Commit 5e6926da authored by Kent Overstreet's avatar Kent Overstreet

bcache: Convert writeback to a kthread

This simplifies the writeback flow control quite a bit - previously, it
was conceptually two coroutines, refill_dirty() and read_dirty(). This
makes the code quite a bit more straightforward.
Signed-off-by: default avatarKent Overstreet <kmo@daterainc.com>
parent 72a44517
...@@ -509,7 +509,7 @@ struct cached_dev { ...@@ -509,7 +509,7 @@ struct cached_dev {
/* Limit number of writeback bios in flight */ /* Limit number of writeback bios in flight */
struct semaphore in_flight; struct semaphore in_flight;
struct closure_with_timer writeback; struct task_struct *writeback_thread;
struct keybuf writeback_keys; struct keybuf writeback_keys;
...@@ -1038,7 +1038,11 @@ static inline void bkey_init(struct bkey *k) ...@@ -1038,7 +1038,11 @@ static inline void bkey_init(struct bkey *k)
#define KEY_START(k) (KEY_OFFSET(k) - KEY_SIZE(k)) #define KEY_START(k) (KEY_OFFSET(k) - KEY_SIZE(k))
#define START_KEY(k) KEY(KEY_INODE(k), KEY_START(k), 0) #define START_KEY(k) KEY(KEY_INODE(k), KEY_START(k), 0)
#define MAX_KEY KEY(~(~0 << 20), ((uint64_t) ~0) >> 1, 0)
#define MAX_KEY_INODE (~(~0 << 20))
#define MAX_KEY_OFFSET (((uint64_t) ~0) >> 1)
#define MAX_KEY KEY(MAX_KEY_INODE, MAX_KEY_OFFSET, 0)
#define ZERO_KEY KEY(0, 0, 0) #define ZERO_KEY KEY(0, 0, 0)
/* /*
...@@ -1214,8 +1218,6 @@ int bch_cache_allocator_init(struct cache *ca); ...@@ -1214,8 +1218,6 @@ int bch_cache_allocator_init(struct cache *ca);
void bch_debug_exit(void); void bch_debug_exit(void);
int bch_debug_init(struct kobject *); int bch_debug_init(struct kobject *);
void bch_writeback_exit(void);
int bch_writeback_init(void);
void bch_request_exit(void); void bch_request_exit(void);
int bch_request_init(void); int bch_request_init(void);
void bch_btree_exit(void); void bch_btree_exit(void);
......
...@@ -1029,6 +1029,7 @@ static void cached_dev_free(struct closure *cl) ...@@ -1029,6 +1029,7 @@ static void cached_dev_free(struct closure *cl)
struct cached_dev *dc = container_of(cl, struct cached_dev, disk.cl); struct cached_dev *dc = container_of(cl, struct cached_dev, disk.cl);
cancel_delayed_work_sync(&dc->writeback_rate_update); cancel_delayed_work_sync(&dc->writeback_rate_update);
kthread_stop(dc->writeback_thread);
mutex_lock(&bch_register_lock); mutex_lock(&bch_register_lock);
...@@ -2006,7 +2007,6 @@ static struct notifier_block reboot = { ...@@ -2006,7 +2007,6 @@ static struct notifier_block reboot = {
static void bcache_exit(void) static void bcache_exit(void)
{ {
bch_debug_exit(); bch_debug_exit();
bch_writeback_exit();
bch_request_exit(); bch_request_exit();
bch_btree_exit(); bch_btree_exit();
if (bcache_kobj) if (bcache_kobj)
...@@ -2039,7 +2039,6 @@ static int __init bcache_init(void) ...@@ -2039,7 +2039,6 @@ static int __init bcache_init(void)
sysfs_create_files(bcache_kobj, files) || sysfs_create_files(bcache_kobj, files) ||
bch_btree_init() || bch_btree_init() ||
bch_request_init() || bch_request_init() ||
bch_writeback_init() ||
bch_debug_init(bcache_kobj)) bch_debug_init(bcache_kobj))
goto err; goto err;
......
...@@ -11,18 +11,11 @@ ...@@ -11,18 +11,11 @@
#include "debug.h" #include "debug.h"
#include "writeback.h" #include "writeback.h"
#include <linux/delay.h>
#include <linux/freezer.h>
#include <linux/kthread.h>
#include <trace/events/bcache.h> #include <trace/events/bcache.h>
static struct workqueue_struct *dirty_wq;
static void read_dirty(struct closure *);
struct dirty_io {
struct closure cl;
struct cached_dev *dc;
struct bio bio;
};
/* Rate limiting */ /* Rate limiting */
static void __update_writeback_rate(struct cached_dev *dc) static void __update_writeback_rate(struct cached_dev *dc)
...@@ -72,9 +65,6 @@ static void __update_writeback_rate(struct cached_dev *dc) ...@@ -72,9 +65,6 @@ static void __update_writeback_rate(struct cached_dev *dc)
dc->writeback_rate_derivative = derivative; dc->writeback_rate_derivative = derivative;
dc->writeback_rate_change = change; dc->writeback_rate_change = change;
dc->writeback_rate_target = target; dc->writeback_rate_target = target;
schedule_delayed_work(&dc->writeback_rate_update,
dc->writeback_rate_update_seconds * HZ);
} }
static void update_writeback_rate(struct work_struct *work) static void update_writeback_rate(struct work_struct *work)
...@@ -90,6 +80,9 @@ static void update_writeback_rate(struct work_struct *work) ...@@ -90,6 +80,9 @@ static void update_writeback_rate(struct work_struct *work)
__update_writeback_rate(dc); __update_writeback_rate(dc);
up_read(&dc->writeback_lock); up_read(&dc->writeback_lock);
schedule_delayed_work(&dc->writeback_rate_update,
dc->writeback_rate_update_seconds * HZ);
} }
static unsigned writeback_delay(struct cached_dev *dc, unsigned sectors) static unsigned writeback_delay(struct cached_dev *dc, unsigned sectors)
...@@ -105,37 +98,11 @@ static unsigned writeback_delay(struct cached_dev *dc, unsigned sectors) ...@@ -105,37 +98,11 @@ static unsigned writeback_delay(struct cached_dev *dc, unsigned sectors)
return min_t(uint64_t, ret, HZ); return min_t(uint64_t, ret, HZ);
} }
/* Background writeback */ struct dirty_io {
struct closure cl;
static bool dirty_pred(struct keybuf *buf, struct bkey *k) struct cached_dev *dc;
{ struct bio bio;
return KEY_DIRTY(k); };
}
static bool dirty_full_stripe_pred(struct keybuf *buf, struct bkey *k)
{
uint64_t stripe = KEY_START(k);
unsigned nr_sectors = KEY_SIZE(k);
struct cached_dev *dc = container_of(buf, struct cached_dev,
writeback_keys);
if (!KEY_DIRTY(k))
return false;
do_div(stripe, dc->disk.stripe_size);
while (1) {
if (atomic_read(dc->disk.stripe_sectors_dirty + stripe) ==
dc->disk.stripe_size)
return true;
if (nr_sectors <= dc->disk.stripe_size)
return false;
nr_sectors -= dc->disk.stripe_size;
stripe++;
}
}
static void dirty_init(struct keybuf_key *w) static void dirty_init(struct keybuf_key *w)
{ {
...@@ -153,132 +120,6 @@ static void dirty_init(struct keybuf_key *w) ...@@ -153,132 +120,6 @@ static void dirty_init(struct keybuf_key *w)
bch_bio_map(bio, NULL); bch_bio_map(bio, NULL);
} }
static void refill_dirty(struct closure *cl)
{
struct cached_dev *dc = container_of(cl, struct cached_dev,
writeback.cl);
struct keybuf *buf = &dc->writeback_keys;
bool searched_from_start = false;
struct bkey end = MAX_KEY;
SET_KEY_INODE(&end, dc->disk.id);
if (!atomic_read(&dc->disk.detaching) &&
