On an SMP x86 with 2 SMT-enabled atom 330 cores  (4 logical CPUs)

Problem is reproducible every time in all kernel releases from
up to 2.6.39-rc5,  but no problem in 2.6.36,  and no problem if I omit the --callgraph=2 option

My opcontrol command lines :

(First I set ksrcdir to my kernel build directory like so)
ksrcdir= symlink target of /lib/modules/`uname -r`/source


opcontrol --init;
opcontrol --setup --vmlinux=${ksrcdir}/vmlinux --separate=cpu --callgraph=2;
opcontrol --start-daemon --vmlinux=${ksrcdir}/vmlinux;
opcontrol --start --separate=cpu;

OOPS occurs soon after.    The OOPS offset is at the line reading
         addr = *stack;
and register EBX (which contains value of stack) is garbaged  -   0x0000000a

I see some other mentions of similar problems elsewhere,  but they are
all for much older kernels,  and in my case the above commands work fine
on 23.6.36

John Lumby
output of ver_linux

Linux lumbyont.localdomain 2.6.39-rc5 #5 SMP Tue May 3 11:04:57 EDT 2011
i686 GNU/Linux

Gnu C                  4.4.4
Gnu make               3.81
binutils               2.19.1
util-linux             2.13-pre7
mount                  support
module-init-tools      3.3-pre1
e2fsprogs              1.41.13
pcmciautils            014
quota-tools            3.13.
PPP                    2.4.4
isdn4k-utils           3.9
Linux C Library        2.11.1
Dynamic linker (ldd)   2.11.1
Procps                 3.2.7
Net-tools              1.60
Kbd                    1.12
oprofile               0.9.7git
Sh-utils               8.9
udev                   095
wireless-tools         29
Modules Loaded         xt_NFLOG nfnetlink_log nfnetlink ppp_synctty
ppp_async crc_ccitt ppp_generic slhc via_rhine snd_hda_intel
snd_hda_codec rtc_cmos rtc_core rtc_lib bridge stp llc i915
drm_kms_helper intel_agp intel_gtt tun xt_state ipt_MASQUERADE
iptable_nat nf_nat nf_conntrack_ipv4 nf_conntrack nf_defrag_ipv4 dm_mod
parport_pc lp parport