An Internet-Ready OS From Scratch in a Week — Rump Kernels on Bare Metal


August 08, 2014 posted by Antti Kantee

The most time-consuming part of operating system development is obtaining enough drivers to enable the OS to run real applications which interact with the real world. NetBSD's rump kernels allow reducing that time to almost zero, for example for developing special-purpose operating systems for the cloud and embedded IoT devices. This article describes an experiment in creating an OS by using a rump kernel for drivers. It attempts to avoid going into full detail on the principles of rump kernels, which are available for interested readers from rumpkernel.org.

[Read More] [8 comments]

 

Interview with Amitai Schlair member of The NetBSD Foundation’s Board of Directors


December 17, 2013 posted by Matthew Sporleder

New interview with schmonz [0 comments]

 

Survey of rump kernel network interfaces


December 17, 2013 posted by Antti Kantee

A cyclic trend in operating systems is moving things in and out of the kernel for better performance. Currently, the pendulum is swinging in the direction of userspace being the locus of high performance. The anykernel architecture of NetBSD ensures that the same kernel drivers work in a monolithic kernel, userspace and beyond. One of those driver stacks is networking. In this article we assume that the NetBSD networking stack is run outside of the monolithic kernel in a rump kernel and survey the open source interface layer options.

[Read More] [0 comments]

 

PCI driver support for rump kernels on Xen


September 18, 2013 posted by Antti Kantee

Yesterday I wrote a serious, user-oriented post about running applications directly on the Xen hypervisor. Today I compensate for the seriousness by writing a why-so-serious, happy-buddha type kernel hacker post. This post is about using NetBSD kernel PCI drivers in rump kernels on Xen, with device access courtesy of Xen PCI passthrough.

[Read More] [2 comments]

 

Running applications on the Xen Hypervisor


September 17, 2013 posted by Antti Kantee

There are a number of motivations for running applications directly on top of the Xen hypervisor without resorting to a full general-purpose OS. For example, one might want to maximally isolate applications with minimal overhead. Leaving the OS out of the picture decreases overhead, since for example the inter-application protection offered normally by virtual memory is already handled once by the Xen hypervisor. However, at the same time problems arise: applications expect and use many services normally provided by the OS, for example files, sockets, event notification and so forth. We were able to set up a production quality environment for running applications as Xen DomU's in a few weeks by reusing hundreds of thousands of lines of unmodified driver and infrastructure code from NetBSD. While the amount of driver code may sound like a lot for running single applications, keep in mind that it involves for example file systems, the TCP/IP stack, stdio, system calls and so forth -- the innocent-looking open() alone accepts over 20 flags which must be properly handled. The remainder of this post looks at the effort in more detail.

[Read More] [5 comments]

 

Google Code-In 2012/2013 (from jdf)


July 02, 2013 posted by Matthew Sporleder

Google Code-In (GCi) is a project like Google Summer Of Code (GSoC), but for younger students. While GSoC is aimed at university students, i.e. for people usually of age 19 or older, GCi wants to recruit pupils for Open Source projects.

When applying for participation, every project had to create a large number of potentially small tasks for students. A task was meant to be two hours of work of an experienced developer, and feasible to be done by a person 13 to 18 years old. Google selected ten participating organisations (this time, NetBSD was the only BSD participating) to insert their tasks into Google Melange (the platform which is used for managing GCi and GSoC).

Then, the students registered at Google Melange, chose a project they wanted to work on, and claimed tasks to do. There were many chats in the NetBSD code channel for students coming in and asking questions about their tasks.

After GCi was over, every organisation had to choose their two favourite students who did the best work. For NetBSD, the choice was difficult, as there were more than two students doing great work, but in the end we chose Mingzhe Wang and Matthew Bauer. These two "grand price winners" were given a trip to Mountain View to visit the Google headquarters and meet with other GCi price winners.

You can see the results on the corresponding wiki page

There were 89 finished tasks, ranging from research tasks (document how other projects manage their documentation), creating howtos, trying out software on NetBSD, writing code (ATF tests and Markdown converters and more), writing manpages and documentation, fixing bugs and converting documentation from the website to the wiki.

Overall, it was a nice experience for NetBSD. On the one hand, some real work was done (for many of them, integration is still pending). On the other hand, it was a stressful time for the NetBSD mentors supervising the students and helping them on their tasks. Especially, we had to learn many lessons (you will find them on the wiki page for GCi 2012), but next time, we will do much better. We will try to apply again next year, but we will need a large bunch of new possible tasks to be chosen again.

So if you think you have a task which doesn't require great prior knowledge, and is solvable within two hours by an experienced developer, but also by a 13-18 year old within finite time, feel free to contact us with an outline, or write it directly to the wiki page for Code-In in the NetBSD wiki.

[1 comment]

 

jdf's Summer of Code project


June 27, 2013 posted by Matthew Sporleder

Julian Djamil Fagir wrote a blog post about his GSoC project

As one of five, I've been chosen for participating in Google Summer Of Code (GSoC) this year for NetBSD. My project is to write a binary upgrade tool for NetBSD, optionally with a “live update” functionality.

Why an upgrade tool? – Yes, updating currently is easy. You download the set tarballs from a mirror, unpack the kernel, reboot, unpack the rest, reboot, and done. But this is an exhausting procedure, and you have to know that there are actually updates, and what they affect.

read more
[0 comments]

 

A Rump Kernel Hypervisor for the Linux Kernel


April 23, 2013 posted by Antti Kantee

Ever since I realized that the anykernel was the best way to construct a modern general purpose operating system kernel, I have been performing experiments by running unmodified NetBSD kernel drivers in rump kernels in various environments (nb. here driver does not mean a hardware device driver, but any driver like a file system driver or TCP driver). These experiments have included userspaces of various platforms, binary kernel modules on Linux and others, and compiling kernel drivers to javascript and running them natively in a web browser. I have also claimed that the anykernel allows harnessing drivers from a general purpose OS onto more specialized embedded computing devices which are becoming the new norm. This is an attractive possibility because while writing drivers is easy, making them handle all the abnormal conditions of the real world is a time-consuming process. Since the above-mentioned experiments were done on POSIX platforms (yes, even the javascript one), the experiments did not fully support the claim. The most interesting, decidedly non-POSIX platform I could think of for experimentation was the Linux kernel. Even though it had been several years since I last worked in the Linux kernel, my hypothesis was that it would be easy and fast to get unmodified NetBSD kernel drivers running in the Linux kernel as rump kernels.

A rump kernel runs on top of the rump kernel hypervisor. The hypervisor provides high level interfaces to host features, such as memory allocation and thread creation. In this case, the Linux kernel is the host. In principle, there are three steps in getting a rump kernel to run in a given environment. In reality, I prefer a more iterative approach, but the development can be divided into three steps all the same.

  1. implement generic rump kernel hypercalls, such as memory allocation, thread creation and synchronization
  2. figure out how to compile and run the rump kernel plus hypervisor in the target environment
  3. implement I/O related hypercalls for whatever I/O you plan to do

Getting basic functionality up and running was a relatively straightforward process. The only issue that required some thinking was an application binary interface (ABI) mismatch. I was testing on x86 where Linux kernel ABI uses -mregparm=3, which means that function arguments are passed in registers where possible. NetBSD always passes arguments on the stack. When two ABIs collide, the code may run, but since function arguments passed between the two ABIs result in garbage, eventually an error will be hit perhaps in the form of accessing invalid memory. The C code was easy enough to "fix" by applying the appropriate compiler flags. In addition to C code, a rump kernel uses a handful of assembly routines from NetBSD, mostly pertaining to optimizations (e.g. ffs()), but also to access the atomic memory operations of the platform. After assembly routines had been handled, it was possible to load a Linux kernel module which bootstraps a rump kernel in the Linux kernel and does some file system operations on the fictional kernfs file system. A screenshot of the resulting dmesg output is shown below.

fs demo screenshot

It is one thing to execute a computation and an entirely different thing to perform I/O. To test I/O capabilities, I ran a rump kernel providing a TCP/IP driver inside the Linux kernel. For a networking stack to be able to do anything sensible, the interface layer needs to be able to shuffle packets. The quickest way to implement the hypercalls for packet shuffling was to use the same method as a userspace virtual TCP/IP stack might use: read/write packets using the tap device. Some might say that doing this from inside the kernel is cheating, but given that the alternative was to copypaste the tuntap driver and edit it slightly, I call my approach constructive laziness.

The demo itself opens a TCP socket to port 80 on vger.kernel.org (IP address 0x43b484d1 if you want to be really precise), does a HTTP get for "/" and displays the last 500 bytes of the result. TCP/IP is handled by the rump kernel, not by the Linux kernel. Think of it as the Linux kernel having two alternative TCP/IP stacks. Again, a screenshot of the resulting dmesg is shown below. Note that unlike in the first screenshot, there is no printout for the root file system because the configuration used here does not include any file system support. Yes, you can ping 10.0.2.17.

net demo screenshot

As hypothesized, a rump kernel hypervisor for the Linux kernel was easy and straightforward to implement. Furthermore, it could be done without making any changes to the existing hypercall interface thereby reinforcing the belief that unmodified NetBSD kernel drivers can run on top of most any embedded firmwares just by implementing a light hypervisor layer.

There were no challenges in the experiment, only annoyances. As Linux does not support rump kernels, I had to revert back to the archaic full OS approach to kernel development. The drawbacks of the full OS approach include for example suffering multi-second reboot cycles during iterative development. The other tangential issue that I spent a disproportionately large amount of time with was thinking about how releasing this code would affect existing NetBSD code due to GPL involvement. My conclusion was that this does not matter since all code used by the current demo is open source anyway, and if someone wants to use my code in a product, it is their problem, not mine.

For people interested in examining the implementation, I put the source code for the hypervisor along with the test code in a git repo here. The repository also contains the demos linked from this article. The NetBSD kernel drivers I used are available from ftp.netbsd.org or by getting buildrump.sh and running ./buildrump.sh checkout.

[2 comments]

 

pointers to rpi docs 2013Q1


March 23, 2013 posted by Matthew Sporleder

We get a lot of comments asking for tips on using the raspberry pi so I thought I would point out some docs:
evbarm/rpi wiki docs
An example of the rpi.img can be found here:
http://nyftp.netbsd.org/pub/NetBSD-daily/HEAD/201303221130Z/evbarm/binary/gzimg/ notice the HEAD (NetBSD -current), datestamp, arch path for future reference

There are also some concerns about building a kernel/img on your own.
building NetBSD
build.sh is one of the best features of NetBSD. You can cross compile from almost any other unix-like system with very little difficulty.

[1 comment]

 

NetBSD binary kernel modules usable on Linux in rump kernels


December 13, 2012 posted by Antti Kantee

Some years ago I wrote about the possibility to load and use standard NetBSD kernel modules in rump kernels on i386 and amd64. With the recent developments in buildrump.sh and the improved ability to host rump kernels on non-NetBSD platforms, I decided to try loading a binary NetBSD kernel module into a rump kernel compiled for and running on Linux. The hypothesis was that the NetBSD kernel modules should just work since both the NetBSD kernel and Linux processes use the ELF calling convention, and all platform details are abstracted by the rump kernel hypercall layer. Sure enough, after two small fixes to the hypervisor I could mount and access a FFS file system on Linux by using ffs.kmod as the driver.

[Read More] [2 comments]

 

gnats, mail-index outage


December 06, 2012 posted by S.P.Zeidler

The machine normally running www.NetBSD.org and also gnats and mail-index.NetBSD.org had a hardware problem. It is now working again on a new chassis. [0 comments]

 

Kernel Drivers Compiled to Javascript and Run in Browser


November 07, 2012 posted by Antti Kantee

The unique anykernel capability of NetBSD allows the creation of rump kernels, which are partially paravirtualized kernels running on top of a high-level hypervisor. This technology e.g. enables running the same file system driver in the monolithic kernel or as a microkernel style server in userspace. POSIX-compatible systems have been more or less supported as rump kernel hypervisors for the past 5 years. A long-time goal has been to extend hypervisor support further, for example to embedded systems. This would bring the solid driverbase of NetBSD available to such systems with only the cost of implementing the hypervisor.

To see how far things can go, last week I started toying with the idea of using a javascript engine as a rump kernel hypervisor. I was planning to compile the NetBSD kernel sources into javascript and manually implement the hypervisor. After some searching for a C->javascript compiler, I found emscripten, which translates C into javascript via LLVM bitcode. Not only is the compiler itself extremely mature, but there is also extensive support for the POSIX API. This meant that I could not only compile the kernel drivers to javascript with emscripten, I could also compile the existing POSIX hypervisor and have it work.

The approach of compiling kernel drivers into javascript allows them to be directly accessed from existing javascript code. Yes, I did add a sys/arch/javascript into the kernel source tree. This contrasts the approach taken by another similar experiment, where an x86 Linux is run inside a x86 machine emulator running in a javascript engine.

I have thrown together a small proof-of-concept demo of how to build a web service with the capability to access file system images using kernel file system drivers compiled to javascript. I compiled a rump kernel with support for the FFS, tmpfs and kernfs file systems. This rump kernel backend is tied to a lightweight web page which passes requests from forms to the rump kernel and displays results. When the javascript is run, it downloads an FFS image (rump.data), bootstraps a rump kernel, and mounts the FFS image r/o at /ffs. The status can be further manipulated with interactive commands.

The demo is available here. I've tested it to work with Firefox and tested it to not work with Internet Explorer. YMMV with other browsers. Note, the javascript and the FFS image together are close to 5.5MB in size, so the page may load for a few moments over a slow link -- javascript is not exactly compact and whitespace removal was the only size reduction technique I used. If you're interested in comparing the generated javascript with the C sources, you can also look at the unoptimized version (14MB).

[16 comments]