NetBSD Problem Report #26687

Received: (qmail 9204 invoked by uid 605); 16 Aug 2004 20:55:40 -0000
Message-Id: <20040816203153.EC52E54B23D@dogfish-head.cs.stevens-tech.edu>
Date: Mon, 16 Aug 2004 16:31:53 -0400 (EDT)
From: jschauma@netmeister.org
Sender: gnats-bugs-owner@NetBSD.org
Reply-To: jschauma@netmeister.org
To: gnats-bugs@gnats.NetBSD.org
Subject: linux emulation / matlab stopped working
X-Send-Pr-Version: 3.95

>Number:         26687
>Category:       port-i386
>Synopsis:       matlab stopped working - linux emulation?
>Confidential:   no
>Severity:       serious
>Priority:       high
>Responsible:    jdolecek
>State:          closed
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Mon Aug 16 20:56:00 +0000 2004
>Closed-Date:    Tue Sep 28 09:03:04 +0000 2004
>Last-Modified:  Fri Oct 01 03:47:00 +0000 2004
>Originator:     Jan Schaumann
>Release:        NetBSD 2.0_BETA
>Organization:

>Environment:


System: NetBSD dogfish-head.cs.stevens-tech.edu 2.0_BETA NetBSD 2.0_BETA (BOCK) #17: Mon Aug 16 13:34:05 EDT 2004 jschauma@doppelbock.hpcf.cs.stevens-tech.edu:/share/src/sys/arch/i386/compile/obj/BOCK i386
Architecture: i386
Machine: i386
>Description:
	matlab used to work under NetBSD 1.6U and newer, including the full
	java desktop.  Some time between 1.6U and 2.0_BETA, the linux emulation
	seems to have changed in a way that now prevents matlab from
	functioning in full desktop mode.

	This is a serious regression and we should not ship NetBSD 2.0 unless
	this has been fixed.

	See threads on port-i386 and current-users:
		http://mail-index.netbsd.org/port-i386/2004/05/14/0003.html
		http://mail-index.netbsd.org/port-i386/2004/07/20/0014.html
		http://mail-index.netbsd.org/current-users/2004/07/21/0002.html
		...

>How-To-Repeat:
	run matlab version 6.5 (R13) and notice that it either hangs the machine
	or segfaults.  R13 works if passed the '-nojvm' switch (ie in
	console only mode).

	run matlab version 7 (R14) and notice that it won't work at all.

	Notice that Sun's JDK 1.4 *does* seem to work fine with other
	applications, however.  (For example, eclipse works fine.)
>Fix:

>Release-Note:
>Audit-Trail:
Responsible-Changed-From-To: port-i386-maintainer->jdolecek 
Responsible-Changed-By: jdolecek 
Responsible-Changed-When: Sat Aug 21 01:08:12 UTC 2004 
Responsible-Changed-Why:  
I'm looking into this. 

From: jdolecek@NetBSD.org (Jaromir Dolecek)
To: jschauma@netmeister.org
Cc: gnats-bugs@gnats.NetBSD.org
Subject: Re: port-i386/26687: linux emulation / matlab stopped working
Date: Sat, 21 Aug 2004 03:07:58 +0200 (CEST)

 jschauma@netmeister.org wrote:
 > System: NetBSD dogfish-head.cs.stevens-tech.edu 2.0_BETA NetBSD 2.0_BETA (BOCK) #17: Mon Aug 16 13:34:05 EDT 2004 jschauma@doppelbock.hpcf.cs.stevens-tech.edu:/share/src/sys/arch/i386/compile/obj/BOCK i386
 >
 > 	matlab used to work under NetBSD 1.6U and newer, including the full
 > 	java desktop.  Some time between 1.6U and 2.0_BETA, the linux emulation
 > 	seems to have changed in a way that now prevents matlab from
 > 	functioning in full desktop mode.
 > 
 > 	This is a serious regression and we should not ship NetBSD 2.0 unless
 > 	this has been fixed.
 > 
 > 	See threads on port-i386 and current-users:
 > 		http://mail-index.netbsd.org/port-i386/2004/05/14/0003.html
 > 		http://mail-index.netbsd.org/port-i386/2004/07/20/0014.html
 > 		http://mail-index.netbsd.org/current-users/2004/07/21/0002.html
 > 		...

 Is this with the up-to-date 2.0_BETA kernel, including linux_misc.c
 rev. 1.122.2.1?

