Broken kretprobe stack traces

[Daniel Xu]

Hi Masami,

Jakub reported a bug with kretprobe stack traces -- wondering if you've gotten
any bug reports related to stack traces being broken for kretprobes.

I think (can't prove) this used to work:

    # bpftrace -e 'kretprobe:__tcp_retransmit_skb { @[kstack()] = count() }'
    Attaching 1 probe...
    ^C

    @[
        kretprobe_trampoline+0
    ]: 1

fentry/fexit probes seem to work:

    # bpftrace -e 'kretfunc:__tcp_retransmit_skb { @[kstack()] = count() }'
    Attaching 1 probe...
    ^C
    
    @[
        ftrace_trampoline+10799
        bpf_get_stackid_raw_tp+121
        ftrace_trampoline+10799
        __tun_chr_ioctl.isra.0.cold+33312
        __tcp_retransmit_skb+5
        tcp_send_loss_probe+254
        tcp_write_timer_handler+394
        tcp_write_timer+149
        call_timer_fn+41
        __run_timers+493
        run_timer_softirq+25
        __softirqentry_text_start+207
        asm_call_sysvec_on_stack+18
        do_softirq_own_stack+55
        irq_exit_rcu+158
        sysvec_apic_timer_interrupt+54
        asm_sysvec_apic_timer_interrupt+18
    ]: 1
    @[
        ftrace_trampoline+10799
        bpf_get_stackid_raw_tp+121
        ftrace_trampoline+10799
        __tun_chr_ioctl.isra.0.cold+33312
        __tcp_retransmit_skb+5
  <...>

which makes me suspect it's a kprobe specific issue.

Thanks,
Daniel