Re: [Valgrind-users] missing syscall: 252

[Benjamin L. <ben...@re...>]

One correction... the Debian package is found in testing not unstable.

testing/main valgrind 1.9.6-1 [785kB]

On Monday, 2003-06-02 at 10:26:20 AM, Benjamin Lee scribbled:
> Stranger still.
> 
> Anyway, I just thought I'd check... but it seems 1.9.6 has support for
> __NR_exit_group.
> 
> [Mon,  2 Jun 2003 10:19:08 +1000] Home Sweet Home
> [ben@pov:/var/local/ben/a/valgrind-1.9.6]$ find -type f | xargs grep -n __NR_exit_group
> ./coregrind/vg_syscalls.c:470:#     if defined(__NR_exit_group)
> ./coregrind/vg_syscalls.c:471:      case __NR_exit_group:
> ./coregrind/vg_scheduler.c:1472:#               if defined(__NR_exit_group)
> ./coregrind/vg_scheduler.c:1473:                || VG_(threads)[tid].m_eax == __NR_exit_group
> ./coregrind/vg_scheduler.c:1511:#                   if defined(__NR_exit_group)
> ./coregrind/vg_scheduler.c:1512:                    && VG_(threads)[tid].m_eax != __NR_exit_group
> [Mon,  2 Jun 2003 10:19:21 +1000] Home Sweet Home
> [ben@pov:/var/local/ben/a/valgrind-1.9.6]$
> 
> 
> I'd suggest using Debian sid/unstable with dselect to install valgrind
> 1.9.6 rather than bother with compiling your own.
> 
> But I'd say the problem will definitely be mismatched headers.
> 
> 
> 
> 
> 
> Also... I checked my CVS files...
> 
> [Mon,  2 Jun 2003 10:15:20 +1000] Home Sweet Home
> [ben@pov:~/cvsremote/valgrind]$ cvs status ./coregrind/vg_syscalls.c ./coregrind/vg_scheduler.c
> ===================================================================
> File: vg_syscalls.c     Status: Up-to-date
> 
>    Working revision:    1.34
>    Repository revision: 1.34    /cvsroot/valgrind/valgrind/coregrind/vg_syscalls.c,v
>    Sticky Tag:          (none)
>    Sticky Date:         (none)
>    Sticky Options:      (none)
> 
> ===================================================================
> File: vg_scheduler.c    Status: Up-to-date
> 
>    Working revision:    1.126
>    Repository revision: 1.126   /cvsroot/valgrind/valgrind/coregrind/vg_scheduler.c,v
>    Sticky Tag:          (none)
>    Sticky Date:         (none)
>    Sticky Options:      (none)
> 
> [Mon,  2 Jun 2003 10:15:39 +1000] Home Sweet Home
> [ben@pov:~/cvsremote/valgrind]$
> 
> 
> [Mon,  2 Jun 2003 10:13:29 +1000] Home Sweet Home
> [ben@pov:~/cvsremote/valgrind]$ grep -n 252 /usr/include/asm/unistd.h
> 259:#define __NR_exit_group             252
> [Mon,  2 Jun 2003 10:13:55 +1000] Home Sweet Home
> [ben@pov:~/cvsremote/valgrind]$ find -type f | xargs grep -n __NR_exit_group
> ./coregrind/vg_scheduler.c:1479:#               if defined(__NR_exit_group)
> ./coregrind/vg_scheduler.c:1480:                || VG_(threads)[tid].m_eax == __NR_exit_group
> ./coregrind/vg_scheduler.c:1518:#                   if defined(__NR_exit_group)
> ./coregrind/vg_scheduler.c:1519:                    && VG_(threads)[tid].m_eax != __NR_exit_group
> ./coregrind/vg_syscalls.c:470:#     if defined(__NR_exit_group)
> ./coregrind/vg_syscalls.c:471:      case __NR_exit_group:
> [Mon,  2 Jun 2003 10:14:17 +1000] Home Sweet Home
> [ben@pov:~/cvsremote/valgrind]$
> 
> 
> On Monday, 2003-06-02 at 06:44:23 AM, Michael Vanier scribbled:
> > 
> > Hmmm... just downloaded and compiled the CVS version of valgrind.  I get
> > the same bug wrt syscall 252.  I looked in vg_syscalls.c and there is no
> > mention of system call 252.
> > 
> > Mike
> > 
> > > Date: Mon, 2 Jun 2003 06:07:49 +1000
> > > From: Benjamin Lee <ben...@re...>
> > > 
> > > Another guess... maybe the compile was using the wrong kernel headers?
> > > 
> > > On Monday, 2003-06-02 at 05:26:49 AM, Michael Vanier scribbled:
> > > > 
> > > > Curious... I *tried* to do that but gave up after many attempts and went
> > > > back to 2.4.17 (at least I thought so...).  Anyway, thanks for the advice.
> > > > 
> > > > Mike
> > > > 
> > > > > From: Benjamin Lee <ben...@re...>
> > > > > Date: Mon, 2 Jun 2003 05:03:52 +1000
> > > > > 
> > > > > Hi,
> > > > > 
> > > > > The new syscall __NR_exit_group / 252 was added to kernel 2.4.20. I'd guess
> > > > > you probably just upgraded your Debian Linux kernel from 2.4.18 to 2.4.20
> > > > > using sid/unstable.
> > > > > 
> > > > > Roll yourself a valgrind from CVS and you'll be fine.
> > > > > 
> > > > > On Sunday, 2003-06-01 at 07:42:43 PM, Michael Vanier scribbled:
> > > > > > 
> > > > > > I just upgraded my valgrind installation to 1.9.6 from 1.9.5, and I
> > > > > > immediately ran into a major problem.  I can't run valgrind any more
> > > > > > because of a missing system call.  To add insult to injury, the number of
> > > > > > the call (252) is not listed in /usr/include/asm/unistd.h.  Recompiling
> > > > > > 1.9.5 fails as well with the same message, so presumably it's happening
> > > > > > because the version of gcc/g++ I'm using has changed (I'm running gcc/g++
> > > > > > 3.2.3 on a Debian Linux system).  Any advice/hints?
> > > > > > 
> > > > > > Mike
> > > > > > 

-- 
Benjamin Lee
Melbourne, Australia             "Always real."    http://realthought.net/

__________________________________________________________________________
	"I assure you the thought never even crossed my mind, lord."
	"Indeed?  Then if I were you I'd sue my face for slander."
		-- Terry Pratchett, "The Colour of Magic"