]>
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> | |
0f8975ec | 14 | #include <linux/mmu_notifier.h> |
e070ad49 | 15 | |
1da177e4 LT |
16 | #include <asm/elf.h> |
17 | #include <asm/uaccess.h> | |
e070ad49 | 18 | #include <asm/tlbflush.h> |
1da177e4 LT |
19 | #include "internal.h" |
20 | ||
df5f8314 | 21 | void task_mem(struct seq_file *m, struct mm_struct *mm) |
1da177e4 | 22 | { |
b084d435 | 23 | unsigned long data, text, lib, swap; |
365e9c87 HD |
24 | unsigned long hiwater_vm, total_vm, hiwater_rss, total_rss; |
25 | ||
26 | /* | |
27 | * Note: to minimize their overhead, mm maintains hiwater_vm and | |
28 | * hiwater_rss only when about to *lower* total_vm or rss. Any | |
29 | * collector of these hiwater stats must therefore get total_vm | |
30 | * and rss too, which will usually be the higher. Barriers? not | |
31 | * worth the effort, such snapshots can always be inconsistent. | |
32 | */ | |
33 | hiwater_vm = total_vm = mm->total_vm; | |
34 | if (hiwater_vm < mm->hiwater_vm) | |
35 | hiwater_vm = mm->hiwater_vm; | |
36 | hiwater_rss = total_rss = get_mm_rss(mm); | |
37 | if (hiwater_rss < mm->hiwater_rss) | |
38 | hiwater_rss = mm->hiwater_rss; | |
1da177e4 LT |
39 | |
40 | data = mm->total_vm - mm->shared_vm - mm->stack_vm; | |
41 | text = (PAGE_ALIGN(mm->end_code) - (mm->start_code & PAGE_MASK)) >> 10; | |
42 | lib = (mm->exec_vm << (PAGE_SHIFT-10)) - text; | |
b084d435 | 43 | swap = get_mm_counter(mm, MM_SWAPENTS); |
df5f8314 | 44 | seq_printf(m, |
365e9c87 | 45 | "VmPeak:\t%8lu kB\n" |
1da177e4 LT |
46 | "VmSize:\t%8lu kB\n" |
47 | "VmLck:\t%8lu kB\n" | |
bc3e53f6 | 48 | "VmPin:\t%8lu kB\n" |
365e9c87 | 49 | "VmHWM:\t%8lu kB\n" |
1da177e4 LT |
50 | "VmRSS:\t%8lu kB\n" |
51 | "VmData:\t%8lu kB\n" | |
52 | "VmStk:\t%8lu kB\n" | |
53 | "VmExe:\t%8lu kB\n" | |
54 | "VmLib:\t%8lu kB\n" | |
b084d435 KH |
55 | "VmPTE:\t%8lu kB\n" |
56 | "VmSwap:\t%8lu kB\n", | |
365e9c87 | 57 | hiwater_vm << (PAGE_SHIFT-10), |
314e51b9 | 58 | total_vm << (PAGE_SHIFT-10), |
1da177e4 | 59 | mm->locked_vm << (PAGE_SHIFT-10), |
bc3e53f6 | 60 | mm->pinned_vm << (PAGE_SHIFT-10), |
365e9c87 HD |
61 | hiwater_rss << (PAGE_SHIFT-10), |
62 | total_rss << (PAGE_SHIFT-10), | |
1da177e4 LT |
63 | data << (PAGE_SHIFT-10), |
64 | mm->stack_vm << (PAGE_SHIFT-10), text, lib, | |
b084d435 KH |
65 | (PTRS_PER_PTE*sizeof(pte_t)*mm->nr_ptes) >> 10, |
66 | swap << (PAGE_SHIFT-10)); | |
1da177e4 LT |
67 | } |
68 | ||
69 | unsigned long task_vsize(struct mm_struct *mm) | |
70 | { | |
71 | return PAGE_SIZE * mm->total_vm; | |
72 | } | |
73 | ||
a2ade7b6 AD |
74 | unsigned long task_statm(struct mm_struct *mm, |
75 | unsigned long *shared, unsigned long *text, | |
76 | unsigned long *data, unsigned long *resident) | |
1da177e4 | 77 | { |
d559db08 | 78 | *shared = get_mm_counter(mm, MM_FILEPAGES); |
1da177e4 LT |
79 | *text = (PAGE_ALIGN(mm->end_code) - (mm->start_code & PAGE_MASK)) |
80 | >> PAGE_SHIFT; | |
81 | *data = mm->total_vm - mm->shared_vm; | |
d559db08 | 82 | *resident = *shared + get_mm_counter(mm, MM_ANONPAGES); |
1da177e4 LT |
83 | return mm->total_vm; |
84 | } | |
85 | ||
1da177e4 LT |
86 | static void pad_len_spaces(struct seq_file *m, int len) |
87 | { | |
88 | len = 25 + sizeof(void*) * 6 - len; | |
89 | if (len < 1) | |
90 | len = 1; | |
91 | seq_printf(m, "%*c", len, ' '); | |
92 | } | |
93 | ||
9e781440 KH |
94 | #ifdef CONFIG_NUMA |
95 | /* | |
96 | * These functions are for numa_maps but called in generic **maps seq_file | |
97 | * ->start(), ->stop() ops. | |
98 | * | |
99 | * numa_maps scans all vmas under mmap_sem and checks their mempolicy. | |
100 | * Each mempolicy object is controlled by reference counting. The problem here | |
101 | * is how to avoid accessing dead mempolicy object. | |
102 | * | |
103 | * Because we're holding mmap_sem while reading seq_file, it's safe to access | |
104 | * each vma's mempolicy, no vma objects will never drop refs to mempolicy. | |
105 | * | |
106 | * A task's mempolicy (task->mempolicy) has different behavior. task->mempolicy | |
107 | * is set and replaced under mmap_sem but unrefed and cleared under task_lock(). | |
108 | * So, without task_lock(), we cannot trust get_vma_policy() because we cannot | |
109 | * gurantee the task never exits under us. But taking task_lock() around | |
110 | * get_vma_plicy() causes lock order problem. | |
111 | * | |
112 | * To access task->mempolicy without lock, we hold a reference count of an | |
113 | * object pointed by task->mempolicy and remember it. This will guarantee | |
114 | * that task->mempolicy points to an alive object or NULL in numa_maps accesses. | |
115 | */ | |
116 | static void hold_task_mempolicy(struct proc_maps_private *priv) | |
117 | { | |
118 | struct task_struct *task = priv->task; | |
119 | ||
120 | task_lock(task); | |
121 | priv->task_mempolicy = task->mempolicy; | |
122 | mpol_get(priv->task_mempolicy); | |
123 | task_unlock(task); | |
124 | } | |
125 | static void release_task_mempolicy(struct proc_maps_private *priv) | |
126 | { | |
127 | mpol_put(priv->task_mempolicy); | |
128 | } | |
129 | #else | |
130 | static void hold_task_mempolicy(struct proc_maps_private *priv) | |
131 | { | |
132 | } | |
133 | static void release_task_mempolicy(struct proc_maps_private *priv) | |
134 | { | |
135 | } | |
136 | #endif | |
137 | ||
a6198797 MM |
138 | static void vma_stop(struct proc_maps_private *priv, struct vm_area_struct *vma) |
139 | { | |
140 | if (vma && vma != priv->tail_vma) { | |
141 | struct mm_struct *mm = vma->vm_mm; | |
9e781440 | 142 | release_task_mempolicy(priv); |
a6198797 MM |
143 | up_read(&mm->mmap_sem); |
144 | mmput(mm); | |
145 | } | |
146 | } | |
ec4dd3eb | 147 | |
a6198797 | 148 | static void *m_start(struct seq_file *m, loff_t *pos) |
e070ad49 | 149 | { |
a6198797 MM |
150 | struct proc_maps_private *priv = m->private; |
151 | unsigned long last_addr = m->version; | |
152 | struct mm_struct *mm; | |
153 | struct vm_area_struct *vma, *tail_vma = NULL; | |
154 | loff_t l = *pos; | |
155 | ||
156 | /* Clear the per syscall fields in priv */ | |
157 | priv->task = NULL; | |
158 | priv->tail_vma = NULL; | |
159 | ||
160 | /* | |
161 | * We remember last_addr rather than next_addr to hit with | |
162 | * mmap_cache most of the time. We have zero last_addr at | |
163 | * the beginning and also after lseek. We will have -1 last_addr | |
164 | * after the end of the vmas. | |
165 | */ | |
166 | ||
167 | if (last_addr == -1UL) | |
168 | return NULL; | |
169 | ||
170 | priv->task = get_pid_task(priv->pid, PIDTYPE_PID); | |
171 | if (!priv->task) | |
ec6fd8a4 | 172 | return ERR_PTR(-ESRCH); |
a6198797 | 173 | |
e7dcd999 | 174 | mm = mm_access(priv->task, PTRACE_MODE_READ); |
ec6fd8a4 AV |
175 | if (!mm || IS_ERR(mm)) |
176 | return mm; | |
00f89d21 | 177 | down_read(&mm->mmap_sem); |
a6198797 | 178 | |
31db58b3 | 179 | tail_vma = get_gate_vma(priv->task->mm); |
a6198797 | 180 | priv->tail_vma = tail_vma; |
9e781440 | 181 | hold_task_mempolicy(priv); |
a6198797 MM |
182 | /* Start with last addr hint */ |
183 | vma = find_vma(mm, last_addr); | |
184 | if (last_addr && vma) { | |
185 | vma = vma->vm_next; | |
186 | goto out; | |
187 | } | |
188 | ||
189 | /* | |
190 | * Check the vma index is within the range and do | |
191 | * sequential scan until m_index. | |
192 | */ | |
193 | vma = NULL; | |
194 | if ((unsigned long)l < mm->map_count) { | |
195 | vma = mm->mmap; | |
196 | while (l-- && vma) | |
197 | vma = vma->vm_next; | |
198 | goto out; | |
199 | } | |
200 | ||
201 | if (l != mm->map_count) | |
202 | tail_vma = NULL; /* After gate vma */ | |
203 | ||
204 | out: | |
205 | if (vma) | |
206 | return vma; | |
207 | ||
9e781440 | 208 | release_task_mempolicy(priv); |
a6198797 MM |
209 | /* End of vmas has been reached */ |
210 | m->version = (tail_vma != NULL)? 0: -1UL; | |
211 | up_read(&mm->mmap_sem); | |
212 | mmput(mm); | |
213 | return tail_vma; | |
214 | } | |
215 | ||
216 | static void *m_next(struct seq_file *m, void *v, loff_t *pos) | |
217 | { | |
218 | struct proc_maps_private *priv = m->private; | |
219 | struct vm_area_struct *vma = v; | |
220 | struct vm_area_struct *tail_vma = priv->tail_vma; | |
221 | ||
222 | (*pos)++; | |
223 | if (vma && (vma != tail_vma) && vma->vm_next) | |
224 | return vma->vm_next; | |
225 | vma_stop(priv, vma); | |
226 | return (vma != tail_vma)? tail_vma: NULL; | |
227 | } | |
228 | ||
229 | static void m_stop(struct seq_file *m, void *v) | |
230 | { | |
231 | struct proc_maps_private *priv = m->private; | |
232 | struct vm_area_struct *vma = v; | |
233 | ||
76597cd3 LT |
234 | if (!IS_ERR(vma)) |
235 | vma_stop(priv, vma); | |
a6198797 MM |
236 | if (priv->task) |
237 | put_task_struct(priv->task); | |
238 | } | |
239 | ||
240 | static int do_maps_open(struct inode *inode, struct file *file, | |
03a44825 | 241 | const struct seq_operations *ops) |
a6198797 MM |
242 | { |
243 | struct proc_maps_private *priv; | |
244 | int ret = -ENOMEM; | |
245 | priv = kzalloc(sizeof(*priv), GFP_KERNEL); | |
246 | if (priv) { | |
247 | priv->pid = proc_pid(inode); | |
248 | ret = seq_open(file, ops); | |
249 | if (!ret) { | |
250 | struct seq_file *m = file->private_data; | |
251 | m->private = priv; | |
252 | } else { | |
253 | kfree(priv); | |
254 | } | |
255 | } | |
256 | return ret; | |
257 | } | |
e070ad49 | 258 | |
b7643757 SP |
259 | static void |
260 | show_map_vma(struct seq_file *m, struct vm_area_struct *vma, int is_pid) | |
1da177e4 | 261 | { |
e070ad49 ML |
262 | struct mm_struct *mm = vma->vm_mm; |
263 | struct file *file = vma->vm_file; | |
b7643757 SP |
264 | struct proc_maps_private *priv = m->private; |
265 | struct task_struct *task = priv->task; | |
ca16d140 | 266 | vm_flags_t flags = vma->vm_flags; |
1da177e4 | 267 | unsigned long ino = 0; |
6260a4b0 | 268 | unsigned long long pgoff = 0; |
a09a79f6 | 269 | unsigned long start, end; |
1da177e4 LT |
270 | dev_t dev = 0; |
271 | int len; | |
b7643757 | 272 | const char *name = NULL; |
1da177e4 LT |
273 | |
274 | if (file) { | |
496ad9aa | 275 | struct inode *inode = file_inode(vma->vm_file); |
1da177e4 LT |
276 | dev = inode->i_sb->s_dev; |
277 | ino = inode->i_ino; | |
6260a4b0 | 278 | pgoff = ((loff_t)vma->vm_pgoff) << PAGE_SHIFT; |
1da177e4 LT |
279 | } |
280 | ||
d7824370 LT |
281 | /* We don't show the stack guard page in /proc/maps */ |
282 | start = vma->vm_start; | |
a09a79f6 MP |
283 | if (stack_guard_page_start(vma, start)) |
284 | start += PAGE_SIZE; | |
285 | end = vma->vm_end; | |
286 | if (stack_guard_page_end(vma, end)) | |
287 | end -= PAGE_SIZE; | |
d7824370 | 288 | |
1804dc6e | 289 | seq_printf(m, "%08lx-%08lx %c%c%c%c %08llx %02x:%02x %lu %n", |
d7824370 | 290 | start, |
a09a79f6 | 291 | end, |
1da177e4 LT |
292 | flags & VM_READ ? 'r' : '-', |
293 | flags & VM_WRITE ? 'w' : '-', | |
294 | flags & VM_EXEC ? 'x' : '-', | |
295 | flags & VM_MAYSHARE ? 's' : 'p', | |
6260a4b0 | 296 | pgoff, |
1da177e4 LT |
297 | MAJOR(dev), MINOR(dev), ino, &len); |
298 | ||
299 | /* | |
300 | * Print the dentry name for named mappings, and a | |
301 | * special [heap] marker for the heap: | |
302 | */ | |
e070ad49 | 303 | if (file) { |
1da177e4 | 304 | pad_len_spaces(m, len); |
c32c2f63 | 305 | seq_path(m, &file->f_path, "\n"); |
b7643757 SP |
306 | goto done; |
307 | } | |
308 | ||
309 | name = arch_vma_name(vma); | |
310 | if (!name) { | |
311 | pid_t tid; | |
312 | ||
313 | if (!mm) { | |
314 | name = "[vdso]"; | |
315 | goto done; | |
316 | } | |
317 | ||
318 | if (vma->vm_start <= mm->brk && | |
319 | vma->vm_end >= mm->start_brk) { | |
320 | name = "[heap]"; | |
321 | goto done; | |
322 | } | |
323 | ||
324 | tid = vm_is_stack(task, vma, is_pid); | |
325 | ||
326 | if (tid != 0) { | |
327 | /* | |
328 | * Thread stack in /proc/PID/task/TID/maps or | |
329 | * the main process stack. | |
330 | */ | |
331 | if (!is_pid || (vma->vm_start <= mm->start_stack && | |
332 | vma->vm_end >= mm->start_stack)) { | |
333 | name = "[stack]"; | |
e6e5494c | 334 | } else { |
b7643757 SP |
335 | /* Thread stack in /proc/PID/maps */ |
336 | pad_len_spaces(m, len); | |
337 | seq_printf(m, "[stack:%d]", tid); | |
1da177e4 | 338 | } |
e6e5494c | 339 | } |
b7643757 SP |
340 | } |
341 | ||
342 | done: | |
343 | if (name) { | |
344 | pad_len_spaces(m, len); | |
345 | seq_puts(m, name); | |
1da177e4 LT |
346 | } |
347 | seq_putc(m, '\n'); | |
7c88db0c JK |
348 | } |
349 | ||
b7643757 | 350 | static int show_map(struct seq_file *m, void *v, int is_pid) |
7c88db0c JK |
351 | { |
352 | struct vm_area_struct *vma = v; | |
353 | struct proc_maps_private *priv = m->private; | |
354 | struct task_struct *task = priv->task; | |
355 | ||
b7643757 | 356 | show_map_vma(m, vma, is_pid); |
e070ad49 | 357 | |
e070ad49 | 358 | if (m->count < m->size) /* vma is copied successfully */ |
31db58b3 SW |
359 | m->version = (vma != get_gate_vma(task->mm)) |
360 | ? vma->vm_start : 0; | |
1da177e4 LT |
361 | return 0; |
362 | } | |
363 | ||
b7643757 SP |
364 | static int show_pid_map(struct seq_file *m, void *v) |
365 | { | |
366 | return show_map(m, v, 1); | |
367 | } | |
368 | ||
369 | static int show_tid_map(struct seq_file *m, void *v) | |
370 | { | |
371 | return show_map(m, v, 0); | |
372 | } | |
373 | ||
03a44825 | 374 | static const struct seq_operations proc_pid_maps_op = { |
a6198797 MM |
375 | .start = m_start, |
376 | .next = m_next, | |
377 | .stop = m_stop, | |
b7643757 SP |
378 | .show = show_pid_map |
379 | }; | |
380 | ||
381 | static const struct seq_operations proc_tid_maps_op = { | |
382 | .start = m_start, | |
383 | .next = m_next, | |
384 | .stop = m_stop, | |
385 | .show = show_tid_map | |
a6198797 MM |
386 | }; |
387 | ||
b7643757 | 388 | static int pid_maps_open(struct inode *inode, struct file *file) |
a6198797 MM |
389 | { |
390 | return do_maps_open(inode, file, &proc_pid_maps_op); | |
391 | } | |
392 | ||
b7643757 SP |
393 | static int tid_maps_open(struct inode *inode, struct file *file) |
394 | { | |
