Re: [PATCH v4 3/7] phy: Add set_mode callback

[Bin Liu]

On Wed, May 04, 2016 at 01:20:36PM -0500, David Lechner wrote:
> On 05/04/2016 01:10 PM, Bin Liu wrote:
> >Hi,
> >
> >On Thu, Apr 14, 2016 at 01:35:14PM -0500, David Lechner wrote:
> >>The initial use for this is for PHYs that have a mode related to USB OTG.
> >>There are several SoCs (e.g. TI OMAP and DA8xx) that have a mode setting
> >>in the USB PHY to override OTG VBUS and ID signals.
> >>
> >>Of course, the enum can be expaned in the future to include modes for
> >>other types of PHYs as well.
> >>
> >>Suggested-by: Kishon Vijay Abraham I <kishon@xxxxxx>
> >>Signed-off-by: David Lechner <david@xxxxxxxxxxxxxx>
> >>---
> >>
> >>v4 changes:
> >>
> >>* This is a new patch to avoid exporting a symbol from the phy-da8xx-usb driver.
> >>
> >>
> >>  drivers/phy/phy-core.c  | 15 +++++++++++++++
> >>  include/linux/phy/phy.h | 15 +++++++++++++++
> >>  2 files changed, 30 insertions(+)
> >>
> >>diff --git a/drivers/phy/phy-core.c b/drivers/phy/phy-core.c
> >>index e7e574d..fe0344c 100644
> >>--- a/drivers/phy/phy-core.c
> >>+++ b/drivers/phy/phy-core.c
> >>@@ -342,6 +342,21 @@ int phy_power_off(struct phy *phy)
> >>  }
> >>  EXPORT_SYMBOL_GPL(phy_power_off);
> >>
> >>+int phy_set_mode(struct phy *phy, enum phy_mode mode)
> >>+{
> >>+	int ret;
> >>+
> >>+	if (!phy || !phy->ops->set_mode)
> >>+		return 0;
> >>+
> >>+	mutex_lock(&phy->mutex);
> >>+	ret = phy->ops->set_mode(phy, mode);
> >>+	mutex_unlock(&phy->mutex);
> >>+
> >>+	return ret;
> >>+}
> >>+EXPORT_SYMBOL_GPL(phy_set_mode);
> >
> >Sorry for my late comments, have been busy on other things.
> >
> >As I commented in v2, isn't it a better idea to not adding this callback
> >and let the da8xx phy driver set the mode register in _probe() based on
> >DT dr_mode?
> >
> >musb core only calls the *optional* _set_mode() during init, so I don't
> >see any problem if the phy set the mode in its probe.
> >
> >Regards,
> >-Bin.
> >
> 
> As was already discussed, the mode can be changed via sysfs as well
> as during probe, so this callback is needed for that case.

Ahh, it seems we have discussed this... I keep forgetting things...

> 
> This is something I actually plan on using because the device I am
> using (LEGO MINDSTORMS EV3) is not wired for OTG, so the callback is
> needed to override the ID and VBUS signals when switching between
> host and peripheral mode.

Have you already tested this? I never tried changing mode via sysfs, but
by quickly reviewing the code, I am wondering how it works. the core
only calls ops->set_mode() but nothing else. To really switch the mode,
the driver has to talk to the root hub, and manipulate the SESSION
bit...

Regards,
-Bin.