[Alpha] 386.2

[gattaca]

^^^ Updated 1 x RT-AC86U WAPs to 386.2_alpha1-gc52b410e89 just uploaded about 8 hours ago. Notify on issues. TY!

 

[Clark Griswald]

+1 From Casa Griswald's

Update to latest alpha was uneventful.
Not sure what testing people are using, but the latest alpha has survived two teenagers not noticing any changes in the LAN/WAN.
Thank You RMerlin
Stay Safe Everyone

 

[maxbraketorque]

New 386.2 alpha builds uploaded 1h ago.

- NEW: Added jittertimer-rngd daemon to HND routers. This will ensure sufficient entropy is generated early on at boot time, reducing boot stalls caused by insufficient entropy for the kernel's random number generator, and also generally improves security related to crypto operations by the router.

Interesting addition. I've not heard the term entropy applied to a kernel, let alone that insufficient entropy can cause a boot stall or affect crypto operations. However, it sounds like a desirable addition to the firmware though and perhaps something that will migrate into the stock FW. I'll be updating my home network as soon as daily backup operations are complete.

 

[John Davis]

Interesting addition. I've not heard the term entropy applied to a kernel, let alone that insufficient entropy can cause a boot stall

it's a know issue on mainstream linux https://forums.gentoo.org/viewtopic-t-1083804-start-0.html

https://ieeexplore.ieee.org/ielx7/6287639/8948470/09050782.pdf?tp=&arnumber=9050782&isnumber=8948470&ref=aHR0cHM6Ly93d3cuZ29vZ2xlLmNvbS8=

 

[RMerlin]

And it was also causing my GT-AX11000 to get stuck at boot time after going from stock firmware to my test build. My firmware has an additional SSH key over stock firmware, so on first boot it was attempting to create that key. The entropy pool was flat out 0 that early in the boot process, dropbear's key generation was pausing the entire boot process as it was waiting to obtain sufficient random numbers, and there wasn't any activity yet to generate sufficient interrupts to fill up the pool, meaning the router was stuck forever at creating this key. That particular issue was resolved by moving the SSH service start at the near end of the boot process. However a few other processes still took a bit longer to initialize (like AiMesh's config management process which relies on generating a key/cert pair), which led to the addition of this entropy generator. Unfortunately it doesn't work on older kernels/uclibc, so it only applies to the RT-AC86U and newer.

This one was picked because it uses the same method Linus added to recent kernel versions (timer jitter) to resolve that issue at the kernel level. And it seems much less bloated than a solution like haveged. This remains experimental at this stage, I haven't decided yet if I will keep it in, or if it will leave it running permanently or only at boot time. That's what alpha stage is for me, a time to experiment with things.

 

[Deleted member 22229]

Just flashed 386.2 A1 AX58U. Working fine.

 

[maxbraketorque]

UPDATED with some additional detail.

So this showed up in the log during the boot sequence of my primary AC86U when starting from a power off state. Its not something that has been present in any recent prior logs of the boot sequence of my primary AC86U, and my AC86U that's serving as an access point does not have this when starting from a power off state.

Router is performing normally as AFAIK.

This event did not show up in the log when performing a soft reboot.

May  4 22:05:14 Mastiff: init
May  4 22:05:16 kernel: pgd = ffffffc015446000
May  4 22:05:16 kernel: [f727e5d0] *pgd=0000000014143003, *pud=0000000014143003, *pmd=00000000155b6003, *pte=0000000000000000
May  4 22:05:16 kernel: CPU: 0 PID: 1959 Comm: amas_lib Tainted: P                4.1.27 #2
May  4 22:05:16 kernel: Hardware name: Broadcom-v8A (DT)
May  4 22:05:16 kernel: task: ffffffc01e02ab40 ti: ffffffc0157a0000 task.ti: ffffffc0157a0000
May  4 22:05:16 kernel: PC is at 0xf6a16a58
May  4 22:05:16 kernel: LR is at 0xf6a14728
May  4 22:05:16 kernel: pc : [<00000000f6a16a58>] lr : [<00000000f6a14728>] pstate: 60080010
May  4 22:05:16 kernel: sp : 00000000f727e5b8
May  4 22:05:16 kernel: x12: 0000000000000000
May  4 22:05:16 kernel: x11: 00000000f7280b94 x10: 0000000000000002
May  4 22:05:16 kernel: x9 : 00000000f6c0a979 x8 : 00000000f6b0f000
May  4 22:05:16 kernel: x7 : 00000000f728bce0 x6 : 00000000f6b0fcc0
May  4 22:05:16 kernel: x5 : 00000000f6b0fcc0 x4 : 00000000f727e680
May  4 22:05:16 kernel: x3 : 00000000000f7db0 x2 : 00000000f7280bac
May  4 22:05:16 kernel: x1 : 00000000f6c0a979 x0 : 00000000fbad8004

 

[john9527]

And it seems much less bloated than a solution like haveged.

Just an FYI....I added haveged to my fork back in 2017 (both arm and mips). Ran into entropy shortage problems when I did the original dnscrypt.

