Merge git://git.kernel.org/pub/scm/linux/kernel/git/tglx/linux-2.6-hrt
[pandora-kernel.git] / kernel / fork.c
index 3fc3c13..ddafdfa 100644 (file)
@@ -29,7 +29,7 @@
 #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>
@@ -50,6 +50,7 @@
 #include <linux/taskstats_kern.h>
 #include <linux/random.h>
 #include <linux/tty.h>
+#include <linux/proc_fs.h>
 
 #include <asm/pgtable.h>
 #include <asm/pgalloc.h>
@@ -107,6 +108,7 @@ static struct kmem_cache *mm_cachep;
 
 void free_task(struct task_struct *tsk)
 {
+       prop_local_destroy_single(&tsk->dirties);
        free_thread_info(tsk->stack);
        rt_mutex_debug_task_free(tsk);
        free_task_struct(tsk);
@@ -115,7 +117,7 @@ EXPORT_SYMBOL(free_task);
 
 void __put_task_struct(struct task_struct *tsk)
 {
-       WARN_ON(!(tsk->exit_state & (EXIT_DEAD | EXIT_ZOMBIE)));
+       WARN_ON(!tsk->exit_state);
        WARN_ON(atomic_read(&tsk->usage));
        WARN_ON(tsk == current);
 
@@ -163,6 +165,7 @@ 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);
 
@@ -178,6 +181,14 @@ static struct task_struct *dup_task_struct(struct task_struct *orig)
 
        *tsk = *orig;
        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
@@ -195,7 +206,7 @@ static struct task_struct *dup_task_struct(struct task_struct *orig)
 }
 
 #ifdef CONFIG_MMU
-static inline int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
+static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
 {
        struct vm_area_struct *mpnt, *tmp, **pprev;
        struct rb_node **rb_link, *rb_parent;
@@ -258,7 +269,7 @@ static inline int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
                        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;
@@ -321,7 +332,7 @@ 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)))
@@ -573,7 +584,7 @@ fail_nomem:
        return retval;
 }
 
-static inline struct fs_struct *__copy_fs_struct(struct fs_struct *old)
+static struct fs_struct *__copy_fs_struct(struct fs_struct *old)
 {
        struct fs_struct *fs = kmem_cache_alloc(fs_cachep, GFP_KERNEL);
        /* We don't need to lock fs - think why ;-) */
@@ -605,7 +616,7 @@ struct fs_struct *copy_fs_struct(struct fs_struct *old)
 
 EXPORT_SYMBOL_GPL(copy_fs_struct);
 
-static inline int copy_fs(unsigned long clone_flags, struct task_struct * tsk)
+static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
 {
        if (clone_flags & CLONE_FS) {
                atomic_inc(&current->fs->count);
@@ -728,8 +739,8 @@ 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_fds > open_files) {
                int left = (new_fdt->max_fds-open_files)/8;
@@ -808,7 +819,7 @@ int unshare_files(void)
 
 EXPORT_SYMBOL(unshare_files);
 
-static inline int copy_sighand(unsigned long clone_flags, struct task_struct * tsk)
+static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
 {
        struct sighand_struct *sig;
 
@@ -831,7 +842,7 @@ void __cleanup_sighand(struct sighand_struct *sighand)
                kmem_cache_free(sighand_cachep, sighand);
 }
 
-static inline int copy_signal(unsigned long clone_flags, struct task_struct * tsk)
+static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
 {
        struct signal_struct *sig;
        int ret;
@@ -913,7 +924,7 @@ void __cleanup_signal(struct signal_struct *sig)
        kmem_cache_free(signal_cachep, sig);
 }
 
-static inline void cleanup_signal(struct task_struct *tsk)
+static void cleanup_signal(struct task_struct *tsk)
 {
        struct signal_struct *sig = tsk->signal;
 
@@ -923,7 +934,7 @@ static inline void cleanup_signal(struct task_struct *tsk)
                __cleanup_signal(sig);
 }
 
-static inline void copy_flags(unsigned long clone_flags, struct task_struct *p)
+static void copy_flags(unsigned long clone_flags, struct task_struct *p)
 {
        unsigned long new_flags = p->flags;
 
@@ -932,16 +943,17 @@ static inline void copy_flags(unsigned long clone_flags, struct task_struct *p)
        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)
 {
        current->clear_child_tid = tidptr;
 
-       return current->pid;
+       return task_pid_vnr(current);
 }
 
-static inline void rt_mutex_init_task(struct task_struct *p)
+static void rt_mutex_init_task(struct task_struct *p)
 {
        spin_lock_init(&p->pi_lock);
 #ifdef CONFIG_RT_MUTEXES
@@ -962,12 +974,12 @@ static struct task_struct *copy_process(unsigned long clone_flags,
                                        unsigned long stack_start,
                                        struct pt_regs *regs,
                                        unsigned long stack_size,
-                                       int __user *parent_tidptr,
                                        int __user *child_tidptr,
                                        struct pid *pid)
 {
        int retval;
-       struct task_struct *p = NULL;
+       struct task_struct *p;
+       int cgroup_callbacks_done = 0;
 
        if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
                return ERR_PTR(-EINVAL);
@@ -1031,12 +1043,6 @@ static struct task_struct *copy_process(unsigned long clone_flags,
        p->did_exec = 0;
        delayacct_tsk_init(p);  /* Must remain after dup_task_struct() */
        copy_flags(clone_flags, p);
-       p->pid = pid_nr(pid);
-       retval = -EFAULT;
-       if (clone_flags & CLONE_PARENT_SETTID)
-               if (put_user(p->pid, parent_tidptr))
-                       goto bad_fork_cleanup_delays_binfmt;
-
        INIT_LIST_HEAD(&p->children);
        INIT_LIST_HEAD(&p->sibling);
        p->vfork_done = NULL;
@@ -1048,6 +1054,8 @@ static struct task_struct *copy_process(unsigned long clone_flags,
        p->utime = cputime_zero;
        p->stime = cputime_zero;
        p->gtime = cputime_zero;
+       p->utimescaled = cputime_zero;
+       p->stimescaled = cputime_zero;
 
 #ifdef CONFIG_TASK_XACCT
        p->rchar = 0;           /* I/O counter: bytes read */
@@ -1058,28 +1066,29 @@ static struct task_struct *copy_process(unsigned long clone_flags,
        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
@@ -1112,10 +1121,6 @@ static struct task_struct *copy_process(unsigned long clone_flags,
        p->blocked_on = NULL; /* not blocked yet */
 #endif
 
-       p->tgid = p->pid;
-       if (clone_flags & CLONE_THREAD)
-               p->tgid = current->tgid;
-
        if ((retval = security_task_alloc(p)))
                goto bad_fork_cleanup_policy;
        if ((retval = audit_alloc(p)))
@@ -1141,18 +1146,37 @@ static struct task_struct *copy_process(unsigned long clone_flags,
        if (retval)
                goto bad_fork_cleanup_namespaces;
 
+       if (pid != &init_struct_pid) {
+               retval = -ENOMEM;
+               pid = alloc_pid(task_active_pid_ns(p));
+               if (!pid)
+                       goto bad_fork_cleanup_namespaces;
+
+               if (clone_flags & CLONE_NEWPID) {
+                       retval = pid_ns_prepare_proc(task_active_pid_ns(p));
+                       if (retval < 0)
+                               goto bad_fork_free_pid;
+               }
+       }
+
+       p->pid = pid_nr(pid);
+       p->tgid = p->pid;
+       if (clone_flags & CLONE_THREAD)
+               p->tgid = current->tgid;
+
        p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
        /*
         * 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
         */
@@ -1189,6 +1213,12 @@ 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);
 
@@ -1226,12 +1256,12 @@ 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);
                retval = -ERESTARTNOINTR;
-               goto bad_fork_cleanup_namespaces;
+               goto bad_fork_free_pid;
        }
 
        if (clone_flags & CLONE_THREAD) {
@@ -1260,11 +1290,22 @@ static struct task_struct *copy_process(unsigned long clone_flags,
                        __ptrace_link(p, current->parent);
 
                if (thread_group_leader(p)) {
-                       p->signal->tty = current->signal->tty;
-                       p->signal->pgrp = process_group(current);
-                       set_signal_session(p->signal, process_session(current));
-                       attach_pid(p, PIDTYPE_PGID, task_pgrp(current));
-                       attach_pid(p, PIDTYPE_SID, task_session(current));
+                       if (clone_flags & CLONE_NEWPID) {
+                               p->nsproxy->pid_ns->child_reaper = p;
+                               p->signal->tty = NULL;
+                               set_task_pgrp(p, p->pid);
+                               set_task_session(p, p->pid);
+                               attach_pid(p, PIDTYPE_PGID, pid);
+                               attach_pid(p, PIDTYPE_SID, pid);
+                       } else {
+                               p->signal->tty = current->signal->tty;
+                               set_task_pgrp(p, 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)++;
@@ -1277,8 +1318,12 @@ 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_free_pid:
+       if (pid != &init_struct_pid)
+               free_pid(pid);
 bad_fork_cleanup_namespaces:
        exit_task_namespaces(p);
 bad_fork_cleanup_keys:
@@ -1303,10 +1348,9 @@ 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);
-bad_fork_cleanup_delays_binfmt:
+       cgroup_exit(p, cgroup_callbacks_done);
        delayacct_tsk_free(p);
        if (p->binfmt)
                module_put(p->binfmt->module);
@@ -1333,7 +1377,7 @@ 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,
+       task = copy_process(CLONE_VM, 0, idle_regs(&regs), 0, NULL,
                                &init_struct_pid);
        if (!IS_ERR(task))
                init_idle(task, cpu);
@@ -1341,7 +1385,7 @@ struct task_struct * __cpuinit fork_idle(int cpu)
        return task;
 }
 
-static inline int fork_traceflag (unsigned clone_flags)
+static int fork_traceflag(unsigned clone_flags)
 {
        if (clone_flags & CLONE_UNTRACED)
                return 0;
@@ -1372,19 +1416,16 @@ 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, pid);
+       p = copy_process(clone_flags, stack_start, regs, stack_size,
+                       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.
@@ -1392,6 +1433,17 @@ long do_fork(unsigned long clone_flags,
        if (!IS_ERR(p)) {
                struct completion vfork;
 
+               /*
+                * this is enough to call pid_nr_ns here, but this if
+                * improves optimisation of regular fork()
+                */
+               nr = (clone_flags & CLONE_NEWPID) ?
+                       task_pid_nr_ns(p, current->nsproxy->pid_ns) :
+                               task_pid_vnr(p);
+
+               if (clone_flags & CLONE_PARENT_SETTID)
+                       put_user(nr, parent_tidptr);
+
                if (clone_flags & CLONE_VFORK) {
                        p->vfork_done = &vfork;
                        init_completion(&vfork);
@@ -1425,7 +1477,6 @@ long do_fork(unsigned long clone_flags,
                        }
                }
        } else {
-               free_pid(pid);
                nr = PTR_ERR(p);
        }
        return nr;
@@ -1435,8 +1486,7 @@ 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;
 
@@ -1471,7 +1521,7 @@ void __init proc_caches_init(void)
  * Check constraints on flags passed to the unshare system call and
  * force unsharing of additional process context as appropriate.
  */
-static inline void check_unshare_flags(unsigned long *flags_ptr)
+static void check_unshare_flags(unsigned long *flags_ptr)
 {
        /*
         * If unsharing a thread from a thread group, must also
@@ -1603,7 +1653,7 @@ asmlinkage long sys_unshare(unsigned long unshare_flags)
        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 nsproxy *new_nsproxy = NULL;
 
        check_unshare_flags(&unshare_flags);
 
@@ -1633,14 +1683,13 @@ asmlinkage long sys_unshare(unsigned long unshare_flags)
 
        if (new_fs ||  new_mm || new_fd || new_ulist || new_nsproxy) {
 
-               task_lock(current);
-
                if (new_nsproxy) {
-                       old_nsproxy = current->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;