Re: linux-next: build failure after merge of the akpm-current tree

[Zi Yan]

On 7 Sep 2017, at 3:46, Andrew Morton wrote:

> On Thu, 7 Sep 2017 15:23:55 +1000 Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx> wrote:
>
>> Hi all,
>>
>> On Wed, 2 Aug 2017 16:31:45 +1000 Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx> wrote:
>>>
>>> On Wed, 2 Aug 2017 15:45:54 +1000 Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx> wrote:
>>>>
>>>> On Tue, 01 Aug 2017 09:08:01 -0400 "Zi Yan" <zi.yan@xxxxxxxxxxxxxx> wrote:
>>>>>
>>>>> I found two possible fixes.
>>>>>
>>>>> 1. This uses C++ zero initializer, GCC is OK with it.
>>>>> I tested with GCC 4.9.3 (has the initialization bug) and GCC 6.4.0.
>>>>>
>>>>> --- a/include/linux/swapops.h~a
>>>>> +++ a/include/linux/swapops.h
>>>>> @@ -217,7 +217,7 @@ static inline swp_entry_t pmd_to_swp_ent
>>>>>
>>>>>  static inline pmd_t swp_entry_to_pmd(swp_entry_t entry)
>>>>>  {
>>>>> -	return (pmd_t){ 0 };
>>>>> +	return (pmd_t){};
>>>>>  }
>>>>
>>>> I have done that for today ... please decide which is best (or find
>>>> something better - maybe every platform really needs to have a __pmd()
>>>> definition) and submit a real fix patch to Andrew.
>>>
>>> OK, that failed for my compiler (gcc 5.2.0) like this:
>>>
>>> In file included from mm/vmscan.c:55:0:
>>> include/linux/swapops.h: In function 'swp_entry_to_pmd':
>>> include/linux/swapops.h:226:16: error: empty scalar initializer
>>>   return (pmd_t){};
>>>                 ^
>>> include/linux/swapops.h:226:16: note: (near initialization for '(anonymous)')
>>
>> This has reappeared today :-( (for the arm multi_v7_defconfig build at
>> least)
>>
>>> So I used the other idea (on top of Andrew's current tree):
>>
>> The fix patch now looks like this:
>>
>> From: Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx>
>> Date: Wed, 2 Aug 2017 15:55:02 +1000
>> Subject: [PATCH] mm-thp-enable-thp-migration-in-generic-path-fix-fix
>>
>> Signed-off-by: Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx>
>> ---
>>  include/linux/swapops.h | 4 +++-
>>  1 file changed, 3 insertions(+), 1 deletion(-)
>>
>> diff --git a/include/linux/swapops.h b/include/linux/swapops.h
>> index 45b092aa6419..61cffa148a79 100644
>> --- a/include/linux/swapops.h
>> +++ b/include/linux/swapops.h
>> @@ -223,7 +223,9 @@ static inline swp_entry_t pmd_to_swp_entry(pmd_t pmd)
>>
>>  static inline pmd_t swp_entry_to_pmd(swp_entry_t entry)
>>  {
>> -	return (pmd_t){};
>> +	pmd_t e;
>> +	memset(&e, 0, sizeof(pmd_t));
>> +	return e;
>>  }
>
> err, yeah.  I didn't want to add that one :(
>
> At the very least we should add a good comment explaining why we had to
> resort to this.

Hi Andrew,

I think __pmd(0) can be used now. I fixed __pmd() in sparc32 at commit
9157259d16a8ee8116a98d32f29b797689327e8d, which is in 4.13 now.
I should have told you this earlier, sorry about that.

Just wonder if any other reason prevents us using __pmd().

Thanks.

â
Best Regards,
Yan Zi

Attachment: signature.asc
Description: OpenPGP digital signature