Changes in kernel_nic_interface.rst

5569dd7d - kni: allow configuring thread granularity

Thu, 20 Jan 2022 12:41:34 +0000

kni: allow configuring thread granularityThe Kni kthreads seem to be re-scheduled at a granularity of roughly1 millisecond right now, which seems to be insufficient for performingtests involving a lot of control plane traffic.Even if KNI_KTHREAD_RESCHEDULE_INTERVAL is set to 5 microseconds, itseems that the existing code cannot reschedule at the desired granularily,due to precision constraints of schedule_timeout_interruptible().In our use case, we leverage the Linux Kernel for control plane, andit is not uncommon to have 60K - 100K pps for some signaling protocols.Since we are not in atomic context, the usleep_range() function seems to bemore appropriate for being able to introduce smaller controlled delays,in the range of 5-10 microseconds. Upon reading the existing code, it wouldseem that this was the original intent. Adding sub-millisecond delays,seems unfeasible with a call to schedule_timeout_interruptible().KNI_KTHREAD_RESCHEDULE_INTERVAL 5 /* us */schedule_timeout_interruptible( usecs_to_jiffies(KNI_KTHREAD_RESCHEDULE_INTERVAL));Below, we attempted a brief comparison between the existing implementation,which uses schedule_timeout_interruptible() and usleep_range().We attempt to measure the CPU usage, and RTT between two Kni interfaces,which are created on top of vmxnet3 adapters, connected by a vSwitch.insmod rte_kni.ko kthread_mode=single carrier=onschedule_timeout_interruptible(usecs_to_jiffies(5))kni_single CPU Usage: 2-4 %[root@localhost ~]# ping 1.1.1.2 -I eth1PING 1.1.1.2 (1.1.1.2) from 1.1.1.1 eth1: 56(84) bytes of data.64 bytes from 1.1.1.2: icmp_seq=1 ttl=64 time=2.70 ms64 bytes from 1.1.1.2: icmp_seq=2 ttl=64 time=1.00 ms64 bytes from 1.1.1.2: icmp_seq=3 ttl=64 time=1.99 ms64 bytes from 1.1.1.2: icmp_seq=4 ttl=64 time=0.985 ms64 bytes from 1.1.1.2: icmp_seq=5 ttl=64 time=1.00 msusleep_range(5, 10)kni_single CPU usage: 50%64 bytes from 1.1.1.2: icmp_seq=1 ttl=64 time=0.338 ms64 bytes from 1.1.1.2: icmp_seq=2 ttl=64 time=0.150 ms64 bytes from 1.1.1.2: icmp_seq=3 ttl=64 time=0.123 ms64 bytes from 1.1.1.2: icmp_seq=4 ttl=64 time=0.139 ms64 bytes from 1.1.1.2: icmp_seq=5 ttl=64 time=0.159 msusleep_range(20, 50)kni_single CPU usage: 24%64 bytes from 1.1.1.2: icmp_seq=1 ttl=64 time=0.202 ms64 bytes from 1.1.1.2: icmp_seq=2 ttl=64 time=0.170 ms64 bytes from 1.1.1.2: icmp_seq=3 ttl=64 time=0.171 ms64 bytes from 1.1.1.2: icmp_seq=4 ttl=64 time=0.248 ms64 bytes from 1.1.1.2: icmp_seq=5 ttl=64 time=0.185 msusleep_range(50, 100)kni_single CPU usage: 13%64 bytes from 1.1.1.2: icmp_seq=1 ttl=64 time=0.537 ms64 bytes from 1.1.1.2: icmp_seq=2 ttl=64 time=0.257 ms64 bytes from 1.1.1.2: icmp_seq=3 ttl=64 time=0.231 ms64 bytes from 1.1.1.2: icmp_seq=4 ttl=64 time=0.143 ms64 bytes from 1.1.1.2: icmp_seq=5 ttl=64 time=0.200 msusleep_range(100, 200)kni_single CPU usage: 7%64 bytes from 1.1.1.2: icmp_seq=1 ttl=64 time=0.716 ms64 bytes from 1.1.1.2: icmp_seq=2 ttl=64 time=0.167 ms64 bytes from 1.1.1.2: icmp_seq=3 ttl=64 time=0.459 ms64 bytes from 1.1.1.2: icmp_seq=4 ttl=64 time=0.455 ms64 bytes from 1.1.1.2: icmp_seq=5 ttl=64 time=0.252 msusleep_range(1000, 1100)kni_single CPU usage: 2%64 bytes from 1.1.1.2: icmp_seq=1 ttl=64 time=2.22 ms64 bytes from 1.1.1.2: icmp_seq=2 ttl=64 time=1.17 ms64 bytes from 1.1.1.2: icmp_seq=3 ttl=64 time=1.17 ms64 bytes from 1.1.1.2: icmp_seq=4 ttl=64 time=1.17 ms64 bytes from 1.1.1.2: icmp_seq=5 ttl=64 time=1.15 msUpon testing, usleep_range(1000, 1100) seems roughly equivalent inlatency and cpu usage to the variant with schedule_timeout_interruptible(),while usleep_range(100, 200) seems to give a decent tradeoff betweenlatency and cpu usage, while allowing users to tweak the limits forimproved precision if they have such use cases.Disabling RTE_KNI_PREEMPT_DEFAULT, interestingly seems to lead to asoftlockup on my kernel.Kernel panic - not syncing: softlockup: hung tasksCPU: 0 PID: 1226 Comm: kni_single Tainted: G W O 3.10 #1 [] dump_stack+0x19/0x1b [] panic+0xcd/0x1e0 [] watchdog_timer_fn+0x160/0x160 [] __run_hrtimer.isra.4+0x42/0xd0 [] hrtimer_interrupt+0xe7/0x1f0 [] smp_apic_timer_interrupt+0x67/0xa0 [] apic_timer_interrupt+0x6d/0x80This patch also attempts to remove this option.References:[1] https://www.kernel.org/doc/Documentation/timers/timers-howto.txtSigned-off-by: Tudor Cornea Acked-by: Padraig Connolly Reviewed-by: Ferruh Yigit List of files: /dpdk/doc/guides/prog_guide/kernel_nic_interface.rst

a1b2558c - kni: restrict bifurcated device support

Tue, 23 Nov 2021 16:46:17 +0000

kni: restrict bifurcated device supportTo enable bifurcated device support, rtnl_lock is released before callinguserspace callbacks and asynchronous requests are enabled.But these changes caused more issues, like bug #809, #816. To reduce thescope of the problems, the bifurcated device support related changes areonly enabled when it is requested explicitly with new 'enable_bifurcated'module parameter.And bifurcated device support is disabled by default.So the bifurcated device related problems are isolated and they can befixed without impacting all use cases.Bugzilla ID: 816Fixes: 631217c76135 ("kni: fix kernel deadlock with bifurcated device")Cc: stable@dpdk.orgSigned-off-by: Ferruh Yigit Acked-by: Igor Ryzhov List of files: /dpdk/doc/guides/prog_guide/kernel_nic_interface.rst

cbd2f21a - doc: fix typo in KNI guide

Mon, 02 Nov 2020 15:23:48 +0000

doc: fix typo in KNI guideThe typo "withe" should have been "with the". This is now fixed.Fixes: 89397a01ce4a ("kni: set default carrier state of interface")Cc: stable@dpdk.orgSigned-off-by: Ciara Power Acked-by: John McNamara List of files: /dpdk/doc/guides/prog_guide/kernel_nic_interface.rst

89c67ae2 - doc: remove references to make from prog guide

Mon, 21 Sep 2020 13:59:17 +0000

doc: remove references to make from prog guideMake is no longer supported for compiling DPDK, references are nowremoved in the documentation.Signed-off-by: Ciara Power Reviewed-by: Kevin Laatz List of files: /dpdk/doc/guides/prog_guide/kernel_nic_interface.rst

de480bbf - kni: fix build with Linux 4.9.x

Thu, 28 Nov 2019 12:15:26 +0000

kni: fix build with Linux 4.9.xThe 'get_user_pages_remote()' API is updated in kernel 4.10.0 [1],but the check added as > 4.9.0,this logic is broken for kernels 4.9.x, because they justify> 4.9.0 check but have the old API.Fixing the check as >= 4.10.0[1]commit 5b56d49fc31d ("mm: add locked parameter to get_user_pages_remote()")Fixes: d965af9e8ae1 ("kni: increase kernel version requirement for VA")Reported-by: Andrew Rybchenko Suggested-by: David Marchand Signed-off-by: Ferruh Yigit Tested-by: Andrew Rybchenko Reviewed-by: David Marchand List of files: /dpdk/doc/guides/prog_guide/kernel_nic_interface.rst

06710448 - remove blank lines at end of file

Mon, 11 Nov 2019 19:50:15 +0000

remove blank lines at end of fileRemove trailing blank lines. They serve no purpose and are justeditor leftovers.These can cause git to complain about whitespace errors during merges.Signed-off-by: Stephen Hemminger Signed-off-by: Thomas Monjalon List of files: /dpdk/doc/guides/prog_guide/kernel_nic_interface.rst

d965af9e - kni: increase kernel version requirement for VA

Wed, 20 Nov 2019 17:22:26 +0000

kni: increase kernel version requirement for VAA build error reported related to the selected 'get_user_pages_remote()'kernel API:.../kernel/linux/kni/kni_dev.h:113:8: error: too few arguments to function ‘get_user_pages_remote’ ret = get_user_pages_remote(tsk, tsk->mm, iova, 1 ^~~~~~~~~~~~~~~~~~~~~Currently there are three versions of the 'get_user_pages_remote()'supported, based on kernel version < 4.9, = 4.9, > 4.9.These version based checks are not working fine with the distro kernelswhich is the cause of reported build error. The error reported by thekernel version 4.8, but it is using API defined in > 4.9.To be able to take control of this, and possible more, related builderror, increasing the minimum supported kernel version for iova=va withKNI to kernel version 4.9.This leaves us with single version of the kernel API and more manageable.Signed-off-by: Ferruh Yigit Reviewed-by: David Marchand List of files: /dpdk/doc/guides/prog_guide/kernel_nic_interface.rst

a0dede62 - eal/linux: remove KNI restriction on IOVA

Sun, 17 Nov 2019 15:12:44 +0000

eal/linux: remove KNI restriction on IOVANow that KNI supports VA (with kernel versions starting 4.6.0), we canaccept IOVA as VA, but KNI must be configured for this.Pass iova_mode when creating KNI netdevs.So far, IOVA detection policy forced IOVA as PA when KNI is loaded,whatever the buses IOVA requirements were.We can now use IOVA as VA, but this comes with a cost in KNI.When no constraint is expressed by the buses, keep the current behaviorof choosing PA.Note: this change supposes that dpdk is built on the same kernel thanthe target system kernel; no objection has been expressed on this topic.Signed-off-by: Vamsi Attunuru Signed-off-by: Kiran Kumar K Signed-off-by: David Marchand Reviewed-by: Jerin Jacob List of files: /dpdk/doc/guides/prog_guide/kernel_nic_interface.rst

f43d3dbb - doc/guides: clean repeated words

Tue, 12 Nov 2019 19:33:41 +0000

doc/guides: clean repeated wordsShoot repeated words in all our guides.Cc: stable@dpdk.orgSigned-off-by: David Marchand Acked-by: Kevin Traynor List of files: /dpdk/doc/guides/prog_guide/kernel_nic_interface.rst

b34801d1 - kni: support allmulticast mode set

Mon, 12 Aug 2019 03:06:30 +0000

kni: support allmulticast mode setThis patch adds support to allow users enable/disable allmulticast mode forkni interface.This requirement comes from bugzilla 312, more details can refer to:https://bugs.dpdk.org/show_bug.cgi?id=312Bugzilla ID: 312Signed-off-by: Xiaolong Ye Acked-by: Ferruh Yigit List of files: /dpdk/doc/guides/prog_guide/kernel_nic_interface.rst

5eb1708e - kni: fix kernel crash with multi-segments

Tue, 25 Jun 2019 15:04:14 +0000

kni: fix kernel crash with multi-segmentsva2pa depends on the physical address and virtual address offset ofcurrent mbuf. It may get the wrong physical address of next mbuf whichallocated in another hugepage segment.In rte_mempool_populate_default(), trying to allocate whole block ofcontiguous memory could be failed. Then, it would reserve memory inseveral memzones that have different physical address and virtual addressoffsets. The rte_mempool_populate_default() is used byrte_pktmbuf_pool_create().Fixes: 8451269e6d7b ("kni: remove continuous memory restriction")Cc: stable@dpdk.orgSigned-off-by: Yangchao Zhou Acked-by: Ferruh Yigit List of files: /dpdk/doc/guides/prog_guide/kernel_nic_interface.rst

89247e1a - doc: update KNI guides

Mon, 24 Jun 2019 16:47:14 +0000

doc: update KNI guidesUpdate KNI documentation to reflect current ethtool support.Replace references to out dated tools (ifconfig) withmodern iproute2. Tshark is a better replacement for tcpdump.Signed-off-by: Stephen Hemminger Acked-by: Ferruh Yigit List of files: /dpdk/doc/guides/prog_guide/kernel_nic_interface.rst

d629b7b5 - doc: fix spelling reported by aspell in guides

Fri, 26 Apr 2019 15:14:21 +0000

doc: fix spelling reported by aspell in guidesFix spelling errors in the guide docs.Signed-off-by: John McNamara Acked-by: Rami Rosen List of files: /dpdk/doc/guides/prog_guide/kernel_nic_interface.rst

ff1e35fb - kni: calculate MTU from mbuf size

Tue, 26 Mar 2019 18:40:10 +0000

kni: calculate MTU from mbuf size- mbuf_size and mtu are now being calculated accordingto the given mb-pool.- max_mtu is now being set according to the given mtuthe above two changes provide the ability to work with jumbo framesSigned-off-by: Liron Himi Acked-by: Ferruh Yigit List of files: /dpdk/doc/guides/prog_guide/kernel_nic_interface.rst

89397a01 - kni: set default carrier state of interface

Wed, 24 Oct 2018 22:26:28 +0000

kni: set default carrier state of interfaceAdd module parameter 'carrier='on|off' to set the default carrier statefor linux network interfaces created by the KNI module. The defaultcarrier state is 'off'.For KNI interfaces which need to reflect the carrier state ofa physical Ethernet port controlled by the DPDK application, thedefault carrier state should be left set to 'off'. The applicationcan set the carrier state of the KNI interface to reflect the stateof the physical Ethernet port using rte_kni_update_link().For KNI interfaces which are purely virtual, the default carrierstate can be set to 'on'. This enables the KNI interface to beused without having to explicity set the carrier state to 'on'using rte_kni_update_link().Signed-off-by: Dan Gora Acked-by: Ferruh Yigit List of files: /dpdk/doc/guides/prog_guide/kernel_nic_interface.rst

5630257f - doc: convert Intel license headers to SPDX tags

Thu, 01 Feb 2018 17:18:17 +0000

doc: convert Intel license headers to SPDX tagsSigned-off-by: Ferruh Yigit Acked-by: Bruce Richardson List of files: /dpdk/doc/guides/prog_guide/kernel_nic_interface.rst

15034c20 - doc: fix typo in KNI guide

Fri, 04 Aug 2017 03:27:06 +0000

doc: fix typo in KNI guideFixes: fc1f2750a3ec ("doc: programmers guide")Signed-off-by: Jerin Jacob Acked-by: John McNamara List of files: /dpdk/doc/guides/prog_guide/kernel_nic_interface.rst

aa0d7c2d - kni: remove KNI vhost support

Mon, 20 Feb 2017 14:30:24 +0000

kni: remove KNI vhost supportSigned-off-by: Ferruh Yigit Acked-by: Bruce Richardson List of files: /dpdk/doc/guides/prog_guide/kernel_nic_interface.rst

d938931b - doc: announce KNI vhost removal

Thu, 17 Nov 2016 13:27:21 +0000

doc: announce KNI vhost removalSigned-off-by: Ferruh Yigit Acked-by: John McNamara Acked-by: Thomas Monjalon Acked-by: Bruce Richardson Acked-by: Pablo de Lara List of files: /dpdk/doc/guides/prog_guide/kernel_nic_interface.rst

93a298b3 - kni: support core id parameter in single threaded mode

Sat, 24 Sep 2016 13:13:02 +0000

kni: support core id parameter in single threaded modeAllow binding KNI thread to specific core in single threaded modeby setting core_id and force_bind config parameters.Signed-off-by: Vladyslav Buslov Acked-by: Ferruh Yigit Tested-by: Ferruh Yigit List of files: /dpdk/doc/guides/prog_guide/kernel_nic_interface.rst