NetBSD on AMD x86-64 architecture

Reinoud Zandijk | 7 Sep 2011 13:54

trap.c lossage (SIGILL)

Hi folks,

i've comitted the following patch to fix the bad address returned in the
siginfo for the SIGILL signal on i386.

There were also some concerns though about the preservation of the %cs2
register in i386/amd64's trap.c since it is not stored in the trapframe. Would
it be a good idea to review/reaudit trap.c ?

With regards,
Reinoud

Tested and comitted patch for i386:
Index: trap.c
===================================================================
RCS file: /cvsroot/src/sys/arch/i386/i386/trap.c,v
retrieving revision 1.261
retrieving revision 1.262
diff -u -r1.261 -r1.262
--- trap.c	3 Apr 2011 22:29:26 -0000	1.261
+++ trap.c	7 Sep 2011 09:24:55 -0000	1.262
 <at>  <at>  -1,4 +1,4  <at>  <at> 
-/*	$NetBSD: trap.c,v 1.261 2011/04/03 22:29:26 dyoung Exp $	*/
+/*	$NetBSD: trap.c,v 1.262 2011/09/07 09:24:55 reinoud Exp $	*/

 /*-
  * Copyright (c) 1998, 2000, 2005, 2006, 2007, 2008 The NetBSD Foundation,
  * Inc.
 <at>  <at>  -68,7 +68,7  <at>  <at> 
  */

(Continue reading)

Nicolas Joly | 27 Sep 2011 19:10

Picon

Picon

FLT_ROUNDS failures ... gcc/binutils regression ?


Hi,

I tracked a new failure in tests/lib/libc/gen/t_fpsetround where
FLT_ROUNDS now returns an incorrect results in the PIC case. And this
seems related to recent gcc/binutils imports.

njoly <at> lanfeust [current:flt_rounds]> cc --version
cc (GCC) 4.1.3 20080704 prerelease (NetBSD nb2 20081120)
njoly <at> lanfeust [current:flt_rounds]> cc -c -fPIC -DPIC -o flt_rounds.old flt_rounds.S

njoly <at> lanfeust [tmp/flt_rounds]> cc --version
cc (NetBSD nb2 20110806) 4.5.3
njoly <at> lanfeust [tmp/flt_rounds]> cc -c -fPIC -DPIC -o flt_rounds.new flt_rounds.S

njoly <at> lanfeust [tmp/flt_rounds]> objdump -d flt_rounds.old >dump.old
njoly <at> lanfeust [tmp/flt_rounds]> objdump -d flt_rounds.new >dump.new
njoly <at> lanfeust [tmp/flt_rounds]> diff dump.old dump.new
2c2
< flt_rounds.old:     file format elf64-x86-64
---
> flt_rounds.new:     file format elf64-x86-64
18c18
<   1e: 48 8d 0d db ff ff ff    lea    -0x25(%rip),%rcx        # 0 <_map>
---
>   1e: 48 8d 0d 00 00 00 00    lea    0x0(%rip),%rcx        # 25 <__flt_rounds+0x15>

The new code, which does not work, looks suspicious to me ...

--

--

(Continue reading)

David Laight | 27 Sep 2011 19:58

Picon

Re: FLT_ROUNDS failures ... gcc/binutils regression ?

On Tue, Sep 27, 2011 at 07:10:30PM +0200, Nicolas Joly wrote:
> 2c2
> < flt_rounds.old:     file format elf64-x86-64
> ---
> > flt_rounds.new:     file format elf64-x86-64
> 18c18
> <   1e: 48 8d 0d db ff ff ff    lea    -0x25(%rip),%rcx        # 0 <_map>
> ---
> >   1e: 48 8d 0d 00 00 00 00    lea    0x0(%rip),%rcx        # 25 <__flt_rounds+0x15>
> 
> The new code, which does not work, looks suspicious to me ...

It is probably worth looking at the output of 'objdump -r' to see
if there is a relocation against 0x21.

	David

--

-- 
David Laight: david <at> l8s.co.uk

Nicolas Joly | 27 Sep 2011 20:05

Picon

Picon

Re: FLT_ROUNDS failures ... gcc/binutils regression ?

