Date:   Tue, 1 Mar 2022 16:03:00 +0000
From:   Robin Murphy <robin.murphy@....com>
To:     Peter Geis <pgwipeout@...il.com>
Cc:     Jaehoon Chung <jh80.chung@...sung.com>,
        Ulf Hansson <ulf.hansson@...aro.org>,
        Heiko Stuebner <heiko@...ech.de>, linux-mmc@...r.kernel.org,
        arm-mail-list <linux-arm-kernel@...ts.infradead.org>,
        "open list:ARM/Rockchip SoC..." <linux-rockchip@...ts.infradead.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] mmc: host: dw-mmc-rockchip: avoid logspam when cd-broken

On 2022-03-01 14:49, Peter Geis wrote:
> On Tue, Mar 1, 2022 at 7:46 AM Peter Geis <pgwipeout@...il.com> wrote:
>>
>> On Tue, Mar 1, 2022 at 7:38 AM Robin Murphy <robin.murphy@....com> wrote:
>>>
>>> On 2022-03-01 11:49, Peter Geis wrote:
>>>> On Tue, Mar 1, 2022 at 6:23 AM Robin Murphy <robin.murphy@....com> wrote:
>>>>>
>>>>> On 2022-02-28 22:36, Peter Geis wrote:
>>>>>> The dw_mmc-rockchip driver drops a large amound of logspam constantly
>>>>>> when the cd-broken flag is enabled.
>>>>>> Set the warning to be debug ratelimited in this case.
>>>>>
>>>>> Isn't this just papering over some fundamental problem with the clock?
>>>>> If it's failing to set the expected rate for communicating with a card,
>>>>> then presumably that's an issue for correct operation in general? The
>>>>> fact that polling for a card makes a lot more of that communication
>>>>> happen seems unrelated :/
>>>>
>>>> Good Morning,
>>>>
>>>> This only happens when a card is not inserted, so communication cannot happen.
>>>
>>> Well, I suppose there's a philosophical question in there about whether
>>> shouting into the void counts as "communication", but AFAIR what the
>>> polling function does is power up the controller, send a command, and
>>> see if it gets a response.
>>>
>>> If the clock can't be set to the proper rate for low-speed discovery,
>>> some or all cards may not be detected properly. Conversely if it is
>>> already at a slow enough rate for discovery but can't be set higher once
>>> a proper communication mode has been established, data transfer
>>> performance will be terrible. Either way, it is not OK in general for
>>> clk_set_rate() to fail, hence the warning. You have a clock driver problem.
>>
>> Alright, I'll look into this.
>> It seems only extremely low clock speeds fail and I know rockchip
>> chips have a hard time with extremely low clock rates.
>> I'll trace out where the failure is happening.
> 
> Okay, I hope you can provide me a direction to go from here, because
> it looks like it's doing exactly what it should do in this situation.
> mmc core is requesting a rate (200k/100k).
> clk core tries to find a parent to provide a clock that low and fails,
> because the lowest possible parent is 750k.
> clk_sdmmc(x) is listed as no-div, so it can't go any lower.
> 
> It seems to me that this error is sane, because other results of
> einval you want to catch.
> But einval in this case is fine, because
> The thing that strikes me weird is currently clk_core thinks the
> lowest possible freq here is 0, when in actuality it should be 750k,
> am I correct here?
> The mmc controller has an internal divider, so if my line of thinking
> is correct here we should be more flexible here and request a rate
> that's acceptable rather than just failing if it doesn't work.
> But that's based on my limited understanding of how mmc core is
> requesting this and what it expects in return.

The downstream solution appears to be just to clamp the rate for 
detection[1][2]. Not sure whether it's feasible to try to be cleverer 
with the local divider to settle on a more in-spec rate for the final 
output :/

Robin.

[1] 
https://github.com/JeffyCN/mirrors/commit/d80d5062b22f9c4a559401bdb7b2727c4ced36c0
[2] 
https://github.com/JeffyCN/mirrors/commit/3f26edfb2392df25efc361ad0a9f41d0917e40ee

Powered by blists - more mailing lists