btrfs: shrink delalloc fixes
Subsystems (5)
cover btrfs: shrink delalloc fixes
Hello,
I backported the patch to switch us to using sync_inode() to our kernel inside Facebook to fix a deadlock when using the async delalloc shrinker threads. This uncovered a bunch of problems with how we shrink delalloc, as we use -o compress-force, and thus everything goes through the async compression threads.
I ripped out the async pages stuff because originally I had switched us to just writing the whole inode. This caused a performance regression, and so I switched us to calling sync_inode() twice to handle the async extent case. The problem is that sync_inode() can skip writing inodes sometimes, and thus we weren't properly waiting on the async helpers. We really do need to wait for the async delalloc pages to go down before continuing to shrink delalloc. There was also a race in how we woke up the async delalloc pages waiter which could be problematic.
And then finally there is our use of sync_inode(). It tries to be too clever for us, when in reality we want to make sure all pages are under writeback before we come back to the shrinking loop. I've added a small helper to give us this flexibilty and have switched us to that helper.
With these patches, and others that will be sent separately, the early ENOSPC problems we were experiencing have been eliminated. Thanks,
Josef Bacik (4):
btrfs: wait on async extents when flushing delalloc
btrfs: wake up async_delalloc_pages waiters after submit
fs: add a filemap_fdatawrite_wbc helper
btrfs: use the filemap_fdatawrite_wbc helper for delalloc shrinking
fs/btrfs/inode.c | 16 ++++++---------- fs/btrfs/space-info.c | 33 +++++++++++++++++++++++++++++++++ include/linux/fs.h | 2 ++ mm/filemap.c | 29 ++++++++++++++++++++++++----- 4 files changed, 65 insertions(+), 15 deletions(-)
Signature · 1 line
2.26.3
PATCH 1/4 btrfs: wait on async extents when flushing delalloc
I've been debugging an early ENOSPC problem in production and finally root caused it to this problem. When we switched to the per-inode flushing code I pulled out the async extent handling, because we were doing the correct thing by calling filemap_flush() if we had async extents set. This would properly wait on any async extents by locking the page in the second flush, thus making sure our ordered extents were properly set up.
However when I switched us back to page based flushing, I used sync_inode(), which allows us to pass in our own wbc. The problem here is that sync_inode() is smarter than the filemap_* helpers, it tries to avoid calling writepages at all. This means that our second call could skip calling do_writepages altogether, and thus not wait on the pagelock for the async helpers. This means we could come back before any ordered extents were created and then simply continue on in our flushing mechanisms and ENOSPC out when we have plenty of space to use.
Fix this by putting back the async pages logic in shrink_delalloc. This allows us to bulk write out everything that we need to, and then we can wait in one place for the async helpers to catch up, and then wait on any ordered extents that are created.
--- fs/btrfs/inode.c | 4 ---- fs/btrfs/space-info.c | 33 +++++++++++++++++++++++++++++++++ 2 files changed, 33 insertions(+), 4 deletions(-)
fs/btrfs/inode.c +0 -4
@@ -9681,10 +9681,6 @@ static int start_delalloc_inodes(struct btrfs_root *root, &work->work); }else { ret =sync_inode (inode, wbc); -if (!ret && -test_bit (BTRFS_INODE_HAS_ASYNC_EXTENT , - &BTRFS_I (inode)->runtime_flags)) - ret =sync_inode (inode, wbc);btrfs_add_delayed_iput (inode);if (ret || wbc->nr_to_write <=0 ) goto out;
fs/btrfs/space-info.c +33 -0
@@ -547,9 +547,42 @@ static void shrink_delalloc(struct btrfs_fs_info *fs_info,while ((delalloc_bytes || ordered_bytes) && loops <3 ) {u64 temp =min (delalloc_bytes, to_reclaim) >>PAGE_SHIFT ;long nr_pages =min_t (u64, temp,LONG_MAX ); +int async_pages;btrfs_start_delalloc_roots (fs_info, nr_pages, true); +/* +* We need to make sure any outstanding async pages are now +* processed before we continue. This is because things like +* sync_inode() try to be smart and skip writing if the inode is +* marked clean. We don't use filemap_fwrite for flushing +* because we want to control how many pages we write out at a +* time, thus this is the only safe way to make sure we've +* waited for outstanding compressed workers to have started +* their jobs and thus have ordered extents set up properly. +* Josef, when you think you are smart enough to remove this in +* the future, read this comment 4 times and explain in the +* commit message why it makes sense to remove it, and then +* delete the commit and don't remove it. +*/ + async_pages =atomic_read (&fs_info->async_delalloc_pages); +if (!async_pages) + goto skip_async; + +/* +* We don't want to wait forever, if we wrote less pages in this +* loop than we have outstanding, only wait for that number of +* pages, otherwise we can wait for all async pages to finish +* before continuing. +*/ +if (async_pages > nr_pages) + async_pages -= nr_pages; +else + async_pages =0 ; +wait_event (fs_info->async_submit_wait, +atomic_read (&fs_info->async_delalloc_pages) <= + async_pages); +skip_async: loops++;if (wait_ordered && !trans) {btrfs_wait_ordered_roots (fs_info, items,0 , (u64)-1 );
Signature · 1 line
2.26.3
Re: [PATCH 1/4] btrfs: wait on async extents when flushing delalloc
Quoted 7 lines · Josef Bacik
I've been debugging an early ENOSPC problem in production and finally root caused it to this problem. When we switched to the per-inode flushing code I pulled out the async extent handling, because we were doing the correct thing by calling
filemap_flush()if we had async extents set. This would properly wait on any async extents by locking the page in the second flush, thus making sure our ordered extents were properly set up.
Which commit is this paragraph referring to? It will be helpful to include a reference so once can actually trace the history of this code.
Quoted 9 lines
However when I switched us back to page based flushing, I used
sync_inode(), which allows us to pass in our own wbc. The problem here is thatsync_inode()is smarter than the filemap_* helpers, it tries to avoid calling writepages at all. This means that our second call could skip callingdo_writepagesaltogether, and thus not wait on the pagelock for the async helpers. This means we could come back before any ordered extents were created and then simply continue on in our flushing mechanisms and ENOSPC out when we have plenty of space to use.
AFAICS in the code and based on your problem statement it would seem the culprit here is really the fact we are calling the 2nd sync_inode with WBC_SYNC_NONE, which is an indicator for sync_inode (respectively. writeback_single_inode) to simply return in case I_SYNC is still set from the first invocation of sync_inode or if the inode doesn't have I_DIRTY_ALL set. So wouldn't a simple fix be to change the wbc's .sync_mode to the second sync_inode to be WB_SYNC_ALL that will ensure that the 2nd sync_inode takes effect?
Quoted 45 lines
Fix this by putting back the async pages logic in
shrink_delalloc. This allows us to bulk write out everything that we need to, and then we can wait in one place for the async helpers to catch up, and then wait on any ordered extents that are created.Fixes e076ab2a2ca7 ("btrfs: shrink delalloc pages instead of full inodes")Signed-off-by Josef Bacik <[email protected]>--- fs/btrfs/inode.c | 4 ---- fs/btrfs/space-info.c | 33 +++++++++++++++++++++++++++++++++ 2 files changed, 33 insertions(+), 4 deletions(-)fs/btrfs/inode.c +0
-4@@ -9681,10 +9681,6 @@ static int start_delalloc_inodes(struct btrfs_root *root, &work->work); }else { ret =sync_inode (inode, wbc); -if (!ret && -test_bit (BTRFS_INODE_HAS_ASYNC_EXTENT , - &BTRFS_I (inode)->runtime_flags)) - ret =sync_inode (inode, wbc);btrfs_add_delayed_iput (inode);if (ret || wbc->nr_to_write <=0 ) goto out;fs/btrfs/space-info.c +4
-0@@ -547,9 +547,42 @@ static void shrink_delalloc(struct btrfs_fs_info *fs_info,while ((delalloc_bytes || ordered_bytes) && loops <3 ) {u64 temp =min (delalloc_bytes, to_reclaim) >>PAGE_SHIFT ;long nr_pages =min_t (u64, temp,LONG_MAX ); +int async_pages;btrfs_start_delalloc_roots (fs_info, nr_pages, true); +/* +* We need to make sure any outstanding async pages are now +* processed before we continue. This is because things like
fs/btrfs/space-info.c static void shrink_delalloc(struct btrfs_fs_info *fs_info, · line 566
Quoted 10 lines
fs/btrfs/space-info.c +10
-0+* sync_inode() try to be smart and skip writing if the inode is +* marked clean. We don't use filemap_fwrite for flushing +* because we want to control how many pages we write out at a +* time, thus this is the only safe way to make sure we've +* waited for outstanding compressed workers to have started +* their jobs and thus have ordered extents set up properly. +* Josef, when you think you are smart enough to remove this in +* the future, read this comment 4 times and explain in the +* commit message why it makes sense to remove it, and then +* delete the commit and don't remove it.
I think the last sentence can be omitted, simply prefix the paragraph with N.B. or put a more generic warning e.g. "Be extra careful and thoughtful when considering changing this logic".
Quoted 8 lines
fs/btrfs/space-info.c +8
-0+*/ + async_pages =atomic_read (&fs_info->async_delalloc_pages); +if (!async_pages) + goto skip_async; + +/* +* We don't want to wait forever, if we wrote less pages in this +* loop than we have outstanding, only wait for that number of
fs/btrfs/space-info.c static void shrink_delalloc(struct btrfs_fs_info *fs_info, · line 584
Quoted 10 lines
fs/btrfs/space-info.c +10
-0+* pages, otherwise we can wait for all async pages to finish +* before continuing. +*/ +if (async_pages > nr_pages) + async_pages -= nr_pages; +else + async_pages =0 ; +wait_event (fs_info->async_submit_wait, +atomic_read (&fs_info->async_delalloc_pages) <= + async_pages);
nit: What's funny is that without this patch async_submit_wait actually doesn't have a single user, it's only woken up in async_cow_submit but nobody actually waited for it. Ideally it should have been removed when its last user was removed. ;)
Quoted 5 lines
fs/btrfs/space-info.c +1
-0+skip_async: loops++;if (wait_ordered && !trans) {btrfs_wait_ordered_roots (fs_info, items,0 , (u64)-1 );
Re: [PATCH 1/4] btrfs: wait on async extents when flushing delalloc
Quoted 23 lines · Nikolay Borisov
Quoted 7 lines · Josef Bacik
I've been debugging an early ENOSPC problem in production and finally root caused it to this problem. When we switched to the per-inode flushing code I pulled out the async extent handling, because we were doing the correct thing by calling
filemap_flush()if we had async extents set. This would properly wait on any async extents by locking the page in the second flush, thus making sure our ordered extents were properly set up.Which commit is this paragraph referring to? It will be helpful to include a reference so once can actually trace the history of this code.
Quoted 9 lines
However when I switched us back to page based flushing, I used
sync_inode(), which allows us to pass in our own wbc. The problem here is thatsync_inode()is smarter than the filemap_* helpers, it tries to avoid calling writepages at all. This means that our second call could skip callingdo_writepagesaltogether, and thus not wait on the pagelock for the async helpers. This means we could come back before any ordered extents were created and then simply continue on in our flushing mechanisms and ENOSPC out when we have plenty of space to use.
Quoted 8 lines
AFAICS in the code and based on your problem statement it would seem the culprit here is really the fact we are calling the 2nd
sync_inodewithWBC_SYNC_NONE, which is an indicator forsync_inode(respectively.writeback_single_inode) to simply return in caseI_SYNCis still set from the first invocation ofsync_inodeor if the inode doesn't haveI_DIRTY_ALLset. So wouldn't a simple fix be to change the wbc's .sync_mode to the secondsync_inodeto beWB_SYNC_ALLthat will ensure that the 2ndsync_inodetakes effect?
It would, except that would make us wait on writeback, which is not what we want necessarily, we want to just wait for the async workers to finish and then continue writing stuff.
Quoted 52 lines
Quoted 52 lines
Fix this by putting back the async pages logic in
shrink_delalloc. This allows us to bulk write out everything that we need to, and then we can wait in one place for the async helpers to catch up, and then wait on any ordered extents that are created.Fixes e076ab2a2ca7 ("btrfs: shrink delalloc pages instead of full inodes")Signed-off-by Josef Bacik <[email protected]>--- fs/btrfs/inode.c | 4 ---- fs/btrfs/space-info.c | 33 +++++++++++++++++++++++++++++++++ 2 files changed, 33 insertions(+), 4 deletions(-)fs/btrfs/inode.c +0
-4@@ -9681,10 +9681,6 @@ static int start_delalloc_inodes(struct btrfs_root *root, &work->work); }else { ret =sync_inode (inode, wbc); -if (!ret && -test_bit (BTRFS_INODE_HAS_ASYNC_EXTENT , - &BTRFS_I (inode)->runtime_flags)) - ret =sync_inode (inode, wbc);btrfs_add_delayed_iput (inode);if (ret || wbc->nr_to_write <=0 ) goto out;fs/btrfs/space-info.c +10
-0@@ -547,9 +547,42 @@ static void shrink_delalloc(struct btrfs_fs_info *fs_info,while ((delalloc_bytes || ordered_bytes) && loops <3 ) {u64 temp =min (delalloc_bytes, to_reclaim) >>PAGE_SHIFT ;long nr_pages =min_t (u64, temp,LONG_MAX ); +int async_pages;btrfs_start_delalloc_roots (fs_info, nr_pages, true); +/* +* We need to make sure any outstanding async pages are now +* processed before we continue. This is because things like +* sync_inode() try to be smart and skip writing if the inode is +* marked clean. We don't use filemap_fwrite for flushing +* because we want to control how many pages we write out at a +* time, thus this is the only safe way to make sure we've +* waited for outstanding compressed workers to have started +* their jobs and thus have ordered extents set up properly.
