* CDDL HEADER END
*/
/*
- * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
* 1<<zfs_vdev_cache_bshift byte reads by the vdev_cache (aka software
* track buffer). At most zfs_vdev_cache_size bytes will be kept in each
* vdev's vdev_cache.
+ *
+ * TODO: Note that with the current ZFS code, it turns out that the
+ * vdev cache is not helpful, and in some cases actually harmful. It
+ * is better if we disable this. Once some time has passed, we should
+ * actually remove this to simplify the code. For now we just disable
+ * it by setting the zfs_vdev_cache_size to zero. Note that Solaris 11
+ * has made these same changes.
*/
int zfs_vdev_cache_max = 1<<14; /* 16KB */
-int zfs_vdev_cache_size = 10ULL << 20; /* 10MB */
+int zfs_vdev_cache_size = 0;
int zfs_vdev_cache_bshift = 16;
#define VCBS (1 << zfs_vdev_cache_bshift) /* 64KB */
vdev_cache_evict(vc, ve);
}
- ve = kmem_zalloc(sizeof (vdev_cache_entry_t), KM_SLEEP);
+ ve = kmem_zalloc(sizeof (vdev_cache_entry_t), KM_PUSHPAGE);
ve->ve_offset = offset;
- ve->ve_lastused = lbolt;
+ ve->ve_lastused = ddi_get_lbolt();
ve->ve_data = zio_buf_alloc(VCBS);
avl_add(&vc->vc_offset_tree, ve);
ASSERT(MUTEX_HELD(&vc->vc_lock));
ASSERT(ve->ve_fill_io == NULL);
- if (ve->ve_lastused != lbolt) {
+ if (ve->ve_lastused != ddi_get_lbolt()) {
avl_remove(&vc->vc_lastused_tree, ve);
- ve->ve_lastused = lbolt;
+ ve->ve_lastused = ddi_get_lbolt();
avl_add(&vc->vc_lastused_tree, ve);
}
* Fill a previously allocated cache entry with data.
*/
static void
-vdev_cache_fill(zio_t *zio)
+vdev_cache_fill(zio_t *fio)
{
- vdev_t *vd = zio->io_vd;
+ vdev_t *vd = fio->io_vd;
vdev_cache_t *vc = &vd->vdev_cache;
- vdev_cache_entry_t *ve = zio->io_private;
- zio_t *dio;
+ vdev_cache_entry_t *ve = fio->io_private;
+ zio_t *pio;
- ASSERT(zio->io_size == VCBS);
+ ASSERT(fio->io_size == VCBS);
/*
* Add data to the cache.
*/
mutex_enter(&vc->vc_lock);
- ASSERT(ve->ve_fill_io == zio);
- ASSERT(ve->ve_offset == zio->io_offset);
- ASSERT(ve->ve_data == zio->io_data);
+ ASSERT(ve->ve_fill_io == fio);
+ ASSERT(ve->ve_offset == fio->io_offset);
+ ASSERT(ve->ve_data == fio->io_data);
ve->ve_fill_io = NULL;
* any reads that were queued up before the missed update are still
* valid, so we can satisfy them from this line before we evict it.
*/
- for (dio = zio->io_delegate_list; dio; dio = dio->io_delegate_next)
- vdev_cache_hit(vc, ve, dio);
+ while ((pio = zio_walk_parents(fio)) != NULL)
+ vdev_cache_hit(vc, ve, pio);
- if (zio->io_error || ve->ve_missed_update)
+ if (fio->io_error || ve->ve_missed_update)
vdev_cache_evict(vc, ve);
mutex_exit(&vc->vc_lock);
-
- while ((dio = zio->io_delegate_list) != NULL) {
- zio->io_delegate_list = dio->io_delegate_next;
- dio->io_delegate_next = NULL;
- dio->io_error = zio->io_error;
- zio_execute(dio);
- }
}
/*
vdev_cache_read(zio_t *zio)
{
vdev_cache_t *vc = &zio->io_vd->vdev_cache;
- vdev_cache_entry_t *ve, ve_search;
+ vdev_cache_entry_t *ve, *ve_search;
uint64_t cache_offset = P2ALIGN(zio->io_offset, VCBS);
- uint64_t cache_phase = P2PHASE(zio->io_offset, VCBS);
+ ASSERTV(uint64_t cache_phase = P2PHASE(zio->io_offset, VCBS);)
zio_t *fio;
ASSERT(zio->io_type == ZIO_TYPE_READ);
mutex_enter(&vc->vc_lock);
- ve_search.ve_offset = cache_offset;
- ve = avl_find(&vc->vc_offset_tree, &ve_search, NULL);
+ ve_search = kmem_alloc(sizeof(vdev_cache_entry_t), KM_PUSHPAGE);
+ ve_search->ve_offset = cache_offset;
+ ve = avl_find(&vc->vc_offset_tree, ve_search, NULL);
+ kmem_free(ve_search, sizeof(vdev_cache_entry_t));
if (ve != NULL) {
if (ve->ve_missed_update) {
}
if ((fio = ve->ve_fill_io) != NULL) {
- zio->io_delegate_next = fio->io_delegate_list;
- fio->io_delegate_list = zio;
zio_vdev_io_bypass(zio);
+ zio_add_child(zio, fio);
mutex_exit(&vc->vc_lock);
VDCSTAT_BUMP(vdc_stat_delegations);
return (0);
zio_vdev_io_bypass(zio);
mutex_exit(&vc->vc_lock);
- zio_execute(zio);
VDCSTAT_BUMP(vdc_stat_hits);
return (0);
}
ZIO_FLAG_DONT_CACHE, vdev_cache_fill, ve);
ve->ve_fill_io = fio;
- fio->io_delegate_list = zio;
zio_vdev_io_bypass(zio);
+ zio_add_child(zio, fio);
mutex_exit(&vc->vc_lock);
zio_nowait(fio);
vdc_ksp = NULL;
}
}
+
+#if defined(_KERNEL) && defined(HAVE_SPL)
+module_param(zfs_vdev_cache_max, int, 0644);
+MODULE_PARM_DESC(zfs_vdev_cache_max, "Inflate reads small than max");
+
+module_param(zfs_vdev_cache_size, int, 0444);
+MODULE_PARM_DESC(zfs_vdev_cache_size, "Total size of the per-disk cache");
+
+module_param(zfs_vdev_cache_bshift, int, 0644);
+MODULE_PARM_DESC(zfs_vdev_cache_bshift, "Shift size to inflate reads too");
+#endif
git://git.camperquake.de / zfs.git / blobdiff