Hi Patrick,
You wrote:
+ fds = (struct pollfd **)kmalloc(nfds * sizeof(struct pollfd *),
GFP_KERNEL);
You knew that I put that calculation in and yet didn't even bother cc'me
in your mail. If you did, I would have told you in private that your
patch is completely wrong.
Why is it wrong? Let's assume we have nfds=1024, then since sizeof(struct
pollfd) = 8 we need 8192 bytes i.e. 2 pages to store all struct pollfd.
This means two full "chunks", fds[0] and fds[1] (nchunks=1), i.e. just two
pointers. And how many pointers does your patch allocate? It allocates
1024 pointers, quite a lot more than needed, no wonder it works for you :)
Anyway, I will have a look at what's going on here as soon as I get a
chance. Can you supply a bit more info, e.g. are sure it is done on 1024
descriptors? Can you duplicate the problem with a simple testcase instead
of squid? (strace'ing the squid may give enough hints as to what
arguments to pass to poll(2) to trigger the bug)
Regards,
------
Tigran A. Aivazian | http://www.sco.com
Escalations Research Group | tel: +44-(0)1923-813796
Santa Cruz Operation Ltd | http://www.ocston.org/~tigran
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Mon Jan 31 2000 - 21:00:27 EST