goggin, edward wrote:
Mike, Can't dec_pending now see different error values for the possibly multiple bio clones of a single original bio?
Yep.
How should it decide which error gets propagated up to the original bio?
Have no idea really. We have this problem if the bio gets partially completed with different errors for different parts too. I think this can happen in SCSI if the front were to get a medium error.
Looks like the old code and your new code just take the error value from the last completed bio which has an error. While this was probably OK when the only error value was -ENXIO, we may now need some logic in dec_pending to decide which error value has more significance.
Yeah all of that was on the TODO under my comment:
Some temporary crap I will fix is the dec_pending fucntions.
- : send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
