Hiya,
I wrote last night telling you all about my experience with a failed drive.
Interestingly, I've pulled up something which is relevant to the ATA-8 implementation.
The drive which failed was setting the top 16 bits of every word in its IDENTIFY DEVICE block to be &EA.
Now, if we look at the current implementation of checksums, this drive would set the top bits to be &EA, rather than &A7, representing 'this checksum is valid'.
Thus, even if my ATA driver checked for checksums, this drive would have gotten around this as it would be interpreted as 'checksum invalid'.
May I propose that we introduce a measure in ATA-8 which allows for potential loss of transmission lines, as well as against total corruption of the identify block? At the simplest measure, two words set as 010101010101 etc and the following word as 101010101010 etc would test this.
This drive would successfully evade all anti-fault measures in ATA-7; definitely an interesting case and worth considering IMHO for ATA-8.
Best wishes,
Drew -- Signature huh??? Hmm...I dunno...
