freebsd-dev/module/zfs/vdev_cache.c

437 lines
12 KiB
C
Raw Normal View History

2008-11-20 20:01:55 +00:00
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
2009-02-18 20:51:31 +00:00
* Copyright 2009 Sun Microsystems, Inc. All rights reserved.
2008-11-20 20:01:55 +00:00
* Use is subject to license terms.
*/
#include <sys/zfs_context.h>
#include <sys/spa.h>
#include <sys/vdev_impl.h>
#include <sys/zio.h>
#include <sys/kstat.h>
/*
* Virtual device read-ahead caching.
*
* This file implements a simple LRU read-ahead cache. When the DMU reads
* a given block, it will often want other, nearby blocks soon thereafter.
* We take advantage of this by reading a larger disk region and caching
* the result. In the best case, this can turn 128 back-to-back 512-byte
* reads into a single 64k read followed by 127 cache hits; this reduces
* latency dramatically. In the worst case, it can turn an isolated 512-byte
* read into a 64k read, which doesn't affect latency all that much but is
* terribly wasteful of bandwidth. A more intelligent version of the cache
* could keep track of access patterns and not do read-ahead unless it sees
* at least two temporally close I/Os to the same region. Currently, only
* metadata I/O is inflated. A futher enhancement could take advantage of
* more semantic information about the I/O. And it could use something
* faster than an AVL tree; that was chosen solely for convenience.
*
* There are five cache operations: allocate, fill, read, write, evict.
*
* (1) Allocate. This reserves a cache entry for the specified region.
* We separate the allocate and fill operations so that multiple threads
* don't generate I/O for the same cache miss.
*
* (2) Fill. When the I/O for a cache miss completes, the fill routine
* places the data in the previously allocated cache entry.
*
* (3) Read. Read data from the cache.
*
* (4) Write. Update cache contents after write completion.
*
* (5) Evict. When allocating a new entry, we evict the oldest (LRU) entry
* if the total cache size exceeds zfs_vdev_cache_size.
*/
/*
* These tunables are for performance analysis.
*/
/*
* All i/os smaller than zfs_vdev_cache_max will be turned into
* 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.
2008-11-20 20:01:55 +00:00
*/
int zfs_vdev_cache_max = 1<<14; /* 16KB */
int zfs_vdev_cache_size = 0;
2008-11-20 20:01:55 +00:00
int zfs_vdev_cache_bshift = 16;
#define VCBS (1 << zfs_vdev_cache_bshift) /* 64KB */
kstat_t *vdc_ksp = NULL;
typedef struct vdc_stats {
kstat_named_t vdc_stat_delegations;
kstat_named_t vdc_stat_hits;
kstat_named_t vdc_stat_misses;
} vdc_stats_t;
static vdc_stats_t vdc_stats = {
{ "delegations", KSTAT_DATA_UINT64 },
{ "hits", KSTAT_DATA_UINT64 },
{ "misses", KSTAT_DATA_UINT64 }
};
#define VDCSTAT_BUMP(stat) atomic_add_64(&vdc_stats.stat.value.ui64, 1);
static int
vdev_cache_offset_compare(const void *a1, const void *a2)
{
const vdev_cache_entry_t *ve1 = a1;
const vdev_cache_entry_t *ve2 = a2;
if (ve1->ve_offset < ve2->ve_offset)
return (-1);
if (ve1->ve_offset > ve2->ve_offset)
return (1);
return (0);
}
static int
vdev_cache_lastused_compare(const void *a1, const void *a2)
{
const vdev_cache_entry_t *ve1 = a1;
const vdev_cache_entry_t *ve2 = a2;
if (ve1->ve_lastused < ve2->ve_lastused)
return (-1);
if (ve1->ve_lastused > ve2->ve_lastused)
return (1);
/*
* Among equally old entries, sort by offset to ensure uniqueness.
*/
return (vdev_cache_offset_compare(a1, a2));
}
/*
* Evict the specified entry from the cache.
