From a5f7a5bb3b7f28ba7e4fa246775b29a0e5537255 Mon Sep 17 00:00:00 2001 From: Jinke Han Date: Tue, 8 Sep 2026 15:37:42 +0800 Subject: x86/kprobes: Fix crash when probing CS CALL instructions When using eBPF to probe CS CALL instructions within a function, a crash can be triggered. The eBPF tool probes offset 257 of the __hrtimer_run_queues() function: <__hrtimer_run_queues+249>: nopl 0x0(%rax,%rax,1) <__hrtimer_run_queues+254>: mov %r14,%rdi <__hrtimer_run_queues+257>: cs call <__x86_indirect_thunk_r12> <__hrtimer_run_queues+263>: mov %eax,%r12d <__hrtimer_run_queues+266>: xchg %ax,%ax <__hrtimer_run_queues+268>: mov %r13,%rdi Which triggers this crash: BUG: unable to handle page fault for address: 00000000000f41c9 #PF: supervisor write access in kernel mode #PF: error_code(0x0002) - not-present page PGD 0 P4D 0 Oops: 0002 [#1] SMP NOPTI CPU: 1 PID: 0 Comm: swapper/1 Kdump: loaded Tainted: P RIP: 0010:__hrtimer_run_queues+0x106/0x230 Note that __hrtimer_run_queues+0x106 is __hrtimer_run_queues+262, which is at the 6th byte of the above CS CALL instruction. Since the CS CALL instruction occupies 6 bytes, the exception occurred in the middle of that call instruction. The root cause is that when using eBPF tools to probe in the middle of a function, a kprobe with INT3 is used as the underlying implementation. During single-step emulation of the original CALL instruction, int3_emulate_call() assumes that the probed CALL instruction is 5 bytes long. However, the actual CS-prefixed CALL instruction occupies 6 bytes, so it constructs an incorrect exception return address. When the CPU returns from the kprobe handler, the next instruction to be executed is at the address of the last byte of that CS CALL instruction. Coincidentally, starting from that address, the CPU fetches and decodes a completely different instruction, which ultimately triggers a kernel crash. Fix the issue by using the actual instruction length obtained from the instruction decoder when constructing the exception return address, rather than relying on the hardcoded CALL_INSN_SIZE macro. [ mingo: Refined the changelog ] Fixes: 6256e668b7af ("x86/kprobes: Use int3 instead of debug trap for single-step") Suggested-by: Masami Hiramatsu (Google) Signed-off-by: Jinke Han Signed-off-by: Ingo Molnar Reviewed-by: Masami Hiramatsu (Google) Acked-by: Yafang Shao Acked-by: Borislav Petkov Cc: Peter Zijlstra Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000 --- drivers/net/netdevsim/hwstats.c | 480 ++++++++++++++++++++++++++++++++++++++++ 1 file changed, 480 insertions(+) create mode 100644 drivers/net/netdevsim/hwstats.c (limited to 'drivers/net/netdevsim/hwstats.c') diff --git a/drivers/net/netdevsim/hwstats.c b/drivers/net/netdevsim/hwstats.c new file mode 100644 index 000000000..14234a4cc --- /dev/null +++ b/drivers/net/netdevsim/hwstats.c @@ -0,0 +1,480 @@ +// SPDX-License-Identifier: GPL-2.0 + +#include + +#include "netdevsim.h" + +#define NSIM_DEV_HWSTATS_TRAFFIC_MS 100 + +static struct list_head * +nsim_dev_hwstats_get_list_head(struct nsim_dev_hwstats *hwstats, + enum netdev_offload_xstats_type type) +{ + switch (type) { + case NETDEV_OFFLOAD_XSTATS_TYPE_L3: + return &hwstats->l3_list; + } + + WARN_ON_ONCE(1); + return NULL; +} + +static void nsim_dev_hwstats_traffic_bump(struct nsim_dev_hwstats *hwstats, + enum netdev_offload_xstats_type type) +{ + struct nsim_dev_hwstats_netdev *hwsdev; + struct list_head *hwsdev_list; + + hwsdev_list = nsim_dev_hwstats_get_list_head(hwstats, type); + if (WARN_ON(!hwsdev_list)) + return; + + list_for_each_entry(hwsdev, hwsdev_list, list) { + if (hwsdev->enabled) { + hwsdev->stats.rx_packets += 1; + hwsdev->stats.tx_packets += 2; + hwsdev->stats.rx_bytes += 100; + hwsdev->stats.tx_bytes += 300; + } + } +} + +static void nsim_dev_hwstats_traffic_work(struct work_struct *work) +{ + struct nsim_dev_hwstats *hwstats; + + hwstats = container_of(work, struct nsim_dev_hwstats, traffic_dw.work); + mutex_lock(&hwstats->hwsdev_list_lock); + nsim_dev_hwstats_traffic_bump(hwstats, NETDEV_OFFLOAD_XSTATS_TYPE_L3); + mutex_unlock(&hwstats->hwsdev_list_lock); + + schedule_delayed_work(&hwstats->traffic_dw, + msecs_to_jiffies(NSIM_DEV_HWSTATS_TRAFFIC_MS)); +} + +static struct nsim_dev_hwstats_netdev * +nsim_dev_hwslist_find_hwsdev(struct list_head *hwsdev_list, + int ifindex) +{ + struct nsim_dev_hwstats_netdev *hwsdev; + + list_for_each_entry(hwsdev, hwsdev_list, list) { + if (hwsdev->netdev->ifindex == ifindex) + return hwsdev; + } + + return NULL; +} + +static int nsim_dev_hwsdev_enable(struct nsim_dev_hwstats_netdev *hwsdev, + struct netlink_ext_ack *extack) +{ + if (hwsdev->fail_enable) { + hwsdev->fail_enable = false; + NL_SET_ERR_MSG_MOD(extack, "Stats enablement set to fail"); + return -ECANCELED; + } + + hwsdev->enabled = true; + return 0; +} + +static void nsim_dev_hwsdev_disable(struct nsim_dev_hwstats_netdev *hwsdev) +{ + hwsdev->enabled = false; + memset(&hwsdev->stats, 0, sizeof(hwsdev->stats)); +} + +static int +nsim_dev_hwsdev_report_delta(struct nsim_dev_hwstats_netdev *hwsdev, + struct netdev_notifier_offload_xstats_info *info) +{ + netdev_offload_xstats_report_delta(info->report_delta, &hwsdev->stats); + memset(&hwsdev->stats, 0, sizeof(hwsdev->stats)); + return 0; +} + +static void +nsim_dev_hwsdev_report_used(struct nsim_dev_hwstats_netdev *hwsdev, + struct netdev_notifier_offload_xstats_info *info) +{ + if (hwsdev->enabled) + netdev_offload_xstats_report_used(info->report_used); +} + +static int nsim_dev_hwstats_event_off_xstats(struct nsim_dev_hwstats *hwstats, + struct net_device *dev, + unsigned long event, void *ptr) +{ + struct netdev_notifier_offload_xstats_info *info; + struct nsim_dev_hwstats_netdev *hwsdev; + struct list_head *hwsdev_list; + int err = 0; + + info = ptr; + hwsdev_list = nsim_dev_hwstats_get_list_head(hwstats, info->type); + if (!hwsdev_list) + return 0; + + mutex_lock(&hwstats->hwsdev_list_lock); + + hwsdev = nsim_dev_hwslist_find_hwsdev(hwsdev_list, dev->ifindex); + if (!hwsdev) + goto out; + + switch (event) { + case NETDEV_OFFLOAD_XSTATS_ENABLE: + err = nsim_dev_hwsdev_enable(hwsdev, info->info.extack); + break; + case NETDEV_OFFLOAD_XSTATS_DISABLE: + nsim_dev_hwsdev_disable(hwsdev); + break; + case NETDEV_OFFLOAD_XSTATS_REPORT_USED: + nsim_dev_hwsdev_report_used(hwsdev, info); + break; + case NETDEV_OFFLOAD_XSTATS_REPORT_DELTA: + err = nsim_dev_hwsdev_report_delta(hwsdev, info); + break; + } + +out: + mutex_unlock(&hwstats->hwsdev_list_lock); + return err; +} + +static void nsim_dev_hwsdev_fini(struct nsim_dev_hwstats_netdev *hwsdev) +{ + dev_put(hwsdev->netdev); + kfree(hwsdev); +} + +static void +__nsim_dev_hwstats_event_unregister(struct nsim_dev_hwstats *hwstats, + struct net_device *dev, + enum netdev_offload_xstats_type type) +{ + struct nsim_dev_hwstats_netdev *hwsdev; + struct list_head *hwsdev_list; + + hwsdev_list = nsim_dev_hwstats_get_list_head(hwstats, type); + if (WARN_ON(!hwsdev_list)) + return; + + hwsdev = nsim_dev_hwslist_find_hwsdev(hwsdev_list, dev->ifindex); + if (!hwsdev) + return; + + list_del(&hwsdev->list); + nsim_dev_hwsdev_fini(hwsdev); +} + +static void nsim_dev_hwstats_event_unregister(struct nsim_dev_hwstats *hwstats, + struct net_device *dev) +{ + mutex_lock(&hwstats->hwsdev_list_lock); + __nsim_dev_hwstats_event_unregister(hwstats, dev, + NETDEV_OFFLOAD_XSTATS_TYPE_L3); + mutex_unlock(&hwstats->hwsdev_list_lock); +} + +static int nsim_dev_hwstats_event(struct nsim_dev_hwstats *hwstats, + struct net_device *dev, + unsigned long event, void *ptr) +{ + switch (event) { + case NETDEV_OFFLOAD_XSTATS_ENABLE: + case NETDEV_OFFLOAD_XSTATS_DISABLE: + case NETDEV_OFFLOAD_XSTATS_REPORT_USED: + case NETDEV_OFFLOAD_XSTATS_REPORT_DELTA: + return nsim_dev_hwstats_event_off_xstats(hwstats, dev, + event, ptr); + case NETDEV_UNREGISTER: + nsim_dev_hwstats_event_unregister(hwstats, dev); + break; + } + + return 0; +} + +static int nsim_dev_netdevice_event(struct notifier_block *nb, + unsigned long event, void *ptr) +{ + struct net_device *dev = netdev_notifier_info_to_dev(ptr); + struct nsim_dev_hwstats *hwstats; + int err = 0; + + hwstats = container_of(nb, struct nsim_dev_hwstats, netdevice_nb); + err = nsim_dev_hwstats_event(hwstats, dev, event, ptr); + if (err) + return notifier_from_errno(err); + + return NOTIFY_OK; +} + +static int +nsim_dev_hwstats_enable_ifindex(struct nsim_dev_hwstats *hwstats, + int ifindex, + enum netdev_offload_xstats_type type, + struct list_head *hwsdev_list) +{ + struct nsim_dev_hwstats_netdev *hwsdev; + struct nsim_dev *nsim_dev; + struct net_device *netdev; + struct net *net; + int err = 0; + + nsim_dev = container_of(hwstats, struct nsim_dev, hwstats); + net = nsim_dev_net(nsim_dev); + + rtnl_lock(); + mutex_lock(&hwstats->hwsdev_list_lock); + hwsdev = nsim_dev_hwslist_find_hwsdev(hwsdev_list, ifindex); + if (hwsdev) + goto out_unlock_list; + + netdev = dev_get_by_index(net, ifindex); + if (!netdev) { + err = -ENODEV; + goto out_unlock_list; + } + + hwsdev = kzalloc_obj(*hwsdev); + if (!hwsdev) { + err = -ENOMEM; + goto out_put_netdev; + } + + hwsdev->netdev = netdev; + list_add_tail(&hwsdev->list, hwsdev_list); + mutex_unlock(&hwstats->hwsdev_list_lock); + + if (netdev_offload_xstats_enabled(netdev, type)) { + nsim_dev_hwsdev_enable(hwsdev, NULL); + rtnl_offload_xstats_notify(netdev); + } + + rtnl_unlock(); + return err; + +out_put_netdev: + dev_put(netdev); +out_unlock_list: + mutex_unlock(&hwstats->hwsdev_list_lock); + rtnl_unlock(); + return err; +} + +static int +nsim_dev_hwstats_disable_ifindex(struct nsim_dev_hwstats *hwstats, + int ifindex, + enum netdev_offload_xstats_type type, + struct list_head *hwsdev_list) +{ + struct nsim_dev_hwstats_netdev *hwsdev; + int err = 0; + + rtnl_lock(); + mutex_lock(&hwstats->hwsdev_list_lock); + hwsdev = nsim_dev_hwslist_find_hwsdev(hwsdev_list, ifindex); + if (hwsdev) + list_del(&hwsdev->list); + mutex_unlock(&hwstats->hwsdev_list_lock); + + if (!hwsdev) { + err = -ENOENT; + goto unlock_out; + } + + if (netdev_offload_xstats_enabled(hwsdev->netdev, type)) { + netdev_offload_xstats_push_delta(hwsdev->netdev, type, + &hwsdev->stats); + rtnl_offload_xstats_notify(hwsdev->netdev); + } + nsim_dev_hwsdev_fini(hwsdev); + +unlock_out: + rtnl_unlock(); + return err; +} + +static int +nsim_dev_hwstats_fail_ifindex(struct nsim_dev_hwstats *hwstats, + int ifindex, + enum netdev_offload_xstats_type type, + struct list_head *hwsdev_list) +{ + struct nsim_dev_hwstats_netdev *hwsdev; + int err = 0; + + mutex_lock(&hwstats->hwsdev_list_lock); + + hwsdev = nsim_dev_hwslist_find_hwsdev(hwsdev_list, ifindex); + if (!hwsdev) { + err = -ENOENT; + goto err_hwsdev_list_unlock; + } + + hwsdev->fail_enable = true; + +err_hwsdev_list_unlock: + mutex_unlock(&hwstats->hwsdev_list_lock); + return err; +} + +enum nsim_dev_hwstats_do { + NSIM_DEV_HWSTATS_DO_DISABLE, + NSIM_DEV_HWSTATS_DO_ENABLE, + NSIM_DEV_HWSTATS_DO_FAIL, +}; + +struct nsim_dev_hwstats_fops { + enum nsim_dev_hwstats_do action; + enum netdev_offload_xstats_type type; +}; + +static ssize_t +nsim_dev_hwstats_do_write(struct file *file, + const char __user *data, + size_t count, loff_t *ppos) +{ + struct nsim_dev_hwstats *hwstats = file->private_data; + const struct nsim_dev_hwstats_fops *hwsfops; + struct list_head *hwsdev_list; + int ifindex; + int err; + + hwsfops = debugfs_get_aux(file); + + err = kstrtoint_from_user(data, count, 0, &ifindex); + if (err) + return err; + + hwsdev_list = nsim_dev_hwstats_get_list_head(hwstats, hwsfops->type); + if (WARN_ON(!hwsdev_list)) + return -EINVAL; + + switch (hwsfops->action) { + case NSIM_DEV_HWSTATS_DO_DISABLE: + err = nsim_dev_hwstats_disable_ifindex(hwstats, ifindex, + hwsfops->type, + hwsdev_list); + break; + case NSIM_DEV_HWSTATS_DO_ENABLE: + err = nsim_dev_hwstats_enable_ifindex(hwstats, ifindex, + hwsfops->type, + hwsdev_list); + break; + case NSIM_DEV_HWSTATS_DO_FAIL: + err = nsim_dev_hwstats_fail_ifindex(hwstats, ifindex, + hwsfops->type, + hwsdev_list); + break; + } + if (err) + return err; + + return count; +} + +static struct debugfs_short_fops debugfs_ops = { + .write = nsim_dev_hwstats_do_write, + .llseek = generic_file_llseek, +}; + +#define NSIM_DEV_HWSTATS_FOPS(ACTION, TYPE) \ + { \ + .action = ACTION, \ + .type = TYPE, \ + } + +static const struct nsim_dev_hwstats_fops nsim_dev_hwstats_l3_disable_fops = + NSIM_DEV_HWSTATS_FOPS(NSIM_DEV_HWSTATS_DO_DISABLE, + NETDEV_OFFLOAD_XSTATS_TYPE_L3); + +static const struct nsim_dev_hwstats_fops nsim_dev_hwstats_l3_enable_fops = + NSIM_DEV_HWSTATS_FOPS(NSIM_DEV_HWSTATS_DO_ENABLE, + NETDEV_OFFLOAD_XSTATS_TYPE_L3); + +static const struct nsim_dev_hwstats_fops nsim_dev_hwstats_l3_fail_fops = + NSIM_DEV_HWSTATS_FOPS(NSIM_DEV_HWSTATS_DO_FAIL, + NETDEV_OFFLOAD_XSTATS_TYPE_L3); + +#undef NSIM_DEV_HWSTATS_FOPS + +int nsim_dev_hwstats_init(struct nsim_dev *nsim_dev) +{ + struct nsim_dev_hwstats *hwstats = &nsim_dev->hwstats; + struct net *net = nsim_dev_net(nsim_dev); + int err; + + mutex_init(&hwstats->hwsdev_list_lock); + INIT_LIST_HEAD(&hwstats->l3_list); + + hwstats->netdevice_nb.notifier_call = nsim_dev_netdevice_event; + err = register_netdevice_notifier_net(net, &hwstats->netdevice_nb); + if (err) + goto err_mutex_destroy; + + hwstats->ddir = debugfs_create_dir("hwstats", nsim_dev->ddir); + if (IS_ERR(hwstats->ddir)) { + err = PTR_ERR(hwstats->ddir); + goto err_unregister_notifier; + } + + hwstats->l3_ddir = debugfs_create_dir("l3", hwstats->ddir); + if (IS_ERR(hwstats->l3_ddir)) { + err = PTR_ERR(hwstats->l3_ddir); + goto err_remove_hwstats_recursive; + } + + debugfs_create_file_aux("enable_ifindex", 0200, hwstats->l3_ddir, hwstats, + &nsim_dev_hwstats_l3_enable_fops, &debugfs_ops); + debugfs_create_file_aux("disable_ifindex", 0200, hwstats->l3_ddir, hwstats, + &nsim_dev_hwstats_l3_disable_fops, &debugfs_ops); + debugfs_create_file_aux("fail_next_enable", 0200, hwstats->l3_ddir, hwstats, + &nsim_dev_hwstats_l3_fail_fops, &debugfs_ops); + + INIT_DELAYED_WORK(&hwstats->traffic_dw, + &nsim_dev_hwstats_traffic_work); + schedule_delayed_work(&hwstats->traffic_dw, + msecs_to_jiffies(NSIM_DEV_HWSTATS_TRAFFIC_MS)); + return 0; + +err_remove_hwstats_recursive: + debugfs_remove_recursive(hwstats->ddir); +err_unregister_notifier: + unregister_netdevice_notifier_net(net, &hwstats->netdevice_nb); +err_mutex_destroy: + mutex_destroy(&hwstats->hwsdev_list_lock); + return err; +} + +static void nsim_dev_hwsdev_list_wipe(struct nsim_dev_hwstats *hwstats, + enum netdev_offload_xstats_type type) +{ + struct nsim_dev_hwstats_netdev *hwsdev, *tmp; + struct list_head *hwsdev_list; + + hwsdev_list = nsim_dev_hwstats_get_list_head(hwstats, type); + if (WARN_ON(!hwsdev_list)) + return; + + mutex_lock(&hwstats->hwsdev_list_lock); + list_for_each_entry_safe(hwsdev, tmp, hwsdev_list, list) { + list_del(&hwsdev->list); + nsim_dev_hwsdev_fini(hwsdev); + } + mutex_unlock(&hwstats->hwsdev_list_lock); +} + +void nsim_dev_hwstats_exit(struct nsim_dev *nsim_dev) +{ + struct nsim_dev_hwstats *hwstats = &nsim_dev->hwstats; + struct net *net = nsim_dev_net(nsim_dev); + + cancel_delayed_work_sync(&hwstats->traffic_dw); + debugfs_remove_recursive(hwstats->ddir); + unregister_netdevice_notifier_net(net, &hwstats->netdevice_nb); + nsim_dev_hwsdev_list_wipe(hwstats, NETDEV_OFFLOAD_XSTATS_TYPE_L3); + mutex_destroy(&hwstats->hwsdev_list_lock); +} -- cgit v1.3.1