[Bug kprobes/2408] New: [RHEL4 U4] Backport Kprobes fixes from mainline

[Bug kprobes/2408] New: [RHEL4 U4] Backport Kprobes fixes from mainline

[anil dot s dot keshavamurthy at intel dot com]

RHEL4 U3 had the kprobes scalability patches in it and since
then we have fixed several issues in Kprobes.

The intent of this bugzilla is that the back port of these
mainline patches gets through testing and review before it
gets into RH distributions.

-- 
           Summary: [RHEL4 U4] Backport Kprobes fixes from mainline
           Product: systemtap
           Version: unspecified
            Status: NEW
          Severity: normal
          Priority: P2
         Component: kprobes
        AssignedTo: systemtap at sources dot redhat dot com
        ReportedBy: anil dot s dot keshavamurthy at intel dot com


http://sourceware.org/bugzilla/show_bug.cgi?id=2408

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

[Bug kprobes/2408] [RHEL4 U4] Backport Kprobes fixes from mainline

[fche at redhat dot com]

------- Additional Comments From fche at redhat dot com  2006-03-01 20:32 -------
I recommend opening a bug on bugzilla.redhat.com against RHEL instead of this
bug here.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2408

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

[Bug kprobes/2408] [RHEL4 U4] Backport Kprobes fixes from mainline

[anil dot s dot keshavamurthy at intel dot com]

------- Additional Comments From anil dot s dot keshavamurthy at intel dot com  2006-03-01 20:34 -------
Created an attachment (id=897)
 --> (http://sourceware.org/bugzilla/attachment.cgi?id=897&action=view)
Back port of kprobes patches from mainline

I am attaching the back ported patches 
along with this bugzilla. The series
files tells the order in which the
patches should be applied. The README file
in here tells the recent back ported patches.

The patches applies cleanly 
onto RHEL4 U3 2.6.9-29.EL version.

I want to see some basic testing done on this 
set of patches before I ask Red Hat for inclusion
in their upcoming updates namely RHEL4 U4.

Please test and report back the test result here.

thanks,
Anil


-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2408

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

[Bug kprobes/2408] [RHEL4 U4] Backport Kprobes fixes from mainline

[anil dot s dot keshavamurthy at intel dot com]

------- Additional Comments From anil dot s dot keshavamurthy at intel dot com  2006-03-01 20:36 -------
(In reply to comment #1)
> I recommend opening a bug on bugzilla.redhat.com against RHEL instead of this
> bug here.

Yes, we have a bugzilla there(180260), but I am not sure if every one can 
access that and hence I opened this.

Once I collect the testing feedback here, I will copy the same there.

thanks,
Anil





-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2408

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

[Bug kprobes/2408] [RHEL4 U4] Backport Kprobes fixes from mainline

[ananth at in dot ibm dot com]

------- Additional Comments From ananth at in dot ibm dot com  2006-03-02 09:00 -------
I have done some basic tests with the patches on a 2.6.9-30.EL on a Power5 LPAR.
All the basic modules (kprobe, jprobe, retprobe) worked fine. The changes I had
to make to the above patchset were:

- Fixup arch/powerpc/kernel/process.c as the patch generated wouldn't apply on
2.6.9-30.EL
- Add the check:

if (args->regs && user_mode(args->regs))
     return ret;

in arch/powerpc/kernel/kprobes.c

Ananth

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2408

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

[Bug kprobes/2408] [RHEL4 U4] Backport Kprobes fixes from mainline

[bibo dot mao at intel dot com]

------- Additional Comments From bibo dot mao at intel dot com  2006-03-02 09:43 -------
I test the patch in IA32 and IA64 platfrom about RHEL4-U3-RC3 2.6.9-34 kernel
version, and in IA32 there is one patch need modification in nx.patch
  +void __kprobes arch_remove_kprobe(struct kprobe *p) 
should be +void arch_remove_kprobe(struct kprobe *p), 
__kprobes is not defined in RHEL4 series.

And in IA64 platform, http://sourceware.org/bugzilla/show_bug.cgi?id=2071 still
unsolved.

And I will test in x86_64 platform.

And as anath pointed out, in iA64 platform it also should be checked.
if (args->regs && user_mode(args->regs))
     return ret;
Now I am sure for IA32, x86_64, for safety I suggest it should be check in all
platform.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2408

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

[Bug kprobes/2408] [RHEL4 U4] Backport Kprobes fixes from mainline

[bibo dot mao at intel dot com]

------- Additional Comments From bibo dot mao at intel dot com  2006-03-02 09:44 -------
> Now I am sure for IA32, x86_64, for safety I suggest it should be check in all
> platform.
sorry, I am not sure for IA32, x86_64 platfrom:(



-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2408

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

[Bug kprobes/2408] [RHEL4 U4] Backport Kprobes fixes from mainline

[joshua dot i dot stone at intel dot com]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |joshua dot i dot stone at
                   |                            |intel dot com


http://sourceware.org/bugzilla/show_bug.cgi?id=2408

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

[Bug kprobes/2408] [RHEL4 U4] Backport Kprobes fixes from mainline

[joshua dot i dot stone at intel dot com]

------- Additional Comments From joshua dot i dot stone at intel dot com  2006-03-03 02:33 -------
I just tested these patches against 2.6.9-34.ELsmp on i686.  They applied
cleanly, but did not compile.  As Bibo pointed out, you need to strip the
'__kprobes' qualifier from arch_remove_kprobe in arch/i386/kprobes.c.  I would
also suggest adding a return value (return 0) at the end of arch_prepare_kprobe
in the same file.

All of the pass-5 tests completed successfully.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2408

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

[Bug kprobes/2408] [RHEL4 U4] Backport Kprobes fixes from mainline

[bibo dot mao at intel dot com]

------- Additional Comments From bibo dot mao at intel dot com  2006-03-03 03:12 -------
I test in x86_64 platform, the systemtap samples and some jprobe(kretprobe) 
test cases can run ok.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2408

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

[Bug kprobes/2408] [RHEL4 U4] Backport Kprobes fixes from mainline

[anil dot s dot keshavamurthy at intel dot com]

------- Additional Comments From anil dot s dot keshavamurthy at intel dot com  2006-03-03 19:42 -------
Created an attachment (id=900)
 --> (http://sourceware.org/bugzilla/attachment.cgi?id=900&action=view)
Single patch to address all the Kprobes backport issues

Changes since previous attachment are:

1) Added the below check in kprobes exception handler code for all
arch.(suggested by Ananth/Yanmin)
 if (args->regs && user_mode(args->regs))
     return ret;
2) Fixed __kprobes macro which is not required for RHEL4(Bibo)
3) Added return 0; from arch_prepare_kprobe (suggested by Josh)
4)Back ported one more Fix of Ananth i,e Fix deadlock in function return probes

5) Removed trailing white space

Pending...Waiting for Prasanna's kprobe fault handler code to be merged.

Change summary:
------------------
 arch/i386/kernel/kprobes.c    |   50 ++++---
 arch/i386/kernel/process.c    |    8 -
 arch/ia64/kernel/jprobes.S    |   27 ++++
 arch/ia64/kernel/kprobes.c    |   75 ++++++++++-
 arch/ia64/kernel/process.c    |    7 -
 arch/ppc64/kernel/kprobes.c   |   28 +++-
 arch/ppc64/kernel/process.c   |    7 -
 arch/sparc64/kernel/kprobes.c |    4
 arch/x86_64/kernel/kprobes.c  |  133 ++------------------
 arch/x86_64/kernel/process.c  |    8 -
 include/asm-i386/kprobes.h    |   14 --
 include/asm-ia64/kprobes.h    |   23 +--
 include/asm-ppc64/kprobes.h   |   12 -
 include/asm-sparc64/kprobes.h |    9 -
 include/asm-x86_64/kprobes.h  |    6
 include/linux/kprobes.h       |    8 -
 include/linux/list.h	       |   21 +++
 kernel/kprobes.c	       |  274
++++++++++++++++++++++++++++++++----------
 kernel/module.c	       |    3
 kernel/sched.c 	       |    9 +
 
20 files changed, 431 insertions(+), 295 deletions(-)

TBD before posting the patch to RedHat:
------------------------------------
1) Delete changes related to sparc64.
2) Include Prasanna's final Kprobes fault handler code.
3) Refresh the patch against latest RH kernel version 2.6.9-34?

Thanks,
Anil


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
 Attachment #897 is|0                           |1
           obsolete|                            |


http://sourceware.org/bugzilla/show_bug.cgi?id=2408

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

[Bug kprobes/2408] [RHEL4 U4] Backport Kprobes fixes from mainline

[joshua dot i dot stone at intel dot com]

------- Additional Comments From joshua dot i dot stone at intel dot com  2006-03-03 21:59 -------
I tried the new patch on i686 2.6.9-34.ELsmp.  All tests in tests/kernel and
tests/testsuite passed.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2408

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

[Bug kprobes/2408] [RHEL4 U4] Backport Kprobes fixes from mainline

[anil dot s dot keshavamurthy at intel dot com]

------- Additional Comments From anil dot s dot keshavamurthy at intel dot com  2006-03-10 22:07 -------
Created an attachment (id=913)
 --> (http://sourceware.org/bugzilla/attachment.cgi?id=913&action=view)
Now includes Prasanna's recent kprobes fault handler code.

Applies on top of RHEL4 U3 (2.6.9-34.EL kernel). Please test 
this final patch. I will likely be posting this internally at RedHat
middle of next week.

Thanks,
Anil

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
 Attachment #900 is|0                           |1
           obsolete|                            |


http://sourceware.org/bugzilla/show_bug.cgi?id=2408

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

[Bug kprobes/2408] [RHEL4 U4] Backport Kprobes fixes from mainline

[joshua dot i dot stone at intel dot com]

------- Additional Comments From joshua dot i dot stone at intel dot com  2006-03-11 01:58 -------
(In reply to comment #11)
> Now includes Prasanna's recent kprobes fault handler code.

Do we have a test for this?

I tested the new patch on i686 SMP - it completes all Kprobes tests and pass-5
SystemTap tests without trouble.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2408

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

[Bug kprobes/2408] [RHEL4 U4] Backport Kprobes fixes from mainline

[anil dot s dot keshavamurthy at intel dot com]

------- Additional Comments From anil dot s dot keshavamurthy at intel dot com  2006-03-11 02:08 -------
(In reply to comment #12)
> (In reply to comment #11)
> > Now includes Prasanna's recent kprobes fault handler code.
> Do we have a test for this?
> I tested the new patch on i686 SMP - it completes all Kprobes tests and pass-
5
> SystemTap tests without trouble.

Josh, thanks for testing. For kprobe fault handler testing you can run the 
below script. Basically this tries to access userspace 
address from kprobe prehandler.

----------------------------- 
probe syscall.*, syscall.*.return {
        if (pid() == target()) {
                if (returnp)
                        printf("%d\n", returnval())
                else
                        printf("%s (%s) = ", name, argstr)
        }
}
-------------------------------

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2408

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

[Bug kprobes/2408] [RHEL4 U4] Backport Kprobes fixes from mainline

[joshua dot i dot stone at intel dot com]

------- Additional Comments From joshua dot i dot stone at intel dot com  2006-03-13 19:19 -------
(In reply to comment #13)
> Josh, thanks for testing. For kprobe fault handler testing you can run the 
> below script. Basically this tries to access userspace 
> address from kprobe prehandler.
> 
> ----------------------------- 
> probe syscall.*, syscall.*.return {
>         if (pid() == target()) {
>                 if (returnp)
>                         printf("%d\n", returnval())
>                 else
>                         printf("%s (%s) = ", name, argstr)
>         }
> }
> -------------------------------

After commenting out some unresolved functions from i686/syscalls.stp, this test
script ran fine.  I don't know if there's a particular target application that's
good for testing this - I used 'du' on my home directory.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2408

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

[Bug kprobes/2408] [RHEL4 U4] Backport Kprobes fixes from mainline

[hien at us dot ibm dot com]

------- Additional Comments From hien at us dot ibm dot com  2006-03-13 21:28 -------
(In reply to comment #14)
> (In reply to comment #13)
> > Josh, thanks for testing. For kprobe fault handler testing you can run the 
> > below script. Basically this tries to access userspace 
> > address from kprobe prehandler.
> > 
> > ----------------------------- 
> > probe syscall.*, syscall.*.return {
> >         if (pid() == target()) {
> >                 if (returnp)
> >                         printf("%d\n", returnval())
> >                 else
> >                         printf("%s (%s) = ", name, argstr)
> >         }
> > }
> > -------------------------------
> 
> After commenting out some unresolved functions from i686/syscalls.stp, this test
> script ran fine.  I don't know if there's a particular target application that's
> good for testing this - I used 'du' on my home directory.

You can try to run this script as 

stap -x <pid> sys.stp

where pid is the pid of a bash shell, then go to that shell and try out commands
like
chdir, rmdir, cat etc... 

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2408

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

[Bug kprobes/2408] [RHEL4 U4] Backport Kprobes fixes from mainline

[hien at us dot ibm dot com]

------- Additional Comments From hien at us dot ibm dot com  2006-03-13 21:31 -------
(In reply to comment #15)
> (In reply to comment #14)
> > (In reply to comment #13)
> > > Josh, thanks for testing. For kprobe fault handler testing you can run the 
> > > below script. Basically this tries to access userspace 
> > > address from kprobe prehandler.
> > > 
> > > ----------------------------- 
> > > probe syscall.*, syscall.*.return {
> > >         if (pid() == target()) {
> > >                 if (returnp)
> > >                         printf("%d\n", returnval())
> > >                 else
> > >                         printf("%s (%s) = ", name, argstr)
> > >         }
> > > }
> > > -------------------------------
> > 
> > After commenting out some unresolved functions from i686/syscalls.stp, this test
> > script ran fine.  I don't know if there's a particular target application that's
> > good for testing this - I used 'du' on my home directory.
> 
> You can try to run this script as 
> 
> stap -x <pid> sys.stp
> 
> where pid is the pid of a bash shell, then go to that shell and try out commands
> like
> chdir, rmdir, cat etc... 

Oops, I meant 
mkdir, rmdir, cat, cd etc...

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2408

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

[Bug kprobes/2408] [RHEL4 U4] Backport Kprobes fixes from mainline

[ananth at in dot ibm dot com]

------- Additional Comments From ananth at in dot ibm dot com  2006-10-12 11:43 -------
RHEL4-U4 has been out for quite sometime now. I see no reason why this bug can't
be closed.

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


http://sourceware.org/bugzilla/show_bug.cgi?id=2408

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