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) <<
71 mapcount = folio_preci <<
72 <<
73 69
74 if (put_user(mapcount, out)) { !! 70 if (!ppage || PageSlab(ppage) || page_has_type(ppage))
>> 71 pcount = 0;
>> 72 else
>> 73 pcount = page_mapcount(ppage);
>> 74
>> 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_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(struct page *page)
110 { 110 {
111 const struct folio *folio; !! 111 u64 k;
112 unsigned long k; !! 112 u64 u;
113 unsigned long mapping; <<
114 bool is_anon; <<
115 u64 u = 0; <<
116 113
117 /* 114 /*
118 * pseudo flag: KPF_NOPAGE 115 * pseudo flag: KPF_NOPAGE
119 * it differentiates a memory hole fro 116 * it differentiates a memory hole from a page with no flags
120 */ 117 */
121 if (!page) 118 if (!page)
122 return 1 << KPF_NOPAGE; 119 return 1 << KPF_NOPAGE;
123 folio = page_folio(page); <<
124 120
125 k = folio->flags; !! 121 k = page->flags;
126 mapping = (unsigned long)folio->mappin !! 122 u = 0;
127 is_anon = mapping & PAGE_MAPPING_ANON; <<
128 123
129 /* 124 /*
130 * pseudo flags for the well known (an 125 * pseudo flags for the well known (anonymous) memory mapped pages
>> 126 *
>> 127 * Note that page->_mapcount is overloaded in SLOB/SLUB/SLQB, so the
>> 128 * simple test in page_mapped() is not enough.
131 */ 129 */
132 if (page_mapped(page)) !! 130 if (!PageSlab(page) && page_mapped(page))
133 u |= 1 << KPF_MMAP; 131 u |= 1 << KPF_MMAP;
134 if (is_anon) { !! 132 if (PageAnon(page))
135 u |= 1 << KPF_ANON; 133 u |= 1 << KPF_ANON;
136 if (mapping & PAGE_MAPPING_KSM !! 134 if (PageKsm(page))
137 u |= 1 << KPF_KSM; !! 135 u |= 1 << KPF_KSM;
138 } <<
139 136
140 /* 137 /*
141 * compound pages: export both head/ta 138 * compound pages: export both head/tail info
142 * they together define a compound pag 139 * they together define a compound page's start/end pos and order
143 */ 140 */
144 if (page == &folio->page) !! 141 if (PageHead(page))
145 u |= kpf_copy_bit(k, KPF_COMPO !! 142 u |= 1 << KPF_COMPOUND_HEAD;
146 else !! 143 if (PageTail(page))
147 u |= 1 << KPF_COMPOUND_TAIL; 144 u |= 1 << KPF_COMPOUND_TAIL;
148 if (folio_test_hugetlb(folio)) !! 145 if (PageHuge(page))
149 u |= 1 << KPF_HUGE; 146 u |= 1 << KPF_HUGE;
150 else if (folio_test_large(folio) && !! 147 /*
151 folio_test_large_rmappable(fo !! 148 * PageTransCompound can be true for non-huge compound pages (slab
152 /* Note: we indicate any THPs !! 149 * pages or pages allocated by drivers with __GFP_COMP) because it
153 u |= 1 << KPF_THP; !! 150 * just checks PG_head/PG_tail, so we need to check PageLRU/PageAnon
154 } else if (is_huge_zero_folio(folio)) !! 151 * to make sure a given page is a thp, not a non-huge compound page.
155 u |= 1 << KPF_ZERO_PAGE; !! 152 */
156 u |= 1 << KPF_THP; !! 153 else if (PageTransCompound(page)) {
157 } else if (is_zero_folio(folio)) { !! 154 struct page *head = compound_head(page);
>> 155
>> 156 if (PageLRU(head) || PageAnon(head))
>> 157 u |= 1 << KPF_THP;
>> 158 else if (is_huge_zero_page(head)) {
>> 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 } !! 164
160 165
161 /* 166 /*
162 * Caveats on high order pages: PG_bud !! 167 * Caveats on high order pages: page->_refcount will only be set
163 * on the head page. !! 168 * -1 on the head page; SLUB/SLQB do the same for PG_slab;
>> 169 * SLOB won't set PG_slab at all on compound pages.
164 */ 170 */
165 if (PageBuddy(page)) 171 if (PageBuddy(page))
166 u |= 1 << KPF_BUDDY; 172 u |= 1 << KPF_BUDDY;
167 else if (page_count(page) == 0 && is_f 173 else if (page_count(page) == 0 && is_free_buddy_page(page))
168 u |= 1 << KPF_BUDDY; 174 u |= 1 << KPF_BUDDY;
169 175
170 if (PageOffline(page)) 176 if (PageOffline(page))
171 u |= 1 << KPF_OFFLINE; 177 u |= 1 << KPF_OFFLINE;
172 if (PageTable(page)) 178 if (PageTable(page))
173 u |= 1 << KPF_PGTABLE; 179 u |= 1 << KPF_PGTABLE;
174 if (folio_test_slab(folio)) <<
175 u |= 1 << KPF_SLAB; <<
176 180
177 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined( !! 181 if (page_is_idle(page))
178 u |= kpf_copy_bit(k, KPF_IDLE, <<
179 #else <<
180 if (folio_test_idle(folio)) <<
181 u |= 1 << KPF_IDLE; 182 u |= 1 << KPF_IDLE;
182 #endif <<
183 183
184 u |= kpf_copy_bit(k, KPF_LOCKED, 184 u |= kpf_copy_bit(k, KPF_LOCKED, PG_locked);
>> 185
>> 186 u |= kpf_copy_bit(k, KPF_SLAB, PG_slab);
>> 187 if (PageTail(page) && PageSlab(compound_head(page)))
>> 188 u |= 1 << KPF_SLAB;
>> 189
>> 190 u |= kpf_copy_bit(k, KPF_ERROR, PG_error);
185 u |= kpf_copy_bit(k, KPF_DIRTY, 191 u |= kpf_copy_bit(k, KPF_DIRTY, PG_dirty);
186 u |= kpf_copy_bit(k, KPF_UPTODATE, 192 u |= kpf_copy_bit(k, KPF_UPTODATE, PG_uptodate);
187 u |= kpf_copy_bit(k, KPF_WRITEBACK, 193 u |= kpf_copy_bit(k, KPF_WRITEBACK, PG_writeback);
188 194
189 u |= kpf_copy_bit(k, KPF_LRU, 195 u |= kpf_copy_bit(k, KPF_LRU, PG_lru);
190 u |= kpf_copy_bit(k, KPF_REFERENCED, 196 u |= kpf_copy_bit(k, KPF_REFERENCED, PG_referenced);
191 u |= kpf_copy_bit(k, KPF_ACTIVE, 197 u |= kpf_copy_bit(k, KPF_ACTIVE, PG_active);
192 u |= kpf_copy_bit(k, KPF_RECLAIM, 198 u |= kpf_copy_bit(k, KPF_RECLAIM, PG_reclaim);
193 199
194 #define SWAPCACHE ((1 << PG_swapbacked) | (1 < !! 200 if (PageSwapCache(page))
195 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 u |= kpf_copy_bit(k, KPF_HWPOISON, PG_hwpoison);
204 u |= kpf_copy_bit(k, KPF_HWPOI !! 209 #endif
205 else !! 210
206 u |= kpf_copy_bit(page->flags, !! 211 #ifdef CONFIG_ARCH_USES_PG_UNCACHED
>> 212 u |= kpf_copy_bit(k, KPF_UNCACHED, PG_uncached);
207 #endif 213 #endif
208 214
209 u |= kpf_copy_bit(k, KPF_RESERVED, 215 u |= kpf_copy_bit(k, KPF_RESERVED, PG_reserved);
210 u |= kpf_copy_bit(k, KPF_OWNER_2, !! 216 u |= kpf_copy_bit(k, KPF_MAPPEDTODISK, PG_mappedtodisk);
211 u |= kpf_copy_bit(k, KPF_PRIVATE, 217 u |= kpf_copy_bit(k, KPF_PRIVATE, PG_private);
212 u |= kpf_copy_bit(k, KPF_PRIVATE_2, 218 u |= kpf_copy_bit(k, KPF_PRIVATE_2, PG_private_2);
213 u |= kpf_copy_bit(k, KPF_OWNER_PRIVATE 219 u |= kpf_copy_bit(k, KPF_OWNER_PRIVATE, PG_owner_priv_1);
214 u |= kpf_copy_bit(k, KPF_ARCH, 220 u |= kpf_copy_bit(k, KPF_ARCH, PG_arch_1);
