ti_armv7_common.h: Add CONFIG_CMD_EXT4_WRITE

Given that with config_distro_defaults.h we always have ext4 read
support, add in write support.

Signed-off-by: Tom Rini <[email protected]>

am335x_evm.h: unsed CONFIG_BOOTDELAY

Now that ti_armv7_common.h uses config_distro_defaults.h we don't need
to include it again and then undef CONFIG_BOOTDELAY

Signed-off-by: Tom Rini <[email protected]>

ext4_common.c: Clean up failure cases in alloc_triple_indirect_block

As noted by Coverity, when we have an error in
alloc_triple_indirect_block we will leak ti_pbuff_start_addr as it's not
being freed.  Further inspection here shows that we could also leak
ti_cbuff_start_addr in one corner case so free that as well.

Reported-by: Coverity (CID 131205, 131206)
Signed-off-by: Tom Rini <[email protected]>

ums: support multiple LUNs at once

Extend the ums command to accept a list of block devices. Each of these
will be exported as a separate LUN. An example use-case would be:

ums 0 mmc 0,0.1,0.2

... which would export LUNs for eMMC 0's user data, boot0, and boot1 HW
partitions. This is useful since it allows the host access to everything
on the eMMC without having to somehow stop the ums command from executing
and restart it with different parameters.

Signed-off-by: Stephen Warren <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

mmc: store hwpart in the block device

This will allow us to have multiple block device structs each referring
to the same eMMC device, yet different HW partitions.

For now, there is still a single block device per eMMC device. As before,
this block device always accesses whichever HW partition was most recently
selected. Clients wishing to make use of multiple block devices referring
to different HW partitions can simply take a copy of this block device
once it points at the correct HW partition, and use each one as they wish.
This feature will be used by the next patch.

In the future, perhaps get_device() could be enhanced to return a
dynamically allocated block device struct, to avoid the client needing to
copy it in order to maintain multiple block devices. However, this would
require all users to be updated to free those block device structs at some
point, which is rather a large change.

Most callers of mmc_switch_part() wish to permanently switch the default
MMC block device's HW partition. Enhance mmc_switch_part() so that it does
this. This removes the need for callers to do this. However,
common/env_mmc.c needs to save and restore the current HW partition. Make
it do this more explicitly.

Replace use of mmc_switch_part() with mmc_select_hwpart() in order to
remove duplicate code that skips the call if that HW partition is already
selected.

Signed-off-by: Stephen Warren <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

block: pass block dev not num to read/write/erase()

This will allow the implementation to make use of data in the block_dev
structure beyond the base device number. This will be useful so that eMMC
block devices can encompass the HW partition ID rather than treating this
out-of-band. Equally, the existence of the priv field is crying out for
this patch to exist.

Signed-off-by: Stephen Warren <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

arm: move gd handling outside of C code

As of gcc 5.2.1 for Thumb-1, it is not possible any
more to assign gd from C code, as gd is mapped to r9,
and r9 may now be saved in the prolog sequence, and
restored in the epilog sequence, of any C functions.

Therefore arch_setup_gd(), which is supposed to set
r9, may actually have no effect, causing U-Boot to
use a bad address to access GD.

Fix this by never calling arch_setup_gd() for ARM,
and instead setting r9 in arch/arm/lib/crt0.S, to
the value returned by board_init_f_alloc_reserve().

Signed-off-by: Albert ARIBAUD <[email protected]>
Reviewed-by: Simon Glass <[email protected]>

Fix board init code to respect the C runtime environment

board_init_f_mem() alters the C runtime environment's
stack it is actually already using. This is not a valid
behaviour within a C runtime environment.

Split board_init_f_mem into C functions which do not alter
their own stack and always behave properly with respect to
their C runtime environment.

Signed-off-by: Albert ARIBAUD <[email protected]>
Acked-by: Thomas Chou <[email protected]>

Merge branch 'master' of git://www.denx.de/git/u-boot-microblaze

Merge git://git.denx.de/u-boot-net

hash.c: Conditionally compile hash_command, static hash_show

The function hash_show is now only called by hash_command, so mark it as
static (and drop from hash.h).  We only call hash_command when any of
CONFIG_CMD_CRC32, CONFIG_CMD_SHA1SUM or CONFIG_CMD_HASH are set.  Since
hash.c is linked in unconditionally we must take extra care with
functions that bring in read-only strings as these will not be
discarded.

Signed-off-by: Tom Rini <[email protected]>

part_dos.c: Don't wrap to negative after 2G sectors

In order to support large IDE disks we need to make certain types be
lbaint_t now.

Reviewed-by: Tom Rini <[email protected]>
Signed-off-by: Stefan Monnier <[email protected]>

image: check "bootm_low" and "bootm_size" if "initrd_high" is missing

To boot Linux, we should prevent Initramdisk and FDT from going too
high.

Currently, boot_relocate_fdt() checks "fdt_high" environment first,
and then falls back to getenv_bootm_mapsize() + getenv_bootm_low()
if "fdt_high" is missing.

On the other hand, boot_ramdisk_high() only checks "initrd_high" to
get the address limit for the Initramdisk.  We also want to let this
case fall back to getenv_bootm_mapsize() + getenv_bootm_low().

Signed-off-by: Masahiro Yamada <[email protected]>

dm: core: Add a new api to get indexed device address

Add new api to get device address based on index.

Signed-off-by: Mugunthan V N <[email protected]>
Acked-by: Jagan Teki <[email protected]>
[Rebased on master]
Signed-off-by: Jagan Teki <[email protected]>

power: pmic.h: Drop include <spi.h>

include/power/pmic.h never used any generic spi code
from include/spi.h, but this has been added in below commit.
"dm: pmic: add implementation of driver model pmic uclass"
(sha1: 4d9057e82be11a862db411c4867e859fe0d4ca2a)

Adding functionalities in include/spi.h will trigger a build
issue as this been added used in include/power/pmic.h

Build issue on trats2 with adding BIT macro on spi.h:
----------------------------------------------------
  CC      lib/asm-offsets.s
In file included from include/power/pmic.h:15:0,
                 from include/power/max77686_pmic.h:11,
                 from include/configs/trats2.h:212,
                 from include/config.h:5,
                 from include/common.h:18,
                 from lib/asm-offsets.c:15:
include/spi.h: In function 'spi_w8r8':
include/spi.h:327:2: warning: implicit declaration of function 'BIT' [-Wimplicit-function-declaration]

Cc: Simon Glass <[email protected]>
Cc: Przemyslaw Marczak <[email protected]>
Reported-by: Tom Rini <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>
Reviewed-by: Tom Rini <[email protected]>
Acked-by: Simon Glass <[email protected]>

defconfig: am437x_sk_evm: enable spi driver model

enable spi driver model for am437x_sk_evm as ti_qspi supports
driver model

Signed-off-by: Mugunthan V N <[email protected]>
Reviewed-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>
Reviewed-by: Jagan Teki <[email protected]>

arm: dts: am4372: add qspi register maps for memory map

Add qspi memory map address to device tree.

Signed-off-by: Mugunthan V N <[email protected]>
Reviewed-by: Simon Glass <[email protected]>
Reviewed-by: Jagan Teki <[email protected]>

am43xx_evm: qspi: do not define DM_SPI and DM_SPI_FLASH for spl

Since OMAP's spl doesn't support DM currently, do not define
DM_SPI and DM_SPI_FLASH for spl build.

Signed-off-by: Mugunthan V N <[email protected]>
Reviewed-by: Tom Rini <[email protected]>
Reviewed-by: Jagan Teki <[email protected]>

defconfig: dra74_evm: enable spi driver model

enable spi driver model for dra74_evm as ti_qspi supports
driver model

Signed-off-by: Mugunthan V N <[email protected]>
Reviewed-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>
Reviewed-by: Jagan Teki <[email protected]>

defconfig: dra72_evm: enable spi driver model

enable mmc driver model for dra72_evm as ti_qspi supports
driver model

Signed-off-by: Mugunthan V N <[email protected]>
Reviewed-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>
Reviewed-by: Jagan Teki <[email protected]>

arm: dts: dra72-evm: add spi-flash comaptible for flash

In U-Boot most flashes uses "spi-flash" as compatible to bind the
device to flash driver, so adding "spi-flash" compatible to
m25p80 node.

