Adding OpenWrt support for QNAP QHora-301W

sppmaster · June 9, 2025, 4:24pm

root@QNAP:~# mdio 90000.mdio-1
 DEV      PHY-ID  LINK
0x00  0x00000000  down
0x08  0x00000000  down
0x10  0x004dd0b1  up
0x11  0x004dd0b1  up
0x12  0x004dd0b1  up
0x13  0x004dd0b1  up
0x14  0x004dd0b1  down
0x15  0x04820a05  down
root@QNAP:~# mdio 90000.mdio-1 mmd 0:4 raw 0xC441
0x0008

root@QNAP:~# mdio 90000.mdio-1 mmd 0x15:4 0xC441
0x0000
root@QNAP:~# mdio 90000.mdio-1 mmd 0x15:4 0xC440
0x0000
root@QNAP:~# mdio 90000.mdio-1 mmd 0x15:4 0xC442
0x0000

@robimarko Is this the needed info?
@rmandrad
With kernel 6.6.92 (latest build I have before switching to kernel 6.12) 10g-2 port doesn't work (device connected to it cannot get link) but there is no error message at all.
With kernel 6.12, 10g ports are reversed 10g-1 doesn't work with message autoneg time out but 10g-2 works OK.
Firmware load fails on 6.12 when I have device connected to the 10g-1 port at boot.
The registers are the same on k6.6 and 6.12.

rmandrad · June 10, 2025, 2:27am

sppmaster:

root@QNAP:~# mdio 90000.mdio-1
 DEV      PHY-ID  LINK
0x00  0x00000000  down
0x08  0x00000000  down
0x10  0x004dd0b1  up
0x11  0x004dd0b1  up
0x12  0x004dd0b1  up
0x13  0x004dd0b1  up
0x14  0x004dd0b1  down
0x15  0x04820a05  down
root@QNAP:~# mdio 90000.mdio-1 mmd 0:4 raw 0xC441
0x0008

root@QNAP:~# mdio 90000.mdio-1 mmd 0x15:4 0xC441
0x0000
root@QNAP:~# mdio 90000.mdio-1 mmd 0x15:4 0xC440
0x0000
root@QNAP:~# mdio 90000.mdio-1 mmd 0x15:4 0xC442
0x0000

those registers show the same on my qnap with 6.12

rmandrad · June 10, 2025, 2:42am

i agree it makes no sense ... but this is the only way I am able to get the ports working! ...

robimarko · June 10, 2025, 8:39am

Hm, its rather weird that autoneg times out since its clearly enabled as 0xC441 is set to 0x8

sppmaster · June 10, 2025, 9:19am

So for now a big warning to those who actively use 10G ports about future upgrades. Too bad if this device is stuck at this point.

robimarko · June 10, 2025, 10:46am

Well, somebody has gotta debug this, I dont have the time currently to work on this

sppmaster · June 10, 2025, 11:05am

Yea, just to put the flag, the issues started around this commit on kernel 6.6.
@Ansuel

suyuan168 · June 11, 2025, 5:00am

I think it's a pcie problem, because I have an ipq8072a using the 6.6 kernel and both PCIs are working fine. One of the pcie connections
root@openwrt:~# lspci
0000:00:00.0 PCI bridge: Qualcomm Technologies, Inc Device 1002
0000:01:00.0 USB controller: Renesas Technology Corp. uPD720201 USB 3.0 Host Controller (rev 03)
0001:00:00.0 PCI bridge: Qualcomm Technologies, Inc MDM9x55 LTE Modem [Snapdragon X16]
root@openwrt:~# lsusb
Bus 001 Device 001: ID 1d6b:0002 Linux 6.6.83 xhci-hcd xHCI Host Controller
Bus 002 Device 001: ID 1d6b:0003 Linux 6.6.83 xhci-hcd xHCI Host Controller
Bus 002 Device 002: ID 2c7c:0801 Quectel RM520N-GL
Bus 002 Device 003: ID 2c7c:0801 Quectel RM520N-GL
Bus 003 Device 001: ID 1d6b:0002 Linux 6.6.83 xhci-hcd xHCI Host Controller
Bus 004 Device 001: ID 1d6b:0003 Linux 6.6.83 xhci-hcd xHCI Host Controller
Bus 004 Device 002: ID 2c7c:0801 Quectel RM520N-GL
Bus 005 Device 001: ID 1d6b:0002 Linux 6.6.83 xhci-hcd xHCI Host Controller
Bus 006 Device 001: ID 1d6b:0003 Linux 6.6.83 xhci-hcd xHCI Host Controller
root@openwrt:~#
But after I used the 6.12 kernel, the Quectel RM520N-GL USB could not be recognized.

