ginal code
--
David Marchand
Darrell Ball (1):
conntrack: Fix FTP seq_skew boundary adjustments.
David Marchand (2):
conntrack: fix tcp seq adjustments when mangling commands
conntrack: fix expectations nat configuration for ftp+DNAT
Vagrantfile | 9 +-
Vagrantfile-FreeBSD |
natting.
The tests are updated so that some ftp+NAT tests send multiple port
commands or other similar commands for a single control connection.
Wget is not able to do this, so switch to lftp.
Co-authored-by: Darrell Ball
Signed-off-by: Darrell Ball
Signed-off-by: David Marchand
---
Changelog
:
0x0040: b715 3232 3720 456e 7465 7269 6e67 2050 ..227.Entering.P
0x0050: 6173 7369 7665 204d 6f64 6520 2831 302c assive.Mode.(10,
0x0060: 312c 312c 3230 302c 3234 312c 3134 3429 1,1,200,241,144)
0x0070: 2e0d 2e ...
Fixes: bd5e81a0e596 ("Userspace Dat
On Tue, Jan 15, 2019 at 8:20 AM Darrell Ball wrote:
>
> On 1/14/19, 4:49 AM, "ovs-dev-boun...@openvswitch.org on behalf of David
> Marchand" david.march...@redhat.com> wrote:
>
> On Mon, Jan 14, 2019 at 7:32 AM Darrell Ball wrote:
>
> > 4/ I adde
- sticked closer to the original code
David Marchand (2):
conntrack: fix tcp seq adjustments when mangling commands
conntrack: fix expectations nat configuration for ftp+DNAT
Vagrantfile | 9 ---
Vagrantfile-FreeBSD | 2 +-
lib/conntrack.c | 64
number by the connection current offset, then prepare for the next
packets by setting an accumulated offset in the ct object.
The tests are updated so that ftp+NAT checks send multiple commands in a
single tcp command connection: wget is not able to do this, so switch to
lftp.
Signed-off-by: David
to reflect they are
dealing with SNAT.
Signed-off-by: David Marchand
---
lib/conntrack.c | 12 ++--
tests/system-traffic.at | 48
2 files changed, 34 insertions(+), 26 deletions(-)
diff --git a/lib/conntrack.c b/lib/conntrack.c
index
- sticked closer to the original code
David Marchand (2):
conntrack: fix tcp seq adjustments when mangling commands
conntrack: fix expectations nat configuration for ftp+DNAT
Vagrantfile | 9 ---
Vagrantfile-FreeBSD | 2 +-
lib/conntrack.c | 64
to reflect they are
dealing with SNAT.
Signed-off-by: David Marchand
---
lib/conntrack.c | 12 ++--
tests/system-traffic.at | 48
2 files changed, 34 insertions(+), 26 deletions(-)
diff --git a/lib/conntrack.c b/lib/conntrack.c
index
number by the connection current offset, then prepare for the next
packets by setting an accumulated offset in the ct object.
The tests are updated so that ftp+NAT checks send multiple commands in a
single tcp command connection: wget is not able to do this, so switch to
lftp.
Signed-off-by: David
On Thu, Dec 20, 2018 at 8:27 PM David Marchand
wrote:
> Some small fixes to make the ftp alg behave a little better.
> I tried to come up with better tests but I am not too sure if we want to
> have a dependency on yet another external tool for it (lftp).
>
Sorry about th
Rather than letting the test framework kill any remaining ofctl monitor,
ask and wait for it to gracefully exit before looking at the log file.
This solves random failures of tests 29, 30 and 50.
Signed-off-by: David Marchand
---
tests/system-traffic.at | 10 --
1 file changed, 8
ion
> - "DPDK 17.11.0"
> + "DPDK 18.08.0"
>
> At this point you can use ovs-vsctl to set up bridges and other Open
> vSwitch
> features. Seeing as we've configured the DPDK datapath, we will use
> DPDK-type
>
>
Just spotted this.
Do you want a patch for this
mangle operations and prepare for
the next packets by setting an accumulated offset.
Signed-off-by: David Marchand
---
lib/conntrack.c | 25 -
1 file changed, 12 insertions(+), 13 deletions(-)
diff --git a/lib/conntrack.c b/lib/conntrack.c
index 974f985..d08d0ea 100644
The ftp alg relies on the attached nat information to the current
connection to trigger the nat operation while it should take the
information from the rule being evaluated.
Signed-off-by: David Marchand
---
lib/conntrack.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git
A lot of small fixes to make the ftp alg behave a little better.
I tried to come up with better tests but I am not too sure if we want to
have a dependency on yet another external tool for it.
--
David Marchand
David Marchand (6):
conntrack: fix multiple tcp seq adjustments
conntrack: apply
it.
Signed-off-by: David Marchand
---
lib/conntrack-private.h | 5 ++---
lib/conntrack.c | 37 +
2 files changed, 23 insertions(+), 19 deletions(-)
diff --git a/lib/conntrack-private.h b/lib/conntrack-private.h
index a344801..6afc6da 100644
--- a/lib
and
positive updates.
Signed-off-by: David Marchand
---
Vagrantfile | 9 ---
Vagrantfile-FreeBSD | 2 +-
tests/system-traffic.at | 64 +
3 files changed, 45 insertions(+), 30 deletions(-)
diff --git a/Vagrantfile b/Vagrantfile
index
the new one and substitute
in one pass like what is done for ipv6.
Signed-off-by: David Marchand
---
lib/conntrack.c | 54 --
1 file changed, 28 insertions(+), 26 deletions(-)
diff --git a/lib/conntrack.c b/lib/conntrack.c
index 96ed8b3..abed3f1
In passive mode, a ftp alg only touches the payload when doing DNAT to
the server, so change the test accordingly and update the active mode
checks test titles to reflect what they test.
Signed-off-by: David Marchand
---
tests/system-traffic.at | 54
ised this is seen with 18.11.
This message has been here for quite some time, I suspect something changed
in the logs but I did not find at a first glance.
--
David Marchand
___
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
) {
-tcp_seq = ec->seq_skew - (UINT32_MAX - tcp_seq);
-} else if ((ec->seq_skew < 0) && (tcp_seq < -ec->seq_skew)) {
-/* Should not be possible; will be marked invalid. */
-tcp_seq = 0;
-} else {
-tcp_seq += ec->
On Thu, Jan 10, 2019 at 10:32 AM Lam, Tiago wrote:
> On 10/01/2019 09:12, David Marchand wrote:
> > On Thu, Jan 10, 2019 at 10:11 AM David Marchand
> > mailto:david.march...@redhat.com>> wrote:
> >
> > This part triggers build (non fatal o_O) warnin
On Thu, Jan 10, 2019 at 10:11 AM David Marchand
wrote:
> This part triggers build (non fatal o_O) warnings on fedora 28.
>
Oops, important.
This is when building without dpdk support.
--
David Marchand
___
dev mailing list
d...@openvswit
~~
In file included from ./lib/string.h:20,
from /vagrant/lib/packets.h:23,
from /vagrant/lib/packets.c:18:
/vagrant/lib/packets.c: In function ‘packet_crc32c’:
/usr/include/string.h:42:14: note: in a call to function ‘memcpy’ declared
here
extern void *memcpy (void *__restrict __dest, const void *__restrict __src,
^~
--
David Marchand
___
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
On Tue, Jan 8, 2019 at 5:49 AM Darrell Ball wrote:
> On Thu, Dec 20, 2018 at 5:33 AM David Marchand
> wrote:
>
>> The ftp alg deals with packets in two ways for the command connection:
>> either they are inspected and can be mangled when nat is enabled
>> (CT_FTP_CT
On Fri, Jan 11, 2019 at 1:49 PM Ian Stokes wrote:
> On 1/11/2019 12:29 PM, Ferruh Yigit wrote:
> > On 1/11/2019 11:20 AM, Andrew Rybchenko wrote:
> >> On 1/11/19 2:11 PM, David Marchand wrote:
> >>> Indeed, the documentation tells it is experimental, but the
>
its form did not
change afaics.
I am not sure it qualifies as experimental anymore.
CC Ferruh, Andrew and Thomas.
--
David Marchand
___
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
rted to 18.11, I don't think the
> comment is needed.
>
I was going to reply that I intended to ask for a backport in 18.11 at
least :-)
--
David Marchand
___
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
- squashed code changes and tests together
- sticked closer to the original code
--
David Marchand
David Marchand (2):
conntrack: fix tcp seq adjustments when mangling commands
conntrack: fix expectations nat configuration for ftp+DNAT
Vagrantfile | 9 ++-
Vagrantfile-FreeBSD
number by the connection current offset, then prepare for the next
packets by setting an accumulated offset in the ct object.
The tests are updated so that ftp+NAT checks send multiple commands in a
single tcp command connection: wget is not able to do this, so switch to
lftp.
Signed-off-by: David
-by: Darrell Ball
Signed-off-by: Darrell Ball
Signed-off-by: David Marchand
---
Changelog since v2:
- following Darrell comment, kept the existing passive snat test as is,
then added new dnat tests for passive and active modes
- updated missing snat tests titles
---
lib/conntrack.c
Hello,
On Wed, Jan 9, 2019 at 4:51 AM Darrell Ball wrote:
> On Tue, Jan 8, 2019 at 2:57 PM Darrell Ball wrote:
>
>> On Tue, Jan 8, 2019 at 2:13 AM David Marchand
>> wrote:
>>
>>>
>>> Let me take an example with the test CHECK_FTP_SNAT_POST_RECIR
Hello and happy new year to all :-)
On Tue, Jan 8, 2019 at 5:24 AM Darrell Ball wrote:
> Thanks for the fix David.
>
> On Thu, Dec 20, 2018 at 5:33 AM David Marchand
> wrote:
>
>> When configuring the nat part of an expectation, care must be taken to
>> lo
On Fri, Jan 11, 2019 at 8:05 PM Darrell Ball wrote:
> On Thu, Jan 10, 2019 at 12:58 PM Darrell Ball wrote:
>
>> On Thu, Jan 10, 2019 at 1:03 AM David Marchand
>> wrote:
>>
>>> Hello,
>>>
>>> Just wondering, can't we rely integer promotion
On Mon, Jan 14, 2019 at 7:32 AM Darrell Ball wrote:
> Thanks for the patch
>
> On Wed, Jan 9, 2019 at 7:33 AM David Marchand
> wrote:
>
> The ftp alg deals with packets in two ways for the command connection:
>> either they are inspected and can be man
orig tuple seqadj FAILED (
> system-traffic.at:4515)
>
Argh, indeed, I guess patch 2 should come first, will check.
--
David Marchand
___
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
On Wed, Dec 19, 2018 at 10:26 AM Darrell Ball wrote:
>
> On Wed, Dec 19, 2018 at 1:10 AM David Marchand
> wrote:
>
>>
>> If you look at the postrecirc seqadj test, you can see the following
>> rules:
>>
>> table=0 ip, action=ct(table=1)
>
On Wed, Dec 19, 2018 at 9:41 AM Darrell Ball wrote:
> Thanks for working on this David.
>
> On Sat, Dec 15, 2018 at 9:39 AM David Marchand
> wrote:
>
>> The ftp alg relies on the attached nat information to the current
>> connection to trigger the nat ope
p.
Hence, should not v6_addr_str be declared as INET6_ADDRSTRLEN bytes long ?
--
David Marchand
___
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
original patch, so I'd prefer to
have a Reported-by tag rather than a Sob.
