Skip to content

x86: cleanups and fixes for irq migration code during cpu offline

Suresh Siddha lkml 1 version In review

Get series with b4
b4 am [email protected]

Versions newest first

  1. v1 Latest6 patches 18 repliesIn review

Patches in v1 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/6x86: unify fixup_irqs() for 32-bit and 64-bit kernelsarch/x86/kernel/irq.c arch/x86/kernel/irq_32.c arch/x86/kernel/irq_64.c+59 -1031 reply
2/6x86, intr-remap: Avoid irq_chip mask/unmask in fixup_irqs() for intr-remappingarch/x86/kernel/irq.c+2 -21 reply
3/6x86: remove move_cleanup_count from irq_cfgarch/x86/include/asm/hw_irq.h arch/x86/kernel/apic/io_apic.c+1 -91 reply
4/6x86: force irq complete move during cpu offlinearch/x86/include/asm/irq.h arch/x86/kernel/apic/io_apic.c arch/x86/kernel/irq.c+23 -31 reply
5/6x86: Use EOI register in io-apic on intel platformsarch/x86/kernel/apic/io_apic.c+53 -2710 replies
6/6x86: remove local_irq_enable()/local_irq_disable() in fixup_irqs()arch/x86/kernel/irq.c+28 -41 reply

Reviews trailers per patch

5 of 6 patches have a Reviewed-by or Acked-by.

PatchEric W. BiedermanGary HadeRajesh SankaranMaciej W. RozyckiSuresh Siddha
1/6
2/6A-b A-b
3/6A-b
4/6A-b
5/6A-b A-b A-b R-b
6/6A-b

Trailers given in replies are filled; those carried in the patches from earlier versions are outlined with dashes. Scroll the table sideways for all 5 reviewers.

Read the v1 thread

Cover letter [patch 0/6] x86: cleanups and fixes for irq migration code during cpu offline

First four patches in the series unify the fixup_irqs() along with couple of cleanups. It also fixes an issue where the interrupt subsystem can point the interrupt to the offlined cpu (for non interrupt-remapping case) causing the device to not work. This was observed by Gary.

These four patches are ready for inclusion and while there was a debate in the past that fixup_irqs() is kind of broken (because of migrating interrupts in the process context for non intr-remapping platforms) we think that these patches enhance the existing code and not introduce any more new races.

The last two patches in the series are titled as RFC. These patches address removing the local_irq_enable()/local_irq_disable() in the fixup_irqs(). All the patches (including these last two RFC patches) are tested on Intel platforms (by myself) and on AMD platforms(by Gary).

We labeled the last two patches as RFC mainly because they need more testing and thoughts from Eric and others who did some unsuccessful experiments in this area (migrating irq in the process context for non intr-remapping platforms) before.

thanks,
suresh