[Prev][Next][Index][Thread]

Kernel Mode Linux with libc support



Dear TYPES colleagues,

We are glad to announce a new release of Kernel Mode Linux (KML), a
patch to the Linux kernel that allows users programs to run in kernel
mode.

  http://www.yl.is.s.u-tokyo.ac.jp/~tosh/kml/

This version of KML supports the standard C library, which means that
almost *all* *standard* user programs---such as ls, cp, find, apache,
mysql, ocaml, java, scheme, or whatever dynamically linked to
libc---can be run in kernel mode with no change at all (provided that
they are trusted to be memory safe) and can benefit from much faster
system calls.  For details, see:

  http://www.yl.is.s.u-tokyo.ac.jp/~tosh/kml/how_to_build_and_use_glibc.html

In addition, since the bug in VMware is fixed, KML can now be run on
VMware (version 4.0 beta or later)!

Feedbacks are welcome.

Best,

    Toshiyuki Maeda
    Eijiro Sumii

> From: Eijiro Sumii <sumii@yl.is.s.u-tokyo.ac.jp>
> Subject: Kernel Mode Linux (KML)
> Date: Mon, 29 Apr 2002 20:25:01 +0900
> 
> [----- The Types Forum, http://www.cis.upenn.edu/~bcpierce/types -----]
> 
> Dear Types readers,
> 
> We are pleased to announce the release of an alpha version of Kernel
> Mode Linux (KML, in short) at:
> 
>     http://www.yl.is.s.u-tokyo.ac.jp/~tosh/kml/
> 
> KML is not a joke.  Of course, Linux is an operating system kernel and
> runs in the kernel mode, but KML enables even *user* programs to run
> in the kernel mode.  Thus, those user programs have direct access to
> the kernel, which is faster than ordinary system calls in an order of
> magnitude.  Unlike kernel modules, KML allows standard mechanisms such
> as preemptive process scheduling and paging to work as usual.
> 
> But isn't this dangerous?  Well, it depends.  As all of you know very
> well, there are technologies to guarantee safety of programs without
> depending on hardware protection, such as static typing, proof
> carrying code, software fault isolation, etc.  We hope that KML will
> be of use for people who want to apply those technologies to OS
> kernels but don't have enough resources to hack them or develop one
> from scratch.
> 
> Comments and questions are welcome.  In particular, if you have
> complaints about the documentation, please tell us - it may not
> improve otherwise!
> 
> Best regards,
> 
>     Toshiyuki Maeda
>     Eijiro Sumii
> 
> P.S.  Unfortunately, KML doesn't work on the current version of VMware
> because of a bug in VMware.  This bug is already reported to the
> vendor, so it will hopefully be fixed in the near future.