Re: more tcpdumpinfo for nfs3 problem: aix-server --- linux 2.4.15pre5 client

Re: more tcpdumpinfo for nfs3 problem: aix-server --- linux 2.4.15pre5 client

[Manfred Spraul]

 
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

> IOW:
>     Either we must demand that CPU 2 uses irq-safe spinlocks in order to 
> protect against sk->write_space(), or we must demand that CPU 1 should drop 
> 'lock1' before being allowed to call dev_kfree_skb_any().

Or dev_kfree_skb_any() should consider disabled local interrupts as
'in_irq()' and call dev_kfree_skb_irq() in this case, or the driver
could call dev_kfree_irq() if it really wants to free while holding an
irq spinlock.

But that's a known problem:
http://groups.google.com/groups?q=dev_kfree_skb_any&hl=en&rnum=1&selm=linux.net.20010905.184245.94554736.davem%40redhat.com

--
	Manfred

Re: more tcpdumpinfo for nfs3 problem: aix-server --- linux 2.4.15pre5

[kuznet]

Hello!

> 				should consider disabled local interrupts as
> 'in_irq()'

How to do this?

Alexey

Re: more tcpdumpinfo for nfs3 problem: aix-server --- linux 2.4.15pre5

[Manfred Spraul]

kuznet@ms2.inr.ac.ru wrote:
> 
> Hello!
> 
> >                               should consider disabled local interrupts as
> > 'in_irq()'
> 
> How to do this?
>
Arch specific, like __global_cli() in arch/i386/kernel/irq.c.
What about a rate-limited warning if dev_kfree_skb_any is called with
disabled interrupts?

--
	Manfred