Discussion:
Anyone able to boot multiple cpus in x86 FS?
(too old to reply)
Ankita (Garg) Goel
2012-02-24 07:23:56 UTC
Permalink
Hi,

Has anyone been able to boot multiple cpus wth FS mode on x86 ? If yes, it
would greatly help if you could share your environmental setup. The issues
I am facing I have already posted in some previous emails.

Thanks for your help!

--
Regards,
Ankita
Gabriel Michael Black
2012-02-24 10:45:23 UTC
Permalink
Yes, I have. You're using O3 though, and that support is fairly new.

Gabe

Quoting "Ankita (Garg) Goel" <***@gmail.com>:

> Hi,
>
> Has anyone been able to boot multiple cpus wth FS mode on x86 ? If yes, it
> would greatly help if you could share your environmental setup. The issues
> I am facing I have already posted in some previous emails.
>
> Thanks for your help!
>
> --
> Regards,
> Ankita
>
Gabe Black
2012-02-26 12:51:56 UTC
Permalink
I looked into this more, and the code that sets up the memory hierarchy
was broken a little while ago. I have a patch that seems to fix it
(attached) but then it gets stuck a little later. I remember a while ago
someone had problems with atomic operations not being supported by the
memory system fully, but I don't remember the specifics. This might be
that problem too.

Gabe

On 02/24/12 02:45, Gabriel Michael Black wrote:
> Yes, I have. You're using O3 though, and that support is fairly new.
>
> Gabe
>
> Quoting "Ankita (Garg) Goel" <***@gmail.com>:
>
>> Hi,
>>
>> Has anyone been able to boot multiple cpus wth FS mode on x86 ? If
>> yes, it
>> would greatly help if you could share your environmental setup. The
>> issues
>> I am facing I have already posted in some previous emails.
>>
>> Thanks for your help!
>>
>> --
>> Regards,
>> Ankita
>>
>
>
>
> _______________________________________________
> gem5-users mailing list
> gem5-***@gem5.org
> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
Jason Clemons
2012-02-27 22:51:54 UTC
Permalink
Gabe,

Thanks for the patch. I did a hg pull and hg update this afternoon
and have made some progress on booting. I found that with a custom
built version of 2.6.28.4 with smp on, the patch and the linux image
files (linux-bigswap2.img and linux-x86.img) from gem5.org I am able
to boot multicore x86 using the default cpu type which I think is
atomic.

After working with Andrea, we noted that the boot problem for x86
happens with the PARSEC root image but not the linux-x86.img from
gem5.org. This was tested using the custom kernel and the default cpu
type.

I am currently attempting to test the detailed cpu-type with 2 cpus.
I will let you know of any errors I run into.


Thanks,

Jason Clemons





Quoting Gabe Black <***@eecs.umich.edu>:

> I looked into this more, and the code that sets up the memory hierarchy
> was broken a little while ago. I have a patch that seems to fix it
> (attached) but then it gets stuck a little later. I remember a while ago
> someone had problems with atomic operations not being supported by the
> memory system fully, but I don't remember the specifics. This might be
> that problem too.
>
> Gabe
>
> On 02/24/12 02:45, Gabriel Michael Black wrote:
>> Yes, I have. You're using O3 though, and that support is fairly new.
>>
>> Gabe
>>
>> Quoting "Ankita (Garg) Goel" <***@gmail.com>:
>>
>>> Hi,
>>>
>>> Has anyone been able to boot multiple cpus wth FS mode on x86 ? If
>>> yes, it
>>> would greatly help if you could share your environmental setup. The
>>> issues
>>> I am facing I have already posted in some previous emails.
>>>
>>> Thanks for your help!
>>>
>>> --
>>> Regards,
>>> Ankita
>>>
>>
>>
>>
>> _______________________________________________
>> gem5-users mailing list
>> gem5-***@gem5.org
>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>
>
Xuhao Chen
2012-02-28 13:40:05 UTC
Permalink
Hi everyone,
The Solaris 10 binary image for SPARC includes "reset.bin, q.bin, openboot.bin" etc.
I copy *.bin and nvram1 from OpenSPARCT1_Arch.1.5/S10image/ to
the /dist/m5/system/binaries/ directory.
Also rename reset.bin, q.bin, and openboot.bin to reset_new.bin, q_new.bin,
and openboot_new.bin, which are the binaries expected by the m5 SPARC_FS scripts. Then copy disk.s10hw2 from the S10image/ directory to the /dist/m5/system/disks/ directory.
But when booting the system, it still cannot find the kernel file:

info: No kernel set for full system simulation. Assuming you know what you're doing...

Anyone know the reason? Should I use the option --kernel to specify the kernel file? But which one is the kernel file?
By the way, how many CPUs can gem5 run in SPARC FS mode?
Thanks.




Xuhao Chen
PhD student
School of Computer
National University of Defense Technology
Changsha, Hunan, P.R.China, 410073
Tel: +86-159-741-03340 (Mobile)
Nilay Vaish
2012-02-28 14:20:52 UTC
Permalink
On Tue, 28 Feb 2012, Xuhao Chen wrote:

> Hi everyone,
> The Solaris 10 binary image for SPARC includes "reset.bin, q.bin, openboot.bin" etc.
> I copy *.bin and nvram1 from OpenSPARCT1_Arch.1.5/S10image/ to
> the /dist/m5/system/binaries/ directory.
> Also rename reset.bin, q.bin, and openboot.bin to reset_new.bin, q_new.bin,
> and openboot_new.bin, which are the binaries expected by the m5 SPARC_FS scripts. Then copy disk.s10hw2 from the S10image/ directory to the /dist/m5/system/disks/ directory.
> But when booting the system, it still cannot find the kernel file:
>
> info: No kernel set for full system simulation. Assuming you know what you're doing...
>
> Anyone know the reason? Should I use the option --kernel to specify the kernel file? But which one is the kernel file?
> By the way, how many CPUs can gem5 run in SPARC FS mode?
> Thanks.
>

I don't think any one has tried booting a full system for SPARC
architecture. gem5 does not provide any kernel for SPARC as of now. You
can compile one on your own and try if it works.

--
Nilay
Xuhao Chen
2012-02-28 16:22:29 UTC
Permalink
Dear Nilay,
I found it here "http://gedare-csphd.blogspot.com/2010/05/week-in-m5.html", but I don't know whether I do it correctly.
Thank you.




Xuhao Chen
PhD student
School of Computer
National University of Defense Technology
Changsha, Hunan, P.R.China, 410073
Tel: +86-159-741-03340 (Mobile)

From: Nilay Vaish
Date: 2012-02-28 22:20
To: gem5 users mailing list
Subject: Re: [gem5-users] How to boot SPARC FS?
On Tue, 28 Feb 2012, Xuhao Chen wrote:

> Hi everyone,
> The Solaris 10 binary image for SPARC includes "reset.bin, q.bin, openboot.bin" etc.
> I copy *.bin and nvram1 from OpenSPARCT1_Arch.1.5/S10image/ to
> the /dist/m5/system/binaries/ directory.
> Also rename reset.bin, q.bin, and openboot.bin to reset_new.bin, q_new.bin,
> and openboot_new.bin, which are the binaries expected by the m5 SPARC_FS scripts. Then copy disk.s10hw2 from the S10image/ directory to the /dist/m5/system/disks/ directory.
> But when booting the system, it still cannot find the kernel file:
>
> info: No kernel set for full system simulation. Assuming you know what you're doing...
>
> Anyone know the reason? Should I use the option --kernel to specify the kernel file? But which one is the kernel file?
> By the way, how many CPUs can gem5 run in SPARC FS mode?
> Thanks.
>

I don't think any one has tried booting a full system for SPARC
architecture. gem5 does not provide any kernel for SPARC as of now. You
can compile one on your own and try if it works.

--
Nilay
_______________________________________________
gem5-users mailing list
gem5-***@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
Gedare Bloom
2012-02-28 17:27:24 UTC
Permalink
On Tue, Feb 28, 2012 at 11:15 AM, Xuhao Chen <***@gmail.com> wrote:
> Hi Gedare,
>
> I have read your blog "a week in m5".  And I have a question about m5.
> The Solaris 10 binary image for SPARC includes "reset.bin, q.bin, openboot.bin" etc.
> I copy *.bin and nvram1 from OpenSPARCT1_Arch.1.5/S10image/ to
> the /dist/m5/system/binaries/ directory.
> Also rename reset.bin, q.bin, and openboot.bin to reset_new.bin, q_new.bin,
> and openboot_new.bin, which are the binaries expected by the m5 SPARC_FS scripts.
> Then
> copy disk.s10hw2 from the S10image/ directory to the /dist/m5/system/disks/ directory.
> But when booting the system, it still cannot find the kernel file:
>
> info: No kernel set for full system simulation. Assuming you know what you're doing...
>
> Do you know the reason? Should I use the option --kernel to specify the
> kernel file? But which one is the kernel file?
> By the way, how many CPUs can gem5 run in SPARC FS mode?  I know that it can
> be up to 64 CPUs in Alpha FS mode.
> Thanks.
>
It has been awhile since I worked with GEM5 so I cannot be certain;
that post is nearly 2 years old so a lot may have changed since then.
I did not have to specify a --kernel option, and the fs.py
(FSConfig.py) had all the options set to point toward the OpenBoot
binary files and example disk containing the OpenSolaris image. The
info message does not necessarily mean there is a problem, and if you
connect with m5term you might still reach the OpenBoot prompt anyway.

When I was using the SPARC_FS only 1 core was supported. Maybe someone
has been developing this port but I am not aware. Unless you are
really committed to the SPARC architecture, I would suggest trying one
of the other CPU models.

Good luck,
Gedare

> ________________________________
>
> Xuhao Chen
>
> PhD student
>
> School of Computer
>
> National University of Defense Technology
>
> Changsha, Hunan, P.R.China, 410073
>
> Tel:  +86-159-741-03340 (Mobile)
Gabe Black
2012-02-28 20:42:35 UTC
Permalink
On 02/28/12 09:27, Gedare Bloom wrote:
> On Tue, Feb 28, 2012 at 11:15 AM, Xuhao Chen <***@gmail.com> wrote:
>> Hi Gedare,
>>
>> I have read your blog "a week in m5". And I have a question about m5.
>> The Solaris 10 binary image for SPARC includes "reset.bin, q.bin, openboot.bin" etc.
>> I copy *.bin and nvram1 from OpenSPARCT1_Arch.1.5/S10image/ to
>> the /dist/m5/system/binaries/ directory.
>> Also rename reset.bin, q.bin, and openboot.bin to reset_new.bin, q_new.bin,
>> and openboot_new.bin, which are the binaries expected by the m5 SPARC_FS scripts.
>> Then
>> copy disk.s10hw2 from the S10image/ directory to the /dist/m5/system/disks/ directory.
>> But when booting the system, it still cannot find the kernel file:
>>
>> info: No kernel set for full system simulation. Assuming you know what you're doing...
>>
>> Do you know the reason? Should I use the option --kernel to specify the
>> kernel file? But which one is the kernel file?
>> By the way, how many CPUs can gem5 run in SPARC FS mode? I know that it can
>> be up to 64 CPUs in Alpha FS mode.
>> Thanks.
>>
> It has been awhile since I worked with GEM5 so I cannot be certain;
> that post is nearly 2 years old so a lot may have changed since then.
> I did not have to specify a --kernel option, and the fs.py
> (FSConfig.py) had all the options set to point toward the OpenBoot
> binary files and example disk containing the OpenSolaris image. The
> info message does not necessarily mean there is a problem, and if you
> connect with m5term you might still reach the OpenBoot prompt anyway.
>
> When I was using the SPARC_FS only 1 core was supported. Maybe someone
> has been developing this port but I am not aware. Unless you are
> really committed to the SPARC architecture, I would suggest trying one
> of the other CPU models.
>
> Good luck,
> Gedare
>
>> ________________________________
>>
>> Xuhao Chen
>>
>> PhD student
>>
>> School of Computer
>>
>> National University of Defense Technology
>>
>> Changsha, Hunan, P.R.China, 410073
>>
>> Tel: +86-159-741-03340 (Mobile)
> _______________________________________________
> gem5-users mailing list
> gem5-***@gem5.org
> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users

Also I should point out that I think image is for booting Solaris, I
think, not Linux. I don't think we got SMP working for SPARC FS, but I
don't really remember.

Gabe
Xuhao Chen
2012-02-29 01:59:38 UTC
Permalink
Thank you for your reply.
Yes, the bin files are from Solaris 10.