!dc->writeback_running)
closure_return(cl);
down_write(&dc->writeback_lock);
if (!atomic_read(&dc->has_dirty)) {
SET_BDEV_STATE(&dc->sb, BDEV_STATE_CLEAN);
bch_write_bdev_super(dc, NULL);
up_write(&dc->writeback_lock);
closure_return(cl);
}
if (bkey_cmp(&buf->last_scanned, &end) >= 0) {
buf->last_scanned = KEY(dc->disk.id, 0, 0);
searched_from_start = true;
}
if (dc->partial_stripes_expensive) {
uint64_t i;
for (i = 0; i < dc->disk.nr_stripes; i++)
if (atomic_read(dc->disk.stripe_sectors_dirty + i) ==
dc->disk.stripe_size)
goto full_stripes;
goto normal_refill;
full_stripes:
searched_from_start = false; /* not searching entire btree */
bch_refill_keybuf(dc->disk.c, buf, &end,
dirty_full_stripe_pred);
} else {
normal_refill:
bch_refill_keybuf(dc->disk.c, buf, &end, dirty_pred);
}
if (bkey_cmp(&buf->last_scanned, &end) >= 0 && searched_from_start) {
/* Searched the entire btree - delay awhile */
if (RB_EMPTY_ROOT(&buf->keys)) {
atomic_set(&dc->has_dirty, 0);
cached_dev_put(dc);
}
if (!atomic_read(&dc->disk.detaching))
closure_delay(&dc->writeback, dc->writeback_delay * HZ);
}
up_write(&dc->writeback_lock);
bch_ratelimit_reset(&dc->writeback_rate);
/* Punt to workqueue only so we don't recurse and blow the stack */
continue_at(cl, read_dirty, dirty_wq);
}
void bch_writeback_queue(struct cached_dev *dc)
{
if (closure_trylock(&dc->writeback.cl, &dc->disk.cl)) {
if (!atomic_read(&dc->disk.detaching))
closure_delay(&dc->writeback, dc->writeback_delay * HZ);
continue_at(&dc->writeback.cl, refill_dirty, dirty_wq);
}
}
void bch_writeback_add(struct cached_dev *dc)
{
if (!atomic_read(&dc->has_dirty) &&
!atomic_xchg(&dc->has_dirty, 1)) {
atomic_inc(&dc->count);
if (BDEV_STATE(&dc->sb) != BDEV_STATE_DIRTY) {
SET_BDEV_STATE(&dc->sb, BDEV_STATE_DIRTY);
/* XXX: should do this synchronously */
bch_write_bdev_super(dc, NULL);
}
bch_writeback_queue(dc);
if (dc->writeback_percent)
schedule_delayed_work(&dc->writeback_rate_update,
dc->writeback_rate_update_seconds * HZ);
}
}
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_offset;
uint64_t stripe = offset;
if (!d)
return;
do_div(stripe, d->stripe_size);
stripe_offset = offset & (d->stripe_size - 1);
while (nr_sectors) {
int s = min_t(unsigned, abs(nr_sectors),
d->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 */
static void dirty_io_destructor(struct closure *cl) static void dirty_io_destructor(struct closure *cl)
{ {
struct dirty_io *io = container_of(cl, struct dirty_io, cl); struct dirty_io *io = container_of(cl, struct dirty_io, cl);
...@@ -378,30 +219,33 @@ static void read_dirty_submit(struct closure *cl) ...@@ -378,30 +219,33 @@ static void read_dirty_submit(struct closure *cl)
continue_at(cl, write_dirty, system_wq); continue_at(cl, write_dirty, system_wq);
} }
static void read_dirty(struct closure *cl) static void read_dirty(struct cached_dev *dc)
{ {
struct cached_dev *dc = container_of(cl, struct cached_dev, unsigned delay = 0;
writeback.cl);
unsigned delay = writeback_delay(dc, 0);
struct keybuf_key *w; struct keybuf_key *w;
struct dirty_io *io; struct dirty_io *io;
struct closure cl;
closure_init_stack(&cl);
/* /*
* XXX: if we error, background writeback just spins. Should use some * XXX: if we error, background writeback just spins. Should use some
* mempools. * mempools.
*/ */
while (1) { while (!kthread_should_stop()) {
try_to_freeze();
w = bch_keybuf_next(&dc->writeback_keys); w = bch_keybuf_next(&dc->writeback_keys);
if (!w) if (!w)
break; break;
BUG_ON(ptr_stale(dc->disk.c, &w->key, 0)); BUG_ON(ptr_stale(dc->disk.c, &w->key, 0));
if (delay > 0 && if (KEY_START(&w->key) != dc->last_read ||
(KEY_START(&w->key) != dc->last_read || jiffies_to_msecs(delay) > 50)
jiffies_to_msecs(delay) > 50)) while (!kthread_should_stop() && delay)
delay = schedule_timeout_uninterruptible(delay); delay = schedule_timeout_interruptible(delay);
dc->last_read = KEY_OFFSET(&w->key); dc->last_read = KEY_OFFSET(&w->key);
...@@ -427,7 +271,7 @@ static void read_dirty(struct closure *cl) ...@@ -427,7 +271,7 @@ static void read_dirty(struct closure *cl)
trace_bcache_writeback(&w->key); trace_bcache_writeback(&w->key);
down(&dc->in_flight); down(&dc->in_flight);
closure_call(&io->cl, read_dirty_submit, NULL, cl); closure_call(&io->cl, read_dirty_submit, NULL, &cl);
delay = writeback_delay(dc, KEY_SIZE(&w->key)); delay = writeback_delay(dc, KEY_SIZE(&w->key));
} }
...@@ -443,7 +287,148 @@ static void read_dirty(struct closure *cl) ...@@ -443,7 +287,148 @@ static void read_dirty(struct closure *cl)
* Wait for outstanding writeback IOs to finish (and keybuf slots to be * Wait for outstanding writeback IOs to finish (and keybuf slots to be
* freed) before refilling again * freed) before refilling again
*/ */
continue_at(cl, refill_dirty, dirty_wq); closure_sync(&cl);
}
/* Scan for dirty data */
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_offset;
uint64_t stripe = offset;
if (!d)
return;
do_div(stripe, d->stripe_size);
stripe_offset = offset & (d->stripe_size - 1);
while (nr_sectors) {
int s = min_t(unsigned, abs(nr_sectors),
d->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++;
}
}
static bool dirty_pred(struct keybuf *buf, struct bkey *k)
{
return KEY_DIRTY(k);
}
static bool dirty_full_stripe_pred(struct keybuf *buf, struct bkey *k)
{
uint64_t stripe = KEY_START(k);
unsigned nr_sectors = KEY_SIZE(k);
struct cached_dev *dc = container_of(buf, struct cached_dev,
writeback_keys);
if (!KEY_DIRTY(k))
return false;
do_div(stripe, dc->disk.stripe_size);
while (1) {
if (atomic_read(dc->disk.stripe_sectors_dirty + stripe) ==
dc->disk.stripe_size)
return true;
if (nr_sectors <= dc->disk.stripe_size)
return false;
nr_sectors -= dc->disk.stripe_size;
stripe++;
}
}
static bool refill_dirty(struct cached_dev *dc)
{
struct keybuf *buf = &dc->writeback_keys;
bool searched_from_start = false;
struct bkey end = KEY(dc->disk.id, MAX_KEY_OFFSET, 0);
if (bkey_cmp(&buf->last_scanned, &end) >= 0) {
buf->last_scanned = KEY(dc->disk.id, 0, 0);
searched_from_start = true;
}
if (dc->partial_stripes_expensive) {
uint64_t i;
for (i = 0; i < dc->disk.nr_stripes; i++)
if (atomic_read(dc->disk.stripe_sectors_dirty + i) ==
dc->disk.stripe_size)
goto full_stripes;
goto normal_refill;
full_stripes:
searched_from_start = false; /* not searching entire btree */
bch_refill_keybuf(dc->disk.c, buf, &end,
dirty_full_stripe_pred);
} else {
normal_refill:
bch_refill_keybuf(dc->disk.c, buf, &end, dirty_pred);
}
return bkey_cmp(&buf->last_scanned, &end) >= 0 && searched_from_start;
}
static int bch_writeback_thread(void *arg)
{
struct cached_dev *dc = arg;
bool searched_full_index;
while (!kthread_should_stop()) {
down_write(&dc->writeback_lock);
if (!atomic_read(&dc->has_dirty) ||
(!atomic_read(&dc->disk.detaching) &&
!dc->writeback_running)) {
up_write(&dc->writeback_lock);
set_current_state(TASK_INTERRUPTIBLE);
if (kthread_should_stop())
return 0;
try_to_freeze();
schedule();
continue;
}
searched_full_index = refill_dirty(dc);
if (searched_full_index &&
RB_EMPTY_ROOT(&dc->writeback_keys.keys)) {
atomic_set(&dc->has_dirty, 0);
cached_dev_put(dc);
SET_BDEV_STATE(&dc->sb, BDEV_STATE_CLEAN);
bch_write_bdev_super(dc, NULL);
}
up_write(&dc->writeback_lock);
bch_ratelimit_reset(&dc->writeback_rate);
read_dirty(dc);
if (searched_full_index) {
unsigned delay = dc->writeback_delay * HZ;
while (delay &&
!kthread_should_stop() &&
!atomic_read(&dc->disk.detaching))
delay = schedule_timeout_interruptible(delay);
}
}
return 0;
} }
/* Init */ /* Init */
...@@ -483,12 +468,10 @@ void bch_sectors_dirty_init(struct cached_dev *dc) ...@@ -483,12 +468,10 @@ void bch_sectors_dirty_init(struct cached_dev *dc)
btree_root(sectors_dirty_init, dc->disk.c, &op, dc); btree_root(sectors_dirty_init, dc->disk.c, &op, dc);
} }
void bch_cached_dev_writeback_init(struct cached_dev *dc) int bch_cached_dev_writeback_init(struct cached_dev *dc)
{ {
sema_init(&dc->in_flight, 64); sema_init(&dc->in_flight, 64);
closure_init_unlocked(&dc->writeback);
init_rwsem(&dc->writeback_lock); init_rwsem(&dc->writeback_lock);
bch_keybuf_init(&dc->writeback_keys); bch_keybuf_init(&dc->writeback_keys);
dc->writeback_metadata = true; dc->writeback_metadata = true;
...@@ -502,22 +485,16 @@ void bch_cached_dev_writeback_init(struct cached_dev *dc) ...@@ -502,22 +485,16 @@ void bch_cached_dev_writeback_init(struct cached_dev *dc)
dc->writeback_rate_p_term_inverse = 64; dc->writeback_rate_p_term_inverse = 64;
dc->writeback_rate_d_smooth = 8; dc->writeback_rate_d_smooth = 8;
dc->writeback_thread = kthread_create(bch_writeback_thread, dc,
"bcache_writeback");
if (IS_ERR(dc->writeback_thread))
return PTR_ERR(dc->writeback_thread);
set_task_state(dc->writeback_thread, TASK_INTERRUPTIBLE);
INIT_DELAYED_WORK(&dc->writeback_rate_update, update_writeback_rate); INIT_DELAYED_WORK(&dc->writeback_rate_update, update_writeback_rate);
schedule_delayed_work(&dc->writeback_rate_update, schedule_delayed_work(&dc->writeback_rate_update,
dc->writeback_rate_update_seconds * HZ); dc->writeback_rate_update_seconds * HZ);
}
void bch_writeback_exit(void)
{
if (dirty_wq)
destroy_workqueue(dirty_wq);
}
int __init bch_writeback_init(void)
{
dirty_wq = create_workqueue("bcache_writeback");
if (!dirty_wq)
return -ENOMEM;
return 0; return 0;
} }
...@@ -56,11 +56,30 @@ static inline bool should_writeback(struct cached_dev *dc, struct bio *bio, ...@@ -56,11 +56,30 @@ static inline bool should_writeback(struct cached_dev *dc, struct bio *bio,
in_use <= CUTOFF_WRITEBACK; in_use <= CUTOFF_WRITEBACK;
} }
static inline void bch_writeback_queue(struct cached_dev *dc)
{
wake_up_process(dc->writeback_thread);
}
static inline void bch_writeback_add(struct cached_dev *dc)
{
if (!atomic_read(&dc->has_dirty) &&
!atomic_xchg(&dc->has_dirty, 1)) {
atomic_inc(&dc->count);
if (BDEV_STATE(&dc->sb) != BDEV_STATE_DIRTY) {
SET_BDEV_STATE(&dc->sb, BDEV_STATE_DIRTY);
/* XXX: should do this synchronously */
bch_write_bdev_super(dc, NULL);
}
bch_writeback_queue(dc);
}
}
void bcache_dev_sectors_dirty_add(struct cache_set *, unsigned, uint64_t, int); 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_sectors_dirty_init(struct cached_dev *dc);
void bch_cached_dev_writeback_init(struct cached_dev *); int bch_cached_dev_writeback_init(struct cached_dev *);
#endif #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