Re: [PATCH 2/3] tracing/kprobes: Make Kconfig dependencies generic

[Frederic Weisbecker <fweisbec@gmail.com>]

On Thu, Feb 18, 2010 at 10:02:52AM -0500, Mike Frysinger wrote:
> On Thu, Feb 18, 2010 at 09:50, Frederic Weisbecker wrote:
> > On Thu, Feb 18, 2010 at 09:01:12AM -0500, Mike Frysinger wrote:
> >> > --- a/arch/Kconfig
> >> > +++ b/arch/Kconfig
> >> > @@ -123,6 +123,10 @@ config USE_GENERIC_SMP_HELPERS
> >> >
> >> >  config HAVE_REGS_AND_STACK_ACCESS_API
> >> >        bool
> >> > +       help
> >> > +         This symbol should be selected by an architecure if it supports
> >> > +         the API needed to access registers and stack entries from pt_regs.
> >> > +         For example the kprobes-based event tracer needs this API.
> >>
> >> a bit vague ... arent there headers/functions people could look at ?
> >> perhaps you're talking about the regset functions (which is an API to
> >> access registers in pt_regs) ?  or you're talking about asm/syscall.h
> >> (which is an API to access registers in pt_regs) ?
> >>
> >> i'm not asking to be a pain, i'm asking because i really havent a
> >> clue.  if i wanted to add support for this stuff to the Blackfin arch,
> >> i wouldnt know where to start.  even after reading this help i'd fall
> >> back to grepping arch/x86/ and trying to divine a starting point from
> >> there.
> >
> > If an arch support kprobes, it just needs to select
> > HAVE_REGS_AND_STACK_ACCESS_API to figure out quickly what is missing,
> > as gcc will barf every missing clues you need.
> 
> so should this new Kconfig option have an appropriate depends on
> KPROBES or something ?



No, kprobes events need KPROBES and the above one, but KPROBES
alone can perfectely work without HAVE_REGS_AND_STACK_ACCESS_API.


 
> > For now it is stored is asm/ptrace.h, but that might be split in
> > the future, especially as ptrace has initially nothing related to
> > that. A documentation that deals with filenames or API enumerations
> > tend to be incidentally async with API evolutions.
> 
> i dont expect there to be per-function documentation here ... such
> things below in the header files themselves (linux/regset.h is an
> example of how to approach this).  but having a tip of reading a file
> or two (like HAVE_ARCH_TRACEHOOK) doesnt bit rot nearly as often.  if
> the common API expected of headers hasnt yet been split, then i guess
> not much left to be done here.



Yeah. Once it's split up from ptrace, we can think about a more dedicated
header file and then it can be reasonable to put the header filename in the
documentation. But not for now IMHO.