FreeBSD 6.0 has problems with Symbios/LSI-HBAs

[Scott Long]

Juergen Dankoweit wrote:
> Hello to the list.
> 
> On my production server (private and at firm) I had the possiblity to
> test the new FreeBSD 6.0 RC1 and REALESE in the downtime.
> 
> Both systems are Siemens Primergy 470, dual PIII 450 MHz, two LSI 83C895
> SCSI-HBAs (one onboard and one on a separate PCI card).
> At the onboard HBA all HDDs are LVD drives.
> At the second only single ended devices are connected.
> The termination is correct.
> 
> On both systems it is imposisible to install FreeBSD 6 because it hangs
> at this position:
> 
> (noperiph:sym0:0:-1:-1): SCSI BUS reset delivered.
> (noperiph:sym1:0:-1:-1): SCSI BUS reset delivered.
> (noperiph:sym1:0:-1:-1): SCSI BUS mode change from SE to SE.
> 
> here stops the system.
> After about five minutes the following messages are shown character for
> character.
> (When the SAF/TE is connected)
> (probe7:sym0:0:8:0): phase change 6-7 6 at 0f76058c resid=4.
> ses0 at sym0 bus 0 target 8 lun 0
> ses0: <SIEMENS STM/L S1 4.1b> Fixed Processor SCSI-2 device
> ses0: 3.300MB/s transfers
> ses0: SAF-TE Compliant Device
> 
> (When the SAF/TE is NOT connected it jumps directly to these messages)
> da0 at sym0 bus 0 target 0 lun 0
> da0: <COMPAQ BD036735C8 B020> Fixed Direct Access SCSI-2 device
> da0: 80.000MB/s transfers (40.000MHz, offset 31, 16bit), Tagged Queueing
> Enabled
> da0: 34732MB (71132000 512 byte sectors: 255H 63S/T 4427C)
> 
> And here nothing goes on. 
> 
> Exactly this behavior was on FreeBSD 5.3 STABLE-1. It was fixed with
> FBSD 5.3 STABLE-2. FreeBSD 5.4 runs also perfect on these machines.
> But in the new version of FreeBSD the same bug comes again. 
> 
> On the production machines it is impossible to install or upgrade the
> new FreeBSD 6 to do more tests.
> 
> Please send an email if more information is needed, but let me say
> again: these machines are production servers. I don't have the
> possibility to compile any kernels or change adapters.
> 
> Many thanks for your help.
> 
> Best regards
> 
> Jürgen
> 

Thanks for the report.  The fact that it existed in 5.3 then was fixed 
might point to some clues.  I'll put it on my TODO list.

Scott