svn commit: r216358 - head/sys/dev/le

[John Baldwin]

On Friday, December 10, 2010 11:36:17 am Tijl Coosemans wrote:
> On Friday 10 December 2010 16:31:36 John Baldwin wrote:
> > On Friday, December 10, 2010 10:05:49 am Tijl Coosemans wrote:
> >> Author: tijl
> >> Date: Fri Dec 10 15:05:49 2010
> >> New Revision: 216358
> >> URL: http://svn.freebsd.org/changeset/base/216358
> >> 
> >> Log:
> >>   Use SYS_RES_IOPORT instead of PCIM_CMD_PORTEN when calling pci_enable_io.
> >>   
> >>   Approved by:	kib (mentor)
> > 
> > To be honest, bus_alloc_resource() with RF_ACTIVE will set PORTEN or MEMEN for 
> > you.  Drivers should no longer be calling pci_enable_io() by hand.
> 
> Looking into this, in dev/sound/pci/atiixp.c and dev/sound/pci/ich.c it
> is used during resume. Can those calls be removed as well?

Yes, we save and restore the command register (along with other standard
header registers like BARs, etc.) across suspend and resume in the PCI bus
layer.

> And in dev/mpt/mpt.c pci_enable_io and pci_disable_io are called in a
> number of places. Those should probably all stay?

Yes, that appears to be a workaround for a quirk in the hardware.

-- 
John Baldwin