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 473e039c5cSMike Rapoport * * ``physmap`` - describes the actual physical memory regardless of 483e039c5cSMike Rapoport * the possible restrictions; the ``physmap`` type is only available 493e039c5cSMike Rapoport * 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 553e039c5cSMike Rapoport * the allocator metadata. The memory types are nicely wrapped with 563e039c5cSMike Rapoport * :c:type:`struct memblock`. This structure is statically initialzed 573e039c5cSMike Rapoport * at build time. The region arrays for the "memory" and "reserved" 583e039c5cSMike Rapoport * types are initially sized to %INIT_MEMBLOCK_REGIONS and for the 593e039c5cSMike Rapoport * "physmap" type to %INIT_PHYSMEM_REGIONS. 606e5af9a8SCao jin * The memblock_allow_resize() enables automatic resizing of the region 616e5af9a8SCao jin * arrays during addition of new regions. This feature should be used 626e5af9a8SCao jin * with care so that memory allocated for the region array will not 636e5af9a8SCao jin * overlap with areas that should be reserved, for example initrd. 643e039c5cSMike Rapoport * 653e039c5cSMike Rapoport * The early architecture setup should tell memblock what the physical 666e5af9a8SCao jin * memory layout is by using memblock_add() or memblock_add_node() 676e5af9a8SCao jin * functions. The first function does not assign the region to a NUMA 686e5af9a8SCao jin * node and it is appropriate for UMA systems. Yet, it is possible to 696e5af9a8SCao jin * use it on NUMA systems as well and assign the region to a NUMA node 706e5af9a8SCao jin * later in the setup process using memblock_set_node(). The 716e5af9a8SCao jin * memblock_add_node() performs such an assignment directly. 723e039c5cSMike Rapoport * 73a2974133SMike Rapoport * Once memblock is setup the memory can be allocated using one of the 74a2974133SMike Rapoport * API variants: 75a2974133SMike Rapoport * 766e5af9a8SCao jin * * memblock_phys_alloc*() - these functions return the **physical** 776e5af9a8SCao jin * address of the allocated memory 786e5af9a8SCao jin * * memblock_alloc*() - these functions return the **virtual** address 796e5af9a8SCao jin * of the allocated memory. 80a2974133SMike Rapoport * 81a2974133SMike Rapoport * Note, that both API variants use implict assumptions about allowed 82a2974133SMike Rapoport * memory ranges and the fallback methods. Consult the documentation 836e5af9a8SCao jin * of memblock_alloc_internal() and memblock_alloc_range_nid() 846e5af9a8SCao jin * functions for more elaborate description. 853e039c5cSMike Rapoport * 866e5af9a8SCao jin * As the system boot progresses, the architecture specific mem_init() 876e5af9a8SCao jin * function frees all the memory to the buddy page allocator. 883e039c5cSMike Rapoport * 896e5af9a8SCao jin * Unless an architecture enables %CONFIG_ARCH_KEEP_MEMBLOCK, the 903e039c5cSMike Rapoport * memblock data structures will be discarded after the system 916e5af9a8SCao jin * initialization completes. 923e039c5cSMike Rapoport */ 933e039c5cSMike Rapoport 94bda49a81SMike Rapoport #ifndef CONFIG_NEED_MULTIPLE_NODES 95bda49a81SMike Rapoport struct pglist_data __refdata contig_page_data; 96bda49a81SMike Rapoport EXPORT_SYMBOL(contig_page_data); 97bda49a81SMike Rapoport #endif 98bda49a81SMike Rapoport 99bda49a81SMike Rapoport unsigned long max_low_pfn; 100bda49a81SMike Rapoport unsigned long min_low_pfn; 101bda49a81SMike Rapoport unsigned long max_pfn; 102bda49a81SMike Rapoport unsigned long long max_possible_pfn; 103bda49a81SMike Rapoport 104fe091c20STejun Heo static struct memblock_region memblock_memory_init_regions[INIT_MEMBLOCK_REGIONS] __initdata_memblock; 1058a5b403dSArd Biesheuvel static struct memblock_region memblock_reserved_init_regions[INIT_MEMBLOCK_RESERVED_REGIONS] __initdata_memblock; 10670210ed9SPhilipp Hachtmann #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP 10770210ed9SPhilipp Hachtmann static struct memblock_region memblock_physmem_init_regions[INIT_PHYSMEM_REGIONS] __initdata_memblock; 10870210ed9SPhilipp Hachtmann #endif 109fe091c20STejun Heo 110fe091c20STejun Heo struct memblock memblock __initdata_memblock = { 111fe091c20STejun Heo .memory.regions = memblock_memory_init_regions, 112fe091c20STejun Heo .memory.cnt = 1, /* empty dummy entry */ 113fe091c20STejun Heo .memory.max = INIT_MEMBLOCK_REGIONS, 1140262d9c8SHeiko Carstens .memory.name = "memory", 115fe091c20STejun Heo 116fe091c20STejun Heo .reserved.regions = memblock_reserved_init_regions, 117fe091c20STejun Heo .reserved.cnt = 1, /* empty dummy entry */ 1188a5b403dSArd Biesheuvel .reserved.max = INIT_MEMBLOCK_RESERVED_REGIONS, 1190262d9c8SHeiko Carstens .reserved.name = "reserved", 120fe091c20STejun Heo 12170210ed9SPhilipp Hachtmann #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP 12270210ed9SPhilipp Hachtmann .physmem.regions = memblock_physmem_init_regions, 12370210ed9SPhilipp Hachtmann .physmem.cnt = 1, /* empty dummy entry */ 12470210ed9SPhilipp Hachtmann .physmem.max = INIT_PHYSMEM_REGIONS, 1250262d9c8SHeiko Carstens .physmem.name = "physmem", 12670210ed9SPhilipp Hachtmann #endif 12770210ed9SPhilipp Hachtmann 12879442ed1STang Chen .bottom_up = false, 129fe091c20STejun Heo .current_limit = MEMBLOCK_ALLOC_ANYWHERE, 130fe091c20STejun Heo }; 13195f72d1eSYinghai Lu 13210d06439SYinghai Lu int memblock_debug __initdata_memblock; 133a3f5bafcSTony Luck static bool system_has_some_mirror __initdata_memblock = false; 1341aadc056STejun Heo static int memblock_can_resize __initdata_memblock; 135181eb394SGavin Shan static int memblock_memory_in_slab __initdata_memblock = 0; 136181eb394SGavin Shan static int memblock_reserved_in_slab __initdata_memblock = 0; 13795f72d1eSYinghai Lu 138c366ea89SMike Rapoport static enum memblock_flags __init_memblock choose_memblock_flags(void) 139a3f5bafcSTony Luck { 140a3f5bafcSTony Luck return system_has_some_mirror ? MEMBLOCK_MIRROR : MEMBLOCK_NONE; 141a3f5bafcSTony Luck } 142a3f5bafcSTony Luck 143eb18f1b5STejun Heo /* adjust *@size so that (@base + *@size) doesn't overflow, return new size */ 144eb18f1b5STejun Heo static inline phys_addr_t memblock_cap_size(phys_addr_t base, phys_addr_t *size) 145eb18f1b5STejun Heo { 1461c4bc43dSStefan Agner return *size = min(*size, PHYS_ADDR_MAX - base); 147eb18f1b5STejun Heo } 148eb18f1b5STejun Heo 1496ed311b2SBenjamin Herrenschmidt /* 1506ed311b2SBenjamin Herrenschmidt * Address comparison utilities 1516ed311b2SBenjamin Herrenschmidt */ 15210d06439SYinghai Lu static unsigned long __init_memblock memblock_addrs_overlap(phys_addr_t base1, phys_addr_t size1, 1532898cc4cSBenjamin Herrenschmidt phys_addr_t base2, phys_addr_t size2) 15495f72d1eSYinghai Lu { 15595f72d1eSYinghai Lu return ((base1 < (base2 + size2)) && (base2 < (base1 + size1))); 15695f72d1eSYinghai Lu } 15795f72d1eSYinghai Lu 15895cf82ecSTang Chen bool __init_memblock memblock_overlaps_region(struct memblock_type *type, 1592d7d3eb2SH Hartley Sweeten phys_addr_t base, phys_addr_t size) 1606ed311b2SBenjamin Herrenschmidt { 1616ed311b2SBenjamin Herrenschmidt unsigned long i; 1626ed311b2SBenjamin Herrenschmidt 163f14516fbSAlexander Kuleshov for (i = 0; i < type->cnt; i++) 164f14516fbSAlexander Kuleshov if (memblock_addrs_overlap(base, size, type->regions[i].base, 165f14516fbSAlexander Kuleshov type->regions[i].size)) 1666ed311b2SBenjamin Herrenschmidt break; 167c5c5c9d1STang Chen return i < type->cnt; 1686ed311b2SBenjamin Herrenschmidt } 1696ed311b2SBenjamin Herrenschmidt 17047cec443SMike Rapoport /** 17179442ed1STang Chen * __memblock_find_range_bottom_up - find free area utility in bottom-up 17279442ed1STang Chen * @start: start of candidate range 17347cec443SMike Rapoport * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or 17447cec443SMike Rapoport * %MEMBLOCK_ALLOC_ACCESSIBLE 17579442ed1STang Chen * @size: size of free area to find 17679442ed1STang Chen * @align: alignment of free area to find 177b1154233SGrygorii Strashko * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 178fc6daaf9STony Luck * @flags: pick from blocks based on memory attributes 17979442ed1STang Chen * 18079442ed1STang Chen * Utility called from memblock_find_in_range_node(), find free area bottom-up. 18179442ed1STang Chen * 18247cec443SMike Rapoport * Return: 18379442ed1STang Chen * Found address on success, 0 on failure. 18479442ed1STang Chen */ 18579442ed1STang Chen static phys_addr_t __init_memblock 18679442ed1STang Chen __memblock_find_range_bottom_up(phys_addr_t start, phys_addr_t end, 187fc6daaf9STony Luck phys_addr_t size, phys_addr_t align, int nid, 188e1720feeSMike Rapoport enum memblock_flags flags) 18979442ed1STang Chen { 19079442ed1STang Chen phys_addr_t this_start, this_end, cand; 19179442ed1STang Chen u64 i; 19279442ed1STang Chen 193fc6daaf9STony Luck for_each_free_mem_range(i, nid, flags, &this_start, &this_end, NULL) { 19479442ed1STang Chen this_start = clamp(this_start, start, end); 19579442ed1STang Chen this_end = clamp(this_end, start, end); 19679442ed1STang Chen 19779442ed1STang Chen cand = round_up(this_start, align); 19879442ed1STang Chen if (cand < this_end && this_end - cand >= size) 19979442ed1STang Chen return cand; 20079442ed1STang Chen } 20179442ed1STang Chen 20279442ed1STang Chen return 0; 20379442ed1STang Chen } 20479442ed1STang Chen 2057bd0b0f0STejun Heo /** 2061402899eSTang Chen * __memblock_find_range_top_down - find free area utility, in top-down 2071402899eSTang Chen * @start: start of candidate range 20847cec443SMike Rapoport * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or 20947cec443SMike Rapoport * %MEMBLOCK_ALLOC_ACCESSIBLE 2101402899eSTang Chen * @size: size of free area to find 2111402899eSTang Chen * @align: alignment of free area to find 212b1154233SGrygorii Strashko * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 213fc6daaf9STony Luck * @flags: pick from blocks based on memory attributes 2141402899eSTang Chen * 2151402899eSTang Chen * Utility called from memblock_find_in_range_node(), find free area top-down. 2161402899eSTang Chen * 21747cec443SMike Rapoport * Return: 21879442ed1STang Chen * Found address on success, 0 on failure. 2191402899eSTang Chen */ 2201402899eSTang Chen static phys_addr_t __init_memblock 2211402899eSTang Chen __memblock_find_range_top_down(phys_addr_t start, phys_addr_t end, 222fc6daaf9STony Luck phys_addr_t size, phys_addr_t align, int nid, 223e1720feeSMike Rapoport enum memblock_flags flags) 2241402899eSTang Chen { 2251402899eSTang Chen phys_addr_t this_start, this_end, cand; 2261402899eSTang Chen u64 i; 2271402899eSTang Chen 228fc6daaf9STony Luck for_each_free_mem_range_reverse(i, nid, flags, &this_start, &this_end, 229fc6daaf9STony Luck NULL) { 2301402899eSTang Chen this_start = clamp(this_start, start, end); 2311402899eSTang Chen this_end = clamp(this_end, start, end); 2321402899eSTang Chen 2331402899eSTang Chen if (this_end < size) 2341402899eSTang Chen continue; 2351402899eSTang Chen 2361402899eSTang Chen cand = round_down(this_end - size, align); 2371402899eSTang Chen if (cand >= this_start) 2381402899eSTang Chen return cand; 2391402899eSTang Chen } 2401402899eSTang Chen 2411402899eSTang Chen return 0; 2421402899eSTang Chen } 2431402899eSTang Chen 2441402899eSTang Chen /** 2457bd0b0f0STejun Heo * memblock_find_in_range_node - find free area in given range and node 2467bd0b0f0STejun Heo * @size: size of free area to find 2477bd0b0f0STejun Heo * @align: alignment of free area to find 24887029ee9SGrygorii Strashko * @start: start of candidate range 24947cec443SMike Rapoport * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or 25047cec443SMike Rapoport * %MEMBLOCK_ALLOC_ACCESSIBLE 251b1154233SGrygorii Strashko * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 252fc6daaf9STony Luck * @flags: pick from blocks based on memory attributes 2537bd0b0f0STejun Heo * 2547bd0b0f0STejun Heo * Find @size free area aligned to @align in the specified range and node. 2557bd0b0f0STejun Heo * 25679442ed1STang Chen * When allocation direction is bottom-up, the @start should be greater 25779442ed1STang Chen * than the end of the kernel image. Otherwise, it will be trimmed. The 25879442ed1STang Chen * reason is that we want the bottom-up allocation just near the kernel 25979442ed1STang Chen * image so it is highly likely that the allocated memory and the kernel 26079442ed1STang Chen * will reside in the same node. 26179442ed1STang Chen * 26279442ed1STang Chen * If bottom-up allocation failed, will try to allocate memory top-down. 26379442ed1STang Chen * 26447cec443SMike Rapoport * Return: 26579442ed1STang Chen * Found address on success, 0 on failure. 2666ed311b2SBenjamin Herrenschmidt */ 267c366ea89SMike Rapoport static phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, 26887029ee9SGrygorii Strashko phys_addr_t align, phys_addr_t start, 269e1720feeSMike Rapoport phys_addr_t end, int nid, 270e1720feeSMike Rapoport enum memblock_flags flags) 271f7210e6cSTang Chen { 2720cfb8f0cSTang Chen phys_addr_t kernel_end, ret; 27379442ed1STang Chen 274f7210e6cSTang Chen /* pump up @end */ 275fed84c78SQian Cai if (end == MEMBLOCK_ALLOC_ACCESSIBLE || 276fed84c78SQian Cai end == MEMBLOCK_ALLOC_KASAN) 277f7210e6cSTang Chen end = memblock.current_limit; 278f7210e6cSTang Chen 279f7210e6cSTang Chen /* avoid allocating the first page */ 280f7210e6cSTang Chen start = max_t(phys_addr_t, start, PAGE_SIZE); 281f7210e6cSTang Chen end = max(start, end); 28279442ed1STang Chen kernel_end = __pa_symbol(_end); 28379442ed1STang Chen 28479442ed1STang Chen /* 28579442ed1STang Chen * try bottom-up allocation only when bottom-up mode 28679442ed1STang Chen * is set and @end is above the kernel image. 28779442ed1STang Chen */ 28879442ed1STang Chen if (memblock_bottom_up() && end > kernel_end) { 28979442ed1STang Chen phys_addr_t bottom_up_start; 29079442ed1STang Chen 29179442ed1STang Chen /* make sure we will allocate above the kernel */ 29279442ed1STang Chen bottom_up_start = max(start, kernel_end); 29379442ed1STang Chen 29479442ed1STang Chen /* ok, try bottom-up allocation first */ 29579442ed1STang Chen ret = __memblock_find_range_bottom_up(bottom_up_start, end, 296fc6daaf9STony Luck size, align, nid, flags); 29779442ed1STang Chen if (ret) 29879442ed1STang Chen return ret; 29979442ed1STang Chen 30079442ed1STang Chen /* 30179442ed1STang Chen * we always limit bottom-up allocation above the kernel, 30279442ed1STang Chen * but top-down allocation doesn't have the limit, so 30379442ed1STang Chen * retrying top-down allocation may succeed when bottom-up 30479442ed1STang Chen * allocation failed. 30579442ed1STang Chen * 30679442ed1STang Chen * bottom-up allocation is expected to be fail very rarely, 30779442ed1STang Chen * so we use WARN_ONCE() here to see the stack trace if 30879442ed1STang Chen * fail happens. 30979442ed1STang Chen */ 310e3d301caSMichal Hocko WARN_ONCE(IS_ENABLED(CONFIG_MEMORY_HOTREMOVE), 311e3d301caSMichal Hocko "memblock: bottom-up allocation failed, memory hotremove may be affected\n"); 31279442ed1STang Chen } 313f7210e6cSTang Chen 314fc6daaf9STony Luck return __memblock_find_range_top_down(start, end, size, align, nid, 315fc6daaf9STony Luck flags); 316f7210e6cSTang Chen } 3176ed311b2SBenjamin Herrenschmidt 3187bd0b0f0STejun Heo /** 3197bd0b0f0STejun Heo * memblock_find_in_range - find free area in given range 3207bd0b0f0STejun Heo * @start: start of candidate range 32147cec443SMike Rapoport * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or 32247cec443SMike Rapoport * %MEMBLOCK_ALLOC_ACCESSIBLE 3237bd0b0f0STejun Heo * @size: size of free area to find 3247bd0b0f0STejun Heo * @align: alignment of free area to find 3257bd0b0f0STejun Heo * 3267bd0b0f0STejun Heo * Find @size free area aligned to @align in the specified range. 3277bd0b0f0STejun Heo * 32847cec443SMike Rapoport * Return: 32979442ed1STang Chen * Found address on success, 0 on failure. 3307bd0b0f0STejun Heo */ 3317bd0b0f0STejun Heo phys_addr_t __init_memblock memblock_find_in_range(phys_addr_t start, 3327bd0b0f0STejun Heo phys_addr_t end, phys_addr_t size, 3337bd0b0f0STejun Heo phys_addr_t align) 3347bd0b0f0STejun Heo { 335a3f5bafcSTony Luck phys_addr_t ret; 336e1720feeSMike Rapoport enum memblock_flags flags = choose_memblock_flags(); 337a3f5bafcSTony Luck 338a3f5bafcSTony Luck again: 339a3f5bafcSTony Luck ret = memblock_find_in_range_node(size, align, start, end, 340a3f5bafcSTony Luck NUMA_NO_NODE, flags); 341a3f5bafcSTony Luck 342a3f5bafcSTony Luck if (!ret && (flags & MEMBLOCK_MIRROR)) { 343a3f5bafcSTony Luck pr_warn("Could not allocate %pap bytes of mirrored memory\n", 344a3f5bafcSTony Luck &size); 345a3f5bafcSTony Luck flags &= ~MEMBLOCK_MIRROR; 346a3f5bafcSTony Luck goto again; 347a3f5bafcSTony Luck } 348a3f5bafcSTony Luck 349a3f5bafcSTony Luck return ret; 3507bd0b0f0STejun Heo } 3517bd0b0f0STejun Heo 35210d06439SYinghai Lu static void __init_memblock memblock_remove_region(struct memblock_type *type, unsigned long r) 35395f72d1eSYinghai Lu { 3541440c4e2STejun Heo type->total_size -= type->regions[r].size; 3557c0caeb8STejun Heo memmove(&type->regions[r], &type->regions[r + 1], 3567c0caeb8STejun Heo (type->cnt - (r + 1)) * sizeof(type->regions[r])); 357e3239ff9SBenjamin Herrenschmidt type->cnt--; 35895f72d1eSYinghai Lu 3598f7a6605SBenjamin Herrenschmidt /* Special case for empty arrays */ 3608f7a6605SBenjamin Herrenschmidt if (type->cnt == 0) { 3611440c4e2STejun Heo WARN_ON(type->total_size != 0); 3628f7a6605SBenjamin Herrenschmidt type->cnt = 1; 3638f7a6605SBenjamin Herrenschmidt type->regions[0].base = 0; 3648f7a6605SBenjamin Herrenschmidt type->regions[0].size = 0; 36566a20757STang Chen type->regions[0].flags = 0; 3667c0caeb8STejun Heo memblock_set_region_node(&type->regions[0], MAX_NUMNODES); 3678f7a6605SBenjamin Herrenschmidt } 36895f72d1eSYinghai Lu } 36995f72d1eSYinghai Lu 370350e88baSMike Rapoport #ifndef CONFIG_ARCH_KEEP_MEMBLOCK 3713010f876SPavel Tatashin /** 37247cec443SMike Rapoport * memblock_discard - discard memory and reserved arrays if they were allocated 3733010f876SPavel Tatashin */ 3743010f876SPavel Tatashin void __init memblock_discard(void) 37529f67386SYinghai Lu { 3763010f876SPavel Tatashin phys_addr_t addr, size; 37729f67386SYinghai Lu 3783010f876SPavel Tatashin if (memblock.reserved.regions != memblock_reserved_init_regions) { 3793010f876SPavel Tatashin addr = __pa(memblock.reserved.regions); 3803010f876SPavel Tatashin size = PAGE_ALIGN(sizeof(struct memblock_region) * 38129f67386SYinghai Lu memblock.reserved.max); 3823010f876SPavel Tatashin __memblock_free_late(addr, size); 38329f67386SYinghai Lu } 38429f67386SYinghai Lu 38591b540f9SPavel Tatashin if (memblock.memory.regions != memblock_memory_init_regions) { 3863010f876SPavel Tatashin addr = __pa(memblock.memory.regions); 3873010f876SPavel Tatashin size = PAGE_ALIGN(sizeof(struct memblock_region) * 3885e270e25SPhilipp Hachtmann memblock.memory.max); 3893010f876SPavel Tatashin __memblock_free_late(addr, size); 3905e270e25SPhilipp Hachtmann } 3913010f876SPavel Tatashin } 3925e270e25SPhilipp Hachtmann #endif 3935e270e25SPhilipp Hachtmann 39448c3b583SGreg Pearson /** 39548c3b583SGreg Pearson * memblock_double_array - double the size of the memblock regions array 39648c3b583SGreg Pearson * @type: memblock type of the regions array being doubled 39748c3b583SGreg Pearson * @new_area_start: starting address of memory range to avoid overlap with 39848c3b583SGreg Pearson * @new_area_size: size of memory range to avoid overlap with 39948c3b583SGreg Pearson * 40048c3b583SGreg Pearson * Double the size of the @type regions array. If memblock is being used to 40148c3b583SGreg Pearson * allocate memory for a new reserved regions array and there is a previously 40248c3b583SGreg Pearson * allocated memory range [@new_area_start, @new_area_start + @new_area_size] 40348c3b583SGreg Pearson * waiting to be reserved, ensure the memory used by the new array does 40448c3b583SGreg Pearson * not overlap. 40548c3b583SGreg Pearson * 40647cec443SMike Rapoport * Return: 40748c3b583SGreg Pearson * 0 on success, -1 on failure. 40848c3b583SGreg Pearson */ 40948c3b583SGreg Pearson static int __init_memblock memblock_double_array(struct memblock_type *type, 41048c3b583SGreg Pearson phys_addr_t new_area_start, 41148c3b583SGreg Pearson phys_addr_t new_area_size) 412142b45a7SBenjamin Herrenschmidt { 413142b45a7SBenjamin Herrenschmidt struct memblock_region *new_array, *old_array; 41429f67386SYinghai Lu phys_addr_t old_alloc_size, new_alloc_size; 415a36aab89SMike Rapoport phys_addr_t old_size, new_size, addr, new_end; 416142b45a7SBenjamin Herrenschmidt int use_slab = slab_is_available(); 417181eb394SGavin Shan int *in_slab; 418142b45a7SBenjamin Herrenschmidt 419142b45a7SBenjamin Herrenschmidt /* We don't allow resizing until we know about the reserved regions 420142b45a7SBenjamin Herrenschmidt * of memory that aren't suitable for allocation 421142b45a7SBenjamin Herrenschmidt */ 422142b45a7SBenjamin Herrenschmidt if (!memblock_can_resize) 423142b45a7SBenjamin Herrenschmidt return -1; 424142b45a7SBenjamin Herrenschmidt 425142b45a7SBenjamin Herrenschmidt /* Calculate new doubled size */ 426142b45a7SBenjamin Herrenschmidt old_size = type->max * sizeof(struct memblock_region); 427142b45a7SBenjamin Herrenschmidt new_size = old_size << 1; 42829f67386SYinghai Lu /* 42929f67386SYinghai Lu * We need to allocated new one align to PAGE_SIZE, 43029f67386SYinghai Lu * so we can free them completely later. 43129f67386SYinghai Lu */ 43229f67386SYinghai Lu old_alloc_size = PAGE_ALIGN(old_size); 43329f67386SYinghai Lu new_alloc_size = PAGE_ALIGN(new_size); 434142b45a7SBenjamin Herrenschmidt 435181eb394SGavin Shan /* Retrieve the slab flag */ 436181eb394SGavin Shan if (type == &memblock.memory) 437181eb394SGavin Shan in_slab = &memblock_memory_in_slab; 438181eb394SGavin Shan else 439181eb394SGavin Shan in_slab = &memblock_reserved_in_slab; 440181eb394SGavin Shan 441a2974133SMike Rapoport /* Try to find some space for it */ 442142b45a7SBenjamin Herrenschmidt if (use_slab) { 443142b45a7SBenjamin Herrenschmidt new_array = kmalloc(new_size, GFP_KERNEL); 4441f5026a7STejun Heo addr = new_array ? __pa(new_array) : 0; 4454e2f0775SGavin Shan } else { 44648c3b583SGreg Pearson /* only exclude range when trying to double reserved.regions */ 44748c3b583SGreg Pearson if (type != &memblock.reserved) 44848c3b583SGreg Pearson new_area_start = new_area_size = 0; 44948c3b583SGreg Pearson 45048c3b583SGreg Pearson addr = memblock_find_in_range(new_area_start + new_area_size, 45148c3b583SGreg Pearson memblock.current_limit, 45229f67386SYinghai Lu new_alloc_size, PAGE_SIZE); 45348c3b583SGreg Pearson if (!addr && new_area_size) 45448c3b583SGreg Pearson addr = memblock_find_in_range(0, 45548c3b583SGreg Pearson min(new_area_start, memblock.current_limit), 45629f67386SYinghai Lu new_alloc_size, PAGE_SIZE); 45748c3b583SGreg Pearson 45815674868SSachin Kamat new_array = addr ? __va(addr) : NULL; 4594e2f0775SGavin Shan } 4601f5026a7STejun Heo if (!addr) { 461142b45a7SBenjamin Herrenschmidt pr_err("memblock: Failed to double %s array from %ld to %ld entries !\n", 4620262d9c8SHeiko Carstens type->name, type->max, type->max * 2); 463142b45a7SBenjamin Herrenschmidt return -1; 464142b45a7SBenjamin Herrenschmidt } 465142b45a7SBenjamin Herrenschmidt 466a36aab89SMike Rapoport new_end = addr + new_size - 1; 467a36aab89SMike Rapoport memblock_dbg("memblock: %s is doubled to %ld at [%pa-%pa]", 468a36aab89SMike Rapoport type->name, type->max * 2, &addr, &new_end); 469ea9e4376SYinghai Lu 470fd07383bSAndrew Morton /* 471fd07383bSAndrew Morton * Found space, we now need to move the array over before we add the 472fd07383bSAndrew Morton * reserved region since it may be our reserved array itself that is 473fd07383bSAndrew Morton * full. 474142b45a7SBenjamin Herrenschmidt */ 475142b45a7SBenjamin Herrenschmidt memcpy(new_array, type->regions, old_size); 476142b45a7SBenjamin Herrenschmidt memset(new_array + type->max, 0, old_size); 477142b45a7SBenjamin Herrenschmidt old_array = type->regions; 478142b45a7SBenjamin Herrenschmidt type->regions = new_array; 479142b45a7SBenjamin Herrenschmidt type->max <<= 1; 480142b45a7SBenjamin Herrenschmidt 481fd07383bSAndrew Morton /* Free old array. We needn't free it if the array is the static one */ 482181eb394SGavin Shan if (*in_slab) 483181eb394SGavin Shan kfree(old_array); 484181eb394SGavin Shan else if (old_array != memblock_memory_init_regions && 485142b45a7SBenjamin Herrenschmidt old_array != memblock_reserved_init_regions) 48629f67386SYinghai Lu memblock_free(__pa(old_array), old_alloc_size); 487142b45a7SBenjamin Herrenschmidt 488fd07383bSAndrew Morton /* 489fd07383bSAndrew Morton * Reserve the new array if that comes from the memblock. Otherwise, we 490fd07383bSAndrew Morton * needn't do it 491181eb394SGavin Shan */ 492181eb394SGavin Shan if (!use_slab) 49329f67386SYinghai Lu BUG_ON(memblock_reserve(addr, new_alloc_size)); 494181eb394SGavin Shan 495181eb394SGavin Shan /* Update slab flag */ 496181eb394SGavin Shan *in_slab = use_slab; 497181eb394SGavin Shan 498142b45a7SBenjamin Herrenschmidt return 0; 499142b45a7SBenjamin Herrenschmidt } 500142b45a7SBenjamin Herrenschmidt 501784656f9STejun Heo /** 502784656f9STejun Heo * memblock_merge_regions - merge neighboring compatible regions 503784656f9STejun Heo * @type: memblock type to scan 504784656f9STejun Heo * 505784656f9STejun Heo * Scan @type and merge neighboring compatible regions. 506784656f9STejun Heo */ 507784656f9STejun Heo static void __init_memblock memblock_merge_regions(struct memblock_type *type) 508784656f9STejun Heo { 509784656f9STejun Heo int i = 0; 510784656f9STejun Heo 511784656f9STejun Heo /* cnt never goes below 1 */ 512784656f9STejun Heo while (i < type->cnt - 1) { 513784656f9STejun Heo struct memblock_region *this = &type->regions[i]; 514784656f9STejun Heo struct memblock_region *next = &type->regions[i + 1]; 515784656f9STejun Heo 5167c0caeb8STejun Heo if (this->base + this->size != next->base || 5177c0caeb8STejun Heo memblock_get_region_node(this) != 51866a20757STang Chen memblock_get_region_node(next) || 51966a20757STang Chen this->flags != next->flags) { 520784656f9STejun Heo BUG_ON(this->base + this->size > next->base); 521784656f9STejun Heo i++; 522784656f9STejun Heo continue; 523784656f9STejun Heo } 524784656f9STejun Heo 525784656f9STejun Heo this->size += next->size; 526c0232ae8SLin Feng /* move forward from next + 1, index of which is i + 2 */ 527c0232ae8SLin Feng memmove(next, next + 1, (type->cnt - (i + 2)) * sizeof(*next)); 528784656f9STejun Heo type->cnt--; 529784656f9STejun Heo } 530784656f9STejun Heo } 531784656f9STejun Heo 532784656f9STejun Heo /** 533784656f9STejun Heo * memblock_insert_region - insert new memblock region 534784656f9STejun Heo * @type: memblock type to insert into 535784656f9STejun Heo * @idx: index for the insertion point 536784656f9STejun Heo * @base: base address of the new region 537784656f9STejun Heo * @size: size of the new region 538209ff86dSTang Chen * @nid: node id of the new region 53966a20757STang Chen * @flags: flags of the new region 540784656f9STejun Heo * 541784656f9STejun Heo * Insert new memblock region [@base, @base + @size) into @type at @idx. 542412d0008SAlexander Kuleshov * @type must already have extra room to accommodate the new region. 543784656f9STejun Heo */ 544784656f9STejun Heo static void __init_memblock memblock_insert_region(struct memblock_type *type, 545784656f9STejun Heo int idx, phys_addr_t base, 54666a20757STang Chen phys_addr_t size, 547e1720feeSMike Rapoport int nid, 548e1720feeSMike Rapoport enum memblock_flags flags) 549784656f9STejun Heo { 550784656f9STejun Heo struct memblock_region *rgn = &type->regions[idx]; 551784656f9STejun Heo 552784656f9STejun Heo BUG_ON(type->cnt >= type->max); 553784656f9STejun Heo memmove(rgn + 1, rgn, (type->cnt - idx) * sizeof(*rgn)); 554784656f9STejun Heo rgn->base = base; 555784656f9STejun Heo rgn->size = size; 55666a20757STang Chen rgn->flags = flags; 5577c0caeb8STejun Heo memblock_set_region_node(rgn, nid); 558784656f9STejun Heo type->cnt++; 5591440c4e2STejun Heo type->total_size += size; 560784656f9STejun Heo } 561784656f9STejun Heo 562784656f9STejun Heo /** 563f1af9d3aSPhilipp Hachtmann * memblock_add_range - add new memblock region 564784656f9STejun Heo * @type: memblock type to add new region into 565784656f9STejun Heo * @base: base address of the new region 566784656f9STejun Heo * @size: size of the new region 5677fb0bc3fSTejun Heo * @nid: nid of the new region 56866a20757STang Chen * @flags: flags of the new region 569784656f9STejun Heo * 570784656f9STejun Heo * Add new memblock region [@base, @base + @size) into @type. The new region 571784656f9STejun Heo * is allowed to overlap with existing ones - overlaps don't affect already 572784656f9STejun Heo * existing regions. @type is guaranteed to be minimal (all neighbouring 573784656f9STejun Heo * compatible regions are merged) after the addition. 574784656f9STejun Heo * 57547cec443SMike Rapoport * Return: 576784656f9STejun Heo * 0 on success, -errno on failure. 577784656f9STejun Heo */ 57802634a44SAnshuman Khandual static int __init_memblock memblock_add_range(struct memblock_type *type, 57966a20757STang Chen phys_addr_t base, phys_addr_t size, 580e1720feeSMike Rapoport int nid, enum memblock_flags flags) 58195f72d1eSYinghai Lu { 582784656f9STejun Heo bool insert = false; 583eb18f1b5STejun Heo phys_addr_t obase = base; 584eb18f1b5STejun Heo phys_addr_t end = base + memblock_cap_size(base, &size); 5858c9c1701SAlexander Kuleshov int idx, nr_new; 5868c9c1701SAlexander Kuleshov struct memblock_region *rgn; 58795f72d1eSYinghai Lu 588b3dc627cSTejun Heo if (!size) 589b3dc627cSTejun Heo return 0; 590b3dc627cSTejun Heo 591784656f9STejun Heo /* special case for empty array */ 592784656f9STejun Heo if (type->regions[0].size == 0) { 5931440c4e2STejun Heo WARN_ON(type->cnt != 1 || type->total_size); 594784656f9STejun Heo type->regions[0].base = base; 595784656f9STejun Heo type->regions[0].size = size; 59666a20757STang Chen type->regions[0].flags = flags; 5977fb0bc3fSTejun Heo memblock_set_region_node(&type->regions[0], nid); 5981440c4e2STejun Heo type->total_size = size; 599784656f9STejun Heo return 0; 600784656f9STejun Heo } 601784656f9STejun Heo repeat: 602784656f9STejun Heo /* 603784656f9STejun Heo * The following is executed twice. Once with %false @insert and 604784656f9STejun Heo * then with %true. The first counts the number of regions needed 605412d0008SAlexander Kuleshov * to accommodate the new area. The second actually inserts them. 606784656f9STejun Heo */ 607784656f9STejun Heo base = obase; 608784656f9STejun Heo nr_new = 0; 609784656f9STejun Heo 61066e8b438SGioh Kim for_each_memblock_type(idx, type, rgn) { 611784656f9STejun Heo phys_addr_t rbase = rgn->base; 612784656f9STejun Heo phys_addr_t rend = rbase + rgn->size; 6138f7a6605SBenjamin Herrenschmidt 614784656f9STejun Heo if (rbase >= end) 6158f7a6605SBenjamin Herrenschmidt break; 616784656f9STejun Heo if (rend <= base) 617784656f9STejun Heo continue; 618784656f9STejun Heo /* 619784656f9STejun Heo * @rgn overlaps. If it separates the lower part of new 620784656f9STejun Heo * area, insert that portion. 6218f7a6605SBenjamin Herrenschmidt */ 622784656f9STejun Heo if (rbase > base) { 623*3f08a302SMike Rapoport #ifdef CONFIG_NEED_MULTIPLE_NODES 624c0a29498SWei Yang WARN_ON(nid != memblock_get_region_node(rgn)); 625c0a29498SWei Yang #endif 6264fcab5f4SWei Yang WARN_ON(flags != rgn->flags); 627784656f9STejun Heo nr_new++; 628784656f9STejun Heo if (insert) 6298c9c1701SAlexander Kuleshov memblock_insert_region(type, idx++, base, 63066a20757STang Chen rbase - base, nid, 63166a20757STang Chen flags); 632784656f9STejun Heo } 633784656f9STejun Heo /* area below @rend is dealt with, forget about it */ 634784656f9STejun Heo base = min(rend, end); 6358f7a6605SBenjamin Herrenschmidt } 6368f7a6605SBenjamin Herrenschmidt 637784656f9STejun Heo /* insert the remaining portion */ 638784656f9STejun Heo if (base < end) { 639784656f9STejun Heo nr_new++; 640784656f9STejun Heo if (insert) 6418c9c1701SAlexander Kuleshov memblock_insert_region(type, idx, base, end - base, 64266a20757STang Chen nid, flags); 6438f7a6605SBenjamin Herrenschmidt } 6448f7a6605SBenjamin Herrenschmidt 645ef3cc4dbSnimisolo if (!nr_new) 646ef3cc4dbSnimisolo return 0; 647ef3cc4dbSnimisolo 648784656f9STejun Heo /* 649784656f9STejun Heo * If this was the first round, resize array and repeat for actual 650784656f9STejun Heo * insertions; otherwise, merge and return. 6518f7a6605SBenjamin Herrenschmidt */ 652784656f9STejun Heo if (!insert) { 653784656f9STejun Heo while (type->cnt + nr_new > type->max) 65448c3b583SGreg Pearson if (memblock_double_array(type, obase, size) < 0) 655784656f9STejun Heo return -ENOMEM; 656784656f9STejun Heo insert = true; 657784656f9STejun Heo goto repeat; 65895f72d1eSYinghai Lu } else { 659784656f9STejun Heo memblock_merge_regions(type); 66095f72d1eSYinghai Lu return 0; 66195f72d1eSYinghai Lu } 662784656f9STejun Heo } 66395f72d1eSYinghai Lu 66448a833ccSMike Rapoport /** 66548a833ccSMike Rapoport * memblock_add_node - add new memblock region within a NUMA node 66648a833ccSMike Rapoport * @base: base address of the new region 66748a833ccSMike Rapoport * @size: size of the new region 66848a833ccSMike Rapoport * @nid: nid of the new region 66948a833ccSMike Rapoport * 67048a833ccSMike Rapoport * Add new memblock region [@base, @base + @size) to the "memory" 67148a833ccSMike Rapoport * type. See memblock_add_range() description for mode details 67248a833ccSMike Rapoport * 67348a833ccSMike Rapoport * Return: 67448a833ccSMike Rapoport * 0 on success, -errno on failure. 67548a833ccSMike Rapoport */ 6767fb0bc3fSTejun Heo int __init_memblock memblock_add_node(phys_addr_t base, phys_addr_t size, 6777fb0bc3fSTejun Heo int nid) 6787fb0bc3fSTejun Heo { 679f1af9d3aSPhilipp Hachtmann return memblock_add_range(&memblock.memory, base, size, nid, 0); 6807fb0bc3fSTejun Heo } 6817fb0bc3fSTejun Heo 68248a833ccSMike Rapoport /** 68348a833ccSMike Rapoport * memblock_add - add new memblock region 68448a833ccSMike Rapoport * @base: base address of the new region 68548a833ccSMike Rapoport * @size: size of the new region 68648a833ccSMike Rapoport * 68748a833ccSMike Rapoport * Add new memblock region [@base, @base + @size) to the "memory" 68848a833ccSMike Rapoport * type. See memblock_add_range() description for mode details 68948a833ccSMike Rapoport * 69048a833ccSMike Rapoport * Return: 69148a833ccSMike Rapoport * 0 on success, -errno on failure. 69248a833ccSMike Rapoport */ 693f705ac4bSAlexander Kuleshov int __init_memblock memblock_add(phys_addr_t base, phys_addr_t size) 6946a4055bcSAlexander Kuleshov { 6955d63f81cSMiles Chen phys_addr_t end = base + size - 1; 6965d63f81cSMiles Chen 697a090d711SAnshuman Khandual memblock_dbg("%s: [%pa-%pa] %pS\n", __func__, 6985d63f81cSMiles Chen &base, &end, (void *)_RET_IP_); 6996a4055bcSAlexander Kuleshov 700f705ac4bSAlexander Kuleshov return memblock_add_range(&memblock.memory, base, size, MAX_NUMNODES, 0); 70195f72d1eSYinghai Lu } 70295f72d1eSYinghai Lu 7036a9ceb31STejun Heo /** 7046a9ceb31STejun Heo * memblock_isolate_range - isolate given range into disjoint memblocks 7056a9ceb31STejun Heo * @type: memblock type to isolate range for 7066a9ceb31STejun Heo * @base: base of range to isolate 7076a9ceb31STejun Heo * @size: size of range to isolate 7086a9ceb31STejun Heo * @start_rgn: out parameter for the start of isolated region 7096a9ceb31STejun Heo * @end_rgn: out parameter for the end of isolated region 7106a9ceb31STejun Heo * 7116a9ceb31STejun Heo * Walk @type and ensure that regions don't cross the boundaries defined by 7126a9ceb31STejun Heo * [@base, @base + @size). Crossing regions are split at the boundaries, 7136a9ceb31STejun Heo * which may create at most two more regions. The index of the first 7146a9ceb31STejun Heo * region inside the range is returned in *@start_rgn and end in *@end_rgn. 7156a9ceb31STejun Heo * 71647cec443SMike Rapoport * Return: 7176a9ceb31STejun Heo * 0 on success, -errno on failure. 7186a9ceb31STejun Heo */ 7196a9ceb31STejun Heo static int __init_memblock memblock_isolate_range(struct memblock_type *type, 7206a9ceb31STejun Heo phys_addr_t base, phys_addr_t size, 7216a9ceb31STejun Heo int *start_rgn, int *end_rgn) 7226a9ceb31STejun Heo { 723eb18f1b5STejun Heo phys_addr_t end = base + memblock_cap_size(base, &size); 7248c9c1701SAlexander Kuleshov int idx; 7258c9c1701SAlexander Kuleshov struct memblock_region *rgn; 7266a9ceb31STejun Heo 7276a9ceb31STejun Heo *start_rgn = *end_rgn = 0; 7286a9ceb31STejun Heo 729b3dc627cSTejun Heo if (!size) 730b3dc627cSTejun Heo return 0; 731b3dc627cSTejun Heo 7326a9ceb31STejun Heo /* we'll create at most two more regions */ 7336a9ceb31STejun Heo while (type->cnt + 2 > type->max) 73448c3b583SGreg Pearson if (memblock_double_array(type, base, size) < 0) 7356a9ceb31STejun Heo return -ENOMEM; 7366a9ceb31STejun Heo 73766e8b438SGioh Kim for_each_memblock_type(idx, type, rgn) { 7386a9ceb31STejun Heo phys_addr_t rbase = rgn->base; 7396a9ceb31STejun Heo phys_addr_t rend = rbase + rgn->size; 7406a9ceb31STejun Heo 7416a9ceb31STejun Heo if (rbase >= end) 7426a9ceb31STejun Heo break; 7436a9ceb31STejun Heo if (rend <= base) 7446a9ceb31STejun Heo continue; 7456a9ceb31STejun Heo 7466a9ceb31STejun Heo if (rbase < base) { 7476a9ceb31STejun Heo /* 7486a9ceb31STejun Heo * @rgn intersects from below. Split and continue 7496a9ceb31STejun Heo * to process the next region - the new top half. 7506a9ceb31STejun Heo */ 7516a9ceb31STejun Heo rgn->base = base; 7521440c4e2STejun Heo rgn->size -= base - rbase; 7531440c4e2STejun Heo type->total_size -= base - rbase; 7548c9c1701SAlexander Kuleshov memblock_insert_region(type, idx, rbase, base - rbase, 75566a20757STang Chen memblock_get_region_node(rgn), 75666a20757STang Chen rgn->flags); 7576a9ceb31STejun Heo } else if (rend > end) { 7586a9ceb31STejun Heo /* 7596a9ceb31STejun Heo * @rgn intersects from above. Split and redo the 7606a9ceb31STejun Heo * current region - the new bottom half. 7616a9ceb31STejun Heo */ 7626a9ceb31STejun Heo rgn->base = end; 7631440c4e2STejun Heo rgn->size -= end - rbase; 7641440c4e2STejun Heo type->total_size -= end - rbase; 7658c9c1701SAlexander Kuleshov memblock_insert_region(type, idx--, rbase, end - rbase, 76666a20757STang Chen memblock_get_region_node(rgn), 76766a20757STang Chen rgn->flags); 7686a9ceb31STejun Heo } else { 7696a9ceb31STejun Heo /* @rgn is fully contained, record it */ 7706a9ceb31STejun Heo if (!*end_rgn) 7718c9c1701SAlexander Kuleshov *start_rgn = idx; 7728c9c1701SAlexander Kuleshov *end_rgn = idx + 1; 7736a9ceb31STejun Heo } 7746a9ceb31STejun Heo } 7756a9ceb31STejun Heo 7766a9ceb31STejun Heo return 0; 7776a9ceb31STejun Heo } 7786a9ceb31STejun Heo 77935bd16a2SAlexander Kuleshov static int __init_memblock memblock_remove_range(struct memblock_type *type, 7808f7a6605SBenjamin Herrenschmidt phys_addr_t base, phys_addr_t size) 78195f72d1eSYinghai Lu { 78271936180STejun Heo int start_rgn, end_rgn; 78371936180STejun Heo int i, ret; 78495f72d1eSYinghai Lu 78571936180STejun Heo ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); 78671936180STejun Heo if (ret) 78771936180STejun Heo return ret; 78895f72d1eSYinghai Lu 78971936180STejun Heo for (i = end_rgn - 1; i >= start_rgn; i--) 79071936180STejun Heo memblock_remove_region(type, i); 79195f72d1eSYinghai Lu return 0; 79295f72d1eSYinghai Lu } 79395f72d1eSYinghai Lu 794581adcbeSTejun Heo int __init_memblock memblock_remove(phys_addr_t base, phys_addr_t size) 79595f72d1eSYinghai Lu { 79625cf23d7SMinchan Kim phys_addr_t end = base + size - 1; 79725cf23d7SMinchan Kim 798a090d711SAnshuman Khandual memblock_dbg("%s: [%pa-%pa] %pS\n", __func__, 79925cf23d7SMinchan Kim &base, &end, (void *)_RET_IP_); 80025cf23d7SMinchan Kim 801f1af9d3aSPhilipp Hachtmann return memblock_remove_range(&memblock.memory, base, size); 80295f72d1eSYinghai Lu } 80395f72d1eSYinghai Lu 8044d72868cSMike Rapoport /** 8054d72868cSMike Rapoport * memblock_free - free boot memory block 8064d72868cSMike Rapoport * @base: phys starting address of the boot memory block 8074d72868cSMike Rapoport * @size: size of the boot memory block in bytes 8084d72868cSMike Rapoport * 8094d72868cSMike Rapoport * Free boot memory block previously allocated by memblock_alloc_xx() API. 8104d72868cSMike Rapoport * The freeing memory will not be released to the buddy allocator. 8114d72868cSMike Rapoport */ 812581adcbeSTejun Heo int __init_memblock memblock_free(phys_addr_t base, phys_addr_t size) 81395f72d1eSYinghai Lu { 8145d63f81cSMiles Chen phys_addr_t end = base + size - 1; 8155d63f81cSMiles Chen 816a090d711SAnshuman Khandual memblock_dbg("%s: [%pa-%pa] %pS\n", __func__, 8175d63f81cSMiles Chen &base, &end, (void *)_RET_IP_); 81824aa0788STejun Heo 8199099daedSCatalin Marinas kmemleak_free_part_phys(base, size); 820f1af9d3aSPhilipp Hachtmann return memblock_remove_range(&memblock.reserved, base, size); 82195f72d1eSYinghai Lu } 82295f72d1eSYinghai Lu 823f705ac4bSAlexander Kuleshov int __init_memblock memblock_reserve(phys_addr_t base, phys_addr_t size) 82495f72d1eSYinghai Lu { 8255d63f81cSMiles Chen phys_addr_t end = base + size - 1; 8265d63f81cSMiles Chen 827a090d711SAnshuman Khandual memblock_dbg("%s: [%pa-%pa] %pS\n", __func__, 8285d63f81cSMiles Chen &base, &end, (void *)_RET_IP_); 82995f72d1eSYinghai Lu 830f705ac4bSAlexander Kuleshov return memblock_add_range(&memblock.reserved, base, size, MAX_NUMNODES, 0); 83195f72d1eSYinghai Lu } 83295f72d1eSYinghai Lu 83302634a44SAnshuman Khandual #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP 83402634a44SAnshuman Khandual int __init_memblock memblock_physmem_add(phys_addr_t base, phys_addr_t size) 83502634a44SAnshuman Khandual { 83602634a44SAnshuman Khandual phys_addr_t end = base + size - 1; 83702634a44SAnshuman Khandual 83802634a44SAnshuman Khandual memblock_dbg("%s: [%pa-%pa] %pS\n", __func__, 83902634a44SAnshuman Khandual &base, &end, (void *)_RET_IP_); 84002634a44SAnshuman Khandual 84102634a44SAnshuman Khandual return memblock_add_range(&memblock.physmem, base, size, MAX_NUMNODES, 0); 84202634a44SAnshuman Khandual } 84302634a44SAnshuman Khandual #endif 84402634a44SAnshuman Khandual 84535fd0808STejun Heo /** 84647cec443SMike Rapoport * memblock_setclr_flag - set or clear flag for a memory region 84747cec443SMike Rapoport * @base: base address of the region 84847cec443SMike Rapoport * @size: size of the region 84947cec443SMike Rapoport * @set: set or clear the flag 85047cec443SMike Rapoport * @flag: the flag to udpate 85166b16edfSTang Chen * 8524308ce17STony Luck * This function isolates region [@base, @base + @size), and sets/clears flag 85366b16edfSTang Chen * 85447cec443SMike Rapoport * Return: 0 on success, -errno on failure. 85566b16edfSTang Chen */ 8564308ce17STony Luck static int __init_memblock memblock_setclr_flag(phys_addr_t base, 8574308ce17STony Luck phys_addr_t size, int set, int flag) 85866b16edfSTang Chen { 85966b16edfSTang Chen struct memblock_type *type = &memblock.memory; 86066b16edfSTang Chen int i, ret, start_rgn, end_rgn; 86166b16edfSTang Chen 86266b16edfSTang Chen ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); 86366b16edfSTang Chen if (ret) 86466b16edfSTang Chen return ret; 86566b16edfSTang Chen 866fe145124SMike Rapoport for (i = start_rgn; i < end_rgn; i++) { 867fe145124SMike Rapoport struct memblock_region *r = &type->regions[i]; 868fe145124SMike Rapoport 8694308ce17STony Luck if (set) 870fe145124SMike Rapoport r->flags |= flag; 8714308ce17STony Luck else 872fe145124SMike Rapoport r->flags &= ~flag; 873fe145124SMike Rapoport } 87466b16edfSTang Chen 87566b16edfSTang Chen memblock_merge_regions(type); 87666b16edfSTang Chen return 0; 87766b16edfSTang Chen } 87866b16edfSTang Chen 87966b16edfSTang Chen /** 8804308ce17STony Luck * memblock_mark_hotplug - Mark hotpluggable memory with flag MEMBLOCK_HOTPLUG. 8814308ce17STony Luck * @base: the base phys addr of the region 8824308ce17STony Luck * @size: the size of the region 8834308ce17STony Luck * 88447cec443SMike Rapoport * Return: 0 on success, -errno on failure. 8854308ce17STony Luck */ 8864308ce17STony Luck int __init_memblock memblock_mark_hotplug(phys_addr_t base, phys_addr_t size) 8874308ce17STony Luck { 8884308ce17STony Luck return memblock_setclr_flag(base, size, 1, MEMBLOCK_HOTPLUG); 8894308ce17STony Luck } 8904308ce17STony Luck 8914308ce17STony Luck /** 89266b16edfSTang Chen * memblock_clear_hotplug - Clear flag MEMBLOCK_HOTPLUG for a specified region. 89366b16edfSTang Chen * @base: the base phys addr of the region 89466b16edfSTang Chen * @size: the size of the region 89566b16edfSTang Chen * 89647cec443SMike Rapoport * Return: 0 on success, -errno on failure. 89766b16edfSTang Chen */ 89866b16edfSTang Chen int __init_memblock memblock_clear_hotplug(phys_addr_t base, phys_addr_t size) 89966b16edfSTang Chen { 9004308ce17STony Luck return memblock_setclr_flag(base, size, 0, MEMBLOCK_HOTPLUG); 90166b16edfSTang Chen } 90266b16edfSTang Chen 90366b16edfSTang Chen /** 904a3f5bafcSTony Luck * memblock_mark_mirror - Mark mirrored memory with flag MEMBLOCK_MIRROR. 905a3f5bafcSTony Luck * @base: the base phys addr of the region 906a3f5bafcSTony Luck * @size: the size of the region 907a3f5bafcSTony Luck * 90847cec443SMike Rapoport * Return: 0 on success, -errno on failure. 909a3f5bafcSTony Luck */ 910a3f5bafcSTony Luck int __init_memblock memblock_mark_mirror(phys_addr_t base, phys_addr_t size) 911a3f5bafcSTony Luck { 912a3f5bafcSTony Luck system_has_some_mirror = true; 913a3f5bafcSTony Luck 914a3f5bafcSTony Luck return memblock_setclr_flag(base, size, 1, MEMBLOCK_MIRROR); 915a3f5bafcSTony Luck } 916a3f5bafcSTony Luck 917bf3d3cc5SArd Biesheuvel /** 918bf3d3cc5SArd Biesheuvel * memblock_mark_nomap - Mark a memory region with flag MEMBLOCK_NOMAP. 919bf3d3cc5SArd Biesheuvel * @base: the base phys addr of the region 920bf3d3cc5SArd Biesheuvel * @size: the size of the region 921bf3d3cc5SArd Biesheuvel * 92247cec443SMike Rapoport * Return: 0 on success, -errno on failure. 923bf3d3cc5SArd Biesheuvel */ 924bf3d3cc5SArd Biesheuvel int __init_memblock memblock_mark_nomap(phys_addr_t base, phys_addr_t size) 925bf3d3cc5SArd Biesheuvel { 926bf3d3cc5SArd Biesheuvel return memblock_setclr_flag(base, size, 1, MEMBLOCK_NOMAP); 927bf3d3cc5SArd Biesheuvel } 928a3f5bafcSTony Luck 929a3f5bafcSTony Luck /** 9304c546b8aSAKASHI Takahiro * memblock_clear_nomap - Clear flag MEMBLOCK_NOMAP for a specified region. 9314c546b8aSAKASHI Takahiro * @base: the base phys addr of the region 9324c546b8aSAKASHI Takahiro * @size: the size of the region 9334c546b8aSAKASHI Takahiro * 93447cec443SMike Rapoport * Return: 0 on success, -errno on failure. 9354c546b8aSAKASHI Takahiro */ 9364c546b8aSAKASHI Takahiro int __init_memblock memblock_clear_nomap(phys_addr_t base, phys_addr_t size) 9374c546b8aSAKASHI Takahiro { 9384c546b8aSAKASHI Takahiro return memblock_setclr_flag(base, size, 0, MEMBLOCK_NOMAP); 9394c546b8aSAKASHI Takahiro } 9404c546b8aSAKASHI Takahiro 9414c546b8aSAKASHI Takahiro /** 9428e7a7f86SRobin Holt * __next_reserved_mem_region - next function for for_each_reserved_region() 9438e7a7f86SRobin Holt * @idx: pointer to u64 loop variable 9448e7a7f86SRobin Holt * @out_start: ptr to phys_addr_t for start address of the region, can be %NULL 9458e7a7f86SRobin Holt * @out_end: ptr to phys_addr_t for end address of the region, can be %NULL 9468e7a7f86SRobin Holt * 9478e7a7f86SRobin Holt * Iterate over all reserved memory regions. 9488e7a7f86SRobin Holt */ 9498e7a7f86SRobin Holt void __init_memblock __next_reserved_mem_region(u64 *idx, 9508e7a7f86SRobin Holt phys_addr_t *out_start, 9518e7a7f86SRobin Holt phys_addr_t *out_end) 9528e7a7f86SRobin Holt { 953567d117bSAlexander Kuleshov struct memblock_type *type = &memblock.reserved; 9548e7a7f86SRobin Holt 955cd33a76bSRichard Leitner if (*idx < type->cnt) { 956567d117bSAlexander Kuleshov struct memblock_region *r = &type->regions[*idx]; 9578e7a7f86SRobin Holt phys_addr_t base = r->base; 9588e7a7f86SRobin Holt phys_addr_t size = r->size; 9598e7a7f86SRobin Holt 9608e7a7f86SRobin Holt if (out_start) 9618e7a7f86SRobin Holt *out_start = base; 9628e7a7f86SRobin Holt if (out_end) 9638e7a7f86SRobin Holt *out_end = base + size - 1; 9648e7a7f86SRobin Holt 9658e7a7f86SRobin Holt *idx += 1; 9668e7a7f86SRobin Holt return; 9678e7a7f86SRobin Holt } 9688e7a7f86SRobin Holt 9698e7a7f86SRobin Holt /* signal end of iteration */ 9708e7a7f86SRobin Holt *idx = ULLONG_MAX; 9718e7a7f86SRobin Holt } 9728e7a7f86SRobin Holt 973c9a688a3SMike Rapoport static bool should_skip_region(struct memblock_region *m, int nid, int flags) 974c9a688a3SMike Rapoport { 975c9a688a3SMike Rapoport int m_nid = memblock_get_region_node(m); 976c9a688a3SMike Rapoport 977c9a688a3SMike Rapoport /* only memory regions are associated with nodes, check it */ 978c9a688a3SMike Rapoport if (nid != NUMA_NO_NODE && nid != m_nid) 979c9a688a3SMike Rapoport return true; 980c9a688a3SMike Rapoport 981c9a688a3SMike Rapoport /* skip hotpluggable memory regions if needed */ 982c9a688a3SMike Rapoport if (movable_node_is_enabled() && memblock_is_hotpluggable(m)) 983c9a688a3SMike Rapoport return true; 984c9a688a3SMike Rapoport 985c9a688a3SMike Rapoport /* if we want mirror memory skip non-mirror memory regions */ 986c9a688a3SMike Rapoport if ((flags & MEMBLOCK_MIRROR) && !memblock_is_mirror(m)) 987c9a688a3SMike Rapoport return true; 988c9a688a3SMike Rapoport 989c9a688a3SMike Rapoport /* skip nomap memory unless we were asked for it explicitly */ 990c9a688a3SMike Rapoport if (!(flags & MEMBLOCK_NOMAP) && memblock_is_nomap(m)) 991c9a688a3SMike Rapoport return true; 992c9a688a3SMike Rapoport 993c9a688a3SMike Rapoport return false; 994c9a688a3SMike Rapoport } 995c9a688a3SMike Rapoport 9968e7a7f86SRobin Holt /** 997a2974133SMike Rapoport * __next_mem_range - next function for for_each_free_mem_range() etc. 99835fd0808STejun Heo * @idx: pointer to u64 loop variable 999b1154233SGrygorii Strashko * @nid: node selector, %NUMA_NO_NODE for all nodes 1000fc6daaf9STony Luck * @flags: pick from blocks based on memory attributes 1001f1af9d3aSPhilipp Hachtmann * @type_a: pointer to memblock_type from where the range is taken 1002f1af9d3aSPhilipp Hachtmann * @type_b: pointer to memblock_type which excludes memory from being taken 1003dad7557eSWanpeng Li * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL 1004dad7557eSWanpeng Li * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL 1005dad7557eSWanpeng Li * @out_nid: ptr to int for nid of the range, can be %NULL 100635fd0808STejun Heo * 1007f1af9d3aSPhilipp Hachtmann * Find the first area from *@idx which matches @nid, fill the out 100835fd0808STejun Heo * parameters, and update *@idx for the next iteration. The lower 32bit of 1009f1af9d3aSPhilipp Hachtmann * *@idx contains index into type_a and the upper 32bit indexes the 1010f1af9d3aSPhilipp Hachtmann * areas before each region in type_b. For example, if type_b regions 101135fd0808STejun Heo * look like the following, 101235fd0808STejun Heo * 101335fd0808STejun Heo * 0:[0-16), 1:[32-48), 2:[128-130) 101435fd0808STejun Heo * 101535fd0808STejun Heo * The upper 32bit indexes the following regions. 101635fd0808STejun Heo * 101735fd0808STejun Heo * 0:[0-0), 1:[16-32), 2:[48-128), 3:[130-MAX) 101835fd0808STejun Heo * 101935fd0808STejun Heo * As both region arrays are sorted, the function advances the two indices 102035fd0808STejun Heo * in lockstep and returns each intersection. 102135fd0808STejun Heo */ 1022e1720feeSMike Rapoport void __init_memblock __next_mem_range(u64 *idx, int nid, 1023e1720feeSMike Rapoport enum memblock_flags flags, 1024f1af9d3aSPhilipp Hachtmann struct memblock_type *type_a, 1025f1af9d3aSPhilipp Hachtmann struct memblock_type *type_b, 102635fd0808STejun Heo phys_addr_t *out_start, 102735fd0808STejun Heo phys_addr_t *out_end, int *out_nid) 102835fd0808STejun Heo { 1029f1af9d3aSPhilipp Hachtmann int idx_a = *idx & 0xffffffff; 1030f1af9d3aSPhilipp Hachtmann int idx_b = *idx >> 32; 1031b1154233SGrygorii Strashko 1032f1af9d3aSPhilipp Hachtmann if (WARN_ONCE(nid == MAX_NUMNODES, 1033f1af9d3aSPhilipp Hachtmann "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) 1034560dca27SGrygorii Strashko nid = NUMA_NO_NODE; 103535fd0808STejun Heo 1036f1af9d3aSPhilipp Hachtmann for (; idx_a < type_a->cnt; idx_a++) { 1037f1af9d3aSPhilipp Hachtmann struct memblock_region *m = &type_a->regions[idx_a]; 1038f1af9d3aSPhilipp Hachtmann 103935fd0808STejun Heo phys_addr_t m_start = m->base; 104035fd0808STejun Heo phys_addr_t m_end = m->base + m->size; 1041f1af9d3aSPhilipp Hachtmann int m_nid = memblock_get_region_node(m); 104235fd0808STejun Heo 1043c9a688a3SMike Rapoport if (should_skip_region(m, nid, flags)) 1044bf3d3cc5SArd Biesheuvel continue; 1045bf3d3cc5SArd Biesheuvel 1046f1af9d3aSPhilipp Hachtmann if (!type_b) { 1047f1af9d3aSPhilipp Hachtmann if (out_start) 1048f1af9d3aSPhilipp Hachtmann *out_start = m_start; 1049f1af9d3aSPhilipp Hachtmann if (out_end) 1050f1af9d3aSPhilipp Hachtmann *out_end = m_end; 1051f1af9d3aSPhilipp Hachtmann if (out_nid) 1052f1af9d3aSPhilipp Hachtmann *out_nid = m_nid; 1053f1af9d3aSPhilipp Hachtmann idx_a++; 1054f1af9d3aSPhilipp Hachtmann *idx = (u32)idx_a | (u64)idx_b << 32; 1055f1af9d3aSPhilipp Hachtmann return; 1056f1af9d3aSPhilipp Hachtmann } 105735fd0808STejun Heo 1058f1af9d3aSPhilipp Hachtmann /* scan areas before each reservation */ 1059f1af9d3aSPhilipp Hachtmann for (; idx_b < type_b->cnt + 1; idx_b++) { 1060f1af9d3aSPhilipp Hachtmann struct memblock_region *r; 1061f1af9d3aSPhilipp Hachtmann phys_addr_t r_start; 1062f1af9d3aSPhilipp Hachtmann phys_addr_t r_end; 1063f1af9d3aSPhilipp Hachtmann 1064f1af9d3aSPhilipp Hachtmann r = &type_b->regions[idx_b]; 1065f1af9d3aSPhilipp Hachtmann r_start = idx_b ? r[-1].base + r[-1].size : 0; 1066f1af9d3aSPhilipp Hachtmann r_end = idx_b < type_b->cnt ? 10671c4bc43dSStefan Agner r->base : PHYS_ADDR_MAX; 1068f1af9d3aSPhilipp Hachtmann 1069f1af9d3aSPhilipp Hachtmann /* 1070f1af9d3aSPhilipp Hachtmann * if idx_b advanced past idx_a, 1071f1af9d3aSPhilipp Hachtmann * break out to advance idx_a 1072f1af9d3aSPhilipp Hachtmann */ 107335fd0808STejun Heo if (r_start >= m_end) 107435fd0808STejun Heo break; 107535fd0808STejun Heo /* if the two regions intersect, we're done */ 107635fd0808STejun Heo if (m_start < r_end) { 107735fd0808STejun Heo if (out_start) 1078f1af9d3aSPhilipp Hachtmann *out_start = 1079f1af9d3aSPhilipp Hachtmann max(m_start, r_start); 108035fd0808STejun Heo if (out_end) 108135fd0808STejun Heo *out_end = min(m_end, r_end); 108235fd0808STejun Heo if (out_nid) 1083f1af9d3aSPhilipp Hachtmann *out_nid = m_nid; 108435fd0808STejun Heo /* 1085f1af9d3aSPhilipp Hachtmann * The region which ends first is 1086f1af9d3aSPhilipp Hachtmann * advanced for the next iteration. 108735fd0808STejun Heo */ 108835fd0808STejun Heo if (m_end <= r_end) 1089f1af9d3aSPhilipp Hachtmann idx_a++; 109035fd0808STejun Heo else 1091f1af9d3aSPhilipp Hachtmann idx_b++; 1092f1af9d3aSPhilipp Hachtmann *idx = (u32)idx_a | (u64)idx_b << 32; 109335fd0808STejun Heo return; 109435fd0808STejun Heo } 109535fd0808STejun Heo } 109635fd0808STejun Heo } 109735fd0808STejun Heo 109835fd0808STejun Heo /* signal end of iteration */ 109935fd0808STejun Heo *idx = ULLONG_MAX; 110035fd0808STejun Heo } 110135fd0808STejun Heo 11027bd0b0f0STejun Heo /** 1103f1af9d3aSPhilipp Hachtmann * __next_mem_range_rev - generic next function for for_each_*_range_rev() 1104f1af9d3aSPhilipp Hachtmann * 11057bd0b0f0STejun Heo * @idx: pointer to u64 loop variable 1106ad5ea8cdSAlexander Kuleshov * @nid: node selector, %NUMA_NO_NODE for all nodes 1107fc6daaf9STony Luck * @flags: pick from blocks based on memory attributes 1108f1af9d3aSPhilipp Hachtmann * @type_a: pointer to memblock_type from where the range is taken 1109f1af9d3aSPhilipp Hachtmann * @type_b: pointer to memblock_type which excludes memory from being taken 1110dad7557eSWanpeng Li * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL 1111dad7557eSWanpeng Li * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL 1112dad7557eSWanpeng Li * @out_nid: ptr to int for nid of the range, can be %NULL 11137bd0b0f0STejun Heo * 111447cec443SMike Rapoport * Finds the next range from type_a which is not marked as unsuitable 111547cec443SMike Rapoport * in type_b. 111647cec443SMike Rapoport * 1117f1af9d3aSPhilipp Hachtmann * Reverse of __next_mem_range(). 11187bd0b0f0STejun Heo */ 1119e1720feeSMike Rapoport void __init_memblock __next_mem_range_rev(u64 *idx, int nid, 1120e1720feeSMike Rapoport enum memblock_flags flags, 1121f1af9d3aSPhilipp Hachtmann struct memblock_type *type_a, 1122f1af9d3aSPhilipp Hachtmann struct memblock_type *type_b, 11237bd0b0f0STejun Heo phys_addr_t *out_start, 11247bd0b0f0STejun Heo phys_addr_t *out_end, int *out_nid) 11257bd0b0f0STejun Heo { 1126f1af9d3aSPhilipp Hachtmann int idx_a = *idx & 0xffffffff; 1127f1af9d3aSPhilipp Hachtmann int idx_b = *idx >> 32; 1128b1154233SGrygorii Strashko 1129560dca27SGrygorii Strashko if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) 1130560dca27SGrygorii Strashko nid = NUMA_NO_NODE; 11317bd0b0f0STejun Heo 11327bd0b0f0STejun Heo if (*idx == (u64)ULLONG_MAX) { 1133f1af9d3aSPhilipp Hachtmann idx_a = type_a->cnt - 1; 1134e47608abSzijun_hu if (type_b != NULL) 1135f1af9d3aSPhilipp Hachtmann idx_b = type_b->cnt; 1136e47608abSzijun_hu else 1137e47608abSzijun_hu idx_b = 0; 11387bd0b0f0STejun Heo } 11397bd0b0f0STejun Heo 1140f1af9d3aSPhilipp Hachtmann for (; idx_a >= 0; idx_a--) { 1141f1af9d3aSPhilipp Hachtmann struct memblock_region *m = &type_a->regions[idx_a]; 1142f1af9d3aSPhilipp Hachtmann 11437bd0b0f0STejun Heo phys_addr_t m_start = m->base; 11447bd0b0f0STejun Heo phys_addr_t m_end = m->base + m->size; 1145f1af9d3aSPhilipp Hachtmann int m_nid = memblock_get_region_node(m); 11467bd0b0f0STejun Heo 1147c9a688a3SMike Rapoport if (should_skip_region(m, nid, flags)) 1148bf3d3cc5SArd Biesheuvel continue; 1149bf3d3cc5SArd Biesheuvel 1150f1af9d3aSPhilipp Hachtmann if (!type_b) { 1151f1af9d3aSPhilipp Hachtmann if (out_start) 1152f1af9d3aSPhilipp Hachtmann *out_start = m_start; 1153f1af9d3aSPhilipp Hachtmann if (out_end) 1154f1af9d3aSPhilipp Hachtmann *out_end = m_end; 1155f1af9d3aSPhilipp Hachtmann if (out_nid) 1156f1af9d3aSPhilipp Hachtmann *out_nid = m_nid; 1157fb399b48Szijun_hu idx_a--; 1158f1af9d3aSPhilipp Hachtmann *idx = (u32)idx_a | (u64)idx_b << 32; 1159f1af9d3aSPhilipp Hachtmann return; 1160f1af9d3aSPhilipp Hachtmann } 11617bd0b0f0STejun Heo 1162f1af9d3aSPhilipp Hachtmann /* scan areas before each reservation */ 1163f1af9d3aSPhilipp Hachtmann for (; idx_b >= 0; idx_b--) { 1164f1af9d3aSPhilipp Hachtmann struct memblock_region *r; 1165f1af9d3aSPhilipp Hachtmann phys_addr_t r_start; 1166f1af9d3aSPhilipp Hachtmann phys_addr_t r_end; 1167f1af9d3aSPhilipp Hachtmann 1168f1af9d3aSPhilipp Hachtmann r = &type_b->regions[idx_b]; 1169f1af9d3aSPhilipp Hachtmann r_start = idx_b ? r[-1].base + r[-1].size : 0; 1170f1af9d3aSPhilipp Hachtmann r_end = idx_b < type_b->cnt ? 11711c4bc43dSStefan Agner r->base : PHYS_ADDR_MAX; 1172f1af9d3aSPhilipp Hachtmann /* 1173f1af9d3aSPhilipp Hachtmann * if idx_b advanced past idx_a, 1174f1af9d3aSPhilipp Hachtmann * break out to advance idx_a 1175f1af9d3aSPhilipp Hachtmann */ 1176f1af9d3aSPhilipp Hachtmann 11777bd0b0f0STejun Heo if (r_end <= m_start) 11787bd0b0f0STejun Heo break; 11797bd0b0f0STejun Heo /* if the two regions intersect, we're done */ 11807bd0b0f0STejun Heo if (m_end > r_start) { 11817bd0b0f0STejun Heo if (out_start) 11827bd0b0f0STejun Heo *out_start = max(m_start, r_start); 11837bd0b0f0STejun Heo if (out_end) 11847bd0b0f0STejun Heo *out_end = min(m_end, r_end); 11857bd0b0f0STejun Heo if (out_nid) 1186f1af9d3aSPhilipp Hachtmann *out_nid = m_nid; 11877bd0b0f0STejun Heo if (m_start >= r_start) 1188f1af9d3aSPhilipp Hachtmann idx_a--; 11897bd0b0f0STejun Heo else 1190f1af9d3aSPhilipp Hachtmann idx_b--; 1191f1af9d3aSPhilipp Hachtmann *idx = (u32)idx_a | (u64)idx_b << 32; 11927bd0b0f0STejun Heo return; 11937bd0b0f0STejun Heo } 11947bd0b0f0STejun Heo } 11957bd0b0f0STejun Heo } 1196f1af9d3aSPhilipp Hachtmann /* signal end of iteration */ 11977bd0b0f0STejun Heo *idx = ULLONG_MAX; 11987bd0b0f0STejun Heo } 11997bd0b0f0STejun Heo 12007c0caeb8STejun Heo /* 120145e79815SChen Chang * Common iterator interface used to define for_each_mem_pfn_range(). 12027c0caeb8STejun Heo */ 12037c0caeb8STejun Heo void __init_memblock __next_mem_pfn_range(int *idx, int nid, 12047c0caeb8STejun Heo unsigned long *out_start_pfn, 12057c0caeb8STejun Heo unsigned long *out_end_pfn, int *out_nid) 12067c0caeb8STejun Heo { 12077c0caeb8STejun Heo struct memblock_type *type = &memblock.memory; 12087c0caeb8STejun Heo struct memblock_region *r; 1209d622abf7SMike Rapoport int r_nid; 12107c0caeb8STejun Heo 12117c0caeb8STejun Heo while (++*idx < type->cnt) { 12127c0caeb8STejun Heo r = &type->regions[*idx]; 1213d622abf7SMike Rapoport r_nid = memblock_get_region_node(r); 12147c0caeb8STejun Heo 12157c0caeb8STejun Heo if (PFN_UP(r->base) >= PFN_DOWN(r->base + r->size)) 12167c0caeb8STejun Heo continue; 1217d622abf7SMike Rapoport if (nid == MAX_NUMNODES || nid == r_nid) 12187c0caeb8STejun Heo break; 12197c0caeb8STejun Heo } 12207c0caeb8STejun Heo if (*idx >= type->cnt) { 12217c0caeb8STejun Heo *idx = -1; 12227c0caeb8STejun Heo return; 12237c0caeb8STejun Heo } 12247c0caeb8STejun Heo 12257c0caeb8STejun Heo if (out_start_pfn) 12267c0caeb8STejun Heo *out_start_pfn = PFN_UP(r->base); 12277c0caeb8STejun Heo if (out_end_pfn) 12287c0caeb8STejun Heo *out_end_pfn = PFN_DOWN(r->base + r->size); 12297c0caeb8STejun Heo if (out_nid) 1230d622abf7SMike Rapoport *out_nid = r_nid; 12317c0caeb8STejun Heo } 12327c0caeb8STejun Heo 12337c0caeb8STejun Heo /** 12347c0caeb8STejun Heo * memblock_set_node - set node ID on memblock regions 12357c0caeb8STejun Heo * @base: base of area to set node ID for 12367c0caeb8STejun Heo * @size: size of area to set node ID for 1237e7e8de59STang Chen * @type: memblock type to set node ID for 12387c0caeb8STejun Heo * @nid: node ID to set 12397c0caeb8STejun Heo * 1240e7e8de59STang Chen * Set the nid of memblock @type regions in [@base, @base + @size) to @nid. 12417c0caeb8STejun Heo * Regions which cross the area boundaries are split as necessary. 12427c0caeb8STejun Heo * 124347cec443SMike Rapoport * Return: 12447c0caeb8STejun Heo * 0 on success, -errno on failure. 12457c0caeb8STejun Heo */ 12467c0caeb8STejun Heo int __init_memblock memblock_set_node(phys_addr_t base, phys_addr_t size, 1247e7e8de59STang Chen struct memblock_type *type, int nid) 12487c0caeb8STejun Heo { 1249*3f08a302SMike Rapoport #ifdef CONFIG_NEED_MULTIPLE_NODES 12506a9ceb31STejun Heo int start_rgn, end_rgn; 12516a9ceb31STejun Heo int i, ret; 12527c0caeb8STejun Heo 12536a9ceb31STejun Heo ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); 12546a9ceb31STejun Heo if (ret) 12556a9ceb31STejun Heo return ret; 12567c0caeb8STejun Heo 12576a9ceb31STejun Heo for (i = start_rgn; i < end_rgn; i++) 1258e9d24ad3SWanpeng Li memblock_set_region_node(&type->regions[i], nid); 12597c0caeb8STejun Heo 12607c0caeb8STejun Heo memblock_merge_regions(type); 1261*3f08a302SMike Rapoport #endif 12627c0caeb8STejun Heo return 0; 12637c0caeb8STejun Heo } 1264*3f08a302SMike Rapoport 1265837566e7SAlexander Duyck #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT 1266837566e7SAlexander Duyck /** 1267837566e7SAlexander Duyck * __next_mem_pfn_range_in_zone - iterator for for_each_*_range_in_zone() 1268837566e7SAlexander Duyck * 1269837566e7SAlexander Duyck * @idx: pointer to u64 loop variable 1270837566e7SAlexander Duyck * @zone: zone in which all of the memory blocks reside 1271837566e7SAlexander Duyck * @out_spfn: ptr to ulong for start pfn of the range, can be %NULL 1272837566e7SAlexander Duyck * @out_epfn: ptr to ulong for end pfn of the range, can be %NULL 1273837566e7SAlexander Duyck * 1274837566e7SAlexander Duyck * This function is meant to be a zone/pfn specific wrapper for the 1275837566e7SAlexander Duyck * for_each_mem_range type iterators. Specifically they are used in the 1276837566e7SAlexander Duyck * deferred memory init routines and as such we were duplicating much of 1277837566e7SAlexander Duyck * this logic throughout the code. So instead of having it in multiple 1278837566e7SAlexander Duyck * locations it seemed like it would make more sense to centralize this to 1279837566e7SAlexander Duyck * one new iterator that does everything they need. 1280837566e7SAlexander Duyck */ 1281837566e7SAlexander Duyck void __init_memblock 1282837566e7SAlexander Duyck __next_mem_pfn_range_in_zone(u64 *idx, struct zone *zone, 1283837566e7SAlexander Duyck unsigned long *out_spfn, unsigned long *out_epfn) 1284837566e7SAlexander Duyck { 1285837566e7SAlexander Duyck int zone_nid = zone_to_nid(zone); 1286837566e7SAlexander Duyck phys_addr_t spa, epa; 1287837566e7SAlexander Duyck int nid; 1288837566e7SAlexander Duyck 1289837566e7SAlexander Duyck __next_mem_range(idx, zone_nid, MEMBLOCK_NONE, 1290837566e7SAlexander Duyck &memblock.memory, &memblock.reserved, 1291837566e7SAlexander Duyck &spa, &epa, &nid); 1292837566e7SAlexander Duyck 1293837566e7SAlexander Duyck while (*idx != U64_MAX) { 1294837566e7SAlexander Duyck unsigned long epfn = PFN_DOWN(epa); 1295837566e7SAlexander Duyck unsigned long spfn = PFN_UP(spa); 1296837566e7SAlexander Duyck 1297837566e7SAlexander Duyck /* 1298837566e7SAlexander Duyck * Verify the end is at least past the start of the zone and 1299837566e7SAlexander Duyck * that we have at least one PFN to initialize. 1300837566e7SAlexander Duyck */ 1301837566e7SAlexander Duyck if (zone->zone_start_pfn < epfn && spfn < epfn) { 1302837566e7SAlexander Duyck /* if we went too far just stop searching */ 1303837566e7SAlexander Duyck if (zone_end_pfn(zone) <= spfn) { 1304837566e7SAlexander Duyck *idx = U64_MAX; 1305837566e7SAlexander Duyck break; 1306837566e7SAlexander Duyck } 1307837566e7SAlexander Duyck 1308837566e7SAlexander Duyck if (out_spfn) 1309837566e7SAlexander Duyck *out_spfn = max(zone->zone_start_pfn, spfn); 1310837566e7SAlexander Duyck if (out_epfn) 1311837566e7SAlexander Duyck *out_epfn = min(zone_end_pfn(zone), epfn); 1312837566e7SAlexander Duyck 1313837566e7SAlexander Duyck return; 1314837566e7SAlexander Duyck } 1315837566e7SAlexander Duyck 1316837566e7SAlexander Duyck __next_mem_range(idx, zone_nid, MEMBLOCK_NONE, 1317837566e7SAlexander Duyck &memblock.memory, &memblock.reserved, 1318837566e7SAlexander Duyck &spa, &epa, &nid); 1319837566e7SAlexander Duyck } 1320837566e7SAlexander Duyck 1321837566e7SAlexander Duyck /* signal end of iteration */ 1322837566e7SAlexander Duyck if (out_spfn) 1323837566e7SAlexander Duyck *out_spfn = ULONG_MAX; 1324837566e7SAlexander Duyck if (out_epfn) 1325837566e7SAlexander Duyck *out_epfn = 0; 1326837566e7SAlexander Duyck } 1327837566e7SAlexander Duyck 1328837566e7SAlexander Duyck #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */ 13297c0caeb8STejun Heo 133092d12f95SMike Rapoport /** 133192d12f95SMike Rapoport * memblock_alloc_range_nid - allocate boot memory block 133292d12f95SMike Rapoport * @size: size of memory block to be allocated in bytes 133392d12f95SMike Rapoport * @align: alignment of the region and block's size 133492d12f95SMike Rapoport * @start: the lower bound of the memory region to allocate (phys address) 133592d12f95SMike Rapoport * @end: the upper bound of the memory region to allocate (phys address) 133692d12f95SMike Rapoport * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 13370ac398b1SYunfeng Ye * @exact_nid: control the allocation fall back to other nodes 133892d12f95SMike Rapoport * 133992d12f95SMike Rapoport * The allocation is performed from memory region limited by 134095830666SCao jin * memblock.current_limit if @end == %MEMBLOCK_ALLOC_ACCESSIBLE. 134192d12f95SMike Rapoport * 13420ac398b1SYunfeng Ye * If the specified node can not hold the requested memory and @exact_nid 13430ac398b1SYunfeng Ye * is false, the allocation falls back to any node in the system. 134492d12f95SMike Rapoport * 134592d12f95SMike Rapoport * For systems with memory mirroring, the allocation is attempted first 134692d12f95SMike Rapoport * from the regions with mirroring enabled and then retried from any 134792d12f95SMike Rapoport * memory region. 134892d12f95SMike Rapoport * 134992d12f95SMike Rapoport * In addition, function sets the min_count to 0 using kmemleak_alloc_phys for 135092d12f95SMike Rapoport * allocated boot memory block, so that it is never reported as leaks. 135192d12f95SMike Rapoport * 135292d12f95SMike Rapoport * Return: 135392d12f95SMike Rapoport * Physical address of allocated memory block on success, %0 on failure. 135492d12f95SMike Rapoport */ 13558676af1fSAslan Bakirov phys_addr_t __init memblock_alloc_range_nid(phys_addr_t size, 13562bfc2862SAkinobu Mita phys_addr_t align, phys_addr_t start, 13570ac398b1SYunfeng Ye phys_addr_t end, int nid, 13580ac398b1SYunfeng Ye bool exact_nid) 135995f72d1eSYinghai Lu { 136092d12f95SMike Rapoport enum memblock_flags flags = choose_memblock_flags(); 13616ed311b2SBenjamin Herrenschmidt phys_addr_t found; 136295f72d1eSYinghai Lu 136392d12f95SMike Rapoport if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) 136492d12f95SMike Rapoport nid = NUMA_NO_NODE; 136592d12f95SMike Rapoport 13662f770806SMike Rapoport if (!align) { 13672f770806SMike Rapoport /* Can't use WARNs this early in boot on powerpc */ 13682f770806SMike Rapoport dump_stack(); 13692f770806SMike Rapoport align = SMP_CACHE_BYTES; 13702f770806SMike Rapoport } 13712f770806SMike Rapoport 137292d12f95SMike Rapoport again: 1373fc6daaf9STony Luck found = memblock_find_in_range_node(size, align, start, end, nid, 1374fc6daaf9STony Luck flags); 137592d12f95SMike Rapoport if (found && !memblock_reserve(found, size)) 137692d12f95SMike Rapoport goto done; 137792d12f95SMike Rapoport 13780ac398b1SYunfeng Ye if (nid != NUMA_NO_NODE && !exact_nid) { 137992d12f95SMike Rapoport found = memblock_find_in_range_node(size, align, start, 138092d12f95SMike Rapoport end, NUMA_NO_NODE, 138192d12f95SMike Rapoport flags); 138292d12f95SMike Rapoport if (found && !memblock_reserve(found, size)) 138392d12f95SMike Rapoport goto done; 138492d12f95SMike Rapoport } 138592d12f95SMike Rapoport 138692d12f95SMike Rapoport if (flags & MEMBLOCK_MIRROR) { 138792d12f95SMike Rapoport flags &= ~MEMBLOCK_MIRROR; 138892d12f95SMike Rapoport pr_warn("Could not allocate %pap bytes of mirrored memory\n", 138992d12f95SMike Rapoport &size); 139092d12f95SMike Rapoport goto again; 139192d12f95SMike Rapoport } 139292d12f95SMike Rapoport 139392d12f95SMike Rapoport return 0; 139492d12f95SMike Rapoport 139592d12f95SMike Rapoport done: 139692d12f95SMike Rapoport /* Skip kmemleak for kasan_init() due to high volume. */ 139792d12f95SMike Rapoport if (end != MEMBLOCK_ALLOC_KASAN) 1398aedf95eaSCatalin Marinas /* 139992d12f95SMike Rapoport * The min_count is set to 0 so that memblock allocated 140092d12f95SMike Rapoport * blocks are never reported as leaks. This is because many 140192d12f95SMike Rapoport * of these blocks are only referred via the physical 140292d12f95SMike Rapoport * address which is not looked up by kmemleak. 1403aedf95eaSCatalin Marinas */ 14049099daedSCatalin Marinas kmemleak_alloc_phys(found, size, 0, 0); 140592d12f95SMike Rapoport 14066ed311b2SBenjamin Herrenschmidt return found; 1407aedf95eaSCatalin Marinas } 140895f72d1eSYinghai Lu 1409a2974133SMike Rapoport /** 1410a2974133SMike Rapoport * memblock_phys_alloc_range - allocate a memory block inside specified range 1411a2974133SMike Rapoport * @size: size of memory block to be allocated in bytes 1412a2974133SMike Rapoport * @align: alignment of the region and block's size 1413a2974133SMike Rapoport * @start: the lower bound of the memory region to allocate (physical address) 1414a2974133SMike Rapoport * @end: the upper bound of the memory region to allocate (physical address) 1415a2974133SMike Rapoport * 1416a2974133SMike Rapoport * Allocate @size bytes in the between @start and @end. 1417a2974133SMike Rapoport * 1418a2974133SMike Rapoport * Return: physical address of the allocated memory block on success, 1419a2974133SMike Rapoport * %0 on failure. 1420a2974133SMike Rapoport */ 14218a770c2aSMike Rapoport phys_addr_t __init memblock_phys_alloc_range(phys_addr_t size, 14228a770c2aSMike Rapoport phys_addr_t align, 14238a770c2aSMike Rapoport phys_addr_t start, 14248a770c2aSMike Rapoport phys_addr_t end) 14252bfc2862SAkinobu Mita { 14260ac398b1SYunfeng Ye return memblock_alloc_range_nid(size, align, start, end, NUMA_NO_NODE, 14270ac398b1SYunfeng Ye false); 14287bd0b0f0STejun Heo } 14297bd0b0f0STejun Heo 1430a2974133SMike Rapoport /** 1431a2974133SMike Rapoport * memblock_phys_alloc_try_nid - allocate a memory block from specified MUMA node 1432a2974133SMike Rapoport * @size: size of memory block to be allocated in bytes 1433a2974133SMike Rapoport * @align: alignment of the region and block's size 1434a2974133SMike Rapoport * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 1435a2974133SMike Rapoport * 1436a2974133SMike Rapoport * Allocates memory block from the specified NUMA node. If the node 1437a2974133SMike Rapoport * has no available memory, attempts to allocated from any node in the 1438a2974133SMike Rapoport * system. 1439a2974133SMike Rapoport * 1440a2974133SMike Rapoport * Return: physical address of the allocated memory block on success, 1441a2974133SMike Rapoport * %0 on failure. 1442a2974133SMike Rapoport */ 14439a8dd708SMike Rapoport phys_addr_t __init memblock_phys_alloc_try_nid(phys_addr_t size, phys_addr_t align, int nid) 14449d1e2492SBenjamin Herrenschmidt { 144533755574SMike Rapoport return memblock_alloc_range_nid(size, align, 0, 14460ac398b1SYunfeng Ye MEMBLOCK_ALLOC_ACCESSIBLE, nid, false); 144795f72d1eSYinghai Lu } 144895f72d1eSYinghai Lu 144926f09e9bSSantosh Shilimkar /** 1450eb31d559SMike Rapoport * memblock_alloc_internal - allocate boot memory block 145126f09e9bSSantosh Shilimkar * @size: size of memory block to be allocated in bytes 145226f09e9bSSantosh Shilimkar * @align: alignment of the region and block's size 145326f09e9bSSantosh Shilimkar * @min_addr: the lower bound of the memory region to allocate (phys address) 145426f09e9bSSantosh Shilimkar * @max_addr: the upper bound of the memory region to allocate (phys address) 145526f09e9bSSantosh Shilimkar * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 14560ac398b1SYunfeng Ye * @exact_nid: control the allocation fall back to other nodes 145726f09e9bSSantosh Shilimkar * 145892d12f95SMike Rapoport * Allocates memory block using memblock_alloc_range_nid() and 145992d12f95SMike Rapoport * converts the returned physical address to virtual. 146092d12f95SMike Rapoport * 146126f09e9bSSantosh Shilimkar * The @min_addr limit is dropped if it can not be satisfied and the allocation 146292d12f95SMike Rapoport * will fall back to memory below @min_addr. Other constraints, such 146392d12f95SMike Rapoport * as node and mirrored memory will be handled again in 146492d12f95SMike Rapoport * memblock_alloc_range_nid(). 146526f09e9bSSantosh Shilimkar * 146647cec443SMike Rapoport * Return: 146726f09e9bSSantosh Shilimkar * Virtual address of allocated memory block on success, NULL on failure. 146826f09e9bSSantosh Shilimkar */ 1469eb31d559SMike Rapoport static void * __init memblock_alloc_internal( 147026f09e9bSSantosh Shilimkar phys_addr_t size, phys_addr_t align, 147126f09e9bSSantosh Shilimkar phys_addr_t min_addr, phys_addr_t max_addr, 14720ac398b1SYunfeng Ye int nid, bool exact_nid) 147326f09e9bSSantosh Shilimkar { 147426f09e9bSSantosh Shilimkar phys_addr_t alloc; 147526f09e9bSSantosh Shilimkar 147626f09e9bSSantosh Shilimkar /* 147726f09e9bSSantosh Shilimkar * Detect any accidental use of these APIs after slab is ready, as at 147826f09e9bSSantosh Shilimkar * this moment memblock may be deinitialized already and its 1479c6ffc5caSMike Rapoport * internal data may be destroyed (after execution of memblock_free_all) 148026f09e9bSSantosh Shilimkar */ 148126f09e9bSSantosh Shilimkar if (WARN_ON_ONCE(slab_is_available())) 148226f09e9bSSantosh Shilimkar return kzalloc_node(size, GFP_NOWAIT, nid); 148326f09e9bSSantosh Shilimkar 1484f3057ad7SMike Rapoport if (max_addr > memblock.current_limit) 1485f3057ad7SMike Rapoport max_addr = memblock.current_limit; 1486f3057ad7SMike Rapoport 14870ac398b1SYunfeng Ye alloc = memblock_alloc_range_nid(size, align, min_addr, max_addr, nid, 14880ac398b1SYunfeng Ye exact_nid); 14892f770806SMike Rapoport 149092d12f95SMike Rapoport /* retry allocation without lower limit */ 149192d12f95SMike Rapoport if (!alloc && min_addr) 14920ac398b1SYunfeng Ye alloc = memblock_alloc_range_nid(size, align, 0, max_addr, nid, 14930ac398b1SYunfeng Ye exact_nid); 149426f09e9bSSantosh Shilimkar 149592d12f95SMike Rapoport if (!alloc) 1496a3f5bafcSTony Luck return NULL; 149726f09e9bSSantosh Shilimkar 149892d12f95SMike Rapoport return phys_to_virt(alloc); 149926f09e9bSSantosh Shilimkar } 150026f09e9bSSantosh Shilimkar 150126f09e9bSSantosh Shilimkar /** 15020ac398b1SYunfeng Ye * memblock_alloc_exact_nid_raw - allocate boot memory block on the exact node 15030ac398b1SYunfeng Ye * without zeroing memory 15040ac398b1SYunfeng Ye * @size: size of memory block to be allocated in bytes 15050ac398b1SYunfeng Ye * @align: alignment of the region and block's size 15060ac398b1SYunfeng Ye * @min_addr: the lower bound of the memory region from where the allocation 15070ac398b1SYunfeng Ye * is preferred (phys address) 15080ac398b1SYunfeng Ye * @max_addr: the upper bound of the memory region from where the allocation 15090ac398b1SYunfeng Ye * is preferred (phys address), or %MEMBLOCK_ALLOC_ACCESSIBLE to 15100ac398b1SYunfeng Ye * allocate only from memory limited by memblock.current_limit value 15110ac398b1SYunfeng Ye * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 15120ac398b1SYunfeng Ye * 15130ac398b1SYunfeng Ye * Public function, provides additional debug information (including caller 15140ac398b1SYunfeng Ye * info), if enabled. Does not zero allocated memory. 15150ac398b1SYunfeng Ye * 15160ac398b1SYunfeng Ye * Return: 15170ac398b1SYunfeng Ye * Virtual address of allocated memory block on success, NULL on failure. 15180ac398b1SYunfeng Ye */ 15190ac398b1SYunfeng Ye void * __init memblock_alloc_exact_nid_raw( 15200ac398b1SYunfeng Ye phys_addr_t size, phys_addr_t align, 15210ac398b1SYunfeng Ye phys_addr_t min_addr, phys_addr_t max_addr, 15220ac398b1SYunfeng Ye int nid) 15230ac398b1SYunfeng Ye { 15240ac398b1SYunfeng Ye void *ptr; 15250ac398b1SYunfeng Ye 15260ac398b1SYunfeng Ye memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=%pa max_addr=%pa %pS\n", 15270ac398b1SYunfeng Ye __func__, (u64)size, (u64)align, nid, &min_addr, 15280ac398b1SYunfeng Ye &max_addr, (void *)_RET_IP_); 15290ac398b1SYunfeng Ye 15300ac398b1SYunfeng Ye ptr = memblock_alloc_internal(size, align, 15310ac398b1SYunfeng Ye min_addr, max_addr, nid, true); 15320ac398b1SYunfeng Ye if (ptr && size > 0) 15330ac398b1SYunfeng Ye page_init_poison(ptr, size); 15340ac398b1SYunfeng Ye 15350ac398b1SYunfeng Ye return ptr; 15360ac398b1SYunfeng Ye } 15370ac398b1SYunfeng Ye 15380ac398b1SYunfeng Ye /** 1539eb31d559SMike Rapoport * memblock_alloc_try_nid_raw - allocate boot memory block without zeroing 1540ea1f5f37SPavel Tatashin * memory and without panicking 1541ea1f5f37SPavel Tatashin * @size: size of memory block to be allocated in bytes 1542ea1f5f37SPavel Tatashin * @align: alignment of the region and block's size 1543ea1f5f37SPavel Tatashin * @min_addr: the lower bound of the memory region from where the allocation 1544ea1f5f37SPavel Tatashin * is preferred (phys address) 1545ea1f5f37SPavel Tatashin * @max_addr: the upper bound of the memory region from where the allocation 154697ad1087SMike Rapoport * is preferred (phys address), or %MEMBLOCK_ALLOC_ACCESSIBLE to 1547ea1f5f37SPavel Tatashin * allocate only from memory limited by memblock.current_limit value 1548ea1f5f37SPavel Tatashin * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 1549ea1f5f37SPavel Tatashin * 1550ea1f5f37SPavel Tatashin * Public function, provides additional debug information (including caller 1551ea1f5f37SPavel Tatashin * info), if enabled. Does not zero allocated memory, does not panic if request 1552ea1f5f37SPavel Tatashin * cannot be satisfied. 1553ea1f5f37SPavel Tatashin * 155447cec443SMike Rapoport * Return: 1555ea1f5f37SPavel Tatashin * Virtual address of allocated memory block on success, NULL on failure. 1556ea1f5f37SPavel Tatashin */ 1557eb31d559SMike Rapoport void * __init memblock_alloc_try_nid_raw( 1558ea1f5f37SPavel Tatashin phys_addr_t size, phys_addr_t align, 1559ea1f5f37SPavel Tatashin phys_addr_t min_addr, phys_addr_t max_addr, 1560ea1f5f37SPavel Tatashin int nid) 1561ea1f5f37SPavel Tatashin { 1562ea1f5f37SPavel Tatashin void *ptr; 1563ea1f5f37SPavel Tatashin 1564d75f773cSSakari Ailus memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=%pa max_addr=%pa %pS\n", 1565a36aab89SMike Rapoport __func__, (u64)size, (u64)align, nid, &min_addr, 1566a36aab89SMike Rapoport &max_addr, (void *)_RET_IP_); 1567ea1f5f37SPavel Tatashin 1568eb31d559SMike Rapoport ptr = memblock_alloc_internal(size, align, 15690ac398b1SYunfeng Ye min_addr, max_addr, nid, false); 1570ea1f5f37SPavel Tatashin if (ptr && size > 0) 1571f682a97aSAlexander Duyck page_init_poison(ptr, size); 1572f682a97aSAlexander Duyck 1573ea1f5f37SPavel Tatashin return ptr; 1574ea1f5f37SPavel Tatashin } 1575ea1f5f37SPavel Tatashin 1576ea1f5f37SPavel Tatashin /** 1577c0dbe825SMike Rapoport * memblock_alloc_try_nid - allocate boot memory block 157826f09e9bSSantosh Shilimkar * @size: size of memory block to be allocated in bytes 157926f09e9bSSantosh Shilimkar * @align: alignment of the region and block's size 158026f09e9bSSantosh Shilimkar * @min_addr: the lower bound of the memory region from where the allocation 158126f09e9bSSantosh Shilimkar * is preferred (phys address) 158226f09e9bSSantosh Shilimkar * @max_addr: the upper bound of the memory region from where the allocation 158397ad1087SMike Rapoport * is preferred (phys address), or %MEMBLOCK_ALLOC_ACCESSIBLE to 158426f09e9bSSantosh Shilimkar * allocate only from memory limited by memblock.current_limit value 158526f09e9bSSantosh Shilimkar * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 158626f09e9bSSantosh Shilimkar * 1587c0dbe825SMike Rapoport * Public function, provides additional debug information (including caller 1588c0dbe825SMike Rapoport * info), if enabled. This function zeroes the allocated memory. 158926f09e9bSSantosh Shilimkar * 159047cec443SMike Rapoport * Return: 159126f09e9bSSantosh Shilimkar * Virtual address of allocated memory block on success, NULL on failure. 159226f09e9bSSantosh Shilimkar */ 1593eb31d559SMike Rapoport void * __init memblock_alloc_try_nid( 159426f09e9bSSantosh Shilimkar phys_addr_t size, phys_addr_t align, 159526f09e9bSSantosh Shilimkar phys_addr_t min_addr, phys_addr_t max_addr, 159626f09e9bSSantosh Shilimkar int nid) 159726f09e9bSSantosh Shilimkar { 159826f09e9bSSantosh Shilimkar void *ptr; 159926f09e9bSSantosh Shilimkar 1600d75f773cSSakari Ailus memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=%pa max_addr=%pa %pS\n", 1601a36aab89SMike Rapoport __func__, (u64)size, (u64)align, nid, &min_addr, 1602a36aab89SMike Rapoport &max_addr, (void *)_RET_IP_); 1603eb31d559SMike Rapoport ptr = memblock_alloc_internal(size, align, 16040ac398b1SYunfeng Ye min_addr, max_addr, nid, false); 1605c0dbe825SMike Rapoport if (ptr) 1606ea1f5f37SPavel Tatashin memset(ptr, 0, size); 160726f09e9bSSantosh Shilimkar 1608c0dbe825SMike Rapoport return ptr; 160926f09e9bSSantosh Shilimkar } 161026f09e9bSSantosh Shilimkar 161126f09e9bSSantosh Shilimkar /** 1612a2974133SMike Rapoport * __memblock_free_late - free pages directly to buddy allocator 161348a833ccSMike Rapoport * @base: phys starting address of the boot memory block 161426f09e9bSSantosh Shilimkar * @size: size of the boot memory block in bytes 161526f09e9bSSantosh Shilimkar * 1616a2974133SMike Rapoport * This is only useful when the memblock allocator has already been torn 161726f09e9bSSantosh Shilimkar * down, but we are still initializing the system. Pages are released directly 1618a2974133SMike Rapoport * to the buddy allocator. 161926f09e9bSSantosh Shilimkar */ 162026f09e9bSSantosh Shilimkar void __init __memblock_free_late(phys_addr_t base, phys_addr_t size) 162126f09e9bSSantosh Shilimkar { 1622a36aab89SMike Rapoport phys_addr_t cursor, end; 162326f09e9bSSantosh Shilimkar 1624a36aab89SMike Rapoport end = base + size - 1; 1625d75f773cSSakari Ailus memblock_dbg("%s: [%pa-%pa] %pS\n", 1626a36aab89SMike Rapoport __func__, &base, &end, (void *)_RET_IP_); 16279099daedSCatalin Marinas kmemleak_free_part_phys(base, size); 162826f09e9bSSantosh Shilimkar cursor = PFN_UP(base); 162926f09e9bSSantosh Shilimkar end = PFN_DOWN(base + size); 163026f09e9bSSantosh Shilimkar 163126f09e9bSSantosh Shilimkar for (; cursor < end; cursor++) { 16327c2ee349SMike Rapoport memblock_free_pages(pfn_to_page(cursor), cursor, 0); 1633ca79b0c2SArun KS totalram_pages_inc(); 163426f09e9bSSantosh Shilimkar } 163526f09e9bSSantosh Shilimkar } 16369d1e2492SBenjamin Herrenschmidt 16379d1e2492SBenjamin Herrenschmidt /* 16389d1e2492SBenjamin Herrenschmidt * Remaining API functions 16399d1e2492SBenjamin Herrenschmidt */ 16409d1e2492SBenjamin Herrenschmidt 16411f1ffb8aSDavid Gibson phys_addr_t __init_memblock memblock_phys_mem_size(void) 164295f72d1eSYinghai Lu { 16431440c4e2STejun Heo return memblock.memory.total_size; 164495f72d1eSYinghai Lu } 164595f72d1eSYinghai Lu 16468907de5dSSrikar Dronamraju phys_addr_t __init_memblock memblock_reserved_size(void) 16478907de5dSSrikar Dronamraju { 16488907de5dSSrikar Dronamraju return memblock.reserved.total_size; 16498907de5dSSrikar Dronamraju } 16508907de5dSSrikar Dronamraju 1651595ad9afSYinghai Lu phys_addr_t __init memblock_mem_size(unsigned long limit_pfn) 1652595ad9afSYinghai Lu { 1653595ad9afSYinghai Lu unsigned long pages = 0; 1654595ad9afSYinghai Lu struct memblock_region *r; 1655595ad9afSYinghai Lu unsigned long start_pfn, end_pfn; 1656595ad9afSYinghai Lu 1657595ad9afSYinghai Lu for_each_memblock(memory, r) { 1658595ad9afSYinghai Lu start_pfn = memblock_region_memory_base_pfn(r); 1659595ad9afSYinghai Lu end_pfn = memblock_region_memory_end_pfn(r); 1660595ad9afSYinghai Lu start_pfn = min_t(unsigned long, start_pfn, limit_pfn); 1661595ad9afSYinghai Lu end_pfn = min_t(unsigned long, end_pfn, limit_pfn); 1662595ad9afSYinghai Lu pages += end_pfn - start_pfn; 1663595ad9afSYinghai Lu } 1664595ad9afSYinghai Lu 166516763230SFabian Frederick return PFN_PHYS(pages); 1666595ad9afSYinghai Lu } 1667595ad9afSYinghai Lu 16680a93ebefSSam Ravnborg /* lowest address */ 16690a93ebefSSam Ravnborg phys_addr_t __init_memblock memblock_start_of_DRAM(void) 16700a93ebefSSam Ravnborg { 16710a93ebefSSam Ravnborg return memblock.memory.regions[0].base; 16720a93ebefSSam Ravnborg } 16730a93ebefSSam Ravnborg 167410d06439SYinghai Lu phys_addr_t __init_memblock memblock_end_of_DRAM(void) 167595f72d1eSYinghai Lu { 167695f72d1eSYinghai Lu int idx = memblock.memory.cnt - 1; 167795f72d1eSYinghai Lu 1678e3239ff9SBenjamin Herrenschmidt return (memblock.memory.regions[idx].base + memblock.memory.regions[idx].size); 167995f72d1eSYinghai Lu } 168095f72d1eSYinghai Lu 1681a571d4ebSDennis Chen static phys_addr_t __init_memblock __find_max_addr(phys_addr_t limit) 168295f72d1eSYinghai Lu { 16831c4bc43dSStefan Agner phys_addr_t max_addr = PHYS_ADDR_MAX; 1684136199f0SEmil Medve struct memblock_region *r; 168595f72d1eSYinghai Lu 1686a571d4ebSDennis Chen /* 1687a571d4ebSDennis Chen * translate the memory @limit size into the max address within one of 1688a571d4ebSDennis Chen * the memory memblock regions, if the @limit exceeds the total size 16891c4bc43dSStefan Agner * of those regions, max_addr will keep original value PHYS_ADDR_MAX 1690a571d4ebSDennis Chen */ 1691136199f0SEmil Medve for_each_memblock(memory, r) { 1692c0ce8fefSTejun Heo if (limit <= r->size) { 1693c0ce8fefSTejun Heo max_addr = r->base + limit; 169495f72d1eSYinghai Lu break; 169595f72d1eSYinghai Lu } 1696c0ce8fefSTejun Heo limit -= r->size; 169795f72d1eSYinghai Lu } 1698c0ce8fefSTejun Heo 1699a571d4ebSDennis Chen return max_addr; 1700a571d4ebSDennis Chen } 1701a571d4ebSDennis Chen 1702a571d4ebSDennis Chen void __init memblock_enforce_memory_limit(phys_addr_t limit) 1703a571d4ebSDennis Chen { 170449aef717SColin Ian King phys_addr_t max_addr; 1705a571d4ebSDennis Chen 1706a571d4ebSDennis Chen if (!limit) 1707a571d4ebSDennis Chen return; 1708a571d4ebSDennis Chen 1709a571d4ebSDennis Chen max_addr = __find_max_addr(limit); 1710a571d4ebSDennis Chen 1711a571d4ebSDennis Chen /* @limit exceeds the total size of the memory, do nothing */ 17121c4bc43dSStefan Agner if (max_addr == PHYS_ADDR_MAX) 1713a571d4ebSDennis Chen return; 1714a571d4ebSDennis Chen 1715c0ce8fefSTejun Heo /* truncate both memory and reserved regions */ 1716f1af9d3aSPhilipp Hachtmann memblock_remove_range(&memblock.memory, max_addr, 17171c4bc43dSStefan Agner PHYS_ADDR_MAX); 1718f1af9d3aSPhilipp Hachtmann memblock_remove_range(&memblock.reserved, max_addr, 17191c4bc43dSStefan Agner PHYS_ADDR_MAX); 172095f72d1eSYinghai Lu } 172195f72d1eSYinghai Lu 1722c9ca9b4eSAKASHI Takahiro void __init memblock_cap_memory_range(phys_addr_t base, phys_addr_t size) 1723c9ca9b4eSAKASHI Takahiro { 1724c9ca9b4eSAKASHI Takahiro int start_rgn, end_rgn; 1725c9ca9b4eSAKASHI Takahiro int i, ret; 1726c9ca9b4eSAKASHI Takahiro 1727c9ca9b4eSAKASHI Takahiro if (!size) 1728c9ca9b4eSAKASHI Takahiro return; 1729c9ca9b4eSAKASHI Takahiro 1730c9ca9b4eSAKASHI Takahiro ret = memblock_isolate_range(&memblock.memory, base, size, 1731c9ca9b4eSAKASHI Takahiro &start_rgn, &end_rgn); 1732c9ca9b4eSAKASHI Takahiro if (ret) 1733c9ca9b4eSAKASHI Takahiro return; 1734c9ca9b4eSAKASHI Takahiro 1735c9ca9b4eSAKASHI Takahiro /* remove all the MAP regions */ 1736c9ca9b4eSAKASHI Takahiro for (i = memblock.memory.cnt - 1; i >= end_rgn; i--) 1737c9ca9b4eSAKASHI Takahiro if (!memblock_is_nomap(&memblock.memory.regions[i])) 1738c9ca9b4eSAKASHI Takahiro memblock_remove_region(&memblock.memory, i); 1739c9ca9b4eSAKASHI Takahiro 1740c9ca9b4eSAKASHI Takahiro for (i = start_rgn - 1; i >= 0; i--) 1741c9ca9b4eSAKASHI Takahiro if (!memblock_is_nomap(&memblock.memory.regions[i])) 1742c9ca9b4eSAKASHI Takahiro memblock_remove_region(&memblock.memory, i); 1743c9ca9b4eSAKASHI Takahiro 1744c9ca9b4eSAKASHI Takahiro /* truncate the reserved regions */ 1745c9ca9b4eSAKASHI Takahiro memblock_remove_range(&memblock.reserved, 0, base); 1746c9ca9b4eSAKASHI Takahiro memblock_remove_range(&memblock.reserved, 17471c4bc43dSStefan Agner base + size, PHYS_ADDR_MAX); 1748c9ca9b4eSAKASHI Takahiro } 1749c9ca9b4eSAKASHI Takahiro 1750a571d4ebSDennis Chen void __init memblock_mem_limit_remove_map(phys_addr_t limit) 1751a571d4ebSDennis Chen { 1752a571d4ebSDennis Chen phys_addr_t max_addr; 1753a571d4ebSDennis Chen 1754a571d4ebSDennis Chen if (!limit) 1755a571d4ebSDennis Chen return; 1756a571d4ebSDennis Chen 1757a571d4ebSDennis Chen max_addr = __find_max_addr(limit); 1758a571d4ebSDennis Chen 1759a571d4ebSDennis Chen /* @limit exceeds the total size of the memory, do nothing */ 17601c4bc43dSStefan Agner if (max_addr == PHYS_ADDR_MAX) 1761a571d4ebSDennis Chen return; 1762a571d4ebSDennis Chen 1763c9ca9b4eSAKASHI Takahiro memblock_cap_memory_range(0, max_addr); 1764a571d4ebSDennis Chen } 1765a571d4ebSDennis Chen 1766cd79481dSYinghai Lu static int __init_memblock memblock_search(struct memblock_type *type, phys_addr_t addr) 176772d4b0b4SBenjamin Herrenschmidt { 176872d4b0b4SBenjamin Herrenschmidt unsigned int left = 0, right = type->cnt; 176972d4b0b4SBenjamin Herrenschmidt 177072d4b0b4SBenjamin Herrenschmidt do { 177172d4b0b4SBenjamin Herrenschmidt unsigned int mid = (right + left) / 2; 177272d4b0b4SBenjamin Herrenschmidt 177372d4b0b4SBenjamin Herrenschmidt if (addr < type->regions[mid].base) 177472d4b0b4SBenjamin Herrenschmidt right = mid; 177572d4b0b4SBenjamin Herrenschmidt else if (addr >= (type->regions[mid].base + 177672d4b0b4SBenjamin Herrenschmidt type->regions[mid].size)) 177772d4b0b4SBenjamin Herrenschmidt left = mid + 1; 177872d4b0b4SBenjamin Herrenschmidt else 177972d4b0b4SBenjamin Herrenschmidt return mid; 178072d4b0b4SBenjamin Herrenschmidt } while (left < right); 178172d4b0b4SBenjamin Herrenschmidt return -1; 178272d4b0b4SBenjamin Herrenschmidt } 178372d4b0b4SBenjamin Herrenschmidt 1784f5a222dcSYueyi Li bool __init_memblock memblock_is_reserved(phys_addr_t addr) 178595f72d1eSYinghai Lu { 178672d4b0b4SBenjamin Herrenschmidt return memblock_search(&memblock.reserved, addr) != -1; 178795f72d1eSYinghai Lu } 178872d4b0b4SBenjamin Herrenschmidt 1789b4ad0c7eSYaowei Bai bool __init_memblock memblock_is_memory(phys_addr_t addr) 179072d4b0b4SBenjamin Herrenschmidt { 179172d4b0b4SBenjamin Herrenschmidt return memblock_search(&memblock.memory, addr) != -1; 179272d4b0b4SBenjamin Herrenschmidt } 179372d4b0b4SBenjamin Herrenschmidt 1794937f0c26SYaowei Bai bool __init_memblock memblock_is_map_memory(phys_addr_t addr) 1795bf3d3cc5SArd Biesheuvel { 1796bf3d3cc5SArd Biesheuvel int i = memblock_search(&memblock.memory, addr); 1797bf3d3cc5SArd Biesheuvel 1798bf3d3cc5SArd Biesheuvel if (i == -1) 1799bf3d3cc5SArd Biesheuvel return false; 1800bf3d3cc5SArd Biesheuvel return !memblock_is_nomap(&memblock.memory.regions[i]); 1801bf3d3cc5SArd Biesheuvel } 1802bf3d3cc5SArd Biesheuvel 1803e76b63f8SYinghai Lu int __init_memblock memblock_search_pfn_nid(unsigned long pfn, 1804e76b63f8SYinghai Lu unsigned long *start_pfn, unsigned long *end_pfn) 1805e76b63f8SYinghai Lu { 1806e76b63f8SYinghai Lu struct memblock_type *type = &memblock.memory; 180716763230SFabian Frederick int mid = memblock_search(type, PFN_PHYS(pfn)); 1808e76b63f8SYinghai Lu 1809e76b63f8SYinghai Lu if (mid == -1) 1810e76b63f8SYinghai Lu return -1; 1811e76b63f8SYinghai Lu 1812f7e2f7e8SFabian Frederick *start_pfn = PFN_DOWN(type->regions[mid].base); 1813f7e2f7e8SFabian Frederick *end_pfn = PFN_DOWN(type->regions[mid].base + type->regions[mid].size); 1814e76b63f8SYinghai Lu 1815d622abf7SMike Rapoport return memblock_get_region_node(&type->regions[mid]); 1816e76b63f8SYinghai Lu } 1817e76b63f8SYinghai Lu 1818eab30949SStephen Boyd /** 1819eab30949SStephen Boyd * memblock_is_region_memory - check if a region is a subset of memory 1820eab30949SStephen Boyd * @base: base of region to check 1821eab30949SStephen Boyd * @size: size of region to check 1822eab30949SStephen Boyd * 1823eab30949SStephen Boyd * Check if the region [@base, @base + @size) is a subset of a memory block. 