195f72d1eSYinghai Lu /* 295f72d1eSYinghai Lu * Procedures for maintaining information about logical memory blocks. 395f72d1eSYinghai Lu * 495f72d1eSYinghai Lu * Peter Bergner, IBM Corp. June 2001. 595f72d1eSYinghai Lu * Copyright (C) 2001 Peter Bergner. 695f72d1eSYinghai Lu * 795f72d1eSYinghai Lu * This program is free software; you can redistribute it and/or 895f72d1eSYinghai Lu * modify it under the terms of the GNU General Public License 995f72d1eSYinghai Lu * as published by the Free Software Foundation; either version 1095f72d1eSYinghai Lu * 2 of the License, or (at your option) any later version. 1195f72d1eSYinghai Lu */ 1295f72d1eSYinghai Lu 1395f72d1eSYinghai Lu #include <linux/kernel.h> 14142b45a7SBenjamin Herrenschmidt #include <linux/slab.h> 1595f72d1eSYinghai Lu #include <linux/init.h> 1695f72d1eSYinghai Lu #include <linux/bitops.h> 17449e8df3SBenjamin Herrenschmidt #include <linux/poison.h> 18c196f76fSBenjamin Herrenschmidt #include <linux/pfn.h> 196d03b885SBenjamin Herrenschmidt #include <linux/debugfs.h> 20514c6032SRandy Dunlap #include <linux/kmemleak.h> 216d03b885SBenjamin Herrenschmidt #include <linux/seq_file.h> 2295f72d1eSYinghai Lu #include <linux/memblock.h> 2395f72d1eSYinghai Lu 24c4c5ad6bSChristoph Hellwig #include <asm/sections.h> 2526f09e9bSSantosh Shilimkar #include <linux/io.h> 2626f09e9bSSantosh Shilimkar 2726f09e9bSSantosh Shilimkar #include "internal.h" 2879442ed1STang Chen 298a5b403dSArd Biesheuvel #define INIT_MEMBLOCK_REGIONS 128 308a5b403dSArd Biesheuvel #define INIT_PHYSMEM_REGIONS 4 318a5b403dSArd Biesheuvel 328a5b403dSArd Biesheuvel #ifndef INIT_MEMBLOCK_RESERVED_REGIONS 338a5b403dSArd Biesheuvel # define INIT_MEMBLOCK_RESERVED_REGIONS INIT_MEMBLOCK_REGIONS 348a5b403dSArd Biesheuvel #endif 358a5b403dSArd Biesheuvel 363e039c5cSMike Rapoport /** 373e039c5cSMike Rapoport * DOC: memblock overview 383e039c5cSMike Rapoport * 393e039c5cSMike Rapoport * Memblock is a method of managing memory regions during the early 403e039c5cSMike Rapoport * boot period when the usual kernel memory allocators are not up and 413e039c5cSMike Rapoport * running. 423e039c5cSMike Rapoport * 433e039c5cSMike Rapoport * Memblock views the system memory as collections of contiguous 443e039c5cSMike Rapoport * regions. There are several types of these collections: 453e039c5cSMike Rapoport * 463e039c5cSMike Rapoport * * ``memory`` - describes the physical memory available to the 473e039c5cSMike Rapoport * kernel; this may differ from the actual physical memory installed 483e039c5cSMike Rapoport * in the system, for instance when the memory is restricted with 493e039c5cSMike Rapoport * ``mem=`` command line parameter 503e039c5cSMike Rapoport * * ``reserved`` - describes the regions that were allocated 513e039c5cSMike Rapoport * * ``physmap`` - describes the actual physical memory regardless of 523e039c5cSMike Rapoport * the possible restrictions; the ``physmap`` type is only available 533e039c5cSMike Rapoport * on some architectures. 543e039c5cSMike Rapoport * 553e039c5cSMike Rapoport * Each region is represented by :c:type:`struct memblock_region` that 563e039c5cSMike Rapoport * defines the region extents, its attributes and NUMA node id on NUMA 573e039c5cSMike Rapoport * systems. Every memory type is described by the :c:type:`struct 583e039c5cSMike Rapoport * memblock_type` which contains an array of memory regions along with 593e039c5cSMike Rapoport * the allocator metadata. The memory types are nicely wrapped with 603e039c5cSMike Rapoport * :c:type:`struct memblock`. This structure is statically initialzed 613e039c5cSMike Rapoport * at build time. The region arrays for the "memory" and "reserved" 623e039c5cSMike Rapoport * types are initially sized to %INIT_MEMBLOCK_REGIONS and for the 633e039c5cSMike Rapoport * "physmap" type to %INIT_PHYSMEM_REGIONS. 643e039c5cSMike Rapoport * The :c:func:`memblock_allow_resize` enables automatic resizing of 653e039c5cSMike Rapoport * the region arrays during addition of new regions. This feature 663e039c5cSMike Rapoport * should be used with care so that memory allocated for the region 673e039c5cSMike Rapoport * array will not overlap with areas that should be reserved, for 683e039c5cSMike Rapoport * example initrd. 693e039c5cSMike Rapoport * 703e039c5cSMike Rapoport * The early architecture setup should tell memblock what the physical 713e039c5cSMike Rapoport * memory layout is by using :c:func:`memblock_add` or 723e039c5cSMike Rapoport * :c:func:`memblock_add_node` functions. The first function does not 733e039c5cSMike Rapoport * assign the region to a NUMA node and it is appropriate for UMA 743e039c5cSMike Rapoport * systems. Yet, it is possible to use it on NUMA systems as well and 753e039c5cSMike Rapoport * assign the region to a NUMA node later in the setup process using 763e039c5cSMike Rapoport * :c:func:`memblock_set_node`. The :c:func:`memblock_add_node` 773e039c5cSMike Rapoport * performs such an assignment directly. 783e039c5cSMike Rapoport * 793e039c5cSMike Rapoport * Once memblock is setup the memory can be allocated using either 803e039c5cSMike Rapoport * memblock or bootmem APIs. 813e039c5cSMike Rapoport * 823e039c5cSMike Rapoport * As the system boot progresses, the architecture specific 833e039c5cSMike Rapoport * :c:func:`mem_init` function frees all the memory to the buddy page 843e039c5cSMike Rapoport * allocator. 853e039c5cSMike Rapoport * 863e039c5cSMike Rapoport * If an architecure enables %CONFIG_ARCH_DISCARD_MEMBLOCK, the 873e039c5cSMike Rapoport * memblock data structures will be discarded after the system 883e039c5cSMike Rapoport * initialization compltes. 893e039c5cSMike Rapoport */ 903e039c5cSMike Rapoport 91bda49a81SMike Rapoport #ifndef CONFIG_NEED_MULTIPLE_NODES 92bda49a81SMike Rapoport struct pglist_data __refdata contig_page_data; 93bda49a81SMike Rapoport EXPORT_SYMBOL(contig_page_data); 94bda49a81SMike Rapoport #endif 95bda49a81SMike Rapoport 96bda49a81SMike Rapoport unsigned long max_low_pfn; 97bda49a81SMike Rapoport unsigned long min_low_pfn; 98bda49a81SMike Rapoport unsigned long max_pfn; 99bda49a81SMike Rapoport unsigned long long max_possible_pfn; 100bda49a81SMike Rapoport 101fe091c20STejun Heo static struct memblock_region memblock_memory_init_regions[INIT_MEMBLOCK_REGIONS] __initdata_memblock; 1028a5b403dSArd Biesheuvel static struct memblock_region memblock_reserved_init_regions[INIT_MEMBLOCK_RESERVED_REGIONS] __initdata_memblock; 10370210ed9SPhilipp Hachtmann #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP 10470210ed9SPhilipp Hachtmann static struct memblock_region memblock_physmem_init_regions[INIT_PHYSMEM_REGIONS] __initdata_memblock; 10570210ed9SPhilipp Hachtmann #endif 106fe091c20STejun Heo 107fe091c20STejun Heo struct memblock memblock __initdata_memblock = { 108fe091c20STejun Heo .memory.regions = memblock_memory_init_regions, 109fe091c20STejun Heo .memory.cnt = 1, /* empty dummy entry */ 110fe091c20STejun Heo .memory.max = INIT_MEMBLOCK_REGIONS, 1110262d9c8SHeiko Carstens .memory.name = "memory", 112fe091c20STejun Heo 113fe091c20STejun Heo .reserved.regions = memblock_reserved_init_regions, 114fe091c20STejun Heo .reserved.cnt = 1, /* empty dummy entry */ 1158a5b403dSArd Biesheuvel .reserved.max = INIT_MEMBLOCK_RESERVED_REGIONS, 1160262d9c8SHeiko Carstens .reserved.name = "reserved", 117fe091c20STejun Heo 11870210ed9SPhilipp Hachtmann #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP 11970210ed9SPhilipp Hachtmann .physmem.regions = memblock_physmem_init_regions, 12070210ed9SPhilipp Hachtmann .physmem.cnt = 1, /* empty dummy entry */ 12170210ed9SPhilipp Hachtmann .physmem.max = INIT_PHYSMEM_REGIONS, 1220262d9c8SHeiko Carstens .physmem.name = "physmem", 12370210ed9SPhilipp Hachtmann #endif 12470210ed9SPhilipp Hachtmann 12579442ed1STang Chen .bottom_up = false, 126fe091c20STejun Heo .current_limit = MEMBLOCK_ALLOC_ANYWHERE, 127fe091c20STejun Heo }; 12895f72d1eSYinghai Lu 12910d06439SYinghai Lu int memblock_debug __initdata_memblock; 130a3f5bafcSTony Luck static bool system_has_some_mirror __initdata_memblock = false; 1311aadc056STejun Heo static int memblock_can_resize __initdata_memblock; 132181eb394SGavin Shan static int memblock_memory_in_slab __initdata_memblock = 0; 133181eb394SGavin Shan static int memblock_reserved_in_slab __initdata_memblock = 0; 13495f72d1eSYinghai Lu 135e1720feeSMike Rapoport enum memblock_flags __init_memblock choose_memblock_flags(void) 136a3f5bafcSTony Luck { 137a3f5bafcSTony Luck return system_has_some_mirror ? MEMBLOCK_MIRROR : MEMBLOCK_NONE; 138a3f5bafcSTony Luck } 139a3f5bafcSTony Luck 140eb18f1b5STejun Heo /* adjust *@size so that (@base + *@size) doesn't overflow, return new size */ 141eb18f1b5STejun Heo static inline phys_addr_t memblock_cap_size(phys_addr_t base, phys_addr_t *size) 142eb18f1b5STejun Heo { 1431c4bc43dSStefan Agner return *size = min(*size, PHYS_ADDR_MAX - base); 144eb18f1b5STejun Heo } 145eb18f1b5STejun Heo 1466ed311b2SBenjamin Herrenschmidt /* 1476ed311b2SBenjamin Herrenschmidt * Address comparison utilities 1486ed311b2SBenjamin Herrenschmidt */ 14910d06439SYinghai Lu static unsigned long __init_memblock memblock_addrs_overlap(phys_addr_t base1, phys_addr_t size1, 1502898cc4cSBenjamin Herrenschmidt phys_addr_t base2, phys_addr_t size2) 15195f72d1eSYinghai Lu { 15295f72d1eSYinghai Lu return ((base1 < (base2 + size2)) && (base2 < (base1 + size1))); 15395f72d1eSYinghai Lu } 15495f72d1eSYinghai Lu 15595cf82ecSTang Chen bool __init_memblock memblock_overlaps_region(struct memblock_type *type, 1562d7d3eb2SH Hartley Sweeten phys_addr_t base, phys_addr_t size) 1576ed311b2SBenjamin Herrenschmidt { 1586ed311b2SBenjamin Herrenschmidt unsigned long i; 1596ed311b2SBenjamin Herrenschmidt 160f14516fbSAlexander Kuleshov for (i = 0; i < type->cnt; i++) 161f14516fbSAlexander Kuleshov if (memblock_addrs_overlap(base, size, type->regions[i].base, 162f14516fbSAlexander Kuleshov type->regions[i].size)) 1636ed311b2SBenjamin Herrenschmidt break; 164c5c5c9d1STang Chen return i < type->cnt; 1656ed311b2SBenjamin Herrenschmidt } 1666ed311b2SBenjamin Herrenschmidt 16747cec443SMike Rapoport /** 16879442ed1STang Chen * __memblock_find_range_bottom_up - find free area utility in bottom-up 16979442ed1STang Chen * @start: start of candidate range 17047cec443SMike Rapoport * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or 17147cec443SMike Rapoport * %MEMBLOCK_ALLOC_ACCESSIBLE 17279442ed1STang Chen * @size: size of free area to find 17379442ed1STang Chen * @align: alignment of free area to find 174b1154233SGrygorii Strashko * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 175fc6daaf9STony Luck * @flags: pick from blocks based on memory attributes 17679442ed1STang Chen * 17779442ed1STang Chen * Utility called from memblock_find_in_range_node(), find free area bottom-up. 17879442ed1STang Chen * 17947cec443SMike Rapoport * Return: 18079442ed1STang Chen * Found address on success, 0 on failure. 18179442ed1STang Chen */ 18279442ed1STang Chen static phys_addr_t __init_memblock 18379442ed1STang Chen __memblock_find_range_bottom_up(phys_addr_t start, phys_addr_t end, 184fc6daaf9STony Luck phys_addr_t size, phys_addr_t align, int nid, 185e1720feeSMike Rapoport enum memblock_flags flags) 18679442ed1STang Chen { 18779442ed1STang Chen phys_addr_t this_start, this_end, cand; 18879442ed1STang Chen u64 i; 18979442ed1STang Chen 190fc6daaf9STony Luck for_each_free_mem_range(i, nid, flags, &this_start, &this_end, NULL) { 19179442ed1STang Chen this_start = clamp(this_start, start, end); 19279442ed1STang Chen this_end = clamp(this_end, start, end); 19379442ed1STang Chen 19479442ed1STang Chen cand = round_up(this_start, align); 19579442ed1STang Chen if (cand < this_end && this_end - cand >= size) 19679442ed1STang Chen return cand; 19779442ed1STang Chen } 19879442ed1STang Chen 19979442ed1STang Chen return 0; 20079442ed1STang Chen } 20179442ed1STang Chen 2027bd0b0f0STejun Heo /** 2031402899eSTang Chen * __memblock_find_range_top_down - find free area utility, in top-down 2041402899eSTang Chen * @start: start of candidate range 20547cec443SMike Rapoport * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or 20647cec443SMike Rapoport * %MEMBLOCK_ALLOC_ACCESSIBLE 2071402899eSTang Chen * @size: size of free area to find 2081402899eSTang Chen * @align: alignment of free area to find 209b1154233SGrygorii Strashko * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 210fc6daaf9STony Luck * @flags: pick from blocks based on memory attributes 2111402899eSTang Chen * 2121402899eSTang Chen * Utility called from memblock_find_in_range_node(), find free area top-down. 2131402899eSTang Chen * 21447cec443SMike Rapoport * Return: 21579442ed1STang Chen * Found address on success, 0 on failure. 2161402899eSTang Chen */ 2171402899eSTang Chen static phys_addr_t __init_memblock 2181402899eSTang Chen __memblock_find_range_top_down(phys_addr_t start, phys_addr_t end, 219fc6daaf9STony Luck phys_addr_t size, phys_addr_t align, int nid, 220e1720feeSMike Rapoport enum memblock_flags flags) 2211402899eSTang Chen { 2221402899eSTang Chen phys_addr_t this_start, this_end, cand; 2231402899eSTang Chen u64 i; 2241402899eSTang Chen 225fc6daaf9STony Luck for_each_free_mem_range_reverse(i, nid, flags, &this_start, &this_end, 226fc6daaf9STony Luck NULL) { 2271402899eSTang Chen this_start = clamp(this_start, start, end); 2281402899eSTang Chen this_end = clamp(this_end, start, end); 2291402899eSTang Chen 2301402899eSTang Chen if (this_end < size) 2311402899eSTang Chen continue; 2321402899eSTang Chen 2331402899eSTang Chen cand = round_down(this_end - size, align); 2341402899eSTang Chen if (cand >= this_start) 2351402899eSTang Chen return cand; 2361402899eSTang Chen } 2371402899eSTang Chen 2381402899eSTang Chen return 0; 2391402899eSTang Chen } 2401402899eSTang Chen 2411402899eSTang Chen /** 2427bd0b0f0STejun Heo * memblock_find_in_range_node - find free area in given range and node 2437bd0b0f0STejun Heo * @size: size of free area to find 2447bd0b0f0STejun Heo * @align: alignment of free area to find 24587029ee9SGrygorii Strashko * @start: start of candidate range 24647cec443SMike Rapoport * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or 24747cec443SMike Rapoport * %MEMBLOCK_ALLOC_ACCESSIBLE 248b1154233SGrygorii Strashko * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 249fc6daaf9STony Luck * @flags: pick from blocks based on memory attributes 2507bd0b0f0STejun Heo * 2517bd0b0f0STejun Heo * Find @size free area aligned to @align in the specified range and node. 2527bd0b0f0STejun Heo * 25379442ed1STang Chen * When allocation direction is bottom-up, the @start should be greater 25479442ed1STang Chen * than the end of the kernel image. Otherwise, it will be trimmed. The 25579442ed1STang Chen * reason is that we want the bottom-up allocation just near the kernel 25679442ed1STang Chen * image so it is highly likely that the allocated memory and the kernel 25779442ed1STang Chen * will reside in the same node. 25879442ed1STang Chen * 25979442ed1STang Chen * If bottom-up allocation failed, will try to allocate memory top-down. 26079442ed1STang Chen * 26147cec443SMike Rapoport * Return: 26279442ed1STang Chen * Found address on success, 0 on failure. 2636ed311b2SBenjamin Herrenschmidt */ 26487029ee9SGrygorii Strashko phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, 26587029ee9SGrygorii Strashko phys_addr_t align, phys_addr_t start, 266e1720feeSMike Rapoport phys_addr_t end, int nid, 267e1720feeSMike Rapoport enum memblock_flags flags) 268f7210e6cSTang Chen { 2690cfb8f0cSTang Chen phys_addr_t kernel_end, ret; 27079442ed1STang Chen 271f7210e6cSTang Chen /* pump up @end */ 272fed84c78SQian Cai if (end == MEMBLOCK_ALLOC_ACCESSIBLE || 273fed84c78SQian Cai end == MEMBLOCK_ALLOC_KASAN) 274f7210e6cSTang Chen end = memblock.current_limit; 275f7210e6cSTang Chen 276f7210e6cSTang Chen /* avoid allocating the first page */ 277f7210e6cSTang Chen start = max_t(phys_addr_t, start, PAGE_SIZE); 278f7210e6cSTang Chen end = max(start, end); 27979442ed1STang Chen kernel_end = __pa_symbol(_end); 28079442ed1STang Chen 28179442ed1STang Chen /* 28279442ed1STang Chen * try bottom-up allocation only when bottom-up mode 28379442ed1STang Chen * is set and @end is above the kernel image. 28479442ed1STang Chen */ 28579442ed1STang Chen if (memblock_bottom_up() && end > kernel_end) { 28679442ed1STang Chen phys_addr_t bottom_up_start; 28779442ed1STang Chen 28879442ed1STang Chen /* make sure we will allocate above the kernel */ 28979442ed1STang Chen bottom_up_start = max(start, kernel_end); 29079442ed1STang Chen 29179442ed1STang Chen /* ok, try bottom-up allocation first */ 29279442ed1STang Chen ret = __memblock_find_range_bottom_up(bottom_up_start, end, 293fc6daaf9STony Luck size, align, nid, flags); 29479442ed1STang Chen if (ret) 29579442ed1STang Chen return ret; 29679442ed1STang Chen 29779442ed1STang Chen /* 29879442ed1STang Chen * we always limit bottom-up allocation above the kernel, 29979442ed1STang Chen * but top-down allocation doesn't have the limit, so 30079442ed1STang Chen * retrying top-down allocation may succeed when bottom-up 30179442ed1STang Chen * allocation failed. 30279442ed1STang Chen * 30379442ed1STang Chen * bottom-up allocation is expected to be fail very rarely, 30479442ed1STang Chen * so we use WARN_ONCE() here to see the stack trace if 30579442ed1STang Chen * fail happens. 30679442ed1STang Chen */ 307e3d301caSMichal Hocko WARN_ONCE(IS_ENABLED(CONFIG_MEMORY_HOTREMOVE), 308e3d301caSMichal Hocko "memblock: bottom-up allocation failed, memory hotremove may be affected\n"); 30979442ed1STang Chen } 310f7210e6cSTang Chen 311fc6daaf9STony Luck return __memblock_find_range_top_down(start, end, size, align, nid, 312fc6daaf9STony Luck flags); 313f7210e6cSTang Chen } 3146ed311b2SBenjamin Herrenschmidt 3157bd0b0f0STejun Heo /** 3167bd0b0f0STejun Heo * memblock_find_in_range - find free area in given range 3177bd0b0f0STejun Heo * @start: start of candidate range 31847cec443SMike Rapoport * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or 31947cec443SMike Rapoport * %MEMBLOCK_ALLOC_ACCESSIBLE 3207bd0b0f0STejun Heo * @size: size of free area to find 3217bd0b0f0STejun Heo * @align: alignment of free area to find 3227bd0b0f0STejun Heo * 3237bd0b0f0STejun Heo * Find @size free area aligned to @align in the specified range. 3247bd0b0f0STejun Heo * 32547cec443SMike Rapoport * Return: 32679442ed1STang Chen * Found address on success, 0 on failure. 3277bd0b0f0STejun Heo */ 3287bd0b0f0STejun Heo phys_addr_t __init_memblock memblock_find_in_range(phys_addr_t start, 3297bd0b0f0STejun Heo phys_addr_t end, phys_addr_t size, 3307bd0b0f0STejun Heo phys_addr_t align) 3317bd0b0f0STejun Heo { 332a3f5bafcSTony Luck phys_addr_t ret; 333e1720feeSMike Rapoport enum memblock_flags flags = choose_memblock_flags(); 334a3f5bafcSTony Luck 335a3f5bafcSTony Luck again: 336a3f5bafcSTony Luck ret = memblock_find_in_range_node(size, align, start, end, 337a3f5bafcSTony Luck NUMA_NO_NODE, flags); 338a3f5bafcSTony Luck 339a3f5bafcSTony Luck if (!ret && (flags & MEMBLOCK_MIRROR)) { 340a3f5bafcSTony Luck pr_warn("Could not allocate %pap bytes of mirrored memory\n", 341a3f5bafcSTony Luck &size); 342a3f5bafcSTony Luck flags &= ~MEMBLOCK_MIRROR; 343a3f5bafcSTony Luck goto again; 344a3f5bafcSTony Luck } 345a3f5bafcSTony Luck 346a3f5bafcSTony Luck return ret; 3477bd0b0f0STejun Heo } 3487bd0b0f0STejun Heo 34910d06439SYinghai Lu static void __init_memblock memblock_remove_region(struct memblock_type *type, unsigned long r) 35095f72d1eSYinghai Lu { 3511440c4e2STejun Heo type->total_size -= type->regions[r].size; 3527c0caeb8STejun Heo memmove(&type->regions[r], &type->regions[r + 1], 3537c0caeb8STejun Heo (type->cnt - (r + 1)) * sizeof(type->regions[r])); 354e3239ff9SBenjamin Herrenschmidt type->cnt--; 35595f72d1eSYinghai Lu 3568f7a6605SBenjamin Herrenschmidt /* Special case for empty arrays */ 3578f7a6605SBenjamin Herrenschmidt if (type->cnt == 0) { 3581440c4e2STejun Heo WARN_ON(type->total_size != 0); 3598f7a6605SBenjamin Herrenschmidt type->cnt = 1; 3608f7a6605SBenjamin Herrenschmidt type->regions[0].base = 0; 3618f7a6605SBenjamin Herrenschmidt type->regions[0].size = 0; 36266a20757STang Chen type->regions[0].flags = 0; 3637c0caeb8STejun Heo memblock_set_region_node(&type->regions[0], MAX_NUMNODES); 3648f7a6605SBenjamin Herrenschmidt } 36595f72d1eSYinghai Lu } 36695f72d1eSYinghai Lu 367354f17e1SPhilipp Hachtmann #ifdef CONFIG_ARCH_DISCARD_MEMBLOCK 3683010f876SPavel Tatashin /** 36947cec443SMike Rapoport * memblock_discard - discard memory and reserved arrays if they were allocated 3703010f876SPavel Tatashin */ 3713010f876SPavel Tatashin void __init memblock_discard(void) 37229f67386SYinghai Lu { 3733010f876SPavel Tatashin phys_addr_t addr, size; 37429f67386SYinghai Lu 3753010f876SPavel Tatashin if (memblock.reserved.regions != memblock_reserved_init_regions) { 3763010f876SPavel Tatashin addr = __pa(memblock.reserved.regions); 3773010f876SPavel Tatashin size = PAGE_ALIGN(sizeof(struct memblock_region) * 37829f67386SYinghai Lu memblock.reserved.max); 3793010f876SPavel Tatashin __memblock_free_late(addr, size); 38029f67386SYinghai Lu } 38129f67386SYinghai Lu 38291b540f9SPavel Tatashin if (memblock.memory.regions != memblock_memory_init_regions) { 3833010f876SPavel Tatashin addr = __pa(memblock.memory.regions); 3843010f876SPavel Tatashin size = PAGE_ALIGN(sizeof(struct memblock_region) * 3855e270e25SPhilipp Hachtmann memblock.memory.max); 3863010f876SPavel Tatashin __memblock_free_late(addr, size); 3875e270e25SPhilipp Hachtmann } 3883010f876SPavel Tatashin } 3895e270e25SPhilipp Hachtmann #endif 3905e270e25SPhilipp Hachtmann 39148c3b583SGreg Pearson /** 39248c3b583SGreg Pearson * memblock_double_array - double the size of the memblock regions array 39348c3b583SGreg Pearson * @type: memblock type of the regions array being doubled 39448c3b583SGreg Pearson * @new_area_start: starting address of memory range to avoid overlap with 39548c3b583SGreg Pearson * @new_area_size: size of memory range to avoid overlap with 39648c3b583SGreg Pearson * 39748c3b583SGreg Pearson * Double the size of the @type regions array. If memblock is being used to 39848c3b583SGreg Pearson * allocate memory for a new reserved regions array and there is a previously 39948c3b583SGreg Pearson * allocated memory range [@new_area_start, @new_area_start + @new_area_size] 40048c3b583SGreg Pearson * waiting to be reserved, ensure the memory used by the new array does 40148c3b583SGreg Pearson * not overlap. 40248c3b583SGreg Pearson * 40347cec443SMike Rapoport * Return: 40448c3b583SGreg Pearson * 0 on success, -1 on failure. 40548c3b583SGreg Pearson */ 40648c3b583SGreg Pearson static int __init_memblock memblock_double_array(struct memblock_type *type, 40748c3b583SGreg Pearson phys_addr_t new_area_start, 40848c3b583SGreg Pearson phys_addr_t new_area_size) 409142b45a7SBenjamin Herrenschmidt { 410142b45a7SBenjamin Herrenschmidt struct memblock_region *new_array, *old_array; 41129f67386SYinghai Lu phys_addr_t old_alloc_size, new_alloc_size; 412a36aab89SMike Rapoport phys_addr_t old_size, new_size, addr, new_end; 413142b45a7SBenjamin Herrenschmidt int use_slab = slab_is_available(); 414181eb394SGavin Shan int *in_slab; 415142b45a7SBenjamin Herrenschmidt 416142b45a7SBenjamin Herrenschmidt /* We don't allow resizing until we know about the reserved regions 417142b45a7SBenjamin Herrenschmidt * of memory that aren't suitable for allocation 418142b45a7SBenjamin Herrenschmidt */ 419142b45a7SBenjamin Herrenschmidt if (!memblock_can_resize) 420142b45a7SBenjamin Herrenschmidt return -1; 421142b45a7SBenjamin Herrenschmidt 422142b45a7SBenjamin Herrenschmidt /* Calculate new doubled size */ 423142b45a7SBenjamin Herrenschmidt old_size = type->max * sizeof(struct memblock_region); 424142b45a7SBenjamin Herrenschmidt new_size = old_size << 1; 42529f67386SYinghai Lu /* 42629f67386SYinghai Lu * We need to allocated new one align to PAGE_SIZE, 42729f67386SYinghai Lu * so we can free them completely later. 42829f67386SYinghai Lu */ 42929f67386SYinghai Lu old_alloc_size = PAGE_ALIGN(old_size); 43029f67386SYinghai Lu new_alloc_size = PAGE_ALIGN(new_size); 431142b45a7SBenjamin Herrenschmidt 432181eb394SGavin Shan /* Retrieve the slab flag */ 433181eb394SGavin Shan if (type == &memblock.memory) 434181eb394SGavin Shan in_slab = &memblock_memory_in_slab; 435181eb394SGavin Shan else 436181eb394SGavin Shan in_slab = &memblock_reserved_in_slab; 437181eb394SGavin Shan 438142b45a7SBenjamin Herrenschmidt /* Try to find some space for it. 439142b45a7SBenjamin Herrenschmidt * 440142b45a7SBenjamin Herrenschmidt * WARNING: We assume that either slab_is_available() and we use it or 441fd07383bSAndrew Morton * we use MEMBLOCK for allocations. That means that this is unsafe to 442fd07383bSAndrew Morton * use when bootmem is currently active (unless bootmem itself is 443fd07383bSAndrew Morton * implemented on top of MEMBLOCK which isn't the case yet) 444142b45a7SBenjamin Herrenschmidt * 445142b45a7SBenjamin Herrenschmidt * This should however not be an issue for now, as we currently only 446fd07383bSAndrew Morton * call into MEMBLOCK while it's still active, or much later when slab 447fd07383bSAndrew Morton * is active for memory hotplug operations 448142b45a7SBenjamin Herrenschmidt */ 449142b45a7SBenjamin Herrenschmidt if (use_slab) { 450142b45a7SBenjamin Herrenschmidt new_array = kmalloc(new_size, GFP_KERNEL); 4511f5026a7STejun Heo addr = new_array ? __pa(new_array) : 0; 4524e2f0775SGavin Shan } else { 45348c3b583SGreg Pearson /* only exclude range when trying to double reserved.regions */ 45448c3b583SGreg Pearson if (type != &memblock.reserved) 45548c3b583SGreg Pearson new_area_start = new_area_size = 0; 45648c3b583SGreg Pearson 45748c3b583SGreg Pearson addr = memblock_find_in_range(new_area_start + new_area_size, 45848c3b583SGreg Pearson memblock.current_limit, 45929f67386SYinghai Lu new_alloc_size, PAGE_SIZE); 46048c3b583SGreg Pearson if (!addr && new_area_size) 46148c3b583SGreg Pearson addr = memblock_find_in_range(0, 46248c3b583SGreg Pearson min(new_area_start, memblock.current_limit), 46329f67386SYinghai Lu new_alloc_size, PAGE_SIZE); 46448c3b583SGreg Pearson 46515674868SSachin Kamat new_array = addr ? __va(addr) : NULL; 4664e2f0775SGavin Shan } 4671f5026a7STejun Heo if (!addr) { 468142b45a7SBenjamin Herrenschmidt pr_err("memblock: Failed to double %s array from %ld to %ld entries !\n", 4690262d9c8SHeiko Carstens type->name, type->max, type->max * 2); 470142b45a7SBenjamin Herrenschmidt return -1; 471142b45a7SBenjamin Herrenschmidt } 472142b45a7SBenjamin Herrenschmidt 473a36aab89SMike Rapoport new_end = addr + new_size - 1; 474a36aab89SMike Rapoport memblock_dbg("memblock: %s is doubled to %ld at [%pa-%pa]", 475a36aab89SMike Rapoport type->name, type->max * 2, &addr, &new_end); 476ea9e4376SYinghai Lu 477fd07383bSAndrew Morton /* 478fd07383bSAndrew Morton * Found space, we now need to move the array over before we add the 479fd07383bSAndrew Morton * reserved region since it may be our reserved array itself that is 480fd07383bSAndrew Morton * full. 481142b45a7SBenjamin Herrenschmidt */ 482142b45a7SBenjamin Herrenschmidt memcpy(new_array, type->regions, old_size); 483142b45a7SBenjamin Herrenschmidt memset(new_array + type->max, 0, old_size); 484142b45a7SBenjamin Herrenschmidt old_array = type->regions; 485142b45a7SBenjamin Herrenschmidt type->regions = new_array; 486142b45a7SBenjamin Herrenschmidt type->max <<= 1; 487142b45a7SBenjamin Herrenschmidt 488fd07383bSAndrew Morton /* Free old array. We needn't free it if the array is the static one */ 489181eb394SGavin Shan if (*in_slab) 490181eb394SGavin Shan kfree(old_array); 491181eb394SGavin Shan else if (old_array != memblock_memory_init_regions && 492142b45a7SBenjamin Herrenschmidt old_array != memblock_reserved_init_regions) 49329f67386SYinghai Lu memblock_free(__pa(old_array), old_alloc_size); 494142b45a7SBenjamin Herrenschmidt 495fd07383bSAndrew Morton /* 496fd07383bSAndrew Morton * Reserve the new array if that comes from the memblock. Otherwise, we 497fd07383bSAndrew Morton * needn't do it 498181eb394SGavin Shan */ 499181eb394SGavin Shan if (!use_slab) 50029f67386SYinghai Lu BUG_ON(memblock_reserve(addr, new_alloc_size)); 501181eb394SGavin Shan 502181eb394SGavin Shan /* Update slab flag */ 503181eb394SGavin Shan *in_slab = use_slab; 504181eb394SGavin Shan 505142b45a7SBenjamin Herrenschmidt return 0; 506142b45a7SBenjamin Herrenschmidt } 507142b45a7SBenjamin Herrenschmidt 508784656f9STejun Heo /** 509784656f9STejun Heo * memblock_merge_regions - merge neighboring compatible regions 510784656f9STejun Heo * @type: memblock type to scan 511784656f9STejun Heo * 512784656f9STejun Heo * Scan @type and merge neighboring compatible regions. 513784656f9STejun Heo */ 514784656f9STejun Heo static void __init_memblock memblock_merge_regions(struct memblock_type *type) 515784656f9STejun Heo { 516784656f9STejun Heo int i = 0; 517784656f9STejun Heo 518784656f9STejun Heo /* cnt never goes below 1 */ 519784656f9STejun Heo while (i < type->cnt - 1) { 520784656f9STejun Heo struct memblock_region *this = &type->regions[i]; 521784656f9STejun Heo struct memblock_region *next = &type->regions[i + 1]; 522784656f9STejun Heo 5237c0caeb8STejun Heo if (this->base + this->size != next->base || 5247c0caeb8STejun Heo memblock_get_region_node(this) != 52566a20757STang Chen memblock_get_region_node(next) || 52666a20757STang Chen this->flags != next->flags) { 527784656f9STejun Heo BUG_ON(this->base + this->size > next->base); 528784656f9STejun Heo i++; 529784656f9STejun Heo continue; 530784656f9STejun Heo } 531784656f9STejun Heo 532784656f9STejun Heo this->size += next->size; 533c0232ae8SLin Feng /* move forward from next + 1, index of which is i + 2 */ 534c0232ae8SLin Feng memmove(next, next + 1, (type->cnt - (i + 2)) * sizeof(*next)); 535784656f9STejun Heo type->cnt--; 536784656f9STejun Heo } 537784656f9STejun Heo } 538784656f9STejun Heo 539784656f9STejun Heo /** 540784656f9STejun Heo * memblock_insert_region - insert new memblock region 541784656f9STejun Heo * @type: memblock type to insert into 542784656f9STejun Heo * @idx: index for the insertion point 543784656f9STejun Heo * @base: base address of the new region 544784656f9STejun Heo * @size: size of the new region 545209ff86dSTang Chen * @nid: node id of the new region 54666a20757STang Chen * @flags: flags of the new region 547784656f9STejun Heo * 548784656f9STejun Heo * Insert new memblock region [@base, @base + @size) into @type at @idx. 549412d0008SAlexander Kuleshov * @type must already have extra room to accommodate the new region. 550784656f9STejun Heo */ 551784656f9STejun Heo static void __init_memblock memblock_insert_region(struct memblock_type *type, 552784656f9STejun Heo int idx, phys_addr_t base, 55366a20757STang Chen phys_addr_t size, 554e1720feeSMike Rapoport int nid, 555e1720feeSMike Rapoport enum memblock_flags flags) 556784656f9STejun Heo { 557784656f9STejun Heo struct memblock_region *rgn = &type->regions[idx]; 558784656f9STejun Heo 559784656f9STejun Heo BUG_ON(type->cnt >= type->max); 560784656f9STejun Heo memmove(rgn + 1, rgn, (type->cnt - idx) * sizeof(*rgn)); 561784656f9STejun Heo rgn->base = base; 562784656f9STejun Heo rgn->size = size; 56366a20757STang Chen rgn->flags = flags; 5647c0caeb8STejun Heo memblock_set_region_node(rgn, nid); 565784656f9STejun Heo type->cnt++; 5661440c4e2STejun Heo type->total_size += size; 567784656f9STejun Heo } 568784656f9STejun Heo 569784656f9STejun Heo /** 570f1af9d3aSPhilipp Hachtmann * memblock_add_range - add new memblock region 571784656f9STejun Heo * @type: memblock type to add new region into 572784656f9STejun Heo * @base: base address of the new region 573784656f9STejun Heo * @size: size of the new region 5747fb0bc3fSTejun Heo * @nid: nid of the new region 57566a20757STang Chen * @flags: flags of the new region 576784656f9STejun Heo * 577784656f9STejun Heo * Add new memblock region [@base, @base + @size) into @type. The new region 578784656f9STejun Heo * is allowed to overlap with existing ones - overlaps don't affect already 579784656f9STejun Heo * existing regions. @type is guaranteed to be minimal (all neighbouring 580784656f9STejun Heo * compatible regions are merged) after the addition. 581784656f9STejun Heo * 58247cec443SMike Rapoport * Return: 583784656f9STejun Heo * 0 on success, -errno on failure. 584784656f9STejun Heo */ 585f1af9d3aSPhilipp Hachtmann int __init_memblock memblock_add_range(struct memblock_type *type, 58666a20757STang Chen phys_addr_t base, phys_addr_t size, 587e1720feeSMike Rapoport int nid, enum memblock_flags flags) 58895f72d1eSYinghai Lu { 589784656f9STejun Heo bool insert = false; 590eb18f1b5STejun Heo phys_addr_t obase = base; 591eb18f1b5STejun Heo phys_addr_t end = base + memblock_cap_size(base, &size); 5928c9c1701SAlexander Kuleshov int idx, nr_new; 5938c9c1701SAlexander Kuleshov struct memblock_region *rgn; 59495f72d1eSYinghai Lu 595b3dc627cSTejun Heo if (!size) 596b3dc627cSTejun Heo return 0; 597b3dc627cSTejun Heo 598784656f9STejun Heo /* special case for empty array */ 599784656f9STejun Heo if (type->regions[0].size == 0) { 6001440c4e2STejun Heo WARN_ON(type->cnt != 1 || type->total_size); 601784656f9STejun Heo type->regions[0].base = base; 602784656f9STejun Heo type->regions[0].size = size; 60366a20757STang Chen type->regions[0].flags = flags; 6047fb0bc3fSTejun Heo memblock_set_region_node(&type->regions[0], nid); 6051440c4e2STejun Heo type->total_size = size; 606784656f9STejun Heo return 0; 607784656f9STejun Heo } 608784656f9STejun Heo repeat: 609784656f9STejun Heo /* 610784656f9STejun Heo * The following is executed twice. Once with %false @insert and 611784656f9STejun Heo * then with %true. The first counts the number of regions needed 612412d0008SAlexander Kuleshov * to accommodate the new area. The second actually inserts them. 613784656f9STejun Heo */ 614784656f9STejun Heo base = obase; 615784656f9STejun Heo nr_new = 0; 616784656f9STejun Heo 61766e8b438SGioh Kim for_each_memblock_type(idx, type, rgn) { 618784656f9STejun Heo phys_addr_t rbase = rgn->base; 619784656f9STejun Heo phys_addr_t rend = rbase + rgn->size; 6208f7a6605SBenjamin Herrenschmidt 621784656f9STejun Heo if (rbase >= end) 6228f7a6605SBenjamin Herrenschmidt break; 623784656f9STejun Heo if (rend <= base) 624784656f9STejun Heo continue; 625784656f9STejun Heo /* 626784656f9STejun Heo * @rgn overlaps. If it separates the lower part of new 627784656f9STejun Heo * area, insert that portion. 6288f7a6605SBenjamin Herrenschmidt */ 629784656f9STejun Heo if (rbase > base) { 630c0a29498SWei Yang #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP 631c0a29498SWei Yang WARN_ON(nid != memblock_get_region_node(rgn)); 632c0a29498SWei Yang #endif 6334fcab5f4SWei Yang WARN_ON(flags != rgn->flags); 634784656f9STejun Heo nr_new++; 635784656f9STejun Heo if (insert) 6368c9c1701SAlexander Kuleshov memblock_insert_region(type, idx++, base, 63766a20757STang Chen rbase - base, nid, 63866a20757STang Chen flags); 639784656f9STejun Heo } 640784656f9STejun Heo /* area below @rend is dealt with, forget about it */ 641784656f9STejun Heo base = min(rend, end); 6428f7a6605SBenjamin Herrenschmidt } 6438f7a6605SBenjamin Herrenschmidt 644784656f9STejun Heo /* insert the remaining portion */ 645784656f9STejun Heo if (base < end) { 646784656f9STejun Heo nr_new++; 647784656f9STejun Heo if (insert) 6488c9c1701SAlexander Kuleshov memblock_insert_region(type, idx, base, end - base, 64966a20757STang Chen nid, flags); 6508f7a6605SBenjamin Herrenschmidt } 6518f7a6605SBenjamin Herrenschmidt 652ef3cc4dbSnimisolo if (!nr_new) 653ef3cc4dbSnimisolo return 0; 654ef3cc4dbSnimisolo 655784656f9STejun Heo /* 656784656f9STejun Heo * If this was the first round, resize array and repeat for actual 657784656f9STejun Heo * insertions; otherwise, merge and return. 6588f7a6605SBenjamin Herrenschmidt */ 659784656f9STejun Heo if (!insert) { 660784656f9STejun Heo while (type->cnt + nr_new > type->max) 66148c3b583SGreg Pearson if (memblock_double_array(type, obase, size) < 0) 662784656f9STejun Heo return -ENOMEM; 663784656f9STejun Heo insert = true; 664784656f9STejun Heo goto repeat; 66595f72d1eSYinghai Lu } else { 666784656f9STejun Heo memblock_merge_regions(type); 66795f72d1eSYinghai Lu return 0; 66895f72d1eSYinghai Lu } 669784656f9STejun Heo } 67095f72d1eSYinghai Lu 67148a833ccSMike Rapoport /** 67248a833ccSMike Rapoport * memblock_add_node - add new memblock region within a NUMA node 67348a833ccSMike Rapoport * @base: base address of the new region 67448a833ccSMike Rapoport * @size: size of the new region 67548a833ccSMike Rapoport * @nid: nid of the new region 67648a833ccSMike Rapoport * 67748a833ccSMike Rapoport * Add new memblock region [@base, @base + @size) to the "memory" 67848a833ccSMike Rapoport * type. See memblock_add_range() description for mode details 67948a833ccSMike Rapoport * 68048a833ccSMike Rapoport * Return: 68148a833ccSMike Rapoport * 0 on success, -errno on failure. 68248a833ccSMike Rapoport */ 6837fb0bc3fSTejun Heo int __init_memblock memblock_add_node(phys_addr_t base, phys_addr_t size, 6847fb0bc3fSTejun Heo int nid) 6857fb0bc3fSTejun Heo { 686f1af9d3aSPhilipp Hachtmann return memblock_add_range(&memblock.memory, base, size, nid, 0); 6877fb0bc3fSTejun Heo } 6887fb0bc3fSTejun Heo 68948a833ccSMike Rapoport /** 69048a833ccSMike Rapoport * memblock_add - add new memblock region 69148a833ccSMike Rapoport * @base: base address of the new region 69248a833ccSMike Rapoport * @size: size of the new region 69348a833ccSMike Rapoport * 69448a833ccSMike Rapoport * Add new memblock region [@base, @base + @size) to the "memory" 69548a833ccSMike Rapoport * type. See memblock_add_range() description for mode details 69648a833ccSMike Rapoport * 69748a833ccSMike Rapoport * Return: 69848a833ccSMike Rapoport * 0 on success, -errno on failure. 69948a833ccSMike Rapoport */ 700f705ac4bSAlexander Kuleshov int __init_memblock memblock_add(phys_addr_t base, phys_addr_t size) 7016a4055bcSAlexander Kuleshov { 7025d63f81cSMiles Chen phys_addr_t end = base + size - 1; 7035d63f81cSMiles Chen 7045d63f81cSMiles Chen memblock_dbg("memblock_add: [%pa-%pa] %pF\n", 7055d63f81cSMiles Chen &base, &end, (void *)_RET_IP_); 7066a4055bcSAlexander Kuleshov 707f705ac4bSAlexander Kuleshov return memblock_add_range(&memblock.memory, base, size, MAX_NUMNODES, 0); 70895f72d1eSYinghai Lu } 70995f72d1eSYinghai Lu 7106a9ceb31STejun Heo /** 7116a9ceb31STejun Heo * memblock_isolate_range - isolate given range into disjoint memblocks 7126a9ceb31STejun Heo * @type: memblock type to isolate range for 7136a9ceb31STejun Heo * @base: base of range to isolate 7146a9ceb31STejun Heo * @size: size of range to isolate 7156a9ceb31STejun Heo * @start_rgn: out parameter for the start of isolated region 7166a9ceb31STejun Heo * @end_rgn: out parameter for the end of isolated region 7176a9ceb31STejun Heo * 7186a9ceb31STejun Heo * Walk @type and ensure that regions don't cross the boundaries defined by 7196a9ceb31STejun Heo * [@base, @base + @size). Crossing regions are split at the boundaries, 7206a9ceb31STejun Heo * which may create at most two more regions. The index of the first 7216a9ceb31STejun Heo * region inside the range is returned in *@start_rgn and end in *@end_rgn. 7226a9ceb31STejun Heo * 72347cec443SMike Rapoport * Return: 7246a9ceb31STejun Heo * 0 on success, -errno on failure. 7256a9ceb31STejun Heo */ 7266a9ceb31STejun Heo static int __init_memblock memblock_isolate_range(struct memblock_type *type, 7276a9ceb31STejun Heo phys_addr_t base, phys_addr_t size, 7286a9ceb31STejun Heo int *start_rgn, int *end_rgn) 7296a9ceb31STejun Heo { 730eb18f1b5STejun Heo phys_addr_t end = base + memblock_cap_size(base, &size); 7318c9c1701SAlexander Kuleshov int idx; 7328c9c1701SAlexander Kuleshov struct memblock_region *rgn; 7336a9ceb31STejun Heo 7346a9ceb31STejun Heo *start_rgn = *end_rgn = 0; 7356a9ceb31STejun Heo 736b3dc627cSTejun Heo if (!size) 737b3dc627cSTejun Heo return 0; 738b3dc627cSTejun Heo 7396a9ceb31STejun Heo /* we'll create at most two more regions */ 7406a9ceb31STejun Heo while (type->cnt + 2 > type->max) 74148c3b583SGreg Pearson if (memblock_double_array(type, base, size) < 0) 7426a9ceb31STejun Heo return -ENOMEM; 7436a9ceb31STejun Heo 74466e8b438SGioh Kim for_each_memblock_type(idx, type, rgn) { 7456a9ceb31STejun Heo phys_addr_t rbase = rgn->base; 7466a9ceb31STejun Heo phys_addr_t rend = rbase + rgn->size; 7476a9ceb31STejun Heo 7486a9ceb31STejun Heo if (rbase >= end) 7496a9ceb31STejun Heo break; 7506a9ceb31STejun Heo if (rend <= base) 7516a9ceb31STejun Heo continue; 7526a9ceb31STejun Heo 7536a9ceb31STejun Heo if (rbase < base) { 7546a9ceb31STejun Heo /* 7556a9ceb31STejun Heo * @rgn intersects from below. Split and continue 7566a9ceb31STejun Heo * to process the next region - the new top half. 7576a9ceb31STejun Heo */ 7586a9ceb31STejun Heo rgn->base = base; 7591440c4e2STejun Heo rgn->size -= base - rbase; 7601440c4e2STejun Heo type->total_size -= base - rbase; 7618c9c1701SAlexander Kuleshov memblock_insert_region(type, idx, rbase, base - rbase, 76266a20757STang Chen memblock_get_region_node(rgn), 76366a20757STang Chen rgn->flags); 7646a9ceb31STejun Heo } else if (rend > end) { 7656a9ceb31STejun Heo /* 7666a9ceb31STejun Heo * @rgn intersects from above. Split and redo the 7676a9ceb31STejun Heo * current region - the new bottom half. 7686a9ceb31STejun Heo */ 7696a9ceb31STejun Heo rgn->base = end; 7701440c4e2STejun Heo rgn->size -= end - rbase; 7711440c4e2STejun Heo type->total_size -= end - rbase; 7728c9c1701SAlexander Kuleshov memblock_insert_region(type, idx--, rbase, end - rbase, 77366a20757STang Chen memblock_get_region_node(rgn), 77466a20757STang Chen rgn->flags); 7756a9ceb31STejun Heo } else { 7766a9ceb31STejun Heo /* @rgn is fully contained, record it */ 7776a9ceb31STejun Heo if (!*end_rgn) 7788c9c1701SAlexander Kuleshov *start_rgn = idx; 7798c9c1701SAlexander Kuleshov *end_rgn = idx + 1; 7806a9ceb31STejun Heo } 7816a9ceb31STejun Heo } 7826a9ceb31STejun Heo 7836a9ceb31STejun Heo return 0; 7846a9ceb31STejun Heo } 7856a9ceb31STejun Heo 78635bd16a2SAlexander Kuleshov static int __init_memblock memblock_remove_range(struct memblock_type *type, 7878f7a6605SBenjamin Herrenschmidt phys_addr_t base, phys_addr_t size) 78895f72d1eSYinghai Lu { 78971936180STejun Heo int start_rgn, end_rgn; 79071936180STejun Heo int i, ret; 79195f72d1eSYinghai Lu 79271936180STejun Heo ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); 79371936180STejun Heo if (ret) 79471936180STejun Heo return ret; 79595f72d1eSYinghai Lu 79671936180STejun Heo for (i = end_rgn - 1; i >= start_rgn; i--) 79771936180STejun Heo memblock_remove_region(type, i); 79895f72d1eSYinghai Lu return 0; 79995f72d1eSYinghai Lu } 80095f72d1eSYinghai Lu 801581adcbeSTejun Heo int __init_memblock memblock_remove(phys_addr_t base, phys_addr_t size) 80295f72d1eSYinghai Lu { 80325cf23d7SMinchan Kim phys_addr_t end = base + size - 1; 80425cf23d7SMinchan Kim 80525cf23d7SMinchan Kim memblock_dbg("memblock_remove: [%pa-%pa] %pS\n", 80625cf23d7SMinchan Kim &base, &end, (void *)_RET_IP_); 80725cf23d7SMinchan Kim 808f1af9d3aSPhilipp Hachtmann return memblock_remove_range(&memblock.memory, base, size); 80995f72d1eSYinghai Lu } 81095f72d1eSYinghai Lu 8114d72868cSMike Rapoport /** 8124d72868cSMike Rapoport * memblock_free - free boot memory block 8134d72868cSMike Rapoport * @base: phys starting address of the boot memory block 8144d72868cSMike Rapoport * @size: size of the boot memory block in bytes 8154d72868cSMike Rapoport * 8164d72868cSMike Rapoport * Free boot memory block previously allocated by memblock_alloc_xx() API. 8174d72868cSMike Rapoport * The freeing memory will not be released to the buddy allocator. 8184d72868cSMike Rapoport */ 819581adcbeSTejun Heo int __init_memblock memblock_free(phys_addr_t base, phys_addr_t size) 82095f72d1eSYinghai Lu { 8215d63f81cSMiles Chen phys_addr_t end = base + size - 1; 8225d63f81cSMiles Chen 8235d63f81cSMiles Chen memblock_dbg(" memblock_free: [%pa-%pa] %pF\n", 8245d63f81cSMiles Chen &base, &end, (void *)_RET_IP_); 82524aa0788STejun Heo 8269099daedSCatalin Marinas kmemleak_free_part_phys(base, size); 827f1af9d3aSPhilipp Hachtmann return memblock_remove_range(&memblock.reserved, base, size); 82895f72d1eSYinghai Lu } 82995f72d1eSYinghai Lu 830f705ac4bSAlexander Kuleshov int __init_memblock memblock_reserve(phys_addr_t base, phys_addr_t size) 83195f72d1eSYinghai Lu { 8325d63f81cSMiles Chen phys_addr_t end = base + size - 1; 8335d63f81cSMiles Chen 8345d63f81cSMiles Chen memblock_dbg("memblock_reserve: [%pa-%pa] %pF\n", 8355d63f81cSMiles Chen &base, &end, (void *)_RET_IP_); 83695f72d1eSYinghai Lu 837f705ac4bSAlexander Kuleshov return memblock_add_range(&memblock.reserved, base, size, MAX_NUMNODES, 0); 83895f72d1eSYinghai Lu } 83995f72d1eSYinghai Lu 84035fd0808STejun Heo /** 84147cec443SMike Rapoport * memblock_setclr_flag - set or clear flag for a memory region 84247cec443SMike Rapoport * @base: base address of the region 84347cec443SMike Rapoport * @size: size of the region 84447cec443SMike Rapoport * @set: set or clear the flag 84547cec443SMike Rapoport * @flag: the flag to udpate 84666b16edfSTang Chen * 8474308ce17STony Luck * This function isolates region [@base, @base + @size), and sets/clears flag 84866b16edfSTang Chen * 84947cec443SMike Rapoport * Return: 0 on success, -errno on failure. 85066b16edfSTang Chen */ 8514308ce17STony Luck static int __init_memblock memblock_setclr_flag(phys_addr_t base, 8524308ce17STony Luck phys_addr_t size, int set, int flag) 85366b16edfSTang Chen { 85466b16edfSTang Chen struct memblock_type *type = &memblock.memory; 85566b16edfSTang Chen int i, ret, start_rgn, end_rgn; 85666b16edfSTang Chen 85766b16edfSTang Chen ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); 85866b16edfSTang Chen if (ret) 85966b16edfSTang Chen return ret; 86066b16edfSTang Chen 86166b16edfSTang Chen for (i = start_rgn; i < end_rgn; i++) 8624308ce17STony Luck if (set) 8634308ce17STony Luck memblock_set_region_flags(&type->regions[i], flag); 8644308ce17STony Luck else 8654308ce17STony Luck memblock_clear_region_flags(&type->regions[i], flag); 86666b16edfSTang Chen 86766b16edfSTang Chen memblock_merge_regions(type); 86866b16edfSTang Chen return 0; 86966b16edfSTang Chen } 87066b16edfSTang Chen 87166b16edfSTang Chen /** 8724308ce17STony Luck * memblock_mark_hotplug - Mark hotpluggable memory with flag MEMBLOCK_HOTPLUG. 8734308ce17STony Luck * @base: the base phys addr of the region 8744308ce17STony Luck * @size: the size of the region 8754308ce17STony Luck * 87647cec443SMike Rapoport * Return: 0 on success, -errno on failure. 8774308ce17STony Luck */ 8784308ce17STony Luck int __init_memblock memblock_mark_hotplug(phys_addr_t base, phys_addr_t size) 8794308ce17STony Luck { 8804308ce17STony Luck return memblock_setclr_flag(base, size, 1, MEMBLOCK_HOTPLUG); 8814308ce17STony Luck } 8824308ce17STony Luck 8834308ce17STony Luck /** 88466b16edfSTang Chen * memblock_clear_hotplug - Clear flag MEMBLOCK_HOTPLUG for a specified region. 88566b16edfSTang Chen * @base: the base phys addr of the region 88666b16edfSTang Chen * @size: the size of the region 88766b16edfSTang Chen * 88847cec443SMike Rapoport * Return: 0 on success, -errno on failure. 88966b16edfSTang Chen */ 89066b16edfSTang Chen int __init_memblock memblock_clear_hotplug(phys_addr_t base, phys_addr_t size) 89166b16edfSTang Chen { 8924308ce17STony Luck return memblock_setclr_flag(base, size, 0, MEMBLOCK_HOTPLUG); 89366b16edfSTang Chen } 89466b16edfSTang Chen 89566b16edfSTang Chen /** 896a3f5bafcSTony Luck * memblock_mark_mirror - Mark mirrored memory with flag MEMBLOCK_MIRROR. 897a3f5bafcSTony Luck * @base: the base phys addr of the region 898a3f5bafcSTony Luck * @size: the size of the region 899a3f5bafcSTony Luck * 90047cec443SMike Rapoport * Return: 0 on success, -errno on failure. 901a3f5bafcSTony Luck */ 902a3f5bafcSTony Luck int __init_memblock memblock_mark_mirror(phys_addr_t base, phys_addr_t size) 903a3f5bafcSTony Luck { 904a3f5bafcSTony Luck system_has_some_mirror = true; 905a3f5bafcSTony Luck 906a3f5bafcSTony Luck return memblock_setclr_flag(base, size, 1, MEMBLOCK_MIRROR); 907a3f5bafcSTony Luck } 908a3f5bafcSTony Luck 909bf3d3cc5SArd Biesheuvel /** 910bf3d3cc5SArd Biesheuvel * memblock_mark_nomap - Mark a memory region with flag MEMBLOCK_NOMAP. 911bf3d3cc5SArd Biesheuvel * @base: the base phys addr of the region 912bf3d3cc5SArd Biesheuvel * @size: the size of the region 913bf3d3cc5SArd Biesheuvel * 91447cec443SMike Rapoport * Return: 0 on success, -errno on failure. 915bf3d3cc5SArd Biesheuvel */ 916bf3d3cc5SArd Biesheuvel int __init_memblock memblock_mark_nomap(phys_addr_t base, phys_addr_t size) 917bf3d3cc5SArd Biesheuvel { 918bf3d3cc5SArd Biesheuvel return memblock_setclr_flag(base, size, 1, MEMBLOCK_NOMAP); 919bf3d3cc5SArd Biesheuvel } 920a3f5bafcSTony Luck 921a3f5bafcSTony Luck /** 9224c546b8aSAKASHI Takahiro * memblock_clear_nomap - Clear flag MEMBLOCK_NOMAP for a specified region. 9234c546b8aSAKASHI Takahiro * @base: the base phys addr of the region 9244c546b8aSAKASHI Takahiro * @size: the size of the region 9254c546b8aSAKASHI Takahiro * 92647cec443SMike Rapoport * Return: 0 on success, -errno on failure. 9274c546b8aSAKASHI Takahiro */ 9284c546b8aSAKASHI Takahiro int __init_memblock memblock_clear_nomap(phys_addr_t base, phys_addr_t size) 9294c546b8aSAKASHI Takahiro { 9304c546b8aSAKASHI Takahiro return memblock_setclr_flag(base, size, 0, MEMBLOCK_NOMAP); 9314c546b8aSAKASHI Takahiro } 9324c546b8aSAKASHI Takahiro 9334c546b8aSAKASHI Takahiro /** 9348e7a7f86SRobin Holt * __next_reserved_mem_region - next function for for_each_reserved_region() 9358e7a7f86SRobin Holt * @idx: pointer to u64 loop variable 9368e7a7f86SRobin Holt * @out_start: ptr to phys_addr_t for start address of the region, can be %NULL 9378e7a7f86SRobin Holt * @out_end: ptr to phys_addr_t for end address of the region, can be %NULL 9388e7a7f86SRobin Holt * 9398e7a7f86SRobin Holt * Iterate over all reserved memory regions. 9408e7a7f86SRobin Holt */ 9418e7a7f86SRobin Holt void __init_memblock __next_reserved_mem_region(u64 *idx, 9428e7a7f86SRobin Holt phys_addr_t *out_start, 9438e7a7f86SRobin Holt phys_addr_t *out_end) 9448e7a7f86SRobin Holt { 945567d117bSAlexander Kuleshov struct memblock_type *type = &memblock.reserved; 9468e7a7f86SRobin Holt 947cd33a76bSRichard Leitner if (*idx < type->cnt) { 948567d117bSAlexander Kuleshov struct memblock_region *r = &type->regions[*idx]; 9498e7a7f86SRobin Holt phys_addr_t base = r->base; 9508e7a7f86SRobin Holt phys_addr_t size = r->size; 9518e7a7f86SRobin Holt 9528e7a7f86SRobin Holt if (out_start) 9538e7a7f86SRobin Holt *out_start = base; 9548e7a7f86SRobin Holt if (out_end) 9558e7a7f86SRobin Holt *out_end = base + size - 1; 9568e7a7f86SRobin Holt 9578e7a7f86SRobin Holt *idx += 1; 9588e7a7f86SRobin Holt return; 9598e7a7f86SRobin Holt } 9608e7a7f86SRobin Holt 9618e7a7f86SRobin Holt /* signal end of iteration */ 9628e7a7f86SRobin Holt *idx = ULLONG_MAX; 9638e7a7f86SRobin Holt } 9648e7a7f86SRobin Holt 9658e7a7f86SRobin Holt /** 966f1af9d3aSPhilipp Hachtmann * __next__mem_range - next function for for_each_free_mem_range() etc. 96735fd0808STejun Heo * @idx: pointer to u64 loop variable 968b1154233SGrygorii Strashko * @nid: node selector, %NUMA_NO_NODE for all nodes 969fc6daaf9STony Luck * @flags: pick from blocks based on memory attributes 970f1af9d3aSPhilipp Hachtmann * @type_a: pointer to memblock_type from where the range is taken 971f1af9d3aSPhilipp Hachtmann * @type_b: pointer to memblock_type which excludes memory from being taken 972dad7557eSWanpeng Li * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL 973dad7557eSWanpeng Li * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL 974dad7557eSWanpeng Li * @out_nid: ptr to int for nid of the range, can be %NULL 97535fd0808STejun Heo * 976f1af9d3aSPhilipp Hachtmann * Find the first area from *@idx which matches @nid, fill the out 97735fd0808STejun Heo * parameters, and update *@idx for the next iteration. The lower 32bit of 978f1af9d3aSPhilipp Hachtmann * *@idx contains index into type_a and the upper 32bit indexes the 979f1af9d3aSPhilipp Hachtmann * areas before each region in type_b. For example, if type_b regions 98035fd0808STejun Heo * look like the following, 98135fd0808STejun Heo * 98235fd0808STejun Heo * 0:[0-16), 1:[32-48), 2:[128-130) 98335fd0808STejun Heo * 98435fd0808STejun Heo * The upper 32bit indexes the following regions. 98535fd0808STejun Heo * 98635fd0808STejun Heo * 0:[0-0), 1:[16-32), 2:[48-128), 3:[130-MAX) 98735fd0808STejun Heo * 98835fd0808STejun Heo * As both region arrays are sorted, the function advances the two indices 98935fd0808STejun Heo * in lockstep and returns each intersection. 99035fd0808STejun Heo */ 991e1720feeSMike Rapoport void __init_memblock __next_mem_range(u64 *idx, int nid, 992e1720feeSMike Rapoport enum memblock_flags flags, 993f1af9d3aSPhilipp Hachtmann struct memblock_type *type_a, 994f1af9d3aSPhilipp Hachtmann struct memblock_type *type_b, 99535fd0808STejun Heo phys_addr_t *out_start, 99635fd0808STejun Heo phys_addr_t *out_end, int *out_nid) 99735fd0808STejun Heo { 998f1af9d3aSPhilipp Hachtmann int idx_a = *idx & 0xffffffff; 999f1af9d3aSPhilipp Hachtmann int idx_b = *idx >> 32; 1000b1154233SGrygorii Strashko 1001f1af9d3aSPhilipp Hachtmann if (WARN_ONCE(nid == MAX_NUMNODES, 1002f1af9d3aSPhilipp Hachtmann "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) 1003560dca27SGrygorii Strashko nid = NUMA_NO_NODE; 100435fd0808STejun Heo 1005f1af9d3aSPhilipp Hachtmann for (; idx_a < type_a->cnt; idx_a++) { 1006f1af9d3aSPhilipp Hachtmann struct memblock_region *m = &type_a->regions[idx_a]; 1007f1af9d3aSPhilipp Hachtmann 100835fd0808STejun Heo phys_addr_t m_start = m->base; 100935fd0808STejun Heo phys_addr_t m_end = m->base + m->size; 1010f1af9d3aSPhilipp Hachtmann int m_nid = memblock_get_region_node(m); 101135fd0808STejun Heo 101235fd0808STejun Heo /* only memory regions are associated with nodes, check it */ 1013f1af9d3aSPhilipp Hachtmann if (nid != NUMA_NO_NODE && nid != m_nid) 101435fd0808STejun Heo continue; 101535fd0808STejun Heo 10160a313a99SXishi Qiu /* skip hotpluggable memory regions if needed */ 10170a313a99SXishi Qiu if (movable_node_is_enabled() && memblock_is_hotpluggable(m)) 10180a313a99SXishi Qiu continue; 10190a313a99SXishi Qiu 1020a3f5bafcSTony Luck /* if we want mirror memory skip non-mirror memory regions */ 1021a3f5bafcSTony Luck if ((flags & MEMBLOCK_MIRROR) && !memblock_is_mirror(m)) 1022a3f5bafcSTony Luck continue; 1023a3f5bafcSTony Luck 1024bf3d3cc5SArd Biesheuvel /* skip nomap memory unless we were asked for it explicitly */ 1025bf3d3cc5SArd Biesheuvel if (!(flags & MEMBLOCK_NOMAP) && memblock_is_nomap(m)) 1026bf3d3cc5SArd Biesheuvel continue; 1027bf3d3cc5SArd Biesheuvel 1028f1af9d3aSPhilipp Hachtmann if (!type_b) { 1029f1af9d3aSPhilipp Hachtmann if (out_start) 1030f1af9d3aSPhilipp Hachtmann *out_start = m_start; 1031f1af9d3aSPhilipp Hachtmann if (out_end) 1032f1af9d3aSPhilipp Hachtmann *out_end = m_end; 1033f1af9d3aSPhilipp Hachtmann if (out_nid) 1034f1af9d3aSPhilipp Hachtmann *out_nid = m_nid; 1035f1af9d3aSPhilipp Hachtmann idx_a++; 1036f1af9d3aSPhilipp Hachtmann *idx = (u32)idx_a | (u64)idx_b << 32; 1037f1af9d3aSPhilipp Hachtmann return; 1038f1af9d3aSPhilipp Hachtmann } 103935fd0808STejun Heo 1040f1af9d3aSPhilipp Hachtmann /* scan areas before each reservation */ 1041f1af9d3aSPhilipp Hachtmann for (; idx_b < type_b->cnt + 1; idx_b++) { 1042f1af9d3aSPhilipp Hachtmann struct memblock_region *r; 1043f1af9d3aSPhilipp Hachtmann phys_addr_t r_start; 1044f1af9d3aSPhilipp Hachtmann phys_addr_t r_end; 1045f1af9d3aSPhilipp Hachtmann 1046f1af9d3aSPhilipp Hachtmann r = &type_b->regions[idx_b]; 1047f1af9d3aSPhilipp Hachtmann r_start = idx_b ? r[-1].base + r[-1].size : 0; 1048f1af9d3aSPhilipp Hachtmann r_end = idx_b < type_b->cnt ? 10491c4bc43dSStefan Agner r->base : PHYS_ADDR_MAX; 1050f1af9d3aSPhilipp Hachtmann 1051f1af9d3aSPhilipp Hachtmann /* 1052f1af9d3aSPhilipp Hachtmann * if idx_b advanced past idx_a, 1053f1af9d3aSPhilipp Hachtmann * break out to advance idx_a 1054f1af9d3aSPhilipp Hachtmann */ 105535fd0808STejun Heo if (r_start >= m_end) 105635fd0808STejun Heo break; 105735fd0808STejun Heo /* if the two regions intersect, we're done */ 105835fd0808STejun Heo if (m_start < r_end) { 105935fd0808STejun Heo if (out_start) 1060f1af9d3aSPhilipp Hachtmann *out_start = 1061f1af9d3aSPhilipp Hachtmann max(m_start, r_start); 106235fd0808STejun Heo if (out_end) 106335fd0808STejun Heo *out_end = min(m_end, r_end); 106435fd0808STejun Heo if (out_nid) 1065f1af9d3aSPhilipp Hachtmann *out_nid = m_nid; 106635fd0808STejun Heo /* 1067f1af9d3aSPhilipp Hachtmann * The region which ends first is 1068f1af9d3aSPhilipp Hachtmann * advanced for the next iteration. 106935fd0808STejun Heo */ 107035fd0808STejun Heo if (m_end <= r_end) 1071f1af9d3aSPhilipp Hachtmann idx_a++; 107235fd0808STejun Heo else 1073f1af9d3aSPhilipp Hachtmann idx_b++; 1074f1af9d3aSPhilipp Hachtmann *idx = (u32)idx_a | (u64)idx_b << 32; 107535fd0808STejun Heo return; 107635fd0808STejun Heo } 107735fd0808STejun Heo } 107835fd0808STejun Heo } 107935fd0808STejun Heo 108035fd0808STejun Heo /* signal end of iteration */ 108135fd0808STejun Heo *idx = ULLONG_MAX; 108235fd0808STejun Heo } 108335fd0808STejun Heo 10847bd0b0f0STejun Heo /** 1085f1af9d3aSPhilipp Hachtmann * __next_mem_range_rev - generic next function for for_each_*_range_rev() 1086f1af9d3aSPhilipp Hachtmann * 10877bd0b0f0STejun Heo * @idx: pointer to u64 loop variable 1088ad5ea8cdSAlexander Kuleshov * @nid: node selector, %NUMA_NO_NODE for all nodes 1089fc6daaf9STony Luck * @flags: pick from blocks based on memory attributes 1090f1af9d3aSPhilipp Hachtmann * @type_a: pointer to memblock_type from where the range is taken 1091f1af9d3aSPhilipp Hachtmann * @type_b: pointer to memblock_type which excludes memory from being taken 1092dad7557eSWanpeng Li * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL 1093dad7557eSWanpeng Li * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL 1094dad7557eSWanpeng Li * @out_nid: ptr to int for nid of the range, can be %NULL 10957bd0b0f0STejun Heo * 109647cec443SMike Rapoport * Finds the next range from type_a which is not marked as unsuitable 109747cec443SMike Rapoport * in type_b. 109847cec443SMike Rapoport * 1099f1af9d3aSPhilipp Hachtmann * Reverse of __next_mem_range(). 11007bd0b0f0STejun Heo */ 1101e1720feeSMike Rapoport void __init_memblock __next_mem_range_rev(u64 *idx, int nid, 1102e1720feeSMike Rapoport enum memblock_flags flags, 1103f1af9d3aSPhilipp Hachtmann struct memblock_type *type_a, 1104f1af9d3aSPhilipp Hachtmann struct memblock_type *type_b, 11057bd0b0f0STejun Heo phys_addr_t *out_start, 11067bd0b0f0STejun Heo phys_addr_t *out_end, int *out_nid) 11077bd0b0f0STejun Heo { 1108f1af9d3aSPhilipp Hachtmann int idx_a = *idx & 0xffffffff; 1109f1af9d3aSPhilipp Hachtmann int idx_b = *idx >> 32; 1110b1154233SGrygorii Strashko 1111560dca27SGrygorii Strashko if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) 1112560dca27SGrygorii Strashko nid = NUMA_NO_NODE; 11137bd0b0f0STejun Heo 11147bd0b0f0STejun Heo if (*idx == (u64)ULLONG_MAX) { 1115f1af9d3aSPhilipp Hachtmann idx_a = type_a->cnt - 1; 1116e47608abSzijun_hu if (type_b != NULL) 1117f1af9d3aSPhilipp Hachtmann idx_b = type_b->cnt; 1118e47608abSzijun_hu else 1119e47608abSzijun_hu idx_b = 0; 11207bd0b0f0STejun Heo } 11217bd0b0f0STejun Heo 1122f1af9d3aSPhilipp Hachtmann for (; idx_a >= 0; idx_a--) { 1123f1af9d3aSPhilipp Hachtmann struct memblock_region *m = &type_a->regions[idx_a]; 1124f1af9d3aSPhilipp Hachtmann 11257bd0b0f0STejun Heo phys_addr_t m_start = m->base; 11267bd0b0f0STejun Heo phys_addr_t m_end = m->base + m->size; 1127f1af9d3aSPhilipp Hachtmann int m_nid = memblock_get_region_node(m); 11287bd0b0f0STejun Heo 11297bd0b0f0STejun Heo /* only memory regions are associated with nodes, check it */ 1130f1af9d3aSPhilipp Hachtmann if (nid != NUMA_NO_NODE && nid != m_nid) 11317bd0b0f0STejun Heo continue; 11327bd0b0f0STejun Heo 113355ac590cSTang Chen /* skip hotpluggable memory regions if needed */ 113455ac590cSTang Chen if (movable_node_is_enabled() && memblock_is_hotpluggable(m)) 113555ac590cSTang Chen continue; 113655ac590cSTang Chen 1137a3f5bafcSTony Luck /* if we want mirror memory skip non-mirror memory regions */ 1138a3f5bafcSTony Luck if ((flags & MEMBLOCK_MIRROR) && !memblock_is_mirror(m)) 1139a3f5bafcSTony Luck continue; 1140a3f5bafcSTony Luck 1141bf3d3cc5SArd Biesheuvel /* skip nomap memory unless we were asked for it explicitly */ 1142bf3d3cc5SArd Biesheuvel if (!(flags & MEMBLOCK_NOMAP) && memblock_is_nomap(m)) 1143bf3d3cc5SArd Biesheuvel continue; 1144bf3d3cc5SArd Biesheuvel 1145f1af9d3aSPhilipp Hachtmann if (!type_b) { 1146f1af9d3aSPhilipp Hachtmann if (out_start) 1147f1af9d3aSPhilipp Hachtmann *out_start = m_start; 1148f1af9d3aSPhilipp Hachtmann if (out_end) 1149f1af9d3aSPhilipp Hachtmann *out_end = m_end; 1150f1af9d3aSPhilipp Hachtmann if (out_nid) 1151f1af9d3aSPhilipp Hachtmann *out_nid = m_nid; 1152fb399b48Szijun_hu idx_a--; 1153f1af9d3aSPhilipp Hachtmann *idx = (u32)idx_a | (u64)idx_b << 32; 1154f1af9d3aSPhilipp Hachtmann return; 1155f1af9d3aSPhilipp Hachtmann } 11567bd0b0f0STejun Heo 1157f1af9d3aSPhilipp Hachtmann /* scan areas before each reservation */ 1158f1af9d3aSPhilipp Hachtmann for (; idx_b >= 0; idx_b--) { 1159f1af9d3aSPhilipp Hachtmann struct memblock_region *r; 1160f1af9d3aSPhilipp Hachtmann phys_addr_t r_start; 1161f1af9d3aSPhilipp Hachtmann phys_addr_t r_end; 1162f1af9d3aSPhilipp Hachtmann 1163f1af9d3aSPhilipp Hachtmann r = &type_b->regions[idx_b]; 1164f1af9d3aSPhilipp Hachtmann r_start = idx_b ? r[-1].base + r[-1].size : 0; 1165f1af9d3aSPhilipp Hachtmann r_end = idx_b < type_b->cnt ? 11661c4bc43dSStefan Agner r->base : PHYS_ADDR_MAX; 1167f1af9d3aSPhilipp Hachtmann /* 1168f1af9d3aSPhilipp Hachtmann * if idx_b advanced past idx_a, 1169f1af9d3aSPhilipp Hachtmann * break out to advance idx_a 1170f1af9d3aSPhilipp Hachtmann */ 1171f1af9d3aSPhilipp Hachtmann 11727bd0b0f0STejun Heo if (r_end <= m_start) 11737bd0b0f0STejun Heo break; 11747bd0b0f0STejun Heo /* if the two regions intersect, we're done */ 11757bd0b0f0STejun Heo if (m_end > r_start) { 11767bd0b0f0STejun Heo if (out_start) 11777bd0b0f0STejun Heo *out_start = max(m_start, r_start); 11787bd0b0f0STejun Heo if (out_end) 11797bd0b0f0STejun Heo *out_end = min(m_end, r_end); 11807bd0b0f0STejun Heo if (out_nid) 1181f1af9d3aSPhilipp Hachtmann *out_nid = m_nid; 11827bd0b0f0STejun Heo if (m_start >= r_start) 1183f1af9d3aSPhilipp Hachtmann idx_a--; 11847bd0b0f0STejun Heo else 1185f1af9d3aSPhilipp Hachtmann idx_b--; 1186f1af9d3aSPhilipp Hachtmann *idx = (u32)idx_a | (u64)idx_b << 32; 11877bd0b0f0STejun Heo return; 11887bd0b0f0STejun Heo } 11897bd0b0f0STejun Heo } 11907bd0b0f0STejun Heo } 1191f1af9d3aSPhilipp Hachtmann /* signal end of iteration */ 11927bd0b0f0STejun Heo *idx = ULLONG_MAX; 11937bd0b0f0STejun Heo } 11947bd0b0f0STejun Heo 11957c0caeb8STejun Heo #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP 11967c0caeb8STejun Heo /* 119745e79815SChen Chang * Common iterator interface used to define for_each_mem_pfn_range(). 11987c0caeb8STejun Heo */ 11997c0caeb8STejun Heo void __init_memblock __next_mem_pfn_range(int *idx, int nid, 12007c0caeb8STejun Heo unsigned long *out_start_pfn, 12017c0caeb8STejun Heo unsigned long *out_end_pfn, int *out_nid) 12027c0caeb8STejun Heo { 12037c0caeb8STejun Heo struct memblock_type *type = &memblock.memory; 12047c0caeb8STejun Heo struct memblock_region *r; 12057c0caeb8STejun Heo 12067c0caeb8STejun Heo while (++*idx < type->cnt) { 12077c0caeb8STejun Heo r = &type->regions[*idx]; 12087c0caeb8STejun Heo 12097c0caeb8STejun Heo if (PFN_UP(r->base) >= PFN_DOWN(r->base + r->size)) 12107c0caeb8STejun Heo continue; 12117c0caeb8STejun Heo if (nid == MAX_NUMNODES || nid == r->nid) 12127c0caeb8STejun Heo break; 12137c0caeb8STejun Heo } 12147c0caeb8STejun Heo if (*idx >= type->cnt) { 12157c0caeb8STejun Heo *idx = -1; 12167c0caeb8STejun Heo return; 12177c0caeb8STejun Heo } 12187c0caeb8STejun Heo 12197c0caeb8STejun Heo if (out_start_pfn) 12207c0caeb8STejun Heo *out_start_pfn = PFN_UP(r->base); 12217c0caeb8STejun Heo if (out_end_pfn) 12227c0caeb8STejun Heo *out_end_pfn = PFN_DOWN(r->base + r->size); 12237c0caeb8STejun Heo if (out_nid) 12247c0caeb8STejun Heo *out_nid = r->nid; 12257c0caeb8STejun Heo } 12267c0caeb8STejun Heo 12277c0caeb8STejun Heo /** 12287c0caeb8STejun Heo * memblock_set_node - set node ID on memblock regions 12297c0caeb8STejun Heo * @base: base of area to set node ID for 12307c0caeb8STejun Heo * @size: size of area to set node ID for 1231e7e8de59STang Chen * @type: memblock type to set node ID for 12327c0caeb8STejun Heo * @nid: node ID to set 12337c0caeb8STejun Heo * 1234e7e8de59STang Chen * Set the nid of memblock @type regions in [@base, @base + @size) to @nid. 12357c0caeb8STejun Heo * Regions which cross the area boundaries are split as necessary. 12367c0caeb8STejun Heo * 123747cec443SMike Rapoport * Return: 12387c0caeb8STejun Heo * 0 on success, -errno on failure. 12397c0caeb8STejun Heo */ 12407c0caeb8STejun Heo int __init_memblock memblock_set_node(phys_addr_t base, phys_addr_t size, 1241e7e8de59STang Chen struct memblock_type *type, int nid) 12427c0caeb8STejun Heo { 12436a9ceb31STejun Heo int start_rgn, end_rgn; 12446a9ceb31STejun Heo int i, ret; 12457c0caeb8STejun Heo 12466a9ceb31STejun Heo ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); 12476a9ceb31STejun Heo if (ret) 12486a9ceb31STejun Heo return ret; 12497c0caeb8STejun Heo 12506a9ceb31STejun Heo for (i = start_rgn; i < end_rgn; i++) 1251e9d24ad3SWanpeng Li memblock_set_region_node(&type->regions[i], nid); 12527c0caeb8STejun Heo 12537c0caeb8STejun Heo memblock_merge_regions(type); 12547c0caeb8STejun Heo return 0; 12557c0caeb8STejun Heo } 12567c0caeb8STejun Heo #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */ 12577c0caeb8STejun Heo 12582bfc2862SAkinobu Mita static phys_addr_t __init memblock_alloc_range_nid(phys_addr_t size, 12592bfc2862SAkinobu Mita phys_addr_t align, phys_addr_t start, 1260e1720feeSMike Rapoport phys_addr_t end, int nid, 1261e1720feeSMike Rapoport enum memblock_flags flags) 126295f72d1eSYinghai Lu { 12636ed311b2SBenjamin Herrenschmidt phys_addr_t found; 126495f72d1eSYinghai Lu 12652f770806SMike Rapoport if (!align) { 12662f770806SMike Rapoport /* Can't use WARNs this early in boot on powerpc */ 12672f770806SMike Rapoport dump_stack(); 12682f770806SMike Rapoport align = SMP_CACHE_BYTES; 12692f770806SMike Rapoport } 12702f770806SMike Rapoport 1271fc6daaf9STony Luck found = memblock_find_in_range_node(size, align, start, end, nid, 1272fc6daaf9STony Luck flags); 1273aedf95eaSCatalin Marinas if (found && !memblock_reserve(found, size)) { 1274aedf95eaSCatalin Marinas /* 1275aedf95eaSCatalin Marinas * The min_count is set to 0 so that memblock allocations are 1276aedf95eaSCatalin Marinas * never reported as leaks. 