Re: ALPHA 2.6.17-rc5 AIC7###: does not boot

[Andrew Morton <akpm@osdl.org>]

Ernst Herzberg <list-lkml@net4u.de> wrote:
>
> moinmoin.
> 
> 2.6.16.18 boots and runs without problems.
> 
> 2.6.17-rc5 with patch
> "[PATCH] alpha: generic hweight (Re: ALPHA 2.6.17-rc5 compile error)"
> http://www.ussg.iu.edu/hypermail/linux/kernel/0605.3/1559.html
> 
> hangs with 
> 
> [....]
> scsi1 : Adaptec AIC7XXX EISA/VLB/PCI SCSI HBA DRIVER, Rev 7.0
>         <Adaptec aic7895 Ultra SCSI adapter>
>         aic7895C: Ultra Wide Channel B, SCSI Id=7, 32/253 SCBs
> 
>  1:0:0:0: Attempting to queue an ABORT message
> CDB:CDB: 0x12 0x12 0x0 0x0 0x0 0x0 0x0 0x0 0x24 0x24 0x0 0x0
>  1:0:0:0: Command already completed
> aic7xxx_abort returns 0x2002
>  1:0:0:0: Attempting to queue an ABORT message
> CDB: 0x12 0x0 0x0 0x0 0x24 0x0
>  1:0:0:0: Command already completed
> aic7xxx_abort returns 0x2002
> [....]
> 
> dmesg (captured with netconsole and netcat, most messages are duplicated
> but better than nothing)
>
> ...
>

James sayeth "Best guess would be lost interrupt ...  especially if there's
no device usually at target1:0:0 (i.e.  the machine doesn't get a reply it
expects doing the initial inquiry)."

But I don't recall us making any changes in the Alpha interrupt-management
code post-2.6.16.  Perhaps it was PCI changes which introduced this
regression.