On Wed, Sep 09, 2015 at 12:55:24PM -0500, Bin Liu wrote:
> On Wed, Sep 9, 2015 at 12:51 PM, Felipe Balbi <balbi@xxxxxx> wrote:
> > On Wed, Sep 09, 2015 at 12:46:43PM -0500, Bin Liu wrote:
> >> Hi,
> >>
> >> On Wed, Sep 9, 2015 at 12:26 PM, Felipe Balbi <balbi@xxxxxx> wrote:
> >> > On Wed, Sep 09, 2015 at 12:18:27PM -0500, Bin Liu wrote:
> >> >> Hi,
> >> >>
> >> >> On 09/09/2015 08:37 AM, Bin Liu wrote:
> >> >> >Set the Power register HSENAB bit based on musb->config->maximum_speed,
> >> >> >so that the glue layer can control MUSB to work in high- or full-speed.
> >> >> >
> >> >> >Signed-off-by: Bin Liu <b-liu@xxxxxx>
> >> >> >---
> >> >> > drivers/usb/musb/musb_core.c | 10 +++++-----
> >> >> > include/linux/usb/musb.h | 2 +-
> >> >> > 2 files changed, 6 insertions(+), 6 deletions(-)
> >> >> >
> >> >> >diff --git a/drivers/usb/musb/musb_core.c b/drivers/usb/musb/musb_core.c
> >> >> >index 6dca3d7..5d8014e 100644
> >> >> >--- a/drivers/usb/musb/musb_core.c
> >> >> >+++ b/drivers/usb/musb/musb_core.c
> >> >> >@@ -1014,6 +1014,7 @@ void musb_start(struct musb *musb)
> >> >> > {
> >> >> > void __iomem *regs = musb->mregs;
> >> >> > u8 devctl = musb_readb(regs, MUSB_DEVCTL);
> >> >> >+ u8 power;
> >> >> >
> >> >> > dev_dbg(musb->controller, "<== devctl %02x\n", devctl);
> >> >> >
> >> >> >@@ -1021,11 +1022,10 @@ void musb_start(struct musb *musb)
> >> >> > musb_writeb(regs, MUSB_TESTMODE, 0);
> >> >> >
> >> >> > /* put into basic highspeed mode and start session */
> >> >> >- musb_writeb(regs, MUSB_POWER, MUSB_POWER_ISOUPDATE
> >> >> >- | MUSB_POWER_HSENAB
> >> >> >- /* ENSUSPEND wedges tusb */
> >> >> >- /* | MUSB_POWER_ENSUSPEND */
> >> >> >- );
> >> >> >+ power = MUSB_POWER_ISOUPDATE;
> >> >> >+ if (musb->config->maximum_speed == USB_SPEED_HIGH)
> >> >> >+ power |= MUSB_POWER_HSENAB;
> >> >>
> >> >> Well, I think this breaks other glue drivers, in which case that
> >> >> musb->config is not allocated, or config->maximum_speed is 0 by default if
> >> >> not set in dts. So, to avoid patching all other glue drivers, is the
> >> >> following change acceptable?
> >> >>
> >> >> + if (musb->config->maximum_speed != USB_SPEED_FULL &&
> >> >> + musb->config->maximum_speed != USB_SPEED_LOW)
> >> >> + power |= MUSB_POWER_HSENAB;
> >> >
> >> > /*
> >> > * treating UNKNONW as unspecified maximum speed, in which case
> >> > * we will default to HIGH speed.
> >> > */
> >> > if (musb->config->maximum_speed == HIGH ||
> >> > musb->config->maximum_speed == UNKNOWN)
> >> > power |= HSENAB;
> >> >
> >> > how about that ?
> >>
> >> Yeah, It is easy to read, but is musb->config NULL in am35x.c,
> >> tusb6010.c, davinci.c, and other old style gules? I am unable to
> >> follow their init routine to see how musb->config is initialized.
> >
> > config is never NULL, if it is we have bigger problems ;-) Note that
> > musb->config is also used to hold fifo configuration for endpoint setup.
>
> I noticed fifo config is in musb->config, but just unable to figure
> out how it is initialized in the old style glues. :( I will read the
> code again whenever I get bored.
see allocate_instance()
--
balbi
Attachment:
signature.asc
Description: Digital signature