Quoted 8 lines
fs/btrfs/space-info.c static void shrink_delalloc(struct btrfs_fs_info *fs_info, · line 566
Quoted 4 lines
fs/btrfs/space-info.c +4
-0+* Josef, when you think you are smart enough to remove this in +* the future, read this comment 4 times and explain in the +* commit message why it makes sense to remove it, and then +* delete the commit and don't remove it. I think the last sentence can be omitted, simply prefix the paragraph with N.B. or put a more generic warning e.g. "Be extra careful and thoughtful when considering changing this logic".
Yeah I meant to go back and change that.
Quoted 16 lines
Quoted 16 lines
fs/btrfs/space-info.c +15
-0+*/ + async_pages =atomic_read (&fs_info->async_delalloc_pages); +if (!async_pages) + goto skip_async; + +/* +* We don't want to wait forever, if we wrote less pages in this +* loop than we have outstanding, only wait for that number of +* pages, otherwise we can wait for all async pages to finish +* before continuing. +*/ +if (async_pages > nr_pages) + async_pages -= nr_pages; +else + async_pages =0 ;
Quoted 9 lines
fs/btrfs/space-info.c static void shrink_delalloc(struct btrfs_fs_info *fs_info, · line 584
Quoted 3 lines
fs/btrfs/space-info.c +3
-0+wait_event (fs_info->async_submit_wait, +atomic_read (&fs_info->async_delalloc_pages) <= + async_pages);nit: What's funny is that without this patch
async_submit_waitactually doesn't have a single user, it's only woken up inasync_cow_submitbut nobody actually waited for it. Ideally it should have been removed when its last user was removed. ;)
Clearly, but hooray I didn't rip it out because now I can just add this bit. Thanks,
Josef
PATCH 2/4 btrfs: wake up async_delalloc_pages waiters after submit
In reply to Josef BacikWe use the async_delalloc_pages mechanism to make sure that we've completed our async work before trying to continue our delalloc flushing. The reason for this is we need to see any ordered extents that were created by our delalloc flushing. However we're waking up before we do the submit work, which is before we create the ordered extents. This is a pretty wide race window where we could potentially think there are no ordered extents and thus exit shrink_delalloc prematurely. Fix this by waking us up after we've done the work to create ordered extents.
--- fs/btrfs/inode.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)
fs/btrfs/inode.c +5 -5
@@ -1271,11 +1271,6 @@ static noinline void async_cow_submit(struct btrfs_work *work) nr_pages = (async_chunk->end - async_chunk->start +PAGE_SIZE ) >>PAGE_SHIFT ; -/* atomic_sub_return implies a barrier */ -if (atomic_sub_return (nr_pages, &fs_info->async_delalloc_pages) < -5 *SZ_1M ) -cond_wake_up_nomb (&fs_info->async_submit_wait); -/* * ->inode could be NULL if async_chunk_start has failed to compress, * in which case we don't have anything to submit, yet we need to
@@ -1284,6 +1279,11 @@ static noinline void async_cow_submit(struct btrfs_work *work)*/ if (async_chunk->inode)submit_compressed_extents (async_chunk); + +/* atomic_sub_return implies a barrier */ +if (atomic_sub_return (nr_pages, &fs_info->async_delalloc_pages) < +5 *SZ_1M ) +cond_wake_up_nomb (&fs_info->async_submit_wait); } static noinline void async_cow_free(struct btrfs_work *work)
Signature · 1 line
2.26.3
Re: [PATCH 2/4] btrfs: wake up async_delalloc_pages waiters after submit
Quoted 10 lines · Josef Bacik
We use the
async_delalloc_pagesmechanism to make sure that we've completed our async work before trying to continue our delalloc flushing. The reason for this is we need to see any ordered extents that were created by our delalloc flushing. However we're waking up before we do the submit work, which is before we create the ordered extents. This is a pretty wide race window where we could potentially think there are no ordered extents and thus exitshrink_delallocprematurely. Fix this by waking us up after we've done the work to create ordered extents.
Quoted 2 lines
Signed-off-by Josef Bacik <[email protected]>
Quoted 33 lines
--- fs/btrfs/inode.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)fs/btrfs/inode.c +5
-5@@ -1271,11 +1271,6 @@ static noinline void async_cow_submit(struct btrfs_work *work) nr_pages = (async_chunk->end - async_chunk->start +PAGE_SIZE ) >>PAGE_SHIFT ; -/* atomic_sub_return implies a barrier */ -if (atomic_sub_return (nr_pages, &fs_info->async_delalloc_pages) < -5 *SZ_1M ) -cond_wake_up_nomb (&fs_info->async_submit_wait); -/* * ->inode could be NULL if async_chunk_start has failed to compress, * in which case we don't have anything to submit, yet we need to @@ -1284,6 +1279,11 @@ static noinline void async_cow_submit(struct btrfs_work *work)*/ if (async_chunk->inode)submit_compressed_extents (async_chunk); + +/* atomic_sub_return implies a barrier */ +if (atomic_sub_return (nr_pages, &fs_info->async_delalloc_pages) < +5 *SZ_1M ) +cond_wake_up_nomb (&fs_info->async_submit_wait); } static noinline void async_cow_free(struct btrfs_work *work)
PATCH 3/4 fs: add a filemap_fdatawrite_wbc helper
In reply to Josef BacikBtrfs sometimes needs to flush dirty pages on a bunch of dirty inodes in order to reclaim metadata reservations. Unfortunately most helpers in this area are too smart for us
1) The normal filemap_fdata* helpers only take range and sync modes, and don't give any indication of how much was written, so we can only flush full inodes, which isn't what we want in most cases.
2) The normal writeback path requires us to have the s_umount sem held, but we can't unconditionally take it in this path because we could deadlock.
3) The normal writeback path also skips inodes with I_SYNC set if we write with WB_SYNC_NONE. This isn't the behavior we want under heavy ENOSPC pressure, we want to actually make sure the pages are under writeback before returning, and if another thread is in the middle of writing the file we may return before they're under writeback and miss our ordered extents and not properly wait for completion.
4) sync_inode() uses the normal writeback path and has the same problem as #3.
What we really want is to call do_writepages() with our wbc. This way we can make sure that writeback is actually started on the pages, and we can control how many pages are written as a whole as we write many inodes using the same wbc. Accomplish this with a new helper that does just that so we can use it for our ENOSPC flushing infrastructure.
--- include/linux/fs.h | 2 ++ mm/filemap.c | 29 ++++++++++++++++++++++++----- 2 files changed, 26 insertions(+), 5 deletions(-)
include/linux/fs.h +2 -0
@@ -2886,6 +2886,8 @@ extern int filemap_fdatawrite_range(struct address_space *mapping, loff_t start, loff_t end);extern int filemap_check_errors (struct address_space *mapping);extern void __filemap_set_wb_err (struct address_space *mapping,int err); +extern int filemap_fdatawrite_wbc (struct address_space *mapping, +struct writeback_control *wbc);static inline int filemap_write_and_wait (struct address_space *mapping) {
mm/filemap.c +24 -5
@@ -376,6 +376,29 @@ static int filemap_check_and_keep_errors(struct address_space *mapping)return -ENOSPC ;return 0 ; } +/** +* filemap_fdatawrite_wbc - start writeback on mapping dirty pages in range +* @mapping: address space structure to write +* @wbc: the writeback_control controlling the writeout +* +* This behaves the same way as __filemap_fdatawrite_range, but simply takes the +* writeback_control as an argument to allow the caller to have more flexibility +* over the writeout parameters, and with no checks around whether the mapping +* has dirty pages or anything, it simply calls writepages. +* +* Return: %0 on success, negative error code otherwise. +*/ +int filemap_fdatawrite_wbc (struct address_space *mapping, +struct writeback_control *wbc) +{ +int ret; + +wbc_attach_fdatawrite_inode (wbc, mapping->host); + ret =do_writepages (mapping, wbc); +wbc_detach_inode (wbc); +return ret; +} +EXPORT_SYMBOL (filemap_fdatawrite_wbc);/** * __filemap_fdatawrite_range - start writeback on mapping dirty pages in range
@@ -397,7 +420,6 @@ static int filemap_check_and_keep_errors(struct address_space *mapping)int __filemap_fdatawrite_range (struct address_space *mapping,loff_t start,loff_t end,int sync_mode) { -int ret;struct writeback_control wbc = { .sync_mode = sync_mode, .nr_to_write =LONG_MAX ,
@@ -409,10 +431,7 @@ int __filemap_fdatawrite_range(struct address_space *mapping, loff_t start, !mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))return 0 ; -wbc_attach_fdatawrite_inode (&wbc, mapping->host); - ret =do_writepages (mapping, &wbc); -wbc_detach_inode (&wbc); -return ret; +return filemap_fdatawrite_wbc (mapping, &wbc); } static inline int __filemap_fdatawrite(struct address_space *mapping,
Signature · 1 line
2.26.3
Re: [PATCH 3/4] fs: add a filemap_fdatawrite_wbc helper
Quoted 49 lines · Josef Bacik
Btrfs sometimes needs to flush dirty pages on a bunch of dirty inodes in order to reclaim metadata reservations. Unfortunately most helpers in this area are too smart for us
1) The normal filemap_fdata* helpers only take range and sync modes, and don't give any indication of how much was written, so we can only flush full inodes, which isn't what we want in most cases.
2) The normal writeback path requires us to have thes_umountsem held, but we can't unconditionally take it in this path because we could deadlock.
3) The normal writeback path also skips inodes withI_SYNCset if we write withWB_SYNC_NONE. This isn't the behavior we want under heavy ENOSPC pressure, we want to actually make sure the pages are under writeback before returning, and if another thread is in the middle of writing the file we may return before they're under writeback and miss our ordered extents and not properly wait for completion.
4)sync_inode()uses the normal writeback path and has the same problem as #3.What we really want is to call
do_writepages()with our wbc. This way we can make sure that writeback is actually started on the pages, and we can control how many pages are written as a whole as we write many inodes using the same wbc. Accomplish this with a new helper that does just that so we can use it for our ENOSPC flushing infrastructure.Signed-off-by Josef Bacik <[email protected]>--- include/linux/fs.h | 2 ++ mm/filemap.c | 29 ++++++++++++++++++++++++----- 2 files changed, 26 insertions(+), 5 deletions(-)include/linux/fs.h +2
-0@@ -2886,6 +2886,8 @@ extern int filemap_fdatawrite_range(struct address_space *mapping, loff_t start, loff_t end);extern int filemap_check_errors (struct address_space *mapping);extern void __filemap_set_wb_err (struct address_space *mapping,int err); +extern int filemap_fdatawrite_wbc (struct address_space *mapping, +struct writeback_control *wbc);static inline int filemap_write_and_wait (struct address_space *mapping) {mm/filemap.c +0
-0
mm/filemap.c static int filemap_check_and_keep_errors(struct address_space *mapping) · line 384
Quoted 9 lines
mm/filemap.c +6
-0@@ -376,6 +376,29 @@ static int filemap_check_and_keep_errors(struct address_space *mapping)return -ENOSPC ;return 0 ; } +/** +* filemap_fdatawrite_wbc - start writeback on mapping dirty pages in range +* @mapping: address space structure to write +* @wbc: the writeback_control controlling the writeout +* +* This behaves the same way as __filemap_fdatawrite_range, but simply takes the
That's not true, because __filemap_fdatawrite_range will only issue writeback in case of PAGECACHE_TAG_DIRTY && the inode's bdi having BDI_CAP_WRITEBACK. So I think those checks should also be moved to fdatawrite_wbc.
In fact what would be good for readability since we have a bunch of __filemap_fdatawrite functions is to have each one call your newly introduced helper and have their body simply setup the correct writeback_control structure. Alternative right now one has to chase up to 3-4 levels of (admittedly very short) functions. I.E
filemap_fdatawrite->__filemap_fdatawrite->__filemap_fdatawrite_range->filemap_fdatawrite_wbc
which is somewhat annoying. Instead I propose having
filemap_fdatawrite->filemap_fdatawrite_wbcfilemap_flush->filemap_fdatawrite_wbc etc...
Quoted 41 lines
mm/filemap.c +18
-5+* writeback_control as an argument to allow the caller to have more flexibility +* over the writeout parameters, and with no checks around whether the mapping +* has dirty pages or anything, it simply calls writepages. +* +* Return: %0 on success, negative error code otherwise. +*/ +int filemap_fdatawrite_wbc (struct address_space *mapping, +struct writeback_control *wbc) +{ +int ret; + +wbc_attach_fdatawrite_inode (wbc, mapping->host); + ret =do_writepages (mapping, wbc); +wbc_detach_inode (wbc); +return ret; +} +EXPORT_SYMBOL (filemap_fdatawrite_wbc);/** * __filemap_fdatawrite_range - start writeback on mapping dirty pages in range @@ -397,7 +420,6 @@ static int filemap_check_and_keep_errors(struct address_space *mapping)int __filemap_fdatawrite_range (struct address_space *mapping,loff_t start,loff_t end,int sync_mode) { -int ret;struct writeback_control wbc = { .sync_mode = sync_mode, .nr_to_write =LONG_MAX ,@@ -409,10 +431,7 @@ int __filemap_fdatawrite_range(struct address_space *mapping, loff_t start, !mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))return 0 ; -wbc_attach_fdatawrite_inode (&wbc, mapping->host); - ret =do_writepages (mapping, &wbc); -wbc_detach_inode (&wbc); -return ret; +return filemap_fdatawrite_wbc (mapping, &wbc); } static inline int __filemap_fdatawrite(struct address_space *mapping,
Re: [PATCH 3/4] fs: add a filemap_fdatawrite_wbc helper
Quoted 58 lines · Nikolay Borisov
Quoted 55 lines · Josef Bacik
Btrfs sometimes needs to flush dirty pages on a bunch of dirty inodes in order to reclaim metadata reservations. Unfortunately most helpers in this area are too smart for us
1) The normal filemap_fdata* helpers only take range and sync modes, and don't give any indication of how much was written, so we can only flush full inodes, which isn't what we want in most cases.
2) The normal writeback path requires us to have thes_umountsem held, but we can't unconditionally take it in this path because we could deadlock.
3) The normal writeback path also skips inodes withI_SYNCset if we write withWB_SYNC_NONE. This isn't the behavior we want under heavy ENOSPC pressure, we want to actually make sure the pages are under writeback before returning, and if another thread is in the middle of writing the file we may return before they're under writeback and miss our ordered extents and not properly wait for completion.
4)sync_inode()uses the normal writeback path and has the same problemas #3.What we really want is to call
do_writepages()with our wbc. This way we can make sure that writeback is actually started on the pages, and we can control how many pages are written as a whole as we write many inodes using the same wbc. Accomplish this with a new helper that does just that so we can use it for our ENOSPC flushing infrastructure.Signed-off-by Josef Bacik <[email protected]>--- include/linux/fs.h | 2 ++ mm/filemap.c | 29 ++++++++++++++++++++++++----- 2 files changed, 26 insertions(+), 5 deletions(-)include/linux/fs.h +2
-0@@ -2886,6 +2886,8 @@ extern int filemap_fdatawrite_range(struct address_space *mapping, loff_t start, loff_t end);extern int filemap_check_errors (struct address_space *mapping);extern void __filemap_set_wb_err (struct address_space *mapping,int err); +extern int filemap_fdatawrite_wbc (struct address_space *mapping, +struct writeback_control *wbc);static inline int filemap_write_and_wait (struct address_space *mapping) {mm/filemap.c +3
-0@@ -376,6 +376,29 @@ static int filemap_check_and_keep_errors(struct address_space *mapping)return -ENOSPC ;return 0 ; } +/** +* filemap_fdatawrite_wbc - start writeback on mapping dirty pages in range +* @mapping: address space structure to write
Quoted 9 lines
mm/filemap.c static int filemap_check_and_keep_errors(struct address_space *mapping) · line 384
Quoted 3 lines
mm/filemap.c +3
-0+* @wbc: the writeback_control controlling the writeout +* +* This behaves the same way as __filemap_fdatawrite_range, but simply takes the That's not true, because
__filemap_fdatawrite_rangewill only issue writeback in case ofPAGECACHE_TAG_DIRTY&& the inode's bdi havingBDI_CAP_WRITEBACK. So I think those checks should also be moved tofdatawrite_wbc.
Yeah I'll move those into _wbc
Quoted 6 lines
In fact what would be good for readability since we have a bunch of
__filemap_fdatawritefunctions is to have each one call your newly introduced helper and have their body simply setup the correctwriteback_controlstructure. Alternative right now one has to chase up to 3-4 levels of (admittedly very short) functions. I.E
Quoted 7 lines
filemap_fdatawrite->__filemap_fdatawrite->__filemap_fdatawrite_range->filemap_fdatawrite_wbcwhich is somewhat annoying. Instead I propose having
filemap_fdatawrite->filemap_fdatawrite_wbcfilemap_flush->filemap_fdatawrite_wbcetc...
Yeah I'd like to clean this up at some point, but that's outside the scope of this patch. I want to get a helper in without needing to run it by everybody, we can take up cleaning this up at a later point with input from everybody else.
Thanks,
Josef
Re: [PATCH 3/4] fs: add a filemap_fdatawrite_wbc helper
Quoted 4 lines · Josef Bacik
Yeah I'd like to clean this up at some point, but that's outside the scope of this patch. I want to get a helper in without needing to run it by everybody, we can take up cleaning this up at a later point with input from everybody else. Thanks,
But you _do_ need to run a new core writeback function past "everybody". Just posting it on the btrfs list like for this series is definitely not enough.
PATCH 4/4 btrfs: use the filemap_fdatawrite_wbc helper for delalloc shrinking
In reply to Josef Baciksync_inode() has some holes that can cause problems if we're under heavy ENOSPC pressure. If there's writeback running on a separate thread sync_inode() will skip writing the inode altogether. What we really want is to make sure writeback has been started on all the pages to make sure we can see the ordered extents and wait on them if appropriate. Switch to this new helper which will allow us to accomplish this and avoid ENOSPC'ing early.
--- fs/btrfs/inode.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
fs/btrfs/inode.c +1 -1
@@ -9680,7 +9680,7 @@ static int start_delalloc_inodes(struct btrfs_root *root,btrfs_queue_work (root->fs_info->flush_workers, &work->work); }else { - ret =sync_inode (inode, wbc); + ret =filemap_fdatawrite_wbc (inode->i_mapping, wbc);btrfs_add_delayed_iput (inode);if (ret || wbc->nr_to_write <=0 ) goto out;
Signature · 1 line
2.26.3
Re: [PATCH 4/4] btrfs: use the filemap_fdatawrite_wbc helper for delalloc shrinking
Quoted 9 lines · Josef Bacik
sync_inode()has some holes that can cause problems if we're under heavy ENOSPC pressure. If there's writeback running on a separate threadsync_inode()will skip writing the inode altogether. What we really want is to make sure writeback has been started on all the pages to make sure we can see the ordered extents and wait on them if appropriate. Switch to this new helper which will allow us to accomplish this and avoid ENOSPC'ing early.Signed-off-by Josef Bacik <[email protected]>
Quoted 18 lines
--- fs/btrfs/inode.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)fs/btrfs/inode.c +1
-1@@ -9680,7 +9680,7 @@ static int start_delalloc_inodes(struct btrfs_root *root,btrfs_queue_work (root->fs_info->flush_workers, &work->work); }else { - ret =sync_inode (inode, wbc); + ret =filemap_fdatawrite_wbc (inode->i_mapping, wbc);btrfs_add_delayed_iput (inode);if (ret || wbc->nr_to_write <=0 ) goto out;
Quoted 30 lines · Josef Bacik
Hello,
I backported the patch to switch us to using
sync_inode()to our kernel inside Facebook to fix a deadlock when using the async delalloc shrinker threads. This uncovered a bunch of problems with how we shrink delalloc, as we use -o compress-force, and thus everything goes through the async compression threads.I ripped out the async pages stuff because originally I had switched us to just writing the whole inode. This caused a performance regression, and so I switched us to calling
sync_inode()twice to handle the async extent case. The problem is thatsync_inode()can skip writing inodes sometimes, and thus we weren't properly waiting on the async helpers. We really do need to wait for the async delalloc pages to go down before continuing to shrink delalloc. There was also a race in how we woke up the async delalloc pages waiter which could be problematic.And then finally there is our use of
sync_inode(). It tries to be too clever for us, when in reality we want to make sure all pages are under writeback before we come back to the shrinking loop. I've added a small helper to give us this flexibilty and have switched us to that helper.With these patches, and others that will be sent separately, the early ENOSPC problems we were experiencing have been eliminated. Thanks,
Josef Bacik (4):
btrfs: wait on async extents when flushing delalloc
btrfs: wake upasync_delalloc_pageswaiters after submit
fs: add afilemap_fdatawrite_wbchelper
btrfs: use thefilemap_fdatawrite_wbchelper for delalloc shrinking
Quoted 6 lines
fs/btrfs/inode.c | 16 ++++++---------- fs/btrfs/space-info.c | 33 +++++++++++++++++++++++++++++++++ include/linux/fs.h | 2 ++ mm/filemap.c | 29 ++++++++++++++++++++++++----- 4 files changed, 65 insertions(+), 15 deletions(-)
I've ran this through xfstests on a zoned device and got the following lockdep splat. I'm not sure if this is your fault or my fault. btrfs_reclaim_bgs_work() very much sounds like my fault. Writing it here for completeness anyways. Otherwise no noticeable deviations from the baseline.
