Re: Updated ec-burst.diff patch

On Tue, 1 Jul 2003, Eirik Oeverby wrote:
> Can you be a bit more specific as to which other problems it fixes? I
> would love to try, because I have had some pretty nasty ACPI problems in
> the past, but I only have this one box to test on (my workstation) so I
> don't feel like going through the compile/install/fail/restore process
> if there isn't a good reason (for myself) to do so ;)
> (Just spent a weekend trying to find another kernel regression, I need
> to get some real work done soon)

There are a ton of problems with the current version, the largest that its
default behavior expects IBF/OBF events to be reported correctly by the
GPE but in fact implementations vary widely in their support.  This can be
worked around by setting "hw.acpi.event_driven=1" but I want to make it
work by default.  The event_driven sysctl is not a full solution because
it uses DELAY heavily.  To reduce the load in DELAY, I implemented burst
mode support.  However, since no open source EC code uses burst mode, it
is unknown how good its hw support is.  So my code has lots of tests of
your EC and I need the dmesg output from it to know how to proceed from
here.  That's why I need wide testing.

The patch is not a complete implementation but it should help identify any
hw problems in burst mode support.  I won't put it into the tree without
making sure it can fall back correctly.

Also, since I forgot the URL the second time:
http://root.org/~nate/freebsd/ec-burst.diff

-Nate