 Jaromir

 > >How-To-Repeat:
 > 	run matlab version 6.5 (R13) and notice that it either hangs the machine
 > 	or segfaults.  R13 works if passed the '-nojvm' switch (ie in
 > 	console only mode).
 > 
 > 	run matlab version 7 (R14) and notice that it won't work at all.
 > 
 > 	Notice that Sun's JDK 1.4 *does* seem to work fine with other
 > 	applications, however.  (For example, eclipse works fine.)
 -- 
 Jaromir Dolecek <jdolecek@NetBSD.org>            http://www.NetBSD.cz/
 -=- We should be mindful of the potential goal, but as the Buddhist -=-
 -=- masters say, ``You may notice during meditation that you        -=-
 -=- sometimes levitate or glow.   Do not let this distract you.''   -=-

From: Jan Schaumann <jschauma@netmeister.org>
To: Jaromir Dolecek <jdolecek@NetBSD.org>
Cc: gnats-bugs@gnats.NetBSD.org
Subject: Re: port-i386/26687: linux emulation / matlab stopped working
Date: Fri, 20 Aug 2004 22:06:07 -0400

 --tKW2IUtsqtDRztdT
 Content-Type: text/plain; charset=us-ascii
 Content-Disposition: inline
 Content-Transfer-Encoding: quoted-printable

 Jaromir Dolecek <jdolecek@NetBSD.org> wrote:
 > jschauma@netmeister.org wrote:
 > > System: NetBSD dogfish-head.cs.stevens-tech.edu 2.0_BETA NetBSD 2.0_BET=
 A (BOCK) #17: Mon Aug 16 13:34:05 EDT 2004 jschauma@doppelbock.hpcf.cs.stev=
 ens-tech.edu:/share/src/sys/arch/i386/compile/obj/BOCK i386
 > >
 > > 	matlab used to work under NetBSD 1.6U and newer, including the full
 > > 	java desktop.  Some time between 1.6U and 2.0_BETA, the linux emulation
 > > 	seems to have changed in a way that now prevents matlab from
 > > 	functioning in full desktop mode.
 > >=20
 > > 	This is a serious regression and we should not ship NetBSD 2.0 unless
 > > 	this has been fixed.
 > >=20
 > > 	See threads on port-i386 and current-users:
 > > 		http://mail-index.netbsd.org/port-i386/2004/05/14/0003.html
 > > 		http://mail-index.netbsd.org/port-i386/2004/07/20/0014.html
 > > 		http://mail-index.netbsd.org/current-users/2004/07/21/0002.html
 > > 		...
 >=20
 > Is this with the up-to-date 2.0_BETA kernel, including linux_misc.c
 > rev. 1.122.2.1?

 Yes.  I have linux_misc.c,v 1.122.2.2.

 -Jan

 --=20
 The Internet interprets censorship as damage and routes around it.
 	-- John Gilmore

 --tKW2IUtsqtDRztdT
 Content-Type: application/pgp-signature
 Content-Disposition: inline

 -----BEGIN PGP SIGNATURE-----
 Version: GnuPG v1.2.3 (NetBSD)

 iD8DBQFBJq4PfFtkr68iakwRAkK8AJ9JlL27qRsY/8e6CGA5OtXHS6cWbgCfTKqj
 awwJ6JdzLPf/Deap+0FYM1s=
 =HBkt
 -----END PGP SIGNATURE-----

 --tKW2IUtsqtDRztdT--

From: jdolecek@NetBSD.org (Jaromir Dolecek)
To: Jan Schaumann <jschauma@netmeister.org>
Cc: gnats-bugs@gnats.NetBSD.org
Subject: Re: port-i386/26687: linux emulation / matlab stopped working
Date: Sat, 21 Aug 2004 17:17:46 +0200 (CEST)

 Jan Schaumann wrote:
 > Yes.  I have linux_misc.c,v 1.122.2.2.

 Can you please do 'ktrace -i' of the (failing) matlab
 and make the result kdump ('kdump -e linux' to ensure the
 syscalls are mapped properly) available somewhere?
 It might give some clue of what's happening.

 Thanks.

 Jaromir
 -- 
 Jaromir Dolecek <jdolecek@NetBSD.org>            http://www.NetBSD.cz/
 -=- We should be mindful of the potential goal, but as the Buddhist -=-
 -=- masters say, ``You may notice during meditation that you        -=-
 -=- sometimes levitate or glow.   Do not let this distract you.''   -=-

From: Jan Schaumann <jschauma@netmeister.org>
To: Jaromir Dolecek <jdolecek@NetBSD.org>
Cc: gnats-bugs@gnats.NetBSD.org
Subject: Re: port-i386/26687: linux emulation / matlab stopped working
Date: Tue, 24 Aug 2004 23:22:00 -0400

 --yLVHuoLXiP9kZBkt
 Content-Type: text/plain; charset=us-ascii
 Content-Disposition: inline
 Content-Transfer-Encoding: quoted-printable

 Jaromir Dolecek <jdolecek@NetBSD.org> wrote:
 > Jan Schaumann wrote:
 > > Yes.  I have linux_misc.c,v 1.122.2.2.
 >=20
 > Can you please do 'ktrace -i' of the (failing) matlab
 > and make the result kdump ('kdump -e linux' to ensure the
 > syscalls are mapped properly) available somewhere?

 http://www.cs.stevens.edu/~jschauma/netbsd/matlab.kernel.ident
 http://www.cs.stevens.edu/~jschauma/netbsd/ktrace.out.gz
 http://www.cs.stevens.edu/~jschauma/netbsd/dump.gz

 -Jan

 --=20
 I'm not even supposed to be here today!

 --yLVHuoLXiP9kZBkt
 Content-Type: application/pgp-signature
 Content-Disposition: inline

 -----BEGIN PGP SIGNATURE-----
 Version: GnuPG v1.2.3 (NetBSD)

 iD8DBQFBLAXYfFtkr68iakwRAnEKAJ9wCONEPH+NDDib9w+2MRamrMqu2gCggW6r
 UJHjeNJ4k7FHyMvEb5Pn2ks=
 =e+07
 -----END PGP SIGNATURE-----

 --yLVHuoLXiP9kZBkt--

From: jdolecek@NetBSD.org (Jaromir Dolecek)
To: Jan Schaumann <jschauma@netmeister.org>
Cc: gnats-bugs@gnats.NetBSD.org
Subject: Re: port-i386/26687: linux emulation / matlab stopped working
Date: Sat, 28 Aug 2004 00:34:46 +0200 (CEST)

 --ELM1093646086-7461-0_
 Content-Transfer-Encoding: 7bit
 Content-Type: text/plain; charset=US-ASCII

 Jan Schaumann wrote:
 -- Start of PGP signed section.
 > Jaromir Dolecek <jdolecek@NetBSD.org> wrote:
 > > Jan Schaumann wrote:
 > > > Yes.  I have linux_misc.c,v 1.122.2.2.
 > > 
 > > Can you please do 'ktrace -i' of the (failing) matlab
 > > and make the result kdump ('kdump -e linux' to ensure the
 > > syscalls are mapped properly) available somewhere?
 > 
 > http://www.cs.stevens.edu/~jschauma/netbsd/matlab.kernel.ident
 > http://www.cs.stevens.edu/~jschauma/netbsd/ktrace.out.gz
 > http://www.cs.stevens.edu/~jschauma/netbsd/dump.gz

 This might be same problem I debugged with Thomas Klausner. Can
 you try the attached patch and let me know if it helped?

 Jaromir
 -- 
 Jaromir Dolecek <jdolecek@NetBSD.org>            http://www.NetBSD.cz/
 -=- We should be mindful of the potential goal, but as the Buddhist -=-
 -=- masters say, ``You may notice during meditation that you        -=-
 -=- sometimes levitate or glow.   Do not let this distract you.''   -=-

 --ELM1093646086-7461-0_
 Content-Transfer-Encoding: 7bit
 Content-Type: text/plain; charset=ISO-8859-2
 Content-Disposition: attachment; filename=linux.diff
 Content-Description: 

 Index: linux_misc.c
 ===================================================================
 RCS file: /cvsroot/src/sys/compat/linux/common/linux_misc.c,v
 retrieving revision 1.127
 diff -u -p -r1.127 linux_misc.c
 --- linux_misc.c	8 Aug 2004 09:40:50 -0000	1.127
 +++ linux_misc.c	26 Aug 2004 22:18:45 -0000
 @@ -454,14 +454,28 @@ linux_sys_mmap(l, v, retval)
  		syscallarg(linux_off_t) offset;
  	} */ *uap = v;
  	struct sys_mmap_args cma;
 +	int error;

  	if (SCARG(uap, offset) & PAGE_MASK)
  		return EINVAL;

 +	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN && SCARG(uap, addr)) {
 +		SCARG(uap, addr) -= SCARG(uap, len);
 +		SCARG(uap, len) *= 2;
 +	}
 +
  	linux_to_bsd_mmap_args(&cma, uap);
  	SCARG(&cma, pos) = (off_t)SCARG(uap, offset);

 -	return sys_mmap(l, &cma, retval);
 +	error = sys_mmap(l, &cma, retval);
 +	if (error)
 +		return (error);
 +
 +	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN && SCARG(uap, addr)) {
 +		retval[0] += SCARG(uap, len) / 2;
 +	}
 +
 +	return (0);
  }

  /*

 --ELM1093646086-7461-0_--
State-Changed-From-To: open->feedback 
State-Changed-By: jdolecek 
State-Changed-When: Sun Aug 29 09:21:49 UTC 2004 
State-Changed-Why:  
Offered patch, waiting for result. 