1277aedf95eaSCatalin Marinas */ 12789099daedSCatalin Marinas kmemleak_alloc_phys(found, size, 0, 0); 12796ed311b2SBenjamin Herrenschmidt return found; 1280aedf95eaSCatalin Marinas } 12816ed311b2SBenjamin Herrenschmidt return 0; 128295f72d1eSYinghai Lu } 128395f72d1eSYinghai Lu 1284*8a770c2aSMike Rapoport phys_addr_t __init memblock_phys_alloc_range(phys_addr_t size, 1285*8a770c2aSMike Rapoport phys_addr_t align, 1286*8a770c2aSMike Rapoport phys_addr_t start, 1287*8a770c2aSMike Rapoport phys_addr_t end) 12882bfc2862SAkinobu Mita { 1289fc6daaf9STony Luck return memblock_alloc_range_nid(size, align, start, end, NUMA_NO_NODE, 1290*8a770c2aSMike Rapoport MEMBLOCK_NONE); 12912bfc2862SAkinobu Mita } 12922bfc2862SAkinobu Mita 12939a8dd708SMike Rapoport phys_addr_t __init memblock_phys_alloc_nid(phys_addr_t size, phys_addr_t align, int nid) 12947bd0b0f0STejun Heo { 1295e1720feeSMike Rapoport enum memblock_flags flags = choose_memblock_flags(); 1296a3f5bafcSTony Luck phys_addr_t ret; 1297a3f5bafcSTony Luck 1298a3f5bafcSTony Luck again: 129953d818d2SMike Rapoport ret = memblock_alloc_range_nid(size, align, 0, 130053d818d2SMike Rapoport MEMBLOCK_ALLOC_ACCESSIBLE, nid, flags); 1301a3f5bafcSTony Luck 1302a3f5bafcSTony Luck if (!ret && (flags & MEMBLOCK_MIRROR)) { 1303a3f5bafcSTony Luck flags &= ~MEMBLOCK_MIRROR; 1304a3f5bafcSTony Luck goto again; 1305a3f5bafcSTony Luck } 1306a3f5bafcSTony Luck return ret; 13077bd0b0f0STejun Heo } 13087bd0b0f0STejun Heo 13097bd0b0f0STejun Heo phys_addr_t __init __memblock_alloc_base(phys_addr_t size, phys_addr_t align, phys_addr_t max_addr) 13107bd0b0f0STejun Heo { 131153d818d2SMike Rapoport return memblock_alloc_range_nid(size, align, 0, max_addr, NUMA_NO_NODE, 1312fc6daaf9STony Luck MEMBLOCK_NONE); 13137bd0b0f0STejun Heo } 13147bd0b0f0STejun Heo 13156ed311b2SBenjamin Herrenschmidt phys_addr_t __init memblock_alloc_base(phys_addr_t size, phys_addr_t align, phys_addr_t max_addr) 131695f72d1eSYinghai Lu { 13176ed311b2SBenjamin Herrenschmidt phys_addr_t alloc; 13186ed311b2SBenjamin Herrenschmidt 13196ed311b2SBenjamin Herrenschmidt alloc = __memblock_alloc_base(size, align, max_addr); 13206ed311b2SBenjamin Herrenschmidt 13216ed311b2SBenjamin Herrenschmidt if (alloc == 0) 13225d63f81cSMiles Chen panic("ERROR: Failed to allocate %pa bytes below %pa.\n", 13235d63f81cSMiles Chen &size, &max_addr); 13246ed311b2SBenjamin Herrenschmidt 13256ed311b2SBenjamin Herrenschmidt return alloc; 132695f72d1eSYinghai Lu } 132795f72d1eSYinghai Lu 13289a8dd708SMike Rapoport phys_addr_t __init memblock_phys_alloc(phys_addr_t size, phys_addr_t align) 132995f72d1eSYinghai Lu { 13306ed311b2SBenjamin Herrenschmidt return memblock_alloc_base(size, align, MEMBLOCK_ALLOC_ACCESSIBLE); 133195f72d1eSYinghai Lu } 133295f72d1eSYinghai Lu 13339a8dd708SMike Rapoport phys_addr_t __init memblock_phys_alloc_try_nid(phys_addr_t size, phys_addr_t align, int nid) 13349d1e2492SBenjamin Herrenschmidt { 13359a8dd708SMike Rapoport phys_addr_t res = memblock_phys_alloc_nid(size, align, nid); 13369d1e2492SBenjamin Herrenschmidt 13379d1e2492SBenjamin Herrenschmidt if (res) 13389d1e2492SBenjamin Herrenschmidt return res; 133915fb0972STejun Heo return memblock_alloc_base(size, align, MEMBLOCK_ALLOC_ACCESSIBLE); 134095f72d1eSYinghai Lu } 134195f72d1eSYinghai Lu 134226f09e9bSSantosh Shilimkar /** 1343eb31d559SMike Rapoport * memblock_alloc_internal - allocate boot memory block 134426f09e9bSSantosh Shilimkar * @size: size of memory block to be allocated in bytes 134526f09e9bSSantosh Shilimkar * @align: alignment of the region and block's size 134626f09e9bSSantosh Shilimkar * @min_addr: the lower bound of the memory region to allocate (phys address) 134726f09e9bSSantosh Shilimkar * @max_addr: the upper bound of the memory region to allocate (phys address) 134826f09e9bSSantosh Shilimkar * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 134926f09e9bSSantosh Shilimkar * 135026f09e9bSSantosh Shilimkar * The @min_addr limit is dropped if it can not be satisfied and the allocation 135126f09e9bSSantosh Shilimkar * will fall back to memory below @min_addr. Also, allocation may fall back 135226f09e9bSSantosh Shilimkar * to any node in the system if the specified node can not 135326f09e9bSSantosh Shilimkar * hold the requested memory. 135426f09e9bSSantosh Shilimkar * 135526f09e9bSSantosh Shilimkar * The allocation is performed from memory region limited by 135697ad1087SMike Rapoport * memblock.current_limit if @max_addr == %MEMBLOCK_ALLOC_ACCESSIBLE. 135726f09e9bSSantosh Shilimkar * 135826f09e9bSSantosh Shilimkar * The phys address of allocated boot memory block is converted to virtual and 135926f09e9bSSantosh Shilimkar * allocated memory is reset to 0. 136026f09e9bSSantosh Shilimkar * 136126f09e9bSSantosh Shilimkar * In addition, function sets the min_count to 0 using kmemleak_alloc for 136226f09e9bSSantosh Shilimkar * allocated boot memory block, so that it is never reported as leaks. 136326f09e9bSSantosh Shilimkar * 136447cec443SMike Rapoport * Return: 136526f09e9bSSantosh Shilimkar * Virtual address of allocated memory block on success, NULL on failure. 136626f09e9bSSantosh Shilimkar */ 1367eb31d559SMike Rapoport static void * __init memblock_alloc_internal( 136826f09e9bSSantosh Shilimkar phys_addr_t size, phys_addr_t align, 136926f09e9bSSantosh Shilimkar phys_addr_t min_addr, phys_addr_t max_addr, 137026f09e9bSSantosh Shilimkar int nid) 137126f09e9bSSantosh Shilimkar { 137226f09e9bSSantosh Shilimkar phys_addr_t alloc; 137326f09e9bSSantosh Shilimkar void *ptr; 1374e1720feeSMike Rapoport enum memblock_flags flags = choose_memblock_flags(); 137526f09e9bSSantosh Shilimkar 1376560dca27SGrygorii Strashko if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) 1377560dca27SGrygorii Strashko nid = NUMA_NO_NODE; 137826f09e9bSSantosh Shilimkar 137926f09e9bSSantosh Shilimkar /* 138026f09e9bSSantosh Shilimkar * Detect any accidental use of these APIs after slab is ready, as at 138126f09e9bSSantosh Shilimkar * this moment memblock may be deinitialized already and its 1382c6ffc5caSMike Rapoport * internal data may be destroyed (after execution of memblock_free_all) 138326f09e9bSSantosh Shilimkar */ 138426f09e9bSSantosh Shilimkar if (WARN_ON_ONCE(slab_is_available())) 138526f09e9bSSantosh Shilimkar return kzalloc_node(size, GFP_NOWAIT, nid); 138626f09e9bSSantosh Shilimkar 13872f770806SMike Rapoport if (!align) { 13882f770806SMike Rapoport dump_stack(); 13892f770806SMike Rapoport align = SMP_CACHE_BYTES; 13902f770806SMike Rapoport } 13912f770806SMike Rapoport 1392f544e14fSYinghai Lu if (max_addr > memblock.current_limit) 1393f544e14fSYinghai Lu max_addr = memblock.current_limit; 139426f09e9bSSantosh Shilimkar again: 139526f09e9bSSantosh Shilimkar alloc = memblock_find_in_range_node(size, align, min_addr, max_addr, 1396a3f5bafcSTony Luck nid, flags); 13977d41c03eSWei Yang if (alloc && !memblock_reserve(alloc, size)) 139826f09e9bSSantosh Shilimkar goto done; 139926f09e9bSSantosh Shilimkar 140026f09e9bSSantosh Shilimkar if (nid != NUMA_NO_NODE) { 140126f09e9bSSantosh Shilimkar alloc = memblock_find_in_range_node(size, align, min_addr, 1402fc6daaf9STony Luck max_addr, NUMA_NO_NODE, 1403a3f5bafcSTony Luck flags); 14047d41c03eSWei Yang if (alloc && !memblock_reserve(alloc, size)) 140526f09e9bSSantosh Shilimkar goto done; 140626f09e9bSSantosh Shilimkar } 140726f09e9bSSantosh Shilimkar 140826f09e9bSSantosh Shilimkar if (min_addr) { 140926f09e9bSSantosh Shilimkar min_addr = 0; 141026f09e9bSSantosh Shilimkar goto again; 141126f09e9bSSantosh Shilimkar } 141226f09e9bSSantosh Shilimkar 1413a3f5bafcSTony Luck if (flags & MEMBLOCK_MIRROR) { 1414a3f5bafcSTony Luck flags &= ~MEMBLOCK_MIRROR; 1415a3f5bafcSTony Luck pr_warn("Could not allocate %pap bytes of mirrored memory\n", 1416a3f5bafcSTony Luck &size); 1417a3f5bafcSTony Luck goto again; 1418a3f5bafcSTony Luck } 1419a3f5bafcSTony Luck 1420a3f5bafcSTony Luck return NULL; 142126f09e9bSSantosh Shilimkar done: 142226f09e9bSSantosh Shilimkar ptr = phys_to_virt(alloc); 142326f09e9bSSantosh Shilimkar 1424fed84c78SQian Cai /* Skip kmemleak for kasan_init() due to high volume. */ 1425fed84c78SQian Cai if (max_addr != MEMBLOCK_ALLOC_KASAN) 142626f09e9bSSantosh Shilimkar /* 1427fed84c78SQian Cai * The min_count is set to 0 so that bootmem allocated 1428fed84c78SQian Cai * blocks are never reported as leaks. This is because many 1429fed84c78SQian Cai * of these blocks are only referred via the physical 1430fed84c78SQian Cai * address which is not looked up by kmemleak. 143126f09e9bSSantosh Shilimkar */ 143226f09e9bSSantosh Shilimkar kmemleak_alloc(ptr, size, 0, 0); 143326f09e9bSSantosh Shilimkar 143426f09e9bSSantosh Shilimkar return ptr; 143526f09e9bSSantosh Shilimkar } 143626f09e9bSSantosh Shilimkar 143726f09e9bSSantosh Shilimkar /** 1438eb31d559SMike Rapoport * memblock_alloc_try_nid_raw - allocate boot memory block without zeroing 1439ea1f5f37SPavel Tatashin * memory and without panicking 1440ea1f5f37SPavel Tatashin * @size: size of memory block to be allocated in bytes 1441ea1f5f37SPavel Tatashin * @align: alignment of the region and block's size 1442ea1f5f37SPavel Tatashin * @min_addr: the lower bound of the memory region from where the allocation 1443ea1f5f37SPavel Tatashin * is preferred (phys address) 1444ea1f5f37SPavel Tatashin * @max_addr: the upper bound of the memory region from where the allocation 144597ad1087SMike Rapoport * is preferred (phys address), or %MEMBLOCK_ALLOC_ACCESSIBLE to 1446ea1f5f37SPavel Tatashin * allocate only from memory limited by memblock.current_limit value 1447ea1f5f37SPavel Tatashin * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 1448ea1f5f37SPavel Tatashin * 1449ea1f5f37SPavel Tatashin * Public function, provides additional debug information (including caller 1450ea1f5f37SPavel Tatashin * info), if enabled. Does not zero allocated memory, does not panic if request 1451ea1f5f37SPavel Tatashin * cannot be satisfied. 1452ea1f5f37SPavel Tatashin * 145347cec443SMike Rapoport * Return: 1454ea1f5f37SPavel Tatashin * Virtual address of allocated memory block on success, NULL on failure. 1455ea1f5f37SPavel Tatashin */ 1456eb31d559SMike Rapoport void * __init memblock_alloc_try_nid_raw( 1457ea1f5f37SPavel Tatashin phys_addr_t size, phys_addr_t align, 1458ea1f5f37SPavel Tatashin phys_addr_t min_addr, phys_addr_t max_addr, 1459ea1f5f37SPavel Tatashin int nid) 1460ea1f5f37SPavel Tatashin { 1461ea1f5f37SPavel Tatashin void *ptr; 1462ea1f5f37SPavel Tatashin 1463a36aab89SMike Rapoport memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=%pa max_addr=%pa %pF\n", 1464a36aab89SMike Rapoport __func__, (u64)size, (u64)align, nid, &min_addr, 1465a36aab89SMike Rapoport &max_addr, (void *)_RET_IP_); 1466ea1f5f37SPavel Tatashin 1467eb31d559SMike Rapoport ptr = memblock_alloc_internal(size, align, 1468ea1f5f37SPavel Tatashin min_addr, max_addr, nid); 1469ea1f5f37SPavel Tatashin if (ptr && size > 0) 1470f682a97aSAlexander Duyck page_init_poison(ptr, size); 1471f682a97aSAlexander Duyck 1472ea1f5f37SPavel Tatashin return ptr; 1473ea1f5f37SPavel Tatashin } 1474ea1f5f37SPavel Tatashin 1475ea1f5f37SPavel Tatashin /** 1476eb31d559SMike Rapoport * memblock_alloc_try_nid_nopanic - allocate boot memory block 147726f09e9bSSantosh Shilimkar * @size: size of memory block to be allocated in bytes 147826f09e9bSSantosh Shilimkar * @align: alignment of the region and block's size 147926f09e9bSSantosh Shilimkar * @min_addr: the lower bound of the memory region from where the allocation 148026f09e9bSSantosh Shilimkar * is preferred (phys address) 148126f09e9bSSantosh Shilimkar * @max_addr: the upper bound of the memory region from where the allocation 148297ad1087SMike Rapoport * is preferred (phys address), or %MEMBLOCK_ALLOC_ACCESSIBLE to 148326f09e9bSSantosh Shilimkar * allocate only from memory limited by memblock.current_limit value 148426f09e9bSSantosh Shilimkar * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 148526f09e9bSSantosh Shilimkar * 1486ea1f5f37SPavel Tatashin * Public function, provides additional debug information (including caller 1487ea1f5f37SPavel Tatashin * info), if enabled. This function zeroes the allocated memory. 148826f09e9bSSantosh Shilimkar * 148947cec443SMike Rapoport * Return: 149026f09e9bSSantosh Shilimkar * Virtual address of allocated memory block on success, NULL on failure. 