Re: [PATCH] ACPI / LPSS: Ignore 10ms delay for Braswell and Baytrail

["Kasagar, Srinidhi" <srinidhi.kasagar@intel.com>]

On Fri, Aug 21, 2015 at 04:20:15PM +0300, Mika Westerberg wrote:
> On Fri, Aug 21, 2015 at 09:36:41AM +0300, Mika Westerberg wrote:
> > On Fri, Aug 21, 2015 at 05:41:52PM +0530, Kasagar, Srinidhi wrote:
> > > On Thu, Aug 20, 2015 at 03:38:05PM +0300, Mika Westerberg wrote:
> > > > +Heikki
> > > > 
> > > > On Thu, Aug 20, 2015 at 10:46:07PM +0530, Srinidhi Kasagar wrote:
> > > > > LPSS devices in Braswell and Baytrail does not need the default
> > > > > 10ms d3_delay imposed by PCI specification. Removing this
> > > > > unnecessary delay significantly reduces the resume time
> > > > > (~200ms on Braswell/Cherrytrail) on these platforms.
> > > > > 
> > > > > Signed-off-by: Srinidhi Kasagar <srinidhi.kasagar@intel.com>
> > > > > Signed-off-by: Kumar P Mahesh <mahesh.kumar.p@intel.com>
> > > > 
> > > > Have you tested this on Asus T100? The delay was actually needed in
> > > > order to restore the context IIRC.
> > > 
> > > Sorry, I do not have T100 h/w :(
> > 
> > OK. We have one and I'm going to test this patch on it. In particular
> > the T100 needed to have these delays otherwise writes failed.
> 
> The patch does not apply on top of v4.2-rc7 or linux-pm/bleeding-edge.
> So I just unconditionally set the delay to 0 for all LPSS devices on
> T100.
> 
> Having msleep(0) seems to be enough and the touch panel including I2C
> host controller runtime resumes just fine :-)
> 
> If I remove msleep() completely then things break apart and runtime
> resuming the touch panel triggers this:
> 
> [   46.143111] i2c_hid i2c-ATML1000:00: i2c_hid_set_power
> [   46.145758] i2c_hid i2c-ATML1000:00: __i2c_hid_command: cmd=fb 00 00 08
> [   46.148483] i2c_designware 80860F41:05: Unknown Synopsys component type: 0x00000000
> [   46.252125] i2c_designware 80860F41:05: timeout in enabling adapter
> [   47.254426] i2c_designware 80860F41:05: controller timed out
> 
> So clearly it needs some delay but it does not have to be 10ms.

Hm..Some internal south devices in Atom platforms does need ~3ms d3_delay
which we have fixed them as PCI fixups. So, do you think it make sense
to selectively pick those kind of devices as fixups and remove the
conditional delay entirely from this part?

Apart from touch/i2c what else breaks in T100?

Srinidhi