|
From: George H. <kh...@op...> - 2004-06-06 06:00:09
|
Tom - Thanks for the info. I downloaded the source from CVS today (6 Jun), and build without doing any patches. Although I am unsure of how to turn of vdso support. Googling for vdso seems to imply that it is a `rather undocumented feature' of Linux. Any help on how to disable vdso? (I am using grub as the bootloader). I kinda figured that debugging valgrind with GDB would be difficult, I at first tried using the --wait-for-gdb=yes flag and then attaching, but when valgrind segfaulted and I tried to attach with GDB there was no valgrind process to attach to. I also tried the command: valgrind --tool=memcheck --help which also seg-faulted on me. Base on this, is why I attempted to debug valgrind directly with GDB. The seg-fault occures within the first thirty instructions being executed. George |
|
From: Tom H. <th...@cy...> - 2004-06-06 07:23:12
|
In message <010f01c44b8b$84a771c0$0201a8c0@aelfin.local>
George Huber <kh...@op...> wrote:
> I downloaded the source from CVS today (6 Jun), and
> build without doing any patches. Although I am unsure
> of how to turn of vdso support. Googling for vdso
> seems to imply that it is a `rather undocumented feature'
> of Linux. Any help on how to disable vdso? (I am
> using grub as the bootloader).
Either add vdso=0 to the boot options in grub or run this command:
sysctl -w kernel.vdso=0
> I kinda figured that debugging valgrind with GDB would
> be difficult, I at first tried using the --wait-for-gdb=yes
> flag and then attaching, but when valgrind segfaulted
> and I tried to attach with GDB there was no valgrind
> process to attach to. I also tried the command:
>
> valgrind --tool=memcheck --help
>
> which also seg-faulted on me. Base on this, is why I
> attempted to debug valgrind directly with GDB. The
> seg-fault occures within the first thirty instructions
> being executed.
If you use wait-for-gdb and it dies before it waits then you
probably do need to run gdb directly. In this case however it
may have been dying between the transfer from phase1 to phase2
and the point where phase2 pauses to allow GDB to be attached
in which case you are pretty much stuffed at the moment as
a directly attached GDB will appear to be at the point where
the transfer to phase2 happens which isn't where the problem
really is.
Tom
--
Tom Hughes (th...@cy...)
Software Engineer, Cyberscience Corporation
http://www.cyberscience.com/
|