Messages in this thread | | | Date | Fri, 18 Jul 2008 09:37:59 -0400 | From | "Frank Ch. Eigler" <> | Subject | Re: [RFC] systemtap: begin the process of using proper kernel APIs (part1: use kprobe symbol_name/offset instead of address) |
| |
Hi -
On Fri, Jul 18, 2008 at 08:21:24AM -0500, James Bottomley wrote:
> [...] I'm not entirely convinced systemtap wants full stack > unwinding in the kernel.
Sure we "want" it if we can get it. It enables richer data gathering. It lets scripts act on the contents of the call stack ("is this probe being run due to a callback from this or that shared library?").
> [...] However, systemtap in userspace sees all the relevant dwarf > information as well ... it could do a much better job of unwinding: > give file and line and arguments for function calls, for instance. > All it really needs is to have the relevant pieces of the stack > relayed back.
The relevant bits of stack for a userspace program could include several megabytes per thread; without unwind info we can't be sure which parts are which. (A full gdb symbolic backtrace takes seconds to compute - something we can't possibly afford in situ.)
- FChE
| |