Make access to task's nsproxy lighter
[safe/jmp/linux-2.6] / kernel / fork.c
index f37980d..d2f4a42 100644 (file)
 #include <linux/slab.h>
 #include <linux/init.h>
 #include <linux/unistd.h>
-#include <linux/smp_lock.h>
 #include <linux/module.h>
 #include <linux/vmalloc.h>
 #include <linux/completion.h>
-#include <linux/namespace.h>
+#include <linux/mnt_namespace.h>
 #include <linux/personality.h>
 #include <linux/mempolicy.h>
 #include <linux/sem.h>
 #include <linux/nsproxy.h>
 #include <linux/capability.h>
 #include <linux/cpu.h>
-#include <linux/cpuset.h>
+#include <linux/cgroup.h>
 #include <linux/security.h>
 #include <linux/swap.h>
 #include <linux/syscalls.h>
 #include <linux/jiffies.h>
 #include <linux/futex.h>
+#include <linux/task_io_accounting_ops.h>
 #include <linux/rcupdate.h>
 #include <linux/ptrace.h>
 #include <linux/mount.h>
 #include <linux/acct.h>
 #include <linux/tsacct_kern.h>
 #include <linux/cn_proc.h>
+#include <linux/freezer.h>
 #include <linux/delayacct.h>
 #include <linux/taskstats_kern.h>
 #include <linux/random.h>
+#include <linux/tty.h>
 
 #include <asm/pgtable.h>
 #include <asm/pgalloc.h>
@@ -105,7 +107,8 @@ static struct kmem_cache *mm_cachep;
 
 void free_task(struct task_struct *tsk)
 {
-       free_thread_info(tsk->thread_info);
+       prop_local_destroy_single(&tsk->dirties);
+       free_thread_info(tsk->stack);
        rt_mutex_debug_task_free(tsk);
        free_task_struct(tsk);
 }
@@ -135,7 +138,7 @@ void __init fork_init(unsigned long mempages)
        /* create a slab on which task_structs can be allocated */
        task_struct_cachep =
                kmem_cache_create("task_struct", sizeof(struct task_struct),
-                       ARCH_MIN_TASKALIGN, SLAB_PANIC, NULL, NULL);
+                       ARCH_MIN_TASKALIGN, SLAB_PANIC, NULL);
 #endif
 
        /*
@@ -161,6 +164,7 @@ static struct task_struct *dup_task_struct(struct task_struct *orig)
 {
        struct task_struct *tsk;
        struct thread_info *ti;
+       int err;
 
        prepare_to_copy(orig);
 
@@ -175,7 +179,15 @@ static struct task_struct *dup_task_struct(struct task_struct *orig)
        }
 
        *tsk = *orig;
-       tsk->thread_info = ti;
+       tsk->stack = ti;
+
+       err = prop_local_init_single(&tsk->dirties);
+       if (err) {
+               free_thread_info(ti);
+               free_task_struct(tsk);
+               return NULL;
+       }
+
        setup_thread_stack(tsk, orig);
 
 #ifdef CONFIG_CC_STACKPROTECTOR
@@ -202,7 +214,7 @@ static inline int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
        struct mempolicy *pol;
 
        down_write(&oldmm->mmap_sem);
-       flush_cache_mm(oldmm);
+       flush_cache_dup_mm(oldmm);
        /*
         * Not linked in yet - no deadlock potential:
         */
@@ -252,11 +264,11 @@ static inline int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
                anon_vma_link(tmp);
                file = tmp->vm_file;
                if (file) {
-                       struct inode *inode = file->f_dentry->d_inode;
+                       struct inode *inode = file->f_path.dentry->d_inode;
                        get_file(file);
                        if (tmp->vm_flags & VM_DENYWRITE)
                                atomic_dec(&inode->i_writecount);
-      
+
                        /* insert tmp into the share list, just after mpnt */
                        spin_lock(&file->f_mapping->i_mmap_lock);
                        tmp->vm_truncate_count = mpnt->vm_truncate_count;
@@ -285,6 +297,8 @@ static inline int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
                if (retval)
                        goto out;
        }
+       /* a new mm has just been created */
+       arch_dup_mmap(oldmm, mm);
        retval = 0;
 out:
        up_write(&mm->mmap_sem);
@@ -317,7 +331,7 @@ static inline void mm_free_pgd(struct mm_struct * mm)
 #define mm_free_pgd(mm)
 #endif /* CONFIG_MMU */
 
- __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
+__cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
 
 #define allocate_mm()  (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
 #define free_mm(mm)    (kmem_cache_free(mm_cachep, (mm)))
@@ -330,6 +344,8 @@ static struct mm_struct * mm_init(struct mm_struct * mm)
        atomic_set(&mm->mm_count, 1);
        init_rwsem(&mm->mmap_sem);
        INIT_LIST_HEAD(&mm->mmlist);
+       mm->flags = (current->mm) ? current->mm->flags
+                                 : MMF_DUMP_FILTER_DEFAULT;
        mm->core_waiters = 0;
        mm->nr_ptes = 0;
        set_mm_counter(mm, file_rss, 0);
@@ -613,7 +629,7 @@ static inline int copy_fs(unsigned long clone_flags, struct task_struct * tsk)
 
 static int count_open_files(struct fdtable *fdt)
 {
-       int size = fdt->max_fdset;
+       int size = fdt->max_fds;
        int i;
 
        /* Find the last open fd */
@@ -640,12 +656,10 @@ static struct files_struct *alloc_files(void)
        newf->next_fd = 0;
        fdt = &newf->fdtab;
        fdt->max_fds = NR_OPEN_DEFAULT;
-       fdt->max_fdset = EMBEDDED_FD_SET_SIZE;
        fdt->close_on_exec = (fd_set *)&newf->close_on_exec_init;
        fdt->open_fds = (fd_set *)&newf->open_fds_init;
        fdt->fd = &newf->fd_array[0];
        INIT_RCU_HEAD(&fdt->rcu);
-       fdt->free_files = NULL;
        fdt->next = NULL;
        rcu_assign_pointer(newf->fdt, fdt);
 out:
@@ -661,7 +675,7 @@ static struct files_struct *dup_fd(struct files_struct *oldf, int *errorp)
 {
        struct files_struct *newf;
        struct file **old_fds, **new_fds;
-       int open_files, size, i, expand;
+       int open_files, size, i;
        struct fdtable *old_fdt, *new_fdt;
 
        *errorp = -ENOMEM;
@@ -672,25 +686,14 @@ static struct files_struct *dup_fd(struct files_struct *oldf, int *errorp)
        spin_lock(&oldf->file_lock);
        old_fdt = files_fdtable(oldf);
        new_fdt = files_fdtable(newf);
-       size = old_fdt->max_fdset;
        open_files = count_open_files(old_fdt);
-       expand = 0;
 
        /*
-        * Check whether we need to allocate a larger fd array or fd set.
-        * Note: we're not a clone task, so the open count won't  change.
+        * Check whether we need to allocate a larger fd array and fd set.
+        * Note: we're not a clone task, so the open count won't change.
         */
-       if (open_files > new_fdt->max_fdset) {
-               new_fdt->max_fdset = 0;
-               expand = 1;
-       }
        if (open_files > new_fdt->max_fds) {
                new_fdt->max_fds = 0;
-               expand = 1;
-       }
-
-       /* if the old fdset gets grown now, we'll only copy up to "size" fds */
-       if (expand) {
                spin_unlock(&oldf->file_lock);
                spin_lock(&newf->file_lock);
                *errorp = expand_files(newf, open_files-1);
@@ -710,8 +713,10 @@ static struct files_struct *dup_fd(struct files_struct *oldf, int *errorp)
        old_fds = old_fdt->fd;
        new_fds = new_fdt->fd;
 
-       memcpy(new_fdt->open_fds->fds_bits, old_fdt->open_fds->fds_bits, open_files/8);
-       memcpy(new_fdt->close_on_exec->fds_bits, old_fdt->close_on_exec->fds_bits, open_files/8);
+       memcpy(new_fdt->open_fds->fds_bits,
+               old_fdt->open_fds->fds_bits, open_files/8);
+       memcpy(new_fdt->close_on_exec->fds_bits,
+               old_fdt->close_on_exec->fds_bits, open_files/8);
 
        for (i = open_files; i != 0; i--) {
                struct file *f = *old_fds++;
@@ -733,25 +738,22 @@ static struct files_struct *dup_fd(struct files_struct *oldf, int *errorp)
        /* compute the remainder to be cleared */
        size = (new_fdt->max_fds - open_files) * sizeof(struct file *);
 
-       /* This is long word aligned thus could use a optimized version */ 
-       memset(new_fds, 0, size); 
+       /* This is long word aligned thus could use a optimized version */
+       memset(new_fds, 0, size);
 
-       if (new_fdt->max_fdset > open_files) {
-               int left = (new_fdt->max_fdset-open_files)/8;
+       if (new_fdt->max_fds > open_files) {
+               int left = (new_fdt->max_fds-open_files)/8;
                int start = open_files / (8 * sizeof(unsigned long));
 
                memset(&new_fdt->open_fds->fds_bits[start], 0, left);
                memset(&new_fdt->close_on_exec->fds_bits[start], 0, left);
        }
 
-out:
        return newf;
 
 out_release:
-       free_fdset (new_fdt->close_on_exec, new_fdt->max_fdset);
-       free_fdset (new_fdt->open_fds, new_fdt->max_fdset);
-       free_fd_array(new_fdt->fd, new_fdt->max_fds);
        kmem_cache_free(files_cachep, newf);
+out:
        return NULL;
 }
 
@@ -847,7 +849,6 @@ static inline int copy_signal(unsigned long clone_flags, struct task_struct * ts
        if (clone_flags & CLONE_THREAD) {
                atomic_inc(&current->signal->count);
                atomic_inc(&current->signal->live);
-               taskstats_tgid_alloc(current);
                return 0;
        }
        sig = kmem_cache_alloc(signal_cachep, GFP_KERNEL);
@@ -872,7 +873,7 @@ static inline int copy_signal(unsigned long clone_flags, struct task_struct * ts
        init_sigpending(&sig->shared_pending);
        INIT_LIST_HEAD(&sig->posix_timers);
 
-       hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_REL);
+       hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
        sig->it_real_incr.tv64 = 0;
        sig->real_timer.function = it_real_fn;
        sig->tsk = tsk;
@@ -883,12 +884,15 @@ static inline int copy_signal(unsigned long clone_flags, struct task_struct * ts
        sig->it_prof_incr = cputime_zero;
 
        sig->leader = 0;        /* session leadership doesn't inherit */
-       sig->tty_old_pgrp = 0;
+       sig->tty_old_pgrp = NULL;
 
        sig->utime = sig->stime = sig->cutime = sig->cstime = cputime_zero;
+       sig->gtime = cputime_zero;
+       sig->cgtime = cputime_zero;
        sig->nvcsw = sig->nivcsw = sig->cnvcsw = sig->cnivcsw = 0;
        sig->min_flt = sig->maj_flt = sig->cmin_flt = sig->cmaj_flt = 0;
-       sig->sched_time = 0;
+       sig->inblock = sig->oublock = sig->cinblock = sig->coublock = 0;
+       sig->sum_sched_runtime = 0;
        INIT_LIST_HEAD(&sig->cpu_timers[0]);
        INIT_LIST_HEAD(&sig->cpu_timers[1]);
        INIT_LIST_HEAD(&sig->cpu_timers[2]);
@@ -908,6 +912,8 @@ static inline int copy_signal(unsigned long clone_flags, struct task_struct * ts
        }
        acct_init_pacct(&sig->pacct);
 
+       tty_audit_fork(sig);
+
        return 0;
 }
 
@@ -931,11 +937,12 @@ static inline void copy_flags(unsigned long clone_flags, struct task_struct *p)
 {
        unsigned long new_flags = p->flags;
 
-       new_flags &= ~(PF_SUPERPRIV | PF_NOFREEZE);
+       new_flags &= ~PF_SUPERPRIV;
        new_flags |= PF_FORKNOEXEC;
        if (!(clone_flags & CLONE_PTRACE))
                p->ptrace = 0;
        p->flags = new_flags;
+       clear_freeze_flag(p);
 }
 
 asmlinkage long sys_set_tid_address(int __user *tidptr)
@@ -947,8 +954,8 @@ asmlinkage long sys_set_tid_address(int __user *tidptr)
 
 static inline void rt_mutex_init_task(struct task_struct *p)
 {
-#ifdef CONFIG_RT_MUTEXES
        spin_lock_init(&p->pi_lock);
+#ifdef CONFIG_RT_MUTEXES
        plist_head_init(&p->pi_waiters, &p->pi_lock);
        p->pi_blocked_on = NULL;
 #endif
@@ -968,10 +975,11 @@ static struct task_struct *copy_process(unsigned long clone_flags,
                                        unsigned long stack_size,
                                        int __user *parent_tidptr,
                                        int __user *child_tidptr,
-                                       int pid)
+                                       struct pid *pid)
 {
        int retval;
        struct task_struct *p = NULL;
+       int cgroup_callbacks_done = 0;
 
        if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
                return ERR_PTR(-EINVAL);
@@ -1010,7 +1018,7 @@ static struct task_struct *copy_process(unsigned long clone_flags,
        if (atomic_read(&p->user->processes) >=
                        p->signal->rlim[RLIMIT_NPROC].rlim_cur) {
                if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) &&
-                               p->user != &root_user)
+                   p->user != current->nsproxy->user_ns->root_user)
                        goto bad_fork_free;
        }
 
@@ -1032,10 +1040,16 @@ static struct task_struct *copy_process(unsigned long clone_flags,
        if (p->binfmt && !try_module_get(p->binfmt->module))
                goto bad_fork_cleanup_put_domain;
 
+       if (pid != &init_struct_pid) {
+               pid = alloc_pid();
+               if (!pid)
+                       goto bad_fork_put_binfmt_module;
+       }
+
        p->did_exec = 0;
        delayacct_tsk_init(p);  /* Must remain after dup_task_struct() */
        copy_flags(clone_flags, p);
-       p->pid = pid;
+       p->pid = pid_nr(pid);
        retval = -EFAULT;
        if (clone_flags & CLONE_PARENT_SETTID)
                if (put_user(p->pid, parent_tidptr))
@@ -1051,33 +1065,42 @@ static struct task_struct *copy_process(unsigned long clone_flags,
 
        p->utime = cputime_zero;
        p->stime = cputime_zero;
-       p->sched_time = 0;
+       p->gtime = cputime_zero;
+       p->utimescaled = cputime_zero;
+       p->stimescaled = cputime_zero;
+
+#ifdef CONFIG_TASK_XACCT
        p->rchar = 0;           /* I/O counter: bytes read */
        p->wchar = 0;           /* I/O counter: bytes written */
        p->syscr = 0;           /* I/O counter: read syscalls */
        p->syscw = 0;           /* I/O counter: write syscalls */
+#endif
+       task_io_accounting_init(p);
        acct_clear_integrals(p);
 
-       p->it_virt_expires = cputime_zero;
+       p->it_virt_expires = cputime_zero;
        p->it_prof_expires = cputime_zero;
-       p->it_sched_expires = 0;
-       INIT_LIST_HEAD(&p->cpu_timers[0]);
-       INIT_LIST_HEAD(&p->cpu_timers[1]);
-       INIT_LIST_HEAD(&p->cpu_timers[2]);
+       p->it_sched_expires = 0;
+       INIT_LIST_HEAD(&p->cpu_timers[0]);
+       INIT_LIST_HEAD(&p->cpu_timers[1]);
+       INIT_LIST_HEAD(&p->cpu_timers[2]);
 
        p->lock_depth = -1;             /* -1 = no lock */
        do_posix_clock_monotonic_gettime(&p->start_time);
+       p->real_start_time = p->start_time;
+       monotonic_to_bootbased(&p->real_start_time);
+#ifdef CONFIG_SECURITY
        p->security = NULL;
+#endif
        p->io_context = NULL;
-       p->io_wait = NULL;
        p->audit_context = NULL;
-       cpuset_fork(p);
+       cgroup_fork(p);
 #ifdef CONFIG_NUMA
        p->mempolicy = mpol_copy(p->mempolicy);
        if (IS_ERR(p->mempolicy)) {
                retval = PTR_ERR(p->mempolicy);
                p->mempolicy = NULL;
-               goto bad_fork_cleanup_cpuset;
+               goto bad_fork_cleanup_cgroup;
        }
        mpol_fix_fork_child_flag(p);
 #endif
@@ -1144,13 +1167,14 @@ static struct task_struct *copy_process(unsigned long clone_flags,
         * Clear TID on mm_release()?
         */
        p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr: NULL;
+#ifdef CONFIG_FUTEX
        p->robust_list = NULL;
 #ifdef CONFIG_COMPAT
        p->compat_robust_list = NULL;
 #endif
        INIT_LIST_HEAD(&p->pi_state_list);
        p->pi_state_cache = NULL;
-
+#endif
        /*
         * sigaltstack should be cleared when sharing the same VM
         */
@@ -1187,6 +1211,12 @@ static struct task_struct *copy_process(unsigned long clone_flags,
        /* Perform scheduler related setup. Assign this task to a CPU. */
        sched_fork(p, clone_flags);
 
+       /* Now that the task is set up, run cgroup callbacks if
+        * necessary. We need to run them before the task is visible
+        * on the tasklist. */
+       cgroup_fork_callbacks(p);
+       cgroup_callbacks_done = 1;
+
        /* Need tasklist lock for parent etc handling! */
        write_lock_irq(&tasklist_lock);
 
@@ -1224,7 +1254,7 @@ static struct task_struct *copy_process(unsigned long clone_flags,
         * A fatal signal pending means that current will exit, so the new
         * thread can't slip out of an OOM kill (or normal SIGKILL).
         */
-       recalc_sigpending();
+       recalc_sigpending();
        if (signal_pending(current)) {
                spin_unlock(&current->sighand->siglock);
                write_unlock_irq(&tasklist_lock);
@@ -1259,15 +1289,15 @@ static struct task_struct *copy_process(unsigned long clone_flags,
 
                if (thread_group_leader(p)) {
                        p->signal->tty = current->signal->tty;
-                       p->signal->pgrp = process_group(current);
-                       p->signal->session = current->signal->session;
-                       attach_pid(p, PIDTYPE_PGID, process_group(p));
-                       attach_pid(p, PIDTYPE_SID, p->signal->session);
+                       p->signal->pgrp = task_pgrp_nr(current);
+                       set_task_session(p, task_session_nr(current));
+                       attach_pid(p, PIDTYPE_PGID, task_pgrp(current));
+                       attach_pid(p, PIDTYPE_SID, task_session(current));
 
                        list_add_tail_rcu(&p->tasks, &init_task.tasks);
                        __get_cpu_var(process_counts)++;
                }
-               attach_pid(p, PIDTYPE_PID, p->pid);
+               attach_pid(p, PIDTYPE_PID, pid);
                nr_threads++;
        }
 
@@ -1275,6 +1305,7 @@ static struct task_struct *copy_process(unsigned long clone_flags,
        spin_unlock(&current->sighand->siglock);
        write_unlock_irq(&tasklist_lock);
        proc_fork_connector(p);
+       cgroup_post_fork(p);
        return p;
 
 bad_fork_cleanup_namespaces:
@@ -1301,11 +1332,14 @@ bad_fork_cleanup_security:
 bad_fork_cleanup_policy:
 #ifdef CONFIG_NUMA
        mpol_free(p->mempolicy);
-bad_fork_cleanup_cpuset:
+bad_fork_cleanup_cgroup:
 #endif
-       cpuset_exit(p);
+       cgroup_exit(p, cgroup_callbacks_done);
 bad_fork_cleanup_delays_binfmt:
        delayacct_tsk_free(p);
+       if (pid != &init_struct_pid)
+               free_pid(pid);
+bad_fork_put_binfmt_module:
        if (p->binfmt)
                module_put(p->binfmt->module);
 bad_fork_cleanup_put_domain:
@@ -1320,18 +1354,19 @@ fork_out:
        return ERR_PTR(retval);
 }
 
-struct pt_regs * __devinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
+noinline struct pt_regs * __devinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
 {
        memset(regs, 0, sizeof(struct pt_regs));
        return regs;
 }
 
-struct task_struct * __devinit fork_idle(int cpu)
+struct task_struct * __cpuinit fork_idle(int cpu)
 {
        struct task_struct *task;
        struct pt_regs regs;
 
-       task = copy_process(CLONE_VM, 0, idle_regs(&regs), 0, NULL, NULL, 0);
+       task = copy_process(CLONE_VM, 0, idle_regs(&regs), 0, NULL, NULL,
+                               &init_struct_pid);
        if (!IS_ERR(task))
                init_idle(task, cpu);
 
@@ -1369,19 +1404,16 @@ long do_fork(unsigned long clone_flags,
 {
        struct task_struct *p;
        int trace = 0;
-       struct pid *pid = alloc_pid();
        long nr;
 
-       if (!pid)
-               return -EAGAIN;
-       nr = pid->nr;
        if (unlikely(current->ptrace)) {
                trace = fork_traceflag (clone_flags);
                if (trace)
                        clone_flags |= CLONE_PTRACE;
        }
 
-       p = copy_process(clone_flags, stack_start, regs, stack_size, parent_tidptr, child_tidptr, nr);
+       p = copy_process(clone_flags, stack_start, regs, stack_size,
+                       parent_tidptr, child_tidptr, NULL);
        /*
         * Do this prior waking up the new thread - the thread pointer
         * might get invalid after that point, if the thread exits quickly.
@@ -1389,6 +1421,8 @@ long do_fork(unsigned long clone_flags,
        if (!IS_ERR(p)) {
                struct completion vfork;
 
+               nr = pid_nr(task_pid(p));
+
                if (clone_flags & CLONE_VFORK) {
                        p->vfork_done = &vfork;
                        init_completion(&vfork);
@@ -1413,14 +1447,15 @@ long do_fork(unsigned long clone_flags,
                }
 
                if (clone_flags & CLONE_VFORK) {
+                       freezer_do_not_count();
                        wait_for_completion(&vfork);
+                       freezer_count();
                        if (unlikely (current->ptrace & PT_TRACE_VFORK_DONE)) {
                                current->ptrace_message = nr;
                                ptrace_notify ((PTRACE_EVENT_VFORK_DONE << 8) | SIGTRAP);
                        }
                }
        } else {
-               free_pid(pid);
                nr = PTR_ERR(p);
        }
        return nr;
@@ -1430,13 +1465,12 @@ long do_fork(unsigned long clone_flags,
 #define ARCH_MIN_MMSTRUCT_ALIGN 0
 #endif
 
-static void sighand_ctor(void *data, struct kmem_cache *cachep, unsigned long flags)
+static void sighand_ctor(struct kmem_cache *cachep, void *data)
 {
        struct sighand_struct *sighand = data;
 
-       if ((flags & (SLAB_CTOR_VERIFY | SLAB_CTOR_CONSTRUCTOR)) ==
-                                       SLAB_CTOR_CONSTRUCTOR)
-               spin_lock_init(&sighand->siglock);
+       spin_lock_init(&sighand->siglock);
+       init_waitqueue_head(&sighand->signalfd_wqh);
 }
 
 void __init proc_caches_init(void)
@@ -1444,25 +1478,24 @@ void __init proc_caches_init(void)
        sighand_cachep = kmem_cache_create("sighand_cache",
                        sizeof(struct sighand_struct), 0,
                        SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU,
-                       sighand_ctor, NULL);
+                       sighand_ctor);
        signal_cachep = kmem_cache_create("signal_cache",
                        sizeof(struct signal_struct), 0,
-                       SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
-       files_cachep = kmem_cache_create("files_cache", 
+                       SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
+       files_cachep = kmem_cache_create("files_cache",
                        sizeof(struct files_struct), 0,
-                       SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
-       fs_cachep = kmem_cache_create("fs_cache", 
+                       SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
+       fs_cachep = kmem_cache_create("fs_cache",
                        sizeof(struct fs_struct), 0,
-                       SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
+                       SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
        vm_area_cachep = kmem_cache_create("vm_area_struct",
                        sizeof(struct vm_area_struct), 0,
-                       SLAB_PANIC, NULL, NULL);
+                       SLAB_PANIC, NULL);
        mm_cachep = kmem_cache_create("mm_struct",
                        sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
-                       SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
+                       SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
 }
 
-
 /*
  * Check constraints on flags passed to the unshare system call and
  * force unsharing of additional process context as appropriate.
@@ -1526,34 +1559,13 @@ static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
 }
 
 /*
- * Unshare the namespace structure if it is being shared
- */
-static int unshare_namespace(unsigned long unshare_flags, struct namespace **new_nsp, struct fs_struct *new_fs)
-{
-       struct namespace *ns = current->nsproxy->namespace;
-
-       if ((unshare_flags & CLONE_NEWNS) && ns) {
-               if (!capable(CAP_SYS_ADMIN))
-                       return -EPERM;
-
-               *new_nsp = dup_namespace(current, new_fs ? new_fs : current->fs);
-               if (!*new_nsp)
-                       return -ENOMEM;
-       }
-
-       return 0;
-}
-
-/*
- * Unsharing of sighand for tasks created with CLONE_SIGHAND is not
- * supported yet
+ * Unsharing of sighand is not supported yet
  */
 static int unshare_sighand(unsigned long unshare_flags, struct sighand_struct **new_sighp)
 {
        struct sighand_struct *sigh = current->sighand;
 
-       if ((unshare_flags & CLONE_SIGHAND) &&
-           (sigh && atomic_read(&sigh->count) > 1))
+       if ((unshare_flags & CLONE_SIGHAND) && atomic_read(&sigh->count) > 1)
                return -EINVAL;
        else
                return 0;
@@ -1604,16 +1616,6 @@ static int unshare_semundo(unsigned long unshare_flags, struct sem_undo_list **n
        return 0;
 }
 
-#ifndef CONFIG_IPC_NS
-static inline int unshare_ipcs(unsigned long flags, struct ipc_namespace **ns)
-{
-       if (flags & CLONE_NEWIPC)
-               return -EINVAL;
-
-       return 0;
-}
-#endif
-
 /*
  * unshare allows a process to 'unshare' part of the process
  * context which was originally shared using clone.  copy_*
@@ -1626,14 +1628,11 @@ asmlinkage long sys_unshare(unsigned long unshare_flags)
 {
        int err = 0;
        struct fs_struct *fs, *new_fs = NULL;
-       struct namespace *ns, *new_ns = NULL;
-       struct sighand_struct *sigh, *new_sigh = NULL;
+       struct sighand_struct *new_sigh = NULL;
        struct mm_struct *mm, *new_mm = NULL, *active_mm = NULL;
        struct files_struct *fd, *new_fd = NULL;
        struct sem_undo_list *new_ulist = NULL;
-       struct nsproxy *new_nsproxy = NULL, *old_nsproxy = NULL;
-       struct uts_namespace *uts, *new_uts = NULL;
-       struct ipc_namespace *ipc, *new_ipc = NULL;
+       struct nsproxy *new_nsproxy = NULL;
 
        check_unshare_flags(&unshare_flags);
 
