Thank you Ravi.
This has great utility.
With all the latest laptop having this NIC in the chipset, the portability
makes DPDK development on laptops very useful.
We have seen cases where with this patch, DPDK development and demonstration
was very portable and beneficial for developers.
Kindly
Hi,
2015-09-24 22:10, Arnon Warshavsky:
> Moving from dpdk 1.5 to 2.0 we observed a PPS performance degradation of
> ~30%.
> After chasing this one for a while we found the problem:
>
> A) Between the 2 versions rte_mbuf was increased in size from 1 to 2 cache
> lines.
> B) The standard (non-vect
2015-09-24 17:23, John McNamara:
> Update the DPDK 2.2 release notes with recent fixes:
>
> 7e01e3 i40e: fix base driver allocation when not using first numa node
> 5e73f4 ixgbe: remove burst size restriction of vector Rx
> 7fcd13 ixgbe: fix X550 DCB
> d49e0f hash: fix memory allocation of
Hi All
Moving from dpdk 1.5 to 2.0 we observed a PPS performance degradation of
~30%.
After chasing this one for a while we found the problem:
A) Between the 2 versions rte_mbuf was increased in size from 1 to 2 cache
lines.
B) The standard (non-vector) rx function does not perform a prefetch fo
2015-09-23 16:03, Aaron Conole:
> The vHost eventlink driver is a kernel module that requires a kernel
> source/build directory to build the ko. Convert the fixed kernel build
> directory specifier to one which may be user specified on the command-line.
>
> Signed-off-by: Aaron Conole
Applied, t
Hi Thomas,
Would you consider to apply these 4 patches please? The 4th patch
affects quite many lines of code, and the sooner we apply it, the
fewer chance it will introduce conflicts for later patches.
Thanks.
--yliu
On Wed, Sep 09, 2015 at 05:44:51AM +, Xie, Huawei wrote:
>
> Ack
Please reword the title (no need to re-send the patch),
so that you describe what the patch do (batch interrupt).
2015-09-21 16:16, Huawei Xie:
> In merge-able RX path, vhost injects interrupts to guest for each packet.
> This should degrade performance a lot.
> This patch fixes this issue by inje
I encountered an issue with DPDK 2.1.0 which occasionally causes the link
status interrupt callback not to be called after the interface is started for
the first time. I traced the problem back to the function
eth_igb_link_update(), which is used to determine if the link has changed state
sinc
Add a new guide doc as part of the Linux Getting Started Guide.
The document is a step-by-step guide on how to get high performance
with DPDK on an Intel platform.
It is designed for users who are not familiar with DPDK but would like
to get the best performance with NICs.
Signed-off-by: Qian Xu
On 9/25/2015 12:36 AM, Stephen Hemminger wrote:
> On Thu, 24 Sep 2015 07:30:41 +
> "Xie, Huawei" wrote:
>
>> On 9/21/2015 11:39 AM, Xie, Huawei wrote:
>> vring_size calculation should consider both used_event_idx at the tail
>> of avail ring and avail_event_idx at the tail of used ring.
>> Wil
Update the DPDK 2.2 release notes with recent fixes:
7e01e3 i40e: fix base driver allocation when not using first numa node
5e73f4 ixgbe: remove burst size restriction of vector Rx
7fcd13 ixgbe: fix X550 DCB
d49e0f hash: fix memory allocation of cuckoo key table
9db649 eal/linux: fix epo
At the original point the rx_pkts[pos( + n)] pointers are not initialized, so
the code is prefetching random data.
Signed-off-by: Zoltan Kiss
diff --git a/drivers/net/ixgbe/ixgbe_rxtx_vec.c
b/drivers/net/ixgbe/ixgbe_rxtx_vec.c
index 3c6d8c5..ccd93c7 100644
--- a/drivers/net/ixgbe/ixgbe_rxtx_vec
Hi,
I?m trying to figure out how to send multiple streams of traffic (say by
changing port/ipaddr) from the peer to SUT to test multiqueue with RSS
functionality. I?m using DPDK PktGen on the SUT as well to bind each
receive queue to different cores for spreading the load.
Can we use pcap option fo
On Thu, 24 Sep 2015 15:50:41 -0700
Ravi Kerur wrote:
> Macros RTE_MBUF_DATA_DMA_ADDR and RTE_MBUF_DATA_DMA_ADDR_DEFAULT
> are defined in each PMD driver file. Move those macros into common
> lib/librte_mbuf/rte_mbuf.h file. All PMD drivers include rte_mbuf.h
> file directly/indirectly hence no ad
> -Original Message-
> From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Qian Xu
> Sent: Thursday, September 24, 2015 12:49 PM
> To: dev at dpdk.org
> Subject: [dpdk-dev] [PATCH v3] doc: add performance test guide to the
> linux gsg
>
> Add a new guide doc as part of the Linux Gettin
Macros RTE_MBUF_DATA_DMA_ADDR and RTE_MBUF_DATA_DMA_ADDR_DEFAULT
are defined in each PMD driver file. Move those macros into common
lib/librte_mbuf/rte_mbuf.h file. All PMD drivers include rte_mbuf.h
file directly/indirectly hence no additionl header file inclusion
is necessary.
Compiled for:
Macros RTE_MBUF_DATA_DMA_ADDR and RTE_MBUF_DATA_DMA_ADDR_DEFAULT
are defined in each PMD driver file. Move those macros into common
lib/librte_mbuf/rte_mbuf.h file. All PMD drivers include rte_mbuf.h
file directly/indirectly hence no additionl header file inclusion
is necessary.
Ravi Kerur (1):
Super discussions today on the community call. Looking forward to continuing
the open and honest dialogue in Dublin in two weeks' time.
Don't forget if you have registered for DPDK Userspace 2015 and can't make it
please let me know as we have a waiting list.
Thanks
Siobh?n
> -Original Mess
Test commands are added to support selecting differnt length of
GRE key.
Signed-off-by: Helin Zhang
Signed-off-by: Andrey Chilikin
---
app/test-pmd/cmdline.c | 52 ++
1 file changed, 52 insertions(+)
diff --git a/app/test-pmd/cmdline.c b/app/test
By default, only first 3 bytes of GRE key will be used for hash or
filter calculation. With these changes, it can select 3 or 4 bytes
of GRE key for hash calculation.
Signed-off-by: Helin Zhang
Signed-off-by: Andrey Chilikin
---
drivers/net/i40e/i40e_ethdev.c | 86 +
By default, only 3 bytes of GRE key will be used for hash or filter
calculation. Here adds a workaround for selecting 3 or 4 bytes of GRE
key for that purpose.
Helin Zhang (2):
i40e: add selecting GRE key length
app/testpmd: add test commands for selecting different GRE key sizes
app/test-pm
This patch adds I217 and I218 Intel chipsets.
Compiled for:
> i686-native-linuxapp-gcc
> x86_64-native-linuxapp-clang
> x86_64-native-linuxapp-gcc
> x86_x32-native-linuxapp-gcc
Tested on:
> x86_64 Ubuntu 14.04 with Intel I218-V and I217-LM chipsets.
Signed-off-by: Ravi Kerur
M. Jay(Jayakumar, Muthurajan) and I
discussed and decided to include support for I217/I218 chipsets since
these chipsets are found everywhere f.e. on laptops, low-end servers and
we found it useful and helpful for testing simple functionality. Hence we
decided to send this patch to be included in t
Test commands are added to support clearing input set, or setting
with new input set per different pctype.
Signed-off-by: Helin Zhang
Signed-off-by: Andrey Chilikin
---
app/test-pmd/cmdline.c | 115 +
1 file changed, 115 insertions(+)
v3 changes:
The default fields of a received packet are loaded from firmware,
which cannot be modified even users want to use different fields
for RSS or filtering. Here adds a workaround to open more
flexibilities of selecting packet fields for hash calculation or
flow director to users.
Signed-off-by: Helin
The default fields of a received packet are loaded from firmware,
which cannot be modified even users want to use different fields
for RSS or filtering. Here adds a workaround to open more
flexibilities of selecting packet fields for hash calculation or
flow director to users. It also includes the
2015-08-20 12:01, Ouyang Changchun:
> This patch originates from the patch:
> [dpdk-dev] [PATCH 1/2] Patch for Qemu wrapper for US-VHost to ensure Qemu
> process ends when VM is shutdown
> http://dpdk.org/ml/archives/dev/2014-June/003606.html
>
> Aslo update the vhost sample guide doc.
>
> Signe
Hi Vinod,
Could you provide more information or probably an example of any "non-standard"
packets?
Regards,
Andrey
-Original Message-
From: dev [mailto:dev-boun...@dpdk.org] On Behalf Of Krishna, Vinod
Sent: Thursday, September 24, 2015 12:55 PM
To: dev at dpdk.org
Subject: [dpdk-dev] R
This patch changes the testpmd DCB forwarding stream to make it
based on traffic class.
It also fixes some coding style issues.
Signed-off-by: Jingjing Wu
---
app/test-pmd/cmdline.c | 39 +++-
app/test-pmd/config.c | 159 +
app/test-pmd/t
This patch implements the get_dcb_info ops in i40e driver.
Signed-off-by: Jingjing Wu
---
drivers/net/i40e/i40e_ethdev.c | 42 ++
1 file changed, 42 insertions(+)
diff --git a/drivers/net/i40e/i40e_ethdev.c b/drivers/net/i40e/i40e_ethdev.c
index 7d252
This patch implements the get_dcb_info ops in ixgbe driver.
Signed-off-by: Jingjing Wu
---
drivers/net/ixgbe/ixgbe_ethdev.c | 77
1 file changed, 77 insertions(+)
diff --git a/drivers/net/ixgbe/ixgbe_ethdev.c b/drivers/net/ixgbe/ixgbe_ethdev.c
index a7dc
This patch adds one new API to get dcb related info.
rte_eth_dev_get_dcb_info
Signed-off-by: Jingjing Wu
---
lib/librte_ether/rte_ethdev.c | 18
lib/librte_ether/rte_ethdev.h | 50 +++
2 files changed, 68 insertions(+)
diff --git a/lib/
This patch enables DCB+RSS multi-queue mode, and also fix some coding
style.
Signed-off-by: Jingjing Wu
---
drivers/net/ixgbe/ixgbe_rxtx.c | 48 +-
1 file changed, 29 insertions(+), 19 deletions(-)
diff --git a/drivers/net/ixgbe/ixgbe_rxtx.c b/drivers/net
This patch enables DCB feature on Intel XL710/X710 NICs. It includes:
Receive queue classification based on traffic class
Round Robin ETS schedule (rx and tx)
Priority flow control
Signed-off-by: Jingjing Wu
---
drivers/net/i40e/i40e_ethdev.c | 532 -
Differnet NIC has its specific constraint on the multi-queue
configuration, so move the checking from ethdev lib to drivers.
Signed-off-by: Jingjing Wu
---
drivers/net/e1000/igb_ethdev.c | 84 -
drivers/net/ixgbe/ixgbe_ethdev.c | 171 +
drivers/
Signed-off-by: Jingjing Wu
---
app/test-pmd/testpmd.c | 8
drivers/net/ixgbe/ixgbe_rxtx.c | 10 +-
examples/vmdq_dcb/main.c | 4 ++--
lib/librte_ether/rte_ethdev.h | 14 +++---
4 files changed, 18 insertions(+), 18 deletions(-)
diff --git a/app/test-pmd/
The patch set enables DCB feature on Intel XL710/X710 NICs, including:
Receive queue classification based on traffic class
Round Robin ETS schedule (rx and tx).
Priority flow control
To make the testpmd and ethdev lib more generic on DCB feature, this patch
set also
adds a new API to get DC
On Thu, 24 Sep 2015 18:35:37 +
"Xie, Huawei" wrote:
> On 9/25/2015 12:36 AM, Stephen Hemminger wrote:
> > On Thu, 24 Sep 2015 07:30:41 +
> > "Xie, Huawei" wrote:
> >
> >> On 9/21/2015 11:39 AM, Xie, Huawei wrote:
> >> vring_size calculation should consider both used_event_idx at the tail
Hi, Zhigang,
I checked error code and got a error information "(No more IOTLB entries)." &
xmit function gxio_mpipe_equeue_try_reserve_fast got "Insufficient DMA
credits"!
Command: run ./testpmd -c 0x -m 6144 -n 1 -r 1 --vdev xgbe1 -- --rx=1
--tx=2 --forward-mode=txonly -a --port-top
Signed-off-by: Cunming Liang
---
doc/guides/rel_notes/release_2_2.rst | 2 ++
1 file changed, 2 insertions(+)
diff --git a/doc/guides/rel_notes/release_2_2.rst
b/doc/guides/rel_notes/release_2_2.rst
index 682f468..73dba47 100644
--- a/doc/guides/rel_notes/release_2_2.rst
+++ b/doc/guides/rel_no
The patch enables rx interrupt support on i40e VF and some necessary change on
PF IOV mode to support VF.
On PF side, running in IOV mode via uio won't allow rx interrupt which is
exclusive with mbox interrupt
in single vector competition.
On VF side, one single vector is shared for all the rx qu
The patch enables rx interrupt support on i40e PF non-IOV mode.
Per queue rx interrupt works on vfio, however on uio, all rx queues share one
interrupt vector.
Signed-off-by: Cunming Liang
---
drivers/net/i40e/i40e_ethdev.c | 319 +++--
drivers/net/i40e/i40e_
When igb runs as a PF, mbox interrupt is prerequisite to make VF start normally.
And PF sometimes won't 'dev_start', so the mbox interrupt register during
'dev_init' is required.
The patch rolls back the interrupt register for mbox,lsc to the 'dev_init'.
As UIO doesn't support multiple vector, mbo
As ixgbe vf doesn't support lsc, the patch removes those unused code.
In addition, it does some tiny cleanup.
Signed-off-by: Cunming Liang
---
drivers/net/ixgbe/ixgbe_ethdev.c | 85 +---
1 file changed, 10 insertions(+), 75 deletions(-)
diff --git a/drivers/n
When ixgbe runs as a PF, mbox interrupt is prerequisite to make VF start
normally.
And PF sometimes won't 'dev_start', so the mbox interrupt register during
'dev_init' is required.
The patch rolls back the interrupt register for mbox,lsc to the 'dev_init'.
As UIO doesn't support multiple vector,
VFIO allows multiple MSI-X vector, others doesn't, but maybe will allow it in
the future.
Device drivers need to be aware of the capability.
It's better to avoid condition check on interrupt type(VFIO) everywhere, instead
a capability api is more flexible for the condition change.
Signed-off-by:
The patch adds condition check to avoid enable nothing.
In disable state, both max_intr and nb_efd are zero.
Signed-off-by: Cunming Liang
---
lib/librte_eal/bsdapp/eal/include/exec-env/rte_interrupts.h | 3 ++-
lib/librte_eal/linuxapp/eal/eal_interrupts.c | 8 +++-
lib/lib
According to the VFIO interrupt mapping, the interrupt vector id for rxq starts
from RX_VEC_START.
It doesn't impact the UIO cases.
Signed-off-by: Cunming Liang
---
drivers/net/e1000/igb_ethdev.c | 19 ++-
1 file changed, 14 insertions(+), 5 deletions(-)
diff --git a/drivers/ne
According to the VFIO interrupt mapping, the interrupt vector id for rxq starts
from RX_VEC_START.
It doesn't impact the UIO cases.
Signed-off-by: Cunming Liang
---
drivers/net/ixgbe/ixgbe_ethdev.c | 14 +-
1 file changed, 9 insertions(+), 5 deletions(-)
diff --git a/drivers/net/ix
During VFIO_DEVICE_SET_IRQS, the previous order is {Q0_fd, ... Qn_fd, misc_fd}.
The vector number of misc is indeterminable which is ugly to some NIC(e.g.
i40e, fm10k).
The patch adjusts the order in {misc_fd, Q0_fd, ... Qn_fd}, always reserve the
first vector to misc interrupt.
Signed-off-by: C
This patch series contains four major parts.
1. always reserve vector zero for misc cause in vfio mapping
2. add api to declare the capability of multiple interrupt vector support
3. fix the rx interrupt compatible issue with mbox in ixgbe/igb IOV-PF
4. add rx interrupt support in i40e PF and VF
On 09/23/2015 06:46 AM, Yuanhan Liu wrote:
> On Tue, Sep 22, 2015 at 05:51:02PM +0300, Marcel Apfelbaum wrote:
[...]
>>> It's proved to work after the fix (at least in my testing), but
>>> it's late here and I'm gonna send a new version tomorrow, including
>>> some other comments addressing. Please
> -Original Message-
> From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Qian Xu
> Sent: Thursday, September 24, 2015 12:49 PM
> To: dev at dpdk.org
> Subject: [dpdk-dev] [PATCH v3] doc: add performance test guide to the
> linux gsg
>
> Add a new guide doc as part of the Linux Gettin
2015-08-18 10:51, Ouyang Changchun:
> This commit removes the dev-index, so update the doc for this change:
> 17b8320a3e11e146868906d0082b6e402d5f2255
> "vhost: remove index parameter"
>
> Signed-off-by: Changchun Ouyang
Applied, thanks
Add integrated MP/MC and fragmentation/reassembly support to SWQs
Signed-off-by: Piotr Azarewicz
---
examples/ip_pipeline/app.h | 14 +++
examples/ip_pipeline/config_check.c | 45 +++-
examples/ip_pipeline/config_parse.c | 195 +--
examples/ip_pi
Bug fixes for ring ports with IPv4/IPv6 reassembly support.
Previous implementation can't work properly due to incorrect choosing
process function.
Also, assuming that, when processing ip packet, ip header is know we can
set l3_len parameter here.
Signed-off-by: Piotr Azarewicz
---
lib/librte_po
ring_multi_reader input port (on top of multi consumer rte_ring)
ring_multi_writer output port (on top of multi producer rte_ring)
Signed-off-by: Piotr Azarewicz
---
lib/librte_port/rte_port_ring.c | 311 +++
lib/librte_port/rte_port_ring.h | 35 -
2 fi
This patch set enhancement ip_pipeline application:
- librte_port: add support for multi-producer/multi-consumer ring ports
- librte_port: bug fixes for ring ports with IPv4/IPv6 reassembly support
- ip_pipeline application: integrate MP/MC and fragmentation/reassembly support
to SWQs
v2 changes:
Hi All,
RSS could me more flexible if it can support non-standard packets.
Thanks,
VK
-Original Message-
From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of dev-requ...@dpdk.org
Sent: Thursday, September 24, 2015 12:39 PM
To: dev at dpdk.org
Subject: dev Digest, Vol 59, Issue 28
Sen
Hi, Zhigang
I used "--hvx hugepagesz=16m --hvx hugepages=256", but testpmd still not
work and exit.
Command: run cat /proc/meminfo
MemTotal:8230464 kB
MemFree: 3872576 kB
Buffers: 0 kB
Cached:12480 kB
SwapCached:0 kB
Active:1120
The validator attempts to disable all kernel modules but since
commit 36080ff96b0eb37a6da8c4fec1a2f8a57dfadf5b fails to do so
for KNI, causing the build stage to fail if kernel headers are missing.
With the introduction of CONFIG_RTE_KNI_KMOD, CONFIG_RTE_LIBRTE_KNI=n
can eventually be dropped but
Signed-off-by: Panu Matilainen
---
scripts/validate-abi.sh | 23 +++
1 file changed, 11 insertions(+), 12 deletions(-)
diff --git a/scripts/validate-abi.sh b/scripts/validate-abi.sh
index b9c9989..12946d4 100755
--- a/scripts/validate-abi.sh
+++ b/scripts/validate-abi.sh
@@ -
Its sometimes necessary to disable warnings etc to get an older
version of code to build.
Signed-off-by: Panu Matilainen
---
scripts/validate-abi.sh | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/scripts/validate-abi.sh b/scripts/validate-abi.sh
index 4476433..b9c9989 10
For giggles, tried running abi-validator between 2.0 and 2.1 on
my Fedora 22 laptop, didn't work due to various build failures.
With this patch series the following now succeeds:
EXTRA_CFLAGS="-Wno-error" scripts/validate-abi.sh v2.0.0 v2.1.0
x86_64-native-linuxapp-gcc
Panu Matilainen (3):
scr
On 09/24/2015 10:34 AM, Panu Matilainen wrote:
> # Checking abi compliance relies on using the dwarf information in
> # The shared objects. Thats only included in the DSO's if we build
> @@ -167,6 +171,8 @@ sed -i -e"$ a\CONFIG_RTE_LIBRTE_KNI=n"
> config/defconfig_$TARGET
> export EXTRA_CF
On 2015-09-24 10:36, Thomas Monjalon wrote:
> 2015-09-23 23:56, St Leger, Jim:
>> This call is aimed to get more open dialogue in the community.
>
> I wonder how a call can "get more open dialogue"?
> Because of being "live", a lot of people cannot attend at this time.
> A call is also a place whe
2015-09-23 23:56, St Leger, Jim:
> This call is aimed to get more open dialogue in the community.
I wonder how a call can "get more open dialogue"?
Because of being "live", a lot of people cannot attend at this time.
A call is also a place where only people having the strongest voice will be
hear
The validator attempts to disable all kernel modules but since
commit 36080ff96b0eb37a6da8c4fec1a2f8a57dfadf5b fails to do so
for KNI, causing the build stage to fail if kernel headers are missing.
With the introduction of CONFIG_RTE_KNI_KMOD, CONFIG_RTE_LIBRTE_KNI=n
can eventually be dropped but
Signed-off-by: Panu Matilainen
---
scripts/validate-abi.sh | 25 -
1 file changed, 12 insertions(+), 13 deletions(-)
diff --git a/scripts/validate-abi.sh b/scripts/validate-abi.sh
index b9c9989..4b555de 100755
--- a/scripts/validate-abi.sh
+++ b/scripts/validate-abi.sh
@@
Its sometimes necessary to disable warnings etc to get an older
version of code to build.
Signed-off-by: Panu Matilainen
---
scripts/validate-abi.sh | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/scripts/validate-abi.sh b/scripts/validate-abi.sh
index 4476433..b9c9989 10
For giggles, tried running abi-validator between 2.0 and 2.1 on
my Fedora 22 laptop, didn't work due to various build failures.
With this patch series the following now succeeds:
EXTRA_CFLAGS="-Wno-error" scripts/validate-abi.sh v2.0.0 v2.1.0
x86_64-native-linuxapp-gcc
Panu Matilainen (3):
scr
> -Original Message-
> From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Piotr Azarewicz
> Sent: Thursday, September 24, 2015 12:56 PM
> To: dev at dpdk.org
> Subject: [dpdk-dev] [PATCH v2 0/3] ip_pipeline: add MP/MC and frag/ras
> support to SWQs
>
>
> v2 changes:
> - rte_port_ri
Hi, Arthas
Seems like something wrong with memseg, could you try with
" --hvx hugepagesz=1G --hvx hugepages=4" to boot Gx card? This
helps reserve hugepages for DPDK, and check there are still such errors.
With regard to MDE-4.3.3, please contact your local tech support.
Thanks
-Zhigang
>-O
Otherwise building with KERNELCC="ccache gcc" will fail:
== Build lib/librte_eal/linuxapp/igb_uio
/usr/src/linux-headers-3.13.0-63-generic/arch/x86/Makefile:98: stack protector
enabled but no compiler support
/usr/src/linux-headers-3.13.0-63-generic/arch/x86/Makefile:113: CONFIG_X86_X32
enabl
On Thu, 24 Sep 2015 07:30:41 +
"Xie, Huawei" wrote:
> On 9/21/2015 11:39 AM, Xie, Huawei wrote:
> vring_size calculation should consider both used_event_idx at the tail
> of avail ring and avail_event_idx at the tail of used ring.
> Will merge those two fixes and send a new patch.
> > used ev
On 9/21/2015 11:39 AM, Xie, Huawei wrote:
vring_size calculation should consider both used_event_idx at the tail
of avail ring and avail_event_idx at the tail of used ring.
Will merge those two fixes and send a new patch.
> used event idx is put at the end of available ring. It isn't taken into
>
On Thu, Sep 24, 2015 at 10:50:56AM +0300, Panu Matilainen wrote:
> For giggles, tried running abi-validator between 2.0 and 2.1 on
> my Fedora 22 laptop, didn't work due to various build failures.
> With this patch series the following now succeeds:
>
> EXTRA_CFLAGS="-Wno-error" scripts/validate-a
77 matches
Mail list logo
