usb: xhci: convert to TRB_LEN() and TRB_INTR_TARGET()
For normal TRB fields:
use TRB_LEN(x) instead of ((x) & TRB_LEN_MASK);
and use TRB_INTR_TARGET(x) instead of
(((x) & TRB_INTR_TARGET_MASK) << TRB_INTR_TARGET_SHIFT)
Signed-off-by: Chunfeng Yun <chunfeng.yun@mediatek.com> Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
usb: xhci: add quirks flag to support MediaTek xHCI 0.96
There some vendor quirks for MTK xHCI 0.96 host controller:
1. It defines some extra SW scheduling parameters for HW
to minimize the scheduling effort for synchronous and
interrupt endpoints. The parameters are put into reserved
DWs of slot context and endpoint context.
2. Its TDS in Normal TRB defines a number of packets that
remains to be transferred for a TD after processing all
Max packets in all previous TRBs.
Signed-off-by: Chunfeng Yun <chunfeng.yun@mediatek.com> Tested-by: Frank Wunderlich <frank-w@public-files.de> Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
usb: xhci: create one unified function to calculate TRB TD remainder
xhci versions 1.0 and later report the untransferred data remaining in a
TD a bit differently than older hosts.
We used to have separate functions for these, and needed to check host
version before calling the right function.
Now Mediatek host has an additional quirk on how it uses the TD Size
field for remaining data. To prevent yet another function for calculating
remainder we instead want to make one quirk friendly unified function.
Porting from the Linux: c840d6ce772d("xhci: create one unified function to calculate TRB TD remainder.") 124c39371114("xhci: use boolean to indicate last trb in td remainder calculation")
Signed-off-by: Chunfeng Yun <chunfeng.yun@mediatek.com> Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Philippe Reynes [Fri, 18 Sep 2020 12:13:02 +0000 (14:13 +0200)]
sntp: use udp framework
This commits update the support of sntp to use
the framework udp. This change allows to remove
all the reference to sntp in the main network
file net/net.c.
Signed-off-by: Philippe Reynes <philippe.reynes@softathome.com> Reviewed-by: Simon Glass <sjg@chromium.org>
Philippe Reynes [Fri, 18 Sep 2020 12:13:00 +0000 (14:13 +0200)]
net: add a generic udp protocol
This commit adds a generic udp protocol framework in the
network loop. So protocol based on udp may be implemented
without modifying the network loop (for example custom
wait magic packet).
Signed-off-by: Philippe Reynes <philippe.reynes@softathome.com> Reviewed-by: Simon Glass <sjg@chromium.org>
Ley Foon Tan [Tue, 25 Aug 2020 02:26:36 +0000 (10:26 +0800)]
net: tftp: Fix store_block offset calculation
tftp_cur_block start with 1 for first block, but tftp_cur_block counter is
start with zero when block number is rollover. The existing code
"tftp_cur_block - 1" will cause the block number become -1 in store_block()
when tftp_cur_block is 0 when tftp_cur_block is rollover.
The fix pass in tftp_cur_block to store_block() and minus the
tftp_block_size when do the offset calculation.
Signed-off-by: Ley Foon Tan <ley.foon.tan@intel.com> Reviewed-By: Ramon Fried <rfried.dev@gmail.com>
Ley Foon Tan [Tue, 25 Aug 2020 02:26:35 +0000 (10:26 +0800)]
net: tftp: Fix tftp_prev_block counter update
Fixes missing update to tftp_prev_block counter before increase
tftp_cur_block counter when do the tftpput operation.
tftp_prev_block counter is used in update_block_number() function to
check whether block number (sequence number) is rollover. This bug
cause the tftpput command fail to upload a large file when block
number is greater than 16-bit (0xFFFF).
Signed-off-by: Ley Foon Tan <ley.foon.tan@intel.com> Reviewed-By: Ramon Fried <rfried.dev@gmail.com>
net: ftgmac100: Add support for board specific PHY interface address
ftgmac100 driver is using hard-coded PHY interface address of zero.
