mm: more likely reclaim MADV_SEQUENTIAL mappings
[linux-2.6.git] / kernel / sys.c
index 18987c7..d356d79 100644 (file)
@@ -32,6 +32,7 @@
 #include <linux/getcpu.h>
 #include <linux/task_io_accounting_ops.h>
 #include <linux/seccomp.h>
+#include <linux/cpu.h>
 
 #include <linux/compat.h>
 #include <linux/syscalls.h>
 #ifndef SET_ENDIAN
 # define SET_ENDIAN(a,b)       (-EINVAL)
 #endif
+#ifndef GET_TSC_CTL
+# define GET_TSC_CTL(a)                (-EINVAL)
+#endif
+#ifndef SET_TSC_CTL
+# define SET_TSC_CTL(a)                (-EINVAL)
+#endif
 
 /*
  * this is where the system-wide overflow UID and GID are defined, for
@@ -100,542 +107,22 @@ struct pid *cad_pid;
 EXPORT_SYMBOL(cad_pid);
 
 /*
- *     Notifier list for kernel code which wants to be called
- *     at shutdown. This is used to stop any idling DMA operations
- *     and the like. 
- */
-
-static BLOCKING_NOTIFIER_HEAD(reboot_notifier_list);
-
-/*
- *     Notifier chain core routines.  The exported routines below
- *     are layered on top of these, with appropriate locking added.
- */
-
-static int notifier_chain_register(struct notifier_block **nl,
-               struct notifier_block *n)
-{
-       while ((*nl) != NULL) {
-               if (n->priority > (*nl)->priority)
-                       break;
-               nl = &((*nl)->next);
-       }
-       n->next = *nl;
-       rcu_assign_pointer(*nl, n);
-       return 0;
-}
-
-static int notifier_chain_unregister(struct notifier_block **nl,
-               struct notifier_block *n)
-{
-       while ((*nl) != NULL) {
-               if ((*nl) == n) {
-                       rcu_assign_pointer(*nl, n->next);
-                       return 0;
-               }
-               nl = &((*nl)->next);
-       }
-       return -ENOENT;
-}
-
-/**
- * notifier_call_chain - Informs the registered notifiers about an event.
- *     @nl:            Pointer to head of the blocking notifier chain
- *     @val:           Value passed unmodified to notifier function
- *     @v:             Pointer passed unmodified to notifier function
- *     @nr_to_call:    Number of notifier functions to be called. Don't care
- *                     value of this parameter is -1.
- *     @nr_calls:      Records the number of notifications sent. Don't care
- *                     value of this field is NULL.
- *     @returns:       notifier_call_chain returns the value returned by the
- *                     last notifier function called.
- */
-
-static int __kprobes notifier_call_chain(struct notifier_block **nl,
-                                       unsigned long val, void *v,
-                                       int nr_to_call, int *nr_calls)
-{
-       int ret = NOTIFY_DONE;
-       struct notifier_block *nb, *next_nb;
-
-       nb = rcu_dereference(*nl);
-
-       while (nb && nr_to_call) {
-               next_nb = rcu_dereference(nb->next);
-               ret = nb->notifier_call(nb, val, v);
-
-               if (nr_calls)
-                       (*nr_calls)++;
-
-               if ((ret & NOTIFY_STOP_MASK) == NOTIFY_STOP_MASK)
-                       break;
-               nb = next_nb;
-               nr_to_call--;
-       }
-       return ret;
-}
-
-/*
- *     Atomic notifier chain routines.  Registration and unregistration
- *     use a spinlock, and call_chain is synchronized by RCU (no locks).
- */
-
-/**
- *     atomic_notifier_chain_register - Add notifier to an atomic notifier chain
- *     @nh: Pointer to head of the atomic notifier chain
- *     @n: New entry in notifier chain
- *
- *     Adds a notifier to an atomic notifier chain.
- *
- *     Currently always returns zero.
- */
-
-int atomic_notifier_chain_register(struct atomic_notifier_head *nh,
-               struct notifier_block *n)
-{
-       unsigned long flags;
-       int ret;
-
-       spin_lock_irqsave(&nh->lock, flags);
-       ret = notifier_chain_register(&nh->head, n);
-       spin_unlock_irqrestore(&nh->lock, flags);
-       return ret;
-}
-
-EXPORT_SYMBOL_GPL(atomic_notifier_chain_register);
-
-/**
- *     atomic_notifier_chain_unregister - Remove notifier from an atomic notifier chain
- *     @nh: Pointer to head of the atomic notifier chain
- *     @n: Entry to remove from notifier chain
- *
- *     Removes a notifier from an atomic notifier chain.
- *
- *     Returns zero on success or %-ENOENT on failure.
- */
-int atomic_notifier_chain_unregister(struct atomic_notifier_head *nh,
-               struct notifier_block *n)
-{
-       unsigned long flags;
-       int ret;
-
-       spin_lock_irqsave(&nh->lock, flags);
-       ret = notifier_chain_unregister(&nh->head, n);
-       spin_unlock_irqrestore(&nh->lock, flags);
-       synchronize_rcu();
-       return ret;
-}
-
-EXPORT_SYMBOL_GPL(atomic_notifier_chain_unregister);
-
-/**
- *     __atomic_notifier_call_chain - Call functions in an atomic notifier chain
- *     @nh: Pointer to head of the atomic notifier chain
- *     @val: Value passed unmodified to notifier function
- *     @v: Pointer passed unmodified to notifier function
- *     @nr_to_call: See the comment for notifier_call_chain.
- *     @nr_calls: See the comment for notifier_call_chain.
- *
- *     Calls each function in a notifier chain in turn.  The functions
- *     run in an atomic context, so they must not block.
- *     This routine uses RCU to synchronize with changes to the chain.
- *
- *     If the return value of the notifier can be and'ed
- *     with %NOTIFY_STOP_MASK then atomic_notifier_call_chain()
- *     will return immediately, with the return value of
- *     the notifier function which halted execution.
- *     Otherwise the return value is the return value
- *     of the last notifier function called.
- */
-int __kprobes __atomic_notifier_call_chain(struct atomic_notifier_head *nh,
-                                       unsigned long val, void *v,
-                                       int nr_to_call, int *nr_calls)
-{
-       int ret;
-
-       rcu_read_lock();
-       ret = notifier_call_chain(&nh->head, val, v, nr_to_call, nr_calls);
-       rcu_read_unlock();
-       return ret;
-}
-
-EXPORT_SYMBOL_GPL(__atomic_notifier_call_chain);
-
-int __kprobes atomic_notifier_call_chain(struct atomic_notifier_head *nh,
-               unsigned long val, void *v)
-{
-       return __atomic_notifier_call_chain(nh, val, v, -1, NULL);
-}
-
-EXPORT_SYMBOL_GPL(atomic_notifier_call_chain);
-/*
- *     Blocking notifier chain routines.  All access to the chain is
- *     synchronized by an rwsem.
- */
-
-/**
- *     blocking_notifier_chain_register - Add notifier to a blocking notifier chain
- *     @nh: Pointer to head of the blocking notifier chain
- *     @n: New entry in notifier chain
- *
- *     Adds a notifier to a blocking notifier chain.
- *     Must be called in process context.
- *
- *     Currently always returns zero.
- */
-int blocking_notifier_chain_register(struct blocking_notifier_head *nh,
-               struct notifier_block *n)
-{
-       int ret;
-
-       /*
-        * This code gets used during boot-up, when task switching is
-        * not yet working and interrupts must remain disabled.  At
-        * such times we must not call down_write().
-        */
-       if (unlikely(system_state == SYSTEM_BOOTING))
-               return notifier_chain_register(&nh->head, n);
-
-       down_write(&nh->rwsem);
-       ret = notifier_chain_register(&nh->head, n);
-       up_write(&nh->rwsem);
-       return ret;
-}
-
-EXPORT_SYMBOL_GPL(blocking_notifier_chain_register);
-
-/**
- *     blocking_notifier_chain_unregister - Remove notifier from a blocking notifier chain
- *     @nh: Pointer to head of the blocking notifier chain
- *     @n: Entry to remove from notifier chain
- *
- *     Removes a notifier from a blocking notifier chain.
- *     Must be called from process context.
- *
- *     Returns zero on success or %-ENOENT on failure.
- */
-int blocking_notifier_chain_unregister(struct blocking_notifier_head *nh,
-               struct notifier_block *n)
-{
-       int ret;
-
-       /*
-        * This code gets used during boot-up, when task switching is
-        * not yet working and interrupts must remain disabled.  At
-        * such times we must not call down_write().
-        */
-       if (unlikely(system_state == SYSTEM_BOOTING))
-               return notifier_chain_unregister(&nh->head, n);
-
-       down_write(&nh->rwsem);
-       ret = notifier_chain_unregister(&nh->head, n);
-       up_write(&nh->rwsem);
-       return ret;
-}
-
-EXPORT_SYMBOL_GPL(blocking_notifier_chain_unregister);
-
-/**
- *     __blocking_notifier_call_chain - Call functions in a blocking notifier chain
- *     @nh: Pointer to head of the blocking notifier chain
- *     @val: Value passed unmodified to notifier function
- *     @v: Pointer passed unmodified to notifier function
- *     @nr_to_call: See comment for notifier_call_chain.
- *     @nr_calls: See comment for notifier_call_chain.
- *
- *     Calls each function in a notifier chain in turn.  The functions
- *     run in a process context, so they are allowed to block.
- *
- *     If the return value of the notifier can be and'ed
- *     with %NOTIFY_STOP_MASK then blocking_notifier_call_chain()
- *     will return immediately, with the return value of
- *     the notifier function which halted execution.
- *     Otherwise the return value is the return value
- *     of the last notifier function called.
- */
-int __blocking_notifier_call_chain(struct blocking_notifier_head *nh,
-                                  unsigned long val, void *v,
-                                  int nr_to_call, int *nr_calls)
-{
-       int ret = NOTIFY_DONE;
-
-       /*
-        * We check the head outside the lock, but if this access is
-        * racy then it does not matter what the result of the test
-        * is, we re-check the list after having taken the lock anyway:
-        */
-       if (rcu_dereference(nh->head)) {
-               down_read(&nh->rwsem);
-               ret = notifier_call_chain(&nh->head, val, v, nr_to_call,
-                                       nr_calls);
-               up_read(&nh->rwsem);
-       }
-       return ret;
-}
-EXPORT_SYMBOL_GPL(__blocking_notifier_call_chain);
-
-int blocking_notifier_call_chain(struct blocking_notifier_head *nh,
-               unsigned long val, void *v)
-{
-       return __blocking_notifier_call_chain(nh, val, v, -1, NULL);
-}
-EXPORT_SYMBOL_GPL(blocking_notifier_call_chain);
-
-/*
- *     Raw notifier chain routines.  There is no protection;
- *     the caller must provide it.  Use at your own risk!
- */
-
-/**
- *     raw_notifier_chain_register - Add notifier to a raw notifier chain
- *     @nh: Pointer to head of the raw notifier chain
- *     @n: New entry in notifier chain
- *
- *     Adds a notifier to a raw notifier chain.
- *     All locking must be provided by the caller.
- *
- *     Currently always returns zero.
+ * If set, this is used for preparing the system to power off.
  */
 
-int raw_notifier_chain_register(struct raw_notifier_head *nh,
-               struct notifier_block *n)
-{
-       return notifier_chain_register(&nh->head, n);
-}
-
-EXPORT_SYMBOL_GPL(raw_notifier_chain_register);
-
-/**
- *     raw_notifier_chain_unregister - Remove notifier from a raw notifier chain
- *     @nh: Pointer to head of the raw notifier chain
- *     @n: Entry to remove from notifier chain
- *
- *     Removes a notifier from a raw notifier chain.
- *     All locking must be provided by the caller.
- *
- *     Returns zero on success or %-ENOENT on failure.
- */
-int raw_notifier_chain_unregister(struct raw_notifier_head *nh,
-               struct notifier_block *n)
-{
-       return notifier_chain_unregister(&nh->head, n);
-}
-
-EXPORT_SYMBOL_GPL(raw_notifier_chain_unregister);
-
-/**
- *     __raw_notifier_call_chain - Call functions in a raw notifier chain
- *     @nh: Pointer to head of the raw notifier chain
- *     @val: Value passed unmodified to notifier function
- *     @v: Pointer passed unmodified to notifier function
- *     @nr_to_call: See comment for notifier_call_chain.
- *     @nr_calls: See comment for notifier_call_chain
- *
- *     Calls each function in a notifier chain in turn.  The functions
- *     run in an undefined context.
- *     All locking must be provided by the caller.
- *
- *     If the return value of the notifier can be and'ed
- *     with %NOTIFY_STOP_MASK then raw_notifier_call_chain()
- *     will return immediately, with the return value of
- *     the notifier function which halted execution.
- *     Otherwise the return value is the return value
- *     of the last notifier function called.
- */
-
-int __raw_notifier_call_chain(struct raw_notifier_head *nh,
-                             unsigned long val, void *v,
-                             int nr_to_call, int *nr_calls)
-{
-       return notifier_call_chain(&nh->head, val, v, nr_to_call, nr_calls);
-}
-
-EXPORT_SYMBOL_GPL(__raw_notifier_call_chain);
-
-int raw_notifier_call_chain(struct raw_notifier_head *nh,
-               unsigned long val, void *v)
-{
-       return __raw_notifier_call_chain(nh, val, v, -1, NULL);
-}
-
-EXPORT_SYMBOL_GPL(raw_notifier_call_chain);
+void (*pm_power_off_prepare)(void);
 
 /*
- *     SRCU notifier chain routines.    Registration and unregistration
- *     use a mutex, and call_chain is synchronized by SRCU (no locks).
+ * set the priority of a task
+ * - the caller must hold the RCU read lock
  */
-
-/**
- *     srcu_notifier_chain_register - Add notifier to an SRCU notifier chain
- *     @nh: Pointer to head of the SRCU notifier chain
- *     @n: New entry in notifier chain
- *
- *     Adds a notifier to an SRCU notifier chain.
- *     Must be called in process context.
- *
- *     Currently always returns zero.
- */
-
-int srcu_notifier_chain_register(struct srcu_notifier_head *nh,
-               struct notifier_block *n)
-{
-       int ret;
-
-       /*
-        * This code gets used during boot-up, when task switching is
-        * not yet working and interrupts must remain disabled.  At
-        * such times we must not call mutex_lock().
-        */
-       if (unlikely(system_state == SYSTEM_BOOTING))
-               return notifier_chain_register(&nh->head, n);
-
-       mutex_lock(&nh->mutex);
-       ret = notifier_chain_register(&nh->head, n);
-       mutex_unlock(&nh->mutex);
-       return ret;
-}
-
-EXPORT_SYMBOL_GPL(srcu_notifier_chain_register);
-
-/**
- *     srcu_notifier_chain_unregister - Remove notifier from an SRCU notifier chain
- *     @nh: Pointer to head of the SRCU notifier chain
- *     @n: Entry to remove from notifier chain
- *
- *     Removes a notifier from an SRCU notifier chain.
- *     Must be called from process context.
- *
- *     Returns zero on success or %-ENOENT on failure.
- */
-int srcu_notifier_chain_unregister(struct srcu_notifier_head *nh,
-               struct notifier_block *n)
-{
-       int ret;
-
-       /*
-        * This code gets used during boot-up, when task switching is
-        * not yet working and interrupts must remain disabled.  At
-        * such times we must not call mutex_lock().
-        */
-       if (unlikely(system_state == SYSTEM_BOOTING))
-               return notifier_chain_unregister(&nh->head, n);
-
-       mutex_lock(&nh->mutex);
-       ret = notifier_chain_unregister(&nh->head, n);
-       mutex_unlock(&nh->mutex);
-       synchronize_srcu(&nh->srcu);
-       return ret;
-}
-
-EXPORT_SYMBOL_GPL(srcu_notifier_chain_unregister);
-
-/**
- *     __srcu_notifier_call_chain - Call functions in an SRCU notifier chain
- *     @nh: Pointer to head of the SRCU notifier chain
- *     @val: Value passed unmodified to notifier function
- *     @v: Pointer passed unmodified to notifier function
- *     @nr_to_call: See comment for notifier_call_chain.
- *     @nr_calls: See comment for notifier_call_chain
- *
- *     Calls each function in a notifier chain in turn.  The functions
- *     run in a process context, so they are allowed to block.
- *
- *     If the return value of the notifier can be and'ed
- *     with %NOTIFY_STOP_MASK then srcu_notifier_call_chain()
- *     will return immediately, with the return value of
- *     the notifier function which halted execution.
- *     Otherwise the return value is the return value
- *     of the last notifier function called.
- */
-
-int __srcu_notifier_call_chain(struct srcu_notifier_head *nh,
-                              unsigned long val, void *v,
-                              int nr_to_call, int *nr_calls)
-{
-       int ret;
-       int idx;
-
-       idx = srcu_read_lock(&nh->srcu);
-       ret = notifier_call_chain(&nh->head, val, v, nr_to_call, nr_calls);
-       srcu_read_unlock(&nh->srcu, idx);
-       return ret;
-}
-EXPORT_SYMBOL_GPL(__srcu_notifier_call_chain);
-
-int srcu_notifier_call_chain(struct srcu_notifier_head *nh,
-               unsigned long val, void *v)
-{
-       return __srcu_notifier_call_chain(nh, val, v, -1, NULL);
-}
-EXPORT_SYMBOL_GPL(srcu_notifier_call_chain);
-
-/**
- *     srcu_init_notifier_head - Initialize an SRCU notifier head
- *     @nh: Pointer to head of the srcu notifier chain
- *
- *     Unlike other sorts of notifier heads, SRCU notifier heads require
- *     dynamic initialization.  Be sure to call this routine before
- *     calling any of the other SRCU notifier routines for this head.
- *
- *     If an SRCU notifier head is deallocated, it must first be cleaned
- *     up by calling srcu_cleanup_notifier_head().  Otherwise the head's
- *     per-cpu data (used by the SRCU mechanism) will leak.
- */
-
-void srcu_init_notifier_head(struct srcu_notifier_head *nh)
-{
-       mutex_init(&nh->mutex);
-       if (init_srcu_struct(&nh->srcu) < 0)
-               BUG();
-       nh->head = NULL;
-}
-
-EXPORT_SYMBOL_GPL(srcu_init_notifier_head);
-
-/**
- *     register_reboot_notifier - Register function to be called at reboot time
- *     @nb: Info about notifier function to be called
- *
- *     Registers a function with the list of functions
- *     to be called at reboot time.
- *
- *     Currently always returns zero, as blocking_notifier_chain_register()
- *     always returns zero.
- */
-int register_reboot_notifier(struct notifier_block * nb)
-{
-       return blocking_notifier_chain_register(&reboot_notifier_list, nb);
-}
-
-EXPORT_SYMBOL(register_reboot_notifier);
-
-/**
- *     unregister_reboot_notifier - Unregister previously registered reboot notifier
- *     @nb: Hook to be unregistered
- *
- *     Unregisters a previously registered reboot
- *     notifier function.
- *
- *     Returns zero on success, or %-ENOENT on failure.
- */
-int unregister_reboot_notifier(struct notifier_block * nb)
-{
-       return blocking_notifier_chain_unregister(&reboot_notifier_list, nb);
-}
-
-EXPORT_SYMBOL(unregister_reboot_notifier);
-
 static int set_one_prio(struct task_struct *p, int niceval, int error)
 {
+       const struct cred *cred = current_cred(), *pcred = __task_cred(p);
        int no_nice;
 
-       if (p->uid != current->euid &&
-               p->euid != current->euid && !capable(CAP_SYS_NICE)) {
+       if (pcred->uid  != cred->euid &&
+           pcred->euid != cred->euid && !capable(CAP_SYS_NICE)) {
                error = -EPERM;
                goto out;
        }
@@ -659,6 +146,7 @@ asmlinkage long sys_setpriority(int which, int who, int niceval)
 {
        struct task_struct *g, *p;
        struct user_struct *user;
+       const struct cred *cred = current_cred();
        int error = -EINVAL;
        struct pid *pgrp;
 
@@ -676,7 +164,7 @@ asmlinkage long sys_setpriority(int which, int who, int niceval)
        switch (which) {
                case PRIO_PROCESS:
                        if (who)
-                               p = find_task_by_pid(who);
+                               p = find_task_by_vpid(who);
                        else
                                p = current;
                        if (p)
@@ -684,26 +172,26 @@ asmlinkage long sys_setpriority(int which, int who, int niceval)
                        break;
                case PRIO_PGRP:
                        if (who)
-                               pgrp = find_pid(who);
+                               pgrp = find_vpid(who);
                        else
                                pgrp = task_pgrp(current);
-                       do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
+                       do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
                                error = set_one_prio(p, niceval, error);
-                       } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
+                       } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
                        break;
                case PRIO_USER:
-                       user = current->user;
+                       user = (struct user_struct *) cred->user;
                        if (!who)
-                               who = current->uid;
-                       else
-                               if ((who != current->uid) && !(user = find_user(who)))
-                                       goto out_unlock;        /* No processes for this user */
+                               who = cred->uid;
+                       else if ((who != cred->uid) &&
+                                !(user = find_user(who)))
+                               goto out_unlock;        /* No processes for this user */
 
                        do_each_thread(g, p)
-                               if (p->uid == who)
+                               if (__task_cred(p)->uid == who)
                                        error = set_one_prio(p, niceval, error);
                        while_each_thread(g, p);
-                       if (who != current->uid)
+                       if (who != cred->uid)
                                free_uid(user);         /* For find_user() */
                        break;
        }
@@ -723,6 +211,7 @@ asmlinkage long sys_getpriority(int which, int who)
 {
        struct task_struct *g, *p;
        struct user_struct *user;
+       const struct cred *cred = current_cred();
        long niceval, retval = -ESRCH;
        struct pid *pgrp;
 
@@ -733,7 +222,7 @@ asmlinkage long sys_getpriority(int which, int who)
        switch (which) {
                case PRIO_PROCESS:
                        if (who)
-                               p = find_task_by_pid(who);
+                               p = find_task_by_vpid(who);
                        else
                                p = current;
                        if (p) {
@@ -744,31 +233,31 @@ asmlinkage long sys_getpriority(int which, int who)
                        break;
                case PRIO_PGRP:
                        if (who)
-                               pgrp = find_pid(who);
+                               pgrp = find_vpid(who);
                        else
                                pgrp = task_pgrp(current);
-                       do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
+                       do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
                                niceval = 20 - task_nice(p);
                                if (niceval > retval)
                                        retval = niceval;
-                       } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
+                       } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
                        break;
                case PRIO_USER:
-                       user = current->user;
+                       user = (struct user_struct *) cred->user;
                        if (!who)
-                               who = current->uid;
-                       else
-                               if ((who != current->uid) && !(user = find_user(who)))
-                                       goto out_unlock;        /* No processes for this user */
+                               who = cred->uid;
+                       else if ((who != cred->uid) &&
+                                !(user = find_user(who)))
+                               goto out_unlock;        /* No processes for this user */
 
                        do_each_thread(g, p)
-                               if (p->uid == who) {
+                               if (__task_cred(p)->uid == who) {
                                        niceval = 20 - task_nice(p);
                                        if (niceval > retval)
                                                retval = niceval;
                                }
                        while_each_thread(g, p);
-                       if (who != current->uid)
+                       if (who != cred->uid)
                                free_uid(user);         /* for find_user() */
                        break;
        }
@@ -792,11 +281,12 @@ void emergency_restart(void)
 }
 EXPORT_SYMBOL_GPL(emergency_restart);
 
-static void kernel_restart_prepare(char *cmd)
+void kernel_restart_prepare(char *cmd)
 {
        blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
        system_state = SYSTEM_RESTART;
        device_shutdown();
+       sysdev_shutdown();
 }
 
 /**
@@ -818,27 +308,7 @@ void kernel_restart(char *cmd)
 }
 EXPORT_SYMBOL_GPL(kernel_restart);
 
-/**
- *     kernel_kexec - reboot the system
- *
- *     Move into place and start executing a preloaded standalone
- *     executable.  If nothing was preloaded return an error.
- */
-static void kernel_kexec(void)
-{
-#ifdef CONFIG_KEXEC
-       struct kimage *image;
-       image = xchg(&kexec_image, NULL);
-       if (!image)
-               return;
-       kernel_restart_prepare(NULL);
-       printk(KERN_EMERG "Starting new kernel\n");
-       machine_shutdown();
-       machine_kexec(image);
-#endif
-}
-
-void kernel_shutdown_prepare(enum system_states state)
+static void kernel_shutdown_prepare(enum system_states state)
 {
        blocking_notifier_call_chain(&reboot_notifier_list,
                (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
@@ -853,6 +323,7 @@ void kernel_shutdown_prepare(enum system_states state)
 void kernel_halt(void)
 {
        kernel_shutdown_prepare(SYSTEM_HALT);
+       sysdev_shutdown();
        printk(KERN_EMERG "System halted.\n");
        machine_halt();
 }
@@ -867,6 +338,10 @@ EXPORT_SYMBOL_GPL(kernel_halt);
 void kernel_power_off(void)
 {
        kernel_shutdown_prepare(SYSTEM_POWER_OFF);
+       if (pm_power_off_prepare)
+               pm_power_off_prepare();
+       disable_nonboot_cpus();
+       sysdev_shutdown();
        printk(KERN_EMERG "Power down.\n");
        machine_power_off();
 }
@@ -937,12 +412,17 @@ asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user
                kernel_restart(buffer);
                break;
 
