我正在尝试使用lttng来分析远程系统上的调度.
但是,使用babeltrace将跟踪显示为文本格式表示所有记录的事件都具有相同的时间戳.因此,TraceCompass中的可视化失败.

thomas@horrorshow: babeltrace ~/temp/lttng/my-trace/MySystem/my-session-20170608-133943/ | head
[13:41:59.444999936] (+?.?????????) MySystem sched_stat_runtime: { cpu_id = 0 }, { comm = "lttng-sessiond", tid = 1533, runtime = 5016120, vruntime = 451268995605 }
[13:41:59.444999936] (+0.000000000) MySystem sched_waking: { cpu_id = 0 }, { comm = "lttng-consumerd", tid = 1547, prio = 20, target_cpu = 0 }
[13:41:59.444999936] (+0.000000000) MySystem sched_stat_runtime: { cpu_id = 0 }, { comm = "lttng-sessiond", tid = 1533, runtime = 276632, vruntime = 451269272237 }
[13:41:59.444999936] (+0.000000000) MySystem sched_wakeup: { cpu_id = 0 }, { comm = "lttng-consumerd", tid = 1547, prio = 20, target_cpu = 0 }
[13:41:59.444999936] (+0.000000000) MySystem sched_switch: { cpu_id = 0 }, { prev_comm = "lttng-sessiond", prev_tid = 1533, prev_prio = 20, prev_state = 4096, next_comm = "lttng-consumerd", next_tid = 1547, next_prio = 20 }
[13:41:59.444999936] (+0.000000000) MySystem syscall_exit_epoll_wait: { cpu_id = 0 }, { ret = 1, fds_length = 1, overflow = 0, fds = [ [0] = { data_union = { u64 = 0x1C, fd = 28 }, raw_events = 0x1, events = { EPOLLIN = 1, EPOLLPRI = 0, EPOLLOUT = 0, EPOLLERR = 0, padding = 0 } } ] }
[13:41:59.444999936] (+0.000000000) MySystem syscall_entry_ioctl: { cpu_id = 0 }, { fd = 28, cmd = 62981, arg = 1547 }
[13:41:59.444999936] (+0.000000000) MySystem syscall_exit_ioctl: { cpu_id = 0 }, { ret = 0, arg = 1547 }
[13:41:59.444999936] (+0.000000000) MySystem syscall_entry_ioctl: { cpu_id = 0 }, { fd = 28, cmd = 2147808776, arg = 3047156396 }
[13:41:59.444999936] (+0.000000000) MySystem syscall_exit_ioctl: { cpu_id = 0 }, { ret = 0, arg = 3047156396 }

thomas@horrorshow: babeltrace ~/temp/lttng/my-trace/MySystem/my-session-20170608-133943/ | tail
[13:41:59.444999936] (+0.000000000) MySystem sched_wakeup: { cpu_id = 0 }, { comm = "rcu_preempt", tid = 7, prio = 20, target_cpu = 0 }
[13:41:59.444999936] (+0.000000000) MySystem sched_switch: { cpu_id = 0 }, { prev_comm = "lttng", prev_tid = 1698, prev_prio = 20, prev_state = 4096, next_comm = "rcu_preempt", next_tid = 7, next_prio = 20 }
[13:41:59.444999936] (+0.000000000) MySystem sched_stat_runtime: { cpu_id = 0 }, { comm = "rcu_preempt", tid = 7, runtime = 128301, vruntime = 456917982563 }
[13:41:59.444999936] (+0.000000000) MySystem sched_switch: { cpu_id = 0 }, { prev_comm = "rcu_preempt", prev_tid = 7, prev_prio = 20, prev_state = 1, next_comm = "lttng", next_tid = 1698, next_prio = 20 }
[13:41:59.444999936] (+0.000000000) MySystem sched_waking: { cpu_id = 0 }, { comm = "lttng-sessiond", tid = 1533, prio = 20, target_cpu = 0 }
[13:41:59.444999936] (+0.000000000) MySystem sched_stat_runtime: { cpu_id = 0 }, { comm = "lttng", tid = 1698, runtime = 209068, vruntime = 456921063330 }
[13:41:59.444999936] (+0.000000000) MySystem sched_wakeup: { cpu_id = 0 }, { comm = "lttng-sessiond", tid = 1533, prio = 20, target_cpu = 0 }
[13:41:59.444999936] (+0.000000000) MySystem sched_switch: { cpu_id = 0 }, { prev_comm = "lttng", prev_tid = 1698, prev_prio = 20, prev_state = 4096, next_comm = "lttng-sessiond", next_tid = 1533, next_prio = 20 }
[13:41:59.444999936] (+0.000000000) MySystem syscall_exit_recvmsg: { cpu_id = 0 }, { ret = 13156, msg = 3046107868 }
[13:41:59.444999936] (+0.000000000) MySystem syscall_entry_ioctl: { cpu_id = 0 }, { fd = 31, cmd = 63063, arg = 17 }

跟踪运行几秒钟,大小为912K.
我使用以下命令在嵌入式Linux设备上启动跟踪,使用lttng-relayd将数据传输到我的主机系统进行分析：

lttng create my-session --set-url=net://10.0.28.1
lttng enable-channel --kernel --subbuf-size=2621440 channel0
lttng enable-event --kernel --syscall --all --channel channel0
lttng enable-event --kernel --channel channel0 sched* 
lttng start
# Wait a bit
lttng stop

我该如何修复时间戳？

解决方法:

The folks at the lttng-dev mailing list solved the problem for me：我意外地将内核模块加载到假的测试时钟,而不是真正的时钟的内核模块.

我的错误是我在lttng手动提供的所有.ko文件上运行了insmod.但是,正确的方法是让lttng-sessiond自动加载模块,这将忽略假的测试时钟模块.自动模块加载失败对我来说因为我在安装内核模块后没有运行depmod -a.

在depmod -a之后,一切正常,时间戳是正确的.