Commit 279afbad authored by Kent Overstreet's avatar Kent Overstreet

bcache: Track dirty data by stripe

To make background writeback aware of raid5/6 stripes, we first need to
track the amount of dirty data within each stripe - we do this by
breaking up the existing sectors_dirty into per stripe atomic_ts
Signed-off-by: default avatarKent Overstreet <koverstreet@google.com>
parent 444fc0b6
...@@ -437,7 +437,10 @@ struct bcache_device { ...@@ -437,7 +437,10 @@ struct bcache_device {
/* If nonzero, we're detaching/unregistering from cache set */ /* If nonzero, we're detaching/unregistering from cache set */
atomic_t detaching; atomic_t detaching;
atomic_long_t sectors_dirty; uint64_t nr_stripes;
unsigned stripe_size_bits;
atomic_t *stripe_sectors_dirty;
unsigned long sectors_dirty_last; unsigned long sectors_dirty_last;
long sectors_dirty_derivative; long sectors_dirty_derivative;
...@@ -1159,9 +1162,6 @@ static inline void wake_up_allocators(struct cache_set *c) ...@@ -1159,9 +1162,6 @@ static inline void wake_up_allocators(struct cache_set *c)
/* Forward declarations */ /* Forward declarations */
void bch_writeback_queue(struct cached_dev *);
void bch_writeback_add(struct cached_dev *, unsigned);
void bch_count_io_errors(struct cache *, int, const char *); void bch_count_io_errors(struct cache *, int, const char *);
void bch_bbio_count_io_errors(struct cache_set *, struct bio *, void bch_bbio_count_io_errors(struct cache_set *, struct bio *,
int, const char *); int, const char *);
...@@ -1224,8 +1224,6 @@ void bch_cache_set_stop(struct cache_set *); ...@@ -1224,8 +1224,6 @@ void bch_cache_set_stop(struct cache_set *);
struct cache_set *bch_cache_set_alloc(struct cache_sb *); struct cache_set *bch_cache_set_alloc(struct cache_sb *);
void bch_btree_cache_free(struct cache_set *); void bch_btree_cache_free(struct cache_set *);
int bch_btree_cache_alloc(struct cache_set *); int bch_btree_cache_alloc(struct cache_set *);
void bch_sectors_dirty_init(struct cached_dev *);
void bch_cached_dev_writeback_init(struct cached_dev *);
void bch_moving_init_cache_set(struct cache_set *); void bch_moving_init_cache_set(struct cache_set *);
int bch_cache_allocator_start(struct cache *ca); int bch_cache_allocator_start(struct cache *ca);
......
...@@ -24,6 +24,7 @@ ...@@ -24,6 +24,7 @@
#include "btree.h" #include "btree.h"
#include "debug.h" #include "debug.h"
#include "request.h" #include "request.h"
#include "writeback.h"
#include <linux/slab.h> #include <linux/slab.h>
#include <linux/bitops.h> #include <linux/bitops.h>
...@@ -1599,14 +1600,14 @@ static bool fix_overlapping_extents(struct btree *b, ...@@ -1599,14 +1600,14 @@ static bool fix_overlapping_extents(struct btree *b,
struct btree_iter *iter, struct btree_iter *iter,
struct btree_op *op) struct btree_op *op)
{ {
void subtract_dirty(struct bkey *k, int sectors) void subtract_dirty(struct bkey *k, uint64_t offset, int sectors)
{ {
struct bcache_device *d = b->c->devices[KEY_INODE(k)]; if (KEY_DIRTY(k))
bcache_dev_sectors_dirty_add(b->c, KEY_INODE(k),
if (KEY_DIRTY(k) && d) offset, -sectors);
atomic_long_sub(sectors, &d->sectors_dirty);
} }
uint64_t old_offset;
unsigned old_size, sectors_found = 0; unsigned old_size, sectors_found = 0;
while (1) { while (1) {
...@@ -1618,6 +1619,7 @@ static bool fix_overlapping_extents(struct btree *b, ...@@ -1618,6 +1619,7 @@ static bool fix_overlapping_extents(struct btree *b,
