xref: /openbmc/linux/fs/proc/page.c (revision be709d48)
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/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>
18 #include "internal.h"
19 
20 #define KPMSIZE sizeof(u64)
21 #define KPMMASK (KPMSIZE - 1)
22 #define KPMBITS (KPMSIZE * BITS_PER_BYTE)
23 
24 /* /proc/kpagecount - an array exposing page counts
25  *
26  * Each entry is a u64 representing the corresponding
27  * physical page count.
28  */
29 static ssize_t kpagecount_read(struct file *file, char __user *buf,
30 			     size_t count, loff_t *ppos)
31 {
32 	u64 __user *out = (u64 __user *)buf;
33 	struct page *ppage;
34 	unsigned long src = *ppos;
35 	unsigned long pfn;
36 	ssize_t ret = 0;
37 	u64 pcount;
38 
39 	pfn = src / KPMSIZE;
40 	count = min_t(size_t, count, (max_pfn * KPMSIZE) - src);
41 	if (src & KPMMASK || count & KPMMASK)
42 		return -EINVAL;
43 
44 	while (count > 0) {
45 		if (pfn_valid(pfn))
46 			ppage = pfn_to_page(pfn);
47 		else
48 			ppage = NULL;
49 		if (!ppage || PageSlab(ppage) || page_has_type(ppage))
50 			pcount = 0;
51 		else
52 			pcount = page_mapcount(ppage);
53 
54 		if (put_user(pcount, out)) {
55 			ret = -EFAULT;
56 			break;
57 		}
58 
59 		pfn++;
60 		out++;
61 		count -= KPMSIZE;
62 
63 		cond_resched();
64 	}
65 
66 	*ppos += (char __user *)out - buf;
67 	if (!ret)
68 		ret = (char __user *)out - buf;
69 	return ret;
70 }
71 
72 static const struct file_operations proc_kpagecount_operations = {
73 	.llseek = mem_lseek,
74 	.read = kpagecount_read,
75 };
76 
77 /* /proc/kpageflags - an array exposing page flags
78  *
79  * Each entry is a u64 representing the corresponding
80  * physical page flags.
81  */
82 
83 static inline u64 kpf_copy_bit(u64 kflags, int ubit, int kbit)
84 {
85 	return ((kflags >> kbit) & 1) << ubit;
86 }
87 
88 u64 stable_page_flags(struct page *page)
89 {
90 	u64 k;
91 	u64 u;
92 
93 	/*
94 	 * pseudo flag: KPF_NOPAGE
95 	 * it differentiates a memory hole from a page with no flags
96 	 */
97 	if (!page)
98 		return 1 << KPF_NOPAGE;
99 
100 	k = page->flags;
101 	u = 0;
102 
103 	/*
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.
108 	 */
109 	if (!PageSlab(page) && page_mapped(page))
110 		u |= 1 << KPF_MMAP;
111 	if (PageAnon(page))
112 		u |= 1 << KPF_ANON;
113 	if (PageKsm(page))
114 		u |= 1 << KPF_KSM;
115 
116 	/*
117 	 * compound pages: export both head/tail info
118 	 * they together define a compound page's start/end pos and order
119 	 */
120 	if (PageHead(page))
121 		u |= 1 << KPF_COMPOUND_HEAD;
122 	if (PageTail(page))
123 		u |= 1 << KPF_COMPOUND_TAIL;
124 	if (PageHuge(page))
125 		u |= 1 << KPF_HUGE;
126 	/*
127 	 * PageTransCompound can be true for non-huge compound pages (slab
128 	 * pages or pages allocated by drivers with __GFP_COMP) because it
129 	 * just checks PG_head/PG_tail, so we need to check PageLRU/PageAnon
130 	 * to make sure a given page is a thp, not a non-huge compound page.
131 	 */
132 	else if (PageTransCompound(page)) {
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)))
142 		u |= 1 << KPF_ZERO_PAGE;
143 
144 
145 	/*
146 	 * Caveats on high order pages: page->_refcount will only be set
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.
149 	 */
150 	if (PageBuddy(page))
151 		u |= 1 << KPF_BUDDY;
152 	else if (page_count(page) == 0 && is_free_buddy_page(page))
153 		u |= 1 << KPF_BUDDY;
154 
155 	if (PageOffline(page))
156 		u |= 1 << KPF_OFFLINE;
157 	if (PageTable(page))
158 		u |= 1 << KPF_PGTABLE;
159 
160 	if (page_is_idle(page))
161 		u |= 1 << KPF_IDLE;
162 
163 	u |= kpf_copy_bit(k, KPF_LOCKED,	PG_locked);
164 
165 	u |= kpf_copy_bit(k, KPF_SLAB,		PG_slab);
166 	if (PageTail(page) && PageSlab(compound_head(page)))
167 		u |= 1 << KPF_SLAB;
168 
169 	u |= kpf_copy_bit(k, KPF_ERROR,		PG_error);
170 	u |= kpf_copy_bit(k, KPF_DIRTY,		PG_dirty);
171 	u |= kpf_copy_bit(k, KPF_UPTODATE,	PG_uptodate);
172 	u |= kpf_copy_bit(k, KPF_WRITEBACK,	PG_writeback);
173 
174 	u |= kpf_copy_bit(k, KPF_LRU,		PG_lru);
175 	u |= kpf_copy_bit(k, KPF_REFERENCED,	PG_referenced);
176 	u |= kpf_copy_bit(k, KPF_ACTIVE,	PG_active);
177 	u |= kpf_copy_bit(k, KPF_RECLAIM,	PG_reclaim);
178 
179 	if (PageSwapCache(page))
180 		u |= 1 << KPF_SWAPCACHE;
181 	u |= kpf_copy_bit(k, KPF_SWAPBACKED,	PG_swapbacked);
182 
183 	u |= kpf_copy_bit(k, KPF_UNEVICTABLE,	PG_unevictable);
184 	u |= kpf_copy_bit(k, KPF_MLOCKED,	PG_mlocked);
185 
186 #ifdef CONFIG_MEMORY_FAILURE
187 	u |= kpf_copy_bit(k, KPF_HWPOISON,	PG_hwpoison);
188 #endif
189 
190 #ifdef CONFIG_ARCH_USES_PG_UNCACHED
191 	u |= kpf_copy_bit(k, KPF_UNCACHED,	PG_uncached);
192 #endif
193 
194 	u |= kpf_copy_bit(k, KPF_RESERVED,	PG_reserved);
195 	u |= kpf_copy_bit(k, KPF_MAPPEDTODISK,	PG_mappedtodisk);
196 	u |= kpf_copy_bit(k, KPF_PRIVATE,	PG_private);
197 	u |= kpf_copy_bit(k, KPF_PRIVATE_2,	PG_private_2);
198 	u |= kpf_copy_bit(k, KPF_OWNER_PRIVATE,	PG_owner_priv_1);
199 	u |= kpf_copy_bit(k, KPF_ARCH,		PG_arch_1);
200 
201 	return u;
202 };
203 
204 static ssize_t kpageflags_read(struct file *file, char __user *buf,
205 			     size_t count, loff_t *ppos)
206 {
207 	u64 __user *out = (u64 __user *)buf;
208 	struct page *ppage;
209 	unsigned long src = *ppos;
210 	unsigned long pfn;
211 	ssize_t ret = 0;
212 
213 	pfn = src / KPMSIZE;
214 	count = min_t(unsigned long, count, (max_pfn * KPMSIZE) - src);
215 	if (src & KPMMASK || count & KPMMASK)
216 		return -EINVAL;
217 
218 	while (count > 0) {
219 		if (pfn_valid(pfn))
220 			ppage = pfn_to_page(pfn);
221 		else
222 			ppage = NULL;
223 
224 		if (put_user(stable_page_flags(ppage), out)) {
225 			ret = -EFAULT;
226 			break;
227 		}
228 
229 		pfn++;
230 		out++;
231 		count -= KPMSIZE;
232 
233 		cond_resched();
234 	}
235 
236 	*ppos += (char __user *)out - buf;
237 	if (!ret)
238 		ret = (char __user *)out - buf;
239 	return ret;
240 }
241 
242 static const struct file_operations proc_kpageflags_operations = {
243 	.llseek = mem_lseek,
244 	.read = kpageflags_read,
245 };
246 
247 #ifdef CONFIG_MEMCG
248 static ssize_t kpagecgroup_read(struct file *file, char __user *buf,
249 				size_t count, loff_t *ppos)
250 {
251 	u64 __user *out = (u64 __user *)buf;
252 	struct page *ppage;
253 	unsigned long src = *ppos;
254 	unsigned long pfn;
255 	ssize_t ret = 0;
256 	u64 ino;
257 
258 	pfn = src / KPMSIZE;
259 	count = min_t(unsigned long, count, (max_pfn * KPMSIZE) - src);
260 	if (src & KPMMASK || count & KPMMASK)
261 		return -EINVAL;
262 
263 	while (count > 0) {
264 		if (pfn_valid(pfn))
265 			ppage = pfn_to_page(pfn);
266 		else
267 			ppage = NULL;
268 
269 		if (ppage)
270 			ino = page_cgroup_ino(ppage);
271 		else
272 			ino = 0;
273 
274 		if (put_user(ino, out)) {
275 			ret = -EFAULT;
276 			break;
277 		}
278 
279 		pfn++;
280 		out++;
281 		count -= KPMSIZE;
282 
283 		cond_resched();
284 	}
285 
286 	*ppos += (char __user *)out - buf;
287 	if (!ret)
288 		ret = (char __user *)out - buf;
289 	return ret;
290 }
291 
292 static const struct file_operations proc_kpagecgroup_operations = {
293 	.llseek = mem_lseek,
294 	.read = kpagecgroup_read,
295 };
296 #endif /* CONFIG_MEMCG */
297 
298 static int __init proc_page_init(void)
299 {
300 	proc_create("kpagecount", S_IRUSR, NULL, &proc_kpagecount_operations);
301 	proc_create("kpageflags", S_IRUSR, NULL, &proc_kpageflags_operations);
302 #ifdef CONFIG_MEMCG
303 	proc_create("kpagecgroup", S_IRUSR, NULL, &proc_kpagecgroup_operations);
304 #endif
305 	return 0;
306 }
307 fs_initcall(proc_page_init);
308