Re: [lwip-devel] Granularity of core locking

[address@hidden]

Tim Cussins wrote:
> I take it that's just an /example/ of something that needs fixing, and
> that there's a bunch of other things too?

I really can't tell you: Noone has ever tried, I guess. The fact that 
the stack uses global variables to keep track of some things doesn't 
help, either ;-)

> A couple of notable efforts to move away from big locks would be Linux
> BKL and Python GIL. Both were/are non-trivial, but represent big leaps
> forward for scalability. Most LWIP deployments probably aren't
> interested in this, but both multi-core and real-time projects benefit
> from finer-grained locking.

I'd love to run httpd on low priority and the industrial ethernet stack 
at high priority without delay other than the netif output function.
ReactOS might also profit from multi-core improvements.

> If I can find some developer time at this end, would you be open to
> discussion about:
>
> LWIP_TCPIP_FINE_GRAINED_LOCKING
>
> for LWIP users that are happy to trade the memory (and some throughput)
> for better real-time performance?

As long as it doesn't turn the whole stack inside-out, I'd love to see 
your ideas for this!
It would have to be a null-change for the standard use case, of course...

Simon