Is the NoisyTCP (CVE-2026-99288) an issue for RouterOS version 7 with kernel version 5.6.3? The Issue introduced in 4.4 with commit 079096f103faca2dd87342cca6f23d4b34da8871.
Yet another thing that needs patching in the ancient, patchwork-like mess of backports found in the kernel used by v7.
It seems v6 escaped unscathed—precisely because it’s so very old.
WTF... I can’t even test a version of RouterOS for more than two weeks without a CVE popping up...
And then, instead of just bumping the version by x.xx.+1,
they take the next "stable" release and force-label it as "long-term", (like 7.23/7.24)
so, instead of just patching the existing long-term version,
I end up having to re-test everything from scratch...
What a PITA, what a STTY way to manage versions and fixes...
For now, I'd make sure tcp-syncookies=no (disabled by default) and that no TCP services on the router are exposed to untrusted networks. Also restrict any necessary access through the input chain.
Disabling SYN cookies (i.e. tcp-syncookies=no) blocks the crash path but not necessarily the entire vulnerability. We still need MikroTik to confirm whether ROS v7 is affected or not.
- Openwall researcher's advisory: https://www.openwall.com/lists/oss-security/2026/11/15/2
- New NoisyTCP vulnerability lets attackers crash Linux servers
Clarification: the CVE-2026-99288 exploit will cause a kernel panic and take down the host (or router) but won't allow any threat actors to alter data or gain arbitrary code execution.
They will act like dead possum and after few days later they will come up with new version with "improve stability".
I can not agree more and IF it was just MT but no every vendor with Linux need to patch constently. It feels like a read only cve emails from vendor these days.
This is such a pain in the ass, 5000 devices updated, bam, done, spend the morning netinstalling and replacing devices that died and bam, have an update at noon, here we go again, and again, and again. Can I have some uptime back pleaseeeee? It's just updates for 3 months and unlike v6, v7 craps out a lot more.
Do you people even understand those CVEs?
Bunch of drama queens.
It is the usual thing. Some wildly unlikely thing that would only be a DoS and everyone panics.
However, it could be a good time to make "kernel" a separate package in RouterOS that can be upgraded independently. So not all those generic Linux kernel "vulnerabilities" do require a full upgrade.
It would be good if Mikrotik could bring back a new implementation of their old bugfix class again.
(on top of long term, stable...)
Mostly only applied to long term chain.
This would be applied to the current popular long-term release, and would only have bug fixes in it.
e.g. CVE's and maybe?? things that used to work earlier but this long-term version in particular has broken.
It might need another level of version or something. e.g. 7.24.6bf1 so they can just bugfix 7.24.6 and keep working on their soon to be released 7.24.7 with its new features/issues. Rather than having to rush it out.
Well, as described above the big problem is that when such a minor issue appears, MikroTik releases a new "stable" version and when the "long-term" version already was more than one major version behind they take the opportunity to move that to a new major. E.g. long-term is 7.23.6 and new stable is 7.25.1, then a new long-term 7.24.6 is released instead of 7.23.7. WRONG!!! it has been written many times, but the folks at MikroTik either do not get it, or they do not care.
Yeah, an LTS that moves every time a new stable drops isn’t really long-term at all. As the Mandalorian would say, this is not the way!
My numbers are a lot less but the problem remains.
Just finished upgrading all devices with my client.
And there we possibly go again ...
Sign of the times and one of the consequences of AI gaining traction.
AI can find problems a lot faster then humans can patch them.
I hope this is like a gartner hype cruve and means it will be worse now but sooner or later the easy to find things will go away and we end up with better and more secure software and all new software I sure hope run trough the same AI models to find issues before it ends up in production. I can be wrong but I live on the hope at least
But I repeat what people say and strap on it is going to be a rollercoaster of a ride for a while.
But having the CMR live would be nice now to better automate the upgrade flow but we are getting closer at least.