> I try to summarize where each lock is used so we can decide if it is ok to
> revert some of them to spin_lock again. This should also resolve a request I
> got from Andrew some time ago.
Hi Sven
Thanks for this.
[...]
> So it is complete correct to use irqsave everywhere (as you have commited
On Mon, Dec 28, 2009 at 04:10:08PM +0100, Simon Wunderlich wrote:
> This patch removes the (ugly and racy) packet receiving thread and the
> kernel socket usage. Instead, packets are received directly by registering
> the ethernet type and handling skbs instead of self-allocated buffers.
>
> Some
Hello,
i've just commited the patch in revision 1517 with some more
adjustments: Spinlocks have been converted to irq saving versions
to avoid recursive lifelocks. I've converted all of them, maybe
we can convert back some of them later. Having them irq saved should
only hurt performance a little.
This patch removes the (ugly and racy) packet receiving thread and the
kernel socket usage. Instead, packets are received directly by registering
the ethernet type and handling skbs instead of self-allocated buffers.
Some consequences and comments:
* we don't copy the payload data when forwarding