149126f09e9bSSantosh Shilimkar */ 1492eb31d559SMike Rapoport void * __init memblock_alloc_try_nid_nopanic( 149326f09e9bSSantosh Shilimkar phys_addr_t size, phys_addr_t align, 149426f09e9bSSantosh Shilimkar phys_addr_t min_addr, phys_addr_t max_addr, 149526f09e9bSSantosh Shilimkar int nid) 149626f09e9bSSantosh Shilimkar { 1497ea1f5f37SPavel Tatashin void *ptr; 1498ea1f5f37SPavel Tatashin 1499a36aab89SMike Rapoport memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=%pa max_addr=%pa %pF\n", 1500a36aab89SMike Rapoport __func__, (u64)size, (u64)align, nid, &min_addr, 1501a36aab89SMike Rapoport &max_addr, (void *)_RET_IP_); 1502ea1f5f37SPavel Tatashin 1503eb31d559SMike Rapoport ptr = memblock_alloc_internal(size, align, 1504ea1f5f37SPavel Tatashin min_addr, max_addr, nid); 1505ea1f5f37SPavel Tatashin if (ptr) 1506ea1f5f37SPavel Tatashin memset(ptr, 0, size); 1507ea1f5f37SPavel Tatashin return ptr; 150826f09e9bSSantosh Shilimkar } 150926f09e9bSSantosh Shilimkar 151026f09e9bSSantosh Shilimkar /** 1511eb31d559SMike Rapoport * memblock_alloc_try_nid - allocate boot memory block with panicking 151226f09e9bSSantosh Shilimkar * @size: size of memory block to be allocated in bytes 151326f09e9bSSantosh Shilimkar * @align: alignment of the region and block's size 151426f09e9bSSantosh Shilimkar * @min_addr: the lower bound of the memory region from where the allocation 151526f09e9bSSantosh Shilimkar * is preferred (phys address) 151626f09e9bSSantosh Shilimkar * @max_addr: the upper bound of the memory region from where the allocation 151797ad1087SMike Rapoport * is preferred (phys address), or %MEMBLOCK_ALLOC_ACCESSIBLE to 151826f09e9bSSantosh Shilimkar * allocate only from memory limited by memblock.current_limit value 151926f09e9bSSantosh Shilimkar * @nid: nid of the free area to find, %NUMA_NO_NODE for any node 152026f09e9bSSantosh Shilimkar * 1521eb31d559SMike Rapoport * Public panicking version of memblock_alloc_try_nid_nopanic() 152226f09e9bSSantosh Shilimkar * which provides debug information (including caller info), if enabled, 152326f09e9bSSantosh Shilimkar * and panics if the request can not be satisfied. 152426f09e9bSSantosh Shilimkar * 152547cec443SMike Rapoport * Return: 152626f09e9bSSantosh Shilimkar * Virtual address of allocated memory block on success, NULL on failure. 152726f09e9bSSantosh Shilimkar */ 1528eb31d559SMike Rapoport void * __init memblock_alloc_try_nid( 152926f09e9bSSantosh Shilimkar phys_addr_t size, phys_addr_t align, 153026f09e9bSSantosh Shilimkar phys_addr_t min_addr, phys_addr_t max_addr, 153126f09e9bSSantosh Shilimkar int nid) 153226f09e9bSSantosh Shilimkar { 153326f09e9bSSantosh Shilimkar void *ptr; 153426f09e9bSSantosh Shilimkar 1535a36aab89SMike Rapoport memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=%pa max_addr=%pa %pF\n", 1536a36aab89SMike Rapoport __func__, (u64)size, (u64)align, nid, &min_addr, 1537a36aab89SMike Rapoport &max_addr, (void *)_RET_IP_); 1538eb31d559SMike Rapoport ptr = memblock_alloc_internal(size, align, 153926f09e9bSSantosh Shilimkar min_addr, max_addr, nid); 1540ea1f5f37SPavel Tatashin if (ptr) { 1541ea1f5f37SPavel Tatashin memset(ptr, 0, size); 154226f09e9bSSantosh Shilimkar return ptr; 1543ea1f5f37SPavel Tatashin } 154426f09e9bSSantosh Shilimkar 1545a36aab89SMike Rapoport panic("%s: Failed to allocate %llu bytes align=0x%llx nid=%d from=%pa max_addr=%pa\n", 1546a36aab89SMike Rapoport __func__, (u64)size, (u64)align, nid, &min_addr, &max_addr); 154726f09e9bSSantosh Shilimkar return NULL; 154826f09e9bSSantosh Shilimkar } 154926f09e9bSSantosh Shilimkar 155026f09e9bSSantosh Shilimkar /** 155126f09e9bSSantosh Shilimkar * __memblock_free_late - free bootmem block pages directly to buddy allocator 155248a833ccSMike Rapoport * @base: phys starting address of the boot memory block 155326f09e9bSSantosh Shilimkar * @size: size of the boot memory block in bytes 155426f09e9bSSantosh Shilimkar * 155526f09e9bSSantosh Shilimkar * This is only useful when the bootmem allocator has already been torn 155626f09e9bSSantosh Shilimkar * down, but we are still initializing the system. Pages are released directly 155726f09e9bSSantosh Shilimkar * to the buddy allocator, no bootmem metadata is updated because it is gone. 155826f09e9bSSantosh Shilimkar */ 155926f09e9bSSantosh Shilimkar void __init __memblock_free_late(phys_addr_t base, phys_addr_t size) 156026f09e9bSSantosh Shilimkar { 1561a36aab89SMike Rapoport phys_addr_t cursor, end; 156226f09e9bSSantosh Shilimkar 1563a36aab89SMike Rapoport end = base + size - 1; 1564a36aab89SMike Rapoport memblock_dbg("%s: [%pa-%pa] %pF\n", 1565a36aab89SMike Rapoport __func__, &base, &end, (void *)_RET_IP_); 15669099daedSCatalin Marinas kmemleak_free_part_phys(base, size); 156726f09e9bSSantosh Shilimkar cursor = PFN_UP(base); 156826f09e9bSSantosh Shilimkar end = PFN_DOWN(base + size); 156926f09e9bSSantosh Shilimkar 157026f09e9bSSantosh Shilimkar for (; cursor < end; cursor++) { 15717c2ee349SMike Rapoport memblock_free_pages(pfn_to_page(cursor), cursor, 0); 1572ca79b0c2SArun KS totalram_pages_inc(); 157326f09e9bSSantosh Shilimkar } 157426f09e9bSSantosh Shilimkar } 15759d1e2492SBenjamin Herrenschmidt 15769d1e2492SBenjamin Herrenschmidt /* 15779d1e2492SBenjamin Herrenschmidt * Remaining API functions 15789d1e2492SBenjamin Herrenschmidt */ 15799d1e2492SBenjamin Herrenschmidt 15801f1ffb8aSDavid Gibson phys_addr_t __init_memblock memblock_phys_mem_size(void) 158195f72d1eSYinghai Lu { 15821440c4e2STejun Heo return memblock.memory.total_size; 158395f72d1eSYinghai Lu } 158495f72d1eSYinghai Lu 15858907de5dSSrikar Dronamraju phys_addr_t __init_memblock memblock_reserved_size(void) 15868907de5dSSrikar Dronamraju { 15878907de5dSSrikar Dronamraju return memblock.reserved.total_size; 15888907de5dSSrikar Dronamraju } 15898907de5dSSrikar Dronamraju 1590595ad9afSYinghai Lu phys_addr_t __init memblock_mem_size(unsigned long limit_pfn) 1591595ad9afSYinghai Lu { 1592595ad9afSYinghai Lu unsigned long pages = 0; 1593595ad9afSYinghai Lu struct memblock_region *r; 1594595ad9afSYinghai Lu unsigned long start_pfn, end_pfn; 1595595ad9afSYinghai Lu 1596595ad9afSYinghai Lu for_each_memblock(memory, r) { 1597595ad9afSYinghai Lu start_pfn = memblock_region_memory_base_pfn(r); 1598595ad9afSYinghai Lu end_pfn = memblock_region_memory_end_pfn(r); 1599595ad9afSYinghai Lu start_pfn = min_t(unsigned long, start_pfn, limit_pfn); 1600595ad9afSYinghai Lu end_pfn = min_t(unsigned long, end_pfn, limit_pfn); 1601595ad9afSYinghai Lu pages += end_pfn - start_pfn; 1602595ad9afSYinghai Lu } 1603595ad9afSYinghai Lu 160416763230SFabian Frederick return PFN_PHYS(pages); 1605595ad9afSYinghai Lu } 1606595ad9afSYinghai Lu 16070a93ebefSSam Ravnborg /* lowest address */ 16080a93ebefSSam Ravnborg phys_addr_t __init_memblock memblock_start_of_DRAM(void) 16090a93ebefSSam Ravnborg { 16100a93ebefSSam Ravnborg return memblock.memory.regions[0].base; 16110a93ebefSSam Ravnborg } 16120a93ebefSSam Ravnborg 161310d06439SYinghai Lu phys_addr_t __init_memblock memblock_end_of_DRAM(void) 161495f72d1eSYinghai Lu { 161595f72d1eSYinghai Lu int idx = memblock.memory.cnt - 1; 161695f72d1eSYinghai Lu 1617e3239ff9SBenjamin Herrenschmidt return (memblock.memory.regions[idx].base + memblock.memory.regions[idx].size); 161895f72d1eSYinghai Lu } 161995f72d1eSYinghai Lu 1620a571d4ebSDennis Chen static phys_addr_t __init_memblock __find_max_addr(phys_addr_t limit) 162195f72d1eSYinghai Lu { 16221c4bc43dSStefan Agner phys_addr_t max_addr = PHYS_ADDR_MAX; 1623136199f0SEmil Medve struct memblock_region *r; 162495f72d1eSYinghai Lu 1625a571d4ebSDennis Chen /* 1626a571d4ebSDennis Chen * translate the memory @limit size into the max address within one of 1627a571d4ebSDennis Chen * the memory memblock regions, if the @limit exceeds the total size 16281c4bc43dSStefan Agner * of those regions, max_addr will keep original value PHYS_ADDR_MAX 1629a571d4ebSDennis Chen */ 1630136199f0SEmil Medve for_each_memblock(memory, r) { 1631c0ce8fefSTejun Heo if (limit <= r->size) { 1632c0ce8fefSTejun Heo max_addr = r->base + limit; 163395f72d1eSYinghai Lu break; 163495f72d1eSYinghai Lu } 1635c0ce8fefSTejun Heo limit -= r->size; 163695f72d1eSYinghai Lu } 1637c0ce8fefSTejun Heo 1638a571d4ebSDennis Chen return max_addr; 1639a571d4ebSDennis Chen } 1640a571d4ebSDennis Chen 1641a571d4ebSDennis Chen void __init memblock_enforce_memory_limit(phys_addr_t limit) 1642a571d4ebSDennis Chen { 16431c4bc43dSStefan Agner phys_addr_t max_addr = PHYS_ADDR_MAX; 1644a571d4ebSDennis Chen 1645a571d4ebSDennis Chen if (!limit) 1646a571d4ebSDennis Chen return; 1647a571d4ebSDennis Chen 1648a571d4ebSDennis Chen max_addr = __find_max_addr(limit); 1649a571d4ebSDennis Chen 1650a571d4ebSDennis Chen /* @limit exceeds the total size of the memory, do nothing */ 16511c4bc43dSStefan Agner if (max_addr == PHYS_ADDR_MAX) 1652a571d4ebSDennis Chen return; 1653a571d4ebSDennis Chen 1654c0ce8fefSTejun Heo /* truncate both memory and reserved regions */ 1655f1af9d3aSPhilipp Hachtmann memblock_remove_range(&memblock.memory, max_addr, 16561c4bc43dSStefan Agner PHYS_ADDR_MAX); 1657f1af9d3aSPhilipp Hachtmann memblock_remove_range(&memblock.reserved, max_addr, 16581c4bc43dSStefan Agner PHYS_ADDR_MAX); 165995f72d1eSYinghai Lu } 166095f72d1eSYinghai Lu 1661c9ca9b4eSAKASHI Takahiro void __init memblock_cap_memory_range(phys_addr_t base, phys_addr_t size) 1662c9ca9b4eSAKASHI Takahiro { 1663c9ca9b4eSAKASHI Takahiro int start_rgn, end_rgn; 1664c9ca9b4eSAKASHI Takahiro int i, ret; 1665c9ca9b4eSAKASHI Takahiro 1666c9ca9b4eSAKASHI Takahiro if (!size) 1667c9ca9b4eSAKASHI Takahiro return; 1668c9ca9b4eSAKASHI Takahiro 1669c9ca9b4eSAKASHI Takahiro ret = memblock_isolate_range(&memblock.memory, base, size, 1670c9ca9b4eSAKASHI Takahiro &start_rgn, &end_rgn); 1671c9ca9b4eSAKASHI Takahiro if (ret) 1672c9ca9b4eSAKASHI Takahiro return; 1673c9ca9b4eSAKASHI Takahiro 1674c9ca9b4eSAKASHI Takahiro /* remove all the MAP regions */ 1675c9ca9b4eSAKASHI Takahiro for (i = memblock.memory.cnt - 1; i >= end_rgn; i--) 1676c9ca9b4eSAKASHI Takahiro if (!memblock_is_nomap(&memblock.memory.regions[i])) 1677c9ca9b4eSAKASHI Takahiro memblock_remove_region(&memblock.memory, i); 1678c9ca9b4eSAKASHI Takahiro 1679c9ca9b4eSAKASHI Takahiro for (i = start_rgn - 1; i >= 0; i--) 1680c9ca9b4eSAKASHI Takahiro if (!memblock_is_nomap(&memblock.memory.regions[i])) 1681c9ca9b4eSAKASHI Takahiro memblock_remove_region(&memblock.memory, i); 1682c9ca9b4eSAKASHI Takahiro 1683c9ca9b4eSAKASHI Takahiro /* truncate the reserved regions */ 1684c9ca9b4eSAKASHI Takahiro memblock_remove_range(&memblock.reserved, 0, base); 1685c9ca9b4eSAKASHI Takahiro memblock_remove_range(&memblock.reserved, 16861c4bc43dSStefan Agner base + size, PHYS_ADDR_MAX); 1687c9ca9b4eSAKASHI Takahiro } 1688c9ca9b4eSAKASHI Takahiro 1689a571d4ebSDennis Chen void __init memblock_mem_limit_remove_map(phys_addr_t limit) 1690a571d4ebSDennis Chen { 1691a571d4ebSDennis Chen phys_addr_t max_addr; 1692a571d4ebSDennis Chen 1693a571d4ebSDennis Chen if (!limit) 1694a571d4ebSDennis Chen return; 1695a571d4ebSDennis Chen 1696a571d4ebSDennis Chen max_addr = __find_max_addr(limit); 1697a571d4ebSDennis Chen 1698a571d4ebSDennis Chen /* @limit exceeds the total size of the memory, do nothing */ 16991c4bc43dSStefan Agner if (max_addr == PHYS_ADDR_MAX) 1700a571d4ebSDennis Chen return; 1701a571d4ebSDennis Chen 1702c9ca9b4eSAKASHI Takahiro memblock_cap_memory_range(0, max_addr); 1703a571d4ebSDennis Chen } 1704a571d4ebSDennis Chen 1705cd79481dSYinghai Lu static int __init_memblock memblock_search(struct memblock_type *type, phys_addr_t addr) 170672d4b0b4SBenjamin Herrenschmidt { 170772d4b0b4SBenjamin Herrenschmidt unsigned int left = 0, right = type->cnt; 170872d4b0b4SBenjamin Herrenschmidt 170972d4b0b4SBenjamin Herrenschmidt do { 171072d4b0b4SBenjamin Herrenschmidt unsigned int mid = (right + left) / 2; 171172d4b0b4SBenjamin Herrenschmidt 171272d4b0b4SBenjamin Herrenschmidt if (addr < type->regions[mid].base) 171372d4b0b4SBenjamin Herrenschmidt right = mid; 171472d4b0b4SBenjamin Herrenschmidt else if (addr >= (type->regions[mid].base + 171572d4b0b4SBenjamin Herrenschmidt type->regions[mid].size)) 171672d4b0b4SBenjamin Herrenschmidt left = mid + 1; 171772d4b0b4SBenjamin Herrenschmidt else 171872d4b0b4SBenjamin Herrenschmidt return mid; 171972d4b0b4SBenjamin Herrenschmidt } while (left < right); 172072d4b0b4SBenjamin Herrenschmidt return -1; 172172d4b0b4SBenjamin Herrenschmidt } 172272d4b0b4SBenjamin Herrenschmidt 1723f5a222dcSYueyi Li bool __init_memblock memblock_is_reserved(phys_addr_t addr) 172495f72d1eSYinghai Lu { 172572d4b0b4SBenjamin Herrenschmidt return memblock_search(&memblock.reserved, addr) != -1; 172695f72d1eSYinghai Lu } 172772d4b0b4SBenjamin Herrenschmidt 1728b4ad0c7eSYaowei Bai bool __init_memblock memblock_is_memory(phys_addr_t addr) 172972d4b0b4SBenjamin Herrenschmidt { 173072d4b0b4SBenjamin Herrenschmidt return memblock_search(&memblock.memory, addr) != -1; 173172d4b0b4SBenjamin Herrenschmidt } 173272d4b0b4SBenjamin Herrenschmidt 1733937f0c26SYaowei Bai bool __init_memblock memblock_is_map_memory(phys_addr_t addr) 1734bf3d3cc5SArd Biesheuvel { 1735bf3d3cc5SArd Biesheuvel int i = memblock_search(&memblock.memory, addr); 1736bf3d3cc5SArd Biesheuvel 1737bf3d3cc5SArd Biesheuvel if (i == -1) 1738bf3d3cc5SArd Biesheuvel return false; 1739bf3d3cc5SArd Biesheuvel return !memblock_is_nomap(&memblock.memory.regions[i]); 1740bf3d3cc5SArd Biesheuvel } 1741bf3d3cc5SArd Biesheuvel 1742e76b63f8SYinghai Lu #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP 1743e76b63f8SYinghai Lu int __init_memblock memblock_search_pfn_nid(unsigned long pfn, 1744e76b63f8SYinghai Lu unsigned long *start_pfn, unsigned long *end_pfn) 1745e76b63f8SYinghai Lu { 1746e76b63f8SYinghai Lu struct memblock_type *type = &memblock.memory; 174716763230SFabian Frederick int mid = memblock_search(type, PFN_PHYS(pfn)); 1748e76b63f8SYinghai Lu 1749e76b63f8SYinghai Lu if (mid == -1) 1750e76b63f8SYinghai Lu return -1; 1751e76b63f8SYinghai Lu 1752f7e2f7e8SFabian Frederick *start_pfn = PFN_DOWN(type->regions[mid].base); 1753f7e2f7e8SFabian Frederick *end_pfn = PFN_DOWN(type->regions[mid].base + type->regions[mid].size); 1754e76b63f8SYinghai Lu 1755e76b63f8SYinghai Lu return type->regions[mid].nid; 1756e76b63f8SYinghai Lu } 1757e76b63f8SYinghai Lu #endif 1758e76b63f8SYinghai Lu 1759eab30949SStephen Boyd /** 1760eab30949SStephen Boyd * memblock_is_region_memory - check if a region is a subset of memory 1761eab30949SStephen Boyd * @base: base of region to check 1762eab30949SStephen Boyd * @size: size of region to check 1763eab30949SStephen Boyd * 1764eab30949SStephen Boyd * Check if the region [@base, @base + @size) is a subset of a memory block. 