[ewg] Re: Continue of "defer skb_orphan() until irqs enabled"

[akepner at sgi.com]

On Wed, Sep 24, 2008 at 07:01:02PM +0300, Moni Shoua wrote:
> Hi,
> On Sep 9 Arthur started a thread with a patch that fixes an issue with a call to skb_orphan() when irqs are disabled.
> Please look at http://lists.openfabrics.org/pipermail/general/2008-September/054053.html
> 
> Although there is an agreement that the fix isn't yet complete I think that it is important enough to include it in
> OFED **until** we have a full fix for it in upstream kernel because it is still a big improvement.
> 
> I want to suggest a small addition to Arthur's original patch that allows to disable call to skb_orphan() totally.
> I hope it's OK with you Arthur.
> 
> 
> Please comment.
> 

Thanks for reviving this discussion, Moni.

Another possibility is to simply remove the call to skb_orphan() 
from ipoib_send(). Eventually the skb->destructor method will be 
called in net_tx_action() (in bottom half context).

But I'm not sure of all the performance implications of that 
change.

-- 
Arthur