Re: Kernel hang with latest snapshot

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

David Boreham wrote:
> The saga continues...
>
> Tonight's installment:
>
> I spun up a totally new machine : P4 with HT, 2G ram.
> Installed FC5, updated to latest fedora mirror state.
> Upgraded elf utils using the rpms from redhat systemtap ftp server.
> Installed kernel devel, kernel debug.
> Pulled systemtap from cvs, built, installed.
> Tried to use LKET. Machine hangs just like the previous
> two machines I tried.

I asked Bibo to test LKET on his machines. He tried several machines 
and all of them works well. Did you uninstall SystemTap shipped with 
FC5 before install from cvs?

> Next, I wanted to test the theory that systemtap in general,
> and not LKET is broken for me. So I went back to my trusty
> vmware machine. Found some systemtap example stp scripts
> in the src/examples/small_demos directory. Basically every
> script I have tried in there works perfectly. No hangs, crashes,
> just works. I've tested so far: kmalloc.stp, close.stp, fileopen.stp.

Can you try to run these stap scripts with -bM option and see if they 
will crash?

> Lastly I tested the theory that perhaps it's only the iosyscall
> stuff in LKET that's causing me trouble. I instead tried network I/O
> like this:
>
> stap -b -M -e "probe addevent.netdev.receive {}" -v
>
> The kernel crashes just as nicely as before :(

Could you please try the following:

1) stap -bMv -e "probe addevent.netdev.receive {}" -D ASCII_TRACE
  if this works, then it should be relayfs that cause the trouble

2) try this script with/without -bM option:
   probe syscall.open
   {
     printf("%0s%4b%4b", filename, flags, mode)
   }

Thanks and looking forward to your reply

- Guanglei