lkml.org 
[lkml]   [1998]   [Aug]   [13]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: kill -9 <pid of X>
On Thu, 13 Aug 1998, Alan Cox wrote:

> > The sound driver doesn't map and copy big chunks of memory around. More to
> > the point, compare the number of crashes from Alan's sound driver work to
>
> The sound driver has to be in kernel space because it makes heavy use of
> DMA and interrupts. We are actually trying to move stuff into user space
> (you'll note that theres a definitely shift with ALSA and the newer drivers
> away from having all the advanced sequencing in kernel space. Things like
> Esound do mixing in user space and keep the kernel side clean. Right now
> sound is a good example of kernel side bloat.

Sounds like KGI/LibGGI, which is doing much the same thing. Don't
put APIs in the kernel that are any more complex than you need to use the
hardware safely. What use is made of the hardware's functionality is a
userspace issue.

> Video as it happens is going IRQ and DMA based more and more so its unlikely
> the PC Xservers will remain entirely user space.

Right. There's lots of stuff that just won't work in userspace.

> The NVidia already screws
> people horribly because of the IRQ problem. So in the longer term a kernel/user
> split is going to have to occur. Thats a foregone conclusion.

Either that or there will be a lot of stuff that Linux just can't
do at all, and a lot of hardware that Linux can't ever use properly.

> Getting the split
> right is very important.

Yes. But as long as you stick to the idea of minimalist kernel
drivers that don't do more than safely export the hardware's functionality
to userspace, the userspace stuff is maximally flexible and so is the
kernel stuff. You keep as much optimization space as possible this way,
too.

> Bizarrely enough the problem is slightly intelligent
> cards. Both dumb frame buffers and the most recent accelerated cards with
> graphic command queues work well with a user/kernel split. Stuff like the
> earlier S3 cards are where the fun is.

Oh yes. That's where you have to have the drivers wrap the
hardware accels in a sanity checking layer because the programming must be
done *just so*. Or you have to give up access to some low-level
functionality because there are hardware bugs that need to be carefully
worked around.

Jon


---
'Cloning and the reprogramming of DNA is the first serious step in
becoming one with God.'
- Scientist G. Richard Seed


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html

\
 
 \ /
  Last update: 2005-03-22 13:44    [W:0.127 / U:0.436 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site