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>
39 #include <linux/memremap.h>
41 #include <asm/pgalloc.h>
44 #include <asm/mmu_context.h>
45 #include <asm/pgtable.h>
48 #include <asm/machdep.h>
49 #include <asm/btext.h>
51 #include <asm/sections.h>
52 #include <asm/sparsemem.h>
54 #include <asm/fixmap.h>
55 #include <asm/swiotlb.h>
60 #ifndef CPU_FTR_COHERENT_ICACHE
61 #define CPU_FTR_COHERENT_ICACHE 0 /* XXX for now */
62 #define CPU_FTR_NOEXECUTE 0
65 unsigned long long memory_limit;
69 EXPORT_SYMBOL(kmap_pte);
71 EXPORT_SYMBOL(kmap_prot);
72 #define TOP_ZONE ZONE_HIGHMEM
74 static inline pte_t *virt_to_kpte(unsigned long vaddr)
76 return pte_offset_kernel(pmd_offset(pud_offset(pgd_offset_k(vaddr),
77 vaddr), vaddr), vaddr);
80 #define TOP_ZONE ZONE_NORMAL
83 int page_is_ram(unsigned long pfn)
85 #ifndef CONFIG_PPC64 /* XXX for now */
88 unsigned long paddr = (pfn << PAGE_SHIFT);
89 struct memblock_region *reg;
91 for_each_memblock(memory, reg)
92 if (paddr >= reg->base && paddr < (reg->base + reg->size))
98 pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
99 unsigned long size, pgprot_t vma_prot)
101 if (ppc_md.phys_mem_access_prot)
102 return ppc_md.phys_mem_access_prot(file, pfn, size, vma_prot);
104 if (!page_is_ram(pfn))
105 vma_prot = pgprot_noncached(vma_prot);
109 EXPORT_SYMBOL(phys_mem_access_prot);
111 #ifdef CONFIG_MEMORY_HOTPLUG
114 int memory_add_physaddr_to_nid(u64 start)
116 return hot_add_scn_to_nid(start);
120 int __weak create_section_mapping(unsigned long start, unsigned long end)
125 int __weak remove_section_mapping(unsigned long start, unsigned long end)
130 int arch_add_memory(int nid, u64 start, u64 size, bool want_memblock)
132 unsigned long start_pfn = start >> PAGE_SHIFT;
133 unsigned long nr_pages = size >> PAGE_SHIFT;
136 resize_hpt_for_hotplug(memblock_phys_mem_size());
138 start = (unsigned long)__va(start);
139 rc = create_section_mapping(start, start + size);
142 "Unable to create mapping for hot added memory 0x%llx..0x%llx: %d\n",
143 start, start + size, rc);
147 return __add_pages(nid, start_pfn, nr_pages, want_memblock);
150 #ifdef CONFIG_MEMORY_HOTREMOVE
151 int arch_remove_memory(u64 start, u64 size)
153 unsigned long start_pfn = start >> PAGE_SHIFT;
154 unsigned long nr_pages = size >> PAGE_SHIFT;
155 struct vmem_altmap *altmap;
160 * If we have an altmap then we need to skip over any reserved PFNs
161 * when querying the zone.
163 page = pfn_to_page(start_pfn);
164 altmap = to_vmem_altmap((unsigned long) page);
166 page += vmem_altmap_offset(altmap);
168 ret = __remove_pages(page_zone(page), start_pfn, nr_pages);
172 /* Remove htab bolted mappings for this section of memory */
173 start = (unsigned long)__va(start);
174 ret = remove_section_mapping(start, start + size);
176 /* Ensure all vmalloc mappings are flushed in case they also
177 * hit that section of memory
181 resize_hpt_for_hotplug(memblock_phys_mem_size());
186 #endif /* CONFIG_MEMORY_HOTPLUG */
189 * walk_memory_resource() needs to make sure there is no holes in a given
190 * memory range. PPC64 does not maintain the memory layout in /proc/iomem.
191 * Instead it maintains it in memblock.memory structures. Walk through the
192 * memory regions, find holes and callback for contiguous regions.
