> -Original Message-
> From: wangyunjian
> Sent: Thursday, May 13, 2021 6:40
> To: dev@dpdk.org
> Cc: Matan Azrad ; Shahaf Shuler
> ; Slava Ovsiienko ;
> dingxiaoxi...@huawei.com; Yunjian Wang ;
> sta...@dpdk.org
> Subject: [dpdk-dev] [PATCH v2 2/2] net/mlx5: fix mem leak when a device
> c
> -Original Message-
> From: wangyunjian
> Sent: Thursday, May 13, 2021 6:39
> To: dev@dpdk.org
> Cc: Matan Azrad ; Shahaf Shuler
> ; Slava Ovsiienko ;
> dingxiaoxi...@huawei.com; Yunjian Wang ;
> sta...@dpdk.org
> Subject: [dpdk-dev] [PATCH v2 1/2] net/mlx4: fix mem leak when a device
> c
> -Original Message-
> From: David Marchand
> Sent: 2021年5月12日 23:20
> To: Wang, Yinan
> Cc: dev@dpdk.org; maxime.coque...@redhat.com;
> olivier.m...@6wind.com; f...@sysclose.org; i.maxim...@ovn.org; Xia,
> Chenbo ; Stokes, Ian ;
> sta...@dpdk.org; Jijiang Liu ; Yuanhan Liu
>
> Subject
Currently, the SVE code is compiled only when -march supports SVE
(e.g. '-march=armv8.2a+sve'), there maybe some problem[1] with this
approach.
The solution:
a. If the minimum instruction set support SVE then compiles it.
b. Else if the compiler support SVE then compiles it.
c. Otherwise don't com
Currently, the soc_kunpeng930 declares '-march=armv8.2-a+crypto+sve',
but some compiler doesn't recognize the march because it doesn't
support sve.
To solve this bug we use the following scheme:
1. Define 'march_base' tuple which defines support march, it should
arrange from lower to higher.
e.g.
This patch set contains two bugfixes for Kunpeng SVE compile.
---
v3:
* fix checkpatch error of 1/2 patch
* modify comments of 1/2 patch
v2:
* support select best -march parameter for kunpeng soc
Chengwen Feng (2):
config/arm: select most suitable -march for kunpeng soc
net/hns3: refactor SVE
Hi, Slava
Please see below.
> -邮件原件-
> 发件人: Slava Ovsiienko
> 发送时间: 2021年5月12日 19:08
> 收件人: Feifei Wang ; Matan Azrad
> ; Shahaf Shuler
> 抄送: dev@dpdk.org; nd ; sta...@dpdk.org; Ruifeng Wang
> ; nd
> 主题: RE: [PATCH v1 3/4] net/mlx5: fix rebuild bug for Memory Region cache
>
> Hi, Feif
On 2021/5/13 7:00, Honnappa Nagarahalli wrote:
>
>
>>
>> On Wed, May 12, 2021 at 2:01 PM Chengwen Feng
>> wrote:
>>>
>>> Currently, the soc_kunpeng930 declares '-march=armv8.2-a+crypto+sve',
>>> but some compiler doesn't recognize the march because it doesn't
>>> support sve.
>>
>> Agree with
This patch set contains two bugfixes for Kunpeng SVE compile.
---
v2:
* support select best -march parameter for kunpeng soc
Chengwen Feng (2):
config/arm: select best -march parameter for kunpeng soc
net/hns3: refactor SVE code compile method
config/arm/meson.build | 28
Currently, the SVE code is compiled only when -march supports SVE
(e.g. '-march=armv8.2a+sve'), there maybe some problem[1] with this
approach.
The solution:
a. If the minimum instruction set support SVE then compiles it.
b. Else if the compiler support SVE then compiles it.
c. Otherwise don't com
Currently, the soc_kunpeng930 declares '-march=armv8.2-a+crypto+sve',
but some compiler doesn't recognize the march because it doesn't
support sve.
To solve this bug we use the following scheme:
1. Define 'march_base' tuple which defines support march, it should
arrange from lower to higher.
e.g.
https://bugs.dpdk.org/show_bug.cgi?id=703
Bug ID: 703
Summary: [21.05]Host testpmd crash when booting VM with
vhost-user 2 queues + viommu
Product: DPDK
Version: unspecified
Hardware: x86
OS: Linux
From: Yunjian Wang
Currently, configuring a mlx device, it will allocate its
own process private in mlx5_proc_priv_init() and only frees
it when closing the device. This will lead to a memory leak,
when a device is configured repeatedly.
For example:
for(...)
do
rte_eth_dev_configure
rte
From: Yunjian Wang
Currently, configuring a mlx device, it will allocate its
own process private in mlx5_proc_priv_init() and only frees
it when closing the device. This will lead to a memory leak,
when a device is configured repeatedly.
For example:
for(...)
do
rte_eth_dev_configure
rte
From: Yunjian Wang
This series include two fixes patches for mlx4/mlx5.
---
v2:
* remove redundant empty line
Yunjian Wang (2):
net/mlx4: fix mem leak when a device configured repeatedly
net/mlx5: fix mem leak when a device configured repeatedly
drivers/net/mlx4/mlx4.c | 1 +
drivers/n
> -Original Message-
> From: Slava Ovsiienko [mailto:viachesl...@nvidia.com]
> Sent: Wednesday, May 12, 2021 10:05 PM
> To: wangyunjian ; dev@dpdk.org
> Cc: Matan Azrad ; Shahaf Shuler ;
> dingxiaoxiong ; sta...@dpdk.org
> Subject: RE: [dpdk-dev] [PATCH 1/2] net/mlx4: fix mem leak when a
On 5/1/2021 7:33 PM, Dmitry Kozlyuk wrote:
Each time a work was scheduled in the interrupt thread,
usually an alarm, a handle was opened but not closed.
Opening a handle is a system call, which harms alarm precision.
Instead of opening and closing a handle each time, open it
when interrupt thr
在 2021/5/13 0:53, Bruce Richardson 写道:
On Fri, Apr 23, 2021 at 03:35:30PM +0800, Min Hu (Connor) wrote:
Variable i is used as a denominator which may be zero, and
this may result in segmentation fault.
This patch fixed it.
Fixes: 948bc3d6d095 ("test: add reciprocal based division")
Cc: sta.
Variable i is used as a denominator which may be zero, and
this may result in segmentation fault.
This patch fixed it.
Fixes: 948bc3d6d095 ("test: add reciprocal based division")
Cc: sta...@dpdk.org
Signed-off-by: Min Hu (Connor)
---
v2:
* to skip the printouts if i == 0.
---
app/test/test_rec
On 2021/5/13 7:21, Honnappa Nagarahalli wrote:
>
>
>>
>>>
>>> Currently, the SVE code is compiled only when -march supports SVE
>>> (e.g. '- march=armv8.2a+sve'), there maybe some problem[1] with this
>> approach.
>>>
>>> The solution:
>>> a. If the minimum instruction set support SVE then com
在 2021/5/12 22:36, Thomas Monjalon 写道:
11/05/2021 02:53, Min Hu (Connor):
Segmentation fault may occur without checking if memzone
reserves succeed or not.
This patch fixed it.
Fixes: 50247fe03fe0 ("test/timer: exercise new APIs in secondary process")
Cc: sta...@dpdk.org
Signed-off-by: Min
>
> >
> > Currently, the SVE code is compiled only when -march supports SVE
> > (e.g. '- march=armv8.2a+sve'), there maybe some problem[1] with this
> approach.
> >
> > The solution:
> > a. If the minimum instruction set support SVE then compiles it.
> > b. Else if the compiler support SVE then
>
> Currently, the SVE code is compiled only when -march supports SVE (e.g. '-
> march=armv8.2a+sve'), there maybe some problem[1] with this approach.
>
> The solution:
> a. If the minimum instruction set support SVE then compiles it.
> b. Else if the compiler support SVE then compiles it.
> c.
>
> On Wed, May 12, 2021 at 2:01 PM Chengwen Feng
> wrote:
> >
> > Currently, the soc_kunpeng930 declares '-march=armv8.2-a+crypto+sve',
> > but some compiler doesn't recognize the march because it doesn't
> > support sve.
>
> Agree with the problem statement.
>
> + @Juraj Linkeš
>
> >
> > T
A new DPDK release candidate is ready for testing:
https://git.dpdk.org/dpdk/tag/?id=v21.05-rc3
There are 114 new patches in this snapshot.
Release notes:
https://doc.dpdk.org/guides/rel_notes/release_21_05.html
Please test and report issues on bugs.dpdk.org.
You may share some r
12/05/2021 22:24, Thomas Monjalon:
> > > Hook check-symbol-maps.sh in the symbol check when in developer mode to
> > > help developers catch issues before submitting their changes.
> > >
> > > Signed-off-by: David Marchand
> > +1
> >
> > Reviewed-by: Ray Kinsella
>
> Series applied, thanks
Oo
> > Hook check-symbol-maps.sh in the symbol check when in developer mode to
> > help developers catch issues before submitting their changes.
> >
> > Signed-off-by: David Marchand
> +1
>
> Reviewed-by: Ray Kinsella
Series applied, thanks
12/05/2021 18:32, David Hunt:
> when turbo is enabled or disabled, the frequency is set to a low non-turbo
> frequency, so we need to set to the frequency expected by the test before
> checking.
>
> Fixes: aeaeaf5f2d62 ("test/power: add cases for turbo feature")
> Cc: sta...@dpdk.org
>
> Signed-o
> -Original Message-
> From: Yigit, Ferruh
> Sent: Tuesday, May 11, 2021 2:21 PM
> To: Dumitrescu, Cristian ; Haifei Luo
> ; ma...@nvidia.com; or...@nvidia.com;
> viachesl...@nvidia.com; Li, Xiaoyun
> Cc: dev@dpdk.org; tho...@monjalon.net; rasl...@nvidia.com; Singh,
> Jasvinder
> Subje
06/05/2021 04:56, Stephen Hemminger:
> On Thu, 6 May 2021 00:54:35 +0300
> Dmitry Kozlyuk wrote:
>
> > This fixes -Wformat warning with clang 10.0.0 on Windows.
> >
> > Fixes: f8244c6399d9 ("ethdev: increase port id range")
> > Cc: sta...@dpdk.org
> >
> > Signed-off-by: Dmitry Kozlyuk
> > Ack
Ping for review please.
21/04/2021 05:53, Min Hu (Connor):
> From: HongBo Zheng
>
> Fix assignment errors of ep_hgh_edpi in function parse_ep_config.
>
> Fixes: a137d012a0dd ("examples/l3fwd-power: support traffic pattern aware
> control")
> Cc: sta...@dpdk.org
>
> Signed-off-by: HongBo Zheng
12/05/2021 18:52, Ferruh Yigit:
> On 4/29/2021 1:50 AM, Min Hu (Connor) wrote:
> > Currently, some examples check that the port is on the same NUMA
> > node as the polling thread for best performance. The method is
> > to compare the socket id of port and that of current core. If the
> > result is
Ping for v3
12/02/2021 01:21, Honnappa Nagarahalli:
>
>
> >
> > 14/01/2021 17:55, Stephen Hemminger:
> > > The implementation of reader/writer locks in DPDK (from first release)
> > > is simple and fast. But it can lead to writer starvation issues.
> > >
> > > It is not easy to fix this without
> > The special fast-path for returning completed descriptors without
> > reporting status or user-handles returns the number of completed ring
> > slots used, rather than the number of actual user-submitted jobs. This
> > means that the counts returned are too high, as the batch descriptor
> > slo
Hello Wenxu,
We've integrated this fix -
https://patchwork.dpdk.org/project/dpdk/patch/20210512102408.7501-1-jiaw...@nvidia.com/
Could you please confirm it resolves your issue?
BTW, have you opened a BZ ticket? If so, could you please send me the link?
Regards,
Asaf Penso
_
Hi Bruce,
On 12/05/2021 19:32, Bruce Richardson wrote:
On Mon, May 10, 2021 at 06:53:19PM +0200, Heinrich Kuhn wrote:
From: "Chaoyong.He"
Set the Rx multi-queue mode to NONE when configuring a port that is
associated with hardware that only supports a single Rx queue.
Signed-off-by: Chaoyong
From: Timothy McDaniel
This commit fixes the following bugs in the vector based
dequeue path:
- extract hw sched type
- update xstats
The default mode of operation was also changed from vector
optimized mode to scalar mode.
Fixes: 000a7b8e7582 ("event/dlb2: optimize dequeue operation")
Cc: timo
On 11-May-2021 11:39 AM, Anatoly Burakov wrote:
> On 11-May-21 4:31 PM, Anatoly Burakov wrote:
> > Previously, the semantics of power monitor were such that we were
> > checking current value against the expected value, and if they matched,
> > then the sleep was aborted. This is somewhat inflexibl
11/05/2021 08:45, Michal Krawczyk:
> From: Igor Chauskin
> --- a/doc/guides/rel_notes/release_21_05.rst
> +++ b/doc/guides/rel_notes/release_21_05.rst
> @@ -209,6 +209,7 @@ New Features
>
>* Changed memcpy mapping to the dpdk-optimized version.
>* Updated ena_com (HAL) to the latest vers
On Fri, Apr 23, 2021 at 03:35:30PM +0800, Min Hu (Connor) wrote:
> Variable i is used as a denominator which may be zero, and
> this may result in segmentation fault.
>
> This patch fixed it.
>
> Fixes: 948bc3d6d095 ("test: add reciprocal based division")
> Cc: sta...@dpdk.org
>
> Signed-off-by:
On 4/29/2021 1:50 AM, Min Hu (Connor) wrote:
> Currently, some examples check that the port is on the same NUMA
> node as the polling thread for best performance. The method is
> to compare the socket id of port and that of current core. If the
> result is different, warning info will be given.
>
On 12/5/2021 4:34 PM, Thomas Monjalon wrote:
22/04/2021 18:13, Burakov, Anatoly:
On 22-Apr-21 3:40 PM, David Hunt wrote:
--- a/app/test/test_power_cpufreq.c
+++ b/app/test/test_power_cpufreq.c
@@ -48,7 +48,7 @@ static uint32_t total_freq_num;
static uint32_t freqs[TEST_POWER_FREQS_NUM_MAX]
when turbo is enabled or disabled, the frequency is set to a low non-turbo
frequency, so we need to set to the frequency expected by the test before
checking.
Fixes: aeaeaf5f2d62 ("test/power: add cases for turbo feature")
Cc: sta...@dpdk.org
Signed-off-by: David Hunt
---
changes in v2
none
With the intel_pstate driver and turbo enabled, indexing is slightly
different to normal, so to get the test to work properly, enable
turbo at the start.
Fixes: ed7c51a6a680 ("app/test: vm power management")
Cc: sta...@dpdk.org
Signed-off-by: David Hunt
---
changes in v2
none
---
app/test/t
With the intel_pstate driver and turbo enabled, the top frequency in
the frequency array is the P1+1, i.e. 231, whereas the frequency
shown in scaling_cur_freq could be a lot higher.
This patch adds a flag to the check_cur_freq function so that we can
specify if a frequency is greater than exp
Different drivers present the current cpu core frequency in different
sysfs iles. Some present it in cpuinfo_cur_freq, some in scaling_cur_freq,
and some actually present it in both.
This patch attempts to open one, if that fails, tries the other.
Fixes: d550a8cc31f3 ("app/test: enhance power man
On Mon, May 10, 2021 at 06:53:19PM +0200, Heinrich Kuhn wrote:
> From: "Chaoyong.He"
>
> Set the Rx multi-queue mode to NONE when configuring a port that is
> associated with hardware that only supports a single Rx queue.
>
> Signed-off-by: Chaoyong He
> Signed-off-by: Heinrich Kuhn
> Signed-o
09/05/2021 18:09, Jerin Jacob Kollanukkaran:
> http://dpdk.org/git/next/dpdk-next-eventdev
Pulled, thanks
> Return value of a function 'rte_zmalloc' is dereferenced without
> checking, and it may call segmentation fault.
>
> This patch fixed it.
>
> Fixes: f8be1786b1b8 ("app/crypto-perf: introduce performance test
> application")
> Cc: sta...@dpdk.org
>
> Signed-off-by: Min Hu (Connor)
> ---
> v3:
> The special fast-path for returning completed descriptors without
> reporting status or user-handles returns the number of completed ring
> slots used, rather than the number of actual user-submitted jobs. This
> means that the counts returned are too high, as the batch descriptor
> slots would b
> When enqueuing a descriptor, when checking that there is at least one
> slot free for the current descriptor and a later batch descriptor, we
> need to test for both two free and one free, in case the last write
> was a batch descriptor which is allowed to use the "spare" slot.
>
> Similarly, wh
> > > This patch fixes the NULL auth generation case where the request
> > > shouldn't contain the authentication result address. Allows to run
> > > ipsec_autotest with a QAT device.
> > >
> > > Fixes: 65beb9abca6d ("crypto/qat: fix null auth when using VFIO")
> > >
> > > Signed-off-by: Adam Dybko
> -Original Message-
> From: Thomas Monjalon
> Sent: Wednesday, May 12, 2021 9:35 PM
> To: Ciara Power ; Akhil Goyal
>
> Cc: dev@dpdk.org; declan.dohe...@intel.com; acon...@redhat.com;
> hemant.agra...@nxp.com; Anoob Joseph ;
> ruifeng.w...@arm.com; asoma...@amd.com;
> ajit.khapa...@bro
PMD updates are expected in alphabetical order based on their vendor
name. Clarify this expectation in the section comment.
Signed-off-by: Ferruh Yigit
---
v2:
* Do not restrict clarification for ethdev
---
doc/guides/rel_notes/release_21_05.rst | 2 +-
1 file changed, 1 insertion(+), 1 deletion
12/05/2021 18:00, Akhil Goyal:
> > I see this patchset is merged in the crypto tree but I guess it is not
> > definitive.
> > Am I right it is too late to do such refactoring after -rc2?
> > What is the intent?
> >
> The case is not failing in my testing. Neither it is reproducible at Ciara's
>
12/05/2021 17:48, Slava Ovsiienko:
> Hi, Thomas
>
> > How do we switch?
>
> Historically mlx5 created Queues and other HW objects with Verbs.
> Now we are migrating ("switched") to DevX - queues are being created with
> DevX calls.
> If there is no E-Switch - the new DevX approach is engaged,
>
> I see this patchset is merged in the crypto tree but I guess it is not
> definitive.
> Am I right it is too late to do such refactoring after -rc2?
> What is the intent?
>
The case is not failing in my testing. Neither it is reproducible at Ciara's
end.
Ciara has asked to rerun the CI to ch
Hi, Thomas
> How do we switch?
Historically mlx5 created Queues and other HW objects with Verbs.
Now we are migrating ("switched") to DevX - queues are being created with DevX
calls.
If there is no E-Switch - the new DevX approach is engaged,
for E-Switch configurations we still use Verbs.
With
I see this patchset is merged in the crypto tree but I guess it is not
definitive.
Am I right it is too late to do such refactoring after -rc2?
What is the intent?
12/05/2021 15:15, Akhil Goyal:
> Hi Ciara,
>
> I see that CI is reporting one error in autotest. Can you please check if it
> is b
22/04/2021 18:13, Burakov, Anatoly:
> On 22-Apr-21 3:40 PM, David Hunt wrote:
> > --- a/app/test/test_power_cpufreq.c
> > +++ b/app/test/test_power_cpufreq.c
> > @@ -48,7 +48,7 @@ static uint32_t total_freq_num;
> > static uint32_t freqs[TEST_POWER_FREQS_NUM_MAX];
> >
> > static int
> > -che
-Original Message-
From: dev On Behalf Of dev-requ...@dpdk.org
Sent: Friday, April 23, 2021 4:31 PM
To: dev@dpdk.org
Subject: dev Digest, Vol 348, Issue 163
Send dev mailing list submissions to
dev@dpdk.org
To subscribe or unsubscribe via the World Wide Web, visit
http
12/05/2021 16:51, Bing Zhao:
> From: Thomas Monjalon
> > 12/05/2021 16:36, Bing Zhao:
> > > When switching to the Devx interface, the kernel driver behavior
> > may
> >
> > What do you mean by "switching"?
>
> I mean when changing the Rxq from Verbs API to Devx API.
How do we switch?
> > > be
On Wed, May 12, 2021 at 5:30 AM Wang, Yinan wrote:
>
> Hi David,
>
> Since vhost tx offload can’t work now, we report a Bugzilla as below, could
> you help to take a look?
> https://bugs.dpdk.org/show_bug.cgi?id=702
(I discovered your mail from 05/08 only today, now that I got a new
mail, might
> > In function power_guest_channel_read_msg, 'lcore_id' is used before
> > validity check, which may cause buffer 'global_fds' accessed by index
> > 'lcore_id' overflow.
> >
> > This patch moves the validity check of 'lcore_id' before the 'lcore_id'
> > being used for the first time.
> >
> > Fixes
11/05/2021 19:21, Ranjit Menon:
> On 5/11/2021 12:59 AM, Dmitry Kozlyuk wrote:
> > 2021-05-11 09:41 (UTC+0200), Thomas Monjalon:
> >> 02/05/2021 04:33, Dmitry Kozlyuk:
> >>> Interrupt manager in Windows EAL allocates on IOCP and starts
> >>> a control thread that runs indefinitely. At DPDK cleanup
11/05/2021 12:41, Chengwen Feng:
> Currently, the mp uses gettimeofday() API to get the time, and used as
> timeout parameter.
>
> But the time which gets from gettimeofday() API isn't monotonically
> increasing. The process may fail if the system time is changed.
>
> This fixes it by using clock
Hi Thomas,
> -Original Message-
> From: Thomas Monjalon
> Sent: Wednesday, May 12, 2021 10:49 PM
> To: Bing Zhao
> Cc: Slava Ovsiienko ; Matan Azrad
> ; dev@dpdk.org; Ori Kam ; Raslan
> Darawsheh ; sta...@dpdk.org
> Subject: Re: [PATCH] net/mlx5: fix loopback for DV queue
>
> External e
The special fast-path for returning completed descriptors without
reporting status or user-handles returns the number of completed ring
slots used, rather than the number of actual user-submitted jobs. This
means that the counts returned are too high, as the batch descriptor
slots would be included
When enqueuing a descriptor, when checking that there is at least one
slot free for the current descriptor and a later batch descriptor, we
need to test for both two free and one free, in case the last write
was a batch descriptor which is allowed to use the "spare" slot.
Similarly, when computing
12/05/2021 16:36, Bing Zhao:
> When switching to the Devx interface, the kernel driver behavior may
What do you mean by "switching"?
> be different from using Verbs. The Tx loopback cannot work properly
> even if the Tx and Rx queues are configured with loopback attribute.
> To support self loopb
> > The documentation is generated in HTML only.
> > The PDF format is abandoned since DPDK 20.11 while dropping support of
> > the make-based build.
> >
> > This decision has been mentioned by the Technical Board:
> > https://mails.dpdk.org/archives/dev/2021-January/195549.html
> >
> > Fixes: 3c
06/05/2021 11:29, Min Hu (Connor):
> From: Chengchang Tang
>
> Timer sample example assumes that the frequency of the timer is about
> 2Ghz to control the period of calling rte_timer_manage(). But this
> assumption is easy to fail. For example. the frequency of tsc on ARM64
> is much less than 2G
11/05/2021 02:53, Min Hu (Connor):
> Segmentation fault may occur without checking if memzone
> reserves succeed or not.
>
> This patch fixed it.
>
> Fixes: 50247fe03fe0 ("test/timer: exercise new APIs in secondary process")
> Cc: sta...@dpdk.org
>
> Signed-off-by: Min Hu (Connor)
You missed r
When switching to the Devx interface, the kernel driver behavior may
be different from using Verbs. The Tx loopback cannot work properly
even if the Tx and Rx queues are configured with loopback attribute.
To support self loopback for Tx, a Verbs dummy queue pair needs to
be created to trigger the
22/04/2021 08:21, Min Hu (Connor):
> This patch adds return value check for setting an attribute.
>
> Fixes: 88a81bcecb7b ("raw/skeleton: remove compile-time constant for device
> id")
> Cc: sta...@dpdk.org
>
> Signed-off-by: Min Hu (Connor)
> ---
> ret = rte_rawdev_set_attr(test_dev_id,
Hi Adam,
> -Original Message-
> From: dev On Behalf Of Zhang, Roy Fan
> Sent: Monday, May 10, 2021 12:42 PM
> To: Dybkowski, AdamX ; dev@dpdk.org;
> Kusztal, ArkadiuszX ; gak...@marvell.com
> Subject: Re: [dpdk-dev] [PATCH] crypto/qat: fix NULL authentication request
>
> > -Original
Hi, Yunjian
Nice and tiny fix, thank you!
Could you, please, remove the empty line after mlx4/5_proc_priv_uninit(dev)
to follow the mlx5 coding style?
With best regards,
Slava
> -Original Message-
> From: wangyunjian
> Sent: Wednesday, May 12, 2021 16:18
> To: dev@dpdk.org
> Cc: Matan
From: Yunjian Wang
Currently, configuring a mlx device, it will allocate its
own process private in mlx5_proc_priv_init() and only frees
it when closing the device. This will lead to a memory leak,
when a device is configured repeatedly.
For example:
for(...)
do
rte_eth_dev_configure
From: Yunjian Wang
Currently, configuring a mlx device, it will allocate its
own process private in mlx5_proc_priv_init() and only frees
it when closing the device. This will lead to a memory leak,
when a device is configured repeatedly.
For example:
for(...)
do
rte_eth_dev_configure
From: Yunjian Wang
This series include two fixes patches for mlx4/mlx5.
Yunjian Wang (2):
net/mlx4: fix mem leak when a device configured repeatedly
net/mlx5: fix mem leak when a device configured repeatedly
drivers/net/mlx4/mlx4.c | 2 ++
drivers/net/mlx5/mlx5.c | 2 ++
2 files changed, 4
Hi Ciara,
I see that CI is reporting one error in autotest. Can you please check if it is
because of your patchset?
9/96 DPDK:fast-tests / cycles_autotestFAIL1.42s
(exit status 255 or signal 127 SIGinvalid)
12:33:15 DPDK_TEST=cycles_autotest MALLOC_PERTURB_=44
/ho
On 5/11/2021 2:14 PM, Ferruh Yigit wrote:
> Reproduced with '--buildtype=debugoptimized' config,
> compiler version: gcc (GCC) 12.0.0 20210509 (experimental)
>
> Build error:
> In file included from ../drivers/net/bnx2x/bnx2x_rxtx.c:8:
> ../drivers/net/bnx2x/bnx2x_rxtx.c: In function ‘bnx2x_upd_rx
12/05/2021 12:47, Kevin Laatz:
> In the function __idxd_completed_ops() we have a parameter shadow warning
> due to a local variable having the same name as one of the function
> parameters. This issue is fixed by simply renaming the local variable.
>
> This warning was caught when additions were
> > When getting meter flow_id bits, there's an issue that not handling
> > correctly
> > if flow_id is 0.
> >
> > This fix this issue that when flow_id is 0, treat it as 1 bit.
> >
> > Fixes: 83306d6c46 ("net/mlx5: fix meter statistics")
> > Cc: sta...@dpdk.org
> >
> > Signed-off-by: Shun Hao
12/05/2021 14:09, Jiawei Wang:
> One of the user parameters for the flow AGE action is the
> action context. This context should be provided back to the
> user when the action is aged-out.
> While this context is NULL, a default value should be provided
> by the PMD: the rte_flow pointer in case of
On 5/12/2021 11:13 AM, Michal Krawczyk wrote:
> Hi,
>
> this version updates the driver to version 2.3.0, which fixes multiple
> bugs, contains part of the work on making the ENA PMD fully MP aware, and
> also updates HAL to the latest version.
>
> More detailed list of changes:
> * memcpy mapp
Currently action RTE_FLOW_ACTION_TYPE_METER_COLOR is defined.
Add the CLI for this aciton: color type (types)
There are three types: green, yellow and red.
Example for the new policy meter CLIs:
add port meter policy 0 1 g_actions color type green / end y_actions
color type yellow / end r
When getting meter flow_id bits, there's an issue that not handling
correctly if flow_id is 0.
This fix this issue that when flow_id is 0, treat it as 1 bit.
Fixes: 83306d6c46 ("net/mlx5: fix meter statistics")
Cc: sta...@dpdk.org
Signed-off-by: Shun Hao
---
drivers/net/mlx5/mlx5_flow.c | 4 ++
Current the ASO age action was supported in the non-root table,
and the counter based age action was be used in the root table.
The FDB table skips group 0 on MLX5 PMD by adding implicit rule
that jump to non-root table, but PMD code use the original group
value for checking.
This patch adds the
One of the user parameters for the flow AGE action is the
action context. This context should be provided back to the
user when the action is aged-out.
While this context is NULL, a default value should be provided
by the PMD: the rte_flow pointer in case of rte_flow_create API
and the action point
On 5/12/2021 11:49 AM, Haifei Luo wrote:
> Currently action RTE_FLOW_ACTION_TYPE_METER_COLOR is defined.
> Add the CLI for this aciton: color type (types)
> There are three types: green, yellow and red.
>
> Example for the new policy meter CLIs:
>add port meter policy 0 1 g_actions color type
> > Currently RSS expansion only supports GRE and GRE KEY.
> > This patch adds RSS expansion for NVGRE item so PMD can expand flow item
> > correctly.
> >
> > Fixes: ea81c1b816f7 ("net/mlx5: fix NVGRE matching")
> > Cc: sta...@dpdk.org
> >
> > Signed-off-by: Jiawei Wang
> > Acked-by: Xiaoyu Min
One of the user parameters for the flow AGE action is the
action context. This context should be provided back to the
user when the action is aged-out.
While this context is NULL, a default value should be provided
by the PMD: the rte_flow pointer in case of rte_flow_create API
and the action point
The flow context in the rte_flow_action_age structure was set by user,
and reported by the MLX5 PMD while calling rte_flow_get_aged_flow API.
If the flow context was NULL while create ASO age action, while flow
aged, the PMD report the NULL context to user.
This patch adds the checking if context
Current the ASO age action was supported in the non-root table,
and the counter based age action was be used in the root table.
The FDB table skips group 0 on MLX5 PMD by adding implicit rule
that jump to non-root table, but PMD code use the original group
value for checking.
This patch uses the
Current the ASO age action was supported in the non-root table,
and the counter based age action was be used in the root table.
The FDB table skips group 0 on MLX5 PMD by adding implicit rule
that jump to non-root table, but PMD code use the original group
value for checking.
This patch adds the
From: Shun Hao
> When getting meter flow_id bits, there's an issue that not handling correctly
> if flow_id is 0.
>
> This fix this issue that when flow_id is 0, treat it as 1 bit.
>
> Fixes: 83306d6c46 ("net/mlx5: fix meter statistics")
> Cc: sta...@dpdk.org
>
> Signed-off-by: Shun Hao
Acke
From: Jiawei(Jonny) Wang
> Currently RSS expansion only supports GRE and GRE KEY.
> This patch adds RSS expansion for NVGRE item so PMD can expand flow item
> correctly.
>
> Fixes: ea81c1b816f7 ("net/mlx5: fix NVGRE matching")
> Cc: sta...@dpdk.org
>
> Signed-off-by: Jiawei Wang
> Acked-by: X
12/05/2021 08:53, Jiawei Wang:
> While there's mirror action prior to the meter action in the E-Switch
> flow, means that the packets should be duplicated into port firstly,
> and then do meter and send to the original destination.
> MLX5 PMD will split the above E-Switch flow into two sub flows,
>
On 5/12/2021 11:49 AM, Haifei Luo wrote:
> Currently action RTE_FLOW_ACTION_TYPE_METER_COLOR is defined.
> Add the CLI for this aciton: color type (types)
> There are three types: green, yellow and red.
>
> Example for the new policy meter CLIs:
>add port meter policy 0 1 g_actions color type
1 - 100 of 158 matches
Mail list logo
