aboutsummaryrefslogtreecommitdiffstats
path: root/fs/proc
diff options
context:
space:
mode:
Diffstat (limited to 'fs/proc')
-rw-r--r--fs/proc/array.c134
-rw-r--r--fs/proc/base.c178
-rw-r--r--fs/proc/inode.c23
-rw-r--r--fs/proc/internal.h12
-rw-r--r--fs/proc/kcore.c8
-rw-r--r--fs/proc/namespaces.c8
-rw-r--r--fs/proc/page.c2
-rw-r--r--fs/proc/proc_sysctl.c1276
-rw-r--r--fs/proc/root.c11
-rw-r--r--fs/proc/stat.c96
-rw-r--r--fs/proc/task_mmu.c370
-rw-r--r--fs/proc/task_nommu.c69
-rw-r--r--fs/proc/vmcore.c23
13 files changed, 1836 insertions, 374 deletions
diff --git a/fs/proc/array.c b/fs/proc/array.c
index c602b8d20f06..dc4c5a7b9ece 100644
--- a/fs/proc/array.c
+++ b/fs/proc/array.c
@@ -81,6 +81,7 @@
81#include <linux/pid_namespace.h> 81#include <linux/pid_namespace.h>
82#include <linux/ptrace.h> 82#include <linux/ptrace.h>
83#include <linux/tracehook.h> 83#include <linux/tracehook.h>
84#include <linux/user_namespace.h>
84 85
85#include <asm/pgtable.h> 86#include <asm/pgtable.h>
86#include <asm/processor.h> 87#include <asm/processor.h>
@@ -161,6 +162,7 @@ static inline const char *get_task_state(struct task_struct *tsk)
161static inline void task_state(struct seq_file *m, struct pid_namespace *ns, 162static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
162 struct pid *pid, struct task_struct *p) 163 struct pid *pid, struct task_struct *p)
163{ 164{
165 struct user_namespace *user_ns = current_user_ns();
164 struct group_info *group_info; 166 struct group_info *group_info;
165 int g; 167 int g;
166 struct fdtable *fdt = NULL; 168 struct fdtable *fdt = NULL;
@@ -189,8 +191,14 @@ static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
189 task_tgid_nr_ns(p, ns), 191 task_tgid_nr_ns(p, ns),
190 pid_nr_ns(pid, ns), 192 pid_nr_ns(pid, ns),
191 ppid, tpid, 193 ppid, tpid,
192 cred->uid, cred->euid, cred->suid, cred->fsuid, 194 from_kuid_munged(user_ns, cred->uid),
193 cred->gid, cred->egid, cred->sgid, cred->fsgid); 195 from_kuid_munged(user_ns, cred->euid),
196 from_kuid_munged(user_ns, cred->suid),
197 from_kuid_munged(user_ns, cred->fsuid),
198 from_kgid_munged(user_ns, cred->gid),
199 from_kgid_munged(user_ns, cred->egid),
200 from_kgid_munged(user_ns, cred->sgid),
201 from_kgid_munged(user_ns, cred->fsgid));
194 202
195 task_lock(p); 203 task_lock(p);
196 if (p->files) 204 if (p->files)
@@ -205,7 +213,8 @@ static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
205 task_unlock(p); 213 task_unlock(p);
206 214
207 for (g = 0; g < min(group_info->ngroups, NGROUPS_SMALL); g++) 215 for (g = 0; g < min(group_info->ngroups, NGROUPS_SMALL); g++)
208 seq_printf(m, "%d ", GROUP_AT(group_info, g)); 216 seq_printf(m, "%d ",
217 from_kgid_munged(user_ns, GROUP_AT(group_info, g)));
209 put_cred(cred); 218 put_cred(cred);
210 219
211 seq_putc(m, '\n'); 220 seq_putc(m, '\n');
@@ -462,59 +471,56 @@ static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
462 /* convert nsec -> ticks */ 471 /* convert nsec -> ticks */
463 start_time = nsec_to_clock_t(start_time); 472 start_time = nsec_to_clock_t(start_time);
464 473
465 seq_printf(m, "%d (%s) %c %d %d %d %d %d %u %lu \ 474 seq_printf(m, "%d (%s) %c", pid_nr_ns(pid, ns), tcomm, state);
466%lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \ 475 seq_put_decimal_ll(m, ' ', ppid);
467%lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu %lu %ld %lu %lu %lu\n", 476 seq_put_decimal_ll(m, ' ', pgid);
468 pid_nr_ns(pid, ns), 477 seq_put_decimal_ll(m, ' ', sid);
469 tcomm, 478 seq_put_decimal_ll(m, ' ', tty_nr);
470 state, 479 seq_put_decimal_ll(m, ' ', tty_pgrp);
471 ppid, 480 seq_put_decimal_ull(m, ' ', task->flags);
472 pgid, 481 seq_put_decimal_ull(m, ' ', min_flt);
473 sid, 482 seq_put_decimal_ull(m, ' ', cmin_flt);
474 tty_nr, 483 seq_put_decimal_ull(m, ' ', maj_flt);
475 tty_pgrp, 484 seq_put_decimal_ull(m, ' ', cmaj_flt);
476 task->flags, 485 seq_put_decimal_ull(m, ' ', cputime_to_clock_t(utime));
477 min_flt, 486 seq_put_decimal_ull(m, ' ', cputime_to_clock_t(stime));
478 cmin_flt, 487 seq_put_decimal_ll(m, ' ', cputime_to_clock_t(cutime));
479 maj_flt, 488 seq_put_decimal_ll(m, ' ', cputime_to_clock_t(cstime));
480 cmaj_flt, 489 seq_put_decimal_ll(m, ' ', priority);
481 cputime_to_clock_t(utime), 490 seq_put_decimal_ll(m, ' ', nice);
482 cputime_to_clock_t(stime), 491 seq_put_decimal_ll(m, ' ', num_threads);
483 cputime_to_clock_t(cutime), 492 seq_put_decimal_ull(m, ' ', 0);
484 cputime_to_clock_t(cstime), 493 seq_put_decimal_ull(m, ' ', start_time);
485 priority, 494 seq_put_decimal_ull(m, ' ', vsize);
486 nice, 495 seq_put_decimal_ll(m, ' ', mm ? get_mm_rss(mm) : 0);
487 num_threads, 496 seq_put_decimal_ull(m, ' ', rsslim);
488 start_time, 497 seq_put_decimal_ull(m, ' ', mm ? (permitted ? mm->start_code : 1) : 0);
489 vsize, 498 seq_put_decimal_ull(m, ' ', mm ? (permitted ? mm->end_code : 1) : 0);
490 mm ? get_mm_rss(mm) : 0, 499 seq_put_decimal_ull(m, ' ', (permitted && mm) ? mm->start_stack : 0);
491 rsslim, 500 seq_put_decimal_ull(m, ' ', esp);
492 mm ? (permitted ? mm->start_code : 1) : 0, 501 seq_put_decimal_ull(m, ' ', eip);
493 mm ? (permitted ? mm->end_code : 1) : 0, 502 /* The signal information here is obsolete.
494 (permitted && mm) ? mm->start_stack : 0, 503 * It must be decimal for Linux 2.0 compatibility.
495 esp, 504 * Use /proc/#/status for real-time signals.
496 eip, 505 */
497 /* The signal information here is obsolete. 506 seq_put_decimal_ull(m, ' ', task->pending.signal.sig[0] & 0x7fffffffUL);
498 * It must be decimal for Linux 2.0 compatibility. 507 seq_put_decimal_ull(m, ' ', task->blocked.sig[0] & 0x7fffffffUL);
499 * Use /proc/#/status for real-time signals. 508 seq_put_decimal_ull(m, ' ', sigign.sig[0] & 0x7fffffffUL);
500 */ 509 seq_put_decimal_ull(m, ' ', sigcatch.sig[0] & 0x7fffffffUL);
501 task->pending.signal.sig[0] & 0x7fffffffUL, 510 seq_put_decimal_ull(m, ' ', wchan);
502 task->blocked.sig[0] & 0x7fffffffUL, 511 seq_put_decimal_ull(m, ' ', 0);
503 sigign .sig[0] & 0x7fffffffUL, 512 seq_put_decimal_ull(m, ' ', 0);
504 sigcatch .sig[0] & 0x7fffffffUL, 513 seq_put_decimal_ll(m, ' ', task->exit_signal);
505 wchan, 514 seq_put_decimal_ll(m, ' ', task_cpu(task));
506 0UL, 515 seq_put_decimal_ull(m, ' ', task->rt_priority);
507 0UL, 516 seq_put_decimal_ull(m, ' ', task->policy);
508 task->exit_signal, 517 seq_put_decimal_ull(m, ' ', delayacct_blkio_ticks(task));
509 task_cpu(task), 518 seq_put_decimal_ull(m, ' ', cputime_to_clock_t(gtime));
510 task->rt_priority, 519 seq_put_decimal_ll(m, ' ', cputime_to_clock_t(cgtime));
511 task->policy, 520 seq_put_decimal_ull(m, ' ', (mm && permitted) ? mm->start_data : 0);
512 (unsigned long long)delayacct_blkio_ticks(task), 521 seq_put_decimal_ull(m, ' ', (mm && permitted) ? mm->end_data : 0);
513 cputime_to_clock_t(gtime), 522 seq_put_decimal_ull(m, ' ', (mm && permitted) ? mm->start_brk : 0);
514 cputime_to_clock_t(cgtime), 523 seq_putc(m, '\n');
515 (mm && permitted) ? mm->start_data : 0,
516 (mm && permitted) ? mm->end_data : 0,
517 (mm && permitted) ? mm->start_brk : 0);
518 if (mm) 524 if (mm)
519 mmput(mm); 525 mmput(mm);
520 return 0; 526 return 0;
@@ -542,8 +548,20 @@ int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
542 size = task_statm(mm, &shared, &text, &data, &resident); 548 size = task_statm(mm, &shared, &text, &data, &resident);
543 mmput(mm); 549 mmput(mm);
544 } 550 }
545 seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n", 551 /*
546 size, resident, shared, text, data); 552 * For quick read, open code by putting numbers directly
553 * expected format is
554 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
555 * size, resident, shared, text, data);
556 */
557 seq_put_decimal_ull(m, 0, size);
558 seq_put_decimal_ull(m, ' ', resident);
559 seq_put_decimal_ull(m, ' ', shared);
560 seq_put_decimal_ull(m, ' ', text);
561 seq_put_decimal_ull(m, ' ', 0);
562 seq_put_decimal_ull(m, ' ', data);
563 seq_put_decimal_ull(m, ' ', 0);
564 seq_putc(m, '\n');
547 565
548 return 0; 566 return 0;
549} 567}
diff --git a/fs/proc/base.c b/fs/proc/base.c
index d4548dd49b02..d7d711876b6a 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -81,6 +81,7 @@
81#include <linux/oom.h> 81#include <linux/oom.h>
82#include <linux/elf.h> 82#include <linux/elf.h>
83#include <linux/pid_namespace.h> 83#include <linux/pid_namespace.h>
84#include <linux/user_namespace.h>
84#include <linux/fs_struct.h> 85#include <linux/fs_struct.h>
85#include <linux/slab.h> 86#include <linux/slab.h>
86#include <linux/flex_array.h> 87#include <linux/flex_array.h>
@@ -410,12 +411,13 @@ static const struct file_operations proc_lstats_operations = {
410 411
411static int proc_oom_score(struct task_struct *task, char *buffer) 412static int proc_oom_score(struct task_struct *task, char *buffer)
412{ 413{
414 unsigned long totalpages = totalram_pages + total_swap_pages;
413 unsigned long points = 0; 415 unsigned long points = 0;
414 416
415 read_lock(&tasklist_lock); 417 read_lock(&tasklist_lock);
416 if (pid_alive(task)) 418 if (pid_alive(task))
417 points = oom_badness(task, NULL, NULL, 419 points = oom_badness(task, NULL, NULL, totalpages) *
418 totalram_pages + total_swap_pages); 420 1000 / totalpages;
419 read_unlock(&tasklist_lock); 421 read_unlock(&tasklist_lock);
420 return sprintf(buffer, "%lu\n", points); 422 return sprintf(buffer, "%lu\n", points);
421} 423}
@@ -1310,8 +1312,7 @@ sched_autogroup_write(struct file *file, const char __user *buf,
1310 if (!p) 1312 if (!p)
1311 return -ESRCH; 1313 return -ESRCH;
1312 1314
1313 err = nice; 1315 err = proc_sched_autogroup_set_nice(p, nice);
1314 err = proc_sched_autogroup_set_nice(p, &err);
1315 if (err) 1316 if (err)
1316 count = err; 1317 count = err;
1317 1318
@@ -1562,8 +1563,8 @@ int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1562 generic_fillattr(inode, stat); 1563 generic_fillattr(inode, stat);
1563 1564
1564 rcu_read_lock(); 1565 rcu_read_lock();
1565 stat->uid = 0; 1566 stat->uid = GLOBAL_ROOT_UID;
1566 stat->gid = 0; 1567 stat->gid = GLOBAL_ROOT_GID;
1567 task = pid_task(proc_pid(inode), PIDTYPE_PID); 1568 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1568 if (task) { 1569 if (task) {
1569 if (!has_pid_permissions(pid, task, 2)) { 1570 if (!has_pid_permissions(pid, task, 2)) {
@@ -1623,8 +1624,8 @@ int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1623 inode->i_gid = cred->egid; 1624 inode->i_gid = cred->egid;
1624 rcu_read_unlock(); 1625 rcu_read_unlock();
1625 } else { 1626 } else {
1626 inode->i_uid = 0; 1627 inode->i_uid = GLOBAL_ROOT_UID;
1627 inode->i_gid = 0; 1628 inode->i_gid = GLOBAL_ROOT_GID;
1628 } 1629 }
1629 inode->i_mode &= ~(S_ISUID | S_ISGID); 1630 inode->i_mode &= ~(S_ISUID | S_ISGID);
1630 security_task_to_inode(task, inode); 1631 security_task_to_inode(task, inode);
@@ -1754,7 +1755,7 @@ static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1754 1755
1755 fdt = files_fdtable(files); 1756 fdt = files_fdtable(files);
1756 f_flags = file->f_flags & ~O_CLOEXEC; 1757 f_flags = file->f_flags & ~O_CLOEXEC;
1757 if (FD_ISSET(fd, fdt->close_on_exec)) 1758 if (close_on_exec(fd, fdt))
1758 f_flags |= O_CLOEXEC; 1759 f_flags |= O_CLOEXEC;
1759 1760
1760 if (path) { 1761 if (path) {
@@ -1800,10 +1801,15 @@ static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1800 if (task) { 1801 if (task) {
1801 files = get_files_struct(task); 1802 files = get_files_struct(task);
1802 if (files) { 1803 if (files) {
1804 struct file *file;
1803 rcu_read_lock(); 1805 rcu_read_lock();
1804 if (fcheck_files(files, fd)) { 1806 file = fcheck_files(files, fd);
1807 if (file) {
1808 unsigned i_mode, f_mode = file->f_mode;
1809
1805 rcu_read_unlock(); 1810 rcu_read_unlock();
1806 put_files_struct(files); 1811 put_files_struct(files);
1812
1807 if (task_dumpable(task)) { 1813 if (task_dumpable(task)) {
1808 rcu_read_lock(); 1814 rcu_read_lock();
1809 cred = __task_cred(task); 1815 cred = __task_cred(task);
@@ -1811,10 +1817,17 @@ static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1811 inode->i_gid = cred->egid; 1817 inode->i_gid = cred->egid;
1812 rcu_read_unlock(); 1818 rcu_read_unlock();
1813 } else { 1819 } else {
1814 inode->i_uid = 0; 1820 inode->i_uid = GLOBAL_ROOT_UID;
1815 inode->i_gid = 0; 1821 inode->i_gid = GLOBAL_ROOT_GID;
1816 } 1822 }
1817 inode->i_mode &= ~(S_ISUID | S_ISGID); 1823
1824 i_mode = S_IFLNK;
1825 if (f_mode & FMODE_READ)
1826 i_mode |= S_IRUSR | S_IXUSR;
1827 if (f_mode & FMODE_WRITE)
1828 i_mode |= S_IWUSR | S_IXUSR;
1829 inode->i_mode = i_mode;
1830
1818 security_task_to_inode(task, inode); 1831 security_task_to_inode(task, inode);
1819 put_task_struct(task); 1832 put_task_struct(task);
1820 return 1; 1833 return 1;
@@ -1838,8 +1851,6 @@ static struct dentry *proc_fd_instantiate(struct inode *dir,
1838 struct dentry *dentry, struct task_struct *task, const void *ptr) 1851 struct dentry *dentry, struct task_struct *task, const void *ptr)
1839{ 1852{
1840 unsigned fd = *(const unsigned *)ptr; 1853 unsigned fd = *(const unsigned *)ptr;
1841 struct file *file;
1842 struct files_struct *files;
1843 struct inode *inode; 1854 struct inode *inode;
1844 struct proc_inode *ei; 1855 struct proc_inode *ei;
1845 struct dentry *error = ERR_PTR(-ENOENT); 1856 struct dentry *error = ERR_PTR(-ENOENT);
@@ -1849,25 +1860,6 @@ static struct dentry *proc_fd_instantiate(struct inode *dir,
1849 goto out; 1860 goto out;
1850 ei = PROC_I(inode); 1861 ei = PROC_I(inode);
1851 ei->fd = fd; 1862 ei->fd = fd;
1852 files = get_files_struct(task);
1853 if (!files)
1854 goto out_iput;
1855 inode->i_mode = S_IFLNK;
1856
1857 /*
1858 * We are not taking a ref to the file structure, so we must
1859 * hold ->file_lock.
1860 */
1861 spin_lock(&files->file_lock);
1862 file = fcheck_files(files, fd);
1863 if (!file)
1864 goto out_unlock;
1865 if (file->f_mode & FMODE_READ)
1866 inode->i_mode |= S_IRUSR | S_IXUSR;
1867 if (file->f_mode & FMODE_WRITE)
1868 inode->i_mode |= S_IWUSR | S_IXUSR;
1869 spin_unlock(&files->file_lock);
1870 put_files_struct(files);
1871 1863
1872 inode->i_op = &proc_pid_link_inode_operations; 1864 inode->i_op = &proc_pid_link_inode_operations;
1873 inode->i_size = 64; 1865 inode->i_size = 64;
@@ -1880,12 +1872,6 @@ static struct dentry *proc_fd_instantiate(struct inode *dir,
1880 1872
1881 out: 1873 out:
1882 return error; 1874 return error;
1883out_unlock:
1884 spin_unlock(&files->file_lock);
1885 put_files_struct(files);
1886out_iput:
1887 iput(inode);
1888 goto out;
1889} 1875}
1890 1876
1891static struct dentry *proc_lookupfd_common(struct inode *dir, 1877static struct dentry *proc_lookupfd_common(struct inode *dir,
@@ -2061,8 +2047,8 @@ static int map_files_d_revalidate(struct dentry *dentry, struct nameidata *nd)
2061 inode->i_gid = cred->egid; 2047 inode->i_gid = cred->egid;
2062 rcu_read_unlock(); 2048 rcu_read_unlock();
2063 } else { 2049 } else {
2064 inode->i_uid = 0; 2050 inode->i_uid = GLOBAL_ROOT_UID;
2065 inode->i_gid = 0; 2051 inode->i_gid = GLOBAL_ROOT_GID;
2066 } 2052 }
2067 security_task_to_inode(task, inode); 2053 security_task_to_inode(task, inode);
2068 status = 1; 2054 status = 1;
@@ -2178,16 +2164,16 @@ static struct dentry *proc_map_files_lookup(struct inode *dir,
2178 goto out; 2164 goto out;
2179 2165
2180 result = ERR_PTR(-EACCES); 2166 result = ERR_PTR(-EACCES);
2181 if (lock_trace(task)) 2167 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2182 goto out_put_task; 2168 goto out_put_task;
2183 2169
2184 result = ERR_PTR(-ENOENT); 2170 result = ERR_PTR(-ENOENT);
2185 if (dname_to_vma_addr(dentry, &vm_start, &vm_end)) 2171 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2186 goto out_unlock; 2172 goto out_put_task;
2187 2173
2188 mm = get_task_mm(task); 2174 mm = get_task_mm(task);
2189 if (!mm) 2175 if (!mm)
2190 goto out_unlock; 2176 goto out_put_task;
2191 2177
2192 down_read(&mm->mmap_sem); 2178 down_read(&mm->mmap_sem);
2193 vma = find_exact_vma(mm, vm_start, vm_end); 2179 vma = find_exact_vma(mm, vm_start, vm_end);
@@ -2199,8 +2185,6 @@ static struct dentry *proc_map_files_lookup(struct inode *dir,
2199out_no_vma: 2185out_no_vma:
2200 up_read(&mm->mmap_sem); 2186 up_read(&mm->mmap_sem);
2201 mmput(mm); 2187 mmput(mm);
2202out_unlock:
2203 unlock_trace(task);
2204out_put_task: 2188out_put_task:
2205 put_task_struct(task); 2189 put_task_struct(task);
2206out: 2190out:
@@ -2234,7 +2218,7 @@ proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
2234 goto out; 2218 goto out;
2235 2219
2236 ret = -EACCES; 2220 ret = -EACCES;
2237 if (lock_trace(task)) 2221 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2238 goto out_put_task; 2222 goto out_put_task;
2239 2223
2240 ret = 0; 2224 ret = 0;
@@ -2242,12 +2226,12 @@ proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
2242 case 0: 2226 case 0:
2243 ino = inode->i_ino; 2227 ino = inode->i_ino;
2244 if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0) 2228 if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
2245 goto out_unlock; 2229 goto out_put_task;
2246 filp->f_pos++; 2230 filp->f_pos++;
2247 case 1: 2231 case 1:
2248 ino = parent_ino(dentry); 2232 ino = parent_ino(dentry);
2249 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0) 2233 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2250 goto out_unlock; 2234 goto out_put_task;
2251 filp->f_pos++; 2235 filp->f_pos++;
2252 default: 2236 default:
2253 { 2237 {
@@ -2258,7 +2242,7 @@ proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
2258 2242
2259 mm = get_task_mm(task); 2243 mm = get_task_mm(task);
2260 if (!mm) 2244 if (!mm)
2261 goto out_unlock; 2245 goto out_put_task;
2262 down_read(&mm->mmap_sem); 2246 down_read(&mm->mmap_sem);
2263 2247
2264 nr_files = 0; 2248 nr_files = 0;
@@ -2288,7 +2272,7 @@ proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
2288 flex_array_free(fa); 2272 flex_array_free(fa);
2289 up_read(&mm->mmap_sem); 2273 up_read(&mm->mmap_sem);
2290 mmput(mm); 2274 mmput(mm);
2291 goto out_unlock; 2275 goto out_put_task;
2292 } 2276 }
2293 for (i = 0, vma = mm->mmap, pos = 2; vma; 2277 for (i = 0, vma = mm->mmap, pos = 2; vma;
2294 vma = vma->vm_next) { 2278 vma = vma->vm_next) {
@@ -2333,8 +2317,6 @@ proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
2333 } 2317 }
2334 } 2318 }
2335 2319
2336out_unlock:
2337 unlock_trace(task);
2338out_put_task: 2320out_put_task:
2339 put_task_struct(task); 2321 put_task_struct(task);
2340out: 2322out:
@@ -2944,6 +2926,74 @@ static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2944} 2926}
2945#endif /* CONFIG_TASK_IO_ACCOUNTING */ 2927#endif /* CONFIG_TASK_IO_ACCOUNTING */
2946 2928
2929#ifdef CONFIG_USER_NS
2930static int proc_id_map_open(struct inode *inode, struct file *file,
2931 struct seq_operations *seq_ops)
2932{
2933 struct user_namespace *ns = NULL;
2934 struct task_struct *task;
2935 struct seq_file *seq;
2936 int ret = -EINVAL;
2937
2938 task = get_proc_task(inode);
2939 if (task) {
2940 rcu_read_lock();
2941 ns = get_user_ns(task_cred_xxx(task, user_ns));
2942 rcu_read_unlock();
2943 put_task_struct(task);
2944 }
2945 if (!ns)
2946 goto err;
2947
2948 ret = seq_open(file, seq_ops);
2949 if (ret)
2950 goto err_put_ns;
2951
2952 seq = file->private_data;
2953 seq->private = ns;
2954
2955 return 0;
2956err_put_ns:
2957 put_user_ns(ns);
2958err:
2959 return ret;
2960}
2961
2962static int proc_id_map_release(struct inode *inode, struct file *file)
2963{
2964 struct seq_file *seq = file->private_data;
2965 struct user_namespace *ns = seq->private;
2966 put_user_ns(ns);
2967 return seq_release(inode, file);
2968}
2969
2970static int proc_uid_map_open(struct inode *inode, struct file *file)
2971{
2972 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2973}
2974
2975static int proc_gid_map_open(struct inode *inode, struct file *file)
2976{
2977 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2978}
2979
2980static const struct file_operations proc_uid_map_operations = {
2981 .open = proc_uid_map_open,
2982 .write = proc_uid_map_write,
2983 .read = seq_read,
2984 .llseek = seq_lseek,
2985 .release = proc_id_map_release,
2986};
2987
2988static const struct file_operations proc_gid_map_operations = {
2989 .open = proc_gid_map_open,
2990 .write = proc_gid_map_write,
2991 .read = seq_read,
2992 .llseek = seq_lseek,
2993 .release = proc_id_map_release,
2994};
2995#endif /* CONFIG_USER_NS */
2996
2947static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns, 2997static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2948 struct pid *pid, struct task_struct *task) 2998 struct pid *pid, struct task_struct *task)
2949{ 2999{
@@ -2990,9 +3040,9 @@ static const struct pid_entry tgid_base_stuff[] = {
2990 INF("cmdline", S_IRUGO, proc_pid_cmdline), 3040 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2991 ONE("stat", S_IRUGO, proc_tgid_stat), 3041 ONE("stat", S_IRUGO, proc_tgid_stat),
2992 ONE("statm", S_IRUGO, proc_pid_statm), 3042 ONE("statm", S_IRUGO, proc_pid_statm),
2993 REG("maps", S_IRUGO, proc_maps_operations), 3043 REG("maps", S_IRUGO, proc_pid_maps_operations),
2994#ifdef CONFIG_NUMA 3044#ifdef CONFIG_NUMA
2995 REG("numa_maps", S_IRUGO, proc_numa_maps_operations), 3045 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
2996#endif 3046#endif
2997 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), 3047 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2998 LNK("cwd", proc_cwd_link), 3048 LNK("cwd", proc_cwd_link),
@@ -3003,7 +3053,7 @@ static const struct pid_entry tgid_base_stuff[] = {
3003 REG("mountstats", S_IRUSR, proc_mountstats_operations), 3053 REG("mountstats", S_IRUSR, proc_mountstats_operations),
3004#ifdef CONFIG_PROC_PAGE_MONITOR 3054#ifdef CONFIG_PROC_PAGE_MONITOR
3005 REG("clear_refs", S_IWUSR, proc_clear_refs_operations), 3055 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3006 REG("smaps", S_IRUGO, proc_smaps_operations), 3056 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
3007 REG("pagemap", S_IRUGO, proc_pagemap_operations), 3057 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3008#endif 3058#endif
3009#ifdef CONFIG_SECURITY 3059#ifdef CONFIG_SECURITY
@@ -3046,6 +3096,10 @@ static const struct pid_entry tgid_base_stuff[] = {
3046#ifdef CONFIG_HARDWALL 3096#ifdef CONFIG_HARDWALL
3047 INF("hardwall", S_IRUGO, proc_pid_hardwall), 3097 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3048#endif 3098#endif
3099#ifdef CONFIG_USER_NS
3100 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3101 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3102#endif
3049}; 3103};
3050 3104
3051static int proc_tgid_base_readdir(struct file * filp, 3105static int proc_tgid_base_readdir(struct file * filp,
@@ -3349,9 +3403,9 @@ static const struct pid_entry tid_base_stuff[] = {
3349 INF("cmdline", S_IRUGO, proc_pid_cmdline), 3403 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3350 ONE("stat", S_IRUGO, proc_tid_stat), 3404 ONE("stat", S_IRUGO, proc_tid_stat),
3351 ONE("statm", S_IRUGO, proc_pid_statm), 3405 ONE("statm", S_IRUGO, proc_pid_statm),
3352 REG("maps", S_IRUGO, proc_maps_operations), 3406 REG("maps", S_IRUGO, proc_tid_maps_operations),
3353#ifdef CONFIG_NUMA 3407#ifdef CONFIG_NUMA
3354 REG("numa_maps", S_IRUGO, proc_numa_maps_operations), 3408 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
3355#endif 3409#endif
3356 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), 3410 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3357 LNK("cwd", proc_cwd_link), 3411 LNK("cwd", proc_cwd_link),
@@ -3361,7 +3415,7 @@ static const struct pid_entry tid_base_stuff[] = {
3361 REG("mountinfo", S_IRUGO, proc_mountinfo_operations), 3415 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3362#ifdef CONFIG_PROC_PAGE_MONITOR 3416#ifdef CONFIG_PROC_PAGE_MONITOR
3363 REG("clear_refs", S_IWUSR, proc_clear_refs_operations), 3417 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3364 REG("smaps", S_IRUGO, proc_smaps_operations), 3418 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
3365 REG("pagemap", S_IRUGO, proc_pagemap_operations), 3419 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3366#endif 3420#endif
3367#ifdef CONFIG_SECURITY 3421#ifdef CONFIG_SECURITY
@@ -3401,6 +3455,10 @@ static const struct pid_entry tid_base_stuff[] = {
3401#ifdef CONFIG_HARDWALL 3455#ifdef CONFIG_HARDWALL
3402 INF("hardwall", S_IRUGO, proc_pid_hardwall), 3456 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3403#endif 3457#endif
3458#ifdef CONFIG_USER_NS
3459 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3460 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3461#endif
3404}; 3462};
3405 3463
3406static int proc_tid_base_readdir(struct file * filp, 3464static int proc_tid_base_readdir(struct file * filp,
diff --git a/fs/proc/inode.c b/fs/proc/inode.c
index 84fd3235a590..7ac817b64a71 100644
--- a/fs/proc/inode.c
+++ b/fs/proc/inode.c
@@ -22,7 +22,6 @@
22#include <linux/slab.h> 22#include <linux/slab.h>
23#include <linux/mount.h> 23#include <linux/mount.h>
24 24
25#include <asm/system.h>
26#include <asm/uaccess.h> 25#include <asm/uaccess.h>
27 26
28#include "internal.h" 27#include "internal.h"
@@ -34,7 +33,7 @@ static void proc_evict_inode(struct inode *inode)
34 const struct proc_ns_operations *ns_ops; 33 const struct proc_ns_operations *ns_ops;
35 34
36 truncate_inode_pages(&inode->i_data, 0); 35 truncate_inode_pages(&inode->i_data, 0);
37 end_writeback(inode); 36 clear_inode(inode);
38 37
39 /* Stop tracking associated processes */ 38 /* Stop tracking associated processes */
40 put_pid(PROC_I(inode)->pid); 39 put_pid(PROC_I(inode)->pid);
@@ -109,8 +108,8 @@ static int proc_show_options(struct seq_file *seq, struct dentry *root)
109 struct super_block *sb = root->d_sb; 108 struct super_block *sb = root->d_sb;
110 struct pid_namespace *pid = sb->s_fs_info; 109 struct pid_namespace *pid = sb->s_fs_info;
111 110
112 if (pid->pid_gid) 111 if (!gid_eq(pid->pid_gid, GLOBAL_ROOT_GID))
113 seq_printf(seq, ",gid=%lu", (unsigned long)pid->pid_gid); 112 seq_printf(seq, ",gid=%u", from_kgid_munged(&init_user_ns, pid->pid_gid));
114 if (pid->hide_pid != 0) 113 if (pid->hide_pid != 0)
115 seq_printf(seq, ",hidepid=%u", pid->hide_pid); 114 seq_printf(seq, ",hidepid=%u", pid->hide_pid);
116 115
@@ -486,8 +485,6 @@ struct inode *proc_get_inode(struct super_block *sb, struct proc_dir_entry *de)
486 485
487int proc_fill_super(struct super_block *s) 486int proc_fill_super(struct super_block *s)
488{ 487{
489 struct inode * root_inode;
490
491 s->s_flags |= MS_NODIRATIME | MS_NOSUID | MS_NOEXEC; 488 s->s_flags |= MS_NODIRATIME | MS_NOSUID | MS_NOEXEC;
492 s->s_blocksize = 1024; 489 s->s_blocksize = 1024;
493 s->s_blocksize_bits = 10; 490 s->s_blocksize_bits = 10;
@@ -496,19 +493,11 @@ int proc_fill_super(struct super_block *s)
496 s->s_time_gran = 1; 493 s->s_time_gran = 1;
497 494
498 pde_get(&proc_root); 495 pde_get(&proc_root);
499 root_inode = proc_get_inode(s, &proc_root); 496 s->s_root = d_make_root(proc_get_inode(s, &proc_root));
500 if (!root_inode) 497 if (s->s_root)
501 goto out_no_root; 498 return 0;
502 root_inode->i_uid = 0;
503 root_inode->i_gid = 0;
504 s->s_root = d_alloc_root(root_inode);
505 if (!s->s_root)
506 goto out_no_root;
507 return 0;
508 499
509out_no_root:
510 printk("proc_read_super: get root inode failed\n"); 500 printk("proc_read_super: get root inode failed\n");
511 iput(root_inode);
512 pde_put(&proc_root); 501 pde_put(&proc_root);
513 return -ENOMEM; 502 return -ENOMEM;
514} 503}
diff --git a/fs/proc/internal.h b/fs/proc/internal.h
index 292577531ad1..5f79bb8b4c60 100644
--- a/fs/proc/internal.h
+++ b/fs/proc/internal.h
@@ -10,12 +10,15 @@
10 */ 10 */
11 11
12#include <linux/proc_fs.h> 12#include <linux/proc_fs.h>
13struct ctl_table_header;
13 14
14extern struct proc_dir_entry proc_root; 15extern struct proc_dir_entry proc_root;
15#ifdef CONFIG_PROC_SYSCTL 16#ifdef CONFIG_PROC_SYSCTL
16extern int proc_sys_init(void); 17extern int proc_sys_init(void);
18extern void sysctl_head_put(struct ctl_table_header *head);
17#else 19#else
18static inline void proc_sys_init(void) { } 20static inline void proc_sys_init(void) { }
21static inline void sysctl_head_put(struct ctl_table_header *head) { }
19#endif 22#endif
20#ifdef CONFIG_NET 23#ifdef CONFIG_NET
21extern int proc_net_init(void); 24extern int proc_net_init(void);
@@ -53,9 +56,12 @@ extern int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
53 struct pid *pid, struct task_struct *task); 56 struct pid *pid, struct task_struct *task);
54extern loff_t mem_lseek(struct file *file, loff_t offset, int orig); 57extern loff_t mem_lseek(struct file *file, loff_t offset, int orig);
55 58
56extern const struct file_operations proc_maps_operations; 59extern const struct file_operations proc_pid_maps_operations;
57extern const struct file_operations proc_numa_maps_operations; 60extern const struct file_operations proc_tid_maps_operations;
58extern const struct file_operations proc_smaps_operations; 61extern const struct file_operations proc_pid_numa_maps_operations;
62extern const struct file_operations proc_tid_numa_maps_operations;
63extern const struct file_operations proc_pid_smaps_operations;
64extern const struct file_operations proc_tid_smaps_operations;
59extern const struct file_operations proc_clear_refs_operations; 65extern const struct file_operations proc_clear_refs_operations;
60extern const struct file_operations proc_pagemap_operations; 66extern const struct file_operations proc_pagemap_operations;
61extern const struct file_operations proc_net_operations; 67extern const struct file_operations proc_net_operations;
diff --git a/fs/proc/kcore.c b/fs/proc/kcore.c
index d245cb23dd72..86c67eee439f 100644
--- a/fs/proc/kcore.c
+++ b/fs/proc/kcore.c
@@ -157,7 +157,8 @@ static int kcore_update_ram(void)
157 157
158#ifdef CONFIG_SPARSEMEM_VMEMMAP 158#ifdef CONFIG_SPARSEMEM_VMEMMAP
159/* calculate vmemmap's address from given system ram pfn and register it */ 159/* calculate vmemmap's address from given system ram pfn and register it */
160int get_sparsemem_vmemmap_info(struct kcore_list *ent, struct list_head *head) 160static int
161get_sparsemem_vmemmap_info(struct kcore_list *ent, struct list_head *head)
161{ 162{
162 unsigned long pfn = __pa(ent->addr) >> PAGE_SHIFT; 163 unsigned long pfn = __pa(ent->addr) >> PAGE_SHIFT;
163 unsigned long nr_pages = ent->size >> PAGE_SHIFT; 164 unsigned long nr_pages = ent->size >> PAGE_SHIFT;
@@ -189,7 +190,8 @@ int get_sparsemem_vmemmap_info(struct kcore_list *ent, struct list_head *head)
189 190
190} 191}
191#else 192#else
192int get_sparsemem_vmemmap_info(struct kcore_list *ent, struct list_head *head) 193static int
194get_sparsemem_vmemmap_info(struct kcore_list *ent, struct list_head *head)
193{ 195{
194 return 1; 196 return 1;
195} 197}
@@ -513,7 +515,7 @@ read_kcore(struct file *file, char __user *buffer, size_t buflen, loff_t *fpos)
513 515
514 n = copy_to_user(buffer, (char *)start, tsz); 516 n = copy_to_user(buffer, (char *)start, tsz);
515 /* 517 /*
516 * We cannot distingush between fault on source 518 * We cannot distinguish between fault on source
517 * and fault on destination. When this happens 519 * and fault on destination. When this happens
518 * we clear too and hope it will trigger the 520 * we clear too and hope it will trigger the
519 * EFAULT again. 521 * EFAULT again.
diff --git a/fs/proc/namespaces.c b/fs/proc/namespaces.c
index 27da860115c6..0d9e23a39e49 100644
--- a/fs/proc/namespaces.c
+++ b/fs/proc/namespaces.c
@@ -53,7 +53,7 @@ static struct dentry *proc_ns_instantiate(struct inode *dir,
53 ei->ns_ops = ns_ops; 53 ei->ns_ops = ns_ops;
54 ei->ns = ns; 54 ei->ns = ns;
55 55
56 dentry->d_op = &pid_dentry_operations; 56 d_set_d_op(dentry, &pid_dentry_operations);
57 d_add(dentry, inode); 57 d_add(dentry, inode);
58 /* Close the race of the process dying before we return the dentry */ 58 /* Close the race of the process dying before we return the dentry */
59 if (pid_revalidate(dentry, NULL)) 59 if (pid_revalidate(dentry, NULL))
@@ -156,15 +156,15 @@ static struct dentry *proc_ns_dir_lookup(struct inode *dir,
156 if (!ptrace_may_access(task, PTRACE_MODE_READ)) 156 if (!ptrace_may_access(task, PTRACE_MODE_READ))
157 goto out; 157 goto out;
158 158
159 last = &ns_entries[ARRAY_SIZE(ns_entries) - 1]; 159 last = &ns_entries[ARRAY_SIZE(ns_entries)];
160 for (entry = ns_entries; entry <= last; entry++) { 160 for (entry = ns_entries; entry < last; entry++) {
161 if (strlen((*entry)->name) != len) 161 if (strlen((*entry)->name) != len)
162 continue; 162 continue;
163 if (!memcmp(dentry->d_name.name, (*entry)->name, len)) 163 if (!memcmp(dentry->d_name.name, (*entry)->name, len))
164 break; 164 break;
165 } 165 }
166 error = ERR_PTR(-ENOENT); 166 error = ERR_PTR(-ENOENT);
167 if (entry > last) 167 if (entry == last)
168 goto out; 168 goto out;
169 169
170 error = proc_ns_instantiate(dir, dentry, task, *entry); 170 error = proc_ns_instantiate(dir, dentry, task, *entry);
diff --git a/fs/proc/page.c b/fs/proc/page.c
index 6d8e6a9e93ab..7fcd0d60a968 100644
--- a/fs/proc/page.c
+++ b/fs/proc/page.c
@@ -115,6 +115,8 @@ u64 stable_page_flags(struct page *page)
115 u |= 1 << KPF_COMPOUND_TAIL; 115 u |= 1 << KPF_COMPOUND_TAIL;
116 if (PageHuge(page)) 116 if (PageHuge(page))
117 u |= 1 << KPF_HUGE; 117 u |= 1 << KPF_HUGE;
118 else if (PageTransCompound(page))
119 u |= 1 << KPF_THP;
118 120
119 /* 121 /*
120 * Caveats on high order pages: page->_count will only be set 122 * Caveats on high order pages: page->_count will only be set
diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index a6b62173d4c3..3476bca8f7af 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -6,7 +6,10 @@
6#include <linux/poll.h> 6#include <linux/poll.h>
7#include <linux/proc_fs.h> 7#include <linux/proc_fs.h>
8#include <linux/security.h> 8#include <linux/security.h>
9#include <linux/sched.h>
9#include <linux/namei.h> 10#include <linux/namei.h>
11#include <linux/mm.h>
12#include <linux/module.h>
10#include "internal.h" 13#include "internal.h"
11 14
12static const struct dentry_operations proc_sys_dentry_operations; 15static const struct dentry_operations proc_sys_dentry_operations;
@@ -24,6 +27,371 @@ void proc_sys_poll_notify(struct ctl_table_poll *poll)
24 wake_up_interruptible(&poll->wait); 27 wake_up_interruptible(&poll->wait);
25} 28}
26 29
30static struct ctl_table root_table[] = {
31 {
32 .procname = "",
33 .mode = S_IFDIR|S_IRUGO|S_IXUGO,
34 },
35 { }
36};
37static struct ctl_table_root sysctl_table_root = {
38 .default_set.dir.header = {
39 {{.count = 1,
40 .nreg = 1,
41 .ctl_table = root_table }},
42 .ctl_table_arg = root_table,
43 .root = &sysctl_table_root,
44 .set = &sysctl_table_root.default_set,
45 },
46};
47
48static DEFINE_SPINLOCK(sysctl_lock);
49
50static void drop_sysctl_table(struct ctl_table_header *header);
51static int sysctl_follow_link(struct ctl_table_header **phead,
52 struct ctl_table **pentry, struct nsproxy *namespaces);
53static int insert_links(struct ctl_table_header *head);
54static void put_links(struct ctl_table_header *header);
55
56static void sysctl_print_dir(struct ctl_dir *dir)
57{
58 if (dir->header.parent)
59 sysctl_print_dir(dir->header.parent);
60 printk(KERN_CONT "%s/", dir->header.ctl_table[0].procname);
61}
62
63static int namecmp(const char *name1, int len1, const char *name2, int len2)
64{
65 int minlen;
66 int cmp;
67
68 minlen = len1;
69 if (minlen > len2)
70 minlen = len2;
71
72 cmp = memcmp(name1, name2, minlen);
73 if (cmp == 0)
74 cmp = len1 - len2;
75 return cmp;
76}
77
78/* Called under sysctl_lock */
79static struct ctl_table *find_entry(struct ctl_table_header **phead,
80 struct ctl_dir *dir, const char *name, int namelen)
81{
82 struct ctl_table_header *head;
83 struct ctl_table *entry;
84 struct rb_node *node = dir->root.rb_node;
85
86 while (node)
87 {
88 struct ctl_node *ctl_node;
89 const char *procname;
90 int cmp;
91
92 ctl_node = rb_entry(node, struct ctl_node, node);
93 head = ctl_node->header;
94 entry = &head->ctl_table[ctl_node - head->node];
95 procname = entry->procname;
96
97 cmp = namecmp(name, namelen, procname, strlen(procname));
98 if (cmp < 0)
99 node = node->rb_left;
100 else if (cmp > 0)
101 node = node->rb_right;
102 else {
103 *phead = head;
104 return entry;
105 }
106 }
107 return NULL;
108}
109
110static int insert_entry(struct ctl_table_header *head, struct ctl_table *entry)
111{
112 struct rb_node *node = &head->node[entry - head->ctl_table].node;
113 struct rb_node **p = &head->parent->root.rb_node;
114 struct rb_node *parent = NULL;
115 const char *name = entry->procname;
116 int namelen = strlen(name);
117
118 while (*p) {
119 struct ctl_table_header *parent_head;
120 struct ctl_table *parent_entry;
121 struct ctl_node *parent_node;
122 const char *parent_name;
123 int cmp;
124
125 parent = *p;
126 parent_node = rb_entry(parent, struct ctl_node, node);
127 parent_head = parent_node->header;
128 parent_entry = &parent_head->ctl_table[parent_node - parent_head->node];
129 parent_name = parent_entry->procname;
130
131 cmp = namecmp(name, namelen, parent_name, strlen(parent_name));
132 if (cmp < 0)
133 p = &(*p)->rb_left;
134 else if (cmp > 0)
135 p = &(*p)->rb_right;
136 else {
137 printk(KERN_ERR "sysctl duplicate entry: ");
138 sysctl_print_dir(head->parent);
139 printk(KERN_CONT "/%s\n", entry->procname);
140 return -EEXIST;
141 }
142 }
143
144 rb_link_node(node, parent, p);
145 return 0;
146}
147
148static void erase_entry(struct ctl_table_header *head, struct ctl_table *entry)
149{
150 struct rb_node *node = &head->node[entry - head->ctl_table].node;
151
152 rb_erase(node, &head->parent->root);
153}
154
155static void init_header(struct ctl_table_header *head,
156 struct ctl_table_root *root, struct ctl_table_set *set,
157 struct ctl_node *node, struct ctl_table *table)
158{
159 head->ctl_table = table;
160 head->ctl_table_arg = table;
161 head->used = 0;
162 head->count = 1;
163 head->nreg = 1;
164 head->unregistering = NULL;
165 head->root = root;
166 head->set = set;
167 head->parent = NULL;
168 head->node = node;
169 if (node) {
170 struct ctl_table *entry;
171 for (entry = table; entry->procname; entry++, node++) {
172 rb_init_node(&node->node);
173 node->header = head;
174 }
175 }
176}
177
178static void erase_header(struct ctl_table_header *head)
179{
180 struct ctl_table *entry;
181 for (entry = head->ctl_table; entry->procname; entry++)
182 erase_entry(head, entry);
183}
184
185static int insert_header(struct ctl_dir *dir, struct ctl_table_header *header)
186{
187 struct ctl_table *entry;
188 int err;
189
190 dir->header.nreg++;
191 header->parent = dir;
192 err = insert_links(header);
193 if (err)
194 goto fail_links;
195 for (entry = header->ctl_table; entry->procname; entry++) {
196 err = insert_entry(header, entry);
197 if (err)
198 goto fail;
199 }
200 return 0;
201fail:
202 erase_header(header);
203 put_links(header);
204fail_links:
205 header->parent = NULL;
206 drop_sysctl_table(&dir->header);
207 return err;
208}
209
210/* called under sysctl_lock */
211static int use_table(struct ctl_table_header *p)
212{
213 if (unlikely(p->unregistering))
214 return 0;
215 p->used++;
216 return 1;
217}
218
219/* called under sysctl_lock */
220static void unuse_table(struct ctl_table_header *p)
221{
222 if (!--p->used)
223 if (unlikely(p->unregistering))
224 complete(p->unregistering);
225}
226
227/* called under sysctl_lock, will reacquire if has to wait */
228static void start_unregistering(struct ctl_table_header *p)
229{
230 /*
231 * if p->used is 0, nobody will ever touch that entry again;
232 * we'll eliminate all paths to it before dropping sysctl_lock
233 */
234 if (unlikely(p->used)) {
235 struct completion wait;
236 init_completion(&wait);
237 p->unregistering = &wait;
238 spin_unlock(&sysctl_lock);
239 wait_for_completion(&wait);
240 spin_lock(&sysctl_lock);
241 } else {
242 /* anything non-NULL; we'll never dereference it */
243 p->unregistering = ERR_PTR(-EINVAL);
244 }
245 /*
246 * do not remove from the list until nobody holds it; walking the
247 * list in do_sysctl() relies on that.
248 */
249 erase_header(p);
250}
251
252static void sysctl_head_get(struct ctl_table_header *head)
253{
254 spin_lock(&sysctl_lock);
255 head->count++;
256 spin_unlock(&sysctl_lock);
257}
258
259void sysctl_head_put(struct ctl_table_header *head)
260{
261 spin_lock(&sysctl_lock);
262 if (!--head->count)
263 kfree_rcu(head, rcu);
264 spin_unlock(&sysctl_lock);
265}
266
267static struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
268{
269 if (!head)
270 BUG();
271 spin_lock(&sysctl_lock);
272 if (!use_table(head))
273 head = ERR_PTR(-ENOENT);
274 spin_unlock(&sysctl_lock);
275 return head;
276}
277
278static void sysctl_head_finish(struct ctl_table_header *head)
279{
280 if (!head)
281 return;
282 spin_lock(&sysctl_lock);
283 unuse_table(head);
284 spin_unlock(&sysctl_lock);
285}
286
287static struct ctl_table_set *
288lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
289{
290 struct ctl_table_set *set = &root->default_set;
291 if (root->lookup)
292 set = root->lookup(root, namespaces);
293 return set;
294}
295
296static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
297 struct ctl_dir *dir,
298 const char *name, int namelen)
299{
300 struct ctl_table_header *head;
301 struct ctl_table *entry;
302
303 spin_lock(&sysctl_lock);
304 entry = find_entry(&head, dir, name, namelen);
305 if (entry && use_table(head))
306 *phead = head;
307 else
308 entry = NULL;
309 spin_unlock(&sysctl_lock);
310 return entry;
311}
312
313static struct ctl_node *first_usable_entry(struct rb_node *node)
314{
315 struct ctl_node *ctl_node;
316
317 for (;node; node = rb_next(node)) {
318 ctl_node = rb_entry(node, struct ctl_node, node);
319 if (use_table(ctl_node->header))
320 return ctl_node;
321 }
322 return NULL;
323}
324
325static void first_entry(struct ctl_dir *dir,
326 struct ctl_table_header **phead, struct ctl_table **pentry)
327{
328 struct ctl_table_header *head = NULL;
329 struct ctl_table *entry = NULL;
330 struct ctl_node *ctl_node;
331
332 spin_lock(&sysctl_lock);
333 ctl_node = first_usable_entry(rb_first(&dir->root));
334 spin_unlock(&sysctl_lock);
335 if (ctl_node) {
336 head = ctl_node->header;
337 entry = &head->ctl_table[ctl_node - head->node];
338 }
339 *phead = head;
340 *pentry = entry;
341}
342
343static void next_entry(struct ctl_table_header **phead, struct ctl_table **pentry)
344{
345 struct ctl_table_header *head = *phead;
346 struct ctl_table *entry = *pentry;
347 struct ctl_node *ctl_node = &head->node[entry - head->ctl_table];
348
349 spin_lock(&sysctl_lock);
350 unuse_table(head);
351
352 ctl_node = first_usable_entry(rb_next(&ctl_node->node));
353 spin_unlock(&sysctl_lock);
354 head = NULL;
355 if (ctl_node) {
356 head = ctl_node->header;
357 entry = &head->ctl_table[ctl_node - head->node];
358 }
359 *phead = head;
360 *pentry = entry;
361}
362
363void register_sysctl_root(struct ctl_table_root *root)
364{
365}
366
367/*
368 * sysctl_perm does NOT grant the superuser all rights automatically, because
369 * some sysctl variables are readonly even to root.
370 */
371
372static int test_perm(int mode, int op)
373{
374 if (uid_eq(current_euid(), GLOBAL_ROOT_UID))
375 mode >>= 6;
376 else if (in_egroup_p(GLOBAL_ROOT_GID))
377 mode >>= 3;
378 if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
379 return 0;
380 return -EACCES;
381}
382
383static int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
384{
385 int mode;
386
387 if (root->permissions)
388 mode = root->permissions(root, current->nsproxy, table);
389 else
390 mode = table->mode;
391
392 return test_perm(mode, op);
393}
394
27static struct inode *proc_sys_make_inode(struct super_block *sb, 395static struct inode *proc_sys_make_inode(struct super_block *sb,
28 struct ctl_table_header *head, struct ctl_table *table) 396 struct ctl_table_header *head, struct ctl_table *table)
29{ 397{
@@ -43,13 +411,12 @@ static struct inode *proc_sys_make_inode(struct super_block *sb,
43 411
44 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; 412 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
45 inode->i_mode = table->mode; 413 inode->i_mode = table->mode;
46 if (!table->child) { 414 if (!S_ISDIR(table->mode)) {
47 inode->i_mode |= S_IFREG; 415 inode->i_mode |= S_IFREG;
48 inode->i_op = &proc_sys_inode_operations; 416 inode->i_op = &proc_sys_inode_operations;
49 inode->i_fop = &proc_sys_file_operations; 417 inode->i_fop = &proc_sys_file_operations;
50 } else { 418 } else {
51 inode->i_mode |= S_IFDIR; 419 inode->i_mode |= S_IFDIR;
52 clear_nlink(inode);
53 inode->i_op = &proc_sys_dir_operations; 420 inode->i_op = &proc_sys_dir_operations;
54 inode->i_fop = &proc_sys_dir_file_operations; 421 inode->i_fop = &proc_sys_dir_file_operations;
55 } 422 }
@@ -57,70 +424,42 @@ out:
57 return inode; 424 return inode;
58} 425}
59 426
60static struct ctl_table *find_in_table(struct ctl_table *p, struct qstr *name)
61{
62 int len;
63 for ( ; p->procname; p++) {
64
65 if (!p->procname)
66 continue;
67
68 len = strlen(p->procname);
69 if (len != name->len)
70 continue;
71
72 if (memcmp(p->procname, name->name, len) != 0)
73 continue;
74
75 /* I have a match */
76 return p;
77 }
78 return NULL;
79}
80
81static struct ctl_table_header *grab_header(struct inode *inode) 427static struct ctl_table_header *grab_header(struct inode *inode)
82{ 428{
83 if (PROC_I(inode)->sysctl) 429 struct ctl_table_header *head = PROC_I(inode)->sysctl;
84 return sysctl_head_grab(PROC_I(inode)->sysctl); 430 if (!head)
85 else 431 head = &sysctl_table_root.default_set.dir.header;
86 return sysctl_head_next(NULL); 432 return sysctl_head_grab(head);
87} 433}
88 434
89static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry, 435static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
90 struct nameidata *nd) 436 struct nameidata *nd)
91{ 437{
92 struct ctl_table_header *head = grab_header(dir); 438 struct ctl_table_header *head = grab_header(dir);
93 struct ctl_table *table = PROC_I(dir)->sysctl_entry;
94 struct ctl_table_header *h = NULL; 439 struct ctl_table_header *h = NULL;
95 struct qstr *name = &dentry->d_name; 440 struct qstr *name = &dentry->d_name;
96 struct ctl_table *p; 441 struct ctl_table *p;
97 struct inode *inode; 442 struct inode *inode;
98 struct dentry *err = ERR_PTR(-ENOENT); 443 struct dentry *err = ERR_PTR(-ENOENT);
444 struct ctl_dir *ctl_dir;
445 int ret;
99 446
100 if (IS_ERR(head)) 447 if (IS_ERR(head))
101 return ERR_CAST(head); 448 return ERR_CAST(head);
102 449
103 if (table && !table->child) { 450 ctl_dir = container_of(head, struct ctl_dir, header);
104 WARN_ON(1);
105 goto out;
106 }
107
108 table = table ? table->child : head->ctl_table;
109
110 p = find_in_table(table, name);
111 if (!p) {
112 for (h = sysctl_head_next(NULL); h; h = sysctl_head_next(h)) {
113 if (h->attached_to != table)
114 continue;
115 p = find_in_table(h->attached_by, name);
116 if (p)
117 break;
118 }
119 }
120 451
452 p = lookup_entry(&h, ctl_dir, name->name, name->len);
121 if (!p) 453 if (!p)
122 goto out; 454 goto out;
123 455
456 if (S_ISLNK(p->mode)) {
457 ret = sysctl_follow_link(&h, &p, current->nsproxy);
458 err = ERR_PTR(ret);
459 if (ret)
460 goto out;
461 }
462
124 err = ERR_PTR(-ENOMEM); 463 err = ERR_PTR(-ENOMEM);
125 inode = proc_sys_make_inode(dir->i_sb, h ? h : head, p); 464 inode = proc_sys_make_inode(dir->i_sb, h ? h : head, p);
126 if (h) 465 if (h)
@@ -188,20 +527,32 @@ static ssize_t proc_sys_write(struct file *filp, const char __user *buf,
188 527
189static int proc_sys_open(struct inode *inode, struct file *filp) 528static int proc_sys_open(struct inode *inode, struct file *filp)
190{ 529{
530 struct ctl_table_header *head = grab_header(inode);
191 struct ctl_table *table = PROC_I(inode)->sysctl_entry; 531 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
192 532
533 /* sysctl was unregistered */
534 if (IS_ERR(head))
535 return PTR_ERR(head);
536
193 if (table->poll) 537 if (table->poll)
194 filp->private_data = proc_sys_poll_event(table->poll); 538 filp->private_data = proc_sys_poll_event(table->poll);
195 539
540 sysctl_head_finish(head);
541
196 return 0; 542 return 0;
197} 543}
198 544
199static unsigned int proc_sys_poll(struct file *filp, poll_table *wait) 545static unsigned int proc_sys_poll(struct file *filp, poll_table *wait)
200{ 546{
201 struct inode *inode = filp->f_path.dentry->d_inode; 547 struct inode *inode = filp->f_path.dentry->d_inode;
548 struct ctl_table_header *head = grab_header(inode);
202 struct ctl_table *table = PROC_I(inode)->sysctl_entry; 549 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
203 unsigned long event = (unsigned long)filp->private_data;
204 unsigned int ret = DEFAULT_POLLMASK; 550 unsigned int ret = DEFAULT_POLLMASK;
551 unsigned long event;
552
553 /* sysctl was unregistered */
554 if (IS_ERR(head))
555 return POLLERR | POLLHUP;
205 556
206 if (!table->proc_handler) 557 if (!table->proc_handler)
207 goto out; 558 goto out;
@@ -209,6 +560,7 @@ static unsigned int proc_sys_poll(struct file *filp, poll_table *wait)
209 if (!table->poll) 560 if (!table->poll)
210 goto out; 561 goto out;
211 562
563 event = (unsigned long)filp->private_data;
212 poll_wait(filp, &table->poll->wait, wait); 564 poll_wait(filp, &table->poll->wait, wait);
213 565
214 if (event != atomic_read(&table->poll->event)) { 566 if (event != atomic_read(&table->poll->event)) {
@@ -217,6 +569,8 @@ static unsigned int proc_sys_poll(struct file *filp, poll_table *wait)
217 } 569 }
218 570
219out: 571out:
572 sysctl_head_finish(head);
573
220 return ret; 574 return ret;
221} 575}
222 576
@@ -258,28 +612,45 @@ static int proc_sys_fill_cache(struct file *filp, void *dirent,
258 return !!filldir(dirent, qname.name, qname.len, filp->f_pos, ino, type); 612 return !!filldir(dirent, qname.name, qname.len, filp->f_pos, ino, type);
259} 613}
260 614
615static int proc_sys_link_fill_cache(struct file *filp, void *dirent,
616 filldir_t filldir,
617 struct ctl_table_header *head,
618 struct ctl_table *table)
619{
620 int err, ret = 0;
621 head = sysctl_head_grab(head);
622
623 if (S_ISLNK(table->mode)) {
624 /* It is not an error if we can not follow the link ignore it */
625 err = sysctl_follow_link(&head, &table, current->nsproxy);
626 if (err)
627 goto out;
628 }
629
630 ret = proc_sys_fill_cache(filp, dirent, filldir, head, table);
631out:
632 sysctl_head_finish(head);
633 return ret;
634}
635
261static int scan(struct ctl_table_header *head, ctl_table *table, 636static int scan(struct ctl_table_header *head, ctl_table *table,
262 unsigned long *pos, struct file *file, 637 unsigned long *pos, struct file *file,
263 void *dirent, filldir_t filldir) 638 void *dirent, filldir_t filldir)
264{ 639{
640 int res;
265 641
266 for (; table->procname; table++, (*pos)++) { 642 if ((*pos)++ < file->f_pos)
267 int res; 643 return 0;
268
269 /* Can't do anything without a proc name */
270 if (!table->procname)
271 continue;
272
273 if (*pos < file->f_pos)
274 continue;
275 644
645 if (unlikely(S_ISLNK(table->mode)))
646 res = proc_sys_link_fill_cache(file, dirent, filldir, head, table);
647 else
276 res = proc_sys_fill_cache(file, dirent, filldir, head, table); 648 res = proc_sys_fill_cache(file, dirent, filldir, head, table);
277 if (res)
278 return res;
279 649
280 file->f_pos = *pos + 1; 650 if (res == 0)
281 } 651 file->f_pos = *pos;
282 return 0; 652
653 return res;
283} 654}
284 655
285static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir) 656static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
@@ -287,20 +658,16 @@ static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
287 struct dentry *dentry = filp->f_path.dentry; 658 struct dentry *dentry = filp->f_path.dentry;
288 struct inode *inode = dentry->d_inode; 659 struct inode *inode = dentry->d_inode;
289 struct ctl_table_header *head = grab_header(inode); 660 struct ctl_table_header *head = grab_header(inode);
290 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
291 struct ctl_table_header *h = NULL; 661 struct ctl_table_header *h = NULL;
662 struct ctl_table *entry;
663 struct ctl_dir *ctl_dir;
292 unsigned long pos; 664 unsigned long pos;
293 int ret = -EINVAL; 665 int ret = -EINVAL;
294 666
295 if (IS_ERR(head)) 667 if (IS_ERR(head))
296 return PTR_ERR(head); 668 return PTR_ERR(head);
297 669
298 if (table && !table->child) { 670 ctl_dir = container_of(head, struct ctl_dir, header);
299 WARN_ON(1);
300 goto out;
301 }
302
303 table = table ? table->child : head->ctl_table;
304 671
305 ret = 0; 672 ret = 0;
306 /* Avoid a switch here: arm builds fail with missing __cmpdi2 */ 673 /* Avoid a switch here: arm builds fail with missing __cmpdi2 */
@@ -318,14 +685,8 @@ static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
318 } 685 }
319 pos = 2; 686 pos = 2;
320 687
321 ret = scan(head, table, &pos, filp, dirent, filldir); 688 for (first_entry(ctl_dir, &h, &entry); h; next_entry(&h, &entry)) {
322 if (ret) 689 ret = scan(h, entry, &pos, filp, dirent, filldir);
323 goto out;
324
325 for (h = sysctl_head_next(NULL); h; h = sysctl_head_next(h)) {
326 if (h->attached_to != table)
327 continue;
328 ret = scan(h, h->attached_by, &pos, filp, dirent, filldir);
329 if (ret) { 690 if (ret) {
330 sysctl_head_finish(h); 691 sysctl_head_finish(h);
331 break; 692 break;
@@ -445,6 +806,21 @@ static int proc_sys_delete(const struct dentry *dentry)
445 return !!PROC_I(dentry->d_inode)->sysctl->unregistering; 806 return !!PROC_I(dentry->d_inode)->sysctl->unregistering;
446} 807}
447 808
809static int sysctl_is_seen(struct ctl_table_header *p)
810{
811 struct ctl_table_set *set = p->set;
812 int res;
813 spin_lock(&sysctl_lock);
814 if (p->unregistering)
815 res = 0;
816 else if (!set->is_seen)
817 res = 1;
818 else
819 res = set->is_seen(set);
820 spin_unlock(&sysctl_lock);
821 return res;
822}
823
448static int proc_sys_compare(const struct dentry *parent, 824static int proc_sys_compare(const struct dentry *parent,
449 const struct inode *pinode, 825 const struct inode *pinode,
450 const struct dentry *dentry, const struct inode *inode, 826 const struct dentry *dentry, const struct inode *inode,
@@ -470,6 +846,753 @@ static const struct dentry_operations proc_sys_dentry_operations = {
470 .d_compare = proc_sys_compare, 846 .d_compare = proc_sys_compare,
471}; 847};
472 848
849static struct ctl_dir *find_subdir(struct ctl_dir *dir,
850 const char *name, int namelen)
851{
852 struct ctl_table_header *head;
853 struct ctl_table *entry;
854
855 entry = find_entry(&head, dir, name, namelen);
856 if (!entry)
857 return ERR_PTR(-ENOENT);
858 if (!S_ISDIR(entry->mode))
859 return ERR_PTR(-ENOTDIR);
860 return container_of(head, struct ctl_dir, header);
861}
862
863static struct ctl_dir *new_dir(struct ctl_table_set *set,
864 const char *name, int namelen)
865{
866 struct ctl_table *table;
867 struct ctl_dir *new;
868 struct ctl_node *node;
869 char *new_name;
870
871 new = kzalloc(sizeof(*new) + sizeof(struct ctl_node) +
872 sizeof(struct ctl_table)*2 + namelen + 1,
873 GFP_KERNEL);
874 if (!new)
875 return NULL;
876
877 node = (struct ctl_node *)(new + 1);
878 table = (struct ctl_table *)(node + 1);
879 new_name = (char *)(table + 2);
880 memcpy(new_name, name, namelen);
881 new_name[namelen] = '\0';
882 table[0].procname = new_name;
883 table[0].mode = S_IFDIR|S_IRUGO|S_IXUGO;
884 init_header(&new->header, set->dir.header.root, set, node, table);
885
886 return new;
887}
888
889/**
890 * get_subdir - find or create a subdir with the specified name.
891 * @dir: Directory to create the subdirectory in
892 * @name: The name of the subdirectory to find or create
893 * @namelen: The length of name
894 *
895 * Takes a directory with an elevated reference count so we know that
896 * if we drop the lock the directory will not go away. Upon success
897 * the reference is moved from @dir to the returned subdirectory.
898 * Upon error an error code is returned and the reference on @dir is
899 * simply dropped.
900 */
901static struct ctl_dir *get_subdir(struct ctl_dir *dir,
902 const char *name, int namelen)
903{
904 struct ctl_table_set *set = dir->header.set;
905 struct ctl_dir *subdir, *new = NULL;
906 int err;
907
908 spin_lock(&sysctl_lock);
909 subdir = find_subdir(dir, name, namelen);
910 if (!IS_ERR(subdir))
911 goto found;
912 if (PTR_ERR(subdir) != -ENOENT)
913 goto failed;
914
915 spin_unlock(&sysctl_lock);
916 new = new_dir(set, name, namelen);
917 spin_lock(&sysctl_lock);
918 subdir = ERR_PTR(-ENOMEM);
919 if (!new)
920 goto failed;
921
922 /* Was the subdir added while we dropped the lock? */
923 subdir = find_subdir(dir, name, namelen);
924 if (!IS_ERR(subdir))
925 goto found;
926 if (PTR_ERR(subdir) != -ENOENT)
927 goto failed;
928
929 /* Nope. Use the our freshly made directory entry. */
930 err = insert_header(dir, &new->header);
931 subdir = ERR_PTR(err);
932 if (err)
933 goto failed;
934 subdir = new;
935found:
936 subdir->header.nreg++;
937failed:
938 if (unlikely(IS_ERR(subdir))) {
939 printk(KERN_ERR "sysctl could not get directory: ");
940 sysctl_print_dir(dir);
941 printk(KERN_CONT "/%*.*s %ld\n",
942 namelen, namelen, name, PTR_ERR(subdir));
943 }
944 drop_sysctl_table(&dir->header);
945 if (new)
946 drop_sysctl_table(&new->header);
947 spin_unlock(&sysctl_lock);
948 return subdir;
949}
950
951static struct ctl_dir *xlate_dir(struct ctl_table_set *set, struct ctl_dir *dir)
952{
953 struct ctl_dir *parent;
954 const char *procname;
955 if (!dir->header.parent)
956 return &set->dir;
957 parent = xlate_dir(set, dir->header.parent);
958 if (IS_ERR(parent))
959 return parent;
960 procname = dir->header.ctl_table[0].procname;
961 return find_subdir(parent, procname, strlen(procname));
962}
963
964static int sysctl_follow_link(struct ctl_table_header **phead,
965 struct ctl_table **pentry, struct nsproxy *namespaces)
966{
967 struct ctl_table_header *head;
968 struct ctl_table_root *root;
969 struct ctl_table_set *set;
970 struct ctl_table *entry;
971 struct ctl_dir *dir;
972 int ret;
973
974 ret = 0;
975 spin_lock(&sysctl_lock);
976 root = (*pentry)->data;
977 set = lookup_header_set(root, namespaces);
978 dir = xlate_dir(set, (*phead)->parent);
979 if (IS_ERR(dir))
980 ret = PTR_ERR(dir);
981 else {
982 const char *procname = (*pentry)->procname;
983 head = NULL;
984 entry = find_entry(&head, dir, procname, strlen(procname));
985 ret = -ENOENT;
986 if (entry && use_table(head)) {
987 unuse_table(*phead);
988 *phead = head;
989 *pentry = entry;
990 ret = 0;
991 }
992 }
993
994 spin_unlock(&sysctl_lock);
995 return ret;
996}
997
998static int sysctl_err(const char *path, struct ctl_table *table, char *fmt, ...)
999{
1000 struct va_format vaf;
1001 va_list args;
1002
1003 va_start(args, fmt);
1004 vaf.fmt = fmt;
1005 vaf.va = &args;
1006
1007 printk(KERN_ERR "sysctl table check failed: %s/%s %pV\n",
1008 path, table->procname, &vaf);
1009
1010 va_end(args);
1011 return -EINVAL;
1012}
1013
1014static int sysctl_check_table(const char *path, struct ctl_table *table)
1015{
1016 int err = 0;
1017 for (; table->procname; table++) {
1018 if (table->child)
1019 err = sysctl_err(path, table, "Not a file");
1020
1021 if ((table->proc_handler == proc_dostring) ||
1022 (table->proc_handler == proc_dointvec) ||
1023 (table->proc_handler == proc_dointvec_minmax) ||
1024 (table->proc_handler == proc_dointvec_jiffies) ||
1025 (table->proc_handler == proc_dointvec_userhz_jiffies) ||
1026 (table->proc_handler == proc_dointvec_ms_jiffies) ||
1027 (table->proc_handler == proc_doulongvec_minmax) ||
1028 (table->proc_handler == proc_doulongvec_ms_jiffies_minmax)) {
1029 if (!table->data)
1030 err = sysctl_err(path, table, "No data");
1031 if (!table->maxlen)
1032 err = sysctl_err(path, table, "No maxlen");
1033 }
1034 if (!table->proc_handler)
1035 err = sysctl_err(path, table, "No proc_handler");
1036
1037 if ((table->mode & (S_IRUGO|S_IWUGO)) != table->mode)
1038 err = sysctl_err(path, table, "bogus .mode 0%o",
1039 table->mode);
1040 }
1041 return err;
1042}
1043
1044static struct ctl_table_header *new_links(struct ctl_dir *dir, struct ctl_table *table,
1045 struct ctl_table_root *link_root)
1046{
1047 struct ctl_table *link_table, *entry, *link;
1048 struct ctl_table_header *links;
1049 struct ctl_node *node;
1050 char *link_name;
1051 int nr_entries, name_bytes;
1052
1053 name_bytes = 0;
1054 nr_entries = 0;
1055 for (entry = table; entry->procname; entry++) {
1056 nr_entries++;
1057 name_bytes += strlen(entry->procname) + 1;
1058 }
1059
1060 links = kzalloc(sizeof(struct ctl_table_header) +
1061 sizeof(struct ctl_node)*nr_entries +
1062 sizeof(struct ctl_table)*(nr_entries + 1) +
1063 name_bytes,
1064 GFP_KERNEL);
1065
1066 if (!links)
1067 return NULL;
1068
1069 node = (struct ctl_node *)(links + 1);
1070 link_table = (struct ctl_table *)(node + nr_entries);
1071 link_name = (char *)&link_table[nr_entries + 1];
1072
1073 for (link = link_table, entry = table; entry->procname; link++, entry++) {
1074 int len = strlen(entry->procname) + 1;
1075 memcpy(link_name, entry->procname, len);
1076 link->procname = link_name;
1077 link->mode = S_IFLNK|S_IRWXUGO;
1078 link->data = link_root;
1079 link_name += len;
1080 }
1081 init_header(links, dir->header.root, dir->header.set, node, link_table);
1082 links->nreg = nr_entries;
1083
1084 return links;
1085}
1086
1087static bool get_links(struct ctl_dir *dir,
1088 struct ctl_table *table, struct ctl_table_root *link_root)
1089{
1090 struct ctl_table_header *head;
1091 struct ctl_table *entry, *link;
1092
1093 /* Are there links available for every entry in table? */
1094 for (entry = table; entry->procname; entry++) {
1095 const char *procname = entry->procname;
1096 link = find_entry(&head, dir, procname, strlen(procname));
1097 if (!link)
1098 return false;
1099 if (S_ISDIR(link->mode) && S_ISDIR(entry->mode))
1100 continue;
1101 if (S_ISLNK(link->mode) && (link->data == link_root))
1102 continue;
1103 return false;
1104 }
1105
1106 /* The checks passed. Increase the registration count on the links */
1107 for (entry = table; entry->procname; entry++) {
1108 const char *procname = entry->procname;
1109 link = find_entry(&head, dir, procname, strlen(procname));
1110 head->nreg++;
1111 }
1112 return true;
1113}
1114
1115static int insert_links(struct ctl_table_header *head)
1116{
1117 struct ctl_table_set *root_set = &sysctl_table_root.default_set;
1118 struct ctl_dir *core_parent = NULL;
1119 struct ctl_table_header *links;
1120 int err;
1121
1122 if (head->set == root_set)
1123 return 0;
1124
1125 core_parent = xlate_dir(root_set, head->parent);
1126 if (IS_ERR(core_parent))
1127 return 0;
1128
1129 if (get_links(core_parent, head->ctl_table, head->root))
1130 return 0;
1131
1132 core_parent->header.nreg++;
1133 spin_unlock(&sysctl_lock);
1134
1135 links = new_links(core_parent, head->ctl_table, head->root);
1136
1137 spin_lock(&sysctl_lock);
1138 err = -ENOMEM;
1139 if (!links)
1140 goto out;
1141
1142 err = 0;
1143 if (get_links(core_parent, head->ctl_table, head->root)) {
1144 kfree(links);
1145 goto out;
1146 }
1147
1148 err = insert_header(core_parent, links);
1149 if (err)
1150 kfree(links);
1151out:
1152 drop_sysctl_table(&core_parent->header);
1153 return err;
1154}
1155
1156/**
1157 * __register_sysctl_table - register a leaf sysctl table
1158 * @set: Sysctl tree to register on
1159 * @path: The path to the directory the sysctl table is in.
1160 * @table: the top-level table structure
1161 *
1162 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1163 * array. A completely 0 filled entry terminates the table.
1164 *
1165 * The members of the &struct ctl_table structure are used as follows:
1166 *
1167 * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1168 * enter a sysctl file
1169 *
1170 * data - a pointer to data for use by proc_handler
1171 *
1172 * maxlen - the maximum size in bytes of the data
1173 *
1174 * mode - the file permissions for the /proc/sys file
1175 *
1176 * child - must be %NULL.
1177 *
1178 * proc_handler - the text handler routine (described below)
1179 *
1180 * extra1, extra2 - extra pointers usable by the proc handler routines
1181 *
1182 * Leaf nodes in the sysctl tree will be represented by a single file
1183 * under /proc; non-leaf nodes will be represented by directories.
1184 *
1185 * There must be a proc_handler routine for any terminal nodes.
1186 * Several default handlers are available to cover common cases -
1187 *
1188 * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1189 * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
1190 * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1191 *
1192 * It is the handler's job to read the input buffer from user memory
1193 * and process it. The handler should return 0 on success.
1194 *
1195 * This routine returns %NULL on a failure to register, and a pointer
1196 * to the table header on success.
1197 */
1198struct ctl_table_header *__register_sysctl_table(
1199 struct ctl_table_set *set,
1200 const char *path, struct ctl_table *table)
1201{
1202 struct ctl_table_root *root = set->dir.header.root;
1203 struct ctl_table_header *header;
1204 const char *name, *nextname;
1205 struct ctl_dir *dir;
1206 struct ctl_table *entry;
1207 struct ctl_node *node;
1208 int nr_entries = 0;
1209
1210 for (entry = table; entry->procname; entry++)
1211 nr_entries++;
1212
1213 header = kzalloc(sizeof(struct ctl_table_header) +
1214 sizeof(struct ctl_node)*nr_entries, GFP_KERNEL);
1215 if (!header)
1216 return NULL;
1217
1218 node = (struct ctl_node *)(header + 1);
1219 init_header(header, root, set, node, table);
1220 if (sysctl_check_table(path, table))
1221 goto fail;
1222
1223 spin_lock(&sysctl_lock);
1224 dir = &set->dir;
1225 /* Reference moved down the diretory tree get_subdir */
1226 dir->header.nreg++;
1227 spin_unlock(&sysctl_lock);
1228
1229 /* Find the directory for the ctl_table */
1230 for (name = path; name; name = nextname) {
1231 int namelen;
1232 nextname = strchr(name, '/');
1233 if (nextname) {
1234 namelen = nextname - name;
1235 nextname++;
1236 } else {
1237 namelen = strlen(name);
1238 }
1239 if (namelen == 0)
1240 continue;
1241
1242 dir = get_subdir(dir, name, namelen);
1243 if (IS_ERR(dir))
1244 goto fail;
1245 }
1246
1247 spin_lock(&sysctl_lock);
1248 if (insert_header(dir, header))
1249 goto fail_put_dir_locked;
1250
1251 drop_sysctl_table(&dir->header);
1252 spin_unlock(&sysctl_lock);
1253
1254 return header;
1255
1256fail_put_dir_locked:
1257 drop_sysctl_table(&dir->header);
1258 spin_unlock(&sysctl_lock);
1259fail:
1260 kfree(header);
1261 dump_stack();
1262 return NULL;
1263}
1264
1265/**
1266 * register_sysctl - register a sysctl table
1267 * @path: The path to the directory the sysctl table is in.
1268 * @table: the table structure
1269 *
1270 * Register a sysctl table. @table should be a filled in ctl_table
1271 * array. A completely 0 filled entry terminates the table.
1272 *
1273 * See __register_sysctl_table for more details.
1274 */
1275struct ctl_table_header *register_sysctl(const char *path, struct ctl_table *table)
1276{
1277 return __register_sysctl_table(&sysctl_table_root.default_set,
1278 path, table);
1279}
1280EXPORT_SYMBOL(register_sysctl);
1281
1282static char *append_path(const char *path, char *pos, const char *name)
1283{
1284 int namelen;
1285 namelen = strlen(name);
1286 if (((pos - path) + namelen + 2) >= PATH_MAX)
1287 return NULL;
1288 memcpy(pos, name, namelen);
1289 pos[namelen] = '/';
1290 pos[namelen + 1] = '\0';
1291 pos += namelen + 1;
1292 return pos;
1293}
1294
1295static int count_subheaders(struct ctl_table *table)
1296{
1297 int has_files = 0;
1298 int nr_subheaders = 0;
1299 struct ctl_table *entry;
1300
1301 /* special case: no directory and empty directory */
1302 if (!table || !table->procname)
1303 return 1;
1304
1305 for (entry = table; entry->procname; entry++) {
1306 if (entry->child)
1307 nr_subheaders += count_subheaders(entry->child);
1308 else
1309 has_files = 1;
1310 }
1311 return nr_subheaders + has_files;
1312}
1313
1314static int register_leaf_sysctl_tables(const char *path, char *pos,
1315 struct ctl_table_header ***subheader, struct ctl_table_set *set,
1316 struct ctl_table *table)
1317{
1318 struct ctl_table *ctl_table_arg = NULL;
1319 struct ctl_table *entry, *files;
1320 int nr_files = 0;
1321 int nr_dirs = 0;
1322 int err = -ENOMEM;
1323
1324 for (entry = table; entry->procname; entry++) {
1325 if (entry->child)
1326 nr_dirs++;
1327 else
1328 nr_files++;
1329 }
1330
1331 files = table;
1332 /* If there are mixed files and directories we need a new table */
1333 if (nr_dirs && nr_files) {
1334 struct ctl_table *new;
1335 files = kzalloc(sizeof(struct ctl_table) * (nr_files + 1),
1336 GFP_KERNEL);
1337 if (!files)
1338 goto out;
1339
1340 ctl_table_arg = files;
1341 for (new = files, entry = table; entry->procname; entry++) {
1342 if (entry->child)
1343 continue;
1344 *new = *entry;
1345 new++;
1346 }
1347 }
1348
1349 /* Register everything except a directory full of subdirectories */
1350 if (nr_files || !nr_dirs) {
1351 struct ctl_table_header *header;
1352 header = __register_sysctl_table(set, path, files);
1353 if (!header) {
1354 kfree(ctl_table_arg);
1355 goto out;
1356 }
1357
1358 /* Remember if we need to free the file table */
1359 header->ctl_table_arg = ctl_table_arg;
1360 **subheader = header;
1361 (*subheader)++;
1362 }
1363
1364 /* Recurse into the subdirectories. */
1365 for (entry = table; entry->procname; entry++) {
1366 char *child_pos;
1367
1368 if (!entry->child)
1369 continue;
1370
1371 err = -ENAMETOOLONG;
1372 child_pos = append_path(path, pos, entry->procname);
1373 if (!child_pos)
1374 goto out;
1375
1376 err = register_leaf_sysctl_tables(path, child_pos, subheader,
1377 set, entry->child);
1378 pos[0] = '\0';
1379 if (err)
1380 goto out;
1381 }
1382 err = 0;
1383out:
1384 /* On failure our caller will unregister all registered subheaders */
1385 return err;
1386}
1387
1388/**
1389 * __register_sysctl_paths - register a sysctl table hierarchy
1390 * @set: Sysctl tree to register on
1391 * @path: The path to the directory the sysctl table is in.
1392 * @table: the top-level table structure
1393 *
1394 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1395 * array. A completely 0 filled entry terminates the table.
1396 *
1397 * See __register_sysctl_table for more details.
1398 */
1399struct ctl_table_header *__register_sysctl_paths(
1400 struct ctl_table_set *set,
1401 const struct ctl_path *path, struct ctl_table *table)
1402{
1403 struct ctl_table *ctl_table_arg = table;
1404 int nr_subheaders = count_subheaders(table);
1405 struct ctl_table_header *header = NULL, **subheaders, **subheader;
1406 const struct ctl_path *component;
1407 char *new_path, *pos;
1408
1409 pos = new_path = kmalloc(PATH_MAX, GFP_KERNEL);
1410 if (!new_path)
1411 return NULL;
1412
1413 pos[0] = '\0';
1414 for (component = path; component->procname; component++) {
1415 pos = append_path(new_path, pos, component->procname);
1416 if (!pos)
1417 goto out;
1418 }
1419 while (table->procname && table->child && !table[1].procname) {
1420 pos = append_path(new_path, pos, table->procname);
1421 if (!pos)
1422 goto out;
1423 table = table->child;
1424 }
1425 if (nr_subheaders == 1) {
1426 header = __register_sysctl_table(set, new_path, table);
1427 if (header)
1428 header->ctl_table_arg = ctl_table_arg;
1429 } else {
1430 header = kzalloc(sizeof(*header) +
1431 sizeof(*subheaders)*nr_subheaders, GFP_KERNEL);
1432 if (!header)
1433 goto out;
1434
1435 subheaders = (struct ctl_table_header **) (header + 1);
1436 subheader = subheaders;
1437 header->ctl_table_arg = ctl_table_arg;
1438
1439 if (register_leaf_sysctl_tables(new_path, pos, &subheader,
1440 set, table))
1441 goto err_register_leaves;
1442 }
1443
1444out:
1445 kfree(new_path);
1446 return header;
1447
1448err_register_leaves:
1449 while (subheader > subheaders) {
1450 struct ctl_table_header *subh = *(--subheader);
1451 struct ctl_table *table = subh->ctl_table_arg;
1452 unregister_sysctl_table(subh);
1453 kfree(table);
1454 }
1455 kfree(header);
1456 header = NULL;
1457 goto out;
1458}
1459
1460/**
1461 * register_sysctl_table_path - register a sysctl table hierarchy
1462 * @path: The path to the directory the sysctl table is in.
1463 * @table: the top-level table structure
1464 *
1465 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1466 * array. A completely 0 filled entry terminates the table.
1467 *
1468 * See __register_sysctl_paths for more details.
1469 */
1470struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1471 struct ctl_table *table)
1472{
1473 return __register_sysctl_paths(&sysctl_table_root.default_set,
1474 path, table);
1475}
1476EXPORT_SYMBOL(register_sysctl_paths);
1477
1478/**
1479 * register_sysctl_table - register a sysctl table hierarchy
1480 * @table: the top-level table structure
1481 *
1482 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1483 * array. A completely 0 filled entry terminates the table.
1484 *
1485 * See register_sysctl_paths for more details.
1486 */
1487struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
1488{
1489 static const struct ctl_path null_path[] = { {} };
1490
1491 return register_sysctl_paths(null_path, table);
1492}
1493EXPORT_SYMBOL(register_sysctl_table);
1494
1495static void put_links(struct ctl_table_header *header)
1496{
1497 struct ctl_table_set *root_set = &sysctl_table_root.default_set;
1498 struct ctl_table_root *root = header->root;
1499 struct ctl_dir *parent = header->parent;
1500 struct ctl_dir *core_parent;
1501 struct ctl_table *entry;
1502
1503 if (header->set == root_set)
1504 return;
1505
1506 core_parent = xlate_dir(root_set, parent);
1507 if (IS_ERR(core_parent))
1508 return;
1509
1510 for (entry = header->ctl_table; entry->procname; entry++) {
1511 struct ctl_table_header *link_head;
1512 struct ctl_table *link;
1513 const char *name = entry->procname;
1514
1515 link = find_entry(&link_head, core_parent, name, strlen(name));
1516 if (link &&
1517 ((S_ISDIR(link->mode) && S_ISDIR(entry->mode)) ||
1518 (S_ISLNK(link->mode) && (link->data == root)))) {
1519 drop_sysctl_table(link_head);
1520 }
1521 else {
1522 printk(KERN_ERR "sysctl link missing during unregister: ");
1523 sysctl_print_dir(parent);
1524 printk(KERN_CONT "/%s\n", name);
1525 }
1526 }
1527}
1528
1529static void drop_sysctl_table(struct ctl_table_header *header)
1530{
1531 struct ctl_dir *parent = header->parent;
1532
1533 if (--header->nreg)
1534 return;
1535
1536 put_links(header);
1537 start_unregistering(header);
1538 if (!--header->count)
1539 kfree_rcu(header, rcu);
1540
1541 if (parent)
1542 drop_sysctl_table(&parent->header);
1543}
1544
1545/**
1546 * unregister_sysctl_table - unregister a sysctl table hierarchy
1547 * @header: the header returned from register_sysctl_table
1548 *
1549 * Unregisters the sysctl table and all children. proc entries may not
1550 * actually be removed until they are no longer used by anyone.
1551 */
1552void unregister_sysctl_table(struct ctl_table_header * header)
1553{
1554 int nr_subheaders;
1555 might_sleep();
1556
1557 if (header == NULL)
1558 return;
1559
1560 nr_subheaders = count_subheaders(header->ctl_table_arg);
1561 if (unlikely(nr_subheaders > 1)) {
1562 struct ctl_table_header **subheaders;
1563 int i;
1564
1565 subheaders = (struct ctl_table_header **)(header + 1);
1566 for (i = nr_subheaders -1; i >= 0; i--) {
1567 struct ctl_table_header *subh = subheaders[i];
1568 struct ctl_table *table = subh->ctl_table_arg;
1569 unregister_sysctl_table(subh);
1570 kfree(table);
1571 }
1572 kfree(header);
1573 return;
1574 }
1575
1576 spin_lock(&sysctl_lock);
1577 drop_sysctl_table(header);
1578 spin_unlock(&sysctl_lock);
1579}
1580EXPORT_SYMBOL(unregister_sysctl_table);
1581
1582void setup_sysctl_set(struct ctl_table_set *set,
1583 struct ctl_table_root *root,
1584 int (*is_seen)(struct ctl_table_set *))
1585{
1586 memset(set, 0, sizeof(*set));
1587 set->is_seen = is_seen;
1588 init_header(&set->dir.header, root, set, NULL, root_table);
1589}
1590
1591void retire_sysctl_set(struct ctl_table_set *set)
1592{
1593 WARN_ON(!RB_EMPTY_ROOT(&set->dir.root));
1594}
1595
473int __init proc_sys_init(void) 1596int __init proc_sys_init(void)
474{ 1597{
475 struct proc_dir_entry *proc_sys_root; 1598 struct proc_dir_entry *proc_sys_root;
@@ -478,5 +1601,6 @@ int __init proc_sys_init(void)
478 proc_sys_root->proc_iops = &proc_sys_dir_operations; 1601 proc_sys_root->proc_iops = &proc_sys_dir_operations;
479 proc_sys_root->proc_fops = &proc_sys_dir_file_operations; 1602 proc_sys_root->proc_fops = &proc_sys_dir_file_operations;
480 proc_sys_root->nlink = 0; 1603 proc_sys_root->nlink = 0;
481 return 0; 1604
1605 return sysctl_init();
482} 1606}
diff --git a/fs/proc/root.c b/fs/proc/root.c
index 46a15d8a29ca..7c30fce037c0 100644
--- a/fs/proc/root.c
+++ b/fs/proc/root.c
@@ -67,7 +67,7 @@ static int proc_parse_options(char *options, struct pid_namespace *pid)
67 case Opt_gid: 67 case Opt_gid:
68 if (match_int(&args[0], &option)) 68 if (match_int(&args[0], &option))
69 return 0; 69 return 0;
70 pid->pid_gid = option; 70 pid->pid_gid = make_kgid(current_user_ns(), option);
71 break; 71 break;
72 case Opt_hidepid: 72 case Opt_hidepid:
73 if (match_int(&args[0], &option)) 73 if (match_int(&args[0], &option))
@@ -115,12 +115,13 @@ static struct dentry *proc_mount(struct file_system_type *fs_type,
115 if (IS_ERR(sb)) 115 if (IS_ERR(sb))
116 return ERR_CAST(sb); 116 return ERR_CAST(sb);
117 117
118 if (!proc_parse_options(options, ns)) {
119 deactivate_locked_super(sb);
120 return ERR_PTR(-EINVAL);
121 }
122
118 if (!sb->s_root) { 123 if (!sb->s_root) {
119 sb->s_flags = flags; 124 sb->s_flags = flags;
120 if (!proc_parse_options(options, ns)) {
121 deactivate_locked_super(sb);
122 return ERR_PTR(-EINVAL);
123 }
124 err = proc_fill_super(sb); 125 err = proc_fill_super(sb);
125 if (err) { 126 if (err) {
126 deactivate_locked_super(sb); 127 deactivate_locked_super(sb);
diff --git a/fs/proc/stat.c b/fs/proc/stat.c
index 121f77cfef76..64c3b3172367 100644
--- a/fs/proc/stat.c
+++ b/fs/proc/stat.c
@@ -18,19 +18,39 @@
18#ifndef arch_irq_stat 18#ifndef arch_irq_stat
19#define arch_irq_stat() 0 19#define arch_irq_stat() 0
20#endif 20#endif
21#ifndef arch_idle_time 21
22#define arch_idle_time(cpu) 0 22#ifdef arch_idle_time
23#endif 23
24static cputime64_t get_idle_time(int cpu)
25{
26 cputime64_t idle;
27
28 idle = kcpustat_cpu(cpu).cpustat[CPUTIME_IDLE];
29 if (cpu_online(cpu) && !nr_iowait_cpu(cpu))
30 idle += arch_idle_time(cpu);
31 return idle;
32}
33
34static cputime64_t get_iowait_time(int cpu)
35{
36 cputime64_t iowait;
37
38 iowait = kcpustat_cpu(cpu).cpustat[CPUTIME_IOWAIT];
39 if (cpu_online(cpu) && nr_iowait_cpu(cpu))
40 iowait += arch_idle_time(cpu);
41 return iowait;
42}
43
44#else
24 45
25static u64 get_idle_time(int cpu) 46static u64 get_idle_time(int cpu)
26{ 47{
27 u64 idle, idle_time = get_cpu_idle_time_us(cpu, NULL); 48 u64 idle, idle_time = get_cpu_idle_time_us(cpu, NULL);
28 49
29 if (idle_time == -1ULL) { 50 if (idle_time == -1ULL)
30 /* !NO_HZ so we can rely on cpustat.idle */ 51 /* !NO_HZ so we can rely on cpustat.idle */
31 idle = kcpustat_cpu(cpu).cpustat[CPUTIME_IDLE]; 52 idle = kcpustat_cpu(cpu).cpustat[CPUTIME_IDLE];
32 idle += arch_idle_time(cpu); 53 else
33 } else
34 idle = usecs_to_cputime64(idle_time); 54 idle = usecs_to_cputime64(idle_time);
35 55
36 return idle; 56 return idle;
@@ -49,6 +69,8 @@ static u64 get_iowait_time(int cpu)
49 return iowait; 69 return iowait;
50} 70}
51 71
72#endif
73
52static int show_stat(struct seq_file *p, void *v) 74static int show_stat(struct seq_file *p, void *v)
53{ 75{
54 int i, j; 76 int i, j;
@@ -89,18 +111,19 @@ static int show_stat(struct seq_file *p, void *v)
89 } 111 }
90 sum += arch_irq_stat(); 112 sum += arch_irq_stat();
91 113
92 seq_printf(p, "cpu %llu %llu %llu %llu %llu %llu %llu %llu %llu " 114 seq_puts(p, "cpu ");
93 "%llu\n", 115 seq_put_decimal_ull(p, ' ', cputime64_to_clock_t(user));
94 (unsigned long long)cputime64_to_clock_t(user), 116 seq_put_decimal_ull(p, ' ', cputime64_to_clock_t(nice));
95 (unsigned long long)cputime64_to_clock_t(nice), 117 seq_put_decimal_ull(p, ' ', cputime64_to_clock_t(system));
96 (unsigned long long)cputime64_to_clock_t(system), 118 seq_put_decimal_ull(p, ' ', cputime64_to_clock_t(idle));
97 (unsigned long long)cputime64_to_clock_t(idle), 119 seq_put_decimal_ull(p, ' ', cputime64_to_clock_t(iowait));
98 (unsigned long long)cputime64_to_clock_t(iowait), 120 seq_put_decimal_ull(p, ' ', cputime64_to_clock_t(irq));
99 (unsigned long long)cputime64_to_clock_t(irq), 121 seq_put_decimal_ull(p, ' ', cputime64_to_clock_t(softirq));
100 (unsigned long long)cputime64_to_clock_t(softirq), 122 seq_put_decimal_ull(p, ' ', cputime64_to_clock_t(steal));
101 (unsigned long long)cputime64_to_clock_t(steal), 123 seq_put_decimal_ull(p, ' ', cputime64_to_clock_t(guest));
102 (unsigned long long)cputime64_to_clock_t(guest), 124 seq_put_decimal_ull(p, ' ', cputime64_to_clock_t(guest_nice));
103 (unsigned long long)cputime64_to_clock_t(guest_nice)); 125 seq_putc(p, '\n');
126
104 for_each_online_cpu(i) { 127 for_each_online_cpu(i) {
105 /* Copy values here to work around gcc-2.95.3, gcc-2.96 */ 128 /* Copy values here to work around gcc-2.95.3, gcc-2.96 */
106 user = kcpustat_cpu(i).cpustat[CPUTIME_USER]; 129 user = kcpustat_cpu(i).cpustat[CPUTIME_USER];
@@ -113,26 +136,24 @@ static int show_stat(struct seq_file *p, void *v)
113 steal = kcpustat_cpu(i).cpustat[CPUTIME_STEAL]; 136 steal = kcpustat_cpu(i).cpustat[CPUTIME_STEAL];
114 guest = kcpustat_cpu(i).cpustat[CPUTIME_GUEST]; 137 guest = kcpustat_cpu(i).cpustat[CPUTIME_GUEST];
115 guest_nice = kcpustat_cpu(i).cpustat[CPUTIME_GUEST_NICE]; 138 guest_nice = kcpustat_cpu(i).cpustat[CPUTIME_GUEST_NICE];
116 seq_printf(p, 139 seq_printf(p, "cpu%d", i);
117 "cpu%d %llu %llu %llu %llu %llu %llu %llu %llu %llu " 140 seq_put_decimal_ull(p, ' ', cputime64_to_clock_t(user));
118 "%llu\n", 141 seq_put_decimal_ull(p, ' ', cputime64_to_clock_t(nice));
119 i, 142 seq_put_decimal_ull(p, ' ', cputime64_to_clock_t(system));
120 (unsigned long long)cputime64_to_clock_t(user), 143 seq_put_decimal_ull(p, ' ', cputime64_to_clock_t(idle));
121 (unsigned long long)cputime64_to_clock_t(nice), 144 seq_put_decimal_ull(p, ' ', cputime64_to_clock_t(iowait));
122 (unsigned long long)cputime64_to_clock_t(system), 145 seq_put_decimal_ull(p, ' ', cputime64_to_clock_t(irq));
123 (unsigned long long)cputime64_to_clock_t(idle), 146 seq_put_decimal_ull(p, ' ', cputime64_to_clock_t(softirq));
124 (unsigned long long)cputime64_to_clock_t(iowait), 147 seq_put_decimal_ull(p, ' ', cputime64_to_clock_t(steal));
125 (unsigned long long)cputime64_to_clock_t(irq), 148 seq_put_decimal_ull(p, ' ', cputime64_to_clock_t(guest));
126 (unsigned long long)cputime64_to_clock_t(softirq), 149 seq_put_decimal_ull(p, ' ', cputime64_to_clock_t(guest_nice));
127 (unsigned long long)cputime64_to_clock_t(steal), 150 seq_putc(p, '\n');
128 (unsigned long long)cputime64_to_clock_t(guest),
129 (unsigned long long)cputime64_to_clock_t(guest_nice));
130 } 151 }
131 seq_printf(p, "intr %llu", (unsigned long long)sum); 152 seq_printf(p, "intr %llu", (unsigned long long)sum);
132 153
133 /* sum again ? it could be updated? */ 154 /* sum again ? it could be updated? */
134 for_each_irq_nr(j) 155 for_each_irq_nr(j)
135 seq_printf(p, " %u", kstat_irqs(j)); 156 seq_put_decimal_ull(p, ' ', kstat_irqs(j));
136 157
137 seq_printf(p, 158 seq_printf(p,
138 "\nctxt %llu\n" 159 "\nctxt %llu\n"
@@ -149,7 +170,7 @@ static int show_stat(struct seq_file *p, void *v)
149 seq_printf(p, "softirq %llu", (unsigned long long)sum_softirq); 170 seq_printf(p, "softirq %llu", (unsigned long long)sum_softirq);
150 171
151 for (i = 0; i < NR_SOFTIRQS; i++) 172 for (i = 0; i < NR_SOFTIRQS; i++)
152 seq_printf(p, " %u", per_softirq_sums[i]); 173 seq_put_decimal_ull(p, ' ', per_softirq_sums[i]);
153 seq_putc(p, '\n'); 174 seq_putc(p, '\n');
154 175
155 return 0; 176 return 0;
@@ -157,11 +178,14 @@ static int show_stat(struct seq_file *p, void *v)
157 178
158static int stat_open(struct inode *inode, struct file *file) 179static int stat_open(struct inode *inode, struct file *file)
159{ 180{
160 unsigned size = 4096 * (1 + num_possible_cpus() / 32); 181 unsigned size = 1024 + 128 * num_possible_cpus();
161 char *buf; 182 char *buf;
162 struct seq_file *m; 183 struct seq_file *m;
163 int res; 184 int res;
164 185
186 /* minimum size to display an interrupt count : 2 bytes */
187 size += 2 * nr_irqs;
188
165 /* don't ask for more than the kmalloc() max size */ 189 /* don't ask for more than the kmalloc() max size */
166 if (size > KMALLOC_MAX_SIZE) 190 if (size > KMALLOC_MAX_SIZE)
167 size = KMALLOC_MAX_SIZE; 191 size = KMALLOC_MAX_SIZE;
@@ -173,7 +197,7 @@ static int stat_open(struct inode *inode, struct file *file)
173 if (!res) { 197 if (!res) {
174 m = file->private_data; 198 m = file->private_data;
175 m->buf = buf; 199 m->buf = buf;
176 m->size = size; 200 m->size = ksize(buf);
177 } else 201 } else
178 kfree(buf); 202 kfree(buf);
179 return res; 203 return res;
diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
index 7dcd2a250495..7faaf2acc570 100644
--- a/fs/proc/task_mmu.c
+++ b/fs/proc/task_mmu.c
@@ -209,16 +209,20 @@ static int do_maps_open(struct inode *inode, struct file *file,
209 return ret; 209 return ret;
210} 210}
211 211
212static void show_map_vma(struct seq_file *m, struct vm_area_struct *vma) 212static void
213show_map_vma(struct seq_file *m, struct vm_area_struct *vma, int is_pid)
213{ 214{
214 struct mm_struct *mm = vma->vm_mm; 215 struct mm_struct *mm = vma->vm_mm;
215 struct file *file = vma->vm_file; 216 struct file *file = vma->vm_file;
217 struct proc_maps_private *priv = m->private;
218 struct task_struct *task = priv->task;
216 vm_flags_t flags = vma->vm_flags; 219 vm_flags_t flags = vma->vm_flags;
217 unsigned long ino = 0; 220 unsigned long ino = 0;
218 unsigned long long pgoff = 0; 221 unsigned long long pgoff = 0;
219 unsigned long start, end; 222 unsigned long start, end;
220 dev_t dev = 0; 223 dev_t dev = 0;
221 int len; 224 int len;
225 const char *name = NULL;
222 226
223 if (file) { 227 if (file) {
224 struct inode *inode = vma->vm_file->f_path.dentry->d_inode; 228 struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
@@ -252,36 +256,57 @@ static void show_map_vma(struct seq_file *m, struct vm_area_struct *vma)
252 if (file) { 256 if (file) {
253 pad_len_spaces(m, len); 257 pad_len_spaces(m, len);
254 seq_path(m, &file->f_path, "\n"); 258 seq_path(m, &file->f_path, "\n");
255 } else { 259 goto done;
256 const char *name = arch_vma_name(vma); 260 }
257 if (!name) { 261
258 if (mm) { 262 name = arch_vma_name(vma);
259 if (vma->vm_start <= mm->brk && 263 if (!name) {
260 vma->vm_end >= mm->start_brk) { 264 pid_t tid;
261 name = "[heap]"; 265
262 } else if (vma->vm_start <= mm->start_stack && 266 if (!mm) {
263 vma->vm_end >= mm->start_stack) { 267 name = "[vdso]";
264 name = "[stack]"; 268 goto done;
265 } 269 }
270
271 if (vma->vm_start <= mm->brk &&
272 vma->vm_end >= mm->start_brk) {
273 name = "[heap]";
274 goto done;
275 }
276
277 tid = vm_is_stack(task, vma, is_pid);
278
279 if (tid != 0) {
280 /*
281 * Thread stack in /proc/PID/task/TID/maps or
282 * the main process stack.
283 */
284 if (!is_pid || (vma->vm_start <= mm->start_stack &&
285 vma->vm_end >= mm->start_stack)) {
286 name = "[stack]";
266 } else { 287 } else {
267 name = "[vdso]"; 288 /* Thread stack in /proc/PID/maps */
289 pad_len_spaces(m, len);
290 seq_printf(m, "[stack:%d]", tid);
268 } 291 }
269 } 292 }
270 if (name) { 293 }
271 pad_len_spaces(m, len); 294
272 seq_puts(m, name); 295done:
273 } 296 if (name) {
297 pad_len_spaces(m, len);
298 seq_puts(m, name);
274 } 299 }
275 seq_putc(m, '\n'); 300 seq_putc(m, '\n');
276} 301}
277 302
278static int show_map(struct seq_file *m, void *v) 303static int show_map(struct seq_file *m, void *v, int is_pid)
279{ 304{
280 struct vm_area_struct *vma = v; 305 struct vm_area_struct *vma = v;
281 struct proc_maps_private *priv = m->private; 306 struct proc_maps_private *priv = m->private;
282 struct task_struct *task = priv->task; 307 struct task_struct *task = priv->task;
283 308
284 show_map_vma(m, vma); 309 show_map_vma(m, vma, is_pid);
285 310
286 if (m->count < m->size) /* vma is copied successfully */ 311 if (m->count < m->size) /* vma is copied successfully */
287 m->version = (vma != get_gate_vma(task->mm)) 312 m->version = (vma != get_gate_vma(task->mm))
@@ -289,20 +314,49 @@ static int show_map(struct seq_file *m, void *v)
289 return 0; 314 return 0;
290} 315}
291 316
317static int show_pid_map(struct seq_file *m, void *v)
318{
319 return show_map(m, v, 1);
320}
321
322static int show_tid_map(struct seq_file *m, void *v)
323{
324 return show_map(m, v, 0);
325}
326
292static const struct seq_operations proc_pid_maps_op = { 327static const struct seq_operations proc_pid_maps_op = {
293 .start = m_start, 328 .start = m_start,
294 .next = m_next, 329 .next = m_next,
295 .stop = m_stop, 330 .stop = m_stop,
296 .show = show_map 331 .show = show_pid_map
297}; 332};
298 333
299static int maps_open(struct inode *inode, struct file *file) 334static const struct seq_operations proc_tid_maps_op = {
335 .start = m_start,
336 .next = m_next,
337 .stop = m_stop,
338 .show = show_tid_map
339};
340
341static int pid_maps_open(struct inode *inode, struct file *file)
300{ 342{
301 return do_maps_open(inode, file, &proc_pid_maps_op); 343 return do_maps_open(inode, file, &proc_pid_maps_op);
302} 344}
303 345
304const struct file_operations proc_maps_operations = { 346static int tid_maps_open(struct inode *inode, struct file *file)
305 .open = maps_open, 347{
348 return do_maps_open(inode, file, &proc_tid_maps_op);
349}
350
351const struct file_operations proc_pid_maps_operations = {
352 .open = pid_maps_open,
353 .read = seq_read,
354 .llseek = seq_lseek,
355 .release = seq_release_private,
356};
357
358const struct file_operations proc_tid_maps_operations = {
359 .open = tid_maps_open,
306 .read = seq_read, 360 .read = seq_read,
307 .llseek = seq_lseek, 361 .llseek = seq_lseek,
308 .release = seq_release_private, 362 .release = seq_release_private,
@@ -394,21 +448,15 @@ static int smaps_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end,
394 pte_t *pte; 448 pte_t *pte;
395 spinlock_t *ptl; 449 spinlock_t *ptl;
396 450
397 spin_lock(&walk->mm->page_table_lock); 451 if (pmd_trans_huge_lock(pmd, vma) == 1) {
398 if (pmd_trans_huge(*pmd)) { 452 smaps_pte_entry(*(pte_t *)pmd, addr, HPAGE_PMD_SIZE, walk);
399 if (pmd_trans_splitting(*pmd)) {
400 spin_unlock(&walk->mm->page_table_lock);
401 wait_split_huge_page(vma->anon_vma, pmd);
402 } else {
403 smaps_pte_entry(*(pte_t *)pmd, addr,
404 HPAGE_PMD_SIZE, walk);
405 spin_unlock(&walk->mm->page_table_lock);
406 mss->anonymous_thp += HPAGE_PMD_SIZE;
407 return 0;
408 }
409 } else {
410 spin_unlock(&walk->mm->page_table_lock); 453 spin_unlock(&walk->mm->page_table_lock);
454 mss->anonymous_thp += HPAGE_PMD_SIZE;
455 return 0;
411 } 456 }
457
458 if (pmd_trans_unstable(pmd))
459 return 0;
412 /* 460 /*
413 * The mmap_sem held all the way back in m_start() is what 461 * The mmap_sem held all the way back in m_start() is what
414 * keeps khugepaged out of here and from collapsing things 462 * keeps khugepaged out of here and from collapsing things
@@ -422,7 +470,7 @@ static int smaps_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end,
422 return 0; 470 return 0;
423} 471}
424 472
425static int show_smap(struct seq_file *m, void *v) 473static int show_smap(struct seq_file *m, void *v, int is_pid)
426{ 474{
427 struct proc_maps_private *priv = m->private; 475 struct proc_maps_private *priv = m->private;
428 struct task_struct *task = priv->task; 476 struct task_struct *task = priv->task;
@@ -440,7 +488,7 @@ static int show_smap(struct seq_file *m, void *v)
440 if (vma->vm_mm && !is_vm_hugetlb_page(vma)) 488 if (vma->vm_mm && !is_vm_hugetlb_page(vma))
441 walk_page_range(vma->vm_start, vma->vm_end, &smaps_walk); 489 walk_page_range(vma->vm_start, vma->vm_end, &smaps_walk);
442 490
443 show_map_vma(m, vma); 491 show_map_vma(m, vma, is_pid);
444 492
445 seq_printf(m, 493 seq_printf(m,
446 "Size: %8lu kB\n" 494 "Size: %8lu kB\n"
@@ -479,20 +527,49 @@ static int show_smap(struct seq_file *m, void *v)
479 return 0; 527 return 0;
480} 528}
481 529
530static int show_pid_smap(struct seq_file *m, void *v)
531{
532 return show_smap(m, v, 1);
533}
534
535static int show_tid_smap(struct seq_file *m, void *v)
536{
537 return show_smap(m, v, 0);
538}
539
482static const struct seq_operations proc_pid_smaps_op = { 540static const struct seq_operations proc_pid_smaps_op = {
483 .start = m_start, 541 .start = m_start,
484 .next = m_next, 542 .next = m_next,
485 .stop = m_stop, 543 .stop = m_stop,
486 .show = show_smap 544 .show = show_pid_smap
545};
546
547static const struct seq_operations proc_tid_smaps_op = {
548 .start = m_start,
549 .next = m_next,
550 .stop = m_stop,
551 .show = show_tid_smap
487}; 552};
488 553
489static int smaps_open(struct inode *inode, struct file *file) 554static int pid_smaps_open(struct inode *inode, struct file *file)
490{ 555{
491 return do_maps_open(inode, file, &proc_pid_smaps_op); 556 return do_maps_open(inode, file, &proc_pid_smaps_op);
492} 557}
493 558
494const struct file_operations proc_smaps_operations = { 559static int tid_smaps_open(struct inode *inode, struct file *file)
495 .open = smaps_open, 560{
561 return do_maps_open(inode, file, &proc_tid_smaps_op);
562}
563
564const struct file_operations proc_pid_smaps_operations = {
565 .open = pid_smaps_open,
566 .read = seq_read,
567 .llseek = seq_lseek,
568 .release = seq_release_private,
569};
570
571const struct file_operations proc_tid_smaps_operations = {
572 .open = tid_smaps_open,
496 .read = seq_read, 573 .read = seq_read,
497 .llseek = seq_lseek, 574 .llseek = seq_lseek,
498 .release = seq_release_private, 575 .release = seq_release_private,
@@ -507,6 +584,8 @@ static int clear_refs_pte_range(pmd_t *pmd, unsigned long addr,
507 struct page *page; 584 struct page *page;
508 585
509 split_huge_page_pmd(walk->mm, pmd); 586 split_huge_page_pmd(walk->mm, pmd);
587 if (pmd_trans_unstable(pmd))
588 return 0;
510 589
511 pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl); 590 pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
512 for (; addr != end; pte++, addr += PAGE_SIZE) { 591 for (; addr != end; pte++, addr += PAGE_SIZE) {
@@ -518,9 +597,6 @@ static int clear_refs_pte_range(pmd_t *pmd, unsigned long addr,
518 if (!page) 597 if (!page)
519 continue; 598 continue;
520 599
521 if (PageReserved(page))
522 continue;
523
524 /* Clear accessed and referenced bits. */ 600 /* Clear accessed and referenced bits. */
525 ptep_test_and_clear_young(vma, addr, pte); 601 ptep_test_and_clear_young(vma, addr, pte);
526 ClearPageReferenced(page); 602 ClearPageReferenced(page);
@@ -598,11 +674,18 @@ const struct file_operations proc_clear_refs_operations = {
598 .llseek = noop_llseek, 674 .llseek = noop_llseek,
599}; 675};
600 676
677typedef struct {
678 u64 pme;
679} pagemap_entry_t;
680
601struct pagemapread { 681struct pagemapread {
602 int pos, len; 682 int pos, len;
603 u64 *buffer; 683 pagemap_entry_t *buffer;
604}; 684};
605 685
686#define PAGEMAP_WALK_SIZE (PMD_SIZE)
687#define PAGEMAP_WALK_MASK (PMD_MASK)
688
606#define PM_ENTRY_BYTES sizeof(u64) 689#define PM_ENTRY_BYTES sizeof(u64)
607#define PM_STATUS_BITS 3 690#define PM_STATUS_BITS 3
608#define PM_STATUS_OFFSET (64 - PM_STATUS_BITS) 691#define PM_STATUS_OFFSET (64 - PM_STATUS_BITS)
@@ -620,10 +703,15 @@ struct pagemapread {
620#define PM_NOT_PRESENT PM_PSHIFT(PAGE_SHIFT) 703#define PM_NOT_PRESENT PM_PSHIFT(PAGE_SHIFT)
621#define PM_END_OF_BUFFER 1 704#define PM_END_OF_BUFFER 1
622 705
623static int add_to_pagemap(unsigned long addr, u64 pfn, 706static inline pagemap_entry_t make_pme(u64 val)
707{
708 return (pagemap_entry_t) { .pme = val };
709}
710
711static int add_to_pagemap(unsigned long addr, pagemap_entry_t *pme,
624 struct pagemapread *pm) 712 struct pagemapread *pm)
625{ 713{
626 pm->buffer[pm->pos++] = pfn; 714 pm->buffer[pm->pos++] = *pme;
627 if (pm->pos >= pm->len) 715 if (pm->pos >= pm->len)
628 return PM_END_OF_BUFFER; 716 return PM_END_OF_BUFFER;
629 return 0; 717 return 0;
@@ -635,8 +723,10 @@ static int pagemap_pte_hole(unsigned long start, unsigned long end,
635 struct pagemapread *pm = walk->private; 723 struct pagemapread *pm = walk->private;
636 unsigned long addr; 724 unsigned long addr;
637 int err = 0; 725 int err = 0;
726 pagemap_entry_t pme = make_pme(PM_NOT_PRESENT);
727
638 for (addr = start; addr < end; addr += PAGE_SIZE) { 728 for (addr = start; addr < end; addr += PAGE_SIZE) {
639 err = add_to_pagemap(addr, PM_NOT_PRESENT, pm); 729 err = add_to_pagemap(addr, &pme, pm);
640 if (err) 730 if (err)
641 break; 731 break;
642 } 732 }
@@ -649,18 +739,40 @@ static u64 swap_pte_to_pagemap_entry(pte_t pte)
649 return swp_type(e) | (swp_offset(e) << MAX_SWAPFILES_SHIFT); 739 return swp_type(e) | (swp_offset(e) << MAX_SWAPFILES_SHIFT);
650} 740}
651 741
652static u64 pte_to_pagemap_entry(pte_t pte) 742static void pte_to_pagemap_entry(pagemap_entry_t *pme, pte_t pte)
653{ 743{
654 u64 pme = 0;
655 if (is_swap_pte(pte)) 744 if (is_swap_pte(pte))
656 pme = PM_PFRAME(swap_pte_to_pagemap_entry(pte)) 745 *pme = make_pme(PM_PFRAME(swap_pte_to_pagemap_entry(pte))
657 | PM_PSHIFT(PAGE_SHIFT) | PM_SWAP; 746 | PM_PSHIFT(PAGE_SHIFT) | PM_SWAP);
658 else if (pte_present(pte)) 747 else if (pte_present(pte))
659 pme = PM_PFRAME(pte_pfn(pte)) 748 *pme = make_pme(PM_PFRAME(pte_pfn(pte))
660 | PM_PSHIFT(PAGE_SHIFT) | PM_PRESENT; 749 | PM_PSHIFT(PAGE_SHIFT) | PM_PRESENT);
661 return pme; 750 else
751 *pme = make_pme(PM_NOT_PRESENT);
662} 752}
663 753
754#ifdef CONFIG_TRANSPARENT_HUGEPAGE
755static void thp_pmd_to_pagemap_entry(pagemap_entry_t *pme,
756 pmd_t pmd, int offset)
757{
758 /*
759 * Currently pmd for thp is always present because thp can not be
760 * swapped-out, migrated, or HWPOISONed (split in such cases instead.)
761 * This if-check is just to prepare for future implementation.
762 */
763 if (pmd_present(pmd))
764 *pme = make_pme(PM_PFRAME(pmd_pfn(pmd) + offset)
765 | PM_PSHIFT(PAGE_SHIFT) | PM_PRESENT);
766 else
767 *pme = make_pme(PM_NOT_PRESENT);
768}
769#else
770static inline void thp_pmd_to_pagemap_entry(pagemap_entry_t *pme,
771 pmd_t pmd, int offset)
772{
773}
774#endif
775
664static int pagemap_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end, 776static int pagemap_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end,
665 struct mm_walk *walk) 777 struct mm_walk *walk)
666{ 778{
@@ -668,29 +780,46 @@ static int pagemap_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end,
668 struct pagemapread *pm = walk->private; 780 struct pagemapread *pm = walk->private;
669 pte_t *pte; 781 pte_t *pte;
670 int err = 0; 782 int err = 0;
671 783 pagemap_entry_t pme = make_pme(PM_NOT_PRESENT);
672 split_huge_page_pmd(walk->mm, pmd);
673 784
674 /* find the first VMA at or above 'addr' */ 785 /* find the first VMA at or above 'addr' */
675 vma = find_vma(walk->mm, addr); 786 vma = find_vma(walk->mm, addr);
787 if (vma && pmd_trans_huge_lock(pmd, vma) == 1) {
788 for (; addr != end; addr += PAGE_SIZE) {
789 unsigned long offset;
790
791 offset = (addr & ~PAGEMAP_WALK_MASK) >>
792 PAGE_SHIFT;
793 thp_pmd_to_pagemap_entry(&pme, *pmd, offset);
794 err = add_to_pagemap(addr, &pme, pm);
795 if (err)
796 break;
797 }
798 spin_unlock(&walk->mm->page_table_lock);
799 return err;
800 }
801
802 if (pmd_trans_unstable(pmd))
803 return 0;
676 for (; addr != end; addr += PAGE_SIZE) { 804 for (; addr != end; addr += PAGE_SIZE) {
677 u64 pfn = PM_NOT_PRESENT;
678 805
679 /* check to see if we've left 'vma' behind 806 /* check to see if we've left 'vma' behind
680 * and need a new, higher one */ 807 * and need a new, higher one */
681 if (vma && (addr >= vma->vm_end)) 808 if (vma && (addr >= vma->vm_end)) {
682 vma = find_vma(walk->mm, addr); 809 vma = find_vma(walk->mm, addr);
810 pme = make_pme(PM_NOT_PRESENT);
811 }
683 812
684 /* check that 'vma' actually covers this address, 813 /* check that 'vma' actually covers this address,
685 * and that it isn't a huge page vma */ 814 * and that it isn't a huge page vma */
686 if (vma && (vma->vm_start <= addr) && 815 if (vma && (vma->vm_start <= addr) &&
687 !is_vm_hugetlb_page(vma)) { 816 !is_vm_hugetlb_page(vma)) {
688 pte = pte_offset_map(pmd, addr); 817 pte = pte_offset_map(pmd, addr);
689 pfn = pte_to_pagemap_entry(*pte); 818 pte_to_pagemap_entry(&pme, *pte);
690 /* unmap before userspace copy */ 819 /* unmap before userspace copy */
691 pte_unmap(pte); 820 pte_unmap(pte);
692 } 821 }
693 err = add_to_pagemap(addr, pfn, pm); 822 err = add_to_pagemap(addr, &pme, pm);
694 if (err) 823 if (err)
695 return err; 824 return err;
696 } 825 }
@@ -701,13 +830,14 @@ static int pagemap_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end,
701} 830}
702 831
703#ifdef CONFIG_HUGETLB_PAGE 832#ifdef CONFIG_HUGETLB_PAGE
704static u64 huge_pte_to_pagemap_entry(pte_t pte, int offset) 833static void huge_pte_to_pagemap_entry(pagemap_entry_t *pme,
834 pte_t pte, int offset)
705{ 835{
706 u64 pme = 0;
707 if (pte_present(pte)) 836 if (pte_present(pte))
708 pme = PM_PFRAME(pte_pfn(pte) + offset) 837 *pme = make_pme(PM_PFRAME(pte_pfn(pte) + offset)
709 | PM_PSHIFT(PAGE_SHIFT) | PM_PRESENT; 838 | PM_PSHIFT(PAGE_SHIFT) | PM_PRESENT);
710 return pme; 839 else
840 *pme = make_pme(PM_NOT_PRESENT);
711} 841}
712 842
713/* This function walks within one hugetlb entry in the single call */ 843/* This function walks within one hugetlb entry in the single call */
@@ -717,12 +847,12 @@ static int pagemap_hugetlb_range(pte_t *pte, unsigned long hmask,
717{ 847{
718 struct pagemapread *pm = walk->private; 848 struct pagemapread *pm = walk->private;
719 int err = 0; 849 int err = 0;
720 u64 pfn; 850 pagemap_entry_t pme;
721 851
722 for (; addr != end; addr += PAGE_SIZE) { 852 for (; addr != end; addr += PAGE_SIZE) {
723 int offset = (addr & ~hmask) >> PAGE_SHIFT; 853 int offset = (addr & ~hmask) >> PAGE_SHIFT;
724 pfn = huge_pte_to_pagemap_entry(*pte, offset); 854 huge_pte_to_pagemap_entry(&pme, *pte, offset);
725 err = add_to_pagemap(addr, pfn, pm); 855 err = add_to_pagemap(addr, &pme, pm);
726 if (err) 856 if (err)
727 return err; 857 return err;
728 } 858 }
@@ -757,8 +887,6 @@ static int pagemap_hugetlb_range(pte_t *pte, unsigned long hmask,
757 * determine which areas of memory are actually mapped and llseek to 887 * determine which areas of memory are actually mapped and llseek to
758 * skip over unmapped regions. 888 * skip over unmapped regions.
