OProfile (Linux profiler) development and users list

Mathieu Desnoyers | 2 Jan 2008 18:43

Picon

Picon

Re: [2.6.24 patch] Fix ARMv6 OProfile support

* Mathieu Desnoyers (mathieu.desnoyers <at> polymtl.ca) wrote:
> Fix ARMv6 OProfile support
> 
> This patch restores the ARMv6 OProfile support that was killed by
> commit 09cadedbdc01f1a4bea1f427d4fb4642eaa19da9.
> 
> It puts the config options in arch/arm/Kconfig.
> 

This patch header should also include :

Problem identified by Adrian Bunk. Patch inspired by the original fix
proposed by him.

Thanks,

Mathieu

> Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers <at> polymtl.ca>
> CC: Adrian Bunk <adrian.bunk <at> movial.fi>
> CC: Randy Dunlap <randy.dunlap <at> oracle.com>
> CC: rmk <at> arm.linux.org.uk
> CC: phil.el <at> wanadoo.fr
> CC: oprofile-list <at> lists.sourceforge.net
> ---
>  arch/arm/Kconfig |   19 +++++++++++++++++++
>  1 file changed, 19 insertions(+)
> 
> Index: linux-2.6-lttng/arch/arm/Kconfig
> ===================================================================

(Continue reading)

Mathieu Desnoyers | 2 Jan 2008 18:47

Picon

Picon

Re: [2.6.24 patch] Fix Cell OProfile support

* Arnd Bergmann (arnd <at> arndb.de) wrote:
> On Saturday 29 December 2007, Mathieu Desnoyers wrote:
> > This patch restores the Cell OProfile support that was killed by
> > commit 09cadedbdc01f1a4bea1f427d4fb4642eaa19da9.
> > 
> > It puts it in arch/powerpc/Kconfig. Since I don't see any good reason to leave
> > this as a supplementary user-selectable option, it is now automatically enabled
> > whenever SPU_FS and OPROFILE are enabled.
> 
> This one has already been superceded by the fix in 
> aed3a8c9bb1a8623a618232087c5ff62718e3b9a, which made CONFIG_OPROFILE_CELL an
> automatically selected option, from arch/powerpc/platforms/cell/Kconfig.
> 

Great, so my patch should be dropped. Thanks!

Mathieu

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

--

-- 
Mathieu Desnoyers
Computer Engineering Ph.D. Student, Ecole Polytechnique de Montreal
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68

(Continue reading)

Maynard Johnson | 4 Jan 2008 22:01

Picon

[PATCH for JIT_SUPPORT branch] symbol versioning support

The main intent of this patch is to set up the JIT support library build 
process for symbol versioning versus library versioning.  This technique 
is used by GLIBC and appears to be gaining popularity as the preferred 
way to handle library versioning issues.  The main idea behind this 
technique is that both old and new symbols coexist in the same binary.  
When a user links an agent lib against our libopagent library, the 
version of the symbols they're linking with is built into their binary.  
Then, if a new libopagent comes out with a new interface (i.e. new 
functions), the agent libs linked against the older libopagent still 
work, and agent libs linked against the new libopagent can use the new 
functions.  As long as an agent library does not use any new functions, 
it can be used with either the old or the new libopagent.  If an agent 
library makes use of a new function, then it must use the new 
libopagent.  An attempt to use (at runtime) an older version would 
result in an error at load time:  e.g., "symbol foo, version OPAGENT_2.0 
not defined in libopagent.so.1".

There is also a technique for changing the signatures of existing 
functions.  This technique allows all versions of the same symbol to 
coexist in the library so that older agent libraries can still use the 
older version of the function.  See the note added to libopagent.c in 
this patch for details.

Aside from the patched files, a new file (attached) will be added to the 
libopagent directory: opagent.ver.  This is a version script file passed 
to the linker that's used for the actual symbol versioning.  
Technically, we would not need to have a version script file for the 
base version of the library since, by default, all the symbols being 
defined now would be given the base version.  But it doesn't hurt and it 
sets up the infrastructure for future needs.

(Continue reading)

John Levon | 5 Jan 2008 03:34

Re: [PATCH for JIT_SUPPORT branch] symbol versioning support

On Fri, Jan 04, 2008 at 03:01:02PM -0600, Maynard Johnson wrote:

> library makes use of a new function, then it must use the new 
> libopagent.  An attempt to use (at runtime) an older version would 
> result in an error at load time:  e.g., "symbol foo, version OPAGENT_2.0 
> not defined in libopagent.so.1".

I agree with symbol versioning in principle.

> Aside from the patched files, a new file (attached) will be added to the 
> libopagent directory: opagent.ver.  This is a version script file passed 

Is this the normal suffix used for such files (I don't remember seeing
such a suffix before?)

> to the linker that's used for the actual symbol versioning.  
> Technically, we would not need to have a version script file for the 
> base version of the library since, by default, all the symbols being 
> defined now would be given the base version.  But it doesn't hurt and it 
> sets up the infrastructure for future needs.

It's absolutely critical that we clearly define this, so it's definitely
important.

> This patch also introduces the use of libtool for building libopagent 

That's fine, I think. We're Linux-specific enough for it not to matter
(it might have done, but the BSD people decided to ignore the work we've
done - oh well).

(Continue reading)

Maynard Johnson | 6 Jan 2008 18:39

Picon

Re: [PATCH for JIT_SUPPORT branch] symbol versioning support

John Levon wrote:
> On Fri, Jan 04, 2008 at 03:01:02PM -0600, Maynard Johnson wrote:
>
>   
>> library makes use of a new function, then it must use the new 
>> libopagent.  An attempt to use (at runtime) an older version would 
>> result in an error at load time:  e.g., "symbol foo, version OPAGENT_2.0 
>> not defined in libopagent.so.1".
>>     
>
> I agree with symbol versioning in principle.
>
>   
>> Aside from the patched files, a new file (attached) will be added to the 
>> libopagent directory: opagent.ver.  This is a version script file passed 
>>     
>
> Is this the normal suffix used for such files (I don't remember seeing
> such a suffix before?)
>   
There doesn't seem to be a standard.  I've seen three different suffixes 
used in different projects:  ".ver" for libstdc++; ".map" for GLIBC; and 
".in" for libtheora.  I'm sure if I kept looking, I could find more.  
But thinking a bit more about the name we should use for our 
version-script file, perhaps something more obvious would be helpful -- 
e.g., opagent_symbols.ver or opagent_version_script.ver, etc.

-Maynard
>   
>> to the linker that's used for the actual symbol versioning.

(Continue reading)

Michael Ellerman | 7 Jan 2008 08:26

Picon

oprofiled crashing on cell?

Hi all,

Running oprofile (0.9.3) on a cell machine (2.6.24-rc7 kernel) I see the
oprofiled intermittently crashing. It only seems to happen when I run an
SPU program.

When it crashes I see this in the log:

oprofiled started Mon Jan  7 18:23:21 2008
kernel pointer size: 8
Read buffer of 98307 entries.
No anon map for pc 0, app anonymous.

Compared to a working run:

oprofiled started Mon Jan  7 18:21:12 2008
kernel pointer size: 8
Read buffer of 11 entries.
Dangling ESCAPE_CODE.
<snip>

I've tried strace'ing oprofiled but that seems to hide the bug. Does
anyone have any ideas?

cheers

--

-- 
Michael Ellerman
OzLabs, IBM Australia Development Lab

(Continue reading)

Maynard Johnson | 7 Jan 2008 16:13

Picon

Re: [Cbe-oss-dev] oprofiled crashing on cell?

Michael Ellerman wrote:
> Hi all,
>
> Running oprofile (0.9.3) on a cell machine (2.6.24-rc7 kernel) I see the
> oprofiled intermittently crashing. It only seems to happen when I run an
> SPU program.
>
> When it crashes I see this in the log:
>
> oprofiled started Mon Jan  7 18:23:21 2008
> kernel pointer size: 8
> Read buffer of 98307 entries.
> No anon map for pc 0, app anonymous.
>   
Well, that's definitely badness, but this, in itself, would not cause 
oprofiled to crash.  Is this the last thing you see in the log?  Does 
the daemon fail both with and without the --verbose option?
> Compared to a working run:
>
> oprofiled started Mon Jan  7 18:21:12 2008
> kernel pointer size: 8
> Read buffer of 11 entries.
> Dangling ESCAPE_CODE.
> <snip>
>   
A dangling ESCAPE code is badness, too.  For Cell, a buffer with 11 
entries could mean 3 entries for profiling start header info + 8 entries 
for SPU context info.  The 11th entry would be the offset of the SPU ELF 
data, if embedded; otherwise 0.  According to the above log snippet, the 
11th entry is an ESCAPE_CODE.  This implies to me that another event

(Continue reading)

Bob Nelson | 7 Jan 2008 20:30

Picon

Re: [Cbe-oss-dev] oprofiled crashing on cell?

On Monday 07 January 2008 09:13:13 am Maynard Johnson wrote:
> Michael Ellerman wrote:
> > Hi all,
> >
> > Running oprofile (0.9.3) on a cell machine (2.6.24-rc7 kernel) I see the
> > oprofiled intermittently crashing. It only seems to happen when I run an
> > SPU program.
> >
> > When it crashes I see this in the log:
> >
> > oprofiled started Mon Jan  7 18:23:21 2008
> > kernel pointer size: 8
> > Read buffer of 98307 entries.
> > No anon map for pc 0, app anonymous.
> >   
> Well, that's definitely badness, but this, in itself, would not cause 
> oprofiled to crash.  Is this the last thing you see in the log?  Does 
> the daemon fail both with and without the --verbose option?
> > Compared to a working run:
> >
> > oprofiled started Mon Jan  7 18:21:12 2008
> > kernel pointer size: 8
> > Read buffer of 11 entries.
> > Dangling ESCAPE_CODE.
> > <snip>
> >   
> A dangling ESCAPE code is badness, too.  For Cell, a buffer with 11 
> entries could mean 3 entries for profiling start header info + 8 entries 
> for SPU context info.  The 11th entry would be the offset of the SPU ELF 
> data, if embedded; otherwise 0.  According to the above log snippet, the

(Continue reading)

Michael Ellerman | 7 Jan 2008 23:30

Picon

Re: [Cbe-oss-dev] oprofiled crashing on cell?

On Mon, 2008-01-07 at 09:13 -0600, Maynard Johnson wrote:
> Michael Ellerman wrote:
> > Hi all,
> >
> > Running oprofile (0.9.3) on a cell machine (2.6.24-rc7 kernel) I see the
> > oprofiled intermittently crashing. It only seems to happen when I run an
> > SPU program.
> >
> > When it crashes I see this in the log:
> >
> > oprofiled started Mon Jan  7 18:23:21 2008
> > kernel pointer size: 8
> > Read buffer of 98307 entries.
> > No anon map for pc 0, app anonymous.

> Well, that's definitely badness, but this, in itself, would not cause 
> oprofiled to crash.  Is this the last thing you see in the log?  Does 
> the daemon fail both with and without the --verbose option?

This is the entire log and that's running with --verbose, which has no
effect on whether it crashes or not. Here's another run I just  did:

oprofiled started Tue Jan  8 09:23:20 2008
kernel pointer size: 8
Read buffer of 76017 entries.
No anon map for pc 0, app anonymous.

So the buffer size is changing but nothing else.

Any hint as to what the message means? "no anon map?"

(Continue reading)

Michael Ellerman | 7 Jan 2008 23:31

Picon

Re: [Cbe-oss-dev] oprofiled crashing on cell?

On Mon, 2008-01-07 at 13:30 -0600, Bob Nelson wrote:
> On Monday 07 January 2008 09:13:13 am Maynard Johnson wrote:
> > Michael Ellerman wrote:
> > > Hi all,
> > >
> > > Running oprofile (0.9.3) on a cell machine (2.6.24-rc7 kernel) I see the
> > > oprofiled intermittently crashing. It only seems to happen when I run an
> > > SPU program.
> > >
> > > When it crashes I see this in the log:
> > >
> > > oprofiled started Mon Jan  7 18:23:21 2008
> > > kernel pointer size: 8
> > > Read buffer of 98307 entries.
> > > No anon map for pc 0, app anonymous.
> > >   
> > Well, that's definitely badness, but this, in itself, would not cause 
> > oprofiled to crash.  Is this the last thing you see in the log?  Does 
> > the daemon fail both with and without the --verbose option?
> > > Compared to a working run:
> > >
> > > oprofiled started Mon Jan  7 18:21:12 2008
> > > kernel pointer size: 8
> > > Read buffer of 11 entries.
> > > Dangling ESCAPE_CODE.
> > > <snip>
> > >   
> > A dangling ESCAPE code is badness, too.  For Cell, a buffer with 11 
> > entries could mean 3 entries for profiling start header info + 8 entries 
> > for SPU context info.  The 11th entry would be the offset of the SPU ELF

(Continue reading)

Group

gmane.linux.oprofile.

