Joe Hershberger [Fri, 30 Mar 2018 16:52:16 +0000 (11:52 -0500)]
net: phy: Don't limit phy addresses by default
Some boards expect to find more than one phy while other boards are old
and need to be limited to a specific phy address. Only limit the phy
address for boards that opt in.
Signed-off-by: Joe Hershberger <joe.hershberger@ni.com> Tested-by: Bin Meng <bmeng.cn@gmail.com> Acked-by: Neil Armstrong <narmstrong@baylibre.com> Reviewed-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Joe Hershberger [Fri, 13 Apr 2018 20:26:40 +0000 (15:26 -0500)]
xilinx: Only enable dist boot pxe when DHCP is enabled
Otherwise, we see this:
In file included from include/configs/zynq-common.h:183:0,
from include/config.h:5,
from include/common.h:21,
from env/common.c:11:
include/config_distro_bootcmd.h:319:2: error: expected ?}? before ?BOOT_TARGET_DEVICES_references_PXE_without_CONFIG_CMD_DHCP_or_PXE?
BOOT_TARGET_DEVICES_references_PXE_without_CONFIG_CMD_DHCP_or_PXE
^
include/config_distro_bootcmd.h:319:2: note: in definition of macro ?BOOTENV_DEV_NAME_PXE?
Signed-off-by: Joe Hershberger <joe.hershberger@ni.com>
Joe Hershberger [Fri, 13 Apr 2018 20:26:37 +0000 (15:26 -0500)]
net: Make the BOOTP options default
The BOOTP options used to be and should still be default for all boards
with CMD_NET enabled. One should not be forced to use DISTRO_DEFAULTS to
get them.
Signed-off-by: Joe Hershberger <joe.hershberger@ni.com> Reviewed-by: Duncan Hare <dh@synoia.com>
Joe Hershberger [Fri, 13 Apr 2018 20:26:35 +0000 (15:26 -0500)]
net: Add the BOOTP_DNS2 option to Kconfig
Commit 3b3ea2c56ec4bc5 ("Kconfig: cmd: Make networking command dependent on NET")
removed the help documentation from the README but didn't add it back to Kconfig.
Signed-off-by: Joe Hershberger <joe.hershberger@ni.com> Reviewed-by: Chris Packham <judge.packham@gmail.com> Reviewed-by: Duncan Hare <dh@synoia.com>
Joe Hershberger [Fri, 13 Apr 2018 20:26:30 +0000 (15:26 -0500)]
net: Make CMD_NET a menuconfig
Previously, CMD_NET was an alias for 2 commands (bootp and tftpboot) and
they we not able to be disabled. Separate out those 2 commands and move
CMD_NET up to the menu level, which more accurately represents the code.
Signed-off-by: Joe Hershberger <joe.hershberger@ni.com> Reviewed-by: Chris Packham <judge.packham@gmail.com> Reviewed-by: Duncan Hare <dh@synoia.com>
Ye Li [Wed, 28 Mar 2018 12:54:16 +0000 (20:54 +0800)]
net: fec: Fix issue in DM probe timeout
Since the probe function has changed to reset FEC controller prior than
setup PHY. If reset FEC controller timeout, the priv->phydev is not
initialized, so can't free it.
Signed-off-by: Ye Li <ye.li@nxp.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Peng Fan [Wed, 28 Mar 2018 12:54:14 +0000 (20:54 +0800)]
net: fec: sharing MDIO for two enet controllers
On i.MX6SX, 6UL and 7D, there are two enet controllers each has a
MDIO port. But Some boards share one MDIO port for the two enets. So
introduce a configuration CONFIG_FEC_MXC_MDIO_BASE to indicate
the MDIO port for sharing.
In Kconfig, user needs enable CONFIG_FEC_MXC_SHARE_MDIO first to enter
the CONFIG_FEC_MXC_MDIO_BASE.
To i.MX28, adapt to use the new config
Signed-off-by: Peng Fan <peng.fan@nxp.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com> Cc: Fabio Estevam <fabio.estevam@nxp.com>
Peng Fan [Wed, 28 Mar 2018 12:54:13 +0000 (20:54 +0800)]
net: fec: set dev->seq to priv->dev_id
To platforms has two enet interface, using dev->seq could
avoid conflict.
i.MX6UL/ULL evk board net get the wrong MAC address from fuse,
eth1 get MAC0 address, eth0 get MAC1 address from fuse. Set the
priv->dev_id to device->seq as the real net interface alias id then
.fec_get_hwaddr() read the related MAC address from fuse.
Signed-off-by: Peng Fan <peng.fan@nxp.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Ye Li [Wed, 28 Mar 2018 12:54:11 +0000 (20:54 +0800)]
net: fec_mxc: Fix DM driver issue in recv
When using ethernet DM driver, the recv interface has a
change with non-DM interface, that driver needs to set
the packet pointer and provide it to upper layer to process.
In fec driver, the fecmxc_recv functions does not handle the
packet pointer parameter. This may cause crash in upper layer
processing because the packet pointer is not set.
This patch allocates a buffer for the packet pointer and free it
through free_pkt interface.
Signed-off-by: Ye Li <ye.li@nxp.com> Reviewed-by: Peng Fan <peng.fan@nxp.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Matt Pelland [Tue, 27 Mar 2018 17:18:25 +0000 (13:18 -0400)]
net: mvneta: support setting hardware address
mvneta already supports setting the MAC address but this was only done
internally when some other part of U-Boot tries to actually use the
interface. This commit exposes this functionality to the ethernet core
code so that the MAC addresses of all interfaces are configured
correctly even if they are not used before loading Linux.
Signed-off-by: Matt Pelland <mpelland@starry.com> Reviewed-by: Stefan Roese <sr@denx.de> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Marek Vasut [Thu, 15 Feb 2018 12:19:33 +0000 (13:19 +0100)]
ARM: rmobile: Fix the memory map on Gen3
Fix up the memory map on Gen3 to match datasheet properly.
This simplifies the memory map setup as well, since we do
no longer need this massive complexity.
Signed-off-by: Marek Vasut <marek.vasut+renesas@gmail.com> Cc: Nobuhiro Iwamatsu <iwamatsu@nigauri.org>
Marek Vasut [Tue, 10 Apr 2018 14:43:47 +0000 (16:43 +0200)]
spi: sh_qspi: Make use of the 32byte FIFO
The QSPI controller on RCar Gen2 has 32byte FIFO. Instead of doing
the SPI transmission 1 byte at time, if there is a 32byte chunk of
data to be transferred, fill the FIFO completely and then transfer
the data to/from the FIFO. This increases the SPI NOR access speed
significantly.
Signed-off-by: Marek Vasut <marek.vasut+renesas@gmail.com> Cc: Nobuhiro Iwamatsu <nobuhiro.iwamatsu.yj@renesas.com>
Marek Vasut [Tue, 10 Apr 2018 14:58:46 +0000 (16:58 +0200)]
spi: sh_qspi: Replace ad hoc waiting with wait_for_bit
Replace the ad-hoc endless loops with wait_for_bit() with
reasonable timeout. Note that the loops had internal 10uS
delays, although there is no reason for those on this HW,
so they are dropped.
Signed-off-by: Marek Vasut <marek.vasut+renesas@gmail.com> Cc: Nobuhiro Iwamatsu <nobuhiro.iwamatsu.yj@renesas.com>
Marek Vasut [Wed, 29 Nov 2017 05:29:46 +0000 (06:29 +0100)]
mtd: spi: Add Renesas RPC SPI-flash driver
Add driver for the RPC block in SPI-flash mode. This driver allows
access to a SPI NOR flash attached to the RPC block and does not
support RPC in Hyperflash mode. Note that this block is extremely
selective when communicating with the SPI NOR.
Signed-off-by: Marek Vasut <marek.vasut+renesas@gmail.com> Cc: Nobuhiro Iwamatsu <iwamatsu@nigauri.org>
Marek Vasut [Sat, 19 Aug 2017 21:24:08 +0000 (23:24 +0200)]
mtd: rpc: Add Renesas RPC Hyperflash driver
Add driver for the RPC block in Hyperflash mode. This driver allows
access to a CFI Hyperflash attached to the RPC block and does not
support RPC in SPI mode.
Signed-off-by: Marek Vasut <marek.vasut+renesas@gmail.com> Cc: Nobuhiro Iwamatsu <iwamatsu@nigauri.org>
Marek Vasut [Mon, 9 Apr 2018 18:47:31 +0000 (20:47 +0200)]
mmc: matsushita-common: Wait for command completion
Make sure to wait for the command to complete altogether, including
the trailing 8 clock cycles. This prevents the driver for accidentally
writing the CMD register too fast before the previous command fully
completed.
Marek Vasut [Mon, 9 Apr 2018 18:47:31 +0000 (20:47 +0200)]
mmc: matsushita-common: Special case only select registers in 16bit
There are only a few registerse used in the 16bit mode which are
32bit internally. Special-case only those in the IO accessors and
always write both halves. Any other register access is protected
from accidentally overwriting neighboring register.
Marek Vasut [Sat, 23 Sep 2017 11:01:20 +0000 (13:01 +0200)]
mmc: matsushita-common: Always check controller version
Handle the controller version even if quirks are set. The controller in
Renesas Gen3 SoCs does provide the version register, which indicates a
controller v10 and the controller does support internal DMA and /1024
divider.
Marek Vasut [Tue, 2 Jan 2018 18:40:49 +0000 (19:40 +0100)]
mmc: matsushita-common: Handle DMA completion flag differences
The DMA READ completion flag position differs on Socionext and Renesas
SoCs. It is bit 20 on Socionext SoCs and using bit 17 is a hardware bug
and forbidden. It is bit 17 on Renesas SoCs and bit 20 does not work on
them.
Marek Vasut [Tue, 26 Sep 2017 18:34:35 +0000 (20:34 +0200)]
mmc: matsushita-common: Handle Renesas div-by-1
On the Renesas version of the IP, the /1 divider is realized by
setting the clock register [7:0] to 0xff instead of setting bit
10 of the register. Check the quirk and handle accordingly.
Marek Vasut [Sun, 8 Apr 2018 16:49:52 +0000 (18:49 +0200)]
mmc: matsushita-common: Add Renesas RCar quirks
Add a quirk to identify that the controller is Renesas RCar variant
of the Matsushita SD IP and another quirk indicating it can support
Renesas RCar HS200/HS400/SDR104 modes.
Marek Vasut [Sun, 8 Apr 2018 16:14:22 +0000 (18:14 +0200)]
mmc: renesas-sdhi: Handle 16bit IP
The Renesas RCar Gen2 chips have a mix of 32bit and 16bit variants
of the IP. There is no DT property which allows discerning those,
so what Linux does is it checks the size of the register area and
if it is 0x100, the IP is 16bit, otherwise the IP is 32bit. Handle
the distinction the same way.
Marek Vasut [Sun, 8 Apr 2018 15:45:23 +0000 (17:45 +0200)]
mmc: uniphier: Allow passing quirks to the probe function
Certain instances of the SD IP require more elaborate digging
in the DT to figure out which variant of the SD IP is in use.
Allow explicit passing of the quirks into the probe function.
Marek Vasut [Sun, 8 Apr 2018 15:41:14 +0000 (17:41 +0200)]
mmc: uniphier: Add support for 16bit variant
Add support for 16bit mutation of the Matsushita SD IP. Since some
registers are internally 32bit, the matsu_sd_{read,write}l() has
to special-case this 16bit variant a bit.
Marek Vasut [Sun, 8 Apr 2018 15:25:49 +0000 (17:25 +0200)]
mmc: uniphier: Drop useless check
Drop useless check in matsu_sd_{read,write}q(), this is only ever
called to read the data from FIFO and only when 64bit variant of
the block is used anyway.
Marek Vasut [Sun, 8 Apr 2018 15:14:42 +0000 (17:14 +0200)]
mmc: uniphier: Factor out FIFO accessors
Add macros to generate the FIFO accessors, since the code is almost
the same with only minor differences. This is done in preparation
for adding 16bit variant of the IP.
