tcp.c::wait_for_tcp_memory() buggy ?

From: Rik van Riel (riel@conectiva.com.br)
Date: Sat Oct 28 2000 - 22:28:38 EST

Hi Davem,

I can't quite put my finger on what wait_for_tcp_memory() is
supposed to do, but the code looks EXTREMELY suspect and I've
had a report of somebody looping in the for(;;) loop in that
function without ever exiting and getting his TCP connection
stuck there...

Also, the locking inside the loop seems fragile, to say the
least.

from tcp.c:

  865 if (tcp_memory_free(sk) && !vm_wait)
  866 break;

  880 release_sock(sk);
  881 if (!tcp_memory_free(sk) || vm_wait)
  882 current_timeo = schedule_timeout(current_timeo);
  883 lock_sock(sk);

Here we hold the lock for the entire loop (meaning that
other people cannot make the exit condition on line 865
come true.

Except for doing a test on tcp_memory_free(sk), where we
do NOT hold the lock we're so dutifully clinging to during
the rest of the loop...

As I said, I can't put my finger down on what exactly is
wrong, but this code looks subtle enough that, together
with the bugreport I got (on IRC), I have the feeling that
it just _can't_ be right ...

regards,

Rik 

--
"What you're running that piece of shit Gnome?!?!"
       -- Miguel de Icaza, UKUUG 2000

http://www.conectiva.com/		http://www.surriel.com/

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

This archive was generated by hypermail 2b29 : Tue Oct 31 2000 - 21:00:24 EST