shrink_folio_list() does not compile on late-model kernels

shrink_folio_list() does not compile on late-model kernels

[Chuck Lever]

Hi-

Some of my CI runs have started failing to build v6.13+ kernels with:

mm/vmscan.c: In function ‘shrink_folio_list’:
mm/vmscan.c:1082:21: error: implicit declaration of function
‘folio_contain_hwpoisoned_page’ [-Wimplicit-function-declaration]
 1082 |                 if (folio_contain_hwpoisoned_page(folio)) {
      |                     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
make[3]: *** [scripts/Makefile.build:196: mm/vmscan.o] Error 1
make[2]: *** [scripts/Makefile.build:442: mm] Error 2
make[2]: *** Waiting for unfinished jobs....
make[1]: *** [/data/linux-stable-rc/Makefile:1989: .] Error 2
make: *** [Makefile:251: __sub-make] Error 2

Is there a .config change I need to prevent this failure?

-- 
Chuck Lever

Re: shrink_folio_list() does not compile on late-model kernels

[Matthew Wilcox]

On Tue, Apr 08, 2025 at 03:21:21PM -0400, Chuck Lever wrote:
> Hi-
> 
> Some of my CI runs have started failing to build v6.13+ kernels with:
> 
> mm/vmscan.c: In function ‘shrink_folio_list’:
> mm/vmscan.c:1082:21: error: implicit declaration of function
> ‘folio_contain_hwpoisoned_page’ [-Wimplicit-function-declaration]
>  1082 |                 if (folio_contain_hwpoisoned_page(folio)) {
>       |                     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> make[3]: *** [scripts/Makefile.build:196: mm/vmscan.o] Error 1
> make[2]: *** [scripts/Makefile.build:442: mm] Error 2
> make[2]: *** Waiting for unfinished jobs....
> make[1]: *** [/data/linux-stable-rc/Makefile:1989: .] Error 2
> make: *** [Makefile:251: __sub-make] Error 2
> 
> Is there a .config change I need to prevent this failure?

I don't see why folio_contain_hwpoisoned_page() wouldn't be declared.
Could you include your .config?

Re: shrink_folio_list() does not compile on late-model kernels

[Chuck Lever]

On 4/8/25 4:21 PM, Matthew Wilcox wrote:
> On Tue, Apr 08, 2025 at 03:21:21PM -0400, Chuck Lever wrote:
>> Hi-
>>
>> Some of my CI runs have started failing to build v6.13+ kernels with:
>>
>> mm/vmscan.c: In function ‘shrink_folio_list’:
>> mm/vmscan.c:1082:21: error: implicit declaration of function
>> ‘folio_contain_hwpoisoned_page’ [-Wimplicit-function-declaration]
>>  1082 |                 if (folio_contain_hwpoisoned_page(folio)) {
>>       |                     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>> make[3]: *** [scripts/Makefile.build:196: mm/vmscan.o] Error 1
>> make[2]: *** [scripts/Makefile.build:442: mm] Error 2
>> make[2]: *** Waiting for unfinished jobs....
>> make[1]: *** [/data/linux-stable-rc/Makefile:1989: .] Error 2
>> make: *** [Makefile:251: __sub-make] Error 2
>>
>> Is there a .config change I need to prevent this failure?
> 
> I don't see why folio_contain_hwpoisoned_page() wouldn't be declared.
> Could you include your .config?

I assumed linux-mm would bounce an attachment. But it turns out I have a
sample .config in a public git repo:

https://github.com/chucklever/kdevops/blob/queue-6-13/playbooks/roles/bootlinux/templates/config-kdevops

The failure seems to be intermittent.


-- 
Chuck Lever

Re: shrink_folio_list() does not compile on late-model kernels

[Chuck Lever]

On 4/8/25 4:21 PM, Matthew Wilcox wrote:
> On Tue, Apr 08, 2025 at 03:21:21PM -0400, Chuck Lever wrote:
>> Hi-
>>
>> Some of my CI runs have started failing to build v6.13+ kernels with:
>>
>> mm/vmscan.c: In function ‘shrink_folio_list’:
>> mm/vmscan.c:1082:21: error: implicit declaration of function
>> ‘folio_contain_hwpoisoned_page’ [-Wimplicit-function-declaration]
>>  1082 |                 if (folio_contain_hwpoisoned_page(folio)) {
>>       |                     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>> make[3]: *** [scripts/Makefile.build:196: mm/vmscan.o] Error 1
>> make[2]: *** [scripts/Makefile.build:442: mm] Error 2
>> make[2]: *** Waiting for unfinished jobs....
>> make[1]: *** [/data/linux-stable-rc/Makefile:1989: .] Error 2
>> make: *** [Makefile:251: __sub-make] Error 2
>>
>> Is there a .config change I need to prevent this failure?
> 
> I don't see why folio_contain_hwpoisoned_page() wouldn't be declared.
> Could you include your .config?

I have not seen a repeat of this issue.


-- 
Chuck Lever