1765eab30949SStephen Boyd * 176647cec443SMike Rapoport * Return: 1767eab30949SStephen Boyd * 0 if false, non-zero if true 1768eab30949SStephen Boyd */ 1769937f0c26SYaowei Bai bool __init_memblock memblock_is_region_memory(phys_addr_t base, phys_addr_t size) 177072d4b0b4SBenjamin Herrenschmidt { 1771abb65272STomi Valkeinen int idx = memblock_search(&memblock.memory, base); 1772eb18f1b5STejun Heo phys_addr_t end = base + memblock_cap_size(base, &size); 177372d4b0b4SBenjamin Herrenschmidt 177472d4b0b4SBenjamin Herrenschmidt if (idx == -1) 1775937f0c26SYaowei Bai return false; 1776ef415ef4SWei Yang return (memblock.memory.regions[idx].base + 1777eb18f1b5STejun Heo memblock.memory.regions[idx].size) >= end; 177895f72d1eSYinghai Lu } 177995f72d1eSYinghai Lu 1780eab30949SStephen Boyd /** 1781eab30949SStephen Boyd * memblock_is_region_reserved - check if a region intersects reserved memory 1782eab30949SStephen Boyd * @base: base of region to check 1783eab30949SStephen Boyd * @size: size of region to check 1784eab30949SStephen Boyd * 178547cec443SMike Rapoport * Check if the region [@base, @base + @size) intersects a reserved 178647cec443SMike Rapoport * memory block. 1787eab30949SStephen Boyd * 178847cec443SMike Rapoport * Return: 1789c5c5c9d1STang Chen * True if they intersect, false if not. 1790eab30949SStephen Boyd */ 1791c5c5c9d1STang Chen bool __init_memblock memblock_is_region_reserved(phys_addr_t base, phys_addr_t size) 179295f72d1eSYinghai Lu { 1793eb18f1b5STejun Heo memblock_cap_size(base, &size); 1794c5c5c9d1STang Chen return memblock_overlaps_region(&memblock.reserved, base, size); 179595f72d1eSYinghai Lu } 179695f72d1eSYinghai Lu 17976ede1fd3SYinghai Lu void __init_memblock memblock_trim_memory(phys_addr_t align) 17986ede1fd3SYinghai Lu { 17996ede1fd3SYinghai Lu phys_addr_t start, end, orig_start, orig_end; 1800136199f0SEmil Medve struct memblock_region *r; 18016ede1fd3SYinghai Lu 1802136199f0SEmil Medve for_each_memblock(memory, r) { 1803136199f0SEmil Medve orig_start = r->base; 1804136199f0SEmil Medve orig_end = r->base + r->size; 18056ede1fd3SYinghai Lu start = round_up(orig_start, align); 18066ede1fd3SYinghai Lu end = round_down(orig_end, align); 18076ede1fd3SYinghai Lu 18086ede1fd3SYinghai Lu if (start == orig_start && end == orig_end) 18096ede1fd3SYinghai Lu continue; 18106ede1fd3SYinghai Lu 18116ede1fd3SYinghai Lu if (start < end) { 1812136199f0SEmil Medve r->base = start; 1813136199f0SEmil Medve r->size = end - start; 18146ede1fd3SYinghai Lu } else { 1815136199f0SEmil Medve memblock_remove_region(&memblock.memory, 1816136199f0SEmil Medve r - memblock.memory.regions); 1817136199f0SEmil Medve r--; 18186ede1fd3SYinghai Lu } 18196ede1fd3SYinghai Lu } 18206ede1fd3SYinghai Lu } 1821e63075a3SBenjamin Herrenschmidt 18223661ca66SYinghai Lu void __init_memblock memblock_set_current_limit(phys_addr_t limit) 1823e63075a3SBenjamin Herrenschmidt { 1824e63075a3SBenjamin Herrenschmidt memblock.current_limit = limit; 1825e63075a3SBenjamin Herrenschmidt } 1826e63075a3SBenjamin Herrenschmidt 1827fec51014SLaura Abbott phys_addr_t __init_memblock memblock_get_current_limit(void) 1828fec51014SLaura Abbott { 1829fec51014SLaura Abbott return memblock.current_limit; 1830fec51014SLaura Abbott } 1831fec51014SLaura Abbott 18320262d9c8SHeiko Carstens static void __init_memblock memblock_dump(struct memblock_type *type) 18336ed311b2SBenjamin Herrenschmidt { 18345d63f81cSMiles Chen phys_addr_t base, end, size; 1835e1720feeSMike Rapoport enum memblock_flags flags; 18368c9c1701SAlexander Kuleshov int idx; 18378c9c1701SAlexander Kuleshov struct memblock_region *rgn; 18386ed311b2SBenjamin Herrenschmidt 18390262d9c8SHeiko Carstens pr_info(" %s.cnt = 0x%lx\n", type->name, type->cnt); 18406ed311b2SBenjamin Herrenschmidt 184166e8b438SGioh Kim for_each_memblock_type(idx, type, rgn) { 18427c0caeb8STejun Heo char nid_buf[32] = ""; 18436ed311b2SBenjamin Herrenschmidt 18447c0caeb8STejun Heo base = rgn->base; 18457c0caeb8STejun Heo size = rgn->size; 18465d63f81cSMiles Chen end = base + size - 1; 184766a20757STang Chen flags = rgn->flags; 18487c0caeb8STejun Heo #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP 18497c0caeb8STejun Heo if (memblock_get_region_node(rgn) != MAX_NUMNODES) 18507c0caeb8STejun Heo snprintf(nid_buf, sizeof(nid_buf), " on node %d", 18517c0caeb8STejun Heo memblock_get_region_node(rgn)); 18527c0caeb8STejun Heo #endif 1853e1720feeSMike Rapoport pr_info(" %s[%#x]\t[%pa-%pa], %pa bytes%s flags: %#x\n", 18540262d9c8SHeiko Carstens type->name, idx, &base, &end, &size, nid_buf, flags); 18556ed311b2SBenjamin Herrenschmidt } 18566ed311b2SBenjamin Herrenschmidt } 18576ed311b2SBenjamin Herrenschmidt 18584ff7b82fSTejun Heo void __init_memblock __memblock_dump_all(void) 18596ed311b2SBenjamin Herrenschmidt { 18606ed311b2SBenjamin Herrenschmidt pr_info("MEMBLOCK configuration:\n"); 18615d63f81cSMiles Chen pr_info(" memory size = %pa reserved size = %pa\n", 18625d63f81cSMiles Chen &memblock.memory.total_size, 18635d63f81cSMiles Chen &memblock.reserved.total_size); 18646ed311b2SBenjamin Herrenschmidt 18650262d9c8SHeiko Carstens memblock_dump(&memblock.memory); 18660262d9c8SHeiko Carstens memblock_dump(&memblock.reserved); 1867409efd4cSHeiko Carstens #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP 18680262d9c8SHeiko Carstens memblock_dump(&memblock.physmem); 1869409efd4cSHeiko Carstens #endif 18706ed311b2SBenjamin Herrenschmidt } 18716ed311b2SBenjamin Herrenschmidt 18721aadc056STejun Heo void __init memblock_allow_resize(void) 18736ed311b2SBenjamin Herrenschmidt { 1874142b45a7SBenjamin Herrenschmidt memblock_can_resize = 1; 18756ed311b2SBenjamin Herrenschmidt } 18766ed311b2SBenjamin Herrenschmidt 18776ed311b2SBenjamin Herrenschmidt static int __init early_memblock(char *p) 18786ed311b2SBenjamin Herrenschmidt { 18796ed311b2SBenjamin Herrenschmidt if (p && strstr(p, "debug")) 18806ed311b2SBenjamin Herrenschmidt memblock_debug = 1; 18816ed311b2SBenjamin Herrenschmidt return 0; 18826ed311b2SBenjamin Herrenschmidt } 18836ed311b2SBenjamin Herrenschmidt early_param("memblock", early_memblock); 18846ed311b2SBenjamin Herrenschmidt 1885bda49a81SMike Rapoport static void __init __free_pages_memory(unsigned long start, unsigned long end) 1886bda49a81SMike Rapoport { 1887bda49a81SMike Rapoport int order; 1888bda49a81SMike Rapoport 1889bda49a81SMike Rapoport while (start < end) { 1890bda49a81SMike Rapoport order = min(MAX_ORDER - 1UL, __ffs(start)); 1891bda49a81SMike Rapoport 1892bda49a81SMike Rapoport while (start + (1UL << order) > end) 1893bda49a81SMike Rapoport order--; 1894bda49a81SMike Rapoport 1895bda49a81SMike Rapoport memblock_free_pages(pfn_to_page(start), start, order); 1896bda49a81SMike Rapoport 1897bda49a81SMike Rapoport start += (1UL << order); 1898bda49a81SMike Rapoport } 1899bda49a81SMike Rapoport } 1900bda49a81SMike Rapoport 1901bda49a81SMike Rapoport static unsigned long __init __free_memory_core(phys_addr_t start, 1902bda49a81SMike Rapoport phys_addr_t end) 1903bda49a81SMike Rapoport { 1904bda49a81SMike Rapoport unsigned long start_pfn = PFN_UP(start); 1905bda49a81SMike Rapoport unsigned long end_pfn = min_t(unsigned long, 1906bda49a81SMike Rapoport PFN_DOWN(end), max_low_pfn); 1907bda49a81SMike Rapoport 1908bda49a81SMike Rapoport if (start_pfn >= end_pfn) 1909bda49a81SMike Rapoport return 0; 1910bda49a81SMike Rapoport 1911bda49a81SMike Rapoport __free_pages_memory(start_pfn, end_pfn); 1912bda49a81SMike Rapoport 1913bda49a81SMike Rapoport return end_pfn - start_pfn; 1914bda49a81SMike Rapoport } 1915bda49a81SMike Rapoport 1916bda49a81SMike Rapoport static unsigned long __init free_low_memory_core_early(void) 1917bda49a81SMike Rapoport { 1918bda49a81SMike Rapoport unsigned long count = 0; 1919bda49a81SMike Rapoport phys_addr_t start, end; 1920bda49a81SMike Rapoport u64 i; 1921bda49a81SMike Rapoport 1922bda49a81SMike Rapoport memblock_clear_hotplug(0, -1); 1923bda49a81SMike Rapoport 1924bda49a81SMike Rapoport for_each_reserved_mem_region(i, &start, &end) 1925bda49a81SMike Rapoport reserve_bootmem_region(start, end); 1926bda49a81SMike Rapoport 1927bda49a81SMike Rapoport /* 1928bda49a81SMike Rapoport * We need to use NUMA_NO_NODE instead of NODE_DATA(0)->node_id 1929bda49a81SMike Rapoport * because in some case like Node0 doesn't have RAM installed 1930bda49a81SMike Rapoport * low ram will be on Node1 1931bda49a81SMike Rapoport */ 1932bda49a81SMike Rapoport for_each_free_mem_range(i, NUMA_NO_NODE, MEMBLOCK_NONE, &start, &end, 1933bda49a81SMike Rapoport NULL) 1934bda49a81SMike Rapoport count += __free_memory_core(start, end); 1935bda49a81SMike Rapoport 1936bda49a81SMike Rapoport return count; 1937bda49a81SMike Rapoport } 1938bda49a81SMike Rapoport 1939bda49a81SMike Rapoport static int reset_managed_pages_done __initdata; 1940bda49a81SMike Rapoport 1941bda49a81SMike Rapoport void reset_node_managed_pages(pg_data_t *pgdat) 1942bda49a81SMike Rapoport { 1943bda49a81SMike Rapoport struct zone *z; 1944bda49a81SMike Rapoport 1945bda49a81SMike Rapoport for (z = pgdat->node_zones; z < pgdat->node_zones + MAX_NR_ZONES; z++) 19469705bea5SArun KS atomic_long_set(&z->managed_pages, 0); 1947bda49a81SMike Rapoport } 1948bda49a81SMike Rapoport 1949bda49a81SMike Rapoport void __init reset_all_zones_managed_pages(void) 1950bda49a81SMike Rapoport { 1951bda49a81SMike Rapoport struct pglist_data *pgdat; 1952bda49a81SMike Rapoport 1953bda49a81SMike Rapoport if (reset_managed_pages_done) 1954bda49a81SMike Rapoport return; 1955bda49a81SMike Rapoport 1956bda49a81SMike Rapoport for_each_online_pgdat(pgdat) 1957bda49a81SMike Rapoport reset_node_managed_pages(pgdat); 1958bda49a81SMike Rapoport 1959bda49a81SMike Rapoport reset_managed_pages_done = 1; 1960bda49a81SMike Rapoport } 1961bda49a81SMike Rapoport 1962bda49a81SMike Rapoport /** 1963bda49a81SMike Rapoport * memblock_free_all - release free pages to the buddy allocator 1964bda49a81SMike Rapoport * 1965bda49a81SMike Rapoport * Return: the number of pages actually released. 1966bda49a81SMike Rapoport */ 1967bda49a81SMike Rapoport unsigned long __init memblock_free_all(void) 1968bda49a81SMike Rapoport { 1969bda49a81SMike Rapoport unsigned long pages; 1970bda49a81SMike Rapoport 1971bda49a81SMike Rapoport reset_all_zones_managed_pages(); 1972bda49a81SMike Rapoport 1973bda49a81SMike Rapoport pages = free_low_memory_core_early(); 1974ca79b0c2SArun KS totalram_pages_add(pages); 1975bda49a81SMike Rapoport 1976bda49a81SMike Rapoport return pages; 1977bda49a81SMike Rapoport } 1978bda49a81SMike Rapoport 1979c378ddd5STejun Heo #if defined(CONFIG_DEBUG_FS) && !defined(CONFIG_ARCH_DISCARD_MEMBLOCK) 19806d03b885SBenjamin Herrenschmidt 19816d03b885SBenjamin Herrenschmidt static int memblock_debug_show(struct seq_file *m, void *private) 19826d03b885SBenjamin Herrenschmidt { 19836d03b885SBenjamin Herrenschmidt struct memblock_type *type = m->private; 19846d03b885SBenjamin Herrenschmidt struct memblock_region *reg; 19856d03b885SBenjamin Herrenschmidt int i; 19865d63f81cSMiles Chen phys_addr_t end; 19876d03b885SBenjamin Herrenschmidt 19886d03b885SBenjamin Herrenschmidt for (i = 0; i < type->cnt; i++) { 19896d03b885SBenjamin Herrenschmidt reg = &type->regions[i]; 19905d63f81cSMiles Chen end = reg->base + reg->size - 1; 19916d03b885SBenjamin Herrenschmidt 19925d63f81cSMiles Chen seq_printf(m, "%4d: ", i); 19935d63f81cSMiles Chen seq_printf(m, "%pa..%pa\n", ®->base, &end); 19946d03b885SBenjamin Herrenschmidt } 19956d03b885SBenjamin Herrenschmidt return 0; 19966d03b885SBenjamin Herrenschmidt } 19975ad35093SAndy Shevchenko DEFINE_SHOW_ATTRIBUTE(memblock_debug); 19986d03b885SBenjamin Herrenschmidt 19996d03b885SBenjamin Herrenschmidt static int __init memblock_init_debugfs(void) 20006d03b885SBenjamin Herrenschmidt { 20016d03b885SBenjamin Herrenschmidt struct dentry *root = debugfs_create_dir("memblock", NULL); 2002d9f7979cSGreg Kroah-Hartman 20030825a6f9SJoe Perches debugfs_create_file("memory", 0444, root, 20040825a6f9SJoe Perches &memblock.memory, &memblock_debug_fops); 20050825a6f9SJoe Perches debugfs_create_file("reserved", 0444, root, 20060825a6f9SJoe Perches &memblock.reserved, &memblock_debug_fops); 200770210ed9SPhilipp Hachtmann #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP 20080825a6f9SJoe Perches debugfs_create_file("physmem", 0444, root, 20090825a6f9SJoe Perches &memblock.physmem, &memblock_debug_fops); 201070210ed9SPhilipp Hachtmann #endif 20116d03b885SBenjamin Herrenschmidt 20126d03b885SBenjamin Herrenschmidt return 0; 20136d03b885SBenjamin Herrenschmidt } 20146d03b885SBenjamin Herrenschmidt __initcall(memblock_init_debugfs); 20156d03b885SBenjamin Herrenschmidt 20166d03b885SBenjamin Herrenschmidt #endif /* CONFIG_DEBUG_FS */ 2017