Hello all,

I have also got severe problems with mvsas, but have managed to at least 
make it usable. My config is a AOC-SASLP-MV8 with an HP SAS expander, 
with WD and Seagate sata disks connected to the expander using a norco 
4020 case that does not have any on board expanders. Kernel version is 
2.6.33 with latest srinivas patch. I also dont raid, the 13 disks have 
their own filesystem.

My experiences are:
The only filesystem that works is JFS. XFS and btrfs crash the 
controller making all the disks unreadable and that can only be solved 
by rebooting. (mkfs succeeds, crashes happen only after mounting or when 
fscking)
JFS works ok as long as i access it via samba or make copies and moves 
with cp and mv. If I try file operations with Thunar file manager, the 
controller crashes. Smartctl has not caused any problems so far, and 
works ok.

Attached are some kernel logs captured from those crashes.

Kind regards,
Konstantinos Skarlatos





On 6/6/2010 3:13 μμ, Jelle de Jong wrote:
> Dear Srini,
>
> I spent a few weeks catering information and did some intensive
> testing the last few days.
>
> Srinivas Naga Venkatasatya Pasagadugula wrote, on 06-05-10 08:01:
>> 1. Is this is the problem with only WD SATA drives? (I don't have WD SATA drives to reproduce this issue.)
>> 2. Whether the problem is with direct attached SATA drives or drives connected in expanders also?
>> 3. Could you please provide me the "dmesg" log or "/var/log/messages" log.
>> 4. How much the capacity of SATA drives connected to controller?
>> 5. Your HBA is having 6440 chipset?
> With a few tricks I managed to boot my OS from the mvsas controller. I
> got eleven different sata disks attached from a 4-port mini-sas
> backplate without expander to the two Marvel 88SE63xx/64xx mvsas
> controllers in my system.
>
> I managed to add five mdadm raid1 arrays without added the actual
> active sync device (so one disk for each array) I build my lvm systems
> on top of this and did a lot of file transfers for testing.
>
> This worked stable enough, there are HDIO_GET_IDENTITY errors during
> boot and operation, but the hard disks seem to be working.
>
> So to debug if issues are related to some brand of harddisk, I started
> to add WDC, Hitachi, SAMSUNG, Maxtor and Seagate disks to the
> respected raid1 array, the disks are of different sizes (320GB, 500GB,
> 1TB.)
>
> The sync starts and will fail directly or a while later on failures in
> the mvsas driver. I attached an failure example as attachment.
>
> The failures are of grave severity, I have lost complete lvm2 volumes
> and raid arrays during testing.
>
> Do you also have a SuperMicro AOC-SASLP-MV8 controller for testing?
>
> I would love to use the controllers for production, but they are
> currently unstable. I hope this information helps to solve the mvsas
> issues?
>
> With kind regards,
>
> Jelle de Jong