cvs commit: src/sys/dev/pci pci.c

[M. Warner Losh]

In message: <432D9F4B.9080902 at root.org>
            Nate Lawson <nate at root.org> writes:
: Joao Barros wrote:
: > On 9/11/05, Warner Losh <imp at freebsd.org> wrote:
: > 
: >>imp         2005-09-11 19:28:31 UTC
: >>
: >>  FreeBSD src repository
: >>
: >>  Modified files:        (Branch: RELENG_6)
: >>    sys/dev/pci          pci.c
: >>  Log:
: >>  Change the default of pci_do_powerstate to 0, per request from re at .
: >>  The number of raid controllers that violate the WHQL seems to be
: >>  growing in number and not isolated to old versions as previously
: >>  thought.  Though the numbers of these seen in the wild is still
: >>  relatively small, they hang the system when parts of their devices are
: >>  powered down.  The one area that these parts appear often are in the
: >>  higher end servers.  As such, be conservative about powering down
: >>  devices that have no driver attached by default.  Until a better
: >>  approach is proven in current, this is the prudent choice.
: >>
: >>  Laptop users wishing the benefits of powering down devices with no
: >>  drivers will now need to set hw.pci.do_powerstate=1 in their
: >>  /boot/loader.conf file.  Some users will have devices that will
: >>  prevent this setting (hence the need to make it default 0).
: >>
: >>  Approved by: re@ (scottl)
: >>
: >>  Revision   Changes    Path
: >>  1.292.2.2  +2 -2      src/sys/dev/pci/pci.c
: > 
: > I was reading the Release Notes for 6.0 on
: > http://www.freebsd.org/relnotes/6-STABLE/relnotes/i386/article.html
: > and noticed:
: > 
: > 2.2.2 Hardware Support
: > 
: > The acpi(4) driver now turns the ACPI and PCI devices off or to a
: > lower power state when suspending, and back on again when resuming.
: > This behavior can be disabled by setting the debug.acpi.do_powerstate
: > and hw.pci.do_powerstate sysctls to 0.
: > 
: > Given this is the same tunable you changed back to 0 by default, does
: > that "when resuming" has anything to do with this last commit? If so,
: > it could still be mentioned the other way around, allowing people
: > wanting to, to enable the tunable :)
: 
: You bring up a different point also, which is that ACPI should probably 
: be set to match this same default for release.

Actually, all that the do_powerstate stuff in pci land does is turn
off the devices when no driver is attached.  We always turn them back
on on resume.

Warner