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