[PATCH 3/4] Consolidate pread/pread64 implementations
Chris Metcalf
cmetcalf@mellanox.com
Wed Feb 24 23:28:00 GMT 2016
On 2/24/2016 5:25 PM, Mike Frysinger wrote:
> On 24 Feb 2016 19:12, Adhemerval Zanella wrote:
>> On 24-02-2016 18:41, Mike Frysinger wrote:
>>> On 23 Feb 2016 15:42, Adhemerval Zanella wrote:
>>>> On 23-02-2016 14:57, Joseph Myers wrote:
>>>>> Also, you're including sysdeps/unix/sysv/linux/generic/sysdep.h for
>>>>> architectures that don't use the generic ABI. If that's safe, that
>>>>> suggests it should just be merged into sysdeps/unix/sysv/linux/sysdep.h
>>>>> (in a separate patch with a careful argument for why the merge is safe)
>>>>> rather than keeping the headers separate.
>>>> I decided to add it on generic sysdep.h because it is where port will get
>>>> the definition of __ALIGNMENT_ARG. However I can split the second path
>>>> in two, one to include generic sysdep.h and another to define SYSCALL_LL{64}.
>>> i don't think splitting the path addresses the contention. Joseph's
>>> point (which i agree with) is that linux/generic/ is for the "generic
>>> ABI" which the kernel is trying to push for new arches. if you have
>>> stuff which works across all linux ABI's, then instead of being in
>>> linux/generic/xxx, it should be in linux/xxx.
>>>
>>> if that means some things get hoisted out of linux/generic/xxx and
>>> into linux/xxx, then that's OK.
>> I think then best approach would be move __ALIGNMENT_{ARG,COUNT} from generic
>> to linux/xxx and add the new SYSCALL_LL on linux/xxx as well. What you think?
> sgtm. seems like many of the users in linux/generic/wordsize-32/ could
> also be merged back if you're feeling proactive :).
> -mike
Agreed, sounds like the right direction.
--
Chris Metcalf, Mellanox Technologies
http://www.mellanox.com
More information about the Libc-alpha
mailing list