Merge branch 'for-next/module'

arm: module: Allow modules outside of bl range ...

Unlike the Linux kernel, barebox does not have a dedicated heap for
storing modules. Therefore, if the system memory configuration places
the general heap further away than can be reached by a 'bl' instruction
(24 bits of address, or 16 MiB), then the module relocations will fail
due to being out of range.

Allocate PLTs when loading modules so that jumps and calls whose
targets are too far away for their relative offsets to be encoded
in the instructions themselves can be bounced via veneers in the
module's PLT. The modules will use slightly more memory, but after
rounding up to page size, the actual memory footprint is usually
the same.

Adoption of Linux commits:

66e94ba3c8ea ARM: kernel: avoid brute force search on PLT generation
1031a7e674d1 ARM: kernel: sort relocation sections before allocating PLTs
05123fef0982 ARM: kernel: allocate PLT entries only for external symbols
35fa91eed817 ARM: kernel: merge core and init PLTs
7d485f647c1f ARM: 8220/1: allow modules outside of bl range

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

arm: makefile: Fix compiler flag variable ...

In order for ARM modules to compile with the correct relocation types,
they must be built without -fPIE. Move -fPIE from KBUILD_CPPFLAGS to
KBUILD_CFLAGS_KERNEL so that the flag is only included when compiling
files being built into the barebox image.

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

kbuild: remove AS variable ...

The assembler files in the barebox are *.S instead of *.s, so they must
be preprocessed. Since 'as' of GNU binutils is not able to preprocess,
we always use $(CC) as an assembler driver.

Remove AS variable.

Linux also removed AS. See Linux commits:

 aa824e0c962b ("kbuild: remove AS variable")
 1ca0c2f61211 ("kbuild: remove unused AS assignment")

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

arch: remove KBUILD_SRC checks from arch Makefile ...

You can always add $(srctree)/ to the mach include path.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

kbuild: prefix compiler flag variables with KBUILD_ ...

In old days, Linux also used to use well-known variables such CFLAGS,
CPPFLAGS, etc.

They were prefixed with KBUILD_ presumably for preventing users from
overriding them accidentally.

Rename as follows:
  CFLAGS          ->  KBUILD_CFLAGS
  AFLAGS          ->  KBUILD_AFLAGS
  CPPFLAGS        ->  KBUILD_CPPFLAGS
  LDFLAGS         ->  KBUILD_LDFLAGS
  HOSTCFLAGS      ->  KBUILD_HOSTCFLAGS
  HOSTCXXFLAGS    ->  KBUILD_HOSTCXXFLAGS
  HOSTLDFLAGS     ->  KBUILD_HOSTLDFLAGS
  HOST_LOADLIBES  ->  KBUILD_HOSTLDLIBS

HOSTCFLAGS, HOSTCXXFLAGS, HOSTLDFLAGS, HOSTLDLIBS are re-used to allow
users to pass-in additional flags to the host compiler.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

images: Use separate LDFLAGS for PBL ...

When setting CONFIG_PBL_RELOCATABLE=y (selected by PBL_MULTI_IMAGES) and
CONFIG_RELOCATABLE=n (trying to reduce image size), the PBL is also
linked with "-static" instead of "-pie". The result is a non-working
PBL.

As a side effect, also get rid of erroneous "-Map barebox.map" when
linking the PBL.

Only changed for ARM, are any other platforms affected?

Signed-off-by: Christian Eggers <ceggers@arri.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

ARM: Compile with -fPIE ...

Early code in barebox often runs at an address the binary is not linked
at. This causes problems for example when simple initializations in a
switch are converted to an array lookup (-ftree-switch-conversion).
These arrays are then addressed where they are linked at. Some code
where this is known to cause problems is already compiled with
-fno-tree-switch-conversion. This however is limited to single files.

This patch compiles barebox with -fPIE enabled. This causes such tables
to be looked up PC relative rather than with its absolute address. This
makes the -fno-tree-switch-conversion and -fno-jump-table options
unnecessary.

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

ARM: remove now unused PBL code ...

Single PBL is now built as part of the multi PBL images, so drop the now
unused single PBL code.

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

ARM: NETX: remove architecture ...

NETX support has been removed from the Kernel, so there's no point in
supporting it in barebox any longer. Remove the architecture.

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

ARM: stm32mp1: rename to stm32mp ...

Serial and clk driver both depend on CONFIG_ARCH_STM32MP1,
so either the Kconfig symbol or their depend needs to change.

Patches posted by the vendor to Linux, U-Boot and their BSP
Yocto-Layer speak of a STM32MP-Family of which the STM32MP1
is the first series, thus rename the arch by dropping the 1.

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

arch: add KBUILD_DEFCONFIG to allow "make defconfig" ...

Currently, "make defconfig" is only supported for sandbox and x86.

I set KBUILD_DEFCONFIG to support "make defconfig" where I was able
to find a suitable one.

nios2 and openrisc have a single generic_defconfig, so it is definitely
good for KBUILD_DEFCONFIG.

For arm, mips, I assume qemu is a preferred choice over a real board.

blackfin has only one defconfig. Since it is inactive (it is gone in
Linux), so no more defconfig will come in.

I am not sure about ppc and riscv, so I did not touch them.

Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

Merge branch 'for-next/stm32'

Merge branch 'for-next/layerscape'

ARM: Makefile: restore kindle3 imxcfg-y setting ...

Commit 45f38c1a846f ("ARM: Makefile: Drop unnecessary imxcfg-y settings")
dropped the imximg generation for the kindle3. This device is not covered
within multiimage support and can be configured with kindle3_defconfig.

Signed-off-by: Alexander Kurz <akurz@blala.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

ARM: Add basic Layerscape support ...

This adds basic Layerscape support:

- Makefile/Kconfig
- Register maps
- errata workarounds

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

ARM: Add initial STM32MP1 support ...

This adds initial STMicroelectronics MP1 support along with support
for the DK2 devel board. Only very basic support:

- UART
- SDRAM memory base/size
- No 1st stage support

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

ARM: Makefile: Drop unnecessary imxcfg-y settings ...

For boards built with multiimage support imxcfg-y is not needed.
Remove it for these boards.

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

Merge branch 'for-next/ultrascale'

ARM: CCMX51: Switch to multiimage support ...

This is a cumulative patch for the Digi ConnectCore CCMX51 SOM.
It includes:
- Switch board to devicetree probe.
- Add MMC update handler.
- Switch to multiimage support.
- Cleanup and optimize board code.

Signed-off-by: Alexander Shiyan <shc_work@mail.ru>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

ARM: zynqmp: add support for Xilinx ZCU104 board ...

Add support for the Xilinx Zynq Ultrascale+ MPSoC architecture (ZynqMP)
and the Xilinx ZCU104 board.

Barebox is booted as BL33 in EL-1 and expects that a BL2 (i.e. the FSBL)
already took care of initializing the RAM. Also for debug_ll, the UART
is expected to be already setup correctly. Thus, you have to add the
Barebox binary to a boot image as described in "Chapter 11: Boot and
Configuration" of "Zynq Ultrascale+ Device Technical Reference Manual".

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

ARM: aarch64: compile with general-regs-only ...

Without this flag, gcc generates code to save the Q/V registers to the
stack for handling the va_list in pr_print(). Saving the registers is
useless, as the registers are never restored, but accessing the
registers to save them hangs the CPU.

Follow the Linux arch/arm64/Makefile and use the general-regs-only flag
to prevent usage of floating point and Advanced SIMD register.

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

Merge branch 'for-next/imx'

ARM: aarch64: disable 32bit boot commands ...

The 32bit boot support is full of legacy cruft. While it should be
possible to merge the aarch64 support there, a fresh start looks more
promising.

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

ARM: aarch64: compile with strict alignment ...

barebox runs with MMU disabled at least during startup. We need
-mstrict-alignment for these parts to avoid alignment aborts.

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

i.MX53/TX53: rework to dts based boot ...

Since nobody likes to use platformcode based machines any more, we also
switch this one to use dts based booting.

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

ARM: Add 64bit compilation alternative ...

Our 64bit file variants have the suffix "_64". This adds a make variable
that is either empty or contains "_64" which can be used to easily
alternatively compile the 32bit or 64bit variant.

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

ARM: Makefile: format fix ...

Signed-off-by: Juergen Borleis <jbe@pengutronix.de>

arm: Add both .lds files to CLEAN_FILES unconditionally ...

'clean' target is listed on 'no-dot-config-targets' list in main
Makefile so that conditional statement would yeild the same result
every time. Given how CLEAN_FILES are rm'ed with -f there should be no
harm in specifying them both unconditionally.

Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

fix ARMv8 interference with ARMv7 ...

CONFIG_CPU_V8 is checked against 'y' and 'n', but the case when the
variable is unset is not considered.
This patch only checks the variable against a single value 'y' so the
logic is always coherent even when the variable is unset.

Signed-off-by: Vicente Bergas <vicencb@gmail.com>
Tested-by: Vicente Bergas <vicencb@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>