linux-kernel - Re: [PATCH next 4/4] iio: sca3000: stop interrupts via devm_add_action_or

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: <464ec1bb-858d-4ff5-8fa0-1c9af9b9945e@oracle.com>
Date: Sat, 31 Jan 2026 23:22:24 +0530
From: Harshit Mogalapalli <harshit.m.mogalapalli@...cle.com>
To: Jonathan Cameron <jic23@...nel.org>
Cc: David Lechner <dlechner@...libre.com>,
        Nuno Sá
 <nuno.sa@...log.com>,
        Andy Shevchenko <andy@...nel.org>,
        Gustavo Bastos <gustavobastos@....br>,
        Andrew Ijano
 <andrew.ijano@...il.com>,
        "open list:IIO SUBSYSTEM AND DRIVERS" <linux-iio@...r.kernel.org>,
        open list <linux-kernel@...r.kernel.org>,
        kernel-janitors@...r.kernel.org,
        Andy Shevchenko <andriy.shevchenko@...el.com>
Subject: Re: [PATCH next 4/4] iio: sca3000: stop interrupts via
 devm_add_action_or_reset()

Hi Jonathan,


On 31/01/26 22:09, Jonathan Cameron wrote:
> On Fri, 30 Jan 2026 13:43:11 -0800
> Harshit Mogalapalli <harshit.m.mogalapalli@...cle.com> wrote:
> 
>> sca3000_stop_all_interrupts() is moved above the probe routine so the
>> new function sca3000_disable_interrupts() used in probe can directly
>> call it without additional declaration.
>>
>> Used devm_add_action_or_reset() for shutting doing the interrupts. After
>> this there is no need to have a remove call back, so got rid of the
>> remove callback.
>>
>> No functional change intended.
>>
>> Suggested-by: Andy Shevchenko <andriy.shevchenko@...el.com>
>> Signed-off-by: Harshit Mogalapalli <harshit.m.mogalapalli@...cle.com>
> 
> Will need an update if you indeed didn't intend to change order in previous
> patch.  

Yes, I didn't intend that change in Patch 3.

> devm_iio_device_register() is last in vast majority of probe
> functions because it opens us up to userspace interaction. We almost
> always want to cut that off on remove before doing anything else.
> 

Sure, thanks a lot!

I have checked in other drivers while thinking about it, but my code 
base changed due to incorrect ordering in PATCH 3 :(

Maybe the best idea is to do squash PATCH3 and PATCH 4 ?

>> ---
>>   drivers/iio/accel/sca3000.c | 58 ++++++++++++++++++++-----------------
>>   1 file changed, 32 insertions(+), 26 deletions(-)
>>
>> diff --git a/drivers/iio/accel/sca3000.c b/drivers/iio/accel/sca3000.c
>> index bf1957c7bc77..7f7d29688a81 100644
>> --- a/drivers/iio/accel/sca3000.c
>> +++ b/drivers/iio/accel/sca3000.c
>> @@ -1437,6 +1437,34 @@ static const struct iio_info sca3000_info = {
>>   	.write_event_config = &sca3000_write_event_config,
>>   };
>>   
>> +static int sca3000_stop_all_interrupts(struct sca3000_state *st)
>> +{
>> +	int ret;
>> +
>> +	mutex_lock(&st->lock);
> 
> Use guard(mutex)(&st->lock); to simplify this. Remember to include cleanup.h
> A future patch could then make use of guard() more widely in this driver.
> 
>> +	ret = sca3000_read_data_short(st, SCA3000_REG_INT_MASK_ADDR, 1);
>> +	if (ret)
>> +		goto error_ret;
> 
> Blank line here.  Keeps each group of action / error check clearly delineated
> if we have a blank line between them. Note this is a case of do as I say
> not as I did nearly 17 years back when I wrote this (younger me did many
> things wrong ;)
> 
> With guard() above, you can just do
> 	if (ret)
> 		return ret;
> 
> here.
>> +	ret = sca3000_write_reg(st, SCA3000_REG_INT_MASK_ADDR,
>> +				(st->rx[0] &
>> +				 ~(SCA3000_REG_INT_MASK_RING_THREE_QUARTER |
>> +				   SCA3000_REG_INT_MASK_RING_HALF |
>> +				   SCA3000_REG_INT_MASK_ALL_INTS)));
> 
> With guard() above, this becomes
> 	return sca3000_write_reg();
> 
> 
>> +error_ret:
>> +	mutex_unlock(&st->lock);
>> +	return ret;
>> +}
>> +


Sure will simplify this function with scoped guards. Thanks for the 
suggestion.

>> +static void sca3000_disable_interrupts(void *data)
>> +{
>> +	struct iio_dev *indio_dev = data;
>> +	struct sca3000_state *st = iio_priv(indio_dev);
>> +
>> +	/* Must ensure no interrupts can be generated after this! */
>> +	sca3000_stop_all_interrupts(st);
>> +}
>> +
>> +
>>   static int sca3000_probe(struct spi_device *spi)
>>   {
>>   	const struct spi_device_id *id = spi_get_device_id(spi);
>> @@ -1481,6 +1509,7 @@ static int sca3000_probe(struct spi_device *spi)
>>   		if (ret)
>>   			return ret;
>>   	}
>> +
> 
> Unrelated change.  Make sure to check for these in patches before
> you send them out.  It adds noise and typically means you'll end
> up doing another version even if everything else is good.
> 

sure, I was actually aware of this new line addition while sending, 
wasn't sure if I had to do it in this or do it in a separate patch. But 
I understood that now to not mix unrelated changes.


>>   	ret = sca3000_clean_setup(st);
>>   	if (ret)
>>   		return ret;
...
>> +		return ret;
>>   
>> -static void sca3000_remove(struct spi_device *spi)
>> -{
>> -	struct iio_dev *indio_dev = spi_get_drvdata(spi);
>> -	struct sca3000_state *st = iio_priv(indio_dev);
>> +	return devm_add_action_or_reset(dev, sca3000_disable_interrupts, indio_dev);
> As mentioned above. This looks unlikely to be a good idea as a reorder of code.
> 
> I'd expect interfaces to go away and then interrupts to be stopped. In general
> that should always be safe unless we have some racey bug somewhere in the IIO
> core or the driver is doing something unusual.
> 

Sure, so iio_device_unregister() is the first thing to happen. Will do a 
v2 trying to address all comments.


Thanks a lot for detailed review and guidance with this! I really
appreciate it.

> Thanks,
> 
> Jonathan
> 

Regards,
Harshit