Marek Vasut [Sun, 8 Apr 2018 13:22:58 +0000 (15:22 +0200)]
mmc: uniphier: Split out SoC specific bits from the driver
Factor out common code from the uniphier SD driver, change the prefix
of the functions from uniphier_sd_ to matsu_sd_ and create separate
renesas-sdhi.c driver. Thus far, all the code is still compiled when
CONFIG_UNIPHIER_MMC is selected and there is no functional change.
This patch is a preparation for further split of the SoC specific
parts of the Matsushita SD driver, used both on Uniphier and R-Car.
Marek Vasut [Sat, 7 Apr 2018 14:16:30 +0000 (16:16 +0200)]
ARM: rmobile: Add JTAG recovery support for M2 Porter
Add JTAG recovery support into the M2 Porter TPL. This allows the
TPL to be loaded over JTAG, initialize the system, wait for the
JTAG debugger to load U-Boot image into RAM and then resume and
start U-Boot from RAM.
The procedure is as follows:
1) Load u-boot-tpl.bin to 0xe6300000
2) Write magic number 0x1337c0de to 0xe6300020
TPL checks for this particular magic and starts JTAG recovery
if this number is present. This is not present by default.
3) Start U-Boot TPL from 0xe6300000
4) Wait for a message from TPL on UART indicating JTAG boot:
"JTAG boot detected!"
5) Halt the system in JTAG debugger
6) Load U-Boot image (u-boot.img) to 0x4fffffc0
7) Write magic number 0xb33fc0de to 0xe6300024
TPL checks for this particular magic to verify that the U-Boot
image was loaded into DRAM by the JTAG debugger.
8) Resume the system in JTAG debugger
Signed-off-by: Marek Vasut <marek.vasut+renesas@gmail.com> Cc: Nobuhiro Iwamatsu <iwamatsu@nigauri.org>
Marek Vasut [Tue, 3 Apr 2018 10:52:48 +0000 (12:52 +0200)]
ARM: rmobile: Add TPL support on R8A7791 M2 Porter
Add and enable TPL on M2 Porter. The TPL must fit into 16 kiB due
to the Gen2 BootROM restriction. The TPL is running from MERAM and
is capable of performing the initial initialization of PFC, Clock,
GPIO, LBSC, DBSC and QSPI NOR. DBSC is responsible for bringing up
the DDR DRAM access. The TPL is capable of loading the next stage,
U-Boot, from either SPI NOR or UART as a fallback. If either does
provide a valid U-Boot uImage, the system stops, which allows the
operator to load U-Boot ie. via JTAG and start it manually.
Signed-off-by: Marek Vasut <marek.vasut+renesas@gmail.com> Cc: Nobuhiro Iwamatsu <iwamatsu@nigauri.org>
Marek Vasut [Sat, 7 Apr 2018 15:03:28 +0000 (17:03 +0200)]
spl: ram: Convert to CONFIG_IS_ENABLED
This patch is a preparation for adding TPL support for RAM loading.
CONFIG_IS_ENABLED allows for proper handling of the U-Boot/SPL/TPL
differences in config symbol names.
Signed-off-by: Marek Vasut <marek.vasut+renesas@gmail.com> Cc: Tom Rini <trini@konsulko.com>
Marek Vasut [Sat, 7 Apr 2018 14:06:36 +0000 (16:06 +0200)]
ARM: Fix Makefile during SPL and TPL build
The tiny variants of memset and memcpy implementations can be
built for TPL as well, check whether a TPL build is in progress
and avoid including the default variants.
Signed-off-by: Marek Vasut <marek.vasut+renesas@gmail.com> Cc: Simon Glass <sjg@chromium.org> Cc: Tom Rini <trini@konsulko.com>
Marek Vasut [Sat, 7 Apr 2018 14:06:22 +0000 (16:06 +0200)]
serial: Fix Makefile during SPL and TPL build
This patch fixes a situation where CONFIG_DM_SERIAL is enabled for
regular U-Boot and SPL, but not for TPL. In that case, the build
will try to include serial-uclass into the TPL nonetheless, because
CONFIG_DM_SERIAL is set.
The solution is to check if the build is for SPL or TPL and in that
case, check if CONFIG_$(SPL_TPL_)DM_SERIAL is also set. Only in that
case, include serial-uclass.c . If the build is for regular U-Boot,
CONFIG_BUILD is not set, so only check if CONFIG_DM_SERIAL is set
and if so, include serial-uclass.c
Signed-off-by: Marek Vasut <marek.vasut+renesas@gmail.com> Cc: Simon Glass <sjg@chromium.org> Cc: Tom Rini <trini@konsulko.com>
The Amlogic Meson GXL and GXM (simple variant) embeds up to 3 USB2 PHYs
and an USB3 PHY. This patch adds drivers for these for the standard generic
PHY interface and supports the power-on/off calls and set the Host mode by
default.
They are based on the excellent work from Martin Blumenstingl merged in linux.
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Neil Armstrong [Wed, 11 Apr 2018 15:08:01 +0000 (17:08 +0200)]
usb: host: dwc3: Add support for multiple PHYs
DWC3 Ips can have more than 1 PHY for USB2 and 1 PHY for USB3, add support
for a generic number of PHYs and adapt the code to handle a generic
number of PHYs.
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
The commit implement a gpio-based software deblocking. The code
extract I2C pins description from device tree, switch pins to GPIO
mode, toggle SCL until slave release SDA, send I2C stop and switch
I2C pins back to I2C mode.
Signed-off-by: Alexander Kochetkov <al.kochet@gmail.com>
The mutex lock and unlock functions are stubbed out and mutex_is_locked
was 0. This caused asserts to fail in ubifs code when checking that the
mutex was locked. For example,
UBIFS assert failed in ubifs_change_lp at 540
UBIFS assert failed in ubifs_release_lprops at 278
Assume that the "mutex" is locked since that is the normal case when it
is checked in the ubifs code.
Neil Armstrong [Tue, 3 Apr 2018 09:40:51 +0000 (11:40 +0200)]
reset: add sandbox test for bulk API
This patch adds the bulk reset API tests for the sandbox test suite.
Unlike the main test, it also check the "other" reset signal using the bulk API
and checks if the resets are correctly asserted/deasserted.
To allow the bulk API to work, and avoid changing the DT, the number of resets
of the sandbox reset controller has been bumped to 101 for the "other" reset
line to be valid.
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> Reviewed-by: Simon Glass <sjg@chromium.org>
Guillaume GARDET [Fri, 30 Mar 2018 08:28:19 +0000 (10:28 +0200)]
mkimage: do not fail if there is no print_header function
Commit 253c60a breaks the exit value of 'mkimage -T rkimage'
and print the following error:
mkimage: Can't print header for Rockchip Boot Image support: Success
It is not a failure to not print headers, so just display the warning message,
and finish the function properly.
Signed-off-by: Guillaume GARDET <guillaume.gardet@free.fr> Cc: Philipp Tomsich <philipp.tomsich@theobroma-systems.com> Cc: Simon Glass <sjg@chromium.org> Cc: Tom Rini <trini@konsulko.com> Reviewed-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Neil Armstrong [Thu, 29 Mar 2018 12:56:02 +0000 (14:56 +0200)]
serial: meson: Update compatible with new Linux bindings
The Amlogic Meson SoCs serial bindings were not written when serial
support was pushed into Linux and U-Boot.
A clean bindings document has been merged into Linux tree to correctly
handle the multiple clocks feeding the serial peripheral.
This update the U-Boot serial_meson driver with the new compatible
string for Amlogic Meson GX Socs.
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> Reviewed-by: Simon Glass <sjg@chromium.org>
Tom Rini [Mon, 9 Apr 2018 15:05:44 +0000 (11:05 -0400)]
Merge tag 'signed-efi-next' of git://github.com/agraf/u-boot
Patch queue for efi - 2018-04-09
Highlights this time around:
- Lots of minor spec compliance fixes
- Support full range of GOP BLT commands
- More fine grained error checking
- Network fixes (init, DP)
- Lots of other bug fixes...
projects / u-boot / log