Skip to content

Add the DS423+ (geminilake, community tier) - #14

Merged
rylos merged 3 commits into
rylos:mainfrom
vlw:ds423plus
Nov 13, 2026
Merged

rylos merged 3 commits into
rylos:mainfrom
vlw:ds423plus

Conversation

@vlw

@vlw vlw commented Nov 12, 2026 •

Copy link
Copy Markdown
Contributor

The DS423+ is a geminilake board very close to the DS920+. This adds it in the community
tier: installed and checked on my own DS423+, results below; the hardware report is attached.

What the DSM 7.4.1 model.dtb, scemd.xml and pad state of this unit show:

  • the same pins as the DS920+: bay power 20-23, presence 35-38 (active low), USB VBUS 29/30;
    GPIO 17 is an output high and 39/40 inputs under DSM, as on the DS920+ (disk LED enable,
    fan tachometers), on INT3453:00 (UID 1)
  • the same microcontroller (LPSS UART 0, 00:18.0) and console (00:18.2), RTC35390, LED3943
    and DOM identity (FAT id 10EE-589C, ext2 UUID 45e5b07d-...)
  • different: bays 1-2 on an ASMedia ASM1061 (01:00.0 behind 00:13.0, ports 0-1), M.2 slot 1
    behind 00:13.3 and slot 2 behind 00:14.0, power_limit = "14.85,9.9", LAN 1 on 02:00.0
    (the reverse of the DS920+), its own fan curves

Changes: boards/ds423plus.conf (community), a ds423plus table in synology_nas.c (DS920+ GPIO
table, own NIC order, DMI vendor "Synology" with BIOS M.901.00), synohwd fan curves, bay and
M.2 maps, nvme-power-limit, README and platform notes. The kernel builds (6.18.60, 7.3.3)
and the synohwd tests pass; tools/sync-boards.py --check passes.

Tested on a DS423+ (BIOS M.901.00), installed over SSH onto the NVMe in M.2 slot 1:

  • 6.18.60 and 7.3.3: DMI match, staggered bay power 2,1,1, RTC, MACs from macs= in DSM's
    order (lan1 02:00.0, lan2 03:00.0), disk LEDs (LP3943) registered, microcontroller id
    P31_1_..., NVMe power limits per slot, watchdog (wdat_wdt), kernel trial and two reboots

  • fans: this BIOS leaves GPIO 39/40 in ACPI mode, so pinctrl-intel refuses their IRQs
    (-EPERM) and synohwd ran the fans at 99 %. The second commit samples a tachometer without
    an IRQ at 1 kHz in a 1 s window per fan cycle; measured about 950 rpm at 20 % and 1710 rpm
    at 99 %, following the duty, so 39/40 are the tachometers. The DS920+ path (IRQ) is
    unchanged.

  • the installer assembled DSM's arrays (md2, md4) and finish wrote them to mdadm.conf
    despite AUTO -all; I removed the two ARRAY lines by hand. Not investigated further.

  • front panel, checked by eye: short and long beep, disk LEDs green and orange in bay order
    1-4, the green LED blinking on reads of its bay (ata3 trigger), the reset button
    (event: reset button from the microcontroller), the power button held 3 s (power LED
    blinking, shutdown started)

  • power-off with systemctl poweroff and power-on by the RTC alarm (rtcwake, INT1): off at
    00:01:16, on by itself at 00:05:09 for an alarm at 00:05:15

The power button starts the shutdown (power LED blinking); the power-off after it is the
synohwd/systemd path checked above with systemctl poweroff. My one run of it could not
finish because the system was already hung by a GPIO probe of mine. Not checked: the serial
console (it needs the board open).

The synodeb package Version is not bumped in this PR.

Boot log (7.3.3, synology-nas and friends, MACs redacted)
[    0.000000] DMI: SMBIOS 3.0.1 present.
[    0.000000] DMI: Synology DS423+/DS423+, BIOS M.901.00 06/13/2022
[    0.000000] DMI: Memory slots populated: 2/2
[    0.356599] rtc-s35390a 0-0030: registered as rtc0
[    0.356916] rtc-s35390a 0-0030: setting system clock to 2026-11-12T21:05:10 UTC (1794517510)
[    0.357932] synology-nas synology-nas: error -EPERM: request_irq(125) syno_tach_irq 0x0 syno-fan1
[    0.357946] synology-nas synology-nas: fan 1 tach irq: -1, polling the gpio
[    0.358050] synology-nas synology-nas: error -EPERM: request_irq(126) syno_tach_irq 0x0 syno-fan2
[    0.358060] synology-nas synology-nas: fan 2 tach irq: -1, polling the gpio
[    0.358106] synology-nas synology-nas: bay 1 power on
[    0.358131] synology-nas synology-nas: bay 2 power on
[    3.401734] synology-nas synology-nas: delay 11 s before spin-up group 2
[   14.409686] synology-nas synology-nas: bay 3 power on
[   17.481680] synology-nas synology-nas: delay 11 s before spin-up group 3
[   28.489676] synology-nas synology-nas: bay 4 power on
[   31.561811] synology-nas synology-nas: Synology DS423+ platform ready
[   35.743378] synology-nas: eth0 (0000:02:00.0): MAC address xx:xx:xx:xx:xx:xx from macs=
[   35.759200] synology-nas: eth1 (0000:03:00.0): MAC address xx:xx:xx:xx:xx:xx from macs=
Hardware report (DSM 7.4.1)

Measured under DSM 7.4.1-90080 Update 1 (kernel 4.4.302+), BIOS M.901.00, 18 GB RAM.
The full tools/collect-report.sh output is available on request.

