concept prove patch for ktrace output to all file types

[Robert Watson]

On Thu, 28 Jun 2007, Howard Su wrote:

> This is a concept prove patch for ktrace. In order to make ktrace write to 
> stdout, /dev/xxx, fifo file, pipe, I changed ktrace syscall to accepting a 
> file handler instead of a file name.
>
> I just get a workable kernel patch. And i make as small as possible changes
> to userland code. Now you can do something like:
> #mkfifo foo
> in session A: #kdump -f foo
> in session B: #ktrace -f foo ls
>
> Then, you can see a real time output like truss did.
>
> Before moving forward (hacking userland), i want to get some feedback if my 
> changes in kernel part is the right direction to do. Appreciate your time.
>
> PS: The patch can be clearly apply to today's current.

What happens to processes associated with the same or other ktrace sessions if 
one ktrace session stalls due to a fifo or pipe buffer filling?

Robert N M Watson
Computer Laboratory
University of Cambridge