Btrfs: fix lockdep deadlock warning due to dev_replace

Xfstests btrfs/011 complains about a deadlock warning,

[ 1226.649039] =========================================================
[ 1226.649039] [ INFO: possible irq lock inversion dependency detected ]
[ 1226.649039] 4.1.0+ #270 Not tainted
[ 1226.649039] ---------------------------------------------------------
[ 1226.652955] kswapd0/46 just changed the state of lock:
[ 1226.652955]  (&delayed_node->mutex){+.+.-.}, at: [<ffffffff81458735>] __btrfs_release_delayed_node+0x45/0x1d0
[ 1226.652955] but this lock took another, RECLAIM_FS-unsafe lock in the past:
[ 1226.652955]  (&fs_info->dev_replace.lock){+.+.+.}

and interrupts could create inverse lock ordering between them.

[ 1226.652955]
other info that might help us debug this:
[ 1226.652955] Chain exists of:
  &delayed_node->mutex --> &found->groups_sem --> &fs_info->dev_replace.lock

[ 1226.652955]  Possible interrupt unsafe locking scenario:

[ 1226.652955]        CPU0                    CPU1
[ 1226.652955]        ----                    ----
[ 1226.652955]   lock(&fs_info->dev_replace.lock);
[ 1226.652955]                                local_irq_disable();
[ 1226.652955]                                lock(&delayed_node->mutex);
[ 1226.652955]                                lock(&found->groups_sem);
[ 1226.652955]   <Interrupt>
[ 1226.652955]     lock(&delayed_node->mutex);
[ 1226.652955]
 *** DEADLOCK ***

Commit 084b6e7c7607 ("btrfs: Fix a lockdep warning when running xfstest.") tried
to fix a similar one that has the exactly same warning, but with that, we still
run to this.

The above lock chain comes from
btrfs_commit_transaction
  ->btrfs_run_delayed_items
    ...
    ->__btrfs_update_delayed_inode
      ...
      ->__btrfs_cow_block
         ...
         ->find_free_extent
            ->cache_block_group
              ->load_free_space_cache
                ->btrfs_readpages
                  ->submit_one_bio
                    ...
                    ->__btrfs_map_block
                      ->btrfs_dev_replace_lock

However, with high memory pressure, tasks which hold dev_replace.lock can
be interrupted by kswapd and then kswapd is intended to release memory occupied
by superblock, inodes and dentries, where we may call evict_inode, and it comes
to

[ 1226.652955]  [<ffffffff81458735>] __btrfs_release_delayed_node+0x45/0x1d0
[ 1226.652955]  [<ffffffff81459e74>] btrfs_remove_delayed_node+0x24/0x30
[ 1226.652955]  [<ffffffff8140c5fe>] btrfs_evict_inode+0x34e/0x700

delayed_node->mutex may be acquired in __btrfs_release_delayed_node(), and it leads
to a ABBA deadlock.

To fix this, we can use "blocking rwlock" used in the case of extent_buffer, but
things are simpler here since we only needs read's spinlock to blocking lock.

With this, btrfs/011 no more produces warnings in dmesg.

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>

1 files changed, 70 insertions, 60 deletions

diff --git a/fs/btrfs/dev-replace.c b/fs/btrfs/dev-replace.c
index cbb7dbfb3fff..8c8b48971bc7 100644
--- a/fs/btrfs/dev-replace.c
+++ b/fs/btrfs/dev-replace.c
@@ -202,13 +202,13 @@ int btrfs_run_dev_replace(struct btrfs_trans_handle *trans,
         struct btrfs_dev_replace_item *ptr;
         struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
 
-        btrfs_dev_replace_lock(dev_replace);
+        btrfs_dev_replace_lock(dev_replace, 0);
         if (!dev_replace->is_valid ||
             !dev_replace->item_needs_writeback) {
-                btrfs_dev_replace_unlock(dev_replace);
+                btrfs_dev_replace_unlock(dev_replace, 0);
                 return 0;
         }
-        btrfs_dev_replace_unlock(dev_replace);
+        btrfs_dev_replace_unlock(dev_replace, 0);
 
         key.objectid = 0;
         key.type = BTRFS_DEV_REPLACE_KEY;
@@ -264,7 +264,7 @@ int btrfs_run_dev_replace(struct btrfs_trans_handle *trans,
         ptr = btrfs_item_ptr(eb, path->slots[0],
                              struct btrfs_dev_replace_item);
 
-        btrfs_dev_replace_lock(dev_replace);
+        btrfs_dev_replace_lock(dev_replace, 1);
         if (dev_replace->srcdev)
                 btrfs_set_dev_replace_src_devid(eb, ptr,
                         dev_replace->srcdev->devid);
@@ -287,7 +287,7 @@ int btrfs_run_dev_replace(struct btrfs_trans_handle *trans,
         btrfs_set_dev_replace_cursor_right(eb, ptr,
                 dev_replace->cursor_right);
         dev_replace->item_needs_writeback = 0;
-        btrfs_dev_replace_unlock(dev_replace);
+        btrfs_dev_replace_unlock(dev_replace, 1);
 
         btrfs_mark_buffer_dirty(eb);
 
@@ -356,7 +356,7 @@ int btrfs_dev_replace_start(struct btrfs_root *root,
                 return PTR_ERR(trans);
         }
 
-        btrfs_dev_replace_lock(dev_replace);
+        btrfs_dev_replace_lock(dev_replace, 1);
         switch (dev_replace->replace_state) {
         case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
         case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
@@ -395,7 +395,7 @@ int btrfs_dev_replace_start(struct btrfs_root *root,
         dev_replace->is_valid = 1;
         dev_replace->item_needs_writeback = 1;
         args->result = BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR;
-        btrfs_dev_replace_unlock(dev_replace);
+        btrfs_dev_replace_unlock(dev_replace, 1);
 
         ret = btrfs_sysfs_add_device_link(tgt_device->fs_devices, tgt_device);
         if (ret)
@@ -407,7 +407,7 @@ int btrfs_dev_replace_start(struct btrfs_root *root,
         trans = btrfs_start_transaction(root, 0);
         if (IS_ERR(trans)) {
                 ret = PTR_ERR(trans);
-                btrfs_dev_replace_lock(dev_replace);
+                btrfs_dev_replace_lock(dev_replace, 1);
                 goto leave;
         }
 
@@ -433,7 +433,7 @@ int btrfs_dev_replace_start(struct btrfs_root *root,
 leave:
         dev_replace->srcdev = NULL;
         dev_replace->tgtdev = NULL;
-        btrfs_dev_replace_unlock(dev_replace);
+        btrfs_dev_replace_unlock(dev_replace, 1);
         btrfs_destroy_dev_replace_tgtdev(fs_info, tgt_device);
         return ret;
 }
@@ -471,18 +471,18 @@ static int btrfs_dev_replace_finishing(struct btrfs_fs_info *fs_info,
         /* don't allow cancel or unmount to disturb the finishing procedure */
         mutex_lock(&dev_replace->lock_finishing_cancel_unmount);
 
-        btrfs_dev_replace_lock(dev_replace);
+        btrfs_dev_replace_lock(dev_replace, 0);
         /* was the operation canceled, or is it finished? */
         if (dev_replace->replace_state !=
             BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED) {
-                btrfs_dev_replace_unlock(dev_replace);
+                btrfs_dev_replace_unlock(dev_replace, 0);
                 mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
                 return 0;
         }
 
         tgt_device = dev_replace->tgtdev;
         src_device = dev_replace->srcdev;
-        btrfs_dev_replace_unlock(dev_replace);
+        btrfs_dev_replace_unlock(dev_replace, 0);
 
         /*
          * flush all outstanding I/O and inode extent mappings before the
@@ -507,7 +507,7 @@ static int btrfs_dev_replace_finishing(struct btrfs_fs_info *fs_info,
         /* keep away write_all_supers() during the finishing procedure */
         mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
         mutex_lock(&root->fs_info->chunk_mutex);
-        btrfs_dev_replace_lock(dev_replace);
+        btrfs_dev_replace_lock(dev_replace, 1);
         dev_replace->replace_state =
                 scrub_ret ? BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
                           : BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED;
@@ -528,7 +528,7 @@ static int btrfs_dev_replace_finishing(struct btrfs_fs_info *fs_info,
                                 rcu_str_deref(src_device->name),
                               src_device->devid,
                               rcu_str_deref(tgt_device->name), scrub_ret);
-                btrfs_dev_replace_unlock(dev_replace);
+                btrfs_dev_replace_unlock(dev_replace, 1);
                 mutex_unlock(&root->fs_info->chunk_mutex);
                 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
                 mutex_unlock(&uuid_mutex);
@@ -565,7 +565,7 @@ static int btrfs_dev_replace_finishing(struct btrfs_fs_info *fs_info,
         list_add(&tgt_device->dev_alloc_list, &fs_info->fs_devices->alloc_list);
         fs_info->fs_devices->rw_devices++;
 
-        btrfs_dev_replace_unlock(dev_replace);
+        btrfs_dev_replace_unlock(dev_replace, 1);
 
         btrfs_rm_dev_replace_blocked(fs_info);
 
@@ -649,7 +649,7 @@ void btrfs_dev_replace_status(struct btrfs_fs_info *fs_info,
         struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
         struct btrfs_device *srcdev;
 
-        btrfs_dev_replace_lock(dev_replace);
+        btrfs_dev_replace_lock(dev_replace, 0);
         /* even if !dev_replace_is_valid, the values are good enough for
          * the replace_status ioctl */
         args->result = BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR;
@@ -675,7 +675,7 @@ void btrfs_dev_replace_status(struct btrfs_fs_info *fs_info,
                         div_u64(btrfs_device_get_total_bytes(srcdev), 1000));
                 break;
         }
-        btrfs_dev_replace_unlock(dev_replace);
+        btrfs_dev_replace_unlock(dev_replace, 0);
 }
 
 int btrfs_dev_replace_cancel(struct btrfs_fs_info *fs_info,
@@ -698,13 +698,13 @@ static u64 __btrfs_dev_replace_cancel(struct btrfs_fs_info *fs_info)
                 return -EROFS;
 
         mutex_lock(&dev_replace->lock_finishing_cancel_unmount);
-        btrfs_dev_replace_lock(dev_replace);
+        btrfs_dev_replace_lock(dev_replace, 1);
         switch (dev_replace->replace_state) {
         case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
         case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
         case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
                 result = BTRFS_IOCTL_DEV_REPLACE_RESULT_NOT_STARTED;
-                btrfs_dev_replace_unlock(dev_replace);
+                btrfs_dev_replace_unlock(dev_replace, 1);
                 goto leave;
         case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
         case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
@@ -717,7 +717,7 @@ static u64 __btrfs_dev_replace_cancel(struct btrfs_fs_info *fs_info)
         dev_replace->replace_state = BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED;
         dev_replace->time_stopped = get_seconds();
         dev_replace->item_needs_writeback = 1;
-        btrfs_dev_replace_unlock(dev_replace);
+        btrfs_dev_replace_unlock(dev_replace, 1);
         btrfs_scrub_cancel(fs_info);
 
         trans = btrfs_start_transaction(root, 0);
@@ -740,7 +740,7 @@ void btrfs_dev_replace_suspend_for_unmount(struct btrfs_fs_info *fs_info)
         struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
 
         mutex_lock(&dev_replace->lock_finishing_cancel_unmount);
-        btrfs_dev_replace_lock(dev_replace);
+        btrfs_dev_replace_lock(dev_replace, 1);
         switch (dev_replace->replace_state) {
         case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
         case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
@@ -756,7 +756,7 @@ void btrfs_dev_replace_suspend_for_unmount(struct btrfs_fs_info *fs_info)
                 break;
         }
 
-        btrfs_dev_replace_unlock(dev_replace);
+        btrfs_dev_replace_unlock(dev_replace, 1);
         mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
 }
 
@@ -766,12 +766,12 @@ int btrfs_resume_dev_replace_async(struct btrfs_fs_info *fs_info)
         struct task_struct *task;
         struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
 
-        btrfs_dev_replace_lock(dev_replace);
+        btrfs_dev_replace_lock(dev_replace, 1);
         switch (dev_replace->replace_state) {
         case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
         case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
         case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
-                btrfs_dev_replace_unlock(dev_replace);
+                btrfs_dev_replace_unlock(dev_replace, 1);
                 return 0;
         case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
                 break;
@@ -784,10 +784,10 @@ int btrfs_resume_dev_replace_async(struct btrfs_fs_info *fs_info)
                 btrfs_info(fs_info, "cannot continue dev_replace, tgtdev is missing");
                 btrfs_info(fs_info,
                         "you may cancel the operation after 'mount -o degraded'");
-                btrfs_dev_replace_unlock(dev_replace);
+                btrfs_dev_replace_unlock(dev_replace, 1);
                 return 0;
         }
-        btrfs_dev_replace_unlock(dev_replace);
+        btrfs_dev_replace_unlock(dev_replace, 1);
 
         WARN_ON(atomic_xchg(
                 &fs_info->mutually_exclusive_operation_running, 1));
@@ -865,48 +865,58 @@ int btrfs_dev_replace_is_ongoing(struct btrfs_dev_replace *dev_replace)
         return 1;
 }
 
-void btrfs_dev_replace_lock(struct btrfs_dev_replace *dev_replace)
+void btrfs_dev_replace_lock(struct btrfs_dev_replace *dev_replace, int rw)
 {
-        /* the beginning is just an optimization for the typical case */
-        if (atomic_read(&dev_replace->nesting_level) == 0) {
-acquire_lock:
-                /* this is not a nested case where the same thread
-                 * is trying to acqurire the same lock twice */
-                mutex_lock(&dev_replace->lock);
-                mutex_lock(&dev_replace->lock_management_lock);
-                dev_replace->lock_owner = current->pid;
-                atomic_inc(&dev_replace->nesting_level);
-                mutex_unlock(&dev_replace->lock_management_lock);
-                return;
+        if (rw == 1) {
+                /* write */
+again:
+                wait_event(dev_replace->read_lock_wq,
+                           atomic_read(&dev_replace->blocking_readers) == 0);
+                write_lock(&dev_replace->lock);
+                if (atomic_read(&dev_replace->blocking_readers)) {
+                        write_unlock(&dev_replace->lock);
+                        goto again;
+                }
+        } else {
+                read_lock(&dev_replace->lock);
+                atomic_inc(&dev_replace->read_locks);
         }
+}
 
-        mutex_lock(&dev_replace->lock_management_lock);
-        if (atomic_read(&dev_replace->nesting_level) > 0 &&
-            dev_replace->lock_owner == current->pid) {
-                WARN_ON(!mutex_is_locked(&dev_replace->lock));
-                atomic_inc(&dev_replace->nesting_level);
-                mutex_unlock(&dev_replace->lock_management_lock);
-                return;
+void btrfs_dev_replace_unlock(struct btrfs_dev_replace *dev_replace, int rw)
+{
+        if (rw == 1) {
+                /* write */
+                ASSERT(atomic_read(&dev_replace->blocking_readers) == 0);
+                write_unlock(&dev_replace->lock);
+        } else {
+                ASSERT(atomic_read(&dev_replace->read_locks) > 0);
+                atomic_dec(&dev_replace->read_locks);
+                read_unlock(&dev_replace->lock);
         }
+}
 
-        mutex_unlock(&dev_replace->lock_management_lock);
-        goto acquire_lock;
+/* inc blocking cnt and release read lock */
+void btrfs_dev_replace_set_lock_blocking(
+                                        struct btrfs_dev_replace *dev_replace)
+{
+        /* only set blocking for read lock */
+        ASSERT(atomic_read(&dev_replace->read_locks) > 0);
+        atomic_inc(&dev_replace->blocking_readers);
+        read_unlock(&dev_replace->lock);
 }
 
-void btrfs_dev_replace_unlock(struct btrfs_dev_replace *dev_replace)
+/* acquire read lock and dec blocking cnt */
+void btrfs_dev_replace_clear_lock_blocking(
+                                        struct btrfs_dev_replace *dev_replace)
 {
-        WARN_ON(!mutex_is_locked(&dev_replace->lock));
-        mutex_lock(&dev_replace->lock_management_lock);
-        WARN_ON(atomic_read(&dev_replace->nesting_level) < 1);
-        WARN_ON(dev_replace->lock_owner != current->pid);
-        atomic_dec(&dev_replace->nesting_level);
-        if (atomic_read(&dev_replace->nesting_level) == 0) {
-                dev_replace->lock_owner = 0;
-                mutex_unlock(&dev_replace->lock_management_lock);
-                mutex_unlock(&dev_replace->lock);
-        } else {
-                mutex_unlock(&dev_replace->lock_management_lock);
-        }
+        /* only set blocking for read lock */
+        ASSERT(atomic_read(&dev_replace->read_locks) > 0);
+        ASSERT(atomic_read(&dev_replace->blocking_readers) > 0);
+        read_lock(&dev_replace->lock);
+        if (atomic_dec_and_test(&dev_replace->blocking_readers) &&
+            waitqueue_active(&dev_replace->read_lock_wq))
+                wake_up(&dev_replace->read_lock_wq);
 }
 
 void btrfs_bio_counter_inc_noblocked(struct btrfs_fs_info *fs_info)