Re: System Hang



When I did !stacks all my treads were blocked most of them on
nt!KiUnexpectedInterrupt+0x29f
I'm not sure what could cause this.

"Frank The Tank" wrote:

I can break in when the system hangs. With the sugestions you guys have given
me I have went from no information to all kinds. Now I have to learn how to
interpret it.
I'm a hardware guy who writes the ocasional driver when needed so I'm a
little slow at figuring out what all these dumps mean.
Thanks

"Mark Roddy" wrote:

On Tue, 2 May 2006 09:11:02 -0700, Frank The Tank
<FrankTheTank@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote:



"Pavel A." wrote:


As usual... windbg with /crashdebug or scroll lock break-in (KB 244139),
driver verifier, checked OS
( since, as you wrote, it worked for years - is this win2000? )

--PA


I'm running it with /debug with windbg hooked up now. When it hangs there
is no indication on the debug screen.

It is unclear from your posts if you can actually break into the
system with the debugger when the system is 'hung'. There are general
two classes of 'hang', one which allows debugger operations and the
other which does not.

If you are in the first category, the debugger works, then follow the
advice given, although I generally start with examining the current
thread stack of each logical processor using '~0 ; kb' substituting
1-N for ) for each logical processor. This gives you a snapshot of
current threads and from that a spinlock deadlock will generally just
appear.

In the second case you need to use runtime tracing (debug prints)
and/or a 'NMI board'. The NMI board is a PCI device that your plug
into your system and that has a button on it that generates an NMI
that will usually get you to a NMI bugcheck from which you can then
proceed with debugging either live or from a crashdump.

My next step was to load the checked OS.
The driver started as a NT4, I modified it for win2000 and most recently for
win2003. It hangs on both 2000 and 2003.

I hangs quicker with 2 cpu's than with 4. The more dma xfers I add the
quicker it hangs. My current test to kill it is 80 dma's every 400ms,
(32bytes, then 56K bytes,32 then 56k....)
Also this driver does not do overlapped i/o, so only one operation should be
happening at a time. Thanks for any input.


=====================
Mark Roddy DDK MVP
Windows Vista/2003/XP/2000 Consulting
Device and Filesystem Drivers
Hollis Technology Solutions 603-321-1032
www.hollistech.com

.

Quantcast