linux-kernel - Re: [PATCH net-next v1] page_pool: check for dma_sync

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <21036339-3eeb-4606-9a84-d36bddba2b31@huawei.com>
Date: Fri, 11 Oct 2024 16:55:34 +0800
From: Yunsheng Lin <linyunsheng@...wei.com>
To: Furong Xu <0x1207@...il.com>, Ilias Apalodimas
	<ilias.apalodimas@...aro.org>
CC: <netdev@...r.kernel.org>, <linux-kernel@...r.kernel.org>, Jesper Dangaard
 Brouer <hawk@...nel.org>, "David S. Miller" <davem@...emloft.net>, Eric
 Dumazet <edumazet@...gle.com>, Jakub Kicinski <kuba@...nel.org>, Paolo Abeni
	<pabeni@...hat.com>, <xfr@...look.com>
Subject: Re: [PATCH net-next v1] page_pool: check for dma_sync_size earlier

On 2024/10/11 14:31, Furong Xu wrote:
> Hi Ilias,
> 
> On Fri, 11 Oct 2024 08:06:04 +0300, Ilias Apalodimas <ilias.apalodimas@...aro.org> wrote:
> 
>> Hi Furong,
>>
>> On Fri, 11 Oct 2024 at 05:15, Furong Xu <0x1207@...il.com> wrote:
>>>
>>> On Thu, 10 Oct 2024 19:53:39 +0800, Yunsheng Lin <linyunsheng@...wei.com> wrote:
>>>  
>>>> Is there any reason that those drivers not to unset the PP_FLAG_DMA_SYNC_DEV
>>>> when calling page_pool_create()?
>>>> Does it only need dma sync for some cases and not need dma sync for other
>>>> cases? if so, why not do the dma sync in the driver instead?  
>>>
>>> The answer is in this commit:
>>> https://git.kernel.org/netdev/net/c/5546da79e6cc  
>>
>> I am not sure I am following. Where does the stmmac driver call a sync
>> with len 0?
> For now, only drivers/net/ethernet/freescale/fec_main.c does.
> And stmmac driver does not yet, but I will send another patch to make it call sync with
> len 0. This is a proper fix as Jakub Kicinski suggested.

In order to support the above use case, it seems there might be two
options here:
1. Driver calls page_pool_create() without PP_FLAG_DMA_SYNC_DEV and
   handle the dma sync itself.
2. Page_pool may provides a non-dma-sync version of page_pool_put_page()
   API even when Driver calls page_pool_create() with PP_FLAG_DMA_SYNC_DEV.

Maybe option 2 is better one in the longer term as it may provide some
flexibility for the user and enable removing of the DMA_SYNC_DEV in the
future?