In netbsd-bugs you write:
> MAXBSIZE was #defined to be MAXPHYS in 4.4BSD-Lite (and Lite2).
> It was changed to be 16k in NetBSD. (If my memory serves me,
> the reason given at the time was to save kernel page table space
> on the i386.
> This drastically reduces the effectiveness of clustering (esp.
> write clustering), because the number of I/Os which can be combined
> is reduced. (Coalsced I/Os must still fit into a buffer with
> maximum size of MAXBSIZE.)
Wait a moment, please.
Setting MAXBSIZE=MAXPHYS=64k on my i386/sup19950927 panics at boot
time.
Setting MAXBSIZE to more that 16 K and up to 57K works and gives
the following bonnie results. The machine is P90, triton, NCR 810,
barracouda 2 GB.
-------Sequential Output-------- ---Sequential Input----Random--
-Per Char- --Block--- -Rewrite-- -Per Char- --Block-----Seeks---
MAXBSIZE=16K (default)
140 1427 99.5 2806 19.6 793 10.2 1625 99.5 4573 32.9 83.6 7.0
140 1432 99.4 2801 18.4 793 9.9 1636 99.7 4589 32.1 84.0 6.8
MAXBSIZE=32K
140 1453 99.4 4276 26.8 1259 13.2 1659 99.7 4593 27.2 84.1 6.8
MAXBSIZE=48K
140 1466 99.5 4215 25.6 1746 17.1 1673 99.7 4592 25.2 83.0 6.7
MAXBSIZE=56K
140 1463 99.5 4209 25.5 1504 14.5 1679 99.5 4592 24.7 84.5 6.9
MAXBSIZE=64K
doesn't boot
FreeBSD-951005-SNAP (no flames please :-)
140 1383 98.8 4557 42.6 1642 18.3 1646 98.3 4595 27.7 92.5 7.9
[ note that FreeBSD needs more CPU cycles ]
Some backtrace points of the panic
_pmap_enter
_vm_fault
...
_kmem_alloc
...
_vm_fork
...
_main
Maybe this is related:
./arch/i386/isa/fdreg.h:
#define FDC_MAXIOSIZE NBPG /* XXX should be MAXBSIZE */
The comment doesn't make sense. NBPG is 4k, not 16k. Setting
#define FDC_MAXIOSIZE MAXBSIZE
doesn't make a working kernel (doesn't find boot disk anymore).
Martin
--
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Martin Cracauer <cracauer@wavehh.hanse.de> - Fax +49 40 522 85 36
BSD User Group Hamburg, Germany - See http://{net,free}bsd.www.org
Copyright 1995. Redistribution via Microsoft Network is prohibited