Fwd: Kexec support for PS3 (ppc64)

Fwd: Kexec support for PS3 (ppc64)

[Rajasekaran P]

[-- Attachment #1.1: Type: text/plain, Size: 1478 bytes --]

Hi all,
I am sorry about the earlier post with disclaimer message.


> On Thu, Mar 06, 2008 at 04:55:04PM +0530, Rajasekaran Periyasamy
wrote:
> > Hi,
> > I put a printk() in the early lines of early_parse_crashk() with
> > KERN_EMERG, but not able to see that on the boot screen. I think the
> > control is not transferred to early_parse_crashk() at all. With out
this
> > "crashkernel" argument, the same kernel boots up properly.
>
> Looks strange.
>
> Could you please include more debug printks in the function
> do_early_param in init/main.c to check whether early_parse_crashk is
> called or not?


I've tried, but no result... Please have a look at the snapshot of boot
process.

I think the kernel is not at all getting initiated to boot up.

Whatever you see in the snapshot,
1. "Modified cmdline:" is from
"kexec-tools-1.101/kexec/arch/ppc64/fs2dt.c" putnode() function,
2. "segment" lines are from
"kexec-tools-1.101/kexec/arch/ppc64/kexec-elf-ppc64.c" elf_ppc64_load()
function.

KEXEC is used by KBOOT to boot the (secondary) linux kernel. But I am
wondering why the kernel is not booting up when I give "crashkernel"
option.

>
> How do you boot the kernel, by yaboot or kboot?

I use kboot latest version (kboot-20080201).
Please refer to
http://www.kernel.org/pub/linux/kernel/people/geoff/cell/CELL-Linux-CL_2
0080201-ADDON/src/<http://www.kernel.org/pub/linux/kernel/people/geoff/cell/CELL-Linux-CL_20080201-ADDON/src/>for
src of kboot.

Regards,
Rajasekaran

[-- Attachment #1.2: Type: text/html, Size: 1846 bytes --]

[-- Attachment #2: ps3_linux_boot_failure1.JPG --]
[-- Type: image/jpeg, Size: 48438 bytes --]

[-- Attachment #3: Type: text/plain, Size: 143 bytes --]

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: Fwd: Kexec support for PS3 (ppc64)

[Geoff Levand]

Hi,

I maintain kexec for PS3, but have just done what was needed to
make it work for kboot.  I am not a kexec expert, and have
never tried to use kdump, or other features of kexec, only the
things needed for kboot.

On 03/06/2008 09:55 PM, Rajasekaran P wrote:
>> On Thu, Mar 06, 2008 at 04:55:04PM +0530, Rajasekaran Periyasamy
> wrote:
>> > Hi,
>> > I put a printk() in the early lines of early_parse_crashk() with
>> > KERN_EMERG, but not able to see that on the boot screen. I think the
>> > control is not transferred to early_parse_crashk() at all. With out
> this
>> > "crashkernel" argument, the same kernel boots up properly.


This kind of thing is what kboot uses, and is known to work:

  kexec -l -t elf-ppc64 --append='root=XXX video=YYY' /boot/vmlinux.strip
  kexec -e

I recommend you get it to work first.


>> Looks strange.
>>
>> Could you please include more debug printks in the function
>> do_early_param in init/main.c to check whether early_parse_crashk is
>> called or not?


That won't work, since it is hanging before the video driver is up,
and you'll never see those.


> I've tried, but no result... Please have a look at the snapshot of boot
> process.
> 
> I think the kernel is not at all getting initiated to boot up.


Your photo shows that the first stage kernel has gone down, and it
hung up either very late going down, or early coming back up in the
second stage.


> Whatever you see in the snapshot,
> 1. "Modified cmdline:" is from
> "kexec-tools-1.101/kexec/arch/ppc64/fs2dt.c" putnode() function,


I guess that version of exec-tools (kexec-tools-1.101) may be too old,
and doesn't have the bug-fixes I needed to make to kexec-tools for it
to work with PS3.  Those fixes were merged into kexec-tools several
month ago.

I recommend you try with an up-to-date kexec from here:

git://git.kernel.org/pub/scm/linux/kernel/git/horms/kexec-tools-testing.git

I have tested both 32 and 64 bit builds.

I also recommend to start with a known good second stage kernel.
I tested Linux 2.6.25-rc4 built with ps3_defconfig and it works.
Do not use an initrd at first try.

-Geoff


_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: Fwd: Kexec support for PS3 (ppc64)

[Rajasekaran P]

[-- Attachment #1.1: Type: text/plain, Size: 3713 bytes --]

Hi Geoff,
Thanks for your information.

I used latest kexec tools with 2.6.23 kernel built with ps3_defconfig.
This kernel boots normally on PS3. But when I  add "crashkernel" argument,
it goes on hang.


Please have a look at the following command outputs:

# kexec -v

kexec-tools-testing 20080227 released 27th February 2008



# kexec --command-line="video=ps3fb:mode:166 rhgb root=/dev/ps3da3 " \

--initrd=/boot/initrd-2.6.23.img \

-l /boot/vmlinux-2.6.23

# kexec -e

 >>>> THIS WORKS FINE, I MEAN, THE NEW KERNEL IS BOOTING UP WITH GUI (FC8)


# kexec --command-line="video=ps3fb:mode:166 rhgb root=/dev/ps3da3
crashkernel=128M@32M" \

--initrd=/boot/initrd-2.6.23.img \

-l /boot/vmlinux-2.6.23


get memory ranges:1

Modified cmdline:video=ps3fb:mode:166 rhgb root=/dev/ps3da3 crashkernel=
128M@32M

segment[0].mem:0x586000 memsz:5787648

segment[1].mem:0xb0b000 memsz:36864

segment[2].mem:0xb14000 memsz:3469312

segment[3].mem:0x7fff000 memsz:4096



# umount –a

umount: /: device is busy

# taskset 1 kexec –e



Nothing happens after this. As you said earlier in your mail, I am not
getting any PRINTK outputs from kernel on screen.

Can anybody help me to find out any other alternate ways to debug?

Thanks in advance,

Rajasekaran


On Fri, Mar 7, 2008 at 12:01 PM, Geoff Levand <geoffrey.levand@am.sony.com>
wrote:

> Hi,
>
> I maintain kexec for PS3, but have just done what was needed to
> make it work for kboot.  I am not a kexec expert, and have
> never tried to use kdump, or other features of kexec, only the
> things needed for kboot.
>
> On 03/06/2008 09:55 PM, Rajasekaran P wrote:
> >> On Thu, Mar 06, 2008 at 04:55:04PM +0530, Rajasekaran Periyasamy
> > wrote:
> >> > Hi,
> >> > I put a printk() in the early lines of early_parse_crashk() with
> >> > KERN_EMERG, but not able to see that on the boot screen. I think the
> >> > control is not transferred to early_parse_crashk() at all. With out
> > this
> >> > "crashkernel" argument, the same kernel boots up properly.
>
>
> This kind of thing is what kboot uses, and is known to work:
>
>  kexec -l -t elf-ppc64 --append='root=XXX video=YYY' /boot/vmlinux.strip
>  kexec -e
>
> I recommend you get it to work first.
>
>
> >> Looks strange.
> >>
> >> Could you please include more debug printks in the function
> >> do_early_param in init/main.c to check whether early_parse_crashk is
> >> called or not?
>
>
> That won't work, since it is hanging before the video driver is up,
> and you'll never see those.
>
>
> > I've tried, but no result... Please have a look at the snapshot of boot
> > process.
> >
> > I think the kernel is not at all getting initiated to boot up.
>
>
> Your photo shows that the first stage kernel has gone down, and it
> hung up either very late going down, or early coming back up in the
> second stage.
>
>
> > Whatever you see in the snapshot,
> > 1. "Modified cmdline:" is from
> > "kexec-tools-1.101/kexec/arch/ppc64/fs2dt.c" putnode() function,
>
>
> I guess that version of exec-tools (kexec-tools-1.101) may be too old,
> and doesn't have the bug-fixes I needed to make to kexec-tools for it
> to work with PS3.  Those fixes were merged into kexec-tools several
> month ago.
>
> I recommend you try with an up-to-date kexec from here:
>
> git://git.kernel.org/pub/scm/linux/kernel/git/horms/kexec-
> tools-testing.git
>
> I have tested both 32 and 64 bit builds.
>
> I also recommend to start with a known good second stage kernel.
> I tested Linux 2.6.25-rc4 built with ps3_defconfig and it works.
> Do not use an initrd at first try.
>
> -Geoff
>
>

[-- Attachment #1.2: Type: text/html, Size: 7431 bytes --]

[-- Attachment #2: Type: text/plain, Size: 143 bytes --]

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: Fwd: Kexec support for PS3 (ppc64)

[Geoff Levand]

Rajasekaran P wrote:
> Hi Geoff,
> Thanks for your information.
> 
> I used latest kexec tools with 2.6.23 kernel built with ps3_defconfig.
> This kernel boots normally on PS3. But when I  add "crashkernel"
> argument, it goes on hang.
> 
> # kexec --command-line="video=ps3fb:mode:166 rhgb root=/dev/ps3da3
> crashkernel=128M@32M" \
> --initrd=/boot/initrd-2.6.23.img \
> -l /boot/vmlinux-2.6.23
> 
> # taskset 1 kexec –e
> 
> Nothing happens after this. As you said earlier in your mail, I am not
> getting any PRINTK outputs from kernel on screen.

As I said, this usage is currently unsupported.  Could you tell me what you
want to do?  I mean, why do you want to use this 'crashkernel=' option?

As a note, the ps3 has 128MiB of boot memory, so I guess 128M@32M won't
work.

-Geoff


_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: Fwd: Kexec support for PS3 (ppc64)

[Geoff Levand]

Geoff Levand wrote:
> Rajasekaran P wrote:
>> Hi Geoff,
>> Thanks for your information.
>> 
>> I used latest kexec tools with 2.6.23 kernel built with ps3_defconfig.
>> This kernel boots normally on PS3. But when I  add "crashkernel"
>> argument, it goes on hang.
>> 
>> # kexec --command-line="video=ps3fb:mode:166 rhgb root=/dev/ps3da3
>> crashkernel=128M@32M" \
>> --initrd=/boot/initrd-2.6.23.img \
>> -l /boot/vmlinux-2.6.23
>> 
>> # taskset 1 kexec –e
>> 
>> Nothing happens after this. As you said earlier in your mail, I am not
>> getting any PRINTK outputs from kernel on screen.
> 
> As I said, this usage is currently unsupported.  Could you tell me what you
> want to do?  I mean, why do you want to use this 'crashkernel=' option?
> 
> As a note, the ps3 has 128MiB of boot memory, so I guess 128M@32M won't
> work.

Just FYI, I tried this and the kernel booted:

./kexec -l -t elf-ppc64 --append='crashkernel=64M@32M root=/dev/nfs rw ip=dhcp video=wuxga' /boot/vmlinux.strip

I saw this output:

kexec: waiting for cpu 1 (physical 1) to go down
Reserving 64MB of memory at 32MB for crashkernel (System RAM: 128MB)
...
Kernel command line: crashkernel=64M@32M root=/dev/nfs rw ip=dhcp video=wuxga

-Geoff


_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: Fwd: Kexec support for PS3 (ppc64)

[Pete/Piet Delaney]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Geoff Levand wrote:
| Geoff Levand wrote:
|> Rajasekaran P wrote:
|>> Hi Geoff,
|>> Thanks for your information.
|>>
|>> I used latest kexec tools with 2.6.23 kernel built with ps3_defconfig.
|>> This kernel boots normally on PS3. But when I  add "crashkernel"
|>> argument, it goes on hang.
|>>
|>> # kexec --command-line="video=ps3fb:mode:166 rhgb root=/dev/ps3da3
|>> crashkernel=128M@32M" \
|>> --initrd=/boot/initrd-2.6.23.img \
|>> -l /boot/vmlinux-2.6.23
|>>
|>> # taskset 1 kexec –e
|>>
|>> Nothing happens after this. As you said earlier in your mail, I am not
|>> getting any PRINTK outputs from kernel on screen.

