[v2] net: restore the iterator to its original state when an error occurs

[PATCH Next V2] net: restore the iterator to its original state when an error occurs

Posted by Edward Adam Davis 3 days, 7 hours ago

In zerocopy_fill_skb_from_iter(), if two copy operations are performed
and the first one succeeds while the second one fails, it returns a
failure but the count in iterator has already been decremented due to
the first successful copy. This ultimately affects the local variable
rest_len in virtio_transport_send_pkt_info(), causing the remaining
count in rest_len to be greater than the actual iterator count. As a
result, packet sending operations continue even when the iterator count
is zero, which further leads to skb->len being 0 and triggers the warning
reported by syzbot [1].

Therefore, if the zerocopy operation fails, we should revert the iterator
to its original state.

The iov_iter_revert() in skb_zerocopy_iter_stream() is no longer needed
and has been removed.

[1]
'send_pkt()' returns 0, but 4096 expected
WARNING: net/vmw_vsock/virtio_transport_common.c:430 at virtio_transport_send_pkt_info+0xd1e/0xef0 net/vmw_vsock/virtio_transport_common.c:428, CPU#1: syz.0.17/5986
Call Trace:
 virtio_transport_stream_enqueue net/vmw_vsock/virtio_transport_common.c:1113 [inline]
 virtio_transport_seqpacket_enqueue+0x143/0x1c0 net/vmw_vsock/virtio_transport_common.c:841
 vsock_connectible_sendmsg+0xabf/0x1040 net/vmw_vsock/af_vsock.c:2158
 sock_sendmsg_nosec net/socket.c:727 [inline]
 __sock_sendmsg+0x21c/0x270 net/socket.c:746

Reported-by: syzbot+28e5f3d207b14bae122a@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=28e5f3d207b14bae122a
Tested-by: syzbot+28e5f3d207b14bae122a@syzkaller.appspotmail.com
Signed-off-by: Edward Adam Davis <eadavis@qq.com>
---
V1 -> V2: Remove iov_iter_revert() in skb_zerocopy_iter_stream()

 net/core/datagram.c | 6 ++++++
 net/core/skbuff.c   | 1 -
 2 files changed, 6 insertions(+), 1 deletion(-)

diff --git a/net/core/datagram.c b/net/core/datagram.c
index c285c6465923..da10465cd8a4 100644
--- a/net/core/datagram.c
+++ b/net/core/datagram.c
@@ -748,10 +748,13 @@ int __zerocopy_sg_from_iter(struct msghdr *msg, struct sock *sk,
 			    size_t length,
 			    struct net_devmem_dmabuf_binding *binding)
 {
+	struct iov_iter_state state;
 	unsigned long orig_size = skb->truesize;
 	unsigned long truesize;
 	int ret;
 
+	iov_iter_save_state(from, &state);
+
 	if (msg && msg->msg_ubuf && msg->sg_from_iter)
 		ret = msg->sg_from_iter(skb, from, length);
 	else if (binding)
@@ -759,6 +762,9 @@ int __zerocopy_sg_from_iter(struct msghdr *msg, struct sock *sk,
 	else
 		ret = zerocopy_fill_skb_from_iter(skb, from, length);
 
+	if (ret)
+		iov_iter_restore(from, &state);
+
 	truesize = skb->truesize - orig_size;
 	if (sk && sk->sk_type == SOCK_STREAM) {
 		sk_wmem_queued_add(sk, truesize);
diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index 5a1d123e7ef7..77ed045c28ff 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -1908,7 +1908,6 @@ int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
 		struct sock *save_sk = skb->sk;
 
 		/* Streams do not free skb on error. Reset to prev state. */
-		iov_iter_revert(&msg->msg_iter, skb->len - orig_len);
 		skb->sk = sk;
 		___pskb_trim(skb, orig_len);
 		skb->sk = save_sk;
-- 
2.43.0

Re: [PATCH Next V2] net: restore the iterator to its original state when an error occurs

Posted by Jakub Kicinski 3 days, 3 hours ago

On Fri, 28 Nov 2025 21:35:57 +0800 Edward Adam Davis wrote:
> In zerocopy_fill_skb_from_iter(), if two copy operations are performed
> and the first one succeeds while the second one fails, it returns a
> failure but the count in iterator has already been decremented due to
> the first successful copy. This ultimately affects the local variable
> rest_len in virtio_transport_send_pkt_info(), causing the remaining
> count in rest_len to be greater than the actual iterator count. As a
> result, packet sending operations continue even when the iterator count
> is zero, which further leads to skb->len being 0 and triggers the warning
> reported by syzbot [1].

Please follow the subsystem guidelines for posting patches:
https://www.kernel.org/doc/html/next/process/maintainer-netdev.html
Your patch breaks zerocopy tests.

Re: [PATCH Next V2] net: restore the iterator to its original state when an error occurs

Posted by Edward Adam Davis 17 hours ago

On Fri, 28 Nov 2025 09:39:46 -0800, Jakub Kicinski wrote:
> > In zerocopy_fill_skb_from_iter(), if two copy operations are performed
> > and the first one succeeds while the second one fails, it returns a
> > failure but the count in iterator has already been decremented due to
> > the first successful copy. This ultimately affects the local variable
> > rest_len in virtio_transport_send_pkt_info(), causing the remaining
> > count in rest_len to be greater than the actual iterator count. As a
> > result, packet sending operations continue even when the iterator count
> > is zero, which further leads to skb->len being 0 and triggers the warning
> > reported by syzbot [1].
> 
> Please follow the subsystem guidelines for posting patches:
> https://www.kernel.org/doc/html/next/process/maintainer-netdev.html
> Your patch breaks zerocopy tests.
I see that they all timed out. I'm not familiar with this test, how can
I get more details about it?

Re: [PATCH Next V2] net: restore the iterator to its original state when an error occurs

Posted by Jakub Kicinski 2 hours ago

On Mon,  1 Dec 2025 11:41:07 +0800 Edward Adam Davis wrote:
> On Fri, 28 Nov 2025 09:39:46 -0800, Jakub Kicinski wrote:
> > > In zerocopy_fill_skb_from_iter(), if two copy operations are performed
> > > and the first one succeeds while the second one fails, it returns a
> > > failure but the count in iterator has already been decremented due to
> > > the first successful copy. This ultimately affects the local variable
> > > rest_len in virtio_transport_send_pkt_info(), causing the remaining
> > > count in rest_len to be greater than the actual iterator count. As a
> > > result, packet sending operations continue even when the iterator count
> > > is zero, which further leads to skb->len being 0 and triggers the warning
> > > reported by syzbot [1].  
> > 
> > Please follow the subsystem guidelines for posting patches:
> > https://www.kernel.org/doc/html/next/process/maintainer-netdev.html
> > Your patch breaks zerocopy tests.  
> I see that they all timed out. I'm not familiar with this test, how can
> I get more details about it?

IIRC its was the packetdrill tests:

tools/testing/selftests/net/packetdrill/tcp_fastopen_server_basic-zero-payload.pkt
tools/testing/selftests/net/packetdrill/tcp_zerocopy_basic.pkt
tools/testing/selftests/net/packetdrill/tcp_zerocopy_batch.pkt
tools/testing/selftests/net/packetdrill/tcp_zerocopy_client.pkt
tools/testing/selftests/net/packetdrill/tcp_zerocopy_closed.pkt
tools/testing/selftests/net/packetdrill/tcp_zerocopy_epoll_edge.pkt
tools/testing/selftests/net/packetdrill/tcp_zerocopy_epoll_exclusive.pkt
tools/testing/selftests/net/packetdrill/tcp_zerocopy_epoll_oneshot.pkt
tools/testing/selftests/net/packetdrill/tcp_zerocopy_fastopen-client.pkt
tools/testing/selftests/net/packetdrill/tcp_zerocopy_fastopen-server.pkt
tools/testing/selftests/net/packetdrill/tcp_zerocopy_maxfrags.pkt
tools/testing/selftests/net/packetdrill/tcp_zerocopy_small.pkt

If you have the packetdrill command installed those _should_ be
relatively easy to run via standard kselftest commands