195 walk_system_ram_range(unsigned long start_pfn, unsigned long nr_pages,
196 void *arg, int (*func)(unsigned long, unsigned long, void *))
198 struct memblock_region *reg;
199 unsigned long end_pfn = start_pfn + nr_pages;
200 unsigned long tstart, tend;
203 for_each_memblock(memory, reg) {
204 tstart = max(start_pfn, memblock_region_memory_base_pfn(reg));
205 tend = min(end_pfn, memblock_region_memory_end_pfn(reg));
208 ret = (*func)(tstart, tend - tstart, arg);
214 EXPORT_SYMBOL_GPL(walk_system_ram_range);
216 #ifndef CONFIG_NEED_MULTIPLE_NODES
217 void __init initmem_init(void)
219 max_low_pfn = max_pfn = memblock_end_of_DRAM() >> PAGE_SHIFT;
220 min_low_pfn = MEMORY_START >> PAGE_SHIFT;
221 #ifdef CONFIG_HIGHMEM
222 max_low_pfn = lowmem_end_addr >> PAGE_SHIFT;
225 /* Place all memblock_regions in the same node and merge contiguous
228 memblock_set_node(0, (phys_addr_t)ULLONG_MAX, &memblock.memory, 0);
230 /* XXX need to clip this if using highmem? */
231 sparse_memory_present_with_active_regions(0);
235 /* mark pages that don't exist as nosave */
236 static int __init mark_nonram_nosave(void)
238 struct memblock_region *reg, *prev = NULL;
240 for_each_memblock(memory, reg) {
242 memblock_region_memory_end_pfn(prev) < memblock_region_memory_base_pfn(reg))
243 register_nosave_region(memblock_region_memory_end_pfn(prev),
244 memblock_region_memory_base_pfn(reg));
249 #else /* CONFIG_NEED_MULTIPLE_NODES */
250 static int __init mark_nonram_nosave(void)
256 static bool zone_limits_final;
259 * The memory zones past TOP_ZONE are managed by generic mm code.
260 * These should be set to zero since that's what every other
263 static unsigned long max_zone_pfns[MAX_NR_ZONES] = {
264 [0 ... TOP_ZONE ] = ~0UL,
265 [TOP_ZONE + 1 ... MAX_NR_ZONES - 1] = 0
269 * Restrict the specified zone and all more restrictive zones
270 * to be below the specified pfn. May not be called after
273 void __init limit_zone_pfn(enum zone_type zone, unsigned long pfn_limit)
277 if (WARN_ON(zone_limits_final))
280 for (i = zone; i >= 0; i--) {
281 if (max_zone_pfns[i] > pfn_limit)
282 max_zone_pfns[i] = pfn_limit;
287 * Find the least restrictive zone that is entirely below the
288 * specified pfn limit. Returns < 0 if no suitable zone is found.
290 * pfn_limit must be u64 because it can exceed 32 bits even on 32-bit
291 * systems -- the DMA limit can be higher than any possible real pfn.
293 int dma_pfn_limit_to_zone(u64 pfn_limit)
297 for (i = TOP_ZONE; i >= 0; i--) {
298 if (max_zone_pfns[i] <= pfn_limit)
306 * paging_init() sets up the page tables - in fact we've already done this.
308 void __init paging_init(void)
310 unsigned long long total_ram = memblock_phys_mem_size();
311 phys_addr_t top_of_ram = memblock_end_of_DRAM();
314 unsigned long v = __fix_to_virt(__end_of_fixed_addresses - 1);
315 unsigned long end = __fix_to_virt(FIX_HOLE);
317 for (; v < end; v += PAGE_SIZE)
318 map_kernel_page(v, 0, 0); /* XXX gross */
321 #ifdef CONFIG_HIGHMEM
322 map_kernel_page(PKMAP_BASE, 0, 0); /* XXX gross */
323 pkmap_page_table = virt_to_kpte(PKMAP_BASE);
325 kmap_pte = virt_to_kpte(__fix_to_virt(FIX_KMAP_BEGIN));
326 kmap_prot = PAGE_KERNEL;
327 #endif /* CONFIG_HIGHMEM */
329 printk(KERN_DEBUG "Top of RAM: 0x%llx, Total RAM: 0x%llx\n",
330 (unsigned long long)top_of_ram, total_ram);
331 printk(KERN_DEBUG "Memory hole size: %ldMB\n",
332 (long int)((top_of_ram - total_ram) >> 20));
334 #ifdef CONFIG_HIGHMEM
335 limit_zone_pfn(ZONE_NORMAL, lowmem_end_addr >> PAGE_SHIFT);
337 limit_zone_pfn(TOP_ZONE, top_of_ram >> PAGE_SHIFT);
338 zone_limits_final = true;
339 free_area_init_nodes(max_zone_pfns);
341 mark_nonram_nosave();
344 void __init mem_init(void)
347 * book3s is limited to 16 page sizes due to encoding this in
348 * a 4-bit field for slices.
350 BUILD_BUG_ON(MMU_PAGE_COUNT > 16);
352 #ifdef CONFIG_SWIOTLB
356 high_memory = (void *) __va(max_low_pfn * PAGE_SIZE);
357 set_max_mapnr(max_pfn);
360 #ifdef CONFIG_HIGHMEM
362 unsigned long pfn, highmem_mapnr;
364 highmem_mapnr = lowmem_end_addr >> PAGE_SHIFT;
365 for (pfn = highmem_mapnr; pfn < max_mapnr; ++pfn) {
366 phys_addr_t paddr = (phys_addr_t)pfn << PAGE_SHIFT;
367 struct page *page = pfn_to_page(pfn);
368 if (!memblock_is_reserved(paddr))
369 free_highmem_page(page);
372 #endif /* CONFIG_HIGHMEM */
374 #if defined(CONFIG_PPC_FSL_BOOK3E) && !defined(CONFIG_SMP)
376 * If smp is enabled, next_tlbcam_idx is initialized in the cpu up
377 * functions.... do it here for the non-smp case.
379 per_cpu(next_tlbcam_idx, smp_processor_id()) =
380 (mfspr(SPRN_TLB1CFG) & TLBnCFG_N_ENTRY) - 1;
383 mem_init_print_info(NULL);
385 pr_info("Kernel virtual memory layout:\n");
386 pr_info(" * 0x%08lx..0x%08lx : fixmap\n", FIXADDR_START, FIXADDR_TOP);
387 #ifdef CONFIG_HIGHMEM
388 pr_info(" * 0x%08lx..0x%08lx : highmem PTEs\n",
389 PKMAP_BASE, PKMAP_ADDR(LAST_PKMAP));
390 #endif /* CONFIG_HIGHMEM */
391 #ifdef CONFIG_NOT_COHERENT_CACHE
392 pr_info(" * 0x%08lx..0x%08lx : consistent mem\n",
393 IOREMAP_TOP, IOREMAP_TOP + CONFIG_CONSISTENT_SIZE);
394 #endif /* CONFIG_NOT_COHERENT_CACHE */
395 pr_info(" * 0x%08lx..0x%08lx : early ioremap\n",
396 ioremap_bot, IOREMAP_TOP);
397 pr_info(" * 0x%08lx..0x%08lx : vmalloc & ioremap\n",
398 VMALLOC_START, VMALLOC_END);
399 #endif /* CONFIG_PPC32 */
402 void free_initmem(void)
404 ppc_md.progress = ppc_printk_progress;
406 free_initmem_default(POISON_FREE_INITMEM);
409 #ifdef CONFIG_BLK_DEV_INITRD
410 void __init free_initrd_mem(unsigned long start, unsigned long end)
412 free_reserved_area((void *)start, (void *)end, -1, "initrd");
417 * This is called when a page has been modified by the kernel.
418 * It just marks the page as not i-cache clean. We do the i-cache
419 * flush later when the page is given to a user process, if necessary.
421 void flush_dcache_page(struct page *page)
423 if (cpu_has_feature(CPU_FTR_COHERENT_ICACHE))
425 /* avoid an atomic op if possible */
426 if (test_bit(PG_arch_1, &page->flags))
427 clear_bit(PG_arch_1, &page->flags);
429 EXPORT_SYMBOL(flush_dcache_page);
431 void flush_dcache_icache_page(struct page *page)
433 #ifdef CONFIG_HUGETLB_PAGE
434 if (PageCompound(page)) {
435 flush_dcache_icache_hugepage(page);
439 #if defined(CONFIG_PPC_8xx) || defined(CONFIG_PPC64)
440 /* On 8xx there is no need to kmap since highmem is not supported */
441 __flush_dcache_icache(page_address(page));
443 if (IS_ENABLED(CONFIG_BOOKE) || sizeof(phys_addr_t) > sizeof(void *)) {
444 void *start = kmap_atomic(page);
445 __flush_dcache_icache(start);
446 kunmap_atomic(start);
448 __flush_dcache_icache_phys(page_to_pfn(page) << PAGE_SHIFT);
452 EXPORT_SYMBOL(flush_dcache_icache_page);
454 void clear_user_page(void *page, unsigned long vaddr, struct page *pg)
459 * We shouldn't have to do this, but some versions of glibc
460 * require it (ld.so assumes zero filled pages are icache clean)
463 flush_dcache_page(pg);
465 EXPORT_SYMBOL(clear_user_page);
467 void copy_user_page(void *vto, void *vfrom, unsigned long vaddr,
470 copy_page(vto, vfrom);
473 * We should be able to use the following optimisation, however
474 * there are two problems.
