12874c5fdSThomas Gleixner // SPDX-License-Identifier: GPL-2.0-or-later 295f72d1eSYinghai Lu /* 395f72d1eSYinghai Lu * Procedures for maintaining information about logical memory blocks. 495f72d1eSYinghai Lu * 595f72d1eSYinghai Lu * Peter Bergner, IBM Corp. June 2001. 695f72d1eSYinghai Lu * Copyright (C) 2001 Peter Bergner. 795f72d1eSYinghai Lu */ 895f72d1eSYinghai Lu 995f72d1eSYinghai Lu #include <linux/kernel.h> 10142b45a7SBenjamin Herrenschmidt #include <linux/slab.h> 1195f72d1eSYinghai Lu #include <linux/init.h> 1295f72d1eSYinghai Lu #include <linux/bitops.h> 13449e8df3SBenjamin Herrenschmidt #include <linux/poison.h> 14c196f76fSBenjamin Herrenschmidt #include <linux/pfn.h> 156d03b885SBenjamin Herrenschmidt #include <linux/debugfs.h> 16514c6032SRandy Dunlap #include <linux/kmemleak.h> 176d03b885SBenjamin Herrenschmidt #include <linux/seq_file.h> 1895f72d1eSYinghai Lu #include <linux/memblock.h> 1995f72d1eSYinghai Lu 20c4c5ad6bSChristoph Hellwig #include <asm/sections.h> 2126f09e9bSSantosh Shilimkar #include <linux/io.h> 2226f09e9bSSantosh Shilimkar 2326f09e9bSSantosh Shilimkar #include "internal.h" 2479442ed1STang Chen 258a5b403dSArd Biesheuvel #define INIT_MEMBLOCK_REGIONS 128 268a5b403dSArd Biesheuvel #define INIT_PHYSMEM_REGIONS 4 278a5b403dSArd Biesheuvel 288a5b403dSArd Biesheuvel #ifndef INIT_MEMBLOCK_RESERVED_REGIONS 298a5b403dSArd Biesheuvel # define INIT_MEMBLOCK_RESERVED_REGIONS INIT_MEMBLOCK_REGIONS 308a5b403dSArd Biesheuvel #endif 318a5b403dSArd Biesheuvel 323e039c5cSMike Rapoport /** 333e039c5cSMike Rapoport * DOC: memblock overview 343e039c5cSMike Rapoport * 353e039c5cSMike Rapoport * Memblock is a method of managing memory regions during the early 363e039c5cSMike Rapoport * boot period when the usual kernel memory allocators are not up and 373e039c5cSMike Rapoport * running. 383e039c5cSMike Rapoport * 393e039c5cSMike Rapoport * Memblock views the system memory as collections of contiguous 403e039c5cSMike Rapoport * regions. There are several types of these collections: 413e039c5cSMike Rapoport * 423e039c5cSMike Rapoport * * ``memory`` - describes the physical memory available to the 433e039c5cSMike Rapoport * kernel; this may differ from the actual physical memory installed 443e039c5cSMike Rapoport * in the system, for instance when the memory is restricted with 453e039c5cSMike Rapoport * ``mem=`` command line parameter 463e039c5cSMike Rapoport * * ``reserved`` - describes the regions that were allocated 4777649905SDavid Hildenbrand * * ``physmem`` - describes the actual physical memory available during 4877649905SDavid Hildenbrand * boot regardless of the possible restrictions and memory hot(un)plug; 4977649905SDavid Hildenbrand * the ``physmem`` type is only available on some architectures. 503e039c5cSMike Rapoport * 513e039c5cSMike Rapoport * Each region is represented by :c:type:`struct memblock_region` that 523e039c5cSMike Rapoport * defines the region extents, its attributes and NUMA node id on NUMA 533e039c5cSMike Rapoport * systems. Every memory type is described by the :c:type:`struct 543e039c5cSMike Rapoport * memblock_type` which contains an array of memory regions along with 5577649905SDavid Hildenbrand * the allocator metadata. The "memory" and "reserved" types are nicely 5677649905SDavid Hildenbrand * wrapped with :c:type:`struct memblock`. This structure is statically 5777649905SDavid Hildenbrand * initialized at build time. The region arrays are initially sized to 5877649905SDavid Hildenbrand * %INIT_MEMBLOCK_REGIONS for "memory" and %INIT_MEMBLOCK_RESERVED_REGIONS 5977649905SDavid Hildenbrand * for "reserved". The region array for "physmem" is initially sized to 6077649905SDavid Hildenbrand * %INIT_PHYSMEM_REGIONS. 616e5af9a8SCao jin * The memblock_allow_resize() enables automatic resizing of the region 626e5af9a8SCao jin * arrays during addition of new regions. This feature should be used 636e5af9a8SCao jin * with care so that memory allocated for the region array will not 646e5af9a8SCao jin * overlap with areas that should be reserved, for example initrd. 653e039c5cSMike Rapoport * 663e039c5cSMike Rapoport * The early architecture setup should tell memblock what the physical 676e5af9a8SCao jin * memory layout is by using memblock_add() or memblock_add_node() 686e5af9a8SCao jin * functions. The first function does not assign the region to a NUMA 696e5af9a8SCao jin * node and it is appropriate for UMA systems. Yet, it is possible to 706e5af9a8SCao jin * use it on NUMA systems as well and assign the region to a NUMA node 716e5af9a8SCao jin * later in the setup process using memblock_set_node(). The 726e5af9a8SCao jin * memblock_add_node() performs such an assignment directly. 733e039c5cSMike Rapoport * 74a2974133SMike Rapoport * Once memblock is setup the memory can be allocated using one of the 75a2974133SMike Rapoport * API variants: 76a2974133SMike Rapoport * 776e5af9a8SCao jin * * memblock_phys_alloc*() - these functions return the **physical** 786e5af9a8SCao jin * address of the allocated memory 796e5af9a8SCao jin * * memblock_alloc*() - these functions return the **virtual** address 806e5af9a8SCao jin * of the allocated memory. 81a2974133SMike Rapoport * 82df1758d9SEthon Paul * Note, that both API variants use implicit assumptions about allowed 83a2974133SMike Rapoport * memory ranges and the fallback methods. Consult the documentation 846e5af9a8SCao jin * of memblock_alloc_internal() and memblock_alloc_range_nid() 856e5af9a8SCao jin * functions for more elaborate description. 863e039c5cSMike Rapoport * 876e5af9a8SCao jin * As the system boot progresses, the architecture specific mem_init() 886e5af9a8SCao jin * function frees all the memory to the buddy page allocator. 893e039c5cSMike Rapoport * 906e5af9a8SCao jin * Unless an architecture enables %CONFIG_ARCH_KEEP_MEMBLOCK, the 9177649905SDavid Hildenbrand * memblock data structures (except "physmem") will be discarded after the 9277649905SDavid Hildenbrand * system initialization completes. 933e039c5cSMike Rapoport */ 943e039c5cSMike Rapoport 95bda49a81SMike Rapoport #ifndef CONFIG_NEED_MULTIPLE_NODES 96bda49a81SMike Rapoport struct pglist_data __refdata contig_page_data; 97bda49a81SMike Rapoport EXPORT_SYMBOL(contig_page_data); 98bda49a81SMike Rapoport #endif 99bda49a81SMike Rapoport 100bda49a81SMike Rapoport unsigned long max_low_pfn; 101bda49a81SMike Rapoport unsigned long min_low_pfn; 102bda49a81SMike Rapoport unsigned long max_pfn; 103bda49a81SMike Rapoport unsigned long long max_possible_pfn; 104bda49a81SMike Rapoport 105fe091c20STejun Heo static struct memblock_region memblock_memory_init_regions[INIT_MEMBLOCK_REGIONS] __initdata_memblock; 1068a5b403dSArd Biesheuvel static struct memblock_region memblock_reserved_init_regions[INIT_MEMBLOCK_RESERVED_REGIONS] __initdata_memblock; 10770210ed9SPhilipp Hachtmann #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP 10877649905SDavid Hildenbrand static struct memblock_region memblock_physmem_init_regions[INIT_PHYSMEM_REGIONS]; 10970210ed9SPhilipp Hachtmann #endif 110fe091c20STejun Heo 111fe091c20STejun Heo struct memblock memblock __initdata_memblock = { 112fe091c20STejun Heo .memory.regions = memblock_memory_init_regions, 113fe091c20STejun Heo .memory.cnt = 1, /* empty dummy entry */ 114fe091c20STejun Heo .memory.max = INIT_MEMBLOCK_REGIONS, 1150262d9c8SHeiko Carstens .memory.name = "memory", 116fe091c20STejun Heo 117fe091c20STejun Heo .reserved.regions = memblock_reserved_init_regions, 118fe091c20STejun Heo .reserved.cnt = 1, /* empty dummy entry */ 1198a5b403dSArd Biesheuvel .reserved.max = INIT_MEMBLOCK_RESERVED_REGIONS, 1200262d9c8SHeiko Carstens .reserved.name = "reserved", 121fe091c20STejun Heo 12279442ed1STang Chen .bottom_up = false, 123fe091c20STejun Heo .current_limit = MEMBLOCK_ALLOC_ANYWHERE, 124fe091c20STejun Heo }; 12595f72d1eSYinghai Lu 12677649905SDavid Hildenbrand #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP 12777649905SDavid Hildenbrand struct memblock_type physmem = { 12877649905SDavid Hildenbrand .regions = memblock_physmem_init_regions, 12977649905SDavid Hildenbrand .cnt = 1, /* empty dummy entry */ 13077649905SDavid Hildenbrand .max = INIT_PHYSMEM_REGIONS, 13177649905SDavid Hildenbrand .name = "physmem", 13277649905SDavid Hildenbrand }; 13377649905SDavid Hildenbrand #endif 13477649905SDavid Hildenbrand 135cd991db8SMike Rapoport #define for_each_memblock_type(i, memblock_type, rgn) \ 136cd991db8SMike Rapoport for (i = 0, rgn = &memblock_type->regions[0]; \ 137cd991db8SMike Rapoport i < memblock_type->cnt; \ 138cd991db8SMike Rapoport i++, rgn = &memblock_type->regions[i]) 139cd991db8SMike Rapoport 140*87c55870SMike Rapoport #define memblock_dbg(fmt, ...) \ 141*87c55870SMike Rapoport do { \ 142*87c55870SMike Rapoport if (memblock_debug) \ 143*87c55870SMike Rapoport pr_info(fmt, ##__VA_ARGS__); \ 144*87c55870SMike Rapoport } while (0) 145*87c55870SMike Rapoport 146*87c55870SMike Rapoport static int memblock_debug __initdata_memblock; 147a3f5bafcSTony Luck static bool system_has_some_mirror __initdata_memblock = false; 1481aadc056STejun Heo static int memblock_can_resize __initdata_memblock; 149181eb394SGavin Shan static int memblock_memory_in_slab __initdata_memblock = 0; 150181eb394SGavin Shan static int memblock_reserved_in_slab __initdata_memblock = 0; 15195f72d1eSYinghai Lu 152c366ea89SMike Rapoport static enum memblock_flags __init_memblock choose_memblock_flags(void) 153a3f5bafcSTony Luck { 154a3f5bafcSTony Luck return system_has_some_mirror ? MEMBLOCK_MIRROR : MEMBLOCK_NONE; 155a3f5bafcSTony Luck } 156a3f5bafcSTony Luck 157eb18f1b5STejun Heo /* adjust *@size so that (@base + *@size) doesn't overflow, return new size */ 158eb18f1b5STejun Heo static inline phys_addr_t memblock_cap_size(phys_addr_t base, phys_addr_t *size) 159eb18f1b5STejun Heo { 1601c4bc43dSStefan Agner return *size = min(*size, PHYS_ADDR_MAX - base); 161eb18f1b5STejun Heo } 162eb18f1b5STejun Heo 1636ed311b2SBenjamin Herrenschmidt /* 1646ed311b2SBenjamin Herrenschmidt * Address comparison utilities 1656ed311b2SBenjamin Herrenschmidt */ 16610d06439SYinghai Lu static unsigned long __init_memblock memblock_addrs_overlap(phys_addr_t base1, phys_addr_t size1, 1672898cc4cSBenjamin Herrenschmidt phys_addr_t base2, phys_addr_t size2) 16895f72d1eSYinghai Lu { 16995f72d1eSYinghai Lu return ((base1 < (base2 + size2)) && (base2 < (base1 + size1))); 17095f72d1eSYinghai Lu } 17195f72d1eSYinghai Lu 17295cf82ecSTang Chen bool __init_memblock memblock_overlaps_region(struct memblock_type *type, 1732d7d3eb2SH Hartley Sweeten phys_addr_t base, phys_addr_t size) 1746ed311b2SBenjamin Herrenschmidt { 1756ed311b2SBenjamin Herrenschmidt unsigned long i; 1766ed311b2SBenjamin Herrenschmidt 177f14516fbSAlexander Kuleshov for (i = 0; i < type->cnt; i++) 178f14516fbSAlexander Kuleshov if (memblock_addrs_overlap(base, size, type->regions[i].base, 179f14516fbSAlexander Kuleshov type->regions[i].size)) 1806ed311b2SBenjamin Herrenschmidt break; 181c5c5c9d1STang Chen return i < type->cnt; 1826ed311b2SBenjamin Herrenschmidt } 1836ed311b2SBenjamin Herrenschmidt 18447cec443SMike Rapoport /** 18579442ed1STang Chen * __memblock_find_range_bottom_up - find free area utility in bottom-up 18679442ed1STang Chen * @start: start of candidate range 18747cec443SMike Rapoport * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or 18847cec443SMike Rapoport * %MEMBLOCK_ALLOC_ACCESSIBLE 18979442ed1STang Chen * @size: size of free area to find 19079442ed1STang Chen * @align: alignment of free area to find 191b1154233SGrygorii Strashko * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 192fc6daaf9STony Luck * @flags: pick from blocks based on memory attributes 19379442ed1STang Chen * 19479442ed1STang Chen * Utility called from memblock_find_in_range_node(), find free area bottom-up. 19579442ed1STang Chen * 19647cec443SMike Rapoport * Return: 19779442ed1STang Chen * Found address on success, 0 on failure. 19879442ed1STang Chen */ 19979442ed1STang Chen static phys_addr_t __init_memblock 20079442ed1STang Chen __memblock_find_range_bottom_up(phys_addr_t start, phys_addr_t end, 201fc6daaf9STony Luck phys_addr_t size, phys_addr_t align, int nid, 202e1720feeSMike Rapoport enum memblock_flags flags) 20379442ed1STang Chen { 20479442ed1STang Chen phys_addr_t this_start, this_end, cand; 20579442ed1STang Chen u64 i; 20679442ed1STang Chen 207fc6daaf9STony Luck for_each_free_mem_range(i, nid, flags, &this_start, &this_end, NULL) { 20879442ed1STang Chen this_start = clamp(this_start, start, end); 20979442ed1STang Chen this_end = clamp(this_end, start, end); 21079442ed1STang Chen 21179442ed1STang Chen cand = round_up(this_start, align); 21279442ed1STang Chen if (cand < this_end && this_end - cand >= size) 21379442ed1STang Chen return cand; 21479442ed1STang Chen } 21579442ed1STang Chen 21679442ed1STang Chen return 0; 21779442ed1STang Chen } 21879442ed1STang Chen 2197bd0b0f0STejun Heo /** 2201402899eSTang Chen * __memblock_find_range_top_down - find free area utility, in top-down 2211402899eSTang Chen * @start: start of candidate range 22247cec443SMike Rapoport * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or 22347cec443SMike Rapoport * %MEMBLOCK_ALLOC_ACCESSIBLE 2241402899eSTang Chen * @size: size of free area to find 2251402899eSTang Chen * @align: alignment of free area to find 226b1154233SGrygorii Strashko * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 227fc6daaf9STony Luck * @flags: pick from blocks based on memory attributes 2281402899eSTang Chen * 2291402899eSTang Chen * Utility called from memblock_find_in_range_node(), find free area top-down. 2301402899eSTang Chen * 23147cec443SMike Rapoport * Return: 23279442ed1STang Chen * Found address on success, 0 on failure. 2331402899eSTang Chen */ 2341402899eSTang Chen static phys_addr_t __init_memblock 2351402899eSTang Chen __memblock_find_range_top_down(phys_addr_t start, phys_addr_t end, 236fc6daaf9STony Luck phys_addr_t size, phys_addr_t align, int nid, 237e1720feeSMike Rapoport enum memblock_flags flags) 2381402899eSTang Chen { 2391402899eSTang Chen phys_addr_t this_start, this_end, cand; 2401402899eSTang Chen u64 i; 2411402899eSTang Chen 242fc6daaf9STony Luck for_each_free_mem_range_reverse(i, nid, flags, &this_start, &this_end, 243fc6daaf9STony Luck NULL) { 2441402899eSTang Chen this_start = clamp(this_start, start, end); 2451402899eSTang Chen this_end = clamp(this_end, start, end); 2461402899eSTang Chen 2471402899eSTang Chen if (this_end < size) 2481402899eSTang Chen continue; 2491402899eSTang Chen 2501402899eSTang Chen cand = round_down(this_end - size, align); 2511402899eSTang Chen if (cand >= this_start) 2521402899eSTang Chen return cand; 2531402899eSTang Chen } 2541402899eSTang Chen 2551402899eSTang Chen return 0; 2561402899eSTang Chen } 2571402899eSTang Chen 2581402899eSTang Chen /** 2597bd0b0f0STejun Heo * memblock_find_in_range_node - find free area in given range and node 2607bd0b0f0STejun Heo * @size: size of free area to find 2617bd0b0f0STejun Heo * @align: alignment of free area to find 26287029ee9SGrygorii Strashko * @start: start of candidate range 26347cec443SMike Rapoport * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or 26447cec443SMike Rapoport * %MEMBLOCK_ALLOC_ACCESSIBLE 265b1154233SGrygorii Strashko * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 266fc6daaf9STony Luck * @flags: pick from blocks based on memory attributes 2677bd0b0f0STejun Heo * 2687bd0b0f0STejun Heo * Find @size free area aligned to @align in the specified range and node. 2697bd0b0f0STejun Heo * 27079442ed1STang Chen * When allocation direction is bottom-up, the @start should be greater 27179442ed1STang Chen * than the end of the kernel image. Otherwise, it will be trimmed. The 27279442ed1STang Chen * reason is that we want the bottom-up allocation just near the kernel 27379442ed1STang Chen * image so it is highly likely that the allocated memory and the kernel 27479442ed1STang Chen * will reside in the same node. 27579442ed1STang Chen * 27679442ed1STang Chen * If bottom-up allocation failed, will try to allocate memory top-down. 27779442ed1STang Chen * 27847cec443SMike Rapoport * Return: 27979442ed1STang Chen * Found address on success, 0 on failure. 2806ed311b2SBenjamin Herrenschmidt */ 281c366ea89SMike Rapoport static phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, 28287029ee9SGrygorii Strashko phys_addr_t align, phys_addr_t start, 283e1720feeSMike Rapoport phys_addr_t end, int nid, 284e1720feeSMike Rapoport enum memblock_flags flags) 285f7210e6cSTang Chen { 2860cfb8f0cSTang Chen phys_addr_t kernel_end, ret; 28779442ed1STang Chen 288f7210e6cSTang Chen /* pump up @end */ 289fed84c78SQian Cai if (end == MEMBLOCK_ALLOC_ACCESSIBLE || 290fed84c78SQian Cai end == MEMBLOCK_ALLOC_KASAN) 291f7210e6cSTang Chen end = memblock.current_limit; 292f7210e6cSTang Chen 293f7210e6cSTang Chen /* avoid allocating the first page */ 294f7210e6cSTang Chen start = max_t(phys_addr_t, start, PAGE_SIZE); 295f7210e6cSTang Chen end = max(start, end); 29679442ed1STang Chen kernel_end = __pa_symbol(_end); 29779442ed1STang Chen 29879442ed1STang Chen /* 29979442ed1STang Chen * try bottom-up allocation only when bottom-up mode 30079442ed1STang Chen * is set and @end is above the kernel image. 30179442ed1STang Chen */ 30279442ed1STang Chen if (memblock_bottom_up() && end > kernel_end) { 30379442ed1STang Chen phys_addr_t bottom_up_start; 30479442ed1STang Chen 30579442ed1STang Chen /* make sure we will allocate above the kernel */ 30679442ed1STang Chen bottom_up_start = max(start, kernel_end); 30779442ed1STang Chen 30879442ed1STang Chen /* ok, try bottom-up allocation first */ 30979442ed1STang Chen ret = __memblock_find_range_bottom_up(bottom_up_start, end, 310fc6daaf9STony Luck size, align, nid, flags); 31179442ed1STang Chen if (ret) 31279442ed1STang Chen return ret; 31379442ed1STang Chen 31479442ed1STang Chen /* 31579442ed1STang Chen * we always limit bottom-up allocation above the kernel, 31679442ed1STang Chen * but top-down allocation doesn't have the limit, so 31779442ed1STang Chen * retrying top-down allocation may succeed when bottom-up 31879442ed1STang Chen * allocation failed. 31979442ed1STang Chen * 32079442ed1STang Chen * bottom-up allocation is expected to be fail very rarely, 32179442ed1STang Chen * so we use WARN_ONCE() here to see the stack trace if 32279442ed1STang Chen * fail happens. 32379442ed1STang Chen */ 324e3d301caSMichal Hocko WARN_ONCE(IS_ENABLED(CONFIG_MEMORY_HOTREMOVE), 325e3d301caSMichal Hocko "memblock: bottom-up allocation failed, memory hotremove may be affected\n"); 32679442ed1STang Chen } 327f7210e6cSTang Chen 328fc6daaf9STony Luck return __memblock_find_range_top_down(start, end, size, align, nid, 329fc6daaf9STony Luck flags); 330f7210e6cSTang Chen } 3316ed311b2SBenjamin Herrenschmidt 3327bd0b0f0STejun Heo /** 3337bd0b0f0STejun Heo * memblock_find_in_range - find free area in given range 3347bd0b0f0STejun Heo * @start: start of candidate range 33547cec443SMike Rapoport * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or 33647cec443SMike Rapoport * %MEMBLOCK_ALLOC_ACCESSIBLE 3377bd0b0f0STejun Heo * @size: size of free area to find 3387bd0b0f0STejun Heo * @align: alignment of free area to find 3397bd0b0f0STejun Heo * 3407bd0b0f0STejun Heo * Find @size free area aligned to @align in the specified range. 3417bd0b0f0STejun Heo * 34247cec443SMike Rapoport * Return: 34379442ed1STang Chen * Found address on success, 0 on failure. 3447bd0b0f0STejun Heo */ 3457bd0b0f0STejun Heo phys_addr_t __init_memblock memblock_find_in_range(phys_addr_t start, 3467bd0b0f0STejun Heo phys_addr_t end, phys_addr_t size, 3477bd0b0f0STejun Heo phys_addr_t align) 3487bd0b0f0STejun Heo { 349a3f5bafcSTony Luck phys_addr_t ret; 350e1720feeSMike Rapoport enum memblock_flags flags = choose_memblock_flags(); 351a3f5bafcSTony Luck 352a3f5bafcSTony Luck again: 353a3f5bafcSTony Luck ret = memblock_find_in_range_node(size, align, start, end, 354a3f5bafcSTony Luck NUMA_NO_NODE, flags); 355a3f5bafcSTony Luck 356a3f5bafcSTony Luck if (!ret && (flags & MEMBLOCK_MIRROR)) { 357a3f5bafcSTony Luck pr_warn("Could not allocate %pap bytes of mirrored memory\n", 358a3f5bafcSTony Luck &size); 359a3f5bafcSTony Luck flags &= ~MEMBLOCK_MIRROR; 360a3f5bafcSTony Luck goto again; 361a3f5bafcSTony Luck } 362a3f5bafcSTony Luck 363a3f5bafcSTony Luck return ret; 3647bd0b0f0STejun Heo } 3657bd0b0f0STejun Heo 36610d06439SYinghai Lu static void __init_memblock memblock_remove_region(struct memblock_type *type, unsigned long r) 36795f72d1eSYinghai Lu { 3681440c4e2STejun Heo type->total_size -= type->regions[r].size; 3697c0caeb8STejun Heo memmove(&type->regions[r], &type->regions[r + 1], 3707c0caeb8STejun Heo (type->cnt - (r + 1)) * sizeof(type->regions[r])); 371e3239ff9SBenjamin Herrenschmidt type->cnt--; 37295f72d1eSYinghai Lu 3738f7a6605SBenjamin Herrenschmidt /* Special case for empty arrays */ 3748f7a6605SBenjamin Herrenschmidt if (type->cnt == 0) { 3751440c4e2STejun Heo WARN_ON(type->total_size != 0); 3768f7a6605SBenjamin Herrenschmidt type->cnt = 1; 3778f7a6605SBenjamin Herrenschmidt type->regions[0].base = 0; 3788f7a6605SBenjamin Herrenschmidt type->regions[0].size = 0; 37966a20757STang Chen type->regions[0].flags = 0; 3807c0caeb8STejun Heo memblock_set_region_node(&type->regions[0], MAX_NUMNODES); 3818f7a6605SBenjamin Herrenschmidt } 38295f72d1eSYinghai Lu } 38395f72d1eSYinghai Lu 384350e88baSMike Rapoport #ifndef CONFIG_ARCH_KEEP_MEMBLOCK 3853010f876SPavel Tatashin /** 38647cec443SMike Rapoport * memblock_discard - discard memory and reserved arrays if they were allocated 3873010f876SPavel Tatashin */ 3883010f876SPavel Tatashin void __init memblock_discard(void) 38929f67386SYinghai Lu { 3903010f876SPavel Tatashin phys_addr_t addr, size; 39129f67386SYinghai Lu 3923010f876SPavel Tatashin if (memblock.reserved.regions != memblock_reserved_init_regions) { 3933010f876SPavel Tatashin addr = __pa(memblock.reserved.regions); 3943010f876SPavel Tatashin size = PAGE_ALIGN(sizeof(struct memblock_region) * 39529f67386SYinghai Lu memblock.reserved.max); 3963010f876SPavel Tatashin __memblock_free_late(addr, size); 39729f67386SYinghai Lu } 39829f67386SYinghai Lu 39991b540f9SPavel Tatashin if (memblock.memory.regions != memblock_memory_init_regions) { 4003010f876SPavel Tatashin addr = __pa(memblock.memory.regions); 4013010f876SPavel Tatashin size = PAGE_ALIGN(sizeof(struct memblock_region) * 4025e270e25SPhilipp Hachtmann memblock.memory.max); 4033010f876SPavel Tatashin __memblock_free_late(addr, size); 4045e270e25SPhilipp Hachtmann } 4053010f876SPavel Tatashin } 4065e270e25SPhilipp Hachtmann #endif 4075e270e25SPhilipp Hachtmann 40848c3b583SGreg Pearson /** 40948c3b583SGreg Pearson * memblock_double_array - double the size of the memblock regions array 41048c3b583SGreg Pearson * @type: memblock type of the regions array being doubled 41148c3b583SGreg Pearson * @new_area_start: starting address of memory range to avoid overlap with 41248c3b583SGreg Pearson * @new_area_size: size of memory range to avoid overlap with 41348c3b583SGreg Pearson * 41448c3b583SGreg Pearson * Double the size of the @type regions array. If memblock is being used to 41548c3b583SGreg Pearson * allocate memory for a new reserved regions array and there is a previously 41648c3b583SGreg Pearson * allocated memory range [@new_area_start, @new_area_start + @new_area_size] 41748c3b583SGreg Pearson * waiting to be reserved, ensure the memory used by the new array does 41848c3b583SGreg Pearson * not overlap. 41948c3b583SGreg Pearson * 42047cec443SMike Rapoport * Return: 42148c3b583SGreg Pearson * 0 on success, -1 on failure. 42248c3b583SGreg Pearson */ 42348c3b583SGreg Pearson static int __init_memblock memblock_double_array(struct memblock_type *type, 42448c3b583SGreg Pearson phys_addr_t new_area_start, 42548c3b583SGreg Pearson phys_addr_t new_area_size) 426142b45a7SBenjamin Herrenschmidt { 427142b45a7SBenjamin Herrenschmidt struct memblock_region *new_array, *old_array; 42829f67386SYinghai Lu phys_addr_t old_alloc_size, new_alloc_size; 429a36aab89SMike Rapoport phys_addr_t old_size, new_size, addr, new_end; 430142b45a7SBenjamin Herrenschmidt int use_slab = slab_is_available(); 431181eb394SGavin Shan int *in_slab; 432142b45a7SBenjamin Herrenschmidt 433142b45a7SBenjamin Herrenschmidt /* We don't allow resizing until we know about the reserved regions 434142b45a7SBenjamin Herrenschmidt * of memory that aren't suitable for allocation 435142b45a7SBenjamin Herrenschmidt */ 436142b45a7SBenjamin Herrenschmidt if (!memblock_can_resize) 437142b45a7SBenjamin Herrenschmidt return -1; 438142b45a7SBenjamin Herrenschmidt 439142b45a7SBenjamin Herrenschmidt /* Calculate new doubled size */ 440142b45a7SBenjamin Herrenschmidt old_size = type->max * sizeof(struct memblock_region); 441142b45a7SBenjamin Herrenschmidt new_size = old_size << 1; 44229f67386SYinghai Lu /* 44329f67386SYinghai Lu * We need to allocated new one align to PAGE_SIZE, 44429f67386SYinghai Lu * so we can free them completely later. 44529f67386SYinghai Lu */ 44629f67386SYinghai Lu old_alloc_size = PAGE_ALIGN(old_size); 44729f67386SYinghai Lu new_alloc_size = PAGE_ALIGN(new_size); 448142b45a7SBenjamin Herrenschmidt 449181eb394SGavin Shan /* Retrieve the slab flag */ 450181eb394SGavin Shan if (type == &memblock.memory) 451181eb394SGavin Shan in_slab = &memblock_memory_in_slab; 452181eb394SGavin Shan else 453181eb394SGavin Shan in_slab = &memblock_reserved_in_slab; 454181eb394SGavin Shan 455a2974133SMike Rapoport /* Try to find some space for it */ 456142b45a7SBenjamin Herrenschmidt if (use_slab) { 457142b45a7SBenjamin Herrenschmidt new_array = kmalloc(new_size, GFP_KERNEL); 4581f5026a7STejun Heo addr = new_array ? __pa(new_array) : 0; 4594e2f0775SGavin Shan } else { 46048c3b583SGreg Pearson /* only exclude range when trying to double reserved.regions */ 46148c3b583SGreg Pearson if (type != &memblock.reserved) 46248c3b583SGreg Pearson new_area_start = new_area_size = 0; 46348c3b583SGreg Pearson 46448c3b583SGreg Pearson addr = memblock_find_in_range(new_area_start + new_area_size, 46548c3b583SGreg Pearson memblock.current_limit, 46629f67386SYinghai Lu new_alloc_size, PAGE_SIZE); 46748c3b583SGreg Pearson if (!addr && new_area_size) 46848c3b583SGreg Pearson addr = memblock_find_in_range(0, 46948c3b583SGreg Pearson min(new_area_start, memblock.current_limit), 47029f67386SYinghai Lu new_alloc_size, PAGE_SIZE); 47148c3b583SGreg Pearson 47215674868SSachin Kamat new_array = addr ? __va(addr) : NULL; 4734e2f0775SGavin Shan } 4741f5026a7STejun Heo if (!addr) { 475142b45a7SBenjamin Herrenschmidt pr_err("memblock: Failed to double %s array from %ld to %ld entries !