linux-kernel - Re: [PATCH v2 2/4] hwrng: stm32 - implement support for STM32MP25x platforms

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]
Message-ID: <90ce2c1e-6e8b-43ba-9313-8b5560cb8e4d@foss.st.com>
Date: Wed, 16 Oct 2024 09:54:26 +0200
From: Gatien CHEVALLIER <gatien.chevallier@...s.st.com>
To: Marek Vasut <marex@...x.de>, Olivia Mackall <olivia@...enic.com>,
        Herbert
 Xu <herbert@...dor.apana.org.au>,
        Rob Herring <robh@...nel.org>,
        Krzysztof
 Kozlowski <krzk+dt@...nel.org>,
        Conor Dooley <conor+dt@...nel.org>,
        Maxime
 Coquelin <mcoquelin.stm32@...il.com>,
        Alexandre Torgue
	<alexandre.torgue@...s.st.com>,
        Lionel Debieve <lionel.debieve@...s.st.com>
CC: <linux-crypto@...r.kernel.org>, <devicetree@...r.kernel.org>,
        <linux-stm32@...md-mailman.stormreply.com>,
        <linux-arm-kernel@...ts.infradead.org>, <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v2 2/4] hwrng: stm32 - implement support for STM32MP25x
 platforms



On 10/15/24 17:39, Marek Vasut wrote:
> On 10/15/24 5:10 PM, Gatien CHEVALLIER wrote:
>>
>>
>> On 10/14/24 20:55, Marek Vasut wrote:
>>> On 10/14/24 2:36 PM, Gatien CHEVALLIER wrote:
>>>>
>>>>
>>>> On 10/14/24 10:52, Marek Vasut wrote:
>>>>> On 10/14/24 10:38 AM, Gatien CHEVALLIER wrote:
>>>>>>
>>>>>>
>>>>>> On 10/11/24 18:17, Marek Vasut wrote:
>>>>>>> On 10/11/24 5:41 PM, Gatien Chevallier wrote:
>>>>>>>
>>>>>>> [...]
>>>>>>>
>>>>>>>> @@ -551,6 +565,41 @@ static int stm32_rng_probe(struct 
>>>>>>>> platform_device *ofdev)
>>>>>>>>       priv->rng.read = stm32_rng_read;
>>>>>>>>       priv->rng.quality = 900;
>>>>>>>> +    if (!priv->data->nb_clock || priv->data->nb_clock > 2)
>>>>>>>> +        return -EINVAL;
>>>>>>>> +
>>>>>>>> +    priv->clk_bulk = devm_kzalloc(dev, priv->data->nb_clock * 
>>>>>>>> sizeof(*priv->clk_bulk),
>>>>>>>> +                      GFP_KERNEL);
>>>>>>>> +    if (!priv->clk_bulk)
>>>>>>>> +        return -ENOMEM;
>>>>>>>
>>>>>>> Try this:
>>>>>>>
>>>>>>> ret = devm_clk_bulk_get(dev, priv->data->nb_clock, priv->clk_bulk);
>>>>>>> ...
>>>>>>> // Swap the clock if they are not in the right order:
>>>>>>> if (priv->data->nb_clock == 2 &&
>>>>>>>      strcmp(__clk_get_name(priv->clk_bulk[0].clk), "core"))
>>>>>>> {
>>>>>>>   const char *id = priv->clk_bulk[1].id;
>>>>>>>   struct clk *clk = priv->clk_bulk[1].clk;
>>>>>>>   priv->clk_bulk[1].id = priv->clk_bulk[0].id;
>>>>>>>   priv->clk_bulk[1].clk = priv->clk_bulk[0].clk;
>>>>>>>   priv->clk_bulk[0].id = id;
>>>>>>>   priv->clk_bulk[0].clk = clk;
>>>>>>> }
>>>>>>>
>>>>>>
>>>>>> Hi Marek,
>>>>>>
>>>>>> This won't work as the name returned by this API is clk->core->name.
>>>>>> AFAICT, it doesn't correspond to the names present in the device tree
>>>>>> under the "clock-names" property.
>>>>>> Any other idea or are you fine with what's below?
>>>>> Hmmm, it is not great, but at least it reduces the changes 
>>>>> throughout the driver, so that is an improvement.
>>>>>
>>>>> I guess one could do some of_clk_get() and clk_is_match() in probe 
>>>>> to look up the clock in OF by name and then compare which clock is 
>>>>> which before swapping them in clk_bulk[] array, but that might be 
>>>>> too convoluted?
>>>>
>>>> Yes, probably too much. What's present in the patch is not close to
>>>> perfection but has the advantage of being straightforward. If we agree
>>>> on that, I'll send a V3 containing the modifications in the bindings
>>>> file.
>>> Errr, I'm sorry, maybe there is a way to do this better. Look at 
>>> drivers/clk/clk-bulk.c :
>>>
>>>   15 static int __must_check of_clk_bulk_get(struct device_node *np, 
>>> int num_clks,
>>>   16                                         struct clk_bulk_data *clks)
>>>   17 {
>>>   18         int ret;
>>>   19         int i;
>>>   20
>>>   21         for (i = 0; i < num_clks; i++) {
>>>   22                 clks[i].id = NULL;
>>>   23                 clks[i].clk = NULL;
>>>   24         }
>>>   25
>>>   26         for (i = 0; i < num_clks; i++) {
>>>   27                 of_property_read_string_index(np, "clock-names", 
>>> i, &clks[i].id);
>>>   28                 clks[i].clk = of_clk_get(np, i);
>>>
>>> If I read this right, then clks[i].id should be the DT clock name. So 
>>> the swap conditional above could use .id to identify whether the 
>>> first position is core clock or not, like this:
>>>
>>> if (priv->data->nb_clock == 2 &&
>>>      strcmp(__clk_get_name(priv->clk_bulk[0].id), "core"))
>>>                                              ^^
>>>
>>> You might need to use devm_clk_bulk_get_all() to access the 
>>> of_clk_bulk_get() .
>>>
>>> Or am I missing something still ?
>>
>> Oooooh I see, devm_clk_bulk_get() and devm_clk_bulk_get_all() use
>> a different path. I don't understand why, to be honest... The doc
>> doesn't state this difference either.
> 
> Indeed, but maybe git log could clarify that ? I learnt about this 
> useful trick at last year Embedded Recipes:
> 
> $ git log -L:clk_bulk_get_all:drivers/clk/clk-bulk.c
> 

Will definitely keep that somewhere, thanks.

>> I'll give this a try while also correcting the issue that the robot
>> highlighted.
> Thank you !