Areca driver arcmsr crashing since 8.3

[ema-freenas]

Hi,
after observing this on two different hosts (same hardware) i suspect a driver bug has been introduced when updating arcmsr in 8.3.0-Beta3.

So, here the details:
Controller: Areca-1160ML,
Disks: 16x 2TB Seagate HDD
FS: raidz1

The machine was running rock-stable for two years running FN8.x (x<3) - after a sudden crash (while storing NFS-data) whe decided to use the downtime to upgrade to the latest stable version of FN and its updated zfs-stack.

Since then, the machine crashed sometimes, after changing some details in the cabling, swapping controllers and so on to rule out hardware-problems, we can now reproduce the failure almost immediatly. The crash-data looks always like this:

Fatal trap 12: page fault while in kernel mode
cpuid = 3, apic id = 03
fault code = supervisor read data, page not found
..
code segement = base 0x0, limit 0xfffff, type 0x1b
= DPL 0, pres 1, long 1, def32 0, gran 1
processor flags = interrupt enabled, resume, IOPL = 0
current process = 12 (irq30: arcmsr0)
[thread pid 12 tid 100045 ]
Stopped at arcmsr_drain_donequeue+0x1b: movq 0x208(%rbx),%rdi

As this function "arcmsr_drain_donequeue" was changed while updating to the latest vendor's drivers version in the last FN-release, i assume the root-cause to be somewhere there.

By starting some I/O (i.e. zfs scrub) the crash occures immediatly when the Areca-controler is running in JBOD-mode. Switching to RAID-mode and having all disks in "pass-through"-mode the crash takes some time to trigger, but will happen after a view days...

So, where do i go with this? Should i talk to areca?

Kind Regards
Marko

 

[cyberjock]

Can you post hardware specs for your machines? I have an areca 1280ML-24 and I haven't had your problem. I do run in JBOD mode and currently have 18x2TB in RAIDZ3. Here's some ideas:

1. I would go check out the FreeBSD forum and see if anyone has anything to say about the issue you are having. Be sure to read their forum rules and they strictly enforce their rule(especially for capitalization and punctuation). The issue doesn't sound like its a FreeNAS issue though.
2. Check your firmware version on your controllers. Perhaps an update would help.
3. When I checked earlier this year the version of Areca drivers included with FreeNAS is the same as FreeBSD 9.1, which also happens to be newer than what is on the Areca website. Go figure! You should verify this if you start digging deeper as I may be mistaken.
4. Check the SMART status of your disks. You will need to use areca-cli to access it.
5. If you are comfortable enough with FreeBSD you could try installing FreeBSD 9.1 and/or 8.3, importing your zpool, then starting a scrub. If the scrub doesn't cause a crash that would definitely increase the "interesting" factor.
6. You mentioned 8.3.0 beta 3, but are you trying to use the release version or did you determine that the issue appeared between beta 2 and 3?
7. Check your BIOS settings. Maybe try resetting to defaults just to rule out some BIOS setting conflicting with Areca that wasn't an issue before.
8. I assume you made no other hardware or software changes except upgrade and that the issues began after the upgrade?

 

[William Grzybowski]

Well, it is too late for this report, 8.3.0-RELEASE has already been tagged.

I suspect this is because the driver is a binary and was compiled for the kernel 8.2.

Try grabbing the 8.3 here http://www.areca.com.tw/support/s_freebsd/freebsd.htm and replacing the .ko in /boot/modules

- - - Updated - - -

Actually looks like those links are down, you'll have to contact areca... =/

 

[cyberjock]

It looks like they may be updating their webpage or something. If you click "support" from their main homepage you get a 404 error. I can't imagine they wouldn't know about this, but who knows.

Edit: I don't believe their drivers are binary, but I could be wrong. I'm tired so everything is a little fuzzy. You can access their FTP at ftp://ftp.areca.com.tw/RaidCards/AP_Drivers/FreeBSD/DRIVER/. I had this link in my bookmarks from last year.

Another edit: I did contact Areca support via email in Dec 2012 and they really didn't seem to have it together. They told me to update drivers and firmware and if that didn't work there must be something wrong but my card isn't supported anymore. Since your card is older than mine I'd expect a similar response.

 

[William Grzybowski]

Well, they provide the source but we ship the binary, no compilation...

But I might be wrong, looking at the logs this driver was introduced in 8.3, it was not in 8.2 as far as I can tell, so I don't know how that worked for you before.

 

[cyberjock]

I didn't try using my controller in FreeNAS until 8.3, so I can't vouch for how it worked in 8.2.

 

[ema-freenas]

Can you post hardware specs for your machines?

Sure, I left those details out to not spoil my first post ;)
Mainboard: Supermicro H8DA8, BIOS 1.1b , 080010 05/22/2006
CPU: 2 Dual Core AMD Opteron(tm) Processor 275 (4 cpu cores) at 2.2GHz
Memory: 16GB ECC (8x2GB)
Freenas is installed on a 2GB Flash-Modul (PATA).

Yes, it's quite old, to todays standards, but was/is sufficient to carry some disks and the cpu-power needed for ZFS.

The memory has been tested using memtest 4.0 while ECC was disabled in the BIOS. No errors after 7 cycles (~16 hours)

I switched the other machine, which had exactly the same crash-issue with FN8.3) back to Centos 6 - this system is running stable (HW-Raid10) since 3-4 weeks. Its used as a NFS-server for 3 vsphere-hosts. BTW: We have another 2 Systems running CentOS and Windows 2008R2 on the very same hardware - without problems. So i'am quite confident, the hardware-combination is solid and stable.

The machine in question here is holding ~10TB zfs-compressed backups (clones of our test-environment VMware-VM's).

Here's some ideas:

Whoa! You guys are quick. I'm deeply impressed with the amount of feedback!

Regarding your ideas:
1. Freebsd-Forum: good thinking, i will do so.
2. firmware-updates. All on latest'n'greatest (Mainboard: 1.1b, areca: 1.49) which means nothing for 6-7 years old components ;)
3. As this driver is included as vendor-driver, it's part of this story, whom to ask ;)
4. getting the full smart-status was our intention in switching to JBOD mode (in RAID/passthrough-mode you get only a subset of smart-attributes returned). So, yes, all disks are OK and finished a short selftest successfully.

5. installing a full freebsd is by all means the best option so far - after having upgraded the zpool, the option to downgrad to FN8.1 is void.

6. I am running the release-version. The README/change-notes for FN 8.3.1-RC1 list "Update the arcmsr driver to version 1.20.00.25" in the section "Changes since 8.3.0-BETA2". So i think this was introduced there, as the driver (or whatever made the controller work in the former FN-release) was changed here. And if i read this right, the troubled function was introduced in this driver-release.

7. BIOS-Settings. I did a "reset to optimal defaults" before - but customized a view settings afterwards. Might be worth a try to keep everything "vanilla".

8. Other changes: no. I swapped the controller (we have two of those systems as "spare") to make sure this is not a bad case of bit-rot. We started seeing crashes after upgrading to 8.3 after many GB (~600) where written during a backup-cycle. The last to weekends always ended in a crashed kernel. The controller was in the RAID-Mode then. We switched to JBOD-mode yesterday and can now "trigger" the crash. As the "crash-point" is always the mentioned "drain_donequeue", im quite sure, both settings only differ in the timing of the crash-cause.

So, for the time beeing, i will go with option 5 and see what happens there.

Whould be great to get in touch with the driver's developer: after solving "my" problem, we might lose the oportunity to debug this issue...

Regards
Marko

PS: Keep up the good work!

 

[cyberjock]

So you were using FreeNAS 8.1 prior to upgrading to 8.3? I might try installing 8.1 on my system temporarily just to see what happens.

I'm not sure I'd go from RAID with passthrough to JBOD without recreating the zpool. I'm wondering if something doesn't quite convert right when you go from one to the other and that's your problem. I'm really grasping at straws at this point just trying to avoid you blowing a week of troubleshooting to find the problem is something simple and stupid.

 

[ema-freenas]

So you were using FreeNAS 8.1 prior to upgrading to 8.3? I might try installing 8.1 on my system temporarily just to see what happens.

The crash also occured on the other machine - which was freshly installed with 8.3, zpool was created using the JBOD mode from start. So - no - upgrading freenas or switching the controler mode isn't the root cause.

Just an idea: Maybe some non-ATA-traffic ist triggering the problem. smartd ? (btw: the freenas 8.3 in its last state yesterday, with an aborted "scrub" from the last crash, crashed repeatable during startup of freenas, as soon as "dbus" was starting).

Side-note: I installed freebsd 9.1 yesterday, imported the pool - which started a scrub. I will give you feedback how this system behaves, as soon as i have done some more tests.

- - - Updated - - -

7. BIOS-Settings. I did a "reset to optimal defaults" before - but customized a view settings afterwards. Might be worth a try to keep everything "vanilla".

Test done: loaded "save defaults" as well as "optimal defaults", nevertheless, FN crashed as usual.

 

[ema-freenas]

So, just to give you feedback how things are currently: FreeBSD 9.1 is scrubbing for 17 houres now, still 5 to go:

Code:

root@freenas:/root # zpool status
  pool: vm-backup
 state: ONLINE
  scan: scrub in progress since Tue Mar 12 11:06:03 2013
        12.3T scanned out of 16.4T at 208M/s, 5h42m to go
        0 repaired, 74.99% done
config:

        NAME                                            STATE     READ WRITE CKSUM
        vm-backup                                       ONLINE       0     0     0
          raidz1-0                                      ONLINE       0     0     0
            gptid/edd32591-2268-11e1-920b-003048772720  ONLINE       0     0     0
            gptid/edec69fe-2268-11e1-920b-003048772720  ONLINE       0     0     0
            gptid/ee03bf95-2268-11e1-920b-003048772720  ONLINE       0     0     0
            gptid/ee1af1a4-2268-11e1-920b-003048772720  ONLINE       0     0     0
            gptid/ee32c4a4-2268-11e1-920b-003048772720  ONLINE       0     0     0
            gptid/ee4a7dc9-2268-11e1-920b-003048772720  ONLINE       0     0     0
            gptid/ee621a09-2268-11e1-920b-003048772720  ONLINE       0     0     0
            gptid/ee79b401-2268-11e1-920b-003048772720  ONLINE       0     0     0
            gptid/ee90d4eb-2268-11e1-920b-003048772720  ONLINE       0     0     0
            gptid/eea80d36-2268-11e1-920b-003048772720  ONLINE       0     0     0
            gptid/eebfe450-2268-11e1-920b-003048772720  ONLINE       0     0     0
            gptid/eed7964f-2268-11e1-920b-003048772720  ONLINE       0     0     0
            gptid/eeef88f7-2268-11e1-920b-003048772720  ONLINE       0     0     0
            gptid/ef071df6-2268-11e1-920b-003048772720  ONLINE       0     0     0
            gptid/ef1e8748-2268-11e1-920b-003048772720  ONLINE       0     0     0
        spares
          da15p2                                        AVAIL

errors: No known data errors

BTW: I don't expect you guys to solve this problem for me - as it seems i'm the only person in the world using this combination of hardware. Should the problem exist in FreeBSD 9.1 as well - then we might dig deeper and get in thouch with areca... If time allows, i could setup a debug-environment using spare-parts we have around.

 

[cyberjock]

Smartd/smartctl doesn't work with the 1100 and 1200 series areca controllers. The only way to obtain SMART data is through the areca-cli from my experience. That's why I have a script that emails me all of the SMART data for all of the drives every night. I can't trust the built-in SMART reporting in FreeNAS to warn me of a disk with problems.

 

[ema-freenas]

Smartd/smartctl doesn't work with the 1100 and 1200 series areca controllers.

In my experience, in RAID-mode (each drive is a passthrough-volume), yes, the values returned from smartctl are rather obscure. The drives are handled as "Fibre-Channel" devices - only a subset of smart-values is returned.

In JBOD-mode the drives are handled as "normal" ATA-devices. In this mode, i think, one can trust the results returned - all the attributes i would expect, are shown.

This is what my smartd.conf looks like (FreeBSD 9.1 currently)

Code:

/dev/arcmsr0 -d areca,1  -H -m alert@domain.net
/dev/arcmsr0 -d areca,2  -H -m alert@domain.net
/dev/arcmsr0 -d areca,3  -H -m alert@domain.net
...

Oh, btw, my "idea" about the smartd triggering the problem has been invalidated: I could start and test smartd on all 16 drives in parallel while the zfs-scrub is (still) running.

 

[cyberjock]

Well, if you are getting any smart information that's more than I get. I get an error along the lines of "device not supported with this version".

 

[ema-freenas]

So, the system is running FreeBSD 9.1 for nearly two days now. It survived a full scrub, two 60GB NFS-writes (VM-cLones), is currently syncing a ~2 TB multi-mio-files-FS...

No crash so far. I think i will keep this setup.

This might lead to the insight that something is wrong with FN 8.3 and its areca-driver (at least in my environment). As no other FN-user is experiencing this issue and me having a workaround, i think its best, to simply ignore this for the time beeing.

Thanks for all your ideas and help.

Kind Regards
Marko Engelmann

- - - Updated - - -

I get an error along the lines of "device not supported with this version".

Hm, thats wierd.... I am sure, i could at least run smart self-test on the individual drives (even in passthrough-mode) using FN8.1. FN8.3, while in JBOD-mode could also access the smart-data. Maybe thats a "special" on my 1160-controller ;)

FYI - just an example of the results for my first drive (it looked the same on FN8.3):

Code:

root@freenas:/ # smartctl -a --device=areca,1 /dev/arcmsr0
smartctl 6.0 2012-10-10 r3643 [FreeBSD 9.1-RELEASE amd64] (local build)
Copyright (C) 2002-12, Bruce Allen, Christian Franke, www.smartmontools.org

=== START OF INFORMATION SECTION ===
Model Family:     Seagate Constellation ES (SATA 6Gb/s)
Device Model:     ST2000NM0011
Serial Number:    Z1P0HGSX
LU WWN Device Id: 5 000c50 03eea1a28
Firmware Version: SN02
User Capacity:    2,000,398,934,016 bytes [2.00 TB]
Sector Size:      512 bytes logical/physical
Rotation Rate:    7202 rpm
Device is:        In smartctl database [for details use: -P show]
ATA Version is:   ATA8-ACS T13/1699-D revision 4
SATA Version is:  SATA 3.0, 6.0 Gb/s (current: 3.0 Gb/s)
Local Time is:    Thu Mar 14 19:43:39 2013 CET
SMART support is: Available - device has SMART capability.
SMART support is: Enabled

=== START OF READ SMART DATA SECTION ===
SMART overall-health self-assessment test result: PASSED

General SMART Values:
Offline data collection status:  (0x82) Offline data collection activity
                                        was completed without error.
                                        Auto Offline Data Collection: Enabled.
Self-test execution status:      (   0) The previous self-test routine completed
                                        without error or no self-test has ever
                                        been run.
Total time to complete Offline
data collection:                (  617) seconds.
Offline data collection
capabilities:                    (0x7b) SMART execute Offline immediate.
                                        Auto Offline data collection on/off support.
                                        Suspend Offline collection upon new
                                        command.
                                        Offline surface scan supported.
                                        Self-test supported.
                                        Conveyance Self-test supported.
                                        Selective Self-test supported.
SMART capabilities:            (0x0003) Saves SMART data before entering
                                        power-saving mode.
                                        Supports SMART auto save timer.
Error logging capability:        (0x01) Error logging supported.
                                        General Purpose Logging supported.
Short self-test routine
recommended polling time:        (   1) minutes.
Extended self-test routine
recommended polling time:        ( 307) minutes.
Conveyance self-test routine
recommended polling time:        (   2) minutes.
SCT capabilities:              (0x10bd) SCT Status supported.
                                        SCT Error Recovery Control supported.
                                        SCT Feature Control supported.
                                        SCT Data Table supported.

SMART Attributes Data Structure revision number: 10
Vendor Specific SMART Attributes with Thresholds:
ID# ATTRIBUTE_NAME          FLAG     VALUE WORST THRESH TYPE      UPDATED  WHEN_FAILED RAW_VALUE
  1 Raw_Read_Error_Rate     0x000f   081   063   044    Pre-fail  Always       -       139304884
  3 Spin_Up_Time            0x0003   091   090   000    Pre-fail  Always       -       0
  4 Start_Stop_Count        0x0032   100   100   020    Old_age   Always       -       20
  5 Reallocated_Sector_Ct   0x0033   100   100   036    Pre-fail  Always       -       0
  7 Seek_Error_Rate         0x000f   078   060   030    Pre-fail  Always       -       61870540
  9 Power_On_Hours          0x0032   087   087   000    Old_age   Always       -       11401
 10 Spin_Retry_Count        0x0013   100   100   097    Pre-fail  Always       -       0
 12 Power_Cycle_Count       0x0032   100   100   020    Old_age   Always       -       19
184 End-to-End_Error        0x0032   100   100   099    Old_age   Always       -       0
187 Reported_Uncorrect      0x0032   100   100   000    Old_age   Always       -       0
188 Command_Timeout         0x0032   100   100   000    Old_age   Always       -       0
189 High_Fly_Writes         0x003a   100   100   000    Old_age   Always       -       0
190 Airflow_Temperature_Cel 0x0022   066   064   045    Old_age   Always       -       34 (Min/Max 27/36)
191 G-Sense_Error_Rate      0x0032   100   100   000    Old_age   Always       -       4
192 Power-Off_Retract_Count 0x0032   100   100   000    Old_age   Always       -       18
193 Load_Cycle_Count        0x0032   100   100   000    Old_age   Always       -       20
194 Temperature_Celsius     0x0022   034   040   000    Old_age   Always       -       34 (0 19 0 0 0)
195 Hardware_ECC_Recovered  0x001a   117   099   000    Old_age   Always       -       139304884
197 Current_Pending_Sector  0x0012   100   100   000    Old_age   Always       -       0
198 Offline_Uncorrectable   0x0010   100   100   000    Old_age   Offline      -       0
199 UDMA_CRC_Error_Count    0x003e   200   200   000    Old_age   Always       -       0

SMART Error Log Version: 1
No Errors Logged

SMART Self-test log structure revision number 1
Num  Test_Description    Status                  Remaining  LifeTime(hours)  LBA_of_first_error
# 1  Short offline       Completed without error       00%     11329         -

SMART Selective self-test log data structure revision number 1
 SPAN  MIN_LBA  MAX_LBA  CURRENT_TEST_STATUS
    1        0        0  Not_testing
    2        0        0  Not_testing
    3        0        0  Not_testing
    4        0        0  Not_testing
    5        0        0  Not_testing
Selective self-test flags (0x0):
  After scanning selected spans, do NOT read-scan remainder of disk.
If Selective self-test is pending on power-up, resume after 0 minute delay.

 

[cyberjock]

smartctl -a --device=areca,1 /dev/arcmsr0

And what do you know.. I just tried that and it worked. Go figure! I'm wondering if I had the parameters wrong before. /shrug