475 * Firstly a bug in some versions of binutils meant PLT sections
476 * were not marked executable.
477 * Secondly the first word in the GOT section is blrl, used
478 * to establish the GOT address. Until recently the GOT was
479 * not marked executable.
483 if (!vma->vm_file && ((vma->vm_flags & VM_EXEC) == 0))
487 flush_dcache_page(pg);
490 void flush_icache_user_range(struct vm_area_struct *vma, struct page *page,
491 unsigned long addr, int len)
495 maddr = (unsigned long) kmap(page) + (addr & ~PAGE_MASK);
496 flush_icache_range(maddr, maddr + len);
499 EXPORT_SYMBOL(flush_icache_user_range);
502 * This is called at the end of handling a user page fault, when the
503 * fault has been handled by updating a PTE in the linux page tables.
504 * We use it to preload an HPTE into the hash table corresponding to
505 * the updated linux PTE.
507 * This must always be called with the pte lock held.
509 void update_mmu_cache(struct vm_area_struct *vma, unsigned long address,
512 #ifdef CONFIG_PPC_STD_MMU
514 * We don't need to worry about _PAGE_PRESENT here because we are
515 * called with either mm->page_table_lock held or ptl lock held
517 unsigned long access, trap;
522 /* We only want HPTEs for linux PTEs that have _PAGE_ACCESSED set */
523 if (!pte_young(*ptep) || address >= TASK_SIZE)
526 /* We try to figure out if we are coming from an instruction
527 * access fault and pass that down to __hash_page so we avoid
528 * double-faulting on execution of fresh text. We have to test
529 * for regs NULL since init will get here first thing at boot
531 * We also avoid filling the hash if not coming from a fault
534 trap = current->thread.regs ? TRAP(current->thread.regs) : 0UL;
546 hash_preload(vma->vm_mm, address, access, trap);
547 #endif /* CONFIG_PPC_STD_MMU */
548 #if (defined(CONFIG_PPC_BOOK3E_64) || defined(CONFIG_PPC_FSL_BOOK3E)) \
549 && defined(CONFIG_HUGETLB_PAGE)
550 if (is_vm_hugetlb_page(vma))
551 book3e_hugetlb_preload(vma, address, *ptep);
556 * System memory should not be in /proc/iomem but various tools expect it
559 static int __init add_system_ram_resources(void)
561 struct memblock_region *reg;
563 for_each_memblock(memory, reg) {
564 struct resource *res;
565 unsigned long base = reg->base;
566 unsigned long size = reg->size;
568 res = kzalloc(sizeof(struct resource), GFP_KERNEL);
572 res->name = "System RAM";
574 res->end = base + size - 1;
575 res->flags = IORESOURCE_SYSTEM_RAM | IORESOURCE_BUSY;
576 WARN_ON(request_resource(&iomem_resource, res) < 0);
582 subsys_initcall(add_system_ram_resources);
584 #ifdef CONFIG_STRICT_DEVMEM
586 * devmem_is_allowed(): check to see if /dev/mem access to a certain address
587 * is valid. The argument is a physical page number.
589 * Access has to be given to non-kernel-ram areas as well, these contain the
590 * PCI mmio resources as well as potential bios/acpi data regions.
592 int devmem_is_allowed(unsigned long pfn)
594 if (page_is_rtas_user_buf(pfn))
596 if (iomem_is_exclusive(PFN_PHYS(pfn)))
598 if (!page_is_ram(pfn))
602 #endif /* CONFIG_STRICT_DEVMEM */