215 #ifdef CONFIG_ARCH_USES_PG_ARCH_2 !! 221 #ifdef CONFIG_64BIT
216 u |= kpf_copy_bit(k, KPF_ARCH_2, 222 u |= kpf_copy_bit(k, KPF_ARCH_2, PG_arch_2);
217 #endif 223 #endif
218 #ifdef CONFIG_ARCH_USES_PG_ARCH_3 <<
219 u |= kpf_copy_bit(k, KPF_ARCH_3, <<
220 #endif <<
221 224
222 return u; 225 return u;
223 }; 226 };
224 227
225 static ssize_t kpageflags_read(struct file *fi 228 static ssize_t kpageflags_read(struct file *file, char __user *buf,
226 size_t count, lof 229 size_t count, loff_t *ppos)
227 { 230 {
228 const unsigned long max_dump_pfn = get 231 const unsigned long max_dump_pfn = get_max_dump_pfn();
229 u64 __user *out = (u64 __user *)buf; 232 u64 __user *out = (u64 __user *)buf;
>> 233 struct page *ppage;
230 unsigned long src = *ppos; 234 unsigned long src = *ppos;
231 unsigned long pfn; 235 unsigned long pfn;
232 ssize_t ret = 0; 236 ssize_t ret = 0;
233 237
234 pfn = src / KPMSIZE; 238 pfn = src / KPMSIZE;
235 if (src & KPMMASK || count & KPMMASK) 239 if (src & KPMMASK || count & KPMMASK)
236 return -EINVAL; 240 return -EINVAL;
237 if (src >= max_dump_pfn * KPMSIZE) 241 if (src >= max_dump_pfn * KPMSIZE)
238 return 0; 242 return 0;
239 count = min_t(unsigned long, count, (m 243 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src);
240 244
241 while (count > 0) { 245 while (count > 0) {
242 /* 246 /*
243 * TODO: ZONE_DEVICE support r 247 * TODO: ZONE_DEVICE support requires to identify
244 * memmaps that were actually 248 * memmaps that were actually initialized.
245 */ 249 */
246 struct page *page = pfn_to_onl !! 250 ppage = pfn_to_online_page(pfn);
247 251
248 if (put_user(stable_page_flags !! 252 if (put_user(stable_page_flags(ppage), out)) {
249 ret = -EFAULT; 253 ret = -EFAULT;
250 break; 254 break;
251 } 255 }
252 256
253 pfn++; 257 pfn++;
254 out++; 258 out++;
255 count -= KPMSIZE; 259 count -= KPMSIZE;
256 260
257 cond_resched(); 261 cond_resched();
258 } 262 }
259 263
260 *ppos += (char __user *)out - buf; 264 *ppos += (char __user *)out - buf;
261 if (!ret) 265 if (!ret)
262 ret = (char __user *)out - buf 266 ret = (char __user *)out - buf;
263 return ret; 267 return ret;
264 } 268 }
265 269
266 static const struct proc_ops kpageflags_proc_o 270 static const struct proc_ops kpageflags_proc_ops = {
267 .proc_flags = PROC_ENTRY_PERMANENT <<
268 .proc_lseek = mem_lseek, 271 .proc_lseek = mem_lseek,
269 .proc_read = kpageflags_read, 272 .proc_read = kpageflags_read,
270 }; 273 };
271 274
272 #ifdef CONFIG_MEMCG 275 #ifdef CONFIG_MEMCG
273 static ssize_t kpagecgroup_read(struct file *f 276 static ssize_t kpagecgroup_read(struct file *file, char __user *buf,
274 size_t count, 277 size_t count, loff_t *ppos)
275 { 278 {
276 const unsigned long max_dump_pfn = get 279 const unsigned long max_dump_pfn = get_max_dump_pfn();
