Re: [PATCH 1/2] xen: make HYPERVISOR_get_debugreg() always_inline

[Andrew Cooper]

On 25/11/2021 15:04, Juergen Gross wrote:
> On 25.11.21 16:00, Andrew Cooper wrote:
>> On 25/11/2021 09:20, Juergen Gross wrote:
>>> HYPERVISOR_get_debugreg() is being called from noinstr code, so it
>>> should be attributed "always_inline".
>>>
>>> Fixes: f4afb713e5c3a4419ba ("x86/xen: Make get_debugreg() noinstr")
>>> Reported-by: kernel test robot <lkp@xxxxxxxxx>
>>> Signed-off-by: Juergen Gross <jgross@xxxxxxxx>
>>> ---
>>>   arch/x86/include/asm/xen/hypercall.h | 2 +-
>>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/arch/x86/include/asm/xen/hypercall.h
>>> b/arch/x86/include/asm/xen/hypercall.h
>>> index 0575f5863b7f..28ca1119606b 100644
>>> --- a/arch/x86/include/asm/xen/hypercall.h
>>> +++ b/arch/x86/include/asm/xen/hypercall.h
>>> @@ -287,7 +287,7 @@ HYPERVISOR_set_debugreg(int reg, unsigned long
>>> value)
>>>       return _hypercall2(int, set_debugreg, reg, value);
>>>   }
>>>   -static inline unsigned long
>>> +static __always_inline unsigned long
>>>   HYPERVISOR_get_debugreg(int reg)
>>>   {
>>>       return _hypercall1(unsigned long, get_debugreg, reg);
>>
>> All this is going to do is push the error one step further in.
>>
>> Next, objtool will complain that hypercall_page isn't noinstr.
>
> I don't think so:
>
> .pushsection .noinstr.text, "ax"
>         .balign PAGE_SIZE
> SYM_CODE_START(hypercall_page)
>         .rept (PAGE_SIZE / 32)
>                 UNWIND_HINT_FUNC
>                 .skip 31, 0x90
>                 ret
>         .endr

/sigh yes.  It would help if I was reading an up to date tree.

Sorry for the noise.

~Andrew