There is a missing in previous
commit 951344b7
(nds32: Convert Makefiles to use COBJS-y style)
will cause compile error.

Signed-off-by: Kuan-Yu Kuo <ken.kuoky@gmail.com>
Cc: Macpaul Lin <macpaul@gmail.com>
Cc: Andes <uboot@andestech.com>
Signed-off-by: Andes <uboot@andestech.com>

U-Boot does not compile for the adp-ag101 boards since
commit a8f9cd18
(net: update FTGMAC100 for MMU/D-cache support)

The driver assumes that the DMA allocation API are provided by all
architectures. This is not the case for nds32 and it causes a
build error. This patch adds DMA allocation API to avoid the errors.

Signed-off-by: Kuan-Yu Kuo <ken.kuoky@gmail.com>
Cc: Macpaul Lin <macpaul@gmail.com>
Cc: Andes <uboot@andestech.com>
Signed-off-by: Andes <uboot@andestech.com>

Signed-off-by: Andes <uboot@andestech.com>
Cc: Macpaul Lin <macpaul@gmail.com>
Cc: Kuan-Yu Kuo <ken.kuoky@gmail.com>

Marek Vasut authored 11 years ago and Heiko Schocher committed 11 years ago

In case only the CONFIG_SYS_I2C_SPEED is set in configuration file,
the CONFIG_SYS_I2C_SOFT_SPEED is defined as CONFIG_SYS_I2C_SPEED.
The CONFIG_SYS_I2C_SOFT_SPEED is then used throughout the driver.

Unfortunatelly, due to a typo in the driver, instead of defining
CONFIG_SYS_I2C_SOFT_SPEED, an CONFIG_SYS_SOFT_I2C_SPEED was defined
and therefore the driver failed to compile. The same applies for
CONFIG_SYS_I2C_SOFT_SLAVE , where the swap happens as well.

This patch fixes the issue.

Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Heiko Schocher <hs@denx.de>

This brings U-Boot's cpp invocation into line with the way the Linux
kernel invokes cpp on device trees. Consistency will be useful to ensure
*.dts is portable between the two.

-undef also has the added advantage of not defining "linux", so DT
property names such as "linux,keymap" don't get mangled.

Signed-off-by: Stephen Warren <swarren@nvidia.com>
Acked-by: Simon Glass <sjg@chromium.org>

Recent dtc supports #line directives in the input source code, and even
uses them to generate useful line numbers in any messages it emits. Stop
passing -P to cpp, since there's no need any more.

Signed-off-by: Stephen Warren <swarren@nvidia.com>
Acked-by: Simon Glass <sjg@chromium.org>

Now that nothing uses CONFIG_ARCH_DEVICE_TREE, stop defining it.

Signed-off-by: Stephen Warren <swarren@nvidia.com>
Acked-by: Simon Glass <sjg@chromium.org>

Now that nothing uses the defines ARCH_CPU_DTS, BOARD_DTS, stop defining
them.

Signed-off-by: Stephen Warren <swarren@nvidia.com>
Acked-by: Simon Glass <sjg@chromium.org>

Now that we assume dtc supports the -i option, we don't need to use
ARCH_CPU_DTS in *.dts{,i}; we simply specify the include filename
directly, and dtc will find it.

Signed-off-by: Stephen Warren <swarren@nvidia.com>
Acked-by: Simon Glass <sjg@chromium.org>

*.dts may use #include (via cpp) or /include/ (via dtc; assuming a newer
dtc). The choice is up to the creator of the DT. Create a common variable
DTC_INCDIRS that lists the paths searched by include statements, and
update cpp and dtc invocation to use them.

For cpp, also specify -nostdinc to ensure the same set of paths is
available to both type of include statement.

For dtc, create a new DTC_FLAGS variable to hold all the flags passed to
dtc.

Signed-off-by: Stephen Warren <swarren@nvidia.com>
Acked-by: Simon Glass <sjg@chromium.org>

The invocation of dtc is significantly more complex that it could be,
in order to work around an issue on old versions of dtc, which print
a message to stdout every time they run.

Remove this workaround, on the assumption that people have or will
upgrade to a newer version of dtc. This simplifies the build rule
significantly.

Related, split the invocation of cpp and dtc into separate commands
rather than a pipeline, so that if either fail, it is detected. This has
the nice benefit of saving off the result of the pre-processing step,
allowing it to be easily inspected.

Assuming a new enough dtc (which an earlier patch enforces), dtc will
parse #line directives in its input file, and generate correct file and
line numbers in error messages, even though cpp is unconditionally
applied to its input file.

Signed-off-by: Stephen Warren <swarren@nvidia.com>
Acked-by: Simon Glass <sjg@chromium.org>

Aside from microblaze, all other SoCs/boards/vendors store their DT files
in board/$vendor/dts/$soc-$board.dts. Move microblaze-generic.dts to this
location for consistency.

Signed-off-by: Stephen Warren <swarren@nvidia.com>
Acked-by: Simon Glass <sjg@chromium.org>
Acked-by: Michal Simek <monstr@monstr.eu>

Subsequent patches assume that dtc supports various recent features.
These are available in dtc 1.4.0. Validate that dtc is at least that
version.

Signed-off-by: Stephen Warren <swarren@nvidia.com>
Acked-by: Simon Glass <sjg@chromium.org>

Use DIV_ROUND_UP to simplify the code.

Signed-off-by: Axel Lin <axel.lin@ingics.com>
Signed-off-by: Scott Jiang <scott.jiang.linux@gmail.com>
Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>

Current code uses gd->baudrate before setting its value.
Besides, I got below build warning which is introduced by
commit ddb5c5be "blackfin: add baudrate to bdinfo".

board.c:235:3: warning: passing argument 1 of 'simple_strtoul' makes pointer from integer without a cast [enabled by default]
include/vsprintf.h:27:7: note: expected 'const char *' but argument is of type 'unsigned int'

This patch ensures we get the baudrate setting before using it.

Signed-off-by: Axel Lin <axel.lin@ingics.com>
Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>

The function return from P_FUNCT2MUX(per) takes 2 bits, however
for BF537_FAMILY with offset != 1 the function is 1 bit.

Also has small refactor for better readability.
In portmux_setup(), it looks odd having "muxreg &= ~(3 << 1);"
while in current code we do muxreg |= (function << offset);.

Signed-off-by: Axel Lin <axel.lin@ingics.com>

In special_gpio_free(), call unreserve() rather than reserve() to release gpio.

Signed-off-by: Axel Lin <axel.lin@ingics.com>
Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>

In special_gpio_free(), call unreserve() rather than reserve() to release gpio.

Signed-off-by: Axel Lin <axel.lin@ingics.com>
Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>

This commit adds some prototypes into include/mtd/cfi_flash.h.
These functions are defined with a weak attribute in
drivers/mtd/cfi_flash.c.
This means they can be overrided by board-specific ones
if necessary.

When defining such functions under board/ directory or
somewhere, cfi_flash.h should be included.
This makes sure that board-specfic cfi functions
are defined in a correct prototype.

Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Signed-off-by: Stefan Roese <sr@denx.de>

usb_lowlevel_init() allocates a new periodic_list each time it is invoked,
without freeing the original list. Since it is initialized later on in the code,
just reuse the first-allocated list in future invocations of usb_lowlevel_init.

Cc: Marek Vasut <marex@denx.de>
Cc: Igor Grinberg <grinberg@compulab.co.il>
Signed-off-by: Nikita Kiryanov <nikita@compulab.co.il>

When power cycling the hub ports, a misbehaving port will prevent all ports
from being powered on because we quit at the first sign of trouble.

Skip problematic ports instead of failing the entire power on.

