LCOV - code coverage report
Current view: top level - fs - file.c (source / functions) Hit Total Coverage
Test: coverage.info Lines: 65 515 12.6 %
Date: 2023-04-06 08:38:28 Functions: 5 54 9.3 %

          Line data    Source code
       1             : // SPDX-License-Identifier: GPL-2.0
       2             : /*
       3             :  *  linux/fs/file.c
       4             :  *
       5             :  *  Copyright (C) 1998-1999, Stephen Tweedie and Bill Hawes
       6             :  *
       7             :  *  Manage the dynamic fd arrays in the process files_struct.
       8             :  */
       9             : 
      10             : #include <linux/syscalls.h>
      11             : #include <linux/export.h>
      12             : #include <linux/fs.h>
      13             : #include <linux/kernel.h>
      14             : #include <linux/mm.h>
      15             : #include <linux/sched/signal.h>
      16             : #include <linux/slab.h>
      17             : #include <linux/file.h>
      18             : #include <linux/fdtable.h>
      19             : #include <linux/bitops.h>
      20             : #include <linux/spinlock.h>
      21             : #include <linux/rcupdate.h>
      22             : #include <linux/close_range.h>
      23             : #include <net/sock.h>
      24             : 
      25             : #include "internal.h"
      26             : 
      27             : unsigned int sysctl_nr_open __read_mostly = 1024*1024;
      28             : unsigned int sysctl_nr_open_min = BITS_PER_LONG;
      29             : /* our min() is unusable in constant expressions ;-/ */
      30             : #define __const_min(x, y) ((x) < (y) ? (x) : (y))
      31             : unsigned int sysctl_nr_open_max =
      32             :         __const_min(INT_MAX, ~(size_t)0/sizeof(void *)) & -BITS_PER_LONG;
      33             : 
      34           0 : static void __free_fdtable(struct fdtable *fdt)
      35             : {
      36           0 :         kvfree(fdt->fd);
      37           0 :         kvfree(fdt->open_fds);
      38           0 :         kfree(fdt);
      39           0 : }
      40             : 
      41           0 : static void free_fdtable_rcu(struct rcu_head *rcu)
      42             : {
      43           0 :         __free_fdtable(container_of(rcu, struct fdtable, rcu));
      44           0 : }
      45             : 
      46             : #define BITBIT_NR(nr)   BITS_TO_LONGS(BITS_TO_LONGS(nr))
      47             : #define BITBIT_SIZE(nr) (BITBIT_NR(nr) * sizeof(long))
      48             : 
      49             : /*
      50             :  * Copy 'count' fd bits from the old table to the new table and clear the extra
      51             :  * space if any.  This does not copy the file pointers.  Called with the files
      52             :  * spinlock held for write.
      53             :  */
      54           1 : static void copy_fd_bitmaps(struct fdtable *nfdt, struct fdtable *ofdt,
      55             :                             unsigned int count)
      56             : {
      57             :         unsigned int cpy, set;
      58             : 
      59           1 :         cpy = count / BITS_PER_BYTE;
      60           1 :         set = (nfdt->max_fds - count) / BITS_PER_BYTE;
      61           1 :         memcpy(nfdt->open_fds, ofdt->open_fds, cpy);
      62           1 :         memset((char *)nfdt->open_fds + cpy, 0, set);
      63           1 :         memcpy(nfdt->close_on_exec, ofdt->close_on_exec, cpy);
      64           1 :         memset((char *)nfdt->close_on_exec + cpy, 0, set);
      65             : 
      66           1 :         cpy = BITBIT_SIZE(count);
      67           1 :         set = BITBIT_SIZE(nfdt->max_fds) - cpy;
      68           1 :         memcpy(nfdt->full_fds_bits, ofdt->full_fds_bits, cpy);
      69           1 :         memset((char *)nfdt->full_fds_bits + cpy, 0, set);
      70           1 : }
      71             : 
      72             : /*
      73             :  * Copy all file descriptors from the old table to the new, expanded table and
      74             :  * clear the extra space.  Called with the files spinlock held for write.
      75             :  */
      76           0 : static void copy_fdtable(struct fdtable *nfdt, struct fdtable *ofdt)
      77             : {
      78             :         size_t cpy, set;
      79             : 
      80           0 :         BUG_ON(nfdt->max_fds < ofdt->max_fds);
      81             : 
      82           0 :         cpy = ofdt->max_fds * sizeof(struct file *);
      83           0 :         set = (nfdt->max_fds - ofdt->max_fds) * sizeof(struct file *);
      84           0 :         memcpy(nfdt->fd, ofdt->fd, cpy);
      85           0 :         memset((char *)nfdt->fd + cpy, 0, set);
      86             : 
      87           0 :         copy_fd_bitmaps(nfdt, ofdt, ofdt->max_fds);
      88           0 : }
      89             : 
      90             : /*
      91             :  * Note how the fdtable bitmap allocations very much have to be a multiple of
      92             :  * BITS_PER_LONG. This is not only because we walk those things in chunks of
      93             :  * 'unsigned long' in some places, but simply because that is how the Linux
      94             :  * kernel bitmaps are defined to work: they are not "bits in an array of bytes",
      95             :  * they are very much "bits in an array of unsigned long".
      96             :  *
      97             :  * The ALIGN(nr, BITS_PER_LONG) here is for clarity: since we just multiplied
      98             :  * by that "1024/sizeof(ptr)" before, we already know there are sufficient
      99             :  * clear low bits. Clang seems to realize that, gcc ends up being confused.
     100             :  *
     101             :  * On a 128-bit machine, the ALIGN() would actually matter. In the meantime,
     102             :  * let's consider it documentation (and maybe a test-case for gcc to improve
     103             :  * its code generation ;)
     104             :  */
     105           0 : static struct fdtable * alloc_fdtable(unsigned int nr)
     106             : {
     107             :         struct fdtable *fdt;
     108             :         void *data;
     109             : 
     110             :         /*
     111             :          * Figure out how many fds we actually want to support in this fdtable.
     112             :          * Allocation steps are keyed to the size of the fdarray, since it
     113             :          * grows far faster than any of the other dynamic data. We try to fit
     114             :          * the fdarray into comfortable page-tuned chunks: starting at 1024B
     115             :          * and growing in powers of two from there on.
     116             :          */
     117           0 :         nr /= (1024 / sizeof(struct file *));
     118           0 :         nr = roundup_pow_of_two(nr + 1);
     119           0 :         nr *= (1024 / sizeof(struct file *));
     120           0 :         nr = ALIGN(nr, BITS_PER_LONG);
     121             :         /*
     122             :          * Note that this can drive nr *below* what we had passed if sysctl_nr_open
     123             :          * had been set lower between the check in expand_files() and here.  Deal
     124             :          * with that in caller, it's cheaper that way.
     125             :          *
     126             :          * We make sure that nr remains a multiple of BITS_PER_LONG - otherwise
     127             :          * bitmaps handling below becomes unpleasant, to put it mildly...