Each board can have different PHY interface address (phy_addr).
This commit modifies the driver to make use of board specific address
by leveraging CONFIG_PHY_ADDR.
Signed-off-by: Thirupathaiah Annapureddy <thiruan@linux.microsoft.com> Reviewed-by: Cédric Le Goater <clg@kaod.org>
Heiko Stuebner [Tue, 9 Jun 2020 13:37:40 +0000 (15:37 +0200)]
net: phy: mscc: sync rx/tx delay settings with Linux on vsc85xx
The Linux kernel does set the clock delays to
- 0.2 ns (their default, and lowest, hardware value) if delays should
not be enabled
- 2.0 ns (which causes the data to be sampled at exactly half way between
clock transitions at 1000 Mbps) if delays should be enabled
depending on the interface mode
See https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/net/phy/mscc/mscc_main.c#n523
So instead of using arbitrary delay values like now, mimic this behaviour.
The behaviour is the same for all of vsc8530/8531/8540/8541 so move that
to a shared function while at it.
Signed-off-by: Heiko Stuebner <heiko.stuebner@theobroma-systems.com> Reviewed-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Heiko Stuebner [Tue, 9 Jun 2020 13:37:39 +0000 (15:37 +0200)]
net: phy: mscc: make clock-output configurable on vsc85xx
The vsc8530/8531/8540/8541 phys have a configurable clock output that
can emit 25, 50 and 125 MHz rates, which in turn may be needed for
stable network connections.
This follows a similar change introduced into the Linux kernel at
https://lore.kernel.org/netdev/20200609133140.1421109-2-heiko@sntech.de
Signed-off-by: Heiko Stuebner <heiko.stuebner@theobroma-systems.com> Reviewed-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
firmware: smci: sandbox test for SCMI reset controllers
Add tests for SCMI reset controllers. A test device driver
sandbox-scmi_devices.c is used to get reset resources, allowing further
resets manipulation.
Change sandbox-smci_agent to emulate 1 reset controller exposed through
an agent. Add DM test scmi_resets to test this reset controller.
Signed-off-by: Etienne Carriere <etienne.carriere@linaro.org> Cc: Simon Glass <sjg@chromium.org> Cc: Peng Fan <peng.fan@nxp.com> Cc: Sudeep Holla <sudeep.holla@arm.com> Reviewed-by: Simon Glass <sjg@chromium.org>
reset: add reset controller driver for SCMI agents
This change introduces a reset controller driver for SCMI agent devices.
When SCMI agent and SCMI reset domain drivers are enabled, SCMI agent
binds a reset controller device for each SCMI reset domain protocol
devices enabled in the FDT.
SCMI reset driver is embedded upon CONFIG_RESET_SCMI=y. If enabled,
CONFIG_SCMI_AGENT is also enabled.
SCMI Reset Domain protocol is defined in the SCMI specification [1].
Links: [1] https://developer.arm.com/architectures/system-architectures/software-standards/scmi Signed-off-by: Etienne Carriere <etienne.carriere@linaro.org> Cc: Simon Glass <sjg@chromium.org> Cc: Peng Fan <peng.fan@nxp.com> Cc: Sudeep Holla <sudeep.holla@arm.com> Reviewed-by: Simon Glass <sjg@chromium.org>
Add tests for SCMI clocks. A test device driver sandbox-scmi_devices.c
is used to get clock resources, allowing further clock manipulation.
Change sandbox-smci_agent to emulate 3 clocks exposed through 2 agents.
Add DM test scmi_clocks to test these 3 clocks.
Update DM test sandbox_scmi_agent with load/remove test sequences
factorized by {load|remove}_sandbox_scmi_test_devices() helper functions.
Signed-off-by: Etienne Carriere <etienne.carriere@linaro.org> Cc: Simon Glass <sjg@chromium.org> Cc: Peng Fan <peng.fan@nxp.com> Cc: Sudeep Holla <sudeep.holla@arm.com> Reviewed-by: Simon Glass <sjg@chromium.org>
This change introduces a clock driver for SCMI agent devices. When
SCMI agent and SCMI clock drivers are enabled, SCMI agent binds a
clock device for each SCMI clock protocol devices enabled in the FDT.
SCMI clock driver is embedded upon CONFIG_CLK_SCMI=y. If enabled,
CONFIG_SCMI_AGENT is also enabled.
SCMI Clock protocol is defined in the SCMI specification [1].
Links: [1] https://developer.arm.com/architectures/system-architectures/software-standards/scmi Signed-off-by: Etienne Carriere <etienne.carriere@linaro.org> Cc: Lukasz Majewski <lukma@denx.de> Cc: Simon Glass <sjg@chromium.org> Cc: Peng Fan <peng.fan@nxp.com> Cc: Sudeep Holla <sudeep.holla@arm.com> Reviewed-by: Simon Glass <sjg@chromium.org>
This change implements a SMCCC transport for SCMI exchanges. This
implementation follows the Linux kernel as references implementation
for SCMI message processing, using the SMT format for communication
channel meta-data.
Use of SMCCC transport in SCMI FDT bindings are defined in the Linux
kernel DT bindings since v5.8. SMCCC with SMT is implemented in OP-TEE
from tag 3.9.0 [2].
Links: [2] https://github.com/OP-TEE/optee_os/commit/a58c4d706d23 Signed-off-by: Etienne Carriere <etienne.carriere@linaro.org> Cc: Simon Glass <sjg@chromium.org> Cc: Peng Fan <peng.fan@nxp.com> Cc: Sudeep Holla <sudeep.holla@arm.com> Reviewed-by: Simon Glass <sjg@chromium.org>
This change implements a mailbox transport using SMT format for SCMI
exchanges. This implementation follows the Linux kernel and
SCP-firmware [1] as references implementation for SCMI message
processing using SMT format for communication channel meta-data.
Use of mailboxes in SCMI FDT bindings are defined in the Linux kernel
DT bindings since v4.17.
Links: [1] https://github.com/ARM-software/SCP-firmware Signed-off-by: Etienne Carriere <etienne.carriere@linaro.org> Cc: Simon Glass <sjg@chromium.org> Cc: Peng Fan <peng.fan@nxp.com> Cc: Sudeep Holla <sudeep.holla@arm.com> Reviewed-by: Simon Glass <sjg@chromium.org>
This change introduces SCMI agent uclass to interact with a firmware
using the SCMI protocols [1].
SCMI agent uclass currently supports a single method to request
processing of the SCMI message by an identified server. A SCMI message
is made of a byte payload associated to a protocol ID and a message ID,
all defined by the SCMI specification [1]. On return from process_msg()
method, the caller gets the service response.
SCMI agent uclass defines a post bind generic sequence for all devices.
The sequence binds all the SCMI protocols listed in the FDT for that
SCMI agent device. Currently none, but later change will introduce
protocols.
This change implements a simple sandbox device for the SCMI agent uclass.
The sandbox nicely answers SCMI_NOT_SUPPORTED to SCMI messages.
To prepare for further test support, the sandbox exposes a architecture
function for test application to read the sandbox emulated devices state.
Currently supports 2 SCMI agents, identified by an ID in the FDT device
name. The simplistic DM test does nothing yet.
SCMI agent uclass is designed for platforms that embed a SCMI server in
a firmware hosted somewhere, for example in a companion co-processor or
in the secure world of the executing processor. SCMI protocols allow an
SCMI agent to discover and access external resources as clock, reset
controllers and more. SCMI agent and server communicate following the
SCMI specification [1]. This SCMI agent implementation complies with
the DT bindings defined in the Linux kernel source tree regarding
SCMI agent description since v5.8.
