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 * PPC44x/36-bit changes by Matt Porter (mporter@mvista.com)
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;
66 bool init_mem_is_free;
70 EXPORT_SYMBOL(kmap_pte);
72 EXPORT_SYMBOL(kmap_prot);
73 #define TOP_ZONE ZONE_HIGHMEM
75 static inline pte_t *virt_to_kpte(unsigned long vaddr)
77 return pte_offset_kernel(pmd_offset(pud_offset(pgd_offset_k(vaddr),
78 vaddr), vaddr), vaddr);
81 #define TOP_ZONE ZONE_NORMAL
84 int page_is_ram(unsigned long pfn)
86 return memblock_is_memory(__pfn_to_phys(pfn));
89 pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
90 unsigned long size, pgprot_t vma_prot)
92 if (ppc_md.phys_mem_access_prot)
93 return ppc_md.phys_mem_access_prot(file, pfn, size, vma_prot);
95 if (!page_is_ram(pfn))
96 vma_prot = pgprot_noncached(vma_prot);
100 EXPORT_SYMBOL(phys_mem_access_prot);
102 #ifdef CONFIG_MEMORY_HOTPLUG
105 int memory_add_physaddr_to_nid(u64 start)
107 return hot_add_scn_to_nid(start);
111 int __weak create_section_mapping(unsigned long start, unsigned long end, int nid)
116 int __weak remove_section_mapping(unsigned long start, unsigned long end)
121 int __ref arch_add_memory(int nid, u64 start, u64 size, struct vmem_altmap *altmap,
124 unsigned long start_pfn = start >> PAGE_SHIFT;
125 unsigned long nr_pages = size >> PAGE_SHIFT;
128 resize_hpt_for_hotplug(memblock_phys_mem_size());
130 start = (unsigned long)__va(start);
131 rc = create_section_mapping(start, start + size, nid);
133 pr_warn("Unable to create mapping for hot added memory 0x%llx..0x%llx: %d\n",
134 start, start + size, rc);
137 flush_inval_dcache_range(start, start + size);
139 return __add_pages(nid, start_pfn, nr_pages, altmap, want_memblock);
142 void __ref arch_remove_memory(int nid, u64 start, u64 size,
143 struct vmem_altmap *altmap)
145 unsigned long start_pfn = start >> PAGE_SHIFT;
146 unsigned long nr_pages = size >> PAGE_SHIFT;
149 __remove_pages(start_pfn, nr_pages, altmap);
151 /* Remove htab bolted mappings for this section of memory */
152 start = (unsigned long)__va(start);
153 flush_inval_dcache_range(start, start + size);
154 ret = remove_section_mapping(start, start + size);
157 /* Ensure all vmalloc mappings are flushed in case they also
158 * hit that section of memory
162 resize_hpt_for_hotplug(memblock_phys_mem_size());
167 * walk_memory_resource() needs to make sure there is no holes in a given
168 * memory range. PPC64 does not maintain the memory layout in /proc/iomem.
169 * Instead it maintains it in memblock.memory structures. Walk through the
170 * memory regions, find holes and callback for contiguous regions.
173 walk_system_ram_range(unsigned long start_pfn, unsigned long nr_pages,
174 void *arg, int (*func)(unsigned long, unsigned long, void *))
176 struct memblock_region *reg;
177 unsigned long end_pfn = start_pfn + nr_pages;
178 unsigned long tstart, tend;
181 for_each_memblock(memory, reg) {
182 tstart = max(start_pfn, memblock_region_memory_base_pfn(reg));
183 tend = min(end_pfn, memblock_region_memory_end_pfn(reg));
186 ret = (*func)(tstart, tend - tstart, arg);
192 EXPORT_SYMBOL_GPL(walk_system_ram_range);
194 #ifndef CONFIG_NEED_MULTIPLE_NODES
195 void __init mem_topology_setup(void)
197 max_low_pfn = max_pfn = memblock_end_of_DRAM() >> PAGE_SHIFT;
198 min_low_pfn = MEMORY_START >> PAGE_SHIFT;
199 #ifdef CONFIG_HIGHMEM
200 max_low_pfn = lowmem_end_addr >> PAGE_SHIFT;
203 /* Place all memblock_regions in the same node and merge contiguous
206 memblock_set_node(0, PHYS_ADDR_MAX, &memblock.memory, 0);
209 void __init initmem_init(void)
211 /* XXX need to clip this if using highmem? */
212 sparse_memory_present_with_active_regions(0);
216 /* mark pages that don't exist as nosave */
217 static int __init mark_nonram_nosave(void)
219 struct memblock_region *reg, *prev = NULL;
221 for_each_memblock(memory, reg) {
223 memblock_region_memory_end_pfn(prev) < memblock_region_memory_base_pfn(reg))
224 register_nosave_region(memblock_region_memory_end_pfn(prev),
225 memblock_region_memory_base_pfn(reg));
230 #else /* CONFIG_NEED_MULTIPLE_NODES */
231 static int __init mark_nonram_nosave(void)
237 static bool zone_limits_final;
240 * The memory zones past TOP_ZONE are managed by generic mm code.
241 * These should be set to zero since that's what every other
244 static unsigned long max_zone_pfns[MAX_NR_ZONES] = {
245 [0 ... TOP_ZONE ] = ~0UL,
246 [TOP_ZONE + 1 ... MAX_NR_ZONES - 1] = 0
250 * Restrict the specified zone and all more restrictive zones
251 * to be below the specified pfn. May not be called after
254 void __init limit_zone_pfn(enum zone_type zone, unsigned long pfn_limit)
258 if (WARN_ON(zone_limits_final))
261 for (i = zone; i >= 0; i--) {
262 if (max_zone_pfns[i] > pfn_limit)
263 max_zone_pfns[i] = pfn_limit;
268 * Find the least restrictive zone that is entirely below the
269 * specified pfn limit. Returns < 0 if no suitable zone is found.
271 * pfn_limit must be u64 because it can exceed 32 bits even on 32-bit
272 * systems -- the DMA limit can be higher than any possible real pfn.
274 int dma_pfn_limit_to_zone(u64 pfn_limit)
278 for (i = TOP_ZONE; i >= 0; i--) {
279 if (max_zone_pfns[i] <= pfn_limit)
287 * paging_init() sets up the page tables - in fact we've already done this.
289 void __init paging_init(void)
291 unsigned long long total_ram = memblock_phys_mem_size();
292 phys_addr_t top_of_ram = memblock_end_of_DRAM();
295 unsigned long v = __fix_to_virt(__end_of_fixed_addresses - 1);
296 unsigned long end = __fix_to_virt(FIX_HOLE);
298 for (; v < end; v += PAGE_SIZE)
299 map_kernel_page(v, 0, 0); /* XXX gross */
302 #ifdef CONFIG_HIGHMEM
303 map_kernel_page(PKMAP_BASE, 0, 0); /* XXX gross */
304 pkmap_page_table = virt_to_kpte(PKMAP_BASE);
306 kmap_pte = virt_to_kpte(__fix_to_virt(FIX_KMAP_BEGIN));
307 kmap_prot = PAGE_KERNEL;
308 #endif /* CONFIG_HIGHMEM */
310 printk(KERN_DEBUG "Top of RAM: 0x%llx, Total RAM: 0x%llx\n",
311 (unsigned long long)top_of_ram, total_ram);
312 printk(KERN_DEBUG "Memory hole size: %ldMB\n",
313 (long int)((top_of_ram - total_ram) >> 20));
315 #ifdef CONFIG_HIGHMEM
316 limit_zone_pfn(ZONE_NORMAL, lowmem_end_addr >> PAGE_SHIFT);
318 limit_zone_pfn(TOP_ZONE, top_of_ram >> PAGE_SHIFT);
319 zone_limits_final = true;
320 free_area_init_nodes(max_zone_pfns);
322 mark_nonram_nosave();
325 void __init mem_init(void)
328 * book3s is limited to 16 page sizes due to encoding this in
329 * a 4-bit field for slices.
331 BUILD_BUG_ON(MMU_PAGE_COUNT > 16);
333 #ifdef CONFIG_SWIOTLB
335 * Some platforms (e.g. 85xx) limit DMA-able memory way below
336 * 4G. We force memblock to bottom-up mode to ensure that the
337 * memory allocated in swiotlb_init() is DMA-able.
338 * As it's the last memblock allocation, no need to reset it
341 memblock_set_bottom_up(true);
345 high_memory = (void *) __va(max_low_pfn * PAGE_SIZE);
346 set_max_mapnr(max_pfn);
349 #ifdef CONFIG_HIGHMEM
351 unsigned long pfn, highmem_mapnr;
353 highmem_mapnr = lowmem_end_addr >> PAGE_SHIFT;
354 for (pfn = highmem_mapnr; pfn < max_mapnr; ++pfn) {
355 phys_addr_t paddr = (phys_addr_t)pfn << PAGE_SHIFT;
356 struct page *page = pfn_to_page(pfn);
357 if (!memblock_is_reserved(paddr))
358 free_highmem_page(page);
361 #endif /* CONFIG_HIGHMEM */
363 #if defined(CONFIG_PPC_FSL_BOOK3E) && !defined(CONFIG_SMP)
365 * If smp is enabled, next_tlbcam_idx is initialized in the cpu up
366 * functions.... do it here for the non-smp case.
368 per_cpu(next_tlbcam_idx, smp_processor_id()) =
369 (mfspr(SPRN_TLB1CFG) & TLBnCFG_N_ENTRY) - 1;
372 mem_init_print_info(NULL);
374 pr_info("Kernel virtual memory layout:\n");
375 pr_info(" * 0x%08lx..0x%08lx : fixmap\n", FIXADDR_START, FIXADDR_TOP);
376 #ifdef CONFIG_HIGHMEM
377 pr_info(" * 0x%08lx..0x%08lx : highmem PTEs\n",
378 PKMAP_BASE, PKMAP_ADDR(LAST_PKMAP));
379 #endif /* CONFIG_HIGHMEM */
380 #ifdef CONFIG_NOT_COHERENT_CACHE
381 pr_info(" * 0x%08lx..0x%08lx : consistent mem\n",
382 IOREMAP_TOP, IOREMAP_TOP + CONFIG_CONSISTENT_SIZE);
383 #endif /* CONFIG_NOT_COHERENT_CACHE */
384 pr_info(" * 0x%08lx..0x%08lx : early ioremap\n",
385 ioremap_bot, IOREMAP_TOP);
386 pr_info(" * 0x%08lx..0x%08lx : vmalloc & ioremap\n",
387 VMALLOC_START, VMALLOC_END);
388 #endif /* CONFIG_PPC32 */
391 void free_initmem(void)
393 ppc_md.progress = ppc_printk_progress;
395 init_mem_is_free = true;
396 free_initmem_default(POISON_FREE_INITMEM);
399 #ifdef CONFIG_BLK_DEV_INITRD
400 void __init free_initrd_mem(unsigned long start, unsigned long end)
402 free_reserved_area((void *)start, (void *)end, -1, "initrd");
407 * This is called when a page has been modified by the kernel.
408 * It just marks the page as not i-cache clean. We do the i-cache
409 * flush later when the page is given to a user process, if necessary.
411 void flush_dcache_page(struct page *page)
413 if (cpu_has_feature(CPU_FTR_COHERENT_ICACHE))
415 /* avoid an atomic op if possible */
416 if (test_bit(PG_arch_1, &page->flags))
417 clear_bit(PG_arch_1, &page->flags);
419 EXPORT_SYMBOL(flush_dcache_page);
421 void flush_dcache_icache_page(struct page *page)
423 #ifdef CONFIG_HUGETLB_PAGE
424 if (PageCompound(page)) {
425 flush_dcache_icache_hugepage(page);
429 #if defined(CONFIG_PPC_8xx) || defined(CONFIG_PPC64)
430 /* On 8xx there is no need to kmap since highmem is not supported */
431 __flush_dcache_icache(page_address(page));
433 if (IS_ENABLED(CONFIG_BOOKE) || sizeof(phys_addr_t) > sizeof(void *)) {
434 void *start = kmap_atomic(page);
435 __flush_dcache_icache(start);
436 kunmap_atomic(start);
438 __flush_dcache_icache_phys(page_to_pfn(page) << PAGE_SHIFT);
442 EXPORT_SYMBOL(flush_dcache_icache_page);
444 void clear_user_page(void *page, unsigned long vaddr, struct page *pg)
449 * We shouldn't have to do this, but some versions of glibc
450 * require it (ld.so assumes zero filled pages are icache clean)
453 flush_dcache_page(pg);
455 EXPORT_SYMBOL(clear_user_page);
457 void copy_user_page(void *vto, void *vfrom, unsigned long vaddr,
460 copy_page(vto, vfrom);
463 * We should be able to use the following optimisation, however
464 * there are two problems.
465 * Firstly a bug in some versions of binutils meant PLT sections
466 * were not marked executable.
467 * Secondly the first word in the GOT section is blrl, used
468 * to establish the GOT address. Until recently the GOT was
469 * not marked executable.
473 if (!vma->vm_file && ((vma->vm_flags & VM_EXEC) == 0))
477 flush_dcache_page(pg);
480 void flush_icache_user_range(struct vm_area_struct *vma, struct page *page,
481 unsigned long addr, int len)
485 maddr = (unsigned long) kmap(page) + (addr & ~PAGE_MASK);
486 flush_icache_range(maddr, maddr + len);
489 EXPORT_SYMBOL(flush_icache_user_range);
492 * This is called at the end of handling a user page fault, when the
493 * fault has been handled by updating a PTE in the linux page tables.
494 * We use it to preload an HPTE into the hash table corresponding to
495 * the updated linux PTE.
497 * This must always be called with the pte lock held.
499 void update_mmu_cache(struct vm_area_struct *vma, unsigned long address,
502 #ifdef CONFIG_PPC_STD_MMU
504 * We don't need to worry about _PAGE_PRESENT here because we are
505 * called with either mm->page_table_lock held or ptl lock held
507 unsigned long access, trap;
509 if (radix_enabled()) {
510 prefetch((void *)address);
514 /* We only want HPTEs for linux PTEs that have _PAGE_ACCESSED set */
515 if (!pte_young(*ptep) || address >= TASK_SIZE)
518 /* We try to figure out if we are coming from an instruction
519 * access fault and pass that down to __hash_page so we avoid
520 * double-faulting on execution of fresh text. We have to test
521 * for regs NULL since init will get here first thing at boot
523 * We also avoid filling the hash if not coming from a fault
526 trap = current->thread.regs ? TRAP(current->thread.regs) : 0UL;
538 hash_preload(vma->vm_mm, address, access, trap);
539 #endif /* CONFIG_PPC_STD_MMU */
540 #if (defined(CONFIG_PPC_BOOK3E_64) || defined(CONFIG_PPC_FSL_BOOK3E)) \
541 && defined(CONFIG_HUGETLB_PAGE)
542 if (is_vm_hugetlb_page(vma))
543 book3e_hugetlb_preload(vma, address, *ptep);
548 * System memory should not be in /proc/iomem but various tools expect it
551 static int __init add_system_ram_resources(void)
553 struct memblock_region *reg;
555 for_each_memblock(memory, reg) {
556 struct resource *res;
557 unsigned long base = reg->base;
558 unsigned long size = reg->size;
560 res = kzalloc(sizeof(struct resource), GFP_KERNEL);
564 res->name = "System RAM";
566 res->end = base + size - 1;
567 res->flags = IORESOURCE_SYSTEM_RAM | IORESOURCE_BUSY;
568 WARN_ON(request_resource(&iomem_resource, res) < 0);
574 subsys_initcall(add_system_ram_resources);
576 #ifdef CONFIG_STRICT_DEVMEM
578 * devmem_is_allowed(): check to see if /dev/mem access to a certain address
579 * is valid. The argument is a physical page number.
581 * Access has to be given to non-kernel-ram areas as well, these contain the
582 * PCI mmio resources as well as potential bios/acpi data regions.
584 int devmem_is_allowed(unsigned long pfn)
586 if (page_is_rtas_user_buf(pfn))
588 if (iomem_is_exclusive(PFN_PHYS(pfn)))
590 if (!page_is_ram(pfn))
594 #endif /* CONFIG_STRICT_DEVMEM */