linux-kernel - Re: [PATCH] ASoC: fsl_sai: Remove unnecessary FIFO reset in ISR

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>] [day] [month] [year] [list]

Message-ID: <CAGoOwPSgPSV2HOqC=t_NVor8YqxaLxmeDvMf8_j25phsPpcd4g@mail.gmail.com>
Date:   Tue, 16 Aug 2022 23:35:22 -0700
From:   Nicolin Chen <nicoleotsuka@...il.com>
To:     Shengjiu Wang <shengjiu.wang@...il.com>
Cc:     Shengjiu Wang <shengjiu.wang@....com>,
        Xiubo Li <Xiubo.Lee@...il.com>,
        Fabio Estevam <festevam@...il.com>,
        Liam Girdwood <lgirdwood@...il.com>,
        Mark Brown <broonie@...nel.org>,
        Jaroslav Kysela <perex@...ex.cz>,
        Takashi Iwai <tiwai@...e.com>, alsa-devel@...a-project.org,
        linuxppc-dev <linuxppc-dev@...ts.ozlabs.org>,
        linux-kernel <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] ASoC: fsl_sai: Remove unnecessary FIFO reset in ISR

On Tue, Aug 16, 2022 at 11:29 PM Shengjiu Wang <shengjiu.wang@...il.com> wrote:

>> > The FIFO reset drops the words in the FIFO, which may cause
>> > channel swap when SAI module is running, especially when the
>> > DMA speed is low. So it is not good to do FIFO reset in ISR,
>> > then remove the operation.
>>
>> I don't recall the details of adding this many years ago, but
>> leaving underrun/overrun errors unhandled does not sound right
>> to me either. Would it result in a channel swap also? Perhaps
>> there needs to be a reset routine that stops and restarts the
>> DMA as well?
>
>
> Remove the reset, the channel swap is gone.

I have no doubt about that :)

> IMO, no need to handle the underrun/overrun in driver, the SAI
> hardware can handle the read/write pointer itself when xrun happen,
> and we don't need reset routine.

That'd be okay then.