yangsong8-a1 opened a new pull request, #22127:
URL: https://github.com/apache/nuttx/pull/22127
## Summary
Register more devices from the flattened device tree:
- SP804 dual timer ("arm,sp804", /dev/timerN) and the CMSDK APB
watchdog ("arm,cmsdk-wdt", /dev/watchdogN), both called from the
qemu-armv8a bringup.
- QEMU fw_cfg: a new read-only driver (drivers/misc/qemu_fwcfg.c) for
the "qemu,fw-cfg-mmio" device. /dev/qemufwcfg gives the named files
that QEMU publishes (-fw_cfg name=...) through the configdata ioctls.
qemu-armv7a and qemu-armv8a register it with CONFIG_QEMU_FWCFG.
- SiFive PWM ("sifive,pwm0", /dev/pwmN) and the i.MX GPT ("fsl,imx-gpt",
OCU and ICU channels): fdt_*_devices_register() helpers that a board
calls.
- CTU CAN FD on a system bus: ctucanfd_register(base, irq) registers a
memory-mapped controller, and CAN_CTUCANFD no longer requires PCI (the
PCI probe path is unchanged). fdt_ctucanfd_devices_register() and
fdt_pl061_devices_register() register the "ctu,ctucanfd" and enabled
"arm,pl061" nodes.
The CAN channel number comes from the node's "can-id" property, or
the discovery order when it is missing, and a node with a
"cpu-affinity" property is only registered on that CPU.
The timer, watchdog, PWM and GPT drivers are already upstream (#21849,
#21850, #21887, #21935).
## Impact
New files in drivers/devicetree and drivers/misc, new prototypes in
include/nuttx/fdt.h and include/nuttx/can/ctucanfd_pci.h, new
CONFIG_QEMU_FWCFG, CAN_CTUCANFD without "depends on PCI". The qemu
bringup files only call the new functions when the matching driver is
enabled, so existing configurations are unchanged.
## Testing
Ubuntu 22.04, gcc 13.2.Rel1 (arm, aarch64). CMake and Make builds of
qemu-armv8a:nsh with all the drivers enabled have no new warnings;
CAN_CTUCANFD also builds with PCI + chardev and without PCI + SocketCAN.
Upstream QEMU has no SP804, CMSDK watchdog, SiFive PWM or sysbus CTU CAN
FD on virt. The run below uses our QEMU 10.2.1 build, whose virt
machine adds these devices to the DTB. qemu-armv8a:nsh without PCI,
with TIMER_SP804, WATCHDOG_CMSDK, PWM_SIFIVE, QEMU_FWCFG,
CAN_CTUCANFD_CHARDEV, IOEXPANDER_PL061 and GPIO_LOWER_HALF:
qemu-system-aarch64 -nic none -cpu cortex-a53 -nographic
-machine virt,virtualization=on,gic-version=3
-object can-bus,id=canbus0
-device ctucan-sysbus,canbus=canbus0
-device ctucan-sysbus,canbus=canbus0
-fw_cfg name=opt/nuttx/test,string=hello-fwcfg -kernel nuttx
The bringup registers SP804, the watchdog and fw_cfg. A test program
calls the SiFive PWM, CTU CAN FD and PL061 helpers with fdt_get(), lists
/dev, reads the fw_cfg file, starts timer0 and watchdog0, and sends a
frame on /dev/can0 that /dev/can1 receives:
fdt_sifive_pwm_devices_register: 0
/dev/pwm0
/dev/qemufwcfg
/dev/timer0
/dev/watchdog0
open /dev/qemufwcfg: ok
CFGDIOC_GETCONFIG opt/nuttx/test: ret 0 len 11 "hello-fwcfg"
timer0: timeout 1000000, timeleft 999940 -> 799205 after 200 ms
watchdog0: flags 0x3 timeout 5000 ms
fdt_ctucanfd_devices_register: 0
fdt_pl061_devices_register: 0
/dev/can0
/dev/can1
gpio nodes: 8
open can0 3 can1 4
write can0: 7
read can1: 7 id 0x123 dlc 4 data 11 22 33 44
qemu-armv7a:nsh with DEVICE_TREE and QEMU_FWCFG on upstream QEMU 6.2
(RAM_START at 0x40200000 as in qemu-armv7a:full, so the DTB is not
overwritten):
/dev/qemufwcfg
open /dev/qemufwcfg: ok
CFGDIOC_GETCONFIG opt/nuttx/test: ret 0 len 11 "hello-fwcfg"
With `-device ctucan-sysbus,canbus=canbus0,can-id=3` and `can-id=5` the
nodes are /dev/can3 and /dev/can5. With `cpu-affinity=0` and
`cpu-affinity=1` (single-core nsh) only the first one is registered:
fdt_ctucanfd_devices_register: 0
/dev/can0
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]