A few changes are made to the Tegra EHCI driver so that it can set
everything up for device-mode operation on the first USB controller.
This can be used in conjunction with ci_udc.c to operate as a USB
device.
Detailed changes are:
* Rename set_host_mode() to set_up_vbus() since that's really what it
does.
* Modify set_up_vbus() to know whether it's initializing in host or
device mode, and:
- Skip the external VBUS check in device mode, since external VBUS is
expected in this case.
- Disable VBUS output in device mode.
* Modify init_phy_mux() to know whether it's initializing in host or
device mode, and hence skip setting USBMODE_CM_HC (which enables host
mode) in device mode. See the comments in that function for why this
is safe w.r.t. the ordering requirements of PHY selection.
* Modify init_utmi_usb_controller() to force "b session valid" in device
mode, since the HW requires this. This is done in UTMI-specific code,
since we only support device mode on the first USB controller, and that
controller can only talk to a UTMI PHY.
* Enhance ehci_hcd_init() to error-check the requested host-/device-mode
vs. the dr_mode (dual-role mode) value present in device tree, and the
HW configurations which support device mode.
* Enhance ehci_hcd_init() not to skip HW initialization when switching
between host and device mode on a controller. This requires remembering
which mode the last initialization used.
Cc: Jim Lin <jilin@nvidia.com>
Cc: Stefan Agner <stefan@agner.ch>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
The HW-defined procedure for booting Tegra requires that
CLAMP_INPUTS_WHEN_TRISTATED be enabled before programming the pinmux.
Add a function to the pinmux driver to allow boards to do this.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
The HW-defined procedure for booting Tegra requires that some pins be
set up as GPIOs immediately at boot in order to avoid glitches on those
pins, when the pinmux is programmed. Add a feature to the GPIO driver
which executes a GPIO configuration table. Board files will use this to
implement the correct HW initialization procedure.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
If CONFIG_API is ever to be enabled on Tegra, this define must be set,
since api/api_storage.c uses it.
A couple of annoyting things about CONFIG_SYS_MMC_MAX_DEVICE
1) It isn't documented in README. The same is true for a lot of similar
defines used by api_storage.c.
2) It doesn't represent MAX_DEVICE but rather NUM_DEVICES, since the
valid values are 0..n-1 not 0..n.
However, I this patch does not address those shortcomings.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Use smaller fields in the Tegra pinmux structures in order to pack the
data tables into a smaller space. This saves around 1-3KB for the SPL
and around 3-8KB for the main build of U-Boot, depending on the board,
which SoC it uses, and how many pinmux table entries there are.
In order to pack PMUX_FUNC_* into a smaller space, don't hard-code the
values of PMUX_FUNC_RSVD* to values which require 16 bits to store them,
but instead let their values be assigned automatically, so they end up
fitting into 8 bits.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Acked-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Combine the Tegra USB header file into one header file for all SoCs.
Use ifdef to account for the difference, especially Tegra20 is quite
different from newer SoCs. This avoids duplication, mainly for
Tegra30 and newer devices.
Reviewed-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Stefan Agner <stefan@agner.ch>
Signed-off-by: Tom Warren <twarren@nvidia.com>
On Tegra30 and later, the PTS (parallel transceiver select) and STS
(serial transceiver select) are part of the HOSTPC1_DEVLC_0 register
rather than PORTSC1_0 register. Since the reset configuration
usually matches the intended configuration, this error did not show
up on Tegra30 devices.
Also use the slightly different bit fields of first USB, (USBD) on
Tegra20 and move those definitions to the Tegra20 specific header
file.
Reviewed-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Stefan Agner <stefan@agner.ch>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Clean up the naming of pinmux-related objects:
* Refer to drive groups rather than pad groups to match the Linux kernel.
* Ensure all pinmux API types are prefixed with pmux_, values (defines)
are prefixed with PMUX_, and functions prefixed with pinmux_.
* Modify a few type names to make their content clearer.
* Minimal changes to SoC-specific .h/.c files are made so the code still
compiles. A separate per-SoC change will be made immediately following,
in order to keep individual patch size down.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Acked-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Remove a few unused functions from the pinmux header. They aren't
currently used, and removing them prevents any new usage from appearing.
This will ease moving to just pinmux_config_table() and
padgrp_config_table() in the future.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Acked-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Much of arch/arm/cpu/tegra*-common/pinmux.c is identical. Remove the
duplication by creating pinmux-common.c for all the identical code.
This leaves:
* arch/arm/include/asm/arch-tegra*/pinmux.h defining only the names of
the various pins/pin groups, drive groups, and mux functions.
* arch/arm/cpu/tegra*-common/pinmux.c containing only the lookup table
stating which pin groups support which mux functions.
The code in pinmux-common.c is semantically identical to that in the
various original pinmux.c, but had some consistency and cleanup fixes
applied during migration.
I removed the definition of struct pmux_tri_ctlr, since this is different
between SoCs (especially Tegra20 vs all others), and it's much simpler to
deal with this via the new REG/MUX_REG/... defines. spl.c, warmboot.c,
and warmboot_avp.c needed updates due to this, since they previously
hijacked this struct to encode the location of some non-pinmux registers.
Now, that code simply calculates these register addresses directly using
simple and obvious math. I like this method better irrespective of the
pinmux code cleanup anyway.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Acked-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Tegra's "APB misc" register region contains various miscellaneous
registers and the Tegra pinmux registers. Some code that touches the
misc registers currently uses struct pmux_tri_ctlr, which is intended to
be a definition of pinmux registers, rather than struct apb_misc_pp_ctrl,
which is intended to be a definition of the miscellaneous registers.
Convert all such code to use struct apb_misc_pp_ctrl, since struct
pmux_tri_ctlr goes away in the next patch.
This requires adding a missing field definition to struct
apb_misc_pp_ctrl, and moving the header into a more common location.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Acked-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Tom Warren <twarren@nvidia.com>
pinmux_init() is a board-level function, not a pinmux driver function.
Move the prototype to a board header rather than the driver header.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Acked-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Tom Warren <twarren@nvidia.com>
The way that struct mmc was implemented was a bit of a mess;
configuration and internal state all jumbled up in a single structure.
On top of that the way initialization is done with mmc_register leads
to a lot of duplicated code in drivers.
Typically the initialization got something like this in every driver.
struct mmc *mmc = malloc(sizeof(struct mmc));
memset(mmc, 0, sizeof(struct mmc);
/* fill in fields of mmc struct */
/* store private data pointer */
mmc_register(mmc);
By using the new mmc_create call one just passes an mmc config struct
and an optional private data pointer like this:
struct mmc = mmc_create(&cfg, priv);
All in tree drivers have been updated to the new form, and expect
mmc_register to go away before long.
Changes since v1:
* Use calloc instead of manually calling memset.
* Mark mmc_register as deprecated.
Signed-off-by: Pantelis Antoniou <panto@antoniou-consulting.com>
Tegra124 moved the CSITE block's base address. Fix U-Boot to use
the correct address.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Register pmc_pwrgate_timer_mult has a different layout on Tegra114 and
Tegra124. Reflect this in pmc.h.
Also, simply write the whole of the register in start_cpu() rather than
doing a read-modify-write; the register is simple enough that the code
can easily construct the entire desired value.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
These headers define the Tegra124 hardware. Add them to the usual
place.
Add Tegra124 chip ID/SKU ID definitions to common headers.
There's no real HW change on Tegra124 for 90% of the toys, so it might
make sense for a future patch to unify some of the content of these
files in a common location.
Signed-off-by: Tom Warren <twarren@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Tested-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Pass just the partition ID to power_partition(), rather than also passing
the partition's status register mask too. This makes it simpler to get
call-sites correct, since they don't need to pass two different values
that define the same thing and must match.
Consequently, we can remove the mask definitions from pmc.h.
Suggested-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Use a named constant for the PLL lock bit in enable_cpu_clocks().
Construct the complete value of pmc_pwrgate_toggle, rather than doing a
read-modify-write; the register is simple enough and doesn't need to
maintain state between operations.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Tested-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Based on the Tegra TRM, the system clock (which is the AVP clock) can
run up to 275MHz. On power on, the default sytem clock source is set to
PLLP_OUT0. In function clock_early_init(), PLLP_OUT0 will be set to
408MHz which is beyond system clock's upper limit.
The fix is to set the system clock to CLK_M before initializing PLLP,
and then switch back to PLLP_OUT4, which has an appropriate divider
configured, after PLLP has been configured
Implement this logic in new function tegra30_set_up_pllp(),
which sets up PLLP and all PLLP_OUT* dividers, and handles the AVP
clock switching. Remove the duplicate PLLP setup from pllx_set_rate()
and adjust_pllp_out_freqs().
Signed-off-by: Jimmy Zhang <jimmzhang@nvidia.com>
[swarren, significantly refactored the change]
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Thierry Reding <treding@nvidia.com>
Tested-by: Thierry Reding <treding@nvidia.com>
Acked-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Tegra114 and later's PMC module removes the pwrgate_timer_on register
and replaces it with a clamp_status register. Adjust pmc.h to reflect
this, and update any code affected by the change.
The cpu.c change in this patch was extracted from a much larger patch
by Jimmy Zhang. The pmc.h change was written from scratch, but inspired
by related changes made by Tom Warren.
There could well be other differences in the PMC register set for chips
after Tegra20/30. However, they don't affect the code in U-Boot at
present, so I haven't attempted an exhaustive update of pmc.h.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Thierry Reding <treding@nvidia.com>
Acked-by: Thierry Reding <treding@nvidia.com>
Tested-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Some clock sources have 3-bit muxes in bits 31:29. Implement core
support for this mux field.
Signed-off-by: Tom Warren <twarren@nvidia.com>
[swarren, extracted from a larger patch by Tom]
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Thierry Reding <treding@nvidia.com>
Tested-by: Thierry Reding <treding@nvidia.com>
Acked-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Since all code that sets or interprets MASK_BITS_* now uses the enums
to define/compare the values, there is no need for MASK_BITS_* to have
a specific integer value. In fact, having a specific integer value may
encourage people to hard-code those values, or interpret the values in
incorrect ways.
As such, remove the logic that assigns a specific value to the enum
values in order to make it completely clear that it's just an enum, not
something that directly represents some integer value.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Thierry Reding <treding@nvidia.com>
Tested-by: Thierry Reding <treding@nvidia.com>
Acked-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
OUT_CLK_SOURCE_ are currently named after the number of bits the mask
they represent includes. However, bit count is not the only possible
variable; bit position may also vary. Rename OUT_CLK_SOURCE_ to
OUT_CLK_SOURCE_31_30_ and OUT_CLK_SOURCE4_ to OUT_CLK_SOURCE_31_28 to
more completely describe exactly what they represent, without having to
go look up the definitions.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Thierry Reding <treding@nvidia.com>
Tested-by: Thierry Reding <treding@nvidia.com>
Acked-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
The only place where the MASK_BITS_* values are used is in
adjust_periph_pll(), which interprets the value 4 (old MASK_BITS_29_28,
new MASK_BITS_31_28) as being associated with mask OUT_CLK_SOURCE4_MASK,
i.e. bits 31:28. Rename the MASK_BITS_ macro to reflect how it's actually
implemented.
Note that no Tegra clock register actually uses all of bits 31:28 as
the mux field. Rather, bits 30:28, 29:28, or 28 are used. However, in
those cases, nothing is stored in the bits above the mux field, so it's
safe to pretend that the mux field extends all the way to the end of the
register. As such, the U-Boot clock driver is currently a bit lazy, and
doesn't distinguish between 31:28, 30:28, 29:28 and 28; it just lumps
them all together and pretends they're all 31:28. This patch doesn't
cause this issue; it was pre-existing. Hopefully, future patches will
clean this up.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Thierry Reding <treding@nvidia.com>
Tested-by: Thierry Reding <treding@nvidia.com>
Acked-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
The enum used to define the set of register bits used to represent a
clock's input mux, MUX_BITS_*, is defined separately for each SoC at
present. Move this definition to a common location to ease fixing up
some issues with the definition, and the code that uses it.
Signed-off-by: Tom Warren <twarren@nvidia.com>
[swarren, extracted from a larger patch by Tom]
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Thierry Reding <treding@nvidia.com>
Tested-by: Thierry Reding <treding@nvidia.com>
Acked-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
The CPU complex reset masks are not matching with the datasheet for
the CLK_RST_CONTROLLER_RST_CPU_CMPLX_SET/CLR_0 registers. For both T20
and T30 the register consist of groups of 4 bits, with one bit for
each CPU core. On T20 the 2 high bits of each group are always stubbed
as there is only 2 cores.
Signed-off-by: Alban Bedel <alban.bedel@avionic-design.de>
Acked-by: Stephen Warren <swarren@nvidia.com>
Tested-by: Stephen Warren <swrren@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Add the Tegra30 SKU b1 and treat it like other Tegra30 chips.
Signed-off-by: Alban Bedel <alban.bedel@avionic-design.de>
Reviewed-by: Julian Scheel <julian.scheel@avionic-design.de>
Signed-off-by: Tom Warren <twarren@nvidia.com>
This commit unifies board-specific USB initialization implementations
under one symbol (usb_board_init), declaration of which is available in
usb.h.
New API allows selective initialization of USB controllers whenever needed.
Signed-off-by: Mateusz Zalega <m.zalega@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Reviewed-by: Lukasz Majewski <l.majewski@samsung.com>
Cc: Marek Vasut <marex@denx.de>
Cc: Lukasz Majewski <l.majewski@samsung.com>
Tegra30 and Tegra114 are compatible except PLL parameters.
Tested on Tegra30 Cardhu, and Tegra114 Dalmore
platforms. All works well.
Signed-off-by: Jim Lin <jilin@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Make U-Boot aware of the Tegra20 SKU 7, and treat it identically
to any other Tegra20.
My Whistler board has a SoC with this SKU.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Make U-Boot aware of the Tegra114 SKU 1, and treat it identically
to any other Tegra114.
This value is used on (at least some) Dalmore boards with a production
rather than engineering chip. Such boards are in the hands of some
partners who want to use upstream U-Boot.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
As suggested by Stephen Warren, use tegra_get_chip() to return
the pure CHIPID for a Tegra SoC (i.e. 0x20 for Tegra20, 0x30 for
Tegra30, etc.) and rename tegra_get_chip_type() to reflect its true
function, i.e. tegra_get_chip_sku(), which returns an ID like
TEGRA_SOC_T25, TEGRA_SOC_T33, etc.
Signed-off-by: Tom Warren <twarren@nvidia.com>
Reviewed-by: Stephen Warren <swarren@nvidia.com>
T114 needs the SYSCTR0 counter initialized so the TSC can be
read by the kernel. Do it in the bootloader since it's a write-once
deal (secure/non-secure mode dependent).
Signed-off-by: Tom Warren <twarren@nvidia.com>
Reviewed-by: Stephen Warren <swarren@nvidia.com>
Without this change, kernel fails at calling function cache_clean_flush
during kernel early boot.
Aprocryphally, intended for T114 only, so I check for a T114 SoC.
Works (i.e. dalmore 3.8 kernel now starts printing to console).
Signed-off-by: Tom Warren <twarren@nvidia.com>
Make U-Boot aware of the T33 SKU of Tegra30, and treat it identically
to any other Tegra30.
An alternative would be to simply remove the SKU checking from
tegra_get_chip_type(); most use of the value most likely simply wants
to know the current chip, not the specific SKU. Or, the function could
be split into separate tegra_get_chip() and tegra_get_sku() for the
cases where differentiation really is required.
I wonder whether tegra_get_chip_type() should printf() whenever any
unkown chip/SKU is found, although perhaps the function is called so
early that the printf() wouldn't actually make it to the UART anyway.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Acked-by: Simon Glass <sjg@chromium.org>
Rename tegra SPI drivers to tegra20_flash and tegra20_slink in
preparation for commonization and addition of tegra114_spi.
Signed-off-by: Allen Martin <amartin@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Reviewed-by: Stephen Warren <swarren@nvidia.com>
This feature was only used for tegra20 seaboard that had a pinmux
conflict on the SPI pins. These boards were never manufactured, so
remove this support to clean up SPI driver.
