From f871abd60f4b67547e62c57c9bec19420052be39 Mon Sep 17 00:00:00 2001 From: Kevin Wolf Date: Tue, 21 May 2019 19:00:25 +0200 Subject: block: Drain source node in bdrv_replace_node() Instead of just asserting that no requests are in flight in bdrv_replace_node(), which is a requirement that most callers ignore, we can just drain the source node right there. This fixes at least starting a commit job while I/O is active on the backing chain, but probably other callers, too. Having requests in flight on the target node isn't a problem because the target just gets new parents, but the call path of running requests isn't modified. So we can just drop this assertion without a replacement. Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1711643 Signed-off-by: Kevin Wolf Reviewed-by: Max Reitz --- block.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'block.c') diff --git a/block.c b/block.c index 1a73e310c1..4c3902508d 100644 --- a/block.c +++ b/block.c @@ -4017,13 +4017,13 @@ void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to, uint64_t perm = 0, shared = BLK_PERM_ALL; int ret; - assert(!atomic_read(&from->in_flight)); - assert(!atomic_read(&to->in_flight)); - /* Make sure that @from doesn't go away until we have successfully attached * all of its parents to @to. */ bdrv_ref(from); + assert(qemu_get_current_aio_context() == qemu_get_aio_context()); + bdrv_drained_begin(from); + /* Put all parents into @list and calculate their cumulative permissions */ QLIST_FOREACH_SAFE(c, &from->parents, next_parent, next) { assert(c->bs == from); @@ -4064,6 +4064,7 @@ void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to, out: g_slist_free(list); + bdrv_drained_end(from); bdrv_unref(from); } -- cgit 1.4.1 From d861ab3acf8dcf817e0c2335979b258847b69564 Mon Sep 17 00:00:00 2001 From: Kevin Wolf Date: Thu, 25 Apr 2019 14:25:10 +0200 Subject: block: Add BlockBackend.ctx This adds a new parameter to blk_new() which requires its callers to declare from which AioContext this BlockBackend is going to be used (or the locks of which AioContext need to be taken anyway). The given context is only stored and kept up to date when changing AioContexts. Actually applying the stored AioContext to the root node is saved for another commit. Signed-off-by: Kevin Wolf --- block.c | 2 +- block/backup.c | 3 ++- block/block-backend.c | 18 +++++++++++++++--- block/commit.c | 11 +++++++---- block/crypto.c | 3 ++- block/mirror.c | 3 ++- block/parallels.c | 3 ++- block/qcow.c | 3 ++- block/qcow2.c | 6 ++++-- block/qed.c | 3 ++- block/sheepdog.c | 3 ++- block/vdi.c | 3 ++- block/vhdx.c | 3 ++- block/vmdk.c | 3 ++- block/vpc.c | 3 ++- blockdev.c | 4 ++-- blockjob.c | 2 +- hmp.c | 3 ++- hw/block/fdc.c | 2 +- hw/block/xen-block.c | 2 +- hw/core/qdev-properties-system.c | 4 +++- hw/ide/qdev.c | 2 +- hw/scsi/scsi-disk.c | 2 +- include/sysemu/block-backend.h | 2 +- migration/block.c | 3 ++- nbd/server.c | 5 +++-- qemu-img.c | 6 ++++-- tests/test-bdrv-drain.c | 30 +++++++++++++++--------------- tests/test-bdrv-graph-mod.c | 5 +++-- tests/test-block-backend.c | 6 ++++-- tests/test-block-iothread.c | 10 +++++----- tests/test-blockjob.c | 2 +- tests/test-throttle.c | 6 +++--- 33 files changed, 102 insertions(+), 64 deletions(-) (limited to 'block.c') diff --git a/block.c b/block.c index 4c3902508d..69ceac6377 100644 --- a/block.c +++ b/block.c @@ -2884,7 +2884,7 @@ static BlockDriverState *bdrv_open_inherit(const char *filename, /* Not requesting BLK_PERM_CONSISTENT_READ because we're only * looking at the header to guess the image format. This works even * in cases where a guest would not see a consistent state. */ - file = blk_new(0, BLK_PERM_ALL); + file = blk_new(bdrv_get_aio_context(file_bs), 0, BLK_PERM_ALL); blk_insert_bs(file, file_bs, &local_err); bdrv_unref(file_bs); if (local_err) { diff --git a/block/backup.c b/block/backup.c index 00f4f8af53..715e1d3be8 100644 --- a/block/backup.c +++ b/block/backup.c @@ -627,7 +627,8 @@ BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs, } /* The target must match the source in size, so no resize here either */ - job->target = blk_new(BLK_PERM_WRITE, + job->target = blk_new(job->common.job.aio_context, + BLK_PERM_WRITE, BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE | BLK_PERM_WRITE_UNCHANGED | BLK_PERM_GRAPH_MOD); ret = blk_insert_bs(job->target, target, errp); diff --git a/block/block-backend.c b/block/block-backend.c index 390fde6f71..f03f14acb0 100644 --- a/block/block-backend.c +++ b/block/block-backend.c @@ -42,6 +42,7 @@ struct BlockBackend { char *name; int refcnt; BdrvChild *root; + AioContext *ctx; DriveInfo *legacy_dinfo; /* null unless created by drive_new() */ QTAILQ_ENTRY(BlockBackend) link; /* for block_backends */ QTAILQ_ENTRY(BlockBackend) monitor_link; /* for monitor_block_backends */ @@ -322,12 +323,13 @@ static const BdrvChildRole child_root = { * * Return the new BlockBackend on success, null on failure. */ -BlockBackend *blk_new(uint64_t perm, uint64_t shared_perm) +BlockBackend *blk_new(AioContext *ctx, uint64_t perm, uint64_t shared_perm) { BlockBackend *blk; blk = g_new0(BlockBackend, 1); blk->refcnt = 1; + blk->ctx = ctx; blk->perm = perm; blk->shared_perm = shared_perm; blk_set_enable_write_cache(blk, true); @@ -347,6 +349,7 @@ BlockBackend *blk_new(uint64_t perm, uint64_t shared_perm) /* * Creates a new BlockBackend, opens a new BlockDriverState, and connects both. + * The new BlockBackend is in the main AioContext. * * Just as with bdrv_open(), after having called this function the reference to * @options belongs to the block layer (even on failure). @@ -382,7 +385,7 @@ BlockBackend *blk_new_open(const char *filename, const char *reference, perm |= BLK_PERM_RESIZE; } - blk = blk_new(perm, BLK_PERM_ALL); + blk = blk_new(qemu_get_aio_context(), perm, BLK_PERM_ALL); bs = bdrv_open(filename, reference, options, flags, errp); if (!bs) { blk_unref(blk); @@ -1856,7 +1859,15 @@ void blk_op_unblock_all(BlockBackend *blk, Error *reason) AioContext *blk_get_aio_context(BlockBackend *blk) { - return bdrv_get_aio_context(blk_bs(blk)); + BlockDriverState *bs = blk_bs(blk); + + /* FIXME The AioContext of bs and blk can be inconsistent. For the moment, + * we prefer the one of bs for compatibility. */ + if (bs) { + return bdrv_get_aio_context(blk_bs(blk)); + } + + return blk->ctx; } static AioContext *blk_aiocb_get_aio_context(BlockAIOCB *acb) @@ -1888,6 +1899,7 @@ static int blk_do_set_aio_context(BlockBackend *blk, AioContext *new_context, } } + blk->ctx = new_context; return 0; } diff --git a/block/commit.c b/block/commit.c index 14e5bb394c..4d519506d6 100644 --- a/block/commit.c +++ b/block/commit.c @@ -338,7 +338,8 @@ void commit_start(const char *job_id, BlockDriverState *bs, goto fail; } - s->base = blk_new(BLK_PERM_CONSISTENT_READ + s->base = blk_new(s->common.job.aio_context, + BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_CONSISTENT_READ @@ -351,7 +352,7 @@ void commit_start(const char *job_id, BlockDriverState *bs, s->base_bs = base; /* Required permissions are already taken with block_job_add_bdrv() */ - s->top = blk_new(0, BLK_PERM_ALL); + s->top = blk_new(s->common.job.aio_context, 0, BLK_PERM_ALL); ret = blk_insert_bs(s->top, top, errp); if (ret < 0) { goto fail; @@ -395,6 +396,7 @@ int bdrv_commit(BlockDriverState *bs) BlockDriverState *backing_file_bs = NULL; BlockDriverState *commit_top_bs = NULL; BlockDriver *drv = bs->drv; + AioContext *ctx; int64_t offset, length, backing_length; int ro; int64_t n; @@ -422,8 +424,9 @@ int bdrv_commit(BlockDriverState *bs) } } - src = blk_new(BLK_PERM_CONSISTENT_READ, BLK_PERM_ALL); - backing = blk_new(BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL); + ctx = bdrv_get_aio_context(bs); + src = blk_new(ctx, BLK_PERM_CONSISTENT_READ, BLK_PERM_ALL); + backing = blk_new(ctx, BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL); ret = blk_insert_bs(src, bs, &local_err); if (ret < 0) { diff --git a/block/crypto.c b/block/crypto.c index 3af46b805f..7351fd479d 100644 --- a/block/crypto.c +++ b/block/crypto.c @@ -257,7 +257,8 @@ static int block_crypto_co_create_generic(BlockDriverState *bs, QCryptoBlock *crypto = NULL; struct BlockCryptoCreateData data; - blk = blk_new(BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL); + blk = blk_new(bdrv_get_aio_context(bs), + BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL); ret = blk_insert_bs(blk, bs, errp); if (ret < 0) { diff --git a/block/mirror.c b/block/mirror.c index ec4bd9f404..eb96b52de9 100644 --- a/block/mirror.c +++ b/block/mirror.c @@ -1584,7 +1584,8 @@ static void mirror_start_job(const char *job_id, BlockDriverState *bs, * We can allow anything except resize there.*/ target_is_backing = bdrv_chain_contains(bs, target); target_graph_mod = (backing_mode != MIRROR_LEAVE_BACKING_CHAIN); - s->target = blk_new(BLK_PERM_WRITE | BLK_PERM_RESIZE | + s->target = blk_new(s->common.job.aio_context, + BLK_PERM_WRITE | BLK_PERM_RESIZE | (target_graph_mod ? BLK_PERM_GRAPH_MOD : 0), BLK_PERM_WRITE_UNCHANGED | (target_is_backing ? BLK_PERM_CONSISTENT_READ | diff --git a/block/parallels.c b/block/parallels.c index 2747400577..00fae125d1 100644 --- a/block/parallels.c +++ b/block/parallels.c @@ -554,7 +554,8 @@ static int coroutine_fn parallels_co_create(BlockdevCreateOptions* opts, return -EIO; } - blk = blk_new(BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL); + blk = blk_new(bdrv_get_aio_context(bs), + BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL); ret = blk_insert_bs(blk, bs, errp); if (ret < 0) { goto out; diff --git a/block/qcow.c b/block/qcow.c index 1bb8fd05e2..6dee5bb792 100644 --- a/block/qcow.c +++ b/block/qcow.c @@ -844,7 +844,8 @@ static int coroutine_fn qcow_co_create(BlockdevCreateOptions *opts, return -EIO; } - qcow_blk = blk_new(BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL); + qcow_blk = blk_new(bdrv_get_aio_context(bs), + BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL); ret = blk_insert_bs(qcow_blk, bs, errp); if (ret < 0) { goto exit; diff --git a/block/qcow2.c b/block/qcow2.c index 14f914117f..9396d490d5 100644 --- a/block/qcow2.c +++ b/block/qcow2.c @@ -3174,7 +3174,8 @@ qcow2_co_create(BlockdevCreateOptions *create_options, Error **errp) } /* Create BlockBackend to write to the image */ - blk = blk_new(BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL); + blk = blk_new(bdrv_get_aio_context(bs), + BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL); ret = blk_insert_bs(blk, bs, errp); if (ret < 0) { goto out; @@ -5006,7 +5007,8 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts, } if (new_size) { - BlockBackend *blk = blk_new(BLK_PERM_RESIZE, BLK_PERM_ALL); + BlockBackend *blk = blk_new(bdrv_get_aio_context(bs), + BLK_PERM_RESIZE, BLK_PERM_ALL); ret = blk_insert_bs(blk, bs, errp); if (ret < 0) { blk_unref(blk); diff --git a/block/qed.c b/block/qed.c index dcdcd62b4a..bb4f5c9863 100644 --- a/block/qed.c +++ b/block/qed.c @@ -649,7 +649,8 @@ static int coroutine_fn bdrv_qed_co_create(BlockdevCreateOptions *opts, return -EIO; } - blk = blk_new(BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL); + blk = blk_new(bdrv_get_aio_context(bs), + BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL); ret = blk_insert_bs(blk, bs, errp); if (ret < 0) { goto out; diff --git a/block/sheepdog.c b/block/sheepdog.c index cbdfe9ab6e..f76d6ddbbc 100644 --- a/block/sheepdog.c +++ b/block/sheepdog.c @@ -1800,7 +1800,8 @@ static int sd_prealloc(BlockDriverState *bs, int64_t old_size, int64_t new_size, void *buf = NULL; int ret; - blk = blk_new(BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE | BLK_PERM_RESIZE, + blk = blk_new(bdrv_get_aio_context(bs), + BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL); ret = blk_insert_bs(blk, bs, errp); diff --git a/block/vdi.c b/block/vdi.c index d7ef6628e7..b9845a4cbd 100644 --- a/block/vdi.c +++ b/block/vdi.c @@ -803,7 +803,8 @@ static int coroutine_fn vdi_co_do_create(BlockdevCreateOptions *create_options, goto exit; } - blk = blk_new(BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL); + blk = blk_new(bdrv_get_aio_context(bs_file), + BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL); ret = blk_insert_bs(blk, bs_file, errp); if (ret < 0) { goto exit; diff --git a/block/vhdx.c b/block/vhdx.c index a143a57657..d6070b6fa8 100644 --- a/block/vhdx.c +++ b/block/vhdx.c @@ -1900,7 +1900,8 @@ static int coroutine_fn vhdx_co_create(BlockdevCreateOptions *opts, return -EIO; } - blk = blk_new(BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL); + blk = blk_new(bdrv_get_aio_context(bs), + BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL); ret = blk_insert_bs(blk, bs, errp); if (ret < 0) { goto delete_and_exit; diff --git a/block/vmdk.c b/block/vmdk.c index de8cb859f8..51067c774f 100644 --- a/block/vmdk.c +++ b/block/vmdk.c @@ -2356,7 +2356,8 @@ static BlockBackend *vmdk_co_create_cb(int64_t size, int idx, if (!bs) { return NULL; } - blk = blk_new(BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE | BLK_PERM_RESIZE, + blk = blk_new(bdrv_get_aio_context(bs), + BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL); if (blk_insert_bs(blk, bs, errp)) { bdrv_unref(bs); diff --git a/block/vpc.c b/block/vpc.c index 0c279b87c8..d4776ee8a5 100644 --- a/block/vpc.c +++ b/block/vpc.c @@ -1011,7 +1011,8 @@ static int coroutine_fn vpc_co_create(BlockdevCreateOptions *opts, return -EIO; } - blk = blk_new(BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL); + blk = blk_new(bdrv_get_aio_context(bs), + BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL); ret = blk_insert_bs(blk, bs, errp); if (ret < 0) { goto out; diff --git a/blockdev.c b/blockdev.c index d88dc115f2..e1ad92c559 100644 --- a/blockdev.c +++ b/blockdev.c @@ -574,7 +574,7 @@ static BlockBackend *blockdev_init(const char *file, QDict *bs_opts, if ((!file || !*file) && !qdict_size(bs_opts)) { BlockBackendRootState *blk_rs; - blk = blk_new(0, BLK_PERM_ALL); + blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL); blk_rs = blk_get_root_state(blk); blk_rs->open_flags = bdrv_flags; blk_rs->read_only = read_only; @@ -3154,7 +3154,7 @@ void qmp_block_resize(bool has_device, const char *device, goto out; } - blk = blk_new(BLK_PERM_RESIZE, BLK_PERM_ALL); + blk = blk_new(bdrv_get_aio_context(bs), BLK_PERM_RESIZE, BLK_PERM_ALL); ret = blk_insert_bs(blk, bs, errp); if (ret < 0) { goto out; diff --git a/blockjob.c b/blockjob.c index cc5f18e7cd..29517ae162 100644 --- a/blockjob.c +++ b/blockjob.c @@ -392,7 +392,7 @@ void *block_job_create(const char *job_id, const BlockJobDriver *driver, job_id = bdrv_get_device_name(bs); } - blk = blk_new(perm, shared_perm); + blk = blk_new(bdrv_get_aio_context(bs), perm, shared_perm); ret = blk_insert_bs(blk, bs, errp); if (ret < 0) { blk_unref(blk); diff --git a/hmp.c b/hmp.c index 56a3ed7375..be5e345c6f 100644 --- a/hmp.c +++ b/hmp.c @@ -2560,7 +2560,8 @@ void hmp_qemu_io(Monitor *mon, const QDict *qdict) if (!blk) { BlockDriverState *bs = bdrv_lookup_bs(NULL, device, &err); if (bs) { - blk = local_blk = blk_new(0, BLK_PERM_ALL); + blk = local_blk = blk_new(bdrv_get_aio_context(bs), + 0, BLK_PERM_ALL); ret = blk_insert_bs(blk, bs, &err); if (ret < 0) { goto fail; diff --git a/hw/block/fdc.c b/hw/block/fdc.c index 6f19f127a5..37ccedc9f7 100644 --- a/hw/block/fdc.c +++ b/hw/block/fdc.c @@ -538,7 +538,7 @@ static void floppy_drive_realize(DeviceState *qdev, Error **errp) if (!dev->conf.blk) { /* Anonymous BlockBackend for an empty drive */ - dev->conf.blk = blk_new(0, BLK_PERM_ALL); + dev->conf.blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL); ret = blk_attach_dev(dev->conf.blk, qdev); assert(ret == 0); } diff --git a/hw/block/xen-block.c b/hw/block/xen-block.c index ef635be4c2..31b0f5ccc8 100644 --- a/hw/block/xen-block.c +++ b/hw/block/xen-block.c @@ -609,7 +609,7 @@ static void xen_cdrom_realize(XenBlockDevice *blockdev, Error **errp) int rc; /* Set up an empty drive */ - conf->blk = blk_new(0, BLK_PERM_ALL); + conf->blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL); rc = blk_attach_dev(conf->blk, DEVICE(blockdev)); if (!rc) { diff --git a/hw/core/qdev-properties-system.c b/hw/core/qdev-properties-system.c index b45a7ef54b..42e048f190 100644 --- a/hw/core/qdev-properties-system.c +++ b/hw/core/qdev-properties-system.c @@ -80,7 +80,9 @@ static void parse_drive(DeviceState *dev, const char *str, void **ptr, if (!blk) { BlockDriverState *bs = bdrv_lookup_bs(NULL, str, NULL); if (bs) { - blk = blk_new(0, BLK_PERM_ALL); + /* BlockBackends of devices start in the main context and are only + * later moved into another context if the device supports that. */ + blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL); blk_created = true; ret = blk_insert_bs(blk, bs, errp); diff --git a/hw/ide/qdev.c b/hw/ide/qdev.c index 573b022e1e..360cd20bd8 100644 --- a/hw/ide/qdev.c +++ b/hw/ide/qdev.c @@ -168,7 +168,7 @@ static void ide_dev_initfn(IDEDevice *dev, IDEDriveKind kind, Error **errp) return; } else { /* Anonymous BlockBackend for an empty drive */ - dev->conf.blk = blk_new(0, BLK_PERM_ALL); + dev->conf.blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL); ret = blk_attach_dev(dev->conf.blk, &dev->qdev); assert(ret == 0); } diff --git a/hw/scsi/scsi-disk.c b/hw/scsi/scsi-disk.c index e7e865ab3b..91c5a8b1ac 100644 --- a/hw/scsi/scsi-disk.c +++ b/hw/scsi/scsi-disk.c @@ -2417,7 +2417,7 @@ static void scsi_cd_realize(SCSIDevice *dev, Error **errp) if (!dev->conf.blk) { /* Anonymous BlockBackend for an empty drive. As we put it into * dev->conf, qdev takes care of detaching on unplug. */ - dev->conf.blk = blk_new(0, BLK_PERM_ALL); + dev->conf.blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL); ret = blk_attach_dev(dev->conf.blk, &dev->qdev); assert(ret == 0); } diff --git a/include/sysemu/block-backend.h b/include/sysemu/block-backend.h index 228fb3fb83..733c4957eb 100644 --- a/include/sysemu/block-backend.h +++ b/include/sysemu/block-backend.h @@ -76,7 +76,7 @@ typedef struct BlockBackendPublic { ThrottleGroupMember throttle_group_member; } BlockBackendPublic; -BlockBackend *blk_new(uint64_t perm, uint64_t shared_perm); +BlockBackend *blk_new(AioContext *ctx, uint64_t perm, uint64_t shared_perm); BlockBackend *blk_new_open(const char *filename, const char *reference, QDict *options, int flags, Error **errp); int blk_get_refcnt(BlockBackend *blk); diff --git a/migration/block.c b/migration/block.c index 83c633fb3f..91f98ef44a 100644 --- a/migration/block.c +++ b/migration/block.c @@ -417,7 +417,8 @@ static int init_blk_migration(QEMUFile *f) } bmds = g_new0(BlkMigDevState, 1); - bmds->blk = blk_new(BLK_PERM_CONSISTENT_READ, BLK_PERM_ALL); + bmds->blk = blk_new(qemu_get_aio_context(), + BLK_PERM_CONSISTENT_READ, BLK_PERM_ALL); bmds->blk_name = g_strdup(bdrv_get_device_name(bs)); bmds->bulk_completed = 0; bmds->total_sectors = sectors; diff --git a/nbd/server.c b/nbd/server.c index d1375350bc..aeca3893fe 100644 --- a/nbd/server.c +++ b/nbd/server.c @@ -1484,8 +1484,9 @@ NBDExport *nbd_export_new(BlockDriverState *bs, uint64_t dev_offset, if ((nbdflags & NBD_FLAG_READ_ONLY) == 0) { perm |= BLK_PERM_WRITE; } - blk = blk_new(perm, BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE_UNCHANGED | - BLK_PERM_WRITE | BLK_PERM_GRAPH_MOD); + blk = blk_new(bdrv_get_aio_context(bs), perm, + BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE_UNCHANGED | + BLK_PERM_WRITE | BLK_PERM_GRAPH_MOD); ret = blk_insert_bs(blk, bs, errp); if (ret < 0) { goto fail; diff --git a/qemu-img.c b/qemu-img.c index b0535919b1..07b6e2a808 100644 --- a/qemu-img.c +++ b/qemu-img.c @@ -3313,7 +3313,8 @@ static int img_rebase(int argc, char **argv) BlockDriverState *base_bs = backing_bs(bs); if (base_bs) { - blk_old_backing = blk_new(BLK_PERM_CONSISTENT_READ, + blk_old_backing = blk_new(qemu_get_aio_context(), + BLK_PERM_CONSISTENT_READ, BLK_PERM_ALL); ret = blk_insert_bs(blk_old_backing, base_bs, &local_err); @@ -3360,7 +3361,8 @@ static int img_rebase(int argc, char **argv) prefix_chain_bs = bdrv_find_backing_image(bs, out_real_path); if (prefix_chain_bs) { g_free(out_real_path); - blk_new_backing = blk_new(BLK_PERM_CONSISTENT_READ, + blk_new_backing = blk_new(qemu_get_aio_context(), + BLK_PERM_CONSISTENT_READ, BLK_PERM_ALL); ret = blk_insert_bs(blk_new_backing, prefix_chain_bs, &local_err); diff --git a/tests/test-bdrv-drain.c b/tests/test-bdrv-drain.c index e86798923f..fabbd52d93 100644 --- a/tests/test-bdrv-drain.c +++ b/tests/test-bdrv-drain.c @@ -206,7 +206,7 @@ static void test_drv_cb_common(enum drain_type drain_type, bool recursive) QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, NULL, 0); - blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL); + blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); bs = bdrv_new_open_driver(&bdrv_test, "test-node", BDRV_O_RDWR, &error_abort); s = bs->opaque; @@ -290,7 +290,7 @@ static void test_quiesce_common(enum drain_type drain_type, bool recursive) BlockBackend *blk; BlockDriverState *bs, *backing; - blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL); + blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); bs = bdrv_new_open_driver(&bdrv_test, "test-node", BDRV_O_RDWR, &error_abort); blk_insert_bs(blk, bs, &error_abort); @@ -353,7 +353,7 @@ static void test_nested(void) BDRVTestState *s, *backing_s; enum drain_type outer, inner; - blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL); + blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); bs = bdrv_new_open_driver(&bdrv_test, "test-node", BDRV_O_RDWR, &error_abort); s = bs->opaque; @@ -402,13 +402,13 @@ static void test_multiparent(void) BlockDriverState *bs_a, *bs_b, *backing; BDRVTestState *a_s, *b_s, *backing_s; - blk_a = blk_new(BLK_PERM_ALL, BLK_PERM_ALL); + blk_a = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); bs_a = bdrv_new_open_driver(&bdrv_test, "test-node-a", BDRV_O_RDWR, &error_abort); a_s = bs_a->opaque; blk_insert_bs(blk_a, bs_a, &error_abort); - blk_b = blk_new(BLK_PERM_ALL, BLK_PERM_ALL); + blk_b = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); bs_b = bdrv_new_open_driver(&bdrv_test, "test-node-b", BDRV_O_RDWR, &error_abort); b_s = bs_b->opaque; @@ -475,13 +475,13 @@ static void test_graph_change_drain_subtree(void) BlockDriverState *bs_a, *bs_b, *backing; BDRVTestState *a_s, *b_s, *backing_s; - blk_a = blk_new(BLK_PERM_ALL, BLK_PERM_ALL); + blk_a = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); bs_a = bdrv_new_open_driver(&bdrv_test, "test-node-a", BDRV_O_RDWR, &error_abort); a_s = bs_a->opaque; blk_insert_bs(blk_a, bs_a, &error_abort); - blk_b = blk_new(BLK_PERM_ALL, BLK_PERM_ALL); + blk_b = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); bs_b = bdrv_new_open_driver(&bdrv_test, "test-node-b", BDRV_O_RDWR, &error_abort); b_s = bs_b->opaque; @@ -555,7 +555,7 @@ static void test_graph_change_drain_all(void) BDRVTestState *a_s, *b_s; /* Create node A with a BlockBackend */ - blk_a = blk_new(BLK_PERM_ALL, BLK_PERM_ALL); + blk_a = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); bs_a = bdrv_new_open_driver(&bdrv_test, "test-node-a", BDRV_O_RDWR, &error_abort); a_s = bs_a->opaque; @@ -571,7 +571,7 @@ static void test_graph_change_drain_all(void) g_assert_cmpint(a_s->drain_count, ==, 1); /* Create node B with a BlockBackend */ - blk_b = blk_new(BLK_PERM_ALL, BLK_PERM_ALL); + blk_b = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); bs_b = bdrv_new_open_driver(&bdrv_test, "test-node-b", BDRV_O_RDWR, &error_abort); b_s = bs_b->opaque; @@ -672,7 +672,7 @@ static void test_iothread_common(enum drain_type drain_type, int drain_thread) goto out; } - blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL); + blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); bs = bdrv_new_open_driver(&bdrv_test, "test-node", BDRV_O_RDWR, &error_abort); s = bs->opaque; @@ -883,7 +883,7 @@ static void test_blockjob_common_drain_node(enum drain_type drain_type, bdrv_set_backing_hd(src, src_backing, &error_abort); bdrv_unref(src_backing); - blk_src = blk_new(BLK_PERM_ALL, BLK_PERM_ALL); + blk_src = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); blk_insert_bs(blk_src, src_overlay, &error_abort); switch (drain_node) { @@ -910,7 +910,7 @@ static void test_blockjob_common_drain_node(enum drain_type drain_type, target = bdrv_new_open_driver(&bdrv_test, "target", BDRV_O_RDWR, &error_abort); - blk_target = blk_new(BLK_PERM_ALL, BLK_PERM_ALL); + blk_target = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); blk_insert_bs(blk_target, target, &error_abort); aio_context_acquire(ctx); @@ -1205,7 +1205,7 @@ static void do_test_delete_by_drain(bool detach_instead_of_delete, &error_abort); bdrv_attach_child(bs, null_bs, "null-child", &child_file, &error_abort); - blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL); + blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); blk_insert_bs(blk, bs, &error_abort); /* Referenced by blk now */ @@ -1368,7 +1368,7 @@ static void test_detach_indirect(bool by_parent_cb) c = bdrv_new_open_driver(&bdrv_test, "c", BDRV_O_RDWR, &error_abort); /* blk is a BB for parent-a */ - blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL); + blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); blk_insert_bs(blk, parent_a, &error_abort); bdrv_unref(parent_a); @@ -1460,7 +1460,7 @@ static void test_append_to_drained(void) BlockDriverState *base, *overlay; BDRVTestState *base_s, *overlay_s; - blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL); + blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); base = bdrv_new_open_driver(&bdrv_test, "base", BDRV_O_RDWR, &error_abort); base_s = base->opaque; blk_insert_bs(blk, base, &error_abort); diff --git a/tests/test-bdrv-graph-mod.c b/tests/test-bdrv-graph-mod.c index 747c0bf8fc..cfeec36566 100644 --- a/tests/test-bdrv-graph-mod.c +++ b/tests/test-bdrv-graph-mod.c @@ -102,7 +102,8 @@ static void test_update_perm_tree(void) { Error *local_err = NULL; - BlockBackend *root = blk_new(BLK_PERM_WRITE | BLK_PERM_CONSISTENT_READ, + BlockBackend *root = blk_new(qemu_get_aio_context(), + BLK_PERM_WRITE | BLK_PERM_CONSISTENT_READ, BLK_PERM_ALL & ~BLK_PERM_WRITE); BlockDriverState *bs = no_perm_node("node"); BlockDriverState *filter = pass_through_node("filter"); @@ -165,7 +166,7 @@ static void test_update_perm_tree(void) */ static void test_should_update_child(void) { - BlockBackend *root = blk_new(0, BLK_PERM_ALL); + BlockBackend *root = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL); BlockDriverState *bs = no_perm_node("node"); BlockDriverState *filter = no_perm_node("filter"); BlockDriverState *target = no_perm_node("target"); diff --git a/tests/test-block-backend.c b/tests/test-block-backend.c index fd59f02bd0..5b5d6845c0 100644 --- a/tests/test-block-backend.c +++ b/tests/test-block-backend.c @@ -37,7 +37,8 @@ static void test_drain_aio_error_flush_cb(void *opaque, int ret) static void test_drain_aio_error(void) { - BlockBackend *blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL); + BlockBackend *blk = blk_new(qemu_get_aio_context(), + BLK_PERM_ALL, BLK_PERM_ALL); BlockAIOCB *acb; bool completed = false; @@ -53,7 +54,8 @@ static void test_drain_aio_error(void) static void test_drain_all_aio_error(void) { - BlockBackend *blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL); + BlockBackend *blk = blk_new(qemu_get_aio_context(), + BLK_PERM_ALL, BLK_PERM_ALL); BlockAIOCB *acb; bool completed = false; diff --git a/tests/test-block-iothread.c b/tests/test-block-iothread.c index 1d47ea9895..2200487d76 100644 --- a/tests/test-block-iothread.c +++ b/tests/test-block-iothread.c @@ -336,7 +336,7 @@ static void test_sync_op(const void *opaque) BlockDriverState *bs; BdrvChild *c; - blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL); + blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); bs = bdrv_new_open_driver(&bdrv_test, "base", BDRV_O_RDWR, &error_abort); bs->total_sectors = 65536 / BDRV_SECTOR_SIZE; blk_insert_bs(blk, bs, &error_abort); @@ -415,7 +415,7 @@ static void test_attach_blockjob(void) BlockDriverState *bs; TestBlockJob *tjob; - blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL); + blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); bs = bdrv_new_open_driver(&bdrv_test, "base", BDRV_O_RDWR, &error_abort); blk_insert_bs(blk, bs, &error_abort); @@ -481,7 +481,7 @@ static void test_propagate_basic(void) QDict *options; /* Create bs_a and its BlockBackend */ - blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL); + blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); bs_a = bdrv_new_open_driver(&bdrv_test, "bs_a", BDRV_O_RDWR, &error_abort); blk_insert_bs(blk, bs_a, &error_abort); @@ -561,7 +561,7 @@ static void test_propagate_diamond(void) qdict_put_str(options, "raw", "bs_c"); bs_verify = bdrv_open(NULL, NULL, options, BDRV_O_RDWR, &error_abort); - blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL); + blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); blk_insert_bs(blk, bs_verify, &error_abort); /* Switch the AioContext */ @@ -628,7 +628,7 @@ static void test_propagate_mirror(void) g_assert(bdrv_get_aio_context(filter) == main_ctx); /* With a BlockBackend on src, changing target must fail */ - blk = blk_new(0, BLK_PERM_ALL); + blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL); blk_insert_bs(blk, src, &error_abort); bdrv_try_set_aio_context(target, ctx, &local_err); diff --git a/tests/test-blockjob.c b/tests/test-blockjob.c index 652d1e8359..8c91980c70 100644 --- a/tests/test-blockjob.c +++ b/tests/test-blockjob.c @@ -68,7 +68,7 @@ static BlockJob *do_test_id(BlockBackend *blk, const char *id, static BlockBackend *create_blk(const char *name) { /* No I/O is performed on this device */ - BlockBackend *blk = blk_new(0, BLK_PERM_ALL); + BlockBackend *blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL); BlockDriverState *bs; bs = bdrv_open("null-co://", NULL, NULL, 0, &error_abort); diff --git a/tests/test-throttle.c b/tests/test-throttle.c index 948a42c991..5644cf95ca 100644 --- a/tests/test-throttle.c +++ b/tests/test-throttle.c @@ -675,9 +675,9 @@ static void test_groups(void) ThrottleGroupMember *tgm1, *tgm2, *tgm3; /* No actual I/O is performed on these devices */ - blk1 = blk_new(0, BLK_PERM_ALL); - blk2 = blk_new(0, BLK_PERM_ALL); - blk3 = blk_new(0, BLK_PERM_ALL); + blk1 = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL); + blk2 = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL); + blk3 = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL); blkp1 = blk_get_public(blk1); blkp2 = blk_get_public(blk2); -- cgit 1.4.1 From 132ada80c4a6fea7b67e8bb0a5fd299994d927c6 Mon Sep 17 00:00:00 2001 From: Kevin Wolf Date: Wed, 24 Apr 2019 17:41:46 +0200 Subject: block: Adjust AioContexts when attaching nodes So far, we only made sure that updating the AioContext of a node affected the whole subtree. However, if a node is newly attached to a new parent, we also need to make sure that both the subtree of the node and the parent are in the same AioContext. This tries to move the new child node to the parent AioContext and returns an error if this isn't possible. BlockBackends now actually apply their AioContext to their root node. Signed-off-by: Kevin Wolf --- block.c | 35 ++++++++++++++++++++++++++++++++++- block/block-backend.c | 9 ++++----- blockjob.c | 10 ++++++++-- include/block/block_int.h | 1 + tests/test-bdrv-drain.c | 6 ++++-- 5 files changed, 51 insertions(+), 10 deletions(-) (limited to 'block.c') diff --git a/block.c b/block.c index 69ceac6377..7d9ed1a724 100644 --- a/block.c +++ b/block.c @@ -2249,14 +2249,19 @@ static void bdrv_replace_child(BdrvChild *child, BlockDriverState *new_bs) * * On failure NULL is returned, errp is set and the reference to * child_bs is also dropped. + * + * The caller must hold the AioContext lock @child_bs, but not that of @ctx + * (unless @child_bs is already in @ctx). */ BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs, const char *child_name, const BdrvChildRole *child_role, + AioContext *ctx, uint64_t perm, uint64_t shared_perm, void *opaque, Error **errp) { BdrvChild *child; + Error *local_err = NULL; int ret; ret = bdrv_check_update_perm(child_bs, NULL, perm, shared_perm, NULL, errp); @@ -2276,6 +2281,31 @@ BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs, .opaque = opaque, }; + /* If the AioContexts don't match, first try to move the subtree of + * child_bs into the AioContext of the new parent. If this doesn't work, + * try moving the parent into the AioContext of child_bs instead. */ + if (bdrv_get_aio_context(child_bs) != ctx) { + ret = bdrv_try_set_aio_context(child_bs, ctx, &local_err); + if (ret < 0 && child_role->can_set_aio_ctx) { + GSList *ignore = g_slist_prepend(NULL, child);; + ctx = bdrv_get_aio_context(child_bs); + if (child_role->can_set_aio_ctx(child, ctx, &ignore, NULL)) { + error_free(local_err); + ret = 0; + g_slist_free(ignore); + ignore = g_slist_prepend(NULL, child);; + child_role->set_aio_ctx(child, ctx, &ignore); + } + g_slist_free(ignore); + } + if (ret < 0) { + error_propagate(errp, local_err); + g_free(child); + bdrv_abort_perm_update(child_bs); + return NULL; + } + } + /* This performs the matching bdrv_set_perm() for the above check. */ bdrv_replace_child(child, child_bs); @@ -2289,6 +2319,9 @@ BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs, * * On failure NULL is returned, errp is set and the reference to * child_bs is also dropped. + * + * If @parent_bs and @child_bs are in different AioContexts, the caller must + * hold the AioContext lock for @child_bs, but not for @parent_bs. */ BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs, BlockDriverState *child_bs, @@ -2302,11 +2335,11 @@ BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs, bdrv_get_cumulative_perm(parent_bs, &perm, &shared_perm); assert(parent_bs->drv); - assert(bdrv_get_aio_context(parent_bs) == bdrv_get_aio_context(child_bs)); bdrv_child_perm(parent_bs, child_bs, NULL, child_role, NULL, perm, shared_perm, &perm, &shared_perm); child = bdrv_root_attach_child(child_bs, child_name, child_role, + bdrv_get_aio_context(parent_bs), perm, shared_perm, parent_bs, errp); if (child == NULL) { return NULL; diff --git a/block/block-backend.c b/block/block-backend.c index f03f14acb0..f5d9407d20 100644 --- a/block/block-backend.c +++ b/block/block-backend.c @@ -392,7 +392,7 @@ BlockBackend *blk_new_open(const char *filename, const char *reference, return NULL; } - blk->root = bdrv_root_attach_child(bs, "root", &child_root, + blk->root = bdrv_root_attach_child(bs, "root", &child_root, blk->ctx, perm, BLK_PERM_ALL, blk, errp); if (!blk->root) { blk_unref(blk); @@ -803,7 +803,7 @@ int blk_insert_bs(BlockBackend *blk, BlockDriverState *bs, Error **errp) { ThrottleGroupMember *tgm = &blk->public.throttle_group_member; bdrv_ref(bs); - blk->root = bdrv_root_attach_child(bs, "root", &child_root, + blk->root = bdrv_root_attach_child(bs, "root", &child_root, blk->ctx, blk->perm, blk->shared_perm, blk, errp); if (blk->root == NULL) { return -EPERM; @@ -1861,10 +1861,9 @@ AioContext *blk_get_aio_context(BlockBackend *blk) { BlockDriverState *bs = blk_bs(blk); - /* FIXME The AioContext of bs and blk can be inconsistent. For the moment, - * we prefer the one of bs for compatibility. */ if (bs) { - return bdrv_get_aio_context(blk_bs(blk)); + AioContext *ctx = bdrv_get_aio_context(blk_bs(blk)); + assert(ctx == blk->ctx); } return blk->ctx; diff --git a/blockjob.c b/blockjob.c index 29517ae162..931d675c0c 100644 --- a/blockjob.c +++ b/blockjob.c @@ -205,8 +205,14 @@ int block_job_add_bdrv(BlockJob *job, const char *name, BlockDriverState *bs, BdrvChild *c; bdrv_ref(bs); - c = bdrv_root_attach_child(bs, name, &child_job, perm, shared_perm, - job, errp); + if (job->job.aio_context != qemu_get_aio_context()) { + aio_context_release(job->job.aio_context); + } + c = bdrv_root_attach_child(bs, name, &child_job, job->job.aio_context, + perm, shared_perm, job, errp); + if (job->job.aio_context != qemu_get_aio_context()) { + aio_context_acquire(job->job.aio_context); + } if (c == NULL) { return -EPERM; } diff --git a/include/block/block_int.h b/include/block/block_int.h index 1eebc7c8f3..06df2bda1b 100644 --- a/include/block/block_int.h +++ b/include/block/block_int.h @@ -1163,6 +1163,7 @@ void hmp_drive_add_node(Monitor *mon, const char *optstr); BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs, const char *child_name, const BdrvChildRole *child_role, + AioContext *ctx, uint64_t perm, uint64_t shared_perm, void *opaque, Error **errp); void bdrv_root_unref_child(BdrvChild *child); diff --git a/tests/test-bdrv-drain.c b/tests/test-bdrv-drain.c index fabbd52d93..16ea2b813f 100644 --- a/tests/test-bdrv-drain.c +++ b/tests/test-bdrv-drain.c @@ -912,6 +912,7 @@ static void test_blockjob_common_drain_node(enum drain_type drain_type, &error_abort); blk_target = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); blk_insert_bs(blk_target, target, &error_abort); + blk_set_allow_aio_context_change(blk_target, true); aio_context_acquire(ctx); tjob = block_job_create("job0", &test_job_driver, NULL, src, @@ -972,7 +973,7 @@ static void test_blockjob_common_drain_node(enum drain_type drain_type, g_assert_false(job->job.paused); g_assert_true(job->job.busy); /* We're in qemu_co_sleep_ns() */ - do_drain_begin(drain_type, target); + do_drain_begin_unlocked(drain_type, target); if (drain_type == BDRV_DRAIN_ALL) { /* bdrv_drain_all() drains both src and target */ @@ -983,7 +984,7 @@ static void test_blockjob_common_drain_node(enum drain_type drain_type, g_assert_true(job->job.paused); g_assert_false(job->job.busy); /* The job is paused */ - do_drain_end(drain_type, target); + do_drain_end_unlocked(drain_type, target); if (use_iothread) { /* paused is reset in the I/O thread, wait for it */ @@ -1002,6 +1003,7 @@ static void test_blockjob_common_drain_node(enum drain_type drain_type, if (use_iothread) { blk_set_aio_context(blk_src, qemu_get_aio_context(), &error_abort); + blk_set_aio_context(blk_target, qemu_get_aio_context(), &error_abort); } aio_context_release(ctx); -- cgit 1.4.1 From ad943dcb226a083a0c433ae73737c4834f8cfe74 Mon Sep 17 00:00:00 2001 From: Kevin Wolf Date: Wed, 24 Apr 2019 18:04:42 +0200 Subject: block: Move node without parents to main AioContext A node should only be in a non-default AioContext if a user is attached to it that requires this. When the last parent of a node is gone, it can move back to the main AioContext. Signed-off-by: Kevin Wolf --- block.c | 4 ++++ tests/test-bdrv-drain.c | 2 +- tests/test-block-iothread.c | 3 +-- 3 files changed, 6 insertions(+), 3 deletions(-) (limited to 'block.c') diff --git a/block.c b/block.c index 7d9ed1a724..6830f2d940 100644 --- a/block.c +++ b/block.c @@ -2235,6 +2235,10 @@ static void bdrv_replace_child(BdrvChild *child, BlockDriverState *new_bs) bdrv_get_cumulative_perm(old_bs, &perm, &shared_perm); bdrv_check_perm(old_bs, NULL, perm, shared_perm, NULL, &error_abort); bdrv_set_perm(old_bs, perm, shared_perm); + + /* 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); } if (new_bs) { diff --git a/tests/test-bdrv-drain.c b/tests/test-bdrv-drain.c index 16ea2b813f..44e658bde0 100644 --- a/tests/test-bdrv-drain.c +++ b/tests/test-bdrv-drain.c @@ -1003,7 +1003,7 @@ static void test_blockjob_common_drain_node(enum drain_type drain_type, if (use_iothread) { blk_set_aio_context(blk_src, qemu_get_aio_context(), &error_abort); - blk_set_aio_context(blk_target, qemu_get_aio_context(), &error_abort); + assert(blk_get_aio_context(blk_target) == qemu_get_aio_context()); } aio_context_release(ctx); diff --git a/tests/test-block-iothread.c b/tests/test-block-iothread.c index f41082e3bd..79d9cf8a57 100644 --- a/tests/test-block-iothread.c +++ b/tests/test-block-iothread.c @@ -713,9 +713,8 @@ static void test_attach_preserve_blk_ctx(void) /* Remove the node again */ blk_remove_bs(blk); - /* TODO bs should move back to main context here */ g_assert(blk_get_aio_context(blk) == ctx); - g_assert(bdrv_get_aio_context(bs) == ctx); + g_assert(bdrv_get_aio_context(bs) == qemu_get_aio_context()); /* Re-attach the node */ blk_insert_bs(blk, bs, &error_abort); -- cgit 1.4.1 From d0ee0204f400956ab429278f1b459d9af969c4a2 Mon Sep 17 00:00:00 2001 From: Kevin Wolf Date: Tue, 7 May 2019 18:19:16 +0200 Subject: block: Remove wrong bdrv_set_aio_context() calls The mirror and commit block jobs use bdrv_set_aio_context() to move their filter node into the right AioContext before hooking it up in the graph. Similarly, bdrv_open_backing_file() explicitly moves the backing file node into the right AioContext first. This isn't necessary any more, they get automatically moved into the right context now when attaching them. However, in the case of bdrv_open_backing_file() with a node reference, it's actually not only unnecessary, but even wrong: The unchecked bdrv_set_aio_context() changes the AioContext of the child node even if other parents require it to retain the old context. So this is not only a simplification, but a bug fix, too. Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1684342 Signed-off-by: Kevin Wolf --- block.c | 1 - block/commit.c | 2 -- block/mirror.c | 1 - 3 files changed, 4 deletions(-) (limited to 'block.c') diff --git a/block.c b/block.c index 6830f2d940..ddfae15d9b 100644 --- a/block.c +++ b/block.c @@ -2554,7 +2554,6 @@ int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options, ret = -EINVAL; goto free_exit; } - bdrv_set_aio_context(backing_hd, bdrv_get_aio_context(bs)); if (implicit_backing) { bdrv_refresh_filename(backing_hd); diff --git a/block/commit.c b/block/commit.c index 4d519506d6..c815def89a 100644 --- a/block/commit.c +++ b/block/commit.c @@ -301,7 +301,6 @@ void commit_start(const char *job_id, BlockDriverState *bs, commit_top_bs->implicit = true; } commit_top_bs->total_sectors = top->total_sectors; - bdrv_set_aio_context(commit_top_bs, bdrv_get_aio_context(top)); bdrv_append(commit_top_bs, top, &local_err); if (local_err) { @@ -443,7 +442,6 @@ int bdrv_commit(BlockDriverState *bs) error_report_err(local_err); goto ro_cleanup; } - bdrv_set_aio_context(commit_top_bs, bdrv_get_aio_context(backing_file_bs)); bdrv_set_backing_hd(commit_top_bs, backing_file_bs, &error_abort); bdrv_set_backing_hd(bs, commit_top_bs, &error_abort); diff --git a/block/mirror.c b/block/mirror.c index eb96b52de9..f8bdb5b21b 100644 --- a/block/mirror.c +++ b/block/mirror.c @@ -1543,7 +1543,6 @@ static void mirror_start_job(const char *job_id, BlockDriverState *bs, BDRV_REQ_NO_FALLBACK; bs_opaque = g_new0(MirrorBDSOpaque, 1); mirror_top_bs->opaque = bs_opaque; - bdrv_set_aio_context(mirror_top_bs, bdrv_get_aio_context(bs)); /* bdrv_append takes ownership of the mirror_top_bs reference, need to keep * it alive until block_job_create() succeeds even if bs has no parent. */ -- cgit 1.4.1 From 42a65f02f9b380bd8074882d5844d4ea033389cc Mon Sep 17 00:00:00 2001 From: Kevin Wolf Date: Tue, 7 May 2019 18:31:38 +0200 Subject: block: Remove bdrv_set_aio_context() All callers of bdrv_set_aio_context() are eliminated now, they have moved to bdrv_try_set_aio_context() and related safe functions. Remove bdrv_set_aio_context(). With this, we can now know that the .set_aio_ctx callback must be present in bdrv_set_aio_context_ignore() because bdrv_can_set_aio_context() would have returned false previously, so instead of checking the condition, we can assert it. Signed-off-by: Kevin Wolf --- block.c | 30 ++++++++++++++---------------- docs/devel/multiple-iothreads.txt | 4 ++-- include/block/block.h | 9 --------- 3 files changed, 16 insertions(+), 27 deletions(-) (limited to 'block.c') diff --git a/block.c b/block.c index ddfae15d9b..e3e77feee0 100644 --- a/block.c +++ b/block.c @@ -5789,8 +5789,17 @@ static void bdrv_attach_aio_context(BlockDriverState *bs, bs->walking_aio_notifiers = false; } -/* @ignore will accumulate all visited BdrvChild object. The caller is - * responsible for freeing the list afterwards. */ +/* + * Changes the AioContext used for fd handlers, timers, and BHs by this + * BlockDriverState and all its children and parents. + * + * 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) { @@ -5813,10 +5822,9 @@ void bdrv_set_aio_context_ignore(BlockDriverState *bs, if (g_slist_find(*ignore, child)) { continue; } - if (child->role->set_aio_ctx) { - *ignore = g_slist_prepend(*ignore, child); - child->role->set_aio_ctx(child, new_context, ignore); - } + assert(child->role->set_aio_ctx); + *ignore = g_slist_prepend(*ignore, child); + child->role->set_aio_ctx(child, new_context, ignore); } bdrv_detach_aio_context(bs); @@ -5830,16 +5838,6 @@ void bdrv_set_aio_context_ignore(BlockDriverState *bs, aio_context_release(new_context); } -/* 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). */ -void bdrv_set_aio_context(BlockDriverState *bs, AioContext *new_context) -{ - GSList *ignore_list = NULL; - bdrv_set_aio_context_ignore(bs, new_context, &ignore_list); - g_slist_free(ignore_list); -} - static bool bdrv_parent_can_set_aio_context(BdrvChild *c, AioContext *ctx, GSList **ignore, Error **errp) { diff --git a/docs/devel/multiple-iothreads.txt b/docs/devel/multiple-iothreads.txt index 4f9012d154..aeb997bed5 100644 --- a/docs/devel/multiple-iothreads.txt +++ b/docs/devel/multiple-iothreads.txt @@ -109,7 +109,7 @@ The AioContext originates from the QEMU block layer, even though nowadays AioContext is a generic event loop that can be used by any QEMU subsystem. The block layer has support for AioContext integrated. Each BlockDriverState -is associated with an AioContext using bdrv_set_aio_context() and +is associated with an AioContext using bdrv_try_set_aio_context() and bdrv_get_aio_context(). This allows block layer code to process I/O inside the right AioContext. Other subsystems may wish to follow a similar approach. @@ -134,5 +134,5 @@ Long-running jobs (usually in the form of coroutines) are best scheduled in the BlockDriverState's AioContext to avoid the need to acquire/release around each bdrv_*() call. The functions bdrv_add/remove_aio_context_notifier, or alternatively blk_add/remove_aio_context_notifier if you use BlockBackends, -can be used to get a notification whenever bdrv_set_aio_context() moves a +can be used to get a notification whenever bdrv_try_set_aio_context() moves a BlockDriverState to a different AioContext. diff --git a/include/block/block.h b/include/block/block.h index 531cf595cf..13ea050a5b 100644 --- a/include/block/block.h +++ b/include/block/block.h @@ -583,15 +583,6 @@ AioContext *bdrv_get_aio_context(BlockDriverState *bs); */ void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co); -/** - * bdrv_set_aio_context: - * - * Changes the #AioContext used for fd handlers, timers, and BHs by this - * BlockDriverState and all its children. - * - * This function must be called with iothread lock held. - */ -void bdrv_set_aio_context(BlockDriverState *bs, AioContext *new_context); void bdrv_set_aio_context_ignore(BlockDriverState *bs, AioContext *new_context, GSList **ignore); int bdrv_try_set_aio_context(BlockDriverState *bs, AioContext *ctx, -- cgit 1.4.1