395 | return do_maps_open(inode, file, &proc_tid_maps_op); | |
396 | } | |
397 | ||
398 | const struct file_operations proc_pid_maps_operations = { | |
399 | .open = pid_maps_open, | |
400 | .read = seq_read, | |
401 | .llseek = seq_lseek, | |
402 | .release = seq_release_private, | |
403 | }; | |
404 | ||
405 | const struct file_operations proc_tid_maps_operations = { | |
406 | .open = tid_maps_open, | |
a6198797 MM |
407 | .read = seq_read, |
408 | .llseek = seq_lseek, | |
409 | .release = seq_release_private, | |
410 | }; | |
411 | ||
412 | /* | |
413 | * Proportional Set Size(PSS): my share of RSS. | |
414 | * | |
415 | * PSS of a process is the count of pages it has in memory, where each | |
416 | * page is divided by the number of processes sharing it. So if a | |
417 | * process has 1000 pages all to itself, and 1000 shared with one other | |
418 | * process, its PSS will be 1500. | |
419 | * | |
420 | * To keep (accumulated) division errors low, we adopt a 64bit | |
421 | * fixed-point pss counter to minimize division errors. So (pss >> | |
422 | * PSS_SHIFT) would be the real byte count. | |
423 | * | |
424 | * A shift of 12 before division means (assuming 4K page size): | |
425 | * - 1M 3-user-pages add up to 8KB errors; | |
426 | * - supports mapcount up to 2^24, or 16M; | |
427 | * - supports PSS up to 2^52 bytes, or 4PB. | |
428 | */ | |
429 | #define PSS_SHIFT 12 | |
430 | ||
1e883281 | 431 | #ifdef CONFIG_PROC_PAGE_MONITOR |
214e471f | 432 | struct mem_size_stats { |
a6198797 MM |
433 | struct vm_area_struct *vma; |
434 | unsigned long resident; | |
435 | unsigned long shared_clean; | |
436 | unsigned long shared_dirty; | |
437 | unsigned long private_clean; | |
438 | unsigned long private_dirty; | |
439 | unsigned long referenced; | |
b40d4f84 | 440 | unsigned long anonymous; |
4031a219 | 441 | unsigned long anonymous_thp; |
214e471f | 442 | unsigned long swap; |
bca15543 | 443 | unsigned long nonlinear; |
a6198797 MM |
444 | u64 pss; |
445 | }; | |
446 | ||
ae11c4d9 DH |
447 | |
448 | static void smaps_pte_entry(pte_t ptent, unsigned long addr, | |
3c9acc78 | 449 | unsigned long ptent_size, struct mm_walk *walk) |
ae11c4d9 DH |
450 | { |
451 | struct mem_size_stats *mss = walk->private; | |
452 | struct vm_area_struct *vma = mss->vma; | |
bca15543 | 453 | pgoff_t pgoff = linear_page_index(vma, addr); |
b1d4d9e0 | 454 | struct page *page = NULL; |
ae11c4d9 DH |
455 | int mapcount; |
456 | ||
b1d4d9e0 KK |
457 | if (pte_present(ptent)) { |
458 | page = vm_normal_page(vma, addr, ptent); | |
459 | } else if (is_swap_pte(ptent)) { | |
460 | swp_entry_t swpent = pte_to_swp_entry(ptent); | |
ae11c4d9 | 461 | |
b1d4d9e0 KK |
462 | if (!non_swap_entry(swpent)) |
463 | mss->swap += ptent_size; | |
464 | else if (is_migration_entry(swpent)) | |
465 | page = migration_entry_to_page(swpent); | |
bca15543 KK |
466 | } else if (pte_file(ptent)) { |
467 | if (pte_to_pgoff(ptent) != pgoff) | |
468 | mss->nonlinear += ptent_size; | |
b1d4d9e0 | 469 | } |
ae11c4d9 | 470 | |
ae11c4d9 DH |
471 | if (!page) |
472 | return; | |
473 | ||
474 | if (PageAnon(page)) | |
3c9acc78 | 475 | mss->anonymous += ptent_size; |
ae11c4d9 | 476 | |
bca15543 KK |
477 | if (page->index != pgoff) |
478 | mss->nonlinear += ptent_size; | |
479 | ||
3c9acc78 | 480 | mss->resident += ptent_size; |
ae11c4d9 DH |
481 | /* Accumulate the size in pages that have been accessed. */ |
482 | if (pte_young(ptent) || PageReferenced(page)) | |
3c9acc78 | 483 | mss->referenced += ptent_size; |
ae11c4d9 DH |
484 | mapcount = page_mapcount(page); |
485 | if (mapcount >= 2) { | |
486 | if (pte_dirty(ptent) || PageDirty(page)) | |
3c9acc78 | 487 | mss->shared_dirty += ptent_size; |
ae11c4d9 | 488 | else |
3c9acc78 DH |
489 | mss->shared_clean += ptent_size; |
490 | mss->pss += (ptent_size << PSS_SHIFT) / mapcount; | |
ae11c4d9 DH |
491 | } else { |
492 | if (pte_dirty(ptent) || PageDirty(page)) | |
3c9acc78 | 493 | mss->private_dirty += ptent_size; |
ae11c4d9 | 494 | else |
3c9acc78 DH |
495 | mss->private_clean += ptent_size; |
496 | mss->pss += (ptent_size << PSS_SHIFT); | |
ae11c4d9 DH |
497 | } |
498 | } | |
499 | ||
b3ae5acb | 500 | static int smaps_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end, |
2165009b | 501 | struct mm_walk *walk) |
e070ad49 | 502 | { |
2165009b | 503 | struct mem_size_stats *mss = walk->private; |
b3ae5acb | 504 | struct vm_area_struct *vma = mss->vma; |
ae11c4d9 | 505 | pte_t *pte; |
705e87c0 | 506 | spinlock_t *ptl; |
e070ad49 | 507 | |
025c5b24 NH |
508 | if (pmd_trans_huge_lock(pmd, vma) == 1) { |
509 | smaps_pte_entry(*(pte_t *)pmd, addr, HPAGE_PMD_SIZE, walk); | |
22e057c5 | 510 | spin_unlock(&walk->mm->page_table_lock); |
025c5b24 NH |
511 | mss->anonymous_thp += HPAGE_PMD_SIZE; |
512 | return 0; | |
22e057c5 | 513 | } |
1a5a9906 AA |
514 | |
515 | if (pmd_trans_unstable(pmd)) | |
516 | return 0; | |
22e057c5 DH |
517 | /* |
518 | * The mmap_sem held all the way back in m_start() is what | |
519 | * keeps khugepaged out of here and from collapsing things | |
520 | * in here. | |
521 | */ | |
705e87c0 | 522 | pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl); |
ae11c4d9 | 523 | for (; addr != end; pte++, addr += PAGE_SIZE) |
3c9acc78 | 524 | smaps_pte_entry(*pte, addr, PAGE_SIZE, walk); |
705e87c0 HD |
525 | pte_unmap_unlock(pte - 1, ptl); |
526 | cond_resched(); | |
b3ae5acb | 527 | return 0; |
e070ad49 ML |
528 | } |
529 | ||
834f82e2 CG |
530 | static void show_smap_vma_flags(struct seq_file *m, struct vm_area_struct *vma) |
531 | { | |
532 | /* | |
533 | * Don't forget to update Documentation/ on changes. | |
534 | */ | |
535 | static const char mnemonics[BITS_PER_LONG][2] = { | |
536 | /* | |
537 | * In case if we meet a flag we don't know about. | |
538 | */ | |
539 | [0 ... (BITS_PER_LONG-1)] = "??", | |
540 | ||
541 | [ilog2(VM_READ)] = "rd", | |
542 | [ilog2(VM_WRITE)] = "wr", | |
543 | [ilog2(VM_EXEC)] = "ex", | |
544 | [ilog2(VM_SHARED)] = "sh", | |
545 | [ilog2(VM_MAYREAD)] = "mr", | |
546 | [ilog2(VM_MAYWRITE)] = "mw", | |
547 | [ilog2(VM_MAYEXEC)] = "me", | |
548 | [ilog2(VM_MAYSHARE)] = "ms", | |
549 | [ilog2(VM_GROWSDOWN)] = "gd", | |
550 | [ilog2(VM_PFNMAP)] = "pf", | |
551 | [ilog2(VM_DENYWRITE)] = "dw", | |
552 | [ilog2(VM_LOCKED)] = "lo", | |
553 | [ilog2(VM_IO)] = "io", | |
554 | [ilog2(VM_SEQ_READ)] = "sr", | |
555 | [ilog2(VM_RAND_READ)] = "rr", | |
556 | [ilog2(VM_DONTCOPY)] = "dc", | |
557 | [ilog2(VM_DONTEXPAND)] = "de", | |
558 | [ilog2(VM_ACCOUNT)] = "ac", | |
559 | [ilog2(VM_NORESERVE)] = "nr", | |
560 | [ilog2(VM_HUGETLB)] = "ht", | |
561 | [ilog2(VM_NONLINEAR)] = "nl", | |
562 | [ilog2(VM_ARCH_1)] = "ar", | |
563 | [ilog2(VM_DONTDUMP)] = "dd", | |
564 | [ilog2(VM_MIXEDMAP)] = "mm", | |
565 | [ilog2(VM_HUGEPAGE)] = "hg", | |
566 | [ilog2(VM_NOHUGEPAGE)] = "nh", | |
567 | [ilog2(VM_MERGEABLE)] = "mg", | |
568 | }; | |
569 | size_t i; | |
570 | ||
571 | seq_puts(m, "VmFlags: "); | |
572 | for (i = 0; i < BITS_PER_LONG; i++) { | |
573 | if (vma->vm_flags & (1UL << i)) { | |
574 | seq_printf(m, "%c%c ", | |
575 | mnemonics[i][0], mnemonics[i][1]); | |
576 | } | |
577 | } | |
578 | seq_putc(m, '\n'); | |
579 | } | |
580 | ||
b7643757 | 581 | static int show_smap(struct seq_file *m, void *v, int is_pid) |
e070ad49 | 582 | { |
7c88db0c JK |
583 | struct proc_maps_private *priv = m->private; |
584 | struct task_struct *task = priv->task; | |
e070ad49 | 585 | struct vm_area_struct *vma = v; |
e070ad49 | 586 | struct mem_size_stats mss; |
2165009b DH |
587 | struct mm_walk smaps_walk = { |
588 | .pmd_entry = smaps_pte_range, | |
589 | .mm = vma->vm_mm, | |
590 | .private = &mss, | |
591 | }; | |
e070ad49 ML |
592 | |
593 | memset(&mss, 0, sizeof mss); | |
b3ae5acb | 594 | mss.vma = vma; |
d82ef020 | 595 | /* mmap_sem is held in m_start */ |
5ddfae16 | 596 | if (vma->vm_mm && !is_vm_hugetlb_page(vma)) |
2165009b | 597 | walk_page_range(vma->vm_start, vma->vm_end, &smaps_walk); |
4752c369 | 598 | |
b7643757 | 599 | show_map_vma(m, vma, is_pid); |
4752c369 MM |
600 | |
601 | seq_printf(m, | |
602 | "Size: %8lu kB\n" | |
603 | "Rss: %8lu kB\n" | |
604 | "Pss: %8lu kB\n" | |
605 | "Shared_Clean: %8lu kB\n" | |
606 | "Shared_Dirty: %8lu kB\n" | |
607 | "Private_Clean: %8lu kB\n" | |
608 | "Private_Dirty: %8lu kB\n" | |
214e471f | 609 | "Referenced: %8lu kB\n" |
b40d4f84 | 610 | "Anonymous: %8lu kB\n" |
4031a219 | 611 | "AnonHugePages: %8lu kB\n" |
08fba699 | 612 | "Swap: %8lu kB\n" |
3340289d | 613 | "KernelPageSize: %8lu kB\n" |
2d90508f NK |
614 | "MMUPageSize: %8lu kB\n" |
615 | "Locked: %8lu kB\n", | |
4752c369 MM |
616 | (vma->vm_end - vma->vm_start) >> 10, |
617 | mss.resident >> 10, | |
618 | (unsigned long)(mss.pss >> (10 + PSS_SHIFT)), | |
619 | mss.shared_clean >> 10, | |
620 | mss.shared_dirty >> 10, | |
621 | mss.private_clean >> 10, | |
622 | mss.private_dirty >> 10, | |
214e471f | 623 | mss.referenced >> 10, |
b40d4f84 | 624 | mss.anonymous >> 10, |
4031a219 | 625 | mss.anonymous_thp >> 10, |
08fba699 | 626 | mss.swap >> 10, |
3340289d | 627 | vma_kernel_pagesize(vma) >> 10, |
2d90508f NK |
628 | vma_mmu_pagesize(vma) >> 10, |
629 | (vma->vm_flags & VM_LOCKED) ? | |
630 | (unsigned long)(mss.pss >> (10 + PSS_SHIFT)) : 0); | |
4752c369 | 631 | |
bca15543 KK |
632 | if (vma->vm_flags & VM_NONLINEAR) |
633 | seq_printf(m, "Nonlinear: %8lu kB\n", | |
634 | mss.nonlinear >> 10); | |
635 | ||
834f82e2 CG |
636 | show_smap_vma_flags(m, vma); |
637 | ||
7c88db0c | 638 | if (m->count < m->size) /* vma is copied successfully */ |
31db58b3 SW |
639 | m->version = (vma != get_gate_vma(task->mm)) |
640 | ? vma->vm_start : 0; | |
7c88db0c | 641 | return 0; |
e070ad49 ML |
642 | } |
643 | ||
b7643757 SP |
644 | static int show_pid_smap(struct seq_file *m, void *v) |
645 | { | |
646 | return show_smap(m, v, 1); | |
647 | } | |
648 | ||
649 | static int show_tid_smap(struct seq_file *m, void *v) | |
650 | { | |
651 | return show_smap(m, v, 0); | |
652 | } | |
653 | ||
03a44825 | 654 | static const struct seq_operations proc_pid_smaps_op = { |
a6198797 MM |
655 | .start = m_start, |
656 | .next = m_next, | |
657 | .stop = m_stop, | |
b7643757 SP |
658 | .show = show_pid_smap |
659 | }; | |
660 | ||
661 | static const struct seq_operations proc_tid_smaps_op = { | |
662 | .start = m_start, | |
663 | .next = m_next, | |
664 | .stop = m_stop, | |
665 | .show = show_tid_smap | |
a6198797 MM |
666 | }; |
667 | ||
b7643757 | 668 | static int pid_smaps_open(struct inode *inode, struct file *file) |
a6198797 MM |
669 | { |
670 | return do_maps_open(inode, file, &proc_pid_smaps_op); | |
671 | } | |
672 | ||
b7643757 SP |
673 | static int tid_smaps_open(struct inode *inode, struct file *file) |
674 | { | |
675 | return do_maps_open(inode, file, &proc_tid_smaps_op); | |
676 | } | |
677 | ||
678 | const struct file_operations proc_pid_smaps_operations = { | |
679 | .open = pid_smaps_open, | |
680 | .read = seq_read, | |
681 | .llseek = seq_lseek, | |
682 | .release = seq_release_private, | |
683 | }; | |
684 | ||
685 | const struct file_operations proc_tid_smaps_operations = { | |
686 | .open = tid_smaps_open, | |
a6198797 MM |
687 | .read = seq_read, |
688 | .llseek = seq_lseek, | |
689 | .release = seq_release_private, | |
690 | }; | |
691 | ||
541c237c PE |
692 | /* |
693 | * We do not want to have constant page-shift bits sitting in | |
694 | * pagemap entries and are about to reuse them some time soon. | |
695 | * | |
696 | * Here's the "migration strategy": | |
697 | * 1. when the system boots these bits remain what they are, | |
698 | * but a warning about future change is printed in log; | |
699 | * 2. once anyone clears soft-dirty bits via clear_refs file, | |
700 | * these flag is set to denote, that user is aware of the | |
701 | * new API and those page-shift bits change their meaning. | |
702 | * The respective warning is printed in dmesg; | |
703 | * 3. In a couple of releases we will remove all the mentions | |
704 | * of page-shift in pagemap entries. | |
705 | */ | |
706 | ||
707 | static bool soft_dirty_cleared __read_mostly; | |
708 | ||
040fa020 PE |
709 | enum clear_refs_types { |
710 | CLEAR_REFS_ALL = 1, | |
711 | CLEAR_REFS_ANON, | |
712 | CLEAR_REFS_MAPPED, | |
0f8975ec | 713 | CLEAR_REFS_SOFT_DIRTY, |
040fa020 PE |
714 | CLEAR_REFS_LAST, |
715 | }; | |
716 | ||
af9de7eb PE |
717 | struct clear_refs_private { |
718 | struct vm_area_struct *vma; | |
0f8975ec | 719 | enum clear_refs_types type; |
af9de7eb PE |
720 | }; |
721 | ||
0f8975ec PE |
722 | static inline void clear_soft_dirty(struct vm_area_struct *vma, |
723 | unsigned long addr, pte_t *pte) | |
724 | { | |
725 | #ifdef CONFIG_MEM_SOFT_DIRTY | |
726 | /* | |
727 | * The soft-dirty tracker uses #PF-s to catch writes | |
728 | * to pages, so write-protect the pte as well. See the | |
729 | * Documentation/vm/soft-dirty.txt for full description | |
730 | * of how soft-dirty works. | |
731 | */ | |
732 | pte_t ptent = *pte; | |
733 | ptent = pte_wrprotect(ptent); | |
734 | ptent = pte_clear_flags(ptent, _PAGE_SOFT_DIRTY); | |
735 | set_pte_at(vma->vm_mm, addr, pte, ptent); | |
736 | #endif | |
737 | } | |
738 | ||
a6198797 | 739 | static int clear_refs_pte_range(pmd_t *pmd, unsigned long addr, |
2165009b | 740 | unsigned long end, struct mm_walk *walk) |
a6198797 | 741 | { |
af9de7eb PE |
742 | struct clear_refs_private *cp = walk->private; |
743 | struct vm_area_struct *vma = cp->vma; | |
a6198797 MM |
744 | pte_t *pte, ptent; |
745 | spinlock_t *ptl; | |
746 | struct page *page; | |
747 | ||
e180377f | 748 | split_huge_page_pmd(vma, addr, pmd); |
1a5a9906 AA |
749 | if (pmd_trans_unstable(pmd)) |
750 | return 0; | |
03319327 | 751 | |
a6198797 MM |
752 | pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl); |
753 | for (; addr != end; pte++, addr += PAGE_SIZE) { | |
754 | ptent = *pte; | |
755 | if (!pte_present(ptent)) | |
756 | continue; | |
757 | ||
0f8975ec PE |
758 | if (cp->type == CLEAR_REFS_SOFT_DIRTY) { |
759 | clear_soft_dirty(vma, addr, pte); | |
760 | continue; | |
761 | } | |
762 | ||
a6198797 MM |
763 | page = vm_normal_page(vma, addr, ptent); |
764 | if (!page) | |
765 | continue; | |
766 | ||
767 | /* Clear accessed and referenced bits. */ | |
768 | ptep_test_and_clear_young(vma, addr, pte); | |
769 | ClearPageReferenced(page); | |
770 | } | |
771 | pte_unmap_unlock(pte - 1, ptl); | |
772 | cond_resched(); | |
773 | return 0; | |
774 | } | |
775 | ||
f248dcb3 MM |
776 | static ssize_t clear_refs_write(struct file *file, const char __user *buf, |
777 | size_t count, loff_t *ppos) | |
b813e931 | 778 | { |
f248dcb3 | 779 | struct task_struct *task; |
fb92a4b0 | 780 | char buffer[PROC_NUMBUF]; |
f248dcb3 | 781 | struct mm_struct *mm; |
b813e931 | 782 | struct vm_area_struct *vma; |
040fa020 PE |
783 | enum clear_refs_types type; |
784 | int itype; | |
0a8cb8e3 | 785 | int rv; |
b813e931 | 786 | |
f248dcb3 MM |
787 | memset(buffer, 0, sizeof(buffer)); |
788 | if (count > sizeof(buffer) - 1) | |
789 | count = sizeof(buffer) - 1; | |
790 | if (copy_from_user(buffer, buf, count)) | |
791 | return -EFAULT; | |
040fa020 | 792 | rv = kstrtoint(strstrip(buffer), 10, &itype); |
0a8cb8e3 AD |
793 | if (rv < 0) |
794 | return rv; | |
040fa020 PE |
795 | type = (enum clear_refs_types)itype; |
796 | if (type < CLEAR_REFS_ALL || type >= CLEAR_REFS_LAST) | |
f248dcb3 | 797 | return -EINVAL; |
541c237c PE |
798 | |
799 | if (type == CLEAR_REFS_SOFT_DIRTY) { | |
800 | soft_dirty_cleared = true; | |
801 | pr_warn_once("The pagemap bits 55-60 has changed their meaning! " | |
802 | "See the linux/Documentation/vm/pagemap.txt for details.\n"); | |
803 | } | |
804 | ||
496ad9aa | 805 | task = get_proc_task(file_inode(file)); |
f248dcb3 MM |
806 | if (!task) |
807 | return -ESRCH; | |
808 | mm = get_task_mm(task); | |
809 | if (mm) { | |
af9de7eb | 810 | struct clear_refs_private cp = { |
0f8975ec | 811 | .type = type, |
af9de7eb | 812 | }; |
20cbc972 AM |
813 | struct mm_walk clear_refs_walk = { |
814 | .pmd_entry = clear_refs_pte_range, | |
815 | .mm = mm, | |
af9de7eb | 816 | .private = &cp, |
20cbc972 | 817 | }; |
f248dcb3 | 818 | down_read(&mm->mmap_sem); |
0f8975ec PE |
819 | if (type == CLEAR_REFS_SOFT_DIRTY) |
820 | mmu_notifier_invalidate_range_start(mm, 0, -1); | |
2165009b | 821 | for (vma = mm->mmap; vma; vma = vma->vm_next) { |
af9de7eb | 822 | cp.vma = vma; |
398499d5 MB |
823 | if (is_vm_hugetlb_page(vma)) |
824 | continue; | |
825 | /* | |
826 | * Writing 1 to /proc/pid/clear_refs affects all pages. | |
827 | * | |
828 | * Writing 2 to /proc/pid/clear_refs only affects | |
829 | * Anonymous pages. | |
830 | * | |
831 | * Writing 3 to /proc/pid/clear_refs only affects file | |
832 | * mapped pages. | |
833 | */ | |
834 | if (type == CLEAR_REFS_ANON && vma->vm_file) | |
835 | continue; | |
836 | if (type == CLEAR_REFS_MAPPED && !vma->vm_file) | |
837 | continue; | |
838 | walk_page_range(vma->vm_start, vma->vm_end, | |
839 | &clear_refs_walk); | |
2165009b | 840 | } |
0f8975ec PE |
841 | if (type == CLEAR_REFS_SOFT_DIRTY) |
842 | mmu_notifier_invalidate_range_end(mm, 0, -1); | |
f248dcb3 MM |
843 | flush_tlb_mm(mm); |
844 | up_read(&mm->mmap_sem); | |
845 | mmput(mm); | |
846 | } | |
847 | put_task_struct(task); | |
fb92a4b0 VL |
848 | |
849 | return count; | |
b813e931 DR |
850 | } |
851 | ||
f248dcb3 MM |
852 | const struct file_operations proc_clear_refs_operations = { |
853 | .write = clear_refs_write, | |
6038f373 | 854 | .llseek = noop_llseek, |
f248dcb3 MM |
855 | }; |
856 | ||
092b50ba NH |
857 | typedef struct { |
858 | u64 pme; | |
859 | } pagemap_entry_t; | |
860 | ||
85863e47 | 861 | struct pagemapread { |
d82ef020 | 862 | int pos, len; |
092b50ba | 863 | pagemap_entry_t *buffer; |
2b0a9f01 | 864 | bool v2; |
85863e47 MM |
865 | }; |
866 | ||
5aaabe83 NH |
867 | #define PAGEMAP_WALK_SIZE (PMD_SIZE) |
868 | #define PAGEMAP_WALK_MASK (PMD_MASK) | |
869 | ||
f16278c6 HR |
870 | #define PM_ENTRY_BYTES sizeof(u64) |
871 | #define PM_STATUS_BITS 3 | |
872 | #define PM_STATUS_OFFSET (64 - PM_STATUS_BITS) | |
873 | #define PM_STATUS_MASK (((1LL << PM_STATUS_BITS) - 1) << PM_STATUS_OFFSET) | |
874 | #define PM_STATUS(nr) (((nr) << PM_STATUS_OFFSET) & PM_STATUS_MASK) | |
875 | #define PM_PSHIFT_BITS 6 | |
876 | #define PM_PSHIFT_OFFSET (PM_STATUS_OFFSET - PM_PSHIFT_BITS) | |
877 | #define PM_PSHIFT_MASK (((1LL << PM_PSHIFT_BITS) - 1) << PM_PSHIFT_OFFSET) | |
2b0a9f01 | 878 | #define __PM_PSHIFT(x) (((u64) (x) << PM_PSHIFT_OFFSET) & PM_PSHIFT_MASK) |
f16278c6 HR |
879 | #define PM_PFRAME_MASK ((1LL << PM_PSHIFT_OFFSET) - 1) |
880 | #define PM_PFRAME(x) ((x) & PM_PFRAME_MASK) | |
2b0a9f01 PE |
881 | /* in "new" pagemap pshift bits are occupied with more status bits */ |
882 | #define PM_STATUS2(v2, x) (__PM_PSHIFT(v2 ? x : PAGE_SHIFT)) | |
f16278c6 | 883 | |
0f8975ec | 884 | #define __PM_SOFT_DIRTY (1LL) |
f16278c6 HR |
885 | #define PM_PRESENT PM_STATUS(4LL) |
886 | #define PM_SWAP PM_STATUS(2LL) | |
052fb0d6 | 887 | #define PM_FILE PM_STATUS(1LL) |
2b0a9f01 | 888 | #define PM_NOT_PRESENT(v2) PM_STATUS2(v2, 0) |
85863e47 MM |
889 | #define PM_END_OF_BUFFER 1 |
890 | ||
092b50ba NH |
891 | static inline pagemap_entry_t make_pme(u64 val) |
892 | { | |
893 | return (pagemap_entry_t) { .pme = val }; | |
894 | } | |
895 | ||
896 | static int add_to_pagemap(unsigned long addr, pagemap_entry_t *pme, | |
85863e47 MM |
897 | struct pagemapread *pm) |
898 | { | |
092b50ba | 899 | pm->buffer[pm->pos++] = *pme; |
d82ef020 | 900 | if (pm->pos >= pm->len) |
aae8679b | 901 | return PM_END_OF_BUFFER; |
85863e47 MM |
902 | return 0; |
903 | } | |
904 | ||
905 | static int pagemap_pte_hole(unsigned long start, unsigned long end, | |
2165009b | 906 | struct mm_walk *walk) |
85863e47 | 907 | { |
2165009b | 908 | struct pagemapread *pm = walk->private; |
85863e47 MM |
909 | unsigned long addr; |
910 | int err = 0; | |
2b0a9f01 | 911 | pagemap_entry_t pme = make_pme(PM_NOT_PRESENT(pm->v2)); |
092b50ba | 912 | |
85863e47 | 913 | for (addr = start; addr < end; addr += PAGE_SIZE) { |
092b50ba | 914 | err = add_to_pagemap(addr, &pme, pm); |
85863e47 MM |
915 | if (err) |
916 | break; | |
917 | } | |
918 | return err; | |
919 | } | |
920 | ||
2b0a9f01 | 921 | static void pte_to_pagemap_entry(pagemap_entry_t *pme, struct pagemapread *pm, |
052fb0d6 | 922 | struct vm_area_struct *vma, unsigned long addr, pte_t pte) |
85863e47 | 923 | { |
052fb0d6 KK |
924 | u64 frame, flags; |
925 | struct page *page = NULL; | |
0f8975ec | 926 | int flags2 = 0; |
85863e47 | 927 | |
052fb0d6 KK |
928 | if (pte_present(pte)) { |
929 | frame = pte_pfn(pte); | |
930 | flags = PM_PRESENT; | |
931 | page = vm_normal_page(vma, addr, pte); | |
932 | } else if (is_swap_pte(pte)) { | |
933 | swp_entry_t entry = pte_to_swp_entry(pte); | |
934 | ||
935 | frame = swp_type(entry) | | |
936 | (swp_offset(entry) << MAX_SWAPFILES_SHIFT); | |
937 | flags = PM_SWAP; | |
938 | if (is_migration_entry(entry)) | |
939 | page = migration_entry_to_page(entry); | |
940 | } else { | |
2b0a9f01 | 941 | *pme = make_pme(PM_NOT_PRESENT(pm->v2)); |
052fb0d6 KK |
942 | return; |
943 | } | |
944 | ||
945 | if (page && !PageAnon(page)) | |
946 | flags |= PM_FILE; | |
0f8975ec PE |
947 | if (pte_soft_dirty(pte)) |
948 | flags2 |= __PM_SOFT_DIRTY; | |
052fb0d6 | 949 | |
0f8975ec | 950 | *pme = make_pme(PM_PFRAME(frame) | PM_STATUS2(pm->v2, flags2) | flags); |
bcf8039e DH |
951 | } |
952 | ||
5aaabe83 | 953 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
2b0a9f01 | 954 | static void thp_pmd_to_pagemap_entry(pagemap_entry_t *pme, struct pagemapread *pm, |
0f8975ec | 955 | pmd_t pmd, int offset, int pmd_flags2) |
5aaabe83 | 956 | { |
5aaabe83 NH |
957 | /* |
958 | * Currently pmd for thp is always present because thp can not be | |
959 | * swapped-out, migrated, or HWPOISONed (split in such cases instead.) | |
960 | * This if-check is just to prepare for future implementation. | |
961 | */ | |
962 | if (pmd_present(pmd)) | |
092b50ba | 963 | *pme = make_pme(PM_PFRAME(pmd_pfn(pmd) + offset) |
0f8975ec | 964 | | PM_STATUS2(pm->v2, pmd_flags2) | PM_PRESENT); |
16fbdce6 | 965 | else |
2b0a9f01 | 966 | *pme = make_pme(PM_NOT_PRESENT(pm->v2)); |
5aaabe83 NH |
967 | } |
968 | #else | |
2b0a9f01 | 969 | static inline void thp_pmd_to_pagemap_entry(pagemap_entry_t *pme, struct pagemapread *pm, |
0f8975ec | 970 | pmd_t pmd, int offset, int pmd_flags2) |
5aaabe83 | 971 | { |
5aaabe83 NH |
972 | } |
973 | #endif | |
974 | ||
85863e47 | 975 | static int pagemap_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end, |
2165009b | 976 | struct mm_walk *walk) |
85863e47 | 977 | { |
bcf8039e | 978 | struct vm_area_struct *vma; |
2165009b | 979 | struct pagemapread *pm = walk->private; |
85863e47 MM |
980 | pte_t *pte; |
981 | int err = 0; | |
2b0a9f01 | 982 | pagemap_entry_t pme = make_pme(PM_NOT_PRESENT(pm->v2)); |
85863e47 | 983 | |
bcf8039e DH |
984 | /* find the first VMA at or above 'addr' */ |
985 | vma = find_vma(walk->mm, addr); | |
08fa29d9 | 986 | if (vma && pmd_trans_huge_lock(pmd, vma) == 1) { |
0f8975ec PE |
987 | int pmd_flags2; |
988 | ||
989 | pmd_flags2 = (pmd_soft_dirty(*pmd) ? __PM_SOFT_DIRTY : 0); | |
025c5b24 NH |
990 | for (; addr != end; addr += PAGE_SIZE) { |
991 | unsigned long offset; | |
992 | ||
993 | offset = (addr & ~PAGEMAP_WALK_MASK) >> | |
994 | PAGE_SHIFT; | |
0f8975ec | 995 | thp_pmd_to_pagemap_entry(&pme, pm, *pmd, offset, pmd_flags2); |
092b50ba | 996 | err = add_to_pagemap(addr, &pme, pm); |
025c5b24 NH |
997 | if (err) |
998 | break; | |
5aaabe83 | 999 | } |
5aaabe83 | 1000 | spin_unlock(&walk->mm->page_table_lock); |
025c5b24 | 1001 | return err; |
5aaabe83 NH |
1002 | } |
1003 | ||
45f83cef AA |
1004 | if (pmd_trans_unstable(pmd)) |
1005 | return 0; | |
85863e47 | 1006 | for (; addr != end; addr += PAGE_SIZE) { |
bcf8039e DH |
1007 | |
1008 | /* check to see if we've left 'vma' behind | |
1009 | * and need a new, higher one */ | |
16fbdce6 | 1010 | if (vma && (addr >= vma->vm_end)) { |
bcf8039e | 1011 | vma = find_vma(walk->mm, addr); |
2b0a9f01 | 1012 | pme = make_pme(PM_NOT_PRESENT(pm->v2)); |
16fbdce6 | 1013 | } |
bcf8039e DH |
1014 | |
1015 | /* check that 'vma' actually covers this address, | |
1016 | * and that it isn't a huge page vma */ | |
1017 | if (vma && (vma->vm_start <= addr) && | |
1018 | !is_vm_hugetlb_page(vma)) { | |
1019 | pte = pte_offset_map(pmd, addr); | |
2b0a9f01 | 1020 | pte_to_pagemap_entry(&pme, pm, vma, addr, *pte); |
bcf8039e DH |
1021 | /* unmap before userspace copy */ |
1022 | pte_unmap(pte); | |
1023 | } | |
092b50ba | 1024 | err = add_to_pagemap(addr, &pme, pm); |
85863e47 MM |
1025 | if (err) |
1026 | return err; | |
1027 | } | |
1028 | ||
1029 | cond_resched(); | |
1030 | ||
1031 | return err; | |
1032 | } | |
1033 | ||
1a5cb814 | 1034 | #ifdef CONFIG_HUGETLB_PAGE |
2b0a9f01 | 1035 | static void huge_pte_to_pagemap_entry(pagemap_entry_t *pme, struct pagemapread *pm, |
092b50ba | 1036 | pte_t pte, int offset) |
5dc37642 | 1037 | { |
5dc37642 | 1038 | if (pte_present(pte)) |
092b50ba | 1039 | *pme = make_pme(PM_PFRAME(pte_pfn(pte) + offset) |
2b0a9f01 | 1040 | | PM_STATUS2(pm->v2, 0) | PM_PRESENT); |
16fbdce6 | 1041 | else |
2b0a9f01 | 1042 | *pme = make_pme(PM_NOT_PRESENT(pm->v2)); |
5dc37642 NH |
1043 | } |
1044 | ||
116354d1 NH |
1045 | /* This function walks within one hugetlb entry in the single call */ |
1046 | static int pagemap_hugetlb_range(pte_t *pte, unsigned long hmask, | |
1047 | unsigned long addr, unsigned long end, | |
1048 | struct mm_walk *walk) | |
5dc37642 | 1049 | { |
5dc37642 | 1050 | struct pagemapread *pm = walk->private; |
5dc37642 | 1051 | int err = 0; |
16fbdce6 | 1052 | pagemap_entry_t pme; |
5dc37642 | 1053 | |
5dc37642 | 1054 | for (; addr != end; addr += PAGE_SIZE) { |
116354d1 | 1055 | int offset = (addr & ~hmask) >> PAGE_SHIFT; |
2b0a9f01 | 1056 | huge_pte_to_pagemap_entry(&pme, pm, *pte, offset); |
092b50ba | 1057 | err = add_to_pagemap(addr, &pme, pm); |
5dc37642 NH |
1058 | if (err) |
1059 | return err; | |
1060 | } | |
1061 | ||
1062 | cond_resched(); | |
1063 | ||
1064 | return err; | |
1065 | } | |
1a5cb814 | 1066 | #endif /* HUGETLB_PAGE */ |
5dc37642 | 1067 | |
85863e47 MM |
1068 | /* |
1069 | * /proc/pid/pagemap - an array mapping virtual pages to pfns | |
1070 | * | |
f16278c6 HR |
1071 | * For each page in the address space, this file contains one 64-bit entry |
1072 | * consisting of the following: | |
1073 | * | |
052fb0d6 | 1074 | * Bits 0-54 page frame number (PFN) if present |
f16278c6 | 1075 | * Bits 0-4 swap type if swapped |
052fb0d6 | 1076 | * Bits 5-54 swap offset if swapped |
f16278c6 | 1077 | * Bits 55-60 page shift (page size = 1<<page shift) |
052fb0d6 | 1078 | * Bit 61 page is file-page or shared-anon |
f16278c6 HR |
1079 | * Bit 62 page swapped |
1080 | * Bit 63 page present | |
1081 | * | |
1082 | * If the page is not present but in swap, then the PFN contains an | |
1083 | * encoding of the swap file number and the page's offset into the | |
1084 | * swap. Unmapped pages return a null PFN. This allows determining | |
85863e47 MM |
1085 | * precisely which pages are mapped (or in swap) and comparing mapped |
1086 | * pages between processes. | |
1087 | * | |
1088 | * Efficient users of this interface will use /proc/pid/maps to | |
1089 | * determine which areas of memory are actually mapped and llseek to | |
1090 | * skip over unmapped regions. | |
1091 | */ | |
1092 | static ssize_t pagemap_read(struct file *file, char __user *buf, | |
1093 | size_t count, loff_t *ppos) | |
1094 | { | |
496ad9aa | 1095 | struct task_struct *task = get_proc_task(file_inode(file)); |
85863e47 MM |
1096 | struct mm_struct *mm; |
1097 | struct pagemapread pm; | |
85863e47 | 1098 | int ret = -ESRCH; |
ee1e6ab6 | 1099 | struct mm_walk pagemap_walk = {}; |
5d7e0d2b AM |
1100 | unsigned long src; |
1101 | unsigned long svpfn; | |
1102 | unsigned long start_vaddr; | |
1103 | unsigned long end_vaddr; | |
