~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/fs/ext4/file.c

Version: ~ [ linux-5.10-rc6 ] ~ [ linux-5.9.12 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.81 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.161 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.210 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.247 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.247 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.85 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /*
  2  *  linux/fs/ext4/file.c
  3  *
  4  * Copyright (C) 1992, 1993, 1994, 1995
  5  * Remy Card (card@masi.ibp.fr)
  6  * Laboratoire MASI - Institut Blaise Pascal
  7  * Universite Pierre et Marie Curie (Paris VI)
  8  *
  9  *  from
 10  *
 11  *  linux/fs/minix/file.c
 12  *
 13  *  Copyright (C) 1991, 1992  Linus Torvalds
 14  *
 15  *  ext4 fs regular file handling primitives
 16  *
 17  *  64-bit file support on 64-bit platforms by Jakub Jelinek
 18  *      (jj@sunsite.ms.mff.cuni.cz)
 19  */
 20 
 21 #include <linux/time.h>
 22 #include <linux/fs.h>
 23 #include <linux/mount.h>
 24 #include <linux/path.h>
 25 #include <linux/dax.h>
 26 #include <linux/quotaops.h>
 27 #include <linux/pagevec.h>
 28 #include <linux/uio.h>
 29 #include "ext4.h"
 30 #include "ext4_jbd2.h"
 31 #include "xattr.h"
 32 #include "acl.h"
 33 
 34 #ifdef CONFIG_FS_DAX
 35 static ssize_t ext4_dax_read_iter(struct kiocb *iocb, struct iov_iter *to)
 36 {
 37         struct inode *inode = file_inode(iocb->ki_filp);
 38         ssize_t ret;
 39 
 40         inode_lock_shared(inode);
 41         /*
 42          * Recheck under inode lock - at this point we are sure it cannot
 43          * change anymore
 44          */
 45         if (!IS_DAX(inode)) {
 46                 inode_unlock_shared(inode);
 47                 /* Fallback to buffered IO in case we cannot support DAX */
 48                 return generic_file_read_iter(iocb, to);
 49         }
 50         ret = dax_iomap_rw(iocb, to, &ext4_iomap_ops);
 51         inode_unlock_shared(inode);
 52 
 53         file_accessed(iocb->ki_filp);
 54         return ret;
 55 }
 56 #endif
 57 
 58 static ssize_t ext4_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
 59 {
 60         if (unlikely(ext4_forced_shutdown(EXT4_SB(file_inode(iocb->ki_filp)->i_sb))))
 61                 return -EIO;
 62 
 63         if (!iov_iter_count(to))
 64                 return 0; /* skip atime */
 65 
 66 #ifdef CONFIG_FS_DAX
 67         if (IS_DAX(file_inode(iocb->ki_filp)))
 68                 return ext4_dax_read_iter(iocb, to);
 69 #endif
 70         return generic_file_read_iter(iocb, to);
 71 }
 72 
 73 /*
 74  * Called when an inode is released. Note that this is different
 75  * from ext4_file_open: open gets called at every open, but release
 76  * gets called only when /all/ the files are closed.
 77  */
 78 static int ext4_release_file(struct inode *inode, struct file *filp)
 79 {
 80         if (ext4_test_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE)) {
 81                 ext4_alloc_da_blocks(inode);
 82                 ext4_clear_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE);
 83         }
 84         /* if we are the last writer on the inode, drop the block reservation */
 85         if ((filp->f_mode & FMODE_WRITE) &&
 86                         (atomic_read(&inode->i_writecount) == 1) &&
 87                         !EXT4_I(inode)->i_reserved_data_blocks)
 88         {
 89                 down_write(&EXT4_I(inode)->i_data_sem);
 90                 ext4_discard_preallocations(inode);
 91                 up_write(&EXT4_I(inode)->i_data_sem);
 92         }
 93         if (is_dx(inode) && filp->private_data)
 94                 ext4_htree_free_dir_info(filp->private_data);
 95 
 96         return 0;
 97 }
 98 
 99 static void ext4_unwritten_wait(struct inode *inode)
100 {
101         wait_queue_head_t *wq = ext4_ioend_wq(inode);
102 
103         wait_event(*wq, (atomic_read(&EXT4_I(inode)->i_unwritten) == 0));
104 }
105 
106 /*
107  * This tests whether the IO in question is block-aligned or not.