Cc: Marek Vasut <marex@denx.de>
Cc: Igor Grinberg <grinberg@compulab.co.il>
Signed-off-by: Nikita Kiryanov <nikita@compulab.co.il>

Fixes NFS root problems with Beagle (3530 ES1.0) when used with
external USB-ethernet adapter and "USB start" command used within
u-boot.

Soft resetting the UHH module causes instability issues on
all OMAPs so we just avoid it.

See OMAP36xx Errata
  i571: USB host EHCI may stall when entering smart-standby mode
  i660: USBHOST Configured In Smart-Idle Can Lead To a Deadlock

On OMAP4/5, soft-resetting the UHH module can put it into
Smart-Idle mode and lead to a deadlock.

On OMAP3 this doesn't seem to be the case but still instabilities
are observed on beagle (3530 ES1.0) if soft-reset is used.
 e.g. NFS root failures with Linux kernel.

Signed-off-by: Roger Quadros <rogerq@ti.com>

This patch extends dfu code to support transmission with -R switch
specified at dfu-util.

When -R is specified, the extra USB_REQ_DFU_DETACH request is sent after
successful data transmission. Then dfu resources are released and reset
command is issued.

Signed-off-by: Lukasz Majewski <l.majewski@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>

The requests sent to the controller are not properly cache aligned
most of the time, thus implement a simple bounce buffer to avoid
problem with cache.

Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Fabio Estevam <fabio.estevam@freescale.com>
Cc: Lei Wen <leiwen@marvell.com>
Cc: Otavio Salvador <otavio@ossystems.com.br>
Cc: Stefano Babic <sbabic@denx.de>

Implement functions to flush/invalidate dcache over QH and qTDs
and make use of them where appropriate. Also use them to replace
the old incorrect cache management attempt. This is the first step
towards making this driver work with data cache enabled.

Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Fabio Estevam <fabio.estevam@freescale.com>
Cc: Lei Wen <leiwen@marvell.com>
Cc: Otavio Salvador <otavio@ossystems.com.br>
Cc: Stefano Babic <sbabic@denx.de>

The code for retrieving qTD item for particular endpoint is hard
to understand, moreover it's duplicated all over the driver. Move
the code into single nice and documented function.

Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Fabio Estevam <fabio.estevam@freescale.com>
Cc: Lei Wen <leiwen@marvell.com>
Cc: Otavio Salvador <otavio@ossystems.com.br>
Cc: Stefano Babic <sbabic@denx.de>

Allocate the qTD items all at once instead of allocating them
separately. Moreover, make sure each qTD is properly aligned
to 32-bytes boundary and that cache can be safely flushed over
each qTD touple.

Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Fabio Estevam <fabio.estevam@freescale.com>
Cc: Lei Wen <leiwen@marvell.com>
Cc: Otavio Salvador <otavio@ossystems.com.br>
Cc: Stefano Babic <sbabic@denx.de>

The code for retrieving QH for particular endpoint is hard to understand,
moreover it's duplicated all over the driver. Move the code into single
nice and documented function.

Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Fabio Estevam <fabio.estevam@freescale.com>
Cc: Lei Wen <leiwen@marvell.com>
Cc: Otavio Salvador <otavio@ossystems.com.br>
Cc: Stefano Babic <sbabic@denx.de>

Check the length of system cacheline at compile-time and fail
if the system uses too long cachelines.

Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Fabio Estevam <fabio.estevam@freescale.com>
Cc: Lei Wen <leiwen@marvell.com>
Cc: Otavio Salvador <otavio@ossystems.com.br>
Cc: Stefano Babic <sbabic@denx.de>

The endpoint QH list has to be aligned to 10-bit boundary. We also have
to make sure the list is aligned on a cacheline boundary. Make sure it
is. Furthermore, check if the memory allocation for the QH list didn't
fail. Moveover, improve the comment about the QH list structure.

Finally, the qTD item list has to be aligned only to 5-bit boundary, not
10-bit as it is now, fix this as well.

Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Fabio Estevam <fabio.estevam@freescale.com>
Cc: Lei Wen <leiwen@marvell.com>
Cc: Otavio Salvador <otavio@ossystems.com.br>
Cc: Stefano Babic <sbabic@denx.de>

Both the endpoint queue head and the endpoint item list is a controller
specific thing. Move them both into controller private data.

Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Fabio Estevam <fabio.estevam@freescale.com>
Cc: Lei Wen <leiwen@marvell.com>
Cc: Otavio Salvador <otavio@ossystems.com.br>
Cc: Stefano Babic <sbabic@denx.de>

There is no need to init this field at runtime, so init it statically.

Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Fabio Estevam <fabio.estevam@freescale.com>
Cc: Lei Wen <leiwen@marvell.com>
Cc: Otavio Salvador <otavio@ossystems.com.br>
Cc: Stefano Babic <sbabic@denx.de>

The QH_MAXNUM is used in absolutelly incorrect manner and is not
even needed. Remove it and correctly replace it's occurance with
2 * NUM_ENDPOINTS .

Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Fabio Estevam <fabio.estevam@freescale.com>
Cc: Lei Wen <leiwen@marvell.com>
Cc: Otavio Salvador <otavio@ossystems.com.br>
Cc: Stefano Babic <sbabic@denx.de>

Clean up the code that checks the validity of a USB gadget driver
in usb_gadget_register_driver(). Moreover, limit the speed of the
driver to either FULL or HIGH, this is more precise and once we
have xHCI support, also more correct.

Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Fabio Estevam <fabio.estevam@freescale.com>
Cc: Lei Wen <leiwen@marvell.com>
Cc: Otavio Salvador <otavio@ossystems.com.br>
Cc: Stefano Babic <sbabic@denx.de>

The usb_lowlevel_init() call already fills and passes back struct
ehci_ctrl , which readily contains correctly determined address of
the port register block address computed from values from controller
configuration registers. Leverage this and make use of this value
as this makes the code mode universal, but also gets us rid of the
CONFIG_USB_REG_BASE configuration option.

Moreover, this patch cleans up the usb_gadget_register_driver() call
a little by correcting the error handling. Note the usb_lowlevel_init()
and mvudc_probe() are now called in reversed order, but this has no
impact on the code.

Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Fabio Estevam <fabio.estevam@freescale.com>
Cc: Lei Wen <leiwen@marvell.com>
Cc: Otavio Salvador <otavio@ossystems.com.br>
Cc: Stefano Babic <sbabic@denx.de>

Move the struct ehci_ctrl defition from ehci-hcd.c into ehci.h
so it can be re-used by drivers. In particular, the mv_udc driver
can benefit from this move.

Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Fabio Estevam <fabio.estevam@freescale.com>
Cc: Lei Wen <leiwen@marvell.com>
Cc: Otavio Salvador <otavio@ossystems.com.br>
Cc: Stefano Babic <sbabic@denx.de>

Move the constant values that are programmed into mv_ep.ep into
separate static const structure so they can be memcpy()'d when
the initialization happens.

Moveover, we only every init NUM_ENDPOINTS, not 2 * NUM_ENDPOINTS,
so fix this bug as well.

Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Fabio Estevam <fabio.estevam@freescale.com>
Cc: Lei Wen <leiwen@marvell.com>
Cc: Otavio Salvador <otavio@ossystems.com.br>
Cc: Stefano Babic <sbabic@denx.de>

The endpoints are operated on a per-controller basis, move the
endpoint array into controller's private data. Also shuffle the
struct mv_ep structure definition just above the definition of
the struct mv_drv so they're well grouped together.

Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Fabio Estevam <fabio.estevam@freescale.com>
Cc: Lei Wen <leiwen@marvell.com>
Cc: Otavio Salvador <otavio@ossystems.com.br>
Cc: Stefano Babic <sbabic@denx.de>