* [PATCH] exfat: hold s_lock for cluster-chain updates in exfat_cont_expand()
@ 2026-11-17 6:00 Youling Tang
0 siblings, 0 replies; only message in thread
From: Youling Tang @ 2026-11-17 6:00 UTC (permalink / raw)
To: Namjae Jeon, Sungjong Seo
Cc: Yuezhang Mo, Chi Zhiling, exfat, linux-kernel, youling.tang,
Youling Tang
From: Youling Tang <tangyouling@kylinos.cn>
exfat_cont_expand() allocates clusters, appends them to the FAT chain
and updates ei->start_clu, ei->flags and i_blocks without holding
sbi->s_lock. Readers such as writeback call exfat_map_cluster() with
s_lock held but without inode_lock, so they are not serialized against
cont_expand() and can observe the chain metadata mid-update.
Hold s_lock around the cluster allocation, chain append and i_blocks
update. i_size is published via exfat_pagecache_isize_extended(), which
takes the straddling folio lock; since writeback takes folio lock ->
s_lock, this must happen after s_lock is released to avoid inverting the
lock order. i_blocks is therefore updated first, under s_lock, which
also ensures a reader never sees i_size beyond the allocated chain.
Signed-off-by: Youling Tang <tangyouling@kylinos.cn>
---
fs/exfat/file.c | 19 ++++++++++++++-----
1 file changed, 14 insertions(+), 5 deletions(-)
diff --git a/fs/exfat/file.c b/fs/exfat/file.c
index 435bbef288b8..de335c278caf 100644
--- a/fs/exfat/file.c
+++ b/fs/exfat/file.c
@@ -95,6 +95,7 @@ static int exfat_cont_expand(struct inode *inode, loff_t size)
if (ret)
return ret;
+ mutex_lock(&sbi->s_lock);
num_clusters = exfat_bytes_to_cluster(sbi, exfat_ondisk_size(inode));
/* integer overflow is already checked in inode_newsize_ok(). */
new_num_clusters = exfat_bytes_to_cluster_round_up(sbi, size);
@@ -106,7 +107,7 @@ static int exfat_cont_expand(struct inode *inode, loff_t size)
exfat_chain_set(&clu, ei->start_clu, num_clusters, ei->flags);
ret = exfat_find_last_cluster(sb, &clu, &last_clu);
if (ret)
- return ret;
+ goto unlock;
clu.dir = last_clu + 1;
} else {
@@ -120,7 +121,7 @@ static int exfat_cont_expand(struct inode *inode, loff_t size)
ret = exfat_alloc_cluster(inode, new_num_clusters - num_clusters,
&clu, inode_needs_sync(inode), false);
if (ret)
- return ret;
+ goto unlock;
/* Append new clusters to chain */
if (num_clusters) {
@@ -137,15 +138,20 @@ static int exfat_cont_expand(struct inode *inode, loff_t size)
ei->flags = clu.flags;
out:
+ /* Update i_blocks before i_size so mappings never exceed the chain. */
+ inode->i_blocks = round_up(size, sbi->cluster_size) >> 9;
inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
+ mutex_unlock(&sbi->s_lock);
+
/*
* When extending file size, call exfat_pagecache_isize_extended()
* to update i_size and ensure the straddling folio is properly
* marked RO so page_mkwrite() is called and post-EOF area is zeroed.
+ *
+ * It takes the folio lock; call it without s_lock (writeback uses
+ * folio lock -> s_lock).
*/
exfat_pagecache_isize_extended(inode, oldsize, size);
-
- inode->i_blocks = round_up(size, sbi->cluster_size) >> 9;
mark_inode_dirty(inode);
if (IS_SYNC(inode))
@@ -155,7 +161,10 @@ static int exfat_cont_expand(struct inode *inode, loff_t size)
free_clu:
exfat_free_cluster(inode, &clu);
- return -EIO;
+ ret = -EIO;
+unlock:
+ mutex_unlock(&sbi->s_lock);
+ return ret;
}
/*
--
2.43.0
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2026-11-17 6:01 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-11-17 6:00 [PATCH] exfat: hold s_lock for cluster-chain updates in exfat_cont_expand() Youling Tang
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®