I confirm. Bug is still there in 2.6.15-25-686 #1 SMP PREEMPT kernel.
On my laptop (Acer Aspire 1692WLMi, Pentium M740 1.73 GHz),
With echo 1 > /sys/module/processor/parameters/max_cstate
top reports tipically:
Cpu(s): 4.3% us, 0.3% sy, 0.0% ni, 95.0% id, 0.0% wa, 0.0% hi, 0.3% si
With echo 2 > /sys/module/processor/parameters/max_cstate
Cpu(s): 54.5% us, 9.1% sy, 0.0% ni, 36.4% id, 0.0% wa, 0.0% hi, 0.0% si
This does not seem to be simply a problem with mis-reporting the CPU usage. The system actually feels slower when max_cstate is >= 2: launching a firefox window is noticeably slower; I can actually see the jumps as the window outline grows on-screen. In fact, that is what prompted me to look for bugs in the first place. I'll try to back this up with some benchmark later. Also, I'll try and check this on a Core Duo processor. Has anyone done that?