Re: [PATCH v2] hwrng: core - Do not read data during PM sleep transition

[Thomas Richard]

On 6/11/26 7:16 AM, Herbert Xu wrote:
> On Mon, Jun 01, 2026 at 03:19:13PM +0200, Thomas Richard (TI) wrote:
>>
>> @@ -538,8 +539,9 @@ static int hwrng_fillfn(void *unused)
>>  		}
>>  
>>  		mutex_lock(&reading_mutex);
>> -		rc = rng_get_data(rng, rng_fillbuf,
>> -				  rng_buffer_size(), 1);
>> +		if (!pm_sleep_transition_in_progress())
>> +			rc = rng_get_data(rng, rng_fillbuf,
>> +					  rng_buffer_size(), 1);
> 
> Sashiko asks how this can be safe given that there is no locking
> at all.  That's exactly what I was going to ask :)

Yes indeed, you are right. I dug again the topic (including the freeze
solution I implemented in a first time), and I may have found the right
solution. The only way is to stop the hwrng_fillfn() kthread before
suspend and restart it after resume. Fortunately there is a mechanism to
do this: pm_notifier. I'll send a patch soon.

Best Regards,
Thomas