I followed your and Jeff's lead about new gdbs being
out-of-sync with UML and downgraded gdb to 5.3.90
(the one that came with FC1). Everything works great
now. Thanks.

But again, even 2.4 kernels *has* frame
pointers (just verified - even if it's not enabled inside .config, but just
from arch/um/Makefile).

Hmmm... strange. When I'm debugging 2.4.27+1um on rh9
I can never see correct backtrace in ddd. But with the gdb
fix above (i.e. fc3 host/5.3.90 gdb/2.6.12+bs9 guest)
backtraces work correctly.

Debugging in SKAS (SKAS0 or SKAS3) mode works much better...

How is it better? So far I only see the extra headache of patching
the host.

Btw, does 2.6.13pre that Jeff mentioned contain both
host and guest skas support, i.e. does it mean if both host and
guest are 2.6.13+ there will be no need to patch either one
for skas (with exception, of course, of bug-fix patches)?

Alexei

From: Blaisorblade <[EMAIL PROTECTED]>
To: user-mode-linux-user@lists.sourceforge.net
CC: "Alexei Potashnik" <[EMAIL PROTECTED]>
Subject: Re: [uml-user] host/guest setup to debug with frame pointers
Date: Wed, 3 Aug 2005 20:21:14 +0200

On Tuesday 02 August 2005 03:07, Alexei Potashnik wrote:
> Could somebody, please, recommend a stable combination of
> the host/guest that was successfully debugged with gdb.
>
> I've looked through blaisorblade's site that this mailing list archive
> and tried the following combinations:
>
> 1. First of all it seems that CONFIG_FRAME_POINTER became
> supported for UML starting with 2.6.11. So I tried 2.6.11+bs4
> and 2.6.12+bs9
It just seems. Actually, at that point it was only separated from DEBUG_INFO;
before, any kernel built with debugging info had frame pointers enabled
(which is reasonable).

I changed that because I wanted to build kernel without debug info but with
frame pointers (that is very useful for getting correct stack traces from the
kernel, for instance on panics or Oopses).
> 2. For the guest file system, I used fc3 image from
> http://www.stearns.org/uml-root/
> Also, I moved /usr/lib/tls to /usr/lib/tls.away inside the image.
The important one is /lib/tls to /lib/tls.away...
> 3. For the host, I've tried FC3, FC4 and FC4 with latest updates.
> Usually I run smp image, but I've tried single-processor versions
> as well.
That shouldn't matter.
> On FC3 I've also tried renaming /lib/tls and /usr/lib/tls.
For the host, this shouldn't be needed.
> On FC4 I tried running guest with
> LD_LIBRARY_PATH=/lib/obsolete/linuxthreads
Again, not needed. With FC4 guests, it may be useful to do this:
./linux <params...> LD_LIBRARY_PATH=/lib/obsolete/linuxthreads

so that the env. var is seen *inside* UML, since there you cannot do
"mv /lib/tls /lib/tls.away".
> 4. I used 6.1 gdb on FC3 and 6.1.1, 6.2.1 and 6.3 gdb on FC4
> (6.3 is really bad... crashes as soon as you try to do list).
> I'm able to debug 2.4.27 guest on RH9, but there is not frame pointers
> in 2.4.27 UML. That is my goal. So any suggestions on getting host/guest
> setup for debugging guest with frame pointer will be greatly appreciated.
Think a 2.6 guest would work as well on RH9 - the only problem is newer GDB
releases, as explained by Jeff. But again, even 2.4 kernels *has* frame
pointers (just verified - even if it's not enabled inside .config, but just
from arch/um/Makefile).

Debugging in SKAS (SKAS0 or SKAS3) mode works much better...
--
Inform me of my mistakes, so I can keep imitating Homer Simpson's "Doh!".
Paolo Giarrusso, aka Blaisorblade (Skype ID "PaoloGiarrusso", ICQ 215621894)
http://www.user-mode-linux.org/~blaisorblade


___________________________________
Yahoo! Mail -  proteggi la tua casella di posta da virus
e posta indesiderata
http://mail.yahoo.it




-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
_______________________________________________
User-mode-linux-user mailing list
User-mode-linux-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-user

Reply via email to