[Crash-utility] Re:[RFC] Crash patch for DWARF CFI based unwind support

[Rachita Kothiyal]

On Mon, Oct 23, 2006 at 02:34:32PM -0400, Dave Anderson wrote:
> Rachita Kothiyal wrote:
> 
> > This is because what crash is reporting is the stack address at which
> > the return address was pushed on stack, while what the dwarf based bt is
> > reporting is the CFA. In most cases, return address is stored at a location
> > (CFA - 8). That is why the offset of 0x8.
> >
> > The low-budget tracer's backtraces are different from the dwarf-tracer
> > because when the low-budget tracer is unwinding the stack by trying to read
> > kernel text addresses, it actually comes across many addresses which were
> > actually not pushed onto stack because of function calls.
> > Specially for the panic task on kdumps, where after 'crash_kexec' is called,
> > the registers are dumped onto stack(for creating NT_PRSTATUS section), this
> > becomes misleading for the low-budget tracer mechanism. Thats why we see
> > multiple crash_kexec entries in the backtrace. Static inline functions can
> > also aggrevate this problem.
> >
> > In other cases, stale frames on the stack can also mislead the low-budget
> > tracer.
> >
> > AFAICT, user_regs_struct register offsets are not the culprits here.
> >
> > Thanks
> > Rachita
> 
> So, in other words, if we hardwire the user_regs_struct so that
> it uses the NT_PRSTATUS registers all the time, then we get
> the second (preferred/better) budget back trace when unwind
> is off.
> 
> That being the case, I argue for hardwiring them all the time.

Yes, we can(should) do that for all the active tasks.

Rachita