Merge branch 'for-next/clk'

Merge branch 'for-next/block'

Merge branch 'for-next/at91'

Merge branch 'for-next/arm'

Release v2020.03.0 ...

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

ARM: at91: add sama5d27-based Groboards Giant Board support ...

The Groboards Giant Board is a ATSAMA5D27C-D1G SiP-based SBC.
The board features a 500MHz ARM Cortex-A5 and 128MB DDR2 SDRAM in the
SiP as well as a MicroSD slot on the PCB.

barebox doesn't yet support the sama5d2 SDHCI-variant, so board support
is limited to toggling the LED and interfacing with the UART, but add
this now till the rest may follow. Device tree is based on the vendor's
device-tree available at https://github.com/Groboards/giantboard-tools

Acked-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

ARM: at91: sama5d27-som1: branch to noinline function after stack setup ...

We need to branch out as soon as possible after setting up the stack.
Do this and thereby avoid an access violation we now run into using
OSELAS.Toolchain-2019.09.1.

Fixes: 8bfb1852cf6a ("ARM: at91: add basic sama5d2-som1-ek1 support")
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

ARM: at91: sama5d27-som1: clean up lowlevel.c ...

In ARM, stack decrements before storing values, so decrementing a number
from the stack base is unnecessary. Existing instances of this were
removed in 6b3dc4abd8 ("ARM: Cleanup stack offset cargo cult"), but the
sama5d27-som1 board was added later than that. Fix the cargo cult here
as well and while at it, turn a macro that doesn't need to be one
into a static inline function.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

ARM: at91: add support for SAM9 SoC reset ...

System reset wasn't supported on the sama5 so far.
Add a driver to remedy this.

Reviewed-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

pinctrl-single: fix parsing of `pinctrl-single,bits` property ...

According to dts/Bindings/pinctrl/pinctrl-single.txt:113 the correct order of
parameters is <offset, value, mask>.

Fixes: f7e0cae43 ('pinctrl-single: Handle "pinctrl-single,function-mask" dt property')
Signed-off-by: Peter Mamonov <pmamonov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: Add extra delay for device scan ...

Since b1d9837182, detection of various USB flash drives stopped working.
While the changes itself look correct, probably the HUB requires some
"recovery time" between the GET_STATUS commands in order to be able to
detect devices.

Our configuration: i.MX6ULL (USB-OTG) with only one port on the root
HUB. This port is connected to a Microchip USB4916 hub (permanently
attached). The USB4916 is configured with it's power on defaults which
means that bPwrOn2PwrGood is set to the maximum (0xFF = 510 ms).

Fixes: b1d9837182 ("usb: Change power-on / scanning timeout handling")
Signed-off-by: Christian Eggers <ceggers@arri.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

Documentation: blspec: drop documentation of devicetree-overlay ...

The devicetree-overlay key is part of the official Boot Loader
Specification. There is no need to document it in Barebox, again.

Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

Documentation: blspec: update link to Boot Loader Specification ...

There is a big warning on the linked page that the page is obsolete and
a link to the current version of the Boot Loader Specification. The
Barebox documentation shall directly link to the current version of the
specification.

Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

Revert "block: Adjust cache sizes" ...

On 12/13/19 2:12 PM, Sascha Hauer wrote:
> On Tue, Dec 10, 2019 at 03:44:52PM +0100, Hubert Feurstein wrote:
>> With v2015.06.0 the indicated progress of the copy command is very
>> smooth. Calling "cp -v /dev/zero /dev/mmc3.root" takes about 80
>> seconds for 256MB. But with v2019.12.0 the progress is very bumpy and
>> the copy takes about 280 seconds.
>>
>> I've tracked this down to this commit which destroys the performance:
>> "block: Adjust cache sizes" (b6fef20)
>
> We could just revert this patch. I can't find any workload that gets
> faster with b6fef20c1215. It's rather the other way round.

Do this by reverting commit b6fef20.

Reported-by: Hubert Feurstein <h.feurstein@gmail.com>
Suggested-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: gadget: fastboot: call discard_range for sparse files as well ...

We can save some time by explicitly telling the block layer that a range
is unused before overwriting it. This brought time my sample write
from 445s to 376s.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

USB: MUSB: defer driver probes where necessary ...

Kernel commit 0782e8572c ("ARM: dts: Probe am335x musb with ti-sysc")
which we pulled in during the v2020.02.0 dts/ sync moved the USB nodes
to be under a ti-sysc bus instead of ti,am33xx-usb.

This new probe order broke am335x USB under barebox, because the MUSB
drivers couldn't cope with the now different device probe order.

Pepper some -EPROBE_DEFER around to make USB work again.

Fixes: 574eed3f6f ("dts: update to v5.5-rc1")
Reported-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Tested-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

USB: MUSB: PHY: scrap singleton am335x_get_usb_phy() ...

am335x_get_usb_phy() retrieves the last probed USB phy. On the BeagleBone
with both PHYs enabled, this means that dependent on probe order, both
MUSB instances could end up with the same PHY.

Remove the global variable and have the MUSB driver parse the "phys"
property instead.

The cleaner way to achieve this would be to migrate phy-am335x.c
and phy-am335x-control.c to the generic phy framework and have MUSB use
of_phy_get, alas, even Linux hasn't done this so far and we need
a short patch for master anyway, thus just do it the easy way.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Tested-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

Documentation: boards: stm32mp: document eMMC partition layout ...

For eMMC, the boot partitions are used for FSBL instead of having
these two partitions in the user area, like with SD-Cards.
Document this.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

ARM64: add support for booting a FIT image ...

Add the image type handler to the ARM64 boot code. The only difference
in the boot handling is that we need to look at the FIT loaded OS image
header to get the image load and text offsets, as the os_header is the
FIT header, not the kernel header.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

clk: print more consistent clock states ...

In clk_dump we only print the state of clk_is_enabled(). Depending on
the clock this can mean different things. When the clock provides an
is_enabled() callback it will print its result, so the enabled state
matches the hardware state of that clock, but doesn't necessarily mean
its parents are enabled. If a clock does not provide an is_enabled()
callback then we rely on our internal enable_count tracker.

Change this to always print the hardware state of the current clock. It
can be:

- disabled: The clock is disabled (as provided by is_enabled())
- enabled: The clock is enabled (as provided by is_enabled())
- always enabled: The clock can't be disabled (no is_enabled callback
                  and can't be enabled/disabled)
- unknown: no is_enabled callback but can be enabled/disabled

Additionally we print the enable_count variable, so from enable_count !=
0 we can know that barebox has enabled the clock.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

clk: Do not let the enable count of critical clocks go below 1 ...

The enable count of critical clocks may not drop below 1. Otherwise
the new parent during a reparenting of a critical clock will not
be enabled.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

clk: initially enable critical clocks ...

Critical clocks have to be enabled initially. We have to do this not
only to enable the clock itself, but also to make sure its parents
stay enabled and also to correctly enable the new parents during
reparenting of the clock.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

Merge branch 'for-master/imx-external-nand'

commands: boot: fixup create boot entries on demand ...

Don't try other boot entries if one succeeds. This is the behavior
before the change, thus reinstate it.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

ARM: i.MX: external NAND boot: Fix passing boarddata ...

We used to pass a device tree pointer to
imx*_barebox_boot_nand_external(). This no longer works since we compile
barebox with -fPIE.

The problem is that we only have the first 2KiB of the image available
when we start from NAND. The pointer to the device tree is not valid
yet. This hasn't been a problem because the pointer is only used later
when we have the full image in SDRAM. With -fPIE though the actual
pointer is derived indirectly through another pointer, and that pointer
to the pointer is also not valid which causes an invalid pointer to
the device tree later.

This is solved by splitting the NAND boot functions up into two
functions. With imx*_nand_relocate_to_sdram() the board code relocates
the image out of the NFC SRAM into SDRAM. From that point on the device
tree pointer can be accessed and the board can load the image using
imx*_nand_load_image() and can afterwards pass control to barebox
passing whatever device tee pointer it likes.

Fixes: b0348d677b ("ARM: Compile with -fPIE")
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

ARM: i.MX: external NAND boot: Leave icache disabled ...

It seems running from the NFC SRAM doesn't work with the instruction
cache enabled, it leads to corruptions on the i.MX27. We stumbled upon
this earlier and the solution at that time was to disable the
instruction cache in the NAND boot code. It is, however, more reliable
to just not enable the instruction cache in the first place.
This is not particularly nice as we have to ifdef this in generic code,
duplicate arm_cpu_lowlevel_init(), or call arm_cpu_lowlevel_init() later
when we are out of NFC SRAM. From the different bad solutions I chose
to ifdef the instruction cache away. It will be enabled later in the
common cache functions.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

ARM: i.MX: external NAND boot: remove unnecessary arguments from imx*_nand_load_image ...

The imx*_nand_load_image functions all take dest, size, nfc_base and
pagesize_2k arguments. All of these can be determined inside the functions,
so remove them.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

net: designware: eqos: stm32: drop no longer needed syscfg-clk ...

The upstream device tree no longer lists "syscfg-clk" and the syscon node
contains a clocks property that the barebox syscon driver now
automatically clocks before access.

Thus drop the requirement that "syscfg-clk" is specified.

Fixes: 4aaceef58 ("dts: update to v5.6-rc1")
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

mfd: syscon: clock peripheral if specified in device tree ...

Linux supports a clocks property in syscon nodes to indicate that access
should only occur with the clock active.

Attach the clock to the regmap if found.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

mfd: syscon: use regmap-mmio as backend ...

Linux supports a clocks property in syscon nodes to indicate that access
should only occur with the clock active. The new regmap-mmio supports
this, thus port syscon over, so it can make use of it in a later commit.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>