Comment 62 for bug 220706

As promised it's me again and... IT'S FIXED FOR ME! :-)

With 2.6.22 (Gutsy 7.10) and with Windows 2000 it worked for years. But kernel 2.6.28 seems to have a more strict timing for IDE. There was an 80pin cable with a DVD drive jumpered as slave at the _LONGER_ end, a harddrive jumpered as master in the middle and the _SHORTER_ end plugged into the mainboard. When I swapped both ends of this cable, the LiveCD of Jaunty 9.04 alpha 3 finally booted!

I even managed to boot Jaunty with the short end of the 80pin cable plugged into the mainboard, but both drives jumpered as "cable select". But then "hdparm -t" showed only 30.2MB/s - instead of 57.4MB/s with correct cabling! Older kernels or Windows do not complain about this. So I think newer kernels indeed should _NOT HANDLE_ such bad cabling, but they should display a more helpful error message than "SRST failed (errno=-16)".

After some reading about "cable select" and some benchmarking I suggest:

1. With 80pin-cables plug the _LONGER_ end into the mainboard! And set jumpers of both drives to "C S" (= "cable select"). Then the drive at the shorter end becomes master, the middle one becomes slave.

2. With 40pin-cables jumper one drive to "M A" (= "master), the other drive to "S L" (= "slave"). Ordering of connectors is not so critical, but doing it like with 80pin cables seems reasonable.

3. With only 2 drives let both become master using 2 cables. With more than 2 drives the fastest two should share one cable. And on each cable the faster drive should become master, the slower slave.

Sorry not to mention SATA - the stubborn, old system here is PATA only.
If you have further results with SATA, please add it here!