active-standby: add cmdline into /proc/net/
[linux-2.6.git] / fs / proc / base.c
index 9cde9ed..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>
@@ -137,6 +139,12 @@ struct pid_entry {
 
 static int proc_fd_permission(struct inode *inode, int mask);
 
+/* 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.
@@ -198,32 +206,12 @@ static int proc_root_link(struct dentry *dentry, struct path *path)
        return result;
 }
 
-static struct mm_struct *mm_access(struct task_struct *task, unsigned int mode)
-{
-       struct mm_struct *mm;
-       int err;
-
-       err =  mutex_lock_killable(&task->signal->cred_guard_mutex);
-       if (err)
-               return ERR_PTR(err);
-
-       mm = get_task_mm(task);
-       if (mm && mm != current->mm &&
-                       !ptrace_may_access(task, mode)) {
-               mmput(mm);
-               mm = ERR_PTR(-EACCES);
-       }
-       mutex_unlock(&task->signal->cred_guard_mutex);
-
-       return mm;
-}
-
 struct mm_struct *mm_for_maps(struct task_struct *task)
 {
        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;
@@ -711,6 +699,13 @@ static int mem_open(struct inode* inode, struct file* file)
        if (IS_ERR(mm))
                return PTR_ERR(mm);
 
+       if (mm) {
+               /* ensure this mm_struct can't be freed */
+               atomic_inc(&mm->mm_count);
+               /* but do not pin its memory */
+               mmput(mm);
+       }
+
        /* OK to pass negative loff_t, we can catch out-of-range */
        file->f_mode |= FMODE_UNSIGNED_OFFSET;
        file->private_data = mm;
@@ -718,57 +713,13 @@ static int mem_open(struct inode* inode, struct file* file)
        return 0;
 }
 
-static ssize_t mem_read(struct file * file, 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 ret;
-       char *page;
-       unsigned long src = *ppos;
        struct mm_struct *mm = file->private_data;
-
-       if (!mm)
-               return 0;
-
-       page = (char *)__get_free_page(GFP_TEMPORARY);
-       if (!page)
-               return -ENOMEM;
-
-       ret = 0;
-       while (count > 0) {
-               int this_len, retval;
-
-               this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
-               retval = access_remote_vm(mm, src, page, this_len, 0);
-               if (!retval) {
-                       if (!ret)
-                               ret = -EIO;
-                       break;
-               }
-
-               if (copy_to_user(buf, page, retval)) {
-                       ret = -EFAULT;
-                       break;
-               }
-               ret += retval;
-               src += retval;
-               buf += retval;
-               count -= retval;
-       }
-       *ppos = src;
-
-       free_page((unsigned long) page);
-       return ret;
-}
-
-static ssize_t mem_write(struct file * file, const char __user *buf,
-                        size_t count, loff_t *ppos)
-{
-       int copied;
+       unsigned long addr = *ppos;
+       ssize_t copied;
        char *page;
-       unsigned long dst = *ppos;
-       struct mm_struct *mm = file->private_data;
 
        if (!mm)
                return 0;
@@ -778,31 +729,54 @@ static ssize_t mem_write(struct file * file, const char __user *buf,
                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_remote_vm(mm, 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);
        return copied;
 }
 
+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)
 {
        switch (orig) {
@@ -822,8 +796,8 @@ loff_t mem_lseek(struct file *file, loff_t offset, int orig)
 static int mem_release(struct inode *inode, struct file *file)
 {
        struct mm_struct *mm = file->private_data;
-
-       mmput(mm);
+       if (mm)
+               mmdrop(mm);
        return 0;
 }
 
@@ -1003,6 +977,35 @@ 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,
@@ -1344,8 +1347,7 @@ sched_autogroup_write(struct file *file, const char __user *buf,
        if (!p)
                return -ESRCH;
 
-       err = nice;
-       err = proc_sched_autogroup_set_nice(p, &err);
+       err = proc_sched_autogroup_set_nice(p, nice);
        if (err)
                count = err;
 
@@ -1788,7 +1790,7 @@ static int proc_fd_info(struct inode *inode, struct path *path, char *info)
 
                        fdt = files_fdtable(files);
                        f_flags = file->f_flags & ~O_CLOEXEC;
-                       if (FD_ISSET(fd, fdt->close_on_exec))
+                       if (close_on_exec(fd, fdt))
                                f_flags |= O_CLOEXEC;
 
                        if (path) {
@@ -1834,10 +1836,15 @@ static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
        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)) {
                                        rcu_read_lock();
                                        cred = __task_cred(task);
@@ -1848,7 +1855,16 @@ static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
                                        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;
@@ -1872,8 +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);
@@ -1883,26 +1897,8 @@ 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 & FMODE_READ)
-               inode->i_mode |= S_IRUSR | S_IXUSR;
-       if (file->f_mode & FMODE_WRITE)
-               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;
@@ -1914,12 +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,
@@ -2212,16 +2202,16 @@ static struct dentry *proc_map_files_lookup(struct inode *dir,
                goto out;
 
        result = ERR_PTR(-EACCES);
-       if (lock_trace(task))
+       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_unlock;
+               goto out_put_task;
 
        mm = get_task_mm(task);
        if (!mm)
-               goto out_unlock;
+               goto out_put_task;
 
        down_read(&mm->mmap_sem);
        vma = find_exact_vma(mm, vm_start, vm_end);
@@ -2233,8 +2223,6 @@ static struct dentry *proc_map_files_lookup(struct inode *dir,
 out_no_vma:
        up_read(&mm->mmap_sem);
        mmput(mm);
-out_unlock:
-       unlock_trace(task);
 out_put_task:
        put_task_struct(task);
 out:
@@ -2268,7 +2256,7 @@ proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
                goto out;
 
        ret = -EACCES;
-       if (lock_trace(task))
+       if (!ptrace_may_access(task, PTRACE_MODE_READ))
                goto out_put_task;
 
        ret = 0;
@@ -2276,12 +2264,12 @@ proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
        case 0:
                ino = inode->i_ino;
                if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
-                       goto out_unlock;
+                       goto out_put_task;
                filp->f_pos++;
        case 1:
                ino = parent_ino(dentry);
                if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
-                       goto out_unlock;
+                       goto out_put_task;
                filp->f_pos++;
        default:
        {
@@ -2292,7 +2280,7 @@ proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
 
                mm = get_task_mm(task);
                if (!mm)
-                       goto out_unlock;
+                       goto out_put_task;
                down_read(&mm->mmap_sem);
 
                nr_files = 0;
@@ -2322,7 +2310,7 @@ proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
                                        flex_array_free(fa);
                                up_read(&mm->mmap_sem);
                                mmput(mm);
-                               goto out_unlock;
+                               goto out_put_task;
                        }
                        for (i = 0, vma = mm->mmap, pos = 2; vma;
                                        vma = vma->vm_next) {
@@ -2367,8 +2355,6 @@ proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
        }
        }
 
-out_unlock:
-       unlock_trace(task);
 out_put_task:
        put_task_struct(task);
 out:
@@ -3024,9 +3010,9 @@ static const struct pid_entry tgid_base_stuff[] = {
        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_maps_operations),
+       REG("maps",       S_IRUGO, proc_pid_maps_operations),
 #ifdef CONFIG_NUMA
-       REG("numa_maps",  S_IRUGO, proc_numa_maps_operations),
+       REG("numa_maps",  S_IRUGO, proc_pid_numa_maps_operations),
 #endif
        REG("mem",        S_IRUSR|S_IWUSR, proc_mem_operations),
        LNK("cwd",        proc_cwd_link),
@@ -3037,7 +3023,7 @@ static const struct pid_entry tgid_base_stuff[] = {
        REG("mountstats", S_IRUSR, proc_mountstats_operations),
 #ifdef CONFIG_PROC_PAGE_MONITOR
        REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
-       REG("smaps",      S_IRUGO, proc_smaps_operations),
+       REG("smaps",      S_IRUGO, proc_pid_smaps_operations),
        REG("pagemap",    S_IRUGO, proc_pagemap_operations),
 #endif
 #ifdef CONFIG_SECURITY
@@ -3062,7 +3048,7 @@ static const struct pid_entry tgid_base_stuff[] = {
        REG("cgroup",  S_IRUGO, proc_cgroup_operations),
 #endif
        INF("oom_score",  S_IRUGO, proc_oom_score),
-       REG("oom_adj",    S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
+       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, proc_loginuid_operations),
@@ -3383,9 +3369,9 @@ static const struct pid_entry tid_base_stuff[] = {
        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_maps_operations),
+       REG("maps",      S_IRUGO, proc_tid_maps_operations),
 #ifdef CONFIG_NUMA
-       REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
+       REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
 #endif
        REG("mem",       S_IRUSR|S_IWUSR, proc_mem_operations),
        LNK("cwd",       proc_cwd_link),
@@ -3395,7 +3381,7 @@ static const struct pid_entry tid_base_stuff[] = {
        REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
 #ifdef CONFIG_PROC_PAGE_MONITOR
        REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
-       REG("smaps",     S_IRUGO, proc_smaps_operations),
+       REG("smaps",     S_IRUGO, proc_tid_smaps_operations),
        REG("pagemap",    S_IRUGO, proc_pagemap_operations),
 #endif
 #ifdef CONFIG_SECURITY