Richard van den Berg wrote:
>
> On Thu, 25 May 2000, Harald Koerfgen wrote:
>
> > On 22-Jul-99 Ralf Baechle wrote:
> > > CVSROOT: /src/ftp/cvs
> > > Module name: linux
> > > Changes by: ralf@linus.linux.sgi.com 99/07/21 18:57:14
> > >
> > > Modified files:
> > > include/asm-mips: Tag: linux_2_2 pgtable.h
> > >
> > > Log message:
> > > Away you demons of memory corruption. Burn, bug burn ...
>
> But they returned:
>
> On Tue, 15 Feb 2000, Harald Koerfgen wrote:
>
> > On 15-Feb-00 Richard van den Berg wrote:
> > > On Sun, 13 Feb 2000, Harald Koerfgen wrote:
> > >
> > >> Log message:
> > >> o bootmem fixes for DECstations
> > >> o declarations in pgtable.h fixed
> > >> o cache management for R3000 improved
> > >
> > > [ravdberg@dolce ravdberg]$ uname -a
> > > Linux dolce 2.3.38 #5 Mon Feb 14 19:13:31 CET 2000 mips unknown
> > > [ravdberg@dolce ravdberg]$ cat /proc/cpuinfo
> > > cpu : MIPS
> > > cpu model : R3000A V3.0
> > > system type : Digital Personal DECstation 5000/xx
> > > BogoMIPS : 24.84
> > > byteorder : little endian
> > > unaligned accesses : 0
> > > wait instruction : no
> > > microsecond timers : no
> > > extra interrupt vector : no
> > > hardware watchpoint : no
> > > VCED exceptions : not available
> > > VCEI exceptions : not available
> >
> > Never mind, but, beware, swapping is broken for 2.3.38 with R3000.
> >
> > While talking about the subject, I've put some assumptions into the
> > optimisation which seem to be valid for R3000As, but I am not exactly shure
> > if
> > they are true for other R2000/R3000 variants. Thus, I'd like to have some
> > feedback from other boxes, for example DS2100/3100 or Bagets.
>
> On 2.3.99-pre8 swapping is still broken... 2.3.21 is the latest version
> with swapping working on R3000 DECstation I know of.
>
> Regards,
> Richard
Looks like bug in swapping was swapped in again ;-)
P.S.
I just applied 3-lines long patch posted by Ralf (SWP_xxx defines from
pgtable.h)
to our 2.2.1 tree.
>From my current point of view it makes swapping Baget working like a lion:
no any program crash occured during one hour of very intensive testing by
few c++ compillation session working in parallel (Baget has only 16Mb RAM).
Any comments and tests are welcoming !
Regards,
Vladimir.
|