1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
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)
11 * linux/fs/minix/dir.c
14 * Copyright (C) 1991, 1992 Linus Torvalds
16 * ext2 directory handling functions
18 * Big-endian to little-endian byte-swapping/bitmaps by
19 * David S. Miller (davem@caip.rutgers.edu), 1995
21 * All code that works with directory layout had been switched to pagecache
24 * Adapted for Lustre Light
25 * Copyright (C) 2002, Cluster File Systems, Inc.
30 #include <linux/ext2_fs.h>
31 #include <linux/pagemap.h>
33 #include <linux/locks.h>
34 #include <asm/uaccess.h>
36 #define DEBUG_SUBSYSTEM S_LLITE
38 #include <linux/obd_support.h>
39 #include <linux/obd_class.h>
40 #include <linux/lustre_lib.h>
41 #include <linux/lustre_idl.h>
42 #include <linux/lustre_mds.h>
43 #include <linux/lustre_lite.h>
45 typedef struct ext2_dir_entry_2 ext2_dirent;
47 #define PageChecked(page) test_bit(PG_checked, &(page)->flags)
48 #define SetPageChecked(page) set_bit(PG_checked, &(page)->flags)
51 static int ll_dir_prepare_write(struct file *file, struct page *page, unsigned from, unsigned to)
56 /* returns the page unlocked, but with a reference */
57 static int ll_dir_readpage(struct file *file, struct page *page)
59 struct inode *inode = page->mapping->host;
60 struct ll_sb_info *sbi = ll_i2sbi(inode);
64 struct ptlrpc_request *request = NULL;
68 if ((inode->i_size + PAGE_CACHE_SIZE - 1) >> PAGE_SHIFT <= page->index){
69 memset(kmap(page), 0, PAGE_CACHE_SIZE);
75 if (Page_Uptodate(page)) {
76 CERROR("Explain this please?\n");
81 offset = page->index << PAGE_SHIFT;
83 rc = mdc_readpage(&sbi->ll_mds_client, sbi->ll_mds_conn, inode->i_ino,
84 S_IFDIR, offset, buf, &request);
86 ptlrpc_free_req(request);
91 SetPageUptodate(page);
95 } /* ll_dir_readpage */
97 struct address_space_operations ll_dir_aops = {
98 readpage: ll_dir_readpage,
99 prepare_write: ll_dir_prepare_write
102 int waitfor_one_page(struct page *page)
105 struct buffer_head *bh, *head = page->buffers;
110 if (buffer_req(bh) && !buffer_uptodate(bh))
112 } while ((bh = bh->b_this_page) != head);
117 * ext2 uses block-sized chunks. Arguably, sector-sized ones would be
118 * more robust, but we have what we have
120 static inline unsigned ext2_chunk_size(struct inode *inode)
122 //return inode->i_sb->s_blocksize;
126 static inline void ext2_put_page(struct page *page)
129 page_cache_release(page);
132 static inline unsigned long dir_pages(struct inode *inode)
134 return (inode->i_size+PAGE_CACHE_SIZE-1)>>PAGE_CACHE_SHIFT;
137 extern void set_page_clean(struct page *page);
139 static int ext2_commit_chunk(struct page *page, unsigned from, unsigned to)
141 struct inode *dir = page->mapping->host;
142 loff_t new_size = (page->index << PAGE_CACHE_SHIFT) + to;
145 dir->i_version = ++event;
146 if (new_size > dir->i_size)
147 dir->i_size = new_size;
148 SetPageUptodate(page);
149 set_page_clean(page);
151 //page->mapping->a_ops->commit_write(NULL, page, from, to);
153 // err = waitfor_one_page(page);
157 static void ext2_check_page(struct page *page)
159 struct inode *dir = page->mapping->host;
160 unsigned chunk_size = ext2_chunk_size(dir);
161 char *kaddr = page_address(page);
162 // u32 max_inumber = le32_to_cpu(sb->u.ext2_sb.s_es->s_inodes_count);
163 unsigned offs, rec_len;
164 unsigned limit = PAGE_CACHE_SIZE;
168 if ((dir->i_size >> PAGE_CACHE_SHIFT) == page->index) {
169 limit = dir->i_size & ~PAGE_CACHE_MASK;
170 if (limit & (chunk_size - 1)) {
171 CERROR("limit %d dir size %lld index %ld\n",
172 limit, dir->i_size, page->index);
175 for (offs = limit; offs<PAGE_CACHE_SIZE; offs += chunk_size) {
176 ext2_dirent *p = (ext2_dirent*)(kaddr + offs);
