active-standby: add cmdline into /proc/net/
[linux-2.6.git] / fs / proc / base.c
index 47338d9..68375eb 100644 (file)
@@ -45,6 +45,8 @@
  *
  *  Paul Mundt <paul.mundt@nokia.com>:
  *  Overall revision about smaps.
+ *
+ * Copyright (c) 2013, NVIDIA CORPORATION.  All rights reserved.
  */
 
 #include <asm/uaccess.h>
 #include <linux/time.h>
 #include <linux/proc_fs.h>
 #include <linux/stat.h>
+#include <linux/task_io_accounting_ops.h>
 #include <linux/init.h>
 #include <linux/capability.h>
 #include <linux/file.h>
+#include <linux/fdtable.h>
 #include <linux/string.h>
 #include <linux/seq_file.h>
 #include <linux/namei.h>
 #include <linux/mnt_namespace.h>
 #include <linux/mm.h>
+#include <linux/swap.h>
 #include <linux/rcupdate.h>
 #include <linux/kallsyms.h>
+#include <linux/stacktrace.h>
 #include <linux/resource.h>
 #include <linux/module.h>
 #include <linux/mount.h>
 #include <linux/security.h>
 #include <linux/ptrace.h>
+#include <linux/tracehook.h>
 #include <linux/cgroup.h>
 #include <linux/cpuset.h>
 #include <linux/audit.h>
 #include <linux/oom.h>
 #include <linux/elf.h>
 #include <linux/pid_namespace.h>
+#include <linux/fs_struct.h>
+#include <linux/slab.h>
+#include <linux/flex_array.h>
+#ifdef CONFIG_HARDWALL
+#include <asm/hardwall.h>
+#endif
+#include <trace/events/oom.h>
 #include "internal.h"
 
 /* NOTE:
 struct pid_entry {
        char *name;
        int len;
-       mode_t mode;
+       umode_t mode;
        const struct inode_operations *iop;
        const struct file_operations *fop;
        union proc_op op;
@@ -106,123 +120,98 @@ struct pid_entry {
        .op   = OP,                                     \
 }
 
-#define DIR(NAME, MODE, OTYPE)                                                 \
-       NOD(NAME, (S_IFDIR|(MODE)),                                             \
-               &proc_##OTYPE##_inode_operations, &proc_##OTYPE##_operations,   \
-               {} )
-#define LNK(NAME, OTYPE)                                       \
+#define DIR(NAME, MODE, iops, fops)    \
+       NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
+#define LNK(NAME, get_link)                                    \
        NOD(NAME, (S_IFLNK|S_IRWXUGO),                          \
                &proc_pid_link_inode_operations, NULL,          \
-               { .proc_get_link = &proc_##OTYPE##_link } )
-#define REG(NAME, MODE, OTYPE)                         \
-       NOD(NAME, (S_IFREG|(MODE)), NULL,               \
-               &proc_##OTYPE##_operations, {})
-#define INF(NAME, MODE, OTYPE)                         \
+               { .proc_get_link = get_link } )
+#define REG(NAME, MODE, fops)                          \
+       NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
+#define INF(NAME, MODE, read)                          \
        NOD(NAME, (S_IFREG|(MODE)),                     \
                NULL, &proc_info_file_operations,       \
-               { .proc_read = &proc_##OTYPE } )
-#define ONE(NAME, MODE, OTYPE)                         \
+               { .proc_read = read } )
+#define ONE(NAME, MODE, show)                          \
        NOD(NAME, (S_IFREG|(MODE)),                     \
                NULL, &proc_single_file_operations,     \
-               { .proc_show = &proc_##OTYPE } )
+               { .proc_show = show } )
 
-int maps_protect;
-EXPORT_SYMBOL(maps_protect);
+static int proc_fd_permission(struct inode *inode, int mask);
 
-static struct fs_struct *get_fs_struct(struct task_struct *task)
+/* ANDROID is for special files in /proc. */
+#define ANDROID(NAME, MODE, OTYPE)                     \
+       NOD(NAME, (S_IFREG|(MODE)),                     \
+               &proc_##OTYPE##_inode_operations,       \
+               &proc_##OTYPE##_operations, {})
+
+/*
+ * Count the number of hardlinks for the pid_entry table, excluding the .
+ * and .. links.
+ */
+static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
+       unsigned int n)
 {
-       struct fs_struct *fs;
-       task_lock(task);
-       fs = task->fs;
-       if(fs)
-               atomic_inc(&fs->count);
-       task_unlock(task);
-       return fs;
+       unsigned int i;
+       unsigned int count;
+
+       count = 0;
+       for (i = 0; i < n; ++i) {
+               if (S_ISDIR(entries[i].mode))
+                       ++count;
+       }
+
+       return count;
 }
 
-static int get_nr_threads(struct task_struct *tsk)
+static int get_task_root(struct task_struct *task, struct path *root)
 {
-       /* Must be called with the rcu_read_lock held */
-       unsigned long flags;
-       int count = 0;
+       int result = -ENOENT;
 
-       if (lock_task_sighand(tsk, &flags)) {
-               count = atomic_read(&tsk->signal->count);
-               unlock_task_sighand(tsk, &flags);
+       task_lock(task);
+       if (task->fs) {
+               get_fs_root(task->fs, root);
+               result = 0;
        }
-       return count;
+       task_unlock(task);
+       return result;
 }
 
-static int proc_cwd_link(struct inode *inode, struct path *path)
+static int proc_cwd_link(struct dentry *dentry, struct path *path)
 {
-       struct task_struct *task = get_proc_task(inode);
-       struct fs_struct *fs = NULL;
+       struct task_struct *task = get_proc_task(dentry->d_inode);
        int result = -ENOENT;
 
        if (task) {
-               fs = get_fs_struct(task);
+               task_lock(task);
+               if (task->fs) {
+                       get_fs_pwd(task->fs, path);
+                       result = 0;
+               }
+               task_unlock(task);
                put_task_struct(task);
        }
-       if (fs) {
-               read_lock(&fs->lock);
-               *path = fs->pwd;
-               path_get(&fs->pwd);
-               read_unlock(&fs->lock);
-               result = 0;
-               put_fs_struct(fs);
-       }
        return result;
 }
 
-static int proc_root_link(struct inode *inode, struct path *path)
+static int proc_root_link(struct dentry *dentry, struct path *path)
 {
-       struct task_struct *task = get_proc_task(inode);
-       struct fs_struct *fs = NULL;
+       struct task_struct *task = get_proc_task(dentry->d_inode);
        int result = -ENOENT;
 
        if (task) {
-               fs = get_fs_struct(task);
+               result = get_task_root(task, path);
                put_task_struct(task);
        }
-       if (fs) {
-               read_lock(&fs->lock);
-               *path = fs->root;
-               path_get(&fs->root);
-               read_unlock(&fs->lock);
-               result = 0;
-               put_fs_struct(fs);
-       }
        return result;
 }
 
-#define MAY_PTRACE(task) \
-       (task == current || \
-       (task->parent == current && \
-       (task->ptrace & PT_PTRACED) && \
-        (task_is_stopped_or_traced(task)) && \
-        security_ptrace(current,task) == 0))
-
 struct mm_struct *mm_for_maps(struct task_struct *task)
 {
-       struct mm_struct *mm = get_task_mm(task);
-       if (!mm)
-               return NULL;
-       down_read(&mm->mmap_sem);
-       task_lock(task);
-       if (task->mm != mm)
-               goto out;
-       if (task->mm != current->mm && __ptrace_may_attach(task) < 0)
-               goto out;
-       task_unlock(task);
-       return mm;
-out:
-       task_unlock(task);
-       up_read(&mm->mmap_sem);
-       mmput(mm);
-       return NULL;
+       return mm_access(task, PTRACE_MODE_READ);
 }
 
-static int proc_pid_cmdline(struct task_struct *task, char * buffer)
+static int proc_pid_cmdline(struct task_struct *task, char *buffer)
 {
        int res = 0;
        unsigned int len;
@@ -261,13 +250,13 @@ out:
 
 static int proc_pid_auxv(struct task_struct *task, char *buffer)
 {
-       int res = 0;
-       struct mm_struct *mm = get_task_mm(task);
-       if (mm) {
+       struct mm_struct *mm = mm_for_maps(task);
+       int res = PTR_ERR(mm);
+       if (mm && !IS_ERR(mm)) {
                unsigned int nwords = 0;
-               do
+               do {
                        nwords += 2;
-               while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
+               } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
                res = nwords * sizeof(mm->saved_auxv[0]);
                if (res > PAGE_SIZE)
                        res = PAGE_SIZE;
@@ -291,12 +280,69 @@ static int proc_pid_wchan(struct task_struct *task, char *buffer)
        wchan = get_wchan(task);
 
        if (lookup_symbol_name(wchan, symname) < 0)
-               return sprintf(buffer, "%lu", wchan);
+               if (!ptrace_may_access(task, PTRACE_MODE_READ))
+                       return 0;
+               else
+                       return sprintf(buffer, "%lu", wchan);
        else
                return sprintf(buffer, "%s", symname);
 }
 #endif /* CONFIG_KALLSYMS */
 
+static int lock_trace(struct task_struct *task)
+{
+       int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
+       if (err)
+               return err;
+       if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
+               mutex_unlock(&task->signal->cred_guard_mutex);
+               return -EPERM;
+       }
+       return 0;
+}
+
+static void unlock_trace(struct task_struct *task)
+{
+       mutex_unlock(&task->signal->cred_guard_mutex);
+}
+
+#ifdef CONFIG_STACKTRACE
+
+#define MAX_STACK_TRACE_DEPTH  64
+
+static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
+                         struct pid *pid, struct task_struct *task)
+{
+       struct stack_trace trace;
+       unsigned long *entries;
+       int err;
+       int i;
+
+       entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
+       if (!entries)
+               return -ENOMEM;
+
+       trace.nr_entries        = 0;
+       trace.max_entries       = MAX_STACK_TRACE_DEPTH;
+       trace.entries           = entries;
+       trace.skip              = 0;
+
+       err = lock_trace(task);
+       if (!err) {
+               save_stack_trace_tsk(task, &trace);
+
+               for (i = 0; i < trace.nr_entries; i++) {
+                       seq_printf(m, "[<%pK>] %pS\n",
+                                  (void *)entries[i], (void *)entries[i]);
+               }
+               unlock_trace(task);
+       }
+       kfree(entries);
+
+       return err;
+}
+#endif
+
 #ifdef CONFIG_SCHEDSTATS
 /*
  * Provides /proc/PID/schedstat
@@ -304,8 +350,8 @@ static int proc_pid_wchan(struct task_struct *task, char *buffer)
 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
 {
        return sprintf(buffer, "%llu %llu %lu\n",
-                       task->sched_info.cpu_time,
-                       task->sched_info.run_delay,
+                       (unsigned long long)task->se.sum_exec_runtime,
+                       (unsigned long long)task->sched_info.run_delay,
                        task->sched_info.pcount);
 }
 #endif
@@ -314,59 +360,48 @@ static int proc_pid_schedstat(struct task_struct *task, char *buffer)
 static int lstats_show_proc(struct seq_file *m, void *v)
 {
        int i;
-       struct task_struct *task = m->private;
-       seq_puts(m, "Latency Top version : v0.1\n");
+       struct inode *inode = m->private;
+       struct task_struct *task = get_proc_task(inode);
 
+       if (!task)
+               return -ESRCH;
+       seq_puts(m, "Latency Top version : v0.1\n");
        for (i = 0; i < 32; i++) {
-               if (task->latency_record[i].backtrace[0]) {
+               struct latency_record *lr = &task->latency_record[i];
+               if (lr->backtrace[0]) {
                        int q;
-                       seq_printf(m, "%i %li %li ",
-                               task->latency_record[i].count,
-                               task->latency_record[i].time,
-                               task->latency_record[i].max);
+                       seq_printf(m, "%i %li %li",
+                                  lr->count, lr->time, lr->max);
                        for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
-                               char sym[KSYM_NAME_LEN];
-                               char *c;
-                               if (!task->latency_record[i].backtrace[q])
+                               unsigned long bt = lr->backtrace[q];
+                               if (!bt)
                                        break;
-                               if (task->latency_record[i].backtrace[q] == ULONG_MAX)
+                               if (bt == ULONG_MAX)
                                        break;
-                               sprint_symbol(sym, task->latency_record[i].backtrace[q]);
-                               c = strchr(sym, '+');
-                               if (c)
-                                       *c = 0;
-                               seq_printf(m, "%s ", sym);
+                               seq_printf(m, " %ps", (void *)bt);
                        }
-                       seq_printf(m, "\n");
+                       seq_putc(m, '\n');
                }
 
