Re: [PATCH] powerpc/paravirt: correct preempt debug splat in vcpu_is_preempted()

[Nathan Lynch <nathanl@linux.ibm.com>]

Srikar Dronamraju <srikar@linux.vnet.ibm.com> writes:
> * Nathan Lynch <nathanl@linux.ibm.com> [2021-09-20 22:12:13]:
>
>> vcpu_is_preempted() can be used outside of preempt-disabled critical
>> sections, yielding warnings such as:
>> 
>> BUG: using smp_processor_id() in preemptible [00000000] code: systemd-udevd/185
>> caller is rwsem_spin_on_owner+0x1cc/0x2d0
>> CPU: 1 PID: 185 Comm: systemd-udevd Not tainted 5.15.0-rc2+ #33
>> Call Trace:
>> [c000000012907ac0] [c000000000aa30a8] dump_stack_lvl+0xac/0x108 (unreliable)
>> [c000000012907b00] [c000000001371f70] check_preemption_disabled+0x150/0x160
>> [c000000012907b90] [c0000000001e0e8c] rwsem_spin_on_owner+0x1cc/0x2d0
>> [c000000012907be0] [c0000000001e1408] rwsem_down_write_slowpath+0x478/0x9a0
>> [c000000012907ca0] [c000000000576cf4] filename_create+0x94/0x1e0
>> [c000000012907d10] [c00000000057ac08] do_symlinkat+0x68/0x1a0
>> [c000000012907d70] [c00000000057ae18] sys_symlink+0x58/0x70
>> [c000000012907da0] [c00000000002e448] system_call_exception+0x198/0x3c0
>> [c000000012907e10] [c00000000000c54c] system_call_common+0xec/0x250
>> 
>> The result of vcpu_is_preempted() is always subject to invalidation by
>> events inside and outside of Linux; it's just a best guess at a point in
>> time. Use raw_smp_processor_id() to avoid such warnings.
>
> Typically smp_processor_id() and raw_smp_processor_id() except for the
> CONFIG_DEBUG_PREEMPT.

Sorry, I don't follow...

> In the CONFIG_DEBUG_PREEMPT case, smp_processor_id()
> is actually debug_smp_processor_id(), which does all the checks.

Yes, OK.

> I believe these checks in debug_smp_processor_id() are only valid for x86
> case (aka cases were they have __smp_processor_id() defined.)

Hmm, I am under the impression that the checks in
debug_smp_processor_id() are valid regardless of whether the arch
overrides __smp_processor_id().

I think the stack trace here correctly identifies an incorrect use of
smp_processor_id(), and the call site needs to be changed. Do you
disagree?