1. 背景
linux 时间管理,包含clocksource,clockevent,timer,tick,timekeeper等等概念 ,这些概念有机地组成了完整的时间代码体系。当然,是代码就会有bug,本文通过一个bug入手,在实战中加深对理论的认识。获取时间,但是crash了。
2. 故障现象
OPPO云内核团队接到连通性告警报障,发现机器复位:
PID: 0 TASK: ffff8d2b3775b0c0 CPU: 1 COMMAND: "swapper/1"
#0 [ffff8d597f6489f0] machine_kexec at ffffffffa5a63b34
#1 [ffff8d597f648a50] __crash_kexec at ffffffffa5b1e242
#2 [ffff8d597f648b20] panic at ffffffffa615d85b
#3 [ffff8d597f648ba0] nmi_panic at ffffffffa5a9859f
#4 [ffff8d597f648bb0] watchdog_overflow_callback at ffffffffa5b4a881
#5 [ffff8d597f648bc8] __perf_event_overflow at ffffffffa5ba26b7
#6 [ffff8d597f648c00] perf_event_overflow at ffffffffa5babd24
#7 [ffff8d597f648c10] intel_pmu_handle_irq at ffffffffa5a0a850
#8 [ffff8d597f648e38] perf_event_nmi_handler at ffffffffa616d031
#9 [ffff8d597f648e58] nmi_handle at ffffffffa616e91c
#10 [ffff8d597f648eb0] do_nmi at ffffffffa616ebf8
#11 [ffff8d597f648ef0] end_repeat_nmi at ffffffffa616dd89
[exception RIP: __getnstimeofday64+144]
RIP: ffffffffa5b03940 RSP: ffff8d597f643c78 RFLAGS: 00000212
RAX: 15b5c8320b8602cd RBX: ffff8d597f643cb0 RCX: 000000005f89ee29
RDX: 00000000ee4479fe RSI: 0000012b5478f3b2 RDI: 0009709c7629b240
RBP: ffff8d597f643c90 R8: 00000000007001de R9: ffff8d596d5c0000
R10: 000000000000007a R11: 000000000000000e R12: ffffffffa662ea80
R13: 000000003ccbcfb6 R14: ffff8d895de08000 R15: 0000000000000081
ORIG_RAX: ffffffffffffffff CS: 0010 SS: 0018
--- <NMI exception stack> ---
#12 [ffff8d597f643c78] __getnstimeofday64 at ffffffffa5b03940
#13 [ffff8d597f643c98] getnstimeofday64 at ffffffffa5b0398e
#14 [ffff8d597f643ca8] ktime_get_real at ffffffffa5b03a45
#15 [ffff8d597f643cd0] netif_receive_skb_internal at ffffffffa603b936
#16 [ffff8d597f643d00] napi_gro_receive at ffffffffa603c588
#17 [ffff8d597f643d28] mlx5e_handle_rx_cqe_mpwrq at ffffffffc052ef1d [mlx5_core]
#18 [ffff8d597f643db8] mlx5e_poll_rx_cq at ffffffffc052f4b8 [mlx5_core]
#19 [ffff8d597f643e08] mlx5e_napi_poll at ffffffffc05304c6 [mlx5_core]
#20 [ffff8d597f643e78] net_rx_action at ffffffffa603bf1f
#21 [ffff8d59