Been wrapping my head around this one...It all boils down to there being something that causes a problem supporting all three (2 work fine in any slot) of my SuperMicro AOC-SAT2-MV8 cards (PCI-X 64bit slots 100/133 mhz). To test, I backed up my flash, installed a fresh 6.2.1 with no modification, remove all three of my MV8s, configure the sd with my pro key...accepts fine and shows my two on boad old cache drives.

Shutdown, install 1 card, boot fine, 2 cards boot fine, 3rd...get hard error. Even moving slots does not change the error and swapping in a spare MV8 does not change it.

 

I've been running 6.2 for a while prior to the 6.2.1 upgrade and nothing but a reboot from the automatic update started the problem.

 

It does not get far enough to dump an error file to the flash, the MV8 bios's show the detection of all drives as normal. The point that the issue happens is just after it appears unraid goes through each drive. I have a video of the problem and a screen shot of the final error with the second screen shot with the cards moved down a slot.

 

I'm in the process of trying 6.1.9 just to verify, if it does work fine, I would love to find a direct link to make sure that the bug is after 6.2.

That depends on which driver unRAID is using for the Marvell 88SX6081. At least, this seems to be the only Marvell chipset for SATA2 8 port. For all I know, they could have a driver that's hard coded to max out with two devices attached to the system and buffer overruns itself when three or more are attached.

Wow...after digging for hours, finally ended up being a drive in the array that detects fine, sounds fine, but the onboard drive controller causes something in either the controller driver or something deeper that caused the hard error. I have never seen a drive cause that much havoc on a system without showing symptoms first. The problem had me worried my backplane, motherboard/mem/proc that was going or something crazy like a bad sata cable.

 

With how many drives I have in the array, the easiest thing to do was pull half the drives, attempt boot, pull half of those, attempt boot....finally found it. In case it matters, it was a Seagate Barracude 3TB ST3000DM001 from 2013.

 

The error does not have anything to do with the version of unraid as I tried with a fresh, 6.1.9, and finally have it up and running minus one drive on 6.2.1. Sucks to have so many variables changing with something so strange.

Has anyone seen a bad controller port or a single port go bad on a backplane without affecting other ports?

 

I swapped out what I thought was a bad drive, plugged it into a windows machine and used diskinternals app to read the contents, and it reads and works fine. Plugged in a new 5TB drive into the same bay and I am still having problems during the preclear. The preclear of the drive gives me a hard error that completely brings down the server (as before), requiring hard reboot. I will probably swap the SATA cable out for that bay just to be sure as well.

 

Perhaps i will swap controller ports for that bay to verify a controller or backplane issue.

I'm still getting a traced error when trying to clear a new 5tb drive. The error crashes the server completely even though the I've changed controller ports, I've changed controllers, it checks out fine on a windows system using the seagate diags (smart).

 

Changed the drive over to an internal sata port on the motherboard, and so far it's preclearing fine.

 

Sooo...researching sata_mv...it is a "rare bug" that I'm dancing around...From the sata_mv.c documentation...https://github.com/yandex/smart/blob/master/drivers/ata/sata_mv.c

 

44  * 80x1-B2 errata PCI#11:

45  *

46  * Users of the 6041/6081 Rev.B2 chips (current is C0)

47  * should be careful to insert those cards only onto PCI-X bus #0,

48  * and only in device slots 0..7, not higher.  The chips may not

49  * work correctly otherwise  (note: this is a pretty rare condition).

 

So now the question is how do I tell if I have Rev B2 chips or C0 chips...I have the Revision 3 version of the cards, and can't see anything on the chips with those designations. The latest bios is 1.0b and they've all been updated. I'll have to check if the cards are able to be on bus #0....I'm thinking they are all on bus 3 from the startup drive checks.

 

Has anyone had any experience with these cards?

Did you ever determine the root cause to this problem??  I'm trying to add new 8TB WD Red's to my system and when hot plugging them, unraid recognized them, but the preclear crashed the system and I can't boot with them installed.

On 11/1/2016 at 12:56 PM, bsim said:

Sooo...researching sata_mv...it is a "rare bug" that I'm dancing around...From the sata_mv.c documentation...https://github.com/yandex/smart/blob/master/drivers/ata/sata_mv.c

 

44  * 80x1-B2 errata PCI#11:

45  *

46  * Users of the 6041/6081 Rev.B2 chips (current is C0)

47  * should be careful to insert those cards only onto PCI-X bus #0,

48  * and only in device slots 0..7, not higher.  The chips may not

49  * work correctly otherwise  (note: this is a pretty rare condition).

 

I just didn't run the last slots of each controller and it did work ok.

 

I've since moved the array to an entirely new mb/proc/mem/raid controller and didn't have to dance around it for too much longer.

Edited October 12, 20178 yr by bsim