[Bug runtime/3379] New: probe on "atomic_notifier_call_chain" crashes the system

[Bug runtime/3379] New: probe on "atomic_notifier_call_chain" crashes the system

[guij at cn dot ibm dot com]

The command line is:
stap -e 'probe kernel.function("atomic_notifier_call_chain"){}'

If on 2.6.18.1/i686, the screen gets dark and the system is rebooted
immediately;

If on 2.6.18/ppc64, the system is crashed immediately and rebooted.
Several lines of error messages are shown on the console, but not
same every time.

Is there any workaround? or should we add it into the blacklist? 
Thanks.

Some examples of error messages on 2.6.18/ppc64:
  BUG: scheduling while atomic: swapper/0x00029032/0
  Unrecoverable exception 901 at 0
  cpu 0x0: Vector: 7265677368657265  at [c0000000058e3780]
  pc: c0000000058e3b20
or:
  cpu 0x3: Vector: 0  at [c00000000fcbd1d0]
  pc: 0000000000000000
  lr: 0000000000000000
  sp: c00000000fcbd450
  msr: 0
  current = 0xc00000007ff81800
  paca    = 0xc000000000569a00
  pid   = 0, comm = swapper
  WARNING: exception is not recoverable, can't continue
or:
  BUG: scheduling while atomic: swapper/0x00029032/0
  Unrecoverable exception 901 at 0
  cpu 0x0: Vector: 7265677368657265  at [c0000000058e3780]
  pc: c0000000058e3b20

-- 
           Summary: probe on "atomic_notifier_call_chain" crashes the system
           Product: systemtap
           Version: unspecified
            Status: NEW
          Severity: normal
          Priority: P2
         Component: runtime
        AssignedTo: systemtap at sources dot redhat dot com
        ReportedBy: guij at cn dot ibm dot com


http://sourceware.org/bugzilla/show_bug.cgi?id=3379

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

[Bug runtime/3379] probe on "atomic_notifier_call_chain" crashes the system

[dsmith at redhat dot com]

------- Additional Comments From dsmith at redhat dot com  2006-10-18 19:11 -------
This is a known problem.  See:

<http://sources.redhat.com/ml/systemtap/2006-q3/msg00637.html>

Bibo Mao volunteered to add kernel patches for the 3 problem areas I found. 
Here's the specific one for atomic_notifier_call_chain(), which marks it with
"__kprobes":

<http://www.ussg.iu.edu/hypermail/linux/kernel/0609.3/0589.html>

A problem we have is that atomic_notifier_call_chain() has been marked as
"__kprobes" in the upstream kernel, but hasn't made it to a released kernel from
my understanding (by either kernel.org or a vendor).

I have no problem with adding "atomic_notifier_call_chain" to the blacklist in
the meantime.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=3379

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

Re: [Bug runtime/3379] probe on "atomic_notifier_call_chain" crashes the system

[Keshavamurthy Anil S]

On Wed, Oct 18, 2006 at 07:11:29PM -0000, dsmith at redhat dot com wrote:
> 
> Bibo Mao volunteered to add kernel patches for the 3 problem areas I found. 
> Here's the specific one for atomic_notifier_call_chain(), which marks it with
> "__kprobes":
> 
> <http://www.ussg.iu.edu/hypermail/linux/kernel/0609.3/0589.html>

BTW, today I ran the 
stap -vv -e 'probe kernel.function("*") { log(".") }' on i386 box running
2.6.19-rc2-git2 and I had no issues. I even was able to do ls -l though the
system was on its knees. I had left this running for ~10 minutes before I killed
(and the stap module got unloaded cleanly).

> 
> A problem we have is that atomic_notifier_call_chain() has been marked as
> "__kprobes" in the upstream kernel, but hasn't made it to a released kernel from
> my understanding (by either kernel.org or a vendor).
Yup, this is in upstream kernel.

> 
> I have no problem with adding "atomic_notifier_call_chain" to the blacklist in
> the meantime.
We need this in any case as the systemtap could be run on older version of the kernel too.

-Anil Keshavamurthy

[Bug runtime/3379] probe on "atomic_notifier_call_chain" crashes the system

[anil dot s dot keshavamurthy at intel dot com]

------- Additional Comments From anil dot s dot keshavamurthy at intel dot com  2006-10-18 21:48 -------
Subject: Re:  probe on "atomic_notifier_call_chain" crashes the system

On Wed, Oct 18, 2006 at 07:11:29PM -0000, dsmith at redhat dot com wrote:
> 
> Bibo Mao volunteered to add kernel patches for the 3 problem areas I found. 
> Here's the specific one for atomic_notifier_call_chain(), which marks it with
> "__kprobes":
> 
> <http://www.ussg.iu.edu/hypermail/linux/kernel/0609.3/0589.html>

BTW, today I ran the 
stap -vv -e 'probe kernel.function("*") { log(".") }' on i386 box running
2.6.19-rc2-git2 and I had no issues. I even was able to do ls -l though the
system was on its knees. I had left this running for ~10 minutes before I killed
(and the stap module got unloaded cleanly).

> 
> A problem we have is that atomic_notifier_call_chain() has been marked as
> "__kprobes" in the upstream kernel, but hasn't made it to a released kernel from
> my understanding (by either kernel.org or a vendor).
Yup, this is in upstream kernel.

> 
> I have no problem with adding "atomic_notifier_call_chain" to the blacklist in
> the meantime.
We need this in any case as the systemtap could be run on older version of the kernel too.

-Anil Keshavamurthy


-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=3379

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

[Bug runtime/3379] probe on "atomic_notifier_call_chain" crashes the system

[dsmith at redhat dot com]

------- Additional Comments From dsmith at redhat dot com  2006-10-24 14:56 -------
I've checked in a patch that adds "atomic_notifier_call_chain" to the blacklist.

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


http://sourceware.org/bugzilla/show_bug.cgi?id=3379

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