On Tue, Sep 27, 2011 at 06:58:14PM +0100, David Laight wrote:
> On Tue, Sep 27, 2011 at 07:10:30PM +0200, Nicolas Joly wrote:
> > 2c2
> > < flt_rounds.old:     file format elf64-x86-64
> > ---
> > > flt_rounds.new:     file format elf64-x86-64
> > 18c18
> > <   1e: 48 8d 0d db ff ff ff    lea    -0x25(%rip),%rcx        # 0 <_map>
> > ---
> > >   1e: 48 8d 0d 00 00 00 00    lea    0x0(%rip),%rcx        # 25 <__flt_rounds+0x15>
> > 
> > The new code, which does not work, looks suspicious to me ...
> 
> It is probably worth looking at the output of 'objdump -r' to see
> if there is a relocation against 0x21.

njoly <at> lanfeust [tmp/flt_rounds]> objdump -r flt_rounds.new 

flt_rounds.new:     file format elf64-x86-64

RELOCATION RECORDS FOR [.text]:
OFFSET           TYPE              VALUE 
0000000000000021 R_X86_64_GOTPCREL  _map+0xfffffffffffffffc

--

-- 
Nicolas Joly

Projects and Developments in Bioinformatics
Institut Pasteur, Paris.

(Continue reading)

David Laight | 27 Sep 2011 20:27

Picon

Re: FLT_ROUNDS failures ... gcc/binutils regression ?

On Tue, Sep 27, 2011 at 08:05:15PM +0200, Nicolas Joly wrote:
> On Tue, Sep 27, 2011 at 06:58:14PM +0100, David Laight wrote:
> > On Tue, Sep 27, 2011 at 07:10:30PM +0200, Nicolas Joly wrote:
> > > 2c2
> > > < flt_rounds.old:     file format elf64-x86-64
> > > ---
> > > > flt_rounds.new:     file format elf64-x86-64
> > > 18c18
> > > <   1e: 48 8d 0d db ff ff ff    lea    -0x25(%rip),%rcx        # 0 <_map>
> > > ---
> > > >   1e: 48 8d 0d 00 00 00 00    lea    0x0(%rip),%rcx        # 25 <__flt_rounds+0x15>
> > > 
> > > The new code, which does not work, looks suspicious to me ...
> > 
...
> 0000000000000021 R_X86_64_GOTPCREL  _map+0xfffffffffffffffc

That is some kind of pc-relative relocation for _map.

So it looks like _map (or rather its address?) has moved sections (or similar)
so that the offset isn't a compile time constant - or gcc is allowing for that.

Dunno it that is the right relocation type.
Might be worth linking the .o into a static program to see what 'ld'
makes of it.
Also link into a shlib, load, and see what the dynamic linker has done.
Might be a bug in the dynamic linker!

	David

(Continue reading)

Matt Thomas | 27 Sep 2011 23:39

Re: FLT_ROUNDS failures ... gcc/binutils regression ?


On Sep 27, 2011, at 11:27 AM, David Laight wrote:

> On Tue, Sep 27, 2011 at 08:05:15PM +0200, Nicolas Joly wrote:
>> On Tue, Sep 27, 2011 at 06:58:14PM +0100, David Laight wrote:
>>> On Tue, Sep 27, 2011 at 07:10:30PM +0200, Nicolas Joly wrote:
>>>> 2c2
>>>> < flt_rounds.old:     file format elf64-x86-64
>>>> ---
>>>>> flt_rounds.new:     file format elf64-x86-64
>>>> 18c18
>>>> <   1e: 48 8d 0d db ff ff ff    lea    -0x25(%rip),%rcx        # 0 <_map>
>>>> ---
>>>>>  1e: 48 8d 0d 00 00 00 00    lea    0x0(%rip),%rcx        # 25 <__flt_rounds+0x15>
>>>> 
>>>> The new code, which does not work, looks suspicious to me ...
>>> 
> ...
>> 0000000000000021 R_X86_64_GOTPCREL  _map+0xfffffffffffffffc
> 
> That is some kind of pc-relative relocation for _map.
> 
> So it looks like _map (or rather its address?) has moved sections (or similar)
> so that the offset isn't a compile time constant - or gcc is allowing for that.
> 
> Dunno it that is the right relocation type.
> Might be worth linking the .o into a static program to see what 'ld'
> makes of it.
> Also link into a shlib, load, and see what the dynamic linker has done.
> Might be a bug in the dynamic linker!

(Continue reading)

Christos Zoulas | 28 Sep 2011 00:47

Re: FLT_ROUNDS failures ... gcc/binutils regression ?

