[prev in list] [next in list] [prev in thread] [next in thread] 

List:       hpux-devtools
Subject:    Re: HPUX-DEVTOOLS: HP-UX kernel preemption question
From:       Stan Sieler <sieler () allegro ! com>
Date:       2008-04-18 22:18:19
Message-ID: 20080418221819.8DDCA52537C () opus ! allegro ! com
[Download RAW message or body]

Re:

> > Most references I can find imply that HP-UX is not a preemptable kernel,
> > but that's clearly not quite true (some HP-UX patches add "preemption points" :)
> 
> Are those in-kernel or more in-system-call-path?

Probably not within system call, unless there are system calls that involve
the paging code or garbage collecting.

However, my driver isn't associated with those kinds of calls.
 
> > I'm trying to track down a problem that might barely be attributable to
> > my driver being preempted and some unlikely other things happening before I
> > get the CPU back.
> 
> You holding a spinlock?  Perhaps a nested interrupt?

In the case of the panic I'm looking at, no. 

Based on a strange stacktrace (or "backtrace") from a system panic, if a process had itself as
its parent (procp->p_ppid having its own pid), that would cause the problem
I saw.  Or, if multiple preemptions came in, changing the ordering of process
family links in just the right exceedingly unlikely order, that could cause
the problem I saw.

I'm changing my code to avoid either situation anyway :)
 
-- 
Stan Sieler
work:     www.allegro.com
personal: www.sieler.com/wanted/index.html  

 _________________________________________________________________
 To leave this mailing list, send mail to majordomo@cxx.cup.hp.com
    with the message UNSUBSCRIBE hpux-devtools
 _________________________________________________________________
[prev in list] [next in list] [prev in thread] [next in thread]