\n", 4760262d9c8SHeiko Carstens type->name, type->max, type->max * 2); 477142b45a7SBenjamin Herrenschmidt return -1; 478142b45a7SBenjamin Herrenschmidt } 479142b45a7SBenjamin Herrenschmidt 480a36aab89SMike Rapoport new_end = addr + new_size - 1; 481a36aab89SMike Rapoport memblock_dbg("memblock: %s is doubled to %ld at [%pa-%pa]", 482a36aab89SMike Rapoport type->name, type->max * 2, &addr, &new_end); 483ea9e4376SYinghai Lu 484fd07383bSAndrew Morton /* 485fd07383bSAndrew Morton * Found space, we now need to move the array over before we add the 486fd07383bSAndrew Morton * reserved region since it may be our reserved array itself that is 487fd07383bSAndrew Morton * full. 488142b45a7SBenjamin Herrenschmidt */ 489142b45a7SBenjamin Herrenschmidt memcpy(new_array, type->regions, old_size); 490142b45a7SBenjamin Herrenschmidt memset(new_array + type->max, 0, old_size); 491142b45a7SBenjamin Herrenschmidt old_array = type->regions; 492142b45a7SBenjamin Herrenschmidt type->regions = new_array; 493142b45a7SBenjamin Herrenschmidt type->max <<= 1; 494142b45a7SBenjamin Herrenschmidt 495fd07383bSAndrew Morton /* Free old array. We needn't free it if the array is the static one */ 496181eb394SGavin Shan if (*in_slab) 497181eb394SGavin Shan kfree(old_array); 498181eb394SGavin Shan else if (old_array != memblock_memory_init_regions && 499142b45a7SBenjamin Herrenschmidt old_array != memblock_reserved_init_regions) 50029f67386SYinghai Lu memblock_free(__pa(old_array), old_alloc_size); 501142b45a7SBenjamin Herrenschmidt 502fd07383bSAndrew Morton /* 503fd07383bSAndrew Morton * Reserve the new array if that comes from the memblock. Otherwise, we 504fd07383bSAndrew Morton * needn't do it 505181eb394SGavin Shan */ 506181eb394SGavin Shan if (!use_slab) 50729f67386SYinghai Lu BUG_ON(memblock_reserve(addr, new_alloc_size)); 508181eb394SGavin Shan 509181eb394SGavin Shan /* Update slab flag */ 510181eb394SGavin Shan *in_slab = use_slab; 511181eb394SGavin Shan 512142b45a7SBenjamin Herrenschmidt return 0; 513142b45a7SBenjamin Herrenschmidt } 514142b45a7SBenjamin Herrenschmidt 515784656f9STejun Heo /** 516784656f9STejun Heo * memblock_merge_regions - merge neighboring compatible regions 517784656f9STejun Heo * @type: memblock type to scan 518784656f9STejun Heo * 519784656f9STejun Heo * Scan @type and merge neighboring compatible regions. 520784656f9STejun Heo */ 521784656f9STejun Heo static void __init_memblock memblock_merge_regions(struct memblock_type *type) 522784656f9STejun Heo { 523784656f9STejun Heo int i = 0; 524784656f9STejun Heo 525784656f9STejun Heo /* cnt never goes below 1 */ 526784656f9STejun Heo while (i < type->cnt - 1) { 527784656f9STejun Heo struct memblock_region *this = &type->regions[i]; 528784656f9STejun Heo struct memblock_region *next = &type->regions[i + 1]; 529784656f9STejun Heo 5307c0caeb8STejun Heo if (this->base + this->size != next->base || 5317c0caeb8STejun Heo memblock_get_region_node(this) != 53266a20757STang Chen memblock_get_region_node(next) || 53366a20757STang Chen this->flags != next->flags) { 534784656f9STejun Heo BUG_ON(this->base + this->size > next->base); 535784656f9STejun Heo i++; 536784656f9STejun Heo continue; 537784656f9STejun Heo } 538784656f9STejun Heo 539784656f9STejun Heo this->size += next->size; 540c0232ae8SLin Feng /* move forward from next + 1, index of which is i + 2 */ 541c0232ae8SLin Feng memmove(next, next + 1, (type->cnt - (i + 2)) * sizeof(*next)); 542784656f9STejun Heo type->cnt--; 543784656f9STejun Heo } 544784656f9STejun Heo } 545784656f9STejun Heo 546784656f9STejun Heo /** 547784656f9STejun Heo * memblock_insert_region - insert new memblock region 548784656f9STejun Heo * @type: memblock type to insert into 549784656f9STejun Heo * @idx: index for the insertion point 550784656f9STejun Heo * @base: base address of the new region 551784656f9STejun Heo * @size: size of the new region 552209ff86dSTang Chen * @nid: node id of the new region 55366a20757STang Chen * @flags: flags of the new region 554784656f9STejun Heo * 555784656f9STejun Heo * Insert new memblock region [@base, @base + @size) into @type at @idx. 556412d0008SAlexander Kuleshov * @type must already have extra room to accommodate the new region. 557784656f9STejun Heo */ 558784656f9STejun Heo static void __init_memblock memblock_insert_region(struct memblock_type *type, 559784656f9STejun Heo int idx, phys_addr_t base, 56066a20757STang Chen phys_addr_t size, 561e1720feeSMike Rapoport int nid, 562e1720feeSMike Rapoport enum memblock_flags flags) 563784656f9STejun Heo { 564784656f9STejun Heo struct memblock_region *rgn = &type->regions[idx]; 565784656f9STejun Heo 566784656f9STejun Heo BUG_ON(type->cnt >= type->max); 567784656f9STejun Heo memmove(rgn + 1, rgn, (type->cnt - idx) * sizeof(*rgn)); 568784656f9STejun Heo rgn->base = base; 569784656f9STejun Heo rgn->size = size; 57066a20757STang Chen rgn->flags = flags; 5717c0caeb8STejun Heo memblock_set_region_node(rgn, nid); 572784656f9STejun Heo type->cnt++; 5731440c4e2STejun Heo type->total_size += size; 574784656f9STejun Heo } 575784656f9STejun Heo 576784656f9STejun Heo /** 577f1af9d3aSPhilipp Hachtmann * memblock_add_range - add new memblock region 578784656f9STejun Heo * @type: memblock type to add new region into 579784656f9STejun Heo * @base: base address of the new region 580784656f9STejun Heo * @size: size of the new region 5817fb0bc3fSTejun Heo * @nid: nid of the new region 58266a20757STang Chen * @flags: flags of the new region 583784656f9STejun Heo * 584784656f9STejun Heo * Add new memblock region [@base, @base + @size) into @type. The new region 585784656f9STejun Heo * is allowed to overlap with existing ones - overlaps don't affect already 586784656f9STejun Heo * existing regions. @type is guaranteed to be minimal (all neighbouring 587784656f9STejun Heo * compatible regions are merged) after the addition. 588784656f9STejun Heo * 58947cec443SMike Rapoport * Return: 590784656f9STejun Heo * 0 on success, -errno on failure. 591784656f9STejun Heo */ 59202634a44SAnshuman Khandual static int __init_memblock memblock_add_range(struct memblock_type *type, 59366a20757STang Chen phys_addr_t base, phys_addr_t size, 594e1720feeSMike Rapoport int nid, enum memblock_flags flags) 59595f72d1eSYinghai Lu { 596784656f9STejun Heo bool insert = false; 597eb18f1b5STejun Heo phys_addr_t obase = base; 598eb18f1b5STejun Heo phys_addr_t end = base + memblock_cap_size(base, &size); 5998c9c1701SAlexander Kuleshov int idx, nr_new; 6008c9c1701SAlexander Kuleshov struct memblock_region *rgn; 60195f72d1eSYinghai Lu 602b3dc627cSTejun Heo if (!size) 603b3dc627cSTejun Heo return 0; 604b3dc627cSTejun Heo 605784656f9STejun Heo /* special case for empty array */ 606784656f9STejun Heo if (type->regions[0].size == 0) { 6071440c4e2STejun Heo WARN_ON(type->cnt != 1 || type->total_size); 608784656f9STejun Heo type->regions[0].base = base; 609784656f9STejun Heo type->regions[0].size = size; 61066a20757STang Chen type->regions[0].flags = flags; 6117fb0bc3fSTejun Heo memblock_set_region_node(&type->regions[0], nid); 6121440c4e2STejun Heo type->total_size = size; 613784656f9STejun Heo return 0; 614784656f9STejun Heo } 615784656f9STejun Heo repeat: 616784656f9STejun Heo /* 617784656f9STejun Heo * The following is executed twice. Once with %false @insert and 618784656f9STejun Heo * then with %true. The first counts the number of regions needed 619412d0008SAlexander Kuleshov * to accommodate the new area. The second actually inserts them. 620784656f9STejun Heo */ 621784656f9STejun Heo base = obase; 622784656f9STejun Heo nr_new = 0; 623784656f9STejun Heo 62466e8b438SGioh Kim for_each_memblock_type(idx, type, rgn) { 625784656f9STejun Heo phys_addr_t rbase = rgn->base; 626784656f9STejun Heo phys_addr_t rend = rbase + rgn->size; 6278f7a6605SBenjamin Herrenschmidt 628784656f9STejun Heo if (rbase >= end) 6298f7a6605SBenjamin Herrenschmidt break; 630784656f9STejun Heo if (rend <= base) 631784656f9STejun Heo continue; 632784656f9STejun Heo /* 633784656f9STejun Heo * @rgn overlaps. If it separates the lower part of new 634784656f9STejun Heo * area, insert that portion. 6358f7a6605SBenjamin Herrenschmidt */ 636784656f9STejun Heo if (rbase > base) { 6373f08a302SMike Rapoport #ifdef CONFIG_NEED_MULTIPLE_NODES 638c0a29498SWei Yang WARN_ON(nid != memblock_get_region_node(rgn)); 639c0a29498SWei Yang #endif 6404fcab5f4SWei Yang WARN_ON(flags != rgn->flags); 641784656f9STejun Heo nr_new++; 642784656f9STejun Heo if (insert) 6438c9c1701SAlexander Kuleshov memblock_insert_region(type, idx++, base, 64466a20757STang Chen rbase - base, nid, 64566a20757STang Chen flags); 646784656f9STejun Heo } 647784656f9STejun Heo /* area below @rend is dealt with, forget about it */ 648784656f9STejun Heo base = min(rend, end); 6498f7a6605SBenjamin Herrenschmidt } 6508f7a6605SBenjamin Herrenschmidt 651784656f9STejun Heo /* insert the remaining portion */ 652784656f9STejun Heo if (base < end) { 653784656f9STejun Heo nr_new++; 654784656f9STejun Heo if (insert) 6558c9c1701SAlexander Kuleshov memblock_insert_region(type, idx, base, end - base, 65666a20757STang Chen nid, flags); 6578f7a6605SBenjamin Herrenschmidt } 6588f7a6605SBenjamin Herrenschmidt 659ef3cc4dbSnimisolo if (!nr_new) 660ef3cc4dbSnimisolo return 0; 661ef3cc4dbSnimisolo 662784656f9STejun Heo /* 663784656f9STejun Heo * If this was the first round, resize array and repeat for actual 664784656f9STejun Heo * insertions; otherwise, merge and return. 6658f7a6605SBenjamin Herrenschmidt */ 666784656f9STejun Heo if (!insert) { 667784656f9STejun Heo while (type->cnt + nr_new > type->max) 66848c3b583SGreg Pearson if (memblock_double_array(type, obase, size) < 0) 669784656f9STejun Heo return -ENOMEM; 670784656f9STejun Heo insert = true; 671784656f9STejun Heo goto repeat; 67295f72d1eSYinghai Lu } else { 673784656f9STejun Heo memblock_merge_regions(type); 67495f72d1eSYinghai Lu return 0; 67595f72d1eSYinghai Lu } 676784656f9STejun Heo } 67795f72d1eSYinghai Lu 67848a833ccSMike Rapoport /** 67948a833ccSMike Rapoport * memblock_add_node - add new memblock region within a NUMA node 68048a833ccSMike Rapoport * @base: base address of the new region 68148a833ccSMike Rapoport * @size: size of the new region 68248a833ccSMike Rapoport * @nid: nid of the new region 68348a833ccSMike Rapoport * 68448a833ccSMike Rapoport * Add new memblock region [@base, @base + @size) to the "memory" 68548a833ccSMike Rapoport * type. See memblock_add_range() description for mode details 68648a833ccSMike Rapoport * 68748a833ccSMike Rapoport * Return: 68848a833ccSMike Rapoport * 0 on success, -errno on failure. 68948a833ccSMike Rapoport */ 6907fb0bc3fSTejun Heo int __init_memblock memblock_add_node(phys_addr_t base, phys_addr_t size, 6917fb0bc3fSTejun Heo int nid) 6927fb0bc3fSTejun Heo { 693f1af9d3aSPhilipp Hachtmann return memblock_add_range(&memblock.memory, base, size, nid, 0); 6947fb0bc3fSTejun Heo } 6957fb0bc3fSTejun Heo 69648a833ccSMike Rapoport /** 69748a833ccSMike Rapoport * memblock_add - add new memblock region 69848a833ccSMike Rapoport * @base: base address of the new region 69948a833ccSMike Rapoport * @size: size of the new region 70048a833ccSMike Rapoport * 70148a833ccSMike Rapoport * Add new memblock region [@base, @base + @size) to the "memory" 70248a833ccSMike Rapoport * type. See memblock_add_range() description for mode details 70348a833ccSMike Rapoport * 70448a833ccSMike Rapoport * Return: 70548a833ccSMike Rapoport * 0 on success, -errno on failure. 70648a833ccSMike Rapoport */ 707f705ac4bSAlexander Kuleshov int __init_memblock memblock_add(phys_addr_t base, phys_addr_t size) 7086a4055bcSAlexander Kuleshov { 7095d63f81cSMiles Chen phys_addr_t end = base + size - 1; 7105d63f81cSMiles Chen 711a090d711SAnshuman Khandual memblock_dbg("%s: [%pa-%pa] %pS\n", __func__, 7125d63f81cSMiles Chen &base, &end, (void *)_RET_IP_); 7136a4055bcSAlexander Kuleshov 714f705ac4bSAlexander Kuleshov return memblock_add_range(&memblock.memory, base, size, MAX_NUMNODES, 0); 71595f72d1eSYinghai Lu } 71695f72d1eSYinghai Lu 7176a9ceb31STejun Heo /** 7186a9ceb31STejun Heo * memblock_isolate_range - isolate given range into disjoint memblocks 7196a9ceb31STejun Heo * @type: memblock type to isolate range for 7206a9ceb31STejun Heo * @base: base of range to isolate 7216a9ceb31STejun Heo * @size: size of range to isolate 7226a9ceb31STejun Heo * @start_rgn: out parameter for the start of isolated region 7236a9ceb31STejun Heo * @end_rgn: out parameter for the end of isolated region 7246a9ceb31STejun Heo * 7256a9ceb31STejun Heo * Walk @type and ensure that regions don't cross the boundaries defined by 7266a9ceb31STejun Heo * [@base, @base + @size). Crossing regions are split at the boundaries, 7276a9ceb31STejun Heo * which may create at most two more regions. The index of the first 7286a9ceb31STejun Heo * region inside the range is returned in *@start_rgn and end in *@end_rgn. 7296a9ceb31STejun Heo * 73047cec443SMike Rapoport * Return: 7316a9ceb31STejun Heo * 0 on success, -errno on failure. 7326a9ceb31STejun Heo */ 7336a9ceb31STejun Heo static int __init_memblock memblock_isolate_range(struct memblock_type *type, 7346a9ceb31STejun Heo phys_addr_t base, phys_addr_t size, 7356a9ceb31STejun Heo int *start_rgn, int *end_rgn) 7366a9ceb31STejun Heo { 737eb18f1b5STejun Heo phys_addr_t end = base + memblock_cap_size(base, &size); 7388c9c1701SAlexander Kuleshov int idx; 7398c9c1701SAlexander Kuleshov struct memblock_region *rgn; 7406a9ceb31STejun Heo 7416a9ceb31STejun Heo *start_rgn = *end_rgn = 0; 7426a9ceb31STejun Heo 743b3dc627cSTejun Heo if (!size) 744b3dc627cSTejun Heo return 0; 745b3dc627cSTejun Heo 7466a9ceb31STejun Heo /* we'll create at most two more regions */ 7476a9ceb31STejun Heo while (type->cnt + 2 > type->max) 74848c3b583SGreg Pearson if (memblock_double_array(type, base, size) < 0) 7496a9ceb31STejun Heo return -ENOMEM; 7506a9ceb31STejun Heo 75166e8b438SGioh Kim for_each_memblock_type(idx, type, rgn) { 7526a9ceb31STejun Heo phys_addr_t rbase = rgn->base; 7536a9ceb31STejun Heo phys_addr_t rend = rbase + rgn->size; 7546a9ceb31STejun Heo 7556a9ceb31STejun Heo if (rbase >= end) 7566a9ceb31STejun Heo break; 7576a9ceb31STejun Heo if (rend <= base) 7586a9ceb31STejun Heo continue; 7596a9ceb31STejun Heo 7606a9ceb31STejun Heo if (rbase < base) { 7616a9ceb31STejun Heo /* 7626a9ceb31STejun Heo * @rgn intersects from below. Split and continue 7636a9ceb31STejun Heo * to process the next region - the new top half. 7646a9ceb31STejun Heo */ 7656a9ceb31STejun Heo rgn->base = base; 7661440c4e2STejun Heo rgn->size -= base - rbase; 7671440c4e2STejun Heo type->total_size -= base - rbase; 7688c9c1701SAlexander Kuleshov memblock_insert_region(type, idx, rbase, base - rbase, 76966a20757STang Chen memblock_get_region_node(rgn), 77066a20757STang Chen rgn->flags); 7716a9ceb31STejun Heo } else if (rend > end) { 7726a9ceb31STejun Heo /* 7736a9ceb31STejun Heo * @rgn intersects from above. Split and redo the 7746a9ceb31STejun Heo * current region - the new bottom half. 7756a9ceb31STejun Heo */ 7766a9ceb31STejun Heo rgn->base = end; 7771440c4e2STejun Heo rgn->size -= end - rbase; 7781440c4e2STejun Heo type->total_size -= end - rbase; 7798c9c1701SAlexander Kuleshov memblock_insert_region(type, idx--, rbase, end - rbase, 78066a20757STang Chen memblock_get_region_node(rgn), 78166a20757STang Chen rgn->flags); 7826a9ceb31STejun Heo } else { 7836a9ceb31STejun Heo /* @rgn is fully contained, record it */ 7846a9ceb31STejun Heo if (!*end_rgn) 7858c9c1701SAlexander Kuleshov *start_rgn = idx; 7868c9c1701SAlexander Kuleshov *end_rgn = idx + 1; 7876a9ceb31STejun Heo } 7886a9ceb31STejun Heo } 7896a9ceb31STejun Heo 7906a9ceb31STejun Heo return 0; 7916a9ceb31STejun Heo } 7926a9ceb31STejun Heo 79335bd16a2SAlexander Kuleshov static int __init_memblock memblock_remove_range(struct memblock_type *type, 7948f7a6605SBenjamin Herrenschmidt phys_addr_t base, phys_addr_t size) 79595f72d1eSYinghai Lu { 79671936180STejun Heo int start_rgn, end_rgn; 79771936180STejun Heo int i, ret; 79895f72d1eSYinghai Lu 79971936180STejun Heo ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); 80071936180STejun Heo if (ret) 80171936180STejun Heo return ret; 80295f72d1eSYinghai Lu 80371936180STejun Heo for (i = end_rgn - 1; i >= start_rgn; i--) 80471936180STejun Heo memblock_remove_region(type, i); 80595f72d1eSYinghai Lu return 0; 80695f72d1eSYinghai Lu } 80795f72d1eSYinghai Lu 808581adcbeSTejun Heo int __init_memblock memblock_remove(phys_addr_t base, phys_addr_t size) 80995f72d1eSYinghai Lu { 81025cf23d7SMinchan Kim phys_addr_t end = base + size - 1; 81125cf23d7SMinchan Kim 812a090d711SAnshuman Khandual memblock_dbg("%s: [%pa-%pa] %pS\n", __func__, 81325cf23d7SMinchan Kim &base, &end, (void *)_RET_IP_); 81425cf23d7SMinchan Kim 815f1af9d3aSPhilipp Hachtmann return memblock_remove_range(&memblock.memory, base, size); 81695f72d1eSYinghai Lu } 81795f72d1eSYinghai Lu 8184d72868cSMike Rapoport /** 8194d72868cSMike Rapoport * memblock_free - free boot memory block 8204d72868cSMike Rapoport * @base: phys starting address of the boot memory block 8214d72868cSMike Rapoport * @size: size of the boot memory block in bytes 8224d72868cSMike Rapoport * 8234d72868cSMike Rapoport * Free boot memory block previously allocated by memblock_alloc_xx() API. 8244d72868cSMike Rapoport * The freeing memory will not be released to the buddy allocator. 8254d72868cSMike Rapoport */ 826581adcbeSTejun Heo int __init_memblock memblock_free(phys_addr_t base, phys_addr_t size) 82795f72d1eSYinghai Lu { 8285d63f81cSMiles Chen phys_addr_t end = base + size - 1; 8295d63f81cSMiles Chen 830a090d711SAnshuman Khandual memblock_dbg("%s: [%pa-%pa] %pS\n", __func__, 8315d63f81cSMiles Chen &base, &end, (void *)_RET_IP_); 83224aa0788STejun Heo 8339099daedSCatalin Marinas kmemleak_free_part_phys(base, size); 834f1af9d3aSPhilipp Hachtmann return memblock_remove_range(&memblock.reserved, base, size); 83595f72d1eSYinghai Lu } 83695f72d1eSYinghai Lu 837f705ac4bSAlexander Kuleshov int __init_memblock memblock_reserve(phys_addr_t base, phys_addr_t size) 83895f72d1eSYinghai Lu { 8395d63f81cSMiles Chen phys_addr_t end = base + size - 1; 8405d63f81cSMiles Chen 841a090d711SAnshuman Khandual memblock_dbg("%s: [%pa-%pa] %pS\n", __func__, 8425d63f81cSMiles Chen &base, &end, (void *)_RET_IP_); 84395f72d1eSYinghai Lu 844f705ac4bSAlexander Kuleshov return memblock_add_range(&memblock.reserved, base, size, MAX_NUMNODES, 0); 84595f72d1eSYinghai Lu } 84695f72d1eSYinghai Lu 84702634a44SAnshuman Khandual #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP 84802634a44SAnshuman Khandual int __init_memblock memblock_physmem_add(phys_addr_t base, phys_addr_t size) 84902634a44SAnshuman Khandual { 85002634a44SAnshuman Khandual phys_addr_t end = base + size - 1; 85102634a44SAnshuman Khandual 85202634a44SAnshuman Khandual memblock_dbg("%s: [%pa-%pa] %pS\n", __func__, 85302634a44SAnshuman Khandual &base, &end, (void *)_RET_IP_); 85402634a44SAnshuman Khandual 85577649905SDavid Hildenbrand return memblock_add_range(&physmem, base, size, MAX_NUMNODES, 0); 85602634a44SAnshuman Khandual } 85702634a44SAnshuman Khandual #endif 85802634a44SAnshuman Khandual 85935fd0808STejun Heo /** 86047cec443SMike Rapoport * memblock_setclr_flag - set or clear flag for a memory region 86147cec443SMike Rapoport * @base: base address of the region 86247cec443SMike Rapoport * @size: size of the region 86347cec443SMike Rapoport * @set: set or clear the flag 86447cec443SMike Rapoport * @flag: the flag to udpate 86566b16edfSTang Chen * 8664308ce17STony Luck * This function isolates region [@base, @base + @size), and sets/clears flag 86766b16edfSTang Chen * 86847cec443SMike Rapoport * Return: 0 on success, -errno on failure. 86966b16edfSTang Chen */ 8704308ce17STony Luck static int __init_memblock memblock_setclr_flag(phys_addr_t base, 8714308ce17STony Luck phys_addr_t size, int set, int flag) 87266b16edfSTang Chen { 87366b16edfSTang Chen struct memblock_type *type = &memblock.memory; 87466b16edfSTang Chen int i, ret, start_rgn, end_rgn; 87566b16edfSTang Chen 87666b16edfSTang Chen ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); 87766b16edfSTang Chen if (ret) 87866b16edfSTang Chen return ret; 87966b16edfSTang Chen 880fe145124SMike Rapoport for (i = start_rgn; i < end_rgn; i++) { 881fe145124SMike Rapoport struct memblock_region *r = &type->regions[i]; 882fe145124SMike Rapoport 8834308ce17STony Luck if (set) 884fe145124SMike Rapoport r->flags |= flag; 8854308ce17STony Luck else 886fe145124SMike Rapoport r->flags &= ~flag; 887fe145124SMike Rapoport } 88866b16edfSTang Chen 88966b16edfSTang Chen memblock_merge_regions(type); 89066b16edfSTang Chen return 0; 89166b16edfSTang Chen } 89266b16edfSTang Chen 89366b16edfSTang Chen /** 8944308ce17STony Luck * memblock_mark_hotplug - Mark hotpluggable memory with flag MEMBLOCK_HOTPLUG. 8954308ce17STony Luck * @base: the base phys addr of the region 8964308ce17STony Luck * @size: the size of the region 8974308ce17STony Luck * 89847cec443SMike Rapoport * Return: 0 on success, -errno on failure. 8994308ce17STony Luck */ 9004308ce17STony Luck int __init_memblock memblock_mark_hotplug(phys_addr_t base, phys_addr_t size) 9014308ce17STony Luck { 9024308ce17STony Luck return memblock_setclr_flag(base, size, 1, MEMBLOCK_HOTPLUG); 9034308ce17STony Luck } 9044308ce17STony Luck 9054308ce17STony Luck /** 90666b16edfSTang Chen * memblock_clear_hotplug - Clear flag MEMBLOCK_HOTPLUG for a specified region. 90766b16edfSTang Chen * @base: the base phys addr of the region 90866b16edfSTang Chen * @size: the size of the region 90966b16edfSTang Chen * 91047cec443SMike Rapoport * Return: 0 on success, -errno on failure. 91166b16edfSTang Chen */ 91266b16edfSTang Chen int __init_memblock memblock_clear_hotplug(phys_addr_t base, phys_addr_t size) 91366b16edfSTang Chen { 9144308ce17STony Luck return memblock_setclr_flag(base, size, 0, MEMBLOCK_HOTPLUG); 91566b16edfSTang Chen } 91666b16edfSTang Chen 91766b16edfSTang Chen /** 918a3f5bafcSTony Luck * memblock_mark_mirror - Mark mirrored memory with flag MEMBLOCK_MIRROR. 919a3f5bafcSTony Luck * @base: the base phys addr of the region 920a3f5bafcSTony Luck * @size: the size of the region 921a3f5bafcSTony Luck * 92247cec443SMike Rapoport * Return: 0 on success, -errno on failure. 923a3f5bafcSTony Luck */ 924a3f5bafcSTony Luck int __init_memblock memblock_mark_mirror(phys_addr_t base, phys_addr_t size) 925a3f5bafcSTony Luck { 926a3f5bafcSTony Luck system_has_some_mirror = true; 927a3f5bafcSTony Luck 928a3f5bafcSTony Luck return memblock_setclr_flag(base, size, 1, MEMBLOCK_MIRROR); 929a3f5bafcSTony Luck } 930a3f5bafcSTony Luck 931bf3d3cc5SArd Biesheuvel /** 932bf3d3cc5SArd Biesheuvel * memblock_mark_nomap - Mark a memory region with flag MEMBLOCK_NOMAP. 933bf3d3cc5SArd Biesheuvel * @base: the base phys addr of the region 934bf3d3cc5SArd Biesheuvel * @size: the size of the region 935bf3d3cc5SArd Biesheuvel * 93647cec443SMike Rapoport * Return: 0 on success, -errno on failure. 937bf3d3cc5SArd Biesheuvel */ 938bf3d3cc5SArd Biesheuvel int __init_memblock memblock_mark_nomap(phys_addr_t base, phys_addr_t size) 939bf3d3cc5SArd Biesheuvel { 940bf3d3cc5SArd Biesheuvel return memblock_setclr_flag(base, size, 1, MEMBLOCK_NOMAP); 941bf3d3cc5SArd Biesheuvel } 942a3f5bafcSTony Luck 943a3f5bafcSTony Luck /** 9444c546b8aSAKASHI Takahiro * memblock_clear_nomap - Clear flag MEMBLOCK_NOMAP for a specified region. 9454c546b8aSAKASHI Takahiro * @base: the base phys addr of the region 9464c546b8aSAKASHI Takahiro * @size: the size of the region 9474c546b8aSAKASHI Takahiro * 94847cec443SMike Rapoport * Return: 0 on success, -errno on failure. 9494c546b8aSAKASHI Takahiro */ 9504c546b8aSAKASHI Takahiro int __init_memblock memblock_clear_nomap(phys_addr_t base, phys_addr_t size) 9514c546b8aSAKASHI Takahiro { 9524c546b8aSAKASHI Takahiro return memblock_setclr_flag(base, size, 0, MEMBLOCK_NOMAP); 9534c546b8aSAKASHI Takahiro } 9544c546b8aSAKASHI Takahiro 9554c546b8aSAKASHI Takahiro /** 9568e7a7f86SRobin Holt * __next_reserved_mem_region - next function for for_each_reserved_region() 9578e7a7f86SRobin Holt * @idx: pointer to u64 loop variable 9588e7a7f86SRobin Holt * @out_start: ptr to phys_addr_t for start address of the region, can be %NULL 9598e7a7f86SRobin Holt * @out_end: ptr to phys_addr_t for end address of the region, can be %NULL 9608e7a7f86SRobin Holt * 9618e7a7f86SRobin Holt * Iterate over all reserved memory regions. 9628e7a7f86SRobin Holt */ 9638e7a7f86SRobin Holt void __init_memblock __next_reserved_mem_region(u64 *idx, 9648e7a7f86SRobin Holt phys_addr_t *out_start, 9658e7a7f86SRobin Holt phys_addr_t *out_end) 9668e7a7f86SRobin Holt { 967567d117bSAlexander Kuleshov struct memblock_type *type = &memblock.reserved; 9688e7a7f86SRobin Holt 969cd33a76bSRichard Leitner if (*idx < type->cnt) { 970567d117bSAlexander Kuleshov struct memblock_region *r = &type->regions[*idx]; 9718e7a7f86SRobin Holt phys_addr_t base = r->base; 9728e7a7f86SRobin Holt phys_addr_t size = r->size; 9738e7a7f86SRobin Holt 9748e7a7f86SRobin Holt if (out_start) 9758e7a7f86SRobin Holt *out_start = base; 9768e7a7f86SRobin Holt if (out_end) 9778e7a7f86SRobin Holt *out_end = base + size - 1; 9788e7a7f86SRobin Holt 9798e7a7f86SRobin Holt *idx += 1; 9808e7a7f86SRobin Holt return; 9818e7a7f86SRobin Holt } 9828e7a7f86SRobin Holt 9838e7a7f86SRobin Holt /* signal end of iteration */ 9848e7a7f86SRobin Holt *idx = ULLONG_MAX; 9858e7a7f86SRobin Holt } 9868e7a7f86SRobin Holt 987c9a688a3SMike Rapoport static bool should_skip_region(struct memblock_region *m, int nid, int flags) 988c9a688a3SMike Rapoport { 989c9a688a3SMike Rapoport int m_nid = memblock_get_region_node(m); 990c9a688a3SMike Rapoport 991c9a688a3SMike Rapoport /* only memory regions are associated with nodes, check it */ 992c9a688a3SMike Rapoport if (nid != NUMA_NO_NODE && nid != m_nid) 993c9a688a3SMike Rapoport return true; 994c9a688a3SMike Rapoport 995c9a688a3SMike Rapoport /* skip hotpluggable memory regions if needed */ 996c9a688a3SMike Rapoport if (movable_node_is_enabled() && memblock_is_hotpluggable(m)) 997c9a688a3SMike Rapoport return true; 998c9a688a3SMike Rapoport 999c9a688a3SMike Rapoport /* if we want mirror memory skip non-mirror memory regions */ 1000c9a688a3SMike Rapoport if ((flags & MEMBLOCK_MIRROR) && !memblock_is_mirror(m)) 1001c9a688a3SMike Rapoport return true; 1002c9a688a3SMike Rapoport 1003c9a688a3SMike Rapoport /* skip nomap memory unless we were asked for it explicitly */ 1004c9a688a3SMike Rapoport if (!(flags & MEMBLOCK_NOMAP) && memblock_is_nomap(m)) 1005c9a688a3SMike Rapoport return true; 1006c9a688a3SMike Rapoport 1007c9a688a3SMike Rapoport return false; 1008c9a688a3SMike Rapoport } 1009c9a688a3SMike Rapoport 10108e7a7f86SRobin Holt /** 1011a2974133SMike Rapoport * __next_mem_range - next function for for_each_free_mem_range() etc. 101235fd0808STejun Heo * @idx: pointer to u64 loop variable 1013b1154233SGrygorii Strashko * @nid: node selector, %NUMA_NO_NODE for all nodes 1014fc6daaf9STony Luck * @flags: pick from blocks based on memory attributes 1015f1af9d3aSPhilipp Hachtmann * @type_a: pointer to memblock_type from where the range is taken 1016f1af9d3aSPhilipp Hachtmann * @type_b: pointer to memblock_type which excludes memory from being taken 1017dad7557eSWanpeng Li * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL 1018dad7557eSWanpeng Li * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL 1019dad7557eSWanpeng Li * @out_nid: ptr to int for nid of the range, can be %NULL 102035fd0808STejun Heo * 1021f1af9d3aSPhilipp Hachtmann * Find the first area from *@idx which matches @nid, fill the out 102235fd0808STejun Heo * parameters, and update *@idx for the next iteration. The lower 32bit of 1023f1af9d3aSPhilipp Hachtmann * *@idx contains index into type_a and the upper 32bit indexes the 1024f1af9d3aSPhilipp Hachtmann * areas before each region in type_b. For example, if type_b regions 102535fd0808STejun Heo * look like the following, 102635fd0808STejun Heo * 102735fd0808STejun Heo * 0:[0-16), 1:[32-48), 2:[128-130) 102835fd0808STejun Heo * 102935fd0808STejun Heo * The upper 32bit indexes the following regions. 103035fd0808STejun Heo * 103135fd0808STejun Heo * 0:[0-0), 1:[16-32), 2:[48-128), 3:[130-MAX) 103235fd0808STejun Heo * 103335fd0808STejun Heo * As both region arrays are sorted, the function advances the two indices 103435fd0808STejun Heo * in lockstep and returns each intersection. 103535fd0808STejun Heo */ 103677649905SDavid Hildenbrand void __next_mem_range(u64 *idx, int nid, enum memblock_flags flags, 1037f1af9d3aSPhilipp Hachtmann struct memblock_type *type_a, 103877649905SDavid Hildenbrand struct memblock_type *type_b, phys_addr_t *out_start, 103935fd0808STejun Heo phys_addr_t *out_end, int *out_nid) 104035fd0808STejun Heo { 1041f1af9d3aSPhilipp Hachtmann int idx_a = *idx & 0xffffffff; 1042f1af9d3aSPhilipp Hachtmann int idx_b = *idx >> 32; 1043b1154233SGrygorii Strashko 1044f1af9d3aSPhilipp Hachtmann if (WARN_ONCE(nid == MAX_NUMNODES, 1045f1af9d3aSPhilipp Hachtmann "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) 1046560dca27SGrygorii Strashko nid = NUMA_NO_NODE; 104735fd0808STejun Heo 1048f1af9d3aSPhilipp Hachtmann for (; idx_a < type_a->cnt; idx_a++) { 1049f1af9d3aSPhilipp Hachtmann struct memblock_region *m = &type_a->regions[idx_a]; 1050f1af9d3aSPhilipp Hachtmann 105135fd0808STejun Heo phys_addr_t m_start = m->base; 105235fd0808STejun Heo phys_addr_t m_end = m->base + m->size; 1053f1af9d3aSPhilipp Hachtmann int m_nid = memblock_get_region_node(m); 105435fd0808STejun Heo 1055c9a688a3SMike Rapoport if (should_skip_region(m, nid, flags)) 1056bf3d3cc5SArd Biesheuvel continue; 1057bf3d3cc5SArd Biesheuvel 1058f1af9d3aSPhilipp Hachtmann if (!type_b) { 1059f1af9d3aSPhilipp Hachtmann if (out_start) 1060f1af9d3aSPhilipp Hachtmann *out_start = m_start; 1061f1af9d3aSPhilipp Hachtmann if (out_end) 1062f1af9d3aSPhilipp Hachtmann *out_end = m_end; 1063f1af9d3aSPhilipp Hachtmann if (out_nid) 1064f1af9d3aSPhilipp Hachtmann *out_nid = m_nid; 1065f1af9d3aSPhilipp Hachtmann idx_a++; 1066f1af9d3aSPhilipp Hachtmann *idx = (u32)idx_a | (u64)idx_b << 32; 1067f1af9d3aSPhilipp Hachtmann return; 1068f1af9d3aSPhilipp Hachtmann } 106935fd0808STejun Heo 1070f1af9d3aSPhilipp Hachtmann /* scan areas before each reservation */ 1071f1af9d3aSPhilipp Hachtmann for (; idx_b < type_b->cnt + 1; idx_b++) { 1072f1af9d3aSPhilipp Hachtmann struct memblock_region *r; 1073f1af9d3aSPhilipp Hachtmann phys_addr_t r_start; 1074f1af9d3aSPhilipp Hachtmann phys_addr_t r_end; 1075f1af9d3aSPhilipp Hachtmann 1076f1af9d3aSPhilipp Hachtmann r = &type_b->regions[idx_b]; 1077f1af9d3aSPhilipp Hachtmann r_start = idx_b ? r[-1].base + r[-1].size : 0; 1078f1af9d3aSPhilipp Hachtmann r_end = idx_b < type_b->cnt ? 10791c4bc43dSStefan Agner r->base : PHYS_ADDR_MAX; 1080f1af9d3aSPhilipp Hachtmann 1081f1af9d3aSPhilipp Hachtmann /* 1082f1af9d3aSPhilipp Hachtmann * if idx_b advanced past idx_a, 1083f1af9d3aSPhilipp Hachtmann * break out to advance idx_a 1084f1af9d3aSPhilipp Hachtmann */ 108535fd0808STejun Heo if (r_start >= m_end) 108635fd0808STejun Heo break; 108735fd0808STejun Heo /* if the two regions intersect, we're done */ 108835fd0808STejun Heo if (m_start < r_end) { 108935fd0808STejun Heo if (out_start) 1090f1af9d3aSPhilipp Hachtmann *out_start = 1091f1af9d3aSPhilipp Hachtmann max(m_start, r_start); 109235fd0808STejun Heo if (out_end) 109335fd0808STejun Heo *out_end = min(m_end, r_end); 109435fd0808STejun Heo if (out_nid) 1095f1af9d3aSPhilipp Hachtmann *out_nid = m_nid; 109635fd0808STejun Heo /* 1097f1af9d3aSPhilipp Hachtmann * The region which ends first is 1098f1af9d3aSPhilipp Hachtmann * advanced for the next iteration. 109935fd0808STejun Heo */ 110035fd0808STejun Heo if (m_end <= r_end) 1101f1af9d3aSPhilipp Hachtmann idx_a++; 110235fd0808STejun Heo else 1103f1af9d3aSPhilipp Hachtmann idx_b++; 1104f1af9d3aSPhilipp Hachtmann *idx = (u32)idx_a | (u64)idx_b << 32; 110535fd0808STejun Heo return; 110635fd0808STejun Heo } 110735fd0808STejun Heo } 110835fd0808STejun Heo } 110935fd0808STejun Heo 111035fd0808STejun Heo /* signal end of iteration */ 111135fd0808STejun Heo *idx = ULLONG_MAX; 111235fd0808STejun Heo } 111335fd0808STejun Heo 11147bd0b0f0STejun Heo /** 1115f1af9d3aSPhilipp Hachtmann * __next_mem_range_rev - generic next function for for_each_*_range_rev() 1116f1af9d3aSPhilipp Hachtmann * 11177bd0b0f0STejun Heo * @idx: pointer to u64 loop variable 1118ad5ea8cdSAlexander Kuleshov * @nid: node selector, %NUMA_NO_NODE for all nodes 1119fc6daaf9STony Luck * @flags: pick from blocks based on memory attributes 1120f1af9d3aSPhilipp Hachtmann * @type_a: pointer to memblock_type from where the range is taken 1121f1af9d3aSPhilipp Hachtmann * @type_b: pointer to memblock_type which excludes memory from being taken 1122dad7557eSWanpeng Li * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL 1123dad7557eSWanpeng Li * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL 1124dad7557eSWanpeng Li * @out_nid: ptr to int for nid of the range, can be %NULL 11257bd0b0f0STejun Heo * 112647cec443SMike Rapoport * Finds the next range from type_a which is not marked as unsuitable 112747cec443SMike Rapoport * in type_b. 112847cec443SMike Rapoport * 1129f1af9d3aSPhilipp Hachtmann * Reverse of __next_mem_range(). 11307bd0b0f0STejun Heo */ 1131e1720feeSMike Rapoport void __init_memblock __next_mem_range_rev(u64 *idx, int nid, 1132e1720feeSMike Rapoport enum memblock_flags flags, 1133f1af9d3aSPhilipp Hachtmann struct memblock_type *type_a, 1134f1af9d3aSPhilipp Hachtmann struct memblock_type *type_b, 11357bd0b0f0STejun Heo phys_addr_t *out_start, 11367bd0b0f0STejun Heo phys_addr_t *out_end, int *out_nid) 11377bd0b0f0STejun Heo { 1138f1af9d3aSPhilipp Hachtmann int idx_a = *idx & 0xffffffff; 1139f1af9d3aSPhilipp Hachtmann int idx_b = *idx >> 32; 1140b1154233SGrygorii Strashko 1141560dca27SGrygorii Strashko if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) 1142560dca27SGrygorii Strashko nid = NUMA_NO_NODE; 11437bd0b0f0STejun Heo 11447bd0b0f0STejun Heo if (*idx == (u64)ULLONG_MAX) { 1145f1af9d3aSPhilipp Hachtmann idx_a = type_a->cnt - 1; 1146e47608abSzijun_hu if (type_b != NULL) 1147f1af9d3aSPhilipp Hachtmann idx_b = type_b->cnt; 1148e47608abSzijun_hu else 1149e47608abSzijun_hu idx_b = 0; 11507bd0b0f0STejun Heo } 11517bd0b0f0STejun Heo 1152f1af9d3aSPhilipp Hachtmann for (; idx_a >= 0; idx_a--) { 1153f1af9d3aSPhilipp Hachtmann struct memblock_region *m = &type_a->regions[idx_a]; 1154f1af9d3aSPhilipp Hachtmann 11557bd0b0f0STejun Heo phys_addr_t m_start = m->base; 11567bd0b0f0STejun Heo phys_addr_t m_end = m->base + m->size; 1157f1af9d3aSPhilipp Hachtmann int m_nid = memblock_get_region_node(m); 11587bd0b0f0STejun Heo 1159c9a688a3SMike Rapoport if (should_skip_region(m, nid, flags)) 1160bf3d3cc5SArd Biesheuvel continue; 1161bf3d3cc5SArd Biesheuvel 1162f1af9d3aSPhilipp Hachtmann if (!type_b) { 1163f1af9d3aSPhilipp Hachtmann if (out_start) 1164f1af9d3aSPhilipp Hachtmann *out_start = m_start; 1165f1af9d3aSPhilipp Hachtmann if (out_end) 1166f1af9d3aSPhilipp Hachtmann *out_end = m_end; 1167f1af9d3aSPhilipp Hachtmann if (out_nid) 1168f1af9d3aSPhilipp Hachtmann *out_nid = m_nid; 1169fb399b48Szijun_hu idx_a--; 1170f1af9d3aSPhilipp Hachtmann *idx = (u32)idx_a | (u64)idx_b << 32; 1171f1af9d3aSPhilipp Hachtmann return; 1172f1af9d3aSPhilipp Hachtmann } 11737bd0b0f0STejun Heo 1174f1af9d3aSPhilipp Hachtmann /* scan areas before each reservation */ 1175f1af9d3aSPhilipp Hachtmann for (; idx_b >= 0; idx_b--) { 1176f1af9d3aSPhilipp Hachtmann struct memblock_region *r; 1177f1af9d3aSPhilipp Hachtmann phys_addr_t r_start; 1178f1af9d3aSPhilipp Hachtmann phys_addr_t r_end; 1179f1af9d3aSPhilipp Hachtmann 1180f1af9d3aSPhilipp Hachtmann r = &type_b->regions[idx_b]; 1181f1af9d3aSPhilipp Hachtmann r_start = idx_b ? r[-1].base + r[-1].size : 0; 1182f1af9d3aSPhilipp Hachtmann r_end = idx_b < type_b->cnt ? 11831c4bc43dSStefan Agner r->base : PHYS_ADDR_MAX; 1184f1af9d3aSPhilipp Hachtmann /* 1185f1af9d3aSPhilipp Hachtmann * if idx_b advanced past idx_a, 1186f1af9d3aSPhilipp Hachtmann * break out to advance idx_a 1187f1af9d3aSPhilipp Hachtmann */ 1188f1af9d3aSPhilipp Hachtmann 11897bd0b0f0STejun Heo if (r_end <= m_start) 11907bd0b0f0STejun Heo break; 11917bd0b0f0STejun Heo /* if the two regions intersect, we're done */ 11927bd0b0f0STejun Heo if (m_end > r_start) { 11937bd0b0f0STejun Heo if (out_start) 11947bd0b0f0STejun Heo *out_start = max(m_start, r_start); 11957bd0b0f0STejun Heo if (out_end) 11967bd0b0f0STejun Heo *out_end = min(m_end, r_end); 11977bd0b0f0STejun Heo if (out_nid) 1198f1af9d3aSPhilipp Hachtmann *out_nid = m_nid; 11997bd0b0f0STejun Heo if (m_start >= r_start) 1200f1af9d3aSPhilipp Hachtmann idx_a--; 12017bd0b0f0STejun Heo else 1202f1af9d3aSPhilipp Hachtmann idx_b--; 1203f1af9d3aSPhilipp Hachtmann *idx = (u32)idx_a | (u64)idx_b << 32; 12047bd0b0f0STejun Heo return; 12057bd0b0f0STejun Heo } 12067bd0b0f0STejun Heo } 12077bd0b0f0STejun Heo } 1208f1af9d3aSPhilipp Hachtmann /* signal end of iteration */ 12097bd0b0f0STejun Heo *idx = ULLONG_MAX; 12107bd0b0f0STejun Heo } 12117bd0b0f0STejun Heo 12127c0caeb8STejun Heo /* 121345e79815SChen Chang * Common iterator interface used to define for_each_mem_pfn_range(). 12147c0caeb8STejun Heo */ 12157c0caeb8STejun Heo void __init_memblock __next_mem_pfn_range(int *idx, int nid, 12167c0caeb8STejun Heo unsigned long *out_start_pfn, 12177c0caeb8STejun Heo unsigned long *out_end_pfn, int *out_nid) 12187c0caeb8STejun Heo { 12197c0caeb8STejun Heo struct memblock_type *type = &memblock.memory; 12207c0caeb8STejun Heo struct memblock_region *r; 1221d622abf7SMike Rapoport int r_nid; 12227c0caeb8STejun Heo 12237c0caeb8STejun Heo while (++*idx < type->cnt) { 12247c0caeb8STejun Heo r = &type->regions[*idx]; 1225d622abf7SMike Rapoport r_nid = memblock_get_region_node(r); 12267c0caeb8STejun Heo 12277c0caeb8STejun Heo if (PFN_UP(r->base) >= PFN_DOWN(r->base + r->size)) 12287c0caeb8STejun Heo continue; 1229d622abf7SMike Rapoport if (nid == MAX_NUMNODES || nid == r_nid) 12307c0caeb8STejun Heo break; 12317c0caeb8STejun Heo } 12327c0caeb8STejun Heo if (*idx >= type->cnt) { 12337c0caeb8STejun Heo *idx = -1; 12347c0caeb8STejun Heo return; 12357c0caeb8STejun Heo } 12367c0caeb8STejun Heo 12377c0caeb8STejun Heo if (out_start_pfn) 12387c0caeb8STejun Heo *out_start_pfn = PFN_UP(r->base); 12397c0caeb8STejun Heo if (out_end_pfn) 12407c0caeb8STejun Heo *out_end_pfn = PFN_DOWN(r->base + r->size); 12417c0caeb8STejun Heo if (out_nid) 1242d622abf7SMike Rapoport *out_nid = r_nid; 12437c0caeb8STejun Heo } 12447c0caeb8STejun Heo 12457c0caeb8STejun Heo /** 12467c0caeb8STejun Heo * memblock_set_node - set node ID on memblock regions 12477c0caeb8STejun Heo * @base: base of area to set node ID for 12487c0caeb8STejun Heo * @size: size of area to set node ID for 1249e7e8de59STang Chen * @type: memblock type to set node ID for 12507c0caeb8STejun Heo * @nid: node ID to set 12517c0caeb8STejun Heo * 1252e7e8de59STang Chen * Set the nid of memblock @type regions in [@base, @base + @size) to @nid. 12537c0caeb8STejun Heo * Regions which cross the area boundaries are split as necessary. 12547c0caeb8STejun Heo * 125547cec443SMike Rapoport * Return: 12567c0caeb8STejun Heo * 0 on success, -errno on failure. 12577c0caeb8STejun Heo */ 12587c0caeb8STejun Heo int __init_memblock memblock_set_node(phys_addr_t base, phys_addr_t size, 1259e7e8de59STang Chen struct memblock_type *type, int nid) 12607c0caeb8STejun Heo { 12613f08a302SMike Rapoport #ifdef CONFIG_NEED_MULTIPLE_NODES 12626a9ceb31STejun Heo int start_rgn, end_rgn; 12636a9ceb31STejun Heo int i, ret; 12647c0caeb8STejun Heo 12656a9ceb31STejun Heo ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); 12666a9ceb31STejun Heo if (ret) 12676a9ceb31STejun Heo return ret; 12687c0caeb8STejun Heo 12696a9ceb31STejun Heo for (i = start_rgn; i < end_rgn; i++) 1270e9d24ad3SWanpeng Li memblock_set_region_node(&type->regions[i], nid); 12717c0caeb8STejun Heo 12727c0caeb8STejun Heo memblock_merge_regions(type); 12733f08a302SMike Rapoport #endif 12747c0caeb8STejun Heo return 0; 12757c0caeb8STejun Heo } 12763f08a302SMike Rapoport 1277837566e7SAlexander Duyck #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT 1278837566e7SAlexander Duyck /** 1279837566e7SAlexander Duyck * __next_mem_pfn_range_in_zone - iterator for for_each_*_range_in_zone() 1280837566e7SAlexander Duyck * 1281837566e7SAlexander Duyck * @idx: pointer to u64 loop variable 1282837566e7SAlexander Duyck * @zone: zone in which all of the memory blocks reside 1283837566e7SAlexander Duyck * @out_spfn: ptr to ulong for start pfn of the range, can be %NULL 1284837566e7SAlexander Duyck * @out_epfn: ptr to ulong for end pfn of the range, can be %NULL 1285837566e7SAlexander Duyck * 1286837566e7SAlexander Duyck * This function is meant to be a zone/pfn specific wrapper for the 1287837566e7SAlexander Duyck * for_each_mem_range type iterators. Specifically they are used in the 1288837566e7SAlexander Duyck * deferred memory init routines and as such we were duplicating much of 1289837566e7SAlexander Duyck * this logic throughout the code. So instead of having it in multiple 1290837566e7SAlexander Duyck * locations it seemed like it would make more sense to centralize this to 1291837566e7SAlexander Duyck * one new iterator that does everything they need. 1292837566e7SAlexander Duyck */ 1293837566e7SAlexander Duyck void __init_memblock 1294837566e7SAlexander Duyck __next_mem_pfn_range_in_zone(u64 *idx, struct zone *zone, 1295837566e7SAlexander Duyck unsigned long *out_spfn, unsigned long *out_epfn) 1296837566e7SAlexander Duyck { 1297837566e7SAlexander Duyck int zone_nid = zone_to_nid(zone); 1298837566e7SAlexander Duyck phys_addr_t spa, epa; 1299837566e7SAlexander Duyck int nid; 1300837566e7SAlexander Duyck 1301837566e7SAlexander Duyck __next_mem_range(idx, zone_nid, MEMBLOCK_NONE, 1302837566e7SAlexander Duyck &memblock.memory, &memblock.reserved, 1303837566e7SAlexander Duyck &spa, &epa, &nid); 1304837566e7SAlexander Duyck 1305837566e7SAlexander Duyck while (*idx != U64_MAX) { 1306837566e7SAlexander Duyck unsigned long epfn = PFN_DOWN(epa); 1307837566e7SAlexander Duyck unsigned long spfn = PFN_UP(spa); 1308837566e7SAlexander Duyck 1309837566e7SAlexander Duyck /* 1310837566e7SAlexander Duyck * Verify the end is at least past the start of the zone and 1311837566e7SAlexander Duyck * that we have at least one PFN to initialize. 