CV-Bowen opened a new pull request, #21765:
URL: https://github.com/apache/nuttx/pull/21765
## Summary
Make the peer reset that rptun does on reboot, poweroff and panic
actually reach the peer and complete.
- drivers/rptun: reset the peers instead of stopping them on reboot
The reboot notifier stopped the peers; a stopped peer only halts.
Reset them, so they reboot or power off with this core.
- drivers/rptun: initialize shmbase and shmlen
Silence a maybe-uninitialized warning.
- drivers/rptun: wait for the peer to reset, report poweroff and panic
Without ops->reset, rptun_dev_reset() set the status bit and returned
at once, so this core could go down before the peer saw it. Poll the
peer status until it echoes the bit, up to
CONFIG_RPTUN_STATUS_TIMEOUT_MS. The reboot notifier passes POWEROFF
for SYS_POWER_OFF, and a peer asked to power off now powers off.
- drivers/rptun: call boardctl() from a work queue
The peer status is handled in the rptun interrupt, and boardctl() runs
reboot notifiers that may sleep, which asserts there (trace below).
Queue the reset/poweroff on HPWORK.
- drivers/rptun: add RPTUN_AUTO_RESET to choose how the peers are reset
Disable, panic the peers from the panic notifier (default, the
current behaviour), or do everything from the reboot notifier.
- drivers/rptun: log the reset reason
- drivers/rptun: sleep instead of spinning while waiting for the peer
So a peer that syncs rpmsgfs while going down is still served.
- drivers/rptun: acknowledge the reset when boardctl cannot do it
Without BOARDCTL_RESET/POWEROFF the peer waited until timeout.
- drivers/rptun: run the panic notifier last
Priority INT_MIN like the reboot notifier, so the coredump runs first.
- drivers/rptun: use PANIC rather than ASSERT as the halt reason
## Impact
- New Kconfig choice `RPTUN_AUTO_RESET`, default
`RPTUN_AUTO_RESET_IN_PANIC_NOTIFIER` which keeps today's behaviour.
- `rptun_reset()` / `RPTUNIOC_RESET` without `ops->reset` now wait up to
`CONFIG_RPTUN_STATUS_TIMEOUT_MS` (5 s) for the peer and return
`-ETIMEDOUT` if it does not answer; before they returned at once.
- On reboot the peers are reset rather than stopped.
- No API or ABI change.
## Testing
Host: Ubuntu 22.04.5 LTS x86_64, QEMU version 10.2.1, sim: cc (Ubuntu
13.1.0-8ubuntu1~22.04) 13.1.0, qemu-armv8a: aarch64-none-elf-gcc (Arm GNU
Toolchain 12.2.Rel1 (Build arm-12.24)) 12.2.1 20221205
qemu-armv8a:rpserver + qemu-armv8a:rpproxy (rptun over ivshmem), test-only
overlay `# CONFIG_LIBC_ENVPATH is not set`, `# CONFIG_BINFMT_CHECKEXEC is
not set`. Each case boots a fresh pair, pings the peer, then runs one
command on the server and watches the proxy console:
| case (server command) | master | this PR |
| --- | --- | --- |
| `rptun reset /dev/rptun/proxy 0` | proxy asserts in `event_wait()` (trace
below) | proxy `FATAL: Reset by remote core: server reason: 0`, reboots |
| `reboot` | proxy `Panic by remote core` | proxy `FATAL: Reset by remote
core: server reason: 0`, reboots |
| `rptun reset /dev/rptun/proxy 2` | proxy `Panic by remote core` | proxy
`Panic by remote core` |
On master the reset runs `boardctl()` from the rptun interrupt, and the
rpmsg_port_uart reboot notifier waits on an event there:
<details><summary>master, proxy, rptun reset 0</summary>
```
Assertion failed event != ((void*)0) && !up_interrupt_context() == 0: at
file: /sched/event/event_wait.c:120 task: Idle_Task process: Kernel 0x404b215c
Assertion failed panic: at file: /drivers/rptun/rptun.c:918 task: Idle_Task
process: Kernel 0x404b215c
sched/event/event_wait.c:120 (discriminator 3)
sched/event/event_wait.c:279
drivers/rpmsg/rpmsg_port_uart.c:184
drivers/rpmsg/rpmsg_port_uart.c:979 (discriminator 3)
sched/misc/reboot_notifier.c:90 (discriminator 10)
boards/boardctl.c:448
drivers/rptun/rptun.c:923
drivers/rptun/rptun.c:932
drivers/rptun/rptun_ivshmem.c:429
sched/wdog/wd_start.c:152
```
</details>
Every commit builds with 0 warnings on sim:rpserver and
qemu-armv8a:rpserver.
- sim:rpserver + sim:rpproxy: PASS
- [x] `uname -a` (server)
- [x] `rpmsg dump all` (server)
- [x] `rpmsg ping all 10 1024 3 0` (server)
- [x] `uname -a` (proxy)
- [x] `rpmsg ping all 10 1024 3 0` (proxy)
- [x] `mount -t rpmsgfs -o cpu=server,fs=/proc /sp` (proxy)
- [x] `cat /sp/version` (proxy)
- [x] `rpmsg dump all` (proxy)
- qemu-armv8a:rpserver + qemu-armv8a:rpproxy: PASS
- [x] `uname -a` (server)
- [x] `uname -a` (proxy)
- [x] `rpmsg dump all` (server)
- [x] `rpmsg ping all 10 1024 3 0` (server)
- [x] `rpmsg ping all 10 1024 3 0` (proxy)
- [x] `mount -t rpmsgfs -o cpu=server,fs=/proc /sp` (proxy)
- [x] `cat /sp/version` (proxy)
No regression in rpmsg: sim:rpserver/rpproxy and
qemu-armv8a:rpserver/rpproxy boot and pass the fixed rpmsg cases (dump, ping,
rpmsgfs; 15/15). qemu-armv8a was built with LIBC_ENVPATH and BINFMT_CHECKEXEC
disabled, which current master needs for nxinit to start nsh there; unrelated
to this change.
--
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]
