[lustre-devel] [PATCH 11/15] lustre: vfs: set_nlink() is not race-safe
James Simmons
jsimmons at infradead.org
Mon Nov 8 07:07:39 PST 2021
From: Andrew Perepechko <andrew.perepechko at hpe.com>
set_nlink() is not atomic wrt race with itself and
the following warning may be triggered by VFS:
WARNING: CPU: 5 PID: 195090 at fs/inode.c:241 __destroy_inode+0xdb/0xf0
It does not seem important what exact nlink value is the result
of the race. However, we need to protect the superblock remove
counter.
HPE-bug-id: LUS-9825
WC-bug-id: https://jira.whamcloud.com/browse/LU-15081
Lustre-commit: 12b05772fdb6d0808 ("LU-15081 vfs: set_nlink() is not race-safe")
Signed-off-by: Andrew Perepechko <andrew.perepechko at hpe.com>
Reviewed-on: https://review.whamcloud.com/45191
Reviewed-by: Andreas Dilger <adilger at whamcloud.com>
Reviewed-by: Alexander Boyko <alexander.boyko at hpe.com>
Reviewed-by: Oleg Drokin <green at whamcloud.com>
Signed-off-by: James Simmons <jsimmons at infradead.org>
---
fs/lustre/llite/file.c | 3 +++
fs/lustre/llite/llite_lib.c | 5 ++++-
fs/lustre/llite/namei.c | 10 ++++++++--
fs/lustre/lmv/lmv_intent.c | 2 ++
4 files changed, 17 insertions(+), 3 deletions(-)
diff --git a/fs/lustre/llite/file.c b/fs/lustre/llite/file.c
index 1e4ff49..6755671 100644
--- a/fs/lustre/llite/file.c
+++ b/fs/lustre/llite/file.c
@@ -5041,7 +5041,10 @@ static int ll_merge_md_attr(struct inode *inode)
if (rc)
return rc;
+ spin_lock(&inode->i_lock);
set_nlink(inode, attr.cat_nlink);
+ spin_unlock(&inode->i_lock);
+
inode->i_blocks = attr.cat_blocks;
i_size_write(inode, attr.cat_size);
diff --git a/fs/lustre/llite/llite_lib.c b/fs/lustre/llite/llite_lib.c
index 51823b5..147e680 100644
--- a/fs/lustre/llite/llite_lib.c
+++ b/fs/lustre/llite/llite_lib.c
@@ -2448,8 +2448,11 @@ int ll_update_inode(struct inode *inode, struct lustre_md *md)
inode->i_gid = make_kgid(&init_user_ns, body->mbo_gid);
if (body->mbo_valid & OBD_MD_FLPROJID)
lli->lli_projid = body->mbo_projid;
- if (body->mbo_valid & OBD_MD_FLNLINK)
+ if (body->mbo_valid & OBD_MD_FLNLINK) {
+ spin_lock(&inode->i_lock);
set_nlink(inode, body->mbo_nlink);
+ spin_unlock(&inode->i_lock);
+ }
if (body->mbo_valid & OBD_MD_FLRDEV)
inode->i_rdev = old_decode_dev(body->mbo_rdev);
diff --git a/fs/lustre/llite/namei.c b/fs/lustre/llite/namei.c
index fe7fdbb..a0192da 100644
--- a/fs/lustre/llite/namei.c
+++ b/fs/lustre/llite/namei.c
@@ -1867,8 +1867,11 @@ static int ll_unlink(struct inode *dir, struct dentry *dchild)
* the link count so the inode can be freed immediately.
*/
body = req_capsule_server_get(&request->rq_pill, &RMF_MDT_BODY);
- if (body->mbo_valid & OBD_MD_FLNLINK)
+ if (body->mbo_valid & OBD_MD_FLNLINK) {
+ spin_lock(&dchild->d_inode->i_lock);
set_nlink(dchild->d_inode, body->mbo_nlink);
+ spin_unlock(&dchild->d_inode->i_lock);
+ }
ll_update_times(request, dir);
ll_stats_ops_tally(ll_i2sbi(dir), LPROC_LL_UNLINK,
@@ -1938,8 +1941,11 @@ static int ll_rmdir(struct inode *dir, struct dentry *dchild)
* immediately.
*/
body = req_capsule_server_get(&request->rq_pill, &RMF_MDT_BODY);
- if (body->mbo_valid & OBD_MD_FLNLINK)
+ if (body->mbo_valid & OBD_MD_FLNLINK) {
+ spin_lock(&dchild->d_inode->i_lock);
set_nlink(dchild->d_inode, body->mbo_nlink);
+ spin_unlock(&dchild->d_inode->i_lock);
+ }
}
ptlrpc_req_finished(request);
diff --git a/fs/lustre/lmv/lmv_intent.c b/fs/lustre/lmv/lmv_intent.c
index 88201e6..93da2b3 100644
--- a/fs/lustre/lmv/lmv_intent.c
+++ b/fs/lustre/lmv/lmv_intent.c
@@ -252,7 +252,9 @@ int lmv_revalidate_slaves(struct obd_export *exp,
i_size_write(inode, body->mbo_size);
inode->i_blocks = body->mbo_blocks;
+ spin_lock(&inode->i_lock);
set_nlink(inode, body->mbo_nlink);
+ spin_unlock(&inode->i_lock);
inode->i_atime.tv_sec = body->mbo_atime;
inode->i_ctime.tv_sec = body->mbo_ctime;
inode->i_mtime.tv_sec = body->mbo_mtime;
--
1.8.3.1
More information about the lustre-devel
mailing list