public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
* Re: 2.5.51 nanosleep fails
@ 2002-12-12 22:27 Jim Houston
  2002-12-13  1:29 ` Stephen Rothwell
  0 siblings, 1 reply; 6+ messages in thread
From: Jim Houston @ 2002-12-12 22:27 UTC (permalink / raw)
  To: Marco d'Itri, julie.n.fleischer, linux-kernel,
	 Linus Torvalds


Hi Linus, Marco, Julie, Everyone,

It really is broken.  I just didn't try enough combinations.

Thanks Julie, the test suite really helps.

The problem is that nanosleep didn't check for the NULL pointer
case and always tries to copy out the remaining time.  The attached
patch will fix this problem.

Jim Houston - Concurrent Computer Corp.

--- linux-2.5.51.orig/kernel/timer.c	Thu Dec 12 17:08:30 2002
+++ linux-2.5.51.new/kernel/timer.c	Thu Dec 12 16:59:10 2002
@@ -1040,7 +1040,7 @@
 		jiffies_to_timespec(expire, &t);
 
 		ret = -ERESTART_RESTARTBLOCK;
-		if (copy_to_user(rmtp, &t, sizeof(t)))
+		if (rmtp && copy_to_user(rmtp, &t, sizeof(t)))
 			ret = -EFAULT;
 		/* The 'restart' block is already filled in */
 	}
@@ -1067,7 +1067,7 @@
 	if (expire) {
 		struct restart_block *restart;
 		jiffies_to_timespec(expire, &t);
-		if (copy_to_user(rmtp, &t, sizeof(t)))
+		if (rmtp && copy_to_user(rmtp, &t, sizeof(t)))
 			return -EFAULT;
 
 		restart = &current_thread_info()->restart_block;

^ permalink raw reply	[flat|nested] 6+ messages in thread
* RE: 2.5.51 nanosleep fails
@ 2002-12-12 19:59 Fleischer, Julie N
  0 siblings, 0 replies; 6+ messages in thread
From: Fleischer, Julie N @ 2002-12-12 19:59 UTC (permalink / raw)
  To: 'jim.houston@attbi.com', Marco d'Itri, linux-kernel

Jim Houston wrote:
> Marco d'Itri wrote:
> > nanosleep fails after being interrupted:
> >
> > [...]
> > nanosleep({1, 0},
> > [1]+  Stopped                 strace tail -f /var/log/uucp/Log
> > md@wonderland:~$ fg
> > strace tail -f /var/log/uucp/Log
> >  <unfinished ...>
> > --- SIGCONT (Continued) ---
> > <... nanosleep resumed> 0)              = -1 ENOSYS 
> (Function not implemented)>
> >
> > This can be reliably reproduced.                            
> I was able to reproduce this issue.  It happens on all the
> kernels I tried including a stock Redhat kernel.  This is just 
> an idiosyncrasy of strace. In this case both the strace and
> tail are sent a SIGTSTP when they are put into the background.
> Its not suprising that this might confuse strace.

I'm not sure this helps or is irrelevant (or already known), but I did try a
similar thing on a 2.5.50 kernel with the high-res-timers patches (all 4),
and it worked.  That same test did fail (as described above) in 2.5.51 (no
patches).

I have a feeling I may be missing something as I think (?) similar things
were discussed in the "[PATCH] compatibility syscall layer (lets try again)"
thread, but I know I didn't keep up.  I also may have misinterpreted the
original report.