Signed-off-by: Mugunthan V N <[email protected]>
Reviewed-by: Jagan Teki <[email protected]>

arm: dts: dra7-evm: add spi-flash comaptible for flash

In U-Boot most flashes uses "spi-flash" as compatible to bind the
device to flash driver, so adding "spi-flash" compatible to
m25p80 node.

Signed-off-by: Mugunthan V N <[email protected]>
Reviewed-by: Jagan Teki <[email protected]>

arm: dts: am437x-gp-evm: add spi-flash comaptible for flash

In U-Boot most flashes uses "spi-flash" as compatible to bind the
device to flash driver, so adding "spi-flash" compatible to
m25p80 node.

Signed-off-by: Mugunthan V N <[email protected]>
Reviewed-by: Jagan Teki <[email protected]>

arm: dts: dra7: add qspi register maps for memory map and control module

Add qspi memory map and control module register maps to device tree.

Signed-off-by: Mugunthan V N <[email protected]>
Reviewed-by: Simon Glass <[email protected]>
Reviewed-by: Jagan Teki <[email protected]>

drivers: spi: ti_qspi: convert driver to adopt device driver model

adopt ti_qspi driver to device driver model

Signed-off-by: Mugunthan V N <[email protected]>
Reviewed-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>
Reviewed-by: Jagan Teki <[email protected]>

dts: dra7: add spi alias for qspi

add spi alias for qspi so that spi probes the device and driver
successfully.

Signed-off-by: Mugunthan V N <[email protected]>
Reviewed-by: Simon Glass <[email protected]>
Reviewed-by: Jagan Teki <[email protected]>

dra7xx_evm: qspi: do not define DM_SPI and DM_SPI_FLASH for spl

Since OMAP's spl doesn't support DM currently, do not define DM_SPI and
DM_SPI_FLASH for spl build.

Signed-off-by: Mugunthan V N <[email protected]>
Reviewed-by: Jagan Teki <[email protected]>

spi: Add support for dual and quad mode

spi bus can support dual and quad wire data transfers for tx and
rx. So defining dual and quad modes for both tx and rx. Also add
support to parse bus width used for spi tx and rx transfers.

Signed-off-by: Mugunthan V N <[email protected]>
Reviewed-by: Jagan Teki <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

drivers: spi: ti_qspi: prepare driver for DM conversion

Prepare driver for DM conversion.

Signed-off-by: Mugunthan V N <[email protected]>
Reviewed-by: Jagan Teki <[email protected]>

drivers: spi:ti_qspi: change ti_qspi_slave to ti_qspi_priv for driver model conversion

Changing the ti_qspi_priv structure and its instance names from
to priv for driver mode conversion.

Signed-off-by: Mugunthan V N <[email protected]>
Reviewed-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>
Reviewed-by: Jagan Teki <[email protected]>

drivers: spi: ti_qspi: do not hard code chip select for memory map configuration

To enable memory map in dra7xx, specific chip select must be
written to control module register. But this hard coded to chip
select 1, fixing it by writing the specific chip select value to
control module register.

Signed-off-by: Mugunthan V N <[email protected]>
Reviewed-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>
Reviewed-by: Jagan Teki <[email protected]>

spi: Add SPI_TX_DUAL mode

Added SPI_TX_DUAL mode flag.

Cc: Bin Meng <[email protected]>
Cc: Mugunthan V N <[email protected]>
Tested-by: Jagan Teki <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

sf: Make IO modes at last in read modes

SLOW, FAST, DUAL, DUAL_IO, QUAD, QUAD_IO changed order to
SLOW, FAST, DUAL, QUAD, DUAL_IO, QUAD_IO

Cc: Simon Glass <[email protected]>
Cc: Michal Simek <[email protected]>
Cc: Siva Durga Prasad Paladugu <[email protected]>
Tested-by: Jagan Teki <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

spi: Minor cleanup

- Add comments on mode_rx
- Tab space's

Cc: Simon Glass <[email protected]>
Cc: Michal Simek <[email protected]>
Cc: Siva Durga Prasad Paladugu <[email protected]>
Tested-by: Mugunthan V N <[email protected]>
Tested-by: Bin Meng <[email protected]>
Tested-by: Jagan Teki <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

spi: Use BIT macro

Used BIT macro like 1 << nr as BIT(nr) where nr is 0...n

Cc: Simon Glass <[email protected]>
Cc: Michal Simek <[email protected]>
Cc: Siva Durga Prasad Paladugu <[email protected]>
Tested-by: Mugunthan V N <[email protected]>
Tested-by: Bin Meng <[email protected]>
Tested-by: Jagan Teki <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

spi: Fix bit assignment with flags

Fixed bit assignment with flags members on spi_slave{}

Cc: Simon Glass <[email protected]>
Cc: Michal Simek <[email protected]>
Cc: Siva Durga Prasad Paladugu <[email protected]>
Tested-by: Mugunthan V N <[email protected]>
Tested-by: Bin Meng <[email protected]>
Tested-by: Jagan Teki <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

spi: Move flags macro's to spi_slave{} members

This patch moves flags macro's to respective member
position on spi_slave{}, for better readabilty and
finding the respective member macro's easily.

Cc: Simon Glass <[email protected]>
Cc: Bin Meng <[email protected]>
Cc: Michal Simek <[email protected]>
Cc: Siva Durga Prasad Paladugu <[email protected]>
Tested-by: Mugunthan V N <[email protected]>
Tested-by: Jagan Teki <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

spi: Rename op_mode_rx to mode_rx

Since spi rx mode macro's are renamed to simple and
meaninfull, this patch will rename the respective
structure members.

Cc: Simon Glass <[email protected]>
Cc: Bin Meng <[email protected]>
Cc: Michal Simek <[email protected]>
Cc: Siva Durga Prasad Paladugu <[email protected]>
Tested-by: Mugunthan V N <[email protected]>
Tested-by: Jagan Teki <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

spi: Rename SPI_OPM_RX_* to SPI_RX_*

SPI_OPM_RX_AS  - SPI_RX_SLOW
SPI_OPM_RX_AF - SPI_RX_FAST
SPI_OPM_RX_DOUT - SPI_RX_DUAL
SPI_OPM_RX_QOF - SPI_RX_QUAD

Cc: Simon Glass <[email protected]>
Cc: Michal Simek <[email protected]>
Cc: Siva Durga Prasad Paladugu <[email protected]>
Tested-by: Mugunthan V N <[email protected]>
Tested-by: Bin Meng <[email protected]>
Tested-by: Jagan Teki <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

spi: Remove SPI_OPM_RX_DIO|QIOF

SPI_OPM_RX_DIO and SPI_OPM_RX_QIOF are rx IO
commands/opmodes for dual and quad. Usually IO
operation's are referred to flash protocol rather
with spi controller protocol, these are still present
in flash side for the usage of spi-nor controllers.

Cc: Simon Glass <[email protected]>
Cc: Michal Simek <[email protected]>
Cc: Siva Durga Prasad Paladugu <[email protected]>
Tested-by: Mugunthan V N <[email protected]>
Tested-by: Bin Meng <[email protected]>
Tested-by: Jagan Teki <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

spi: Remove SPI_OPM_RX_EXTN

SPI_OPM_RX_EXTN is a combination of all rx opmode's
and spi driver shall use any one of the rx mode at
a time not the combination and it is true in case of
flash where spi_flash_table mention combination of
supported read opmodes so-that the required one
will pick based on the rx mode from spi driver.

