machine: add the 2G5F4_10G2_V1.01 (HYWS-SGT0204S) target

The hongyavision HYWS-SGT0204S is sold as an unmanaged 4x2.5G + 2x10G SFP+
switch and carries the 2G5F4_10G2_V1.01 marking on the PCB. It is a sibling of
the KP-9000-6XH-X2 (2M-PCB43-V1.1), so machine/2G5F4_10G2_V1_01.c is that board
definition with these changes:

- The two SFP+ cages (physical ports 5 and 6) are wired the other way round:
  physical 5 is on SDS1 with GPIO30/GPIO37, physical 6 on SDS0 with
  GPIO50/GPIO10. Copper ports 1-4 match the KP-9000 layout.
- No reset button is fitted and the case has no opening for one, so reset_pin
  is GPIO_NA. GPIO54, the reset pin on the KP-9000 layout, reads as permanently
  pressed here.
- The board definition sets flash_sio, so the flash is read with single-IO: the
  fitted 512 KB Bright Moon T25S40 cannot be read in DIO mode on this board, an
  image that leaves DIO enabled does not come up at all, while the same image
  with single-IO boots and runs normally (two units).
- The board is added to the machine_custom_init() group that sets
  LED_GLB_IO_EN bit 6, otherwise the RTL837x "solid" conditions (LINK/speed)
  never drive the LEDs and the 2.5G LED glows faintly at 100M/1000M.

flash_sio is a new field of struct machine that rtlplayground.c hands to
flash_init(); only this board sets it, every other machine keeps DIO. It costs
management-plane speed only, the ASIC keeps forwarding in hardware.

The stock firmware runs a factory test console on the serial header at 9600
baud, whose silkscreen labels are shifted by one pin.

Verified: warning-free build, `make machine_check` passes for every target, the
host tests pass, and a control build of this target without the LED hook differs
from the KP-9000-6XH_X2_V1_1 image in 44 bytes: all inside the machine
description (including flash_sio) and the image CRC16, nothing else in the
firmware moved. On the board the single-IO image boots and runs on two units;
port order, SFP cages and LED behaviour come from the machine definition
verified on this board earlier.

doc/devices/2G5F4_10G2_V1.01.md

@@ -0,0 +1,147 @@
+hongyavision HYWS-SGT0204S
+
+RTL8372-based 4×2.5G + 2×SFP+ unmanaged switch.
+
+Also marked on the PCB as `2G5F4_10G2_V1.01`.
+
+Stock firmware runs a factory test console on UART at 9600 baud 8N1 (banner
+`RTL8371B start`, prompt `RTL8371B:`). There is no management CLI and no Web
+interface.
+
+Using SPI clamp in-board is the only method for initial installation and update.
+Erase the whole chip and verify the write (`flashrom -v`) before powering the
+board, so an incomplete write is caught before the board is powered up.
+
+Stock flash chip is 512 KB (Bright Moon `T25S40`, 4 Mbit, SOP-8), which the
+RTLPlayground image fills exactly.
+
+### Label specifications
+
+- **Manufacturer**: hongyavision
+- **Model**: HYWS-SGT0204S
+- **Ports**:
+  - 4 × RJ45: 10/100/1000/2500 Mbps
+  - 2 × SFP+: 1G / 2.5G / 10G
+
+### What works
+
+- All four 2.5GBASE-T RJ45 ports at 10/100/1000/2500 Mbps
+- Both SFP+ ports supporting 1G, 2.5G and 10G modules
+- Copper LEDs: white (2.5G) and blue (link/activity)
+- SFP LEDs: one single LED per port (link/activity)
+- SFP EEPROM and DDM readable on both cages, verified with a 10G module
+  (`Rate: 0x67`, module identity and DDM values via the `sfp` console command)
+- Serial console at 115200 8N1 (RTLPlayground) / 9600 8N1 (stock)
+
+### Not verified
+
+- The reset button footprint is unpopulated and the case has no opening for it,
+  so it could not be tested.
+
+### PCB overview
+
+**Board markings**
+- Top silkscreen: `2G5F4_10G2_V1.01`
+
+**Key components**
+- Main switch: RTL8372 (RTL8371B family), reported as `RTL8372` (non-N) by
+  RTLPlayground. The chip sits under a heatsink and its package marking is not
+  readable.
+- U1: SPI NOR flash, Bright Moon `T25S40`, 4 Mbit / 512 KB, SOP-8
+- A 3-hole reset button footprint is present but unpopulated, and the case has
+  no opening for it, so there is no usable reset button
+
+### Port layout
+
+| Front panel position | Logical port | Physical port | Type    |
+|----------------------|--------------|---------------|---------|
+| SFP left             | 8            | 5             | SFP+    |
+| RJ45 1               | 4            | 1             | Copper  |
+| RJ45 2               | 5            | 2             | Copper  |
+| RJ45 3               | 6            | 3             | Copper  |
+| RJ45 4               | 7            | 4             | Copper  |
+| SFP right            | 3            | 6             | SFP+    |
+
+Copper ports 1–4 match the KP-9000-6XH-X2 (2M-PCB43-V1.1) layout, which this
+board is a sibling of. The two SFP+ cages are wired the other way round
+compared to that board: physical port 5 is on SDS1 and physical port 6 is on
+SDS0.
+
+### Serial console
+
+The board has three through-holes used as the serial console. **The silkscreen
+labels are wrong**: they are shifted relative to the actual signals.
+
+| Silkscreen label | Actual signal |
+|------------------|---------------|
+| TX               | GND           |
+| RX               | TX            |
+| GND              | RX            |
+
+Wiring (following the measured signals, not the labels):
+
+| USB-TTL adapter | Board pad            |
+|-----------------|----------------------|
+| GND             | pad labelled `TX`    |
+| RX              | pad labelled `RX`    |
+| TX              | pad labelled `GND`   |
+
+- **Settings**: 115200 baud / 8N1 / 3.3V TTL (RTLPlayground), 9600 baud for the
+  stock firmware
+- The stock firmware only prints a few bytes and a factory test menu
+  (`?`, `rst`, `f`, `r`, `10`, `100`, `g1`, `sd`, `eye`, …); it is not an
+  interactive management console
+
+### LED configuration
+
+Case silkscreen:
+
+```
+White : 2.5G
+Blue  : 1000M/2.5G L/A
+```
+
+| Port       | LED          | Linked, idle | With traffic |
+|------------|--------------|--------------|--------------|
+| Copper     | white        | 2.5G only    | same         |
+| Copper     | blue         | solid        | blinks (L/A) |
+| SFP        | single LED   | solid        | blinks (L/A) |
+
+Measured LED-ID mapping (from the RTL837x LED set registers):
+
+| LED-ID | Function                        |
+|--------|---------------------------------|
+| LED-ID0| copper blue LED                 |
+| LED-ID1| copper white LED                |
+| LED-ID2| SFP single LED                  |
+
+**Important**: `RTL837X_REG_LED_GLB_IO_EN` bit 6 must be set in
+`machine_custom_init()`. Without it the RTL837x "solid" conditions (LINK/speed)
+do not drive the LEDs at all and only `ACT` does, which makes the white LED
+glow faintly at non-2.5G speeds.
+
+### SFP GPIO assignments
+
+| SFP               | pin\_detect (ModAbs)          | pin\_los         | SerDes | I2C SDA                 | I2C SCL                  |
+|-------------------|-------------------------------|------------------|--------|-------------------------|--------------------------|
+| Left (logical 8, physical 5) | GPIO30\_ACL\_BIT3\_EN | GPIO37   | SDS1   | GPIO39\_I2C\_SDA4       | GPIO40\_I2C\_SCL3\_MDC1  |
+| Right (logical 3, physical 6)| GPIO50\_I2C\_SCL2\_UART1\_TX | GPIO10\_LED10 | SDS0 | GPIO41\_I2C\_SDA3\_MDIO1 | GPIO40\_I2C\_SCL3\_MDC1 |
+
+`pin_tx_disable` is GPIO\_NA on both ports. Both I2C buses are confirmed
+working: with the same 10G module inserted, the `sfp` console command reads its
+vendor and model from the left cage (`Slot 1`) and from the right cage
+(`Slot 2`), DDM values included. The insertion also shows which SerDes drives
+each cage - the left cage reports `sds_config sds: 01` and the right cage
+`sds: 00`, which matches the table above.
+
+### Machine target
+
+Build for this device with `make MACHINE=2G5F4_10G2_V1_01`, which selects the
+`machine/2G5F4_10G2_V1_01.c` board definition. It is derived from
+`machine/KP_9000_6XHML_X2.c` with the SFP cage order and the reset pin changed.
+The fitted T25S40 cannot be read in DIO mode on this board: an image that leaves
+DIO enabled does not come up at all, both at the default read clock divider of 4
+and at 8 (two boards tested), while the same image with single-IO boots and runs
+normally. The board definition therefore sets `flash_sio`, which makes
+`flash_init()` use `CMD_FREAD` at divider 8. Only the management plane is
+slower, the ASIC keeps forwarding in hardware.