        }
+       put_task_struct(task);
        return 0;
 }
 
 static int lstats_open(struct inode *inode, struct file *file)
 {
-       int ret;
-       struct seq_file *m;
-       struct task_struct *task = get_proc_task(inode);
-
-       ret = single_open(file, lstats_show_proc, NULL);
-       if (!ret) {
-               m = file->private_data;
-               m->private = task;
-       }
-       return ret;
+       return single_open(file, lstats_show_proc, inode);
 }
 
 static ssize_t lstats_write(struct file *file, const char __user *buf,
                            size_t count, loff_t *offs)
 {
-       struct seq_file *m;
-       struct task_struct *task;
+       struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
 
-       m = file->private_data;
-       task = m->private;
+       if (!task)
+               return -ESRCH;
        clear_all_latency_tracing(task);
+       put_task_struct(task);
 
        return count;
 }
@@ -381,16 +416,14 @@ static const struct file_operations proc_lstats_operations = {
 
 #endif
 
-/* The badness from the OOM killer */
-unsigned long badness(struct task_struct *p, unsigned long uptime);
 static int proc_oom_score(struct task_struct *task, char *buffer)
 {
-       unsigned long points;
-       struct timespec uptime;
+       unsigned long points = 0;
 
-       do_posix_clock_monotonic_gettime(&uptime);
        read_lock(&tasklist_lock);
-       points = badness(task, uptime.tv_sec);
+       if (pid_alive(task))
+               points = oom_badness(task, NULL, NULL,
+                                       totalram_pages + total_swap_pages);
        read_unlock(&tasklist_lock);
        return sprintf(buffer, "%lu\n", points);
 }
@@ -401,7 +434,7 @@ struct limit_names {
 };
 
 static const struct limit_names lnames[RLIM_NLIMITS] = {
-       [RLIMIT_CPU] = {"Max cpu time", "ms"},
+       [RLIMIT_CPU] = {"Max cpu time", "seconds"},
        [RLIMIT_FSIZE] = {"Max file size", "bytes"},
        [RLIMIT_DATA] = {"Max data size", "bytes"},
        [RLIMIT_STACK] = {"Max stack size", "bytes"},
@@ -416,6 +449,7 @@ static const struct limit_names lnames[RLIM_NLIMITS] = {
        [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
        [RLIMIT_NICE] = {"Max nice priority", NULL},
        [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
+       [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
 };
 
 /* Display limits for a process */
@@ -428,14 +462,10 @@ static int proc_pid_limits(struct task_struct *task, char *buffer)
 
        struct rlimit rlim[RLIM_NLIMITS];
 
-       rcu_read_lock();
-       if (!lock_task_sighand(task,&flags)) {
-               rcu_read_unlock();
+       if (!lock_task_sighand(task, &flags))
                return 0;
-       }
        memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
        unlock_task_sighand(task, &flags);
-       rcu_read_unlock();
 
        /*
         * print the file header
@@ -467,6 +497,30 @@ static int proc_pid_limits(struct task_struct *task, char *buffer)
        return count;
 }
 
+#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
+static int proc_pid_syscall(struct task_struct *task, char *buffer)
+{
+       long nr;
+       unsigned long args[6], sp, pc;
+       int res = lock_trace(task);
+       if (res)
+               return res;
+
+       if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
+               res = sprintf(buffer, "running\n");
+       else if (nr < 0)
+               res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
+       else
+               res = sprintf(buffer,
+                      "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
+                      nr,
+                      args[0], args[1], args[2], args[3], args[4], args[5],
+                      sp, pc);
+       unlock_trace(task);
+       return res;
+}
+#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
+
 /************************************************************************/
 /*                       Here the fs part begins                        */
 /************************************************************************/
@@ -482,13 +536,13 @@ static int proc_fd_access_allowed(struct inode *inode)
         */
        task = get_proc_task(inode);
        if (task) {
-               allowed = ptrace_may_attach(task);
+               allowed = ptrace_may_access(task, PTRACE_MODE_READ);
                put_task_struct(task);
        }
        return allowed;
 }
 
-static int proc_setattr(struct dentry *dentry, struct iattr *attr)
+int proc_setattr(struct dentry *dentry, struct iattr *attr)
 {
        int error;
        struct inode *inode = dentry->d_inode;
@@ -497,133 +551,69 @@ static int proc_setattr(struct dentry *dentry, struct iattr *attr)
                return -EPERM;
 
        error = inode_change_ok(inode, attr);
-       if (!error)
-               error = inode_setattr(inode, attr);
-       return error;
-}
-
-static const struct inode_operations proc_def_inode_operations = {
-       .setattr        = proc_setattr,
-};
-
-extern const struct seq_operations mounts_op;
-struct proc_mounts {
-       struct seq_file m;
-       int event;
-};
-
-static int mounts_open(struct inode *inode, struct file *file)
-{
-       struct task_struct *task = get_proc_task(inode);
-       struct nsproxy *nsp;
-       struct mnt_namespace *ns = NULL;
-       struct proc_mounts *p;
-       int ret = -EINVAL;
-
-       if (task) {
-               rcu_read_lock();
-               nsp = task_nsproxy(task);
-               if (nsp) {
-                       ns = nsp->mnt_ns;
-                       if (ns)
-                               get_mnt_ns(ns);
-               }
-               rcu_read_unlock();
-
-               put_task_struct(task);
-       }
+       if (error)
+               return error;
 
-       if (ns) {
-               ret = -ENOMEM;
-               p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
-               if (p) {
-                       file->private_data = &p->m;
-                       ret = seq_open(file, &mounts_op);
-                       if (!ret) {
-                               p->m.private = ns;
-                               p->event = ns->event;
-                               return 0;
-                       }
-                       kfree(p);
-               }
-               put_mnt_ns(ns);
+       if ((attr->ia_valid & ATTR_SIZE) &&
+           attr->ia_size != i_size_read(inode)) {
+               error = vmtruncate(inode, attr->ia_size);
+               if (error)
+                       return error;
        }
-       return ret;
-}
 
-static int mounts_release(struct inode *inode, struct file *file)
-{
-       struct seq_file *m = file->private_data;
-       struct mnt_namespace *ns = m->private;
-       put_mnt_ns(ns);
-       return seq_release(inode, file);
+       setattr_copy(inode, attr);
+       mark_inode_dirty(inode);
+       return 0;
 }
 
-static unsigned mounts_poll(struct file *file, poll_table *wait)
+/*
+ * May current process learn task's sched/cmdline info (for hide_pid_min=1)
+ * or euid/egid (for hide_pid_min=2)?
+ */
+static bool has_pid_permissions(struct pid_namespace *pid,
+                                struct task_struct *task,
+                                int hide_pid_min)
 {
-       struct proc_mounts *p = file->private_data;
-       struct mnt_namespace *ns = p->m.private;
-       unsigned res = 0;
-
-       poll_wait(file, &ns->poll, wait);
-
-       spin_lock(&vfsmount_lock);
-       if (p->event != ns->event) {
-               p->event = ns->event;
-               res = POLLERR;
-       }
-       spin_unlock(&vfsmount_lock);
-
-       return res;
+       if (pid->hide_pid < hide_pid_min)
+               return true;
+       if (in_group_p(pid->pid_gid))
+               return true;
+       return ptrace_may_access(task, PTRACE_MODE_READ);
 }
 
-static const struct file_operations proc_mounts_operations = {
-       .open           = mounts_open,
-       .read           = seq_read,
-       .llseek         = seq_lseek,
-       .release        = mounts_release,
-       .poll           = mounts_poll,
-};
 
-extern const struct seq_operations mountstats_op;
-static int mountstats_open(struct inode *inode, struct file *file)
+static int proc_pid_permission(struct inode *inode, int mask)
 {
-       int ret = seq_open(file, &mountstats_op);
-
-       if (!ret) {
-               struct seq_file *m = file->private_data;
-               struct nsproxy *nsp;
-               struct mnt_namespace *mnt_ns = NULL;
-               struct task_struct *task = get_proc_task(inode);
+       struct pid_namespace *pid = inode->i_sb->s_fs_info;
+       struct task_struct *task;
+       bool has_perms;
 
-               if (task) {
-                       rcu_read_lock();
-                       nsp = task_nsproxy(task);
-                       if (nsp) {
-                               mnt_ns = nsp->mnt_ns;
-                               if (mnt_ns)
-                                       get_mnt_ns(mnt_ns);
-                       }
-                       rcu_read_unlock();
+       task = get_proc_task(inode);
+       if (!task)
+               return -ESRCH;
+       has_perms = has_pid_permissions(pid, task, 1);
+       put_task_struct(task);
 
-                       put_task_struct(task);
+       if (!has_perms) {
+               if (pid->hide_pid == 2) {
+                       /*
+                        * Let's make getdents(), stat(), and open()
+                        * consistent with each other.  If a process
+                        * may not stat() a file, it shouldn't be seen
+                        * in procfs at all.
+                        */
+                       return -ENOENT;
                }
 
-               if (mnt_ns)
-                       m->private = mnt_ns;
-               else {
-                       seq_release(inode, file);
-                       ret = -EINVAL;
-               }
+               return -EPERM;
        }
-       return ret;
+       return generic_permission(inode, mask);
 }
 
-static const struct file_operations proc_mountstats_operations = {
-       .open           = mountstats_open,
-       .read           = seq_read,
-       .llseek         = seq_lseek,
-       .release        = mounts_release,
+
+
+static const struct inode_operations proc_def_inode_operations = {
+       .setattr        = proc_setattr,
 };
 
 #define PROC_BLOCK_SIZE        (3*1024)                /* 4K page size but our output routines use some slack for overruns */
@@ -660,6 +650,7 @@ out_no_task:
 
 static const struct file_operations proc_info_file_operations = {
        .read           = proc_info_read,
+       .llseek         = generic_file_llseek,
 };
 
 static int proc_single_show(struct seq_file *m, void *v)
@@ -684,14 +675,7 @@ static int proc_single_show(struct seq_file *m, void *v)
 
 static int proc_single_open(struct inode *inode, struct file *filp)
 {
-       int ret;
-       ret = single_open(filp, proc_single_show, NULL);
-       if (!ret) {
-               struct seq_file *m = filp->private_data;
-
-               m->private = inode;
-       }
-       return ret;
+       return single_open(filp, proc_single_show, inode);
 }
 
 static const struct file_operations proc_single_file_operations = {
@@ -703,128 +687,95 @@ static const struct file_operations proc_single_file_operations = {
 
 static int mem_open(struct inode* inode, struct file* file)
 {
-       file->private_data = (void*)((long)current->self_exec_id);
-       return 0;
-}
-
-static ssize_t mem_read(struct file * file, char __user * buf,
-                       size_t count, loff_t *ppos)
-{
        struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
-       char *page;
-       unsigned long src = *ppos;
-       int ret = -ESRCH;
        struct mm_struct *mm;
 
        if (!task)
-               goto out_no_task;
-
-       if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
-               goto out;
-
-       ret = -ENOMEM;
-       page = (char *)__get_free_page(GFP_TEMPORARY);
-       if (!page)
-               goto out;
-
-       ret = 0;
-       mm = get_task_mm(task);
-       if (!mm)
-               goto out_free;
-
-       ret = -EIO;
-       if (file->private_data != (void*)((long)current->self_exec_id))
-               goto out_put;
+               return -ESRCH;
 
-       ret = 0;
-       while (count > 0) {
-               int this_len, retval;
+       mm = mm_access(task, PTRACE_MODE_ATTACH);
+       put_task_struct(task);
 
-               this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
-               retval = access_process_vm(task, src, page, this_len, 0);
-               if (!retval || !MAY_PTRACE(task) || !ptrace_may_attach(task)) {
-                       if (!ret)
-                               ret = -EIO;
-                       break;
-               }
+       if (IS_ERR(mm))
+               return PTR_ERR(mm);
 
-               if (copy_to_user(buf, page, retval)) {
-                       ret = -EFAULT;
-                       break;
-               }
-               ret += retval;
-               src += retval;
-               buf += retval;
-               count -= retval;
+       if (mm) {
+               /* ensure this mm_struct can't be freed */
+               atomic_inc(&mm->mm_count);
+               /* but do not pin its memory */
+               mmput(mm);
        }
-       *ppos = src;
 
-out_put:
-       mmput(mm);
-out_free:
-       free_page((unsigned long) page);
-out:
-       put_task_struct(task);
-out_no_task:
-       return ret;
-}
+       /* OK to pass negative loff_t, we can catch out-of-range */
+       file->f_mode |= FMODE_UNSIGNED_OFFSET;
+       file->private_data = mm;
 
-#define mem_write NULL
+       return 0;
+}
 
-#ifndef mem_write
-/* This is a security hazard */
-static ssize_t mem_write(struct file * file, const char __user *buf,
-                        size_t count, loff_t *ppos)
+static ssize_t mem_rw(struct file *file, char __user *buf,
+                       size_t count, loff_t *ppos, int write)
 {
-       int copied;
+       struct mm_struct *mm = file->private_data;
+       unsigned long addr = *ppos;
+       ssize_t copied;
        char *page;
-       struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
-       unsigned long dst = *ppos;
-
-       copied = -ESRCH;
-       if (!task)
-               goto out_no_task;
 
-       if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
-               goto out;
+       if (!mm)
+               return 0;
 
-       copied = -ENOMEM;
        page = (char *)__get_free_page(GFP_TEMPORARY);
        if (!page)
-               goto out;
+               return -ENOMEM;
 
        copied = 0;
+       if (!atomic_inc_not_zero(&mm->mm_users))
+               goto free;
+
        while (count > 0) {
-               int this_len, retval;
+               int this_len = min_t(int, count, PAGE_SIZE);
 
-               this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
-               if (copy_from_user(page, buf, this_len)) {
+               if (write && copy_from_user(page, buf, this_len)) {
                        copied = -EFAULT;
                        break;
                }
-               retval = access_process_vm(task, dst, page, this_len, 1);
-               if (!retval) {
+
+               this_len = access_remote_vm(mm, addr, page, this_len, write);
+               if (!this_len) {
                        if (!copied)
                                copied = -EIO;
                        break;
                }
-               copied += retval;
-               buf += retval;
-               dst += retval;
-               count -= retval;                        
+
+               if (!write && copy_to_user(buf, page, this_len)) {
+                       copied = -EFAULT;
+                       break;
+               }
+
+               buf += this_len;
+               addr += this_len;
+               copied += this_len;
+               count -= this_len;
        }
-       *ppos = dst;
+       *ppos = addr;
+
+       mmput(mm);
+free:
        free_page((unsigned long) page);
-out:
-       put_task_struct(task);
-out_no_task:
        return copied;
 }
-#endif
+
+static ssize_t mem_read(struct file *file, char __user *buf,
+                       size_t count, loff_t *ppos)
+{
+       return mem_rw(file, buf, count, ppos, 0);
+}
+
+static ssize_t mem_write(struct file *file, const char __user *buf,
+                        size_t count, loff_t *ppos)
+{
+       return mem_rw(file, (char __user*)buf, count, ppos, 1);
+}
 
 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
 {
@@ -842,11 +793,20 @@ loff_t mem_lseek(struct file *file, loff_t offset, int orig)
        return file->f_pos;
 }
 
+static int mem_release(struct inode *inode, struct file *file)
+{
+       struct mm_struct *mm = file->private_data;
+       if (mm)
+               mmdrop(mm);
+       return 0;
+}
+
 static const struct file_operations proc_mem_operations = {
        .llseek         = mem_lseek,
        .read           = mem_read,
        .write          = mem_write,
        .open           = mem_open,
+       .release        = mem_release,
 };
 
 static ssize_t environ_read(struct file *file, char __user *buf,
@@ -861,20 +821,18 @@ static ssize_t environ_read(struct file *file, char __user *buf,
        if (!task)
                goto out_no_task;
 
-       if (!ptrace_may_attach(task))
-               goto out;
-
        ret = -ENOMEM;
        page = (char *)__get_free_page(GFP_TEMPORARY);
        if (!page)
                goto out;
 
-       ret = 0;
 
-       mm = get_task_mm(task);
-       if (!mm)
+       mm = mm_for_maps(task);
+       ret = PTR_ERR(mm);
+       if (!mm || IS_ERR(mm))
                goto out_free;
 
+       ret = 0;
        while (count > 0) {
                int this_len, retval, max_len;
 
@@ -917,6 +875,7 @@ out_no_task:
 
 static const struct file_operations proc_environ_operations = {
        .read           = environ_read,
+       .llseek         = generic_file_llseek,
 };
 
 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
@@ -925,11 +884,17 @@ static ssize_t oom_adjust_read(struct file *file, char __user *buf,
        struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
        char buffer[PROC_NUMBUF];
        size_t len;
-       int oom_adjust;
+       int oom_adjust = OOM_DISABLE;
+       unsigned long flags;
 
        if (!task)
                return -ESRCH;
-       oom_adjust = task->oomkilladj;
+
+       if (lock_task_sighand(task, &flags)) {
+               oom_adjust = task->signal->oom_adj;
+               unlock_task_sighand(task, &flags);
+       }
+
        put_task_struct(task);
 
        len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
@@ -941,37 +906,207 @@ static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
                                size_t count, loff_t *ppos)
 {
        struct task_struct *task;
-       char buffer[PROC_NUMBUF], *end;
+       char buffer[PROC_NUMBUF];
        int oom_adjust;
+       unsigned long flags;
+       int err;
 
        memset(buffer, 0, sizeof(buffer));
        if (count > sizeof(buffer) - 1)
                count = sizeof(buffer) - 1;
-       if (copy_from_user(buffer, buf, count))
-               return -EFAULT;
-       oom_adjust = simple_strtol(buffer, &end, 0);
+       if (copy_from_user(buffer, buf, count)) {
+               err = -EFAULT;
+               goto out;
+       }
+
+       err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
+       if (err)
+               goto out;
        if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
-            oom_adjust != OOM_DISABLE)
-               return -EINVAL;
-       if (*end == '\n')
-               end++;
+            oom_adjust != OOM_DISABLE) {
+               err = -EINVAL;
+               goto out;
+       }
+
        task = get_proc_task(file->f_path.dentry->d_inode);
-       if (!task)
-               return -ESRCH;
-       if (oom_adjust < task->oomkilladj && !capable(CAP_SYS_RESOURCE)) {
-               put_task_struct(task);
-               return -EACCES;
+       if (!task) {
+               err = -ESRCH;
+               goto out;
+       }
+
+       task_lock(task);
+       if (!task->mm) {
+               err = -EINVAL;
+               goto err_task_lock;
        }
-       task->oomkilladj = oom_adjust;
+
+       if (!lock_task_sighand(task, &flags)) {
+               err = -ESRCH;
+               goto err_task_lock;
+       }
+
+       if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
+               err = -EACCES;
+               goto err_sighand;
+       }
+
+       /*
+        * Warn that /proc/pid/oom_adj is deprecated, see
+        * Documentation/feature-removal-schedule.txt.
+        */
+       printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
+                 current->comm, task_pid_nr(current), task_pid_nr(task),
+                 task_pid_nr(task));
+       task->signal->oom_adj = oom_adjust;
+       /*
+        * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
+        * value is always attainable.
+        */
+       if (task->signal->oom_adj == OOM_ADJUST_MAX)
+               task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
+       else
+               task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
+                                                               -OOM_DISABLE;
+       trace_oom_score_adj_update(task);
+err_sighand:
+       unlock_task_sighand(task, &flags);
+err_task_lock:
+       task_unlock(task);
        put_task_struct(task);
-       if (end - buffer == 0)
-               return -EIO;
-       return end - buffer;
+out:
+       return err < 0 ? err : count;
+}
+
+static int oom_adjust_permission(struct inode *inode, int mask)
+{
+       uid_t uid;
+       struct task_struct *p;
+
+       p = get_proc_task(inode);
+       if(p) {
+               uid = task_uid(p);
+               put_task_struct(p);
+       }
+
+       /*
+        * System Server (uid == 1000) is granted access to oom_adj of all 
+        * android applications (uid > 10000) as and services (uid >= 1000)
+        */
+       if (p && (current_fsuid() == 1000) && (uid >= 1000)) {
+               if (inode->i_mode >> 6 & mask) {
+                       return 0;
+               }
+       }
+
+       /* Fall back to default. */
+       return generic_permission(inode, mask);
 }
 
+static const struct inode_operations proc_oom_adjust_inode_operations = {
+       .permission     = oom_adjust_permission,
+};
+
 static const struct file_operations proc_oom_adjust_operations = {
        .read           = oom_adjust_read,
        .write          = oom_adjust_write,
+       .llseek         = generic_file_llseek,
+};
+
+static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
+                                       size_t count, loff_t *ppos)
+{
+       struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
+       char buffer[PROC_NUMBUF];
+       int oom_score_adj = OOM_SCORE_ADJ_MIN;
+       unsigned long flags;
+       size_t len;
+
+       if (!task)
+               return -ESRCH;
+       if (lock_task_sighand(task, &flags)) {
+               oom_score_adj = task->signal->oom_score_adj;
+               unlock_task_sighand(task, &flags);
+       }
+       put_task_struct(task);
+       len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
+       return simple_read_from_buffer(buf, count, ppos, buffer, len);
+}
+
+static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
+                                       size_t count, loff_t *ppos)
+{
+       struct task_struct *task;
+       char buffer[PROC_NUMBUF];
+       unsigned long flags;
+       int oom_score_adj;
+       int err;
+
+       memset(buffer, 0, sizeof(buffer));
+       if (count > sizeof(buffer) - 1)
+               count = sizeof(buffer) - 1;
+       if (copy_from_user(buffer, buf, count)) {
+               err = -EFAULT;
+               goto out;
+       }
+
+       err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
+       if (err)
+               goto out;
+       if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
+                       oom_score_adj > OOM_SCORE_ADJ_MAX) {
+               err = -EINVAL;
+               goto out;
+       }
+
+       task = get_proc_task(file->f_path.dentry->d_inode);
+       if (!task) {
+               err = -ESRCH;
+               goto out;
+       }
+
+       task_lock(task);
+       if (!task->mm) {
+               err = -EINVAL;
+               goto err_task_lock;
+       }
+
+       if (!lock_task_sighand(task, &flags)) {
+               err = -ESRCH;
+               goto err_task_lock;
+       }
+
+       if (oom_score_adj < task->signal->oom_score_adj_min &&
+                       !capable(CAP_SYS_RESOURCE)) {
+               err = -EACCES;
+               goto err_sighand;
+       }
+
+       task->signal->oom_score_adj = oom_score_adj;
+       if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
+               task->signal->oom_score_adj_min = oom_score_adj;
+       trace_oom_score_adj_update(task);
+       /*
+        * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
+        * always attainable.
+        */
+       if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
+               task->signal->oom_adj = OOM_DISABLE;
+       else
+               task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
+                                                       OOM_SCORE_ADJ_MAX;
+err_sighand:
+       unlock_task_sighand(task, &flags);
+err_task_lock:
+       task_unlock(task);
+       put_task_struct(task);
+out:
+       return err < 0 ? err : count;
+}
+
+static const struct file_operations proc_oom_score_adj_operations = {
+       .read           = oom_score_adj_read,
+       .write          = oom_score_adj_write,
+       .llseek         = default_llseek,
 };
 
 #ifdef CONFIG_AUDITSYSCALL
@@ -1000,11 +1135,12 @@ static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
        ssize_t length;
        uid_t loginuid;
 
-       if (!capable(CAP_AUDIT_CONTROL))
-               return -EPERM;
-
-       if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
+       rcu_read_lock();
+       if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
+               rcu_read_unlock();
                return -EPERM;
+       }
+       rcu_read_unlock();
 
        if (count >= PAGE_SIZE)
                count = PAGE_SIZE - 1;
@@ -1027,7 +1163,7 @@ static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
                goto out_free_page;
 
        }
-       length = audit_set_loginuid(current, loginuid);
+       length = audit_set_loginuid(loginuid);
        if (likely(length == 0))
                length = count;
 
@@ -1039,6 +1175,28 @@ out_free_page:
 static const struct file_operations proc_loginuid_operations = {
        .read           = proc_loginuid_read,
        .write          = proc_loginuid_write,
+       .llseek         = generic_file_llseek,
+};
+
+static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
+                                 size_t count, loff_t *ppos)
+{
+       struct inode * inode = file->f_path.dentry->d_inode;
+       struct task_struct *task = get_proc_task(inode);
+       ssize_t length;
+       char tmpbuf[TMPBUFLEN];
+
+       if (!task)
+               return -ESRCH;
+       length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
+                               audit_get_sessionid(task));
+       put_task_struct(task);
+       return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
+}
+
+static const struct file_operations proc_sessionid_operations = {
+       .read           = proc_sessionid_read,
+       .llseek         = generic_file_llseek,
 };
 #endif
 
@@ -1075,22 +1233,22 @@ static ssize_t proc_fault_inject_write(struct file * file,
                count = sizeof(buffer) - 1;
        if (copy_from_user(buffer, buf, count))
                return -EFAULT;
-       make_it_fail = simple_strtol(buffer, &end, 0);
-       if (*end == '\n')
-               end++;
+       make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
+       if (*end)
+               return -EINVAL;
        task = get_proc_task(file->f_dentry->d_inode);
        if (!task)
                return -ESRCH;
        task->make_it_fail = make_it_fail;
        put_task_struct(task);
-       if (end - buffer == 0)
-               return -EIO;
-       return end - buffer;
+
+       return count;
 }
 
 static const struct file_operations proc_fault_inject_operations = {
        .read           = proc_fault_inject_read,
        .write          = proc_fault_inject_write,
+       .llseek         = generic_file_llseek,
 };
 #endif
 
@@ -1104,8 +1262,6 @@ static int sched_show(struct seq_file *m, void *v)
        struct inode *inode = m->private;
        struct task_struct *p;
 
-       WARN_ON(!inode);
-
        p = get_proc_task(inode);
        if (!p)
                return -ESRCH;
@@ -1123,8 +1279,6 @@ sched_write(struct file *file, const char __user *buf,
        struct inode *inode = file->f_path.dentry->d_inode;
        struct task_struct *p;
 
-       WARN_ON(!inode);
-
        p = get_proc_task(inode);
        if (!p)
                return -ESRCH;
@@ -1137,9 +1291,76 @@ sched_write(struct file *file, const char __user *buf,
 
 static int sched_open(struct inode *inode, struct file *filp)
 {
+       return single_open(filp, sched_show, inode);
+}
+
+static const struct file_operations proc_pid_sched_operations = {
+       .open           = sched_open,
+       .read           = seq_read,
+       .write          = sched_write,
+       .llseek         = seq_lseek,
+       .release        = single_release,
+};
+
+#endif
+
+#ifdef CONFIG_SCHED_AUTOGROUP
+/*
+ * Print out autogroup related information:
+ */
+static int sched_autogroup_show(struct seq_file *m, void *v)
+{
+       struct inode *inode = m->private;
+       struct task_struct *p;
+
+       p = get_proc_task(inode);
+       if (!p)
+               return -ESRCH;
+       proc_sched_autogroup_show_task(p, m);
+
+       put_task_struct(p);
+
+       return 0;
+}
+
+static ssize_t
+sched_autogroup_write(struct file *file, const char __user *buf,
+           size_t count, loff_t *offset)
+{
+       struct inode *inode = file->f_path.dentry->d_inode;
+       struct task_struct *p;
+       char buffer[PROC_NUMBUF];
+       int nice;
+       int err;
+
+       memset(buffer, 0, sizeof(buffer));
+       if (count > sizeof(buffer) - 1)
+               count = sizeof(buffer) - 1;
+       if (copy_from_user(buffer, buf, count))
+               return -EFAULT;
+
+       err = kstrtoint(strstrip(buffer), 0, &nice);
+       if (err < 0)
+               return err;
+
+       p = get_proc_task(inode);
+       if (!p)
+               return -ESRCH;
+
+       err = proc_sched_autogroup_set_nice(p, nice);
+       if (err)
+               count = err;
+
+       put_task_struct(p);
+
+       return count;
+}
+
+static int sched_autogroup_open(struct inode *inode, struct file *filp)
+{
        int ret;
 
-       ret = single_open(filp, sched_show, NULL);
+       ret = single_open(filp, sched_autogroup_show, NULL);
        if (!ret) {
                struct seq_file *m = filp->private_data;
 
@@ -1148,15 +1369,97 @@ static int sched_open(struct inode *inode, struct file *filp)
        return ret;
 }
 
-static const struct file_operations proc_pid_sched_operations = {
-       .open           = sched_open,
+static const struct file_operations proc_pid_sched_autogroup_operations = {
+       .open           = sched_autogroup_open,
        .read           = seq_read,
-       .write          = sched_write,
+       .write          = sched_autogroup_write,
        .llseek         = seq_lseek,
        .release        = single_release,
 };
 
-#endif
+#endif /* CONFIG_SCHED_AUTOGROUP */
+
+static ssize_t comm_write(struct file *file, const char __user *buf,
+                               size_t count, loff_t *offset)
+{
+       struct inode *inode = file->f_path.dentry->d_inode;
+       struct task_struct *p;
+       char buffer[TASK_COMM_LEN];
+
+       memset(buffer, 0, sizeof(buffer));
+       if (count > sizeof(buffer) - 1)
+               count = sizeof(buffer) - 1;
+       if (copy_from_user(buffer, buf, count))
+               return -EFAULT;
+
+       p = get_proc_task(inode);
+       if (!p)
+               return -ESRCH;
+
+       if (same_thread_group(current, p))
+               set_task_comm(p, buffer);
+       else
+               count = -EINVAL;
+
+       put_task_struct(p);
+
+       return count;
+}
+
+static int comm_show(struct seq_file *m, void *v)
+{
+       struct inode *inode = m->private;
+       struct task_struct *p;
+
+       p = get_proc_task(inode);
+       if (!p)
+               return -ESRCH;
+
+       task_lock(p);
+       seq_printf(m, "%s\n", p->comm);
+       task_unlock(p);
+
+       put_task_struct(p);
+
+       return 0;
+}
+
+static int comm_open(struct inode *inode, struct file *filp)
+{
+       return single_open(filp, comm_show, inode);
+}
+
+static const struct file_operations proc_pid_set_comm_operations = {
+       .open           = comm_open,
+       .read           = seq_read,
+       .write          = comm_write,
+       .llseek         = seq_lseek,
+       .release        = single_release,
+};
+
+static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
+{
+       struct task_struct *task;
+       struct mm_struct *mm;
+       struct file *exe_file;
+
+       task = get_proc_task(dentry->d_inode);
+       if (!task)
+               return -ENOENT;
+       mm = get_task_mm(task);
+       put_task_struct(task);
+       if (!mm)
+               return -ENOENT;
+       exe_file = get_mm_exe_file(mm);
+       mmput(mm);
+       if (exe_file) {
+               *exe_path = exe_file->f_path;
+               path_get(&exe_file->f_path);
+               fput(exe_file);
+               return 0;
+       } else
+               return -ENOENT;
+}
 
 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
 {
@@ -1170,8 +1473,7 @@ static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
        if (!proc_fd_access_allowed(inode))
                goto out;
 
-       error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
-       nd->last_type = LAST_BIND;
+       error = PROC_I(inode)->op.proc_get_link(dentry, &nd->path);
 out:
        return ERR_PTR(error);
 }
@@ -1185,7 +1487,7 @@ static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
        if (!tmp)
                return -ENOMEM;
 
-       pathname = d_path(path->dentry, path->mnt, tmp, PAGE_SIZE);
+       pathname = d_path(path, tmp, PAGE_SIZE);
        len = PTR_ERR(pathname);
        if (IS_ERR(pathname))
                goto out;
@@ -1210,7 +1512,7 @@ static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int b
        if (!proc_fd_access_allowed(inode))
                goto out;
 
-       error = PROC_I(inode)->op.proc_get_link(inode, &path);
+       error = PROC_I(inode)->op.proc_get_link(dentry, &path);
        if (error)
                goto out;
 
@@ -1244,11 +1546,11 @@ static int task_dumpable(struct task_struct *task)
        return 0;
 }
 
-
-static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
+struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
 {
        struct inode * inode;
        struct proc_inode *ei;
+       const struct cred *cred;
 
        /* We need a new inode */
 
@@ -1258,6 +1560,7 @@ static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_st
 
        /* Common stuff */
        ei = PROC_I(inode);
+       inode->i_ino = get_next_ino();
        inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
        inode->i_op = &proc_def_inode_operations;
 
@@ -1268,11 +1571,12 @@ static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_st
        if (!ei->pid)
                goto out_unlock;
 
-       inode->i_uid = 0;
-       inode->i_gid = 0;
        if (task_dumpable(task)) {
-               inode->i_uid = task->euid;
-               inode->i_gid = task->egid;
+               rcu_read_lock();
+               cred = __task_cred(task);
+               inode->i_uid = cred->euid;
+               inode->i_gid = cred->egid;
+               rcu_read_unlock();
        }
        security_task_to_inode(task, inode);
 
@@ -1284,10 +1588,13 @@ out_unlock:
        return NULL;
 }
 
-static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
+int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
 {
        struct inode *inode = dentry->d_inode;
        struct task_struct *task;
+       const struct cred *cred;
+       struct pid_namespace *pid = dentry->d_sb->s_fs_info;
+
        generic_fillattr(inode, stat);
 
        rcu_read_lock();
@@ -1295,10 +1602,19 @@ static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat
        stat->gid = 0;
        task = pid_task(proc_pid(inode), PIDTYPE_PID);
        if (task) {
+               if (!has_pid_permissions(pid, task, 2)) {
+                       rcu_read_unlock();
+                       /*
+                        * This doesn't prevent learning whether PID exists,
+                        * it only makes getattr() consistent with readdir().
+                        */
+                       return -ENOENT;
+               }
                if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
                    task_dumpable(task)) {
-                       stat->uid = task->euid;
-                       stat->gid = task->egid;
+                       cred = __task_cred(task);
+                       stat->uid = cred->euid;
+                       stat->gid = cred->egid;
                }
        }
        rcu_read_unlock();
@@ -1322,15 +1638,26 @@ static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat
  * made this apply to all per process world readable and executable
  * directories.
  */
-static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
+int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
 {
-       struct inode *inode = dentry->d_inode;
-       struct task_struct *task = get_proc_task(inode);
+       struct inode *inode;
+       struct task_struct *task;
+       const struct cred *cred;
+
+       if (nd && nd->flags & LOOKUP_RCU)
+               return -ECHILD;
+
+       inode = dentry->d_inode;
+       task = get_proc_task(inode);
+
        if (task) {
                if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
                    task_dumpable(task)) {
-                       inode->i_uid = task->euid;
-                       inode->i_gid = task->egid;
+                       rcu_read_lock();
+                       cred = __task_cred(task);
+                       inode->i_uid = cred->euid;
+                       inode->i_gid = cred->egid;
+                       rcu_read_unlock();
                } else {
                        inode->i_uid = 0;
                        inode->i_gid = 0;
@@ -1344,7 +1671,7 @@ static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
        return 0;
 }
 
-static int pid_delete_dentry(struct dentry * dentry)
+static int pid_delete_dentry(const struct dentry * dentry)
 {
        /* Is the task we represent dead?
         * If so, then don't put the dentry on the lru list,
@@ -1353,7 +1680,7 @@ static int pid_delete_dentry(struct dentry * dentry)
        return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
 }
 
-static struct dentry_operations pid_dentry_operations =
+const struct dentry_operations pid_dentry_operations =
 {
        .d_revalidate   = pid_revalidate,
        .d_delete       = pid_delete_dentry,
@@ -1361,9 +1688,6 @@ static struct dentry_operations pid_dentry_operations =
 
 /* Lookups */
 
-typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
-                               struct task_struct *, const void *);
-
 /*
  * Fill a directory entry.
  *
@@ -1376,8 +1700,8 @@ typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
  * reported by readdir in sync with the inode numbers reported
  * by stat.
  */
-static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
-       char *name, int len,
+int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
+       const char *name, int len,
        instantiate_t instantiate, struct task_struct *task, const void *ptr)
 {
        struct dentry *child, *dir = filp->f_path.dentry;
@@ -1461,6 +1785,14 @@ static int proc_fd_info(struct inode *inode, struct path *path, char *info)
                spin_lock(&files->file_lock);
                file = fcheck_files(files, fd);
                if (file) {
+                       unsigned int f_flags;
+                       struct fdtable *fdt;
+
+                       fdt = files_fdtable(files);
+                       f_flags = file->f_flags & ~O_CLOEXEC;
+                       if (close_on_exec(fd, fdt))
+                               f_flags |= O_CLOEXEC;
+
                        if (path) {
                                *path = file->f_path;
                                path_get(&file->f_path);
@@ -1470,7 +1802,7 @@ static int proc_fd_info(struct inode *inode, struct path *path, char *info)
                                         "pos:\t%lli\n"
                                         "flags:\t0%o\n",
                                         (long long) file->f_pos,
-                                        file->f_flags);
+                                        f_flags);
                        spin_unlock(&files->file_lock);
                        put_files_struct(files);
                        return 0;
@@ -1481,33 +1813,58 @@ static int proc_fd_info(struct inode *inode, struct path *path, char *info)
        return -ENOENT;
 }
 
-static int proc_fd_link(struct inode *inode, struct path *path)
+static int proc_fd_link(struct dentry *dentry, struct path *path)
 {
-       return proc_fd_info(inode, path, NULL);
+       return proc_fd_info(dentry->d_inode, path, NULL);
 }
 
 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
 {
-       struct inode *inode = dentry->d_inode;
-       struct task_struct *task = get_proc_task(inode);
-       int fd = proc_fd(inode);
+       struct inode *inode;
+       struct task_struct *task;
+       int fd;
        struct files_struct *files;
+       const struct cred *cred;
+
+       if (nd && nd->flags & LOOKUP_RCU)
+               return -ECHILD;
+
+       inode = dentry->d_inode;
+       task = get_proc_task(inode);
+       fd = proc_fd(inode);
 
        if (task) {
                files = get_files_struct(task);
                if (files) {
+                       struct file *file;
                        rcu_read_lock();
-                       if (fcheck_files(files, fd)) {
+                       file = fcheck_files(files, fd);
+                       if (file) {
+                               unsigned f_mode = file->f_mode;
+
                                rcu_read_unlock();
                                put_files_struct(files);
+
                                if (task_dumpable(task)) {
-                                       inode->i_uid = task->euid;
-                                       inode->i_gid = task->egid;
+                                       rcu_read_lock();
+                                       cred = __task_cred(task);
+                                       inode->i_uid = cred->euid;
+                                       inode->i_gid = cred->egid;
+                                       rcu_read_unlock();
                                } else {
                                        inode->i_uid = 0;
                                        inode->i_gid = 0;
                                }
-                               inode->i_mode &= ~(S_ISUID | S_ISGID);
+
+                               if (S_ISLNK(inode->i_mode)) {
+                                       unsigned i_mode = S_IFLNK;
+                                       if (f_mode & FMODE_READ)
+                                               i_mode |= S_IRUSR | S_IXUSR;
+                                       if (f_mode & FMODE_WRITE)
+                                               i_mode |= S_IWUSR | S_IXUSR;
+                                       inode->i_mode = i_mode;
+                               }
+
                                security_task_to_inode(task, inode);
                                put_task_struct(task);
                                return 1;
@@ -1521,7 +1878,7 @@ static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
        return 0;
 }
 
-static struct dentry_operations tid_fd_dentry_operations =
+static const struct dentry_operations tid_fd_dentry_operations =
 {
        .d_revalidate   = tid_fd_revalidate,
        .d_delete       = pid_delete_dentry,
@@ -1531,8 +1888,6 @@ static struct dentry *proc_fd_instantiate(struct inode *dir,
        struct dentry *dentry, struct task_struct *task, const void *ptr)
 {
        unsigned fd = *(const unsigned *)ptr;
-       struct file *file;
-       struct files_struct *files;
        struct inode *inode;
        struct proc_inode *ei;
        struct dentry *error = ERR_PTR(-ENOENT);
@@ -1542,30 +1897,12 @@ static struct dentry *proc_fd_instantiate(struct inode *dir,
                goto out;
        ei = PROC_I(inode);
        ei->fd = fd;
-       files = get_files_struct(task);
-       if (!files)
-               goto out_iput;
-       inode->i_mode = S_IFLNK;
-
-       /*
-        * We are not taking a ref to the file structure, so we must
-        * hold ->file_lock.
-        */
-       spin_lock(&files->file_lock);
-       file = fcheck_files(files, fd);
-       if (!file)
-               goto out_unlock;
-       if (file->f_mode & 1)
-               inode->i_mode |= S_IRUSR | S_IXUSR;
-       if (file->f_mode & 2)
-               inode->i_mode |= S_IWUSR | S_IXUSR;
-       spin_unlock(&files->file_lock);
-       put_files_struct(files);
 
+       inode->i_mode = S_IFLNK;
        inode->i_op = &proc_pid_link_inode_operations;
        inode->i_size = 64;
        ei->op.proc_get_link = proc_fd_link;
-       dentry->d_op = &tid_fd_dentry_operations;
+       d_set_d_op(dentry, &tid_fd_dentry_operations);
        d_add(dentry, inode);
        /* Close the race of the process dying before we return the dentry */
        if (tid_fd_revalidate(dentry, NULL))
@@ -1573,12 +1910,6 @@ static struct dentry *proc_fd_instantiate(struct inode *dir,
 
  out:
        return error;
-out_unlock:
-       spin_unlock(&files->file_lock);
-       put_files_struct(files);
-out_iput:
-       iput(inode);
-       goto out;
 }
 
 static struct dentry *proc_lookupfd_common(struct inode *dir,
@@ -1610,7 +1941,6 @@ static int proc_readfd_common(struct file * filp, void * dirent,
        unsigned int fd, ino;
        int retval;
        struct files_struct * files;
-       struct fdtable *fdt;
 
        retval = -ENOENT;
        if (!p)
@@ -1633,9 +1963,8 @@ static int proc_readfd_common(struct file * filp, void * dirent,
                        if (!files)
                                goto out;
                        rcu_read_lock();
-                       fdt = files_fdtable(files);
                        for (fd = filp->f_pos-2;
-                            fd < fdt->max_fds;
+                            fd < files_fdtable(files)->max_fds;
                             fd++, filp->f_pos++) {
                                char name[PROC_NUMBUF];
                                int len;
@@ -1684,25 +2013,369 @@ static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
 }
 
 static const struct file_operations proc_fdinfo_file_operations = {
-       .open           = nonseekable_open,
+       .open           = nonseekable_open,
        .read           = proc_fdinfo_read,
+       .llseek         = no_llseek,
 };
 
 static const struct file_operations proc_fd_operations = {
        .read           = generic_read_dir,
        .readdir        = proc_readfd,
+       .llseek         = default_llseek,
 };
 
+#ifdef CONFIG_CHECKPOINT_RESTORE
+
+/*
+ * dname_to_vma_addr - maps a dentry name into two unsigned longs
+ * which represent vma start and end addresses.
+ */
+static int dname_to_vma_addr(struct dentry *dentry,
+                            unsigned long *start, unsigned long *end)
+{
+       if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
+               return -EINVAL;
+
+       return 0;
+}
+
+static int map_files_d_revalidate(struct dentry *dentry, struct nameidata *nd)
+{
+       unsigned long vm_start, vm_end;
+       bool exact_vma_exists = false;
+       struct mm_struct *mm = NULL;
+       struct task_struct *task;
+       const struct cred *cred;
+       struct inode *inode;
+       int status = 0;
+
+       if (nd && nd->flags & LOOKUP_RCU)
+               return -ECHILD;
+
+       if (!capable(CAP_SYS_ADMIN)) {
+               status = -EACCES;
+               goto out_notask;
+       }
+
+       inode = dentry->d_inode;
+       task = get_proc_task(inode);
+       if (!task)
+               goto out_notask;
+
+       if (!ptrace_may_access(task, PTRACE_MODE_READ))
+               goto out;
+
+       mm = get_task_mm(task);
+       if (!mm)
+               goto out;
+
+       if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
+               down_read(&mm->mmap_sem);
+               exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
+               up_read(&mm->mmap_sem);
+       }
+
+       mmput(mm);
+
+       if (exact_vma_exists) {
+               if (task_dumpable(task)) {
+                       rcu_read_lock();
+                       cred = __task_cred(task);
+                       inode->i_uid = cred->euid;
+                       inode->i_gid = cred->egid;
+                       rcu_read_unlock();
+               } else {
+                       inode->i_uid = 0;
+                       inode->i_gid = 0;
+               }
+               security_task_to_inode(task, inode);
+               status = 1;
+       }
+
+out:
+       put_task_struct(task);
+
+out_notask:
+       if (status <= 0)
+               d_drop(dentry);
+
+       return status;
+}
+
+static const struct dentry_operations tid_map_files_dentry_operations = {
+       .d_revalidate   = map_files_d_revalidate,
+       .d_delete       = pid_delete_dentry,
+};
+
+static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
+{
+       unsigned long vm_start, vm_end;
+       struct vm_area_struct *vma;
+       struct task_struct *task;
+       struct mm_struct *mm;
+       int rc;
+
+       rc = -ENOENT;
+       task = get_proc_task(dentry->d_inode);
+       if (!task)
+               goto out;
+
+       mm = get_task_mm(task);
+       put_task_struct(task);
+       if (!mm)
+               goto out;
+
+       rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
+       if (rc)
+               goto out_mmput;
+
+       down_read(&mm->mmap_sem);
+       vma = find_exact_vma(mm, vm_start, vm_end);
+       if (vma && vma->vm_file) {
+               *path = vma->vm_file->f_path;
+               path_get(path);
+               rc = 0;
+       }
+       up_read(&mm->mmap_sem);
+
+out_mmput:
+       mmput(mm);
+out:
+       return rc;
+}
+
+struct map_files_info {
+       struct file     *file;
+       unsigned long   len;
+       unsigned char   name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
+};
+
+static struct dentry *
+proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
+                          struct task_struct *task, const void *ptr)
+{
+       const struct file *file = ptr;
+       struct proc_inode *ei;
+       struct inode *inode;
+
+       if (!file)
+               return ERR_PTR(-ENOENT);
+
+       inode = proc_pid_make_inode(dir->i_sb, task);
+       if (!inode)
+               return ERR_PTR(-ENOENT);
+
+       ei = PROC_I(inode);
+       ei->op.proc_get_link = proc_map_files_get_link;
+
+       inode->i_op = &proc_pid_link_inode_operations;
+       inode->i_size = 64;
+       inode->i_mode = S_IFLNK;
+
+       if (file->f_mode & FMODE_READ)
+               inode->i_mode |= S_IRUSR;
+       if (file->f_mode & FMODE_WRITE)
+               inode->i_mode |= S_IWUSR;
+
+       d_set_d_op(dentry, &tid_map_files_dentry_operations);
+       d_add(dentry, inode);
+
+       return NULL;
+}
+
+static struct dentry *proc_map_files_lookup(struct inode *dir,
+               struct dentry *dentry, struct nameidata *nd)
+{
+       unsigned long vm_start, vm_end;
+       struct vm_area_struct *vma;
+       struct task_struct *task;
+       struct dentry *result;
+       struct mm_struct *mm;
+
+       result = ERR_PTR(-EACCES);
+       if (!capable(CAP_SYS_ADMIN))
+               goto out;
+
+       result = ERR_PTR(-ENOENT);
+       task = get_proc_task(dir);
+       if (!task)
+               goto out;
+
+       result = ERR_PTR(-EACCES);
+       if (!ptrace_may_access(task, PTRACE_MODE_READ))
+               goto out_put_task;
+
+       result = ERR_PTR(-ENOENT);
+       if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
+               goto out_put_task;
+
+       mm = get_task_mm(task);
+       if (!mm)
+               goto out_put_task;
+
+       down_read(&mm->mmap_sem);
+       vma = find_exact_vma(mm, vm_start, vm_end);
+       if (!vma)
+               goto out_no_vma;
+
+       result = proc_map_files_instantiate(dir, dentry, task, vma->vm_file);
+
+out_no_vma:
+       up_read(&mm->mmap_sem);
+       mmput(mm);
+out_put_task:
+       put_task_struct(task);
+out:
+       return result;
+}
+
+static const struct inode_operations proc_map_files_inode_operations = {
+       .lookup         = proc_map_files_lookup,
+       .permission     = proc_fd_permission,
+       .setattr        = proc_setattr,
+};
+
+static int
+proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
+{
+       struct dentry *dentry = filp->f_path.dentry;
+       struct inode *inode = dentry->d_inode;
+       struct vm_area_struct *vma;
+       struct task_struct *task;
+       struct mm_struct *mm;
+       ino_t ino;
+       int ret;
+
+       ret = -EACCES;
+       if (!capable(CAP_SYS_ADMIN))
+               goto out;
+
+       ret = -ENOENT;
+       task = get_proc_task(inode);
+       if (!task)
+               goto out;
+
+       ret = -EACCES;
+       if (!ptrace_may_access(task, PTRACE_MODE_READ))
+               goto out_put_task;
+
+       ret = 0;
+       switch (filp->f_pos) {
+       case 0:
+               ino = inode->i_ino;
+               if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
+                       goto out_put_task;
+               filp->f_pos++;
+       case 1:
+               ino = parent_ino(dentry);
+               if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
+                       goto out_put_task;
+               filp->f_pos++;
+       default:
+       {
+               unsigned long nr_files, pos, i;
+               struct flex_array *fa = NULL;
+               struct map_files_info info;
+               struct map_files_info *p;
+
+               mm = get_task_mm(task);
+               if (!mm)
+                       goto out_put_task;
+               down_read(&mm->mmap_sem);
+
+               nr_files = 0;
+
+               /*
+                * We need two passes here:
+                *
+                *  1) Collect vmas of mapped files with mmap_sem taken
+                *  2) Release mmap_sem and instantiate entries
+                *
+                * otherwise we get lockdep complained, since filldir()
+                * routine might require mmap_sem taken in might_fault().
+                */
+
+               for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
+                       if (vma->vm_file && ++pos > filp->f_pos)
+                               nr_files++;
+               }
+
+               if (nr_files) {
+                       fa = flex_array_alloc(sizeof(info), nr_files,
+                                               GFP_KERNEL);
+                       if (!fa || flex_array_prealloc(fa, 0, nr_files,
+                                                       GFP_KERNEL)) {
+                               ret = -ENOMEM;
+                               if (fa)
+                                       flex_array_free(fa);
+                               up_read(&mm->mmap_sem);
+                               mmput(mm);
+                               goto out_put_task;
+                       }
+                       for (i = 0, vma = mm->mmap, pos = 2; vma;
+                                       vma = vma->vm_next) {
+                               if (!vma->vm_file)
+                                       continue;
+                               if (++pos <= filp->f_pos)
+                                       continue;
+
+                               get_file(vma->vm_file);
+                               info.file = vma->vm_file;
+                               info.len = snprintf(info.name,
+                                               sizeof(info.name), "%lx-%lx",
+                                               vma->vm_start, vma->vm_end);
+                               if (flex_array_put(fa, i++, &info, GFP_KERNEL))
+                                       BUG();
+                       }
+               }
+               up_read(&mm->mmap_sem);
+
+               for (i = 0; i < nr_files; i++) {
+                       p = flex_array_get(fa, i);
+                       ret = proc_fill_cache(filp, dirent, filldir,
+                                             p->name, p->len,
+                                             proc_map_files_instantiate,
+                                             task, p->file);
+                       if (ret)
+                               break;
+                       filp->f_pos++;
+                       fput(p->file);
+               }
+               for (; i < nr_files; i++) {
+                       /*
+                        * In case of error don't forget
+                        * to put rest of file refs.
+                        */
+                       p = flex_array_get(fa, i);
+                       fput(p->file);
+               }
+               if (fa)
+                       flex_array_free(fa);
+               mmput(mm);
+       }
+       }
+
+out_put_task:
+       put_task_struct(task);
+out:
+       return ret;
+}
+
+static const struct file_operations proc_map_files_operations = {
+       .read           = generic_read_dir,
+       .readdir        = proc_map_files_readdir,
+       .llseek         = default_llseek,
+};
+
+#endif /* CONFIG_CHECKPOINT_RESTORE */
+
 /*
  * /proc/pid/fd needs a special permission handler so that a process can still
  * access /proc/self/fd after it has executed a setuid().
  */
-static int proc_fd_permission(struct inode *inode, int mask,
-                               struct nameidata *nd)
+static int proc_fd_permission(struct inode *inode, int mask)
 {
-       int rv;
-
-       rv = generic_permission(inode, mask, NULL);
+       int rv = generic_permission(inode, mask);
        if (rv == 0)
                return 0;
        if (task_pid(current) == proc_pid(inode))
@@ -1734,7 +2407,7 @@ static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
        ei->fd = fd;
        inode->i_mode = S_IFREG | S_IRUSR;
        inode->i_fop = &proc_fdinfo_file_operations;
-       dentry->d_op = &tid_fd_dentry_operations;
+       d_set_d_op(dentry, &tid_fd_dentry_operations);
        d_add(dentry, inode);
        /* Close the race of the process dying before we return the dentry */
        if (tid_fd_revalidate(dentry, NULL))
@@ -1760,6 +2433,7 @@ static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
 static const struct file_operations proc_fdinfo_operations = {
        .read           = generic_read_dir,
        .readdir        = proc_readfdinfo,
+       .llseek         = default_llseek,
 };
 
 /*
@@ -1777,7 +2451,7 @@ static struct dentry *proc_pident_instantiate(struct inode *dir,
        const struct pid_entry *p = ptr;
        struct inode *inode;
        struct proc_inode *ei;
-       struct dentry *error = ERR_PTR(-EINVAL);
+       struct dentry *error = ERR_PTR(-ENOENT);
 
        inode = proc_pid_make_inode(dir->i_sb, task);
        if (!inode)
@@ -1786,13 +2460,13 @@ static struct dentry *proc_pident_instantiate(struct inode *dir,
        ei = PROC_I(inode);
        inode->i_mode = p->mode;
        if (S_ISDIR(inode->i_mode))
-               inode->i_nlink = 2;     /* Use getattr to fix if necessary */
+               set_nlink(inode, 2);    /* Use getattr to fix if necessary */
        if (p->iop)
                inode->i_op = p->iop;
        if (p->fop)
                inode->i_fop = p->fop;
        ei->op = p->op;
-       dentry->d_op = &pid_dentry_operations;
+       d_set_d_op(dentry, &pid_dentry_operations);
        d_add(dentry, inode);
        /* Close the race of the process dying before we return the dentry */
        if (pid_revalidate(dentry, NULL))
@@ -1806,13 +2480,11 @@ static struct dentry *proc_pident_lookup(struct inode *dir,
                                         const struct pid_entry *ents,
                                         unsigned int nents)
 {
-       struct inode *inode;
        struct dentry *error;
        struct task_struct *task = get_proc_task(dir);
        const struct pid_entry *p, *last;
 
        error = ERR_PTR(-ENOENT);
-       inode = NULL;
 
        if (!task)
                goto out_no_task;
@@ -1951,9 +2623,15 @@ static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
        if (copy_from_user(page, buf, count))
                goto out_free;
 
+       /* Guard against adverse ptrace interaction */
+       length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
+       if (length < 0)
+               goto out_free;
+
        length = security_setprocattr(task,
                                      (char*)file->f_path.dentry->d_name.name,
                                      (void*)page, count);
+       mutex_unlock(&task->signal->cred_guard_mutex);
 out_free:
        free_page((unsigned long) page);
 out:
@@ -1965,15 +2643,16 @@ out_no_task:
 static const struct file_operations proc_pid_attr_operations = {
        .read           = proc_pid_attr_read,
        .write          = proc_pid_attr_write,
+       .llseek         = generic_file_llseek,
 };
 
 static const struct pid_entry attr_dir_stuff[] = {
-       REG("current",    S_IRUGO|S_IWUGO, pid_attr),
-       REG("prev",       S_IRUGO,         pid_attr),
-       REG("exec",       S_IRUGO|S_IWUGO, pid_attr),
-       REG("fscreate",   S_IRUGO|S_IWUGO, pid_attr),
-       REG("keycreate",  S_IRUGO|S_IWUGO, pid_attr),
-       REG("sockcreate", S_IRUGO|S_IWUGO, pid_attr),
+       REG("current",    S_IRUGO|S_IWUGO, proc_pid_attr_operations),
+       REG("prev",       S_IRUGO,         proc_pid_attr_operations),
+       REG("exec",       S_IRUGO|S_IWUGO, proc_pid_attr_operations),
+       REG("fscreate",   S_IRUGO|S_IWUGO, proc_pid_attr_operations),
+       REG("keycreate",  S_IRUGO|S_IWUGO, proc_pid_attr_operations),
+       REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
 };
 
 static int proc_attr_dir_readdir(struct file * filp,
@@ -1986,6 +2665,7 @@ static int proc_attr_dir_readdir(struct file * filp,
 static const struct file_operations proc_attr_dir_operations = {
        .read           = generic_read_dir,
        .readdir        = proc_attr_dir_readdir,
+       .llseek         = default_llseek,
 };
 
 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
@@ -2003,7 +2683,7 @@ static const struct inode_operations proc_attr_dir_inode_operations = {
 
 #endif
 
-#if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
+#ifdef CONFIG_ELF_CORE
 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
                                         size_t count, loff_t *ppos)
 {
@@ -2085,6 +2765,7 @@ static ssize_t proc_coredump_filter_write(struct file *file,
 static const struct file_operations proc_coredump_filter_operations = {
        .read           = proc_coredump_filter_read,
        .write          = proc_coredump_filter_write,
+       .llseek         = generic_file_llseek,
 };
 #endif
 
@@ -2107,16 +2788,30 @@ static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
 {
        struct pid_namespace *ns = dentry->d_sb->s_fs_info;
        pid_t tgid = task_tgid_nr_ns(current, ns);
-       char tmp[PROC_NUMBUF];
-       if (!tgid)
-               return ERR_PTR(-ENOENT);
-       sprintf(tmp, "%d", task_tgid_nr_ns(current, ns));
-       return ERR_PTR(vfs_follow_link(nd,tmp));
+       char *name = ERR_PTR(-ENOENT);
+       if (tgid) {
+               name = __getname();
+               if (!name)
+                       name = ERR_PTR(-ENOMEM);
+               else
+                       sprintf(name, "%d", tgid);
+       }
+       nd_set_link(nd, name);
+       return NULL;
+}
+
+static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
+                               void *cookie)
+{
+       char *s = nd_get_link(nd);
+       if (!IS_ERR(s))
+               __putname(s);
 }
 
 static const struct inode_operations proc_self_inode_operations = {
        .readlink       = proc_self_readlink,
        .follow_link    = proc_self_follow_link,
+       .put_link       = proc_self_put_link,
 };
 
 /*
@@ -2131,36 +2826,13 @@ static const struct pid_entry proc_base_stuff[] = {
                &proc_self_inode_operations, NULL, {}),
 };
 
-/*
- *     Exceptional case: normally we are not allowed to unhash a busy
- * directory. In this case, however, we can do it - no aliasing problems
- * due to the way we treat inodes.
- */
-static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
-{
-       struct inode *inode = dentry->d_inode;
-       struct task_struct *task = get_proc_task(inode);
-       if (task) {
-               put_task_struct(task);
-               return 1;
-       }
-       d_drop(dentry);
-       return 0;
-}
-
-static struct dentry_operations proc_base_dentry_operations =
-{
-       .d_revalidate   = proc_base_revalidate,
-       .d_delete       = pid_delete_dentry,
-};
-
 static struct dentry *proc_base_instantiate(struct inode *dir,
        struct dentry *dentry, struct task_struct *task, const void *ptr)
 {
        const struct pid_entry *p = ptr;
        struct inode *inode;
        struct proc_inode *ei;
-       struct dentry *error = ERR_PTR(-EINVAL);
+       struct dentry *error;
 
        /* Allocate the inode */
        error = ERR_PTR(-ENOMEM);
@@ -2170,6 +2842,7 @@ static struct dentry *proc_base_instantiate(struct inode *dir,
 
        /* Initialize the inode */
        ei = PROC_I(inode);
+       inode->i_ino = get_next_ino();
        inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
 
        /*
@@ -2179,11 +2852,9 @@ static struct dentry *proc_base_instantiate(struct inode *dir,
        if (!ei->pid)
                goto out_iput;
 
-       inode->i_uid = 0;
-       inode->i_gid = 0;
        inode->i_mode = p->mode;
        if (S_ISDIR(inode->i_mode))
-               inode->i_nlink = 2;
+               set_nlink(inode, 2);
        if (S_ISLNK(inode->i_mode))
                inode->i_size = 64;
        if (p->iop)
@@ -2191,7 +2862,6 @@ static struct dentry *proc_base_instantiate(struct inode *dir,
        if (p->fop)
                inode->i_fop = p->fop;
        ei->op = p->op;
-       dentry->d_op = &proc_base_dentry_operations;
        d_add(dentry, inode);
        error = NULL;
 out:
@@ -2239,29 +2909,71 @@ static int proc_base_fill_cache(struct file *filp, void *dirent,
 }
 
 #ifdef CONFIG_TASK_IO_ACCOUNTING
-static int proc_pid_io_accounting(struct task_struct *task, char *buffer)
+static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
 {
-       return sprintf(buffer,
-#ifdef CONFIG_TASK_XACCT
+       struct task_io_accounting acct = task->ioac;
+       unsigned long flags;
+       int result;
+
+       result = mutex_lock_killable(&task->signal->cred_guard_mutex);
+       if (result)
+               return result;
+
+       if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
+               result = -EACCES;
+               goto out_unlock;
+       }
+
+       if (whole && lock_task_sighand(task, &flags)) {
+               struct task_struct *t = task;
+
+               task_io_accounting_add(&acct, &task->signal->ioac);
+               while_each_thread(task, t)
+                       task_io_accounting_add(&acct, &t->ioac);
+
+               unlock_task_sighand(task, &flags);
+       }
+       result = sprintf(buffer,
                        "rchar: %llu\n"
                        "wchar: %llu\n"
                        "syscr: %llu\n"
                        "syscw: %llu\n"
-#endif
                        "read_bytes: %llu\n"
                        "write_bytes: %llu\n"
                        "cancelled_write_bytes: %llu\n",
-#ifdef CONFIG_TASK_XACCT
-                       (unsigned long long)task->rchar,
-                       (unsigned long long)task->wchar,
-                       (unsigned long long)task->syscr,
-                       (unsigned long long)task->syscw,
-#endif
-                       (unsigned long long)task->ioac.read_bytes,
-                       (unsigned long long)task->ioac.write_bytes,
-                       (unsigned long long)task->ioac.cancelled_write_bytes);
+                       (unsigned long long)acct.rchar,
+                       (unsigned long long)acct.wchar,
+                       (unsigned long long)acct.syscr,
+                       (unsigned long long)acct.syscw,
+                       (unsigned long long)acct.read_bytes,
+                       (unsigned long long)acct.write_bytes,
+                       (unsigned long long)acct.cancelled_write_bytes);
+out_unlock:
+       mutex_unlock(&task->signal->cred_guard_mutex);
+       return result;
+}
+
+static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
+{
+       return do_io_accounting(task, buffer, 0);
+}
+
+static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
+{
+       return do_io_accounting(task, buffer, 1);
+}
+#endif /* CONFIG_TASK_IO_ACCOUNTING */
+
+static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
+                               struct pid *pid, struct task_struct *task)
+{
+       int err = lock_trace(task);
+       if (!err) {
+               seq_printf(m, "%08x\n", task->personality);
+               unlock_trace(task);
+       }
+       return err;
 }
-#endif
 
 /*
  * Thread groups
@@ -2270,65 +2982,89 @@ static const struct file_operations proc_task_operations;
 static const struct inode_operations proc_task_inode_operations;
 
 static const struct pid_entry tgid_base_stuff[] = {
-       DIR("task",       S_IRUGO|S_IXUGO, task),
-       DIR("fd",         S_IRUSR|S_IXUSR, fd),
-       DIR("fdinfo",     S_IRUSR|S_IXUSR, fdinfo),
-       REG("environ",    S_IRUSR, environ),
-       INF("auxv",       S_IRUSR, pid_auxv),
-       ONE("status",     S_IRUGO, pid_status),
-       INF("limits",     S_IRUSR, pid_limits),
+       DIR("task",       S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
+       DIR("fd",         S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
+#ifdef CONFIG_CHECKPOINT_RESTORE
+       DIR("map_files",  S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
+#endif
+       DIR("fdinfo",     S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
+       DIR("ns",         S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
+#ifdef CONFIG_NET
+       DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
+#endif
+       REG("environ",    S_IRUSR, proc_environ_operations),
+       INF("auxv",       S_IRUSR, proc_pid_auxv),
+       ONE("status",     S_IRUGO, proc_pid_status),
+       ONE("personality", S_IRUGO, proc_pid_personality),
+       INF("limits",     S_IRUGO, proc_pid_limits),
 #ifdef CONFIG_SCHED_DEBUG
-       REG("sched",      S_IRUGO|S_IWUSR, pid_sched),
+       REG("sched",      S_IRUGO|S_IWUSR, proc_pid_sched_operations),
 #endif
-       INF("cmdline",    S_IRUGO, pid_cmdline),
-       ONE("stat",       S_IRUGO, tgid_stat),
-       ONE("statm",      S_IRUGO, pid_statm),
-       REG("maps",       S_IRUGO, maps),
+#ifdef CONFIG_SCHED_AUTOGROUP
+       REG("autogroup",  S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
+#endif
+       REG("comm",      S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
+#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
+       INF("syscall",    S_IRUGO, proc_pid_syscall),
+#endif
+       INF("cmdline",    S_IRUGO, proc_pid_cmdline),
+       ONE("stat",       S_IRUGO, proc_tgid_stat),
+       ONE("statm",      S_IRUGO, proc_pid_statm),
+       REG("maps",       S_IRUGO, proc_pid_maps_operations),
 #ifdef CONFIG_NUMA
-       REG("numa_maps",  S_IRUGO, numa_maps),
+       REG("numa_maps",  S_IRUGO, proc_pid_numa_maps_operations),
 #endif
-       REG("mem",        S_IRUSR|S_IWUSR, mem),
-       LNK("cwd",        cwd),
-       LNK("root",       root),
-       LNK("exe",        exe),
-       REG("mounts",     S_IRUGO, mounts),
-       REG("mountstats", S_IRUSR, mountstats),
+       REG("mem",        S_IRUSR|S_IWUSR, proc_mem_operations),
+       LNK("cwd",        proc_cwd_link),
+       LNK("root",       proc_root_link),
+       LNK("exe",        proc_exe_link),
+       REG("mounts",     S_IRUGO, proc_mounts_operations),
+       REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
+       REG("mountstats", S_IRUSR, proc_mountstats_operations),
 #ifdef CONFIG_PROC_PAGE_MONITOR
-       REG("clear_refs", S_IWUSR, clear_refs),
-       REG("smaps",      S_IRUGO, smaps),
-       REG("pagemap",    S_IRUSR, pagemap),
+       REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
+       REG("smaps",      S_IRUGO, proc_pid_smaps_operations),
+       REG("pagemap",    S_IRUGO, proc_pagemap_operations),
 #endif
 #ifdef CONFIG_SECURITY
-       DIR("attr",       S_IRUGO|S_IXUGO, attr_dir),
+       DIR("attr",       S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
 #endif
 #ifdef CONFIG_KALLSYMS
-       INF("wchan",      S_IRUGO, pid_wchan),
+       INF("wchan",      S_IRUGO, proc_pid_wchan),
+#endif
+#ifdef CONFIG_STACKTRACE
+       ONE("stack",      S_IRUGO, proc_pid_stack),
 #endif
 #ifdef CONFIG_SCHEDSTATS
-       INF("schedstat",  S_IRUGO, pid_schedstat),
+       INF("schedstat",  S_IRUGO, proc_pid_schedstat),
 #endif
 #ifdef CONFIG_LATENCYTOP
-       REG("latency",  S_IRUGO, lstats),
+       REG("latency",  S_IRUGO, proc_lstats_operations),
 #endif
 #ifdef CONFIG_PROC_PID_CPUSET
-       REG("cpuset",     S_IRUGO, cpuset),
+       REG("cpuset",     S_IRUGO, proc_cpuset_operations),
 #endif
 #ifdef CONFIG_CGROUPS
-       REG("cgroup",  S_IRUGO, cgroup),
+       REG("cgroup",  S_IRUGO, proc_cgroup_operations),
 #endif
-       INF("oom_score",  S_IRUGO, oom_score),
-       REG("oom_adj",    S_IRUGO|S_IWUSR, oom_adjust),
+       INF("oom_score",  S_IRUGO, proc_oom_score),
+       ANDROID("oom_adj",S_IRUGO|S_IWUSR, oom_adjust),
+       REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
 #ifdef CONFIG_AUDITSYSCALL
-       REG("loginuid",   S_IWUSR|S_IRUGO, loginuid),
+       REG("loginuid",   S_IWUSR|S_IRUGO, proc_loginuid_operations),
+       REG("sessionid",  S_IRUGO, proc_sessionid_operations),
 #endif
 #ifdef CONFIG_FAULT_INJECTION
-       REG("make-it-fail", S_IRUGO|S_IWUSR, fault_inject),
+       REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
 #endif
-#if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
-       REG("coredump_filter", S_IRUGO|S_IWUSR, coredump_filter),
+#ifdef CONFIG_ELF_CORE
+       REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
 #endif
 #ifdef CONFIG_TASK_IO_ACCOUNTING
-       INF("io",       S_IRUGO, pid_io_accounting),
+       INF("io",       S_IRUSR, proc_tgid_io_accounting),
+#endif
+#ifdef CONFIG_HARDWALL
+       INF("hardwall",   S_IRUGO, proc_pid_hardwall),
 #endif
 };
 
@@ -2342,6 +3078,7 @@ static int proc_tgid_base_readdir(struct file * filp,
 static const struct file_operations proc_tgid_base_operations = {
        .read           = generic_read_dir,
        .readdir        = proc_tgid_base_readdir,
+       .llseek         = default_llseek,
 };
 
 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
@@ -2353,6 +3090,7 @@ static const struct inode_operations proc_tgid_base_inode_operations = {
        .lookup         = proc_tgid_base_lookup,
        .getattr        = pid_getattr,
        .setattr        = proc_setattr,
+       .permission     = proc_pid_permission,
 };
 
 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
@@ -2365,15 +3103,11 @@ static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
        name.len = snprintf(buf, sizeof(buf), "%d", pid);
        dentry = d_hash_and_lookup(mnt->mnt_root, &name);
        if (dentry) {
-               if (!(current->flags & PF_EXITING))
-                       shrink_dcache_parent(dentry);
+               shrink_dcache_parent(dentry);
                d_drop(dentry);
                dput(dentry);
        }
 
-       if (tgid == 0)
-               goto out;
-
        name.name = buf;
        name.len = snprintf(buf, sizeof(buf), "%d", tgid);
        leader = d_hash_and_lookup(mnt->mnt_root, &name);
@@ -2430,17 +3164,16 @@ out:
 void proc_flush_task(struct task_struct *task)
 {
        int i;
-       struct pid *pid, *tgid = NULL;
+       struct pid *pid, *tgid;
        struct upid *upid;
 
        pid = task_pid(task);
-       if (thread_group_leader(task))
-               tgid = task_tgid(task);
+       tgid = task_tgid(task);
 
        for (i = 0; i <= pid->level; i++) {
                upid = &pid->numbers[i];
                proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
-                       tgid ? tgid->numbers[i].nr : 0);
+                                       tgid->numbers[i].nr);
        }
 
        upid = &pid->numbers[pid->level];
@@ -2463,12 +3196,11 @@ static struct dentry *proc_pid_instantiate(struct inode *dir,
        inode->i_op = &proc_tgid_base_inode_operations;
        inode->i_fop = &proc_tgid_base_operations;
        inode->i_flags|=S_IMMUTABLE;
-       inode->i_nlink = 5;
-#ifdef CONFIG_SECURITY
-       inode->i_nlink += 1;
-#endif
 
-       dentry->d_op = &pid_dentry_operations;
+       set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
+                                                 ARRAY_SIZE(tgid_base_stuff)));
+
+       d_set_d_op(dentry, &pid_dentry_operations);
 
        d_add(dentry, inode);
        /* Close the race of the process dying before we return the dentry */
@@ -2480,7 +3212,7 @@ out:
 
 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
 {
-       struct dentry *result = ERR_PTR(-ENOENT);
+       struct dentry *result;
        struct task_struct *task;
        unsigned tgid;
        struct pid_namespace *ns;
@@ -2562,14 +3294,26 @@ static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldi
                                proc_pid_instantiate, iter.task, NULL);
 }
 
+static int fake_filldir(void *buf, const char *name, int namelen,
+                       loff_t offset, u64 ino, unsigned d_type)
+{
+       return 0;
+}
+
 /* for the /proc/ directory itself, after non-process stuff has been done */
 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
 {
-       unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
-       struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
+       unsigned int nr;
+       struct task_struct *reaper;
        struct tgid_iter iter;
        struct pid_namespace *ns;
+       filldir_t __filldir;
+
+       if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
+               goto out_no_task;
+       nr = filp->f_pos - FIRST_PROCESS_ENTRY;
 
+       reaper = get_proc_task(filp->f_path.dentry->d_inode);
        if (!reaper)
                goto out_no_task;
 
@@ -2585,8 +3329,13 @@ int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
        for (iter = next_tgid(ns, iter);
             iter.task;
             iter.tgid += 1, iter = next_tgid(ns, iter)) {
+               if (has_pid_permissions(ns, iter.task, 2))
+                       __filldir = filldir;
+               else
+                       __filldir = fake_filldir;
+
                filp->f_pos = iter.tgid + TGID_OFFSET;
-               if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
+               if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) {
                        put_task_struct(iter.task);
                        goto out;
                }
@@ -2602,57 +3351,75 @@ out_no_task:
  * Tasks
  */
 static const struct pid_entry tid_base_stuff[] = {
-       DIR("fd",        S_IRUSR|S_IXUSR, fd),
-       DIR("fdinfo",    S_IRUSR|S_IXUSR, fdinfo),
-       REG("environ",   S_IRUSR, environ),
-       INF("auxv",      S_IRUSR, pid_auxv),
-       ONE("status",    S_IRUGO, pid_status),
-       INF("limits",    S_IRUSR, pid_limits),
+       DIR("fd",        S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
+       DIR("fdinfo",    S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
+       DIR("ns",        S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
+       REG("environ",   S_IRUSR, proc_environ_operations),
+       INF("auxv",      S_IRUSR, proc_pid_auxv),
+       ONE("status",    S_IRUGO, proc_pid_status),
+       ONE("personality", S_IRUGO, proc_pid_personality),
+       INF("limits",    S_IRUGO, proc_pid_limits),
 #ifdef CONFIG_SCHED_DEBUG
-       REG("sched",     S_IRUGO|S_IWUSR, pid_sched),
+       REG("sched",     S_IRUGO|S_IWUSR, proc_pid_sched_operations),
 #endif
-       INF("cmdline",   S_IRUGO, pid_cmdline),
-       ONE("stat",      S_IRUGO, tid_stat),
-       ONE("statm",     S_IRUGO, pid_statm),
-       REG("maps",      S_IRUGO, maps),
+       REG("comm",      S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
+#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
+       INF("syscall",   S_IRUGO, proc_pid_syscall),
+#endif
+       INF("cmdline",   S_IRUGO, proc_pid_cmdline),
+       ONE("stat",      S_IRUGO, proc_tid_stat),
+       ONE("statm",     S_IRUGO, proc_pid_statm),
+       REG("maps",      S_IRUGO, proc_tid_maps_operations),
 #ifdef CONFIG_NUMA
-       REG("numa_maps", S_IRUGO, numa_maps),
+       REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
 #endif
-       REG("mem",       S_IRUSR|S_IWUSR, mem),
-       LNK("cwd",       cwd),
-       LNK("root",      root),
-       LNK("exe",       exe),
-       REG("mounts",    S_IRUGO, mounts),
+       REG("mem",       S_IRUSR|S_IWUSR, proc_mem_operations),
+       LNK("cwd",       proc_cwd_link),
+       LNK("root",      proc_root_link),
+       LNK("exe",       proc_exe_link),
+       REG("mounts",    S_IRUGO, proc_mounts_operations),
+       REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
 #ifdef CONFIG_PROC_PAGE_MONITOR
-       REG("clear_refs", S_IWUSR, clear_refs),
-       REG("smaps",     S_IRUGO, smaps),
-       REG("pagemap",    S_IRUSR, pagemap),
+       REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
+       REG("smaps",     S_IRUGO, proc_tid_smaps_operations),
+       REG("pagemap",    S_IRUGO, proc_pagemap_operations),
 #endif
 #ifdef CONFIG_SECURITY
-       DIR("attr",      S_IRUGO|S_IXUGO, attr_dir),
+       DIR("attr",      S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
 #endif
 #ifdef CONFIG_KALLSYMS
-       INF("wchan",     S_IRUGO, pid_wchan),
+       INF("wchan",     S_IRUGO, proc_pid_wchan),
+#endif
+#ifdef CONFIG_STACKTRACE
+       ONE("stack",      S_IRUGO, proc_pid_stack),
 #endif
 #ifdef CONFIG_SCHEDSTATS
-       INF("schedstat", S_IRUGO, pid_schedstat),
+       INF("schedstat", S_IRUGO, proc_pid_schedstat),
 #endif
 #ifdef CONFIG_LATENCYTOP
-       REG("latency",  S_IRUGO, lstats),
+       REG("latency",  S_IRUGO, proc_lstats_operations),
 #endif
 #ifdef CONFIG_PROC_PID_CPUSET
-       REG("cpuset",    S_IRUGO, cpuset),
+       REG("cpuset",    S_IRUGO, proc_cpuset_operations),
 #endif
 #ifdef CONFIG_CGROUPS
-       REG("cgroup",  S_IRUGO, cgroup),
+       REG("cgroup",  S_IRUGO, proc_cgroup_operations),
 #endif
-       INF("oom_score", S_IRUGO, oom_score),
-       REG("oom_adj",   S_IRUGO|S_IWUSR, oom_adjust),
+       INF("oom_score", S_IRUGO, proc_oom_score),
+       REG("oom_adj",   S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
+       REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
 #ifdef CONFIG_AUDITSYSCALL
-       REG("loginuid",  S_IWUSR|S_IRUGO, loginuid),
+       REG("loginuid",  S_IWUSR|S_IRUGO, proc_loginuid_operations),
+       REG("sessionid",  S_IRUGO, proc_sessionid_operations),
 #endif
 #ifdef CONFIG_FAULT_INJECTION
-       REG("make-it-fail", S_IRUGO|S_IWUSR, fault_inject),
+       REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
+#endif
+#ifdef CONFIG_TASK_IO_ACCOUNTING
+       INF("io",       S_IRUSR, proc_tid_io_accounting),
+#endif
+#ifdef CONFIG_HARDWALL
+       INF("hardwall",   S_IRUGO, proc_pid_hardwall),
 #endif
 };
 
@@ -2671,6 +3438,7 @@ static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *den
 static const struct file_operations proc_tid_base_operations = {
        .read           = generic_read_dir,
        .readdir        = proc_tid_base_readdir,
+       .llseek         = default_llseek,
 };
 
 static const struct inode_operations proc_tid_base_inode_operations = {
@@ -2692,12 +3460,11 @@ static struct dentry *proc_task_instantiate(struct inode *dir,
        inode->i_op = &proc_tid_base_inode_operations;
        inode->i_fop = &proc_tid_base_operations;
        inode->i_flags|=S_IMMUTABLE;
-       inode->i_nlink = 4;
-#ifdef CONFIG_SECURITY
-       inode->i_nlink += 1;
-#endif
 
-       dentry->d_op = &pid_dentry_operations;
+       set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
+                                                 ARRAY_SIZE(tid_base_stuff)));
+
+       d_set_d_op(dentry, &pid_dentry_operations);
 
        d_add(dentry, inode);
        /* Close the race of the process dying before we return the dentry */
@@ -2830,7 +3597,6 @@ static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldi
        int retval = -ENOENT;
        ino_t ino;
        int tid;
-       unsigned long pos = filp->f_pos;  /* avoiding "long long" filp->f_pos */
        struct pid_namespace *ns;
 
        task = get_proc_task(inode);
@@ -2847,18 +3613,18 @@ static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldi
                goto out_no_task;
        retval = 0;
 
-       switch (pos) {
+       switch ((unsigned long)filp->f_pos) {
        case 0:
                ino = inode->i_ino;
-               if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
+               if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
                        goto out;
-               pos++;
+               filp->f_pos++;
                /* fall through */
        case 1:
                ino = parent_ino(dentry);
-               if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
+               if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
                        goto out;
-               pos++;
+               filp->f_pos++;
                /* fall through */
        }
 
@@ -2868,9 +3634,9 @@ static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldi
        ns = filp->f_dentry->d_sb->s_fs_info;
        tid = (int)filp->f_version;
        filp->f_version = 0;
-       for (task = first_tid(leader, tid, pos - 2, ns);
+       for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
             task;
-            task = next_tid(task), pos++) {
+            task = next_tid(task), filp->f_pos++) {
                tid = task_pid_nr_ns(task, ns);
                if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
                        /* returning this tgid failed, save it as the first
@@ -2881,7 +3647,6 @@ static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldi
                }
        }
 out:
-       filp->f_pos = pos;
        put_task_struct(leader);
 out_no_task:
        return retval;
@@ -2894,9 +3659,7 @@ static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct
        generic_fillattr(inode, stat);
 
        if (p) {
-               rcu_read_lock();
                stat->nlink += get_nr_threads(p);
-               rcu_read_unlock();
                put_task_struct(p);
        }
 
@@ -2907,9 +3670,11 @@ static const struct inode_operations proc_task_inode_operations = {
        .lookup         = proc_task_lookup,
        .getattr        = proc_task_getattr,
        .setattr        = proc_setattr,
+       .permission     = proc_pid_permission,
 };
 
 static const struct file_operations proc_task_operations = {
        .read           = generic_read_dir,
        .readdir        = proc_task_readdir,
+       .llseek         = default_llseek,
 };