Date: Mon, 22 Mar 1999 14:18:25 +0800 From: Peter Wemm <peter@netplex.com.au> To: Ollivier Robert <roberto@keltia.freenix.fr> Cc: "FreeBSD Hackers' list" <freebsd-hackers@FreeBSD.ORG> Subject: Re: Fwd: Re: cron problem? Message-ID: <199903220618.OAA53483@spinner.netplex.com.au> In-Reply-To: Your message of "Mon, 22 Mar 1999 00:13:09 %2B0100." <19990322001309.A53335@keltia.freenix.fr>
next in thread | previous in thread | raw e-mail | index | archive | help
Ollivier Robert wrote:
> Wietse found the following while writing Postfix. Can anyone look at it ?
A followup is on the way, I tracked down what is causing it.
cron(8) is passed stdout etc pointing to /dev/console at startup.
[some time later at the end of the boot sequence]
init fires up a getty on /dev/ttyv0
revoke("/dev/ttyv0") is redirecting the /dev/console fd's to deadfs
[later]
cron does a popen() to sendmail (postfix in this case)
stdin is a pipe, stdout/err are inherited deadfs descriptors formerly to
/dev/console
fstat() on the inherited descriptor returns EBADF.
methods of testing open fd's no longer work.
> ----- Forwarded message from Wietse Venema <wietse@porcupine.org> -----
>
> Subject: Re: cron problem?
> Date: Fri, 19 Mar 1999 14:59:34 -0500 (EST)
> From: wietse@porcupine.org (Wietse Venema)
>
> Found it. I'll post a patch later. This must be a *BSD oddity.
>
> Wietse
>
> Problem:
>
> Vixie cron output from an /etc/crontab entry is not mailed,
> unless the crontab line ends with an explicit mail command
>
> System:
>
> FreeBSD 2.x, BSD/OS 2.x ... (yes I'm not running the latest)
>
> Analysis:
>
> For the sake of security and robustness, Postfix tries to
> ensure that stdin/out/err are open before opening anything
> else. The following code fragment appears in many programs
> that I wrote in the course of time:
>
> for (fd = 0; fd < 3; fd++)
> if (fstat(fd, &st) == -1 && open("/dev/null", O_WRONLY) != fd)
> msg_fatal("open /dev/null: %m");
>
> This fragment examines the status of stdin/stdout/stderr.
> When fstat() returns an error, the code assumes the file
> descriptor is not in use, and expects that the next open()
> will allocate that file descriptor.
>
> However, the kernel trace shows a small surprise:
>
> 3817 sendmail CALL fstat(0,0xefbfddc8)
> 3817 sendmail RET fstat 0
> 3817 sendmail CALL fstat(0x1,0xefbfddc8)
> 3817 sendmail RET fstat -1 errno 9 Bad file descriptor
> 3817 sendmail CALL open(0x1c8c,0x1,0xefbfde4c)
> 3817 sendmail NAMI "/dev/null"
> 3817 sendmail RET open 3
>
> So: fstat(1) returns EBADF, and open() returns 3.
>
> Conclusion: file descriptor 1 is still in use.
>
> *** Why on earth does fstat() return EBADF on an open file
> descriptor? ***
>
> Workaround:
>
> Close the file descriptor after error return from fstat().
>
> The following code does the job.
>
> if (fstat(fd, &st) == -1
> && (close(fd), open("/dev/null", O_WRONLY)) != fd)
> msg_fatal("open /dev/null: %m");
>
>
>
>
> ----- End forwarded message -----
>
> --
> Ollivier ROBERT -=- FreeBSD: The Power to Serve! -=- roberto@keltia.freenix.f
r
> FreeBSD keltia.freenix.fr 4.0-CURRENT #70: Sat Feb 27 09:43:08 CET 1999
>
>
>
> To Unsubscribe: send mail to majordomo@FreeBSD.org
> with "unsubscribe freebsd-hackers" in the body of the message
>
Cheers,
-Peter
--
Peter Wemm <peter@netplex.com.au> Netplex Consulting
"No coffee, No workee!" :-)
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199903220618.OAA53483>