1 // SPDX-License-Identifier: GPL-2.0 1 // SPDX-License-Identifier: GPL-2.0
2 #include <linux/memblock.h> 2 #include <linux/memblock.h>
3 #include <linux/compiler.h> 3 #include <linux/compiler.h>
4 #include <linux/fs.h> 4 #include <linux/fs.h>
5 #include <linux/init.h> 5 #include <linux/init.h>
6 #include <linux/ksm.h> 6 #include <linux/ksm.h>
7 #include <linux/mm.h> 7 #include <linux/mm.h>
8 #include <linux/mmzone.h> 8 #include <linux/mmzone.h>
9 #include <linux/huge_mm.h> 9 #include <linux/huge_mm.h>
10 #include <linux/proc_fs.h> 10 #include <linux/proc_fs.h>
11 #include <linux/seq_file.h> 11 #include <linux/seq_file.h>
12 #include <linux/hugetlb.h> 12 #include <linux/hugetlb.h>
13 #include <linux/memremap.h> 13 #include <linux/memremap.h>
14 #include <linux/memcontrol.h> 14 #include <linux/memcontrol.h>
15 #include <linux/mmu_notifier.h> 15 #include <linux/mmu_notifier.h>
16 #include <linux/page_idle.h> 16 #include <linux/page_idle.h>
17 #include <linux/kernel-page-flags.h> 17 #include <linux/kernel-page-flags.h>
18 #include <linux/uaccess.h> 18 #include <linux/uaccess.h>
19 #include "internal.h" 19 #include "internal.h"
20 20
21 #define KPMSIZE sizeof(u64) 21 #define KPMSIZE sizeof(u64)
22 #define KPMMASK (KPMSIZE - 1) 22 #define KPMMASK (KPMSIZE - 1)
23 #define KPMBITS (KPMSIZE * BITS_PER_BYTE) 23 #define KPMBITS (KPMSIZE * BITS_PER_BYTE)
24 24
25 static inline unsigned long get_max_dump_pfn(v 25 static inline unsigned long get_max_dump_pfn(void)
26 { 26 {
27 #ifdef CONFIG_SPARSEMEM 27 #ifdef CONFIG_SPARSEMEM
28 /* 28 /*
29 * The memmap of early sections is com 29 * The memmap of early sections is completely populated and marked
30 * online even if max_pfn does not fal 30 * online even if max_pfn does not fall on a section boundary -
31 * pfn_to_online_page() will succeed o 31 * pfn_to_online_page() will succeed on all pages. Allow inspecting
32 * these memmaps. 32 * these memmaps.
33 */ 33 */
34 return round_up(max_pfn, PAGES_PER_SEC 34 return round_up(max_pfn, PAGES_PER_SECTION);
35 #else 35 #else
36 return max_pfn; 36 return max_pfn;
37 #endif 37 #endif
38 } 38 }
39 39
40 /* /proc/kpagecount - an array exposing page m 40 /* /proc/kpagecount - an array exposing page mapcounts
41 * 41 *
42 * Each entry is a u64 representing the corres 42 * Each entry is a u64 representing the corresponding
43 * physical page mapcount. 43 * physical page mapcount.
44 */ 44 */
45 static ssize_t kpagecount_read(struct file *fi 45 static ssize_t kpagecount_read(struct file *file, char __user *buf,
46 size_t count, lof 46 size_t count, loff_t *ppos)
47 { 47 {
48 const unsigned long max_dump_pfn = get 48 const unsigned long max_dump_pfn = get_max_dump_pfn();
49 u64 __user *out = (u64 __user *)buf; 49 u64 __user *out = (u64 __user *)buf;
50 unsigned long src = *ppos; 50 unsigned long src = *ppos;
51 unsigned long pfn; 51 unsigned long pfn;
52 ssize_t ret = 0; 52 ssize_t ret = 0;
53 53
54 pfn = src / KPMSIZE; 54 pfn = src / KPMSIZE;
55 if (src & KPMMASK || count & KPMMASK) 55 if (src & KPMMASK || count & KPMMASK)
56 return -EINVAL; 56 return -EINVAL;
57 if (src >= max_dump_pfn * KPMSIZE) 57 if (src >= max_dump_pfn * KPMSIZE)
58 return 0; 58 return 0;
59 count = min_t(unsigned long, count, (m 59 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src);
60 60
61 while (count > 0) { 61 while (count > 0) {
62 struct page *page; 62 struct page *page;
63 u64 mapcount = 0; 63 u64 mapcount = 0;
64 64
65 /* 65 /*
66 * TODO: ZONE_DEVICE support r 66 * TODO: ZONE_DEVICE support requires to identify
67 * memmaps that were actually 67 * memmaps that were actually initialized.
68 */ 68 */
69 page = pfn_to_online_page(pfn) 69 page = pfn_to_online_page(pfn);
70 if (page) 70 if (page)
71 mapcount = folio_preci 71 mapcount = folio_precise_page_mapcount(page_folio(page),
72 72 page);
73 73
74 if (put_user(mapcount, out)) { 74 if (put_user(mapcount, out)) {
75 ret = -EFAULT; 75 ret = -EFAULT;
76 break; 76 break;
77 } 77 }
78 78
79 pfn++; 79 pfn++;
80 out++; 80 out++;
81 count -= KPMSIZE; 81 count -= KPMSIZE;
82 82
83 cond_resched(); 83 cond_resched();
84 } 84 }
85 85
86 *ppos += (char __user *)out - buf; 86 *ppos += (char __user *)out - buf;
87 if (!ret) 87 if (!ret)
88 ret = (char __user *)out - buf 88 ret = (char __user *)out - buf;
89 return ret; 89 return ret;
90 } 90 }
91 91
92 static const struct proc_ops kpagecount_proc_o 92 static const struct proc_ops kpagecount_proc_ops = {
93 .proc_flags = PROC_ENTRY_PERMANENT 93 .proc_flags = PROC_ENTRY_PERMANENT,
94 .proc_lseek = mem_lseek, 94 .proc_lseek = mem_lseek,
95 .proc_read = kpagecount_read, 95 .proc_read = kpagecount_read,
96 }; 96 };
97 97
98 /* /proc/kpageflags - an array exposing page f 98 /* /proc/kpageflags - an array exposing page flags
99 * 99 *
100 * Each entry is a u64 representing the corres 100 * Each entry is a u64 representing the corresponding
101 * physical page flags. 101 * physical page flags.
102 */ 102 */
103 103
104 static inline u64 kpf_copy_bit(u64 kflags, int 104 static inline u64 kpf_copy_bit(u64 kflags, int ubit, int kbit)
105 { 105 {
106 return ((kflags >> kbit) & 1) << ubit; 106 return ((kflags >> kbit) & 1) << ubit;
107 } 107 }
108 108
109 u64 stable_page_flags(const struct page *page) 109 u64 stable_page_flags(const struct page *page)
110 { 110 {
111 const struct folio *folio; 111 const struct folio *folio;
112 unsigned long k; 112 unsigned long k;
113 unsigned long mapping; 113 unsigned long mapping;
114 bool is_anon; 114 bool is_anon;
115 u64 u = 0; 115 u64 u = 0;
116 116
117 /* 117 /*
118 * pseudo flag: KPF_NOPAGE 118 * pseudo flag: KPF_NOPAGE
119 * it differentiates a memory hole fro 119 * it differentiates a memory hole from a page with no flags
120 */ 120 */
121 if (!page) 121 if (!page)
122 return 1 << KPF_NOPAGE; 122 return 1 << KPF_NOPAGE;
123 folio = page_folio(page); 123 folio = page_folio(page);
124 124
125 k = folio->flags; 125 k = folio->flags;
126 mapping = (unsigned long)folio->mappin 126 mapping = (unsigned long)folio->mapping;
127 is_anon = mapping & PAGE_MAPPING_ANON; 127 is_anon = mapping & PAGE_MAPPING_ANON;
128 128
129 /* 129 /*
130 * pseudo flags for the well known (an 130 * pseudo flags for the well known (anonymous) memory mapped pages
131 */ 131 */
132 if (page_mapped(page)) 132 if (page_mapped(page))
133 u |= 1 << KPF_MMAP; 133 u |= 1 << KPF_MMAP;
134 if (is_anon) { 134 if (is_anon) {
135 u |= 1 << KPF_ANON; 135 u |= 1 << KPF_ANON;
136 if (mapping & PAGE_MAPPING_KSM 136 if (mapping & PAGE_MAPPING_KSM)
137 u |= 1 << KPF_KSM; 137 u |= 1 << KPF_KSM;
138 } 138 }
139 139
140 /* 140 /*
141 * compound pages: export both head/ta 141 * compound pages: export both head/tail info
142 * they together define a compound pag 142 * they together define a compound page's start/end pos and order
143 */ 143 */
144 if (page == &folio->page) 144 if (page == &folio->page)
145 u |= kpf_copy_bit(k, KPF_COMPO 145 u |= kpf_copy_bit(k, KPF_COMPOUND_HEAD, PG_head);
146 else 146 else
147 u |= 1 << KPF_COMPOUND_TAIL; 147 u |= 1 << KPF_COMPOUND_TAIL;
148 if (folio_test_hugetlb(folio)) 148 if (folio_test_hugetlb(folio))
149 u |= 1 << KPF_HUGE; 149 u |= 1 << KPF_HUGE;
150 else if (folio_test_large(folio) && 150 else if (folio_test_large(folio) &&
151 folio_test_large_rmappable(fo 151 folio_test_large_rmappable(folio)) {
152 /* Note: we indicate any THPs 152 /* Note: we indicate any THPs here, not just PMD-sized ones */
153 u |= 1 << KPF_THP; 153 u |= 1 << KPF_THP;
154 } else if (is_huge_zero_folio(folio)) 154 } else if (is_huge_zero_folio(folio)) {
155 u |= 1 << KPF_ZERO_PAGE; 155 u |= 1 << KPF_ZERO_PAGE;
156 u |= 1 << KPF_THP; 156 u |= 1 << KPF_THP;
157 } else if (is_zero_folio(folio)) { 157 } else if (is_zero_folio(folio)) {
158 u |= 1 << KPF_ZERO_PAGE; 158 u |= 1 << KPF_ZERO_PAGE;
159 } 159 }
160 160
161 /* 161 /*
162 * Caveats on high order pages: PG_bud 162 * Caveats on high order pages: PG_buddy and PG_slab will only be set
163 * on the head page. 163 * on the head page.
