Hi,

I've received second-hand reports about a problem which may
depend on this.  This is on a PPro-200 system with 512MB memory.
Currently our main problem is:

# cksum foo
3674204387 4354408 foo
# cp foo bar
# cksum bar
2265597358 4354408 bar
#

On bootup it says:

cpu0: family 6 model 1 step 9
cpu0: Intel Pentium Pro (686-class)
real mem  =3D 536477696
avail mem =3D 499548160
using 1388 buffers containing 26927104 bytes of memory

I don't think NBUF or BUFPAGES have been tweaked, so this is what
NetBSD defaults to in this configuration.  Is the default
"reasonable", or does it need to be tweaked?

First, if my calculations aren't way off this makes for on the
average 19399 bytes per buffer, which seems a little odd.

Second, I'm not sure I understand the implication of this
statement:

>Good luck. You can't have more than 1388 buffers on the i386 at
>the moment, because it runs out of address space for them: each
>buffer uses 64K of address space. (Which means that on most
>systems--with 8K blocks--there's not much point in having more than
>11 MB of buffer memory allocated.)

What, if anything, needs to be tweaked on the above machine, or
is the only problem that I'm wasting about 15MB of memory?  (1388
buffers seems to consume approx. 90MB of KVA space, which is
"close enough" to the 128MB KVA limit, right?)

We're a little suspicious of the memory installed in this machine
which is 8 64MB SIMMs, it currently only has partity checking
enabled, and reports no problems when the above test is run.

The I/O configuration looks like this:

bha0 at pci0 dev 13 function 0: BusLogic 9xxC SCSI
bha0: interrupting at irq 15
bha0: model BT-958, firmware 5.06I
bha0: sync, parity
bha0 targ 0: sync, offset 15, period 100nsec
bha0 targ 5: sync, offset 8, period 100nsec
bha0 targ 6: sync, offset 15, period 100nsec
scsibus0 at bha0: 16 targets
sd0 at scsibus0 targ 0 lun 0: <SEAGATE, ST34371W, 0484> SCSI2 0/direct fixed
sd0: 4148MB, 5172 cyl, 10 head, 164 sec, 512 bytes/sec x 8496884 sectors
cd0 at scsibus0 targ 5 lun 0: <TOSHIBA, CD-ROM XM-5701TA, 0167> SCSI2 5/cdrom removable
st0 at scsibus0 targ 6 lun 0: <HP, C1533A, A612> SCSI2 1/sequential removable
st0: drive empty

Shouldn't be anything weird here, right?

BTW, the kernel doesn't spit out funny messages when the
copy/compare steps are done.

Any hints wrt. what we should be looking for?


- H=E5vard