     128             :          */
     129           0 :         if (unlikely(nr > sysctl_nr_open))
     130           0 :                 nr = ((sysctl_nr_open - 1) | (BITS_PER_LONG - 1)) + 1;
     131             : 
     132           0 :         fdt = kmalloc(sizeof(struct fdtable), GFP_KERNEL_ACCOUNT);
     133           0 :         if (!fdt)
     134             :                 goto out;
     135           0 :         fdt->max_fds = nr;
     136           0 :         data = kvmalloc_array(nr, sizeof(struct file *), GFP_KERNEL_ACCOUNT);
     137           0 :         if (!data)
     138             :                 goto out_fdt;
     139           0 :         fdt->fd = data;
     140             : 
     141           0 :         data = kvmalloc(max_t(size_t,
     142             :                                  2 * nr / BITS_PER_BYTE + BITBIT_SIZE(nr), L1_CACHE_BYTES),
     143             :                                  GFP_KERNEL_ACCOUNT);
     144           0 :         if (!data)
     145             :                 goto out_arr;
     146           0 :         fdt->open_fds = data;
     147           0 :         data += nr / BITS_PER_BYTE;
     148           0 :         fdt->close_on_exec = data;
     149           0 :         data += nr / BITS_PER_BYTE;
     150           0 :         fdt->full_fds_bits = data;
     151             : 
     152           0 :         return fdt;
     153             : 
     154             : out_arr:
     155           0 :         kvfree(fdt->fd);
     156             : out_fdt:
     157           0 :         kfree(fdt);
     158             : out:
     159             :         return NULL;
     160             : }
     161             : 
     162             : /*
     163             :  * Expand the file descriptor table.
     164             :  * This function will allocate a new fdtable and both fd array and fdset, of
     165             :  * the given size.
     166             :  * Return <0 error code on error; 1 on successful completion.
     167             :  * The files->file_lock should be held on entry, and will be held on exit.
     168             :  */
     169           0 : static int expand_fdtable(struct files_struct *files, unsigned int nr)
     170             :         __releases(files->file_lock)
     171             :         __acquires(files->file_lock)
     172             : {
     173             :         struct fdtable *new_fdt, *cur_fdt;
     174             : 
     175           0 :         spin_unlock(&files->file_lock);
     176           0 :         new_fdt = alloc_fdtable(nr);
     177             : 
     178             :         /* make sure all fd_install() have seen resize_in_progress
     179             :          * or have finished their rcu_read_lock_sched() section.
     180             :          */
     181           0 :         if (atomic_read(&files->count) > 1)
     182           0 :                 synchronize_rcu();
     183             : 
     184           0 :         spin_lock(&files->file_lock);
     185           0 :         if (!new_fdt)
     186             :                 return -ENOMEM;
     187             :         /*
     188             :          * extremely unlikely race - sysctl_nr_open decreased between the check in
     189             :          * caller and alloc_fdtable().  Cheaper to catch it here...
     190             :          */
     191           0 :         if (unlikely(new_fdt->max_fds <= nr)) {
     192           0 :                 __free_fdtable(new_fdt);
     193           0 :                 return -EMFILE;
     194             :         }
     195           0 :         cur_fdt = files_fdtable(files);
     196           0 :         BUG_ON(nr < cur_fdt->max_fds);
     197           0 :         copy_fdtable(new_fdt, cur_fdt);
     198           0 :         rcu_assign_pointer(files->fdt, new_fdt);
     199           0 :         if (cur_fdt != &files->fdtab)
     200           0 :                 call_rcu(&cur_fdt->rcu, free_fdtable_rcu);
     201             :         /* coupled with smp_rmb() in fd_install() */
     202           0 :         smp_wmb();
     203           0 :         return 1;
     204             : }
     205             : 
     206             : /*
     207             :  * Expand files.
     208             :  * This function will expand the file structures, if the requested size exceeds
     209             :  * the current capacity and there is room for expansion.
     210             :  * Return <0 error code on error; 0 when nothing done; 1 when files were
     211             :  * expanded and execution may have blocked.
     212             :  * The files->file_lock should be held on entry, and will be held on exit.
     213             :  */
     214           0 : static int expand_files(struct files_struct *files, unsigned int nr)
     215             :         __releases(files->file_lock)
     216             :         __acquires(files->file_lock)
     217             : {
     218             :         struct fdtable *fdt;
     219           0 :         int expanded = 0;
     220             : 
     221             : repeat:
     222           0 :         fdt = files_fdtable(files);
     223             : 
     224             :         /* Do we need to expand? */
     225           0 :         if (nr < fdt->max_fds)
     226             :                 return expanded;
     227             : 
     228             :         /* Can we expand? */
     229           0 :         if (nr >= sysctl_nr_open)
     230             :                 return -EMFILE;
     231             : 
     232           0 :         if (unlikely(files->resize_in_progress)) {
     233           0 :                 spin_unlock(&files->file_lock);
     234           0 :                 expanded = 1;
     235           0 :                 wait_event(files->resize_wait, !files->resize_in_progress);
     236           0 :                 spin_lock(&files->file_lock);
     237             :                 goto repeat;
     238             :         }
     239             : 
     240             :         /* All good, so we try */
     241           0 :         files->resize_in_progress = true;
     242           0 :         expanded = expand_fdtable(files, nr);
     243           0 :         files->resize_in_progress = false;
     244             : 
     245           0 :         wake_up_all(&files->resize_wait);
     246           0 :         return expanded;
     247             : }
     248             : 
     249             : static inline void __set_close_on_exec(unsigned int fd, struct fdtable *fdt)
     250             : {
     251           0 :         __set_bit(fd, fdt->close_on_exec);
     252             : }
     253             : 
     254           0 : static inline void __clear_close_on_exec(unsigned int fd, struct fdtable *fdt)
     255             : {
     256           0 :         if (test_bit(fd, fdt->close_on_exec))
     257           0 :                 __clear_bit(fd, fdt->close_on_exec);
     258           0 : }
     259             : 
     260           0 : static inline void __set_open_fd(unsigned int fd, struct fdtable *fdt)
     261             : {
     262           0 :         __set_bit(fd, fdt->open_fds);
     263           0 :         fd /= BITS_PER_LONG;
     264           0 :         if (!~fdt->open_fds[fd])
     265           0 :                 __set_bit(fd, fdt->full_fds_bits);
     266           0 : }
     267             : 
     268          64 : static inline void __clear_open_fd(unsigned int fd, struct fdtable *fdt)
     269             : {
     270         128 :         __clear_bit(fd, fdt->open_fds);
     271         128 :         __clear_bit(fd / BITS_PER_LONG, fdt->full_fds_bits);
     272          64 : }
     273             : 
     274             : static unsigned int count_open_files(struct fdtable *fdt)
     275             : {
     276           1 :         unsigned int size = fdt->max_fds;
     277             :         unsigned int i;
     278             : 
     279             :         /* Find the last open fd */
     280           3 :         for (i = size / BITS_PER_LONG; i > 0; ) {
     281           1 :                 if (fdt->open_fds[--i])
     282             :                         break;
     283             :         }
     284           1 :         i = (i + 1) * BITS_PER_LONG;
     285             :         return i;
     286             : }
     287             : 
     288             : /*
     289             :  * Note that a sane fdtable size always has to be a multiple of
     290             :  * BITS_PER_LONG, since we have bitmaps that are sized by this.
     291             :  *
     292             :  * 'max_fds' will normally already be properly aligned, but it
     293             :  * turns out that in the close_range() -> __close_range() ->
     294             :  * unshare_fd() -> dup_fd() -> sane_fdtable_size() we can end
     295             :  * up having a 'max_fds' value that isn't already aligned.
