NetBSD Problem Report #21292

Received: (qmail 21120 invoked by uid 605); 23 Apr 2003 19:53:21 -0000
Message-Id: <20030423195319.92B1711152@narn.netbsd.org>
Date: Wed, 23 Apr 2003 19:53:19 +0000 (UTC)
From: paulpp@jvl-ent.com
Sender: gnats-bugs-owner@netbsd.org
Reply-To: paulpp@jvl-ent.com
To: gnats-bugs@gnats.netbsd.org
Subject: ping -s10000 -f    brokes IP-stack
X-Send-Pr-Version: www-1.0

>Number:         21292
>Category:       kern
>Synopsis:       ping -s10000 -f    brokes IP-stack
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Wed Apr 23 19:54:00 +0000 2003
>Closed-Date:    
>Last-Modified:  Mon Apr 28 14:40:01 +0000 2003
>Originator:     Pavel Podvoiski
>Release:        1.6.1
>Organization:
JVL Corporation
>Environment:
NetBSD cmynop 1.6.1 NetBSD 1.6.1 (GENERIC) #0: Tue Apr  8 12:05:52 UTC 2003
autobuild@tgm.daemon.org:/autobuild/netbsd-1-6/i386/OBJ/autobuild/netbsd-1-6/src/sys/arch/i386/compile/GENERIC i386
>Description:
Hardware
    CPU: P3-1GHz
    RAM: 256MB PC-133
    NIC: RTL8139C connected to 10Mb hub

Shortly after start 'ping -s10000 -f' series of messages
"ping: sendto: No buffer space available"
appears on screen. Then it's turns to a constant stream of such messages.

After terminating ping by ^C, computer can't ping anything on network
and can't be pinged.
>How-To-Repeat:

>Fix:

>Release-Note:
>Audit-Trail:

From: "Perry E. Metzger" <perry@piermont.com>
To: paulpp@jvl-ent.com
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: kern/21292: ping -s10000 -f    brokes IP-stack
Date: 23 Apr 2003 23:31:47 -0400

 paulpp@jvl-ent.com writes:
 > Shortly after start 'ping -s10000 -f' series of messages
 > "ping: sendto: No buffer space available"
 > appears on screen. Then it's turns to a constant stream of such messages.

 That means you've run out of buffers. You've instructed ping to send
 out pings as fast as the program can, and the program is capable of
 sending them faster than your driver will output them. The constant
 stream of the messages is completely understandable -- you asked it to
 try sending as fast as possible and it is giving you errors on each
 packet you are trying to send!

 A common cause of being out of buffer space is that you've lost your
 network connection.

 > After terminating ping by ^C, computer can't ping anything on network
 > and can't be pinged.

 Are you sure you can consistently reproduce this and that the problem
 doesn't go away when your buffers drain?

 What happens when you ifconfig the interface after trying this? What
 does it say about the interface's status?

 Perry

From: Pavel Podvoiski <paulpp@jvl-ent.com>
To: "Perry E. Metzger" <perry@piermont.com>
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: kern/21292: ping -s10000 -f    brokes IP-stack
Date: Thu, 24 Apr 2003 08:11:45 -0400

 On 23 Apr 2003 23:31:47 -0400, Perry E. Metzger <perry@piermont.com> wrote:

 > A common cause of being out of buffer space is that you've lost your
 > network connection.

 Nope, I'm runing it on LAN. Note, that NIC-10/100 connected to HUB-10

 >> After terminating ping by ^C, computer can't ping anything on network
 >> and can't be pinged.
 >
 > Are you sure you can consistently reproduce this and that the problem
 > doesn't go away when your buffers drain?

 It does not "drain" -- LED shows no activity.

 Wow! Looks like it depends on what machine I'm blasting! ???
 I think it's better to draw diagram:

 NetBSD 1.6.1 ----\
 RTL8139C         |
 10.0.0.221       |                                      Win2000             
 HUB-10 --- SWITCH-10/100 --- HUB-100 --- 3C905B
 Win98SE          |                                      10.0.0.20
 3C905B           |
 10.0.0.22 -------/

 	ping -s10000 -f 10.0.0.22
 Bursts of "No buffer space". System holding.

 	ping -s10000 -f 255.255.255.255
 Larger bursts of "No buffer space". System holding.

 	ping -s60000 -f 10.0.0.22
 	ping -s60000 -f 255.255.255.255
 System holding.

 	ping -s10000 -f 10.0.0.20
 Bursts of "No buffer space" became larger and larger and soon it's a
 constant stream of it (at this point LED goes from constant blinking to
 sporadic blinks). After ^C -- net is dead.

 	ifconfig rtk0 down; ifconfig rtk0 up
 Brings net back to life.

 > What happens when you ifconfig the interface after trying this?
 > What does it say about the interface's status?

 **** BEFORE DEAD ****
 rtk0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
 	address: 00:e0:4c:39:00:0a
 	media: Ethernet autoselect (none)
 	status: active
 	inet 10.0.0.221 netmask 0xffffff00 broadcast 10.0.0.255
 	inet6 fe80::2e0:4cff:fe39:a%rtk0 prefixlen 64 scopeid 0x1

 **** AFTER DEAD ****
 rtk0: flags=8c43<UP,BROADCAST,RUNNING,OACTIVE,SIMPLEX,MULTICAST> mtu 1500
 	address: 00:e0:4c:39:00:0a
 	media: Ethernet autoselect (none)
 	status: active
 	inet 10.0.0.221 netmask 0xffffff00 broadcast 10.0.0.255
 	inet6 fe80::2e0:4cff:fe39:a%rtk0 prefixlen 64 scopeid 0x1

 The only difference I can see is OACTIVE flag.

 -- 
 Pavel Podvoiski
 Software Engineer


From: "Perry E. Metzger" <perry@piermont.com>
To: Pavel Podvoiski <paulpp@jvl-ent.com>
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: kern/21292: ping -s10000 -f    brokes IP-stack
Date: 24 Apr 2003 14:30:40 -0400

 Pavel Podvoiski <paulpp@jvl-ent.com> writes:
 > 	ifconfig rtk0 down; ifconfig rtk0 up
 > Brings net back to life.

 Okay, so it appears that somehow excess activity is jamming up the
 interface but downing it and upping fixes it.

 > > What happens when you ifconfig the interface after trying this?
 > > What does it say about the interface's status?
 > 
 > **** BEFORE DEAD ****
 > rtk0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500

 > **** AFTER DEAD ****
 > rtk0: flags=8c43<UP,BROADCAST,RUNNING,OACTIVE,SIMPLEX,MULTICAST> mtu 1500

 > The only difference I can see is OACTIVE flag.

 I believe OACTIVE means "transmission in progress". The card still
 believes it is transmitting.

 If you do this with something other than a Realtek does it work okay?
 That will narrow this down to being a realtek driver bug.

 Also, what version of the rtk driver are you running exactly?

 -- 
 Perry E. Metzger		perry@piermont.com

From: "Perry E. Metzger" <perry@piermont.com>
To: Pavel Podvoiski <paulpp@jvl-ent.com>
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: kern/21292: ping -s10000 -f    brokes IP-stack
Date: 24 Apr 2003 15:36:46 -0400

 Pavel Podvoiski <paulpp@jvl-ent.com> writes:
 > On 24 Apr 2003 14:30:40 -0400, Perry E. Metzger <perry@piermont.com> wrote:
 > 
 > > If you do this with something other than a Realtek does it work okay?
 > > That will narrow this down to being a realtek driver bug.
 > 
 > Was unable to reproduce it with 3C905C.

 Okay, so it is certainly a realtek problem.

 > > Also, what version of the rtk driver are you running exactly?
 > 
 > NetBSD: if_rtk_pci.c,v 1.12 2002/02/06 12:43:05 tsutsui

 Could you check if the problem is still present in -current?
 The current driver is many revisions newer. Even just booting a
 -current kernel from the last few days should be enough to test the
 problem.

 Perry

From: Pavel Podvoiski <paulpp@jvl-ent.com>
To: "Perry E. Metzger" <perry@piermont.com>
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: kern/21292: ping -s10000 -f    brokes IP-stack
Date: Thu, 24 Apr 2003 10:33:21 -0400

 On 24 Apr 2003 14:30:40 -0400, Perry E. Metzger <perry@piermont.com> wrote:

 > If you do this with something other than a Realtek does it work okay?
 > That will narrow this down to being a realtek driver bug.

 Was unable to reproduce it with 3C905C.

 > Also, what version of the rtk driver are you running exactly?

 NetBSD: if_rtk_pci.c,v 1.12 2002/02/06 12:43:05 tsutsui

 -- 
 Pavel Podvoiski
 Software Engineer


From: Pavel Podvoiski <paulpp@jvl-ent.com>
To: "Perry E. Metzger" <perry@piermont.com>
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: kern/21292: ping -s10000 -f    brokes IP-stack
Date: Fri, 25 Apr 2003 05:36:37 -0400

 On 24 Apr 2003 15:36:46 -0400, Perry E. Metzger <perry@piermont.com> wrote:

 > Could you check if the problem is still present in -current?

 Tried GENERIC kernel from 2003/04/17. Problem is still there.

 if_rtk_pci.c,v 1.18 2003/01/29 12:57:36 kanaoka

 -- 
 Pavel Podvoiski
 Software Engineer


From: "Perry E. Metzger" <perry@piermont.com>
To: Pavel Podvoiski <paulpp@jvl-ent.com>
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: kern/21292: ping -s10000 -f    brokes IP-stack
Date: 25 Apr 2003 10:46:03 -0400

 Pavel Podvoiski <paulpp@jvl-ent.com> writes:
 > On 24 Apr 2003 15:36:46 -0400, Perry E. Metzger <perry@piermont.com> wrote:
 > 
 > > Could you check if the problem is still present in -current?
 > 
 > Tried GENERIC kernel from 2003/04/17. Problem is still there.

 You should do more recent than that -- I think a couple of other
 things have changed. There should be a GENERIC from a couple of days
 ago available.

 > if_rtk_pci.c,v 1.18 2003/01/29 12:57:36 kanaoka

 -- 
 Perry E. Metzger		perry@piermont.com

From: Pavel Podvoiski <paulpp@jvl-ent.com>
To: "Perry E. Metzger" <perry@piermont.com>
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: kern/21292: ping -s10000 -f    brokes IP-stack
Date: Mon, 28 Apr 2003 04:26:35 -0400

 On 25 Apr 2003 10:46:03 -0400, Perry E. Metzger <perry@piermont.com> wrote:

 >> Tried GENERIC kernel from 2003/04/17. Problem is still there.
 >
 > You should do more recent than that -- I think a couple of other
 > things have changed. There should be a GENERIC from a couple of days
 > ago available.

 Tried GENERIC kernel from 2003-04-25 -- same thing.

 -- 
 Pavel Podvoiski
 Software Engineer

>Unformatted:

Home	PR Database Search