> It now seems to me that this is not a iwl4965 bug.
> The 'fix' no longer seems to work.
> 
> However on the same hardware Ubuntu (2.6.24) and Rawhide (2.6.26) do not
> exhibit the problem (regardless of the power_level setting) but Fedora 9
> (2.6.25) does have the problem.

This appears to me to be yet another example of a problem that has been
reported in the past in various places including this mailing list.
(See http://www.bughost.org/pipermail/power/2007-May/000337.html
http://www.bughost.org/pipermail/power/2007-October/000998.html
http://www.bughost.org/pipermail/power/2008-March/001327.html and
others.)

It doesn't seem likely that the strange behaviour is related to beacon
responses, because these responses should generate an interrupt that is
picked up by powertop.

However, it is entirely possible that the iwl4965 (and iwl3965) are the
sources of the strange behaviour, through some other mechanism, perhaps
abberant DMA activity (possibly even aberrant DMA requests) or some
other mechanism that causes a wakeup that is not counted by powertop.
Of course, it is also entirely possible that some other part of the
computer is responsible.

I've seen lots of attempts at explaining the problem, but nearly all of
them either can't explain how there are so many wakeups (sometimes over
40K/s) or why powertop doesn't pick up any reason for the wakeups.  This
is why I'm leaning to the DMA explanation, even though I know very
little about Linux internals.

It sure would be nice to have some knowledgeable person take a look at
the issue and start on a resolution.  I'm willing to try out various
things on my machine, but I haven't noticed the behaviour for a while
now.  My experience is that the problem comes and goes, so I wouldn't
exonerate Ubuntu and Rawhide just yet.


Peter F. Patel-Schneider

_______________________________________________
Power mailing list
[email protected]
http://www.bughost.org/mailman/listinfo/power

Reply via email to