Re: [PATCH][RFC] io_uring: Fix io_uring_mmu_get_unmapped_area() for IA-64

On 7/20/23 4:38?PM, Helge Deller wrote:
> On 7/21/23 00:30, Jens Axboe wrote:
>> On 7/19/23 3:20?PM, Helge Deller wrote:
>>> The io_uring testcase is broken on IA-64 since commit d808459b2e31
>>> ("io_uring: Adjust mapping wrt architecture aliasing requirements").
>>>
>>> The reason is, that this commit introduced an own architecture
>>> independend get_unmapped_area() search algorithm which doesn't suite the
>>> memory region requirements for IA-64.
>>>
>>> To avoid similar problems in the future it's better to switch back to
>>> the architecture-provided get_unmapped_area() function and adjust the
>>> needed input parameters before the call.  Additionally
>>> io_uring_mmu_get_unmapped_area() will now become easier to understand
>>> and maintain.
>>>
>>> This patch has been tested on physical IA-64 and PA-RISC machines,
>>> without any failures in the io_uring testcases. On PA-RISC the
>>> LTP mmmap testcases did not report any regressions either.
>>>
>>> I don't expect issues for other architectures, but it would be nice if
>>> this patch could be tested on other machines too.
>>
>> Any comments from the IA64 folks?
> 
> matoro tested it on ia64 at least...
> 
>> Helge, should this be split into three patches? One for hppa, one for
>> ia64, and then the io_uring one?
> 
> If we split up, I would prefer to split it into 2 patches: One for
> io_uring together with the hppa patch, since they should go in
> together.

OK, that makes sense. Want to spin a new version done like that?

> The ia64 patch is probably unrelated, and can go seperately. But there
> doesn't seem to be any ia64 maintainer...?

Yeah, not too worried about IA64, and matoro having tested it is great.

> Do you have a chance to test this patch on the other io_uring
> platforms, without applying it into your tree? I think some testing
> would be good.

Yep, I can run it on arm64 and x86-64 pretty easily. Will do so.

-- 
Jens Axboe