netdev - Re: [PATCH net-next v6] page_pool: disable dma mapping support for 32-bit arch with 64-bit DMA

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: <YZKrfvt01uTosWc8@apalos.home>
Date:   Mon, 15 Nov 2021 20:48:30 +0200
From:   Ilias Apalodimas <ilias.apalodimas@...aro.org>
To:     Christoph Hellwig <hch@...radead.org>
Cc:     Yunsheng Lin <linyunsheng@...wei.com>,
        Guillaume Tucker <guillaume.tucker@...labora.com>,
        davem@...emloft.net, kuba@...nel.org, netdev@...r.kernel.org,
        linux-kernel@...r.kernel.org, linuxarm@...neuler.org,
        hawk@...nel.org, akpm@...ux-foundation.org, peterz@...radead.org,
        will@...nel.org, jhubbard@...dia.com, yuzhao@...gle.com,
        mcroce@...rosoft.com, fenghua.yu@...el.com, feng.tang@...el.com,
        jgg@...pe.ca, aarcange@...hat.com, guro@...com,
        "kernelci@...ups.io" <kernelci@...ups.io>
Subject: Re: [PATCH net-next v6] page_pool: disable dma mapping support for
 32-bit arch with 64-bit DMA

Hi Christoph,

On Mon, Nov 15, 2021 at 09:59:50AM -0800, Christoph Hellwig wrote:
> This is just popping out of nowhere on lkml, but yes ARM LPAE
> uses a 64-bit dma_addr_t, as a 64-bit phys_addr_t implies that.
> Same for x86-32 PAE and similar cases on MIPS and probably a few
> other architectures.
> 
> But what is the actual problem with a 32-bit virtual and 64-bit
> pysical/dma address?  That is a pretty common setup and absolutely
> needs to be deal with in drivers and inrastructure.

page_pool (the API in question), apart from allocating memory can manage
the mappings for you.  However while doing so it stores some parts (incl
the dma addr) in struct page.  The code in there could be simplified if 
we skipped support of the 'mapping' feature for 32-bit architectures with 
64-bit DMA.  We thought no driver was using the mapping feature (on 32bits)
and cleaned up that part, but apparently we missed 
'32-bit -- LPAE -- page pool manages DMA mappings'

Regards
/Ilias