1824eab30949SStephen Boyd * 182547cec443SMike Rapoport * Return: 1826eab30949SStephen Boyd * 0 if false, non-zero if true 1827eab30949SStephen Boyd */ 1828937f0c26SYaowei Bai bool __init_memblock memblock_is_region_memory(phys_addr_t base, phys_addr_t size) 182972d4b0b4SBenjamin Herrenschmidt { 1830abb65272STomi Valkeinen int idx = memblock_search(&memblock.memory, base); 1831eb18f1b5STejun Heo phys_addr_t end = base + memblock_cap_size(base, &size); 183272d4b0b4SBenjamin Herrenschmidt 183372d4b0b4SBenjamin Herrenschmidt if (idx == -1) 1834937f0c26SYaowei Bai return false; 1835ef415ef4SWei Yang return (memblock.memory.regions[idx].base + 1836eb18f1b5STejun Heo memblock.memory.regions[idx].size) >= end; 183795f72d1eSYinghai Lu } 183895f72d1eSYinghai Lu 1839eab30949SStephen Boyd /** 1840eab30949SStephen Boyd * memblock_is_region_reserved - check if a region intersects reserved memory 1841eab30949SStephen Boyd * @base: base of region to check 1842eab30949SStephen Boyd * @size: size of region to check 1843eab30949SStephen Boyd * 184447cec443SMike Rapoport * Check if the region [@base, @base + @size) intersects a reserved 184547cec443SMike Rapoport * memory block. 1846eab30949SStephen Boyd * 184747cec443SMike Rapoport * Return: 1848c5c5c9d1STang Chen * True if they intersect, false if not. 1849eab30949SStephen Boyd */ 1850c5c5c9d1STang Chen bool __init_memblock memblock_is_region_reserved(phys_addr_t base, phys_addr_t size) 185195f72d1eSYinghai Lu { 1852eb18f1b5STejun Heo memblock_cap_size(base, &size); 1853c5c5c9d1STang Chen return memblock_overlaps_region(&memblock.reserved, base, size); 185495f72d1eSYinghai Lu } 185595f72d1eSYinghai Lu 18566ede1fd3SYinghai Lu void __init_memblock memblock_trim_memory(phys_addr_t align) 18576ede1fd3SYinghai Lu { 18586ede1fd3SYinghai Lu phys_addr_t start, end, orig_start, orig_end; 1859136199f0SEmil Medve struct memblock_region *r; 18606ede1fd3SYinghai Lu 1861136199f0SEmil Medve for_each_memblock(memory, r) { 1862136199f0SEmil Medve orig_start = r->base; 1863136199f0SEmil Medve orig_end = r->base + r->size; 18646ede1fd3SYinghai Lu start = round_up(orig_start, align); 18656ede1fd3SYinghai Lu end = round_down(orig_end, align); 18666ede1fd3SYinghai Lu 18676ede1fd3SYinghai Lu if (start == orig_start && end == orig_end) 18686ede1fd3SYinghai Lu continue; 18696ede1fd3SYinghai Lu 18706ede1fd3SYinghai Lu if (start < end) { 1871136199f0SEmil Medve r->base = start; 1872136199f0SEmil Medve r->size = end - start; 18736ede1fd3SYinghai Lu } else { 1874136199f0SEmil Medve memblock_remove_region(&memblock.memory, 1875136199f0SEmil Medve r - memblock.memory.regions); 1876136199f0SEmil Medve r--; 18776ede1fd3SYinghai Lu } 18786ede1fd3SYinghai Lu } 18796ede1fd3SYinghai Lu } 1880e63075a3SBenjamin Herrenschmidt 18813661ca66SYinghai Lu void __init_memblock memblock_set_current_limit(phys_addr_t limit) 1882e63075a3SBenjamin Herrenschmidt { 1883e63075a3SBenjamin Herrenschmidt memblock.current_limit = limit; 1884e63075a3SBenjamin Herrenschmidt } 1885e63075a3SBenjamin Herrenschmidt 1886fec51014SLaura Abbott phys_addr_t __init_memblock memblock_get_current_limit(void) 1887fec51014SLaura Abbott { 1888fec51014SLaura Abbott return memblock.current_limit; 1889fec51014SLaura Abbott } 1890fec51014SLaura Abbott 18910262d9c8SHeiko Carstens static void __init_memblock memblock_dump(struct memblock_type *type) 18926ed311b2SBenjamin Herrenschmidt { 18935d63f81cSMiles Chen phys_addr_t base, end, size; 1894e1720feeSMike Rapoport enum memblock_flags flags; 18958c9c1701SAlexander Kuleshov int idx; 18968c9c1701SAlexander Kuleshov struct memblock_region *rgn; 18976ed311b2SBenjamin Herrenschmidt 18980262d9c8SHeiko Carstens pr_info(" %s.cnt = 0x%lx\n", type->name, type->cnt); 18996ed311b2SBenjamin Herrenschmidt 190066e8b438SGioh Kim for_each_memblock_type(idx, type, rgn) { 19017c0caeb8STejun Heo char nid_buf[32] = ""; 19026ed311b2SBenjamin Herrenschmidt 19037c0caeb8STejun Heo base = rgn->base; 19047c0caeb8STejun Heo size = rgn->size; 19055d63f81cSMiles Chen end = base + size - 1; 190666a20757STang Chen flags = rgn->flags; 1907*3f08a302SMike Rapoport #ifdef CONFIG_NEED_MULTIPLE_NODES 19087c0caeb8STejun Heo if (memblock_get_region_node(rgn) != MAX_NUMNODES) 19097c0caeb8STejun Heo snprintf(nid_buf, sizeof(nid_buf), " on node %d", 19107c0caeb8STejun Heo memblock_get_region_node(rgn)); 19117c0caeb8STejun Heo #endif 1912e1720feeSMike Rapoport pr_info(" %s[%#x]\t[%pa-%pa], %pa bytes%s flags: %#x\n", 19130262d9c8SHeiko Carstens type->name, idx, &base, &end, &size, nid_buf, flags); 19146ed311b2SBenjamin Herrenschmidt } 19156ed311b2SBenjamin Herrenschmidt } 19166ed311b2SBenjamin Herrenschmidt 19174ff7b82fSTejun Heo void __init_memblock __memblock_dump_all(void) 19186ed311b2SBenjamin Herrenschmidt { 19196ed311b2SBenjamin Herrenschmidt pr_info("MEMBLOCK configuration:\n"); 19205d63f81cSMiles Chen pr_info(" memory size = %pa reserved size = %pa\n", 19215d63f81cSMiles Chen &memblock.memory.total_size, 19225d63f81cSMiles Chen &memblock.reserved.total_size); 19236ed311b2SBenjamin Herrenschmidt 19240262d9c8SHeiko Carstens memblock_dump(&memblock.memory); 19250262d9c8SHeiko Carstens memblock_dump(&memblock.reserved); 1926409efd4cSHeiko Carstens #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP 19270262d9c8SHeiko Carstens memblock_dump(&memblock.physmem); 1928409efd4cSHeiko Carstens #endif 19296ed311b2SBenjamin Herrenschmidt } 19306ed311b2SBenjamin Herrenschmidt 19311aadc056STejun Heo void __init memblock_allow_resize(void) 19326ed311b2SBenjamin Herrenschmidt { 1933142b45a7SBenjamin Herrenschmidt memblock_can_resize = 1; 19346ed311b2SBenjamin Herrenschmidt } 19356ed311b2SBenjamin Herrenschmidt 19366ed311b2SBenjamin Herrenschmidt static int __init early_memblock(char *p) 19376ed311b2SBenjamin Herrenschmidt { 19386ed311b2SBenjamin Herrenschmidt if (p && strstr(p, "debug")) 19396ed311b2SBenjamin Herrenschmidt memblock_debug = 1; 19406ed311b2SBenjamin Herrenschmidt return 0; 19416ed311b2SBenjamin Herrenschmidt } 19426ed311b2SBenjamin Herrenschmidt early_param("memblock", early_memblock); 19436ed311b2SBenjamin Herrenschmidt 1944bda49a81SMike Rapoport static void __init __free_pages_memory(unsigned long start, unsigned long end) 1945bda49a81SMike Rapoport { 1946bda49a81SMike Rapoport int order; 1947bda49a81SMike Rapoport 1948bda49a81SMike Rapoport while (start < end) { 1949bda49a81SMike Rapoport order = min(MAX_ORDER - 1UL, __ffs(start)); 1950bda49a81SMike Rapoport 1951bda49a81SMike Rapoport while (start + (1UL << order) > end) 1952bda49a81SMike Rapoport order--; 1953bda49a81SMike Rapoport 1954bda49a81SMike Rapoport memblock_free_pages(pfn_to_page(start), start, order); 1955bda49a81SMike Rapoport 1956bda49a81SMike Rapoport start += (1UL << order); 1957bda49a81SMike Rapoport } 1958bda49a81SMike Rapoport } 1959bda49a81SMike Rapoport 1960bda49a81SMike Rapoport static unsigned long __init __free_memory_core(phys_addr_t start, 1961bda49a81SMike Rapoport phys_addr_t end) 1962bda49a81SMike Rapoport { 1963bda49a81SMike Rapoport unsigned long start_pfn = PFN_UP(start); 1964bda49a81SMike Rapoport unsigned long end_pfn = min_t(unsigned long, 1965bda49a81SMike Rapoport PFN_DOWN(end), max_low_pfn); 1966bda49a81SMike Rapoport 1967bda49a81SMike Rapoport if (start_pfn >= end_pfn) 1968bda49a81SMike Rapoport return 0; 1969bda49a81SMike Rapoport 1970bda49a81SMike Rapoport __free_pages_memory(start_pfn, end_pfn); 1971bda49a81SMike Rapoport 1972bda49a81SMike Rapoport return end_pfn - start_pfn; 1973bda49a81SMike Rapoport } 1974bda49a81SMike Rapoport 1975bda49a81SMike Rapoport static unsigned long __init free_low_memory_core_early(void) 1976bda49a81SMike Rapoport { 1977bda49a81SMike Rapoport unsigned long count = 0; 1978bda49a81SMike Rapoport phys_addr_t start, end; 1979bda49a81SMike Rapoport u64 i; 1980bda49a81SMike Rapoport 1981bda49a81SMike Rapoport memblock_clear_hotplug(0, -1); 1982bda49a81SMike Rapoport 1983bda49a81SMike Rapoport for_each_reserved_mem_region(i, &start, &end) 1984bda49a81SMike Rapoport reserve_bootmem_region(start, end); 1985bda49a81SMike Rapoport 1986bda49a81SMike Rapoport /* 1987bda49a81SMike Rapoport * We need to use NUMA_NO_NODE instead of NODE_DATA(0)->node_id 1988bda49a81SMike Rapoport * because in some case like Node0 doesn't have RAM installed 1989bda49a81SMike Rapoport * low ram will be on Node1 1990bda49a81SMike Rapoport */ 1991bda49a81SMike Rapoport for_each_free_mem_range(i, NUMA_NO_NODE, MEMBLOCK_NONE, &start, &end, 1992bda49a81SMike Rapoport NULL) 1993bda49a81SMike Rapoport count += __free_memory_core(start, end); 1994bda49a81SMike Rapoport 1995bda49a81SMike Rapoport return count; 1996bda49a81SMike Rapoport } 1997bda49a81SMike Rapoport 1998bda49a81SMike Rapoport static int reset_managed_pages_done __initdata; 1999bda49a81SMike Rapoport 2000bda49a81SMike Rapoport void reset_node_managed_pages(pg_data_t *pgdat) 2001bda49a81SMike Rapoport { 2002bda49a81SMike Rapoport struct zone *z; 2003bda49a81SMike Rapoport 2004bda49a81SMike Rapoport for (z = pgdat->node_zones; z < pgdat->node_zones + MAX_NR_ZONES; z++) 20059705bea5SArun KS atomic_long_set(&z->managed_pages, 0); 2006bda49a81SMike Rapoport } 2007bda49a81SMike Rapoport 2008bda49a81SMike Rapoport void __init reset_all_zones_managed_pages(void) 2009bda49a81SMike Rapoport { 2010bda49a81SMike Rapoport struct pglist_data *pgdat; 2011bda49a81SMike Rapoport 2012bda49a81SMike Rapoport if (reset_managed_pages_done) 2013bda49a81SMike Rapoport return; 2014bda49a81SMike Rapoport 2015bda49a81SMike Rapoport for_each_online_pgdat(pgdat) 2016bda49a81SMike Rapoport reset_node_managed_pages(pgdat); 2017bda49a81SMike Rapoport 2018bda49a81SMike Rapoport reset_managed_pages_done = 1; 2019bda49a81SMike Rapoport } 2020bda49a81SMike Rapoport 2021bda49a81SMike Rapoport /** 2022bda49a81SMike Rapoport * memblock_free_all - release free pages to the buddy allocator 2023bda49a81SMike Rapoport * 2024bda49a81SMike Rapoport * Return: the number of pages actually released. 2025bda49a81SMike Rapoport */ 2026bda49a81SMike Rapoport unsigned long __init memblock_free_all(void) 2027bda49a81SMike Rapoport { 2028bda49a81SMike Rapoport unsigned long pages; 2029bda49a81SMike Rapoport 2030bda49a81SMike Rapoport reset_all_zones_managed_pages(); 2031bda49a81SMike Rapoport 2032bda49a81SMike Rapoport pages = free_low_memory_core_early(); 2033ca79b0c2SArun KS totalram_pages_add(pages); 2034bda49a81SMike Rapoport 2035bda49a81SMike Rapoport return pages; 2036bda49a81SMike Rapoport } 2037bda49a81SMike Rapoport 2038350e88baSMike Rapoport #if defined(CONFIG_DEBUG_FS) && defined(CONFIG_ARCH_KEEP_MEMBLOCK) 20396d03b885SBenjamin Herrenschmidt 20406d03b885SBenjamin Herrenschmidt static int memblock_debug_show(struct seq_file *m, void *private) 20416d03b885SBenjamin Herrenschmidt { 20426d03b885SBenjamin Herrenschmidt struct memblock_type *type = m->private; 20436d03b885SBenjamin Herrenschmidt struct memblock_region *reg; 20446d03b885SBenjamin Herrenschmidt int i; 20455d63f81cSMiles Chen phys_addr_t end; 20466d03b885SBenjamin Herrenschmidt 20476d03b885SBenjamin Herrenschmidt for (i = 0; i < type->cnt; i++) { 20486d03b885SBenjamin Herrenschmidt reg = &type->regions[i]; 20495d63f81cSMiles Chen end = reg->base + reg->size - 1; 20506d03b885SBenjamin Herrenschmidt 20515d63f81cSMiles Chen seq_printf(m, "%4d: ", i); 20525d63f81cSMiles Chen seq_printf(m, "%pa..%pa\n", ®->base, &end); 20536d03b885SBenjamin Herrenschmidt } 20546d03b885SBenjamin Herrenschmidt return 0; 20556d03b885SBenjamin Herrenschmidt } 20565ad35093SAndy Shevchenko DEFINE_SHOW_ATTRIBUTE(memblock_debug); 20576d03b885SBenjamin Herrenschmidt 20586d03b885SBenjamin Herrenschmidt static int __init memblock_init_debugfs(void) 20596d03b885SBenjamin Herrenschmidt { 20606d03b885SBenjamin Herrenschmidt struct dentry *root = debugfs_create_dir("memblock", NULL); 2061d9f7979cSGreg Kroah-Hartman 20620825a6f9SJoe Perches debugfs_create_file("memory", 0444, root, 20630825a6f9SJoe Perches &memblock.memory, &memblock_debug_fops); 20640825a6f9SJoe Perches debugfs_create_file("reserved", 0444, root, 20650825a6f9SJoe Perches &memblock.reserved, &memblock_debug_fops); 206670210ed9SPhilipp Hachtmann #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP 20670825a6f9SJoe Perches debugfs_create_file("physmem", 0444, root, 20680825a6f9SJoe Perches &memblock.physmem, &memblock_debug_fops); 206970210ed9SPhilipp Hachtmann #endif 20706d03b885SBenjamin Herrenschmidt 20716d03b885SBenjamin Herrenschmidt return 0; 20726d03b885SBenjamin Herrenschmidt } 20736d03b885SBenjamin Herrenschmidt __initcall(memblock_init_debugfs); 20746d03b885SBenjamin Herrenschmidt 20756d03b885SBenjamin Herrenschmidt #endif /* CONFIG_DEBUG_FS */ 2076