Bug 4553 - i686 RHEL4u5 systemcall tapset argument issues with 2.6.9-55.ELsmp kernel
Summary: i686 RHEL4u5 systemcall tapset argument issues with 2.6.9-55.ELsmp kernel
Status: RESOLVED WONTFIX
Alias: None
Product: systemtap
Classification: Unclassified
Component: tapsets (show other bugs)
Version: unspecified
: P2 normal
Target Milestone: ---
Assignee: Unassigned
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2007-05-25 21:53 UTC by William Cohen
Modified: 2010-01-14 17:03 UTC (History)
1 user (show)

See Also:
Host: i686
Target:
Build:
Last reconfirmed:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description William Cohen 2007-05-25 21:53:56 UTC
A number of the systemtap runtime tests fail on the 2.6.9-55.ELsmp kernel
because the the arguments for the systemcalls are not be generated correctly for
the kernel.

FAIL: 32-bit access
FAIL: 32-bit alarm
FAIL: 32-bit chmod
FAIL: 32-bit clock
FAIL: 32-bit dir
FAIL: 32-bit link
FAIL: 32-bit mmap
FAIL: 32-bit mount
FAIL: 32-bit net1
FAIL: 32-bit openclose
FAIL: 32-bit readwrite
FAIL: 32-bit rt_signal
FAIL: 32-bit signal
FAIL: 32-bit stat
FAIL: 32-bit statfs
FAIL: 32-bit swap
FAIL: 32-bit sync
FAIL: 32-bit timer
FAIL: 32-bit trunc
FAIL: 32-bit uid
FAIL: 32-bit uid16
FAIL: 32-bit unlink


Looking through the systemtap.log shows that many of the arguments are listed as
"<unknown>". For example output of 32-bit link includes the following output:

link: open (<unknown>,
O_RDONLY|O_APPEND|O_DIRECTORY|O_EXCL|O_LARGEFILE|O_NONBLOCK) = 4
link: close (4) = 0
link: link (<unknown>, <unknown>) = 0
link: link (<unknown>, <unknown>) = -17 (EEXIST)
link: link (<unknown>, <unknown>) = -2 (ENOENT)
link: symlink (<unknown>, <unknown>) = 0
link: readlink (<unknown>, 0x3a080300, 128) = 6
link: exit_group (0) =
Comment 1 William Cohen 2007-06-07 16:29:11 UTC
Did some searching through the email archive of test results. It looks like
something changed between the following dates to make it so that the tests fails.

worked
Date: 200612200930
User: wcohen
Kernel: Linux 2.6.9-42.0.3.EL #1 Mon Sep 25 17:14:19 EDT 2006 i686 athlon i386
GNU/Linux



failed
Date: 200701120930
User: wcohen
Kernel: Linux 2.6.9-42.0.3.EL #1 Mon Sep 25 17:14:19 EDT 2006 i686 athlon i386
GNU/Linux

Comment 2 David Smith 2007-06-07 21:01:45 UTC
It looks like RHEL4u5 (or any other system with gcc version < 4) needs the new
"--enable-prologues" configure option.
Comment 3 Frank Ch. Eigler 2007-06-07 21:54:34 UTC
(In reply to comment #2)
> It looks like RHEL4u5 (or any other system with gcc version < 4) needs the new
> "--enable-prologues" configure option.

Yes, that's what the src/NEWS file says too.  It might be
automated (parse "gcc -V" during configure?), but I'm not
sure that this would be appropriate.  The spec file for
RHEL4 would of course have to do it one way or another.
Comment 4 David Smith 2007-06-08 18:14:41 UTC
I've checked in a change to configure.ac/configure that automatically enables
prologue searching if:
- we're running on an x86
- gcc version is less than 4

This change reduces the cases of system calls returning "<unknown>".  However,
the syscall tests still don't pass.  It appears that on the two rhel4 systems I
tried this on probes get skipped which changes the output the tests look for.

I'm unsure as to why this is happening.
Comment 5 David Smith 2010-01-14 17:03:08 UTC
On current rhel5 kernels (2.6.18-164.9.1.el5) we don't see this problem, so I'm
closing this as WONTFIX.