Comment 110 for bug 438136

For all those with very large counts of bad sectors (i.e. >1000) take a look at my comment here:
http://bugs.freedesktop.org/show_bug.cgi?id=25772#c2

Basically it looks like the manufacturers are using the raw value in a different way. i.e. the 48 bit value is likely split into 3 x 16bit values that are least significant byte first. (It could be 6 x 8bit values, but I suspect not, as 256 would probably be too low - anyone know how many are normally made available for reallocations?)

From Jose Mirco's collection of Hitachi drives:
  0x01000F000000 65551
  0x010007000000 65543
  0x07002E000000 458798
  0x050011000000 327697
  0xB4001D000000 1900724

This would tie up with my old IBM/Hitachi in my comment linked above. The fact is that minus specs from manufacturers this raw value can obviously not be relied upon, and to do so is a FAIL!

If I had to hazard a guess for the first of the ones above:
0100 = 0x0001 = 1: probably equals the event count of attribute 196.
0F00 = 0x000F = 15: is the number of bad sectors
i.e. there was one event where the disk found 15 bad sectors
The last one doesn't make so much sense though:
B400 = 0x00B4 = 180
1D00 = 0x001D = 29
i.e. there were 180 events where a total of 29 bad sectors were found?!? So perhaps we can't even rely on consistency within a brand.