Re: [PATCH] ARM i.MX6SL: Add Kindle generation 6/7 support

[Sascha Hauer <s.hauer@pengutronix.de>]

On Tue, Nov 11, 2025 at 07:32:13AM +0000, Alexander Kurz wrote:
> Hi Sascha,
> 
> On Mon, 10 Nov 2025, Sascha Hauer wrote:
> 
> > 
> > Hi Alexander,
> > 
> > On Sun, Nov 09, 2025 at 06:03:37PM +0000, Alexander Kurz wrote:
> > > Four Amazon Kindle 6th and 7th generation e-book readers were considered
> ...
> > > diff --git a/Documentation/boards/imx/amazon-kindle-6-7.rst b/Documentation/boards/imx/amazon-kindle-6-7.rst
> > 
> > Thanks for the good documentation, very appreciated :)
> Thanks!
> 
> > 
> > > new file mode 100644
> > > index 0000000000..c7a5ef9424
> > > +ENTRY_FUNCTION(start_imx6sl_kindle6_dp75sdi, r0, r1, r2)
> > > +{
> > > +	void *fdt;
> > > +
> > > +	imx6_cpu_lowlevel_init();
> > > +
> > > +	arm_setup_stack(0x00920000);
> > > +
> > > +	if (IS_ENABLED(CONFIG_DEBUG_LL)) {
> > > +		writel(0x4, 0x020e016c);
> > > +		imx6_uart_setup_ll();
> > > +	}
> > > +
> > > +	fdt = __dtb_imx6sl_kindle6_dp75sdi_start + get_runtime_offset();
> > > +	barebox_arm_entry(0x80000000, SZ_256M, fdt);
> > 
> > Does imx6ul_barebox_entry() work here instead? In that case you could
> > drop the memory nodes from the device trees.
> Replacing barebox_arm_entry() with imx6ul_barebox_entry() seems to work
> so far (have only tested it on one device), but will this really make the 
> memory nodes from the device trees obsolete as well? Some devices have 256 
> while others 512MB RAM, when the amount of memory is not hardcoded, is it 
> auto-determined?

Yes, the memory nodes will be autogenerated from the configuration read
back from the DRAM controller.

> 
> > 
> > > +++ b/arch/arm/dts/imx6sl-kindle-nm460gz.dts
> > > @@ -0,0 +1,69 @@
> > > +// SPDX-License-Identifier: GPL-2.0-or-later
> > > +/*
> > > + * Copyright (C) 2025 Alexander Kurz <akurz@blala.de>
> > > + */
> > > +
> > > +/dts-v1/;
> > > +#include "imx6sl-kindle-common.dtsi"
> > > +#include <dt-bindings/interrupt-controller/irq.h>
> > > +#include <dt-bindings/gpio/gpio.h>
> > > +#include <dt-bindings/input/input.h>
> > > +
> > > +/ {
> > > +	model = "kindle-nm460gz";
> > > +	compatible = "amazon,imx6sl-kindle-nm460gz", "fsl,imx6sl";
> > > +	barebox,disable-deep-probe;
> > 
> > Is this intentional? What doesn't work with deep probe enabled?
> With barebox,deep-probe mmc1 will not be available when kindle_rev_init()
> is called at late_initcall(kindle_mx6sl_late_init).
> This way, the required ATAGs will be missing.
> Will deep-probe influence device_initcall(kindle6_device_init)?

deep probe will probe devices on demand for you. What's left to do for
you is to say that you need the mmc1 device by doing a:

of_device_ensure_probed_by_alias("mmc1");

This will handle the dependencies like probing regulators or clocks
regardless of the initcall level you are in.

You also have this in your patch:

+       /* Probe the eMMC to allow reading the board serial and revision */
+       dev = get_device_by_name("mmc1");
+       if (dev)
+               dev_set_param(dev, "probe", "1");

which can be replaced by a:

device_detect_by_name("mmc1");

The different initcall levels in your board code shouldn't be necessary
anmore with this.

Sascha

-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |