xen-netback: refactor guest rx
authorDavid Vrabel <david.vrabel@citrix.com>
Tue, 4 Oct 2016 09:29:14 +0000 (10:29 +0100)
committerDavid S. Miller <davem@davemloft.net>
Fri, 7 Oct 2016 00:37:35 +0000 (20:37 -0400)
commiteb1723a29b9a75dd787510a39096a68dba6cc200
tree7f6c52ddcb5b98accb591c01b94def17d100007e
parentfedbc8c132bcf836358103195d8b6df6c03d9daf
xen-netback: refactor guest rx

Refactor the to-guest (rx) path to:

1. Push responses for completed skbs earlier, reducing latency.

2. Reduce the per-queue memory overhead by greatly reducing the
   maximum number of grant copy ops in each hypercall (from 4352 to
   64).  Each struct xenvif_queue is now only 44 kB instead of 220 kB.

3. Make the code more maintainable.

Signed-off-by: David Vrabel <david.vrabel@citrix.com>
[re-based]
Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/xen-netback/common.h
drivers/net/xen-netback/rx.c