I was wondering about bringing in the printf() code from purgatory
and using that to print info about what's going on in the kexec code.
The standalone printf() might come in handy in other places; Ex: kgdb
stub. If it was in a library it could also be used during kernel
decompression to inform on problems occurring, like checksum errors.

- -piet


|> As I said, this usage is currently unsupported.  Could you tell me
what you
|> want to do?  I mean, why do you want to use this 'crashkernel=' option?
|>
|> As a note, the ps3 has 128MiB of boot memory, so I guess 128M@32M won't
|> work.
|
| Just FYI, I tried this and the kernel booted:
|
| ./kexec -l -t elf-ppc64 --append='crashkernel=64M@32M root=/dev/nfs rw
ip=dhcp video=wuxga' /boot/vmlinux.strip
|
| I saw this output:
|
| kexec: waiting for cpu 1 (physical 1) to go down
| Reserving 64MB of memory at 32MB for crashkernel (System RAM: 128MB)
| ...
| Kernel command line: crashkernel=64M@32M root=/dev/nfs rw ip=dhcp
video=wuxga
|
| -Geoff
|
|
| _______________________________________________
| kexec mailing list
| kexec@lists.infradead.org
| http://lists.infradead.org/mailman/listinfo/kexec
|

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFH1zACJICwm/rv3hoRAtyQAJ9E+nFUBAtTpmRCYoXTCNXNeuRm7ACfXOUa
YX3Ipj6hDj9Cfwli6H/E1Gg=
=SMaU
-----END PGP SIGNATURE-----

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: Fwd: Kexec support for PS3 (ppc64)

[Geoff Levand]

Pete/Piet Delaney wrote:
> Geoff Levand wrote:
> | Geoff Levand wrote:
> |> Rajasekaran P wrote:
> |>> Hi Geoff,
> |>> Thanks for your information.
> |>>
> |>> I used latest kexec tools with 2.6.23 kernel built with ps3_defconfig.
> |>> This kernel boots normally on PS3. But when I  add "crashkernel"
> |>> argument, it goes on hang.
> |>>
> |>> # kexec --command-line="video=ps3fb:mode:166 rhgb root=/dev/ps3da3
> |>> crashkernel=128M@32M" \
> |>> --initrd=/boot/initrd-2.6.23.img \
> |>> -l /boot/vmlinux-2.6.23
> |>>
> |>> # taskset 1 kexec \x13e
> |>>
> |>> Nothing happens after this. As you said earlier in your mail, I am not
> |>> getting any PRINTK outputs from kernel on screen.
> 
> I was wondering about bringing in the printf() code from purgatory
> and using that to print info about what's going on in the kexec code.
> The standalone printf() might come in handy in other places; Ex: kgdb
> stub. If it was in a library it could also be used during kernel
> decompression to inform on problems occurring, like checksum errors.

When the first stage kernel goes down it releases all hypervisor
resources so that whatever boots next can successful open those
resources.  The frame buffer is one of those resources that are
released, and after it is, there will be no more output to the
display until the second stage image brings up its display.  It
is not a matter of adding print statements or string formatting.

-Geoff


_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: Fwd: Kexec support for PS3 (ppc64)

[Piet Delaney]

[-- Attachment #1.1: Type: text/plain, Size: 1784 bytes --]

Geoff Levand wrote:
> Pete/Piet Delaney wrote:
>   
>> Geoff Levand wrote:
>> | Geoff Levand wrote:
>> |> Rajasekaran P wrote:
>> |>> Hi Geoff,
>> |>> Thanks for your information.
>> |>>
>> |>> I used latest kexec tools with 2.6.23 kernel built with ps3_defconfig.
>> |>> This kernel boots normally on PS3. But when I  add "crashkernel"
>> |>> argument, it goes on hang.
>> |>>
>> |>> # kexec --command-line="video=ps3fb:mode:166 rhgb root=/dev/ps3da3
>> |>> crashkernel=128M@32M" \
>> |>> --initrd=/boot/initrd-2.6.23.img \
>> |>> -l /boot/vmlinux-2.6.23
>> |>>
>> |>> # taskset 1 kexec \x13e
>> |>>
>> |>> Nothing happens after this. As you said earlier in your mail, I am not
>> |>> getting any PRINTK outputs from kernel on screen.
>>
>> I was wondering about bringing in the printf() code from purgatory
>> and using that to print info about what's going on in the kexec code.
>> The standalone printf() might come in handy in other places; Ex: kgdb
>> stub. If it was in a library it could also be used during kernel
>> decompression to inform on problems occurring, like checksum errors.
>>     
>
> When the first stage kernel goes down it releases all hypervisor
> resources so that whatever boots next can successful open those
> resources.  The frame buffer is one of those resources that are
> released, and after it is, there will be no more output to the
> display until the second stage image brings up its display.  It
> is not a matter of adding print statements or string formatting.
>   

Oh, I thought you experienced the same problem I had
with printk() calls in some of the kexec functions making
it no longer work. I was considering changing my KEXEC_DEBUG()
macros to using printf() instead of printk() and importing
the purgatory printf() code.
> -Geoff
>
>   


[-- Attachment #1.2: Type: text/html, Size: 2339 bytes --]

[-- Attachment #2: Type: text/plain, Size: 143 bytes --]

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: Fwd: Kexec support for PS3 (ppc64)

[Mohan Kumar M]

On Tue, Mar 11, 2008 at 06:40:46PM -0700, Geoff Levand wrote:
> When the first stage kernel goes down it releases all hypervisor
> resources so that whatever boots next can successful open those
> resources.  The frame buffer is one of those resources that are
> released, and after it is, there will be no more output to the
> display until the second stage image brings up its display.  It
> is not a matter of adding print statements or string formatting.
>

Geoff,

What about using Hypervisor call H_PUT_TERM_CHAR?

I used this call to debug the first kernel kexec shutdown sequence.

Regards,
Mohan.

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: Fwd: Kexec support for PS3 (ppc64)

[Geoff Levand]

On 03/11/2008 10:25 PM, Mohan Kumar M wrote:
> On Tue, Mar 11, 2008 at 06:40:46PM -0700, Geoff Levand wrote:
>> When the first stage kernel goes down it releases all hypervisor
>> resources so that whatever boots next can successful open those
>> resources.  The frame buffer is one of those resources that are
>> released, and after it is, there will be no more output to the
>> display until the second stage image brings up its display.  It
>> is not a matter of adding print statements or string formatting.
>>
> 
> Geoff,
> 
> What about using Hypervisor call H_PUT_TERM_CHAR?
> 
> I used this call to debug the first kernel kexec shutdown sequence.

That is for the IBM pSeries.  I can't see how it would work
on the ps3.  ps3 hvcalls are in lv1calls.h.

Just out of curiosity, how exactly did you 'use' it?

-Geoff


_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: Fwd: Kexec support for PS3 (ppc64)

[Mohan Kumar M]

On Tue, Mar 11, 2008 at 10:40:25PM -0700, Geoff Levand wrote:
> That is for the IBM pSeries.  I can't see how it would work
> on the ps3.  ps3 hvcalls are in lv1calls.h.

Oh, okay.

> 
> Just out of curiosity, how exactly did you 'use' it?
> 

We used the function plpar_put_term_char

Regards,
Mohan.

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: Fwd: Kexec support for PS3 (ppc64)

[Rajasekaran P]

[-- Attachment #1.1: Type: text/plain, Size: 2317 bytes --]

On Tue, Mar 11, 2008 at 11:48 AM, Geoff Levand <geoffrey.levand@am.sony.com>
wrote:

> Geoff Levand wrote:
>
>
> Just FYI, I tried this and the kernel booted:
>
> ./kexec -l -t elf-ppc64 --append='crashkernel=64M@32M root=/dev/nfs rw
> ip=dhcp video=wuxga' /boot/vmlinux.strip
>
> I saw this output:
>
> kexec: waiting for cpu 1 (physical 1) to go down
> Reserving 64MB of memory at 32MB for crashkernel (System RAM: 128MB)
> ...
> Kernel command line: crashkernel=64M@32M root=/dev/nfs rw ip=dhcp
> video=wuxga
>
> -Geoff
>
> This works properly. Thanks Geoff.

Initially I made mistake by referring /proc/meminfo and thought the total
RAM was > 200M iB (actually it shows around 211MiB in /proc/meminfo
excluding swap space). But actual available memory is 128MiB only. Thats why
reserving 128MiB on 32MiB failed and kernel was not booting up.

Now "crashkernel" argument works properly for 64MB@64MB and kernel reserves
64MB for crash dump capture kernel.

This memory is reserved for "dump capture kernel" which will be loaded later
by using kexec. Please refer linux_src/Documentation/kdump/kdump.txt for
more information.

I tried to load dump capture configured kernel and facing similar kind of
problem. After successful loading, Dump capture kernel is NOT BOOTING UP
upon kernel crash.
I used the following command to load:

# kexec -t elf-ppc64 --append="video=ps3fb:mode:166 rhgb root=/dev/ps3da3
single maxcpus=1 noirqdistrib" \
 --initrd=/boot/initrd-2.6.23-kdump.img \
 -p vmlinux.strip.kdump
*
Output:*
*setup_memory_ranges memory_range[0] start:585001, end:8000000
get memory ranges:1
setup_memory_ranges memory_range[0] start:585001, end:8000000
Modified cmdline:video=ps3fb:mode:166 rhgb root=/dev/ps3da3 single maxcpus=1
noirqdistrib elfcorehdr=71068K savemaxmem=64M
segment[0].mem:0x4000000 memsz:5599232
segment[1].mem:0x4557000 memsz:65536
segment[2].mem:0x4567000 memsz:4096
segment[3].mem:0x4568000 memsz:36864
segment[4].mem:0x4571000 memsz:3469312
segment[5].mem:0x7fff000 memsz:4096
*
One more thing I confirmed was, this same dump capture kernel boots up
properly when I load it on normal mode (kexec -l and then kexec -e). I am
having no clue on this..

If anybody know how to debug a kernel failure during boot up, it will be
very helpful. Please help.

Thanks in advance,
Rajasekaran.P

[-- Attachment #1.2: Type: text/html, Size: 2842 bytes --]

[-- Attachment #2: Type: text/plain, Size: 143 bytes --]

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: Fwd: Kexec support for PS3 (ppc64)

[Mohan Kumar M]

On Fri, Mar 14, 2008 at 03:36:58AM -0700, Rajasekaran P wrote:
> On Tue, Mar 11, 2008 at 11:48 AM, Geoff Levand <geoffrey.levand@am.sony.com>
> wrote:
> 
> Initially I made mistake by referring /proc/meminfo and thought the total
> RAM was > 200M iB (actually it shows around 211MiB in /proc/meminfo
> excluding swap space). But actual available memory is 128MiB only. Thats why
> reserving 128MiB on 32MiB failed and kernel was not booting up.
> 
> Now "crashkernel" argument works properly for 64MB@64MB and kernel reserves
> 64MB for crash dump capture kernel.

In PPC64, crashkernel base address is hard coded i.e it is always loaded
at 32MB, even though you have passed 64MB@64MB. Did you get any warning
message like crashkernel reserved at 32MB?

