Page MenuHomeFreeBSD

LinuxKPI: Allow spin_lock_irqsave to be called within a critical section
ClosedPublic

Authored by wulf on Jan 13 2022, 11:48 PM.

Details

Summary

with spinning on spin_trylock. dma-buf part of drm-kmod depends on this
property and absence of it support results in "mi_switch: switch in a
critical section" assertions [1][2].

[1] https://github.com/freebsd/drm-kmod/issues/116
[2] https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=261166

MFC after: 1 week

Diff Detail

Repository
rG FreeBSD src repository
Lint
Automatic diff as part of commit; lint not applicable.
Unit
Automatic diff as part of commit; unit tests not applicable.

Event Timeline

wulf requested review of this revision.Jan 13 2022, 11:48 PM
This revision is now accepted and ready to land.Jan 14 2022, 6:28 AM
sys/compat/linuxkpi/common/include/linux/spinlock.h
120

Is this a property of spin_lock_irqsave() only or should other spin_lock() functions be updated aswell?

sys/compat/linuxkpi/common/include/linux/spinlock.h
120

AFAIK no, they should not. Ordinary spin_lock() requires fully preemptible context at least on RT-kernels. See PREEMPT_RT caveats section of https://www.kernel.org/doc/html/latest/locking/locktypes.html

May be some other spin_lock_XXX functions e.g. spin_lock_irqsave_nested() can be updated too, but I did not dig in to this direction.

sys/compat/linuxkpi/common/include/linux/spinlock.h
120

Please updsate spin_lock_irqsave_nested() aswell and let's get moving :-)