sppmaster · June 13, 2025, 12:53pm

I created an issue on github.

github.com/openwrt/openwrt

10G ports of QNAP QHora-301w Qualcommax not working properly with Main branch

opened 12:52PM - 13 Jun 25 UTC

sppmasterspp

QNAP QHora-301w (Qualcommax) has two 10G ports Aquantia AQR113c. This is dedicat…ed device forum thread - https://forum.openwrt.org/t/adding-openwrt-support-for-qnap-qhora-301w/96934/922 There are several reports from different users that one or the other 10G ports just don't work properly and that changes with different commits of OpenWrt Main branch. They cause different issues and produce different error messages. There were/are different issues with the AQR 10G Firmware not loading or simply the 10G-1 port gives errors. Recently there was a commit that broke 10g-2 port with kernel 6.6 and later 10g-1 port with kernel 6.12. https://github.com/openwrt/openwrt/commit/cc50cac8a0fd0e12fc9e2a7fa7d14f99a2a5ffae reported here - https://forum.openwrt.org/t/adding-openwrt-support-for-qnap-qhora-301w/96934/868?u=sppmaster and confirmed by several users. Some examples from the forum. ``` Error messages that can be seen: [ 30.117803] Aquantia AQR113C 90000.mdio-1:08: CRC mismatch: calculated 0x7011 PHY 0xffff [ 30.117853] Aquantia AQR113C 90000.mdio-1:08: firmware loading failed: -22 [ 30.135101] Aquantia AQR113C 90000.mdio-1:08: Direct firmware load for marvell/AQR-G4_v5.4.C-AQR_CIG_WF-1945_0x8_ID44776_VER1630.cld failed with error -2 [ 30.135166] Aquantia AQR113C 90000.mdio-1:08: Falling back to sysfs fallback for: marvell/AQR-G4_v5.4.C-AQR_CIG_WF-1945_0x8_ID44776_VER1630.cld [ 30.382519] kmodloader: loading kernel modules from /etc/modules.d/* [ 30.528283] urngd: v1.0.2 started. [ 40.853318] Aquantia AQR113C 90000.mdio-1:08: failed to find FW file marvell/AQR-G4_v5.4.C-AQR_CIG_WF-1945_0x8_ID44776_VER1630.cld (-12) [ 40.853383] Aquantia AQR113C 90000.mdio-1:08: probe with driver Aquantia AQR113C failed with error -12 ``` ``` [ 267.771052] nss-dp 3a007000.dp6-syn 10g-2: PHY Link is down [ 267.771413] br-lan: port 2(10g-2) entered disabled state [ 302.672476] uniphy autoneg time out! [ 303.131636] nss-dp 3a001800.dp5 10g-1: PHY Link up speed: 2500 [ 303.131902] br-lan: port 1(10g-1) entered blocking state [ 303.136387] br-lan: port 1(10g-1) entered forwarding state [ 309.371330] nss-dp 3a001800.dp5 10g-1: PHY Link is down [ 309.371690] br-lan: port 1(10g-1) entered disabled state [ 315.612185] nss-dp 3a001800.dp5 10g-1: PHY Link up speed: 2500 [ 315.612451] br-lan: port 1(10g-1) entered blocking state [ 315.616932] br-lan: port 1(10g-1) entered forwarding state [ 315.632585] uniphy autoneg time out! [ 316.651357] nss-dp 3a001800.dp5 10g-1: PHY Link is down [ 316.651718] br-lan: port 1(10g-1) entered disabled state [ 322.891653] nss-dp 3a001800.dp5 10g-1: PHY Link up speed: 2500 [ 322.891968] br-lan: port 1(10g-1) entered blocking state [ 322.896406] br-lan: port 1(10g-1) entered forwarding state [ 323.003107] uniphy autoneg time out! ``` Or even there was a case with the 10g-2 port cannot pass any traffic but there was no error message at all.

rmandrad · August 17, 2025, 5:39pm

I have loaded the firmware from @kirdes https://github.com/kirdesde/nbg7815_gpl/blob/master/target/linux/ipq/ipq807x_64/prebuilt_images/AQR_ethphyfw.mbn repo into the mtd10

and have bootcmd=aq_load_fw 0 && aq_load_fw 8 && bootipq

got both 10g-1 and 10g-2 to work … note this is qosmio nss version

[    2.503188] ssdk_dt_parse_interrupt[942]:INFO:intr-gpio does not exist
[    2.545397] aquantia_phy_api_ops_init[1485]:INFO:qca probe aquantia phy driver succeeded!
[    4.393241] regi_init[2525]:INFO:Initializing HPPE Done!!
[    4.393370] regi_init[2574]:INFO:qca-ssdk module init succeeded!
[    4.400080] EDMA ver 1 hw init
[    4.403962] EDMA HW Reset completed succesfully
[    4.406604] Num rings - TxDesc:1 (23-23) TxCmpl:1 (7-7)
[    4.410998] RxDesc:1 (15-15) RxFill:1 (7-7)
[    4.594033] Qualcomm QCA8075 90000.mdio-1:10: attached PHY driver (mii_bus:phy_addr=90000.mdio-1:10, irq=POLL)
[    4.663530] Qualcomm QCA8075 90000.mdio-1:11: attached PHY driver (mii_bus:phy_addr=90000.mdio-1:11, irq=POLL)
[    4.733515] Qualcomm QCA8075 90000.mdio-1:12: attached PHY driver (mii_bus:phy_addr=90000.mdio-1:12, irq=POLL)
[    4.813545] Qualcomm QCA8075 90000.mdio-1:13: attached PHY driver (mii_bus:phy_addr=90000.mdio-1:13, irq=POLL)
[    4.823258] Aquantia AQR113C 90000.mdio-1:08: attached PHY driver (mii_bus:phy_addr=90000.mdio-1:08, irq=POLL)
[    4.833288] Aquantia AQR113C 90000.mdio-1:00: attached PHY driver (mii_bus:phy_addr=90000.mdio-1:00, irq=POLL)

i haven’t made any changes apart from loading from uboot… don’t really know what is going on

sppmaster · August 17, 2025, 5:47pm

Is the NSS build a current one with all the latest commits?

@rmandrad

Can you check this mbn firmware with the command?

dd if=/dev/mtd10 | strings | grep v5

rmandrad · August 18, 2025, 3:56am

v5.4.4 DNI NBG7815 112019 17:28:07

yes latest

Znullptr · August 26, 2025, 6:40pm

ugh, main 10-g phy stopped working today, see this has become a thing… it’s been functional for 3 years.

+1 for no errors, phy reports connection, no sign it’s not functional, no DHCP and static fails

it’s the nss version only? (that was a lot of annoyance and made little to no difference for me, will try switching back).

^ same fw / v5.4.4 DNI NBG7815 112019

Znullptr · August 29, 2025, 12:37pm

It is not consistent, powering it off for some time && removing both connections from 10g phy’s during boot and it works now..

sppmaster · August 29, 2025, 1:09pm

Have you always/peviously used that firmware or you used stock or any other 10G FW before @rmandrad suggested workaround using v5.4.4 DNI NBG7815 112019?

Znullptr · September 1, 2025, 12:06am

I have updated it in the past, but not recently.. it’s been using that firmware for some time.

sppmaster · October 6, 2025, 5:31pm

@rmandrad Thanks a lot for providing this workaround. I've just tried it and It really works. Latest commits from the main branch.
Just a blind shot - maybe a firmware bug in other 10G FW versions.
There are at least several others 10G FW versions floating around the web. Maybe they deserve a try too.
@robimarko Maybe you can try it if that can give you any clue about the actual issue.

robimarko · October 6, 2025, 5:34pm

I mean, I have used the same FW version from the start and only recently this started.

It really screams some kind of MDIO issue to me, but I cannot narrow it down

sppmaster · October 6, 2025, 5:39pm

But I don't understand why the v5.4.4 DNI NBG7815 112019 17:28:07 version of the firmware works for @rmandrad and me.
Does that one work for you if loaded by U-boot?

robimarko · October 6, 2025, 6:14pm

Look, I dont care whether it works via U-Boot, it must work when loaded by the kernel as it worked just fine