On Wed, Oct 12, 2005 at 03:58:02PM -0400, Andrew Gallatin wrote: > > Frank Hofmann - Solaris Sustaining writes: > > > > In kernel debugging, you usually pull an address out of thin air. > > > In (gdb/dbx) I typically pull an address off the stack in the form of > local variables or function arguments. In gdb parlance, a typical > debugging session for me will be something like: > > backtrace > frame 4 > print *local_variable > print *function_arg0 > <...>
The problem is that in kernel debugging, you don't have all of the debugging information a typical user-land program compiled with '-g' has; we only have the type definitions (and C function argument/return types). There isn't enough information available for the debugger to figure it all out a priori. If kmem debugging is on (kmem_flags is 0xf), you have a stack trace from every allocation, which makes figuring out the type fairly easy, given the source code. Without kmem debugging, ::typegraph does a very good job. Cheers, - jonathan -- Jonathan Adams, Solaris Kernel Development
