[PATCH v5] powerpc/kdump: Add support for crashkernel CMA reservation

[Sourabh Jain]

On 04/11/25 15:54, Ritesh Harjani (IBM) wrote:
> Sourabh Jain <sourabhjain at linux.ibm.com> writes:
>
>> On 04/11/25 10:48, Sourabh Jain wrote:
>>>
>>> On 03/11/25 15:40, Ritesh Harjani (IBM) wrote:
>>>> Sourabh Jain <sourabhjain at linux.ibm.com> writes:
>>>>
>>>>> Commit 35c18f2933c5 ("Add a new optional ",cma" suffix to the
>>>>> crashkernel= command line option") and commit ab475510e042 ("kdump:
>>>>> implement reserve_crashkernel_cma") added CMA support for kdump
>>>>> crashkernel reservation.
>>>>>
>>>>> Extend crashkernel CMA reservation support to powerpc.
>>>>>
>>>>> The following changes are made to enable CMA reservation on powerpc:
>>>>>
>>>>> - Parse and obtain the CMA reservation size along with other
>>>>> crashkernel
>>>>>     parameters
>>>>> - Call reserve_crashkernel_cma() to allocate the CMA region for kdump
>>>>> - Include the CMA-reserved ranges in the usable memory ranges for the
>>>>>     kdump kernel to use.
>>>>> - Exclude the CMA-reserved ranges from the crash kernel memory to
>>>>>     prevent them from being exported through /proc/vmcore.
>>>>>
>>>>> With the introduction of the CMA crashkernel regions,
>>>>> crash_exclude_mem_range() needs to be called multiple times to exclude
>>>>> both crashk_res and crashk_cma_ranges from the crash memory ranges. To
>>>>> avoid repetitive logic for validating mem_ranges size and handling
>>>>> reallocation when required, this functionality is moved to a new
>>>>> wrapper
>>>>> function crash_exclude_mem_range_guarded().
>>>>>
>>>>> To ensure proper CMA reservation, reserve_crashkernel_cma() is called
>>>>> after pageblock_order is initialized.
>>>>>
>>>>> Update kernel-parameters.txt to document CMA support for crashkernel on
>>>>> powerpc architecture.
>>>>>
>>>>> Cc: Baoquan he <bhe at redhat.com>
>>>>> Cc: Jiri Bohac <jbohac at suse.cz>
>>>>> Cc: Hari Bathini <hbathini at linux.ibm.com>
>>>>> Cc: Madhavan Srinivasan <maddy at linux.ibm.com>
>>>>> Cc: Mahesh Salgaonkar <mahesh at linux.ibm.com>
>>>>> Cc: Michael Ellerman <mpe at ellerman.id.au>
>>>>> Cc: Ritesh Harjani (IBM) <ritesh.list at gmail.com>
>>>>> Cc: Shivang Upadhyay <shivangu at linux.ibm.com>
>>>>> Cc: kexec at lists.infradead.org
>>>>> Signed-off-by: Sourabh Jain <sourabhjain at linux.ibm.com>
>>>>> ---
>>>>> Changlog:
>>>>>
>>>>> v3 -> v4
>>>>>    - Removed repeated initialization to tmem in
>>>>>      crash_exclude_mem_range_guarded()
>>>>>    - Call crash_exclude_mem_range() with right crashk ranges
>>>>>
>>>>> v4 -> v5:
>>>>>    - Document CMA-based crashkernel support for ppc64 in
>>>>> kernel-parameters.txt
>>>>> ---
>>>>>    .../admin-guide/kernel-parameters.txt         |  2 +-
>>>>>    arch/powerpc/include/asm/kexec.h              |  2 +
>>>>>    arch/powerpc/kernel/setup-common.c            |  4 +-
>>>>>    arch/powerpc/kexec/core.c                     | 10 ++++-
>>>>>    arch/powerpc/kexec/ranges.c                   | 43
>>>>> ++++++++++++++-----
>>>>>    5 files changed, 47 insertions(+), 14 deletions(-)
>>>>>
>>>>> diff --git a/Documentation/admin-guide/kernel-parameters.txt
>>>>> b/Documentation/admin-guide/kernel-parameters.txt
>>>>> index 6c42061ca20e..0f386b546cec 100644
>>>>> --- a/Documentation/admin-guide/kernel-parameters.txt
>>>>> +++ b/Documentation/admin-guide/kernel-parameters.txt
>>>>> @@ -1013,7 +1013,7 @@
>>>>>                It will be ignored when crashkernel=X,high is not used
>>>>>                or memory reserved is below 4G.
>>>>>        crashkernel=size[KMG],cma
>>>>> -            [KNL, X86] Reserve additional crash kernel memory from
>>>>> +            [KNL, X86, ppc64] Reserve additional crash kernel
>>>>> memory from
>>>> Shouldn't this be PPC and not ppc64?
>>>>
>>>> If I see the crash_dump support...
>>>>
>>>> config ARCH_SUPPORTS_CRASH_DUMP
>>>>      def_bool PPC64 || PPC_BOOK3S_32 || PPC_85xx || (44x && !SMP)
>>>>
>>>> The changes below aren't specific to ppc64 correct?
>>> The thing is this feature is only supported with KEXEC_FILE and which
>>> only supported on PPC64:
>>>
>>> config ARCH_SUPPORTS_KEXEC_FILE
>>>      def_bool PPC64
>>>
>>> Hence I kept it as ppc64.
>>>
> I am not much familiar with the kexec_load v/s kexec_file_load
> internals. Maybe because of that I am unable to clearly understand your
> above points.
>
> But let me try and explain what I think you meant :)
>
> We first call "get_usable_memory_ranges(&umem)" which updates the usable
> memory ranges in "umem". We then call "update_usable_mem_fdt(fdt, umem)"
> which updates the FDT for the kdump kernel's fdt to inform about these
> usable memory ranges to the kdump kernel.
>
> Now since your patch only does that in get_usable_memory_range(), this
> extra CMA reservation is mainly only useful when the kdump load happens
> via kexec_file_load(), (because get_usable_memory_range() only gets
> called from kexec_file_load() path)
>
> Is this what you meant here?


