Re: What Linux lacks : better in-kernel user interface

From: Saturday7 (
Date: 09/02/04

Date: 1 Sep 2004 17:51:45 -0700

Todd Knarr <> wrote in message news:<L2eYc.159811

> This has been suggested before, and rejected for a very good reason:
> interface code doesn't belong in the kernel where it can crash the
> whole system. Low-level framebuffer code is there, ditto the bits
> of the video drivers that access the hardware, but the UI belongs
> out in userland where if it dies it doesn't take down the whole
> rest of the system.

I wasn't very clear about what I am proposing. I am not suggesting
putting X Windows in the kernel....that would be crazy.

Nor would I ever suggest putting a windowing system in the kernel,
since its memory demands are potentially very large.

Rather I am suggesting a simple unified interface like SVGAlib,
optionally accelerated but founded on the framebuffer code,
that supports the ideas of :

1. mouse events and mouse configuration
2. basic drawing primitives e.g. line, arc, point, rectangle, fill-area,
3. a basic font-drawing primitive using a single compiled-in proportional font,
   the rest being in user space

All of this would be organized around a single device i.e.
IOCTL interface.

The reason I am suggesting this is that the most recent SVGAlib
is broken for kernel 2.4 and newer -- at least with my system --
and it seems to me silly to have two similar types of code base --
framebuffer/mouse in the kernel plus SVGAlib (also GRX and
other similar fullscreen interfaces) being updated. I say,
why not just update the kernel.

Relevant Pages

  • [BUG] panic 2.6.20-rc3 in nf_conntrack
    ... When I shut down my ppp0 interface the kernel ... This kernel had the ipp2p patch from patch-o-matic-ng applied, ... # Firmware Drivers ... # ACPI Support ...
  • [PATCH 3/3] Add documentation about why the in-kernel api is the way it is.
    ... +The Linux Kernel Driver Interface ... +kernel interface, nor does it have a stable kernel interface. ...
  • Re: [PATCH 0/7] dlm: overview
    ... > aren't just unique within a single cluster (think clusters of clusters, ... How the configuration gets from the config file to kernel is a mystery to me ... By a message over a socket, ... Let's have no magical filesystems in the core interface please. ...
  • [RFC] New kernel-message logging API
    ... Backwards compatibility with printk(), syslog, etc. ... way the whole kernel can be converted to a new interface in one go. ... const char *format; ...
  • Re: [take24 0/6] kevent: Generic event handling mechanism.
    ... this would also require the kernel to enforce an ordering. ... And it seems to be much more complicated then simply have an kevent call which tells the kernel "wake up N or 1 more threads since I cannot handle it". ... Instead the existing POSIX timer syscalls should be modified to handle SIGEV_KEVENT notification. ... the POSIX timer interface is very flexible. ...