[BUG] Fusion MPT 2.6.8 driver and multi LUN device

[BUG] Fusion MPT 2.6.8 driver and multi LUN device

[Manu]

Under a fresh Mandrake 10.1 with a 2.6.8 kernel, the Fusion MPT driver 
do recognize the HP tape autoloader. The autoloader is composed of tape 
drive (on LUN 0) and the changer (on LUN 1).
The message comes from scsi_scan.c :
     printk(KERN_WARNING "scsi: %s lun%d has a LUN larger"
	       " than allowed by the host adapter\n",
	       devname, lun);

Playing a bit with printk it looks like the num_luns is well computed 
(the value is 2), but the 2 LUN values seems to be miss-aligned.
When displaying the 8 bytes raw buffer (through lunp) in the loop :
  00 00 00 00 00 00 00 00
  00 00 01 00 00 00 00 00

I think the second buffer should be :
  00 01 00 00 00 00 00 00

Which, when computed, gives 1 instead of 16777216 (0x1000000).

The autoloader works like a charm under the 2.4.27 kernel provided with 
the Mandrake 10.1.

- Emmanuel Varagnat

PS: the 2 harddrives in RAID1 are well working, even with the 2.6.8 kernel.

Re: [BUG] Fusion MPT 2.6.8 driver and multi LUN device

[Douglas Gilbert]

Manu wrote:
> 
> Under a fresh Mandrake 10.1 with a 2.6.8 kernel, the Fusion MPT driver 
> do recognize the HP tape autoloader. The autoloader is composed of tape 
> drive (on LUN 0) and the changer (on LUN 1).
> The message comes from scsi_scan.c :
>     printk(KERN_WARNING "scsi: %s lun%d has a LUN larger"
>            " than allowed by the host adapter\n",
>            devname, lun);
> 
> Playing a bit with printk it looks like the num_luns is well computed 
> (the value is 2), but the 2 LUN values seems to be miss-aligned.
> When displaying the 8 bytes raw buffer (through lunp) in the loop :
>  00 00 00 00 00 00 00 00
>  00 00 01 00 00 00 00 00
> 
> I think the second buffer should be :
>  00 01 00 00 00 00 00 00
> 
> Which, when computed, gives 1 instead of 16777216 (0x1000000).
> 
> The autoloader works like a charm under the 2.4.27 kernel provided with 
> the Mandrake 10.1.
> 
> - Emmanuel Varagnat
> 
> PS: the 2 harddrives in RAID1 are well working, even with the 2.6.8 kernel.

Perhaps you could try the sg_luns utility from the
sg3_utils package ( http://www.torque.net/sg )
with either the 2.4 or 2.6 kernel. It should be sent
to lun 0 (i.e. the tape drive), for example:
   sg_luns /dev/st0

[I'm not sure whether st devices support the SG_IO
ioctl in lk 2.6.8; they do now. If that fails use the
sg_map utility to find the corresponding sg device and
apply sg_luns to that sg device.]

If the sg_luns response shows an offset result for the
tape changer (lun 1), then you should be looking to see
if HP has new firmware for that tape autoloader.

The difference between lk 2.4 and lk 2.6 is that 2.6
uses the REPORT LUNS SCSI command whereas 2.4 just polled
the possible luns (with INQUIRY).

Doug Gilbert

Re: [BUG] Fusion MPT 2.6.8 driver and multi LUN device

[Manu]

Douglas Gilbert wrote:
> 
> Perhaps you could try the sg_luns utility from the
> sg3_utils package ( http://www.torque.net/sg )
> with either the 2.4 or 2.6 kernel. It should be sent
> to lun 0 (i.e. the tape drive), for example:
>   sg_luns /dev/st0
> 
> [I'm not sure whether st devices support the SG_IO
> ioctl in lk 2.6.8; they do now. If that fails use the
> sg_map utility to find the corresponding sg device and
> apply sg_luns to that sg device.]

It doesn't work with /dev/st0 or /dev/nst0 but either work with /dev/sg1 
and /dev/sg2 which are respectively the drive and the changer. But it 
reports the same things that I reported.

> If the sg_luns response shows an offset result for the
> tape changer (lun 1), then you should be looking to see
> if HP has new firmware for that tape autoloader.

OK.
I tried to force the kernel to register the LUN 1 (by modifying the 
scsi_scan.c code) but the kernel crashes...

See you.