News for hackers. Stuff that matters.
Re:Debian 9 not including support for UEFI Secure Boot by giancarlostoro on Friday April 28, 2017 @07:44PM (#14224085)
Anyone know what other distros support UEFI? I hope Debian supports it in a later release I wouldn't mind going back to my second distro ever (first one was Slackware, not an option for me these days).
Re:Debian 9 not including support for UEFI Secure Boot by tadfisher on Friday April 28, 2017 @07:54PM (#14224137)
Debian still supports UEFI, just not Secure Boot (which is not very secure these days if you're using a Microsoft-signed key).
Re:Debian 9 not including support for UEFI Secure Boot by MichaelGG on Friday April 28, 2017 @08:28PM (#14224263)
Can you elaborate on why it's not secure? Too many people have MS signed keys that just work out of the box ?
Re:Debian 9 not including support for UEFI Secure Boot by JonathonW on Friday April 28, 2017 @09:13PM (#14224419)
Microsoft's secure boot key is compromised* on ARM, and it's not feasible to revoke and replace it.

AFAIK, Secure Boot is still secure on x86_64, at least if you trust Microsoft to not be doing something (either deliberately or unintentionally) malicious.

* The key isn't actually exposed, but a signed binary that will run unsigned code has been-- which allows secure boot to be bypassed without turning it off on affected systems.

Re:Debian 9 not including support for UEFI Secure Boot by devrandomguy on Friday April 28, 2017 @09:51PM (#14224572)
Found the story: https://arstechnica.com/security/2016/08/microsoft-secure-bo...

It's things like this that make me want to try using an FPGA dev kit as my next motherboard. Not because the FPGA is relevant to this, but because it demands a lot of the high speed components (CPU cores, RAM controller) that a desktop computer would have.

Mind you, having an FPGA running the show would be the ultimate in extensibility.

Re:Debian 9 not including support for UEFI Secure Boot by NeutronBoy on Friday April 28, 2017 @08:56PM (#14224359)
Debian will still support UEFI, just not SB. I use Arch and UEFI. With a kernel supporting EFISTUB, a UEFI system can boot the kernel directly, without the use of a bootloader - it's awesome!
Re:Debian 9 not including support for UEFI Secure Boot by mixmastamyk on Friday April 28, 2017 @11:49PM (#14224937)
Does that mean you can skip the efi partition?
Re:Debian 9 not including support for UEFI Secure Boot by josteink on Saturday April 29, 2017 @12:49AM (#14225096)
No. The kernel is its own bootloader and that must be on the EFI partition.

The UEFI spec pretty much says it must be there. What's your big problem with that, anyway? :)

Re:Debian 9 not including support for UEFI Secure Boot by mixmastamyk on Saturday April 29, 2017 @07:22PM (#14229346)
Waste of space, and extra complexity. This is a thread about over engineering, right?
Re:Debian 9 not including support for UEFI Secure Boot by josteink on Sunday April 30, 2017 @01:59PM (#14232890)
I'll agree lots of bits of the UEFI-spec represent solid over-engineering, but the EFI partition is IMO not one of those bits.

It standardizes how to deploy a bootloader in a multi-boot friendly way without any cheating, tricks or risk of "competing" OSes stepping on each other's toes.

It significantly reduces complexity in all cases except for the simplest of them all, and even in those cases it represents a reliable standardization so it's possible for people (like you and me!) to inspect what actually gets booted, how and why.

Try that with ad-hoc MBR-written voodoo-tracks outside the FS and partitions. Now try it again as a OS-vendor who doesn't want to risk bad PR by accidentally overwriting existing bootloaders.

I honestly think this part of the UEFI spec is among those which makes most sense and is least problematic.

Re:Debian 9 not including support for UEFI Secure Boot by mixmastamyk on Tuesday May 02, 2017 @06:22PM (#14251165)
Ok. But as a single OS user I'd prefer the space used be in the writeable area of the firmware rather than taking up my disk.
Re:Debian 9 not including support for UEFI Secure Boot by josteink on Wednesday May 03, 2017 @06:18AM (#14254403)
> But as a single OS user I'd prefer the space used be in the writeable area of the firmware rather than taking up my disk.

OK. That's your preference. But now consider this use-case:

You are installing a new OS, your first OS, on a blank PC, on a new disk. All is grand. No toes to step on. How does that OS deploy its bootloader?

You vote it should chose a simplistic way which is optimized for the single-OS use-case. It's primary objective: To save space, around 500MB of space, because you think that's something which matters.

But let's say you were actually planning for a dual-boot scenario. Now you're in the process of installing your second OS. How should that OS deploy its bootloader?

To determine that it is indeed a single-OS use-case (and this that it should use the single-OS simplistic approach, which differs from the multi-OS approach), it needs to find out if another OS and another bootloader is deployed. How should it know that? How should it be certain it doesn't overwrite something which doesn't belong to it?

And since we are now moving from a single-OS solution to a multi-OS solution... Should we migrate the existing simplistically deployed bootloader to a multi-OS config and wipe the existing single-OS configuration?

To put it bluntly: Should Windows move Linux's bootloader? Can you imagine the outrage? Any and every honest bug in such code would be deemed malicious by internet hotheads and be talked about for decades to come in threads filled with "M$".

And this is all getting very complex, very fast, isn't it? It wasn't so simple after all, was it?

With the EFI-partition there is a simple solution and simple answers for all those questions, at any point. Inspecting the state of bootloaders is simple. Deploying OSes is simple. Avoiding collisions is the default because different OSes namespace their bootloaders inside the EFI partition. Windows doesn't even need to know or care if Linux is installed!

Everything just works. At a cost. The cost of allocating a few MBs in the age when Terabytes rules.

And honestly, I find that a pretty good trade-off.

Re:Debian 9 not including support for UEFI Secure Boot by mixmastamyk on Wednesday May 03, 2017 @11:39AM (#14256665)
There's no reason these days something like an efi partition couldn't be put in flash next to the firmware, cheap. SSDs on the other hand are still expensive, after many years I was just able to afford a 512GB.

In the 90s I used the ibm? (I think) boot manager with 4 or 5 OSs, it took a one or two megabytes and worked quite well. That is more choices than are practical, honestly. And with VMs and containers less necessary than ever. (I do like GPT.) Doesn't have to be that small today, but shows what is possible. Xosl too.

Thirdly developing a completely new proprietary OS (uefi) when embedded Linux exists is about as cynical and wasteful as they come.

Re:Debian 9 not including support for UEFI Secure Boot by NeutronBoy on Wednesday May 03, 2017 @10:58PM (#14262080)
> There's no reason these days something like an efi partition couldn't be put in flash next to the firmware, cheap. SSDs on the other hand are still expensive, after many years I was just able to afford a 512GB.

Unless you want to switch your HDD to another computer and have it boot properly?

Re:Debian 9 not including support for UEFI Secure Boot by mixmastamyk on Thursday May 04, 2017 @11:21AM (#14265469)
That doesn't happen very often, and when it does you could copy the bootloader as well. Probably each OS should have a copy of it, for recovery purposes.
Re:Debian 9 not including support for UEFI Secure Boot by josteink on Thursday May 04, 2017 @01:49PM (#14267033)
So now your computer needs to figure out if it should boot the bootloader either from firmware OR from disk. How should it know that? What would be a 100% reliable way to determine this?

And if there is a conflict (because now there can be conflicts. Oh boy!)... How should it be resolved?

Again, this is a source of even more complexity. All in the name of the "simple" solution you wanted. Just because a few MBs on disk.

Can't you see where this is heading? Do you really think this is worth all that extra bullshit?

While UEFI is clearly over-engineered in many aspects, what you're asking for is under-engineering. And we all know that leads to nasty stuff down the road too.

Re:Debian 9 not including support for UEFI Secure Boot by mixmastamyk on Thursday May 04, 2017 @03:28PM (#14267916)
>So now your computer needs to figure out if it should boot the bootloader either from firmware OR from disk.

That's not what I wrote, pls read again.

Re:Debian 9 not including support for UEFI Secure Boot by cyphar on Saturday April 29, 2017 @12:41AM (#14225082)
openSUSE, Fedora and Ubuntu all work. In fact openSUSE and (I think) Fedora will enrol their own distribution keys (after prompting you) to ensure that Microsoft can't blacklist them in the future.
Re:Debian 9 not including support for UEFI Secure Boot by blcknight on Friday April 28, 2017 @07:48PM (#14224104)
AFAIK both Fedora and Ubuntu support it.
Re:Debian 9 not including support for UEFI Secure Boot by btgeekboy on Friday April 28, 2017 @08:28PM (#14224262)
I can vouch that Fedora does, decently well. My only issue is that if you're using Nvidia-vended graphics drivers, you'll need to sign the kernel modules after every kernel update & subsequent akmod build. And since you don't have RedHat's key, you need to generate your own key, upload it into your BIOS/EFI control panel, and run sign-file against each of the .ko that akmod generates. If you don't do this before rebooting, the drivers will fail to load and GDM will crash.

This is a rough overview of the process, though it's slightly outdated: http://www.pellegrino.link/2015/11/29/signing-nvidia-proprie... (I believe there's 4 modules to sign now.)

Re:Debian 9 not including support for UEFI Secure Boot by cmurf on Saturday April 29, 2017 @12:00AM (#14224977)
Fedora is using their own patches. Upstream are basically doing the ignorant asshole "ostrich maneuver" and refusing to settle the Secure Boot issue by upstreaming a solution. So every distro has their own way of doing it and it's not sustainable is basically how I view Debian's decision.

https://lists.fedoraproject.org/archives/list/[email protected]...

Re:Debian 9 not including support for UEFI Secure Boot by giancarlostoro on Saturday April 29, 2017 @03:25PM (#14228247)
Not sure why the downvotes, I asked a legitimate question. Oh well.
Re:Debian 9 not including support for UEFI Secure Boot by JdeBP on Sunday April 30, 2017 @06:28AM (#14230904)
See https://news.ycombinator.com/item?id=14224297 for the error in your question.
Re:Debian 9 not including support for UEFI Secure Boot by giancarlostoro on Sunday April 30, 2017 @12:02PM (#14232145)
I rather be downvoted and told why than left clueless. I guess my mistake is in seeing UEFI and UEFI Secure Boot as the same problem altogether. I'm pretty sure I'm not alone in this.