diff options
author | Kevin Wolf <kwolf@redhat.com> | 2018-06-21 17:54:35 +0200 |
---|---|---|
committer | Kevin Wolf <kwolf@redhat.com> | 2018-06-29 14:20:56 +0200 |
commit | 061ca8a368165fae300748c17971824a089f521f (patch) | |
tree | 94e2fb34999021ead20698bf7399ae151d13047b /block.c | |
parent | ae5475e82fd1ebb24f4f77cf28f59ca6548c6136 (diff) |
block: Convert .bdrv_truncate callback to coroutine_fn
bdrv_truncate() is an operation that can block (even for a quite long
time, depending on the PreallocMode) in I/O paths that shouldn't block.
Convert it to a coroutine_fn so that we have the infrastructure for
drivers to make their .bdrv_co_truncate implementation asynchronous.
This change could potentially introduce new race conditions because
bdrv_truncate() isn't necessarily executed atomically any more. Whether
this is a problem needs to be evaluated for each block driver that
supports truncate:
* file-posix/win32, gluster, iscsi, nfs, rbd, ssh, sheepdog: The
protocol drivers are trivially safe because they don't actually yield
yet, so there is no change in behaviour.
* copy-on-read, crypto, raw-format: Essentially just filter drivers that
pass the request to a child node, no problem.
* qcow2: The implementation modifies metadata, so it needs to hold
s->lock to be safe with concurrent I/O requests. In order to avoid
double locking, this requires pulling the locking out into
preallocate_co() and using qcow2_write_caches() instead of
bdrv_flush().
* qed: Does a single header update, this is fine without locking.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Diffstat (limited to 'block.c')
-rw-r--r-- | block.c | 63 |
1 files changed, 55 insertions, 8 deletions
@@ -3788,8 +3788,8 @@ exit: /** * Truncate file to 'offset' bytes (needed only for file protocols) */ -int bdrv_truncate(BdrvChild *child, int64_t offset, PreallocMode prealloc, - Error **errp) +int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset, + PreallocMode prealloc, Error **errp) { BlockDriverState *bs = child->bs; BlockDriver *drv = bs->drv; @@ -3807,23 +3807,28 @@ int bdrv_truncate(BdrvChild *child, int64_t offset, PreallocMode prealloc, return -EINVAL; } - if (!drv->bdrv_truncate) { + bdrv_inc_in_flight(bs); + + if (!drv->bdrv_co_truncate) { if (bs->file && drv->is_filter) { - return bdrv_truncate(bs->file, offset, prealloc, errp); + ret = bdrv_co_truncate(bs->file, offset, prealloc, errp); + goto out; } error_setg(errp, "Image format driver does not support resize"); - return -ENOTSUP; + ret = -ENOTSUP; + goto out; } if (bs->read_only) { error_setg(errp, "Image is read-only"); - return -EACCES; + ret = -EACCES; + goto out; } assert(!(bs->open_flags & BDRV_O_INACTIVE)); - ret = drv->bdrv_truncate(bs, offset, prealloc, errp); + ret = drv->bdrv_co_truncate(bs, offset, prealloc, errp); if (ret < 0) { - return ret; + goto out; } ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS); if (ret < 0) { @@ -3834,9 +3839,51 @@ int bdrv_truncate(BdrvChild *child, int64_t offset, PreallocMode prealloc, bdrv_dirty_bitmap_truncate(bs, offset); bdrv_parent_cb_resize(bs); atomic_inc(&bs->write_gen); + +out: + bdrv_dec_in_flight(bs); return ret; } +typedef struct TruncateCo { + BdrvChild *child; + int64_t offset; + PreallocMode prealloc; + Error **errp; + int ret; +} TruncateCo; + +static void coroutine_fn bdrv_truncate_co_entry(void *opaque) +{ + TruncateCo *tco = opaque; + tco->ret = bdrv_co_truncate(tco->child, tco->offset, tco->prealloc, + tco->errp); +} + +int bdrv_truncate(BdrvChild *child, int64_t offset, PreallocMode prealloc, + Error **errp) +{ + Coroutine *co; + TruncateCo tco = { + .child = child, + .offset = offset, + .prealloc = prealloc, + .errp = errp, + .ret = NOT_DONE, + }; + + if (qemu_in_coroutine()) { + /* Fast-path if already in coroutine context */ + bdrv_truncate_co_entry(&tco); + } else { + co = qemu_coroutine_create(bdrv_truncate_co_entry, &tco); + qemu_coroutine_enter(co); + BDRV_POLL_WHILE(child->bs, tco.ret == NOT_DONE); + } + + return tco.ret; +} + /** * Length of a allocated file in bytes. Sparse files are counted by actual * allocated space. Return < 0 if error or unknown. |