This is the mail archive of the
systemtap@sourceware.org
mailing list for the systemtap project.
RE: [ltt-dev] LTTng 0.10 for Linux 2.6.26
- From: "Takashi Nishiie" <t-nishiie at np dot css dot fujitsu dot com>
- To: "'Wenji Huang'" <wenji dot huang at oracle dot com>, "'Mathieu Desnoyers'" <compudj at krystal dot dyndns dot org>
- Cc: <ltt-dev at lists dot casi dot polymtl dot ca>, <systemtap at sources dot redhat dot com>
- Date: Thu, 17 Jul 2008 17:04:10 +0900
- Subject: RE: [ltt-dev] LTTng 0.10 for Linux 2.6.26
- References: <20080715024754.GA26471@Krystal> <000001c8e7aa$b9548600$2bfd9200$@css.fujitsu.com> <20080717024550.GA22869@Krystal> <487ED503.2060503@oracle.com>
Wenji Huang wrote:
>
>Mathieu Desnoyers wrote:
>> * Takashi Nishiie (t-nishiie@np.css.fujitsu.com) wrote:
>>> Hi.
>>>
>>>> http://ltt.polymtl.ca/lttng/patch-2.6.26-0.10.tar.bz2
>>> I found a suspicious point though I compiled by the x86 architecture.
>>>
>>> Distribution: Red Hat Enterprise Linux 5
>>> gcc: gcc-4.1.2-42.el5
>>>
>>> In Module.markers
>>> <snip>
>>> kernel_arch_trap_exit vmlinux
>>> kernel_irq_entry vmlinux irq_id %u kernel_mode %u
>>> kernel_irq_exit vmlinux
>>> kernel_irq_exit vmlinux handled #1u%u
>>> kernel_kthread_stop vmlinux pid %d
>>> <snip>
>>>
>>> Why are two kernel_irq_exit list?
>>>
>>
>> That's because you have an old Modules.markers in your tree. Try
>> deleting the old one and to recompile the kernel. This should probably
>> be fixed in modpost.c.
>>
>Remember it had been discussed in
>http://sources.redhat.com/ml/systemtap/2008-q1/msg00466.html
>
It is a mysterious problem.
Env.1
Architecture: x86
Distribution: Red Hat Enterprise Linux 5
gcc: gcc-4.1.2-42.el5
In this environment, other marker is normally listed though two
kernel_irq_exit is listed even if Modules.markers is deleted
before it compiles. The problem was solved by using Roland 's
patch.
Env.2
Architecture: x86_64
Distribution: Fedora 9
gcc: gcc-4.3.0-8
In this environment, Module.markers is normally made. But,when Roland's
patch was used oppositely, kernel marker that existed in source
codes other than the samples directory and the ltt directory was not
listed.
There might be some problems in my compilation environment.
Is Module.markers normally generated to you?
>Roland's patch plus removing/emptying the file before vmlinux-modpost
>in scripts/Makefile.modpost like:
>
> vmlinux.o: FORCE
>+ @rm -fr $(kernelmarkersfile)
> $(call cmd,kernel-mod)
>
>will solve the problem.
I think that Module.markers should delete it when "Make clean" is executed.
Makefile:
clean: archclean $(clean-dirs)
$(call cmd,rmdirs)
$(call cmd,rmfiles)
@find . $(RCS_FIND_IGNORE) \
\( -name '*.[oas]' -o -name '*.ko' -o -name '.*.cmd' \
-o -name '.*.d' -o -name '.*.tmp' -o -name '*.mod.c' \
- -o -name '*.symtypes' -o -name 'modules.order' \) \
+ -o -name '*.symtypes' -o -name 'modules.order' \
+ -o -name 'Module.markers' \) \
-type f -print | xargs rm -f
Thank you,
--
Takashi Nishiie