From: Jaromir Dolecek <jdolecek@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc:  
Subject: pr/26687 CVS commit: src/sys/compat/linux/common
Date: Sun, 29 Aug 2004 14:08:06 +0000 (UTC)

 Module Name:	src
 Committed By:	jdolecek
 Date:		Sun Aug 29 14:08:06 UTC 2004

 Modified Files:
 	src/sys/compat/linux/common: linux_misc.c

 Log Message:
 handle mmap() request with MAP_GROWSDOWN flag (request for stack-like
 grow-down auto extend segment) by allocating segment sized at
 current stack size limit, and offsetting requested/returned address
 as required

 due to how normal virtual memory management work, allocating the
 full sized stack memory segment up-front actually requires exactly same
 amount of VA space and physical memory as the Linux 'grow' scheme and the
 'grow' scheme is quite a lot more difficult to use in applications correctly,
 so it's not very apparent why Linux introduced this feature at all

 this fixes Thomas Klausner's Heroes3 crash, and might also
 fix PR 26687 by Jan Schaumann


 To generate a diff of this commit:
 cvs rdiff -r1.127 -r1.128 src/sys/compat/linux/common/linux_misc.c

 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.


From: Jan Schaumann <jschauma@netmeister.org>
To: Jaromir Dolecek <jdolecek@NetBSD.org>
Cc: gnats-bugs@gnats.NetBSD.org
Subject: Re: port-i386/26687: linux emulation / matlab stopped working
Date: Mon, 30 Aug 2004 13:02:22 -0400

 --xHFwDpU9dbj6ez1V
 Content-Type: text/plain; charset=us-ascii
 Content-Disposition: inline
 Content-Transfer-Encoding: quoted-printable

 Jaromir Dolecek <jdolecek@NetBSD.org> wrote:
 > Jan Schaumann wrote:
 =20
 > > http://www.cs.stevens.edu/~jschauma/netbsd/matlab.kernel.ident
 > > http://www.cs.stevens.edu/~jschauma/netbsd/ktrace.out.gz
 > > http://www.cs.stevens.edu/~jschauma/netbsd/dump.gz
 >=20
 > This might be same problem I debugged with Thomas Klausner. Can
 > you try the attached patch and let me know if it helped?

 Sorry, no luck.  Same behaviour.  Do you want a new ktrace?

 -Jan

 --=20
 'I have reached an age where my main purpose is not to receive
 messages.' --- Umberto Eco, quoted in the New Yorker

 --xHFwDpU9dbj6ez1V
 Content-Type: application/pgp-signature
 Content-Disposition: inline

 -----BEGIN PGP SIGNATURE-----
 Version: GnuPG v1.2.3 (NetBSD)

 iD8DBQFBM12efFtkr68iakwRAr5ZAJ9kM7PVQsGmU+OqlAEfhZctjOgJyQCguQtk
 TteWwvMurckMDrtT7HOqhR4=
 =Gv+8
 -----END PGP SIGNATURE-----

 --xHFwDpU9dbj6ez1V--

From: Jan Schaumann <jschauma@netmeister.org>
To: Jaromir Dolecek <jdolecek@NetBSD.org>
Cc: gnats-bugs@NetBSD.org
Subject: Re: port-i386/26687: linux emulation / matlab stopped working
Date: Wed, 1 Sep 2004 09:41:42 -0400

 --GpGaEY17fSl8rd50
 Content-Type: text/plain; charset=us-ascii
 Content-Disposition: inline
 Content-Transfer-Encoding: quoted-printable

 Jaromir Dolecek <jdolecek@NetBSD.org> wrote:
 > Jan Schaumann wrote:
 > > Well, that specific JVM _used_ to work just fine.  As my PR states, I
 > > was using matlab in that version with that JVM for several months.
 > > Until I upgraded to 2.0_BETA, in fact, which would lead me to conclude
 > > that there was indeed some regression in the Linux emulation.
 >=20
 > Hmm, true.
 >=20
 > Can you please make a ktrace -i & kdump -e linux with the old
 > (working) kernel and the new (non-working) kernel. Can you also
 > please check if the old (working) kernel had any option KERNBASE
 > in it's kernel config?

 Hmm, I'll see what I can do, but this may be difficult:  All my machines
 have been updated to 2.0_BETA a long time ago (I only noticed the matlab
 regression a few weeks ago).  I'll try to track down the last working
 version.

 -Jan

 --=20
 Time is an illusion, lunchtime doubly so.

 --GpGaEY17fSl8rd50
 Content-Type: application/pgp-signature
 Content-Disposition: inline

 -----BEGIN PGP SIGNATURE-----
 Version: GnuPG v1.2.3 (NetBSD)

 iD8DBQFBNdGWfFtkr68iakwRAjHlAJ9Vs3/TRdwP9a0bvHNkQjeEaVTTEQCfd1Wr
 c28PUZCXKOBVi8g9tALugV4=
 =J51y
 -----END PGP SIGNATURE-----

 --GpGaEY17fSl8rd50--

From: Matthias Scheler <tron@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc:  
Subject: pr/26687 CVS commit: [netbsd-2-0] src/sys/compat/linux/common
Date: Sun,  5 Sep 2004 10:49:57 +0000 (UTC)

 Module Name:	src
 Committed By:	tron
 Date:		Sun Sep  5 10:49:57 UTC 2004

 Modified Files:
 	src/sys/compat/linux/common [netbsd-2-0]: linux_misc.c

 Log Message:
 Pull up revision 1.128 (requested by jdolecek in ticket #810):
 handle mmap() request with MAP_GROWSDOWN flag (request for stack-like
 grow-down auto extend segment) by allocating segment sized at
 current stack size limit, and offsetting requested/returned address
 as required
 due to how normal virtual memory management work, allocating the
 full sized stack memory segment up-front actually requires exactly same
 amount of VA space and physical memory as the Linux 'grow' scheme and the
 'grow' scheme is quite a lot more difficult to use in applications correctly,
 so it's not very apparent why Linux introduced this feature at all
 this fixes Thomas Klausner's Heroes3 crash, and might also
 fix PR 26687 by Jan Schaumann


 To generate a diff of this commit:
 cvs rdiff -r1.122.2.2 -r1.122.2.3 src/sys/compat/linux/common/linux_misc.c

 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.


From: jdolecek@NetBSD.org (Jaromir Dolecek)
To: Jan Schaumann <jschauma@netmeister.org>
Cc: gnats-bugs@netbsd.org
Subject: Re: port-i386/26687: linux emulation / matlab stopped working
Date: Wed, 15 Sep 2004 07:36:07 +0200 (CEST)

 Hi,

 The only difference I can spot is SIGSEGV handling. It causes java
 crash with 2.0, but is handled normally with 1.6ZH. AFAIK such
 mechanism is used for memory garbage collection.

 Can you please find kernel version where it breaks? The window
 between 1.6ZH and 2-0 branch is about two months, so shouldn't be
 too hard.

 Kernel version was bumped to 1.6ZH on '2004/01/10 14:46:24',
 2.0_BETA was created '2004/03/29 05:21:33'.

 First of all, I'd check whether kernel build from checkout with tag
 netbsd-2-0-base works (to rule out any later pullup broke it),
 then check time-based checkout at date just before each kernel
 version bump between 1.6ZH and 1.6ZL.

 I'd do this myself, but I don't have access to Matlab...

 Thanks,

 Jaromir
 -- 
 Jaromir Dolecek <jdolecek@NetBSD.org>            http://www.NetBSD.cz/
 -=- We should be mindful of the potential goal, but as the Buddhist -=-
 -=- masters say, ``You may notice during meditation that you        -=-
 -=- sometimes levitate or glow.   Do not let this distract you.''   -=-

From: Mark Davies <mark@mcs.vuw.ac.nz>
To: Jan Schaumann <jschauma@netmeister.org>
Cc: Jaromir Dolecek <jdolecek@netbsd.org>, gnats-bugs@netbsd.org
Subject: Re: port-i386/26687: linux emulation / matlab stopped working
Date: Thu, 16 Sep 2004 18:03:48 +1200

 On Thursday 16 September 2004 06:00, you wrote:
 > FYI.  If you can help out trying to nail down the kernel versions that
 > worked last, that would be reat.

 OK this is what I've found out so far.  In each of the following the userland 
 is 1.6ZH

 Kernel from tag netbsd-2-0-base   -  Matlab works
 Kernel from date 2004.04.20.12.00.00 (2.0C) - Matlab segfaults after splash 
 screen
 Kernel from date 2004.04.05.12.00.00 (2.0C) - Matlab segfaults after splash 
 screen

 I'm trying a 2004.03.31.12.00.00 (2.0B) at the moment but I need to go home 
 and I'm not sure how easily I'll be able to get at this test box from there.

 cheers
 mark

From: Mark Davies <mark@mcs.vuw.ac.nz>
To: Jan Schaumann <jschauma@netmeister.org>
Cc: Jaromir Dolecek <jdolecek@netbsd.org>, gnats-bugs@netbsd.org
Subject: Re: [jdolecek@NetBSD.org: Re: port-i386/26687: linux emulation / matlab stopped working]
Date: Thu, 16 Sep 2004 18:30:14 +1200

 I wrote:
 > Kernel from date 2004.04.05.12.00.00 (2.0C) - Matlab segfaults after splash 
 > screen

 > I'm trying a 2004.03.31.12.00.00 (2.0B) at the moment but I need to go home 
 > and I'm not sure how easily I'll be able to get at this test box from there.

 So I didn't quite get home yet.  The 2004.03.31.12.00.00 kernel works so the 
 problem is somewhere in those 5 days.

 cheers
 mark

From: Mark Davies <mark@mcs.vuw.ac.nz>
To: Jan Schaumann <jschauma@netmeister.org>
Cc: Jaromir Dolecek <jdolecek@netbsd.org>, gnats-bugs@netbsd.org
Subject: Re: port-i386/26687: linux emulation / matlab stopped working
Date: Thu, 16 Sep 2004 22:35:24 +1200

 On Thursday 16 September 2004 18:30, Mark Davies wrote:
 > So I didn't quite get home yet.  The 2004.03.31.12.00.00 kernel works so
 > the problem is somewhere in those 5 days.

 2004.04.02.12.00.00 - matlab segfaults, so its down to 2 days.

 mark

From: Mark Davies <mark@mcs.vuw.ac.nz>
To: Jan Schaumann <jschauma@netmeister.org>
Cc: Jaromir Dolecek <jdolecek@netbsd.org>, gnats-bugs@netbsd.org
Subject: Re: port-i386/26687: linux emulation / matlab stopped working
Date: Thu, 16 Sep 2004 23:55:57 +1200

 OK, I've tracked it down.
 Prior to the below change matlab works, after it it Seg faults.

   From: Matt Thomas <matt@netbsd.org>
   List: source-changes
   Date: 04/01/2004 16:56:45 
   Module Name:    src
   Committed By:   matt
   Date:           Thu Apr  1 16:56:45 UTC 2004

   Modified Files:
         src/sys/kern: kern_sig.c

   Log Message:
   If a signal is the result of trap, only invoke a supplied handler if it's
   not blocked.  Otherwise (it if it blocked or the hanlder is set to SIG_IGN)
   reset the signal back to its default settings so that a coredump can be
   generated.

   To generate a diff of this commit:
   cvs rdiff -r1.189 -r1.190 src/sys/kern/kern_sig.c

 Now how to fix?

 cheers
 mark

From: jdolecek@NetBSD.org (Jaromir Dolecek)
To: Mark Davies <mark@mcs.vuw.ac.nz>
Cc: Jan Schaumann <jschauma@netmeister.org>, gnats-bugs@netbsd.org
Subject: Re: port-i386/26687: linux emulation / matlab stopped working
Date: Thu, 16 Sep 2004 20:33:46 +0200 (CEST)

 Mark Davies wrote:
 > OK, I've tracked it down.
 > Prior to the below change matlab works, after it it Seg faults.

 Thanks a lot! I will start discussion on tech-kern@ to resolve this.

 Jaromir

 >   From: Matt Thomas <matt@netbsd.org>
 >   List: source-changes
 >   Date: 04/01/2004 16:56:45 
 >   Module Name:    src
 >   Committed By:   matt
 >   Date:           Thu Apr  1 16:56:45 UTC 2004
 > 
 >   Modified Files:
 >         src/sys/kern: kern_sig.c
 > 
 >   Log Message:
 >   If a signal is the result of trap, only invoke a supplied handler if it's
 >   not blocked.  Otherwise (it if it blocked or the hanlder is set to SIG_IGN)
 >   reset the signal back to its default settings so that a coredump can be
 >   generated.
 > 
 >   To generate a diff of this commit:
 >   cvs rdiff -r1.189 -r1.190 src/sys/kern/kern_sig.c
 > 
 > Now how to fix?
 > 
 > cheers
 > mark
 > 

 -- 
 Jaromir Dolecek <jdolecek@NetBSD.org>            http://www.NetBSD.cz/
 -=- We should be mindful of the potential goal, but as the Buddhist -=-
 -=- masters say, ``You may notice during meditation that you        -=-
 -=- sometimes levitate or glow.   Do not let this distract you.''   -=-

From: jdolecek@NetBSD.org (Jaromir Dolecek)
To: mark@mcs.vuw.ac.nz
Cc: jschauma@netmeister.org, gnats-bugs@netbsd.org
Subject: Re: port-i386/26687: linux emulation / matlab stopped working
Date: Mon, 27 Sep 2004 21:12:48 +0200 (CEST)

 Hi,

 please try following patch if it makes any difference. It basically
 keeps what current code does, but untangles the code and
 fixes handling of the traps not hitting the reset condition.

 Index: kern_sig.c
 ===================================================================
 RCS file: /cvsroot/src/sys/kern/kern_sig.c,v
 retrieving revision 1.197
 diff -u -p -r1.197 kern_sig.c
 --- kern_sig.c	8 Jun 2004 19:35:30 -0000	1.197
 +++ kern_sig.c	27 Sep 2004 19:07:22 -0000
 @@ -1035,44 +1035,51 @@ kpsignal2(struct proc *p, const ksiginfo
  	/*
  	 * If proc is traced, always give parent a chance.
  	 */
 -	action = SIG_DFL;
 -	if ((p->p_flag & P_TRACED) == 0) {
 -		if (KSI_TRAP_P(ksi)) {
 -			/*
 -			 * If the signal was the result of a trap, only catch
 -			 * the signal if it isn't masked and there is a
 -			 * non-default non-ignore handler installed for it.
 -			 * Otherwise take the default action.
 -			 */
 -			if (!sigismember(&p->p_sigctx.ps_sigmask, signum) &&
 -			    sigismember(&p->p_sigctx.ps_sigcatch, signum))
 -				action = SIG_CATCH;
 +	if (p->p_flag & P_TRACED) {
 +		action = SIG_DFL;
 +
 +		/*
 +		 * If the process is being traced and the signal is being
 +		 * caught, make sure to save any ksiginfo.
 +		 */
 +		if (sigismember(&p->p_sigctx.ps_sigcatch, signum))
 +			ksiginfo_put(p, ksi);
 +	} else {
 +		/*
 +		 * If the signal was the result of a trap, only catch
 +		 * the signal if it isn't masked and there is a
 +		 * non-default non-ignore handler installed for it.
 +		 * Otherwise take the default action.
 +		 */
 +		if (KSI_TRAP_P(ksi)
 +		    && sigismember(&p->p_sigctx.ps_sigmask, signum)
 +		    && !sigismember(&p->p_sigctx.ps_sigcatch, signum)) {
  			/*
  			 * If we are to take the default action, reset the
  			 * signal back to its defaults.
  			 */
 -			if (action == SIG_DFL) {
 -				sigdelset(&p->p_sigctx.ps_sigignore, signum);
 -				sigdelset(&p->p_sigctx.ps_sigcatch, signum);
 -				sigdelset(&p->p_sigctx.ps_sigmask, signum);
 -				SIGACTION(p, signum).sa_handler = SIG_DFL;
 -			}
 -		} else {
 -			/*
 -			 * If the signal is being ignored,
 -			 * then we forget about it immediately.
 -			 * (Note: we don't set SIGCONT in p_sigctx.ps_sigignore,
 -			 * and if it is set to SIG_IGN,
 -			 * action will be SIG_DFL here.)
 -			 */
 -			if (sigismember(&p->p_sigctx.ps_sigignore, signum))
 -				return;
 -			if (sigismember(&p->p_sigctx.ps_sigmask, signum))
 -				action = SIG_HOLD;
 -			else if (sigismember(&p->p_sigctx.ps_sigcatch, signum))
 -				action = SIG_CATCH;
 +			sigdelset(&p->p_sigctx.ps_sigignore, signum);
 +			sigdelset(&p->p_sigctx.ps_sigcatch, signum);
 +			sigdelset(&p->p_sigctx.ps_sigmask, signum);
 +			SIGACTION(p, signum).sa_handler = SIG_DFL;
  		}
 -		if (action == SIG_DFL) {
 +
 +		/*
 +		 * If the signal is being ignored,
 +		 * then we forget about it immediately.
 +		 * (Note: we don't set SIGCONT in p_sigctx.ps_sigignore,
 +		 * and if it is set to SIG_IGN,
 +		 * action will be SIG_DFL here.)
 +		 */
 +		if (sigismember(&p->p_sigctx.ps_sigignore, signum))
 +			return;
 +		if (sigismember(&p->p_sigctx.ps_sigmask, signum))
 +			action = SIG_HOLD;
 +		else if (sigismember(&p->p_sigctx.ps_sigcatch, signum))
 +			action = SIG_CATCH;
 +		else {
 +			action = SIG_DFL;
 +
  			if (prop & SA_KILL && p->p_nice > NZERO)
  				p->p_nice = NZERO;

