On Mon, Oct 31, 2022 at 06:56 PM +01, Jakub Sitnicki wrote:

On Sat, Oct 29, 2022 at 07:44 AM +08, Wang Yufen wrote:

quoted

When running `test_sockmap` selftests, got the following warning:

WARNING: CPU: 2 PID: 197 at net/core/stream.c:205 sk_stream_kill_queues+0xd3/0xf0
Call Trace:
  <TASK>
  inet_csk_destroy_sock+0x55/0x110
  tcp_rcv_state_process+0xd28/0x1380
  ? tcp_v4_do_rcv+0x77/0x2c0
  tcp_v4_do_rcv+0x77/0x2c0
  __release_sock+0x106/0x130
  __tcp_close+0x1a7/0x4e0
  tcp_close+0x20/0x70
  inet_release+0x3c/0x80
  __sock_release+0x3a/0xb0
  sock_close+0x14/0x20
  __fput+0xa3/0x260
  task_work_run+0x59/0xb0
  exit_to_user_mode_prepare+0x1b3/0x1c0
  syscall_exit_to_user_mode+0x19/0x50
  do_syscall_64+0x48/0x90
  entry_SYSCALL_64_after_hwframe+0x44/0xae

The root case is: In commit 84472b436e76 ("bpf, sockmap: Fix more
uncharged while msg has more_data") , I used msg->sg.size replace
tosend rudely, which break the
   if (msg->apply_bytes && msg->apply_bytes < send)
scene.

Fixes: 84472b436e76 ("bpf, sockmap: Fix more uncharged while msg has more_data")
Reported-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: Wang Yufen <redacted>
Acked-by: John Fastabend <john.fastabend@gmail.com>
---
v1 -> v2: typo fixup
 net/ipv4/tcp_bpf.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/net/ipv4/tcp_bpf.c b/net/ipv4/tcp_bpf.c
index a1626af..774d481 100644
--- a/net/ipv4/tcp_bpf.c
+++ b/net/ipv4/tcp_bpf.c

@@ -278,7 +278,7 @@ static int tcp_bpf_send_verdict(struct sock *sk, struct sk_psock *psock,
 {
 	bool cork = false, enospc = sk_msg_full(msg);
 	struct sock *sk_redir;
-	u32 tosend, delta = 0;
+	u32 tosend, orgsize, sent, delta = 0;
 	u32 eval = __SK_NONE;
 	int ret;
 

@@ -333,10 +333,12 @@ static int tcp_bpf_send_verdict(struct sock *sk, struct sk_psock *psock,
 			cork = true;
 			psock->cork = NULL;
 		}
-		sk_msg_return(sk, msg, msg->sg.size);
+		sk_msg_return(sk, msg, tosend);
 		release_sock(sk);
 
+		orgsize = msg->sg.size;
 		ret = tcp_bpf_sendmsg_redir(sk_redir, msg, tosend, flags);
+		sent = orgsize - msg->sg.size;

If I'm reading the code right, it's the same as:

                sent = tosend - msg->sg.size;

If so, no need for orgsize.

Sorry, that doesn't make any sense. I misread the code.

The fix is correct.

If I can have a small ask to rename `orgsize` to something more common.

We have `orig_size` or `origsize` in use today, but no `orgsize`:

$ git grep -c '\<orig_size\>' -- net
net/core/sysctl_net_core.c:3
net/psample/psample.c:1
net/tls/tls_device.c:5
net/tls/tls_sw.c:7
$ git grep -c '\<origsize\>' -- net
net/bridge/netfilter/ebtables.c:5
net/ipv4/netfilter/arp_tables.c:10
net/ipv4/netfilter/ip_tables.c:10
net/ipv6/netfilter/ip6_tables.c:10

It reads a bit better, IMHO.

Thanks for fixing it so quickly.

Acked-by: Jakub Sitnicki <jakub@cloudflare.com>