This is the mail archive of the systemtap@sourceware.org mailing list for the systemtap project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Why did control channel closes itself?


Thank you Masami,

On Tue, Apr 22, 2008 at 12:47 AM, Masami Hiramatsu <mhiramat@redhat.com> wrote:
> Hi Peter,
>
>
>  Peter Teoh wrote:
>  > WARNING: Number of errors: 1, skipped probes: 0
>  > WARNING: There were 56939 transport failures.
>
>  hmm, you might have an error message. I guess that message
>  has been washed away by other outputs.
>
>  could you run that script with "-o outfile" option?
>
>  And also, it seems that a lot of transport failuers occurred.
>  I think you can reduce the transport failures by -s option.
>


stapio:stp_main_loop:282 nb=46
ERROR: probe overhead exceeded threshold
stapio:stp_main_loop:282 nb=53
WARNING: Number of errors: 1, skipped probes: 2
stapio:stp_main_loop:282 nb=51
WARNING: There were 45618 transport failures.
stapio:stp_main_loop:282 nb=4
stapio:stp_main_loop:319 got STP_EXIT
stapio:cleanup_and_exit:234 detach=0
stapio:close_relayfs:221 closing
stapio:close_relayfs:240 done
stapio:cleanup_and_exit:247 closing control channel
stapio:cleanup_and_exit:253 removing
stap_437787d15cda3e4c2d917c2b55f74c0e_108785

Notice the error "probe overhead exceeded threshold".   And
subsequent control-C does not stop the execution, it just hanged
without any response.

My version is the latest:

1482d30eb166b566e99fa21f9cd697abb711c30e                branch 'master' of git:/
/sources.redhat.com/git/systemtap

lsmod |grep stap
stap_437787d15cda3e4c2d917c2b55f74c0e_108785   528444  1

Most likely irrelevant  but the last  few lines in  the output file are:

60165
  60166    857 Xorg(3118):    -> __resched_task
  60167  0xc041a18b : __resched_task+0x1/0x63
  60168  0xc06377a6 : kretprobe_trampoline_holder+0x3/0x33
  60169  0xc06377a6 : kretprobe_trampoline_holder+0x3/0x33
  60170  0xc04370e2 : hrtimer_interrupt+0xe0/0x13f
  60171  0xc0413206 : smp_apic_timer_interrupt+0x6c/0x82
  60172  0xc0405394 : apic_timer_interrupt+0x28/0x30
  60173  0xc041c237 : __wake_up_sync+0x3a/0x44
  60174  0xc06377a6 : kretprobe_trampoline_holder+0x3/0x33
  60175  0xc05c204c : sock_wfree+0x22/0x37
  60176  0xc05c3ee4 : skb_release_all+0x69/0xbc
  60177  0xc05c3804 : __kfree_skb+0xb/0x66
  60178  0xc05c388d : kfree_skb+0x2e/0x30
  60179  0xc061f558 : unix_stream_recvmsg+0x357/0x486
  60180  0xc05be7e1 : sock_aio_read+0xed/0xfb
  60181  0xc0473c52 : do_sync_read+0xab/0xe9
  60182  0xc04743f8 : vfs_read+0x9b/0x131
  60183  0xc047486a : sys_read+0x3b/0x60
  60184  0xc04048c6 : ia32_sysenter_target+0x66/0x8c
  60185
  60186    880 Xorg(3118):    <- __resched_task
  60187
  60188    883 Xorg(3118):   <- task_tick_fair
  60189
  60190    886 Xorg(3118):  <- hrtick
  60191

All the above  names have been repeated many times before, I must emphasize.

Funny thing is this - in  the stap_xxxxx.c auto-generated in the /tmp/
directory, the following lines were observed:

 1529         : (STP_OVERLOAD_INTERVAL + 1);
   1530       c->cycles_sum += cycles_elapsed;
   1531       if (interval > STP_OVERLOAD_INTERVAL) {
   1532         if (c->cycles_sum > STP_OVERLOAD_THRESHOLD) {
   1533           _stp_error ("probe overhead exceeded threshold");
   1534           atomic_set (&session_state, STAP_SESSION_ERROR);
   1535           atomic_inc (&error_count);
   1536         }
   1537         c->cycles_base = cycles_atend;
   1538         c->cycles_sum = 0;
   1539       }

Here c->cycles_sum is incremented.   But nowhere can I find it being
initialized - uninitialized cycles_sum variable - could  this be a
bug?

>  Thank you,
>
>  --
>  Masami Hiramatsu
>
>  Software Engineer
>  Hitachi Computer Products (America) Inc.
>  Software Solutions Division
>
>  e-mail: mhiramat@redhat.com
>
>



-- 
Regards,
Peter Teoh


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]