tjh.dev
Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1// SPDX-License-Identifier: GPL-2.0
2#include <linux/memblock.h>
3#include <linux/compiler.h>
4#include <linux/fs.h>
5#include <linux/init.h>
6#include <linux/ksm.h>
7#include <linux/mm.h>
8#include <linux/mmzone.h>
9#include <linux/huge_mm.h>
10#include <linux/proc_fs.h>
11#include <linux/seq_file.h>
12#include <linux/hugetlb.h>
13#include <linux/memremap.h>
14#include <linux/memcontrol.h>
15#include <linux/mmu_notifier.h>
16#include <linux/page_idle.h>
17#include <linux/kernel-page-flags.h>
18#include <linux/uaccess.h>
19#include "internal.h"
20
21#define KPMSIZE sizeof(u64)
22#define KPMMASK (KPMSIZE - 1)
23#define KPMBITS (KPMSIZE * BITS_PER_BYTE)
24
25static inline unsigned long get_max_dump_pfn(void)
26{
27#ifdef CONFIG_SPARSEMEM
28 /*
29 * The memmap of early sections is completely populated and marked
30 * online even if max_pfn does not fall on a section boundary -
31 * pfn_to_online_page() will succeed on all pages. Allow inspecting
32 * these memmaps.
33 */
34 return round_up(max_pfn, PAGES_PER_SECTION);
35#else
36 return max_pfn;
37#endif
38}
39
40/* /proc/kpagecount - an array exposing page counts
41 *
42 * Each entry is a u64 representing the corresponding
43 * physical page count.
44 */
45static ssize_t kpagecount_read(struct file *file, char __user *buf,
46 size_t count, loff_t *ppos)
47{
48 const unsigned long max_dump_pfn = get_max_dump_pfn();
49 u64 __user *out = (u64 __user *)buf;
50 struct page *ppage;
51 unsigned long src = *ppos;
52 unsigned long pfn;
53 ssize_t ret = 0;
54 u64 pcount;
55
56 pfn = src / KPMSIZE;
57 if (src & KPMMASK || count & KPMMASK)
58 return -EINVAL;
59 if (src >= max_dump_pfn * KPMSIZE)
60 return 0;
61 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src);
62
63 while (count > 0) {
64 /*
65 * TODO: ZONE_DEVICE support requires to identify
66 * memmaps that were actually initialized.
67 */
68 ppage = pfn_to_online_page(pfn);
69
70 if (!ppage)
71 pcount = 0;
72 else
73 pcount = page_mapcount(ppage);
74
75 if (put_user(pcount, out)) {
76 ret = -EFAULT;
77 break;
78 }
79
80 pfn++;
81 out++;
82 count -= KPMSIZE;
83
84 cond_resched();
85 }
86
87 *ppos += (char __user *)out - buf;
88 if (!ret)
89 ret = (char __user *)out - buf;
90 return ret;
91}
92
93static const struct proc_ops kpagecount_proc_ops = {
94 .proc_flags = PROC_ENTRY_PERMANENT,
95 .proc_lseek = mem_lseek,
96 .proc_read = kpagecount_read,
97};
98
99/* /proc/kpageflags - an array exposing page flags
100 *
101 * Each entry is a u64 representing the corresponding
102 * physical page flags.
103 */
104
105static inline u64 kpf_copy_bit(u64 kflags, int ubit, int kbit)
106{
107 return ((kflags >> kbit) & 1) << ubit;
108}
109
110u64 stable_page_flags(const struct page *page)
111{
112 const struct folio *folio;
113 unsigned long k;
114 unsigned long mapping;
115 bool is_anon;
116 u64 u = 0;
117
118 /*
119 * pseudo flag: KPF_NOPAGE
120 * it differentiates a memory hole from a page with no flags
121 */
122 if (!page)
123 return 1 << KPF_NOPAGE;
124 folio = page_folio(page);
125
126 k = folio->flags;
127 mapping = (unsigned long)folio->mapping;
128 is_anon = mapping & PAGE_MAPPING_ANON;
129
130 /*
131 * pseudo flags for the well known (anonymous) memory mapped pages
132 */
133 if (page_mapped(page))
134 u |= 1 << KPF_MMAP;
135 if (is_anon) {
136 u |= 1 << KPF_ANON;
137 if (mapping & PAGE_MAPPING_KSM)
138 u |= 1 << KPF_KSM;
139 }
140
141 /*
142 * compound pages: export both head/tail info
143 * they together define a compound page's start/end pos and order
144 */
145 if (page == &folio->page)
146 u |= kpf_copy_bit(k, KPF_COMPOUND_HEAD, PG_head);
147 else
148 u |= 1 << KPF_COMPOUND_TAIL;
149 if (folio_test_hugetlb(folio))
150 u |= 1 << KPF_HUGE;
151 /*
152 * We need to check PageLRU/PageAnon
153 * to make sure a given page is a thp, not a non-huge compound page.
154 */
155 else if (folio_test_large(folio)) {
156 if ((k & (1 << PG_lru)) || is_anon)
157 u |= 1 << KPF_THP;
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)))
163 u |= 1 << KPF_ZERO_PAGE;
164
165 /*
166 * Caveats on high order pages: PG_buddy and PG_slab will only be set
167 * on the head page.
168 */
169 if (PageBuddy(page))
170 u |= 1 << KPF_BUDDY;
171 else if (page_count(page) == 0 && is_free_buddy_page(page))
172 u |= 1 << KPF_BUDDY;
173
174 if (PageOffline(page))
175 u |= 1 << KPF_OFFLINE;
176 if (PageTable(page))
177 u |= 1 << KPF_PGTABLE;
178 if (folio_test_slab(folio))
179 u |= 1 << KPF_SLAB;
180
181#if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT)
182 u |= kpf_copy_bit(k, KPF_IDLE, PG_idle);
183#else
184 if (folio_test_idle(folio))
185 u |= 1 << KPF_IDLE;
186#endif
187
188 u |= kpf_copy_bit(k, KPF_LOCKED, PG_locked);
189 u |= kpf_copy_bit(k, KPF_ERROR, PG_error);
190 u |= kpf_copy_bit(k, KPF_DIRTY, PG_dirty);
191 u |= kpf_copy_bit(k, KPF_UPTODATE, PG_uptodate);
192 u |= kpf_copy_bit(k, KPF_WRITEBACK, PG_writeback);
193
194 u |= kpf_copy_bit(k, KPF_LRU, PG_lru);
195 u |= kpf_copy_bit(k, KPF_REFERENCED, PG_referenced);
196 u |= kpf_copy_bit(k, KPF_ACTIVE, PG_active);
197 u |= kpf_copy_bit(k, KPF_RECLAIM, PG_reclaim);
198
199#define SWAPCACHE ((1 << PG_swapbacked) | (1 << PG_swapcache))
200 if ((k & SWAPCACHE) == SWAPCACHE)
201 u |= 1 << KPF_SWAPCACHE;
202 u |= kpf_copy_bit(k, KPF_SWAPBACKED, PG_swapbacked);
203
204 u |= kpf_copy_bit(k, KPF_UNEVICTABLE, PG_unevictable);
205 u |= kpf_copy_bit(k, KPF_MLOCKED, PG_mlocked);
206
207#ifdef CONFIG_MEMORY_FAILURE
208 if (u & (1 << KPF_HUGE))
209 u |= kpf_copy_bit(k, KPF_HWPOISON, PG_hwpoison);
210 else
211 u |= kpf_copy_bit(page->flags, KPF_HWPOISON, PG_hwpoison);
212#endif
213
214#ifdef CONFIG_ARCH_USES_PG_UNCACHED
215 u |= kpf_copy_bit(k, KPF_UNCACHED, PG_uncached);
216#endif
217
218 u |= kpf_copy_bit(k, KPF_RESERVED, PG_reserved);