 @@ -1085,13 +1092,6 @@ kpsignal2(struct proc *p, const ksiginfo
  			if (prop & SA_TTYSTOP && p->p_pgrp->pg_jobc == 0)
  				return;
  		}
 -	} else {
 -		/*
 -		 * If the process is being traced and the signal is being
 -		 * caught, make sure to save any ksiginfo.
 -		 */
 -		if (sigismember(&p->p_sigctx.ps_sigcatch, signum))
 -			ksiginfo_put(p, ksi);
  	}

  	if (prop & SA_CONT)
 -- 
 Jaromir Dolecek <jdolecek@NetBSD.org>            http://www.NetBSD.cz/
 -=- We should be mindful of the potential goal, but as the Buddhist -=-
 -=- masters say, ``You may notice during meditation that you        -=-
 -=- sometimes levitate or glow.   Do not let this distract you.''   -=-

From: Mark Davies <mark@mcs.vuw.ac.nz>
To: Jaromir Dolecek <jdolecek@netbsd.org>
Cc: jschauma@netmeister.org, gnats-bugs@netbsd.org
Subject: Re: port-i386/26687: linux emulation / matlab stopped working
Date: Tue, 28 Sep 2004 11:31:24 +1200

 On Tuesday 28 September 2004 07:12, Jaromir Dolecek wrote:
 > please try following patch if it makes any difference. It basically
 > keeps what current code does, but untangles the code and
 > fixes handling of the traps not hitting the reset condition.

 That does appear to fix it.  No more seg fault after the splash screen and the 
 gui comes up.

 cheers
 mark

From: Mark Davies <mark@mcs.vuw.ac.nz>
To: Jaromir Dolecek <jdolecek@netbsd.org>
Cc: jschauma@netmeister.org, gnats-bugs@netbsd.org
Subject: Re: port-i386/26687: linux emulation / matlab stopped working
Date: Tue, 28 Sep 2004 13:29:07 +1200

 And with the patch plus the SUSE9.1 libs MatLab 7 works.

 cheers
 mark

From: Jaromir Dolecek <jdolecek@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc:  
Subject: pr/26687 CVS commit: src/sys/kern
Date: Tue, 28 Sep 2004 08:59:20 +0000 (UTC)

 Module Name:	src
 Committed By:	jdolecek
 Date:		Tue Sep 28 08:59:20 UTC 2004

 Modified Files:
 	src/sys/kern: kern_sig.c

 Log Message:
 adjust the change of rev. 1.190 so that trap signals not matching
 the reset condition are processed properly; this fixes PR#26687 by
 Jan Schaumann

 many thanks to Mark Davies, who tracked the offending change down
 and helped test patches

 while here, g/c unused sigtrapmask and rearrange some code to pre-r1.190 form
 for better readability


 To generate a diff of this commit:
 cvs rdiff -r1.197 -r1.198 src/sys/kern/kern_sig.c

 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.

State-Changed-From-To: feedback->closed 
State-Changed-By: jdolecek 
State-Changed-When: Tue Sep 28 09:02:42 UTC 2004 
State-Changed-Why:  
Change committed. Thanks a lot to all involved! 

From: James Chacon <jmc@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc:  
Subject: pr/26687 CVS commit: [netbsd-2-0] src/sys/kern
Date: Fri,  1 Oct 2004 03:46:37 +0000 (UTC)

 Module Name:	src
 Committed By:	jmc
 Date:		Fri Oct  1 03:46:37 UTC 2004

 Modified Files:
 	src/sys/kern [netbsd-2-0]: kern_sig.c

 Log Message:
 Pullup rev 1.198 (requested by jdolecek in ticket #881)

 Adjust the change of rev. 1.190 so that trap signals not matching
 the reset condition are processed properly. PR#26687


 To generate a diff of this commit:
 cvs rdiff -r1.189.2.5 -r1.189.2.6 src/sys/kern/kern_sig.c

 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.

>Unformatted:

NetBSD Home
NetBSD PR Database Search

(Contact us) $NetBSD: query-full-pr,v 1.51 2026/08/10 02:28:17 riastradh Exp $
$NetBSD: gnats_config.sh,v 1.10 2026/05/13 22:00:09 riastradh Exp $
Copyright © 1994-2026 The NetBSD Foundation, Inc. ALL RIGHTS RESERVED.