OpenWrt 23.05.5 - Service Release

XA-Hydra · 2024-09-25T17:36:30.339Z

I've noticed it on both my mediatek and non-mediatek devices

acn · 2024-09-25T17:37:56.551Z

Xiaomi Redmi AX6S / AX3200, upgraded via Luci Attended Sysupgrade, no problems.
Thanks developers

grauerfuchs · 2024-09-25T17:47:56.545Z

It looks like the package list for the builder has been reverted to an earlier point. Some firmware files are present, but those recently added or changed are missing.

wahflz · 2024-09-25T17:53:14.978Z

@XA-Hydra @grauerfuchs

Yes, it was the custom build. Stock firmware worked fine.

XA-Hydra · 2024-09-25T18:15:20.255Z

The E8450 list still has kmod-mt7622-firmware, but not kmod-mt7915-firmware (Hence one of the radios not working). Other packages like kmod-usb3 not there either. Weird.

Synerworks · 2024-09-25T18:15:57.630Z

Checking the pre-built image package contents against those generated from the firmware selector shows that the '/lib/firmware' folder and kernel modules are missing for all devices. The nasty side effect would be to effectively render the box as a 'soft-brick' after installation or following an upgrade without all the required components. Manually adding the missing package components to the build list for each target and versions from SNAPSHOT down does generate the correct image identical to the existing pre-built images.

skynet · 2024-09-25T18:26:46.459Z

Just used the "Flash new firmware image" page in luci to update Linksys WRT3200ACM from 23.05.04 to 23.05.05 using an image with customized packages list (add a few to the base).

Install went fine, and I'm getting ~600Mbps on wifi vs ~100Mbps before.

grauerfuchs · 2024-09-25T18:28:06.637Z

kmod-mt7915-firmware was added as a dependency for the e8450 when the firmware was split into its own package, and that change was somewhat recent.

This is definitely going to cause some really ugly issues for people who are now missing critical packages.

dimitris · 2024-09-25T20:00:59.536Z

Thanks to developers, upgrades from .4 and previous 23.05-snapshot builds went smoothly here with no apparent regressions.

Buffalo WZR-HP-AG300H and Netgear WAX206 with some non-default VLAN config and services.

Re: the MT7622/MT8915E Netgear and firmware files: I built my images locally with imagebuilder which still includes the necessary firmware as default dependencies:

make image PROFILE=netgear_wax206 PACKAGES="wpad-mbedtls mtr luci-ssl tcpdump htop -wpad-basic-mbedtls luci-app-openvpn luci-app-sqm luci-app-tinyproxy openvpn tinyproxy sqm-scripts"

dannil · 2024-09-25T20:43:27.595Z

Looking at the JSON responses, it looks like the firmware selector isn't parsing the specific device packages correctly and adding that to the package list, so it affects multiple devices, irregardless of target. This explains why for example ath10k-firmware-qca9984-ct and kmod-ramoops is missing for R7800 when you compare with https://downloads.openwrt.org/releases/23.05.5/targets/ipq806x/generic/profiles.json.

I know that there's at least a few devices which configure their switch with device packages, so I guess this in worst case could lead to a brick if serial isn't available/the user don't know how to use it (just guessing tho).

GetVladimir · 2024-09-26T00:39:20.222Z

You probably already checked this, but do you happen to have Hardware Offload enabled on the Xiaomi AX3200 after the upgrade? That usually causes the discrepancy in the accounted packets on WAN

elatllat · 2024-09-26T01:12:05.377Z

Linksys EA8500 continues to work well with OpenWrt version 23.05.5.

timur.davletshin · 2024-09-26T04:00:20.666Z

Do we have bugtracker link so I can read more about it?

GetVladimir · 2024-09-26T04:36:12.633Z

If that is indeed the cause, it's not really a bug, as hardware offload usually works by offloading a lot of the NAT away from the CPU.

This also means that the offloaded traffic can't really be tracked or accounted as properly as before.

I don't know if there is an official bug tracker for it, but you can check the shortlog for the updates added since the previous version: https://git.openwrt.org/?p=openwrt/openwrt.git;a=shortlog;h=refs/tags/v23.05.5

Axxel · 2024-09-26T05:31:27.521Z

AP-only WAX202 upgrading via luci to 23.05.5 seems to fail to boot with my existing config. No WiFi networks, no response to the configured IPs on Ethernet. Using the reset button doesn't consistently work, 9 times out of 10 pressing the reset has main LED turn off instead of flashing orange to indicate a reset.

After a lot of power cycling I got a reset, and it boots to 23.05.5, but restoring my config backup takes it right back to the boot issue. A rollback to 23.05.4 with the same restored backup is working fine.

Not sure how to debug this, it seem like the thread thinks some packages are missing in the default 23.05.5 image for WAX202?

fkl7834456 · 2024-09-26T10:20:26.564Z

yes this is a large-scale failure, do not update through the selector yet

lleachii · 2024-09-26T10:45:00.244Z

You missed my post:

OpenWrt 23.05.5 - Service Release Release and security announcements

Same for the WAX202. I had to add kmod-mt7915-firmware to my Firmware Selector package recipe. These recipe changes should be noted somewhere.

Also see:

The OpenWrt Firmware Selector Release and security announcements

It's an issue specific with the firmware selector since the device packages are in the JSON responses, but they're not parsed correctly I guess. I also wrote about this at OpenWrt 23.05.5 - Service Release - #55 by dannil. Note the missing device packages in the list. [Screenshot_21]

axreios · 2024-09-26T13:01:14.210Z

23.05.5 failed to boot on my WRT1900ACS-mvebu-cortexa9 Linksys router after sysupgrade. The 26 Sept development release is up and running.

vortex-5 · 2024-09-26T13:04:59.346Z

Working great on WNDR4300v1 upgraded from 23.05.4

trinidude4 · 2024-09-26T13:32:57.266Z

Building my custom image for WRT32X went fine using the imagebuilder. I still have to test it, but there were no errors building it.

Compiling custom packages using the SDK had errors. I had to copy the files from feeds/base/package to feeds/base to get the custom package to compile. I haven't had time to look into it much, but I think it's related to commit 10cc5fc, but I'm not sure.