Re: Ping: cacheflush.2

[Martin Sebor <msebor@gmail.com>]

On 12/18/20 3:42 AM, Alejandro Colomar (man-pages) wrote:
> Hi Martin,
> 
> I sent you an email, but I received a "delivery failure".
> If you're reading this from a list, could you answer, please?
> 
> Thanks,
> 
> Alex
> 
> On 12/14/20 11:34 PM, Alejandro Colomar (man-pages) wrote:
>> Hello Martin,
>>
>> Thanks for the correction!
>> Then the prototypes that changes from 'char *' to 'void *' in r269082
>> were not exposed to the user, right?
>> I guess then those are just internal implementation where GCC did use
>> 'char *'.

__builtin___clear_cache was added to GCC in r126535 (the __builtin_
prefix is added by the macro):

+DEF_EXT_LIB_BUILTIN    (BUILT_IN_CLEAR_CACHE, "__clear_cache", 
BT_FN_VOID_PTR_PTR, ATTR_NOTHROW_LIST)

The BT_FN_VOID_PTR_PTR macro describes its signature as returning
void and taking two void pointer arguments.  AFAIK, this has never
changed.  Contrary to that, the manual entry for the built-in added
in the same revision documented it as taking two char*.  That was
corrected to void* in r269082 to match.

There's a GCC internal declaration of __clear_cache (apparently
provided in libgcc for VxWorks).  It was added in r264479 and
it also used char*.  This was also changed to void* in r269082
to match the built-in.  Looks like this __clear_cache has just
been removed from libgcc in GCC 11:
https://gcc.gnu.org/pipermail/gcc-cvs/2020-December/338478.html

>>
>> Where is the actual prototype exposed to the user declared?

Built-in functions are declared implicitly by GCC.  They have no
explicit declarations like user-defined functions.  The implicit
internal "declarations" are specified in the GCC internal file
gcc/builtins.def, where they are hidden behind layers of macros.
For example, on the GCC 10 branch, the declaration for
__builtin___clear_cache is here:

https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=gcc/builtins.def;h=fa8b0641ab13b36f983c591a7020f6b432e5fb3d;hb=refs/heads/releases/gcc-10#l837

Martin

>>
>> Thanks,
>>
>> Alex
>>
>> P.S.: Michael, wait for a patch revision (v6).
>>
>> On 12/14/20 10:13 PM, Martin Sebor wrote:
>>> On 12/11/20 11:14 AM, Alejandro Colomar (man-pages) via Gcc wrote:
>>>> It looks like GCC recently moved from 'char *' to 'void *'.
>>>> This SO question[1] (4 years ago) quotes the GCC docs
>>>> and they had 'char *'.
>>>
>>> __builtin___clear_cache in GCC has always been declared to take
>>> void*.  The signature in the manual was recently corrected to match
>>> the implementation, i.e., from char* to void*, in r269082.
>>>
>>> Martin
>>>
>>>> Maybe Clang hasn't noticed the change.
>>>> I'll report a bug.
>>>>
>>>> [1]: https://stackoverflow.com/q/35741814/6872717
>>>>
>>>> On 12/9/20 8:15 PM, Alejandro Colomar (man-pages) wrote:
>>>>> Hi Heinrich,
>>>>>
>>>>> It looks like a bug (or at least an undocumented divergence from GCC) in
>>>>> Clang/LLVM.  Or I couldn't find the documentation for it.
>>>>>
>>>>> Clang uses 'char *':
>>>>> https://github.com/llvm/llvm-project/blob/7faf62a80bfc3a9dfe34133681fcc31f8e8d658b/clang/include/clang/Basic/Builtins.def#L583
>>>>>
>>>>>
>>>>> GCC uses 'void *':
>>>>> https://gcc.gnu.org/onlinedocs/gcc/Other-Builtins.html
>>>>>
>>>>> I CCd Clang and GCC lists; maybe they know about that divergence.
>>>>>
>>>>> Cheers,
>>>>>
>>>>> Alex
>>>>>
>>>>> On 12/9/20 7:48 PM, Heinrich Schuchardt wrote:
>>>>>> On 12/9/20 7:34 PM, Alejandro Colomar (man-pages) wrote:
>>>>>>> Hi Heinrich & Michael,
>>>>>>>
>>>>>>> What about the following?:
>>>>>>>
>>>>>>> [
>>>>>>> NOTES
>>>>>>>           GCC provides a similar function, which may be useful on
>>>>>>> archi‐
>>>>>>>           tectures that lack this system call:
>>>>>>>
>>>>>>>               void __builtin___clear_cache(void *begin, void *end);
>>>>>>> ]
>>>>>>
>>>>>> I just checked building with Clang/LLVM. There the arguments are of
>>>>>> type
>>>>>> (char *). See the following error output:
>>>>>>
>>>>>> +arch/sandbox/cpu/cache.c:19:26: error: passing 'uint8_t *' (aka
>>>>>> 'unsigned char *') to parameter of type 'char *' converts between
>>>>>> pointers to integer types with different sign [-Werror,-Wpointer-sign]
>>>>>> +        __builtin___clear_cache(state->ram_buf,
>>>>>> +                                ^~~~~~~~~~~~~~
>>>>>> +arch/sandbox/cpu/cache.c:20:12: error: passing 'uint8_t *' (aka
>>>>>> 'unsigned char *') to parameter of type 'char *' converts between
>>>>>> pointers to integer types with different sign [-Werror,-Wpointer-sign]
>>>>>> +                                state->ram_buf + state->ram_size);
>>>>>> +                                ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>>>>>
>>>>>> Best regards
>>>>>>
>>>>>> Heinrich
>>>>>>
>>>>>>>
>>>>>>> Cheers,
>>>>>>>
>>>>>>> Alex
>>>>>>>
>>>>>>> On 12/9/20 7:04 PM, Heinrich Schuchardt wrote:
>>>>>>>> Hello Michael,
>>>>>>>>
>>>>>>>> function cacheflush() does not exist on many architectures.
>>>>>>>>
>>>>>>>> It would have saved me a lot of time if the man-page had referenced
>>>>>>>> GCC's
>>>>>>>>
>>>>>>>> void __builtin___clear_cache(void *begin, void *end)
>>>>>>>>
>>>>>>>> Maybe you can add it to NOTES.
>>>>>>>>
>>>>>>>> Best regards
>>>>>>>>
>>>>>>>> heirnich
>>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
>