Subject: kern/29528: NFS locks on UDP transport
To: None <kern-bug-people@netbsd.org, gnats-admin@netbsd.org,>
From: None <mishka@apk.od.ua>
List: netbsd-bugs
Date: 02/25/2005 12:40:00
>Number:         29528
>Category:       kern
>Synopsis:       NFS locks on UDP transport
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Fri Feb 25 12:40:00 +0000 2005
>Originator:     Mike M. Volokhov
>Release:        NetBSD 2.0
>Organization:
ISPK
>Environment:
NetBSD betty.holding.local 2.0_STABLE NetBSD 2.0_STABLE (BETTY-$Revision: 1.7 $) #1: Fri Feb 18 12:30:25 EET 2005 mishka@nostromo.holding.local:/build/netbsd/src-2/sys/arch/i386/compile/BETTY i386
Architecture: i386
Machine: i386
>Description:

	Greetings!

	After recent upgarde to 2.0_STABLE (18 Feb 2005) from 2.0
	(27 Jan 2005) my NFS mounts on updated machine very often
	(several times per day or even per hour) hangs with "nfsrcvlk"
	state. This host used as NFS server and client, and deadlocks
	happens for both remote and exported local mounts.

	Seems it occurs when data are sent to remotes only, i.e.
	during reading by clients or writing to server, but this
	is not exactly.

	A "-bis" flags are used for all mounts.

	NFS server and clients are all on local network.

>How-To-Repeat:

	Upgrade OS to latest netbsd-2 branch. The run NFS server
	or mount remote filesystem using UDP transport. Create a
	huge traffic in way from NetBSD 2.0 to remote peer.

>Fix:

	Seems this problem may be workarounded by switching to TCP
	transport, i.e. using "-Tbis" flags to mount_nfs.

	Please let me know if you will need any additional info.

	--
	Best regards,
	Mishka.