Linux kernel & device driver programming

Cross-Referenced Linux and Device Driver Code

[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ]
Version: [ 2.6.11.8 ] [ 2.6.25 ] [ 2.6.25.8 ] [ 2.6.31.13 ] Architecture: [ i386 ]
  1 #include <linux/mm.h>
  2 #include <linux/hugetlb.h>
  3 #include <linux/mount.h>
  4 #include <linux/seq_file.h>
  5 #include <linux/highmem.h>
  6 #include <linux/ptrace.h>
  7 #include <linux/pagemap.h>
  8 #include <linux/ptrace.h>
  9 #include <linux/mempolicy.h>
 10 #include <linux/swap.h>
 11 #include <linux/swapops.h>
 12 #include <linux/seq_file.h>
 13 
 14 #include <asm/elf.h>
 15 #include <asm/uaccess.h>
 16 #include <asm/tlbflush.h>
 17 #include "internal.h"
 18 
 19 void task_mem(struct seq_file *m, struct mm_struct *mm)
 20 {
 21         unsigned long data, text, lib;
 22         unsigned long hiwater_vm, total_vm, hiwater_rss, total_rss;
 23 
 24         /*
 25          * Note: to minimize their overhead, mm maintains hiwater_vm and
 26          * hiwater_rss only when about to *lower* total_vm or rss.  Any
 27          * collector of these hiwater stats must therefore get total_vm
 28          * and rss too, which will usually be the higher.  Barriers? not
 29          * worth the effort, such snapshots can always be inconsistent.
 30          */
 31         hiwater_vm = total_vm = mm->total_vm;
 32         if (hiwater_vm < mm->hiwater_vm)
 33                 hiwater_vm = mm->hiwater_vm;
 34         hiwater_rss = total_rss = get_mm_rss(mm);
 35         if (hiwater_rss < mm->hiwater_rss)
 36                 hiwater_rss = mm->hiwater_rss;
 37 
 38         data = mm->total_vm - mm->shared_vm - mm->stack_vm;
 39         text = (PAGE_ALIGN(mm->end_code) - (mm->start_code & PAGE_MASK)) >> 10;
 40         lib = (mm->exec_vm << (PAGE_SHIFT-10)) - text;
 41         seq_printf(m,
 42                 "VmPeak:\t%8lu kB\n"
 43                 "VmSize:\t%8lu kB\n"
 44                 "VmLck:\t%8lu kB\n"
 45                 "VmHWM:\t%8lu kB\n"
 46                 "VmRSS:\t%8lu kB\n"
 47                 "VmData:\t%8lu kB\n"
 48                 "VmStk:\t%8lu kB\n"
 49                 "VmExe:\t%8lu kB\n"
 50                 "VmLib:\t%8lu kB\n"
 51                 "VmPTE:\t%8lu kB\n",
 52                 hiwater_vm << (PAGE_SHIFT-10),
 53                 (total_vm - mm->reserved_vm) << (PAGE_SHIFT-10),
 54                 mm->locked_vm << (PAGE_SHIFT-10),
 55                 hiwater_rss << (PAGE_SHIFT-10),
 56                 total_rss << (PAGE_SHIFT-10),
 57                 data << (PAGE_SHIFT-10),
 58                 mm->stack_vm << (PAGE_SHIFT-10), text, lib,
 59                 (PTRS_PER_PTE*sizeof(pte_t)*mm->nr_ptes) >> 10);
 60 }
 61 
 62 unsigned long task_vsize(struct mm_struct *mm)
 63 {
 64         return PAGE_SIZE * mm->total_vm;
 65 }
 66 
 67 int task_statm(struct mm_struct *mm, int *shared, int *text,
 68                int *data, int *resident)
 69 {
 70         *shared = get_mm_counter(mm, file_rss);
 71         *text = (PAGE_ALIGN(mm->end_code) - (mm->start_code & PAGE_MASK))
 72                                                                 >> PAGE_SHIFT;
 73         *data = mm->total_vm - mm->shared_vm;
 74         *resident = *shared + get_mm_counter(mm, anon_rss);
 75         return mm->total_vm;
 76 }
 77 
 78 int proc_exe_link(struct inode *inode, struct path *path)
 79 {
 80         struct vm_area_struct * vma;
 81         int result = -ENOENT;
 82         struct task_struct *task = get_proc_task(inode);
 83         struct mm_struct * mm = NULL;
 84 
 85         if (task) {
 86                 mm = get_task_mm(task);
 87                 put_task_struct(task);
 88         }
 89         if (!mm)
 90                 goto out;
 91         down_read(&mm->mmap_sem);
 92 
 93         vma = mm->mmap;
 94         while (vma) {
 95                 if ((vma->vm_flags & VM_EXECUTABLE) && vma->vm_file)
 96                         break;
 97                 vma = vma->vm_next;
 98         }
 99 
100         if (vma) {
101                 *path = vma->vm_file->f_path;
102                 path_get(&vma->vm_file->f_path);
103                 result = 0;
104         }
105 
106         up_read(&mm->mmap_sem);
107         mmput(mm);
108 out:
109         return result;
110 }
111 
112 static void pad_len_spaces(struct seq_file *m, int len)
113 {
114         len = 25 + sizeof(void*) * 6 - len;
115         if (len < 1)
116                 len = 1;
117         seq_printf(m, "%*c", len, ' ');
118 }
119 
120 static void vma_stop(struct proc_maps_private *priv, struct vm_area_struct *vma)
121 {
122         if (vma && vma != priv->tail_vma) {
123                 struct mm_struct *mm = vma->vm_mm;
124                 up_read(&mm->mmap_sem);
125                 mmput(mm);
126         }
127 }
128 
129 static void *m_start(struct seq_file *m, loff_t *pos)
130 {
131         struct proc_maps_private *priv = m->private;
132         unsigned long last_addr = m->version;
133         struct mm_struct *mm;
134         struct vm_area_struct *vma, *tail_vma = NULL;
135         loff_t l = *pos;
136 
137         /* Clear the per syscall fields in priv */
138         priv->task = NULL;
139         priv->tail_vma = NULL;
140 
141         /*
142          * We remember last_addr rather than next_addr to hit with
143          * mmap_cache most of the time. We have zero last_addr at
144          * the beginning and also after lseek. We will have -1 last_addr
145          * after the end of the vmas.
146          */
147 
148         if (last_addr == -1UL)
149                 return NULL;
150 
151         priv->task = get_pid_task(priv->pid, PIDTYPE_PID);
152         if (!priv->task)
153                 return NULL;
154 
155         mm = mm_for_maps(priv->task);
156         if (!mm)
157                 return NULL;
158 
159         tail_vma = get_gate_vma(priv->task);
160         priv->tail_vma = tail_vma;
161 
162         /* Start with last addr hint */
163         vma = find_vma(mm, last_addr);
164         if (last_addr && vma) {
165                 vma = vma->vm_next;
166                 goto out;
167         }
168 
169         /*
170          * Check the vma index is within the range and do
171          * sequential scan until m_index.
172          */
173         vma = NULL;
174         if ((unsigned long)l < mm->map_count) {
175                 vma = mm->mmap;
176                 while (l-- && vma) {
177                         vma = vma->vm_next;
178                         cond_resched();
179                 }
180                 goto out;
181         }
182 
183         if (l != mm->map_count)
184                 tail_vma = NULL; /* After gate vma */
185 
186 out:
187         if (vma)
188                 return vma;
189 
190         /* End of vmas has been reached */
191         m->version = (tail_vma != NULL)? 0: -1UL;
192         up_read(&mm->mmap_sem);
193         mmput(mm);
194         return tail_vma;
195 }
196 
197 static void *m_next(struct seq_file *m, void *v, loff_t *pos)
198 {
199         struct proc_maps_private *priv = m->private;
200         struct vm_area_struct *vma = v;
201         struct vm_area_struct *tail_vma = priv->tail_vma;
202 
203         (*pos)++;
204         if (vma && (vma != tail_vma) && vma->vm_next)
205                 return vma->vm_next;
206         vma_stop(priv, vma);
207         return (vma != tail_vma)? tail_vma: NULL;
208 }
209 
210 static void m_stop(struct seq_file *m, void *v)
211 {
212         struct proc_maps_private *priv = m->private;
213         struct vm_area_struct *vma = v;
214 
215         vma_stop(priv, vma);
216         if (priv->task)
217                 put_task_struct(priv->task);
218 }
219 
220 static int do_maps_open(struct inode *inode, struct file *file,
221                         const struct seq_operations *ops)
222 {
223         struct proc_maps_private *priv;
224         int ret = -ENOMEM;
225         priv = kzalloc(sizeof(*priv), GFP_KERNEL);
226         if (priv) {
227                 priv->pid = proc_pid(inode);
228                 ret = seq_open(file, ops);
229                 if (!ret) {
230                         struct seq_file *m = file->private_data;
231                         m->private = priv;
232                 } else {
233                         kfree(priv);
234                 }
235         }
236         return ret;
237 }
238 
239 static int show_map(struct seq_file *m, void *v)
240 {
241         struct proc_maps_private *priv = m->private;
242         struct task_struct *task = priv->task;
243         struct vm_area_struct *vma = v;
244         struct mm_struct *mm = vma->vm_mm;
245         struct file *file = vma->vm_file;
246         int flags = vma->vm_flags;
247         unsigned long ino = 0;
248         dev_t dev = 0;
249         int len;
250 
251         if (maps_protect && !ptrace_may_attach(task))
252                 return -EACCES;
253 
254         if (file) {
255                 struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
256                 dev = inode->i_sb->s_dev;
257                 ino = inode->i_ino;
258         }
259 
260         seq_printf(m, "%08lx-%08lx %c%c%c%c %08lx %02x:%02x %lu %n",
261                         vma->vm_start,
262                         vma->vm_end,
263                         flags & VM_READ ? 'r' : '-',
264                         flags & VM_WRITE ? 'w' : '-',
265                         flags & VM_EXEC ? 'x' : '-',
266                         flags & VM_MAYSHARE ? 's' : 'p',
267                         vma->vm_pgoff << PAGE_SHIFT,
268                         MAJOR(dev), MINOR(dev), ino, &len);
269 
270         /*
271          * Print the dentry name for named mappings, and a
272          * special [heap] marker for the heap:
273          */
274         if (file) {
275                 pad_len_spaces(m, len);
276                 seq_path(m, &file->f_path, "\n");
277         } else {
278                 const char *name = arch_vma_name(vma);
279                 if (!name) {
280                         if (mm) {
281                                 if (vma->vm_start <= mm->start_brk &&
282                                                 vma->vm_end >= mm->brk) {
283                                         name = "[heap]";
284                                 } else if (vma->vm_start <= mm->start_stack &&
285                                            vma->vm_end >= mm->start_stack) {
286                                         name = "[stack]";
287                                 }
288                         } else {
289                                 name = "[vdso]";
290                         }
291                 }
292                 if (name) {
293                         pad_len_spaces(m, len);
294                         seq_puts(m, name);
295                 }
296         }
297         seq_putc(m, '\n');
298 
299         if (m->count < m->size)  /* vma is copied successfully */
300                 m->version = (vma != get_gate_vma(task))? vma->vm_start: 0;
301         return 0;
302 }
303 
304 static const struct seq_operations proc_pid_maps_op = {
305         .start  = m_start,
306         .next   = m_next,
307         .stop   = m_stop,
308         .show   = show_map
309 };
310 
311 static int maps_open(struct inode *inode, struct file *file)
312 {
313         return do_maps_open(inode, file, &proc_pid_maps_op);
314 }
315 
316 const struct file_operations proc_maps_operations = {
317         .open           = maps_open,
318         .read           = seq_read,
319         .llseek         = seq_lseek,
320         .release        = seq_release_private,
321 };
322 
323 /*
324  * Proportional Set Size(PSS): my share of RSS.
325  *
326  * PSS of a process is the count of pages it has in memory, where each
327  * page is divided by the number of processes sharing it.  So if a
328  * process has 1000 pages all to itself, and 1000 shared with one other
329  * process, its PSS will be 1500.
330  *
331  * To keep (accumulated) division errors low, we adopt a 64bit
332  * fixed-point pss counter to minimize division errors. So (pss >>
333  * PSS_SHIFT) would be the real byte count.
334  *
335  * A shift of 12 before division means (assuming 4K page size):
336  *      - 1M 3-user-pages add up to 8KB errors;
337  *      - supports mapcount up to 2^24, or 16M;
338  *      - supports PSS up to 2^52 bytes, or 4PB.
339  */
340 #define PSS_SHIFT 12
341 
342 #ifdef CONFIG_PROC_PAGE_MONITOR
343 struct mem_size_stats
344 {
345         struct vm_area_struct *vma;
346         unsigned long resident;
347         unsigned long shared_clean;
348         unsigned long shared_dirty;
349         unsigned long private_clean;
350         unsigned long private_dirty;
351         unsigned long referenced;
352         u64 pss;
353 };
354 
355 static int smaps_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end,
356                            void *private)
357 {
358         struct mem_size_stats *mss = private;
359         struct vm_area_struct *vma = mss->vma;
360         pte_t *pte, ptent;
361         spinlock_t *ptl;
362         struct page *page;
363         int mapcount;
364 
365         pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
366         for (; addr != end; pte++, addr += PAGE_SIZE) {
367                 ptent = *pte;
368                 if (!pte_present(ptent))
369                         continue;
370 
371                 mss->resident += PAGE_SIZE;
372 
373                 page = vm_normal_page(vma, addr, ptent);
374                 if (!page)
375                         continue;
376 
377                 /* Accumulate the size in pages that have been accessed. */
378                 if (pte_young(ptent) || PageReferenced(page))
379                         mss->referenced += PAGE_SIZE;
380                 mapcount = page_mapcount(page);
381                 if (mapcount >= 2) {
382                         if (pte_dirty(ptent))
383                                 mss->shared_dirty += PAGE_SIZE;
384                         else
385                                 mss->shared_clean += PAGE_SIZE;
386                         mss->pss += (PAGE_SIZE << PSS_SHIFT) / mapcount;
387                 } else {
388                         if (pte_dirty(ptent))
389                                 mss->private_dirty += PAGE_SIZE;
390                         else
391                                 mss->private_clean += PAGE_SIZE;
392                         mss->pss += (PAGE_SIZE << PSS_SHIFT);
393                 }
394         }
395         pte_unmap_unlock(pte - 1, ptl);
396         cond_resched();
397         return 0;
398 }
399 
400 static struct mm_walk smaps_walk = { .pmd_entry = smaps_pte_range };
401 
402 static int show_smap(struct seq_file *m, void *v)
403 {
404         struct vm_area_struct *vma = v;
405         struct mem_size_stats mss;
406         int ret;
407 
408         memset(&mss, 0, sizeof mss);
409         mss.vma = vma;
410         if (vma->vm_mm && !is_vm_hugetlb_page(vma))
411                 walk_page_range(vma->vm_mm, vma->vm_start, vma->vm_end,
412                                 &smaps_walk, &mss);
413 
414         ret = show_map(m, v);
415         if (ret)
416                 return ret;
417 
418         seq_printf(m,
419                    "Size:           %8lu kB\n"
420                    "Rss:            %8lu kB\n"
421                    "Pss:            %8lu kB\n"
422                    "Shared_Clean:   %8lu kB\n"
423                    "Shared_Dirty:   %8lu kB\n"
424                    "Private_Clean:  %8lu kB\n"
425                    "Private_Dirty:  %8lu kB\n"
426                    "Referenced:     %8lu kB\n",
427                    (vma->vm_end - vma->vm_start) >> 10,
428                    mss.resident >> 10,
429                    (unsigned long)(mss.pss >> (10 + PSS_SHIFT)),
430                    mss.shared_clean  >> 10,
431                    mss.shared_dirty  >> 10,
432                    mss.private_clean >> 10,
433                    mss.private_dirty >> 10,
434                    mss.referenced >> 10);
435 
436         return ret;
437 }
438 
439 static const struct seq_operations proc_pid_smaps_op = {
440         .start  = m_start,
441         .next   = m_next,
442         .stop   = m_stop,
443         .show   = show_smap
444 };
445 
446 static int smaps_open(struct inode *inode, struct file *file)
447 {
448         return do_maps_open(inode, file, &proc_pid_smaps_op);
449 }
450 
451 const struct file_operations proc_smaps_operations = {
452         .open           = smaps_open,
453         .read           = seq_read,
454         .llseek         = seq_lseek,
455         .release        = seq_release_private,
456 };
457 
458 static int clear_refs_pte_range(pmd_t *pmd, unsigned long addr,
459                                 unsigned long end, void *private)
460 {
461         struct vm_area_struct *vma = private;
462         pte_t *pte, ptent;
463         spinlock_t *ptl;
464         struct page *page;
465 
466         pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
467         for (; addr != end; pte++, addr += PAGE_SIZE) {
468                 ptent = *pte;
469                 if (!pte_present(ptent))
470                         continue;
471 
472                 page = vm_normal_page(vma, addr, ptent);
473                 if (!page)
474                         continue;
475 
476                 /* Clear accessed and referenced bits. */
477                 ptep_test_and_clear_young(vma, addr, pte);
478                 ClearPageReferenced(page);
479         }
480         pte_unmap_unlock(pte - 1, ptl);
481         cond_resched();
482         return 0;
483 }
484 
485 static struct mm_walk clear_refs_walk = { .pmd_entry = clear_refs_pte_range };
486 
487 static ssize_t clear_refs_write(struct file *file, const char __user *buf,
488                                 size_t count, loff_t *ppos)
489 {
490         struct task_struct *task;
491         char buffer[PROC_NUMBUF], *end;
492         struct mm_struct *mm;
493         struct vm_area_struct *vma;
494 
495         memset(buffer, 0, sizeof(buffer));
496         if (count > sizeof(buffer) - 1)
497                 count = sizeof(buffer) - 1;
498         if (copy_from_user(buffer, buf, count))
499                 return -EFAULT;
500         if (!simple_strtol(buffer, &end, 0))
501                 return -EINVAL;
502         if (*end == '\n')
503                 end++;
504         task = get_proc_task(file->f_path.dentry->d_inode);
505         if (!task)
506                 return -ESRCH;
507         mm = get_task_mm(task);
508         if (mm) {
509                 down_read(&mm->mmap_sem);
510                 for (vma = mm->mmap; vma; vma = vma->vm_next)
511                         if (!is_vm_hugetlb_page(vma))
512                                 walk_page_range(mm, vma->vm_start, vma->vm_end,
513                                                 &clear_refs_walk, vma);
514                 flush_tlb_mm(mm);
515                 up_read(&mm->mmap_sem);
516                 mmput(mm);
517         }
518         put_task_struct(task);
519         if (end - buffer == 0)
520                 return -EIO;
521         return end - buffer;
522 }
523 
524 const struct file_operations proc_clear_refs_operations = {
525         .write          = clear_refs_write,
526 };
527 
528 struct pagemapread {
529         u64 __user *out, *end;
530 };
531 
532 #define PM_ENTRY_BYTES      sizeof(u64)
533 #define PM_STATUS_BITS      3
534 #define PM_STATUS_OFFSET    (64 - PM_STATUS_BITS)
535 #define PM_STATUS_MASK      (((1LL << PM_STATUS_BITS) - 1) << PM_STATUS_OFFSET)
536 #define PM_STATUS(nr)       (((nr) << PM_STATUS_OFFSET) & PM_STATUS_MASK)
537 #define PM_PSHIFT_BITS      6
538 #define PM_PSHIFT_OFFSET    (PM_STATUS_OFFSET - PM_PSHIFT_BITS)
539 #define PM_PSHIFT_MASK      (((1LL << PM_PSHIFT_BITS) - 1) << PM_PSHIFT_OFFSET)
540 #define PM_PSHIFT(x)        (((u64) (x) << PM_PSHIFT_OFFSET) & PM_PSHIFT_MASK)
541 #define PM_PFRAME_MASK      ((1LL << PM_PSHIFT_OFFSET) - 1)
542 #define PM_PFRAME(x)        ((x) & PM_PFRAME_MASK)
543 
544 #define PM_PRESENT          PM_STATUS(4LL)
545 #define PM_SWAP             PM_STATUS(2LL)
546 #define PM_NOT_PRESENT      PM_PSHIFT(PAGE_SHIFT)
547 #define PM_END_OF_BUFFER    1
548 
549 static int add_to_pagemap(unsigned long addr, u64 pfn,
550                           struct pagemapread *pm)
551 {
552         if (put_user(pfn, pm->out))
553                 return -EFAULT;
554         pm->out++;
555         if (pm->out >= pm->end)
556                 return PM_END_OF_BUFFER;
557         return 0;
558 }
559 
560 static int pagemap_pte_hole(unsigned long start, unsigned long end,
561                                 void *private)
562 {
563         struct pagemapread *pm = private;
564         unsigned long addr;
565         int err = 0;
566         for (addr = start; addr < end; addr += PAGE_SIZE) {
567                 err = add_to_pagemap(addr, PM_NOT_PRESENT, pm);
568                 if (err)
569                         break;
570         }
571         return err;
572 }
573 
574 u64 swap_pte_to_pagemap_entry(pte_t pte)
575 {
576         swp_entry_t e = pte_to_swp_entry(pte);
577         return swp_type(e) | (swp_offset(e) << MAX_SWAPFILES_SHIFT);
578 }
579 
580 static int pagemap_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end,
581                              void *private)
582 {
583         struct pagemapread *pm = private;
584         pte_t *pte;
585         int err = 0;
586 
587         for (; addr != end; addr += PAGE_SIZE) {
588                 u64 pfn = PM_NOT_PRESENT;
589                 pte = pte_offset_map(pmd, addr);
590                 if (is_swap_pte(*pte))
591                         pfn = PM_PFRAME(swap_pte_to_pagemap_entry(*pte))
592                                 | PM_PSHIFT(PAGE_SHIFT) | PM_SWAP;
593                 else if (pte_present(*pte))
594                         pfn = PM_PFRAME(pte_pfn(*pte))
595                                 | PM_PSHIFT(PAGE_SHIFT) | PM_PRESENT;
596                 /* unmap so we're not in atomic when we copy to userspace */
597                 pte_unmap(pte);
598                 err = add_to_pagemap(addr, pfn, pm);
599                 if (err)
600                         return err;
601         }
602 
603         cond_resched();
604 
605         return err;
606 }
607 
608 static struct mm_walk pagemap_walk = {
609         .pmd_entry = pagemap_pte_range,
610         .pte_hole = pagemap_pte_hole
611 };
612 
613 /*
614  * /proc/pid/pagemap - an array mapping virtual pages to pfns
615  *
616  * For each page in the address space, this file contains one 64-bit entry
617  * consisting of the following:
618  *
619  * Bits 0-55  page frame number (PFN) if present
620  * Bits 0-4   swap type if swapped
621  * Bits 5-55  swap offset if swapped
622  * Bits 55-60 page shift (page size = 1<<page shift)
623  * Bit  61    reserved for future use
624  * Bit  62    page swapped
625  * Bit  63    page present
626  *
627  * If the page is not present but in swap, then the PFN contains an
628  * encoding of the swap file number and the page's offset into the
629  * swap. Unmapped pages return a null PFN. This allows determining
630  * precisely which pages are mapped (or in swap) and comparing mapped
631  * pages between processes.
632  *
633  * Efficient users of this interface will use /proc/pid/maps to
634  * determine which areas of memory are actually mapped and llseek to
635  * skip over unmapped regions.
636  */
637 static ssize_t pagemap_read(struct file *file, char __user *buf,
638                             size_t count, loff_t *ppos)
639 {
640         struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
641         struct page **pages, *page;
642         unsigned long uaddr, uend;
643         struct mm_struct *mm;
644         struct pagemapread pm;
645         int pagecount;
646         int ret = -ESRCH;
647 
648         if (!task)
649                 goto out;
650 
651         ret = -EACCES;
652         if (!ptrace_may_attach(task))
653                 goto out_task;
654 
655         ret = -EINVAL;
656         /* file position must be aligned */
657         if ((*ppos % PM_ENTRY_BYTES) || (count % PM_ENTRY_BYTES))
658                 goto out_task;
659 
660         ret = 0;
661         mm = get_task_mm(task);
662         if (!mm)
663                 goto out_task;
664 
665         ret = -ENOMEM;
666         uaddr = (unsigned long)buf & PAGE_MASK;
667         uend = (unsigned long)(buf + count);
668         pagecount = (PAGE_ALIGN(uend) - uaddr) / PAGE_SIZE;
669         pages = kmalloc(pagecount * sizeof(struct page *), GFP_KERNEL);
670         if (!pages)
671                 goto out_mm;
672 
673         down_read(&current->mm->mmap_sem);
674         ret = get_user_pages(current, current->mm, uaddr, pagecount,
675                              1, 0, pages, NULL);
676         up_read(&current->mm->mmap_sem);
677 
678         if (ret < 0)
679                 goto out_free;
680 
681         if (ret != pagecount) {
682                 pagecount = ret;
683                 ret = -EFAULT;
684                 goto out_pages;
685         }
686 
687         pm.out = (u64 *)buf;
688         pm.end = (u64 *)(buf + count);
689 
690         if (!ptrace_may_attach(task)) {
691                 ret = -EIO;
692         } else {
693                 unsigned long src = *ppos;
694                 unsigned long svpfn = src / PM_ENTRY_BYTES;
695                 unsigned long start_vaddr = svpfn << PAGE_SHIFT;
696                 unsigned long end_vaddr = TASK_SIZE_OF(task);
697 
698                 /* watch out for wraparound */
699                 if (svpfn > TASK_SIZE_OF(task) >> PAGE_SHIFT)
700                         start_vaddr = end_vaddr;
701 
702                 /*
703                  * The odds are that this will stop walking way
704                  * before end_vaddr, because the length of the
705                  * user buffer is tracked in "pm", and the walk
706                  * will stop when we hit the end of the buffer.
707                  */
708                 ret = walk_page_range(mm, start_vaddr, end_vaddr,
709                                         &pagemap_walk, &pm);
710                 if (ret == PM_END_OF_BUFFER)
711                         ret = 0;
712                 /* don't need mmap_sem for these, but this looks cleaner */
713                 *ppos += (char *)pm.out - buf;
714                 if (!ret)
715                         ret = (char *)pm.out - buf;
716         }
717 
718 out_pages:
719         for (; pagecount; pagecount--) {
720                 page = pages[pagecount-1];
721                 if (!PageReserved(page))
722                         SetPageDirty(page);
723                 page_cache_release(page);
724         }
725 out_free:
726         kfree(pages);
727 out_mm:
728         mmput(mm);
729 out_task:
730         put_task_struct(task);
731 out:
732         return ret;
733 }
734 
735 const struct file_operations proc_pagemap_operations = {
736         .llseek         = mem_lseek, /* borrow this */
737         .read           = pagemap_read,
738 };
739 #endif /* CONFIG_PROC_PAGE_MONITOR */
740 
741 #ifdef CONFIG_NUMA
742 extern int show_numa_map(struct seq_file *m, void *v);
743 
744 static int show_numa_map_checked(struct seq_file *m, void *v)
745 {
746         struct proc_maps_private *priv = m->private;
747         struct task_struct *task = priv->task;
748 
749         if (maps_protect && !ptrace_may_attach(task))
750                 return -EACCES;
751 
752         return show_numa_map(m, v);
753 }
754 
755 static const struct seq_operations proc_pid_numa_maps_op = {
756         .start  = m_start,
757         .next   = m_next,
758         .stop   = m_stop,
759         .show   = show_numa_map_checked
760 };
761 
762 static int numa_maps_open(struct inode *inode, struct file *file)
763 {
764         return do_maps_open(inode, file, &proc_pid_numa_maps_op);
765 }
766 
767 const struct file_operations proc_numa_maps_operations = {
768         .open           = numa_maps_open,
769         .read           = seq_read,
770         .llseek         = seq_lseek,
771         .release        = seq_release_private,
772 };
773 #endif
774 
  This page was automatically generated by the LXR engine.