Menu
▾
▴
Re: OProfile
|
From: Venkat K. <ven...@en...> - 2008-01-17 15:40:14
|
Despite of setting the path of vmlinux through opcontrol --vmlinux=3D"Path" command, Opreport is not generating data related to this image file. It's been configured with "--kernel-support" option. It's generating data for Dynamically loaded modules & some binaries but not for Vmlinux. What's wrong. Processor - ARM/Xscale3 Oprofile Version: 0.9.3 Thanks, Venkat ----- Original Message ----- From: "Maynard Johnson" To: "Venkat Kumar" Subject: Re: OProfile Date: Wed, 09 Jan 2008 08:58:35 -0600 Venkat Kumar wrote: > Hi John, > It's best to reply to the entire oprofile-list for your best chance of getting an answer to questions. I cc'ed the list in my reply P.S. My name is "Maynard". > I had succesfully installed OProfile on my ARM/XScale3. > I am able to profile the Kernel and applications as well. > But if I want to see the output of only symbols of a particular > dynamic kernel module ( opreport -l > --image-path=3DPATH/kernelmodule.ko ). The --image-path option should just specify a directory, not the complete pathname (i.e., do not include the module name). But if you actually ran this command above with the complete pathname, you would have gotten an error telling you this. > It shows the following error. > error: no sample files found: profile specification too strict ? > Is it because the kernel module hasn't run for long time to > generate sample files That's a possibility. If you don't run opreport with the --image-path option, do you see a warning sent to stderr that says: "warning: / could not be found." If not, then you didn't have samples for that module in the profile. > or is there any other problem? > > At the same time, If I program the counters with some events, How > are these programmed events and the output generated by opreport > related to each other? The output lists the events at the top of the report. The column order of the events is in the same order as that list. -Maynard > The documents available online has not explained this part indetail I think. > > Please clarify. > > Thanks, > Venkat > > ----- Original Message ----- > From: "Maynard Johnson" > To: "Venkat Kumar" > Subject: Re: OProfile > Date: Mon, 12 Nov 2007 08:53:24 -0600 > > > Venkat Kumar wrote: > > Hi Folks, > > > > > > I had installed OProfile 0.9.3 version ( latest ) on my workstation. > > > > The Processor is an ARM/Xscale3 processor. > > > > I am running Linux version 2.6.18.3 > > > > I browsed through the code of OProfile Kernel driver and found > > that there is support for XScale3. > > > > But with the user space tools I found that there is no support > for XScale3. > > > > I browsed through that code as well and figured out that there > > are few structures which needs to be updated with XScale3 > > information other wise it throws an error > > saying "cpu_type unset is not valid, Force OProfile in timer > > mode" I updated the structures with XScale3 information and when > > opcontrol --start is run this error is not shown. > > (With out these updates when tried to run opcontrol --start, > > cpu_type error was coming) > > > > After the above changes I am able to profile the kernel and > > applications as well. > Hello, Venkat, > Difficult to discern exactly what's wrong without seeing your > patch. But these are the things you need to add to userspace > oprofile to support a new processor that the kernel already > supports: > > - Do 'opcontrol --init' and see what the kernel assigns to > /dev/oprofile/cpu_type. This typically consists of two segments: > / > - Create new directory under the existing 'events' directory: > events//, where '/' > match what the kernel reported as cpu_type. > - Add events file and unit_masks file to the new > directory > - Update events/Makefile.am to include new directory > - Update libop/op_cpu_type.c[h], libop/op_events.c, and > utils/ophelp.c as needed > > Good luck. > > Regards, > -Maynard > > > > But after profiling if I run the command "opreport -l" > > It shows the following two lines followed by the data in the > > format given below. The output shows the speed as 0 MHz. What > > does this mean? > > and it also shows that profiling through timer interrupt ( Deos > > this mean that the processor though has support performance > > counters doesn't have enough hardware support to generate the > > interrupt ). > > Please Clarify... > > > > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D > > CPU: CPU with timer interrupt, speed 0 MHz (estimated) > > Profiling through timer interrupt > > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D > > > > samples % app name symbol name > > ................ > > ................ > > ................ > > > > How will I make sure that the OProfile is profiling precisely? > > > > > > Regards, > > Venkat > > > > -- Want an e-mail address like mine? > > Get a *free e-mail *account today at www.mail.com > > ! > > > ------------------------------------------------------------------------ > > > > > ------------------------------------------------------------------------- > > This SF.net email is sponsored by: Splunk Inc. > > Still grepping through log files to find problems? Stop. > > Now Search log events and configuration files using AJAX and a > browser. > > Download your FREE copy of Splunk now >> http://get.splunk.com/ > > > ------------------------------------------------------------------------ > > > > _______________________________________________ > > oprofile-list mailing list > > opr...@li... > > https://lists.sourceforge.net/lists/listinfo/oprofile-list > > > > > -- Are we headed for a recession? * Read more* > on the Money Portal. > mail.com Money --=20 Want an e-mail address like mine? Get a free e-mail account today at www.mail.com! |