]>
Commit | Line | Data |
---|---|---|
1da177e4 | 1 | #include <linux/mm.h> |
615d6e87 | 2 | #include <linux/vmacache.h> |
1da177e4 | 3 | #include <linux/hugetlb.h> |
22e057c5 | 4 | #include <linux/huge_mm.h> |
1da177e4 LT |
5 | #include <linux/mount.h> |
6 | #include <linux/seq_file.h> | |
e070ad49 | 7 | #include <linux/highmem.h> |
5096add8 | 8 | #include <linux/ptrace.h> |
5a0e3ad6 | 9 | #include <linux/slab.h> |
6e21c8f1 CL |
10 | #include <linux/pagemap.h> |
11 | #include <linux/mempolicy.h> | |
22e057c5 | 12 | #include <linux/rmap.h> |
85863e47 MM |
13 | #include <linux/swap.h> |
14 | #include <linux/swapops.h> | |
0f8975ec | 15 | #include <linux/mmu_notifier.h> |
e070ad49 | 16 | |
1da177e4 LT |
17 | #include <asm/elf.h> |
18 | #include <asm/uaccess.h> | |
e070ad49 | 19 | #include <asm/tlbflush.h> |
1da177e4 LT |
20 | #include "internal.h" |
21 | ||
df5f8314 | 22 | void task_mem(struct seq_file *m, struct mm_struct *mm) |
1da177e4 | 23 | { |
dc6c9a35 | 24 | unsigned long data, text, lib, swap, ptes, pmds; |
365e9c87 HD |
25 | unsigned long hiwater_vm, total_vm, hiwater_rss, total_rss; |
26 | ||
27 | /* | |
28 | * Note: to minimize their overhead, mm maintains hiwater_vm and | |
29 | * hiwater_rss only when about to *lower* total_vm or rss. Any | |
30 | * collector of these hiwater stats must therefore get total_vm | |
31 | * and rss too, which will usually be the higher. Barriers? not | |
32 | * worth the effort, such snapshots can always be inconsistent. | |
33 | */ | |
34 | hiwater_vm = total_vm = mm->total_vm; | |
35 | if (hiwater_vm < mm->hiwater_vm) | |
36 | hiwater_vm = mm->hiwater_vm; | |
37 | hiwater_rss = total_rss = get_mm_rss(mm); | |
38 | if (hiwater_rss < mm->hiwater_rss) | |
39 | hiwater_rss = mm->hiwater_rss; | |
1da177e4 LT |
40 | |
41 | data = mm->total_vm - mm->shared_vm - mm->stack_vm; | |
42 | text = (PAGE_ALIGN(mm->end_code) - (mm->start_code & PAGE_MASK)) >> 10; | |
43 | lib = (mm->exec_vm << (PAGE_SHIFT-10)) - text; | |
b084d435 | 44 | swap = get_mm_counter(mm, MM_SWAPENTS); |
dc6c9a35 KS |
45 | ptes = PTRS_PER_PTE * sizeof(pte_t) * atomic_long_read(&mm->nr_ptes); |
46 | pmds = PTRS_PER_PMD * sizeof(pmd_t) * mm_nr_pmds(mm); | |
df5f8314 | 47 | seq_printf(m, |
365e9c87 | 48 | "VmPeak:\t%8lu kB\n" |
1da177e4 LT |
49 | "VmSize:\t%8lu kB\n" |
50 | "VmLck:\t%8lu kB\n" | |
bc3e53f6 | 51 | "VmPin:\t%8lu kB\n" |
365e9c87 | 52 | "VmHWM:\t%8lu kB\n" |
1da177e4 LT |
53 | "VmRSS:\t%8lu kB\n" |
54 | "VmData:\t%8lu kB\n" | |
55 | "VmStk:\t%8lu kB\n" | |
56 | "VmExe:\t%8lu kB\n" | |
57 | "VmLib:\t%8lu kB\n" | |
b084d435 | 58 | "VmPTE:\t%8lu kB\n" |
dc6c9a35 | 59 | "VmPMD:\t%8lu kB\n" |
b084d435 | 60 | "VmSwap:\t%8lu kB\n", |
365e9c87 | 61 | hiwater_vm << (PAGE_SHIFT-10), |
314e51b9 | 62 | total_vm << (PAGE_SHIFT-10), |
1da177e4 | 63 | mm->locked_vm << (PAGE_SHIFT-10), |
bc3e53f6 | 64 | mm->pinned_vm << (PAGE_SHIFT-10), |
365e9c87 HD |
65 | hiwater_rss << (PAGE_SHIFT-10), |
66 | total_rss << (PAGE_SHIFT-10), | |
1da177e4 LT |
67 | data << (PAGE_SHIFT-10), |
68 | mm->stack_vm << (PAGE_SHIFT-10), text, lib, | |
dc6c9a35 KS |
69 | ptes >> 10, |
70 | pmds >> 10, | |
b084d435 | 71 | swap << (PAGE_SHIFT-10)); |
1da177e4 LT |
72 | } |
73 | ||
74 | unsigned long task_vsize(struct mm_struct *mm) | |
75 | { | |
76 | return PAGE_SIZE * mm->total_vm; | |
77 | } | |
78 | ||
a2ade7b6 AD |
79 | unsigned long task_statm(struct mm_struct *mm, |
80 | unsigned long *shared, unsigned long *text, | |
81 | unsigned long *data, unsigned long *resident) | |
1da177e4 | 82 | { |
d559db08 | 83 | *shared = get_mm_counter(mm, MM_FILEPAGES); |
1da177e4 LT |
84 | *text = (PAGE_ALIGN(mm->end_code) - (mm->start_code & PAGE_MASK)) |
85 | >> PAGE_SHIFT; | |
86 | *data = mm->total_vm - mm->shared_vm; | |
d559db08 | 87 | *resident = *shared + get_mm_counter(mm, MM_ANONPAGES); |
1da177e4 LT |
88 | return mm->total_vm; |
89 | } | |
90 | ||
9e781440 KH |
91 | #ifdef CONFIG_NUMA |
92 | /* | |
498f2371 | 93 | * Save get_task_policy() for show_numa_map(). |
9e781440 KH |
94 | */ |
95 | static void hold_task_mempolicy(struct proc_maps_private *priv) | |
96 | { | |
97 | struct task_struct *task = priv->task; | |
98 | ||
99 | task_lock(task); | |
498f2371 | 100 | priv->task_mempolicy = get_task_policy(task); |
9e781440 KH |
101 | mpol_get(priv->task_mempolicy); |
102 | task_unlock(task); | |
103 | } | |
104 | static void release_task_mempolicy(struct proc_maps_private *priv) | |
105 | { | |
106 | mpol_put(priv->task_mempolicy); | |
107 | } | |
108 | #else | |
109 | static void hold_task_mempolicy(struct proc_maps_private *priv) | |
110 | { | |
111 | } | |
112 | static void release_task_mempolicy(struct proc_maps_private *priv) | |
113 | { | |
114 | } | |
115 | #endif | |
116 | ||
59b4bf12 | 117 | static void vma_stop(struct proc_maps_private *priv) |
a6198797 | 118 | { |
59b4bf12 ON |
119 | struct mm_struct *mm = priv->mm; |
120 | ||
121 | release_task_mempolicy(priv); | |
122 | up_read(&mm->mmap_sem); | |
123 | mmput(mm); | |
a6198797 | 124 | } |
ec4dd3eb | 125 | |
ad2a00e4 ON |
126 | static struct vm_area_struct * |
127 | m_next_vma(struct proc_maps_private *priv, struct vm_area_struct *vma) | |
128 | { | |
129 | if (vma == priv->tail_vma) | |
130 | return NULL; | |
131 | return vma->vm_next ?: priv->tail_vma; | |
132 | } | |
133 | ||
b8c20a9b ON |
134 | static void m_cache_vma(struct seq_file *m, struct vm_area_struct *vma) |
135 | { | |
136 | if (m->count < m->size) /* vma is copied successfully */ | |
137 | m->version = m_next_vma(m->private, vma) ? vma->vm_start : -1UL; | |
138 | } | |
139 | ||
0c255321 | 140 | static void *m_start(struct seq_file *m, loff_t *ppos) |
e070ad49 | 141 | { |
a6198797 | 142 | struct proc_maps_private *priv = m->private; |
b8c20a9b | 143 | unsigned long last_addr = m->version; |
a6198797 | 144 | struct mm_struct *mm; |
0c255321 ON |
145 | struct vm_area_struct *vma; |
146 | unsigned int pos = *ppos; | |
a6198797 | 147 | |
b8c20a9b ON |
148 | /* See m_cache_vma(). Zero at the start or after lseek. */ |
149 | if (last_addr == -1UL) | |
150 | return NULL; | |
151 | ||
2c03376d | 152 | priv->task = get_proc_task(priv->inode); |
a6198797 | 153 | if (!priv->task) |
ec6fd8a4 | 154 | return ERR_PTR(-ESRCH); |
a6198797 | 155 | |
29a40ace ON |
156 | mm = priv->mm; |
157 | if (!mm || !atomic_inc_not_zero(&mm->mm_users)) | |
158 | return NULL; | |
a6198797 | 159 | |
0c255321 | 160 | down_read(&mm->mmap_sem); |
9e781440 | 161 | hold_task_mempolicy(priv); |
0c255321 | 162 | priv->tail_vma = get_gate_vma(mm); |
a6198797 | 163 | |
b8c20a9b ON |
164 | if (last_addr) { |
165 | vma = find_vma(mm, last_addr); | |
166 | if (vma && (vma = m_next_vma(priv, vma))) | |
167 | return vma; | |
168 | } | |
169 | ||
170 | m->version = 0; | |
0c255321 | 171 | if (pos < mm->map_count) { |
557c2d8a ON |
172 | for (vma = mm->mmap; pos; pos--) { |
173 | m->version = vma->vm_start; | |
a6198797 | 174 | vma = vma->vm_next; |
557c2d8a | 175 | } |
a6198797 | 176 | return vma; |
0c255321 | 177 | } |
a6198797 | 178 | |
557c2d8a | 179 | /* we do not bother to update m->version in this case */ |
0c255321 ON |
180 | if (pos == mm->map_count && priv->tail_vma) |
181 | return priv->tail_vma; | |
59b4bf12 ON |
182 | |
183 | vma_stop(priv); | |
184 | return NULL; | |
a6198797 MM |
185 | } |
186 | ||
187 | static void *m_next(struct seq_file *m, void *v, loff_t *pos) | |
188 | { | |
189 | struct proc_maps_private *priv = m->private; | |
ad2a00e4 | 190 | struct vm_area_struct *next; |
a6198797 MM |
191 | |
192 | (*pos)++; | |
ad2a00e4 | 193 | next = m_next_vma(priv, v); |
59b4bf12 ON |
194 | if (!next) |
195 | vma_stop(priv); | |
196 | return next; | |
a6198797 MM |
197 | } |
198 | ||
199 | static void m_stop(struct seq_file *m, void *v) | |
200 | { | |
201 | struct proc_maps_private *priv = m->private; | |
a6198797 | 202 | |
59b4bf12 ON |
203 | if (!IS_ERR_OR_NULL(v)) |
204 | vma_stop(priv); | |
0d5f5f45 | 205 | if (priv->task) { |
a6198797 | 206 | put_task_struct(priv->task); |
0d5f5f45 ON |
207 | priv->task = NULL; |
208 | } | |
a6198797 MM |
209 | } |
210 | ||
4db7d0ee ON |
211 | static int proc_maps_open(struct inode *inode, struct file *file, |
212 | const struct seq_operations *ops, int psize) | |
213 | { | |
214 | struct proc_maps_private *priv = __seq_open_private(file, ops, psize); | |
215 | ||
216 | if (!priv) | |
217 | return -ENOMEM; | |
218 | ||
2c03376d | 219 | priv->inode = inode; |
29a40ace ON |
220 | priv->mm = proc_mem_open(inode, PTRACE_MODE_READ); |
221 | if (IS_ERR(priv->mm)) { | |
222 | int err = PTR_ERR(priv->mm); | |
223 | ||
224 | seq_release_private(inode, file); | |
225 | return err; | |
226 | } | |
227 | ||
4db7d0ee ON |
228 | return 0; |
229 | } | |
230 | ||
29a40ace ON |
231 | static int proc_map_release(struct inode *inode, struct file *file) |
232 | { | |
233 | struct seq_file *seq = file->private_data; | |
234 | struct proc_maps_private *priv = seq->private; | |
235 | ||
236 | if (priv->mm) | |
237 | mmdrop(priv->mm); | |
238 | ||
239 | return seq_release_private(inode, file); | |
240 | } | |
241 | ||
a6198797 | 242 | static int do_maps_open(struct inode *inode, struct file *file, |
03a44825 | 243 | const struct seq_operations *ops) |
a6198797 | 244 | { |
4db7d0ee ON |
245 | return proc_maps_open(inode, file, ops, |
246 | sizeof(struct proc_maps_private)); | |
a6198797 | 247 | } |
e070ad49 | 248 | |
58cb6548 ON |
249 | static pid_t pid_of_stack(struct proc_maps_private *priv, |
250 | struct vm_area_struct *vma, bool is_pid) | |
251 | { | |
252 | struct inode *inode = priv->inode; | |
253 | struct task_struct *task; | |
254 | pid_t ret = 0; | |
255 | ||
256 | rcu_read_lock(); | |
257 | task = pid_task(proc_pid(inode), PIDTYPE_PID); | |
258 | if (task) { | |
259 | task = task_of_stack(task, vma, is_pid); | |
260 | if (task) | |
261 | ret = task_pid_nr_ns(task, inode->i_sb->s_fs_info); | |
262 | } | |
263 | rcu_read_unlock(); | |
264 | ||
265 | return ret; | |
266 | } | |
267 | ||
b7643757 SP |
268 | static void |
269 | show_map_vma(struct seq_file *m, struct vm_area_struct *vma, int is_pid) | |
1da177e4 | 270 | { |
e070ad49 ML |
271 | struct mm_struct *mm = vma->vm_mm; |
272 | struct file *file = vma->vm_file; | |
b7643757 | 273 | struct proc_maps_private *priv = m->private; |
ca16d140 | 274 | vm_flags_t flags = vma->vm_flags; |
1da177e4 | 275 | unsigned long ino = 0; |
6260a4b0 | 276 | unsigned long long pgoff = 0; |
a09a79f6 | 277 | unsigned long start, end; |
1da177e4 | 278 | dev_t dev = 0; |
b7643757 | 279 | const char *name = NULL; |
1da177e4 LT |
280 | |
281 | if (file) { | |
496ad9aa | 282 | struct inode *inode = file_inode(vma->vm_file); |
1da177e4 LT |
283 | dev = inode->i_sb->s_dev; |
284 | ino = inode->i_ino; | |
6260a4b0 | 285 | pgoff = ((loff_t)vma->vm_pgoff) << PAGE_SHIFT; |
1da177e4 LT |
286 | } |
287 | ||
d7824370 LT |
288 | /* We don't show the stack guard page in /proc/maps */ |
289 | start = vma->vm_start; | |
a09a79f6 MP |
290 | if (stack_guard_page_start(vma, start)) |
291 | start += PAGE_SIZE; | |
292 | end = vma->vm_end; | |
293 | if (stack_guard_page_end(vma, end)) | |
294 | end -= PAGE_SIZE; | |
d7824370 | 295 | |
652586df TH |
296 | seq_setwidth(m, 25 + sizeof(void *) * 6 - 1); |
297 | seq_printf(m, "%08lx-%08lx %c%c%c%c %08llx %02x:%02x %lu ", | |
d7824370 | 298 | start, |
a09a79f6 | 299 | end, |
1da177e4 LT |
300 | flags & VM_READ ? 'r' : '-', |
301 | flags & VM_WRITE ? 'w' : '-', | |
302 | flags & VM_EXEC ? 'x' : '-', | |
303 | flags & VM_MAYSHARE ? 's' : 'p', | |
6260a4b0 | 304 | pgoff, |
652586df | 305 | MAJOR(dev), MINOR(dev), ino); |
1da177e4 LT |
306 | |
307 | /* | |
308 | * Print the dentry name for named mappings, and a | |
309 | * special [heap] marker for the heap: | |
310 | */ | |
e070ad49 | 311 | if (file) { |
652586df | 312 | seq_pad(m, ' '); |
2726d566 | 313 | seq_file_path(m, file, "\n"); |
b7643757 SP |
314 | goto done; |
315 | } | |
316 | ||
78d683e8 AL |
317 | if (vma->vm_ops && vma->vm_ops->name) { |
318 | name = vma->vm_ops->name(vma); | |
319 | if (name) | |
320 | goto done; | |
321 | } | |
322 | ||
b7643757 SP |
323 | name = arch_vma_name(vma); |
324 | if (!name) { | |
325 | pid_t tid; | |
326 | ||
327 | if (!mm) { | |
328 | name = "[vdso]"; | |
329 | goto done; | |
330 | } | |
331 | ||
332 | if (vma->vm_start <= mm->brk && | |
333 | vma->vm_end >= mm->start_brk) { | |
334 | name = "[heap]"; | |
335 | goto done; | |
336 | } | |
337 | ||
58cb6548 | 338 | tid = pid_of_stack(priv, vma, is_pid); |
b7643757 SP |
339 | if (tid != 0) { |
340 | /* | |
341 | * Thread stack in /proc/PID/task/TID/maps or | |
342 | * the main process stack. | |
343 | */ | |
344 | if (!is_pid || (vma->vm_start <= mm->start_stack && | |
345 | vma->vm_end >= mm->start_stack)) { | |
346 | name = "[stack]"; | |
e6e5494c | 347 | } else { |
b7643757 | 348 | /* Thread stack in /proc/PID/maps */ |
652586df | 349 | seq_pad(m, ' '); |
b7643757 | 350 | seq_printf(m, "[stack:%d]", tid); |
1da177e4 | 351 | } |
e6e5494c | 352 | } |
b7643757 SP |
353 | } |
354 | ||
355 | done: | |
356 | if (name) { | |
652586df | 357 | seq_pad(m, ' '); |
b7643757 | 358 | seq_puts(m, name); |
1da177e4 LT |
359 | } |
360 | seq_putc(m, '\n'); | |
7c88db0c JK |
361 | } |
362 | ||
b7643757 | 363 | static int show_map(struct seq_file *m, void *v, int is_pid) |
7c88db0c | 364 | { |
ebb6cdde | 365 | show_map_vma(m, v, is_pid); |
b8c20a9b | 366 | m_cache_vma(m, v); |
1da177e4 LT |
367 | return 0; |
368 | } | |
369 | ||
b7643757 SP |
370 | static int show_pid_map(struct seq_file *m, void *v) |
371 | { | |
372 | return show_map(m, v, 1); | |
373 | } | |
374 | ||
375 | static int show_tid_map(struct seq_file *m, void *v) | |
376 | { | |
377 | return show_map(m, v, 0); | |
378 | } | |
379 | ||
03a44825 | 380 | static const struct seq_operations proc_pid_maps_op = { |
a6198797 MM |
381 | .start = m_start, |
382 | .next = m_next, | |
383 | .stop = m_stop, | |
b7643757 SP |
384 | .show = show_pid_map |
385 | }; | |
386 | ||
387 | static const struct seq_operations proc_tid_maps_op = { | |
388 | .start = m_start, | |
389 | .next = m_next, | |
390 | .stop = m_stop, | |
391 | .show = show_tid_map | |
a6198797 MM |
392 | }; |
393 | ||
b7643757 | 394 | static int pid_maps_open(struct inode *inode, struct file *file) |
a6198797 MM |
395 | { |
396 | return do_maps_open(inode, file, &proc_pid_maps_op); | |
397 | } | |
398 | ||
b7643757 SP |
399 | static int tid_maps_open(struct inode *inode, struct file *file) |
400 | { | |
401 | return do_maps_open(inode, file, &proc_tid_maps_op); | |
402 | } | |
403 | ||
404 | const struct file_operations proc_pid_maps_operations = { | |
405 | .open = pid_maps_open, | |
406 | .read = seq_read, | |
407 | .llseek = seq_lseek, | |
29a40ace | 408 | .release = proc_map_release, |
b7643757 SP |
409 | }; |
410 | ||
411 | const struct file_operations proc_tid_maps_operations = { | |
412 | .open = tid_maps_open, | |
a6198797 MM |
413 | .read = seq_read, |
414 | .llseek = seq_lseek, | |
29a40ace | 415 | .release = proc_map_release, |
a6198797 MM |
416 | }; |
417 | ||
418 | /* | |
419 | * Proportional Set Size(PSS): my share of RSS. | |
420 | * | |
421 | * PSS of a process is the count of pages it has in memory, where each | |
422 | * page is divided by the number of processes sharing it. So if a | |
423 | * process has 1000 pages all to itself, and 1000 shared with one other | |
424 | * process, its PSS will be 1500. | |
425 | * | |
426 | * To keep (accumulated) division errors low, we adopt a 64bit | |
427 | * fixed-point pss counter to minimize division errors. So (pss >> | |
428 | * PSS_SHIFT) would be the real byte count. | |
429 | * | |
430 | * A shift of 12 before division means (assuming 4K page size): | |
431 | * - 1M 3-user-pages add up to 8KB errors; | |
432 | * - supports mapcount up to 2^24, or 16M; | |
433 | * - supports PSS up to 2^52 bytes, or 4PB. | |
434 | */ | |
435 | #define PSS_SHIFT 12 | |
436 | ||
1e883281 | 437 | #ifdef CONFIG_PROC_PAGE_MONITOR |
214e471f | 438 | struct mem_size_stats { |
a6198797 MM |
439 | unsigned long resident; |
440 | unsigned long shared_clean; | |
441 | unsigned long shared_dirty; | |
442 | unsigned long private_clean; | |
443 | unsigned long private_dirty; | |
444 | unsigned long referenced; | |
b40d4f84 | 445 | unsigned long anonymous; |
4031a219 | 446 | unsigned long anonymous_thp; |
214e471f | 447 | unsigned long swap; |
a6198797 MM |
448 | u64 pss; |
449 | }; | |
450 | ||
c164e038 KS |
451 | static void smaps_account(struct mem_size_stats *mss, struct page *page, |
452 | unsigned long size, bool young, bool dirty) | |
453 | { | |
454 | int mapcount; | |
455 | ||
456 | if (PageAnon(page)) | |
457 | mss->anonymous += size; | |
458 | ||
459 | mss->resident += size; | |
460 | /* Accumulate the size in pages that have been accessed. */ | |
461 | if (young || PageReferenced(page)) | |
462 | mss->referenced += size; | |
463 | mapcount = page_mapcount(page); | |
464 | if (mapcount >= 2) { | |
465 | u64 pss_delta; | |
466 | ||
467 | if (dirty || PageDirty(page)) | |
468 | mss->shared_dirty += size; | |
469 | else | |
470 | mss->shared_clean += size; | |
471 | pss_delta = (u64)size << PSS_SHIFT; | |
472 | do_div(pss_delta, mapcount); | |
473 | mss->pss += pss_delta; | |
474 | } else { | |
475 | if (dirty || PageDirty(page)) | |
476 | mss->private_dirty += size; | |
477 | else | |
478 | mss->private_clean += size; | |
479 | mss->pss += (u64)size << PSS_SHIFT; | |
480 | } | |
481 | } | |
ae11c4d9 | 482 | |
c164e038 KS |
483 | static void smaps_pte_entry(pte_t *pte, unsigned long addr, |
484 | struct mm_walk *walk) | |
ae11c4d9 DH |
485 | { |
486 | struct mem_size_stats *mss = walk->private; | |
14eb6fdd | 487 | struct vm_area_struct *vma = walk->vma; |
b1d4d9e0 | 488 | struct page *page = NULL; |
ae11c4d9 | 489 | |
c164e038 KS |
490 | if (pte_present(*pte)) { |
491 | page = vm_normal_page(vma, addr, *pte); | |
492 | } else if (is_swap_pte(*pte)) { | |
493 | swp_entry_t swpent = pte_to_swp_entry(*pte); | |
ae11c4d9 | 494 | |
b1d4d9e0 | 495 | if (!non_swap_entry(swpent)) |
c164e038 | 496 | mss->swap += PAGE_SIZE; |
b1d4d9e0 KK |
497 | else if (is_migration_entry(swpent)) |
498 | page = migration_entry_to_page(swpent); | |
499 | } | |
ae11c4d9 | 500 | |
ae11c4d9 DH |
501 | if (!page) |
502 | return; | |
c164e038 | 503 | smaps_account(mss, page, PAGE_SIZE, pte_young(*pte), pte_dirty(*pte)); |
ae11c4d9 DH |
504 | } |
505 | ||
c164e038 KS |
506 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
507 | static void smaps_pmd_entry(pmd_t *pmd, unsigned long addr, | |
508 | struct mm_walk *walk) | |
509 | { | |
510 | struct mem_size_stats *mss = walk->private; | |
14eb6fdd | 511 | struct vm_area_struct *vma = walk->vma; |
c164e038 KS |
512 | struct page *page; |
513 | ||
514 | /* FOLL_DUMP will return -EFAULT on huge zero page */ | |
515 | page = follow_trans_huge_pmd(vma, addr, pmd, FOLL_DUMP); | |
516 | if (IS_ERR_OR_NULL(page)) | |
517 | return; | |
518 | mss->anonymous_thp += HPAGE_PMD_SIZE; | |
519 | smaps_account(mss, page, HPAGE_PMD_SIZE, | |
520 | pmd_young(*pmd), pmd_dirty(*pmd)); | |
521 | } | |
522 | #else | |
523 | static void smaps_pmd_entry(pmd_t *pmd, unsigned long addr, | |
524 | struct mm_walk *walk) | |
525 | { | |
526 | } | |
527 | #endif | |
528 | ||
b3ae5acb | 529 | static int smaps_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end, |
2165009b | 530 | struct mm_walk *walk) |
e070ad49 | 531 | { |
14eb6fdd | 532 | struct vm_area_struct *vma = walk->vma; |
ae11c4d9 | 533 | pte_t *pte; |
705e87c0 | 534 | spinlock_t *ptl; |
e070ad49 | 535 | |
bf929152 | 536 | if (pmd_trans_huge_lock(pmd, vma, &ptl) == 1) { |
c164e038 | 537 | smaps_pmd_entry(pmd, addr, walk); |
bf929152 | 538 | spin_unlock(ptl); |
025c5b24 | 539 | return 0; |
22e057c5 | 540 | } |
1a5a9906 AA |
541 | |
542 | if (pmd_trans_unstable(pmd)) | |
543 | return 0; | |
22e057c5 DH |
544 | /* |
545 | * The mmap_sem held all the way back in m_start() is what | |
546 | * keeps khugepaged out of here and from collapsing things | |
547 | * in here. | |
548 | */ | |
705e87c0 | 549 | pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl); |
ae11c4d9 | 550 | for (; addr != end; pte++, addr += PAGE_SIZE) |
c164e038 | 551 | smaps_pte_entry(pte, addr, walk); |
705e87c0 HD |
552 | pte_unmap_unlock(pte - 1, ptl); |
553 | cond_resched(); | |
b3ae5acb | 554 | return 0; |
e070ad49 ML |
555 | } |
556 | ||
834f82e2 CG |
557 | static void show_smap_vma_flags(struct seq_file *m, struct vm_area_struct *vma) |
558 | { | |
559 | /* | |
560 | * Don't forget to update Documentation/ on changes. | |
561 | */ | |
562 | static const char mnemonics[BITS_PER_LONG][2] = { | |
563 | /* | |
564 | * In case if we meet a flag we don't know about. | |
565 | */ | |
566 | [0 ... (BITS_PER_LONG-1)] = "??", | |
567 | ||
568 | [ilog2(VM_READ)] = "rd", | |
569 | [ilog2(VM_WRITE)] = "wr", | |
570 | [ilog2(VM_EXEC)] = "ex", | |
571 | [ilog2(VM_SHARED)] = "sh", | |
572 | [ilog2(VM_MAYREAD)] = "mr", | |
573 | [ilog2(VM_MAYWRITE)] = "mw", | |
574 | [ilog2(VM_MAYEXEC)] = "me", | |
575 | [ilog2(VM_MAYSHARE)] = "ms", | |
576 | [ilog2(VM_GROWSDOWN)] = "gd", | |
577 | [ilog2(VM_PFNMAP)] = "pf", | |
578 | [ilog2(VM_DENYWRITE)] = "dw", | |
4aae7e43 QR |
579 | #ifdef CONFIG_X86_INTEL_MPX |
580 | [ilog2(VM_MPX)] = "mp", | |
581 | #endif | |
834f82e2 CG |
582 | [ilog2(VM_LOCKED)] = "lo", |
583 | [ilog2(VM_IO)] = "io", | |
584 | [ilog2(VM_SEQ_READ)] = "sr", | |
585 | [ilog2(VM_RAND_READ)] = "rr", | |
586 | [ilog2(VM_DONTCOPY)] = "dc", | |
587 | [ilog2(VM_DONTEXPAND)] = "de", | |
588 | [ilog2(VM_ACCOUNT)] = "ac", | |
589 | [ilog2(VM_NORESERVE)] = "nr", | |
590 | [ilog2(VM_HUGETLB)] = "ht", | |
834f82e2 CG |
591 | [ilog2(VM_ARCH_1)] = "ar", |
592 | [ilog2(VM_DONTDUMP)] = "dd", | |
ec8e41ae NH |
593 | #ifdef CONFIG_MEM_SOFT_DIRTY |
594 | [ilog2(VM_SOFTDIRTY)] = "sd", | |
595 | #endif | |
834f82e2 CG |
596 | [ilog2(VM_MIXEDMAP)] = "mm", |
597 | [ilog2(VM_HUGEPAGE)] = "hg", | |
598 | [ilog2(VM_NOHUGEPAGE)] = "nh", | |
599 | [ilog2(VM_MERGEABLE)] = "mg", | |
16ba6f81 AA |
600 | [ilog2(VM_UFFD_MISSING)]= "um", |
601 | [ilog2(VM_UFFD_WP)] = "uw", | |
834f82e2 CG |
602 | }; |
603 | size_t i; | |
604 | ||
605 | seq_puts(m, "VmFlags: "); | |
606 | for (i = 0; i < BITS_PER_LONG; i++) { | |
607 | if (vma->vm_flags & (1UL << i)) { | |
608 | seq_printf(m, "%c%c ", | |
609 | mnemonics[i][0], mnemonics[i][1]); | |
610 | } | |
611 | } | |
612 | seq_putc(m, '\n'); | |
613 | } | |
614 | ||
b7643757 | 615 | static int show_smap(struct seq_file *m, void *v, int is_pid) |
e070ad49 ML |
616 | { |
617 | struct vm_area_struct *vma = v; | |
e070ad49 | 618 | struct mem_size_stats mss; |
2165009b DH |
619 | struct mm_walk smaps_walk = { |
620 | .pmd_entry = smaps_pte_range, | |
621 | .mm = vma->vm_mm, | |
622 | .private = &mss, | |
623 | }; | |
e070ad49 ML |
624 | |
625 | memset(&mss, 0, sizeof mss); | |
d82ef020 | 626 | /* mmap_sem is held in m_start */ |
14eb6fdd | 627 | walk_page_vma(vma, &smaps_walk); |
4752c369 | 628 | |
b7643757 | 629 | show_map_vma(m, vma, is_pid); |
4752c369 MM |
630 | |
631 | seq_printf(m, | |
632 | "Size: %8lu kB\n" | |
633 | "Rss: %8lu kB\n" | |
634 | "Pss: %8lu kB\n" | |
635 | "Shared_Clean: %8lu kB\n" | |
636 | "Shared_Dirty: %8lu kB\n" | |
637 | "Private_Clean: %8lu kB\n" | |
638 | "Private_Dirty: %8lu kB\n" | |
214e471f | 639 | "Referenced: %8lu kB\n" |
b40d4f84 | 640 | "Anonymous: %8lu kB\n" |
4031a219 | 641 | "AnonHugePages: %8lu kB\n" |
08fba699 | 642 | "Swap: %8lu kB\n" |
3340289d | 643 | "KernelPageSize: %8lu kB\n" |
2d90508f NK |
644 | "MMUPageSize: %8lu kB\n" |
645 | "Locked: %8lu kB\n", | |
4752c369 MM |
646 | (vma->vm_end - vma->vm_start) >> 10, |
647 | mss.resident >> 10, | |
648 | (unsigned long)(mss.pss >> (10 + PSS_SHIFT)), | |
649 | mss.shared_clean >> 10, | |
650 | mss.shared_dirty >> 10, | |
651 | mss.private_clean >> 10, | |
652 | mss.private_dirty >> 10, | |
214e471f | 653 | mss.referenced >> 10, |
b40d4f84 | 654 | mss.anonymous >> 10, |
4031a219 | 655 | mss.anonymous_thp >> 10, |
08fba699 | 656 | mss.swap >> 10, |
3340289d | 657 | vma_kernel_pagesize(vma) >> 10, |
2d90508f NK |
658 | vma_mmu_pagesize(vma) >> 10, |
659 | (vma->vm_flags & VM_LOCKED) ? | |
660 | (unsigned long)(mss.pss >> (10 + PSS_SHIFT)) : 0); | |
4752c369 | 661 | |
834f82e2 | 662 | show_smap_vma_flags(m, vma); |
b8c20a9b | 663 | m_cache_vma(m, vma); |
7c88db0c | 664 | return 0; |
e070ad49 ML |
665 | } |
666 | ||
b7643757 SP |
667 | static int show_pid_smap(struct seq_file *m, void *v) |
668 | { | |
669 | return show_smap(m, v, 1); | |
670 | } | |
671 | ||
672 | static int show_tid_smap(struct seq_file *m, void *v) | |
673 | { | |
674 | return show_smap(m, v, 0); | |
675 | } | |
676 | ||
03a44825 | 677 | static const struct seq_operations proc_pid_smaps_op = { |
a6198797 MM |
678 | .start = m_start, |
679 | .next = m_next, | |
680 | .stop = m_stop, | |
b7643757 SP |
681 | .show = show_pid_smap |
682 | }; | |
683 | ||
684 | static const struct seq_operations proc_tid_smaps_op = { | |
685 | .start = m_start, | |
686 | .next = m_next, | |
687 | .stop = m_stop, | |
688 | .show = show_tid_smap | |
a6198797 MM |
689 | }; |
690 | ||
b7643757 | 691 | static int pid_smaps_open(struct inode *inode, struct file *file) |
a6198797 MM |
692 | { |
693 | return do_maps_open(inode, file, &proc_pid_smaps_op); | |
694 | } | |
695 | ||
b7643757 SP |
696 | static int tid_smaps_open(struct inode *inode, struct file *file) |
697 | { | |
698 | return do_maps_open(inode, file, &proc_tid_smaps_op); | |
699 | } | |
700 | ||
701 | const struct file_operations proc_pid_smaps_operations = { | |
702 | .open = pid_smaps_open, | |
703 | .read = seq_read, | |
704 | .llseek = seq_lseek, | |
29a40ace | 705 | .release = proc_map_release, |
b7643757 SP |
706 | }; |
707 | ||
708 | const struct file_operations proc_tid_smaps_operations = { | |
709 | .open = tid_smaps_open, | |
a6198797 MM |
710 | .read = seq_read, |
711 | .llseek = seq_lseek, | |
29a40ace | 712 | .release = proc_map_release, |
a6198797 MM |
713 | }; |
714 | ||
040fa020 PE |
715 | enum clear_refs_types { |
716 | CLEAR_REFS_ALL = 1, | |
717 | CLEAR_REFS_ANON, | |
718 | CLEAR_REFS_MAPPED, | |
0f8975ec | 719 | CLEAR_REFS_SOFT_DIRTY, |
695f0559 | 720 | CLEAR_REFS_MM_HIWATER_RSS, |
040fa020 PE |
721 | CLEAR_REFS_LAST, |
722 | }; | |
723 | ||
af9de7eb | 724 | struct clear_refs_private { |
0f8975ec | 725 | enum clear_refs_types type; |
af9de7eb PE |
726 | }; |
727 | ||
7d5b3bfa | 728 | #ifdef CONFIG_MEM_SOFT_DIRTY |
0f8975ec PE |
729 | static inline void clear_soft_dirty(struct vm_area_struct *vma, |
730 | unsigned long addr, pte_t *pte) | |
731 | { | |
0f8975ec PE |
732 | /* |
733 | * The soft-dirty tracker uses #PF-s to catch writes | |
734 | * to pages, so write-protect the pte as well. See the | |
735 | * Documentation/vm/soft-dirty.txt for full description | |
736 | * of how soft-dirty works. | |
737 | */ | |
738 | pte_t ptent = *pte; | |
179ef71c CG |
739 | |
740 | if (pte_present(ptent)) { | |
741 | ptent = pte_wrprotect(ptent); | |
742 | ptent = pte_clear_flags(ptent, _PAGE_SOFT_DIRTY); | |
743 | } else if (is_swap_pte(ptent)) { | |
744 | ptent = pte_swp_clear_soft_dirty(ptent); | |
745 | } | |
746 | ||
0f8975ec | 747 | set_pte_at(vma->vm_mm, addr, pte, ptent); |
0f8975ec PE |
748 | } |
749 | ||
7d5b3bfa KS |
750 | static inline void clear_soft_dirty_pmd(struct vm_area_struct *vma, |
751 | unsigned long addr, pmd_t *pmdp) | |
752 | { | |
753 | pmd_t pmd = *pmdp; | |
754 | ||
755 | pmd = pmd_wrprotect(pmd); | |
756 | pmd = pmd_clear_flags(pmd, _PAGE_SOFT_DIRTY); | |
757 | ||
758 | if (vma->vm_flags & VM_SOFTDIRTY) | |
759 | vma->vm_flags &= ~VM_SOFTDIRTY; | |
760 | ||
761 | set_pmd_at(vma->vm_mm, addr, pmdp, pmd); | |
762 | } | |
763 | ||
764 | #else | |
765 | ||
766 | static inline void clear_soft_dirty(struct vm_area_struct *vma, | |
767 | unsigned long addr, pte_t *pte) | |
768 | { | |
769 | } | |
770 | ||
771 | static inline void clear_soft_dirty_pmd(struct vm_area_struct *vma, | |
772 | unsigned long addr, pmd_t *pmdp) | |
773 | { | |
774 | } | |
775 | #endif | |
776 | ||
a6198797 | 777 | static int clear_refs_pte_range(pmd_t *pmd, unsigned long addr, |
2165009b | 778 | unsigned long end, struct mm_walk *walk) |
a6198797 | 779 | { |
af9de7eb | 780 | struct clear_refs_private *cp = walk->private; |
5c64f52a | 781 | struct vm_area_struct *vma = walk->vma; |
a6198797 MM |
782 | pte_t *pte, ptent; |
783 | spinlock_t *ptl; | |
784 | struct page *page; | |
785 | ||
7d5b3bfa KS |
786 | if (pmd_trans_huge_lock(pmd, vma, &ptl) == 1) { |
787 | if (cp->type == CLEAR_REFS_SOFT_DIRTY) { | |
788 | clear_soft_dirty_pmd(vma, addr, pmd); | |
789 | goto out; | |
790 | } | |
791 | ||
792 | page = pmd_page(*pmd); | |
793 | ||
794 | /* Clear accessed and referenced bits. */ | |
795 | pmdp_test_and_clear_young(vma, addr, pmd); | |
796 | ClearPageReferenced(page); | |
797 | out: | |
798 | spin_unlock(ptl); | |
799 | return 0; | |
800 | } | |
801 | ||
1a5a9906 AA |
802 | if (pmd_trans_unstable(pmd)) |
803 | return 0; | |
03319327 | 804 | |
a6198797 MM |
