Re: [PATCH] generic show_stack facility

[Andrew Morton]

Keith Owens wrote:
> 
> On Fri, 29 Mar 2002 10:41:11 -0800,
> Andrew Morton <[e-mail]> wrote:
> >Christoph Hellwig wrote:
> >>
> >> On Fri, Mar 29, 2002 at 09:36:26AM -0800, Andrew Morton wrote:
> >> > Here's the diff.  Comments?
> >>
> >> I don't see who having to independand declaration in the same kernel
> >> image are supposed to work..
> >
> >It goes in lib/lib.a.  The linker will only pick up
> >the default version if the architecture doesn't
> >have its own dump_stack().
> >
> >bust_spinlocks() has worked that way for quite some time.
> 
> I have a problem with putting routines in lib.a and relying on the
> linker to pull them out by default.  It does not work for routines
> called from modules, modules do not include lib.a.  Remember the recent
> problems with crc32.o?
> 
> bust_spinlocks() is not an issue because it is only called from built
> in code.  show_stack() has been used as a debugging facility and it
> could be called from a module.

Yes, that's a good point.  We're safe as long as core kernel
always contains a call to dump_stack(). Which is the case,
but that's a bit subtle for general usage.