Is there still sufficient reason for hw.ata.atapi_dma being 0 by default?

[Chuck Swiger]

Søren Schmidt wrote:
> Chuck Swiger wrote:
>> [ FWIW, I've got a 16/10/40x Yamaha burner which just predates the 
>> first generation of burners with underrun protection-- this affects me 
>> directly. ]
> 
> Hmn, you should be able to burn 16x in PIO4 mode...

It's like trying to burn at 4x over a USB 1 channel-- 4x times 300KBs *fits* 
into 10Mbs, but not by much, and not if anything else is taking some 
bandwidth.  I almost always have another device (generally another CD-ROM or 
DVD-ROM drive) sharing the secondary IDE channel with a burner.

For what it's worth, enabling atapi_dma seems to help a significant fraction 
of the people using the dvd+rw-tools port, although issues with ATAPICAM are 
more common.

>> Oh.  Ewww.  Could chipsets which do that be added to a "quirks" table 
>> similar to the way USB devices are being handled?  Or is it not just 
>> the chipset, but some more complex interaction between ATAPI DMA and 
>> other devices in the system which want to do DMA which causes the lockup?
> 
> Right, its a combination of chipset and device, the matrix would be 
> impossible to maintain.

I would feel the same way about trying to make the ATA code work on devices I 
didn't have available to test, but you're doing a remarkable job of handling 
impossible tasks right now, Soren.  :-)

-- 
-Chuck