summaryrefslogtreecommitdiffstats
path: root/block.c
diff options
context:
space:
mode:
authorStefan Hajnoczi2022-10-30 20:15:12 +0100
committerStefan Hajnoczi2022-10-30 20:15:12 +0100
commitd5ab9490cdf3ebcd97793982d4664f951816f6db (patch)
tree4d16d89f72ae91858dc0a7a89f1b7d407797364b /block.c
parentMerge tag 'misc-next-pull-request' of https://gitlab.com/berrange/qemu into s... (diff)
parentblock/block-backend: blk_set_enable_write_cache is IO_CODE (diff)
downloadqemu-d5ab9490cdf3ebcd97793982d4664f951816f6db.tar.gz
qemu-d5ab9490cdf3ebcd97793982d4664f951816f6db.tar.xz
qemu-d5ab9490cdf3ebcd97793982d4664f951816f6db.zip
Merge tag 'for-upstream' of https://repo.or.cz/qemu/kevin into staging
Block layer patches - Cleanup bs->backing and bs->file handling - Refactor bdrv_try_set_aio_context using transactions - Changes for improved coroutine_fn consistency - vhost-user-blk: fix the resize crash - io_uring: Use of io_uring_register_ring_fd() led to breakage, revert - vvfat: Fix some problems with r/w mode - Code cleanup - MAINTAINERS: Fold "Block QAPI, monitor, ..." into "Block layer core" # -----BEGIN PGP SIGNATURE----- # # iQJFBAABCAAvFiEE3D3rFZqa+V09dFb+fwmycsiPL9YFAmNazhIRHGt3b2xmQHJl # ZGhhdC5jb20ACgkQfwmycsiPL9ZyTw/8Dfck/SuxfyeLlnQItkjaV4cnqWOU8vHs # 9x0KhlptCs+HXdF/3iicpA0lHojn7mNnbdFGjPRY4E0LriQv91TQ5ycdEmrseFPf # sgeQlgdKCVU/pHjZ2wYarm2pE43Cx85a5xuufmw+7w49dNNZn14l4t+DgviuClVM # nuVaogfZFbYyetre+Qd2TgLl+gJ+0d4o7Zs5lSWLrT8t0L9AGkcWPA7Nrbl6loIE # dOautV4G7jLjuMiCeJZOGcnuRVe3gCQ5rCGBFzzH4DUtz4BmiYx4hd3LMEsP0PMM # CrsfDZS04Ztybl9M7TmJuwkAm1gx1JDMOuJuh18lbJocIOBvhkKKxY2wI5LIdZVI # ZntmU36RowkX+GGu/PYpYyMjBDClJppZCl7vnjyLYsVt6r0Vu6SmlHpJhcRYabhe # 96Kv1LXH9A6+ogKPU3Layw6JGjg01GNr1ALuT7PO3pGto/JshmOuBEJJDucoF84M # 5AfxFCohMROVldwblA6M0eKnlQBgtr5BvtgbV54BBo88VlFJgDJFQn7R09cTFUEo # UwaJoS+nIaiZ0bQQVZhZloVppUaTdVJojzfVRCZZctga96/tu1HSFnGLnbEFpUN3 # KOf+XnVNS6Ro+nPSDf9bMjbIom2JicGFfV+6yMgIoxY/d5UA2dTZfefil4TAlSod # 6PsTgg+jrm8= # =/Fw0 # -----END PGP SIGNATURE----- # gpg: Signature made Thu 27 Oct 2022 14:29:38 EDT # gpg: using RSA key DC3DEB159A9AF95D3D7456FE7F09B272C88F2FD6 # gpg: issuer "kwolf@redhat.com" # gpg: Good signature from "Kevin Wolf <kwolf@redhat.com>" [full] # Primary key fingerprint: DC3D EB15 9A9A F95D 3D74 56FE 7F09 B272 C88F 2FD6 * tag 'for-upstream' of https://repo.or.cz/qemu/kevin: (58 commits) block/block-backend: blk_set_enable_write_cache is IO_CODE monitor: switch to *_co_* functions vmdk: switch to *_co_* functions vhdx: switch to *_co_* functions vdi: switch to *_co_* functions qed: switch to *_co_* functions qcow2: switch to *_co_* functions qcow: switch to *_co_* functions parallels: switch to *_co_* functions mirror: switch to *_co_* functions block: switch to *_co_* functions commit: switch to *_co_* functions vmdk: manually add more coroutine_fn annotations qcow2: manually add more coroutine_fn annotations qcow: manually add more coroutine_fn annotations blkdebug: add missing coroutine_fn annotation for indirect-called functions qcow2: add coroutine_fn annotation for indirect-called functions block: add missing coroutine_fn annotation to BlockDriverState callbacks coroutine-io: add missing coroutine_fn annotation to prototypes coroutine-lock: add missing coroutine_fn annotation to prototypes ... Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Diffstat (limited to 'block.c')
-rw-r--r--block.c855
1 files changed, 380 insertions, 475 deletions
diff --git a/block.c b/block.c
index c69be2cfe3..3bd594eb2a 100644
--- a/block.c
+++ b/block.c
@@ -90,13 +90,9 @@ static BlockDriverState *bdrv_open_inherit(const char *filename,
static bool bdrv_recurse_has_child(BlockDriverState *bs,
BlockDriverState *child);
-static void bdrv_child_free(BdrvChild *child);
-static void bdrv_replace_child_noperm(BdrvChild **child,
- BlockDriverState *new_bs,
- bool free_empty_child);
-static void bdrv_remove_file_or_backing_child(BlockDriverState *bs,
- BdrvChild *child,
- Transaction *tran);
+static void bdrv_replace_child_noperm(BdrvChild *child,
+ BlockDriverState *new_bs);
+static void bdrv_remove_child(BdrvChild *child, Transaction *tran);
static void bdrv_remove_filter_or_cow_child(BlockDriverState *bs,
Transaction *tran);
@@ -108,6 +104,10 @@ 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);
+
/* If non-zero, use only whitelisted block drivers */
static int use_bdrv_whitelist;
@@ -643,7 +643,7 @@ create_file_fallback_zero_first_sector(BlockBackend *blk,
bytes_to_clear = MIN(current_size, BDRV_SECTOR_SIZE);
if (bytes_to_clear) {
- ret = blk_pwrite_zeroes(blk, 0, bytes_to_clear, BDRV_REQ_MAY_UNMAP);
+ ret = blk_co_pwrite_zeroes(blk, 0, bytes_to_clear, BDRV_REQ_MAY_UNMAP);
if (ret < 0) {
error_setg_errno(errp, -ret,
"Failed to clear the new image's first sector");
@@ -861,38 +861,42 @@ int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo)
/*
* Create a uniquely-named empty temporary file.