1312837566e7SAlexander Duyck */ 1313837566e7SAlexander Duyck if (zone->zone_start_pfn < epfn && spfn < epfn) { 1314837566e7SAlexander Duyck /* if we went too far just stop searching */ 1315837566e7SAlexander Duyck if (zone_end_pfn(zone) <= spfn) { 1316837566e7SAlexander Duyck *idx = U64_MAX; 1317837566e7SAlexander Duyck break; 1318837566e7SAlexander Duyck } 1319837566e7SAlexander Duyck 1320837566e7SAlexander Duyck if (out_spfn) 1321837566e7SAlexander Duyck *out_spfn = max(zone->zone_start_pfn, spfn); 1322837566e7SAlexander Duyck if (out_epfn) 1323837566e7SAlexander Duyck *out_epfn = min(zone_end_pfn(zone), epfn); 1324837566e7SAlexander Duyck 1325837566e7SAlexander Duyck return; 1326837566e7SAlexander Duyck } 1327837566e7SAlexander Duyck 1328837566e7SAlexander Duyck __next_mem_range(idx, zone_nid, MEMBLOCK_NONE, 1329837566e7SAlexander Duyck &memblock.memory, &memblock.reserved, 1330837566e7SAlexander Duyck &spa, &epa, &nid); 1331837566e7SAlexander Duyck } 1332837566e7SAlexander Duyck 1333837566e7SAlexander Duyck /* signal end of iteration */ 1334837566e7SAlexander Duyck if (out_spfn) 1335837566e7SAlexander Duyck *out_spfn = ULONG_MAX; 1336837566e7SAlexander Duyck if (out_epfn) 1337837566e7SAlexander Duyck *out_epfn = 0; 1338837566e7SAlexander Duyck } 1339837566e7SAlexander Duyck 1340837566e7SAlexander Duyck #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */ 13417c0caeb8STejun Heo 134292d12f95SMike Rapoport /** 134392d12f95SMike Rapoport * memblock_alloc_range_nid - allocate boot memory block 134492d12f95SMike Rapoport * @size: size of memory block to be allocated in bytes 134592d12f95SMike Rapoport * @align: alignment of the region and block's size 134692d12f95SMike Rapoport * @start: the lower bound of the memory region to allocate (phys address) 134792d12f95SMike Rapoport * @end: the upper bound of the memory region to allocate (phys address) 134892d12f95SMike Rapoport * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 13490ac398b1SYunfeng Ye * @exact_nid: control the allocation fall back to other nodes 135092d12f95SMike Rapoport * 135192d12f95SMike Rapoport * The allocation is performed from memory region limited by 135295830666SCao jin * memblock.current_limit if @end == %MEMBLOCK_ALLOC_ACCESSIBLE. 135392d12f95SMike Rapoport * 13540ac398b1SYunfeng Ye * If the specified node can not hold the requested memory and @exact_nid 13550ac398b1SYunfeng Ye * is false, the allocation falls back to any node in the system. 135692d12f95SMike Rapoport * 135792d12f95SMike Rapoport * For systems with memory mirroring, the allocation is attempted first 135892d12f95SMike Rapoport * from the regions with mirroring enabled and then retried from any 135992d12f95SMike Rapoport * memory region. 136092d12f95SMike Rapoport * 136192d12f95SMike Rapoport * In addition, function sets the min_count to 0 using kmemleak_alloc_phys for 136292d12f95SMike Rapoport * allocated boot memory block, so that it is never reported as leaks. 136392d12f95SMike Rapoport * 136492d12f95SMike Rapoport * Return: 136592d12f95SMike Rapoport * Physical address of allocated memory block on success, %0 on failure. 136692d12f95SMike Rapoport */ 13678676af1fSAslan Bakirov phys_addr_t __init memblock_alloc_range_nid(phys_addr_t size, 13682bfc2862SAkinobu Mita phys_addr_t align, phys_addr_t start, 13690ac398b1SYunfeng Ye phys_addr_t end, int nid, 13700ac398b1SYunfeng Ye bool exact_nid) 137195f72d1eSYinghai Lu { 137292d12f95SMike Rapoport enum memblock_flags flags = choose_memblock_flags(); 13736ed311b2SBenjamin Herrenschmidt phys_addr_t found; 137495f72d1eSYinghai Lu 137592d12f95SMike Rapoport if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) 137692d12f95SMike Rapoport nid = NUMA_NO_NODE; 137792d12f95SMike Rapoport 13782f770806SMike Rapoport if (!align) { 13792f770806SMike Rapoport /* Can't use WARNs this early in boot on powerpc */ 13802f770806SMike Rapoport dump_stack(); 13812f770806SMike Rapoport align = SMP_CACHE_BYTES; 13822f770806SMike Rapoport } 13832f770806SMike Rapoport 138492d12f95SMike Rapoport again: 1385fc6daaf9STony Luck found = memblock_find_in_range_node(size, align, start, end, nid, 1386fc6daaf9STony Luck flags); 138792d12f95SMike Rapoport if (found && !memblock_reserve(found, size)) 138892d12f95SMike Rapoport goto done; 138992d12f95SMike Rapoport 13900ac398b1SYunfeng Ye if (nid != NUMA_NO_NODE && !exact_nid) { 139192d12f95SMike Rapoport found = memblock_find_in_range_node(size, align, start, 139292d12f95SMike Rapoport end, NUMA_NO_NODE, 139392d12f95SMike Rapoport flags); 139492d12f95SMike Rapoport if (found && !memblock_reserve(found, size)) 139592d12f95SMike Rapoport goto done; 139692d12f95SMike Rapoport } 139792d12f95SMike Rapoport 139892d12f95SMike Rapoport if (flags & MEMBLOCK_MIRROR) { 139992d12f95SMike Rapoport flags &= ~MEMBLOCK_MIRROR; 140092d12f95SMike Rapoport pr_warn("Could not allocate %pap bytes of mirrored memory\n", 140192d12f95SMike Rapoport &size); 140292d12f95SMike Rapoport goto again; 140392d12f95SMike Rapoport } 140492d12f95SMike Rapoport 140592d12f95SMike Rapoport return 0; 140692d12f95SMike Rapoport 140792d12f95SMike Rapoport done: 140892d12f95SMike Rapoport /* Skip kmemleak for kasan_init() due to high volume. */ 140992d12f95SMike Rapoport if (end != MEMBLOCK_ALLOC_KASAN) 1410aedf95eaSCatalin Marinas /* 141192d12f95SMike Rapoport * The min_count is set to 0 so that memblock allocated 141292d12f95SMike Rapoport * blocks are never reported as leaks. This is because many 141392d12f95SMike Rapoport * of these blocks are only referred via the physical 141492d12f95SMike Rapoport * address which is not looked up by kmemleak. 1415aedf95eaSCatalin Marinas */ 14169099daedSCatalin Marinas kmemleak_alloc_phys(found, size, 0, 0); 141792d12f95SMike Rapoport 14186ed311b2SBenjamin Herrenschmidt return found; 1419aedf95eaSCatalin Marinas } 142095f72d1eSYinghai Lu 1421a2974133SMike Rapoport /** 1422a2974133SMike Rapoport * memblock_phys_alloc_range - allocate a memory block inside specified range 1423a2974133SMike Rapoport * @size: size of memory block to be allocated in bytes 1424a2974133SMike Rapoport * @align: alignment of the region and block's size 1425a2974133SMike Rapoport * @start: the lower bound of the memory region to allocate (physical address) 1426a2974133SMike Rapoport * @end: the upper bound of the memory region to allocate (physical address) 1427a2974133SMike Rapoport * 1428a2974133SMike Rapoport * Allocate @size bytes in the between @start and @end. 1429a2974133SMike Rapoport * 1430a2974133SMike Rapoport * Return: physical address of the allocated memory block on success, 1431a2974133SMike Rapoport * %0 on failure. 1432a2974133SMike Rapoport */ 14338a770c2aSMike Rapoport phys_addr_t __init memblock_phys_alloc_range(phys_addr_t size, 14348a770c2aSMike Rapoport phys_addr_t align, 14358a770c2aSMike Rapoport phys_addr_t start, 14368a770c2aSMike Rapoport phys_addr_t end) 14372bfc2862SAkinobu Mita { 14380ac398b1SYunfeng Ye return memblock_alloc_range_nid(size, align, start, end, NUMA_NO_NODE, 14390ac398b1SYunfeng Ye false); 14407bd0b0f0STejun Heo } 14417bd0b0f0STejun Heo 1442a2974133SMike Rapoport /** 1443a2974133SMike Rapoport * memblock_phys_alloc_try_nid - allocate a memory block from specified MUMA node 1444a2974133SMike Rapoport * @size: size of memory block to be allocated in bytes 1445a2974133SMike Rapoport * @align: alignment of the region and block's size 1446a2974133SMike Rapoport * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 1447a2974133SMike Rapoport * 1448a2974133SMike Rapoport * Allocates memory block from the specified NUMA node. If the node 1449a2974133SMike Rapoport * has no available memory, attempts to allocated from any node in the 1450a2974133SMike Rapoport * system. 1451a2974133SMike Rapoport * 1452a2974133SMike Rapoport * Return: physical address of the allocated memory block on success, 1453a2974133SMike Rapoport * %0 on failure. 1454a2974133SMike Rapoport */ 14559a8dd708SMike Rapoport phys_addr_t __init memblock_phys_alloc_try_nid(phys_addr_t size, phys_addr_t align, int nid) 14569d1e2492SBenjamin Herrenschmidt { 145733755574SMike Rapoport return memblock_alloc_range_nid(size, align, 0, 14580ac398b1SYunfeng Ye MEMBLOCK_ALLOC_ACCESSIBLE, nid, false); 145995f72d1eSYinghai Lu } 146095f72d1eSYinghai Lu 146126f09e9bSSantosh Shilimkar /** 1462eb31d559SMike Rapoport * memblock_alloc_internal - allocate boot memory block 146326f09e9bSSantosh Shilimkar * @size: size of memory block to be allocated in bytes 146426f09e9bSSantosh Shilimkar * @align: alignment of the region and block's size 146526f09e9bSSantosh Shilimkar * @min_addr: the lower bound of the memory region to allocate (phys address) 146626f09e9bSSantosh Shilimkar * @max_addr: the upper bound of the memory region to allocate (phys address) 146726f09e9bSSantosh Shilimkar * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 14680ac398b1SYunfeng Ye * @exact_nid: control the allocation fall back to other nodes 146926f09e9bSSantosh Shilimkar * 147092d12f95SMike Rapoport * Allocates memory block using memblock_alloc_range_nid() and 147192d12f95SMike Rapoport * converts the returned physical address to virtual. 147292d12f95SMike Rapoport * 147326f09e9bSSantosh Shilimkar * The @min_addr limit is dropped if it can not be satisfied and the allocation 147492d12f95SMike Rapoport * will fall back to memory below @min_addr. Other constraints, such 147592d12f95SMike Rapoport * as node and mirrored memory will be handled again in 147692d12f95SMike Rapoport * memblock_alloc_range_nid(). 147726f09e9bSSantosh Shilimkar * 147847cec443SMike Rapoport * Return: 147926f09e9bSSantosh Shilimkar * Virtual address of allocated memory block on success, NULL on failure. 148026f09e9bSSantosh Shilimkar */ 1481eb31d559SMike Rapoport static void * __init memblock_alloc_internal( 148226f09e9bSSantosh Shilimkar phys_addr_t size, phys_addr_t align, 148326f09e9bSSantosh Shilimkar phys_addr_t min_addr, phys_addr_t max_addr, 14840ac398b1SYunfeng Ye int nid, bool exact_nid) 148526f09e9bSSantosh Shilimkar { 148626f09e9bSSantosh Shilimkar phys_addr_t alloc; 148726f09e9bSSantosh Shilimkar 148826f09e9bSSantosh Shilimkar /* 148926f09e9bSSantosh Shilimkar * Detect any accidental use of these APIs after slab is ready, as at 149026f09e9bSSantosh Shilimkar * this moment memblock may be deinitialized already and its 1491c6ffc5caSMike Rapoport * internal data may be destroyed (after execution of memblock_free_all) 149226f09e9bSSantosh Shilimkar */ 149326f09e9bSSantosh Shilimkar if (WARN_ON_ONCE(slab_is_available())) 149426f09e9bSSantosh Shilimkar return kzalloc_node(size, GFP_NOWAIT, nid); 149526f09e9bSSantosh Shilimkar 1496f3057ad7SMike Rapoport if (max_addr > memblock.current_limit) 1497f3057ad7SMike Rapoport max_addr = memblock.current_limit; 1498f3057ad7SMike Rapoport 14990ac398b1SYunfeng Ye alloc = memblock_alloc_range_nid(size, align, min_addr, max_addr, nid, 15000ac398b1SYunfeng Ye exact_nid); 15012f770806SMike Rapoport 150292d12f95SMike Rapoport /* retry allocation without lower limit */ 150392d12f95SMike Rapoport if (!alloc && min_addr) 15040ac398b1SYunfeng Ye alloc = memblock_alloc_range_nid(size, align, 0, max_addr, nid, 15050ac398b1SYunfeng Ye exact_nid); 150626f09e9bSSantosh Shilimkar 150792d12f95SMike Rapoport if (!alloc) 1508a3f5bafcSTony Luck return NULL; 150926f09e9bSSantosh Shilimkar 151092d12f95SMike Rapoport return phys_to_virt(alloc); 151126f09e9bSSantosh Shilimkar } 151226f09e9bSSantosh Shilimkar 151326f09e9bSSantosh Shilimkar /** 15140ac398b1SYunfeng Ye * memblock_alloc_exact_nid_raw - allocate boot memory block on the exact node 15150ac398b1SYunfeng Ye * without zeroing memory 15160ac398b1SYunfeng Ye * @size: size of memory block to be allocated in bytes 15170ac398b1SYunfeng Ye * @align: alignment of the region and block's size 15180ac398b1SYunfeng Ye * @min_addr: the lower bound of the memory region from where the allocation 15190ac398b1SYunfeng Ye * is preferred (phys address) 15200ac398b1SYunfeng Ye * @max_addr: the upper bound of the memory region from where the allocation 15210ac398b1SYunfeng Ye * is preferred (phys address), or %MEMBLOCK_ALLOC_ACCESSIBLE to 15220ac398b1SYunfeng Ye * allocate only from memory limited by memblock.current_limit value 15230ac398b1SYunfeng Ye * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 15240ac398b1SYunfeng Ye * 15250ac398b1SYunfeng Ye * Public function, provides additional debug information (including caller 15260ac398b1SYunfeng Ye * info), if enabled. Does not zero allocated memory. 15270ac398b1SYunfeng Ye * 15280ac398b1SYunfeng Ye * Return: 15290ac398b1SYunfeng Ye * Virtual address of allocated memory block on success, NULL on failure. 15300ac398b1SYunfeng Ye */ 15310ac398b1SYunfeng Ye void * __init memblock_alloc_exact_nid_raw( 15320ac398b1SYunfeng Ye phys_addr_t size, phys_addr_t align, 15330ac398b1SYunfeng Ye phys_addr_t min_addr, phys_addr_t max_addr, 15340ac398b1SYunfeng Ye int nid) 15350ac398b1SYunfeng Ye { 15360ac398b1SYunfeng Ye void *ptr; 15370ac398b1SYunfeng Ye 15380ac398b1SYunfeng Ye memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=%pa max_addr=%pa %pS\n", 15390ac398b1SYunfeng Ye __func__, (u64)size, (u64)align, nid, &min_addr, 15400ac398b1SYunfeng Ye &max_addr, (void *)_RET_IP_); 15410ac398b1SYunfeng Ye 15420ac398b1SYunfeng Ye ptr = memblock_alloc_internal(size, align, 15430ac398b1SYunfeng Ye min_addr, max_addr, nid, true); 15440ac398b1SYunfeng Ye if (ptr && size > 0) 15450ac398b1SYunfeng Ye page_init_poison(ptr, size); 15460ac398b1SYunfeng Ye 15470ac398b1SYunfeng Ye return ptr; 15480ac398b1SYunfeng Ye } 15490ac398b1SYunfeng Ye 15500ac398b1SYunfeng Ye /** 1551eb31d559SMike Rapoport * memblock_alloc_try_nid_raw - allocate boot memory block without zeroing 1552ea1f5f37SPavel Tatashin * memory and without panicking 1553ea1f5f37SPavel Tatashin * @size: size of memory block to be allocated in bytes 1554ea1f5f37SPavel Tatashin * @align: alignment of the region and block's size 1555ea1f5f37SPavel Tatashin * @min_addr: the lower bound of the memory region from where the allocation 1556ea1f5f37SPavel Tatashin * is preferred (phys address) 1557ea1f5f37SPavel Tatashin * @max_addr: the upper bound of the memory region from where the allocation 155897ad1087SMike Rapoport * is preferred (phys address), or %MEMBLOCK_ALLOC_ACCESSIBLE to 1559ea1f5f37SPavel Tatashin * allocate only from memory limited by memblock.current_limit value 1560ea1f5f37SPavel Tatashin * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 1561ea1f5f37SPavel Tatashin * 1562ea1f5f37SPavel Tatashin * Public function, provides additional debug information (including caller 1563ea1f5f37SPavel Tatashin * info), if enabled. Does not zero allocated memory, does not panic if request 1564ea1f5f37SPavel Tatashin * cannot be satisfied. 