d82ef020 | 1104 | int copied = 0; |
85863e47 MM |
1105 | |
1106 | if (!task) | |
1107 | goto out; | |
1108 | ||
85863e47 MM |
1109 | ret = -EINVAL; |
1110 | /* file position must be aligned */ | |
aae8679b | 1111 | if ((*ppos % PM_ENTRY_BYTES) || (count % PM_ENTRY_BYTES)) |
fb39380b | 1112 | goto out_task; |
85863e47 MM |
1113 | |
1114 | ret = 0; | |
08161786 VM |
1115 | if (!count) |
1116 | goto out_task; | |
1117 | ||
541c237c | 1118 | pm.v2 = soft_dirty_cleared; |
d82ef020 KH |
1119 | pm.len = PM_ENTRY_BYTES * (PAGEMAP_WALK_SIZE >> PAGE_SHIFT); |
1120 | pm.buffer = kmalloc(pm.len, GFP_TEMPORARY); | |
5d7e0d2b | 1121 | ret = -ENOMEM; |
d82ef020 | 1122 | if (!pm.buffer) |
98bc93e5 KM |
1123 | goto out_task; |
1124 | ||
e7dcd999 | 1125 | mm = mm_access(task, PTRACE_MODE_READ); |
98bc93e5 KM |
1126 | ret = PTR_ERR(mm); |
1127 | if (!mm || IS_ERR(mm)) | |
1128 | goto out_free; | |
85863e47 | 1129 | |
5d7e0d2b AM |
1130 | pagemap_walk.pmd_entry = pagemap_pte_range; |
1131 | pagemap_walk.pte_hole = pagemap_pte_hole; | |
1a5cb814 | 1132 | #ifdef CONFIG_HUGETLB_PAGE |
5dc37642 | 1133 | pagemap_walk.hugetlb_entry = pagemap_hugetlb_range; |
1a5cb814 | 1134 | #endif |
5d7e0d2b AM |
1135 | pagemap_walk.mm = mm; |
1136 | pagemap_walk.private = ± | |
1137 | ||
1138 | src = *ppos; | |
1139 | svpfn = src / PM_ENTRY_BYTES; | |
1140 | start_vaddr = svpfn << PAGE_SHIFT; | |
1141 | end_vaddr = TASK_SIZE_OF(task); | |
1142 | ||
1143 | /* watch out for wraparound */ | |
1144 | if (svpfn > TASK_SIZE_OF(task) >> PAGE_SHIFT) | |
1145 | start_vaddr = end_vaddr; | |
1146 | ||
1147 | /* | |
1148 | * The odds are that this will stop walking way | |
1149 | * before end_vaddr, because the length of the | |
1150 | * user buffer is tracked in "pm", and the walk | |
1151 | * will stop when we hit the end of the buffer. | |
1152 | */ | |
d82ef020 KH |
1153 | ret = 0; |
1154 | while (count && (start_vaddr < end_vaddr)) { | |
1155 | int len; | |
1156 | unsigned long end; | |
1157 | ||
1158 | pm.pos = 0; | |
ea251c1d | 1159 | end = (start_vaddr + PAGEMAP_WALK_SIZE) & PAGEMAP_WALK_MASK; |
d82ef020 KH |
1160 | /* overflow ? */ |
1161 | if (end < start_vaddr || end > end_vaddr) | |
1162 | end = end_vaddr; | |
1163 | down_read(&mm->mmap_sem); | |
1164 | ret = walk_page_range(start_vaddr, end, &pagemap_walk); | |
1165 | up_read(&mm->mmap_sem); | |
1166 | start_vaddr = end; | |
1167 | ||
1168 | len = min(count, PM_ENTRY_BYTES * pm.pos); | |
309361e0 | 1169 | if (copy_to_user(buf, pm.buffer, len)) { |
d82ef020 | 1170 | ret = -EFAULT; |
98bc93e5 | 1171 | goto out_mm; |
d82ef020 KH |
1172 | } |
1173 | copied += len; | |
1174 | buf += len; | |
1175 | count -= len; | |
85863e47 | 1176 | } |
d82ef020 KH |
1177 | *ppos += copied; |
1178 | if (!ret || ret == PM_END_OF_BUFFER) | |
1179 | ret = copied; | |
1180 | ||
fb39380b MT |
1181 | out_mm: |
1182 | mmput(mm); | |
98bc93e5 KM |
1183 | out_free: |
1184 | kfree(pm.buffer); | |
85863e47 MM |
1185 | out_task: |
1186 | put_task_struct(task); | |
1187 | out: | |
1188 | return ret; | |
1189 | } | |
1190 | ||
541c237c PE |
1191 | static int pagemap_open(struct inode *inode, struct file *file) |
1192 | { | |
1193 | pr_warn_once("Bits 55-60 of /proc/PID/pagemap entries are about " | |
1194 | "to stop being page-shift some time soon. See the " | |
1195 | "linux/Documentation/vm/pagemap.txt for details.\n"); | |
1196 | return 0; | |
1197 | } | |
1198 | ||
85863e47 MM |
1199 | const struct file_operations proc_pagemap_operations = { |
1200 | .llseek = mem_lseek, /* borrow this */ | |
1201 | .read = pagemap_read, | |
541c237c | 1202 | .open = pagemap_open, |
85863e47 | 1203 | }; |
1e883281 | 1204 | #endif /* CONFIG_PROC_PAGE_MONITOR */ |
85863e47 | 1205 | |
6e21c8f1 | 1206 | #ifdef CONFIG_NUMA |
6e21c8f1 | 1207 | |
f69ff943 SW |
1208 | struct numa_maps { |
1209 | struct vm_area_struct *vma; | |
1210 | unsigned long pages; | |
1211 | unsigned long anon; | |
1212 | unsigned long active; | |
1213 | unsigned long writeback; | |
1214 | unsigned long mapcount_max; | |
1215 | unsigned long dirty; | |
1216 | unsigned long swapcache; | |
1217 | unsigned long node[MAX_NUMNODES]; | |
1218 | }; | |
1219 | ||
5b52fc89 SW |
1220 | struct numa_maps_private { |
1221 | struct proc_maps_private proc_maps; | |
1222 | struct numa_maps md; | |
1223 | }; | |
1224 | ||
eb4866d0 DH |
1225 | static void gather_stats(struct page *page, struct numa_maps *md, int pte_dirty, |
1226 | unsigned long nr_pages) | |
f69ff943 SW |
1227 | { |
1228 | int count = page_mapcount(page); | |
1229 | ||
eb4866d0 | 1230 | md->pages += nr_pages; |
f69ff943 | 1231 | if (pte_dirty || PageDirty(page)) |
eb4866d0 | 1232 | md->dirty += nr_pages; |
f69ff943 SW |
1233 | |
1234 | if (PageSwapCache(page)) | |
eb4866d0 | 1235 | md->swapcache += nr_pages; |
f69ff943 SW |
1236 | |
1237 | if (PageActive(page) || PageUnevictable(page)) | |
eb4866d0 | 1238 | md->active += nr_pages; |
f69ff943 SW |
1239 | |
1240 | if (PageWriteback(page)) | |
eb4866d0 | 1241 | md->writeback += nr_pages; |
f69ff943 SW |
1242 | |
1243 | if (PageAnon(page)) | |
eb4866d0 | 1244 | md->anon += nr_pages; |
f69ff943 SW |
1245 | |
1246 | if (count > md->mapcount_max) | |
1247 | md->mapcount_max = count; | |
1248 | ||
eb4866d0 | 1249 | md->node[page_to_nid(page)] += nr_pages; |
f69ff943 SW |
1250 | } |
1251 | ||
3200a8aa DH |
1252 | static struct page *can_gather_numa_stats(pte_t pte, struct vm_area_struct *vma, |
1253 | unsigned long addr) | |
1254 | { | |
1255 | struct page *page; | |
1256 | int nid; | |
1257 | ||
1258 | if (!pte_present(pte)) | |
1259 | return NULL; | |
1260 | ||
1261 | page = vm_normal_page(vma, addr, pte); | |
1262 | if (!page) | |
1263 | return NULL; | |
1264 | ||
1265 | if (PageReserved(page)) | |
1266 | return NULL; | |
1267 | ||
1268 | nid = page_to_nid(page); | |
4ff1b2c2 | 1269 | if (!node_isset(nid, node_states[N_MEMORY])) |
3200a8aa DH |
1270 | return NULL; |
1271 | ||
1272 | return page; | |
1273 | } | |
1274 | ||
f69ff943 SW |
1275 | static int gather_pte_stats(pmd_t *pmd, unsigned long addr, |
1276 | unsigned long end, struct mm_walk *walk) | |
1277 | { | |
1278 | struct numa_maps *md; | |
1279 | spinlock_t *ptl; | |
1280 | pte_t *orig_pte; | |
1281 | pte_t *pte; | |
1282 | ||
1283 | md = walk->private; | |
025c5b24 NH |
1284 | |
1285 | if (pmd_trans_huge_lock(pmd, md->vma) == 1) { | |
1286 | pte_t huge_pte = *(pte_t *)pmd; | |
1287 | struct page *page; | |
1288 | ||
1289 | page = can_gather_numa_stats(huge_pte, md->vma, addr); | |
1290 | if (page) | |
1291 | gather_stats(page, md, pte_dirty(huge_pte), | |
1292 | HPAGE_PMD_SIZE/PAGE_SIZE); | |
32ef4384 | 1293 | spin_unlock(&walk->mm->page_table_lock); |
025c5b24 | 1294 | return 0; |
32ef4384 DH |
1295 | } |
1296 | ||
1a5a9906 AA |
1297 | if (pmd_trans_unstable(pmd)) |
1298 | return 0; | |
f69ff943 SW |
1299 | orig_pte = pte = pte_offset_map_lock(walk->mm, pmd, addr, &ptl); |
1300 | do { | |
3200a8aa | 1301 | struct page *page = can_gather_numa_stats(*pte, md->vma, addr); |
f69ff943 SW |
1302 | if (!page) |
1303 | continue; | |
eb4866d0 | 1304 | gather_stats(page, md, pte_dirty(*pte), 1); |
f69ff943 SW |
1305 | |
1306 | } while (pte++, addr += PAGE_SIZE, addr != end); | |
1307 | pte_unmap_unlock(orig_pte, ptl); | |
1308 | return 0; | |
1309 | } | |
1310 | #ifdef CONFIG_HUGETLB_PAGE | |
1311 | static int gather_hugetbl_stats(pte_t *pte, unsigned long hmask, | |
1312 | unsigned long addr, unsigned long end, struct mm_walk *walk) | |
1313 | { | |
1314 | struct numa_maps *md; | |
1315 | struct page *page; | |
1316 | ||
1317 | if (pte_none(*pte)) | |
1318 | return 0; | |
1319 | ||
1320 | page = pte_page(*pte); | |
1321 | if (!page) | |
1322 | return 0; | |
1323 | ||
1324 | md = walk->private; | |
eb4866d0 | 1325 | gather_stats(page, md, pte_dirty(*pte), 1); |
f69ff943 SW |
1326 | return 0; |
1327 | } | |
1328 | ||
1329 | #else | |
1330 | static int gather_hugetbl_stats(pte_t *pte, unsigned long hmask, | |
1331 | unsigned long addr, unsigned long end, struct mm_walk *walk) | |
1332 | { | |
1333 | return 0; | |
1334 | } | |
1335 | #endif | |
1336 | ||
1337 | /* | |
1338 | * Display pages allocated per node and memory policy via /proc. | |
1339 | */ | |
b7643757 | 1340 | static int show_numa_map(struct seq_file *m, void *v, int is_pid) |
f69ff943 | 1341 | { |
5b52fc89 SW |
1342 | struct numa_maps_private *numa_priv = m->private; |
1343 | struct proc_maps_private *proc_priv = &numa_priv->proc_maps; | |
f69ff943 | 1344 | struct vm_area_struct *vma = v; |
5b52fc89 | 1345 | struct numa_maps *md = &numa_priv->md; |
f69ff943 | 1346 | struct file *file = vma->vm_file; |
32f8516a | 1347 | struct task_struct *task = proc_priv->task; |
f69ff943 SW |
1348 | struct mm_struct *mm = vma->vm_mm; |
1349 | struct mm_walk walk = {}; | |
1350 | struct mempolicy *pol; | |
1351 | int n; | |
1352 | char buffer[50]; | |
1353 | ||
1354 | if (!mm) | |
1355 | return 0; | |
1356 | ||
5b52fc89 SW |
1357 | /* Ensure we start with an empty set of numa_maps statistics. */ |
1358 | memset(md, 0, sizeof(*md)); | |
f69ff943 SW |
1359 | |
1360 | md->vma = vma; | |
1361 | ||
1362 | walk.hugetlb_entry = gather_hugetbl_stats; | |
1363 | walk.pmd_entry = gather_pte_stats; | |
1364 | walk.private = md; | |
1365 | walk.mm = mm; | |
1366 | ||
32f8516a | 1367 | pol = get_vma_policy(task, vma, vma->vm_start); |
a7a88b23 | 1368 | mpol_to_str(buffer, sizeof(buffer), pol); |
f69ff943 SW |
1369 | mpol_cond_put(pol); |
1370 | ||
1371 | seq_printf(m, "%08lx %s", vma->vm_start, buffer); | |
1372 | ||
1373 | if (file) { | |
1374 | seq_printf(m, " file="); | |
1375 | seq_path(m, &file->f_path, "\n\t= "); | |
1376 | } else if (vma->vm_start <= mm->brk && vma->vm_end >= mm->start_brk) { | |
1377 | seq_printf(m, " heap"); | |
b7643757 | 1378 | } else { |
32f8516a | 1379 | pid_t tid = vm_is_stack(task, vma, is_pid); |
b7643757 SP |
1380 | if (tid != 0) { |
1381 | /* | |
1382 | * Thread stack in /proc/PID/task/TID/maps or | |
1383 | * the main process stack. | |
1384 | */ | |
1385 | if (!is_pid || (vma->vm_start <= mm->start_stack && | |
1386 | vma->vm_end >= mm->start_stack)) | |
1387 | seq_printf(m, " stack"); | |
1388 | else | |
1389 | seq_printf(m, " stack:%d", tid); | |
1390 | } | |
f69ff943 SW |
1391 | } |
1392 | ||
fc360bd9 AM |
1393 | if (is_vm_hugetlb_page(vma)) |
1394 | seq_printf(m, " huge"); | |
1395 | ||
f69ff943 SW |
1396 | walk_page_range(vma->vm_start, vma->vm_end, &walk); |
1397 | ||
1398 | if (!md->pages) | |
1399 | goto out; | |
1400 | ||
1401 | if (md->anon) | |
1402 | seq_printf(m, " anon=%lu", md->anon); | |
1403 | ||
1404 | if (md->dirty) | |
1405 | seq_printf(m, " dirty=%lu", md->dirty); | |
1406 | ||
1407 | if (md->pages != md->anon && md->pages != md->dirty) | |
1408 | seq_printf(m, " mapped=%lu", md->pages); | |
1409 | ||
1410 | if (md->mapcount_max > 1) | |
1411 | seq_printf(m, " mapmax=%lu", md->mapcount_max); | |
1412 | ||
1413 | if (md->swapcache) | |
1414 | seq_printf(m, " swapcache=%lu", md->swapcache); | |
1415 | ||
1416 | if (md->active < md->pages && !is_vm_hugetlb_page(vma)) | |
1417 | seq_printf(m, " active=%lu", md->active); | |
1418 | ||
1419 | if (md->writeback) | |
1420 | seq_printf(m, " writeback=%lu", md->writeback); | |
1421 | ||
4ff1b2c2 | 1422 | for_each_node_state(n, N_MEMORY) |
f69ff943 SW |
1423 | if (md->node[n]) |
1424 | seq_printf(m, " N%d=%lu", n, md->node[n]); | |
1425 | out: | |
1426 | seq_putc(m, '\n'); | |
f69ff943 SW |
1427 | |
1428 | if (m->count < m->size) | |
5b52fc89 | 1429 | m->version = (vma != proc_priv->tail_vma) ? vma->vm_start : 0; |
f69ff943 SW |
1430 | return 0; |
1431 | } | |
5b52fc89 | 1432 | |
b7643757 SP |
1433 | static int show_pid_numa_map(struct seq_file *m, void *v) |
1434 | { | |
1435 | return show_numa_map(m, v, 1); | |
1436 | } | |
1437 | ||
1438 | static int show_tid_numa_map(struct seq_file *m, void *v) | |
1439 | { | |
1440 | return show_numa_map(m, v, 0); | |
1441 | } | |
1442 | ||
03a44825 | 1443 | static const struct seq_operations proc_pid_numa_maps_op = { |
b7643757 SP |
1444 | .start = m_start, |
1445 | .next = m_next, | |
1446 | .stop = m_stop, | |
1447 | .show = show_pid_numa_map, | |
6e21c8f1 | 1448 | }; |
662795de | 1449 | |
b7643757 SP |
1450 | static const struct seq_operations proc_tid_numa_maps_op = { |
1451 | .start = m_start, | |
1452 | .next = m_next, | |
1453 | .stop = m_stop, | |
1454 | .show = show_tid_numa_map, | |
1455 | }; | |
1456 | ||
1457 | static int numa_maps_open(struct inode *inode, struct file *file, | |
1458 | const struct seq_operations *ops) | |
662795de | 1459 | { |
5b52fc89 SW |
1460 | struct numa_maps_private *priv; |
1461 | int ret = -ENOMEM; | |
1462 | priv = kzalloc(sizeof(*priv), GFP_KERNEL); | |
1463 | if (priv) { | |
1464 | priv->proc_maps.pid = proc_pid(inode); | |
b7643757 | 1465 | ret = seq_open(file, ops); |
5b52fc89 SW |
1466 | if (!ret) { |
1467 | struct seq_file *m = file->private_data; | |
1468 | m->private = priv; | |
1469 | } else { | |
1470 | kfree(priv); | |
1471 | } | |
1472 | } | |
1473 | return ret; | |
662795de EB |
1474 | } |
1475 | ||
b7643757 SP |
1476 | static int pid_numa_maps_open(struct inode *inode, struct file *file) |
1477 | { | |
1478 | return numa_maps_open(inode, file, &proc_pid_numa_maps_op); | |
1479 | } | |
1480 | ||
1481 | static int tid_numa_maps_open(struct inode *inode, struct file *file) | |
1482 | { | |
1483 | return numa_maps_open(inode, file, &proc_tid_numa_maps_op); | |
1484 | } | |
1485 | ||
1486 | const struct file_operations proc_pid_numa_maps_operations = { | |
1487 | .open = pid_numa_maps_open, | |
1488 | .read = seq_read, | |
1489 | .llseek = seq_lseek, | |
1490 | .release = seq_release_private, | |
1491 | }; | |
1492 | ||
1493 | const struct file_operations proc_tid_numa_maps_operations = { | |
1494 | .open = tid_numa_maps_open, | |
662795de EB |
1495 | .read = seq_read, |
1496 | .llseek = seq_lseek, | |
99f89551 | 1497 | .release = seq_release_private, |
662795de | 1498 | }; |
f69ff943 | 1499 | #endif /* CONFIG_NUMA */ |