177 p->rec_len = cpu_to_le16(chunk_size);
182 for (offs = 0; offs <= limit - EXT2_DIR_REC_LEN(1); offs += rec_len) {
183 p = (ext2_dirent *)(kaddr + offs);
184 rec_len = le16_to_cpu(p->rec_len);
186 if (rec_len < EXT2_DIR_REC_LEN(1))
190 if (rec_len < EXT2_DIR_REC_LEN(p->name_len))
192 if (((offs + rec_len - 1) ^ offs) & ~(chunk_size-1))
194 // if (le32_to_cpu(p->inode) > max_inumber)
200 SetPageChecked(page);
203 /* Too bad, we had an error */
206 CERROR("ext2_check_page"
207 "size of directory #%lu is not a multiple of chunk size\n",
212 error = "rec_len is smaller than minimal";
215 error = "unaligned directory entry";
218 error = "rec_len is too small for name_len";
221 error = "directory entry across blocks";
224 // error = "inode out of bounds";
226 CERROR("ext2_check_page: bad entry in directory #%lu: %s - "
227 "offset=%lu, inode=%lu, rec_len=%d, name_len=%d",
228 dir->i_ino, error, (page->index<<PAGE_CACHE_SHIFT)+offs,
229 (unsigned long) le32_to_cpu(p->inode),
230 rec_len, p->name_len);
233 p = (ext2_dirent *)(kaddr + offs);
234 CERROR("ext2_check_page"
235 "entry in directory #%lu spans the page boundary"
236 "offset=%lu, inode=%lu",
237 dir->i_ino, (page->index<<PAGE_CACHE_SHIFT)+offs,
238 (unsigned long) le32_to_cpu(p->inode));
240 SetPageChecked(page);
245 static struct page * ext2_get_page(struct inode *dir, unsigned long n)
247 struct address_space *mapping = dir->i_mapping;
248 struct page *page = read_cache_page(mapping, n,
249 (filler_t*)mapping->a_ops->readpage, NULL);
253 if (!Page_Uptodate(page))
255 if (!PageChecked(page))
256 ext2_check_page(page);
264 return ERR_PTR(-EIO);
268 * NOTE! unlike strncmp, ext2_match returns 1 for success, 0 for failure.
270 * len <= EXT2_NAME_LEN and de != NULL are guaranteed by caller.
272 static inline int ext2_match (int len, const char * const name,
273 struct ext2_dir_entry_2 * de)
275 if (len != de->name_len)
279 return !memcmp(name, de->name, len);
283 * p is at least 6 bytes before the end of page
285 static inline ext2_dirent *ext2_next_entry(ext2_dirent *p)
287 return (ext2_dirent *)((char*)p + le16_to_cpu(p->rec_len));
290 static inline unsigned
291 ext2_validate_entry(char *base, unsigned offset, unsigned mask)
293 ext2_dirent *de = (ext2_dirent*)(base + offset);
294 ext2_dirent *p = (ext2_dirent*)(base + (offset&mask));
295 while ((char*)p < (char*)de)
296 p = ext2_next_entry(p);
297 return (char *)p - base;
300 static unsigned char ext2_filetype_table[EXT2_FT_MAX] = {
301 [EXT2_FT_UNKNOWN] DT_UNKNOWN,
302 [EXT2_FT_REG_FILE] DT_REG,
303 [EXT2_FT_DIR] DT_DIR,
304 [EXT2_FT_CHRDEV] DT_CHR,
305 [EXT2_FT_BLKDEV] DT_BLK,
306 [EXT2_FT_FIFO] DT_FIFO,
307 [EXT2_FT_SOCK] DT_SOCK,
308 [EXT2_FT_SYMLINK] DT_LNK,
311 static unsigned int ll_dt2fmt[DT_WHT + 1] = {
313 [EXT2_FT_REG_FILE] S_IFREG,
314 [EXT2_FT_DIR] S_IFDIR,
315 [EXT2_FT_CHRDEV] S_IFCHR,
316 [EXT2_FT_BLKDEV] S_IFBLK,
317 [EXT2_FT_FIFO] S_IFIFO,
318 [EXT2_FT_SOCK] S_IFSOCK,
319 [EXT2_FT_SYMLINK] S_IFLNK
323 static unsigned char ext2_type_by_mode[S_IFMT >> S_SHIFT] = {
324 [S_IFREG >> S_SHIFT] EXT2_FT_REG_FILE,
325 [S_IFDIR >> S_SHIFT] EXT2_FT_DIR,
326 [S_IFCHR >> S_SHIFT] EXT2_FT_CHRDEV,
327 [S_IFBLK >> S_SHIFT] EXT2_FT_BLKDEV,
328 [S_IFIFO >> S_SHIFT] EXT2_FT_FIFO,
329 [S_IFSOCK >> S_SHIFT] EXT2_FT_SOCK,
330 [S_IFLNK >> S_SHIFT] EXT2_FT_SYMLINK,
333 static inline void ext2_set_de_type(ext2_dirent *de, struct inode *inode)
335 mode_t mode = inode->i_mode;
336 de->file_type = ext2_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
340 new_ll_readdir (struct file * filp, void * dirent, filldir_t filldir)
342 loff_t pos = filp->f_pos;
