On May 13, 2021 / 12:37, Mike Christie wrote: > On 5/13/21 11:03 AM, Bart Van Assche wrote: > > On 5/13/21 8:36 AM, Mike Christie wrote: > >> Is there something else that's used normally? raw_smp_processor_id? > > > > +1 for raw_smp_processor_id() since my understanding is that > > SCF_USE_CPUID is a performance optimization only and correctness of the > > target code is not affected if the running thread is moved to another > > CPU core by the scheduler. > > > > Ok. > > Shin'ichiro if we all are in agreement then I made the same mistake in > iblock_plug_device. I can send a patch for that, or if you want to get > your patch commit count up, feel free to when you resend this patch. Thank you for the discussion. I also agree to use raw_smp_processor_id() instead of get_cpu()/put_cpu(). Will post v2. Regarding iblock_plug_device, may I ask Mike to send a patch? I'm a bit caught up in other tasks. -- Best Regards, Shin'ichiro Kawasaki
