With ACPI, and Now With UEFI 'Secure' Boot, Linux Users Are Made Less Confident Their System Can Boot or Wake Up (Just Like Bill Gates Demanded)
Reprinted with permission from Ryan Farmer.
“One thing I find myself wondering about is whether we shouldn’t try and make the “ACPI” extensions somehow Windows specific.
It seems unfortunate if we do this work and get our partners to do the work and the result is that Linux works great without having to do the work.
Maybe there is no way Io avoid this problem but it does bother me. Maybe we could define the APIs so that they work well with NT and not the others even if they are open.
Or maybe we could patent something related to this.
-Bill Gates, 1999
Linus Torvalds on the Lenovo UEFI Bugs. “When You Can’t Trust Kernel Updates, People Will Stop Updating the Kernel.”
In my post yesterday, I mentioned the nightmare that’s going on in Lenovo laptops from the last 3-4 years due to a Chinese developer at Loongson bumbling around in some really hacky ACPI (power management and device info) code in the Linux kernel.
Let’s face it, Lenovo is never not a nightmare.
They’re so filthy and corrupt, and they put such cheap shit (keyboards are always a problem after a few years in my observation, I also had to have them replace the entire mainboard in this laptop when it was less than a year old due to USB port malfunctions), that they’re always being sued for SOMETHING even if it isn’t Linux-related.
Like the Superfish malware they pre-installed to get ad money on some computers, or the time they abused a Windows anti-theft feature by having a BIOS that re-installed all the OEM “crapware” every time a user uninstalled it, or for a different adware incident where they ended up paying me some money, or all of their defective laptop monitors (which you can still claim money from if you have those systems.)
ACPI, though, is a Microsoft standard that started out in the 90s.
From the beginning, Bill Gates (who presided over Microsoft during its high water mark of making money through criminal activities) was E-Mailing people at Microsoft trying to figure out how they could make the ACPI standard so bad that it was either Windows-only or at least difficult to get working right in a competently-designed OS.
So it is not the fault of Linux that the PC has so many problems.
Most of the actual, worst, parts of the PC’s Legacy BIOS. The parts, like ACPI, which have brought so much swearing and cursing from users when their computers don’t work, was hashed out by Microsoft, and for the purpose of deliberately bricking non-Windows systems, or at least to cause annoying malfunctions.
Then, because re-writing things costs money, this crap was basically copied and pasted, verbatim, into the even bigger PC firmware trash fire, called UEFI.
As bad as ACPI in general is, it’s more of a problem with some manufacturers, mainly ones that use a particularly heinous supplier of UEFI firmware. (Lenovo tends to use Insyde.)
The ACPI code in the Linux kernel is some of the worst code because it deals with some of the worst firmware, PC firmware.
Microsoft designed it to sabotage other systems to maintain a Windows monopoly.
So Torvalds was right when he says if you do something to fix one thing, it often breaks something else. It’s pissing him off, it’s pissing me off. I’d imagine that it pisses off anyone who is not in Microsoft’s orbit, to be honest.
And it’s why I’ll either buy a System76 x86 laptop next time, with open source firmware, that isn’t some binary blob designed by Lenovo, a Chinese company that only barely tests to see if Windows boots and then calls it, or just start over with something like the 8 GB RAM Raspberry Pi 5 with Linux on some flash memory.
(Check your Windows system logs sometime if you have a typical Lenovo system. The firmware certainly isn’t harmless even on Windows. It fills the logs with errors, Windows hides them. It’s a fantastic arrangement they have.)
I’m smart enough to make an RPi 5 work, and they only set you back about $100-ish plus maybe some peripheral expenses. It’s not a hole in the bank account like some $1,000 lulzy laptop made out of Chinesium.
(I do wonder whether ZStd or lzo-rle would be the way to go for the ZRam device though. You’d definitely want the fastest algorithm for a compressed RAM device on a Pi 5 even if it is 4-5 times as fast as the Pi 4. So far, the only things going for an x86 PC are compatibility with proprietary Windows software in Wine, Steam for some people, and the fact that Windows coming with it made it bog standard for cheap mass produced garbage to throw Linux on, until now anyway.)
Linus Torvalds brought up a good point though. Even though these ACPI disasters are not the fault of Linux, he’s also not allowed to do interviews and bring this to public attention.
Linus Torvalds can’t tell you how pissed he is at this, because the Microsoft-controlled (buying influence and voting seats, along with partners) “Linux Foundation” is his paycheck.
Abuse has historically been hurled at Linus by criminals, monsters, and mobsters.
Sometimes abuse comes from incompetent fucktards who may not be criminals, but they’re at least doing bad work. Often from IBM/Red Hat (related to systemd and the idea to put dbus in the Linux kernel, among their other greatest hits), which kept bringing him bad code and even worse ideas.
I could see why Red Hat wanted dbus in the kernel. I really can. If it’s in the kernel, it makes the problem everyone’s problem. They were also hoping to lob it in there like a grenade without fixing any of their bugs, like random disconnects from the bus, and then run away without fixing anything. A drive-by dbus-ing.
These types of toxic people were the reason Linus was forced into “therapy” to keep his job (on a project he started), but although he can’t be as straightforward as he once was, he did at least let on that he’s badly annoyed that if things like the Lenovo incidents keep occurring nobody will trust kernel updates.
And he’s right.
Fedora once broke power management on my 2016 Yoga 900 ISK2 for over 3 months when Intel turned it off to investigate a security hole in their graphics card. Then they turned it back on without ever fixing that hole three months later. I had to version lock an older kernel with DNF and let dozens of much more serious CVEs pile up.
My mistake was that it was easier to keep Fedora then and do that than undertake the transition away from Fedora, which I eventually did years later because they’re in such a bad shape now that they’re ruining the distribution by dropping packages, making incompetent design decisions, and can’t even manage a release anymore without multiple delays.
Now that I am on Debian 12 and Linux 6.1 LTS will just keep getting bug fix backports indefinitely from upstream, this laptop is going to use Linux 6.1 until Debian 12 doesn’t work anymore or the hardware konks out.
I’m done pulling in major component updates that nobody can support because they don’t know what my exact computer will do when they get them. If I wanted broken shit every month, I could just grab the Windows 11 ISO and install that.
Thankfully, Debian has long term releases. They do what they do the day you install them, and it probably just gets better later on because it’s low risk stability patches and security patches.
Some people have chided me and say I should be using something even “edgier” than Fedora, like Manjaro. They claim that with little or no testing, I can just deploy what someone managed to run through a compiler yesterday that has thousands of major changes, deploy it, and it will never ever break down on me.
No thank you. You’re basically throwing darts at a board and hoping they at least land somewhere on the board when you run a distribution that doesn’t commit to conservative, semi-frozen releases, that are supported for years.
Debian 12 doesn’t have to be completely stuffy.
There are ways to target and backport individual releases of newer software to it.
Even Mozilla figured out the other day how to run an Apt repository, apparently, after a tradition that ran back to Netscape Navigator of putting it in a tarball and saying “Linux”. *slow claps*
There’s Flatpaks, there’s Debian Backports.
I’ve found out, at some great burden on myself, that it’s better to learn how to administer Debian than get something like Fedora installed really quickly by slapping “next” a bunch of times and then finding out that you “passed it to find out what’s in it”, as former House Speaker Nancy Pelosi put it with Obamacare.
In America, Congress passes gigantic bills in the middle of the night which contain 3,000 pages of new laws, which no voting member has even read, then you need hundreds of pages of “trailer bills” to fix the mistakes and the stuff that nobody even knew was there.
Then there are court cases to determine what Congress even meant because they used the wrong words in a draft and it made it into the law, and someone thinks they can bring the law down because of a draft error.
While all this is playing out, they’ve done 100 more laws just like that one.
That’s sort of what trying to deal with a distribution that’s constantly bringing new software in is like. Problems come and go. There’s no time to even figure out what’s gone wrong and where at.
The cost of slapping next a bunch of times and having a “system that functions” is that it will turn out to not function especially well for anyone.
They have to kind of guess what a core user will want to do, and they can guess wrong.
Every decision they make can either bloat or leave something out of a “live installer”, or make a setting that works for some people, and not so well for others.
Lately, Fedora has been a really big WTF for me. First they drop LibreOffice, which I need, because that’s what IBM demanded. Then they put in a systemd-oomd that kills browser tabs when you have a ton of free memory.
Now they can’t even figure out how to make a release without at least a two week delay, and they say maybe they’ll just give up fixing the blockers and cut an ISO.
Fedora has always been an uneven distribution, but now I don’t even trust it on a laptop I mainly use 10 applications on.
Again, it comes back to making changes to the kernel, and sending them out if they compile at all. I’ve had problems with Fedora I haven’t had with any longterm distribution, ever, including “upgrade your kernel and get a panic in the Ext4 driver”.
There is a very real risk that marking things stable and compiling them without anyone really checking what’s going on, people won’t trust kernel updates anymore.
Many of these problems are happening upstream because nobody checks the work of companies like Intel. They’ve been given a license to talk in something rather like “Fedspeak” and not document what their code really does, and there’s a very “laissez faire, laissez passer“, or let them do whatever they want, with companies and their driver code. So by far, the thing that scares me the most about upgrading the kernel is not core code.
It’s the x86 branch, especially ACPI, and the drivers, from some hardware makers more than others.
I’ve been around Linux long enough that unlike some of the critics of Techrights, I know how to use git on my computer and I once spent years with my own kernel series because I got so fed up with my distributions taking forever and packaging some rather awful releases. When you own the fork, the patchset, the compile time options and compiler, you get to manage things. My kernels almost always worked better for me than the ones the distributions made.
There are lots of reasons, from slimming them down, to turning off a bunch of really godawful shit that is useful to almost nobody that makes the entire thing flakier, to making sure you’re pulling the latest features in the hardware drivers into a kernel series you know isn’t bad.
Many of the people arguing that Microsoft has become a new company that likes Linux are either sockpuppets, or really are so stupid that they don’t realize that the kind of shit Microsoft does to the PC now is a million times worse than trying to deal with a PC 20 years ago.
Maybe, hopefully, some of these people ran Arch or something and yanked in a broken kernel and got a taste of what I’ve been going through for years.
*takes a sip of coffee* Told you so.
Debian 12 keeps updating Linux 6.1 LTS, so I really don’t know what later kernels are doing on this hardware, and I don’t care, as this is the most unbroken stretch I’ve had with a computer that was not doing something ridiculous.
The “pass it to find out what’s in it” approach is a bad way to run a country, and it’s a bad way to even run some laptops.
The PC situation is rapidly becoming untenable. It’s possible that it could fall apart completely on a technical level long before Microsoft has a chance to try to mitigate further erosion of the Windows operating system by making it impossible to turn “Secure Boot” off.
The only reason they ever gave you that switch was because the state of the PC industry in the Windows 8 era meant that some hardware needed Legacy Boot (BIOS compatibility mode) and there were people with downgrade rights to Windows 7.
BIOS mode is gone and Windows 7 is out of support. Time’s almost up. █