Re: [PATCH 2.4.34-pre1] fix x86_64 etc build failure due to memchr change

[Willy Tarreau <w@1wt.eu>]

On Sun, Aug 20, 2006 at 12:24:04PM +0200, Mikael Pettersson wrote:
> On Sun, 20 Aug 2006 03:31:09 +0200, Willy Tarreau wrote:
> >On Sun, Aug 20, 2006 at 12:19:06AM +0200, Mikael Pettersson wrote:
> >> 2.4.34-pre1 doesn't build on x86_64:
> >> 
> >> kernel/kernel.o(__ksymtab+0x1c10): multiple definition of `__ksymtab_memchr'
> >> arch/x86_64/kernel/kernel.o(__ksymtab+0x3f0): first defined here
> >> kernel/kernel.o(.kstrtab+0x3960): multiple definition of `__kstrtab_memchr'
> >> arch/x86_64/kernel/kernel.o(.kstrtab+0x5fd): first defined here
> >> ld: Warning: size of symbol `__kstrtab_memchr' changed from 7 in arch/x86_64/kernel/kernel.o to 17 in kernel/kernel.o
> >> make: *** [vmlinux] Error 1
> >> 
> >> This is because the 'export memchr() which is used by smbfs and lp driver'
> >> change in 2.4.34-pre1 added an EXPORT_SYMBOL of memchr to kernel/ksyms.c
> >> without also removing the existing one in arch/x86_64/kernel/x8664_ksyms.c.
> >> Alpha, ARM, ppc32, and SH also have EXPORTs of memchr so they probably
> >> also broke.
> >> 
> >> This patch removes the EXPORTs of memchr under arch/, which fixes x86_64
> >> and should fix the other architectures as well.
> >
> >OK Mikael,
> >
> >I have fixed sparc and sparc64 instead, and it's OK now without having to
> >export memchr() in kernel/ksyms.c. So the fix is shorter and more logical.
> >It brings non-sparc architectures to their state before my wrong fix. However,
> >sparc needs to export memchr() for smbfs and lp.
> 
> Works for me. Tested on i386, x86_64, and ppc32. I'll try to test on sparc64
> later today.

Perfect. I have already checked sparc64 last night, so I think the fix was
the right thing to do.

> /Mikael

Thanks Mikael,
Willy