The DDR PHY training function, ddrphy_prepare_training() would not
work if compiled with GCC 4.9.

The struct ddrphy (arch/arm/include/asm/arch-uniphier/ddrphy-regs.h)
is specified with __packed because it represents a hardware register
mapping, but it turned out to cause a problem on GCC 4.9.

If -mno-unaligned-access is specified (yes, it is in
arch/arm/cpu/armv7/config.mk), GCC 4.9 is aware of the
__attribute__((packed)) and generates extra instructions to perform
the memory access in a way that does not cause unaligned access.
(Actually it is not need here because the register base, the first
argument of the ddrphy_prepare_training(), is always given with a
4-byte aligned address.)

Anyway, as a result, readl() / writel() is divided into byte-wise
accesses.  The problem is that this hardware only accepts 4-byte
register access.  Byte-wise accesses lead to unexpected behavior.

There are some options to avoid this problem.

[1] Remove -mno-unaligned-access
[2] Add __aligned(4) along with __packed to struct ddrphy
[3] Remove __packed from struct ddrphy

[1] solves the problem for ARMv7, but it does not for pre-ARMv6 and
ARMv6-M architectures where -mno-unaligned-access is default.
So, [1] does not seem reasonable in terms of code portability.

Both [2] and [3] work well, but [2] seems too much.  All the members
of struct ddrphy have the u32 type.  No padding would be inserted
even if __packed is dropped.

Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Reviewed-by: Tom Rini <trini@ti.com>

2 recent sunxi changes have removed the usage of lowlevel_init by moving some
code around and then setting CONFIG_SKIP_LOWLEVEL_INIT.
This is problematic for 2 reasons:

1) It does not just stop s_init from being called, it also stops
cpu_init_cp15 from getting called, which is undesirable.

2) We want u-boot.bin to be usable standalone, without SPL, some people e.g.
use an upstream u-boot.bin together with Allwinner's boot0 loader. So
u-boot.bin must (re)initialize the gpios, timer, etc.

This commit restores the lowlevel_init / s_init usage, while keeping the
changes to no longer use the global-data (gd) struct in the SPL.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>

Signed-off-by: Sinan Akman <sinan@writeme.com>
Cc: Tom Rini <trini@ti.com>

Signed-off-by: Sinan Akman <sinan@writeme.com>
Cc: kim.phillips@freescale.com

The global_data pointer (gd) has already been set before board_init_f()
is called. We should not assign it again. We should also not use gdata since
it is going away.

Signed-off-by: Simon Glass <sjg@chromium.org>

The global_data pointer (gd) has already been set before board_init_f()
is called. We should not assign it again. We should also not use gdata since
it is going away.

Signed-off-by: Simon Glass <sjg@chromium.org>
Acked-by: Stefano Babic <sbabic@denx.de>

The global_data pointer (gd) has already been set before board_init_f()
is called. We should not assign it again. We should also not use gdata since
it is going away.

Signed-off-by: Simon Glass <sjg@chromium.org>
Acked-by: Igor Grinberg <grinberg@compulab.co.il>
Tested-by: Nikita Kiryanov <nikita@compulab.co.il>
Acked-by: Nikita Kiryanov <nikita@compulab.co.il>

The global_data pointer (gd) has already been set before board_init_f()
is called. We should not assign it again. We should also not use gdata since
it is going away.

Signed-off-by: Simon Glass <sjg@chromium.org>

Wider bus widths (larger than default 1 bit) appeared in MMC standard
version 4.0. So, for MMC cards of any earlier version trying to change
the bus width (including ext_csd comparison) does not make any sense.
It may work incorrectly and at least cause unnecessary timeouts.
So, just skip the entire bus width related activity for earlier versions.

