Those dont matter at all, cause this is C45 only PHY, there are vendor specific registers that are read for the real status.
Check out the aquantia PHY driver
1 Like
What speeds were the AQR-s running at?
Basically, did you just advertise 2.5 and 5 or actually used them?
10g-1 (WAN) advertised 2.5G and negotiated to my cable modem at 2.5G (it’s a 2.5g modem).
10g-2 (LAN) advertised 2.5g but negotiated at 10g to my 10g switch.
My LAN speeds were normal. Bottleneck was on the 10g-1 WAN link when operating at 2.5g speeds
Ok, now I know how to reproduce it
1 Like
@robimarko now that we're getting snapshots built officially (and finally can use attended-sysupgrade), I'm a bit more ready to delve into the depths of this device.
What exactly would it take to reformat the eMMC completely, and use it as one big unified UBI layout? I understand that not everyone might want to do that, but I think it would be worth the time to implement it, even as a new target (similar to how the Belkin RT3200 has two targets, although one is practically defunct now due to the stock firmware shipping with ECC faults that the new flash driver won't handle gracefully), so that people can utilise the full 4GB storage of the device.
Well, since it uses eMMC you would need to start by modifying the GPT partition table first, probably even modify U-boot.
Other than that somebody would have to invest some time into it, but your ECC points make no sense since this is not NAND and any error correction or bad block management is done internally in the eMMC itself
That makes sense, I presume U-Boot would contain a fixed boot location (one for each kernel entry, for dual-boot support), and doesn't dynamically resolve it from the GPT table.
The ECC bit was to point out that the RT3200 target has some issues that made the switch to the converted UBI layout necessary. I wasn't trying to draw parallels here, beyond the fact that the RT3200 has two maintained targets - one for the stock layout, and one for the new partition layout that utilises the whole NAND.
Well, its easy with fixed-partitions, not so with GPT
1 Like
Need your help with the best branch to build for this device, would the "openwrt-22.03" branch contain the most features?
Since the branch is fairly new, Is there likely to be an issue if I build from openwrt-22.03 ?
kirdes
515
ipq807x is master branch only.
Thanks @kirdes, do you mean that only the master highlighted below works?
kirdes
517
I'm talking about openwrt master, since ipq807x was merged lately.
Please clarify, can I build directly from OpenWrt master, would that work for this device?
Strange, where have I been? I just checked and noticed that this device is now officially supported.
Great work guys, could you confirm if the installation process remains the same?
kirdes
521
Welcome back from the island 
It depends, if you run an old build including Ethernet interface name eth0 - eth5 you have to manually update your network config before flashing the official build.
eth3 -> lan1
eth2 -> lan2
eth1 -> lan3
eth0 -> lan4
eth4 -> 10g-1
eth5 -> 10g-2
2 Likes
Ended up upgrading my unit and pushing it into production (i.e. making it my home gateway).
Got everything working properly, except... The two 10g interfaces seem to have died in the process.
Here's the relevant dmesg lines:
root@yggdrasil:~# dmesg | grep -i "aqua\|10g"
[ 2.048364] aquantia_phy_api_ops_init[2241]:INFO:qca probe aquantia phy driver succeeded!
[ 6.607220] Aquantia AQR113C 90000.mdio-1:08: aqr107_wait_reset_complete failed: -110
[ 6.616302] Aquantia AQR113C 90000.mdio-1:08: attached PHY driver (mii_bus:phy_addr=90000.mdio-1:08, irq=POLL)
[ 6.617061] nss-dp 3a001800.dp5 10g-1: Registered netdev 10g-1(qcom-id:5)
[ 8.634165] Aquantia AQR113C 90000.mdio-1:00: aqr107_wait_reset_complete failed: -110
[ 8.643238] Aquantia AQR113C 90000.mdio-1:00: attached PHY driver (mii_bus:phy_addr=90000.mdio-1:00, irq=POLL)
[ 8.643994] nss-dp 3a007000.dp6-syn 10g-2: Registered netdev 10g-2(qcom-id:6)
[ 14.600443] switch: port 1(10g-1) entered blocking state
[ 14.600488] switch: port 1(10g-1) entered disabled state
[ 14.621738] switch: port 2(10g-2) entered blocking state
[ 14.621786] switch: port 2(10g-2) entered disabled state
[ 14.626572] device 10g-2 entered promiscuous mode
[ 14.631580] device 10g-1 entered promiscuous mode
I'm fairly certain the Aquantia firmware wasn't wiped in the update process. Could it be that a hard reset (which I executed through the reset button on the back) wiped it?
I just upgraded mine too and have not run into any hiccups.
I built from OpenWrt master and flashed using sysupgrade -n. Both 10G ports seem to work in my case
root@OpenWrt:~# dmesg | grep -i "aqua\|10g"
[ 1.982190] aquantia_phy_api_ops_init[2241]:INFO:qca probe aquantia phy driver succeeded!
[ 4.302821] Aquantia AQR113C 90000.mdio-1:08: FW 5.4, Build 4, Provisioning 1
[ 4.310766] Aquantia AQR113C 90000.mdio-1:08: attached PHY driver (mii_bus:phy_addr=90000.mdio-1:08, irq=POLL)
[ 4.316791] nss-dp 3a001800.dp5 10g-1: Registered netdev 10g-1(qcom-id:5)
[ 4.326410] Aquantia AQR113C 90000.mdio-1:00: FW 5.4, Build 4, Provisioning 1
[ 4.338401] Aquantia AQR113C 90000.mdio-1:00: attached PHY driver (mii_bus:phy_addr=90000.mdio-1:00, irq=POLL)
[ 4.340630] nss-dp 3a007000.dp6-syn 10g-2: Registered netdev 10g-2(qcom-id:6)
[ 17.048274] br-lan: port 5(10g-2) entered blocking state
[ 17.048320] br-lan: port 5(10g-2) entered disabled state
[ 17.053281] device 10g-2 entered promiscuous mode
[ 310.311790] nss-dp 3a001800.dp5 10g-1: PHY Link up speed: 1000
[ 310.312005] IPv6: ADDRCONF(NETDEV_CHANGE): 10g-1: link becomes ready
[ 340.471696] nss-dp 3a001800.dp5 10g-1: PHY Link is down
[ 345.672153] nss-dp 3a007000.dp6-syn 10g-2: PHY Link up speed: 1000
[ 345.672246] br-lan: port 5(10g-2) entered blocking state
[ 345.677239] br-lan: port 5(10g-2) entered forwarding state
[ 349.831981] nss-dp 3a001800.dp5 10g-1: PHY Link up speed: 1000
Ah it was indeed a missing firmware case... So fair warning people, if you reset the router with the physical reset button, don't forget to flash it back!
@robimarko do you know if it would be against the licensing of the Aquantia firmware to create an OpenWRT package that would take care of the firmware flashing? I'm thinking of a straightforward first-boot-only flash, and the package would be included by default with the 301w target. That way the manual flashing could be avoided.
There is no license attached, so by default its not redistributable
