Re: [syzbot] WARNING in c_start

[Tetsuo Handa]

On 2022/10/16 9:28, Randy Dunlap wrote:
>> +/*
>> + * We want to avoid passing -1 as a valid cpu argument.
>> + * But we should not crash the kernel until all in-tree callers are fixed.
>> + */
> 
> Why not say that any negative cpu argument is invalid?

Currently only -1 is accepted as exception.

>>  	if (n != -1)
>>  		cpumask_check(n);
>> +	else
>> +		report_negative_cpuid();

> Or is it OK to pass -2 as the cpu arg?

Passing -2 will hit WARN_ON_ONCE(cpu >= nr_cpumask_bits) path.