Re: [patch 11/11] ftrace plugin for kernel symbol tracing using HW Breakpoint interfaces

[Ingo Molnar <mingo@elte.hu>]

* K.Prasad <prasad@linux.vnet.ibm.com> wrote:

> On Thu, Mar 05, 2009 at 10:16:11AM +0100, Ingo Molnar wrote:
> > 
> > * Frederic Weisbecker <fweisbec@gmail.com> wrote:
> > 
> > > On Thu, Mar 05, 2009 at 10:13:33AM +0530, prasad@linux.vnet.ibm.com wrote:
> > > > This patch adds an ftrace plugin to detect and profile memory access over
> > > > kernel variables. It uses HW Breakpoint interfaces to 'watch memory
> > > > addresses.
> > > > 
> > > > Signed-off-by: K.Prasad <prasad@linux.vnet.ibm.com> 
> > > > ---
> > > 
> > > 
> > > Hi,
> > > 
> > > Nice feature. And moreover the standardized hardware 
> > > breakpoints could be helpful for tracing.
> > 
> > yeah. The feature is much more alive now.
> > 
> > > Just some comments below.
> > 
> > One other thing:
> > 
> > +#ifdef CONFIG_FTRACE_SELFTEST
> > +int trace_selftest_startup_ksym(struct tracer *trace, struct trace_array *tr)
> > +{
> > +       /* TODO: Will be implemented later */
> > +       return 0;
> > +}
> > +#endif /* CONFIG_FTRACE_SELFTEST */
> > 
> > This needs to be implemented before i can pick the code up into 
> > tip:tracing, as otherwise we will not notice it fast enough if 
> > some of this stuff breaks.
> > 
> > Basically the ftrace plugin will be the main usage vector of 
> > this facility, so the self-test is a must-have.
> > 
> > Looks very nice otherwise.
> > 
> > 	Ingo
> 
> Thanks for the comments.
> 
> Test-cases for the hardware breakpoint interfaces can be the following:
> 
> - Basic sanity test to check if the API is intact
> - Perform various types of memory accesses, like read, write (I/O and 
>   others when implemented) on a dummy kernel variable and verify the 
>   trigger of the exception handler.
> 
> While the above can be a part of trace_selftest_startup_ksym(),
> rigorous testing would involve:
> 
> i) stressing the HW breakpoint infrastructure to confirm sane behaviour
> when interoperated with other users of a)breakpoint register b)the
> do_debug() exception. This will involve simultaneous use of kprobes,
> hardware breakpoint interface and requests from user-space (say through
> GDB).
> ii) Verifying successful HB_NUM number of register_ requests.
> iii) Verifying right priority resolution, and handling user-space
> requests.
> 
> These, in my opinion, would better fit in a full-featured 
> test-suite such as LTP, as opposed to startup testing in 
> ftrace.

sure. It's just a quick self-test to make sure basic 
functionality is ok.

> I will implement trace_selftest_startup_ksym() to contain the 
> first two test-cases in the next iteration of this code.

Thanks.

	Ingo