Xuhao Chen
PhD student
School of Computer
National University of Defense Technology
Changsha, Hunan, P.R.China, 410073
Tel: +86-159-741-03340 (Mobile)

From: Gabe Black
Date: 2012-02-29 04:42
To: gem5-users
Subject: Re: [gem5-users] How to boot SPARC FS?
On 02/28/12 09:27, Gedare Bloom wrote:
> On Tue, Feb 28, 2012 at 11:15 AM, Xuhao Chen <***@gmail.com> wrote:
>> Hi Gedare,
>>
>> I have read your blog "a week in m5". And I have a question about m5.
>> The Solaris 10 binary image for SPARC includes "reset.bin, q.bin, openboot.bin" etc.
>> I copy *.bin and nvram1 from OpenSPARCT1_Arch.1.5/S10image/ to
>> the /dist/m5/system/binaries/ directory.
>> Also rename reset.bin, q.bin, and openboot.bin to reset_new.bin, q_new.bin,
>> and openboot_new.bin, which are the binaries expected by the m5 SPARC_FS scripts.
>> Then
>> copy disk.s10hw2 from the S10image/ directory to the /dist/m5/system/disks/ directory.
>> But when booting the system, it still cannot find the kernel file:
>>
>> info: No kernel set for full system simulation. Assuming you know what you're doing...
>>
>> Do you know the reason? Should I use the option --kernel to specify the
>> kernel file? But which one is the kernel file?
>> By the way, how many CPUs can gem5 run in SPARC FS mode? I know that it can
>> be up to 64 CPUs in Alpha FS mode.
>> Thanks.
>>
> It has been awhile since I worked with GEM5 so I cannot be certain;
> that post is nearly 2 years old so a lot may have changed since then.
> I did not have to specify a --kernel option, and the fs.py
> (FSConfig.py) had all the options set to point toward the OpenBoot
> binary files and example disk containing the OpenSolaris image. The
> info message does not necessarily mean there is a problem, and if you
> connect with m5term you might still reach the OpenBoot prompt anyway.
>
> When I was using the SPARC_FS only 1 core was supported. Maybe someone
> has been developing this port but I am not aware. Unless you are
> really committed to the SPARC architecture, I would suggest trying one
> of the other CPU models.
>
> Good luck,
> Gedare
>
>> ________________________________
>>
>> Xuhao Chen
>>
>> PhD student
>>
>> School of Computer
>>
>> National University of Defense Technology
>>
>> Changsha, Hunan, P.R.China, 410073
>>
>> Tel: +86-159-741-03340 (Mobile)
> _______________________________________________
> gem5-users mailing list
> gem5-***@gem5.org
> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users

Also I should point out that I think image is for booting Solaris, I
think, not Linux. I don't think we got SMP working for SPARC FS, but I
don't really remember.

Gabe
_______________________________________________
gem5-users mailing list
gem5-***@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
Xuhao Chen
2012-02-29 04:58:04 UTC
Permalink
Hi Gabe,
You mentioned that there's a configuration option for the kernel that sets how many CPUs it supports in X86 FS mode. Do you mean the config files "config-x86.tar" at http://www.m5sim.org/Download. How to apply such configuration and set it to support 64-core?
Thanks.



Xuhao Chen
PhD student
School of Computer
National University of Defense Technology
Changsha, Hunan, P.R.China, 410073
Tel: +86-159-741-03340 (Mobile)

From: Gabe Black
Date: 2012-02-29 04:42
To: gem5-users
Subject: Re: [gem5-users] How to boot SPARC FS?
On 02/28/12 09:27, Gedare Bloom wrote:
> On Tue, Feb 28, 2012 at 11:15 AM, Xuhao Chen <***@gmail.com> wrote:
>> Hi Gedare,
>>
>> I have read your blog "a week in m5". And I have a question about m5.
>> The Solaris 10 binary image for SPARC includes "reset.bin, q.bin, openboot.bin" etc.
>> I copy *.bin and nvram1 from OpenSPARCT1_Arch.1.5/S10image/ to
>> the /dist/m5/system/binaries/ directory.
>> Also rename reset.bin, q.bin, and openboot.bin to reset_new.bin, q_new.bin,
>> and openboot_new.bin, which are the binaries expected by the m5 SPARC_FS scripts.
>> Then
>> copy disk.s10hw2 from the S10image/ directory to the /dist/m5/system/disks/ directory.
>> But when booting the system, it still cannot find the kernel file:
>>
>> info: No kernel set for full system simulation. Assuming you know what you're doing...
>>
>> Do you know the reason? Should I use the option --kernel to specify the
>> kernel file? But which one is the kernel file?
>> By the way, how many CPUs can gem5 run in SPARC FS mode? I know that it can
>> be up to 64 CPUs in Alpha FS mode.
>> Thanks.
>>
> It has been awhile since I worked with GEM5 so I cannot be certain;
> that post is nearly 2 years old so a lot may have changed since then.
> I did not have to specify a --kernel option, and the fs.py
> (FSConfig.py) had all the options set to point toward the OpenBoot
> binary files and example disk containing the OpenSolaris image. The
> info message does not necessarily mean there is a problem, and if you
> connect with m5term you might still reach the OpenBoot prompt anyway.
>
> When I was using the SPARC_FS only 1 core was supported. Maybe someone
> has been developing this port but I am not aware. Unless you are
> really committed to the SPARC architecture, I would suggest trying one
> of the other CPU models.
>
> Good luck,
> Gedare
> _______________________________________________
> gem5-users mailing list
> gem5-***@gem5.org
> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users

Also I should point out that I think image is for booting Solaris, I
think, not Linux. I don't think we got SMP working for SPARC FS, but I
don't really remember.

Gabe
_______________________________________________
gem5-users mailing list
gem5-***@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
Gabriel Michael Black
2012-03-01 11:47:33 UTC
Permalink
First off, when you send emails to the list, you should make sure the
subject reflects what your asking about. Second, you shouldn't reply
to an unrelated email to ask your own question. You should start a new
thread specifically for your question.

As far as configuring things for more CPUs, you'd do that using
standard Linux kernel configuration mechanisms. In that tar ball are
regular Linux config files which you can copy into a kernel source
tree and use in the normal way. I'm sure there's lots of documentation
about that online somewhere. There are only two things you should be
careful about, getting the version of the kernel to match the config
(mostly to make life simpler for yourself) and also that exotic
options that use special hardware or platform features may not work.

Gabe

Quoting Xuhao Chen <***@gmail.com>:

> Hi Gabe,
> You mentioned that there's a configuration option for the kernel
> that sets how many CPUs it supports in X86 FS mode. Do you mean the
> config files "config-x86.tar" at http://www.m5sim.org/Download. How
> to apply such configuration and set it to support 64-core?
> Thanks.
>
>
>
> Xuhao Chen
> PhD student
> School of Computer
> National University of Defense Technology
> Changsha, Hunan, P.R.China, 410073
> Tel: +86-159-741-03340 (Mobile)
>
> From: Gabe Black
> Date: 2012-02-29 04:42
> To: gem5-users
> Subject: Re: [gem5-users] How to boot SPARC FS?
> On 02/28/12 09:27, Gedare Bloom wrote:
>> On Tue, Feb 28, 2012 at 11:15 AM, Xuhao Chen <***@gmail.com> wrote:
>>> Hi Gedare,
>>>
>>> I have read your blog "a week in m5". And I have a question about m5.
>>> The Solaris 10 binary image for SPARC includes "reset.bin, q.bin,
>>> openboot.bin" etc.
>>> I copy *.bin and nvram1 from OpenSPARCT1_Arch.1.5/S10image/ to
>>> the /dist/m5/system/binaries/ directory.
>>> Also rename reset.bin, q.bin, and openboot.bin to reset_new.bin, q_new.bin,
>>> and openboot_new.bin, which are the binaries expected by the m5
>>> SPARC_FS scripts.
>>> Then
>>> copy disk.s10hw2 from the S10image/ directory to the
>>> /dist/m5/system/disks/ directory.
>>> But when booting the system, it still cannot find the kernel file:
>>>
>>> info: No kernel set for full system simulation. Assuming you know
>>> what you're doing...
>>>
>>> Do you know the reason? Should I use the option --kernel to specify the
>>> kernel file? But which one is the kernel file?
>>> By the way, how many CPUs can gem5 run in SPARC FS mode? I know
>>> that it can
>>> be up to 64 CPUs in Alpha FS mode.
>>> Thanks.
>>>
>> It has been awhile since I worked with GEM5 so I cannot be certain;
>> that post is nearly 2 years old so a lot may have changed since then.
>> I did not have to specify a --kernel option, and the fs.py
>> (FSConfig.py) had all the options set to point toward the OpenBoot
>> binary files and example disk containing the OpenSolaris image. The
>> info message does not necessarily mean there is a problem, and if you
>> connect with m5term you might still reach the OpenBoot prompt anyway.
>>
>> When I was using the SPARC_FS only 1 core was supported. Maybe someone
>> has been developing this port but I am not aware. Unless you are
>> really committed to the SPARC architecture, I would suggest trying one
>> of the other CPU models.
>>
>> Good luck,
>> Gedare
>> _______________________________________________
>> gem5-users mailing list
>> gem5-***@gem5.org
>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>
> Also I should point out that I think image is for booting Solaris, I
> think, not Linux. I don't think we got SMP working for SPARC FS, but I
> don't really remember.
>
> Gabe
> _______________________________________________
> gem5-users mailing list
> gem5-***@gem5.org
> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
Xuhao Chen
2012-03-01 13:39:28 UTC
Permalink
Hi Gabe,
Sorry about my mistake.
Thank you for your help. I just used the "Linux 2.6.28.4 Kernel Image (x86_64)" from "http://www.cs.utexas.edu/~parsec_m5/", and it works. But when I compile applications (e.g. HelloWorld or PARSEC apps) on the host machine (with Ubuntu-10.04), and copy the binary into this image and run it, there is a segmentation fault. I then compile apps on Redhat E5 with gcc-4.1.2 and copy into the image, but there is still this error. Is it because the incompatible kernel version of the host (compile) and target (execute) platform? Thanks.




Xuhao Chen
PhD student
School of Computer
National University of Defense Technology
Changsha, Hunan, P.R.China, 410073
Tel: +86-159-741-03340 (Mobile)

From: Gabriel Michael Black
Date: 2012-03-01 19:47
To: gem5-users
Subject: Re: [gem5-users] How to boot SPARC FS?
First off, when you send emails to the list, you should make sure the
subject reflects what your asking about. Second, you shouldn't reply
to an unrelated email to ask your own question. You should start a new
thread specifically for your question.

As far as configuring things for more CPUs, you'd do that using
standard Linux kernel configuration mechanisms. In that tar ball are
regular Linux config files which you can copy into a kernel source
tree and use in the normal way. I'm sure there's lots of documentation
about that online somewhere. There are only two things you should be
careful about, getting the version of the kernel to match the config
(mostly to make life simpler for yourself) and also that exotic
options that use special hardware or platform features may not work.
Xuhao Chen
2012-03-03 15:45:59 UTC
Permalink
Hi everyone,
When I compile applications (e.g. HelloWorld or PARSEC apps) on the host machine (with Ubuntu-10.04), and copy the binary into the disk image (by mount and umount), then launch the x86 FS and run these apps, there is a segmentation fault shown in the m5term console.

(none) / # ./hello
hello[976]: segfault at c8 ip 00000000f7f43ac7 sp 00000000fff5551c error 6 in ld-2.6.1.so[f7f3c000+1a000]
Segmentation fault
(none) / #

I then compile apps on Redhat E5 with gcc-4.1.2 and copy into the image, but there is still this error. Is it because the incompatible kernel version of the host (compile) and target (execute) platform? Thanks.




Xuhao Chen
PhD student
School of Computer
National University of Defense Technology
Changsha, Hunan, P.R.China, 410073
Tel: +86-159-741-03340 (Mobile)
Nilay Vaish
2012-03-03 15:52:11 UTC
Permalink
On Sat, 3 Mar 2012, Xuhao Chen wrote:

> Hi everyone,
> When I compile applications (e.g. HelloWorld or PARSEC apps) on the host machine (with Ubuntu-10.04), and copy the binary into the disk image (by mount and umount), then launch the x86 FS and run these apps, there is a segmentation fault shown in the m5term console.
>
> (none) / # ./hello
> hello[976]: segfault at c8 ip 00000000f7f43ac7 sp 00000000fff5551c error 6 in ld-2.6.1.so[f7f3c000+1a000]
> Segmentation fault
> (none) / #
>
> I then compile apps on Redhat E5 with gcc-4.1.2 and copy into the image,
> but there is still this error. Is it because the incompatible kernel
> version of the host (compile) and target (execute) platform? Thanks.
>
>

It seems to me that you are linking with glibc that is not available on
the disk image.

--
Nilay
Xuhao Chen
2012-03-03 16:30:05 UTC
Permalink
I use the disk image in "x86-system.tar" from http://gem5.org/Download , in which I think gcc-4.1.2 has been installed. It may be due to the different version of glibc, right?




Xuhao Chen
PhD student
School of Computer
National University of Defense Technology
Changsha, Hunan, P.R.China, 410073
Tel: +86-159-741-03340 (Mobile)

From: Nilay Vaish
Date: 2012-03-03 23:52
To: gem5 users mailing list
Subject: Re: [gem5-users] booting 64-core in X86 FS mode
On Sat, 3 Mar 2012, Xuhao Chen wrote:

> Hi everyone,
> When I compile applications (e.g. HelloWorld or PARSEC apps) on the host machine (with Ubuntu-10.04), and copy the binary into the disk image (by mount and umount), then launch the x86 FS and run these apps, there is a segmentation fault shown in the m5term console.
>
> (none) / # ./hello
> hello[976]: segfault at c8 ip 00000000f7f43ac7 sp 00000000fff5551c error 6 in ld-2.6.1.so[f7f3c000+1a000]
> Segmentation fault
> (none) / #
>
> I then compile apps on Redhat E5 with gcc-4.1.2 and copy into the image,
> but there is still this error. Is it because the incompatible kernel
> version of the host (compile) and target (execute) platform? Thanks.
>
>

It seems to me that you are linking with glibc that is not available on
the disk image.

--
Nilay
_______________________________________________
gem5-users mailing list
gem5-***@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
Nilay Vaish
2012-03-03 16:39:22 UTC
Permalink
On Sun, 4 Mar 2012, Xuhao Chen wrote:

> I use the disk image in "x86-system.tar" from http://gem5.org/Download ,
> in which I think gcc-4.1.2 has been installed. It may be due to the
> different version of glibc, right?
>
>
>
>
> Xuhao Chen
> PhD student
> School of Computer
> National University of Defense Technology
> Changsha, Hunan, P.R.China, 410073
> Tel: +86-159-741-03340 (Mobile)
>
> From: Nilay Vaish
> Date: 2012-03-03 23:52
> To: gem5 users mailing list
> Subject: Re: [gem5-users] booting 64-core in X86 FS mode
> On Sat, 3 Mar 2012, Xuhao Chen wrote:
>
>> Hi everyone,
>> When I compile applications (e.g. HelloWorld or PARSEC apps) on the host machine (with Ubuntu-10.04), and copy the binary into the disk image (by mount and umount), then launch the x86 FS and run these apps, there is a segmentation fault shown in the m5term console.
>>
>> (none) / # ./hello
>> hello[976]: segfault at c8 ip 00000000f7f43ac7 sp 00000000fff5551c error 6 in ld-2.6.1.so[f7f3c000+1a000]
>> Segmentation fault
>> (none) / #
>>
>> I then compile apps on Redhat E5 with gcc-4.1.2 and copy into the image,
>> but there is still this error. Is it because the incompatible kernel
>> version of the host (compile) and target (execute) platform? Thanks.
>>
>>
>
> It seems to me that you are linking with glibc that is not available on
> the disk image.
>

It certainly seems like so library was not found on the disk image, this
could be due to different version, or just that the library locations are
different on the disk image and the host machine. Run the same program
using a VM, this experiment should be able to confirm whether the
hypothesis of being linked to wrong library is correct or not.

--
Nilay
Xuhao Chen
2012-03-04 14:51:41 UTC
Permalink
I compile the application with -static -static-libgcc options but there is still error:

(none) home # ./hello
hello[840]: segfault at 000000000000000c rip 000000000805cd10 rsp 00000000ffff3f34 error 4
Segmentation fault
(none) home #

If I want to create a disk image which is the same as my host platform, I should firstly create an empty image, and then copy all the files into this image, right? But the file system in the host platform is too large, I wonder how you prune it (such as the linux-x86.img in x86-system.tar) but still keep the same compiling and runtime environment. Thanks.

Best,



Xuhao Chen
PhD student
School of Computer
National University of Defense Technology
Changsha, Hunan, P.R.China, 410073
Tel: +86-159-741-03340 (Mobile)

From: Nilay Vaish
Date: 2012-03-04 00:39
To: gem5 users mailing list
Subject: Re: [gem5-users] booting 64-core in X86 FS mode


On Sun, 4 Mar 2012, Xuhao Chen wrote:

> I use the disk image in "x86-system.tar" from http://gem5.org/Download ,
> in which I think gcc-4.1.2 has been installed. It may be due to the
> different version of glibc, right?
>
> Xuhao Chen
> PhD student
> School of Computer
> National University of Defense Technology
> Changsha, Hunan, P.R.China, 410073
> Tel: +86-159-741-03340 (Mobile)
>
> From: Nilay Vaish
> Date: 2012-03-03 23:52
> To: gem5 users mailing list
> Subject: Re: [gem5-users] booting 64-core in X86 FS mode
> On Sat, 3 Mar 2012, Xuhao Chen wrote:
>
>> Hi everyone,
>> When I compile applications (e.g. HelloWorld or PARSEC apps) on the host machine (with Ubuntu-10.04), and copy the binary into the disk image (by mount and umount), then launch the x86 FS and run these apps, there is a segmentation fault shown in the m5term console.
>>
>> (none) / # ./hello
>> hello[976]: segfault at c8 ip 00000000f7f43ac7 sp 00000000fff5551c error 6 in ld-2.6.1.so[f7f3c000+1a000]
>> Segmentation fault
>> (none) / #
>>
>> I then compile apps on Redhat E5 with gcc-4.1.2 and copy into the image,
>> but there is still this error. Is it because the incompatible kernel
>> version of the host (compile) and target (execute) platform? Thanks.
>
> It seems to me that you are linking with glibc that is not available on
> the disk image.
>

It certainly seems like so library was not found on the disk image, this
could be due to different version, or just that the library locations are
different on the disk image and the host machine. Run the same program
using a VM, this experiment should be able to confirm whether the
hypothesis of being linked to wrong library is correct or not.

--
Nilay
_______________________________________________
gem5-users mailing list
gem5-***@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
Nilay Vaish
2012-03-04 15:17:46 UTC
Permalink
On Sun, 4 Mar 2012, Xuhao Chen wrote:

> I compile the application with -static -static-libgcc options but there is still error:
>
> (none) home # ./hello
> hello[840]: segfault at 000000000000000c rip 000000000805cd10 rsp 00000000ffff3f34 error 4
> Segmentation fault
> (none) home #
>
> If I want to create a disk image which is the same as my host platform,
> I should firstly create an empty image, and then copy all the files into
> this image, right? But the file system in the host platform is too
> large, I wonder how you prune it (such as the linux-x86.img in
> x86-system.tar) but still keep the same compiling and runtime
> environment. Thanks.
>

I think the disk image can be created using virtual machine environment.
Anyways, from the printed rip value, you can try to figure out which
instruction is causing the segmentation fault and why. Looking through the
trace of executed instructions should help.

--
Nilay
Xuhao Chen
2012-03-07 14:25:16 UTC
Permalink
Hi,
Do you mean Xen virtual machine?
Is there detailed tutorial for customizing a disk image,
I find the related document on gem5 wiki is very sketchy. Thanks.




Xuhao Chen
PhD student
School of Computer
National University of Defense Technology
Changsha, Hunan, P.R.China, 410073
Tel: +86-159-741-03340 (Mobile)

From: Nilay Vaish
Date: 2012-03-04 23:17
To: gem5 users mailing list
Subject: Re: [gem5-users] booting 64-core in X86 FS mode
On Sun, 4 Mar 2012, Xuhao Chen wrote:

> I compile the application with -static -static-libgcc options but there is still error:
>
> (none) home # ./hello
> hello[840]: segfault at 000000000000000c rip 000000000805cd10 rsp 00000000ffff3f34 error 4
> Segmentation fault
> (none) home #
>
> If I want to create a disk image which is the same as my host platform,
> I should firstly create an empty image, and then copy all the files into
> this image, right? But the file system in the host platform is too
> large, I wonder how you prune it (such as the linux-x86.img in
> x86-system.tar) but still keep the same compiling and runtime
> environment. Thanks.
>

I think the disk image can be created using virtual machine environment.
Anyways, from the printed rip value, you can try to figure out which
instruction is causing the segmentation fault and why. Looking through the
trace of executed instructions should help.

--
Nilay
_______________________________________________
gem5-users mailing list
gem5-***@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
Xuhao Chen
2012-03-18 01:25:35 UTC
Permalink
Hi,
I find that there is mismatch of the gem5 X86 FS disk image (where I run the benchmarks) with my host environment (where I compile the benchmarks).
So I want to customize a disk image just as my host machine.
I tried to do this following the instructions at http://gem5.org/Disk_images and <Running PARSEC 2.1 on M5>, but I failed.
Is there any detailed tutorial for customizing a disk image,
I find the related document on gem5 wiki is very sketchy. Thanks.

Best regards,



Xuhao Chen
PhD student
School of Computer
National University of Defense Technology
Changsha, Hunan, P.R.China, 410073
Tel: +86-159-741-03340 (Mobile)
Ali Saidi
2012-02-28 23:20:17 UTC
Permalink
On 28.02.2012 11:27, Gedare Bloom wrote:

> On Tue, Feb 28, 2012 at
11:15 AM, Xuhao Chen <***@gmail.com> wrote:
>
>> Hi Gedare, I have
read your blog "a week in m5". And I have a question about m5. The
Solaris 10 binary image for SPARC includes "reset.bin, q.bin,
openboot.bin" etc. I copy *.bin and nvram1 from
OpenSPARCT1_Arch.1.5/S10image/ to the /dist/m5/system/binaries/
directory. Also rename reset.bin, q.bin, and openboot.bin to
reset_new.bin, q_new.bin, and openboot_new.bin, which are the binaries
expected by the m5 SPARC_FS scripts. Then copy disk.s10hw2 from the
S10image/ directory to the /dist/m5/system/disks/ directory. But when
booting the system, it still cannot find the kernel file: info: No
kernel set for full system simulation. Assuming you know what you're
doing... Do you know the reason? Should I use the option --kernel to
specify the kernel file? But which one is the kernel file? By the way,
how many CPUs can gem5 run in SPARC FS mode? I know that it can be up to
64 CPUs in Alpha FS mode. Thanks.
>
> It has been awhile since I worked
with GEM5 so I cannot be certain;
> that post is nearly 2 years old so a
lot may have changed since then.
> I did not have to specify a --kernel
option, and the fs.py
> (FSConfig.py) had all the options set to point
toward the OpenBoot
> binary files and example disk containing the
OpenSolaris image. The
> info message does not necessarily mean there is
a problem, and if you
> connect with m5term you might still reach the
OpenBoot prompt anyway.
>
> When I was using the SPARC_FS only 1 core
was supported. Maybe someone
> has been developing this port but I am
not aware. Unless you are
> really committed to the SPARC architecture,
I would suggest trying one
> of the other CPU models.
>
> Good luck,
>
Gedare
>
>> ________________________________ Xuhao Chen PhD student
School of Computer National University of Defense Technology Changsha,
Hunan, P.R.China, 410073 Tel: +86-159-741-03340 (Mobile)
>
>
_______________________________________________
> gem5-users mailing
list
> gem5-***@gem5.org
>
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users

The message
printed it simply a message that you need to know what you're doing. In
SPARC/Solaris unlike other architectures we run the full bare-metal boot
loader which pulls the kernel off the disk image. For SPARC no kernel
should be specified on the command line. I imagine the system is booting
just fine. We only ever got Solaris/SPARC booting on a single core in
atomic mode. While some of the peripherals that needed to exist were
written for multi-cores it was never tested and debugged.

Ali
Xuhao Chen
2012-02-29 15:05:03 UTC
Permalink
Hi Ali,
Thank you very much.
The problem is that there is nothing printed out in m5term console.




Xuhao Chen
PhD student
School of Computer
National University of Defense Technology
Changsha, Hunan, P.R.China, 410073
Tel: +86-159-741-03340 (Mobile)

From: Ali Saidi
Date: 2012-02-29 07:20
To: gem5 users mailing list
Subject: Re: [gem5-users]How to boot SPARC FS?
On 28.02.2012 11:27, Gedare Bloom wrote:
On Tue, Feb 28, 2012 at 11:15 AM, Xuhao Chen <***@gmail.com> wrote:
Hi Gedare, I have read your blog "a week in m5". And I have a question about m5. The Solaris 10 binary image for SPARC includes "reset.bin, q.bin, openboot.bin" etc. I copy *.bin and nvram1 from OpenSPARCT1_Arch.1.5/S10image/ to the /dist/m5/system/binaries/ directory. Also rename reset.bin, q.bin, and openboot.bin to reset_new.bin, q_new.bin, and openboot_new.bin, which are the binaries expected by the m5 SPARC_FS scripts. Then copy disk.s10hw2 from the S10image/ directory to the /dist/m5/system/disks/ directory. But when booting the system, it still cannot find the kernel file: info: No kernel set for full system simulation. Assuming you know what you're doing... Do you know the reason? Should I use the option --kernel to specify the kernel file? But which one is the kernel file? By the way, how many CPUs can gem5 run in SPARC FS mode? I know that it can be up to 64 CPUs in Alpha FS mode. Thanks.It has been awhile since I worked with GEM5 so I cannot be certain;
that post is nearly 2 years old so a lot may have changed since then.
I did not have to specify a --kernel option, and the fs.py
(FSConfig.py) had all the options set to point toward the OpenBoot
binary files and example disk containing the OpenSolaris image. The
info message does not necessarily mean there is a problem, and if you
connect with m5term you might still reach the OpenBoot prompt anyway.

