Although there are several patches improving the extent map shrinker, there are still reports of too frequent shrinker behavior, taking too much CPU for the kswapd process.
So let's only enable extent shrinker for now, until we got more comprehensive understanding and a better solution.
Link: https://lore.kernel.org/linux-btrfs/3df4acd616a07ef4d2dc6bad668701504b412ffc... Link: https://lore.kernel.org/linux-btrfs/c30fd6b3-ca7a-4759-8a53-d42878bf84f7@gma... Fixes: 956a17d9d050 ("btrfs: add a shrinker for extent maps") CC: stable@vger.kernel.org # 6.10+ Signed-off-by: Qu Wenruo wqu@suse.com --- I also checked how XFS (the only other fs implemented the free_cached_objects callback) implemented the callback.
They did two things:
- Make sure there is only one queued reclaim Currently we only do the reclaim for kswapd, but for multi-node systems, we can still have multiple kswapd processes.
But I do not think that's the root cause.
- With an extra delay of 60% of xfs_syncd_centiseccs The default value for xfs_syncd_centiseccs is 3000 centiseconds (30s), with a minimal 100 centiseconds (1s). This results the reclaim work only to be executed at most every 18 seconds by default (or 0.6s for the minimal interval).
I believe this is the root cause, we have no extra delay and that makes btrfs to shrink extent maps too frequently. --- fs/btrfs/super.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 11044e9e2cb1..98fa0f382480 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -2402,7 +2402,13 @@ static long btrfs_nr_cached_objects(struct super_block *sb, struct shrink_contro
trace_btrfs_extent_map_shrinker_count(fs_info, nr);
- return nr; + /* + * Only report the real number for DEBUG builds, as there are reports of + * serious performance degradation caused by too frequent shrinks. + */ + if (IS_ENABLED(CONFIG_BTRFS_DEBUG)) + return nr; + return 0; }
static long btrfs_free_cached_objects(struct super_block *sb, struct shrink_control *sc)