Signed-off-by: Andrew Gabbasov <andrew_gabbasov@mentor.com>
Tested-by: Alexey Brodkin <abrodkin@synopsys.com>

If all the commands switching an MMC card to 4- or 8-bit bus width fail,
and the bus width for the controller and the driver is still set
to default 1 bit, there is no need to send one more command to switch
the card to 1-bit bus width. Also, if the card or host controller do not
support wider bus widths, there is no need to send a switch command at all.

However, if one of switch commands succeeds, but the subsequent ext_csd
fields comparison fails, the card should be switched to some other bus width
(next in the list for the loop), or to default 1-bit bus width as a last
resort. That's why it would be incorrect to just remove the 1-bit bus width
case from the list, it should still be processed in some cases.

panto: Minor cosmetic edit removing superfluous parentheses.

Signed-off-by: Andrew Gabbasov <andrew_gabbasov@mentor.com>
Tested-by: Alexey Brodkin <abrodkin@synopsys.com>
Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com>

This extends the mmcinfo hardware partition info output to show
partitions with write reliability enabled with the "WRREL" string.
If the partition does not have write reliability enabled the "WRREL"
string is omitted; this is analogous to the ehhanced attribute.

Example output:

Device: OMAP SD/MMC
Manufacturer ID: fe
OEM: 14e
Name: MMC16
Tran Speed: 52000000
Rd Block Len: 512
MMC version 4.41
High Capacity: Yes
Capacity: 13.8 GiB
Bus Width: 4-bit
Erase Group Size: 8 MiB
HC WP Group Size: 16 MiB
User Capacity: 13.8 GiB ENH WRREL
User Enhanced Start: 0 Bytes
User Enhanced Size: 512 MiB
Boot Capacity: 16 MiB ENH
RPMB Capacity: 128 KiB ENH
GP1 Capacity: 64 MiB ENH WRREL
GP2 Capacity: 64 MiB ENH WRREL

Signed-off-by: Diego Santa Cruz <Diego.SantaCruz@spinetix.com>

This change extends the mmc hwpartition sub-command to change the
per-partition write reliability settings. It also changes the
syntax used for the enhanced user data area slightly to better
accomodate the write reliability option.

Signed-off-by: Diego Santa Cruz <Diego.SantaCruz@spinetix.com>

The eMMC partition write reliability settings are to be set while
partitioning a device, as per the eMMC spec, so changes to these
attributes needs to be done in the hardware partitioning API.
This commit adds such support.

Signed-off-by: Diego Santa Cruz <Diego.SantaCruz@spinetix.com>

Adds the mmc hwpartition sub-command to perform eMMC hardware
partitioning on an mmc device. The number of arguments can be
large for a complex partitioning, but as the partitioning has
to be done in one go it is difficult to make it simpler.

Signed-off-by: Diego Santa Cruz <Diego.SantaCruz@spinetix.com>

This adds an API to do hardware partitioning on eMMC devices. The
new mmc_hwpart_config() function does the partitioning in one go.
As the different attributes and partitioning options on eMMC may
be interdependent validation has to be done based on the complete
partitioning configuration. The function accepts three modes:

- MMC_HWPART_CONF_CHECK: just validates that the configuration
  is valid.
- MMC_HWPART_CONF_SET: validates and sets all the fields in
  EXT_CSD but without setting the "partitioning completed" bit,
  and thus is reversible.
- MMC_HWPART_CONF_COMPLETE: does everything and is thus not
  reversible.

Signed-off-by: Diego Santa Cruz <Diego.SantaCruz@spinetix.com>

The mmc_startup() function uses the ext_csd data even if reading it
from the mmc device failed. This bug was introduced in commit
bc897b1d. We now bail out if
reading it fails, this should not be a problem as ext_csd was
introduced in MMC 4.0 and this code is conditional on MMC >= 4.0.

Signed-off-by: Diego Santa Cruz <Diego.SantaCruz@spinetix.com>

The eMMC spec says that partitioning is only effective after the
PARTITION_SETTING_COMPLETED is set in EXT_CSD (and a power cycle was done,
but that we cannot know). Thus the partition sizes and attributes should
be ignored when that bit is not set, otherwise the various capacities
are not coherent (e.g., the user data capacity will be that of the
unpartitioned device while partition sizes would be non-zero).

Prescence of non-zero partitioning data is nevertheless still used to
activate the high-capacity size definitions (EXT_CSD_ERASE_GROUP_DEF)
as it is necessary to set that to write any of the partitioning fields
in EXT_CSD, so having partitioning data means someone previously
activated that and we should keep it activated.

Signed-off-by: Diego Santa Cruz <Diego.SantaCruz@spinetix.com>

This adds the erase group size and high-capacity WP group size to
mmcinfo's output. The erase group size is necessary to properly align
erase requests on eMMC. The high-capacity WP group size is necessary
to properly align partitions on eMMC.

Signed-off-by: Diego Santa Cruz <Diego.SantaCruz@spinetix.com>

Read the eMMC high capacity write protect group size at mmc device
initialization. This is useful to correctly partition an eMMC device,
as partitions need to be aligned to this size.

Signed-off-by: Diego Santa Cruz <Diego.SantaCruz@spinetix.com>

The erase_grp_size in struct mmc is to be a size in 512-byte sectors
but the code used to compute it for eMMC when EXT_CSD_ERASE_GROUP_DEF is
enabled computed it as bytes, leading to erase sizes and alignment
much larger than what is actually required by the mmc device.

Signed-off-by: Diego Santa Cruz <Diego.SantaCruz@spinetix.com>

This adds output to show the eMMC enhanced user data area size and offset
along with the partition sizes in mmcinfo's output.

Signed-off-by: Diego Santa Cruz <Diego.SantaCruz@spinetix.com>

This modification reads the size of the eMMC enhanced user data area
upon initialization of an mmc device, it will be used later by
mmcinfo.

Signed-off-by: Diego Santa Cruz <Diego.SantaCruz@spinetix.com>

Signed-off-by: Diego Santa Cruz <Diego.SantaCruz@spinetix.com>

The eMMC spec mandates that the high-capacity group size definitions
should be enabled when the device is partitioned (by setting
ERASE_GROUP_DEF in EXT_CSD). The current test to determine when this is
required misses a few cases. In particular a device may have been
partitioned without setting the enhanced attribute on any partition
or partitioning may be completed without creating any extra partitions.

This change moves the code to set ERASE_GROUP_DEF to after reading
all partition information. It is also enabled when
PARTITIONING_SETTING_COMPLETED is set as it is necessary to enable
ERASE_GROUP_DEF before setting that bit, so it means that the user
previously switched to the high capacity definitions.

Signed-off-by: Diego Santa Cruz <Diego.SantaCruz@spinetix.com>

eMMC partitions are defined as of eMMC 4.41, but mmcinfo process
partition info for eMMC >= 4.0, change it to do it for >= 4.41

Signed-off-by: Diego Santa Cruz <Diego.SantaCruz@spinetix.com>

The eMMC spec numbers general purpose partitions starting at 1, but
the mmcinfo output follows the internal numbering which starts at 0.
Make the mmcinfo command output number partitions as in the eMMC
spec to avoid confusion.

Signed-off-by: Diego Santa Cruz <Diego.SantaCruz@spinetix.com>

This extends the mmcinfo command's output to show which eMMC partitions
have the enhanced attribute set. Note that the eMMC spec says that
if the enhanced attribute is supported then the boot and RPMB
partitions are of the enhanced type.

The output of mmcinfo becomes:
Device: OMAP SD/MMC
Manufacturer ID: fe
OEM: 14e
Name: MMC16
Tran Speed: 52000000
Rd Block Len: 512
MMC version 4.41
High Capacity: Yes
Capacity: 13.8 GiB
Bus Width: 4-bit
User Capacity: 13.8 GiB ENH
Boot Capacity: 16 MiB ENH
RPMB Capacity: 128 KiB ENH
GP1 Capacity: 64 MiB ENH
GP2 Capacity: 64 MiB ENH

Signed-off-by: Diego Santa Cruz <Diego.SantaCruz@spinetix.com>

There is currently no command that will provide an overview of the hardware
partitions present on an eMMC device, one has to switch to every partition
via "mmc dev" and run mmcinfo for each to get the partition's capacity.
This commit adds a few lines of output to mmcinfo with the sizes of the
present partitions, like this:

Device: OMAP SD/MMC
Manufacturer ID: fe
OEM: 14e
Name: MMC16
Tran Speed: 52000000
Rd Block Len: 512
MMC version 4.41
High Capacity: Yes
Capacity: 13.8 GiB
Bus Width: 4-bit
User Capacity: 13.8 GiB
Boot Capacity: 16 MiB
RPMB Capacity: 128 KiB
GP1 Capacity: 64 MiB
GP2 Capacity: 64 MiB

panto: Minor edit removing superfluous parentheses.

Signed-off-by: Diego Santa Cruz <Diego.SantaCruz@spinetix.com>
Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com>

This adds Renesas rmobile ARM SoC's SD/MMC host support.
This drivers tested with Gose board and Koelsch board.

Signed-off-by: Nobuhiro Iwamatsu <nobuhiro.iwamatsu.yj@renesas.com>
Acked-by: Pantelis Antoniou <panto@antoniou-consulting.com>

This patch enables U-Boot to modify the MAC address of the AX88179.
Tested on RECS5250 (similar to Arndale5250)

Signed-off-by: Rene Griessl <rgriessl@cit-ec.uni-bielefeld.de>

Add interrupt queue support, so that a usb keyboard can be used without
causing huge latencies.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>

Make construct_urb take an urb and hep parameter, rather then having it always
operate on the file global urb and hep structs. This is a preperation patch
for adding interrupt queue support.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>

If a transfer / urb times-out, properly remove it from the schedule, rather
then letting it sit on the ep head. This stops the musb code from getting
confused and refusing to queue further transfers after a timeout.

Tested by unplugging a usb-keyboard, replugging it and doing a usb-reset,
before this commit the keyboard would not work after the usb-reset.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>

This commit fixes a number of issues with the reset sequence of musb-new
in host mode:

1) Our usb device probe relies on a second device reset being done after the
first descriptors read. Factor the musb reset code into a usb_reset_root_port
function (and add this as an empty define for other controllers), and call
this when a device has no parent.

2) Just like with normal usb controllers there needs to be a delay after
reset, for normal usb controllers, this is handled in hub_port_reset, add a
delay to usb_reset_root_port.

3) Sync the musb reset sequence with the upstream kernel, clear all bits of
power except bits 4-7, and increase the time reset is asserted to 50 ms.

With these fixes an usb keyboard I have now always enumerates properly, where
as earlier it would only enumerare properly once every 5 tries.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>

For bulk and ctrl transfers common/usb.c sets udev->status = USB_ST_NOT_PROC,
but it does not do so for interrupt transfers.

musb_uboot.c: submit_urb() however was waiting for USB_ST_NOT_PROC to become 0,
and thus without anyone setting USB_ST_NOT_PROC would exit immediately for
interrupt urbs, returning the urb status of EINPROGRESS as error.

This commit fixes this, thereby also making usb_kbd.c work together with
musb_new and CONFIG_SYS_USB_EVENT_POLL.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>

CPU cycle based timeouts are no good, because how long they use depends on
CPU speed. Instead use time based timeouts, and wait one second for a
device connection to show up (per the USB spec), and wait USB_TIMEOUT_MS
for various urbs to complete.

This fixes "usb start" taking for ever when no device is plugged into the
otg port.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>

This is based on Jussi Kivilinna's work for the linux-sunxi-3.4 kernel to use
the kernels musb driver instead of Allwinners own custom driver.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>