Post 13.0 switch from hd* --> sd* in -current with 2.6.32+

marnold · 01-09-2010, 09:29 AM

Quote:

Originally Posted by wroom

Is this yet another example of people meddling with things they shouldn't touch, with no other obvious reason than to feed the "symmetry devil"? Or am i missing something important? Is there a particularly good reason to start naming the old ide devices as scsi?

I was wondering the exact same thing. I assume that there must be some method to the madness, but I'm not sure what that method might be at this point.

GazL · 01-09-2010, 10:09 AM

Abstraction is probably what it's all about. If you write a scsi layer above the ide/sata drivers then the upside is that people only have to write apps to support scsi. The downside is you tend to find you lose a little functionality due to the need to limit yourself to a common feature set.

MannyNix · 01-09-2010, 11:05 PM

Quote:

Originally Posted by rworkman

I suspect a few people will eventually run into this problem, and while it's easy to get around for the more experienced users (the ones that *should* be running -current), we all know that everyone isn't in that group :-)
http://rlworkman.net/howtos/libata-switchover

Thank you for taking the time to write it withing a few days of being mentioned in the Changelog, also thanks to the Slackers who helped to fine tune it. It's really helpful, I'm even booting my Open and Free BSD partitions now changed to sdaX. Painless. Thanks.

I'm still trying to figure out what are the correct kernel config options to make my custom .config working. gentoo's kernel-config (3. Common problems and areas of confusion) seems to be a starting point, I'm at my 2nd try now ;-)

wroom · 01-10-2010, 03:26 AM

Quote:

Originally Posted by GazL

Abstraction is probably what it's all about. If you write a scsi layer above the ide/sata drivers then the upside is that people only have to write apps to support scsi. The downside is you tend to find you lose a little functionality due to the need to limit yourself to a common feature set.

Yes. But the common interface subset is still there between hd and sd, so one does not have to write anything extra to support both of them within the common subset.

And yes, the ioctl list for the sd devices are a little shorter than the hd ioctl list. But removing functionality for old ide disks may not be what we wanted.

The bottom line is that there is an "agreed standard" for naming devices, that should not be deviated from unless some particularly good reasons:

http://www.lanana.org/docs/device-list/devices.txt