Re:Adaptec AIC7xxx driver broken with patch. (Score 1)

Ian (mccalli) writes:

Certainly not. A driver's response to broken hardware should never be to segfault.

You are wrong Ian. The flaw in your logic is your failure to consider that the SCSI driver is responsible for causing executable code to be loaded. When the driver in question loads executable code, bad hardware can cause segmentation and other faults. Even a perfect, totally bug-free driver is susceptible.

It helps to remember that the SCSI bus is a transmission line, not an ideal digital bus. A bad hardware configuration causes a segfault when some of the data from the disk is corrupted as it passes through the SCSI bus. When the corrupted data happens to be executable code it can fault for several reasons:

• corruption results in illegal pointer.
• corruption results in an illegal jump target.
• corruption results in an illegal instruction.

With SCSI, one of the first symptoms of improper termination is spurious segmentation faults. Since there are no widespread reports of serious AIC7xxx problems, and since SCSI hardware misconfiguration is by and large the most common cause for SCSI problems, the odds are stacked against this being a driver problem.