Re: [patch 00/20] x86, ptrace, bts, hw-branch-tracer: fixes and cleanups

[Ingo Molnar <mingo@elte.hu>]

* Markus Metzger <markus.t.metzger@googlemail.com> wrote:

> On Fri, 2009-04-03 at 17:36 +0200, Ingo Molnar wrote: 
> > * markus.t.metzger@intel.com <markus.t.metzger@intel.com> wrote:
> > 
> > > Patches 1-5 fix races with context switching code when the branch 
> > > traced task is currently running.
> > > 
> > > In the worst case, this might cause context switch code to access 
> > > freed memory or the tracing hardware to continue tracing into a 
> > > freed buffer. Both might crash the kernel.
> > > 
> > > The first 4 patches apply to .29 using the below preparation 
> > > patch.
> > > 
> > > The remaining patches fix bugs in the context of per-cpu tracing 
> > > (i.e. hw-branch-tracer) and pebs, add more selftest code, and do 
> > > some cleanups.
> > 
> > Thanks - this looks pretty acceptable. Latest -git changed a lot of 
> > code in the same area (-mm bits went upstream), creating a lot of 
> > conflicts.
> > 
> > To not prolongue this any longer (20 patches are difficult enough 
> > already to handle) i picked up your patches and resolved the 
> > conflicts in situ - mind having a look at the resulting 
> > tip:tracing/hw-branch-tracing branch - does the end result look sane 
> > to you?
> 
> Thanks,
> 
> The branch tracing bits in tip/tracing/hw-branch-tracing look good.
> The kernel boots (after changes, see below) and my tests pass.
> 
> A defconfig build fails with a compile error, though:
> In file included from net/core/skbuff.c:68:
> include/trace/skb.h:7: error: expected ‘)’ before ‘(’ token
> .....
> 
> Changing TP~ to TP_~ (as in tip/master) seems to do the trick.

Ah. Darn - i should have merged tracing/core into it. That will ruin 
bisectability... I'll redo it.

	Ingo