11da177e4SLinus Torvalds /*
21da177e4SLinus Torvalds * Copyright 2000 by Hans Reiser, licensing governed by reiserfs/README
31da177e4SLinus Torvalds */
41da177e4SLinus Torvalds
51da177e4SLinus Torvalds #include <linux/time.h>
61da177e4SLinus Torvalds #include <linux/fs.h>
7f466c6fdSAl Viro #include "reiserfs.h"
8a3063ab8SAl Viro #include "acl.h"
9c45ac888SAl Viro #include "xattr.h"
10a5694255SChristoph Hellwig #include <linux/exportfs.h>
111da177e4SLinus Torvalds #include <linux/pagemap.h>
121da177e4SLinus Torvalds #include <linux/highmem.h>
135a0e3ad6STejun Heo #include <linux/slab.h>
1417093991SFabian Frederick #include <linux/uaccess.h>
151da177e4SLinus Torvalds #include <asm/unaligned.h>
161da177e4SLinus Torvalds #include <linux/buffer_head.h>
171da177e4SLinus Torvalds #include <linux/mpage.h>
181da177e4SLinus Torvalds #include <linux/writeback.h>
191da177e4SLinus Torvalds #include <linux/quotaops.h>
20ba9d8cecSVladimir Saveliev #include <linux/swap.h>
21e2e40f2cSChristoph Hellwig #include <linux/uio.h>
22be297968SChristoph Hellwig #include <linux/bio.h>
231da177e4SLinus Torvalds
24ba9d8cecSVladimir Saveliev int reiserfs_commit_write(struct file *f, struct page *page,
251da177e4SLinus Torvalds unsigned from, unsigned to);
261da177e4SLinus Torvalds
reiserfs_evict_inode(struct inode * inode)27845a2cc0SAl Viro void reiserfs_evict_inode(struct inode *inode)
281da177e4SLinus Torvalds {
29098297b2SJeff Mahoney /*
30098297b2SJeff Mahoney * We need blocks for transaction + (user+group) quota
31098297b2SJeff Mahoney * update (possibly delete)
32098297b2SJeff Mahoney */
33bd4c625cSLinus Torvalds int jbegin_count =
34bd4c625cSLinus Torvalds JOURNAL_PER_BALANCE_CNT * 2 +
35bd4c625cSLinus Torvalds 2 * REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb);
361da177e4SLinus Torvalds struct reiserfs_transaction_handle th;
3724996049SJeff Mahoney int err;
381da177e4SLinus Torvalds
39845a2cc0SAl Viro if (!inode->i_nlink && !is_bad_inode(inode))
40871a2931SChristoph Hellwig dquot_initialize(inode);
41907f4554SChristoph Hellwig
4291b0abe3SJohannes Weiner truncate_inode_pages_final(&inode->i_data);
43845a2cc0SAl Viro if (inode->i_nlink)
44845a2cc0SAl Viro goto no_delete;
45fef26658SMark Fasheh
46098297b2SJeff Mahoney /*
47098297b2SJeff Mahoney * The = 0 happens when we abort creating a new inode
48098297b2SJeff Mahoney * for some reason like lack of space..
49098297b2SJeff Mahoney * also handles bad_inode case
50098297b2SJeff Mahoney */
51098297b2SJeff Mahoney if (!(inode->i_state & I_NEW) && INODE_PKEY(inode)->k_objectid != 0) {
524c05141dSJeff Mahoney
531da177e4SLinus Torvalds reiserfs_delete_xattrs(inode);
541da177e4SLinus Torvalds
55278f6679SJeff Mahoney reiserfs_write_lock(inode->i_sb);
564c05141dSJeff Mahoney
57b0b33deeSAlexander Zarochentsev if (journal_begin(&th, inode->i_sb, jbegin_count))
581da177e4SLinus Torvalds goto out;
591da177e4SLinus Torvalds reiserfs_update_inode_transaction(inode);
601da177e4SLinus Torvalds
61eb35c218SJeff Mahoney reiserfs_discard_prealloc(&th, inode);
62eb35c218SJeff Mahoney
6324996049SJeff Mahoney err = reiserfs_delete_object(&th, inode);
641da177e4SLinus Torvalds
65098297b2SJeff Mahoney /*
66098297b2SJeff Mahoney * Do quota update inside a transaction for journaled quotas.
67098297b2SJeff Mahoney * We must do that after delete_object so that quota updates
68098297b2SJeff Mahoney * go into the same transaction as stat data deletion
69098297b2SJeff Mahoney */
70d2d0395fSJeff Mahoney if (!err) {
71d2d0395fSJeff Mahoney int depth = reiserfs_write_unlock_nested(inode->i_sb);
7263936ddaSChristoph Hellwig dquot_free_inode(inode);
73d2d0395fSJeff Mahoney reiserfs_write_lock_nested(inode->i_sb, depth);
74d2d0395fSJeff Mahoney }
751da177e4SLinus Torvalds
7658d85426SJeff Mahoney if (journal_end(&th))
771da177e4SLinus Torvalds goto out;
781da177e4SLinus Torvalds
79098297b2SJeff Mahoney /*
80098297b2SJeff Mahoney * check return value from reiserfs_delete_object after
8124996049SJeff Mahoney * ending the transaction
8224996049SJeff Mahoney */
8324996049SJeff Mahoney if (err)
8424996049SJeff Mahoney goto out;
8524996049SJeff Mahoney
86098297b2SJeff Mahoney /*
87098297b2SJeff Mahoney * all items of file are deleted, so we can remove
88098297b2SJeff Mahoney * "save" link
89098297b2SJeff Mahoney * we can't do anything about an error here
90098297b2SJeff Mahoney */
91098297b2SJeff Mahoney remove_save_link(inode, 0 /* not truncate */);
924c05141dSJeff Mahoney out:
93278f6679SJeff Mahoney reiserfs_write_unlock(inode->i_sb);
941da177e4SLinus Torvalds } else {
951da177e4SLinus Torvalds /* no object items are in the tree */
961da177e4SLinus Torvalds ;
971da177e4SLinus Torvalds }
98098297b2SJeff Mahoney
99098297b2SJeff Mahoney /* note this must go after the journal_end to prevent deadlock */
100098297b2SJeff Mahoney clear_inode(inode);
101098297b2SJeff Mahoney
102845a2cc0SAl Viro dquot_drop(inode);
1031da177e4SLinus Torvalds inode->i_blocks = 0;
104f4ae2faaSSergey Senozhatsky return;
105845a2cc0SAl Viro
106845a2cc0SAl Viro no_delete:
107dbd5768fSJan Kara clear_inode(inode);
108845a2cc0SAl Viro dquot_drop(inode);
1091da177e4SLinus Torvalds }
1101da177e4SLinus Torvalds
_make_cpu_key(struct cpu_key * key,int version,__u32 dirid,__u32 objectid,loff_t offset,int type,int length)111bd4c625cSLinus Torvalds static void _make_cpu_key(struct cpu_key *key, int version, __u32 dirid,
112bd4c625cSLinus Torvalds __u32 objectid, loff_t offset, int type, int length)
1131da177e4SLinus Torvalds {
1141da177e4SLinus Torvalds key->version = version;
1151da177e4SLinus Torvalds
1161da177e4SLinus Torvalds key->on_disk_key.k_dir_id = dirid;
1171da177e4SLinus Torvalds key->on_disk_key.k_objectid = objectid;
1181da177e4SLinus Torvalds set_cpu_key_k_offset(key, offset);
1191da177e4SLinus Torvalds set_cpu_key_k_type(key, type);
1201da177e4SLinus Torvalds key->key_length = length;
1211da177e4SLinus Torvalds }
1221da177e4SLinus Torvalds
123098297b2SJeff Mahoney /*
124098297b2SJeff Mahoney * take base of inode_key (it comes from inode always) (dirid, objectid)
125098297b2SJeff Mahoney * and version from an inode, set offset and type of key
126098297b2SJeff Mahoney */
make_cpu_key(struct cpu_key * key,struct inode * inode,loff_t offset,int type,int length)1271da177e4SLinus Torvalds void make_cpu_key(struct cpu_key *key, struct inode *inode, loff_t offset,
1281da177e4SLinus Torvalds int type, int length)
1291da177e4SLinus Torvalds {
130bd4c625cSLinus Torvalds _make_cpu_key(key, get_inode_item_key_version(inode),
131bd4c625cSLinus Torvalds le32_to_cpu(INODE_PKEY(inode)->k_dir_id),
132bd4c625cSLinus Torvalds le32_to_cpu(INODE_PKEY(inode)->k_objectid), offset, type,
133bd4c625cSLinus Torvalds length);
1341da177e4SLinus Torvalds }
1351da177e4SLinus Torvalds
136098297b2SJeff Mahoney /* when key is 0, do not set version and short key */
make_le_item_head(struct item_head * ih,const struct cpu_key * key,int version,loff_t offset,int type,int length,int entry_count)1371da177e4SLinus Torvalds inline void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
1381da177e4SLinus Torvalds int version,
1391da177e4SLinus Torvalds loff_t offset, int type, int length,
1401da177e4SLinus Torvalds int entry_count /*or ih_free_space */ )
1411da177e4SLinus Torvalds {
1421da177e4SLinus Torvalds if (key) {
1431da177e4SLinus Torvalds ih->ih_key.k_dir_id = cpu_to_le32(key->on_disk_key.k_dir_id);
144bd4c625cSLinus Torvalds ih->ih_key.k_objectid =
145bd4c625cSLinus Torvalds cpu_to_le32(key->on_disk_key.k_objectid);
1461da177e4SLinus Torvalds }
1471da177e4SLinus Torvalds put_ih_version(ih, version);
1481da177e4SLinus Torvalds set_le_ih_k_offset(ih, offset);
1491da177e4SLinus Torvalds set_le_ih_k_type(ih, type);
1501da177e4SLinus Torvalds put_ih_item_len(ih, length);
1511da177e4SLinus Torvalds /* set_ih_free_space (ih, 0); */
152098297b2SJeff Mahoney /*
153098297b2SJeff Mahoney * for directory items it is entry count, for directs and stat
154098297b2SJeff Mahoney * datas - 0xffff, for indirects - 0
155098297b2SJeff Mahoney */
1561da177e4SLinus Torvalds put_ih_entry_count(ih, entry_count);
1571da177e4SLinus Torvalds }
1581da177e4SLinus Torvalds
159098297b2SJeff Mahoney /*
160098297b2SJeff Mahoney * FIXME: we might cache recently accessed indirect item
161098297b2SJeff Mahoney * Ugh. Not too eager for that....
162098297b2SJeff Mahoney * I cut the code until such time as I see a convincing argument (benchmark).
163098297b2SJeff Mahoney * I don't want a bloated inode struct..., and I don't like code complexity....
1641da177e4SLinus Torvalds */
1651da177e4SLinus Torvalds
166098297b2SJeff Mahoney /*
167098297b2SJeff Mahoney * cutting the code is fine, since it really isn't in use yet and is easy
168098297b2SJeff Mahoney * to add back in. But, Vladimir has a really good idea here. Think
169098297b2SJeff Mahoney * about what happens for reading a file. For each page,
1702c69e205SMatthew Wilcox (Oracle) * The VFS layer calls reiserfs_read_folio, who searches the tree to find
171098297b2SJeff Mahoney * an indirect item. This indirect item has X number of pointers, where
172098297b2SJeff Mahoney * X is a big number if we've done the block allocation right. But,
1732c69e205SMatthew Wilcox (Oracle) * we only use one or two of these pointers during each call to read_folio,
174098297b2SJeff Mahoney * needlessly researching again later on.
175098297b2SJeff Mahoney *
176098297b2SJeff Mahoney * The size of the cache could be dynamic based on the size of the file.
177098297b2SJeff Mahoney *
178098297b2SJeff Mahoney * I'd also like to see us cache the location the stat data item, since
179098297b2SJeff Mahoney * we are needlessly researching for that frequently.
180098297b2SJeff Mahoney *
181098297b2SJeff Mahoney * --chris
182098297b2SJeff Mahoney */
183098297b2SJeff Mahoney
184098297b2SJeff Mahoney /*
185098297b2SJeff Mahoney * If this page has a file tail in it, and
186098297b2SJeff Mahoney * it was read in by get_block_create_0, the page data is valid,
187098297b2SJeff Mahoney * but tail is still sitting in a direct item, and we can't write to
188098297b2SJeff Mahoney * it. So, look through this page, and check all the mapped buffers
189098297b2SJeff Mahoney * to make sure they have valid block numbers. Any that don't need
190098297b2SJeff Mahoney * to be unmapped, so that __block_write_begin will correctly call
191098297b2SJeff Mahoney * reiserfs_get_block to convert the tail into an unformatted node
1921da177e4SLinus Torvalds */
fix_tail_page_for_writing(struct page * page)193bd4c625cSLinus Torvalds static inline void fix_tail_page_for_writing(struct page *page)
194bd4c625cSLinus Torvalds {
1951da177e4SLinus Torvalds struct buffer_head *head, *next, *bh;
1961da177e4SLinus Torvalds
1971da177e4SLinus Torvalds if (page && page_has_buffers(page)) {
1981da177e4SLinus Torvalds head = page_buffers(page);
1991da177e4SLinus Torvalds bh = head;
2001da177e4SLinus Torvalds do {
2011da177e4SLinus Torvalds next = bh->b_this_page;
2021da177e4SLinus Torvalds if (buffer_mapped(bh) && bh->b_blocknr == 0) {
2031da177e4SLinus Torvalds reiserfs_unmap_buffer(bh);
2041da177e4SLinus Torvalds }
2051da177e4SLinus Torvalds bh = next;
2061da177e4SLinus Torvalds } while (bh != head);
2071da177e4SLinus Torvalds }
2081da177e4SLinus Torvalds }
2091da177e4SLinus Torvalds
210098297b2SJeff Mahoney /*
211098297b2SJeff Mahoney * reiserfs_get_block does not need to allocate a block only if it has been
212098297b2SJeff Mahoney * done already or non-hole position has been found in the indirect item
213098297b2SJeff Mahoney */
allocation_needed(int retval,b_blocknr_t allocated,struct item_head * ih,__le32 * item,int pos_in_item)2141da177e4SLinus Torvalds static inline int allocation_needed(int retval, b_blocknr_t allocated,
2151da177e4SLinus Torvalds struct item_head *ih,
2163e8962beSAl Viro __le32 * item, int pos_in_item)
2171da177e4SLinus Torvalds {
2181da177e4SLinus Torvalds if (allocated)
2191da177e4SLinus Torvalds return 0;
2201da177e4SLinus Torvalds if (retval == POSITION_FOUND && is_indirect_le_ih(ih) &&
2211da177e4SLinus Torvalds get_block_num(item, pos_in_item))
2221da177e4SLinus Torvalds return 0;
2231da177e4SLinus Torvalds return 1;
2241da177e4SLinus Torvalds }
2251da177e4SLinus Torvalds
indirect_item_found(int retval,struct item_head * ih)2261da177e4SLinus Torvalds static inline int indirect_item_found(int retval, struct item_head *ih)
2271da177e4SLinus Torvalds {
2281da177e4SLinus Torvalds return (retval == POSITION_FOUND) && is_indirect_le_ih(ih);
2291da177e4SLinus Torvalds }
2301da177e4SLinus Torvalds
set_block_dev_mapped(struct buffer_head * bh,b_blocknr_t block,struct inode * inode)2311da177e4SLinus Torvalds static inline void set_block_dev_mapped(struct buffer_head *bh,
2321da177e4SLinus Torvalds b_blocknr_t block, struct inode *inode)
2331da177e4SLinus Torvalds {
2341da177e4SLinus Torvalds map_bh(bh, inode->i_sb, block);
2351da177e4SLinus Torvalds }
2361da177e4SLinus Torvalds
237098297b2SJeff Mahoney /*
238098297b2SJeff Mahoney * files which were created in the earlier version can not be longer,
239098297b2SJeff Mahoney * than 2 gb
240098297b2SJeff Mahoney */
file_capable(struct inode * inode,sector_t block)2413ee16670SJeff Mahoney static int file_capable(struct inode *inode, sector_t block)
2421da177e4SLinus Torvalds {
243098297b2SJeff Mahoney /* it is new file. */
244098297b2SJeff Mahoney if (get_inode_item_key_version(inode) != KEY_FORMAT_3_5 ||
245098297b2SJeff Mahoney /* old file, but 'block' is inside of 2gb */
246098297b2SJeff Mahoney block < (1 << (31 - inode->i_sb->s_blocksize_bits)))
2471da177e4SLinus Torvalds return 1;
2481da177e4SLinus Torvalds
2491da177e4SLinus Torvalds return 0;
2501da177e4SLinus Torvalds }
2511da177e4SLinus Torvalds
restart_transaction(struct reiserfs_transaction_handle * th,struct inode * inode,struct treepath * path)252deba0f49SAdrian Bunk static int restart_transaction(struct reiserfs_transaction_handle *th,
253fec6d055SJosef "Jeff" Sipek struct inode *inode, struct treepath *path)
254bd4c625cSLinus Torvalds {
2551da177e4SLinus Torvalds struct super_block *s = th->t_super;
2561da177e4SLinus Torvalds int err;
2571da177e4SLinus Torvalds
2581da177e4SLinus Torvalds BUG_ON(!th->t_trans_id);
2591da177e4SLinus Torvalds BUG_ON(!th->t_refcount);
2601da177e4SLinus Torvalds
26187b4126fSSuzuki K P pathrelse(path);
26287b4126fSSuzuki K P
2631da177e4SLinus Torvalds /* we cannot restart while nested */
2641da177e4SLinus Torvalds if (th->t_refcount > 1) {
2651da177e4SLinus Torvalds return 0;
2661da177e4SLinus Torvalds }
2671da177e4SLinus Torvalds reiserfs_update_sd(th, inode);
26858d85426SJeff Mahoney err = journal_end(th);
2691da177e4SLinus Torvalds if (!err) {
2701da177e4SLinus Torvalds err = journal_begin(th, s, JOURNAL_PER_BALANCE_CNT * 6);
2711da177e4SLinus Torvalds if (!err)
2721da177e4SLinus Torvalds reiserfs_update_inode_transaction(inode);
2731da177e4SLinus Torvalds }
2741da177e4SLinus Torvalds return err;
2751da177e4SLinus Torvalds }
2761da177e4SLinus Torvalds
277098297b2SJeff Mahoney /*
278098297b2SJeff Mahoney * it is called by get_block when create == 0. Returns block number
279098297b2SJeff Mahoney * for 'block'-th logical block of file. When it hits direct item it
280098297b2SJeff Mahoney * returns 0 (being called from bmap) or read direct item into piece
281098297b2SJeff Mahoney * of page (bh_result)
282098297b2SJeff Mahoney * Please improve the english/clarity in the comment above, as it is
283098297b2SJeff Mahoney * hard to understand.
284098297b2SJeff Mahoney */
_get_block_create_0(struct inode * inode,sector_t block,struct buffer_head * bh_result,int args)2853ee16670SJeff Mahoney static int _get_block_create_0(struct inode *inode, sector_t block,
286bd4c625cSLinus Torvalds struct buffer_head *bh_result, int args)
2871da177e4SLinus Torvalds {
2881da177e4SLinus Torvalds INITIALIZE_PATH(path);
2891da177e4SLinus Torvalds struct cpu_key key;
2901da177e4SLinus Torvalds struct buffer_head *bh;
2911da177e4SLinus Torvalds struct item_head *ih, tmp_ih;
2923ee16670SJeff Mahoney b_blocknr_t blocknr;
293bd6e21a9SZeng Jingxiang char *p;
2941da177e4SLinus Torvalds int chars;
2951da177e4SLinus Torvalds int ret;
2966283d58eSQu Fuping int result;
2971da177e4SLinus Torvalds int done = 0;
2981da177e4SLinus Torvalds unsigned long offset;
2991da177e4SLinus Torvalds
300098297b2SJeff Mahoney /* prepare the key to look for the 'block'-th block of file */
3011da177e4SLinus Torvalds make_cpu_key(&key, inode,
302bd4c625cSLinus Torvalds (loff_t) block * inode->i_sb->s_blocksize + 1, TYPE_ANY,
303bd4c625cSLinus Torvalds 3);
3041da177e4SLinus Torvalds
3056283d58eSQu Fuping result = search_for_position_by_key(inode->i_sb, &key, &path);
3066283d58eSQu Fuping if (result != POSITION_FOUND) {
3071da177e4SLinus Torvalds pathrelse(&path);
3086283d58eSQu Fuping if (result == IO_ERROR)
3096283d58eSQu Fuping return -EIO;
310098297b2SJeff Mahoney /*
311098297b2SJeff Mahoney * We do not return -ENOENT if there is a hole but page is
312098297b2SJeff Mahoney * uptodate, because it means that there is some MMAPED data
313098297b2SJeff Mahoney * associated with it that is yet to be written to disk.
314098297b2SJeff Mahoney */
315bd4c625cSLinus Torvalds if ((args & GET_BLOCK_NO_HOLE)
316bd4c625cSLinus Torvalds && !PageUptodate(bh_result->b_page)) {
3171da177e4SLinus Torvalds return -ENOENT;
3181da177e4SLinus Torvalds }
3191da177e4SLinus Torvalds return 0;
3201da177e4SLinus Torvalds }
321098297b2SJeff Mahoney
3221da177e4SLinus Torvalds bh = get_last_bh(&path);
3234cf5f7adSJeff Mahoney ih = tp_item_head(&path);
3241da177e4SLinus Torvalds if (is_indirect_le_ih(ih)) {
3254cf5f7adSJeff Mahoney __le32 *ind_item = (__le32 *) ih_item_body(bh, ih);
3261da177e4SLinus Torvalds
327098297b2SJeff Mahoney /*
328098297b2SJeff Mahoney * FIXME: here we could cache indirect item or part of it in
329098297b2SJeff Mahoney * the inode to avoid search_by_key in case of subsequent
330098297b2SJeff Mahoney * access to file
331098297b2SJeff Mahoney */
3321da177e4SLinus Torvalds blocknr = get_block_num(ind_item, path.pos_in_item);
3331da177e4SLinus Torvalds ret = 0;
3341da177e4SLinus Torvalds if (blocknr) {
3351da177e4SLinus Torvalds map_bh(bh_result, inode->i_sb, blocknr);
336bd4c625cSLinus Torvalds if (path.pos_in_item ==
337bd4c625cSLinus Torvalds ((ih_item_len(ih) / UNFM_P_SIZE) - 1)) {
3381da177e4SLinus Torvalds set_buffer_boundary(bh_result);
3391da177e4SLinus Torvalds }
3401da177e4SLinus Torvalds } else
341098297b2SJeff Mahoney /*
342098297b2SJeff Mahoney * We do not return -ENOENT if there is a hole but
343098297b2SJeff Mahoney * page is uptodate, because it means that there is
344098297b2SJeff Mahoney * some MMAPED data associated with it that is
345098297b2SJeff Mahoney * yet to be written to disk.
346098297b2SJeff Mahoney */
347bd4c625cSLinus Torvalds if ((args & GET_BLOCK_NO_HOLE)
348bd4c625cSLinus Torvalds && !PageUptodate(bh_result->b_page)) {
3491da177e4SLinus Torvalds ret = -ENOENT;
3501da177e4SLinus Torvalds }
3511da177e4SLinus Torvalds
3521da177e4SLinus Torvalds pathrelse(&path);
3531da177e4SLinus Torvalds return ret;
3541da177e4SLinus Torvalds }
355098297b2SJeff Mahoney /* requested data are in direct item(s) */
3561da177e4SLinus Torvalds if (!(args & GET_BLOCK_READ_DIRECT)) {
357098297b2SJeff Mahoney /*
358098297b2SJeff Mahoney * we are called by bmap. FIXME: we can not map block of file
359098297b2SJeff Mahoney * when it is stored in direct item(s)
360098297b2SJeff Mahoney */
3611da177e4SLinus Torvalds pathrelse(&path);
3621da177e4SLinus Torvalds return -ENOENT;
3631da177e4SLinus Torvalds }
3641da177e4SLinus Torvalds
365098297b2SJeff Mahoney /*
366098297b2SJeff Mahoney * if we've got a direct item, and the buffer or page was uptodate,
367098297b2SJeff Mahoney * we don't want to pull data off disk again. skip to the
368098297b2SJeff Mahoney * end, where we map the buffer and return
3691da177e4SLinus Torvalds */
3701da177e4SLinus Torvalds if (buffer_uptodate(bh_result)) {
3711da177e4SLinus Torvalds goto finished;
3721da177e4SLinus Torvalds } else
3731da177e4SLinus Torvalds /*
374098297b2SJeff Mahoney * grab_tail_page can trigger calls to reiserfs_get_block on
375098297b2SJeff Mahoney * up to date pages without any buffers. If the page is up
376098297b2SJeff Mahoney * to date, we don't want read old data off disk. Set the up
377098297b2SJeff Mahoney * to date bit on the buffer instead and jump to the end
3781da177e4SLinus Torvalds */
3791da177e4SLinus Torvalds if (!bh_result->b_page || PageUptodate(bh_result->b_page)) {
3801da177e4SLinus Torvalds set_buffer_uptodate(bh_result);
3811da177e4SLinus Torvalds goto finished;
3821da177e4SLinus Torvalds }
383098297b2SJeff Mahoney /* read file tail into part of page */
38409cbfeafSKirill A. Shutemov offset = (cpu_key_k_offset(&key) - 1) & (PAGE_SIZE - 1);
3851da177e4SLinus Torvalds copy_item_head(&tmp_ih, ih);
3861da177e4SLinus Torvalds
387098297b2SJeff Mahoney /*
388098297b2SJeff Mahoney * we only want to kmap if we are reading the tail into the page.
389098297b2SJeff Mahoney * this is not the common case, so we don't kmap until we are
390098297b2SJeff Mahoney * sure we need to. But, this means the item might move if
391098297b2SJeff Mahoney * kmap schedules
3921da177e4SLinus Torvalds */
3931da177e4SLinus Torvalds p = (char *)kmap(bh_result->b_page);
3941da177e4SLinus Torvalds p += offset;
3951da177e4SLinus Torvalds memset(p, 0, inode->i_sb->s_blocksize);
3961da177e4SLinus Torvalds do {
3971da177e4SLinus Torvalds if (!is_direct_le_ih(ih)) {
3981da177e4SLinus Torvalds BUG();
3991da177e4SLinus Torvalds }
400098297b2SJeff Mahoney /*
401098297b2SJeff Mahoney * make sure we don't read more bytes than actually exist in
402098297b2SJeff Mahoney * the file. This can happen in odd cases where i_size isn't
403098297b2SJeff Mahoney * correct, and when direct item padding results in a few
404098297b2SJeff Mahoney * extra bytes at the end of the direct item
4051da177e4SLinus Torvalds */
4061da177e4SLinus Torvalds if ((le_ih_k_offset(ih) + path.pos_in_item) > inode->i_size)
4071da177e4SLinus Torvalds break;
4081da177e4SLinus Torvalds if ((le_ih_k_offset(ih) - 1 + ih_item_len(ih)) > inode->i_size) {
409bd4c625cSLinus Torvalds chars =
410bd4c625cSLinus Torvalds inode->i_size - (le_ih_k_offset(ih) - 1) -
411bd4c625cSLinus Torvalds path.pos_in_item;
4121da177e4SLinus Torvalds done = 1;
4131da177e4SLinus Torvalds } else {
4141da177e4SLinus Torvalds chars = ih_item_len(ih) - path.pos_in_item;
4151da177e4SLinus Torvalds }
4164cf5f7adSJeff Mahoney memcpy(p, ih_item_body(bh, ih) + path.pos_in_item, chars);
4171da177e4SLinus Torvalds
4181da177e4SLinus Torvalds if (done)
4191da177e4SLinus Torvalds break;
4201da177e4SLinus Torvalds
4211da177e4SLinus Torvalds p += chars;
4221da177e4SLinus Torvalds
423098297b2SJeff Mahoney /*
424098297b2SJeff Mahoney * we done, if read direct item is not the last item of
425098297b2SJeff Mahoney * node FIXME: we could try to check right delimiting key
426098297b2SJeff Mahoney * to see whether direct item continues in the right
427098297b2SJeff Mahoney * neighbor or rely on i_size
428098297b2SJeff Mahoney */
4291da177e4SLinus Torvalds if (PATH_LAST_POSITION(&path) != (B_NR_ITEMS(bh) - 1))
4301da177e4SLinus Torvalds break;
4311da177e4SLinus Torvalds
432098297b2SJeff Mahoney /* update key to look for the next piece */
4331da177e4SLinus Torvalds set_cpu_key_k_offset(&key, cpu_key_k_offset(&key) + chars);
4346283d58eSQu Fuping result = search_for_position_by_key(inode->i_sb, &key, &path);
4356283d58eSQu Fuping if (result != POSITION_FOUND)
436098297b2SJeff Mahoney /* i/o error most likely */
4371da177e4SLinus Torvalds break;
4381da177e4SLinus Torvalds bh = get_last_bh(&path);
4394cf5f7adSJeff Mahoney ih = tp_item_head(&path);
4401da177e4SLinus Torvalds } while (1);
4411da177e4SLinus Torvalds
4421da177e4SLinus Torvalds flush_dcache_page(bh_result->b_page);
4431da177e4SLinus Torvalds kunmap(bh_result->b_page);
4441da177e4SLinus Torvalds
4451da177e4SLinus Torvalds finished:
4461da177e4SLinus Torvalds pathrelse(&path);
4476283d58eSQu Fuping
4486283d58eSQu Fuping if (result == IO_ERROR)
4496283d58eSQu Fuping return -EIO;
4506283d58eSQu Fuping
451098297b2SJeff Mahoney /*
452098297b2SJeff Mahoney * this buffer has valid data, but isn't valid for io. mapping it to
4531da177e4SLinus Torvalds * block #0 tells the rest of reiserfs it just has a tail in it
4541da177e4SLinus Torvalds */
4551da177e4SLinus Torvalds map_bh(bh_result, inode->i_sb, 0);
4561da177e4SLinus Torvalds set_buffer_uptodate(bh_result);
4571da177e4SLinus Torvalds return 0;
4581da177e4SLinus Torvalds }
4591da177e4SLinus Torvalds
460098297b2SJeff Mahoney /*
461098297b2SJeff Mahoney * this is called to create file map. So, _get_block_create_0 will not
462098297b2SJeff Mahoney * read direct item
463098297b2SJeff Mahoney */
reiserfs_bmap(struct inode * inode,sector_t block,struct buffer_head * bh_result,int create)4641da177e4SLinus Torvalds static int reiserfs_bmap(struct inode *inode, sector_t block,
4651da177e4SLinus Torvalds struct buffer_head *bh_result, int create)
4661da177e4SLinus Torvalds {
4671da177e4SLinus Torvalds if (!file_capable(inode, block))
4681da177e4SLinus Torvalds return -EFBIG;
4691da177e4SLinus Torvalds
4701da177e4SLinus Torvalds reiserfs_write_lock(inode->i_sb);
4711da177e4SLinus Torvalds /* do not read the direct item */
4721da177e4SLinus Torvalds _get_block_create_0(inode, block, bh_result, 0);
4731da177e4SLinus Torvalds reiserfs_write_unlock(inode->i_sb);
4741da177e4SLinus Torvalds return 0;
4751da177e4SLinus Torvalds }
4761da177e4SLinus Torvalds
477098297b2SJeff Mahoney /*
478098297b2SJeff Mahoney * special version of get_block that is only used by grab_tail_page right
479098297b2SJeff Mahoney * now. It is sent to __block_write_begin, and when you try to get a
480098297b2SJeff Mahoney * block past the end of the file (or a block from a hole) it returns
481098297b2SJeff Mahoney * -ENOENT instead of a valid buffer. __block_write_begin expects to
482098297b2SJeff Mahoney * be able to do i/o on the buffers returned, unless an error value
483098297b2SJeff Mahoney * is also returned.
