linux-kernel - Re: [PATCH] sound/soc/adi/axi-spdif.c: Support programmable master clock

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: <54806D4E.8040504@topic.nl>
Date:	Thu, 4 Dec 2014 15:18:54 +0100
From:	Mike Looijmans <mike.looijmans@...ic.nl>
To:	Lars-Peter Clausen <lars@...afoo.de>
CC:	<alsa-devel@...a-project.org>, <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] sound/soc/adi/axi-spdif.c: Support programmable master
 clock

On 12/04/2014 01:45 PM, Lars-Peter Clausen wrote:
> On 12/04/2014 07:52 AM, Mike Looijmans wrote:
>> If the master clock supports programmable rates, program it to generate
>> the desired frequency. Only apply constraints when the clock is fixed.
>> This allows proper clock generation for both 44100 and 48000 Hz based
>> sampling rates if the platform supports it.
>>
>> The clock frequency must be set before enabling it. Enabling the clock
>> was done in "startup", but that occurs before "hw_params" where the rate
>> is known. Move the clock start to the hw_params routine, and keep track
>> of whether the clock has been started, because shutdown may be called
>> without having called hw_params first.
>
> Usually that shouldn't be a problem. If your clock chip requires it to be
> disabled in order to be reprogrammed than the CLK_SET_RATE_GATE flag should be
> set. This will tell the core to disable the clock before changing it.

The issue here is not the clock's capabilities, but the order in which things 
are being done. If the driver just enables the clock without ever having set a 
rate, the clock will run at whatever happens to be the default. That default 
may have unpredictable results or even be harmful to the system. So the driver 
should first set a valid clock rate before enabling the clock. Suggestions on 
rewording my comments to better reflect that are welcome.

>
> [...]
>>   static const struct snd_soc_dai_ops axi_spdif_dai_ops = {
>> @@ -216,14 +227,17 @@ static int axi_spdif_probe(struct platform_device *pdev)
>>       spdif->dma_data.addr_width = 4;
>>       spdif->dma_data.maxburst = 1;
>>
>> -    spdif->ratnum.num = clk_get_rate(spdif->clk_ref) / 128;
>> -    spdif->ratnum.den_step = 1;
>> -    spdif->ratnum.den_min = 1;
>> -    spdif->ratnum.den_max = 64;
>> -
>> -    spdif->rate_constraints.rats = &spdif->ratnum;
>> -    spdif->rate_constraints.nrats = 1;
>> +    /* Determine if the clock rate is fixed. If it cannot change frequency,
>> +     * it returns an error here. */
>> +    if (clk_round_rate(spdif->clk_ref, 128 * 44100) < 0) {
>
> I don't think this works. For a fixed clock clk_round_rate() will return the
> fixed rate rather than an error. I tried the patch and even though I have a
> fixed clock the constraints are no longer set.
>
> There is unfortunately no good way to enumerate which frequencies are
> supported by a clock other than just calling round_rate for all possible rates.
>
> I think the best way to implement this for now is to try e.g. 32000 * 128,
> 44100 * 128, 48000 * 128 and then check if clk_round_rate returns the expected
> rate and if it does set up a rate constraint for that rate.

For what I could see, ALSA never reported or limited the sample rate correctly 
even before this patch, so maybe the even simpler approach is better: Just 
remove the constraint. I wonder how you concluded that the constraint didn't 
get added?

Actually, the SPDIF logic wants a clock that is an "integer multiple of 
128*samplerate in the range of 1..64". Other than just looping through them 
all, there's also no way to request the clock framework for such a requirement.

Your suggestion is quite robust though, but the fixed clock may be set to any 
integer multiple of the desired frequency, so the algorithm would not work if 
the fixed clock is running at 48000*256 (12.8MHz like in the reference 
design). So I guess the best I could do here is just check that:

clk_round_rate(spdif->clk_ref, 128 * 44100) != clk_round_rate(spdif->clk_ref, 
128 * 48000)

Only a configurable clock that is compatible with the patch would do that.

Met vriendelijke groet / kind regards,

Mike Looijmans
System Expert

TOPIC Embedded Systems
Eindhovenseweg 32-C, NL-5683 KH Best
Postbus 440, NL-5680 AK Best
Telefoon: (+31) (0) 499 33 69 79
Telefax:  (+31) (0) 499 33 69 70
E-mail: mike.looijmans@...ic.nl
Website: www.topic.nl

Please consider the environment before printing this e-mail

Topic zoekt gedreven (embedded) software specialisten!
http://topic.nl/vacatures/topic-zoekt-software-engineers/

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/