Kernel 2.6.22 changes /dev/sda to /dev/hda

[Kocil]

FYI, kernel 2.6.22 has changed something.

In my laptop (Toshiba Satelite M55 series),
kernel 2.6.21 and older reports the harddisk as /dev/sda.
Now kernel 2.6.22 lists it as /dev/hda, the same as kernel 2.4.31.

[easuter]

I'm guessing that your laptop has a sata drive (from sda), right?
There were quite a few changes made to both libata and libsata in 2.6.22 acording to KernelTrap, so maybe that could be an explanation for the name change.

http://pcburn.com/article.php?sid=2343

[Kocil]

I know.
 
Just to let you know, because it is potential to screw up your machine.
I got 3 different partitions (VL, VL, Slack).
After installing the new kernel in VL and rerun vliloconf,
other partitions that have the older kernel won't boot.

The solution was to reinstall the new kernel to all partitions.

[apock627]

This was a real problem for me as I still boot from the (Windows) MBR to lilo on the boot sector of the linux partition. After editing /etc/lilo.conf under the running kernel (2.6.21), lilo refused to run after changing the root =  option from /dev/sdax to /dev/hdax. I also couldn't see how to have an older kernel and the newer kernel in lilo at the same time.

The solution found here http://www.dslreports.com/forum/r18682293-22-rant is to re-compile the 2.6.22 kernel with CONFIG_IDE_GENERIC not set, then it all goes back to /dev/sda and problem solved. There are a couple of other solutions there too, but this seemed the neatest.

Cheers

[Vxt]

Please don't cuss the kernel - it is one of best tested
(& least understood) portions of any Linux distribution

Potential  >  In Linux there is always alternate solutions:
The "standard practice" is to edit the loader
to boot any (uniquely named) kernel version desired.
Just as long as the path is correct.

I.E. The loader may have several "test" versions
concurrent to your (stable/production) favourite.

Another reason-  it's not wise to use "make CLEAN or MRPROPER"
at least until  certain old headers/etal are no longer desired.

What better way to have a fail-safe_fall-back escape route?

RFE ? >

it did whatever it is that it does while it is doing what it does when one doesn't know what it is doing. So that worked. I guess.

As I have & can personally    v/ouch 
Automation will never replace backups (or outguess user)