Aren
/
NetBSD
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Add "kvtophys()", which maps MIPS R2000 kernel-virtual addresses to physical addresses, 

so they can (e.g.) be written to DMA mapping registers.
This commit is contained in:
jonathan 1995-09-25 20:36:23 +00:00
parent b47de9fb69
commit cd887a9930
2 changed files with 134 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

68
sys/arch/mips/mips/vm_machdep.c
View File

@ -1,4 +1,4 @@
/* $NetBSD: vm_machdep.c,v 1.6 1995/01/18 06:52:46 mellon Exp $ */
/* $NetBSD: vm_machdep.c,v 1.7 1995/09/25 20:36:23 jonathan Exp $ */
/*
* Copyright (c) 1988 University of Utah.
@ -57,6 +57,8 @@
#include <vm/vm_page.h>
#include <machine/pte.h>
#include <machine/vmparam.h>
#include <machine/machConst.h>
/*
* Finish a fork operation, with process p2 nearly set up.
@ -318,3 +320,67 @@ vunmapbuf(bp)
bp->b_un.b_addr = bp->b_saveaddr;
bp->b_saveaddr = NULL;
}
/*XXX*/
/*
* Map a (kernel) virtual address to a physical address.
* There are four cases:
* A kseg0 kernel "virtual address" for the cached physical address space;
* A kseg1 kernel "virtual address" for the uncached physical address space;
* A kseg2 normal kernel "virtual address" for the kernel stack or
* "u area". These ARE NOT necessarily in sysmap, since processes 0
* and 1 are handcrafted before the sysmap is set up.
* A kseg2 normal kernel "virtual address" mapped via the TLB, which
* IS NOT in Sysmap (eg., an mbuf).
* The first two are so cheap they could just be macros. The last two
* overlap, so we must check for UADDR pages first.
*
* XXX the u-area mappng should all change anyway.
*/
vm_offset_t
kvtophys(vm_offset_t kva)
{
pt_entry_t *pte;
vm_offset_t phys;
if (kva >= MACH_CACHED_MEMORY_ADDR && kva < MACH_UNCACHED_MEMORY_ADDR)
{
return (MACH_CACHED_TO_PHYS(kva));
}
else if (kva >= MACH_UNCACHED_MEMORY_ADDR && kva < MACH_KSEG2_ADDR) {
return (MACH_UNCACHED_TO_PHYS(kva));
}
else if (kva >= UADDR && kva < KERNELSTACK) {
int upage = (kva - UADDR) >> PGSHIFT;
pte = (pt_entry_t *)&curproc->p_md.md_upte[upage];
phys = (pte->pt_entry & PG_FRAME) |
(kva & PGOFSET);
}
else if (kva >= MACH_KSEG2_ADDR /*&& kva < VM_MAX_KERNEL_ADDRESS*/) {
pte = kvtopte(kva);
if ((pte - Sysmap) > Sysmapsize) {
printf("oops: Sysmap overrun, max %d index %d\n",
Sysmapsize, pte - Sysmap);
}
kernelmapped:
if ((pte->pt_entry & PG_V) == 0) {
printf("kvtophys: pte not valid for %x\n", kva);
}
phys = (pte->pt_entry & PG_FRAME) |
(kva & PGOFSET);
#ifdef DEBUG_VIRTUAL_TO_PHYSICAL
printf("kvtophys: kv %x, phys %x", kva, phys);
#endif
}
else {
printf("Virtual address %x: cannot map to physical\n",
kva);
phys = 0;
/*panic("non-kernel address to kvtophys\n");*/
return(kva); /* XXX -- while debugging ASC */
}
return(phys);
}

68
sys/arch/pmax/pmax/vm_machdep.c
View File

@ -1,4 +1,4 @@
/* $NetBSD: vm_machdep.c,v 1.6 1995/01/18 06:52:46 mellon Exp $ */
/* $NetBSD: vm_machdep.c,v 1.7 1995/09/25 20:36:23 jonathan Exp $ */
/*
* Copyright (c) 1988 University of Utah.
@ -57,6 +57,8 @@
#include <vm/vm_page.h>
#include <machine/pte.h>
#include <machine/vmparam.h>
#include <machine/machConst.h>
/*
* Finish a fork operation, with process p2 nearly set up.
@ -318,3 +320,67 @@ vunmapbuf(bp)
bp->b_un.b_addr = bp->b_saveaddr;
bp->b_saveaddr = NULL;
}
/*XXX*/
/*
* Map a (kernel) virtual address to a physical address.
* There are four cases:
* A kseg0 kernel "virtual address" for the cached physical address space;
* A kseg1 kernel "virtual address" for the uncached physical address space;
* A kseg2 normal kernel "virtual address" for the kernel stack or
* "u area". These ARE NOT necessarily in sysmap, since processes 0
* and 1 are handcrafted before the sysmap is set up.
* A kseg2 normal kernel "virtual address" mapped via the TLB, which
* IS NOT in Sysmap (eg., an mbuf).
* The first two are so cheap they could just be macros. The last two
* overlap, so we must check for UADDR pages first.
*
* XXX the u-area mappng should all change anyway.
*/
vm_offset_t
kvtophys(vm_offset_t kva)
{
pt_entry_t *pte;
vm_offset_t phys;
if (kva >= MACH_CACHED_MEMORY_ADDR && kva < MACH_UNCACHED_MEMORY_ADDR)
{
return (MACH_CACHED_TO_PHYS(kva));
}
else if (kva >= MACH_UNCACHED_MEMORY_ADDR && kva < MACH_KSEG2_ADDR) {
return (MACH_UNCACHED_TO_PHYS(kva));
}
else if (kva >= UADDR && kva < KERNELSTACK) {
int upage = (kva - UADDR) >> PGSHIFT;
pte = (pt_entry_t *)&curproc->p_md.md_upte[upage];
phys = (pte->pt_entry & PG_FRAME) |
(kva & PGOFSET);
}
else if (kva >= MACH_KSEG2_ADDR /*&& kva < VM_MAX_KERNEL_ADDRESS*/) {
pte = kvtopte(kva);
if ((pte - Sysmap) > Sysmapsize) {
printf("oops: Sysmap overrun, max %d index %d\n",
Sysmapsize, pte - Sysmap);
}
kernelmapped:
if ((pte->pt_entry & PG_V) == 0) {
printf("kvtophys: pte not valid for %x\n", kva);
}
phys = (pte->pt_entry & PG_FRAME) |
(kva & PGOFSET);
#ifdef DEBUG_VIRTUAL_TO_PHYSICAL
printf("kvtophys: kv %x, phys %x", kva, phys);
#endif
}
else {
printf("Virtual address %x: cannot map to physical\n",
kva);
phys = 0;
/*panic("non-kernel address to kvtophys\n");*/
return(kva); /* XXX -- while debugging ASC */
}
return(phys);
}