[ 634.999371] BTRFS info (device nullb1): reclaiming chunk 3489660928 with 22% used [ 634.999445] BTRFS info (device nullb1): relocating block group 3489660928 flags metadata
[ 642.127246] ======================================================
[ 642.127876] WARNING: possible circular locking dependency detected
[ 642.128526] 5.13.0-rc5-josef-delalloc #1080 Not tainted
[ 642.129060] ------------------------------------------------------
[ 642.129699] kworker/u4:5/11096 is trying to acquire lock:
[ 642.130259] ffff888278839438 (btrfs-treloc-02#2){+.+.}-{3:3}, at: __btrfs_tree_lock+0x28/0x120 [btrfs]
[ 642.131321]
but task is already holding lock:
[ 642.131921] ffff8882bdc45cf8 (btrfs-tree-01#2/1){+.+.}-{3:3}, at: __btrfs_tree_lock+0x28/0x120 [btrfs]
[ 642.132926]
which lock already depends on the new lock.[ 642.133767]
the existing dependency chain (in reverse order) is:
[ 642.134529]
-> #2 (btrfs-tree-01#2/1){+.+.}-{3:3}:
[ 642.135181] down_write_nested+0x23/0x60
[ 642.135647] __btrfs_tree_lock+0x28/0x120 [btrfs]
[ 642.136238] btrfs_init_new_buffer+0x7d/0x2a0 [btrfs]
[ 642.136844] btrfs_alloc_tree_block+0x113/0x320 [btrfs]
[ 642.137460] alloc_tree_block_no_bg_flush+0x4f/0x60 [btrfs]
[ 642.138115] __btrfs_cow_block+0x13b/0x5e0 [btrfs]
[ 642.138688] btrfs_cow_block+0x114/0x220 [btrfs]
[ 642.139296] btrfs_search_slot+0x587/0x950 [btrfs]
[ 642.139907] btrfs_lookup_inode+0x2a/0x90 [btrfs]
[ 642.140501] __btrfs_update_delayed_inode+0x80/0x2d0 [btrfs]
[ 642.141177] btrfs_async_run_delayed_root+0x174/0x240 [btrfs]
[ 642.141878] btrfs_work_helper+0xfe/0x530 [btrfs]
[ 642.142454] process_one_work+0x24f/0x570
[ 642.142931] worker_thread+0x4f/0x3e0
[ 642.143357] kthread+0x12c/0x150
[ 642.143748] ret_from_fork+0x22/0x30
[ 642.144181]
-> #1 (btrfs-tree-01#2){++++}-{3:3}:
[ 642.144807] down_write_nested+0x23/0x60
[ 642.145260] __btrfs_tree_lock+0x28/0x120 [btrfs]
[ 642.145839] btrfs_search_slot+0x259/0x950 [btrfs]
[ 642.146413] do_relocation+0xf9/0x5d0 [btrfs]
[ 642.146960] relocate_tree_blocks+0x293/0x610 [btrfs]
[ 642.147579] relocate_block_group+0x1f2/0x560 [btrfs]
[ 642.148201] btrfs_relocate_block_group+0x16c/0x320 [btrfs]
[ 642.148869] btrfs_relocate_chunk+0x38/0x120 [btrfs]
[ 642.149479] btrfs_reclaim_bgs_work.cold+0x5d/0x10f [btrfs]
[ 642.150162] process_one_work+0x24f/0x570
[ 642.150628] worker_thread+0x4f/0x3e0
[ 642.151062] kthread+0x12c/0x150
[ 642.151450] ret_from_fork+0x22/0x30
[ 642.151872]
-> #0 (btrfs-treloc-02#2){+.+.}-{3:3}:
[ 642.152518] __lock_acquire+0x1235/0x2320
[ 642.152985] lock_acquire+0xab/0x340
[ 642.153409] down_write_nested+0x23/0x60
[ 642.153872] __btrfs_tree_lock+0x28/0x120 [btrfs]
[ 642.154455] btrfs_lock_root_node+0x31/0x40 [btrfs]
[ 642.155064] btrfs_search_slot+0x6bc/0x950 [btrfs]
[ 642.155639] replace_path+0x56f/0xa30 [btrfs]
[ 642.156181] merge_reloc_root+0x258/0x600 [btrfs]
[ 642.156773] merge_reloc_roots+0xdd/0x210 [btrfs]
[ 642.157368] relocate_block_group+0x2c9/0x560 [btrfs]
[ 642.157986] btrfs_relocate_block_group+0x16c/0x320 [btrfs]
[ 642.158660] btrfs_relocate_chunk+0x38/0x120 [btrfs]
[ 642.159272] btrfs_reclaim_bgs_work.cold+0x5d/0x10f [btrfs]
[ 642.159951] process_one_work+0x24f/0x570
[ 642.160418] worker_thread+0x4f/0x3e0
[ 642.160850] kthread+0x12c/0x150
[ 642.161243] ret_from_fork+0x22/0x30
[ 642.161664]
other info that might help us debug this:[ 642.162480] Chain exists of:
btrfs-treloc-02#2 --> btrfs-tree-01#2 --> btrfs-tree-01#2/1[ 642.163616] Possible unsafe locking scenario:
[ 642.164226] CPU0 CPU1
[ 642.164699] ---- ----
[ 642.165164] lock(btrfs-tree-01#2/1);
[ 642.165553] lock(btrfs-tree-01#2);
[ 642.166183] lock(btrfs-tree-01#2/1);
[ 642.166820] lock(btrfs-treloc-02#2);
[ 642.167209]
*** DEADLOCK ***[ 642.167812] 6 locks held by kworker/u4:5/11096:
[ 642.168288] #0: ffff8881000c4938 ((wq_completion)events_unbound){+.+.}-{0:0}, at: process_one_work+0x1c6/0x570
[ 642.169317] #1: ffffc900046e3e78 ((work_completion)(&fs_info->reclaim_bgs_work)){+.+.}-{0:0}, at: process_one_work+0x1c6/0x570
[ 642.170507] #2: ffff888136c6a0a0 (&fs_info->reclaim_bgs_lock){+.+.}-{3:3}, at: btrfs_reclaim_bgs_work+0x5d/0x1b0 [btrfs]
[ 642.171687] #3: ffff888136c68838 (&fs_info->cleaner_mutex){+.+.}-{3:3}, at: btrfs_relocate_block_group+0x164/0x320 [btrfs]
[ 642.172877] #4: ffff88811d11f620 (sb_internal){.+.+}-{0:0}, at: merge_reloc_root+0x178/0x600 [btrfs]
[ 642.173871] #5: ffff8882bdc45cf8 (btrfs-tree-01#2/1){+.+.}-{3:3}, at: __btrfs_tree_lock+0x28/0x120 [btrfs]
[ 642.174906]
[ 642.175362] CPU: 0 PID: 11096 Comm: kworker/u4:5 Not tainted 5.13.0-rc5-josef-delalloc #1080
[ 642.176215] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a-rebuilt.opensuse.org 04/01/2014
[ 642.177366] Workqueue: events_unbound btrfs_reclaim_bgs_work [btrfs]
[ 642.178065] Call Trace:
[ 642.178323] dump_stack+0x6d/0x89
[ 642.178670] check_noncircular+0xcf/0xf0
[ 642.179085] __lock_acquire+0x1235/0x2320
[ 642.179509] lock_acquire+0xab/0x340
[ 642.179889] ? __btrfs_tree_lock+0x28/0x120 [btrfs]
