>On Sun, 28 Jun 1998, Chris Evans wrote:
>
>>Jun 27 14:18:03 ferret kernel: esi: 0057e8d0 edi: 00000000 ebp: 00000001 esp: 037f8f2c
> ^^^^^^^^^^^^^
>
>This patch should avoid the Oops, but it' s not the fix, the bug is still
>somewhere (I really don' t know where ;-).
I seen that 2.0.34 adds some xxx_bh_atomic() around release_sock(). With
this patch I moved lock_socket() and release_sock() in section bh_atomic,
but I really don' t know if it could make sense (nor I have compiled or
booted the new kernel). This patch is only a _suggestion_ for kernel
hackers that knows how the TCP layer works so _don' t_ apply it to your
kernel!
===================================================================
RCS file: linux/net/ipv4/tcp.c,v
retrieving revision 1.1
diff -u -r1.1 linux/net/ipv4/tcp.c
--- linux/net/ipv4/tcp.c 1998/06/29 00:02:44 1.1
+++ linux/net/ipv4/tcp.c 1998/06/29 00:05:47
@@ -2066,12 +2066,12 @@
add_wait_queue(sk->sleep, &wait);
for (;;) {
+ release_sock(sk);
current->state = TASK_INTERRUPTIBLE;
end_bh_atomic();
- release_sock(sk);
schedule();
- lock_sock(sk);
start_bh_atomic();
+ lock_sock(sk);
skb = tcp_find_established(sk);
if (skb)
break;
@@ -2103,7 +2103,8 @@
if (sk->state != TCP_LISTEN)
goto no_listen;
- lock_sock(sk);start_bh_atomic();
+ start_bh_atomic();
+ lock_sock(sk);
skb = tcp_find_established(sk);
if (skb) {
@@ -2114,8 +2115,8 @@
sk->ack_backlog--;
error = 0;
out:
- end_bh_atomic();
release_sock(sk);
+ end_bh_atomic();
no_listen:
sk->err = error;
return newsk;
Andrea[s] Arcangeli
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu