SSD dropping off bus

[Peter Jeremy]

Several months ago, I added a OCZ Agility4 SSD to my 8.3 system. It
had been running without problem but yesterday the system reported:

Oct 28 08:54:36 server kernel: ahcich3: Timeout on slot 25 port 0
Oct 28 08:54:36 server kernel: ahcich3: is 00000000 cs 00000000 ss ffc1ffff rs ffc1ffff tfd 50 serr 00000800 cmd 0004b017
Oct 28 08:57:40 server kernel: ahcich3: AHCI reset: device not ready after 31000ms (tfd = 000000d0)
Oct 28 08:57:40 server kernel: ahcich3: Timeout on slot 16 port 0
Oct 28 08:57:40 server kernel: ahcich3: is 00000000 cs 00010000 ss 00000000 rs 00010000 tfd d0 serr 00000000 cmd 0004b017
Oct 28 08:57:40 server kernel: ahcich3: AHCI reset: device not ready after 31000ms (tfd = 000000d0)
Oct 28 08:57:40 server kernel: ahcich3: Timeout on slot 16 port 0
Oct 28 08:57:40 server kernel: ahcich3: is 00000000 cs 00010000 ss 00000000 rs 00010000 tfd d0 serr 00000000 cmd 0004b017
Oct 28 08:57:40 server kernel: (ada6:ahcich3:0:0:0): lost device
Oct 28 08:57:40 server kernel: ahcich3: AHCI reset: device not ready after 31000ms (tfd = 000000d0)
Oct 28 08:57:40 server kernel: ahcich3: Timeout on slot 16 port 0
Oct 28 08:57:40 server kernel: ahcich3: is 00000000 cs ffff07ff ss ffff07ff rs ffff07ff tfd d0 serr 00000000 cmd 0004b017

Power-cycling the SSD recovered it and smartctl shows no problems.
This makes me suspect it's an incompatibility between the SSD and
the SATA controller or FreeBSD driver.

The SSD is directly connected to a ATI IXP600 (one of the onboard
ports on my Gigabyte GA-MA770-DS3) and is partitioned 3 ways:
- a mirror in my ZFS root
- ZIL for tha main data pool
- L2ARC for tha main data pool

Has anyone else seen this behaviour?

The SSD firmware is currently 1.4.1.3 but the latest firmware on the
OCZ site is 1.5.2 (though the release notes don't describe the
differences). Has anyone tried either firmware version?

[nate keegan]

I tried all sorts of things and the solution for me was to ditch
Crucial SSD for straight SATA disks.

http://lists.freebsd.org/pipermail/freebsd-hardware/2012-October/007135.html

Post by Peter Jeremy
Several months ago, I added a OCZ Agility4 SSD to my 8.3 system. It
Oct 28 08:54:36 server kernel: ahcich3: Timeout on slot 25 port 0
Oct 28 08:54:36 server kernel: ahcich3: is 00000000 cs 00000000 ss ffc1ffff rs ffc1ffff tfd 50 serr 00000800 cmd 0004b017
Oct 28 08:57:40 server kernel: ahcich3: AHCI reset: device not ready after 31000ms (tfd = 000000d0)
Oct 28 08:57:40 server kernel: ahcich3: Timeout on slot 16 port 0
Oct 28 08:57:40 server kernel: ahcich3: is 00000000 cs 00010000 ss 00000000 rs 00010000 tfd d0 serr 00000000 cmd 0004b017
Oct 28 08:57:40 server kernel: ahcich3: AHCI reset: device not ready after 31000ms (tfd = 000000d0)
Oct 28 08:57:40 server kernel: ahcich3: Timeout on slot 16 port 0
Oct 28 08:57:40 server kernel: ahcich3: is 00000000 cs 00010000 ss 00000000 rs 00010000 tfd d0 serr 00000000 cmd 0004b017
Oct 28 08:57:40 server kernel: (ada6:ahcich3:0:0:0): lost device
Oct 28 08:57:40 server kernel: ahcich3: AHCI reset: device not ready after 31000ms (tfd = 000000d0)
Oct 28 08:57:40 server kernel: ahcich3: Timeout on slot 16 port 0
Oct 28 08:57:40 server kernel: ahcich3: is 00000000 cs ffff07ff ss ffff07ff rs ffff07ff tfd d0 serr 00000000 cmd 0004b017
Power-cycling the SSD recovered it and smartctl shows no problems.
This makes me suspect it's an incompatibility between the SSD and
the SATA controller or FreeBSD driver.
The SSD is directly connected to a ATI IXP600 (one of the onboard
- a mirror in my ZFS root
- ZIL for tha main data pool
- L2ARC for tha main data pool
Has anyone else seen this behaviour?
The SSD firmware is currently 1.4.1.3 but the latest firmware on the
OCZ site is 1.5.2 (though the release notes don't describe the
differences). Has anyone tried either firmware version?
--
Peter Jeremy

[Peter Jeremy]

Post by nate keegan
I tried all sorts of things and the solution for me was to ditch
Crucial SSD for straight SATA disks.
http://lists.freebsd.org/pipermail/freebsd-hardware/2012-October/007135.html

Yes, I saw that thread (and even provided input). You have different
brand SSDs, attached to different SATA controllers and a different
FreeBSD version (though I'm not sure if the ahci(4) driver is different).

My suspicion is the SSD firmware, though it's possible there's an issue
in the driver that is triggered by something unique to SSDs in general.