Re: Kernel hang with latest snapshot

[Li Guanglei <guanglei at cn dot ibm dot com>]

David Boreham wrote:
> Hi, I'm interested in using LKET to analyse the behavior of a large
> application that has disk i/o-bound performance. My top level goal
> is to be able to record system-wide traces of all i/o activity at a
> system call level (open,close,write,fsync etc). I am using Fedora Core 5
> (although if it would help I'd be happy to use any distribution/custom 
> kernel).
>
> The systemtap that comes with the FC5 distribution is too old to
> include LKET. The LKET from the latest systemtap snapshot
> (20061007) does not work with the systemtap binary in the distribution.
>
> So I removed the distribution systemtap, installed the dependencies
> and built/installed the latest snapshot systemtap.
>
> So far so good.
>
> Unfortunately when I try to use LKET it would appear that my
> kernel hangs (machine totally unresponsive after the kernel module
> is loaded).
>
> This is the command line I'm using :
>
> /usr/local/bin/stap -e "probe addevent.iosyscall.flock.* {}" -bM -vv -k
>
> (btw addevent.iosyscall.* produces C code that will not compile)
>
> It gets this far and then hangs:
>
>  ld -m elf_i386 -m elf_i386 -r -o /tmp/stapA607FG/stap_3312.ko 
> /tmp/stapA607FG/stap_3312.o /tmp/stapA607FG/stap_3312.mod.o
> make: Leaving directory `/usr/src/kernels/2.6.17-1.2187_FC5-i686'
> Pass 4: compiled C into "stap_3312.ko" in 23930usr/2630sys/28585real ms.
> Pass 5: starting run.
> Running sudo /usr/local/bin/staprun -u root -d 3312 
> /tmp/stapA607FG/stap_3312.ko
>
> I'm using this kernel : 2.6.17-1.2187_FC5 (i386/i686) and vmware.
>
> Any advice on how to proceed is most welcome.
>
> Thanks for your time.

I will take a look at this tomorrow. Although I usually develop & test 
LKET in PPC64, I think I should also make sure LKET will run on other 
platform, especially i386(I don't have other platforms :( ).

I also noticed that some trace hooks in LKET have compile errors on 
i386, and most of them are due to the 32-bit/64-bit environment 
difference since machine I am using is a 64-bit environment. I will 
take a look at these errors using my thinkpad tomorrow.

BTW, I checked in a lot of LKET changes, especially lket-b2a today. I 
will have to take some time to update the document. The biggest 
enhancement is that lket-b2a could dump the trace data into MySQL 
database, which makes the trace data analyzing much easier.

- Guanglei