Project Web Page

OProfile (Linux profiler) development and users list

Search Archive

Page

1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11

Articles in period: 104

January 2008

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

[PATCH] Fix bug if .oprofile/dameonrc doesn't exists (23 May 2013)
[PATCH] Fix Qt4 error messages at oprof_start startup (23 May 2013)
oprofile on blackfin uClinux system (23 May 2013)
[PATCH] Fix recent regression involving unit mask values of '0' (22 May 2013)
Revive this old patch (21 May 2013)
[PATCH] Fix Coverity errors found on May 20, 2013 git snapshot (21 May 2013)
[PATCH] Add support for architected events for IBM ppc64 architecture (17 May 2013)
[PATCH] Add support for Intel Netburst (e.g., Pentium P4) to operf (15 May 2013)
New Coverity scan (8 May 2013)
[ oprofile-Bugs-3612911 ] Coverity scan results: May 8, 2013 (8 May 2013)
[PATCH] Fix Coverity issues identified against oprofile 0.9.8 release (8 May 2013)
[ oprofile-Bugs-3612843 ] Feature request: oprofile should have an eve (7 May 2013)
RFC: events counting with operf (7 May 2013)
Some help needed about operf (7 May 2013)
Regarding performance on shared library (2 May 2013)
problem related to oprofile (30 Apr 2013)
'precise_ip' compile error from OProfile Git tree (30 Apr 2013)
Oprofine-0.9.8 install error (27 Apr 2013)
[PATCH] Use PMC5/PMC6 on ppc64 arch for run cycles/run instructions (26 Apr 2013)
Fwd: no sample when profiling ARM Cortex-A9 with Linux kernel 3.3 (26 Apr 2013)
no sample when profiling ARM Cortex-A9 with Linux kernel 3.3 (26 Apr 2013)

Archives

50	May 2013
128	April 2013
65	March 2013
63	February 2013
94	January 2013
89	December 2012
42	November 2012
45	October 2012
58	September 2012
72	August 2012
151	July 2012
89	June 2012
103	May 2012
36	April 2012
39	March 2012
62	February 2012
105	January 2012
95	December 2011
88	November 2011
115	October 2011
58	September 2011
54	August 2011
55	July 2011
82	June 2011
143	May 2011
121	April 2011
121	March 2011
54	February 2011
63	January 2011
57	December 2010
50	November 2010
84	October 2010
83	September 2010
69	August 2010
35	July 2010
65	June 2010
85	May 2010
50	April 2010
88	March 2010
59	February 2010
71	January 2010
79	December 2009
88	November 2009
147	October 2009
80	September 2009
75	August 2009
155	July 2009
153	June 2009
203	May 2009
197	April 2009
157	March 2009
85	February 2009
131	January 2009
113	December 2008
114	November 2008
162	October 2008
88	September 2008
89	August 2008
206	July 2008
99	June 2008
114	May 2008
125	April 2008
48	March 2008
67	February 2008
104	January 2008
52	December 2007
111	November 2007
102	October 2007
40	September 2007
54	August 2007
184	July 2007
105	June 2007
145	May 2007
124	April 2007
43	March 2007
119	February 2007
63	January 2007
105	December 2006
101	November 2006
87	October 2006
99	September 2006
44	August 2006
36	July 2006
60	June 2006
66	May 2006
95	April 2006
65	March 2006
117	February 2006
102	January 2006
60	December 2005
77	November 2005
55	October 2005
16	September 2005
87	August 2005
35	July 2005
35	June 2005
69	May 2005
151	April 2005
120	March 2005
42	February 2005
25	January 2005
33	December 2004
114	November 2004
71	October 2004
87	September 2004
89	August 2004
32	July 2004
34	June 2004
92	May 2004
72	April 2004
58	March 2004
93	February 2004
99	January 2004
49	December 2003
99	November 2003
102	October 2003
41	September 2003
95	August 2003
86	July 2003
58	June 2003
83	May 2003
103	April 2003
187	March 2003
131	February 2003

Design

Zawodny | Right Menu | Left Menu

Your Own Design

Paste the URL of your CSS below. Download CSS template