Adjusted the code as suggested, and applied to master~ !
Thx very much for the help and explanation throughout the time,
Alex Wang,
On Mon, Apr 14, 2014 at 10:24 PM, Pravin Shelar <pshe...@nicira.com> wrote:
> On Tue, Apr 15, 2014 at 6:42 AM, Alex Wang <al...@nicira.com> wrote:
> > In order to allow handlers directly read upcalls from datapath,
> > we need to support per-handler netlink socket for each vport in
> > datapath. This commit makes this happen. Also, it is guaranteed
> > backward and forward compatibility with previous branch.
> >
> > Signed-off-by: Alex Wang <al...@nicira.com>
> > Acked-by: Thomas Graf <tg...@redhat.com>
> >
> > ---
> > V7 -> V8:
> > - declare "u32 *ids" as "u32 ids[]" to save space and pointer
> indirection.
> > - use rcu_dereference() when the dereference only happens within the
> > rcu read-side critical section.
> > - fix the comments.
> >
> > V6 -> V7:
> > - rebase.
> >
> > V5 -> V6:
> > - '__force' cast 'struct vport_portids *' in ovs_vport_free().
> >
> > V4 -> V5:
> > - use nla_memcpy() instead of memcpy().
> > - move 'struct rcu_head' back in the 'struct vport_portids'
> > for cacheline efficiency.
> >
> > V3 -> V4:
> > - define the type of OVS_VPORT_ATTR_UPCALL_PID to NLA_U32.
> > - add check of the OVS_VPORT_ATTR_UPCALL_PID in attr array in
> > ovs_vport_cmd_set().
> > - use reciprocal_div() in the modular computation, to reduce the
> > overhead.
> >
> > V2 -> V3:
> > - use OVS_DP_ATTR_USER_FEATURES to control the type of
> > OVS_VPORT_ATTR_UPCALL_PID attribute returned by the
> ovs_vport_cmd_get().
> > - directly free the 'vport->upcall_portids' in ovs_vport_free().
> > - add check to guarantee pids array length is always positive.
> >
> > PATCH -> V2:
> > - rebase.
> >
> > major changes since RFC:
> > - guarantee the compatibility with branch-2.1 userspace.
> > - use rcu to protect upcall port_id array multi-access.
> > - use skb_get_rxhash() to select the port id to send upcall.
> ....
>
> > diff --git a/datapath/vport.h b/datapath/vport.h
> > index 18b723e..09ecf7c 100644
> > --- a/datapath/vport.h
> > +++ b/datapath/vport.h
> > @@ -23,6 +23,7 @@
> > #include <linux/list.h>
> > #include <linux/netlink.h>
> > #include <linux/openvswitch.h>
> > +#include <linux/reciprocal_div.h>
> > #include <linux/skbuff.h>
> > #include <linux/spinlock.h>
> > #include <linux/u64_stats_sync.h>
> > @@ -50,6 +51,11 @@ void ovs_vport_get_stats(struct vport *, struct
> ovs_vport_stats *);
> > int ovs_vport_set_options(struct vport *, struct nlattr *options);
> > int ovs_vport_get_options(const struct vport *, struct sk_buff *);
> >
> > +int ovs_vport_set_upcall_portids(struct vport *, struct nlattr *pids);
> > +int ovs_vport_get_upcall_portids(const struct vport *, struct sk_buff
> *);
> > +
> extra blank line.
> > +u32 ovs_vport_find_portid(const struct vport *, struct sk_buff *);
> > +
> Lets rename to make the consistent.
> s/ovs_vport_find_portid/ovs_vport_find_upcall_portid
>
> > int ovs_vport_send(struct vport *, struct sk_buff *);
> >
> > /* The following definitions are for implementers of vport devices: */
> > @@ -60,13 +66,27 @@ struct vport_err_stats {
> > u64 tx_dropped;
> > u64 tx_errors;
> > };
>
> Looks good to me.
> Acked-by: Pravin B Shelar <pshe...@nicira.com>
>
> Thanks.
>
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
