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 counts
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 count.
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 struct page *ppage;
50 unsigned long src = *ppos; 51 unsigned long src = *ppos;
51 unsigned long pfn; 52 unsigned long pfn;
52 ssize_t ret = 0; 53 ssize_t ret = 0;
>> 54 u64 pcount;
53 55
54 pfn = src / KPMSIZE; 56 pfn = src / KPMSIZE;
55 if (src & KPMMASK || count & KPMMASK) 57 if (src & KPMMASK || count & KPMMASK)
56 return -EINVAL; 58 return -EINVAL;
57 if (src >= max_dump_pfn * KPMSIZE) 59 if (src >= max_dump_pfn * KPMSIZE)
58 return 0; 60 return 0;
59 count = min_t(unsigned long, count, (m 61 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src);
60 62
61 while (count > 0) { 63 while (count > 0) {
62 struct page *page; <<
63 u64 mapcount = 0; <<
64 <<
65 /* 64 /*
66 * TODO: ZONE_DEVICE support r 65 * TODO: ZONE_DEVICE support requires to identify
67 * memmaps that were actually 66 * memmaps that were actually initialized.
68 */ 67 */
69 page = pfn_to_online_page(pfn) !! 68 ppage = pfn_to_online_page(pfn);
70 if (page) !! 69
71 mapcount = folio_preci !! 70 if (!ppage)
72 !! 71 pcount = 0;
>> 72 else
>> 73 pcount = page_mapcount(ppage);
73 74
74 if (put_user(mapcount, out)) { !! 75 if (put_user(pcount, out)) {
75 ret = -EFAULT; 76 ret = -EFAULT;
76 break; 77 break;
77 } 78 }
78 79
79 pfn++; 80 pfn++;
80 out++; 81 out++;
81 count -= KPMSIZE; 82 count -= KPMSIZE;
82 83
83 cond_resched(); 84 cond_resched();
84 } 85 }
85 86
86 *ppos += (char __user *)out - buf; 87 *ppos += (char __user *)out - buf;
87 if (!ret) 88 if (!ret)
88 ret = (char __user *)out - buf 89 ret = (char __user *)out - buf;
89 return ret; 90 return ret;
90 } 91 }
91 92
92 static const struct proc_ops kpagecount_proc_o 93 static const struct proc_ops kpagecount_proc_ops = {
93 .proc_flags = PROC_ENTRY_PERMANENT 94 .proc_flags = PROC_ENTRY_PERMANENT,
94 .proc_lseek = mem_lseek, 95 .proc_lseek = mem_lseek,
95 .proc_read = kpagecount_read, 96 .proc_read = kpagecount_read,
96 }; 97 };
97 98
98 /* /proc/kpageflags - an array exposing page f 99 /* /proc/kpageflags - an array exposing page flags
99 * 100 *
100 * Each entry is a u64 representing the corres 101 * Each entry is a u64 representing the corresponding
101 * physical page flags. 102 * physical page flags.
102 */ 103 */
103 104
104 static inline u64 kpf_copy_bit(u64 kflags, int 105 static inline u64 kpf_copy_bit(u64 kflags, int ubit, int kbit)
105 { 106 {
106 return ((kflags >> kbit) & 1) << ubit; 107 return ((kflags >> kbit) & 1) << ubit;
107 } 108 }
108 109
109 u64 stable_page_flags(const struct page *page) 110 u64 stable_page_flags(const struct page *page)
110 { 111 {
111 const struct folio *folio; 112 const struct folio *folio;
112 unsigned long k; 113 unsigned long k;
113 unsigned long mapping; 114 unsigned long mapping;
114 bool is_anon; 115 bool is_anon;
115 u64 u = 0; 116 u64 u = 0;
116 117
117 /* 118 /*
118 * pseudo flag: KPF_NOPAGE 119 * pseudo flag: KPF_NOPAGE
119 * it differentiates a memory hole fro 120 * it differentiates a memory hole from a page with no flags
120 */ 121 */
121 if (!page) 122 if (!page)
122 return 1 << KPF_NOPAGE; 123 return 1 << KPF_NOPAGE;
123 folio = page_folio(page); 124 folio = page_folio(page);
124 125
125 k = folio->flags; 126 k = folio->flags;
126 mapping = (unsigned long)folio->mappin 127 mapping = (unsigned long)folio->mapping;
127 is_anon = mapping & PAGE_MAPPING_ANON; 128 is_anon = mapping & PAGE_MAPPING_ANON;
128 129
129 /* 130 /*
130 * pseudo flags for the well known (an 131 * pseudo flags for the well known (anonymous) memory mapped pages
131 */ 132 */
132 if (page_mapped(page)) 133 if (page_mapped(page))
133 u |= 1 << KPF_MMAP; 134 u |= 1 << KPF_MMAP;
134 if (is_anon) { 135 if (is_anon) {
135 u |= 1 << KPF_ANON; 136 u |= 1 << KPF_ANON;
136 if (mapping & PAGE_MAPPING_KSM 137 if (mapping & PAGE_MAPPING_KSM)
137 u |= 1 << KPF_KSM; 138 u |= 1 << KPF_KSM;
138 } 139 }
139 140
140 /* 141 /*
141 * compound pages: export both head/ta 142 * compound pages: export both head/tail info
142 * they together define a compound pag 143 * they together define a compound page's start/end pos and order
143 */ 144 */
144 if (page == &folio->page) 145 if (page == &folio->page)
145 u |= kpf_copy_bit(k, KPF_COMPO 146 u |= kpf_copy_bit(k, KPF_COMPOUND_HEAD, PG_head);
146 else 147 else
147 u |= 1 << KPF_COMPOUND_TAIL; 148 u |= 1 << KPF_COMPOUND_TAIL;
148 if (folio_test_hugetlb(folio)) 149 if (folio_test_hugetlb(folio))
149 u |= 1 << KPF_HUGE; 150 u |= 1 << KPF_HUGE;
150 else if (folio_test_large(folio) && !! 151 /*
151 folio_test_large_rmappable(fo !! 152 * We need to check PageLRU/PageAnon
152 /* Note: we indicate any THPs !! 153 * to make sure a given page is a thp, not a non-huge compound page.
153 u |= 1 << KPF_THP; !! 154 */
154 } else if (is_huge_zero_folio(folio)) !! 155 else if (folio_test_large(folio)) {
155 u |= 1 << KPF_ZERO_PAGE; !! 156 if ((k & (1 << PG_lru)) || is_anon)
156 u |= 1 << KPF_THP; !! 157 u |= 1 << KPF_THP;
157 } else if (is_zero_folio(folio)) { !! 158 else if (is_huge_zero_folio(folio)) {
>> 159 u |= 1 << KPF_ZERO_PAGE;
>> 160 u |= 1 << KPF_THP;
>> 161 }
>> 162 } else if (is_zero_pfn(page_to_pfn(page)))
158 u |= 1 << KPF_ZERO_PAGE; 163 u |= 1 << KPF_ZERO_PAGE;
159 } <<
160 164
161 /* 165 /*
162 * Caveats on high order pages: PG_bud 166 * Caveats on high order pages: PG_buddy and PG_slab will only be set
163 * on the head page. 167 * on the head page.
