f2fs: avoid deadlock during evict after f2fs_gc
o Deadlock case #1
Thread 1:
- writeback_sb_inodes
- do_writepages
- f2fs_write_data_pages
- write_cache_pages
- f2fs_write_data_page
- f2fs_balance_fs
- wait mutex_lock(gc_mutex)
Thread 2:
- f2fs_balance_fs
- mutex_lock(gc_mutex)
- f2fs_gc
- f2fs_iget
- wait iget_locked(inode->i_lock)
Thread 3:
- do_unlinkat
- iput
- lock(inode->i_lock)
- evict
- inode_wait_for_writeback
o Deadlock case #2
Thread 1:
- __writeback_single_inode
: set I_SYNC
- do_writepages
- f2fs_write_data_page
- f2fs_balance_fs
- f2fs_gc
- iput
- evict
- inode_wait_for_writeback(I_SYNC)
In order to avoid this, even though iput is called with the zero-reference
count, we need to stop the eviction procedure if the inode is on writeback.
So this patch links f2fs_drop_inode which checks the I_SYNC flag.
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index 2db9380..61b4454 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -577,6 +577,7 @@
{
struct inode *inode = mapping->host;
struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
+ bool locked = false;
int ret;
long excess_nrtw = 0, desired_nrtw;
@@ -590,10 +591,12 @@
wbc->nr_to_write = desired_nrtw;
}
- if (!S_ISDIR(inode->i_mode))
+ if (!S_ISDIR(inode->i_mode)) {
mutex_lock(&sbi->writepages);
+ locked = true;
+ }
ret = write_cache_pages(mapping, wbc, __f2fs_writepage, mapping);
- if (!S_ISDIR(inode->i_mode))
+ if (locked)
mutex_unlock(&sbi->writepages);
f2fs_submit_bio(sbi, DATA, (wbc->sync_mode == WB_SYNC_ALL));
diff --git a/fs/f2fs/namei.c b/fs/f2fs/namei.c
index 4aa26e5..47abc97 100644
--- a/fs/f2fs/namei.c
+++ b/fs/f2fs/namei.c
@@ -72,6 +72,7 @@
unlock_new_inode(inode);
fail:
trace_f2fs_new_inode(inode, err);
+ make_bad_inode(inode);
iput(inode);
if (nid_free)
alloc_nid_failed(sbi, ino);
@@ -155,6 +156,7 @@
out:
clear_nlink(inode);
unlock_new_inode(inode);
+ make_bad_inode(inode);
iput(inode);
alloc_nid_failed(sbi, ino);
return err;
@@ -190,6 +192,7 @@
return 0;
out:
clear_inode_flag(F2FS_I(inode), FI_INC_LINK);
+ make_bad_inode(inode);
iput(inode);
return err;
}
@@ -295,6 +298,7 @@
out:
clear_nlink(inode);
unlock_new_inode(inode);
+ make_bad_inode(inode);
iput(inode);
alloc_nid_failed(sbi, inode->i_ino);
return err;
@@ -335,6 +339,7 @@
clear_inode_flag(F2FS_I(inode), FI_INC_LINK);
clear_nlink(inode);
unlock_new_inode(inode);
+ make_bad_inode(inode);
iput(inode);
alloc_nid_failed(sbi, inode->i_ino);
return err;
@@ -382,6 +387,7 @@
out:
clear_nlink(inode);
unlock_new_inode(inode);
+ make_bad_inode(inode);
iput(inode);
alloc_nid_failed(sbi, inode->i_ino);
return err;
diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index 5835aaf..cd0e89a 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -98,6 +98,20 @@
return &fi->vfs_inode;
}
+static int f2fs_drop_inode(struct inode *inode)
+{
+ /*
+ * This is to avoid a deadlock condition like below.
+ * writeback_single_inode(inode)
+ * - f2fs_write_data_page
+ * - f2fs_gc -> iput -> evict
+ * - inode_wait_for_writeback(inode)
+ */
+ if (!inode_unhashed(inode) && inode->i_state & I_SYNC)
+ return 0;
+ return generic_drop_inode(inode);
+}
+
static void f2fs_i_callback(struct rcu_head *head)
{
struct inode *inode = container_of(head, struct inode, i_rcu);
@@ -232,6 +246,7 @@
static struct super_operations f2fs_sops = {
.alloc_inode = f2fs_alloc_inode,
+ .drop_inode = f2fs_drop_inode,
.destroy_inode = f2fs_destroy_inode,
.write_inode = f2fs_write_inode,
.show_options = f2fs_show_options,