Re: SCSI status 0x3

[Emmanuel Dreyfus <manu%netbsd.org@localhost>]

On Mon, Dec 30, 2013 at 06:45:51PM +0100, Edgar Fuß wrote:
> a SCSI code 0x3.
> You mean scsireq->status, i.e. the SCSI status code/byte?

I thought it was, but I messed up my debug messages. I do not get
an odd SCSI status after all, it is just the ioc_status of the mpii
driver:

mpii0: mpii_scsi_cmd_done xs cmd: 0x0a len: 524288 xs_control 0x42010
mpii0:  dev_handle: 9 msg_length: 12 function: 0x00
mpii0:  vp_id: 0x00 vf_id: 0x00
mpii0:  scsi_status: 0x00 scsi_state: 0x0c ioc_status: 0x8003
mpii0:  ioc_loginfo: 0x31120322
mpii0:  transfer_count: 0
mpii0:  sense_count: 0
mpii0:  response_info: 0x00000000
mpii0:  task_tag: 0x02aa
mpii0:  bidirectional_transfer_count: 0x00000000

ioc_status = 0x8003 is 0x8000 | MPII_IOCSTATUS_INVALID_SGL

According to the Linux driver, 0x8000 means log info is available. I have 
no idea what SGL stands for.

Another strage problem: right before this error, I get an NMI. I was
notaware a PCI card could do that:

NMI ... going to debugger
Stopped in pid 0.2 (system) at  netbsd:x86_mwait+0xc:   leave
db{0}> btx86_mwait(0,0,0,dc4c8d04,c05a50d5,c4ac1d20,dc4c8d2c,c058b76e,0,0) at 
netbsd:x86_mwait+0xcx
86_cpu_idle_mwait(0,0,deadbeaf,deadbeaf,deadbeaf,c4ac1d20,c058b5e0,c4ac1d20,0,c0100321)
 at netbsd:x86_cpu_idle_mwait+0x54
idle_loop(c4ac1d20,ed6000,edf000,0,c0100307,0,0,0,0,0) at netbsd:idle_loop+0x18e

-- 
Emmanuel Dreyfus
manu%netbsd.org@localhost