*/
static void
vdev_cache_evict(vdev_cache_t *vc, vdev_cache_entry_t *ve)
{
ASSERT(MUTEX_HELD(&vc->vc_lock));
ASSERT(ve->ve_fill_io == NULL);
ASSERT(ve->ve_data != NULL);
avl_remove(&vc->vc_lastused_tree, ve);
avl_remove(&vc->vc_offset_tree, ve);
zio_buf_free(ve->ve_data, VCBS);
kmem_free(ve, sizeof (vdev_cache_entry_t));
}
/*
* Allocate an entry in the cache. At the point we don't have the data,
* we're just creating a placeholder so that multiple threads don't all
* go off and read the same blocks.
*/
static vdev_cache_entry_t *
vdev_cache_allocate(zio_t *zio)
{
vdev_cache_t *vc = &zio->io_vd->vdev_cache;
uint64_t offset = P2ALIGN(zio->io_offset, VCBS);
vdev_cache_entry_t *ve;
ASSERT(MUTEX_HELD(&vc->vc_lock));
if (zfs_vdev_cache_size == 0)
return (NULL);
/*
* If adding a new entry would exceed the cache size,
* evict the oldest entry (LRU).
*/
if ((avl_numnodes(&vc->vc_lastused_tree) << zfs_vdev_cache_bshift) >
zfs_vdev_cache_size) {
ve = avl_first(&vc->vc_lastused_tree);
if (ve->ve_fill_io != NULL)
2008-11-20 20:01:55 +00:00
return (NULL);
ASSERT(ve->ve_hits != 0);
vdev_cache_evict(vc, ve);
}
ve = kmem_zalloc(sizeof (vdev_cache_entry_t), KM_PUSHPAGE);
2008-11-20 20:01:55 +00:00
ve->ve_offset = offset;
ve->ve_lastused = ddi_get_lbolt();
2008-11-20 20:01:55 +00:00
ve->ve_data = zio_buf_alloc(VCBS);
avl_add(&vc->vc_offset_tree, ve);
avl_add(&vc->vc_lastused_tree, ve);
return (ve);
}
static void
vdev_cache_hit(vdev_cache_t *vc, vdev_cache_entry_t *ve, zio_t *zio)
{
uint64_t cache_phase = P2PHASE(zio->io_offset, VCBS);
ASSERT(MUTEX_HELD(&vc->vc_lock));
ASSERT(ve->ve_fill_io == NULL);
if (ve->ve_lastused != ddi_get_lbolt()) {
2008-11-20 20:01:55 +00:00
avl_remove(&vc->vc_lastused_tree, ve);
ve->ve_lastused = ddi_get_lbolt();
2008-11-20 20:01:55 +00:00
avl_add(&vc->vc_lastused_tree, ve);
}
ve->ve_hits++;
bcopy(ve->ve_data + cache_phase, zio->io_data, zio->io_size);
}
/*
* Fill a previously allocated cache entry with data.
*/
static void
2009-02-18 20:51:31 +00:00
vdev_cache_fill(zio_t *fio)
2008-11-20 20:01:55 +00:00
{
2009-02-18 20:51:31 +00:00
vdev_t *vd = fio->io_vd;
2008-11-20 20:01:55 +00:00
vdev_cache_t *vc = &vd->vdev_cache;
2009-02-18 20:51:31 +00:00
vdev_cache_entry_t *ve = fio->io_private;
zio_t *pio;
2008-11-20 20:01:55 +00:00
2009-02-18 20:51:31 +00:00
ASSERT(fio->io_size == VCBS);
2008-11-20 20:01:55 +00:00
/*
* Add data to the cache.
*/
mutex_enter(&vc->vc_lock);
2009-02-18 20:51:31 +00:00
ASSERT(ve->ve_fill_io == fio);
ASSERT(ve->ve_offset == fio->io_offset);
ASSERT(ve->ve_data == fio->io_data);
2008-11-20 20:01:55 +00:00
ve->ve_fill_io = NULL;
/*
* Even if this cache line was invalidated by a missed write update,
* 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.
*/
2009-02-18 20:51:31 +00:00
while ((pio = zio_walk_parents(fio)) != NULL)
vdev_cache_hit(vc, ve, pio);
2008-11-20 20:01:55 +00:00
2009-02-18 20:51:31 +00:00
if (fio->io_error || ve->ve_missed_update)
2008-11-20 20:01:55 +00:00
vdev_cache_evict(vc, ve);
mutex_exit(&vc->vc_lock);
}
/*
* Read data from the cache. Returns 0 on cache hit, errno on a miss.
*/
int
vdev_cache_read(zio_t *zio)
{
vdev_cache_t *vc = &zio->io_vd->vdev_cache;
vdev_cache_entry_t *ve, *ve_search;
2008-11-20 20:01:55 +00:00
uint64_t cache_offset = P2ALIGN(zio->io_offset, VCBS);
ASSERTV(uint64_t cache_phase = P2PHASE(zio->io_offset, VCBS);)
2008-11-20 20:01:55 +00:00
zio_t *fio;
ASSERT(zio->io_type == ZIO_TYPE_READ);
if (zio->io_flags & ZIO_FLAG_DONT_CACHE)
return (EINVAL);
if (zio->io_size > zfs_vdev_cache_max)
return (EOVERFLOW);
/*
* If the I/O straddles two or more cache blocks, don't cache it.
*/
if (P2BOUNDARY(zio->io_offset, zio->io_size, VCBS))
2008-11-20 20:01:55 +00:00
return (EXDEV);
ASSERT(cache_phase + zio->io_size <= VCBS);
mutex_enter(&vc->vc_lock);
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));
2008-11-20 20:01:55 +00:00
if (ve != NULL) {
if (ve->ve_missed_update) {
mutex_exit(&vc->vc_lock);
return (ESTALE);
}
if ((fio = ve->ve_fill_io) != NULL) {
zio_vdev_io_bypass(zio);
2009-02-18 20:51:31 +00:00
zio_add_child(zio, fio);
2008-11-20 20:01:55 +00:00
mutex_exit(&vc->vc_lock);
VDCSTAT_BUMP(vdc_stat_delegations);
return (0);
}
vdev_cache_hit(vc, ve, zio);
zio_vdev_io_bypass(zio);
mutex_exit(&vc->vc_lock);
VDCSTAT_BUMP(vdc_stat_hits);
return (0);
}
ve = vdev_cache_allocate(zio);
if (ve == NULL) {
mutex_exit(&vc->vc_lock);
return (ENOMEM);
}
fio = zio_vdev_delegated_io(zio->io_vd, cache_offset,
2008-11-20 20:01:55 +00:00
ve->ve_data, VCBS, ZIO_TYPE_READ, ZIO_PRIORITY_CACHE_FILL,
ZIO_FLAG_DONT_CACHE, vdev_cache_fill, ve);
2008-11-20 20:01:55 +00:00
ve->ve_fill_io = fio;
zio_vdev_io_bypass(zio);
2009-02-18 20:51:31 +00:00
zio_add_child(zio, fio);
2008-11-20 20:01:55 +00:00
mutex_exit(&vc->vc_lock);
zio_nowait(fio);
VDCSTAT_BUMP(vdc_stat_misses);
return (0);
}
/*
* Update cache contents upon write completion.
