After month of usage I can confirm that this patch resolve this issue.
--
Best regards,
Konstantin Shalygin
From: Tang Junhui <tang.junhui@xxxxxxxxxx>
gc and write-back get raced (see the email "bcache get stucked" I sended
before):
gc thread write-back thread
| |bch_writeback_thread()
|bch_gc_thread() |
| |==>read_dirty()
|==>bch_btree_gc() |
|==>btree_root() //get btree root |
| node write locker |
|==>bch_btree_gc_root() |
| |==>read_dirty_submit()
| |==>write_dirty()
| |==>continue_at(cl, write_dirty_finish, system_wq);
| |==>write_dirty_finish()//excute in system_wq
| |==>bch_btree_insert()
| |==>bch_btree_map_leaf_nodes()
| |==>__bch_btree_map_nodes()
| |==>btree_root //try to get btree root node read lock
| |-----stuck here
|==>bch_btree_set_root() |
|==>bch_journal_meta() |
|==>bch_journal() |
|==>journal_try_write() |
|==>journal_write_unlocked() //journal_full(&c->journal) condition satisfied
|==>continue_at(cl, journal_write, system_wq); //try to excute journal_write in system_wq
| //but work queue is excuting write_dirty_finish()
|==>closure_sync(); //wait journal_write execute over and wake up gc,
| --stuck here
|==>release root node write locker
This patch alloc a separate work-queue for write-back thread to avoid such
race.
Signed-off-by: Tang Junhui <tang.junhui@xxxxxxxxxx>
---
diff -Naupr linux-4.11_orig/drivers/md/bcache/bcache.h linux-4.11/drivers/md/bcache/bcache.h
--- linux-4.11_orig/drivers/md/bcache/bcache.h 2017-05-01 09:47:48.000000000 +0700
+++ linux-4.11/drivers/md/bcache/bcache.h 2017-07-08 11:49:25.044671097 +0700
@@ -333,6 +333,7 @@ struct cached_dev {
/* Limit number of writeback bios in flight */
struct semaphore in_flight;
struct task_struct *writeback_thread;
+ struct workqueue_struct *writeback_write_wq;
struct keybuf writeback_keys;
diff -Naupr linux-4.11_orig/drivers/md/bcache/super.c linux-4.11/drivers/md/bcache/super.c
--- linux-4.11_orig/drivers/md/bcache/super.c 2017-05-01 09:47:48.000000000 +0700
+++ linux-4.11/drivers/md/bcache/super.c 2017-07-08 11:50:16.574952148 +0700
@@ -1061,6 +1061,8 @@ static void cached_dev_free(struct closu
cancel_delayed_work_sync(&dc->writeback_rate_update);
if (!IS_ERR_OR_NULL(dc->writeback_thread))
kthread_stop(dc->writeback_thread);
+ if (dc->writeback_write_wq)
+ destroy_workqueue(dc->writeback_write_wq);
mutex_lock(&bch_register_lock);
diff -Naupr linux-4.11_orig/drivers/md/bcache/writeback.c linux-4.11/drivers/md/bcache/writeback.c
--- linux-4.11_orig/drivers/md/bcache/writeback.c 2017-05-01 09:47:48.000000000 +0700
+++ linux-4.11/drivers/md/bcache/writeback.c 2017-07-08 11:52:21.799010810 +0700
@@ -186,7 +186,7 @@ static void write_dirty(struct closure *
closure_bio_submit(&io->bio, cl);
- continue_at(cl, write_dirty_finish, system_wq);
+ continue_at(cl, write_dirty_finish, io->dc->writeback_write_wq);
}
static void read_dirty_endio(struct bio *bio)
@@ -206,7 +206,7 @@ static void read_dirty_submit(struct clo
closure_bio_submit(&io->bio, cl);
- continue_at(cl, write_dirty, system_wq);
+ continue_at(cl, write_dirty, io->dc->writeback_write_wq);
}
static void read_dirty(struct cached_dev *dc)
@@ -516,6 +516,10 @@ void bch_cached_dev_writeback_init(struc
int bch_cached_dev_writeback_start(struct cached_dev *dc)
{
+ dc->writeback_write_wq = alloc_workqueue("bcache_writeback_wq", WQ_MEM_RECLAIM, 0);
+ if (!dc->writeback_write_wq)
+ return -ENOMEM;
+
dc->writeback_thread = kthread_create(bch_writeback_thread, dc,
"bcache_writeback");
if (IS_ERR(dc->writeback_thread))