108  * Ext4 utilizes unwritten extents when hole-filling during direct IO, and they
109  * are converted to written only after the IO is complete.  Until they are
110  * mapped, these blocks appear as holes, so dio_zero_block() will assume that
111  * it needs to zero out portions of the start and/or end block.  If 2 AIO
112  * threads are at work on the same unwritten block, they must be synchronized
113  * or one thread will zero the other's data, causing corruption.
114  */
115 static int
116 ext4_unaligned_aio(struct inode *inode, struct iov_iter *from, loff_t pos)
117 {
118         struct super_block *sb = inode->i_sb;
119         int blockmask = sb->s_blocksize - 1;
120 
121         if (pos >= i_size_read(inode))
122                 return 0;
123 
124         if ((pos | iov_iter_alignment(from)) & blockmask)
125                 return 1;
126 
127         return 0;
128 }
129 
130 /* Is IO overwriting allocated and initialized blocks? */
131 static bool ext4_overwrite_io(struct inode *inode, loff_t pos, loff_t len)
132 {
133         struct ext4_map_blocks map;
134         unsigned int blkbits = inode->i_blkbits;
135         int err, blklen;
136 
137         if (pos + len > i_size_read(inode))
138                 return false;
139 
140         map.m_lblk = pos >> blkbits;
141         map.m_len = EXT4_MAX_BLOCKS(len, pos, blkbits);
142         blklen = map.m_len;
143 
144         err = ext4_map_blocks(NULL, inode, &map, 0);
145         /*
146          * 'err==len' means that all of the blocks have been preallocated,
147          * regardless of whether they have been initialized or not. To exclude
148          * unwritten extents, we need to check m_flags.
149          */
150         return err == blklen && (map.m_flags & EXT4_MAP_MAPPED);
151 }
152 
153 static ssize_t ext4_write_checks(struct kiocb *iocb, struct iov_iter *from)
154 {
155         struct inode *inode = file_inode(iocb->ki_filp);
156         ssize_t ret;
157 
158         ret = generic_write_checks(iocb, from);
159         if (ret <= 0)
160                 return ret;
161         /*
162          * If we have encountered a bitmap-format file, the size limit
163          * is smaller than s_maxbytes, which is for extent-mapped files.
164          */
165         if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) {
166                 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
167 
168                 if (iocb->ki_pos >= sbi->s_bitmap_maxbytes)
169                         return -EFBIG;
170                 iov_iter_truncate(from, sbi->s_bitmap_maxbytes - iocb->ki_pos);
171         }
172         return iov_iter_count(from);
173 }
174 
175 #ifdef CONFIG_FS_DAX
176 static ssize_t
177 ext4_dax_write_iter(struct kiocb *iocb, struct iov_iter *from)
178 {
179         struct inode *inode = file_inode(iocb->ki_filp);
180         ssize_t ret;
181 
182         inode_lock(inode);
183         ret = ext4_write_checks(iocb, from);
184         if (ret <= 0)
185                 goto out;
186         ret = file_remove_privs(iocb->ki_filp);
187         if (ret)
188                 goto out;
189         ret = file_update_time(iocb->ki_filp);
190         if (ret)
191                 goto out;
192 
193         ret = dax_iomap_rw(iocb, from, &ext4_iomap_ops);
194 out:
195         inode_unlock(inode);
196         if (ret > 0)
197                 ret = generic_write_sync(iocb, ret);
198         return ret;
199 }
200 #endif
201 
202 static ssize_t
203 ext4_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
204 {
205         struct inode *inode = file_inode(iocb->ki_filp);
206         int o_direct = iocb->ki_flags & IOCB_DIRECT;
207         int unaligned_aio = 0;
208         int overwrite = 0;
209         ssize_t ret;
210 
211         if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
212                 return -EIO;
213 
214 #ifdef CONFIG_FS_DAX
215         if (IS_DAX(inode))
216                 return ext4_dax_write_iter(iocb, from);
217 #endif
218 
219         inode_lock(inode);
220         ret = ext4_write_checks(iocb, from);
221         if (ret <= 0)
222                 goto out;
223 
224         /*
225          * Unaligned direct AIO must be serialized among each other as zeroing
226          * of partial blocks of two competing unaligned AIOs can result in data
227          * corruption.
228          */
229         if (o_direct && ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS) &&
230             !is_sync_kiocb(iocb) &&
231             ext4_unaligned_aio(inode, from, iocb->ki_pos)) {
232                 unaligned_aio = 1;
233                 ext4_unwritten_wait(inode);
234         }
235 
236         iocb->private = &overwrite;
237         /* Check whether we do a DIO overwrite or not */
238         if (o_direct && ext4_should_dioread_nolock(inode) && !unaligned_aio &&
239             ext4_overwrite_io(inode, iocb->ki_pos, iov_iter_count(from)))
240                 overwrite = 1;
241 
242         ret = __generic_file_write_iter(iocb, from);
243         inode_unlock(inode);
244 
245         if (ret > 0)
246                 ret = generic_write_sync(iocb, ret);
247 
248         return ret;
249 
250 out:
251         inode_unlock(inode);
252         return ret;
253 }
254 
255 #ifdef CONFIG_FS_DAX
256 static int ext4_dax_huge_fault(struct vm_fault *vmf,
257                 enum page_entry_size pe_size)
258 {
259         int result;
260         handle_t *handle = NULL;
261         struct inode *inode = file_inode(vmf->vma->vm_file);
262         struct super_block *sb = inode->i_sb;
263         bool write = vmf->flags & FAULT_FLAG_WRITE;
264 
265         if (write) {
266                 sb_start_pagefault(sb);
267                 file_update_time(vmf->vma->vm_file);
268                 down_read(&EXT4_I(inode)->i_mmap_sem);
269                 handle = ext4_journal_start_sb(sb, EXT4_HT_WRITE_PAGE,
270                                                EXT4_DATA_TRANS_BLOCKS(sb));
271         } else {
272                 down_read(&EXT4_I(inode)->i_mmap_sem);
273         }
274         if (!IS_ERR(handle))
275                 result = dax_iomap_fault(vmf, pe_size, &ext4_iomap_ops);
276         else
277                 result = VM_FAULT_SIGBUS;
278         if (write) {
279                 if (!IS_ERR(handle))
280                         ext4_journal_stop(handle);
281                 up_read(&EXT4_I(inode)->i_mmap_sem);
282                 sb_end_pagefault(sb);
283         } else {
284                 up_read(&EXT4_I(inode)->i_mmap_sem);
285         }
286 
287         return result;
288 }
289 
290 static int ext4_dax_fault(struct vm_fault *vmf)
291 {
292         return ext4_dax_huge_fault(vmf, PE_SIZE_PTE);
293 }
294 
295 /*
296  * Handle write fault for VM_MIXEDMAP mappings. Similarly to ext4_dax_fault()
297  * handler we check for races agaist truncate. Note that since we cycle through
298  * i_mmap_sem, we are sure that also any hole punching that began before we
299  * were called is finished by now and so if it included part of the file we
300  * are working on, our pte will get unmapped and the check for pte_same() in
301  * wp_pfn_shared() fails. Thus fault gets retried and things work out as
302  * desired.
303  */
304 static int ext4_dax_pfn_mkwrite(struct vm_fault *vmf)
305 {
306         struct inode *inode = file_inode(vmf->vma->vm_file);
307         struct super_block *sb = inode->i_sb;
308         loff_t size;
309         int ret;
310 
311         sb_start_pagefault(sb);
312         file_update_time(vmf->vma->vm_file);
313         down_read(&EXT4_I(inode)->i_mmap_sem);
314         size = (i_size_read(inode) + PAGE_SIZE - 1) >> PAGE_SHIFT;
315         if (vmf->pgoff >= size)
316                 ret = VM_FAULT_SIGBUS;
317         else
318                 ret = dax_pfn_mkwrite(vmf);
319         up_read(&EXT4_I(inode)->i_mmap_sem);
320         sb_end_pagefault(sb);
321 
322         return ret;
323 }
324 
325 static const struct vm_operations_struct ext4_dax_vm_ops = {
326         .fault          = ext4_dax_fault,
327         .huge_fault     = ext4_dax_huge_fault,
328         .page_mkwrite   = ext4_dax_fault,
329         .pfn_mkwrite    = ext4_dax_pfn_mkwrite,
330 };
331 #else
332 #define ext4_dax_vm_ops ext4_file_vm_ops
333 #endif
334 
335 static const struct vm_operations_struct ext4_file_vm_ops = {
336         .fault          = ext4_filemap_fault,
337         .map_pages      = filemap_map_pages,
338         .page_mkwrite   = ext4_page_mkwrite,
339 };
340 
341 static int ext4_file_mmap(struct file *file, struct vm_area_struct *vma)
342 {
343         struct inode *inode = file->f_mapping->host;
344 
345         if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
346                 return -EIO;
347 
348         if (ext4_encrypted_inode(inode)) {
349                 int err = fscrypt_get_encryption_info(inode);
350                 if (err)
351                         return 0;
352                 if (!fscrypt_has_encryption_key(inode))
353                         return -ENOKEY;
354         }
355         file_accessed(file);
356         if (IS_DAX(file_inode(file))) {
357                 vma->vm_ops = &ext4_dax_vm_ops;
358                 vma->vm_flags |= VM_MIXEDMAP | VM_HUGEPAGE;
359         } else {
360                 vma->vm_ops = &ext4_file_vm_ops;
361         }
362         return 0;
363 }
364 
365 static int ext4_file_open(struct inode * inode, struct file * filp)
366 {
367         struct super_block *sb = inode->i_sb;
368         struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
369         struct vfsmount *mnt = filp->f_path.mnt;
370         struct dentry *dir;
371         struct path path;
372         char buf[64], *cp;
373         int ret;
374 
375         if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
376                 return -EIO;
377 
378         if (unlikely(!(sbi->s_mount_flags & EXT4_MF_MNTDIR_SAMPLED) &&
379                      !(sb->s_flags & MS_RDONLY))) {
380                 sbi->s_mount_flags |= EXT4_MF_MNTDIR_SAMPLED;
381                 /*
382                  * Sample where the filesystem has been mounted and
383                  * store it in the superblock for sysadmin convenience
384                  * when trying to sort through large numbers of block
385                  * devices or filesystem images.
