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> <<
14 #include <linux/memcontrol.h> 13 #include <linux/memcontrol.h>
15 #include <linux/mmu_notifier.h> 14 #include <linux/mmu_notifier.h>
16 #include <linux/page_idle.h> 15 #include <linux/page_idle.h>
17 #include <linux/kernel-page-flags.h> 16 #include <linux/kernel-page-flags.h>
18 #include <linux/uaccess.h> 17 #include <linux/uaccess.h>
19 #include "internal.h" 18 #include "internal.h"
20 19
21 #define KPMSIZE sizeof(u64) 20 #define KPMSIZE sizeof(u64)
22 #define KPMMASK (KPMSIZE - 1) 21 #define KPMMASK (KPMSIZE - 1)
23 #define KPMBITS (KPMSIZE * BITS_PER_BYTE) 22 #define KPMBITS (KPMSIZE * BITS_PER_BYTE)
24 23
25 static inline unsigned long get_max_dump_pfn(v 24 static inline unsigned long get_max_dump_pfn(void)
26 { 25 {
27 #ifdef CONFIG_SPARSEMEM 26 #ifdef CONFIG_SPARSEMEM
28 /* 27 /*
29 * The memmap of early sections is com 28 * The memmap of early sections is completely populated and marked
30 * online even if max_pfn does not fal 29 * online even if max_pfn does not fall on a section boundary -
31 * pfn_to_online_page() will succeed o 30 * pfn_to_online_page() will succeed on all pages. Allow inspecting
32 * these memmaps. 31 * these memmaps.
33 */ 32 */
34 return round_up(max_pfn, PAGES_PER_SEC 33 return round_up(max_pfn, PAGES_PER_SECTION);
35 #else 34 #else
36 return max_pfn; 35 return max_pfn;
37 #endif 36 #endif
38 } 37 }
39 38
40 /* /proc/kpagecount - an array exposing page m !! 39 /* /proc/kpagecount - an array exposing page counts
41 * 40 *
42 * Each entry is a u64 representing the corres 41 * Each entry is a u64 representing the corresponding
43 * physical page mapcount. !! 42 * physical page count.
44 */ 43 */
45 static ssize_t kpagecount_read(struct file *fi 44 static ssize_t kpagecount_read(struct file *file, char __user *buf,
46 size_t count, lof 45 size_t count, loff_t *ppos)
47 { 46 {
48 const unsigned long max_dump_pfn = get 47 const unsigned long max_dump_pfn = get_max_dump_pfn();
49 u64 __user *out = (u64 __user *)buf; 48 u64 __user *out = (u64 __user *)buf;
>> 49 struct page *ppage;
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 u64 pcount;
53 54
54 pfn = src / KPMSIZE; 55 pfn = src / KPMSIZE;
55 if (src & KPMMASK || count & KPMMASK) 56 if (src & KPMMASK || count & KPMMASK)
56 return -EINVAL; 57 return -EINVAL;
57 if (src >= max_dump_pfn * KPMSIZE) 58 if (src >= max_dump_pfn * KPMSIZE)
58 return 0; 59 return 0;
59 count = min_t(unsigned long, count, (m 60 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src);
60 61
61 while (count > 0) { 62 while (count > 0) {
62 struct page *page; <<
63 u64 mapcount = 0; <<
64 <<
65 /* 63 /*
66 * TODO: ZONE_DEVICE support r 64 * TODO: ZONE_DEVICE support requires to identify
67 * memmaps that were actually 65 * memmaps that were actually initialized.
68 */ 66 */
69 page = pfn_to_online_page(pfn) !! 67 ppage = pfn_to_online_page(pfn);
70 if (page) <<
71 mapcount = folio_preci <<
72 <<
73 68
74 if (put_user(mapcount, out)) { !! 69 if (!ppage || PageSlab(ppage) || page_has_type(ppage))
>> 70 pcount = 0;
>> 71 else
>> 72 pcount = page_mapcount(ppage);
>> 73
>> 74 if (put_user(pcount, 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 <<
94 .proc_lseek = mem_lseek, 93 .proc_lseek = mem_lseek,
95 .proc_read = kpagecount_read, 94 .proc_read = kpagecount_read,
96 }; 95 };
97 96
98 /* /proc/kpageflags - an array exposing page f 97 /* /proc/kpageflags - an array exposing page flags
99 * 98 *
100 * Each entry is a u64 representing the corres 99 * Each entry is a u64 representing the corresponding
101 * physical page flags. 100 * physical page flags.
