sata_mv problems on Marvell MV88SX5081

sata_mv problems on Marvell MV88SX5081

[Jiri Kosina]

Hi,

I have this piece:

01:06.0 SCSI storage controller: Marvell Technology Group Ltd. MV88SX5081 
8-port SATA I PCI-X Controller (rev 03)

It's RocketRAID 1820A controller.

With sata_mv libata driver from 2.6.18, I am getting a load of these:

Buffer I/O error on device sda1, logical block 122126336

After a few minutes of using the device, the filesystem gets corrupted - 
write()s to some directories start returning -EIO, kernel gives

EXT2-fs error (device sda1): read_inode_bitmap: Cannot read inode bitmap - block_group = 10664, inode_bitmap = 349437953

etc. 

The proprietary open-source driver from 
http://www.highpoint-tech.com/USA/bios_rr1820a.htm doesn't seem to have 
this problem and works well so it's likely not a HW problem.

-- 
Jiri Kosina

Re: sata_mv problems on Marvell MV88SX5081

[Mark Lord]

Jiri Kosina wrote:
> Hi,
> 
> I have this piece:
> 
> 01:06.0 SCSI storage controller: Marvell Technology Group Ltd. MV88SX5081 
> 8-port SATA I PCI-X Controller (rev 03)
> 
> It's RocketRAID 1820A controller.
> 
> With sata_mv libata driver from 2.6.18, I am getting a load of these:
> 
> Buffer I/O error on device sda1, logical block 122126336

Scroll WAY BACK in the log, and find the error messages which preceeded
the first occurance of that one.  (with a bad sector, it is very likely
you will see a zillion "Buffer I/O errors", as the SCSI mid-layer is not
at all nice about how it handles them).

Cheers

Re: sata_mv problems on Marvell MV88SX5081

[Jiri Kosina]

On Wed, 18 Oct 2006, Mark Lord wrote:

> > Buffer I/O error on device sda1, logical block 122126336
> Scroll WAY BACK in the log, and find the error messages which preceeded
> the first occurance of that one.  (with a bad sector, it is very likely
> you will see a zillion "Buffer I/O errors", as the SCSI mid-layer is not
> at all nice about how it handles them).

Actually the "Buffer I/O error" is the very first error message I am 
receiving, a few minutes after the driver is initialized and there is some 
disk activity happening.

Thanks,

-- 
Jiri Kosina

Re: sata_mv problems on Marvell MV88SX5081

[Roger Heflin]

Jiri Kosina wrote:
> On Wed, 18 Oct 2006, Mark Lord wrote:
> 
>>> Buffer I/O error on device sda1, logical block 122126336
>> Scroll WAY BACK in the log, and find the error messages which preceeded
>> the first occurance of that one.  (with a bad sector, it is very likely
>> you will see a zillion "Buffer I/O errors", as the SCSI mid-layer is not
>> at all nice about how it handles them).
> 
> Actually the "Buffer I/O error" is the very first error message I am 
> receiving, a few minutes after the driver is initialized and there is some 
> disk activity happening.
> 
> Thanks,
> 

Jiri,

Make sure that you delete the arrays in the highpoint bios before
trying to use the open source driver.

I have seen alot of really funny things on several of the different
hw/sw raid controllers (not just highpoint's) if you attempt to use
a disk that someone has already tagged as raid disks, even on the
same controller it gets interesting if you attempt to use them
non-raid.

I would be interested in knowing if the open source driver can successfully
driver those boards.

                           Roger