diff options
Diffstat (limited to 'fs/proc')
| -rw-r--r-- | fs/proc/array.c | 85 | ||||
| -rw-r--r-- | fs/proc/base.c | 67 | ||||
| -rw-r--r-- | fs/proc/kcore.c | 335 | ||||
| -rw-r--r-- | fs/proc/meminfo.c | 4 | ||||
| -rw-r--r-- | fs/proc/nommu.c | 2 | ||||
| -rw-r--r-- | fs/proc/page.c | 5 | ||||
| -rw-r--r-- | fs/proc/task_mmu.c | 57 |
7 files changed, 459 insertions, 96 deletions
diff --git a/fs/proc/array.c b/fs/proc/array.c index 725a650bbbb8..0c6bc602e6c4 100644 --- a/fs/proc/array.c +++ b/fs/proc/array.c | |||
| @@ -82,6 +82,7 @@ | |||
| 82 | #include <linux/pid_namespace.h> | 82 | #include <linux/pid_namespace.h> |
| 83 | #include <linux/ptrace.h> | 83 | #include <linux/ptrace.h> |
| 84 | #include <linux/tracehook.h> | 84 | #include <linux/tracehook.h> |
| 85 | #include <linux/swapops.h> | ||
| 85 | 86 | ||
| 86 | #include <asm/pgtable.h> | 87 | #include <asm/pgtable.h> |
| 87 | #include <asm/processor.h> | 88 | #include <asm/processor.h> |
| @@ -321,6 +322,87 @@ static inline void task_context_switch_counts(struct seq_file *m, | |||
| 321 | p->nivcsw); | 322 | p->nivcsw); |
| 322 | } | 323 | } |
| 323 | 324 | ||
| 325 | struct stack_stats { | ||
| 326 | struct vm_area_struct *vma; | ||
| 327 | unsigned long startpage; | ||
| 328 | unsigned long usage; | ||
| 329 | }; | ||
| 330 | |||
| 331 | static int stack_usage_pte_range(pmd_t *pmd, unsigned long addr, | ||
| 332 | unsigned long end, struct mm_walk *walk) | ||
| 333 | { | ||
| 334 | struct stack_stats *ss = walk->private; | ||
| 335 | struct vm_area_struct *vma = ss->vma; | ||
| 336 | pte_t *pte, ptent; | ||
| 337 | spinlock_t *ptl; | ||
| 338 | int ret = 0; | ||
| 339 | |||
| 340 | pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl); | ||
| 341 | for (; addr != end; pte++, addr += PAGE_SIZE) { | ||
| 342 | ptent = *pte; | ||
| 343 | |||
| 344 | #ifdef CONFIG_STACK_GROWSUP | ||
| 345 | if (pte_present(ptent) || is_swap_pte(ptent)) | ||
| 346 | ss->usage = addr - ss->startpage + PAGE_SIZE; | ||
| 347 | #else | ||
| 348 | if (pte_present(ptent) || is_swap_pte(ptent)) { | ||
| 349 | ss->usage = ss->startpage - addr + PAGE_SIZE; | ||
| 350 | pte++; | ||
| 351 | ret = 1; | ||
| 352 | break; | ||
| 353 | } | ||
| 354 | #endif | ||
| 355 | } | ||
| 356 | pte_unmap_unlock(pte - 1, ptl); | ||
| 357 | cond_resched(); | ||
| 358 | return ret; | ||
| 359 | } | ||
| 360 | |||
| 361 | static inline unsigned long get_stack_usage_in_bytes(struct vm_area_struct *vma, | ||
| 362 | struct task_struct *task) | ||
| 363 | { | ||
| 364 | struct stack_stats ss; | ||
| 365 | struct mm_walk stack_walk = { | ||
| 366 | .pmd_entry = stack_usage_pte_range, | ||
| 367 | .mm = vma->vm_mm, | ||
| 368 | .private = &ss, | ||
| 369 | }; | ||
| 370 | |||
| 371 | if (!vma->vm_mm || is_vm_hugetlb_page(vma)) | ||
| 372 | return 0; | ||
| 373 | |||
| 374 | ss.vma = vma; | ||
| 375 | ss.startpage = task->stack_start & PAGE_MASK; | ||
| 376 | ss.usage = 0; | ||
| 377 | |||
| 378 | #ifdef CONFIG_STACK_GROWSUP | ||
| 379 | walk_page_range(KSTK_ESP(task) & PAGE_MASK, vma->vm_end, | ||
| 380 | &stack_walk); | ||
| 381 | #else | ||
| 382 | walk_page_range(vma->vm_start, (KSTK_ESP(task) & PAGE_MASK) + PAGE_SIZE, | ||
| 383 | &stack_walk); | ||
| 384 | #endif | ||
| 385 | return ss.usage; | ||
| 386 | } | ||
| 387 | |||
| 388 | static inline void task_show_stack_usage(struct seq_file *m, | ||
| 389 | struct task_struct *task) | ||
| 390 | { | ||
| 391 | struct vm_area_struct *vma; | ||
| 392 | struct mm_struct *mm = get_task_mm(task); | ||
| 393 | |||
| 394 | if (mm) { | ||
| 395 | down_read(&mm->mmap_sem); | ||
| 396 | vma = find_vma(mm, task->stack_start); | ||
| 397 | if (vma) | ||
| 398 | seq_printf(m, "Stack usage:\t%lu kB\n", | ||
| 399 | get_stack_usage_in_bytes(vma, task) >> 10); | ||
| 400 | |||
| 401 | up_read(&mm->mmap_sem); | ||
| 402 | mmput(mm); | ||
| 403 | } | ||
| 404 | } | ||
| 405 | |||
| 324 | int proc_pid_status(struct seq_file *m, struct pid_namespace *ns, | 406 | int proc_pid_status(struct seq_file *m, struct pid_namespace *ns, |
| 325 | struct pid *pid, struct task_struct *task) | 407 | struct pid *pid, struct task_struct *task) |
| 326 | { | 408 | { |
| @@ -340,6 +422,7 @@ int proc_pid_status(struct seq_file *m, struct pid_namespace *ns, | |||
| 340 | task_show_regs(m, task); | 422 | task_show_regs(m, task); |
| 341 | #endif | 423 | #endif |
| 342 | task_context_switch_counts(m, task); | 424 | task_context_switch_counts(m, task); |
| 425 | task_show_stack_usage(m, task); | ||
| 343 | return 0; | 426 | return 0; |
| 344 | } | 427 | } |
| 345 | 428 | ||
| @@ -481,7 +564,7 @@ static int do_task_stat(struct seq_file *m, struct pid_namespace *ns, | |||
| 481 | rsslim, | 564 | rsslim, |
| 482 | mm ? mm->start_code : 0, | 565 | mm ? mm->start_code : 0, |
| 483 | mm ? mm->end_code : 0, | 566 | mm ? mm->end_code : 0, |
| 484 | (permitted && mm) ? mm->start_stack : 0, | 567 | (permitted) ? task->stack_start : 0, |
| 485 | esp, | 568 | esp, |
| 486 | eip, | 569 | eip, |
| 487 | /* The signal information here is obsolete. | 570 | /* The signal information here is obsolete. |
diff --git a/fs/proc/base.c b/fs/proc/base.c index 6f742f6658a9..837469a96598 100644 --- a/fs/proc/base.c +++ b/fs/proc/base.c | |||
| @@ -447,7 +447,7 @@ static int proc_oom_score(struct task_struct *task, char *buffer) | |||
| 447 | 447 | ||
| 448 | do_posix_clock_monotonic_gettime(&uptime); | 448 | do_posix_clock_monotonic_gettime(&uptime); |
| 449 | read_lock(&tasklist_lock); | 449 | read_lock(&tasklist_lock); |
| 450 | points = badness(task, uptime.tv_sec); | 450 | points = badness(task->group_leader, uptime.tv_sec); |
| 451 | read_unlock(&tasklist_lock); | 451 | read_unlock(&tasklist_lock); |
| 452 | return sprintf(buffer, "%lu\n", points); | 452 | return sprintf(buffer, "%lu\n", points); |
| 453 | } | 453 | } |
| @@ -458,7 +458,7 @@ struct limit_names { | |||
| 458 | }; | 458 | }; |
| 459 | 459 | ||
| 460 | static const struct limit_names lnames[RLIM_NLIMITS] = { | 460 | static const struct limit_names lnames[RLIM_NLIMITS] = { |
| 461 | [RLIMIT_CPU] = {"Max cpu time", "ms"}, | 461 | [RLIMIT_CPU] = {"Max cpu time", "seconds"}, |
| 462 | [RLIMIT_FSIZE] = {"Max file size", "bytes"}, | 462 | [RLIMIT_FSIZE] = {"Max file size", "bytes"}, |
| 463 | [RLIMIT_DATA] = {"Max data size", "bytes"}, | 463 | [RLIMIT_DATA] = {"Max data size", "bytes"}, |
| 464 | [RLIMIT_STACK] = {"Max stack size", "bytes"}, | 464 | [RLIMIT_STACK] = {"Max stack size", "bytes"}, |
| @@ -999,11 +999,17 @@ static ssize_t oom_adjust_read(struct file *file, char __user *buf, | |||
| 999 | struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode); | 999 | struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode); |
| 1000 | char buffer[PROC_NUMBUF]; | 1000 | char buffer[PROC_NUMBUF]; |
| 1001 | size_t len; | 1001 | size_t len; |
| 1002 | int oom_adjust; | 1002 | int oom_adjust = OOM_DISABLE; |
| 1003 | unsigned long flags; | ||
| 1003 | 1004 | ||
| 1004 | if (!task) | 1005 | if (!task) |
| 1005 | return -ESRCH; | 1006 | return -ESRCH; |
| 1006 | oom_adjust = task->oomkilladj; | 1007 | |
| 1008 | if (lock_task_sighand(task, &flags)) { | ||
| 1009 | oom_adjust = task->signal->oom_adj; | ||
| 1010 | unlock_task_sighand(task, &flags); | ||
| 1011 | } | ||
| 1012 | |||
| 1007 | put_task_struct(task); | 1013 | put_task_struct(task); |
| 1008 | 1014 | ||
| 1009 | len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust); | 1015 | len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust); |
| @@ -1015,32 +1021,44 @@ static ssize_t oom_adjust_write(struct file *file, const char __user *buf, | |||
| 1015 | size_t count, loff_t *ppos) | 1021 | size_t count, loff_t *ppos) |
| 1016 | { | 1022 | { |
| 1017 | struct task_struct *task; | 1023 | struct task_struct *task; |
| 1018 | char buffer[PROC_NUMBUF], *end; | 1024 | char buffer[PROC_NUMBUF]; |
| 1019 | int oom_adjust; | 1025 | long oom_adjust; |
| 1026 | unsigned long flags; | ||
| 1027 | int err; | ||
| 1020 | 1028 | ||
| 1021 | memset(buffer, 0, sizeof(buffer)); | 1029 | memset(buffer, 0, sizeof(buffer)); |
| 1022 | if (count > sizeof(buffer) - 1) | 1030 | if (count > sizeof(buffer) - 1) |
| 1023 | count = sizeof(buffer) - 1; | 1031 | count = sizeof(buffer) - 1; |
| 1024 | if (copy_from_user(buffer, buf, count)) | 1032 | if (copy_from_user(buffer, buf, count)) |
| 1025 | return -EFAULT; | 1033 | return -EFAULT; |
| 1026 | oom_adjust = simple_strtol(buffer, &end, 0); | 1034 | |
| 1035 | err = strict_strtol(strstrip(buffer), 0, &oom_adjust); | ||
| 1036 | if (err) | ||
| 1037 | return -EINVAL; | ||
| 1027 | if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) && | 1038 | if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) && |
| 1028 | oom_adjust != OOM_DISABLE) | 1039 | oom_adjust != OOM_DISABLE) |
| 1029 | return -EINVAL; | 1040 | return -EINVAL; |
| 1030 | if (*end == '\n') | 1041 | |
| 1031 | end++; | ||
| 1032 | task = get_proc_task(file->f_path.dentry->d_inode); | 1042 | task = get_proc_task(file->f_path.dentry->d_inode); |
| 1033 | if (!task) | 1043 | if (!task) |
| 1034 | return -ESRCH; | 1044 | return -ESRCH; |
| 1035 | if (oom_adjust < task->oomkilladj && !capable(CAP_SYS_RESOURCE)) { | 1045 | if (!lock_task_sighand(task, &flags)) { |
| 1046 | put_task_struct(task); | ||
| 1047 | return -ESRCH; | ||
| 1048 | } | ||
| 1049 | |||
| 1050 | if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) { | ||
| 1051 | unlock_task_sighand(task, &flags); | ||
| 1036 | put_task_struct(task); | 1052 | put_task_struct(task); |
| 1037 | return -EACCES; | 1053 | return -EACCES; |
| 1038 | } | 1054 | } |
| 1039 | task->oomkilladj = oom_adjust; | 1055 | |
| 1056 | task->signal->oom_adj = oom_adjust; | ||
| 1057 | |||
| 1058 | unlock_task_sighand(task, &flags); | ||
| 1040 | put_task_struct(task); | 1059 | put_task_struct(task); |
| 1041 | if (end - buffer == 0) | 1060 | |
| 1042 | return -EIO; | 1061 | return count; |
| 1043 | return end - buffer; | ||
| 1044 | } | 1062 | } |
| 1045 | 1063 | ||
| 1046 | static const struct file_operations proc_oom_adjust_operations = { | 1064 | static const struct file_operations proc_oom_adjust_operations = { |
| @@ -1169,17 +1187,16 @@ static ssize_t proc_fault_inject_write(struct file * file, | |||
| 1169 | count = sizeof(buffer) - 1; | 1187 | count = sizeof(buffer) - 1; |
| 1170 | if (copy_from_user(buffer, buf, count)) | 1188 | if (copy_from_user(buffer, buf, count)) |
| 1171 | return -EFAULT; | 1189 | return -EFAULT; |
| 1172 | make_it_fail = simple_strtol(buffer, &end, 0); | 1190 | make_it_fail = simple_strtol(strstrip(buffer), &end, 0); |
| 1173 | if (*end == '\n') | 1191 | if (*end) |
| 1174 | end++; | 1192 | return -EINVAL; |
| 1175 | task = get_proc_task(file->f_dentry->d_inode); | 1193 | task = get_proc_task(file->f_dentry->d_inode); |
| 1176 | if (!task) | 1194 | if (!task) |
| 1177 | return -ESRCH; | 1195 | return -ESRCH; |
| 1178 | task->make_it_fail = make_it_fail; | 1196 | task->make_it_fail = make_it_fail; |
| 1179 | put_task_struct(task); | 1197 | put_task_struct(task); |
| 1180 | if (end - buffer == 0) | 1198 | |
| 1181 | return -EIO; | 1199 | return count; |
| 1182 | return end - buffer; | ||
| 1183 | } | 1200 | } |
| 1184 | 1201 | ||
| 1185 | static const struct file_operations proc_fault_inject_operations = { | 1202 | static const struct file_operations proc_fault_inject_operations = { |
| @@ -2586,9 +2603,6 @@ static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid) | |||
| 2586 | dput(dentry); | 2603 | dput(dentry); |
| 2587 | } | 2604 | } |
| 2588 | 2605 | ||
| 2589 | if (tgid == 0) | ||
| 2590 | goto out; | ||
| 2591 | |||
| 2592 | name.name = buf; | 2606 | name.name = buf; |
| 2593 | name.len = snprintf(buf, sizeof(buf), "%d", tgid); | 2607 | name.len = snprintf(buf, sizeof(buf), "%d", tgid); |
| 2594 | leader = d_hash_and_lookup(mnt->mnt_root, &name); | 2608 | leader = d_hash_and_lookup(mnt->mnt_root, &name); |
| @@ -2645,17 +2659,16 @@ out: | |||
| 2645 | void proc_flush_task(struct task_struct *task) | 2659 | void proc_flush_task(struct task_struct *task) |
| 2646 | { | 2660 | { |
| 2647 | int i; | 2661 | int i; |
| 2648 | struct pid *pid, *tgid = NULL; | 2662 | struct pid *pid, *tgid; |
| 2649 | struct upid *upid; | 2663 | struct upid *upid; |
| 2650 | 2664 | ||
| 2651 | pid = task_pid(task); | 2665 | pid = task_pid(task); |
| 2652 | if (thread_group_leader(task)) | 2666 | tgid = task_tgid(task); |
| 2653 | tgid = task_tgid(task); | ||
| 2654 | 2667 | ||
| 2655 | for (i = 0; i <= pid->level; i++) { | 2668 | for (i = 0; i <= pid->level; i++) { |
| 2656 | upid = &pid->numbers[i]; | 2669 | upid = &pid->numbers[i]; |
| 2657 | proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr, | 2670 | proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr, |
| 2658 | tgid ? tgid->numbers[i].nr : 0); | 2671 | tgid->numbers[i].nr); |
| 2659 | } | 2672 | } |
| 2660 | 2673 | ||
| 2661 | upid = &pid->numbers[pid->level]; | 2674 | upid = &pid->numbers[pid->level]; |
diff --git a/fs/proc/kcore.c b/fs/proc/kcore.c index 59b43a068872..56013371f9f3 100644 --- a/fs/proc/kcore.c +++ b/fs/proc/kcore.c | |||
| @@ -17,9 +17,15 @@ | |||
| 17 | #include <linux/elfcore.h> | 17 | #include <linux/elfcore.h> |
| 18 | #include <linux/vmalloc.h> | 18 | #include <linux/vmalloc.h> |
| 19 | #include <linux/highmem.h> | 19 | #include <linux/highmem.h> |
| 20 | #include <linux/bootmem.h> | ||
| 20 | #include <linux/init.h> | 21 | #include <linux/init.h> |
| 21 | #include <asm/uaccess.h> | 22 | #include <asm/uaccess.h> |
| 22 | #include <asm/io.h> | 23 | #include <asm/io.h> |
| 24 | #include <linux/list.h> | ||
| 25 | #include <linux/ioport.h> | ||
| 26 | #include <linux/mm.h> | ||
| 27 | #include <linux/memory.h> | ||
| 28 | #include <asm/sections.h> | ||
| 23 | 29 | ||
| 24 | #define CORE_STR "CORE" | 30 | #define CORE_STR "CORE" |
| 25 | 31 | ||
| @@ -29,17 +35,6 @@ | |||
| 29 | 35 | ||
| 30 | static struct proc_dir_entry *proc_root_kcore; | 36 | static struct proc_dir_entry *proc_root_kcore; |
| 31 | 37 | ||
| 32 | static int open_kcore(struct inode * inode, struct file * filp) | ||
| 33 | { | ||
| 34 | return capable(CAP_SYS_RAWIO) ? 0 : -EPERM; | ||
| 35 | } | ||
| 36 | |||
| 37 | static ssize_t read_kcore(struct file *, char __user *, size_t, loff_t *); | ||
| 38 | |||
| 39 | static const struct file_operations proc_kcore_operations = { | ||
| 40 | .read = read_kcore, | ||
| 41 | .open = open_kcore, | ||
| 42 | }; | ||
| 43 | 38 | ||
| 44 | #ifndef kc_vaddr_to_offset | 39 | #ifndef kc_vaddr_to_offset |
| 45 | #define kc_vaddr_to_offset(v) ((v) - PAGE_OFFSET) | 40 | #define kc_vaddr_to_offset(v) ((v) - PAGE_OFFSET) |
| @@ -57,18 +52,19 @@ struct memelfnote | |||
| 57 | void *data; | 52 | void *data; |
| 58 | }; | 53 | }; |
| 59 | 54 | ||
| 60 | static struct kcore_list *kclist; | 55 | static LIST_HEAD(kclist_head); |
| 61 | static DEFINE_RWLOCK(kclist_lock); | 56 | static DEFINE_RWLOCK(kclist_lock); |
| 57 | static int kcore_need_update = 1; | ||
| 62 | 58 | ||
| 63 | void | 59 | void |
| 64 | kclist_add(struct kcore_list *new, void *addr, size_t size) | 60 | kclist_add(struct kcore_list *new, void *addr, size_t size, int type) |
| 65 | { | 61 | { |
| 66 | new->addr = (unsigned long)addr; | 62 | new->addr = (unsigned long)addr; |
| 67 | new->size = size; | 63 | new->size = size; |
| 64 | new->type = type; | ||
| 68 | 65 | ||
| 69 | write_lock(&kclist_lock); | 66 | write_lock(&kclist_lock); |
| 70 | new->next = kclist; | 67 | list_add_tail(&new->list, &kclist_head); |
| 71 | kclist = new; | ||
| 72 | write_unlock(&kclist_lock); | 68 | write_unlock(&kclist_lock); |
| 73 | } | 69 | } |
| 74 | 70 | ||
| @@ -80,7 +76,7 @@ static size_t get_kcore_size(int *nphdr, size_t *elf_buflen) | |||
| 80 | *nphdr = 1; /* PT_NOTE */ | 76 | *nphdr = 1; /* PT_NOTE */ |
| 81 | size = 0; | 77 | size = 0; |
| 82 | 78 | ||
| 83 | for (m=kclist; m; m=m->next) { | 79 | list_for_each_entry(m, &kclist_head, list) { |
| 84 | try = kc_vaddr_to_offset((size_t)m->addr + m->size); | 80 | try = kc_vaddr_to_offset((size_t)m->addr + m->size); |
| 85 | if (try > size) | 81 | if (try > size) |
| 86 | size = try; | 82 | size = try; |
| @@ -97,6 +93,177 @@ static size_t get_kcore_size(int *nphdr, size_t *elf_buflen) | |||
| 97 | return size + *elf_buflen; | 93 | return size + *elf_buflen; |
| 98 | } | 94 | } |
| 99 | 95 | ||
| 96 | static void free_kclist_ents(struct list_head *head) | ||
| 97 | { | ||
| 98 | struct kcore_list *tmp, *pos; | ||
| 99 | |||
| 100 | list_for_each_entry_safe(pos, tmp, head, list) { | ||
| 101 | list_del(&pos->list); | ||
| 102 | kfree(pos); | ||
| 103 | } | ||
| 104 | } | ||
| 105 | /* | ||
| 106 | * Replace all KCORE_RAM/KCORE_VMEMMAP information with passed list. | ||
| 107 | */ | ||
| 108 | static void __kcore_update_ram(struct list_head *list) | ||
| 109 | { | ||
| 110 | int nphdr; | ||
| 111 | size_t size; | ||
| 112 | struct kcore_list *tmp, *pos; | ||
| 113 | LIST_HEAD(garbage); | ||
| 114 | |||
| 115 | write_lock(&kclist_lock); | ||
| 116 | if (kcore_need_update) { | ||
| 117 | list_for_each_entry_safe(pos, tmp, &kclist_head, list) { | ||
| 118 | if (pos->type == KCORE_RAM | ||
| 119 | || pos->type == KCORE_VMEMMAP) | ||
| 120 | list_move(&pos->list, &garbage); | ||
| 121 | } | ||
| 122 | list_splice_tail(list, &kclist_head); | ||
| 123 | } else | ||
| 124 | list_splice(list, &garbage); | ||
| 125 | kcore_need_update = 0; | ||
| 126 | proc_root_kcore->size = get_kcore_size(&nphdr, &size); | ||
| 127 | write_unlock(&kclist_lock); | ||
| 128 | |||
| 129 | free_kclist_ents(&garbage); | ||
| 130 | } | ||
| 131 | |||
| 132 | |||
| 133 | #ifdef CONFIG_HIGHMEM | ||
| 134 | /* | ||
| 135 | * If no highmem, we can assume [0...max_low_pfn) continuous range of memory | ||
| 136 | * because memory hole is not as big as !HIGHMEM case. | ||
| 137 | * (HIGHMEM is special because part of memory is _invisible_ from the kernel.) | ||
| 138 | */ | ||
| 139 | static int kcore_update_ram(void) | ||
| 140 | { | ||
| 141 | LIST_HEAD(head); | ||
| 142 | struct kcore_list *ent; | ||
| 143 | int ret = 0; | ||
| 144 | |||
| 145 | ent = kmalloc(sizeof(*ent), GFP_KERNEL); | ||
| 146 | if (!ent) | ||
| 147 | return -ENOMEM; | ||
| 148 | ent->addr = (unsigned long)__va(0); | ||
| 149 | ent->size = max_low_pfn << PAGE_SHIFT; | ||
| 150 | ent->type = KCORE_RAM; | ||
| 151 | list_add(&ent->list, &head); | ||
| 152 | __kcore_update_ram(&head); | ||
| 153 | return ret; | ||
| 154 | } | ||
| 155 | |||
| 156 | #else /* !CONFIG_HIGHMEM */ | ||
| 157 | |||
| 158 | #ifdef CONFIG_SPARSEMEM_VMEMMAP | ||
| 159 | /* calculate vmemmap's address from given system ram pfn and register it */ | ||
| 160 | int get_sparsemem_vmemmap_info(struct kcore_list *ent, struct list_head *head) | ||
| 161 | { | ||
| 162 | unsigned long pfn = __pa(ent->addr) >> PAGE_SHIFT; | ||
| 163 | unsigned long nr_pages = ent->size >> PAGE_SHIFT; | ||
| 164 | unsigned long start, end; | ||
| 165 | struct kcore_list *vmm, *tmp; | ||
| 166 | |||
| 167 | |||
| 168 | start = ((unsigned long)pfn_to_page(pfn)) & PAGE_MASK; | ||
| 169 | end = ((unsigned long)pfn_to_page(pfn + nr_pages)) - 1; | ||
| 170 | end = ALIGN(end, PAGE_SIZE); | ||
| 171 | /* overlap check (because we have to align page */ | ||
| 172 | list_for_each_entry(tmp, head, list) { | ||
| 173 | if (tmp->type != KCORE_VMEMMAP) | ||
| 174 | continue; | ||
| 175 | if (start < tmp->addr + tmp->size) | ||
| 176 | if (end > tmp->addr) | ||
| 177 | end = tmp->addr; | ||
| 178 | } | ||
| 179 | if (start < end) { | ||
| 180 | vmm = kmalloc(sizeof(*vmm), GFP_KERNEL); | ||
| 181 | if (!vmm) | ||
| 182 | return 0; | ||
| 183 | vmm->addr = start; | ||
| 184 | vmm->size = end - start; | ||
| 185 | vmm->type = KCORE_VMEMMAP; | ||
| 186 | list_add_tail(&vmm->list, head); | ||
| 187 | } | ||
| 188 | return 1; | ||
| 189 | |||
| 190 | } | ||
| 191 | #else | ||
| 192 | int get_sparsemem_vmemmap_info(struct kcore_list *ent, struct list_head *head) | ||
| 193 | { | ||
| 194 | return 1; | ||
| 195 | } | ||
| 196 | |||
| 197 | #endif | ||
| 198 | |||
| 199 | static int | ||
| 200 | kclist_add_private(unsigned long pfn, unsigned long nr_pages, void *arg) | ||
| 201 | { | ||
| 202 | struct list_head *head = (struct list_head *)arg; | ||
| 203 | struct kcore_list *ent; | ||
| 204 | |||
| 205 | ent = kmalloc(sizeof(*ent), GFP_KERNEL); | ||
| 206 | if (!ent) | ||
| 207 | return -ENOMEM; | ||
| 208 | ent->addr = (unsigned long)__va((pfn << PAGE_SHIFT)); | ||
| 209 | ent->size = nr_pages << PAGE_SHIFT; | ||
| 210 | |||
| 211 | /* Sanity check: Can happen in 32bit arch...maybe */ | ||
| 212 | if (ent->addr < (unsigned long) __va(0)) | ||
| 213 | goto free_out; | ||
| 214 | |||
| 215 | /* cut not-mapped area. ....from ppc-32 code. */ | ||
| 216 | if (ULONG_MAX - ent->addr < ent->size) | ||
| 217 | ent->size = ULONG_MAX - ent->addr; | ||
| 218 | |||
| 219 | /* cut when vmalloc() area is higher than direct-map area */ | ||
| 220 | if (VMALLOC_START > (unsigned long)__va(0)) { | ||
| 221 | if (ent->addr > VMALLOC_START) | ||
| 222 | goto free_out; | ||
| 223 | if (VMALLOC_START - ent->addr < ent->size) | ||
| 224 | ent->size = VMALLOC_START - ent->addr; | ||
| 225 | } | ||
| 226 | |||
| 227 | ent->type = KCORE_RAM; | ||
| 228 | list_add_tail(&ent->list, head); | ||
| 229 | |||
| 230 | if (!get_sparsemem_vmemmap_info(ent, head)) { | ||
| 231 | list_del(&ent->list); | ||
| 232 | goto free_out; | ||
| 233 | } | ||
| 234 | |||
| 235 | return 0; | ||
| 236 | free_out: | ||
| 237 | kfree(ent); | ||
| 238 | return 1; | ||
| 239 | } | ||
| 240 | |||
| 241 | static int kcore_update_ram(void) | ||
| 242 | { | ||
| 243 | int nid, ret; | ||
| 244 | unsigned long end_pfn; | ||
| 245 | LIST_HEAD(head); | ||
| 246 | |||
| 247 | /* Not inialized....update now */ | ||
| 248 | /* find out "max pfn" */ | ||
| 249 | end_pfn = 0; | ||
| 250 | for_each_node_state(nid, N_HIGH_MEMORY) { | ||
| 251 | unsigned long node_end; | ||
| 252 | node_end = NODE_DATA(nid)->node_start_pfn + | ||
| 253 | NODE_DATA(nid)->node_spanned_pages; | ||
| 254 | if (end_pfn < node_end) | ||
| 255 | end_pfn = node_end; | ||
| 256 | } | ||
| 257 | /* scan 0 to max_pfn */ | ||
| 258 | ret = walk_system_ram_range(0, end_pfn, &head, kclist_add_private); | ||
| 259 | if (ret) { | ||
| 260 | free_kclist_ents(&head); | ||
| 261 | return -ENOMEM; | ||
| 262 | } | ||
| 263 | __kcore_update_ram(&head); | ||
| 264 | return ret; | ||
| 265 | } | ||
| 266 | #endif /* CONFIG_HIGHMEM */ | ||
| 100 | 267 | ||
| 101 | /*****************************************************************************/ | 268 | /*****************************************************************************/ |
| 102 | /* | 269 | /* |
| @@ -192,7 +359,7 @@ static void elf_kcore_store_hdr(char *bufp, int nphdr, int dataoff) | |||
| 192 | nhdr->p_align = 0; | 359 | nhdr->p_align = 0; |
| 193 | 360 | ||
| 194 | /* setup ELF PT_LOAD program header for every area */ | 361 | /* setup ELF PT_LOAD program header for every area */ |
| 195 | for (m=kclist; m; m=m->next) { | 362 | list_for_each_entry(m, &kclist_head, list) { |
| 196 | phdr = (struct elf_phdr *) bufp; | 363 | phdr = (struct elf_phdr *) bufp; |
| 197 | bufp += sizeof(struct elf_phdr); | 364 | bufp += sizeof(struct elf_phdr); |
| 198 | offset += sizeof(struct elf_phdr); | 365 | offset += sizeof(struct elf_phdr); |
| @@ -265,7 +432,8 @@ read_kcore(struct file *file, char __user *buffer, size_t buflen, loff_t *fpos) | |||
| 265 | unsigned long start; | 432 | unsigned long start; |
| 266 | 433 | ||
| 267 | read_lock(&kclist_lock); | 434 | read_lock(&kclist_lock); |
| 268 | proc_root_kcore->size = size = get_kcore_size(&nphdr, &elf_buflen); | 435 | size = get_kcore_size(&nphdr, &elf_buflen); |
| 436 | |||
| 269 | if (buflen == 0 || *fpos >= size) { | 437 | if (buflen == 0 || *fpos >= size) { |
| 270 | read_unlock(&kclist_lock); | 438 | read_unlock(&kclist_lock); |
| 271 | return 0; | 439 | return 0; |
| @@ -317,7 +485,7 @@ read_kcore(struct file *file, char __user *buffer, size_t buflen, loff_t *fpos) | |||
| 317 | struct kcore_list *m; | 485 | struct kcore_list *m; |
| 318 | 486 | ||
| 319 | read_lock(&kclist_lock); | 487 | read_lock(&kclist_lock); |
| 320 | for (m=kclist; m; m=m->next) { | 488 | list_for_each_entry(m, &kclist_head, list) { |
| 321 | if (start >= m->addr && start < (m->addr+m->size)) | 489 | if (start >= m->addr && start < (m->addr+m->size)) |
| 322 | break; | 490 | break; |
| 323 | } | 491 | } |
| @@ -326,45 +494,14 @@ read_kcore(struct file *file, char __user *buffer, size_t buflen, loff_t *fpos) | |||
| 326 | if (m == NULL) { | 494 | if (m == NULL) { |
| 327 | if (clear_user(buffer, tsz)) | 495 | if (clear_user(buffer, tsz)) |
| 328 | return -EFAULT; | 496 | return -EFAULT; |
| 329 | } else if (is_vmalloc_addr((void *)start)) { | 497 | } else if (is_vmalloc_or_module_addr((void *)start)) { |
| 330 | char * elf_buf; | 498 | char * elf_buf; |
| 331 | struct vm_struct *m; | ||
| 332 | unsigned long curstart = start; | ||
| 333 | unsigned long cursize = tsz; | ||
| 334 | 499 | ||
| 335 | elf_buf = kzalloc(tsz, GFP_KERNEL); | 500 | elf_buf = kzalloc(tsz, GFP_KERNEL); |
| 336 | if (!elf_buf) | 501 | if (!elf_buf) |
| 337 | return -ENOMEM; | 502 | return -ENOMEM; |
| 338 | 503 | vread(elf_buf, (char *)start, tsz); | |
| 339 | read_lock(&vmlist_lock); | 504 | /* we have to zero-fill user buffer even if no read */ |
| 340 | for (m=vmlist; m && cursize; m=m->next) { | ||
| 341 | unsigned long vmstart; | ||
| 342 | unsigned long vmsize; | ||
| 343 | unsigned long msize = m->size - PAGE_SIZE; | ||
| 344 | |||
| 345 | if (((unsigned long)m->addr + msize) < | ||
| 346 | curstart) | ||
| 347 | continue; | ||
| 348 | if ((unsigned long)m->addr > (curstart + | ||
| 349 | cursize)) | ||
| 350 | break; | ||
| 351 | vmstart = (curstart < (unsigned long)m->addr ? | ||
| 352 | (unsigned long)m->addr : curstart); | ||
| 353 | if (((unsigned long)m->addr + msize) > | ||
| 354 | (curstart + cursize)) | ||
| 355 | vmsize = curstart + cursize - vmstart; | ||
| 356 | else | ||
| 357 | vmsize = (unsigned long)m->addr + | ||
| 358 | msize - vmstart; | ||
| 359 | curstart = vmstart + vmsize; | ||
| 360 | cursize -= vmsize; | ||
| 361 | /* don't dump ioremap'd stuff! (TA) */ | ||
| 362 | if (m->flags & VM_IOREMAP) | ||
| 363 | continue; | ||
| 364 | memcpy(elf_buf + (vmstart - start), | ||
| 365 | (char *)vmstart, vmsize); | ||
| 366 | } | ||
| 367 | read_unlock(&vmlist_lock); | ||
| 368 | if (copy_to_user(buffer, elf_buf, tsz)) { | 505 | if (copy_to_user(buffer, elf_buf, tsz)) { |
| 369 | kfree(elf_buf); | 506 | kfree(elf_buf); |
| 370 | return -EFAULT; | 507 | return -EFAULT; |
| @@ -402,12 +539,96 @@ read_kcore(struct file *file, char __user *buffer, size_t buflen, loff_t *fpos) | |||
| 402 | return acc; | 539 | return acc; |
| 403 | } | 540 | } |
| 404 | 541 | ||
| 542 | |||
| 543 | static int open_kcore(struct inode *inode, struct file *filp) | ||
| 544 | { | ||
| 545 | if (!capable(CAP_SYS_RAWIO)) | ||
| 546 | return -EPERM; | ||
| 547 | if (kcore_need_update) | ||
| 548 | kcore_update_ram(); | ||
| 549 | if (i_size_read(inode) != proc_root_kcore->size) { | ||
| 550 | mutex_lock(&inode->i_mutex); | ||
| 551 | i_size_write(inode, proc_root_kcore->size); | ||
| 552 | mutex_unlock(&inode->i_mutex); | ||
| 553 | } | ||
| 554 | return 0; | ||
| 555 | } | ||
| 556 | |||
| 557 | |||
| 558 | static const struct file_operations proc_kcore_operations = { | ||
| 559 | .read = read_kcore, | ||
| 560 | .open = open_kcore, | ||
| 561 | }; | ||
| 562 | |||
| 563 | #ifdef CONFIG_MEMORY_HOTPLUG | ||
| 564 | /* just remember that we have to update kcore */ | ||
| 565 | static int __meminit kcore_callback(struct notifier_block *self, | ||
| 566 | unsigned long action, void *arg) | ||
| 567 | { | ||
| 568 | switch (action) { | ||
| 569 | case MEM_ONLINE: | ||
| 570 | case MEM_OFFLINE: | ||
| 571 | write_lock(&kclist_lock); | ||
| 572 | kcore_need_update = 1; | ||
| 573 | write_unlock(&kclist_lock); | ||
| 574 | } | ||
| 575 | return NOTIFY_OK; | ||
| 576 | } | ||
| 577 | #endif | ||
| 578 | |||
| 579 | |||
| 580 | static struct kcore_list kcore_vmalloc; | ||
| 581 | |||
| 582 | #ifdef CONFIG_ARCH_PROC_KCORE_TEXT | ||
| 583 | static struct kcore_list kcore_text; | ||
| 584 | /* | ||
| 585 | * If defined, special segment is used for mapping kernel text instead of | ||
| 586 | * direct-map area. We need to create special TEXT section. | ||
| 587 | */ | ||
| 588 | static void __init proc_kcore_text_init(void) | ||
| 589 | { | ||
| 590 | kclist_add(&kcore_text, _stext, _end - _stext, KCORE_TEXT); | ||
| 591 | } | ||
| 592 | #else | ||
| 593 | static void __init proc_kcore_text_init(void) | ||
| 594 | { | ||
| 595 | } | ||
| 596 | #endif | ||
| 597 | |||
| 598 | #if defined(CONFIG_MODULES) && defined(MODULES_VADDR) | ||
| 599 | /* | ||
| 600 | * MODULES_VADDR has no intersection with VMALLOC_ADDR. | ||
| 601 | */ | ||
| 602 | struct kcore_list kcore_modules; | ||
| 603 | static void __init add_modules_range(void) | ||
| 604 | { | ||
| 605 | kclist_add(&kcore_modules, (void *)MODULES_VADDR, | ||
| 606 | MODULES_END - MODULES_VADDR, KCORE_VMALLOC); | ||
| 607 | } | ||
| 608 | #else | ||
| 609 | static void __init add_modules_range(void) | ||
| 610 | { | ||
| 611 | } | ||
| 612 | #endif | ||
| 613 | |||
| 405 | static int __init proc_kcore_init(void) | 614 | static int __init proc_kcore_init(void) |
| 406 | { | 615 | { |
| 407 | proc_root_kcore = proc_create("kcore", S_IRUSR, NULL, &proc_kcore_operations); | 616 | proc_root_kcore = proc_create("kcore", S_IRUSR, NULL, |
| 408 | if (proc_root_kcore) | 617 | &proc_kcore_operations); |
| 409 | proc_root_kcore->size = | 618 | if (!proc_root_kcore) { |
| 410 | (size_t)high_memory - PAGE_OFFSET + PAGE_SIZE; | 619 | printk(KERN_ERR "couldn't create /proc/kcore\n"); |
| 620 | return 0; /* Always returns 0. */ | ||
| 621 | } | ||
| 622 | /* Store text area if it's special */ | ||
| 623 | proc_kcore_text_init(); | ||
| 624 | /* Store vmalloc area */ | ||
| 625 | kclist_add(&kcore_vmalloc, (void *)VMALLOC_START, | ||
| 626 | VMALLOC_END - VMALLOC_START, KCORE_VMALLOC); | ||
| 627 | add_modules_range(); | ||
| 628 | /* Store direct-map area from physical memory map */ | ||
| 629 | kcore_update_ram(); | ||
| 630 | hotplug_memory_notifier(kcore_callback, 0); | ||
| 631 | |||
| 411 | return 0; | 632 | return 0; |
| 412 | } | 633 | } |
| 413 | module_init(proc_kcore_init); | 634 | module_init(proc_kcore_init); |
diff --git a/fs/proc/meminfo.c b/fs/proc/meminfo.c index d5c410d47fae..171e052c07b3 100644 --- a/fs/proc/meminfo.c +++ b/fs/proc/meminfo.c | |||
| @@ -81,9 +81,11 @@ static int meminfo_proc_show(struct seq_file *m, void *v) | |||
| 81 | "Writeback: %8lu kB\n" | 81 | "Writeback: %8lu kB\n" |
| 82 | "AnonPages: %8lu kB\n" | 82 | "AnonPages: %8lu kB\n" |
| 83 | "Mapped: %8lu kB\n" | 83 | "Mapped: %8lu kB\n" |
| 84 | "Shmem: %8lu kB\n" | ||
| 84 | "Slab: %8lu kB\n" | 85 | "Slab: %8lu kB\n" |
| 85 | "SReclaimable: %8lu kB\n" | 86 | "SReclaimable: %8lu kB\n" |
| 86 | "SUnreclaim: %8lu kB\n" | 87 | "SUnreclaim: %8lu kB\n" |
| 88 | "KernelStack: %8lu kB\n" | ||
| 87 | "PageTables: %8lu kB\n" | 89 | "PageTables: %8lu kB\n" |
| 88 | #ifdef CONFIG_QUICKLIST | 90 | #ifdef CONFIG_QUICKLIST |
| 89 | "Quicklists: %8lu kB\n" | 91 | "Quicklists: %8lu kB\n" |
| @@ -124,10 +126,12 @@ static int meminfo_proc_show(struct seq_file *m, void *v) | |||
| 124 | K(global_page_state(NR_WRITEBACK)), | 126 | K(global_page_state(NR_WRITEBACK)), |
| 125 | K(global_page_state(NR_ANON_PAGES)), | 127 | K(global_page_state(NR_ANON_PAGES)), |
| 126 | K(global_page_state(NR_FILE_MAPPED)), | 128 | K(global_page_state(NR_FILE_MAPPED)), |
| 129 | K(global_page_state(NR_SHMEM)), | ||
| 127 | K(global_page_state(NR_SLAB_RECLAIMABLE) + | 130 | K(global_page_state(NR_SLAB_RECLAIMABLE) + |
| 128 | global_page_state(NR_SLAB_UNRECLAIMABLE)), | 131 | global_page_state(NR_SLAB_UNRECLAIMABLE)), |
| 129 | K(global_page_state(NR_SLAB_RECLAIMABLE)), | 132 | K(global_page_state(NR_SLAB_RECLAIMABLE)), |
| 130 | K(global_page_state(NR_SLAB_UNRECLAIMABLE)), | 133 | K(global_page_state(NR_SLAB_UNRECLAIMABLE)), |
| 134 | global_page_state(NR_KERNEL_STACK) * THREAD_SIZE / 1024, | ||
| 131 | K(global_page_state(NR_PAGETABLE)), | 135 | K(global_page_state(NR_PAGETABLE)), |
| 132 | #ifdef CONFIG_QUICKLIST | 136 | #ifdef CONFIG_QUICKLIST |
| 133 | K(quicklist_total_size()), | 137 | K(quicklist_total_size()), |
diff --git a/fs/proc/nommu.c b/fs/proc/nommu.c index 7e14d1a04001..9fe7d7ebe115 100644 --- a/fs/proc/nommu.c +++ b/fs/proc/nommu.c | |||
| @@ -109,7 +109,7 @@ static void *nommu_region_list_next(struct seq_file *m, void *v, loff_t *pos) | |||
| 109 | return rb_next((struct rb_node *) v); | 109 | return rb_next((struct rb_node *) v); |
| 110 | } | 110 | } |
| 111 | 111 | ||
| 112 | static struct seq_operations proc_nommu_region_list_seqop = { | 112 | static const struct seq_operations proc_nommu_region_list_seqop = { |
| 113 | .start = nommu_region_list_start, | 113 | .start = nommu_region_list_start, |
| 114 | .next = nommu_region_list_next, | 114 | .next = nommu_region_list_next, |
| 115 | .stop = nommu_region_list_stop, | 115 | .stop = nommu_region_list_stop, |
diff --git a/fs/proc/page.c b/fs/proc/page.c index 2707c6c7a20f..2281c2cbfe2b 100644 --- a/fs/proc/page.c +++ b/fs/proc/page.c | |||
| @@ -2,6 +2,7 @@ | |||
| 2 | #include <linux/compiler.h> | 2 | #include <linux/compiler.h> |
| 3 | #include <linux/fs.h> | 3 | #include <linux/fs.h> |
| 4 | #include <linux/init.h> | 4 | #include <linux/init.h> |
| 5 | #include <linux/ksm.h> | ||
| 5 | #include <linux/mm.h> | 6 | #include <linux/mm.h> |
| 6 | #include <linux/mmzone.h> | 7 | #include <linux/mmzone.h> |
| 7 | #include <linux/proc_fs.h> | 8 | #include <linux/proc_fs.h> |
| @@ -95,6 +96,8 @@ static const struct file_operations proc_kpagecount_operations = { | |||
| 95 | #define KPF_UNEVICTABLE 18 | 96 | #define KPF_UNEVICTABLE 18 |
| 96 | #define KPF_NOPAGE 20 | 97 | #define KPF_NOPAGE 20 |
| 97 | 98 | ||
| 99 | #define KPF_KSM 21 | ||
| 100 | |||
| 98 | /* kernel hacking assistances | 101 | /* kernel hacking assistances |
| 99 | * WARNING: subject to change, never rely on them! | 102 | * WARNING: subject to change, never rely on them! |
| 100 | */ | 103 | */ |
| @@ -137,6 +140,8 @@ static u64 get_uflags(struct page *page) | |||
| 137 | u |= 1 << KPF_MMAP; | 140 | u |= 1 << KPF_MMAP; |
| 138 | if (PageAnon(page)) | 141 | if (PageAnon(page)) |
| 139 | u |= 1 << KPF_ANON; | 142 | u |= 1 << KPF_ANON; |
| 143 | if (PageKsm(page)) | ||
| 144 | u |= 1 << KPF_KSM; | ||
| 140 | 145 | ||
| 141 | /* | 146 | /* |
| 142 | * compound pages: export both head/tail info | 147 | * compound pages: export both head/tail info |
diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c index 9bd8be1d235c..2a1bef9203c6 100644 --- a/fs/proc/task_mmu.c +++ b/fs/proc/task_mmu.c | |||
| @@ -243,6 +243,25 @@ static void show_map_vma(struct seq_file *m, struct vm_area_struct *vma) | |||
| 243 | } else if (vma->vm_start <= mm->start_stack && | 243 | } else if (vma->vm_start <= mm->start_stack && |
| 244 | vma->vm_end >= mm->start_stack) { | 244 | vma->vm_end >= mm->start_stack) { |
| 245 | name = "[stack]"; | 245 | name = "[stack]"; |
| 246 | } else { | ||
| 247 | unsigned long stack_start; | ||
| 248 | struct proc_maps_private *pmp; | ||
| 249 | |||
| 250 | pmp = m->private; | ||
| 251 | stack_start = pmp->task->stack_start; | ||
| 252 | |||
| 253 | if (vma->vm_start <= stack_start && | ||
| 254 | vma->vm_end >= stack_start) { | ||
| 255 | pad_len_spaces(m, len); | ||
| 256 | seq_printf(m, | ||
| 257 | "[threadstack:%08lx]", | ||
| 258 | #ifdef CONFIG_STACK_GROWSUP | ||
| 259 | vma->vm_end - stack_start | ||
| 260 | #else | ||
| 261 | stack_start - vma->vm_start | ||
| 262 | #endif | ||
| 263 | ); | ||
| 264 | } | ||
| 246 | } | 265 | } |
| 247 | } else { | 266 | } else { |
| 248 | name = "[vdso]"; | 267 | name = "[vdso]"; |
| @@ -465,23 +484,28 @@ static int clear_refs_pte_range(pmd_t *pmd, unsigned long addr, | |||
| 465 | return 0; | 484 | return 0; |
| 466 | } | 485 | } |
| 467 | 486 | ||
| 487 | #define CLEAR_REFS_ALL 1 | ||
| 488 | #define CLEAR_REFS_ANON 2 | ||
| 489 | #define CLEAR_REFS_MAPPED 3 | ||
| 490 | |||
| 468 | static ssize_t clear_refs_write(struct file *file, const char __user *buf, | 491 | static ssize_t clear_refs_write(struct file *file, const char __user *buf, |
| 469 | size_t count, loff_t *ppos) | 492 | size_t count, loff_t *ppos) |
| 470 | { | 493 | { |
| 471 | struct task_struct *task; | 494 | struct task_struct *task; |
| 472 | char buffer[PROC_NUMBUF], *end; | 495 | char buffer[PROC_NUMBUF]; |
| 473 | struct mm_struct *mm; | 496 | struct mm_struct *mm; |
| 474 | struct vm_area_struct *vma; | 497 | struct vm_area_struct *vma; |
| 498 | long type; | ||
| 475 | 499 | ||
| 476 | memset(buffer, 0, sizeof(buffer)); | 500 | memset(buffer, 0, sizeof(buffer)); |
| 477 | if (count > sizeof(buffer) - 1) | 501 | if (count > sizeof(buffer) - 1) |
| 478 | count = sizeof(buffer) - 1; | 502 | count = sizeof(buffer) - 1; |
| 479 | if (copy_from_user(buffer, buf, count)) | 503 | if (copy_from_user(buffer, buf, count)) |
| 480 | return -EFAULT; | 504 | return -EFAULT; |
| 481 | if (!simple_strtol(buffer, &end, 0)) | 505 | if (strict_strtol(strstrip(buffer), 10, &type)) |
| 506 | return -EINVAL; | ||
| 507 | if (type < CLEAR_REFS_ALL || type > CLEAR_REFS_MAPPED) | ||
| 482 | return -EINVAL; | 508 | return -EINVAL; |
| 483 | if (*end == '\n') | ||
| 484 | end++; | ||
| 485 | task = get_proc_task(file->f_path.dentry->d_inode); | 509 | task = get_proc_task(file->f_path.dentry->d_inode); |
| 486 | if (!task) | 510 | if (!task) |
| 487 | return -ESRCH; | 511 | return -ESRCH; |
| @@ -494,18 +518,31 @@ static ssize_t clear_refs_write(struct file *file, const char __user *buf, | |||
| 494 | down_read(&mm->mmap_sem); | 518 | down_read(&mm->mmap_sem); |
| 495 | for (vma = mm->mmap; vma; vma = vma->vm_next) { | 519 | for (vma = mm->mmap; vma; vma = vma->vm_next) { |
| 496 | clear_refs_walk.private = vma; | 520 | clear_refs_walk.private = vma; |
| 497 | if (!is_vm_hugetlb_page(vma)) | 521 | if (is_vm_hugetlb_page(vma)) |
| 498 | walk_page_range(vma->vm_start, vma->vm_end, | 522 | continue; |
| 499 | &clear_refs_walk); | 523 | /* |
| 524 | * Writing 1 to /proc/pid/clear_refs affects all pages. | ||
| 525 | * | ||
| 526 | * Writing 2 to /proc/pid/clear_refs only affects | ||
| 527 | * Anonymous pages. | ||
| 528 | * | ||
| 529 | * Writing 3 to /proc/pid/clear_refs only affects file | ||
| 530 | * mapped pages. | ||
| 531 | */ | ||
| 532 | if (type == CLEAR_REFS_ANON && vma->vm_file) | ||
| 533 | continue; | ||
| 534 | if (type == CLEAR_REFS_MAPPED && !vma->vm_file) | ||
| 535 | continue; | ||
| 536 | walk_page_range(vma->vm_start, vma->vm_end, | ||
| 537 | &clear_refs_walk); | ||
| 500 | } | 538 | } |
| 501 | flush_tlb_mm(mm); | 539 | flush_tlb_mm(mm); |
| 502 | up_read(&mm->mmap_sem); | 540 | up_read(&mm->mmap_sem); |
| 503 | mmput(mm); | 541 | mmput(mm); |
| 504 | } | 542 | } |
| 505 | put_task_struct(task); | 543 | put_task_struct(task); |
| 506 | if (end - buffer == 0) | 544 | |
| 507 | return -EIO; | 545 | return count; |
| 508 | return end - buffer; | ||
| 509 | } | 546 | } |
| 510 | 547 | ||
| 511 | const struct file_operations proc_clear_refs_operations = { | 548 | const struct file_operations proc_clear_refs_operations = { |
