Re: [uml-devel] actually useful backtrace from a CPU-chewing hang

[Jeff Dike <jdike@addtoit.com>]

On Sat, May 03, 2008 at 12:21:15AM +0100, Nix wrote:
> With this patch (migrating most of the work into os_nsecs(), with a
> non-NO_HZ version doing skew computations too, atop your first patch,
> and making a couple of variables static for good measure), I still had
> no luck:

From a quick look, this seems right.  And this would be the gold
standard of preventing UML from seeing time going backwards.

> *slap* why didn't I think of that?
  ....
> ... unfortunately it's not terribly informative. Here, with an earlier
> gettimeofday() included for context:
> 
> So there's basically nothing unusual here. The first gettimeofday() call
> after the jump, and *whoompf* we're off into the magic land of looping
> with no extra syscalls to speak of at all.

This was with a solid hang?  The thing is still handling page faults
from something at the end of that trace.  I would try again, making
sure it's a solid hang, and including enough of the trace so that you
stop seeing page faults:
     waitpid(9963, [{WIFSTOPPED(s) && WSTOPSIG(s) == SIGSEGV}])
and system calls (WSTOPSIG(s) == 133) and see what's happening at that
point.

And if you never stop seeing page faults or system calls, then that's
interesting too.

				Jeff

-- 
Work email - jdike at linux dot intel dot com

-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference 
Don't miss this year's exciting event. There's still time to save $100. 
Use priority code J8TL2D2. 
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel