Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | #include <linux/mm.h> |
2 | #include <linux/hugetlb.h> | |
22e057c5 | 3 | #include <linux/huge_mm.h> |
1da177e4 LT |
4 | #include <linux/mount.h> |
5 | #include <linux/seq_file.h> | |
e070ad49 | 6 | #include <linux/highmem.h> |
5096add8 | 7 | #include <linux/ptrace.h> |
5a0e3ad6 | 8 | #include <linux/slab.h> |
6e21c8f1 CL |
9 | #include <linux/pagemap.h> |
10 | #include <linux/mempolicy.h> | |
22e057c5 | 11 | #include <linux/rmap.h> |
85863e47 MM |
12 | #include <linux/swap.h> |
13 | #include <linux/swapops.h> | |
e070ad49 | 14 | |
1da177e4 LT |
15 | #include <asm/elf.h> |
16 | #include <asm/uaccess.h> | |
e070ad49 | 17 | #include <asm/tlbflush.h> |
1da177e4 LT |
18 | #include "internal.h" |
19 | ||
df5f8314 | 20 | void task_mem(struct seq_file *m, struct mm_struct *mm) |
1da177e4 | 21 | { |
b084d435 | 22 | unsigned long data, text, lib, swap; |
365e9c87 HD |
23 | unsigned long hiwater_vm, total_vm, hiwater_rss, total_rss; |
24 | ||
25 | /* | |
26 | * Note: to minimize their overhead, mm maintains hiwater_vm and | |
27 | * hiwater_rss only when about to *lower* total_vm or rss. Any | |
28 | * collector of these hiwater stats must therefore get total_vm | |
29 | * and rss too, which will usually be the higher. Barriers? not | |
30 | * worth the effort, such snapshots can always be inconsistent. | |
31 | */ | |
32 | hiwater_vm = total_vm = mm->total_vm; | |
33 | if (hiwater_vm < mm->hiwater_vm) | |
34 | hiwater_vm = mm->hiwater_vm; | |
35 | hiwater_rss = total_rss = get_mm_rss(mm); | |
36 | if (hiwater_rss < mm->hiwater_rss) | |
37 | hiwater_rss = mm->hiwater_rss; | |
1da177e4 LT |
38 | |
39 | data = mm->total_vm - mm->shared_vm - mm->stack_vm; | |
40 | text = (PAGE_ALIGN(mm->end_code) - (mm->start_code & PAGE_MASK)) >> 10; | |
41 | lib = (mm->exec_vm << (PAGE_SHIFT-10)) - text; | |
b084d435 | 42 | swap = get_mm_counter(mm, MM_SWAPENTS); |
df5f8314 | 43 | seq_printf(m, |
365e9c87 | 44 | "VmPeak:\t%8lu kB\n" |
1da177e4 LT |
45 | "VmSize:\t%8lu kB\n" |
46 | "VmLck:\t%8lu kB\n" | |
bc3e53f6 | 47 | "VmPin:\t%8lu kB\n" |
365e9c87 | 48 | "VmHWM:\t%8lu kB\n" |
1da177e4 LT |
49 | "VmRSS:\t%8lu kB\n" |
50 | "VmData:\t%8lu kB\n" | |
51 | "VmStk:\t%8lu kB\n" | |
52 | "VmExe:\t%8lu kB\n" | |
53 | "VmLib:\t%8lu kB\n" | |
b084d435 KH |
54 | "VmPTE:\t%8lu kB\n" |
55 | "VmSwap:\t%8lu kB\n", | |
365e9c87 HD |
56 | hiwater_vm << (PAGE_SHIFT-10), |
57 | (total_vm - mm->reserved_vm) << (PAGE_SHIFT-10), | |
1da177e4 | 58 | mm->locked_vm << (PAGE_SHIFT-10), |
bc3e53f6 | 59 | mm->pinned_vm << (PAGE_SHIFT-10), |
365e9c87 HD |
60 | hiwater_rss << (PAGE_SHIFT-10), |
61 | total_rss << (PAGE_SHIFT-10), | |
1da177e4 LT |
62 | data << (PAGE_SHIFT-10), |
63 | mm->stack_vm << (PAGE_SHIFT-10), text, lib, | |
b084d435 KH |
64 | (PTRS_PER_PTE*sizeof(pte_t)*mm->nr_ptes) >> 10, |
65 | swap << (PAGE_SHIFT-10)); | |
1da177e4 LT |
66 | } |
67 | ||
68 | unsigned long task_vsize(struct mm_struct *mm) | |
69 | { | |
70 | return PAGE_SIZE * mm->total_vm; | |
71 | } | |
72 | ||
a2ade7b6 AD |
73 | unsigned long task_statm(struct mm_struct *mm, |
74 | unsigned long *shared, unsigned long *text, | |
75 | unsigned long *data, unsigned long *resident) | |
1da177e4 | 76 | { |
d559db08 | 77 | *shared = get_mm_counter(mm, MM_FILEPAGES); |
1da177e4 LT |
78 | *text = (PAGE_ALIGN(mm->end_code) - (mm->start_code & PAGE_MASK)) |
79 | >> PAGE_SHIFT; | |
80 | *data = mm->total_vm - mm->shared_vm; | |
d559db08 | 81 | *resident = *shared + get_mm_counter(mm, MM_ANONPAGES); |
1da177e4 LT |
82 | return mm->total_vm; |
83 | } | |
84 | ||
1da177e4 LT |
85 | static void pad_len_spaces(struct seq_file *m, int len) |
86 | { | |
87 | len = 25 + sizeof(void*) * 6 - len; | |
88 | if (len < 1) | |
89 | len = 1; | |
90 | seq_printf(m, "%*c", len, ' '); | |
91 | } | |
92 | ||
a6198797 MM |
93 | static void vma_stop(struct proc_maps_private *priv, struct vm_area_struct *vma) |
94 | { | |
95 | if (vma && vma != priv->tail_vma) { | |
96 | struct mm_struct *mm = vma->vm_mm; | |
97 | up_read(&mm->mmap_sem); | |
98 | mmput(mm); | |
99 | } | |
100 | } | |
ec4dd3eb | 101 | |
a6198797 | 102 | static void *m_start(struct seq_file *m, loff_t *pos) |
e070ad49 | 103 | { |
a6198797 MM |
104 | struct proc_maps_private *priv = m->private; |
105 | unsigned long last_addr = m->version; | |
106 | struct mm_struct *mm; | |
107 | struct vm_area_struct *vma, *tail_vma = NULL; | |
108 | loff_t l = *pos; | |
109 | ||
110 | /* Clear the per syscall fields in priv */ | |
111 | priv->task = NULL; | |
112 | priv->tail_vma = NULL; | |
113 | ||
114 | /* | |
115 | * We remember last_addr rather than next_addr to hit with | |
116 | * mmap_cache most of the time. We have zero last_addr at | |
117 | * the beginning and also after lseek. We will have -1 last_addr | |
118 | * after the end of the vmas. | |
119 | */ | |
120 | ||
121 | if (last_addr == -1UL) | |
122 | return NULL; | |
123 | ||
124 | priv->task = get_pid_task(priv->pid, PIDTYPE_PID); | |
125 | if (!priv->task) | |
ec6fd8a4 | 126 | return ERR_PTR(-ESRCH); |
a6198797 | 127 | |
e7dcd999 | 128 | mm = mm_access(priv->task, PTRACE_MODE_READ); |
ec6fd8a4 AV |
129 | if (!mm || IS_ERR(mm)) |
130 | return mm; | |
00f89d21 | 131 | down_read(&mm->mmap_sem); |
a6198797 | 132 | |
31db58b3 | 133 | tail_vma = get_gate_vma(priv->task->mm); |
a6198797 MM |
134 | priv->tail_vma = tail_vma; |
135 | ||
136 | /* Start with last addr hint */ | |
137 | vma = find_vma(mm, last_addr); | |
138 | if (last_addr && vma) { | |
139 | vma = vma->vm_next; | |
140 | goto out; | |
141 | } | |
142 | ||
143 | /* | |
144 | * Check the vma index is within the range and do | |
145 | * sequential scan until m_index. | |
146 | */ | |
147 | vma = NULL; | |
148 | if ((unsigned long)l < mm->map_count) { | |
149 | vma = mm->mmap; | |
150 | while (l-- && vma) | |
151 | vma = vma->vm_next; | |
152 | goto out; | |
153 | } | |
154 | ||
155 | if (l != mm->map_count) | |
156 | tail_vma = NULL; /* After gate vma */ | |
157 | ||
158 | out: | |
159 | if (vma) | |
160 | return vma; | |
161 | ||
162 | /* End of vmas has been reached */ | |
163 | m->version = (tail_vma != NULL)? 0: -1UL; | |
164 | up_read(&mm->mmap_sem); | |
165 | mmput(mm); | |
166 | return tail_vma; | |
167 | } | |
168 | ||
169 | static void *m_next(struct seq_file *m, void *v, loff_t *pos) | |
170 | { | |
171 | struct proc_maps_private *priv = m->private; | |
172 | struct vm_area_struct *vma = v; | |
173 | struct vm_area_struct *tail_vma = priv->tail_vma; | |
174 | ||
175 | (*pos)++; | |
176 | if (vma && (vma != tail_vma) && vma->vm_next) | |
177 | return vma->vm_next; | |
178 | vma_stop(priv, vma); | |
179 | return (vma != tail_vma)? tail_vma: NULL; | |
180 | } | |
181 | ||
182 | static void m_stop(struct seq_file *m, void *v) | |
183 | { | |
184 | struct proc_maps_private *priv = m->private; | |
185 | struct vm_area_struct *vma = v; | |
186 | ||
76597cd3 LT |
187 | if (!IS_ERR(vma)) |
188 | vma_stop(priv, vma); | |
a6198797 MM |
189 | if (priv->task) |
190 | put_task_struct(priv->task); | |
191 | } | |
192 | ||
193 | static int do_maps_open(struct inode *inode, struct file *file, | |
03a44825 | 194 | const struct seq_operations *ops) |
a6198797 MM |
195 | { |
196 | struct proc_maps_private *priv; | |
197 | int ret = -ENOMEM; | |
198 | priv = kzalloc(sizeof(*priv), GFP_KERNEL); | |
199 | if (priv) { | |
200 | priv->pid = proc_pid(inode); | |
201 | ret = seq_open(file, ops); | |
202 | if (!ret) { | |
203 | struct seq_file *m = file->private_data; | |
204 | m->private = priv; | |
205 | } else { | |
206 | kfree(priv); | |
207 | } | |
208 | } | |
209 | return ret; | |
210 | } | |
e070ad49 | 211 | |
b7643757 SP |
212 | static void |
213 | show_map_vma(struct seq_file *m, struct vm_area_struct *vma, int is_pid) | |
1da177e4 | 214 | { |
e070ad49 ML |
215 | struct mm_struct *mm = vma->vm_mm; |
216 | struct file *file = vma->vm_file; | |
b7643757 SP |
217 | struct proc_maps_private *priv = m->private; |
218 | struct task_struct *task = priv->task; | |
ca16d140 | 219 | vm_flags_t flags = vma->vm_flags; |
1da177e4 | 220 | unsigned long ino = 0; |
6260a4b0 | 221 | unsigned long long pgoff = 0; |
a09a79f6 | 222 | unsigned long start, end; |
1da177e4 LT |
223 | dev_t dev = 0; |
224 | int len; | |
b7643757 | 225 | const char *name = NULL; |
1da177e4 LT |
226 | |
227 | if (file) { | |
2fddfeef | 228 | struct inode *inode = vma->vm_file->f_path.dentry->d_inode; |
1da177e4 LT |
229 | dev = inode->i_sb->s_dev; |
230 | ino = inode->i_ino; | |
6260a4b0 | 231 | pgoff = ((loff_t)vma->vm_pgoff) << PAGE_SHIFT; |
1da177e4 LT |
232 | } |
233 | ||
d7824370 LT |
234 | /* We don't show the stack guard page in /proc/maps */ |
235 | start = vma->vm_start; | |
a09a79f6 MP |
236 | if (stack_guard_page_start(vma, start)) |
237 | start += PAGE_SIZE; | |
238 | end = vma->vm_end; | |
239 | if (stack_guard_page_end(vma, end)) | |
240 | end -= PAGE_SIZE; | |
d7824370 | 241 | |
1804dc6e | 242 | seq_printf(m, "%08lx-%08lx %c%c%c%c %08llx %02x:%02x %lu %n", |
d7824370 | 243 | start, |
a09a79f6 | 244 | end, |
1da177e4 LT |
245 | flags & VM_READ ? 'r' : '-', |
246 | flags & VM_WRITE ? 'w' : '-', | |
247 | flags & VM_EXEC ? 'x' : '-', | |
248 | flags & VM_MAYSHARE ? 's' : 'p', | |
6260a4b0 | 249 | pgoff, |
1da177e4 LT |
250 | MAJOR(dev), MINOR(dev), ino, &len); |
251 | ||
252 | /* | |
253 | * Print the dentry name for named mappings, and a | |
254 | * special [heap] marker for the heap: | |
255 | */ | |
e070ad49 | 256 | if (file) { |
1da177e4 | 257 | pad_len_spaces(m, len); |
c32c2f63 | 258 | seq_path(m, &file->f_path, "\n"); |
b7643757 SP |
259 | goto done; |
260 | } | |
261 | ||
262 | name = arch_vma_name(vma); | |
263 | if (!name) { | |
264 | pid_t tid; | |
265 | ||
266 | if (!mm) { | |
267 | name = "[vdso]"; | |
268 | goto done; | |
269 | } | |
270 | ||
271 | if (vma->vm_start <= mm->brk && | |
272 | vma->vm_end >= mm->start_brk) { | |
273 | name = "[heap]"; | |
274 | goto done; | |
275 | } | |
276 | ||
277 | tid = vm_is_stack(task, vma, is_pid); | |
278 | ||
279 | if (tid != 0) { | |
280 | /* | |
281 | * Thread stack in /proc/PID/task/TID/maps or | |
282 | * the main process stack. | |
283 | */ | |
284 | if (!is_pid || (vma->vm_start <= mm->start_stack && | |
285 | vma->vm_end >= mm->start_stack)) { | |
286 | name = "[stack]"; | |
e6e5494c | 287 | } else { |
b7643757 SP |
288 | /* Thread stack in /proc/PID/maps */ |
289 | pad_len_spaces(m, len); | |
290 | seq_printf(m, "[stack:%d]", tid); | |
1da177e4 | 291 | } |
e6e5494c | 292 | } |
b7643757 SP |
293 | } |
294 | ||
295 | done: | |
296 | if (name) { | |
297 | pad_len_spaces(m, len); | |
298 | seq_puts(m, name); | |
1da177e4 LT |
299 | } |
300 | seq_putc(m, '\n'); | |
7c88db0c JK |
301 | } |
302 | ||
b7643757 | 303 | static int show_map(struct seq_file *m, void *v, int is_pid) |
7c88db0c JK |
304 | { |
305 | struct vm_area_struct *vma = v; | |
306 | struct proc_maps_private *priv = m->private; | |
307 | struct task_struct *task = priv->task; | |
308 | ||
b7643757 | 309 | show_map_vma(m, vma, is_pid); |
e070ad49 | 310 | |
e070ad49 | 311 | if (m->count < m->size) /* vma is copied successfully */ |
31db58b3 SW |
312 | m->version = (vma != get_gate_vma(task->mm)) |
313 | ? vma->vm_start : 0; | |
1da177e4 LT |
314 | return 0; |
315 | } | |
316 | ||
b7643757 SP |
317 | static int show_pid_map(struct seq_file *m, void *v) |
318 | { | |
319 | return show_map(m, v, 1); | |
320 | } | |
321 | ||
322 | static int show_tid_map(struct seq_file *m, void *v) | |
323 | { | |
324 | return show_map(m, v, 0); | |
325 | } | |
326 | ||
03a44825 | 327 | static const struct seq_operations proc_pid_maps_op = { |
a6198797 MM |
328 | .start = m_start, |
329 | .next = m_next, | |
330 | .stop = m_stop, | |
b7643757 SP |
331 | .show = show_pid_map |
332 | }; | |
333 | ||
334 | static const struct seq_operations proc_tid_maps_op = { | |
335 | .start = m_start, | |
336 | .next = m_next, | |
337 | .stop = m_stop, | |
338 | .show = show_tid_map | |
a6198797 MM |
339 | }; |
340 | ||
b7643757 | 341 | static int pid_maps_open(struct inode *inode, struct file *file) |
a6198797 MM |
342 | { |
343 | return do_maps_open(inode, file, &proc_pid_maps_op); | |
344 | } | |
345 | ||
b7643757 SP |
346 | static int tid_maps_open(struct inode *inode, struct file *file) |
347 | { | |
348 | return do_maps_open(inode, file, &proc_tid_maps_op); | |
349 | } | |
350 | ||
351 | const struct file_operations proc_pid_maps_operations = { | |
352 | .open = pid_maps_open, | |
353 | .read = seq_read, | |
354 | .llseek = seq_lseek, | |
355 | .release = seq_release_private, | |
356 | }; | |
357 | ||
358 | const struct file_operations proc_tid_maps_operations = { | |
359 | .open = tid_maps_open, | |
a6198797 MM |
360 | .read = seq_read, |
361 | .llseek = seq_lseek, | |
362 | .release = seq_release_private, | |
363 | }; | |
364 | ||
365 | /* | |
366 | * Proportional Set Size(PSS): my share of RSS. | |
367 | * | |
368 | * PSS of a process is the count of pages it has in memory, where each | |
369 | * page is divided by the number of processes sharing it. So if a | |
370 | * process has 1000 pages all to itself, and 1000 shared with one other | |
371 | * process, its PSS will be 1500. | |
372 | * | |
373 | * To keep (accumulated) division errors low, we adopt a 64bit | |
374 | * fixed-point pss counter to minimize division errors. So (pss >> | |
375 | * PSS_SHIFT) would be the real byte count. | |
376 | * | |
377 | * A shift of 12 before division means (assuming 4K page size): | |
378 | * - 1M 3-user-pages add up to 8KB errors; | |
379 | * - supports mapcount up to 2^24, or 16M; | |
380 | * - supports PSS up to 2^52 bytes, or 4PB. | |
381 | */ | |
382 | #define PSS_SHIFT 12 | |
383 | ||
1e883281 | 384 | #ifdef CONFIG_PROC_PAGE_MONITOR |
214e471f | 385 | struct mem_size_stats { |
a6198797 MM |
386 | struct vm_area_struct *vma; |
387 | unsigned long resident; | |
388 | unsigned long shared_clean; | |
389 | unsigned long shared_dirty; | |
390 | unsigned long private_clean; | |
391 | unsigned long private_dirty; | |
392 | unsigned long referenced; | |
b40d4f84 | 393 | unsigned long anonymous; |
4031a219 | 394 | unsigned long anonymous_thp; |
214e471f | 395 | unsigned long swap; |
a6198797 MM |
396 | u64 pss; |
397 | }; | |
398 | ||
ae11c4d9 DH |
399 | |
400 | static void smaps_pte_entry(pte_t ptent, unsigned long addr, | |
3c9acc78 | 401 | unsigned long ptent_size, struct mm_walk *walk) |
ae11c4d9 DH |
402 | { |
403 | struct mem_size_stats *mss = walk->private; | |
404 | struct vm_area_struct *vma = mss->vma; | |
b1d4d9e0 | 405 | struct page *page = NULL; |
ae11c4d9 DH |
406 | int mapcount; |
407 | ||
b1d4d9e0 KK |
408 | if (pte_present(ptent)) { |
409 | page = vm_normal_page(vma, addr, ptent); | |
410 | } else if (is_swap_pte(ptent)) { | |
411 | swp_entry_t swpent = pte_to_swp_entry(ptent); | |
ae11c4d9 | 412 | |
b1d4d9e0 KK |
413 | if (!non_swap_entry(swpent)) |
414 | mss->swap += ptent_size; | |
415 | else if (is_migration_entry(swpent)) | |
416 | page = migration_entry_to_page(swpent); | |
417 | } | |
ae11c4d9 | 418 | |
ae11c4d9 DH |
419 | if (!page) |
420 | return; | |
421 | ||
422 | if (PageAnon(page)) | |
3c9acc78 | 423 | mss->anonymous += ptent_size; |
ae11c4d9 | 424 | |
3c9acc78 | 425 | mss->resident += ptent_size; |
ae11c4d9 DH |
426 | /* Accumulate the size in pages that have been accessed. */ |
427 | if (pte_young(ptent) || PageReferenced(page)) | |
3c9acc78 | 428 | mss->referenced += ptent_size; |
ae11c4d9 DH |
429 | mapcount = page_mapcount(page); |
430 | if (mapcount >= 2) { | |
431 | if (pte_dirty(ptent) || PageDirty(page)) | |
3c9acc78 | 432 | mss->shared_dirty += ptent_size; |
ae11c4d9 | 433 | else |
3c9acc78 DH |
434 | mss->shared_clean += ptent_size; |
435 | mss->pss += (ptent_size << PSS_SHIFT) / mapcount; | |
ae11c4d9 DH |
436 | } else { |
437 | if (pte_dirty(ptent) || PageDirty(page)) | |
3c9acc78 | 438 | mss->private_dirty += ptent_size; |
ae11c4d9 | 439 | else |
3c9acc78 DH |
440 | mss->private_clean += ptent_size; |
441 | mss->pss += (ptent_size << PSS_SHIFT); | |
ae11c4d9 DH |
442 | } |
443 | } | |
444 | ||
b3ae5acb | 445 | static int smaps_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end, |
2165009b | 446 | struct mm_walk *walk) |
e070ad49 | 447 | { |
2165009b | 448 | struct mem_size_stats *mss = walk->private; |
b3ae5acb | 449 | struct vm_area_struct *vma = mss->vma; |
ae11c4d9 | 450 | pte_t *pte; |
705e87c0 | 451 | spinlock_t *ptl; |
e070ad49 | 452 | |
025c5b24 NH |
453 | if (pmd_trans_huge_lock(pmd, vma) == 1) { |
454 | smaps_pte_entry(*(pte_t *)pmd, addr, HPAGE_PMD_SIZE, walk); | |
22e057c5 | 455 | spin_unlock(&walk->mm->page_table_lock); |
025c5b24 NH |
456 | mss->anonymous_thp += HPAGE_PMD_SIZE; |
457 | return 0; | |
22e057c5 | 458 | } |
1a5a9906 AA |
459 | |
460 | if (pmd_trans_unstable(pmd)) | |
461 | return 0; | |
22e057c5 DH |
462 | /* |
463 | * The mmap_sem held all the way back in m_start() is what | |
464 | * keeps khugepaged out of here and from collapsing things | |
465 | * in here. | |
466 | */ | |
705e87c0 | 467 | pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl); |
ae11c4d9 | 468 | for (; addr != end; pte++, addr += PAGE_SIZE) |
3c9acc78 | 469 | smaps_pte_entry(*pte, addr, PAGE_SIZE, walk); |
705e87c0 HD |
470 | pte_unmap_unlock(pte - 1, ptl); |
471 | cond_resched(); | |
b3ae5acb | 472 | return 0; |
e070ad49 ML |
473 | } |
474 | ||
b7643757 | 475 | static int show_smap(struct seq_file *m, void *v, int is_pid) |
e070ad49 | 476 | { |
7c88db0c JK |
477 | struct proc_maps_private *priv = m->private; |
478 | struct task_struct *task = priv->task; | |
e070ad49 | 479 | struct vm_area_struct *vma = v; |
e070ad49 | 480 | struct mem_size_stats mss; |
2165009b DH |
481 | struct mm_walk smaps_walk = { |
482 | .pmd_entry = smaps_pte_range, | |
483 | .mm = vma->vm_mm, | |
484 | .private = &mss, | |
485 | }; | |
e070ad49 ML |
486 | |
487 | memset(&mss, 0, sizeof mss); | |
b3ae5acb | 488 | mss.vma = vma; |
d82ef020 | 489 | /* mmap_sem is held in m_start */ |
5ddfae16 | 490 | if (vma->vm_mm && !is_vm_hugetlb_page(vma)) |
2165009b | 491 | walk_page_range(vma->vm_start, vma->vm_end, &smaps_walk); |
4752c369 | 492 | |
b7643757 | 493 | show_map_vma(m, vma, is_pid); |
4752c369 MM |
494 | |
495 | seq_printf(m, | |
496 | "Size: %8lu kB\n" | |
497 | "Rss: %8lu kB\n" | |
498 | "Pss: %8lu kB\n" | |
499 | "Shared_Clean: %8lu kB\n" | |
500 | "Shared_Dirty: %8lu kB\n" | |
501 | "Private_Clean: %8lu kB\n" | |
502 | "Private_Dirty: %8lu kB\n" | |
214e471f | 503 | "Referenced: %8lu kB\n" |
b40d4f84 | 504 | "Anonymous: %8lu kB\n" |
4031a219 | 505 | "AnonHugePages: %8lu kB\n" |
08fba699 | 506 | "Swap: %8lu kB\n" |
3340289d | 507 | "KernelPageSize: %8lu kB\n" |
2d90508f NK |
508 | "MMUPageSize: %8lu kB\n" |
509 | "Locked: %8lu kB\n", | |
4752c369 MM |
510 | (vma->vm_end - vma->vm_start) >> 10, |
511 | mss.resident >> 10, | |
512 | (unsigned long)(mss.pss >> (10 + PSS_SHIFT)), | |
513 | mss.shared_clean >> 10, | |
514 | mss.shared_dirty >> 10, | |
515 | mss.private_clean >> 10, | |
516 | mss.private_dirty >> 10, | |
214e471f | 517 | mss.referenced >> 10, |
b40d4f84 | 518 | mss.anonymous >> 10, |
4031a219 | 519 | mss.anonymous_thp >> 10, |
08fba699 | 520 | mss.swap >> 10, |
3340289d | 521 | vma_kernel_pagesize(vma) >> 10, |
2d90508f NK |
522 | vma_mmu_pagesize(vma) >> 10, |
523 | (vma->vm_flags & VM_LOCKED) ? | |
524 | (unsigned long)(mss.pss >> (10 + PSS_SHIFT)) : 0); | |
4752c369 | 525 | |
7c88db0c | 526 | if (m->count < m->size) /* vma is copied successfully */ |
31db58b3 SW |
527 | m->version = (vma != get_gate_vma(task->mm)) |
528 | ? vma->vm_start : 0; | |
7c88db0c | 529 | return 0; |
e070ad49 ML |
530 | } |
531 | ||
b7643757 SP |
532 | static int show_pid_smap(struct seq_file *m, void *v) |
533 | { | |
534 | return show_smap(m, v, 1); | |
535 | } | |
536 | ||
537 | static int show_tid_smap(struct seq_file *m, void *v) | |
538 | { | |
539 | return show_smap(m, v, 0); | |
540 | } | |
541 | ||
03a44825 | 542 | static const struct seq_operations proc_pid_smaps_op = { |
a6198797 MM |
543 | .start = m_start, |
544 | .next = m_next, | |
545 | .stop = m_stop, | |
b7643757 SP |
546 | .show = show_pid_smap |
547 | }; | |
548 | ||
549 | static const struct seq_operations proc_tid_smaps_op = { | |
550 | .start = m_start, | |
551 | .next = m_next, | |
552 | .stop = m_stop, | |
553 | .show = show_tid_smap | |
a6198797 MM |
554 | }; |
555 | ||
b7643757 | 556 | static int pid_smaps_open(struct inode *inode, struct file *file) |
a6198797 MM |
557 | { |
558 | return do_maps_open(inode, file, &proc_pid_smaps_op); | |
559 | } | |
560 | ||
b7643757 SP |
561 | static int tid_smaps_open(struct inode *inode, struct file *file) |
562 | { | |
563 | return do_maps_open(inode, file, &proc_tid_smaps_op); | |
564 | } | |
565 | ||
566 | const struct file_operations proc_pid_smaps_operations = { | |
567 | .open = pid_smaps_open, | |
568 | .read = seq_read, | |
569 | .llseek = seq_lseek, | |
570 | .release = seq_release_private, | |
571 | }; | |
572 | ||
573 | const struct file_operations proc_tid_smaps_operations = { | |
574 | .open = tid_smaps_open, | |
a6198797 MM |
575 | .read = seq_read, |
576 | .llseek = seq_lseek, | |
577 | .release = seq_release_private, | |
578 | }; | |
579 | ||
580 | static int clear_refs_pte_range(pmd_t *pmd, unsigned long addr, | |
2165009b | 581 | unsigned long end, struct mm_walk *walk) |
a6198797 | 582 | { |
2165009b | 583 | struct vm_area_struct *vma = walk->private; |
a6198797 MM |
584 | pte_t *pte, ptent; |
585 | spinlock_t *ptl; | |
586 | struct page *page; | |
587 | ||
03319327 | 588 | split_huge_page_pmd(walk->mm, pmd); |
1a5a9906 AA |
589 | if (pmd_trans_unstable(pmd)) |
590 | return 0; | |
03319327 | 591 | |
a6198797 MM |
592 | pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl); |
593 | for (; addr != end; pte++, addr += PAGE_SIZE) { | |
594 | ptent = *pte; | |
595 | if (!pte_present(ptent)) | |
596 | continue; | |
597 | ||
598 | page = vm_normal_page(vma, addr, ptent); | |
599 | if (!page) | |
600 | continue; | |
601 | ||
602 | /* Clear accessed and referenced bits. */ | |
603 | ptep_test_and_clear_young(vma, addr, pte); | |
604 | ClearPageReferenced(page); | |
605 | } | |
606 | pte_unmap_unlock(pte - 1, ptl); | |
607 | cond_resched(); | |
608 | return 0; | |
609 | } | |
610 | ||
398499d5 MB |
611 | #define CLEAR_REFS_ALL 1 |
612 | #define CLEAR_REFS_ANON 2 | |
613 | #define CLEAR_REFS_MAPPED 3 | |
614 | ||
f248dcb3 MM |
615 | static ssize_t clear_refs_write(struct file *file, const char __user *buf, |
616 | size_t count, loff_t *ppos) | |
b813e931 | 617 | { |
f248dcb3 | 618 | struct task_struct *task; |
fb92a4b0 | 619 | char buffer[PROC_NUMBUF]; |
f248dcb3 | 620 | struct mm_struct *mm; |
b813e931 | 621 | struct vm_area_struct *vma; |
0a8cb8e3 AD |
622 | int type; |
623 | int rv; | |
b813e931 | 624 | |
f248dcb3 MM |
625 | memset(buffer, 0, sizeof(buffer)); |
626 | if (count > sizeof(buffer) - 1) | |
627 | count = sizeof(buffer) - 1; | |
628 | if (copy_from_user(buffer, buf, count)) | |
629 | return -EFAULT; | |
0a8cb8e3 AD |
630 | rv = kstrtoint(strstrip(buffer), 10, &type); |
631 | if (rv < 0) | |
632 | return rv; | |
398499d5 | 633 | if (type < CLEAR_REFS_ALL || type > CLEAR_REFS_MAPPED) |
f248dcb3 | 634 | return -EINVAL; |
f248dcb3 MM |
635 | task = get_proc_task(file->f_path.dentry->d_inode); |
636 | if (!task) | |
637 | return -ESRCH; | |
638 | mm = get_task_mm(task); | |
639 | if (mm) { | |
20cbc972 AM |
640 | struct mm_walk clear_refs_walk = { |
641 | .pmd_entry = clear_refs_pte_range, | |
642 | .mm = mm, | |
643 | }; | |
f248dcb3 | 644 | down_read(&mm->mmap_sem); |
2165009b DH |
645 | for (vma = mm->mmap; vma; vma = vma->vm_next) { |
646 | clear_refs_walk.private = vma; | |
398499d5 MB |
647 | if (is_vm_hugetlb_page(vma)) |
648 | continue; | |
649 | /* | |
650 | * Writing 1 to /proc/pid/clear_refs affects all pages. | |
651 | * | |
652 | * Writing 2 to /proc/pid/clear_refs only affects | |
653 | * Anonymous pages. | |
654 | * | |
655 | * Writing 3 to /proc/pid/clear_refs only affects file | |
656 | * mapped pages. | |
657 | */ | |
658 | if (type == CLEAR_REFS_ANON && vma->vm_file) | |
659 | continue; | |
660 | if (type == CLEAR_REFS_MAPPED && !vma->vm_file) | |
661 | continue; | |
662 | walk_page_range(vma->vm_start, vma->vm_end, | |
663 | &clear_refs_walk); | |
2165009b | 664 | } |
f248dcb3 MM |
665 | flush_tlb_mm(mm); |
666 | up_read(&mm->mmap_sem); | |
667 | mmput(mm); | |
668 | } | |
669 | put_task_struct(task); | |
fb92a4b0 VL |
670 | |
671 | return count; | |
b813e931 DR |
672 | } |
673 | ||
f248dcb3 MM |
674 | const struct file_operations proc_clear_refs_operations = { |
675 | .write = clear_refs_write, | |
6038f373 | 676 | .llseek = noop_llseek, |
f248dcb3 MM |
677 | }; |
678 | ||
092b50ba NH |
679 | typedef struct { |
680 | u64 pme; | |
681 | } pagemap_entry_t; | |
682 | ||
85863e47 | 683 | struct pagemapread { |
d82ef020 | 684 | int pos, len; |
092b50ba | 685 | pagemap_entry_t *buffer; |
85863e47 MM |
686 | }; |
687 | ||
5aaabe83 NH |
688 | #define PAGEMAP_WALK_SIZE (PMD_SIZE) |
689 | #define PAGEMAP_WALK_MASK (PMD_MASK) | |
690 | ||
f16278c6 HR |
691 | #define PM_ENTRY_BYTES sizeof(u64) |
692 | #define PM_STATUS_BITS 3 | |
693 | #define PM_STATUS_OFFSET (64 - PM_STATUS_BITS) | |
694 | #define PM_STATUS_MASK (((1LL << PM_STATUS_BITS) - 1) << PM_STATUS_OFFSET) | |
695 | #define PM_STATUS(nr) (((nr) << PM_STATUS_OFFSET) & PM_STATUS_MASK) | |
696 | #define PM_PSHIFT_BITS 6 | |
697 | #define PM_PSHIFT_OFFSET (PM_STATUS_OFFSET - PM_PSHIFT_BITS) | |
698 | #define PM_PSHIFT_MASK (((1LL << PM_PSHIFT_BITS) - 1) << PM_PSHIFT_OFFSET) | |
699 | #define PM_PSHIFT(x) (((u64) (x) << PM_PSHIFT_OFFSET) & PM_PSHIFT_MASK) | |
700 | #define PM_PFRAME_MASK ((1LL << PM_PSHIFT_OFFSET) - 1) | |
701 | #define PM_PFRAME(x) ((x) & PM_PFRAME_MASK) | |
702 | ||
703 | #define PM_PRESENT PM_STATUS(4LL) | |
704 | #define PM_SWAP PM_STATUS(2LL) | |
052fb0d6 | 705 | #define PM_FILE PM_STATUS(1LL) |
f16278c6 | 706 | #define PM_NOT_PRESENT PM_PSHIFT(PAGE_SHIFT) |
85863e47 MM |
707 | #define PM_END_OF_BUFFER 1 |
708 | ||
092b50ba NH |
709 | static inline pagemap_entry_t make_pme(u64 val) |
710 | { | |
711 | return (pagemap_entry_t) { .pme = val }; | |
712 | } | |
713 | ||
714 | static int add_to_pagemap(unsigned long addr, pagemap_entry_t *pme, | |
85863e47 MM |
715 | struct pagemapread *pm) |
716 | { | |
092b50ba | 717 | pm->buffer[pm->pos++] = *pme; |
d82ef020 | 718 | if (pm->pos >= pm->len) |
aae8679b | 719 | return PM_END_OF_BUFFER; |
85863e47 MM |
720 | return 0; |
721 | } | |
722 | ||
723 | static int pagemap_pte_hole(unsigned long start, unsigned long end, | |
2165009b | 724 | struct mm_walk *walk) |
85863e47 | 725 | { |
2165009b | 726 | struct pagemapread *pm = walk->private; |
85863e47 MM |
727 | unsigned long addr; |
728 | int err = 0; | |
092b50ba NH |
729 | pagemap_entry_t pme = make_pme(PM_NOT_PRESENT); |
730 | ||
85863e47 | 731 | for (addr = start; addr < end; addr += PAGE_SIZE) { |
092b50ba | 732 | err = add_to_pagemap(addr, &pme, pm); |
85863e47 MM |
733 | if (err) |
734 | break; | |
735 | } | |
736 | return err; | |
737 | } | |
738 | ||
052fb0d6 KK |
739 | static void pte_to_pagemap_entry(pagemap_entry_t *pme, |
740 | struct vm_area_struct *vma, unsigned long addr, pte_t pte) | |
85863e47 | 741 | { |
052fb0d6 KK |
742 | u64 frame, flags; |
743 | struct page *page = NULL; | |
85863e47 | 744 | |
052fb0d6 KK |
745 | if (pte_present(pte)) { |
746 | frame = pte_pfn(pte); | |
747 | flags = PM_PRESENT; | |
748 | page = vm_normal_page(vma, addr, pte); | |
749 | } else if (is_swap_pte(pte)) { | |
750 | swp_entry_t entry = pte_to_swp_entry(pte); | |
751 | ||
752 | frame = swp_type(entry) | | |
753 | (swp_offset(entry) << MAX_SWAPFILES_SHIFT); | |
754 | flags = PM_SWAP; | |
755 | if (is_migration_entry(entry)) | |
756 | page = migration_entry_to_page(entry); | |
757 | } else { | |
16fbdce6 | 758 | *pme = make_pme(PM_NOT_PRESENT); |
052fb0d6 KK |
759 | return; |
760 | } | |
761 | ||
762 | if (page && !PageAnon(page)) | |
763 | flags |= PM_FILE; | |
764 | ||
765 | *pme = make_pme(PM_PFRAME(frame) | PM_PSHIFT(PAGE_SHIFT) | flags); | |
bcf8039e DH |
766 | } |
767 | ||
5aaabe83 | 768 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
092b50ba NH |
769 | static void thp_pmd_to_pagemap_entry(pagemap_entry_t *pme, |
770 | pmd_t pmd, int offset) | |
5aaabe83 | 771 | { |
5aaabe83 NH |
772 | /* |
773 | * Currently pmd for thp is always present because thp can not be | |
774 | * swapped-out, migrated, or HWPOISONed (split in such cases instead.) | |
775 | * This if-check is just to prepare for future implementation. | |
776 | */ | |
777 | if (pmd_present(pmd)) | |
092b50ba NH |
778 | *pme = make_pme(PM_PFRAME(pmd_pfn(pmd) + offset) |
779 | | PM_PSHIFT(PAGE_SHIFT) | PM_PRESENT); | |
16fbdce6 KK |
780 | else |
781 | *pme = make_pme(PM_NOT_PRESENT); | |
5aaabe83 NH |
782 | } |
783 | #else | |
092b50ba NH |
784 | static inline void thp_pmd_to_pagemap_entry(pagemap_entry_t *pme, |
785 | pmd_t pmd, int offset) | |
5aaabe83 | 786 | { |
5aaabe83 NH |
787 | } |
788 | #endif | |
789 | ||
85863e47 | 790 | static int pagemap_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end, |
2165009b | 791 | struct mm_walk *walk) |
85863e47 | 792 | { |
bcf8039e | 793 | struct vm_area_struct *vma; |
2165009b | 794 | struct pagemapread *pm = walk->private; |
85863e47 MM |
795 | pte_t *pte; |
796 | int err = 0; | |
092b50ba | 797 | pagemap_entry_t pme = make_pme(PM_NOT_PRESENT); |
85863e47 | 798 | |
bcf8039e DH |
799 | /* find the first VMA at or above 'addr' */ |
800 | vma = find_vma(walk->mm, addr); | |
08fa29d9 | 801 | if (vma && pmd_trans_huge_lock(pmd, vma) == 1) { |
025c5b24 NH |
802 | for (; addr != end; addr += PAGE_SIZE) { |
803 | unsigned long offset; | |
804 | ||
805 | offset = (addr & ~PAGEMAP_WALK_MASK) >> | |
806 | PAGE_SHIFT; | |
092b50ba NH |
807 | thp_pmd_to_pagemap_entry(&pme, *pmd, offset); |
808 | err = add_to_pagemap(addr, &pme, pm); | |
025c5b24 NH |
809 | if (err) |
810 | break; | |
5aaabe83 | 811 | } |
5aaabe83 | 812 | spin_unlock(&walk->mm->page_table_lock); |
025c5b24 | 813 | return err; |
5aaabe83 NH |
814 | } |
815 | ||
45f83cef AA |
816 | if (pmd_trans_unstable(pmd)) |
817 | return 0; | |
85863e47 | 818 | for (; addr != end; addr += PAGE_SIZE) { |
bcf8039e DH |
819 | |
820 | /* check to see if we've left 'vma' behind | |
821 | * and need a new, higher one */ | |
16fbdce6 | 822 | if (vma && (addr >= vma->vm_end)) { |
bcf8039e | 823 | vma = find_vma(walk->mm, addr); |
16fbdce6 KK |
824 | pme = make_pme(PM_NOT_PRESENT); |
825 | } | |
bcf8039e DH |
826 | |
827 | /* check that 'vma' actually covers this address, | |
828 | * and that it isn't a huge page vma */ | |
829 | if (vma && (vma->vm_start <= addr) && | |
830 | !is_vm_hugetlb_page(vma)) { | |
831 | pte = pte_offset_map(pmd, addr); | |
052fb0d6 | 832 | pte_to_pagemap_entry(&pme, vma, addr, *pte); |
bcf8039e DH |
833 | /* unmap before userspace copy */ |
834 | pte_unmap(pte); | |
835 | } | |
092b50ba | 836 | err = add_to_pagemap(addr, &pme, pm); |
85863e47 MM |
837 | if (err) |
838 | return err; | |
839 | } | |
840 | ||
841 | cond_resched(); | |
842 | ||
843 | return err; | |
844 | } | |
845 | ||
1a5cb814 | 846 | #ifdef CONFIG_HUGETLB_PAGE |
092b50ba NH |
847 | static void huge_pte_to_pagemap_entry(pagemap_entry_t *pme, |
848 | pte_t pte, int offset) | |
5dc37642 | 849 | { |
5dc37642 | 850 | if (pte_present(pte)) |
092b50ba NH |
851 | *pme = make_pme(PM_PFRAME(pte_pfn(pte) + offset) |
852 | | PM_PSHIFT(PAGE_SHIFT) | PM_PRESENT); | |
16fbdce6 KK |
853 | else |
854 | *pme = make_pme(PM_NOT_PRESENT); | |
5dc37642 NH |
855 | } |
856 | ||
116354d1 NH |
857 | /* This function walks within one hugetlb entry in the single call */ |
858 | static int pagemap_hugetlb_range(pte_t *pte, unsigned long hmask, | |
859 | unsigned long addr, unsigned long end, | |
860 | struct mm_walk *walk) | |
5dc37642 | 861 | { |
5dc37642 | 862 | struct pagemapread *pm = walk->private; |
5dc37642 | 863 | int err = 0; |
16fbdce6 | 864 | pagemap_entry_t pme; |
5dc37642 | 865 | |
5dc37642 | 866 | for (; addr != end; addr += PAGE_SIZE) { |
116354d1 | 867 | int offset = (addr & ~hmask) >> PAGE_SHIFT; |
092b50ba NH |
868 | huge_pte_to_pagemap_entry(&pme, *pte, offset); |
869 | err = add_to_pagemap(addr, &pme, pm); | |
5dc37642 NH |
870 | if (err) |
871 | return err; | |
872 | } | |
873 | ||
874 | cond_resched(); | |
875 | ||
876 | return err; | |
877 | } | |
1a5cb814 | 878 | #endif /* HUGETLB_PAGE */ |
5dc37642 | 879 | |
85863e47 MM |
880 | /* |
881 | * /proc/pid/pagemap - an array mapping virtual pages to pfns | |
882 | * | |
f16278c6 HR |
883 | * For each page in the address space, this file contains one 64-bit entry |
884 | * consisting of the following: | |
885 | * | |
052fb0d6 | 886 | * Bits 0-54 page frame number (PFN) if present |
f16278c6 | 887 | * Bits 0-4 swap type if swapped |
052fb0d6 | 888 | * Bits 5-54 swap offset if swapped |
f16278c6 | 889 | * Bits 55-60 page shift (page size = 1<<page shift) |
052fb0d6 | 890 | * Bit 61 page is file-page or shared-anon |
f16278c6 HR |
891 | * Bit 62 page swapped |
892 | * Bit 63 page present | |
893 | * | |
894 | * If the page is not present but in swap, then the PFN contains an | |
895 | * encoding of the swap file number and the page's offset into the | |
896 | * swap. Unmapped pages return a null PFN. This allows determining | |
85863e47 MM |
897 | * precisely which pages are mapped (or in swap) and comparing mapped |
898 | * pages between processes. | |
899 | * | |
900 | * Efficient users of this interface will use /proc/pid/maps to | |
901 | * determine which areas of memory are actually mapped and llseek to | |
902 | * skip over unmapped regions. | |
903 | */ | |
904 | static ssize_t pagemap_read(struct file *file, char __user *buf, | |
905 | size_t count, loff_t *ppos) | |
906 | { | |
907 | struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode); | |
85863e47 MM |
908 | struct mm_struct *mm; |
909 | struct pagemapread pm; | |
85863e47 | 910 | int ret = -ESRCH; |
ee1e6ab6 | 911 | struct mm_walk pagemap_walk = {}; |
5d7e0d2b AM |
912 | unsigned long src; |
913 | unsigned long svpfn; | |
914 | unsigned long start_vaddr; | |
915 | unsigned long end_vaddr; | |
d82ef020 | 916 | int copied = 0; |
85863e47 MM |
917 | |
918 | if (!task) | |
919 | goto out; | |
920 | ||
85863e47 MM |
921 | ret = -EINVAL; |
922 | /* file position must be aligned */ | |
aae8679b | 923 | if ((*ppos % PM_ENTRY_BYTES) || (count % PM_ENTRY_BYTES)) |
fb39380b | 924 | goto out_task; |
85863e47 MM |
925 | |
926 | ret = 0; | |
08161786 VM |
927 | if (!count) |
928 | goto out_task; | |
929 | ||
d82ef020 KH |
930 | pm.len = PM_ENTRY_BYTES * (PAGEMAP_WALK_SIZE >> PAGE_SHIFT); |
931 | pm.buffer = kmalloc(pm.len, GFP_TEMPORARY); | |
5d7e0d2b | 932 | ret = -ENOMEM; |
d82ef020 | 933 | if (!pm.buffer) |
98bc93e5 KM |
934 | goto out_task; |
935 | ||
e7dcd999 | 936 | mm = mm_access(task, PTRACE_MODE_READ); |
98bc93e5 KM |
937 | ret = PTR_ERR(mm); |
938 | if (!mm || IS_ERR(mm)) | |
939 | goto out_free; | |
85863e47 | 940 | |
5d7e0d2b AM |
941 | pagemap_walk.pmd_entry = pagemap_pte_range; |
942 | pagemap_walk.pte_hole = pagemap_pte_hole; | |
1a5cb814 | 943 | #ifdef CONFIG_HUGETLB_PAGE |
5dc37642 | 944 | pagemap_walk.hugetlb_entry = pagemap_hugetlb_range; |
1a5cb814 | 945 | #endif |
5d7e0d2b AM |
946 | pagemap_walk.mm = mm; |
947 | pagemap_walk.private = ± | |
948 | ||
949 | src = *ppos; | |
950 | svpfn = src / PM_ENTRY_BYTES; | |
951 | start_vaddr = svpfn << PAGE_SHIFT; | |
952 | end_vaddr = TASK_SIZE_OF(task); | |
953 | ||
954 | /* watch out for wraparound */ | |
955 | if (svpfn > TASK_SIZE_OF(task) >> PAGE_SHIFT) | |
956 | start_vaddr = end_vaddr; | |
957 | ||
958 | /* | |
959 | * The odds are that this will stop walking way | |
960 | * before end_vaddr, because the length of the | |
961 | * user buffer is tracked in "pm", and the walk | |
962 | * will stop when we hit the end of the buffer. | |
963 | */ | |
d82ef020 KH |
964 | ret = 0; |
965 | while (count && (start_vaddr < end_vaddr)) { | |
966 | int len; | |
967 | unsigned long end; | |
968 | ||
969 | pm.pos = 0; | |
ea251c1d | 970 | end = (start_vaddr + PAGEMAP_WALK_SIZE) & PAGEMAP_WALK_MASK; |
d82ef020 KH |
971 | /* overflow ? */ |
972 | if (end < start_vaddr || end > end_vaddr) | |
973 | end = end_vaddr; | |
974 | down_read(&mm->mmap_sem); | |
975 | ret = walk_page_range(start_vaddr, end, &pagemap_walk); | |
976 | up_read(&mm->mmap_sem); | |
977 | start_vaddr = end; | |
978 | ||
979 | len = min(count, PM_ENTRY_BYTES * pm.pos); | |
309361e0 | 980 | if (copy_to_user(buf, pm.buffer, len)) { |
d82ef020 | 981 | ret = -EFAULT; |
98bc93e5 | 982 | goto out_mm; |
d82ef020 KH |
983 | } |
984 | copied += len; | |
985 | buf += len; | |
986 | count -= len; | |
85863e47 | 987 | } |
d82ef020 KH |
988 | *ppos += copied; |
989 | if (!ret || ret == PM_END_OF_BUFFER) | |
990 | ret = copied; | |
991 | ||
fb39380b MT |
992 | out_mm: |
993 | mmput(mm); | |
98bc93e5 KM |
994 | out_free: |
995 | kfree(pm.buffer); | |
85863e47 MM |
996 | out_task: |
997 | put_task_struct(task); | |
998 | out: | |
999 | return ret; | |
1000 | } | |
1001 | ||
1002 | const struct file_operations proc_pagemap_operations = { | |
1003 | .llseek = mem_lseek, /* borrow this */ | |
1004 | .read = pagemap_read, | |
1005 | }; | |
1e883281 | 1006 | #endif /* CONFIG_PROC_PAGE_MONITOR */ |
85863e47 | 1007 | |
6e21c8f1 | 1008 | #ifdef CONFIG_NUMA |
6e21c8f1 | 1009 | |
f69ff943 SW |
1010 | struct numa_maps { |
1011 | struct vm_area_struct *vma; | |
1012 | unsigned long pages; | |
1013 | unsigned long anon; | |
1014 | unsigned long active; | |
1015 | unsigned long writeback; | |
1016 | unsigned long mapcount_max; | |
1017 | unsigned long dirty; | |
1018 | unsigned long swapcache; | |
1019 | unsigned long node[MAX_NUMNODES]; | |
1020 | }; | |
1021 | ||
5b52fc89 SW |
1022 | struct numa_maps_private { |
1023 | struct proc_maps_private proc_maps; | |
1024 | struct numa_maps md; | |
1025 | }; | |
1026 | ||
eb4866d0 DH |
1027 | static void gather_stats(struct page *page, struct numa_maps *md, int pte_dirty, |
1028 | unsigned long nr_pages) | |
f69ff943 SW |
1029 | { |
1030 | int count = page_mapcount(page); | |
1031 | ||
eb4866d0 | 1032 | md->pages += nr_pages; |
f69ff943 | 1033 | if (pte_dirty || PageDirty(page)) |
eb4866d0 | 1034 | md->dirty += nr_pages; |
f69ff943 SW |
1035 | |
1036 | if (PageSwapCache(page)) | |
eb4866d0 | 1037 | md->swapcache += nr_pages; |
f69ff943 SW |
1038 | |
1039 | if (PageActive(page) || PageUnevictable(page)) | |
eb4866d0 | 1040 | md->active += nr_pages; |
f69ff943 SW |
1041 | |
1042 | if (PageWriteback(page)) | |
eb4866d0 | 1043 | md->writeback += nr_pages; |
f69ff943 SW |
1044 | |
1045 | if (PageAnon(page)) | |
eb4866d0 | 1046 | md->anon += nr_pages; |
f69ff943 SW |
1047 | |
1048 | if (count > md->mapcount_max) | |
1049 | md->mapcount_max = count; | |
1050 | ||
eb4866d0 | 1051 | md->node[page_to_nid(page)] += nr_pages; |
f69ff943 SW |
1052 | } |
1053 | ||
3200a8aa DH |
1054 | static struct page *can_gather_numa_stats(pte_t pte, struct vm_area_struct *vma, |
1055 | unsigned long addr) | |
1056 | { | |
1057 | struct page *page; | |
1058 | int nid; | |
1059 | ||
1060 | if (!pte_present(pte)) | |
1061 | return NULL; | |
1062 | ||
1063 | page = vm_normal_page(vma, addr, pte); | |
1064 | if (!page) | |
1065 | return NULL; | |
1066 | ||
1067 | if (PageReserved(page)) | |
1068 | return NULL; | |
1069 | ||
1070 | nid = page_to_nid(page); | |
1071 | if (!node_isset(nid, node_states[N_HIGH_MEMORY])) | |
1072 | return NULL; | |
1073 | ||
1074 | return page; | |
1075 | } | |
1076 | ||
f69ff943 SW |
1077 | static int gather_pte_stats(pmd_t *pmd, unsigned long addr, |
1078 | unsigned long end, struct mm_walk *walk) | |
1079 | { | |
1080 | struct numa_maps *md; | |
1081 | spinlock_t *ptl; | |
1082 | pte_t *orig_pte; | |
1083 | pte_t *pte; | |
1084 | ||
1085 | md = walk->private; | |
025c5b24 NH |
1086 | |
1087 | if (pmd_trans_huge_lock(pmd, md->vma) == 1) { | |
1088 | pte_t huge_pte = *(pte_t *)pmd; | |
1089 | struct page *page; | |
1090 | ||
1091 | page = can_gather_numa_stats(huge_pte, md->vma, addr); | |
1092 | if (page) | |
1093 | gather_stats(page, md, pte_dirty(huge_pte), | |
1094 | HPAGE_PMD_SIZE/PAGE_SIZE); | |
32ef4384 | 1095 | spin_unlock(&walk->mm->page_table_lock); |
025c5b24 | 1096 | return 0; |
32ef4384 DH |
1097 | } |
1098 | ||
1a5a9906 AA |
1099 | if (pmd_trans_unstable(pmd)) |
1100 | return 0; | |
f69ff943 SW |
1101 | orig_pte = pte = pte_offset_map_lock(walk->mm, pmd, addr, &ptl); |
1102 | do { | |
3200a8aa | 1103 | struct page *page = can_gather_numa_stats(*pte, md->vma, addr); |
f69ff943 SW |
1104 | if (!page) |
1105 | continue; | |
eb4866d0 | 1106 | gather_stats(page, md, pte_dirty(*pte), 1); |
f69ff943 SW |
1107 | |
1108 | } while (pte++, addr += PAGE_SIZE, addr != end); | |
1109 | pte_unmap_unlock(orig_pte, ptl); | |
1110 | return 0; | |
1111 | } | |
1112 | #ifdef CONFIG_HUGETLB_PAGE | |
1113 | static int gather_hugetbl_stats(pte_t *pte, unsigned long hmask, | |
1114 | unsigned long addr, unsigned long end, struct mm_walk *walk) | |
1115 | { | |
1116 | struct numa_maps *md; | |
1117 | struct page *page; | |
1118 | ||
1119 | if (pte_none(*pte)) | |
1120 | return 0; | |
1121 | ||
1122 | page = pte_page(*pte); | |
1123 | if (!page) | |
1124 | return 0; | |
1125 | ||
1126 | md = walk->private; | |
eb4866d0 | 1127 | gather_stats(page, md, pte_dirty(*pte), 1); |
f69ff943 SW |
1128 | return 0; |
1129 | } | |
1130 | ||
1131 | #else | |
1132 | static int gather_hugetbl_stats(pte_t *pte, unsigned long hmask, | |
