d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

zfs: add zio_buf_alloc_nowait and use it in vdev_queue_aggregate

Browse Source 
This way we can avoid blocking the whole queue in the low memory
situations.  It's better to sacrifice some I/O performance by not doing
the aggregation than to add an indefinite wait for more memory.

Reviewed by:	smh
MFC after:	2 weeks
Sponsored by:	Panzura
Differential Revision: https://reviews.freebsd.org/D9999
This commit is contained in:
Andriy Gapon 2017-03-23 08:59:17 +00:00
parent afa0a46cfd
commit 3d775e193e
3 changed files with 28 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zio.h
View File

@ -567,6 +567,7 @@ extern zio_t *zio_unique_parent(zio_t *cio);
extern void zio_add_child(zio_t *pio, zio_t *cio);
extern void *zio_buf_alloc(size_t size);
extern void *zio_buf_alloc_nowait(size_t size);
extern void zio_buf_free(void *buf, size_t size);
extern void *zio_data_buf_alloc(size_t size);
extern void zio_data_buf_free(void *buf, size_t size);

7
sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_queue.c
View File

@ -647,6 +647,7 @@ static zio_t *
vdev_queue_aggregate(vdev_queue_t *vq, zio_t *zio)
{
zio_t *first, *last, *aio, *dio, *mandatory, *nio;
void *abuf;
uint64_t maxgap = 0;
uint64_t size;
boolean_t stretch;
@ -755,8 +756,12 @@ vdev_queue_aggregate(vdev_queue_t *vq, zio_t *zio)
size = IO_SPAN(first, last);
ASSERT3U(size, <=, zfs_vdev_aggregation_limit);
abuf = zio_buf_alloc_nowait(size);
if (abuf == NULL)
return (NULL);
aio = zio_vdev_delegated_io(first->io_vd, first->io_offset,
zio_buf_alloc(size), size, first->io_type, zio->io_priority,
abuf, size, first->io_type, zio->io_priority,
flags | ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_QUEUE,
vdev_queue_agg_io_done, NULL);
aio->io_timestamp = first->io_timestamp;

27
sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zio.c
View File

@ -272,18 +272,33 @@ zio_fini(void)
* useful to inspect ZFS metadata, but if possible, we should avoid keeping
* excess / transient data in-core during a crashdump.
*/
void *
zio_buf_alloc(size_t size)
static void *
zio_buf_alloc_impl(size_t size, boolean_t canwait)
{
size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;
int flags = zio_exclude_metadata ? KM_NODEBUG : 0;
VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
if (zio_use_uma)
return (kmem_cache_alloc(zio_buf_cache[c], KM_PUSHPAGE));
else
return (kmem_alloc(size, KM_SLEEP|flags));
if (zio_use_uma) {
return (kmem_cache_alloc(zio_buf_cache[c],
canwait ? KM_PUSHPAGE : KM_NOSLEEP));
} else {
return (kmem_alloc(size,
(canwait ? KM_SLEEP : KM_NOSLEEP) | flags));
}
}
void *
zio_buf_alloc(size_t size)
{
return (zio_buf_alloc_impl(size, B_TRUE));
}
void *
zio_buf_alloc_nowait(size_t size)
{
return (zio_buf_alloc_impl(size, B_FALSE));
}
/*