[Bug translator/12498] New: systemtap.examples/process/noptrace hangs on RHEL4 because arguments are wrong

[Bug translator/12498] New: systemtap.examples/process/noptrace hangs on RHEL4 because arguments are wrong

[wcohen at redhat dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=12498

           Summary: systemtap.examples/process/noptrace hangs on RHEL4
                    because arguments are wrong
           Product: systemtap
           Version: unspecified
            Status: NEW
          Severity: normal
          Priority: P2
         Component: translator
        AssignedTo: systemtap@sources.redhat.com
        ReportedBy: wcohen@redhat.com


I was looking at why the tests were not completing on a i386 rhel4 system with
the linux 2.6.9-100.ELsmp kernel. Looking through the systemtap.log found that
noptrace.stp was the last running test. The end of systemtap.log would have the
following:

meta taglines 'test_installcheck: stap -g noptrace.stp -c 'strace -d ls'' tag
'test_installcheck' value 'stap -g noptrace.stp -c 'strace ls''


The tests hang there. If the "make installcheck" on the RHEL4 was put in the
background with a ^z and then put back in the foreground with "fg", the tests
would continue with:


attempting command stap -g noptrace.stp -c 'strace ls'
OUT strace[6259] ptrace(13082612) blocked: rc=0
strace[6257] ptrace(13082612) blocked: rc=0
strace[6257] ptrace(13082612) blocked: rc=0
strace[6257] ptrace(13082612) blocked: rc=0
chng_cpu.meta
chng_cpu.stp
cycle_thief.meta
cycle_thief.stp
...
RC 0
PASS: systemtap.examples/process/noptrace run


Commenting out the following line in syscall.ptrace seems to allow things to
continue:

         $request=0xbeef # anything invalid should do

Get a lot of output like:

strace[11378] ptrace(13082612) blocked: rc=0
strace[11376] ptrace(13082612) blocked: rc=0
strace[11376] ptrace(13082612) blocked: rc=0
strace[11376] ptrace(13082612) blocked: rc=0

the 13082612 value looks suspect.

ran following on RHEL4:

stap -e 'probe syscall.ptrace{printf("%s\n", $$parms)}' -c 'strace ls'

The $$parms look very wrong. lots of things like the following in the output

request=0xc79ff4 pid=0xffffffffbff1cdb8 addr=0x1 data=0x0
request=0xc79ff4 pid=0xffffffffbff1cd08 addr=0x2c data=0xffffffffbff1ccf8
request=0xc79ff4 pid=0xffffffffbff1cd28 addr=0x18 data=0xffffffffbff1cd18
request=0xc79ff4 pid=0xffffffffbff1cd28 addr=0x0 data=0xffffffffbff1cd18
request=0xc79ff4 pid=0xffffffffbff1cd28 addr=0x4 data=0xffffffffbff1cd18
request=0xc79ff4 pid=0xffffffffbff1cdb8 addr=0x1 data=0x0

On i386 rhel5 system the output looks much more reasonable:

request=0x3 pid=0x32d4 addr=0x2c data=0xffffffffbf8c63a8
request=0x3 pid=0x32d4 addr=0x18 data=0xffffffffbf8c63d8
request=0x3 pid=0x32d4 addr=0x0 data=0xffffffffbf8c63d8
request=0x3 pid=0x32d4 addr=0x4 data=0xffffffffbf8c63d8
request=0x18 pid=0x32d4 addr=0x1 data=0x0
request=0x3 pid=0x32d4 addr=0x2c data=0xffffffffbf8c63d8

Maybe this is due to the debuginfo being bad on the RHEL4 kernel.

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

[Bug translator/12498] systemtap.examples/process/noptrace hangs on RHEL4 because arguments are wrong

[fche at redhat dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=12498

Frank Ch. Eigler <fche at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |fche at redhat dot com

--- Comment #1 from Frank Ch. Eigler <fche at redhat dot com> 2011-02-18 15:19:02 UTC ---
Perhaps the script could sanity-check incoming parameters and abort and/or 
suppress its "$request=0xbeef" assignment.

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

[Bug translator/12498] systemtap.examples/process/noptrace hangs on RHEL4 because arguments are wrong

[wcohen at redhat dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=12498

--- Comment #2 from William Cohen <wcohen at redhat dot com> 2011-02-18 20:59:17 UTC ---
On RHEL5 puts the probe at the first instruction of sys_ptrace function. On the
i686 RHEL4 system the probe in put farther into the function. RHEL4:

[wcohen@slingshot obj]$ stap -p2 -e 'probe syscall.ptrace{printf("%s\n",
$$parms)}' -c 'strace ls'
# functions
_dwarf_tvar_get_addr_6:long ()
_dwarf_tvar_get_data_7:long ()
_dwarf_tvar_get_pid_5:long ()
_dwarf_tvar_get_request_4:long ()
# probes
kernel.function("sys_ptrace@arch/i386/kernel/ptrace.c:235").call? /*
pc=_stext+0x985c */ /* <- syscall.ptrace = kernel.function("sys_ptrace").call?
<- syscall.ptrace */

using the _stext from System.Map-`uname -r`
pc=_stext+0x985c = c0100298  + 0x985c = C0109AF4

c0109af2 <sys_ptrace>:
c0109af2:    55                       push   %ebp
c0109af3:    57                       push   %edi
c0109af4:    bf 00 f0 ff ff           mov    $0xfffff000,%edi #<-probe here
c0109af9:    56                       push   %esi
c0109afa:    53                       push   %ebx
c0109afb:    53                       push   %ebx
c0109afc:    53                       push   %ebx
c0109afd:    8b 5c 24 24              mov    0x24(%esp),%ebx
c0109b01:    21 e7                    and    %esp,%edi


Due to the eu-readelf being really old on RHEL-4, I copied the vmlinux over to
RHEL-5 and did the following:

eu-readelf --debug-dump=info  vmlinux > /tmp/debug

section of interest:

[ 862dd]    subprogram
             sibling              [ 86b9f]
             external             
             name                 "sys_ptrace"
             decl_file            1
             decl_line            235
             prototyped           
             type                 [ 7d202]
             low_pc               0xc0109af2 <sys_ptrace>
             high_pc              0xc010a0ea <do_syscall_trace>
             frame_base           location list [ 13a48]
 [ 862fa]      formal_parameter
               name                 "request"
               decl_file            1
               decl_line            234
               type                 [ 7d231]
               location             2 byte block
                [   0] breg4 28
 [ 86308]      formal_parameter
               name                 "pid"
               decl_file            1
               decl_line            234
               type                 [ 7d231]
               location             2 byte block
                [   0] breg4 32
 [ 86316]      formal_parameter
               name                 "addr"
               decl_file            1
               decl_line            234
               type                 [ 7d231]
               location             location list [ 13b40]
 [ 86325]      formal_parameter
               name                 "data"
               decl_file            1
               decl_line            234
               type                 [ 7d231]
               location             location list [ 13bc7]

Are the "location" fields for request and pid parameters reasonable?

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

[Bug translator/12498] systemtap.examples/process/noptrace hangs on RHEL4 because arguments are wrong

[wcohen at redhat dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=12498

William Cohen <wcohen at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED

--- Comment #3 from William Cohen <wcohen at redhat dot com> 2011-02-25 16:17:53 UTC ---
Added work around to avoid running the test on RHEL4 system.

commit ca23376536effd5ed935cde660c06d51b7334657
Author: William Cohen <wcohen@redhat.com>
Date:   Thu Feb 24 14:02:21 2011 -0500

    Disable systemtap.examples/process/noptrace.stp on RHEL4 systems (PR12498)

    The debug information on RHEL4 kernels appears to be wrong. This causes
    the noptrace.stp example to write into the wrong location. Rather than the
    ptrace syscall exiting immediately, the ptrace syscall continues and the
    test hangs on RHEL4 systems.  The noptrace.stp test is disabled on RHEL4
    systems to work around this issue.

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.