Re: [PATCH v2] arm64: entry.S: Do not preempt from IRQ before all cpufeatures are enabled

[James Morse]

Hi Will,

On 15/10/2019 21:07, Will Deacon wrote:
> Patch looks good apart from one thing...
> 
> On Tue, Oct 15, 2019 at 06:25:44PM +0100, James Morse wrote:
>> diff --git a/include/linux/sched.h b/include/linux/sched.h
>> index 2c2e56bd8913..67a1d86981a9 100644
>> --- a/include/linux/sched.h
>> +++ b/include/linux/sched.h
>> @@ -223,6 +223,7 @@ extern long schedule_timeout_uninterruptible(long timeout);
>>  extern long schedule_timeout_idle(long timeout);
>>  asmlinkage void schedule(void);
>>  extern void schedule_preempt_disabled(void);
>> +asmlinkage void preempt_schedule_irq(void);
> 
> I don't understand the need for this hunk, since we're only calling the
> function from C now. Please could you explain?

(A prototype is needed to make the thing build[0], but)

you mean the asmlinkage?

The definition in kernel/sched/core.c has asmlinkage. It does nothing on arm64, but if
another architecture does add a C call, and uses asmlinkage to tinker with the calling
convention, it would need to be here so callers use the correct convention.

e.g. for X86_32 defines asmlinkage in arch/x86/include/asm/linkage.h:
| #define asmlinkage CPP_ASMLINKAGE __attribute__((regparm(0)))

This forces all arguments out of registers and onto the stack [1].

Without this annotation, asm->preempt_schedule_irq() callers would put arguments on the
stack, but C->preempt_schedule_irq() callers would use whatever the C->C calling
convention is, which might not match.

schedule() further up the hunk does the same.

I agree it doesn't matter today, but omitting it would be a bug for the next user to debug!


Thanks,

James

[0] Without that hunk,
../arch/arm64/kernel/process.c: In function âarm64_preempt_schedule_irqâ:
../arch/arm64/kernel/process.c:650:3: error: implicit declaration of function
âpreempt_schedule_irqâ; did you mean âpreempt_scheduleâ?
[-Werror=implicit-function-declaration]
   preempt_schedule_irq();
   ^~~~~~~~~~~~~~~~~~~~
   preempt_schedule
cc1: some warnings being treated as errors
make[3]: *** [../scripts/Makefile.build:266: arch/arm64/kernel/process.o] Error 1
make[3]: *** Waiting for unfinished jobs....
make[2]: *** [../scripts/Makefile.build:509: arch/arm64/kernel] Error 2
make[2]: *** Waiting for unfinished jobs....
make[1]: *** [/home/morse/kernel/linux/Makefile:1649: arch/arm64] Error 2
make[1]: *** Waiting for unfinished jobs....
make: *** [../Makefile:179: sub-make] Error 2

[1] https://gcc.gnu.org/onlinedocs/gcc/x86-Function-Attributes.html