Date: Wed, 28 Jul 1999 12:29:15 -0400 (EDT) From: Mikhail Teterin <mi@misha.cisco.com> To: FreeBSD-gnats-submit@freebsd.org Cc: eischen@VIGRID.COM Subject: bin/12853: resend: uthread_init.c PANICs in case of (legally) close stdin Message-ID: <199907281629.MAA09267@misha.cisco.com>
next in thread | raw e-mail | index | archive | help
>Number: 12853
>Category: bin
>Synopsis: uthread_init.c PANICs in case of (legally) close stdin
>Confidential: no
>Severity: serious
>Priority: high
>Responsible: freebsd-bugs
>State: open
>Quarter:
>Keywords:
>Date-Required:
>Class: sw-bug
>Submitter-Id: current-users
>Arrival-Date: Wed Jul 28 09:40:01 PDT 1999
>Closed-Date:
>Last-Modified:
>Originator: Mikhail Teterin
>Release: FreeBSD 3.2-STABLE i386
>Organization:
Virtual Estates, Inc.
>Environment:
>Description:
It appears, a thread-using (linked with -lc_r) application
must have a valid stdin at startup (and stdout, and stderr).
Otherwise, it will PANIC("Cannot get stdio flags") and may
then seg-fault (because of the closed stderr too).
The discussion of the matter with John Birrell and Dan Eischen
yielded a patch below, which fixed our problem.
>How-To-Repeat:
We are replacing the HylaFAX's ${PREFIX}/sbin/faxsend with our
own thread-using application. The app works fine when started
from command line, but dies a horrible death when started
normally -- by HylaFAX's ${PREFIX}/sbin/faxq.
I managed to produce the ktrace (had to make /usr/bin/ktrace
suid), which helped me identify the problem and, most
importantly, helped the "thread people" come up with the fix:
1157 ktrace RET ktrace 0
1157 ktrace CALL execve(0xbfbfdc23,0xbfbfdaec,0xbfbfdb00)
1157 ktrace NAMI "/usr/local/sbin/faxsend.bin"
1157 faxsend.bin RET execve 0
1157 faxsend.bin CALL getpid
1157 faxsend.bin RET getpid 1157/0x485
1157 faxsend.bin CALL fcntl(0,0x3,0)
1157 faxsend.bin RET fcntl -1 errno 9 Bad file descriptor
1157 faxsend.bin CALL write(0x2,0xbfbfd9ac,0x6e)
1157 faxsend.bin GIO fd 2 wrote 110 bytes
"Fatal error 'Cannot get stdio flags' at line ? in file /usr/src/lib/libc_r/ut\
hread/uthread_init.c (errno = ?)
"
1157 faxsend.bin RET write 110/0x6e
1157 faxsend.bin CALL setitimer(0x2,0xbfbfd968,0)
1157 faxsend.bin RET setitimer 0
1157 faxsend.bin CALL close(0xffffffff)
1157 faxsend.bin RET close -1 errno 9 Bad file descriptor
1157 faxsend.bin CALL close(0xffffffff)
1157 faxsend.bin RET close -1 errno 9 Bad file descriptor
1157 faxsend.bin PSIG SIGSEGV SIG_DFL
>Fix:
This fix does NOT fix the seg-fault itself, but only prevents
the panic from happening in this legal case.
I hope, this gets commited before the seg-fault is hunted down,
as it is a different problem.
--- uthread_init.c.orig Wed Jul 28 10:48:25 1999
+++ uthread_init.c Wed Jul 28 10:33:50 1999
@@ -118,5 +118,13 @@
if ((_pthread_stdio_flags[i] =
- _thread_sys_fcntl(i,F_GETFL, NULL)) == -1)
- PANIC("Cannot get stdio flags");
-
+ _thread_sys_fcntl(i,F_GETFL, NULL)) == -1 && errno != EBADF) {
+ const char *panics[] = {
+ "Cannot initialize " "stdin" " file descriptor "
+ "table entry",
+ "Cannot initialize " "stdout" " file descriptor "
+ "table entry",
+ "Cannot initialize " "stderr" " file descriptor "
+ "table entry"
+ };
+ PANIC(panics[i]);
+ }
/*
>Release-Note:
>Audit-Trail:
>Unformatted:
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-bugs" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199907281629.MAA09267>