- * Return 0 upon success, otherwise a negative errno value.
+ * Return the actual file name used upon success, otherwise NULL.
+ * This string should be freed with g_free() when not needed any longer.
+ *
+ * Note: creating a temporary file for the caller to (re)open is
+ * inherently racy. Use g_file_open_tmp() instead whenever practical.
*/
-int get_tmp_filename(char *filename, int size)
+char *create_tmp_file(Error **errp)
{
-#ifdef _WIN32
- char temp_dir[MAX_PATH];
- /* GetTempFileName requires that its output buffer (4th param)
- have length MAX_PATH or greater. */
- assert(size >= MAX_PATH);
- return (GetTempPath(MAX_PATH, temp_dir)
- && GetTempFileName(temp_dir, "qem", 0, filename)
- ? 0 : -GetLastError());
-#else
int fd;
const char *tmpdir;
- tmpdir = getenv("TMPDIR");
- if (!tmpdir) {
+ g_autofree char *filename = NULL;
+
+ tmpdir = g_get_tmp_dir();
+#ifndef _WIN32
+ /*
+ * See commit 69bef79 ("block: use /var/tmp instead of /tmp for -snapshot")
+ *
+ * This function is used to create temporary disk images (like -snapshot),
+ * so the files can become very large. /tmp is often a tmpfs where as
+ * /var/tmp is usually on a disk, so more appropriate for disk images.
+ */
+ if (!g_strcmp0(tmpdir, "/tmp")) {
tmpdir = "/var/tmp";
}
- if (snprintf(filename, size, "%s/vl.XXXXXX", tmpdir) >= size) {
- return -EOVERFLOW;
- }
- fd = mkstemp(filename);
+#endif
+
+ filename = g_strdup_printf("%s/vl.XXXXXX", tmpdir);
+ fd = g_mkstemp(filename);
if (fd < 0) {
- return -errno;
- }
- if (close(fd) != 0) {
- unlink(filename);
- return -errno;
+ error_setg_errno(errp, errno, "Could not open temporary file '%s'",
+ filename);
+ return NULL;
}
- return 0;
-#endif
+ close(fd);
+
+ return g_steal_pointer(&filename);
}
/*
@@ -1238,18 +1242,12 @@ static int bdrv_child_cb_inactivate(BdrvChild *child)
return 0;
}
-static bool bdrv_child_cb_can_set_aio_ctx(BdrvChild *child, AioContext *ctx,
- GSList **ignore, Error **errp)
-{
- BlockDriverState *bs = child->opaque;
- return bdrv_can_set_aio_context(bs, ctx, ignore, errp);
-}
-
-static void bdrv_child_cb_set_aio_ctx(BdrvChild *child, AioContext *ctx,
- GSList **ignore)
+static bool bdrv_child_cb_change_aio_ctx(BdrvChild *child, AioContext *ctx,
+ GHashTable *visited, Transaction *tran,
+ Error **errp)
{
BlockDriverState *bs = child->opaque;
- return bdrv_set_aio_context_ignore(bs, ctx, ignore);
+ return bdrv_change_aio_context(bs, ctx, visited, tran, errp);
}
/*
@@ -1443,9 +1441,39 @@ static void bdrv_child_cb_attach(BdrvChild *child)
assert_bdrv_graph_writable(bs);
QLIST_INSERT_HEAD(&bs->children, child, next);
-
- if (child->role & BDRV_CHILD_COW) {
+ if (bs->drv->is_filter || (child->role & BDRV_CHILD_FILTERED)) {
+ /*
+ * Here we handle filters and block/raw-format.c when it behave like
+ * filter. They generally have a single PRIMARY child, which is also the
+ * FILTERED child, and that they may have multiple more children, which
+ * are neither PRIMARY nor FILTERED. And never we have a COW child here.
+ * So bs->file will be the PRIMARY child, unless the PRIMARY child goes
+ * into bs->backing on exceptional cases; and bs->backing will be
+ * nothing else.
+ */
+ assert(!(child->role & BDRV_CHILD_COW));
+ if (child->role & BDRV_CHILD_PRIMARY) {
+ assert(child->role & BDRV_CHILD_FILTERED);
+ assert(!bs->backing);
+ assert(!bs->file);
+
+ if (bs->drv->filtered_child_is_backing) {
+ bs->backing = child;
+ } else {
+ bs->file = child;
+ }
+ } else {
+ assert(!(child->role & BDRV_CHILD_FILTERED));
+ }
+ } else if (child->role & BDRV_CHILD_COW) {
+ assert(bs->drv->supports_backing);
+ assert(!(child->role & BDRV_CHILD_PRIMARY));
+ assert(!bs->backing);
+ bs->backing = child;
bdrv_backing_attach(child);
+ } else if (child->role & BDRV_CHILD_PRIMARY) {
+ assert(!bs->file);
+ bs->file = child;
}
bdrv_apply_subtree_drain(child, bs);
@@ -1463,6 +1491,12 @@ static void bdrv_child_cb_detach(BdrvChild *child)
assert_bdrv_graph_writable(bs);
QLIST_REMOVE(child, next);
+ if (child == bs->backing) {
+ assert(child != bs->file);
+ bs->backing = NULL;
+ } else if (child == bs->file) {
+ bs->file = NULL;
+ }
}
static int bdrv_child_cb_update_filename(BdrvChild *c, BlockDriverState *base,
@@ -1492,8 +1526,7 @@ const BdrvChildClass child_of_bds = {
.attach = bdrv_child_cb_attach,
.detach = bdrv_child_cb_detach,
.inactivate = bdrv_child_cb_inactivate,
- .can_set_aio_ctx = bdrv_child_cb_can_set_aio_ctx,
- .set_aio_ctx = bdrv_child_cb_set_aio_ctx,
+ .change_aio_ctx = bdrv_child_cb_change_aio_ctx,
.update_filename = bdrv_child_cb_update_filename,
.get_parent_aio_context = child_of_bds_get_parent_aio_context,
};
@@ -1682,7 +1715,7 @@ open_failed:
bs->drv = NULL;
if (bs->file != NULL) {
bdrv_unref_child(bs, bs->file);
- bs->file = NULL;
+ assert(!bs->file);
}
g_free(bs->opaque);
bs->opaque = NULL;
@@ -2353,9 +2386,7 @@ static int bdrv_drv_set_perm(BlockDriverState *bs, uint64_t perm,
typedef struct BdrvReplaceChildState {
BdrvChild *child;
- BdrvChild **childp;
BlockDriverState *old_bs;
- bool free_empty_child;
} BdrvReplaceChildState;
static void bdrv_replace_child_commit(void *opaque)
@@ -2363,9 +2394,6 @@ static void bdrv_replace_child_commit(void *opaque)
BdrvReplaceChildState *s = opaque;
GLOBAL_STATE_CODE();
- if (s->free_empty_child && !s->child->bs) {
- bdrv_child_free(s->child);
- }
bdrv_unref(s->old_bs);
}
@@ -2375,34 +2403,8 @@ static void bdrv_replace_child_abort(void *opaque)
BlockDriverState *new_bs = s->child->bs;
GLOBAL_STATE_CODE();
- /*
- * old_bs reference is transparently moved from @s to s->child.
