Debian and firmware
blog.einval.com
blog.einval.com
I know I'm in the minority, and expect my opinion to be weighted accordingly, but principle is a key factor I use when choosing anything - including software. My loyalty to Debian would be significantly impacted by such a change.
Is Debian missing projected growth KPI's or shedding users? Is this a key hurdle to onboarding new contributors? I'm aware there may be other considerations that those actively participating in the community or leadership may be aware of that I am not, and I might be ignorant to the full value of the proposition. Debian is amazing and I have the upmost respect for everyone involved over the decades, but a move in a direction that weakens its position on non-free components would most likely drive me to using Guix exclusively.
And many of the other top Linux distributions are Debian derivatives which both participate in the community and provide the alternate experience.
Upholding its principles is the way Debian distinguishes itself. If you don't like it, use Mint or something.
Have been using Debian as my daily driver for 5+ years on both desktops and laptops. No issues at all though I did spend some time on gnome shell to get it closer to what I prefer (multiple desktops, alt-tab behaviour, etc)
I've been burned a couple times too many on Ubuntu's regressions. They'll ship a newer version of a package that breaks some use case which isn't super common but not super obscure either, and then I have to figure out what broke. I've stopped trusting Ubuntu's packages so I don't use Mint either even though I like the distro.
Debian continues to impress me with its stability. I run the testing branch at home and breakage is still extremely rare. Yes, there's things that Debian makes harder to set up, and running Debian means you basically resign yourself to running a userspace that's 1+ year old, but the overall experience just works better in my experience.
Maybe I should give Fedora a serious try one day. I haven't properly used any Red Hat based desktop since Mandrake but I did like CentOS a lot for servers.
I don't care about FOSS, so the principles are not a motivation for me.
The installation process, that I have to run when a local disk eventually fails is insignificant compared to constantly fixing shit. (And yes, it's not a nice process - even worse because every time I run it, it changed.)
If we look at popcon[1], it currently says 11.84% of users who opted into popcon regularly use the "firmware-misc-nonfree" package, along with 9.57% for the realtek package, 8.76% for the modern Intel wifi firmware package (iwlwifi), and 7.07% for firmware-amd-graphics.
I can't claim that's a statistically significant percentage of Debian users, I don't have that data, but of the ones who volunteered their information, even assuming perfect overlap, a not insignificant number of users are happily opting into the "not officially Debian, we swear" portion of the world.
[1] - https://qa.debian.org/popcon.php?package=firmware-nonfree
I've made a conscious choice to install the non free packages. I'm quite happy I had that choice, but I'm even happier that they weren't forced on me.
I certainly hope nobody will point to me and say "look here, non-free should obviously be the default!"
Personally, while I would prefer to be able to change and replace the software running on the tiny computers that make up a modern system, focusing on the distinction between "the vendor shipped it on a flash chip so we don't have to think about it" and "we need to load it at runtime" seemed a bit like spending too much effort on too little reward, to me - if we convince the manufacturer to up the cost by a few cents to hide the firmware blob from us, is that really a victory for Free Software, simply because we don't have to think about it? (Particularly if it sometimes results in never having ready access to a mechanism to replace the firmware blob, should someone sufficiently motivated either convince them to Free it or develop a replacement?)
It's great whenever someone develops replacements bits or, even more rarely, convinces a company to release their firmware bits permissively, but without a list of examples of Debian's stance changing the policies of other organizations, I think it's not that effective a tool for changing behavior in this case, and instead primarily inconveniences people who want to use Debian.
Whether they end up changing the default behavior or not, people in this thread replying "wait, there are firmware-bundled images?" makes me think Debian should either stop making them or stop making them so unintuitive to discover.
The Web has plenty of tutorials explaining the installation process and the difference between the official ISO and the non-free ISO, even though I would love a better Debian Wiki.
Maybe it's a bit of gatekeeping, but popularity by itself shouldn't be the goal for Debian. You don't know how to install Debian, you don't want to search for a tutorial first, you won't let someone more knowledgeable to help you? Well, I believe this is not the best place for you, you'll face worse things on the road if you keep using Linux.
(But, of course, this isn't an uncompromisable principle, as evidenced by the existence of the non-free repository. Packaging basic firmware on the same image as the rest of the system isn't the unthinkable action that people are talking about, but silently running them would be.)
Having said that, my reason for preferring firmware freedom mainly isn't principle; it's that non-free firmware potentially deprives me of control of my own equipment. By turning it's nose up at non-free firmware, Debian increases incentives for people to buy equipment that depends less on these blobs. That in turn incentivises manufacturers to offer equipment that doesn't depend on blobbiness.
I realise that these incentives are strictly at the margin; but Debian is the only leading distro that takes a hard line on these matters. If Debian were to retire from the battlefield, nobody would be left standing except the closed-firmware hardware makers. That would make me sad.
Actually, it only incentives people to buy and manufacturers to offer equipment which has the exact same blobs on a NAND chip on the equipment itself. It's more convenient, sure, since there's no need to find a copy of (the correct version of) the blob and load it together with the driver, but it still depends on these blobs.
I found in the end, proceeding to install and then post-install patching was easier but tedious.
If you have access to USB slots, this is a non-problem. If it were possible to find and make and maintain "CD with a bit of the blobset glued in" I'd be happier frankly.
As mentioned in my original comment, those images already exist, and are linked to on the Debian installation page.
https://cdimage.debian.org/cdimage/unofficial/non-free/image...
For the longest time (since fixed), trying to use Wi-Fi in debian-installer simply didn't work, regardless whether I used the unofficial image or dumped the firmware debs on disk. Can't complain about that anymore since it's fixed.
Now that Wi-Fi works when available at install-time, let's try to install a bog standard desktop system. Did you run the text installer or the graphical installer? If you used the text installer, you might be surprised to get an unescapable gray screen on an otherwise fully working booted system! Your video out doesn't work properly because kernel modesetting doesn't work without firmware or some mumbo jumbo like that. Your options are:
- Run the graphical installer in the first place so that the GPU firmware is used/installed by the installer
- Run a shell in the text installer and do "apt-install firmware-blah", a command which is only documented in a dusty unstyled html in a file cabinet hidden by a beware of leopard sign copyrighted 2010
- On boot up, use some relatively arcane GRUB magic (I'm just a user, so I have to look it up every time, okay?) to boot into some kind of safe non-graphical mode and install the GPU firmware
(Bonus: the unofficial firmware-included disc image symlinks the firmware debs, so the image only works when you DD it. The most popular Windows disc writing software Rufus recommends writing disc images in ISO mode (file-by-file) so that you can use excess free space. If you're wondering why the firmware directory has a bunch of 0 byte files, that's why, and you need to switch to DD mode.)
Enabling it should display a big scary warning about handing over partial control of your machine to greedy faceless megacorps who will throw you under the bus in a heartbeat if that makes them money. Maybe add an illustration of a robot terminator, to really drive the point home.
Yes, I agree that a default Debian configuration should be clean. But I also know the reality of needing at least Atheros (Wifi), Intel (Ethernet), and NVIDIA firmware blobs to get my tower to boot correctly. Not including any firmware repository on the CD or on a fully installed system is just a recipe for users to get frustrated while they are trying to follow a half-outdated tutorial on one of those spammy tech support blogs.
So in my opinion, the most common blobs should be cached by default, but not activated by default. If the machine boots without them, they can be deleted upon first reboot. If not, we can let the user decide if they are willing to compromise their principles in order to boot the system ;)
Back in my early Linux days I had an issue where the default Ubuntu/Debian installer would come up with a garbled display and I'd have to use `vga=...` or `nomodeset` (or similar, can't remember) to get a proper console going. Once installed I would install a package and it would "just work".
A way to directly boot into an environment with non-free firmware available could help and the user doesn't lose any more freedom if they're already determined to use non-free firmware post-install anyway.
Then select "expert mode", enable "non-free" repo, maybe "backports", go through base install only, reboot, re-enable USB-tethering a last time on the phone, install needed network card firmware, reboot, disconnect phone, run "tasksel", select whatever you want to install to finish install.
Having done this "detour" for many years now, it's "automatic" for me, and takes barely 3 minutes longer than the "normal" install from the IsoWithNonFree. I usually want "backports" enabled, so going into "expert mode" is required anyway.
Like I said, each thing is just "fix this one thing by compromising principles." If you do that with all the UX problems you literally end up with ChromeOS. Different distros are on different places along the spectrum between Debian and ChromeOS. Ubuntu is further along for example, so you get applications in containers and other things like that.
This is an immediate and direct contradiction.
The choice is literally between the non free firmware or your wifi and ethernet cards will not work.
That's why I believe the most common firmware .deb packages should be cached during installation so that they are on-disk in case they are needed.
This has been the case since the SI chipsets in 2012. I'm not sure if the fault lies fully with the driver, maybe the hardware doesn't allow querying the state of the engine firmware before enabling the engine.
At /usr/share/X11/xorg.conf.d/10-radeon.conf:
Section "Device"
identifier "Radeon"
Driver "modesetting"
EndSection
TTY works well before and after KMS.
There's already a "non-free" section under official "dists" ([0] for stable). It's just not added to the ISOs. Netinstall directly asks that whether you want non-free enabled during installation too.
The only thing is whether building CDs with it or not. Actually there are Zip files containing all firmware for a release, which can be added to a USB drive and added during install, but it's not well documented.
Maybe, this Zip files can be made more prominent, and using them can be better documented. Yes, it's not including firmware in the CD, but it's a good compromise.
What about a nice tool (akin to GRML2USB) which writes the ISO to USB and leaves a mountable FAT32 partition where user can drop in the firmware.zip file, so it becomes self contained if user wants?
https://cdimage.debian.org/cdimage/unofficial/non-free/image...
The blog post is putting the idea of adding the firmware blobs into the official ISO files, and as a Debian user, I'm not comfortable with that.
So, I'm offering another trade-off: "Make inclusion of firmware blobs to installation media on the fly easier, keep firmware off the official ISOs". It's a bit of a DIY solution, but it doesn't need flexing policies, DFSG and the values Debian is standing for.
It's a thin but important line to cross for me.
Wouldn't this imply per-seat IP licenses for things like embedded H.264 software fallback decoding in GPU drivers? I.e. the reason that Ubuntu (that doesn't really care about "proprietary" software by itself) still doesn't ship media codecs on the CD, nor driver blobs that could include them?
I think you got it backwards. Users will seek out SOFTWARE alternatives. They're easier to change than the hardware. They're likelier to go the Ubuntu or Linux Mint route than to die on a Debian hill.
Life is too short to depend on people that don't want to support your use-case.
It's so funny that this doesn't affect you at all but merely making the platform more accessible "significantly impacts" your loyalty.
Doesn't that mean your loyalty was conditional anyway? You are ready to jump ship if anything you don't like changes, even if it doesn't have anything to do with your system. Some loyalty.
Switched to Windows server (and was blown away by how polished that was, but that's another story.)
I get the advantages of free software, I really do, but this sort of neglect of usefulness in the name of ideological purity does not win converts.
On the other hand who cares? Free Software does not need to grow, it fulfils a purpose for those who want to be ideological pure, and I accept that is a valuable thing. Chasing converts at the cost of purity may just end up being no-man's-land.
Perhaps what the author is looking for I'd a debian derivitive, not debian itself?
Sure, there's no VC breathing down the neck of Debian, but arguably the purpose of free software is to provide software freedom for users. If people aren't able to use that free software, it's a bit pointless. Further, a (small) fraction of users grow into developers creating more free software.
To the extent non-free firmware allows more people to use free software, I don't think it's clear at all that a puritan approach to firmware is a net win for software freedom.
If the goal is to create converts then a puritanical approach is a poor approach, IMO.
Personally I've never seen the FSF approach as being "conversional" - rather it acts as a purity touchstone.
Debian needs to articulate its specific goals to the author, there are many paths they can take, but unless the goals are clearly articulated its hard to see what the right path is.
Note that some goals which we take for granted (everything should be free, get folk to switch) are not necessarily compatible. I think that is the root of this discussion.
Perhaps not purest, and perhaps not in stone, but in the closest thing to a digital equivalent it's at least written that Debian has to be pure.
The very first bullet of Debian's Social contract is: "Debian will remain 100% free". You could argue that when that changes, it stops being Debian.
Anyway, for FW one can certainly argue it's not part of the OS, as it's executing in some auxiliary core on the device rather than being part of the OS itself executing on the main CPU's. And further, if that same non-free FW would be stored in ROM or on flash, Debian has no problems with it as long as open source drivers are present in the kernel. So the only problem here is devices that want the driver to provide the FW blob to them upon initialization instead of the FW blob being persistently stored on the device. Not really any meaningful difference from a software freedom perspective either way.
