Fishwaldo opened a new pull request, #20062:
URL: https://github.com/apache/nuttx/pull/20062
## Summary
The Clock and Reset Generator holds the reset line for every block on the
SoC,
and nothing in this port could see or move one. This registers all 324 lines
with the NuttX reset framework and lists them in `/proc/reset`.
- `arch: Add the CRG reset controller.` — `assert`, `deassert`, `reset`,
`status` over 61 control registers
- `boards: Report the reset lines at startup.` — one line of boot output
- `arch: Name the reset lines through procfs.` — `get_line`, so the listing
carries names
- `boards: Enable the reset procfs entry.` — the two board configurations
## Design notes
**Addressing.** A line is its register index times thirty two plus its bit,
so
decoding needs no table. The ids are sparse: 324 lines in a space of 1952.
**Three masks per register.** `valid` names the bits that are lines, `rdonly`
those the hardware will not let software drive, and `critical` those whose
assertion takes down the system asserting them: interconnect, DDR path, U84
cluster, and the configuration path back to this block. Critical lines are
still readable and releasable; only `assert` and `reset` refuse, with
`-EPERM`.
**The lines are active low, which the manual never states.** Inferred from
the
field naming (`_rstn`, `_arstn`, `_prstn`, `_hrstn`), the reset defaults, and
both vendor Linux drivers. Get it wrong and `deassert` asserts, so the
evidence
is written out beside the table. It matches the hardware: `uart0` reads
released while the console is printing.
**Lines the manual omits** are described rather than skipped: the GPIO resets
at `0x438`, which the register table jumps straight past, and the translation
buffer and NPU E31 lines given only as reserved. Where an instance mapping is
missing, the assumption is recorded at the table row.
Registration writes nothing to the hardware.
## Output
```
[CPU0] clk: registered 264 clocks, 0 failed
[CPU0] reset: 324 lines, 117 held
```
`/proc/reset`, abridged from 324 rows:
```
eic7700x-crg:
0 noc_nsp state:released reg:0x400 bit:0
224 hsp_axi state:released reg:0x41c bit:0
243 hsp_emmc_p state:released reg:0x41c bit:19
288 i2c0 state:asserted reg:0x424 bit:0
416 uart0 state:released reg:0x434 bit:0
448 gpio0 state:asserted reg:0x438 bit:0
1921 spi_slv state:released reg:0x4f0 bit:1
```
Ids naming no line return `-ENODEV` and the framework skips them, which is
what
keeps the listing dense. The name table costs about 8 KiB and is built only
when `CONFIG_RESET_PROCFS` is.
## Testing
Booted on the ESWIN EIC7700 EVB, TFTP loaded from U-Boot.
Compiled with `-Wno-cpp -Werror` in three configurations, the reset code
being
option gated:
| configuration | result |
|---|---|
| `RESET_PROCFS=y` | clean |
| `RESET_PROCFS=n` | clean |
| `DEBUG_RESET=n`, `RESET_PROCFS=y` | clean |
`tools/checkpatch.sh -c -u -m -g master..HEAD` passes.
`CONFIG_RESET_PROCFS` depends on `FS_PROCFS_REGISTER`, so both board
configurations enable it; without it the symbol is dropped when the
configuration is regenerated and the entry never appears.
--
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]
