After waiting for the command completion event, the interrupt status
bits, that occured to be set by that time, are cleared by writing them
back. It is supposed, that it should be command related bits (command
complete and may be command errors).

However, in some cases the DMA already completes by that time before
the full transaction completes. The corresponding DINT bit gets set
and then cleared before even entering the loop, waiting for data part
completion. That waiting loop never gets this bit set, causing the
operation to hang. This is reported to happen, for example, for write
operation of 1 sector to upper area (block #7400000) of SanDisk Ultra II
8GB card.

The solution could be to explicitly clear only command related interrupt
status bits. However, since subsequent processing does not rely on
any command bits state, it could be easier just to remove clearing
of any bits at that point, leaving them all until all data processing
completes. After that the whole register will be cleared at once.

Also, on occasion, interrupts masking moved to before writing the command,
just for the case there should be no chance of interrupt between the first
command and interrupts masking.
Reported-by: NDirk Behme <dirk.behme@de.bosch.com>
Signed-off-by: NAndrew Gabbasov <andrew_gabbasov@mentor.com>
Acked-by: NDirk Behme <dirk.behme@de.bosch.com>
Signed-off-by: NAndy Fleming <afleming@freescale.com>

Since commit 48e0b2bd (powerpc/esdhc: Correct judgement for DATA PIO mode)
we see mx6 systems to hang after doing a 'save' command.

Revert this commit since the original 'ifdef' logic from 7b43db92
(drivers/mmc/fsl_esdhc.c: fix compiler warnings) was the correct one.
Reported-by: NTapani Utriainen <tapani@technexion.com>
Signed-off-by: NFabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: NAndy Fleming <afleming@freescale.com>

This patch fixes a bug related to mmc writes.

When doing fatwrites on an SD-Card, MMC bus problems can occur. Depending
on the size of the file, "MMC0: Bus busy timeout!" is reported, resulting
in an SD-Card that is no longer responding.
It appears to be, that set_cluster can be called with a size being zero.
That can be with a file that has a size being an exact multiple
(including 0) of the clustersize, but also for files that are smaller than
the size of one cluster.
The same problem occurs if the "mmc write" command is given with a block
count being 0.

By adding a check for the block count being zero in mmc_write_blocks
(drivers/mmc.c), this problem is solved.
Signed-off-by: NRuud Commandeur <rcommandeur@clb.nl>
Cc: Tom Rini <trini@ti.com>
Cc: Benoît Thébaudeau <benoit.thebaudeau@advansee.com>
Cc: Mats Karrman <Mats.Karrman@tritech.se>
Cc: Andy Fleming <afleming@gmail.com>
Signed-off-by: NAndy Fleming <afleming@freescale.com>

CAP register don't have any information for 8-bit buswidth support
on 2.0 sdhci spec, only from 3.0 onwards bit[18] got this information.

Due to this misassignment in sdhci, mmc is setting 8-bit buswidth using
mmc_set_bus_width even if controller doesn't support.
Below change has code information.
"mmc: Properly determine maximum supported bus width"
(sha1: 7798f6db)

Bug log: <mmc plus and emmc cards)
-------
zynq-uboot> mmcinfo
Error detected in status(0x208100)!
Device: zynq_sdhci
Manufacturer ID: fe
.....

So enable 8-bit support only for 3.0 spec using CAP and for below 3.0
assign mmc->host_caps = MMC_MODE_8BIT on respective platform driver
if host have a support.
Signed-off-by: NJagannadha Sutradharudu Teki <jaganna@xilinx.com>
Signed-off-by: NAndy Fleming <afleming@freescale.com>

The commit d196bd88 (env_mmc: add support for redundant environment)
introduce the following compile error when enable redundant
environment support with MMC
---8<---
env_mmc.c:149: error: 'env_t' has no member named 'flags'
env_mmc.c:248: error: 'env_t' has no member named 'flags'
env_mmc.c:248: error: 'env_t' has no member named 'flags'
env_mmc.c:250: error: 'env_t' has no member named 'flags'
env_mmc.c:250: error: 'env_t' has no member named 'flags'
env_mmc.c:252: error: 'env_t' has no member named 'flags'
env_mmc.c:252: error: 'env_t' has no member named 'flags'
env_mmc.c:254: error: 'env_t' has no member named 'flags'
env_mmc.c:254: error: 'env_t' has no member named 'flags'
env_mmc.c:267: error: 'env_t' has no member named 'flags'
make[1]: *** [env_mmc.o] Error 1
--->8---

Add this patch to fix it
Signed-off-by: NBo Shen <voice.shen@atmel.com>
Reviewed-by: NMichael Heimpold <mhei@heimpold.de>
Signed-off-by: NAndy Fleming <afleming@freescale.com>

The flash_info_t->start[] field is limited in size by CONFIG_SYS_MAX_FLASH_SECT
macro, which is set to 19 for this board in the board config file. If we inspect
the board/ppmc7xx/flash.c closely, especially the flash_get_size() function, we
can notice the "switch ((long)flashtest)" at around line 80 having a few results
which will set flash_info_t->sector_count to value higher than 19, for example
"case AMD_ID_LV640U" will set it to 128. Notice that right underneath, iteration
over flash_info_t->start[] happens and the upper bound for the interation is
flash_info_t->sector_count. Now if the sector_count is 128 as it is for the
AMD_ID_LV640U case, but the CONFIG_SYS_MAX_FLASH_SECT limiting the start[] is
only 19, an access past the start[] array much happen. Moreover, during this
iteration, the field is written to, so memory corruption is inevitable.
Signed-off-by: NMarek Vasut <marex@denx.de>
Cc: Wolfgang Denk <wd@denx.de>
Cc: Tom Rini <trini@ti.com>
Cc: Richard Danter <richard.danter@windriver.com>

C99's strict aliasing rules are insane to use in low-level code such as a
bootloader, but as Wolfgang has rejected -fno-strict-aliasing in the
past, add a union so that 16-bit accesses can be performed.

Compile-tested only.
Signed-off-by: NScott Wood <scottwood@freescale.com>
Acked-by: NWolfgang Denk <wd@denx.de>

This fixes several warnings like

In file included from ./u-boot/include/linux/mtd/mtd.h:13:0,
                 from env_onenand.c:37:
./u-boot/build/vct_platinumavc_onenand_small/include2/asm/errno.h:52:0: warning: "ENOMSG" redefined [enabled by default]
Signed-off-by: NDaniel Schwierzeck <daniel.schwierzeck@gmail.com>

The purpose of the __raw* IO accessors is to provide
IO access in native-endian order. However in the current
MIPS implementation, the 16 and 32 bit variants of the
__raw accessors are swapping the values on big-endian
systems if the CONFIG_SWAP_IO_SPACE option is enabled.

The patch changes the IO accessor macros to fix this
broken behaviour.
Signed-off-by: NGabor Juhos <juhosg@openwrt.org>
Cc: Daniel Schwierzeck <daniel.schwierzeck@googlemail.com>

The pci_indirect.c file is always compiled when
CONFIG_PCI is defined although the indirect PCI
bridge support is not needed by every board.

Introduce a new CONFIG_PCI_INDIRECT_BRIDGE
config option and only compile indirect PCI
bridge support if this options is enabled.

Also add the new option into the configuration
files of the boards which needs that.

Compile tested for powerpc, x86, arm and nds32.
MAKEALL results:

powerpc:
  --------------------- SUMMARY ----------------------------
  Boards compiled: 641
  Boards with warnings but no errors: 2 ( ELPPC MPC8323ERDB )
  ----------------------------------------------------------
  Note: the warnings for ELPPC and MPC8323ERDB are present even
  without the actual patch.

x86:
  --------------------- SUMMARY ----------------------------
  Boards compiled: 1
  ----------------------------------------------------------

arm:
  --------------------- SUMMARY ----------------------------
  Boards compiled: 311
  ----------------------------------------------------------

nds32:
  --------------------- SUMMARY ----------------------------
  Boards compiled: 3
  ----------------------------------------------------------

Cc: Tom Rini <trini@ti.com>
Cc: Daniel Schwierzeck <daniel.schwierzeck@googlemail.com>
Signed-off-by: NGabor Juhos <juhosg@openwrt.org>

Some ARM compilers may emit code that makes unaligned accesses when
faced with constructs such as:

char mac[16] = "ethaddr";

Replace this with a strcpy() call instead to avoid this. strcpy() is
used here, rather than replacing all usage of the mac variable with the
string itself, since the loop itself sprintf()s to the variable each
iteration, so strcpy() is doing basically the same thing.

Reported-by: Florian Meier
Signed-off-by: NStephen Warren <swarren@wwwdotorg.org>

This commit refactors common/board_f.c and common/board_r.c
in order to delete the dest_addr and dest_addr_sp from
gd_t struct.

As mentioned as follows in include/asm-generic/global_data.h,

  /* TODO: is this the same as relocaddr, or something else? */
  unsigned long dest_addr;        /* Post-relocation address of U-Boot */

dest_addr is the same as relocaddr.
Likewise, dest_addr_sp is the same as start_addr_sp.

It seemed dest_addr/dest_addr_sp was used only as a scratch variable
to calculate relocaddr/start_addr_sp, respectively.

With a little refactoring, we can delete dest_addr and dest_addr_sp.
Signed-off-by: NMasahiro Yamada <yamada.m@jp.panasonic.com>
Cc: Simon Glass <sjg@chromium.org>

Jump into full u-boot mode if a 'c' character is received on the uart.

We need to adjust the spl bss/malloc area to not overlap with the
loadaddr of the kernel (sdram + 32k), so move it past u-boot instead.

For raw mmc, we store the kernel parameter area in the free space after
the MBR (if used). For nand, we use the last sector of the partition
reserved for u-boot.

This also enables the spl command in the full u-boot so the kernel
parameter area snapshot can be created.
Signed-off-by: NPeter Korsgaard <peter.korsgaard@barco.com>

If Falcon mode support is enabled (and the system isn't directed into
booting u-boot), it will instead try to load kernel from sector
CONFIG_SYS_MMCSD_RAW_MODE_KERNEL_SECTOR and
CONFIG_SYS_MMCSD_RAW_MODE_ARGS_SECTORS of kernel argument parameters
starting from sector CONFIG_SYS_MMCSD_RAW_MODE_ARGS_SECTOR.
Signed-off-by: NPeter Korsgaard <peter.korsgaard@barco.com>

So we can use it for falcon mode as well.
Signed-off-by: NPeter Korsgaard <peter.korsgaard@barco.com>

If Falcon mode support is enabled (and the system isn't directed into
booting u-boot), it will instead try to load kernel from
CONFIG_SPL_FAT_LOAD_KERNEL_NAME file and kernel argument parameters from
CONFIG_SPL_FAT_LOAD_ARGS_NAME, both from the same partition as u-boot.
Signed-off-by: NPeter Korsgaard <peter.korsgaard@barco.com>

Signed-off-by: NTom Rini <trini@ti.com>

So we can use it for falcon mode as well.
Signed-off-by: NPeter Korsgaard <peter.korsgaard@barco.com>

So we can instead fallback to doing something else on errors.
Signed-off-by: NPeter Korsgaard <peter.korsgaard@barco.com>

[trini: Applied v1 of the series rather than v2, this commit is the
delta from v1 to v2]
Signed-off-by: NStephen Warren <swarren@nvidia.com>
Signed-off-by: NTom Rini <trini@ti.com>

Signed-off-by: NTom Rini <trini@ti.com>

The location of valid scratch space is dependent on SoC, so move that
there.  On OMAP4+ we continue to use SRAM_SCRATCH_SPACE_ADDR.  On
am33xx/ti814x we want to use what the ROM defines as "public stack"
which is the area after our defined download image space.  Correct the
comment about and location of CONFIG_SPL_TEXT_BASE.
Signed-off-by: NTom Rini <trini@ti.com>

Add a DT simple-framebuffer node to DT when booting the Linux kernel.
This will allow the kernel to inherit the framebuffer configuration from
U-Boot, and display a graphical boot console, and even run a full SW-
rendered X server.
Signed-off-by: NStephen Warren <swarren@wwwdotorg.org>
Acked-by: NSimon Glass <sjg@chromium.org>

simple-framebuffer is a new device tree binding that describes a pre-
configured frame-buffer memory region and its format. The Linux kernel
contains a driver that supports this binding. Implement functions to
create a DT node (or fill in an existing node) with parameters that
describe the framebuffer format that U-Boot is using.

This will be immediately used by the Raspberry Pi board in U-Boot, and
likely will be used by the Samsung ARM ChromeBook support soon too. It
could well be used by many other boards (e.g. Tegra boards with built-in
LCD panels, which aren't yet supported by the Linux kernel).
Signed-off-by: NStephen Warren <swarren@wwwdotorg.org>
Acked-by: NSimon Glass <sjg@chromium.org>

We need to call the save_omap_boot_params function on am33xx/ti81xx and
other newer TI SoCs, so move the function to boot-common.  Only OMAP4+
has the omap_hw_init_context function so add ifdefs to not call it on
am33xx/ti81xx.  Call save_omap_boot_params from s_init on am33xx/ti81xx
boards.
Reviewed-by: NR Sricharan <r.sricharan@ti.com>
Signed-off-by: NTom Rini <trini@ti.com>

Prior to Sricharan's cleanup of the boot parameter saving code, we
did not make use of NON_SECURE_SRAM_START on am33xx, so it wasn't a
problem that the address was pointing to the middle of our running SPL.
Correct to point to the base location of the download image area.
Increase CONFIG_SPL_TEXT_BASE to account for this scratch area being
used.  As part of correcting these tests, make use of the fact that
we've always been placing our stack outside of the download image area
(which is fine, once the downloaded image is run, ROM is gone) so
correct the max size test to be the ROM defined top of the download area
to where we link/load at.
Signed-off-by: NTom Rini <trini@ti.com>

---
Changes in v2:
- Fix typo noted by Peter Korsgaard

Only called in this file, mark as static.
Signed-off-by: NTom Rini <trini@ti.com>

This can be useful to force bootcmd to execute as soon as U-Boot has
started.

My use-case is: An SoC-specific tool pushes U-Boot into RAM, along with
an image to be written to device boot flash, with the DT config property
"bootcmd" set to contain a command to write that image to flash. In this
scenario, we don't want to allow any stale bootdelay value taken from
the current flash content to affect how long it takes before the
flashing process starts.
Signed-off-by: NStephen Warren <swarren@nvidia.com>
Acked-by: NSimon Glass <sjg@chromium.org>
Acked-by: NGerald Van Baren <vanbaren@cideas.com>

We know the exact property names that the code wants to process. Look
these up directly with fdt_get_property(), rather than iterating over
all properties within the node, and checking each property's name, in
a convoluted fashion, against the expected name.
Signed-off-by: NStephen Warren <swarren@nvidia.com>

Initialized character arrays on the stack can cause gcc to emit code that
performs unaligned accessess. Make the data static to avoid this.

Note that the unaligned accesses are made when copying data to prefix[] on
the stack from .rodata. By making the data static, the copy is completely
avoided. All explicitly written code treats the data as u8[], so will never
cause any unaligned accesses.
Signed-off-by: NStephen Warren <swarren@nvidia.com>
Acked-by: NSimon Glass <sjg@chromium.org>

The generic-board board_init_f function called board_postclk_init twice.

The first one came from arch/arm/lib/board.c, while the second one
from arch/powerpc/lib/board.c.

This commit deletes the first occurrence.
In addition, the second get_clocks call is moved after
board_postclk_init in order to keep the function call order
both for ARM and PowerPC.
ARM board calles get_clocks function after board_postclk_init.
Signed-off-by: NMasahiro Yamada <yamada.m@jp.panasonic.com>

Make sure to never access beyond bounds of either EFI partition name
or DOS partition name. This situation is happening:

part.h:     disk_partition_t->name is 32-byte long
part_efi.h: gpt_entry->partition_name is 36-bytes long

The loop in part_efi.c copies over 36 bytes and thus accesses beyond
the disk_partition_t->name .

Fix this by picking the shortest of source and destination arrays and
make sure the destination array is cleared so the trailing bytes are
zeroed-out and don't cause issues with string manipulation.
Signed-off-by: NMarek Vasut <marex@denx.de>
Cc: Tom Rini <trini@ti.com>
Cc: Simon Glass <sjg@chromium.org>

The image code is fairly complex with various different options. It would
be useful to have comprehensive tests for this.

As a start, create a script which tries out loading a kernel/ramdisk/fdt
from a FIT and checks that the images appear in the right place in memory.

This uses sandbox which now supports bootm and related features.
Signed-off-by: NSimon Glass <sjg@chromium.org>

Now that the code for loading these three images from a FIT is common, we
don't need individual boostage IDs for each of them.

Note: there are some minor changes in the bootstage numbering, particuarly
for kernel loading. I don't believe this matters.
Signed-off-by: NSimon Glass <sjg@chromium.org>

Use map_sysmem() to convert from address to pointer, so that sandbox can
print FIT information without crashing.
Signed-off-by: NSimon Glass <sjg@chromium.org>