOCK_GUARD_MAINLOOP(); if (bdrv_has_bds_parent(bs, true)) { error_setg(errp, "Node has active parent node"); @@ -7082,7 +7085,9 @@ int bdrv_inactivate_all(void) int ret = 0; GLOBAL_STATE_CODE(); - GRAPH_RDLOCK_GUARD_MAINLOOP(); + + bdrv_drain_all_begin(); + bdrv_graph_rdlock_main_loop(); for (bs = bdrv_first(&it); bs; bs = bdrv_next(&it)) { /* Nodes with BDS parents are covered by recursion from the last @@ -7098,6 +7103,9 @@ int bdrv_inactivate_all(void) } } + bdrv_graph_rdunlock_main_loop(); + bdrv_drain_all_end(); + return ret; } @@ -7278,10 +7286,6 @@ bool bdrv_op_blocker_is_empty(BlockDriverState *bs) return true; } -/* - * Must not be called while holding the lock of an AioContext other than the - * current one. - */ void bdrv_img_create(const char *filename, const char *fmt, const char *base_filename, const char *base_fmt, char *options, uint64_t img_size, int flags, bool quiet, @@ -7568,10 +7572,21 @@ typedef struct BdrvStateSetAioContext { BlockDriverState *bs; } BdrvStateSetAioContext; -static bool bdrv_parent_change_aio_context(BdrvChild *c, AioContext *ctx, - GHashTable *visited, - Transaction *tran, - Error **errp) +/* + * Changes the AioContext of @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. + */ +static bool GRAPH_RDLOCK +bdrv_parent_change_aio_context(BdrvChild *c, AioContext *ctx, + GHashTable *visited, Transaction *tran, + Error **errp) { GLOBAL_STATE_CODE(); if (g_hash_table_contains(visited, c)) { @@ -7596,6 +7611,17 @@ static bool bdrv_parent_change_aio_context(BdrvChild *c, AioContext *ctx, return true; } +/* + * Changes the AioContext of @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. + */ bool bdrv_child_change_aio_context(BdrvChild *c, AioContext *ctx, GHashTable *visited, Transaction *tran, Error **errp) @@ -7611,10 +7637,6 @@ bool bdrv_child_change_aio_context(BdrvChild *c, AioContext *ctx, static void bdrv_set_aio_context_clean(void *opaque) { BdrvStateSetAioContext *state = (BdrvStateSetAioContext *) opaque; - BlockDriverState *bs = (BlockDriverState *) state->bs; - - /* Paired with bdrv_drained_begin in bdrv_change_aio_context() */ - bdrv_drained_end(bs); g_free(state); } @@ -7642,10 +7664,12 @@ static TransactionActionDrv set_aio_context = { * * @visited will accumulate all visited BdrvChild objects. The caller is * responsible for freeing the list afterwards. + * + * @bs must be drained. */ -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) { BdrvChild *c; BdrvStateSetAioContext *state; @@ -7656,21 +7680,17 @@ static bool bdrv_change_aio_context(BlockDriverState *bs, AioContext *ctx, return true; } - bdrv_graph_rdlock_main_loop(); QLIST_FOREACH(c, &bs->parents, next_parent) { if (!bdrv_parent_change_aio_context(c, ctx, visited, tran, errp)) { - bdrv_graph_rdunlock_main_loop(); return false; } } QLIST_FOREACH(c, &bs->children, next) { if (!bdrv_child_change_aio_context(c, ctx, visited, tran, errp)) { - bdrv_graph_rdunlock_main_loop(); return false; } } - bdrv_graph_rdunlock_main_loop(); state = g_new(BdrvStateSetAioContext, 1); *state = (BdrvStateSetAioContext) { @@ -7678,8 +7698,7 @@ static bool bdrv_change_aio_context(BlockDriverState *bs, AioContext *ctx, .bs = bs, }; - /* Paired with bdrv_drained_end in bdrv_set_aio_context_clean() */ - bdrv_drained_begin(bs); + assert(bs->quiesce_counter > 0); tran_add(tran, &set_aio_context, state); @@ -7692,9 +7711,13 @@ static bool bdrv_change_aio_context(BlockDriverState *bs, AioContext *ctx, * * If ignore_child is not NULL, that child (and its subgraph) will not * be touched. + * + * Called with the graph lock held. + * + * Called while all bs are drained. */ -int bdrv_try_change_aio_context(BlockDriverState *bs, AioContext *ctx, - BdrvChild *ignore_child, Error **errp) +int bdrv_try_change_aio_context_locked(BlockDriverState *bs, AioContext *ctx, + BdrvChild *ignore_child, Error **errp) { Transaction *tran; GHashTable *visited; @@ -7703,9 +7726,9 @@ int bdrv_try_change_aio_context(BlockDriverState *bs, AioContext *ctx, /* * Recursion phase: go through all nodes of the graph. - * Take care of checking that all nodes support changing AioContext - * and drain them, building a linear list of callbacks to run if everything - * is successful (the transaction itself). + * Take care of checking that all nodes support changing AioContext, + * building a linear list of callbacks to run if everything is successful + * (the transaction itself). */ tran = tran_new(); visited = g_hash_table_new(NULL, NULL); @@ -7732,6 +7755,29 @@ int bdrv_try_change_aio_context(BlockDriverState *bs, AioContext *ctx, return 0; } +/* + * Change bs's and recursively all of its parents' and children's AioContext + * to the given new context, returning an error if that isn't possible. + * + * If ignore_child is not NULL, that child (and its subgraph) will not + * be touched. + */ +int bdrv_try_change_aio_context(BlockDriverState *bs, AioContext *ctx, + BdrvChild *ignore_child, Error **errp) +{ + int ret; + + GLOBAL_STATE_CODE(); + + bdrv_drain_all_begin(); + bdrv_graph_rdlock_main_loop(); + ret = bdrv_try_change_aio_context_locked(bs, ctx, ignore_child, errp); + bdrv_graph_rdunlock_main_loop(); + bdrv_drain_all_end(); + + return ret; +} + void bdrv_add_aio_context_notifier(BlockDriverState *bs, void (*attached_aio_context)(AioContext *new_context, void *opaque), void (*detach_aio_context)(void *opaque), void *opaque) @@ -8159,8 +8205,10 @@ char *bdrv_dirname(BlockDriverState *bs, Error **errp) } /* - * Hot add/remove a BDS's child. So the user can take a child offline when - * it is broken and take a new child online + * Hot add a BDS's child. Used in combination with bdrv_del_child, so the user + * can take a child offline when it is broken and take a new child online. + * + * All block nodes must be drained. */ void bdrv_add_child(BlockDriverState *parent_bs, BlockDriverState *child_bs, Error **errp) @@ -8200,6 +8248,12 @@ void bdrv_add_child(BlockDriverState *parent_bs, BlockDriverState *child_bs, parent_bs->drv->bdrv_add_child(parent_bs, child_bs, errp); } +/* + * Hot remove a BDS's child. Used in combination with bdrv_add_child, so the + * user can take a child offline when it is broken and take a new child online. + * + * All block nodes must be drained. + */ void bdrv_del_child(BlockDriverState *parent_bs, BdrvChild *child, Error **errp) { BdrvChild *tmp; |