     296             :  *
     297             :  * Rather than make close_range() have to worry about this,
     298             :  * just make that BITS_PER_LONG alignment be part of a sane
     299             :  * fdtable size. Becuase that's really what it is.
     300             :  */
     301             : static unsigned int sane_fdtable_size(struct fdtable *fdt, unsigned int max_fds)
     302             : {
     303             :         unsigned int count;
     304             : 
     305           2 :         count = count_open_files(fdt);
     306           1 :         if (max_fds < NR_OPEN_DEFAULT)
     307           0 :                 max_fds = NR_OPEN_DEFAULT;
     308           1 :         return ALIGN(min(count, max_fds), BITS_PER_LONG);
     309             : }
     310             : 
     311             : /*
     312             :  * Allocate a new files structure and copy contents from the
     313             :  * passed in files structure.
     314             :  * errorp will be valid only when the returned files_struct is NULL.
     315             :  */
     316           1 : struct files_struct *dup_fd(struct files_struct *oldf, unsigned int max_fds, int *errorp)
     317             : {
     318             :         struct files_struct *newf;
     319             :         struct file **old_fds, **new_fds;
     320             :         unsigned int open_files, i;
     321             :         struct fdtable *old_fdt, *new_fdt;
     322             : 
     323           1 :         *errorp = -ENOMEM;
     324           1 :         newf = kmem_cache_alloc(files_cachep, GFP_KERNEL);
     325           1 :         if (!newf)
     326             :                 goto out;
     327             : 
     328           2 :         atomic_set(&newf->count, 1);
     329             : 
     330           1 :         spin_lock_init(&newf->file_lock);
     331           1 :         newf->resize_in_progress = false;
     332           1 :         init_waitqueue_head(&newf->resize_wait);
     333           1 :         newf->next_fd = 0;
     334           1 :         new_fdt = &newf->fdtab;
     335           1 :         new_fdt->max_fds = NR_OPEN_DEFAULT;
     336           1 :         new_fdt->close_on_exec = newf->close_on_exec_init;
     337           1 :         new_fdt->open_fds = newf->open_fds_init;
     338           1 :         new_fdt->full_fds_bits = newf->full_fds_bits_init;
     339           1 :         new_fdt->fd = &newf->fd_array[0];
     340             : 
     341           2 :         spin_lock(&oldf->file_lock);
     342           1 :         old_fdt = files_fdtable(oldf);
     343           1 :         open_files = sane_fdtable_size(old_fdt, max_fds);
     344             : 
     345             :         /*
     346             :          * Check whether we need to allocate a larger fd array and fd set.
     347             :          */
     348           2 :         while (unlikely(open_files > new_fdt->max_fds)) {
     349           0 :                 spin_unlock(&oldf->file_lock);
     350             : 
     351           0 :                 if (new_fdt != &newf->fdtab)
     352           0 :                         __free_fdtable(new_fdt);
     353             : 
     354           0 :                 new_fdt = alloc_fdtable(open_files - 1);
     355           0 :                 if (!new_fdt) {
     356           0 :                         *errorp = -ENOMEM;
     357           0 :                         goto out_release;
     358             :                 }
     359             : 
     360             :                 /* beyond sysctl_nr_open; nothing to do */
     361           0 :                 if (unlikely(new_fdt->max_fds < open_files)) {
     362           0 :                         __free_fdtable(new_fdt);
     363           0 :                         *errorp = -EMFILE;
     364           0 :                         goto out_release;
     365             :                 }
     366             : 
     367             :                 /*
     368             :                  * Reacquire the oldf lock and a pointer to its fd table
     369             :                  * who knows it may have a new bigger fd table. We need
     370             :                  * the latest pointer.
     371             :                  */
     372           0 :                 spin_lock(&oldf->file_lock);
     373           0 :                 old_fdt = files_fdtable(oldf);
     374           0 :                 open_files = sane_fdtable_size(old_fdt, max_fds);
     375             :         }
     376             : 
     377           1 :         copy_fd_bitmaps(new_fdt, old_fdt, open_files);
     378             : 
     379           1 :         old_fds = old_fdt->fd;
     380           1 :         new_fds = new_fdt->fd;
     381             : 
     382          65 :         for (i = open_files; i != 0; i--) {
     383          64 :                 struct file *f = *old_fds++;
     384          64 :                 if (f) {
     385             :                         get_file(f);
     386             :                 } else {
     387             :                         /*
     388             :                          * The fd may be claimed in the fd bitmap but not yet
     389             :                          * instantiated in the files array if a sibling thread
     390             :                          * is partway through open().  So make sure that this
     391             :                          * fd is available to the new process.
     392             :                          */
     393          64 :                         __clear_open_fd(open_files - i, new_fdt);
     394             :                 }
     395          64 :                 rcu_assign_pointer(*new_fds++, f);
     396             :         }
     397           2 :         spin_unlock(&oldf->file_lock);
     398             : 
     399             :         /* clear the remainder */
     400           1 :         memset(new_fds, 0, (new_fdt->max_fds - open_files) * sizeof(struct file *));
     401             : 
     402           1 :         rcu_assign_pointer(newf->fdt, new_fdt);
     403             : 
     404           1 :         return newf;
     405             : 
     406             : out_release:
     407           0 :         kmem_cache_free(files_cachep, newf);
     408             : out:
     409             :         return NULL;
     410             : }
     411             : 
     412           0 : static struct fdtable *close_files(struct files_struct * files)
     413             : {
     414             :         /*
     415             :          * It is safe to dereference the fd table without RCU or
     416             :          * ->file_lock because this is the last reference to the
     417             :          * files structure.