Links: [1] https://developer.arm.com/architectures/system-architectures/software-standards/scmi Signed-off-by: Etienne Carriere <etienne.carriere@linaro.org> Cc: Simon Glass <sjg@chromium.org> Cc: Peng Fan <peng.fan@nxp.com> Cc: Sudeep Holla <sudeep.holla@arm.com> Reviewed-by: Simon Glass <sjg@chromium.org>
test: dm: Add tests for regmap managed API and regmap fields
The tests rely on a dummy driver to allocate and initialize the regmaps
and the regmap fields using the managed API. The first test checks if
the regmap config fields like width, reg_offset_shift, range specifiers,
etc work. The second test checks if regmap fields behave properly (mask
and shift are ok) by peeking into the regmap.
When more nodes are added for a uclass the index might go into two or
more digits. This means that there are less spaces printed because they
are used up by the extra digits. Update the regular expression to allow
variable-length spacing between the class name and and index.
This was discovered when adding a simple_bus node in test.dts made
test_bind_unbind_with_uclass() fail because the index went up to 10.
regmap: Allow devices to specify regmap range start and size in config
Some devices need to calculate the regmap base address at runtime. This
makes it impossible to use device tree to get the regmap base. Instead,
allow devices to specify it in the regmap config. This will create a
regmap with a single range that corresponds to the start and size given
by the driver.
Signed-off-by: Pratyush Yadav <p.yadav@ti.com> Reviewed-by: Simon Glass <sjg@chromium.org>
Right now, the base of a regmap can only be obtained from the device
tree. This makes it impossible for devices which calculate the base at
runtime to use a regmap. An example of such a device is the Cadence
Sierra PHY.
Allow creating a regmap with one range whose start and size can be
specified by the driver based on calculations at runtime.
Signed-off-by: Pratyush Yadav <p.yadav@ti.com> Reviewed-by: Simon Glass <sjg@chromium.org>
Right now, regmap_read() and regmap_write() read/write a 32-bit value
only. To write other lengths, regmap_raw_read() and regmap_raw_write()
need to be used.
This means that any driver ported from Linux that relies on
regmap_{read,write}() to know the size already has to be updated at each
callsite. This makes the port harder to maintain.
So, allow specifying the read/write width to make it easier to port the
drivers, since now the only change needed is when initializing the
regmap.
Signed-off-by: Pratyush Yadav <p.yadav@ti.com> Reviewed-by: Simon Glass <sjg@chromium.org>
Some fields will be introduced in the regmap structure that should be
set to 0 by default. So, once we allocate a regmap, make sure it is
zeroed out to avoid unexpected defaults for those values.
Signed-off-by: Pratyush Yadav <p.yadav@ti.com> Reviewed-by: Simon Glass <sjg@chromium.org>
Most of new linux drivers are using managed-API to allocate resources. To
ease porting drivers from linux to U-Boot, introduce devm_regmap_init() as
a managed API to get a regmap from the device tree.
Add a test to verify that GPIOs can be acquired/released using the managed
API. Also check that the GPIOs are released when the consumer device is
removed.
drivers: reset: Add a managed API to get reset controllers from the DT
Add managed functions to get a reset_ctl from the device-tree, based on a
name or an index.
Also add a managed functions to get a reset_ctl_bulk (array of reset_ctl)
from the device-tree.
When the device is unbound, the reset controllers are automatically
released and the data structure is freed.
Tom Rini [Wed, 30 Sep 2020 13:21:43 +0000 (09:21 -0400)]
Merge branch 'next' of https://gitlab.denx.de/u-boot/custodians/u-boot-riscv into next
- Disable CMD_IRQ for RISC-V.
- Update sipeed/maix doc
- Obtain reg of SiFive RAM via dev_read_addr_index() instead of regmap API.
- Cleans up RISC-V timer drivers and converts them to DM.
- Correctly handle IPIs already pending upon prior stage bootloader (on the K210)
Sean Anderson [Tue, 15 Sep 2020 14:45:22 +0000 (10:45 -0400)]
dm: Print device name in dev_xxx like Linux
This adorns messages generated by dev_xxx with the device and driver
names. It also redirects dev_xxx to log when it is available. The names
of these functions very roughly take inspiration from Linux, but there is
no deeper correlation.
Both struct udevice and struct device are supported when logging, though
logging with struct device is no better than using log_xxx. The latter is
supported because of the large amount of existing code which logs with
struct device.
Signed-off-by: Sean Anderson <seanga2@gmail.com> Tested-by: Patrick Delaunay <patrick.delaunay@st.com>
Sean Anderson [Tue, 15 Sep 2020 14:45:20 +0000 (10:45 -0400)]
video: stm32: Fix not calling dev_xxx with a device
There is no member `dev` in dw_mipi_dsi, but there is one in mipi_dsi_host,
so use that.
Signed-off-by: Sean Anderson <seanga2@gmail.com> Reviewed-by: Patrick Delaunay <patrick.delaunay@st.com> Tested-by: Patrick Delaunay <patrick.delaunay@st.com>
Sean Anderson [Tue, 15 Sep 2020 14:45:15 +0000 (10:45 -0400)]
usb: dwc2: Fix not calling dev_xxx with a device
This adds a dev argument to some functions so dev_xxx always has a device
to log with. In one instance we must use use a different log function when
we are compiled without DM_USB.
Signed-off-by: Sean Anderson <seanga2@gmail.com> Reviewed-by: Patrice Chotard <patrice.chotard@st.com> Reviewed-by: Patrick Delaunay <patrick.delaunay@st.com> Tested-by: Patrick Delaunay <patrick.delaunay@st.com>
Sean Anderson [Tue, 15 Sep 2020 14:45:12 +0000 (10:45 -0400)]
spi: zynqmp_gqspi: Fix not calling dev_err with a device
Use `bus` instead of `dev`.
Signed-off-by: Sean Anderson <seanga2@gmail.com> Reviewed-by: Michal Simek <michal.simek@xilinx.com> Tested-by: Patrick Delaunay <patrick.delaunay@st.com>
Sean Anderson [Tue, 15 Sep 2020 14:45:06 +0000 (10:45 -0400)]
phy: usbphyc: Fix not calling dev_err with a device
Use the phy's device.
Signed-off-by: Sean Anderson <seanga2@gmail.com> Reviewed-by: Patrice Chotard <patrice.chotard@st.com> Reviewed-by: Patrick Delaunay <patrick.delaunay@st.com> Tested-by: Patrick Delaunay <patrick.delaunay@st.com>
Sean Anderson [Tue, 15 Sep 2020 14:45:03 +0000 (10:45 -0400)]
phy: rockchip: Fix not calling dev_err with a device
Get the device from phy, or pass the phy in.
Signed-off-by: Sean Anderson <seanga2@gmail.com> Reviewed-by: Kever Yang <kever.yang@rock-chips.com> Tested-by: Patrick Delaunay <patrick.delaunay@st.com>
Sean Anderson [Tue, 15 Sep 2020 14:44:57 +0000 (10:44 -0400)]
net: mvpp2: Convert netdev_xxx to dev_xxx
netdev_xxx evaluates to printf in U-Boot, so there is no extra info
printed. mvpp2 one of only two drivers which use these functions in U-Boot.
Convert these functions to dev_xxx where possible (and to log_xxx where
not).
Signed-off-by: Sean Anderson <seanga2@gmail.com> Reviewed-by: Simon Glass <sjg@chromium.org> Tested-by: Patrick Delaunay <patrick.delaunay@st.com>
Sean Anderson [Tue, 15 Sep 2020 14:44:55 +0000 (10:44 -0400)]
net: mvneta: Convert netdev_xxx to dev_xxx
netdev_xxx evaluates to printf in U-Boot, so there is no extra info
printed. mvneta is one of two drivers which use these functions in U-Boot.
Convert these functions to dev_xxx where possible (and to log_xxx where
not).
Signed-off-by: Sean Anderson <seanga2@gmail.com> Reviewed-by: Simon Glass <sjg@chromium.org> Tested-by: Patrick Delaunay <patrick.delaunay@st.com>
Sean Anderson [Tue, 15 Sep 2020 14:44:51 +0000 (10:44 -0400)]
nand: vybrid: Re-introduce vf610_nfc.dev
This member was presumably dropped when this driver was converted from
Linux. However, it is still used in log statements during initialization.
This patch adds the member back. In addition, allocation of struct
vf610_nfc has been moved to the callers of vf610_nfc_nand_init. This allows
it to be allocated by DM (if it is being used) and for dev to be
initialized.
Signed-off-by: Sean Anderson <seanga2@gmail.com> Tested-by: Patrick Delaunay <patrick.delaunay@st.com>
Sean Anderson [Tue, 15 Sep 2020 14:44:50 +0000 (10:44 -0400)]
nand: brcmnand: Fix not calling dev_err() with a device
There are too many levels of indirection when calling dev_err. This is an
artifact of the conversion of brcmnand_host.pdev from a struct
platform_device (which has a member `dev` pointing to a struct device) to
struct udevice.
Signed-off-by: Sean Anderson <seanga2@gmail.com> Tested-by: Patrick Delaunay <patrick.delaunay@st.com>
Sean Anderson [Tue, 15 Sep 2020 14:44:43 +0000 (10:44 -0400)]
mtd: spi: Fix logging in spi-nor-tiny
This fixes dev_xxx() not always being called with a device. In
spi_nor_reg_read, a the slave device may not always be available, so we use
bus and cs instead.
Signed-off-by: Sean Anderson <seanga2@gmail.com> Tested-by: Patrick Delaunay <patrick.delaunay@st.com>
Sean Anderson [Tue, 15 Sep 2020 14:44:41 +0000 (10:44 -0400)]
mtd: nand: sunxi: Fix not calling dev_err with a device
Usually the device is gotten from sunxi_nfc. This is a struct device and
not a struct udevice, but the whole driver seems to be written wihout DM
anyway...
In a few instances, this patch modifies functions to take an nfc to log
with. In once instance we use mtd_info's device since there is no nfc.
Signed-off-by: Sean Anderson <seanga2@gmail.com> Tested-by: Patrick Delaunay <patrick.delaunay@st.com>
Sean Anderson [Tue, 15 Sep 2020 14:44:39 +0000 (10:44 -0400)]
i2c: mxc: Fix dev_err being called on a nonexistant variable
The udevice we are working with is called `bus` and not `dev`.
Signed-off-by: Sean Anderson <seanga2@gmail.com> Reviewed-by: Simon Glass <sjg@chromium.org> Reviewed-by: Heiko Schocher <hs@denx.de> Tested-by: Patrick Delaunay <patrick.delaunay@st.com>
Sean Anderson [Tue, 15 Sep 2020 14:44:37 +0000 (10:44 -0400)]
dm: syscon: Fix calling dev_dbg with an uninitialized device
We can't use dev_dbg here because we haven't bound to the device yet. Use
log_debug instead.
Signed-off-by: Sean Anderson <seanga2@gmail.com> Reviewed-by: Patrick Delaunay <patrick.delaunay@st.com> Tested-by: Patrick Delaunay <patrick.delaunay@st.com>
Sean Anderson [Mon, 21 Sep 2020 11:51:41 +0000 (07:51 -0400)]
riscv: Add some comments to start.S
This adds comments regarding the ordering and purpose of certain
instructions as I understand them.
Signed-off-by: Sean Anderson <seanga2@gmail.com> Reviewed-by: Bin Meng <bin.meng@windriver.com> Reviewed-by: Rick Chen <rick@andestech.com> Reviewed-by: Leo Liang <ycliang@andestech.com>
Sean Anderson [Mon, 21 Sep 2020 11:51:40 +0000 (07:51 -0400)]
riscv: Ensure gp is NULL or points to valid data
This ensures constructs like `if (gd & gd->...) { ... }` work when
accessing the global data pointer. Without this change, it was possible for
a very early trap to cause _exit_trap to directly or indirectly (through
printf) to read arbitrary memory. This could cause a second trap,
preventing show_regs from being printed.
printf (and specifically puts) uses gd to determine what function to print
with. These functions in turn use gd to find the serial device, etc.
However, before accessing gd, puts first checks to see if it is non-NULL.
This indicates an existing (perhaps undocumented) assumption that either gd
is NULL or it is completely valid.
Before this patch, gd either points to unexpected data (because it retains
the value it did from the prior-stage) or points to uninitialized data
(because it has not yet been initialized by board_init_f_init_reserve)
until the hart has acquired available_harts_lock. This can cause two
problems, depending on the value of gd->flags. If GD_FLG_SERIAL_READY is
unset, then some garbage data will be printed to stdout, but there will not
be a second trap. However, if GD_FLG_SERIAL_READY is set, then puts will
try to print with serial_puts, which will likely cause a second trap.
After this patch, gd is zero up until either a hart has set it in
wait_for_gd_init, or until it is set by arch_init_gd. This prevents its
usage before its data is initialized because both handle_trap and puts
ensure that gd is nonzero before using it. After gd has been set, it is OK
to access it because its data has been cleared (and so flags is valid).
XIP cannot use locks because flash is not writable. This leaves it
vulnerable to the same class of bugs regarding already-pending IPIs as
before this series. Fixing that would require finding another method of
synchronization, which is outside the scope of this series.
Fixes: 7c6ca03eae ("riscv: additional crash information") Signed-off-by: Sean Anderson <seanga2@gmail.com> Reviewed-by: Bin Meng <bin.meng@windriver.com> Reviewed-by: Rick Chen <rick@andestech.com>
Sean Anderson [Mon, 21 Sep 2020 11:51:38 +0000 (07:51 -0400)]
riscv: Clear pending IPIs on initialization
Even though we no longer call smp_function if an IPI was not sent by
U-Boot, we still need to clear any IPIs which were pending from the
execution environment. Otherwise, secondary harts will busy-wait in
secondary_hart_loop, instead of relaxing.
Along with the previous commit ("riscv: Use a valid bit to ignore
already-pending IPIs"), this fixes SMP booting on the Kendryte K210.
Signed-off-by: Sean Anderson <seanga2@gmail.com> Reviewed-by: Bin Meng <bin.meng@windriver.com> Reviewed-by: Rick Chen <rick@andestech.com>
Sean Anderson [Mon, 21 Sep 2020 11:51:37 +0000 (07:51 -0400)]
riscv: Use a valid bit to ignore already-pending IPIs
Some IPIs may already be pending when U-Boot is started. This could be a
problem if a secondary hart tries to handle an IPI before the boot hart has
initialized the IPI device.
To be specific, the Kendryte K210 ROM-based bootloader does not clear IPIs
before passing control to U-Boot. Without this patch, the secondary hart
jumps to address 0x0 as soon as it enters secondary_hart_loop, and then
hangs in its trap handler.
This commit introduces a valid bit so secondary harts know when and IPI
originates from U-Boot, and it is safe to use the IPI API. The valid bit is
initialized to 0 by board_init_f_init_reserve. Before this, secondary harts
wait in wait_for_gd_init.
Signed-off-by: Sean Anderson <seanga2@gmail.com> Reviewed-by: Bin Meng <bin.meng@windriver.com> Reviewed-by: Rick Chen <rick@andestech.com> Reviewed-by: Leo Liang <ycliang@andestech.com>
Sean Anderson [Mon, 21 Sep 2020 11:51:36 +0000 (07:51 -0400)]
riscv: Match memory barriers between send_ipi_many and handle_ipi
Without a matching barrier on the write side, the barrier in handle_ipi
does nothing. It was entirely possible for the boot hart to write to addr,
arg0, and arg1 *after* sending the IPI, because there was no barrier on the
sending side.
Fixes: 90ae281437 ("riscv: add option to wait for ack from secondary harts in smp functions") Signed-off-by: Sean Anderson <seanga2@gmail.com> Reviewed-by: Bin Meng <bin.meng@windriver.com> Reviewed-by: Rick Chen <rick@andestech.com> Reviewed-by: Leo Liang <ycliang@andestech.com>
projects / u-boot.git / log