On Sun, 2002-11-17 at 02:26, Rob Weir wrote: > On Sat, Nov 16, 2002 at 05:20:05AM -0500, Mark L. Kahnt wrote: > > That said, unless you *really* need any of these and they are compiled > > in, framebuffer is a re-implementation of text mode in the kernel with > > slower methods and more overhead, and as such isn't exactly the most > > useful. Or is this one of those parts of the kernel that got included > > because some potential 2% of all eventual kernel users might be able to > > make use of this (similar to the kernel http server) and somebody was > > able to lobby Linus Torvalds enough to get it past his good judgement? > > I think Linus sees it as a necessary evil; non-x86 platforms don't all > have actual VGA hardware, so you need someway to get a text console and > fb is it. I remember someone trying to get a patch to improve it (on > x86) into Linus' tree, and the response was something along the lines of > 'either use a text-mode VT or use X and be happy', i.e. it's a dirty > sucky hack that you should avoid if at all possible. That doesn't surprise me - it was my understanding of why it was created in the first place (needed on some non i386 systems,) and that is why it has stayed experimental in so many ways on i386 type devices. I have to admit that I'm surprised that it is apparently the default on some Debian i386-based kernels as a result. > > Another interesting point is that the Linux Low-Latency howto says to > never ever switch VTs or scroll them when using fb if you're trying to > record any sort of audio, since the fb code disables all interupts on > your system for some large fraction of a second. Most of the system hangs I've had over the past year have come from using a framebuffer console while X was running, and the console was scrolling. Only <Alt><SysRq><B> would work in that situation (ick!) Screen redraws were often bad until the most recent kernels (2.4.16), and I often had to hop to another vt and back to get the screen to display properly. It made X11 far more the core of my functioning than would otherwise be the case (I go back to GCOS on Honeywell, VMS on Vax, RMX on PDP-11s, and 4.1BSD on DEC - long before these Gooey GUIs. > > > In theory, it could be very useful with X11 as the X Server for > > FrameBuffer would allow X11 on systems where there was framebuffer > > support, but not X11 video card modules, except that at present, X11 > > supports more video cards. > > You can already do this; give X the 'useframebuffer' option and you can > make use of video cards that X has never heard of. Of course, as you > say, X supports many more video cards than the kernel does. > > -rob -- Mark L. Kahnt, FLMI/M, ALHC, HIA, AIAA, ACS, MHP ML Kahnt New Markets Consulting Tel: (613) 531-8684 / (613) 539-0935 Email: kahnt@hosehead.dyndns.org
Attachment:
signature.asc
Description: This is a digitally signed message part