Have you told this story anywhere else already? I'd love to read it. In a past life I worked very close to Windows Server, but I mostly selfhost Linux at home now.
What is it mostly identity related? I miss AD sometimes...or maybe storage? The SMB stack is pretty good too.
after booting windows looked at the disk layout and asked me what I planned to do. It suggested some quick options, one of which was exactly what I planned to do (raid the spindle pair, and raid the SSD pair) and one even better (create a volume putting the SSD and spindle together, and then raid that.)
Yes thanks I said, and it did everything. By contrast setting up raid on the previous Linux box, albeit 5 years earlier) took days of searching for docs, specific magic incantations on the command line, and lots of rebooting.
So yes, I get the freedom thing, but sometimes a computer is just a tool, and I want it to just work.
Even the incantation to do it manually isn't that magic. I'm guessing you could read the manual for the options to mdadm --create in minuted rather than days. If you want to be really fancy zpool create mirror foo mirror bar.
The point is to turn the tables in software. Free Software needs to grow and expand until it makes sense to say:
> who cares? Proprietary software does not need to grow, it fulfills a purpose for those who want to be ideological pure as propertarian/capitalist
and the question is whether to completely ignore anything involving non-free software or not.
"What is the likelihood of closed-firmware-blob hardware manufacturers changing their behavior because of Debian?"
vs
"What is the likelihood of fewer people using Debian because their installer is more difficult?"
IMHO, the odds aren't good that ideological purity is going to have any net result on hardware manufacturers. So you're really just shooting yourself in the foot so you can feel painfully just.
That said, there are absolutely fights Debian can and should focus on, even if they're painful: ones where there's a chance of victory, and the rewards of victory balance with the risks of the attempt.
How about all three? We have plenty of very good "open with proprietary firmware" distros, many very similar to debian (Ubuntu, PopOS, Mint, Manjaro...), but not that many good "fully open" ones. That is the whole point of distros. If we dilute Debian's fully open philosophy, we don't really gain anything in middle category, but lose one of the few remaining good fully open distros.
To fix your example: "Oh, you say this Debian thing doesn't support my network card? I see why everyone recommended that Pop thing instead, I'll use that then."
Debian is not a friendly distro for users who don't know what they're getting themselves into and that's fine. That's why nobody should (and nobody does) recommend it to Windows users looking for their first Linux distro.
The problem with accepting "Debian is not a friendly distro and that's fine" is that unlesss the snags are addressed, the number of reasons for people to run off to Pop and Ubuntu will only increase over time. Having a friendlier on-ramp to something that is actually Debian is good, but I suspect you could satisfy the "keep Debian pure" requirement with a fairly thin branding layer.
That sounds like Debian's non-free ISO.
(You can tell from the blurbs on the readme and download page).
We currently have a strange situation where the devices that are easy to use are the most open and the most locked down ones, and the middle-ground is punished by policy. If that policy were successful at promoting open firmware I think that would be acceptable, but I don't think I even need to argue that is not the case.
Not disagreeing that policies should be made with success in mind. But by the same reasoning "vegan" should include the top 5 most popular meat dishes as exceptions. This would easily make it more popular, more accessible and probably much more successful at reducing meat usage over all. The issue is that the definition of "success" might be entirely unacceptable. Just like "entirely free but...." might be unacceptable.
The first time I tried becoming vegetarian I failed. I often had no reasonable options in restaurants nor social events. Also I live in a country where meat is cheap and barbecues are a key social activity.
Nowadays I still eat meat 0-2 times per week when my choices are limited, but that's much better than eating 14 times per week.
Yes, but that is a result of legal framework, where handling and using independent firmware requires accepting appropriate licences, while handling and using burned-in firmware does not.
In my personal case nVidia stopped supporting my graphics card and eventually their binary blob stopped working with the newer Linux kernels so I had to chuck it away.
Now I've learnt the hard way that I'll never by nVidia again (or if I do I'll make sure it's open first). At the time nVidia was getting a lot of good press supporting Linux and I didn't realise that meant closed source blobs that would eventually stop working.
I know it's not pure freedom but I'm picking my battles these days and if I can keep something working by bridging the gap between an evolving kernel and static closed firmware then that's good enough for me.
My experience is the same as the author's: there's more today that needs non-free support to use Linux on the desktop than there was in the past.
So the friction isn't doing its job.
So why bother if the friction is just pushing users to distros which bundle the blobs more directly, instead of pushing manufacturers at all?
Is that a function of Linux Desktop simply being more capable nowadays?
Also, there is a world of difference between not publishing signed closed-source firmware for a driver that does not exist vs having an open source driver with published firmware.
It's more the examples provided in the original article: cpu firmware blobs, network firmware blobs, same as always, just even more.
I don't think so - it's more a function of hardware becoming more complicated, and vendors choosing to put a lot of the logic in firmware rather than putting it in ROM because there's a much lower chance of them getting it right first time now than there was in the past.
ROM to firmware creep is understandable, given flexibility that you mentioned. But to an end-user, is there a difference between having a blob in ROM vs firmware, if both are closed source?
I don't think so. Every component on my system that runs firmware is significantly more complicated than the iterations that didn't. 802.11ax is much harder than 802.11b, for instance.
> But to an end-user, is there a difference between having a blob in ROM vs firmware, if both are closed source?
Yeah, bugs that are discovered after release can be fixed.
I definitely understand that real benefit.
But I was questioning the premise that having blob in ROM is somehow better than needing same blob in firmware, simply because that one doesn’t need/want to distribute non-free firmware on media.
If one wanted libre hardware, then both cases are not great.
Oh right! I absolutely believe that having non-free firmware on media is preferable - in some cases this has led to free implementations that replace the non-free firmware, and even outside that, given the choice between non-free code that doesn't work and non-free code that does, I'd prefer to take the latter.
IMHO it would be far better to put effort into change at a different level, like perhaps a broader push with right to repair legislation that expands it to include releasing firmware in addition to schematics, spare parts, etc. Change is going to have to come from a different source, you're just abusing users today for no benefit.
But I don't think punishing potential users who just want to get their laptop to work is particularly effective free software advocacy. Most likely the user has no clue what wifi chip their laptop has.
... but there are two possible long term changes, right?
In one of them, the friction incrementally pushes manufacturers not to have proprietary firmware and the number of machines supported by the Debian official install improves.
In the other one, the friction incrementally discourages users from installing Debian, and they either give up on Linux or adopt a distribution that isn't quite so user-hostile.
As opposed to making it clear that this is the one you need if you're doing something outrageous like ... using a laptop or something.
The only friction you're causing is to Debian users. My current laptop is Arch after 5 laptops that ran Debian. My Debian desktop is barely functional with the friendliest AMD GPU to Linux I could find.
By my next update cycle I will be Debian free for the first time in 20 years. Debian needs less friction if it wants to keep me as a user.
You can only force anyone to do anything if you have power. If you don't have power you're just a weird guy screaming from a soapbox and people are going to ignore you. No offense but do you think nvidia cares about your boycott? There's no shortage of graphics card buyers.
I assume the non-free version of Debian sees significant amounts of downloads, maybe at this point more than the free version, so the situation is obviously quite awful or this post would not exist.
I still live back in the days when Fedora would give you a pop up to install non-free drivers ("additional software" or something like that).
I have been using Debian as my daily driver for almost 6 months now and Linux in general for years - I still have no idea how the video drivers work. I have no idea how to install them, which ones mean what, what's free, non-free, optimal, professional - I have no idea how any of it works and it's a source of frustration for me.
What is mesa? Why do people tell me not to install the video drivers from the Radeon website. When I download and install the drivers from the Radeon website there are lots of errors but it still seems to work - did I install them wrong? How to I update them?
When I install the `apt install firmware-amd-graphics` I get loads of warning, but it works so I ignore them.
What is going on and why isn't it as simple as Window's where I run an executable and a setup wizard does the rest?
Why would you want to? This is not something I want to actually have to care about. And also I absolutely don't have to.
> What is going on and why isn't it as simple as Window's where I run an executable and a setup wizard does the rest?
Do you mean the included drivers? Or the OEM ones? Or the Nvidia ones? Current or beta? Also, it is even simpler than that on distros that support nonfree stuff, e.g. on ubuntu it is just click here for which nvidia stuff you want and done. As opposed to navigate websites, versions and wizards in windows.
Generally you don't have to touch video drivers at all in linux except for arm stuff and nvidia. (that's not always the case,but mostly) Just keep your kernel up to date. That's the main problem with debian and wanting to say play games or have newer hardware.
This is why there's no need to install anything from the Radeon website -- you already have a driver.
As for the `firmware-amd-graphics` package, it provides proprietary firmware for your GPU. That could be as insignificant as microcode updates or an important part of the interface that the driver expects. It's probably a good idea to install (and maybe try and fix the warnings.)
While ROCm doesn't officially support consumer GPUs, the supported W6800 is Navi 21. The consumer Navi 21 cards thus also work. Those are the 6800 / 6800 XT and 6900 XT.
If any Debian contributors would like to help package the rest of the ROCm stack, the Debian AI mailing list is where most of the action happens.
Disclosure: I work for AMD on ROCm, but all opinions are my own.
Now a bit of cold water for real world use cases currently...
The bulk of the market is on 6700 XT and lower so ROCm is currently effectively saying 'we don't support the vast majority of our shipped cards for compute'[2] which sends a bad message to the consumer/prosumer market. Also, I do hope AMD reconsiders the requirement for PCIe 3.0 atomics support for compute as it seriously limits the utility of its cards in the consumer space.[3] It limits the compute support for most systems to a single slot (i.e. the processor direct PCIe connection) which will result in a 'no-sale' for many people.[4] I recently upgraded to a B550 system and imagine my surprise when I found out the chipset doesn't support it, but the compute drivers require it, so only one AMD card for me... this pushes me back to nVidia, or possibly Intel if they ever ship, for most of my compute needs for foreseeable future.
I don't write this to discourage you as it would be great to see AMD be successful on Linux. However, I think it's important for AMD folks to hear real world feedback on the current situation and understand you've got a long ways to go still: I'm on a full AMD system and my experience has been that it (GPU support) is far from ready for prime time so it probably won't be a full AMD system for much longer.
[1] Where key packages can disappear for months at a time and serious breakage can and does occur. I understand the need for this in unstable which is why I steer clear of it for my daily driver.
[2] Saying 'well just get a 6800 or better' isn't realistic given availability and pricing of the higher end cards to date. I had planned on getting a 6800 [XT] but was not willing to pay the markups due to scalpers/AIB markups. Even the AMD weekly drops didn't help on these cards. So I settled on a 6700 XT until prices return to a more sane level.
[3] I tried to raise this issue on the AMD forums but my message got marked as spam. <sigh>
[4] Multi-GPU setups for compute are not uncommon on consumer hardware, especially rendering and ML.
The other 5000 and 6000 series cards work to varying degrees. Unfortunately, since AMD doesn't officially support them, I can't easily get hardware to test them myself.
My understanding is that the RX 6700 XT (gfx1031) mostly works, with some caveats. Gentoo did some performance tuning for the BLAS libraries with that GPU. The AMD GPU libraries aren't built for that architecture by default, but I'd be happy to help anyone compile them from source. There's an extensive open source test suite that can be used to validate GPU functionality outside of the official hardware support list.
I've gotten both the 6700 XT and 6600 to work (that's how I discovered the PCIe atomics issue: trying to run both cards at the same time with OpenCL) which is why I initially said 'can be made to work'. It took quite a bit of fiddling and I'm sure having the drivers packaged in the Debian repos will be better still as I don't know what I don't know and knowledge on the ground re: AMD compute drivers still appears a bit thin.
[1] The part of this story I find so depressing is that AMD should be eating nVidia's lunch re: Linux support. But since at a management level AMD doesn't seem to care about consumer Linux they actually manage to make nVidia's drivers look good in comparison. It would be nice if they would take a page out of Intel's playbook and give engineers like yourself their support to actually do consumer (open source) Linux support work as your day job. I can dream...
It is the OpenGL and Vulkan libraries. Code calls the Vulkan API, and mesa translates that into something the video driver can handle, which in turn passes the command to the video card.
> Why do people tell me not to install the video drivers from the Radeon website.
Because AMD have spent a lot of time and effort getting the drivers in the kernel, and the website drivers are old and no longer the best. There is "AMDGPU Pro" video drivers, which gives you the AMD OpenGL/Vulkan libraries, but they are generally not as good as Mesa as Mesa has everyone working on it.
> How to I update them?
AMD have spent a lot of time getting them in the kernel. You update by getting the latest version of Debian, or switching to a distribution that uses a newer kernel.
> What is going on and why isn't it as simple as Window's where I run an executable and a setup wizard does the rest?
It is a lot more simple. It is all built in. This is the beauty of collaborative development, AMD and Intel get improvements based upon each others work and they concentrate on giving us better hardware. This is also the reason that AMD works so much better on Linux than Windows, and generally AMD gives Nvidia a run for its money on Linux, at least according to the last benchmarks I saw.
> Because AMD have spent a lot of time and effort getting the drivers in the kernel
Does that mean the drivers are open source?
> It is a lot more simple. It is all built in.
When it works. When you have no context (like me) and you are presented with the wrong resolution and no GPU acceleration - It's really hard to find a tl;dr of how to get your graphics card working.
Your comment was literally more valuable than a few hours on the internet trying to disambiguate the situation.
Yes, they are. Of the big three (Intel, AMD, Nvidia), only the Nvidia drivers aren't open source (there's an open source driver for Nvidia, called "nouveau", but it doesn't come from the manufacturer and isn't that good).
Define good?
I’m not gaming and nouveau works seamlessly OOB and get updated via apt. IMO that’s the best you can get.
What more do I need?
There are other points that aren't directly applicable to gaming which just aren't as great. Nouveau are pretty open and honest about how their driver will never be that great on the modern cards because Nvidia doesnt want it to be.
Yes, Intel and AMDs video drivers are opensource.
> It's really hard to find a tl;dr of how to get your graphics card working.
So I don't know what video card you have, or what the problem is. With something like Debian it could easily be that your video card is too new for the kernel version if you don't have backports enabled.
That is when the AMD website drivers come in, as they are normally for older kernels, so they are used to enable to video card while your distribution gets up to date.
This is why Ubuntu became so popular, it took Debian and made everything slightly more up to date, plus made the Nvidia drivers accessible. Though if you find Debian not quite new enough for your usecase, try something like OpenSuse or Fedora. Make a bootable USB and see if they are any better for you. Or enable backports, I found that didn't work for my usecase as I had kernel panics, and could not get to enable them.
Let me know if you have any other questions or queries.
sudo apt install -t bullseye-backports linux-image-amd64
As someone who is currently dealing with proprietary AMD drivers (and previously proprietary nVidia drivers) on Debian testing, it's pretty much only a matter of time before things will break... you've been warned but I understand if you need to do it anyways (as I do for compute support with AMD currently.) I'm very much looking forward to the day when we have Debian packages (at least in the non-free repo) for current AMD GPU compute stuff.
- router
- NAS
- thin and light laptop
- cheap laptop
- beast of a CPU laptop
- gaming laptop
- ml laptop
- low wattage compact desktop / htpc
- workstation
- gaming desktop
- tablets, watches, phones, etc.
- handheld portable gaming system (e.g. rk2020, steam deck, odin pro)
To get on the list, manufacturers would have to commit to making a Linux SKU that does not change (modulo hardware bugfixes) for 3-5 years. PC Engines alrady does this for routers; Pine does it for cheap laptops; AMD GPU gaming desktops are also an easy target.
They'd also need to donate a few dev / continuous integration machines to the project (or operate them themselves). Each night, these machines would automatically confirm that all hardware devices are working right in top of trunk and all contemporary supported releases.
There could be revenue sharing, or not. Don't care.
What OpenBSD seems better, but in many cases, you may need a wired connection for on First Install. Upgrading, no issues but I like to be plugged in anyway just in case wireless drops.
There isn't AFAIK an organisation or project dedicated to producing free firmware to replace the blobs. The most you see is the occasional project for making firmware for a single chip. Debian's page lists these: https://wiki.debian.org/Firmware/Open It's a depressingly short list.
I guess the problem is in part that there's no money to fund anyone for doing such (very hard) work, and no incentive for companies to pay for it.
Other firmware the cpu microcode is handled via packaging.
There has been a lot of work packaging peripheral firmware for things like mice... it's not like it is going to go away. This is also distributed in packaging in Fedora now.
It's not ideal that these are proprietary blobs but some accommodation is needed I think.
While it is great, it doesn't solve the problem that this post is talking about; when devices have no pre-installed firmware and expect you to upload (proprietary) firmware on every boot. This is usually things like WiFi chips, GPUs, Ethernet etc.
Same as Fedora, Debian seems to package linux-firmware?
There isn't even enough manpower to get the Raspberry Pi, the world's most used embedded system for tinkerers, a fully blob-free experience - it's madness to expect that for systems like GPUs that are orders of magnitude more complex than that.
> Modern laptops normally don't come with wired ethernet now.
Both then and now, on my laptop, I use Ethernet occasionally but rarely, only when I specifically need its speed or reliability advantages. The rest of the time I use Wi-Fi, which did and does require non-free firmware. Ethernet moved from a built-in port to a dongle, but my actual usage patterns have barely changed.
That said, there used to be some Wi-Fi chips that didn't require non-free firmware (just not the ones I had), whereas now there are none.
For desktops, on the other hand, it was and is worth hooking up Ethernet, making the point largely moot. (I think I used Wi-Fi on my desktop at the time, but I wished I didn't have to. Now I don't.)
> There won't be any usable graphics on the laptop's screen.
Even at the time, I saw hardware-accelerated graphics as essential, both for windowing (Compiz) and for gaming. So I used the 'nvidia' non-free driver on the PC… just as I do now. (I don't remember what the Mac had.)
Incidentally, for those who don't prioritize graphics performance over freedom, free drivers are better than they used to be. Unfortunately, they require non-free firmware – and I know this post is about firmware. But still… that seems like a win of some sort.
I hope Debian ignores them.
> Going back 10 years or so, most computers only needed firmware uploads to make WiFi hardware work.
You can actually go back to almost 30 years with dialup modems, most notably the winmodem/softmodem popularizing this concept in the pc market. If there is anything to blame on this path of binary blobs, I'd fault them.
Maybe hardware companies should make an effort to contribute and make their hardware work better on Linux
When I buy a laptop I make sure it is Linux/Debian/Arch... compatible
Extra I go on eBay and get a cheap compatible open source driver WiFi card, like QCNFA335 to replace the original
Slightly apart from the main issue here I would suggest that there is an opportunity for communication with some of the parties that matter most. Producing media that shows how much happier customers are with products that have freely available firmware with updates might convince some manufacturers to change their ways. If it were clear that power users would direct purchasing for whole organizations based in part on availability of firmware then there would be a clear profit motive for doing that. At this time makers see providing firmware as an expense and a risk that only diminishes their position.
That said, who cares if some users want it? Debian doesn’t aim to serve all users’ needs. Those who want non-free firmware bundled with the operating system can get that with one of the many Linux-based (or non Linux-based) operating systems out there. Debian just isn’t for those users. That’s fine.
https://wiki.debian.org/Firmware/Open
As you can see, it is very minimal.
For the most part this annoyance affects the installation step. After your system is in some semi-working state where you at least can install packages or have access to network, you can add any firmware you want already.
I read around and used an installer which apparently had proprietary drivers. Same issue, couldn't install.
At this stage I returned to Ubuntu.
InstallingDebianOn HP Envy 14 Beats Edition 2020ep https://wiki.debian.org/InstallingDebianOn/HP/Envy%2014%20Be...
> The installer isn't able to detect the Wifi or LAN network cards. To do a net install from the LAN card you need to drop to a shell and do
modprobe atl1c echo "1969 1083" > /sys/bus/pci/drivers/atl1c/new_id
One thing that could start paving a path forward would be to make contributing to Debian more approachable... Not saying they should get on GH/GL or anything, but there's huge room for improvement in their existing tools and docs.
Edit: Here we go. They're "unoffical" images for sure, but they seem to be maintained similarly to the official ones. https://cdimage.debian.org/cdimage/unofficial/non-free/cd-in...
Can anyone say with a straight face that there is anything wrong with Debian's vision or principles merely existing somewhere in the world?
Ok, so they exist somewhere in the world, in Debian.
And it predates you. Debian didn't come along and fuck up your life. There is no basis for any kind of "take it back" idea to "fix" Debian.
If you don't subscribe to those principles, great news! The entire rest of the world agrees with you! Go use Ubuntu or Mint or really anything.
If your argument is that Debian is an important base for countless users of other distros, and so it's inconsiderate or irresponsible or something not to serve all those indirect users better (according to your personal definition of "better"), well then why ARE so many other distros choosing Debian's shoulders to stand on if it's so unhelpful? All those other distros are perfectly free to provide their own base the "right" way.
You in fact are free yourself to launch a new better os to replace Debian. Go ahead. No one is stopping you. Show those backwards clods how it's done by practical people with their heads screwed on straight.