When I was using the SPARC_FS only 1 core was supported. Maybe someone
has been developing this port but I am not aware. Unless you are
really committed to the SPARC architecture, I would suggest trying one
of the other CPU models.

Good luck,
Gedare
________________________________ Xuhao Chen PhD student School of Computer National University of Defense Technology Changsha, Hunan, P.R.China, 410073 Tel: +86-159-741-03340 (Mobile)_______________________________________________
gem5-users mailing list
gem5-***@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users



The message printed it simply a message that you need to know what you're doing. In SPARC/Solaris unlike other architectures we run the full bare-metal boot loader which pulls the kernel off the disk image. For SPARC no kernel should be specified on the command line. I imagine the system is booting just fine. We only ever got Solaris/SPARC booting on a single core in atomic mode. While some of the peripherals that needed to exist were written for multi-cores it was never tested and debugged.Ali
Ali Saidi
2012-02-29 16:27:34 UTC
Permalink
Make sure you're attaching to the correct console.

Ali

On
29.02.2012 09:05, Xuhao Chen wrote:

> Hi Ali,
> Thank you very much.

> The problem is that there is nothing printed out in m5term console.

>
> -------------------------
>
> Xuhao Chen
>
> PhD student
>
>
School of Computer
>
> National University of Defense Technology
>
>
Changsha, Hunan, P.R.China, 410073
>
> Tel: +86-159-741-03340 (Mobile)

>
> FROM: Ali Saidi [1]
> DATE: 2012-02-29 07:20
> TO: gem5 users
mailing list [2]
> SUBJECT: Re: [gem5-users]How to boot SPARC FS?
>
>
On 28.02.2012 11:27, Gedare Bloom wrote:
>
>> On Tue, Feb 28, 2012 at
11:15 AM, Xuhao Chen <***@gmail.com> wrote:
>>
>>> Hi Gedare, I
have read your blog "a week in m5". And I have a question about m5. The
Solaris 10 binary image for SPARC includes "reset.bin, q.bin,
openboot.bin" etc. I copy *.bin and nvram1 from
OpenSPARCT1_Arch.1.5/S10image/ to the /dist/m5/system/binaries/
directory. Also rename reset.bin, q.bin, and openboot.bin to
reset_new.bin, q_new.bin, and openboot_new.bin, which are the binaries
expected by the m5 SPARC_FS scripts. Then copy disk.s10hw2 from the
S10image/ directory to the /dist/m5/system/disks/ directory. But when
booting the system, it still cannot find the kernel file: info: No
kernel set for full system simulation. Assuming you know what you're
doing... Do you know the reason? Should I use the option --kernel to
specify the kernel file? But which one is the kernel file? By the way,
how many CPUs can gem5 run in SPARC FS mode? I know that it can be up to
64 CPUs in Alpha FS mode. Thanks.
>>
>> It has been awhile since I
worked with GEM5 so I cannot be certain;
>> that post is nearly 2 years
old so a lot may have changed since then.
>> I did not have to specify a
--kernel option, and the fs.py
>> (FSConfig.py) had all the options set
to point toward the OpenBoot
>> binary files and example disk containing
the OpenSolaris image. The
>> info message does not necessarily mean
there is a problem, and if you
>> connect with m5term you might still
reach the OpenBoot prompt anyway.
>>
>> When I was using the SPARC_FS
only 1 core was supported. Maybe someone
>> has been developing this
port but I am not aware. Unless you are
>> really committed to the SPARC
architecture, I would suggest trying one
>> of the other CPU models.
>>

>> Good luck,
>> Gedare
>>
>>> ________________________________ Xuhao
Chen PhD student School of Computer National University of Defense
Technology Changsha, Hunan, P.R.China, 410073 Tel: +86-159-741-03340
(Mobile)
>>
>> _______________________________________________
>>
gem5-users mailing list
>> gem5-***@gem5.org
>>
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>
> The message
printed it simply a message that you need to know what you're doing.
>
In SPARC/Solaris unlike other architectures we run the full bare-metal
boot loader which pulls the kernel off the disk image.
> For SPARC no
kernel should be specified on the command line.
> I imagine the system
is booting just fine. We only ever got Solaris/SPARC booting on a single
core in atomic mode.
> While some of the peripherals that needed to
exist were written for multi-cores it was never tested and debugged.
>
Ali



Links:
------
[1] mailto:***@umich.edu
[2]
mailto:gem5-***@gem5.org
Xiao Guo
2012-02-28 19:38:57 UTC
Permalink
Is it possible to boot more than 4 cores in x86 FS?


On Feb 27, 2012, at 5:51 PM, Jason Clemons wrote:

> Gabe,
>
> Thanks for the patch. I did a hg pull and hg update this afternoon and have made some progress on booting. I found that with a custom built version of 2.6.28.4 with smp on, the patch and the linux image files (linux-bigswap2.img and linux-x86.img) from gem5.org I am able to boot multicore x86 using the default cpu type which I think is atomic.
>
> After working with Andrea, we noted that the boot problem for x86 happens with the PARSEC root image but not the linux-x86.img from gem5.org. This was tested using the custom kernel and the default cpu type.
>
> I am currently attempting to test the detailed cpu-type with 2 cpus. I will let you know of any errors I run into.
>
>
> Thanks,
>
> Jason Clemons
>
>
>
> Quoting Gabe Black <***@eecs.umich.edu>:
>
>> I looked into this more, and the code that sets up the memory hierarchy
>> was broken a little while ago. I have a patch that seems to fix it
>> (attached) but then it gets stuck a little later. I remember a while ago
>> someone had problems with atomic operations not being supported by the
>> memory system fully, but I don't remember the specifics. This might be
>> that problem too.
>>
>> Gabe
>>
>> On 02/24/12 02:45, Gabriel Michael Black wrote:
>>> Yes, I have. You're using O3 though, and that support is fairly new.
>>>
>>> Gabe
>>>
>>> Quoting "Ankita (Garg) Goel" <***@gmail.com>:
>>>
>>>> Hi,
>>>>
>>>> Has anyone been able to boot multiple cpus wth FS mode on x86 ? If
>>>> yes, it
>>>> would greatly help if you could share your environmental setup. The
>>>> issues
>>>> I am facing I have already posted in some previous emails.
>>>>
>>>> Thanks for your help!
>>>>
>>>> --
>>>> Regards,
>>>> Ankita
>>>>
>>>
Gabe Black
2012-03-03 04:53:28 UTC
Permalink
I dug into this some more, and it looks like it really is because the
memory system doesn't enforce locking a particular address. Specifically
here:

5049193421500: system.cpu0 T0 : @_spin_lock_irqsave+3.1 :
DEC_LOCKED_M : ldstl t1d, DS:[rdi] : MemRead : D=0x0000000000000000
A=0xffffffff80768400
5049193435000: system.cpu1 T0 : @generic_unplug_device+24.0 : MOV_R_M
: ld rax, DS:[rbx + 0x180] : MemRead : D=0xffffffff80768400
A=0xffff8100075f62c8
5049193436000: system.cpu1 T0 : @generic_unplug_device+31.0 : MOV_M_I
: limm t1d, 0x1 : IntAlu : D=0x0000000000000001
5049193421500: system.cpu0 T0 : @_spin_lock_irqsave+3.2 :
DEC_LOCKED_M : subi t1d, t1d, 0x1 : IntAlu : D=0x0000000000000094
5049193421500: system.cpu0 T0 : @_spin_lock_irqsave+3.3 :
DEC_LOCKED_M : stul t1d, DS:[rdi] : MemWrite : D=0x00000000ffffffff
A=0xffffffff80768400
5049193436000: system.cpu1 T0 : @generic_unplug_device+31.1 : MOV_M_I
: st t1d, DS:[rax] : MemWrite : D=0x0000000000000001 A=0xffffffff80768400


You can more or less see where accesses to a spin lock are getting
jumbled and the spin lock getting set to -1 because one CPU is
clobbering the other's store. Shortly after that point, CPU 1 seems to
go to sleep, and CPU 0 spins forever waiting for that -1 to become a 0.


\/\/\/\/
So if any of you memory system folks are reading this, the memory system
not respecting the lock request flags is a fairly serious problem that
really deserves some attention.
/\/\/\/\

Gabe



On 02/26/12 04:51, Gabe Black wrote:
> I looked into this more, and the code that sets up the memory hierarchy
> was broken a little while ago. I have a patch that seems to fix it
> (attached) but then it gets stuck a little later. I remember a while ago
> someone had problems with atomic operations not being supported by the
> memory system fully, but I don't remember the specifics. This might be
> that problem too.
>
> Gabe
>
> On 02/24/12 02:45, Gabriel Michael Black wrote:
>> Yes, I have. You're using O3 though, and that support is fairly new.
>>
>> Gabe
>>
>> Quoting "Ankita (Garg) Goel" <***@gmail.com>:
>>
>>> Hi,
>>>
>>> Has anyone been able to boot multiple cpus wth FS mode on x86 ? If
>>> yes, it
>>> would greatly help if you could share your environmental setup. The
>>> issues
>>> I am facing I have already posted in some previous emails.
>>>
>>> Thanks for your help!
>>>
>>> --
>>> Regards,
>>> Ankita
>>>
>>
>>
>> _______________________________________________
>> gem5-users mailing list
>> gem5-***@gem5.org
>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>
>
> _______________________________________________
> gem5-users mailing list
> gem5-***@gem5.org
> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
Ali Saidi
2012-03-03 16:44:18 UTC
Permalink
How does x86 implement locked transactions? Are they done via load locked/store conditionals?

Ali

Sent from my ARM powered mobile device

On Mar 2, 2012, at 10:53 PM, Gabe Black <***@eecs.umich.edu> wrote:

> I dug into this some more, and it looks like it really is because the memory system doesn't enforce locking a particular address. Specifically here:
>
> 5049193421500: system.cpu0 T0 : @_spin_lock_irqsave+3.1 : DEC_LOCKED_M : ldstl t1d, DS:[rdi] : MemRead : D=0x0000000000000000 A=0xffffffff80768400
> 5049193435000: system.cpu1 T0 : @generic_unplug_device+24.0 : MOV_R_M : ld rax, DS:[rbx + 0x180] : MemRead : D=0xffffffff80768400 A=0xffff8100075f62c8
> 5049193436000: system.cpu1 T0 : @generic_unplug_device+31.0 : MOV_M_I : limm t1d, 0x1 : IntAlu : D=0x0000000000000001
> 5049193421500: system.cpu0 T0 : @_spin_lock_irqsave+3.2 : DEC_LOCKED_M : subi t1d, t1d, 0x1 : IntAlu : D=0x0000000000000094
> 5049193421500: system.cpu0 T0 : @_spin_lock_irqsave+3.3 : DEC_LOCKED_M : stul t1d, DS:[rdi] : MemWrite : D=0x00000000ffffffff A=0xffffffff80768400
> 5049193436000: system.cpu1 T0 : @generic_unplug_device+31.1 : MOV_M_I : st t1d, DS:[rax] : MemWrite : D=0x0000000000000001 A=0xffffffff80768400
>
>
> You can more or less see where accesses to a spin lock are getting jumbled and the spin lock getting set to -1 because one CPU is clobbering the other's store. Shortly after that point, CPU 1 seems to go to sleep, and CPU 0 spins forever waiting for that -1 to become a 0.
>
>
> \/\/\/\/
> So if any of you memory system folks are reading this, the memory system not respecting the lock request flags is a fairly serious problem that really deserves some attention.
> /\/\/\/\
>
> Gabe
>
>
>
> On 02/26/12 04:51, Gabe Black wrote:
>>
>> I looked into this more, and the code that sets up the memory hierarchy
>> was broken a little while ago. I have a patch that seems to fix it
>> (attached) but then it gets stuck a little later. I remember a while ago
>> someone had problems with atomic operations not being supported by the
>> memory system fully, but I don't remember the specifics. This might be
>> that problem too.
>>
>> Gabe
>>
>> On 02/24/12 02:45, Gabriel Michael Black wrote:
>>> Yes, I have. You're using O3 though, and that support is fairly new.
>>>
>>> Gabe
>>>
>>> Quoting "Ankita (Garg) Goel" <***@gmail.com>:
>>>
>>>> Hi,
>>>>
>>>> Has anyone been able to boot multiple cpus wth FS mode on x86 ? If
>>>> yes, it
>>>> would greatly help if you could share your environmental setup. The
>>>> issues
>>>> I am facing I have already posted in some previous emails.
>>>>
>>>> Thanks for your help!
>>>>
>>>> --
>>>> Regards,
>>>> Ankita
>>>>
>>>
>>>
>>> _______________________________________________
>>> gem5-users mailing list
>>> gem5-***@gem5.org
>>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>>
>> _______________________________________________
>> gem5-users mailing list
>> gem5-***@gem5.org
>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>
> _______________________________________________
> gem5-users mailing list
> gem5-***@gem5.org
> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
Steve Reinhardt
2012-03-03 17:50:13 UTC
Permalink
They're locked RMW ops in microcode.

