Re: [lng-odp] [API-NEXT/PATCHv1 4/4] validation: classificaiton: remove redundant sequence number check

On 21.10.15 16:54, Bala Manoharan wrote: Whenever a packet contains an error in L3 layer the HW will not parse the L4 layer and since we are reading the sequence number in the L4 data checking the sequence number in this case of error packet is invalid. Regards, Bala Yep. It should be

Re: [lng-odp] [API-NEXT/PATCHv1 4/4] validation: classificaiton: remove redundant sequence number check

Whenever a packet contains an error in L3 layer the HW will not parse the L4 layer and since we are reading the sequence number in the L4 data checking the sequence number in this case of error packet is invalid. Regards, Bala On 21 October 2015 at 18:15, Ivan Khoronzhuk

Re: [lng-odp] [odp-lng][PATCH v0] test: performance: add crypto test

On 19 October 2015 at 03:24, wrote: > From: Alexandru Badicioiu > > Test program to measure crypto operation performance. > Measures the time required to launch and get the result > of ODP crypto operations in async and sync API

Re: [lng-odp] [API-NEXT PATCHv4 1/4] api: packet reference count support

On 21 October 2015 at 14:46, Maxim Uvarov wrote: > On 10/21/2015 15:27, Ola Liljedahl wrote: > >> On 20 October 2015 at 17:41, Maxim Uvarov > > wrote: >> >> On 10/15/2015 23:49, Ola Liljedahl wrote: >> >>

Re: [lng-odp] [PATCH 5/8] add fbk hash table

On 20 October 2015 at 18:29, HePeng wrote: > > 在 2015年10月19日，下午11:31，Ola Liljedahl 写道： > > On 16 October 2015 at 04:59, HePeng wrote: > >> >> 在 2015年10月16日，上午4:19，Ola Liljedahl 写道： >> >> >> >> On 12

[lng-odp] [API-NEXT PATCHv5 0/4] Egress Traffic Manager

Changes in v5 - Add include file odp_traffic_mngr_internal.h - Add support for odp_tm_enq() from packets originating from ordered queues Changes in v4 - Incorporate API changes suggested by Petri Changes in v3 - Fix checkpatch errors (Bill) Changes in v2 - Full patch submission (Barry) Barry

[lng-odp] [API-NEXT PATCHv5 4/4] example: tm: traffic manager example

From: Barry Spinney This commit includes all of the changes to build the traffic_mgr example application. Signed-off-by: Barry Spinney Signed-off-by: Bill Fischofer --- configure.ac| 1 +

[lng-odp] [API-NEXT PATCHv5 1/4] api: tm: add tm API definitions

From: Barry Spinney This introduces an API for configuring and using Traffic Management systems. The purpose of this API is as a general packet scheduling system that accepts packets from input queues and applies strict priority scheduling, weighted fair queuing scheduling

[lng-odp] [API-NEXT PATCHv5 3/4] linux-generic: tm: add tm to build

From: Barry Spinney This commit causes the traffic_mgr to become part of the ODP linux-generic build. Signed-off-by: Barry Spinney Signed-off-by: Bill Fischofer --- platform/linux-generic/Makefile.am| 13

Re: [lng-odp] [PATCH v2] linux-generic: pktio: fill in L2 parse results by default

On Wed, Oct 21, 2015 at 7:36 AM, Petri Savolainen < petri.savolai...@nokia.com> wrote: > Optimize and simplify packet parsing. Fill in L2 metadata > allways in packet input. Perform full packet parsing only > if other than L2 metadata is requested. Perform parsing only > for packets received from

[lng-odp] [PATCH] helper: linux: do not call odp_term_global on thread exit

Remove the call to odp_term_global when a thread returns. This causes some serious side effects (and bugs) when thread are created and joined "dynamically". Signed-off-by: Nicolas Morey-Chaisemartin --- helper/linux.c | 2 -- 1 file changed, 2 deletions(-) diff --git

Re: [lng-odp] [PATCH 2/2] linux-generic: pktio: replace pktio ticketlock by a rwlock

Matias just pointed out to me that the current netmap implementation do not support tx and rx at the same time. This patch should either wait for his patch that adds tx/rx support or be changed to use a single lock. I'll wait for some feedback before posting an updated version On 10/20/2015

[lng-odp] [API-NEXT PATCH] api: pktio: add odp_pktio_print() API

Add API function for printing implementation specific pktio debug information. Signed-off-by: Matias Elo --- include/odp/api/packet_io.h| 9 platform/linux-generic/odp_packet_io.c | 40 ++ 2 files changed, 49

[lng-odp] [PATCH v4 0/6] l2fwd test improvements

Added various improvements to the l2fwd test application: - Supports using odd number of ports - Scheduler queue type can be selected - Added options for enabling/disabling eth address filling New options: -d, --dst_change <0/1>: Enable/disable changing packets' dst eth addresses -s,

[lng-odp] [PATCH v4 2/6] test: l2fwd: add option to select scheduler queue type

Previously only atomic scheduler queues where supported. This was a bottleneck when the number of worker threads was increased. Added an option to choose scheduler queue sync mode (none, atomic, ordered). Signed-off-by: Matias Elo --- test/performance/odp_l2fwd.c | 94

[lng-odp] [PATCH v4 6/6] test: l2fwd: word copy ethernet addresses

Optimize ethernet address filling by using word copy. Signed-off-by: Matias Elo --- test/performance/odp_l2fwd.c | 50 +--- 1 file changed, 38 insertions(+), 12 deletions(-) diff --git a/test/performance/odp_l2fwd.c

Re: [lng-odp] [API-NEXT/PATCHv1 1/4] validation: classification: move destroy_inq() to common file

On 21.10.15 08:14, Balasubramanian Manoharan wrote: Common in-queue destroy function destroy_inq() is moved to odp_classification_common.c file Signed-off-by: Balasubramanian Manoharan Reviewed-by: Ivan Khoronzhuk ---

Re: [lng-odp] [API-NEXT/PATCHv1 4/4] validation: classificaiton: remove redundant sequence number check

I skipped this coz this as I just realised that since this was an error packet and it might not be required for platforms to parse an error packet and hence I removed the sequence number check for this packet and just checked whether this packet was received on error CoS queue. Regards, Bala On

Re: [lng-odp] [API-NEXT PATCHv4 1/4] api: packet reference count support

On 20 October 2015 at 17:41, Maxim Uvarov wrote: > On 10/15/2015 23:49, Ola Liljedahl wrote: > >> On 9 October 2015 at 08:59, Maxim Uvarov > > wrote: >> >> Add api for packet reference count support. Which is

[lng-odp] [PATCH v2] linux-generic: pktio: fill in L2 parse results by default

Optimize and simplify packet parsing. Fill in L2 metadata allways in packet input. Perform full packet parsing only if other than L2 metadata is requested. Perform parsing only for packets received from the network (disable parsing of user created packets). Signed-off-by: Petri Savolainen

Re: [lng-odp] [API-NEXT/PATCHv1 4/4] validation: classificaiton: remove redundant sequence number check

On 21.10.15 15:14, Bala Manoharan wrote: I skipped this coz this as I just realised that since this was an error packet and it might not be required for platforms to parse an error packet and hence I removed the sequence number check for this packet and just checked whether this packet was

Re: [lng-odp] [API-NEXT PATCHv4 1/4] api: packet reference count support

On 10/21/2015 15:27, Ola Liljedahl wrote: On 20 October 2015 at 17:41, Maxim Uvarov > wrote: On 10/15/2015 23:49, Ola Liljedahl wrote: On 9 October 2015 at 08:59, Maxim Uvarov

[lng-odp] [PATCH] configure: move HAVE_PCAP AM_CONDITIONAL to configure.ac

Platform specific m4 files cannot define AM_CONDITIONAL. Signed-off-by: Nicolas Morey-Chaisemartin --- configure.ac | 1 + platform/linux-generic/m4/odp_pcap.m4 | 1 - 2 files changed, 1 insertion(+), 1 deletion(-) diff --git a/configure.ac

Re: [lng-odp] [API-NEXTv4 0/7] Add functions to free multiple packets/buffers at once

Any more feedback on this ? On 10/13/2015 10:47 AM, Nicolas Morey-Chaisemartin wrote: > As a lot of ODP calls allows the user to receive/poll/send multiple packets > at once, I think it makes sense to provide a call > to free a bunch of packets/buffers at the same time. > > Although

Re: [lng-odp] [API-NEXT PATCH] api: pktio: add odp_pktio_print() API

it has to be odp helper and some function like: char * odp_pktio_name(pktio); Maxim. On 10/21/2015 13:28, Matias Elo wrote: Add API function for printing implementation specific pktio debug information. Signed-off-by: Matias Elo --- include/odp/api/packet_io.h

Re: [lng-odp] [API-NEXT/PATCHv1 1/4] validation: classification: move destroy_inq() to common file

For this series: Reviewed-by: Bill Fischofer Part 4 has a typo in the commit msg, however (s/classificaiton/classification/) On Wed, Oct 21, 2015 at 12:14 AM, Balasubramanian Manoharan < bala.manoha...@linaro.org> wrote: > Common in-queue destroy function

Re: [lng-odp] [PATCH] linux-generic: pktio: fill in L2 parse results by default

OK, did found the reply. It improves l2fwd throughput about 20% if errors are not check (which would need full parse). E.g. DPDK L2 fwd does not check errors, and I’m planning create command line param for error checks. Development can continue with L3 and L4 levels also, but L2 is the most

Re: [lng-odp] [API-NEXT/PATCHv1 2/4] validation: classification: use tcp data offset field to calculate data offset

On 21.10.15 08:14, Balasubramanian Manoharan wrote: TCP data offset field is used to calculate the data offset for accessing the sequence number field in the packet. Signed-off-by: Balasubramanian Manoharan Reviewed-by: Ivan Khoronzhuk

Re: [lng-odp] [API-NEXT/PATCHv1 4/4] validation: classificaiton: remove redundant sequence number check

Seems in last patch the following was added to test_pktio_error_cos(): + seqno = cls_pkt_get_seq(pkt); + CU_ASSERT(seqno != TEST_SEQ_INVALID); ... + CU_ASSERT(seqno == cls_pkt_get_seq(pkt)); Why did you skip it here? I expected patch with name "Align..." including this change.

Re: [lng-odp] [PATCH 2/2] linux-generic: netmap: use separate rx and tx descriptors

On Wed, Oct 21, 2015 at 06:26:39AM +, Elo, Matias (Nokia - FI/Espoo) wrote: > Hi, > > I'm still unable to reproduce the problem. Have you tried reloading the > netmap kernel module and driver? > I found that if I apply only this patch and not the link state patch, it works. The problem

Re: [lng-odp] [PATCH] linux-generic: pktio: fill in L2 parse results by default

My comment from 6 days ago hasn't been addressed. There's a small typo in the patch but the larger issue is dow we have any evidence that this improves OVS performance since that was the main reason lazy parsing was introduced? Bill On Wed, Oct 21, 2015 at 2:38 AM, Savolainen, Petri (Nokia -

[lng-odp] [PATCH v4 1/6] test: l2fwd: add option to change destination eth addresses

If the application is being run on a host connected to a switch, not changing the ethernet destination may cause loops. Added an option to change the destination ethernet address. The destination addresses follow the format 02:00:00:00:00:XX where the final octet is the output port number.

[lng-odp] [PATCH v4 3/6] test: l2fwd: fix crash when accuracy is set to 0

Application crashes if accuracy option is set to zero (division by zero). Disable statistics printing if accuracy <= 0. Signed-off-by: Matias Elo --- test/performance/odp_l2fwd.c | 59 +++- 1 file changed, 37 insertions(+), 22

[lng-odp] [PATCH v4 4/6] test: l2fwd: add support for using odd number of ports

Previously only even numbers of ports were supported. Added support for odd numbers of ports (including 1). Src-dst port mappings are saved during initialization to minimize per packet operations. Signed-off-by: Matias Elo --- test/performance/odp_l2fwd.c | 35

[lng-odp] [PATCH v4 5/6] test: l2fwd: add option to disable filling eth addresses

By default every packet's source MAC address is filled to match the output port. Add option to disable this to enable testing packet forwarding without touching the packets. Signed-off-by: Matias Elo --- test/performance/odp_l2fwd.c | 17 +++-- 1 file changed,

Re: [lng-odp] [API-NEXT PATCH] api: pktio: add odp_pktio_print() API

> -Original Message- > From: lng-odp [mailto:lng-odp-boun...@lists.linaro.org] On Behalf Of EXT Maxim > Uvarov > Sent: Wednesday, October 21, 2015 1:58 PM > To: lng-odp@lists.linaro.org > Subject: Re: [lng-odp] [API-NEXT PATCH] api: pktio: add odp_pktio_print() API > > it has to be odp

Re: [lng-odp] [API-NEXT/PATCHv1 3/4] validation: classification: remove redundant pool lookup function

On 21.10.15 08:14, Balasubramanian Manoharan wrote: pool lookup function is removed and pool_default variable is updated directly during pool create Signed-off-by: Balasubramanian Manoharan Reviewed-by: Ivan Khoronzhuk ---

Re: [lng-odp] [PATCH 1/2] linux-generic: netmap: wait for the interface to become active

Ping. > -Original Message- > From: lng-odp [mailto:lng-odp-boun...@lists.linaro.org] On Behalf Of EXT Elo, > Matias (Nokia - FI/Espoo) > Sent: Monday, October 19, 2015 1:32 PM > To: EXT Stuart Haslam ; Maxim Uvarov > > Cc:

Re: [lng-odp] [PATCH 2/2] linux-generic: netmap: use separate rx and tx descriptors

Hi, I'm still unable to reproduce the problem. Have you tried reloading the netmap kernel module and driver? I rebased the particular two patches to the HEAD and this is the output from the pktio validation test: $ sudo ODP_PKTIO_IF0=em49 ODP_PKTIO_IF1=em50 ./test/validation/pktio/pktio_main

Re: [lng-odp] [PATCH 2/2] linux-generic: pktio: replace pktio ticketlock by a rwlock

It would be better to move send / recv locking inside the callbacks. E.g. netmap with multiple queues could reserve single thread per input queue and remove need for any receive side locking (instead of the double locking of this patch). Locking for pktio status data and send/recv could be

Re: [lng-odp] [PATCH] helper: linux: do not call odp_term_global on thread exit

The last ODP thread should call term global. Local term returns which thread was the last one. As long as you leave one thread not terminated, you can create and term threads multiple times. Global term is called only when you have 0 ODP threads left. -Petri > -Original Message- >

Re: [lng-odp] [PATCHv3 0/4] linux-generic: add pktio pcap type

Merged. it will be good to skip other pktio init if prefix for that was pcap: Something like open() can return 0, -1 , and error/skip next. Maxim. On 10/14/2015 15:06, Stuart Haslam wrote: Add a pcap pktio type as a convenient method of providing test traffic to an application. This is

Re: [lng-odp] [PATCH] linux-generic: pktio: fill in L2 parse results by default

Ping. > -Original Message- > From: Savolainen, Petri (Nokia - FI/Espoo) > Sent: Thursday, October 15, 2015 9:12 AM > To: lng-odp@lists.linaro.org > Subject: RE: [lng-odp] [PATCH] linux-generic: pktio: fill in L2 parse > results by default > > Ping. > > > -Original Message- > >

Re: [lng-odp] [PATCH] helper: linux: do not call odp_term_global on thread exit

Shouldn't the main thread be the one to call odp_term_global ? The main thread is not counted in the thread global struct. Nicolas On 10/21/2015 09:54 AM, Savolainen, Petri (Nokia - FI/Espoo) wrote: > The last ODP thread should call term global. Local term returns which thread > was the last

Re: [lng-odp] [PATCH 2/2] linux-generic: pktio: replace pktio ticketlock by a rwlock

On 10/21/2015 09:35 AM, Savolainen, Petri (Nokia - FI/Espoo) wrote: > It would be better to move send / recv locking inside the callbacks. E.g. > netmap with multiple queues could reserve single thread per input queue and > remove need for any receive side locking (instead of the double

Re: [lng-odp] [PATCH] helper: linux: do not call odp_term_global on thread exit

There's no concept of "main thread". All threads need to call init_local(), which should increment ODP thread count. Term_local() decrements the count and when that reach zero it returns (0) that this was the last one. Any thread can call init_global() (be the first one) and any can call

Re: [lng-odp] [PATCH 2/2] linux-generic: pktio: replace pktio ticketlock by a rwlock

> -Original Message- > From: EXT Nicolas Morey-Chaisemartin [mailto:nmo...@kalray.eu] > Sent: Wednesday, October 21, 2015 11:03 AM > To: Savolainen, Petri (Nokia - FI/Espoo); lng-odp@lists.linaro.org > Subject: Re: [lng-odp] [PATCH 2/2] linux-generic: pktio: replace pktio > ticketlock by

Re: [lng-odp] [PATCH 2/2] linux-generic: pktio: replace pktio ticketlock by a rwlock

On 10/21/2015 10:28 AM, Savolainen, Petri (Nokia - FI/Espoo) wrote: > >> -Original Message- >> From: EXT Nicolas Morey-Chaisemartin [mailto:nmo...@kalray.eu] >> Sent: Wednesday, October 21, 2015 11:03 AM >> To: Savolainen, Petri (Nokia - FI/Espoo); lng-odp@lists.linaro.org >> Subject: