Signed-off-by: Jarno Rajahalme
---
ODP ports are now 32-bit, so OVSP_LOCAL should be too.
(Internally, kernel module still keeps port numbers 16-bit, though.)
---
include/linux/openvswitch.h |2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/include/linux/openvswitch.h b/incl
On Sat, Jan 05, 2013 at 11:25:57AM -0800, Justin Pettit wrote:
> Userspace-kernel communication is a possible bottleneck when OVS is
> receiving a large number of flow set up requests. To help prevent a bad
> actor from consuming too much of this resource, we introduced channels
> to segegrate tra
On Jan 7, 2013, at 3:11 AM, Isaku Yamahata wrote:
>> +if (++dpif->event_offset >= dpif->n_events) {
>> +dpif->event_offset = dpif->n_events = 0;
>> +}
>
> dpfi->event_offset isn't initialized as zero.
Perhaps I missed a spot, but I didn't see any issues. The dpif st
On Sat, Jan 05, 2013 at 11:25:55AM -0800, Justin Pettit wrote:
> The compose_slow_path() function used UINT16_MAX for
> dpif_port_get_pid(), when it should be UINT32_MAX to get the proper
> reserved PID.
>
> Signed-off-by: Justin Pettit
Looks good, thanks.
___
On Sat, Jan 05, 2013 at 11:25:56AM -0800, Justin Pettit wrote:
> When adding a port, the code previously logged the requested port number
> (which is generally UINT32_MAX) instead of the assigned port number.
>
> Signed-off-by: Justin Pettit
Looks good, thanks.
__
On Mon, Jan 7, 2013 at 2:45 AM, Jarno Rajahalme wrote:
>
> Signed-off-by: Jarno Rajahalme
> ---
> ODP ports are now 32-bit, so OVSP_LOCAL should be too.
> (Internally, kernel module still keeps port numbers 16-bit, though.)
Justin, I see two direct uses of OVSP_LOCAL in the ofproto directory.
I
On Mon, Jan 7, 2013 at 2:45 AM, Jarno Rajahalme wrote:
>
> Signed-off-by: Jarno Rajahalme
> ---
> ODP ports are now 32-bit, so OVSP_LOCAL should be too.
> (Internally, kernel module still keeps port numbers 16-bit, though.)
Applied, thank you. I included your comments about the ODP ports in
the
On Jan 7, 2013, at 9:50 AM, Jesse Gross wrote:
> On Mon, Jan 7, 2013 at 2:45 AM, Jarno Rajahalme
> wrote:
>>
>> Signed-off-by: Jarno Rajahalme
>> ---
>> ODP ports are now 32-bit, so OVSP_LOCAL should be too.
>> (Internally, kernel module still keeps port numbers 16-bit, though.)
>
> Justin,
On Sat, Jan 05, 2013 at 11:25:57AM -0800, Justin Pettit wrote:
> +- The Linux datapath implementation creates a different kernel-
> + userspace channel for each port instead of sharing a static 16
> + channels to provide better performance isolation.
I think that's going to prompt mo
On Sat, Jan 05, 2013 at 09:42:16PM +, Zoltan Kiss wrote:
> The hash entry tag connects to facet(s), not slaves.
>
> Signed-off-by: Zoltan Kiss
> ---
> lib/bond.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/lib/bond.c b/lib/bond.c
> index a95c9b0..2446d7e 100644
Signed-off-by: Ben Pfaff
CC: Zoltan Kiss
---
lib/bond.c |9 +
1 files changed, 5 insertions(+), 4 deletions(-)
diff --git a/lib/bond.c b/lib/bond.c
index 2c59f9d..462ebd8 100644
--- a/lib/bond.c
+++ b/lib/bond.c
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2008, 2009, 2010, 2011, 2012 Nici
On Sat, Jan 05, 2013 at 09:42:02PM +, Zoltan Kiss wrote:
> During rebalancing there could be hash entries with 0 bytes traffic
> on them. It is unnecessary to move them.
>
> Signed-off-by: Zoltan Kiss
Can this ever trigger? I don't see how these hash entries would end
up on the list iterate
Seems fine to me. I might change 'Pick' to 'Picks' in the comment.
Acked-by: Ethan Jackson
On Mon, Jan 7, 2013 at 11:34 AM, Ben Pfaff wrote:
> Signed-off-by: Ben Pfaff
> CC: Zoltan Kiss
> ---
> lib/bond.c |9 +
> 1 files changed, 5 insertions(+), 4 deletions(-)
>
> diff --git
Thanks, I made that fix and applied this to master.
On Mon, Jan 07, 2013 at 12:20:56PM -0800, Ethan Jackson wrote:
> Seems fine to me. I might change 'Pick' to 'Picks' in the comment.
>
> Acked-by: Ethan Jackson
>
>
> On Mon, Jan 7, 2013 at 11:34 AM, Ben Pfaff wrote:
>
> > Signed-off-by: Be
Ethan, do you have any thoughts about this patch? It seems
reasonable, but I'd like another set of eyes on it before I apply it.
On Sat, Jan 05, 2013 at 09:41:19PM +, Zoltan Kiss wrote:
> The old algorithm tries to converge to 0, despite it would mean a very
> unbalanced situation. 1 is a mor
Signed-off-by: Ben Pfaff
---
ovsdb/ovsdb-doc.in |2 ++
1 files changed, 2 insertions(+), 0 deletions(-)
diff --git a/ovsdb/ovsdb-doc.in b/ovsdb/ovsdb-doc.in
index aa4fae2..b328326 100755
--- a/ovsdb/ovsdb-doc.in
+++ b/ovsdb/ovsdb-doc.in
@@ -212,6 +212,8 @@ def columnGroupToNroff(table, group
Oops, I've missed the "#include " from this second version of
the patch. To avoid warnings, I will resend it.
Zoli
On 07/01/13 20:27, Ben Pfaff wrote:
Ethan, do you have any thoughts about this patch? It seems
reasonable, but I'd like another set of eyes on it before I apply it.
On Sat, Jan
The old algorithm tries to converge to 0, despite it would mean a very
unbalanced situation. 1 is a more desirable goal.
Signed-off-by: Zoltan Kiss
---
lib/bond.c | 6 --
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/lib/bond.c b/lib/bond.c
index 2c59f9d..2f8afdd 100644
--- a
OK, thanks. I'd still like some comments from Ethan or another person
who understands the bonding code.
On Mon, Jan 07, 2013 at 10:03:00PM +, Zoltan Kiss wrote:
> Oops, I've missed the "#include " from this second version
> of the patch. To avoid warnings, I will resend it.
>
> Zoli
>
> On
This seems fine to me. My only comment is that we should probably add a
bit more explanation. It took me 20 minutes just to figure out what this
code is intended to do.
> +/* Ideal ratio is 1, move it if the new ratio is closer to
> it. */
>
Perhaps we could expand this comment a b
Hi,
Originally I've intended to avoid such kind of messages:
Jan 3 17:18:13 localhost ovs-vswitchd: 00095|bond|INFO|bond bond0:
shift 0kB of load (with hash 79) from eth3 to eth1 (now carrying
1153733kB and 0kB load, respectively)
But now I see the log shows tx_bytes / 1024, which could be zer
On 07/01/13 22:12, Ethan Jackson wrote:
This seems fine to me. My only comment is that we should probably add a
bit more explanation. It took me 20 minutes just to figure out what
this code is intended to do.
> +/* Ideal ratio is 1, move it if the new ratio is
closer to it
On Mon, Jan 07, 2013 at 10:23:25PM +, Zoltan Kiss wrote:
> On 07/01/13 22:12, Ethan Jackson wrote:
> >This seems fine to me. My only comment is that we should probably add a
> >bit more explanation. It took me 20 minutes just to figure out what
> >this code is intended to do.
> >
> > > +
Acked-by: Ethan Jackson
On Fri, Jan 4, 2013 at 1:48 PM, Ben Pfaff wrote:
> Bug #14357.
> Reported-by: Luca Giraudo
> Signed-off-by: Ben Pfaff
> ---
> This leak has been around forever. I expect to backport it all the
> way to 1.4.
>
> lib/rconn.c | 11 +--
> 1 files changed, 5 in
On 07/01/13 19:33, Ben Pfaff wrote:
On Sat, Jan 05, 2013 at 09:42:16PM +, Zoltan Kiss wrote:
The hash entry tag connects to facet(s), not slaves.
Signed-off-by: Zoltan Kiss
---
lib/bond.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/lib/bond.c b/lib/bond.c
index a
The old algorithm tries to converge to 0, despite it would mean a very
unbalanced situation. We're aiming for an ideal ratio of 1, meaning both
the 'from' and 'to' slave have the same load. Therefore, we only move an
entry if it decreases the load on from, and brings us closer to equal
traffic loa
Thanks. I applied this to master and branch-1.[789]. I take back the
claim about this bug going back farther than that; it only extends to
1.7.
On Mon, Jan 07, 2013 at 02:33:28PM -0800, Ethan Jackson wrote:
> Acked-by: Ethan Jackson
>
>
> On Fri, Jan 4, 2013 at 1:48 PM, Ben Pfaff wrote:
>
>
Signed-off-by: Ethan Jackson
---
utilities/bugtool/.gitignore |1 +
1 file changed, 1 insertion(+)
diff --git a/utilities/bugtool/.gitignore b/utilities/bugtool/.gitignore
index 4d31ed6..c242628 100644
--- a/utilities/bugtool/.gitignore
+++ b/utilities/bugtool/.gitignore
@@ -1 +1,2 @@
/ovs-
On Mon, Jan 07, 2013 at 03:13:50PM -0800, Ethan Jackson wrote:
> Signed-off-by: Ethan Jackson
Looks good, thanks.
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
On Mon, Jan 07, 2013 at 10:47:51PM +, Zoltan Kiss wrote:
> The old algorithm tries to converge to 0, despite it would mean a very
> unbalanced situation. We're aiming for an ideal ratio of 1, meaning both
> the 'from' and 'to' slave have the same load. Therefore, we only move an
> entry if it
Thanks, I've merged this.
Ethan
On Mon, Jan 7, 2013 at 3:15 PM, Ben Pfaff wrote:
> On Mon, Jan 07, 2013 at 03:13:50PM -0800, Ethan Jackson wrote:
> > Signed-off-by: Ethan Jackson
>
> Looks good, thanks.
>
___
dev mailing list
dev@openvswitch.org
http
On Jan 7, 2013, at 11:06 AM, Ben Pfaff wrote:
> We can assume that that is the common case, but it looks like your
> code goes farther than that and actually dereferences beyond array
> bounds if the assumption is bad. If port_no is 10 but there are only
> 5 ports, it looks very much to me that
Userspace-kernel communication is a possible bottleneck when OVS is
receiving a large number of flow set up requests. To help prevent a bad
actor from consuming too much of this resource, we introduced channels
to segegrate traffic. Previously, we created 17 channels and
round-robin assigned port
From: Pravin Shelar
Following patch series restructure GRE and IPIP tunneling code
to make it modular. It adds ip_tunnel module which acts as
generic tunneling layer which has common code. I have patch
to do same for VXLAN too.
In Addidtion to restructuring it adds demultiplexer for ipgre
protoc
From: Pravin Shelar
Signed-off-by: Pravin B Shelar
---
include/linux/skbuff.h | 12 ++
include/net/gre.h |6 +++
net/ipv4/af_inet.c |1 +
net/ipv4/gre.c | 101
net/ipv4/tcp.c |1 +
net/ipv4/udp.c
From: Pravin Shelar
Currently gre demultiplexer uses its own spin-lock. We should
rather RTNL lock.
Suggested-by: Stephen Hemminger
Signed-off-by: Pravin B Shelar
---
drivers/net/ppp/pptp.c |4
net/ipv4/gre.c | 18 +-
net/ipv4/ip_gre.c |4
3 fi
From: Pravin Shelar
I reference patch posted for IPGRE:
http://permalink.gmane.org/gmane.linux.network/226946
--8<--cut here-->8--
IP tunnels like other layered devices should propogate
carrier and state from lower device to tunnel.
Following patc
From: Pravin Shelar
Reuse common ip-tunneling code which is re-factored from GRE
module.
Signed-off-by: Pravin B Shelar
---
net/ipv4/Kconfig |1 +
net/ipv4/ipip.c | 778 ++
2 files changed, 84 insertions(+), 695 deletions(-)
diff --git
On Fri, Jan 04, 2013 at 09:19:57AM -0800, Ben Pfaff wrote:
> Since Jesse is perpetually backlogged: if there are any parts of the
> series that can be usefully applied without changing the datapath, let
> me know, and I'll take care of it.
Thanks.
I believe that with a small amount of reshuffling
Allow datapath to recognize and extract MPLS labels into flow keys
and execute actions which push, pop, and set labels on packets.
Based heavily on work by Leo Alterman and Ravi K.
Cc: Ravi K
Cc: Leo Alterman
Reviewed-by: Isaku Yamahata
Signed-off-by: Simon Horman
---
v2.14
* Remove include
Use the innermost dl_type when decoding L3 and L4 data from a packet.
Signed-off-by: Simon Horman
---
v2.14
* No change
v2.13
* No change
v2.12
* Use flow_innermost_dl_type helper
v2.11
* First post
---
lib/odp-util.c |6 --
1 file changed, 4 insertions(+), 2 deletions(-)
diff --gi
Allow repeated extraction of flow from packets
Split the L3 and above portion of ovs_flow_extract() out into
ovs_flow_extract_l3_onwards() and call ovs_flow_extract_l3_onwards()
from ovs_flow_extract().
This is to allow re-extraction of L3 and higher information using
an ethernet type supplied by
This adds support for the OpenFlow 1.1+ set_mpls_ttl action.
And also adds an NX set_mpls_ttl action.
The handling of the TTL modification is entirely handled in userspace.
Reviewed-by: Isaku Yamahata
Signed-off-by: Simon Horman
---
v2.14
* No change
v2.13
* No change
v2.12
* Use eth_type_m
Allow copy_ttl_in to function in the case where:
a) The outer header is MPLS;
b) The next-outer header is IP and;
c) The resulting nw ttl will be different from its existing value
In this circumstance a set ipv4 action will be used to adjust the TTL
but the dl_type of the match will be ETH_P_MPLS_
The ethernet type of an encapsulated frame may be obtained from
actions. If so it should be used to correct decoding of L3 and L4
packet data.
Signed-off-by: Simon Horman
---
v2.14
* No change
v2.13
* As suggested by Jarno Rajahalme
- Initialise encap_eth_type as 0 in ovs_dp_process_received
Actions may provide information to allow further decoding of
a flow.
For example:
In the case of MPLS, L3 and L4 information may not initially be decoded
from the frame as the ethernet type of the frame is an MPLS type and no
information is known about the type of the inner frame.
However, the t
This adds support for the OpenFlow 1.1+ copy_ttl_in action.
And also adds an NX copy_ttl_out action.
The implementation does not support copying in the case where the outermost
header is IP as it is unclear to me that Open vSwtich has a notion of an
inner IP header to copy the TLL from.
The imple
The ethernet type of an encapsulated frame may be obtained from
actions. If so it should be used to allow a richer conversion
of a flow to nlattrs.
Signed-off-by: Simon Horman
---
v2.14
* No change
v2.13
* No change
v2.12
* No change
v2.11
* First post
---
datapath/datapath.c |4 ++--
d
This adds support for the OpenFlow 1.1+ copy_ttl_out action.
And also adds an NX copy_ttl_out action.
The implementation does not support copying in the case where the outermost
header is IP as it is unclear to me that Open vSwtich has a notion of an
inner IP header to copy the TLL from.
The hand
Use the innermost dl_type when decoding L3 and L4 data from a packet.
Signed-off-by: Simon Horman
---
v2.14
* No change
v2.13
* No change
v2.12
* Use flow_innermost_dl_type helper
v2.11
* First post
---
lib/packets.c |4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a
If the actions of a flow allow further decoding of L3 and L4 data to
provide a richer match then use this richer match.
For example:
In the case of MPLS, L3 and L4 information may not initially be decoded
from the frame as the ethernet type of the frame is an MPLS type and no
information is known
This adds support for the OpenFlow 1.1+ dec_mpls_ttl action.
And also adds an NX dec_mpls_ttl action.
The handling of the TTL modification is entirely handled in userspace.
Reviewed-by: Isaku Yamahata
Signed-off-by: Simon Horman
---
v2.14
* No change
v2.13
* No change
v2.12
* Use eth_type_m
Hi,
This series implements basic MPLS actions and matches based on work by
Ravi K, Leo Alterman and Yamahata-san.
The main limitation of this implementation is that it only
supports manipulating the outer-most MPLS label.
Key differences between the v2.13 and v2.14:
* Reshuffle patches moving ke
This is the user-space datapath component of this change
Allow a flow to be visible in two forms, an outer flow and an inner flow.
This may occur if the actions allow further decoding of a packet to
provide a more fine-grained match. In this case the first-pass
coarse-grained match will hit the ou
Allow a flow to be visible in two forms, an outer flow and an inner flow.
This may occur if the actions allow further decoding of a packet to
provide a more fine-grained match. In this case the first-pass
coarse-grained match will hit the outer-flow and the second-pass
fined-grained match will hit
On Jan 7, 2013 9:41 PM, "Simon Horman" wrote:
>
> On Fri, Jan 04, 2013 at 09:19:57AM -0800, Ben Pfaff wrote:
> > Since Jesse is perpetually backlogged: if there are any parts of the
> > series that can be usefully applied without changing the datapath, let
> > me know, and I'll take care of it.
>
56 matches
Mail list logo