759 */ 889 */
760#define PAGEMAP_WALK_SIZE (PMD_SIZE)
761#define PAGEMAP_WALK_MASK (PMD_MASK)
762static ssize_t pagemap_read(struct file *file, char __user *buf, 890static ssize_t pagemap_read(struct file *file, char __user *buf,
763 size_t count, loff_t *ppos) 891 size_t count, loff_t *ppos)
764{ 892{
@@ -941,26 +1069,21 @@ static int gather_pte_stats(pmd_t *pmd, unsigned long addr,
941 pte_t *pte; 1069 pte_t *pte;
942 1070
943 md = walk->private; 1071 md = walk->private;
944 spin_lock(&walk->mm->page_table_lock); 1072
945 if (pmd_trans_huge(*pmd)) { 1073 if (pmd_trans_huge_lock(pmd, md->vma) == 1) {
946 if (pmd_trans_splitting(*pmd)) { 1074 pte_t huge_pte = *(pte_t *)pmd;
947 spin_unlock(&walk->mm->page_table_lock); 1075 struct page *page;
948 wait_split_huge_page(md->vma->anon_vma, pmd); 1076
949 } else { 1077 page = can_gather_numa_stats(huge_pte, md->vma, addr);
950 pte_t huge_pte = *(pte_t *)pmd; 1078 if (page)
951 struct page *page; 1079 gather_stats(page, md, pte_dirty(huge_pte),
952 1080 HPAGE_PMD_SIZE/PAGE_SIZE);
953 page = can_gather_numa_stats(huge_pte, md->vma, addr);
954 if (page)
955 gather_stats(page, md, pte_dirty(huge_pte),
956 HPAGE_PMD_SIZE/PAGE_SIZE);
957 spin_unlock(&walk->mm->page_table_lock);
958 return 0;
959 }
960 } else {
961 spin_unlock(&walk->mm->page_table_lock); 1081 spin_unlock(&walk->mm->page_table_lock);
1082 return 0;
962 } 1083 }
963 1084
1085 if (pmd_trans_unstable(pmd))
1086 return 0;
964 orig_pte = pte = pte_offset_map_lock(walk->mm, pmd, addr, &ptl); 1087 orig_pte = pte = pte_offset_map_lock(walk->mm, pmd, addr, &ptl);
965 do { 1088 do {
966 struct page *page = can_gather_numa_stats(*pte, md->vma, addr); 1089 struct page *page = can_gather_numa_stats(*pte, md->vma, addr);
@@ -1002,7 +1125,7 @@ static int gather_hugetbl_stats(pte_t *pte, unsigned long hmask,
1002/* 1125/*
1003 * Display pages allocated per node and memory policy via /proc. 1126 * Display pages allocated per node and memory policy via /proc.
1004 */ 1127 */
1005static int show_numa_map(struct seq_file *m, void *v) 1128static int show_numa_map(struct seq_file *m, void *v, int is_pid)
1006{ 1129{
1007 struct numa_maps_private *numa_priv = m->private; 1130 struct numa_maps_private *numa_priv = m->private;
1008 struct proc_maps_private *proc_priv = &numa_priv->proc_maps; 1131 struct proc_maps_private *proc_priv = &numa_priv->proc_maps;
@@ -1039,9 +1162,19 @@ static int show_numa_map(struct seq_file *m, void *v)
1039 seq_path(m, &file->f_path, "\n\t= "); 1162 seq_path(m, &file->f_path, "\n\t= ");
1040 } else if (vma->vm_start <= mm->brk && vma->vm_end >= mm->start_brk) { 1163 } else if (vma->vm_start <= mm->brk && vma->vm_end >= mm->start_brk) {
1041 seq_printf(m, " heap"); 1164 seq_printf(m, " heap");
1042 } else if (vma->vm_start <= mm->start_stack && 1165 } else {
1043 vma->vm_end >= mm->start_stack) { 1166 pid_t tid = vm_is_stack(proc_priv->task, vma, is_pid);
1044 seq_printf(m, " stack"); 1167 if (tid != 0) {
1168 /*
1169 * Thread stack in /proc/PID/task/TID/maps or
1170 * the main process stack.
1171 */
1172 if (!is_pid || (vma->vm_start <= mm->start_stack &&
1173 vma->vm_end >= mm->start_stack))
1174 seq_printf(m, " stack");
1175 else
1176 seq_printf(m, " stack:%d", tid);
1177 }
1045 } 1178 }
1046 1179
1047 if (is_vm_hugetlb_page(vma)) 1180 if (is_vm_hugetlb_page(vma))
@@ -1084,21 +1217,39 @@ out:
1084 return 0; 1217 return 0;
1085} 1218}
1086 1219
1220static int show_pid_numa_map(struct seq_file *m, void *v)
1221{
1222 return show_numa_map(m, v, 1);
1223}
1224
1225static int show_tid_numa_map(struct seq_file *m, void *v)
1226{
1227 return show_numa_map(m, v, 0);
1228}
1229
1087static const struct seq_operations proc_pid_numa_maps_op = { 1230static const struct seq_operations proc_pid_numa_maps_op = {
1088 .start = m_start, 1231 .start = m_start,
1089 .next = m_next, 1232 .next = m_next,
1090 .stop = m_stop, 1233 .stop = m_stop,
1091 .show = show_numa_map, 1234 .show = show_pid_numa_map,
1235};
1236
1237static const struct seq_operations proc_tid_numa_maps_op = {
1238 .start = m_start,
1239 .next = m_next,
1240 .stop = m_stop,
1241 .show = show_tid_numa_map,
1092}; 1242};
1093 1243
1094static int numa_maps_open(struct inode *inode, struct file *file) 1244static int numa_maps_open(struct inode *inode, struct file *file,
1245 const struct seq_operations *ops)
1095{ 1246{
1096 struct numa_maps_private *priv; 1247 struct numa_maps_private *priv;
1097 int ret = -ENOMEM; 1248 int ret = -ENOMEM;
1098 priv = kzalloc(sizeof(*priv), GFP_KERNEL); 1249 priv = kzalloc(sizeof(*priv), GFP_KERNEL);
1099 if (priv) { 1250 if (priv) {
1100 priv->proc_maps.pid = proc_pid(inode); 1251 priv->proc_maps.pid = proc_pid(inode);
1101 ret = seq_open(file, &proc_pid_numa_maps_op); 1252 ret = seq_open(file, ops);
1102 if (!ret) { 1253 if (!ret) {
1103 struct seq_file *m = file->private_data; 1254 struct seq_file *m = file->private_data;
1104 m->private = priv; 1255 m->private = priv;
@@ -1109,8 +1260,25 @@ static int numa_maps_open(struct inode *inode, struct file *file)
1109 return ret; 1260 return ret;
1110} 1261}
1111 1262
1112const struct file_operations proc_numa_maps_operations = { 1263static int pid_numa_maps_open(struct inode *inode, struct file *file)
1113 .open = numa_maps_open, 1264{
1265 return numa_maps_open(inode, file, &proc_pid_numa_maps_op);
1266}
1267
1268static int tid_numa_maps_open(struct inode *inode, struct file *file)
1269{
1270 return numa_maps_open(inode, file, &proc_tid_numa_maps_op);
1271}
1272
1273const struct file_operations proc_pid_numa_maps_operations = {
1274 .open = pid_numa_maps_open,
1275 .read = seq_read,
1276 .llseek = seq_lseek,
1277 .release = seq_release_private,
1278};
1279
1280const struct file_operations proc_tid_numa_maps_operations = {
1281 .open = tid_numa_maps_open,
1114 .read = seq_read, 1282 .read = seq_read,
1115 .llseek = seq_lseek, 1283 .llseek = seq_lseek,
1116 .release = seq_release_private, 1284 .release = seq_release_private,
diff --git a/fs/proc/task_nommu.c b/fs/proc/task_nommu.c
index 980de547c070..74fe164d1b23 100644
--- a/fs/proc/task_nommu.c
+++ b/fs/proc/task_nommu.c
@@ -134,9 +134,11 @@ static void pad_len_spaces(struct seq_file *m, int len)
134/* 134/*
135 * display a single VMA to a sequenced file 135 * display a single VMA to a sequenced file
136 */ 136 */
137static int nommu_vma_show(struct seq_file *m, struct vm_area_struct *vma) 137static int nommu_vma_show(struct seq_file *m, struct vm_area_struct *vma,
138 int is_pid)
138{ 139{
139 struct mm_struct *mm = vma->vm_mm; 140 struct mm_struct *mm = vma->vm_mm;
141 struct proc_maps_private *priv = m->private;
140 unsigned long ino = 0; 142 unsigned long ino = 0;
141 struct file *file; 143 struct file *file;
142 dev_t dev = 0; 144 dev_t dev = 0;
@@ -168,10 +170,19 @@ static int nommu_vma_show(struct seq_file *m, struct vm_area_struct *vma)
168 pad_len_spaces(m, len); 170 pad_len_spaces(m, len);
169 seq_path(m, &file->f_path, ""); 171 seq_path(m, &file->f_path, "");
170 } else if (mm) { 172 } else if (mm) {
171 if (vma->vm_start <= mm->start_stack && 173 pid_t tid = vm_is_stack(priv->task, vma, is_pid);
172 vma->vm_end >= mm->start_stack) { 174
175 if (tid != 0) {
173 pad_len_spaces(m, len); 176 pad_len_spaces(m, len);
174 seq_puts(m, "[stack]"); 177 /*
178 * Thread stack in /proc/PID/task/TID/maps or
179 * the main process stack.
180 */
181 if (!is_pid || (vma->vm_start <= mm->start_stack &&
182 vma->vm_end >= mm->start_stack))
183 seq_printf(m, "[stack]");
184 else
185 seq_printf(m, "[stack:%d]", tid);
175 } 186 }
176 } 187 }
177 188
@@ -182,11 +193,22 @@ static int nommu_vma_show(struct seq_file *m, struct vm_area_struct *vma)
182/* 193/*
183 * display mapping lines for a particular process's /proc/pid/maps 194 * display mapping lines for a particular process's /proc/pid/maps
184 */ 195 */
185static int show_map(struct seq_file *m, void *_p) 196static int show_map(struct seq_file *m, void *_p, int is_pid)
186{ 197{
187 struct rb_node *p = _p; 198 struct rb_node *p = _p;
188 199
189 return nommu_vma_show(m, rb_entry(p, struct vm_area_struct, vm_rb)); 200 return nommu_vma_show(m, rb_entry(p, struct vm_area_struct, vm_rb),
201 is_pid);
202}
203
204static int show_pid_map(struct seq_file *m, void *_p)
205{
206 return show_map(m, _p, 1);
207}
208
209static int show_tid_map(struct seq_file *m, void *_p)
210{
211 return show_map(m, _p, 0);
190} 212}
191 213
192static void *m_start(struct seq_file *m, loff_t *pos) 214static void *m_start(struct seq_file *m, loff_t *pos)
@@ -240,10 +262,18 @@ static const struct seq_operations proc_pid_maps_ops = {
240 .start = m_start, 262 .start = m_start,
241 .next = m_next, 263 .next = m_next,
242 .stop = m_stop, 264 .stop = m_stop,
243 .show = show_map 265 .show = show_pid_map
266};
267
268static const struct seq_operations proc_tid_maps_ops = {
269 .start = m_start,
270 .next = m_next,
271 .stop = m_stop,
272 .show = show_tid_map
244}; 273};
245 274
246static int maps_open(struct inode *inode, struct file *file) 275static int maps_open(struct inode *inode, struct file *file,
276 const struct seq_operations *ops)
247{ 277{
248 struct proc_maps_private *priv; 278 struct proc_maps_private *priv;
249 int ret = -ENOMEM; 279 int ret = -ENOMEM;
@@ -251,7 +281,7 @@ static int maps_open(struct inode *inode, struct file *file)
251 priv = kzalloc(sizeof(*priv), GFP_KERNEL); 281 priv = kzalloc(sizeof(*priv), GFP_KERNEL);
252 if (priv) { 282 if (priv) {
253 priv->pid = proc_pid(inode); 283 priv->pid = proc_pid(inode);
254 ret = seq_open(file, &proc_pid_maps_ops); 284 ret = seq_open(file, ops);
255 if (!ret) { 285 if (!ret) {
256 struct seq_file *m = file->private_data; 286 struct seq_file *m = file->private_data;
257 m->private = priv; 287 m->private = priv;
@@ -262,8 +292,25 @@ static int maps_open(struct inode *inode, struct file *file)
262 return ret; 292 return ret;
263} 293}
264 294
265const struct file_operations proc_maps_operations = { 295static int pid_maps_open(struct inode *inode, struct file *file)
266 .open = maps_open, 296{
297 return maps_open(inode, file, &proc_pid_maps_ops);
298}
299
300static int tid_maps_open(struct inode *inode, struct file *file)
301{
302 return maps_open(inode, file, &proc_tid_maps_ops);
303}
304
305const struct file_operations proc_pid_maps_operations = {
306 .open = pid_maps_open,
307 .read = seq_read,
308 .llseek = seq_lseek,
309 .release = seq_release_private,
310};
311
312const struct file_operations proc_tid_maps_operations = {
313 .open = tid_maps_open,
267 .read = seq_read, 314 .read = seq_read,
268 .llseek = seq_lseek, 315 .llseek = seq_lseek,
269 .release = seq_release_private, 316 .release = seq_release_private,
diff --git a/fs/proc/vmcore.c b/fs/proc/vmcore.c
index b0f450a2bb7c..0d5071d29985 100644
--- a/fs/proc/vmcore.c
+++ b/fs/proc/vmcore.c
@@ -700,3 +700,26 @@ static int __init vmcore_init(void)
700 return 0; 700 return 0;
701} 701}
702module_init(vmcore_init) 702module_init(vmcore_init)
703
704/* Cleanup function for vmcore module. */
705void vmcore_cleanup(void)
706{
707 struct list_head *pos, *next;
708
709 if (proc_vmcore) {
710 remove_proc_entry(proc_vmcore->name, proc_vmcore->parent);
711 proc_vmcore = NULL;
712 }
713
714 /* clear the vmcore list. */
715 list_for_each_safe(pos, next, &vmcore_list) {
716 struct vmcore *m;
717
718 m = list_entry(pos, struct vmcore, list);
719 list_del(&m->list);
720 kfree(m);
721 }
722 kfree(elfcorebuf);
723 elfcorebuf = NULL;
724}
725EXPORT_SYMBOL_GPL(vmcore_cleanup);
generated by cgit v1.2.2 (git 2.25.0) at 2025-12-25 14:19:27 -0500