Re: [PATCH 2/2] ARM: tegra: remove copy-and-pasted usb platform data from boards

["Heiko Stübner" <heiko@sntech.de>]

Am Freitag 15 Juli 2011, 18:45:53 schrieb Colin Cross:
> On Fri, Jul 15, 2011 at 2:55 AM, Heiko Stübner <heiko@sntech.de> wrote:
> > Am Freitag, 15. Juli 2011, 11:13:48 schrieb Marc Dietrich:
> >> see comment for PATCH 1/2
> >> 
> >> Am Sun, 10 Jul 2011 15:19:34 -0700
> >> 
> >> schrieb Colin Cross <ccross@android.com>:
> >> > trimslice and paz00 both have functionally identical platform
> >> > data for the tegra-ehci driver.  Move the platform data into
> >> > devices.c, and remove it from all the board files.
> >> > 
> >> > Signed-off-by: Colin Cross <ccross@android.com>
> >> 
> >> Acked-By: Marc Dietrich <marvin24@gmx.de>
> >> 
> >> > ---
> >> > 
> >> >  arch/arm/mach-tegra/board-paz00.c     |   26
> >> > 
> >> > 
> >> > 
> >> > -static struct tegra_ulpi_config ulpi_phy_config = {
> >> > -           .reset_gpio = TEGRA_ULPI_RST,
> >> > -           .clk = "cdev2",
> >> > -};
> >> > -
> >> > -static struct tegra_ehci_platform_data tegra_ehci_pdata[] = {
> >> > -           [0] = {
> >> > -                   .operating_mode = TEGRA_USB_OTG,
> >> > -                   .power_down_on_bus_suspend = 1,
> >> > -           },
> >> > -           [1] = {
> >> > -                   .phy_config = &ulpi_phy_config,
> >> > -                   .operating_mode = TEGRA_USB_HOST,
> >> > -                   .power_down_on_bus_suspend = 1,
> >> > -           },
> >> > -           [2] = {
> >> > -                   .operating_mode = TEGRA_USB_HOST,
> >> > -                   .power_down_on_bus_suspend = 1,
> >> > -           },
> >> > -};
> > 
> > I'm not sure about this.
> > 
> > For paz00 tegra_ehci3_device also leads to an external usb port. With
> > power_down_on_bus_suspend = 1 hotplugging usb devices doesn't work (as
> > the bus might be suspended at the time).
> > 
> > From the initial commit message of ehci-tegra on android.git.kernel.org:
> > "For additional power saving, the driver supports powering down the
> > phy on bus suspend when it is used, for example, to connect an internal
> > device that use an out-of-band remote wakeup mechanism (e.g. a gpio)."
> > 
> > So my guess was, that the correct handling of this was setting
> > power_down_on_bus_suspend to 0 - which did fix the problem.
> > 
> > With this data moving to devices.c, setting this would also change the
> > trimslice behaviour which would be a problem.
> > 
> > In the end I'm not sure if setting power_down_on_bus_suspend to 0 is the
> > correct solution to the hotplug problem or if a better solution exists.
> > 
> > Thanks
> > Heiko
> 
> Can you or someone else test if power_down_on_bus_suspend is causing a
> problem?  If it is, I'll change the default in devices.c to
> power_down_on_bus_suspend=0, and devices that want
> power_down_on_bus_suspend=1 will have to set it.

With power_down_on_bus_suspend = 1, paz00 fails to recognize newly attached 
devices, i.e. only devices connected during boot are accessible.
But it doesn't cause problems with the devices themselfs.

Am Freitag 15 Juli 2011, 20:30:18 schrieb Olof Johansson:
> (Note: I'll post a patch in a bit that keeps the IP clocked even if phy is
> powered down on bus suspend, since otherwise we don't get hotplug events
> for hub-less systems).
This might also fix the hotplug problem on paz00 *fingers crossed*.

Heiko