@@ -1641,65 +1640,41 @@ asmlinkage long sys_unshare(unsigned long unshare_flags)
        err = -EINVAL;
        if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
                                CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
-                               CLONE_NEWUTS|CLONE_NEWIPC))
+                               CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWUSER|
+                               CLONE_NEWNET))
                goto bad_unshare_out;
 
        if ((err = unshare_thread(unshare_flags)))
                goto bad_unshare_out;
        if ((err = unshare_fs(unshare_flags, &new_fs)))
                goto bad_unshare_cleanup_thread;
-       if ((err = unshare_namespace(unshare_flags, &new_ns, new_fs)))
-               goto bad_unshare_cleanup_fs;
        if ((err = unshare_sighand(unshare_flags, &new_sigh)))
-               goto bad_unshare_cleanup_ns;
+               goto bad_unshare_cleanup_fs;
        if ((err = unshare_vm(unshare_flags, &new_mm)))
                goto bad_unshare_cleanup_sigh;
        if ((err = unshare_fd(unshare_flags, &new_fd)))
                goto bad_unshare_cleanup_vm;
        if ((err = unshare_semundo(unshare_flags, &new_ulist)))
                goto bad_unshare_cleanup_fd;
-       if ((err = unshare_utsname(unshare_flags, &new_uts)))
+       if ((err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy,
+                       new_fs)))
                goto bad_unshare_cleanup_semundo;
-       if ((err = unshare_ipcs(unshare_flags, &new_ipc)))
-               goto bad_unshare_cleanup_uts;
-
-       if (new_ns || new_uts || new_ipc) {
-               old_nsproxy = current->nsproxy;
-               new_nsproxy = dup_namespaces(old_nsproxy);
-               if (!new_nsproxy) {
-                       err = -ENOMEM;
-                       goto bad_unshare_cleanup_ipc;
-               }
-       }
-
-       if (new_fs || new_ns || new_sigh || new_mm || new_fd || new_ulist ||
-                               new_uts || new_ipc) {
 
-               task_lock(current);
+       if (new_fs ||  new_mm || new_fd || new_ulist || new_nsproxy) {
 
                if (new_nsproxy) {
-                       current->nsproxy = new_nsproxy;
-                       new_nsproxy = old_nsproxy;
+                       switch_task_namespaces(current, new_nsproxy);
+                       new_nsproxy = NULL;
                }
 
+               task_lock(current);
+
                if (new_fs) {
                        fs = current->fs;
                        current->fs = new_fs;
                        new_fs = fs;
                }
 
-               if (new_ns) {
-                       ns = current->nsproxy->namespace;
-                       current->nsproxy->namespace = new_ns;
-                       new_ns = ns;
-               }
-
-               if (new_sigh) {
-                       sigh = current->sighand;
-                       rcu_assign_pointer(current->sighand, new_sigh);
-                       new_sigh = sigh;
-               }
-
                if (new_mm) {
                        mm = current->mm;
                        active_mm = current->active_mm;
@@ -1715,32 +1690,12 @@ asmlinkage long sys_unshare(unsigned long unshare_flags)
                        new_fd = fd;
                }
 
-               if (new_uts) {
-                       uts = current->nsproxy->uts_ns;
-                       current->nsproxy->uts_ns = new_uts;
-                       new_uts = uts;
-               }
-
-               if (new_ipc) {
-                       ipc = current->nsproxy->ipc_ns;
-                       current->nsproxy->ipc_ns = new_ipc;
-                       new_ipc = ipc;
-               }
-
                task_unlock(current);
        }
 
        if (new_nsproxy)
                put_nsproxy(new_nsproxy);
 
-bad_unshare_cleanup_ipc:
-       if (new_ipc)
-               put_ipc_ns(new_ipc);
-
-bad_unshare_cleanup_uts:
-       if (new_uts)
-               put_uts_ns(new_uts);
-
 bad_unshare_cleanup_semundo:
 bad_unshare_cleanup_fd:
        if (new_fd)
@@ -1755,10 +1710,6 @@ bad_unshare_cleanup_sigh:
                if (atomic_dec_and_test(&new_sigh->count))
                        kmem_cache_free(sighand_cachep, new_sigh);
 
-bad_unshare_cleanup_ns:
-       if (new_ns)
-               put_namespace(new_ns);
-
 bad_unshare_cleanup_fs:
        if (new_fs)
                put_fs_struct(new_fs);