Marek Vasut [Fri, 31 Aug 2012 16:18:10 +0000 (16:18 +0000)]
MX28: MMC: Avoid DMA DCache race condition
This patch prevents dcache-related problem. The problem manifested
itself on the SPI driver, this is just a port to the MMC driver.
The scenario is the same. In case an "mmc read" is issued to a
buffer which was written right before it and data cache is enabled,
the cache eviction might happen during the DMA transfer into the
buffer, therefore corrupting the buffer. Clear any cache lines that
might contain the buffer to prevent such issue.
Signed-off-by: Marek Vasut <marex@denx.de> Cc: Fabio Estevam <festevam@gmail.com> Cc: Otavio Salvador <otavio@ossystems.com.br> Cc: Stefano Babic <sbabic@denx.de>
Marek Vasut [Fri, 31 Aug 2012 16:08:00 +0000 (16:08 +0000)]
MX28: SPI: Fix the DMA chaining
It turns out that in order for the SPI DMA to properly support
continuous transfers longer than 65280 bytes, there are some very
important parts that were left out from the documentation.
Firstly, the XFER_SIZE register is not written with the whole length
of a transfer, but is written by each and every chained descriptor
with the length of the descriptors data buffer.
Next, unlike the demo code supplied by FSL, which only writes one PIO
word per descriptor, this does not apply if the descriptors are chained,
since the XFER_SIZE register must be written. Therefore, it is essential
to use four PIO words, CTRL0, CMD0, CMD1, XFER_SIZE. CMD0 and CMD1 are
written with zero, since they don't apply. The DMA programs the PIO words
in an incrementing order, so four PIO words.
Finally, unlike the demo code supplied by FSL, the SSP_CTRL0_IGNORE_CRC
must not be set during the whole transfer, but it must be set only on the
last descriptor in the chain.
Signed-off-by: Marek Vasut <marex@denx.de> Cc: Fabio Estevam <festevam@gmail.com> Cc: Otavio Salvador <otavio@ossystems.com.br> Cc: Stefano Babic <sbabic@denx.de>
Marek Vasut [Fri, 31 Aug 2012 16:07:59 +0000 (16:07 +0000)]
MX28: SPI: Fix the DMA DCache race condition
This patch fixes dcache-related problem. The problem manifested
when dcache was enabled and the following command issued twice:
mw 0x42000000 0 0x4000 ; sf probe ; sf read 0x42000000 0x0 0x10000 ; sha1sum 0x42000000 0x10000
The SHA1 checksum was correct during the first call. Yet with
every subsequent call of the above command, it differed and was
wrong.
It turns out this was because of a race condition. On the first
time the command was called, no cacheline contained any data from
the destination memory location. The DMA transfered data into the
location and the cache above the location was invalidated. Then the
checksum was computed, but that meant the data were loaded into data
cache.
On any subsequent call, the DMA again transfered data into the same
destination. Yet during the transfer, some of the DCache lines were
evicted and written back into the main memory. Once the DMA transfer
completed, the data cache was invalidated over the memory location as
usual. But the data that were to be loaded back into the data cache
by subsequent SHA1 checksuming were corrupted.
Signed-off-by: Marek Vasut <marex@denx.de> Cc: Fabio Estevam <festevam@gmail.com> Cc: Otavio Salvador <otavio@ossystems.com.br> Cc: Stefano Babic <sbabic@denx.de>
Switch the mx35 timer driver to the 32-kHz clock source to avoid calling
mxc_get_clock() again and again, and to be consistent with the timer drivers of
other i.MX SoCs.
The clock dividers that were used do not match at all the reference manual. They
were either completely broken, or came from an early silicon revision
incompatible with the current one.
Matt Sealey [Thu, 23 Aug 2012 04:52:33 +0000 (04:52 +0000)]
efikamx: refine USB support
Because of the way USB pad settings are handled it doesn't make sense to
be able to build the Efika MX board support without CONFIG_CMD_USB turned
on. So, we change the build to always compile in USB support.
We do not need to check for CONFIG_CMD_USB like we do with CONFIG_MXC_SPI
since the USB subsystem will error out of the compile for us.
