Public bug reported: SRU Justification
[Impact] net: mana: Fix perf regression: remove rx_cqes, tx_cqes counters https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net.git/commit/?id=1919b39fc6eabb9a6f9a51706ff6d03865f5df29 It resolves a big perf regression. More details: The apc->eth_stats.rx_cqes is one per NIC (vport), and it's on the frequent and parallel code path of all queues. So, r/w into this single shared variable by many threads on different CPUs creates a lot caching and memory overhead, hence perf regression. And, it's not accurate due to the high volume concurrent r/w. For example, a workload is iperf with 128 threads, and with RPS enabled. We saw perf regression of 25% with the previous patch adding the counters. And this patch eliminates the regression. Since the error path of mana_poll_rx_cq() already has warnings, so keeping the counter and convert it to a per-queue variable is not necessary. So, just remove this counter from this high frequency code path. Also, remove the tx_cqes counter for the same reason. We have warnings & other counters for errors on that path, and don't need to count every normal cqe processing. [Test Plan] MSFT tested [Regression potential] Counters are disappearing that may be in use by user space programs. [Other Info] SF: #00361807 ** Affects: linux-azure (Ubuntu) Importance: Undecided Status: Fix Released ** Affects: linux-azure (Ubuntu Lunar) Importance: Medium Assignee: Tim Gardner (timg-tpi) Status: In Progress ** Package changed: linux (Ubuntu) => linux-azure (Ubuntu) ** Also affects: linux-azure (Ubuntu Lunar) Importance: Undecided Status: New ** Changed in: linux-azure (Ubuntu Lunar) Importance: Undecided => Medium ** Changed in: linux-azure (Ubuntu Lunar) Status: New => In Progress ** Changed in: linux-azure (Ubuntu Lunar) Assignee: (unassigned) => Tim Gardner (timg-tpi) ** Changed in: linux-azure (Ubuntu) Status: New => Fix Released -- You received this bug notification because you are a member of Kernel Packages, which is subscribed to linux-azure in Ubuntu. https://bugs.launchpad.net/bugs/2022940 Title: Azure: Fix perf regression: remove rx_cqes, tx_cqes counters for MANA Status in linux-azure package in Ubuntu: Fix Released Status in linux-azure source package in Lunar: In Progress Bug description: SRU Justification [Impact] net: mana: Fix perf regression: remove rx_cqes, tx_cqes counters https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net.git/commit/?id=1919b39fc6eabb9a6f9a51706ff6d03865f5df29 It resolves a big perf regression. More details: The apc->eth_stats.rx_cqes is one per NIC (vport), and it's on the frequent and parallel code path of all queues. So, r/w into this single shared variable by many threads on different CPUs creates a lot caching and memory overhead, hence perf regression. And, it's not accurate due to the high volume concurrent r/w. For example, a workload is iperf with 128 threads, and with RPS enabled. We saw perf regression of 25% with the previous patch adding the counters. And this patch eliminates the regression. Since the error path of mana_poll_rx_cq() already has warnings, so keeping the counter and convert it to a per-queue variable is not necessary. So, just remove this counter from this high frequency code path. Also, remove the tx_cqes counter for the same reason. We have warnings & other counters for errors on that path, and don't need to count every normal cqe processing. [Test Plan] MSFT tested [Regression potential] Counters are disappearing that may be in use by user space programs. [Other Info] SF: #00361807 To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux-azure/+bug/2022940/+subscriptions -- Mailing list: https://launchpad.net/~kernel-packages Post to : kernel-packages@lists.launchpad.net Unsubscribe : https://launchpad.net/~kernel-packages More help : https://help.launchpad.net/ListHelp
