27 Jul, 2019
4 commits
-
Rather than simply parking the R5 core in WFE after starting up ATF
on A53 instead use SYSFW API to properly shut down the R5 CPU cores
as well as associated timer resources that were pre-allocated. This
allows software further downstream to properly and gracefully bring
the R5 cores back online if desired.Signed-off-by: Andreas Dannenberg
Signed-off-by: Lokesh Vutla -
Any host while requesting for a device can request for its exclusive
access. If an exclusive permission is obtained then it is the host's
responsibility to release the device before the software entity on
the host completes its execution. Else any other host's request for
the device will be nacked. So add a command that releases all the
exclusive devices that is acquired by the current host. This should
be used with utmost care and can be called only at the end of the
execution.Signed-off-by: Lokesh Vutla
-
Add and expose a new processor shutdown API that wraps the two TISCI
messages involved in initiating a core shutdown. The API will first
queue a message to have the DMSC wait for a certain processor boot
status to happen followed by a message to trigger the actual shutdown-
with both messages being sent without waiting or requesting for a
response. Note that the processor shutdown API call will need to be
followed up by user software placing the respective core into either
WFE or WFI mode.Signed-off-by: Andreas Dannenberg
-
Sysfw provides an option for requesting exclusive access for a
device using the flags MSG_FLAG_DEVICE_EXCLUSIVE. If this flag is
not used, the device is meant to be shared across hosts. Once a device
is requested from a host with this flag set, any request to this
device from a different host will be nacked by sysfw. Current tisci
driver enables this flag for every device requests. But this may not
be true for all the devices. So provide a separate commands in driver
for exclusive and shared device requests.Signed-off-by: Lokesh Vutla
26 Jul, 2019
32 commits
-
- DaVinci emac DM work
- NXP driver work
- macb updates for RISC-V -
H3/H5 can either use the internal phy or an external one.
Before getting clock and resets for the internal phy,
test that we are using it because otherwise it break emac
when using an external phy.Tested-on: OrangePi PC2 (H5)
Fixes: 2348453c41 (net: sun8i_emac: Add EPHY CLK and RESET support)
Signed-off-by: Emmanuel Vadot
Acked-by: Joe Hershberger -
Now that we removed all legacy boards selecting TI_EMAC we can
completely convert the driver code to using the driver model.
This patch also updates all remaining users of davinci_emac.Signed-off-by: Bartosz Golaszewski
Tested-by: Adam Ford #am3517-evm & da850-evm
Reviewed-by: Ramon Fried -
We typically use same set of distro images (yocto, debian, fedora, etc.)
on both QEMU RISC-V virt machine and SiFive Unleashed board.With growing kernel and ramdisk images, we need to re-adjust default
U-Boot environment variables. The config header for QEMU RISC-V virt
machine has been already updated to handle bigger kernel and ramdisk
images hence this patch updates SiFive FU540 config header accordingly.Signed-off-by: Anup Patel
Reviewed-by: Bin Meng
Reviewed-by: Joe Hershberger
Reviewed-by: David Abdurachmanov
Tested-by: David Abdurachmanov -
Instead of depending on CONFIG_SYS_LITTLE_ENDIAN, we check at runtime
whether underlying system is little-endian or big-endian. This way
we are not dependent on any U-Boot specific OR compiler specific macro
to check system endianness.Signed-off-by: Anup Patel
Reviewed-by: Bin Meng
Reviewed-by: Ramon Fried
Acked-by: Joe Hershberger -
The SiFive MACB ethernet has a custom TX_CLK_SEL register to select
different TX clock for 1000mbps vs 10/100mbps.This patch adds SiFive MACB compatible string and extends the MACB
ethernet driver to change TX clock using TX_CLK_SEL register for
SiFive MACB.Signed-off-by: Anup Patel
Reviewed-by: Bin Meng
Reviewed-by: Ramon Fried
Acked-by: Joe Hershberger -
The LS1021A-TSN is a development board built by VVDN/Argonboards in
partnership with NXP.It features the LS1021A SoC and the first-generation SJA1105T Ethernet
switch for prototyping implementations of a subset of IEEE 802.1 TSN
standards.Supported boot media: microSD card (via SPL), QSPI flash.
Rev. A of the board uses a Spansion S25FL512S_256K serial flash, which
is 64 MB in size and has an erase sector size of 256KB (therefore,
flashing the RCW would erase part of U-Boot).Rev. B and C of the board use a Spansion S25FL256S1 serial flash, which
is only 32 MB in size but has an erase sector size of 64KB (therefore
the RCW image can be flashed without erasing U-Boot).To avoid the problems above, the U-Boot base address has been selected
at 0x100000 (the start of the 5th 256KB erase sector), which works for
all board revisions. Actually 0x40000 would have been enough, but
0x100000 is common for all Layerscape devices.eTSEC3 is connecting directly to SJA1105 via an RGMII fixed-link, but
SJA1105 is currently not supported by uboot. Therefore, eTSEC3 is
disabled.Signed-off-by: Xiaoliang Yang
Signed-off-by: Mingkai Hu
Signed-off-by: Jianchao Wang
Signed-off-by: Changming Huang
Signed-off-by: Vladimir Oltean[Vladimir] Code taken from https://github.com/openil/u-boot (which
itself is mostly copied from ls1021a-iot) and adapted with the following
changes:- Add a008850 errata workaround
- Converted eTSEC, MMC to DM to avoid all build warnings
- Plugged in distro boot feature, including support for extlinux.conf
- Added defconfig for QSPI boot
- Added the board/freescale/ls1021atsn/README.rst for initial setup
- Increased CONFIG_SYS_MONITOR_LEN so that the SPL malloc pool does not
get overwritten during copying of the u-boot.bin payload from MMC to
DDR.
Acked-by: Joe Hershberger
Reviewed-by: Bin Meng -
Due to a typo, "run qspi_bootcmd" and "env exists secureboot" got
concatenated instead of being separated by a semicolon.Signed-off-by: Vladimir Oltean
Reviewed-by: Bin Meng
Acked-by: Joe Hershberger -
Now that we have added driver model support to the TSEC driver,
convert ls1021atwr board to use it.This depends on previous DM series for ls1021atwr:
http://patchwork.ozlabs.org/patch/561855/Signed-off-by: Bin Meng
Signed-off-by: Vladimir Oltean
Acked-by: Joe Hershberger[Vladimir] Made the following changes:
- Added 'status = "disabled";' for all Ethernet ports in ls1021a.dtsi
- Fixed the confusion between the SGMII/TBI PCS for enet0 and enet1 -
a mistake ported over from Linux. Each SGMII PCS lies on the private
MDIO bus of the interface (and the RGMII enet2 has no SGMII PCS).
- Added CONFIG_DM_ETH to all ls1021atwr_* defconfigs
- Completely removed non-DM_ETH support from ls1021atwr
- Changed "compatible" string from "fsl,tsec-mdio" to "fsl,etsec2-mdio"
and from "fsl,tsec" to "fsl,etsec2" to match Linux -
In the case of the tsec network driver, so far there has been no
mainline user of DM_ETH where the DT bindings get used.In the case of the mdio bus, it looks like the "fsl,tsec-mdio" string
was made up for the documentation, but there is no mainline code that
parses the "compatible" property anyway.In both cases, there are no DT blobs that contain the old strings.
So change the documentation to "fsl,etsec2" for the Ethernet ports and
"fsl,etsec2-mdio" for the MDIO buses, which are strings that Linux also
uses, at least for LS1021A. More compatible strings can be added once
other (PowerPC) SoCs are migrated to DM_ETH.The current ls1021a.dtsi doesn't match what was documented for the MDIO
buses anyway (the "compatible" is "gianfar" currently). This will be
fixed in the next patch.Fixes: 69a00875e3db ("doc: dt-bindings: Describe Freescale TSEC ethernet controller")
Signed-off-by: Vladimir Oltean
Reviewed-by: Bin Meng
Acked-by: Joe Hershberger -
In tsec_init, the MAC address is retrieved from 2 different structures
depending on whether DM_ETH is enabled or not.But since the field name is the same inside both structures, we can
conditionally define the structure of the correct type and simplify the
assignments.Signed-off-by: Vladimir Oltean
Acked-by: Joe Hershberger
Reviewed-by: Bin Meng -
This replaces debug() calls with printf() so that it is immediately
obvious from the console that something is wrong.Signed-off-by: Vladimir Oltean
Acked-by: Joe Hershberger
Reviewed-by: Bin Meng -
This is a cosmetic patch that reorders variable definitions in the
inverse order of their line length, where possible.Signed-off-by: Vladimir Oltean
Acked-by: Joe Hershberger
Reviewed-by: Bin Meng -
By convention, the eTSEC MDIO controller nodes are defined in DT at
0x2d24000 and 0x2d50000, but actually U-Boot does not touch the
interrupt portion of the register map (MDIO_IEVENTM, MDIO_IMASKM,
MDIO_EMAPM).That leaves only the MDIO bus registers (MDIO_MIIMCFG, MDIO_MIIMCOM,
MDIO_MIIMADD, MDIO_MIIMADD, MDIO_MIIMCON, MDIO_MIIMSTAT) which start at
the 0x520 offset.So shift the DT-defined register map by the offset of MDIO_MIIMCFG when
mapping the MDIO bus registers.Signed-off-by: Vladimir Oltean
Acked-by: Joe Hershberger
Reviewed-by: Bin Meng -
The point of this patch is to eliminate the use of the locally-defined
"reg" variable (which interferes with next patch) and simplify the
fallback to the default CONFIG_SYS_TBIPA_VALUE in case "tbi-handle" is
missing.Signed-off-by: Vladimir Oltean
Acked-by: Joe Hershberger
Reviewed-by: Bin Meng -
Macb Ethernet controller requires a RX buffer of 128 bytes. It is
highly sub-optimal for Gigabit-capable GEM that is able to use
a bigger DMA buffer. Change this constant and associated macros
with data stored in the private structure.
RX DMA buffer size has to be multiple of 64 bytes as indicated in
DMA Configuration Register specification.Signed-off-by: Ramon Fried
Acked-by: Joe Hershberger -
DMA configuration was heavily dependent on the HW
defaults, add function to properly set the required
fields, including the new dma_burst_length.Signed-off-by: Ramon Fried
Reviewed-by: Anup Patel
Tested-by: Anup Patel
Acked-by: Joe Hershberger -
GEM support higher DMA burst writes/reads than the default (4).
add configuration structure with dma burst length so it could be
applied later to DMA configuration.Signed-off-by: Ramon Fried
Reviewed-by: Anup Patel
Tested-by: Anup Patel
Acked-by: Joe Hershberger -
This patch adds support for the sgmii phy interface,
available only to DM users, dictated by current driver
design.Signed-off-by: Ramon Fried
Reviewed-by: Anup Patel
Tested-by: Anup Patel
Acked-by: Joe Hershberger -
macb.h provides macros for reading/setting bitfields,
in macb registers and descriptors. use that instead
of redefining them in the source file.Signed-off-by: Ramon Fried
Reviewed-by: Anup Patel
Tested-by: Anup Patel
Acked-by: Joe Hershberger -
add support for clock rates higher than 2.4Mhz
Signed-off-by: Ramon Fried
Reviewed-by: Anup Patel
Tested-by: Anup Patel
Acked-by: Joe Hershberger -
Few registers and bits were added by Cadence and
they were not updated in the headers.
Take the latest definitions as defined in Linux
header (5.1) that also includes some comments
about existing registers.One register was improperly named (UR), fix that.
Signed-off-by: Ramon Fried
Reviewed-by: Anup Patel
Tested-by: Anup Patel
Acked-by: Joe Hershberger -
This binding documents two properties that describe the registers used to
perform MUX selection.Signed-off-by: Alex Marginean
Acked-by: Joe Hershberger
Reviewed-by: Bin Meng -
This driver is used for MDIO muxes driven over I2C. This is currently
used on Freescale LS1028A QDS board, on which the physical MDIO MUX is
controlled by an on-board FPGA which in turn is configured through I2C.Signed-off-by: Alex Marginean
Acked-by: Joe Hershberger
Reviewed-by: Bin Meng -
Using 'phy_connect' instead of 'phy_find_by_mask' and 'phy_connect_dev'
both deduplicates code and adds support for 'fixed-link'.Signed-off-by: Simon Goldschmidt
Acked-by: Joe Hershberger -
SGMII 2500 as supported on NXP SoCs requires AN to be disabled, handle
this case in the enetc sgmii init code.Signed-off-by: Alex Marginean
Acked-by: Joe Hershberger -
Enables ethernet, MDIO, PHY drivers for LS1028A RDB and QDS.
Signed-off-by: Alex Marginean
Reviewed-by: Bin Meng
Acked-by: Joe Hershberger -
Defines LS1028A RDB SGMII port, QDS RGMII port.
Signed-off-by: Alex Marginean
Reviewed-by: Bin Meng
Acked-by: Joe Hershberger -
Ethernet interfaces using serial protocols go through the serdes block
integrated in the SoC. This is accessed over dedicated internal MDIOs
which are part of the Ethernet PCI functions. Set up serdes at _start,
along with other protocol specific port/MAC configuration.
MDIO code is shared with enetc_mdio, read/write functions are exported
from fsl_enetc_mdio for this reason.Signed-off-by: Alex Marginean
Reviewed-by: Bin Meng
Acked-by: Joe Hershberger -
Adds a driver for the MDIO interface currently integrated in LS1028A SoC.
This MDIO interface is shared by multiple ethernet interfaces and is
presented as a stand-alone PCI function on the SoC ECAM.
Ethernet has a functional dependency on MDIO, for simplicity there is a
single config option for both.Signed-off-by: Alex Marginean
Reviewed-by: Bin Meng
Acked-by: Joe Hershberger -
Adds a driver for NXP ENETC ethernet controller currently integrated in
LS1028A. ENETC is a fairly straight-forward BD ring device and interfaces
are presented as PCI EPs on the SoC ECAM.Signed-off-by: Catalin Horghidan
Signed-off-by: Alex Marginean
Reviewed-by: Bin Meng
Acked-by: Joe Hershberger -
LS1028A ethernet interfaces work with at least 8 BDs, set number of buffers
to match that.Signed-off-by: Alex Marginean
Reviewed-by: Bin Meng
Acked-by: Joe Hershberger
25 Jul, 2019
4 commits
-
- Add in lzma-alone for current binman tests
- Update to Ubuntu's xenial-20190720 tag (latest).Signed-off-by: Tom Rini
Reviewed-by: Simon Glass -
We need pyelftools here to run rather than skip some tests.
Signed-off-by: Tom Rini
Reviewed-by: Simon Glass -
Minor driver-model fixes and tweaks
A few device-tree fixes
Binman support for extracting files from an image -
Reduce power domain calls when CONFIG_POWER_DOMAIN is disabled.
With gcc v8.2, this change saves 104 bytes.Signed-off-by: Anatolij Gustschin
Reviewed-by: Peng Fan