Fix length calculation in scsi_command_util::ModeSelect

[kkaempf]

OpenVMS Alpha sends a strange ModeSelect payload, apparently one byte too large. This was 'fixed' by a (wrong) length calculation in #1405, breaking #1427.

This PR

• fixes the wrong length calculation
• improves the loop test in scsi_command_util::ModeSelect to prevent a buffer overflow. (Remaining length was checked for > 0, but buffer access is at offset and offset + 1, effectively requiring 2 bytes.)
• the loop test fix makes ModeSelect6 fails with DEC Alpha system #1402 pass
• adds a testcase for ModeSelect6 fails with DEC Alpha system #1402
• adds a testcase for MODE SELECT ignores mode pages in develop branch, v23.11.01 works #1427

Fixes issue #1427

[uweseimet]

@kkaempf I may be wrong, but on first sight this change might not be correct. If the last byte of the payload is a page code (without further potentially required mandatory bytes following) the change seems to ignore this page code. This may result in a wrong (or none at all) SCSI error to be reported for the incomplete page data (incomplete for pages other than page 0).

[kkaempf]

How about c3f8d94 ?

This just breaks the loop when page code 0 is detected (it's supposed to be last anyway) and checks the (remaining) buffer length before accessing buffer data.

[uweseimet]

That's probably better, but it's up to you to test this and check that it is compliant with the specification, and that in particular there is the correct SCSI error message for pages other than 0. What's definitely not compliant (but maybe not that relevant) is that MODE SELECT supports page 0, but MODE SENSE does not. I have already elaborated on this elsewhere.

I am sorry, but I don't want to spend more time on this.
@kkaempf IMHO you are too focused on getting piscsi to work with your special hardware, but you seem to forget that any new bug may result in piscsi not working anymore for many other users.
@rdmark Regarding the two bug tickets I raised, I would like to bail out ;-). This has become an endless story, with one bug being replaced by the next. Please deal with the bug tickets as you please.

[kkaempf]

I'm sorry. It was my intention to improve piscsi.

Maybe with better contributors documentation and better test cases things will improve in the future.

[kkaempf]

@rdmark please remove my contributions to develop. I'll continue in my own branch.

Feel free to reach out in case anyone else has questions on DEC hardware.

[sonarcloud]

Quality Gate failed

Failed conditions
45.5% Coverage on New Code (required ≥ 60%)

See analysis details on SonarCloud

[rdmark]

@kkaempf I'm sorry to hear that you're withdrawing your support. I think we were close to fully cross-platform compatible code. The question what to do with vendor specific pages in a compatible way is still open, and will come up again the future I think.

I would appreciate your help reverting the changes, since you're the most familiar with that code. It's up to you if you want to assist with this.