Slashdot Mirror
Debian GNU/Hurd 2013 Released
jrepin writes "The GNU Hurd is the GNU project's replacement for the Unix kernel. It is a collection of servers that run on the Mach microkernel to implement file systems, network protocols, file access control, and other features that are implemented by the Unix kernel or similar kernels (such as Linux). The Debian GNU/Hurd team announces the release of Debian GNU/Hurd 2013. This is a snapshot of Debian 'sid' at the time of the Debian 'wheezy' release (May 2013), so it is mostly based on the same sources. Debian GNU/Hurd is currently available for the i386 architecture with more than 10,000 software packages available (more than 75% of the Debian archive)."
Would anyone mind explaining to me the key differences between Debian Wheezy & Debian GNU/Hurd 2013? What are the benefits of using GNU/Hurd 2013?
Libera te ex Inferis!
Oh come on. April 1st is over. Everyone knows Hurd is a running gag. It's an ancient meme.
Ha, indeed! Someone once tried to convince me that Duke Nukem Forever had been released too. I'm not so stupid that I'd fall for that!
If you're interested in understanding microkernel OS architectures, then Hurd might be useful to experiment with. Other than that its pretty close to unusuable as there isn't even basic SATA and USB support (IE you're going to have to install on OLD hardware, or much more likely in a VM where you can supply virtualized IDE).
Honestly, while I certainly don't want to rain on anyone's pet project Hurd has mostly become pointless. Its user space really offers nothing beyond what Linux or other POSIX *nix user spaces offer, and while microkernels are interesting concepts they've never proven to be terribly practical in most applications. Even in terms of microkernel design Hurd is dated. I'd think it would be much more interesting to work on future-looking OSes, say something with a Plan 9-like user space and some more modern experimental kernal with features designed around high core counts and heterogeneous compute resources. Not sure what that is, but I'm sure there are people out there working on stuff like that.
"Malo periculosam, libertatem quam quietam servitutem." -- Jefferson
Microkernel operating systems aren't inherently difficult to construct but there's a very noticeable tradeoff between the performance of a hybrid/monolithic kernel and the security/stability of a microkernel.
The performance hit comes from the hardware isolated process model used by modern microprocessors. Whenever an application needs to do something outside of its own scope, such as request additional memory, access shared resources, or interface with a device driver it makes a system call. In a monolithic system this requires the processor to switch from the running task to the kernel task, perform the requested action, and then switch back to the running task. If the kernel needs to access the tasks memory, it can access it through segmentation or shared memory with ease because the kernel in a monolithic system has no access constraints.
In a microkernel system the processor switches from the running task to an interprocess messaging task (part of the microkernel), which then copies the message to the requested server's buffer, switches to the server task, processes the message, switches back to the messaging task, copies the response back to the original client's buffer, and then switches back to the client task.
Task switches are very expensive in terms of CPU cycles, so minimizing them is key to obtaining performance. Hybrid and Monolithic kernels have a massive performance edge on modern processors because they perform a fraction as many task switches and memory operations whenever a system call is performed.
I looked at it a while back with an eye towards doing some work on it, but I'm interested in file systems and large storage and Hurd was limited to a max of 4GB per file because all files were memory mapped all the time and Hurd only runs on 32-bit architectures. So, for me, the amount of work before I could do something interesting was pretty steep.
I think the main reason that microkernels don't have great performance is because not much work has been put into them. I worked on Apple's Copland OS back in the mid-90's (the "failed" OS before OS X). Copland was a true microkernel and there were a number of performance optimizations that we'd put in. Had it shipped, we probably would have started making some modifications to the CPUs to support the microkernel better as well.
A big issue for performance is switching between processes. If you have to make multiple process switches for each kernel call that can get slow due to things like reloading the MMU tables, etc. There are a lot of different paths that could be taken. I could imagine a micro kernel, for example, written in Java or similar language running in a VM that enforced fine-grained memory controls, e.g. at the object level. If you used this for memory protection between trusted (e.g. OS level) servers you could avoid the hit of reloading the CPU's page maps. User space separations could be enforced by the CPU for better security.
>I could imagine a micro kernel, for example, written in Java or similar language running in a VM that enforced fine-grained memory controls, e.g. at the object level. If you used this for memory protection between trusted (e.g. OS level) servers you could avoid the hit of reloading the CPU's page maps. User space separations could be enforced by the CPU for better security.
Microsoft Research has done a lot of work on this exact idea. They even produced a usable operating system
http://en.wikipedia.org/wiki/Singularity_(operating_system)
Oh come on. April 1st is over. Everyone knows Hurd is a running gag. It's an ancient meme.
You mean this is "Debian does Dallas"?
Build a Man a Fire, and He'll Be Warm for a Day. Set a Man on Fire, and He'll Be Warm for the Rest of His Life.
Managing the trust graph is why it's hard. Security is always hard. On a monolithic kernel we just say: Uhm, yeah, I trust all these drivers and whatever, even though I probably shouldn't because... well... That's how it works. GNU/HURD/HIRD has a more modular approach that pushes the drivers out of kernel space, but it has some design flaws ( letting a directory node provide its own ".." -- Yikes! ), and the number of developers is next to non-existent.
Furthermore modern processors are designed for monolithic kernels. Just like x86 has a bunch of cruft from when ASM coders wanted more complex instructions (for less / easier coding), Features like Multiple Execution Ring Levels are missing. ARM gives me Two Rings. AMD x86 gives me Two Rings. Intel x86 gives me 4 rings! A ring level essentially is a hardware supported security level. Each ring allows another "mode" of security. So, with only two rings, I can create an OS that has userspace and kernel mode. With 3 rings I can have Kernel, Trusted Driver/Module/Interface, and Userspace. The barriers required to easily create a secure microkernel don't exist. With only 2 rings we have to decide if userspace or kernel mode is where a module belongs -- They don't belong in either! We Need The One Ring to be an intermediary between Ring Zero (which rules them all) and give Ring 2 to the userland, and in the darkness bind them.
Everyone's using monoliths, hardware makers give us 2 rings to make that happen. Hell the hardware even prevents adoption of new (more secure) programming paradigms. Even the virtual memory addressing system in modern chipsets is designed to work best with C. I'm working on a more secure language with separate call and data stacks, and code-pointer overwrite protections for heap data, but the x86 / x64 / ARM platforms I'm working on are built for single stacks, and thus stack smashing or buffer overflow is an inherit design flaw. Segmented memory would be great for securing functions on a per call basis -- Swapping Stacks at will, Super easy Co-Routines... but those bits were sacrificed to the More Memory God, and the registers became a part of the virtual addressing system. On 16 bit code I can do some neat things that I can't do on 32bit mode code without a huge headache, because the hardware doesn't support me doing it.
So, that's why it takes so long. Because we're trying to do stuff in software that the hardware doesn't support. These things are more secure and are great for modularity, but the hardware's designed to do it faster the monolith / C way. Note that to a program it won't matter about whether the filesystem is uber modular, or the device drivers are not in ring 0. Hell, eventually I'll port a C compiler to the multi-stack code.
Note: I don't work on GNU/HURD/HIRD, just develop my own OSs. Yeah, I could work on Linux or other POSIX OSs, but why? That's not going to advance the state of the art in Operating Systems at all. A reliable design is grand for production systems, but to make the leap from the 80's, we're going to need some new hardware to help us out. Got Viruses? Blame the Chip Maker, Language Implementer (not designer), and Operating System. Seriously, they're all doing it WRONG if security is the goal. With a separate call and data stacks on chip, One Ring more, you could actually have the damn security you want.
Stallman has already announced that he is content with Linux as the GNU kernel and that he has lost interest in the Hurd project. Hurd is kept alive now by hobbyists interested in alternative kernel designs, not Free Software demagogues.
If they dumped Hurd now it would be a complete loss of face
Yay it's the daily make shit up about the FSF/RMS thread!
http://blog.reddit.com/2010/07/rms-ama.html
TL;DR
http://lists.gnu.org/archive/html/bug-hurd/2010-08/msg00000.html
Seriously, is it hard to google RMS Hurd before posting crap?
SJW n. One who posts facts.
I suppose there's a first time for everything.
This was modded informative because the argument is so old, it's coming back as 'vintage' by hipsters.
Flexible bare-metal recovery for Linux/UNIX
There is Debian, with its GNU/BSD version:
http://www.debian.org/ports/kfreebsd-gnu/
And Gentoo has their variant:
http://www.gentoo.org/proj/en/gentoo-alt/bsd/fbsd/
Those are the only two I know about :)
They aren't. There are many microkernel OSes out there that are successful, like QNX (which has made plenty of noise about how it runs nuclear reactors and such). Hell, even Windows was completely microkernel at one point.
The main problem is performance. This comes from two problems - repeated kernel requests, and IPC.
Kernel requests happen because device drivers are run at application level (which provides great isolation). However, device drivers tend to require a lot of stuff at the kernel level (which is why they're typically in the kernel...) - things like interrupts, physical memory access, DMA, memory allocations (both physical and virtual), and such. Each one of those things it can't do alone (because well, it's an application - if applications can do those things, your microkernel is no better than DOS - the goal is to isolate things from each other). So it becomes an kernel API call to request an interrupt, to register an event object (the interrupt handler runs in the driver server as an interrupt thread), to get memory mappings installed, etc. Each API call is a system call in the end, which are generally expensive things because they require context saving and switching (some microkernel OSes use "thread migration" to mitigate this) and so forth.
The second problem is IPC. All the servers are isolated from each other and can only communicate through IPC mechanisms. So a microkernel has to end up being a message routing and forwarding service as well. Let's say an application wants to read a file it has open. It calls read(), which traps into the kernel (system call, after all), which the kernel then sends a message to the server which can handle the call (filesystem), so it passes the message to the filesystem server and then switches back to user mode so the filesystem server can handle it. The filesystem server then translates it to a block and issues a read to the partition driver (which if it's a separate server is yet another user-kernel-user transition), which then goes to the disk driver (u-k-u). From here, it goes to the bus handler (because said disk can be on SATA, IDE, USB, Firewire) where the transfer actually happens, and then the message winds it way back to the disk driver, the partition driver, the filesystem driver, then the application.
Switching from user to kernel is expensive - generally requires generating a software interrupt (system call) which triggers into the kernel's exception handler which then has to decode the request. Switching back is generally cheaper (usually just a return instruction which sets the proper mode bits), but you're still taking several mode switches per API call.
No big surprise, these things add up into a ton of cycles.
Microkernel OSes have developed means to alleviate the issue - thread migration being a big one (typically a server is implemented as a thread waiting on a mailbox, it gets the message then handles it). Thread migration means the application's thread context isn't saved, but migrated to the kernel, then passed onto the servers as necessary so instead of having to wake up threads and run the server loop, it becomes more expensive function calls, almost like RPC except the thread that called it is where it's executing on.
In a monolithic OS like Linux, all those messages and IPC are reduced down to function calls (usually through function pointers) - so the application making the system call becomes the only transition - the virtual filesystem handles the call, calls the filesystem driver, which calls the partition driver, which calls the disk driver, which calls the bus driver, ... and then they return up the stack like a typical subroutine call.
Oh, and Windows NT 3.51 did this as well. Guess what? Graphics performance sucked, which is why in NT 4, Microsoft moved the graphics driver into ring0 (kernel mode), thus creating the ability for poorly written graphics drivers to crash the entire OS. But, graphics are faster because you're not shuffling so much messages around. I think Windows has steadily put more and more of the graphics stack in the kernel since then, as well.
The Mayans were pointing to the dawn of a new era, the age of Linux on the desktop, which supposedly will last for the next B'ak'tun until Hurd is up and running on the L4 microkernel.
Lets be clear: the performance hit comes from the expensive x86/x64 trap handling. RISC processors are on the order of 30 cycles. x86/x64 is on the order of 2,000-3,000 cycles. The braindead x86 architecture is the only reason microkernels haven't already "taken over the world".
The L4 and EROS/CapROS microkernels did a lot of small hacks to reduce the above overhead, and they got some pretty decent performance even out of x86. But contrary to your previous claim, x86 makes good microkernels very difficult to construct.
Higher Logics: where programming meets science.
This is good news. I am glad people are working on BSD(s), Hurd, Minix, and other systems because it ensures technological diversity. It would be a sad state if only GNU/Linux and proprietary systems were developed. If we have a thriving ecosystem of vaious operating systems and kernels, that bodes far better for advancement than a monoculture.
"SO we bide our time, waiting for a purer kick to bloom and the future is still bleak, uncertain and beautiful" -GSYBE
Main problem for Hurd would be support for hardware who needs closed parts (firmware, binary drivers) as Hurd propably is GPL3 which essentially forbids usage of such things without disclosure to user, essentially killing any chances of having binary Nvidia driver supported. Still, most of open source stuff can be ported to be used with it.
Yeah, that is what would make it pretty much a non starter on the desktop, since it's probably GPL3 - or else, its rationale for existence separate from Linux is as strong as the rationale for East Germany or North Korea existing. Since binary blobs would be banned here, they'd be limited to Intel & AMD GPUs, however bad, and then, on top of that, run X, and GNOME 3.whatever in fallback mode, or in real mode if the drivers are liberated. In short, the best use of HURD, where it would be almost guaranteed to work right, is in CLI mode, if one is like RMS and lives in an emacs world. In which case, the login script could just as well include that, one goes into emacs, and then is off doing everything that one does there.
Just wondering if the "Libre-" Linux crowd will celebrate this, or release a list of 50 reasons why Debian doesn't pass the purity test and therefore, Debian Hurd can't be endorsed? I sure wish gNewSense comes up w/ a HURD distro based on this one.