- *
- * Pass &s->child here instead of s->childp, because:
- * (1) s->old_bs must be non-NULL, so bdrv_replace_child_noperm() will not
- * modify the BdrvChild * pointer we indirectly pass to it, i.e. it
- * will not modify s->child. From that perspective, it does not matter
- * whether we pass s->childp or &s->child.
- * (2) If new_bs is not NULL, s->childp will be NULL. We then cannot use
- * it here.
- * (3) If new_bs is NULL, *s->childp will have been NULLed by
- * bdrv_replace_child_tran()'s bdrv_replace_child_noperm() call, and we
- * must not pass a NULL *s->childp here.
- *
- * So whether new_bs was NULL or not, we cannot pass s->childp here; and in
- * any case, there is no reason to pass it anyway.
- */
- bdrv_replace_child_noperm(&s->child, s->old_bs, true);
- /*
- * The child was pre-existing, so s->old_bs must be non-NULL, and
- * s->child thus must not have been freed
- */
- assert(s->child != NULL);
- if (!new_bs) {
- /* As described above, *s->childp was cleared, so restore it */
- assert(s->childp != NULL);
- *s->childp = s->child;
- }
+ /* old_bs reference is transparently moved from @s to @s->child */
+ bdrv_replace_child_noperm(s->child, s->old_bs);
bdrv_unref(new_bs);
}
@@ -2418,46 +2420,22 @@ static TransactionActionDrv bdrv_replace_child_drv = {
* Note: real unref of old_bs is done only on commit.
*
* The function doesn't update permissions, caller is responsible for this.
- *
- * (*childp)->bs must not be NULL.
- *
- * Note that if new_bs == NULL, @childp is stored in a state object attached
- * to @tran, so that the old child can be reinstated in the abort handler.
- * Therefore, if @new_bs can be NULL, @childp must stay valid until the
- * transaction is committed or aborted.
- *
- * If @free_empty_child is true and @new_bs is NULL, the BdrvChild is
- * freed (on commit). @free_empty_child should only be false if the
- * caller will free the BDrvChild themselves (which may be important
- * if this is in turn called in another transactional context).
*/
-static void bdrv_replace_child_tran(BdrvChild **childp,
- BlockDriverState *new_bs,
- Transaction *tran,
- bool free_empty_child)
+static void bdrv_replace_child_tran(BdrvChild *child, BlockDriverState *new_bs,
+ Transaction *tran)
{
BdrvReplaceChildState *s = g_new(BdrvReplaceChildState, 1);
*s = (BdrvReplaceChildState) {
- .child = *childp,
- .childp = new_bs == NULL ? childp : NULL,
- .old_bs = (*childp)->bs,
- .free_empty_child = free_empty_child,
+ .child = child,
+ .old_bs = child->bs,
};
tran_add(tran, &bdrv_replace_child_drv, s);
- /* The abort handler relies on this */
- assert(s->old_bs != NULL);
-
if (new_bs) {
bdrv_ref(new_bs);
}
- /*
- * Pass free_empty_child=false, we will free the child (if
- * necessary) in bdrv_replace_child_commit() (if our
- * @free_empty_child parameter was true).
- */
- bdrv_replace_child_noperm(childp, new_bs, false);
- /* old_bs reference is transparently moved from *childp to @s */
+ bdrv_replace_child_noperm(child, new_bs);
+ /* old_bs reference is transparently moved from @child to @s */
}
/*
@@ -2838,24 +2816,9 @@ uint64_t bdrv_qapi_perm_to_blk_perm(BlockPermission qapi_perm)
return permissions[qapi_perm];
}
-/**
- * Replace (*childp)->bs by @new_bs.
- *
- * If @new_bs is NULL, *childp will be set to NULL, too: BDS parents
- * generally cannot handle a BdrvChild with .bs == NULL, so clearing
- * BdrvChild.bs should generally immediately be followed by the
- * BdrvChild pointer being cleared as well.
- *
- * If @free_empty_child is true and @new_bs is NULL, the BdrvChild is
- * freed. @free_empty_child should only be false if the caller will
- * free the BdrvChild themselves (this may be important in a
- * transactional context, where it may only be freed on commit).
- */
-static void bdrv_replace_child_noperm(BdrvChild **childp,
- BlockDriverState *new_bs,
- bool free_empty_child)
+static void bdrv_replace_child_noperm(BdrvChild *child,
+ BlockDriverState *new_bs)
{
- BdrvChild *child = *childp;
BlockDriverState *old_bs = child->bs;
int new_bs_quiesce_counter;
int drain_saldo;
@@ -2892,9 +2855,6 @@ static void bdrv_replace_child_noperm(BdrvChild **childp,
}
child->bs = new_bs;
- if (!new_bs) {
- *childp = NULL;
- }
if (new_bs) {
assert_bdrv_graph_writable(new_bs);
@@ -2925,10 +2885,6 @@ static void bdrv_replace_child_noperm(BdrvChild **childp,
bdrv_parent_drained_end_single(child);
drain_saldo++;
}
-
- if (free_empty_child && !child->bs) {
- bdrv_child_free(child);
- }
}
/**
@@ -2948,7 +2904,7 @@ static void bdrv_child_free(BdrvChild *child)
}
typedef struct BdrvAttachChildCommonState {
- BdrvChild **child;
+ BdrvChild *child;
AioContext *old_parent_ctx;
AioContext *old_child_ctx;
} BdrvAttachChildCommonState;
@@ -2956,39 +2912,35 @@ typedef struct BdrvAttachChildCommonState {
static void bdrv_attach_child_common_abort(void *opaque)
{
BdrvAttachChildCommonState *s = opaque;
- BdrvChild *child = *s->child;
- BlockDriverState *bs = child->bs;
+ BlockDriverState *bs = s->child->bs;
GLOBAL_STATE_CODE();
- /*
- * Pass free_empty_child=false, because we still need the child
- * for the AioContext operations on the parent below; those
- * BdrvChildClass methods all work on a BdrvChild object, so we
- * need to keep it as an empty shell (after this function, it will
- * not be attached to any parent, and it will not have a .bs).
- */
- bdrv_replace_child_noperm(s->child, NULL, false);
+ bdrv_replace_child_noperm(s->child, NULL);
if (bdrv_get_aio_context(bs) != s->old_child_ctx) {
- bdrv_try_set_aio_context(bs, s->old_child_ctx, &error_abort);
+ bdrv_try_change_aio_context(bs, s->old_child_ctx, NULL, &error_abort);
}
- if (bdrv_child_get_parent_aio_context(child) != s->old_parent_ctx) {
- GSList *ignore;
+ if (bdrv_child_get_parent_aio_context(s->child) != s->old_parent_ctx) {
+ Transaction *tran;
+ GHashTable *visited;
+ bool ret;
- /* No need to ignore `child`, because it has been detached already */
- ignore = NULL;
- child->klass->can_set_aio_ctx(child, s->old_parent_ctx, &ignore,
- &error_abort);
- g_slist_free(ignore);
+ tran = tran_new();
- ignore = NULL;
- child->klass->set_aio_ctx(child, s->old_parent_ctx, &ignore);
- g_slist_free(ignore);
+ /* No need to visit `child`, because it has been detached already */
+ visited = g_hash_table_new(NULL, NULL);
+ ret = s->child->klass->change_aio_ctx(s->child, s->old_parent_ctx,
+ visited, tran, &error_abort);
+ g_hash_table_destroy(visited);
+
+ /* transaction is supposed to always succeed */
+ assert(ret == true);
+ tran_commit(tran);
}
bdrv_unref(bs);
- bdrv_child_free(child);
+ bdrv_child_free(s->child);
}
static TransactionActionDrv bdrv_attach_child_common_drv = {
@@ -2999,28 +2951,22 @@ static TransactionActionDrv bdrv_attach_child_common_drv = {
/*
* Common part of attaching bdrv child to bs or to blk or to job
*
- * Resulting new child is returned through @child.
- * At start *@child must be NULL.
- * @child is saved to a new entry of @tran, so that *@child could be reverted to
- * NULL on abort(). So referenced variable must live at least until transaction
- * end.
- *
* Function doesn't update permissions, caller is responsible for this.
+ *
+ * Returns new created child.
*/
-static int bdrv_attach_child_common(BlockDriverState *child_bs,
- const char *child_name,
- const BdrvChildClass *child_class,
- BdrvChildRole child_role,
- uint64_t perm, uint64_t shared_perm,
- void *opaque, BdrvChild **child,
- Transaction *tran, Error **errp)
+static BdrvChild *bdrv_attach_child_common(BlockDriverState *child_bs,
+ const char *child_name,
+ const BdrvChildClass *child_class,
+ BdrvChildRole child_role,
+ uint64_t perm, uint64_t shared_perm,
+ void *opaque,
+ Transaction *tran, Error **errp)
{
BdrvChild *new_child;
AioContext *parent_ctx;
AioContext *child_ctx = bdrv_get_aio_context(child_bs);
- assert(child);
- assert(*child == NULL);
assert(child_class->get_parent_desc);
GLOBAL_STATE_CODE();
@@ -3043,63 +2989,57 @@ static int 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_set_aio_context(child_bs, parent_ctx, &local_err);
-
- if (ret < 0 && child_class->can_set_aio_ctx) {
- GSList *ignore = g_slist_prepend(NULL, new_child);
- if (child_class->can_set_aio_ctx(new_child, child_ctx, &ignore,
- NULL))
- {
+ int ret = bdrv_try_change_aio_context(child_bs, parent_ctx, NULL,
+ &local_err);
+
+ if (ret < 0 && child_class->change_aio_ctx) {
+ Transaction *tran = tran_new();
+ GHashTable *visited = g_hash_table_new(NULL, NULL);
+ bool ret_child;
+
+ g_hash_table_add(visited, new_child);
+ ret_child = child_class->change_aio_ctx(new_child, child_ctx,
+ visited, tran, NULL);
+ if (ret_child == true) {
error_free(local_err);
ret = 0;
- g_slist_free(ignore);
- ignore = g_slist_prepend(NULL, new_child);
- child_class->set_aio_ctx(new_child, child_ctx, &ignore);
}
- g_slist_free(ignore);
+ tran_finalize(tran, ret_child == true ? 0 : -1);
+ g_hash_table_destroy(visited);
}
if (ret < 0) {
error_propagate(errp, local_err);
bdrv_child_free(new_child);
- return ret;
+ return NULL;
}
}
bdrv_ref(child_bs);
- bdrv_replace_child_noperm(&new_child, child_bs, true);
- /* child_bs was non-NULL, so new_child must not have been freed */
- assert(new_child != NULL);
-
- *child = new_child;
+ bdrv_replace_child_noperm(new_child, child_bs);
BdrvAttachChildCommonState *s = g_new(BdrvAttachChildCommonState, 1);
*s = (BdrvAttachChildCommonState) {
- .child = child,
+ .child = new_child,
.old_parent_ctx = parent_ctx,
.old_child_ctx = child_ctx,
};
tran_add(tran, &bdrv_attach_child_common_drv, s);
- return 0;
+ return new_child;
}
/*
- * Variable referenced by @child must live at least until transaction end.
- * (see bdrv_attach_child_common() doc for details)
- *
* Function doesn't update permissions, caller is responsible for this.