Identity

  • DMI: sys_vendor Synology, product_name DS423+, board_name DS423+
  • syno_hw_version=DS423+, model.dtb model = "synology_geminilake_423+"
  • DOM: synoboot1 FAT id 10EE-589C, synoboot2 ext2 UUID
    45e5b07d-4783-4867-a369-f99c0cd1e610, 245760 sectors (the same as the DS920+)
  • DSM command line: syno_ttyS0=pciserial,0x0:0x18.0x2,115200
    syno_ttyS1=pciserial,0x0:0x18.0x0,115200 (console on 00:18.2, microcontroller on 00:18.0)

PCI

function device role
00:12.0 8086:31e3 SoC AHCI bays 3-4 (ata1, ata2)
00:13.0 -> 01:00.0 1b21:0612 ASMedia ASM1061 bays 1-2 (ata3, ata4); the DS920+ has a Marvell 88SE9235 here
00:13.1 -> 02:00.0 10ec:8168 eth0 (LAN 1) under DSM
00:13.2 -> 03:00.0 10ec:8168 eth1 (LAN 2) under DSM (the reverse of the DS920+)
00:13.3 -> 04:00.0 NVMe M.2 slot 1 (model.dtb nvme_slot@1)
00:14.0 -> 05:00.0 NVMe M.2 slot 2 (model.dtb nvme_slot@2)
00:18.0 8086:31bc LPSS UART 0 front-panel microcontroller
00:18.2 8086:31c0 LPSS UART 2 serial console

model.dtb (decoded)

  • syno_spinup_group = <2 1 1>, syno_spinup_group_delay = <10>
  • power_limit = "14.85,9.9", syno_image_config = "TOWER_4_Bay-M2X1"
  • internal_slot@1..4: power_pin_gpio 20, 21, 22, 23; detect_pin_gpio 35, 36, 37, 38
    (active low); led_type = "lp3943", LEDs syno_led0..7
  • slot 1/2: pcie_root = "00:13.0,00.0" ata_port 0/1; slot 3/4: pcie_root = "00:12.0"
    ata_port 0/1
  • usb_slot@1/2: vbus syno_gpio 29, 30 (active high), ports 1-1/2-1 and 1-2/2-2
  • nvme_slot@1 00:13.3, nvme_slot@2 00:14.0

GPIO (INT3453:00, UID 1, community northwest), pad state under DSM

  • outputs high: 17, 20, 21, 22, 23, 27, 28, 29, 30, 62, 67
  • inputs: 19, 35, 36, 37, 38 (all four bays populated: read 0), 39 (1), 40 (0), 42
  • 39/40 are inputs as on the DS920+ (fan tachometers there); not in model.dtb

Other devices

  • ACPI RTC35390 (rtc-s35390a), LED3943, INT3453:00..03 — as on the DS920+
  • scemd.xml fan curves (DUAL_MODE_HIGH/LOW), in synohwd form:
    • high: disk (0,20) (41,30) (46,50) (50,70) (53,99), shutdown 61; m2 (0,20) (55,50)
      (60,70) (65,99), shutdown 70; cpu (0,20) (45,30) (55,50) (65,70) (75,99), shutdown 90
    • low: disk (0,20) (46,30) (52,50) (55,70) (58,99), shutdown 61; m2 as high; cpu (0,20)
      (50,30) (60,50) (70,70) (80,99), shutdown 90

Checked on Debian

Fan tachometers (39/40, 2 pulses), restart and power-off through the microcontroller, LED
order: see the results above. Serial console address: not checked.

🤖 Generated with Claude Code

vlw added 2 commits November 13, 2026 00:12
Board file, platform driver table (DS920+ pins, its own NIC order), synohwd fan curves,
bay and M.2 maps from the DSM 7.4.1 model.dtb and scemd.xml, NVMe power limits.
The DS423+ BIOS (M.901.00) leaves GPIO 39/40 in ACPI mode, where
pinctrl-intel refuses IRQs (-EPERM): the fans were never measured and synohwd
ran them at 99 %. Those tachometers are now sampled at 1 kHz in a 1 s window
per fan cycle.
@vlw vlw changed the title Add the DS423+ (geminilake, work in progress) Add the DS423+ (geminilake, community tier) Nov 12, 2026
@rylos
rylos merged commit 498c437 into rylos:main Nov 13, 2026
@rylos

rylos commented Nov 13, 2026

Copy link
Copy Markdown
Owner

Thanks again! This is now published: synodeb 4.9 (DS423+ fan curves, bay and M.2 maps, NVMe power limits, board file) and kernel 6.18.63-11 with the polled tachometers, from the synodeb apt repository; synodeb-setup.sh lists the DS423+ in the community tier.

The mdadm.conf issue you found is fixed too (6ec8b81): the installer now writes only the arrays the new system uses, and leaves the other ones (such as DSM's) commented out. It is in the new installer, which is published together with the kernel.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants