linux-kernel - Re: [PATCH 2/4] ASoC: codecs: lpass-rx-macro: Keep static regmap

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]

Date: Fri, 28 Jun 2024 10:39:07 +0200
From: Krzysztof Kozlowski <krzysztof.kozlowski@...aro.org>
To: Dmitry Baryshkov <dmitry.baryshkov@...aro.org>
Cc: Srinivas Kandagatla <srinivas.kandagatla@...aro.org>,
 Banajit Goswami <bgoswami@...cinc.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,
 linux-arm-msm@...r.kernel.org, linux-sound@...r.kernel.org,
 linux-kernel@...r.kernel.org
Subject: Re: [PATCH 2/4] ASoC: codecs: lpass-rx-macro: Keep static
 regmap_config as const

On 28/06/2024 10:34, Dmitry Baryshkov wrote:
> On Thu, Jun 27, 2024 at 05:23:44PM GMT, Krzysztof Kozlowski wrote:
>> The driver has static 'struct regmap_config', which is then customized
>> depending on device version.  This works fine, because there should not
>> be two devices in a system simultaneously and even less likely that such
>> two devices would have different versions, thus different regmap config.
>> However code is cleaner and more obvious when static data in the driver
>> is also const - it serves as a template.
>>
>> Mark the 'struct regmap_config' as const and duplicate it in the probe()
>> with devm_kmemdup to allow customizing per detected device variant.
>>
>> Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@...aro.org>
>> ---
>>  sound/soc/codecs/lpass-rx-macro.c | 17 +++++++++++++----
>>  1 file changed, 13 insertions(+), 4 deletions(-)
> 
> Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@...aro.org>
> 
>>
>> diff --git a/sound/soc/codecs/lpass-rx-macro.c b/sound/soc/codecs/lpass-rx-macro.c
>> index 59fe76b13cdb..3d8149665439 100644
>> --- a/sound/soc/codecs/lpass-rx-macro.c
>> +++ b/sound/soc/codecs/lpass-rx-macro.c
>> @@ -1662,7 +1662,7 @@ static bool rx_is_readable_register(struct device *dev, unsigned int reg)
>>  	return rx_is_rw_register(dev, reg);
>>  }
>>  
>> -static struct regmap_config rx_regmap_config = {
>> +static const struct regmap_config rx_regmap_config = {
>>  	.name = "rx_macro",
>>  	.reg_bits = 16,
>>  	.val_bits = 32, /* 8 but with 32 bit read/write */
>> @@ -3765,6 +3765,7 @@ static const struct snd_soc_component_driver rx_macro_component_drv = {
>>  static int rx_macro_probe(struct platform_device *pdev)
>>  {
>>  	struct reg_default *reg_defaults;
>> +	struct regmap_config *reg_config;
>>  	struct device *dev = &pdev->dev;
>>  	kernel_ulong_t flags;
>>  	struct rx_macro *rx;
>> @@ -3851,14 +3852,22 @@ static int rx_macro_probe(struct platform_device *pdev)
>>  		goto err;
>>  	}
>>  
>> -	rx_regmap_config.reg_defaults = reg_defaults;
>> -	rx_regmap_config.num_reg_defaults = def_count;
>> +	reg_config = devm_kmemdup(dev, &rx_regmap_config, sizeof(*reg_config),
>> +				  GFP_KERNEL);
>> +	if (!reg_config) {
>> +		ret = -ENOMEM;
>> +		goto err;
>> +	}
>>  
>> -	rx->regmap = devm_regmap_init_mmio(dev, base, &rx_regmap_config);
>> +	reg_config->reg_defaults = reg_defaults;
>> +	reg_config->num_reg_defaults = def_count;
>> +
>> +	rx->regmap = devm_regmap_init_mmio(dev, base, reg_config);
>>  	if (IS_ERR(rx->regmap)) {
>>  		ret = PTR_ERR(rx->regmap);
>>  		goto err;
>>  	}
>> +	devm_kfree(dev, reg_config);
>>  	devm_kfree(dev, reg_defaults);
> 
> Seeing devm_kfree in the non-error path makes me feel strange. Maybe
> it's one of the rare occasions when I can say that __free is suitable
> here.

These would have a bit different meaning in such case. The __free would
not clean it in this spot, but on exit from the scope. I wanted to
kfree() here, because the config (and reg_defaults) are not used by past
regmap_init. I mentioned it briefly in previous patch msg.

To me this code is readable and obvious - past this point nothing uses
that allocation. However maybe instead of devm(), the code would be
easier to read if non-devm-malloc + __free()?

Best regards,
Krzysztof