Hello Pravin,
I added some new comments inline, may you take a look? Please ingore the firest
reply if you haven't looked at it.
And I have sent out patch v5, with the addressed comments except two points I
have explained in comments.
Bests,
Wenyu
-Original Message-
From: Wenyu Zhang
On Sun, Jul 20, 2014 at 12:43:41AM -0700, Alex Wang wrote:
> From: Ryan Wilson
>
> Since the use of xcache, the netdev struct can be freed by the
> revalidator threads. This fact also makes the following race possible:
>
> 1. Consider there is a gre tunnel, and datapath flows that go through
>
Thx, backported to branch-2.3 with suggested changes,
On Mon, Jul 21, 2014 at 8:57 AM, Ben Pfaff wrote:
> On Sun, Jul 20, 2014 at 12:43:41AM -0700, Alex Wang wrote:
> > From: Ryan Wilson
> >
> > Since the use of xcache, the netdev struct can be freed by the
> > revalidator threads. This fact
On Sun, Jul 20, 2014 at 7:30 PM, Joe Stringer wrote:
> Converting the flow key and mask back into netlink format during each
> flow dump is fairly expensive. By caching the netlink versions of these
> the first time a flow is dumped, and copying the memory directly during
> subsequent dumps, we ar
On Wed, Jul 16, 2014 at 11:42 PM, Andy Zhou wrote:
> Remove get_dp() API, it becomes get_dp__() which is for internal
> use only. Added the following APIs to make it clear of the locking
> requirement.
>
> o get_dp_rcu() requires its caller to hold rcu lock.
> o get_dp_ovsl() requires i
An error from nl_sock_recv() could mean that there issues with the netlink
socket (EBADF, ENOTSOCK, ...). Keeping calling nl_sock_recv() in this case is
harmful: nln_run() will never return and, since we are calling it from the main
thread, vswitchd will become unresponsive.
Suggested-by: Alex Wan
Bonjour
Je suis tombée sur votre contact E-mail suite à une recherche personnelle alors
j'ai décidé de vous écrire.
Je
suis madame BUSHMANN ENRIKA , ex- administrateur d'une multinationale
du diamant à la retraite. Allemande de nationalité je suis hospitalisée
en Afrique pour raison de santé
On Mon, Jul 21, 2014 at 11:13:28AM -0700, Daniele Di Proietto wrote:
> An error from nl_sock_recv() could mean that there issues with the netlink
> socket (EBADF, ENOTSOCK, ...). Keeping calling nl_sock_recv() in this case is
> harmful: nln_run() will never return and, since we are calling it from
o get_dp_rcu() requires its caller to hold rcu lock.
o get_dp() requires its caller to hold either ovs lock or
rcu lock.
Signed-off-by: Andy Zhou
---
v1->v2: Reduce APIs from 4 to 2.
---
datapath/datapath.c | 27 +++
1 file changed,
O.K. I have send out a V2 that reduced the number of APIs from 4 to 2.
get_dp() as before, and get_dp_rcu(). V2 is a simpler change.
On Mon, Jul 21, 2014 at 11:04 AM, Pravin Shelar wrote:
> On Wed, Jul 16, 2014 at 11:42 PM, Andy Zhou wrote:
>> Remove get_dp() API, it becomes get_dp__() which is
On Fri, Jul 18, 2014 at 5:35 PM, Saurabh Shah wrote:
>
> From: Ben Pfaff
> Date: Wednesday, June 25, 2014 at 2:57 PM
> To: Saurabh Shah
> Cc: "dev@openvswitch.org" , Guolin Yang
>
> Subject: Re: [ovs-dev] [PATCH 1/4] datapath-windows: Base code for
> developing the Hyper-V switch entension.
>
>
This patch fixes following kernel crash that could happen, if geneve
vport was not added yet, but revalidator thread attempted to dump flows.
To reproduce:
1. switch tunnel type between geneve and gre; and
2. run ping.
BUG: unable to handle kernel NULL pointer dereference at 0048
IP: [
On Fri, Jul 18, 2014 at 5:27 PM, Saurabh Shah wrote:
> Ben Pfaff writes:
> > The logging in dpif_windows_flow_del() seems like a debugging stray.
>
> Do you mean in do_put?
I mean the one here:
+static int
+dpif_windows_flow_del(struct dpif *dpif, const struct dpif_flow_del *del)
+{
+VLOG_IN
On Thu, Jul 17, 2014 at 4:08 PM, Jesse Gross wrote:
> On Thu, Jul 17, 2014 at 3:46 PM, Ansis Atteka wrote:
>> diff --git a/datapath/flow_netlink.c b/datapath/flow_netlink.c
>> index 5f975a1..a4108c0 100644
>> --- a/datapath/flow_netlink.c
>> +++ b/datapath/flow_netlink.c
>> @@ -1034,7 +1034,7 @@
I think we could take this further by getting rid of the flow put and
execute, and having exec_upcalls instead just return the necessary
actions and mask. At any rate, this is a good staring point.
In exec_upcalls, we assert that cnt is less that UPCALL_MAX_BATCH
which seems wrong. It's much les
On Fri, Jul 18, 2014 at 01:49:52PM -0700, Alex Wang wrote:
> This commit adds more logging points in stp module for debugging.
> Also, it makes the log print out the port name.
>
> Signed-off-by: Alex Wang
This is not safe against simultaneous calls for the same port, but it
would be if the assi
↗℡ⅱぃ①Ⅵò∝ぜ▲￥◥ξさば
服 务
销 售
设 备 禾兑
王 生 Q扣: 1 2 1 1 6 0 1 6 5 1 電:1 8 3 2 0 7 9 3 1 6
5
す①※ね—︷◆￣┣╅づΨ*:み
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
Thanks for the review! Pushed to master,
Jarno
On Jul 18, 2014, at 10:40 PM, Ben Pfaff wrote:
> On Fri, Jul 18, 2014 at 09:05:48PM -0700, Jarno Rajahalme wrote:
>> As ovsrcu_get() looks like a function call, it is reasonable for the
>> callers to expect that the arguments are evaluated only o
Thanks, pushed,
Jarno
On Jul 18, 2014, at 11:36 PM, Ben Pfaff wrote:
> On Fri, Jul 18, 2014 at 09:05:50PM -0700, Jarno Rajahalme wrote:
>> ovs-atomic-gcc4+ did expand arguments again, if locked set/get was
>> called.
>>
>> Also fix atomic_is_lock_free().
>>
>> Signed-off-by: Jarno Rajahalme
Pushed with the proposed change, thanks for the review!
Jarno
On Jul 18, 2014, at 11:49 PM, Ben Pfaff wrote:
> On Fri, Jul 18, 2014 at 09:05:51PM -0700, Jarno Rajahalme wrote:
>> The check for the need of default values was in the wrong place,
>> causing no prefix tracking to be used when dat
As you noted earlier, CMAP_FOR_EACH and CMAP_FOR_EACH_SAFE now are basically
the same thing, the only difference being on that safe does the advance before
calling the body. You considered merging them before?
Acked-by: Jarno Rajahalme
On Jul 19, 2014, at 10:28 AM, Ben Pfaff wrote:
> CMAP_FO
Acked-by: Jarno Rajahalme
On Jul 19, 2014, at 10:28 AM, Ben Pfaff wrote:
> Signed-off-by: Ben Pfaff
> ---
> lib/classifier.c | 49 +
> 1 file changed, 21 insertions(+), 28 deletions(-)
>
> diff --git a/lib/classifier.c b/lib/classifier.c
> index
Thanks,
After an offline discussion with Alex we decided that it might be worth to go
one
step further and avoid calling the notification callback in case of error.
This (for example) would prevent lib/route-table.c from believing that the
routing
table has changed on every iteration
I’m about
An error from nl_sock_recv() could mean that there are some issues with the
netlink socket (EBADF, ENOTSOCK, ...). Calling nl_sock_recv() in this case is
harmful: nln_run() will never return and since we are calling it from the main
thread, vswitchd becomes unresponsive.
Also, with this commit we a
v2 seems fine to me too.
I'll leave it to Alex to apply this in case he has further comments.
On Mon, Jul 21, 2014 at 03:48:14PM -0700, Daniele Di Proietto wrote:
> Thanks,
>
> After an offline discussion with Alex we decided that it might be worth to go
> one
> step further and avoid calling t
Reported-by: Madhu Venugopal
Signed-off-by: Ben Pfaff
---
ovsdb/ovsdb-server.1.in | 16
1 file changed, 16 insertions(+)
diff --git a/ovsdb/ovsdb-server.1.in b/ovsdb/ovsdb-server.1.in
index 15382b4..ec408a6 100644
--- a/ovsdb/ovsdb-server.1.in
+++ b/ovsdb/ovsdb-server.1.in
@@
Signed-off-by: Daniele Di Proietto
---
I've noticed a testcase (ofproto-dpif - controller, 758) failure on master.
I believe that the problem is in the testcase itself, in which case this commit
fixes it.
Can someone, please, confirm that the issue is in the testcase and that the fix
makes sense?
Thanks for the review, my comments inline.
Bests,
Wenyu
-Original Message-
From: Pravin Shelar [mailto:pshe...@nicira.com]
Sent: Tuesday, July 22, 2014 6:56 AM
To: Wenyu Zhang
Cc: Romain Lenglet; Ben Pfaff; Jesse Gross; dev@openvswitch.org
Subject: Re: [PATCH v5] Extend OVS IPFIX exporter
Thanks for the reviews. I applied these to master.
On Mon, Jul 21, 2014 at 03:29:35PM -0700, Jarno Rajahalme wrote:
> Acked-by: Jarno Rajahalme
>
> On Jul 19, 2014, at 10:28 AM, Ben Pfaff wrote:
>
> > Signed-off-by: Ben Pfaff
> > ---
> > lib/classifier.c | 49 +---
There isn't any significant downside to making cmap iteration "safe" all
the time, so this drops the _SAFE variant.
Similar changes to CMAP_CURSOR_FOR_EACH and CMAP_CURSOR_FOR_EACH_CONTINUE.
Signed-off-by: Ben Pfaff
---
lib/classifier.c | 4 ++--
lib/cmap.h| 57 ---
On Mon, Jul 21, 2014 at 03:25:58PM -0700, Jarno Rajahalme wrote:
> As you noted earlier, CMAP_FOR_EACH and CMAP_FOR_EACH_SAFE now are
> basically the same thing, the only difference being on that safe does
> the advance before calling the body. You considered merging them
> before?
Sure. Here's a
On Mon, Jul 21, 2014 at 05:11:49PM -0700, Daniele Di Proietto wrote:
> Signed-off-by: Daniele Di Proietto
> ---
> I've noticed a testcase (ofproto-dpif - controller, 758) failure on master.
> I believe that the problem is in the testcase itself, in which case this
> commit
> fixes it.
> Can someo
32 matches
Mail list logo