484098297b2SJeff Mahoney *
485098297b2SJeff Mahoney * So, this allows __block_write_begin to be used for reading a single block
486098297b2SJeff Mahoney * in a page. Where it does not produce a valid page for holes, or past the
487098297b2SJeff Mahoney * end of the file. This turns out to be exactly what we need for reading
488098297b2SJeff Mahoney * tails for conversion.
489098297b2SJeff Mahoney *
490098297b2SJeff Mahoney * The point of the wrapper is forcing a certain value for create, even
491098297b2SJeff Mahoney * though the VFS layer is calling this function with create==1. If you
492098297b2SJeff Mahoney * don't want to send create == GET_BLOCK_NO_HOLE to reiserfs_get_block,
493098297b2SJeff Mahoney * don't use this function.
4941da177e4SLinus Torvalds */
reiserfs_get_block_create_0(struct inode * inode,sector_t block,struct buffer_head * bh_result,int create)4951da177e4SLinus Torvalds static int reiserfs_get_block_create_0(struct inode *inode, sector_t block,
496bd4c625cSLinus Torvalds struct buffer_head *bh_result,
497bd4c625cSLinus Torvalds int create)
498bd4c625cSLinus Torvalds {
4991da177e4SLinus Torvalds return reiserfs_get_block(inode, block, bh_result, GET_BLOCK_NO_HOLE);
5001da177e4SLinus Torvalds }
5011da177e4SLinus Torvalds
502098297b2SJeff Mahoney /*
503098297b2SJeff Mahoney * This is special helper for reiserfs_get_block in case we are executing
504098297b2SJeff Mahoney * direct_IO request.
505098297b2SJeff Mahoney */
reiserfs_get_blocks_direct_io(struct inode * inode,sector_t iblock,struct buffer_head * bh_result,int create)5061da177e4SLinus Torvalds static int reiserfs_get_blocks_direct_io(struct inode *inode,
5071da177e4SLinus Torvalds sector_t iblock,
5081da177e4SLinus Torvalds struct buffer_head *bh_result,
5091da177e4SLinus Torvalds int create)
5101da177e4SLinus Torvalds {
5111da177e4SLinus Torvalds int ret;
5121da177e4SLinus Torvalds
5131da177e4SLinus Torvalds bh_result->b_page = NULL;
5141da177e4SLinus Torvalds
515098297b2SJeff Mahoney /*
516098297b2SJeff Mahoney * We set the b_size before reiserfs_get_block call since it is
517098297b2SJeff Mahoney * referenced in convert_tail_for_hole() that may be called from
518098297b2SJeff Mahoney * reiserfs_get_block()
519098297b2SJeff Mahoney */
52093407472SFabian Frederick bh_result->b_size = i_blocksize(inode);
5211da177e4SLinus Torvalds
5221da177e4SLinus Torvalds ret = reiserfs_get_block(inode, iblock, bh_result,
5231da177e4SLinus Torvalds create | GET_BLOCK_NO_DANGLE);
5241da177e4SLinus Torvalds if (ret)
5251da177e4SLinus Torvalds goto out;
5261da177e4SLinus Torvalds
5271da177e4SLinus Torvalds /* don't allow direct io onto tail pages */
5281da177e4SLinus Torvalds if (buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
529098297b2SJeff Mahoney /*
530098297b2SJeff Mahoney * make sure future calls to the direct io funcs for this
531098297b2SJeff Mahoney * offset in the file fail by unmapping the buffer
5321da177e4SLinus Torvalds */
5331da177e4SLinus Torvalds clear_buffer_mapped(bh_result);
5341da177e4SLinus Torvalds ret = -EINVAL;
5351da177e4SLinus Torvalds }
536098297b2SJeff Mahoney
537098297b2SJeff Mahoney /*
538098297b2SJeff Mahoney * Possible unpacked tail. Flush the data before pages have
539098297b2SJeff Mahoney * disappeared
540098297b2SJeff Mahoney */
5411da177e4SLinus Torvalds if (REISERFS_I(inode)->i_flags & i_pack_on_close_mask) {
5421da177e4SLinus Torvalds int err;
5438ebc4232SFrederic Weisbecker
5448ebc4232SFrederic Weisbecker reiserfs_write_lock(inode->i_sb);
5458ebc4232SFrederic Weisbecker
5461da177e4SLinus Torvalds err = reiserfs_commit_for_inode(inode);
5471da177e4SLinus Torvalds REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
5488ebc4232SFrederic Weisbecker
5498ebc4232SFrederic Weisbecker reiserfs_write_unlock(inode->i_sb);
5508ebc4232SFrederic Weisbecker
5511da177e4SLinus Torvalds if (err < 0)
5521da177e4SLinus Torvalds ret = err;
5531da177e4SLinus Torvalds }
5541da177e4SLinus Torvalds out:
5551da177e4SLinus Torvalds return ret;
5561da177e4SLinus Torvalds }
5571da177e4SLinus Torvalds
5581da177e4SLinus Torvalds /*
559098297b2SJeff Mahoney * helper function for when reiserfs_get_block is called for a hole
560098297b2SJeff Mahoney * but the file tail is still in a direct item
561098297b2SJeff Mahoney * bh_result is the buffer head for the hole
562098297b2SJeff Mahoney * tail_offset is the offset of the start of the tail in the file
563098297b2SJeff Mahoney *
564098297b2SJeff Mahoney * This calls prepare_write, which will start a new transaction
565098297b2SJeff Mahoney * you should not be in a transaction, or have any paths held when you
566098297b2SJeff Mahoney * call this.
5671da177e4SLinus Torvalds */
convert_tail_for_hole(struct inode * inode,struct buffer_head * bh_result,loff_t tail_offset)5681da177e4SLinus Torvalds static int convert_tail_for_hole(struct inode *inode,
5691da177e4SLinus Torvalds struct buffer_head *bh_result,
570bd4c625cSLinus Torvalds loff_t tail_offset)
571bd4c625cSLinus Torvalds {
5721da177e4SLinus Torvalds unsigned long index;
5731da177e4SLinus Torvalds unsigned long tail_end;
5741da177e4SLinus Torvalds unsigned long tail_start;
5751da177e4SLinus Torvalds struct page *tail_page;
5761da177e4SLinus Torvalds struct page *hole_page = bh_result->b_page;
5771da177e4SLinus Torvalds int retval = 0;
5781da177e4SLinus Torvalds
5791da177e4SLinus Torvalds if ((tail_offset & (bh_result->b_size - 1)) != 1)
5801da177e4SLinus Torvalds return -EIO;
5811da177e4SLinus Torvalds
5821da177e4SLinus Torvalds /* always try to read until the end of the block */
58309cbfeafSKirill A. Shutemov tail_start = tail_offset & (PAGE_SIZE - 1);
5841da177e4SLinus Torvalds tail_end = (tail_start | (bh_result->b_size - 1)) + 1;
5851da177e4SLinus Torvalds
58609cbfeafSKirill A. Shutemov index = tail_offset >> PAGE_SHIFT;
587098297b2SJeff Mahoney /*
588098297b2SJeff Mahoney * hole_page can be zero in case of direct_io, we are sure
589098297b2SJeff Mahoney * that we cannot get here if we write with O_DIRECT into tail page
590098297b2SJeff Mahoney */
5911da177e4SLinus Torvalds if (!hole_page || index != hole_page->index) {
5921da177e4SLinus Torvalds tail_page = grab_cache_page(inode->i_mapping, index);
5931da177e4SLinus Torvalds retval = -ENOMEM;
5941da177e4SLinus Torvalds if (!tail_page) {
5951da177e4SLinus Torvalds goto out;
5961da177e4SLinus Torvalds }
5971da177e4SLinus Torvalds } else {
5981da177e4SLinus Torvalds tail_page = hole_page;
5991da177e4SLinus Torvalds }
6001da177e4SLinus Torvalds
601098297b2SJeff Mahoney /*
602098297b2SJeff Mahoney * we don't have to make sure the conversion did not happen while
603098297b2SJeff Mahoney * we were locking the page because anyone that could convert
604098297b2SJeff Mahoney * must first take i_mutex.
605098297b2SJeff Mahoney *
606098297b2SJeff Mahoney * We must fix the tail page for writing because it might have buffers
607098297b2SJeff Mahoney * that are mapped, but have a block number of 0. This indicates tail
608098297b2SJeff Mahoney * data that has been read directly into the page, and
609098297b2SJeff Mahoney * __block_write_begin won't trigger a get_block in this case.
6101da177e4SLinus Torvalds */
6111da177e4SLinus Torvalds fix_tail_page_for_writing(tail_page);
612ebdec241SChristoph Hellwig retval = __reiserfs_write_begin(tail_page, tail_start,
613ebdec241SChristoph Hellwig tail_end - tail_start);
6141da177e4SLinus Torvalds if (retval)
6151da177e4SLinus Torvalds goto unlock;
6161da177e4SLinus Torvalds
6171da177e4SLinus Torvalds /* tail conversion might change the data in the page */
6181da177e4SLinus Torvalds flush_dcache_page(tail_page);
6191da177e4SLinus Torvalds
6201da177e4SLinus Torvalds retval = reiserfs_commit_write(NULL, tail_page, tail_start, tail_end);
6211da177e4SLinus Torvalds
6221da177e4SLinus Torvalds unlock:
6231da177e4SLinus Torvalds if (tail_page != hole_page) {
6241da177e4SLinus Torvalds unlock_page(tail_page);
62509cbfeafSKirill A. Shutemov put_page(tail_page);
6261da177e4SLinus Torvalds }
6271da177e4SLinus Torvalds out:
6281da177e4SLinus Torvalds return retval;
6291da177e4SLinus Torvalds }
6301da177e4SLinus Torvalds
_allocate_block(struct reiserfs_transaction_handle * th,sector_t block,struct inode * inode,b_blocknr_t * allocated_block_nr,struct treepath * path,int flags)6311da177e4SLinus Torvalds static inline int _allocate_block(struct reiserfs_transaction_handle *th,
6323ee16670SJeff Mahoney sector_t block,
6331da177e4SLinus Torvalds struct inode *inode,
6341da177e4SLinus Torvalds b_blocknr_t * allocated_block_nr,
635fec6d055SJosef "Jeff" Sipek struct treepath *path, int flags)
636bd4c625cSLinus Torvalds {
6371da177e4SLinus Torvalds BUG_ON(!th->t_trans_id);
6381da177e4SLinus Torvalds
6391da177e4SLinus Torvalds #ifdef REISERFS_PREALLOCATE
6401b1dcc1bSJes Sorensen if (!(flags & GET_BLOCK_NO_IMUX)) {
641bd4c625cSLinus Torvalds return reiserfs_new_unf_blocknrs2(th, inode, allocated_block_nr,
642bd4c625cSLinus Torvalds path, block);
6431da177e4SLinus Torvalds }
6441da177e4SLinus Torvalds #endif
645bd4c625cSLinus Torvalds return reiserfs_new_unf_blocknrs(th, inode, allocated_block_nr, path,
646bd4c625cSLinus Torvalds block);
6471da177e4SLinus Torvalds }
6481da177e4SLinus Torvalds
reiserfs_get_block(struct inode * inode,sector_t block,struct buffer_head * bh_result,int create)6491da177e4SLinus Torvalds int reiserfs_get_block(struct inode *inode, sector_t block,
6501da177e4SLinus Torvalds struct buffer_head *bh_result, int create)
6511da177e4SLinus Torvalds {
6521da177e4SLinus Torvalds int repeat, retval = 0;
653098297b2SJeff Mahoney /* b_blocknr_t is (unsigned) 32 bit int*/
654098297b2SJeff Mahoney b_blocknr_t allocated_block_nr = 0;
6551da177e4SLinus Torvalds INITIALIZE_PATH(path);
6561da177e4SLinus Torvalds int pos_in_item;
6571da177e4SLinus Torvalds struct cpu_key key;
6581da177e4SLinus Torvalds struct buffer_head *bh, *unbh = NULL;
6591da177e4SLinus Torvalds struct item_head *ih, tmp_ih;
6603e8962beSAl Viro __le32 *item;
6611da177e4SLinus Torvalds int done;
6621da177e4SLinus Torvalds int fs_gen;
6631da177e4SLinus Torvalds struct reiserfs_transaction_handle *th = NULL;
664098297b2SJeff Mahoney /*
665098297b2SJeff Mahoney * space reserved in transaction batch:
666098297b2SJeff Mahoney * . 3 balancings in direct->indirect conversion
667098297b2SJeff Mahoney * . 1 block involved into reiserfs_update_sd()
668098297b2SJeff Mahoney * XXX in practically impossible worst case direct2indirect()
669098297b2SJeff Mahoney * can incur (much) more than 3 balancings.
670098297b2SJeff Mahoney * quota update for user, group
671098297b2SJeff Mahoney */
672bd4c625cSLinus Torvalds int jbegin_count =
673bd4c625cSLinus Torvalds JOURNAL_PER_BALANCE_CNT * 3 + 1 +
674bd4c625cSLinus Torvalds 2 * REISERFS_QUOTA_TRANS_BLOCKS(inode->i_sb);
6751da177e4SLinus Torvalds int version;
6761da177e4SLinus Torvalds int dangle = 1;
677bd4c625cSLinus Torvalds loff_t new_offset =
678bd4c625cSLinus Torvalds (((loff_t) block) << inode->i_sb->s_blocksize_bits) + 1;
6791da177e4SLinus Torvalds
680278f6679SJeff Mahoney reiserfs_write_lock(inode->i_sb);
6811da177e4SLinus Torvalds version = get_inode_item_key_version(inode);
6821da177e4SLinus Torvalds
6831da177e4SLinus Torvalds if (!file_capable(inode, block)) {
684278f6679SJeff Mahoney reiserfs_write_unlock(inode->i_sb);
6851da177e4SLinus Torvalds return -EFBIG;
6861da177e4SLinus Torvalds }
6871da177e4SLinus Torvalds
688098297b2SJeff Mahoney /*
689098297b2SJeff Mahoney * if !create, we aren't changing the FS, so we don't need to
690098297b2SJeff Mahoney * log anything, so we don't need to start a transaction
6911da177e4SLinus Torvalds */
6921da177e4SLinus Torvalds if (!(create & GET_BLOCK_CREATE)) {
6931da177e4SLinus Torvalds int ret;
6941da177e4SLinus Torvalds /* find number of block-th logical block of the file */
6951da177e4SLinus Torvalds ret = _get_block_create_0(inode, block, bh_result,
6961da177e4SLinus Torvalds create | GET_BLOCK_READ_DIRECT);
697278f6679SJeff Mahoney reiserfs_write_unlock(inode->i_sb);
6981da177e4SLinus Torvalds return ret;
6991da177e4SLinus Torvalds }
700098297b2SJeff Mahoney
7011da177e4SLinus Torvalds /*
7021da177e4SLinus Torvalds * if we're already in a transaction, make sure to close
7031da177e4SLinus Torvalds * any new transactions we start in this func
7041da177e4SLinus Torvalds */
7051da177e4SLinus Torvalds if ((create & GET_BLOCK_NO_DANGLE) ||
7061da177e4SLinus Torvalds reiserfs_transaction_running(inode->i_sb))
7071da177e4SLinus Torvalds dangle = 0;
7081da177e4SLinus Torvalds
709098297b2SJeff Mahoney /*
710098297b2SJeff Mahoney * If file is of such a size, that it might have a tail and
711098297b2SJeff Mahoney * tails are enabled we should mark it as possibly needing
712098297b2SJeff Mahoney * tail packing on close
7131da177e4SLinus Torvalds */
714bd4c625cSLinus Torvalds if ((have_large_tails(inode->i_sb)
715bd4c625cSLinus Torvalds && inode->i_size < i_block_size(inode) * 4)
716bd4c625cSLinus Torvalds || (have_small_tails(inode->i_sb)
717bd4c625cSLinus Torvalds && inode->i_size < i_block_size(inode)))
7181da177e4SLinus Torvalds REISERFS_I(inode)->i_flags |= i_pack_on_close_mask;
7191da177e4SLinus Torvalds
7201da177e4SLinus Torvalds /* set the key of the first byte in the 'block'-th block of file */
721bd4c625cSLinus Torvalds make_cpu_key(&key, inode, new_offset, TYPE_ANY, 3 /*key length */ );
7221da177e4SLinus Torvalds if ((new_offset + inode->i_sb->s_blocksize - 1) > inode->i_size) {
7231da177e4SLinus Torvalds start_trans:
7241da177e4SLinus Torvalds th = reiserfs_persistent_transaction(inode->i_sb, jbegin_count);
7251da177e4SLinus Torvalds if (!th) {
7261da177e4SLinus Torvalds retval = -ENOMEM;
7271da177e4SLinus Torvalds goto failure;
7281da177e4SLinus Torvalds }
7291da177e4SLinus Torvalds reiserfs_update_inode_transaction(inode);
7301da177e4SLinus Torvalds }
7311da177e4SLinus Torvalds research:
7321da177e4SLinus Torvalds
7331da177e4SLinus Torvalds retval = search_for_position_by_key(inode->i_sb, &key, &path);
7341da177e4SLinus Torvalds if (retval == IO_ERROR) {
7351da177e4SLinus Torvalds retval = -EIO;
7361da177e4SLinus Torvalds goto failure;
7371da177e4SLinus Torvalds }
7381da177e4SLinus Torvalds
7391da177e4SLinus Torvalds bh = get_last_bh(&path);
7404cf5f7adSJeff Mahoney ih = tp_item_head(&path);
7414cf5f7adSJeff Mahoney item = tp_item_body(&path);
7421da177e4SLinus Torvalds pos_in_item = path.pos_in_item;
7431da177e4SLinus Torvalds
7441da177e4SLinus Torvalds fs_gen = get_generation(inode->i_sb);
7451da177e4SLinus Torvalds copy_item_head(&tmp_ih, ih);
7461da177e4SLinus Torvalds
747bd4c625cSLinus Torvalds if (allocation_needed
748bd4c625cSLinus Torvalds (retval, allocated_block_nr, ih, item, pos_in_item)) {
7491da177e4SLinus Torvalds /* we have to allocate block for the unformatted node */
7501da177e4SLinus Torvalds if (!th) {
7511da177e4SLinus Torvalds pathrelse(&path);
7521da177e4SLinus Torvalds goto start_trans;
7531da177e4SLinus Torvalds }
7541da177e4SLinus Torvalds
755bd4c625cSLinus Torvalds repeat =
756bd4c625cSLinus Torvalds _allocate_block(th, block, inode, &allocated_block_nr,
757bd4c625cSLinus Torvalds &path, create);
7581da177e4SLinus Torvalds
759098297b2SJeff Mahoney /*
760098297b2SJeff Mahoney * restart the transaction to give the journal a chance to free
761098297b2SJeff Mahoney * some blocks. releases the path, so we have to go back to
762098297b2SJeff Mahoney * research if we succeed on the second try
7631da177e4SLinus Torvalds */
764098297b2SJeff Mahoney if (repeat == NO_DISK_SPACE || repeat == QUOTA_EXCEEDED) {
7651da177e4SLinus Torvalds SB_JOURNAL(inode->i_sb)->j_next_async_flush = 1;
7661da177e4SLinus Torvalds retval = restart_transaction(th, inode, &path);
7671da177e4SLinus Torvalds if (retval)
7681da177e4SLinus Torvalds goto failure;
769bd4c625cSLinus Torvalds repeat =
770bd4c625cSLinus Torvalds _allocate_block(th, block, inode,
771bd4c625cSLinus Torvalds &allocated_block_nr, NULL, create);
7721da177e4SLinus Torvalds
7731da177e4SLinus Torvalds if (repeat != NO_DISK_SPACE && repeat != QUOTA_EXCEEDED) {
7741da177e4SLinus Torvalds goto research;
7751da177e4SLinus Torvalds }
7761da177e4SLinus Torvalds if (repeat == QUOTA_EXCEEDED)
7771da177e4SLinus Torvalds retval = -EDQUOT;
7781da177e4SLinus Torvalds else
7791da177e4SLinus Torvalds retval = -ENOSPC;
7801da177e4SLinus Torvalds goto failure;
7811da177e4SLinus Torvalds }
7821da177e4SLinus Torvalds
783bd4c625cSLinus Torvalds if (fs_changed(fs_gen, inode->i_sb)
784bd4c625cSLinus Torvalds && item_moved(&tmp_ih, &path)) {
7851da177e4SLinus Torvalds goto research;
7861da177e4SLinus Torvalds }
7871da177e4SLinus Torvalds }
7881da177e4SLinus Torvalds
7891da177e4SLinus Torvalds if (indirect_item_found(retval, ih)) {
7901da177e4SLinus Torvalds b_blocknr_t unfm_ptr;
791098297b2SJeff Mahoney /*
792098297b2SJeff Mahoney * 'block'-th block is in the file already (there is
793098297b2SJeff Mahoney * corresponding cell in some indirect item). But it may be
794098297b2SJeff Mahoney * zero unformatted node pointer (hole)
795098297b2SJeff Mahoney */
7961da177e4SLinus Torvalds unfm_ptr = get_block_num(item, pos_in_item);
7971da177e4SLinus Torvalds if (unfm_ptr == 0) {
7981da177e4SLinus Torvalds /* use allocated block to plug the hole */
7991da177e4SLinus Torvalds reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
800bd4c625cSLinus Torvalds if (fs_changed(fs_gen, inode->i_sb)
801bd4c625cSLinus Torvalds && item_moved(&tmp_ih, &path)) {
802bd4c625cSLinus Torvalds reiserfs_restore_prepared_buffer(inode->i_sb,
803bd4c625cSLinus Torvalds bh);
8041da177e4SLinus Torvalds goto research;
8051da177e4SLinus Torvalds }
8061da177e4SLinus Torvalds set_buffer_new(bh_result);
807bd4c625cSLinus Torvalds if (buffer_dirty(bh_result)
808bd4c625cSLinus Torvalds && reiserfs_data_ordered(inode->i_sb))
8091da177e4SLinus Torvalds reiserfs_add_ordered_list(inode, bh_result);
8101da177e4SLinus Torvalds put_block_num(item, pos_in_item, allocated_block_nr);
8111da177e4SLinus Torvalds unfm_ptr = allocated_block_nr;
81209f1b80bSJeff Mahoney journal_mark_dirty(th, bh);
8131da177e4SLinus Torvalds reiserfs_update_sd(th, inode);
8141da177e4SLinus Torvalds }
8151da177e4SLinus Torvalds set_block_dev_mapped(bh_result, unfm_ptr, inode);
8161da177e4SLinus Torvalds pathrelse(&path);
8171da177e4SLinus Torvalds retval = 0;
8181da177e4SLinus Torvalds if (!dangle && th)
8191da177e4SLinus Torvalds retval = reiserfs_end_persistent_transaction(th);
8201da177e4SLinus Torvalds
821278f6679SJeff Mahoney reiserfs_write_unlock(inode->i_sb);
8221da177e4SLinus Torvalds
823098297b2SJeff Mahoney /*
824098297b2SJeff Mahoney * the item was found, so new blocks were not added to the file
825098297b2SJeff Mahoney * there is no need to make sure the inode is updated with this
826098297b2SJeff Mahoney * transaction
8271da177e4SLinus Torvalds */
8281da177e4SLinus Torvalds return retval;
8291da177e4SLinus Torvalds }
8301da177e4SLinus Torvalds
8311da177e4SLinus Torvalds if (!th) {
8321da177e4SLinus Torvalds pathrelse(&path);
8331da177e4SLinus Torvalds goto start_trans;
8341da177e4SLinus Torvalds }
8351da177e4SLinus Torvalds
836098297b2SJeff Mahoney /*
837098297b2SJeff Mahoney * desired position is not found or is in the direct item. We have
838098297b2SJeff Mahoney * to append file with holes up to 'block'-th block converting
839098297b2SJeff Mahoney * direct items to indirect one if necessary
840098297b2SJeff Mahoney */
8411da177e4SLinus Torvalds done = 0;
8421da177e4SLinus Torvalds do {
8431da177e4SLinus Torvalds if (is_statdata_le_ih(ih)) {
8443e8962beSAl Viro __le32 unp = 0;
8451da177e4SLinus Torvalds struct cpu_key tmp_key;
8461da177e4SLinus Torvalds
8471da177e4SLinus Torvalds /* indirect item has to be inserted */
848bd4c625cSLinus Torvalds make_le_item_head(&tmp_ih, &key, version, 1,
849bd4c625cSLinus Torvalds TYPE_INDIRECT, UNFM_P_SIZE,
850bd4c625cSLinus Torvalds 0 /* free_space */ );
8511da177e4SLinus Torvalds
852098297b2SJeff Mahoney /*
853098297b2SJeff Mahoney * we are going to add 'block'-th block to the file.
854098297b2SJeff Mahoney * Use allocated block for that
855098297b2SJeff Mahoney */
8561da177e4SLinus Torvalds if (cpu_key_k_offset(&key) == 1) {
8571da177e4SLinus Torvalds unp = cpu_to_le32(allocated_block_nr);
858bd4c625cSLinus Torvalds set_block_dev_mapped(bh_result,
859bd4c625cSLinus Torvalds allocated_block_nr, inode);
8601da177e4SLinus Torvalds set_buffer_new(bh_result);
8611da177e4SLinus Torvalds done = 1;
8621da177e4SLinus Torvalds }
863098297b2SJeff Mahoney tmp_key = key; /* ;) */
8641da177e4SLinus Torvalds set_cpu_key_k_offset(&tmp_key, 1);
8651da177e4SLinus Torvalds PATH_LAST_POSITION(&path)++;
8661da177e4SLinus Torvalds
867bd4c625cSLinus Torvalds retval =
868bd4c625cSLinus Torvalds reiserfs_insert_item(th, &path, &tmp_key, &tmp_ih,
869bd4c625cSLinus Torvalds inode, (char *)&unp);
8701da177e4SLinus Torvalds if (retval) {
871bd4c625cSLinus Torvalds reiserfs_free_block(th, inode,
872bd4c625cSLinus Torvalds allocated_block_nr, 1);
873098297b2SJeff Mahoney /*
874098297b2SJeff Mahoney * retval == -ENOSPC, -EDQUOT or -EIO
875098297b2SJeff Mahoney * or -EEXIST
876098297b2SJeff Mahoney */
877098297b2SJeff Mahoney goto failure;
8781da177e4SLinus Torvalds }
8791da177e4SLinus Torvalds } else if (is_direct_le_ih(ih)) {
8801da177e4SLinus Torvalds /* direct item has to be converted */
8811da177e4SLinus Torvalds loff_t tail_offset;
8821da177e4SLinus Torvalds
883bd4c625cSLinus Torvalds tail_offset =
884bd4c625cSLinus Torvalds ((le_ih_k_offset(ih) -
885bd4c625cSLinus Torvalds 1) & ~(inode->i_sb->s_blocksize - 1)) + 1;
886098297b2SJeff Mahoney
887098297b2SJeff Mahoney /*
888098297b2SJeff Mahoney * direct item we just found fits into block we have
889098297b2SJeff Mahoney * to map. Convert it into unformatted node: use
890098297b2SJeff Mahoney * bh_result for the conversion
891098297b2SJeff Mahoney */
8921da177e4SLinus Torvalds if (tail_offset == cpu_key_k_offset(&key)) {
893bd4c625cSLinus Torvalds set_block_dev_mapped(bh_result,
894bd4c625cSLinus Torvalds allocated_block_nr, inode);
8951da177e4SLinus Torvalds unbh = bh_result;
8961da177e4SLinus Torvalds done = 1;
8971da177e4SLinus Torvalds } else {
898098297b2SJeff Mahoney /*
899098297b2SJeff Mahoney * we have to pad file tail stored in direct
900098297b2SJeff Mahoney * item(s) up to block size and convert it
901098297b2SJeff Mahoney * to unformatted node. FIXME: this should
902098297b2SJeff Mahoney * also get into page cache
903098297b2SJeff Mahoney */
9041da177e4SLinus Torvalds
9051da177e4SLinus Torvalds pathrelse(&path);
9061da177e4SLinus Torvalds /*
9071da177e4SLinus Torvalds * ugly, but we can only end the transaction if
9081da177e4SLinus Torvalds * we aren't nested
9091da177e4SLinus Torvalds */
9101da177e4SLinus Torvalds BUG_ON(!th->t_refcount);
9111da177e4SLinus Torvalds if (th->t_refcount == 1) {
912bd4c625cSLinus Torvalds retval =
913bd4c625cSLinus Torvalds reiserfs_end_persistent_transaction
914bd4c625cSLinus Torvalds (th);
9151da177e4SLinus Torvalds th = NULL;
9161da177e4SLinus Torvalds if (retval)
9171da177e4SLinus Torvalds goto failure;
9181da177e4SLinus Torvalds }
9191da177e4SLinus Torvalds
920bd4c625cSLinus Torvalds retval =
921bd4c625cSLinus Torvalds convert_tail_for_hole(inode, bh_result,
922bd4c625cSLinus Torvalds tail_offset);
9231da177e4SLinus Torvalds if (retval) {
9241da177e4SLinus Torvalds if (retval != -ENOSPC)
9250030b645SJeff Mahoney reiserfs_error(inode->i_sb,
92645b03d5eSJeff Mahoney "clm-6004",
92745b03d5eSJeff Mahoney "convert tail failed "
92845b03d5eSJeff Mahoney "inode %lu, error %d",
929bd4c625cSLinus Torvalds inode->i_ino,
930bd4c625cSLinus Torvalds retval);
9311da177e4SLinus Torvalds if (allocated_block_nr) {
932098297b2SJeff Mahoney /*
933098297b2SJeff Mahoney * the bitmap, the super,
934098297b2SJeff Mahoney * and the stat data == 3
935098297b2SJeff Mahoney */
9361da177e4SLinus Torvalds if (!th)
9371da177e4SLinus Torvalds th = reiserfs_persistent_transaction(inode->i_sb, 3);
9381da177e4SLinus Torvalds if (th)
939bd4c625cSLinus Torvalds reiserfs_free_block(th,
940bd4c625cSLinus Torvalds inode,
941bd4c625cSLinus Torvalds allocated_block_nr,
942bd4c625cSLinus Torvalds 1);
9431da177e4SLinus Torvalds }
9441da177e4SLinus Torvalds goto failure;
9451da177e4SLinus Torvalds }
9461da177e4SLinus Torvalds goto research;
9471da177e4SLinus Torvalds }
948bd4c625cSLinus Torvalds retval =
949bd4c625cSLinus Torvalds direct2indirect(th, inode, &path, unbh,
950bd4c625cSLinus Torvalds tail_offset);
9511da177e4SLinus Torvalds if (retval) {
9521da177e4SLinus Torvalds reiserfs_unmap_buffer(unbh);
953bd4c625cSLinus Torvalds reiserfs_free_block(th, inode,
954bd4c625cSLinus Torvalds allocated_block_nr, 1);
9551da177e4SLinus Torvalds goto failure;
9561da177e4SLinus Torvalds }
957098297b2SJeff Mahoney /*
958098297b2SJeff Mahoney * it is important the set_buffer_uptodate is done
959098297b2SJeff Mahoney * after the direct2indirect. The buffer might
960098297b2SJeff Mahoney * contain valid data newer than the data on disk
9612c69e205SMatthew Wilcox (Oracle) * (read by read_folio, changed, and then sent here by
962098297b2SJeff Mahoney * writepage). direct2indirect needs to know if unbh
963098297b2SJeff Mahoney * was already up to date, so it can decide if the
964098297b2SJeff Mahoney * data in unbh needs to be replaced with data from
965098297b2SJeff Mahoney * the disk
9661da177e4SLinus Torvalds */
9671da177e4SLinus Torvalds set_buffer_uptodate(unbh);
9681da177e4SLinus Torvalds
969098297b2SJeff Mahoney /*
970098297b2SJeff Mahoney * unbh->b_page == NULL in case of DIRECT_IO request,
971098297b2SJeff Mahoney * this means buffer will disappear shortly, so it
972098297b2SJeff Mahoney * should not be added to
9731da177e4SLinus Torvalds */
9741da177e4SLinus Torvalds if (unbh->b_page) {
975098297b2SJeff Mahoney /*
976098297b2SJeff Mahoney * we've converted the tail, so we must
977098297b2SJeff Mahoney * flush unbh before the transaction commits
9781da177e4SLinus Torvalds */
9791da177e4SLinus Torvalds reiserfs_add_tail_list(inode, unbh);
9801da177e4SLinus Torvalds
981098297b2SJeff Mahoney /*
982098297b2SJeff Mahoney * mark it dirty now to prevent commit_write
983098297b2SJeff Mahoney * from adding this buffer to the inode's
984098297b2SJeff Mahoney * dirty buffer list
9851da177e4SLinus Torvalds */
9861da177e4SLinus Torvalds /*
987098297b2SJeff Mahoney * AKPM: changed __mark_buffer_dirty to
988098297b2SJeff Mahoney * mark_buffer_dirty(). It's still atomic,
989098297b2SJeff Mahoney * but it sets the page dirty too, which makes
990098297b2SJeff Mahoney * it eligible for writeback at any time by the
991098297b2SJeff Mahoney * VM (which was also the case with
992098297b2SJeff Mahoney * __mark_buffer_dirty())
9931da177e4SLinus Torvalds */
9941da177e4SLinus Torvalds mark_buffer_dirty(unbh);
9951da177e4SLinus Torvalds }
9961da177e4SLinus Torvalds } else {
997098297b2SJeff Mahoney /*
998098297b2SJeff Mahoney * append indirect item with holes if needed, when
999098297b2SJeff Mahoney * appending pointer to 'block'-th block use block,
1000098297b2SJeff Mahoney * which is already allocated
1001098297b2SJeff Mahoney */
10021da177e4SLinus Torvalds struct cpu_key tmp_key;
1003098297b2SJeff Mahoney /*
1004098297b2SJeff Mahoney * We use this in case we need to allocate
1005098297b2SJeff Mahoney * only one block which is a fastpath
1006098297b2SJeff Mahoney */
1007098297b2SJeff Mahoney unp_t unf_single = 0;
10081da177e4SLinus Torvalds unp_t *un;
1009bd4c625cSLinus Torvalds __u64 max_to_insert =
1010bd4c625cSLinus Torvalds MAX_ITEM_LEN(inode->i_sb->s_blocksize) /
1011bd4c625cSLinus Torvalds UNFM_P_SIZE;
10121da177e4SLinus Torvalds __u64 blocks_needed;
10131da177e4SLinus Torvalds
10141da177e4SLinus Torvalds RFALSE(pos_in_item != ih_item_len(ih) / UNFM_P_SIZE,
10151da177e4SLinus Torvalds "vs-804: invalid position for append");
1016098297b2SJeff Mahoney /*
1017098297b2SJeff Mahoney * indirect item has to be appended,
1018098297b2SJeff Mahoney * set up key of that position
1019098297b2SJeff Mahoney * (key type is unimportant)
1020098297b2SJeff Mahoney */
10211da177e4SLinus Torvalds make_cpu_key(&tmp_key, inode,
1022bd4c625cSLinus Torvalds le_key_k_offset(version,
1023a228bf8fSJeff Mahoney &ih->ih_key) +
1024bd4c625cSLinus Torvalds op_bytes_number(ih,
1025bd4c625cSLinus Torvalds inode->i_sb->s_blocksize),
1026098297b2SJeff Mahoney TYPE_INDIRECT, 3);
10271da177e4SLinus Torvalds
1028c499ec24SVladimir V. Saveliev RFALSE(cpu_key_k_offset(&tmp_key) > cpu_key_k_offset(&key),
1029c499ec24SVladimir V. Saveliev "green-805: invalid offset");
1030bd4c625cSLinus Torvalds blocks_needed =
1031bd4c625cSLinus Torvalds 1 +
1032bd4c625cSLinus Torvalds ((cpu_key_k_offset(&key) -
1033bd4c625cSLinus Torvalds cpu_key_k_offset(&tmp_key)) >> inode->i_sb->
1034bd4c625cSLinus Torvalds s_blocksize_bits);
10351da177e4SLinus Torvalds
10361da177e4SLinus Torvalds if (blocks_needed == 1) {
10371da177e4SLinus Torvalds un = &unf_single;
10381da177e4SLinus Torvalds } else {
10396396bb22SKees Cook un = kcalloc(min(blocks_needed, max_to_insert),
10406396bb22SKees Cook UNFM_P_SIZE, GFP_NOFS);
10411da177e4SLinus Torvalds if (!un) {
10421da177e4SLinus Torvalds un = &unf_single;
10431da177e4SLinus Torvalds blocks_needed = 1;
10441da177e4SLinus Torvalds max_to_insert = 0;
104501afb213SYan Burman }
10461da177e4SLinus Torvalds }
10471da177e4SLinus Torvalds if (blocks_needed <= max_to_insert) {
1048098297b2SJeff Mahoney /*
1049098297b2SJeff Mahoney * we are going to add target block to
1050098297b2SJeff Mahoney * the file. Use allocated block for that
1051098297b2SJeff Mahoney */
1052bd4c625cSLinus Torvalds un[blocks_needed - 1] =
1053bd4c625cSLinus Torvalds cpu_to_le32(allocated_block_nr);
1054bd4c625cSLinus Torvalds set_block_dev_mapped(bh_result,
1055bd4c625cSLinus Torvalds allocated_block_nr, inode);
10561da177e4SLinus Torvalds set_buffer_new(bh_result);
10571da177e4SLinus Torvalds done = 1;
10581da177e4SLinus Torvalds } else {
10591da177e4SLinus Torvalds /* paste hole to the indirect item */
1060098297b2SJeff Mahoney /*
10615626de1eSLiao Pingfang * If kcalloc failed, max_to_insert becomes
1062098297b2SJeff Mahoney * zero and it means we only have space for
1063098297b2SJeff Mahoney * one block
1064098297b2SJeff Mahoney */
1065bd4c625cSLinus Torvalds blocks_needed =
1066bd4c625cSLinus Torvalds max_to_insert ? max_to_insert : 1;
10671da177e4SLinus Torvalds }
1068bd4c625cSLinus Torvalds retval =
1069bd4c625cSLinus Torvalds reiserfs_paste_into_item(th, &path, &tmp_key, inode,
1070bd4c625cSLinus Torvalds (char *)un,
1071bd4c625cSLinus Torvalds UNFM_P_SIZE *
1072bd4c625cSLinus Torvalds blocks_needed);
10731da177e4SLinus Torvalds
10741da177e4SLinus Torvalds if (blocks_needed != 1)
10751da177e4SLinus Torvalds kfree(un);
10761da177e4SLinus Torvalds
10771da177e4SLinus Torvalds if (retval) {
1078bd4c625cSLinus Torvalds reiserfs_free_block(th, inode,
1079bd4c625cSLinus Torvalds allocated_block_nr, 1);
10801da177e4SLinus Torvalds goto failure;
10811da177e4SLinus Torvalds }
10821da177e4SLinus Torvalds if (!done) {
1083098297b2SJeff Mahoney /*
1084098297b2SJeff Mahoney * We need to mark new file size in case
1085098297b2SJeff Mahoney * this function will be interrupted/aborted
1086098297b2SJeff Mahoney * later on. And we may do this only for
1087098297b2SJeff Mahoney * holes.
1088098297b2SJeff Mahoney */
1089bd4c625cSLinus Torvalds inode->i_size +=
1090bd4c625cSLinus Torvalds inode->i_sb->s_blocksize * blocks_needed;
10911da177e4SLinus Torvalds }
10921da177e4SLinus Torvalds }
10931da177e4SLinus Torvalds
10941da177e4SLinus Torvalds if (done == 1)
10951da177e4SLinus Torvalds break;
10961da177e4SLinus Torvalds
1097098297b2SJeff Mahoney /*
1098098297b2SJeff Mahoney * this loop could log more blocks than we had originally
1099098297b2SJeff Mahoney * asked for. So, we have to allow the transaction to end
1100098297b2SJeff Mahoney * if it is too big or too full. Update the inode so things
1101098297b2SJeff Mahoney * are consistent if we crash before the function returns
1102098297b2SJeff Mahoney * release the path so that anybody waiting on the path before
1103098297b2SJeff Mahoney * ending their transaction will be able to continue.
11041da177e4SLinus Torvalds */
11051da177e4SLinus Torvalds if (journal_transaction_should_end(th, th->t_blocks_allocated)) {
11061da177e4SLinus Torvalds retval = restart_transaction(th, inode, &path);
11071da177e4SLinus Torvalds if (retval)
11081da177e4SLinus Torvalds goto failure;
11091da177e4SLinus Torvalds }
11108ebc4232SFrederic Weisbecker /*
11118ebc4232SFrederic Weisbecker * inserting indirect pointers for a hole can take a
11128ebc4232SFrederic Weisbecker * long time. reschedule if needed and also release the write
11138ebc4232SFrederic Weisbecker * lock for others.
11141da177e4SLinus Torvalds */
1115278f6679SJeff Mahoney reiserfs_cond_resched(inode->i_sb);
11161da177e4SLinus Torvalds
11171da177e4SLinus Torvalds retval = search_for_position_by_key(inode->i_sb, &key, &path);
11181da177e4SLinus Torvalds if (retval == IO_ERROR) {
11191da177e4SLinus Torvalds retval = -EIO;
11201da177e4SLinus Torvalds goto failure;
11211da177e4SLinus Torvalds }
11221da177e4SLinus Torvalds if (retval == POSITION_FOUND) {
112345b03d5eSJeff Mahoney reiserfs_warning(inode->i_sb, "vs-825",
11241da177e4SLinus Torvalds "%K should not be found", &key);
11251da177e4SLinus Torvalds retval = -EEXIST;
11261da177e4SLinus Torvalds if (allocated_block_nr)
1127bd4c625cSLinus Torvalds reiserfs_free_block(th, inode,
1128bd4c625cSLinus Torvalds allocated_block_nr, 1);
11291da177e4SLinus Torvalds pathrelse(&path);
11301da177e4SLinus Torvalds goto failure;
11311da177e4SLinus Torvalds }
11321da177e4SLinus Torvalds bh = get_last_bh(&path);
11334cf5f7adSJeff Mahoney ih = tp_item_head(&path);
11344cf5f7adSJeff Mahoney item = tp_item_body(&path);
11351da177e4SLinus Torvalds pos_in_item = path.pos_in_item;
11361da177e4SLinus Torvalds } while (1);
11371da177e4SLinus Torvalds
11381da177e4SLinus Torvalds retval = 0;
11391da177e4SLinus Torvalds
11401da177e4SLinus Torvalds failure:
11411da177e4SLinus Torvalds if (th && (!dangle || (retval && !th->t_trans_id))) {
11421da177e4SLinus Torvalds int err;
11431da177e4SLinus Torvalds if (th->t_trans_id)
11441da177e4SLinus Torvalds reiserfs_update_sd(th, inode);
11451da177e4SLinus Torvalds err = reiserfs_end_persistent_transaction(th);
11461da177e4SLinus Torvalds if (err)
11471da177e4SLinus Torvalds retval = err;
11481da177e4SLinus Torvalds }
11491da177e4SLinus Torvalds
1150278f6679SJeff Mahoney reiserfs_write_unlock(inode->i_sb);
11511da177e4SLinus Torvalds reiserfs_check_path(&path);
11521da177e4SLinus Torvalds return retval;
11531da177e4SLinus Torvalds }
11541da177e4SLinus Torvalds
reiserfs_readahead(struct readahead_control * rac)1155d4388340SMatthew Wilcox (Oracle) static void reiserfs_readahead(struct readahead_control *rac)
11561da177e4SLinus Torvalds {
1157d4388340SMatthew Wilcox (Oracle) mpage_readahead(rac, reiserfs_get_block);
11581da177e4SLinus Torvalds }
11591da177e4SLinus Torvalds
1160098297b2SJeff Mahoney /*
1161098297b2SJeff Mahoney * Compute real number of used bytes by file
1162098297b2SJeff Mahoney * Following three functions can go away when we'll have enough space in
1163098297b2SJeff Mahoney * stat item
11641da177e4SLinus Torvalds */
real_space_diff(struct inode * inode,int sd_size)11651da177e4SLinus Torvalds static int real_space_diff(struct inode *inode, int sd_size)
11661da177e4SLinus Torvalds {
11671da177e4SLinus Torvalds int bytes;
11681da177e4SLinus Torvalds loff_t blocksize = inode->i_sb->s_blocksize;
11691da177e4SLinus Torvalds
11701da177e4SLinus Torvalds if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode))
11711da177e4SLinus Torvalds return sd_size;
11721da177e4SLinus Torvalds
1173098297b2SJeff Mahoney /*
1174098297b2SJeff Mahoney * End of file is also in full block with indirect reference, so round
1175098297b2SJeff Mahoney * up to the next block.
1176098297b2SJeff Mahoney *
1177098297b2SJeff Mahoney * there is just no way to know if the tail is actually packed
1178098297b2SJeff Mahoney * on the file, so we have to assume it isn't. When we pack the
1179098297b2SJeff Mahoney * tail, we add 4 bytes to pretend there really is an unformatted
1180098297b2SJeff Mahoney * node pointer
11811da177e4SLinus Torvalds */
1182bd4c625cSLinus Torvalds bytes =
1183bd4c625cSLinus Torvalds ((inode->i_size +
1184bd4c625cSLinus Torvalds (blocksize - 1)) >> inode->i_sb->s_blocksize_bits) * UNFM_P_SIZE +
1185bd4c625cSLinus Torvalds sd_size;
11861da177e4SLinus Torvalds return bytes;
11871da177e4SLinus Torvalds }
11881da177e4SLinus Torvalds
to_real_used_space(struct inode * inode,ulong blocks,int sd_size)11891da177e4SLinus Torvalds static inline loff_t to_real_used_space(struct inode *inode, ulong blocks,
11901da177e4SLinus Torvalds int sd_size)
11911da177e4SLinus Torvalds {
11921da177e4SLinus Torvalds if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1193bd4c625cSLinus Torvalds return inode->i_size +
1194bd4c625cSLinus Torvalds (loff_t) (real_space_diff(inode, sd_size));
11951da177e4SLinus Torvalds }
1196bd4c625cSLinus Torvalds return ((loff_t) real_space_diff(inode, sd_size)) +
1197bd4c625cSLinus Torvalds (((loff_t) blocks) << 9);
11981da177e4SLinus Torvalds }
11991da177e4SLinus Torvalds
12001da177e4SLinus Torvalds /* Compute number of blocks used by file in ReiserFS counting */
to_fake_used_blocks(struct inode * inode,int sd_size)12011da177e4SLinus Torvalds static inline ulong to_fake_used_blocks(struct inode *inode, int sd_size)
12021da177e4SLinus Torvalds {
12031da177e4SLinus Torvalds loff_t bytes = inode_get_bytes(inode);
12041da177e4SLinus Torvalds loff_t real_space = real_space_diff(inode, sd_size);
12051da177e4SLinus Torvalds
12061da177e4SLinus Torvalds /* keeps fsck and non-quota versions of reiserfs happy */
12071da177e4SLinus Torvalds if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
12081da177e4SLinus Torvalds bytes += (loff_t) 511;
12091da177e4SLinus Torvalds }
12101da177e4SLinus Torvalds
1211098297b2SJeff Mahoney /*
1212098297b2SJeff Mahoney * files from before the quota patch might i_blocks such that
1213098297b2SJeff Mahoney * bytes < real_space. Deal with that here to prevent it from
1214098297b2SJeff Mahoney * going negative.
12151da177e4SLinus Torvalds */
12161da177e4SLinus Torvalds if (bytes < real_space)
12171da177e4SLinus Torvalds return 0;
12181da177e4SLinus Torvalds return (bytes - real_space) >> 9;
12191da177e4SLinus Torvalds }
12201da177e4SLinus Torvalds
1221098297b2SJeff Mahoney /*
1222098297b2SJeff Mahoney * BAD: new directories have stat data of new type and all other items
1223098297b2SJeff Mahoney * of old type. Version stored in the inode says about body items, so
1224098297b2SJeff Mahoney * in update_stat_data we can not rely on inode, but have to check
1225098297b2SJeff Mahoney * item version directly
1226098297b2SJeff Mahoney */
12271da177e4SLinus Torvalds
1228098297b2SJeff Mahoney /* called by read_locked_inode */
init_inode(struct inode * inode,struct treepath * path)1229fec6d055SJosef "Jeff" Sipek static void init_inode(struct inode *inode, struct treepath *path)
12301da177e4SLinus Torvalds {
12311da177e4SLinus Torvalds struct buffer_head *bh;
12321da177e4SLinus Torvalds struct item_head *ih;
12331da177e4SLinus Torvalds __u32 rdev;
12341da177e4SLinus Torvalds
12351da177e4SLinus Torvalds bh = PATH_PLAST_BUFFER(path);
12364cf5f7adSJeff Mahoney ih = tp_item_head(path);
12371da177e4SLinus Torvalds
1238a228bf8fSJeff Mahoney copy_key(INODE_PKEY(inode), &ih->ih_key);
12391da177e4SLinus Torvalds
1240a228bf8fSJeff Mahoney INIT_LIST_HEAD(&REISERFS_I(inode)->i_prealloc_list);
12411da177e4SLinus Torvalds REISERFS_I(inode)->i_flags = 0;
12421da177e4SLinus Torvalds REISERFS_I(inode)->i_prealloc_block = 0;
12431da177e4SLinus Torvalds REISERFS_I(inode)->i_prealloc_count = 0;
12441da177e4SLinus Torvalds REISERFS_I(inode)->i_trans_id = 0;
12451da177e4SLinus Torvalds REISERFS_I(inode)->i_jl = NULL;
1246068fbb31SAlexey Dobriyan reiserfs_init_xattr_rwsem(inode);
12471da177e4SLinus Torvalds
12481da177e4SLinus Torvalds if (stat_data_v1(ih)) {
1249bd4c625cSLinus Torvalds struct stat_data_v1 *sd =
12504cf5f7adSJeff Mahoney (struct stat_data_v1 *)ih_item_body(bh, ih);
12511da177e4SLinus Torvalds unsigned long blocks;
12521da177e4SLinus Torvalds
12531da177e4SLinus Torvalds set_inode_item_key_version(inode, KEY_FORMAT_3_5);
12541da177e4SLinus Torvalds set_inode_sd_version(inode, STAT_DATA_V1);
12551da177e4SLinus Torvalds inode->i_mode = sd_v1_mode(sd);
1256bfe86848SMiklos Szeredi set_nlink(inode, sd_v1_nlink(sd));
1257df814654SEric W. Biederman i_uid_write(inode, sd_v1_uid(sd));
1258df814654SEric W. Biederman i_gid_write(inode, sd_v1_gid(sd));
12591da177e4SLinus Torvalds inode->i_size = sd_v1_size(sd);
12601da177e4SLinus Torvalds inode->i_atime.tv_sec = sd_v1_atime(sd);
12611da177e4SLinus Torvalds inode->i_mtime.tv_sec = sd_v1_mtime(sd);
1262*ae834901SJeff Layton inode_set_ctime(inode, sd_v1_ctime(sd), 0);
12631da177e4SLinus Torvalds inode->i_atime.tv_nsec = 0;
12641da177e4SLinus Torvalds inode->i_mtime.tv_nsec = 0;
12651da177e4SLinus Torvalds
12661da177e4SLinus Torvalds inode->i_blocks = sd_v1_blocks(sd);
12671da177e4SLinus Torvalds inode->i_generation = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
12681da177e4SLinus Torvalds blocks = (inode->i_size + 511) >> 9;
12691da177e4SLinus Torvalds blocks = _ROUND_UP(blocks, inode->i_sb->s_blocksize >> 9);
1270098297b2SJeff Mahoney
1271098297b2SJeff Mahoney /*
1272098297b2SJeff Mahoney * there was a bug in <=3.5.23 when i_blocks could take
1273098297b2SJeff Mahoney * negative values. Starting from 3.5.17 this value could
1274098297b2SJeff Mahoney * even be stored in stat data. For such files we set
1275098297b2SJeff Mahoney * i_blocks based on file size. Just 2 notes: this can be
1276098297b2SJeff Mahoney * wrong for sparse files. On-disk value will be only
1277098297b2SJeff Mahoney * updated if file's inode will ever change
1278098297b2SJeff Mahoney */
12791da177e4SLinus Torvalds if (inode->i_blocks > blocks) {
12801da177e4SLinus Torvalds inode->i_blocks = blocks;
12811da177e4SLinus Torvalds }
12821da177e4SLinus Torvalds
12831da177e4SLinus Torvalds rdev = sd_v1_rdev(sd);
1284bd4c625cSLinus Torvalds REISERFS_I(inode)->i_first_direct_byte =
1285bd4c625cSLinus Torvalds sd_v1_first_direct_byte(sd);
1286098297b2SJeff Mahoney
1287098297b2SJeff Mahoney /*
1288098297b2SJeff Mahoney * an early bug in the quota code can give us an odd
1289098297b2SJeff Mahoney * number for the block count. This is incorrect, fix it here.