219 u |= kpf_copy_bit(k, KPF_MAPPEDTODISK, PG_mappedtodisk);
220 u |= kpf_copy_bit(k, KPF_PRIVATE, PG_private);
221 u |= kpf_copy_bit(k, KPF_PRIVATE_2, PG_private_2);
222 u |= kpf_copy_bit(k, KPF_OWNER_PRIVATE, PG_owner_priv_1);
223 u |= kpf_copy_bit(k, KPF_ARCH, PG_arch_1);
224#ifdef CONFIG_ARCH_USES_PG_ARCH_X
225 u |= kpf_copy_bit(k, KPF_ARCH_2, PG_arch_2);
226 u |= kpf_copy_bit(k, KPF_ARCH_3, PG_arch_3);
227#endif
228
229 return u;
230};
231
232static ssize_t kpageflags_read(struct file *file, char __user *buf,
233 size_t count, loff_t *ppos)
234{
235 const unsigned long max_dump_pfn = get_max_dump_pfn();
236 u64 __user *out = (u64 __user *)buf;
237 unsigned long src = *ppos;
238 unsigned long pfn;
239 ssize_t ret = 0;
240
241 pfn = src / KPMSIZE;
242 if (src & KPMMASK || count & KPMMASK)
243 return -EINVAL;
244 if (src >= max_dump_pfn * KPMSIZE)
245 return 0;
246 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src);
247
248 while (count > 0) {
249 /*
250 * TODO: ZONE_DEVICE support requires to identify
251 * memmaps that were actually initialized.
252 */
253 struct page *page = pfn_to_online_page(pfn);
254
255 if (put_user(stable_page_flags(page), out)) {
256 ret = -EFAULT;
257 break;
258 }
259
260 pfn++;
261 out++;
262 count -= KPMSIZE;
263
264 cond_resched();
265 }
266
267 *ppos += (char __user *)out - buf;
268 if (!ret)
269 ret = (char __user *)out - buf;
270 return ret;
271}
272
273static const struct proc_ops kpageflags_proc_ops = {
274 .proc_flags = PROC_ENTRY_PERMANENT,
275 .proc_lseek = mem_lseek,
276 .proc_read = kpageflags_read,
277};
278
279#ifdef CONFIG_MEMCG
280static ssize_t kpagecgroup_read(struct file *file, char __user *buf,
281 size_t count, loff_t *ppos)
282{
283 const unsigned long max_dump_pfn = get_max_dump_pfn();
284 u64 __user *out = (u64 __user *)buf;
285 struct page *ppage;
286 unsigned long src = *ppos;
287 unsigned long pfn;
288 ssize_t ret = 0;
289 u64 ino;
290
291 pfn = src / KPMSIZE;
292 if (src & KPMMASK || count & KPMMASK)
293 return -EINVAL;
294 if (src >= max_dump_pfn * KPMSIZE)
295 return 0;
296 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src);
297
298 while (count > 0) {
299 /*
300 * TODO: ZONE_DEVICE support requires to identify
301 * memmaps that were actually initialized.
302 */
303 ppage = pfn_to_online_page(pfn);
304
305 if (ppage)
306 ino = page_cgroup_ino(ppage);
307 else
308 ino = 0;
309
310 if (put_user(ino, out)) {
311 ret = -EFAULT;
312 break;
313 }
314
315 pfn++;
316 out++;
317 count -= KPMSIZE;
318
319 cond_resched();
320 }
321
322 *ppos += (char __user *)out - buf;
323 if (!ret)
324 ret = (char __user *)out - buf;
325 return ret;
326}
327
328static const struct proc_ops kpagecgroup_proc_ops = {
329 .proc_flags = PROC_ENTRY_PERMANENT,
330 .proc_lseek = mem_lseek,
331 .proc_read = kpagecgroup_read,
332};
333#endif /* CONFIG_MEMCG */
334
335static int __init proc_page_init(void)
336{
337 proc_create("kpagecount", S_IRUSR, NULL, &kpagecount_proc_ops);
338 proc_create("kpageflags", S_IRUSR, NULL, &kpageflags_proc_ops);
339#ifdef CONFIG_MEMCG
340 proc_create("kpagecgroup", S_IRUSR, NULL, &kpagecgroup_proc_ops);
341#endif
342 return 0;
343}
344fs_initcall(proc_page_init);