In my test case, I set up a nanosleep to sleep for 5 seconds while I
interrupted it after 1 second with SIGSTOP and then SIGCONT.
On 2.5.50 w/HRT, it output:
  nanosleep() returned success [i.e., exitted w/no errors]
  Start 1039720777 sec; End 1039720782 sec
  Test PASSED

On 2.5.51 w/no HRT, it output:
  nanosleep() did not return success [i.e., exitted w/-1 and set errno]
  Start 1039721052 sec; End 1039721053 sec
  nanosleep() did not sleep long enough

>From strace of the 2.5.51 run:
nanosleep({1, 0}, {1, 0})               = 0
kill(1233, SIGSTOP)                     = 0
--- SIGCHLD (Child exited) ---
kill(1233, SIGCONTnanosleep() did not return success
)                     = 0
--- SIGCHLD (Child exited) ---
wait4(-1, [WIFEXITED(s) && WEXITSTATUS(s) == 1], 0, NULL) = 1233
gettimeofday({1039721104, 153936}, NULL) = 0

The test case was a variant of nanosleep/3-2.c on posixtest.sf.net.

- Julie
**These views are not necessarily those of my employer.**

^ permalink raw reply	[flat|nested] 6+ messages in thread
* Re: 2.5.51 nanosleep fails
@ 2002-12-11 21:05 Jim Houston
  2002-12-12 20:57 ` Marco d'Itri
  0 siblings, 1 reply; 6+ messages in thread
From: Jim Houston @ 2002-12-11 21:05 UTC (permalink / raw)
  To: Marco d'Itri, linux-kernel

Marco d'Itri wrote:
> nanosleep fails after being interrupted:
>
> [...]
> nanosleep({1, 0},
> [1]+  Stopped                 strace tail -f /var/log/uucp/Log
> md@wonderland:~$ fg
> strace tail -f /var/log/uucp/Log
>  <unfinished ...>
> --- SIGCONT (Continued) ---
> <... nanosleep resumed> 0)              = -1 ENOSYS (Function not implemented)>
>
> This can be reliably reproduced.                                              

Hi Marco, Everyone,

I was able to reproduce this issue.  It happens on all the
kernels I tried including a stock Redhat kernel.  This is just 
an idiosyncrasy of strace. In this case both the strace and
tail are sent a SIGTSTP when they are put into the background.
Its not suprising that this might confuse strace.

I also tried this with a program which just does a long
nanosleep.  Strace still shows a return of ENOSYS but the
program actually gets the correct EINTR.

Jim Houston - Concurrent Computer Corp.

^ permalink raw reply	[flat|nested] 6+ messages in thread
* 2.5.51 nanosleep fails
@ 2002-12-10 23:57 Marco d'Itri
  0 siblings, 0 replies; 6+ messages in thread
From: Marco d'Itri @ 2002-12-10 23:57 UTC (permalink / raw)
  To: linux-kernel

[-- Attachment #1: Type: text/plain, Size: 725 bytes --]

nanosleep fails after being interrupted:

[...]
fstat64(3, {st_mode=S_IFREG|0644, st_size=5444, ...}) = 0
gettimeofday({1039564416, 703895}, NULL) = 0
nanosleep({1, 0}, NULL)                 = 0
fstat64(3, {st_mode=S_IFREG|0644, st_size=5444, ...}) = 0
gettimeofday({1039564417, 777452}, NULL) = 0
nanosleep({1, 0}, 
[1]+  Stopped                 strace tail -f /var/log/uucp/Log
md@wonderland:~$ fg
strace tail -f /var/log/uucp/Log
 <unfinished ...>
--- SIGCONT (Continued) ---
<... nanosleep resumed> 0)              = -1 ENOSYS (Function not implemented)


This can be reliably reproduced.


Linux wonderland 2.5.51 #13 Tue Dec 10 14:15:49 CET 2002 i686 unknown unknown GNU/Linux

-- 
ciao,
Marco

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

^ permalink raw reply	[flat|nested] 6+ messages in thread
end of thread, other threads:[~2002-12-13  1:21 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2002-12-12 22:27 2.5.51 nanosleep fails Jim Houston
2002-12-13  1:29 ` Stephen Rothwell
  -- strict thread matches above, loose matches on Subject: below --
2002-12-12 19:59 Fleischer, Julie N
2002-12-11 21:05 Jim Houston
2002-12-12 20:57 ` Marco d'Itri
2002-12-10 23:57 Marco d'Itri

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox