ifconfig up/down problem

ifconfig up/down problem

[Sven-Haegar Koch]

hi...

Short: ifconfig ethX down locks

Kernel: 2.4.22-bk12 (same problem with 2.4.23-pre3)
eth0: eepro100
eth1: orinoco_cs (orinoco mini-pci)
System: Toshiba Satellite Pro 4600 Laptop, P3 700Mhz

Just after booting, no X startet yet, interface not yet initialized:

aurora:~# ifconfig eth1 down
aurora:~# ifconfig eth1 up
aurora:~# ifconfig eth1 down
aurora:~# ifconfig eth1 up
aurora:~# ifconfig eth1 down
<--lock here, shell does not return, even ctrl-c does not help

haegar@aurora:~$ ps axl|grep ifconfig
4     0  1041  1035   9   0  1448  404 dev_cl S    pts/0      0:00 ifconfig eth1

top shows ifconfig consuming 100% cpu, 100% system

The same happens with eth0, there it takes only two up/down cycles,
perhaps because it is already configured with ipv4+ipv6 addresses, and the
same happens using '/sbin/ip link set eth0 up/down' too.

Kernel 2.4.20-pre2-ac3 is ok (my last kernel, running for month')

c'ya
sven

ps: kernel config available under http://sdinet.de/tmp/config-2.4.22-bk12
ps2: loaded modules:
	Module                  Size  Used by    Not tainted
	input                   3328   0 (autoclean)
	ipt_limit                888  10 (autoclean)
	ipt_REJECT              3480   2
	ipt_LOG                 3416  10
	ipt_state                568  12
	ip_conntrack_ftp        4176   0 (unused)
	ip_conntrack           18116   3 [ipt_state ip_conntrack_ftp]
	iptable_filter          1740   1
	ip_tables              12448   5 [ipt_limit ipt_REJECT ipt_LOG ipt_state iptable_filter]
	orinoco_cs              4596   1
	orinoco                36116   0 [orinoco_cs]
	hermes                  7044   0 [orinoco_cs orinoco]
	ds                      6996   3 [orinoco_cs]
	acm                     6464   0 (unused)
	pwc                    44432   0 (unused)
	videodev                6048   0 [pwc]
	audio                  43256   0 (unused)
	soundcore               3652   0 [audio]
	usb-uhci               23696   0 (unused)
	ide-scsi               10320   0
	i810_rng                3200   0 (unused)
	serial                 48228   0 (autoclean)
	isa-pnp                32560   0 (autoclean) [serial]
	parport_pc             23784   1 (autoclean)
	lp                      7040   0 (autoclean)
	parport                25512   1 (autoclean) [parport_pc lp]
	sd_mod                 11180   0 (unused)
	scsi_mod               87216   2 [ide-scsi sd_mod]
	usbcore                62956   1 [acm pwc audio usb-uhci]
	yenta_socket           12928   3
	pcmcia_core            43520   0 [orinoco_cs ds yenta_socket]
	agpgart                25320   0 (unused)
	ipv6                  174100  -1
	eepro100               21972   1
	mii                     2432   0 [eepro100]
	rtc                     6792   0 (autoclean)

--

The Internet treats censorship as a routing problem, and routes around it.
(John Gilmore on http://www.cygnus.com/~gnu/)

[PATCH] Re: ifconfig up/down problem

[Jeff Garzik]

[-- Attachment #1: Type: text/plain, Size: 1035 bytes --]

Sven-Haegar Koch wrote:
> hi...
> 
> Short: ifconfig ethX down locks
> 
> Kernel: 2.4.22-bk12 (same problem with 2.4.23-pre3)
> eth0: eepro100
> eth1: orinoco_cs (orinoco mini-pci)
> System: Toshiba Satellite Pro 4600 Laptop, P3 700Mhz
> 
> Just after booting, no X startet yet, interface not yet initialized:
> 
> aurora:~# ifconfig eth1 down
> aurora:~# ifconfig eth1 up
> aurora:~# ifconfig eth1 down
> aurora:~# ifconfig eth1 up
> aurora:~# ifconfig eth1 down
> <--lock here, shell does not return, even ctrl-c does not help
> 
> haegar@aurora:~$ ps axl|grep ifconfig
> 4     0  1041  1035   9   0  1448  404 dev_cl S    pts/0      0:00 ifconfig eth1
> 
> top shows ifconfig consuming 100% cpu, 100% system
> 
> The same happens with eth0, there it takes only two up/down cycles,
> perhaps because it is already configured with ipv4+ipv6 addresses, and the
> same happens using '/sbin/ip link set eth0 up/down' too.
> 
> Kernel 2.4.20-pre2-ac3 is ok (my last kernel, running for month')



Does the attached patch fix it?

	Jeff



[-- Attachment #2: patch --]
[-- Type: text/plain, Size: 510 bytes --]

diff -Nru a/net/core/dev.c b/net/core/dev.c
--- a/net/core/dev.c	Mon Sep  8 18:14:36 2003
+++ b/net/core/dev.c	Mon Sep  8 18:14:36 2003
@@ -851,7 +851,11 @@
 	 * engine, but this requires more changes in devices. */
 
 	smp_mb__after_clear_bit(); /* Commit netif_running(). */
-	netif_poll_disable(dev);
+	while (test_bit(__LINK_STATE_RX_SCHED, &dev->state)) {
+		/* No hurry. */
+		current->state = TASK_INTERRUPTIBLE;
+		schedule_timeout(1);
+	}
 
 	/*
 	 *	Call the device specific close. This cannot fail.

Re: [PATCH] Re: ifconfig up/down problem

[Sven-Haegar Koch]

On Mon, 8 Sep 2003, Jeff Garzik wrote:

> Sven-Haegar Koch wrote:
> > Short: ifconfig ethX down locks
> Does the attached patch fix it?

Yes, it does - thanks a lot!

c'ya
sven

-- 

The Internet treats censorship as a routing problem, and routes around it.
(John Gilmore on http://www.cygnus.com/~gnu/)

Re: [PATCH] Re: ifconfig up/down problem

[Fedor Karpelevitch]

Jeff Garzik wrote:
> Sven-Haegar Koch wrote:
> > hi...
> >
> > Short: ifconfig ethX down locks
> >
> > Kernel: 2.4.22-bk12 (same problem with 2.4.23-pre3)
> > eth0: eepro100
> > eth1: orinoco_cs (orinoco mini-pci)
> > System: Toshiba Satellite Pro 4600 Laptop, P3 700Mhz
> >
> > Just after booting, no X startet yet, interface not yet
> > initialized:
> >
> > aurora:~# ifconfig eth1 down
> > aurora:~# ifconfig eth1 up
> > aurora:~# ifconfig eth1 down
> > aurora:~# ifconfig eth1 up
> > aurora:~# ifconfig eth1 down
> > <--lock here, shell does not return, even ctrl-c does not help
> >
> > haegar@aurora:~$ ps axl|grep ifconfig
> > 4     0  1041  1035   9   0  1448  404 dev_cl S    pts/0     
> > 0:00 ifconfig eth1
> >
> > top shows ifconfig consuming 100% cpu, 100% system
> >
> > The same happens with eth0, there it takes only two up/down
> > cycles, perhaps because it is already configured with ipv4+ipv6
> > addresses, and the same happens using '/sbin/ip link set eth0
> > up/down' too.
> >
> > Kernel 2.4.20-pre2-ac3 is ok (my last kernel, running for month')
>
> Does the attached patch fix it?
>
> 	Jeff


does not help me (assuming I have the same problem). I have a total 
lockup a few seconds after setting up the interface (not 
immidiately).

Fedor.

Re: [PATCH] Re: ifconfig up/down problem

[Fedor Karpelevitch]

Fedor Karpelevitch wrote:
> Jeff Garzik wrote:
> > Sven-Haegar Koch wrote:
> > > Kernel 2.4.20-pre2-ac3 is ok (my last kernel, running for
> > > month')
> >
> > Does the attached patch fix it?
> >
> > 	Jeff
>
> does not help me (assuming I have the same problem). I have a total
> lockup a few seconds after setting up the interface (not
> immidiately).
>
> Fedor.

actually it seemed to have helped with 2.6.0-test5 where I was 
apparently having the same issue. Not with 2.4.23-pre3 however...

Fedor.

Re: [PATCH] Re: ifconfig up/down problem

[Fedor Karpelevitch]

Fedor Karpelevitch wrote:
> Fedor Karpelevitch wrote:
> > Jeff Garzik wrote:
> > > Sven-Haegar Koch wrote:
> > > > Kernel 2.4.20-pre2-ac3 is ok (my last kernel, running for
> > > > month')
> > >
> > > Does the attached patch fix it?
> > >
> > > 	Jeff
> >
> > does not help me (assuming I have the same problem). I have a
> > total lockup a few seconds after setting up the interface (not
> > immidiately).
> >
> > Fedor.
>
> actually it seemed to have helped with 2.6.0-test5 where I was
> apparently having the same issue. Not with 2.4.23-pre3 however...
>
> Fedor.

I am really sorry for giving all this misleading information but now 
it works fine for me with 2.4.23-pre3 as well, but I bet it did lock 
up the first time I tried it. So there may be something wrong with me 
or it is some other random problem I am seeing...

Fedor

Re: [PATCH] Re: ifconfig up/down problem

[Fedor Karpelevitch]

Fedor Karpelevitch wrote:
> Fedor Karpelevitch wrote:
> > Fedor Karpelevitch wrote:
> > > Jeff Garzik wrote:
> > > > Sven-Haegar Koch wrote:
> > > > > Kernel 2.4.20-pre2-ac3 is ok (my last kernel, running for
> > > > > month')
> > > >
> > > > Does the attached patch fix it?
> > > >
> > > > 	Jeff
> > >
> > > does not help me (assuming I have the same problem). I have a
> > > total lockup a few seconds after setting up the interface (not
> > > immidiately).
> > >
> > > Fedor.
> >
> > actually it seemed to have helped with 2.6.0-test5 where I was
> > apparently having the same issue. Not with 2.4.23-pre3 however...
> >
> > Fedor.
>
> I am really sorry for giving all this misleading information but
> now it works fine for me with 2.4.23-pre3 as well, but I bet it did
> lock up the first time I tried it. So there may be something wrong
> with me or it is some other random problem I am seeing...
>
> Fedor

shit, it DOES happen. somehow, when network cable is unplugged it 
seems to never happen, when I am plugged in at home I believe I could 
have gotten that lockup once or twice, but when I am plugged in to 
the network at my office it seems to happen 100% of the time. I use 
DHCP in either case. I wonder if it could be related to the noise 
traffic on the network or what. I get it with both 2.4.23-pre3 and 
2.6.0-test5 with your patch. It does not seem to occur at any 
particular time - just some time (seconds to minute or two) after the 
boot up.

Could you suggest any way to at least trace this problem down?

Fedor.