Cc: Simon Glass <[email protected]>
Cc: Michal Simek <[email protected]>
Cc: Siva Durga Prasad Paladugu <[email protected]>
Tested-by: Mugunthan V N <[email protected]>
Tested-by: Bin Meng <[email protected]>
Tested-by: Jagan Teki <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

sf: Write quad bit along with read status

While setting quad bit on spansion, macronix code
is writing only particular quad bit this may give
wrong functionality with other register bits,
So this patch fix the issue where it with write
previous read reg status along  particular quad bit.

Cc: Vignesh R <[email protected]>
Cc: Mugunthan V N <[email protected]>
Cc: Simon Glass <[email protected]>
Cc: Bin Meng <[email protected]>
Cc: Michal Simek <[email protected]>
Cc: Siva Durga Prasad Paladugu <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

sf: Read back and check once macronix quad bit set

One macronix quad bit set using SR, it's good to
read back and check the written bit and also if
it's already been set check for the bit and return.

Cc: Vignesh R <[email protected]>
Cc: Simon Glass <[email protected]>
Cc: Bin Meng <[email protected]>
Tested-by: Mugunthan V N <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

sf: Read back and check once spansion quad bit set

One spansion quad bit set using CR, it's good to
read back and check the written bit and also if
it's already been set check for the bit and return.

Cc: Vignesh R <[email protected]>
Cc: Simon Glass <[email protected]>
Cc: Bin Meng <[email protected]>
Cc: Michal Simek <[email protected]>
Cc: Siva Durga Prasad Paladugu <[email protected]>
Tested-by: Mugunthan V N <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

sf: Fix quad bit set for micron devices

Setting up quad bit for micron devices need to do the
same way as other flash devices like spansion, winbond
etc does using enhanced volatile config register so this
patch adds this support instead of printing "QEB is volatile"

Cc: Simon Glass <[email protected]>
Cc: Bin Meng <[email protected]>
Cc: Peter Pan <[email protected]>
Cc: Fabio Estevam <[email protected]>
Cc: Michal Simek <[email protected]>
Cc: Siva Durga Prasad Paladugu <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

sf: Use BIT macro

Used BIT macro like 1 << nr as BIT(nr) where nr is 0...n

Reviewed-by: Bin Meng <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

sf: Minor cleanup

- Tab space
- Place all read commands at one place.
- Re-arrange write commands.

Reviewed-by: Bin Meng <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

sf: Remove spi_flash_remove

Use direct call to device_remove instead of exctra
spi_flash_remove defination.

Reviewed-by: Simon Glass <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

sf: Rename bank_end to bar_end in read_bar

bar_end gives more meaningfull compared to bank_end and
spi_flash_write_bar uses bar_end so replaced bank_end with
bar_end in spi_flash_read_bar

Signed-off-by: Jagan Teki <[email protected]>

sf: Rename spi_flash_set_* functions

Since quad_mode functions are local to spi flash core,
rename them to a meaningful and readable names.

Signed-off-by: Jagan Teki <[email protected]>

sf: Move spi_read_cmds_array locally

Since spi_read_cmds_array is used locally in
spi_flash_scan, so move array to locally used
function instead of defining global array.

Tested-by: Jagan Teki <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

spi: Rename SPI_TX_BP|QPP to SPI_TX_BYTE|QUAD

Since SPI_TX_* are spi_slave{} members so use spi protocol
notation instead spi flash programming, like

SPI_TX_BP  => SPI_TX_BYTE
SPI_TX_QPP => SPI_TX_QUAD

Cc: Simon Glass <[email protected]>
Tested-by: Jagan Teki <[email protected]>
Tested-by: Bin Meng <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

spi: Use mode instead of op_mode_tx

Used mode member from spi_slave{} instead of op_mode_tx.

Cc: Simon Glass <[email protected]>
Tested-by: Jagan Teki <[email protected]>
Tested-by: Bin Meng <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

spi: make mode visible to both dm and non-dm

Couldn't find the exact reason to define 'mode' for dm,
probably it is not using in non-dm drivers but it need
to visible both dm and non-dm as mode data is getting
dereferred in spi flash core ie common to both.

Cc: Simon Glass <[email protected]>
Tested-by: Jagan Teki <[email protected]>
Tested-by: Bin Meng <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

sf: Get spi locally from spi_flash

For better code readabilty, get the spi pointer from
spi_flash{} locally and use it instead of direct
dereferring spi pinter as flash->spi->*

Tested-by: Jagan Teki <[email protected]>
Tested-by: Bin Meng <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Signed-off-by: Jagan Teki <[email protected]>

arm: socfpga: Enable SPL_DM_SEQ_ALIAS for all SOCFPGA configs

This feature is required in SPL to enable support for loading from SPI
flash.

Also clean up the #define in socfpga_common.h.

Signed-off-by: Nathan Rossi <[email protected]>
Reviewed-by: Stefan Roese <[email protected]>
Acked: Marek Vasut <[email protected]>
Cc: Dinh Nguyen <[email protected]>
Cc: Chin-Liang See <[email protected]>
Cc: Marek Vasut <[email protected]>
Cc: Stefan Roese <[email protected]>
Signed-off-by: Michal Simek <[email protected]>

arm: mvebu: Select SPL_DM_SEQ_ALIAS

Select SPL_DM_SEQ_ALIAS which is required for certain uclasses,
specifically SPI Flash.

Signed-off-by: Nathan Rossi <[email protected]>
Reviewed-by: Stefan Roese <[email protected]>
Cc: Albert Aribaud <[email protected]>
Cc: Stefan Roese <[email protected]>
Acked-by: Stefan Roese <[email protected]>
Signed-off-by: Michal Simek <[email protected]>

arm: zynq: Enable SPL_DM_SEQ_ALIAS for all Zynq configs

This feature is required in SPL to enable support for loading from SPI
flash when the device is booted from QSPI.

Signed-off-by: Nathan Rossi <[email protected]>
Cc: Michal Simek <[email protected]>
Signed-off-by: Michal Simek <[email protected]>

spl: dm: Add SPL_DM_SEQ_ALIAS config option

The Device Model sequence alias feature is required by some Uclasses.
Instead of disabling the feature for all SPL targets allow it to be
configured.

The config option is disabled by default to reduce code size for targets
that are not interested or do not require this feature.

Signed-off-by: Nathan Rossi <[email protected]>
Acked-by: Simon Glass <[email protected]>
Acked-by: Michal Simek <[email protected]>
Cc: Simon Glass <[email protected]>
Cc: Masahiro Yamada <[email protected]>
Cc: Linus Walleij <[email protected]>
Cc: Marek Vasut <[email protected]>
Cc: Michal Simek <[email protected]>
Cc: Stefan Roese <[email protected]>
Signed-off-by: Michal Simek <[email protected]>

Merge git://git.denx.de/u-boot-dm

sandbox: eth-raw-os.c: Ensure that our interface name is not too long

Coverity notes that we do not ensure when we copy ifname we still have
space left to ensure NULL termination.  As cannot control the size of
ifr_name we must make sure that our argument will not overflow the
buffer.

Reported-by: Coverity (CID 131094)
Cc: Simon Glass <[email protected]>
Signed-off-by: Tom Rini <[email protected]>
Reviewed-by: Simon Glass <[email protected]>

dm: core: Add option to configure an offset for the address translation

Some platforms need to ability to configure an offset to the standard
addresses extracted from the device-tree. This patch allows this by
adding a function to DM to configure this offset (if needed).

Signed-off-by: Stefan Roese <[email protected]>
Acked-by: Simon Glass <[email protected]>
Cc: Simon Glass <[email protected]>
Fixed space before tab:
Signed-off-by: Simon Glass <[email protected]>

rockchip: Use the debug UART on rk3036

Rather than using a new debug UART implementation, use the standard one
provided by U-Boot.
Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Thomas Chou <[email protected]>

dm: ns16550: Allow the driver to be omitted if requested

Allow the ns16550 debug UART to be used without the full driver.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Thomas Chou <[email protected]>

dm: serial: Allow the UART driver to be dropped from the image

In very very space-constrained devices even the full UART driver is too
large. In this case the debug UART can still be used in some cases.

Add options to enable the UART driver in SPL and U-Boot proper. Enable both
options by default.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Thomas Chou <[email protected]>

dm: net: usb: Convert mcs7830 driver to support driver model

Adjust this driver to support driver model for Ethernet.

Signed-off-by: Simon Glass <[email protected]>
Acked-by: Joe Hershberger <[email protected]>

dm: net: usb: Refactor mcs7830 driver ready for DM conversion

Remove stamp data and create common functions for the main Ethernet
operations. This will make it easier to convert this driver to support
driver model.

Signed-off-by: Simon Glass <[email protected]>
Acked-by: Joe Hershberger <[email protected]>

pci: Tidy up comments in pci_bind_bus_devices()

The current comments are confusing. We don't actually bind a generic device
when the device tree has no information. We try to scan available PCI
drivers. Update the comments to reflect this.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>

dm: Convert PCI MMC over to use DM PCI API

At present pci_mmc_init() does not correctly use the PCI function since the
list it passes is not terminated. The array size passed to pci_mmc_init() is
actually not used correctly. Fix this and adjust the pci_mmc_init() to scan
all available MMC devices.

Adjust this code to use the new driver model PCI API.

This should move over to the new MMC uclass at some point.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>

dm: usb: Convert echi-pci to use new DM PCI API

Convert this driver to use the new driver model PCI API.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Acked-by: Marek Vasut <[email protected]>

dm: ahci: Convert to use new DM PCI API

Convert this driver to use the new driver model PCI API.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>

dm: pci: Switch to DM API for PCI address mapping

We should use the new address mapping functions unless we are in
compatibility mode. Disable the old functions by default.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>

dm: net: Convert rtl8169 to use DM PCI API

Update this driver to use the proper driver-model PCI API functions.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Acked-by: Joe Hershberger <[email protected]>

dm: pci: Add driver model API functions for address mapping

At present the PCI address map functions use the old API. Add new functions
for this so that drivers can be converted.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>

dm: test: Convert PCI tests to use the DM PCI API

Update these tests to use driver model API functions.

Signed-off-by: Simon Glass <[email protected]>

dm: tegra: net: Convert tegra boards to driver model for Ethernet

Adjust all Tegra boards to use driver model for Ethernet, now that the
required drivers are converted.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Acked-by: Joe Hershberger <[email protected]>

dm: pci: Drop the old version of pci_find_class()

Move this function into the compatibility file so that it is not available
by default.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>

dm: pci: Drop the old version of pci_find_device/s()

Move these functions into the compatibility file so that they are not
available by default.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>

dm: x86: pci: Adjust bios_run_on_x86() to use the DM PCI API

This function should take a struct udevice rather than pci_dev_t. Update it.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Tested-by: Bin Meng <[email protected]>

dm: pci: video: Convert video and pci_rom to use DM PCI API

Adjust these files to use the driver-model PCI API instead of the legacy
functions.

Signed-off-by: Simon Glass <[email protected]>
Acked-by: Anatolij Gustschin <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Tested-by: Bin Meng <[email protected]>

dm: Convert bios_interrupts to use DM PCI API

Adjust this code to use driver model for devices where possible. Since
existing users have not been converted the old code must remain.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Tested-by: Bin Meng <[email protected]>

dm: x86: ivybridge: Convert graphics init to use DM PCI API

Use the driver-model PCI functions here where possible. For now we have to
search for the device with pci_bus_find_bdf() but at some point we can put
this in a proper driver and avoid this.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>

dm: serial: Convert ns16550 driver to use driver model PCI API

Use the driver model version of the function to find the BAR. This updates
the fdtdec function, of which ns16550 is the only user.

The fdtdec_get_pci_bdf() function is dropped for several reasons:
- with driver model we should use 'struct udevice *' rather than passing the
   device tree offset explicitly
- there are no other users in the tree
- the function parses for information which is already available in the PCI
device structure (specifically struct pci_child_platdata which is available
at dev_get_parent_platdata(dev)

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Tested-by: Bin Meng <[email protected]>

dm: pci: Add a function to read a PCI BAR

Add a driver-model function for reading the PCI BAR from a device.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>

dm: pci: Add a driver-model version of pci_find_class()

Add a function which scans the driver model device information rather
than scanning the PCI bus again.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Tested-by: Bin Meng <[email protected]>

dm: pci: scsi: Use driver-model PCI API

Adjust the SCSI command to use driver model for its PCI interface.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Tested-by: Bin Meng <[email protected]>

dm: pci: Add a driver-model version of pci_find_device()

Add a function which scans the driver model device information rather
than scanning the PCI bus again.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Tested-by: Bin Meng <[email protected]>

dm: pci: Use driver model PCI API in auto-config

At present we are using legacy functions even in the auto-configuration code
used by driver model. Add a new pci_auto.c version which uses the correct
API.

Create a new pci_internal.h header to hold functions that are used within
the PCI subsystem, but are not exported to other drivers.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>
Tested-by: Bin Meng <[email protected]>

dm: pci: Add a dm_ prefix to pci_bus_find_bdf()

Most driver model PCI functions have a dm_ prefix. At some point, when the
old code is converted to driver model and the old functions are removed, we
will drop that prefix.

For consistency, we should use the dm_ prefix for all driver model
functions. Update pci_bus_find_bdf() accordingly.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>

dm: pci: Add a dm_ prefix to pci_get_bdf()

Most driver model PCI functions have a dm_ prefix. At some point, when the
old code is converted to driver model and the old functions are removed, we
will drop that prefix.

For consistency, we should use the dm_ prefix for all driver model
functions. Update pci_get_bdf() accordingly.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>

dm: pci: Mark legacy files as such

We don't want people changing the legacy PCI files while migration is in
progress. Update the file headers to indicate that.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>

ARM: uniphier: fix recommended board setting in document

The mem_is_flash() in arch/arm/mach-uniphier/micro-support_card.c
writes/reads the tail of each NOR flash bank to check if the device
really resides there.

If CS1_SPLIT were enabled, the support card would always require two
NOR flash devices to be inserted for the correct NOR detection.
This is not probably what we want.

Signed-off-by: Masahiro Yamada <[email protected]>

ARM: uniphier: add dump command for DDR Multi PHY registers

The ProXstream2/PH1-LD6b is integrated with a new IP for DDR PHY
which is not register-compatible with the former SoCs.
Add a new command to support the register dump of this IP.

Signed-off-by: Masahiro Yamada <[email protected]>

ARM: uniphier: add DRAM init code for ProXstream2/PH1-LD6b

As mentioned in the log of commit 019df879a93e2 (ARM: uniphier: add
ProXstream2 and PH1-LD6b support), the DRAM init code was missing
for a long time.  Finally, here it is.  SPL works now.

Signed-off-by: Masahiro Yamada <[email protected]>

ARM: uniphier: move headers out of include/mach directory

These headers are only included locally in arch/arm/mach-uniphier/.
There is no reason to export them by putting in the mach/ directory.

Signed-off-by: Masahiro Yamada <[email protected]>

ARM: uniphier: move cmd_pinmon.c to boot-mode subdirectory

This command shows the boot mode pins, so it would be more at home
in the boot-mode subdirectory.

Signed-off-by: Masahiro Yamada <[email protected]>

ARM: uniphier: move cmd_ddrphy.c to dram subdirectory

This command shows the register dump of the DDR PHY, so it would be
more at home in the dram subdirectory.

Signed-off-by: Masahiro Yamada <[email protected]>

ARM: uniphier: kill bogus header includes

These includes are not necessary.

Signed-off-by: Masahiro Yamada <[email protected]>

Prepare v2016.01

Signed-off-by: Tom Rini <[email protected]>

dm: test: usb: Update the USB tests so that they all pass

Due to a limitation removed in an earlier patch, USB tests were not seeing
all the devices. Update the tests to pass now that all devices are visible.

Signed-off-by: Simon Glass <[email protected]>