In article <20110927171030.GA4402 <at> lynche.sis.pasteur.fr>,
Nicolas Joly  <njoly <at> pasteur.fr> wrote:
>
>Hi,
>
>I tracked a new failure in tests/lib/libc/gen/t_fpsetround where
>FLT_ROUNDS now returns an incorrect results in the PIC case. And this
>seems related to recent gcc/binutils imports.
>
>njoly <at> lanfeust [current:flt_rounds]> cc --version
>cc (GCC) 4.1.3 20080704 prerelease (NetBSD nb2 20081120)
>njoly <at> lanfeust [current:flt_rounds]> cc -c -fPIC -DPIC -o flt_rounds.old
>flt_rounds.S
>
>njoly <at> lanfeust [tmp/flt_rounds]> cc --version
>cc (NetBSD nb2 20110806) 4.5.3
>njoly <at> lanfeust [tmp/flt_rounds]> cc -c -fPIC -DPIC -o flt_rounds.new
>flt_rounds.S
>
>njoly <at> lanfeust [tmp/flt_rounds]> objdump -d flt_rounds.old >dump.old
>njoly <at> lanfeust [tmp/flt_rounds]> objdump -d flt_rounds.new >dump.new
>njoly <at> lanfeust [tmp/flt_rounds]> diff dump.old dump.new
>2c2
>< flt_rounds.old:     file format elf64-x86-64
>---
>> flt_rounds.new:     file format elf64-x86-64
>18c18
><   1e: 48 8d 0d db ff ff ff    lea    -0x25(%rip),%rcx        # 0 <_map>
>---
>>   1e: 48 8d 0d 00 00 00 00    lea    0x0(%rip),%rcx        # 25

(Continue reading)

Nicolas Joly | 29 Sep 2011 19:11

Picon

Picon

Re: FLT_ROUNDS failures ... gcc/binutils regression ?

On Tue, Sep 27, 2011 at 10:47:06PM +0000, Christos Zoulas wrote:
> In article <20110927171030.GA4402 <at> lynche.sis.pasteur.fr>,
> Nicolas Joly  <njoly <at> pasteur.fr> wrote:
> >
> >Hi,
> >
> >I tracked a new failure in tests/lib/libc/gen/t_fpsetround where
> >FLT_ROUNDS now returns an incorrect results in the PIC case. And this
> >seems related to recent gcc/binutils imports.
> >
> >njoly <at> lanfeust [current:flt_rounds]> cc --version
> >cc (GCC) 4.1.3 20080704 prerelease (NetBSD nb2 20081120)
> >njoly <at> lanfeust [current:flt_rounds]> cc -c -fPIC -DPIC -o flt_rounds.old
> >flt_rounds.S
> >
> >njoly <at> lanfeust [tmp/flt_rounds]> cc --version
> >cc (NetBSD nb2 20110806) 4.5.3
> >njoly <at> lanfeust [tmp/flt_rounds]> cc -c -fPIC -DPIC -o flt_rounds.new
> >flt_rounds.S
> >
> >njoly <at> lanfeust [tmp/flt_rounds]> objdump -d flt_rounds.old >dump.old
> >njoly <at> lanfeust [tmp/flt_rounds]> objdump -d flt_rounds.new >dump.new
> >njoly <at> lanfeust [tmp/flt_rounds]> diff dump.old dump.new
> >2c2
> >< flt_rounds.old:     file format elf64-x86-64
> >---
> >> flt_rounds.new:     file format elf64-x86-64
> >18c18
> ><   1e: 48 8d 0d db ff ff ff    lea    -0x25(%rip),%rcx        # 0 <_map>
> >---

(Continue reading)

Nicolas Joly | 29 Sep 2011 19:13

Picon

Picon

Re: FLT_ROUNDS failures ... gcc/binutils regression ?

On Tue, Sep 27, 2011 at 02:39:40PM -0700, Matt Thomas wrote:
[...]
> you could recode it to extract two bits from 0x78 (01.11.10.00) which
> avoid a memory reference.  Something like:
> 
>         fnstcw	-4(%rsp)
>         movl	-4(%rsp),  %ecx
> 	shrl	$9, %ecx
>         andl    $6, %ecx
>         movl    $120, %eax
>         sarl    %cl, %eax
>         andl    $3, %eax
>         ret

It works, but with the inverted bits sequence :

   00.10.11.01 = 0x2d = 45

Thanks.

--

-- 
Nicolas Joly

Projects and Developments in Bioinformatics
Institut Pasteur, Paris.

Christos Zoulas | 30 Sep 2011 00:57

Re: FLT_ROUNDS failures ... gcc/binutils regression ?

In article <20110929171342.GB24332 <at> lynche.sis.pasteur.fr>,
Nicolas Joly  <njoly <at> pasteur.fr> wrote:
>On Tue, Sep 27, 2011 at 02:39:40PM -0700, Matt Thomas wrote:
>[...]
>> you could recode it to extract two bits from 0x78 (01.11.10.00) which
>> avoid a memory reference.  Something like:
>> 
>>         fnstcw	-4(%rsp)
>>         movl	-4(%rsp),  %ecx
>> 	shrl	$9, %ecx
>>         andl    $6, %ecx
>>         movl    $120, %eax
>>         sarl    %cl, %eax
>>         andl    $3, %eax
>>         ret
>
>It works, but with the inverted bits sequence :
>
>   00.10.11.01 = 0x2d = 45
>

Or just:
        fnstcw  -4(%rsp)
        movl    -4(%rsp), %eax
        shrl    $10, %eax
        andl    $3, %eax
        xorl    $1, %eax        /* map 0,1,2,3 -> 1,0,3,2 */
        ret

Which is the version I committed.

(Continue reading)

Group

gmane.os.netbsd.ports.x86-64.

Project Web Page

NetBSD on AMD x86-64 architecture

Search Archive

Page

1 | 2

Articles in period: 13

September 2011

Language

Change language

Options

Current view: Threads only / Showing only 20 lines / Not hiding cited text.
Change to All messages, whole messages, or hide cited text.

Post a message
NNTP Newsgroup
Classic Gmane web interface
XML

XML

RSS Feed
List Information

About Gmane

Recent entries

Changing the default i387 precision (4 May 2013)
port-amd64/47782 - ifconfig SIOCSIFMEDIA error with bge (1 May 2013)
Intermittent SpeedStep? (7 Mar 2013)
Does that work? Intel C204 PCH Chipset on /AMD (20 Feb 2013)
32bit linux emulation on amd64 (19 Feb 2013)
Hello (29 Jan 2013)
RAID5 3x 3TB SATA drives, help! (29 Jan 2013)
de vendedor a cerrador (25 Jan 2013)
Boot failure while installing on ASUS UEFI (7 Jan 2013)
BETA6.0 - AMD Opetron 6272 (16 core) multiprocessor config crashes (24 Dec 2012)
running i386 1.0 binaries on amd64 (10 Dec 2012)
x86 pcitag_t change proposal/patch (5 Dec 2012)
Temperature sensors on HP N40L? (26 Nov 2012)
i5 and invalid instructions (13 Nov 2012)
fbsd binaries on amd64? or native 3ware 9000 series mgmt? (7 Sep 2012)
ACPI S3 suspend/resume troubleshooting (26 Aug 2012)
VM_PHYSSEG_MAX (14 Aug 2012)
Conect a modem using com port (18 Jul 2012)
HP DL380 Gen8 / G8 Boot Log (12 Jul 2012)
FYI: -current sysinst in amd64 domu dumps core (24 Jun 2012)
fail to boot 6.0 BETA2 (20 Jun 2012)

Archives

11	May 2013
2	April 2013
1	March 2013
13	February 2013
18	January 2013
10	December 2012
19	November 2012
6	September 2012
11	August 2012
11	July 2012
10	June 2012
25	May 2012
36	April 2012
1	March 2012
22	February 2012
12	January 2012
32	December 2011
28	November 2011
15	October 2011
13	September 2011
6	August 2011
19	July 2011
46	June 2011
12	May 2011
7	April 2011
11	March 2011
10	February 2011
28	January 2011
6	December 2010
7	November 2010
1	October 2010
1	September 2010
19	August 2010
26	July 2010
16	June 2010
33	May 2010
4	April 2010
7	March 2010
16	February 2010
25	January 2010
9	December 2009
20	November 2009
16	October 2009
19	September 2009
20	August 2009
18	July 2009
17	June 2009
34	May 2009
10	April 2009
24	March 2009
92	February 2009
41	January 2009
58	December 2008
133	November 2008
43	October 2008
54	September 2008
27	August 2008
35	July 2008
51	June 2008
74	May 2008
33	April 2008
51	March 2008
50	February 2008
86	January 2008
117	December 2007
76	November 2007
51	October 2007
33	September 2007
61	August 2007
25	July 2007
43	June 2007
24	May 2007
52	April 2007
104	March 2007
57	February 2007
34	January 2007
27	December 2006
20	November 2006
38	October 2006
56	September 2006
115	August 2006
74	July 2006
22	June 2006
43	May 2006
43	April 2006
109	March 2006
80	February 2006
133	January 2006
62	December 2005
52	November 2005
31	October 2005
110	September 2005
61	August 2005
96	July 2005
86	June 2005
29	May 2005
41	April 2005
66	March 2005
67	February 2005
51	January 2005
51	December 2004
39	November 2004
21	October 2004
13	September 2004
10	August 2004
18	July 2004
41	June 2004
21	May 2004
46	April 2004
66	March 2004
42	February 2004
20	January 2004
24	December 2003
51	November 2003
56	October 2003
14	September 2003
2	August 2003
7	May 2003
12	April 2003
1	March 2003
2	February 2003

Design

Zawodny | Right Menu | Left Menu

Your Own Design

Paste the URL of your CSS below. Download CSS template