> 
> This memory is reserved for "dump capture kernel" which will be loaded later
> by using kexec. Please refer linux_src/Documentation/kdump/kdump.txt for
> more information.
> 
> I tried to load dump capture configured kernel and facing similar kind of
> problem. After successful loading, Dump capture kernel is NOT BOOTING UP
> upon kernel crash.

Do you get any message from the dump capture kernel? How do you trigger
the dump?

> I used the following command to load:
> 
> # kexec -t elf-ppc64 --append="video=ps3fb:mode:166 rhgb root=/dev/ps3da3
> single maxcpus=1 noirqdistrib" \
>  --initrd=/boot/initrd-2.6.23-kdump.img \
>  -p vmlinux.strip.kdump

	^^ Is this kernel compiled with CRASH_DUMP enabled? If yes, it
should build the kernel for 32MB.

> *
> Output:*
> *setup_memory_ranges memory_range[0] start:585001, end:8000000
> get memory ranges:1
> setup_memory_ranges memory_range[0] start:585001, end:8000000
> Modified cmdline:video=ps3fb:mode:166 rhgb root=/dev/ps3da3 single maxcpus=1
> noirqdistrib elfcorehdr=71068K savemaxmem=64M
> segment[0].mem:0x4000000 memsz:5599232

			^^ Says the vmlinux is loaded at 64MB
> segment[1].mem:0x4557000 memsz:65536
> segment[2].mem:0x4567000 memsz:4096
> segment[3].mem:0x4568000 memsz:36864
> segment[4].mem:0x4571000 memsz:3469312
> segment[5].mem:0x7fff000 memsz:4096
> *
> One more thing I confirmed was, this same dump capture kernel boots up
> properly when I load it on normal mode (kexec -l and then kexec -e). I am
> having no clue on this..

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: Fwd: Kexec support for PS3 (ppc64)

[Rajasekaran P]

[-- Attachment #1.1: Type: text/plain, Size: 3034 bytes --]

On Fri, Mar 14, 2008 at 4:02 AM, Mohan Kumar M <mohan@in.ibm.com> wrote:

> On Fri, Mar 14, 2008 at 03:36:58AM -0700, Rajasekaran P wrote:
> > On Tue, Mar 11, 2008 at 11:48 AM, Geoff Levand <
> geoffrey.levand@am.sony.com>
> > wrote:
> >
> > Initially I made mistake by referring /proc/meminfo and thought the
> total
> > RAM was > 200M iB (actually it shows around 211MiB in /proc/meminfo
> > excluding swap space). But actual available memory is 128MiB only. Thats
> why
> > reserving 128MiB on 32MiB failed and kernel was not booting up.
> >
> > Now "crashkernel" argument works properly for 64MB@64MB and kernel
> reserves
> > 64MB for crash dump capture kernel.
>
> In PPC64, crashkernel base address is hard coded i.e it is always loaded
> at 32MB, even though you have passed 64MB@64MB. Did you get any warning
> message like crashkernel reserved at 32MB?
>
> >
> > This memory is reserved for "dump capture kernel" which will be loaded
> later
> > by using kexec. Please refer linux_src/Documentation/kdump/kdump.txt for
> > more information.
> >
> > I tried to load dump capture configured kernel and facing similar kind
> of
> > problem. After successful loading, Dump capture kernel is NOT BOOTING UP
> > upon kernel crash.
>
> Do you get any message from the dump capture kernel? How do you trigger
> the dump?
>
> > I used the following command to load:
> >
> > # kexec -t elf-ppc64 --append="video=ps3fb:mode:166 rhgb
> root=/dev/ps3da3
> > single maxcpus=1 noirqdistrib" \
> >  --initrd=/boot/initrd-2.6.23-kdump.img \
> >  -p vmlinux.strip.kdump
>
>         ^^ Is this kernel compiled with CRASH_DUMP enabled? If yes, it
> should build the kernel for 32MB.