Additionally, the following behaviors have changed;
* Smartbook "preboot" should not set input and output to USB keyboard as
there is no display support
* board_eth_init is implemented such that it does not cause U-Boot to
report an explicit failure ("CPU Net Initialization Failed").
Since Ethernet is implemented via USB (fixed on Smarttop, pluggable on
Smartbook, and handled by "usb start") - the warning that is left
("No ethernet found") is perfectly reasonable at the point it is printed
since the USB system hasn't been started and nothing has been probed yet.
Signed-off-by: Matt Sealey <matt@genesi-usa.com> Cc: Stefano Babic <sbabic@denx.de> Cc: Marek Vasut <marex@denx.de>
Marek Vasut [Tue, 28 Aug 2012 15:12:48 +0000 (15:12 +0000)]
M28: Fix the use of gpmi-nand in mtdparts
The mtd name of the NAND in Linux is "gpmi-nand", not "gpmi-nand.0" as
it would be expected, since the controller doesn't support multiple NANDs
attached to it as of now. Rectify this flub by adjusting default mtdparts.
Signed-off-by: Marek Vasut <marex@denx.de> Cc: Fabio Estevam <fabio.estevam@freescale.com> Cc: Stefano Babic <sbabic@denx.de> Cc: Otavio Salvador <otavio@ossystems.com.br> Acked-by: Stefano Babic <sbabic@denx.de>
Michael Walle [Mon, 30 Jul 2012 10:47:12 +0000 (10:47 +0000)]
lsxl: support power switch
This patch restores the Linkstation's original behaviour when powering off.
Once the (soft) power switch is turned off, linux will reboot and the
bootloader turns off HDD and USB power. Then it loops as long as the switch
is in the off position, before continuing the boot process again.
Additionally, this patch fixes the board function set_led(LED_OFF).
Signed-off-by: Michael Walle <michael@walle.cc> Cc: Prafulla Wadaskar <prafulla@marvell.com>
Karl O. Pinc [Thu, 2 Aug 2012 16:51:56 +0000 (16:51 +0000)]
cosmetic: Better explain how to use the kirkwood kwbimage.cfg file.
Hi,
This adds to the documenation to explain how to use the
kwbimage.cfg file necessary to generate an image with
prefixed board setup values necessary for the kirkwood
boards.
Holger Brunck [Thu, 9 Aug 2012 01:37:47 +0000 (01:37 +0000)]
arm/km: remove unused code
For some reasons we had an own implementaion of dram_init and
dram_init_banksize. This is not needed anymore, use the standard
kirkwood functions instead.
km/ivm: fix string len check to support 7 char board names
The fanless boards now have a 7-digit (XXXXX-F) board name. This
triggers a border condition when reading this string in the IVM although
this string is smaller than the currenly read string size, but only by 1
character.
This patch corrects this by changing the size check condition for string
length. It is the same change that was done in the platform for this
same bug.
The computation was not correct with low clock values: setting a 1MHz
clock would result in an overlap that would then configure a 25Mhz
clock.
This patch implements a correct computation method according to the
kirkwood functionnal spec. table 600 (Serial Memory Interface
Configuration Register).
Wolfgang Denk [Sat, 1 Sep 2012 22:44:09 +0000 (00:44 +0200)]
Merge branch 'agust@denx.de' of git://git.denx.de/u-boot-staging
* 'agust@denx.de' of git://git.denx.de/u-boot-staging:
tx25: Use generic gpio_* calls
config: Always use GNU ld
tools: add kwboot binary to .gitignore file
fdt: Include arch specific gpio.h instead of asm-generic/gpio.h
serial: CONSOLE macro is not used
Wu, Josh [Thu, 23 Aug 2012 00:05:36 +0000 (00:05 +0000)]
at91: atmel_nand: Update driver to support Programmable Multibit ECC controller
The Programmable Multibit ECC (PMECC) controller is a programmable binary
BCH(Bose, Chaudhuri and Hocquenghem) encoder and decoder. This controller
can be used to support both SLC and MLC NAND Flash devices. It supports to
generate ECC to correct 2, 4, 8, 12 or 24 bits of error per sector of data.
To use PMECC in this driver, the user needs to set the PMECC correction
capability, the sector size and ROM lookup table offsets in board config file.
This driver is ported from Linux kernel atmel_nand PMECC patch. The main difference
is in this version it uses registers structure access hardware instead of using macros.
It is tested in 9x5 serial boards.
Signed-off-by: Josh Wu <josh.wu@atmel.com>
[rebase] Signed-off-by: Andreas Bießmann <andreas.devel@googlemail.com>
Wu, Josh [Thu, 23 Aug 2012 00:05:34 +0000 (00:05 +0000)]
at91: atmel_nand: extract HWECC initialization code into one function: atmel_hw_nand_init_param().
This patch
1. extract the hwecc initialization code into one function. It is a preparation for adding atmel PMECC support.
2. enable CONFIG_SYS_NAND_SELF_INIT. Which make us can configurate the ecc parameters between nand_scan_ident() and nand_scan_tail().
Signed-off-by: Josh Wu <josh.wu@atmel.com>
[fix empty newline at EOF error and move return value check into ifdef] Signed-off-by: Andreas Bießmann <andreas.devel@googlemail.com>
Bo Shen [Sun, 19 Aug 2012 20:32:22 +0000 (20:32 +0000)]
spi: atmel: add WDRBT bit to avoid receive overrun
The atmel at91sam9x5 series spi has feature to avoid receive overren
Using the patch to enable it
Signed-off-by: Bo Shen <voice.shen@atmel.com> Acked-by: Andreas Bießmann <andreas.devel@googlemail.com> Signed-off-by: Andreas Bießmann <andreas.devel@googlemail.com>
Xu, Hong [Tue, 2 Aug 2011 01:05:04 +0000 (01:05 +0000)]
AT91: Small fix on AT91 USART initialization code
Before reset dbgu transmitter, we just wait TXEMPTY to drain the
transmitter register(Just in case). If not doing this, we may sometimes
see several weird characters from DBGU.
A short delay is also added to make sure the new serial settings are
settled.
Signed-off-by: Hong Xu <hong.xu@atmel.com>
[cherry-picked from u-boot-atmel/old-next] Signed-off-by: Andreas Bießmann <andreas.devel@googlemail.com>
The cache snooping feature of Freescale's eSDHC IP is not available on i.MX, so
disable it globally for this architecture. This avoids setting no_snoop for all
i.MX boards, and it prevents setting a reserved bit of a reserved register if
fsl_esdhc_mmc_init() is used on i.MX, like in
arch/arm/cpu/armv7/imx-common/cpu.c/cpu_mmc_init().
Since no_snoop was only used on i.MX, get rid of it BTW.
Signed-off-by: Benoît Thébaudeau <benoit.thebaudeau@advansee.com> Cc: Andy Fleming <afleming@gmail.com> Cc: Stefano Babic <sbabic@denx.de> Cc: Kim Phillips <kim.phillips@freescale.com>
Matt Sealey [Fri, 24 Aug 2012 06:44:24 +0000 (06:44 +0000)]
efikamx: sync Smartbook DDR settings in DCD with those found in Genesi's production U-Boot
We have no idea where the DCD was derived from for Smartbook support, but they
differ from the Smarttop settings, MX51EVK settings and certainly don't
correspond to any shipped or development version of U-Boot that Genesi has ever
had on any Smartbook.
So, copy the calibrated, verified settings from the U-Boot as shipped with every
Smartbook since retail production. Remove those few settings that just set the
POR defaults which have already been confirmed for the previous Smarttop DCD
change.
One of the lines is specific to i.MX51 TO3 designs and therefore TO2 Smartbooks
will possibly not work so reliably with this new DCD; that said, TO2 Smartbooks
basically don't exist at retail and the number of units in the world is less
than 5 (3 of which are at the Genesi office or owned by Genesi employees).
Many hours of memory testing confirms the new settings are stable.
Patch v2:
* picked the correct commit from our development tree, correcting tuned DDR ODF setting
(which was correct anyway)
Signed-off-by: Matt Sealey <matt@genesi-usa.com> Cc: Stefano Babic <sbabic@denx.de> Cc: Marek Vasut <marex@denx.de>
Matt Sealey [Mon, 27 Aug 2012 05:58:30 +0000 (05:58 +0000)]
efikamx: update to Efika MX Smarttop and Smartbook boards
This is a rework of a previously submitted patchset and bundles the
main board support and USB support into a single commit.
It requires the patch "mx5: add iomux-mx51.h include"
* Use iomux-mx51.h include to simplify board configuration.
* Simplify LED support (remove efikamx_toggle_led, change lit LEDs).
* Simplify MMC support for CD and WP pin differences.
* Fix broken CPU voltage setting - comment said 1.1V but the code set to
1.2V. It should never have been set to 1.2V even on i.MX51 TO2 and
all available Linux kernels would drop the voltage to 1.1V anyway and
work reliably. This should lower power consumption during the boot
process.
* Function renames for readability.
* Some board identification string changes to match actual product names.
* Passes checkpatch (v2)
Signed-off-by: Matt Sealey <matt@genesi-usa.com> Cc: Stefano Babic <sbabic@denx.de>
Matt Sealey [Wed, 22 Aug 2012 09:25:40 +0000 (09:25 +0000)]
efikamx: configure Smarttop PCBID and LED pads in DCD for convenience
PCBID pads seem to need time to settle due to external pulldowns, otherwise
we are reading floating GPIO pins with implicit pad pullups and get the wrong
data. However we can't "wait" at the time we need them before relocation,
since timers are not available. The time taken to get from DCD to the code
requiring the pads set seems to be more than long enough (even with caches
enabled).
We have space in the DCD due to the DDR settings changes to configure all
the pad settings we need for this, plus the LED pad settings too which
reduces the amount of code required later on.
Signed-off-by: Matt Sealey <matt@genesi-usa.com> Acked-by: Stefano Babic <sbabic@denx.de>
Matt Sealey [Wed, 22 Aug 2012 09:25:39 +0000 (09:25 +0000)]
efikamx: remove drive strength function and roll its functionality into the DCD
Efika MX boards configure their DDR pad settings twice, one in the DCD generated
from imximage_*.cfg and again in init_drive_strength called before relocation.
Rather than doing this, roll the changes it makes into the DCD so DDR is set up
before a single line of code in U-Boot is run.
The settings are identical with this DCD block which is shorter (by 7 entries)
than the old one, and after the output of init_drive_strength since a lot of the
functionality in the existing DCD and init_drive_strength function was just
setting the POR defaults. This goes to explain some now-missing entries.
Several hundred rounds of mtest have been run to test the settings before and
after to confirm DDR is stable and no ill-effects have been found.
Signed-off-by: Matt Sealey <matt@genesi-usa.com> Acked-by: Stefano Babic <sbabic@denx.de>
Matt Sealey [Wed, 22 Aug 2012 09:25:38 +0000 (09:25 +0000)]
efikamx: move and rename Efika MX directories and config files to prepare for new boards
* Move Efika MX Smarttop and Smartbook boards into a "genesi" vendor directory
* Rename efikamx -> mx51_efikamx since there is an mx53_efikamx and mx6_efikamx to come
Marek Vasut [Tue, 21 Aug 2012 16:17:26 +0000 (16:17 +0000)]
MX28: DMA: Prolong the DMA timeout
Load from SPI flash can create a long DMA chain, which can take long
time to transfer. Change the DMA timeout to roughly 10s to prevent
such long chains misreporting errors.
Signed-off-by: Marek Vasut <marex@denx.de> Cc: Fabio Estevam <festevam@gmail.com> Cc: Otavio Salvador <otavio@ossystems.com.br> Cc: Stefano Babic <sbabic@denx.de>
Matt Sealey [Wed, 22 Aug 2012 09:24:06 +0000 (09:24 +0000)]
mx5: add iomux-mx51.h include
Allow usage of the imx-common/iomux-v3.h framework by including pad settings
for the i.MX51. The content of the file is taken from Linux kernel at
commit 5d23b39 plus the required changes to make it work in U-Boot.
The contained pad settings are the minimum required to make an Efika MX boot
and get all the currently-implemented peripherals working in U-Boot.
It is recommended that this file not be just a dumping ground for pins but
only contain the settings required for all the boards using it.
Changes for v2:
* reference commit id from Linux kernel
* additionally roll in the USB pads
* removed GPIO_NUMBER define
Signed-off-by: Matt Sealey <matt@genesi-usa.com> Acked-by: Stefano Babic <sbabic@denx.de>
Otavio Salvador [Sun, 19 Aug 2012 04:58:29 +0000 (04:58 +0000)]
mxs: Only build internal Ethernet controller for i.MX28
The internal Ethernet controller is only available on i.MX28
processors so it needs to use CONFIG_MX28 guardian to avoid having
this code called in others.
Signed-off-by: Otavio Salvador <otavio@ossystems.com.br> Acked-by: Marek Vasut <marex@denx.de>
gpio_get_value() should use PSR like Linux, not DR, because DR does not always
reflect the pin state, while PSR does. This is especially useful to detect a
short circuit on a GPIO pin configured as output, or to read the level of a pin
controlled by a non-GPIO IOMUX function.
Stefano Babic [Sun, 19 Aug 2012 21:33:50 +0000 (21:33 +0000)]
MX: Set a common gpio.h for all i.MX
Each i.MX has its own gpio.h, defining the same structure.
The internal GPIO controller has the same layout
(at least for the register used by u-boot) and can be shared.
Signed-off-by: Stefano Babic <sbabic@denx.de> Tested-by: Matt Sealey <matt@genesi-usa.com>
Do not pretend to have initialized mmc successfully if CONFIG_FSL_ESDHC is not
defined. Instead, only implement a custom cpu_mmc_init() when it does something.
Matt Sealey [Fri, 17 Aug 2012 08:15:11 +0000 (08:15 +0000)]
spi: fix mxs_spi_slave structure allocation to clear memory
Use calloc() instead of malloc() to allocate the mxs_spi_slave structure.
Clearing the memory is necessary since most of the time this gets done
super early in boot, but on warm reboots, and when SPI probing is done
long after the init stages it could actually pick up previously used memory,
and things like the chipselect polarity and other data end up being filled
with trash data if not explicitly set by the board files.
This solves a semi-random, almost unreproducable error whereby SPI devices
act very, very strangly on boot.
Signed-off-by: Matt Sealey <matt@genesi-usa.com> Acked-by: Stefano Babic <sbabic@denx.de>
Matt Sealey [Fri, 17 Aug 2012 08:15:10 +0000 (08:15 +0000)]
spi: fix mxc_spi_slave structure allocation to clear memory
Use calloc() instead of malloc() to allocate the mxc_spi_slave structure.
Clearing the memory is necessary since most of the time this gets done
super early in boot, but on warm reboots, and when SPI probing is done
long after the init stages it could actually pick up previously used memory,
and things like the chipselect polarity and other data end up being filled
with trash data if not explicitly set by the board files.
This solves a semi-random, almost unreproducable error whereby SPI devices
act very, very strangly on boot. Tested on Efika MX over several years..
Signed-off-by: Matt Sealey <matt@genesi-usa.com> Acked-by: Stefano Babic <sbabic@denx.de>
This is useful for forcing the ROM's
usb downloader to activate upon a watchdog reset.
Or, you can boot from either SD Card.
Currently, support added for MX53 and MX6Q Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
Note: MX53 support untested. Acked-by: Stefano Babic <sbabic@denx.de>
Round up tick_to_time() value instead of truncating it. This avoids stopping
waits instantly for low usec values, and this generally guarantees that the code
always waits for at least the requested duration.
Since the input frequency of the API is a maximum that should not be exceeded in
order for the devices to operate properly, the SPI clock divider should be
rounded up, not truncated.
Signed-off-by: Benoît Thébaudeau <benoit.thebaudeau@advansee.com> Cc: Wolfgang Denk <wd@denx.de> Cc: Stefano Babic <sbabic@denx.de> Acked-by: Stefano Babic <sbabic@denx.de>
projects / karo-tx-uboot.git / log