[RFC][PATCH 02/10] fs: Avoid looking at i_count without i_lock held

From: Peter Zijlstra
Date: Fri Feb 24 2017 - 13:44:35 EST


In general i_count will only stay 0 if we have i_lock held. I realize
this is evict, so having MS_ACTIVE cleared might avoid the race
against find_inode_fast() in other ways.

Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
---
fs/inode.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)

--- a/fs/inode.c
+++ b/fs/inode.c
@@ -604,10 +604,12 @@ void evict_inodes(struct super_block *sb
again:
spin_lock(&sb->s_inode_list_lock);
list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) {
- if (atomic_read(&inode->i_count))
+ spin_lock(&inode->i_lock);
+ if (atomic_read(&inode->i_count)) {
+ spin_unlock(&inode->i_lock);
continue;
+ }

- spin_lock(&inode->i_lock);
if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) {
spin_unlock(&inode->i_lock);
continue;