Signed-off-by: Allen Martin <amartin@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Reviewed-by: Stephen Warren <swarren@nvidia.com>
Removed SDMMC base addresses from tegra.h since they're no longer used.
Added additional vendor-specific SD/MMC registers and bus power defines.
Signed-off-by: Tom Warren <twarren@nvidia.com>
Reviewed-by: Stephen Warren <swarren@nvidia.com>
tegra_mmc_init() now parses the DT info for bus width, WP/CD GPIOs, etc.
Tested on Seaboard, fully functional.
Tamonten boards (medcom-wide, plutux, and tec) use a different/new
dtsi file w/common settings.
Signed-off-by: Tom Warren <twarren@nvidia.com>
Signed-off-by: Thierry Reding <thierry.reding@avionic-design.de>
Reviewed-by: Stephen Warren <swarren@nvidia.com>
T114 has a slightly different I2C clock, with a new (extra) divisor
in standard/fast mode and HS mode. Tested on my Dalmore, and the I2C
clock is 100KHz +/- 3Hz on my Saleae Logic analyzer.
Added a new entry in compat_names for T114 I2C since it differs
from the previous Tegra SoCs. A flag is set when T114 I2C HW is
found so new features like the extra clock divisor can be used.
Signed-off-by: Tom Warren <twarren@nvidia.com>
Acked-by: Laxman Dewangan <ldewangan@nvidia.com>
The ehci_hcd entry points were just calling into the Tegra USB
functions. Now that they are in the same file we can just move over the
implementation.
Signed-off-by: Lucas Stach <dev@lynxeye.de>
Acked-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Tom Warren <twarren@nvidia.com>
This moves the Tegra USB implementation into the drivers/usb/host
directory. Note that this merges the old
/arch/arm/cpu/armv7/tegra20/usb.c file into ehci-tegra.c. No code
changes, just moving stuff around.
v2: While at it also move some defines and the usb.h header file to make
usb driver usable for Tegra30.
NOTE: A lot more work is required to properly init the PHYs and PLL_U on
Tegra30, this is just to make porting easier and it does no harm here.
Signed-off-by: Lucas Stach <dev@lynxeye.de>
Acked-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Common Tegra files are in arch-tegra, shared between T20/T30/T114.
Tegra114-specific headers are in arch-tegra114. Note that some of
these will be filled in as more T114 support is added (drivers,
WB/LP0 support, etc.).
Signed-off-by: Tom Warren <twarren@nvidia.com>
Reviewed-by: Stephen Warren <swarren@nvidia.com>
Add driver for tegra SPI "SLINK" style driver. This controller is
similar to the tegra20 SPI "SFLASH" controller. The difference is
that the SLINK controller is a genernal purpose SPI controller and the
SFLASH controller is special purpose and can only talk to FLASH
devices. In addition there are potentially many instances of an SLINK
controller on tegra and only a single instance of SFLASH. Tegra20 is
currently ths only version of tegra that instantiates an SFLASH
controller.
This driver supports basic PIO mode of operation and is configurable
(CONFIG_OF_CONTROL) to be driven off devicetree bindings. Up to 4
devices per controller may be attached, although typically only a
single chip select line is exposed from tegra per controller so in
reality this is usually limited to 1.
To enable this driver, use CONFIG_TEGRA_SLINK
Signed-off-by: Allen Martin <amartin@nvidia.com>
Signed-off-by: Tom Warren <twarren@nvidia.com>
Add I/O addresses of SPI SLINK controllers 1-6
Signed-off-by: Allen Martin <amartin@nvidia.com>
Acked-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Tom Warren <twarren@nvidia.com>
This 'commonizes' much of the clock/pll code. SoC-dependent code
and tables are left in arch/cpu/tegraXXX-common/clock.c
Some T30 tables needed whitespace fixes due to checkpatch complaints.
Signed-off-by: Tom Warren <twarren@nvidia.com>
These files are used by both SPL and main U-Boot.
Also made minor changes to shared Tegra code to support
T30 differences.
Signed-off-by: Tom Warren <twarren@nvidia.com>
Reviewed-by: Stephen Warren <swarren@nvidia.com>
