Re: [PATCH net-next v7 1/6] page_pool: frag API support for 32-bit arch with 64-bit DMA

[Yunsheng Lin <linyunsheng@huawei.com>]

On 2023/8/24 2:00, Alexander Duyck wrote:
> On Wed, Aug 23, 2023 at 7:25 AM Jakub Kicinski <kuba@kernel.org> wrote:
>>
>> On Wed, 23 Aug 2023 11:03:31 +0800 Yunsheng Lin wrote:
>>> On 2023/8/22 23:38, Jakub Kicinski wrote:
>>>> On Tue, 22 Aug 2023 17:21:35 +0800 Yunsheng Lin wrote:
>>>>> As the CONFIG_PHYS_ADDR_T_64BIT seems to used widely in x86/arm/mips/powerpc,
>>>>> I am not sure if we can really make the above assumption.
>>>>>
>>>>> https://elixir.free-electrons.com/linux/v6.4-rc6/K/ident/CONFIG_PHYS_ADDR_T_64BIT
>>>>
>>>> Huh, it's actually used a lot less than I anticipated!
>>>>
>>>> None of the x86/arm/mips/powerpc systems matter IMHO - the only _real_
>>>
>>> Is there any particular reason that you think that the above systems does
>>> not really matter?
>>
>> Not the systems themselves but the combination of a 32b arch with
>> an address space >16TB. All those arches have 64b equivalent, seems
>> logical to use the 64b version for a system with a large address space.
>> If we're talking about a system which ends up running Linux.
>>
>>> As we have made a similar wrong assumption about those arches before, I am
>>> really trying to be more cautious about it.
>>>
>>> I searched through the web, some seems to be claiming that "32-bits is DEAD",
>>> I am not sure if there is some common agreement among the kernel community,
>>> is there any previous discussion about that?
>>
>> My suspicion/claim is that 32 + PAGE_SHIFT should be enough bits for
>> any 32b platform.
> 
> One additional thing we could consider would be to simply look at
> having page_pool enforce a DMA mask for the device to address any
> cases where we might not be able to fit the address. Then in the
> unlikely event that somebody is running a 32b system with over 16
> terabytes of RAM. With that the DMA subsystem would handle it for us
> and we wouldn't have to worry so much about it.

It seems there is a API to acquire the DMA mask used by the device:
https://elixir.free-electrons.com/linux/v6.4-rc6/source/include/linux/dma-mapping.h#L434

Is it possible to use that to check if DMA mask used by the device is
within 32 + PAGE_SHIFT limit, if yes, we use jakub's proposal to reduce
reduce the dma address bit, if no, we fail the page_pool creation?

> .
>