Add stack traces to trace.py

[palmtenor]

As mentioned in #621 , this would be very useful for debugging. This Diffs adds -U and -K options to trace.py so that it would print out user / kernel stack traces on each event.

Sample generated BPF program and outpout:

qinteng@devbig473:~ {}$ sudo LD_LIBRARY_PATH=/data/users/qinteng/toolchains/bcc-dev/lib64: /data/users/qinteng/toolchains/python-2.7.11/bin/python /data/users/qinteng/bcc/tools/trace.py finish_task_switch -K -U -M 3 -v

#include <linux/ptrace.h>
#include <linux/sched.h>        /* For TASK_COMM_LEN */


struct probe_finish_task_switch_1_data_t
{
        u64 timestamp_ns;
        u32 pid;
        char comm[TASK_COMM_LEN];

       int kernel_stack_id;
       int user_stack_id;
};

BPF_PERF_OUTPUT(probe_finish_task_switch_1_events);
BPF_STACK_TRACE(probe_finish_task_switch_1_stacks, 1024);


int probe_finish_task_switch_1(struct pt_regs *ctx)
{

        u32 __pid = bpf_get_current_pid_tgid();
        if (__pid == 911113) { return 0; }



        if (!(1)) return 0;

        struct probe_finish_task_switch_1_data_t __data = {0};
        __data.timestamp_ns = bpf_ktime_get_ns();
        __data.pid = bpf_get_current_pid_tgid();
        bpf_get_current_comm(&__data.comm, sizeof(__data.comm));


        __data.user_stack_id = probe_finish_task_switch_1_stacks.get_stackid(
          ctx, BPF_F_REUSE_STACKID | BPF_F_USER_STACK
        );
        __data.kernel_stack_id = probe_finish_task_switch_1_stacks.get_stackid(
          ctx, BPF_F_REUSE_STACKID
        );
        probe_finish_task_switch_1_events.perf_submit(ctx, &__data, sizeof(__data));
        return 0;
}

p::finish_task_switch FLT=1 ACT=[]/[]
TIME     PID    COMM         FUNC
21:19:20 7565   TierUpdNotif finish_task_switch
    User Stack Trace:
        00007fbf96ec0a83 [unknown]
        000000000247d071 event_base_loop
        0000000002190046 folly::EventBase::loopBody(int)
        0000000000f9df02 facebook::services::TierUpdatesObserver::notifierThread()
        00007fbf96b44ec1 execute_native_thread_routine
        00007fbf978af7f1 start_thread
        00007fbf96ec046d __clone
    Kernel Stack Trace:
        ffffffff8108e0f1 finish_task_switch
        ffffffff81774cf7 schedule
        ffffffff817777dc schedule_hrtimeout_range_clock
        ffffffff81777823 schedule_hrtimeout_range
        ffffffff811fd3fc ep_poll
        ffffffff811fe7f5 sys_epoll_wait
        ffffffff81778232 system_call_fastpath

21:19:20 0      swapper/11   finish_task_switch
    User Stack Trace:
        -14
    Kernel Stack Trace:
        ffffffff8108e0f1 finish_task_switch
        ffffffff81774cf7 schedule
        ffffffff81774f9e schedule_preempt_disabled
        ffffffff810a6ca8 cpu_startup_entry
        ffffffff81031760 start_secondary

21:19:20 862    fct0-smp     finish_task_switch
    User Stack Trace:
        -14
    Kernel Stack Trace:
        ffffffff8108e0f1 finish_task_switch
        ffffffff81774cf7 schedule
        ffffffff817773a7 schedule_timeout
        ffffffff810c0460 msleep_interruptible
        ffffffffa000328e kfio_msleep_noload
        ffffffffa009f635 ifio_8a30f.3f931e2037196fc46b5b711d5ae377442c8.3.2.4.1086

[goldshtn]

Looks good. Great work, this will be very useful!