1565ea1f5f37SPavel Tatashin * 156647cec443SMike Rapoport * Return: 1567ea1f5f37SPavel Tatashin * Virtual address of allocated memory block on success, NULL on failure. 1568ea1f5f37SPavel Tatashin */ 1569eb31d559SMike Rapoport void * __init memblock_alloc_try_nid_raw( 1570ea1f5f37SPavel Tatashin phys_addr_t size, phys_addr_t align, 1571ea1f5f37SPavel Tatashin phys_addr_t min_addr, phys_addr_t max_addr, 1572ea1f5f37SPavel Tatashin int nid) 1573ea1f5f37SPavel Tatashin { 1574ea1f5f37SPavel Tatashin void *ptr; 1575ea1f5f37SPavel Tatashin 1576d75f773cSSakari Ailus memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=%pa max_addr=%pa %pS\n", 1577a36aab89SMike Rapoport __func__, (u64)size, (u64)align, nid, &min_addr, 1578a36aab89SMike Rapoport &max_addr, (void *)_RET_IP_); 1579ea1f5f37SPavel Tatashin 1580eb31d559SMike Rapoport ptr = memblock_alloc_internal(size, align, 15810ac398b1SYunfeng Ye min_addr, max_addr, nid, false); 1582ea1f5f37SPavel Tatashin if (ptr && size > 0) 1583f682a97aSAlexander Duyck page_init_poison(ptr, size); 1584f682a97aSAlexander Duyck 1585ea1f5f37SPavel Tatashin return ptr; 1586ea1f5f37SPavel Tatashin } 1587ea1f5f37SPavel Tatashin 1588ea1f5f37SPavel Tatashin /** 1589c0dbe825SMike Rapoport * memblock_alloc_try_nid - allocate boot memory block 159026f09e9bSSantosh Shilimkar * @size: size of memory block to be allocated in bytes 159126f09e9bSSantosh Shilimkar * @align: alignment of the region and block's size 159226f09e9bSSantosh Shilimkar * @min_addr: the lower bound of the memory region from where the allocation 159326f09e9bSSantosh Shilimkar * is preferred (phys address) 159426f09e9bSSantosh Shilimkar * @max_addr: the upper bound of the memory region from where the allocation 159597ad1087SMike Rapoport * is preferred (phys address), or %MEMBLOCK_ALLOC_ACCESSIBLE to 159626f09e9bSSantosh Shilimkar * allocate only from memory limited by memblock.current_limit value 159726f09e9bSSantosh Shilimkar * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 159826f09e9bSSantosh Shilimkar * 1599c0dbe825SMike Rapoport * Public function, provides additional debug information (including caller 1600c0dbe825SMike Rapoport * info), if enabled. This function zeroes the allocated memory. 160126f09e9bSSantosh Shilimkar * 160247cec443SMike Rapoport * Return: 160326f09e9bSSantosh Shilimkar * Virtual address of allocated memory block on success, NULL on failure. 160426f09e9bSSantosh Shilimkar */ 1605eb31d559SMike Rapoport void * __init memblock_alloc_try_nid( 160626f09e9bSSantosh Shilimkar phys_addr_t size, phys_addr_t align, 160726f09e9bSSantosh Shilimkar phys_addr_t min_addr, phys_addr_t max_addr, 160826f09e9bSSantosh Shilimkar int nid) 160926f09e9bSSantosh Shilimkar { 161026f09e9bSSantosh Shilimkar void *ptr; 161126f09e9bSSantosh Shilimkar 1612d75f773cSSakari Ailus memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=%pa max_addr=%pa %pS\n", 1613a36aab89SMike Rapoport __func__, (u64)size, (u64)align, nid, &min_addr, 1614a36aab89SMike Rapoport &max_addr, (void *)_RET_IP_); 1615eb31d559SMike Rapoport ptr = memblock_alloc_internal(size, align, 16160ac398b1SYunfeng Ye min_addr, max_addr, nid, false); 1617c0dbe825SMike Rapoport if (ptr) 1618ea1f5f37SPavel Tatashin memset(ptr, 0, size); 161926f09e9bSSantosh Shilimkar 1620c0dbe825SMike Rapoport return ptr; 162126f09e9bSSantosh Shilimkar } 162226f09e9bSSantosh Shilimkar 162326f09e9bSSantosh Shilimkar /** 1624a2974133SMike Rapoport * __memblock_free_late - free pages directly to buddy allocator 162548a833ccSMike Rapoport * @base: phys starting address of the boot memory block 162626f09e9bSSantosh Shilimkar * @size: size of the boot memory block in bytes 162726f09e9bSSantosh Shilimkar * 1628a2974133SMike Rapoport * This is only useful when the memblock allocator has already been torn 162926f09e9bSSantosh Shilimkar * down, but we are still initializing the system. Pages are released directly 1630a2974133SMike Rapoport * to the buddy allocator. 163126f09e9bSSantosh Shilimkar */ 163226f09e9bSSantosh Shilimkar void __init __memblock_free_late(phys_addr_t base, phys_addr_t size) 163326f09e9bSSantosh Shilimkar { 1634a36aab89SMike Rapoport phys_addr_t cursor, end; 163526f09e9bSSantosh Shilimkar 1636a36aab89SMike Rapoport end = base + size - 1; 1637d75f773cSSakari Ailus memblock_dbg("%s: [%pa-%pa] %pS\n", 1638a36aab89SMike Rapoport __func__, &base, &end, (void *)_RET_IP_); 16399099daedSCatalin Marinas kmemleak_free_part_phys(base, size); 164026f09e9bSSantosh Shilimkar cursor = PFN_UP(base); 164126f09e9bSSantosh Shilimkar end = PFN_DOWN(base + size); 164226f09e9bSSantosh Shilimkar 164326f09e9bSSantosh Shilimkar for (; cursor < end; cursor++) { 16447c2ee349SMike Rapoport memblock_free_pages(pfn_to_page(cursor), cursor, 0); 1645ca79b0c2SArun KS totalram_pages_inc(); 164626f09e9bSSantosh Shilimkar } 164726f09e9bSSantosh Shilimkar } 16489d1e2492SBenjamin Herrenschmidt 16499d1e2492SBenjamin Herrenschmidt /* 16509d1e2492SBenjamin Herrenschmidt * Remaining API functions 16519d1e2492SBenjamin Herrenschmidt */ 16529d1e2492SBenjamin Herrenschmidt 16531f1ffb8aSDavid Gibson phys_addr_t __init_memblock memblock_phys_mem_size(void) 165495f72d1eSYinghai Lu { 16551440c4e2STejun Heo return memblock.memory.total_size; 165695f72d1eSYinghai Lu } 165795f72d1eSYinghai Lu 16588907de5dSSrikar Dronamraju phys_addr_t __init_memblock memblock_reserved_size(void) 16598907de5dSSrikar Dronamraju { 16608907de5dSSrikar Dronamraju return memblock.reserved.total_size; 16618907de5dSSrikar Dronamraju } 16628907de5dSSrikar Dronamraju 1663595ad9afSYinghai Lu phys_addr_t __init memblock_mem_size(unsigned long limit_pfn) 1664595ad9afSYinghai Lu { 1665595ad9afSYinghai Lu unsigned long pages = 0; 1666595ad9afSYinghai Lu struct memblock_region *r; 1667595ad9afSYinghai Lu unsigned long start_pfn, end_pfn; 1668595ad9afSYinghai Lu 1669595ad9afSYinghai Lu for_each_memblock(memory, r) { 1670595ad9afSYinghai Lu start_pfn = memblock_region_memory_base_pfn(r); 1671595ad9afSYinghai Lu end_pfn = memblock_region_memory_end_pfn(r); 1672595ad9afSYinghai Lu start_pfn = min_t(unsigned long, start_pfn, limit_pfn); 1673595ad9afSYinghai Lu end_pfn = min_t(unsigned long, end_pfn, limit_pfn); 1674595ad9afSYinghai Lu pages += end_pfn - start_pfn; 1675595ad9afSYinghai Lu } 1676595ad9afSYinghai Lu 167716763230SFabian Frederick return PFN_PHYS(pages); 1678595ad9afSYinghai Lu } 1679595ad9afSYinghai Lu 16800a93ebefSSam Ravnborg /* lowest address */ 16810a93ebefSSam Ravnborg phys_addr_t __init_memblock memblock_start_of_DRAM(void) 16820a93ebefSSam Ravnborg { 16830a93ebefSSam Ravnborg return memblock.memory.regions[0].base; 16840a93ebefSSam Ravnborg } 16850a93ebefSSam Ravnborg 168610d06439SYinghai Lu phys_addr_t __init_memblock memblock_end_of_DRAM(void) 168795f72d1eSYinghai Lu { 168895f72d1eSYinghai Lu int idx = memblock.memory.cnt - 1; 168995f72d1eSYinghai Lu 1690e3239ff9SBenjamin Herrenschmidt return (memblock.memory.regions[idx].base + memblock.memory.regions[idx].size); 169195f72d1eSYinghai Lu } 169295f72d1eSYinghai Lu 1693a571d4ebSDennis Chen static phys_addr_t __init_memblock __find_max_addr(phys_addr_t limit) 169495f72d1eSYinghai Lu { 16951c4bc43dSStefan Agner phys_addr_t max_addr = PHYS_ADDR_MAX; 1696136199f0SEmil Medve struct memblock_region *r; 169795f72d1eSYinghai Lu 1698a571d4ebSDennis Chen /* 1699a571d4ebSDennis Chen * translate the memory @limit size into the max address within one of 1700a571d4ebSDennis Chen * the memory memblock regions, if the @limit exceeds the total size 17011c4bc43dSStefan Agner * of those regions, max_addr will keep original value PHYS_ADDR_MAX 1702a571d4ebSDennis Chen */ 1703136199f0SEmil Medve for_each_memblock(memory, r) { 1704c0ce8fefSTejun Heo if (limit <= r->size) { 1705c0ce8fefSTejun Heo max_addr = r->base + limit; 170695f72d1eSYinghai Lu break; 170795f72d1eSYinghai Lu } 1708c0ce8fefSTejun Heo limit -= r->size; 170995f72d1eSYinghai Lu } 1710c0ce8fefSTejun Heo 1711a571d4ebSDennis Chen return max_addr; 1712a571d4ebSDennis Chen } 1713a571d4ebSDennis Chen 1714a571d4ebSDennis Chen void __init memblock_enforce_memory_limit(phys_addr_t limit) 1715a571d4ebSDennis Chen { 171649aef717SColin Ian King phys_addr_t max_addr; 1717a571d4ebSDennis Chen 1718a571d4ebSDennis Chen if (!limit) 1719a571d4ebSDennis Chen return; 1720a571d4ebSDennis Chen 1721a571d4ebSDennis Chen max_addr = __find_max_addr(limit); 1722a571d4ebSDennis Chen 1723a571d4ebSDennis Chen /* @limit exceeds the total size of the memory, do nothing */ 17241c4bc43dSStefan Agner if (max_addr == PHYS_ADDR_MAX) 1725a571d4ebSDennis Chen return; 1726a571d4ebSDennis Chen 1727c0ce8fefSTejun Heo /* truncate both memory and reserved regions */ 1728f1af9d3aSPhilipp Hachtmann memblock_remove_range(&memblock.memory, max_addr, 17291c4bc43dSStefan Agner PHYS_ADDR_MAX); 1730f1af9d3aSPhilipp Hachtmann memblock_remove_range(&memblock.reserved, max_addr, 17311c4bc43dSStefan Agner PHYS_ADDR_MAX); 173295f72d1eSYinghai Lu } 173395f72d1eSYinghai Lu 1734c9ca9b4eSAKASHI Takahiro void __init memblock_cap_memory_range(phys_addr_t base, phys_addr_t size) 1735c9ca9b4eSAKASHI Takahiro { 1736c9ca9b4eSAKASHI Takahiro int start_rgn, end_rgn; 1737c9ca9b4eSAKASHI Takahiro int i, ret; 1738c9ca9b4eSAKASHI Takahiro 1739c9ca9b4eSAKASHI Takahiro if (!size) 1740c9ca9b4eSAKASHI Takahiro return; 1741c9ca9b4eSAKASHI Takahiro 1742c9ca9b4eSAKASHI Takahiro ret = memblock_isolate_range(&memblock.memory, base, size, 1743c9ca9b4eSAKASHI Takahiro &start_rgn, &end_rgn); 1744c9ca9b4eSAKASHI Takahiro if (ret) 1745c9ca9b4eSAKASHI Takahiro return; 1746c9ca9b4eSAKASHI Takahiro 1747c9ca9b4eSAKASHI Takahiro /* remove all the MAP regions */ 1748c9ca9b4eSAKASHI Takahiro for (i = memblock.memory.cnt - 1; i >= end_rgn; i--) 1749c9ca9b4eSAKASHI Takahiro if (!memblock_is_nomap(&memblock.memory.regions[i])) 1750c9ca9b4eSAKASHI Takahiro memblock_remove_region(&memblock.memory, i); 1751c9ca9b4eSAKASHI Takahiro 1752c9ca9b4eSAKASHI Takahiro for (i = start_rgn - 1; i >= 0; i--) 1753c9ca9b4eSAKASHI Takahiro if (!memblock_is_nomap(&memblock.memory.regions[i])) 1754c9ca9b4eSAKASHI Takahiro memblock_remove_region(&memblock.memory, i); 1755c9ca9b4eSAKASHI Takahiro 1756c9ca9b4eSAKASHI Takahiro /* truncate the reserved regions */ 1757c9ca9b4eSAKASHI Takahiro memblock_remove_range(&memblock.reserved, 0, base); 1758c9ca9b4eSAKASHI Takahiro memblock_remove_range(&memblock.reserved, 17591c4bc43dSStefan Agner base + size, PHYS_ADDR_MAX); 1760c9ca9b4eSAKASHI Takahiro } 1761c9ca9b4eSAKASHI Takahiro 1762a571d4ebSDennis Chen void __init memblock_mem_limit_remove_map(phys_addr_t limit) 1763a571d4ebSDennis Chen { 1764a571d4ebSDennis Chen phys_addr_t max_addr; 1765a571d4ebSDennis Chen 1766a571d4ebSDennis Chen if (!limit) 1767a571d4ebSDennis Chen return; 1768a571d4ebSDennis Chen 1769a571d4ebSDennis Chen max_addr = __find_max_addr(limit); 1770a571d4ebSDennis Chen 1771a571d4ebSDennis Chen /* @limit exceeds the total size of the memory, do nothing */ 17721c4bc43dSStefan Agner if (max_addr == PHYS_ADDR_MAX) 1773a571d4ebSDennis Chen return; 1774a571d4ebSDennis Chen 1775c9ca9b4eSAKASHI Takahiro memblock_cap_memory_range(0, max_addr); 1776a571d4ebSDennis Chen } 1777a571d4ebSDennis Chen 1778cd79481dSYinghai Lu static int __init_memblock memblock_search(struct memblock_type *type, phys_addr_t addr) 177972d4b0b4SBenjamin Herrenschmidt { 178072d4b0b4SBenjamin Herrenschmidt unsigned int left = 0, right = type->cnt; 178172d4b0b4SBenjamin Herrenschmidt 178272d4b0b4SBenjamin Herrenschmidt do { 178372d4b0b4SBenjamin Herrenschmidt unsigned int mid = (right + left) / 2; 178472d4b0b4SBenjamin Herrenschmidt 178572d4b0b4SBenjamin Herrenschmidt if (addr < type->regions[mid].base) 178672d4b0b4SBenjamin Herrenschmidt right = mid; 178772d4b0b4SBenjamin Herrenschmidt else if (addr >= (type->regions[mid].base + 178872d4b0b4SBenjamin Herrenschmidt type->regions[mid].size)) 178972d4b0b4SBenjamin Herrenschmidt left = mid + 1; 179072d4b0b4SBenjamin Herrenschmidt else 179172d4b0b4SBenjamin Herrenschmidt return mid; 179272d4b0b4SBenjamin Herrenschmidt } while (left < right); 179372d4b0b4SBenjamin Herrenschmidt return -1; 179472d4b0b4SBenjamin Herrenschmidt } 179572d4b0b4SBenjamin Herrenschmidt 1796f5a222dcSYueyi Li bool __init_memblock memblock_is_reserved(phys_addr_t addr) 179795f72d1eSYinghai Lu { 179872d4b0b4SBenjamin Herrenschmidt return memblock_search(&memblock.reserved, addr) != -1; 179995f72d1eSYinghai Lu } 180072d4b0b4SBenjamin Herrenschmidt 1801b4ad0c7eSYaowei Bai bool __init_memblock memblock_is_memory(phys_addr_t addr) 180272d4b0b4SBenjamin Herrenschmidt { 180372d4b0b4SBenjamin Herrenschmidt return memblock_search(&memblock.memory, addr) != -1; 180472d4b0b4SBenjamin Herrenschmidt } 180572d4b0b4SBenjamin Herrenschmidt 1806937f0c26SYaowei Bai bool __init_memblock memblock_is_map_memory(phys_addr_t addr) 1807bf3d3cc5SArd Biesheuvel { 1808bf3d3cc5SArd Biesheuvel int i = memblock_search(&memblock.memory, addr); 1809bf3d3cc5SArd Biesheuvel 1810bf3d3cc5SArd Biesheuvel if (i == -1) 1811bf3d3cc5SArd Biesheuvel return false; 1812bf3d3cc5SArd Biesheuvel return !memblock_is_nomap(&memblock.memory.regions[i]); 1813bf3d3cc5SArd Biesheuvel } 1814bf3d3cc5SArd Biesheuvel 1815e76b63f8SYinghai Lu int __init_memblock memblock_search_pfn_nid(unsigned long pfn, 1816e76b63f8SYinghai Lu unsigned long *start_pfn, unsigned long *end_pfn) 1817e76b63f8SYinghai Lu { 1818e76b63f8SYinghai Lu struct memblock_type *type = &memblock.memory; 181916763230SFabian Frederick int mid = memblock_search(type, PFN_PHYS(pfn)); 1820e76b63f8SYinghai Lu 1821e76b63f8SYinghai Lu if (mid == -1) 1822e76b63f8SYinghai Lu return -1; 1823e76b63f8SYinghai Lu 1824f7e2f7e8SFabian Frederick *start_pfn = PFN_DOWN(type->regions[mid].base); 1825f7e2f7e8SFabian Frederick *end_pfn = PFN_DOWN(type->regions[mid].base + type->regions[mid].size); 1826e76b63f8SYinghai Lu 1827d622abf7SMike Rapoport return memblock_get_region_node(&type->regions[mid]); 1828e76b63f8SYinghai Lu } 1829e76b63f8SYinghai Lu 1830eab30949SStephen Boyd /** 1831eab30949SStephen Boyd * memblock_is_region_memory - check if a region is a subset of memory 1832eab30949SStephen Boyd * @base: base of region to check 1833eab30949SStephen Boyd * @size: size of region to check 1834eab30949SStephen Boyd * 1835eab30949SStephen Boyd * Check if the region [@base, @base + @size) is a subset of a memory block. 1836eab30949SStephen Boyd * 183747cec443SMike Rapoport * Return: 1838eab30949SStephen Boyd * 0 if false, non-zero if true 1839eab30949SStephen Boyd */ 1840937f0c26SYaowei Bai bool __init_memblock memblock_is_region_memory(phys_addr_t base, phys_addr_t size) 184172d4b0b4SBenjamin Herrenschmidt { 1842abb65272STomi Valkeinen int idx = memblock_search(&memblock.memory, base); 1843eb18f1b5STejun Heo phys_addr_t end = base + memblock_cap_size(base, &size); 184472d4b0b4SBenjamin Herrenschmidt 184572d4b0b4SBenjamin Herrenschmidt if (idx == -1) 1846937f0c26SYaowei Bai return false; 1847ef415ef4SWei Yang return (memblock.memory.regions[idx].base + 1848eb18f1b5STejun Heo memblock.memory.regions[idx].size) >= end; 184995f72d1eSYinghai Lu } 185095f72d1eSYinghai Lu 1851eab30949SStephen Boyd /** 1852eab30949SStephen Boyd * memblock_is_region_reserved - check if a region intersects reserved memory 1853eab30949SStephen Boyd * @base: base of region to check 1854eab30949SStephen Boyd * @size: size of region to check 1855eab30949SStephen Boyd * 185647cec443SMike Rapoport * Check if the region [@base, @base + @size) intersects a reserved 185747cec443SMike Rapoport * memory block. 1858eab30949SStephen Boyd * 185947cec443SMike Rapoport * Return: 1860c5c5c9d1STang Chen * True if they intersect, false if not. 1861eab30949SStephen Boyd */ 1862c5c5c9d1STang Chen bool __init_memblock memblock_is_region_reserved(phys_addr_t base, phys_addr_t size) 186395f72d1eSYinghai Lu { 1864eb18f1b5STejun Heo memblock_cap_size(base, &size); 1865c5c5c9d1STang Chen return memblock_overlaps_region(&memblock.reserved, base, size); 186695f72d1eSYinghai Lu } 186795f72d1eSYinghai Lu 18686ede1fd3SYinghai Lu void __init_memblock memblock_trim_memory(phys_addr_t align) 18696ede1fd3SYinghai Lu { 18706ede1fd3SYinghai Lu phys_addr_t start, end, orig_start, orig_end; 1871136199f0SEmil Medve struct memblock_region *r; 18726ede1fd3SYinghai Lu 1873136199f0SEmil Medve for_each_memblock(memory, r) { 1874136199f0SEmil Medve orig_start = r->base; 1875136199f0SEmil Medve orig_end = r->base + r->size; 18766ede1fd3SYinghai Lu start = round_up(orig_start, align); 18776ede1fd3SYinghai Lu end = round_down(orig_end, align); 18786ede1fd3SYinghai Lu 18796ede1fd3SYinghai Lu if (start == orig_start && end == orig_end) 18806ede1fd3SYinghai Lu continue; 18816ede1fd3SYinghai Lu 18826ede1fd3SYinghai Lu if (start < end) { 1883136199f0SEmil Medve r->base = start; 1884136199f0SEmil Medve r->size = end - start; 18856ede1fd3SYinghai Lu } else { 1886136199f0SEmil Medve memblock_remove_region(&memblock.memory, 1887136199f0SEmil Medve r - memblock.memory.regions); 1888136199f0SEmil Medve r--; 18896ede1fd3SYinghai Lu } 18906ede1fd3SYinghai Lu } 18916ede1fd3SYinghai Lu } 1892e63075a3SBenjamin Herrenschmidt 18933661ca66SYinghai Lu void __init_memblock memblock_set_current_limit(phys_addr_t limit) 1894e63075a3SBenjamin Herrenschmidt { 1895e63075a3SBenjamin Herrenschmidt memblock.current_limit = limit; 1896e63075a3SBenjamin Herrenschmidt } 1897e63075a3SBenjamin Herrenschmidt 1898fec51014SLaura Abbott phys_addr_t __init_memblock memblock_get_current_limit(void) 1899fec51014SLaura Abbott { 1900fec51014SLaura Abbott return memblock.current_limit; 1901fec51014SLaura Abbott } 1902fec51014SLaura Abbott 19030262d9c8SHeiko Carstens static void __init_memblock memblock_dump(struct memblock_type *type) 19046ed311b2SBenjamin Herrenschmidt { 19055d63f81cSMiles Chen phys_addr_t base, end, size; 1906e1720feeSMike Rapoport enum memblock_flags flags; 19078c9c1701SAlexander Kuleshov int idx; 19088c9c1701SAlexander Kuleshov struct memblock_region *rgn; 19096ed311b2SBenjamin Herrenschmidt 19100262d9c8SHeiko Carstens pr_info(" %s.cnt = 0x%lx\n", type->name, type->cnt); 19116ed311b2SBenjamin Herrenschmidt 191266e8b438SGioh Kim for_each_memblock_type(idx, type, rgn) { 19137c0caeb8STejun Heo char nid_buf[32] = ""; 19146ed311b2SBenjamin Herrenschmidt 19157c0caeb8STejun Heo base = rgn->base; 19167c0caeb8STejun Heo size = rgn->size; 19175d63f81cSMiles Chen end = base + size - 1; 191866a20757STang Chen flags = rgn->flags; 19193f08a302SMike Rapoport #ifdef CONFIG_NEED_MULTIPLE_NODES 19207c0caeb8STejun Heo if (memblock_get_region_node(rgn) != MAX_NUMNODES) 19217c0caeb8STejun Heo snprintf(nid_buf, sizeof(nid_buf), " on node %d", 19227c0caeb8STejun Heo memblock_get_region_node(rgn)); 19237c0caeb8STejun Heo #endif 1924e1720feeSMike Rapoport pr_info(" %s[%#x]\t[%pa-%pa], %pa bytes%s flags: %#x\n", 19250262d9c8SHeiko Carstens type->name, idx, &base, &end, &size, nid_buf, flags); 19266ed311b2SBenjamin Herrenschmidt } 19276ed311b2SBenjamin Herrenschmidt } 19286ed311b2SBenjamin Herrenschmidt 1929*87c55870SMike Rapoport static void __init_memblock __memblock_dump_all(void) 19306ed311b2SBenjamin Herrenschmidt { 19316ed311b2SBenjamin Herrenschmidt pr_info("MEMBLOCK configuration:\n"); 19325d63f81cSMiles Chen pr_info(" memory size = %pa reserved size = %pa\n", 19335d63f81cSMiles Chen &memblock.memory.total_size, 19345d63f81cSMiles Chen &memblock.reserved.total_size); 19356ed311b2SBenjamin Herrenschmidt 19360262d9c8SHeiko Carstens memblock_dump(&memblock.memory); 19370262d9c8SHeiko Carstens memblock_dump(&memblock.reserved); 1938409efd4cSHeiko Carstens #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP 193977649905SDavid Hildenbrand memblock_dump(&physmem); 1940409efd4cSHeiko Carstens #endif 19416ed311b2SBenjamin Herrenschmidt } 19426ed311b2SBenjamin Herrenschmidt 1943*87c55870SMike Rapoport void __init_memblock memblock_dump_all(void) 1944*87c55870SMike Rapoport { 1945*87c55870SMike Rapoport if (memblock_debug) 1946*87c55870SMike Rapoport __memblock_dump_all(); 1947*87c55870SMike Rapoport } 1948*87c55870SMike Rapoport 19491aadc056STejun Heo void __init memblock_allow_resize(void) 19506ed311b2SBenjamin Herrenschmidt { 1951142b45a7SBenjamin Herrenschmidt memblock_can_resize = 1; 19526ed311b2SBenjamin Herrenschmidt } 19536ed311b2SBenjamin Herrenschmidt 19546ed311b2SBenjamin Herrenschmidt static int __init early_memblock(char *p) 19556ed311b2SBenjamin Herrenschmidt { 19566ed311b2SBenjamin Herrenschmidt if (p && strstr(p, "debug")) 19576ed311b2SBenjamin Herrenschmidt memblock_debug = 1; 19586ed311b2SBenjamin Herrenschmidt return 0; 19596ed311b2SBenjamin Herrenschmidt } 19606ed311b2SBenjamin Herrenschmidt early_param("memblock", early_memblock); 19616ed311b2SBenjamin Herrenschmidt 1962bda49a81SMike Rapoport static void __init __free_pages_memory(unsigned long start, unsigned long end) 1963bda49a81SMike Rapoport { 1964bda49a81SMike Rapoport int order; 1965bda49a81SMike Rapoport 1966bda49a81SMike Rapoport while (start < end) { 1967bda49a81SMike Rapoport order = min(MAX_ORDER - 1UL, __ffs(start)); 1968bda49a81SMike Rapoport 1969bda49a81SMike Rapoport while (start + (1UL << order) > end) 1970bda49a81SMike Rapoport order--; 1971bda49a81SMike Rapoport 1972bda49a81SMike Rapoport memblock_free_pages(pfn_to_page(start), start, order); 1973bda49a81SMike Rapoport 1974bda49a81SMike Rapoport start += (1UL << order); 1975bda49a81SMike Rapoport } 1976bda49a81SMike Rapoport } 1977bda49a81SMike Rapoport 1978bda49a81SMike Rapoport static unsigned long __init __free_memory_core(phys_addr_t start, 1979bda49a81SMike Rapoport phys_addr_t end) 1980bda49a81SMike Rapoport { 1981bda49a81SMike Rapoport unsigned long start_pfn = PFN_UP(start); 1982bda49a81SMike Rapoport unsigned long end_pfn = min_t(unsigned long, 1983bda49a81SMike Rapoport PFN_DOWN(end), max_low_pfn); 1984bda49a81SMike Rapoport 1985bda49a81SMike Rapoport if (start_pfn >= end_pfn) 1986bda49a81SMike Rapoport return 0; 1987bda49a81SMike Rapoport 1988bda49a81SMike Rapoport __free_pages_memory(start_pfn, end_pfn); 1989bda49a81SMike Rapoport 1990bda49a81SMike Rapoport return end_pfn - start_pfn; 1991bda49a81SMike Rapoport } 1992bda49a81SMike Rapoport 1993bda49a81SMike Rapoport static unsigned long __init free_low_memory_core_early(void) 1994bda49a81SMike Rapoport { 1995bda49a81SMike Rapoport unsigned long count = 0; 1996bda49a81SMike Rapoport phys_addr_t start, end; 1997bda49a81SMike Rapoport u64 i; 1998bda49a81SMike Rapoport 1999bda49a81SMike Rapoport memblock_clear_hotplug(0, -1); 2000bda49a81SMike Rapoport 2001bda49a81SMike Rapoport for_each_reserved_mem_region(i, &start, &end) 2002bda49a81SMike Rapoport reserve_bootmem_region(start, end); 2003bda49a81SMike Rapoport 2004bda49a81SMike Rapoport /* 2005bda49a81SMike Rapoport * We need to use NUMA_NO_NODE instead of NODE_DATA(0)->node_id 2006bda49a81SMike Rapoport * because in some case like Node0 doesn't have RAM installed 2007bda49a81SMike Rapoport * low ram will be on Node1 2008bda49a81SMike Rapoport */ 2009bda49a81SMike Rapoport for_each_free_mem_range(i, NUMA_NO_NODE, MEMBLOCK_NONE, &start, &end, 2010bda49a81SMike Rapoport NULL) 2011bda49a81SMike Rapoport count += __free_memory_core(start, end); 2012bda49a81SMike Rapoport 2013bda49a81SMike Rapoport return count; 2014bda49a81SMike Rapoport } 2015bda49a81SMike Rapoport 2016bda49a81SMike Rapoport static int reset_managed_pages_done __initdata; 2017bda49a81SMike Rapoport 2018bda49a81SMike Rapoport void reset_node_managed_pages(pg_data_t *pgdat) 2019bda49a81SMike Rapoport { 2020bda49a81SMike Rapoport struct zone *z; 2021bda49a81SMike Rapoport 2022bda49a81SMike Rapoport for (z = pgdat->node_zones; z < pgdat->node_zones + MAX_NR_ZONES; z++) 20239705bea5SArun KS atomic_long_set(&z->managed_pages, 0); 2024bda49a81SMike Rapoport } 2025bda49a81SMike Rapoport 2026bda49a81SMike Rapoport void __init reset_all_zones_managed_pages(void) 2027bda49a81SMike Rapoport { 2028bda49a81SMike Rapoport struct pglist_data *pgdat; 2029bda49a81SMike Rapoport 2030bda49a81SMike Rapoport if (reset_managed_pages_done) 2031bda49a81SMike Rapoport return; 2032bda49a81SMike Rapoport 2033bda49a81SMike Rapoport for_each_online_pgdat(pgdat) 2034bda49a81SMike Rapoport reset_node_managed_pages(pgdat); 2035bda49a81SMike Rapoport 2036bda49a81SMike Rapoport reset_managed_pages_done = 1; 2037bda49a81SMike Rapoport } 2038bda49a81SMike Rapoport 2039bda49a81SMike Rapoport /** 2040bda49a81SMike Rapoport * memblock_free_all - release free pages to the buddy allocator 2041bda49a81SMike Rapoport * 2042bda49a81SMike Rapoport * Return: the number of pages actually released. 2043bda49a81SMike Rapoport */ 2044bda49a81SMike Rapoport unsigned long __init memblock_free_all(void) 2045bda49a81SMike Rapoport { 2046bda49a81SMike Rapoport unsigned long pages; 2047bda49a81SMike Rapoport 2048bda49a81SMike Rapoport reset_all_zones_managed_pages(); 2049bda49a81SMike Rapoport 2050bda49a81SMike Rapoport pages = free_low_memory_core_early(); 2051ca79b0c2SArun KS totalram_pages_add(pages); 2052bda49a81SMike Rapoport 2053bda49a81SMike Rapoport return pages; 2054bda49a81SMike Rapoport } 2055bda49a81SMike Rapoport 2056350e88baSMike Rapoport #if defined(CONFIG_DEBUG_FS) && defined(CONFIG_ARCH_KEEP_MEMBLOCK) 20576d03b885SBenjamin Herrenschmidt 20586d03b885SBenjamin Herrenschmidt static int memblock_debug_show(struct seq_file *m, void *private) 20596d03b885SBenjamin Herrenschmidt { 20606d03b885SBenjamin Herrenschmidt struct memblock_type *type = m->private; 20616d03b885SBenjamin Herrenschmidt struct memblock_region *reg; 20626d03b885SBenjamin Herrenschmidt int i; 20635d63f81cSMiles Chen phys_addr_t end; 20646d03b885SBenjamin Herrenschmidt 20656d03b885SBenjamin Herrenschmidt for (i = 0; i < type->cnt; i++) { 20666d03b885SBenjamin Herrenschmidt reg = &type->regions[i]; 20675d63f81cSMiles Chen end = reg->base + reg->size - 1; 20686d03b885SBenjamin Herrenschmidt 20695d63f81cSMiles Chen seq_printf(m, "%4d: ", i); 20705d63f81cSMiles Chen seq_printf(m, "%pa..%pa\n", ®->base, &end); 20716d03b885SBenjamin Herrenschmidt } 20726d03b885SBenjamin Herrenschmidt return 0; 20736d03b885SBenjamin Herrenschmidt } 20745ad35093SAndy Shevchenko DEFINE_SHOW_ATTRIBUTE(memblock_debug); 20756d03b885SBenjamin Herrenschmidt 20766d03b885SBenjamin Herrenschmidt static int __init memblock_init_debugfs(void) 20776d03b885SBenjamin Herrenschmidt { 20786d03b885SBenjamin Herrenschmidt struct dentry *root = debugfs_create_dir("memblock", NULL); 2079d9f7979cSGreg Kroah-Hartman 20800825a6f9SJoe Perches debugfs_create_file("memory", 0444, root, 20810825a6f9SJoe Perches &memblock.memory, &memblock_debug_fops); 20820825a6f9SJoe Perches debugfs_create_file("reserved", 0444, root, 20830825a6f9SJoe Perches &memblock.reserved, &memblock_debug_fops); 208470210ed9SPhilipp Hachtmann #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP 208577649905SDavid Hildenbrand debugfs_create_file("physmem", 0444, root, &physmem, 208677649905SDavid Hildenbrand &memblock_debug_fops); 208770210ed9SPhilipp Hachtmann #endif 20886d03b885SBenjamin Herrenschmidt 20896d03b885SBenjamin Herrenschmidt return 0; 20906d03b885SBenjamin Herrenschmidt } 20916d03b885SBenjamin Herrenschmidt __initcall(memblock_init_debugfs); 20926d03b885SBenjamin Herrenschmidt 20936d03b885SBenjamin Herrenschmidt #endif /* CONFIG_DEBUG_FS */ 2094