Hi Ilya et al
I can do that, but it may take a few days to get to it. Thanks for the
detailed analysis, I'll try to get my head around that too and contribute
to a fix.
Tony
On Thu, Aug 12, 2021 at 10:51 AM Ilya Maximets wrote:
> Hi.
>
> It's been some time since the last email in this
On Fri, Aug 13, 2021 at 4:07 AM Ilya Maximets wrote:
>
> On 8/12/21 8:33 AM, Sriharsha Basavapatna via dev wrote:
> > In netdev_offload_dpdk_flow_create() when an offload request fails,
> > dump_flow() is called to log a warning message. The 's_tnl' string
> > in flow_patterns gets initialized in
Hi,
I provided an error clue. The kernel version I used at compile time was
4.9.188-35.el7.x86_ 64. Unfortunately, I have encountered this error and
haven't solved it yet. If you are interested in this problem, I can provide
more information.
| |
mythosmonkeyking
|
|
If there are snat entries on the router, and some logical_ip are set to
network instead of an IP address then given SNAT is masquerade. In such
case ct_snat action is used in lr_in_unsnat table to ensure that the
packet is matched against conntrack and destination IP is replaced with
one from
On 8/12/21 8:33 AM, Sriharsha Basavapatna via dev wrote:
> In netdev_offload_dpdk_flow_create() when an offload request fails,
> dump_flow() is called to log a warning message. The 's_tnl' string
> in flow_patterns gets initialized in vport_to_rte_tunnel() conditionally
> via ds_put_format(). If
If there are snat entries on the router, and some logical_ip are set to
network instead of an IP address then given SNAT is masquerade. In such
case ct_snat action is used in lr_in_unsnat table to ensure that the
packet is matched against conntrack and destination IP is replaced with
one from
+ovs-discuss for a wider audience.
Numan
On Thu, Aug 12, 2021 at 10:23 AM Frode Nordahl
wrote:
>
> On Mon, Aug 9, 2021 at 8:33 PM Mark Michelson wrote:
> >
> > Hi everyone,
> >
> > If you've paid attention to ovn commits lately, you'll know that we've
> > been putting in a lot of performance
Hi,
the patch looks good to me, there's only one remark.
Timothy Redaelli writes:
> Currently, there are some patches with the tags wrongly written (with
> space instead of dash ) and this may prevent some automatic system or CI
> to detect them correctly.
>
> This commit adds a check in
This has little effect on performance in the current benchmark, but it
seems like the cleanest approach to me.
Signed-off-by: Ben Pfaff
---
northd/ovn_northd.dl | 38 ++
1 file changed, 22 insertions(+), 16 deletions(-)
diff --git a/northd/ovn_northd.dl
With the benchmark at
https://mail.openvswitch.org/pipermail/ovs-dev/2021-July/385333.html,
this reduces memory consumption by 300 MB and elapsed time by a few
seconds.
Signed-off-by: Ben Pfaff
---
northd/ipam.dl | 2 +-
northd/lswitch.dl| 36 ++--
The stage_hint only shows 32 bits of the uuid, so it's cheaper to omit
the rest for the internal representation. Also, this is just a hint, so
we might as well use zero to mean None and save the cost of the Option
wrapper.
With the benchmark at
It was possible for these rules to generate multiple Flow records that
differed only in their stage_hint. This caused a lot of duplication for
the load balancer benchmark.
With the benchmark at
https://mail.openvswitch.org/pipermail/ovs-dev/2021-July/385333.html,
this reduces memory consumption
With the benchmark at
https://mail.openvswitch.org/pipermail/ovs-dev/2021-July/385333.html,
this reduces memory consumption from 72 GB to 66 GB and elapsed time by
a few seconds.
Signed-off-by: Ben Pfaff
---
northd/ovn_northd.dl | 140 +--
1 file changed,
Apologies for repost; I forgot to tag this as "ovn" in the first
posting.
For the benchmark at
https://mail.openvswitch.org/pipermail/ovs-dev/2021-July/385333.html,
this reduces memory consumption from 115 GB to 17 GB and elapsed time
from 19 minutes to 5 minutes.
I think there's headroom for
Hi,
thanks for working on this
we...@ucloud.cn writes:
> From: wenxu
>
> The ipf collect original fragment packets and reass a new pkt
> to do the conntrack logic. After finsh the conntrack things
> copy the ct meta info to each orignal packet and modify the
> l4 header in the first fragment.
On Thu, Aug 12, 2021 at 6:18 AM Mark Gray wrote:
>
> On 11/08/2021 21:00, Mark Gray wrote:
> > Conjunctive flows require a unique integer as an identifier.
> > In a future patch, we will use conjunctive flows for LB
> > hairpin flows. This patch generates a unique integer id for each
> > LB.
> >
On Tue, Aug 10, 2021 at 05:17:56PM -0700, Sergey Madaminov wrote:
> Checking if '__has_builtin' is defined and then defining OVS_PREFETCH to
> be '__builtin_prefetch' if it is available. To preserve backwards
> compatibility, the previous way to define OVS_PREFETCH macro is also
> there.
Thanks
On Wed, Aug 11, 2021 at 11:34 AM Priyankar Jain
wrote:
>
> Hi Numan,
>
> In case the patch looks good to you, can you please merge it in master?
Done. Applied to the main branch.
Thanks
Numan
>
> Thanks and Regards,
> Priyankar Jain
>
> On 02/08/21 2:02 pm, Priyankar Jain wrote:
> > Allow
On 12/08/2021 16:57, Harry van Haaren wrote:
> This patch fixes a memory leak in the commands for DPIF and MFEX
> get and set. In order to operate the commands require a pmd_list,
> which is currently not freed after it has been used. This issue
> was identified by a static analysis tool.
>
>
On 12/08/2021 16:57, Harry van Haaren wrote:
> This patch fixes a memory leak when the command
> "dpif-netdev/subtable-lookup-prio-set" is run, the pmd_list
> required to iterate the PMD threads was not being freed.
> This issue was identified by a static analysis tool.
>
> Fixes: 3d018c3e
we...@ucloud.cn writes:
> From: wenxu
>
> The ipf collect original fragment packets and reass a new pkt
> to do the conntrack logic. After finsh the conntrack things
> copy the ct meta info to each orignal packet and modify the
> l4 header in the first fragment. It should modify the ip src/
>
> -Original Message-
> From: Ilya Maximets
> Sent: Thursday, August 12, 2021 4:27 PM
> To: Van Haaren, Harry ; d...@openvswitch.org
> Cc: i.maxim...@ovn.org; amber.ku...@intel.com; Stokes, Ian
> ; Flavio Leitner
> Subject: Re: [ovs-dev] [PATCH] dpif: fix memory leak of pmd_list after
On 12/08/2021 16:27, Ilya Maximets wrote:
> On 8/12/21 5:11 PM, Harry van Haaren wrote:
>> This commit fixes a memory leak when a pmd_list is retrieved
>> from the sorted_poll_thread_list() function. Inside the function,
>> the pmd list is allocated, but it was not freed once no longer
>> required
This patch fixes a memory leak in the commands for DPIF and MFEX
get and set. In order to operate the commands require a pmd_list,
which is currently not freed after it has been used. This issue
was identified by a static analysis tool.
Fixes: 3d8f47bc ("dpif-netdev: Add command line and function
This patch fixes a memory leak when the command
"dpif-netdev/subtable-lookup-prio-set" is run, the pmd_list
required to iterate the PMD threads was not being freed.
This issue was identified by a static analysis tool.
Fixes: 3d018c3e ("dpif-netdev: Add subtable lookup prio set command.")
On 12/08/2021 16:11, Harry van Haaren wrote:
> This commit fixes a memory leak when a pmd_list is retrieved
> from the sorted_poll_thread_list() function. Inside the function,
> the pmd list is allocated, but it was not freed once no longer
> required for the command functionality. These leaks
Bleep bloop. Greetings Ben Pfaff, I am a robot and I have tried out your patch.
Thanks for your contribution.
I encountered some error that I wasn't expecting. See the details below.
Patch skipped due to previous failure.
Please check this out. If you feel there has been an error, please
Bleep bloop. Greetings Ben Pfaff, I am a robot and I have tried out your patch.
Thanks for your contribution.
I encountered some error that I wasn't expecting. See the details below.
Patch skipped due to previous failure.
Please check this out. If you feel there has been an error, please
Bleep bloop. Greetings Ben Pfaff, I am a robot and I have tried out your patch.
Thanks for your contribution.
I encountered some error that I wasn't expecting. See the details below.
Patch skipped due to previous failure.
Please check this out. If you feel there has been an error, please
Bleep bloop. Greetings Ben Pfaff, I am a robot and I have tried out your patch.
Thanks for your contribution.
I encountered some error that I wasn't expecting. See the details below.
Patch skipped due to previous failure.
Please check this out. If you feel there has been an error, please
Bleep bloop. Greetings Ben Pfaff, I am a robot and I have tried out your patch.
Thanks for your contribution.
I encountered some error that I wasn't expecting. See the details below.
git-am:
error: sha1 information is lacking or useless (northd/ovn_northd.dl).
error: could not build fake
Bleep bloop. Greetings Ben Pfaff, I am a robot and I have tried out your patch.
Thanks for your contribution.
I encountered some error that I wasn't expecting. See the details below.
Patch skipped due to previous failure.
Please check this out. If you feel there has been an error, please
Bleep bloop. Greetings Ben Pfaff, I am a robot and I have tried out your patch.
Thanks for your contribution.
I encountered some error that I wasn't expecting. See the details below.
Patch skipped due to previous failure.
Please check this out. If you feel there has been an error, please
On Thu, Aug 12, 2021 at 08:53:51AM -0700, Ben Pfaff wrote:
> For the benchmark at
> https://mail.openvswitch.org/pipermail/ovs-dev/2021-July/385333.html,
> this reduces memory consumption from 115 GB to 17 GB and elapsed time
> from 19 minutes to 5 minutes.
>
> I think there's headroom for more
From: Ben Pfaff
This has little effect on performance in the current benchmark, but it
seems like the cleanest approach to me.
Signed-off-by: Ben Pfaff
---
northd/ovn_northd.dl | 38 ++
1 file changed, 22 insertions(+), 16 deletions(-)
diff --git
From: Ben Pfaff
With the benchmark at
https://mail.openvswitch.org/pipermail/ovs-dev/2021-July/385333.html,
this reduces memory consumption by 300 MB and elapsed time by a few
seconds.
Signed-off-by: Ben Pfaff
---
northd/ipam.dl | 2 +-
northd/lswitch.dl| 36
From: Ben Pfaff
With the benchmark at
https://mail.openvswitch.org/pipermail/ovs-dev/2021-July/385333.html,
this reduces memory consumption from 72 GB to 66 GB and elapsed time by
a few seconds.
Signed-off-by: Ben Pfaff
Signed-off-by: Ben Pfaff
---
northd/ovn_northd.dl | 140
From: Ben Pfaff
It was possible for these rules to generate multiple Flow records that
differed only in their stage_hint. This caused a lot of duplication for
the load balancer benchmark.
With the benchmark at
https://mail.openvswitch.org/pipermail/ovs-dev/2021-July/385333.html,
this reduces
From: Ben Pfaff
The stage_hint only shows 32 bits of the uuid, so it's cheaper to omit
the rest for the internal representation. Also, this is just a hint, so
we might as well use zero to mean None and save the cost of the Option
wrapper.
With the benchmark at
For the benchmark at
https://mail.openvswitch.org/pipermail/ovs-dev/2021-July/385333.html,
this reduces memory consumption from 115 GB to 17 GB and elapsed time
from 19 minutes to 5 minutes.
I think there's headroom for more improvement, because Leonid has some
cleverness that I haven't been able
On 8/12/21 5:11 PM, Harry van Haaren wrote:
> This commit fixes a memory leak when a pmd_list is retrieved
> from the sorted_poll_thread_list() function. Inside the function,
> the pmd list is allocated, but it was not freed once no longer
> required for the command functionality. These leaks were
This commit fixes a memory leak when a pmd_list is retrieved
from the sorted_poll_thread_list() function. Inside the function,
the pmd list is allocated, but it was not freed once no longer
required for the command functionality. These leaks were found
by a static analysis tool.
Fixes: 3d8f47bc04
> The "name" parameter isn't optional so don't use brackets around it.
>
Thanks for the patch Cian, comment below.
> Fixes: 5c5c98cec21b ("docs/dpdk/bridge: Add miniflow extract section.")
> Signed-off-by: Cian Ferriter
> ---
> Documentation/topics/dpdk/bridge.rst | 4 ++--
> 1 file changed,
> Fixes: 3d8f47bc041a ("dpif-netdev: Add command line and function pointer for
> miniflow extract")
> Signed-off-by: Cian Ferriter
Hi Cian, as before a one line summary line in the message body would be
appreciated.
> ---
> lib/dpif-netdev-unixctl.man | 10 ++
> 1 file changed, 10
> Fixes: 9ff7cabfd78d ("dpif-netdev: add subtable-lookup-prio-get command.")
> Fixes: 3d018c3ea79d ("dpif-netdev: add subtable lookup prio set command.")
> Signed-off-by: Cian Ferriter
Thanks for the patch Cian. Can you add a one line summary to the commit message
above also?
Few comments
On Mon, Aug 9, 2021 at 8:33 PM Mark Michelson wrote:
>
> Hi everyone,
>
> If you've paid attention to ovn commits lately, you'll know that we've
> been putting in a lot of performance improvements. At Red Hat, this is
> because we've been engaging our scale team to try to help us to figure
> out
On Fri, Aug 6, 2021 at 9:37 AM Frode Nordahl
wrote:
>
> On Thu, Aug 5, 2021 at 9:19 PM Mark Michelson wrote:
> >
> > Hello,
> >
> > I was having a look at the calendar and realized that tomorrow (6
> > August) is our scheduled soft freeze date for OVN 21.09. Since I did a
> > poor job of
If process title is not initialized, the monitoring process can not
change it, so the current status of the monitored process is not visible
in the output of 'ps -aux' or other utilities.
Before:
$ ps -aux | grep nbctl
2926 ovn-nbctl --detach --monitor --log-file --unixctl=./ovn-nbctl.ctl
On 11/08/2021 21:00, Mark Gray wrote:
> Conjunctive flows require a unique integer as an identifier.
> In a future patch, we will use conjunctive flows for LB
> hairpin flows. This patch generates a unique integer id for each
> LB.
>
> Signed-off-by: Mark Gray
> ---
> controller/lflow.c
Hi,
Up to now, I still haven't found the reason for this problem. Who can help
me see it? Thank you very much.
| |
mythosmonkeyking
|
|
mythosmonkeyk...@163.com
|
签名由网易邮箱大师定制
On 8/12/2021 10:13,mythosmonkeyking wrote:
Hi,
When I compiled OVS, I found that ovs-ctl and other scripts were
In netdev_offload_dpdk_flow_create() when an offload request fails,
dump_flow() is called to log a warning message. The 's_tnl' string
in flow_patterns gets initialized in vport_to_rte_tunnel() conditionally
via ds_put_format(). If it is not initialized, it crashes later in
51 matches
Mail list logo