Hello!
On Mon, 2007-05-07 at 19:04 -0700, Ehud Gavron wrote:
> 2.6.21.1 ifup eth1 panics and is repeatable.
I understand it's Fedora Core 6. /sbin/ifup should be a script. You
can trace it by running it as
bash -x /sbin/ifup eth1
> Unfortunately I can't get netconsole working because the 2.6.21.1 kernel
> doesn't like my hardwired Ethernet card for no good reason. TXcount
> increases, RXcount does not. tcpdump shows nothing. Rebooting in
> 2.6.20-1.2948.fc6 restores full functionality.
That should go to netdev list or to a list specifically about that
driver. But I don't think netconsole is the best tool to debug other
networking drivers.
Please see Documentation/oops-tracing.txt in the Linux sources. If you
have serial console, please use it. If you can use local console,
that's the second best choice. Use "dmesg -n 8" to enable all messages
on the console. Switch to the first virtual console by Ctrl-Alt-F1 if X
Window System is running. To maximize the number of lines, set the
smallest font, e.g.
setfont drdos8x8
Reproduce the panic and write down the messages. Or you can make a
photograph of the screen.
Documentation/oops-tracing.txt also suggests kdump. I haven't tried it,
but I think it would work.
> If I do the following it works flawlessly:
> ifconfig eth1 up
> iwconfig eth1 key wep-key-here
> iwconfig eth1 essid any (or "real-essid-here")
> ifconfig eth1 10.1.1.5 netmask 255.255.255.0
> route add -net 0.0.0.0/0 gw 10.1.1.1
>
>
> If I do the following I get a kernel panic:
> ifup eth1
That's a clear sign that it's caused by something other that bringing
the interface up.
> [EMAIL PROTECTED] ~]# more /etc/sysconfig/network-scripts/ifcfg-eth1
> # Broadcom Corporation BCM4306 802.11b/g Wireless LAN Controller
> TYPE=wireless
> DEVICE=eth1
> ONBOOT=no
> BOOTPROTO=static
> ADDRESS=10.1.1.5
> NETMASK=255.255.255.255
> GATEWAY=10.1.1.1
> ESSID=wetwork
> KEY=xxxa0055ae775e80fd864exxxx
> #RATE=24
> #CHANNEL=6
This looks good.
> While I can't use netconsole, I can use strace. Accordingly attached is
> the output from
> strace ifup eth1 > nothing.txt 2>output.txt
That's not very useful. The interesting stuff is done in other
processes. You could try to follow them by adding "-f" and "-F" flags.
But I think the first priority should be getting the kernel messages.
> If anyone has any ideas what's wrong with the wired ethernet...
> 09:00.0 Ethernet controller: Broadcom Corporation NetXtreme BCM5752
> Gigabit Ethernet PCI Express (rev 02)
Please ask in netdev.
> If I don't get any better suggestions I'll remake oldconfig but answer N
> to all the new stuff and see if that makes a difference. There are a
> couple of new GE drivers, some experimental, and I'm guessing (with no
> basis for the guess) that there may be interference there.
I would not change anything unless all possibilities for tracking down
the bug are exhausted. The bug can just disappear temporarily to appear
later.
> access("/sbin/ip", X_OK) = 0
> access("/sbin/ip", R_OK) = 0
> rt_sigprocmask(SIG_BLOCK, [INT CHLD], [], 8) = 0
> _llseek(255, -1124, [9428], SEEK_CUR) = 0
> clone(child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|
> SIGCHLD, child_tidptr=0xb7f5d918) = 3400
> rt_sigprocmask(SIG_SETMASK, [], NULL, 8) = 0
> rt_sigprocmask(SIG_BLOCK, [CHLD], [], 8) = 0
> rt_sigprocmask(SIG_SETMASK, [], NULL, 8) = 0
> rt_sigprocmask(SIG_BLOCK, [CHLD], [], 8) = 0
> rt_sigaction(SIGINT, {0x807a8c0, [], 0}, {SIG_DFL}, 8) = 0
> waitpid(-1,
Either it was /sbin/ip that triggered the crash, or the end of the log
wasn't written to the hard drive. Generally, the console is designed to
be more survivable than the storage devices in case of a panic, so using
the console is preferred.
--
Regards,
Pavel Roskin
_______________________________________________
Bcm43xx-dev mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/bcm43xx-dev
