Experimental build of Gargoyle 1.12.0 for ipq806x architecture routers

[pythonic]

Now that support for the ipq806x architecture has been added to Gargoyle's official sources and @ispyisail and @lantis have made builds available (in the 1.13.x series) I've withdrawn this build to make the space available for another project.

[Lantis]

Gargoyle related?

[Maniac]

Thanks for support of this architecture @pythonic @Lantis! I've been meaning to install 1.13.x soon.

[pythonic]

Gargoyle related?

Thanks for support of this architecture @pythonic @Lantis!

Glad it helped you out!

[d3fz]

Greetings!

First of all, @pythonic thanks a lot for your time with the initial support for the ipq806x architecture, and @Lantis for being in active development for so long and thankfully bringing official support with the 1.13.X build(s).

I've got an Archer C2600 last week, flashed with the latest 1.13.X (Built 20210319-0339 git@396836cd) build, and unfortunately, I'm having issues with this build/architecture. The Wi-Fi is really, really unstable, and the range coverage was also quite bad.

For comparison, my main one is an Archer C7 v2, which is running Gargoyle 1.12.X (Built 20200312-2129 git@f9df5c02) flawlessly for a long time. Great Wi-Fi range and performance. But due to a slow CPU, can't handle more than 75Mbps with QoS enabled.

After research, I've discovered that the ath10k driver has changed (to ath10k-CT) with the lastest OpenWrt releases, and that has caused wifi stability and performance issues with a lot of devices using the QCA998X/9984 radio (D7800, R7500, EA8500, C2600, R7800, etc).

My question is, is there an easy way to test the non-CT ath10k-firmware driver with the 1.13.X build, or do I have to make a full build?

Any feedback is welcome,

Thank you!

[Lantis]

I think you'd need to do your own build, but I'll let Pythonic comment on whether he has a quicker option for you.
The CT/non-CT debate is ongoing and it seems very much dependant on the device and the environment it is operating in (plus clients).

If people felt strongly about it, I'd consider building the non-CT version as well so that it was easier to install and switch to it. However I'd still want to keep CT as the default (follow OpenWrt).
I think we have that ability with the way the Configs are written now.

[pythonic]

@defz:
If you search the threads on the OpenWrt forums you should find some instructions for substituting the non-ct firmware for the ct firmware. My understanding from previous reading there is that the non-ct firmware should work with the ct driver so updating the firmware should at least give you some hints as to whether it helps. The firmware packages from OpenWrt 19.07.7 would be the ones to use (but d/l the packages, copy to the device and install directly - via ssh - rather than try to install over the net) as that is what recent Gargoyle builds would be using (+/- a handful of commits).

@Lantis:
I think that building both the non-ct driver and firmware packages for optional installation would be a good idea as there definitely seem to be situations where one or the other is a better choice. I concur that ct should be the default as with OpenWrt.

[d3fz]

@defz:
If you search the threads on the OpenWrt forums you should find some instructions for substituting the non-ct firmware for the ct firmware. My understanding from previous reading there is that the non-ct firmware should work with the ct driver so updating the firmware should at least give you some hints as to whether it helps. The firmware packages from OpenWrt 19.07.7 would be the ones to use (but d/l the packages, copy to the device and install directly - via ssh - rather than try to install over the net) as that is what recent Gargoyle builds would be using (+/- a handful of commits).

So you're proposing that I could switch between ath10k-CT to non-CT firmware drivers, just by package manipulation? That's really handy. It also makes a lot easier for future testing.

I'm following the guidelines from this similiar issue, and I'll post here how did it go.

Quick question:

My C7 (ath79) ships with the QCA988X radio, and after this command:

Code: Select all

gpkg list-installed | grep ath10k
ath10k-firmware-qca4019 - 2018-05-12-952afa49-1
ath10k-firmware-qca9887 - 2018-05-12-952afa49-1
ath10k-firmware-qca9888 - 2018-05-12-952afa49-1
ath10k-firmware-qca988x - 2018-05-12-952afa49-1
ath10k-firmware-qca9984 - 2018-05-12-952afa49-1
ath10k-firmware-qca99x0 - 2018-05-12-952afa49-1
kmod-ath10k - 4.9.212+2017-11-01-10-2c1e9ee38752675db627fdc1ec2163c1

I see that all ath10k-firmware drivers packages from ath79 architecture were included as well. Since that's not something likely to change, is it safe to uninstall those unused packages to free some space on the router?

Sorry if that's a noob question.

If people felt strongly about it, I'd consider building the non-CT version as well so that it was easier to install and switch to it. However I'd still want to keep CT as the default (follow OpenWrt).
I think we have that ability with the way the Configs are written now.

@Lantis:
I think that building both the non-ct driver and firmware packages for optional installation would be a good idea as there definitely seem to be situations where one or the other is a better choice. I concur that ct should be the default as with OpenWrt.

That's a great idea.

Maybe even giving the alternative ath10k driver option as a plugin install for easing the process to the user?

Thank you for the tip.

[pythonic]

Quick question:

My C7 (ath79) ships with the QCA988X radio, and after this command:

Code: Select all

gpkg list-installed | grep ath10k
ath10k-firmware-qca4019 - 2018-05-12-952afa49-1
ath10k-firmware-qca9887 - 2018-05-12-952afa49-1
ath10k-firmware-qca9888 - 2018-05-12-952afa49-1
ath10k-firmware-qca988x - 2018-05-12-952afa49-1
ath10k-firmware-qca9984 - 2018-05-12-952afa49-1
ath10k-firmware-qca99x0 - 2018-05-12-952afa49-1
kmod-ath10k - 4.9.212+2017-11-01-10-2c1e9ee38752675db627fdc1ec2163c1

I see that all ath10k-firmware drivers packages from ath79 architecture were included as well. Since that's not something likely to change, is it safe to uninstall those unused packages to free some space on the router?

Safe? Probably... Will it actually free flash space? Not if they were included in the original Gargoyle firmware image, because the files are in the SquashFS part of the flash not the overlay file system. If you want to reduce the firmware footprint you'd need to build a custom image, or try a 1.13.x image (which incorporates a feature of OpenWrt 19.07 that removes such superfluous files from the firmware images).

[d3fz]

Safe? Probably... Will it actually free flash space? Not if they were included in the original Gargoyle firmware image, because the files are in the SquashFS part of the flash not the overlay file system. If you want to reduce the firmware footprint you'd need to build a custom image, or try a 1.13.x image (which incorporates a feature of OpenWrt 19.07 that removes such superfluous files from the firmware images).

Understood. Thank you for clarifying.