if (bkey_cmp(k, &START_KEY(insert)) <= 0) if (bkey_cmp(k, &START_KEY(insert)) <= 0)
continue; continue;
old_offset = KEY_START(k);
old_size = KEY_SIZE(k); old_size = KEY_SIZE(k);
/* /*
...@@ -1673,7 +1675,7 @@ static bool fix_overlapping_extents(struct btree *b, ...@@ -1673,7 +1675,7 @@ static bool fix_overlapping_extents(struct btree *b,
struct bkey *top; struct bkey *top;
subtract_dirty(k, KEY_SIZE(insert)); subtract_dirty(k, KEY_START(insert), KEY_SIZE(insert));
if (bkey_written(b, k)) { if (bkey_written(b, k)) {
/* /*
...@@ -1720,7 +1722,7 @@ static bool fix_overlapping_extents(struct btree *b, ...@@ -1720,7 +1722,7 @@ static bool fix_overlapping_extents(struct btree *b,
} }
} }
subtract_dirty(k, old_size - KEY_SIZE(k)); subtract_dirty(k, old_offset, old_size - KEY_SIZE(k));
} }
check_failed: check_failed:
...@@ -1796,6 +1798,10 @@ static bool btree_insert_key(struct btree *b, struct btree_op *op, ...@@ -1796,6 +1798,10 @@ static bool btree_insert_key(struct btree *b, struct btree_op *op,
insert: shift_keys(b, m, k); insert: shift_keys(b, m, k);
copy: bkey_copy(m, k); copy: bkey_copy(m, k);
merged: merged:
if (KEY_DIRTY(k))
bcache_dev_sectors_dirty_add(b->c, KEY_INODE(k),
KEY_START(k), KEY_SIZE(k));
bch_check_keys(b, "%u for %s", status, op_type(op)); bch_check_keys(b, "%u for %s", status, op_type(op));
if (b->level && !KEY_OFFSET(k)) if (b->level && !KEY_OFFSET(k))
......
...@@ -10,6 +10,7 @@ ...@@ -10,6 +10,7 @@
#include "btree.h" #include "btree.h"
#include "debug.h" #include "debug.h"
#include "request.h" #include "request.h"
#include "writeback.h"
#include <linux/cgroup.h> #include <linux/cgroup.h>
#include <linux/module.h> #include <linux/module.h>
...@@ -1044,7 +1045,7 @@ static void request_write(struct cached_dev *dc, struct search *s) ...@@ -1044,7 +1045,7 @@ static void request_write(struct cached_dev *dc, struct search *s)
closure_bio_submit(bio, cl, s->d); closure_bio_submit(bio, cl, s->d);
} else { } else {
s->op.cache_bio = bio; s->op.cache_bio = bio;
bch_writeback_add(dc, bio_sectors(bio)); bch_writeback_add(dc);
} }
out: out:
closure_call(&s->op.cl, bch_insert_data, NULL, cl); closure_call(&s->op.cl, bch_insert_data, NULL, cl);
......
...@@ -10,6 +10,7 @@ ...@@ -10,6 +10,7 @@
#include "btree.h" #include "btree.h"
#include "debug.h" #include "debug.h"
#include "request.h" #include "request.h"
#include "writeback.h"
#include <linux/blkdev.h> #include <linux/blkdev.h>
#include <linux/buffer_head.h> #include <linux/buffer_head.h>
...@@ -744,13 +745,35 @@ static void bcache_device_free(struct bcache_device *d) ...@@ -744,13 +745,35 @@ static void bcache_device_free(struct bcache_device *d)
mempool_destroy(d->unaligned_bvec); mempool_destroy(d->unaligned_bvec);
if (d->bio_split) if (d->bio_split)
bioset_free(d->bio_split); bioset_free(d->bio_split);
if (is_vmalloc_addr(d->stripe_sectors_dirty))
vfree(d->stripe_sectors_dirty);
else
kfree(d->stripe_sectors_dirty);
closure_debug_destroy(&d->cl); closure_debug_destroy(&d->cl);
} }
static int bcache_device_init(struct bcache_device *d, unsigned block_size) static int bcache_device_init(struct bcache_device *d, unsigned block_size,
sector_t sectors)
{ {
struct request_queue *q; struct request_queue *q;
size_t n;
if (!d->stripe_size_bits)
d->stripe_size_bits = 31;
d->nr_stripes = round_up(sectors, 1 << d->stripe_size_bits) >>