386                  */
387                 memset(buf, 0, sizeof(buf));
388                 path.mnt = mnt;
389                 path.dentry = mnt->mnt_root;
390                 cp = d_path(&path, buf, sizeof(buf));
391                 if (!IS_ERR(cp)) {
392                         handle_t *handle;
393                         int err;
394 
395                         handle = ext4_journal_start_sb(sb, EXT4_HT_MISC, 1);
396                         if (IS_ERR(handle))
397                                 return PTR_ERR(handle);
398                         BUFFER_TRACE(sbi->s_sbh, "get_write_access");
399                         err = ext4_journal_get_write_access(handle, sbi->s_sbh);
400                         if (err) {
401                                 ext4_journal_stop(handle);
402                                 return err;
403                         }
404                         strlcpy(sbi->s_es->s_last_mounted, cp,
405                                 sizeof(sbi->s_es->s_last_mounted));
406                         ext4_handle_dirty_super(handle, sb);
407                         ext4_journal_stop(handle);
408                 }
409         }
410         if (ext4_encrypted_inode(inode)) {
411                 ret = fscrypt_get_encryption_info(inode);
412                 if (ret)
413                         return -EACCES;
414                 if (!fscrypt_has_encryption_key(inode))
415                         return -ENOKEY;
416         }
417 
418         dir = dget_parent(file_dentry(filp));
419         if (ext4_encrypted_inode(d_inode(dir)) &&
420                         !fscrypt_has_permitted_context(d_inode(dir), inode)) {
421                 ext4_warning(inode->i_sb,
422                              "Inconsistent encryption contexts: %lu/%lu",
423                              (unsigned long) d_inode(dir)->i_ino,
424                              (unsigned long) inode->i_ino);
425                 dput(dir);
426                 return -EPERM;
427         }
428         dput(dir);
429         /*
430          * Set up the jbd2_inode if we are opening the inode for
431          * writing and the journal is present
432          */
433         if (filp->f_mode & FMODE_WRITE) {
434                 ret = ext4_inode_attach_jinode(inode);
435                 if (ret < 0)
436                         return ret;
437         }
438         return dquot_file_open(inode, filp);
439 }
440 
441 /*
442  * Here we use ext4_map_blocks() to get a block mapping for a extent-based
443  * file rather than ext4_ext_walk_space() because we can introduce
444  * SEEK_DATA/SEEK_HOLE for block-mapped and extent-mapped file at the same
445  * function.  When extent status tree has been fully implemented, it will
446  * track all extent status for a file and we can directly use it to
447  * retrieve the offset for SEEK_DATA/SEEK_HOLE.
448  */
449 
450 /*
451  * When we retrieve the offset for SEEK_DATA/SEEK_HOLE, we would need to
452  * lookup page cache to check whether or not there has some data between
453  * [startoff, endoff] because, if this range contains an unwritten extent,
454  * we determine this extent as a data or a hole according to whether the
455  * page cache has data or not.
456  */
457 static int ext4_find_unwritten_pgoff(struct inode *inode,
458                                      int whence,
459                                      ext4_lblk_t end_blk,
460                                      loff_t *offset)
461 {
462         struct pagevec pvec;
463         unsigned int blkbits;
464         pgoff_t index;
465         pgoff_t end;
466         loff_t endoff;
467         loff_t startoff;
468         loff_t lastoff;
469         int found = 0;
470 
471         blkbits = inode->i_sb->s_blocksize_bits;
472         startoff = *offset;
473         lastoff = startoff;
474         endoff = (loff_t)end_blk << blkbits;
475 
476         index = startoff >> PAGE_SHIFT;
477         end = (endoff - 1) >> PAGE_SHIFT;
478 
479         pagevec_init(&pvec, 0);
480         do {
481                 int i, num;
482                 unsigned long nr_pages;
483 
484                 num = min_t(pgoff_t, end - index, PAGEVEC_SIZE - 1) + 1;
485                 nr_pages = pagevec_lookup(&pvec, inode->i_mapping, index,
486                                           (pgoff_t)num);
487                 if (nr_pages == 0)
488                         break;
489 
490                 for (i = 0; i < nr_pages; i++) {
491                         struct page *page = pvec.pages[i];
492                         struct buffer_head *bh, *head;
493 
494                         /*
495                          * If current offset is smaller than the page offset,
496                          * there is a hole at this offset.
497                          */
498                         if (whence == SEEK_HOLE && lastoff < endoff &&
499                             lastoff < page_offset(pvec.pages[i])) {
500                                 found = 1;
501                                 *offset = lastoff;
502                                 goto out;
503                         }
504 
505                         if (page->index > end)
506                                 goto out;
507 
508                         lock_page(page);
509 
510                         if (unlikely(page->mapping != inode->i_mapping)) {
511                                 unlock_page(page);
512                                 continue;
513                         }
514 
515                         if (!page_has_buffers(page)) {
516                                 unlock_page(page);
517                                 continue;
518                         }
519 
520                         if (page_has_buffers(page)) {
521                                 lastoff = page_offset(page);
522                                 bh = head = page_buffers(page);
523                                 do {
524                                         if (lastoff + bh->b_size <= startoff)
525                                                 goto next;
526                                         if (buffer_uptodate(bh) ||
527                                             buffer_unwritten(bh)) {
528                                                 if (whence == SEEK_DATA)
529                                                         found = 1;
530                                         } else {
531                                                 if (whence == SEEK_HOLE)
532                                                         found = 1;
533                                         }
534                                         if (found) {
535                                                 *offset = max_t(loff_t,
536                                                         startoff, lastoff);
537                                                 unlock_page(page);
538                                                 goto out;
539                                         }
540 next:
541                                         lastoff += bh->b_size;
542                                         bh = bh->b_this_page;
543                                 } while (bh != head);
544                         }
545 
546                         lastoff = page_offset(page) + PAGE_SIZE;
547                         unlock_page(page);
548                 }
549 
550                 /* The no. of pages is less than our desired, we are done. */
551                 if (nr_pages < num)
552                         break;
553 
554                 index = pvec.pages[i - 1]->index + 1;
555                 pagevec_release(&pvec);
556         } while (index <= end);
557 
558         if (whence == SEEK_HOLE && lastoff < endoff) {
559                 found = 1;
560                 *offset = lastoff;
561         }
562 out:
563         pagevec_release(&pvec);
564         return found;
565 }
566 
567 /*
568  * ext4_seek_data() retrieves the offset for SEEK_DATA.
569  */
570 static loff_t ext4_seek_data(struct file *file, loff_t offset, loff_t maxsize)
571 {
572         struct inode *inode = file->f_mapping->host;
573         struct extent_status es;
574         ext4_lblk_t start, last, end;
575         loff_t dataoff, isize;
576         int blkbits;
577         int ret;
578 
579         inode_lock(inode);
580 
581         isize = i_size_read(inode);
582         if (offset >= isize) {
583                 inode_unlock(inode);
584                 return -ENXIO;
585         }
586 
587         blkbits = inode->i_sb->s_blocksize_bits;
588         start = offset >> blkbits;
589         last = start;
590         end = isize >> blkbits;
591         dataoff = offset;
592 
593         do {
594                 ret = ext4_get_next_extent(inode, last, end - last + 1, &es);
595                 if (ret <= 0) {
596                         /* No extent found -> no data */
597                         if (ret == 0)
598                                 ret = -ENXIO;
599                         inode_unlock(inode);
600                         return ret;
601                 }
602 
603                 last = es.es_lblk;
604                 if (last != start)
605                         dataoff = (loff_t)last << blkbits;
606                 if (!ext4_es_is_unwritten(&es))
607                         break;
608 
609                 /*
610                  * If there is a unwritten extent at this offset,
611                  * it will be as a data or a hole according to page
612                  * cache that has data or not.
613                  */
614                 if (ext4_find_unwritten_pgoff(inode, SEEK_DATA,
615                                               es.es_lblk + es.es_len, &dataoff))
616                         break;
617                 last += es.es_len;
618                 dataoff = (loff_t)last << blkbits;
619                 cond_resched();
620         } while (last <= end);
621 
622         inode_unlock(inode);
623 
624         if (dataoff > isize)
625                 return -ENXIO;
626 
627         return vfs_setpos(file, dataoff, maxsize);
628 }
629 
630 /*
631  * ext4_seek_hole() retrieves the offset for SEEK_HOLE.
