Re: [perfmon2] [PATCH] 1/1 - fix Power6 sampling timer implementation for counters 5 & 6, v2

[Corey J A. <cja...@us...>]

Ok, I'll give this another go tomorrow.  Thanks.

- Corey

stephane eranian <er...@go...> wrote on 01/19/2009 02:54:49 PM:

> Corey,
> 
> Sorry for late reply.
> 
> The reason you're getting the warning is because perfmon_init.c is
> missing #include <linux/modules.h>, thus the macro 
EXPORT_PER_CPU_SYMBOL()
> is unknown.
> 
> On Sat, Jan 17, 2009 at 12:41 AM, Corey J Ashford <cja...@us...> 
wrote:
> > Hi Stephane,
> >
> > I tried the simplification you suggested below, but there is a 
problem.
> > The pmu_ctx variable is not exported from the kernel, so the
> > perfmon_power6.c module does not have visibility to it.  As an 
experiment,
> > I tried adding
> >
> > EXPORT_PER_CPU_SYMBOL(pmu_ctx);
> >
> > in perfmon_init.c, but I get these warning messages during compile, 
which
> > I don't understand:
> >
> > make -j8 vmlinux modules
> >  CHK     include/linux/version.h
> >  CHK     include/linux/utsrelease.h
> >  SYMLINK include/asm -> include/asm-powerpc
> >  CALL    scripts/checksyscalls.sh
> >  CHK     include/linux/compile.h
> >  CALL    arch/powerpc/kernel/systbl_chk.sh
> >  CALL    arch/powerpc/kernel/prom_init_check.sh
> >  CC      perfmon/perfmon_init.o
> > perfmon/perfmon_init.c:52: warning: data definition has no type or 
storage
> > class
> > perfmon/perfmon_init.c:52: warning: type defaults to 'int' in 
declaration
> > of 'EXPORT_SYMBOL'
> > perfmon/perfmon_init.c:52: warning: parameter names (without types) in
> > function declaration
> >  LD      perfmon/built-in.o
> >  LD      vmlinux.o
> >  MODPOST vmlinux.o
> >
> > Line 52 is the "EXPORT_PER_CPU_SYMBOL(pmu_ctx);" line.
> >
> > and when it goes to build the modules, I still get this error message:
> >
> > ERROR: "per_cpu__pmu_ctx" [arch/powerpc/perfmon/perfmon_power6.ko]
> > undefined!
> >
> > Any ideas?
> >
> > - Corey
> >
> > stephane eranian <er...@go...> wrote on 01/15/2009 02:43:01 
PM:
> >
> >> Corey,
> >>
> >> I looked at the patch. It is better than the first version with the
> >> per-cpu info.
> >>
> >> I think it could be simplified a bit more by dropping:
> >>         struct pfm_arch_context *ctx_arch;
> >>
> >> >From the struct pmc5_6_per_cpu. You can get to the context,
> >> even in the hrtimer handler, via
> >>
> >>        ctx = __get_cpu_var(pmu_ctx);
> >>        ctx_arch = pfm_ctx_arch(ctx);
> >
> >