Skip to contentSkip to conversation

Setting /proc/irq/[n]/smp_affinity failure on 8 cpu systems

RFC PATCH John Blackwood 1 participant 1 message Dec 3, 2009

lore
Share & tools
lore
#0 lore raw
Quoted 26 lines · John Blackwood
Quoted 21 lines
Quoted 21 lines

I would like to mention an irq set affinity issue occurs on an older Dell PowerEdge 6650 system:
- 8 cpus (4 are hyper-threaded),
- uses the struct apic 'apic_default' support
- 32 bit (x86) kernel

This issue is that once you change the irq smp_affinity of an irq to be less than all cpus in the system, you can never change really the
irq smp_affinity back to be all cpus in the system (0xff) again, even though no error status is returned on the "/bin/echo ff > /proc/irq/[n]/smp_affinity" operation.

This is due to that fact that BAD_APICID has the same value as all cpus (0xff) on 32bit kernels, and thus the value returned from set_desc_affinity() via the cpu_mask_to_apicid_and() function is treated
as a failure in set_ioapic_affinity_irq_desc(), and no affinity changes
are made.

Quoted 3 lines
Quoted 3 lines

It seems like changing the BAD_APICID value from 0xFF to 0xFFFF for CONFIG_X86_32 might be one possible solution, and it seems to function properly for this Dell PowerEdge 6650 system.

Quoted 6 lines

John, Thanks for the root cause and analysis. As you noticed, 0xff is a valid logical apicid. Similarly 0xffff is a valid logical x2apic cluster id. So while your patch fixes your issue, it doesn't completely fix the issue. Can you please check if the appended patch works and ack if it passes your test?

Hi Suresh,

Yes, your patch works on the Dell PowerEdge 6650 system.

Thanks for your help and quick reply.