Re: [patches] Re: 2.4 module on amd64

[Will C. <wc...@re...>]

I have tried some other things and I haven't had much success with 
getting the AMD64 2.4 oprofile driver working. I am attempting to run 
things on a dual-processor newisys 2100 machine 
(http://www.newisys.com/products/2100.html).

I started with a RHEL 3 amd64 smp config file and did a "make oldconfig" 
in each of the cases.  Here are the thing I have tried and all seemed to 
fail for the oprofile driver for 2.4 linux:

2.4.21.SUSE kernel with RHEL 3 amd64 smp config, oprofile 0.5.4 (2.4 module)

2.4.21.SUSE kernel with RHEL 3 amd64 smp config, oprofile 0.7cvs 
snapshot 10/13/2003 (2.4 module)

2.4.21.SUSE kernel with RHEL 3 amd64 smp config, oprofile 0.7cvs 
snapshot 10/13/2003 (2.4 module) with phe's tgid patch reverted

2.4.23-pre7 from x86-64.org cvs with RHEL 3 amd64 smp config, oprof 0.7 cvs

Maybe problem is due to the config file I am build the kernel with or 
the hardware itself.

The RHEL 3 kernel which uses a back port of the 2.6 linux kernel 
oprofile support works fine on the machine with oprofile 0.7 cvs.

-Will


Andi Kleen wrote:
> On Tue, Oct 14, 2003 at 04:38:17PM -0400, Will Cohen wrote:
> 
>>Andi Kleen wrote:
>>
>>>>AMD Hammer support requires a recent (>= 2.4.19) kernel with the line 
>>>>EXPORT_SYMBOL(do_fork); present in kernel/ksyms.c. Such a kernel is 
>>>>present in the x86-64.org CVS repository.
>>>
>>>
>>>Indeed it got lost. I will readd it.
>>>
>>>BTW we're using the 2.4 oprofile without problems with a 2.4.21 based
>>>kernel on x86-64.
>>>
>>>2.6 works too.
>>>
>>>-Andi
>>
>>Where can I get the source for the x86-64 2.4.21 kernel that oprofile is 
>>known to work with?
> 
> 
> It's the SuSE kernel (ftp://ftp.suse.com/pub/people/mantel/next/ has a version)
> 
> 
> 
>>It looks like there are some other problems with the 2.4.23-pre7 kernel. 
>>I manually added the export, so the oprofile module would load. The 
>>module loaded fine. When I tried to start the data collection. Lots of 
>>error messages like the following started showing up.
>>
>>
>>Oct 14 11:25:18 katzj-amd64 kernel: bash[3296]: segfault at 
>>fffffffffffffff8 rip 0000002a9591d0cb rsp 0000000000000000 error 6
>>Oct 14 11:25:23 katzj-amd64 kernel: bash[3297] bad frame in signal 
>>deliver frame:fffffffffffffdb8 rip:2a9591d0cb rsp:0 orax:ffffffffffffffff
>>Oct 14 11:25:23 katzj-amd64 kernel: bash[3297]: segfault at 
>>fffffffffffffff8 rip 0000002a9591d0cb rsp 0000000000000000 error 6
>>Oct 14 11:25:25 katzj-amd64 kernel: bash[3298] bad frame in signal 
>>deliver frame:fffffffffffffdb8 rip:2a9591d0cb rsp:0 orax:ffffffffffffffff
> 
> 
> Looks like your module is corrupting the user space context.
> 
> -Andi