+#ifdef CONFIG_KEXEC
        case LINUX_REBOOT_CMD_KEXEC:
-               kernel_kexec();
-               unlock_kernel();
-               return -EINVAL;
+               {
+                       int ret;
+                       ret = kernel_kexec();
+                       unlock_kernel();
+                       return ret;
+               }
+#endif
 
-#ifdef CONFIG_SOFTWARE_SUSPEND
+#ifdef CONFIG_HIBERNATION
        case LINUX_REBOOT_CMD_SW_SUSPEND:
                {
                        int ret = hibernate();
@@ -999,46 +479,48 @@ void ctrl_alt_del(void)
  */
 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
 {
-       int old_rgid = current->gid;
-       int old_egid = current->egid;
-       int new_rgid = old_rgid;
-       int new_egid = old_egid;
+       const struct cred *old;
+       struct cred *new;
        int retval;
 
+       new = prepare_creds();
+       if (!new)
+               return -ENOMEM;
+       old = current_cred();
+
        retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
        if (retval)
-               return retval;
+               goto error;
 
+       retval = -EPERM;
        if (rgid != (gid_t) -1) {
-               if ((old_rgid == rgid) ||
-                   (current->egid==rgid) ||
+               if (old->gid == rgid ||
+                   old->egid == rgid ||
                    capable(CAP_SETGID))
-                       new_rgid = rgid;
+                       new->gid = rgid;
                else
-                       return -EPERM;
+                       goto error;
        }
        if (egid != (gid_t) -1) {
-               if ((old_rgid == egid) ||
-                   (current->egid == egid) ||
-                   (current->sgid == egid) ||
+               if (old->gid == egid ||
+                   old->egid == egid ||
+                   old->sgid == egid ||
                    capable(CAP_SETGID))
-                       new_egid = egid;
+                       new->egid = egid;
                else
-                       return -EPERM;
-       }
-       if (new_egid != old_egid) {
-               current->mm->dumpable = suid_dumpable;
-               smp_wmb();
+                       goto error;
        }
+
        if (rgid != (gid_t) -1 ||
-           (egid != (gid_t) -1 && egid != old_rgid))
-               current->sgid = new_egid;
-       current->fsgid = new_egid;
-       current->egid = new_egid;
-       current->gid = new_rgid;
-       key_fsgid_changed(current);
-       proc_id_connector(current, PROC_EVENT_GID);
-       return 0;
+           (egid != (gid_t) -1 && egid != old->gid))
+               new->sgid = new->egid;
+       new->fsgid = new->egid;
+
+       return commit_creds(new);
+
+error:
+       abort_creds(new);
+       return retval;
 }
 
 /*
@@ -1048,56 +530,54 @@ asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
  */
 asmlinkage long sys_setgid(gid_t gid)
 {
-       int old_egid = current->egid;
+       const struct cred *old;
+       struct cred *new;
        int retval;
 
+       new = prepare_creds();
+       if (!new)
+               return -ENOMEM;
+       old = current_cred();
+
        retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
        if (retval)
-               return retval;
+               goto error;
 
-       if (capable(CAP_SETGID)) {
-               if (old_egid != gid) {
-                       current->mm->dumpable = suid_dumpable;
-                       smp_wmb();
-               }
-               current->gid = current->egid = current->sgid = current->fsgid = gid;
-       } else if ((gid == current->gid) || (gid == current->sgid)) {
-               if (old_egid != gid) {
-                       current->mm->dumpable = suid_dumpable;
-                       smp_wmb();
-               }
-               current->egid = current->fsgid = gid;
-       }
+       retval = -EPERM;
+       if (capable(CAP_SETGID))
+               new->gid = new->egid = new->sgid = new->fsgid = gid;
+       else if (gid == old->gid || gid == old->sgid)
+               new->egid = new->fsgid = gid;
        else
-               return -EPERM;
+               goto error;
 
-       key_fsgid_changed(current);
-       proc_id_connector(current, PROC_EVENT_GID);
-       return 0;
+       return commit_creds(new);
+
+error:
+       abort_creds(new);
+       return retval;
 }
   
-static int set_user(uid_t new_ruid, int dumpclear)
+/*
+ * change the user struct in a credentials set to match the new UID
+ */
+static int set_user(struct cred *new)
 {
        struct user_struct *new_user;
 
-       new_user = alloc_uid(current->nsproxy->user_ns, new_ruid);
+       new_user = alloc_uid(current_user_ns(), new->uid);
        if (!new_user)
                return -EAGAIN;
 
        if (atomic_read(&new_user->processes) >=
                                current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
-                       new_user != current->nsproxy->user_ns->root_user) {
+                       new_user != INIT_USER) {
                free_uid(new_user);
                return -EAGAIN;
        }
 
-       switch_uid(new_user);
-
-       if (dumpclear) {
-               current->mm->dumpable = suid_dumpable;
-               smp_wmb();
-       }
-       current->uid = new_ruid;
+       free_uid(new->user);
+       new->user = new_user;
        return 0;
 }
 
@@ -1118,54 +598,56 @@ static int set_user(uid_t new_ruid, int dumpclear)
  */
 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
 {
-       int old_ruid, old_euid, old_suid, new_ruid, new_euid;
+       const struct cred *old;
+       struct cred *new;
        int retval;
 
+       new = prepare_creds();
+       if (!new)
+               return -ENOMEM;
+       old = current_cred();
+
        retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
        if (retval)
-               return retval;
-
-       new_ruid = old_ruid = current->uid;
-       new_euid = old_euid = current->euid;
-       old_suid = current->suid;
+               goto error;
 
+       retval = -EPERM;
        if (ruid != (uid_t) -1) {
-               new_ruid = ruid;
-               if ((old_ruid != ruid) &&
-                   (current->euid != ruid) &&
+               new->uid = ruid;
+               if (old->uid != ruid &&
+                   old->euid != ruid &&
                    !capable(CAP_SETUID))
-                       return -EPERM;
+                       goto error;
        }
 
        if (euid != (uid_t) -1) {
-               new_euid = euid;
-               if ((old_ruid != euid) &&
-                   (current->euid != euid) &&
-                   (current->suid != euid) &&
+               new->euid = euid;
+               if (old->uid != euid &&
+                   old->euid != euid &&
+                   old->suid != euid &&
                    !capable(CAP_SETUID))
-                       return -EPERM;
+                       goto error;
        }
 
-       if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
-               return -EAGAIN;
+       retval = -EAGAIN;
+       if (new->uid != old->uid && set_user(new) < 0)
+               goto error;
 
-       if (new_euid != old_euid) {
-               current->mm->dumpable = suid_dumpable;
-               smp_wmb();
-       }
-       current->fsuid = current->euid = new_euid;
        if (ruid != (uid_t) -1 ||
-           (euid != (uid_t) -1 && euid != old_ruid))
-               current->suid = current->euid;
-       current->fsuid = current->euid;
-
-       key_fsuid_changed(current);
-       proc_id_connector(current, PROC_EVENT_UID);
+           (euid != (uid_t) -1 && euid != old->uid))
+               new->suid = new->euid;
+       new->fsuid = new->euid;
 
-       return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
-}
+       retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
+       if (retval < 0)
+               goto error;
 
+       return commit_creds(new);
 
+error:
+       abort_creds(new);
+       return retval;
+}
                
 /*
  * setuid() is implemented like SysV with SAVED_IDS 
@@ -1180,36 +662,41 @@ asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
  */
 asmlinkage long sys_setuid(uid_t uid)
 {
-       int old_euid = current->euid;
-       int old_ruid, old_suid, new_suid;
+       const struct cred *old;
+       struct cred *new;
        int retval;
 
+       new = prepare_creds();
+       if (!new)
+               return -ENOMEM;
+       old = current_cred();
+
        retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
        if (retval)
-               return retval;
+               goto error;
 
-       old_ruid = current->uid;
-       old_suid = current->suid;
-       new_suid = old_suid;
-       
+       retval = -EPERM;
        if (capable(CAP_SETUID)) {
-               if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
-                       return -EAGAIN;
-               new_suid = uid;
-       } else if ((uid != current->uid) && (uid != new_suid))
-               return -EPERM;
-
-       if (old_euid != uid) {
-               current->mm->dumpable = suid_dumpable;
-               smp_wmb();
+               new->suid = new->uid = uid;
+               if (uid != old->uid && set_user(new) < 0) {
+                       retval = -EAGAIN;
+                       goto error;
+               }
+       } else if (uid != old->uid && uid != new->suid) {
+               goto error;
        }
-       current->fsuid = current->euid = uid;
-       current->suid = new_suid;
 
-       key_fsuid_changed(current);
-       proc_id_connector(current, PROC_EVENT_UID);
+       new->fsuid = new->euid = uid;
+
+       retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
+       if (retval < 0)
+               goto error;
 
-       return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
+       return commit_creds(new);
+
+error:
+       abort_creds(new);
+       return retval;
 }
 
 
@@ -1219,54 +706,63 @@ asmlinkage long sys_setuid(uid_t uid)
  */
 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
 {
-       int old_ruid = current->uid;
-       int old_euid = current->euid;
-       int old_suid = current->suid;
+       const struct cred *old;
+       struct cred *new;
        int retval;
 
+       new = prepare_creds();
+       if (!new)
+               return -ENOMEM;
+
        retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
        if (retval)
-               return retval;
+               goto error;
+       old = current_cred();
 
+       retval = -EPERM;
        if (!capable(CAP_SETUID)) {
-               if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
-                   (ruid != current->euid) && (ruid != current->suid))
-                       return -EPERM;
-               if ((euid != (uid_t) -1) && (euid != current->uid) &&
-                   (euid != current->euid) && (euid != current->suid))
-                       return -EPERM;
-               if ((suid != (uid_t) -1) && (suid != current->uid) &&
-                   (suid != current->euid) && (suid != current->suid))
-                       return -EPERM;
+               if (ruid != (uid_t) -1 && ruid != old->uid &&
+                   ruid != old->euid  && ruid != old->suid)
+                       goto error;
+               if (euid != (uid_t) -1 && euid != old->uid &&
+                   euid != old->euid  && euid != old->suid)
+                       goto error;
+               if (suid != (uid_t) -1 && suid != old->uid &&
+                   suid != old->euid  && suid != old->suid)
+                       goto error;
        }
+
+       retval = -EAGAIN;
        if (ruid != (uid_t) -1) {
-               if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
-                       return -EAGAIN;
+               new->uid = ruid;
+               if (ruid != old->uid && set_user(new) < 0)
+                       goto error;
        }
-       if (euid != (uid_t) -1) {
-               if (euid != current->euid) {
-                       current->mm->dumpable = suid_dumpable;
-                       smp_wmb();
-               }
-               current->euid = euid;
-       }
-       current->fsuid = current->euid;
+       if (euid != (uid_t) -1)
+               new->euid = euid;
        if (suid != (uid_t) -1)
-               current->suid = suid;
+               new->suid = suid;
+       new->fsuid = new->euid;
+
+       retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
+       if (retval < 0)
+               goto error;
 
-       key_fsuid_changed(current);
-       proc_id_connector(current, PROC_EVENT_UID);
+       return commit_creds(new);
 
-       return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
+error:
+       abort_creds(new);
+       return retval;
 }
 
 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
 {
+       const struct cred *cred = current_cred();
        int retval;
 
-       if (!(retval = put_user(current->uid, ruid)) &&
-           !(retval = put_user(current->euid, euid)))
-               retval = put_user(current->suid, suid);
+       if (!(retval   = put_user(cred->uid,  ruid)) &&
+           !(retval   = put_user(cred->euid, euid)))
+               retval = put_user(cred->suid, suid);
 
        return retval;
 }
@@ -1276,48 +772,55 @@ asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __us
  */
 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
 {
+       const struct cred *old;
+       struct cred *new;
        int retval;
 
+       new = prepare_creds();
+       if (!new)
+               return -ENOMEM;
+       old = current_cred();
+
        retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
        if (retval)
-               return retval;
+               goto error;
 
+       retval = -EPERM;
        if (!capable(CAP_SETGID)) {
-               if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
-                   (rgid != current->egid) && (rgid != current->sgid))
-                       return -EPERM;
-               if ((egid != (gid_t) -1) && (egid != current->gid) &&
-                   (egid != current->egid) && (egid != current->sgid))
-                       return -EPERM;
-               if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
-                   (sgid != current->egid) && (sgid != current->sgid))
-                       return -EPERM;
-       }
-       if (egid != (gid_t) -1) {
-               if (egid != current->egid) {
-                       current->mm->dumpable = suid_dumpable;
-                       smp_wmb();
-               }
-               current->egid = egid;
+               if (rgid != (gid_t) -1 && rgid != old->gid &&
+                   rgid != old->egid  && rgid != old->sgid)
+                       goto error;
+               if (egid != (gid_t) -1 && egid != old->gid &&
+                   egid != old->egid  && egid != old->sgid)
+                       goto error;
+               if (sgid != (gid_t) -1 && sgid != old->gid &&
+                   sgid != old->egid  && sgid != old->sgid)
+                       goto error;
        }
-       current->fsgid = current->egid;
+
        if (rgid != (gid_t) -1)
-               current->gid = rgid;
+               new->gid = rgid;
+       if (egid != (gid_t) -1)
+               new->egid = egid;
        if (sgid != (gid_t) -1)
-               current->sgid = sgid;
+               new->sgid = sgid;
+       new->fsgid = new->egid;
 
-       key_fsgid_changed(current);
-       proc_id_connector(current, PROC_EVENT_GID);
-       return 0;
+       return commit_creds(new);
+
+error:
+       abort_creds(new);
+       return retval;
 }
 
 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
 {
+       const struct cred *cred = current_cred();
        int retval;
 
-       if (!(retval = put_user(current->gid, rgid)) &&
-           !(retval = put_user(current->egid, egid)))
-               retval = put_user(current->sgid, sgid);
+       if (!(retval   = put_user(cred->gid,  rgid)) &&
+           !(retval   = put_user(cred->egid, egid)))
+               retval = put_user(cred->sgid, sgid);
 
        return retval;
 }
@@ -1331,27 +834,35 @@ asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __us
  */
 asmlinkage long sys_setfsuid(uid_t uid)
 {
-       int old_fsuid;
+       const struct cred *old;
+       struct cred *new;
+       uid_t old_fsuid;
 
-       old_fsuid = current->fsuid;
-       if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
-               return old_fsuid;
+       new = prepare_creds();
+       if (!new)
+               return current_fsuid();
+       old = current_cred();
+       old_fsuid = old->fsuid;
 
-       if (uid == current->uid || uid == current->euid ||
-           uid == current->suid || uid == current->fsuid || 
+       if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS) < 0)
+               goto error;
+
+       if (uid == old->uid  || uid == old->euid  ||
+           uid == old->suid || uid == old->fsuid ||
            capable(CAP_SETUID)) {
                if (uid != old_fsuid) {
-                       current->mm->dumpable = suid_dumpable;
-                       smp_wmb();
+                       new->fsuid = uid;
+                       if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
+                               goto change_okay;
                }
-               current->fsuid = uid;
        }
 
-       key_fsuid_changed(current);
-       proc_id_connector(current, PROC_EVENT_UID);
-
-       security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
+error:
+       abort_creds(new);
+       return old_fsuid;
 
+change_okay:
+       commit_creds(new);
        return old_fsuid;
 }
 
@@ -1360,58 +871,59 @@ asmlinkage long sys_setfsuid(uid_t uid)
  */
 asmlinkage long sys_setfsgid(gid_t gid)
 {
-       int old_fsgid;
+       const struct cred *old;
+       struct cred *new;
+       gid_t old_fsgid;
+
+       new = prepare_creds();
+       if (!new)
+               return current_fsgid();
+       old = current_cred();
+       old_fsgid = old->fsgid;
 
-       old_fsgid = current->fsgid;
        if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
-               return old_fsgid;
+               goto error;
 
-       if (gid == current->gid || gid == current->egid ||
-           gid == current->sgid || gid == current->fsgid || 
+       if (gid == old->gid  || gid == old->egid  ||
+           gid == old->sgid || gid == old->fsgid ||
            capable(CAP_SETGID)) {
                if (gid != old_fsgid) {
-                       current->mm->dumpable = suid_dumpable;
-                       smp_wmb();
+                       new->fsgid = gid;
+                       goto change_okay;
                }
-               current->fsgid = gid;
-               key_fsgid_changed(current);
-               proc_id_connector(current, PROC_EVENT_GID);
        }
+
+error:
+       abort_creds(new);
+       return old_fsgid;
+
+change_okay:
+       commit_creds(new);
        return old_fsgid;
 }
 
+void do_sys_times(struct tms *tms)
+{
+       struct task_cputime cputime;
+       cputime_t cutime, cstime;
+
+       thread_group_cputime(current, &cputime);
+       spin_lock_irq(&current->sighand->siglock);
+       cutime = current->signal->cutime;
+       cstime = current->signal->cstime;
+       spin_unlock_irq(&current->sighand->siglock);
+       tms->tms_utime = cputime_to_clock_t(cputime.utime);
+       tms->tms_stime = cputime_to_clock_t(cputime.stime);
+       tms->tms_cutime = cputime_to_clock_t(cutime);
+       tms->tms_cstime = cputime_to_clock_t(cstime);
+}
+
 asmlinkage long sys_times(struct tms __user * tbuf)
 {
-       /*
-        *      In the SMP world we might just be unlucky and have one of
-        *      the times increment as we use it. Since the value is an
-        *      atomically safe type this is just fine. Conceptually its
-        *      as if the syscall took an instant longer to occur.
-        */
        if (tbuf) {
                struct tms tmp;
-               struct task_struct *tsk = current;
-               struct task_struct *t;
-               cputime_t utime, stime, cutime, cstime;
-
-               spin_lock_irq(&tsk->sighand->siglock);
-               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);
-
-               cutime = tsk->signal->cutime;
-               cstime = tsk->signal->cstime;
-               spin_unlock_irq(&tsk->sighand->siglock);
-
-               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);
+
+               do_sys_times(&tmp);
                if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
                        return -EFAULT;
        }
@@ -1430,15 +942,15 @@ asmlinkage long sys_times(struct tms __user * tbuf)
  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
  * LBT 04.03.94
  */
-
 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
 {
        struct task_struct *p;
        struct task_struct *group_leader = current->group_leader;
-       int err = -EINVAL;
+       struct pid *pgrp;
+       int err;
 
        if (!pid)
-               pid = group_leader->pid;
+               pid = task_pid_vnr(group_leader);
        if (!pgid)
                pgid = pid;
        if (pgid < 0)
@@ -1450,7 +962,7 @@ asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
        write_lock_irq(&tasklist_lock);
 
        err = -ESRCH;
-       p = find_task_by_pid(pid);
+       p = find_task_by_vpid(pid);
        if (!p)
                goto out;
 
@@ -1458,7 +970,7 @@ asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
        if (!thread_group_leader(p))
                goto out;
 
-       if (p->real_parent == group_leader) {
+       if (same_thread_group(p->real_parent, group_leader)) {
                err = -EPERM;
                if (task_session(p) != task_session(group_leader))
                        goto out;
@@ -1475,10 +987,12 @@ asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
        if (p->signal->leader)
                goto out;
 
+       pgrp = task_pid(p);
        if (pgid != pid) {
-               struct task_struct *g =
-                       find_task_by_pid_type(PIDTYPE_PGID, pgid);
+               struct task_struct *g;
 
+               pgrp = find_vpid(pgid);
+               g = pid_task(pgrp, PIDTYPE_PGID);
                if (!g || task_session(g) != task_session(group_leader))
                        goto out;
        }
@@ -1487,10 +1001,9 @@ asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
        if (err)
                goto out;
 
-       if (process_group(p) != pgid) {
-               detach_pid(p, PIDTYPE_PGID);
-               p->signal->pgrp = pgid;
-               attach_pid(p, PIDTYPE_PGID, find_pid(pgid));
+       if (task_pgrp(p) != pgrp) {
+               change_pid(p, PIDTYPE_PGID, pgrp);
+               set_task_pgrp(p, pid_nr(pgrp));
        }
 
        err = 0;
@@ -1502,89 +1015,93 @@ out:
 
 asmlinkage long sys_getpgid(pid_t pid)
 {
+       struct task_struct *p;
+       struct pid *grp;
+       int retval;
+
+       rcu_read_lock();
        if (!pid)
-               return process_group(current);
+               grp = task_pgrp(current);
        else {
-               int retval;
-               struct task_struct *p;
-
-               read_lock(&tasklist_lock);
-               p = find_task_by_pid(pid);
-
                retval = -ESRCH;
-               if (p) {
-                       retval = security_task_getpgid(p);
-                       if (!retval)
-                               retval = process_group(p);
-               }
-               read_unlock(&tasklist_lock);
-               return retval;
+               p = find_task_by_vpid(pid);
+               if (!p)
+                       goto out;
+               grp = task_pgrp(p);
+               if (!grp)
+                       goto out;
+
+               retval = security_task_getpgid(p);
+               if (retval)
+                       goto out;
        }
+       retval = pid_vnr(grp);
+out:
+       rcu_read_unlock();
+       return retval;
 }
 
 #ifdef __ARCH_WANT_SYS_GETPGRP
 
 asmlinkage long sys_getpgrp(void)
 {
-       /* SMP - assuming writes are word atomic this is fine */
-       return process_group(current);
+       return sys_getpgid(0);
 }
 
 #endif
 
 asmlinkage long sys_getsid(pid_t pid)
 {
+       struct task_struct *p;
+       struct pid *sid;
+       int retval;
+
+       rcu_read_lock();
        if (!pid)
-               return process_session(current);
+               sid = task_session(current);
        else {
-               int retval;
-               struct task_struct *p;
-
-               read_lock(&tasklist_lock);
-               p = find_task_by_pid(pid);
-
                retval = -ESRCH;
-               if (p) {
-                       retval = security_task_getsid(p);
-                       if (!retval)
-                               retval = process_session(p);
-               }
-               read_unlock(&tasklist_lock);
-               return retval;
+               p = find_task_by_vpid(pid);
+               if (!p)
+                       goto out;
+               sid = task_session(p);
+               if (!sid)
+                       goto out;
+
+               retval = security_task_getsid(p);
+               if (retval)
+                       goto out;
        }
+       retval = pid_vnr(sid);
+out:
+       rcu_read_unlock();
+       return retval;
 }
 
 asmlinkage long sys_setsid(void)
 {
        struct task_struct *group_leader = current->group_leader;
-       pid_t session;
+       struct pid *sid = task_pid(group_leader);
+       pid_t session = pid_vnr(sid);
        int err = -EPERM;
 
        write_lock_irq(&tasklist_lock);
-
        /* Fail if I am already a session leader */
        if (group_leader->signal->leader)
                goto out;
 
-       session = group_leader->pid;
        /* Fail if a process group id already exists that equals the
         * proposed session id.
-        *
-        * Don't check if session id == 1 because kernel threads use this
-        * session id and so the check will always fail and make it so
-        * init cannot successfully call setsid.
         */
-       if (session > 1 && find_task_by_pid_type(PIDTYPE_PGID, session))
+       if (pid_task(sid, PIDTYPE_PGID))
                goto out;
 
        group_leader->signal->leader = 1;
-       __set_special_pids(session, session);
+       __set_special_pids(sid);
 
-       spin_lock(&group_leader->sighand->siglock);
-       group_leader->signal->tty = NULL;
-       spin_unlock(&group_leader->sighand->siglock);
+       proc_clear_tty(group_leader);
 
-       err = process_group(group_leader);
+       err = session;
 out:
        write_unlock_irq(&tasklist_lock);
        return err;
@@ -1650,19 +1167,19 @@ EXPORT_SYMBOL(groups_free);
 
 /* export the group_info to a user-space array */
 static int groups_to_user(gid_t __user *grouplist,
-    struct group_info *group_info)
+                         const struct group_info *group_info)
 {
        int i;
-       int count = group_info->ngroups;
+       unsigned int count = group_info->ngroups;
 
        for (i = 0; i < group_info->nblocks; i++) {
-               int cp_count = min(NGROUPS_PER_BLOCK, count);
-               int off = i * NGROUPS_PER_BLOCK;
-               int len = cp_count * sizeof(*grouplist);
+               unsigned int cp_count = min(NGROUPS_PER_BLOCK, count);
+               unsigned int len = cp_count * sizeof(*grouplist);
 
-               if (copy_to_user(grouplist+off, group_info->blocks[i], len))
+               if (copy_to_user(grouplist, group_info->blocks[i], len))
                        return -EFAULT;
 
+               grouplist += NGROUPS_PER_BLOCK;
                count -= cp_count;
        }
        return 0;
@@ -1673,16 +1190,16 @@ static int groups_from_user(struct group_info *group_info,
     gid_t __user *grouplist)
 {
        int i;
-       int count = group_info->ngroups;
+       unsigned int count = group_info->ngroups;
 
        for (i = 0; i < group_info->nblocks; i++) {
-               int cp_count = min(NGROUPS_PER_BLOCK, count);
-               int off = i * NGROUPS_PER_BLOCK;
-               int len = cp_count * sizeof(*grouplist);
+               unsigned int cp_count = min(NGROUPS_PER_BLOCK, count);
+               unsigned int len = cp_count * sizeof(*grouplist);
 
-               if (copy_from_user(group_info->blocks[i], grouplist+off, len))
+               if (copy_from_user(group_info->blocks[i], grouplist, len))
                        return -EFAULT;
 
+               grouplist += NGROUPS_PER_BLOCK;
                count -= cp_count;
        }
        return 0;
@@ -1718,7 +1235,7 @@ static void groups_sort(struct group_info *group_info)
 }
 
 /* a simple bsearch */
-int groups_search(struct group_info *group_info, gid_t grp)
+int groups_search(const struct group_info *group_info, gid_t grp)
 {
        unsigned int left, right;
 
@@ -1740,51 +1257,74 @@ int groups_search(struct group_info *group_info, gid_t grp)
        return 0;
 }
 
-/* validate and set current->group_info */
-int set_current_groups(struct group_info *group_info)
+/**
+ * set_groups - Change a group subscription in a set of credentials
+ * @new: The newly prepared set of credentials to alter
+ * @group_info: The group list to install
+ *
+ * Validate a group subscription and, if valid, insert it into a set
+ * of credentials.
+ */
+int set_groups(struct cred *new, struct group_info *group_info)
 {
        int retval;
-       struct group_info *old_info;
 
        retval = security_task_setgroups(group_info);
        if (retval)
                return retval;
 
+       put_group_info(new->group_info);
        groups_sort(group_info);
        get_group_info(group_info);
+       new->group_info = group_info;
+       return 0;
+}
 
-       task_lock(current);
-       old_info = current->group_info;
-       current->group_info = group_info;
-       task_unlock(current);
+EXPORT_SYMBOL(set_groups);
 
-       put_group_info(old_info);
+/**
+ * set_current_groups - Change current's group subscription
+ * @group_info: The group list to impose
+ *
+ * Validate a group subscription and, if valid, impose it upon current's task
+ * security record.
+ */
+int set_current_groups(struct group_info *group_info)
+{
+       struct cred *new;
+       int ret;
 
-       return 0;
+       new = prepare_creds();
+       if (!new)
+               return -ENOMEM;
+
+       ret = set_groups(new, group_info);
+       if (ret < 0) {
+               abort_creds(new);
+               return ret;
+       }
+
+       return commit_creds(new);
 }
 
 EXPORT_SYMBOL(set_current_groups);
 
 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
 {
-       int i = 0;
-
-       /*
-        *      SMP: Nobody else can change our grouplist. Thus we are
-        *      safe.
-        */
+       const struct cred *cred = current_cred();
+       int i;
 
        if (gidsetsize < 0)
                return -EINVAL;
 
        /* no need to grab task_lock here; it cannot change */
-       i = current->group_info->ngroups;
+       i = cred->group_info->ngroups;
        if (gidsetsize) {
                if (i > gidsetsize) {
                        i = -EINVAL;
                        goto out;
                }
-               if (groups_to_user(grouplist, current->group_info)) {
+               if (groups_to_user(grouplist, cred->group_info)) {
                        i = -EFAULT;
                        goto out;
                }
@@ -1828,9 +1368,11 @@ asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
  */
 int in_group_p(gid_t grp)
 {
+       const struct cred *cred = current_cred();
        int retval = 1;
-       if (grp != current->fsgid)
-               retval = groups_search(current->group_info, grp);
+
+       if (grp != cred->fsgid)
+               retval = groups_search(cred->group_info, grp);
        return retval;
 }
 
@@ -1838,9 +1380,11 @@ EXPORT_SYMBOL(in_group_p);
 
 int in_egroup_p(gid_t grp)
 {
+       const struct cred *cred = current_cred();
        int retval = 1;
-       if (grp != current->egid)
-               retval = groups_search(current->group_info, grp);
+
+       if (grp != cred->egid)
+               retval = groups_search(cred->group_info, grp);
        return retval;
 }
 
@@ -1848,8 +1392,6 @@ EXPORT_SYMBOL(in_egroup_p);
 
 DECLARE_RWSEM(uts_sem);
 
-EXPORT_SYMBOL(uts_sem);
-
 asmlinkage long sys_newuname(struct new_utsname __user * name)
 {
        int errno = 0;
@@ -1873,8 +1415,10 @@ asmlinkage long sys_sethostname(char __user *name, int len)
        down_write(&uts_sem);
        errno = -EFAULT;
        if (!copy_from_user(tmp, name, len)) {
-               memcpy(utsname()->nodename, tmp, len);
-               utsname()->nodename[len] = 0;
+               struct new_utsname *u = utsname();
+
+               memcpy(u->nodename, tmp, len);
+               memset(u->nodename + len, 0, sizeof(u->nodename) - len);
                errno = 0;
        }
        up_write(&uts_sem);
@@ -1886,15 +1430,17 @@ asmlinkage long sys_sethostname(char __user *name, int len)
 asmlinkage long sys_gethostname(char __user *name, int len)
 {
        int i, errno;
+       struct new_utsname *u;
 
        if (len < 0)
                return -EINVAL;
        down_read(&uts_sem);
-       i = 1 + strlen(utsname()->nodename);
+       u = utsname();
+       i = 1 + strlen(u->nodename);
        if (i > len)
                i = len;
        errno = 0;
-       if (copy_to_user(name, utsname()->nodename, i))
+       if (copy_to_user(name, u->nodename, i))
                errno = -EFAULT;
        up_read(&uts_sem);
        return errno;
@@ -1919,8 +1465,10 @@ asmlinkage long sys_setdomainname(char __user *name, int len)
        down_write(&uts_sem);
        errno = -EFAULT;
        if (!copy_from_user(tmp, name, len)) {
-               memcpy(utsname()->domainname, tmp, len);
-               utsname()->domainname[len] = 0;
+               struct new_utsname *u = utsname();
+
+               memcpy(u->domainname, tmp, len);
+               memset(u->domainname + len, 0, sizeof(u->domainname) - len);
                errno = 0;
        }
        up_write(&uts_sem);
@@ -1967,21 +1515,28 @@ asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *r
 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
 {
        struct rlimit new_rlim, *old_rlim;
-       unsigned long it_prof_secs;
        int retval;
 
        if (resource >= RLIM_NLIMITS)
                return -EINVAL;
        if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
                return -EFAULT;
-       if (new_rlim.rlim_cur > new_rlim.rlim_max)
-               return -EINVAL;
        old_rlim = current->signal->rlim + resource;
        if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
            !capable(CAP_SYS_RESOURCE))
                return -EPERM;
-       if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
-               return -EPERM;
+
+       if (resource == RLIMIT_NOFILE) {
+               if (new_rlim.rlim_max == RLIM_INFINITY)
+                       new_rlim.rlim_max = sysctl_nr_open;
+               if (new_rlim.rlim_cur == RLIM_INFINITY)
+                       new_rlim.rlim_cur = sysctl_nr_open;
+               if (new_rlim.rlim_max > sysctl_nr_open)
+                       return -EPERM;
+       }
+
+       if (new_rlim.rlim_cur > new_rlim.rlim_max)
+               return -EINVAL;
 
        retval = security_task_setrlimit(resource, &new_rlim);
        if (retval)
@@ -2013,18 +1568,7 @@ asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
        if (new_rlim.rlim_cur == RLIM_INFINITY)
                goto out;
 
-       it_prof_secs = cputime_to_secs(current->signal->it_prof_expires);
-       if (it_prof_secs == 0 || new_rlim.rlim_cur <= it_prof_secs) {
-               unsigned long rlim_cur = new_rlim.rlim_cur;
-               cputime_t cputime;
-
-               cputime = secs_to_cputime(rlim_cur);
-               read_lock(&tasklist_lock);
-               spin_lock_irq(&current->sighand->siglock);
-               set_process_cpu_timer(current, CPUCLOCK_PROF, &cputime, NULL);
-               spin_unlock_irq(&current->sighand->siglock);
-               read_unlock(&tasklist_lock);
-       }
+       update_rlimit_cpu(new_rlim.rlim_cur);
 out:
        return 0;
 }
@@ -2062,21 +1606,34 @@ out:
  *
  */
 
+static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
+{
+       r->ru_nvcsw += t->nvcsw;
+       r->ru_nivcsw += t->nivcsw;
+       r->ru_minflt += t->min_flt;
+       r->ru_majflt += t->maj_flt;
+       r->ru_inblock += task_io_get_inblock(t);
+       r->ru_oublock += task_io_get_oublock(t);
+}
+
 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
 {
        struct task_struct *t;
        unsigned long flags;
        cputime_t utime, stime;
+       struct task_cputime cputime;
 
        memset((char *) r, 0, sizeof *r);
        utime = stime = cputime_zero;
 
-       rcu_read_lock();
-       if (!lock_task_sighand(p, &flags)) {
-               rcu_read_unlock();
-               return;
+       if (who == RUSAGE_THREAD) {
+               accumulate_thread_rusage(p, r);
+               goto out;
        }
 
+       if (!lock_task_sighand(p, &flags))
+               return;
+
        switch (who) {
                case RUSAGE_BOTH:
                case RUSAGE_CHILDREN:
@@ -2093,8 +1650,9 @@ static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
                                break;
 
                case RUSAGE_SELF:
-                       utime = cputime_add(utime, p->signal->utime);
-                       stime = cputime_add(stime, p->signal->stime);
+                       thread_group_cputime(p, &cputime);
+                       utime = cputime_add(utime, cputime.utime);
+                       stime = cputime_add(stime, cputime.stime);
                        r->ru_nvcsw += p->signal->nvcsw;
                        r->ru_nivcsw += p->signal->nivcsw;
                        r->ru_minflt += p->signal->min_flt;
@@ -2103,14 +1661,7 @@ static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
                        r->ru_oublock += p->signal->oublock;
                        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;
-                               r->ru_inblock += task_io_get_inblock(t);
-                               r->ru_oublock += task_io_get_oublock(t);
+                               accumulate_thread_rusage(t, r);
                                t = next_thread(t);
                        } while (t != p);
                        break;
@@ -2118,10 +1669,9 @@ static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
                default:
                        BUG();
        }
-
        unlock_task_sighand(p, &flags);
-       rcu_read_unlock();
 
+out:
        cputime_to_timeval(utime, &r->ru_utime);
        cputime_to_timeval(stime, &r->ru_stime);
 }
@@ -2135,7 +1685,8 @@ int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
 
 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
 {
-       if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
+       if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
+           who != RUSAGE_THREAD)
                return -EINVAL;
        return getrusage(current, who, ru);
 }
@@ -2145,102 +1696,89 @@ asmlinkage long sys_umask(int mask)
        mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
        return mask;
 }
-    
+
 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
                          unsigned long arg4, unsigned long arg5)
 {
+       struct task_struct *me = current;
+       unsigned char comm[sizeof(me->comm)];
        long error;
 
        error = security_task_prctl(option, arg2, arg3, arg4, arg5);
-       if (error)
+       if (error != -ENOSYS)
                return error;
 
+       error = 0;
        switch (option) {
                case PR_SET_PDEATHSIG:
                        if (!valid_signal(arg2)) {
                                error = -EINVAL;
                                break;
                        }
-                       current->pdeath_signal = arg2;
+                       me->pdeath_signal = arg2;
+                       error = 0;
                        break;
                case PR_GET_PDEATHSIG:
-                       error = put_user(current->pdeath_signal, (int __user *)arg2);
+                       error = put_user(me->pdeath_signal, (int __user *)arg2);
                        break;
                case PR_GET_DUMPABLE:
-                       error = current->mm->dumpable;
+                       error = get_dumpable(me->mm);
                        break;
                case PR_SET_DUMPABLE:
                        if (arg2 < 0 || arg2 > 1) {
                                error = -EINVAL;
                                break;
                        }
-                       current->mm->dumpable = arg2;
+                       set_dumpable(me->mm, arg2);
+                       error = 0;
                        break;
 
                case PR_SET_UNALIGN:
-                       error = SET_UNALIGN_CTL(current, arg2);
+                       error = SET_UNALIGN_CTL(me, arg2);
                        break;
                case PR_GET_UNALIGN:
-                       error = GET_UNALIGN_CTL(current, arg2);
+                       error = GET_UNALIGN_CTL(me, arg2);
                        break;
                case PR_SET_FPEMU:
-                       error = SET_FPEMU_CTL(current, arg2);
+                       error = SET_FPEMU_CTL(me, arg2);
                        break;
                case PR_GET_FPEMU:
-                       error = GET_FPEMU_CTL(current, arg2);
+                       error = GET_FPEMU_CTL(me, arg2);
                        break;
                case PR_SET_FPEXC:
-                       error = SET_FPEXC_CTL(current, arg2);
+                       error = SET_FPEXC_CTL(me, arg2);
                        break;
                case PR_GET_FPEXC:
-                       error = GET_FPEXC_CTL(current, arg2);
+                       error = GET_FPEXC_CTL(me, arg2);
                        break;
                case PR_GET_TIMING:
                        error = PR_TIMING_STATISTICAL;
                        break;
                case PR_SET_TIMING:
-                       if (arg2 == PR_TIMING_STATISTICAL)
-                               error = 0;
-                       else
+                       if (arg2 != PR_TIMING_STATISTICAL)
                                error = -EINVAL;
+                       else
+                               error = 0;
                        break;
 
-               case PR_GET_KEEPCAPS:
-                       if (current->keep_capabilities)
-                               error = 1;
-                       break;
-               case PR_SET_KEEPCAPS:
-                       if (arg2 != 0 && arg2 != 1) {
-                               error = -EINVAL;
-                               break;
-                       }
-                       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)
+               case PR_SET_NAME:
+                       comm[sizeof(me->comm)-1] = 0;
+                       if (strncpy_from_user(comm, (char __user *)arg2,
+                                             sizeof(me->comm) - 1) < 0)
                                return -EFAULT;
-                       set_task_comm(me, ncomm);
+                       set_task_comm(me, comm);
                        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)))
+               case PR_GET_NAME:
+                       get_task_comm(comm, me);
+                       if (copy_to_user((char __user *)arg2, comm,
+                                        sizeof(comm)))
                                return -EFAULT;
                        return 0;
-               }
                case PR_GET_ENDIAN:
-                       error = GET_ENDIAN(current, arg2);
+                       error = GET_ENDIAN(me, arg2);
                        break;
                case PR_SET_ENDIAN:
-                       error = SET_ENDIAN(current, arg2);
+                       error = SET_ENDIAN(me, arg2);
                        break;
 
                case PR_GET_SECCOMP:
@@ -2249,7 +1787,23 @@ asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
                case PR_SET_SECCOMP:
                        error = prctl_set_seccomp(arg2);
                        break;
-
+               case PR_GET_TSC:
+                       error = GET_TSC_CTL(arg2);
+                       break;
+               case PR_SET_TSC:
+                       error = SET_TSC_CTL(arg2);
+                       break;
+               case PR_GET_TIMERSLACK:
+                       error = current->timer_slack_ns;
+                       break;
+               case PR_SET_TIMERSLACK:
+                       if (arg2 <= 0)
+                               current->timer_slack_ns =
+                                       current->default_timer_slack_ns;
+                       else
+                               current->timer_slack_ns = arg2;
+                       error = 0;
+                       break;
                default:
                        error = -EINVAL;
                        break;
@@ -2258,7 +1812,7 @@ asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
 }
 
 asmlinkage long sys_getcpu(unsigned __user *cpup, unsigned __user *nodep,
-                          struct getcpu_cache __user *cache)
+                          struct getcpu_cache __user *unused)
 {
        int err = 0;
        int cpu = raw_smp_processor_id();
@@ -2266,24 +1820,6 @@ asmlinkage long sys_getcpu(unsigned __user *cpup, unsigned __user *nodep,
                err |= put_user(cpu, cpup);
        if (nodep)
                err |= put_user(cpu_to_node(cpu), nodep);
-       if (cache) {
-               /*
-                * The cache is not needed for this implementation,
-                * but make sure user programs pass something
-                * valid. vsyscall implementations can instead make
-                * good use of the cache. Only use t0 and t1 because
-                * these are available in both 32bit and 64bit ABI (no
-                * need for a compat_getcpu). 32bit has enough
-                * padding
-                */
-               unsigned long t0, t1;
-               get_user(t0, &cache->blob[0]);
-               get_user(t1, &cache->blob[1]);
-               t0++;
-               t1++;
-               put_user(t0, &cache->blob[0]);
-               put_user(t1, &cache->blob[1]);
-       }
        return err ? -EFAULT : 0;
 }
 
@@ -2319,7 +1855,7 @@ int orderly_poweroff(bool force)
                goto out;
        }
 
-       info = call_usermodehelper_setup(argv[0], argv, envp);
+       info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
        if (info == NULL) {
                argv_free(argv);
                goto out;