Re: errors using new Samsung 870 EVO SSD

On Sun, 15 Sep 2024 02:40:44 -0400, Winston wrote:

I previously wrote (in part):

It looks like there's some kind of compatibility or interface problem.

 

Aug 21 03:01:24: (ada0:ahcich0:0:0:0): CAM status: Auto-Sense Retrieval Failed
Aug 21 03:01:24: (ada0:ahcich0:0:0:0): Error 5, Unretryable error

 ...

Aug 21 03:01:25: (ada0:ahcich0:0:0:0): CAM status: ATA Status Error

 ...

Aug 21 03:01:26: (ada0:ahcich0:0:0:0): CAM status: Uncorrectable parity/CRC error
----------
Here's smartctl -x output, keeping only what looked "interesting"/relevant:

 ...

0x06  0x018  4              64  ---  Number of Interface CRC Errors

 
Careful reading of the error messages and check of smartctl info
indicated to me some kind of data/interface/compatibility error, not a
read/write problem in the SSD.
 
Google search turned up various Samsung 860 and 870 EVO articles.
 
What I've determined:
 
* It's not a bad data cable:
 
   Some of the articles suggested the problem might be a bad cable.
   I ordered two new ones (SATA III).
   Result: no improvement.
   That's as I expected, since I was pretty sure my cables were good,
   but it was cheap to try.
 
* Fix 1:
 
   Based on the articles I read, this problem with Samsung 870 EVO and
   860 EVO SSDs appears to affect not just FreeBSD, but *BSD and at
   least some Linuxes.
 
   The solution/workaround (at least for FreeBSD) is to disable command
   queueing ("camcontrol negotiate $theSSD -T disable").
 
* Fix 2:
 
   Connect the SSD with a USB-to-SATA adapter cable.
   I found such a cable in stock at Best Buy for $12.
   Perhaps this works because there's no command queueing over USB.
 
As a side note, instead of turning off tagged queueing, I also tried
reducing the number of tags from 32 to 2.  Didn't help: the errors
continued to happen.
 
Maybe some day Samsung will come out with new firmware that fixes this
problem.
 
Is this something I should post to bugzilla (it's not a FreeBSD bug,
though) or to some FreeBSD forum (which one)?  Google no longer gets
USENET, so I don't expect this article would be found by Google search.

Report your findings in bugzilla.  A developer may add a quirk to
the scsi subsystem to automatically detect the drive and "do the
right thing".  I would use somthing like "SCSI tag-queue error with
Samsung 870 EVO SSD" as the title.

--
steve