RE: [patch v4 13/13] btrace, x86: restrict to Atom

["Metzger, Markus T" <markus.t.metzger@intel.com>]

> -----Original Message-----
> From: Mark Kettenis [mailto:mark.kettenis@xs4all.nl]
> Sent: Tuesday, November 27, 2012 3:42 PM
> To: Metzger, Markus T
> Cc: gdb-patches@sourceware.org; markus.t.metzger@gmail.com; jan.kratochvil@redhat.com; palves@redhat.com;
> tromey@redhat.com
> Subject: Re: [patch v4 13/13] btrace, x86: restrict to Atom
> 
> > From: "Metzger, Markus T" <markus.t.metzger@intel.com>
> > Date: Tue, 27 Nov 2012 11:49:09 +0000
> >
> > > > From: <markus.t.metzger@intel.com>
> > > > Date: Tue, 27 Nov 2012 11:49:11 +0100
> > > >
> > > > From: Markus Metzger <markus.t.metzger@intel.com>
> > > >
> > > > Restrict branch tracing support to Atom processors.
> > > >
> > > > 2012-11-27 Markus Metzger <markus.t.metzger@intel.com>
> > > >
> > > > 	* amd64-linux-nat.c (amd64_linux_supports_btrace): New.
> > > > 	(_initialize_amd64_linux_nat): Change supports_btrace method.
> > > > 	* i386-linux-nat.c (i386_linux_supports_btrace): New.
> > > > 	(_initialize_i386_linux_nat): Change supports_btrace method.
> > > >
> > > > gdbserver/
> > > > 	* linux-low.h (linux_target_ops): Add supports_btrace method.
> > > > 	* linux-low.c (linux_low_supports_btrace): New.
> > > > 	(linux_target_ops): Change supports_btrace method.
> > > > 	* linux-x86-low.c (x86_supports_btrace): New.
> > > > 	(the_low_target): Add supports_btrace method.
> > >
> > > This is still unacceptable.  This needs to be fixed in the Linux kernel.
> >
> > We will still need this patch in gdb so we handle kernels that do
> > not contain the fix correctly.
> 
> Frankly I don't see why.  I just consider this to be a bug in a
> recently introduced new kernel feature.  People should just upgrade to
> a kernel that has this bug fixed.  We can't add workarounds for every
> Linux kernel bug.  And in this case it is especially bad since you're
> restricting GDB to supporting this feature on specific CPU models even
> on systems that support the feature on other CPU types.  I feel
> (rather strongly) that this is bad.

For most people, I think it's easier to update gdb than it is to update the kernel. For some, it may not even be possible to update the kernel.

GDB would only be more restrictive if an old version of GDB that did not know about a new Atom processor were used on a new kernel that already knows this new Atom processor (and if it would run on this new processor). While this is scenario is certainly possible, I think it rather unlikely.

I do agree that this needs to be fixed in the kernel. But I also think that it needs to be addressed in GDB.

Would you be OK if I sent out a respective kernel patch? It would still take a long time before a kernel containing that patch would be available. I don't think that we should wait with the review of this patch series until then.

Regards,
Markus.
Intel GmbH
Dornacher Strasse 1
85622 Feldkirchen/Muenchen, Deutschland
Sitz der Gesellschaft: Feldkirchen bei Muenchen
Geschaeftsfuehrer: Peter Gleissner, Christian Lamprechter, Hannes Schwaderer, Douglas Lusk
Registergericht: Muenchen HRB 47456
Ust.-IdNr./VAT Registration No.: DE129385895
Citibank Frankfurt a.M. (BLZ 502 109 00) 600119052