aboutsummaryrefslogtreecommitdiffstats
path: root/kernel/sys.c
diff options
context:
space:
mode:
Diffstat (limited to 'kernel/sys.c')
-rw-r--r--kernel/sys.c489
1 files changed, 342 insertions, 147 deletions
diff --git a/kernel/sys.c b/kernel/sys.c
index ce8129192a26..dfce4debd138 100644
--- a/kernel/sys.c
+++ b/kernel/sys.c
@@ -62,28 +62,28 @@
62#include <asm/unistd.h> 62#include <asm/unistd.h>
63 63
64#ifndef SET_UNALIGN_CTL 64#ifndef SET_UNALIGN_CTL
65# define SET_UNALIGN_CTL(a,b) (-EINVAL) 65# define SET_UNALIGN_CTL(a, b) (-EINVAL)
66#endif 66#endif
67#ifndef GET_UNALIGN_CTL 67#ifndef GET_UNALIGN_CTL
68# define GET_UNALIGN_CTL(a,b) (-EINVAL) 68# define GET_UNALIGN_CTL(a, b) (-EINVAL)
69#endif 69#endif
70#ifndef SET_FPEMU_CTL 70#ifndef SET_FPEMU_CTL
71# define SET_FPEMU_CTL(a,b) (-EINVAL) 71# define SET_FPEMU_CTL(a, b) (-EINVAL)
72#endif 72#endif
73#ifndef GET_FPEMU_CTL 73#ifndef GET_FPEMU_CTL
74# define GET_FPEMU_CTL(a,b) (-EINVAL) 74# define GET_FPEMU_CTL(a, b) (-EINVAL)
75#endif 75#endif
76#ifndef SET_FPEXC_CTL 76#ifndef SET_FPEXC_CTL
77# define SET_FPEXC_CTL(a,b) (-EINVAL) 77# define SET_FPEXC_CTL(a, b) (-EINVAL)
78#endif 78#endif
79#ifndef GET_FPEXC_CTL 79#ifndef GET_FPEXC_CTL
80# define GET_FPEXC_CTL(a,b) (-EINVAL) 80# define GET_FPEXC_CTL(a, b) (-EINVAL)
81#endif 81#endif
82#ifndef GET_ENDIAN 82#ifndef GET_ENDIAN
83# define GET_ENDIAN(a,b) (-EINVAL) 83# define GET_ENDIAN(a, b) (-EINVAL)
84#endif 84#endif
85#ifndef SET_ENDIAN 85#ifndef SET_ENDIAN
86# define SET_ENDIAN(a,b) (-EINVAL) 86# define SET_ENDIAN(a, b) (-EINVAL)
87#endif 87#endif
88#ifndef GET_TSC_CTL 88#ifndef GET_TSC_CTL
89# define GET_TSC_CTL(a) (-EINVAL) 89# define GET_TSC_CTL(a) (-EINVAL)
@@ -182,39 +182,40 @@ SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
182 rcu_read_lock(); 182 rcu_read_lock();
183 read_lock(&tasklist_lock); 183 read_lock(&tasklist_lock);
184 switch (which) { 184 switch (which) {
185 case PRIO_PROCESS: 185 case PRIO_PROCESS:
186 if (who) 186 if (who)
187 p = find_task_by_vpid(who); 187 p = find_task_by_vpid(who);
188 else 188 else
189 p = current; 189 p = current;
190 if (p) 190 if (p)
191 error = set_one_prio(p, niceval, error); 191 error = set_one_prio(p, niceval, error);
192 break; 192 break;
193 case PRIO_PGRP: 193 case PRIO_PGRP:
194 if (who) 194 if (who)
195 pgrp = find_vpid(who); 195 pgrp = find_vpid(who);
196 else 196 else
197 pgrp = task_pgrp(current); 197 pgrp = task_pgrp(current);
198 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) { 198 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
199 error = set_one_prio(p, niceval, error); 199 error = set_one_prio(p, niceval, error);
200 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p); 200 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
201 break; 201 break;
202 case PRIO_USER: 202 case PRIO_USER:
203 uid = make_kuid(cred->user_ns, who); 203 uid = make_kuid(cred->user_ns, who);
204 user = cred->user; 204 user = cred->user;
205 if (!who) 205 if (!who)
206 uid = cred->uid; 206 uid = cred->uid;
207 else if (!uid_eq(uid, cred->uid) && 207 else if (!uid_eq(uid, cred->uid)) {
208 !(user = find_user(uid))) 208 user = find_user(uid);
209 if (!user)
209 goto out_unlock; /* No processes for this user */ 210 goto out_unlock; /* No processes for this user */
210 211 }
211 do_each_thread(g, p) { 212 do_each_thread(g, p) {
212 if (uid_eq(task_uid(p), uid)) 213 if (uid_eq(task_uid(p), uid))
213 error = set_one_prio(p, niceval, error); 214 error = set_one_prio(p, niceval, error);
214 } while_each_thread(g, p); 215 } while_each_thread(g, p);
215 if (!uid_eq(uid, cred->uid)) 216 if (!uid_eq(uid, cred->uid))
216 free_uid(user); /* For find_user() */ 217 free_uid(user); /* For find_user() */
217 break; 218 break;
218 } 219 }
219out_unlock: 220out_unlock:
220 read_unlock(&tasklist_lock); 221 read_unlock(&tasklist_lock);
@@ -244,47 +245,48 @@ SYSCALL_DEFINE2(getpriority, int, which, int, who)
244 rcu_read_lock(); 245 rcu_read_lock();
245 read_lock(&tasklist_lock); 246 read_lock(&tasklist_lock);
246 switch (which) { 247 switch (which) {
247 case PRIO_PROCESS: 248 case PRIO_PROCESS:
248 if (who) 249 if (who)
249 p = find_task_by_vpid(who); 250 p = find_task_by_vpid(who);
250 else 251 else
251 p = current; 252 p = current;
252 if (p) { 253 if (p) {
254 niceval = nice_to_rlimit(task_nice(p));
255 if (niceval > retval)
256 retval = niceval;
257 }
258 break;
259 case PRIO_PGRP:
260 if (who)
261 pgrp = find_vpid(who);
262 else
263 pgrp = task_pgrp(current);
264 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
265 niceval = nice_to_rlimit(task_nice(p));
266 if (niceval > retval)
267 retval = niceval;
268 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
269 break;
270 case PRIO_USER:
271 uid = make_kuid(cred->user_ns, who);
272 user = cred->user;
273 if (!who)
274 uid = cred->uid;
275 else if (!uid_eq(uid, cred->uid)) {
276 user = find_user(uid);
277 if (!user)
278 goto out_unlock; /* No processes for this user */
279 }
280 do_each_thread(g, p) {
281 if (uid_eq(task_uid(p), uid)) {
253 niceval = nice_to_rlimit(task_nice(p)); 282 niceval = nice_to_rlimit(task_nice(p));
254 if (niceval > retval) 283 if (niceval > retval)
255 retval = niceval; 284 retval = niceval;
256 } 285 }
257 break; 286 } while_each_thread(g, p);
258 case PRIO_PGRP: 287 if (!uid_eq(uid, cred->uid))
259 if (who) 288 free_uid(user); /* for find_user() */
260 pgrp = find_vpid(who); 289 break;
261 else
262 pgrp = task_pgrp(current);
263 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
264 niceval = nice_to_rlimit(task_nice(p));
265 if (niceval > retval)
266 retval = niceval;
267 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
268 break;
269 case PRIO_USER:
270 uid = make_kuid(cred->user_ns, who);
271 user = cred->user;
272 if (!who)
273 uid = cred->uid;
274 else if (!uid_eq(uid, cred->uid) &&
275 !(user = find_user(uid)))
276 goto out_unlock; /* No processes for this user */
277
278 do_each_thread(g, p) {
279 if (uid_eq(task_uid(p), uid)) {
280 niceval = nice_to_rlimit(task_nice(p));
281 if (niceval > retval)
282 retval = niceval;
283 }
284 } while_each_thread(g, p);
285 if (!uid_eq(uid, cred->uid))
286 free_uid(user); /* for find_user() */
287 break;
288 } 290 }
289out_unlock: 291out_unlock:
290 read_unlock(&tasklist_lock); 292 read_unlock(&tasklist_lock);
@@ -306,7 +308,7 @@ out_unlock:
306 * 308 *
307 * The general idea is that a program which uses just setregid() will be 309 * The general idea is that a program which uses just setregid() will be
308 * 100% compatible with BSD. A program which uses just setgid() will be 310 * 100% compatible with BSD. A program which uses just setgid() will be
309 * 100% compatible with POSIX with saved IDs. 311 * 100% compatible with POSIX with saved IDs.
310 * 312 *
311 * SMP: There are not races, the GIDs are checked only by filesystem 313 * SMP: There are not races, the GIDs are checked only by filesystem
312 * operations (as far as semantic preservation is concerned). 314 * operations (as far as semantic preservation is concerned).
@@ -364,7 +366,7 @@ error:
364} 366}
365 367
366/* 368/*
367 * setgid() is implemented like SysV w/ SAVED_IDS 369 * setgid() is implemented like SysV w/ SAVED_IDS
368 * 370 *
369 * SMP: Same implicit races as above. 371 * SMP: Same implicit races as above.
370 */ 372 */
@@ -442,7 +444,7 @@ static int set_user(struct cred *new)
442 * 444 *
443 * The general idea is that a program which uses just setreuid() will be 445 * The general idea is that a program which uses just setreuid() will be
444 * 100% compatible with BSD. A program which uses just setuid() will be 446 * 100% compatible with BSD. A program which uses just setuid() will be
445 * 100% compatible with POSIX with saved IDs. 447 * 100% compatible with POSIX with saved IDs.
446 */ 448 */
447SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid) 449SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
448{ 450{
@@ -503,17 +505,17 @@ error:
503 abort_creds(new); 505 abort_creds(new);
504 return retval; 506 return retval;
505} 507}
506 508
507/* 509/*
508 * setuid() is implemented like SysV with SAVED_IDS 510 * setuid() is implemented like SysV with SAVED_IDS
509 * 511 *
510 * Note that SAVED_ID's is deficient in that a setuid root program 512 * Note that SAVED_ID's is deficient in that a setuid root program
511 * like sendmail, for example, cannot set its uid to be a normal 513 * like sendmail, for example, cannot set its uid to be a normal
512 * user and then switch back, because if you're root, setuid() sets 514 * user and then switch back, because if you're root, setuid() sets
513 * the saved uid too. If you don't like this, blame the bright people 515 * the saved uid too. If you don't like this, blame the bright people
514 * in the POSIX committee and/or USG. Note that the BSD-style setreuid() 516 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
515 * will allow a root program to temporarily drop privileges and be able to 517 * will allow a root program to temporarily drop privileges and be able to
516 * regain them by swapping the real and effective uid. 518 * regain them by swapping the real and effective uid.
517 */ 519 */
518SYSCALL_DEFINE1(setuid, uid_t, uid) 520SYSCALL_DEFINE1(setuid, uid_t, uid)
519{ 521{
@@ -637,10 +639,12 @@ SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t _
637 euid = from_kuid_munged(cred->user_ns, cred->euid); 639 euid = from_kuid_munged(cred->user_ns, cred->euid);
638 suid = from_kuid_munged(cred->user_ns, cred->suid); 640 suid = from_kuid_munged(cred->user_ns, cred->suid);
639 641
640 if (!(retval = put_user(ruid, ruidp)) && 642 retval = put_user(ruid, ruidp);
641 !(retval = put_user(euid, euidp))) 643 if (!retval) {
642 retval = put_user(suid, suidp); 644 retval = put_user(euid, euidp);
643 645 if (!retval)
646 return put_user(suid, suidp);
647 }
644 return retval; 648 return retval;
645} 649}
646 650
@@ -709,9 +713,12 @@ SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t _
709 egid = from_kgid_munged(cred->user_ns, cred->egid); 713 egid = from_kgid_munged(cred->user_ns, cred->egid);
710 sgid = from_kgid_munged(cred->user_ns, cred->sgid); 714 sgid = from_kgid_munged(cred->user_ns, cred->sgid);
711 715
712 if (!(retval = put_user(rgid, rgidp)) && 716 retval = put_user(rgid, rgidp);
713 !(retval = put_user(egid, egidp))) 717 if (!retval) {
714 retval = put_user(sgid, sgidp); 718 retval = put_user(egid, egidp);
719 if (!retval)
720 retval = put_user(sgid, sgidp);
721 }
715 722
716 return retval; 723 return retval;
717} 724}
@@ -1284,7 +1291,6 @@ SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1284/* 1291/*
1285 * Back compatibility for getrlimit. Needed for some apps. 1292 * Back compatibility for getrlimit. Needed for some apps.
1286 */ 1293 */
1287
1288SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource, 1294SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1289 struct rlimit __user *, rlim) 1295 struct rlimit __user *, rlim)
1290{ 1296{
@@ -1299,7 +1305,7 @@ SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1299 x.rlim_cur = 0x7FFFFFFF; 1305 x.rlim_cur = 0x7FFFFFFF;
1300 if (x.rlim_max > 0x7FFFFFFF) 1306 if (x.rlim_max > 0x7FFFFFFF)
1301 x.rlim_max = 0x7FFFFFFF; 1307 x.rlim_max = 0x7FFFFFFF;
1302 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0; 1308 return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0;
1303} 1309}
1304 1310
1305#endif 1311#endif
@@ -1527,7 +1533,7 @@ static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1527 cputime_t tgutime, tgstime, utime, stime; 1533 cputime_t tgutime, tgstime, utime, stime;
1528 unsigned long maxrss = 0; 1534 unsigned long maxrss = 0;
1529 1535
1530 memset((char *) r, 0, sizeof *r); 1536 memset((char *)r, 0, sizeof (*r));
1531 utime = stime = 0; 1537 utime = stime = 0;
1532 1538
1533 if (who == RUSAGE_THREAD) { 1539 if (who == RUSAGE_THREAD) {
@@ -1541,41 +1547,41 @@ static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1541 return; 1547 return;
1542 1548
1543 switch (who) { 1549 switch (who) {
1544 case RUSAGE_BOTH: 1550 case RUSAGE_BOTH:
1545 case RUSAGE_CHILDREN: 1551 case RUSAGE_CHILDREN:
1546 utime = p->signal->cutime; 1552 utime = p->signal->cutime;
1547 stime = p->signal->cstime; 1553 stime = p->signal->cstime;
1548 r->ru_nvcsw = p->signal->cnvcsw; 1554 r->ru_nvcsw = p->signal->cnvcsw;
1549 r->ru_nivcsw = p->signal->cnivcsw; 1555 r->ru_nivcsw = p->signal->cnivcsw;
1550 r->ru_minflt = p->signal->cmin_flt; 1556 r->ru_minflt = p->signal->cmin_flt;
1551 r->ru_majflt = p->signal->cmaj_flt; 1557 r->ru_majflt = p->signal->cmaj_flt;
1552 r->ru_inblock = p->signal->cinblock; 1558 r->ru_inblock = p->signal->cinblock;
1553 r->ru_oublock = p->signal->coublock; 1559 r->ru_oublock = p->signal->coublock;
1554 maxrss = p->signal->cmaxrss; 1560 maxrss = p->signal->cmaxrss;
1555 1561
1556 if (who == RUSAGE_CHILDREN) 1562 if (who == RUSAGE_CHILDREN)
1557 break;
1558
1559 case RUSAGE_SELF:
1560 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1561 utime += tgutime;
1562 stime += tgstime;
1563 r->ru_nvcsw += p->signal->nvcsw;
1564 r->ru_nivcsw += p->signal->nivcsw;
1565 r->ru_minflt += p->signal->min_flt;
1566 r->ru_majflt += p->signal->maj_flt;
1567 r->ru_inblock += p->signal->inblock;
1568 r->ru_oublock += p->signal->oublock;
1569 if (maxrss < p->signal->maxrss)
1570 maxrss = p->signal->maxrss;
1571 t = p;
1572 do {
1573 accumulate_thread_rusage(t, r);
1574 } while_each_thread(p, t);
1575 break; 1563 break;
1576 1564
1577 default: 1565 case RUSAGE_SELF:
1578 BUG(); 1566 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1567 utime += tgutime;
1568 stime += tgstime;
1569 r->ru_nvcsw += p->signal->nvcsw;
1570 r->ru_nivcsw += p->signal->nivcsw;
1571 r->ru_minflt += p->signal->min_flt;
1572 r->ru_majflt += p->signal->maj_flt;
1573 r->ru_inblock += p->signal->inblock;
1574 r->ru_oublock += p->signal->oublock;
1575 if (maxrss < p->signal->maxrss)
1576 maxrss = p->signal->maxrss;
1577 t = p;
1578 do {
1579 accumulate_thread_rusage(t, r);
1580 } while_each_thread(p, t);
1581 break;
1582
1583 default:
1584 BUG();
1579 } 1585 }
1580 unlock_task_sighand(p, &flags); 1586 unlock_task_sighand(p, &flags);
1581 1587
@@ -1585,6 +1591,7 @@ out:
1585 1591
1586 if (who != RUSAGE_CHILDREN) { 1592 if (who != RUSAGE_CHILDREN) {
1587 struct mm_struct *mm = get_task_mm(p); 1593 struct mm_struct *mm = get_task_mm(p);
1594
1588 if (mm) { 1595 if (mm) {
1589 setmax_mm_hiwater_rss(&maxrss, mm); 1596 setmax_mm_hiwater_rss(&maxrss, mm);
1590 mmput(mm); 1597 mmput(mm);
@@ -1596,6 +1603,7 @@ out:
1596int getrusage(struct task_struct *p, int who, struct rusage __user *ru) 1603int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1597{ 1604{
1598 struct rusage r; 1605 struct rusage r;
1606
1599 k_getrusage(p, who, &r); 1607 k_getrusage(p, who, &r);
1600 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0; 1608 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1601} 1609}
@@ -1628,12 +1636,14 @@ SYSCALL_DEFINE1(umask, int, mask)
1628 return mask; 1636 return mask;
1629} 1637}
1630 1638
1631static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd) 1639static int prctl_set_mm_exe_file_locked(struct mm_struct *mm, unsigned int fd)
1632{ 1640{
1633 struct fd exe; 1641 struct fd exe;
1634 struct inode *inode; 1642 struct inode *inode;
1635 int err; 1643 int err;
1636 1644
1645 VM_BUG_ON_MM(!rwsem_is_locked(&mm->mmap_sem), mm);
1646
1637 exe = fdget(fd); 1647 exe = fdget(fd);
1638 if (!exe.file) 1648 if (!exe.file)
1639 return -EBADF; 1649 return -EBADF;
@@ -1654,8 +1664,6 @@ static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1654 if (err) 1664 if (err)
1655 goto exit; 1665 goto exit;
1656 1666
1657 down_write(&mm->mmap_sem);
1658
1659 /* 1667 /*
1660 * Forbid mm->exe_file change if old file still mapped. 1668 * Forbid mm->exe_file change if old file still mapped.
1661 */ 1669 */
@@ -1667,7 +1675,7 @@ static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1667 if (vma->vm_file && 1675 if (vma->vm_file &&
1668 path_equal(&vma->vm_file->f_path, 1676 path_equal(&vma->vm_file->f_path,
1669 &mm->exe_file->f_path)) 1677 &mm->exe_file->f_path))
1670 goto exit_unlock; 1678 goto exit;
1671 } 1679 }
1672 1680
1673 /* 1681 /*
@@ -1678,34 +1686,222 @@ static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1678 */ 1686 */
1679 err = -EPERM; 1687 err = -EPERM;
1680 if (test_and_set_bit(MMF_EXE_FILE_CHANGED, &mm->flags)) 1688 if (test_and_set_bit(MMF_EXE_FILE_CHANGED, &mm->flags))
1681 goto exit_unlock; 1689 goto exit;
1682 1690
1683 err = 0; 1691 err = 0;
1684 set_mm_exe_file(mm, exe.file); /* this grabs a reference to exe.file */ 1692 set_mm_exe_file(mm, exe.file); /* this grabs a reference to exe.file */
1685exit_unlock:
1686 up_write(&mm->mmap_sem);
1687
1688exit: 1693exit:
1689 fdput(exe); 1694 fdput(exe);
1690 return err; 1695 return err;
1691} 1696}
1692 1697
1698#ifdef CONFIG_CHECKPOINT_RESTORE
1699/*
1700 * WARNING: we don't require any capability here so be very careful
1701 * in what is allowed for modification from userspace.
1702 */
1703static int validate_prctl_map(struct prctl_mm_map *prctl_map)
1704{
1705 unsigned long mmap_max_addr = TASK_SIZE;
1706 struct mm_struct *mm = current->mm;
1707 int error = -EINVAL, i;
1708
1709 static const unsigned char offsets[] = {
1710 offsetof(struct prctl_mm_map, start_code),
1711 offsetof(struct prctl_mm_map, end_code),
1712 offsetof(struct prctl_mm_map, start_data),
1713 offsetof(struct prctl_mm_map, end_data),
1714 offsetof(struct prctl_mm_map, start_brk),
1715 offsetof(struct prctl_mm_map, brk),
1716 offsetof(struct prctl_mm_map, start_stack),
1717 offsetof(struct prctl_mm_map, arg_start),
1718 offsetof(struct prctl_mm_map, arg_end),
1719 offsetof(struct prctl_mm_map, env_start),
1720 offsetof(struct prctl_mm_map, env_end),
1721 };
1722
1723 /*
1724 * Make sure the members are not somewhere outside
1725 * of allowed address space.
1726 */
1727 for (i = 0; i < ARRAY_SIZE(offsets); i++) {
1728 u64 val = *(u64 *)((char *)prctl_map + offsets[i]);
1729
1730 if ((unsigned long)val >= mmap_max_addr ||
1731 (unsigned long)val < mmap_min_addr)
1732 goto out;
1733 }
1734
1735 /*
1736 * Make sure the pairs are ordered.
1737 */
1738#define __prctl_check_order(__m1, __op, __m2) \
1739 ((unsigned long)prctl_map->__m1 __op \
1740 (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1741 error = __prctl_check_order(start_code, <, end_code);
1742 error |= __prctl_check_order(start_data, <, end_data);
1743 error |= __prctl_check_order(start_brk, <=, brk);
1744 error |= __prctl_check_order(arg_start, <=, arg_end);
1745 error |= __prctl_check_order(env_start, <=, env_end);
1746 if (error)
1747 goto out;
1748#undef __prctl_check_order
1749
1750 error = -EINVAL;
1751
1752 /*
1753 * @brk should be after @end_data in traditional maps.
1754 */
1755 if (prctl_map->start_brk <= prctl_map->end_data ||
1756 prctl_map->brk <= prctl_map->end_data)
1757 goto out;
1758
1759 /*
1760 * Neither we should allow to override limits if they set.
1761 */
1762 if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk,
1763 prctl_map->start_brk, prctl_map->end_data,
1764 prctl_map->start_data))
1765 goto out;
1766
1767 /*
1768 * Someone is trying to cheat the auxv vector.
1769 */
1770 if (prctl_map->auxv_size) {
1771 if (!prctl_map->auxv || prctl_map->auxv_size > sizeof(mm->saved_auxv))
1772 goto out;
1773 }
1774
1775 /*
1776 * Finally, make sure the caller has the rights to
1777 * change /proc/pid/exe link: only local root should
1778 * be allowed to.
1779 */
1780 if (prctl_map->exe_fd != (u32)-1) {
1781 struct user_namespace *ns = current_user_ns();
1782 const struct cred *cred = current_cred();
1783
1784 if (!uid_eq(cred->uid, make_kuid(ns, 0)) ||
1785 !gid_eq(cred->gid, make_kgid(ns, 0)))
1786 goto out;
1787 }
1788
1789 error = 0;
1790out:
1791 return error;
1792}
1793
1794static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size)
1795{
1796 struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, };
1797 unsigned long user_auxv[AT_VECTOR_SIZE];
1798 struct mm_struct *mm = current->mm;
1799 int error;
1800
1801 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
1802 BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256);
1803
1804 if (opt == PR_SET_MM_MAP_SIZE)
1805 return put_user((unsigned int)sizeof(prctl_map),
1806 (unsigned int __user *)addr);
1807
1808 if (data_size != sizeof(prctl_map))
1809 return -EINVAL;
1810
1811 if (copy_from_user(&prctl_map, addr, sizeof(prctl_map)))
1812 return -EFAULT;
1813
1814 error = validate_prctl_map(&prctl_map);
1815 if (error)
1816 return error;
1817
1818 if (prctl_map.auxv_size) {
1819 memset(user_auxv, 0, sizeof(user_auxv));
1820 if (copy_from_user(user_auxv,
1821 (const void __user *)prctl_map.auxv,
1822 prctl_map.auxv_size))
1823 return -EFAULT;
1824
1825 /* Last entry must be AT_NULL as specification requires */
1826 user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL;
1827 user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL;
1828 }
1829
1830 down_write(&mm->mmap_sem);
1831 if (prctl_map.exe_fd != (u32)-1)
1832 error = prctl_set_mm_exe_file_locked(mm, prctl_map.exe_fd);
1833 downgrade_write(&mm->mmap_sem);
1834 if (error)
1835 goto out;
1836
1837 /*
1838 * We don't validate if these members are pointing to
1839 * real present VMAs because application may have correspond
1840 * VMAs already unmapped and kernel uses these members for statistics
1841 * output in procfs mostly, except
1842 *
1843 * - @start_brk/@brk which are used in do_brk but kernel lookups
1844 * for VMAs when updating these memvers so anything wrong written
1845 * here cause kernel to swear at userspace program but won't lead
1846 * to any problem in kernel itself
1847 */
1848
1849 mm->start_code = prctl_map.start_code;
1850 mm->end_code = prctl_map.end_code;
1851 mm->start_data = prctl_map.start_data;
1852 mm->end_data = prctl_map.end_data;
1853 mm->start_brk = prctl_map.start_brk;
1854 mm->brk = prctl_map.brk;
1855 mm->start_stack = prctl_map.start_stack;
1856 mm->arg_start = prctl_map.arg_start;
1857 mm->arg_end = prctl_map.arg_end;
1858 mm->env_start = prctl_map.env_start;
1859 mm->env_end = prctl_map.env_end;
1860
1861 /*
1862 * Note this update of @saved_auxv is lockless thus
1863 * if someone reads this member in procfs while we're
1864 * updating -- it may get partly updated results. It's
1865 * known and acceptable trade off: we leave it as is to
1866 * not introduce additional locks here making the kernel
1867 * more complex.
