>Number:         23869
>Category:       kern
>Synopsis:       Deadlocks when writing to /dev/wd*
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Wed Dec 24 08:27:00 UTC 2003
>Closed-Date:
>Last-Modified:
>Originator:     Christian Biere
>Release:        NetBSD 1.6ZG
>Organization:
>Environment:
System: NetBSD cyclonus 1.6ZG NetBSD 1.6ZG (STARSCREAM) #0: Tue Dec 16 17:11:50 CET 2003 bin@cyclonus:/usr/build/arch/i386/compile/STARSCREAM i386
Architecture: i386
Machine: i386
>Description:

When I zero-out a disk with dd without using the raw device, newly started
processes hang in state "getnewbuf" and after a while (kernel) processes will
hang in "vnlock". I've found a description of this problem which shows that
this isn't a current one:

	http://mail-index.netbsd.org/current-users/2001/03/07/0003.html

The first time this happened mutt also hung in state "getblk". I don't
know whether that was just a strange coincident. I successfully tried
to reproduce the problem as described above several times. Strangely,
after a reboot vim immediately hung in state "getblk" although I had
just rebooted and had not try to reproduce the problem. So maybe that's
another (new) problem with cgd(4) on which the $HOME is hosted.

>How-To-Repeat:

# dd if=/dev/zero of=/dev/wdXY bs=64k

Now try something from another vt or put dd in the background:

$ ls

Recognize how slow it starts. Try 'top' and see that new processes
hang in "getnewbuf" for quite a while.

>Fix:
>Release-Note:
>Audit-Trail:
>Unformatted: