On Mon, Apr 28, 2008 at 6:54 AM, danny webster
<dannywebster at googlemail.com> wrote:
> Hi - i'd like to work on the following bug.
> Bug ID: 6518130
> Synopsis: pfiles(1) should list the processes listening on each socket
> Reported: SunOS 5.10

Hi, Dan,

I'm curious how you propose to solve this problem, as I've also
requested a sponsor for this.  (You can check this page to see if a
bug already has a sponsor request:
http://opensolaris.org/os/bug_reports/request_sponsor/.  I'm more than
happy to withdraw my request if you want this, though.)

You also might want to check out an earlier thread on this
specifically Mike Shapiro's comments on this:

> I'm glad to see this: it's long overdue. The major issue I have though is
> one of implementation: namely, it's way way too heavyweight to use the
> global /proc hammer for this. This really needs to be implemented as
> a kernel service that queries sockfs data structures or whatever
> such that it is just a system call and a hash lookup, and not
> globally stopping and starting every process. The problem with the /proc
> approach is that it means that if you were to have a bunch of junior
> admins run this in production a few times while learning the arguments
> and trying to poke around at some problem, they could literally kill
> performance of the box. Pgrab is *very* heavyweight: you're stopping
> *every* thread of *every process*, kicking it off CPU, making it run
> system calls, etc. This is fine for debugging one bad thing, but
> it's like dropping an atomic bomb on a fully utilized busy machine.

Chad Mynhier

Reply via email to