fs/file_table.c

   1 /*
   2  *  linux/fs/file_table.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  *  Copyright (C) 1997 David S. Miller (davem@caip.rutgers.edu)
   6  */
   7
   8 #include <linux/string.h>
   9 #include <linux/slab.h>
  10 #include <linux/file.h>
  11 #include <linux/init.h>
  12 #include <linux/module.h>
  13 #include <linux/smp_lock.h>
  14 #include <linux/fs.h>
  15 #include <linux/security.h>
  16 #include <linux/eventpoll.h>
  17 #include <linux/mount.h>
  18 #include <linux/cdev.h>
  19
  20 /* sysctl tunables... */
  21 struct files_stat_struct files_stat = {
  22         .max_files = NR_FILE
  23 };
  24
  25 EXPORT_SYMBOL(files_stat); /* Needed by unix.o */
  26
  27 /* public *and* exported. Not pretty! */
  28 spinlock_t __cacheline_aligned_in_smp files_lock = SPIN_LOCK_UNLOCKED;
  29
  30 EXPORT_SYMBOL(files_lock);
  31
  32 static spinlock_t filp_count_lock = SPIN_LOCK_UNLOCKED;
  33
  34 /* slab constructors and destructors are called from arbitrary
  35  * context and must be fully threaded - use a local spinlock
  36  * to protect files_stat.nr_files
  37  */
  38 void filp_ctor(void * objp, struct kmem_cache_s *cachep, unsigned long cflags)
  39 {
  40         if ((cflags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
  41             SLAB_CTOR_CONSTRUCTOR) {
  42                 unsigned long flags;
  43                 spin_lock_irqsave(&filp_count_lock, flags);
  44                 files_stat.nr_files++;
  45                 spin_unlock_irqrestore(&filp_count_lock, flags);
  46         }
  47 }
  48
  49 void filp_dtor(void * objp, struct kmem_cache_s *cachep, unsigned long dflags)
  50 {
  51         unsigned long flags;
  52         spin_lock_irqsave(&filp_count_lock, flags);
  53         files_stat.nr_files--;
  54         spin_unlock_irqrestore(&filp_count_lock, flags);
  55 }
  56
  57 static inline void file_free(struct file *f)
  58 {
  59         kmem_cache_free(filp_cachep, f);
  60 }
  61
  62 /* Find an unused file structure and return a pointer to it.
  63  * Returns NULL, if there are no more free file structures or
  64  * we run out of memory.
  65  */
  66 struct file *get_empty_filp(void)
  67 {
  68 static int old_max;
  69         struct file * f;
  70
  71         /*
  72          * Privileged users can go above max_files
  73          */
  74         if (files_stat.nr_files < files_stat.max_files ||
  75                                 capable(CAP_SYS_ADMIN)) {
  76                 f = kmem_cache_alloc(filp_cachep, GFP_KERNEL);
  77                 if (f) {
  78                         memset(f, 0, sizeof(*f));
  79                         if (security_file_alloc(f)) {
  80                                 file_free(f);
  81                                 goto fail;
  82                         }
  83                         eventpoll_init_file(f);
  84                         atomic_set(&f->f_count, 1);
  85                         f->f_uid = current->fsuid;
  86                         f->f_gid = current->fsgid;
  87                         f->f_owner.lock = RW_LOCK_UNLOCKED;
  88                         /* f->f_version: 0 */
  89                         INIT_LIST_HEAD(&f->f_list);
  90                         return f;
  91                 }
  92         }
  93
  94         /* Ran out of filps - report that */
  95         if (files_stat.max_files >= old_max) {
  96                 printk(KERN_INFO "VFS: file-max limit %d reached\n",
  97                                         files_stat.max_files);
  98                 old_max = files_stat.max_files;
  99         } else {
 100                 /* Big problems... */
 101                 printk(KERN_WARNING "VFS: filp allocation failed\n");
 102         }
 103 fail:
 104         return NULL;
 105 }
 106
 107 EXPORT_SYMBOL(get_empty_filp);
 108
 109 /*
 110  * Clear and initialize a (private) struct file for the given dentry,
 111  * allocate the security structure, and call the open function (if any).
 112  * The file should be released using close_private_file.
 113  */
 114 int open_private_file(struct file *filp, struct dentry *dentry, int flags)
 115 {
 116         int error;
 117         memset(filp, 0, sizeof(*filp));
 118         eventpoll_init_file(filp);
 119         filp->f_flags  = flags;
 120         filp->f_mode   = (flags+1) & O_ACCMODE;
 121         atomic_set(&filp->f_count, 1);
 122         filp->f_dentry = dentry;
 123         filp->f_mapping = dentry->d_inode->i_mapping;
 124         filp->f_uid    = current->fsuid;
 125         filp->f_gid    = current->fsgid;
 126         filp->f_op     = dentry->d_inode->i_fop;
 127         INIT_LIST_HEAD(&filp->f_list);
 128         error = security_file_alloc(filp);
 129         if (!error)
 130                 if (filp->f_op && filp->f_op->open) {
 131                         error = filp->f_op->open(dentry->d_inode, filp);
 132                         if (error)
 133                                 security_file_free(filp);
 134                 }
 135         return error;
 136 }
 137
 138 EXPORT_SYMBOL(open_private_file);
 139
 140 /*
 141  * Release a private file by calling the release function (if any) and
 142  * freeing the security structure.
 143  */
 144 void close_private_file(struct file *file)
 145 {
 146         struct inode * inode = file->f_dentry->d_inode;
 147
 148         if (file->f_op && file->f_op->release)
 149                 file->f_op->release(inode, file);
 150         security_file_free(file);
 151 }
 152
 153 EXPORT_SYMBOL(close_private_file);
 154
 155 void fput(struct file *file)
 156 {
 157         if (atomic_dec_and_test(&file->f_count))
 158                 __fput(file);
 159 }
 160
 161 EXPORT_SYMBOL(fput);
 162
 163 /* __fput is called from task context when aio completion releases the last
 164  * last use of a struct file *.  Do not use otherwise.
 165  */
 166 void __fput(struct file *file)
 167 {
 168         struct dentry *dentry = file->f_dentry;
 169         struct vfsmount *mnt = file->f_vfsmnt;
 170         struct inode *inode = dentry->d_inode;
 171
 172         /*
 173          * The function eventpoll_release() should be the first called
 174          * in the file cleanup chain.
 175          */
 176         eventpoll_release(file);
 177         locks_remove_flock(file);
 178
 179         if (file->f_op && file->f_op->release)
 180                 file->f_op->release(inode, file);
 181         security_file_free(file);
 182         if (unlikely(inode->i_cdev != NULL))
 183                 cdev_put(inode->i_cdev);
 184         fops_put(file->f_op);
 185         if (file->f_mode & FMODE_WRITE)
 186                 put_write_access(inode);
 187         file_kill(file);
 188         file->f_dentry = NULL;
 189         file->f_vfsmnt = NULL;
 190         file_free(file);
 191         dput(dentry);
 192         mntput(mnt);
 193 }
 194
 195 struct file *fget(unsigned int fd)
 196 {
 197         struct file *file;
 198         struct files_struct *files = current->files;
 199
 200         spin_lock(&files->file_lock);
 201         file = fcheck(fd);
 202         if (file)
 203                 get_file(file);
 204         spin_unlock(&files->file_lock);
 205         return file;
 206 }
 207
 208 EXPORT_SYMBOL(fget);
 209
 210 /*
 211  * Lightweight file lookup - no refcnt increment if fd table isn't shared.
 212  * You can use this only if it is guranteed that the current task already
 213  * holds a refcnt to that file. That check has to be done at fget() only
 214  * and a flag is returned to be passed to the corresponding fput_light().
 215  * There must not be a cloning between an fget_light/fput_light pair.
 216  */
 217 struct file *fget_light(unsigned int fd, int *fput_needed)
 218 {
 219         struct file *file;
 220         struct files_struct *files = current->files;
 221
 222         *fput_needed = 0;
 223         if (likely((atomic_read(&files->count) == 1))) {
 224                 file = fcheck(fd);
 225         } else {
 226                 spin_lock(&files->file_lock);
 227                 file = fcheck(fd);
 228                 if (file) {
 229                         get_file(file);
 230                         *fput_needed = 1;
 231                 }
 232                 spin_unlock(&files->file_lock);
 233         }
 234         return file;
 235 }
 236
 237
 238 void put_filp(struct file *file)
 239 {
 240         if (atomic_dec_and_test(&file->f_count)) {
 241                 security_file_free(file);
 242                 file_kill(file);
 243                 file_free(file);
 244         }
 245 }
 246
 247 EXPORT_SYMBOL(put_filp);
 248
 249 void file_move(struct file *file, struct list_head *list)
 250 {
 251         if (!list)
 252                 return;
 253         file_list_lock();
 254         list_move(&file->f_list, list);
 255         file_list_unlock();
 256 }
 257
 258 void file_kill(struct file *file)
 259 {
 260         if (!list_empty(&file->f_list)) {
 261                 file_list_lock();
 262                 list_del_init(&file->f_list);
 263                 file_list_unlock();
 264         }
 265 }
 266
 267 int fs_may_remount_ro(struct super_block *sb)
 268 {
 269         struct list_head *p;
 270
 271         /* Check that no files are currently opened for writing. */
 272         file_list_lock();
 273         list_for_each(p, &sb->s_files) {
 274                 struct file *file = list_entry(p, struct file, f_list);
 275                 struct inode *inode = file->f_dentry->d_inode;
 276
 277                 /* File with pending delete? */
 278                 if (inode->i_nlink == 0)
 279                         goto too_bad;
 280
 281                 /* Writeable file? */
 282                 if (S_ISREG(inode->i_mode) && (file->f_mode & FMODE_WRITE))
 283                         goto too_bad;
 284         }
 285         file_list_unlock();
 286         return 1; /* Tis' cool bro. */
 287 too_bad:
 288         file_list_unlock();
 289         return 0;
 290 }
 291
 292 void __init files_init(unsigned long mempages)
 293 {
 294         int n;
 295         /* One file with associated inode and dcache is very roughly 1K.
 296          * Per default don't use more than 10% of our memory for files.
 297          */
 298
 299         n = (mempages * (PAGE_SIZE / 1024)) / 10;
 300         files_stat.max_files = n;
 301         if (files_stat.max_files < NR_FILE)
 302                 files_stat.max_files = NR_FILE;
 303 }