7 * Copyright (C) 1996 Paul Mackerras
10 * Derived from "arch/i386/mm/init.c"
11 * Copyright (C) 1991, 1992, 1993, 1994 Linus Torvalds
13 * This program is free software; you can redistribute it and/or
14 * modify it under the terms of the GNU General Public License
15 * as published by the Free Software Foundation; either version
16 * 2 of the License, or (at your option) any later version.
20 #include <linux/export.h>
21 #include <linux/sched.h>
22 #include <linux/kernel.h>
23 #include <linux/errno.h>
24 #include <linux/string.h>
25 #include <linux/gfp.h>
26 #include <linux/types.h>
28 #include <linux/stddef.h>
29 #include <linux/init.h>
30 #include <linux/bootmem.h>
31 #include <linux/highmem.h>
32 #include <linux/initrd.h>
33 #include <linux/pagemap.h>
34 #include <linux/suspend.h>
35 #include <linux/memblock.h>
36 #include <linux/hugetlb.h>
37 #include <linux/slab.h>
38 #include <linux/vmalloc.h>
40 #include <asm/pgalloc.h>
43 #include <asm/mmu_context.h>
44 #include <asm/pgtable.h>
47 #include <asm/machdep.h>
48 #include <asm/btext.h>
50 #include <asm/sections.h>
51 #include <asm/sparsemem.h>
53 #include <asm/fixmap.h>
54 #include <asm/swiotlb.h>
59 #ifndef CPU_FTR_COHERENT_ICACHE
60 #define CPU_FTR_COHERENT_ICACHE 0 /* XXX for now */
61 #define CPU_FTR_NOEXECUTE 0
64 unsigned long long memory_limit;
68 EXPORT_SYMBOL(kmap_pte);
70 EXPORT_SYMBOL(kmap_prot);
71 #define TOP_ZONE ZONE_HIGHMEM
73 static inline pte_t *virt_to_kpte(unsigned long vaddr)
75 return pte_offset_kernel(pmd_offset(pud_offset(pgd_offset_k(vaddr),
76 vaddr), vaddr), vaddr);
79 #define TOP_ZONE ZONE_NORMAL
82 int page_is_ram(unsigned long pfn)
84 #ifndef CONFIG_PPC64 /* XXX for now */
87 unsigned long paddr = (pfn << PAGE_SHIFT);
88 struct memblock_region *reg;
90 for_each_memblock(memory, reg)
91 if (paddr >= reg->base && paddr < (reg->base + reg->size))
97 pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
98 unsigned long size, pgprot_t vma_prot)
100 if (ppc_md.phys_mem_access_prot)
101 return ppc_md.phys_mem_access_prot(file, pfn, size, vma_prot);
103 if (!page_is_ram(pfn))
104 vma_prot = pgprot_noncached(vma_prot);
108 EXPORT_SYMBOL(phys_mem_access_prot);
110 #ifdef CONFIG_MEMORY_HOTPLUG
113 int memory_add_physaddr_to_nid(u64 start)
115 return hot_add_scn_to_nid(start);
119 int __weak create_section_mapping(unsigned long start, unsigned long end)
124 int __weak remove_section_mapping(unsigned long start, unsigned long end)
129 int arch_add_memory(int nid, u64 start, u64 size, bool for_device)
131 unsigned long start_pfn = start >> PAGE_SHIFT;
132 unsigned long nr_pages = size >> PAGE_SHIFT;
135 resize_hpt_for_hotplug(memblock_phys_mem_size());
137 start = (unsigned long)__va(start);
138 rc = create_section_mapping(start, start + size);
141 "Unable to create mapping for hot added memory 0x%llx..0x%llx: %d\n",
142 start, start + size, rc);
146 return __add_pages(nid, start_pfn, nr_pages, !for_device);
149 #ifdef CONFIG_MEMORY_HOTREMOVE
150 int arch_remove_memory(u64 start, u64 size)
152 unsigned long start_pfn = start >> PAGE_SHIFT;
153 unsigned long nr_pages = size >> PAGE_SHIFT;
157 zone = page_zone(pfn_to_page(start_pfn));
158 ret = __remove_pages(zone, start_pfn, nr_pages);
162 /* Remove htab bolted mappings for this section of memory */
163 start = (unsigned long)__va(start);
164 ret = remove_section_mapping(start, start + size);
166 /* Ensure all vmalloc mappings are flushed in case they also
167 * hit that section of memory
171 resize_hpt_for_hotplug(memblock_phys_mem_size());
176 #endif /* CONFIG_MEMORY_HOTPLUG */
179 * walk_memory_resource() needs to make sure there is no holes in a given
180 * memory range. PPC64 does not maintain the memory layout in /proc/iomem.
181 * Instead it maintains it in memblock.memory structures. Walk through the
182 * memory regions, find holes and callback for contiguous regions.
185 walk_system_ram_range(unsigned long start_pfn, unsigned long nr_pages,
186 void *arg, int (*func)(unsigned long, unsigned long, void *))
188 struct memblock_region *reg;
189 unsigned long end_pfn = start_pfn + nr_pages;
190 unsigned long tstart, tend;
193 for_each_memblock(memory, reg) {
194 tstart = max(start_pfn, memblock_region_memory_base_pfn(reg));
195 tend = min(end_pfn, memblock_region_memory_end_pfn(reg));
198 ret = (*func)(tstart, tend - tstart, arg);
204 EXPORT_SYMBOL_GPL(walk_system_ram_range);
206 #ifndef CONFIG_NEED_MULTIPLE_NODES
207 void __init initmem_init(void)
209 max_low_pfn = max_pfn = memblock_end_of_DRAM() >> PAGE_SHIFT;
210 min_low_pfn = MEMORY_START >> PAGE_SHIFT;
211 #ifdef CONFIG_HIGHMEM
212 max_low_pfn = lowmem_end_addr >> PAGE_SHIFT;
215 /* Place all memblock_regions in the same node and merge contiguous
218 memblock_set_node(0, (phys_addr_t)ULLONG_MAX, &memblock.memory, 0);
220 /* XXX need to clip this if using highmem? */
221 sparse_memory_present_with_active_regions(0);
225 /* mark pages that don't exist as nosave */
226 static int __init mark_nonram_nosave(void)
228 struct memblock_region *reg, *prev = NULL;
230 for_each_memblock(memory, reg) {
232 memblock_region_memory_end_pfn(prev) < memblock_region_memory_base_pfn(reg))
233 register_nosave_region(memblock_region_memory_end_pfn(prev),
234 memblock_region_memory_base_pfn(reg));
239 #else /* CONFIG_NEED_MULTIPLE_NODES */
240 static int __init mark_nonram_nosave(void)
246 static bool zone_limits_final;
249 * The memory zones past TOP_ZONE are managed by generic mm code.
250 * These should be set to zero since that's what every other
253 static unsigned long max_zone_pfns[MAX_NR_ZONES] = {
254 [0 ... TOP_ZONE ] = ~0UL,
255 [TOP_ZONE + 1 ... MAX_NR_ZONES - 1] = 0
259 * Restrict the specified zone and all more restrictive zones
260 * to be below the specified pfn. May not be called after
263 void __init limit_zone_pfn(enum zone_type zone, unsigned long pfn_limit)
267 if (WARN_ON(zone_limits_final))
270 for (i = zone; i >= 0; i--) {
271 if (max_zone_pfns[i] > pfn_limit)
272 max_zone_pfns[i] = pfn_limit;
277 * Find the least restrictive zone that is entirely below the
278 * specified pfn limit. Returns < 0 if no suitable zone is found.
280 * pfn_limit must be u64 because it can exceed 32 bits even on 32-bit
281 * systems -- the DMA limit can be higher than any possible real pfn.
283 int dma_pfn_limit_to_zone(u64 pfn_limit)
287 for (i = TOP_ZONE; i >= 0; i--) {
288 if (max_zone_pfns[i] <= pfn_limit)
296 * paging_init() sets up the page tables - in fact we've already done this.
298 void __init paging_init(void)
300 unsigned long long total_ram = memblock_phys_mem_size();
301 phys_addr_t top_of_ram = memblock_end_of_DRAM();
304 unsigned long v = __fix_to_virt(__end_of_fixed_addresses - 1);
305 unsigned long end = __fix_to_virt(FIX_HOLE);
307 for (; v < end; v += PAGE_SIZE)
308 map_page(v, 0, 0); /* XXX gross */
311 #ifdef CONFIG_HIGHMEM
312 map_page(PKMAP_BASE, 0, 0); /* XXX gross */
313 pkmap_page_table = virt_to_kpte(PKMAP_BASE);
315 kmap_pte = virt_to_kpte(__fix_to_virt(FIX_KMAP_BEGIN));
316 kmap_prot = PAGE_KERNEL;
317 #endif /* CONFIG_HIGHMEM */
319 printk(KERN_DEBUG "Top of RAM: 0x%llx, Total RAM: 0x%llx\n",
320 (unsigned long long)top_of_ram, total_ram);
321 printk(KERN_DEBUG "Memory hole size: %ldMB\n",
322 (long int)((top_of_ram - total_ram) >> 20));
324 #ifdef CONFIG_HIGHMEM
325 limit_zone_pfn(ZONE_NORMAL, lowmem_end_addr >> PAGE_SHIFT);
327 limit_zone_pfn(TOP_ZONE, top_of_ram >> PAGE_SHIFT);
328 zone_limits_final = true;
329 free_area_init_nodes(max_zone_pfns);
331 mark_nonram_nosave();
334 void __init mem_init(void)
337 * book3s is limited to 16 page sizes due to encoding this in
338 * a 4-bit field for slices.
340 BUILD_BUG_ON(MMU_PAGE_COUNT > 16);
342 #ifdef CONFIG_SWIOTLB
346 high_memory = (void *) __va(max_low_pfn * PAGE_SIZE);
347 set_max_mapnr(max_pfn);
350 #ifdef CONFIG_HIGHMEM
352 unsigned long pfn, highmem_mapnr;
354 highmem_mapnr = lowmem_end_addr >> PAGE_SHIFT;
355 for (pfn = highmem_mapnr; pfn < max_mapnr; ++pfn) {
356 phys_addr_t paddr = (phys_addr_t)pfn << PAGE_SHIFT;
357 struct page *page = pfn_to_page(pfn);
358 if (!memblock_is_reserved(paddr))
359 free_highmem_page(page);
362 #endif /* CONFIG_HIGHMEM */
364 #if defined(CONFIG_PPC_FSL_BOOK3E) && !defined(CONFIG_SMP)
366 * If smp is enabled, next_tlbcam_idx is initialized in the cpu up
367 * functions.... do it here for the non-smp case.
369 per_cpu(next_tlbcam_idx, smp_processor_id()) =
370 (mfspr(SPRN_TLB1CFG) & TLBnCFG_N_ENTRY) - 1;
373 mem_init_print_info(NULL);
375 pr_info("Kernel virtual memory layout:\n");
376 pr_info(" * 0x%08lx..0x%08lx : fixmap\n", FIXADDR_START, FIXADDR_TOP);
377 #ifdef CONFIG_HIGHMEM
378 pr_info(" * 0x%08lx..0x%08lx : highmem PTEs\n",
379 PKMAP_BASE, PKMAP_ADDR(LAST_PKMAP));
380 #endif /* CONFIG_HIGHMEM */
381 #ifdef CONFIG_NOT_COHERENT_CACHE
382 pr_info(" * 0x%08lx..0x%08lx : consistent mem\n",
383 IOREMAP_TOP, IOREMAP_TOP + CONFIG_CONSISTENT_SIZE);
384 #endif /* CONFIG_NOT_COHERENT_CACHE */
385 pr_info(" * 0x%08lx..0x%08lx : early ioremap\n",
386 ioremap_bot, IOREMAP_TOP);
387 pr_info(" * 0x%08lx..0x%08lx : vmalloc & ioremap\n",
388 VMALLOC_START, VMALLOC_END);
389 #endif /* CONFIG_PPC32 */
392 void free_initmem(void)
394 ppc_md.progress = ppc_printk_progress;
395 free_initmem_default(POISON_FREE_INITMEM);
398 #ifdef CONFIG_BLK_DEV_INITRD
399 void __init free_initrd_mem(unsigned long start, unsigned long end)
401 free_reserved_area((void *)start, (void *)end, -1, "initrd");
406 * This is called when a page has been modified by the kernel.
407 * It just marks the page as not i-cache clean. We do the i-cache
408 * flush later when the page is given to a user process, if necessary.
410 void flush_dcache_page(struct page *page)
412 if (cpu_has_feature(CPU_FTR_COHERENT_ICACHE))
414 /* avoid an atomic op if possible */
415 if (test_bit(PG_arch_1, &page->flags))
416 clear_bit(PG_arch_1, &page->flags);
418 EXPORT_SYMBOL(flush_dcache_page);
420 void flush_dcache_icache_page(struct page *page)
422 #ifdef CONFIG_HUGETLB_PAGE
423 if (PageCompound(page)) {
424 flush_dcache_icache_hugepage(page);
428 #if defined(CONFIG_8xx) || defined(CONFIG_PPC64)
429 /* On 8xx there is no need to kmap since highmem is not supported */
430 __flush_dcache_icache(page_address(page));
432 if (IS_ENABLED(CONFIG_BOOKE) || sizeof(phys_addr_t) > sizeof(void *)) {
433 void *start = kmap_atomic(page);
434 __flush_dcache_icache(start);
435 kunmap_atomic(start);
437 __flush_dcache_icache_phys(page_to_pfn(page) << PAGE_SHIFT);
441 EXPORT_SYMBOL(flush_dcache_icache_page);
443 void clear_user_page(void *page, unsigned long vaddr, struct page *pg)
448 * We shouldn't have to do this, but some versions of glibc
449 * require it (ld.so assumes zero filled pages are icache clean)
452 flush_dcache_page(pg);
454 EXPORT_SYMBOL(clear_user_page);
456 void copy_user_page(void *vto, void *vfrom, unsigned long vaddr,
459 copy_page(vto, vfrom);
462 * We should be able to use the following optimisation, however
463 * there are two problems.
464 * Firstly a bug in some versions of binutils meant PLT sections
465 * were not marked executable.
466 * Secondly the first word in the GOT section is blrl, used
467 * to establish the GOT address. Until recently the GOT was
468 * not marked executable.
472 if (!vma->vm_file && ((vma->vm_flags & VM_EXEC) == 0))
476 flush_dcache_page(pg);
479 void flush_icache_user_range(struct vm_area_struct *vma, struct page *page,
480 unsigned long addr, int len)
484 maddr = (unsigned long) kmap(page) + (addr & ~PAGE_MASK);
485 flush_icache_range(maddr, maddr + len);
488 EXPORT_SYMBOL(flush_icache_user_range);
491 * This is called at the end of handling a user page fault, when the
492 * fault has been handled by updating a PTE in the linux page tables.
493 * We use it to preload an HPTE into the hash table corresponding to
494 * the updated linux PTE.
496 * This must always be called with the pte lock held.
498 void update_mmu_cache(struct vm_area_struct *vma, unsigned long address,
501 #ifdef CONFIG_PPC_STD_MMU
503 * We don't need to worry about _PAGE_PRESENT here because we are
504 * called with either mm->page_table_lock held or ptl lock held
506 unsigned long access, trap;
511 /* We only want HPTEs for linux PTEs that have _PAGE_ACCESSED set */
512 if (!pte_young(*ptep) || address >= TASK_SIZE)
515 /* We try to figure out if we are coming from an instruction
516 * access fault and pass that down to __hash_page so we avoid
517 * double-faulting on execution of fresh text. We have to test
518 * for regs NULL since init will get here first thing at boot
520 * We also avoid filling the hash if not coming from a fault
523 trap = current->thread.regs ? TRAP(current->thread.regs) : 0UL;
535 hash_preload(vma->vm_mm, address, access, trap);
536 #endif /* CONFIG_PPC_STD_MMU */
537 #if (defined(CONFIG_PPC_BOOK3E_64) || defined(CONFIG_PPC_FSL_BOOK3E)) \
538 && defined(CONFIG_HUGETLB_PAGE)
539 if (is_vm_hugetlb_page(vma))
540 book3e_hugetlb_preload(vma, address, *ptep);
545 * System memory should not be in /proc/iomem but various tools expect it
548 static int __init add_system_ram_resources(void)
550 struct memblock_region *reg;
552 for_each_memblock(memory, reg) {
553 struct resource *res;
554 unsigned long base = reg->base;
555 unsigned long size = reg->size;
557 res = kzalloc(sizeof(struct resource), GFP_KERNEL);
561 res->name = "System RAM";
563 res->end = base + size - 1;
564 res->flags = IORESOURCE_SYSTEM_RAM | IORESOURCE_BUSY;
565 WARN_ON(request_resource(&iomem_resource, res) < 0);
571 subsys_initcall(add_system_ram_resources);
573 #ifdef CONFIG_STRICT_DEVMEM
575 * devmem_is_allowed(): check to see if /dev/mem access to a certain address
576 * is valid. The argument is a physical page number.
578 * Access has to be given to non-kernel-ram areas as well, these contain the
579 * PCI mmio resources as well as potential bios/acpi data regions.
581 int devmem_is_allowed(unsigned long pfn)
583 if (page_is_rtas_user_buf(pfn))
585 if (iomem_is_exclusive(PFN_PHYS(pfn)))
587 if (!page_is_ram(pfn))
591 #endif /* CONFIG_STRICT_DEVMEM */