*/
-static int bdrv_attach_child_noperm(BlockDriverState *parent_bs,
- BlockDriverState *child_bs,
- const char *child_name,
- const BdrvChildClass *child_class,
- BdrvChildRole child_role,
- BdrvChild **child,
- Transaction *tran,
- Error **errp)
+static BdrvChild *bdrv_attach_child_noperm(BlockDriverState *parent_bs,
+ BlockDriverState *child_bs,
+ const char *child_name,
+ const BdrvChildClass *child_class,
+ BdrvChildRole child_role,
+ Transaction *tran,
+ Error **errp)
{
- int ret;
uint64_t perm, shared_perm;
assert(parent_bs->drv);
@@ -3108,29 +3048,25 @@ static int bdrv_attach_child_noperm(BlockDriverState *parent_bs,
if (bdrv_recurse_has_child(child_bs, parent_bs)) {
error_setg(errp, "Making '%s' a %s child of '%s' would create a cycle",
child_bs->node_name, child_name, parent_bs->node_name);
- return -EINVAL;
+ return NULL;
}
bdrv_get_cumulative_perm(parent_bs, &perm, &shared_perm);
bdrv_child_perm(parent_bs, child_bs, NULL, child_role, NULL,
perm, shared_perm, &perm, &shared_perm);
- ret = bdrv_attach_child_common(child_bs, child_name, child_class,
- child_role, perm, shared_perm, parent_bs,
- child, tran, errp);
- if (ret < 0) {
- return ret;
- }
-
- return 0;
+ return bdrv_attach_child_common(child_bs, child_name, child_class,
+ child_role, perm, shared_perm, parent_bs,
+ tran, errp);
}
-static void bdrv_detach_child(BdrvChild **childp)
+static void bdrv_detach_child(BdrvChild *child)
{
- BlockDriverState *old_bs = (*childp)->bs;
+ BlockDriverState *old_bs = child->bs;
GLOBAL_STATE_CODE();
- bdrv_replace_child_noperm(childp, NULL, true);
+ bdrv_replace_child_noperm(child, NULL);
+ bdrv_child_free(child);
if (old_bs) {
/*
@@ -3144,7 +3080,7 @@ static void bdrv_detach_child(BdrvChild **childp)
* When the parent requiring a non-default AioContext is removed, the
* node moves back to the main AioContext
*/
- bdrv_try_set_aio_context(old_bs, qemu_get_aio_context(), NULL);
+ bdrv_try_change_aio_context(old_bs, qemu_get_aio_context(), NULL, NULL);
}
}
@@ -3166,15 +3102,16 @@ BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
void *opaque, Error **errp)
{
int ret;
- BdrvChild *child = NULL;
+ BdrvChild *child;
Transaction *tran = tran_new();
GLOBAL_STATE_CODE();
- ret = bdrv_attach_child_common(child_bs, child_name, child_class,
+ child = bdrv_attach_child_common(child_bs, child_name, child_class,
child_role, perm, shared_perm, opaque,
- &child, tran, errp);
- if (ret < 0) {
+ tran, errp);
+ if (!child) {
+ ret = -EINVAL;
goto out;
}
@@ -3182,11 +3119,10 @@ BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
out:
tran_finalize(tran, ret);
- /* child is unset on failure by bdrv_attach_child_common_abort() */
- assert((ret < 0) == !child);
bdrv_unref(child_bs);
- return child;
+
+ return ret < 0 ? NULL : child;
}
/*
@@ -3208,14 +3144,15 @@ BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
Error **errp)
{
int ret;
- BdrvChild *child = NULL;
+ BdrvChild *child;
Transaction *tran = tran_new();
GLOBAL_STATE_CODE();
- ret = bdrv_attach_child_noperm(parent_bs, child_bs, child_name, child_class,
- child_role, &child, tran, errp);
- if (ret < 0) {
+ child = bdrv_attach_child_noperm(parent_bs, child_bs, child_name,
+ child_class, child_role, tran, errp);
+ if (!child) {
+ ret = -EINVAL;
goto out;
}
@@ -3226,12 +3163,10 @@ BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
out:
tran_finalize(tran, ret);
- /* child is unset on failure by bdrv_attach_child_common_abort() */
- assert((ret < 0) == !child);
bdrv_unref(child_bs);
- return child;
+ return ret < 0 ? NULL : child;
}
/* Callers must ensure that child->frozen is false. */
@@ -3242,7 +3177,7 @@ void bdrv_root_unref_child(BdrvChild *child)
GLOBAL_STATE_CODE();
child_bs = child->bs;
- bdrv_detach_child(&child);
+ bdrv_detach_child(child);
bdrv_unref(child_bs);
}
@@ -3373,7 +3308,6 @@ static int bdrv_set_file_or_backing_noperm(BlockDriverState *parent_bs,
bool is_backing,
Transaction *tran, Error **errp)
{
- int ret = 0;
bool update_inherits_from =
bdrv_inherits_from_recursive(child_bs, parent_bs);
BdrvChild *child = is_backing ? parent_bs->backing : parent_bs->file;
@@ -3424,21 +3358,19 @@ static int bdrv_set_file_or_backing_noperm(BlockDriverState *parent_bs,
if (child) {
bdrv_unset_inherits_from(parent_bs, child, tran);
- bdrv_remove_file_or_backing_child(parent_bs, child, tran);
+ bdrv_remove_child(child, tran);
}
if (!child_bs) {
goto out;
}
- ret = bdrv_attach_child_noperm(parent_bs, child_bs,
- is_backing ? "backing" : "file",
- &child_of_bds, role,
- is_backing ? &parent_bs->backing :
- &parent_bs->file,
- tran, errp);
- if (ret < 0) {
- return ret;
+ child = bdrv_attach_child_noperm(parent_bs, child_bs,
+ is_backing ? "backing" : "file",
+ &child_of_bds, role,
+ tran, errp);
+ if (!child) {
+ return -EINVAL;
}
@@ -3684,6 +3616,29 @@ BdrvChild *bdrv_open_child(const char *filename,
}
/*
+ * Wrapper on bdrv_open_child() for most popular case: open primary child of bs.
+ */
+int bdrv_open_file_child(const char *filename,
+ QDict *options, const char *bdref_key,
+ BlockDriverState *parent, Error **errp)
+{
+ BdrvChildRole role;
+
+ /* commit_top and mirror_top don't use this function */
+ assert(!parent->drv->filtered_child_is_backing);
+ role = parent->drv->is_filter ?
+ (BDRV_CHILD_FILTERED | BDRV_CHILD_PRIMARY) : BDRV_CHILD_IMAGE;
+
+ if (!bdrv_open_child(filename, options, bdref_key, parent,
+ &child_of_bds, role, false, errp))
+ {
+ return -EINVAL;
+ }
+
+ return 0;
+}
+
+/*
* TODO Future callers may need to specify parent/child_class in order for
* option inheritance to work. Existing callers use it for the root node.
