On Fri, 10 Nov 2000 10:12:55 -0800 (PST),
ajones@xxxxxxxxxx wrote:
>I'm frustrated by the backtrace function in kdb for the 2.2.13 kernel.
>The module I'm working with only seems to compile with -O2. However,
>kdb's bt seems to be confused by the call stacks used with -O2 and
>it gives up early. Has there been any work to improve the backtrace
>function since the 2.2.13 release?
There is an unofficial backport of kdb v1.5 to 2.2.18-pre15 in
ftp://oss.sgi.com/projects/kdb/download/ix86/kdb-v1.5-2.2.18-pre15.gz
It is not being supported nor updated to new 2.2 kernels.
>With the Solaris kdb I used to hunt for stack pointers in the stack
>bejond a trap were kdb gives up. It might be useful to automate this
>function; have bt <addr> scan a couple of hundred bytes to look for
>signs of a stack frame; pointers near their address next (into the
>stack) next to pointers near known symbols (return pointers).
Read the comments in kdb about ix86 stack frames, they are horrible.
"mds %esp" is useful to manually back track.
|