- 09 Feb, 2021 40 commits
-
-
git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queueDavid S. Miller authored
Tony Nguyen says: ==================== 100GbE Intel Wired LAN Driver Updates 2021-02-08 This series contains updates to the ice driver and documentation. Brett adds a log message when a trusted VF goes in and out of promiscuous for consistency with i40e driver. Dave implements a new LLDP command that allows adding VSI destinations to existing filters and adds support for netdev bonding events, current support is software based. Michal refactors code to move from VSI stored xsk_buff_pools to netdev-provided ones. Kiran implements the creation scheduler aggregator nodes and distributing VSIs within the nodes. Ben modifies rate limit calculations to use clock frequency from the hardware instead of using a hardcoded one. Jesse adds support for user to control writeback frequency. Chinh refactors DCB variables out of the ice_port_info struct. Bruce removes some unnecessary casting. Mitch fixes an error message that was reported as if_up instead of if_down. Tony adjusts fallback allocation for MSI-X to use all given vectors instead of using only the minimum configuration and updates documentation for the ice driver. ==================== Reviewed-by:
Alexander Duyck <alexanderduyck@fb.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
David S. Miller authored
Huazhong Tan says: ==================== net: hns3: some cleanups for -next There are some cleanups for the HNS3 ethernet driver. change log: V2: remove previous #3 which should target net. previous version: V1: https://patchwork.kernel.org/project/netdevbpf/cover/1612784382-27262-1-git-send-email-tanhuazhong@huawei.com/ ==================== Acked-by:
Jakub Kicinski <kuba@kernel.org> Signed-off-by:
-
Jian Shen authored
Currently the RSS commands of VF are using host byte order. According to the user manual, it should use little endian in the command to firmware. For the host and firmware are both using little endian, so it can work well in this case. Do cleanup to make it more explicitly. Signed-off-by:
Jian Shen <shenjian15@huawei.com> Signed-off-by:
Huazhong tan <tanhuazhong@huawei.com> Signed-off-by:
-
Peng Li authored
Some macros are defined but unused, so remove them. Signed-off-by:
Peng Li <lipeng321@huawei.com> Signed-off-by:
-
Huazhong Tan authored
Parameter vf in hclge_vf_rate_param_check() is unused now, so remove it. Signed-off-by:
-
Huazhong Tan authored
Since hns3_uninit_all_ring() only returns 0, so remove this redundant return value and function declaration in hns3_enet.h. Signed-off-by:
-
Peng Li authored
The type of parameter mpf_bd_num and pf_bd_num in hclge_query_bd_num() should be u32* instead of int*, so change them. Signed-off-by:
-
Peng Li authored
The type of parameters in hclge_parse_speed() should be unsigned type, so change them. Signed-off-by:
-
Jiaran Zhang authored
Fix an issue where the formatting symbol of the formatting input and output function does not match the actual type. Signed-off-by:
Jiaran Zhang <zhangjiaran@huawei.com> Signed-off-by:
-
Yufeng Mo authored
In macro definitions, parentheses are unnecessary in some cases, such as the calling parameter of a function, the left variable of the equal sign, and so on. So remove these unnecessary parentheses according to these rules. Signed-off-by:
Yufeng Mo <moyufeng@huawei.com> Signed-off-by:
-
Peng Li authored
To make the code more readable, this patch adds a definition for the magic number 126 used for the default shaper param ir_b, and rename macro DIVISOR_IR_B_126. No functional change. Signed-off-by:
-
Jian Shen authored
Since the real tx queue number and real rx queue number always be updated when netdev opens, it's redundant to call hclge_client_setup_tc to do the same thing. So remove it. Signed-off-by:
-
Yonglong Liu authored
queue_id, qset_id and other IDs are unsigned type, so modify the corresponding local variables' type in hclge_dbg_dump_tm_map() from signed to unsigned. kstrtouint() and the print format should be updated as well. Signed-off-by:
Yonglong Liu <liuyonglong@huawei.com> Signed-off-by:
-
David S. Miller authored
