[edk2-devel] [PATCH v1 1/1] CryptoPkg/IntrinsicLib: Fix CLANG38 IA32 build problem
Andrew Fish via Groups.Io
afish=apple.com at groups.io
Thu Jun 6 03:38:31 UTC 2019
> On Jun 4, 2019, at 11:33 PM, Xiaoyu Lu <xiaoyux.lu at intel.com> wrote:
>
> Hi Liming,
>
>> -----Original Message-----
>> From: Gao, Liming
>> Sent: Wednesday, June 5, 2019 1:57 PM
>> To: devel at edk2.groups.io <mailto:devel at edk2.groups.io>; Lu, XiaoyuX <xiaoyux.lu at intel.com <mailto:xiaoyux.lu at intel.com>>
>> Cc: Bi, Dandan <dandan.bi at intel.com <mailto:dandan.bi at intel.com>>; Wang, Jian J <jian.j.wang at intel.com <mailto:jian.j.wang at intel.com>>
>> Subject: RE: [edk2-devel] [PATCH v1 1/1] CryptoPkg/IntrinsicLib: Fix
>> CLANG38 IA32 build problem
>>
>> Xiaoyu:
>>
>>> -----Original Message-----
>>> From: devel at edk2.groups.io [mailto:devel at edk2.groups.io] On Behalf Of
>>> Xiaoyu Lu
>>> Sent: Wednesday, June 05, 2019 1:25 PM
>>> To: devel at edk2.groups.io
>>> Cc: Lu, XiaoyuX <xiaoyux.lu at intel.com>; Bi, Dandan
>> <dandan.bi at intel.com>;
>>> Wang, Jian J <jian.j.wang at intel.com>
>>> Subject: [edk2-devel] [PATCH v1 1/1] CryptoPkg/IntrinsicLib: Fix CLANG38
>>> IA32 build problem
>>>
>>> When use clang-3.8 to build the NetworkPkg, compiler optimization
>>> may use memcpy for memory copy. For example:
>>>
>>> CryptoPkg/Library/OpensslLib/openssl/ssl/ssl_rsa.c:918: undefined
>>> reference to `memcpy'`
>>>
>>> Compiler optimization is sophisticated, but we can work around it
>>> use __attribute__((__used__)) to informs the compiler that symbol
>>> should be retained in the object file, even if it may be
>>> unreferenced.
>>>
>>> Cc: Jian J Wang <jian.j.wang at intel.com>
>>> Cc: Dandan Bi <dandan.bi at intel.com>
>>> Signed-off-by: Xiaoyu Lu <xiaoyux.lu at intel.com>
>>> ---
>>> CryptoPkg/Library/IntrinsicLib/CopyMem.c | 13 +++++++++++++
>>> 1 file changed, 13 insertions(+)
>>>
>>> diff --git a/CryptoPkg/Library/IntrinsicLib/CopyMem.c
>>> b/CryptoPkg/Library/IntrinsicLib/CopyMem.c
>>> index e29b4918d200..7faf5a34d8c1 100644
>>> --- a/CryptoPkg/Library/IntrinsicLib/CopyMem.c
>>> +++ b/CryptoPkg/Library/IntrinsicLib/CopyMem.c
>>> @@ -10,8 +10,21 @@ SPDX-License-Identifier: BSD-2-Clause-Patent
>>> #include <Base.h>
>>> #include <Library/BaseMemoryLib.h>
>>>
>>> +#if defined(__clang__) && !defined(__APPLE__)
>>
>> So, this change is only for CLANG tool chain.
>>
>>> +
>>> +/* Copies bytes between buffers */
>>> +static __attribute__((__used__))
>>
>> What purpose for static?
>>
>
> Because I want __memcpy only use in this file scope.
>
>>> +void * __memcpy (void *dest, const void *src, unsigned int count)
>>> +{
>>> + return CopyMem (dest, src, (UINTN)count);
>>> +}
>>> +__attribute__((__alias__("__memcpy")))
>>> +void * memcpy (void *dest, const void *src, unsigned int count);
>>
>> __memcpy is IA32 Intrinsic API, memcpy is X64 Intrinsic API, right?
>>
>
> __memcpy isn't IA32 Intrinsic API, only memcpy is intrinsic API for both IA32 and X64.
>
> The reason I alias memcpy and use __attribute__((__used__)) is let compiler retain symbol in object file,
> So it can link correct.
>
> Is this correct?
>
I think this is a bug in clang that requires the __used__, we hit something like this with Xcode too. If the compiler emits the intrinsic it should tell the linker and some how that was getting missed. Thus the __used__ is a work around.
Thanks,
Andrew Fish
>> Thanks
>> Liming
>
>>> +
>>> +#else
>>> /* Copies bytes between buffers */
>>> void * memcpy (void *dest, const void *src, unsigned int count)
>>> {
>>> return CopyMem (dest, src, (UINTN)count);
>>> }
>>> +#endif
>>> --
>>> 2.7.4
>>>
>>>
>>>
>
>
>
-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#41974): https://edk2.groups.io/g/devel/message/41974
Mute This Topic: https://groups.io/mt/31934411/1813853
Group Owner: devel+owner at edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [edk2-devel-archive at redhat.com]
-=-=-=-=-=-=-=-=-=-=-=-
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/edk2-devel-archive/attachments/20190605/a87e2cee/attachment.htm>
More information about the edk2-devel-archive
mailing list