Hello friends, I’m new to the forum. I love Linux, math, electronics, piano, and also EndeavourOS!
The internet problem we’re experiencing is due to a regression introduced in commit 9e30ecf23b1b affecting IPv4 traffic, which impacts kernel versions 6.16.2-arch1-1 and 6.16.3-arch1-1. Other distros, such as Nobara and CachyOS, already fixed it a couple of days ago with their latest kernels.
There are several fixes available, the first by Brett Sheffield, which was then superseded by Oscar Maes’s fix, dated August 26, which has already been approved.
If you want, you can download the Arch kernel source code and directly modify the /net/ipv4/route.c file. Then compile the kernel and you will have a 6.16.3-arch1-1 working perfectly. I already did it and it solved all my internet problems.
Wow, thanks for the update, @albersc2! I just came back to report that downgrading to LTS 6.12.43-1 fixed the problem – but looks like waiting for the next release will work, too. (I’m not savvy enough to patch my kernel yet. I hang my head in shame.)
That could possibly explain why I didn’t see any problems here—I have everything dual-stack (including gateway & ISP), and Linux seems to prefer IPv6 if it can.
This makes me wondering would there be a reason not to enable IPv6 , because I always enable it, ever since I read about the fact we are running out of IPv4 addresses some years ago. When somehow gets decided to phase out IPv4, and had missed the memo about it, I would be covered up front. But ofcourse that is just me.
Hello again. Unfortunately, Arch has not fixed the ipv4 bug in today’s patch. Very disappointing. I just ran the tests, and 6.16.4-arch1-1 still reproduces the same ipv4 internet issues as 6.16.3 and 6.16.2.
The fix has not yet landed upstream, although, as we have already seen, other distributions are patching it into their custom kernels.
It’s a shame Arch doesn’t do the same, because the problem is critical. In fact, internet failures could affect the update process itself and leave it half-finished, causing potential disasters. Use Timeshift before any upgrade and be careful.
And if you can make yourself a custom kernel with the fix applied, that’s best. Because from what we’ve seen, it doesn’t look like it’s going to be fixed in two or three days.
Yes, 6.16.1 works well; it’s the most practical solution. But imho, it shouldn’t be the case that from 6.16.2 the kernel has been updated two more times and it still isn’t fixed. It’s not a minor issue.
Also, be careful with LTS 6.12.43-1 (which Mike09 mentioned yesterday) because afaik it has the same ipv4 problem.
I hope they will fix it a bit faster. All my machines are affected (PC and Notebook). My Notebook is now my system to test because i blocked Kernelupdate on PC and waiting for fix
Basically this helped best, Kernel 6.16.1 runs perfectly, while today’s kernel releases, main and lts, still have the same problem for me. Now all is stable again, well, after I made a classic mistake of a … less experienced user: please do not forget to downgrade the kernel-headers package as well. Was a funky result without that
After updating last night before bed, and testing all three this morning, all seems well in the following kernels: linux-zen 6.16.4.zen1-1, linux-lqx 6.16.4.lqx1-1, linux 6.16.4.arch1-1
Sweet Jesus, Joseph and Mary.. I’ve been banging my head against a wall for a week now, trying to figure out what’s wrong. Downgraded dnscrypt-proxy, switched to systemd-resolved, tried different DNS servers and configs, adjusted MTU..
Glad to see that this bug has finally been identified! Perhaps this topic should indeed be pinned to save others from unnecessary headaches.
Bug still exists in 6.16.4… My Notebook is the testing machine for now. Sometimes the bug is completely random. Yesterday late at night i got 0 issues with packet loss tests. This morning issues with packet loss tests. If u have dual stack and IPv6 prefered then its logical u haven’t issues…
I have been using a RPi 5 with EnOS, KDE Plasma, and Firefox for my web browsing. Works like before. No problems.
The latest RPi kernel is linux-rpi-16k.6.12.43-1. I assume that the older kernel is the difference and doesn’t have the current kernel’s bug. mesa is the same version as EnOS x86_64.
Yes, I can confirm that the bug still exists in 6.16.4. In fact, nothing has changed at all. I just checked the hash of /net/ipv4/route.c (the file containing the bug) in the source code for 6.16.4 from Arch, and it is exactly the same as in 6.16.3. No fix has been applied despite being available for days (the regression was detected on the 22nd, Brett’s first secure fix is from the 23rd, and Oscar’s is from the 25-26th). I am already compiling a fixed 6.16.4.
This is interesting. Apparently, some people suffer from this problem much more intensely than others. In my case, with very standard equipment—an Asus Tuf Gaming motherboard with an i5 10400 and an Nvidia 4060—there are times when the internet connection becomes a nightmare. Even updates fail. In my case, it could be said that the computer becomes almost unusable for everyday tasks.
