Re: [PATCH 2/3] Add tracing_off_event() calls to BUG() and WARN() paths

[Frederic Weisbecker <fweisbec@gmail.com>]

On Thu, Apr 15, 2010 at 02:49:05PM -0700, Chase Douglas wrote:
> On Thu, Apr 15, 2010 at 1:57 PM, Frederic Weisbecker <fweisbec@gmail.com> wrote:
> > On Wed, Apr 14, 2010 at 12:20:15PM -0400, Chase Douglas wrote:
> >> This change adds tracing_off_event() calls, which stop debug tracing,
> >> when a BUG() or WARN() function is called. The stoppage depends on
> >> commandline paramenter tracing_off={bug,warn,none}. The default is
> >> "bug", so only the BUG() paths will stop tracing.
> >>
> >> Signed-off-by: Chase Douglas <chase.douglas@canonical.com>
> >
> >
> > ACK
> >
> > But I can't find where we call oops_enter() from panic()
> > I see it called in x86 from oops_begin() that is called
> > in various places like die(), but couldn't find how it
> > ends up beeing called from panic().
> 
> I did not test the panic() path, but when that happens the system goes
> down anyways. You won't be able to get the ftrace unless you enable
> the ftrace on oops option, in which case the end of the trace will
> likely be close enough to what you are looking for anyways.


Yeah but there is also a sysrq that can trigger an ftrace dump.
So if you forgot ftrace dump on oops and you want to use the sysrq
then you don't want the tracing to continue between the panic and
the time for your fingers to reach the keyboard (that notwithstanding
the time to remember the right combination...which I always forget).