3 * Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
5 * Modifications by Paul Mackerras (PowerMac) (paulus@cs.anu.edu.au)
6 * and Cort Dougan (PReP) (cort@cs.nmt.edu)
7 * Copyright (C) 1996 Paul Mackerras
8 * Amiga/APUS changes by Jesper Skov (jskov@cygnus.co.uk).
9 * PPC44x/36-bit changes by Matt Porter (mporter@mvista.com)
11 * Derived from "arch/i386/mm/init.c"
12 * Copyright (C) 1991, 1992, 1993, 1994 Linus Torvalds
14 * This program is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU General Public License
16 * as published by the Free Software Foundation; either version
17 * 2 of the License, or (at your option) any later version.
21 #include <linux/config.h>
22 #include <linux/module.h>
23 #include <linux/sched.h>
24 #include <linux/kernel.h>
25 #include <linux/errno.h>
26 #include <linux/string.h>
27 #include <linux/types.h>
29 #include <linux/stddef.h>
30 #include <linux/init.h>
31 #include <linux/bootmem.h>
32 #include <linux/highmem.h>
33 #include <linux/initrd.h>
35 #include <asm/pgalloc.h>
38 #include <asm/mmu_context.h>
39 #include <asm/pgtable.h>
42 #include <asm/machdep.h>
43 #include <asm/btext.h>
45 #include <asm/bootinfo.h>
47 #include "mem_pieces.h"
50 #if defined(CONFIG_KERNEL_START_BOOL) || defined(CONFIG_LOWMEM_SIZE_BOOL)
51 /* The ammount of lowmem must be within 0xF0000000 - KERNELBASE. */
52 #if (CONFIG_LOWMEM_SIZE > (0xF0000000 - KERNELBASE))
53 #error "You must adjust CONFIG_LOWMEM_SIZE or CONFIG_START_KERNEL"
56 #define MAX_LOW_MEM CONFIG_LOWMEM_SIZE
58 DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
60 unsigned long total_memory;
61 unsigned long total_lowmem;
63 unsigned long ppc_memstart;
64 unsigned long ppc_memoffset = PAGE_OFFSET;
67 int init_bootmem_done;
69 #ifdef CONFIG_PPC_PMAC
70 unsigned long agp_special_page;
74 extern char etext[], _stext[];
75 extern char __init_begin, __init_end;
76 extern char __prep_begin, __prep_end;
77 extern char __chrp_begin, __chrp_end;
78 extern char __pmac_begin, __pmac_end;
79 extern char __openfirmware_begin, __openfirmware_end;
85 EXPORT_SYMBOL(kmap_prot);
86 EXPORT_SYMBOL(kmap_pte);
90 void set_phys_avail(unsigned long total_ram);
92 /* XXX should be in current.h -- paulus */
93 extern struct task_struct *current_set[NR_CPUS];
96 struct mem_pieces phys_avail;
99 extern unsigned long sysmap_size;
102 * this tells the system to map all of ram with the segregs
103 * (i.e. page tables) instead of the bats.
106 int __map_without_bats;
108 /* max amount of RAM to use */
109 unsigned long __max_memory;
110 /* max amount of low RAM to map in */
111 unsigned long __max_low_memory = MAX_LOW_MEM;
115 int i,free = 0,total = 0,reserved = 0;
116 int shared = 0, cached = 0;
119 printk("Mem-info:\n");
121 printk("Free swap: %6dkB\n",nr_swap_pages<<(PAGE_SHIFT-10));
125 if (PageHighMem(mem_map+i))
127 if (PageReserved(mem_map+i))
129 else if (PageSwapCache(mem_map+i))
131 else if (!page_count(mem_map+i))
134 shared += page_count(mem_map+i) - 1;
136 printk("%d pages of RAM\n",total);
137 printk("%d pages of HIGHMEM\n", highmem);
138 printk("%d free pages\n",free);
139 printk("%d reserved pages\n",reserved);
140 printk("%d pages shared\n",shared);
141 printk("%d pages swap cached\n",cached);
144 /* Free up now-unused memory */
145 static void free_sec(unsigned long start, unsigned long end, const char *name)
147 unsigned long cnt = 0;
149 while (start < end) {
150 ClearPageReserved(virt_to_page(start));
151 set_page_count(virt_to_page(start), 1);
157 printk(" %ldk %s", cnt << (PAGE_SHIFT - 10), name);
158 totalram_pages += cnt;
162 void free_initmem(void)
164 #define FREESEC(TYPE) \
165 free_sec((unsigned long)(&__ ## TYPE ## _begin), \
166 (unsigned long)(&__ ## TYPE ## _end), \
169 printk ("Freeing unused kernel memory:");
171 if (_machine != _MACH_Pmac)
173 if (_machine != _MACH_chrp)
175 if (_machine != _MACH_prep)
178 FREESEC(openfirmware);
183 #ifdef CONFIG_BLK_DEV_INITRD
184 void free_initrd_mem(unsigned long start, unsigned long end)
186 printk ("Freeing initrd memory: %ldk freed\n", (end - start) >> 10);
188 for (; start < end; start += PAGE_SIZE) {
189 ClearPageReserved(virt_to_page(start));
190 set_page_count(virt_to_page(start), 1);
198 * Check for command-line options that affect what MMU_init will do.
202 /* Check for nobats option (used in mapin_ram). */
203 if (strstr(cmd_line, "nobats")) {
204 __map_without_bats = 1;
207 /* Look for mem= option on command line */
208 if (strstr(cmd_line, "mem=")) {
210 unsigned long maxmem = 0;
212 for (q = cmd_line; (p = strstr(q, "mem=")) != 0; ) {
214 if (p > cmd_line && p[-1] != ' ')
216 maxmem = simple_strtoul(q, &q, 0);
217 if (*q == 'k' || *q == 'K') {
220 } else if (*q == 'm' || *q == 'M') {
225 __max_memory = maxmem;
230 * MMU_init sets up the basic memory mappings for the kernel,
231 * including both RAM and possibly some I/O regions,
232 * and sets up the page tables and the MMU hardware ready to go.
234 void __init MMU_init(void)
237 ppc_md.progress("MMU:enter", 0x111);
239 /* parse args from command line */
243 * Figure out how much memory we have, how much
244 * is lowmem, and how much is highmem. If we were
245 * passed the total memory size from the bootloader,
249 total_memory = boot_mem_size;
251 total_memory = ppc_md.find_end_of_memory();
253 if (__max_memory && total_memory > __max_memory)
254 total_memory = __max_memory;
255 total_lowmem = total_memory;
256 if (total_lowmem > __max_low_memory) {
257 total_lowmem = __max_low_memory;
258 #ifndef CONFIG_HIGHMEM
259 total_memory = total_lowmem;
260 #endif /* CONFIG_HIGHMEM */
262 set_phys_avail(total_lowmem);
264 /* Initialize the MMU hardware */
266 ppc_md.progress("MMU:hw init", 0x300);
269 /* Map in all of RAM starting at KERNELBASE */
271 ppc_md.progress("MMU:mapin", 0x301);
274 #ifdef CONFIG_HIGHMEM
275 ioremap_base = PKMAP_BASE;
277 ioremap_base = 0xfe000000UL; /* for now, could be 0xfffff000 */
278 #endif /* CONFIG_HIGHMEM */
279 ioremap_bot = ioremap_base;
281 /* Map in I/O resources */
283 ppc_md.progress("MMU:setio", 0x302);
284 if (ppc_md.setup_io_mappings)
285 ppc_md.setup_io_mappings();
287 /* Initialize the context management stuff */
291 ppc_md.progress("MMU:exit", 0x211);
293 #ifdef CONFIG_BOOTX_TEXT
294 /* By default, we are no longer mapped */
295 boot_text_mapped = 0;
296 /* Must be done last, or ppc_md.progress will die. */
301 /* This is only called until mem_init is done. */
302 void __init *early_get_page(void)
306 if (init_bootmem_done) {
307 p = alloc_bootmem_pages(PAGE_SIZE);
309 p = mem_pieces_find(PAGE_SIZE, PAGE_SIZE);
315 * Initialize the bootmem system and give it all the memory we
318 void __init do_init_bootmem(void)
320 unsigned long start, size;
324 * Find an area to use for the bootmem bitmap.
325 * We look for the first area which is at least
326 * 128kB in length (128kB is enough for a bitmap
327 * for 4GB of memory, using 4kB pages), plus 1 page
328 * (in case the address isn't page-aligned).
332 for (i = 0; i < phys_avail.n_regions; ++i) {
333 unsigned long a = phys_avail.regions[i].address;
334 unsigned long s = phys_avail.regions[i].size;
339 if (s >= 33 * PAGE_SIZE)
342 start = PAGE_ALIGN(start);
344 min_low_pfn = start >> PAGE_SHIFT;
345 max_low_pfn = (PPC_MEMSTART + total_lowmem) >> PAGE_SHIFT;
346 max_pfn = (PPC_MEMSTART + total_memory) >> PAGE_SHIFT;
347 boot_mapsize = init_bootmem_node(&contig_page_data, min_low_pfn,
348 PPC_MEMSTART >> PAGE_SHIFT,
351 /* remove the bootmem bitmap from the available memory */
352 mem_pieces_remove(&phys_avail, start, boot_mapsize, 1);
354 /* add everything in phys_avail into the bootmem map */
355 for (i = 0; i < phys_avail.n_regions; ++i)
356 free_bootmem(phys_avail.regions[i].address,
357 phys_avail.regions[i].size);
359 init_bootmem_done = 1;
363 * paging_init() sets up the page tables - in fact we've already done this.
365 void __init paging_init(void)
367 unsigned long zones_size[MAX_NR_ZONES], i;
369 #ifdef CONFIG_HIGHMEM
370 map_page(PKMAP_BASE, 0, 0); /* XXX gross */
371 pkmap_page_table = pte_offset_kernel(pmd_offset(pgd_offset_k
372 (PKMAP_BASE), PKMAP_BASE), PKMAP_BASE);
373 map_page(KMAP_FIX_BEGIN, 0, 0); /* XXX gross */
374 kmap_pte = pte_offset_kernel(pmd_offset(pgd_offset_k
375 (KMAP_FIX_BEGIN), KMAP_FIX_BEGIN), KMAP_FIX_BEGIN);
376 kmap_prot = PAGE_KERNEL;
377 #endif /* CONFIG_HIGHMEM */
380 * All pages are DMA-able so we put them all in the DMA zone.
382 zones_size[ZONE_DMA] = total_lowmem >> PAGE_SHIFT;
383 for (i = 1; i < MAX_NR_ZONES; i++)
386 #ifdef CONFIG_HIGHMEM
387 zones_size[ZONE_HIGHMEM] = (total_memory - total_lowmem) >> PAGE_SHIFT;
388 #endif /* CONFIG_HIGHMEM */
390 free_area_init(zones_size);
393 void __init mem_init(void)
399 #ifdef CONFIG_HIGHMEM
400 unsigned long highmem_mapnr;
402 highmem_mapnr = total_lowmem >> PAGE_SHIFT;
403 highmem_start_page = mem_map + highmem_mapnr;
404 #endif /* CONFIG_HIGHMEM */
405 max_mapnr = total_memory >> PAGE_SHIFT;
407 high_memory = (void *) __va(PPC_MEMSTART + total_lowmem);
408 num_physpages = max_mapnr; /* RAM is assumed contiguous */
410 totalram_pages += free_all_bootmem();
412 #ifdef CONFIG_BLK_DEV_INITRD
413 /* if we are booted from BootX with an initial ramdisk,
414 make sure the ramdisk pages aren't reserved. */
416 for (addr = initrd_start; addr < initrd_end; addr += PAGE_SIZE)
417 ClearPageReserved(virt_to_page(addr));
419 #endif /* CONFIG_BLK_DEV_INITRD */
422 /* mark the RTAS pages as reserved */
424 for (addr = (ulong)__va(rtas_data);
425 addr < PAGE_ALIGN((ulong)__va(rtas_data)+rtas_size) ;
427 SetPageReserved(virt_to_page(addr));
429 #ifdef CONFIG_PPC_PMAC
430 if (agp_special_page)
431 SetPageReserved(virt_to_page(agp_special_page));
434 for (addr = (unsigned long)sysmap;
435 addr < PAGE_ALIGN((unsigned long)sysmap+sysmap_size) ;
437 SetPageReserved(virt_to_page(addr));
439 for (addr = PAGE_OFFSET; addr < (unsigned long)high_memory;
441 if (!PageReserved(virt_to_page(addr)))
443 if (addr < (ulong) etext)
445 else if (addr >= (unsigned long)&__init_begin
446 && addr < (unsigned long)&__init_end)
448 else if (addr < (ulong) klimit)
452 #ifdef CONFIG_HIGHMEM
456 for (pfn = highmem_mapnr; pfn < max_mapnr; ++pfn) {
457 struct page *page = mem_map + pfn;
459 ClearPageReserved(page);
460 set_bit(PG_highmem, &page->flags);
461 set_page_count(page, 1);
465 totalram_pages += totalhigh_pages;
467 #endif /* CONFIG_HIGHMEM */
469 printk("Memory: %luk available (%dk kernel code, %dk data, %dk init, %ldk highmem)\n",
470 (unsigned long)nr_free_pages()<< (PAGE_SHIFT-10),
471 codepages<< (PAGE_SHIFT-10), datapages<< (PAGE_SHIFT-10),
472 initpages<< (PAGE_SHIFT-10),
473 (unsigned long) (totalhigh_pages << (PAGE_SHIFT-10)));
475 printk("System.map loaded at 0x%08x for debugger, size: %ld bytes\n",
476 (unsigned int)sysmap, sysmap_size);
477 #ifdef CONFIG_PPC_PMAC
478 if (agp_special_page)
479 printk(KERN_INFO "AGP special page: 0x%08lx\n", agp_special_page);
482 /* Make sure all our pagetable pages have page->mapping
483 and page->index set correctly. */
484 for (addr = KERNELBASE; addr != 0; addr += PGDIR_SIZE) {
486 pmd_t *pmd = pmd_offset(pgd_offset_k(addr), addr);
487 if (pmd_present(*pmd)) {
489 pg->mapping = (void *) &init_mm;
498 * Set phys_avail to the amount of physical memory,
499 * less the kernel text/data/bss.
502 set_phys_avail(unsigned long total_memory)
504 unsigned long kstart, ksize;
507 * Initially, available physical memory is equivalent to all
511 phys_avail.regions[0].address = PPC_MEMSTART;
512 phys_avail.regions[0].size = total_memory;
513 phys_avail.n_regions = 1;
516 * Map out the kernel text/data/bss from the available physical
520 kstart = __pa(_stext); /* should be 0 */
521 ksize = PAGE_ALIGN(klimit - _stext);
523 mem_pieces_remove(&phys_avail, kstart, ksize, 0);
524 mem_pieces_remove(&phys_avail, 0, 0x4000, 0);
526 #if defined(CONFIG_BLK_DEV_INITRD)
527 /* Remove the init RAM disk from the available memory. */
529 mem_pieces_remove(&phys_avail, __pa(initrd_start),
530 initrd_end - initrd_start, 1);
532 #endif /* CONFIG_BLK_DEV_INITRD */
534 /* remove the RTAS pages from the available memory */
536 mem_pieces_remove(&phys_avail, rtas_data, rtas_size, 1);
538 /* remove the sysmap pages from the available memory */
540 mem_pieces_remove(&phys_avail, __pa(sysmap), sysmap_size, 1);
541 #ifdef CONFIG_PPC_PMAC
542 /* Because of some uninorth weirdness, we need a page of
543 * memory as high as possible (it must be outside of the
544 * bus address seen as the AGP aperture). It will be used
545 * by the r128 DRM driver
547 * FIXME: We need to make sure that page doesn't overlap any of the\
548 * above. This could be done by improving mem_pieces_find to be able
549 * to do a backward search from the end of the list.
551 if (_machine == _MACH_Pmac && find_devices("uni-north-agp")) {
552 agp_special_page = (total_memory - PAGE_SIZE);
553 mem_pieces_remove(&phys_avail, agp_special_page, PAGE_SIZE, 0);
554 agp_special_page = (unsigned long)__va(agp_special_page);
556 #endif /* CONFIG_PPC_PMAC */
559 /* Mark some memory as reserved by removing it from phys_avail. */
560 void __init reserve_phys_mem(unsigned long start, unsigned long size)
562 mem_pieces_remove(&phys_avail, start, size, 1);
566 * This is called when a page has been modified by the kernel.
567 * It just marks the page as not i-cache clean. We do the i-cache
568 * flush later when the page is given to a user process, if necessary.
570 void flush_dcache_page(struct page *page)
572 clear_bit(PG_arch_1, &page->flags);
575 void flush_dcache_icache_page(struct page *page)
578 __flush_dcache_icache(kmap(page));
581 __flush_dcache_icache_phys(page_to_pfn(page) << PAGE_SHIFT);
585 void clear_user_page(void *page, unsigned long vaddr, struct page *pg)
588 clear_bit(PG_arch_1, &pg->flags);
591 void copy_user_page(void *vto, void *vfrom, unsigned long vaddr,
594 copy_page(vto, vfrom);
595 clear_bit(PG_arch_1, &pg->flags);
598 void flush_icache_user_range(struct vm_area_struct *vma, struct page *page,
599 unsigned long addr, int len)
603 maddr = (unsigned long) kmap(page) + (addr & ~PAGE_MASK);
604 flush_icache_range(maddr, maddr + len);
609 * This is called at the end of handling a user page fault, when the
610 * fault has been handled by updating a PTE in the linux page tables.
611 * We use it to preload an HPTE into the hash table corresponding to
612 * the updated linux PTE.
614 void update_mmu_cache(struct vm_area_struct *vma, unsigned long address,
617 /* handle i-cache coherency */
618 unsigned long pfn = pte_pfn(pte);
620 if (pfn_valid(pfn)) {
621 struct page *page = pfn_to_page(pfn);
622 if (!PageReserved(page)
623 && !test_bit(PG_arch_1, &page->flags)) {
624 if (vma->vm_mm == current->active_mm)
625 __flush_dcache_icache((void *) address);
627 flush_dcache_icache_page(page);
628 set_bit(PG_arch_1, &page->flags);
632 #ifdef CONFIG_PPC_STD_MMU
633 /* We only want HPTEs for linux PTEs that have _PAGE_ACCESSED set */
634 if (Hash != 0 && pte_young(pte)) {
635 struct mm_struct *mm;
638 mm = (address < TASK_SIZE)? vma->vm_mm: &init_mm;
639 pmd = pmd_offset(pgd_offset(mm, address), address);
641 add_hash_page(mm->context, address, pmd_val(*pmd));
646 int page_is_ram (unsigned long pagenr)