XR700 (R9000 equiv HW) lost all WiFi after FW upgrade

[user123]

I am a R9000 user and long been using Voxel FW and Kamoj addon.
I also have a XR700 which has identical HW as the R9000 that runs the Duma OS.
I just upgraded the XR700 to this new major release Duma 3.3 FW and lost all WiFi in a few minutes. I actually was checking out the new OS for a while until the interface suddenly froze. I reboot and found all WiFi’s were gone. The router was working perfectly before this FW upgrade. I realize the R9000 and the XR700 are commonly subjected to FW loading corruption. I have been unable to bring back the XR700 WiFi.

The XR700 uses the same basic netgear kernel as the R9000.
I have tried simple reset, factory reset, downgrading to an older bin, and reloading bin through TFTP, toggling wifi off/on switch at the advanced wireless setting, tried the wireless on/off button.
Also I have tried Kamoj’s methods of flash_erase nvram, wlan down up toggle
None worked to bring WiFi back. I have also tried loading in the ddwrt bin and no wifi was detected and then tftp back to the netgear bin. Voxel bin will not load in on this

The 2.4/5/60 wifi lights immediately turn on after power button is pushed on, with the 2.4/5 lights dimmer at about 50% intensity. They stay same throughout and the antenna red lights turn on after booting. The WiFi lights coming immediately on is abnormal. System Info tells all WiFi are disabled and the temperature reporting is blank for all Wifi in the debug.htm. It is unlikely that both Atheros controllers and the 60G controller all failed together in HW.

iwconfig did not find any wifi connection, that wlan0, ath1 and ath0 are missing.
lspci showed that the Atheros controllers are not loaded at sysinit.

I think the sysinit section pertaining to wireless must be corrupted, perhaps the phy address or some parameter. Is there a section of the nvram pertaining to the wireless that can be reset to fix this or something else? Is the EEPROM going bad? I have re-flashed a few time before this upgrade but generally not many times on the XR700. All other router functions seem ok.

Anyone has any ideas to bring the Wifi back?
Voxel please help too.
Thx.


Here are some of the errors I saw:
phy for wifi device wifi1 not found
wifi1(qcawifi): disable failed
phy for wifi device wifi0 not found
wifi0(qcawifi): disable failed
(same enable errors)
wifi1 no private ioctls
/sbin/wlan: eval: 1: cannot create /sys/class/net/wifi1/led: Directory nonexistent
…

 

[sfx2000]

I am a R9000 user and long been using Voxel FW and Kamoj addon.
I also have a XR700 which has identical HW as the R9000 that runs the Duma OS.
I just upgraded the XR700 to this new major release Duma 3.3 FW and lost all WiFi in a few minutes. I actually was checking out the new OS for a while until the interface suddenly froze. I reboot and found all WiFi’s were gone. The router was working perfectly before this FW upgrade. I realize the R9000 and the XR700 are commonly subjected to FW loading corruption. I have been unable to bring back the XR700 WiFi.

You may have nuk'ed the ART partition where the radio calibration is located...

Have you tried rolling back to stock firmware?

 

[user123]

You may have nuk'ed the ART partition where the radio calibration is located...

Have you tried rolling back to stock firmware?

Yes, rolled back to much earlier bin mentioned above. No go.

 

[sfx2000]

DumaOS is based on OpenWRT, and OpenWRT on Sysupgrade clears the partitions it is aware of...

You go across builds, even though it's the same HW, partition mapping can change...

QCA radios use ART - so if you can login to the shell via SSH/Telnet, do a DF and see if the ART partition is still available...

(you might also look at the console/dmesg output - athXK outputs to stdout during boot/init)

 

[user123]

DumaOS is based on OpenWRT, and OpenWRT on Sysupgrade clears the partitions it is aware of...

You go across builds, even though it's the same HW, partition mapping can change...

QCA radios use ART - so if you can login to the shell via SSH/Telnet, do a DF and see if the ART partition is still available...

(you might also look at the console/dmesg output - athXK outputs to stdout during boot/init)

Which is the ART partition?
All 5 partitions are there under df: devtmpfs, tmpfs, ubi0verlay ubi1lex, overlayfs
I previously looked at the console debug msg and noted that ath0/1 are not there, and the Atheros controllers are not loaded. I had the debug zip file logged during boot.

Does the ART contain the mac address to load first before any calibration? It seemed to skip everything as it didn't load the radio controllers.

How do I look at the ART ? Does it require serial port?

 

[sfx2000]

ART should be one of the MTD partions in flash - so if you don't see it, it's likely gone...

This would would explain why ath0/ath1 are not present - drivers won't load if the calibration isn't present...

 

[sfx2000]

Does the ART contain the mac address to load first before any calibration? It seemed to skip everything as it didn't load the radio controllers.

MAC is a reference point in ART, the RF cal is what ath is looking for...

ART is Atheros Radio Test - it's basically the info that the chipset is looking for during boot of the WiFi chips.

 

[user123]

I must not be looking for the ART in the right place. I did df on my R9000 and I see the same disk partition as in XR700, nothing says ART.
Where do I look?

How do I read out and restore the ART if I need to?

 

[user123]

You go across builds, even though it's the same HW, partition mapping can change...

You are right. I did go between v1.0.20 which was very old to v3.3.363 (the first v3 beta release) and mapping may have changed. Most people are on V1.0.xx though.

 

[sfx2000]

Watch the bootlogs - UBFFS mounts volumes...

 

[user123]

I found ART, also ART.bak. I will dump them out to compare.

Also the bin version I mentioned was incorrect. It was from v1.0.1.20 to v1.0.1.50. But the new bin is the first Duma 3.3.363 beta release in probably 5 yrs. It is a major OS change. Probably changed the ART.

 

[user123]

How do I know if ART.bak is good to copy back to replace ART? They both have similar structure but some varied content.
It seems to easy to write it back under openwrt. But if it doesn't boot up after, it would be more challenging to replace back in uboot from TFTP.

 

[user123]

Does anyone know when the router first boot up doing sysinit, is there a list in the mtd that it looks to find the devices? If so, where is it? Apparently all my Atheros Wifi controllers & the 60 are not loaded initially. Apparently it is not in the ART as I changed it and nothing changed.

 

[sfx2000]

You'll need to look at uboot and see how the flash is partitioned out... as for sorting out your device, it's possible I'm sure, but you'll likely need some assistance from the NetDUMA folks to rebuild your device.

FWIW - art.bak likely will differ from art - art is the running version, and may differ from art.bak for things like MAC addresses etc...

this is stock R7800 as an example...

cat /proc/mtd 
dev:    size   erasesize  name
mtd0: 00c80000 00020000 "qcadata"
mtd1: 00500000 00020000 "APPSBL"
mtd2: 00080000 00020000 "APPSBLENV"
mtd3: 00140000 00020000 "ART"
mtd4: 00140000 00020000 "ART.bak"
mtd5: 00220000 00020000 "kernel"
mtd6: 01de0000 00020000 "rootfs"
mtd7: 04480000 00020000 "netgear"
mtd8: 02000000 00020000 "firmware"
mtd9: 00080000 00020000 "crashdump"
mtd10: 00380000 00020000 "language"
mtd11: 00120000 00020000 "config"
mtd12: 00120000 00020000 "pot"
mtd13: 00010000 00001000 "m25p80"
mtd14: 0001f000 0001f000 "cert"
mtd15: 0005d000 0001f000 "pot.bak"
mtd16: 001b2000 0001f000 "traffic_meter"
mtd17: 001b2000 0001f000 "traffic_meter.bak"
mtd18: 001b2000 0001f000 "dongle"
mtd19: 037b4000 0001f000 "overlay_volume"

you might find this helpful - Netgear on the QCA devices, usually builds on QSDK, which in turn is based on an older OpenWRT (CC)...

How to restore ART partition

How to restore ART partition Backing up your ART data is a wise precaution as it is irreplaceable. This procedure does not apply to copying ART data from a current router to another! Current routers' ART data are usually specific to the instance of the router and contain calibration data.

openwrt.org

Good luck!

 

[user123]

Mine is partition exactly like the R9000 similar to yours. I have already read out the act.bak partition and restored it on art and used the earliest FW bin. Also md5sum checked both same. Sysinit never loaded the wifi controllers. So I don't think changing the ART did anything. Sysinit was not using the mac address to load all the devices first.

 

[user123]

I meant sysinit was not using the mac address in ART to load.

 

[xTe22-pT]

Hi there, curious to know but did you manage to get your router's wifi working back ?

 

[user123]

Hi there, curious to know but did you manage to get your router's wifi working back ?

The short answer is no.

Here are some details. I loaded an earlier version of the dd-wrt bin and amazing enough the 2.4GHz came back to live working with full power! But no 5GHz, as there seemed to be known issues at the 5G with that bin. Later to recent ddwrt bins won’t even load as they are experimental and something must have changed. There was actually a lot of good help in the other forum. Now going back to the Netgear/Duma bin there is no 2.4G but the 5G was then detected but at very low unusable power. At least it is now detected but calibration power level is not right. No help at all from the NG forum as the moderator only told me it is time to buy their new router. With the above exercises, they told me there is absolutely no problem with the router hardware as both radios are detected, but there was definitely unreliable FW issues. I should have never listened to the upgrade to this new great Duma bin, which rendered my expensive unit useless.

I have spent enough time on this and now moved on to the ASUS AX-88U Pro that does everything the NG did but now with 2.5G ethernet port which makes a big difference; not enough 2.5G ports but is expandable with my 2.5G QNAP switch. No more NG router for me.