632  */
633 static loff_t ext4_seek_hole(struct file *file, loff_t offset, loff_t maxsize)
634 {
635         struct inode *inode = file->f_mapping->host;
636         struct extent_status es;
637         ext4_lblk_t start, last, end;
638         loff_t holeoff, isize;
639         int blkbits;
640         int ret;
641 
642         inode_lock(inode);
643 
644         isize = i_size_read(inode);
645         if (offset >= isize) {
646                 inode_unlock(inode);
647                 return -ENXIO;
648         }
649 
650         blkbits = inode->i_sb->s_blocksize_bits;
651         start = offset >> blkbits;
652         last = start;
653         end = isize >> blkbits;
654         holeoff = offset;
655 
656         do {
657                 ret = ext4_get_next_extent(inode, last, end - last + 1, &es);
658                 if (ret < 0) {
659                         inode_unlock(inode);
660                         return ret;
661                 }
662                 /* Found a hole? */
663                 if (ret == 0 || es.es_lblk > last) {
664                         if (last != start)
665                                 holeoff = (loff_t)last << blkbits;
666                         break;
667                 }
668                 /*
669                  * If there is a unwritten extent at this offset,
670                  * it will be as a data or a hole according to page
671                  * cache that has data or not.
672                  */
673                 if (ext4_es_is_unwritten(&es) &&
674                     ext4_find_unwritten_pgoff(inode, SEEK_HOLE,
675                                               last + es.es_len, &holeoff))
676                         break;
677 
678                 last += es.es_len;
679                 holeoff = (loff_t)last << blkbits;
680                 cond_resched();
681         } while (last <= end);
682 
683         inode_unlock(inode);
684 
685         if (holeoff > isize)
686                 holeoff = isize;
687 
688         return vfs_setpos(file, holeoff, maxsize);
689 }
690 
691 /*
692  * ext4_llseek() handles both block-mapped and extent-mapped maxbytes values
693  * by calling generic_file_llseek_size() with the appropriate maxbytes
694  * value for each.
695  */
696 loff_t ext4_llseek(struct file *file, loff_t offset, int whence)
697 {
698         struct inode *inode = file->f_mapping->host;
699         loff_t maxbytes;
700 
701         if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
702                 maxbytes = EXT4_SB(inode->i_sb)->s_bitmap_maxbytes;
703         else
704                 maxbytes = inode->i_sb->s_maxbytes;
705 
706         switch (whence) {
707         case SEEK_SET:
708         case SEEK_CUR:
709         case SEEK_END:
710                 return generic_file_llseek_size(file, offset, whence,
711                                                 maxbytes, i_size_read(inode));
712         case SEEK_DATA:
713                 return ext4_seek_data(file, offset, maxbytes);
714         case SEEK_HOLE:
715                 return ext4_seek_hole(file, offset, maxbytes);
716         }
717 
718         return -EINVAL;
719 }
720 
721 const struct file_operations ext4_file_operations = {
722         .llseek         = ext4_llseek,
723         .read_iter      = ext4_file_read_iter,
724         .write_iter     = ext4_file_write_iter,
725         .unlocked_ioctl = ext4_ioctl,
726 #ifdef CONFIG_COMPAT
727         .compat_ioctl   = ext4_compat_ioctl,
728 #endif
729         .mmap           = ext4_file_mmap,
730         .open           = ext4_file_open,
731         .release        = ext4_release_file,
732         .fsync          = ext4_sync_file,
733         .get_unmapped_area = thp_get_unmapped_area,
734         .splice_read    = generic_file_splice_read,
735         .splice_write   = iter_file_splice_write,
736         .fallocate      = ext4_fallocate,
737 };
738 
739 const struct inode_operations ext4_file_inode_operations = {
740         .setattr        = ext4_setattr,
741         .getattr        = ext4_file_getattr,
742         .listxattr      = ext4_listxattr,
743         .get_acl        = ext4_get_acl,
744         .set_acl        = ext4_set_acl,
745         .fiemap         = ext4_fiemap,
746 };
747 
748 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp