Subject: Re: aic7899 support ?
To: None <current-users@netbsd.org>
From: Greg A. Woods <woods@weird.com>
List: current-users
Date: 05/11/2001 15:44:30
[ On Friday, May 11, 2001 at 20:40:48 (+0200), Frank van der Linden wrote: ]
> Subject: Re: aic7899 support ?
>
> On Fri, May 11, 2001 at 08:32:12PM +0200, Manuel Bouyer wrote:
> > does someone run NetBSD with a aic7899 SCSI adapter (that is,
> > "Adaptec 3960D Ultra160 SCSI adapter" or "Adaptec (Compaq OEM) 3960D Ultra160
> > SCSI adapter") ?
> > Someone gets a uvm_fault in the driver at init time with such an adapter,
> > and I don't see why ...
> 
> I have a 7899 in one of my machines:

That's what's on the Intel STL2 motherboard on which I installed a 1.5T
kernel and system on yesterday (soon to run Squid along with an
identical twin of itself).

ahc0 at pci1 dev 4 function 0
OptionMode = 3
ahc0: interrupting at irq 9
ahc0: aic7899 Wide Channel A, SCSI Id=7, 16/255 SCBs
scsibus0 at ahc0 channel 0: 16 targets, 8 luns per target
ahc1 at pci1 dev 4 function 1
OptionMode = 3
ahc1: interrupting at irq 11
ahc1: aic7899 Wide Channel B, SCSI Id=7, 16/255 SCBs
scsibus1 at ahc1 channel 0: 16 targets, 8 luns per target

> It's been working fine so far

Yes, so far, so good!

> (not in 160 mode, since the ahc driver
> needs an update for that).

I was wondering about that....

scsibus0: waiting 2 seconds for devices to settle...
ahc0: target 0 using 16bit transfers
ahc0: target 0 synchronous at 40.0MHz, offset = 0x3f
ahc0: target 0 using tagged queuing
sd0 at scsibus0 target 0 lun 0: <SEAGATE, ST39205LW, 5063> SCSI3 0/direct fixed
sd0: 8750 MB, 19036 cyl, 2 head, 470 sec, 512 bytes/sect x 17921835 sectors
ahc0: target 1 using 16bit transfers
ahc0: target 1 synchronous at 40.0MHz, offset = 0x3f
ahc0: target 1 using tagged queuing
sd1 at scsibus0 target 1 lun 0: <SEAGATE, ST39205LW, 5063> SCSI3 0/direct fixed
sd1: 8750 MB, 19036 cyl, 2 head, 470 sec, 512 bytes/sect x 17921835 sectors
ahc0: target 2 using 16bit transfers
ahc0: target 2 synchronous at 40.0MHz, offset = 0x3f
ahc0: target 2 using tagged queuing
sd2 at scsibus0 target 2 lun 0: <SEAGATE, ST39205LW, 5063> SCSI3 0/direct fixed
sd2: 8750 MB, 19036 cyl, 2 head, 470 sec, 512 bytes/sect x 17921835 sectors
ahc0: target 3 using 16bit transfers
ahc0: target 3 synchronous at 40.0MHz, offset = 0x3f
ahc0: target 3 using tagged queuing
sd3 at scsibus0 target 3 lun 0: <SEAGATE, ST39205LW, 5063> SCSI3 0/direct fixed
sd3: 8750 MB, 19036 cyl, 2 head, 470 sec, 512 bytes/sect x 17921835 sectors

> I've been pounding on the disk fairly hard,
> of course, with only one device attached, that's not really a
> stresstest.

from a single bonnie run on a bsize=16384/fsize=2048 filesystem (there
512MB RAM in the system too, BTW):

              -------Sequential Output-------- ---Sequential Input-- --Random--
              -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
STL2     2000 36139 88.5 36386 32.6  5169  4.1 30893 91.3 36815 15.7 170.9  1.5

and from four simultaneous bonnie runs, one on each disk:

              -------Sequential Output-------- ---Sequential Input-- --Random--
              -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
         1000  9948 23.9 19403 18.1  4846  4.9  7817 23.9 18956  9.9 149.1  1.8
         1000  9861 23.9 20129 19.2  4666  5.0  7682 23.7 18231 10.3 165.8  1.9
         1000  9718 23.4 19799 18.7  4851  5.2  7704 23.7 18602 10.0 154.5  1.9
         1000  9645 23.4 19729 18.7  4801  5.4  7656 23.8 18696 10.3 154.4  1.8


The CPU is a P-III 733MHz, which does seem to feel the load of the ahc
driver.  According to 'systat vmstat' the %sys load was a significant
part of the above %CPU stats.

The "squid -z" run, with standard 16x256 levels, completed the build of
over 4000 directories on each disk, simultaneously, in less than 10
seconds, and that's without softdep.  I should probably do it again just
to get an accurate timing!  ;-)

(I'm not sure whether to try using softdep for the cache, or not....)


-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods@acm.org>     <woods@robohack.ca>
Planix, Inc. <woods@planix.com>;   Secrets of the Weird <woods@weird.com>