linux-kernel - Re: [PATCH RFC 2/3] mmc: sdhci-omap: Add using external 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] [day] [month] [year] [list]

Date:   Mon, 19 Nov 2018 17:32:25 +0100
From:   Ulf Hansson <ulf.hansson@...aro.org>
To:     Kishon Vijay Abraham I <kishon@...com>,
        Arnd Bergmann <arnd@...db.de>,
        Chunyan Zhang <zhang.chunyan@...aro.org>
Cc:     Adrian Hunter <adrian.hunter@...el.com>,
        "linux-mmc@...r.kernel.org" <linux-mmc@...r.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Mark Brown <broonie@...nel.org>, Sekhar Nori <nsekhar@...com>,
        Chunyan Zhang <zhang.lyra@...il.com>
Subject: Re: [PATCH RFC 2/3] mmc: sdhci-omap: Add using external dma

On 9 November 2018 at 06:27, Kishon Vijay Abraham I <kishon@...com> wrote:
> Hi Arnd,
>
> On 06/11/18 6:21 PM, Arnd Bergmann wrote:
>> On 11/5/18, Kishon Vijay Abraham I <kishon@...com> wrote:
>>> On 05/11/18 8:46 AM, Chunyan Zhang wrote:
>>>>
>>>> +   sdhci_switch_extdma(host, true);
>>>
>>> A number of devices using sdhci-omap supports ADMA. So switching to
>>> external
>>> DMA shouldn't be unconditional.
>>>
>>> IMHO sdhci.c should see if the device supports ADMA or SDMA. If not it
>>> should
>>> try switching to external DMA and if external DMA too is not supported, it
>>> should use PIO.
>>
>> What's the reasoning for preferring ADMA/SDMA over external DMA if
>> both are supported?
>
> Generally from our experiments we've found ADMA gives better throughput than
> DMA. I have to ascertain the actual reasons for that.
>>
>> I'd expect that if the external DMA for some reason is worse than
>> ADMA, we just wouldn't list it in the DT at all, but if it's listed and
>> ADMA also works, then the driver should try to use it before falling
>> back to ADMA.
>
> I would agree that for newer dtbs. However if you consider omap_hsmmc, external
> DMA bindings are already added in dt. If we try to switch to sdhci-omap with
> the same bindings, systems with older dtbs will use external DMA and give
> lesser throughput.

I was under the impression that the internal DMA (for old versions) is
rather poor.

However, in the end this is a software policy decision of what to use.
I have no strong opinion, so feel free to pick what you think makes
best sense.

Maybe Adrian also have some thoughts?

Kind regards
Uffe