NetBSD Problem Report #10278

Received: (qmail 25547 invoked from network); 4 Jun 2000 15:53:22 -0000
Message-Id: <200006041553.AAA01250@miyuki.my.domain>
Date: Mon, 5 Jun 2000 00:53:10 +0900 (JST)
From: IWAMOTO Toshihiro <iwamoto@sat.t.u-tokyo.ac.jp>
Reply-To: IWAMOTO Toshihiro <iwamoto@sat.t.u-tokyo.ac.jp>
To: gnats-bugs@gnats.netbsd.org
Subject: kernfs may hang kernel
X-Send-Pr-Version: 3.95

>Number:         10278
>Category:       kern
>Synopsis:       kernfs may hang kernel
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    jdolecek
>State:          closed
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Sun Jun 04 15:54:00 +0000 2000
>Closed-Date:    Fri Jul 19 18:38:07 +0000 2002
>Last-Modified:  Fri Jul 19 18:38:07 +0000 2002
>Originator:     IWAMOTO Toshihiro
>Release:        2 days old -current
>Organization:

>Environment:

System: NetBSD 1.4Z i386


>Description:
	kernfs_getattr goes to an infinite loop if attributes of
	/kern/msgbuf is requested and the kernel msgbuf is broken.
	An infinite loop also happens if the hostname is 256-bytes
	long and attributes of /kern/hostname is requested.

	This happens because kernfs_getattr calls kernfs_xread in
	a loop until it returns 0 (around line 498 of kernfs_vnops.c),
	and kernfs_xread repeatedly returns non-zero values as error
	codes in the above situations.


>How-To-Repeat:
	build a kernel with a broken msgbuf or set its hostname
	to 256-bytes long, and do "ls -l /kern"
>Fix:
	The following code should work around the problem.
	I don't know if better error handling is necessary.

--- kernfs_vnops.c.orig Thu Aug 26 07:54:17 1999
+++ kernfs_vnops.c      Mon Jun  5 00:40:50 2000
@@ -234,7 +234,7 @@
                 */
                if (!msgbufenabled || msgbufp->msg_magic != MSG_MAGIC) {
                        msgbufenabled = 0;
-                       return (ENXIO);
+                       return (0);
                }

                /*
@@ -261,7 +261,7 @@
                int xlen = hostnamelen;

                if (xlen >= (len-2))
-                       return (EINVAL);
+                       return (0);

                memcpy(*bufp, cp, xlen);
                (*bufp)[xlen] = '\n';
>Release-Note:
>Audit-Trail:

From: Martin Husemann <martin@rumolt.teuto.de>
To: iwamoto@sat.t.u-tokyo.ac.jp (IWAMOTO Toshihiro)
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: kern/10278: kernfs may hang kernel
Date: Sun, 4 Jun 2000 18:30:24 +0200 (MEST)

 Wouldn't it be preferable to fix the loops in kernfs_getattr to check
 for the error return codes?


 Martin
Responsible-Changed-From-To: kern-bug-people->jdolecek 
Responsible-Changed-By: jdolecek 
Responsible-Changed-When: Tue Jul 9 16:00:58 PDT 2002 
Responsible-Changed-Why:  
I'll look at this. 
State-Changed-From-To: open->closed 
State-Changed-By: jdolecek 
State-Changed-When: Fri Jul 19 11:36:15 PDT 2002 
State-Changed-Why:  
The bug was fixed, by changing kernfs_xread() to return errno or 0, and 
passing len via separate argument. 
This has been fixed in rev. 1.82 of syssrc/sys/miscfs/kernfs/kernfs_vnops.c 
Thanks for report. 
>Unformatted:

NetBSD Home
NetBSD PR Database Search

(Contact us) $NetBSD: query-full-pr,v 1.51 2026/08/10 02:28:17 riastradh Exp $
$NetBSD: gnats_config.sh,v 1.10 2026/05/13 22:00:09 riastradh Exp $
Copyright © 1994-2026 The NetBSD Foundation, Inc. ALL RIGHTS RESERVED.