[ 642.180442] ? find_held_lock+0x2b/0x80
[ 642.180843] ? btrfs_root_node+0x93/0x1d0 [btrfs]
[ 642.181359] down_write_nested+0x23/0x60
[ 642.181781] ? __btrfs_tree_lock+0x28/0x120 [btrfs]
[ 642.182327] __btrfs_tree_lock+0x28/0x120 [btrfs]
[ 642.182862] btrfs_lock_root_node+0x31/0x40 [btrfs]
[ 642.183411] btrfs_search_slot+0x6bc/0x950 [btrfs]
[ 642.183940] ? release_extent_buffer+0x111/0x160 [btrfs]
[ 642.184526] replace_path+0x56f/0xa30 [btrfs]
[ 642.185019] merge_reloc_root+0x258/0x600 [btrfs]
[ 642.185557] merge_reloc_roots+0xdd/0x210 [btrfs]
[ 642.186087] relocate_block_group+0x2c9/0x560 [btrfs]
[ 642.186658] btrfs_relocate_block_group+0x16c/0x320 [btrfs]
[ 642.187283] btrfs_relocate_chunk+0x38/0x120 [btrfs]
[ 642.187839] btrfs_reclaim_bgs_work.cold+0x5d/0x10f [btrfs]
[ 642.188472] process_one_work+0x24f/0x570
[ 642.188894] worker_thread+0x4f/0x3e0
[ 642.189273] ? process_one_work+0x570/0x570
[ 642.189714] kthread+0x12c/0x150
[ 642.190053] ? __kthread_bind_mask+0x60/0x60
[ 642.190492] ret_from_fork+0x22/0x30
[ 644.827436] BTRFS info (device nullb1): found 3645 extents, stage: move data extents
[ 645.132633] BTRFS info (device nullb1): reclaiming chunk 4026531840 with 22% usedQuoted 161 lines · Johannes Thumshirn
Quoted 30 lines · Josef Bacik
Hello,
I backported the patch to switch us to using
sync_inode()to our kernel inside Facebook to fix a deadlock when using the async delalloc shrinker threads. This uncovered a bunch of problems with how we shrink delalloc, as we use -o compress-force, and thus everything goes through the async compression threads.I ripped out the async pages stuff because originally I had switched us to just writing the whole inode. This caused a performance regression, and so I switched us to calling
sync_inode()twice to handle the async extent case. The problem is thatsync_inode()can skip writing inodes sometimes, and thus we weren't properly waiting on the async helpers. We really do need to wait for the async delalloc pages to go down before continuing to shrink delalloc. There was also a race in how we woke up the async delalloc pages waiter which could be problematic.And then finally there is our use of
sync_inode(). It tries to be too clever for us, when in reality we want to make sure all pages are under writeback before we come back to the shrinking loop. I've added a small helper to give us this flexibilty and have switched us to that helper.With these patches, and others that will be sent separately, the early ENOSPC problems we were experiencing have been eliminated. Thanks,
Josef Bacik (4):
btrfs: wait on async extents when flushing delalloc
btrfs: wake upasync_delalloc_pageswaiters after submit
fs: add afilemap_fdatawrite_wbchelper
btrfs: use thefilemap_fdatawrite_wbchelper for delalloc shrinkingQuoted 6 lines
fs/btrfs/inode.c | 16 ++++++---------- fs/btrfs/space-info.c | 33 +++++++++++++++++++++++++++++++++ include/linux/fs.h | 2 ++ mm/filemap.c | 29 ++++++++++++++++++++++++----- 4 files changed, 65 insertions(+), 15 deletions(-)I've ran this through xfstests on a zoned device and got the following lockdep splat. I'm not sure if this is your fault or my fault.
btrfs_reclaim_bgs_work()very much sounds like my fault. Writing it here for completeness anyways. Otherwise no noticeable deviations from the baseline.[ 634.999371] BTRFS info (device nullb1): reclaiming chunk 3489660928 with 22% used [ 634.999445] BTRFS info (device nullb1): relocating block group 3489660928 flags metadata[ 642.127246] ====================================================== [ 642.127876] WARNING: possible circular locking dependency detected [ 642.128526] 5.13.0-rc5-josef-delalloc #1080 Not tainted [ 642.129060] ------------------------------------------------------ [ 642.129699] kworker/u4:5/11096 is trying to acquire lock: [ 642.130259] ffff888278839438 (btrfs-treloc-02#2){+.+.}-{3:3}, at: __btrfs_tree_lock+0x28/0x120 [btrfs] [ 642.131321] but task is already holding lock: [ 642.131921] ffff8882bdc45cf8 (btrfs-tree-01#2/1){+.+.}-{3:3}, at: __btrfs_tree_lock+0x28/0x120 [btrfs] [ 642.132926] which lock already depends on the new lock.[ 642.133767] the existing dependency chain (in reverse order) is: [ 642.134529] -> #2 (btrfs-tree-01#2/1){+.+.}-{3:3}: [ 642.135181] down_write_nested+0x23/0x60 [ 642.135647] __btrfs_tree_lock+0x28/0x120 [btrfs] [ 642.136238] btrfs_init_new_buffer+0x7d/0x2a0 [btrfs] [ 642.136844] btrfs_alloc_tree_block+0x113/0x320 [btrfs] [ 642.137460] alloc_tree_block_no_bg_flush+0x4f/0x60 [btrfs] [ 642.138115] __btrfs_cow_block+0x13b/0x5e0 [btrfs] [ 642.138688] btrfs_cow_block+0x114/0x220 [btrfs] [ 642.139296] btrfs_search_slot+0x587/0x950 [btrfs] [ 642.139907] btrfs_lookup_inode+0x2a/0x90 [btrfs] [ 642.140501] __btrfs_update_delayed_inode+0x80/0x2d0 [btrfs] [ 642.141177] btrfs_async_run_delayed_root+0x174/0x240 [btrfs] [ 642.141878] btrfs_work_helper+0xfe/0x530 [btrfs] [ 642.142454] process_one_work+0x24f/0x570 [ 642.142931] worker_thread+0x4f/0x3e0 [ 642.143357] kthread+0x12c/0x150 [ 642.143748] ret_from_fork+0x22/0x30 [ 642.144181] -> #1 (btrfs-tree-01#2){++++}-{3:3}: [ 642.144807] down_write_nested+0x23/0x60 [ 642.145260] __btrfs_tree_lock+0x28/0x120 [btrfs] [ 642.145839] btrfs_search_slot+0x259/0x950 [btrfs] [ 642.146413] do_relocation+0xf9/0x5d0 [btrfs] [ 642.146960] relocate_tree_blocks+0x293/0x610 [btrfs] [ 642.147579] relocate_block_group+0x1f2/0x560 [btrfs] [ 642.148201] btrfs_relocate_block_group+0x16c/0x320 [btrfs] [ 642.148869] btrfs_relocate_chunk+0x38/0x120 [btrfs] [ 642.149479] btrfs_reclaim_bgs_work.cold+0x5d/0x10f [btrfs] [ 642.150162] process_one_work+0x24f/0x570 [ 642.150628] worker_thread+0x4f/0x3e0 [ 642.151062] kthread+0x12c/0x150 [ 642.151450] ret_from_fork+0x22/0x30 [ 642.151872] -> #0 (btrfs-treloc-02#2){+.+.}-{3:3}: [ 642.152518] __lock_acquire+0x1235/0x2320 [ 642.152985] lock_acquire+0xab/0x340 [ 642.153409] down_write_nested+0x23/0x60 [ 642.153872] __btrfs_tree_lock+0x28/0x120 [btrfs] [ 642.154455] btrfs_lock_root_node+0x31/0x40 [btrfs] [ 642.155064] btrfs_search_slot+0x6bc/0x950 [btrfs] [ 642.155639] replace_path+0x56f/0xa30 [btrfs] [ 642.156181] merge_reloc_root+0x258/0x600 [btrfs] [ 642.156773] merge_reloc_roots+0xdd/0x210 [btrfs] [ 642.157368] relocate_block_group+0x2c9/0x560 [btrfs] [ 642.157986] btrfs_relocate_block_group+0x16c/0x320 [btrfs] [ 642.158660] btrfs_relocate_chunk+0x38/0x120 [btrfs] [ 642.159272] btrfs_reclaim_bgs_work.cold+0x5d/0x10f [btrfs] [ 642.159951] process_one_work+0x24f/0x570 [ 642.160418] worker_thread+0x4f/0x3e0 [ 642.160850] kthread+0x12c/0x150 [ 642.161243] ret_from_fork+0x22/0x30 [ 642.161664] other info that might help us debug this:[ 642.162480] Chain exists of: btrfs-treloc-02#2 --> btrfs-tree-01#2 --> btrfs-tree-01#2/1[ 642.163616] Possible unsafe locking scenario:[ 642.164226] CPU0 CPU1 [ 642.164699] ---- ---- [ 642.165164] lock(btrfs-tree-01#2/1); [ 642.165553] lock(btrfs-tree-01#2); [ 642.166183] lock(btrfs-tree-01#2/1); [ 642.166820] lock(btrfs-treloc-02#2); [ 642.167209] *** DEADLOCK ***[ 642.167812] 6 locks held by kworker/u4:5/11096: [ 642.168288] #0: ffff8881000c4938 ((wq_completion)events_unbound){+.+.}-{0:0}, at: process_one_work+0x1c6/0x570 [ 642.169317] #1: ffffc900046e3e78 ((work_completion)(&fs_info->reclaim_bgs_work)){+.+.}-{0:0}, at: process_one_work+0x1c6/0x570 [ 642.170507] #2: ffff888136c6a0a0 (&fs_info->reclaim_bgs_lock){+.+.}-{3:3}, at: btrfs_reclaim_bgs_work+0x5d/0x1b0 [btrfs] [ 642.171687] #3: ffff888136c68838 (&fs_info->cleaner_mutex){+.+.}-{3:3}, at: btrfs_relocate_block_group+0x164/0x320 [btrfs] [ 642.172877] #4: ffff88811d11f620 (sb_internal){.+.+}-{0:0}, at: merge_reloc_root+0x178/0x600 [btrfs] [ 642.173871] #5: ffff8882bdc45cf8 (btrfs-tree-01#2/1){+.+.}-{3:3}, at: __btrfs_tree_lock+0x28/0x120 [btrfs] [ 642.174906] [ 642.175362] CPU: 0 PID: 11096 Comm: kworker/u4:5 Not tainted 5.13.0-rc5-josef-delalloc #1080 [ 642.176215] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a-rebuilt.opensuse.org 04/01/2014 [ 642.177366] Workqueue: events_unbound btrfs_reclaim_bgs_work [btrfs] [ 642.178065] Call Trace: [ 642.178323] dump_stack+0x6d/0x89 [ 642.178670] check_noncircular+0xcf/0xf0 [ 642.179085] __lock_acquire+0x1235/0x2320 [ 642.179509] lock_acquire+0xab/0x340 [ 642.179889] ? __btrfs_tree_lock+0x28/0x120 [btrfs] [ 642.180442] ? find_held_lock+0x2b/0x80 [ 642.180843] ? btrfs_root_node+0x93/0x1d0 [btrfs] [ 642.181359] down_write_nested+0x23/0x60 [ 642.181781] ? __btrfs_tree_lock+0x28/0x120 [btrfs] [ 642.182327] __btrfs_tree_lock+0x28/0x120 [btrfs] [ 642.182862] btrfs_lock_root_node+0x31/0x40 [btrfs] [ 642.183411] btrfs_search_slot+0x6bc/0x950 [btrfs] [ 642.183940] ? release_extent_buffer+0x111/0x160 [btrfs] [ 642.184526] replace_path+0x56f/0xa30 [btrfs] [ 642.185019] merge_reloc_root+0x258/0x600 [btrfs] [ 642.185557] merge_reloc_roots+0xdd/0x210 [btrfs] [ 642.186087] relocate_block_group+0x2c9/0x560 [btrfs] [ 642.186658] btrfs_relocate_block_group+0x16c/0x320 [btrfs] [ 642.187283] btrfs_relocate_chunk+0x38/0x120 [btrfs] [ 642.187839] btrfs_reclaim_bgs_work.cold+0x5d/0x10f [btrfs]
Quoted 9 lines
[ 642.188472] process_one_work+0x24f/0x570 [ 642.188894] worker_thread+0x4f/0x3e0 [ 642.189273] ? process_one_work+0x570/0x570 [ 642.189714] kthread+0x12c/0x150 [ 642.190053] ? __kthread_bind_mask+0x60/0x60 [ 642.190492] ret_from_fork+0x22/0x30 [ 644.827436] BTRFS info (device nullb1): found 3645 extents, stage: move data extents [ 645.132633] BTRFS info (device nullb1): reclaiming chunk 4026531840 with 22% used
This doesn't seem to be either of our faults, unless you changed the locking in replace_path. The locking order is
reloc root -> real root
I'm actually not sure how this happens looking at the code, we clearly come in with the path pointing at the reloc root, and then we by hand walk down the real root. At the point that we unlock the path and search down the real root again we shouldn't be holding any more reloc root things. So this requires more investigation, because something fishy is happening. Thanks,
Josef