[U-Boot] Regarding patch: http://patchwork.ozlabs.org/patch/373593/

[Ramneek Mehresh]

> -----Original Message-----
> From: Marek Vasut [mailto:marex at denx.de]
> Sent: Monday, April 06, 2015 8:42 PM
> To: Mehresh Ramneek-B31383
> Cc: u-boot; sr at denx.de
> Subject: Re: Regarding patch: http://patchwork.ozlabs.org/patch/373593/
> 
> On Monday, April 06, 2015 at 10:58:30 AM, Ramneek Mehresh wrote:
> 
> [...]
> 
> > > > > The static void fsl_xhci_core_exit(struct fsl_xhci *fsl_xhci)
> > > > > must shut down the controller, which I don't see happening. Why?
> > > >
> > > > I could not locate any such requirement in IP documentation. Have
> > > > contacted local IP/PHY team for the same. Waiting for response
> > > > from them
> > >
> > > This is needed, so you don't start Linux with a running USB controller.
> >
> > xhci controller is already stopped in
> > usb_stop->usb_lowlevel_stop->xhci_reset(). I could see CMD_RUN bit
> > getting reset in this function before the controller is reset. So,
> > from your previously stated requirement, controller is halted when Linux is
> started.
> >
> > Other people are shutting down PHY as part of xhci_core_exit(), not
> > the controller!! We would not like to re-start and re-configure PHY
> > inside Linux, and take phy initialization inside bootloader. I got
> > word from hw team that they do not support phy shutting down from sw.
> > hence, we do not have any sequence for current socs to shut down phy
> > from sw. if required, I'll put forward a request to provide this control in
> future socs.
> 
> Hi,
> 
> Is this similar hardware bug to the one which MX6 PCIe is suffering ? On MX6,
> the bug is that you cannot reset the PCIe and PCIe PHY from software, which
> means that if you start PCIe in U-Boot, you cannot reliably use it in Linux.
> Furthermore, if you reset the MX6 via WDT, you cannot start PCIe reliably in
> Linux even if PCIe is not used in U-Boot at all.
> 
> Is this the same type of hardware screwup where the design team didn't
> think reset was necessary?
> 
I have raised request for this feature in up-coming socs, and they have agreed to
provide phy shut-down in future revs. However, this support is not available in current
revs for which the code is sent. What do you suggest we should do for current socs?

> Best regards,
> Marek Vasut