 

[SomeWhereOverTheRainBow]

And it was also causing my GT-AX11000 to get stuck at boot time after going from stock firmware to my test build. My firmware has an additional SSH key over stock firmware, so on first boot it was attempting to create that key. The entropy pool was flat out 0 that early in the boot process, dropbear's key generation was pausing the entire boot process as it was waiting to obtain sufficient random numbers, and there wasn't any activity yet to generate sufficient interrupts to fill up the pool, meaning the router was stuck forever at creating this key. That particular issue was resolved by moving the SSH service start at the near end of the boot process. However a few other processes still took a bit longer to initialize (like AiMesh's config management process which relies on generating a key/cert pair), which led to the addition of this entropy generator. Unfortunately it doesn't work on older kernels/uclibc, so it only applies to the RT-AC86U and newer.

This one was picked because it uses the same method Linus added to recent kernel versions (timer jitter) to resolve that issue at the kernel level. And it seems much less bloated than a solution like haveged. This remains experimental at this stage, I haven't decided yet if I will keep it in, or if it will leave it running permanently or only at boot time. That's what alpha stage is for me, a time to experiment with things.

Alpha on GT-AX11000 seems to be working as expected. Aside from listing 5 lan ports(figured such) and the extra 5G band, I would not be able to tell the difference in the firmware GUI-Layout than the RTAX88U.Has any one been able to test the 2G port functionality. I was able to get WAN connection with it though I haven't been able to test for the max uplink due to equipment limitations.

 

[Slawek P]

Morning, I observed several things with 386.2 alpha. Any comments/suggestions?

1/ AX88U default setting for 2.4GHz OFDMA has changed to Disabled (before it was DL OFDMA only as for 5GHz)

My AX clients are using 5Ghz anyway, but it causes one difference for AiMesh node actually
- if I activate DL OFDMA then it creates additional 2.4GHz connection to main router - see <unknown> below (though it chooses 5GHz for uplink)
- disabling it (as per factory reset defaults) leaves only 5Ghz connection between nodes

2/ Some clients are struggling to get their IPv6 address - I noticed there is now a tendency that occasionally IPv6 appears quite late after reboots in many cases and some clients do not get it at all - including Alexa and AiMesh node. Checked ifconfig on AiMesh node and tried traceroute -6, ping -6 with no luck. I did factory reset twice on both router and AiMesh node to be sure.

 

[RMerlin]

 

[maxbraketorque]

Took me a minute to find the relevant tidbit of info on that page. Rather fascinating to see an AX68U. Wondering how it differs from an AX86U.

 

[isr25]

Took me a minute to find the relevant tidbit of info on that page. Rather fascinating to see an AX68U. Wondering how it differs from an AX86U.

2 core vs 4 core CPU, 512MB vs 1GB RAM, AX86U has an extra internal antenna, AX86U has a 2.5Gbps port, 2 x 5Gbps USB Type-A vs 5Gbps USB Type-A + 480Mbps USB Type-A, Theoretical WiFi throughput on AX86U is 2x higher

 

[SomeWhereOverTheRainBow]

Took me a minute to find the relevant tidbit of info on that page. Rather fascinating to see an AX68U. Wondering how it differs from an AX86U.

Fascinating, as no one has pointed out yet... It seems this is an upgraded AC68U (the AX version). Just as the AX86U (upgraded AC86U?)...

 

[heysoundude]

Fascinating, as no one has pointed out yet... It seems this is an upgraded AC68U (the AX version). Just as the AX86U (upgraded AC86U?)...

Eagle-eyed forum members have been waiting patiently for a sign... lol
I'd wager moderately that it spells the end of the venerable AC68 in these parts sooner than later, unfortunately. Frankly, I'm surprised it lasted this long if I'm anywhere near right.

 

[L&LD]

The price for the new RT-AX68U is the price the RT-AX86U was on sale for many times previously.

There is no comparison between the two (hardware), I would not recommend it at the current price (the RT-AC86U is still the best bang for the buck entry level 'router').

If it also goes on sale for less than $200.00, then it would be worth considering. Otherwise, it is a curiosity right now. I have no doubt the Wi-Fi is superior, but the rest of the hardware seems lacking to me.

 

[RMerlin]

If it also goes on sale for less than $200.00, then it would be worth considering.

The RT-AX68U MSRP is 199$ USD, so you don't need for it to be on sale. Just to not be sold by a scalper...

 

[L&LD]

Different currency. The RT-AX86U was available for $200 USD on sale too. The RT-AX68U does not justify that price with half the resources it has vs the much better model up.

 

[Slawek P]

Different currency. The RT-AX86U was available for $200 USD on sale too. The RT-AX68U does not justify that price with half the resources it has vs the much better model up.

I was lucky to get both RT-AX88Us on Amazon sale events for £200. Different currency.

 

[heysoundude]

There is no comparison between the two (hardware), I would not recommend it at the current price (the RT-AC86U is still the best bang for the buck entry level 'router').

you may have misplaced the quotes there. I would have placed them thus: "entry level"

As far as pricing on the ax68, perhaps if Asus was to put together an AiMesh-AX bundle for $300 for an ax86 and an ax68, or $400 for an ax86 and two ax68s...else, I can wait for 6e and prices to drop; my ac86 has plenty of life left in it, and my devices haven't exceeded its capabilities