When hitting the NFS_LAYOUT_RETURN branch in pnfs_update_layout(),
the code calls pnfs_prepare_to_retry_layoutget(lo). If it succeeds,
pnfs_put_layout_hdr(lo) is called before trace_pnfs_update_layout(),
which still references 'lo'. This results in a use-after-free when the
tracepoint accesses lo's fields.

Fix this by moving the tracepoint call before pnfs_put_layout_hdr(lo).

Fixes: 2c8d5fc37fe2 ("pNFS: Stricter ordering of layoutget and layoutreturn")
Cc: stable@vger.kernel.org
Signed-off-by: Wentao Liang <vulab@iscas.ac.cn>
---
 fs/nfs/pnfs.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index a3135b5af7ee..42e5f8ec31d4 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -2191,11 +2191,11 @@ pnfs_update_layout(struct inode *ino,
 		dprintk("%s wait for layoutreturn\n", __func__);
 		lseg = ERR_PTR(pnfs_prepare_to_retry_layoutget(lo));
 		if (!IS_ERR(lseg)) {
-			pnfs_put_layout_hdr(lo);
 			dprintk("%s retrying\n", __func__);
 			trace_pnfs_update_layout(ino, pos, count, iomode, lo,
 						 lseg,
 						 PNFS_UPDATE_LAYOUT_RETRY);
+			pnfs_put_layout_hdr(lo);
 			goto lookup_again;
 		}
 		trace_pnfs_update_layout(ino, pos, count, iomode, lo, lseg,
-- 
2.34.1

…
> Fix this by moving the tracepoint call before pnfs_put_layout_hdr(lo).

Or:
Thus move a pnfs_put_layout_hdr(lo) call between a trace_pnfs_update_layout() call
and a goto statement.


How do you think about to avoid a bit of duplicate source code
in the implementation of the function “pnfs_update_layout” then?
https://elixir.bootlin.com/linux/v7.1-rc4/source/fs/nfs/pnfs.c#L2128-L2385
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/coding-style.rst?h=v7.1-rc4#n526


Were any source code analysis tools involved in the detection of improvable places?

Regards,
Markus