Steve

On Sat, Mar 3, 2012 at 8:44 AM, Ali Saidi <***@umich.edu> wrote:

> How does x86 implement locked transactions? Are they done via load
> locked/store conditionals?
>
> Ali
>
> Sent from my ARM powered mobile device
>
> On Mar 2, 2012, at 10:53 PM, Gabe Black <***@eecs.umich.edu> wrote:
>
> I dug into this some more, and it looks like it really is because the
> memory system doesn't enforce locking a particular address. Specifically
> here:
>
> 5049193421500: system.cpu0 T0 : @_spin_lock_irqsave+3.1 : DEC_LOCKED_M
> : ldstl t1d, DS:[rdi] : MemRead : D=0x0000000000000000
> A=0xffffffff80768400
> 5049193435000: system.cpu1 T0 : @generic_unplug_device+24.0 : MOV_R_M :
> ld rax, DS:[rbx + 0x180] : MemRead : D=0xffffffff80768400
> A=0xffff8100075f62c8
> 5049193436000: system.cpu1 T0 : @generic_unplug_device+31.0 : MOV_M_I :
> limm t1d, 0x1 : IntAlu : D=0x0000000000000001
> 5049193421500: system.cpu0 T0 : @_spin_lock_irqsave+3.2 : DEC_LOCKED_M
> : subi t1d, t1d, 0x1 : IntAlu : D=0x0000000000000094
> 5049193421500: system.cpu0 T0 : @_spin_lock_irqsave+3.3 : DEC_LOCKED_M
> : stul t1d, DS:[rdi] : MemWrite : D=0x00000000ffffffff
> A=0xffffffff80768400
> 5049193436000: system.cpu1 T0 : @generic_unplug_device+31.1 : MOV_M_I :
> st t1d, DS:[rax] : MemWrite : D=0x0000000000000001 A=0xffffffff80768400
>
>
> You can more or less see where accesses to a spin lock are getting jumbled
> and the spin lock getting set to -1 because one CPU is clobbering the
> other's store. Shortly after that point, CPU 1 seems to go to sleep, and
> CPU 0 spins forever waiting for that -1 to become a 0.
>
>
> \/\/\/\/
> So if any of you memory system folks are reading this, the memory system
> not respecting the lock request flags is a fairly serious problem that
> really deserves some attention.
> /\/\/\/\
>
> Gabe
>
>
>
> On 02/26/12 04:51, Gabe Black wrote:
>
> I looked into this more, and the code that sets up the memory hierarchy
> was broken a little while ago. I have a patch that seems to fix it
> (attached) but then it gets stuck a little later. I remember a while ago
> someone had problems with atomic operations not being supported by the
> memory system fully, but I don't remember the specifics. This might be
> that problem too.
>
> Gabe
>
> On 02/24/12 02:45, Gabriel Michael Black wrote:
>
> Yes, I have. You're using O3 though, and that support is fairly new.
>
> Gabe
>
> Quoting "Ankita (Garg) Goel" <***@gmail.com> <***@gmail.com>:
>
>
> Hi,
>
> Has anyone been able to boot multiple cpus wth FS mode on x86 ? If
> yes, it
> would greatly help if you could share your environmental setup. The
> issues
> I am facing I have already posted in some previous emails.
>
> Thanks for your help!
>
> --
> Regards,
> Ankita
>
>
> _______________________________________________
> gem5-users mailing listgem5-***@gem5.orghttp://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>
>
> _______________________________________________
> gem5-users mailing listgem5-***@gem5.orghttp://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>
>
> _______________________________________________
> gem5-users mailing list
> gem5-***@gem5.org
> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>
>
> _______________________________________________
> gem5-users mailing list
> gem5-***@gem5.org
> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>
Steve Reinhardt
2012-03-03 17:48:17 UTC
Permalink
On Fri, Mar 2, 2012 at 8:53 PM, Gabe Black <***@eecs.umich.edu> wrote:

> **
> \/\/\/\/
> So if any of you memory system folks are reading this, the memory system
> not respecting the lock request flags is a fairly serious problem that
> really deserves some attention.
> /\/\/\/\
>

This is a known (though apparently not well remembered) issue. See note 5:
http://gem5.org/Status_Matrix <http://gem5.org/Status_Matrix#x86>

and this thread:

http://www.mail-archive.com/gem5-***@gem5.org/msg00831.html

I think Nilay has taken care of note 3, is that correct? If so, Nilay,
would you do the honor of updating the wiki page?

Steve
Nilay Vaish
2012-03-07 14:18:09 UTC
Permalink
On Sat, 3 Mar 2012, Steve Reinhardt wrote:

> On Fri, Mar 2, 2012 at 8:53 PM, Gabe Black <***@eecs.umich.edu> wrote:
>
>> **
>> \/\/\/\/
>> So if any of you memory system folks are reading this, the memory system
>> not respecting the lock request flags is a fairly serious problem that
>> really deserves some attention.
>> /\/\/\/\
>>
>
> This is a known (though apparently not well remembered) issue. See note 5:
> http://gem5.org/Status_Matrix <http://gem5.org/Status_Matrix#x86>
>
> and this thread:
>
> http://www.mail-archive.com/gem5-***@gem5.org/msg00831.html
>
> I think Nilay has taken care of note 3, is that correct? If so, Nilay,
> would you do the honor of updating the wiki page?
>

Done!

--
Nilay
Jason Clemons
2012-02-28 20:20:42 UTC
Permalink
I was able to boot 8 cores using atomic but I am still working on multicore timing or detailed models. I currently get a error when init calls fork at the end of booting. I am trying a couple of patches from the website to see if they will help. 

Later,

Jason


Sent from my Samsung Epic™ 4G

Xiao Guo <***@gmail.com> wrote:

Is it possible to boot more than 4 cores in x86 FS?


On Feb 27, 2012, at 5:51 PM, Jason Clemons wrote:

> Gabe,
>
> Thanks for the patch.  I did a hg pull and hg update this afternoon and have made some progress on booting.  I found that with a custom built version of 2.6.28.4 with smp on, the patch and the linux image files (linux-bigswap2.img and linux-x86.img) from gem5.org I am able to boot multicore x86 using the default cpu type which I think is atomic.
>
> After working with Andrea, we noted that the boot problem for x86 happens with the PARSEC root image but not the linux-x86.img from gem5.org.  This was tested using the custom kernel and the default cpu type.
>
> I am currently attempting to test the detailed cpu-type with 2 cpus.  I will let you know of any errors I run into.
>
>
> Thanks,
>
> Jason Clemons
>
>
>
> Quoting Gabe Black <***@eecs.umich.edu>:
>
>> I looked into this more, and the code that sets up the memory hierarchy
>> was broken a little while ago. I have a patch that seems to fix it
>> (attached) but then it gets stuck a little later. I remember a while ago
>> someone had problems with atomic operations not being supported by the
>> memory system fully, but I don't remember the specifics. This might be
>> that problem too.
>>
>> Gabe
>>
>> On 02/24/12 02:45, Gabriel Michael Black wrote:
>>> Yes, I have. You're using O3 though, and that support is fairly new.
>>>
>>> Gabe
>>>
>>> Quoting "Ankita (Garg) Goel" <***@gmail.com>:
>>>
>>>> Hi,
>>>>
>>>> Has anyone been able to boot multiple cpus wth FS mode on x86 ? If
>>>> yes, it
>>>> would greatly help if you could share your environmental setup. The
>>>> issues
>>>> I am facing I have already posted in some previous emails.
>>>>
>>>> Thanks for your help!
>>>>
>>>> --
>>>> Regards,
>>>> Ankita
>>>>
>>>
Nilay Vaish
2012-02-28 20:37:47 UTC
Permalink
I have booted a 64 core system using timing simple cpu.

--
Nilay


On Tue, 28 Feb 2012, Jason Clemons wrote:

> I was able to boot 8 cores using atomic but I am still working on multicore timing or detailed models. I currently get a error when init calls fork at the end of booting. I am trying a couple of patches from the website to see if they will help. 

Later,

Jason


Sent from my Samsung Epic™ 4G

Xiao Guo <***@gmail.com> wrote:

Is it possible to boot more than 4 cores in x86 FS?


On Feb 27, 2012, at 5:51 PM, Jason Clemons wrote:

> Gabe,
>
> Thanks for the patch.  I did a hg pull and hg update this afternoon and have made some progress on booting.  I found that with a custom built version of 2.6.28.4 with smp on, the patch and the linux image files (linux-bigswap2.img and linux-x86.img) from gem5.org I am able to boot multicore x86 using the default cpu type which I think is atomic.
>
> After working with Andrea, we noted that the boot problem for x86 happens with the PARSEC root image but not the linux-x86.img from gem5.org.  This was tested using the custom kernel and the default cpu type.
>
> I am currently attempting to test the detailed cpu-type with 2 cpus.  I will let you know of any errors I run into.
>
>
> Thanks,
>
> Jason Clemons
>
>
>
> Quoting Gabe Black <***@eecs.umich.edu>:
>
>> I looked into this more, and the code that sets up the memory hierarchy
>> was broken a little while ago. I have a patch that seems to fix it
>> (attached) but then it gets stuck a little later. I remember a while ago
>> someone had problems with atomic operations not being supported by the
>> memory system fully, but I don't remember the specifics. This might be
>> that problem too.
>>
>> Gabe
>>
>> On 02/24/12 02:45, Gabriel Michael Black wrote:
>>> Yes, I have. You're using O3 though, and that support is fairly new.
>>>
>>> Gabe
>>>
>>> Quoting "Ankita (Garg) Goel" <***@gmail.com>:
>>>
>>>> Hi,
>>>>
>>>> Has anyone been able to boot multiple cpus wth FS mode on x86 ? If
>>>> yes, it
>>>> would greatly help if you could share your environmental setup. The
>>>> issues
>>>> I am facing I have already posted in some previous emails.
>>>>
>>>> Thanks for your help!
>>>>
>>>> --
>>>> Regards,
>>>> Ankita
>>>>
>>>
Ankita (Garg) Goel
2012-02-28 23:31:14 UTC
Permalink
I am facing the same issue as Jason mentioned with detailed CPUs. The
system boot-up seems to be going fine, but seems stuck at running INIT,
even after 2-3 hrs.

On Tue, Feb 28, 2012 at 2:37 PM, Nilay Vaish <***@cs.wisc.edu> wrote:

> I have booted a 64 core system using timing simple cpu.
>
> --
> Nilay
>
>
>
> On Tue, 28 Feb 2012, Jason Clemons wrote:
>
> I was able to boot 8 cores using atomic but I am still working on
>> multicore timing or detailed models. I currently get a error when init
>> calls fork at the end of booting. I am trying a couple of patches from the
>> website to see if they will help.
>>
>
> Later,
>
> Jason
>
>
> Sent from my Samsung Epic™ 4G
>
> Xiao Guo <***@gmail.com> wrote:
>
> Is it possible to boot more than 4 cores in x86 FS?
>
>
> On Feb 27, 2012, at 5:51 PM, Jason Clemons wrote:
>
> Gabe,
>>
>> Thanks for the patch. I did a hg pull and hg update this afternoon and
>> have made some progress on booting. I found that with a custom built
>> version of 2.6.28.4 with smp on, the patch and the linux image files
>> (linux-bigswap2.img and linux-x86.img) from gem5.org I am able to boot
>> multicore x86 using the default cpu type which I think is atomic.
>>
>> After working with Andrea, we noted that the boot problem for x86 happens
>> with the PARSEC root image but not the linux-x86.img from gem5.org.
>> This was tested using the custom kernel and the default cpu type.
>>
>> I am currently attempting to test the detailed cpu-type with 2 cpus. I
>> will let you know of any errors I run into.
>>
>>
>> Thanks,
>>
>> Jason Clemons
>>
>>
>>
>> Quoting Gabe Black <***@eecs.umich.edu>:
>>
>> I looked into this more, and the code that sets up the memory hierarchy
>>> was broken a little while ago. I have a patch that seems to fix it
>>> (attached) but then it gets stuck a little later. I remember a while ago
>>> someone had problems with atomic operations not being supported by the
>>> memory system fully, but I don't remember the specifics. This might be
>>> that problem too.
>>>
>>> Gabe
>>>
>>> On 02/24/12 02:45, Gabriel Michael Black wrote:
>>>
>>>> Yes, I have. You're using O3 though, and that support is fairly new.
>>>>
>>>> Gabe
>>>>
>>>> Quoting "Ankita (Garg) Goel" <***@gmail.com>:
>>>>
>>>> Hi,
>>>>>
>>>>> Has anyone been able to boot multiple cpus wth FS mode on x86 ? If
>>>>> yes, it
>>>>> would greatly help if you could share your environmental setup. The
>>>>> issues
>>>>> I am facing I have already posted in some previous emails.
>>>>>
>>>>> Thanks for your help!
>>>>>
>>>>> --
>>>>> Regards,
>>>>> Ankita
>>>>>
>>>>>
>>>>
> ______________________________**_________________
> gem5-users mailing list
> gem5-***@gem5.org
> http://m5sim.org/cgi-bin/**mailman/listinfo/gem5-users<http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users>
>
> _______________________________________________
> gem5-users mailing list
> gem5-***@gem5.org
> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>



--
Regards,
Ankita
Gabriel Michael Black
2012-02-29 01:44:54 UTC
Permalink
I saw this problem as well. I don't know what's causing it. If you're
looking for ways to debug it, I'd suggest figuring out what the CPU is
doing while it's hung. Is it just waiting idle? Is it stuck in a loop?
Is it stuck at one point in time and not progressing the simulation
(unlikely)? Then try to find out what it's waiting for, if anything,
and why that hasn't happened or the CPU hasn't noticed that it happened.

