Signed-off-by: Ben Pfaff
---
lib/dpif-netdev.c | 150 --
1 file changed, 43 insertions(+), 107 deletions(-)
diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c
index e97e4ce..efff796 100644
--- a/lib/dpif-netdev.c
+++ b/lib/dpif-netdev.c
@@ -49,
This allows clients to do more than just increment a counter. The
following commit will make the first use of that feature.
Signed-off-by: Ben Pfaff
---
lib/dpif-netdev.c | 75 +++--
lib/ovs-thread.c | 91 +++--
RCU allows multiple threads to read objects in parallel without any
performance penalty. The following commit will introduce the first use.
Signed-off-by: Ben Pfaff
---
lib/automake.mk | 2 +
lib/ovs-rcu.c | 293 ++
lib/ovs
This factors code out of fat-rwlock, making it easily usable by other code.
Signed-off-by: Ben Pfaff
---
lib/fat-rwlock.c | 29 ++---
lib/util.c | 65 +++-
lib/util.h | 6 +-
3 files changed, 71 insertions(+
This should scale better than a single mutex, though still not
ideally.
Signed-off-by: Ben Pfaff
---
lib/dpif-netdev.c | 99 +--
1 file changed, 74 insertions(+), 25 deletions(-)
diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c
index cb1beb7
The upcoming RCU library will use these.
Signed-off-by: Ben Pfaff
---
lib/ovs-atomic-clang.h| 4 +++-
lib/ovs-atomic-gcc4+.h| 2 ++
lib/ovs-atomic-gcc4.7+.h | 4 +++-
lib/ovs-atomic-pthreads.h | 1 +
lib/ovs-atomic.h | 23 +++
5 files changed, 32 inserti
This makes it possible to add new uses of the underlying types for
these declarations without the "typedef" wrappers. The following commit
will make use of that feature.
Signed-off-by: Ben Pfaff
---
lib/ovs-atomic-gcc4+.h| 56 +--
lib/ovs-atomic-pthre
Signed-off-by: Ben Pfaff
---
ofproto/connmgr.c| 5 +-
ofproto/ofproto-dpif-xlate.c | 2 -
ofproto/ofproto-dpif.c | 2 -
ofproto/ofproto-provider.h | 19 ---
ofproto/ofproto.c| 120 ++-
5 files changed, 63 insertion
The first three patches in this series implement an RCU library
within Open vSwitch. The library is self-contained; it does not
depend on liburcu or another external library. (This is the
primary difference between this series and the previous RFC
version.)
The fourth patch adds one simple user
On Wed, Mar 5, 2014 at 10:43 AM, Ben Pfaff wrote:
> On Thu, Feb 27, 2014 at 11:44:24AM -0800, Alex Wang 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. Als
Hey Ben,
> It would be nice to add a sentence or two to the commit message
> explaining why we think it's OK to drop the distinction between the
> two queues we had before (as already discussed).
>
Yes, I'll do that,
> The choice of hash function, in flow_hash_5tuple(), seems curious. I
>
This patch adds a new configuration option, "max-idle" to the
Open_vSwitch "other-config" column. This sets how long datapath flows
are cached in the datapath before revalidators expire them.
Signed-off-by: Joe Stringer
---
v6: Shift documentation to ofproto-provider.h.
v5: Rebase.
v4: Remove ex
This commit changes the API in 'dpif-provider.h' to allow multiple
handler threads call dpif_recv() simultaneously.
Signed-off-by: Alex Wang
---
PATCH -> V2:
- fold in Ben's clarification.
- refine the comments.
- invoke dpif_handlers_set() in udpif_set_threads(). this is a bug.
the previous
From: Ethan Jackson
Previously, we had a separate flow_dumper thread that fetched flows from
the datapath to distribute to revalidator threads. This patch takes the
logic for dumping and pushes it into the revalidator threads, resulting
in simpler code with similar performance to the current code
On Wed, Mar 05, 2014 at 04:09:38PM -0800, Alex Wang wrote:
> For some reason, the context didn't show up. But I think I know where you
> are talking
> about. lib/dpif.c
>
> Please let me know if there is need to modify and refine. I'd like to go
> over and over
> to make the API and comments com
Thanks Ben for the review~,
For some reason, the context didn't show up. But I think I know where you
are talking
about. lib/dpif.c
Please let me know if there is need to modify and refine. I'd like to go
over and over
to make the API and comments comprehensible. ;D
The comment on dpif_port_g
On 5 March 2014 11:08, Ben Pfaff wrote:
> That's fine with me.
>
OK, I'll send a fresh revision.
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
On Wed, Mar 05, 2014 at 11:03:24AM -0800, Joe Stringer wrote:
> On 5 March 2014 10:50, Ben Pfaff wrote:
>
> > On Wed, Mar 05, 2014 at 10:47:33AM -0800, Joe Stringer wrote:
> > > On 4 March 2014 14:49, Ben Pfaff wrote:
> > > > If we were to maintain the ukeys hmaps separately, then we wouldn't
>
On 5 March 2014 10:50, Ben Pfaff wrote:
> On Wed, Mar 05, 2014 at 10:47:33AM -0800, Joe Stringer wrote:
> > On 4 March 2014 14:49, Ben Pfaff wrote:
> > > If we were to maintain the ukeys hmaps separately, then we wouldn't
> > > have to have the same number of hmaps as threads. That seems,
> > >
On Wed, Mar 05, 2014 at 10:47:33AM -0800, Joe Stringer wrote:
> On 4 March 2014 14:49, Ben Pfaff wrote:
>
> > After studying the code for a while, I think I see something odd.
> > Each 'ukeys' hmap is associated with a revalidator thread, but the
> > hmaps have barely any relationship with their
On 4 March 2014 14:49, Ben Pfaff wrote:
> After studying the code for a while, I think I see something odd.
> Each 'ukeys' hmap is associated with a revalidator thread, but the
> hmaps have barely any relationship with their owning thread. Except
> during the "sweep", it's not more likely, for e
On Thu, Feb 27, 2014 at 11:44:25AM -0800, Alex Wang wrote:
> Signed-off-by: Alex Wang
This one has a few rejects against master. Since it's a pretty
sophisticated patch, I'd prefer to wait until it can be rebased to
take a look.
Thanks,
Ben.
___
dev
On Thu, Feb 27, 2014 at 11:44:24AM -0800, Alex Wang 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
> that the newer datapath is compatible
On Thu, Feb 27, 2014 at 11:44:23AM -0800, Alex Wang wrote:
> Signed-off-by: Alex Wang
It would be nice to add a sentence or two to the commit message
explaining why we think it's OK to drop the distinction between the
two queues we had before (as already discussed).
The choice of hash function,
On Thu, Feb 27, 2014 at 11:44:22AM -0800, Alex Wang wrote:
> This commit changes the API in 'dpif-provider.h' to allow multiple
> handler threads call dpif_recv() simultaneously.
>
> Signed-off-by: Alex Wang
The comment on dpif_port_get_pid() says that the 'hash' parameter
should be a 5-tuple ha
On Tue, Mar 04, 2014 at 09:32:52AM -0800, Gurucharan Shetty wrote:
> QueryPerformanceCounter() retrieves the current value of the performance
> counter, which is a high resolution (<1us) time stamp that can be used for
> time-interval measurements. So, use it for MONOTONIC clock.
>
> The GetSystem
26 matches
Mail list logo