Yeah, for kexec_file_load, the FDT for the kdump kernel is prepared in 
the Linux
kernel (using the functions you mentioned), whereas for kexec_load, it 
is prepared
in the kexec tool (userspace). Hence, these changes are not sufficient 
to support
this feature with the kexec_load syscall.

The kexec tool must be updated to ensure that the FDT is prepared in a 
way that
marks the crashkernel CMA reservation as usable in the kdump FDT for the
kexec_load system call.

Anyway, it makes more sense to say that crashkernel=xM,cma support is 
available
on ppc rather than ppc64, since restricting crashkernel CMA reservation to
KEXEC_FILE does not help. The details are explained below.


>
>
>>> I think I should update that in the commit message.
>>>
>>> Also do you think is it good to restrict this feature to KEXEC_FILE?
>> Putting this under KEXEC_FILE may not help much because KEXEC_FILE is
>> enabled
>> by default in most configurations. Once it is enabled, the CMA
>> reservation will
>> happen regardless of which system call is used to load the kdump kernel
>> (kexec_load or kexec_file_load).
>>
> What I understood from the feature was that, on the normal production
> kernel this feature crashkernel=xM,cma allows to reserve an extra xMB of
> memory as a CMA region for kdump kernel's memory allocations. But this
> CMA reservation would happen in the normal kernel itself during
> setup_arch() -> kdump_cma_reserve()..
>
> And this CMA reservation happens irrespective of whether the kdump
> kernel will get loaded via whichever system call.

Yeah that's right.

>
>> However, not restricting this feature to KEXEC_FILE will allow the kexec
>> tool to
>> independently add support for this feature in the future for the kexec_load
>> system call.
> Sure.
>
>> With that logic, I think if we do not restrict this feature to
>> KEXEC_FILE, the support
>> will be available for ppc and not limited to ppc64.
>>
> Yes, that make sense.
>
> If one doesn't want to make the CMA reservation, then we need not pass
> the extra cmdline argument and no reservation will be made. So, no need
> to restrict this to PPC64 by making it available only for KEXEC_FILE.

Agree.