102 */ 101 */
103 102
104 static inline u64 kpf_copy_bit(u64 kflags, int 103 static inline u64 kpf_copy_bit(u64 kflags, int ubit, int kbit)
105 { 104 {
106 return ((kflags >> kbit) & 1) << ubit; 105 return ((kflags >> kbit) & 1) << ubit;
107 } 106 }
108 107
109 u64 stable_page_flags(const struct page *page) !! 108 u64 stable_page_flags(struct page *page)
110 { 109 {
111 const struct folio *folio; !! 110 u64 k;
112 unsigned long k; !! 111 u64 u;
113 unsigned long mapping; <<
114 bool is_anon; <<
115 u64 u = 0; <<
116 112
117 /* 113 /*
118 * pseudo flag: KPF_NOPAGE 114 * pseudo flag: KPF_NOPAGE
119 * it differentiates a memory hole fro 115 * it differentiates a memory hole from a page with no flags
120 */ 116 */
121 if (!page) 117 if (!page)
122 return 1 << KPF_NOPAGE; 118 return 1 << KPF_NOPAGE;
123 folio = page_folio(page); <<
124 119
125 k = folio->flags; !! 120 k = page->flags;
126 mapping = (unsigned long)folio->mappin !! 121 u = 0;
127 is_anon = mapping & PAGE_MAPPING_ANON; <<
128 122
129 /* 123 /*
130 * pseudo flags for the well known (an 124 * pseudo flags for the well known (anonymous) memory mapped pages
>> 125 *
>> 126 * Note that page->_mapcount is overloaded in SLOB/SLUB/SLQB, so the
>> 127 * simple test in page_mapped() is not enough.
131 */ 128 */
132 if (page_mapped(page)) !! 129 if (!PageSlab(page) && page_mapped(page))
133 u |= 1 << KPF_MMAP; 130 u |= 1 << KPF_MMAP;
134 if (is_anon) { !! 131 if (PageAnon(page))
135 u |= 1 << KPF_ANON; 132 u |= 1 << KPF_ANON;
136 if (mapping & PAGE_MAPPING_KSM !! 133 if (PageKsm(page))
137 u |= 1 << KPF_KSM; !! 134 u |= 1 << KPF_KSM;
138 } <<
139 135
140 /* 136 /*
141 * compound pages: export both head/ta 137 * compound pages: export both head/tail info
142 * they together define a compound pag 138 * they together define a compound page's start/end pos and order
143 */ 139 */
144 if (page == &folio->page) !! 140 if (PageHead(page))
145 u |= kpf_copy_bit(k, KPF_COMPO !! 141 u |= 1 << KPF_COMPOUND_HEAD;
146 else !! 142 if (PageTail(page))
147 u |= 1 << KPF_COMPOUND_TAIL; 143 u |= 1 << KPF_COMPOUND_TAIL;
148 if (folio_test_hugetlb(folio)) !! 144 if (PageHuge(page))
149 u |= 1 << KPF_HUGE; 145 u |= 1 << KPF_HUGE;
150 else if (folio_test_large(folio) && !! 146 /*
151 folio_test_large_rmappable(fo !! 147 * PageTransCompound can be true for non-huge compound pages (slab
152 /* Note: we indicate any THPs !! 148 * pages or pages allocated by drivers with __GFP_COMP) because it
153 u |= 1 << KPF_THP; !! 149 * just checks PG_head/PG_tail, so we need to check PageLRU/PageAnon
154 } else if (is_huge_zero_folio(folio)) !! 150 * to make sure a given page is a thp, not a non-huge compound page.
155 u |= 1 << KPF_ZERO_PAGE; !! 151 */
156 u |= 1 << KPF_THP; !! 152 else if (PageTransCompound(page)) {
157 } else if (is_zero_folio(folio)) { !! 153 struct page *head = compound_head(page);
>> 154
>> 155 if (PageLRU(head) || PageAnon(head))
>> 156 u |= 1 << KPF_THP;
>> 157 else if (is_huge_zero_page(head)) {
>> 158 u |= 1 << KPF_ZERO_PAGE;
>> 159 u |= 1 << KPF_THP;
>> 160 }
>> 161 } else if (is_zero_pfn(page_to_pfn(page)))
158 u |= 1 << KPF_ZERO_PAGE; 162 u |= 1 << KPF_ZERO_PAGE;
159 } !! 163
160 164
161 /* 165 /*
162 * Caveats on high order pages: PG_bud !! 166 * Caveats on high order pages: page->_refcount will only be set
163 * on the head page. !! 167 * -1 on the head page; SLUB/SLQB do the same for PG_slab;
>> 168 * SLOB won't set PG_slab at all on compound pages.
164 */ 169 */
165 if (PageBuddy(page)) 170 if (PageBuddy(page))
166 u |= 1 << KPF_BUDDY; 171 u |= 1 << KPF_BUDDY;
167 else if (page_count(page) == 0 && is_f 172 else if (page_count(page) == 0 && is_free_buddy_page(page))
168 u |= 1 << KPF_BUDDY; 173 u |= 1 << KPF_BUDDY;
169 174
170 if (PageOffline(page)) 175 if (PageOffline(page))
171 u |= 1 << KPF_OFFLINE; 176 u |= 1 << KPF_OFFLINE;
172 if (PageTable(page)) 177 if (PageTable(page))
173 u |= 1 << KPF_PGTABLE; 178 u |= 1 << KPF_PGTABLE;
174 if (folio_test_slab(folio)) <<
175 u |= 1 << KPF_SLAB; <<
176 179
177 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined( !! 180 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; 181 u |= 1 << KPF_IDLE;
182 #endif <<
183 182
184 u |= kpf_copy_bit(k, KPF_LOCKED, 183 u |= kpf_copy_bit(k, KPF_LOCKED, PG_locked);
>> 184
>> 185 u |= kpf_copy_bit(k, KPF_SLAB, PG_slab);
>> 186 if (PageTail(page) && PageSlab(compound_head(page)))
>> 187 u |= 1 << KPF_SLAB;
>> 188
>> 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 if (PageSwapCache(page))
195 if ((k & SWAPCACHE) == SWAPCACHE) <<
196 u |= 1 << KPF_SWAPCACHE; 200 u |= 1 << KPF_SWAPCACHE;
197 u |= kpf_copy_bit(k, KPF_SWAPBACKED, 201 u |= kpf_copy_bit(k, KPF_SWAPBACKED, PG_swapbacked);
198 202
199 u |= kpf_copy_bit(k, KPF_UNEVICTABLE, 203 u |= kpf_copy_bit(k, KPF_UNEVICTABLE, PG_unevictable);
200 u |= kpf_copy_bit(k, KPF_MLOCKED, 204 u |= kpf_copy_bit(k, KPF_MLOCKED, PG_mlocked);
201 205
202 #ifdef CONFIG_MEMORY_FAILURE 206 #ifdef CONFIG_MEMORY_FAILURE
203 if (u & (1 << KPF_HUGE)) !! 207 u |= kpf_copy_bit(k, KPF_HWPOISON, PG_hwpoison);
204 u |= kpf_copy_bit(k, KPF_HWPOI !! 208 #endif
205 else !! 209
206 u |= kpf_copy_bit(page->flags, !! 210 #ifdef CONFIG_ARCH_USES_PG_UNCACHED
>> 211 u |= kpf_copy_bit(k, KPF_UNCACHED, PG_uncached);
207 #endif 212 #endif
208 213
209 u |= kpf_copy_bit(k, KPF_RESERVED, 214 u |= kpf_copy_bit(k, KPF_RESERVED, PG_reserved);
210 u |= kpf_copy_bit(k, KPF_OWNER_2, !! 215 u |= kpf_copy_bit(k, KPF_MAPPEDTODISK, PG_mappedtodisk);
211 u |= kpf_copy_bit(k, KPF_PRIVATE, 216 u |= kpf_copy_bit(k, KPF_PRIVATE, PG_private);
212 u |= kpf_copy_bit(k, KPF_PRIVATE_2, 217 u |= kpf_copy_bit(k, KPF_PRIVATE_2, PG_private_2);
213 u |= kpf_copy_bit(k, KPF_OWNER_PRIVATE 218 u |= kpf_copy_bit(k, KPF_OWNER_PRIVATE, PG_owner_priv_1);
214 u |= kpf_copy_bit(k, KPF_ARCH, 219 u |= kpf_copy_bit(k, KPF_ARCH, PG_arch_1);
215 #ifdef CONFIG_ARCH_USES_PG_ARCH_2 !! 220 #ifdef CONFIG_64BIT
216 u |= kpf_copy_bit(k, KPF_ARCH_2, 221 u |= kpf_copy_bit(k, KPF_ARCH_2, PG_arch_2);
217 #endif 222 #endif
218 #ifdef CONFIG_ARCH_USES_PG_ARCH_3 <<
219 u |= kpf_copy_bit(k, KPF_ARCH_3, <<
220 #endif <<
221 223
222 return u; 224 return u;
223 }; 225 };
224 226
225 static ssize_t kpageflags_read(struct file *fi 227 static ssize_t kpageflags_read(struct file *file, char __user *buf,
226 size_t count, lof 228 size_t count, loff_t *ppos)
227 { 229 {
228 const unsigned long max_dump_pfn = get 230 const unsigned long max_dump_pfn = get_max_dump_pfn();
229 u64 __user *out = (u64 __user *)buf; 231 u64 __user *out = (u64 __user *)buf;
>> 232 struct page *ppage;
230 unsigned long src = *ppos; 233 unsigned long src = *ppos;
231 unsigned long pfn; 234 unsigned long pfn;
232 ssize_t ret = 0; 235 ssize_t ret = 0;
233 236
234 pfn = src / KPMSIZE; 237 pfn = src / KPMSIZE;
235 if (src & KPMMASK || count & KPMMASK) 238 if (src & KPMMASK || count & KPMMASK)
236 return -EINVAL; 239 return -EINVAL;
237 if (src >= max_dump_pfn * KPMSIZE) 240 if (src >= max_dump_pfn * KPMSIZE)
238 return 0; 241 return 0;
239 count = min_t(unsigned long, count, (m 242 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src);
240 243
241 while (count > 0) { 244 while (count > 0) {
242 /* 245 /*
243 * TODO: ZONE_DEVICE support r 246 * TODO: ZONE_DEVICE support requires to identify
244 * memmaps that were actually 247 * memmaps that were actually initialized.
245 */ 248 */
246 struct page *page = pfn_to_onl !! 249 ppage = pfn_to_online_page(pfn);
247 250
248 if (put_user(stable_page_flags !! 251 if (put_user(stable_page_flags(ppage), out)) {
249 ret = -EFAULT; 252 ret = -EFAULT;
250 break; 253 break;
251 } 254 }
252 255
253 pfn++; 256 pfn++;
254 out++; 257 out++;
255 count -= KPMSIZE; 258 count -= KPMSIZE;
256 259
257 cond_resched(); 260 cond_resched();
258 } 261 }
259 262
260 *ppos += (char __user *)out - buf; 263 *ppos += (char __user *)out - buf;
261 if (!ret) 264 if (!ret)
262 ret = (char __user *)out - buf 265 ret = (char __user *)out - buf;
263 return ret; 266 return ret;
264 } 267 }
265 268
266 static const struct proc_ops kpageflags_proc_o 269 static const struct proc_ops kpageflags_proc_ops = {
267 .proc_flags = PROC_ENTRY_PERMANENT <<
268 .proc_lseek = mem_lseek, 270 .proc_lseek = mem_lseek,
269 .proc_read = kpageflags_read, 271 .proc_read = kpageflags_read,
270 }; 272 };
271 273
272 #ifdef CONFIG_MEMCG 274 #ifdef CONFIG_MEMCG
273 static ssize_t kpagecgroup_read(struct file *f 275 static ssize_t kpagecgroup_read(struct file *file, char __user *buf,