*/
@@ -3732,8 +3687,7 @@ static BlockDriverState *bdrv_append_temp_snapshot(BlockDriverState *bs,
QDict *snapshot_options,
Error **errp)
{
- /* TODO: extra byte is a hack to ensure MAX_PATH space on Windows. */
- char *tmp_filename = g_malloc0(PATH_MAX + 1);
+ g_autofree char *tmp_filename = NULL;
int64_t total_size;
QemuOpts *opts = NULL;
BlockDriverState *bs_snapshot = NULL;
@@ -3752,9 +3706,8 @@ static BlockDriverState *bdrv_append_temp_snapshot(BlockDriverState *bs,
}
/* Create the temporary image */
- ret = get_tmp_filename(tmp_filename, PATH_MAX + 1);
- if (ret < 0) {
- error_setg_errno(errp, -ret, "Could not get temporary filename");
+ tmp_filename = create_tmp_file(errp);
+ if (!tmp_filename) {
goto out;
}
@@ -3788,7 +3741,6 @@ static BlockDriverState *bdrv_append_temp_snapshot(BlockDriverState *bs,
out:
qobject_unref(snapshot_options);
- g_free(tmp_filename);
return bs_snapshot;
}
@@ -4955,8 +4907,8 @@ static void bdrv_close(BlockDriverState *bs)
bdrv_unref_child(bs, child);
}
- bs->backing = NULL;
- bs->file = NULL;
+ assert(!bs->backing);
+ assert(!bs->file);
g_free(bs->opaque);
bs->opaque = NULL;
qatomic_set(&bs->copy_on_read, 0);
@@ -5087,96 +5039,28 @@ static bool should_update_child(BdrvChild *c, BlockDriverState *to)
return ret;
}
-typedef struct BdrvRemoveFilterOrCowChild {
- BdrvChild *child;
- BlockDriverState *bs;
- bool is_backing;
-} BdrvRemoveFilterOrCowChild;
-
-static void bdrv_remove_filter_or_cow_child_abort(void *opaque)
+static void bdrv_remove_child_commit(void *opaque)
{
- BdrvRemoveFilterOrCowChild *s = opaque;
- BlockDriverState *parent_bs = s->child->opaque;
-
- if (s->is_backing) {
- parent_bs->backing = s->child;
- } else {
- parent_bs->file = s->child;
- }
-
- /*
- * We don't have to restore child->bs here to undo bdrv_replace_child_tran()
- * because that function is transactionable and it registered own completion
- * entries in @tran, so .abort() for bdrv_replace_child_safe() will be
- * called automatically.
- */
-}
-
-static void bdrv_remove_filter_or_cow_child_commit(void *opaque)
-{
- BdrvRemoveFilterOrCowChild *s = opaque;
GLOBAL_STATE_CODE();
- bdrv_child_free(s->child);
-}
-
-static void bdrv_remove_filter_or_cow_child_clean(void *opaque)
-{
- BdrvRemoveFilterOrCowChild *s = opaque;
-
- /* Drop the bs reference after the transaction is done */
- bdrv_unref(s->bs);
- g_free(s);
+ bdrv_child_free(opaque);
}
-static TransactionActionDrv bdrv_remove_filter_or_cow_child_drv = {
- .abort = bdrv_remove_filter_or_cow_child_abort,
- .commit = bdrv_remove_filter_or_cow_child_commit,
- .clean = bdrv_remove_filter_or_cow_child_clean,
+static TransactionActionDrv bdrv_remove_child_drv = {
+ .commit = bdrv_remove_child_commit,
};
-/*
- * A function to remove backing or file child of @bs.
- * Function doesn't update permissions, caller is responsible for this.
- */
-static void bdrv_remove_file_or_backing_child(BlockDriverState *bs,
- BdrvChild *child,
- Transaction *tran)
+/* Function doesn't update permissions, caller is responsible for this. */
+static void bdrv_remove_child(BdrvChild *child, Transaction *tran)
{
- BdrvChild **childp;
- BdrvRemoveFilterOrCowChild *s;
-
if (!child) {
return;
}
- /*
- * Keep a reference to @bs so @childp will stay valid throughout the
- * transaction (required by bdrv_replace_child_tran())
- */
- bdrv_ref(bs);
- if (child == bs->backing) {
- childp = &bs->backing;
- } else if (child == bs->file) {
- childp = &bs->file;
- } else {
- g_assert_not_reached();
- }
-
if (child->bs) {
- /*
- * Pass free_empty_child=false, we will free the child in
- * bdrv_remove_filter_or_cow_child_commit()
- */
- bdrv_replace_child_tran(childp, NULL, tran, false);
+ bdrv_replace_child_tran(child, NULL, tran);
}
- s = g_new(BdrvRemoveFilterOrCowChild, 1);
- *s = (BdrvRemoveFilterOrCowChild) {
- .child = child,
- .bs = bs,
- .is_backing = (childp == &bs->backing),
- };
- tran_add(tran, &bdrv_remove_filter_or_cow_child_drv, s);
+ tran_add(tran, &bdrv_remove_child_drv, child);
}
/*
@@ -5187,7 +5071,7 @@ static void bdrv_remove_file_or_backing_child(BlockDriverState *bs,
static void bdrv_remove_filter_or_cow_child(BlockDriverState *bs,
Transaction *tran)
{
- bdrv_remove_file_or_backing_child(bs, bdrv_filter_or_cow_child(bs), tran);
+ bdrv_remove_child(bdrv_filter_or_cow_child(bs), tran);
}
static int bdrv_replace_node_noperm(BlockDriverState *from,
@@ -5197,7 +5081,6 @@ static int bdrv_replace_node_noperm(BlockDriverState *from,
{
BdrvChild *c, *next;
- assert(to != NULL);
GLOBAL_STATE_CODE();
QLIST_FOREACH_SAFE(c, &from->parents, next_parent, next) {
@@ -5215,12 +5098,7 @@ static int bdrv_replace_node_noperm(BlockDriverState *from,
c->name, from->node_name);
return -EPERM;
}
-
- /*
- * Passing a pointer to the local variable @c is fine here, because
- * @to is not NULL, and so &c will not be attached to the transaction.
- */
- bdrv_replace_child_tran(&c, to, tran, true);
+ bdrv_replace_child_tran(c, to, tran);
}
return 0;
@@ -5235,8 +5113,6 @@ static int bdrv_replace_node_noperm(BlockDriverState *from,
*
* With @detach_subchain=true @to must be in a backing chain of @from. In this
* case backing link of the cow-parent of @to is removed.
- *
- * @to must not be NULL.
*/
static int bdrv_replace_node_common(BlockDriverState *from,
BlockDriverState *to,
@@ -5250,7 +5126,6 @@ static int bdrv_replace_node_common(BlockDriverState *from,
int ret;
GLOBAL_STATE_CODE();
- assert(to != NULL);
if (detach_subchain) {
assert(bdrv_chain_contains(from, to));
@@ -5307,9 +5182,6 @@ out:
return ret;
}
-/**
- * Replace node @from by @to (where neither may be NULL).
