Re: Burning audio disc literally 30x the system load of data disc
On Tue, Aug 19, 2003 at 05:24:26PM -0700, Dave Platt wrote:
> My recollection is that the ATAPI-bus implementation on many
> kernel versions will not do DMA for block sizes other than
> 2048 bytes/block. When you burn audio, the larger block/sector
> size used for these discs causes DMA to be disabled, and the
> kernel is forced to PIO all of the data. This pushes your
> kernel CPU load up through the roof.
>
> If it's what I think it is, it's a kernel issue of long-standing,
> and I don't know when/if it's ever likely to be fixed.
Ahhhh! I even remember that thread. Silly me. I missed the point it
was making. I even responded saying 'gee, data discs work fine'.
OK, well, the current situation is not acceptible. Are we talking
'2048 is a dependency assumed by and riddled through 1000 kernel
files, drivers and sixteen abstraction layers', or is it
'unimplemented because the current infrastructure has no way to
represent it' or just 'for any one chipset you need to make six
changes, we're just lazy.'?
Jens, Doug, point me at some kernel source. I have my goggles on and
chainsaw running. Ahem. Scalpel running. Ignore the two-stroke
noise.
Monty
Reply to: