This commit was manufactured by cvs2svn to create branch 'vserver'.
[linux-2.6.git] / kernel / sys.c
index e7a076d..4b471fd 100644 (file)
@@ -5,7 +5,6 @@
  */
 
 #include <linux/config.h>
-#include <linux/compat.h>
 #include <linux/module.h>
 #include <linux/mm.h>
 #include <linux/utsname.h>
 #include <linux/fs.h>
 #include <linux/workqueue.h>
 #include <linux/device.h>
+#include <linux/key.h>
 #include <linux/times.h>
 #include <linux/security.h>
 #include <linux/dcookies.h>
 #include <linux/suspend.h>
-#include <linux/vs_base.h>
+#include <linux/tty.h>
+
+#include <linux/compat.h>
+#include <linux/syscalls.h>
 #include <linux/vs_cvirt.h>
 
 #include <asm/uaccess.h>
@@ -88,7 +91,7 @@ int cad_pid = 1;
  */
 
 static struct notifier_block *reboot_notifier_list;
-rwlock_t notifier_lock = RW_LOCK_UNLOCKED;
+DEFINE_RWLOCK(notifier_lock);
 
 /**
  *     notifier_chain_register - Add notifier to a notifier chain
@@ -215,75 +218,6 @@ int unregister_reboot_notifier(struct notifier_block * nb)
 }
 
 EXPORT_SYMBOL(unregister_reboot_notifier);
-
-asmlinkage long sys_ni_syscall(void)
-{
-       return -ENOSYS;
-}
-
-cond_syscall(sys_nfsservctl)
-cond_syscall(sys_quotactl)
-cond_syscall(sys_acct)
-cond_syscall(sys_lookup_dcookie)
-cond_syscall(sys_swapon)
-cond_syscall(sys_swapoff)
-cond_syscall(sys_init_module)
-cond_syscall(sys_delete_module)
-cond_syscall(sys_socketpair)
-cond_syscall(sys_bind)
-cond_syscall(sys_listen)
-cond_syscall(sys_accept)
-cond_syscall(sys_connect)
-cond_syscall(sys_getsockname)
-cond_syscall(sys_getpeername)
-cond_syscall(sys_sendto)
-cond_syscall(sys_send)
-cond_syscall(sys_recvfrom)
-cond_syscall(sys_recv)
-cond_syscall(sys_socket)
-cond_syscall(sys_setsockopt)
-cond_syscall(sys_getsockopt)
-cond_syscall(sys_shutdown)
-cond_syscall(sys_sendmsg)
-cond_syscall(sys_recvmsg)
-cond_syscall(sys_socketcall)
-cond_syscall(sys_futex)
-cond_syscall(compat_sys_futex)
-cond_syscall(sys_epoll_create)
-cond_syscall(sys_epoll_ctl)
-cond_syscall(sys_epoll_wait)
-cond_syscall(sys_semget)
-cond_syscall(sys_semop)
-cond_syscall(sys_semtimedop)
-cond_syscall(sys_semctl)
-cond_syscall(sys_msgget)
-cond_syscall(sys_msgsnd)
-cond_syscall(sys_msgrcv)
-cond_syscall(sys_msgctl)
-cond_syscall(sys_shmget)
-cond_syscall(sys_shmdt)
-cond_syscall(sys_shmctl)
-cond_syscall(sys_mq_open)
-cond_syscall(sys_mq_unlink)
-cond_syscall(sys_mq_timedsend)
-cond_syscall(sys_mq_timedreceive)
-cond_syscall(sys_mq_notify)
-cond_syscall(sys_mq_getsetattr)
-cond_syscall(compat_sys_mq_open)
-cond_syscall(compat_sys_mq_timedsend)
-cond_syscall(compat_sys_mq_timedreceive)
-cond_syscall(compat_sys_mq_notify)
-cond_syscall(compat_sys_mq_getsetattr)
-cond_syscall(sys_mbind)
-cond_syscall(sys_get_mempolicy)
-cond_syscall(sys_set_mempolicy)
-cond_syscall(compat_get_mempolicy)
-
-/* arch-specific weak syscall entries */
-cond_syscall(sys_pciconfig_read)
-cond_syscall(sys_pciconfig_write)
-cond_syscall(sys_pciconfig_iobase)
-
 static int set_one_prio(struct task_struct *p, int niceval, int error)
 {
        int no_nice;
@@ -294,7 +228,10 @@ static int set_one_prio(struct task_struct *p, int niceval, int error)
                goto out;
        }
        if (niceval < task_nice(p) && !capable(CAP_SYS_NICE)) {
-               error = -EACCES;
+               if (vx_flags(VXF_IGNEG_NICE, 0))
+                       error = 0;
+               else
+                       error = -EACCES;
                goto out;
        }
        no_nice = security_task_setnice(p, niceval);
@@ -313,8 +250,6 @@ asmlinkage long sys_setpriority(int which, int who, int niceval)
 {
        struct task_struct *g, *p;
        struct user_struct *user;
-       struct pid *pid;
-       struct list_head *l;
        int error = -EINVAL;
 
        if (which > 2 || which < 0)
@@ -339,23 +274,24 @@ asmlinkage long sys_setpriority(int which, int who, int niceval)
                case PRIO_PGRP:
                        if (!who)
                                who = process_group(current);
-                       for_each_task_pid(who, PIDTYPE_PGID, p, l, pid)
+                       do_each_task_pid(who, PIDTYPE_PGID, p) {
                                error = set_one_prio(p, niceval, error);
+                       } while_each_task_pid(who, PIDTYPE_PGID, p);
                        break;
                case PRIO_USER:
+                       user = current->user;
                        if (!who)
-                               user = current->user;
+                               who = current->uid;
                        else
-                               user = find_user(vx_current_xid(), who);
-
-                       if (!user)
-                               goto out_unlock;
+                               if ((who != current->uid) &&
+                                       !(user = find_user(vx_current_xid(), who)))
+                                       goto out_unlock;        /* No processes for this user */
 
                        do_each_thread(g, p)
                                if (p->uid == who)
                                        error = set_one_prio(p, niceval, error);
                        while_each_thread(g, p);
-                       if (who)
+                       if (who != current->uid)
                                free_uid(user);         /* For find_user() */
                        break;
        }
@@ -374,8 +310,6 @@ out:
 asmlinkage long sys_getpriority(int which, int who)
 {
        struct task_struct *g, *p;
-       struct list_head *l;
-       struct pid *pid;
        struct user_struct *user;
        long niceval, retval = -ESRCH;
 
@@ -397,20 +331,20 @@ asmlinkage long sys_getpriority(int which, int who)
                case PRIO_PGRP:
                        if (!who)
                                who = process_group(current);
-                       for_each_task_pid(who, PIDTYPE_PGID, p, l, pid) {
+                       do_each_task_pid(who, PIDTYPE_PGID, p) {
                                niceval = 20 - task_nice(p);
                                if (niceval > retval)
                                        retval = niceval;
-                       }
+                       } while_each_task_pid(who, PIDTYPE_PGID, p);
                        break;
                case PRIO_USER:
+                       user = current->user;
                        if (!who)
-                               user = current->user;
+                               who = current->uid;
                        else
-                               user = find_user(vx_current_xid(), who);
-
-                       if (!user)
-                               goto out_unlock;
+                               if ((who != current->uid) &&
+                                       !(user = find_user(vx_current_xid(), who)))
+                                       goto out_unlock;        /* No processes for this user */
 
                        do_each_thread(g, p)
                                if (p->uid == who) {
@@ -419,7 +353,7 @@ asmlinkage long sys_getpriority(int which, int who)
                                                retval = niceval;
                                }
                        while_each_thread(g, p);
-                       if (who)
+                       if (who != current->uid)
                                free_uid(user);         /* for find_user() */
                        break;
        }
@@ -608,6 +542,7 @@ asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
        current->fsgid = new_egid;
        current->egid = new_egid;
        current->gid = new_rgid;
+       key_fsgid_changed(current);
        return 0;
 }
 
@@ -645,6 +580,8 @@ asmlinkage long sys_setgid(gid_t gid)
        }
        else
                return -EPERM;
+
+       key_fsgid_changed(current);
        return 0;
 }
   
@@ -657,7 +594,7 @@ static int set_user(uid_t new_ruid, int dumpclear)
                return -EAGAIN;
 
        if (atomic_read(&new_user->processes) >=
-                               current->rlim[RLIMIT_NPROC].rlim_cur &&
+                               current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
                        new_user != &root_user) {
                free_uid(new_user);
                return -EAGAIN;
@@ -733,6 +670,8 @@ asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
                current->suid = current->euid;
        current->fsuid = current->euid;
 
+       key_fsuid_changed(current);
+
        return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
 }
 
@@ -778,6 +717,8 @@ asmlinkage long sys_setuid(uid_t uid)
        current->fsuid = current->euid = uid;
        current->suid = new_suid;
 
+       key_fsuid_changed(current);
+
        return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
 }
 
@@ -824,6 +765,8 @@ asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
        if (suid != (uid_t) -1)
                current->suid = suid;
 
+       key_fsuid_changed(current);
+
        return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
 }
 
@@ -873,6 +816,8 @@ asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
                current->gid = rgid;
        if (sgid != (gid_t) -1)
                current->sgid = sgid;
+
+       key_fsgid_changed(current);
        return 0;
 }
 
@@ -914,6 +859,8 @@ asmlinkage long sys_setfsuid(uid_t uid)
                current->fsuid = uid;
        }
 
+       key_fsuid_changed(current);
+
        security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
 
        return old_fsuid;
@@ -940,6 +887,7 @@ asmlinkage long sys_setfsgid(gid_t gid)
                        wmb();
                }
                current->fsgid = gid;
+               key_fsgid_changed(current);
        }
        return old_fsgid;
 }
@@ -954,10 +902,39 @@ asmlinkage long sys_times(struct tms __user * tbuf)
         */
        if (tbuf) {
                struct tms tmp;
-               tmp.tms_utime = jiffies_to_clock_t(current->utime);
-               tmp.tms_stime = jiffies_to_clock_t(current->stime);
-               tmp.tms_cutime = jiffies_to_clock_t(current->cutime);
-               tmp.tms_cstime = jiffies_to_clock_t(current->cstime);
+               struct task_struct *tsk = current;
+               struct task_struct *t;
+               cputime_t utime, stime, cutime, cstime;
+
+               read_lock(&tasklist_lock);
+               utime = tsk->signal->utime;
+               stime = tsk->signal->stime;
+               t = tsk;
+               do {
+                       utime = cputime_add(utime, t->utime);
+                       stime = cputime_add(stime, t->stime);
+                       t = next_thread(t);
+               } while (t != tsk);
+
+               /*
+                * While we have tasklist_lock read-locked, no dying thread
+                * can be updating current->signal->[us]time.  Instead,
+                * we got their counts included in the live thread loop.
+                * However, another thread can come in right now and
+                * do a wait call that updates current->signal->c[us]time.
+                * To make sure we always see that pair updated atomically,
+                * we take the siglock around fetching them.
+                */
+               spin_lock_irq(&tsk->sighand->siglock);
+               cutime = tsk->signal->cutime;
+               cstime = tsk->signal->cstime;
+               spin_unlock_irq(&tsk->sighand->siglock);
+               read_unlock(&tasklist_lock);
+
+               tmp.tms_utime = cputime_to_clock_t(utime);
+               tmp.tms_stime = cputime_to_clock_t(stime);
+               tmp.tms_cutime = cputime_to_clock_t(cutime);
+               tmp.tms_cstime = cputime_to_clock_t(cstime);
                if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
                        return -EFAULT;
        }
@@ -981,14 +958,17 @@ asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
 {
        struct task_struct *p;
        int err = -EINVAL;
+       pid_t rpgid;
 
        if (!pid)
-               pid = current->pid;
+               pid = vx_map_pid(current->pid);
        if (!pgid)
                pgid = pid;
        if (pgid < 0)
                return -EINVAL;
 
+       rpgid = vx_rmap_pid(pgid);
+
        /* From this point forward we keep holding onto the tasklist lock
         * so that our parent does not change from under us. -DaveM
         */
@@ -1022,24 +1002,23 @@ asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
 
        if (pgid != pid) {
                struct task_struct *p;
-               struct pid *pid;
-               struct list_head *l;
 
-               for_each_task_pid(pgid, PIDTYPE_PGID, p, l, pid)
+               do_each_task_pid(rpgid, PIDTYPE_PGID, p) {
                        if (p->signal->session == current->signal->session)
                                goto ok_pgid;
+               } while_each_task_pid(rpgid, PIDTYPE_PGID, p);
                goto out;
        }
 
 ok_pgid:
-       err = security_task_setpgid(p, pgid);
+       err = security_task_setpgid(p, rpgid);
        if (err)
                goto out;
 
-       if (process_group(p) != pgid) {
+       if (process_group(p) != rpgid) {
                detach_pid(p, PIDTYPE_PGID);
-               p->signal->pgrp = pgid;
-               attach_pid(p, PIDTYPE_PGID, pgid);
+               p->signal->pgrp = rpgid;
+               attach_pid(p, PIDTYPE_PGID, rpgid);
        }
 
        err = 0;
@@ -1052,7 +1031,7 @@ out:
 asmlinkage long sys_getpgid(pid_t pid)
 {
        if (!pid) {
-               return process_group(current);
+               return vx_rmap_pid(process_group(current));
        } else {
                int retval;
                struct task_struct *p;
@@ -1064,7 +1043,7 @@ asmlinkage long sys_getpgid(pid_t pid)
                if (p) {
                        retval = security_task_getpgid(p);
                        if (!retval)
-                               retval = process_group(p);
+                               retval = vx_rmap_pid(process_group(p));
                }
                read_unlock(&tasklist_lock);
                return retval;
@@ -1111,6 +1090,7 @@ asmlinkage long sys_setsid(void)
        if (!thread_group_leader(current))
                return -EINVAL;
 
+       down(&tty_sem);
        write_lock_irq(&tasklist_lock);
 
        pid = find_pid(PIDTYPE_PGID, current->pid);
@@ -1124,6 +1104,7 @@ asmlinkage long sys_setsid(void)
        err = process_group(current);
 out:
        write_unlock_irq(&tasklist_lock);
+       up(&tty_sem);
        return err;
 }
 
@@ -1482,9 +1463,13 @@ asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
 {
        if (resource >= RLIM_NLIMITS)
                return -EINVAL;
-       else
-               return copy_to_user(rlim, current->rlim + resource, sizeof(*rlim))
-                       ? -EFAULT : 0;
+       else {
+               struct rlimit value;
+               task_lock(current->group_leader);
+               value = current->signal->rlim[resource];
+               task_unlock(current->group_leader);
+               return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
+       }
 }
 
 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
@@ -1499,7 +1484,9 @@ asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *r
        if (resource >= RLIM_NLIMITS)
                return -EINVAL;
 
-       memcpy(&x, current->rlim + resource, sizeof(*rlim));
+       task_lock(current->group_leader);
+       x = current->signal->rlim[resource];
+       task_unlock(current->group_leader);
        if(x.rlim_cur > 0x7FFFFFFF)
                x.rlim_cur = 0x7FFFFFFF;
        if(x.rlim_max > 0x7FFFFFFF)
@@ -1520,21 +1507,20 @@ asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
                return -EFAULT;
        if (new_rlim.rlim_cur > new_rlim.rlim_max)
                return -EINVAL;
-       old_rlim = current->rlim + resource;
-       if (((new_rlim.rlim_cur > old_rlim->rlim_max) ||
-            (new_rlim.rlim_max > old_rlim->rlim_max)) &&
+       old_rlim = current->signal->rlim + resource;
+       if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
            !capable(CAP_SYS_RESOURCE) && !vx_ccaps(VXC_SET_RLIMIT))
                return -EPERM;
-       if (resource == RLIMIT_NOFILE) {
-               if (new_rlim.rlim_cur > NR_OPEN || new_rlim.rlim_max > NR_OPEN)
+       if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
                        return -EPERM;
-       }
 
        retval = security_task_setrlimit(resource, &new_rlim);
        if (retval)
                return retval;
 
+       task_lock(current->group_leader);
        *old_rlim = new_rlim;
+       task_unlock(current->group_leader);
        return 0;
 }
 
@@ -1546,44 +1532,86 @@ asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
  * a lot simpler!  (Which we're not doing right now because we're not
  * measuring them yet).
  *
- * This is SMP safe.  Either we are called from sys_getrusage on ourselves
- * below (we know we aren't going to exit/disappear and only we change our
- * rusage counters), or we are called from wait4() on a process which is
- * either stopped or zombied.  In the zombied case the task won't get
- * reaped till shortly after the call to getrusage(), in both cases the
- * task being examined is in a frozen state so the counters won't change.
+ * This expects to be called with tasklist_lock read-locked or better,
+ * and the siglock not locked.  It may momentarily take the siglock.
+ *
+ * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
+ * races with threads incrementing their own counters.  But since word
+ * reads are atomic, we either get new values or old values and we don't
+ * care which for the sums.  We always take the siglock to protect reading
+ * the c* fields from p->signal from races with exit.c updating those
+ * fields when reaping, so a sample either gets all the additions of a
+ * given child after it's reaped, or none so this sample is before reaping.
  */
-int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
+
+void k_getrusage(struct task_struct *p, int who, struct rusage *r)
 {
-       struct rusage r;
+       struct task_struct *t;
+       unsigned long flags;
+       cputime_t utime, stime;
+
+       memset((char *) r, 0, sizeof *r);
+
+       if (unlikely(!p->signal))
+               return;
 
-       memset((char *) &r, 0, sizeof(r));
        switch (who) {
-               case RUSAGE_SELF:
-                       jiffies_to_timeval(p->utime, &r.ru_utime);
-                       jiffies_to_timeval(p->stime, &r.ru_stime);
-                       r.ru_nvcsw = p->nvcsw;
-                       r.ru_nivcsw = p->nivcsw;
-                       r.ru_minflt = p->min_flt;
-                       r.ru_majflt = p->maj_flt;
-                       break;
                case RUSAGE_CHILDREN:
-                       jiffies_to_timeval(p->cutime, &r.ru_utime);
-                       jiffies_to_timeval(p->cstime, &r.ru_stime);
-                       r.ru_nvcsw = p->cnvcsw;
-                       r.ru_nivcsw = p->cnivcsw;
-                       r.ru_minflt = p->cmin_flt;
-                       r.ru_majflt = p->cmaj_flt;
+                       spin_lock_irqsave(&p->sighand->siglock, flags);
+                       utime = p->signal->cutime;
+                       stime = p->signal->cstime;
+                       r->ru_nvcsw = p->signal->cnvcsw;
+                       r->ru_nivcsw = p->signal->cnivcsw;
+                       r->ru_minflt = p->signal->cmin_flt;
+                       r->ru_majflt = p->signal->cmaj_flt;
+                       spin_unlock_irqrestore(&p->sighand->siglock, flags);
+                       cputime_to_timeval(utime, &r->ru_utime);
+                       cputime_to_timeval(stime, &r->ru_stime);
                        break;
-               default:
-                       jiffies_to_timeval(p->utime + p->cutime, &r.ru_utime);
-                       jiffies_to_timeval(p->stime + p->cstime, &r.ru_stime);
-                       r.ru_nvcsw = p->nvcsw + p->cnvcsw;
-                       r.ru_nivcsw = p->nivcsw + p->cnivcsw;
-                       r.ru_minflt = p->min_flt + p->cmin_flt;
-                       r.ru_majflt = p->maj_flt + p->cmaj_flt;
+               case RUSAGE_SELF:
+                       spin_lock_irqsave(&p->sighand->siglock, flags);
+                       utime = stime = cputime_zero;
+                       goto sum_group;
+               case RUSAGE_BOTH:
+                       spin_lock_irqsave(&p->sighand->siglock, flags);
+                       utime = p->signal->cutime;
+                       stime = p->signal->cstime;
+                       r->ru_nvcsw = p->signal->cnvcsw;
+                       r->ru_nivcsw = p->signal->cnivcsw;
+                       r->ru_minflt = p->signal->cmin_flt;
+                       r->ru_majflt = p->signal->cmaj_flt;
+               sum_group:
+                       utime = cputime_add(utime, p->signal->utime);
+                       stime = cputime_add(stime, p->signal->stime);
+                       r->ru_nvcsw += p->signal->nvcsw;
+                       r->ru_nivcsw += p->signal->nivcsw;
+                       r->ru_minflt += p->signal->min_flt;
+                       r->ru_majflt += p->signal->maj_flt;
+                       t = p;
+                       do {
+                               utime = cputime_add(utime, t->utime);
+                               stime = cputime_add(stime, t->stime);
+                               r->ru_nvcsw += t->nvcsw;
+                               r->ru_nivcsw += t->nivcsw;
+                               r->ru_minflt += t->min_flt;
+                               r->ru_majflt += t->maj_flt;
+                               t = next_thread(t);
+                       } while (t != p);
+                       spin_unlock_irqrestore(&p->sighand->siglock, flags);
+                       cputime_to_timeval(utime, &r->ru_utime);
+                       cputime_to_timeval(stime, &r->ru_stime);
                        break;
+               default:
+                       BUG();
        }
+}
+
+int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
+{
+       struct rusage r;
+       read_lock(&tasklist_lock);
+       k_getrusage(p, who, &r);
+       read_unlock(&tasklist_lock);
        return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
 }
 
@@ -1603,7 +1631,7 @@ asmlinkage long sys_umask(int mask)
 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
                          unsigned long arg4, unsigned long arg5)
 {
-       int error;
+       long error;
        int sig;
 
        error = security_task_prctl(option, arg2, arg3, arg4, arg5);
@@ -1673,6 +1701,26 @@ asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
                        }
                        current->keep_capabilities = arg2;
                        break;
+               case PR_SET_NAME: {
+                       struct task_struct *me = current;
+                       unsigned char ncomm[sizeof(me->comm)];
+
+                       ncomm[sizeof(me->comm)-1] = 0;
+                       if (strncpy_from_user(ncomm, (char __user *)arg2,
+                                               sizeof(me->comm)-1) < 0)
+                               return -EFAULT;
+                       set_task_comm(me, ncomm);
+                       return 0;
+               }
+               case PR_GET_NAME: {
+                       struct task_struct *me = current;
+                       unsigned char tcomm[sizeof(me->comm)];
+
+                       get_task_comm(tcomm, me);
+                       if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))
+                               return -EFAULT;
+                       return 0;
+               }
                default:
                        error = -EINVAL;
                        break;