- */
int bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
Error **errp)
{
@@ -5342,16 +5214,18 @@ int bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
Error **errp)
{
int ret;
+ BdrvChild *child;
Transaction *tran = tran_new();
GLOBAL_STATE_CODE();
assert(!bs_new->backing);
- ret = bdrv_attach_child_noperm(bs_new, bs_top, "backing",
- &child_of_bds, bdrv_backing_role(bs_new),
- &bs_new->backing, tran, errp);
- if (ret < 0) {
+ child = bdrv_attach_child_noperm(bs_new, bs_top, "backing",
+ &child_of_bds, bdrv_backing_role(bs_new),
+ tran, errp);
+ if (!child) {
+ ret = -EINVAL;
goto out;
}
@@ -5385,9 +5259,7 @@ int bdrv_replace_child_bs(BdrvChild *child, BlockDriverState *new_bs,
bdrv_drained_begin(old_bs);
bdrv_drained_begin(new_bs);
- bdrv_replace_child_tran(&child, new_bs, tran, true);
- /* @new_bs must have been non-NULL, so @child must not have been freed */
- assert(child != NULL);
+ bdrv_replace_child_tran(child, new_bs, tran);
found = g_hash_table_new(NULL, NULL);
refresh_list = bdrv_topological_dfs(refresh_list, found, old_bs);
@@ -7299,6 +7171,7 @@ static void bdrv_detach_aio_context(BlockDriverState *bs)
if (bs->quiesce_counter) {
aio_enable_external(bs->aio_context);
}
+ assert_bdrv_graph_writable(bs);
bs->aio_context = NULL;
}
@@ -7312,6 +7185,7 @@ static void bdrv_attach_aio_context(BlockDriverState *bs,
aio_disable_external(new_context);
}
+ assert_bdrv_graph_writable(bs);
bs->aio_context = new_context;
if (bs->drv && bs->drv->bdrv_attach_aio_context) {
@@ -7330,191 +7204,222 @@ static void bdrv_attach_aio_context(BlockDriverState *bs,
bs->walking_aio_notifiers = false;
}
-/*
- * Changes the AioContext used for fd handlers, timers, and BHs by this
- * BlockDriverState and all its children and parents.
- *
- * Must be called from the main AioContext.
- *
- * The caller must own the AioContext lock for the old AioContext of bs, but it
- * must not own the AioContext lock for new_context (unless new_context is the
- * same as the current context of bs).
- *
- * @ignore will accumulate all visited BdrvChild object. The caller is
- * responsible for freeing the list afterwards.
- */
-void bdrv_set_aio_context_ignore(BlockDriverState *bs,
- AioContext *new_context, GSList **ignore)
-{
- AioContext *old_context = bdrv_get_aio_context(bs);
- GSList *children_to_process = NULL;
- GSList *parents_to_process = NULL;
- GSList *entry;
- BdrvChild *child, *parent;
-
- g_assert(qemu_get_current_aio_context() == qemu_get_aio_context());
- GLOBAL_STATE_CODE();
-
- if (old_context == new_context) {
- return;
- }
-
- bdrv_drained_begin(bs);
-
- QLIST_FOREACH(child, &bs->children, next) {
- if (g_slist_find(*ignore, child)) {
- continue;
- }
- *ignore = g_slist_prepend(*ignore, child);
- children_to_process = g_slist_prepend(children_to_process, child);
- }
-
- QLIST_FOREACH(parent, &bs->parents, next_parent) {
- if (g_slist_find(*ignore, parent)) {
- continue;
- }
- *ignore = g_slist_prepend(*ignore, parent);
- parents_to_process = g_slist_prepend(parents_to_process, parent);
- }
-
- for (entry = children_to_process;
- entry != NULL;
- entry = g_slist_next(entry)) {
- child = entry->data;
- bdrv_set_aio_context_ignore(child->bs, new_context, ignore);
- }
- g_slist_free(children_to_process);
-
- for (entry = parents_to_process;
- entry != NULL;
- entry = g_slist_next(entry)) {
- parent = entry->data;
- assert(parent->klass->set_aio_ctx);
- parent->klass->set_aio_ctx(parent, new_context, ignore);
- }
- g_slist_free(parents_to_process);
-
- bdrv_detach_aio_context(bs);
-
- /* Acquire the new context, if necessary */
- if (qemu_get_aio_context() != new_context) {
- aio_context_acquire(new_context);
- }
-
- bdrv_attach_aio_context(bs, new_context);
-
- /*
- * If this function was recursively called from
- * bdrv_set_aio_context_ignore(), there may be nodes in the
- * subtree that have not yet been moved to the new AioContext.
- * Release the old one so bdrv_drained_end() can poll them.
- */
- if (qemu_get_aio_context() != old_context) {
- aio_context_release(old_context);
- }
-
- bdrv_drained_end(bs);
-
- if (qemu_get_aio_context() != old_context) {
- aio_context_acquire(old_context);
- }
- if (qemu_get_aio_context() != new_context) {
- aio_context_release(new_context);
- }
-}
+typedef struct BdrvStateSetAioContext {
+ AioContext *new_ctx;
+ BlockDriverState *bs;
+} BdrvStateSetAioContext;
-static bool bdrv_parent_can_set_aio_context(BdrvChild *c, AioContext *ctx,
- GSList **ignore, Error **errp)
+static bool bdrv_parent_change_aio_context(BdrvChild *c, AioContext *ctx,
+ GHashTable *visited,
+ Transaction *tran,
+ Error **errp)
{
GLOBAL_STATE_CODE();
- if (g_slist_find(*ignore, c)) {
+ if (g_hash_table_contains(visited, c)) {
return true;
}
- *ignore = g_slist_prepend(*ignore, c);
+ g_hash_table_add(visited, c);
/*
* A BdrvChildClass that doesn't handle AioContext changes cannot
* tolerate any AioContext changes
*/
- if (!c->klass->can_set_aio_ctx) {
+ if (!c->klass->change_aio_ctx) {
char *user = bdrv_child_user_desc(c);
error_setg(errp, "Changing iothreads is not supported by %s", user);
g_free(user);
return false;
}
- if (!c->klass->can_set_aio_ctx(c, ctx, ignore, errp)) {
+ if (!c->klass->change_aio_ctx(c, ctx, visited, tran, errp)) {
assert(!errp || *errp);
return false;
}
return true;
}
-bool bdrv_child_can_set_aio_context(BdrvChild *c, AioContext *ctx,
- GSList **ignore, Error **errp)
+bool bdrv_child_change_aio_context(BdrvChild *c, AioContext *ctx,
+ GHashTable *visited, Transaction *tran,
+ Error **errp)
{
GLOBAL_STATE_CODE();
- if (g_slist_find(*ignore, c)) {
+ if (g_hash_table_contains(visited, c)) {
return true;
}
- *ignore = g_slist_prepend(*ignore, c);
- return bdrv_can_set_aio_context(c->bs, ctx, ignore, errp);
+ g_hash_table_add(visited, c);
+ return bdrv_change_aio_context(c->bs, ctx, visited, tran, errp);
+}
+
+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);
+}
+
+static void bdrv_set_aio_context_commit(void *opaque)
+{
+ BdrvStateSetAioContext *state = (BdrvStateSetAioContext *) opaque;
+ BlockDriverState *bs = (BlockDriverState *) state->bs;
+ AioContext *new_context = state->new_ctx;
+ AioContext *old_context = bdrv_get_aio_context(bs);
+ assert_bdrv_graph_writable(bs);
+
+ /*
+ * Take the old AioContex when detaching it from bs.
