Re: [ovs-dev] [PATCH 40/41] ofproto-dpif-xlate: Put recirc_state, not recirc_id_node, in xlate_in.

I think I'd be more comfortable letting you figure that out, if that's OK, and just stick to the compromise we have here for the moment. On Wed, Jan 20, 2016 at 05:46:31PM -0800, Jarno Rajahalme wrote: > I did not look at the details yet, but basically the reference would be > released if gotten

Re: [ovs-dev] [PATCH 40/41] ofproto-dpif-xlate: Put recirc_state, not recirc_id_node, in xlate_in.

Thanks. I applied this to master. On Thu, Jan 21, 2016 at 01:09:29PM -0800, Jarno Rajahalme wrote: > Sounds like a plan, > > Jarno > > > On Jan 21, 2016, at 8:42 AM, Ben Pfaff wrote: > > > > I think I'd be more comfortable letting you figure that out, if that's > > OK, and

Re: [ovs-dev] [PATCH 40/41] ofproto-dpif-xlate: Put recirc_state, not recirc_id_node, in xlate_in.

Sounds like a plan, Jarno > On Jan 21, 2016, at 8:42 AM, Ben Pfaff wrote: > > I think I'd be more comfortable letting you figure that out, if that's > OK, and just stick to the compromise we have here for the moment. > > On Wed, Jan 20, 2016 at 05:46:31PM -0800, Jarno

Re: [ovs-dev] [PATCH 40/41] ofproto-dpif-xlate: Put recirc_state, not recirc_id_node, in xlate_in.

I did not look at the details yet, but basically the reference would be released if gotten initially and not taken for the flow install. Jarno > On Jan 20, 2016, at 4:59 PM, Ben Pfaff wrote: > > I think I like the idea of only needing a recirc_state in that code, but > I don't

Re: [ovs-dev] [PATCH 40/41] ofproto-dpif-xlate: Put recirc_state, not recirc_id_node, in xlate_in.

I think I like the idea of only needing a recirc_state in that code, but I don't think I fully understand yet. I guess it's pretty easy to make xlate_in_init() take the reference and pass back the recirc_state instead of the recirc_id_node, but where would the eventual release of the reference

Re: [ovs-dev] [PATCH 40/41] ofproto-dpif-xlate: Put recirc_state, not recirc_id_node, in xlate_in.

Thanks, I folded it in. On Wed, Jan 20, 2016 at 04:41:24PM -0800, Jarno Rajahalme wrote: > I was a bit too fast, I had to add this incremental to make this patch > compile: > > diff --git a/ofproto/ofproto-dpif-xlate.c b/ofproto/ofproto-dpif-xlate.c > index 184eb46..6549191 100644 > ---

Re: [ovs-dev] [PATCH 40/41] ofproto-dpif-xlate: Put recirc_state, not recirc_id_node, in xlate_in.

Maybe ofproto-dpif-upcall.c could be refactored to not need the recirc_id_node pointer. It is only used for getting the recirc ID and managing recirc_id_node reference counting. If the reference counting could be moved upstream, the pointer could be removed. But the idea was to take a reference

Re: [ovs-dev] [PATCH 40/41] ofproto-dpif-xlate: Put recirc_state, not recirc_id_node, in xlate_in.

I was a bit too fast, I had to add this incremental to make this patch compile: diff --git a/ofproto/ofproto-dpif-xlate.c b/ofproto/ofproto-dpif-xlate.c index 184eb46..6549191 100644 --- a/ofproto/ofproto-dpif-xlate.c +++ b/ofproto/ofproto-dpif-xlate.c @@ -5143,7 +5143,7 @@ xlate_actions(struct

[ovs-dev] [PATCH 40/41] ofproto-dpif-xlate: Put recirc_state, not recirc_id_node, in xlate_in.

This will make it possible, in an upcoming commit, to construct a recirc_state locally on the stack to pass to xlate_actions(). It would also be possible to construct and pass a recirc_id_node on the stack, but the translation process only uses the recirc_state anyway. The alternative here of