Re: [peterz-queue:perf/core 26/39] arch/x86/include/asm/perf_event.h:290:19: error: flexible array member in a struct with no named members

[Liang, Kan <kan.liang@linux.intel.com>]

[-- Attachment #1: Type: text/plain, Size: 1977 bytes --]



On 7/7/2020 10:44 AM, Peter Zijlstra wrote:
> On Tue, Jul 07, 2020 at 08:51:10AM -0500, Gustavo A. R. Silva wrote:
>> On Tue, Jul 07, 2020 at 12:24:43PM +0200, Peter Zijlstra wrote:
>>> On Mon, Jul 06, 2020 at 09:08:41PM -0400, Liang, Kan wrote:
>>>
>>>>>      288	
>>>>>      289	struct pebs_lbr {
>>>>>    > 290		struct lbr_entry lbr[]; /* Variable length */
>>>>>      291	};
>>>>>      292	
>>>>
>>>> The struct pebs_lbr only has one member. To fix the warning, it seems we
>>>> have to define the struct as below.
>>>>
>>>>
>>>> diff --git a/arch/x86/include/asm/perf_event.h
>>>> b/arch/x86/include/asm/perf_event.h
>>>> index 2338200..a387e14 100644
>>>> --- a/arch/x86/include/asm/perf_event.h
>>>> +++ b/arch/x86/include/asm/perf_event.h
>>>> @@ -283,7 +283,7 @@ struct pebs_xmm {
>>>>   };
>>>>
>>>>   struct pebs_lbr {
>>>> -	struct lbr_entry lbr[]; /* Variable length */
>>>> +	struct lbr_entry lbr[0]; /* Variable length */
>>>>   };
>>>>
>>>
>>> Right, but then we get trouble like the below again. There's basically
>>
>> Yep, zero-length and one-element arrays are being deprecated[1].
>>
>>> no good solution here :-(
>>>
>>> Gustavo, any clues?
>>>
>>
>> Yep; the issue is that "Flexible array members may only appear as the
>> last member of a struct that is otherwise non-empty"[2].
>>
>> The solution is to declare something like this:
>>
>> struct pebs_lbr {
>> 	size_t num_lbr; /* count for number of lbr entries */
>> 	struct lbr_entry lbr[];
>> };
>>
> 
> Can't.. this is hardware provided layout. The thing is that the array
> size is variable, so [0] or [] would be correct.
> 
> That said, Kan, I suppose we can simply remove struct pebs_lbr and use
> struct lbr_entry * directly. There is very little actual point in having
> struct pebs_lbr.
> 

Yes, I will do more tests with the patch I just sent in another thread.

> I'll frob the patches and push out again.
> 

Thanks.

Kan