+ * At this point, new_context lock is already acquired, and we are now
+ * also taking old_context. This is safe as long as bdrv_detach_aio_context
+ * does not call AIO_POLL_WHILE().
+ */
+ if (old_context != qemu_get_aio_context()) {
+ aio_context_acquire(old_context);
+ }
+ bdrv_detach_aio_context(bs);
+ if (old_context != qemu_get_aio_context()) {
+ aio_context_release(old_context);
+ }
+ bdrv_attach_aio_context(bs, new_context);
}
-/* @ignore will accumulate all visited BdrvChild object. The caller is
- * responsible for freeing the list afterwards. */
-bool bdrv_can_set_aio_context(BlockDriverState *bs, AioContext *ctx,
- GSList **ignore, Error **errp)
+static TransactionActionDrv set_aio_context = {
+ .commit = bdrv_set_aio_context_commit,
+ .clean = bdrv_set_aio_context_clean,
+};
+
+/*
+ * Changes the AioContext used for fd handlers, timers, and BHs by this
+ * BlockDriverState and all its children and parents.
+ *
+ * Must be called from the main AioContext.
+ *
+ * The caller must own the AioContext lock for the old AioContext of bs, but it
+ * must not own the AioContext lock for new_context (unless new_context is the
+ * same as the current context of bs).
+ *
+ * @visited will accumulate all visited BdrvChild objects. The caller is
+ * responsible for freeing the list afterwards.
+ */
+static bool bdrv_change_aio_context(BlockDriverState *bs, AioContext *ctx,
+ GHashTable *visited, Transaction *tran,
+ Error **errp)
{
BdrvChild *c;
+ BdrvStateSetAioContext *state;
+
+ GLOBAL_STATE_CODE();
if (bdrv_get_aio_context(bs) == ctx) {
return true;
}
- GLOBAL_STATE_CODE();
-
QLIST_FOREACH(c, &bs->parents, next_parent) {
- if (!bdrv_parent_can_set_aio_context(c, ctx, ignore, errp)) {
+ if (!bdrv_parent_change_aio_context(c, ctx, visited, tran, errp)) {
return false;
}
}
+
QLIST_FOREACH(c, &bs->children, next) {
- if (!bdrv_child_can_set_aio_context(c, ctx, ignore, errp)) {
+ if (!bdrv_child_change_aio_context(c, ctx, visited, tran, errp)) {
return false;
}
}
+ state = g_new(BdrvStateSetAioContext, 1);
+ *state = (BdrvStateSetAioContext) {
+ .new_ctx = ctx,
+ .bs = bs,
+ };
+
+ /* Paired with bdrv_drained_end in bdrv_set_aio_context_clean() */
+ bdrv_drained_begin(bs);
+
+ tran_add(tran, &set_aio_context, state);
+
return true;
}
-int bdrv_child_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
- BdrvChild *ignore_child, Error **errp)
+/*
+ * 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.
+ *
+ * This function still requires the caller to take the bs current
+ * AioContext lock, otherwise draining will fail since AIO_WAIT_WHILE
+ * assumes the lock is always held if bs is in another AioContext.
+ * For the same reason, it temporarily also holds the new AioContext, since
+ * bdrv_drained_end calls BDRV_POLL_WHILE that assumes the lock is taken too.
+ * Therefore the new AioContext lock must not be taken by the caller.
+ */
+int bdrv_try_change_aio_context(BlockDriverState *bs, AioContext *ctx,
+ BdrvChild *ignore_child, Error **errp)
{
- GSList *ignore;
- bool ret;
-
+ Transaction *tran;
+ GHashTable *visited;
+ int ret;
+ AioContext *old_context = bdrv_get_aio_context(bs);
GLOBAL_STATE_CODE();
- ignore = ignore_child ? g_slist_prepend(NULL, ignore_child) : NULL;
- ret = bdrv_can_set_aio_context(bs, ctx, &ignore, errp);
- g_slist_free(ignore);
+ /*
+ * Recursion phase: go through all nodes of the graph.
+ * Take care of checking that all nodes support changing AioContext
+ * and drain them, builing a linear list of callbacks to run if everything
+ * is successful (the transaction itself).
+ */
+ tran = tran_new();
+ visited = g_hash_table_new(NULL, NULL);
+ if (ignore_child) {
+ g_hash_table_add(visited, ignore_child);
+ }
+ ret = bdrv_change_aio_context(bs, ctx, visited, tran, errp);
+ g_hash_table_destroy(visited);
+
+ /*
+ * Linear phase: go through all callbacks collected in the transaction.
+ * Run all callbacks collected in the recursion to switch all nodes
+ * AioContext lock (transaction commit), or undo all changes done in the
+ * recursion (transaction abort).
+ */
if (!ret) {
+ /* Just run clean() callbacks. No AioContext changed. */
+ tran_abort(tran);
return -EPERM;
}
- ignore = ignore_child ? g_slist_prepend(NULL, ignore_child) : NULL;
- bdrv_set_aio_context_ignore(bs, ctx, &ignore);
- g_slist_free(ignore);
+ /*
+ * Release old AioContext, it won't be needed anymore, as all
+ * bdrv_drained_begin() have been called already.
+ */
+ if (qemu_get_aio_context() != old_context) {
+ aio_context_release(old_context);
+ }
+
+ /*
+ * Acquire new AioContext since bdrv_drained_end() is going to be called
+ * after we switched all nodes in the new AioContext, and the function
+ * assumes that the lock of the bs is always taken.
+ */
+ if (qemu_get_aio_context() != ctx) {
+ aio_context_acquire(ctx);
+ }
- return 0;
-}
+ tran_commit(tran);
-int bdrv_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
- Error **errp)
-{
- GLOBAL_STATE_CODE();
- return bdrv_child_try_set_aio_context(bs, ctx, NULL, errp);
+ if (qemu_get_aio_context() != ctx) {
+ aio_context_release(ctx);
+ }
+
+ /* Re-acquire the old AioContext, since the caller takes and releases it. */
+ if (qemu_get_aio_context() != old_context) {
+ aio_context_acquire(old_context);
+ }
+
+ return 0;
}
void bdrv_add_aio_context_notifier(BlockDriverState *bs,