Re: [PATCH net v3] af_unix: Read with MSG_PEEK loops if the first unread byte is OOB

[Kuniyuki Iwashima <kuniyu@amazon.com>]

From: Rao Shoaib <rao.shoaib@oracle.com>
Date: Tue, 23 Apr 2024 18:18:24 -0700
> On 4/23/24 17:15, Kuniyuki Iwashima wrote:
> > From: Rao Shoaib <Rao.Shoaib@oracle.com>
> > Date: Mon, 22 Apr 2024 02:25:03 -0700
> >> Read with MSG_PEEK flag loops if the first byte to read is an OOB byte.
> >> commit 22dd70eb2c3d ("af_unix: Don't peek OOB data without MSG_OOB.")
> >> addresses the loop issue but does not address the issue that no data
> >> beyond OOB byte can be read.
> >>
> >>>>> from socket import *
> >>>>> c1, c2 = socketpair(AF_UNIX, SOCK_STREAM)
> >>>>> c1.send(b'a', MSG_OOB)
> >> 1
> >>>>> c1.send(b'b')
> >> 1
> >>>>> c2.recv(1, MSG_PEEK | MSG_DONTWAIT)
> >> b'b'
> >>
> >> Fixes: 314001f0bf92 ("af_unix: Add OOB support")
> >> Signed-off-by: Rao Shoaib <Rao.Shoaib@oracle.com>
> >> ---
> >>  net/unix/af_unix.c | 26 ++++++++++++++------------
> >>  1 file changed, 14 insertions(+), 12 deletions(-)
> >>
> >> diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c
> >> index 9a6ad5974dff..ed5f70735435 100644
> >> --- a/net/unix/af_unix.c
> >> +++ b/net/unix/af_unix.c
> >> @@ -2658,19 +2658,19 @@ static struct sk_buff *manage_oob(struct sk_buff *skb, struct sock *sk,
> >>  		if (skb == u->oob_skb) {
> >>  			if (copied) {
> >>  				skb = NULL;
> >> -			} else if (sock_flag(sk, SOCK_URGINLINE)) {
> >> -				if (!(flags & MSG_PEEK)) {
> >> +			} else if (!(flags & MSG_PEEK)) {
> >> +				if (sock_flag(sk, SOCK_URGINLINE)) {
> >>  					WRITE_ONCE(u->oob_skb, NULL);
> >>  					consume_skb(skb);
> >> +				} else {
> >> +					skb_unlink(skb, &sk->sk_receive_queue);
> >> +					WRITE_ONCE(u->oob_skb, NULL);
> >> +					if (!WARN_ON_ONCE(skb_unref(skb)))
> >> +						kfree_skb(skb);
> >> +					skb = skb_peek(&sk->sk_receive_queue);
> > 
> > I added a comment about this case.
> 
> OK. I will sync up.
> > 
> > 
> >>  				}
> >> -			} else if (flags & MSG_PEEK) {
> >> -				skb = NULL;
> >> -			} else {
> >> -				skb_unlink(skb, &sk->sk_receive_queue);
> >> -				WRITE_ONCE(u->oob_skb, NULL);
> >> -				if (!WARN_ON_ONCE(skb_unref(skb)))
> >> -					kfree_skb(skb);
> >> -				skb = skb_peek(&sk->sk_receive_queue);
> >> +			} else if (!sock_flag(sk, SOCK_URGINLINE)) {
> >> +				skb = skb_peek_next(skb, &sk->sk_receive_queue);
> >>  			}
> >>  		}
> >>  	}
> >> @@ -2747,9 +2747,11 @@ static int unix_stream_read_generic(struct unix_stream_read_state *state,
> >>  #if IS_ENABLED(CONFIG_AF_UNIX_OOB)
> >>  		if (skb) {
> >>  			skb = manage_oob(skb, sk, flags, copied);
> >> -			if (!skb && copied) {
> >> +			if (!skb) {
> >>  				unix_state_unlock(sk);
> >> -				break;
> >> +				if (copied || (flags & MSG_PEEK))
> >> +					break;
> >> +				goto redo;
> > 
> > Here, copied == 0 && !(flags & MSG_PEEK) && skb == NULL, so it means
> > skb_peek(&sk->sk_receive_queue) above returned NULL.  Then, we need
> > not jump to the redo label, where we call the same skb_peek().
> > 
> > Instead, we can just fall through the if (!skb) clause below.
> > 
> > Thanks!
> 
> Yes that makes sense. I will submit a new version with the jump to redo
> removed.

If skb_peek_next() returns NULL, should it also fall down to the
!skb case ?

TCP is blocked in the situation.

So, I think this hunk in unix_stream_read_generic() is not needed.

---8<---
>>> from socket import *
>>> 
>>> s = socket()
>>> s.listen()
>>> 
>>> c1 = socket()
>>> c1.connect(s.getsockname())
>>> c2, _ = s.accept()
>>> 
>>> c1.send(b'h', MSG_OOB)
1
>>> c2.recv(5, MSG_PEEK)
^C
---8<---