Re: [PATCH v3 16/22] ACPI: platform_profile: Set profile for all registered handlers

[Mario Limonciello <mario.limonciello@amd.com>]

On 10/31/2024 09:37, Ilpo Järvinen wrote:
> On Thu, 31 Oct 2024, Mario Limonciello wrote:
> 
>> On 10/31/2024 05:25, Ilpo Järvinen wrote:
>>> On Wed, 30 Oct 2024, Mario Limonciello wrote:
>>>
>>>> If multiple platform profile handlers have been registered then when
>>>> setting a profile verify that all profile handlers support the requested
>>>> profile and set it to each handler.
>>>>
>>>> If this fails for any given handler, revert all profile handlers back to
>>>> balanced and log an error into the kernel ring buffer.
>>>>
>>>> Tested-by: Matthew Schwartz <matthew.schwartz@linux.dev>
>>>> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
>>>> ---
>>>>    drivers/acpi/platform_profile.c | 25 ++++++++++++++++++++-----
>>>>    1 file changed, 20 insertions(+), 5 deletions(-)
>>>>
>>>> diff --git a/drivers/acpi/platform_profile.c
>>>> b/drivers/acpi/platform_profile.c
>>>> index 90cbc0de4d5bc..c2bb325ba531c 100644
>>>> --- a/drivers/acpi/platform_profile.c
>>>> +++ b/drivers/acpi/platform_profile.c
>>>> @@ -99,6 +99,8 @@ static ssize_t platform_profile_store(struct device
>>>> *dev,
>>>>    			    struct device_attribute *attr,
>>>>    			    const char *buf, size_t count)
>>>>    {
>>>> +	struct platform_profile_handler *handler;
>>>> +	unsigned long choices;
>>>>    	int err, i;
>>>>      	/* Scan for a matching profile */
>>>> @@ -107,16 +109,29 @@ static ssize_t platform_profile_store(struct device
>>>> *dev,
>>>>    		return -EINVAL;
>>>>      	scoped_cond_guard(mutex_intr, return -ERESTARTSYS,
>>>> &profile_lock) {
>>>> -		if (!cur_profile)
>>>> +		if (!platform_profile_is_registered())
>>>>    			return -ENODEV;
>>>>    -		/* Check that platform supports this profile choice */
>>>> -		if (!test_bit(i, cur_profile->choices))
>>>> +		/* Check that all handlers support this profile choice */
>>>> +		choices = platform_profile_get_choices();
>>>> +		if (!test_bit(i, &choices))
>>>>    			return -EOPNOTSUPP;
>>>>    -		err = cur_profile->profile_set(cur_profile, i);
>>>> -		if (err)
>>>> +		list_for_each_entry(handler, &platform_profile_handler_list,
>>>> list) {
>>>> +			err = handler->profile_set(handler, i);
>>>> +			if (err) {
>>>> +				pr_err("Failed to set profile for handler
>>>> %s\n", handler->name);
>>>> +				break;
>>>> +			}
>>>> +		}
>>>> +		if (err) {
>>>> +			list_for_each_entry_continue_reverse(handler,
>>>> &platform_profile_handler_list, list) {
>>>
>>> Too long line.
>>>
>>> This looks an error rollback though so instead of break inside the loop
>>> you could goto into a label at the end of the function and have much less
>>> indentation to begin with.
>>
>> How does the scoped_cond_guard interact with a goto?  With the jump I had
>> guessed it goes out of scope, but I wasn't really sure what the compiler does.
>>
>> I guess in the goto label I'll need another scoped_cond_guard()?
> 
> Ah, the scope problem is a good point.
> 
> Perhaps you could instead add e.g. platform_profile_reset_default() and
> call that before break, both patches that had the rollback did the same
> thing anyway so it can be reused too.
> 

Sounds like a good idea, thanks.