i.MX7 clock support broken

[s.hauer@pengutronix.de (Sascha Hauer)]

On Wed, Feb 08, 2017 at 04:47:06PM +0000, A.S. Dong wrote:
> Hi Sascha,
> 
> > -----Original Message-----
> > From: Sascha Hauer [mailto:s.hauer at pengutronix.de]
> > Sent: Tuesday, February 07, 2017 10:22 PM
> > To: linux-arm-kernel at lists.infradead.org
> > Cc: Shawn Guo; A.S. Dong; Fabio Estevam; kernel at pengutronix.de
> > Subject: i.MX7 clock support broken
> > 
> > Hi All,
> > 
> > I recently adopted the i.MX7 clock tree from Linux to barebox and wondered
> > why ethernet doesn't work.
> 
> I just reproduced the issue on MX7D SDB board with latest upstream kernel
> That kernel NULL point crash during boot on
> fec_enet_open->..->phy_attach_direct.
> 
> Not sure if the same as yours.
> (Paste log at last)
> 
> Andy,
> Can you help check it?
> 
> > I found out that the clock tree in Linux is
> > quite different from what the reference manual describes and also quite
> > different from the hardware I have on my desk. There are several clocks
> > that only exist in the driver, but not in the actual hardware and also
> > several clocks which are in hardware, but not in the driver.
> > 
> > These clocks exist in the driver only, but not in the hardware:
> > 
> 
> Please refer to secion 5.2.5 in RM.
> They're LPCG gates.
> 
> But I did notice some of them seem not find in RM.
> e.g. IMX7D_ARM_A7_ROOT_CLK, IMX7D_ARM_M0_ROOT_CLK, IMX7D_SPDIF_ROOT_CLK ...
> 
> It might be true that not all listed in RM as I see a Note
> " Not all CCGRs are mapped " right upon Table 5-19.

Ok, I try to be more explicit:

The current clock driver controls ethernet gates at CCM_BASE + 0x44e0 and
CCM_BASE + 0x44f0. These registers do not have any effect on the
hardware. The real gate for enet1 is at CCM_BASE + 0x4700, also
documented in the reference manual. Changing this register really has
an effect on the hardware. Or to put it different: Ethernet won't work
in the kernel when the (real) gates have not been previously enabled in
the bootloader.


> 
> Overally the current imx7d clk driver in upstream seems intend to be
> an initial version and may miss many local fix patches.
> 
> You can also refer to:
> http://git.freescale.com/git/cgit.cgi/imx/linux-imx.git/?h=imx_4.1.15_2.0.0_ga

This tree has the same bugs.

Sascha

-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |