linux-kernel - Re: [PATCH v3 3/8] ASoC: qcom: lpass: Use 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]
Message-ID: <ac634d4d-527a-7262-2744-203c04ee0982@linaro.org>
Date:   Thu, 9 Jul 2020 11:06:52 +0100
From:   Srinivas Kandagatla <srinivas.kandagatla@...aro.org>
To:     Rohit Kumar <rohitkr@...eaurora.org>, agross@...nel.org,
        bjorn.andersson@...aro.org, lgirdwood@...il.com,
        broonie@...nel.org, robh+dt@...nel.org, plai@...eaurora.org,
        bgoswami@...eaurora.org, perex@...ex.cz, tiwai@...e.com,
        linux-arm-msm@...r.kernel.org, alsa-devel@...a-project.org,
        devicetree@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v3 3/8] ASoC: qcom: lpass: Use regmap_field for i2sctl and
 dmactl registers



On 09/07/2020 10:57, Rohit Kumar wrote:
> Thanks Srini for reviewing.
> 
> On 7/9/2020 2:56 PM, Srinivas Kandagatla wrote:
>>
>>
>> On 08/07/2020 06:08, Rohit kumar wrote:
>>> I2SCTL and DMACTL registers has different bits alignment for newer
>>> LPASS variants of SC7180 soc. Use REG_FIELD_ID() to define the
>>> reg_fields in platform specific file and removed shifts and mask
>>> macros for such registers from header file.
>>>
>>> Signed-off-by: Rohit kumar <rohitkr@...eaurora.org>
>>
>> Thanks Rohit for doing this, this looks much better now!
>> I have few minor comments..
>>
>>> ---
>>>   sound/soc/qcom/lpass-apq8016.c   |  24 ++++++
>>>   sound/soc/qcom/lpass-cpu.c       | 163 
>>> +++++++++++++++++++++++----------------
>>>   sound/soc/qcom/lpass-ipq806x.c   |  24 ++++++
>>>   sound/soc/qcom/lpass-lpaif-reg.h | 157 
>>> +++++++++++++++++++------------------
>>>   sound/soc/qcom/lpass-platform.c  | 151 
>>> +++++++++++++++++++++++++++---------
>>>   sound/soc/qcom/lpass.h           |  53 +++++++++++++
>>>   6 files changed, 398 insertions(+), 174 deletions(-)
>>>
>>
>> index f0c7e93..f358d12 100644
>>> --- a/sound/soc/qcom/lpass-cpu.c
>>> +++ b/sound/soc/qcom/lpass-cpu.c
>>> @@ -29,6 +29,32 @@
>>>   #define LPASS_CPU_I2S_SD0_1_2_MASK    GENMASK(2, 0)
>>>   #define LPASS_CPU_I2S_SD0_1_2_3_MASK    GENMASK(3, 0)
>>
>>
>>>   }
>>>   -static int lpass_cpu_daiops_hw_free(struct snd_pcm_substream 
>>> *substream,
>>> -        struct snd_soc_dai *dai)
>>> -{
>>> -    struct lpass_data *drvdata = snd_soc_dai_get_drvdata(dai);
>>> -    int ret;
>>> -
>>> -    ret = regmap_write(drvdata->lpaif_map,
>>> -               LPAIF_I2SCTL_REG(drvdata->variant, dai->driver->id),
>>> -               0);
>>> -    if (ret)
>>> -        dev_err(dai->dev, "error writing to i2sctl reg: %d\n", ret);
>>> -
>>> -    return ret;
>>> -}
>>
>> Any particular reason why this function remove
> 
> This was causing issue in playback/capture concurrency. It sets I2SCTL 
> register value to 0
> 
> when usecase ends. However, playback/capture specific bits are already 
> cleared during trigger() stop
> 
> function. So, this is not needed.

This should be sent as separate fix with fixes tag!

> 
>
>>
>>> diff --git a/sound/soc/qcom/lpass-lpaif-reg.h 
>>> b/sound/soc/qcom/lpass-lpaif-reg.h
>>> index 72a3e2f..5258e60 100644
>>> --- a/sound/soc/qcom/lpass-lpaif-reg.h
>>> +++ b/sound/soc/qcom/lpass-lpaif-reg.h
>>> @@ -12,15 +12,12 @@
>> ...
>>>   #endif /* __LPASS_LPAIF_REG_H__ */
>>> diff --git a/sound/soc/qcom/lpass-platform.c 
>>> b/sound/soc/qcom/lpass-platform.c
>>> index 34f7fd1..445ca193 100644
>>> --- a/sound/soc/qcom/lpass-platform.c
>>> +++ b/sound/soc/qcom/lpass-platform.c
>>> @@ -50,6 +50,53 @@ static const struct snd_pcm_hardware 
>>> lpass_platform_pcm_hardware = {
>>>       .fifo_size        =    0,
>>>   };
>> ...
>>>   static int lpass_platform_pcmops_open(struct snd_soc_component 
>>> *component,
>>>                         struct snd_pcm_substream *substream)
>>>   {
>>> @@ -59,9 +106,9 @@ static int lpass_platform_pcmops_open(struct 
>>> snd_soc_component *component,
>>>       struct lpass_data *drvdata = 
>>> snd_soc_component_get_drvdata(component);
>>>       struct lpass_variant *v = drvdata->variant;
>>>       int ret, dma_ch, dir = substream->stream;
>>> -    struct lpass_pcm_data *data;
>>> +    struct lpass_pcm_data *data = NULL;
>>>   -    data = devm_kzalloc(soc_runtime->dev, sizeof(*data), GFP_KERNEL);
>>> +    data = kzalloc(sizeof(*data), GFP_KERNEL);
>>
>> Does this change belong in this patch?
> 
> 
> As part of this change, I fixed memory leak too by adding kfree() in 
> close()
> 
> However, this was causing issue as memory was allocated using 
> devm_kzalloc().
> 
> Should I move it to different patch?

That would be the right thing to do, can also add fixes tag!


> 
>>
>>>       if (!data)
>>>           return -ENOMEM;
>>>   @@ -111,13 +158,13 @@ static int lpass_platform_pcmops_close(struct 
>>> snd_soc_component *component,
>>>       struct snd_pcm_runtime *runtime = substream->runtime;
>>>       struct lpass_data *drvdata = 
>>> snd_soc_component_get_drvdata(component);
>>>       struct lpass_variant *v = drvdata->variant;

>>
>> Above two along with rddma members can be removed, these become 
>> redundant after adding regmap field!
>>
> wrdma_channels is used in alloc_dma_channel() to get the channel id.
> 
> Also, both are used for other DMA registers such as LPAIF_RDMABASE_REG,
> 
> LPAIF_RDMABUFF_REG, LPAIF_RDMACURR_REG, etc.
> 
Ah I see we are still using this in lpass_cpu_regmap_writeable!
ignore my previous comments about removing them!

--srini