Wei Wang says: ==================== implement kthread based napi polle The idea of moving the napi poll process out of softirq context to a kernel thread based context is not new. Paolo Abeni and Hannes Frederic Sowa have proposed patches to move napi poll to kthread back in 2016. And Felix Fietkau has also proposed patches of similar ideas to use workqueue to process napi poll just a few weeks ago. The main reason we'd like to push forward with this idea is that the scheduler has poor visibility into cpu cycles spent in softirq context, and is not able to make optimal scheduling decisions of the user threads. For example, we see in one of the application benchmark where network load is high, the CPUs handling network softirqs has ~80% cpu util. And user threads are still scheduled on those CPUs, despite other more idle cpus available in the system. And we see very high tail latencies. In this case, we have to explicitly pin away user threads from the CPUs handling network softirqs to ensure good performance. With napi poll moved to kthread, scheduler is in charge of scheduling both the kthreads handling network load, and the user threads, and is able to make better decisions. In the previous benchmark, if we do this and we pin the kthreads processing napi poll to specific CPUs, scheduler is able to schedule user threads away from these CPUs automatically. And the reason we prefer 1 kthread per napi, instead of 1 workqueue entity per host, is that kthread is more configurable than workqueue, and we could leverage existing tuning tools for threads, like taskset, chrt, etc to tune scheduling class and cpu set, etc. Another reason is if we eventually want to provide busy poll feature using kernel threads for napi poll, kthread seems to be more suitable than workqueue. Furthermore, for large platforms with 2 NICs attached to 2 sockets, kthread is more flexible to be pinned to different sets of CPUs. In this patch series, I revived Paolo and Hannes's patch in 2016 and made modifications. Then there are changes proposed by Felix, Jakub, Paolo and myself on top of those, with suggestions from Eric Dumazet. In terms of performance, I ran tcp_rr tests with 1000 flows with various request/response sizes, with RFS/RPS disabled, and compared performance between softirq vs kthread vs workqueue (patchset proposed by Felix Fietkau). Host has 56 hyper threads and 100Gbps nic, 8 rx queues and only 1 numa node. All threads are unpinned. req/resp QPS 50%tile 90%tile 99%tile 99.9%tile softirq 1B/1B 2.75M 337us 376us 1.04ms 3.69ms kthread 1B/1B 2.67M 371us 408us 455us 550us workq 1B/1B 2.56M 384us 435us 673us 822us softirq 5KB/5KB 1.46M 678us 750us 969us 2.78ms kthread 5KB/5KB 1.44M 695us 789us 891us 1.06ms workq 5KB/5KB 1.34M 720us 905us 1.06ms 1.57ms softirq 1MB/1MB 11.0K 79ms 166ms 306ms 630ms kthread 1MB/1MB 11.0K 75ms 177ms 303ms 596ms workq 1MB/1MB 11.0K 79ms 180ms 303ms 587ms When running workqueue implementation, I found the number of threads used is usually twice as much as kthread implementation. This probably introduces higher scheduling cost, which results in higher tail latencies in most cases. I also ran an application benchmark, which performs fixed qps remote SSD read/write operations, with various sizes. Again, both with RFS/RPS disabled. The result is as follows: op_size QPS 50%tile 95%tile 99%tile 99.9%tile softirq 4K 572.6K 385us 1.5ms 3.16ms 6.41ms kthread 4K 572.6K 390us 803us 2.21ms 6.83ms workq 4k 572.6K 384us 763us 3.12ms 6.87ms softirq 64K 157.9K 736us 1.17ms 3.40ms 13.75ms kthread 64K 157.9K 745us 1.23ms 2.76ms 9.87ms workq 64K 157.9K 746us 1.23ms 2.76ms 9.96ms softirq 1M 10.98K 2.03ms 3.10ms 3.7ms 11.56ms kthread 1M 10.98K 2.13ms 3.21ms 4.02ms 13.3ms workq 1M 10.98K 2.13ms 3.20ms 3.99ms 14.12ms In this set of tests, the latency is predominant by the SSD operation. Also, the user threads are much busier compared to tcp_rr tests. We have to pin the kthreads/workqueue threads to limit to a few CPUs, to not disturb user threads, and provide some isolation. Changes since v9: Small change in napi_poll() in patch 1. Split napi_kthread_stop() functionality to add separately in napi_disable() and netif_napi_del() in patch 2. Add description for napi_set_threaded() and return dev->threaded when dev->napi_list is empty for threaded sysfs in patch 3. Changes since v8: Added description for threaded param in struct net_device in patch 2. Changes since v7: Break napi_set_threaded() into 2 parts, one to create kthread called from netif_napi_add(), the other to set threaded bit in napi_enable(), to get rid of inconsistency through all napi in 1 dev. Added documentation for /sys/class/net/<dev>/threaded. Changes since v6: Added memory barrier in napi_set_threaded(). Changed /sys/class/net/<dev>/thread to a ternary value. Change dev->threaded to a bit instead of bool. Changes since v5: Removed ASSERT_RTNL() from napi_set_threaded() and removed rtnl_lock() operation from napi_enable(). Changes since v4: Recorded the threaded setting in dev and restore it in napi_enable(). Changes since v3: Merged and rearranged patches in a logical order for easier review. Changed sysfs control to be per device. Changes since v2: Corrected typo in patch 1, and updated the cover letter with more detailed and updated test results. Changes since v1: Replaced kthread_create() with kthread_run() in patch 5 as suggested by Felix Fietkau. Changes since RFC: Renamed the kthreads to be napi/<dev>-<napi_id> in patch 5 as suggested by Hannes Frederic Sowa. ==================== Signed-off-by: -
Wei Wang authored
This patch adds a new sysfs attribute to the network device class. Said attribute provides a per-device control to enable/disable the threaded mode for all the napi instances of the given network device, without the need for a device up/down. User sets it to 1 or 0 to enable or disable threaded mode. Note: when switching between threaded and the current softirq based mode for a napi instance, it will not immediately take effect if the napi is currently being polled. The mode switch will happen for the next time napi_schedule() is called. Co-developed-by:
Paolo Abeni <pabeni@redhat.com> Signed-off-by:
Hannes Frederic Sowa <hannes@stressinduktion.org> Signed-off-by:
Felix Fietkau <nbd@nbd.name> Signed-off-by:
Wei Wang <weiwan@google.com> Reviewed-by:
-
Wei Wang authored
This patch allows running each napi poll loop inside its own kernel thread. The kthread is created during netif_napi_add() if dev->threaded is set. And threaded mode is enabled in napi_enable(). We will provide a way to set dev->threaded and enable threaded mode without a device up/down in the following patch. Once that threaded mode is enabled and the kthread is started, napi_schedule() will wake-up such thread instead of scheduling the softirq. The threaded poll loop behaves quite likely the net_rx_action, but it does not have to manipulate local irqs and uses an explicit scheduling point based on netdev_budget. Co-developed-by:
-
Felix Fietkau authored
This commit introduces a new function __napi_poll() which does the main logic of the existing napi_poll() function, and will be called by other functions in later commits. This idea and implementation is done by Felix Fietkau <nbd@nbd.name> and is proposed as part of the patch to move napi work to work_queue context. This commit by itself is a code restructure. Signed-off-by:
-
Rafał Miłecki authored
BCM4908 SoCs family uses Ethernel controller that includes UniMAC but uses different DMA engine (than other controllers) and requires different programming. Signed-off-by:
Rafał Miłecki <rafal@milecki.pl> Signed-off-by:
-
Rafał Miłecki authored
BCM4908 is a family of SoCs with integrated Ethernet controller. Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec-nextDavid S. Miller authored
Steffen Klassert says: ==================== pull request (net-next): ipsec-next 2021-02-09 1) Support TSO on xfrm interfaces. From Eyal Birger. 2) Variable calculation simplifications in esp4/esp6. From Jiapeng Chong / Jiapeng Zhong. 3) Fix a return code in xfrm_do_migrate. From Zheng Yongjun. Please pull or let me know if there are problems. ==================== Signed-off-by: -
Jay Vosburgh authored
Provide documentation for src_valid_mark sysctl, which was added in commit 28f6aeea ("net: restore ip source validation"). Signed-off-by:
Jay Vosburgh <jay.vosburgh@canonical.com> Signed-off-by:
-
Michael Walle authored
It is nowhere used in the kernel. It also seems to be lacking the proper fiber advertise flags. Remove it. Signed-off-by:
Michael Walle <michael@walle.cc> Acked-by:
Florian Fainelli <f.fainelli@gmail.com> Reviewed-by:
Andrew Lunn <andrew@lunn.ch> Signed-off-by:
-
Michael Walle authored
genphy_read_status() is already the default for the .read_status() op. Drop the unnecessary references. Signed-off-by:
-
David S. Miller authored
net: Add support for route offload failure notifications Ido Schimmel says: ==================== This is a complementary series to the one merged in commit 389cb1ec ("Merge branch 'add-notifications-when-route-hardware-flags-change'"). The previous series added RTM_NEWROUTE notifications to user space whenever a route was successfully installed in hardware or when its state in hardware changed. This allows routing daemons to delay advertisement of routes until they are installed in hardware. However, if route installation failed, a routing daemon will wait indefinitely for a notification that will never come. The aim of this series is to provide a failure notification via a new flag (RTM_F_OFFLOAD_FAILED) in the RTM_NEWROUTE message. Upon such a notification a routing daemon may decide to withdraw the route from the FIB. Series overview: Patch #1 adds the new RTM_F_OFFLOAD_FAILED flag Patches #2-#3 and #4-#5 add failure notifications to IPv4 and IPv6, respectively Patches #6-#8 teach netdevsim to fail route installation via a new knob in debugfs Patch #9 extends mlxsw to mark routes with the new flag Patch #10 adds test cases for the new notification over netdevsim ==================== Signed-off-by:
-
Amit Cohen authored
Add cases to verify that when debugfs variable "fail_route_offload" is set, notification with "rt_offload_failed" flag is received. Extend the existing cases to verify that when sysctl "fib_notify_on_flag_change" is set to 2, the kernel emits notifications only for failed route installation. $ ./fib_notifications.sh TEST: IPv4 route addition [ OK ] TEST: IPv4 route deletion [ OK ] TEST: IPv4 route replacement [ OK ] TEST: IPv4 route offload failed [ OK ] TEST: IPv6 route addition [ OK ] TEST: IPv6 route deletion [ OK ] TEST: IPv6 route replacement [ OK ] TEST: IPv6 route offload failed [ OK ] Signed-off-by:
Amit Cohen <amcohen@nvidia.com> Signed-off-by:
Ido Schimmel <idosch@nvidia.com> Signed-off-by:
-
Amit Cohen authored
When FIB_EVENT_ENTRY_{REPLACE, APPEND} are triggered and route insertion fails, FIB abort is triggered. After aborting, set the appropriate hardware flag to make the kernel emit RTM_NEWROUTE notification with RTM_F_OFFLOAD_FAILED flag. Signed-off-by: -
Amit Cohen authored
Add "fail_route_offload" flag to disallow offloading routes. It is needed to test "offload failed" notifications. Create the flag as part of nsim_fib_create() under fib directory and set it to false by default. When FIB_EVENT_ENTRY_{REPLACE, APPEND} are triggered and "fail_route_offload" value is true, set the appropriate hardware flag to make the kernel emit RTM_NEWROUTE notification with RTM_F_OFFLOAD_FAILED flag. Signed-off-by: -
Ido Schimmel authored
Initialize the dummy FIB offload module after debugfs, so that the FIB module could create its own directory there. Signed-off-by:
-
Amit Cohen authored
The next patch will add the ability to fail route offload controlled by debugfs variable called "fail_route_offload". If we vetoed the addition, we might get a delete or append notification for a route we do not have. Therefore, do not warn if route was not found. Signed-off-by:
-
Amit Cohen authored
Add the value '2' to 'fib_notify_on_flag_change' to allow sending notifications only for failed route installation. Separate value is added for such notifications because there are less of them, so they do not impact performance and some users will find them more important. Signed-off-by:
-
Amit Cohen authored
After installing a route to the kernel, user space receives an acknowledgment, which means the route was installed in the kernel, but not necessarily in hardware. The asynchronous nature of route installation in hardware can lead to a routing daemon advertising a route before it was actually installed in hardware. This can result in packet loss or mis-routed packets until the route is installed in hardware. To avoid such cases, previous patch set added the ability to emit RTM_NEWROUTE notifications whenever RTM_F_OFFLOAD/RTM_F_TRAP flags are changed, this behavior is controlled by sysctl. With the above mentioned behavior, it is possible to know from user-space if the route was offloaded, but if the offload fails there is no indication to user-space. Following a failure, a routing daemon will wait indefinitely for a notification that will never come. This patch adds an "offload_failed" indication to IPv6 routes, so that users will have better visibility into the offload process. 'struct fib6_info' is extended with new field that indicates if route offload failed. Note that the new field is added using unused bit and therefore there is no need to increase struct size. Signed-off-by:
-
Amit Cohen authored
Add the value '2' to 'fib_notify_on_flag_change' to allow sending notifications only for failed route installation. Separate value is added for such notifications because there are less of them, so they do not impact performance and some users will find them more important. Signed-off-by:
-
Amit Cohen authored
After installing a route to the kernel, user space receives an acknowledgment, which means the route was installed in the kernel, but not necessarily in hardware. The asynchronous nature of route installation in hardware can lead to a routing daemon advertising a route before it was actually installed in hardware. This can result in packet loss or mis-routed packets until the route is installed in hardware. To avoid such cases, previous patch set added the ability to emit RTM_NEWROUTE notifications whenever RTM_F_OFFLOAD/RTM_F_TRAP flags are changed, this behavior is controlled by sysctl. With the above mentioned behavior, it is possible to know from user-space if the route was offloaded, but if the offload fails there is no indication to user-space. Following a failure, a routing daemon will wait indefinitely for a notification that will never come. This patch adds an "offload_failed" indication to IPv4 routes, so that users will have better visibility into the offload process. 'struct fib_alias', and 'struct fib_rt_info' are extended with new field that indicates if route offload failed. Note that the new field is added using unused bit and therefore there is no need to increase structs size. Signed-off-by:
-
Amit Cohen authored
The flag indicates to user space that route offload failed. Previous patch set added the ability to emit RTM_NEWROUTE notifications whenever RTM_F_OFFLOAD/RTM_F_TRAP flags are changed, but if the offload fails there is no indication to user-space. The flag will be used in subsequent patches by netdevsim and mlxsw to indicate to user space that route offload failed, so that users will have better visibility into the offload process. Signed-off-by:
-
Tony Nguyen authored
The ice documentation has not been updated since the initial commits of the driver. Update the documentation with features and information that are now available. Signed-off-by:Tony Nguyen <anthony.l.nguyen@intel.com>
-
Tony Nguyen authored
Currently if the driver is unable to get all the MSI-X vectors it wants, it falls back to the minimum configuration which equates to a single Tx/Rx traffic queue pair. Instead of using the minimum configuration, if given more vectors than the minimum, utilize those vectors for additional traffic queues after accounting for other interrupts. Signed-off-by:
Tony Brelinski <tonyx.brelinski@intel.com>
-
Mitch Williams authored
This message indicates an error on close, not open. Signed-off-by:
Mitch Williams <mitch.a.williams@intel.com> Tested-by:
-
Bruce Allan authored
Casting a void * rvalue in an assignment is unnecessary in C; remove the casts. Signed-off-by:
Bruce Allan <bruce.w.allan@intel.com> Tested-by:
-
Chinh T Cao authored
Refactor the DCB related variables out of the ice_port_info_struct. The goal is to make the ice_port_info struct cleaner. Signed-off-by:
Chinh T Cao <chinh.t.cao@intel.com> Co-developed-by:
Dave Ertman <david.m.ertman@intel.com> Signed-off-by:
-
Jesse Brandeburg authored
The writeback enable logic was incorrectly implemented (due to misunderstanding what the side effects of the implementation would be during polling). Fix this logic issue, while implementing a new feature allowing the user to control the writeback frequency using the knobs for controlling interrupt throttling that we already have. Basically if you leave adaptive interrupts enabled, the writeback frequency will be varied even if busy_polling or if napi-poll is in use. If the interrupt rates are set to a fixed value by ethtool -C and adaptive is off, the driver will allow the user-set interrupt rate to guide how frequently the hardware will complete descriptors to the driver. Effectively the user will get a control over the hardware efficiency, allowing the choice between immediate interrupts or delayed up to a maximum of the interrupt rate, even when interrupts are disabled during polling. Signed-off-by:
Jesse Brandeburg <jesse.brandeburg@intel.com> Co-developed-by:
Brett Creeley <brett.creeley@intel.com> Signed-off-by:
-
