Netgear R7800 23.05-rc2 poor gigabit network performance

What I proposed happened in the past on this project: LEDE forked OpenWrt because people were unhappy with the development of the project and wanted a different direction. Afterwards LEDE took over the OpenWrt name. Over simplified but this happens all the time.

There are countless other examples for this in Linux distributions and open source software projects.

openwrt supports over 1700 ! devices - how do you see regression testing being performed ?

1 Like

Seek for feedback from the user base. But most importantly, listen to the user base.
I am constantly building OpenWRT for 6 different router models, the most popular in Europe because they have the best specs and because they are easily found either on eBay or Amazon for a good price. Linksys EA8300, Linksys MR8300, Linksys EA8500, Linksys WRT 32x, Netgear R7800, BT Home Hub 5 and I give feedback when I find something has broken. I don't feel like developers are willing to listen however. They are quite 'touchy'. Linksys WRT 32x and Netgear R7800 now are still on OpenWRT 22.03. I upgraded my WRT32x to 22.03 only last month after years on 19. And it still has issues with WiFi. OpenWRT 23.05 introduced also routing issues for WRT32x, which I reported, but again, nobody appeared to be interested.

Developers should take the advice to find a working kernel for the release and stick to it.

well - openwrt is built with code - not with feedback. While I agree that feedback is important for the well being of a project ultimately code is what drive it further and from 6 devices to 1700 is a long gap that needs to be filled.
As for popularity this is quite subjective - what's popular in Europe doesn't and most likely isn't in Asia or South America and openwrt targets all those 1700+ devices.

@ezplanet The bugs with wifi on the wrt32xx routers are not the falt of OpenWrt mate. The mwlwifi drivers can not be worked on by any of the devs of openwrt. https://github.com/kaloz/mwlwifi/commits/master

in @ezplanet's world, the mwlwifi wifi target should have stayed on 19.07 kernel level forever, so the wifi would keep working flawlessly ...

@frollic reading @ezplanet post, wifi is not working fine in 19.07 and 22.03, seems it's nothing related to @ezplanet 's world.

@Klingon this looks like what @frollic was tring to emfasize [quote="ezplanet, post:14, topic:169035"]
Maintenance for OpenWRT 19 should be resumed until the newer releases are properly fixed.
[/quote]

No, @frollic deliberately posted a sarcastic remark and stopped any chance to have a constructive debate.
There is no point to try and reason further.

rc3 is released - could you test it ?

I think the same about your thread. You want developers of this open source software project to change their architecture choice for your existing devices, so that you get acceptable performance with older kernels.

I understand your idea, but it does not make sense. As already explained: kernel 4.14 from OpenWrt 19.07 will finally be end of life and out of support from upstream in January 2024.

There will be nobody to fix even the worst security problems in this kernel. You will have to support this kernel by yourself. I still think you don’t understand what this means what you are asking for.

But: you are free to fork OpenWrt code and create your own EzplanetWrt 19.07 project, fix upcoming security issues in the unsupported Linux kernel 4.14 beginning from February 2024. This is possible and you can do it. It’s not a childish idea or a mental issue.

If you would not like to do it, why should someone else here take the burden to support end of life Linux kernels from OpenWrt 19.07? This does not make sense.

And to not receive further insults from others: I am also affected by this performance issue on my R7800. But: I see the need to go forward and run newer Linux kernels because nobody here is able to pay for or support end of life Linux kernels.

There were already attempts to fix kernel 5.15 performance for ipq806x, with the new default of the performance governor. This is what we have for now. A performance fix cannot be forced. A solution will not be found if you ask enough, only by doing the work of analyzing the issue and provide code that helps fixing the issue. This is constructive help. There is a need of doing more analysis and attempt work for fixing the performance issues with kernels 5.15 for 23.05 and 6.1 for master. But it does not make sense to complain about OpenWrt updating to newer kernels and ask for updates of the already abandoned end of life 19.07 branch.

4 Likes

The linux kernel maintainers will drop support for older Kernel version however some are still maintained and will get security patches e.g. 4.4 and 4.19 according to:

https://wiki.linuxfoundation.org/civilinfrastructureplatform/start

I know DDWRT for its older Broadcom routers uses K4.4. and it is getting very regular updates via CIP

I am not saying we should stay on these older kernels though :slight_smile:

1 Like

If enough people agree to start riding dead horses and continue with kernel 4.14 with third party extended long term support beginning February 2024: why not? In theory it’s possible.

I just don’t see that somebody here in OpenWrt likes to go this way for now and continue to support the already abandoned end of life 19.07 branch. I only see that some users would like to get this support but nobody actually working on this. OpenWrt 19.07:
https://git.openwrt.org/?p=openwrt/openwrt.git;a=shortlog;h=refs/heads/openwrt-19.07

Maybe, just maybe there is some good news for the lacklustre performance of the R7800 with 23.05, see:

1 Like

The kernel is the foundation over which OpenWRT or any other operating system are built, but it does not require constant changes. OpenWRT takes its shape and form from the ecosystem of configuration management and services integrated together to deliver what we conventionally call a router.

The kernel plays a fundamental role in supporting the hardware, the bare metal over which OpenWRT is built. So I understand that in order to support new hardware in some cases it might be easier to upgrade to newer kernel releases for which that hardware support was built. But that isn't always the case. SOC manufacturers do not chase the latest kernel, often they provide drivers for well established kernel releases first. And in most cases, that will be that, no more updates.

From our recent experience we have seen that newer kernel releases break the hardware support that we need for some well established and popular devices. In some cases because of design choices. So we know that in these cases, there isn't going to be a future fix.

So, what are the values do you believe new kernel releases add, that OpenWRT cannot move forward without?

We need to weigh the value of new kernels against OpenWRT platform stability and reliability. What am I missing?

1 Like

Kernel 5.10 and 5.10 EOL is in 3 years, in fact EOL for 5.10 is 1 month later than 5.15. What was the rationale for the upgrade to 5.15?

1 Like

Well...security...
Would you like your router to be of a botnet just because....it runs some ...well established kernel...and it's just that...
I would preffer to have a speed penalty and to run a patced kernel and I believe this is what most of the people doing openwrt updates.
The alternative is to keep to that well establieshed performing but abandoned release like 19.07 or 18.06...

1 Like

The list of changes of Linux kernel 5.15 is overwhelming, please do this research by yourself: https://cdn.kernel.org/pub/linux/kernel/v5.x/ChangeLog-5.15

For OpenWrt 23.05 with kernel 5.15: https://openwrt.org/releases/23.05/notes-23.05.0-rc3#highlights_in_openwrt_23050

Many new devices added

OpenWrt 23.05 supports over 1794 devices. Support for over 200 new devices was added in addition to the device support by OpenWrt 22.03.

  • The ipq807x target for the Qualcomm IPQ807x Wifi 6 SoCs was added

  • The mediatek/filogic subtarget for the Mediatek Filogic 830 and 630 SoCs was added

  • The sifiveu target for the HiFive RISC-V Unleashed and Unmatched boards

Highlights of device support

  • Switched ipq40xx target to DSA

  • VDSL support on AVM FRITZ!Box 7530

  • Support for devices with 2.5G PHYs

    • Acer Predator W6 (MT7986A), Mercusys MR90X v1 (MT7986BLA), Netgear WAX206 (MT7622), Netgear WAX220 (MT7986), ZyXEL NWA50AX Pro (MT7981) Asus (TUF Gaming) AX4200 (MT7986A), Netgear WAX218 (IPQ8074), Xiaomi AX9000 (IPQ8074), Dynalink DL-WRX36 (IPQ8074)
  • 2 Gbps WAN/LAN NAT Routing on ramips MT7621 devices (See OpenWrt forum)

  • Improved DSL statistics on ubus and in LuCI

The support for ipq807x Qualcomm AX and Mediatek Filogic 830 AX for me alone is worth this update. 2.5 Gbps Ethernet will be a benefit for users with more than 1 Gbps needs.

2 Likes

…and there have been quite a few RPi and x86 users for whose 2.5GBASE-T USB ethernet cards kernel v5.10 era rtl8150.ko was too old to support their card. As outlined by odrt, the same applies to a lot of other new- and/or actively developed targets and -devices (e.g. anything with 802.11ax).

1 Like

I look forward to trying a 2.5G WAN Ethernet with my 26Mbit/s British Telecom Superfast Broadband (a "fabulous" FTTC that still shares wires with POT). I think I will need a new 2.5G WAN Ethernet router right now to get ready when in about 20 years British Telecom will have provided FTTP (likely the lesser 1G version) to most homes in Sussex. Oh, and BTW, since Brits always have to be different from anyone else, what the rest of the world call FTTH they call it Fibre To The "Premises"

1 Like