FAIL in 2.6.24-rc3

FAIL in 2.6.24-rc3

[Wenji Huang]

Hi,
 
   There are new failures in test on 2.6.24-rc3 kernel(i386).

   semok/twentyfour.stp:
       It complains Segmentation fault.
   Possible reason: $fd couldn't be accessed in sys_read return probe.

   buildok/fortyfive.stp
       It complains Segmentation fault.
   Possible reason: target_set_pid() caused that.

It seems that accessing parameters in return probe results in the above 
failures.

Is that related to kernel config? Do I miss something?

Regards,
Wenji
  

Re: FAIL in 2.6.24-rc3

[Srinivasa Ds]

Wenji Huang wrote:

> It seems that accessing parameters in return probe results in the above 
> failures.

I could reproduce the problem on x86_64 and ppc64. On debugging with GDB, It looks like reason for failure in both the testcases are similar.
=====================================================
Loaded symbols for /home/systemtap/tmp/stap_testing_200711260838/install/lib/systemtap/libebl_x86_64.so
Core was generated by `./stap -vvvv ../src/testsuite/buildok/fortyfive.stp'.
Program terminated with signal 11, Segmentation fault.
#0  0x00000000004519c9 in require<expression*> (v=0x7fffa9be2860,
    dst=0x7fffa9be28d8, src=0x202c642522286674)
    at /home/systemtap/tmp/stap_testing_200711260838/src/staptree.h:845
845           src->visit(v);
(gdb)
(gdb)
(gdb) t
[Current thread is 1 (process 14563)]
(gdb) bt
#0  0x00000000004519c9 in require<expression*> (v=0x7fffa9be2860,
    dst=0x7fffa9be28d8, src=0x202c642522286674)
    at /home/systemtap/tmp/stap_testing_200711260838/src/staptree.h:845
#1  0x0000000000451aa2 in deep_copy_visitor::deep_copy (s=0x202c642522286674)
    at /home/systemtap/tmp/stap_testing_200711260838/src/staptree.cxx:2389
#2  0x00000000004837e0 in derived_probe (this=0x2776d40, p=0xc3d210, l=0xc3d080)
    at /home/systemtap/tmp/stap_testing_200711260838/src/elaborate.cxx:56
#3  0x0000000000532aa2 in be_derived_probe (this=0x2776d40, p=0xc3d210,
    l=0xc3d080, t=BEGIN, pr=0)
    at /home/systemtap/tmp/stap_testing_200711260838/src/tapsets.cxx:85
#4  0x00000000005331e6 in be_builder::build (this=0x82b360, base=0xc3d210,
    location=0xc3d080, parameters=@0x7fffa9be2b70,
    finished_results=@0x7fffa9be3530)
========================================

Problem is not there in kretprobes, but may be in processing return value.

> 
> Is that related to kernel config? Do I miss something?
> 
  No, Its a bug.

Thanks
 Srinivasa DS

Re: FAIL in 2.6.24-rc3

[Frank Ch. Eigler]

Srinivasa Ds <srinivasa@in.ibm.com> writes:

> I could reproduce the problem on x86_64 and ppc64. On debugging with
> GDB, It looks like reason for failure in both the testcases are
> similar. [...]

I believe I fixed this small thinko in CVS.

- FChE