On 8/22/2017 5:45 AM, zimeiw wrote:
> hi,
> My test env is dell T430 server, NIC is intel 82599ES.
> I integrate my app with dpdk-17.05.1 version. after do some tcp traffic
> testing, app can receive packets from the 82599 NIC, but 82599 NIC send
> packets failed.
> But my app with dpdk-17.02 wor
On Thu, Aug 17, 2017 at 03:38:05PM +0100, Ferruh Yigit wrote:
> On 8/1/2017 1:09 PM, Nelio Laranjeiro wrote:
> > This version of MLNX_OFED is no more supported.
>
> Does it make sense to document minimum supported MLNX_OFED version?
>
> >
> > Signed-off-by: Nelio Laranjeiro
> > Acked-by: Shahaf
On 8/22/2017 2:00 AM, Hu, Jiayu wrote:
> Hi,
>> -Original Message-
>> From: Yigit, Ferruh
>> Sent: Monday, August 21, 2017 7:04 PM
>> To: Hu, Jiayu ; dev@dpdk.org
>> Cc: Ananyev, Konstantin ; Tan, Jianfeng
>> ; tho...@monjalon.net; Wu, Jingjing
>> ; Yao, Lei A
>> Subject: Re: [PATCH v2 1/2
hi,
no use rte_eth_tx_prepare()?
I set txq_flags value as below, may it cause this issue?
.txq_flags = ~ETH_TXQ_FLAGS_NOXSUMS
--
Best Regards,
zimeiw
At 2017-08-22 15:49:56, "Ferruh Yigit" wrote:
>On 8/22/2017 5:45 AM, zimeiw wrote:
>> hi,
>> My test env is dell T430 server, NIC is
On Thu, Aug 17, 2017 at 03:38:13PM +0100, Ferruh Yigit wrote:
> On 8/1/2017 1:09 PM, Nelio Laranjeiro wrote:
> > Since MLNX_OFED 4.1 this code is no more useful.
>
> Can you please elaborate what is changed with MLNX_OFED 4.1?
Hardware drop queues are supported since this version, before that, th
On Thu, Aug 17, 2017 at 03:38:22PM +0100, Ferruh Yigit wrote:
> On 8/1/2017 1:09 PM, Nelio Laranjeiro wrote:
> > Secondary process is a copy/paste of the mlx4 drivers, it was never
> > tested and it even segfault at the secondary process start in the
> > mlx5_pci_probe().
>
> Does this means multi
On Thu, Aug 17, 2017 at 03:38:33PM +0100, Ferruh Yigit wrote:
> On 8/1/2017 1:09 PM, Nelio Laranjeiro wrote:
> > Those are useless since DPDK headers have been cleaned up.
> >
> > Signed-off-by: Nelio Laranjeiro
> > Acked-by: Shahaf Shuler
>
> <...>
>
> >
> > -/* DPDK headers don't like -ped
On Fri, Aug 18, 2017 at 02:44:53PM +0100, Ferruh Yigit wrote:
> On 8/2/2017 3:10 PM, Nelio Laranjeiro wrote:
> > This series cleanups the control plane part and the way it uses the
> > different
> > kind of objects (DPDK queues, Verbs Queues, ...). It has three goals:
> >
> > 1. Reduce the memo
On 8/22/2017 9:30 AM, zimeiw wrote:
> hi,
>
> no use rte_eth_tx_prepare()?
>
> I set txq_flags value as below, may it cause this issue?
>
> .txq_flags = ~ETH_TXQ_FLAGS_NOXSUMS
So you are enabling all csum offloading, for Tx path this requires
packets to be prepared before sending to the NIC
Hi Jens,
> I'm a little uncertain but my gut feeling is that in this context a
> connection is
> something between two sockets, not between devices.
What do you mean?
This is a unix domain socket connection. DPDK can create the socket, then the
client may connect to it via connect(2).
> I woul
On Mon, 2017-08-21 at 17:19 -0700, Yipeng Wang wrote:
> Membership library is an extension and generalization of a
> traditional
> filter (for example Bloom Filter) structure. In general, the
> Membership
> library is a data structure that provides a "set-summary" and
> responds
> to set-membership
A DPDK app could, whatever the reason, send packets with size 0.
The PMD is not sending those packets, which does make sense,
but the problem is the mbuf is not released either. That leads
to mbufs not being available, because the app trusts the
PMD will do it.
Although this is a problem related t
On Tue, Aug 22, 2017 at 09:55:19AM +, Stojaczyk, DariuszX wrote:
Hi Jens,
I'm a little uncertain but my gut feeling is that in this context a connection
is
something between two sockets, not between devices.
What do you mean?
This is a unix domain socket connection. DPDK can create the s
Use rte_bsf32 and fast bit unset operation to optimize the softrss computation.
The following measurements shows improvement over the default
softrss computation function.
tuple lens old(cycles) new(cycles)
31225 337
93743 992
Signed-off-by: Yangchao Zhou
On Tue, Aug 22, 2017 at 01:58:44PM +0200, Jens Freimann wrote:
On Tue, Aug 22, 2017 at 09:55:19AM +, Stojaczyk, DariuszX wrote:
Do we need a documentation change for this?
To answer my own question, I think doc/guides/prog_guide/vhost_lib.rst
needs an update.
regards,
Jens
regards,
Jen
When user receives destroy_device signal, he does not know *why* that
event happened. He does not differ between socket shutdown and virtio
processing pause. User could completely delete device during transition
from BIOS to kernel, causing freeze or possibly kernel panic. Instead
of changing new_d
This patch adds MSI IRQ mode and in a way, that should
also work on older kernel versions. The base for my patch
was an attempt to do this in cf705bc36c which was later reverted in
d8ee82745a. Compilation was tested on Linux 3.2, 4.10 and 4.12.
MSI(X) setup was already using pci_alloc_irq_vectors
Added new capability to the list of rx offloads for hw timestamp
The PMDs how expose this capability will always have it enabled.
But, if the following API got accepted applications can choose
between disable/enable this API.
http://dpdk.org/dev/patchwork/patch/27470/
Signed-off-by: Raslan Daraws
Added new print in case a PMD exposes Rx timestamp.
Also, added a print for timestamp value in rxonly mode
in case the packet was timestamped.
Signed-off-by: Raslan Darawsheh
---
app/test-pmd/config.c | 3 +++
app/test-pmd/rxonly.c | 2 ++
2 files changed, 5 insertions(+)
diff --git a/app/test-
Expose a new capapilty of Rx hw timestamp and
added new device args to enable it hw_timestamp.
It will add the raw hw timestamp into the packets.
Its expected that it will lower down the performance since using it
will disable the cqe comprission, and will add extra checkes in
the vec rx path.
Si
The names of rte_gro_ctx_create() and rte_gro_ctx_destroy() in
rte_gro_version.map are incorrect. This patch is to fix this issue.
Fixes: e996506a1c07 ("lib/gro: add Generic Receive Offload API framework")
Signed-off-by: Jiayu Hu
---
lib/librte_gro/rte_gro_version.map | 4 ++--
1 file changed,
We need to support how firmware metadata was handled until now and also
the new api, since NFP NFD 3.0 firmware versions. The new metadata api
adds flexibility for working with different metadata types and, mainly,
to allow adding metadata from different firmware components independently.
Although
Hi Ferruh,
On Tue, Aug 22, 2017 at 09:30:32AM +0100, Ferruh Yigit wrote:
> On 8/22/2017 2:00 AM, Hu, Jiayu wrote:
> > Hi,
> >> -Original Message-
> >> From: Yigit, Ferruh
> >> Sent: Monday, August 21, 2017 7:04 PM
> >> To: Hu, Jiayu ; dev@dpdk.org
> >> Cc: Ananyev, Konstantin ; Tan, Jianfe
On 8/22/2017 2:58 PM, Jiayu Hu wrote:
> The names of rte_gro_ctx_create() and rte_gro_ctx_destroy() in
> rte_gro_version.map are incorrect. This patch is to fix this issue.
>
> Fixes: e996506a1c07 ("lib/gro: add Generic Receive Offload API framework")
>
> Signed-off-by: Jiayu Hu
Reviewed-by: Fe
a. about uevent mechanism
As we know, uevent is come from the kobject of the kernel side, every kobject
would have its own uevent, and a sysfs folder identify a kobject,
such as cpu,usb,pci,pci-express,virio, these bus component all have uevent. I
agree that uevent would be the best if it could
Hi,
2017-08-22 15:02 GMT+03:00 Yangchao Zhou :
> Use rte_bsf32 and fast bit unset operation to optimize the softrss
> computation.
> The following measurements shows improvement over the default
> softrss computation function.
>
> tuple lens old(cycles) new(cycles)
> 31225 337
On 8/21/2017 7:17 AM, Shijith Thotton wrote:
> igb_uio and vfio-pci does pci reset during open and release of device.
> So FLR request to LiquidIO PF driver during init and close in PMD is not
> required.
>
> See commit b58eedfc7dd5 ("igb_uio: issue FLR during open and release of
> device file")
>
On 8/21/2017 7:21 AM, Wei Dai wrote:
> This patch align with PF kerenl driver version 5.1.3 to add the
> number of queues to transmit VLAN packets in msg of queue info
> to VF. If DCB is enabled, it is the number of DCB traffic
> classes. If DCB is not enabled and default VLAN is enabled,
> it is
Recent generations of the Intel® Xeon® family processors allow Turbo Boost
to be enabled/disabled on a per-core basis.
This patch set introduces additional API calls to the librte_power library
to allow users to enable/disable Turbo Boost on particular cores.
Additionally, the use of the library
Adds a new set of APIs to allow per-core turbo
enable-disable.
Signed-off-by: David Hunt
---
lib/librte_power/channel_commands.h | 2 +
lib/librte_power/rte_power.c | 9 ++
lib/librte_power/rte_power.h | 41 +
lib/librte_power/rte_power_acpi_cpufreq.c
Add extra commands to command line to allow enable/disable of
per-core turbo.
Signed-off-by: David Hunt
---
examples/vm_power_manager/channel_monitor.c | 12 +++
examples/vm_power_manager/power_manager.c | 36
examples/vm_power_manager/power_manager.h | 52 ++
Add extra commands to guest cli to allow enable/disable of
per-core turbo. Includes messages to vm_power_mgr in host.
Signed-off-by: David Hunt
---
examples/vm_power_manager/guest_cli/vm_power_cli_guest.c | 6 +-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/examples/vm_power_
The per-core turbo functionality is only available on specific models
of CPU, so this patch limits it to those models.
Signed-off-by: David Hunt
---
lib/librte_power/rte_power_acpi_cpufreq.c | 28
1 file changed, 28 insertions(+)
diff --git a/lib/librte_power/rte_po
On 8/22/2017 11:41 AM, Alejandro Lucero wrote:
> A DPDK app could, whatever the reason, send packets with size 0.
> The PMD is not sending those packets, which does make sense,
> but the problem is the mbuf is not released either. That leads
> to mbufs not being available, because the app trusts th
On 8/22/2017 3:00 PM, Alejandro Lucero wrote:
> We need to support how firmware metadata was handled until now and also
> the new api, since NFP NFD 3.0 firmware versions. The new metadata api
> adds flexibility for working with different metadata types and, mainly,
> to allow adding metadata from
This commit extends the testpmd application with new forwarding engine
that demonstrates the use of ethdev traffic management APIs and softnic
PMD for QoS traffic management.
In this mode, 5-level hierarchical tree of the QoS scheduler is built
with the help of ethdev TM APIs such as shaper profil
Add following CLIs in testpmd application;
- commands to build hierarchical tree for the QoS Scheduler.
- commands for runtime update of the hierarchical tree.
- commands to display TM capability information.
(per port, per hierarchy level and per hierarchy node)
- command to set the packet field
On Mon, 21 Aug 2017 19:33:45 +0200
Markus Theil wrote:
> This patch adds MSI IRQ mode and in a way, that should
> also work on older kernel versions. The base for my patch
> was an attempt to do this in cf705bc36c which was later reverted in
> d8ee82745a. Compilation was tested on Linux 3.2, 4.10
On 22.08.2017 18:55, Stephen Hemminger wrote:
> On Mon, 21 Aug 2017 19:33:45 +0200
> Markus Theil wrote:
>
>> This patch adds MSI IRQ mode and in a way, that should
>> also work on older kernel versions. The base for my patch
>> was an attempt to do this in cf705bc36c which was later reverted in
>
Hi All,
With DPDK Stable 17.05 branch, we are seeing an issue with the link state of
10G ports on Broadwell.
Link status of one of the ports keeps flapping at regular intervals, typically
starts after few seconds after the link appeared to settle down.
Other behaviors include not honoring a link
The i40e maintains a single MAC filter table for both
unicast and multicast addresses. The i40e_validate_mac_addr
function was preventing multicast addresses from being added
to the table via i40evf_add_mac_addr. Fixed the issue by
removing the multicast address check in i40e_validate_mac_addr.
In the function ixgbe_flow_create(), the value ntuple_filter_ptr,
ethertype_filter_ptr,syn_filter_ptr,fdir_rule_ptr and l2_tn_filter_ptr
use rte_zmalloc() malloc memory may return NULL,so, we should
add judge the return is NULL or success.
Signed-off-by: Rongqiang XIE
---
drivers/net/ixgbe/ixgbe
eth_ena_dev_init() was not initializing all of the common
pci dev info for the rte_eth_dev. Added call to
rte_eth_copy_pci_info() to complete the init particularly
the driver name.
Signed-off-by: David Harton
---
drivers/net/ena/ena_ethdev.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/d
We are seeing similar behavior with testPmd as well.
1. Keep port 0/1/2/3 in Up/Up/Down/Down status to start the actual test.
1. Following commands will do link status change twice on each port. Hence
we should see the same status as in (1).
1. Testpmd: port stop 0;port start 0;por
rte_eth_stats_get() unconditonally would set rx_nombuf
even if the device was setting the value. A check has
been added in rte_eth_stats_get() to leave the device
value in-tact when non-zero.
Signed-off-by: David Harton
---
lib/librte_ether/rte_ethdev.c | 5 -
1 file changed, 4 insertions(+
An application may want to manipulate the MTU settings of
a device without having to start the device first.
In order to remove the need to start the device the
ixgbe/ixgbevf drivers need to initialize the scattered_rx
value during dev_configure.
Signed-off-by: David Harton
---
drivers/net/ixgbe
When use bond function in mellanox driver environment, we call the
find_port_id_by_pci_addr() function,if we don't add mellanox kernel
driver type in enum rte_kernel_driver, the function will return -1
because kdrv unknow, so we add the mellanox driver type, and when
scan the pci, fill the kdrv to
When use bond function in mellanox driver environment, we call the
find_port_id_by_pci_addr() function,if we don't add mellanox kernel
driver type in enum rte_kernel_driver, the function will return -1
because kdrv unknown, so we add the mellanox driver type, and when
scan the pci, fill the kdrv to
rte_eth_stats_get() unconditonally would set rx_nombuf
even if the device was setting the value. A check has
been added in rte_eth_stats_get() to leave the device
value in-tact when non-zero.
Signed-off-by: David Harton
---
v2: Fixed braces complaint required by other coding standards.
lib/li
Hi Stephen,
Thanks for the comments. We will remove the needless return initialization in
next version.
For the unified API concern, we think the current implementation has two
benefits: 1. It is easier to extend the library
with new types of filters without adding a lot of top level APIs every
rte_pktmbuf_pool_init and rte_pktmbuf_init callback caused memory
corruption on a message memory pool, remove both.
On the other hand, add rte_pktmbuf_pool assertion of private data size
in function rte_pktmbuf_pool_init() to avoid initializing none mbuf
memory pool.
Fixes: 95e8005a56e8 ("example
From: xuemingl
DPDK uses it's own memory management, few regular memory profiler tool
support DPDK now. Malloc cookie check provides limited memory corruption
check, better than nothing. This patch fixes the following:
1. Replace broken generated configuration macro RTE_LIBRTE_MALLOC_DEBUG
with R
Thanks, Cristian
-Original Message-
From: Dumitrescu, Cristian [mailto:cristian.dumitre...@intel.com]
Sent: Friday, August 18, 2017 7:24 PM
To: Kevin Yan ; dev@dpdk.org
Subject: RE: DPDK qos support for 40G port
Hi Kevin,
> Hi Cristian,
> Sorry to bother again, could you give suggestions
Acked-by: Michal Krawczyk
2017-08-23 2:41 GMT+02:00 David Harton :
> eth_ena_dev_init() was not initializing all of the common
> pci dev info for the rte_eth_dev. Added call to
> rte_eth_copy_pci_info() to complete the init particularly
> the driver name.
>
> Signed-off-by: David Harton
> ---
Hi,
I would like to get some inputs on the below.
This is a big (and important) work which I want to include on 17.11. I need to
understand the current approach is acceptable before I continue.
Monday, August 7, 2017 1:54 PM, Shahaf Shuler:
> Tx offloads configuration is per queue. Tx offloads
55 matches
Mail list logo
