Re: [Fastboot] IA64 Kdump patch V3

Re: [Fastboot] IA64 Kdump patch V3

[Jay Lan]

Jay Lan wrote:
> Hi Nanhai,
> 
> There is a problem in sn2. The sn2 does not have any memory with
> a predictible offset. It is not possible to specify a memory location
> at boot.
> 
> I modified the efi_initialize_iomem_resources() routine in
> arch/ia64/kernel/efi.c to calculate crashk_res.start and .end on the
> fly and uses the "A" of crashkernel=A@B as size to find a region
> that fits.
> 
> Maybe we should separate A and B as two options?

The "B" part of the crashkernel boot option is meaningless to sn2
machine and can be very confusing to sn2 users.

> 
> I now have memory reserved, but kernal hanged when executing 'kexec'
> command. There must be similar issues in kexec.

On further debugging, the kexec dumped core instead at this statement:
+        elf_rel_set_symbol(&info->rhdr, "__boot_param_base",
+                        &boot_param_base, sizeof(long));

of elf_ia64_load() in kexec/arch/ia64/kexec-elf-ia64.c.

Examing the memory i found "__boot_param_base" indeed not present in
the symbol table.

It seems that purgatory/arch/ia64/entry.S does not have a line
+DECLARE_DATAB(____boot_param_base).
Is it not needed in other ia64 machines? How does it work in HP
machines?

Thanks,
  - jay





> 
> Regards,
>  - jay

Re: [Fastboot] IA64 Kdump patch V3

[Jay Lan]

Jay Lan wrote:
> Jay Lan wrote:
> 
>> Hi Nanhai,
>>
>> There is a problem in sn2. The sn2 does not have any memory with
>> a predictible offset. It is not possible to specify a memory location
>> at boot.
>>
>> I modified the efi_initialize_iomem_resources() routine in
>> arch/ia64/kernel/efi.c to calculate crashk_res.start and .end on the
>> fly and uses the "A" of crashkernel=A@B as size to find a region
>> that fits.
>>
>> Maybe we should separate A and B as two options?
> 
> 
> The "B" part of the crashkernel boot option is meaningless to sn2
> machine and can be very confusing to sn2 users.
> 
>>
>> I now have memory reserved, but kernal hanged when executing 'kexec'
>> command. There must be similar issues in kexec.
> 
> 
> On further debugging, the kexec dumped core instead at this statement:
> +        elf_rel_set_symbol(&info->rhdr, "__boot_param_base",
> +                        &boot_param_base, sizeof(long));
> 
> of elf_ia64_load() in kexec/arch/ia64/kexec-elf-ia64.c.

Well, need a bit of clarification. :) The code hits a die() since
elf_rel_find_symbol() returned -1. The die() routine would do an
exit(1), but i changed die() to force a core dump to examine the
memory.

Regards,
  - jay


> 
> Examing the memory i found "__boot_param_base" indeed not present in
> the symbol table.
> 
> It seems that purgatory/arch/ia64/entry.S does not have a line
> +DECLARE_DATAB(____boot_param_base).
> Is it not needed in other ia64 machines? How does it work in HP
> machines?
> 
> Thanks,
>  - jay
> 
> 

RE: [Fastboot] IA64 Kdump patch V3

[Zou, Nanhai]

> -----Original Message-----
> From: Jay Lan [mailto:jlan@sgi.com]
> Sent: 2006Äê8ÔÂ30ÈÕ 9:27
> To: Zou, Nanhai
> Cc: Linux-IA64; fastboot; Luck, Tony; Dave Anderson; Eric W. Biederman; Horms
> Subject: Re: [Fastboot] IA64 Kdump patch V3
> 
> Jay Lan wrote:
> > Hi Nanhai,
> >
> > There is a problem in sn2. The sn2 does not have any memory with
> > a predictible offset. It is not possible to specify a memory location
> > at boot.
> >
> > I modified the efi_initialize_iomem_resources() routine in
> > arch/ia64/kernel/efi.c to calculate crashk_res.start and .end on the
> > fly and uses the "A" of crashkernel=A@B as size to find a region
> > that fits.
> >
> > Maybe we should separate A and B as two options?
> 
> The "B" part of the crashkernel boot option is meaningless to sn2
> machine and can be very confusing to sn2 users.
>

 Hi, Jay
	It is possible to have only A option and let kernel determine the B address at startup time. Horms has patch to do that, I will merge it.

> >
> > I now have memory reserved, but kernal hanged when executing 'kexec'
> > command. There must be similar issues in kexec.
> 
> On further debugging, the kexec dumped core instead at this statement:
> +        elf_rel_set_symbol(&info->rhdr, "__boot_param_base",
> +                        &boot_param_base, sizeof(long));
> 
> of elf_ia64_load() in kexec/arch/ia64/kexec-elf-ia64.c.
> 
> Examing the memory i found "__boot_param_base" indeed not present in
> the symbol table.
> 
> It seems that purgatory/arch/ia64/entry.S does not have a line
> +DECLARE_DATAB(____boot_param_base).
> Is it not needed in other ia64 machines? How does it work in HP
> machines?
  There is a DECLARE_DATA8(__boot_param_base) in purgatory/arch/ia64/entry.S
  Are you using kexec-tools-kdump10 patch?
 
 Thanks
 Zou Nan hai
> 
> Thanks,
>   - jay
> 

Re: [Fastboot] IA64 Kdump patch V3

[Jay Lan]

Zou, Nanhai wrote:
>>-----Original Message-----
>>From: Jay Lan [mailto:jlan@sgi.com]
>>Sent: 2006年8月30日 9:27
>>To: Zou, Nanhai
>>Cc: Linux-IA64; fastboot; Luck, Tony; Dave Anderson; Eric W. Biederman; Horms
>>Subject: Re: [Fastboot] IA64 Kdump patch V3
>>
>>Jay Lan wrote:
>>
>>>Hi Nanhai,
>>>
>>>There is a problem in sn2. The sn2 does not have any memory with
>>>a predictible offset. It is not possible to specify a memory location
>>>at boot.
>>>
>>>I modified the efi_initialize_iomem_resources() routine in
>>>arch/ia64/kernel/efi.c to calculate crashk_res.start and .end on the
>>>fly and uses the "A" of crashkernel=A@B as size to find a region
>>>that fits.
>>>
>>>Maybe we should separate A and B as two options?
>>
>>The "B" part of the crashkernel boot option is meaningless to sn2
>>machine and can be very confusing to sn2 users.
>>
> 
> 
>  Hi, Jay
> 	It is possible to have only A option and let kernel determine the B address at startup time. Horms has patch to do that, I will merge it.
> 
That would be great!

> 
>>>I now have memory reserved, but kernal hanged when executing 'kexec'
>>>command. There must be similar issues in kexec.
>>
>>On further debugging, the kexec dumped core instead at this statement:
>>+        elf_rel_set_symbol(&info->rhdr, "__boot_param_base",
>>+                        &boot_param_base, sizeof(long));
>>
>>of elf_ia64_load() in kexec/arch/ia64/kexec-elf-ia64.c.
>>
>>Examing the memory i found "__boot_param_base" indeed not present in
>>the symbol table.
>>
>>It seems that purgatory/arch/ia64/entry.S does not have a line
>>+DECLARE_DATAB(____boot_param_base).
>>Is it not needed in other ia64 machines? How does it work in HP
>>machines?
> 
>   There is a DECLARE_DATA8(__boot_param_base) in purgatory/arch/ia64/entry.S
>   Are you using kexec-tools-kdump10 patch?

Yes, i apply the kexec-tools-kdump10 patch and the "reserve efi memmap
and boot param" patch from you. Only these two. Did i miss any?

Thanks,
  - jay


>  
>  Thanks
>  Zou Nan hai
> 
>>Thanks,
>>  - jay
>>

RE: [Fastboot] IA64 Kdump patch V3

[Zou, Nanhai]

> -----Original Message-----
> From: Jay Lan [mailto:jlan@sgi.com]
> Sent: 2006年8月30日 9:46
> To: Zou, Nanhai
> Cc: Linux-IA64; fastboot; Luck, Tony; Dave Anderson; Eric W. Biederman; Horms
> Subject: Re: [Fastboot] IA64 Kdump patch V3
> 
> Zou, Nanhai wrote:
> >>-----Original Message-----
> >>From: Jay Lan [mailto:jlan@sgi.com]
> >>Sent: 2006年8月30日 9:27
> >>To: Zou, Nanhai
> >>Cc: Linux-IA64; fastboot; Luck, Tony; Dave Anderson; Eric W. Biederman; Horms
> >>Subject: Re: [Fastboot] IA64 Kdump patch V3
> >>
> >>Jay Lan wrote:
> >>
> >>>Hi Nanhai,
> >>>
> >>>There is a problem in sn2. The sn2 does not have any memory with
> >>>a predictible offset. It is not possible to specify a memory location
> >>>at boot.
> >>>
> >>>I modified the efi_initialize_iomem_resources() routine in
> >>>arch/ia64/kernel/efi.c to calculate crashk_res.start and .end on the
> >>>fly and uses the "A" of crashkernel=A@B as size to find a region
> >>>that fits.
> >>>
> >>>Maybe we should separate A and B as two options?
> >>
> >>The "B" part of the crashkernel boot option is meaningless to sn2
> >>machine and can be very confusing to sn2 users.
> >>
> >
> >
> >  Hi, Jay
> > 	It is possible to have only A option and let kernel determine the B address
> at startup time. Horms has patch to do that, I will merge it.
> >
> That would be great!
> 
> >
> >>>I now have memory reserved, but kernal hanged when executing 'kexec'
> >>>command. There must be similar issues in kexec.
> >>
> >>On further debugging, the kexec dumped core instead at this statement:
> >>+        elf_rel_set_symbol(&info->rhdr, "__boot_param_base",
> >>+                        &boot_param_base, sizeof(long));
> >>
> >>of elf_ia64_load() in kexec/arch/ia64/kexec-elf-ia64.c.
> >>
> >>Examing the memory i found "__boot_param_base" indeed not present in
> >>the symbol table.
> >>
> >>It seems that purgatory/arch/ia64/entry.S does not have a line
> >>+DECLARE_DATAB(____boot_param_base).
> >>Is it not needed in other ia64 machines? How does it work in HP
> >>machines?
> >
> >   There is a DECLARE_DATA8(__boot_param_base) in
> purgatory/arch/ia64/entry.S
> >   Are you using kexec-tools-kdump10 patch?
> 
> Yes, i apply the kexec-tools-kdump10 patch and the "reserve efi memmap
> and boot param" patch from you. Only these two. Did i miss any?
> 
 Then there should be a line 
DECLARE_DATA8(__boot_param_base) in purgatory/arch/ia64/entry.S...

 If it is hang in kernel, maybe there is something wrong with the physical destination address of reserved region, have you checked the /proc/iomem?

Thanks
Zou Nan hai

> Thanks,
>   - jay
> 
> 
> >
> >  Thanks
> >  Zou Nan hai
> >
> >>Thanks,
> >>  - jay
> >>
-
To unsubscribe from this list: send the line "unsubscribe linux-ia64" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html