This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: Adding cache shape info to powerpc
- From: Steven Munroe <munroesj at linux dot vnet dot ibm dot com>
- To: Benjamin Herrenschmidt <benh at kernel dot crashing dot org>
- Cc: Tulio Magno Quites Machado Filho <tuliom at linux dot vnet dot ibm dot com>, Florian Weimer <fweimer at redhat dot com>, libc-alpha at sourceware dot org, Paul Mackerras <pmac at au1 dot ibm dot com>
- Date: Thu, 01 Dec 2016 16:04:13 -0600
- Subject: Re: Adding cache shape info to powerpc
- Authentication-results: sourceware.org; auth=none
- References: <1471408952.19495.29.camel@kernel.crashing.org> <d350e721-0410-3c3a-86c2-a9f3d90b8f0d@redhat.com> <1471479455.19495.36.camel@kernel.crashing.org> <1480605033.7377.9.camel@oc7878010663> <1480623003.3459.21.camel@kernel.crashing.org>
- Reply-to: munroesj at linux dot vnet dot ibm dot com
On Fri, 2016-12-02 at 07:10 +1100, Benjamin Herrenschmidt wrote:
> On Thu, 2016-12-01 at 09:10 -0600, Steven Munroe wrote:
> > > The old format was only used by other architectures, not powerpc, do I
> > > still need to pick different numbers in that case ?
> > >
> > > As for IFUNC I don't think so but I'll let Steve reply as I believe he
> > > is the original requestor for the functionality.
> > >
> >
> > Ben were are we on this?
> >
> > This is needed to fix a customer reported bug for lack of
> > sysconf(_SC_*CACHE_SIZE) support on powerpc.
> >
> > Adding this information to the cputable and then auxv fits into the
> > current sysconf infrastructure best.
> >
> > When we know the AT_ info format and upstream Kernel version we can
> > update kernel-features.h and enable a sysconf.c fragment to return these
> > values for powerpc
>
> I was waiting for a review by you guys to tell me what you want in the
> AT_* entries. People weren't too happy with my first sketch that used
> the same entries used by a couple of other archs but with a slightly
> different format.
>
> Once we agree on an encoding/format I can resend the patches.
The AT_*_CACHESHAPE where encoding more information then strictly
required for _SC_*CACHE_SIZE.
So the alternative is to define 4 new/different AT_*_CACHESIZE entries
where each is just the respective cache size in KB.
This should handle current and projected values in 32-bits for the
foreseeable future.
Thanks