[Bug tapsets/6732] New: task_struct parent field removed from RHEL-5 and fedora 9 kernels

[Bug tapsets/6732] New: task_struct parent field removed from RHEL-5 and fedora 9 kernels

[mwielaard at redhat dot com]

From the thread starting at:
http://sourceware.org/ml/systemtap/2008-q2/msg00390.html

William Cohen:
The ppid() and pexecname() functions in tapset/context.stp and the task_parent()
function in tapset/task.stp use the parent field in task_struct. It appears that
the parent field in task_struct has been removed by linux-2.6-utrace.patch in
RHEL-5 and F-9 kernels. All that is left is the real_parent field.

Roland McGrath:
Using real_parent is correct in all kernels if you are concerned with
anything like the parent of the process.  The task_struct.parent field in
vanilla kernels in fact only really relates to ptrace.  

In RHEL5 and F[678] kernels, the utrace patch removed the ptrace-related
parent field and renamed real_parent to parent.  In future Fedora kernels,
there may or may not be a ptrace-related parent field, but the real useful
field will go back to being called real_parent (so as to minimize the
cosmetic divergence from upstream code).  (No statements about RHEL5>.2
kernels at this time.)

-- 
           Summary: task_struct parent field removed from RHEL-5 and fedora
                    9 kernels
           Product: systemtap
           Version: unspecified
            Status: NEW
          Severity: normal
          Priority: P2
         Component: tapsets
        AssignedTo: systemtap at sources dot redhat dot com
        ReportedBy: mwielaard at redhat dot com


http://sourceware.org/bugzilla/show_bug.cgi?id=6732

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug tapsets/6732] task_struct parent field removed from RHEL-5 and fedora 9 kernels

[mwielaard at redhat dot com]

------- Additional Comments From mwielaard at redhat dot com  2008-07-09 10:54 -------
Also runtime/task_finder.c currently contains this workaround:

#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,25)
#define real_parent parent
#endif

Which isn't entirely the correct check.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=6732

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug tapsets/6732] task_struct parent field removed from RHEL-5 and fedora 9 kernels

[dsmith at redhat dot com]

------- Additional Comments From dsmith at redhat dot com  2008-07-09 13:59 -------
(In reply to comment #1)
> Also runtime/task_finder.c currently contains this workaround:
> 
> #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,25)
> #define real_parent parent
> #endif
> 
> Which isn't entirely the correct check.

Since I'm the guilty party with that code, I'll confess.  I now understand that
it isn't the correct fix.

It looks like to me from a quick look through the tapsets that all uses of the
task_struct parent field are in embedded-C code.  I think this means that if we
add an "autoconf" test, such as was done for STAPCONF_HRTIMER_REL, that should
fix the problem.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=6732

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug tapsets/6732] task_struct parent field removed from RHEL-5 and fedora 9 kernels

[mwielaard at redhat dot com]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|systemtap at sources dot    |mwielaard at redhat dot com
                   |redhat dot com              |
             Status|NEW                         |ASSIGNED


http://sourceware.org/bugzilla/show_bug.cgi?id=6732

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.