12901da177e4SLinus Torvalds */
12911da177e4SLinus Torvalds if (inode->i_blocks & 1) {
12921da177e4SLinus Torvalds inode->i_blocks++;
12931da177e4SLinus Torvalds }
1294bd4c625cSLinus Torvalds inode_set_bytes(inode,
1295bd4c625cSLinus Torvalds to_real_used_space(inode, inode->i_blocks,
12961da177e4SLinus Torvalds SD_V1_SIZE));
1297098297b2SJeff Mahoney /*
1298098297b2SJeff Mahoney * nopack is initially zero for v1 objects. For v2 objects,
1299098297b2SJeff Mahoney * nopack is initialised from sd_attrs
1300098297b2SJeff Mahoney */
13011da177e4SLinus Torvalds REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
13021da177e4SLinus Torvalds } else {
1303098297b2SJeff Mahoney /*
1304098297b2SJeff Mahoney * new stat data found, but object may have old items
1305098297b2SJeff Mahoney * (directories and symlinks)
1306098297b2SJeff Mahoney */
13074cf5f7adSJeff Mahoney struct stat_data *sd = (struct stat_data *)ih_item_body(bh, ih);
13081da177e4SLinus Torvalds
13091da177e4SLinus Torvalds inode->i_mode = sd_v2_mode(sd);
1310bfe86848SMiklos Szeredi set_nlink(inode, sd_v2_nlink(sd));
1311df814654SEric W. Biederman i_uid_write(inode, sd_v2_uid(sd));
13121da177e4SLinus Torvalds inode->i_size = sd_v2_size(sd);
1313df814654SEric W. Biederman i_gid_write(inode, sd_v2_gid(sd));
13141da177e4SLinus Torvalds inode->i_mtime.tv_sec = sd_v2_mtime(sd);
13151da177e4SLinus Torvalds inode->i_atime.tv_sec = sd_v2_atime(sd);
1316*ae834901SJeff Layton inode_set_ctime(inode, sd_v2_ctime(sd), 0);
13171da177e4SLinus Torvalds inode->i_mtime.tv_nsec = 0;
13181da177e4SLinus Torvalds inode->i_atime.tv_nsec = 0;
13191da177e4SLinus Torvalds inode->i_blocks = sd_v2_blocks(sd);
13201da177e4SLinus Torvalds rdev = sd_v2_rdev(sd);
13211da177e4SLinus Torvalds if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1322bd4c625cSLinus Torvalds inode->i_generation =
1323bd4c625cSLinus Torvalds le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
13241da177e4SLinus Torvalds else
13251da177e4SLinus Torvalds inode->i_generation = sd_v2_generation(sd);
13261da177e4SLinus Torvalds
13271da177e4SLinus Torvalds if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
13281da177e4SLinus Torvalds set_inode_item_key_version(inode, KEY_FORMAT_3_5);
13291da177e4SLinus Torvalds else
13301da177e4SLinus Torvalds set_inode_item_key_version(inode, KEY_FORMAT_3_6);
13311da177e4SLinus Torvalds REISERFS_I(inode)->i_first_direct_byte = 0;
13321da177e4SLinus Torvalds set_inode_sd_version(inode, STAT_DATA_V2);
1333bd4c625cSLinus Torvalds inode_set_bytes(inode,
1334bd4c625cSLinus Torvalds to_real_used_space(inode, inode->i_blocks,
13351da177e4SLinus Torvalds SD_V2_SIZE));
1336098297b2SJeff Mahoney /*
1337098297b2SJeff Mahoney * read persistent inode attributes from sd and initialise
1338098297b2SJeff Mahoney * generic inode flags from them
1339098297b2SJeff Mahoney */
13401da177e4SLinus Torvalds REISERFS_I(inode)->i_attrs = sd_v2_attrs(sd);
13411da177e4SLinus Torvalds sd_attrs_to_i_attrs(sd_v2_attrs(sd), inode);
13421da177e4SLinus Torvalds }
13431da177e4SLinus Torvalds
13441da177e4SLinus Torvalds pathrelse(path);
13451da177e4SLinus Torvalds if (S_ISREG(inode->i_mode)) {
13461da177e4SLinus Torvalds inode->i_op = &reiserfs_file_inode_operations;
13471da177e4SLinus Torvalds inode->i_fop = &reiserfs_file_operations;
13481da177e4SLinus Torvalds inode->i_mapping->a_ops = &reiserfs_address_space_operations;
13491da177e4SLinus Torvalds } else if (S_ISDIR(inode->i_mode)) {
13501da177e4SLinus Torvalds inode->i_op = &reiserfs_dir_inode_operations;
13511da177e4SLinus Torvalds inode->i_fop = &reiserfs_dir_operations;
13521da177e4SLinus Torvalds } else if (S_ISLNK(inode->i_mode)) {
13531da177e4SLinus Torvalds inode->i_op = &reiserfs_symlink_inode_operations;
135421fc61c7SAl Viro inode_nohighmem(inode);
13551da177e4SLinus Torvalds inode->i_mapping->a_ops = &reiserfs_address_space_operations;
13561da177e4SLinus Torvalds } else {
13571da177e4SLinus Torvalds inode->i_blocks = 0;
13581da177e4SLinus Torvalds inode->i_op = &reiserfs_special_inode_operations;
13591da177e4SLinus Torvalds init_special_inode(inode, inode->i_mode, new_decode_dev(rdev));
13601da177e4SLinus Torvalds }
13611da177e4SLinus Torvalds }
13621da177e4SLinus Torvalds
1363098297b2SJeff Mahoney /* update new stat data with inode fields */
inode2sd(void * sd,struct inode * inode,loff_t size)13641da177e4SLinus Torvalds static void inode2sd(void *sd, struct inode *inode, loff_t size)
13651da177e4SLinus Torvalds {
13661da177e4SLinus Torvalds struct stat_data *sd_v2 = (struct stat_data *)sd;
13671da177e4SLinus Torvalds
13681da177e4SLinus Torvalds set_sd_v2_mode(sd_v2, inode->i_mode);
13691da177e4SLinus Torvalds set_sd_v2_nlink(sd_v2, inode->i_nlink);
1370df814654SEric W. Biederman set_sd_v2_uid(sd_v2, i_uid_read(inode));
13711da177e4SLinus Torvalds set_sd_v2_size(sd_v2, size);
1372df814654SEric W. Biederman set_sd_v2_gid(sd_v2, i_gid_read(inode));
13731da177e4SLinus Torvalds set_sd_v2_mtime(sd_v2, inode->i_mtime.tv_sec);
13741da177e4SLinus Torvalds set_sd_v2_atime(sd_v2, inode->i_atime.tv_sec);
1375*ae834901SJeff Layton set_sd_v2_ctime(sd_v2, inode_get_ctime(inode).tv_sec);
13761da177e4SLinus Torvalds set_sd_v2_blocks(sd_v2, to_fake_used_blocks(inode, SD_V2_SIZE));
13771da177e4SLinus Torvalds if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
13781da177e4SLinus Torvalds set_sd_v2_rdev(sd_v2, new_encode_dev(inode->i_rdev));
13791da177e4SLinus Torvalds else
13801da177e4SLinus Torvalds set_sd_v2_generation(sd_v2, inode->i_generation);
1381a480b5beSJan Kara set_sd_v2_attrs(sd_v2, REISERFS_I(inode)->i_attrs);
13821da177e4SLinus Torvalds }
13831da177e4SLinus Torvalds
1384098297b2SJeff Mahoney /* used to copy inode's fields to old stat data */
inode2sd_v1(void * sd,struct inode * inode,loff_t size)13851da177e4SLinus Torvalds static void inode2sd_v1(void *sd, struct inode *inode, loff_t size)
13861da177e4SLinus Torvalds {
13871da177e4SLinus Torvalds struct stat_data_v1 *sd_v1 = (struct stat_data_v1 *)sd;
13881da177e4SLinus Torvalds
13891da177e4SLinus Torvalds set_sd_v1_mode(sd_v1, inode->i_mode);
1390df814654SEric W. Biederman set_sd_v1_uid(sd_v1, i_uid_read(inode));
1391df814654SEric W. Biederman set_sd_v1_gid(sd_v1, i_gid_read(inode));
13921da177e4SLinus Torvalds set_sd_v1_nlink(sd_v1, inode->i_nlink);
13931da177e4SLinus Torvalds set_sd_v1_size(sd_v1, size);
13941da177e4SLinus Torvalds set_sd_v1_atime(sd_v1, inode->i_atime.tv_sec);
1395*ae834901SJeff Layton set_sd_v1_ctime(sd_v1, inode_get_ctime(inode).tv_sec);
13961da177e4SLinus Torvalds set_sd_v1_mtime(sd_v1, inode->i_mtime.tv_sec);
13971da177e4SLinus Torvalds
13981da177e4SLinus Torvalds if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
13991da177e4SLinus Torvalds set_sd_v1_rdev(sd_v1, new_encode_dev(inode->i_rdev));
14001da177e4SLinus Torvalds else
14011da177e4SLinus Torvalds set_sd_v1_blocks(sd_v1, to_fake_used_blocks(inode, SD_V1_SIZE));
14021da177e4SLinus Torvalds
1403098297b2SJeff Mahoney /* Sigh. i_first_direct_byte is back */
1404bd4c625cSLinus Torvalds set_sd_v1_first_direct_byte(sd_v1,
1405bd4c625cSLinus Torvalds REISERFS_I(inode)->i_first_direct_byte);
14061da177e4SLinus Torvalds }
14071da177e4SLinus Torvalds
1408098297b2SJeff Mahoney /*
1409098297b2SJeff Mahoney * NOTE, you must prepare the buffer head before sending it here,
1410098297b2SJeff Mahoney * and then log it after the call
14111da177e4SLinus Torvalds */
update_stat_data(struct treepath * path,struct inode * inode,loff_t size)1412fec6d055SJosef "Jeff" Sipek static void update_stat_data(struct treepath *path, struct inode *inode,
14131da177e4SLinus Torvalds loff_t size)
14141da177e4SLinus Torvalds {
14151da177e4SLinus Torvalds struct buffer_head *bh;
14161da177e4SLinus Torvalds struct item_head *ih;
14171da177e4SLinus Torvalds
14181da177e4SLinus Torvalds bh = PATH_PLAST_BUFFER(path);
14194cf5f7adSJeff Mahoney ih = tp_item_head(path);
14201da177e4SLinus Torvalds
14211da177e4SLinus Torvalds if (!is_statdata_le_ih(ih))
1422c3a9c210SJeff Mahoney reiserfs_panic(inode->i_sb, "vs-13065", "key %k, found item %h",
14231da177e4SLinus Torvalds INODE_PKEY(inode), ih);
14241da177e4SLinus Torvalds
1425098297b2SJeff Mahoney /* path points to old stat data */
14261da177e4SLinus Torvalds if (stat_data_v1(ih)) {
14274cf5f7adSJeff Mahoney inode2sd_v1(ih_item_body(bh, ih), inode, size);
14281da177e4SLinus Torvalds } else {
14294cf5f7adSJeff Mahoney inode2sd(ih_item_body(bh, ih), inode, size);
14301da177e4SLinus Torvalds }
14311da177e4SLinus Torvalds
14321da177e4SLinus Torvalds return;
14331da177e4SLinus Torvalds }
14341da177e4SLinus Torvalds
reiserfs_update_sd_size(struct reiserfs_transaction_handle * th,struct inode * inode,loff_t size)14351da177e4SLinus Torvalds void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
14361da177e4SLinus Torvalds struct inode *inode, loff_t size)
14371da177e4SLinus Torvalds {
14381da177e4SLinus Torvalds struct cpu_key key;
14391da177e4SLinus Torvalds INITIALIZE_PATH(path);
14401da177e4SLinus Torvalds struct buffer_head *bh;
14411da177e4SLinus Torvalds int fs_gen;
14421da177e4SLinus Torvalds struct item_head *ih, tmp_ih;
14431da177e4SLinus Torvalds int retval;
14441da177e4SLinus Torvalds
14451da177e4SLinus Torvalds BUG_ON(!th->t_trans_id);
14461da177e4SLinus Torvalds
1447098297b2SJeff Mahoney /* key type is unimportant */
1448098297b2SJeff Mahoney make_cpu_key(&key, inode, SD_OFFSET, TYPE_STAT_DATA, 3);
14491da177e4SLinus Torvalds
14501da177e4SLinus Torvalds for (;;) {
14511da177e4SLinus Torvalds int pos;
14521da177e4SLinus Torvalds /* look for the object's stat data */
14531da177e4SLinus Torvalds retval = search_item(inode->i_sb, &key, &path);
14541da177e4SLinus Torvalds if (retval == IO_ERROR) {
14550030b645SJeff Mahoney reiserfs_error(inode->i_sb, "vs-13050",
145645b03d5eSJeff Mahoney "i/o failure occurred trying to "
14570030b645SJeff Mahoney "update %K stat data", &key);
14581da177e4SLinus Torvalds return;
14591da177e4SLinus Torvalds }
14601da177e4SLinus Torvalds if (retval == ITEM_NOT_FOUND) {
14611da177e4SLinus Torvalds pos = PATH_LAST_POSITION(&path);
14621da177e4SLinus Torvalds pathrelse(&path);
14631da177e4SLinus Torvalds if (inode->i_nlink == 0) {
14641da177e4SLinus Torvalds /*reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: i_nlink == 0, stat data not found"); */
14651da177e4SLinus Torvalds return;
14661da177e4SLinus Torvalds }
146745b03d5eSJeff Mahoney reiserfs_warning(inode->i_sb, "vs-13060",
146845b03d5eSJeff Mahoney "stat data of object %k (nlink == %d) "
146945b03d5eSJeff Mahoney "not found (pos %d)",
1470bd4c625cSLinus Torvalds INODE_PKEY(inode), inode->i_nlink,
1471bd4c625cSLinus Torvalds pos);
14721da177e4SLinus Torvalds reiserfs_check_path(&path);
14731da177e4SLinus Torvalds return;
14741da177e4SLinus Torvalds }
14751da177e4SLinus Torvalds
1476098297b2SJeff Mahoney /*
1477098297b2SJeff Mahoney * sigh, prepare_for_journal might schedule. When it
1478098297b2SJeff Mahoney * schedules the FS might change. We have to detect that,
1479098297b2SJeff Mahoney * and loop back to the search if the stat data item has moved
14801da177e4SLinus Torvalds */
14811da177e4SLinus Torvalds bh = get_last_bh(&path);
14824cf5f7adSJeff Mahoney ih = tp_item_head(&path);
14831da177e4SLinus Torvalds copy_item_head(&tmp_ih, ih);
14841da177e4SLinus Torvalds fs_gen = get_generation(inode->i_sb);
14851da177e4SLinus Torvalds reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
1486098297b2SJeff Mahoney
1487098297b2SJeff Mahoney /* Stat_data item has been moved after scheduling. */
1488bd4c625cSLinus Torvalds if (fs_changed(fs_gen, inode->i_sb)
1489bd4c625cSLinus Torvalds && item_moved(&tmp_ih, &path)) {
14901da177e4SLinus Torvalds reiserfs_restore_prepared_buffer(inode->i_sb, bh);
1491098297b2SJeff Mahoney continue;
14921da177e4SLinus Torvalds }
14931da177e4SLinus Torvalds break;
14941da177e4SLinus Torvalds }
14951da177e4SLinus Torvalds update_stat_data(&path, inode, size);
149609f1b80bSJeff Mahoney journal_mark_dirty(th, bh);
14971da177e4SLinus Torvalds pathrelse(&path);
14981da177e4SLinus Torvalds return;
14991da177e4SLinus Torvalds }
15001da177e4SLinus Torvalds
1501098297b2SJeff Mahoney /*
1502098297b2SJeff Mahoney * reiserfs_read_locked_inode is called to read the inode off disk, and it
1503098297b2SJeff Mahoney * does a make_bad_inode when things go wrong. But, we need to make sure
1504098297b2SJeff Mahoney * and clear the key in the private portion of the inode, otherwise a
1505098297b2SJeff Mahoney * corresponding iput might try to delete whatever object the inode last
1506098297b2SJeff Mahoney * represented.
15071da177e4SLinus Torvalds */
reiserfs_make_bad_inode(struct inode * inode)1508bd4c625cSLinus Torvalds static void reiserfs_make_bad_inode(struct inode *inode)
1509bd4c625cSLinus Torvalds {
15101da177e4SLinus Torvalds memset(INODE_PKEY(inode), 0, KEY_SIZE);
15111da177e4SLinus Torvalds make_bad_inode(inode);
15121da177e4SLinus Torvalds }
15131da177e4SLinus Torvalds
1514098297b2SJeff Mahoney /*
1515098297b2SJeff Mahoney * initially this function was derived from minix or ext2's analog and
1516098297b2SJeff Mahoney * evolved as the prototype did
1517098297b2SJeff Mahoney */
reiserfs_init_locked_inode(struct inode * inode,void * p)15181da177e4SLinus Torvalds int reiserfs_init_locked_inode(struct inode *inode, void *p)
15191da177e4SLinus Torvalds {
15201da177e4SLinus Torvalds struct reiserfs_iget_args *args = (struct reiserfs_iget_args *)p;
15211da177e4SLinus Torvalds inode->i_ino = args->objectid;
15221da177e4SLinus Torvalds INODE_PKEY(inode)->k_dir_id = cpu_to_le32(args->dirid);
15231da177e4SLinus Torvalds return 0;
15241da177e4SLinus Torvalds }
15251da177e4SLinus Torvalds
1526098297b2SJeff Mahoney /*
1527098297b2SJeff Mahoney * looks for stat data in the tree, and fills up the fields of in-core
1528098297b2SJeff Mahoney * inode stat data fields
1529098297b2SJeff Mahoney */
reiserfs_read_locked_inode(struct inode * inode,struct reiserfs_iget_args * args)1530bd4c625cSLinus Torvalds void reiserfs_read_locked_inode(struct inode *inode,
1531bd4c625cSLinus Torvalds struct reiserfs_iget_args *args)
15321da177e4SLinus Torvalds {
15331da177e4SLinus Torvalds INITIALIZE_PATH(path_to_sd);
15341da177e4SLinus Torvalds struct cpu_key key;
15351da177e4SLinus Torvalds unsigned long dirino;
15361da177e4SLinus Torvalds int retval;
15371da177e4SLinus Torvalds
15381da177e4SLinus Torvalds dirino = args->dirid;
15391da177e4SLinus Torvalds
1540098297b2SJeff Mahoney /*
1541098297b2SJeff Mahoney * set version 1, version 2 could be used too, because stat data
1542098297b2SJeff Mahoney * key is the same in both versions
1543098297b2SJeff Mahoney */
15444443390eSJan Kara _make_cpu_key(&key, KEY_FORMAT_3_5, dirino, inode->i_ino, 0, 0, 3);
15451da177e4SLinus Torvalds
15461da177e4SLinus Torvalds /* look for the object's stat data */
15471da177e4SLinus Torvalds retval = search_item(inode->i_sb, &key, &path_to_sd);
15481da177e4SLinus Torvalds if (retval == IO_ERROR) {
15490030b645SJeff Mahoney reiserfs_error(inode->i_sb, "vs-13070",
155045b03d5eSJeff Mahoney "i/o failure occurred trying to find "
155145b03d5eSJeff Mahoney "stat data of %K", &key);
15521da177e4SLinus Torvalds reiserfs_make_bad_inode(inode);
15531da177e4SLinus Torvalds return;
15541da177e4SLinus Torvalds }
1555098297b2SJeff Mahoney
15561da177e4SLinus Torvalds /* a stale NFS handle can trigger this without it being an error */
1557098297b2SJeff Mahoney if (retval != ITEM_FOUND) {
15581da177e4SLinus Torvalds pathrelse(&path_to_sd);
15591da177e4SLinus Torvalds reiserfs_make_bad_inode(inode);
15606d6b77f1SMiklos Szeredi clear_nlink(inode);
15611da177e4SLinus Torvalds return;
15621da177e4SLinus Torvalds }
15631da177e4SLinus Torvalds
15641da177e4SLinus Torvalds init_inode(inode, &path_to_sd);
15651da177e4SLinus Torvalds
1566098297b2SJeff Mahoney /*
1567098297b2SJeff Mahoney * It is possible that knfsd is trying to access inode of a file
1568098297b2SJeff Mahoney * that is being removed from the disk by some other thread. As we
1569098297b2SJeff Mahoney * update sd on unlink all that is required is to check for nlink
1570098297b2SJeff Mahoney * here. This bug was first found by Sizif when debugging
1571098297b2SJeff Mahoney * SquidNG/Butterfly, forgotten, and found again after Philippe
1572098297b2SJeff Mahoney * Gramoulle <philippe.gramoulle@mmania.com> reproduced it.
15731da177e4SLinus Torvalds
1574098297b2SJeff Mahoney * More logical fix would require changes in fs/inode.c:iput() to
1575098297b2SJeff Mahoney * remove inode from hash-table _after_ fs cleaned disk stuff up and
1576098297b2SJeff Mahoney * in iget() to return NULL if I_FREEING inode is found in
1577098297b2SJeff Mahoney * hash-table.
1578098297b2SJeff Mahoney */
1579098297b2SJeff Mahoney
1580098297b2SJeff Mahoney /*
1581098297b2SJeff Mahoney * Currently there is one place where it's ok to meet inode with
1582098297b2SJeff Mahoney * nlink==0: processing of open-unlinked and half-truncated files
1583098297b2SJeff Mahoney * during mount (fs/reiserfs/super.c:finish_unfinished()).
1584098297b2SJeff Mahoney */
15851da177e4SLinus Torvalds if ((inode->i_nlink == 0) &&
15861da177e4SLinus Torvalds !REISERFS_SB(inode->i_sb)->s_is_unlinked_ok) {
158745b03d5eSJeff Mahoney reiserfs_warning(inode->i_sb, "vs-13075",
15881da177e4SLinus Torvalds "dead inode read from disk %K. "
15891da177e4SLinus Torvalds "This is likely to be race with knfsd. Ignore",
15901da177e4SLinus Torvalds &key);
15911da177e4SLinus Torvalds reiserfs_make_bad_inode(inode);
15921da177e4SLinus Torvalds }
15931da177e4SLinus Torvalds
1594098297b2SJeff Mahoney /* init inode should be relsing */
1595098297b2SJeff Mahoney reiserfs_check_path(&path_to_sd);
15961da177e4SLinus Torvalds
15974482a087SChristoph Hellwig /*
15984482a087SChristoph Hellwig * Stat data v1 doesn't support ACLs.
15994482a087SChristoph Hellwig */
16004482a087SChristoph Hellwig if (get_inode_sd_version(inode) == STAT_DATA_V1)
16014482a087SChristoph Hellwig cache_no_acl(inode);
16021da177e4SLinus Torvalds }
16031da177e4SLinus Torvalds
1604098297b2SJeff Mahoney /*
16051da177e4SLinus Torvalds * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked().
16061da177e4SLinus Torvalds *
16071da177e4SLinus Torvalds * @inode: inode from hash table to check
16081da177e4SLinus Torvalds * @opaque: "cookie" passed to iget5_locked(). This is &reiserfs_iget_args.
16091da177e4SLinus Torvalds *
16101da177e4SLinus Torvalds * This function is called by iget5_locked() to distinguish reiserfs inodes
16111da177e4SLinus Torvalds * having the same inode numbers. Such inodes can only exist due to some
16121da177e4SLinus Torvalds * error condition. One of them should be bad. Inodes with identical
16131da177e4SLinus Torvalds * inode numbers (objectids) are distinguished by parent directory ids.
16141da177e4SLinus Torvalds *
16151da177e4SLinus Torvalds */
reiserfs_find_actor(struct inode * inode,void * opaque)16161da177e4SLinus Torvalds int reiserfs_find_actor(struct inode *inode, void *opaque)
16171da177e4SLinus Torvalds {
16181da177e4SLinus Torvalds struct reiserfs_iget_args *args;
16191da177e4SLinus Torvalds
16201da177e4SLinus Torvalds args = opaque;
16211da177e4SLinus Torvalds /* args is already in CPU order */
16221da177e4SLinus Torvalds return (inode->i_ino == args->objectid) &&
16231da177e4SLinus Torvalds (le32_to_cpu(INODE_PKEY(inode)->k_dir_id) == args->dirid);
16241da177e4SLinus Torvalds }
16251da177e4SLinus Torvalds
reiserfs_iget(struct super_block * s,const struct cpu_key * key)16261da177e4SLinus Torvalds struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key)
16271da177e4SLinus Torvalds {
16281da177e4SLinus Torvalds struct inode *inode;
16291da177e4SLinus Torvalds struct reiserfs_iget_args args;
1630278f6679SJeff Mahoney int depth;
16311da177e4SLinus Torvalds
16321da177e4SLinus Torvalds args.objectid = key->on_disk_key.k_objectid;
16331da177e4SLinus Torvalds args.dirid = key->on_disk_key.k_dir_id;
1634278f6679SJeff Mahoney depth = reiserfs_write_unlock_nested(s);
16351da177e4SLinus Torvalds inode = iget5_locked(s, key->on_disk_key.k_objectid,
1636bd4c625cSLinus Torvalds reiserfs_find_actor, reiserfs_init_locked_inode,
1637bd4c625cSLinus Torvalds (void *)(&args));
1638278f6679SJeff Mahoney reiserfs_write_lock_nested(s, depth);
16391da177e4SLinus Torvalds if (!inode)
16401da177e4SLinus Torvalds return ERR_PTR(-ENOMEM);
16411da177e4SLinus Torvalds
16421da177e4SLinus Torvalds if (inode->i_state & I_NEW) {
16431da177e4SLinus Torvalds reiserfs_read_locked_inode(inode, &args);
16441da177e4SLinus Torvalds unlock_new_inode(inode);
16451da177e4SLinus Torvalds }
16461da177e4SLinus Torvalds
16471da177e4SLinus Torvalds if (comp_short_keys(INODE_PKEY(inode), key) || is_bad_inode(inode)) {
16481da177e4SLinus Torvalds /* either due to i/o error or a stale NFS handle */
16491da177e4SLinus Torvalds iput(inode);
16501da177e4SLinus Torvalds inode = NULL;
16511da177e4SLinus Torvalds }
16521da177e4SLinus Torvalds return inode;
16531da177e4SLinus Torvalds }
16541da177e4SLinus Torvalds
reiserfs_get_dentry(struct super_block * sb,u32 objectid,u32 dir_id,u32 generation)1655be55caf1SChristoph Hellwig static struct dentry *reiserfs_get_dentry(struct super_block *sb,
1656be55caf1SChristoph Hellwig u32 objectid, u32 dir_id, u32 generation)
1657be55caf1SChristoph Hellwig
16581da177e4SLinus Torvalds {
16591da177e4SLinus Torvalds struct cpu_key key;
16601da177e4SLinus Torvalds struct inode *inode;
16611da177e4SLinus Torvalds
1662be55caf1SChristoph Hellwig key.on_disk_key.k_objectid = objectid;
1663be55caf1SChristoph Hellwig key.on_disk_key.k_dir_id = dir_id;
16641da177e4SLinus Torvalds reiserfs_write_lock(sb);
16651da177e4SLinus Torvalds inode = reiserfs_iget(sb, &key);
1666be55caf1SChristoph Hellwig if (inode && !IS_ERR(inode) && generation != 0 &&
1667be55caf1SChristoph Hellwig generation != inode->i_generation) {
16681da177e4SLinus Torvalds iput(inode);
16691da177e4SLinus Torvalds inode = NULL;
16701da177e4SLinus Torvalds }
16711da177e4SLinus Torvalds reiserfs_write_unlock(sb);
167244003728SChristoph Hellwig
167344003728SChristoph Hellwig return d_obtain_alias(inode);
16741da177e4SLinus Torvalds }
16751da177e4SLinus Torvalds
reiserfs_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)1676be55caf1SChristoph Hellwig struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1677be55caf1SChristoph Hellwig int fh_len, int fh_type)
1678bd4c625cSLinus Torvalds {
1679098297b2SJeff Mahoney /*
1680098297b2SJeff Mahoney * fhtype happens to reflect the number of u32s encoded.
16811da177e4SLinus Torvalds * due to a bug in earlier code, fhtype might indicate there
16821da177e4SLinus Torvalds * are more u32s then actually fitted.
16831da177e4SLinus Torvalds * so if fhtype seems to be more than len, reduce fhtype.
16841da177e4SLinus Torvalds * Valid types are:
16851da177e4SLinus Torvalds * 2 - objectid + dir_id - legacy support
16861da177e4SLinus Torvalds * 3 - objectid + dir_id + generation
16871da177e4SLinus Torvalds * 4 - objectid + dir_id + objectid and dirid of parent - legacy
16881da177e4SLinus Torvalds * 5 - objectid + dir_id + generation + objectid and dirid of parent
16891da177e4SLinus Torvalds * 6 - as above plus generation of directory
16901da177e4SLinus Torvalds * 6 does not fit in NFSv2 handles
16911da177e4SLinus Torvalds */
1692be55caf1SChristoph Hellwig if (fh_type > fh_len) {
1693be55caf1SChristoph Hellwig if (fh_type != 6 || fh_len != 5)
169445b03d5eSJeff Mahoney reiserfs_warning(sb, "reiserfs-13077",
1695bd4c625cSLinus Torvalds "nfsd/reiserfs, fhtype=%d, len=%d - odd",
1696be55caf1SChristoph Hellwig fh_type, fh_len);
169735c2a7f4SHugh Dickins fh_type = fh_len;
16981da177e4SLinus Torvalds }
169935c2a7f4SHugh Dickins if (fh_len < 2)
170035c2a7f4SHugh Dickins return NULL;
17011da177e4SLinus Torvalds
1702be55caf1SChristoph Hellwig return reiserfs_get_dentry(sb, fid->raw[0], fid->raw[1],
1703be55caf1SChristoph Hellwig (fh_type == 3 || fh_type >= 5) ? fid->raw[2] : 0);
17041da177e4SLinus Torvalds }
1705be55caf1SChristoph Hellwig
reiserfs_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)1706be55caf1SChristoph Hellwig struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,
1707be55caf1SChristoph Hellwig int fh_len, int fh_type)
1708be55caf1SChristoph Hellwig {
170935c2a7f4SHugh Dickins if (fh_type > fh_len)
171035c2a7f4SHugh Dickins fh_type = fh_len;
1711be55caf1SChristoph Hellwig if (fh_type < 4)
1712be55caf1SChristoph Hellwig return NULL;
1713be55caf1SChristoph Hellwig
1714be55caf1SChristoph Hellwig return reiserfs_get_dentry(sb,
1715be55caf1SChristoph Hellwig (fh_type >= 5) ? fid->raw[3] : fid->raw[2],
1716be55caf1SChristoph Hellwig (fh_type >= 5) ? fid->raw[4] : fid->raw[3],
1717be55caf1SChristoph Hellwig (fh_type == 6) ? fid->raw[5] : 0);
17181da177e4SLinus Torvalds }
17191da177e4SLinus Torvalds
reiserfs_encode_fh(struct inode * inode,__u32 * data,int * lenp,struct inode * parent)1720b0b0382bSAl Viro int reiserfs_encode_fh(struct inode *inode, __u32 * data, int *lenp,
1721b0b0382bSAl Viro struct inode *parent)
1722bd4c625cSLinus Torvalds {
17231da177e4SLinus Torvalds int maxlen = *lenp;
17241da177e4SLinus Torvalds
1725b0b0382bSAl Viro if (parent && (maxlen < 5)) {
17265fe0c237SAneesh Kumar K.V *lenp = 5;
172794e07a75SNamjae Jeon return FILEID_INVALID;
17285fe0c237SAneesh Kumar K.V } else if (maxlen < 3) {
17295fe0c237SAneesh Kumar K.V *lenp = 3;
173094e07a75SNamjae Jeon return FILEID_INVALID;
17315fe0c237SAneesh Kumar K.V }
17321da177e4SLinus Torvalds
17331da177e4SLinus Torvalds data[0] = inode->i_ino;
17341da177e4SLinus Torvalds data[1] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
17351da177e4SLinus Torvalds data[2] = inode->i_generation;
17361da177e4SLinus Torvalds *lenp = 3;
1737b0b0382bSAl Viro if (parent) {
1738b0b0382bSAl Viro data[3] = parent->i_ino;
1739b0b0382bSAl Viro data[4] = le32_to_cpu(INODE_PKEY(parent)->k_dir_id);
17401da177e4SLinus Torvalds *lenp = 5;
17411da177e4SLinus Torvalds if (maxlen >= 6) {
1742b0b0382bSAl Viro data[5] = parent->i_generation;
17431da177e4SLinus Torvalds *lenp = 6;
17441da177e4SLinus Torvalds }
1745b0b0382bSAl Viro }
17461da177e4SLinus Torvalds return *lenp;
17471da177e4SLinus Torvalds }
17481da177e4SLinus Torvalds
1749098297b2SJeff Mahoney /*
1750098297b2SJeff Mahoney * looks for stat data, then copies fields to it, marks the buffer
1751098297b2SJeff Mahoney * containing stat data as dirty
1752098297b2SJeff Mahoney */
1753098297b2SJeff Mahoney /*
1754098297b2SJeff Mahoney * reiserfs inodes are never really dirty, since the dirty inode call
1755098297b2SJeff Mahoney * always logs them. This call allows the VFS inode marking routines
1756098297b2SJeff Mahoney * to properly mark inodes for datasync and such, but only actually
1757098297b2SJeff Mahoney * does something when called for a synchronous update.
17581da177e4SLinus Torvalds */
reiserfs_write_inode(struct inode * inode,struct writeback_control * wbc)1759a9185b41SChristoph Hellwig int reiserfs_write_inode(struct inode *inode, struct writeback_control *wbc)
1760bd4c625cSLinus Torvalds {
17611da177e4SLinus Torvalds struct reiserfs_transaction_handle th;
17621da177e4SLinus Torvalds int jbegin_count = 1;
17631da177e4SLinus Torvalds
1764bc98a42cSDavid Howells if (sb_rdonly(inode->i_sb))
17651da177e4SLinus Torvalds return -EROFS;
1766098297b2SJeff Mahoney /*
1767098297b2SJeff Mahoney * memory pressure can sometimes initiate write_inode calls with
1768098297b2SJeff Mahoney * sync == 1,
1769098297b2SJeff Mahoney * these cases are just when the system needs ram, not when the
1770098297b2SJeff Mahoney * inode needs to reach disk for safety, and they can safely be
1771098297b2SJeff Mahoney * ignored because the altered inode has already been logged.
17721da177e4SLinus Torvalds */
1773a9185b41SChristoph Hellwig if (wbc->sync_mode == WB_SYNC_ALL && !(current->flags & PF_MEMALLOC)) {
17741da177e4SLinus Torvalds reiserfs_write_lock(inode->i_sb);
17751da177e4SLinus Torvalds if (!journal_begin(&th, inode->i_sb, jbegin_count)) {
17761da177e4SLinus Torvalds reiserfs_update_sd(&th, inode);
177758d85426SJeff Mahoney journal_end_sync(&th);
17781da177e4SLinus Torvalds }
17791da177e4SLinus Torvalds reiserfs_write_unlock(inode->i_sb);
17801da177e4SLinus Torvalds }
17811da177e4SLinus Torvalds return 0;
17821da177e4SLinus Torvalds }
17831da177e4SLinus Torvalds
1784098297b2SJeff Mahoney /*
1785098297b2SJeff Mahoney * stat data of new object is inserted already, this inserts the item
1786098297b2SJeff Mahoney * containing "." and ".." entries
1787098297b2SJeff Mahoney */
reiserfs_new_directory(struct reiserfs_transaction_handle * th,struct inode * inode,struct item_head * ih,struct treepath * path,struct inode * dir)17881da177e4SLinus Torvalds static int reiserfs_new_directory(struct reiserfs_transaction_handle *th,
17891da177e4SLinus Torvalds struct inode *inode,
1790fec6d055SJosef "Jeff" Sipek struct item_head *ih, struct treepath *path,
17911da177e4SLinus Torvalds struct inode *dir)
17921da177e4SLinus Torvalds {
17931da177e4SLinus Torvalds struct super_block *sb = th->t_super;
17941da177e4SLinus Torvalds char empty_dir[EMPTY_DIR_SIZE];
17951da177e4SLinus Torvalds char *body = empty_dir;
17961da177e4SLinus Torvalds struct cpu_key key;
17971da177e4SLinus Torvalds int retval;
17981da177e4SLinus Torvalds
17991da177e4SLinus Torvalds BUG_ON(!th->t_trans_id);
18001da177e4SLinus Torvalds
18011da177e4SLinus Torvalds _make_cpu_key(&key, KEY_FORMAT_3_5, le32_to_cpu(ih->ih_key.k_dir_id),
1802bd4c625cSLinus Torvalds le32_to_cpu(ih->ih_key.k_objectid), DOT_OFFSET,
1803bd4c625cSLinus Torvalds TYPE_DIRENTRY, 3 /*key length */ );
18041da177e4SLinus Torvalds
1805098297b2SJeff Mahoney /*
1806098297b2SJeff Mahoney * compose item head for new item. Directories consist of items of
1807098297b2SJeff Mahoney * old type (ITEM_VERSION_1). Do not set key (second arg is 0), it
1808098297b2SJeff Mahoney * is done by reiserfs_new_inode
1809098297b2SJeff Mahoney */
18101da177e4SLinus Torvalds if (old_format_only(sb)) {
1811bd4c625cSLinus Torvalds make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1812bd4c625cSLinus Torvalds TYPE_DIRENTRY, EMPTY_DIR_SIZE_V1, 2);
18131da177e4SLinus Torvalds
1814bd4c625cSLinus Torvalds make_empty_dir_item_v1(body, ih->ih_key.k_dir_id,
1815bd4c625cSLinus Torvalds ih->ih_key.k_objectid,
18161da177e4SLinus Torvalds INODE_PKEY(dir)->k_dir_id,
18171da177e4SLinus Torvalds INODE_PKEY(dir)->k_objectid);
18181da177e4SLinus Torvalds } else {
1819bd4c625cSLinus Torvalds make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1820bd4c625cSLinus Torvalds TYPE_DIRENTRY, EMPTY_DIR_SIZE, 2);
18211da177e4SLinus Torvalds
1822bd4c625cSLinus Torvalds make_empty_dir_item(body, ih->ih_key.k_dir_id,
1823bd4c625cSLinus Torvalds ih->ih_key.k_objectid,
18241da177e4SLinus Torvalds INODE_PKEY(dir)->k_dir_id,
18251da177e4SLinus Torvalds INODE_PKEY(dir)->k_objectid);
18261da177e4SLinus Torvalds }
18271da177e4SLinus Torvalds
18281da177e4SLinus Torvalds /* look for place in the tree for new item */
18291da177e4SLinus Torvalds retval = search_item(sb, &key, path);
18301da177e4SLinus Torvalds if (retval == IO_ERROR) {
18310030b645SJeff Mahoney reiserfs_error(sb, "vs-13080",
18321da177e4SLinus Torvalds "i/o failure occurred creating new directory");
18331da177e4SLinus Torvalds return -EIO;
18341da177e4SLinus Torvalds }
18351da177e4SLinus Torvalds if (retval == ITEM_FOUND) {
18361da177e4SLinus Torvalds pathrelse(path);
183745b03d5eSJeff Mahoney reiserfs_warning(sb, "vs-13070",
1838bd4c625cSLinus Torvalds "object with this key exists (%k)",
1839bd4c625cSLinus Torvalds &(ih->ih_key));
18401da177e4SLinus Torvalds return -EEXIST;
18411da177e4SLinus Torvalds }
18421da177e4SLinus Torvalds
18431da177e4SLinus Torvalds /* insert item, that is empty directory item */
18441da177e4SLinus Torvalds return reiserfs_insert_item(th, path, &key, ih, inode, body);
18451da177e4SLinus Torvalds }
18461da177e4SLinus Torvalds
1847098297b2SJeff Mahoney /*
1848098297b2SJeff Mahoney * stat data of object has been inserted, this inserts the item
1849098297b2SJeff Mahoney * containing the body of symlink
1850098297b2SJeff Mahoney */
reiserfs_new_symlink(struct reiserfs_transaction_handle * th,struct inode * inode,struct item_head * ih,struct treepath * path,const char * symname,int item_len)1851098297b2SJeff Mahoney static int reiserfs_new_symlink(struct reiserfs_transaction_handle *th,
1852098297b2SJeff Mahoney struct inode *inode,
18531da177e4SLinus Torvalds struct item_head *ih,
1854fec6d055SJosef "Jeff" Sipek struct treepath *path, const char *symname,
1855bd4c625cSLinus Torvalds int item_len)
18561da177e4SLinus Torvalds {
18571da177e4SLinus Torvalds struct super_block *sb = th->t_super;
18581da177e4SLinus Torvalds struct cpu_key key;
18591da177e4SLinus Torvalds int retval;
18601da177e4SLinus Torvalds
18611da177e4SLinus Torvalds BUG_ON(!th->t_trans_id);
18621da177e4SLinus Torvalds
18631da177e4SLinus Torvalds _make_cpu_key(&key, KEY_FORMAT_3_5,
18641da177e4SLinus Torvalds le32_to_cpu(ih->ih_key.k_dir_id),
18651da177e4SLinus Torvalds le32_to_cpu(ih->ih_key.k_objectid),
18661da177e4SLinus Torvalds 1, TYPE_DIRECT, 3 /*key length */ );
18671da177e4SLinus Torvalds
1868bd4c625cSLinus Torvalds make_le_item_head(ih, NULL, KEY_FORMAT_3_5, 1, TYPE_DIRECT, item_len,
1869bd4c625cSLinus Torvalds 0 /*free_space */ );
18701da177e4SLinus Torvalds
18711da177e4SLinus Torvalds /* look for place in the tree for new item */
18721da177e4SLinus Torvalds retval = search_item(sb, &key, path);
18731da177e4SLinus Torvalds if (retval == IO_ERROR) {
18740030b645SJeff Mahoney reiserfs_error(sb, "vs-13080",
18751da177e4SLinus Torvalds "i/o failure occurred creating new symlink");
18761da177e4SLinus Torvalds return -EIO;
18771da177e4SLinus Torvalds }
18781da177e4SLinus Torvalds if (retval == ITEM_FOUND) {
18791da177e4SLinus Torvalds pathrelse(path);
188045b03d5eSJeff Mahoney reiserfs_warning(sb, "vs-13080",
1881bd4c625cSLinus Torvalds "object with this key exists (%k)",
1882bd4c625cSLinus Torvalds &(ih->ih_key));
18831da177e4SLinus Torvalds return -EEXIST;
18841da177e4SLinus Torvalds }
18851da177e4SLinus Torvalds
18861da177e4SLinus Torvalds /* insert item, that is body of symlink */
18871da177e4SLinus Torvalds return reiserfs_insert_item(th, path, &key, ih, inode, symname);
18881da177e4SLinus Torvalds }
18891da177e4SLinus Torvalds
1890098297b2SJeff Mahoney /*
1891098297b2SJeff Mahoney * inserts the stat data into the tree, and then calls
1892098297b2SJeff Mahoney * reiserfs_new_directory (to insert ".", ".." item if new object is
1893098297b2SJeff Mahoney * directory) or reiserfs_new_symlink (to insert symlink body if new
1894098297b2SJeff Mahoney * object is symlink) or nothing (if new object is regular file)
18951da177e4SLinus Torvalds
1896098297b2SJeff Mahoney * NOTE! uid and gid must already be set in the inode. If we return
1897098297b2SJeff Mahoney * non-zero due to an error, we have to drop the quota previously allocated
1898098297b2SJeff Mahoney * for the fresh inode. This can only be done outside a transaction, so
1899098297b2SJeff Mahoney * if we return non-zero, we also end the transaction.
1900098297b2SJeff Mahoney *
1901098297b2SJeff Mahoney * @th: active transaction handle
1902098297b2SJeff Mahoney * @dir: parent directory for new inode
1903098297b2SJeff Mahoney * @mode: mode of new inode
1904098297b2SJeff Mahoney * @symname: symlink contents if inode is symlink
1905098297b2SJeff Mahoney * @isize: 0 for regular file, EMPTY_DIR_SIZE for dirs, strlen(symname) for
1906098297b2SJeff Mahoney * symlinks
1907098297b2SJeff Mahoney * @inode: inode to be filled
1908098297b2SJeff Mahoney * @security: optional security context to associate with this inode
1909098297b2SJeff Mahoney */
reiserfs_new_inode(struct reiserfs_transaction_handle * th,struct inode * dir,umode_t mode,const char * symname,loff_t i_size,struct dentry * dentry,struct inode * inode,struct reiserfs_security_handle * security)19101da177e4SLinus Torvalds int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
19118e071892SAl Viro struct inode *dir, umode_t mode, const char *symname,
19121da177e4SLinus Torvalds /* 0 for regular, EMTRY_DIR_SIZE for dirs,
19131da177e4SLinus Torvalds strlen (symname) for symlinks) */
19141da177e4SLinus Torvalds loff_t i_size, struct dentry *dentry,
191557fe60dfSJeff Mahoney struct inode *inode,
191657fe60dfSJeff Mahoney struct reiserfs_security_handle *security)
19171da177e4SLinus Torvalds {
1918d2d0395fSJeff Mahoney struct super_block *sb = dir->i_sb;
1919c1eaa26bSAl Viro struct reiserfs_iget_args args;
19201da177e4SLinus Torvalds INITIALIZE_PATH(path_to_key);
19211da177e4SLinus Torvalds struct cpu_key key;
19221da177e4SLinus Torvalds struct item_head ih;
19231da177e4SLinus Torvalds struct stat_data sd;
19241da177e4SLinus Torvalds int retval;
19251da177e4SLinus Torvalds int err;
1926278f6679SJeff Mahoney int depth;
19271da177e4SLinus Torvalds
19281da177e4SLinus Torvalds BUG_ON(!th->t_trans_id);
19291da177e4SLinus Torvalds
1930d2d0395fSJeff Mahoney depth = reiserfs_write_unlock_nested(sb);
193163936ddaSChristoph Hellwig err = dquot_alloc_inode(inode);
1932d2d0395fSJeff Mahoney reiserfs_write_lock_nested(sb, depth);
193363936ddaSChristoph Hellwig if (err)
19341da177e4SLinus Torvalds goto out_end_trans;
1935585b7747SEric Sesterhenn if (!dir->i_nlink) {
19361da177e4SLinus Torvalds err = -EPERM;
19371da177e4SLinus Torvalds goto out_bad_inode;
19381da177e4SLinus Torvalds }
19391da177e4SLinus Torvalds
19401da177e4SLinus Torvalds /* item head of new item */
19411da177e4SLinus Torvalds ih.ih_key.k_dir_id = reiserfs_choose_packing(dir);
19421da177e4SLinus Torvalds ih.ih_key.k_objectid = cpu_to_le32(reiserfs_get_unused_objectid(th));
19431da177e4SLinus Torvalds if (!ih.ih_key.k_objectid) {
19441da177e4SLinus Torvalds err = -ENOMEM;
19451da177e4SLinus Torvalds goto out_bad_inode;
19461da177e4SLinus Torvalds }
1947c1eaa26bSAl Viro args.objectid = inode->i_ino = le32_to_cpu(ih.ih_key.k_objectid);
19482f1169e2SAl Viro if (old_format_only(sb))
19492f1169e2SAl Viro make_le_item_head(&ih, NULL, KEY_FORMAT_3_5, SD_OFFSET,
19502f1169e2SAl Viro TYPE_STAT_DATA, SD_V1_SIZE, MAX_US_INT);
19512f1169e2SAl Viro else
19522f1169e2SAl Viro make_le_item_head(&ih, NULL, KEY_FORMAT_3_6, SD_OFFSET,
19532f1169e2SAl Viro TYPE_STAT_DATA, SD_SIZE, MAX_US_INT);
1954a228bf8fSJeff Mahoney memcpy(INODE_PKEY(inode), &ih.ih_key, KEY_SIZE);
1955c1eaa26bSAl Viro args.dirid = le32_to_cpu(ih.ih_key.k_dir_id);
1956a1457c0cSJeff Mahoney
1957278f6679SJeff Mahoney depth = reiserfs_write_unlock_nested(inode->i_sb);
1958a1457c0cSJeff Mahoney err = insert_inode_locked4(inode, args.objectid,
1959a1457c0cSJeff Mahoney reiserfs_find_actor, &args);
1960278f6679SJeff Mahoney reiserfs_write_lock_nested(inode->i_sb, depth);
1961a1457c0cSJeff Mahoney if (err) {
1962c1eaa26bSAl Viro err = -EINVAL;
1963c1eaa26bSAl Viro goto out_bad_inode;
1964c1eaa26bSAl Viro }
1965a1457c0cSJeff Mahoney
19661da177e4SLinus Torvalds if (old_format_only(sb))
1967098297b2SJeff Mahoney /*
1968098297b2SJeff Mahoney * not a perfect generation count, as object ids can be reused,
1969098297b2SJeff Mahoney * but this is as good as reiserfs can do right now.
1970098297b2SJeff Mahoney * note that the private part of inode isn't filled in yet,
1971098297b2SJeff Mahoney * we have to use the directory.
19721da177e4SLinus Torvalds */
19731da177e4SLinus Torvalds inode->i_generation = le32_to_cpu(INODE_PKEY(dir)->k_objectid);
19741da177e4SLinus Torvalds else
19751da177e4SLinus Torvalds #if defined( USE_INODE_GENERATION_COUNTER )
1976bd4c625cSLinus Torvalds inode->i_generation =
1977bd4c625cSLinus Torvalds le32_to_cpu(REISERFS_SB(sb)->s_rs->s_inode_generation);
19781da177e4SLinus Torvalds #else
19791da177e4SLinus Torvalds inode->i_generation = ++event;
19801da177e4SLinus Torvalds #endif
19811da177e4SLinus Torvalds
19821da177e4SLinus Torvalds /* fill stat data */
1983bfe86848SMiklos Szeredi set_nlink(inode, (S_ISDIR(mode) ? 2 : 1));
19841da177e4SLinus Torvalds
19851da177e4SLinus Torvalds /* uid and gid must already be set by the caller for quota init */
19861da177e4SLinus Torvalds
1987*ae834901SJeff Layton inode->i_mtime = inode->i_atime = inode_set_ctime_current(inode);
19881da177e4SLinus Torvalds inode->i_size = i_size;
19891da177e4SLinus Torvalds inode->i_blocks = 0;
19901da177e4SLinus Torvalds inode->i_bytes = 0;
19911da177e4SLinus Torvalds REISERFS_I(inode)->i_first_direct_byte = S_ISLNK(mode) ? 1 :
19921da177e4SLinus Torvalds U32_MAX /*NO_BYTES_IN_DIRECT_ITEM */ ;
19931da177e4SLinus Torvalds
1994a228bf8fSJeff Mahoney INIT_LIST_HEAD(&REISERFS_I(inode)->i_prealloc_list);
19951da177e4SLinus Torvalds REISERFS_I(inode)->i_flags = 0;
19961da177e4SLinus Torvalds REISERFS_I(inode)->i_prealloc_block = 0;
19971da177e4SLinus Torvalds REISERFS_I(inode)->i_prealloc_count = 0;
19981da177e4SLinus Torvalds REISERFS_I(inode)->i_trans_id = 0;
19991da177e4SLinus Torvalds REISERFS_I(inode)->i_jl = NULL;
20001da177e4SLinus Torvalds REISERFS_I(inode)->i_attrs =
20011da177e4SLinus Torvalds REISERFS_I(dir)->i_attrs & REISERFS_INHERIT_MASK;
20021da177e4SLinus Torvalds sd_attrs_to_i_attrs(REISERFS_I(inode)->i_attrs, inode);
2003068fbb31SAlexey Dobriyan reiserfs_init_xattr_rwsem(inode);
20041da177e4SLinus Torvalds
20051da177e4SLinus Torvalds /* key to search for correct place for new stat data */
20061da177e4SLinus Torvalds _make_cpu_key(&key, KEY_FORMAT_3_6, le32_to_cpu(ih.ih_key.k_dir_id),
2007bd4c625cSLinus Torvalds le32_to_cpu(ih.ih_key.k_objectid), SD_OFFSET,
2008bd4c625cSLinus Torvalds TYPE_STAT_DATA, 3 /*key length */ );
20091da177e4SLinus Torvalds
20101da177e4SLinus Torvalds /* find proper place for inserting of stat data */
20111da177e4SLinus Torvalds retval = search_item(sb, &key, &path_to_key);
20121da177e4SLinus Torvalds if (retval == IO_ERROR) {
20131da177e4SLinus Torvalds err = -EIO;
20141da177e4SLinus Torvalds goto out_bad_inode;
20151da177e4SLinus Torvalds }
20161da177e4SLinus Torvalds if (retval == ITEM_FOUND) {
20171da177e4SLinus Torvalds pathrelse(&path_to_key);
20181da177e4SLinus Torvalds err = -EEXIST;
20191da177e4SLinus Torvalds goto out_bad_inode;
20201da177e4SLinus Torvalds }
20211da177e4SLinus Torvalds if (old_format_only(sb)) {
2022098297b2SJeff Mahoney /* i_uid or i_gid is too big to be stored in stat data v3.5 */
2023df814654SEric W. Biederman if (i_uid_read(inode) & ~0xffff || i_gid_read(inode) & ~0xffff) {
20241da177e4SLinus Torvalds pathrelse(&path_to_key);
20251da177e4SLinus Torvalds err = -EINVAL;
20261da177e4SLinus Torvalds goto out_bad_inode;
20271da177e4SLinus Torvalds }
20281da177e4SLinus Torvalds inode2sd_v1(&sd, inode, inode->i_size);
20291da177e4SLinus Torvalds } else {
20301da177e4SLinus Torvalds inode2sd(&sd, inode, inode->i_size);
20311da177e4SLinus Torvalds }
2032098297b2SJeff Mahoney /*
2033098297b2SJeff Mahoney * store in in-core inode the key of stat data and version all
2034098297b2SJeff Mahoney * object items will have (directory items will have old offset
2035098297b2SJeff Mahoney * format, other new objects will consist of new items)
2036098297b2SJeff Mahoney */
20371da177e4SLinus Torvalds if (old_format_only(sb) || S_ISDIR(mode) || S_ISLNK(mode))
20381da177e4SLinus Torvalds set_inode_item_key_version(inode, KEY_FORMAT_3_5);
20391da177e4SLinus Torvalds else
20401da177e4SLinus Torvalds set_inode_item_key_version(inode, KEY_FORMAT_3_6);
20411da177e4SLinus Torvalds if (old_format_only(sb))
20421da177e4SLinus Torvalds set_inode_sd_version(inode, STAT_DATA_V1);
20431da177e4SLinus Torvalds else
20441da177e4SLinus Torvalds set_inode_sd_version(inode, STAT_DATA_V2);
20451da177e4SLinus Torvalds
20461da177e4SLinus Torvalds /* insert the stat data into the tree */
20471da177e4SLinus Torvalds #ifdef DISPLACE_NEW_PACKING_LOCALITIES
20481da177e4SLinus Torvalds if (REISERFS_I(dir)->new_packing_locality)
20491da177e4SLinus Torvalds th->displace_new_blocks = 1;
20501da177e4SLinus Torvalds #endif
2051bd4c625cSLinus Torvalds retval =
2052bd4c625cSLinus Torvalds reiserfs_insert_item(th, &path_to_key, &key, &ih, inode,
2053bd4c625cSLinus Torvalds (char *)(&sd));
20541da177e4SLinus Torvalds if (retval) {
20551da177e4SLinus Torvalds err = retval;
20561da177e4SLinus Torvalds reiserfs_check_path(&path_to_key);
20571da177e4SLinus Torvalds goto out_bad_inode;
20581da177e4SLinus Torvalds }
20591da177e4SLinus Torvalds #ifdef DISPLACE_NEW_PACKING_LOCALITIES
20601da177e4SLinus Torvalds if (!th->displace_new_blocks)
20611da177e4SLinus Torvalds REISERFS_I(dir)->new_packing_locality = 0;
20621da177e4SLinus Torvalds #endif
20631da177e4SLinus Torvalds if (S_ISDIR(mode)) {
20641da177e4SLinus Torvalds /* insert item with "." and ".." */
2065bd4c625cSLinus Torvalds retval =
2066bd4c625cSLinus Torvalds reiserfs_new_directory(th, inode, &ih, &path_to_key, dir);
20671da177e4SLinus Torvalds }
20681da177e4SLinus Torvalds
20691da177e4SLinus Torvalds if (S_ISLNK(mode)) {
20701da177e4SLinus Torvalds /* insert body of symlink */
20711da177e4SLinus Torvalds if (!old_format_only(sb))
20721da177e4SLinus Torvalds i_size = ROUND_UP(i_size);
2073bd4c625cSLinus Torvalds retval =
2074bd4c625cSLinus Torvalds reiserfs_new_symlink(th, inode, &ih, &path_to_key, symname,
2075bd4c625cSLinus Torvalds i_size);
20761da177e4SLinus Torvalds }
20771da177e4SLinus Torvalds if (retval) {
20781da177e4SLinus Torvalds err = retval;
20791da177e4SLinus Torvalds reiserfs_check_path(&path_to_key);
208058d85426SJeff Mahoney journal_end(th);
20811da177e4SLinus Torvalds goto out_inserted_sd;
20821da177e4SLinus Torvalds }
20831da177e4SLinus Torvalds
208460e4cf67SJeff Mahoney /*
208560e4cf67SJeff Mahoney * Mark it private if we're creating the privroot
208660e4cf67SJeff Mahoney * or something under it.
208760e4cf67SJeff Mahoney */
2088d9f892b9SChristian Brauner if (IS_PRIVATE(dir) || dentry == REISERFS_SB(sb)->priv_root)
2089d9f892b9SChristian Brauner reiserfs_init_priv_inode(inode);
209060e4cf67SJeff Mahoney
20911da177e4SLinus Torvalds if (reiserfs_posixacl(inode->i_sb)) {
20924c05141dSJeff Mahoney reiserfs_write_unlock(inode->i_sb);
20930ab2621eSJeff Mahoney retval = reiserfs_inherit_default_acl(th, dir, dentry, inode);
20944c05141dSJeff Mahoney reiserfs_write_lock(inode->i_sb);
20951da177e4SLinus Torvalds if (retval) {
20961da177e4SLinus Torvalds err = retval;
20971da177e4SLinus Torvalds reiserfs_check_path(&path_to_key);
209858d85426SJeff Mahoney journal_end(th);
20991da177e4SLinus Torvalds goto out_inserted_sd;
21001da177e4SLinus Torvalds }
21011751e8a6SLinus Torvalds } else if (inode->i_sb->s_flags & SB_POSIXACL) {
210245b03d5eSJeff Mahoney reiserfs_warning(inode->i_sb, "jdm-13090",
210345b03d5eSJeff Mahoney "ACLs aren't enabled in the fs, "
21041da177e4SLinus Torvalds "but vfs thinks they are!");
210560e4cf67SJeff Mahoney }
21061da177e4SLinus Torvalds
210757fe60dfSJeff Mahoney if (security->name) {
21084c05141dSJeff Mahoney reiserfs_write_unlock(inode->i_sb);
210957fe60dfSJeff Mahoney retval = reiserfs_security_write(th, inode, security);
21104c05141dSJeff Mahoney reiserfs_write_lock(inode->i_sb);
211157fe60dfSJeff Mahoney if (retval) {
211257fe60dfSJeff Mahoney err = retval;
211357fe60dfSJeff Mahoney reiserfs_check_path(&path_to_key);
211458d85426SJeff Mahoney retval = journal_end(th);
211557fe60dfSJeff Mahoney if (retval)
211657fe60dfSJeff Mahoney err = retval;
211757fe60dfSJeff Mahoney goto out_inserted_sd;
211857fe60dfSJeff Mahoney }
21191da177e4SLinus Torvalds }
21201da177e4SLinus Torvalds
21211da177e4SLinus Torvalds reiserfs_update_sd(th, inode);
21221da177e4SLinus Torvalds reiserfs_check_path(&path_to_key);
21231da177e4SLinus Torvalds
21241da177e4SLinus Torvalds return 0;
21251da177e4SLinus Torvalds
21261da177e4SLinus Torvalds out_bad_inode:
21271da177e4SLinus Torvalds /* Invalidate the object, nothing was inserted yet */
21281da177e4SLinus Torvalds INODE_PKEY(inode)->k_objectid = 0;
21291da177e4SLinus Torvalds
21301da177e4SLinus Torvalds /* Quota change must be inside a transaction for journaling */
2131d2d0395fSJeff Mahoney depth = reiserfs_write_unlock_nested(inode->i_sb);
213263936ddaSChristoph Hellwig dquot_free_inode(inode);
2133d2d0395fSJeff Mahoney reiserfs_write_lock_nested(inode->i_sb, depth);
21341da177e4SLinus Torvalds
21351da177e4SLinus Torvalds out_end_trans:
213658d85426SJeff Mahoney journal_end(th);
2137098297b2SJeff Mahoney /*
2138098297b2SJeff Mahoney * Drop can be outside and it needs more credits so it's better
2139098297b2SJeff Mahoney * to have it outside
2140098297b2SJeff Mahoney */
2141d2d0395fSJeff Mahoney depth = reiserfs_write_unlock_nested(inode->i_sb);
21429f754758SChristoph Hellwig dquot_drop(inode);
2143d2d0395fSJeff Mahoney reiserfs_write_lock_nested(inode->i_sb, depth);
21441da177e4SLinus Torvalds inode->i_flags |= S_NOQUOTA;
21451da177e4SLinus Torvalds make_bad_inode(inode);
21461da177e4SLinus Torvalds
21471da177e4SLinus Torvalds out_inserted_sd:
21486d6b77f1SMiklos Szeredi clear_nlink(inode);
21491da177e4SLinus Torvalds th->t_trans_id = 0; /* so the caller can't use this handle later */
21508859bf2bSEric Biggers if (inode->i_state & I_NEW)
21518859bf2bSEric Biggers unlock_new_inode(inode);
21521da177e4SLinus Torvalds iput(inode);
21531da177e4SLinus Torvalds return err;
21541da177e4SLinus Torvalds }
21551da177e4SLinus Torvalds
21561da177e4SLinus Torvalds /*
2157098297b2SJeff Mahoney * finds the tail page in the page cache,
2158098297b2SJeff Mahoney * reads the last block in.
2159098297b2SJeff Mahoney *
2160098297b2SJeff Mahoney * On success, page_result is set to a locked, pinned page, and bh_result
2161098297b2SJeff Mahoney * is set to an up to date buffer for the last block in the file. returns 0.
2162098297b2SJeff Mahoney *
2163098297b2SJeff Mahoney * tail conversion is not done, so bh_result might not be valid for writing
2164098297b2SJeff Mahoney * check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
2165098297b2SJeff Mahoney * trying to write the block.
2166098297b2SJeff Mahoney *
2167098297b2SJeff Mahoney * on failure, nonzero is returned, page_result and bh_result are untouched.
21681da177e4SLinus Torvalds */
grab_tail_page(struct inode * inode,struct page ** page_result,struct buffer_head ** bh_result)2169995c762eSJeff Mahoney static int grab_tail_page(struct inode *inode,
21701da177e4SLinus Torvalds struct page **page_result,
2171bd4c625cSLinus Torvalds struct buffer_head **bh_result)
2172bd4c625cSLinus Torvalds {
21731da177e4SLinus Torvalds
2174098297b2SJeff Mahoney /*
2175098297b2SJeff Mahoney * we want the page with the last byte in the file,
2176098297b2SJeff Mahoney * not the page that will hold the next byte for appending
21771da177e4SLinus Torvalds */
217809cbfeafSKirill A. Shutemov unsigned long index = (inode->i_size - 1) >> PAGE_SHIFT;
21791da177e4SLinus Torvalds unsigned long pos = 0;
21801da177e4SLinus Torvalds unsigned long start = 0;
2181995c762eSJeff Mahoney unsigned long blocksize = inode->i_sb->s_blocksize;
218209cbfeafSKirill A. Shutemov unsigned long offset = (inode->i_size) & (PAGE_SIZE - 1);
21831da177e4SLinus Torvalds struct buffer_head *bh;
21841da177e4SLinus Torvalds struct buffer_head *head;
21851da177e4SLinus Torvalds struct page *page;
21861da177e4SLinus Torvalds int error;
21871da177e4SLinus Torvalds
2188098297b2SJeff Mahoney /*
2189098297b2SJeff Mahoney * we know that we are only called with inode->i_size > 0.
2190098297b2SJeff Mahoney * we also know that a file tail can never be as big as a block
2191098297b2SJeff Mahoney * If i_size % blocksize == 0, our file is currently block aligned
2192098297b2SJeff Mahoney * and it won't need converting or zeroing after a truncate.
21931da177e4SLinus Torvalds */
21941da177e4SLinus Torvalds if ((offset & (blocksize - 1)) == 0) {
21951da177e4SLinus Torvalds return -ENOENT;
21961da177e4SLinus Torvalds }
2197995c762eSJeff Mahoney page = grab_cache_page(inode->i_mapping, index);
21981da177e4SLinus Torvalds error = -ENOMEM;
21991da177e4SLinus Torvalds if (!page) {
22001da177e4SLinus Torvalds goto out;
22011da177e4SLinus Torvalds }
22021da177e4SLinus Torvalds /* start within the page of the last block in the file */
22031da177e4SLinus Torvalds start = (offset / blocksize) * blocksize;
22041da177e4SLinus Torvalds
2205ebdec241SChristoph Hellwig error = __block_write_begin(page, start, offset - start,
22061da177e4SLinus Torvalds reiserfs_get_block_create_0);
22071da177e4SLinus Torvalds if (error)
22081da177e4SLinus Torvalds goto unlock;
22091da177e4SLinus Torvalds
22101da177e4SLinus Torvalds head = page_buffers(page);
22111da177e4SLinus Torvalds bh = head;
22121da177e4SLinus Torvalds do {
22131da177e4SLinus Torvalds if (pos >= start) {
22141da177e4SLinus Torvalds break;
22151da177e4SLinus Torvalds }
22161da177e4SLinus Torvalds bh = bh->b_this_page;
22171da177e4SLinus Torvalds pos += blocksize;
22181da177e4SLinus Torvalds } while (bh != head);
22191da177e4SLinus Torvalds
22201da177e4SLinus Torvalds if (!buffer_uptodate(bh)) {
2221098297b2SJeff Mahoney /*
2222098297b2SJeff Mahoney * note, this should never happen, prepare_write should be
2223098297b2SJeff Mahoney * taking care of this for us. If the buffer isn't up to
2224098297b2SJeff Mahoney * date, I've screwed up the code to find the buffer, or the
2225098297b2SJeff Mahoney * code to call prepare_write
22261da177e4SLinus Torvalds */
2227995c762eSJeff Mahoney reiserfs_error(inode->i_sb, "clm-6000",
22280030b645SJeff Mahoney "error reading block %lu", bh->b_blocknr);
22291da177e4SLinus Torvalds error = -EIO;
22301da177e4SLinus Torvalds goto unlock;
22311da177e4SLinus Torvalds }
22321da177e4SLinus Torvalds *bh_result = bh;
22331da177e4SLinus Torvalds *page_result = page;
22341da177e4SLinus Torvalds
22351da177e4SLinus Torvalds out:
22361da177e4SLinus Torvalds return error;
22371da177e4SLinus Torvalds
22381da177e4SLinus Torvalds unlock:
22391da177e4SLinus Torvalds unlock_page(page);
224009cbfeafSKirill A. Shutemov put_page(page);
22411da177e4SLinus Torvalds return error;
22421da177e4SLinus Torvalds }
22431da177e4SLinus Torvalds
22441da177e4SLinus Torvalds /*
2245098297b2SJeff Mahoney * vfs version of truncate file. Must NOT be called with
2246098297b2SJeff Mahoney * a transaction already started.
2247098297b2SJeff Mahoney *
2248098297b2SJeff Mahoney * some code taken from block_truncate_page
22491da177e4SLinus Torvalds */
reiserfs_truncate_file(struct inode * inode,int update_timestamps)2250995c762eSJeff Mahoney int reiserfs_truncate_file(struct inode *inode, int update_timestamps)
2251bd4c625cSLinus Torvalds {
22521da177e4SLinus Torvalds struct reiserfs_transaction_handle th;
22531da177e4SLinus Torvalds /* we want the offset for the first byte after the end of the file */
225409cbfeafSKirill A. Shutemov unsigned long offset = inode->i_size & (PAGE_SIZE - 1);
2255995c762eSJeff Mahoney unsigned blocksize = inode->i_sb->s_blocksize;
22561da177e4SLinus Torvalds unsigned length;
22571da177e4SLinus Torvalds struct page *page = NULL;
22581da177e4SLinus Torvalds int error;
22591da177e4SLinus Torvalds struct buffer_head *bh = NULL;
226024996049SJeff Mahoney int err2;
22611da177e4SLinus Torvalds
2262278f6679SJeff Mahoney reiserfs_write_lock(inode->i_sb);
22631da177e4SLinus Torvalds
2264995c762eSJeff Mahoney if (inode->i_size > 0) {
2265995c762eSJeff Mahoney error = grab_tail_page(inode, &page, &bh);
2266995c762eSJeff Mahoney if (error) {
2267098297b2SJeff Mahoney /*
2268098297b2SJeff Mahoney * -ENOENT means we truncated past the end of the
2269098297b2SJeff Mahoney * file, and get_block_create_0 could not find a
2270098297b2SJeff Mahoney * block to read in, which is ok.
2271098297b2SJeff Mahoney */
22721da177e4SLinus Torvalds if (error != -ENOENT)
2273995c762eSJeff Mahoney reiserfs_error(inode->i_sb, "clm-6001",
227445b03d5eSJeff Mahoney "grab_tail_page failed %d",
22751da177e4SLinus Torvalds error);
22761da177e4SLinus Torvalds page = NULL;
22771da177e4SLinus Torvalds bh = NULL;
22781da177e4SLinus Torvalds }
22791da177e4SLinus Torvalds }
22801da177e4SLinus Torvalds
2281098297b2SJeff Mahoney /*
2282098297b2SJeff Mahoney * so, if page != NULL, we have a buffer head for the offset at
2283098297b2SJeff Mahoney * the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
2284098297b2SJeff Mahoney * then we have an unformatted node. Otherwise, we have a direct item,
2285098297b2SJeff Mahoney * and no zeroing is required on disk. We zero after the truncate,
2286098297b2SJeff Mahoney * because the truncate might pack the item anyway
2287098297b2SJeff Mahoney * (it will unmap bh if it packs).
2288098297b2SJeff Mahoney *
2289098297b2SJeff Mahoney * it is enough to reserve space in transaction for 2 balancings:
2290098297b2SJeff Mahoney * one for "save" link adding and another for the first
2291098297b2SJeff Mahoney * cut_from_item. 1 is for update_sd
22921da177e4SLinus Torvalds */
2293995c762eSJeff Mahoney error = journal_begin(&th, inode->i_sb,
22941da177e4SLinus Torvalds JOURNAL_PER_BALANCE_CNT * 2 + 1);
22951da177e4SLinus Torvalds if (error)
22961da177e4SLinus Torvalds goto out;
2297995c762eSJeff Mahoney reiserfs_update_inode_transaction(inode);
22981da177e4SLinus Torvalds if (update_timestamps)
2299098297b2SJeff Mahoney /*
2300098297b2SJeff Mahoney * we are doing real truncate: if the system crashes
2301098297b2SJeff Mahoney * before the last transaction of truncating gets committed
2302098297b2SJeff Mahoney * - on reboot the file either appears truncated properly
2303098297b2SJeff Mahoney * or not truncated at all
2304098297b2SJeff Mahoney */
2305995c762eSJeff Mahoney add_save_link(&th, inode, 1);
2306995c762eSJeff Mahoney err2 = reiserfs_do_truncate(&th, inode, page, update_timestamps);
230758d85426SJeff Mahoney error = journal_end(&th);
23081da177e4SLinus Torvalds if (error)
23091da177e4SLinus Torvalds goto out;
23101da177e4SLinus Torvalds
231124996049SJeff Mahoney /* check reiserfs_do_truncate after ending the transaction */
231224996049SJeff Mahoney if (err2) {
231324996049SJeff Mahoney error = err2;
231424996049SJeff Mahoney goto out;
231524996049SJeff Mahoney }
231624996049SJeff Mahoney
23171da177e4SLinus Torvalds if (update_timestamps) {
2318995c762eSJeff Mahoney error = remove_save_link(inode, 1 /* truncate */);
23191da177e4SLinus Torvalds if (error)
23201da177e4SLinus Torvalds goto out;
23211da177e4SLinus Torvalds }
23221da177e4SLinus Torvalds
23231da177e4SLinus Torvalds if (page) {
23241da177e4SLinus Torvalds length = offset & (blocksize - 1);
23251da177e4SLinus Torvalds /* if we are not on a block boundary */
23261da177e4SLinus Torvalds if (length) {
23271da177e4SLinus Torvalds length = blocksize - length;
2328eebd2aa3SChristoph Lameter zero_user(page, offset, length);
23291da177e4SLinus Torvalds if (buffer_mapped(bh) && bh->b_blocknr != 0) {
23301da177e4SLinus Torvalds mark_buffer_dirty(bh);
23311da177e4SLinus Torvalds }
23321da177e4SLinus Torvalds }
23331da177e4SLinus Torvalds unlock_page(page);
233409cbfeafSKirill A. Shutemov put_page(page);
23351da177e4SLinus Torvalds }
23361da177e4SLinus Torvalds
2337278f6679SJeff Mahoney reiserfs_write_unlock(inode->i_sb);
233822c963adSFrederic Weisbecker
23391da177e4SLinus Torvalds return 0;
23401da177e4SLinus Torvalds out:
23411da177e4SLinus Torvalds if (page) {
23421da177e4SLinus Torvalds unlock_page(page);
234309cbfeafSKirill A. Shutemov put_page(page);
23441da177e4SLinus Torvalds }
234522c963adSFrederic Weisbecker
2346278f6679SJeff Mahoney reiserfs_write_unlock(inode->i_sb);
234722c963adSFrederic Weisbecker
23481da177e4SLinus Torvalds return error;
23491da177e4SLinus Torvalds }
23501da177e4SLinus Torvalds
map_block_for_writepage(struct inode * inode,struct buffer_head * bh_result,unsigned long block)23511da177e4SLinus Torvalds static int map_block_for_writepage(struct inode *inode,
23521da177e4SLinus Torvalds struct buffer_head *bh_result,
2353bd4c625cSLinus Torvalds unsigned long block)
2354bd4c625cSLinus Torvalds {
23551da177e4SLinus Torvalds struct reiserfs_transaction_handle th;
23561da177e4SLinus Torvalds int fs_gen;
23571da177e4SLinus Torvalds struct item_head tmp_ih;
23581da177e4SLinus Torvalds struct item_head *ih;
23591da177e4SLinus Torvalds struct buffer_head *bh;
23603e8962beSAl Viro __le32 *item;
23611da177e4SLinus Torvalds struct cpu_key key;
23621da177e4SLinus Torvalds INITIALIZE_PATH(path);
23631da177e4SLinus Torvalds int pos_in_item;
23641da177e4SLinus Torvalds int jbegin_count = JOURNAL_PER_BALANCE_CNT;
23657729ac5eSOleg Drokin loff_t byte_offset = ((loff_t)block << inode->i_sb->s_blocksize_bits)+1;
23661da177e4SLinus Torvalds int retval;
23671da177e4SLinus Torvalds int use_get_block = 0;
23681da177e4SLinus Torvalds int bytes_copied = 0;
23691da177e4SLinus Torvalds int copy_size;
23701da177e4SLinus Torvalds int trans_running = 0;
23711da177e4SLinus Torvalds
2372098297b2SJeff Mahoney /*
2373098297b2SJeff Mahoney * catch places below that try to log something without
2374098297b2SJeff Mahoney * starting a trans
2375098297b2SJeff Mahoney */
23761da177e4SLinus Torvalds th.t_trans_id = 0;
23771da177e4SLinus Torvalds
23781da177e4SLinus Torvalds if (!buffer_uptodate(bh_result)) {
23791da177e4SLinus Torvalds return -EIO;
23801da177e4SLinus Torvalds }
23811da177e4SLinus Torvalds
23821da177e4SLinus Torvalds kmap(bh_result->b_page);
23831da177e4SLinus Torvalds start_over:
23841da177e4SLinus Torvalds reiserfs_write_lock(inode->i_sb);
23851da177e4SLinus Torvalds make_cpu_key(&key, inode, byte_offset, TYPE_ANY, 3);
23861da177e4SLinus Torvalds
23871da177e4SLinus Torvalds research:
23881da177e4SLinus Torvalds retval = search_for_position_by_key(inode->i_sb, &key, &path);
23891da177e4SLinus Torvalds if (retval != POSITION_FOUND) {
23901da177e4SLinus Torvalds use_get_block = 1;
23911da177e4SLinus Torvalds goto out;
23921da177e4SLinus Torvalds }
23931da177e4SLinus Torvalds
23941da177e4SLinus Torvalds bh = get_last_bh(&path);
23954cf5f7adSJeff Mahoney ih = tp_item_head(&path);
23964cf5f7adSJeff Mahoney item = tp_item_body(&path);
23971da177e4SLinus Torvalds pos_in_item = path.pos_in_item;
23981da177e4SLinus Torvalds
23991da177e4SLinus Torvalds /* we've found an unformatted node */
24001da177e4SLinus Torvalds if (indirect_item_found(retval, ih)) {
24011da177e4SLinus Torvalds if (bytes_copied > 0) {
240245b03d5eSJeff Mahoney reiserfs_warning(inode->i_sb, "clm-6002",
240345b03d5eSJeff Mahoney "bytes_copied %d", bytes_copied);
24041da177e4SLinus Torvalds }
24051da177e4SLinus Torvalds if (!get_block_num(item, pos_in_item)) {
24061da177e4SLinus Torvalds /* crap, we are writing to a hole */
24071da177e4SLinus Torvalds use_get_block = 1;
24081da177e4SLinus Torvalds goto out;
24091da177e4SLinus Torvalds }
2410bd4c625cSLinus Torvalds set_block_dev_mapped(bh_result,
2411bd4c625cSLinus Torvalds get_block_num(item, pos_in_item), inode);
24121da177e4SLinus Torvalds } else if (is_direct_le_ih(ih)) {
24131da177e4SLinus Torvalds char *p;
24141da177e4SLinus Torvalds p = page_address(bh_result->b_page);
241509cbfeafSKirill A. Shutemov p += (byte_offset - 1) & (PAGE_SIZE - 1);
24161da177e4SLinus Torvalds copy_size = ih_item_len(ih) - pos_in_item;
24171da177e4SLinus Torvalds
24181da177e4SLinus Torvalds fs_gen = get_generation(inode->i_sb);
24191da177e4SLinus Torvalds copy_item_head(&tmp_ih, ih);
24201da177e4SLinus Torvalds
24211da177e4SLinus Torvalds if (!trans_running) {
24221da177e4SLinus Torvalds /* vs-3050 is gone, no need to drop the path */
24231da177e4SLinus Torvalds retval = journal_begin(&th, inode->i_sb, jbegin_count);
24241da177e4SLinus Torvalds if (retval)
24251da177e4SLinus Torvalds goto out;
24261da177e4SLinus Torvalds reiserfs_update_inode_transaction(inode);
24271da177e4SLinus Torvalds trans_running = 1;
2428bd4c625cSLinus Torvalds if (fs_changed(fs_gen, inode->i_sb)
2429bd4c625cSLinus Torvalds && item_moved(&tmp_ih, &path)) {
2430bd4c625cSLinus Torvalds reiserfs_restore_prepared_buffer(inode->i_sb,
2431bd4c625cSLinus Torvalds bh);
24321da177e4SLinus Torvalds goto research;
24331da177e4SLinus Torvalds }
24341da177e4SLinus Torvalds }
24351da177e4SLinus Torvalds
24361da177e4SLinus Torvalds reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
24371da177e4SLinus Torvalds
2438bd4c625cSLinus Torvalds if (fs_changed(fs_gen, inode->i_sb)
2439bd4c625cSLinus Torvalds && item_moved(&tmp_ih, &path)) {
24401da177e4SLinus Torvalds reiserfs_restore_prepared_buffer(inode->i_sb, bh);
24411da177e4SLinus Torvalds goto research;
24421da177e4SLinus Torvalds }
24431da177e4SLinus Torvalds
24444cf5f7adSJeff Mahoney memcpy(ih_item_body(bh, ih) + pos_in_item, p + bytes_copied,
2445bd4c625cSLinus Torvalds copy_size);
24461da177e4SLinus Torvalds
244709f1b80bSJeff Mahoney journal_mark_dirty(&th, bh);
24481da177e4SLinus Torvalds bytes_copied += copy_size;
24491da177e4SLinus Torvalds set_block_dev_mapped(bh_result, 0, inode);
24501da177e4SLinus Torvalds
24511da177e4SLinus Torvalds /* are there still bytes left? */
24521da177e4SLinus Torvalds if (bytes_copied < bh_result->b_size &&
24531da177e4SLinus Torvalds (byte_offset + bytes_copied) < inode->i_size) {
2454bd4c625cSLinus Torvalds set_cpu_key_k_offset(&key,
2455bd4c625cSLinus Torvalds cpu_key_k_offset(&key) +
2456bd4c625cSLinus Torvalds copy_size);
24571da177e4SLinus Torvalds goto research;
24581da177e4SLinus Torvalds }
24591da177e4SLinus Torvalds } else {
246045b03d5eSJeff Mahoney reiserfs_warning(inode->i_sb, "clm-6003",
246145b03d5eSJeff Mahoney "bad item inode %lu", inode->i_ino);
24621da177e4SLinus Torvalds retval = -EIO;
24631da177e4SLinus Torvalds goto out;
24641da177e4SLinus Torvalds }
24651da177e4SLinus Torvalds retval = 0;
24661da177e4SLinus Torvalds
24671da177e4SLinus Torvalds out:
24681da177e4SLinus Torvalds pathrelse(&path);
24691da177e4SLinus Torvalds if (trans_running) {
247058d85426SJeff Mahoney int err = journal_end(&th);
24711da177e4SLinus Torvalds if (err)
24721da177e4SLinus Torvalds retval = err;
24731da177e4SLinus Torvalds trans_running = 0;
24741da177e4SLinus Torvalds }
24751da177e4SLinus Torvalds reiserfs_write_unlock(inode->i_sb);
24761da177e4SLinus Torvalds
24771da177e4SLinus Torvalds /* this is where we fill in holes in the file. */
24781da177e4SLinus Torvalds if (use_get_block) {
24791da177e4SLinus Torvalds retval = reiserfs_get_block(inode, block, bh_result,
24801b1dcc1bSJes Sorensen GET_BLOCK_CREATE | GET_BLOCK_NO_IMUX
2481bd4c625cSLinus Torvalds | GET_BLOCK_NO_DANGLE);
24821da177e4SLinus Torvalds if (!retval) {
2483bd4c625cSLinus Torvalds if (!buffer_mapped(bh_result)
2484bd4c625cSLinus Torvalds || bh_result->b_blocknr == 0) {
24851da177e4SLinus Torvalds /* get_block failed to find a mapped unformatted node. */
24861da177e4SLinus Torvalds use_get_block = 0;
24871da177e4SLinus Torvalds goto start_over;
24881da177e4SLinus Torvalds }
24891da177e4SLinus Torvalds }
24901da177e4SLinus Torvalds }
24911da177e4SLinus Torvalds kunmap(bh_result->b_page);
24921da177e4SLinus Torvalds
24931da177e4SLinus Torvalds if (!retval && buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
2494098297b2SJeff Mahoney /*
2495098297b2SJeff Mahoney * we've copied data from the page into the direct item, so the
24961da177e4SLinus Torvalds * buffer in the page is now clean, mark it to reflect that.
24971da177e4SLinus Torvalds */
24981da177e4SLinus Torvalds lock_buffer(bh_result);
24991da177e4SLinus Torvalds clear_buffer_dirty(bh_result);
25001da177e4SLinus Torvalds unlock_buffer(bh_result);
25011da177e4SLinus Torvalds }
25021da177e4SLinus Torvalds return retval;
25031da177e4SLinus Torvalds }
25041da177e4SLinus Torvalds
25051da177e4SLinus Torvalds /*
25061da177e4SLinus Torvalds * mason@suse.com: updated in 2.5.54 to follow the same general io
250753418a18SMatthew Wilcox (Oracle) * start/recovery path as __block_write_full_folio, along with special
25081da177e4SLinus Torvalds * code to handle reiserfs tails.
25091da177e4SLinus Torvalds */
reiserfs_write_full_page(struct page * page,struct writeback_control * wbc)2510bd4c625cSLinus Torvalds static int reiserfs_write_full_page(struct page *page,
2511bd4c625cSLinus Torvalds struct writeback_control *wbc)
2512bd4c625cSLinus Torvalds {
25131da177e4SLinus Torvalds struct inode *inode = page->mapping->host;
251409cbfeafSKirill A. Shutemov unsigned long end_index = inode->i_size >> PAGE_SHIFT;
25151da177e4SLinus Torvalds int error = 0;
25161da177e4SLinus Torvalds unsigned long block;
2517b4c76fa7SChris Mason sector_t last_block;
25181da177e4SLinus Torvalds struct buffer_head *head, *bh;
25191da177e4SLinus Torvalds int partial = 0;
25201da177e4SLinus Torvalds int nr = 0;
25211da177e4SLinus Torvalds int checked = PageChecked(page);
25221da177e4SLinus Torvalds struct reiserfs_transaction_handle th;
25231da177e4SLinus Torvalds struct super_block *s = inode->i_sb;
252409cbfeafSKirill A. Shutemov int bh_per_page = PAGE_SIZE / s->s_blocksize;
25251da177e4SLinus Torvalds th.t_trans_id = 0;
25261da177e4SLinus Torvalds
2527e0e851cfSChris Mason /* no logging allowed when nonblocking or from PF_MEMALLOC */
2528e0e851cfSChris Mason if (checked && (current->flags & PF_MEMALLOC)) {
2529e0e851cfSChris Mason redirty_page_for_writepage(wbc, page);
2530e0e851cfSChris Mason unlock_page(page);
2531e0e851cfSChris Mason return 0;
2532e0e851cfSChris Mason }
2533e0e851cfSChris Mason
2534098297b2SJeff Mahoney /*
2535098297b2SJeff Mahoney * The page dirty bit is cleared before writepage is called, which
25361da177e4SLinus Torvalds * means we have to tell create_empty_buffers to make dirty buffers
25371da177e4SLinus Torvalds * The page really should be up to date at this point, so tossing
25381da177e4SLinus Torvalds * in the BH_Uptodate is just a sanity check.
25391da177e4SLinus Torvalds */
25401da177e4SLinus Torvalds if (!page_has_buffers(page)) {
25411da177e4SLinus Torvalds create_empty_buffers(page, s->s_blocksize,
25421da177e4SLinus Torvalds (1 << BH_Dirty) | (1 << BH_Uptodate));
25431da177e4SLinus Torvalds }
25441da177e4SLinus Torvalds head = page_buffers(page);
25451da177e4SLinus Torvalds
2546098297b2SJeff Mahoney /*
2547098297b2SJeff Mahoney * last page in the file, zero out any contents past the
2548098297b2SJeff Mahoney * last byte in the file
25491da177e4SLinus Torvalds */
25501da177e4SLinus Torvalds if (page->index >= end_index) {
25511da177e4SLinus Torvalds unsigned last_offset;
25521da177e4SLinus Torvalds
255309cbfeafSKirill A. Shutemov last_offset = inode->i_size & (PAGE_SIZE - 1);
25541da177e4SLinus Torvalds /* no file contents in this page */
25551da177e4SLinus Torvalds if (page->index >= end_index + 1 || !last_offset) {
25561da177e4SLinus Torvalds unlock_page(page);
25571da177e4SLinus Torvalds return 0;
25581da177e4SLinus Torvalds }
255909cbfeafSKirill A. Shutemov zero_user_segment(page, last_offset, PAGE_SIZE);
25601da177e4SLinus Torvalds }
25611da177e4SLinus Torvalds bh = head;
256209cbfeafSKirill A. Shutemov block = page->index << (PAGE_SHIFT - s->s_blocksize_bits);
2563b4c76fa7SChris Mason last_block = (i_size_read(inode) - 1) >> inode->i_blkbits;
25641da177e4SLinus Torvalds /* first map all the buffers, logging any direct items we find */
25651da177e4SLinus Torvalds do {
2566b4c76fa7SChris Mason if (block > last_block) {
2567b4c76fa7SChris Mason /*
2568b4c76fa7SChris Mason * This can happen when the block size is less than
2569b4c76fa7SChris Mason * the page size. The corresponding bytes in the page
2570b4c76fa7SChris Mason * were zero filled above
2571b4c76fa7SChris Mason */
2572b4c76fa7SChris Mason clear_buffer_dirty(bh);
2573b4c76fa7SChris Mason set_buffer_uptodate(bh);
2574b4c76fa7SChris Mason } else if ((checked || buffer_dirty(bh)) &&
257521e4e15aSYueHaibing (!buffer_mapped(bh) || bh->b_blocknr == 0)) {
2576098297b2SJeff Mahoney /*
2577098297b2SJeff Mahoney * not mapped yet, or it points to a direct item, search
25781da177e4SLinus Torvalds * the btree for the mapping info, and log any direct
25791da177e4SLinus Torvalds * items found
25801da177e4SLinus Torvalds */
25811da177e4SLinus Torvalds if ((error = map_block_for_writepage(inode, bh, block))) {
25821da177e4SLinus Torvalds goto fail;
25831da177e4SLinus Torvalds }
25841da177e4SLinus Torvalds }
25851da177e4SLinus Torvalds bh = bh->b_this_page;
25861da177e4SLinus Torvalds block++;
25871da177e4SLinus Torvalds } while (bh != head);
25881da177e4SLinus Torvalds
25891da177e4SLinus Torvalds /*
25901da177e4SLinus Torvalds * we start the transaction after map_block_for_writepage,
25911da177e4SLinus Torvalds * because it can create holes in the file (an unbounded operation).
