Merge branch 'for-next/cpsw'

Merge branch 'for-next/compile-test'

Release v2020.05.0 ...

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

drivers: add COMPILE_TEST prompts for some off-by-default options ...

Some Kconfig options we have are promptless and off-by-default and
instead can only be enabled by being selected from platform options.

For some of those that aren't compile testable, add a new
COMPILE_TEST-only prompt.

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

drivers: add COMPILE_TEST prompts to some on-by-default drivers ...

Many clock source and GPIO controller drivers don't have a prompt, but
are instead default y if their respective platform is enabled. Maintain
this behavior, but add a prompt for when COMPILE_TEST is enabled, so
they can be included in test builds.

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

drivers: mark first batch of compilable drivers for COMPILE_TEST ...

All of these drivers have a runtime dependency on SoC peripherals, but
can nevertheless be compile-tested. Add COMPILE_TEST as an alternate
dependency.

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

common: introduce COMPILE_TEST option for build-testing ...

Compile-time analysis may help us catch latent bugs in barebox. For this
to be most effective, we need an easy way to compile as much of barebox
as possible. Giving all driver options prompts would do this, but at the
cost of making user experience worse, by asking them about drivers for HW
that's clearly not relevant to the platform they selected.

Do as Linux does and provide a default-off COMPILE_TEST option, which we
can use to make extra drivers selectable.

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

drivers: explicitly select dependency STMP_DEVICE for i.MX23/28 drivers ...

Some of the i.MXs drivers can be build tested even if building for other
platforms. Support adding a future COMPILE_TEST dependency to some of
these drivers by explicitly having their Kconfig symbol select STMP_DEVICE
if the driver calls stmp_reset_block.

This has not been necessary for the normal build, because the ARCH
symbols for the IMX23 and IMX28 already selected STMP_DEVICE.

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

ARM: at91: Move mci platform data to include/ ...

To be able to add the Atmel mci driver to COMPILE_TEST move the
definition of struct atmel_mci_platform_data to include/ where it
can be reached from foreign SoCs as well.

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

fs: Fix inode size setting in ftruncate ...

When changing a file size the size of the corresponding inode has to be
changed as well.

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

net: cpsw: make cpsw_send directional ...

For dual_emac mode, the driver needs to send on the individual
ports. We set the directional bits on each packet so the controller
knows where to send its.

From TI am335x Reference Manual:

14.3.2.10.2 Dual Mac Mode

- Packets from the host (port 0) to ports 1 and 2 should be directed.

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

net: cpsw: add vlan handling for dual_emac mode ...

From TI am335x Reference Manual:

14.3.2.10.2 Dual Mac Mode

When operating in dual mac mode the intention is to transfer packets
between ports 0 and 1 and ports 0 and 2, but not between ports 1 and 2.

Each CPGMAC_SL appears as a single MAC with no bridging between MAC’s.
Each CPGMAC_SL has at least one unique (not the same) mac address.

Dual mac mode is configured as described below:

- Set the ale_vlan_aware bit in the ALE_Control register. This bit
configures the ALE to process in vlan aware mode.

With X can be Port 1 or Port 2:

- Configure the Port X to Port 0 VLAN.
Add a VLAN Table Entry with ports 0 and X as members (clear the flood
masks).

Add a VLAN/Unicast Address Table Entry with the PortX/0 VLAN and
a port number of 0.

- Select the dual mac mode on the port 0 FIFO by setting
tx_in_sel[1:0] = 01 in P0_Tx_In_Ctl.

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

net: cpsw: cpsw_process should only handle rx channels for its own port ...

The driver is currently processing the whole channel list, so it is
possible that it handles a channel of the wrong interface. We limit the
rx_chan handling for its desired port.

The cpsw_send is removing finished tx_chan transfers on each send, so this
has no directional limitation.

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

net: cpsw: move generic setup code to probe ...

All generic cpsw code can move to probe, so the controller is only reset
and configured once.

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

net: cpsw: fix cpsw_slave_regs register offsets ...

The cpsw_slave_regs struct also start with max_blks register, so here we
also have to skip the first two bytes like for .host_port_reg_ofs

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

kbuild: add $(always-y) to 'targets' ...

I missed to sync this line in commit 421108c51da4 ("kbuild: rename
hostprogs-y/always to hostprogs/always-y").

Since then, the files in always(-y) are needlessly rebuilt.

Fixes: 421108c51da4 ("kbuild: rename hostprogs-y/always to hostprogs/always-y")
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

video: ssd1307fb: fix null pointer dereference on error ...

On error, a goto may jump over the initialization of vmem and free the
uninitialized value during clean up. Fix this.

Found by compiling the code with clang.

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

ARM: Phytec phyFLEX i.MX6: delete wrong memory node ...

The phyFLEX i.MX6 has two memory nodes in the device tree. /memory comes
from the barebox dtsi file and contains dummy values. This is
unnecessary, barebox will create a memory node when booting the kernel.
The other one comes from the Kernel dtsi and causes more harm. The
memory in the node has a size of 2GiB. Not all boards have so much
memory, since 8226f7f909 ("of: base: parse all available memory nodes")
this node is no longer ignored and now all boards are assumed to have
2GiB of memory. Remove the node using /delete-node/

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Reviewed-by: Stefan Riedmüller <s.riedmueller@phytec.de>

ppc: Test for compiler options ...

The -mno-spe and -mspe=no options are not available on all compilers.
Call cc-option to set them only when they exist.

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

dtc: remove redundant YYLOC global declaration ...

Same as the upstream fix for building dtc with gcc 10.

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

Documentation: kvx: add first documentation ...

Add documentation on kvx architecture and kalray k200 board.

Signed-off-by: Clement Leger <cleger@kalray.eu>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

ARM: dts. i.MX7: add stopgap solution for barebox UART clock breakage ...

Commit 8b2104d740 ("driver: Call of_clk_set_defaults for each probed device")
made barebox act on the assigned-clock-parents property everywhere, not
only for clock provider nodes. This breaks at least the serial console on
the i.MX7 SabreSD board; only garbage is output if the patch is not
reverted. As stopgap measure, override these properties for all enabled
uart nodes in the barebox i.MX7 device tree. This reverts the behavior
to what it was like before the offending commit.

This is of course no real solution, as obviously Linux was fine dealing
with these properties before. But this hack should at least fix the
serial console for the boards that were broken in v2020.03.0.

When a proper fix follows, this patch can be reverted.

Cc: Andrey Smirnov <andrew.smirnov@gmail.com>
Cc: Lucas Stach <l.stach@pengutronix.de>
Cc: Juergen Borleis <jbe@pengutronix.de>
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

Merge branch 'for-next/zynq'

Merge branch 'for-next/zii'

Merge branch 'for-next/usb'

Merge branch 'for-next/spi'

Merge branch 'for-next/pwm'

Merge branch 'for-next/nfs'

Merge branch 'for-next/net'

Merge branch 'for-next/misc'