ipvs-0.9.8pre1 (Re: Ipvs 0.9.3 : panic on heavy load : SOLVED)

[Wensong Zhang <wensong@xxxxxxxxxxxx>]

Hello,

On Tue, 4 Dec 2001, Lionel Bringuier wrote:

> On lun, déc 03, 2001 at 07:53:50 +0100, Lionel Bringuier wrote:
> > >   The networking code avoids disabling the hardware IRQs. You
> > > should be aware of interrupting the user space from softirq/bh (on the
> > > same CPU). sltimer_handler is started from the timer bh.
> > OK... I think, we're touching a beginning of a solution. I replaced all the
> > 'write_(un)lock' stuff with 'write_(un)lock_bh'... and it seems to be much
> > better ! I'll let a test run over the night and I tell you.
> Yes, that's it, you were right. It ran perfectly over the night.
>
> Thank you very much Julian !
>
> Gosh, those LVS people are really tough guys :)
>

Thanks a lot for reporting and finding the problem.

I made 0.9.8pre1.diff (http://linux-vs.org/~wensong/ipvs-0.9.8pre1.diff),
which includes adding nat_icmp_send sysctl variable, changing to use
spin_lock_bh to guard data access between network softirqs and bottom half
in the slow timer handling, etc.

Please test it if you have time.

Thanks again,

Wensong