1133 | unsigned long addr, unsigned long end, struct mm_walk *walk) | |
1134 | { | |
1135 | return 0; | |
1136 | } | |
1137 | #endif | |
1138 | ||
1139 | /* | |
1140 | * Display pages allocated per node and memory policy via /proc. | |
1141 | */ | |
b7643757 | 1142 | static int show_numa_map(struct seq_file *m, void *v, int is_pid) |
f69ff943 | 1143 | { |
5b52fc89 SW |
1144 | struct numa_maps_private *numa_priv = m->private; |
1145 | struct proc_maps_private *proc_priv = &numa_priv->proc_maps; | |
f69ff943 | 1146 | struct vm_area_struct *vma = v; |
5b52fc89 | 1147 | struct numa_maps *md = &numa_priv->md; |
f69ff943 SW |
1148 | struct file *file = vma->vm_file; |
1149 | struct mm_struct *mm = vma->vm_mm; | |
1150 | struct mm_walk walk = {}; | |
1151 | struct mempolicy *pol; | |
1152 | int n; | |
1153 | char buffer[50]; | |
1154 | ||
1155 | if (!mm) | |
1156 | return 0; | |
1157 | ||
5b52fc89 SW |
1158 | /* Ensure we start with an empty set of numa_maps statistics. */ |
1159 | memset(md, 0, sizeof(*md)); | |
f69ff943 SW |
1160 | |
1161 | md->vma = vma; | |
1162 | ||
1163 | walk.hugetlb_entry = gather_hugetbl_stats; | |
1164 | walk.pmd_entry = gather_pte_stats; | |
1165 | walk.private = md; | |
1166 | walk.mm = mm; | |
1167 | ||
5b52fc89 | 1168 | pol = get_vma_policy(proc_priv->task, vma, vma->vm_start); |
f69ff943 SW |
1169 | mpol_to_str(buffer, sizeof(buffer), pol, 0); |
1170 | mpol_cond_put(pol); | |
1171 | ||
1172 | seq_printf(m, "%08lx %s", vma->vm_start, buffer); | |
1173 | ||
1174 | if (file) { | |
1175 | seq_printf(m, " file="); | |
1176 | seq_path(m, &file->f_path, "\n\t= "); | |
1177 | } else if (vma->vm_start <= mm->brk && vma->vm_end >= mm->start_brk) { | |
1178 | seq_printf(m, " heap"); | |
b7643757 SP |
1179 | } else { |
1180 | pid_t tid = vm_is_stack(proc_priv->task, vma, is_pid); | |
1181 | if (tid != 0) { | |
1182 | /* | |
1183 | * Thread stack in /proc/PID/task/TID/maps or | |
1184 | * the main process stack. | |
1185 | */ | |
1186 | if (!is_pid || (vma->vm_start <= mm->start_stack && | |
1187 | vma->vm_end >= mm->start_stack)) | |
1188 | seq_printf(m, " stack"); | |
1189 | else | |
1190 | seq_printf(m, " stack:%d", tid); | |
1191 | } | |
f69ff943 SW |
1192 | } |
1193 | ||
fc360bd9 AM |
1194 | if (is_vm_hugetlb_page(vma)) |
1195 | seq_printf(m, " huge"); | |
1196 | ||
f69ff943 SW |
1197 | walk_page_range(vma->vm_start, vma->vm_end, &walk); |
1198 | ||
1199 | if (!md->pages) | |
1200 | goto out; | |
1201 | ||
1202 | if (md->anon) | |
1203 | seq_printf(m, " anon=%lu", md->anon); | |
1204 | ||
1205 | if (md->dirty) | |
1206 | seq_printf(m, " dirty=%lu", md->dirty); | |
1207 | ||
1208 | if (md->pages != md->anon && md->pages != md->dirty) | |
1209 | seq_printf(m, " mapped=%lu", md->pages); | |
1210 | ||
1211 | if (md->mapcount_max > 1) | |
1212 | seq_printf(m, " mapmax=%lu", md->mapcount_max); | |
1213 | ||
1214 | if (md->swapcache) | |
1215 | seq_printf(m, " swapcache=%lu", md->swapcache); | |
1216 | ||
1217 | if (md->active < md->pages && !is_vm_hugetlb_page(vma)) | |
1218 | seq_printf(m, " active=%lu", md->active); | |
1219 | ||
1220 | if (md->writeback) | |
1221 | seq_printf(m, " writeback=%lu", md->writeback); | |
1222 | ||
1223 | for_each_node_state(n, N_HIGH_MEMORY) | |
1224 | if (md->node[n]) | |
1225 | seq_printf(m, " N%d=%lu", n, md->node[n]); | |
1226 | out: | |
1227 | seq_putc(m, '\n'); | |
f69ff943 SW |
1228 | |
1229 | if (m->count < m->size) | |
5b52fc89 | 1230 | m->version = (vma != proc_priv->tail_vma) ? vma->vm_start : 0; |
f69ff943 SW |
1231 | return 0; |
1232 | } | |
5b52fc89 | 1233 | |
b7643757 SP |
1234 | static int show_pid_numa_map(struct seq_file *m, void *v) |
1235 | { | |
1236 | return show_numa_map(m, v, 1); | |
1237 | } | |
1238 | ||
1239 | static int show_tid_numa_map(struct seq_file *m, void *v) | |
1240 | { | |
1241 | return show_numa_map(m, v, 0); | |
1242 | } | |
1243 | ||
03a44825 | 1244 | static const struct seq_operations proc_pid_numa_maps_op = { |
b7643757 SP |
1245 | .start = m_start, |
1246 | .next = m_next, | |
1247 | .stop = m_stop, | |
1248 | .show = show_pid_numa_map, | |
6e21c8f1 | 1249 | }; |
662795de | 1250 | |
b7643757 SP |
1251 | static const struct seq_operations proc_tid_numa_maps_op = { |
1252 | .start = m_start, | |
1253 | .next = m_next, | |
1254 | .stop = m_stop, | |
1255 | .show = show_tid_numa_map, | |
1256 | }; | |
1257 | ||
1258 | static int numa_maps_open(struct inode *inode, struct file *file, | |
1259 | const struct seq_operations *ops) | |
662795de | 1260 | { |
5b52fc89 SW |
1261 | struct numa_maps_private *priv; |
1262 | int ret = -ENOMEM; | |
1263 | priv = kzalloc(sizeof(*priv), GFP_KERNEL); | |
1264 | if (priv) { | |
1265 | priv->proc_maps.pid = proc_pid(inode); | |
b7643757 | 1266 | ret = seq_open(file, ops); |
5b52fc89 SW |
1267 | if (!ret) { |
1268 | struct seq_file *m = file->private_data; | |
1269 | m->private = priv; | |
1270 | } else { | |
1271 | kfree(priv); | |
1272 | } | |
1273 | } | |
1274 | return ret; | |
662795de EB |
1275 | } |
1276 | ||
b7643757 SP |
1277 | static int pid_numa_maps_open(struct inode *inode, struct file *file) |
1278 | { | |
1279 | return numa_maps_open(inode, file, &proc_pid_numa_maps_op); | |
1280 | } | |
1281 | ||
1282 | static int tid_numa_maps_open(struct inode *inode, struct file *file) | |
1283 | { | |
1284 | return numa_maps_open(inode, file, &proc_tid_numa_maps_op); | |
1285 | } | |
1286 | ||
1287 | const struct file_operations proc_pid_numa_maps_operations = { | |
1288 | .open = pid_numa_maps_open, | |
1289 | .read = seq_read, | |
1290 | .llseek = seq_lseek, | |
1291 | .release = seq_release_private, | |
1292 | }; | |
1293 | ||
1294 | const struct file_operations proc_tid_numa_maps_operations = { | |
1295 | .open = tid_numa_maps_open, | |
662795de EB |
1296 | .read = seq_read, |
1297 | .llseek = seq_lseek, | |
99f89551 | 1298 | .release = seq_release_private, |
662795de | 1299 | }; |
f69ff943 | 1300 | #endif /* CONFIG_NUMA */ |