164 */ 168 */
165 if (PageBuddy(page)) 169 if (PageBuddy(page))
166 u |= 1 << KPF_BUDDY; 170 u |= 1 << KPF_BUDDY;
167 else if (page_count(page) == 0 && is_f 171 else if (page_count(page) == 0 && is_free_buddy_page(page))
168 u |= 1 << KPF_BUDDY; 172 u |= 1 << KPF_BUDDY;
169 173
170 if (PageOffline(page)) 174 if (PageOffline(page))
171 u |= 1 << KPF_OFFLINE; 175 u |= 1 << KPF_OFFLINE;
172 if (PageTable(page)) 176 if (PageTable(page))
173 u |= 1 << KPF_PGTABLE; 177 u |= 1 << KPF_PGTABLE;
174 if (folio_test_slab(folio)) 178 if (folio_test_slab(folio))
175 u |= 1 << KPF_SLAB; 179 u |= 1 << KPF_SLAB;
176 180
177 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined( 181 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT)
178 u |= kpf_copy_bit(k, KPF_IDLE, 182 u |= kpf_copy_bit(k, KPF_IDLE, PG_idle);
179 #else 183 #else
180 if (folio_test_idle(folio)) 184 if (folio_test_idle(folio))
181 u |= 1 << KPF_IDLE; 185 u |= 1 << KPF_IDLE;
182 #endif 186 #endif
183 187
184 u |= kpf_copy_bit(k, KPF_LOCKED, 188 u |= kpf_copy_bit(k, KPF_LOCKED, PG_locked);
>> 189 u |= kpf_copy_bit(k, KPF_ERROR, PG_error);
185 u |= kpf_copy_bit(k, KPF_DIRTY, 190 u |= kpf_copy_bit(k, KPF_DIRTY, PG_dirty);
186 u |= kpf_copy_bit(k, KPF_UPTODATE, 191 u |= kpf_copy_bit(k, KPF_UPTODATE, PG_uptodate);
187 u |= kpf_copy_bit(k, KPF_WRITEBACK, 192 u |= kpf_copy_bit(k, KPF_WRITEBACK, PG_writeback);
188 193
189 u |= kpf_copy_bit(k, KPF_LRU, 194 u |= kpf_copy_bit(k, KPF_LRU, PG_lru);
190 u |= kpf_copy_bit(k, KPF_REFERENCED, 195 u |= kpf_copy_bit(k, KPF_REFERENCED, PG_referenced);
191 u |= kpf_copy_bit(k, KPF_ACTIVE, 196 u |= kpf_copy_bit(k, KPF_ACTIVE, PG_active);
192 u |= kpf_copy_bit(k, KPF_RECLAIM, 197 u |= kpf_copy_bit(k, KPF_RECLAIM, PG_reclaim);
193 198
194 #define SWAPCACHE ((1 << PG_swapbacked) | (1 < 199 #define SWAPCACHE ((1 << PG_swapbacked) | (1 << PG_swapcache))
195 if ((k & SWAPCACHE) == SWAPCACHE) 200 if ((k & SWAPCACHE) == SWAPCACHE)
196 u |= 1 << KPF_SWAPCACHE; 201 u |= 1 << KPF_SWAPCACHE;
197 u |= kpf_copy_bit(k, KPF_SWAPBACKED, 202 u |= kpf_copy_bit(k, KPF_SWAPBACKED, PG_swapbacked);
198 203
199 u |= kpf_copy_bit(k, KPF_UNEVICTABLE, 204 u |= kpf_copy_bit(k, KPF_UNEVICTABLE, PG_unevictable);
200 u |= kpf_copy_bit(k, KPF_MLOCKED, 205 u |= kpf_copy_bit(k, KPF_MLOCKED, PG_mlocked);
201 206
202 #ifdef CONFIG_MEMORY_FAILURE 207 #ifdef CONFIG_MEMORY_FAILURE
203 if (u & (1 << KPF_HUGE)) 208 if (u & (1 << KPF_HUGE))
204 u |= kpf_copy_bit(k, KPF_HWPOI 209 u |= kpf_copy_bit(k, KPF_HWPOISON, PG_hwpoison);
205 else 210 else
206 u |= kpf_copy_bit(page->flags, 211 u |= kpf_copy_bit(page->flags, KPF_HWPOISON, PG_hwpoison);
207 #endif 212 #endif
208 213
>> 214 #ifdef CONFIG_ARCH_USES_PG_UNCACHED
>> 215 u |= kpf_copy_bit(k, KPF_UNCACHED, PG_uncached);
>> 216 #endif
>> 217
209 u |= kpf_copy_bit(k, KPF_RESERVED, 218 u |= kpf_copy_bit(k, KPF_RESERVED, PG_reserved);
210 u |= kpf_copy_bit(k, KPF_OWNER_2, !! 219 u |= kpf_copy_bit(k, KPF_MAPPEDTODISK, PG_mappedtodisk);
211 u |= kpf_copy_bit(k, KPF_PRIVATE, 220 u |= kpf_copy_bit(k, KPF_PRIVATE, PG_private);
212 u |= kpf_copy_bit(k, KPF_PRIVATE_2, 221 u |= kpf_copy_bit(k, KPF_PRIVATE_2, PG_private_2);
213 u |= kpf_copy_bit(k, KPF_OWNER_PRIVATE 222 u |= kpf_copy_bit(k, KPF_OWNER_PRIVATE, PG_owner_priv_1);
214 u |= kpf_copy_bit(k, KPF_ARCH, 223 u |= kpf_copy_bit(k, KPF_ARCH, PG_arch_1);
215 #ifdef CONFIG_ARCH_USES_PG_ARCH_2 !! 224 #ifdef CONFIG_ARCH_USES_PG_ARCH_X
216 u |= kpf_copy_bit(k, KPF_ARCH_2, 225 u |= kpf_copy_bit(k, KPF_ARCH_2, PG_arch_2);
217 #endif <<
218 #ifdef CONFIG_ARCH_USES_PG_ARCH_3 <<
219 u |= kpf_copy_bit(k, KPF_ARCH_3, 226 u |= kpf_copy_bit(k, KPF_ARCH_3, PG_arch_3);
220 #endif 227 #endif
221 228
222 return u; 229 return u;
223 }; 230 };
224 231
225 static ssize_t kpageflags_read(struct file *fi 232 static ssize_t kpageflags_read(struct file *file, char __user *buf,
226 size_t count, lof 233 size_t count, loff_t *ppos)
227 { 234 {
228 const unsigned long max_dump_pfn = get 235 const unsigned long max_dump_pfn = get_max_dump_pfn();
229 u64 __user *out = (u64 __user *)buf; 236 u64 __user *out = (u64 __user *)buf;
230 unsigned long src = *ppos; 237 unsigned long src = *ppos;
231 unsigned long pfn; 238 unsigned long pfn;
232 ssize_t ret = 0; 239 ssize_t ret = 0;
233 240
234 pfn = src / KPMSIZE; 241 pfn = src / KPMSIZE;
235 if (src & KPMMASK || count & KPMMASK) 242 if (src & KPMMASK || count & KPMMASK)
236 return -EINVAL; 243 return -EINVAL;
237 if (src >= max_dump_pfn * KPMSIZE) 244 if (src >= max_dump_pfn * KPMSIZE)
238 return 0; 245 return 0;
239 count = min_t(unsigned long, count, (m 246 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src);
240 247
241 while (count > 0) { 248 while (count > 0) {
242 /* 249 /*
243 * TODO: ZONE_DEVICE support r 250 * TODO: ZONE_DEVICE support requires to identify
244 * memmaps that were actually 251 * memmaps that were actually initialized.
245 */ 252 */
246 struct page *page = pfn_to_onl 253 struct page *page = pfn_to_online_page(pfn);
247 254
248 if (put_user(stable_page_flags 255 if (put_user(stable_page_flags(page), out)) {
249 ret = -EFAULT; 256 ret = -EFAULT;
250 break; 257 break;
251 } 258 }
252 259
253 pfn++; 260 pfn++;
254 out++; 261 out++;
255 count -= KPMSIZE; 262 count -= KPMSIZE;
256 263
257 cond_resched(); 264 cond_resched();
258 } 265 }
259 266
260 *ppos += (char __user *)out - buf; 267 *ppos += (char __user *)out - buf;
261 if (!ret) 268 if (!ret)
262 ret = (char __user *)out - buf 269 ret = (char __user *)out - buf;
263 return ret; 270 return ret;
264 } 271 }
265 272
266 static const struct proc_ops kpageflags_proc_o 273 static const struct proc_ops kpageflags_proc_ops = {
267 .proc_flags = PROC_ENTRY_PERMANENT 274 .proc_flags = PROC_ENTRY_PERMANENT,
268 .proc_lseek = mem_lseek, 275 .proc_lseek = mem_lseek,
269 .proc_read = kpageflags_read, 276 .proc_read = kpageflags_read,
270 }; 277 };
271 278
272 #ifdef CONFIG_MEMCG 279 #ifdef CONFIG_MEMCG
273 static ssize_t kpagecgroup_read(struct file *f 280 static ssize_t kpagecgroup_read(struct file *file, char __user *buf,
274 size_t count, 281 size_t count, loff_t *ppos)
275 { 282 {
276 const unsigned long max_dump_pfn = get 283 const unsigned long max_dump_pfn = get_max_dump_pfn();
277 u64 __user *out = (u64 __user *)buf; 284 u64 __user *out = (u64 __user *)buf;
278 struct page *ppage; 285 struct page *ppage;
279 unsigned long src = *ppos; 286 unsigned long src = *ppos;
280 unsigned long pfn; 287 unsigned long pfn;
281 ssize_t ret = 0; 288 ssize_t ret = 0;
282 u64 ino; 289 u64 ino;
283 290
284 pfn = src / KPMSIZE; 291 pfn = src / KPMSIZE;
285 if (src & KPMMASK || count & KPMMASK) 292 if (src & KPMMASK || count & KPMMASK)
286 return -EINVAL; 293 return -EINVAL;
287 if (src >= max_dump_pfn * KPMSIZE) 294 if (src >= max_dump_pfn * KPMSIZE)
288 return 0; 295 return 0;
289 count = min_t(unsigned long, count, (m 296 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src);
290 297
291 while (count > 0) { 298 while (count > 0) {
292 /* 299 /*
293 * TODO: ZONE_DEVICE support r 300 * TODO: ZONE_DEVICE support requires to identify
294 * memmaps that were actually 301 * memmaps that were actually initialized.
295 */ 302 */
296 ppage = pfn_to_online_page(pfn 303 ppage = pfn_to_online_page(pfn);
297 304
298 if (ppage) 305 if (ppage)
299 ino = page_cgroup_ino( 306 ino = page_cgroup_ino(ppage);
300 else 307 else
301 ino = 0; 308 ino = 0;
302 309
303 if (put_user(ino, out)) { 310 if (put_user(ino, out)) {
304 ret = -EFAULT; 311 ret = -EFAULT;
305 break; 312 break;
306 } 313 }
307 314
308 pfn++; 315 pfn++;
309 out++; 316 out++;
310 count -= KPMSIZE; 317 count -= KPMSIZE;
311 318
312 cond_resched(); 319 cond_resched();
313 } 320 }
314 321
315 *ppos += (char __user *)out - buf; 322 *ppos += (char __user *)out - buf;
316 if (!ret) 323 if (!ret)
317 ret = (char __user *)out - buf 324 ret = (char __user *)out - buf;
318 return ret; 325 return ret;
319 } 326 }
320 327
321 static const struct proc_ops kpagecgroup_proc_ 328 static const struct proc_ops kpagecgroup_proc_ops = {
322 .proc_flags = PROC_ENTRY_PERMANENT 329 .proc_flags = PROC_ENTRY_PERMANENT,
323 .proc_lseek = mem_lseek, 330 .proc_lseek = mem_lseek,
324 .proc_read = kpagecgroup_read, 331 .proc_read = kpagecgroup_read,
325 }; 332 };
326 #endif /* CONFIG_MEMCG */ 333 #endif /* CONFIG_MEMCG */
327 334
328 static int __init proc_page_init(void) 335 static int __init proc_page_init(void)
329 { 336 {
330 proc_create("kpagecount", S_IRUSR, NUL 337 proc_create("kpagecount", S_IRUSR, NULL, &kpagecount_proc_ops);
331 proc_create("kpageflags", S_IRUSR, NUL 338 proc_create("kpageflags", S_IRUSR, NULL, &kpageflags_proc_ops);
332 #ifdef CONFIG_MEMCG 339 #ifdef CONFIG_MEMCG
333 proc_create("kpagecgroup", S_IRUSR, NU 340 proc_create("kpagecgroup", S_IRUSR, NULL, &kpagecgroup_proc_ops);
334 #endif 341 #endif
335 return 0; 342 return 0;
336 } 343 }
337 fs_initcall(proc_page_init); 344 fs_initcall(proc_page_init);
338 345
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.