164 */ 164 */
165 if (PageBuddy(page)) 165 if (PageBuddy(page))
166 u |= 1 << KPF_BUDDY; 166 u |= 1 << KPF_BUDDY;
167 else if (page_count(page) == 0 && is_f 167 else if (page_count(page) == 0 && is_free_buddy_page(page))
168 u |= 1 << KPF_BUDDY; 168 u |= 1 << KPF_BUDDY;
169 169
170 if (PageOffline(page)) 170 if (PageOffline(page))
171 u |= 1 << KPF_OFFLINE; 171 u |= 1 << KPF_OFFLINE;
172 if (PageTable(page)) 172 if (PageTable(page))
173 u |= 1 << KPF_PGTABLE; 173 u |= 1 << KPF_PGTABLE;
174 if (folio_test_slab(folio)) 174 if (folio_test_slab(folio))
175 u |= 1 << KPF_SLAB; 175 u |= 1 << KPF_SLAB;
176 176
177 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined( 177 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT)
178 u |= kpf_copy_bit(k, KPF_IDLE, 178 u |= kpf_copy_bit(k, KPF_IDLE, PG_idle);
179 #else 179 #else
180 if (folio_test_idle(folio)) 180 if (folio_test_idle(folio))
181 u |= 1 << KPF_IDLE; 181 u |= 1 << KPF_IDLE;
182 #endif 182 #endif
183 183
184 u |= kpf_copy_bit(k, KPF_LOCKED, 184 u |= kpf_copy_bit(k, KPF_LOCKED, PG_locked);
185 u |= kpf_copy_bit(k, KPF_DIRTY, 185 u |= kpf_copy_bit(k, KPF_DIRTY, PG_dirty);
186 u |= kpf_copy_bit(k, KPF_UPTODATE, 186 u |= kpf_copy_bit(k, KPF_UPTODATE, PG_uptodate);
187 u |= kpf_copy_bit(k, KPF_WRITEBACK, 187 u |= kpf_copy_bit(k, KPF_WRITEBACK, PG_writeback);
188 188
189 u |= kpf_copy_bit(k, KPF_LRU, 189 u |= kpf_copy_bit(k, KPF_LRU, PG_lru);
190 u |= kpf_copy_bit(k, KPF_REFERENCED, 190 u |= kpf_copy_bit(k, KPF_REFERENCED, PG_referenced);
191 u |= kpf_copy_bit(k, KPF_ACTIVE, 191 u |= kpf_copy_bit(k, KPF_ACTIVE, PG_active);
192 u |= kpf_copy_bit(k, KPF_RECLAIM, 192 u |= kpf_copy_bit(k, KPF_RECLAIM, PG_reclaim);
193 193
194 #define SWAPCACHE ((1 << PG_swapbacked) | (1 < 194 #define SWAPCACHE ((1 << PG_swapbacked) | (1 << PG_swapcache))
195 if ((k & SWAPCACHE) == SWAPCACHE) 195 if ((k & SWAPCACHE) == SWAPCACHE)
196 u |= 1 << KPF_SWAPCACHE; 196 u |= 1 << KPF_SWAPCACHE;
197 u |= kpf_copy_bit(k, KPF_SWAPBACKED, 197 u |= kpf_copy_bit(k, KPF_SWAPBACKED, PG_swapbacked);
198 198
199 u |= kpf_copy_bit(k, KPF_UNEVICTABLE, 199 u |= kpf_copy_bit(k, KPF_UNEVICTABLE, PG_unevictable);
200 u |= kpf_copy_bit(k, KPF_MLOCKED, 200 u |= kpf_copy_bit(k, KPF_MLOCKED, PG_mlocked);
201 201
202 #ifdef CONFIG_MEMORY_FAILURE 202 #ifdef CONFIG_MEMORY_FAILURE
203 if (u & (1 << KPF_HUGE)) 203 if (u & (1 << KPF_HUGE))
204 u |= kpf_copy_bit(k, KPF_HWPOI 204 u |= kpf_copy_bit(k, KPF_HWPOISON, PG_hwpoison);
205 else 205 else
206 u |= kpf_copy_bit(page->flags, 206 u |= kpf_copy_bit(page->flags, KPF_HWPOISON, PG_hwpoison);
207 #endif 207 #endif
208 208
209 u |= kpf_copy_bit(k, KPF_RESERVED, 209 u |= kpf_copy_bit(k, KPF_RESERVED, PG_reserved);
210 u |= kpf_copy_bit(k, KPF_OWNER_2, 210 u |= kpf_copy_bit(k, KPF_OWNER_2, PG_owner_2);
211 u |= kpf_copy_bit(k, KPF_PRIVATE, 211 u |= kpf_copy_bit(k, KPF_PRIVATE, PG_private);
212 u |= kpf_copy_bit(k, KPF_PRIVATE_2, 212 u |= kpf_copy_bit(k, KPF_PRIVATE_2, PG_private_2);
213 u |= kpf_copy_bit(k, KPF_OWNER_PRIVATE 213 u |= kpf_copy_bit(k, KPF_OWNER_PRIVATE, PG_owner_priv_1);
214 u |= kpf_copy_bit(k, KPF_ARCH, 214 u |= kpf_copy_bit(k, KPF_ARCH, PG_arch_1);
215 #ifdef CONFIG_ARCH_USES_PG_ARCH_2 215 #ifdef CONFIG_ARCH_USES_PG_ARCH_2
216 u |= kpf_copy_bit(k, KPF_ARCH_2, 216 u |= kpf_copy_bit(k, KPF_ARCH_2, PG_arch_2);
217 #endif 217 #endif
218 #ifdef CONFIG_ARCH_USES_PG_ARCH_3 218 #ifdef CONFIG_ARCH_USES_PG_ARCH_3
219 u |= kpf_copy_bit(k, KPF_ARCH_3, 219 u |= kpf_copy_bit(k, KPF_ARCH_3, PG_arch_3);
220 #endif 220 #endif
221 221
222 return u; 222 return u;
223 }; 223 };
224 224
225 static ssize_t kpageflags_read(struct file *fi 225 static ssize_t kpageflags_read(struct file *file, char __user *buf,
226 size_t count, lof 226 size_t count, loff_t *ppos)
227 { 227 {
228 const unsigned long max_dump_pfn = get 228 const unsigned long max_dump_pfn = get_max_dump_pfn();
229 u64 __user *out = (u64 __user *)buf; 229 u64 __user *out = (u64 __user *)buf;
230 unsigned long src = *ppos; 230 unsigned long src = *ppos;
231 unsigned long pfn; 231 unsigned long pfn;
232 ssize_t ret = 0; 232 ssize_t ret = 0;
233 233
234 pfn = src / KPMSIZE; 234 pfn = src / KPMSIZE;
235 if (src & KPMMASK || count & KPMMASK) 235 if (src & KPMMASK || count & KPMMASK)
236 return -EINVAL; 236 return -EINVAL;
237 if (src >= max_dump_pfn * KPMSIZE) 237 if (src >= max_dump_pfn * KPMSIZE)
238 return 0; 238 return 0;
239 count = min_t(unsigned long, count, (m 239 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src);
240 240
241 while (count > 0) { 241 while (count > 0) {
242 /* 242 /*
243 * TODO: ZONE_DEVICE support r 243 * TODO: ZONE_DEVICE support requires to identify
244 * memmaps that were actually 244 * memmaps that were actually initialized.
245 */ 245 */
246 struct page *page = pfn_to_onl 246 struct page *page = pfn_to_online_page(pfn);
247 247
248 if (put_user(stable_page_flags 248 if (put_user(stable_page_flags(page), out)) {
249 ret = -EFAULT; 249 ret = -EFAULT;
250 break; 250 break;
251 } 251 }
252 252
253 pfn++; 253 pfn++;
254 out++; 254 out++;
255 count -= KPMSIZE; 255 count -= KPMSIZE;
256 256
257 cond_resched(); 257 cond_resched();
258 } 258 }
259 259
260 *ppos += (char __user *)out - buf; 260 *ppos += (char __user *)out - buf;
261 if (!ret) 261 if (!ret)
262 ret = (char __user *)out - buf 262 ret = (char __user *)out - buf;
263 return ret; 263 return ret;
264 } 264 }
265 265
266 static const struct proc_ops kpageflags_proc_o 266 static const struct proc_ops kpageflags_proc_ops = {
267 .proc_flags = PROC_ENTRY_PERMANENT 267 .proc_flags = PROC_ENTRY_PERMANENT,
268 .proc_lseek = mem_lseek, 268 .proc_lseek = mem_lseek,
269 .proc_read = kpageflags_read, 269 .proc_read = kpageflags_read,
270 }; 270 };
271 271
272 #ifdef CONFIG_MEMCG 272 #ifdef CONFIG_MEMCG
273 static ssize_t kpagecgroup_read(struct file *f 273 static ssize_t kpagecgroup_read(struct file *file, char __user *buf,
274 size_t count, 274 size_t count, loff_t *ppos)
275 { 275 {
276 const unsigned long max_dump_pfn = get 276 const unsigned long max_dump_pfn = get_max_dump_pfn();
277 u64 __user *out = (u64 __user *)buf; 277 u64 __user *out = (u64 __user *)buf;
278 struct page *ppage; 278 struct page *ppage;
279 unsigned long src = *ppos; 279 unsigned long src = *ppos;
280 unsigned long pfn; 280 unsigned long pfn;
281 ssize_t ret = 0; 281 ssize_t ret = 0;
282 u64 ino; 282 u64 ino;
283 283
284 pfn = src / KPMSIZE; 284 pfn = src / KPMSIZE;
285 if (src & KPMMASK || count & KPMMASK) 285 if (src & KPMMASK || count & KPMMASK)
286 return -EINVAL; 286 return -EINVAL;
287 if (src >= max_dump_pfn * KPMSIZE) 287 if (src >= max_dump_pfn * KPMSIZE)
288 return 0; 288 return 0;
289 count = min_t(unsigned long, count, (m 289 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src);
290 290
291 while (count > 0) { 291 while (count > 0) {
292 /* 292 /*
293 * TODO: ZONE_DEVICE support r 293 * TODO: ZONE_DEVICE support requires to identify
294 * memmaps that were actually 294 * memmaps that were actually initialized.
295 */ 295 */
296 ppage = pfn_to_online_page(pfn 296 ppage = pfn_to_online_page(pfn);
297 297
298 if (ppage) 298 if (ppage)
299 ino = page_cgroup_ino( 299 ino = page_cgroup_ino(ppage);
300 else 300 else
301 ino = 0; 301 ino = 0;
302 302
303 if (put_user(ino, out)) { 303 if (put_user(ino, out)) {
304 ret = -EFAULT; 304 ret = -EFAULT;
305 break; 305 break;
306 } 306 }
307 307
308 pfn++; 308 pfn++;
309 out++; 309 out++;
310 count -= KPMSIZE; 310 count -= KPMSIZE;
311 311
312 cond_resched(); 312 cond_resched();
313 } 313 }
314 314
315 *ppos += (char __user *)out - buf; 315 *ppos += (char __user *)out - buf;
316 if (!ret) 316 if (!ret)
317 ret = (char __user *)out - buf 317 ret = (char __user *)out - buf;
318 return ret; 318 return ret;
319 } 319 }
320 320
321 static const struct proc_ops kpagecgroup_proc_ 321 static const struct proc_ops kpagecgroup_proc_ops = {
322 .proc_flags = PROC_ENTRY_PERMANENT 322 .proc_flags = PROC_ENTRY_PERMANENT,
323 .proc_lseek = mem_lseek, 323 .proc_lseek = mem_lseek,
324 .proc_read = kpagecgroup_read, 324 .proc_read = kpagecgroup_read,
325 }; 325 };
326 #endif /* CONFIG_MEMCG */ 326 #endif /* CONFIG_MEMCG */
327 327
328 static int __init proc_page_init(void) 328 static int __init proc_page_init(void)
329 { 329 {
330 proc_create("kpagecount", S_IRUSR, NUL 330 proc_create("kpagecount", S_IRUSR, NULL, &kpagecount_proc_ops);
331 proc_create("kpageflags", S_IRUSR, NUL 331 proc_create("kpageflags", S_IRUSR, NULL, &kpageflags_proc_ops);
332 #ifdef CONFIG_MEMCG 332 #ifdef CONFIG_MEMCG
333 proc_create("kpagecgroup", S_IRUSR, NU 333 proc_create("kpagecgroup", S_IRUSR, NULL, &kpagecgroup_proc_ops);
334 #endif 334 #endif
335 return 0; 335 return 0;
336 } 336 }
337 fs_initcall(proc_page_init); 337 fs_initcall(proc_page_init);
338 338
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.