277 u64 __user *out = (u64 __user *)buf; 280 u64 __user *out = (u64 __user *)buf;
278 struct page *ppage; 281 struct page *ppage;
279 unsigned long src = *ppos; 282 unsigned long src = *ppos;
280 unsigned long pfn; 283 unsigned long pfn;
281 ssize_t ret = 0; 284 ssize_t ret = 0;
282 u64 ino; 285 u64 ino;
283 286
284 pfn = src / KPMSIZE; 287 pfn = src / KPMSIZE;
285 if (src & KPMMASK || count & KPMMASK) 288 if (src & KPMMASK || count & KPMMASK)
286 return -EINVAL; 289 return -EINVAL;
287 if (src >= max_dump_pfn * KPMSIZE) 290 if (src >= max_dump_pfn * KPMSIZE)
288 return 0; 291 return 0;
289 count = min_t(unsigned long, count, (m 292 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src);
290 293
291 while (count > 0) { 294 while (count > 0) {
292 /* 295 /*
293 * TODO: ZONE_DEVICE support r 296 * TODO: ZONE_DEVICE support requires to identify
294 * memmaps that were actually 297 * memmaps that were actually initialized.
295 */ 298 */
296 ppage = pfn_to_online_page(pfn 299 ppage = pfn_to_online_page(pfn);
297 300
298 if (ppage) 301 if (ppage)
299 ino = page_cgroup_ino( 302 ino = page_cgroup_ino(ppage);
300 else 303 else
301 ino = 0; 304 ino = 0;
302 305
303 if (put_user(ino, out)) { 306 if (put_user(ino, out)) {
304 ret = -EFAULT; 307 ret = -EFAULT;
305 break; 308 break;
306 } 309 }
307 310
308 pfn++; 311 pfn++;
309 out++; 312 out++;
310 count -= KPMSIZE; 313 count -= KPMSIZE;
311 314
312 cond_resched(); 315 cond_resched();
313 } 316 }
314 317
315 *ppos += (char __user *)out - buf; 318 *ppos += (char __user *)out - buf;
316 if (!ret) 319 if (!ret)
317 ret = (char __user *)out - buf 320 ret = (char __user *)out - buf;
318 return ret; 321 return ret;
319 } 322 }
320 323
321 static const struct proc_ops kpagecgroup_proc_ 324 static const struct proc_ops kpagecgroup_proc_ops = {
322 .proc_flags = PROC_ENTRY_PERMANENT <<
323 .proc_lseek = mem_lseek, 325 .proc_lseek = mem_lseek,
324 .proc_read = kpagecgroup_read, 326 .proc_read = kpagecgroup_read,
325 }; 327 };
326 #endif /* CONFIG_MEMCG */ 328 #endif /* CONFIG_MEMCG */
327 329
328 static int __init proc_page_init(void) 330 static int __init proc_page_init(void)
329 { 331 {
330 proc_create("kpagecount", S_IRUSR, NUL 332 proc_create("kpagecount", S_IRUSR, NULL, &kpagecount_proc_ops);
331 proc_create("kpageflags", S_IRUSR, NUL 333 proc_create("kpageflags", S_IRUSR, NULL, &kpageflags_proc_ops);
332 #ifdef CONFIG_MEMCG 334 #ifdef CONFIG_MEMCG
333 proc_create("kpagecgroup", S_IRUSR, NU 335 proc_create("kpagecgroup", S_IRUSR, NULL, &kpagecgroup_proc_ops);
334 #endif 336 #endif
335 return 0; 337 return 0;
336 } 338 }
337 fs_initcall(proc_page_init); 339 fs_initcall(proc_page_init);
338 340
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.