805 | pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl); |
806 | for (; addr != end; pte++, addr += PAGE_SIZE) { | |
807 | ptent = *pte; | |
a6198797 | 808 | |
0f8975ec PE |
809 | if (cp->type == CLEAR_REFS_SOFT_DIRTY) { |
810 | clear_soft_dirty(vma, addr, pte); | |
811 | continue; | |
812 | } | |
813 | ||
179ef71c CG |
814 | if (!pte_present(ptent)) |
815 | continue; | |
816 | ||
a6198797 MM |
817 | page = vm_normal_page(vma, addr, ptent); |
818 | if (!page) | |
819 | continue; | |
820 | ||
821 | /* Clear accessed and referenced bits. */ | |
822 | ptep_test_and_clear_young(vma, addr, pte); | |
823 | ClearPageReferenced(page); | |
824 | } | |
825 | pte_unmap_unlock(pte - 1, ptl); | |
826 | cond_resched(); | |
827 | return 0; | |
828 | } | |
829 | ||
5c64f52a NH |
830 | static int clear_refs_test_walk(unsigned long start, unsigned long end, |
831 | struct mm_walk *walk) | |
832 | { | |
833 | struct clear_refs_private *cp = walk->private; | |
834 | struct vm_area_struct *vma = walk->vma; | |
835 | ||
48684a65 NH |
836 | if (vma->vm_flags & VM_PFNMAP) |
837 | return 1; | |
838 | ||
5c64f52a NH |
839 | /* |
840 | * Writing 1 to /proc/pid/clear_refs affects all pages. | |
841 | * Writing 2 to /proc/pid/clear_refs only affects anonymous pages. | |
842 | * Writing 3 to /proc/pid/clear_refs only affects file mapped pages. | |
843 | * Writing 4 to /proc/pid/clear_refs affects all pages. | |
844 | */ | |
845 | if (cp->type == CLEAR_REFS_ANON && vma->vm_file) | |
846 | return 1; | |
847 | if (cp->type == CLEAR_REFS_MAPPED && !vma->vm_file) | |
848 | return 1; | |
849 | return 0; | |
850 | } | |
851 | ||
f248dcb3 MM |
852 | static ssize_t clear_refs_write(struct file *file, const char __user *buf, |
853 | size_t count, loff_t *ppos) | |
b813e931 | 854 | { |
f248dcb3 | 855 | struct task_struct *task; |
fb92a4b0 | 856 | char buffer[PROC_NUMBUF]; |
f248dcb3 | 857 | struct mm_struct *mm; |
b813e931 | 858 | struct vm_area_struct *vma; |
040fa020 PE |
859 | enum clear_refs_types type; |
860 | int itype; | |
0a8cb8e3 | 861 | int rv; |
b813e931 | 862 | |
f248dcb3 MM |
863 | memset(buffer, 0, sizeof(buffer)); |
864 | if (count > sizeof(buffer) - 1) | |
865 | count = sizeof(buffer) - 1; | |
866 | if (copy_from_user(buffer, buf, count)) | |
867 | return -EFAULT; | |
040fa020 | 868 | rv = kstrtoint(strstrip(buffer), 10, &itype); |
0a8cb8e3 AD |
869 | if (rv < 0) |
870 | return rv; | |
040fa020 PE |
871 | type = (enum clear_refs_types)itype; |
872 | if (type < CLEAR_REFS_ALL || type >= CLEAR_REFS_LAST) | |
f248dcb3 | 873 | return -EINVAL; |
541c237c | 874 | |
496ad9aa | 875 | task = get_proc_task(file_inode(file)); |
f248dcb3 MM |
876 | if (!task) |
877 | return -ESRCH; | |
878 | mm = get_task_mm(task); | |
879 | if (mm) { | |
af9de7eb | 880 | struct clear_refs_private cp = { |
0f8975ec | 881 | .type = type, |
af9de7eb | 882 | }; |
20cbc972 AM |
883 | struct mm_walk clear_refs_walk = { |
884 | .pmd_entry = clear_refs_pte_range, | |
5c64f52a | 885 | .test_walk = clear_refs_test_walk, |
20cbc972 | 886 | .mm = mm, |
af9de7eb | 887 | .private = &cp, |
20cbc972 | 888 | }; |
695f0559 PC |
889 | |
890 | if (type == CLEAR_REFS_MM_HIWATER_RSS) { | |
891 | /* | |
892 | * Writing 5 to /proc/pid/clear_refs resets the peak | |
893 | * resident set size to this mm's current rss value. | |
894 | */ | |
895 | down_write(&mm->mmap_sem); | |
896 | reset_mm_hiwater_rss(mm); | |
897 | up_write(&mm->mmap_sem); | |
898 | goto out_mm; | |
899 | } | |
900 | ||
f248dcb3 | 901 | down_read(&mm->mmap_sem); |
64e45507 PF |
902 | if (type == CLEAR_REFS_SOFT_DIRTY) { |
903 | for (vma = mm->mmap; vma; vma = vma->vm_next) { | |
904 | if (!(vma->vm_flags & VM_SOFTDIRTY)) | |
905 | continue; | |
906 | up_read(&mm->mmap_sem); | |
907 | down_write(&mm->mmap_sem); | |
908 | for (vma = mm->mmap; vma; vma = vma->vm_next) { | |
909 | vma->vm_flags &= ~VM_SOFTDIRTY; | |
910 | vma_set_page_prot(vma); | |
911 | } | |
912 | downgrade_write(&mm->mmap_sem); | |
913 | break; | |
914 | } | |
0f8975ec | 915 | mmu_notifier_invalidate_range_start(mm, 0, -1); |
64e45507 | 916 | } |
5c64f52a | 917 | walk_page_range(0, ~0UL, &clear_refs_walk); |
0f8975ec PE |
918 | if (type == CLEAR_REFS_SOFT_DIRTY) |
919 | mmu_notifier_invalidate_range_end(mm, 0, -1); | |
f248dcb3 MM |
920 | flush_tlb_mm(mm); |
921 | up_read(&mm->mmap_sem); | |
695f0559 | 922 | out_mm: |
f248dcb3 MM |
923 | mmput(mm); |
924 | } | |
925 | put_task_struct(task); | |
fb92a4b0 VL |
926 | |
927 | return count; | |
b813e931 DR |
928 | } |
929 | ||
f248dcb3 MM |
930 | const struct file_operations proc_clear_refs_operations = { |
931 | .write = clear_refs_write, | |
6038f373 | 932 | .llseek = noop_llseek, |
f248dcb3 MM |
933 | }; |
934 | ||
092b50ba NH |
935 | typedef struct { |
936 | u64 pme; | |
937 | } pagemap_entry_t; | |
938 | ||
85863e47 | 939 | struct pagemapread { |
8c829622 | 940 | int pos, len; /* units: PM_ENTRY_BYTES, not bytes */ |
092b50ba | 941 | pagemap_entry_t *buffer; |
1c90308e | 942 | bool show_pfn; |
85863e47 MM |
943 | }; |
944 | ||
5aaabe83 NH |
945 | #define PAGEMAP_WALK_SIZE (PMD_SIZE) |
946 | #define PAGEMAP_WALK_MASK (PMD_MASK) | |
947 | ||
deb94544 KK |
948 | #define PM_ENTRY_BYTES sizeof(pagemap_entry_t) |
949 | #define PM_PFRAME_BITS 55 | |
950 | #define PM_PFRAME_MASK GENMASK_ULL(PM_PFRAME_BITS - 1, 0) | |
951 | #define PM_SOFT_DIRTY BIT_ULL(55) | |
952 | #define PM_FILE BIT_ULL(61) | |
953 | #define PM_SWAP BIT_ULL(62) | |
954 | #define PM_PRESENT BIT_ULL(63) | |
955 | ||
85863e47 MM |
956 | #define PM_END_OF_BUFFER 1 |
957 | ||
deb94544 | 958 | static inline pagemap_entry_t make_pme(u64 frame, u64 flags) |
092b50ba | 959 | { |
deb94544 | 960 | return (pagemap_entry_t) { .pme = (frame & PM_PFRAME_MASK) | flags }; |
092b50ba NH |
961 | } |
962 | ||
963 | static int add_to_pagemap(unsigned long addr, pagemap_entry_t *pme, | |
85863e47 MM |
964 | struct pagemapread *pm) |
965 | { | |
092b50ba | 966 | pm->buffer[pm->pos++] = *pme; |
d82ef020 | 967 | if (pm->pos >= pm->len) |
aae8679b | 968 | return PM_END_OF_BUFFER; |
85863e47 MM |
969 | return 0; |
970 | } | |
971 | ||
972 | static int pagemap_pte_hole(unsigned long start, unsigned long end, | |
2165009b | 973 | struct mm_walk *walk) |
85863e47 | 974 | { |
2165009b | 975 | struct pagemapread *pm = walk->private; |
68b5a652 | 976 | unsigned long addr = start; |
85863e47 | 977 | int err = 0; |
092b50ba | 978 | |
68b5a652 PF |
979 | while (addr < end) { |
980 | struct vm_area_struct *vma = find_vma(walk->mm, addr); | |
deb94544 | 981 | pagemap_entry_t pme = make_pme(0, 0); |
87e6d49a PF |
982 | /* End of address space hole, which we mark as non-present. */ |
983 | unsigned long hole_end; | |
68b5a652 | 984 | |
87e6d49a PF |
985 | if (vma) |
986 | hole_end = min(end, vma->vm_start); | |
987 | else | |
988 | hole_end = end; | |
989 | ||
990 | for (; addr < hole_end; addr += PAGE_SIZE) { | |
991 | err = add_to_pagemap(addr, &pme, pm); | |
992 | if (err) | |
993 | goto out; | |
68b5a652 PF |
994 | } |
995 | ||
87e6d49a PF |
996 | if (!vma) |
997 | break; | |
998 | ||
999 | /* Addresses in the VMA. */ | |
1000 | if (vma->vm_flags & VM_SOFTDIRTY) | |
deb94544 | 1001 | pme = make_pme(0, PM_SOFT_DIRTY); |
87e6d49a | 1002 | for (; addr < min(end, vma->vm_end); addr += PAGE_SIZE) { |
68b5a652 PF |
1003 | err = add_to_pagemap(addr, &pme, pm); |
1004 | if (err) | |
1005 | goto out; | |
1006 | } | |
85863e47 | 1007 | } |
68b5a652 | 1008 | out: |
85863e47 MM |
1009 | return err; |
1010 | } | |
1011 | ||
deb94544 | 1012 | static pagemap_entry_t pte_to_pagemap_entry(struct pagemapread *pm, |
052fb0d6 | 1013 | struct vm_area_struct *vma, unsigned long addr, pte_t pte) |
85863e47 | 1014 | { |
deb94544 | 1015 | u64 frame = 0, flags = 0; |
052fb0d6 | 1016 | struct page *page = NULL; |
85863e47 | 1017 | |
052fb0d6 | 1018 | if (pte_present(pte)) { |
1c90308e KK |
1019 | if (pm->show_pfn) |
1020 | frame = pte_pfn(pte); | |
deb94544 | 1021 | flags |= PM_PRESENT; |
052fb0d6 | 1022 | page = vm_normal_page(vma, addr, pte); |
e9cdd6e7 | 1023 | if (pte_soft_dirty(pte)) |
deb94544 | 1024 | flags |= PM_SOFT_DIRTY; |
052fb0d6 | 1025 | } else if (is_swap_pte(pte)) { |
179ef71c CG |
1026 | swp_entry_t entry; |
1027 | if (pte_swp_soft_dirty(pte)) | |
deb94544 | 1028 | flags |= PM_SOFT_DIRTY; |
179ef71c | 1029 | entry = pte_to_swp_entry(pte); |
052fb0d6 KK |
1030 | frame = swp_type(entry) | |
1031 | (swp_offset(entry) << MAX_SWAPFILES_SHIFT); | |
deb94544 | 1032 | flags |= PM_SWAP; |
052fb0d6 KK |
1033 | if (is_migration_entry(entry)) |
1034 | page = migration_entry_to_page(entry); | |
052fb0d6 KK |
1035 | } |
1036 | ||
1037 | if (page && !PageAnon(page)) | |
1038 | flags |= PM_FILE; | |
deb94544 KK |
1039 | if (vma->vm_flags & VM_SOFTDIRTY) |
1040 | flags |= PM_SOFT_DIRTY; | |
052fb0d6 | 1041 | |
deb94544 | 1042 | return make_pme(frame, flags); |
bcf8039e DH |
1043 | } |
1044 | ||
356515e7 | 1045 | static int pagemap_pmd_range(pmd_t *pmdp, unsigned long addr, unsigned long end, |
2165009b | 1046 | struct mm_walk *walk) |
85863e47 | 1047 | { |
f995ece2 | 1048 | struct vm_area_struct *vma = walk->vma; |
2165009b | 1049 | struct pagemapread *pm = walk->private; |
bf929152 | 1050 | spinlock_t *ptl; |
05fbf357 | 1051 | pte_t *pte, *orig_pte; |
85863e47 MM |
1052 | int err = 0; |
1053 | ||
356515e7 KK |
1054 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
1055 | if (pmd_trans_huge_lock(pmdp, vma, &ptl) == 1) { | |
1056 | u64 flags = 0, frame = 0; | |
1057 | pmd_t pmd = *pmdp; | |
0f8975ec | 1058 | |
356515e7 | 1059 | if ((vma->vm_flags & VM_SOFTDIRTY) || pmd_soft_dirty(pmd)) |
deb94544 | 1060 | flags |= PM_SOFT_DIRTY; |
d9104d1c | 1061 | |
356515e7 KK |
1062 | /* |
1063 | * Currently pmd for thp is always present because thp | |
1064 | * can not be swapped-out, migrated, or HWPOISONed | |
1065 | * (split in such cases instead.) | |
1066 | * This if-check is just to prepare for future implementation. | |
1067 | */ | |
1068 | if (pmd_present(pmd)) { | |
1069 | flags |= PM_PRESENT; | |
1c90308e KK |
1070 | if (pm->show_pfn) |
1071 | frame = pmd_pfn(pmd) + | |
1072 | ((addr & ~PMD_MASK) >> PAGE_SHIFT); | |
356515e7 KK |
1073 | } |
1074 | ||
025c5b24 | 1075 | for (; addr != end; addr += PAGE_SIZE) { |
356515e7 | 1076 | pagemap_entry_t pme = make_pme(frame, flags); |
025c5b24 | 1077 | |
092b50ba | 1078 | err = add_to_pagemap(addr, &pme, pm); |
025c5b24 NH |
1079 | if (err) |
1080 | break; | |
1c90308e | 1081 | if (pm->show_pfn && (flags & PM_PRESENT)) |
356515e7 | 1082 | frame++; |
5aaabe83 | 1083 | } |
bf929152 | 1084 | spin_unlock(ptl); |
025c5b24 | 1085 | return err; |
5aaabe83 NH |
1086 | } |
1087 | ||
356515e7 | 1088 | if (pmd_trans_unstable(pmdp)) |
45f83cef | 1089 | return 0; |
356515e7 | 1090 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE */ |
81d0fa62 | 1091 | |
f995ece2 NH |
1092 | /* |
1093 | * We can assume that @vma always points to a valid one and @end never | |
1094 | * goes beyond vma->vm_end. | |
1095 | */ | |
356515e7 | 1096 | orig_pte = pte = pte_offset_map_lock(walk->mm, pmdp, addr, &ptl); |
f995ece2 NH |
1097 | for (; addr < end; pte++, addr += PAGE_SIZE) { |
1098 | pagemap_entry_t pme; | |
05fbf357 | 1099 | |
deb94544 | 1100 | pme = pte_to_pagemap_entry(pm, vma, addr, *pte); |
f995ece2 | 1101 | err = add_to_pagemap(addr, &pme, pm); |
05fbf357 | 1102 | if (err) |
81d0fa62 | 1103 | break; |
85863e47 | 1104 | } |
f995ece2 | 1105 | pte_unmap_unlock(orig_pte, ptl); |
85863e47 MM |
1106 | |
1107 | cond_resched(); | |
1108 | ||
1109 | return err; | |
1110 | } | |
1111 | ||
1a5cb814 | 1112 | #ifdef CONFIG_HUGETLB_PAGE |
116354d1 | 1113 | /* This function walks within one hugetlb entry in the single call */ |
356515e7 | 1114 | static int pagemap_hugetlb_range(pte_t *ptep, unsigned long hmask, |
116354d1 NH |
1115 | unsigned long addr, unsigned long end, |
1116 | struct mm_walk *walk) | |
5dc37642 | 1117 | { |
5dc37642 | 1118 | struct pagemapread *pm = walk->private; |
f995ece2 | 1119 | struct vm_area_struct *vma = walk->vma; |
356515e7 | 1120 | u64 flags = 0, frame = 0; |
5dc37642 | 1121 | int err = 0; |
356515e7 | 1122 | pte_t pte; |
5dc37642 | 1123 | |
f995ece2 | 1124 | if (vma->vm_flags & VM_SOFTDIRTY) |
deb94544 | 1125 | flags |= PM_SOFT_DIRTY; |
d9104d1c | 1126 | |
356515e7 KK |
1127 | pte = huge_ptep_get(ptep); |
1128 | if (pte_present(pte)) { | |
1129 | struct page *page = pte_page(pte); | |
1130 | ||
1131 | if (!PageAnon(page)) | |
1132 | flags |= PM_FILE; | |
1133 | ||
1134 | flags |= PM_PRESENT; | |
1c90308e KK |
1135 | if (pm->show_pfn) |
1136 | frame = pte_pfn(pte) + | |
1137 | ((addr & ~hmask) >> PAGE_SHIFT); | |
356515e7 KK |
1138 | } |
1139 | ||
5dc37642 | 1140 | for (; addr != end; addr += PAGE_SIZE) { |
356515e7 KK |
1141 | pagemap_entry_t pme = make_pme(frame, flags); |
1142 | ||
092b50ba | 1143 | err = add_to_pagemap(addr, &pme, pm); |
5dc37642 NH |
1144 | if (err) |
1145 | return err; | |
1c90308e | 1146 | if (pm->show_pfn && (flags & PM_PRESENT)) |
356515e7 | 1147 | frame++; |
5dc37642 NH |
1148 | } |
1149 | ||
1150 | cond_resched(); | |
1151 | ||
1152 | return err; | |
1153 | } | |
1a5cb814 | 1154 | #endif /* HUGETLB_PAGE */ |
5dc37642 | 1155 | |
85863e47 MM |
1156 | /* |
1157 | * /proc/pid/pagemap - an array mapping virtual pages to pfns | |
1158 | * | |
f16278c6 HR |
1159 | * For each page in the address space, this file contains one 64-bit entry |
1160 | * consisting of the following: | |
1161 | * | |
052fb0d6 | 1162 | * Bits 0-54 page frame number (PFN) if present |
f16278c6 | 1163 | * Bits 0-4 swap type if swapped |
052fb0d6 | 1164 | * Bits 5-54 swap offset if swapped |
deb94544 KK |
1165 | * Bit 55 pte is soft-dirty (see Documentation/vm/soft-dirty.txt) |
1166 | * Bits 56-60 zero | |
052fb0d6 | 1167 | * Bit 61 page is file-page or shared-anon |
f16278c6 HR |
1168 | * Bit 62 page swapped |
1169 | * Bit 63 page present | |
1170 | * | |
1171 | * If the page is not present but in swap, then the PFN contains an | |
1172 | * encoding of the swap file number and the page's offset into the | |
1173 | * swap. Unmapped pages return a null PFN. This allows determining | |
85863e47 MM |
1174 | * precisely which pages are mapped (or in swap) and comparing mapped |
1175 | * pages between processes. | |
1176 | * | |
1177 | * Efficient users of this interface will use /proc/pid/maps to | |
1178 | * determine which areas of memory are actually mapped and llseek to | |
1179 | * skip over unmapped regions. | |
1180 | */ | |
1181 | static ssize_t pagemap_read(struct file *file, char __user *buf, | |
1182 | size_t count, loff_t *ppos) | |
1183 | { | |
a06db751 | 1184 | struct mm_struct *mm = file->private_data; |
85863e47 | 1185 | struct pagemapread pm; |
ee1e6ab6 | 1186 | struct mm_walk pagemap_walk = {}; |
5d7e0d2b AM |
1187 | unsigned long src; |
1188 | unsigned long svpfn; | |
1189 | unsigned long start_vaddr; | |
1190 | unsigned long end_vaddr; | |
a06db751 | 1191 | int ret = 0, copied = 0; |
85863e47 | 1192 | |
a06db751 | 1193 | if (!mm || !atomic_inc_not_zero(&mm->mm_users)) |
85863e47 MM |
1194 | goto out; |
1195 | ||
85863e47 MM |
1196 | ret = -EINVAL; |
1197 | /* file position must be aligned */ | |
aae8679b | 1198 | if ((*ppos % PM_ENTRY_BYTES) || (count % PM_ENTRY_BYTES)) |
a06db751 | 1199 | goto out_mm; |
85863e47 MM |
1200 | |
1201 | ret = 0; | |
08161786 | 1202 | if (!count) |
a06db751 | 1203 | goto out_mm; |
08161786 | 1204 | |
1c90308e KK |
1205 | /* do not disclose physical addresses: attack vector */ |
1206 | pm.show_pfn = file_ns_capable(file, &init_user_ns, CAP_SYS_ADMIN); | |
1207 | ||
8c829622 | 1208 | pm.len = (PAGEMAP_WALK_SIZE >> PAGE_SHIFT); |
1209 | pm.buffer = kmalloc(pm.len * PM_ENTRY_BYTES, GFP_TEMPORARY); | |
5d7e0d2b | 1210 | ret = -ENOMEM; |
d82ef020 | 1211 | if (!pm.buffer) |
a06db751 | 1212 | goto out_mm; |
85863e47 | 1213 | |
356515e7 | 1214 | pagemap_walk.pmd_entry = pagemap_pmd_range; |
5d7e0d2b | 1215 | pagemap_walk.pte_hole = pagemap_pte_hole; |
1a5cb814 | 1216 | #ifdef CONFIG_HUGETLB_PAGE |
5dc37642 | 1217 | pagemap_walk.hugetlb_entry = pagemap_hugetlb_range; |
1a5cb814 | 1218 | #endif |
5d7e0d2b AM |
1219 | pagemap_walk.mm = mm; |
1220 | pagemap_walk.private = ± | |
1221 | ||
1222 | src = *ppos; | |
1223 | svpfn = src / PM_ENTRY_BYTES; | |
1224 | start_vaddr = svpfn << PAGE_SHIFT; | |
a06db751 | 1225 | end_vaddr = mm->task_size; |
5d7e0d2b AM |
1226 | |
1227 | /* watch out for wraparound */ | |
a06db751 | 1228 | if (svpfn > mm->task_size >> PAGE_SHIFT) |
5d7e0d2b AM |
1229 | start_vaddr = end_vaddr; |
1230 | ||
1231 | /* | |
1232 | * The odds are that this will stop walking way | |
1233 | * before end_vaddr, because the length of the | |
1234 | * user buffer is tracked in "pm", and the walk | |
1235 | * will stop when we hit the end of the buffer. | |
1236 | */ | |
d82ef020 KH |
1237 | ret = 0; |
1238 | while (count && (start_vaddr < end_vaddr)) { | |
1239 | int len; | |
1240 | unsigned long end; | |
1241 | ||
1242 | pm.pos = 0; | |
ea251c1d | 1243 | end = (start_vaddr + PAGEMAP_WALK_SIZE) & PAGEMAP_WALK_MASK; |
d82ef020 KH |
1244 | /* overflow ? */ |
1245 | if (end < start_vaddr || end > end_vaddr) | |
1246 | end = end_vaddr; | |
1247 | down_read(&mm->mmap_sem); | |
1248 | ret = walk_page_range(start_vaddr, end, &pagemap_walk); | |
1249 | up_read(&mm->mmap_sem); | |
1250 | start_vaddr = end; | |
1251 | ||
1252 | len = min(count, PM_ENTRY_BYTES * pm.pos); | |
309361e0 | 1253 | if (copy_to_user(buf, pm.buffer, len)) { |
d82ef020 | 1254 | ret = -EFAULT; |
a06db751 | 1255 | goto out_free; |
d82ef020 KH |
1256 | } |
1257 | copied += len; | |
1258 | buf += len; | |
1259 | count -= len; | |
85863e47 | 1260 | } |
d82ef020 KH |
1261 | *ppos += copied; |
1262 | if (!ret || ret == PM_END_OF_BUFFER) | |
1263 | ret = copied; | |
1264 | ||
98bc93e5 KM |
1265 | out_free: |
1266 | kfree(pm.buffer); | |
a06db751 KK |
1267 | out_mm: |
1268 | mmput(mm); | |
85863e47 MM |
1269 | out: |
1270 | return ret; | |
1271 | } | |
1272 | ||
541c237c PE |
1273 | static int pagemap_open(struct inode *inode, struct file *file) |
1274 | { | |
a06db751 KK |
1275 | struct mm_struct *mm; |
1276 | ||
a06db751 KK |
1277 | mm = proc_mem_open(inode, PTRACE_MODE_READ); |
1278 | if (IS_ERR(mm)) | |
1279 | return PTR_ERR(mm); | |
1280 | file->private_data = mm; | |
1281 | return 0; | |
1282 | } | |
1283 | ||
1284 | static int pagemap_release(struct inode *inode, struct file *file) | |
1285 | { | |
1286 | struct mm_struct *mm = file->private_data; | |
1287 | ||
1288 | if (mm) | |
1289 | mmdrop(mm); | |
541c237c PE |
1290 | return 0; |
1291 | } | |
1292 | ||
85863e47 MM |
1293 | const struct file_operations proc_pagemap_operations = { |
1294 | .llseek = mem_lseek, /* borrow this */ | |
1295 | .read = pagemap_read, | |
541c237c | 1296 | .open = pagemap_open, |
a06db751 | 1297 | .release = pagemap_release, |
85863e47 | 1298 | }; |
1e883281 | 1299 | #endif /* CONFIG_PROC_PAGE_MONITOR */ |
85863e47 | 1300 | |
6e21c8f1 | 1301 | #ifdef CONFIG_NUMA |
6e21c8f1 | 1302 | |
f69ff943 | 1303 | struct numa_maps { |
f69ff943 SW |
1304 | unsigned long pages; |
1305 | unsigned long anon; | |
1306 | unsigned long active; | |
1307 | unsigned long writeback; | |
1308 | unsigned long mapcount_max; | |
1309 | unsigned long dirty; | |
1310 | unsigned long swapcache; | |
1311 | unsigned long node[MAX_NUMNODES]; | |
1312 | }; | |
1313 | ||
5b52fc89 SW |
1314 | struct numa_maps_private { |
1315 | struct proc_maps_private proc_maps; | |
1316 | struct numa_maps md; | |
1317 | }; | |
1318 | ||
eb4866d0 DH |
1319 | static void gather_stats(struct page *page, struct numa_maps *md, int pte_dirty, |
1320 | unsigned long nr_pages) | |
f69ff943 SW |
1321 | { |
1322 | int count = page_mapcount(page); | |
1323 | ||
eb4866d0 | 1324 | md->pages += nr_pages; |
f69ff943 | 1325 | if (pte_dirty || PageDirty(page)) |
eb4866d0 | 1326 | md->dirty += nr_pages; |
f69ff943 SW |
1327 | |
1328 | if (PageSwapCache(page)) | |
eb4866d0 | 1329 | md->swapcache += nr_pages; |
f69ff943 SW |
1330 | |
1331 | if (PageActive(page) || PageUnevictable(page)) | |
eb4866d0 | 1332 | md->active += nr_pages; |
f69ff943 SW |
1333 | |
1334 | if (PageWriteback(page)) | |
eb4866d0 | 1335 | md->writeback += nr_pages; |
f69ff943 SW |
1336 | |
1337 | if (PageAnon(page)) | |
eb4866d0 | 1338 | md->anon += nr_pages; |
f69ff943 SW |
1339 | |
1340 | if (count > md->mapcount_max) | |
1341 | md->mapcount_max = count; | |
1342 | ||
eb4866d0 | 1343 | md->node[page_to_nid(page)] += nr_pages; |
f69ff943 SW |
1344 | } |
1345 | ||
3200a8aa DH |
1346 | static struct page *can_gather_numa_stats(pte_t pte, struct vm_area_struct *vma, |
1347 | unsigned long addr) | |
1348 | { | |
1349 | struct page *page; | |
1350 | int nid; | |
1351 | ||
1352 | if (!pte_present(pte)) | |
1353 | return NULL; | |
1354 | ||
1355 | page = vm_normal_page(vma, addr, pte); | |
1356 | if (!page) | |
1357 | return NULL; | |
1358 | ||
1359 | if (PageReserved(page)) | |
1360 | return NULL; | |
1361 | ||
1362 | nid = page_to_nid(page); | |
4ff1b2c2 | 1363 | if (!node_isset(nid, node_states[N_MEMORY])) |
3200a8aa DH |
1364 | return NULL; |
1365 | ||
1366 | return page; | |
1367 | } | |
1368 | ||
f69ff943 SW |
1369 | static int gather_pte_stats(pmd_t *pmd, unsigned long addr, |
1370 | unsigned long end, struct mm_walk *walk) | |
1371 | { | |
d85f4d6d NH |
1372 | struct numa_maps *md = walk->private; |
1373 | struct vm_area_struct *vma = walk->vma; | |
f69ff943 SW |
1374 | spinlock_t *ptl; |
1375 | pte_t *orig_pte; | |
1376 | pte_t *pte; | |
1377 | ||
d85f4d6d | 1378 | if (pmd_trans_huge_lock(pmd, vma, &ptl) == 1) { |
025c5b24 NH |
1379 | pte_t huge_pte = *(pte_t *)pmd; |
1380 | struct page *page; | |
1381 | ||
d85f4d6d | 1382 | page = can_gather_numa_stats(huge_pte, vma, addr); |
025c5b24 NH |
1383 | if (page) |
1384 | gather_stats(page, md, pte_dirty(huge_pte), | |
1385 | HPAGE_PMD_SIZE/PAGE_SIZE); | |
bf929152 | 1386 | spin_unlock(ptl); |
025c5b24 | 1387 | return 0; |
32ef4384 DH |
1388 | } |
1389 | ||
1a5a9906 AA |
1390 | if (pmd_trans_unstable(pmd)) |
1391 | return 0; | |
f69ff943 SW |
1392 | orig_pte = pte = pte_offset_map_lock(walk->mm, pmd, addr, &ptl); |
1393 | do { | |
d85f4d6d | 1394 | struct page *page = can_gather_numa_stats(*pte, vma, addr); |
f69ff943 SW |
1395 | if (!page) |
1396 | continue; | |
eb4866d0 | 1397 | gather_stats(page, md, pte_dirty(*pte), 1); |
f69ff943 SW |
1398 | |
1399 | } while (pte++, addr += PAGE_SIZE, addr != end); | |
1400 | pte_unmap_unlock(orig_pte, ptl); | |
1401 | return 0; | |
1402 | } | |
1403 | #ifdef CONFIG_HUGETLB_PAGE | |
632fd60f | 1404 | static int gather_hugetlb_stats(pte_t *pte, unsigned long hmask, |
f69ff943 SW |
1405 | unsigned long addr, unsigned long end, struct mm_walk *walk) |
1406 | { | |
1407 | struct numa_maps *md; | |
1408 | struct page *page; | |
1409 | ||
d4c54919 | 1410 | if (!pte_present(*pte)) |
f69ff943 SW |
1411 | return 0; |
1412 | ||
1413 | page = pte_page(*pte); | |
1414 | if (!page) | |
1415 | return 0; | |
1416 | ||
1417 | md = walk->private; | |
eb4866d0 | 1418 | gather_stats(page, md, pte_dirty(*pte), 1); |
f69ff943 SW |
1419 | return 0; |
1420 | } | |
1421 | ||
1422 | #else | |
632fd60f | 1423 | static int gather_hugetlb_stats(pte_t *pte, unsigned long hmask, |
f69ff943 SW |
1424 | unsigned long addr, unsigned long end, struct mm_walk *walk) |
1425 | { | |
1426 | return 0; | |
1427 | } | |
1428 | #endif | |
1429 | ||
1430 | /* | |
1431 | * Display pages allocated per node and memory policy via /proc. | |
1432 | */ | |
b7643757 | 1433 | static int show_numa_map(struct seq_file *m, void *v, int is_pid) |
f69ff943 | 1434 | { |
5b52fc89 SW |
1435 | struct numa_maps_private *numa_priv = m->private; |
1436 | struct proc_maps_private *proc_priv = &numa_priv->proc_maps; | |
f69ff943 | 1437 | struct vm_area_struct *vma = v; |
5b52fc89 | 1438 | struct numa_maps *md = &numa_priv->md; |
f69ff943 SW |
1439 | struct file *file = vma->vm_file; |
1440 | struct mm_struct *mm = vma->vm_mm; | |
d85f4d6d NH |
1441 | struct mm_walk walk = { |
1442 | .hugetlb_entry = gather_hugetlb_stats, | |
1443 | .pmd_entry = gather_pte_stats, | |
1444 | .private = md, | |
1445 | .mm = mm, | |
1446 | }; | |
f69ff943 | 1447 | struct mempolicy *pol; |
948927ee DR |
1448 | char buffer[64]; |
1449 | int nid; | |
f69ff943 SW |
1450 | |
1451 | if (!mm) | |
1452 | return 0; | |
1453 | ||
5b52fc89 SW |
1454 | /* Ensure we start with an empty set of numa_maps statistics. */ |
1455 | memset(md, 0, sizeof(*md)); | |
f69ff943 | 1456 | |
498f2371 ON |
1457 | pol = __get_vma_policy(vma, vma->vm_start); |
1458 | if (pol) { | |
1459 | mpol_to_str(buffer, sizeof(buffer), pol); | |
1460 | mpol_cond_put(pol); | |
1461 | } else { | |
1462 | mpol_to_str(buffer, sizeof(buffer), proc_priv->task_mempolicy); | |
1463 | } | |
f69ff943 SW |
1464 | |
1465 | seq_printf(m, "%08lx %s", vma->vm_start, buffer); | |
1466 | ||
1467 | if (file) { | |
17c2b4ee | 1468 | seq_puts(m, " file="); |
2726d566 | 1469 | seq_file_path(m, file, "\n\t= "); |
f69ff943 | 1470 | } else if (vma->vm_start <= mm->brk && vma->vm_end >= mm->start_brk) { |
17c2b4ee | 1471 | seq_puts(m, " heap"); |
b7643757 | 1472 | } else { |
58cb6548 | 1473 | pid_t tid = pid_of_stack(proc_priv, vma, is_pid); |
b7643757 SP |
1474 | if (tid != 0) { |
1475 | /* | |
1476 | * Thread stack in /proc/PID/task/TID/maps or | |
1477 | * the main process stack. | |
1478 | */ | |
1479 | if (!is_pid || (vma->vm_start <= mm->start_stack && | |
1480 | vma->vm_end >= mm->start_stack)) | |
17c2b4ee | 1481 | seq_puts(m, " stack"); |
b7643757 SP |
1482 | else |
1483 | seq_printf(m, " stack:%d", tid); | |
1484 | } | |
f69ff943 SW |
1485 | } |
1486 | ||
fc360bd9 | 1487 | if (is_vm_hugetlb_page(vma)) |
17c2b4ee | 1488 | seq_puts(m, " huge"); |
fc360bd9 | 1489 | |
d85f4d6d NH |
1490 | /* mmap_sem is held by m_start */ |
1491 | walk_page_vma(vma, &walk); | |
f69ff943 SW |
1492 | |
1493 | if (!md->pages) | |
1494 | goto out; | |
1495 | ||
1496 | if (md->anon) | |
1497 | seq_printf(m, " anon=%lu", md->anon); | |
1498 | ||
1499 | if (md->dirty) | |
1500 | seq_printf(m, " dirty=%lu", md->dirty); | |
1501 | ||
1502 | if (md->pages != md->anon && md->pages != md->dirty) | |
1503 | seq_printf(m, " mapped=%lu", md->pages); | |
1504 | ||
1505 | if (md->mapcount_max > 1) | |
1506 | seq_printf(m, " mapmax=%lu", md->mapcount_max); | |
1507 | ||
1508 | if (md->swapcache) | |
1509 | seq_printf(m, " swapcache=%lu", md->swapcache); | |
1510 | ||
1511 | if (md->active < md->pages && !is_vm_hugetlb_page(vma)) | |
1512 | seq_printf(m, " active=%lu", md->active); | |
1513 | ||
1514 | if (md->writeback) | |
1515 | seq_printf(m, " writeback=%lu", md->writeback); | |
1516 | ||
948927ee DR |
1517 | for_each_node_state(nid, N_MEMORY) |
1518 | if (md->node[nid]) | |
1519 | seq_printf(m, " N%d=%lu", nid, md->node[nid]); | |
198d1597 RA |
1520 | |
1521 | seq_printf(m, " kernelpagesize_kB=%lu", vma_kernel_pagesize(vma) >> 10); | |
f69ff943 SW |
1522 | out: |
1523 | seq_putc(m, '\n'); | |
b8c20a9b | 1524 | m_cache_vma(m, vma); |
f69ff943 SW |
1525 | return 0; |
1526 | } | |
5b52fc89 | 1527 | |
b7643757 SP |
1528 | static int show_pid_numa_map(struct seq_file *m, void *v) |
1529 | { | |
1530 | return show_numa_map(m, v, 1); | |
1531 | } | |
1532 | ||
1533 | static int show_tid_numa_map(struct seq_file *m, void *v) | |
1534 | { | |
1535 | return show_numa_map(m, v, 0); | |
1536 | } | |
1537 | ||
03a44825 | 1538 | static const struct seq_operations proc_pid_numa_maps_op = { |
b7643757 SP |
1539 | .start = m_start, |
1540 | .next = m_next, | |
1541 | .stop = m_stop, | |
1542 | .show = show_pid_numa_map, | |
6e21c8f1 | 1543 | }; |
662795de | 1544 | |
b7643757 SP |
1545 | static const struct seq_operations proc_tid_numa_maps_op = { |
1546 | .start = m_start, | |
1547 | .next = m_next, | |
1548 | .stop = m_stop, | |
1549 | .show = show_tid_numa_map, | |
1550 | }; | |
1551 | ||
1552 | static int numa_maps_open(struct inode *inode, struct file *file, | |
1553 | const struct seq_operations *ops) | |
662795de | 1554 | { |
4db7d0ee ON |
1555 | return proc_maps_open(inode, file, ops, |
1556 | sizeof(struct numa_maps_private)); | |
662795de EB |
1557 | } |
1558 | ||
b7643757 SP |
1559 | static int pid_numa_maps_open(struct inode *inode, struct file *file) |
1560 | { | |
1561 | return numa_maps_open(inode, file, &proc_pid_numa_maps_op); | |
1562 | } | |
1563 | ||
1564 | static int tid_numa_maps_open(struct inode *inode, struct file *file) | |
1565 | { | |
1566 | return numa_maps_open(inode, file, &proc_tid_numa_maps_op); | |
1567 | } | |
1568 | ||
1569 | const struct file_operations proc_pid_numa_maps_operations = { | |
1570 | .open = pid_numa_maps_open, | |
1571 | .read = seq_read, | |
1572 | .llseek = seq_lseek, | |
29a40ace | 1573 | .release = proc_map_release, |
b7643757 SP |
1574 | }; |
1575 | ||
1576 | const struct file_operations proc_tid_numa_maps_operations = { | |
1577 | .open = tid_numa_maps_open, | |
662795de EB |
1578 | .read = seq_read, |
1579 | .llseek = seq_lseek, | |
29a40ace | 1580 | .release = proc_map_release, |
662795de | 1581 | }; |
f69ff943 | 1582 | #endif /* CONFIG_NUMA */ |