Release v2020.01.0 ...

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

usb: ehci: Do not use memset on dma coherent memory ...

memset is an optimized operation that at least on ARM64 may only be
called on cached memory, see 32e8842c40 ("ARM: lib64: Make string
functions aware of MMU configuration"). To avoid crashes in the ehci
driver we no longer call memset on memory allocated with
dma_alloc_coherent(), but use a simple memzero32 function instead.

Reported-by: Elmar Albert <EAlbert@data-modul.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

lds: align RO_DATA_SECTION members ...

When using the RO_DATA_SECTION macro, all the individual components
are put into the rodata section. As they now end up in the same
section there is no alignment enforced between them anymore. This
leads to unaligned start pointers when any of the components have a
unaligned size. Enforce some alignment between them.

Fixes: c5d38e920101 (lds: Add and use RO_DATA_SECTION macro)
Signed-off-by: Lucas Stach <dev@lynxeye.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

ARM: i.MX: Fix PBL size test ...

MAX_PBL_MEMORY_SIZE needs the start symbol name without any suffix. Fix
this to make the PBL size tests effective.

Reported-by: Ulrich Ölmann <u.oelmann@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

Merge branch 'for-next/zynq'

Merge branch 'for-next/watchdog'

Merge branch 'for-next/usb'

Merge branch 'for-next/stm32'

Merge branch 'for-next/shebang'

Merge branch 'for-next/sdhci'

Merge branch 'for-next/randfixes'

Merge branch 'for-next/pci'

Merge branch 'for-next/misc'

Merge branch 'for-next/layerscape'

Merge branch 'for-next/imx'

Merge branch 'for-next/gpio'

Merge branch 'for-next/fsl-fman'

Merge branch 'for-next/dts'

Merge branch 'for-next/arm'

misc: add ACPI test driver ...

We have a bus driver for ACPI, but no drivers yet. Add a simple (rather
useless) driver that exercises the API, which future drivers can be
based on. As soon as we have proper ACPI drivers, this driver could be
removed again.

Signed-off-by: Ahmad Fatoum <ahmad@a3f.at>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

bus: efi: add basic ACPI bus infrastructure ...

It makes sense to support some ACPI tables like the WDAT (Watchdog
Action Table) in barebox. To facilitate this add a ACPI bus and
the necessary bits to integrate ACPI tables into the barebox
device/driver model as devices. These devices are identified by the
four byte signature, which drivers can then match against and the
system description table (SDT) of the device is then available as a
device memory resource.

Even without drivers, with this patch, devinfo and md can now be used
to view the ACPI system description tables, which can be useful
during UEFI payload development.

Support for the ACPI Machine Language and ACPI 5.1 _DSD (Device
Specific Data) is not implemented.

Signed-off-by: Ahmad Fatoum <ahmad@a3f.at>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

serial: add support for PCI NS16550 UARTs ...

This ports over the Linux v5.4 8250_pci driver. Unlike Linux, the
barebox ns16550 implementation provides less hooks for covering all the
quirky behavior of all the different 8250-"compatible" hardware blocks.

The barebox driver matches against all serial port the Linux driver does,
but for hardware with quirks that can't be expressed in the current barebox
n16550 API, the probe fails with -ENOSYS and an appropriate message
telling that the quirk handling is missing. This should make future
extension of the driver straight-forward.

Unlike the kernel variant of the driver, most ->exit quirk callbacks can
be dropped in barebox as they often disable IRQs, which we do in the
->init anyway.

This was tested with the EFI PCI driver on qemu with the options:

  -device pci-serial,chardev=cdev0 -chardev file,id=cdev0,path=COM

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

Release v2019.12.0 ...

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

Revert "mci: core: make mci_register -EPROBE_DEFER sensitive" ...

The patch does the right thing: When a regulator can't be found and
-EPROBE_DEFER is returned we forward it. However, we have several boards
in the tree which do not have any driver for the regulator, but still
work because the regulator is enabled by default or by board code.

Revert this patch until a solution for these board is found.

This reverts commit 2e16c52.

clk: i.MX6ul: Disable GPMI_IO clk before reparenting ...

Before reparenting the ENFC_SEL mux we have to make sure the GPMI_IO clk
is disabled as recommended in the reference manual. Otherwise glitches
may occur on the GPMI clk which result in a later NAND failure:

MXS: Timeout resetting block via register 0x01806000
mxs_nand 1806000.gpmi-nand@1806000.of: probe failed: Connection timed out

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

mtd: spi-nor: fix wrong value for CR_QUAD_EN_SPAN ...

Durign a previous patch, 0x2 was mistakenly converted to BIT(2) whereas
it should be BIT(1).

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

efi: turn set of defines into enumerations ...

This allows us to use enum types instead of plain integers ttypes to
hold the values, which is done in a follow up commit.

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

PCI: copy over some Linux PCI helpers ...

Linux PCI drivers, like the incoming 8250_pci, make use of these
helpers. Port them over from Linux v5.4.

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

PCI: add driver_data member to struct pci_device_id ...

Linux drivers have a driver data member in the struct to associate
variant specific driver data with each device id. Do likewise in
barebox.

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

x86: efi: lds: don't discard any relocation sections ...

The incoming EFI PCI root bridge IO protocol driver will register
PCI fixups. Executing them will fail because the hook function's
relocation information is stripped from the final barebox.efi binary.

Instead of adding each section by name, just keep all .rela* sections
in the final binary.

This doesn't yet increase the size of the resulting barebox (yet).

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