25921da177e4SLinus Torvalds * starting it here, we can make a reliable estimate for how many
25931da177e4SLinus Torvalds * blocks we're going to log
25941da177e4SLinus Torvalds */
25951da177e4SLinus Torvalds if (checked) {
25961da177e4SLinus Torvalds ClearPageChecked(page);
25971da177e4SLinus Torvalds reiserfs_write_lock(s);
25981da177e4SLinus Torvalds error = journal_begin(&th, s, bh_per_page + 1);
25991da177e4SLinus Torvalds if (error) {
26001da177e4SLinus Torvalds reiserfs_write_unlock(s);
26011da177e4SLinus Torvalds goto fail;
26021da177e4SLinus Torvalds }
26031da177e4SLinus Torvalds reiserfs_update_inode_transaction(inode);
26041da177e4SLinus Torvalds }
26051da177e4SLinus Torvalds /* now go through and lock any dirty buffers on the page */
26061da177e4SLinus Torvalds do {
26071da177e4SLinus Torvalds get_bh(bh);
26081da177e4SLinus Torvalds if (!buffer_mapped(bh))
26091da177e4SLinus Torvalds continue;
26101da177e4SLinus Torvalds if (buffer_mapped(bh) && bh->b_blocknr == 0)
26111da177e4SLinus Torvalds continue;
26121da177e4SLinus Torvalds
26131da177e4SLinus Torvalds if (checked) {
26141da177e4SLinus Torvalds reiserfs_prepare_for_journal(s, bh, 1);
261509f1b80bSJeff Mahoney journal_mark_dirty(&th, bh);
26161da177e4SLinus Torvalds continue;
26171da177e4SLinus Torvalds }
2618098297b2SJeff Mahoney /*
2619098297b2SJeff Mahoney * from this point on, we know the buffer is mapped to a
26201da177e4SLinus Torvalds * real block and not a direct item
26211da177e4SLinus Torvalds */
26221b430beeSWu Fengguang if (wbc->sync_mode != WB_SYNC_NONE) {
26231da177e4SLinus Torvalds lock_buffer(bh);
26241da177e4SLinus Torvalds } else {
2625ca5de404SNick Piggin if (!trylock_buffer(bh)) {
26261da177e4SLinus Torvalds redirty_page_for_writepage(wbc, page);
26271da177e4SLinus Torvalds continue;
26281da177e4SLinus Torvalds }
26291da177e4SLinus Torvalds }
26301da177e4SLinus Torvalds if (test_clear_buffer_dirty(bh)) {
26311da177e4SLinus Torvalds mark_buffer_async_write(bh);
26321da177e4SLinus Torvalds } else {
26331da177e4SLinus Torvalds unlock_buffer(bh);
26341da177e4SLinus Torvalds }
26351da177e4SLinus Torvalds } while ((bh = bh->b_this_page) != head);
26361da177e4SLinus Torvalds
26371da177e4SLinus Torvalds if (checked) {
263858d85426SJeff Mahoney error = journal_end(&th);
26391da177e4SLinus Torvalds reiserfs_write_unlock(s);
26401da177e4SLinus Torvalds if (error)
26411da177e4SLinus Torvalds goto fail;
26421da177e4SLinus Torvalds }
26431da177e4SLinus Torvalds BUG_ON(PageWriteback(page));
26441da177e4SLinus Torvalds set_page_writeback(page);
26451da177e4SLinus Torvalds unlock_page(page);
26461da177e4SLinus Torvalds
26471da177e4SLinus Torvalds /*
26481da177e4SLinus Torvalds * since any buffer might be the only dirty buffer on the page,
26491da177e4SLinus Torvalds * the first submit_bh can bring the page out of writeback.
26501da177e4SLinus Torvalds * be careful with the buffers.
26511da177e4SLinus Torvalds */
26521da177e4SLinus Torvalds do {
26531da177e4SLinus Torvalds struct buffer_head *next = bh->b_this_page;
26541da177e4SLinus Torvalds if (buffer_async_write(bh)) {
26551420c4a5SBart Van Assche submit_bh(REQ_OP_WRITE, bh);
26561da177e4SLinus Torvalds nr++;
26571da177e4SLinus Torvalds }
26581da177e4SLinus Torvalds put_bh(bh);
26591da177e4SLinus Torvalds bh = next;
26601da177e4SLinus Torvalds } while (bh != head);
26611da177e4SLinus Torvalds
26621da177e4SLinus Torvalds error = 0;
26631da177e4SLinus Torvalds done:
26641da177e4SLinus Torvalds if (nr == 0) {
26651da177e4SLinus Torvalds /*
26661da177e4SLinus Torvalds * if this page only had a direct item, it is very possible for
26671da177e4SLinus Torvalds * no io to be required without there being an error. Or,
26681da177e4SLinus Torvalds * someone else could have locked them and sent them down the
26691da177e4SLinus Torvalds * pipe without locking the page
26701da177e4SLinus Torvalds */
26711da177e4SLinus Torvalds bh = head;
26721da177e4SLinus Torvalds do {
26731da177e4SLinus Torvalds if (!buffer_uptodate(bh)) {
26741da177e4SLinus Torvalds partial = 1;
26751da177e4SLinus Torvalds break;
26761da177e4SLinus Torvalds }
26771da177e4SLinus Torvalds bh = bh->b_this_page;
26781da177e4SLinus Torvalds } while (bh != head);
26791da177e4SLinus Torvalds if (!partial)
26801da177e4SLinus Torvalds SetPageUptodate(page);
26811da177e4SLinus Torvalds end_page_writeback(page);
26821da177e4SLinus Torvalds }
26831da177e4SLinus Torvalds return error;
26841da177e4SLinus Torvalds
26851da177e4SLinus Torvalds fail:
2686098297b2SJeff Mahoney /*
2687098297b2SJeff Mahoney * catches various errors, we need to make sure any valid dirty blocks
26881da177e4SLinus Torvalds * get to the media. The page is currently locked and not marked for
26891da177e4SLinus Torvalds * writeback
26901da177e4SLinus Torvalds */
26911da177e4SLinus Torvalds ClearPageUptodate(page);
26921da177e4SLinus Torvalds bh = head;
26931da177e4SLinus Torvalds do {
26941da177e4SLinus Torvalds get_bh(bh);
26951da177e4SLinus Torvalds if (buffer_mapped(bh) && buffer_dirty(bh) && bh->b_blocknr) {
26961da177e4SLinus Torvalds lock_buffer(bh);
26971da177e4SLinus Torvalds mark_buffer_async_write(bh);
26981da177e4SLinus Torvalds } else {
26991da177e4SLinus Torvalds /*
2700098297b2SJeff Mahoney * clear any dirty bits that might have come from
2701098297b2SJeff Mahoney * getting attached to a dirty page
27021da177e4SLinus Torvalds */
27031da177e4SLinus Torvalds clear_buffer_dirty(bh);
27041da177e4SLinus Torvalds }
27051da177e4SLinus Torvalds bh = bh->b_this_page;
27061da177e4SLinus Torvalds } while (bh != head);
27071da177e4SLinus Torvalds SetPageError(page);
27081da177e4SLinus Torvalds BUG_ON(PageWriteback(page));
27091da177e4SLinus Torvalds set_page_writeback(page);
27101da177e4SLinus Torvalds unlock_page(page);
27111da177e4SLinus Torvalds do {
27121da177e4SLinus Torvalds struct buffer_head *next = bh->b_this_page;
27131da177e4SLinus Torvalds if (buffer_async_write(bh)) {
27141da177e4SLinus Torvalds clear_buffer_dirty(bh);
27151420c4a5SBart Van Assche submit_bh(REQ_OP_WRITE, bh);
27161da177e4SLinus Torvalds nr++;
27171da177e4SLinus Torvalds }
27181da177e4SLinus Torvalds put_bh(bh);
27191da177e4SLinus Torvalds bh = next;
27201da177e4SLinus Torvalds } while (bh != head);
27211da177e4SLinus Torvalds goto done;
27221da177e4SLinus Torvalds }
27231da177e4SLinus Torvalds
reiserfs_read_folio(struct file * f,struct folio * folio)27242c69e205SMatthew Wilcox (Oracle) static int reiserfs_read_folio(struct file *f, struct folio *folio)
27251da177e4SLinus Torvalds {
27262c69e205SMatthew Wilcox (Oracle) return block_read_full_folio(folio, reiserfs_get_block);
27271da177e4SLinus Torvalds }
27281da177e4SLinus Torvalds
reiserfs_writepage(struct page * page,struct writeback_control * wbc)27291da177e4SLinus Torvalds static int reiserfs_writepage(struct page *page, struct writeback_control *wbc)
27301da177e4SLinus Torvalds {
27311da177e4SLinus Torvalds struct inode *inode = page->mapping->host;
27321da177e4SLinus Torvalds reiserfs_wait_on_write_block(inode->i_sb);
27331da177e4SLinus Torvalds return reiserfs_write_full_page(page, wbc);
27341da177e4SLinus Torvalds }
27351da177e4SLinus Torvalds
reiserfs_truncate_failed_write(struct inode * inode)2736ec8e2f74SJan Kara static void reiserfs_truncate_failed_write(struct inode *inode)
2737ec8e2f74SJan Kara {
2738ec8e2f74SJan Kara truncate_inode_pages(inode->i_mapping, inode->i_size);
2739ec8e2f74SJan Kara reiserfs_truncate_file(inode, 0);
2740ec8e2f74SJan Kara }
2741ec8e2f74SJan Kara
reiserfs_write_begin(struct file * file,struct address_space * mapping,loff_t pos,unsigned len,struct page ** pagep,void ** fsdata)2742ba9d8cecSVladimir Saveliev static int reiserfs_write_begin(struct file *file,
2743ba9d8cecSVladimir Saveliev struct address_space *mapping,
27449d6b0cd7SMatthew Wilcox (Oracle) loff_t pos, unsigned len,
2745ba9d8cecSVladimir Saveliev struct page **pagep, void **fsdata)
2746ba9d8cecSVladimir Saveliev {
2747ba9d8cecSVladimir Saveliev struct inode *inode;
2748ba9d8cecSVladimir Saveliev struct page *page;
2749ba9d8cecSVladimir Saveliev pgoff_t index;
2750ba9d8cecSVladimir Saveliev int ret;
2751ba9d8cecSVladimir Saveliev int old_ref = 0;
2752ba9d8cecSVladimir Saveliev
2753f7557e8fSVladimir Saveliev inode = mapping->host;
275409cbfeafSKirill A. Shutemov index = pos >> PAGE_SHIFT;
2755b7446e7cSMatthew Wilcox (Oracle) page = grab_cache_page_write_begin(mapping, index);
2756ba9d8cecSVladimir Saveliev if (!page)
2757ba9d8cecSVladimir Saveliev return -ENOMEM;
2758ba9d8cecSVladimir Saveliev *pagep = page;
2759ba9d8cecSVladimir Saveliev
2760ba9d8cecSVladimir Saveliev reiserfs_wait_on_write_block(inode->i_sb);
2761ba9d8cecSVladimir Saveliev fix_tail_page_for_writing(page);
2762ba9d8cecSVladimir Saveliev if (reiserfs_transaction_running(inode->i_sb)) {
2763ba9d8cecSVladimir Saveliev struct reiserfs_transaction_handle *th;
2764ba9d8cecSVladimir Saveliev th = (struct reiserfs_transaction_handle *)current->
2765ba9d8cecSVladimir Saveliev journal_info;
2766ba9d8cecSVladimir Saveliev BUG_ON(!th->t_refcount);
2767ba9d8cecSVladimir Saveliev BUG_ON(!th->t_trans_id);
2768ba9d8cecSVladimir Saveliev old_ref = th->t_refcount;
2769ba9d8cecSVladimir Saveliev th->t_refcount++;
2770ba9d8cecSVladimir Saveliev }
27716e1db88dSChristoph Hellwig ret = __block_write_begin(page, pos, len, reiserfs_get_block);
2772ba9d8cecSVladimir Saveliev if (ret && reiserfs_transaction_running(inode->i_sb)) {
2773ba9d8cecSVladimir Saveliev struct reiserfs_transaction_handle *th = current->journal_info;
2774098297b2SJeff Mahoney /*
2775098297b2SJeff Mahoney * this gets a little ugly. If reiserfs_get_block returned an
2776098297b2SJeff Mahoney * error and left a transacstion running, we've got to close
2777098297b2SJeff Mahoney * it, and we've got to free handle if it was a persistent
2778098297b2SJeff Mahoney * transaction.
2779ba9d8cecSVladimir Saveliev *
2780ba9d8cecSVladimir Saveliev * But, if we had nested into an existing transaction, we need
2781ba9d8cecSVladimir Saveliev * to just drop the ref count on the handle.
2782ba9d8cecSVladimir Saveliev *
2783ba9d8cecSVladimir Saveliev * If old_ref == 0, the transaction is from reiserfs_get_block,
2784098297b2SJeff Mahoney * and it was a persistent trans. Otherwise, it was nested
2785098297b2SJeff Mahoney * above.
2786ba9d8cecSVladimir Saveliev */
2787ba9d8cecSVladimir Saveliev if (th->t_refcount > old_ref) {
2788ba9d8cecSVladimir Saveliev if (old_ref)
2789ba9d8cecSVladimir Saveliev th->t_refcount--;
2790ba9d8cecSVladimir Saveliev else {
2791ba9d8cecSVladimir Saveliev int err;
2792ba9d8cecSVladimir Saveliev reiserfs_write_lock(inode->i_sb);
2793ba9d8cecSVladimir Saveliev err = reiserfs_end_persistent_transaction(th);
2794ba9d8cecSVladimir Saveliev reiserfs_write_unlock(inode->i_sb);
2795ba9d8cecSVladimir Saveliev if (err)
2796ba9d8cecSVladimir Saveliev ret = err;
2797ba9d8cecSVladimir Saveliev }
2798ba9d8cecSVladimir Saveliev }
2799ba9d8cecSVladimir Saveliev }
2800ba9d8cecSVladimir Saveliev if (ret) {
2801ba9d8cecSVladimir Saveliev unlock_page(page);
280209cbfeafSKirill A. Shutemov put_page(page);
2803ec8e2f74SJan Kara /* Truncate allocated blocks */
2804ec8e2f74SJan Kara reiserfs_truncate_failed_write(inode);
2805ba9d8cecSVladimir Saveliev }
2806ba9d8cecSVladimir Saveliev return ret;
2807ba9d8cecSVladimir Saveliev }
2808ba9d8cecSVladimir Saveliev
__reiserfs_write_begin(struct page * page,unsigned from,unsigned len)2809ebdec241SChristoph Hellwig int __reiserfs_write_begin(struct page *page, unsigned from, unsigned len)
2810bd4c625cSLinus Torvalds {
28111da177e4SLinus Torvalds struct inode *inode = page->mapping->host;
28121da177e4SLinus Torvalds int ret;
28131da177e4SLinus Torvalds int old_ref = 0;
2814278f6679SJeff Mahoney int depth;
28151da177e4SLinus Torvalds
2816278f6679SJeff Mahoney depth = reiserfs_write_unlock_nested(inode->i_sb);
28171da177e4SLinus Torvalds reiserfs_wait_on_write_block(inode->i_sb);
2818278f6679SJeff Mahoney reiserfs_write_lock_nested(inode->i_sb, depth);
28198ebc4232SFrederic Weisbecker
28201da177e4SLinus Torvalds fix_tail_page_for_writing(page);
28211da177e4SLinus Torvalds if (reiserfs_transaction_running(inode->i_sb)) {
28221da177e4SLinus Torvalds struct reiserfs_transaction_handle *th;
2823bd4c625cSLinus Torvalds th = (struct reiserfs_transaction_handle *)current->
2824bd4c625cSLinus Torvalds journal_info;
28251da177e4SLinus Torvalds BUG_ON(!th->t_refcount);
28261da177e4SLinus Torvalds BUG_ON(!th->t_trans_id);
28271da177e4SLinus Torvalds old_ref = th->t_refcount;
28281da177e4SLinus Torvalds th->t_refcount++;
28291da177e4SLinus Torvalds }
28301da177e4SLinus Torvalds
2831ebdec241SChristoph Hellwig ret = __block_write_begin(page, from, len, reiserfs_get_block);
28321da177e4SLinus Torvalds if (ret && reiserfs_transaction_running(inode->i_sb)) {
28331da177e4SLinus Torvalds struct reiserfs_transaction_handle *th = current->journal_info;
2834098297b2SJeff Mahoney /*
2835098297b2SJeff Mahoney * this gets a little ugly. If reiserfs_get_block returned an
2836098297b2SJeff Mahoney * error and left a transacstion running, we've got to close
2837098297b2SJeff Mahoney * it, and we've got to free handle if it was a persistent
2838098297b2SJeff Mahoney * transaction.
28391da177e4SLinus Torvalds *
28401da177e4SLinus Torvalds * But, if we had nested into an existing transaction, we need
28411da177e4SLinus Torvalds * to just drop the ref count on the handle.
28421da177e4SLinus Torvalds *
28431da177e4SLinus Torvalds * If old_ref == 0, the transaction is from reiserfs_get_block,
2844098297b2SJeff Mahoney * and it was a persistent trans. Otherwise, it was nested
2845098297b2SJeff Mahoney * above.
28461da177e4SLinus Torvalds */
28471da177e4SLinus Torvalds if (th->t_refcount > old_ref) {
28481da177e4SLinus Torvalds if (old_ref)
28491da177e4SLinus Torvalds th->t_refcount--;
28501da177e4SLinus Torvalds else {
28511da177e4SLinus Torvalds int err;
28521da177e4SLinus Torvalds reiserfs_write_lock(inode->i_sb);
28531da177e4SLinus Torvalds err = reiserfs_end_persistent_transaction(th);
28541da177e4SLinus Torvalds reiserfs_write_unlock(inode->i_sb);
28551da177e4SLinus Torvalds if (err)
28561da177e4SLinus Torvalds ret = err;
28571da177e4SLinus Torvalds }
28581da177e4SLinus Torvalds }
28591da177e4SLinus Torvalds }
28601da177e4SLinus Torvalds return ret;
28611da177e4SLinus Torvalds
28621da177e4SLinus Torvalds }
28631da177e4SLinus Torvalds
reiserfs_aop_bmap(struct address_space * as,sector_t block)2864bd4c625cSLinus Torvalds static sector_t reiserfs_aop_bmap(struct address_space *as, sector_t block)
2865bd4c625cSLinus Torvalds {
28661da177e4SLinus Torvalds return generic_block_bmap(as, block, reiserfs_bmap);
28671da177e4SLinus Torvalds }
28681da177e4SLinus Torvalds
reiserfs_write_end(struct file * file,struct address_space * mapping,loff_t pos,unsigned len,unsigned copied,struct page * page,void * fsdata)2869ba9d8cecSVladimir Saveliev static int reiserfs_write_end(struct file *file, struct address_space *mapping,
2870ba9d8cecSVladimir Saveliev loff_t pos, unsigned len, unsigned copied,
2871ba9d8cecSVladimir Saveliev struct page *page, void *fsdata)
2872ba9d8cecSVladimir Saveliev {
28734a9622f2SMatthew Wilcox (Oracle) struct folio *folio = page_folio(page);
2874ba9d8cecSVladimir Saveliev struct inode *inode = page->mapping->host;
2875ba9d8cecSVladimir Saveliev int ret = 0;
2876ba9d8cecSVladimir Saveliev int update_sd = 0;
2877ba9d8cecSVladimir Saveliev struct reiserfs_transaction_handle *th;
2878ba9d8cecSVladimir Saveliev unsigned start;
2879d6f5b0aaSFrederic Weisbecker bool locked = false;
2880ba9d8cecSVladimir Saveliev
2881ba9d8cecSVladimir Saveliev reiserfs_wait_on_write_block(inode->i_sb);
2882ba9d8cecSVladimir Saveliev if (reiserfs_transaction_running(inode->i_sb))
2883ba9d8cecSVladimir Saveliev th = current->journal_info;
2884ba9d8cecSVladimir Saveliev else
2885ba9d8cecSVladimir Saveliev th = NULL;
2886ba9d8cecSVladimir Saveliev
288709cbfeafSKirill A. Shutemov start = pos & (PAGE_SIZE - 1);
2888ba9d8cecSVladimir Saveliev if (unlikely(copied < len)) {
28894a9622f2SMatthew Wilcox (Oracle) if (!folio_test_uptodate(folio))
2890ba9d8cecSVladimir Saveliev copied = 0;
2891ba9d8cecSVladimir Saveliev
28924a9622f2SMatthew Wilcox (Oracle) folio_zero_new_buffers(folio, start + copied, start + len);
2893ba9d8cecSVladimir Saveliev }
28944a9622f2SMatthew Wilcox (Oracle) flush_dcache_folio(folio);
2895ba9d8cecSVladimir Saveliev
2896ba9d8cecSVladimir Saveliev reiserfs_commit_page(inode, page, start, start + copied);
2897ba9d8cecSVladimir Saveliev
2898098297b2SJeff Mahoney /*
2899098297b2SJeff Mahoney * generic_commit_write does this for us, but does not update the
2900098297b2SJeff Mahoney * transaction tracking stuff when the size changes. So, we have
2901098297b2SJeff Mahoney * to do the i_size updates here.
2902ba9d8cecSVladimir Saveliev */
2903ec8e2f74SJan Kara if (pos + copied > inode->i_size) {
2904ba9d8cecSVladimir Saveliev struct reiserfs_transaction_handle myth;
2905278f6679SJeff Mahoney reiserfs_write_lock(inode->i_sb);
2906d6f5b0aaSFrederic Weisbecker locked = true;
2907098297b2SJeff Mahoney /*
2908098297b2SJeff Mahoney * If the file have grown beyond the border where it
2909098297b2SJeff Mahoney * can have a tail, unmark it as needing a tail
2910098297b2SJeff Mahoney * packing
2911098297b2SJeff Mahoney */
2912ba9d8cecSVladimir Saveliev if ((have_large_tails(inode->i_sb)
2913ba9d8cecSVladimir Saveliev && inode->i_size > i_block_size(inode) * 4)
2914ba9d8cecSVladimir Saveliev || (have_small_tails(inode->i_sb)
2915ba9d8cecSVladimir Saveliev && inode->i_size > i_block_size(inode)))
2916ba9d8cecSVladimir Saveliev REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
2917ba9d8cecSVladimir Saveliev
2918ba9d8cecSVladimir Saveliev ret = journal_begin(&myth, inode->i_sb, 1);
2919d6f5b0aaSFrederic Weisbecker if (ret)
2920ba9d8cecSVladimir Saveliev goto journal_error;
2921d6f5b0aaSFrederic Weisbecker
2922ba9d8cecSVladimir Saveliev reiserfs_update_inode_transaction(inode);
2923ec8e2f74SJan Kara inode->i_size = pos + copied;
2924ba9d8cecSVladimir Saveliev /*
2925ba9d8cecSVladimir Saveliev * this will just nest into our transaction. It's important
2926098297b2SJeff Mahoney * to use mark_inode_dirty so the inode gets pushed around on
2927098297b2SJeff Mahoney * the dirty lists, and so that O_SYNC works as expected
2928ba9d8cecSVladimir Saveliev */
2929ba9d8cecSVladimir Saveliev mark_inode_dirty(inode);
2930ba9d8cecSVladimir Saveliev reiserfs_update_sd(&myth, inode);
2931ba9d8cecSVladimir Saveliev update_sd = 1;
293258d85426SJeff Mahoney ret = journal_end(&myth);
2933ba9d8cecSVladimir Saveliev if (ret)
2934ba9d8cecSVladimir Saveliev goto journal_error;
2935ba9d8cecSVladimir Saveliev }
2936ba9d8cecSVladimir Saveliev if (th) {
2937d6f5b0aaSFrederic Weisbecker if (!locked) {
2938278f6679SJeff Mahoney reiserfs_write_lock(inode->i_sb);
2939d6f5b0aaSFrederic Weisbecker locked = true;
2940d6f5b0aaSFrederic Weisbecker }
2941ba9d8cecSVladimir Saveliev if (!update_sd)
2942ba9d8cecSVladimir Saveliev mark_inode_dirty(inode);
2943ba9d8cecSVladimir Saveliev ret = reiserfs_end_persistent_transaction(th);
2944ba9d8cecSVladimir Saveliev if (ret)
2945ba9d8cecSVladimir Saveliev goto out;
2946ba9d8cecSVladimir Saveliev }
2947ba9d8cecSVladimir Saveliev
2948ba9d8cecSVladimir Saveliev out:
2949d6f5b0aaSFrederic Weisbecker if (locked)
2950278f6679SJeff Mahoney reiserfs_write_unlock(inode->i_sb);
2951ba9d8cecSVladimir Saveliev unlock_page(page);
295209cbfeafSKirill A. Shutemov put_page(page);
2953ec8e2f74SJan Kara
2954ec8e2f74SJan Kara if (pos + len > inode->i_size)
2955ec8e2f74SJan Kara reiserfs_truncate_failed_write(inode);
2956ec8e2f74SJan Kara
2957ba9d8cecSVladimir Saveliev return ret == 0 ? copied : ret;
2958ba9d8cecSVladimir Saveliev
2959ba9d8cecSVladimir Saveliev journal_error:
2960278f6679SJeff Mahoney reiserfs_write_unlock(inode->i_sb);
2961d6f5b0aaSFrederic Weisbecker locked = false;
2962ba9d8cecSVladimir Saveliev if (th) {
2963ba9d8cecSVladimir Saveliev if (!update_sd)
2964ba9d8cecSVladimir Saveliev reiserfs_update_sd(th, inode);
2965ba9d8cecSVladimir Saveliev ret = reiserfs_end_persistent_transaction(th);
2966ba9d8cecSVladimir Saveliev }
2967ba9d8cecSVladimir Saveliev goto out;
2968ba9d8cecSVladimir Saveliev }
2969ba9d8cecSVladimir Saveliev
reiserfs_commit_write(struct file * f,struct page * page,unsigned from,unsigned to)2970ba9d8cecSVladimir Saveliev int reiserfs_commit_write(struct file *f, struct page *page,
2971bd4c625cSLinus Torvalds unsigned from, unsigned to)
2972bd4c625cSLinus Torvalds {
29731da177e4SLinus Torvalds struct inode *inode = page->mapping->host;
297409cbfeafSKirill A. Shutemov loff_t pos = ((loff_t) page->index << PAGE_SHIFT) + to;
29751da177e4SLinus Torvalds int ret = 0;
29761da177e4SLinus Torvalds int update_sd = 0;
29771da177e4SLinus Torvalds struct reiserfs_transaction_handle *th = NULL;
2978278f6679SJeff Mahoney int depth;
29791da177e4SLinus Torvalds
2980278f6679SJeff Mahoney depth = reiserfs_write_unlock_nested(inode->i_sb);
29811da177e4SLinus Torvalds reiserfs_wait_on_write_block(inode->i_sb);
2982278f6679SJeff Mahoney reiserfs_write_lock_nested(inode->i_sb, depth);
29838ebc4232SFrederic Weisbecker
29841da177e4SLinus Torvalds if (reiserfs_transaction_running(inode->i_sb)) {
29851da177e4SLinus Torvalds th = current->journal_info;
29861da177e4SLinus Torvalds }
29871da177e4SLinus Torvalds reiserfs_commit_page(inode, page, from, to);
29881da177e4SLinus Torvalds
2989098297b2SJeff Mahoney /*
2990098297b2SJeff Mahoney * generic_commit_write does this for us, but does not update the
2991098297b2SJeff Mahoney * transaction tracking stuff when the size changes. So, we have
2992098297b2SJeff Mahoney * to do the i_size updates here.