Gabe

Quoting "Ankita (Garg) Goel" <***@gmail.com>:

> I am facing the same issue as Jason mentioned with detailed CPUs. The
> system boot-up seems to be going fine, but seems stuck at running INIT,
> even after 2-3 hrs.
>
> On Tue, Feb 28, 2012 at 2:37 PM, Nilay Vaish <***@cs.wisc.edu> wrote:
>
>> I have booted a 64 core system using timing simple cpu.
>>
>> --
>> Nilay
>>
>>
>>
>> On Tue, 28 Feb 2012, Jason Clemons wrote:
>>
>> I was able to boot 8 cores using atomic but I am still working on
>>> multicore timing or detailed models. I currently get a error when init
>>> calls fork at the end of booting. I am trying a couple of patches from the
>>> website to see if they will help.
>>>
>>
>> Later,
>>
>> Jason
>>
>>
>> Sent from my Samsung Epic? 4G
>>
>> Xiao Guo <***@gmail.com> wrote:
>>
>> Is it possible to boot more than 4 cores in x86 FS?
>>
>>
>> On Feb 27, 2012, at 5:51 PM, Jason Clemons wrote:
>>
>> Gabe,
>>>
>>> Thanks for the patch. I did a hg pull and hg update this afternoon and
>>> have made some progress on booting. I found that with a custom built
>>> version of 2.6.28.4 with smp on, the patch and the linux image files
>>> (linux-bigswap2.img and linux-x86.img) from gem5.org I am able to boot
>>> multicore x86 using the default cpu type which I think is atomic.
>>>
>>> After working with Andrea, we noted that the boot problem for x86 happens
>>> with the PARSEC root image but not the linux-x86.img from gem5.org.
>>> This was tested using the custom kernel and the default cpu type.
>>>
>>> I am currently attempting to test the detailed cpu-type with 2 cpus. I
>>> will let you know of any errors I run into.
>>>
>>>
>>> Thanks,
>>>
>>> Jason Clemons
>>>
>>>
>>>
>>> Quoting Gabe Black <***@eecs.umich.edu>:
>>>
>>> I looked into this more, and the code that sets up the memory hierarchy
>>>> was broken a little while ago. I have a patch that seems to fix it
>>>> (attached) but then it gets stuck a little later. I remember a while ago
>>>> someone had problems with atomic operations not being supported by the
>>>> memory system fully, but I don't remember the specifics. This might be
>>>> that problem too.
>>>>
>>>> Gabe
>>>>
>>>> On 02/24/12 02:45, Gabriel Michael Black wrote:
>>>>
>>>>> Yes, I have. You're using O3 though, and that support is fairly new.
>>>>>
>>>>> Gabe
>>>>>
>>>>> Quoting "Ankita (Garg) Goel" <***@gmail.com>:
>>>>>
>>>>> Hi,
>>>>>>
>>>>>> Has anyone been able to boot multiple cpus wth FS mode on x86 ? If
>>>>>> yes, it
>>>>>> would greatly help if you could share your environmental setup. The
>>>>>> issues
>>>>>> I am facing I have already posted in some previous emails.
>>>>>>
>>>>>> Thanks for your help!
>>>>>>
>>>>>> --
>>>>>> Regards,
>>>>>> Ankita
>>>>>>
>>>>>>
>>>>>
>> ______________________________**_________________
>> gem5-users mailing list
>> gem5-***@gem5.org
>> http://m5sim.org/cgi-bin/**mailman/listinfo/gem5-users<http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users>
>>
>> _______________________________________________
>> gem5-users mailing list
>> gem5-***@gem5.org
>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>>
>
>
>
> --
> Regards,
> Ankita
>
Xiao Guo
2012-03-02 04:41:51 UTC
Permalink
I got kernel panic when booting 4 cores using linux kernel 2.6.27, does anybody have a solution for this? Thanks!

> Root-NFS: No NFS server available, giving up.
> VFS: Unable to mount root fs via NFS, trying floppy.
> VFS: Insert root floppy and press ENTER
> VFS: Cannot open root device "<NULL>" or unknown-block(2,0)
> Please append a correct "root=" boot option; here are the available partitions:
> Kernel panic - not syncing: VFS: Unable to mount root fs on unknown-block(2,0)

xiao


On Feb 28, 2012, at 8:44 PM, Gabriel Michael Black wrote:

> I saw this problem as well. I don't know what's causing it. If you're
> looking for ways to debug it, I'd suggest figuring out what the CPU is
> doing while it's hung. Is it just waiting idle? Is it stuck in a loop?
> Is it stuck at one point in time and not progressing the simulation
> (unlikely)? Then try to find out what it's waiting for, if anything,
> and why that hasn't happened or the CPU hasn't noticed that it happened.
>
> Gabe
>
> Quoting "Ankita (Garg) Goel" <***@gmail.com>:
>
>> I am facing the same issue as Jason mentioned with detailed CPUs. The
>> system boot-up seems to be going fine, but seems stuck at running INIT,
>> even after 2-3 hrs.
>>
>> On Tue, Feb 28, 2012 at 2:37 PM, Nilay Vaish <***@cs.wisc.edu> wrote:
>>
>>> I have booted a 64 core system using timing simple cpu.
>>>
>>> --
>>> Nilay
>>>
>>>
>>>
>>> On Tue, 28 Feb 2012, Jason Clemons wrote:
>>>
>>> I was able to boot 8 cores using atomic but I am still working on
>>>> multicore timing or detailed models. I currently get a error when init
>>>> calls fork at the end of booting. I am trying a couple of patches from the
>>>> website to see if they will help.
>>>>
>>>
>>> Later,
>>>
>>> Jason
>>>
>>>
>>> Sent from my Samsung Epic? 4G
>>>
>>> Xiao Guo <***@gmail.com> wrote:
>>>
>>> Is it possible to boot more than 4 cores in x86 FS?
>>>
>>>
>>> On Feb 27, 2012, at 5:51 PM, Jason Clemons wrote:
>>>
>>> Gabe,
>>>>
>>>> Thanks for the patch. I did a hg pull and hg update this afternoon and
>>>> have made some progress on booting. I found that with a custom built
>>>> version of 2.6.28.4 with smp on, the patch and the linux image files
>>>> (linux-bigswap2.img and linux-x86.img) from gem5.org I am able to boot
>>>> multicore x86 using the default cpu type which I think is atomic.
>>>>
>>>> After working with Andrea, we noted that the boot problem for x86 happens
>>>> with the PARSEC root image but not the linux-x86.img from gem5.org.
>>>> This was tested using the custom kernel and the default cpu type.
>>>>
>>>> I am currently attempting to test the detailed cpu-type with 2 cpus. I
>>>> will let you know of any errors I run into.
>>>>
>>>>
>>>> Thanks,
>>>>
>>>> Jason Clemons
>>>>
>>>>
>>>>
>>>> Quoting Gabe Black <***@eecs.umich.edu>:
>>>>
>>>> I looked into this more, and the code that sets up the memory hierarchy
>>>>> was broken a little while ago. I have a patch that seems to fix it
>>>>> (attached) but then it gets stuck a little later. I remember a while ago
>>>>> someone had problems with atomic operations not being supported by the
>>>>> memory system fully, but I don't remember the specifics. This might be
>>>>> that problem too.
>>>>>
>>>>> Gabe
>>>>>
>>>>> On 02/24/12 02:45, Gabriel Michael Black wrote:
>>>>>
>>>>>> Yes, I have. You're using O3 though, and that support is fairly new.
>>>>>>
>>>>>> Gabe
>>>>>>
>>>>>> Quoting "Ankita (Garg) Goel" <***@gmail.com>:
>>>>>>
>>>>>> Hi,
>>>>>>>
>>>>>>> Has anyone been able to boot multiple cpus wth FS mode on x86 ? If
>>>>>>> yes, it
>>>>>>> would greatly help if you could share your environmental setup. The
>>>>>>> issues
>>>>>>> I am facing I have already posted in some previous emails.
>>>>>>>
>>>>>>> Thanks for your help!
>>>>>>>
>>>>>>> --
>>>>>>> Regards,
>>>>>>> Ankita
>>>>>>>
>>>>>>>
>>>>>>
>>> ______________________________**_________________
>>> gem5-users mailing list
>>> gem5-***@gem5.org
>>> http://m5sim.org/cgi-bin/**mailman/listinfo/gem5-users<http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users>
>>>
>>> _______________________________________________
>>> gem5-users mailing list
>>> gem5-***@gem5.org
>>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>>>
>>
>>
>>
>> --
>> Regards,
>> Ankita
>>
>
>
> _______________________________________________
> gem5-users mailing list
> gem5-***@gem5.org
> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
Ankita (Garg) Goel
2012-03-02 07:07:21 UTC
Permalink
Hi Xiao,

What config file did you use to build the kernel ? The error probably could
be that the fs drivers like ext3 or ext4 were not built into the kernel so
the root filesystem in the image provided could not be mounted.

There are a few sample config files provided in the X86 FS download section
on the wiki.

Regards,
Ankita

On Thu, Mar 1, 2012 at 10:41 PM, Xiao Guo <***@gmail.com> wrote:

> I got kernel panic when booting 4 cores using linux kernel 2.6.27, does
> anybody have a solution for this? Thanks!
>
> > Root-NFS: No NFS server available, giving up.
> > VFS: Unable to mount root fs via NFS, trying floppy.
> > VFS: Insert root floppy and press ENTER
> > VFS: Cannot open root device "<NULL>" or unknown-block(2,0)
> > Please append a correct "root=" boot option; here are the available
> partitions:
> > Kernel panic - not syncing: VFS: Unable to mount root fs on
> unknown-block(2,0)
>
> xiao
>
>
> On Feb 28, 2012, at 8:44 PM, Gabriel Michael Black wrote:
>
> > I saw this problem as well. I don't know what's causing it. If you're
> > looking for ways to debug it, I'd suggest figuring out what the CPU is
> > doing while it's hung. Is it just waiting idle? Is it stuck in a loop?
> > Is it stuck at one point in time and not progressing the simulation
> > (unlikely)? Then try to find out what it's waiting for, if anything,
> > and why that hasn't happened or the CPU hasn't noticed that it happened.
> >
> > Gabe
> >
> > Quoting "Ankita (Garg) Goel" <***@gmail.com>:
> >
> >> I am facing the same issue as Jason mentioned with detailed CPUs. The
> >> system boot-up seems to be going fine, but seems stuck at running INIT,
> >> even after 2-3 hrs.
> >>
> >> On Tue, Feb 28, 2012 at 2:37 PM, Nilay Vaish <***@cs.wisc.edu> wrote:
> >>
> >>> I have booted a 64 core system using timing simple cpu.
> >>>
> >>> --
> >>> Nilay
> >>>
> >>>
> >>>
> >>> On Tue, 28 Feb 2012, Jason Clemons wrote:
> >>>
> >>> I was able to boot 8 cores using atomic but I am still working on
> >>>> multicore timing or detailed models. I currently get a error when init
> >>>> calls fork at the end of booting. I am trying a couple of patches
> from the
> >>>> website to see if they will help.
> >>>>
> >>>
> >>> Later,
> >>>
> >>> Jason
> >>>
> >>>
> >>> Sent from my Samsung Epic? 4G
> >>>
> >>> Xiao Guo <***@gmail.com> wrote:
> >>>
> >>> Is it possible to boot more than 4 cores in x86 FS?
> >>>
> >>>
> >>> On Feb 27, 2012, at 5:51 PM, Jason Clemons wrote:
> >>>
> >>> Gabe,
> >>>>
> >>>> Thanks for the patch. I did a hg pull and hg update this afternoon
> and
> >>>> have made some progress on booting. I found that with a custom built
> >>>> version of 2.6.28.4 with smp on, the patch and the linux image files
> >>>> (linux-bigswap2.img and linux-x86.img) from gem5.org I am able to
> boot
> >>>> multicore x86 using the default cpu type which I think is atomic.
> >>>>
> >>>> After working with Andrea, we noted that the boot problem for x86
> happens
> >>>> with the PARSEC root image but not the linux-x86.img from gem5.org.
> >>>> This was tested using the custom kernel and the default cpu type.
> >>>>
> >>>> I am currently attempting to test the detailed cpu-type with 2 cpus.
> I
> >>>> will let you know of any errors I run into.
> >>>>
> >>>>
> >>>> Thanks,
> >>>>
> >>>> Jason Clemons
> >>>>
> >>>>
> >>>>
> >>>> Quoting Gabe Black <***@eecs.umich.edu>:
> >>>>
> >>>> I looked into this more, and the code that sets up the memory
> hierarchy
> >>>>> was broken a little while ago. I have a patch that seems to fix it
> >>>>> (attached) but then it gets stuck a little later. I remember a while
> ago
> >>>>> someone had problems with atomic operations not being supported by
> the
> >>>>> memory system fully, but I don't remember the specifics. This might
> be
> >>>>> that problem too.
> >>>>>
> >>>>> Gabe
> >>>>>
> >>>>> On 02/24/12 02:45, Gabriel Michael Black wrote:
> >>>>>
> >>>>>> Yes, I have. You're using O3 though, and that support is fairly new.
> >>>>>>
> >>>>>> Gabe
> >>>>>>
> >>>>>> Quoting "Ankita (Garg) Goel" <***@gmail.com>:
> >>>>>>
> >>>>>> Hi,
> >>>>>>>
> >>>>>>> Has anyone been able to boot multiple cpus wth FS mode on x86 ? If
> >>>>>>> yes, it
> >>>>>>> would greatly help if you could share your environmental setup. The
> >>>>>>> issues
> >>>>>>> I am facing I have already posted in some previous emails.
> >>>>>>>
> >>>>>>> Thanks for your help!
> >>>>>>>
> >>>>>>> --
> >>>>>>> Regards,
> >>>>>>> Ankita
> >>>>>>>
> >>>>>>>
> >>>>>>
> >>> ______________________________**_________________
> >>> gem5-users mailing list
> >>> gem5-***@gem5.org
> >>> http://m5sim.org/cgi-bin/**mailman/listinfo/gem5-users<
> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users>
> >>>
> >>> _______________________________________________
> >>> gem5-users mailing list
> >>> gem5-***@gem5.org
> >>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
> >>>
> >>
> >>
> >>
> >> --
> >> Regards,
> >> Ankita
> >>
> >
> >
> > _______________________________________________
> > gem5-users mailing list
> > gem5-***@gem5.org
> > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>
>
>
>
> _______________________________________________
> gem5-users mailing list
> gem5-***@gem5.org
> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>



--
Regards,
Ankita
Graduate Student
Department of Computer Science
University of Texas at Austin
Andrea Pellegrini
2012-03-02 16:39:02 UTC
Permalink
I can boot the O3 model multicore, but the system is highly unstable and
often it gets stuck or crashes. I attached a couple of logs from my
experiments.
The single core detailed model seems to work well.

-Andrea

--------------------------------------------------------
x86_64-vmlinux-2.6.28.4.smp


Script from M5 readfile is empty, starting bash shell...
(none) / # cd parsec/install
(none) install # /etc/init.d/rcS: line 19: rcS[965]: segfault at
ffffffff8020b7c0 ip ffffffff8020b7d5 sp 00007fff47048b78 error 15 9
70 Segmentation fault /bin/bash

Enter runle


--------------------------------------------------------
x86_64-vmlinux-2.6.22.9.smp


Unable to handle kernel NULL pointer dereference at 00000000000000ac RIP:
[<ffffffff80596900>] _spin_unlock_irqrestore+0x0/0x10
PGD 0
Oops: 0002 [1] SMP
CPU 0
Modules linked in:
Pid: 0, comm: swapper Not tainted 2.6.22.9 #12
RIP: e200:[<ffffffff80596900>] [<ffffffff80596900>]
_spin_unlock_irqrestore+0x0/0x10
RSP: 6633:ffffffff8023e138 EFLAGS: 000001a8
RAX: ffff81000121d3c0 RBX: 65fb8948078b4853 RCX: ffff810001369150
RDX: 00000000000002a8 RSI: ffff81000121d3d8 RDI: 00000000000000ac
RBP: 0000000000000001 R08: ffffffff80245691 R09: ffffffff80225b53
R10: ffffffff807beed0 R11: 0000000000000003 R12: 480000002425348b
R13: e838788b48fce083 R14: 0000000300000000 R15: ffffffff802254f7
FS: 00002b639e267dd0(0000) GS:ffffffff80757000(0000) knlGS:0000000000000000
CS: 0010 DS: 0018 ES: 180018 CR0: 000000008005003b
CR2: 00000000000000ac CR3: 0000000000201000 CR4: 00000000000006e0
Process swapper (pid: 0, threadinfo ffffffff80770000, task ffffffff806f13c0)
Stack: 48de8948fffff934 2e66668beb5bc789 0000000000841f0f 2eba0ff0f3894853
74c085d231c01900 468d48c3d0895b04 65217508463b4808 e80000002425348b
48de8948fffff8f4 5bffffff49e8c789 c3d08900000001ba 2e666666feeb0b0f
Call Trace:


Code: c7 07 01 00 00 00 56 9d c3 0f 1f 80 00 00 00 00 c7 07 01 00
RIP [<ffffffff80596900>] _spin_unlock_irqrestore+0x0/0x10
RSP <ffffffff8023e138>
CR2: 00000000000000ac
Kernel panic - not syncing: Aiee, killing interrupt handler!


---------------------------------------------------------


On Fri, Mar 2, 2012 at 2:07 AM, Ankita (Garg) Goel <***@gmail.com>wrote:

> Hi Xiao,
>
> What config file did you use to build the kernel ? The error probably
> could be that the fs drivers like ext3 or ext4 were not built into the
> kernel so the root filesystem in the image provided could not be mounted.
>
> There are a few sample config files provided in the X86 FS download
> section on the wiki.
>
> Regards,
> Ankita
>
>
> On Thu, Mar 1, 2012 at 10:41 PM, Xiao Guo <***@gmail.com> wrote:
>
>> I got kernel panic when booting 4 cores using linux kernel 2.6.27, does
>> anybody have a solution for this? Thanks!
>>
>> > Root-NFS: No NFS server available, giving up.
>> > VFS: Unable to mount root fs via NFS, trying floppy.
>> > VFS: Insert root floppy and press ENTER
>> > VFS: Cannot open root device "<NULL>" or unknown-block(2,0)
>> > Please append a correct "root=" boot option; here are the available
>> partitions:
>> > Kernel panic - not syncing: VFS: Unable to mount root fs on
>> unknown-block(2,0)
>>
>> xiao
>>
>>
>> On Feb 28, 2012, at 8:44 PM, Gabriel Michael Black wrote:
>>
>> > I saw this problem as well. I don't know what's causing it. If you're
>> > looking for ways to debug it, I'd suggest figuring out what the CPU is
>> > doing while it's hung. Is it just waiting idle? Is it stuck in a loop?
>> > Is it stuck at one point in time and not progressing the simulation
>> > (unlikely)? Then try to find out what it's waiting for, if anything,
>> > and why that hasn't happened or the CPU hasn't noticed that it happened.
>> >
>> > Gabe
>> >
>> > Quoting "Ankita (Garg) Goel" <***@gmail.com>:
>> >
>> >> I am facing the same issue as Jason mentioned with detailed CPUs. The
>> >> system boot-up seems to be going fine, but seems stuck at running INIT,
>> >> even after 2-3 hrs.
>> >>
>> >> On Tue, Feb 28, 2012 at 2:37 PM, Nilay Vaish <***@cs.wisc.edu>
>> wrote:
>> >>
>> >>> I have booted a 64 core system using timing simple cpu.
>> >>>
>> >>> --
>> >>> Nilay
>> >>>
>> >>>
>> >>>
>> >>> On Tue, 28 Feb 2012, Jason Clemons wrote:
>> >>>
>> >>> I was able to boot 8 cores using atomic but I am still working on
>> >>>> multicore timing or detailed models. I currently get a error when
>> init
>> >>>> calls fork at the end of booting. I am trying a couple of patches
>> from the
>> >>>> website to see if they will help.
>> >>>>
>> >>>
>> >>> Later,
>> >>>
>> >>> Jason
>> >>>
>> >>>
>> >>> Sent from my Samsung Epic? 4G
>> >>>
>> >>> Xiao Guo <***@gmail.com> wrote:
>> >>>
>> >>> Is it possible to boot more than 4 cores in x86 FS?
>> >>>
>> >>>
>> >>> On Feb 27, 2012, at 5:51 PM, Jason Clemons wrote:
>> >>>
>> >>> Gabe,
>> >>>>
>> >>>> Thanks for the patch. I did a hg pull and hg update this afternoon
>> and
>> >>>> have made some progress on booting. I found that with a custom built
>> >>>> version of 2.6.28.4 with smp on, the patch and the linux image files
>> >>>> (linux-bigswap2.img and linux-x86.img) from gem5.org I am able to
>> boot
>> >>>> multicore x86 using the default cpu type which I think is atomic.
>> >>>>
>> >>>> After working with Andrea, we noted that the boot problem for x86
>> happens
>> >>>> with the PARSEC root image but not the linux-x86.img from gem5.org.
>> >>>> This was tested using the custom kernel and the default cpu type.
>> >>>>
>> >>>> I am currently attempting to test the detailed cpu-type with 2 cpus.
>> I
>> >>>> will let you know of any errors I run into.
>> >>>>
>> >>>>
>> >>>> Thanks,
>> >>>>
>> >>>> Jason Clemons
>> >>>>
>> >>>>
>> >>>>
>> >>>> Quoting Gabe Black <***@eecs.umich.edu>:
>> >>>>
>> >>>> I looked into this more, and the code that sets up the memory
>> hierarchy
>> >>>>> was broken a little while ago. I have a patch that seems to fix it
>> >>>>> (attached) but then it gets stuck a little later. I remember a
>> while ago
>> >>>>> someone had problems with atomic operations not being supported by
>> the
>> >>>>> memory system fully, but I don't remember the specifics. This might
>> be
>> >>>>> that problem too.
>> >>>>>
>> >>>>> Gabe
>> >>>>>
>> >>>>> On 02/24/12 02:45, Gabriel Michael Black wrote:
>> >>>>>
>> >>>>>> Yes, I have. You're using O3 though, and that support is fairly
>> new.
>> >>>>>>
>> >>>>>> Gabe
>> >>>>>>
>> >>>>>> Quoting "Ankita (Garg) Goel" <***@gmail.com>:
>> >>>>>>
>> >>>>>> Hi,
>> >>>>>>>
>> >>>>>>> Has anyone been able to boot multiple cpus wth FS mode on x86 ? If
>> >>>>>>> yes, it
>> >>>>>>> would greatly help if you could share your environmental setup.
>> The
>> >>>>>>> issues
>> >>>>>>> I am facing I have already posted in some previous emails.
>> >>>>>>>
>> >>>>>>> Thanks for your help!
>> >>>>>>>
>> >>>>>>> --
>> >>>>>>> Regards,
>> >>>>>>> Ankita
>> >>>>>>>
>> >>>>>>>
>> >>>>>>
>> >>> ______________________________**_________________
>> >>> gem5-users mailing list
>> >>> gem5-***@gem5.org
>> >>> http://m5sim.org/cgi-bin/**mailman/listinfo/gem5-users<
>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users>
>> >>>
>> >>> _______________________________________________
>> >>> gem5-users mailing list
>> >>> gem5-***@gem5.org
>> >>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>> >>>
>> >>
>> >>
>> >>
>> >> --
>> >> Regards,
>> >> Ankita
>> >>
>> >
>> >
>> > _______________________________________________
>> > gem5-users mailing list
>> > gem5-***@gem5.org
>> > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>>
>>
>>
>>
>> _______________________________________________
>> gem5-users mailing list
>> gem5-***@gem5.org
>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>>
>
>
>
> --
> Regards,
> Ankita
> Graduate Student
> Department of Computer Science
> University of Texas at Austin
>
>
>
> _______________________________________________
> gem5-users mailing list
> gem5-***@gem5.org
> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>
Nilay Vaish
2012-03-02 19:18:42 UTC
Permalink
Which gem5 repo are you using? stable or dev? I suggest using dev.

