Re: Results of systemtap-20071215 snapshot on ia64

[William Cohen <wcohen at redhat dot com>]

Masami Hiramatsu wrote:
> Hi Will,
>
> William Cohen wrote:
> > Masami Hiramatsu wrote:
> > > > FAIL: backtrace of yyy_func4.return (0)
> > > > FAIL: print_stack of yyy_func4.return (0)
> > > Could you show me the error log in systemtap.log about this testcase?
> > > I saw a patch which fixes this bug was merged into 2.6.24-rc5.
> > > So I expected it would pass on 2.6.24-rc5.
> > Do you have a pointer to the patch that is suppose to fix this? Below is the 
> > output from the systemtap.log file related to those tests.
>
> Thank you,
> Here is the (kernel) patch which can fix this fails.
> http://marc.info/?l=linux-ia64&m=119493734125209&w=2

Masami,

Thanks for the pointer to the kernel patch. The patch is definitely in the
2.6.24-rc5 kernel used for the tests.

> > backtrace from 
> > module("systemtap_test_module2").function("yyy_func4@/home/wcohen/stap_snap_200712152157/src/testsuite/systemtap.context/stapte: 
> >
> > Returning from: 0xa000000200af0000 : yyy_func4+0x0/0x20 [systemtap_test_module2]
> > Returning to  : 0xa000000200af0050 : yyy_func3+0x30/0x60 [systemtap_test_module2]
> >   0xa000000200af0050 : yyy_func3+0x30/0x60 [systemtap_test_module2]
> >   0xa000000200af00b0 : yyy_func2+0x30/0x60 [systemtap_test_module2]
> >   0xa000000200af0110 : yyy_func1+0x30/0x60 [systemtap_test_module2]
> >   0xa000000200ac8220 : stm_write_cmd+0x100/0x0 [systemtap_test_module1]
> >   0xa0000001001ebaa0 : proc_reg_write+0x120/0x160 []
> >   0xa0000001001664d0 : vfs_write+0x1d0/0x320 []
> >   0xa000000100166f50 : sys_write+0x70/0xe0 []
> >   0xa00000010000b2d0 : ia64_trace_syscall+0xd0/0x110 []
> >   0xa000000000010720
> > --------
> > the return stack is 0xa000000200af0050 0xa000000200af00b0 0xa000000200af0110 
> > 0xa000000200ac8220 0xa0000001001ebaa0 0xa0000001001664d0 0xa0000001001
> > --------
> >   0xa000000200af0050 : yyy_func3+0x30/0x60 [systemtap_test_module2]
> >   0xa000000200af00b0 : yyy_func2+0x30/0x60 [systemtap_test_module2]
> >   0xa000000200af0110 : yyy_func1+0x30/0x60 [systemtap_test_module2]
> >   0xa000000200ac8220 : stm_write_cmd+0x100/0x0 [systemtap_test_module1]
> >   0xa0000001001ebaa0 : proc_reg_write+0x120/0x160 []
> >   0xa0000001001664d0 : vfs_write+0x1d0/0x320 []
> >   0x00000a0000001001
>
> Hmm, interesting. That output is a bit different from x86-64, but works fine.
> I think it is a kind of architecture dependent specification. In that case,
> we should fix that testcase.
>
> backtrace from module("systemtap_test_module2").function("yyy_func4@/home/mhiramat/Work/systemtap/testsuite/systemtap.context/staptest11202/sy:
> Returning from: 0xffffffff8821d000 : yyy_func4+0x0/0x4 [systemtap_test_module2]
> Returning to  : 0xffffffff8821d009 : yyy_func3+0x5/0x8 [systemtap_test_module2]
>  0xffffffff8821d011 : yyy_func2+0x5/0x8 [systemtap_test_module2]
>  0xffffffff8821d019 : yyy_func1+0x5/0x8 [systemtap_test_module2]
>  0xffffffff8822008a : stm_write_cmd+0x3d/0x0 [systemtap_test_module1]
>  0xffffffff810da370 : proc_reg_write+0x80/0x9b []
>  0xffffffff810a1811 : vfs_write+0xc6/0x16f []
>  0xffffffff810a1dce : sys_write+0x45/0x6e []
>  0xffffffff8100c09a : tracesys+0xdc/0xe1 []
> --------
> the return stack is 0xffffffff81024585 0xffffffff8821d011 0xffffffff8821d019 0xffffffff8822008a 0xffffffff810da370 0xffffffff810a1811 0xffffffff810
> --------
>  0xffffffff81024585 : kretprobe_trampoline_holder+0x1/0x4 []
>  0xffffffff8821d011 : yyy_func2+0x5/0x8 [systemtap_test_module2]
>  0xffffffff8821d019 : yyy_func1+0x5/0x8 [systemtap_test_module2]
>  0xffffffff8822008a : stm_write_cmd+0x3d/0x0 [systemtap_test_module1]
>  0xffffffff810da370 : proc_reg_write+0x80/0x9b []
>  0xffffffff810a1811 : vfs_write+0xc6/0x16f []
>  0x00000ffffffff810
>
> Thanks,

It looks like there needs to be some tweaking on the test. Both of the tests 
return (0) so nothing is matched up (variables m3 and m4 in 
testsuite/systemtap.context/backtrace.tcl)

FAIL: backtrace of yyy_func4.return (0)
FAIL: print_stack of yyy_func4.return (0)

Noticed that the print_stack expect kretprobe_trampoline_holder to be in the 
output and it is missing from the ia64 version, so "print_stack of 
yyy_func4.return" is not going to pass.

-Will