Re: [PATCH 0/3] tools/mm: show_page_info.py enhancements and fixes

[Ye Liu]

在 2026/10/15 08:05, Andrew Morton 写道:
> On Wed, 14 Oct 2026 15:04:27 +0800 Ye Liu <ye.liu@xxxxxxxxx> wrote:
> 
>>
>> This series improves the show_page_info.py drgn script with bug fixes
>> and new features:
>>
>> Patch 1 fixes memory cgroup information retrieval by updating the
>> obj_cgroup type casting and handling slabobj_ext field name changes
>> for compatibility with newer kernel versions.
>>
>> Patch 2 removes the incorrect Page Index field display, as pages
>> don't have indexes - only folios do.
>>
>> Patch 3 adds VMA flags decoding and VMA name display to provide
>> more detailed information about the virtual memory area containing
>> the page.
>>
>> These changes ensure the script works correctly with current kernel
>> memory management structures and provides more comprehensive page
>> state information.
> 
> Thanks for continuing to maintain this.
> 
> I'm curious.  Does this work mean that you're using this tool in
> real-world situations?
> 
Hi Andrew,

Yes. We've been using this tool to investigate real issues on our
production systems at Kylinsoft.

One concrete case: we hit list_del corruption warnings in
isolate_lru_pages() during memory reclaim (shrink_inactive_list /
shrink_active_list path). The system stayed running, but the LRU
list was clearly corrupted for some pages. We needed to inspect the
affected pages' state.

I also use it during kernel development as a general-purpose page
state inspector. There's an rmap analysis feature I haven't posted
yet — it currently only works on our internal kernel version. I also
plan to add a mode that accepts a page address directly rather than
requiring a PID + virtual address, which would be useful for analyzing
pages found in dmesg crash traces.

> 
> Sashiko had quite a lot to say.
> 
> 	https://sashiko.dev/#/patchset/20261014-show_page_info-v1-0-7a68aee0114a@xxxxxxxxx
I'll address Sashiko's comments in a v2.

-- 
Thanks,
Ye Liu