klibc - Oct 2013 - Unsupported system calls

On 10/16/2013 12:42 AM, Anil Singhar wrote:
> Hi HPA,
> 
> Regarding the architecture un-supported system calls, I wrapped them with
> #ifndef __NR_<foo> .. #endif /* __NR_<foo> */ switches. I have
already done
> that. But you also mentioned that in SYSCALLS.def all these calls should be
> present and marked with '?'.. Is that right..? Could you please
give little
> more detail about the change you want.
> 
It's very simple: if the system call has <?> before it, it will simply
be omitted if it isn't present on a specific architecture.  If it
*doesn't* have <?> in front of it, then it is an error for it to not
exist.

	-hpa

Hi HPA,

Thanks for your reply. One confusion I have is:

e.g. 'access' is not supported in aarch64, so should the declaration be:

<!aarch64> int access(const char *, int);  <-- This is how it is
currently
indicating that only aarch64 doesn't support it.

OR

<?> int access(const char *, int); <-- if I do this instead, it will be
optional for every architecture, right..?

Thanks and Regards,
Anil



On 16 October 2013 13:14, H. Peter Anvin <hpa at zytor.com> wrote:
> On 10/16/2013 12:42 AM, Anil Singhar wrote:
> > Hi HPA,
> >
> > Regarding the architecture un-supported system calls, I wrapped them
with
> > #ifndef __NR_<foo> .. #endif /* __NR_<foo> */ switches. I
have already
> done
> > that. But you also mentioned that in SYSCALLS.def all these calls
should
> be
> > present and marked with '?'.. Is that right..? Could you
please give
> little
> > more detail about the change you want.
> >
>
> It's very simple: if the system call has <?> before it, it will
simply
> be omitted if it isn't present on a specific architecture.  If it
> *doesn't* have <?> in front of it, then it is an error for it to
not exist.
>
>         -hpa
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://www.zytor.com/pipermail/klibc/attachments/20131016/2ca8fd74/attachment.html>

On 10/16/2013 12:52 AM, Anil Singhar wrote:
> Hi HPA,
> 
> Thanks for your reply. One confusion I have is:
> 
> e.g. 'access' is not supported in aarch64, so should the
declaration be:
> 
> <!aarch64> int access(const char *, int);  <-- This is how it is
currently
> indicating that only aarch64 doesn't support it.
> 
> OR
> 
> <?> int access(const char *, int); <-- if I do this instead, it
will be
> optional for every architecture, right..?
> 
Yes, and that is the right way to do it if there is a generic,
non-architecture specific fallback in terms of another system call.

	-hpa