Linux 5.6 Ships With Broken Intel WiFi Driver After Network Security Fixes Go Awry

[phoronix]

Phoronix: Linux 5.6 Ships With Broken Intel WiFi Driver After Network Security Fixes Go Awry

For those that are normally spinning their own kernels and punctually upgrading to new releases, you will want to hold off on the new Linux 5.6 kernel for the moment if you use the Intel "IWLWIFI" WiFi driver...

http://www.phoronix.com/scan.php?page=news_item&px=Linux-5.6-Broken-Intel-IWLWIFI

[birdie]

In the past I was brave enough to migrate to point zero kernel releases as soon they got released. For the past three years however, I've only upgraded to point four or five releases, so that I don't have to deal with the most critical regressions as kernel QA/QC is still horrible. Please don't tell me to use "stable" distros - it's idiocy. It just means other people have tested the kernel for a considerable amount of time to resolve all the corner cases and, more often than not, it's people like me who find regressions, identify the bad commit(s) and then test the fix which then finds itself in "stable" kernels. It's not like Ubuntu or RHEL have extra special people who test the kernel on all possible hardware - no, they do not do that. They are pure software companies.

[nranger]

QA for Intel iwlwifi seems to have been pretty poor lately. I've been bit by https://bugzilla.kernel.org/show_bug.cgi?id=206479 and https://bugzilla.kernel.org/show_bug.cgi?id=206329 .

I rarely compile my own kernel these days, but after a few failed patches, I ended up just copying the whole iwlwifi folder out of the 5.4 tree to get a working 5.5 build.

[Volta]

In the past I was brave enough to migrate to point zero kernel releases as soon they got released. For the past three years however, I've only upgraded to point four or five releases, so that I don't have to deal with the most critical regressions as kernel QA/QC is still horrible. Please don't tell me to use "stable" distros - it's idiocy. It just means other people have tested the kernel for a considerable amount of time to resolve all the corner cases and, more often than not, it's people like me who find regressions, identify the bad commit(s) and then test the fix which then finds itself in "stable" kernels. It's not like Ubuntu or RHEL have extra special people who test the kernel on all possible hardware - no, they do not do that. They are pure software companies.

Terrible. I still worry to make my Windows 10 upgrades, because this damn thing never gets stable. Ps. last times kernel was the least of your concerns:

https://www.phoronix.com/forums/foru...04#post1168204

It seems you're just a troll. It's Intel driver maintained by Intel developers, so blame Intel.

[ddriver]

Intel is too busy code-blocking amd drivers to handle its own stuff properly....

[pete910]

Terrible. I still worry to make my Windows 10 upgrades, because this damn thing never gets stable. Ps. last times kernel was the least of your concerns:

https://www.phoronix.com/forums/foru...04#post1168204

It seems you're just a fucking troll. It's Intel driver maintained by Intel developers, so blame Intel hypocrite.

Any need to swear like that on a public forum! ?

[Quetzalcoatlus]

Yeah intel's been doing quite a bit of the ol' "breaking their own drivers" lately. Whew, dodged a bullet on that one with my Broadcom(R) Wifi and driver that i also had to port into my own kernel fork because it's so not GPL compatible even the source code comes with a binary blob.

I think maybe the kernel should go back to something like the old even/odd minor number system for explicitly stable and unstable branches, these type of shenanigans used to be confined to the odd number versions and anyone not doing active developent could stay on the safe even numbered playground without so many sharp edges. I feel like it's only a matter of time until this ends up getting accidentally downstreamed into some distro long after being fixed in master like the i915 driver breakage was

[TemplarGR]

Intel is too busy code-blocking amd drivers to handle its own stuff properly....

This. Intel and Nvidia MO. Why worry making your own drivers good when you can block the competition's hardware from working? And i am not talking only about Linux.

[birdie]

Any need to swear like that on a public forum! ?

It's best to put him in your ignore list. He's a rabid open source fanatic who does exactly nothing for Open Source while at the same time exaggerating and outright lying about W10 issues. I've been running W10 for over 5 years now (yeah, I do play games occasionally and Windows remains the best PC OS for them). I haven't had a single issue with it.

For instance the critical issue I personally discovered in the kernel has nothing to do with any HW company but he will insist anyone but kernel developers are to be blamed for this crap.

[geearf]

In the past I was brave enough to migrate to point zero kernel releases as soon they got released. For the past three years however, I've only upgraded to point four or five releases, so that I don't have to deal with the most critical regressions as kernel QA/QC is still horrible. Please don't tell me to use "stable" distros - it's idiocy. It just means other people have tested the kernel for a considerable amount of time to resolve all the corner cases and, more often than not, it's people like me who find regressions, identify the bad commit(s) and then test the fix which then finds itself in "stable" kernels. It's not like Ubuntu or RHEL have extra special people who test the kernel on all possible hardware - no, they do not do that. They are pure software companies.

I'm not as drastic as you but I agree, these days I start with at best .1 instead of .0 or even rcs in the old days.
Well, there's also the fact that there is no particular new feature that I need so there is no point risking it.