29931da177e4SLinus Torvalds */
29941da177e4SLinus Torvalds if (pos > inode->i_size) {
29951da177e4SLinus Torvalds struct reiserfs_transaction_handle myth;
2996098297b2SJeff Mahoney /*
2997098297b2SJeff Mahoney * If the file have grown beyond the border where it
2998098297b2SJeff Mahoney * can have a tail, unmark it as needing a tail
2999098297b2SJeff Mahoney * packing
3000098297b2SJeff Mahoney */
3001bd4c625cSLinus Torvalds if ((have_large_tails(inode->i_sb)
3002bd4c625cSLinus Torvalds && inode->i_size > i_block_size(inode) * 4)
3003bd4c625cSLinus Torvalds || (have_small_tails(inode->i_sb)
3004bd4c625cSLinus Torvalds && inode->i_size > i_block_size(inode)))
30051da177e4SLinus Torvalds REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
30061da177e4SLinus Torvalds
30071da177e4SLinus Torvalds ret = journal_begin(&myth, inode->i_sb, 1);
30087e942770SFrederic Weisbecker if (ret)
30091da177e4SLinus Torvalds goto journal_error;
30107e942770SFrederic Weisbecker
30111da177e4SLinus Torvalds reiserfs_update_inode_transaction(inode);
30121da177e4SLinus Torvalds inode->i_size = pos;
30139f03783cSChris Mason /*
30149f03783cSChris Mason * this will just nest into our transaction. It's important
3015098297b2SJeff Mahoney * to use mark_inode_dirty so the inode gets pushed around
3016098297b2SJeff Mahoney * on the dirty lists, and so that O_SYNC works as expected
30179f03783cSChris Mason */
30189f03783cSChris Mason mark_inode_dirty(inode);
30191da177e4SLinus Torvalds reiserfs_update_sd(&myth, inode);
30201da177e4SLinus Torvalds update_sd = 1;
302158d85426SJeff Mahoney ret = journal_end(&myth);
30221da177e4SLinus Torvalds if (ret)
30231da177e4SLinus Torvalds goto journal_error;
30241da177e4SLinus Torvalds }
30251da177e4SLinus Torvalds if (th) {
30261da177e4SLinus Torvalds if (!update_sd)
30279f03783cSChris Mason mark_inode_dirty(inode);
30281da177e4SLinus Torvalds ret = reiserfs_end_persistent_transaction(th);
30291da177e4SLinus Torvalds if (ret)
30301da177e4SLinus Torvalds goto out;
30311da177e4SLinus Torvalds }
30321da177e4SLinus Torvalds
30331da177e4SLinus Torvalds out:
30341da177e4SLinus Torvalds return ret;
30351da177e4SLinus Torvalds
30361da177e4SLinus Torvalds journal_error:
30371da177e4SLinus Torvalds if (th) {
30381da177e4SLinus Torvalds if (!update_sd)
30391da177e4SLinus Torvalds reiserfs_update_sd(th, inode);
30401da177e4SLinus Torvalds ret = reiserfs_end_persistent_transaction(th);
30411da177e4SLinus Torvalds }
30421da177e4SLinus Torvalds
30431da177e4SLinus Torvalds return ret;
30441da177e4SLinus Torvalds }
30451da177e4SLinus Torvalds
sd_attrs_to_i_attrs(__u16 sd_attrs,struct inode * inode)30461da177e4SLinus Torvalds void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode)
30471da177e4SLinus Torvalds {
30481da177e4SLinus Torvalds if (reiserfs_attrs(inode->i_sb)) {
30491da177e4SLinus Torvalds if (sd_attrs & REISERFS_SYNC_FL)
30501da177e4SLinus Torvalds inode->i_flags |= S_SYNC;
30511da177e4SLinus Torvalds else
30521da177e4SLinus Torvalds inode->i_flags &= ~S_SYNC;
30531da177e4SLinus Torvalds if (sd_attrs & REISERFS_IMMUTABLE_FL)
30541da177e4SLinus Torvalds inode->i_flags |= S_IMMUTABLE;
30551da177e4SLinus Torvalds else
30561da177e4SLinus Torvalds inode->i_flags &= ~S_IMMUTABLE;
30571da177e4SLinus Torvalds if (sd_attrs & REISERFS_APPEND_FL)
30581da177e4SLinus Torvalds inode->i_flags |= S_APPEND;
30591da177e4SLinus Torvalds else
30601da177e4SLinus Torvalds inode->i_flags &= ~S_APPEND;
30611da177e4SLinus Torvalds if (sd_attrs & REISERFS_NOATIME_FL)
30621da177e4SLinus Torvalds inode->i_flags |= S_NOATIME;
30631da177e4SLinus Torvalds else
30641da177e4SLinus Torvalds inode->i_flags &= ~S_NOATIME;
30651da177e4SLinus Torvalds if (sd_attrs & REISERFS_NOTAIL_FL)
30661da177e4SLinus Torvalds REISERFS_I(inode)->i_flags |= i_nopack_mask;
30671da177e4SLinus Torvalds else
30681da177e4SLinus Torvalds REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
30691da177e4SLinus Torvalds }
30701da177e4SLinus Torvalds }
30711da177e4SLinus Torvalds
3072098297b2SJeff Mahoney /*
3073098297b2SJeff Mahoney * decide if this buffer needs to stay around for data logging or ordered
3074098297b2SJeff Mahoney * write purposes
30751da177e4SLinus Torvalds */
invalidate_folio_can_drop(struct inode * inode,struct buffer_head * bh)3076d97dfc94SMatthew Wilcox (Oracle) static int invalidate_folio_can_drop(struct inode *inode, struct buffer_head *bh)
30771da177e4SLinus Torvalds {
30781da177e4SLinus Torvalds int ret = 1;
30791da177e4SLinus Torvalds struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
30801da177e4SLinus Torvalds
3081d62b1b87SChris Mason lock_buffer(bh);
30821da177e4SLinus Torvalds spin_lock(&j->j_dirty_buffers_lock);
30831da177e4SLinus Torvalds if (!buffer_mapped(bh)) {
30841da177e4SLinus Torvalds goto free_jh;
30851da177e4SLinus Torvalds }
3086098297b2SJeff Mahoney /*
3087098297b2SJeff Mahoney * the page is locked, and the only places that log a data buffer
30881da177e4SLinus Torvalds * also lock the page.
30891da177e4SLinus Torvalds */
30901da177e4SLinus Torvalds if (reiserfs_file_data_log(inode)) {
30911da177e4SLinus Torvalds /*
30921da177e4SLinus Torvalds * very conservative, leave the buffer pinned if
30931da177e4SLinus Torvalds * anyone might need it.
30941da177e4SLinus Torvalds */
30951da177e4SLinus Torvalds if (buffer_journaled(bh) || buffer_journal_dirty(bh)) {
30961da177e4SLinus Torvalds ret = 0;
30971da177e4SLinus Torvalds }
3098d62b1b87SChris Mason } else if (buffer_dirty(bh)) {
30991da177e4SLinus Torvalds struct reiserfs_journal_list *jl;
31001da177e4SLinus Torvalds struct reiserfs_jh *jh = bh->b_private;
31011da177e4SLinus Torvalds
3102098297b2SJeff Mahoney /*
3103098297b2SJeff Mahoney * why is this safe?
31041da177e4SLinus Torvalds * reiserfs_setattr updates i_size in the on disk
31051da177e4SLinus Torvalds * stat data before allowing vmtruncate to be called.
31061da177e4SLinus Torvalds *
31071da177e4SLinus Torvalds * If buffer was put onto the ordered list for this
31081da177e4SLinus Torvalds * transaction, we know for sure either this transaction
31091da177e4SLinus Torvalds * or an older one already has updated i_size on disk,
31101da177e4SLinus Torvalds * and this ordered data won't be referenced in the file
31111da177e4SLinus Torvalds * if we crash.
31121da177e4SLinus Torvalds *
31131da177e4SLinus Torvalds * if the buffer was put onto the ordered list for an older
31141da177e4SLinus Torvalds * transaction, we need to leave it around
31151da177e4SLinus Torvalds */
3116bd4c625cSLinus Torvalds if (jh && (jl = jh->jl)
3117bd4c625cSLinus Torvalds && jl != SB_JOURNAL(inode->i_sb)->j_current_jl)
31181da177e4SLinus Torvalds ret = 0;
31191da177e4SLinus Torvalds }
31201da177e4SLinus Torvalds free_jh:
31211da177e4SLinus Torvalds if (ret && bh->b_private) {
31221da177e4SLinus Torvalds reiserfs_free_jh(bh);
31231da177e4SLinus Torvalds }
31241da177e4SLinus Torvalds spin_unlock(&j->j_dirty_buffers_lock);
3125d62b1b87SChris Mason unlock_buffer(bh);
31261da177e4SLinus Torvalds return ret;
31271da177e4SLinus Torvalds }
31281da177e4SLinus Torvalds
3129d97dfc94SMatthew Wilcox (Oracle) /* clm -- taken from fs/buffer.c:block_invalidate_folio */
reiserfs_invalidate_folio(struct folio * folio,size_t offset,size_t length)3130d97dfc94SMatthew Wilcox (Oracle) static void reiserfs_invalidate_folio(struct folio *folio, size_t offset,
3131d97dfc94SMatthew Wilcox (Oracle) size_t length)
31321da177e4SLinus Torvalds {
31331da177e4SLinus Torvalds struct buffer_head *head, *bh, *next;
3134d97dfc94SMatthew Wilcox (Oracle) struct inode *inode = folio->mapping->host;
31351da177e4SLinus Torvalds unsigned int curr_off = 0;
3136bad54831SLukas Czerner unsigned int stop = offset + length;
3137d97dfc94SMatthew Wilcox (Oracle) int partial_page = (offset || length < folio_size(folio));
31381da177e4SLinus Torvalds int ret = 1;
31391da177e4SLinus Torvalds
3140d97dfc94SMatthew Wilcox (Oracle) BUG_ON(!folio_test_locked(folio));
31411da177e4SLinus Torvalds
3142bad54831SLukas Czerner if (!partial_page)
3143d97dfc94SMatthew Wilcox (Oracle) folio_clear_checked(folio);
31441da177e4SLinus Torvalds
3145d97dfc94SMatthew Wilcox (Oracle) head = folio_buffers(folio);
3146d97dfc94SMatthew Wilcox (Oracle) if (!head)
31471da177e4SLinus Torvalds goto out;
31481da177e4SLinus Torvalds
31491da177e4SLinus Torvalds bh = head;
31501da177e4SLinus Torvalds do {
31511da177e4SLinus Torvalds unsigned int next_off = curr_off + bh->b_size;
31521da177e4SLinus Torvalds next = bh->b_this_page;
31531da177e4SLinus Torvalds
3154bad54831SLukas Czerner if (next_off > stop)
3155bad54831SLukas Czerner goto out;
3156bad54831SLukas Czerner
31571da177e4SLinus Torvalds /*
31581da177e4SLinus Torvalds * is this block fully invalidated?
31591da177e4SLinus Torvalds */
31601da177e4SLinus Torvalds if (offset <= curr_off) {
3161d97dfc94SMatthew Wilcox (Oracle) if (invalidate_folio_can_drop(inode, bh))
31621da177e4SLinus Torvalds reiserfs_unmap_buffer(bh);
31631da177e4SLinus Torvalds else
31641da177e4SLinus Torvalds ret = 0;
31651da177e4SLinus Torvalds }
31661da177e4SLinus Torvalds curr_off = next_off;
31671da177e4SLinus Torvalds bh = next;
31681da177e4SLinus Torvalds } while (bh != head);
31691da177e4SLinus Torvalds
31701da177e4SLinus Torvalds /*
31711da177e4SLinus Torvalds * We release buffers only if the entire page is being invalidated.
31721da177e4SLinus Torvalds * The get_block cached value has been unconditionally invalidated,
31731da177e4SLinus Torvalds * so real IO is not possible anymore.
31741da177e4SLinus Torvalds */
3175bad54831SLukas Czerner if (!partial_page && ret) {
3176d97dfc94SMatthew Wilcox (Oracle) ret = filemap_release_folio(folio, 0);
31772ff28e22SNeilBrown /* maybe should BUG_ON(!ret); - neilb */
31782ff28e22SNeilBrown }
31791da177e4SLinus Torvalds out:
31802ff28e22SNeilBrown return;
31811da177e4SLinus Torvalds }
31821da177e4SLinus Torvalds
reiserfs_dirty_folio(struct address_space * mapping,struct folio * folio)3183e621900aSMatthew Wilcox (Oracle) static bool reiserfs_dirty_folio(struct address_space *mapping,
3184e621900aSMatthew Wilcox (Oracle) struct folio *folio)
3185bd4c625cSLinus Torvalds {
3186e621900aSMatthew Wilcox (Oracle) if (reiserfs_file_data_log(mapping->host)) {
3187e621900aSMatthew Wilcox (Oracle) folio_set_checked(folio);
3188e621900aSMatthew Wilcox (Oracle) return filemap_dirty_folio(mapping, folio);
31891da177e4SLinus Torvalds }
3190e621900aSMatthew Wilcox (Oracle) return block_dirty_folio(mapping, folio);
31911da177e4SLinus Torvalds }
31921da177e4SLinus Torvalds
31931da177e4SLinus Torvalds /*
3194dc2e58b2SMatthew Wilcox (Oracle) * Returns true if the folio's buffers were dropped. The folio is locked.
31951da177e4SLinus Torvalds *
31961da177e4SLinus Torvalds * Takes j_dirty_buffers_lock to protect the b_assoc_buffers list_heads
3197dc2e58b2SMatthew Wilcox (Oracle) * in the buffers at folio_buffers(folio).
31981da177e4SLinus Torvalds *
31991da177e4SLinus Torvalds * even in -o notail mode, we can't be sure an old mount without -o notail
32001da177e4SLinus Torvalds * didn't create files with tails.
32011da177e4SLinus Torvalds */
reiserfs_release_folio(struct folio * folio,gfp_t unused_gfp_flags)3202dc2e58b2SMatthew Wilcox (Oracle) static bool reiserfs_release_folio(struct folio *folio, gfp_t unused_gfp_flags)
32031da177e4SLinus Torvalds {
3204dc2e58b2SMatthew Wilcox (Oracle) struct inode *inode = folio->mapping->host;
32051da177e4SLinus Torvalds struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
32061da177e4SLinus Torvalds struct buffer_head *head;
32071da177e4SLinus Torvalds struct buffer_head *bh;
3208dc2e58b2SMatthew Wilcox (Oracle) bool ret = true;
32091da177e4SLinus Torvalds
3210dc2e58b2SMatthew Wilcox (Oracle) WARN_ON(folio_test_checked(folio));
32111da177e4SLinus Torvalds spin_lock(&j->j_dirty_buffers_lock);
3212dc2e58b2SMatthew Wilcox (Oracle) head = folio_buffers(folio);
32131da177e4SLinus Torvalds bh = head;
32141da177e4SLinus Torvalds do {
32151da177e4SLinus Torvalds if (bh->b_private) {
32161da177e4SLinus Torvalds if (!buffer_dirty(bh) && !buffer_locked(bh)) {
32171da177e4SLinus Torvalds reiserfs_free_jh(bh);
32181da177e4SLinus Torvalds } else {
3219dc2e58b2SMatthew Wilcox (Oracle) ret = false;
32201da177e4SLinus Torvalds break;
32211da177e4SLinus Torvalds }
32221da177e4SLinus Torvalds }
32231da177e4SLinus Torvalds bh = bh->b_this_page;
32241da177e4SLinus Torvalds } while (bh != head);
32251da177e4SLinus Torvalds if (ret)
322668189fefSMatthew Wilcox (Oracle) ret = try_to_free_buffers(folio);
32271da177e4SLinus Torvalds spin_unlock(&j->j_dirty_buffers_lock);
32281da177e4SLinus Torvalds return ret;
32291da177e4SLinus Torvalds }
32301da177e4SLinus Torvalds
3231098297b2SJeff Mahoney /*
3232098297b2SJeff Mahoney * We thank Mingming Cao for helping us understand in great detail what
3233098297b2SJeff Mahoney * to do in this section of the code.
3234098297b2SJeff Mahoney */
reiserfs_direct_IO(struct kiocb * iocb,struct iov_iter * iter)3235c8b8e32dSChristoph Hellwig static ssize_t reiserfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
32361da177e4SLinus Torvalds {
32371da177e4SLinus Torvalds struct file *file = iocb->ki_filp;
32381da177e4SLinus Torvalds struct inode *inode = file->f_mapping->host;
3239a6cbcd4aSAl Viro size_t count = iov_iter_count(iter);
3240eafdc7d1SChristoph Hellwig ssize_t ret;
32411da177e4SLinus Torvalds
3242c8b8e32dSChristoph Hellwig ret = blockdev_direct_IO(iocb, inode, iter,
3243aacfc19cSChristoph Hellwig reiserfs_get_blocks_direct_io);
3244eafdc7d1SChristoph Hellwig
3245eafdc7d1SChristoph Hellwig /*
3246eafdc7d1SChristoph Hellwig * In case of error extending write may have instantiated a few
3247eafdc7d1SChristoph Hellwig * blocks outside i_size. Trim these off again.
3248eafdc7d1SChristoph Hellwig */
32496f673763SOmar Sandoval if (unlikely(iov_iter_rw(iter) == WRITE && ret < 0)) {
3250eafdc7d1SChristoph Hellwig loff_t isize = i_size_read(inode);
3251c8b8e32dSChristoph Hellwig loff_t end = iocb->ki_pos + count;
3252eafdc7d1SChristoph Hellwig
3253cfac4b47SMarco Stornelli if ((end > isize) && inode_newsize_ok(inode, isize) == 0) {
3254cfac4b47SMarco Stornelli truncate_setsize(inode, isize);
3255cfac4b47SMarco Stornelli reiserfs_vfs_truncate_file(inode);
3256cfac4b47SMarco Stornelli }
3257eafdc7d1SChristoph Hellwig }
3258eafdc7d1SChristoph Hellwig
3259eafdc7d1SChristoph Hellwig return ret;
32601da177e4SLinus Torvalds }
32611da177e4SLinus Torvalds
reiserfs_setattr(struct mnt_idmap * idmap,struct dentry * dentry,struct iattr * attr)3262c1632a0fSChristian Brauner int reiserfs_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
3263549c7297SChristian Brauner struct iattr *attr)
3264bd4c625cSLinus Torvalds {
32652b0143b5SDavid Howells struct inode *inode = d_inode(dentry);
3266cdd6fe6eSJeff Layton unsigned int ia_valid;
32675fe1533fSFrederic Weisbecker int error;
3268cdd6fe6eSJeff Layton
3269c1632a0fSChristian Brauner error = setattr_prepare(&nop_mnt_idmap, dentry, attr);
3270db78b877SChristoph Hellwig if (error)
3271db78b877SChristoph Hellwig return error;
3272db78b877SChristoph Hellwig
3273cdd6fe6eSJeff Layton /* must be turned off for recursive notify_change calls */
3274cdd6fe6eSJeff Layton ia_valid = attr->ia_valid &= ~(ATTR_KILL_SUID|ATTR_KILL_SGID);
3275cdd6fe6eSJeff Layton
3276f861646aSChristian Brauner if (is_quota_modification(&nop_mnt_idmap, inode, attr)) {
32772e6c97eaSJan Kara error = dquot_initialize(inode);
32782e6c97eaSJan Kara if (error)
32792e6c97eaSJan Kara return error;
32802e6c97eaSJan Kara }
3281278f6679SJeff Mahoney reiserfs_write_lock(inode->i_sb);
328212755627SDmitry Monakhov if (attr->ia_valid & ATTR_SIZE) {
3283098297b2SJeff Mahoney /*
3284098297b2SJeff Mahoney * version 2 items will be caught by the s_maxbytes check
3285098297b2SJeff Mahoney * done for us in vmtruncate
32861da177e4SLinus Torvalds */
32871da177e4SLinus Torvalds if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5 &&
32881da177e4SLinus Torvalds attr->ia_size > MAX_NON_LFS) {
3289278f6679SJeff Mahoney reiserfs_write_unlock(inode->i_sb);
32901da177e4SLinus Torvalds error = -EFBIG;
32911da177e4SLinus Torvalds goto out;
32921da177e4SLinus Torvalds }
3293562c72aaSChristoph Hellwig
3294562c72aaSChristoph Hellwig inode_dio_wait(inode);
3295562c72aaSChristoph Hellwig
32961da177e4SLinus Torvalds /* fill in hole pointers in the expanding truncate case. */
32971da177e4SLinus Torvalds if (attr->ia_size > inode->i_size) {
329848b0e011SEdward Shishkin loff_t pos = attr->ia_size;
329948b0e011SEdward Shishkin
330048b0e011SEdward Shishkin if ((pos & (inode->i_sb->s_blocksize - 1)) == 0)
330148b0e011SEdward Shishkin pos++;
330248b0e011SEdward Shishkin error = generic_cont_expand_simple(inode, pos);
33031da177e4SLinus Torvalds if (REISERFS_I(inode)->i_prealloc_count > 0) {
33041da177e4SLinus Torvalds int err;
33051da177e4SLinus Torvalds struct reiserfs_transaction_handle th;
33061da177e4SLinus Torvalds /* we're changing at most 2 bitmaps, inode + super */
33071da177e4SLinus Torvalds err = journal_begin(&th, inode->i_sb, 4);
33081da177e4SLinus Torvalds if (!err) {
33091da177e4SLinus Torvalds reiserfs_discard_prealloc(&th, inode);
331058d85426SJeff Mahoney err = journal_end(&th);
33111da177e4SLinus Torvalds }
33121da177e4SLinus Torvalds if (err)
33131da177e4SLinus Torvalds error = err;
33141da177e4SLinus Torvalds }
33154c05141dSJeff Mahoney if (error) {
3316278f6679SJeff Mahoney reiserfs_write_unlock(inode->i_sb);
33171da177e4SLinus Torvalds goto out;
33184c05141dSJeff Mahoney }
3319dd535a59SVladimir Saveliev /*
3320dd535a59SVladimir Saveliev * file size is changed, ctime and mtime are
3321dd535a59SVladimir Saveliev * to be updated
3322dd535a59SVladimir Saveliev */
3323dd535a59SVladimir Saveliev attr->ia_valid |= (ATTR_MTIME | ATTR_CTIME);
33241da177e4SLinus Torvalds }
33251da177e4SLinus Torvalds }
3326278f6679SJeff Mahoney reiserfs_write_unlock(inode->i_sb);
33271da177e4SLinus Torvalds
3328df814654SEric W. Biederman if ((((attr->ia_valid & ATTR_UID) && (from_kuid(&init_user_ns, attr->ia_uid) & ~0xffff)) ||
3329df814654SEric W. Biederman ((attr->ia_valid & ATTR_GID) && (from_kgid(&init_user_ns, attr->ia_gid) & ~0xffff))) &&
33301da177e4SLinus Torvalds (get_inode_sd_version(inode) == STAT_DATA_V1)) {
33311da177e4SLinus Torvalds /* stat data of format v3.5 has 16 bit uid and gid */
33321da177e4SLinus Torvalds error = -EINVAL;
33331da177e4SLinus Torvalds goto out;
33341da177e4SLinus Torvalds }
33351da177e4SLinus Torvalds
3336df814654SEric W. Biederman if ((ia_valid & ATTR_UID && !uid_eq(attr->ia_uid, inode->i_uid)) ||
3337df814654SEric W. Biederman (ia_valid & ATTR_GID && !gid_eq(attr->ia_gid, inode->i_gid))) {
33381da177e4SLinus Torvalds struct reiserfs_transaction_handle th;
3339bd4c625cSLinus Torvalds int jbegin_count =
3340bd4c625cSLinus Torvalds 2 *
3341bd4c625cSLinus Torvalds (REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb) +
3342bd4c625cSLinus Torvalds REISERFS_QUOTA_DEL_BLOCKS(inode->i_sb)) +
3343bd4c625cSLinus Torvalds 2;
33441da177e4SLinus Torvalds
33451025774cSChristoph Hellwig error = reiserfs_chown_xattrs(inode, attr);
33461025774cSChristoph Hellwig
33471025774cSChristoph Hellwig if (error)
33481025774cSChristoph Hellwig return error;
33491025774cSChristoph Hellwig
3350098297b2SJeff Mahoney /*
3351098297b2SJeff Mahoney * (user+group)*(old+new) structure - we count quota
3352098297b2SJeff Mahoney * info and , inode write (sb, inode)
3353098297b2SJeff Mahoney */
3354278f6679SJeff Mahoney reiserfs_write_lock(inode->i_sb);
33551025774cSChristoph Hellwig error = journal_begin(&th, inode->i_sb, jbegin_count);
3356278f6679SJeff Mahoney reiserfs_write_unlock(inode->i_sb);
3357bd6a1f16SJan Kara if (error)
3358bd6a1f16SJan Kara goto out;
3359f861646aSChristian Brauner error = dquot_transfer(&nop_mnt_idmap, inode, attr);
3360278f6679SJeff Mahoney reiserfs_write_lock(inode->i_sb);
33611da177e4SLinus Torvalds if (error) {
336258d85426SJeff Mahoney journal_end(&th);
3363278f6679SJeff Mahoney reiserfs_write_unlock(inode->i_sb);
33641da177e4SLinus Torvalds goto out;
33651da177e4SLinus Torvalds }
33661025774cSChristoph Hellwig
3367098297b2SJeff Mahoney /*
3368098297b2SJeff Mahoney * Update corresponding info in inode so that everything
3369098297b2SJeff Mahoney * is in one transaction
3370098297b2SJeff Mahoney */
33711da177e4SLinus Torvalds if (attr->ia_valid & ATTR_UID)
33721da177e4SLinus Torvalds inode->i_uid = attr->ia_uid;
33731da177e4SLinus Torvalds if (attr->ia_valid & ATTR_GID)
33741da177e4SLinus Torvalds inode->i_gid = attr->ia_gid;
33751da177e4SLinus Torvalds mark_inode_dirty(inode);
337658d85426SJeff Mahoney error = journal_end(&th);
3377278f6679SJeff Mahoney reiserfs_write_unlock(inode->i_sb);
33781025774cSChristoph Hellwig if (error)
33791025774cSChristoph Hellwig goto out;
33801da177e4SLinus Torvalds }
33811025774cSChristoph Hellwig
33821025774cSChristoph Hellwig if ((attr->ia_valid & ATTR_SIZE) &&
3383cfac4b47SMarco Stornelli attr->ia_size != i_size_read(inode)) {
3384cfac4b47SMarco Stornelli error = inode_newsize_ok(inode, attr->ia_size);
3385cfac4b47SMarco Stornelli if (!error) {
338622e7478dSJeff Mahoney /*
338722e7478dSJeff Mahoney * Could race against reiserfs_file_release
338822e7478dSJeff Mahoney * if called from NFS, so take tailpack mutex.
338922e7478dSJeff Mahoney */
339022e7478dSJeff Mahoney mutex_lock(&REISERFS_I(inode)->tailpack);
3391cfac4b47SMarco Stornelli truncate_setsize(inode, attr->ia_size);
339222e7478dSJeff Mahoney reiserfs_truncate_file(inode, 1);
339322e7478dSJeff Mahoney mutex_unlock(&REISERFS_I(inode)->tailpack);
3394cfac4b47SMarco Stornelli }
3395cfac4b47SMarco Stornelli }
33961025774cSChristoph Hellwig
33971025774cSChristoph Hellwig if (!error) {
3398c1632a0fSChristian Brauner setattr_copy(&nop_mnt_idmap, inode, attr);
33991025774cSChristoph Hellwig mark_inode_dirty(inode);
34001025774cSChristoph Hellwig }
34011da177e4SLinus Torvalds
34021da177e4SLinus Torvalds if (!error && reiserfs_posixacl(inode->i_sb)) {
34031da177e4SLinus Torvalds if (attr->ia_valid & ATTR_MODE)
3404138060baSChristian Brauner error = reiserfs_acl_chmod(dentry);
34051da177e4SLinus Torvalds }
34061da177e4SLinus Torvalds
34071da177e4SLinus Torvalds out:
34081da177e4SLinus Torvalds return error;
34091da177e4SLinus Torvalds }
34101da177e4SLinus Torvalds
3411f5e54d6eSChristoph Hellwig const struct address_space_operations reiserfs_address_space_operations = {
34121da177e4SLinus Torvalds .writepage = reiserfs_writepage,
34132c69e205SMatthew Wilcox (Oracle) .read_folio = reiserfs_read_folio,
3414d4388340SMatthew Wilcox (Oracle) .readahead = reiserfs_readahead,
3415dc2e58b2SMatthew Wilcox (Oracle) .release_folio = reiserfs_release_folio,
3416d97dfc94SMatthew Wilcox (Oracle) .invalidate_folio = reiserfs_invalidate_folio,
3417ba9d8cecSVladimir Saveliev .write_begin = reiserfs_write_begin,
3418ba9d8cecSVladimir Saveliev .write_end = reiserfs_write_end,
34191da177e4SLinus Torvalds .bmap = reiserfs_aop_bmap,
34201da177e4SLinus Torvalds .direct_IO = reiserfs_direct_IO,
3421e621900aSMatthew Wilcox (Oracle) .dirty_folio = reiserfs_dirty_folio,
34221da177e4SLinus Torvalds };
3423