1 // SPDX-License-Identifier: GPL-2.0 2 #include <linux/memblock.h> 3 #include <linux/compiler.h> 4 #include <linux/fs.h> 5 #include <linux/init.h> 6 #include <linux/ksm.h> 7 #include <linux/mm.h> 8 #include <linux/mmzone.h> 9 #include <linux/huge_mm.h> 10 #include <linux/proc_fs.h> 11 #include <linux/seq_file.h> 12 #include <linux/hugetlb.h> 13 #include <linux/memremap.h> 14 #include <linux/memcontrol.h> 15 #include <linux/mmu_notifier.h> 16 #include <linux/page_idle.h> 17 #include <linux/kernel-page-flags.h> 18 #include <linux/uaccess.h> 19 #include "internal.h" 20 21 #define KPMSIZE sizeof(u64) 22 #define KPMMASK (KPMSIZE - 1) 23 #define KPMBITS (KPMSIZE * BITS_PER_BYTE) 24 25 static inline unsigned long get_max_dump_pfn(void) 26 { 27 #ifdef CONFIG_SPARSEMEM 28 /* 29 * The memmap of early sections is completely populated and marked 30 * online even if max_pfn does not fall on a section boundary - 31 * pfn_to_online_page() will succeed on all pages. Allow inspecting 32 * these memmaps. 33 */ 34 return round_up(max_pfn, PAGES_PER_SECTION); 35 #else 36 return max_pfn; 37 #endif 38 } 39 40 /* /proc/kpagecount - an array exposing page mapcounts 41 * 42 * Each entry is a u64 representing the corresponding 43 * physical page mapcount. 44 */ 45 static ssize_t kpagecount_read(struct file *file, char __user *buf, 46 size_t count, loff_t *ppos) 47 { 48 const unsigned long max_dump_pfn = get_max_dump_pfn(); 49 u64 __user *out = (u64 __user *)buf; 50 unsigned long src = *ppos; 51 unsigned long pfn; 52 ssize_t ret = 0; 53 54 pfn = src / KPMSIZE; 55 if (src & KPMMASK || count & KPMMASK) 56 return -EINVAL; 57 if (src >= max_dump_pfn * KPMSIZE) 58 return 0; 59 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src); 60 61 while (count > 0) { 62 struct page *page; 63 u64 mapcount = 0; 64 65 /* 66 * TODO: ZONE_DEVICE support requires to identify 67 * memmaps that were actually initialized. 68 */ 69 page = pfn_to_online_page(pfn); 70 if (page) 71 mapcount = folio_precise_page_mapcount(page_folio(page), 72 page); 73 74 if (put_user(mapcount, out)) { 75 ret = -EFAULT; 76 break; 77 } 78 79 pfn++; 80 out++; 81 count -= KPMSIZE; 82 83 cond_resched(); 84 } 85 86 *ppos += (char __user *)out - buf; 87 if (!ret) 88 ret = (char __user *)out - buf; 89 return ret; 90 } 91 92 static const struct proc_ops kpagecount_proc_ops = { 93 .proc_flags = PROC_ENTRY_PERMANENT, 94 .proc_lseek = mem_lseek, 95 .proc_read = kpagecount_read, 96 }; 97 98 /* /proc/kpageflags - an array exposing page flags 99 * 100 * Each entry is a u64 representing the corresponding 101 * physical page flags. 102 */ 103 104 static inline u64 kpf_copy_bit(u64 kflags, int ubit, int kbit) 105 { 106 return ((kflags >> kbit) & 1) << ubit; 107 } 108 109 u64 stable_page_flags(const struct page *page) 110 { 111 const struct folio *folio; 112 unsigned long k; 113 unsigned long mapping; 114 bool is_anon; 115 u64 u = 0; 116 117 /* 118 * pseudo flag: KPF_NOPAGE 119 * it differentiates a memory hole from a page with no flags 120 */ 121 if (!page) 122 return 1 << KPF_NOPAGE; 123 folio = page_folio(page); 124 125 k = folio->flags; 126 mapping = (unsigned long)folio->mapping; 127 is_anon = mapping & PAGE_MAPPING_ANON; 128 129 /* 130 * pseudo flags for the well known (anonymous) memory mapped pages 131 */ 132 if (page_mapped(page)) 133 u |= 1 << KPF_MMAP; 134 if (is_anon) { 135 u |= 1 << KPF_ANON; 136 if (mapping & PAGE_MAPPING_KSM) 137 u |= 1 << KPF_KSM; 138 } 139 140 /* 141 * compound pages: export both head/tail info 142 * they together define a compound page's start/end pos and order 143 */ 144 if (page == &folio->page) 145 u |= kpf_copy_bit(k, KPF_COMPOUND_HEAD, PG_head); 146 else 147 u |= 1 << KPF_COMPOUND_TAIL; 148 if (folio_test_hugetlb(folio)) 149 u |= 1 << KPF_HUGE; 150 else if (folio_test_large(folio) && 151 folio_test_large_rmappable(folio)) { 152 /* Note: we indicate any THPs here, not just PMD-sized ones */ 153 u |= 1 << KPF_THP; 154 } else if (is_huge_zero_folio(folio)) { 155 u |= 1 << KPF_ZERO_PAGE; 156 u |= 1 << KPF_THP; 157 } else if (is_zero_folio(folio)) { 158 u |= 1 << KPF_ZERO_PAGE; 159 } 160 161 /* 162 * Caveats on high order pages: PG_buddy and PG_slab will only be set 163 * on the head page. 164 */ 165 if (PageBuddy(page)) 166 u |= 1 << KPF_BUDDY; 167 else if (page_count(page) == 0 && is_free_buddy_page(page)) 168 u |= 1 << KPF_BUDDY; 169 170 if (PageOffline(page)) 171 u |= 1 << KPF_OFFLINE; 172 if (PageTable(page)) 173 u |= 1 << KPF_PGTABLE; 174 if (folio_test_slab(folio)) 175 u |= 1 << KPF_SLAB; 176 177 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT) 178 u |= kpf_copy_bit(k, KPF_IDLE, PG_idle); 179 #else 180 if (folio_test_idle(folio)) 181 u |= 1 << KPF_IDLE; 182 #endif 183 184 u |= kpf_copy_bit(k, KPF_LOCKED, PG_locked); 185 u |= kpf_copy_bit(k, KPF_ERROR, PG_error); 186 u |= kpf_copy_bit(k, KPF_DIRTY, PG_dirty); 187 u |= kpf_copy_bit(k, KPF_UPTODATE, PG_uptodate); 188 u |= kpf_copy_bit(k, KPF_WRITEBACK, PG_writeback); 189 190 u |= kpf_copy_bit(k, KPF_LRU, PG_lru); 191 u |= kpf_copy_bit(k, KPF_REFERENCED, PG_referenced); 192 u |= kpf_copy_bit(k, KPF_ACTIVE, PG_active); 193 u |= kpf_copy_bit(k, KPF_RECLAIM, PG_reclaim); 194 195 #define SWAPCACHE ((1 << PG_swapbacked) | (1 << PG_swapcache)) 196 if ((k & SWAPCACHE) == SWAPCACHE) 197 u |= 1 << KPF_SWAPCACHE; 198 u |= kpf_copy_bit(k, KPF_SWAPBACKED, PG_swapbacked); 199 200 u |= kpf_copy_bit(k, KPF_UNEVICTABLE, PG_unevictable); 201 u |= kpf_copy_bit(k, KPF_MLOCKED, PG_mlocked); 202 203 #ifdef CONFIG_MEMORY_FAILURE 204 if (u & (1 << KPF_HUGE)) 205 u |= kpf_copy_bit(k, KPF_HWPOISON, PG_hwpoison); 206 else 207 u |= kpf_copy_bit(page->flags, KPF_HWPOISON, PG_hwpoison); 208 #endif 209 210 #ifdef CONFIG_ARCH_USES_PG_UNCACHED 211 u |= kpf_copy_bit(k, KPF_UNCACHED, PG_uncached); 212 #endif 213 214 u |= kpf_copy_bit(k, KPF_RESERVED, PG_reserved); 215 u |= kpf_copy_bit(k, KPF_MAPPEDTODISK, PG_mappedtodisk); 216 u |= kpf_copy_bit(k, KPF_PRIVATE, PG_private); 217 u |= kpf_copy_bit(k, KPF_PRIVATE_2, PG_private_2); 218 u |= kpf_copy_bit(k, KPF_OWNER_PRIVATE, PG_owner_priv_1); 219 u |= kpf_copy_bit(k, KPF_ARCH, PG_arch_1); 220 #ifdef CONFIG_ARCH_USES_PG_ARCH_X 221 u |= kpf_copy_bit(k, KPF_ARCH_2, PG_arch_2); 222 u |= kpf_copy_bit(k, KPF_ARCH_3, PG_arch_3); 223 #endif 224 225 return u; 226 }; 227 228 static ssize_t kpageflags_read(struct file *file, char __user *buf, 229 size_t count, loff_t *ppos) 230 { 231 const unsigned long max_dump_pfn = get_max_dump_pfn(); 232 u64 __user *out = (u64 __user *)buf; 233 unsigned long src = *ppos; 234 unsigned long pfn; 235 ssize_t ret = 0; 236 237 pfn = src / KPMSIZE; 238 if (src & KPMMASK || count & KPMMASK) 239 return -EINVAL; 240 if (src >= max_dump_pfn * KPMSIZE) 241 return 0; 242 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src); 243 244 while (count > 0) { 245 /* 246 * TODO: ZONE_DEVICE support requires to identify 247 * memmaps that were actually initialized. 248 */ 249 struct page *page = pfn_to_online_page(pfn); 250 251 if (put_user(stable_page_flags(page), out)) { 252 ret = -EFAULT; 253 break; 254 } 255 256 pfn++; 257 out++; 258 count -= KPMSIZE; 259 260 cond_resched(); 261 } 262 263 *ppos += (char __user *)out - buf; 264 if (!ret) 265 ret = (char __user *)out - buf; 266 return ret; 267 } 268 269 static const struct proc_ops kpageflags_proc_ops = { 270 .proc_flags = PROC_ENTRY_PERMANENT, 271 .proc_lseek = mem_lseek, 272 .proc_read = kpageflags_read, 273 }; 274 275 #ifdef CONFIG_MEMCG 276 static ssize_t kpagecgroup_read(struct file *file, char __user *buf, 277 size_t count, loff_t *ppos) 278 { 279 const unsigned long max_dump_pfn = get_max_dump_pfn(); 280 u64 __user *out = (u64 __user *)buf; 281 struct page *ppage; 282 unsigned long src = *ppos; 283 unsigned long pfn; 284 ssize_t ret = 0; 285 u64 ino; 286 287 pfn = src / KPMSIZE; 288 if (src & KPMMASK || count & KPMMASK) 289 return -EINVAL; 290 if (src >= max_dump_pfn * KPMSIZE) 291 return 0; 292 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src); 293 294 while (count > 0) { 295 /* 296 * TODO: ZONE_DEVICE support requires to identify 297 * memmaps that were actually initialized. 298 */ 299 ppage = pfn_to_online_page(pfn); 300 301 if (ppage) 302 ino = page_cgroup_ino(ppage); 303 else 304 ino = 0; 305 306 if (put_user(ino, out)) { 307 ret = -EFAULT; 308 break; 309 } 310 311 pfn++; 312 out++; 313 count -= KPMSIZE; 314 315 cond_resched(); 316 } 317 318 *ppos += (char __user *)out - buf; 319 if (!ret) 320 ret = (char __user *)out - buf; 321 return ret; 322 } 323 324 static const struct proc_ops kpagecgroup_proc_ops = { 325 .proc_flags = PROC_ENTRY_PERMANENT, 326 .proc_lseek = mem_lseek, 327 .proc_read = kpagecgroup_read, 328 }; 329 #endif /* CONFIG_MEMCG */ 330 331 static int __init proc_page_init(void) 332 { 333 proc_create("kpagecount", S_IRUSR, NULL, &kpagecount_proc_ops); 334 proc_create("kpageflags", S_IRUSR, NULL, &kpageflags_proc_ops); 335 #ifdef CONFIG_MEMCG 336 proc_create("kpagecgroup", S_IRUSR, NULL, &kpagecgroup_proc_ops); 337 #endif 338 return 0; 339 } 340 fs_initcall(proc_page_init); 341