343 struct inode *inode = filp->f_dentry->d_inode;
344 // XXX struct super_block *sb = inode->i_sb;
345 unsigned offset = pos & ~PAGE_CACHE_MASK;
346 unsigned long n = pos >> PAGE_CACHE_SHIFT;
347 unsigned long npages = dir_pages(inode);
348 unsigned chunk_mask = ~(ext2_chunk_size(inode)-1);
349 unsigned char *types = NULL;
350 int need_revalidate = (filp->f_version != inode->i_version);
352 if (pos > inode->i_size - EXT2_DIR_REC_LEN(1))
355 types = ext2_filetype_table;
357 for ( ; n < npages; n++, offset = 0) {
360 struct page *page = ext2_get_page(inode, n);
364 kaddr = page_address(page);
365 if (need_revalidate) {
366 offset = ext2_validate_entry(kaddr, offset, chunk_mask);
369 de = (ext2_dirent *)(kaddr+offset);
370 limit = kaddr + PAGE_CACHE_SIZE - EXT2_DIR_REC_LEN(1);
371 for ( ;(char*)de <= limit; de = ext2_next_entry(de))
374 unsigned char d_type = DT_UNKNOWN;
376 if (types && de->file_type < EXT2_FT_MAX)
377 d_type = types[de->file_type];
379 offset = (char *)de - kaddr;
380 over = filldir(dirent, de->name, de->name_len,
381 (n<<PAGE_CACHE_SHIFT) | offset,
382 le32_to_cpu(de->inode), d_type);
392 filp->f_pos = (n << PAGE_CACHE_SHIFT) | offset;
393 filp->f_version = inode->i_version;
401 * finds an entry in the specified directory with the wanted name. It
402 * returns the page in which the entry was found, and the entry itself
403 * (as a parameter - res_dir). Page is returned mapped and unlocked.
404 * Entry is guaranteed to be valid.
406 struct ext2_dir_entry_2 * ext2_find_entry (struct inode * dir,
407 struct dentry *dentry, struct page ** res_page)
409 const char *name = dentry->d_name.name;
410 int namelen = dentry->d_name.len;
411 unsigned reclen = EXT2_DIR_REC_LEN(namelen);
412 unsigned long start, n;
413 unsigned long npages = dir_pages(dir);
414 struct page *page = NULL;
420 // start = dir->u.ext2_i.i_dir_start_lookup;
427 page = ext2_get_page(dir, n);
429 kaddr = page_address(page);
430 de = (ext2_dirent *) kaddr;
431 kaddr += PAGE_CACHE_SIZE - reclen;
432 while ((char *) de <= kaddr) {
433 if (ext2_match (namelen, name, de))
435 de = ext2_next_entry(de);
441 } while (n != start);
446 // dir->u.ext2_i.i_dir_start_lookup = n;
450 struct ext2_dir_entry_2 * ext2_dotdot (struct inode *dir, struct page **p)
452 struct page *page = ext2_get_page(dir, 0);
453 ext2_dirent *de = NULL;
456 de = ext2_next_entry((ext2_dirent *) page_address(page));
462 ino_t ll_inode_by_name(struct inode * dir, struct dentry *dentry, int *type)
465 struct ext2_dir_entry_2 * de;
468 de = ext2_find_entry (dir, dentry, &page);
470 res = le32_to_cpu(de->inode);
471 *type = ll_dt2fmt[de->file_type];
473 page_cache_release(page);
478 /* Releases the page */
479 void ext2_set_link(struct inode *dir, struct ext2_dir_entry_2 *de,
480 struct page *page, struct inode *inode)
482 unsigned from = (char *) de - (char *) page_address(page);
483 unsigned to = from + le16_to_cpu(de->rec_len);
487 err = page->mapping->a_ops->prepare_write(NULL, page, from, to);
490 de->inode = cpu_to_le32(inode->i_ino);
491 ext2_set_de_type (de, inode);
492 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
493 err = ext2_commit_chunk(page, from, to);
501 int ll_add_link (struct dentry *dentry, struct inode *inode)
503 struct inode *dir = dentry->d_parent->d_inode;
504 const char *name = dentry->d_name.name;
505 int namelen = dentry->d_name.len;
506 unsigned reclen = EXT2_DIR_REC_LEN(namelen);
507 unsigned short rec_len, name_len;
508 struct page *page = NULL;
510 unsigned long npages = dir_pages(dir);
516 /* We take care of directory expansion in the same loop */
517 for (n = 0; n <= npages; n++) {
518 page = ext2_get_page(dir, n);
522 kaddr = page_address(page);
523 de = (ext2_dirent *)kaddr;
524 kaddr += PAGE_CACHE_SIZE - reclen;
525 while ((char *)de <= kaddr) {
527 if (ext2_match (namelen, name, de))
529 name_len = EXT2_DIR_REC_LEN(de->name_len);
530 rec_len = le16_to_cpu(de->rec_len);
531 if ( n==npages && rec_len == 0) {
532 CERROR("Fatal dir behaviour\n");
535 if (!de->inode && rec_len >= reclen)
537 if (rec_len >= name_len + reclen)
539 de = (ext2_dirent *) ((char *) de + rec_len);
547 from = (char*)de - (char*)page_address(page);
550 err = page->mapping->a_ops->prepare_write(NULL, page, from, to);
554 ext2_dirent *de1 = (ext2_dirent *) ((char *) de + name_len);
555 de1->rec_len = cpu_to_le16(rec_len - name_len);
556 de->rec_len = cpu_to_le16(name_len);
559 de->name_len = namelen;
560 memcpy (de->name, name, namelen);
561 de->inode = cpu_to_le32(inode->i_ino);
562 ext2_set_de_type (de, inode);
563 CDEBUG(D_INODE, "type set to %o\n", de->file_type);
564 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
565 err = ext2_commit_chunk(page, from, to);
567 // change_inode happens with the commit_chunk
568 /* XXX OFFSET_CACHE */
579 * ext2_delete_entry deletes a directory entry by merging it with the
580 * previous entry. Page is up-to-date. Releases the page.
582 int ext2_delete_entry (struct ext2_dir_entry_2 * dir, struct page * page )
584 struct address_space *mapping = page->mapping;
585 struct inode *inode = mapping->host;
586 char *kaddr = page_address(page);
587 unsigned from = ((char*)dir - kaddr) & ~(ext2_chunk_size(inode)-1);
588 unsigned to = ((char*)dir - kaddr) + le16_to_cpu(dir->rec_len);
589 ext2_dirent * pde = NULL;
590 ext2_dirent * de = (ext2_dirent *) (kaddr + from);
593 while ((char*)de < (char*)dir) {
595 de = ext2_next_entry(de);
598 from = (char*)pde - (char*)page_address(page);
600 err = mapping->a_ops->prepare_write(NULL, page, from, to);
604 pde->rec_len = cpu_to_le16(to-from);
606 inode->i_ctime = inode->i_mtime = CURRENT_TIME;
607 err = ext2_commit_chunk(page, from, to);
614 * Set the first fragment of directory.
616 int ext2_make_empty(struct inode *inode, struct inode *parent)
618 struct address_space *mapping = inode->i_mapping;
619 struct page *page = grab_cache_page(mapping, 0);
620 unsigned chunk_size = ext2_chunk_size(inode);
621 struct ext2_dir_entry_2 * de;
628 err = mapping->a_ops->prepare_write(NULL, page, 0, chunk_size);
632 base = page_address(page);
634 de = (struct ext2_dir_entry_2 *) base;
636 de->rec_len = cpu_to_le16(EXT2_DIR_REC_LEN(1));
637 memcpy (de->name, ".\0\0", 4);
638 de->inode = cpu_to_le32(inode->i_ino);
639 ext2_set_de_type (de, inode);
641 de = (struct ext2_dir_entry_2 *) (base + EXT2_DIR_REC_LEN(1));
643 de->rec_len = cpu_to_le16(chunk_size - EXT2_DIR_REC_LEN(1));
644 de->inode = cpu_to_le32(parent->i_ino);
645 memcpy (de->name, "..\0", 4);
646 ext2_set_de_type (de, inode);
648 err = ext2_commit_chunk(page, 0, chunk_size);
651 page_cache_release(page);
657 * routine to check that the specified directory is empty (for rmdir)
659 int ext2_empty_dir (struct inode * inode)
661 struct page *page = NULL;
662 unsigned long i, npages = dir_pages(inode);
664 for (i = 0; i < npages; i++) {
667 page = ext2_get_page(inode, i);
672 kaddr = page_address(page);
673 de = (ext2_dirent *)kaddr;
674 kaddr += PAGE_CACHE_SIZE-EXT2_DIR_REC_LEN(1);
676 while ((char *)de <= kaddr) {
677 if (de->inode != 0) {
678 /* check for . and .. */
679 if (de->name[0] != '.')
681 if (de->name_len > 2)
683 if (de->name_len < 2) {
685 cpu_to_le32(inode->i_ino))
687 } else if (de->name[1] != '.')
690 de = ext2_next_entry(de);
701 struct file_operations ll_dir_operations = {
702 read: generic_read_dir,
703 readdir: new_ll_readdir