On Wed, May 07, 2003 at 11:11:17PM -0700, Keith M Wesolowski wrote: > > Following patch builds whole RAM map based of MC's memory configuration > > registers, does some samity checks adds high system memory (if any) to > > bootmem. > > > +static void init_bootmem(void) > ... > > + init_bootmem(); > > This is a pretty unfortunate choice of names for this function. See > mm/bootmem.c. > > Other than that, your patch works fine for me; my Indy has 192MB > memory and it's detected properly. I do get an oops in do_be from > xdm, but I get that without the patch also. > > Determined physical RAM map: > memory: 00001000 @ 00000000 (reserved) > memory: 00001000 @ 00001000 (reserved) > memory: 001e1000 @ 08002000 (reserved) > memory: 0055d000 @ 081e3000 (usable) > memory: 000c0000 @ 08740000 (ROM data) > memory: 0b800000 @ 08800000 (usable) > > I need to do the same kind of thing for ip32 as the ARC memory > detection has the same shortcoming on that platform. No sense having > a machine support 1GB memory and only looking for 256MB of it, > especially in a 64-bit kernel. ARC[S] really does seem to be useless. That's what I'm saying since '94. ARC was a commitee approach of the ACE consortium which soon died. The firmware part of the ARC standard was also published as the Microsoft Portable Bootloader Standard but Today every ARC implementation has some major deviations from the standard rendering the term standard into nothing but a cynic demonation ... Ralf
