IDE driver updates: cd-changers? 3rd/4th ports; DMA

mark (
Tue, 8 Aug 1995 10:46:00 -0400

>From: Albert Cahalan <>
>Date: Sun, 6 Aug 1995 06:53:33 -0400
>Subject: Re: Development of ATAPI CDROM Code
>> I am preparing to modify ide-cd.c and supporting routines in order to
>> access the 3 platters in my Sanyo 3 disk changer, and was looking for a
>> few hints. Should I treat the three platters as partitions (ie hdc1,
>> hdc2, hdc3, etc) or would it be more correct to define new devices for
>> this unit. I am fairly new to kernel modifications, and are there any
>> good documents available describing the multiple levels of driver code?
>As partitions, you cause problems with Mac (and other?) CDs.
>As new devices, you run out of devices.

There are 64 minor devs (partitions) per IDE drive.
There are two IDE drives per major device (to avoid re-implementing
request-queuing, ala SCSI).

However, there are still 128 minor device numbers *unused*
for every IDE major number. Perhaps these could be taken
for changers.

>You could change the way the IDE driver works so that it is the same
>as the SCSI driver (which supports CD changers). The SCSI driver uses
>dynamic device numbers so that it doesn't run out of device numbers.

Ugh. Too much overhead and people have enough trouble identifying
their drives already, without having them hop around depending
on what disks happen to be inserted at any given time.
Maybe IDE users are not as sophisticated as SCSI users.

On another note, the reorganization of the IDE driver to dynamically
handle 1,2,3 or 4 IDE interfaces is underway.. I expect to have it
ready for release within the next couple of weeks. Just about everything
is being touched, so I'd advise anyone with new IDE schemes to get in
touch Real Soon Now.

Once that is out of the way, I'll be adding bus-mastering DMA support
for the IDE ports on the Intel Triton chipset.