d->stripe_size_bits;
if (!d->nr_stripes || d->nr_stripes > SIZE_MAX / sizeof(atomic_t))
return -ENOMEM;
n = d->nr_stripes * sizeof(atomic_t);
d->stripe_sectors_dirty = n < PAGE_SIZE << 6
? kzalloc(n, GFP_KERNEL)
: vzalloc(n);
if (!d->stripe_sectors_dirty)
return -ENOMEM;
if (!(d->bio_split = bioset_create(4, offsetof(struct bbio, bio))) || if (!(d->bio_split = bioset_create(4, offsetof(struct bbio, bio))) ||
!(d->unaligned_bvec = mempool_create_kmalloc_pool(1, !(d->unaligned_bvec = mempool_create_kmalloc_pool(1,
...@@ -760,6 +783,7 @@ static int bcache_device_init(struct bcache_device *d, unsigned block_size) ...@@ -760,6 +783,7 @@ static int bcache_device_init(struct bcache_device *d, unsigned block_size)
!(q = blk_alloc_queue(GFP_KERNEL))) !(q = blk_alloc_queue(GFP_KERNEL)))
return -ENOMEM; return -ENOMEM;
set_capacity(d->disk, sectors);
snprintf(d->disk->disk_name, DISK_NAME_LEN, "bcache%i", bcache_minor); snprintf(d->disk->disk_name, DISK_NAME_LEN, "bcache%i", bcache_minor);
d->disk->major = bcache_major; d->disk->major = bcache_major;
...@@ -1047,7 +1071,8 @@ static int cached_dev_init(struct cached_dev *dc, unsigned block_size) ...@@ -1047,7 +1071,8 @@ static int cached_dev_init(struct cached_dev *dc, unsigned block_size)
hlist_add_head(&io->hash, dc->io_hash + RECENT_IO); hlist_add_head(&io->hash, dc->io_hash + RECENT_IO);
} }
ret = bcache_device_init(&dc->disk, block_size); ret = bcache_device_init(&dc->disk, block_size,
dc->bdev->bd_part->nr_sects - dc->sb.data_offset);
if (ret) if (ret)
return ret; return ret;
...@@ -1146,11 +1171,10 @@ static int flash_dev_run(struct cache_set *c, struct uuid_entry *u) ...@@ -1146,11 +1171,10 @@ static int flash_dev_run(struct cache_set *c, struct uuid_entry *u)
kobject_init(&d->kobj, &bch_flash_dev_ktype); kobject_init(&d->kobj, &bch_flash_dev_ktype);
if (bcache_device_init(d, block_bytes(c))) if (bcache_device_init(d, block_bytes(c), u->sectors))
goto err; goto err;
bcache_device_attach(d, c, u - c->uuids); bcache_device_attach(d, c, u - c->uuids);
set_capacity(d->disk, u->sectors);
bch_flash_dev_request_init(d); bch_flash_dev_request_init(d);
add_disk(d->disk); add_disk(d->disk);
......
...@@ -9,6 +9,7 @@ ...@@ -9,6 +9,7 @@
#include "sysfs.h" #include "sysfs.h"
#include "btree.h" #include "btree.h"
#include "request.h" #include "request.h"
#include "writeback.h"
#include <linux/blkdev.h> #include <linux/blkdev.h>
#include <linux/sort.h> #include <linux/sort.h>
...@@ -128,7 +129,7 @@ SHOW(__bch_cached_dev) ...@@ -128,7 +129,7 @@ SHOW(__bch_cached_dev)
char derivative[20]; char derivative[20];
char target[20]; char target[20];
bch_hprint(dirty, bch_hprint(dirty,
atomic_long_read(&dc->disk.sectors_dirty) << 9); bcache_dev_sectors_dirty(&dc->disk) << 9);
bch_hprint(derivative, dc->writeback_rate_derivative << 9); bch_hprint(derivative, dc->writeback_rate_derivative << 9);
bch_hprint(target, dc->writeback_rate_target << 9); bch_hprint(target, dc->writeback_rate_target << 9);
...@@ -144,7 +145,7 @@ SHOW(__bch_cached_dev) ...@@ -144,7 +145,7 @@ SHOW(__bch_cached_dev)
} }
sysfs_hprint(dirty_data, sysfs_hprint(dirty_data,
atomic_long_read(&dc->disk.sectors_dirty) << 9); bcache_dev_sectors_dirty(&dc->disk) << 9);
var_printf(sequential_merge, "%i"); var_printf(sequential_merge, "%i");
var_hprint(sequential_cutoff); var_hprint(sequential_cutoff);
......
