[PATCH v2] NFSD: restart ssc_expire_umount walk after dropping nfsd_ssc_lock

[Michael Bommarito <michael.bommarito@gmail.com>]

nfsd4_ssc_expire_umount() walks nn->nfsd_ssc_mount_list with
list_for_each_entry_safe(ni, tmp, ...).  For each expired entry it
sets nsui_busy = true, drops nfsd_ssc_lock to run mntput() on the
source vfsmount, then reacquires the lock to list_del + kfree the
entry and continue iterating via the macro's saved tmp pointer.

The nsui_busy flag protects the current ni from concurrent
nfsd4_ssc_setup_dul() finders during the lock-drop window, but it
does not pin tmp.  Another nfsd RPC thread that fails its source-
server mount and reaches nfsd4_ssc_cancel_dul() will, during that
same window, take nfsd_ssc_lock, list_del + kfree its own ssc_umount
item, and release the lock.  If that item is the saved tmp of the
expire walk, the next iteration dereferences a freed
nfsd4_ssc_umount_item.

Restart the walk from the head after the mntput() unlock window so
no saved next pointer survives the lock-drop.  The list is bounded
by the number of active inter-server source mounts (typically small)
and the expire delayed-work runs periodically rather than per-IO,
so the restart is cheap.

Cc: stable@vger.kernel.org
Fixes: f4e44b393389 ("NFSD: delay unmount source's export after inter-server copy completed.")
Assisted-by: Claude:claude-opus-4-7
Signed-off-by: Michael Bommarito <michael.bommarito@gmail.com>
---
 fs/nfsd/nfs4state.c | 44 +++++++++++++++++++++++++-------------------

Changes since v1:
 - Rewrote the in-code comment above goto-restart to describe the
   current code (concurrent cancel_dul can free any list item while
   the lock is dropped) rather than referencing the removed
   list_for_each_entry_safe() iterator.  (Chuck Lever.)
 1 file changed, 25 insertions(+), 19 deletions(-)

diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
index 6b9c399b89dfb..33f72fcf65458 100644
--- a/fs/nfsd/nfs4state.c
+++ b/fs/nfsd/nfs4state.c
@@ -6728,30 +6728,36 @@ static void nfsd4_ssc_shutdown_umount(struct nfsd_net *nn)
 static void nfsd4_ssc_expire_umount(struct nfsd_net *nn)
 {
 	bool do_wakeup = false;
-	struct nfsd4_ssc_umount_item *ni = NULL;
-	struct nfsd4_ssc_umount_item *tmp;
+	struct nfsd4_ssc_umount_item *ni;
 
+restart:
 	spin_lock(&nn->nfsd_ssc_lock);
-	list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
-		if (time_after(jiffies, ni->nsui_expire)) {
-			if (refcount_read(&ni->nsui_refcnt) > 1)
-				continue;
+	list_for_each_entry(ni, &nn->nfsd_ssc_mount_list, nsui_list) {
+		if (!time_after(jiffies, ni->nsui_expire))
+			break;
+		if (refcount_read(&ni->nsui_refcnt) > 1)
+			continue;
 
-			/* mark being unmount */
-			ni->nsui_busy = true;
-			spin_unlock(&nn->nfsd_ssc_lock);
-			mntput(ni->nsui_vfsmount);
-			spin_lock(&nn->nfsd_ssc_lock);
+		/* mark being unmount */
+		ni->nsui_busy = true;
+		spin_unlock(&nn->nfsd_ssc_lock);
+		mntput(ni->nsui_vfsmount);
+		spin_lock(&nn->nfsd_ssc_lock);
 
-			/* waiters need to start from begin of list */
-			list_del(&ni->nsui_list);
-			kfree(ni);
+		/* waiters need to start from begin of list */
+		list_del(&ni->nsui_list);
+		kfree(ni);
 
-			/* wakeup ssc_connect waiters */
-			do_wakeup = true;
-			continue;
-		}
-		break;
+		/* wakeup ssc_connect waiters */
+		do_wakeup = true;
+		/*
+		 * Concurrent nfsd4_ssc_cancel_dul() can free any item
+		 * on the list under nfsd_ssc_lock while mntput() runs
+		 * above.  Restart from the head; the list is short and
+		 * the expire worker is periodic, so this is cheap.
+		 */
+		spin_unlock(&nn->nfsd_ssc_lock);
+		goto restart;
 	}
 	if (do_wakeup)
 		wake_up_all(&nn->nfsd_ssc_waitq);
-- 
2.53.0