Re: [PATCH] pinctrl: renesas: rzg2l: Fix double unlock in rzg2l_dt_subnode_to_map()

[claudiu beznea <claudiu.beznea@xxxxxxxxx>]

On 12.01.2024 11:53, Dan Carpenter wrote:
> On Fri, Jan 12, 2024 at 10:55:40AM +0200, claudiu beznea wrote:
>> Hi, Dan,
>>
>> Thanks for your patch!
>>
>> On 10.01.2024 20:41, Dan Carpenter wrote:
>>> If rzg2l_map_add_config() fails then the error handling calls
>>> mutex_unlock(&pctrl->mutex) but we're not holding that mutex.  Move
>>> the unlocks to before the gotos to avoid this situation.
>>>
>>> Fixes: d3aaa7203a17 ("pinctrl: renesas: rzg2l: Add pin configuration support for pinmux groups")
>>> Signed-off-by: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
>>> ---
>>> (Not tested).
>>
>> I've tested it on RZ/G3S SoC and all is good.
>>
>> However, I think, to keep the locking scheme unchanged and simpler (FMPOV),
>> commit d3aaa7203a17 ("pinctrl: renesas: rzg2l: Add pin configuration
>> support for pinmux groups") should have been call rzg2l_map_add_config()
>> just before the mutex is locked. That would be the following diff:
>>
>> --- a/drivers/pinctrl/renesas/pinctrl-rzg2l.c
>> +++ b/drivers/pinctrl/renesas/pinctrl-rzg2l.c
>> @@ -447,6 +447,16 @@ static int rzg2l_dt_subnode_to_map(struct pinctrl_dev
>> *pctldev,
>>  		name = np->name;
>>  	}
>>
>> +	if (num_configs) {
>> +		ret = rzg2l_map_add_config(&maps[idx], name,
>> +					   PIN_MAP_TYPE_CONFIGS_GROUP,
>> +					   configs, num_configs);
>> +		if (ret < 0)
>> +			goto done;
>> +
>> +		idx++;
>> +	}
>> +
>>  	mutex_lock(&pctrl->mutex);
>>
>>  	/* Register a single pin group listing all the pins we read from DT */
>> @@ -474,16 +484,6 @@ static int rzg2l_dt_subnode_to_map(struct pinctrl_dev
>> *pctldev,
>>  	maps[idx].data.mux.function = name;
>>  	idx++;
>         ^^^^^

This needs to be here for subsequent calls of rzg2l_dt_subnode_to_map() to
know which entry in maps[] to be populated next time.

> 
>>
>> -	if (num_configs) {
>> -		ret = rzg2l_map_add_config(&maps[idx], name,
>> -					   PIN_MAP_TYPE_CONFIGS_GROUP,
>> -					   configs, num_configs);
>> -		if (ret < 0)
>> -			goto remove_group;
>> -
>> -		idx++;
>> -	}
> 
> Does the ordering of the maps[] not matter?

It doesn't matter, AFAIKT. The core code checks for map type (e.g.
PIN_MAP_TYPE_CONFIGS_GROUP) when processes the data from maps[].

> 
>> -
>>  	dev_dbg(pctrl->dev, "Parsed %pOF with %d pins\n", np, num_pinmux);
>>  	ret = 0;
>>  	goto done;
>>
>> Would you mind doing it like this?
>>
>> Please, let me know if you want me to handle it.
> 
> Either way is fine.  Whatever is easiest.

Ok, I'll prepare a patch as I already tested it on my side on multiple
platforms.

Thank you,
Claudiu Beznea

> 
> regards,
> dan carpenter
>