Date: Mon, 05 Jul 1999 18:18:08 PDT From: Craig Leres <leres@ee.lbl.gov> To: FreeBSD-gnats-submit@freebsd.org Cc: leres@ee.lbl.gov (Craig Leres) Subject: bin/12528: [PATCH] tip's "tipout" child doesn't always exit Message-ID: <199907060118.SAA03813@hot.ee.lbl.gov>
next in thread | raw e-mail | index | archive | help
>Number: 12528
>Category: bin
>Synopsis: [PATCH] tip's "tipout" child doesn't always exit
>Confidential: no
>Severity: non-critical
>Priority: low
>Responsible: freebsd-bugs
>State: open
>Quarter:
>Keywords:
>Date-Required:
>Class: sw-bug
>Submitter-Id: current-users
>Arrival-Date: Mon Jul 5 18:20:02 PDT 1999
>Closed-Date:
>Last-Modified:
>Originator: Craig Leres
>Release: FreeBSD 3.2-RELEASE i386
>Organization:
Lawrence Berkeley National Laboratory
>Environment:
>Description:
If the user exists abruptly, tip's "tipout" child can hang
around forever. Since the lock file doesn't get cleaned up,
this prevents other users from accessing the target device.
>How-To-Repeat:
Here's how I found this:
- slogin to a host
- tip to a device
- terminate slogin with the "~." sequence
- notice that the 2nd tip process still exists
- notice that the lockfile still exists
I used truss on the 2nd process. Before terminating we have:
syscall sigprocmask(0x3,0x0)
returns 18528 (0x4860)
syscall read(0x3,0xbfbfd26c,0x400)
returns 7 (0x7)
syscall sigprocmask(0x1,0x4860)
returns 0 (0x0)
syscall write(1,0xbfbfd26c,7)
returns 7 (0x7)
syscall sigprocmask(0x3,0x0)
returns 18528 (0x4860)
After exiting from slogin:
syscall read(0x3,0xbfbfd26c,0x400)
returns 1 (0x1)
syscall sigprocmask(0x1,0x4860)
returns 0 (0x0)
syscall write(1,0xbfbfd26c,1)
errno 5 'Input/output error'
syscall sigprocmask(0x3,0x0)
returns 18528 (0x4860)
syscall read(0x3,0xbfbfd26c,0x400)
returns 1 (0x1)
syscall sigprocmask(0x1,0x4860)
returns 0 (0x0)
syscall write(1,0xbfbfd26c,1)
errno 5 'Input/output error'
syscall sigprocmask(0x3,0x0)
returns 18528 (0x4860)
syscall read(0x3,0xbfbfd26c,0x400)
returns 7 (0x7)
syscall sigprocmask(0x1,0x4860)
returns 0 (0x0)
This repeats until all the characters from the device are
processed and then the process becomes dormant.
>Fix:
The tipout child needs to exit if write() fails with EIO. But
if the device does not output any characters, we won't get to
this exit so the parent should also arrange for the child to be
killed when it exits.
RCS file: RCS/tip.c,v
retrieving revision 1.1
diff -c -r1.1 tip.c
*** /tmp/,RCSt1G48409 Mon Jul 5 18:13:45 1999
--- tip.c Mon Jul 5 18:13:13 1999
***************
*** 80,85 ****
--- 80,86 ----
void intprompt();
void timeout();
+ void killchild();
void cleanup();
void tipdone();
char *sname();
***************
*** 285,290 ****
--- 286,300 ----
}
void
+ killchild()
+ {
+ if (pid != 0) {
+ kill(pid, SIGTERM);
+ pid = 0;
+ }
+ }
+
+ void
cleanup()
{
***************
*** 425,430 ****
--- 435,442 ----
{
int i;
char gch, bol = 1;
+
+ atexit(killchild);
/*
* Kinda klugey here...
RCS file: RCS/tipout.c,v
retrieving revision 1.1
diff -c -r1.1 tipout.c
*** /tmp/,RCSt1E48414 Mon Jul 5 18:13:46 1999
--- tipout.c Mon Jul 5 17:49:07 1999
***************
*** 160,166 ****
omask = sigblock(ALLSIGS);
for (cp = buf; cp < buf + cnt; cp++)
*cp &= 0177;
! write(1, buf, cnt);
if (boolean(value(SCRIPT)) && fscript != NULL) {
if (!boolean(value(BEAUTIFY))) {
fwrite(buf, 1, cnt, fscript);
--- 160,167 ----
omask = sigblock(ALLSIGS);
for (cp = buf; cp < buf + cnt; cp++)
*cp &= 0177;
! if (write(1, buf, cnt) < 0)
! exit(1);
if (boolean(value(SCRIPT)) && fscript != NULL) {
if (!boolean(value(BEAUTIFY))) {
fwrite(buf, 1, cnt, fscript);
>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?199907060118.SAA03813>