Re: [PATCH v3] ASoC: amd: acp: Initialize list to store acp_stream during pcm_open

["Prasad, Prasad" <venkataprasad.potturu@amd.com>]

On 7/29/2022 4:19 PM, Takashi Iwai wrote:
Thanks for your time.
> [CAUTION: External Email]
>
> On Fri, 29 Jul 2022 12:34:51 +0200,
> Venkata Prasad Potturu wrote:
>> On 7/28/22 18:19, Mark Brown wrote:
>> Thanks for your time.
>>
>>      On Thu, Jul 28, 2022 at 06:10:50PM +0530, Venkata Prasad Potturu wrote:
>>
>>          @@ -104,14 +105,13 @@ static irqreturn_t i2s_irq_handler(int irq, void *data)
>>
>>                ext_intr_stat = readl(ACP_EXTERNAL_INTR_STAT(adata, rsrc->irqp_used));
>>
>>          -     for (i = 0; i < ACP_MAX_STREAM; i++) {
>>          -                           stream = adata->stream[i];
>>          +     spin_lock_irqsave(&adata->acp_lock, flags);
>>          +     list_for_each_entry(stream, &adata->stream_list, list) {
>>
>>      If we're already in an interrupt handler here (presumably not a threaded
>>      one) why are we using irqsave?
>>
>> Yes, your statement make sense, I have followed below statement in kernel
>> document. so used irqsave in interrupt context as well.
>>
>> We will change it to spin_lock() and send it in the next version.
>>
>> statement:- spin_lock_irqsave() will turn off interrupts if they are on,
>> otherwise does nothing (if we are already in an interrupt handler), hence
>> these functions are safe to call from any context.
> Also the open and close callbacks are certainly non-irq context, hence
> you can use spin_lock_irq() instead of irqsave(), too.

One doubt, if we use *spin_lock_irq()* it will do *local_irq_disable() 
*even if interrupts are already disabled, and when we call 
*spin_unlock_irq() *after critical section it  will forcibly re-enable 
interrupts in a potentially unwanted manner.

If we use *spin_lock_irqsave();* and *spin_unlock_irqrestore(); *will 
save and restore the interrupt state.

My understanding is *spin_lock_irqsave();* is better instead of improper 
interrupt enable and disable as like *spin_lock_irq();*

Could you please let me know is there any other thing to use 
*spin_lock_irq(); *particularly here.

> Takashi