1

In my research on developing a low-cost external drive bay for my server, I was presented with a critical issue in attaching commodity SATA drives through any SAS expander based on the LSI controller chip. Is anyone else aware of this issue?

Is there more information available on exactly which chips are affected and avoided?

The specific issue is highlighted by the script hosted here on GitHub.

I do not know if the above example is the only way to make this fault occur or is just the easiest way. The problem seems to be something like a deadlock that is permanently written to the drive, making its data inaccessible. I can't tell if that means a bit of drive surgery with a donor board would allow data recovery or if the drive data is permanently scrambled or destroyed. The script from github is expanded below and appears to me a method of secure erasure so it is unsurprising that data is lost, but that does not appear to be the issue being highlighted.

echo --security-set-pass
hdparm --user-master u --security-set-pass 123 $1
sleep 1
echo --security-erase
hdparm --user-master u --security-erase 123 $1 &
sleep 1
echo --security-set-pass
hdparm --user-master u --security-set-pass 123 $1 &

Update:

A present configuration I have uses an 8 bay Netstor drive enclosure with two MiniSAS connectors, each of which are connected to ports on a PCIe LSI SAS card in the server. This has for several years hosted SATA drives, and to the best of my knowledge the Netstor box does not feature the interposers mentioned by @davidgo. Is this a practical use case that essentially validates the configuration or is it not?

J Collins
  • 191
  • 7
  • 1
    I'm not an LSI expert - I do vaguely recall a strange issue a long time ago on a dell server which required devices called interposes to allow SATA drives to be used on its SAS backplane. This was unusual because sata drives normally work fine when connected to SAS. – davidgo Apr 20 '20 at 20:36
  • 1
    https://www.dell.com/community/PowerEdge-HDD-SCSI-RAID/Interposers-what-are-they-for/td-p/3752135 – davidgo Apr 20 '20 at 20:38
  • @davidgo would you say that my LSI SAS host card would have the same limitation that is implied by this issue? – J Collins Apr 22 '20 at 15:07
  • On rereading I can't understand what you are saying the issue us, other then, possibly, that your data drives don't support secure erase? Many don't. Have you read https://ata.wiki.kernel.org/index.php/ATA_Secure_Erase ? – davidgo Apr 22 '20 at 19:01
  • @davidgo I was happy to go with SATA drives on the expander however another forum member said it was a really bad idea because of the issue presented by that github script. So now I am trying to learn more about the issue and whether I should be worried about it. – J Collins Apr 22 '20 at 19:57
  • Still don't understand. Is the following comments helpful? - Not all drives support secure erase. It is fine to use a drive that does not support extended features like secure erase in a NAS/with an LSI controller. Using full disk encryption is more secure and can often be used as a work around to lack of secure erase. – davidgo Apr 22 '20 at 20:03
  • @davidgo sorry haven't been clear. I have no interest in using secure erase on the disks, I have just been making the guess that those commands are just an example of the types of commands that could lock a disk out, not that that was the *only* way it can be done, so hoping to find out whether I should be worried. – J Collins Apr 22 '20 at 21:22
  • 1
    you do not need to be worried on that count. – davidgo Apr 22 '20 at 22:47
  • @davidgo cheers for the positive comment. How do you figure that other commands are unlikely to do the same thing? – J Collins Apr 24 '20 at 10:47
  • Because secure erase is not supported by all SATA disks. Its failing because the disks don't support it, not the controller. – davidgo Apr 24 '20 at 10:52
  • Ah so it is just not gracefully handling unsupported commands. I expect there would be a list of unsupported commands that would all lead to the same issue. – J Collins Apr 24 '20 at 10:58

1 Answers1

0

According to Seagate technical sales, "...with the new technologies[,] issues like these have been resolved."

If this can be widely accepted, then the issue is not apparent with modern SATA drives.

J Collins
  • 191
  • 7