Re: [PATCH 02/16] perf: Unified API to record selective sets of arch registers

[Jiri Olsa <jolsa@redhat.com>]

On Mon, Apr 23, 2012 at 12:10:57PM +0200, Stephane Eranian wrote:
> On Tue, Apr 17, 2012 at 1:17 PM, Jiri Olsa <jolsa@redhat.com> wrote:
> > This brings a new API to help the selective dump of registers on
> > event sampling, and its implementation in x86.
> >
> > - The informations about the desired registers will be passed
> >  to a single u64 mask. It's up to the architecture to map the
> >  registers into the mask bits.
> >
> > - The architecture must provide a non-zero and unique id to
> >  identify the origin of a register set because interpreting a
> >  register dump requires to know from which architecture it comes.
> >  The achitecture is considered different between the 32 and 64 bits
> >  version. x86-32 has the id 1, x86-64 has the id 2.
> >
> > Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
> > Signed-off-by: Jiri Olsa <jolsa@redhat.com>
> > ---
> >  arch/x86/include/asm/perf_regs.h    |   16 ++++++
> >  arch/x86/include/asm/perf_regs_32.h |   84 +++++++++++++++++++++++++++++
> >  arch/x86/include/asm/perf_regs_64.h |  101 +++++++++++++++++++++++++++++++++++
> >  include/asm-generic/perf_regs.h     |   23 ++++++++
> >  4 files changed, 224 insertions(+), 0 deletions(-)
> >  create mode 100644 arch/x86/include/asm/perf_regs.h
> >  create mode 100644 arch/x86/include/asm/perf_regs_32.h
> >  create mode 100644 arch/x86/include/asm/perf_regs_64.h
> >  create mode 100644 include/asm-generic/perf_regs.h
> >
> > diff --git a/arch/x86/include/asm/perf_regs.h b/arch/x86/include/asm/perf_regs.h
> > new file mode 100644
> > index 0000000..80b7fbe
> > --- /dev/null
> > +++ b/arch/x86/include/asm/perf_regs.h
> > @@ -0,0 +1,16 @@
> > +#ifndef _ASM_X86_PERF_REGS_H
> > +#define _ASM_X86_PERF_REGS_H
> > +
> > +enum {
> > +       PERF_REGS_VERSION_NONE   = 0UL,
> > +       PERF_REGS_VERSION_X86_32 = 1UL,
> > +       PERF_REGS_VERSION_X86_64 = 2UL,
> > +};
> > +
> I don't really like the term VERSION here. It's not a versioning
> problem you're trying to solve. It's an ABI problem, unless I am
> mistaken. You should rename to PERF_REGS_ABI_X86_32 and
> PERF_REGS_ABI_X86_64.
> 
> I assume the NONE is here to cover the case where you don't
> have a user machine state, i.e., hit a kernel thread. Is that right?

right

> 
> 
> > +#ifdef CONFIG_X86_32
> > +#include "perf_regs_32.h"
> > +#else
> > +#include "perf_regs_64.h"
> > +#endif
> > +
> How are you going to deal with 32-bit binaries sampled on a 64-bit system?

I dont have the solution right now... but seems like compat tasks need more
thinking even before go ahead with this patchset.. since it's going affect
the perf_event_attr and could bite us in future.

I'll see what I can do about that and send out new patchset

jirka