*/
void
vdev_cache_write(zio_t *zio)
{
vdev_cache_t *vc = &zio->io_vd->vdev_cache;
vdev_cache_entry_t *ve, ve_search;
uint64_t io_start = zio->io_offset;
uint64_t io_end = io_start + zio->io_size;
uint64_t min_offset = P2ALIGN(io_start, VCBS);
uint64_t max_offset = P2ROUNDUP(io_end, VCBS);
avl_index_t where;
ASSERT(zio->io_type == ZIO_TYPE_WRITE);
mutex_enter(&vc->vc_lock);
ve_search.ve_offset = min_offset;
ve = avl_find(&vc->vc_offset_tree, &ve_search, &where);
if (ve == NULL)
ve = avl_nearest(&vc->vc_offset_tree, where, AVL_AFTER);
while (ve != NULL && ve->ve_offset < max_offset) {
uint64_t start = MAX(ve->ve_offset, io_start);
uint64_t end = MIN(ve->ve_offset + VCBS, io_end);
if (ve->ve_fill_io != NULL) {
ve->ve_missed_update = 1;
} else {
bcopy((char *)zio->io_data + start - io_start,
ve->ve_data + start - ve->ve_offset, end - start);
}
ve = AVL_NEXT(&vc->vc_offset_tree, ve);
}
mutex_exit(&vc->vc_lock);
}
void
vdev_cache_purge(vdev_t *vd)
{
vdev_cache_t *vc = &vd->vdev_cache;
vdev_cache_entry_t *ve;
mutex_enter(&vc->vc_lock);
while ((ve = avl_first(&vc->vc_offset_tree)) != NULL)
vdev_cache_evict(vc, ve);
mutex_exit(&vc->vc_lock);
}
void
vdev_cache_init(vdev_t *vd)
{
vdev_cache_t *vc = &vd->vdev_cache;
mutex_init(&vc->vc_lock, NULL, MUTEX_DEFAULT, NULL);
avl_create(&vc->vc_offset_tree, vdev_cache_offset_compare,
sizeof (vdev_cache_entry_t),
offsetof(struct vdev_cache_entry, ve_offset_node));
avl_create(&vc->vc_lastused_tree, vdev_cache_lastused_compare,
sizeof (vdev_cache_entry_t),
offsetof(struct vdev_cache_entry, ve_lastused_node));
}
void
vdev_cache_fini(vdev_t *vd)
{
vdev_cache_t *vc = &vd->vdev_cache;
vdev_cache_purge(vd);
avl_destroy(&vc->vc_offset_tree);
avl_destroy(&vc->vc_lastused_tree);
mutex_destroy(&vc->vc_lock);
}
void
vdev_cache_stat_init(void)
{
vdc_ksp = kstat_create("zfs", 0, "vdev_cache_stats", "misc",
KSTAT_TYPE_NAMED, sizeof (vdc_stats) / sizeof (kstat_named_t),
KSTAT_FLAG_VIRTUAL);
if (vdc_ksp != NULL) {
vdc_ksp->ks_data = &vdc_stats;
kstat_install(vdc_ksp);
}
}
void
vdev_cache_stat_fini(void)
{
if (vdc_ksp != NULL) {
kstat_delete(vdc_ksp);
vdc_ksp = NULL;
}
}
Add missing ZFS tunables This commit adds module options for all existing zfs tunables. Ideally the average user should never need to modify any of these values. However, in practice sometimes you do need to tweak these values for one reason or another. In those cases it's nice not to have to resort to rebuilding from source. All tunables are visable to modinfo and the list is as follows: $ modinfo module/zfs/zfs.ko filename: module/zfs/zfs.ko license: CDDL author: Sun Microsystems/Oracle, Lawrence Livermore National Laboratory description: ZFS srcversion: 8EAB1D71DACE05B5AA61567 depends: spl,znvpair,zcommon,zunicode,zavl vermagic: 2.6.32-131.0.5.el6.x86_64 SMP mod_unload modversions parm: zvol_major:Major number for zvol device (uint) parm: zvol_threads:Number of threads for zvol device (uint) parm: zio_injection_enabled:Enable fault injection (int) parm: zio_bulk_flags:Additional flags to pass to bulk buffers (int) parm: zio_delay_max:Max zio millisec delay before posting event (int) parm: zio_requeue_io_start_cut_in_line:Prioritize requeued I/O (bool) parm: zil_replay_disable:Disable intent logging replay (int) parm: zfs_nocacheflush:Disable cache flushes (bool) parm: zfs_read_chunk_size:Bytes to read per chunk (long) parm: zfs_vdev_max_pending:Max pending per-vdev I/Os (int) parm: zfs_vdev_min_pending:Min pending per-vdev I/Os (int) parm: zfs_vdev_aggregation_limit:Max vdev I/O aggregation size (int) parm: zfs_vdev_time_shift:Deadline time shift for vdev I/O (int) parm: zfs_vdev_ramp_rate:Exponential I/O issue ramp-up rate (int) parm: zfs_vdev_read_gap_limit:Aggregate read I/O over gap (int) parm: zfs_vdev_write_gap_limit:Aggregate write I/O over gap (int) parm: zfs_vdev_scheduler:I/O scheduler (charp) parm: zfs_vdev_cache_max:Inflate reads small than max (int) parm: zfs_vdev_cache_size:Total size of the per-disk cache (int) parm: zfs_vdev_cache_bshift:Shift size to inflate reads too (int) parm: zfs_scrub_limit:Max scrub/resilver I/O per leaf vdev (int) parm: zfs_recover:Set to attempt to recover from fatal errors (int) parm: spa_config_path:SPA config file (/etc/zfs/zpool.cache) (charp) parm: zfs_zevent_len_max:Max event queue length (int) parm: zfs_zevent_cols:Max event column width (int) parm: zfs_zevent_console:Log events to the console (int) parm: zfs_top_maxinflight:Max I/Os per top-level (int) parm: zfs_resilver_delay:Number of ticks to delay resilver (int) parm: zfs_scrub_delay:Number of ticks to delay scrub (int) parm: zfs_scan_idle:Idle window in clock ticks (int) parm: zfs_scan_min_time_ms:Min millisecs to scrub per txg (int) parm: zfs_free_min_time_ms:Min millisecs to free per txg (int) parm: zfs_resilver_min_time_ms:Min millisecs to resilver per txg (int) parm: zfs_no_scrub_io:Set to disable scrub I/O (bool) parm: zfs_no_scrub_prefetch:Set to disable scrub prefetching (bool) parm: zfs_txg_timeout:Max seconds worth of delta per txg (int) parm: zfs_no_write_throttle:Disable write throttling (int) parm: zfs_write_limit_shift:log2(fraction of memory) per txg (int) parm: zfs_txg_synctime_ms:Target milliseconds between tgx sync (int) parm: zfs_write_limit_min:Min tgx write limit (ulong) parm: zfs_write_limit_max:Max tgx write limit (ulong) parm: zfs_write_limit_inflated:Inflated tgx write limit (ulong) parm: zfs_write_limit_override:Override tgx write limit (ulong) parm: zfs_prefetch_disable:Disable all ZFS prefetching (int) parm: zfetch_max_streams:Max number of streams per zfetch (uint) parm: zfetch_min_sec_reap:Min time before stream reclaim (uint) parm: zfetch_block_cap:Max number of blocks to fetch at a time (uint) parm: zfetch_array_rd_sz:Number of bytes in a array_read (ulong) parm: zfs_pd_blks_max:Max number of blocks to prefetch (int) parm: zfs_dedup_prefetch:Enable prefetching dedup-ed blks (int) parm: zfs_arc_min:Min arc size (ulong) parm: zfs_arc_max:Max arc size (ulong) parm: zfs_arc_meta_limit:Meta limit for arc size (ulong) parm: zfs_arc_reduce_dnlc_percent:Meta reclaim percentage (int) parm: zfs_arc_grow_retry:Seconds before growing arc size (int) parm: zfs_arc_shrink_shift:log2(fraction of arc to reclaim) (int) parm: zfs_arc_p_min_shift:arc_c shift to calc min/max arc_p (int)
2011-05-03 22:09:28 +00:00
#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