...@@ -9,6 +9,7 @@ ...@@ -9,6 +9,7 @@
#include "bcache.h" #include "bcache.h"
#include "btree.h" #include "btree.h"
#include "debug.h" #include "debug.h"
#include "writeback.h"
#include <trace/events/bcache.h> #include <trace/events/bcache.h>
...@@ -38,7 +39,7 @@ static void __update_writeback_rate(struct cached_dev *dc) ...@@ -38,7 +39,7 @@ static void __update_writeback_rate(struct cached_dev *dc)
int change = 0; int change = 0;
int64_t error; int64_t error;
int64_t dirty = atomic_long_read(&dc->disk.sectors_dirty); int64_t dirty = bcache_dev_sectors_dirty(&dc->disk);
int64_t derivative = dirty - dc->disk.sectors_dirty_last; int64_t derivative = dirty - dc->disk.sectors_dirty_last;
dc->disk.sectors_dirty_last = dirty; dc->disk.sectors_dirty_last = dirty;
...@@ -183,10 +184,8 @@ void bch_writeback_queue(struct cached_dev *dc) ...@@ -183,10 +184,8 @@ void bch_writeback_queue(struct cached_dev *dc)
} }
} }
void bch_writeback_add(struct cached_dev *dc, unsigned sectors) void bch_writeback_add(struct cached_dev *dc)
{ {
atomic_long_add(sectors, &dc->disk.sectors_dirty);
if (!atomic_read(&dc->has_dirty) && if (!atomic_read(&dc->has_dirty) &&
!atomic_xchg(&dc->has_dirty, 1)) { !atomic_xchg(&dc->has_dirty, 1)) {
atomic_inc(&dc->count); atomic_inc(&dc->count);
...@@ -205,6 +204,34 @@ void bch_writeback_add(struct cached_dev *dc, unsigned sectors) ...@@ -205,6 +204,34 @@ void bch_writeback_add(struct cached_dev *dc, unsigned sectors)
} }
} }
void bcache_dev_sectors_dirty_add(struct cache_set *c, unsigned inode,
uint64_t offset, int nr_sectors)
{
struct bcache_device *d = c->devices[inode];
unsigned stripe_size, stripe_offset;
uint64_t stripe;
if (!d)
return;
stripe_size = 1 << d->stripe_size_bits;
stripe = offset >> d->stripe_size_bits;
stripe_offset = offset & (stripe_size - 1);
while (nr_sectors) {
int s = min_t(unsigned, abs(nr_sectors),
stripe_size - stripe_offset);
if (nr_sectors < 0)
s = -s;
atomic_add(s, d->stripe_sectors_dirty + stripe);
nr_sectors -= s;
stripe_offset = 0;
stripe++;
}
}
/* Background writeback - IO loop */ /* Background writeback - IO loop */
static void dirty_io_destructor(struct closure *cl) static void dirty_io_destructor(struct closure *cl)
...@@ -392,8 +419,9 @@ static int bch_btree_sectors_dirty_init(struct btree *b, struct btree_op *op, ...@@ -392,8 +419,9 @@ static int bch_btree_sectors_dirty_init(struct btree *b, struct btree_op *op,
break; break;
if (KEY_DIRTY(k)) if (KEY_DIRTY(k))
atomic_long_add(KEY_SIZE(k), bcache_dev_sectors_dirty_add(b->c, dc->disk.id,
&dc->disk.sectors_dirty); KEY_START(k),
KEY_SIZE(k));
} else { } else {
btree(sectors_dirty_init, k, b, op, dc); btree(sectors_dirty_init, k, b, op, dc);
if (KEY_INODE(k) > dc->disk.id) if (KEY_INODE(k) > dc->disk.id)
......
#ifndef _BCACHE_WRITEBACK_H
#define _BCACHE_WRITEBACK_H
static inline uint64_t bcache_dev_sectors_dirty(struct bcache_device *d)
{
uint64_t i, ret = 0;
for (i = 0; i < d->nr_stripes; i++)
ret += atomic_read(d->stripe_sectors_dirty + i);
return ret;
}
void bcache_dev_sectors_dirty_add(struct cache_set *, unsigned, uint64_t, int);
void bch_writeback_queue(struct cached_dev *);
void bch_writeback_add(struct cached_dev *);
void bch_sectors_dirty_init(struct cached_dev *dc);
void bch_cached_dev_writeback_init(struct cached_dev *);
#endif
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment