From 0d6526f82c3cdefcca47f73f5fc08dc6f335eac6 Mon Sep 17 00:00:00 2001 From: Tim Chen Date: Mon, 21 Sep 2026 17:37:22 -0700 Subject: sched/cache: Keep nr_pref_llc_running in the runnable domain, to fix LLC mis-scheduling bug alb_break_llc() decides whether to break LLC preference during active load balance. It does so by testing that every runnable fair task on the source rq prefers its LLC: env->src_rq->nr_pref_llc_running == env->src_rq->cfs.h_nr_runnable But the two counters cover different sets. nr_pref_llc_running is updated in account_llc_enqueue()/account_llc_dequeue(), next to cfs_rq->nr_queued, so it follows queued tasks. h_nr_runnable is updated in set_delayed()/ clear_delayed() and drops delay-dequeued tasks. So under DELAY_DEQUEUE, a preferring task that goes to sleep stays counted in nr_pref_llc_running while h_nr_runnable falls. The equality then breaks, alb_break_llc() returns false, and active balance is free to pull a task off its preferred LLC. Active balance only moves runnable tasks, and this is the only LLC check it consults: once the stopper runs, LBF_ACTIVE_LB skips the per-task test in can_migrate_task(). The runnable set is the one we want. Fix it on the counter side. A task should be counted in nr_pref_llc_running exactly while it is both queued on its preferred LLC (pref_llc_queued) and runnable (!sched_delayed). Define that membership once in task_pref_llc_runnable(), and adjust the counter only through pref_llc_running_inc()/pref_llc_running_dec() from the four sites that change either input: account_llc_enqueue(), account_llc_dequeue(), set_delayed() and clear_delayed(). Gating every update on the same predicate keeps the delay, wake and dequeue paths from double-counting or underflowing; see the comments at those sites for the ordering. nr_llc_running and sd->llc_counts are not touched and stay on queued semantics. Fixes: 714059f79ff0 ("sched/cache: Handle moving single tasks to/from their preferred LLC") Closes: https://lore.kernel.org/lkml/20260827135000.735138-1-zhanxusheng@xiaomi.com/ Reported-by: Zhan Xusheng Suggested-by: Chen Yu Signed-off-by: Tim Chen Signed-off-by: Peter Zijlstra (Intel) Signed-off-by: Ingo Molnar Reviewed-by: Kayra Cizmeci Cc: # v7.2.x Link: https://patch.msgid.link/06af61afedac32e6477f57feb4d658f6c411c3af.1790035273.git.tim.c.chen@linux.intel.com --- drivers/net/macvtap.c | 256 ++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 256 insertions(+) create mode 100644 drivers/net/macvtap.c (limited to 'drivers/net/macvtap.c') diff --git a/drivers/net/macvtap.c b/drivers/net/macvtap.c new file mode 100644 index 000000000..cc975dfb7 --- /dev/null +++ b/drivers/net/macvtap.c @@ -0,0 +1,256 @@ +// SPDX-License-Identifier: GPL-2.0-only +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include +#include + +struct macvtap_dev { + struct macvlan_dev vlan; + struct tap_dev tap; +}; + +/* + * Variables for dealing with macvtaps device numbers. + */ +static dev_t macvtap_major; + +static const struct ns_common *macvtap_net_namespace(const struct device *d) +{ + const struct net_device *dev = to_net_dev(d->parent); + + return to_ns_common(dev_net(dev)); +} + +static struct class macvtap_class = { + .name = "macvtap", + .ns_type = &net_ns_type_operations, + .namespace = macvtap_net_namespace, +}; +static struct cdev macvtap_cdev; + +#define TUN_OFFLOADS (NETIF_F_HW_CSUM | NETIF_F_TSO_ECN | NETIF_F_TSO | \ + NETIF_F_TSO6) + +static void macvtap_count_tx_dropped(struct tap_dev *tap) +{ + struct macvtap_dev *vlantap = container_of(tap, struct macvtap_dev, tap); + struct macvlan_dev *vlan = &vlantap->vlan; + + this_cpu_inc(vlan->pcpu_stats->tx_dropped); +} + +static void macvtap_count_rx_dropped(struct tap_dev *tap) +{ + struct macvtap_dev *vlantap = container_of(tap, struct macvtap_dev, tap); + struct macvlan_dev *vlan = &vlantap->vlan; + + macvlan_count_rx(vlan, 0, 0, 0); +} + +static void macvtap_update_features(struct tap_dev *tap, + netdev_features_t features) +{ + struct macvtap_dev *vlantap = container_of(tap, struct macvtap_dev, tap); + struct macvlan_dev *vlan = &vlantap->vlan; + + vlan->set_features = features; + netdev_update_features(vlan->dev); +} + +static int macvtap_newlink(struct net_device *dev, + struct rtnl_newlink_params *params, + struct netlink_ext_ack *extack) +{ + struct macvtap_dev *vlantap = netdev_priv(dev); + int err; + + INIT_LIST_HEAD(&vlantap->tap.queue_list); + + /* Since macvlan supports all offloads by default, make + * tap support all offloads also. + */ + vlantap->tap.tap_features = TUN_OFFLOADS; + + /* Register callbacks for rx/tx drops accounting and updating + * net_device features + */ + vlantap->tap.count_tx_dropped = macvtap_count_tx_dropped; + vlantap->tap.count_rx_dropped = macvtap_count_rx_dropped; + vlantap->tap.update_features = macvtap_update_features; + + err = netdev_rx_handler_register(dev, tap_handle_frame, &vlantap->tap); + if (err) + return err; + + /* Don't put anything that may fail after macvlan_common_newlink + * because we can't undo what it does. + */ + err = macvlan_common_newlink(dev, params, extack); + if (err) { + netdev_rx_handler_unregister(dev); + return err; + } + + vlantap->tap.dev = vlantap->vlan.dev; + + return 0; +} + +static void macvtap_dellink(struct net_device *dev, + struct list_head *head) +{ + struct macvtap_dev *vlantap = netdev_priv(dev); + + netdev_rx_handler_unregister(dev); + tap_del_queues(&vlantap->tap); + macvlan_dellink(dev, head); +} + +static void macvtap_setup(struct net_device *dev) +{ + macvlan_common_setup(dev); + dev->tx_queue_len = TUN_READQ_SIZE; +} + +static struct net *macvtap_link_net(const struct net_device *dev) +{ + return dev_net(macvlan_dev_real_dev(dev)); +} + +static struct rtnl_link_ops macvtap_link_ops __read_mostly = { + .kind = "macvtap", + .setup = macvtap_setup, + .newlink = macvtap_newlink, + .dellink = macvtap_dellink, + .get_link_net = macvtap_link_net, + .priv_size = sizeof(struct macvtap_dev), +}; + +static int macvtap_device_event(struct notifier_block *unused, + unsigned long event, void *ptr) +{ + struct net_device *dev = netdev_notifier_info_to_dev(ptr); + struct macvtap_dev *vlantap; + struct device *classdev; + dev_t devt; + int err; + char tap_name[IFNAMSIZ]; + + if (dev->rtnl_link_ops != &macvtap_link_ops) + return NOTIFY_DONE; + + snprintf(tap_name, IFNAMSIZ, "tap%d", dev->ifindex); + vlantap = netdev_priv(dev); + + switch (event) { + case NETDEV_REGISTER: + /* Create the device node here after the network device has + * been registered but before register_netdevice has + * finished running. + */ + err = tap_get_minor(macvtap_major, &vlantap->tap); + if (err) + return notifier_from_errno(err); + + devt = MKDEV(MAJOR(macvtap_major), vlantap->tap.minor); + classdev = device_create(&macvtap_class, &dev->dev, devt, + dev, "%s", tap_name); + if (IS_ERR(classdev)) { + tap_free_minor(macvtap_major, &vlantap->tap); + return notifier_from_errno(PTR_ERR(classdev)); + } + err = sysfs_create_link(&dev->dev.kobj, &classdev->kobj, + tap_name); + if (err) + return notifier_from_errno(err); + break; + case NETDEV_UNREGISTER: + /* vlan->minor == 0 if NETDEV_REGISTER above failed */ + if (vlantap->tap.minor == 0) + break; + sysfs_remove_link(&dev->dev.kobj, tap_name); + devt = MKDEV(MAJOR(macvtap_major), vlantap->tap.minor); + device_destroy(&macvtap_class, devt); + tap_free_minor(macvtap_major, &vlantap->tap); + break; + case NETDEV_CHANGE_TX_QUEUE_LEN: + if (tap_queue_resize(&vlantap->tap)) + return NOTIFY_BAD; + break; + } + + return NOTIFY_DONE; +} + +static struct notifier_block macvtap_notifier_block __read_mostly = { + .notifier_call = macvtap_device_event, +}; + +static int __init macvtap_init(void) +{ + int err; + + err = tap_create_cdev(&macvtap_cdev, &macvtap_major, "macvtap", + THIS_MODULE); + if (err) + goto out1; + + err = class_register(&macvtap_class); + if (err) + goto out2; + + err = register_netdevice_notifier(&macvtap_notifier_block); + if (err) + goto out3; + + err = macvlan_link_register(&macvtap_link_ops); + if (err) + goto out4; + + return 0; + +out4: + unregister_netdevice_notifier(&macvtap_notifier_block); +out3: + class_unregister(&macvtap_class); +out2: + tap_destroy_cdev(macvtap_major, &macvtap_cdev); +out1: + return err; +} +module_init(macvtap_init); + +static void __exit macvtap_exit(void) +{ + rtnl_link_unregister(&macvtap_link_ops); + unregister_netdevice_notifier(&macvtap_notifier_block); + class_unregister(&macvtap_class); + tap_destroy_cdev(macvtap_major, &macvtap_cdev); +} +module_exit(macvtap_exit); + +MODULE_ALIAS_RTNL_LINK("macvtap"); +MODULE_DESCRIPTION("MAC-VLAN based tap driver"); +MODULE_AUTHOR("Arnd Bergmann "); +MODULE_LICENSE("GPL"); -- cgit v1.3.1