module: Implement HAVE_MOD_ARCH_SPECIFIC ...

Implement HAVE_MOD_ARCH_SPECIFIC Kconfig and module_frob_arch_sections()
function prototype from Linux module subsystem.
module_frob_arch_sections() should be implemented by any architecture
that selects HAVE_MOD_ARCH_SPECIFIC, and is called from load_module()

Signed-off-by: David Dgien <dgienda125@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

module: Fix module command registration ...

Commit 1e4a948673d7 ("command: Use array of pointers to commands")
updated the command linker sections to contain an array of pointers to
the command structs, rather than the command structs themselves.
However, the load_module function was not updated to account for the
change to the linker section, so fix that here.

Additionally, commit 22bdecc9c591 ("lds: Move start/end address
variables into defines") moved the _start and _end symbols out of the
linker script and into the section macro defines, causing them to show
up in the module symbol table as well, so we take can advantage of those
to find the array of pointers.

Fixes: 1e4a948673d7 ("command: Use array of pointers to commands")
Signed-off-by: David Dgien <dgienda125@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

module: Fix adding module to list after layout ...

During load_module(), the 'this_module' section is relocated, but the
pointer to the module struct isn't updated to account account for the
move. Do so before adding the module to the module_list.

As a side effect of properly pointing to the relocated module struct, we
no longer need to manually search for and fixup the init_module symbol,
so remove that code.

Signed-off-by: David Dgien <dgienda125@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

module: Add init macros to module.h ...

Port module initialization macros and initcall redefinition macros from
Linux

Signed-off-by: David Dgien <dgienda125@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

Makefile: Initialize and export KBUILD variables ...

Initialize and export KBUILD KERNEL and MODULE variables. Otherwise,
appending to them elsewhere in the srctree will not be seen by the build
commands in Makefile.lib

Signed-off-by: David Dgien <dgienda125@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

kbuild: fix typo in --no-whole-archive flag ...

"whole" was replaced with "while" during a previous patch. Fix this flag
to allow building correctly.

Fixes 0a78b33954 ("kbuild: switch over to thin archive")
Signed-off-by: Clement Leger <cleger@kalray.eu>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

ARM: vexpress: dts: fix barebox partitioning breakage ...

There is no /smb@4000000 node upstream any longer as it was
renamed with the 8d158e1a4091 ("dts: update to v5.7-rc1") sync.

Adjust the barebox device tree accordingly. It's likely that this
resulted in barebox v2020.06.0 state, environment and update breakage on
the vexpress boards.

Cc: Roland Hieber <rhi@pengutronix.de>
Fixes: 8d158e1a4091 ("dts: update to v5.7-rc1")
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

ARM: dts: align all applicable memory nodes with spec and convention ...

The spec[1] says that top level memory nodes `shall have device_type =
"memory"'. Since 8226f7f9 ("of: base: parse all available memory nodes"),
we parse only nodes with device_type = "memory", so we should add the
device_type property where ever it's missing. When the nodes were first
added, this wasn't an issue because the skeleton.dtsi defined a memory
skeleton with correct device_type, but this was removed upstream and
imported into barebox along with 1dc748b3 ("dts: update to v5.1-rc1").

Upstream device trees with a memory node were adapted to have
device_type = "memory" in-line. This wasn't the case for nodes defined
only within barebox. Rectify this now and rename the nodes to feature a
unit name as well to be aligned with the upstream trend.

[1]: https://github.com/devicetree-org/devicetree-specification/releases/download/v0.3/devicetree-specification-v0.3.pdf

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

ARM: dts: fix now duplicate memory nodes ...

With the upstream migration to memory@BASE_ADDRESS naming scheme in the
device tree of some SoC families, we now have multiple overlapping memory
nodes in some of the barebox device trees. Rename the defunct overrides, so
they once again override the upstream nodes.

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

ARM: i.MX50: fix renamed node reference ...

There is no /soc/aips@50000000 node upstream any longer as it was
renamed with the 8d158e1a4091 ("dts: update to v5.7-rc1") sync.

Adjust the barebox device tree accordingly. It's likely that this
resulted in barebox v2020.06.0 USB connectivity breakage for the
i.MX50's second USB PHY.

Fixes: 8d158e1a4091 ("dts: update to v5.7-rc1")
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

ARM: rpi: dts: fix MMIO resources clash breaking boot ...

The rpi3b device tree specifies 1G of RAM in the memory node, but maps
the highest 16M of RAM's physical address range for peripheral use.

When support for device tree probing was first added for the rpi3b
in e8fab18835a1 ("ARM: rpi: switch to DT probe and multi-image build"),
it had a memory { } node overriding the kernel node size with 0, so the
size is computed later correctly.

1dc748b3b202 ("dts: update to v5.1-rc1") came and renamed the upstream
memory { } node to memory@0. So instead of overriding the old, we
now created a second dummy along with the broken node.
barebox did only parse the dummy, so it continued to limp along.

8226f7f90973 ("of: base: parse all available memory nodes") started
considering both nodes and rendered barebox v2020.05.0 unable to start
up on the rpi3b because of the conflict between RAM range and peripherals
like the bcm2835_mbox caused by parsing the broken memory node.

At least one board also was broken for the same reason[1], so a more
robust approach is in order. For nodes that lack a phandle, the
&{/path} syntax can be used instead. This has the benefit that should
the path disappear in future, a compile failure would result.

[1] 0d26ce1c11a1 ("ARM: Phytec phyFLEX i.MX6: delete wrong memory node")

Fixes: 1dc748b3b202 ("dts: update to v5.1-rc1")
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

Merge branch 'for-next/spdx'

Merge branch 'for-next/sandbox-compile-test'

Merge branch 'for-next/ratp'

Merge branch 'for-next/mmc'

Merge branch 'for-next/misc'

Merge branch 'for-next/kvx'

Merge branch 'for-next/kbuild'

Merge branch 'for-next/imx'

Merge branch 'for-next/fastboot'

Merge branch 'for-next/dts'

Release v2020.06.0 ...

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

mci: sdhci: Fix swapped argument for sdhci_write32 ...

sdhci_write32 accepts the register as second argument, not third.
Fix this up.

Fixes: 075179763a ("mci: sdhci: add Atmel SDHCI (sama5d2, sam9x60) support")
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

dts: update to v5.7 ...

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

dts: update to v5.7-rc7 ...

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

dts: update to v5.7-rc6 ...

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

dts: update to v5.7-rc5 ...

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

MIPS: correctly work with disabled CONFIG_MMU ...

There are some issues with CONFIG_MMU=n in MIPS configuration
(e.g. qemu-malta_defconfig):

  * "Cannot request SDRAM region for stack" error message on start;
  * empty kseg1_ram0 region in the iomem command output.

The patch fixes the problems by forcing KSEG1 memory region usage
for CONFIG_MMU=n MIPS configurations.

Signed-off-by: Antony Pavlov <antonynpavlov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

mtd: nand: base: fix use of uninitialized struct member ...

ooblen is read in nand_do_read_ops, despite never having been assigned a
value. Because ooblen is used to indicate how many bytes of oobbuf are
usable and oobbuf is NULL, use 0 as initial value.

Found on ARCH=sandbox using clang-analyzer.

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

sandbox: os: add_image: fix memory leak ...

devname is strdup'd few lines later, so remove the earlier strdup.
The pointer isn't stored anywhere persistent in between.

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