1 // SPDX-License-Identifier: GPL-2.0
2 #include <linux/memblock.h>
3 #include <linux/compiler.h>
5 #include <linux/init.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>
21 #define KPMSIZE sizeof(u64)
22 #define KPMMASK (KPMSIZE - 1)
23 #define KPMBITS (KPMSIZE * BITS_PER_BYTE)
25 static inline unsigned long get_max_dump_pfn(void)
27 #ifdef CONFIG_SPARSEMEM
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
34 return round_up(max_pfn, PAGES_PER_SECTION);
40 /* /proc/kpagecount - an array exposing page counts
42 * Each entry is a u64 representing the corresponding
43 * physical page count.
45 static ssize_t kpagecount_read(struct file *file, char __user *buf,
46 size_t count, loff_t *ppos)
48 const unsigned long max_dump_pfn = get_max_dump_pfn();
49 u64 __user *out = (u64 __user *)buf;
51 unsigned long src = *ppos;
57 if (src & KPMMASK || count & KPMMASK)
59 if (src >= max_dump_pfn * KPMSIZE)
61 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src);
65 * TODO: ZONE_DEVICE support requires to identify
66 * memmaps that were actually initialized.
68 ppage = pfn_to_online_page(pfn);
73 pcount = page_mapcount(ppage);
75 if (put_user(pcount, out)) {
87 *ppos += (char __user *)out - buf;
89 ret = (char __user *)out - buf;
93 static const struct proc_ops kpagecount_proc_ops = {
94 .proc_flags = PROC_ENTRY_PERMANENT,
95 .proc_lseek = mem_lseek,
96 .proc_read = kpagecount_read,
99 /* /proc/kpageflags - an array exposing page flags
101 * Each entry is a u64 representing the corresponding
102 * physical page flags.
105 static inline u64 kpf_copy_bit(u64 kflags, int ubit, int kbit)
107 return ((kflags >> kbit) & 1) << ubit;
110 u64 stable_page_flags(const struct page *page)
112 const struct folio *folio;
114 unsigned long mapping;
119 * pseudo flag: KPF_NOPAGE
120 * it differentiates a memory hole from a page with no flags
123 return 1 << KPF_NOPAGE;
124 folio = page_folio(page);
127 mapping = (unsigned long)folio->mapping;
128 is_anon = mapping & PAGE_MAPPING_ANON;
131 * pseudo flags for the well known (anonymous) memory mapped pages
133 if (page_mapped(page))
137 if (mapping & PAGE_MAPPING_KSM)
142 * compound pages: export both head/tail info
143 * they together define a compound page's start/end pos and order
145 if (page == &folio->page)
146 u |= kpf_copy_bit(k, KPF_COMPOUND_HEAD, PG_head);
148 u |= 1 << KPF_COMPOUND_TAIL;
149 if (folio_test_hugetlb(folio))
152 * We need to check PageLRU/PageAnon
153 * to make sure a given page is a thp, not a non-huge compound page.
155 else if (folio_test_large(folio)) {
156 if ((k & (1 << PG_lru)) || is_anon)
158 else if (is_huge_zero_folio(folio)) {
159 u |= 1 << KPF_ZERO_PAGE;
162 } else if (is_zero_pfn(page_to_pfn(page)))
163 u |= 1 << KPF_ZERO_PAGE;
166 * Caveats on high order pages: PG_buddy and PG_slab will only be set
171 else if (page_count(page) == 0 && is_free_buddy_page(page))
174 if (PageOffline(page))
175 u |= 1 << KPF_OFFLINE;
177 u |= 1 << KPF_PGTABLE;
178 if (folio_test_slab(folio))
181 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT)
182 u |= kpf_copy_bit(k, KPF_IDLE, PG_idle);
184 if (folio_test_idle(folio))
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);
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);
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);
204 u |= kpf_copy_bit(k, KPF_UNEVICTABLE, PG_unevictable);
205 u |= kpf_copy_bit(k, KPF_MLOCKED, PG_mlocked);
207 #ifdef CONFIG_MEMORY_FAILURE
208 if (u & (1 << KPF_HUGE))
209 u |= kpf_copy_bit(k, KPF_HWPOISON, PG_hwpoison);
211 u |= kpf_copy_bit(page->flags, KPF_HWPOISON, PG_hwpoison);
214 #ifdef CONFIG_ARCH_USES_PG_UNCACHED
215 u |= kpf_copy_bit(k, KPF_UNCACHED, PG_uncached);
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);
232 static ssize_t kpageflags_read(struct file *file, char __user *buf,
233 size_t count, loff_t *ppos)
235 const unsigned long max_dump_pfn = get_max_dump_pfn();
236 u64 __user *out = (u64 __user *)buf;
237 unsigned long src = *ppos;
242 if (src & KPMMASK || count & KPMMASK)
244 if (src >= max_dump_pfn * KPMSIZE)
246 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src);
250 * TODO: ZONE_DEVICE support requires to identify
251 * memmaps that were actually initialized.
253 struct page *page = pfn_to_online_page(pfn);
255 if (put_user(stable_page_flags(page), out)) {
267 *ppos += (char __user *)out - buf;
269 ret = (char __user *)out - buf;
273 static const struct proc_ops kpageflags_proc_ops = {
274 .proc_flags = PROC_ENTRY_PERMANENT,
275 .proc_lseek = mem_lseek,
276 .proc_read = kpageflags_read,
280 static ssize_t kpagecgroup_read(struct file *file, char __user *buf,
281 size_t count, loff_t *ppos)
283 const unsigned long max_dump_pfn = get_max_dump_pfn();
284 u64 __user *out = (u64 __user *)buf;
286 unsigned long src = *ppos;
292 if (src & KPMMASK || count & KPMMASK)
294 if (src >= max_dump_pfn * KPMSIZE)
296 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src);
300 * TODO: ZONE_DEVICE support requires to identify
301 * memmaps that were actually initialized.
303 ppage = pfn_to_online_page(pfn);
306 ino = page_cgroup_ino(ppage);
310 if (put_user(ino, out)) {
322 *ppos += (char __user *)out - buf;
324 ret = (char __user *)out - buf;
328 static const struct proc_ops kpagecgroup_proc_ops = {
329 .proc_flags = PROC_ENTRY_PERMANENT,
330 .proc_lseek = mem_lseek,
331 .proc_read = kpagecgroup_read,
333 #endif /* CONFIG_MEMCG */
335 static int __init proc_page_init(void)
337 proc_create("kpagecount", S_IRUSR, NULL, &kpagecount_proc_ops);
338 proc_create("kpageflags", S_IRUSR, NULL, &kpageflags_proc_ops);
340 proc_create("kpagecgroup", S_IRUSR, NULL, &kpagecgroup_proc_ops);
344 fs_initcall(proc_page_init);