1868 */
1869 if (prctl_map.auxv_size)
1870 memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv));
1871
1872 error = 0;
1873out:
1874 up_read(&mm->mmap_sem);
1875 return error;
1876}
1877#endif /* CONFIG_CHECKPOINT_RESTORE */
1878
1693static int prctl_set_mm(int opt, unsigned long addr, 1879static int prctl_set_mm(int opt, unsigned long addr,
1694 unsigned long arg4, unsigned long arg5) 1880 unsigned long arg4, unsigned long arg5)
1695{ 1881{
1696 unsigned long rlim = rlimit(RLIMIT_DATA);
1697 struct mm_struct *mm = current->mm; 1882 struct mm_struct *mm = current->mm;
1698 struct vm_area_struct *vma; 1883 struct vm_area_struct *vma;
1699 int error; 1884 int error;
1700 1885
1701 if (arg5 || (arg4 && opt != PR_SET_MM_AUXV)) 1886 if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV &&
1887 opt != PR_SET_MM_MAP &&
1888 opt != PR_SET_MM_MAP_SIZE)))
1702 return -EINVAL; 1889 return -EINVAL;
1703 1890
1891#ifdef CONFIG_CHECKPOINT_RESTORE
1892 if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE)
1893 return prctl_set_mm_map(opt, (const void __user *)addr, arg4);
1894#endif
1895
1704 if (!capable(CAP_SYS_RESOURCE)) 1896 if (!capable(CAP_SYS_RESOURCE))
1705 return -EPERM; 1897 return -EPERM;
1706 1898
1707 if (opt == PR_SET_MM_EXE_FILE) 1899 if (opt == PR_SET_MM_EXE_FILE) {
1708 return prctl_set_mm_exe_file(mm, (unsigned int)addr); 1900 down_write(&mm->mmap_sem);
1901 error = prctl_set_mm_exe_file_locked(mm, (unsigned int)addr);
1902 up_write(&mm->mmap_sem);
1903 return error;
1904 }
1709 1905
1710 if (addr >= TASK_SIZE || addr < mmap_min_addr) 1906 if (addr >= TASK_SIZE || addr < mmap_min_addr)
1711 return -EINVAL; 1907 return -EINVAL;
@@ -1733,9 +1929,8 @@ static int prctl_set_mm(int opt, unsigned long addr,
1733 if (addr <= mm->end_data) 1929 if (addr <= mm->end_data)
1734 goto out; 1930 goto out;
1735 1931
1736 if (rlim < RLIM_INFINITY && 1932 if (check_data_rlimit(rlimit(RLIMIT_DATA), mm->brk, addr,
1737 (mm->brk - addr) + 1933 mm->end_data, mm->start_data))
1738 (mm->end_data - mm->start_data) > rlim)
1739 goto out; 1934 goto out;
1740 1935
1741 mm->start_brk = addr; 1936 mm->start_brk = addr;
@@ -1745,9 +1940,8 @@ static int prctl_set_mm(int opt, unsigned long addr,
1745 if (addr <= mm->end_data) 1940 if (addr <= mm->end_data)
1746 goto out; 1941 goto out;
1747 1942
1748 if (rlim < RLIM_INFINITY && 1943 if (check_data_rlimit(rlimit(RLIMIT_DATA), addr, mm->start_brk,
1749 (addr - mm->start_brk) + 1944 mm->end_data, mm->start_data))
1750 (mm->end_data - mm->start_data) > rlim)
1751 goto out; 1945 goto out;
1752 1946
1753 mm->brk = addr; 1947 mm->brk = addr;
@@ -2023,6 +2217,7 @@ SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2023{ 2217{
2024 int err = 0; 2218 int err = 0;
2025 int cpu = raw_smp_processor_id(); 2219 int cpu = raw_smp_processor_id();
2220
2026 if (cpup) 2221 if (cpup)
2027 err |= put_user(cpu, cpup); 2222 err |= put_user(cpu, cpup);
2028 if (nodep) 2223 if (nodep)
@@ -2135,7 +2330,7 @@ COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2135 /* Check to see if any memory value is too large for 32-bit and scale 2330 /* Check to see if any memory value is too large for 32-bit and scale
2136 * down if needed 2331 * down if needed
2137 */ 2332 */
2138 if ((s.totalram >> 32) || (s.totalswap >> 32)) { 2333 if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) {
2139 int bitcount = 0; 2334 int bitcount = 0;
2140 2335
2141 while (s.mem_unit < PAGE_SIZE) { 2336 while (s.mem_unit < PAGE_SIZE) {