On Sat, Mar 02 2024 at 16:44, Thomas Gleixner wrote:
> On Sat, Mar 02 2024 at 12:37, Thomas Gleixner wrote:
> The below addresses _all_ percpu related sparse warnings except
> the ones in arch/x86/cpu/bugs.o but that's a sparse problem:
>
> The following is handled correctly:
>
> DECLARE_PER_CPU(u64, foo);
> this_cpu_read(foo);
>
> But this is not:
>
> DECLARE_PER_CPU(u64, foo);
> DEFINE_PER_CPU(u64, foo);
> this_cpu_read(foo);
>
> arch/x86/kernel/cpu/bugs.c:71:9: sparse: warning: incorrect type in initializer (different address spaces)
> arch/x86/kernel/cpu/bugs.c:71:9: sparse: expected void const [noderef] __percpu *__vpp_verify
> arch/x86/kernel/cpu/bugs.c:71:9: sparse: got unsigned long long *
>
> Commenting out the DEFINE_PER_CPU(u64, x86_spec_ctrl_current) in that
> file makes sparse happy, but that's obviously not a solution :)
Correction. I found the real issue:
DEFINE_PER_CPU(u64, x86_spec_ctrl_current);
EXPORT_SYMBOL_GPL(x86_spec_ctrl_current);
I had commented out both. But the real reason is the EXPORT_SYMBOL,
which obviously wants to be EXPORT_PER_CPU_SYMBOL_GPL...
So sparse was right. Nothing to see here.
Thanks,
tglx
