Re: [Xen-devel] [PATCH 1/1] x86/xen: fix section of xen_init_time_ops() in header

[Jan Beulich]

>>> On 04.09.17 at 12:35, <andrew.cooper3@xxxxxxxxxx> wrote:
> On 04/09/17 11:15, Jan Beulich wrote:
>>>>> On 04.09.17 at 10:17, <jgross@xxxxxxxx> wrote:
>>> On 03/09/17 10:38, Nicolas Iooss wrote:
>>>> Commit d162809f85b4 ("xen/x86: Do not call xen_init_time_ops() until
>>>> shared_info is initialized") moved xen_init_time_ops() from __init to
>>>> __ref without updating xen-ops.h accordingly. Fix this.
>>>>
>>>> Fixes: d162809f85b4 ("xen/x86: Do not call xen_init_time_ops() until
>>>> shared_info is initialized")
>>>> Signed-off-by: Nicolas Iooss <nicolas.iooss_linux@xxxxxxx>
>>>> ---
>>>>  arch/x86/xen/xen-ops.h | 2 +-
>>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/arch/x86/xen/xen-ops.h b/arch/x86/xen/xen-ops.h
>>>> index 0d5004477db6..b2a5d48a2c2a 100644
>>>> --- a/arch/x86/xen/xen-ops.h
>>>> +++ b/arch/x86/xen/xen-ops.h
>>>> @@ -69,7 +69,7 @@ void xen_setup_runstate_info(int cpu);
>>>>  void xen_teardown_timer(int cpu);
>>>>  u64 xen_clocksource_read(void);
>>>>  void xen_setup_cpu_clockevents(void);
>>>> -void __init xen_init_time_ops(void);
>>>> +void __ref xen_init_time_ops(void);
>>>>  void __init xen_hvm_init_time_ops(void);
>>> When correcting this could you please modify the prototypes to comply to
>>> the intended form as noted in include/linux/init.h (the __ref or __init
>>> annotations should be just before the ending semicolon)?
>> Why would these annotations be kept on the declarations anyway?
>> Attributes affecting code/data placement generally belong on the
>> definitions only.
> 
> Because:
> 
> a) Thatâs what the coding style says, and
> 
> b) So various static analysis can be done (e.g. sparse) on an individual
> translation unit basis.
> 
> 
> Your objection to having annotations on declarations is why I've never
> got around to adding sparse to the hypervisor build.

Would you mind educating me what use these annotations can be
for static analysis? If they're useful in headers, I would stop
objecting to them being added there, but I'd then demand for them
to never be present on non-static definitions (unless there are
attributes where the compiler requires them to be repeated, but I
think all attributes are cumulative).

Jan