From 00d873c17e29cc32d90ca852b82685f1673acaa5 Mon Sep 17 00:00:00 2001
From: Jan Kara <jack@suse.cz>
Date: Thu, 4 May 2023 14:47:23 +0200
Subject: [PATCH] ext4: avoid deadlock in fs reclaim with page writeback
Git-commit: 00d873c17e29cc32d90ca852b82685f1673acaa5
Patch-mainline: v6.4-rc2
References: bsc#1213016
Ext4 has a filesystem wide lock protecting ext4_writepages() calls to
avoid races with switching of journalled data flag or inode format. This
lock can however cause a deadlock like:
CPU0 CPU1
ext4_writepages()
percpu_down_read(sbi->s_writepages_rwsem);
ext4_change_inode_journal_flag()
percpu_down_write(sbi->s_writepages_rwsem);
- blocks, all readers block from now on
ext4_do_writepages()
ext4_init_io_end()
kmem_cache_zalloc(io_end_cachep, GFP_KERNEL)
fs_reclaim frees dentry...
dentry_unlink_inode()
iput() - last ref =>
iput_final() - inode dirty =>
write_inode_now()...
ext4_writepages() tries to acquire sbi->s_writepages_rwsem
and blocks forever
Make sure we cannot recurse into filesystem reclaim from writeback code
to avoid the deadlock.
Reported-by: syzbot+6898da502aef574c5f8a@syzkaller.appspotmail.com
Link: https://lore.kernel.org/all/0000000000004c66b405fa108e27@google.com
Fixes: c8585c6fcaf2 ("ext4: fix races between changing inode journal mode and ext4_writepages")
Cc: stable@vger.kernel.org
Signed-off-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230504124723.20205-1-jack@suse.cz
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Acked-by: Jan Kara <jack@suse.cz>
---
fs/ext4/ext4.h | 25 +++++++++++++++++++++++++
fs/ext4/inode.c | 19 ++++++++++---------
fs/ext4/migrate.c | 11 ++++++-----
3 files changed, 41 insertions(+), 14 deletions(-)
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -40,6 +40,7 @@
#endif
#include <linux/falloc.h>
#include <linux/percpu-rwsem.h>
+#include <linux/sched/mm.h>
#ifdef __KERNEL__
#include <linux/compat.h>
#endif
@@ -1548,6 +1549,30 @@ static inline struct ext4_inode_info *EX
return container_of(inode, struct ext4_inode_info, vfs_inode);
}
+static inline int ext4_writepages_down_read(struct super_block *sb)
+{
+ percpu_down_read(&EXT4_SB(sb)->s_writepages_rwsem);
+ return memalloc_nofs_save();
+}
+
+static inline void ext4_writepages_up_read(struct super_block *sb, int ctx)
+{
+ memalloc_nofs_restore(ctx);
+ percpu_up_read(&EXT4_SB(sb)->s_writepages_rwsem);
+}
+
+static inline int ext4_writepages_down_write(struct super_block *sb)
+{
+ percpu_down_write(&EXT4_SB(sb)->s_writepages_rwsem);
+ return memalloc_nofs_save();
+}
+
+static inline void ext4_writepages_up_write(struct super_block *sb, int ctx)
+{
+ memalloc_nofs_restore(ctx);
+ percpu_up_write(&EXT4_SB(sb)->s_writepages_rwsem);
+}
+
static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
{
return ino == EXT4_ROOT_INO ||
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -2688,11 +2688,12 @@ static int ext4_writepages(struct addres
bool done;
struct blk_plug plug;
bool give_up_on_write = false;
+ int alloc_ctx;
if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
return -EIO;
- percpu_down_read(&sbi->s_writepages_rwsem);
+ alloc_ctx = ext4_writepages_down_read(inode->i_sb);
trace_ext4_writepages(inode, wbc);
/*
@@ -2916,7 +2917,7 @@ unplug:
out_writepages:
trace_ext4_writepages_result(inode, wbc, ret,
nr_to_write - wbc->nr_to_write);
- percpu_up_read(&sbi->s_writepages_rwsem);
+ ext4_writepages_up_read(inode->i_sb, alloc_ctx);
return ret;
}
@@ -2926,18 +2927,18 @@ static int ext4_dax_writepages(struct ad
int ret;
long nr_to_write = wbc->nr_to_write;
struct inode *inode = mapping->host;
- struct ext4_sb_info *sbi = EXT4_SB(mapping->host->i_sb);
+ int alloc_ctx;
if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
return -EIO;
- percpu_down_read(&sbi->s_writepages_rwsem);
+ alloc_ctx = ext4_writepages_down_read(inode->i_sb);
trace_ext4_writepages(inode, wbc);
ret = dax_writeback_mapping_range(mapping, inode->i_sb->s_bdev, wbc);
trace_ext4_writepages_result(inode, wbc, ret,
nr_to_write - wbc->nr_to_write);
- percpu_up_read(&sbi->s_writepages_rwsem);
+ ext4_writepages_up_read(inode->i_sb, alloc_ctx);
return ret;
}
@@ -6018,7 +6019,7 @@ int ext4_change_inode_journal_flag(struc
journal_t *journal;
handle_t *handle;
int err;
- struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
+ int alloc_ctx;
/*
* We have to be very careful here: changing a data block's
@@ -6058,7 +6059,7 @@ int ext4_change_inode_journal_flag(struc
}
}
- percpu_down_write(&sbi->s_writepages_rwsem);
+ alloc_ctx = ext4_writepages_down_write(inode->i_sb);
jbd2_journal_lock_updates(journal);
/*
@@ -6075,7 +6076,7 @@ int ext4_change_inode_journal_flag(struc
err = jbd2_journal_flush(journal);
if (err < 0) {
jbd2_journal_unlock_updates(journal);
- percpu_up_write(&sbi->s_writepages_rwsem);
+ ext4_writepages_up_write(inode->i_sb, alloc_ctx);
ext4_inode_resume_unlocked_dio(inode);
return err;
}
@@ -6084,7 +6085,7 @@ int ext4_change_inode_journal_flag(struc
ext4_set_aops(inode);
jbd2_journal_unlock_updates(journal);
- percpu_up_write(&sbi->s_writepages_rwsem);
+ ext4_writepages_up_write(inode->i_sb, alloc_ctx);
if (val)
up_write(&EXT4_I(inode)->i_mmap_sem);
--- a/fs/ext4/migrate.c
+++ b/fs/ext4/migrate.c
@@ -434,7 +434,6 @@ static int free_ext_block(handle_t *hand
int ext4_ext_migrate(struct inode *inode)
{
- struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
handle_t *handle;
int retval = 0, i;
__le32 *i_data;
@@ -444,6 +443,7 @@ int ext4_ext_migrate(struct inode *inode
unsigned long max_entries;
__u32 goal, tmp_csum_seed;
uid_t owner[2];
+ int alloc_ctx;
/*
* If the filesystem does not support extents, or the inode
@@ -460,7 +460,7 @@ int ext4_ext_migrate(struct inode *inode
*/
return retval;
- percpu_down_write(&sbi->s_writepages_rwsem);
+ alloc_ctx = ext4_writepages_down_write(inode->i_sb);
/*
* Worst case we can touch the allocation bitmaps and a block
@@ -611,7 +611,7 @@ out_tmp_inode:
unlock_new_inode(tmp_inode);
iput(tmp_inode);
out_unlock:
- percpu_up_write(&sbi->s_writepages_rwsem);
+ ext4_writepages_up_write(inode->i_sb, alloc_ctx);
return retval;
}
@@ -630,6 +630,7 @@ int ext4_ind_migrate(struct inode *inode
ext4_fsblk_t blk;
handle_t *handle;
int ret;
+ int alloc_ctx;
if (!ext4_has_feature_extents(inode->i_sb) ||
(!ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
@@ -646,7 +647,7 @@ int ext4_ind_migrate(struct inode *inode
if (test_opt(inode->i_sb, DELALLOC))
ext4_alloc_da_blocks(inode);
- percpu_down_write(&sbi->s_writepages_rwsem);
+ alloc_ctx = ext4_writepages_down_write(inode->i_sb);
handle = ext4_journal_start(inode, EXT4_HT_MIGRATE, 1);
if (IS_ERR(handle)) {
@@ -688,6 +689,6 @@ errout:
ext4_journal_stop(handle);
up_write(&EXT4_I(inode)->i_data_sem);
out_unlock:
- percpu_up_write(&sbi->s_writepages_rwsem);
+ ext4_writepages_up_write(inode->i_sb, alloc_ctx);
return ret;
}