Re: lp8000 and 2.6.18-rc6

[Douglas Gilbert <dougg@torque.net>]

James Bottomley wrote:
> On Tue, 2006-09-12 at 08:45 +0200, Alberto Cammozzo wrote:
>> novantatre:~# dmesg
>> [...]
>> st 0:0:5:0: Attached scsi generic sg0 type 1
>> ch 0:0:5:1: Attached scsi generic sg1 type 8
>> sd 1:0:0:0: Attached scsi generic sg2 type 0
>> scsi 1:0:15:0: Attached scsi generic sg3 type 3
>> scsi 1:1:9:0: Attached scsi generic sg4 type 3
>> scsi 1:2:8:0: Attached scsi generic sg5 type 3
>> scsi 2:0:0:0: Attached scsi generic sg6 type 0
>> scsi 2:0:1:0: Attached scsi generic sg7 type 0
>> scsi 2:0:2:0: Attached scsi generic sg8 type 0
>> scsi 2:0:3:0: Attached scsi generic sg9 type 0
> 
> This shows it found LUNS 0 1 2 3

Wouldn't that be target identifiers 0, 1, 2 and 3
all with luns 0 (the fourth element in the scsi tuple)?

>> sd 3:0:0:0: Attached scsi generic sg10 type 0
>>
>> novantatre:~# sg_luns -v sg6
>> sg_luns: open error: sg6: No such file or directory
>> novantatre:~# sg_luns -v /dev/sg6
>>     report luns cdb: a0 00 00 00 00 00 00 00 04 00 00 00 
>>     report luns: requested 1024 bytes but got 32 bytes
>> Lun list length = 24 which imples 3 lun entries
>>
>> Output response in hex
>>  00     00 00 00 18 00 00 00 00  00 01 00 00 00 00 00 00                    
>>  10     00 03 00 00 00 00 00 00  00 04 00 00 00 00 00 00                    
>> Report luns [select_report=0]:
>>     0001000000000000
>>     0003000000000000
>>     0004000000000000
> 
> This says the system only has three luns: 1 3 4 (i.e. no 2)

No lun 0 either hence the PQual=1 for the inquiry
response for /dev/sg6 ('sg_inq /dev/sg6'). I suspect
'sg_inq /dev/sg7', sg8 and sg9 also have PQual=1.
There may be a similar pattern to 'sg_luns /dev/sg7',
sg8 and sg9 as was reported for 'sg_luns /dev/sg6'.

> So something really strange is going on here ... The linux code should
> have executed a probe and add luns for the contents of report lun, which
> should have found the correct set, unless something went wrong during
> the reporting.
> 
> Can you post the full dmesg from the time the Emulex is detected to the
> time sg attaches?  report luns is supposed to issue a failure message if
> something goes wrong (additionally, using the latest -mm would be
> helpful since that prints enhanced inquiry information from scsi-misc).

Yes dmesg would be useful. Is there something stopping
the mid level scanning luns where lun>0 (perhaps something
in the black list)?

Doug Gilbert