274 size_t count, 276 size_t count, loff_t *ppos)
275 { 277 {
276 const unsigned long max_dump_pfn = get 278 const unsigned long max_dump_pfn = get_max_dump_pfn();
277 u64 __user *out = (u64 __user *)buf; 279 u64 __user *out = (u64 __user *)buf;
278 struct page *ppage; 280 struct page *ppage;
279 unsigned long src = *ppos; 281 unsigned long src = *ppos;
280 unsigned long pfn; 282 unsigned long pfn;
281 ssize_t ret = 0; 283 ssize_t ret = 0;
282 u64 ino; 284 u64 ino;
283 285
284 pfn = src / KPMSIZE; 286 pfn = src / KPMSIZE;
285 if (src & KPMMASK || count & KPMMASK) 287 if (src & KPMMASK || count & KPMMASK)
286 return -EINVAL; 288 return -EINVAL;
287 if (src >= max_dump_pfn * KPMSIZE) 289 if (src >= max_dump_pfn * KPMSIZE)
288 return 0; 290 return 0;
289 count = min_t(unsigned long, count, (m 291 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src);
290 292
291 while (count > 0) { 293 while (count > 0) {
292 /* 294 /*
293 * TODO: ZONE_DEVICE support r 295 * TODO: ZONE_DEVICE support requires to identify
294 * memmaps that were actually 296 * memmaps that were actually initialized.
295 */ 297 */
296 ppage = pfn_to_online_page(pfn 298 ppage = pfn_to_online_page(pfn);
297 299
298 if (ppage) 300 if (ppage)
299 ino = page_cgroup_ino( 301 ino = page_cgroup_ino(ppage);
300 else 302 else
301 ino = 0; 303 ino = 0;
302 304
303 if (put_user(ino, out)) { 305 if (put_user(ino, out)) {
304 ret = -EFAULT; 306 ret = -EFAULT;
305 break; 307 break;
306 } 308 }
307 309
308 pfn++; 310 pfn++;
309 out++; 311 out++;
310 count -= KPMSIZE; 312 count -= KPMSIZE;
311 313
312 cond_resched(); 314 cond_resched();
313 } 315 }
314 316
315 *ppos += (char __user *)out - buf; 317 *ppos += (char __user *)out - buf;
316 if (!ret) 318 if (!ret)
317 ret = (char __user *)out - buf 319 ret = (char __user *)out - buf;
318 return ret; 320 return ret;
319 } 321 }
320 322
321 static const struct proc_ops kpagecgroup_proc_ 323 static const struct proc_ops kpagecgroup_proc_ops = {
322 .proc_flags = PROC_ENTRY_PERMANENT <<
323 .proc_lseek = mem_lseek, 324 .proc_lseek = mem_lseek,
324 .proc_read = kpagecgroup_read, 325 .proc_read = kpagecgroup_read,
325 }; 326 };
326 #endif /* CONFIG_MEMCG */ 327 #endif /* CONFIG_MEMCG */
327 328
328 static int __init proc_page_init(void) 329 static int __init proc_page_init(void)
329 { 330 {
330 proc_create("kpagecount", S_IRUSR, NUL 331 proc_create("kpagecount", S_IRUSR, NULL, &kpagecount_proc_ops);
331 proc_create("kpageflags", S_IRUSR, NUL 332 proc_create("kpageflags", S_IRUSR, NULL, &kpageflags_proc_ops);
332 #ifdef CONFIG_MEMCG 333 #ifdef CONFIG_MEMCG
333 proc_create("kpagecgroup", S_IRUSR, NU 334 proc_create("kpagecgroup", S_IRUSR, NULL, &kpagecgroup_proc_ops);
334 #endif 335 #endif
335 return 0; 336 return 0;
336 } 337 }
337 fs_initcall(proc_page_init); 338 fs_initcall(proc_page_init);
338 339
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.