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/memcontrol.h>
14 #include <linux/mmu_notifier.h>
15 #include <linux/page_idle.h>
16 #include <linux/kernel-page-flags.h>
17 #include <linux/uaccess.h>
20 #define KPMSIZE sizeof(u64)
21 #define KPMMASK (KPMSIZE - 1)
22 #define KPMBITS (KPMSIZE * BITS_PER_BYTE)
24 static inline unsigned long get_max_dump_pfn(void)
26 #ifdef CONFIG_SPARSEMEM
28 * The memmap of early sections is completely populated and marked
29 * online even if max_pfn does not fall on a section boundary -
30 * pfn_to_online_page() will succeed on all pages. Allow inspecting
33 return round_up(max_pfn, PAGES_PER_SECTION);
39 /* /proc/kpagecount - an array exposing page counts
41 * Each entry is a u64 representing the corresponding
42 * physical page count.
44 static ssize_t kpagecount_read(struct file *file, char __user *buf,
45 size_t count, loff_t *ppos)
47 const unsigned long max_dump_pfn = get_max_dump_pfn();
48 u64 __user *out = (u64 __user *)buf;
50 unsigned long src = *ppos;
56 if (src & KPMMASK || count & KPMMASK)
58 if (src >= max_dump_pfn * KPMSIZE)
60 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src);
64 * TODO: ZONE_DEVICE support requires to identify
65 * memmaps that were actually initialized.
67 ppage = pfn_to_online_page(pfn);
69 if (!ppage || PageSlab(ppage) || page_has_type(ppage))
72 pcount = page_mapcount(ppage);
74 if (put_user(pcount, out)) {
86 *ppos += (char __user *)out - buf;
88 ret = (char __user *)out - buf;
92 static const struct proc_ops kpagecount_proc_ops = {
93 .proc_lseek = mem_lseek,
94 .proc_read = kpagecount_read,
97 /* /proc/kpageflags - an array exposing page flags
99 * Each entry is a u64 representing the corresponding
100 * physical page flags.
103 static inline u64 kpf_copy_bit(u64 kflags, int ubit, int kbit)
105 return ((kflags >> kbit) & 1) << ubit;
108 u64 stable_page_flags(struct page *page)
114 * pseudo flag: KPF_NOPAGE
115 * it differentiates a memory hole from a page with no flags
118 return 1 << KPF_NOPAGE;
124 * pseudo flags for the well known (anonymous) memory mapped pages
126 * Note that page->_mapcount is overloaded in SLOB/SLUB/SLQB, so the
127 * simple test in page_mapped() is not enough.
129 if (!PageSlab(page) && page_mapped(page))
137 * compound pages: export both head/tail info
138 * they together define a compound page's start/end pos and order
141 u |= 1 << KPF_COMPOUND_HEAD;
143 u |= 1 << KPF_COMPOUND_TAIL;
147 * PageTransCompound can be true for non-huge compound pages (slab
148 * pages or pages allocated by drivers with __GFP_COMP) because it
149 * just checks PG_head/PG_tail, so we need to check PageLRU/PageAnon
150 * to make sure a given page is a thp, not a non-huge compound page.
152 else if (PageTransCompound(page)) {
153 struct page *head = compound_head(page);
155 if (PageLRU(head) || PageAnon(head))
157 else if (is_huge_zero_page(head)) {
158 u |= 1 << KPF_ZERO_PAGE;
161 } else if (is_zero_pfn(page_to_pfn(page)))
162 u |= 1 << KPF_ZERO_PAGE;
166 * Caveats on high order pages: page->_refcount will only be set
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.
172 else if (page_count(page) == 0 && is_free_buddy_page(page))
175 if (PageOffline(page))
176 u |= 1 << KPF_OFFLINE;
178 u |= 1 << KPF_PGTABLE;
180 if (page_is_idle(page))
183 u |= kpf_copy_bit(k, KPF_LOCKED, PG_locked);
185 u |= kpf_copy_bit(k, KPF_SLAB, PG_slab);
186 if (PageTail(page) && PageSlab(compound_head(page)))
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 if (PageSwapCache(page))
200 u |= 1 << KPF_SWAPCACHE;
201 u |= kpf_copy_bit(k, KPF_SWAPBACKED, PG_swapbacked);
203 u |= kpf_copy_bit(k, KPF_UNEVICTABLE, PG_unevictable);
204 u |= kpf_copy_bit(k, KPF_MLOCKED, PG_mlocked);
206 #ifdef CONFIG_MEMORY_FAILURE
207 u |= kpf_copy_bit(k, KPF_HWPOISON, PG_hwpoison);
210 #ifdef CONFIG_ARCH_USES_PG_UNCACHED
211 u |= kpf_copy_bit(k, KPF_UNCACHED, PG_uncached);
214 u |= kpf_copy_bit(k, KPF_RESERVED, PG_reserved);
215 u |= kpf_copy_bit(k, KPF_MAPPEDTODISK, PG_mappedtodisk);
216 u |= kpf_copy_bit(k, KPF_PRIVATE, PG_private);
217 u |= kpf_copy_bit(k, KPF_PRIVATE_2, PG_private_2);
218 u |= kpf_copy_bit(k, KPF_OWNER_PRIVATE, PG_owner_priv_1);
219 u |= kpf_copy_bit(k, KPF_ARCH, PG_arch_1);
221 u |= kpf_copy_bit(k, KPF_ARCH_2, PG_arch_2);
227 static ssize_t kpageflags_read(struct file *file, char __user *buf,
228 size_t count, loff_t *ppos)
230 const unsigned long max_dump_pfn = get_max_dump_pfn();
231 u64 __user *out = (u64 __user *)buf;
233 unsigned long src = *ppos;
238 if (src & KPMMASK || count & KPMMASK)
240 if (src >= max_dump_pfn * KPMSIZE)
242 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src);
246 * TODO: ZONE_DEVICE support requires to identify
247 * memmaps that were actually initialized.
249 ppage = pfn_to_online_page(pfn);
251 if (put_user(stable_page_flags(ppage), out)) {
263 *ppos += (char __user *)out - buf;
265 ret = (char __user *)out - buf;
269 static const struct proc_ops kpageflags_proc_ops = {
270 .proc_lseek = mem_lseek,
271 .proc_read = kpageflags_read,
275 static ssize_t kpagecgroup_read(struct file *file, char __user *buf,
276 size_t count, loff_t *ppos)
278 const unsigned long max_dump_pfn = get_max_dump_pfn();
279 u64 __user *out = (u64 __user *)buf;
281 unsigned long src = *ppos;
287 if (src & KPMMASK || count & KPMMASK)
289 if (src >= max_dump_pfn * KPMSIZE)
291 count = min_t(unsigned long, count, (max_dump_pfn * KPMSIZE) - src);
295 * TODO: ZONE_DEVICE support requires to identify
296 * memmaps that were actually initialized.
298 ppage = pfn_to_online_page(pfn);
301 ino = page_cgroup_ino(ppage);
305 if (put_user(ino, out)) {
317 *ppos += (char __user *)out - buf;
319 ret = (char __user *)out - buf;
323 static const struct proc_ops kpagecgroup_proc_ops = {
324 .proc_lseek = mem_lseek,
325 .proc_read = kpagecgroup_read,
327 #endif /* CONFIG_MEMCG */
329 static int __init proc_page_init(void)
331 proc_create("kpagecount", S_IRUSR, NULL, &kpagecount_proc_ops);
332 proc_create("kpageflags", S_IRUSR, NULL, &kpageflags_proc_ops);
334 proc_create("kpagecgroup", S_IRUSR, NULL, &kpagecgroup_proc_ops);
338 fs_initcall(proc_page_init);