     418             :          */
     419           0 :         struct fdtable *fdt = rcu_dereference_raw(files->fdt);
     420           0 :         unsigned int i, j = 0;
     421             : 
     422             :         for (;;) {
     423             :                 unsigned long set;
     424           0 :                 i = j * BITS_PER_LONG;
     425           0 :                 if (i >= fdt->max_fds)
     426             :                         break;
     427           0 :                 set = fdt->open_fds[j++];
     428           0 :                 while (set) {
     429           0 :                         if (set & 1) {
     430           0 :                                 struct file * file = xchg(&fdt->fd[i], NULL);
     431           0 :                                 if (file) {
     432           0 :                                         filp_close(file, files);
     433           0 :                                         cond_resched();
     434             :                                 }
     435             :                         }
     436           0 :                         i++;
     437           0 :                         set >>= 1;
     438             :                 }
     439             :         }
     440             : 
     441           0 :         return fdt;
     442             : }
     443             : 
     444         333 : void put_files_struct(struct files_struct *files)
     445             : {
     446         666 :         if (atomic_dec_and_test(&files->count)) {
     447           0 :                 struct fdtable *fdt = close_files(files);
     448             : 
     449             :                 /* free the arrays if they are not embedded */
     450           0 :                 if (fdt != &files->fdtab)
     451           0 :                         __free_fdtable(fdt);
     452           0 :                 kmem_cache_free(files_cachep, files);
     453             :         }
     454         333 : }
     455             : 
     456         333 : void exit_files(struct task_struct *tsk)
     457             : {
     458         333 :         struct files_struct * files = tsk->files;
     459             : 
     460         333 :         if (files) {
     461         333 :                 task_lock(tsk);
     462         333 :                 tsk->files = NULL;
     463         333 :                 task_unlock(tsk);
     464         333 :                 put_files_struct(files);
     465             :         }
     466         333 : }
     467             : 
     468             : struct files_struct init_files = {
     469             :         .count          = ATOMIC_INIT(1),
     470             :         .fdt            = &init_files.fdtab,
     471             :         .fdtab          = {
     472             :                 .max_fds        = NR_OPEN_DEFAULT,
     473             :                 .fd             = &init_files.fd_array[0],
     474             :                 .close_on_exec  = init_files.close_on_exec_init,
     475             :                 .open_fds       = init_files.open_fds_init,
     476             :                 .full_fds_bits  = init_files.full_fds_bits_init,
     477             :         },
     478             :         .file_lock      = __SPIN_LOCK_UNLOCKED(init_files.file_lock),
     479             :         .resize_wait    = __WAIT_QUEUE_HEAD_INITIALIZER(init_files.resize_wait),
     480             : };
     481             : 
     482           0 : static unsigned int find_next_fd(struct fdtable *fdt, unsigned int start)
     483             : {
     484           0 :         unsigned int maxfd = fdt->max_fds;
     485           0 :         unsigned int maxbit = maxfd / BITS_PER_LONG;
     486           0 :         unsigned int bitbit = start / BITS_PER_LONG;
     487             : 
     488           0 :         bitbit = find_next_zero_bit(fdt->full_fds_bits, maxbit, bitbit) * BITS_PER_LONG;
     489           0 :         if (bitbit > maxfd)
     490             :                 return maxfd;
     491           0 :         if (bitbit > start)
     492           0 :                 start = bitbit;
     493           0 :         return find_next_zero_bit(fdt->open_fds, maxfd, start);
     494             : }
     495             : 
     496             : /*
     497             :  * allocate a file descriptor, mark it busy.
     498             :  */
     499           0 : static int alloc_fd(unsigned start, unsigned end, unsigned flags)
     500             : {
     501           0 :         struct files_struct *files = current->files;
     502             :         unsigned int fd;
     503             :         int error;
     504             :         struct fdtable *fdt;
     505             : 
     506           0 :         spin_lock(&files->file_lock);
     507             : repeat:
     508           0 :         fdt = files_fdtable(files);
     509           0 :         fd = start;
     510           0 :         if (fd < files->next_fd)
     511           0 :                 fd = files->next_fd;
     512             : 
     513           0 :         if (fd < fdt->max_fds)
     514           0 :                 fd = find_next_fd(fdt, fd);
     515             : 
     516             :         /*
     517             :          * N.B. For clone tasks sharing a files structure, this test
     518             :          * will limit the total number of files that can be opened.
     519             :          */
     520           0 :         error = -EMFILE;
     521           0 :         if (fd >= end)
     522             :                 goto out;
     523             : 
     524           0 :         error = expand_files(files, fd);
     525           0 :         if (error < 0)
     526             :                 goto out;
     527             : 
     528             :         /*
     529             :          * If we needed to expand the fs array we
     530             :          * might have blocked - try again.
     531             :          */
     532           0 :         if (error)
     533             :                 goto repeat;
     534             : 
     535           0 :         if (start <= files->next_fd)
     536           0 :                 files->next_fd = fd + 1;
     537             : 
     538           0 :         __set_open_fd(fd, fdt);
     539           0 :         if (flags & O_CLOEXEC)
     540           0 :                 __set_close_on_exec(fd, fdt);
     541             :         else
     542           0 :                 __clear_close_on_exec(fd, fdt);
     543           0 :         error = fd;
     544             : #if 1
     545             :         /* Sanity check */
     546           0 :         if (rcu_access_pointer(fdt->fd[fd]) != NULL) {
     547           0 :                 printk(KERN_WARNING "alloc_fd: slot %d not NULL!\n", fd);
     548           0 :                 rcu_assign_pointer(fdt->fd[fd], NULL);
     549             :         }
     550             : #endif
     551             : 
     552             : out:
     553           0 :         spin_unlock(&files->file_lock);
     554           0 :         return error;
     555             : }
     556             : 
     557           0 : int __get_unused_fd_flags(unsigned flags, unsigned long nofile)
     558             : {
     559           0 :         return alloc_fd(0, nofile, flags);
     560             : }
     561             : 
     562           0 : int get_unused_fd_flags(unsigned flags)
     563             : {
     564           0 :         return __get_unused_fd_flags(flags, rlimit(RLIMIT_NOFILE));
     565             : }
     566             : EXPORT_SYMBOL(get_unused_fd_flags);
     567             : 
     568             : static void __put_unused_fd(struct files_struct *files, unsigned int fd)
     569             : {
     570           0 :         struct fdtable *fdt = files_fdtable(files);
     571           0 :         __clear_open_fd(fd, fdt);
     572           0 :         if (fd < files->next_fd)
     573           0 :                 files->next_fd = fd;
     574             : }
     575             : 
     576           0 : void put_unused_fd(unsigned int fd)
     577             : {
     578           0 :         struct files_struct *files = current->files;
     579           0 :         spin_lock(&files->file_lock);
     580           0 :         __put_unused_fd(files, fd);
     581           0 :         spin_unlock(&files->file_lock);
     582           0 : }
     583             : 
     584             : EXPORT_SYMBOL(put_unused_fd);
     585             : 
     586             : /*
     587             :  * Install a file pointer in the fd array.
     588             :  *
     589             :  * The VFS is full of places where we drop the files lock between
     590             :  * setting the open_fds bitmap and installing the file in the file
     591             :  * array.  At any such point, we are vulnerable to a dup2() race
     592             :  * installing a file in the array before us.  We need to detect this and
     593             :  * fput() the struct file we are about to overwrite in this case.
     594             :  *
     595             :  * It should never happen - if we allow dup2() do it, _really_ bad things
     596             :  * will follow.
     597             :  *
     598             :  * This consumes the "file" refcount, so callers should treat it
     599             :  * as if they had called fput(file).
     600             :  */
     601             : 
     602           0 : void fd_install(unsigned int fd, struct file *file)
     603             : {
     604           0 :         struct files_struct *files = current->files;
     605             :         struct fdtable *fdt;
     606             : 
     607             :         rcu_read_lock_sched();
     608             : 
     609           0 :         if (unlikely(files->resize_in_progress)) {
     610             :                 rcu_read_unlock_sched();
     611           0 :                 spin_lock(&files->file_lock);
     612           0 :                 fdt = files_fdtable(files);
     613           0 :                 BUG_ON(fdt->fd[fd] != NULL);
     614           0 :                 rcu_assign_pointer(fdt->fd[fd], file);
     615           0 :                 spin_unlock(&files->file_lock);
     616             :                 return;
     617             :         }
     618             :         /* coupled with smp_wmb() in expand_fdtable() */
     619           0 :         smp_rmb();
     620           0 :         fdt = rcu_dereference_sched(files->fdt);
     621           0 :         BUG_ON(fdt->fd[fd] != NULL);
     622           0 :         rcu_assign_pointer(fdt->fd[fd], file);
     623             :         rcu_read_unlock_sched();
     624             : }
     625             : 
     626             : EXPORT_SYMBOL(fd_install);
     627             : 
     628             : /**
     629             :  * pick_file - return file associatd with fd
     630             :  * @files: file struct to retrieve file from
     631             :  * @fd: file descriptor to retrieve file for
     632             :  *
     633             :  * Context: files_lock must be held.
     634             :  *
     635             :  * Returns: The file associated with @fd (NULL if @fd is not open)
     636             :  */
     637           0 : static struct file *pick_file(struct files_struct *files, unsigned fd)
     638             : {
     639           0 :         struct fdtable *fdt = files_fdtable(files);
     640             :         struct file *file;
     641             : 
     642           0 :         if (fd >= fdt->max_fds)
     643             :                 return NULL;
     644             : 
     645           0 :         file = fdt->fd[fd];
     646           0 :         if (file) {
     647           0 :                 rcu_assign_pointer(fdt->fd[fd], NULL);
     648             :                 __put_unused_fd(files, fd);
     649             :         }
     650             :         return file;
     651             : }
     652             : 
     653           0 : int close_fd(unsigned fd)
     654             : {
     655           0 :         struct files_struct *files = current->files;
     656             :         struct file *file;
     657             : 
     658           0 :         spin_lock(&files->file_lock);
     659           0 :         file = pick_file(files, fd);
     660           0 :         spin_unlock(&files->file_lock);
     661           0 :         if (!file)
     662             :                 return -EBADF;
     663             : 
     664           0 :         return filp_close(file, files);
     665             : }
     666             : EXPORT_SYMBOL(close_fd); /* for ksys_close() */
     667             : 
     668             : /**
     669             :  * last_fd - return last valid index into fd table
     670             :  * @cur_fds: files struct
     671             :  *
     672             :  * Context: Either rcu read lock or files_lock must be held.
     673             :  *
     674             :  * Returns: Last valid index into fdtable.
     675             :  */
     676             : static inline unsigned last_fd(struct fdtable *fdt)
     677             : {
     678           0 :         return fdt->max_fds - 1;
     679             : }
     680             : 
     681           0 : static inline void __range_cloexec(struct files_struct *cur_fds,
     682             :                                    unsigned int fd, unsigned int max_fd)
     683             : {
     684             :         struct fdtable *fdt;
     685             : 
     686             :         /* make sure we're using the correct maximum value */
     687           0 :         spin_lock(&cur_fds->file_lock);
     688           0 :         fdt = files_fdtable(cur_fds);
     689           0 :         max_fd = min(last_fd(fdt), max_fd);
     690           0 :         if (fd <= max_fd)
     691           0 :                 bitmap_set(fdt->close_on_exec, fd, max_fd - fd + 1);
     692           0 :         spin_unlock(&cur_fds->file_lock);
     693           0 : }
     694             : 
     695           0 : static inline void __range_close(struct files_struct *cur_fds, unsigned int fd,
     696             :                                  unsigned int max_fd)
     697             : {
     698             :         unsigned n;
     699             : 
     700             :         rcu_read_lock();
     701           0 :         n = last_fd(files_fdtable(cur_fds));
     702             :         rcu_read_unlock();
     703           0 :         max_fd = min(max_fd, n);
     704             : 
     705           0 :         while (fd <= max_fd) {
     706             :                 struct file *file;
     707             : 
     708           0 :                 spin_lock(&cur_fds->file_lock);
     709           0 :                 file = pick_file(cur_fds, fd++);
     710           0 :                 spin_unlock(&cur_fds->file_lock);
     711             : 
     712           0 :                 if (file) {
     713             :                         /* found a valid file to close */
     714           0 :                         filp_close(file, cur_fds);
     715           0 :                         cond_resched();
     716             :                 }
     717             :         }
     718           0 : }
     719             : 
     720             : /**
     721             :  * __close_range() - Close all file descriptors in a given range.
     722             :  *
     723             :  * @fd:     starting file descriptor to close
     724             :  * @max_fd: last file descriptor to close
     725             :  *
     726             :  * This closes a range of file descriptors. All file descriptors
     727             :  * from @fd up to and including @max_fd are closed.
     728             :  */
     729           0 : int __close_range(unsigned fd, unsigned max_fd, unsigned int flags)
     730             : {
     731           0 :         struct task_struct *me = current;
     732           0 :         struct files_struct *cur_fds = me->files, *fds = NULL;
     733             : 
     734           0 :         if (flags & ~(CLOSE_RANGE_UNSHARE | CLOSE_RANGE_CLOEXEC))
     735             :                 return -EINVAL;
     736             : 
     737           0 :         if (fd > max_fd)
     738             :                 return -EINVAL;
     739             : 
     740           0 :         if (flags & CLOSE_RANGE_UNSHARE) {
     741             :                 int ret;
     742           0 :                 unsigned int max_unshare_fds = NR_OPEN_MAX;
     743             : 
     744             :                 /*
     745             :                  * If the caller requested all fds to be made cloexec we always
     746             :                  * copy all of the file descriptors since they still want to
     747             :                  * use them.
     748             :                  */
     749           0 :                 if (!(flags & CLOSE_RANGE_CLOEXEC)) {
     750             :                         /*
     751             :                          * If the requested range is greater than the current
     752             :                          * maximum, we're closing everything so only copy all
     753             :                          * file descriptors beneath the lowest file descriptor.
     754             :                          */
     755             :                         rcu_read_lock();
     756           0 :                         if (max_fd >= last_fd(files_fdtable(cur_fds)))
     757           0 :                                 max_unshare_fds = fd;
     758             :                         rcu_read_unlock();
     759             :                 }
     760             : 
     761           0 :                 ret = unshare_fd(CLONE_FILES, max_unshare_fds, &fds);
     762           0 :                 if (ret)
     763             :                         return ret;
     764             : 
     765             :                 /*
     766             :                  * We used to share our file descriptor table, and have now
     767             :                  * created a private one, make sure we're using it below.
     768             :                  */
     769           0 :                 if (fds)
     770           0 :                         swap(cur_fds, fds);
     771             :         }
     772             : 
     773           0 :         if (flags & CLOSE_RANGE_CLOEXEC)
     774           0 :                 __range_cloexec(cur_fds, fd, max_fd);
     775             :         else
     776           0 :                 __range_close(cur_fds, fd, max_fd);
     777             : 
     778           0 :         if (fds) {
     779             :                 /*
     780             :                  * We're done closing the files we were supposed to. Time to install
     781             :                  * the new file descriptor table and drop the old one.
     782             :                  */
     783           0 :                 task_lock(me);
     784           0 :                 me->files = cur_fds;
     785           0 :                 task_unlock(me);
     786           0 :                 put_files_struct(fds);
     787             :         }
     788             : 
     789             :         return 0;
     790             : }
     791             : 
     792             : /*
     793             :  * See close_fd_get_file() below, this variant assumes current->files->file_lock
     794             :  * is held.
     795             :  */
     796           0 : struct file *__close_fd_get_file(unsigned int fd)
     797             : {
     798           0 :         return pick_file(current->files, fd);
     799             : }
     800             : 
     801             : /*
     802             :  * variant of close_fd that gets a ref on the file for later fput.
     803             :  * The caller must ensure that filp_close() called on the file.
     804             :  */
     805           0 : struct file *close_fd_get_file(unsigned int fd)
     806             : {
     807           0 :         struct files_struct *files = current->files;
     808             :         struct file *file;
     809             : 
     810           0 :         spin_lock(&files->file_lock);
     811           0 :         file = pick_file(files, fd);
     812           0 :         spin_unlock(&files->file_lock);
     813             : 
     814           0 :         return file;
     815             : }
     816             : 
     817           0 : void do_close_on_exec(struct files_struct *files)
     818             : {
     819             :         unsigned i;
     820             :         struct fdtable *fdt;
     821             : 
     822             :         /* exec unshares first */
     823           0 :         spin_lock(&files->file_lock);
     824           0 :         for (i = 0; ; i++) {
     825             :                 unsigned long set;
     826           0 :                 unsigned fd = i * BITS_PER_LONG;
     827           0 :                 fdt = files_fdtable(files);
     828           0 :                 if (fd >= fdt->max_fds)
     829             :                         break;
     830           0 :                 set = fdt->close_on_exec[i];
     831           0 :                 if (!set)
     832           0 :                         continue;
     833           0 :                 fdt->close_on_exec[i] = 0;
     834           0 :                 for ( ; set ; fd++, set >>= 1) {
     835             :                         struct file *file;
     836           0 :                         if (!(set & 1))
     837           0 :                                 continue;
     838           0 :                         file = fdt->fd[fd];
     839           0 :                         if (!file)
     840           0 :                                 continue;
     841           0 :                         rcu_assign_pointer(fdt->fd[fd], NULL);
     842           0 :                         __put_unused_fd(files, fd);
     843           0 :                         spin_unlock(&files->file_lock);
     844           0 :                         filp_close(file, files);
     845           0 :                         cond_resched();
     846           0 :                         spin_lock(&files->file_lock);
     847             :                 }
     848             : 
     849             :         }
     850           0 :         spin_unlock(&files->file_lock);
     851           0 : }
     852             : 
     853           0 : static inline struct file *__fget_files_rcu(struct files_struct *files,
     854             :         unsigned int fd, fmode_t mask)
     855             : {
     856             :         for (;;) {
     857             :                 struct file *file;
     858           0 :                 struct fdtable *fdt = rcu_dereference_raw(files->fdt);
     859             :                 struct file __rcu **fdentry;
     860             : 
     861           0 :                 if (unlikely(fd >= fdt->max_fds))
     862             :                         return NULL;
     863             : 
     864           0 :                 fdentry = fdt->fd + array_index_nospec(fd, fdt->max_fds);
     865           0 :                 file = rcu_dereference_raw(*fdentry);
     866           0 :                 if (unlikely(!file))
     867             :                         return NULL;
     868             : 
     869           0 :                 if (unlikely(file->f_mode & mask))
     870             :                         return NULL;
     871             : 
     872             :                 /*
     873             :                  * Ok, we have a file pointer. However, because we do
     874             :                  * this all locklessly under RCU, we may be racing with
     875             :                  * that file being closed.
     876             :                  *
     877             :                  * Such a race can take two forms:
     878             :                  *
     879             :                  *  (a) the file ref already went down to zero,
     880             :                  *      and get_file_rcu() fails. Just try again:
     881             :                  */
     882           0 :                 if (unlikely(!get_file_rcu(file)))
     883           0 :                         continue;
     884             : 
     885             :                 /*
     886             :                  *  (b) the file table entry has changed under us.
     887             :                  *       Note that we don't need to re-check the 'fdt->fd'
     888             :                  *       pointer having changed, because it always goes
     889             :                  *       hand-in-hand with 'fdt'.
     890             :                  *
     891             :                  * If so, we need to put our ref and try again.
     892             :                  */
     893           0 :                 if (unlikely(rcu_dereference_raw(files->fdt) != fdt) ||
     894           0 :                     unlikely(rcu_dereference_raw(*fdentry) != file)) {
     895           0 :                         fput(file);
     896           0 :                         continue;
     897             :                 }
     898             : 
     899             :                 /*
     900             :                  * Ok, we have a ref to the file, and checked that it
     901             :                  * still exists.
     902             :                  */
     903             :                 return file;
     904             :         }
     905             : }
     906             : 
     907             : static struct file *__fget_files(struct files_struct *files, unsigned int fd,
     908             :                                  fmode_t mask)
     909             : {
     910             :         struct file *file;
     911             : 
     912             :         rcu_read_lock();
     913           0 :         file = __fget_files_rcu(files, fd, mask);
     914             :         rcu_read_unlock();
     915             : 
     916             :         return file;
     917             : }
     918             : 
     919             : static inline struct file *__fget(unsigned int fd, fmode_t mask)
     920             : {
     921           0 :         return __fget_files(current->files, fd, mask);
     922             : }
     923             : 
     924           0 : struct file *fget(unsigned int fd)
     925             : {
     926           0 :         return __fget(fd, FMODE_PATH);
     927             : }
     928             : EXPORT_SYMBOL(fget);
     929             : 
     930           0 : struct file *fget_raw(unsigned int fd)
     931             : {
     932           0 :         return __fget(fd, 0);
     933             : }
     934             : EXPORT_SYMBOL(fget_raw);
     935             : 
     936           0 : struct file *fget_task(struct task_struct *task, unsigned int fd)
     937             : {
     938           0 :         struct file *file = NULL;
     939             : 
     940           0 :         task_lock(task);
     941           0 :         if (task->files)
     942           0 :                 file = __fget_files(task->files, fd, 0);
     943           0 :         task_unlock(task);
     944             : 
     945           0 :         return file;
     946             : }
     947             : 
     948           0 : struct file *task_lookup_fd_rcu(struct task_struct *task, unsigned int fd)
     949             : {
     950             :         /* Must be called with rcu_read_lock held */
     951             :         struct files_struct *files;
     952           0 :         struct file *file = NULL;
     953             : 
     954           0 :         task_lock(task);
     955           0 :         files = task->files;
     956           0 :         if (files)
     957             :                 file = files_lookup_fd_rcu(files, fd);
     958           0 :         task_unlock(task);
     959             : 
     960           0 :         return file;
     961             : }
     962             : 
     963           0 : struct file *task_lookup_next_fd_rcu(struct task_struct *task, unsigned int *ret_fd)
     964             : {
     965             :         /* Must be called with rcu_read_lock held */
     966             :         struct files_struct *files;
     967           0 :         unsigned int fd = *ret_fd;
     968           0 :         struct file *file = NULL;
     969             : 
     970           0 :         task_lock(task);
     971           0 :         files = task->files;
     972           0 :         if (files) {
     973           0 :                 for (; fd < files_fdtable(files)->max_fds; fd++) {
     974           0 :                         file = files_lookup_fd_rcu(files, fd);
     975           0 :                         if (file)
     976             :                                 break;
     977             :                 }
     978             :         }
     979           0 :         task_unlock(task);
     980           0 :         *ret_fd = fd;
     981           0 :         return file;
     982             : }
     983             : EXPORT_SYMBOL(task_lookup_next_fd_rcu);
     984             : 
     985             : /*
     986             :  * Lightweight file lookup - no refcnt increment if fd table isn't shared.
     987             :  *
     988             :  * You can use this instead of fget if you satisfy all of the following
     989             :  * conditions:
     990             :  * 1) You must call fput_light before exiting the syscall and returning control
     991             :  *    to userspace (i.e. you cannot remember the returned struct file * after
     992             :  *    returning to userspace).
     993             :  * 2) You must not call filp_close on the returned struct file * in between
     994             :  *    calls to fget_light and fput_light.
     995             :  * 3) You must not clone the current task in between the calls to fget_light
     996             :  *    and fput_light.
     997             :  *
     998             :  * The fput_needed flag returned by fget_light should be passed to the
     999             :  * corresponding fput_light.
    1000             :  */
    1001           0 : static unsigned long __fget_light(unsigned int fd, fmode_t mask)
    1002             : {
    1003           0 :         struct files_struct *files = current->files;
    1004             :         struct file *file;
    1005             : 
    1006             :         /*
    1007             :          * If another thread is concurrently calling close_fd() followed
    1008             :          * by put_files_struct(), we must not observe the old table
    1009             :          * entry combined with the new refcount - otherwise we could
    1010             :          * return a file that is concurrently being freed.
    1011             :          *
    1012             :          * atomic_read_acquire() pairs with atomic_dec_and_test() in
    1013             :          * put_files_struct().
    1014             :          */
    1015           0 :         if (atomic_read_acquire(&files->count) == 1) {
    1016           0 :                 file = files_lookup_fd_raw(files, fd);
    1017           0 :                 if (!file || unlikely(file->f_mode & mask))
    1018             :                         return 0;
    1019           0 :                 return (unsigned long)file;
    1020             :         } else {
    1021           0 :                 file = __fget(fd, mask);
    1022           0 :                 if (!file)
    1023             :                         return 0;
    1024           0 :                 return FDPUT_FPUT | (unsigned long)file;
    1025             :         }
    1026             : }
    1027           0 : unsigned long __fdget(unsigned int fd)
    1028             : {
    1029           0 :         return __fget_light(fd, FMODE_PATH);
    1030             : }
    1031             : EXPORT_SYMBOL(__fdget);
    1032             : 
    1033           0 : unsigned long __fdget_raw(unsigned int fd)
    1034             : {
    1035           0 :         return __fget_light(fd, 0);
    1036             : }
    1037             : 
    1038           0 : unsigned long __fdget_pos(unsigned int fd)
    1039             : {
    1040           0 :         unsigned long v = __fdget(fd);
    1041           0 :         struct file *file = (struct file *)(v & ~3);
    1042             : 
    1043           0 :         if (file && (file->f_mode & FMODE_ATOMIC_POS)) {
    1044           0 :                 if (file_count(file) > 1) {
    1045           0 :                         v |= FDPUT_POS_UNLOCK;
    1046           0 :                         mutex_lock(&file->f_pos_lock);
    1047             :                 }
    1048             :         }
    1049           0 :         return v;
    1050             : }
    1051             : 
    1052           0 : void __f_unlock_pos(struct file *f)
    1053             : {
    1054           0 :         mutex_unlock(&f->f_pos_lock);
    1055           0 : }
    1056             : 
    1057             : /*
    1058             :  * We only lock f_pos if we have threads or if the file might be
    1059             :  * shared with another process. In both cases we'll have an elevated
    1060             :  * file count (done either by fdget() or by fork()).
    1061             :  */
    1062             : 
    1063           0 : void set_close_on_exec(unsigned int fd, int flag)
    1064             : {
    1065           0 :         struct files_struct *files = current->files;
    1066             :         struct fdtable *fdt;
    1067           0 :         spin_lock(&files->file_lock);
    1068           0 :         fdt = files_fdtable(files);
    1069           0 :         if (flag)
    1070           0 :                 __set_close_on_exec(fd, fdt);
    1071             :         else
    1072           0 :                 __clear_close_on_exec(fd, fdt);
    1073           0 :         spin_unlock(&files->file_lock);
    1074           0 : }
    1075             : 
    1076           0 : bool get_close_on_exec(unsigned int fd)
    1077             : {
    1078           0 :         struct files_struct *files = current->files;
    1079             :         struct fdtable *fdt;
    1080             :         bool res;
    1081             :         rcu_read_lock();
    1082           0 :         fdt = files_fdtable(files);
    1083           0 :         res = close_on_exec(fd, fdt);
    1084             :         rcu_read_unlock();
    1085           0 :         return res;
    1086             : }
    1087             : 
    1088           0 : static int do_dup2(struct files_struct *files,
    1089             :         struct file *file, unsigned fd, unsigned flags)
    1090             : __releases(&files->file_lock)
    1091             : {
    1092             :         struct file *tofree;
    1093             :         struct fdtable *fdt;
    1094             : 
    1095             :         /*
    1096             :          * We need to detect attempts to do dup2() over allocated but still
    1097             :          * not finished descriptor.  NB: OpenBSD avoids that at the price of
    1098             :          * extra work in their equivalent of fget() - they insert struct
    1099             :          * file immediately after grabbing descriptor, mark it larval if
    1100             :          * more work (e.g. actual opening) is needed and make sure that
    1101             :          * fget() treats larval files as absent.  Potentially interesting,
    1102             :          * but while extra work in fget() is trivial, locking implications
    1103             :          * and amount of surgery on open()-related paths in VFS are not.
    1104             :          * FreeBSD fails with -EBADF in the same situation, NetBSD "solution"
    1105             :          * deadlocks in rather amusing ways, AFAICS.  All of that is out of
    1106             :          * scope of POSIX or SUS, since neither considers shared descriptor
    1107             :          * tables and this condition does not arise without those.
    1108             :          */
    1109           0 :         fdt = files_fdtable(files);
    1110           0 :         tofree = fdt->fd[fd];
    1111           0 :         if (!tofree && fd_is_open(fd, fdt))
    1112             :                 goto Ebusy;
    1113           0 :         get_file(file);
    1114           0 :         rcu_assign_pointer(fdt->fd[fd], file);
    1115           0 :         __set_open_fd(fd, fdt);
    1116           0 :         if (flags & O_CLOEXEC)
    1117           0 :                 __set_close_on_exec(fd, fdt);
    1118             :         else
    1119           0 :                 __clear_close_on_exec(fd, fdt);
    1120           0 :         spin_unlock(&files->file_lock);
    1121             : 
    1122           0 :         if (tofree)
    1123           0 :                 filp_close(tofree, files);
    1124             : 
    1125           0 :         return fd;
    1126             : 
    1127             : Ebusy:
    1128           0 :         spin_unlock(&files->file_lock);
    1129           0 :         return -EBUSY;
    1130             : }
    1131             : 
    1132           0 : int replace_fd(unsigned fd, struct file *file, unsigned flags)
    1133             : {
    1134             :         int err;
    1135           0 :         struct files_struct *files = current->files;
    1136             : 
    1137           0 :         if (!file)
    1138           0 :                 return close_fd(fd);
    1139             : 
    1140           0 :         if (fd >= rlimit(RLIMIT_NOFILE))
    1141             :                 return -EBADF;
    1142             : 
    1143           0 :         spin_lock(&files->file_lock);
    1144           0 :         err = expand_files(files, fd);
    1145           0 :         if (unlikely(err < 0))
    1146             :                 goto out_unlock;
    1147           0 :         return do_dup2(files, file, fd, flags);
    1148             : 
    1149             : out_unlock:
    1150           0 :         spin_unlock(&files->file_lock);
    1151           0 :         return err;
    1152             : }
    1153             : 
    1154             : /**
    1155             :  * __receive_fd() - Install received file into file descriptor table
    1156             :  * @file: struct file that was received from another process
    1157             :  * @ufd: __user pointer to write new fd number to
    1158             :  * @o_flags: the O_* flags to apply to the new fd entry
    1159             :  *
    1160             :  * Installs a received file into the file descriptor table, with appropriate
    1161             :  * checks and count updates. Optionally writes the fd number to userspace, if
    1162             :  * @ufd is non-NULL.
    1163             :  *
    1164             :  * This helper handles its own reference counting of the incoming
    1165             :  * struct file.
    1166             :  *
    1167             :  * Returns newly install fd or -ve on error.
    1168             :  */
    1169           0 : int __receive_fd(struct file *file, int __user *ufd, unsigned int o_flags)
    1170             : {
    1171             :         int new_fd;
    1172             :         int error;
    1173             : 
    1174           0 :         error = security_file_receive(file);
    1175             :         if (error)
    1176             :                 return error;
    1177             : 
    1178           0 :         new_fd = get_unused_fd_flags(o_flags);
    1179           0 :         if (new_fd < 0)
    1180             :                 return new_fd;
    1181             : 
    1182           0 :         if (ufd) {
    1183           0 :                 error = put_user(new_fd, ufd);
    1184           0 :                 if (error) {
    1185           0 :                         put_unused_fd(new_fd);
    1186           0 :                         return error;
    1187             :                 }
    1188             :         }
    1189             : 
    1190           0 :         fd_install(new_fd, get_file(file));
    1191           0 :         __receive_sock(file);
    1192           0 :         return new_fd;
    1193             : }
    1194             : 
    1195           0 : int receive_fd_replace(int new_fd, struct file *file, unsigned int o_flags)
    1196             : {
    1197             :         int error;
    1198             : 
    1199           0 :         error = security_file_receive(file);
    1200             :         if (error)
    1201             :                 return error;
    1202           0 :         error = replace_fd(new_fd, file, o_flags);
    1203           0 :         if (error)
    1204             :                 return error;
    1205           0 :         __receive_sock(file);
    1206           0 :         return new_fd;
    1207             : }
    1208             : 
    1209           0 : int receive_fd(struct file *file, unsigned int o_flags)
    1210             : {
    1211           0 :         return __receive_fd(file, NULL, o_flags);
    1212             : }
    1213             : EXPORT_SYMBOL_GPL(receive_fd);
    1214             : 
    1215           0 : static int ksys_dup3(unsigned int oldfd, unsigned int newfd, int flags)
    1216             : {
    1217           0 :         int err = -EBADF;
    1218             :         struct file *file;
    1219           0 :         struct files_struct *files = current->files;
    1220             : 
    1221           0 :         if ((flags & ~O_CLOEXEC) != 0)
    1222             :                 return -EINVAL;
    1223             : 
    1224           0 :         if (unlikely(oldfd == newfd))
    1225             :                 return -EINVAL;
    1226             : 
    1227           0 :         if (newfd >= rlimit(RLIMIT_NOFILE))
    1228             :                 return -EBADF;
    1229             : 
    1230           0 :         spin_lock(&files->file_lock);
    1231           0 :         err = expand_files(files, newfd);
    1232           0 :         file = files_lookup_fd_locked(files, oldfd);
    1233           0 :         if (unlikely(!file))
    1234             :                 goto Ebadf;
    1235           0 :         if (unlikely(err < 0)) {
    1236           0 :                 if (err == -EMFILE)
    1237             :                         goto Ebadf;
    1238             :                 goto out_unlock;
    1239             :         }
    1240           0 :         return do_dup2(files, file, newfd, flags);
    1241             : 
    1242             : Ebadf:
    1243           0 :         err = -EBADF;
    1244             : out_unlock:
    1245           0 :         spin_unlock(&files->file_lock);
    1246           0 :         return err;
    1247             : }
    1248             : 
    1249           0 : SYSCALL_DEFINE3(dup3, unsigned int, oldfd, unsigned int, newfd, int, flags)
    1250             : {
    1251           0 :         return ksys_dup3(oldfd, newfd, flags);
    1252             : }
    1253             : 
    1254           0 : SYSCALL_DEFINE2(dup2, unsigned int, oldfd, unsigned int, newfd)
    1255             : {
    1256           0 :         if (unlikely(newfd == oldfd)) { /* corner case */
    1257           0 :                 struct files_struct *files = current->files;
    1258           0 :                 int retval = oldfd;
    1259             : 
    1260             :                 rcu_read_lock();
    1261           0 :                 if (!files_lookup_fd_rcu(files, oldfd))
    1262           0 :                         retval = -EBADF;
    1263             :                 rcu_read_unlock();
    1264           0 :                 return retval;
    1265             :         }
    1266           0 :         return ksys_dup3(oldfd, newfd, 0);
    1267             : }
    1268             : 
    1269           0 : SYSCALL_DEFINE1(dup, unsigned int, fildes)
    1270             : {
    1271           0 :         int ret = -EBADF;
    1272           0 :         struct file *file = fget_raw(fildes);
    1273             : 
    1274           0 :         if (file) {
    1275           0 :                 ret = get_unused_fd_flags(0);
    1276           0 :                 if (ret >= 0)
    1277           0 :                         fd_install(ret, file);
    1278             :                 else
    1279           0 :                         fput(file);
    1280             :         }
    1281           0 :         return ret;
    1282             : }
    1283             : 
    1284           0 : int f_dupfd(unsigned int from, struct file *file, unsigned flags)
    1285             : {
    1286           0 :         unsigned long nofile = rlimit(RLIMIT_NOFILE);
    1287             :         int err;
    1288           0 :         if (from >= nofile)
    1289             :                 return -EINVAL;
    1290           0 :         err = alloc_fd(from, nofile, flags);
    1291           0 :         if (err >= 0) {
    1292           0 :                 get_file(file);
    1293           0 :                 fd_install(err, file);
    1294             :         }
    1295             :         return err;
    1296             : }
    1297             : 
    1298           0 : int iterate_fd(struct files_struct *files, unsigned n,
    1299             :                 int (*f)(const void *, struct file *, unsigned),
    1300             :                 const void *p)
    1301             : {
    1302             :         struct fdtable *fdt;
    1303           0 :         int res = 0;
    1304           0 :         if (!files)
    1305             :                 return 0;
    1306           0 :         spin_lock(&files->file_lock);
    1307           0 :         for (fdt = files_fdtable(files); n < fdt->max_fds; n++) {
    1308             :                 struct file *file;
    1309           0 :                 file = rcu_dereference_check_fdtable(files, fdt->fd[n]);
    1310           0 :                 if (!file)
    1311           0 :                         continue;
    1312           0 :                 res = f(p, file, n);
    1313           0 :                 if (res)
    1314             :                         break;
    1315             :         }
    1316           0 :         spin_unlock(&files->file_lock);
    1317           0 :         return res;
    1318             : }
    1319             : EXPORT_SYMBOL(iterate_fd);

Generated by: LCOV version 1.14