Date: Fri, 28 Jun 1996 18:42:18 +1000 From: Bruce Evans <bde@zeta.org.au> To: cimaxp1!jb@werple.net.au, haldjas.folklore.ee!narvi@melb.werple.net.au Cc: FreeBSD.org!hackers@melb.werple.net.au Subject: Re: Threads for -stable Message-ID: <199606280842.SAA20014@godzilla.zeta.org.au>
next in thread | raw e-mail | index | archive | help
>libc_r from -current should work with -stable. We use it with 2.1R. We
>*can't* use it with -current because of the pipe problems.
This reminds me that your pipe problem is supposed to be fixed (by
disabling an optimization) in -current, but isn't because the wrong NBIO
flag is checked.
The per-pipe NBIO flag is still wrong for another reason. O_NONBLOCK is
supposed to be per-fd. fcntl() converts O_NONBLOCK to FIONBIO for the
benefit of old drivers but it is possible for the per-fd flag to differ
from the per-device flag and the behaviour when the per-device flag is
modified while another process is blocked waiting for i/o is unpredictable.
This problem affects pipes, sockets, tty_snoop, bpf and tun.
Bruce
*** sys_pipe.c~ Mon Jun 17 23:27:59 1996
--- sys_pipe.c Wed Jun 19 15:40:04 1996
***************
*** 737,741 ****
* direct write mechanism.
*/
! if ((wpipe->pipe_state & PIPE_NBIO) == 0 &&
(amountpipekva < LIMITPIPEKVA) &&
(uio->uio_iov->iov_len >= PIPE_MINDIRECT)) {
--- 748,752 ----
* direct write mechanism.
*/
! if (!nbio &&
(amountpipekva < LIMITPIPEKVA) &&
(uio->uio_iov->iov_len >= PIPE_MINDIRECT)) {
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199606280842.SAA20014>