From: Thomas Graf
Date: Fri, 8 Nov 2013 11:47:16 +0100
> Respin of the zerocopy patches for the openvswitch upcall.
>
> V4: - Daniel Borkmann pointed out that the style in skbuff.h has changed
> in net-next, adapted to no longer using extern in headers.
> V3: - Removed unneeded alignment
On Wed, Nov 13, 2013 at 06:21:51PM -0800, Ben Pfaff wrote:
> On Thu, Nov 14, 2013 at 11:19:07AM +0900, Simon Horman wrote:
> > * MPLS BoS match is supported as much as other MPLS matches.
> > That is, all that is missing is the kernel datapath side
> > which is part my pending patchset.
> >
>
On Wed, Nov 13, 2013 at 10:45:13AM -0800, Ben Pfaff wrote:
> On Tue, Nov 05, 2013 at 05:34:34PM +0900, Simon Horman wrote:
> > Some different versions of OpenFlow require different consistency
> > checking. This patch allows for three different variants to be checked.
> >
> > 1. OpenFlow1.0
> >
>
On Wed, Nov 13, 2013 at 03:39:05PM -0800, Ben Pfaff wrote:
> On Wed, Nov 13, 2013 at 02:12:58PM -0800, Jarno Rajahalme wrote:
> > Action translation to/from OpenFlow 1.2 set-field actions is done.
> >
> > Signed-off-by: Jarno Rajahalme
>
> Acked-by: Ben Pfaff
Reviewed-by: Simon Horman
__
At a high level, I don't think this patch is tracking the right thing.
We don't actually care how many times the bfd state changes. What we
actually care about is how many times the result of bfd_forwarding()
changes. That's because bfd_forwarding() is the function which
actually changes how tra
On Thu, Nov 14, 2013 at 11:19:07AM +0900, Simon Horman wrote:
> * MPLS BoS match is supported as much as other MPLS matches.
> That is, all that is missing is the kernel datapath side
> which is part my pending patchset.
>
> * The rework of tag order does require work but it is
> covered by
* MPLS BoS match is supported as much as other MPLS matches.
That is, all that is missing is the kernel datapath side
which is part my pending patchset.
* The rework of tag order does require work but it is
covered by my pending patchset.
Signed-off-by: Simon Horman
---
OPENFLOW-1.1+ | 7
This commit adds a new key "flap_count" to "bfd_status". It is to
count the number of bfd state flapping since start.
Signed-off-by: Alex Wang
---
v2 -> v3:
- rebase to master
v1 -> v2:
- rebase to master
---
lib/bfd.c| 14 ++
tests/bfd.at | 50 +++
On Wed, Nov 13, 2013 at 01:57:35PM -0800, Jarno Rajahalme wrote:
>
> On Nov 12, 2013, at 10:06 PM, Ben Pfaff wrote:
>
> > On Wed, Nov 13, 2013 at 03:06:04PM +0900, Simon Horman wrote:
> >> I have a question regarding the following entry in OPENFLOW-1.1+:
> >>
> >>* Action translation needs
On Wed, Nov 13, 2013 at 02:12:58PM -0800, Jarno Rajahalme wrote:
> Action translation to/from OpenFlow 1.2 set-field actions is done.
>
> Signed-off-by: Jarno Rajahalme
Acked-by: Ben Pfaff
___
dev mailing list
dev@openvswitch.org
http://openvswitch.or
Action translation to/from OpenFlow 1.2 set-field actions is done.
Signed-off-by: Jarno Rajahalme
---
OPENFLOW-1.1+ |5 -
1 file changed, 5 deletions(-)
diff --git a/OPENFLOW-1.1+ b/OPENFLOW-1.1+
index d7c2b5f..48ba441 100644
--- a/OPENFLOW-1.1+
+++ b/OPENFLOW-1.1+
@@ -86,11 +86,6 @@ ad
On Nov 12, 2013, at 10:06 PM, Ben Pfaff wrote:
> On Wed, Nov 13, 2013 at 03:06:04PM +0900, Simon Horman wrote:
>> I have a question regarding the following entry in OPENFLOW-1.1+:
>>
>>* Action translation needs some work to transform OpenFlow 1.1
>> field modification actions into Ope
Please ignore this patch. I have sent a v2 with more content in the commit
message.
On Wed, Nov 13, 2013 at 12:11 PM, Andy Zhou wrote:
> API changes only. No functional chnages.
>
> Reported by: Ben Pfaff
>
> Signed-off-by: Andy Zhou
> ---
> datapath/datapath.c | 16
> d
API changes only for code readability. No functional chnages.
This patch removes the underscored version. Added a new API
ovs_flow_tbl_lookup_stats() that
returns the n_mask_hits.
Reported by: Ben Pfaff
Signed-off-by: Andy Zhou
---
v1->v2:
Add more content to the commit message.
"fieldspec" is a string map with the following keys:
"prefix" A list of field names for which prefix lookup should be used.
The fields for which prefix lookup can be enabled are:
- tun_id, tun_src, tun_dst
- metadata
- reg0 -- reg7
- ip_src, ip_dst (or aliases nw_src and nw_dst)
- ipv6_src, ipv6
There is a bit more bookkeeping, and preliminary measurements did
not show significant benefit. Results may differ with larger
flow tables with more address space, though.
Signed-off-by: Jarno Rajahalme
---
lib/classifier.c | 158 +-
1 file c
Simplifies code a bit and makes it more understandable. We are
currently at 44 or so uint32s, so there is room for extending struct
flow before we run out of bits in the map.
Signed-off-by: Jarno Rajahalme
---
lib/flow.c | 233 ++-
lib/f
This series implements two techniques for better wildcarding:
1. Staged sub-table matching: Each sub-table is matched in segments,
starting from metadata and lower protocol layer fields, progressing
towards higher layers only as needed. That is, if we can determine
that there is no match
Signed-off-by: Jarno Rajahalme
---
lib/util.c |2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/lib/util.c b/lib/util.c
index a32c80a..75d443d 100644
--- a/lib/util.c
+++ b/lib/util.c
@@ -892,7 +892,7 @@ log_2_ceil(uint32_t n)
#elif __GNUC__ >= 4 && UINT_MAX == UINT32_MAX
/
Add a prefix tree (trie) structure for tracking the used IP address
space, enabling skipping classifier tables containing longer masks
than necessary for the given address. This enables more wildcarding
for datapath flows in ,e.g., parts of the address space without host
routes. In fact, the trie
Add raw_ctz64(), ctz64(), and popcount64() using builtins when
available.
I'm not sure if the "UINT64_MAX == ~0UL" and "UINT64_MAX == ~0ULL"
work in all cases as I imagine they would.
Signed-off-by: Jarno Rajahalme
---
lib/util.h| 51 +---
tests/tes
Each sub-table of a classifier is matched in segments, starting from
metadata and lower protocol layer fields, progressing towards higher
layers only as needed. That is, if we can determine that there is no
match on a specific segment, the higher layer fields need not be
looked at, and hence will
Make comparison done after matching hash values faster by optimizing
for the expected success.
Signed-off-by: Jarno Rajahalme
---
lib/classifier.c |2 +-
lib/match.c | 22 ++
lib/match.h |2 ++
3 files changed, 25 insertions(+), 1 deletion(-)
diff --git a
Count leading zeroes using builtins if available.
Signed-off-by: Jarno Rajahalme
---
lib/util.c| 25 -
lib/util.h| 35 +++
tests/library.at |1 +
tests/test-util.c | 53
Somehow it got pushed without the Signed-off-by. There's supposed to
be a hook to prevent that, I wonder what happened.
On Wed, Nov 13, 2013 at 12:43:47PM -0800, Jarno Rajahalme wrote:
> Thanks, applied to master.
>
> On Nov 12, 2013, at 8:37 AM, Ben Pfaff wrote:
>
> > On Tue, Nov 12, 2013 at
Thanks, applied to master.
On Nov 12, 2013, at 8:37 AM, Ben Pfaff wrote:
> On Tue, Nov 12, 2013 at 07:36:23AM -0800, Jarno Rajahalme wrote:
>> Add a "Performance" FAQ category. So far the only entry addresses the
>> issue with using a new kernel module with an older (pre-megaflows)
>> userspace
API changes only. No functional chnages.
Reported by: Ben Pfaff
Signed-off-by: Andy Zhou
---
datapath/datapath.c | 16
datapath/flow_table.c | 10 +-
datapath/flow_table.h |4 +++-
3 files changed, 16 insertions(+), 14 deletions(-)
diff --git a/datapath/data
On Wed, Nov 13, 2013 at 03:15:37PM +0900, Simon Horman wrote:
> Signed-off-by: Simon Horman
Applied, thanks!
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
On Wed, Nov 13, 2013 at 10:54:17AM -0800, Ben Pfaff wrote:
> On Wed, Nov 13, 2013 at 03:19:17PM +0900, Simon Horman wrote:
> > Signed-off-by: Simon Horman
> > ---
> > utilities/ovs-ofctl.8.in | 9 +
> > 1 file changed, 9 insertions(+)
> >
> > diff --git a/utilities/ovs-ofctl.8.in b/utili
On Wed, Nov 13, 2013 at 03:18:44PM +0900, Simon Horman wrote:
> Signed-off-by: Simon Horman
Applied, thanks!
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
On Wed, Nov 13, 2013 at 03:16:18PM +0900, Simon Horman wrote:
> Signed-off-by: Simon Horman
Applied, thanks!
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
On Wed, Nov 13, 2013 at 03:13:42PM +0900, Simon Horman wrote:
> Document masked versions of arp_sha and arp_tha matches.
> Also update documentation of unmasked versions of these
> matches to include an example address as is the case with
> the documentation of dl_src and dl_dst.
>
> Signed-off-by
On Wed, Nov 13, 2013 at 03:19:17PM +0900, Simon Horman wrote:
> Signed-off-by: Simon Horman
> ---
> utilities/ovs-ofctl.8.in | 9 +
> 1 file changed, 9 insertions(+)
>
> diff --git a/utilities/ovs-ofctl.8.in b/utilities/ovs-ofctl.8.in
> index f8feb6d..7c31386 100644
> --- a/utilities/ovs
The number of masks in the datapath should not be large. But saving memory
is a good thing at any time. The change looks good to me.
Acked-by: Andy Zhou
On Wed, Nov 13, 2013 at 9:54 AM, Ben Pfaff wrote:
> We won't normally have a ton of flow masks but using a size_t to store
> values no bigg
On Wed, Nov 13, 2013 at 03:17:54PM +0900, Simon Horman wrote:
> * MPLS BoS match is supported as much as other MPLS matches.
> That is, all that is missing is the kernel datapath side
> which is part my pending patchset.
>
> * The rework of tag order does require work but it is
> covered by
The comment change is fine.
Not related to the comment, I am not happy with the abstraction of
flow_mask_insert() -- passing in flow, and setting up flow->mask should not
be part of the insert() function.
A better function name may help. I did not come up with a better one.
On the other hand, ho
On Tue, Nov 05, 2013 at 05:34:34PM +0900, Simon Horman wrote:
> Some different versions of OpenFlow require different consistency
> checking. This patch allows for three different variants to be checked.
>
> 1. OpenFlow1.0
>
>This variant implicitly pushes a VLAN tag if one doesn't exist
>
Ben is right, the other way around will read better. I will send out a
patch to fix this.
On Wed, Nov 13, 2013 at 10:00 AM, Justin Pettit wrote:
> On Nov 13, 2013, at 9:54 AM, Ben Pfaff wrote:
>
> > La la la, does anyone ever read part 0 of a patch series?
>
> That's the best part. Everythin
On Wed, Nov 13, 2013 at 08:32:32AM -0800, Gurucharan Shetty wrote:
> On Tue, Nov 12, 2013 at 3:59 PM, Ben Pfaff wrote:
> > On Mon, Nov 04, 2013 at 06:23:54AM -0800, Gurucharan Shetty wrote:
> >> Instead of an exact match flow table, we introduce a classifier.
> >> This enables mega-flows in usersp
On Nov 13, 2013, at 9:54 AM, Ben Pfaff wrote:
> La la la, does anyone ever read part 0 of a patch series?
That's the best part. Everything else is just details.
--Justin
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listin
Signed-off-by: Ben Pfaff
---
datapath/flow_table.c |6 +-
1 file changed, 1 insertion(+), 5 deletions(-)
diff --git a/datapath/flow_table.c b/datapath/flow_table.c
index ddb14da..4804db0 100644
--- a/datapath/flow_table.c
+++ b/datapath/flow_table.c
@@ -519,11 +519,7 @@ static struct sw_
Just a couple of things I noticed while reviewing Guru's userspace
datapath patch yesterday.
La la la, does anyone ever read part 0 of a patch series? Just in
case, I noticed another oddity in the datapath but I don't have a
patch to propose: normally, __function() is a wrapper around
function(),
We won't normally have a ton of flow masks but using a size_t to store
values no bigger than sizeof(struct sw_flow_key) seems excessive.
This reduces sw_flow_key_range and sw_flow_mask by 4 bytes on 32-bit
systems. On 64-bit systems it shrinks sw_flow_key_range by 12 bytes but
sw_flow_mask only b
On Tue, Nov 12, 2013 at 3:59 PM, Ben Pfaff wrote:
> On Mon, Nov 04, 2013 at 06:23:54AM -0800, Gurucharan Shetty wrote:
>> Instead of an exact match flow table, we introduce a classifier.
>> This enables mega-flows in userspace datapath.
>>
>> Signed-off-by: Gurucharan Shetty
>
> dp_netdev_lookup_
On 11/13/2013 07:11 AM, Jesse Gross wrote:
On Mon, Nov 11, 2013 at 11:53 PM, Thomas Graf wrote:
On 11/11/2013 04:50 PM, Ben Pfaff wrote:
On Mon, Nov 11, 2013 at 04:36:24PM +0100, Thomas Graf wrote:
Following commit (''netlink: Do not enforce alignment of last Netlink
attribute''), signal th
On Wed, Nov 13, 2013 at 03:14:35PM +0900, Simon Horman wrote:
> Document masked versions of arp_sha and arp_tha matches.
>
> Also update documentation of unmasked versions of these
> matches to include an example address as is the case with
> the documentation of dl_src and dl_dst.
Sorry, I manag
Signed-off-by: Simon Horman
---
utilities/ovs-ofctl.8.in | 11 +++
1 file changed, 11 insertions(+)
diff --git a/utilities/ovs-ofctl.8.in b/utilities/ovs-ofctl.8.in
index f8feb6d..27c85b5 100644
--- a/utilities/ovs-ofctl.8.in
+++ b/utilities/ovs-ofctl.8.in
@@ -923,6 +923,17 @@ The \fBip_
On Wed, Nov 06, 2013 at 10:46:12PM +0800, Alexander Wu wrote:
>
> V2:
> No change.
>
> V1:
> Make the enums allowed
>
> Signed-off-by: Alexander Wu
Reviewed-by: Simon Horman
> ---
> lib/rconn.c |4 ++--
> 1 files changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/lib/rconn.c
On Wed, Nov 06, 2013 at 10:46:10PM +0800, Alexander Wu wrote:
>
> V2:
> No change
>
> V1:
> Now the cli we implement is very crude. Maybe it could
> display better.
>
> Signed-off-by: Alexander Wu
Reviewed-by: Simon Horman
> ---
> utilities/ovs-ofctl.c | 17 +
> 1 fi
On Wed, Nov 06, 2013 at 10:46:07PM +0800, Alexander Wu wrote:
>
> V2:
> Change calls of ofputil.
> Fix CodingStyle accoring to Simon Horman's suggestions.
>
> V1:
> Add function to print OFPMP_TABLE_FEATURES.
> But now the print is crude and dirty.
> Fix it to bitmap or more desc later.
On Wed, Nov 06, 2013 at 10:45:31PM +0800, Alexander Wu wrote:
>
> V2:
> Restructure implement of OFPMP_TABLE_FEATURES
> Change decode_*_raw to normalized pull functions
>
> 1. add macros and funcs to en/decode table features
> 2. restructure OFPMP_TABLE_FEATURES en/decode function
>
On Wed, Nov 06, 2013 at 10:46:04PM +0800, Alexander Wu wrote:
>
> V2:
> 1. fix function align, align to 79 bytes, fix BE64 to UINT64, etc.
> 2. fix n_tables_miss init type.
> 3. Change calls from ofputil.
>
> Simon Horman suggestions:
> 1. Fix CodingStyle, some coding error.
> 2. Fix th
I think it would be quite helpful if you added some text describing
the relationship between tf and otf:
a) in the changelog
b) above otf
c) above tf
On Wed, Nov 06, 2013 at 10:45:52PM +0800, Alexander Wu wrote:
>
> V2:
> No change.
>
> V1:
> Add table_feature structs in ofproto && oftable
53 matches
Mail list logo
