Re: [PATCH] af_unix: Don't set err in unix_stream_read_generic unless there was an error

The present unix_stream_read_generic contains various code sequences of
the form

err = -EDISASTER;
if (<test>)
	goto out;

This has the unfortunate side effect of possibly causing the error code
to bleed through to the final

out:
	return copied ? : err;

and then to be wrongly returned if no data was copied because the caller
didn't supply a data buffer, as demonstrated by the program available at

http://pad.lv/1540731

Change it such that err is only set if an error condition was detected.

Signed-off-by: Rainer Weikusat <redacted>
---

With unlikely() added to the two leading error checks. I've also checked
that this actually generates better code (for me at least). Without an
annotation, the non-blocking read case is treated like the others, ie, a
jump to some code at the end of the function loading the error code into
a register, followed by a 2nd jump to the termination sequence.

diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c
index 49d5093..c1e4dd7 100644
--- a/net/unix/af_unix.c
+++ b/net/unix/af_unix.c

@@ -2277,13 +2277,15 @@ static int unix_stream_read_generic(struct unix_stream_read_state *state)
 	size_t size = state->size;
 	unsigned int last_len;
 
-	err = -EINVAL;
-	if (sk->sk_state != TCP_ESTABLISHED)
+	if (unlikely(sk->sk_state != TCP_ESTABLISHED)) {
+		err = -EINVAL;
 		goto out;
+	}
 
-	err = -EOPNOTSUPP;
-	if (flags & MSG_OOB)
+	if (unlikely(flags & MSG_OOB)) {
+		err = -EOPNOTSUPP;
 		goto out;
+	}
 
 	target = sock_rcvlowat(sk, flags & MSG_WAITALL, size);
 	timeo = sock_rcvtimeo(sk, noblock);

@@ -2329,9 +2331,11 @@ again:
 				goto unlock;
 
 			unix_state_unlock(sk);
-			err = -EAGAIN;
-			if (!timeo)
+			if (!timeo) {
+				err = -EAGAIN;
 				break;
+			}
+
 			mutex_unlock(&u->readlock);
 
 			timeo = unix_stream_data_wait(sk, timeo, last,