Subject: Re: uvm_fp2 deadlock in 1.6B4
To: Jaromir Dolecek <>
From: tld <>
List: current-users
Date: 08/07/2002 19:50:21
Jaromir Dolecek wrote:
> tld wrote:
>>I am using NetBSD 1.6_BETA4 (jul 27) on i386 and I often find my 
>>processes which access the ext2 partitions (like gzip) stuck with a 
>>uvm_fp2 status. Every other access to the disk gets caught in the 
>>uvm_fp2 lock, and only hard-rebooting helps.
>>Is this a known bug? Am I the only one having this? Is it already 
>>corrected? Is this not the correct ML to report this (if so, please le 
>>me know!)?
> Does this happen with ext2fs only, or any other filesystem as well?
> Would it be possible to break to DDB at this point (via Ctrl-Alt-Esc
> if you have DDB enabled in your kernel), and provide kernel
> traceback of gzip process?

So far it happened 3 times (out of 3 tries to backup the drives), and 
only on ext2fs partitions. Similar problems did not arise with ffs 
(softdep off) when doing the very same things.
One thing I should probably mention is that the ext2 drives suffer from 
soft errors (they are old, and often miss a read or two). But previous 
(1.5.2) version never gave me errors like that.

As for the DDB, I can compile that, find a keyboard and a monitor (I 
can't do that via network, can I?).
I don't know about "provide kernel traceback of gzip process". If given 
appropriate instructions I will gladly follow them and provide with 
anything you might need.

Thank you for your support!

--- TLD
Was I aware whom I had slain I fear I was:the Faceless,the Nameless,the 
Bush set on fire