Looked at the patch, nothing wrong to me in it, you can add my
Reviewed-by: David Marchand
Thanks.
--
David Marchand
___
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
).
A consequence of this is that before a device is connected to the vhost
port, no rxq is polled at all.
Signed-off-by: David Marchand
---
We tried to lower the number of rebalances but we don't have a
satisfying solution at the moment, so this patch rebalances on each
update.
---
lib/dpif-netdev.c
gt; --
> 2.20.1
>
>
I have been using git-pw for quite some time for ovs and dpdk.
I personnally use an API token for auth, but apart from that it looks good.
Reviewed-by: David Marchand
--
David Marchand
___
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
Hello Ilya,
On Mon, Apr 8, 2019 at 10:27 AM Ilya Maximets
wrote:
> On 04.04.2019 22:49, David Marchand wrote:
> > We tried to lower the number of rebalances but we don't have a
> > satisfying solution at the moment, so this patch rebalances on each
> > update.
&
Fix merge conflict introduced when backporting the original commit
253e4dc0683b ("conntrack: fix tcp seq adjustments when mangling
commands.")
Fixes: addbb52274ae ("conntrack: fix tcp seq adjustments when mangling
commands.")
Signed-off-by: David Marchand
---
Vagrantfile
destroy(>server);
> ovsdb_idl_db_destroy(>data);
> json_destroy(idl->request_id);
> +if (idl->remote) {
> +free(idl->remote);
> +}
>
Testing for null is unnecessary.
free(idl);
> }
> }
>
--
David Mar
On Wed, Mar 6, 2019 at 7:23 PM Han Zhou wrote:
> From: Han Zhou
>
> Reported by Address Sanitizer.
>
> Fixes: 5e07b8f93f03 ("ovsdb-idl: New function
> ovsdb_idl_create_unconnected().")
> Signed-off-by: Han Zhou
> ---
>
> Notes:
> v1->v2: addres
submit backported versions.
>
Thanks Ben.
Darrell, can these two fixes be backported to 2.9 ?
I can send the backports if you don't have the bandwidth.
--
David Marchand
___
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
avoid a race condition on monitor log. This patch adds
> back this line but omits the log file because this test doesn't
> dpends on the log file.
>
> Fixes: e8833217914f9c071c49 ("system-traffic.at: avoid a race condition
> on monitor log")
> CC: David Marchand
> Signed
and FTP.")
> Signed-off-by: Darrell Ball
> ---
>
> v2: s/IPV6_SCAN_LEN/INET6_ADDRSTRLEN/ in
> char v6_addr_str[IPV6_SCAN_LEN] = {0};
> per review by David Marchand.
>
> lib/conntrack.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git
le_string_is_equal(ifname, dev->vhost_id)) {
> +uint32_t qp_num = NR_QUEUE;
> +
> +/* Restore the number of queue pairs to default. */
> +if (dev->requested_n_txq != qp_num
> +|| dev->requested_n_rxq != qp_num) {
> +dev->requested_n_
On Thu, Apr 11, 2019 at 4:02 PM David Marchand
wrote:
> We currently poll all available queues based on the max queue count
> exchanged with the vhost peer and rely on the vhost library in DPDK to
> check the vring status beneath.
> This can lead to some overhead when we have a l
...
# While rebooting the vm
total: 66, enabled: 4
total: 66, enabled: 2
...
total: 66, enabled: 66
...
# After shutting down the vm
total: 66, enabled: 66
total: 66, enabled: 2
Signed-off-by: David Marchand
---
lib/dpif-netdev.c | 27 ++
hange will do nothing but have openvswitch complain
that the vhost device is unknown:
dpdk|INFO|VHOST_CONFIG: vhost peer closed
dpdk|ERR|VHOST_CONFIG: (0) device not found.
dpdk|INFO|VHOST_CONFIG: vhost peer closed
dpdk|ERR|VHOST_CONFIG: (1) device not found.
Signed-off-by: David Marchan
patch reduced the observed cost of polling disabled queues,
so the only cost is memory.
Signed-off-by: David Marchand
---
lib/netdev-dpdk.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/lib/netdev-dpdk.c b/lib/netdev-dpdk.c
index 9088ff4..8a9723e 100644
--- a/lib/netdev
On Wed, Apr 10, 2019 at 4:24 PM Ilya Maximets
wrote:
> On 10.04.2019 17:10, David Marchand wrote:
> > Incorporated this change, and it works fine, I get similar numbers
> with/without a bitmap, so I dropped the bitmap.
> > I would say we are only missing some information in
>
On Thu, Apr 18, 2019 at 1:51 PM Ilya Maximets
wrote:
>
>
> On 18.04.2019 14:26, David Marchand wrote:
> > On Wed, Apr 17, 2019 at 4:27 PM Kevin Traynor <mailto:ktray...@redhat.com>> wrote:
> >
> > On 16/04/2019 10:45, David Marchand wrote:
> >
On Wed, Apr 17, 2019 at 12:21 PM Ilya Maximets
wrote:
> On 17.04.2019 11:37, David Marchand wrote:
> > On Tue, Apr 16, 2019 at 4:01 PM Ilya Maximets <mailto:i.maxim...@samsung.com>> wrote:
> > > +LIST_FOR_EACH (dev, list_node, _list) {
> >
On Wed, Apr 17, 2019 at 4:27 PM Kevin Traynor wrote:
> On 16/04/2019 10:45, David Marchand wrote:
>
> > Note: this patch requires a fix for the vhost library submitted here:
> > http://patchwork.dpdk.org/patch/52680/
> >
> > Without it, this change will do nothing
me rule is applicable to the comments in the code, but this is
> documented in coding-style.
>
Another thing different from my dpdk habits.
Need to focus when hacking ovs :-).
On 16.04.2019 12:45, David Marchand wrote:
> > diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c
> > index 4d6
On Tue, Apr 16, 2019 at 4:01 PM Ilya Maximets
wrote:
> On 16.04.2019 12:45, David Marchand wrote:
> > Rather than poll all disabled queues and waste some memory for vms that
> > have been shutdown, we can reconfigure when receiving a destroy
> > connection notification f
On Tue, Apr 16, 2019 at 3:51 PM Ilya Maximets
wrote:
> On 16.04.2019 12:45, David Marchand wrote:
> > At the moment, a malicious guest might negotiate VIRTIO_NET_F_MQ and
> > !VIRTIO_NET_F_MQ in a loop which would be seen as qp_num going from 1 to
> > n and n to 1 con
On Fri, Apr 12, 2019 at 2:12 PM Ilya Maximets
wrote:
> On 11.04.2019 17:00, David Marchand wrote:
> > We currently poll all available queues based on the max queue count
> > exchanged with the vhost peer and rely on the vhost library in DPDK to
> > check the vring status
On Fri, Apr 12, 2019 at 12:00 PM Ilya Maximets
wrote:
> On 11.04.2019 17:13, David Marchand wrote:
> >
> >
> > On Thu, Apr 11, 2019 at 4:02 PM David Marchand <
> david.march...@redhat.com <mailto:david.march...@redhat.com>> wrote:
> >
> >
Hello,
On Wed, May 15, 2019 at 12:04 PM Ilya Maximets
wrote:
> On 25.04.2019 18:22, David Marchand wrote:
> > We currently poll all available queues based on the max queue count
> > exchanged with the vhost peer and rely on the vhost library in DPDK to
> > check the
On Tue, Jun 4, 2019 at 11:29 AM David Marchand
wrote:
> Following a rework of dpdk network structures names [1], update the
> concerned parts.
>
> Ran Olivier script:
> sh prefix-net-rte.sh $(find -name "*dpdk*.c")
> sh prefix-net-rte.sh $(find -name "*dpdk*.
quot;
old=RTE_ETHER_TYPE_IPv6
new=RTE_ETHER_TYPE_IPV6
git grep -lw $old | xargs sed -i -e "s/\<$old\>/$new/g"
1: http://mails.dpdk.org/archives/dev/2019-May/132612.html
2: https://git.dpdk.org/dpdk/commit/?id=0c9da7555da8
Signed-off-by: David Marchand
---
Small note on checkpa
erstand everyone is busy with their own stuff.
Thanks.
--
David Marchand
___
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
On Thu, Jun 6, 2019 at 1:25 PM Ian Stokes wrote:
> On 6/4/2019 12:14 PM, David Marchand wrote:
> > On Tue, Jun 4, 2019 at 11:29 AM David Marchand
> > mailto:david.march...@redhat.com>> wrote:
> >
> > Following a rework of dpdk network structures names [1], u
Hello guys,
On Thu, May 23, 2019 at 4:27 PM David Marchand
wrote:
> We have been testing the rebalance code in different situations while
> having traffic going through OVS.
> Those tests have shown that part of the observed packets losses is due to
> some time wasted in signa
to this
reference (to save some disk),
- else, any other string which is understood as an official release.
This triggers a tarball download on dpdk.org.
Signed-off-by: David Marchand
---
.travis/linux-build.sh | 13 -
1 file changed, 8 insertions(+), 5 deletions(-)
diff --git
Rather than patch the dpdk makefile and a template config file, we can
pass the -fPIC flag via EXTRA_CFLAGS.
This is more reliable than expecting the dpdk file names to be kept
unchanged.
Signed-off-by: David Marchand
---
This patch applies on top of Ilya proposed change [1].
1: https
This section is about checking dpdk, we can focus on just its version.
Then update the version to 18.11.1 in the commands output.
Fixes: 03f3f9c0faf8 ("dpdk: Update to use DPDK 18.11.")
Fixes: b5355b0d6e66 ("dpdk: Use DPDK 18.11.1 release.")
Signed-off-by: David Marchand
On Tue, Jun 11, 2019 at 11:31 AM Ilya Maximets
wrote:
> On 11.06.2019 11:45, David Marchand wrote:
> > I noticed that OVS CI [1] patches the dpdk sources to force some
> inlining parameters and get kni and igb_uio to build fine.
> >
> > Looking at it in dpd
ARY_PATH=$LD_LIBRARY_PATH:$(pwd)/$TARGET/lib
> fi
> +
> +# Disable building DPDK kernel modules. Not needed for OVS build or
> tests.
> + sed -i '/CONFIG_RTE_EAL_IGB_UIO=y/s/=y/=n/' config/common_linuxapp
> +sed -i '/CONFIG_RTE_KNI_KMOD=y/s/=y/=n/' config/common_l
onfig
>
We could also disable the librte_kni but we are fine with just disabling
the kmods.
+
> +make -j4 CC=gcc
> echo "Installed DPDK source in $(pwd)"
> cd ..
> }
> @@ -97,7 +98,7 @@ function configure_ovs()
> ./boot.sh && ./configu
On Tue, Jun 11, 2019 at 6:30 PM Ilya Maximets
wrote:
> On 11.06.2019 19:10, David Marchand wrote:
> >
> >
> > On Tue, Jun 11, 2019 at 5:36 PM Ilya Maximets <mailto:i.maxim...@samsung.com>> wrote:
> >
> > On 11.06.2019 18:21, Ian Stokes wrote:
&g
On Tue, Jun 11, 2019 at 5:36 PM Ilya Maximets
wrote:
> On 11.06.2019 18:21, Ian Stokes wrote:
> > On 6/11/2019 3:40 PM, Aaron Conole wrote:
> >> Ian Stokes writes:
> >>
> >>> On 6/10/2019 3:57 PM, Ian Stokes wrote:
> >&
On Tue, Jun 18, 2019 at 5:24 PM Ilya Maximets
wrote:
> On 12.06.2019 12:21, David Marchand wrote:
> > Rework the build script so that we can pass branches and tags.
> >
> > With this, DPDK_VER can be passed as:
> > - a string starting with refs/ which is un
to this
reference (to save some disk),
- else, any other string which is understood as an official release.
This triggers a tarball download on dpdk.org.
Signed-off-by: David Marchand
---
Changelog since v1:
- removed (now unneeded) directory renames
- added a "git log" so th
Rather than patch the dpdk makefile and a template config file, we can
pass the -fPIC flag via EXTRA_CFLAGS.
This is more reliable than expecting the dpdk file names to be kept
unchanged.
Signed-off-by: David Marchand
---
Changelog since v1:
- rebased on master
---
.travis/linux-build.sh | 4
On Mon, Jun 24, 2019 at 9:14 PM Ian Stokes wrote:
> On 5/23/2019 3:23 PM, David Marchand wrote:
> > Looking at the reload flag only every 1024 loops can be a long time
> > under load, since we might be handling 32 packets per polled rxq, per
> > iteration, which means up to
On Tue, Jun 25, 2019 at 3:40 PM Ilya Maximets
wrote:
> On 23.05.2019 17:23, David Marchand wrote:
> > pmd reloads are currently serialised in each steps calling
> > reload_affected_pmds.
> > Any pmd processing packets, waiting on a mutex etc... will make other
> > p
On Tue, Jun 25, 2019 at 3:47 PM Ilya Maximets
wrote:
> On 23.05.2019 17:23, David Marchand wrote:
> > When swapping queues from a pmd thread to another (q0 polled by pmd0/q1
> > polled by pmd1 -> q1 polled by pmd0/q0 polled by pmd1), the current
> > "Step 5&quo
On Wed, Jun 12, 2019 at 2:55 PM Ian Stokes wrote:
> On 6/12/2019 12:49 PM, David Marchand wrote:
> > This section is about checking dpdk, we can focus on just its version.
> > Then update the version to 18.11.1 in the commands output.
> >
> > Fixes: 03f3f9c0faf8 (&quo
On Mon, Jun 10, 2019 at 4:58 PM Ian Stokes wrote:
> On 6/6/2019 12:36 PM, David Marchand wrote:
> >
> >
> > On Thu, Jun 6, 2019 at 1:25 PM Ian Stokes > <mailto:ian.sto...@intel.com>> wrote:
> > I'll be applying this to dpdk-latest and dpdk-hwol branch
On Wed, Jun 19, 2019 at 1:22 PM Ilya Maximets
wrote:
> On 19.06.2019 10:26, David Marchand wrote:
> > Rework the build script so that we can pass branches and tags.
> >
> > With this, DPDK_VER can be passed as:
> > - a string starting with refs/ which is un
On Thu, Jun 20, 2019 at 1:57 PM Stokes, Ian wrote:
> > > -Original Message-
> > > From: ovs-dev-boun...@openvswitch.org [mailto:ovs-dev-
> > > boun...@openvswitch.org] On Behalf Of Stokes, Ian
> > > Sent: Thursday, June 20, 2019 11:53 AM
>
On Wed, Jun 19, 2019 at 3:40 PM Ian Stokes wrote:
> On 5/23/2019 3:23 PM, David Marchand wrote:
> > When swapping queues from a pmd thread to another (q0 polled by pmd0/q1
> > polled by pmd1 -> q1 polled by pmd0/q0 polled by pmd1), the current
> > "Step 5&quo
dpdk-socket-limit' to limit amount of
> >> hugepage memory that can be used by DPDK.
> >> * Add support for vHost Post-copy Live Migration (experimental).
> >> - * OVS validated with DPDK 18.11.1 which is recommended to be used.
> >> + * DPDK:
> >
> > Just spotted that the additional * DPDK above is unneeded, will spin a
> v2 if there any other comments.
> >
> > Ian
> >> + - DPDK 18.11.2 is the new minimal supported version.
> >> + - DPDK 18.11.1 and lower is no longer supported.
>
> Not a strong opinion, but maybe this should be part of David's patch?
> i.e. for this patch we'll update "validated" version with 18.11.2,
> and David will bump the "minimal supported" version in his patch along
> with the actual changes to callback usage. This looks more consistent.
> This will also allow us to not have strong 18.11.2 requirement (only
> recommendation) for 2.11 branch where we'll not backport "enabled vhost
> queues" patch-set.
>
> What do you think?
>
Sounds good this way yes.
--
David Marchand
___
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
Looking at the reload flag only every 1024 loops can be a long time
under load, since we might be handling 32 packets per iteration, which
means 32k packets.
Look at the flag every loop, no major performance impact seen.
Signed-off-by: David Marchand
---
lib/dpif-netdev.c | 10 +-
1
void sleeping an
undeterministic amount of time.
Signed-off-by: David Marchand
---
lib/dpif-netdev.c | 47 ++-
1 file changed, 42 insertions(+), 5 deletions(-)
diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c
index 4febd8b..254a50e 100644
--- a/lib/dpif-netde
count.
The memory order on this atomic is rel-acq to have an explicit
synchronisation between the pmd threads and the control thread.
Signed-off-by: David Marchand
---
lib/dpif-netdev.c | 50 +-
1 file changed, 37 insertions(+), 13 deletions
Hello Ilya,
On Mon, May 6, 2019 at 5:37 PM Ilya Maximets wrote:
> On 30.04.2019 15:17, David Marchand wrote:
> > No need for a latch here since we don't have to wait.
> > A simple boolean flag is enough.
> >
> > Fixes: e4cfed38b159 ("dpif-netdev: Add poll-mode
On Tue, May 14, 2019 at 11:53 AM Ilya Maximets
wrote:
>
> On 14.05.2019 10:38, David Marchand wrote:
> > Hello Ilya,
> > On Mon, May 6, 2019 at 5:37 PM Ilya Maximets <mailto:i.maxim...@samsung.com>> wrote:
> >
> > On 30.04.2019 15:17, David Marchand wr
On Wed, May 22, 2019 at 3:26 PM Kevin Traynor wrote:
> On 14/05/2019 17:33, David Marchand wrote:
> > Looking at the reload flag only every 1024 loops can be a long time
> > under load, since we might be handling 32 packets per iteration, which
> > means 32k packets.
>
&
On Wed, May 22, 2019 at 3:10 PM Kevin Traynor wrote:
> On 14/05/2019 17:33, David Marchand wrote:
> > No need for a latch here since we don't have to wait.
> > A simple boolean flag is enough.
> >
> > The memory order on the reload flag is changed to re
between threads
in patch 1 and patch 2
--
David Marchand
David Marchand (5):
dpif-netdev: Convert exit latch to flag.
dpif-netdev: Trigger parallel pmd reloads.
dpif-netdev: Do not sleep when swapping queues.
dpif-netdev: Only reload static tx qid when needed.
dpif-netdev: Catch reloads
;dpif-netdev: Add poll-mode-device thread.")
Signed-off-by: David Marchand
Acked-by: Eelco Chaudron
---
lib/dpif-netdev.c | 14 ++
1 file changed, 6 insertions(+), 8 deletions(-)
---
Changelog since RFC v2:
- removed now unused latch.h inclusion
Changelog since RFC v1:
- added memor
Looking at the reload flag only every 1024 loops can be a long time
under load, since we might be handling 32 packets per polled rxq, per
iteration, which means up to poll_cnt * 32 * 1024 packets.
Look at the flag every loop, no major performance impact seen.
Signed-off-by: David Marchand
Acked
1 - 100 of 967 matches
Mail list logo
