System lockup

Hello,

I am using linux channel bonding for 2.4.19 kernel and trying to
bond 4 interfaces. The system has 4 processors. Now I have an
application that tries to take an interface out of bonding at times.
Without this change, everything seems to be working fine. But with
this changes, the system locks up and nothing works. I tried to catch
it using the NMI watchdog timer and the KDB shows the following info:

eip = 0x80107c68 ebp = 0xf89687f0

and the backtrace shows like this:

__read_lock_failed +0x08(0x1d2, 0x202, ..., ..., ...)

bond_xmit_roundrobin +0x62(..., ..., ..., ...)

ip_output +0x104 (..., ..., ..., 0x0, 0x0)

ip_queue_xmit +0x37b (..., 0x1, ..., 0x2bd, 0x0)


When I tried to go thouugh the objdump of the bonding driver, it
looked like it was pointing to the locks in bond_xmit_roundrobin().
Could there be any possible issue with the bonding driver and the
system being SMP?? Any ideas and suggestions would be appreciated.

Thanks,
Shanthi.

Shanthi Paladugu <(E-Mail Removed)> wrote:
> Hello,

> I am using linux channel bonding for 2.4.19 kernel and trying to
> bond 4 interfaces. The system has 4 processors. Now I have an
> application that tries to take an interface out of bonding at times.

What app?

> Without this change, everything seems to be working fine. But with
> this changes, the system locks up and nothing works. I tried to catch
> it using the NMI watchdog timer and the KDB shows the following info:

[..]

> When I tried to go thouugh the objdump of the bonding driver, it
> looked like it was pointing to the locks in bond_xmit_roundrobin().
> Could there be any possible issue with the bonding driver and the
> system being SMP?? Any ideas and suggestions would be appreciated.

There have been some changes with the bonding driver in recent
kernels (AFAIK), I'd take a look at the CHANGELOG and upgrade.

--
Michael Heiming

Remove +SIGNS and www. if you expect an answer, sorry for
inconvenience, but I get tons of SPAM

Shanthi Paladugu <(E-Mail Removed)> wrote:
> Hello,

> I am using linux channel bonding for 2.4.19 kernel and trying to
> bond 4 interfaces. The system has 4 processors. Now I have an
> application that tries to take an interface out of bonding at times.

What app?

> Without this change, everything seems to be working fine. But with
> this changes, the system locks up and nothing works. I tried to catch
> it using the NMI watchdog timer and the KDB shows the following info:

[..]

> When I tried to go thouugh the objdump of the bonding driver, it
> looked like it was pointing to the locks in bond_xmit_roundrobin().
> Could there be any possible issue with the bonding driver and the
> system being SMP?? Any ideas and suggestions would be appreciated.

There have been some changes with the bonding driver in recent
kernels (AFAIK), I'd take a look at the CHANGELOG and upgrade if
you need to run a vendor kernel, check there for updates.


--
Michael Heiming

Remove +SIGNS and www. if you expect an answer, sorry for
inconvenience, but I get tons of SPAM

Michael Heiming <michael+(E-Mail Removed)> wrote in message news:<(E-Mail Removed)>...
> Shanthi Paladugu <(E-Mail Removed)> wrote:
> > Hello,
>
> > I am using linux channel bonding for 2.4.19 kernel and trying to
> > bond 4 interfaces. The system has 4 processors. Now I have an
> > application that tries to take an interface out of bonding at times.
>
> What app?
>

This is a C program that I wrote and issues IFBONDRELEASE ioctl whenever needed.

> > Without this change, everything seems to be working fine. But with
> > this changes, the system locks up and nothing works. I tried to catch
> > it using the NMI watchdog timer and the KDB shows the following info:
>
> [..]
>
> > When I tried to go thouugh the objdump of the bonding driver, it
> > looked like it was pointing to the locks in bond_xmit_roundrobin().
> > Could there be any possible issue with the bonding driver and the
> > system being SMP?? Any ideas and suggestions would be appreciated.
>
> There have been some changes with the bonding driver in recent
> kernels (AFAIK), I'd take a look at the CHANGELOG and upgrade if
> you need to run a vendor kernel, check there for updates.