Linux NFS development
 help / color / mirror / Atom feed
From: Jeff Layton <jlayton@kernel.org>
To: Chuck Lever <chuck.lever@oracle.com>, NeilBrown <neil@brown.name>,
	 Olga Kornievskaia <okorniev@redhat.com>,
	Dai Ngo <Dai.Ngo@oracle.com>,  Tom Talpey <tom@talpey.com>,
	Lorenzo Bianconi <lorenzo@kernel.org>,
	 Anna Schumaker <anna.schumaker@oracle.com>,
	 Trond Myklebust <trondmy@kernel.org>,
	Anna Schumaker <anna@kernel.org>,
	 Mike Snitzer <snitzer@kernel.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>, Chris Mason <clm@meta.com>,
	 linux-nfs@vger.kernel.org, linux-kernel@vger.kernel.org,
	 Trond Myklebust <trond.myklebust@hammerspace.com>,
	 Jeff Layton <jlayton@kernel.org>
Subject: [PATCH v2 8/9] nfsd: hold net namespace reference for delayed-dispose nfsd_files
Date: Tue, 02 Jun 2026 12:23:20 -0400	[thread overview]
Message-ID: <20260602-nfsd-testing-v2-8-e4ea62e3cd5c@kernel.org> (raw)
In-Reply-To: <20260602-nfsd-testing-v2-0-e4ea62e3cd5c@kernel.org>

Take a net-namespace reference in nfsd_file_dispose_list_delayed()
(get_net) and release it in nfsd_file_free() (put_net), so that
nf_net is always valid for files that the GC or shrinker has isolated
from the hash table and LRU -- which __nfsd_file_cache_purge() cannot
see.

Without this, nf_net can dangle for in-flight files whose net namespace
is torn down concurrently, causing a use-after-free when
nfsd_file_dispose_list_delayed() calls net_generic(nf->nf_net, ...).

Only files entering the delayed-dispose path need the net reference;
files freed synchronously (non-GC stateful opens, purge, direct put)
are always freed while the net namespace is still alive, so they skip
get_net()/put_net() entirely.  A new NFSD_FILE_NET_HELD flag tracks
whether a given nfsd_file holds a net reference.

Because nfsd_file_free() may now call put_net(nf->nf_net), the old
nfsd_file_put_local() pattern of returning nf->nf_net after
nfsd_file_put() is unsafe -- put_net() could theoretically drop the
last net namespace reference, leaving the returned pointer stale.
Fix this by moving the nfsd_net_put() call into nfsd_file_put_local()
itself, before the nfsd_file_put() that may trigger nfsd_file_free().
The function now returns void and the caller no longer needs to handle
the net reference.

Fixes: 43fd953fa7e2 ("nfsd: simplify the delayed disposal list code")
Assisted-by: Claude:claude-opus-4-6
Signed-off-by: Jeff Layton <jlayton@kernel.org>
---
 fs/nfsd/filecache.c        | 34 ++++++++++++++++++++++++++--------
 fs/nfsd/filecache.h        |  3 ++-
 fs/nfsd/localio.c          |  4 ++--
 include/linux/nfslocalio.h |  9 ++-------
 4 files changed, 32 insertions(+), 18 deletions(-)

diff --git a/fs/nfsd/filecache.c b/fs/nfsd/filecache.c
index 03f01a0beced..957fe57be063 100644
--- a/fs/nfsd/filecache.c
+++ b/fs/nfsd/filecache.c
@@ -295,6 +295,9 @@ nfsd_file_free(struct nfsd_file *nf)
 	if (WARN_ON_ONCE(!list_empty(&nf->nf_lru)))
 		return;
 
+	if (test_bit(NFSD_FILE_NET_HELD, &nf->nf_flags))
+		put_net(nf->nf_net);
+
 	call_rcu(&nf->nf_rcu, nfsd_file_slab_free);
 }
 
