Re: CVS commit: syssrc/sys/arch/arc/conf
Noriyuki Soda <[email protected]> Wed, 4 Dec 2002 19:02:02 +0900 (JST)
| Newsgroups | gmane.os.netbsd.ports.arc |
|---|---|
| Message-ID | <[email protected]> |
>>>>> On Wed, 04 Dec 2002 01:49:18 -0700 (MST), "M. Warner Losh" <[email protected]> said: > However, there are two problems. First problem is that the magic > number should be 0407 not 0413 in order for the BIOS' run command to > be able to run the binary. Perhaps using usr.bin/elf2ecoff (or, even old OpenBSD elf2ecoff source) may be safer way to load the kernel to rPC44 in this point, although I'm not really sure. > However, shortly after we load the OS, we get a panic accessing > either address 0x00000004 or addres 0x00a000a4[sic]. I'd have > expected it to be 0x000a0000 or so since that's where the video RAM > is (well, we really should be accessing it via 0xa00a0000 because it > should be treated, I think, as uncached physical memory). Hmmm, strange. pccons for rPC44 is intialized via following path: kern/init_main.c:main() -> arch/arc/arc/machdep.c:consinit() -> arch/arc/arc/c_isa.c:c_isa_cons_init() via (*platform->cons_init)() -> arch/arc/isa/pccons_isa.c:pccons_isa_cnattach() -> arch/arc/dev/pccons.c:pccons_common_cnattach() In this point, arc_bus_io and arc_bus_mem, which are initialized in arch/arc/arc/p_dti_arcstation.c:p_dti_arcstation_init(), should be passed to pccons_common_cnattach(), so that pccons must access framebuffer via arc_bus_mem with RPC44_V_ISA_MEM offset (== 0xa0000000). But perhaps I made some mistakes in this rPC44 code path. Anyway, thanks for the work! -- soda