Skip to contentSkip to conversation

select, jiffies, and SIGALRM

DISCUSSION Robert Hancock 2 participants 2 messages Oct 14, 2004 1 message not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw

I see calls to getitimer, so I'm assuming it's also using setitimer. SIGALRM is what you get when those timers go off - if it's not handling that, that's a bug, but presumably the timer is in there for a reason..

----- Original Message ----- From: "Vx Glenn" <[email protected]>
Newsgroups: fa.linux.kernel
To: <[email protected]>; <[email protected]>
Sent: Wednesday, October 13, 2004 10:13 AM
Subject: select, jiffies, and SIGALRM

Quoted 42 lines

Hi all,

I am seeing an issue relating to the jiffies counter wrapping around at 0x7FFFFFFF.

This is a legacy application, and when it runs on 32-bit Unix-Like OS's, the application silently dies without leaving core after 248 days.

I was able to manipulate the jiffies counter and run the application. I was able to reproduce the problem. I captured an strace log, and I see that SIGALRM (alarm clock) is raised after select times out (because of no data).

I can add a signal handler to intercept the SIGALRM. But my question is, why should the signal be raised?

---[ strace.log ]---
select(1024, [3 4 5 6], NULL, NULL, {0, 320000}) = 0 (Timeout) getitimer(ITIMER_REAL, {it_interval={2147157, 520}, it_value={0, 684895}}) = 0 adjtimex({modes=32769, offset=0, freq=0, maxerror=16384000, esterror=16384000, status=64, constant=2, precision=1, tolerance=33554432, time={1097551596, 43475}}) = 5 getitimer(ITIMER_REAL, {it_interval={2147157, 520}, it_value={0, 684895}}) = 0 select(1024, [3 4 5 6], NULL, NULL, {1, 0}) = ? ERESTARTNOHAND (To be restarted) --- SIGALRM (Alarm clock) @ 0 (0) ---
Process 4881 detached
---[ eof strace.log ]---

Anyone have any ideas?

Signature · 8 lines
You're not your Job;
You're not the contents of your wallet.
You're the all singing all dancing crap of the world
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/ 
#2 lore raw

Thanks for the insight. I looked more closely at the trace I have and I see the POSIX timer is adversely affected by the wrap-around of the jiffies counter.

getitimer(ITIMER_REAL, {it_interval={2147157, 520}, it_value={2146931, 982728}}) = 0
getitimer(ITIMER_REAL, {it_interval={2147157, 520}, it_value={2146931, 962731}}) = 0
getitimer(ITIMER_REAL, {it_interval={2147157, 520}, it_value={2146931, 962731}}) = 0
getitimer(ITIMER_REAL, {it_interval={2147157, 520}, it_value={131, 601993}}) = 0 getitimer(ITIMER_REAL, {it_interval={2147157, 520}, it_value={131, 601993}}) = 0 getitimer(ITIMER_REAL, {it_interval={2147157, 520}, it_value={131, 601993}}) = 0 getitimer(ITIMER_REAL, {it_interval={2147157, 520}, it_value={131, 599993}}) = 0 getitimer(ITIMER_REAL, {it_interval={2147157, 520}, it_value={131, 599993}}) = 0

And when the SIGALRM fires, the app does not handle it.

My next question is, should the POSIX timer be affected like this? I guess if it uses the jiffies counter, like everything else, it probably would.

Quoted 66 lines · Robert Hancock

I see calls to getitimer, so I'm assuming it's also using setitimer. SIGALRM is what you get when those timers go off - if it's not handling that, that's a bug, but presumably the timer is in there for a reason..

----- Original Message -----
From: "Vx Glenn" <[email protected]>
Newsgroups: fa.linux.kernel
To: <[email protected]>; <[email protected]>
Sent: Wednesday, October 13, 2004 10:13 AM
Subject: select, jiffies, and SIGALRM

Quoted 36 lines

Hi all,

I am seeing an issue relating to the jiffies counter wrapping around at 0x7FFFFFFF.

This is a legacy application, and when it runs on 32-bit Unix-Like OS's, the application silently dies without leaving core after 248 days.

I was able to manipulate the jiffies counter and run the application. I was able to reproduce the problem. I captured an strace log, and I see that SIGALRM (alarm clock) is raised after select times out (because of no data).

I can add a signal handler to intercept the SIGALRM. But my question is, why should the signal be raised?

---[ strace.log ]---
select(1024, [3 4 5 6], NULL, NULL, {0, 320000}) = 0 (Timeout) getitimer(ITIMER_REAL, {it_interval={2147157, 520}, it_value={0, 684895}}) = 0
adjtimex({modes=32769, offset=0, freq=0, maxerror=16384000, esterror=16384000, status=64, constant=2, precision=1, tolerance=33554432, time={1097551596, 43475}}) = 5 getitimer(ITIMER_REAL, {it_interval={2147157, 520}, it_value={0, 684895}}) = 0
select(1024, [3 4 5 6], NULL, NULL, {1, 0}) = ? ERESTARTNOHAND (To be restarted)
--- SIGALRM (Alarm clock) @ 0 (0) ---
Process 4881 detached
---[ eof strace.log ]---

Anyone have any ideas?

Quoted 9 lines

--
You're not your Job;
You're not the contents of your wallet.
You're the all singing all dancing crap of the world
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Signature · 2 lines
Get Firefox 
http://getfirefox.com