Dual 2940UW/v1.32 on linux 2.0.33/aic7xxx-5.0.8 looses bigtime

[Leif Johansson]

Hi,

I am in desperate need of help and I hope that this is the right
forum. I am running linux 2.0.33 with the aic7xxx-5.0.8 driver
using two 2940UW cards (the SCSI BIOS version is 1.32). I am
using four fujitsu mab3091 9GB disks (two per controller) at 16MHz. 
When trying to copy large amounts of data (~ 800M) either from the net
or between the disks I get "Underflows" and resets. (I have included
an outtake from /var/log/messages below). I have swaped cables,
verified that the disks are correctly terminated, etc etc etc. I have
tried to use older version of the driver (specifically the version
usually shiped in 2.0.33 and 2.1.90). I have used both 2.0.33 and 
2.1.90 (only with older dirver). I have even tried to disable tagged 
command queuing on these older versions. I use the following boot-
options (from /etc/lilo.conf):
	
	mem=128M aic7xxx=extended,verbose,ultra

Any help would be appreciated. BTW I have never had any problems 
with mke2fs. It is only when I try to use the filesystem (I have also 
tried striping two disks with the same result) that it breaks down.

After having written the above I decided to try on final thing: I tried
to set all disks to 10MHz in bios. The disks on the scsi0-controller 
all show up as 10MHz but the disks on the second controller show up as
20MHz!! (yes, I have checked the bios settings twice). When I try to mount
a filesystem on the second controller this fails with timeouts. See the 
second outtake below.  Please help!

=======================================================================

Mar 18 13:41:50 fs1 kernel: (scsi1:0:8:0) Underflow - Wanted at least 124928, g
ot 1024, residual SG count 82.
Mar 18 13:41:51 fs1 kernel: scsi1 channel 0 : resetting for second half of retr
ies.
Mar 18 13:41:51 fs1 kernel: SCSI bus is being reset for host 1 channel 0.
Mar 18 13:41:51 fs1 kernel: (scsi1:0:8:0) Reset called, no scb, flags 0x1
Mar 18 13:41:51 fs1 kernel: (scsi1:0:-1:-1) Channel reset
Mar 18 13:41:51 fs1 kernel: SCSI disk error : host 1 channel 0 id 8 lun 0 retur
n code = 80000
Mar 18 13:41:51 fs1 kernel: scsidisk I/O error: dev 08:21, sector 44446, absolu
te sector 44509
Mar 18 13:41:52 fs1 kernel: (scsi1:0:8:0) Using 16 bit (Wide)transfers.
Mar 18 13:41:52 fs1 kernel: (scsi1:0:8:0) Synchronous at 16.0MHz, offset 8.
Mar 18 13:42:49 fs1 kernel: (scsi1:0:9:0) Using 16 bit (Wide)transfers.
Mar 18 13:42:49 fs1 kernel: (scsi1:0:9:0) Synchronous at 16.0MHz, offset 8.
Mar 18 13:44:02 fs1 kernel: (scsi0:0:9:0) Underflow - Wanted at least 98304, go
t 48640, residual SG count 34.
Mar 18 13:44:05 fs1 kernel: mand for SCB 5, QOUTCNT 0, QINCNT 0, SCB flags 0x0,
 cmd 0x0
Mar 18 13:44:05 fs1 kernel: (scsi0:-1:-1:-1) CMDCMPLT without command for SCB 4
, QOUTCNT 0, QINCNT 0, SCB flags 0x0, cmd 0x0
Mar 18 13:44:05 fs1 kernel: (scsi0:-1:-1:-1) CMDCMPLT without command for SCB 0
, QOUTCNT 0, QINCNT 0, SCB flags 0x0, cmd 0x0
Mar 18 13:44:05 fs1 kernel: (scsi0:-1:-1:-1) CMDCMPLT without command for SCB 1
4, QOUTCNT 0, QINCNT 0, SCB flags 0x0, cmd 0x0
Mar 18 13:44:05 fs1 kernel: (scsi0:-1:-1:-1) CMDCMPLT without command for SCB 1
1, QOUTCNT 0, QINCNT 0, SCB flags 0x0, cmd 0x0
Mar 18 13:44:05 fs1 kernel: (scsi0:-1:-1:-1) CMDCMPLT without command for SCB 1
5, QOUTCNT 0, QINCNT 0, SCB flags 0x0, cmd 0x0
Mar 18 13:44:05 fs1 kernel: (scsi0:-1:-1:-1) CMDCMPLT without command for SCB 1
2, QOUTCNT 0, QINCNT 0, SCB flags 0x0, cmd 0x0
Mar 18 13:44:05 fs1 kernel: (scsi0:-1:-1:-1) CMDCMPLT without command for SCB 1
0, QOUTCNT 0, QINCNT 0, SCB flags 0x0, cmd 0x0                Mar 18 13:41:50 fs1 kernel: (scsi1:0:8:0) Underflow - Wanted at least 124928, g
ot 1024, residual SG count 82.
Mar 18 13:41:51 fs1 kernel: scsi1 channel 0 : resetting for second half of retr
ies.
Mar 18 13:41:51 fs1 kernel: SCSI bus is being reset for host 1 channel 0.
Mar 18 13:41:51 fs1 kernel: (scsi1:0:8:0) Reset called, no scb, flags 0x1
Mar 18 13:41:51 fs1 kernel: (scsi1:0:-1:-1) Channel reset
Mar 18 13:41:51 fs1 kernel: SCSI disk error : host 1 channel 0 id 8 lun 0 retur
n code = 80000
Mar 18 13:41:51 fs1 kernel: scsidisk I/O error: dev 08:21, sector 44446, absolu
te sector 44509
Mar 18 13:41:52 fs1 kernel: (scsi1:0:8:0) Using 16 bit (Wide)transfers.
Mar 18 13:41:52 fs1 kernel: (scsi1:0:8:0) Synchronous at 16.0MHz, offset 8.
Mar 18 13:42:49 fs1 kernel: (scsi1:0:9:0) Using 16 bit (Wide)transfers.
Mar 18 13:42:49 fs1 kernel: (scsi1:0:9:0) Synchronous at 16.0MHz, offset 8.
Mar 18 13:44:02 fs1 kernel: (scsi0:0:9:0) Underflow - Wanted at least 98304, go
t 48640, residual SG count 34.
Mar 18 13:44:05 fs1 kernel: mand for SCB 5, QOUTCNT 0, QINCNT 0, SCB flags 0x0,
 cmd 0x0
Mar 18 13:44:05 fs1 kernel: (scsi0:-1:-1:-1) CMDCMPLT without command for SCB 4
, QOUTCNT 0, QINCNT 0, SCB flags 0x0, cmd 0x0
Mar 18 13:44:05 fs1 kernel: (scsi0:-1:-1:-1) CMDCMPLT without command for SCB 0
, QOUTCNT 0, QINCNT 0, SCB flags 0x0, cmd 0x0
Mar 18 13:44:05 fs1 kernel: (scsi0:-1:-1:-1) CMDCMPLT without command for SCB 1
4, QOUTCNT 0, QINCNT 0, SCB flags 0x0, cmd 0x0
Mar 18 13:44:05 fs1 kernel: (scsi0:-1:-1:-1) CMDCMPLT without command for SCB 1
1, QOUTCNT 0, QINCNT 0, SCB flags 0x0, cmd 0x0
Mar 18 13:44:05 fs1 kernel: (scsi0:-1:-1:-1) CMDCMPLT without command for SCB 1
5, QOUTCNT 0, QINCNT 0, SCB flags 0x0, cmd 0x0
Mar 18 13:44:05 fs1 kernel: (scsi0:-1:-1:-1) CMDCMPLT without command for SCB 1
2, QOUTCNT 0, QINCNT 0, SCB flags 0x0, cmd 0x0
Mar 18 13:44:05 fs1 kernel: (scsi0:-1:-1:-1) CMDCMPLT without command for SCB 1
0, QOUTCNT 0, QINCNT 0, SCB flags 0x0, cmd 0x0                

....

the list goes on like this (at a screen-flickering speed) until I hit 
reset. The box is completely frozen.	