--
Nilay

On Fri, 2 Mar 2012, Andrea Pellegrini wrote:

> I can boot the O3 model multicore, but the system is highly unstable and
> often it gets stuck or crashes. I attached a couple of logs from my
> experiments.
> The single core detailed model seems to work well.
>
> -Andrea
>
> --------------------------------------------------------
> x86_64-vmlinux-2.6.28.4.smp
>
>
> Script from M5 readfile is empty, starting bash shell...
> (none) / # cd parsec/install
> (none) install # /etc/init.d/rcS: line 19: rcS[965]: segfault at
> ffffffff8020b7c0 ip ffffffff8020b7d5 sp 00007fff47048b78 error 15 9
> 70 Segmentation fault /bin/bash
>
> Enter runle
>
>
> --------------------------------------------------------
> x86_64-vmlinux-2.6.22.9.smp
>
>
> Unable to handle kernel NULL pointer dereference at 00000000000000ac RIP:
> [<ffffffff80596900>] _spin_unlock_irqrestore+0x0/0x10
> PGD 0
> Oops: 0002 [1] SMP
> CPU 0
> Modules linked in:
> Pid: 0, comm: swapper Not tainted 2.6.22.9 #12
> RIP: e200:[<ffffffff80596900>] [<ffffffff80596900>]
> _spin_unlock_irqrestore+0x0/0x10
> RSP: 6633:ffffffff8023e138 EFLAGS: 000001a8
> RAX: ffff81000121d3c0 RBX: 65fb8948078b4853 RCX: ffff810001369150
> RDX: 00000000000002a8 RSI: ffff81000121d3d8 RDI: 00000000000000ac
> RBP: 0000000000000001 R08: ffffffff80245691 R09: ffffffff80225b53
> R10: ffffffff807beed0 R11: 0000000000000003 R12: 480000002425348b
> R13: e838788b48fce083 R14: 0000000300000000 R15: ffffffff802254f7
> FS: 00002b639e267dd0(0000) GS:ffffffff80757000(0000) knlGS:0000000000000000
> CS: 0010 DS: 0018 ES: 180018 CR0: 000000008005003b
> CR2: 00000000000000ac CR3: 0000000000201000 CR4: 00000000000006e0
> Process swapper (pid: 0, threadinfo ffffffff80770000, task ffffffff806f13c0)
> Stack: 48de8948fffff934 2e66668beb5bc789 0000000000841f0f 2eba0ff0f3894853
> 74c085d231c01900 468d48c3d0895b04 65217508463b4808 e80000002425348b
> 48de8948fffff8f4 5bffffff49e8c789 c3d08900000001ba 2e666666feeb0b0f
> Call Trace:
>
>
> Code: c7 07 01 00 00 00 56 9d c3 0f 1f 80 00 00 00 00 c7 07 01 00
> RIP [<ffffffff80596900>] _spin_unlock_irqrestore+0x0/0x10
> RSP <ffffffff8023e138>
> CR2: 00000000000000ac
> Kernel panic - not syncing: Aiee, killing interrupt handler!
>
>
> ---------------------------------------------------------
>
>
> On Fri, Mar 2, 2012 at 2:07 AM, Ankita (Garg) Goel <***@gmail.com>wrote:
>
>> Hi Xiao,
>>
>> What config file did you use to build the kernel ? The error probably
>> could be that the fs drivers like ext3 or ext4 were not built into the
>> kernel so the root filesystem in the image provided could not be mounted.
>>
>> There are a few sample config files provided in the X86 FS download
>> section on the wiki.
>>
>> Regards,
>> Ankita
>>
>>
>> On Thu, Mar 1, 2012 at 10:41 PM, Xiao Guo <***@gmail.com> wrote:
>>
>>> I got kernel panic when booting 4 cores using linux kernel 2.6.27, does
>>> anybody have a solution for this? Thanks!
>>>
>>>> Root-NFS: No NFS server available, giving up.
>>>> VFS: Unable to mount root fs via NFS, trying floppy.
>>>> VFS: Insert root floppy and press ENTER
>>>> VFS: Cannot open root device "<NULL>" or unknown-block(2,0)
>>>> Please append a correct "root=" boot option; here are the available
>>> partitions:
>>>> Kernel panic - not syncing: VFS: Unable to mount root fs on
>>> unknown-block(2,0)
>>>
>>> xiao
>>>
>>>
>>> On Feb 28, 2012, at 8:44 PM, Gabriel Michael Black wrote:
>>>
>>>> I saw this problem as well. I don't know what's causing it. If you're
>>>> looking for ways to debug it, I'd suggest figuring out what the CPU is
>>>> doing while it's hung. Is it just waiting idle? Is it stuck in a loop?
>>>> Is it stuck at one point in time and not progressing the simulation
>>>> (unlikely)? Then try to find out what it's waiting for, if anything,
>>>> and why that hasn't happened or the CPU hasn't noticed that it happened.
>>>>
>>>> Gabe
>>>>
>>>> Quoting "Ankita (Garg) Goel" <***@gmail.com>:
>>>>
>>>>> I am facing the same issue as Jason mentioned with detailed CPUs. The
>>>>> system boot-up seems to be going fine, but seems stuck at running INIT,
>>>>> even after 2-3 hrs.
>>>>>
>>>>> On Tue, Feb 28, 2012 at 2:37 PM, Nilay Vaish <***@cs.wisc.edu>
>>> wrote:
>>>>>
>>>>>> I have booted a 64 core system using timing simple cpu.
>>>>>>
>>>>>> --
>>>>>> Nilay
>>>>>>
>>>>>>
>>>>>>
>>>>>> On Tue, 28 Feb 2012, Jason Clemons wrote:
>>>>>>
>>>>>> I was able to boot 8 cores using atomic but I am still working on
>>>>>>> multicore timing or detailed models. I currently get a error when
>>> init
>>>>>>> calls fork at the end of booting. I am trying a couple of patches
>>> from the
>>>>>>> website to see if they will help.
>>>>>>>
>>>>>>
>>>>>> Later,
>>>>>>
>>>>>> Jason
>>>>>>
>>>>>>
>>>>>> Sent from my Samsung Epic? 4G
>>>>>>
>>>>>> Xiao Guo <***@gmail.com> wrote:
>>>>>>
>>>>>> Is it possible to boot more than 4 cores in x86 FS?
>>>>>>
>>>>>>
>>>>>> On Feb 27, 2012, at 5:51 PM, Jason Clemons wrote:
>>>>>>
>>>>>> Gabe,
>>>>>>>
>>>>>>> Thanks for the patch. I did a hg pull and hg update this afternoon
>>> and
>>>>>>> have made some progress on booting. I found that with a custom built
>>>>>>> version of 2.6.28.4 with smp on, the patch and the linux image files
>>>>>>> (linux-bigswap2.img and linux-x86.img) from gem5.org I am able to
>>> boot
>>>>>>> multicore x86 using the default cpu type which I think is atomic.
>>>>>>>
>>>>>>> After working with Andrea, we noted that the boot problem for x86
>>> happens
>>>>>>> with the PARSEC root image but not the linux-x86.img from gem5.org.
>>>>>>> This was tested using the custom kernel and the default cpu type.
>>>>>>>
>>>>>>> I am currently attempting to test the detailed cpu-type with 2 cpus.
>>> I
>>>>>>> will let you know of any errors I run into.
>>>>>>>
>>>>>>>
>>>>>>> Thanks,
>>>>>>>
>>>>>>> Jason Clemons
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> Quoting Gabe Black <***@eecs.umich.edu>:
>>>>>>>
>>>>>>> I looked into this more, and the code that sets up the memory
>>> hierarchy
>>>>>>>> was broken a little while ago. I have a patch that seems to fix it
>>>>>>>> (attached) but then it gets stuck a little later. I remember a
>>> while ago
>>>>>>>> someone had problems with atomic operations not being supported by
>>> the
>>>>>>>> memory system fully, but I don't remember the specifics. This might
>>> be
>>>>>>>> that problem too.
>>>>>>>>
>>>>>>>> Gabe
>>>>>>>>
>>>>>>>> On 02/24/12 02:45, Gabriel Michael Black wrote:
>>>>>>>>
>>>>>>>>> Yes, I have. You're using O3 though, and that support is fairly
>>> new.
>>>>>>>>>
>>>>>>>>> Gabe
>>>>>>>>>
>>>>>>>>> Quoting "Ankita (Garg) Goel" <***@gmail.com>:
>>>>>>>>>
>>>>>>>>> Hi,
>>>>>>>>>>
>>>>>>>>>> Has anyone been able to boot multiple cpus wth FS mode on x86 ? If
>>>>>>>>>> yes, it
>>>>>>>>>> would greatly help if you could share your environmental setup.
>>> The
>>>>>>>>>> issues
>>>>>>>>>> I am facing I have already posted in some previous emails.
>>>>>>>>>>
>>>>>>>>>> Thanks for your help!
>>>>>>>>>>
>>>>>>>>>> --
>>>>>>>>>> Regards,
>>>>>>>>>> Ankita
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>
>>>>>> ______________________________**_________________
>>>>>> gem5-users mailing list
>>>>>> gem5-***@gem5.org
>>>>>> http://m5sim.org/cgi-bin/**mailman/listinfo/gem5-users<
>>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users>
>>>>>>
>>>>>> _______________________________________________
>>>>>> gem5-users mailing list
>>>>>> gem5-***@gem5.org
>>>>>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> Regards,
>>>>> Ankita
>>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> gem5-users mailing list
>>>> gem5-***@gem5.org
>>>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> gem5-users mailing list
>>> gem5-***@gem5.org
>>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>>>
>>
>>
>>
>> --
>> Regards,
>> Ankita
>> Graduate Student
>> Department of Computer Science
>> University of Texas at Austin
>>
>>
>>
>> _______________________________________________
>> gem5-users mailing list
>> gem5-***@gem5.org
>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>>
>
Xuhao Chen
2012-02-29 02:03:34 UTC
Permalink
I can only boot 8 core in X86 FS mode, and the simulator shows that 8 core is the maximum limit. Do I need to modify some config file, or is there any patch for booting 64 core?




Xuhao Chen
PhD student
School of Computer
National University of Defense Technology
Changsha, Hunan, P.R.China, 410073
Tel: +86-159-741-03340 (Mobile)

From: Nilay Vaish
Date: 2012-02-29 04:37
To: gem5 users mailing list
Subject: Re: [gem5-users] Anyone able to boot multiple cpus in x86 FS?
I have booted a 64 core system using timing simple cpu.

--
Nilay


On Tue, 28 Feb 2012, Jason Clemons wrote:

> I was able to boot 8 cores using atomic but I am still working on multicore timing or detailed models. I currently get a error when init calls fork at the end of booting. I am trying a couple of patches from the website to see if they will help. 

Later,

Jason


Sent from my Samsung Epicâ„?4G

Xiao Guo <***@gmail.com> wrote:

Is it possible to boot more than 4 cores in x86 FS?


On Feb 27, 2012, at 5:51 PM, Jason Clemons wrote:

> Gabe,
>
> Thanks for the patch.  I did a hg pull and hg update this afternoon and have made some progress on booting.  I found that with a custom built version of 2.6.28.4 with smp on, the patch and the linux image files (linux-bigswap2.img and linux-x86.img) from gem5.org I am able to boot multicore x86 using the default cpu type which I think is atomic.
>
> After working with Andrea, we noted that the boot problem for x86 happens with the PARSEC root image but not the linux-x86.img from gem5.org.  This was tested using the custom kernel and the default cpu type.
>
> I am currently attempting to test the detailed cpu-type with 2 cpus.  I will let you know of any errors I run into.
>
>
> Thanks,
>
> Jason Clemons
>
>
>
> Quoting Gabe Black <***@eecs.umich.edu>:
>
>> I looked into this more, and the code that sets up the memory hierarchy
>> was broken a little while ago. I have a patch that seems to fix it
>> (attached) but then it gets stuck a little later. I remember a while ago
>> someone had problems with atomic operations not being supported by the
>> memory system fully, but I don't remember the specifics. This might be
>> that problem too.
>>
>> Gabe
>>
>> On 02/24/12 02:45, Gabriel Michael Black wrote:
>>> Yes, I have. You're using O3 though, and that support is fairly new.
>>>
>>> Gabe
>>>
>>> Quoting "Ankita (Garg) Goel" <***@gmail.com>:
>>>
>>>> Hi,
>>>>
>>>> Has anyone been able to boot multiple cpus wth FS mode on x86 ? If
>>>> yes, it
>>>> would greatly help if you could share your environmental setup. The
>>>> issues
>>>> I am facing I have already posted in some previous emails.
>>>>
>>>> Thanks for your help!
>>>>
>>>> --
>>>> Regards,
>>>> Ankita
>>>>
>>>

_______________________________________________
gem5-users mailing list
gem5-***@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users


_______________________________________________
gem5-users mailing list
gem5-***@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
Continue reading on narkive:
Loading...