Fabio Estevam [Mon, 7 May 2012 10:56:00 +0000 (10:56 +0000)]
mx5: Add workaround for ARM erratum ID 468414
Add the software workaround for ARM erratum ID 468414.
According to mx53/mx51 errata document:
"ENGcm11133 - ARM: NEON load data can be incorrectly forwarded to a
subsequent request
Description:
Under very specific set of conditions, data from a Neon load request can be incorrectly forwarded
to a subsequent, unrelated memory request.
The conditions are as follows:
• Neon loads and stores must be in use
• Neon L1 caching must be disabled
• Trustzone must be configured and in use
• The secure memory address space and the non-secure memory address space both use the same
physical addresses, either as an alias or the same memory location or for separate memory
locations
The issue is reported by ARM, erratum ID 468414, Category 2"
Dirk Behme [Wed, 8 Aug 2012 02:26:31 +0000 (02:26 +0000)]
mx6qsabrelite: enable DCache and MMC bounce buffer
The recent U-Boot version 2012.07 has improved drivers
(e.g. MMC and network/FEC) regarding DCache handling.
So it should be safe to use the DCache on the i.MX6, now.
Fabio Estevam [Mon, 15 Oct 2012 05:37:16 +0000 (05:37 +0000)]
mx5: lowlevel_init.S: Fix PLL settings for mx53
Currently PLL2 is not explicitely configured for mx53 and it runs at 333MHz.
Since PLL2 is the parent clock for DDR2, IPU, VPU, we should set it at 400MHz
instead.
Without doing so, it is not possible to use a 2.6.35 FSL kernel and display HDMI
at 1080p because the IPU clock cannot reach the requested frequency.
Set PLL2 to 400MHz, so that 1080p can be played and the DDR2 can run at its
maximum frequency.
Also, setup the other PLL's as done in FSL U-boot and re-arrange the code a little
bit to allow easier comparison with the original clock setup from FSL U-boot.
Eric Nelson [Wed, 3 Oct 2012 07:28:43 +0000 (07:28 +0000)]
i.MX6: mx6qsabrelite: Add splash screen support
Adds support for HDMI, two LVDS panels and one RGB panel to
the SABRE-Lite board.
Displays supported:
HDMI - 1024 x 768 for maximum compatibility
Hannstar-XGA - 1024 x 768 LVDS (Freescale part number MCIMX-LVDS1)
wsvga-lvds - 1024 x 600 LVDS (Boundary p/n Nit6X_1024x600)
wvga-rgb - 800 x 480 RGB (Boundary p/n Nit6X_800x480)
Since the ipuv3_fb display driver currently supports only a single display,
this code auto-detects panel by probing the HDMI Phy for Hot Plug Detect
or the I2C touch controller of the LVDS and RGB displays in the priority
listed above.
Setting 'panel' environment variable to one of the names above will
override auto-detection.
Signed-off-by: Eric Nelson <eric.nelson@boundarydevices.com>
On mxc, each SDHC instance has a dedicated clock, so gd->sdhc_clk is not
suitable for the multi-instance use case (initialization made directly with
fsl_esdhc_initialize()).
This patch fixes this issue by adding a configuration field for the SDHC input
clock frequency.
Signed-off-by: Benoît Thébaudeau <benoit.thebaudeau@advansee.com> Cc: Stefano Babic <sbabic@denx.de> Cc: Eric Bénard <eric@eukrea.com> Cc: Otavio Salvador <otavio@ossystems.com.br> Cc: Fabio Estevam <fabio.estevam@freescale.com> Cc: Jason Liu <r64343@freescale.com> Cc: Matt Sealey <matt@genesi-usa.com> Cc: Andy Fleming <afleming@gmail.com>
gd->sdhc_clk must also be set accordingly. This is good for the case only a
single SDHC instance is used (initialization made with fsl_esdhc_mmc_init()). A
future patch will fix the multi-instance use case (initialization made directly
with fsl_esdhc_initialize()).
Signed-off-by: Benoît Thébaudeau <benoit.thebaudeau@advansee.com> Cc: Stefano Babic <sbabic@denx.de> Cc: Eric Bénard <eric@eukrea.com> Cc: Otavio Salvador <otavio@ossystems.com.br>
Clean up mx35 lowlevel_init:
- Indent with tabs.
- Fix comments.
- Use defined values instead of literal constants.
- Use defined macros instead of duplicating code.
- Use macro parameters with default values instead of #define'd configs.
Marek Vasut [Sat, 6 Oct 2012 14:05:00 +0000 (14:05 +0000)]
kerneldoc: Implement "Example" section handling
The default kernel-doc strips starting spaces from every single
line in the Example section. This makes the code look bad. Thus,
implement special handling for this section.
Marek Vasut [Sat, 6 Oct 2012 14:04:59 +0000 (14:04 +0000)]
kerneldoc: Add nicer CSS stylesheet for HTML docs
Import basic CSS stylesheet for the HTML documentation. The base for
the stylesheet is taken from:
http://ds9a.nl/docbook/minimal-page.html
I customized the CSS a bit further, for example to add curvy corners
to example section and change the tint of gray. The HTML documentation
does not look that crude anymore.
Marek Vasut [Sat, 6 Oct 2012 14:04:58 +0000 (14:04 +0000)]
kerneldoc: Implant DocBook from Linux kernel
Pull slightly modified version of Documentation/DocBook, the related perl
script scripts/kernel-doc and the scripts/docproc.c from Linux kernel and
implant it into U-Boot. This will allow smooth generation of kerneldoc
style documentation.
It was necessary to modify the DocBook/Makefile to work with U-Boot build
system. The changes were only minor though and involved replacing the kbuild
specific parts.
It was also necessary to replace use of variables like KERNEL_VERSION with
U_BOOT_VERSION, strings like Linux kernel with U-Boot Bootloader etc. so
the generated result actually matches.
Finally, it was necessary to adjust docproc.c, since the documentation in
U-Boot is located in doc/DocBook instead of Documentation/DocBook as is in
case of the Linux kernel.
Some parts of the DocBook Makefile are unused, but to allow easier sync with
Linux kernel, these parts are still left in. The targets enabled now are
"htmldocs" "pdfdocs" "psdocs" "xmldocs" and "cleandocs" to remove the results
of documentation build.
Clean up mx25 lowlevel_init:
- Add comments.
- Do not use write32 repeatedly with the same value in order not to increase
code size.
- Make register values configurable.
- Use macro parameters with default values instead of literal constants.
- Use defined macros instead of duplicating code.
The CSI PODF bit-field used by the previous code for the i.MX31 CCM PDR0
register is actually composed of two bit-fields: one pre-divider and one
post-divider. This patch fixes the CCM access macros and the code using them
accordingly.
Optimize mx5 lowlevel_init.S code size:
- Compute values at compile time rather than at runtime where possible.
- Assign r4 to hold the zero value rather than setting registers to 0 again and
again.
- Associate a function to setup_pll rather than expanding its large macro code
multiple times.
- Allocate constant values in section only if used.
The MMC driver appropriate for the i.MX25 is fsl_esdhc, which has nothing to do
with mxcmmc.
Also, each eSDHC instance has a dedicated clock, so gd->sdhc_clk must be set
accordingly. This is good for the case only a single SDHC instance is used
(initialization made with fsl_esdhc_mmc_init()). A future patch will fix the
multi-instance use case (initialization made directly with
fsl_esdhc_initialize()).
Signed-off-by: Benoît Thébaudeau <benoit.thebaudeau@advansee.com> Cc: Stefano Babic <sbabic@denx.de> Cc: Eric Bénard <eric@eukrea.com> Cc: Otavio Salvador <otavio@ossystems.com.br>
The i.MX5 eSDHC clocks were considered as coming from the IPG clock although
they have dedicated clock paths.
Also, on i.MX5/6, each SDHC instance has a dedicated clock, so gd->sdhc_clk must
be set accordingly. This is good for the case only a single SDHC instance is
used (initialization made with fsl_esdhc_mmc_init()). A future patch will fix
the multi-instance use case (initialization made directly with
fsl_esdhc_initialize()).
Signed-off-by: Benoît Thébaudeau <benoit.thebaudeau@advansee.com> Cc: Stefano Babic <sbabic@denx.de> Cc: Eric Bénard <eric@eukrea.com> Cc: Otavio Salvador <otavio@ossystems.com.br>
The FEC clock does not come from PLL1, but from the IPG clock. The previous code
was even inconsistent with itself, returning the IPG clock as expected for
imx_get_fecclk(), but the PLL1 clock for mxc_get_clock(MXC_FEC_CLK).
The code handling the dividers was duplicated for each possible input clock, and
this function can benefit from the newly introduced get_standard_pll_sel_clk()
function instead of duplicating this mux handling code.
This function returned 66500000 instead of the correct lp_apm clock frequency if
the CCM.CSCMR1.uart_clk_sel mux is set to 3.
This patch fixes this issue by introducing the get_standard_pll_sel_clk()
function that will be used by future patches to handle identical muxes used by
many other clocks.
This fixes the "IPG PERCLK" frequency printed by the clocks command. The issue
was that get_ipg_per_clk() used periph_clk instead of lp_apm in the case
CCM.CBCMR.perclk_lp_apm_sel is set.
In the case periph_clk comes from periph_apm_clk, the latter is selected by the
CCM.CBCMR.periph_apm_sel mux, which can source the lp_apm clock from its
input ♯2. get_periph_clk() returned 0 instead of the lp_apm clock frequency in
this case.
If CCM.CCSR.lp_apm is set, the lp_apm clock is not necessarily 32768 Hz x 1024.
In that case:
- on i.MX51, this clock comes from the output of the FPM,
- on i.MX53, this clock comes from the output of PLL4.
The clock gate values are 2-bit bit-fields. Hence, setting or clearing only one
of these bits like what was done is wrong and can lead to unpredictable behavior
depending on the original value of these bit-fields.
Clean up the i.MX5 clock driver:
- Use readl() and writel() instead of their __raw_ counterparts.
- Use the clr/setbits_le32() family of macros rather than expanding code.
- Use accessor macros for bit-fields instead of _MASK and _OFFSET.
mx6qsabreauto is a board based on mx6q SoC with the following features:
- 2GB of DDR3
- 2 USB ports
- 1 HDMI output port
- SPI NOR
- 2 LVDS LCD ports
- Gigabit Ethernet
- Camera
- eMMC and SD card slot
- Multichannel Audio
- CAN
- SATA
- NAND
- PCIE
- Video Input
mx6qsabresd is a board based on mx6q SoC with the following features:
- 1GB of DDR3
- 1 USB OTG port
- 1 HDMI output port
- SPI NOR
- LVDS panel
- Gigabit Ethernet
- Camera Connector
- eMMC and SD card slot
- Audio
Lucas Stach [Sat, 29 Sep 2012 10:02:08 +0000 (10:02 +0000)]
tegra: clean up board include hell
The prototypes used in board files were all scattered out, which lead to
code duplication between SPL and normal U-Boot and some prototypes not actually
being used. Consolidate this in a common board header.
Signed-off-by: Lucas Stach <dev@lynxeye.de> Acked-by: Stephen Warren <swarren@nvidia.com> Signed-off-by: Tom Warren <twarren@nvidia.com>
Lucas Stach [Tue, 25 Sep 2012 20:21:14 +0000 (20:21 +0000)]
tegra20: rework UART GPIO handling
Rename board provided gpio_config_uart() to
gpio_early_init_uart() as it does the same thing as the equally
called function provided by the uart-switch code. This allows
to simply call this function in early board init whether or not
we are building with CONFIG_UART_SWITCH defined.
Also provide a weak symbol for this function, to avoid the
need to provide this function for boards that don't need any
fixup.
This patch supersedes the earlier posted
"tegra: convert gpio_config_uart to weak symbol".
Build tested with MAKEALL -s tegra20
Signed-off-by: Lucas Stach <dev@lynxeye.de> Acked-by: Simon Glass <sjg@chromium.org> Signed-off-by: Tom Warren <twarren@nvidia.com>
Lucas Stach [Tue, 25 Sep 2012 20:21:13 +0000 (20:21 +0000)]
tegra20: add clock_set_pllout function
Common practice on Tegra 2 boards is to use the pllp_out4 FO
to generate the ULPI reference clock. For this to work we have
to override the default hardware generated output divider.
This function adds a clean way to do so.
Signed-off-by: Lucas Stach <dev@lynxeye.de> Signed-off-by: Tom Warren <twarren@nvidia.com>
Lucas Stach [Tue, 25 Sep 2012 09:59:12 +0000 (09:59 +0000)]
tegra20: complete periph_id enum
Most Tegra boards output the ULPI reference clock on pad DEV2.
Complete the periph_id enum so that we are able to enable this
clock output circuit.
Signed-off-by: Lucas Stach <dev@lynxeye.de> Acked-by: Stephen Warren <swarren@wwwdotorg.org> Acked-by: Simon Glass <sjg@chromium.org> Signed-off-by: Tom Warren <twarren@nvidia.com>
Tom Warren [Wed, 19 Sep 2012 22:50:56 +0000 (15:50 -0700)]
Tegra20: Move some include files to arch-tegra for sharing with Tegra30
The move is pretty straight-forward. ap20.h and tegra20.h were renamed to ap.h and tegra.h.
Some files remain in arch-tegra20 but 'include' a file in 'arch-tegra' with #defines & structs
that will be common between T20 and T30 HW. HW-specific #defines, etc. stay in the 'arch-tegra20'
'root' file.
All boards build OK w/MAKEALL -s tegra20. Checkpatch.pl runs clean. Seaboard works OK.
Tom Warren [Wed, 19 Sep 2012 21:08:52 +0000 (14:08 -0700)]
Tegra20: Move some code files to common directories for upcoming Tegra30 patches.
Move files that are going to be common between T20 and T30 into 'tegra-common'
subdirs in AVP (arm720t), CPU (armv7), and shared (arch/arm/cpu/.) areas. Any
files that are left behind in '/tegra20' will be copied to '/tegra30' subdirs
and modified for that SoC. The 'common' files should need only minor changes.
Include files (arch/arm/include/asm/arch-tegra/tegra20) will be done in a
follow-on patch.
Builds fine w/MAKEALL -s tegra20. Checkpatch.pl is clean.
Medcom is the marketing name for an older, PXA-based version of the same
device. In order to avoid confusion, rename the Tegra-based version to
the new marketing name.
Signed-off-by: Thierry Reding <thierry.reding@avionic-design.de> Acked-by: Stephen Warren <swarren@nvidia.com> Signed-off-by: Tom Warren <twarren@nvidia.com>
Joe Hershberger [Wed, 3 Oct 2012 09:38:47 +0000 (09:38 +0000)]
tools/env: Remove unneeded complexity
The length included the name length, and then it was subtracted back
out on each use. Now we don't include it in the first place. Also
realloc as we process arguments and eliminate memset. Use memcpy
instead of manually copying each byte.
Signed-off-by: Joe Hershberger <joe.hershberger@ni.com>
Before this patch, there was a hard-coded env that was used as default
if the env in flash is detected as invalid. Now this tool (compiled
for a given board) will share the default env with the u-boot for the
board.
Fix include of config.h
Need to define "TEXT_BASE" when building the fw_env tool so that the
default env will be correct for environments which use it.
Define __ASSEMBLY__ when calling #include <config.h> so that we only
get #defines (all we're interested in).
Signed-off-by: Joe Hershberger <joe.hershberger@ni.com>
common: cmd_bdinfo: fix type of value in print_lnum
This fixes a warning when compiling with ELDK-5.2.1 for MIPS64:
cmd_bdinfo.c: In function 'print_lnum':
cmd_bdinfo.c:56:2: warning: format '%llX' expects argument of type 'long long unsigned int', but argument 3 has type 'u64' [-Wformat]
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Simon Glass [Thu, 27 Sep 2012 15:33:46 +0000 (15:33 +0000)]
patman: Handle checkpatch.pl not providing file/line info
Sometimes we don't get a valid filename or line number from checkpatch.pl,
for example if the patch is in a bad format. Deal with this by using a
default value, rather than a stack trace.
Simon Glass [Thu, 27 Sep 2012 15:06:02 +0000 (15:06 +0000)]
patman: Support Series-name tag to name a series
Sometimes it is possible to forget the name of the branch you used to
generate an upstream series. To assist with this, add an optional
patman does not use this.
Kim Phillips [Fri, 21 Sep 2012 12:28:17 +0000 (12:28 +0000)]
add check infrastructure, default sparse
Add support for running source code checkers on u-boot source, e.g.,
using sparse to aid with typechecking. This comes in especially
handy as SoC vendors mix and match cores and devices with different
endianness, thus here we add CHECK_ENDIAN to the otherwise linux
kernel default CHECKFLAGS.
Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
projects / u-boot / log