[PATCH v9 0/8] kdump: reduce vmcore size and capture time

[Wandun]

On 10/21/26 21:23, Rob Herring wrote:
> On Wed, Oct 21, 2026 at 04:30:03PM +0800, Wandun Chen wrote:
>> From: Wandun Chen <chenwandun at lixiang.com>
>>
>> On SoCs that carve out large firmware-owned reserved memory (GPU,
>> camera ISP, ...), kdump currently dumps those carveouts as part of
>> system RAM even though their contents are firmware state that is not
>> useful for kernel crash analysis.
>>
>> ACPI systems already filter reserved memory out of the vmcore through
>> their existing path: firmware reports those carveouts as unusable in
>> the EFI memory map, they get MEMBLOCK_NOMAP set in memblock.memory
>> (is_usable_memory(), efi-init.c), and the ELF header builder has
>> always skipped nomap regions; only DT-based systems currently fail
>> to filter these regions, which is what this series addresses.
>>
>> This series introduces a MEMBLOCK_NODUMP flag (suggested by Rob) in
>> memblock to filter reserved memory on DT-based architectures. A new
>> flag is needed because MEMBLOCK_NOMAP can only express regions without
>> a linear mapping, while MEMBLOCK_NODUMP can filter regions that are
>> linearly mapped. Reserved regions are marked MEMBLOCK_NODUMP by default
>> so kdump omits them; reusable CMA regions are different because their
>> pages are handed back to the buddy allocator and may carry crash-relevant
>> data.
> 
> There is little point in continuing to send new versions until Mike 
> agrees with MEMBLOCK_NODUMP being added. You didn't answer him on v7.

Hi Rob,

Thanks for the reminder. On v7 your reply to Mike stated exactly what
I wanted to say, after a week without further objections, so I assumed
the question was settled and went ahead with next version. That was a
mistake on my part: I should have gotten Mike's agreement on the flag
before sending a new version. I'll hold off further versions until then.



Hi Mike,

Sorry for not following up on your v7 question. The reason a new flag
is needed is described in the v9 cover letter:

    A new flag is needed because MEMBLOCK_NOMAP can only express
    regions without a linear mapping, while MEMBLOCK_NODUMP can
    filter regions that are linearly mapped.

Put differently: today, whether a region ends up in the vmcore is a
side effect of how it is mapped, mapped regions are always dumped,
nomap regions never are. Nothing can express "mapped, but not useful
in a crash dump". MEMBLOCK_NODUMP decouples the two, and DT is simply
the first user.



> 
> Also, I intend to apply this[1] before any further reserved memory 
> changes which will greatly simplify the code.

Understood, and it makes sense. It removes the two-pass reserved_mem
initialization and indeed simplify the code, I'll rebase on top of
it once it lands;


Best regards,
Wandun




> 
> Rob
> 
> [1] https://lore.kernel.org/all/20261015082315.56876-1-m.szyprowski@samsung.com/