[Crash-utility] user-space enhancements

[Bradshaw, James]

Right. To be able to examine user space, you'd have to build an elf core file by processing the desired task structure in the kdump file, find all the user pages, etc.--essentially what elf_core_dump() does in a running kernel. Then you could use gdb offline or the embedded gdb.

I understand your desire not to burden crash with user space stuff, although the extensions facility seems to provide a mechanism for cleanly excluding such functionality from the standard configuration. Just a thought.

-----Original Message-----
From: crash-utility-bounces at redhat.com [mailto:crash-utility-bounces at redhat.com] On Behalf Of Dave Anderson
Sent: Thursday, December 04, 2008 3:25 PM
To: Discussion list for crash utility usage, maintenance and development
Subject: Re: [Crash-utility] user-space enhancements


----- "James Bradshaw" <jbradsha at enterasys.com> wrote:

> One of the items in the bug list is the following:
>
> DESCRIPTION:
> User space enhancements:
> - show user space stack backtrace, if present in the dump file,
> - ability to link user space namelist (debug object files),
>
> RESOLUTION STATUS: TBD
>
> Is anyone currently working on this?

The items in the TODO list, with the exception of the first
one about the "search" command, are all essentially "wish-list"
items.  They were originally requested to be put there by IBM
several years ago when the http://people.redhat.com/anderson
site was instantiated as the "upstream" source of the crash
utility.

The only item that I'm aware of that somebody is actually looking
into is the one regarding "local variables", where I believe the
guy looking into it is part of the IBM LTC in India.  I don't
expect much to come out of it, though, because for one thing
it presumes that the crash utility's backtrace frame information
is etched in stone -- and with the exception of ia64 which has
unwind information actually built into the kernel -- the backtrace
is essentially a "best-guess" operation.  So trying to pull local
arguments (or function arguments for that matter) from a
dubious source doesn't make a whole lot of sense.

As far as a user-space backtrace, I still think the way to go
is to work on creating a core dump file of the requested task,
and then use gdb externally on that core file, completely outside
of the crash utility.  Trying to overload the crash utility with
a bunch of user-space stuff is something I don't have a lot of
interest in.

Dave





--
Crash-utility mailing list
Crash-utility at redhat.com
https://www.redhat.com/mailman/listinfo/crash-utility