Netgear R7800 exploration (IPQ8065, QCA9984)

Is that a 32bit CPU architecture and instruction set?
What is the size of a pointer (and data) in this architecture?

P.S.
Are there R7800-specific build instructions for the OpenWrt source-code anywhere ?

Yes, it's bog standard ARMv7.

32 bit.

A subset of ipq806x is already part of the mainline kernel, for anything OpenWrt specific, check target/linux/ipq806x/.


The 802.11ax capable ipq807x SOCs are ARMv8/ Cortex A53 (64 bit) based, with everything that entails (and yes, they're typically also running a 64 bit ARM64/ AARCH OS image).

i wonder if the new platform will also use nss core... or finally switch to use standard hardware accelerator...

http://www.bitswrt.com/11AX.html (this page was considerably more verbose a few months back)
http://en.techinfodepot.shoutwiki.com/wiki/Qualcomm#ax_2
That sounds very similar to the situation with ipq806x, with two 1.7 GHz NPU (NSS) cores (instead of 800 MHz).

pure joke then... or we need to start working on nss core for real

IPQ8074 SoC: CPU Cores: 4x ARM Cortex A53. There is no mentioning of the nss cores.

Unfortunately it looks like the IPQ8074 will have nss cores: https://patchwork.kernel.org/patch/10110137/

Nothing wrong with nss cores, as long as they add decent upstream support for it...

1 Like

And that is the problem, I dont see them ever adding upstream support for those.
They dont upstream networking except for wireless drivers.

1 Like

Is the plan for our R7800 to get kernel 4.19 in the official builds? Hopefully that will fix the random reboots/freezes I see with my device.

The pr is up in github
Well tested and doesn't look to cause any trouble to anyone so.... We just need to give more attention to it.

1 Like

The ipq806x kernel 4.19 PR has been some time in the staging tree of @chunkeey and I have not seen any negative comments so far, so I hope that it may get pushed officially to the main repo rather soon. ipq806x is one of the laggards in that front.

I feel that it would be good to get wider testing via the master snapshots so that the possible negative items can be found. So hopefully 4.19 gets pushed soon.

1 Like

@hnyman do you have any test builds with 4.19? Ideally for my case I'd like it with the "old" firmware as "ct" just doesn't work for my clients. Not sure if 4.19 also fixes the "old" crashes your recent "old" builds experience or it's not related.

master-kernel4.19-test-r11597-2fedf023e4-20191127

I made a test build with the kernel 4.19 PR from @Ansuel
( https://github.com/openwrt/openwrt/pull/2472 )

2 Likes

Is it possible to make a build with the "old" firmware or does it have the same crash issues as master-r11526-f84d6d2f3a-20191116-ath10k-crashes ("old" mainline ath10k)?

You can switch to whatever firmware you want if you're using ath10k-ct driver, just upload the firmware to the router and reboot. But if you want the ath10k driver you need a new build and only use one of the official firmware's..

1 Like

The ath10k issues have not been fixed as far as I know, so only ath10k-ct works at this time.

1 Like

i don't have time at the moment to look into it, but as i said on the PR, with 4.19 the memory bandwidth is significantly worse in some scenarios.

it's visible using this very basic memory bandwidth test: https://github.com/raas/mbw, i've linked to a binary I built here: https://github.com/facboy/openwrt-r7800-freq-test.

Could be due to vulnerability mitigation?

Ah interesting didn't realize ct driver would work with "old" firmware. Might be a good test for my problems, see if the issues are in the driver or in the firmware.