TCP rpc processing is now pushed back to a bh handler. There are
some minor socket locking issues to check through and clean up -
its possible this bh handler needs to cover UDP too and be a
thread. Whats there is at least most of the code needed in either
case
Maybe I'm missing something, but one thing has occurred to me.
We have the rpciod kernel thread, why don't we use it for what it was
designed, to get into a context and do things such as creating and
opening sockets etc. ie. things which we need to do outside of a
plain interrupt context.
If I am right here, the best solution to all these problems is to
queue the "unsafe in bh" operations as yet another task for rpciod to
perform. Things should just work as they are supposed to, without the
races and problems, if this is implemented.
Just some thoughts...
Later,
David S. Miller
davem@dm.cobaltmicro.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html