==========================================================================

Mar 19 16:25:40 fs1 kernel: scsi : aborting command due to timeout : pid 48, scs
i1, channel 0, id 8, lun 0 Read (6) 00 00 45 02 00
Mar 19 16:25:40 fs1 kernel: (scsi1:0:8:0) Aborting scb 29, flags 0x4
Mar 19 16:25:40 fs1 kernel: (scsi1:0:8:0) SCB is currently active.  Waiting on c
ompletion.
Mar 19 16:25:42 fs1 kernel: SCSI host 1 abort (pid 48) timed out - resetting
Mar 19 16:25:42 fs1 kernel: SCSI bus is being reset for host 1 channel 0.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:8:0) Reset called, scb 29, flags 0x84
Mar 19 16:25:42 fs1 kernel: (scsi1:0:8:0) Bus device reset stupid when other act
ion has failed.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:-1:-1) Reset channel called, will initiate
reset.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:-1:-1) Resetting currently active channel.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:-1:-1) Channel reset
Mar 19 16:25:42 fs1 kernel: (scsi1:0:-1:-1) Reset device, active_scb 16
Mar 19 16:25:42 fs1 kernel: (scsi1:0:0:-1) Cleaning up status information and de
layed_scbs.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:1:-1) Cleaning up status information and de
layed_scbs.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:2:-1) Cleaning up status information and de
layed_scbs.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:3:-1) Cleaning up status information and de
layed_scbs.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:4:-1) Cleaning up status information and de
layed_scbs.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:5:-1) Cleaning up status information and de
layed_scbs.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:6:-1) Cleaning up status information and de
layed_scbs.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:7:-1) Cleaning up status information and de
layed_scbs.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:8:-1) Cleaning up status information and de
layed_scbs.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:9:-1) Cleaning up status information and de
layed_scbs.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:10:-1) Cleaning up status information and d
elayed_scbs.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:11:-1) Cleaning up status information and d
elayed_scbs.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:12:-1) Cleaning up status information and d
elayed_scbs.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:13:-1) Cleaning up status information and d
elayed_scbs.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:14:-1) Cleaning up status information and d
elayed_scbs.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:15:-1) Cleaning up status information and d
elayed_scbs.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:-1:-1) Cleaning QINFIFO.  
Mar 19 16:25:42 fs1 kernel: (scsi1:0:-1:-1) Cleaning waiting_scbs.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:-1:-1) Cleaning waiting for selection list.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:8:0:tag29) matches search criteria (scsi1:0
:-1:-1:tag255)
Mar 19 16:25:42 fs1 kernel: (scsi1:0:8:0:tag29) matches search criteria (scsi1:0
:-1:-1:tag255)
Mar 19 16:25:42 fs1 kernel: (scsi1:0:8:0) Aborting scb 29
Mar 19 16:25:42 fs1 kernel: (scsi1:-1:-1:-1) 1 commands found and queued for com
pletion.
Mar 19 16:25:45 fs1 kernel: (scsi1:0:8:0) Using wide (16 bit) transfers.
Mar 19 16:25:45 fs1 kernel: (scsi1:0:8:0) Synchronous at 20.0MHz, offset 8. Mar 19 16:25:42 fs1 kernel: (scsi1:0:-1:-1) Cleaning waiting_scbs.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:-1:-1) Cleaning waiting for selection list.
Mar 19 16:25:42 fs1 kernel: (scsi1:0:8:0:tag29) matches search criteria (scsi1:0
:-1:-1:tag255)
Mar 19 16:25:42 fs1 kernel: (scsi1:0:8:0:tag29) matches search criteria (scsi1:0
:-1:-1:tag255)
Mar 19 16:25:42 fs1 kernel: (scsi1:0:8:0) Aborting scb 29
Mar 19 16:25:42 fs1 kernel: (scsi1:-1:-1:-1) 1 commands found and queued for com
pletion.
Mar 19 16:25:45 fs1 kernel: (scsi1:0:8:0) Using wide (16 bit) transfers.
Mar 19 16:25:45 fs1 kernel: (scsi1:0:8:0) Synchronous at 20.0MHz, offset 8. 


... at this point the entire message is repeated again after a few
seconds. hitting reset is the only remedy.

==========================================================================

	Best Regards
	Leif Johansson

Leif Johansson				Phone: +46 8 164541		
Department of Mathematics		Fax  : +46 8 6126717		
Stockholm University 			email: leifj at matematik.su.se 	

    <This space is left blank for quotational and disclamatory purposes.>


To Unsubscribe: send mail to majordomo at FreeBSD.org
with "unsubscribe aic7xxx" in the body of the message