linux-kernel - Re: About irq_create_affinity_masks() for a platform device driver

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: <871rrazp31.fsf@nanos.tec.linutronix.de>
Date:   Tue, 04 Feb 2020 09:20:02 +0000
From:   Thomas Gleixner <tglx@...utronix.de>
To:     John Garry <john.garry@...wei.com>
Cc:     Marc Zyngier <maz@...nel.org>,
        "linux-kernel\@vger.kernel.org" <linux-kernel@...r.kernel.org>,
        chenxiang <chenxiang66@...ilicon.com>
Subject: Re: About irq_create_affinity_masks() for a platform device driver

John,

John Garry <john.garry@...wei.com> writes:
>> I wouldn't mind to expose a function which allows you to switch the
>> allocated interrupts to managed. The reason why we do it in one go in
>> the PCI code is that we get automatically the irq descriptors allocated
>> on the correct node. So if the node aware allocation is not a
>> showstopper 
>
> I wouldn't say so for now.

Good.

> for this then your function would do:
>> 
>> 	...
>> 	for (i = 0; i < count; i++) {
>> 		pirqs[i] = platform_get_irq(dev, i);
>> 
>>                  irq_update_affinity_desc(pirqs[i], affdescs + i);
>> 
>>          }
>> 
>> int irq_update_affinity_desc(unsigned int irq, irq_affinity_desc *affinity)
>> {
>> 	unsigned long flags;
>> 	struct irq_desc *desc = irq_get_desc_lock(irq, &flags, 0);
>> 
>>          if (!desc)
>>          	return -EINVAL;
>> 
>>          if (affinity->is_managed) {
>>          	irqd_set(&desc->irq_data, IRQD_IRQ_DISABLED);
>> 	        irqd_set(&desc->irq_data, IRQD_IRQ_MASKED);
>
> Are these correct? I assume we want to follow alloc_descs() here.

Yeah, copied the wrong chunk :)

>>          }
>>          cpumask_copy(desc->irq_common_data.affinity, affinity);
>>          return 0;
>> }
>
> I see. So I made a couple of changes and it did work:
>
> int irq_update_affinity_desc(unsigned int irq, struct irq_affinity_desc 
> *affinity)
> {
> 	unsigned long flags;
> 	struct irq_desc *desc = irq_get_desc_lock(irq, &flags, 0);
>
> 	if (!desc)
> 		return -EINVAL;
>
> 	if (affinity->is_managed) {
> 		irqd_set(&desc->irq_data, IRQD_AFFINITY_MANAGED);
> 		irqd_set(&desc->irq_data, IRQD_MANAGED_SHUTDOWN);
> 	}
>
> 	cpumask_copy(desc->irq_common_data.affinity, &affinity->mask);
> 	irq_put_desc_unlock(desc, flags);
> 	return 0;
> }

Looks correct.

> And if we were to go this way, then we don't need to add the pointer in 
> struct platform_device to hold affinity mask descriptors as we're using 
> them immediately. Or even have a single function to do it all in the irq 
> code (create the masks and update the affinity desc).
>
> And since we're just updating the masks, I figure we shouldn't need to 
> add acpi_irq_get_count(), which I invented to get the irq count (without 
> creating the IRQ mapping).

Yes, you can create and apply the masks after setting up the interrupts.

Thanks,

        tglx