>
> > *
> > Output:*
> > *setup_memory_ranges memory_range[0] start:585001, end:8000000
> > get memory ranges:1
> > setup_memory_ranges memory_range[0] start:585001, end:8000000
> > Modified cmdline:video=ps3fb:mode:166 rhgb root=/dev/ps3da3 single
> maxcpus=1
> > noirqdistrib elfcorehdr=71068K savemaxmem=64M
> > segment[0].mem:0x4000000 memsz:5599232
>
>                         ^^ Says the vmlinux is loaded at 64MB
> > segment[1].mem:0x4557000 memsz:65536
> > segment[2].mem:0x4567000 memsz:4096
> > segment[3].mem:0x4568000 memsz:36864
> > segment[4].mem:0x4571000 memsz:3469312
> > segment[5].mem:0x7fff000 memsz:4096
> > *
> > One more thing I confirmed was, this same dump capture kernel boots up
> > properly when I load it on normal mode (kexec -l and then kexec -e). I
> am
> > having no clue on this..
>

Yes,I built this kernel with CRASH_DUMP and VMCORE file system  enabled.
Moreover, I disabled SMP as well.
My system kernel is (First kernel) is SMP enabled, does it matter?

I dont get any message from dump capture kernel.

I trigger crash dump using :

echo c > /proc/sysrq-trigger
Initially capture kernel was loaded into 32MB only, but still it was not
booting. So I tweaked the KDUMP_KERNELBASE macro to load capture kernel to
64MB ( My idea is, first 64MB for system kernel and second 64MB for capture
kernel). Should it be 32MB only?

[-- Attachment #1.2: Type: text/html, Size: 4486 bytes --]

[-- Attachment #2: Type: text/plain, Size: 143 bytes --]

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: Fwd: Kexec support for PS3 (ppc64)

[Mohan Kumar M]

On Fri, Mar 14, 2008 at 04:18:52AM -0700, Rajasekaran P wrote:
> Yes,I built this kernel with CRASH_DUMP and VMCORE file system  enabled.
> Moreover, I disabled SMP as well.
> My system kernel is (First kernel) is SMP enabled, does it matter?

It should not matter.

> 
> I dont get any message from dump capture kernel.
> 
> I trigger crash dump using :
> 
> echo c > /proc/sysrq-trigger

While triggering the dump, don't you get any message, like sending IPI
to cpu?

> Initially capture kernel was loaded into 32MB only, but still it was not
> booting. So I tweaked the KDUMP_KERNELBASE macro to load capture kernel to
> 64MB ( My idea is, first 64MB for system kernel and second 64MB for capture
> kernel). Should it be 32MB only?

You can't do that. It should be at 32MB only.

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: Fwd: Kexec support for PS3 (ppc64)

[Rajasekaran P]

[-- Attachment #1.1: Type: text/plain, Size: 974 bytes --]

On Fri, Mar 14, 2008 at 4:52 PM, Mohan Kumar M <mohan@in.ibm.com> wrote:

> On Fri, Mar 14, 2008 at 04:18:52AM -0700, Rajasekaran P wrote:
> > Yes,I built this kernel with CRASH_DUMP and VMCORE file system  enabled.
> > Moreover, I disabled SMP as well.
> > My system kernel is (First kernel) is SMP enabled, does it matter?
>
> It should not matter.
>
> >
> > I dont get any message from dump capture kernel.
> >
> > I trigger crash dump using :
> >
> > echo c > /proc/sysrq-trigger
>
> While triggering the dump, don't you get any message, like sending IPI
> to cpu?
>

No, I dont  get anything  on terminal.

>
> > Initially capture kernel was loaded into 32MB only, but still it was not
> > booting. So I tweaked the KDUMP_KERNELBASE macro to load capture kernel
> to
> > 64MB ( My idea is, first 64MB for system kernel and second 64MB for
> capture
> > kernel). Should it be 32MB only?
>
> You can't do that. It should be at 32MB only.

Ok, I will try this and update.

[-- Attachment #1.2: Type: text/html, Size: 1595 bytes --]

[-- Attachment #2: Type: text/plain, Size: 143 bytes --]

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec