Commit | Line | Data |
---|---|---|
6d80e53f AD |
1 | #include <linux/bootmem.h> |
2 | #include <linux/compiler.h> | |
3 | #include <linux/fs.h> | |
4 | #include <linux/init.h> | |
9a840895 | 5 | #include <linux/ksm.h> |
6d80e53f AD |
6 | #include <linux/mm.h> |
7 | #include <linux/mmzone.h> | |
56873f43 | 8 | #include <linux/huge_mm.h> |
6d80e53f AD |
9 | #include <linux/proc_fs.h> |
10 | #include <linux/seq_file.h> | |
20a0307c | 11 | #include <linux/hugetlb.h> |
80ae2fdc | 12 | #include <linux/memcontrol.h> |
33c3fc71 VD |
13 | #include <linux/mmu_notifier.h> |
14 | #include <linux/page_idle.h> | |
1a9b5b7f | 15 | #include <linux/kernel-page-flags.h> |
6d80e53f AD |
16 | #include <asm/uaccess.h> |
17 | #include "internal.h" | |
18 | ||
19 | #define KPMSIZE sizeof(u64) | |
20 | #define KPMMASK (KPMSIZE - 1) | |
33c3fc71 | 21 | #define KPMBITS (KPMSIZE * BITS_PER_BYTE) |
ed7ce0f1 | 22 | |
6d80e53f AD |
23 | /* /proc/kpagecount - an array exposing page counts |
24 | * | |
25 | * Each entry is a u64 representing the corresponding | |
26 | * physical page count. | |
27 | */ | |
28 | static ssize_t kpagecount_read(struct file *file, char __user *buf, | |
29 | size_t count, loff_t *ppos) | |
30 | { | |
31 | u64 __user *out = (u64 __user *)buf; | |
32 | struct page *ppage; | |
33 | unsigned long src = *ppos; | |
34 | unsigned long pfn; | |
35 | ssize_t ret = 0; | |
36 | u64 pcount; | |
37 | ||
38 | pfn = src / KPMSIZE; | |
39 | count = min_t(size_t, count, (max_pfn * KPMSIZE) - src); | |
40 | if (src & KPMMASK || count & KPMMASK) | |
41 | return -EINVAL; | |
42 | ||
43 | while (count > 0) { | |
6d80e53f AD |
44 | if (pfn_valid(pfn)) |
45 | ppage = pfn_to_page(pfn); | |
ed7ce0f1 WF |
46 | else |
47 | ppage = NULL; | |
a6fc86d2 | 48 | if (!ppage || PageSlab(ppage)) |
6d80e53f AD |
49 | pcount = 0; |
50 | else | |
51 | pcount = page_mapcount(ppage); | |
52 | ||
ed7ce0f1 | 53 | if (put_user(pcount, out)) { |
6d80e53f AD |
54 | ret = -EFAULT; |
55 | break; | |
56 | } | |
57 | ||
ed7ce0f1 WF |
58 | pfn++; |
59 | out++; | |
6d80e53f AD |
60 | count -= KPMSIZE; |
61 | } | |
62 | ||
63 | *ppos += (char __user *)out - buf; | |
64 | if (!ret) | |
65 | ret = (char __user *)out - buf; | |
66 | return ret; | |
67 | } | |
68 | ||
69 | static const struct file_operations proc_kpagecount_operations = { | |
70 | .llseek = mem_lseek, | |
71 | .read = kpagecount_read, | |
72 | }; | |
73 | ||
74 | /* /proc/kpageflags - an array exposing page flags | |
75 | * | |
76 | * Each entry is a u64 representing the corresponding | |
77 | * physical page flags. | |
78 | */ | |
79 | ||
17797549 WF |
80 | static inline u64 kpf_copy_bit(u64 kflags, int ubit, int kbit) |
81 | { | |
82 | return ((kflags >> kbit) & 1) << ubit; | |
83 | } | |
84 | ||
1a9b5b7f | 85 | u64 stable_page_flags(struct page *page) |
17797549 WF |
86 | { |
87 | u64 k; | |
88 | u64 u; | |
89 | ||
90 | /* | |
91 | * pseudo flag: KPF_NOPAGE | |
92 | * it differentiates a memory hole from a page with no flags | |
93 | */ | |
94 | if (!page) | |
95 | return 1 << KPF_NOPAGE; | |
96 | ||
97 | k = page->flags; | |
98 | u = 0; | |
99 | ||
100 | /* | |
101 | * pseudo flags for the well known (anonymous) memory mapped pages | |
102 | * | |
103 | * Note that page->_mapcount is overloaded in SLOB/SLUB/SLQB, so the | |
104 | * simple test in page_mapped() is not enough. | |
105 | */ | |
106 | if (!PageSlab(page) && page_mapped(page)) | |
107 | u |= 1 << KPF_MMAP; | |
108 | if (PageAnon(page)) | |
109 | u |= 1 << KPF_ANON; | |
9a840895 HD |
110 | if (PageKsm(page)) |
111 | u |= 1 << KPF_KSM; | |
17797549 WF |
112 | |
113 | /* | |
114 | * compound pages: export both head/tail info | |
115 | * they together define a compound page's start/end pos and order | |
116 | */ | |
117 | if (PageHead(page)) | |
118 | u |= 1 << KPF_COMPOUND_HEAD; | |
119 | if (PageTail(page)) | |
120 | u |= 1 << KPF_COMPOUND_TAIL; | |
121 | if (PageHuge(page)) | |
122 | u |= 1 << KPF_HUGE; | |
7a71932d NH |
123 | /* |
124 | * PageTransCompound can be true for non-huge compound pages (slab | |
125 | * pages or pages allocated by drivers with __GFP_COMP) because it | |
e3bba3c3 NH |
126 | * just checks PG_head/PG_tail, so we need to check PageLRU/PageAnon |
127 | * to make sure a given page is a thp, not a non-huge compound page. | |
7a71932d | 128 | */ |
56873f43 WY |
129 | else if (PageTransCompound(page)) { |
130 | struct page *head = compound_head(page); | |
131 | ||
132 | if (PageLRU(head) || PageAnon(head)) | |
133 | u |= 1 << KPF_THP; | |
134 | else if (is_huge_zero_page(head)) { | |
135 | u |= 1 << KPF_ZERO_PAGE; | |
136 | u |= 1 << KPF_THP; | |
137 | } | |
138 | } else if (is_zero_pfn(page_to_pfn(page))) | |
139 | u |= 1 << KPF_ZERO_PAGE; | |
140 | ||
17797549 | 141 | |
17797549 | 142 | /* |
5f24ce5f AA |
143 | * Caveats on high order pages: page->_count will only be set |
144 | * -1 on the head page; SLUB/SLQB do the same for PG_slab; | |
145 | * SLOB won't set PG_slab at all on compound pages. | |
17797549 | 146 | */ |
5f24ce5f AA |
147 | if (PageBuddy(page)) |
148 | u |= 1 << KPF_BUDDY; | |
149 | ||
09316c09 KK |
150 | if (PageBalloon(page)) |
151 | u |= 1 << KPF_BALLOON; | |
152 | ||
5f24ce5f AA |
153 | u |= kpf_copy_bit(k, KPF_LOCKED, PG_locked); |
154 | ||
17797549 | 155 | u |= kpf_copy_bit(k, KPF_SLAB, PG_slab); |
17797549 WF |
156 | |
157 | u |= kpf_copy_bit(k, KPF_ERROR, PG_error); | |
158 | u |= kpf_copy_bit(k, KPF_DIRTY, PG_dirty); | |
159 | u |= kpf_copy_bit(k, KPF_UPTODATE, PG_uptodate); | |
160 | u |= kpf_copy_bit(k, KPF_WRITEBACK, PG_writeback); | |
161 | ||
162 | u |= kpf_copy_bit(k, KPF_LRU, PG_lru); | |
163 | u |= kpf_copy_bit(k, KPF_REFERENCED, PG_referenced); | |
164 | u |= kpf_copy_bit(k, KPF_ACTIVE, PG_active); | |
165 | u |= kpf_copy_bit(k, KPF_RECLAIM, PG_reclaim); | |
166 | ||
167 | u |= kpf_copy_bit(k, KPF_SWAPCACHE, PG_swapcache); | |
168 | u |= kpf_copy_bit(k, KPF_SWAPBACKED, PG_swapbacked); | |
169 | ||
17797549 WF |
170 | u |= kpf_copy_bit(k, KPF_UNEVICTABLE, PG_unevictable); |
171 | u |= kpf_copy_bit(k, KPF_MLOCKED, PG_mlocked); | |
17797549 | 172 | |
253fb02d WF |
173 | #ifdef CONFIG_MEMORY_FAILURE |
174 | u |= kpf_copy_bit(k, KPF_HWPOISON, PG_hwpoison); | |
175 | #endif | |
176 | ||
ed430fec | 177 | #ifdef CONFIG_ARCH_USES_PG_UNCACHED |
17797549 WF |
178 | u |= kpf_copy_bit(k, KPF_UNCACHED, PG_uncached); |
179 | #endif | |
180 | ||
181 | u |= kpf_copy_bit(k, KPF_RESERVED, PG_reserved); | |
182 | u |= kpf_copy_bit(k, KPF_MAPPEDTODISK, PG_mappedtodisk); | |
183 | u |= kpf_copy_bit(k, KPF_PRIVATE, PG_private); | |
184 | u |= kpf_copy_bit(k, KPF_PRIVATE_2, PG_private_2); | |
185 | u |= kpf_copy_bit(k, KPF_OWNER_PRIVATE, PG_owner_priv_1); | |
186 | u |= kpf_copy_bit(k, KPF_ARCH, PG_arch_1); | |
187 | ||
188 | return u; | |
189 | }; | |
6d80e53f AD |
190 | |
191 | static ssize_t kpageflags_read(struct file *file, char __user *buf, | |
192 | size_t count, loff_t *ppos) | |
193 | { | |
194 | u64 __user *out = (u64 __user *)buf; | |
195 | struct page *ppage; | |
196 | unsigned long src = *ppos; | |
197 | unsigned long pfn; | |
198 | ssize_t ret = 0; | |
6d80e53f AD |
199 | |
200 | pfn = src / KPMSIZE; | |
201 | count = min_t(unsigned long, count, (max_pfn * KPMSIZE) - src); | |
202 | if (src & KPMMASK || count & KPMMASK) | |
203 | return -EINVAL; | |
204 | ||
205 | while (count > 0) { | |
6d80e53f AD |
206 | if (pfn_valid(pfn)) |
207 | ppage = pfn_to_page(pfn); | |
ed7ce0f1 WF |
208 | else |
209 | ppage = NULL; | |
17797549 | 210 | |
1a9b5b7f | 211 | if (put_user(stable_page_flags(ppage), out)) { |
6d80e53f AD |
212 | ret = -EFAULT; |
213 | break; | |
214 | } | |
215 | ||
ed7ce0f1 WF |
216 | pfn++; |
217 | out++; | |
6d80e53f AD |
218 | count -= KPMSIZE; |
219 | } | |
220 | ||
221 | *ppos += (char __user *)out - buf; | |
222 | if (!ret) | |
223 | ret = (char __user *)out - buf; | |
224 | return ret; | |
225 | } | |
226 | ||
227 | static const struct file_operations proc_kpageflags_operations = { | |
228 | .llseek = mem_lseek, | |
229 | .read = kpageflags_read, | |
230 | }; | |
231 | ||
80ae2fdc VD |
232 | #ifdef CONFIG_MEMCG |
233 | static ssize_t kpagecgroup_read(struct file *file, char __user *buf, | |
234 | size_t count, loff_t *ppos) | |
235 | { | |
236 | u64 __user *out = (u64 __user *)buf; | |
237 | struct page *ppage; | |
238 | unsigned long src = *ppos; | |
239 | unsigned long pfn; | |
240 | ssize_t ret = 0; | |
241 | u64 ino; | |
242 | ||
243 | pfn = src / KPMSIZE; | |
244 | count = min_t(unsigned long, count, (max_pfn * KPMSIZE) - src); | |
245 | if (src & KPMMASK || count & KPMMASK) | |
246 | return -EINVAL; | |
247 | ||
248 | while (count > 0) { | |
249 | if (pfn_valid(pfn)) | |
250 | ppage = pfn_to_page(pfn); | |
251 | else | |
252 | ppage = NULL; | |
253 | ||
254 | if (ppage) | |
255 | ino = page_cgroup_ino(ppage); | |
256 | else | |
257 | ino = 0; | |
258 | ||
259 | if (put_user(ino, out)) { | |
260 | ret = -EFAULT; | |
261 | break; | |
262 | } | |
263 | ||
264 | pfn++; | |
265 | out++; | |
266 | count -= KPMSIZE; | |
267 | } | |
268 | ||
269 | *ppos += (char __user *)out - buf; | |
270 | if (!ret) | |
271 | ret = (char __user *)out - buf; | |
272 | return ret; | |
273 | } | |
274 | ||
275 | static const struct file_operations proc_kpagecgroup_operations = { | |
276 | .llseek = mem_lseek, | |
277 | .read = kpagecgroup_read, | |
278 | }; | |
279 | #endif /* CONFIG_MEMCG */ | |
280 | ||
6d80e53f AD |
281 | static int __init proc_page_init(void) |
282 | { | |
283 | proc_create("kpagecount", S_IRUSR, NULL, &proc_kpagecount_operations); | |
284 | proc_create("kpageflags", S_IRUSR, NULL, &proc_kpageflags_operations); | |
80ae2fdc VD |
285 | #ifdef CONFIG_MEMCG |
286 | proc_create("kpagecgroup", S_IRUSR, NULL, &proc_kpagecgroup_operations); | |
287 | #endif | |
6d80e53f AD |
288 | return 0; |
289 | } | |
abaf3787 | 290 | fs_initcall(proc_page_init); |