On Fri, Dec 13, 2013 at 09:53:25PM -0500, Kyle Mestery wrote:
> Fixes the OVS builds on Fedora, which are currently broken upstream.
>
> Signed-off-by: Kyle Mestery
Guru, will you review this (and apply it if you're happy with it)?
Thanks,
Ben.
___
d
Fixes the OVS builds on Fedora, which are currently broken upstream.
Signed-off-by: Kyle Mestery
---
rhel/openvswitch-fedora.spec.in | 2 +-
rhel/openvswitch-kmod-fedora.spec.in | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/rhel/openvswitch-fedora.spec.in b/rhel/ope
Currently, every time a monitoring port is added or reconfigured, the
main thread notifies the monitoring thread to wake up immediately using
monitor_seq. When adding a large number of ports at once, this causes
contention as the threads fight over access to the monitor heap---one
thread adding new
I realised that this no longer applies on master. I'll send a rebased version.
On 13 December 2013 15:00, Alex Wang wrote:
> Hey Ben,
>
> Could you help review and push it?
>
> Thanks,
> Alex Wang,
>
>
> On Fri, Dec 13, 2013 at 11:29 AM, Alex Wang wrote:
>>
>> Looks good to me, ran the cfm flap_
This patch is to add for secure pseudorandom number generator on windows.
Signed-off-by: Alin Serdean
---
lib/entropy.c | 13 +
1 files changed, 13 insertions(+), 0 deletions(-)
---
diff --git a/lib/entropy.c b/lib/entropy.c
index 02f56e0..45e83ec 100644
--- a/lib/entropy.c
+++ b/
On Fri, Dec 13, 2013 at 04:20:42PM -0800, Justin Pettit wrote:
> Signed-off-by: Justin Pettit
Acked-by: Ben Pfaff
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
Signed-off-by: Justin Pettit
---
NEWS |4
configure.ac |2 +-
debian/changelog |7 +++
3 files changed, 12 insertions(+), 1 deletions(-)
diff --git a/NEWS b/NEWS
index e0fafd0..35a9cd0 100644
--- a/NEWS
+++ b/NEWS
@@ -1,3 +1,7 @@
+v2.0.1 - 13 Dec 2013
+--
On Fri, Nov 15, 2013 at 03:12:18PM -0800, Jarno Rajahalme wrote:
> Allowing the packet to be modified by execution allows less data
> copying for userspace action execution. Some users of the
> dpif_execute already expect that the packet may be modified. This
> patch makes this behavior uniform a
Looks good to me, tested on bare metal and vm,
Could you add commit message to summarize the test? (especially that it
checks the cfm_fault_status update in database)
On Fri, Dec 13, 2013 at 2:46 PM, Joe Stringer wrote:
> Signed-off-by: Joe Stringer
> ---
> tests/cfm.at | 38 ++
Thanks, I fixed the commit message and pushed.
On Fri, Dec 13, 2013 at 02:48:48PM -0800, Alex Wang wrote:
> In the commit message, 's/UINT_MAX_1/UINT_MAX-1',
>
> Looks good to me,
>
>
>
> On Fri, Dec 13, 2013 at 2:31 PM, Ben Pfaff wrote:
>
> > ofproto_set_threads() uses the calculation MAX(c
On Fri, Dec 06, 2013 at 03:36:04PM -0800, Joe Stringer wrote:
> Looks good to me.
Thanks, I'll apply this soon.
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
Hey Ben,
Could you help review and push it?
Thanks,
Alex Wang,
On Fri, Dec 13, 2013 at 11:29 AM, Alex Wang wrote:
> Looks good to me, ran the cfm flap_count test 300 (clang and sparse) times
> on my vm
> and didn't see the failure any more,
>
>
> On Fri, Dec 13, 2013 at 10:52 AM, Joe Stringer
On Tue, Nov 05, 2013 at 05:35:03PM +0900, Simon Horman wrote:
> ovs-ofctl needs to mask the allowed protocols with the usable protocols
> after calling parse_ofp_str() so that only usable protocols will be
> negotiated.
>
> This problem became apparent to me after 8621547cc94ad910
> ("lib/ofp-acti
Thanks for the update,
On Fri, Dec 13, 2013 at 2:52 PM, Ben Pfaff wrote:
> On Mon, Nov 25, 2013 at 12:49:45PM -0800, Alex Wang wrote:
> > LGTM, only one thing,
> >
> > I did the following test:
> >
> > """
> > root@server329:~/ben-reviews# ovs-vsctl --version
> > ovs-vsctl (Open vSwitch) 2.0.90
On Mon, Nov 25, 2013 at 12:49:45PM -0800, Alex Wang wrote:
> LGTM, only one thing,
>
> I did the following test:
>
> """
> root@server329:~/ben-reviews# ovs-vsctl --version
> ovs-vsctl (Open vSwitch) 2.0.90
> Compiled Nov 25 2013 11:44:29
> root@server329:~/ben-reviews# lsof -p `pidof ovs-vswitch
In the commit message, 's/UINT_MAX_1/UINT_MAX-1/',
Looks good to me,
On Fri, Dec 13, 2013 at 2:31 PM, Ben Pfaff wrote:
> ofproto_set_threads() uses the calculation MAX(count_cpu_cores() - 2, 1)
> to decide on the default thread count. However, count_cpu_cores() returns
> 0 if it can't count
In the commit message, 's/UINT_MAX_1/UINT_MAX-1',
Looks good to me,
On Fri, Dec 13, 2013 at 2:31 PM, Ben Pfaff wrote:
> ofproto_set_threads() uses the calculation MAX(count_cpu_cores() - 2, 1)
> to decide on the default thread count. However, count_cpu_cores() returns
> 0 if it can't count t
Signed-off-by: Joe Stringer
---
tests/cfm.at | 38 ++
1 file changed, 38 insertions(+)
diff --git a/tests/cfm.at b/tests/cfm.at
index 9e351d0..8e5917d 100644
--- a/tests/cfm.at
+++ b/tests/cfm.at
@@ -96,4 +96,42 @@ CFM_VSCTL_LIST_IFACE([p0], [cfm_flap_count]
We allow zero 'values' in a miniflow for it to have the same map
as the corresponding minimask. Minimasks themselves never have
zero data values, though. Document this and optimize the code
accordingly.
Signed-off-by: Jarno Rajahalme
---
lib/flow.c | 16 +++-
lib/flow.h |9
ofproto_set_threads() uses the calculation MAX(count_cpu_cores() - 2, 1)
to decide on the default thread count. However, count_cpu_cores() returns
0 if it can't count the number of cores, or 1 if there's only one core,
and that causes the calculation to come out as UINT_MAX-2 or UINT_MAX_1,
respec
Thanks for the explanation, I'm good with the patch,
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
Thanks, I applied these patches.
On Mon, Dec 02, 2013 at 10:16:28AM -0800, Alex Wang wrote:
> LGTM, thx
>
>
> On Sat, Nov 2, 2013 at 10:02 PM, Ben Pfaff wrote:
>
> > Signed-off-by: Ben Pfaff
> > ---
> > tests/ovsdb-server.at | 85
> > +
> > 1
On Wed, Nov 27, 2013 at 02:58:13PM -0800, Alex Wang wrote:
> +if (get_detach()) {
> > +int error;
> > +
> > +error = unixctl_server_create(NULL, &unixctl);
> > +if (error) {
> > +ovs_fatal(error, "failed to create unixctl server");
> > +}
> > +
> > +
This patch depends on
http://openvswitch.org/pipermail/dev/2013-December/035103.html
I'll rebase and resend it later, other two patches are unaffected,
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
On Fri, Oct 25, 2013 at 05:18:02PM -0700, Alex Wang wrote:
> Commit 348f01e3e3 (cfm: Eight byte MPIDs in extended mode.) allows
> eight byte MPIDs when running in extended mode. This commit explains
> this change in the vswitch.conf.db.
>
> Signed-off-by: Alex Wang
Applied, thanks.
On Fri, Dec 13, 2013 at 1:55 AM, Francesco Fusco wrote:
> On 12/12/2013 09:20 PM, Jesse Gross wrote:
>>
>> Out of curiosity, did you try using crc32q? OVS data structures are
>> already aligned to 8 bytes. It would also be interesting to know if a
>> parallelized implementation is worthwhile, alth
I first sent these fixes on September 30. This version is rebased,
with conflicts fixed.
Ben Pfaff (2):
bridge: Let ofprotos run once before reporting configuration
complete.
lacp: Give LACP a moment to initialize before testing its state, in
tests.
AUTHORS |1 +
tests
Occasionally in the unit tests the following race can happen:
1. ovs-vsctl updates database
2. ovs-vswitchd reconfigures, notifies ovs-vsctl that it is complete
3. ovs-appctl ofproto/trace fails to see newly added port
4. ovs-vswitchd main loop calls ofproto's ->type_run(), making the
These tests configured LACP and then immediately dumped out its state.
Most of the time, this worked, but there was a brief race window in which
the "negotiated" flag could be missing because this took one pass through
the main loop. This fixes the problem.
This race may be seen in the failures o
Not at all. I want to do a quick review myself over them before I send it in
though.
Thanks for the explanation, I am quite new to this "world" :-).
Kind Regards,
Alin.
From: Gurucharan Shetty [shet...@nicira.com]
Sent: Friday, December 13, 2013 8:40 PM
T
Looks good to me, ran the cfm flap_count test 300 (clang and sparse) times
on my vm
and didn't see the failure any more,
On Fri, Dec 13, 2013 at 10:52 AM, Joe Stringer wrote:
> Currently, every time a monitoring port is added or reconfigured, the
> main thread notifies the monitoring thread to w
With ovs multithreading implementation, the bfd decay test
becomes fragile due to its high dependency on timing sequence.
This commit removes these dependencies and makes the test robust.
Signed-off-by: Alex Wang
---
tests/bfd.at | 141 --
Commit 307464a1 (ofproto-dpif-monitor: Use heap to order the mport
wakeup time.) re-heapifies the heap in monitor_run(). So the
monitor_run() should be protected by the write lock, rather than
the read lock.
This commit fixes the issue.
Signed-off-by: Alex Wang
---
ofproto/ofproto-dpif-monitor
Commit 307464a11 (ofproto-dpif-monitor: Use heap to order the mport
wakeup time.) makes bfd only send packet at specified periodic instant.
This fails to meet the RFC5880 requirement, which requires bfd send
FINAL immediately after receiving POLL.
This commit fixes the above issue by scheduling bf
I don't have a strong preference about the structuring. Anything that is
intuitive is fine with me. With regards to keeping all pragma comments
separate, I actually like keeping them close to the files that use them.
However, since these are DLLs, I don't see any major downside in declaring them
Currently, every time a monitoring port is added or reconfigured, the
main thread notifies the monitoring thread to wake up immediately using
monitor_seq. When adding a large number of ports at once, this causes
contention as the threads fight over access to the monitor heap---one
thread adding new
yeah, I can only reproduce it on my VM setup by running command like this:
for i in `seq 1 1000`; do make check TESTSUITEFLAGS=12 1>/dev/null
2>file_make_check; if cat file_make_check | grep "ERROR: 1 test was run";
then echo FAIL; break; fi; done
and wait a long time.
But the issue is valid. L
Would you mind sending in a proper patch which includes your changes
added to the mentioned patch for a review with a commit message and
rationale. (The reviewers usually do a "git am 'patch.txt' and then
review. And if there is a review comment, you re-spin the patch with
the subject edited with a
Ben Pfaff schrieb:
>On Fri, Dec 13, 2013 at 10:30:56AM -0800, Ben Pfaff wrote:
>> On Fri, Dec 13, 2013 at 06:54:27PM +0100, Helmut Schaa wrote:
>> > Reduces binary size. Use a static inline function instead of
>> > a macro to not get "unused variable" warnings everywhere.
>> >
>> > Making ovs
On Fri, Dec 13, 2013 at 10:30:56AM -0800, Ben Pfaff wrote:
> On Fri, Dec 13, 2013 at 06:54:27PM +0100, Helmut Schaa wrote:
> > Reduces binary size. Use a static inline function instead of
> > a macro to not get "unused variable" warnings everywhere.
> >
> > Making ovs_assert an empty inline in the
Curious. I can't seem to reproduce the occasional fault in my VM
setup. I'm happy to fold in the change that you suggested though Alex.
(Incremental below)
diff --git a/tests/cfm.at b/tests/cfm.at
index 620e3e0..fe6778a 100644
--- a/tests/cfm.at
+++ b/tests/cfm.at
@@ -90,12 +90,9 @@ CFM_VSCTL_LIST
On Fri, Dec 13, 2013 at 06:54:27PM +0100, Helmut Schaa wrote:
> Reduces binary size. Use a static inline function instead of
> a macro to not get "unused variable" warnings everywhere.
>
> Making ovs_assert an empty inline in the header file effectively
> optimizes out all calls to ovs_assert.
>
I was thinking to include all pragmas and other needed defines in a common
header used through all of the source
files(http://openvswitch.org/pipermail/dev/2013-December/034983.html), but this
is just my thought.
The way you want to structure it is up too you :-).
Alin.
___
On Fri, Dec 13, 2013 at 10:11:41AM -0800, Gurucharan Shetty wrote:
> On Fri, Dec 13, 2013 at 9:43 AM, Ben Pfaff wrote:
> > On Wed, Dec 11, 2013 at 04:11:24PM -0800, Gurucharan Shetty wrote:
> >> For systems that do not use linker sections and also do not
> >> have either HAVE_THREAD_LOCAL or HAVE_
In case of flow-miss kernel sends skb to userspace and then
userspace sends execute netlink msg to switch the packet. But
in this process skb rxhash is not restored. skb rxhash is used
for calculating tunnel source port for vxlan. If rxhash is not
restored packet which is sent to miss call handler
On Fri, Dec 13, 2013 at 06:45:14PM +0100, Helmut Schaa wrote:
> On Fri, Dec 13, 2013 at 6:08 PM, Ben Pfaff wrote:
> > On Fri, Dec 13, 2013 at 02:17:16PM +0100, Helmut Schaa wrote:
> >> Reduces binary size. Use a static inline function instead of
> >> a macro to not get "unused variable" warning ev
Commit f23d157c ("ofproto-dpif: Don't poll ports when nothing changes"
changed how cfm information is propagated to ovsdb, but did not ensure
that this happened when users made "appctl cfm/set_fault" calls. This
patch fixes this case.
Signed-off-by: Joe Stringer
---
lib/cfm.c |1 +
1 file ch
On Fri, Dec 13, 2013 at 9:43 AM, Ben Pfaff wrote:
> On Wed, Dec 11, 2013 at 04:11:24PM -0800, Gurucharan Shetty wrote:
>> For systems that do not use linker sections and also do not
>> have either HAVE_THREAD_LOCAL or HAVE___THREAD (ex: windows
>> using MSVC), a COVERAGE_INC() calls xmalloc which
On Fri, Dec 13, 2013 at 09:59:36AM -0800, Ethan Jackson wrote:
> Fixes the following warning on my system.
>
> ../lib/coverage.c:54:40: error: incompatible pointer types passing
> 'unsigned int *' to parameter of type 'size_t *' (aka 'unsigned long
> *')
>
> Signed-off-by: Ethan Jackson
Acked-b
Fixes the following warning on my system.
../lib/coverage.c:54:40: error: incompatible pointer types passing
'unsigned int *' to parameter of type 'size_t *' (aka 'unsigned long
*')
Signed-off-by: Ethan Jackson
---
lib/coverage.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff -
Thanks Guru and Alin, I applied these.
On Thu, Dec 12, 2013 at 08:38:48PM +, Alin Serdean wrote:
> I applied the patches and everything seems in order.
>
> The same remark about adding string.h to lib/.gitignore
>
> Thanks,
> Alin.
>
> From: dev-boun.
Currently openvswitch builds all libraries static only. However,
libopenvswitch is linked into nearly all openvswitch executables
making it hardly possible to run openvswitch on embedded devices
(for example running OpenWrt).
Convert openvswitch to use libtool for building its internal libs.
This
Reduces binary size. Use a static inline function instead of
a macro to not get "unused variable" warnings everywhere.
Making ovs_assert an empty inline in the header file effectively
optimizes out all calls to ovs_assert.
Signed-off-by: Helmut Schaa
---
lib/util.h | 7 ++-
1 file changed,
On Thu, Dec 12, 2013 at 11:30:49AM -0800, Gurucharan Shetty wrote:
> Between, what is the difference between NOTICE and COPYING?
COPYING is the original file we had when we started the project.
NOTICE is required by the Apache 2.0 license.
I'm not sure we need both, and the distinction isn't cle
Your message dated Fri, 13 Dec 2013 17:48:48 +
with message-id
and subject line Bug#732070: fixed in openvswitch 1.9.3+git20131029-1.1
has caused the Debian Bug report #732070,
regarding openvswitch-switch: installation fails
to be marked as done.
This means that you claim that the problem ha
On Thu, Dec 12, 2013 at 11:27:03AM -0800, Gurucharan Shetty wrote:
> On Wed, Dec 11, 2013 at 3:04 PM, Ben Pfaff wrote:
> > This will be used for the #include_next in string.h in the following
> > commit.
> >
> > Signed-off-by: Ben Pfaff
> > ---
> > NOTICE|5 +-
> > build-aux/
Accepted:
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA256
Format: 1.8
Date: Fri, 13 Dec 2013 16:53:41 +
Source: openvswitch
Binary: openvswitch-datapath-source openvswitch-datapath-dkms
openvswitch-common openvswitch-switch openvswitch-ipsec openvswitch-pki
openvswitch-controller openvswi
openvswitch_1.9.3+git20131029-1.1_amd64.changes uploaded successfully to
localhost
along with the files:
openvswitch_1.9.3+git20131029-1.1.dsc
openvswitch_1.9.3+git20131029-1.1.debian.tar.gz
openvswitch-common_1.9.3+git20131029-1.1_amd64.deb
openvswitch-switch_1.9.3+git20131029-1.1_amd64.d
On Fri, Dec 13, 2013 at 6:08 PM, Ben Pfaff wrote:
> On Fri, Dec 13, 2013 at 02:17:16PM +0100, Helmut Schaa wrote:
>> Reduces binary size. Use a static inline function instead of
>> a macro to not get "unused variable" warning everywhere.
>>
>> Signed-off-by: Helmut Schaa
>> ---
>>
>> Not sure if
On Wed, Dec 11, 2013 at 04:11:24PM -0800, Gurucharan Shetty wrote:
> For systems that do not use linker sections and also do not
> have either HAVE_THREAD_LOCAL or HAVE___THREAD (ex: windows
> using MSVC), a COVERAGE_INC() calls xmalloc which inturn calls
> COVERAGE_INC() creating a recursion that
On Fri, Dec 13, 2013 at 5:35 PM, Ben Pfaff wrote:
> On Fri, Dec 13, 2013 at 02:17:13PM +0100, Helmut Schaa wrote:
>> Avoid "holes" in structs by moving "small" members together.
>>
>> Signed-off-by: Helmut Schaa
>
> I don't think any of these is particularly useful. Most notably, the
> two "ofpu
On Fri, Dec 13, 2013 at 06:35:27PM +0100, Helmut Schaa wrote:
> On Fri, Dec 13, 2013 at 6:27 PM, Ben Pfaff wrote:
> > On Fri, Dec 13, 2013 at 02:05:04PM +0100, Helmut Schaa wrote:
> >> Currently openvswitch builds all libraries static only. However,
> >> libopenvswitch is linked into nearly all op
On Fri, Dec 13, 2013 at 6:27 PM, Ben Pfaff wrote:
> On Fri, Dec 13, 2013 at 02:05:04PM +0100, Helmut Schaa wrote:
>> Currently openvswitch builds all libraries static only. However,
>> libopenvswitch is linked into nearly all openvswitch executables
>> making it hardly possible to run openvswitch
Yep, you didn't give me an acked on the last round so I resent it.
Thanks for the reviews.
Ethan
On Thu, Dec 12, 2013 at 11:07 PM, Ben Pfaff wrote:
> On Thu, Dec 12, 2013 at 08:24:56PM -0800, Ethan Jackson wrote:
>> There's no particular reason for the function controlling the number
>> of thre
On Fri, Dec 13, 2013 at 02:05:03PM +0100, Helmut Schaa wrote:
> They are unused and can be removed.
>
> Signed-off-by: Helmut Schaa
Thanks.
I applied patches 1 through 4.
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo
On Fri, Dec 13, 2013 at 02:05:05PM +0100, Helmut Schaa wrote:
> Signed-off-by: Helmut Schaa
I'd fold this into patch 5.
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
On Fri, Dec 13, 2013 at 02:05:04PM +0100, Helmut Schaa wrote:
> Currently openvswitch builds all libraries static only. However,
> libopenvswitch is linked into nearly all openvswitch executables
> making it hardly possible to run openvswitch on embedded devices
> (for example running OpenWrt).
>
On Fri, Dec 13, 2013 at 02:17:17PM +0100, Helmut Schaa wrote:
> str_to_uint, str_to_ulong and str_to_ullong are just wrappers
> around the corresponding signed functions. Move these to util.h
> and make them inline saving some library exports and letting
> the compiler do some more magic.
>
> Sign
On Fri, Dec 13, 2013 at 02:17:16PM +0100, Helmut Schaa wrote:
> Reduces binary size. Use a static inline function instead of
> a macro to not get "unused variable" warning everywhere.
>
> Signed-off-by: Helmut Schaa
> ---
>
> Not sure if this is the best idea but in my builds it reduces libopenv
On Fri, Dec 13, 2013 at 02:17:15PM +0100, Helmut Schaa wrote:
> Does not matter as long as ovs_assert is a simple macro but if ovs_assert
> is converted to a function this will fail.
>
> Signed-off-by: Helmut Schaa
Good catch.
___
dev mailing list
dev@
Thanks for the detail, I'll look into it.
On 12 December 2013 21:50, Alex Wang wrote:
> Yes, I could reproduce it on my VM setup,
>
> It is the timing issue again, if the p1 starts sending ccm first, it will
> have flap count = 0.
>
> If p1 receives ccm from p0 first, it will go to [rdi] and then
On Dec 13, 2013, at 5:17 AM, Helmut Schaa wrote:
> Reduces binary size. Use a static inline function instead of
> a macro to not get "unused variable" warning everywhere.
>
> Signed-off-by: Helmut Schaa
> ---
>
> Not sure if this is the best idea but in my builds it reduces libopenvswitch
> b
Hi Ben,
it's always a pleasure to see that you are ok with help. I will NMU shortly.
Thomas
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
On Fri, Dec 13, 2013 at 02:17:13PM +0100, Helmut Schaa wrote:
> Avoid "holes" in structs by moving "small" members together.
>
> Signed-off-by: Helmut Schaa
I don't think any of these is particularly useful. Most notably, the
two "ofputil_*" structures are only ever allocated one at a time, on
On Sat, Dec 14, 2013 at 12:22:46AM +0800, Thomas Goirand wrote:
> Since this completely blocks the apt-get install process, and makes it
> impossible to install openvswitch-switch without manual intervention, I
> have set this bug severity to "grave", because it fits this definition:
>
> "makes th
Hi,
Since this completely blocks the apt-get install process, and makes it
impossible to install openvswitch-switch without manual intervention, I
have set this bug severity to "grave", because it fits this definition:
"makes the package in question unusable by most or all users."
By the way, it
Processing commands for cont...@bugs.debian.org:
> severity 732070 grave
Bug #732070 [openvswitch-switch] openvswitch-switch: installation fails
Severity set to 'grave' from 'important'
>
End of message, stopping processing here.
Please contact me if you need assistance.
--
732070: http://bugs.d
Package: openvswitch-switch
Version: 1.9.3+git20131029-1
Severity: important
Dear Maintainer,
root@t430gone:/home/goneri# apt-get install openvswitch-switch
Reading package lists... Done
Building dependency tree
Reading state information... Done
The following packages were automatically in
On 12/13/2013 11:01 AM, David Laight wrote:
My thoughts exactly.
Given this is a hash it could crc alternate words into separate
accumulators and the combine the values at the end.
That way you are still doing sequential accesses to the data.
(The crc instruction might be better than an xor for t
Make the skb zerocopy logic written for nfnetlink queue available for
use by other modules.
Signed-off-by: Thomas Graf
Reviewed-by: Daniel Borkmann
---
include/linux/skbuff.h | 3 ++
net/core/skbuff.c| 85
net/netfilter/nfn
Allows removing the net and dp_ifindex argument and simplify the
code.
Signed-off-by: Thomas Graf
---
net/openvswitch/datapath.c | 34 ++
1 file changed, 14 insertions(+), 20 deletions(-)
diff --git a/net/openvswitch/datapath.c b/net/openvswitch/datapath.c
index
Use of skb_zerocopy() can avoid the expensive call to memcpy()
when copying the packet data into the Netlink skb. Completes
checksum through skb_checksum_help() if not already done in
GSO segmentation.
Zerocopy is only performed if user space supported unaligned
Netlink messages. memory mapped net
Signed-off-by: Thomas Graf
Reviewed-by: Daniel Borkmann
---
include/uapi/linux/openvswitch.h | 4
net/openvswitch/datapath.c | 14 ++
net/openvswitch/datapath.h | 2 ++
3 files changed, 20 insertions(+)
diff --git a/include/uapi/linux/openvswitch.h b/include/uapi/
The copy & csum optimization is no longer present with zerocopy
enabled. Compute the checksum in skb_gso_segment() directly by
dropping the HW CSUM capability from the features passed in.
Signed-off-by: Thomas Graf
---
net/openvswitch/datapath.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion
Drop user features if an outdated user space instance that does not
understand the concept of user_features attempted to create a new
datapath.
Signed-off-by: Thomas Graf
---
include/uapi/linux/openvswitch.h | 10 +-
net/openvswitch/datapath.c | 21 +
2 files ch
Series implementing a zerocopy method for OVS upcall messages. Based
on top of commit: (''openvswitch: Enable memory mapped Netlink i/o'')
Thomas Graf (6):
net: Export skb_zerocopy() to zerocopy from one skb to another
openvswitch: Allow user space to announce ability to accept unaligned
N
Helmut Schaa (5):
lib/ofp: Consolidate some struct field orders
lib/random: Make some functions in random.c inline
lib/mac-learning: Add missing semicolon
lib/util.h: Disable ovs_assert when build with NDEBUG
lib/util: Make some functions in util.c inline
lib/mac-learning.c | 2 +-
lib
Does not matter as long as ovs_assert is a simple macro but if ovs_assert
is converted to a function this will fail.
Signed-off-by: Helmut Schaa
---
lib/mac-learning.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/lib/mac-learning.c b/lib/mac-learning.c
index fe06744..c9c1a
Avoid "holes" in structs by moving "small" members together.
Signed-off-by: Helmut Schaa
---
lib/ofp-actions.h | 2 +-
lib/ofp-util.h| 8
2 files changed, 5 insertions(+), 5 deletions(-)
diff --git a/lib/ofp-actions.h b/lib/ofp-actions.h
index 470a371..00cba6a 100644
--- a/lib/ofp-
random_range, random_uint8 and random_uint16 are basically just wrappers
around random_uint32. Hence, don't export these functions so the compiler
can be smarter when it comes to optimizations.
Signed-off-by: Helmut Schaa
---
lib/random.c | 17 -
lib/random.h | 21 +++
str_to_uint, str_to_ulong and str_to_ullong are just wrappers
around the corresponding signed functions. Move these to util.h
and make them inline saving some library exports and letting
the compiler do some more magic.
Signed-off-by: Helmut Schaa
---
lib/util.c | 18 --
lib/util
Reduces binary size. Use a static inline function instead of
a macro to not get "unused variable" warning everywhere.
Signed-off-by: Helmut Schaa
---
Not sure if this is the best idea but in my builds it reduces libopenvswitch
by around 10%.
Thoughts?
Thanks,
Helmut
lib/util.h | 5 +
1 f
Use a static array in coverage.c that gets initialized by constructor
functions per coverage definition.
Signed-off-by: Helmut Schaa
---
lib/coverage.c | 46 +++---
lib/coverage.h | 20 ++--
2 files changed, 21 insertions(+), 45 deletions(-
They are unused and can be removed.
Signed-off-by: Helmut Schaa
---
configure.ac | 1 -
lib/automake.mk | 22
m4/openvswitch.m4 | 77 ---
3 files changed, 100 deletions(-)
diff --git a/configure.ac b/configure.ac
index
Signed-off-by: Helmut Schaa
---
This is untested since I don't use/have RHEL or Fedora.
I'd be glad if anyone could give that change a try.
Thanks,
Helmut
rhel/openvswitch-fedora.spec.in | 3 +++
rhel/openvswitch.spec.in| 1 +
2 files changed, 4 insertions(+)
diff --git a/rhel/openvsw
This allows to get rid of some special segment handling to allow
distributed registering of vlog modules.
Instead use a global list and vlog module constructor functions to
build up the list. That means vlog modules reside within the
compilation unit they are defined in and can be iterated upon
by
Currently openvswitch builds all libraries static only. However,
libopenvswitch is linked into nearly all openvswitch executables
making it hardly possible to run openvswitch on embedded devices
(for example running OpenWrt).
Convert openvswitch to use libtool for building its internal libs.
This
This series converts OVS to use shared libs for its internal stuff instead
of linking everything static.
Helmut Schaa (6):
compiler: Add OVS_CONSTRUCTOR to mark functions as init functions
vlog: Use OVS_CONSTRUCTOR for vlog initialization
coverage: Use OVS_CONSTRUCTOR to initialize the cover
Functions marked with OVS_CONSTRUCTOR are called unconditionally
before main.
Tested with GCC. Untested with MSVC.
Signed-off-by: Helmut Schaa
---
lib/compiler.h | 13 +
1 file changed, 13 insertions(+)
diff --git a/lib/compiler.h b/lib/compiler.h
index 0dbacbf..daca32d 100644
---
> From: Jesse Gross
...
> Out of curiosity, did you try using crc32q? OVS data structures are
> already aligned to 8 bytes. It would also be interesting to know if a
> parallelized implementation is worthwhile, although my guess is that
> the OVS flow key is not quite long enough.
My thoughts exac
1 - 100 of 102 matches
Mail list logo