@@ -375,24 +378,28 @@ nfsd_file_put(struct nfsd_file *nf)
 }
 
 /**
- * nfsd_file_put_local - put nfsd_file reference and arm nfsd_net_put in caller
+ * nfsd_file_put_local - put nfsd_file reference and release nfsd_net ref
  * @pnf: nfsd_file of which to put the reference
  *
- * First save the associated net to return to caller, then put
- * the reference of the nfsd_file.
+ * Drops both the nfsd_file reference and the associated nfsd_net
+ * reference.  The nfsd_net ref is released before the file ref so
+ * that put_net() inside nfsd_file_free() cannot drop the last net
+ * namespace reference while the caller still needs it.
  */
-struct net *
+void
 nfsd_file_put_local(struct nfsd_file __rcu **pnf)
 {
 	struct nfsd_file *nf;
-	struct net *net = NULL;
 
 	nf = unrcu_pointer(xchg(pnf, NULL));
 	if (nf) {
-		net = nf->nf_net;
+		struct net *net = nf->nf_net;
+
+		rcu_read_lock();
+		nfsd_net_put(net);
+		rcu_read_unlock();
 		nfsd_file_put(nf);
 	}
-	return net;
 }
 
 /**
@@ -433,9 +440,20 @@ nfsd_file_dispose_list_delayed(struct list_head *dispose)
 	while (!list_empty(dispose)) {
 		struct nfsd_file *nf = list_first_entry(dispose,
 						struct nfsd_file, nf_gc);
-		struct nfsd_net *nn = net_generic(nf->nf_net, nfsd_net_id);
+		struct nfsd_net *nn;
 		struct svc_serv *serv;
 
+		/*
+		 * Pin the net namespace so nf_net stays valid while the
+		 * file sits on the per-net dispose list.  Callers (the GC
+		 * worker, shrinker, and fsnotify callbacks) always run
+		 * before nfsd_net_exit(), so nf_net is still live here.
+		 * The matching put_net() is in nfsd_file_free().
+		 */
+		get_net(nf->nf_net);
+		set_bit(NFSD_FILE_NET_HELD, &nf->nf_flags);
+
+		nn = net_generic(nf->nf_net, nfsd_net_id);
 		spin_lock(&nn->fcache_dispose_lock);
 		list_move_tail(&nf->nf_gc, &nn->fcache_dispose_list);
 		spin_unlock(&nn->fcache_dispose_lock);
diff --git a/fs/nfsd/filecache.h b/fs/nfsd/filecache.h
index 683b6437cacc..7ae3c0ea0a2a 100644
--- a/fs/nfsd/filecache.h
+++ b/fs/nfsd/filecache.h
@@ -45,6 +45,7 @@ struct nfsd_file {
 #define NFSD_FILE_REFERENCED	(2)
 #define NFSD_FILE_GC		(3)
 #define NFSD_FILE_RECENT	(4)
+#define NFSD_FILE_NET_HELD	(5)
 	unsigned long		nf_flags;
 	refcount_t		nf_ref;
 	unsigned char		nf_may;
@@ -66,7 +67,7 @@ void nfsd_file_cache_shutdown(void);
 int nfsd_file_cache_start_net(struct net *net);
 void nfsd_file_cache_shutdown_net(struct net *net);
 void nfsd_file_put(struct nfsd_file *nf);
-struct net *nfsd_file_put_local(struct nfsd_file __rcu **nf);
+void nfsd_file_put_local(struct nfsd_file __rcu **nf);
 struct nfsd_file *nfsd_file_get(struct nfsd_file *nf);
 struct file *nfsd_file_file(struct nfsd_file *nf);
 void nfsd_file_close_inode_sync(struct inode *inode);
diff --git a/fs/nfsd/localio.c b/fs/nfsd/localio.c
index c3eb0557b3e1..e3295bae75a4 100644
--- a/fs/nfsd/localio.c
+++ b/fs/nfsd/localio.c
@@ -40,8 +40,8 @@
  * avoid all the NFS overhead with reads, writes and commits.
  *
  * On successful return, returned nfsd_file will have its nf_net member
- * set. Caller (NFS client) is responsible for calling nfsd_net_put and
- * nfsd_file_put (via nfs_to_nfsd_file_put_local).
+ * set. Caller (NFS client) is responsible for calling nfsd_file_put
+ * (via nfs_to_nfsd_file_put_local), which also releases the nfsd_net ref.
  */
 static struct nfsd_file *
 nfsd_open_local_fh(struct net *net, struct auth_domain *dom,
diff --git a/include/linux/nfslocalio.h b/include/linux/nfslocalio.h
index 3d91043254e6..7267a69092d1 100644
--- a/include/linux/nfslocalio.h
+++ b/include/linux/nfslocalio.h
@@ -62,7 +62,7 @@ struct nfsd_localio_operations {
 						const struct nfs_fh *,
 						struct nfsd_file __rcu **pnf,
 						const fmode_t);
-	struct net *(*nfsd_file_put_local)(struct nfsd_file __rcu **);
+	void (*nfsd_file_put_local)(struct nfsd_file __rcu **);
 	struct file *(*nfsd_file_file)(struct nfsd_file *);
 	void (*nfsd_file_dio_alignment)(struct nfsd_file *,
 					u32 *, u32 *, u32 *);
@@ -96,12 +96,7 @@ static inline void nfs_to_nfsd_file_put_local(struct nfsd_file __rcu **localio)
 	 * must prevent nfsd shutdown from completing as nfs_close_local_fh()
 	 * does by blocking the nfs_uuid from being finally put.
 	 */
-	struct net *net;
-
-	net = nfs_to->nfsd_file_put_local(localio);
-
-	if (net)
-		nfs_to_nfsd_net_put(net);
+	nfs_to->nfsd_file_put_local(localio);
 }
 
 #else   /* CONFIG_NFS_LOCALIO */

-- 
2.54.0


  parent reply	other threads:[~2026-06-02 16:23 UTC|newest]

Thread overview: 15+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-06-02 16:23 [PATCH v2 0/9] nfsd: fixes for locally-triggerable bugs Jeff Layton
2026-06-02 16:23 ` [PATCH v2 1/9] nfsd: defer vfree of compound ops to fix rpc_status UAF Jeff Layton
2026-06-02 16:23 ` [PATCH v2 2/9] nfsd: hold rcu across localio cmpxchg retry Jeff Layton
2026-06-02 16:23 ` [PATCH v2 3/9] nfs/localio: fix ref leak on nfs_uuid_add_file failure Jeff Layton
2026-06-02 16:23 ` [PATCH v2 4/9] nfsd: guard nfsd_serv deref in nfsd_file_net_dispose Jeff Layton
2026-06-02 16:23 ` [PATCH v2 5/9] nfsd: widen nfsd_genl_rqstp address fields to sockaddr_storage Jeff Layton
2026-06-02 16:23 ` [PATCH v2 6/9] nfsd: fix refcount leak in nfsd_file_lru_add on insertion failure Jeff Layton
2026-06-02 16:23 ` [PATCH v2 7/9] nfsd: fix fcache_disposal UAF by inlining dispose state into nfsd_net Jeff Layton
2026-06-02 16:23 ` Jeff Layton [this message]
2026-06-03 17:33   ` [PATCH v2 8/9] nfsd: hold net namespace reference for delayed-dispose nfsd_files Chuck Lever
2026-06-03 17:50     ` Jeff Layton
2026-06-03 18:20       ` Chuck Lever
2026-06-03 19:15         ` Jeff Layton
2026-06-02 16:23 ` [PATCH v2 9/9] nfsd: unify cleanups in nfsd_cross_mnt() exits Jeff Layton
2026-06-03 20:30 ` [PATCH v2 0/9] nfsd: fixes for locally-triggerable bugs Chuck Lever

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20260602-nfsd-testing-v2-8-e4ea62e3cd5c@kernel.org \
    --to=jlayton@kernel.org \
    --cc=Dai.Ngo@oracle.com \
    --cc=anna.schumaker@oracle.com \
    --cc=anna@kernel.org \
    --cc=chuck.lever@oracle.com \
    --cc=clm@meta.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-nfs@vger.kernel.org \
    --cc=lorenzo@kernel.org \
    --cc=neil@brown.name \
    --cc=okorniev@redhat.com \
    --cc=snitzer@kernel.org \
    --cc=tom@talpey.com \
    --cc=trond.myklebust@hammerspace.com \
    --cc=trondmy@kernel.org \
    --cc=viro@zeniv.linux.org.uk \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox