blob: 9dfd768ed9f8d9044b55fe8b9fe888e3cd6794b9 [file] [log] [blame]
Greg Kroah-Hartmanb2441312017-11-01 15:07:57 +01001// SPDX-License-Identifier: GPL-2.0
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002/*
Mingming Cao617ba132006-10-11 01:20:53 -07003 * linux/fs/ext4/ialloc.c
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004 *
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
9 *
10 * BSD ufs-inspired inode and directory allocation by
11 * Stephen Tweedie (sct@redhat.com), 1993
12 * Big-endian to little-endian byte-swapping/bitmaps by
13 * David S. Miller (davem@caip.rutgers.edu), 1995
14 */
15
16#include <linux/time.h>
17#include <linux/fs.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070018#include <linux/stat.h>
19#include <linux/string.h>
20#include <linux/quotaops.h>
21#include <linux/buffer_head.h>
22#include <linux/random.h>
23#include <linux/bitops.h>
Mingming Cao3a5b2ec2006-10-11 01:21:05 -070024#include <linux/blkdev.h>
Ingo Molnar5b825c32017-02-02 17:54:15 +010025#include <linux/cred.h>
26
Dave Kleikampac27a0e2006-10-11 01:20:50 -070027#include <asm/byteorder.h>
Theodore Ts'o9bffad12009-06-17 11:48:11 -040028
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040029#include "ext4.h"
30#include "ext4_jbd2.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070031#include "xattr.h"
32#include "acl.h"
33
Theodore Ts'o9bffad12009-06-17 11:48:11 -040034#include <trace/events/ext4.h>
35
Dave Kleikampac27a0e2006-10-11 01:20:50 -070036/*
37 * ialloc.c contains the inodes allocation and deallocation routines
38 */
39
40/*
41 * The free inodes are managed by bitmaps. A file system contains several
42 * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap
43 * block for inodes, N blocks for the inode table and data blocks.
44 *
45 * The file system contains group descriptors which are located after the
46 * super block. Each descriptor contains the number of the bitmap block and
47 * the free blocks count in the block.
48 */
49
Andreas Dilger717d50e2007-10-16 18:38:25 -040050/*
51 * To avoid calling the atomic setbit hundreds or thousands of times, we only
52 * need to use it within a single byte (to ensure we get endianness right).
53 * We can use memset for the rest of the bitmap as there are no other users.
54 */
Theodore Ts'o61d08672010-10-27 21:30:15 -040055void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap)
Andreas Dilger717d50e2007-10-16 18:38:25 -040056{
57 int i;
58
59 if (start_bit >= end_bit)
60 return;
61
62 ext4_debug("mark end bits +%d through +%d used\n", start_bit, end_bit);
63 for (i = start_bit; i < ((start_bit + 7) & ~7UL); i++)
64 ext4_set_bit(i, bitmap);
65 if (i < end_bit)
66 memset(bitmap + (i >> 3), 0xff, (end_bit - i) >> 3);
67}
68
Theodore Ts'o813e5722012-02-20 17:52:46 -050069void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate)
70{
71 if (uptodate) {
72 set_buffer_uptodate(bh);
73 set_bitmap_uptodate(bh);
74 }
75 unlock_buffer(bh);
76 put_bh(bh);
77}
78
Darrick J. Wong9008a582015-10-17 21:33:24 -040079static int ext4_validate_inode_bitmap(struct super_block *sb,
80 struct ext4_group_desc *desc,
81 ext4_group_t block_group,
82 struct buffer_head *bh)
83{
84 ext4_fsblk_t blk;
Harshad Shirwadkar8016e292020-10-15 13:37:59 -070085 struct ext4_group_info *grp;
86
87 if (EXT4_SB(sb)->s_mount_state & EXT4_FC_REPLAY)
88 return 0;
89
90 grp = ext4_get_group_info(sb, block_group);
Darrick J. Wong9008a582015-10-17 21:33:24 -040091
92 if (buffer_verified(bh))
93 return 0;
Theodore Ts'o5354b2a2023-04-29 00:06:28 -040094 if (!grp || EXT4_MB_GRP_IBITMAP_CORRUPT(grp))
Darrick J. Wong9008a582015-10-17 21:33:24 -040095 return -EFSCORRUPTED;
96
97 ext4_lock_group(sb, block_group);
Theodore Ts'o8d5a8032018-07-12 19:08:05 -040098 if (buffer_verified(bh))
99 goto verified;
Darrick J. Wong9008a582015-10-17 21:33:24 -0400100 blk = ext4_inode_bitmap(sb, desc);
Kemeng Shib83acc72023-02-22 04:30:24 +0800101 if (!ext4_inode_bitmap_csum_verify(sb, desc, bh,
Theodore Ts'o46f870d62019-11-21 13:09:43 -0500102 EXT4_INODES_PER_GROUP(sb) / 8) ||
103 ext4_simulate_fail(sb, EXT4_SIM_IBITMAP_CRC)) {
Darrick J. Wong9008a582015-10-17 21:33:24 -0400104 ext4_unlock_group(sb, block_group);
105 ext4_error(sb, "Corrupt inode bitmap - block_group = %u, "
106 "inode_bitmap = %llu", block_group, blk);
Wang Shilongdb79e6d2018-05-12 11:39:40 -0400107 ext4_mark_group_bitmap_corrupted(sb, block_group,
108 EXT4_GROUP_INFO_IBITMAP_CORRUPT);
Darrick J. Wong9008a582015-10-17 21:33:24 -0400109 return -EFSBADCRC;
110 }
111 set_buffer_verified(bh);
Theodore Ts'o8d5a8032018-07-12 19:08:05 -0400112verified:
Darrick J. Wong9008a582015-10-17 21:33:24 -0400113 ext4_unlock_group(sb, block_group);
114 return 0;
115}
116
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700117/*
118 * Read the inode allocation bitmap for a given block_group, reading
119 * into the specified slot in the superblock's bitmap cache.
120 *
Josh Triplett90337832020-03-29 13:21:41 -0700121 * Return buffer_head of bitmap on success, or an ERR_PTR on error.
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700122 */
123static struct buffer_head *
Eric Sandeene29d1cd2008-08-02 21:21:02 -0400124ext4_read_inode_bitmap(struct super_block *sb, ext4_group_t block_group)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700125{
Mingming Cao617ba132006-10-11 01:20:53 -0700126 struct ext4_group_desc *desc;
Theodore Ts'o7dac4a12018-03-26 23:54:10 -0400127 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700128 struct buffer_head *bh = NULL;
Eric Sandeene29d1cd2008-08-02 21:21:02 -0400129 ext4_fsblk_t bitmap_blk;
Darrick J. Wong9008a582015-10-17 21:33:24 -0400130 int err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700131
Mingming Cao617ba132006-10-11 01:20:53 -0700132 desc = ext4_get_group_desc(sb, block_group, NULL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700133 if (!desc)
Darrick J. Wong9008a582015-10-17 21:33:24 -0400134 return ERR_PTR(-EFSCORRUPTED);
Lukas Czernerbfff6872010-10-27 21:30:05 -0400135
Eric Sandeene29d1cd2008-08-02 21:21:02 -0400136 bitmap_blk = ext4_inode_bitmap(sb, desc);
Theodore Ts'o7dac4a12018-03-26 23:54:10 -0400137 if ((bitmap_blk <= le32_to_cpu(sbi->s_es->s_first_data_block)) ||
138 (bitmap_blk >= ext4_blocks_count(sbi->s_es))) {
139 ext4_error(sb, "Invalid inode bitmap blk %llu in "
140 "block_group %u", bitmap_blk, block_group);
Wang Shilong206f6d52018-05-12 12:15:21 -0400141 ext4_mark_group_bitmap_corrupted(sb, block_group,
142 EXT4_GROUP_INFO_IBITMAP_CORRUPT);
Theodore Ts'o7dac4a12018-03-26 23:54:10 -0400143 return ERR_PTR(-EFSCORRUPTED);
144 }
Eric Sandeene29d1cd2008-08-02 21:21:02 -0400145 bh = sb_getblk(sb, bitmap_blk);
146 if (unlikely(!bh)) {
Wang Shilong5ef2a692018-08-01 12:02:31 -0400147 ext4_warning(sb, "Cannot read inode bitmap - "
148 "block_group = %u, inode_bitmap = %llu",
149 block_group, bitmap_blk);
Wang Shilong0db9fde2018-05-12 11:35:01 -0400150 return ERR_PTR(-ENOMEM);
Eric Sandeene29d1cd2008-08-02 21:21:02 -0400151 }
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500152 if (bitmap_uptodate(bh))
Darrick J. Wong41a246d2012-04-29 18:33:10 -0400153 goto verify;
Eric Sandeene29d1cd2008-08-02 21:21:02 -0400154
Frederic Bohec806e682008-10-10 08:09:18 -0400155 lock_buffer(bh);
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500156 if (bitmap_uptodate(bh)) {
157 unlock_buffer(bh);
Darrick J. Wong41a246d2012-04-29 18:33:10 -0400158 goto verify;
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500159 }
Lukas Czernerbfff6872010-10-27 21:30:05 -0400160
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -0400161 ext4_lock_group(sb, block_group);
Theodore Ts'o88446182018-06-14 00:58:00 -0400162 if (ext4_has_group_desc_csum(sb) &&
163 (desc->bg_flags & cpu_to_le16(EXT4_BG_INODE_UNINIT))) {
164 if (block_group == 0) {
165 ext4_unlock_group(sb, block_group);
166 unlock_buffer(bh);
167 ext4_error(sb, "Inode bitmap for bg 0 marked "
168 "uninitialized");
169 err = -EFSCORRUPTED;
170 goto out;
171 }
Theodore Ts'o044e6e32018-02-19 14:16:47 -0500172 memset(bh->b_data, 0, (EXT4_INODES_PER_GROUP(sb) + 7) / 8);
173 ext4_mark_bitmap_end(EXT4_INODES_PER_GROUP(sb),
174 sb->s_blocksize * 8, bh->b_data);
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500175 set_bitmap_uptodate(bh);
Eric Sandeene29d1cd2008-08-02 21:21:02 -0400176 set_buffer_uptodate(bh);
Darrick J. Wong41a246d2012-04-29 18:33:10 -0400177 set_buffer_verified(bh);
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -0400178 ext4_unlock_group(sb, block_group);
Aneesh Kumar K.V3300bed2009-01-03 22:33:39 -0500179 unlock_buffer(bh);
Eric Sandeene29d1cd2008-08-02 21:21:02 -0400180 return bh;
181 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -0400182 ext4_unlock_group(sb, block_group);
Lukas Czernerbfff6872010-10-27 21:30:05 -0400183
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500184 if (buffer_uptodate(bh)) {
185 /*
186 * if not uninit if bh is uptodate,
187 * bitmap is also uptodate
188 */
189 set_bitmap_uptodate(bh);
190 unlock_buffer(bh);
Darrick J. Wong41a246d2012-04-29 18:33:10 -0400191 goto verify;
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500192 }
193 /*
Theodore Ts'o813e5722012-02-20 17:52:46 -0500194 * submit the buffer_head for reading
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500195 */
Jiaying Zhang0562e0b2011-03-21 21:38:05 -0400196 trace_ext4_load_inode_bitmap(sb, block_group);
zhangyi (F)2d069c02020-09-24 15:33:33 +0800197 ext4_read_bh(bh, REQ_META | REQ_PRIO, ext4_end_bitmap_read);
Theodore Ts'o46f870d62019-11-21 13:09:43 -0500198 ext4_simulate_fail_bh(sb, bh, EXT4_SIM_IBITMAP_EIO);
Theodore Ts'o813e5722012-02-20 17:52:46 -0500199 if (!buffer_uptodate(bh)) {
Eric Sandeene29d1cd2008-08-02 21:21:02 -0400200 put_bh(bh);
Theodore Ts'o54d3adb2020-03-28 19:33:43 -0400201 ext4_error_err(sb, EIO, "Cannot read inode bitmap - "
202 "block_group = %u, inode_bitmap = %llu",
203 block_group, bitmap_blk);
Wang Shilong206f6d52018-05-12 12:15:21 -0400204 ext4_mark_group_bitmap_corrupted(sb, block_group,
205 EXT4_GROUP_INFO_IBITMAP_CORRUPT);
Darrick J. Wong9008a582015-10-17 21:33:24 -0400206 return ERR_PTR(-EIO);
Eric Sandeene29d1cd2008-08-02 21:21:02 -0400207 }
Darrick J. Wong41a246d2012-04-29 18:33:10 -0400208
209verify:
Darrick J. Wong9008a582015-10-17 21:33:24 -0400210 err = ext4_validate_inode_bitmap(sb, desc, block_group, bh);
211 if (err)
212 goto out;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700213 return bh;
Darrick J. Wong9008a582015-10-17 21:33:24 -0400214out:
215 put_bh(bh);
216 return ERR_PTR(err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700217}
218
219/*
220 * NOTE! When we get the inode, we're the only people
221 * that have access to it, and as such there are no
222 * race conditions we have to worry about. The inode
223 * is not on the hash-lists, and it cannot be reached
224 * through the filesystem because the directory entry
225 * has been deleted earlier.
226 *
227 * HOWEVER: we must make sure that we get no aliases,
228 * which means that we have to call "clear_inode()"
229 * _before_ we mark the inode not in use in the inode
230 * bitmaps. Otherwise a newly created file might use
231 * the same inode number (not actually the same pointer
232 * though), and then we'd have two inodes sharing the
233 * same inode number and space on the harddisk.
234 */
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400235void ext4_free_inode(handle_t *handle, struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700236{
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400237 struct super_block *sb = inode->i_sb;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700238 int is_directory;
239 unsigned long ino;
240 struct buffer_head *bitmap_bh = NULL;
241 struct buffer_head *bh2;
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500242 ext4_group_t block_group;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700243 unsigned long bit;
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400244 struct ext4_group_desc *gdp;
245 struct ext4_super_block *es;
Mingming Cao617ba132006-10-11 01:20:53 -0700246 struct ext4_sb_info *sbi;
Eric Sandeen7ce9d5d2009-03-04 18:38:18 -0500247 int fatal = 0, err, count, cleared;
Darrick J. Wong87a39382013-08-28 18:32:58 -0400248 struct ext4_group_info *grp;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700249
Theodore Ts'o92b978162012-03-19 23:41:49 -0400250 if (!sb) {
251 printk(KERN_ERR "EXT4-fs: %s:%d: inode on "
252 "nonexistent device\n", __func__, __LINE__);
253 return;
254 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700255 if (atomic_read(&inode->i_count) > 1) {
Theodore Ts'o92b978162012-03-19 23:41:49 -0400256 ext4_msg(sb, KERN_ERR, "%s:%d: inode #%lu: count=%d",
257 __func__, __LINE__, inode->i_ino,
258 atomic_read(&inode->i_count));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700259 return;
260 }
261 if (inode->i_nlink) {
Theodore Ts'o92b978162012-03-19 23:41:49 -0400262 ext4_msg(sb, KERN_ERR, "%s:%d: inode #%lu: nlink=%d\n",
263 __func__, __LINE__, inode->i_ino, inode->i_nlink);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700264 return;
265 }
Mingming Cao617ba132006-10-11 01:20:53 -0700266 sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700267
268 ino = inode->i_ino;
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400269 ext4_debug("freeing inode %lu\n", ino);
Theodore Ts'o9bffad12009-06-17 11:48:11 -0400270 trace_ext4_free_inode(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700271
Christoph Hellwig871a2932010-03-03 09:05:07 -0500272 dquot_initialize(inode);
Christoph Hellwig63936dd2010-03-03 09:05:01 -0500273 dquot_free_inode(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700274
275 is_directory = S_ISDIR(inode->i_mode);
276
277 /* Do this BEFORE marking the inode not in use or returning an error */
Al Viro0930fcc2010-06-07 13:16:22 -0400278 ext4_clear_inode(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700279
Jun Piao49598e02018-01-11 13:17:49 -0500280 es = sbi->s_es;
Mingming Cao617ba132006-10-11 01:20:53 -0700281 if (ino < EXT4_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) {
Eric Sandeen12062dd2010-02-15 14:19:27 -0500282 ext4_error(sb, "reserved or nonexistent inode %lu", ino);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700283 goto error_return;
284 }
Mingming Cao617ba132006-10-11 01:20:53 -0700285 block_group = (ino - 1) / EXT4_INODES_PER_GROUP(sb);
286 bit = (ino - 1) % EXT4_INODES_PER_GROUP(sb);
Eric Sandeene29d1cd2008-08-02 21:21:02 -0400287 bitmap_bh = ext4_read_inode_bitmap(sb, block_group);
Darrick J. Wong87a39382013-08-28 18:32:58 -0400288 /* Don't bother if the inode bitmap is corrupt. */
Darrick J. Wong9008a582015-10-17 21:33:24 -0400289 if (IS_ERR(bitmap_bh)) {
290 fatal = PTR_ERR(bitmap_bh);
291 bitmap_bh = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700292 goto error_return;
Darrick J. Wong9008a582015-10-17 21:33:24 -0400293 }
Harshad Shirwadkar8016e292020-10-15 13:37:59 -0700294 if (!(sbi->s_mount_state & EXT4_FC_REPLAY)) {
295 grp = ext4_get_group_info(sb, block_group);
Theodore Ts'o5354b2a2023-04-29 00:06:28 -0400296 if (!grp || unlikely(EXT4_MB_GRP_IBITMAP_CORRUPT(grp))) {
Harshad Shirwadkar8016e292020-10-15 13:37:59 -0700297 fatal = -EFSCORRUPTED;
298 goto error_return;
299 }
Darrick J. Wong9008a582015-10-17 21:33:24 -0400300 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700301
302 BUFFER_TRACE(bitmap_bh, "get_write_access");
Jan Kara188c2992021-08-16 11:57:04 +0200303 fatal = ext4_journal_get_write_access(handle, sb, bitmap_bh,
304 EXT4_JTR_NONE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700305 if (fatal)
306 goto error_return;
307
Dmitry Monakhovd17413c2010-05-16 07:00:00 -0400308 fatal = -ESRCH;
309 gdp = ext4_get_group_desc(sb, block_group, &bh2);
310 if (gdp) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700311 BUFFER_TRACE(bh2, "get_write_access");
Jan Kara188c2992021-08-16 11:57:04 +0200312 fatal = ext4_journal_get_write_access(handle, sb, bh2,
313 EXT4_JTR_NONE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700314 }
Dmitry Monakhovd17413c2010-05-16 07:00:00 -0400315 ext4_lock_group(sb, block_group);
Akinobu Mita597d5082011-12-28 20:32:07 -0500316 cleared = ext4_test_and_clear_bit(bit, bitmap_bh->b_data);
Dmitry Monakhovd17413c2010-05-16 07:00:00 -0400317 if (fatal || !cleared) {
318 ext4_unlock_group(sb, block_group);
319 goto out;
320 }
321
322 count = ext4_free_inodes_count(sb, gdp) + 1;
323 ext4_free_inodes_set(sb, gdp, count);
324 if (is_directory) {
325 count = ext4_used_dirs_count(sb, gdp) - 1;
326 ext4_used_dirs_set(sb, gdp, count);
Ritesh Harjanib45f1892021-04-29 16:13:44 +0530327 if (percpu_counter_initialized(&sbi->s_dirs_counter))
328 percpu_counter_dec(&sbi->s_dirs_counter);
Dmitry Monakhovd17413c2010-05-16 07:00:00 -0400329 }
Kemeng Shi4fd873c2023-02-22 04:30:25 +0800330 ext4_inode_bitmap_csum_set(sb, gdp, bitmap_bh,
Darrick J. Wong41a246d2012-04-29 18:33:10 -0400331 EXT4_INODES_PER_GROUP(sb) / 8);
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -0400332 ext4_group_desc_csum_set(sb, block_group, gdp);
Dmitry Monakhovd17413c2010-05-16 07:00:00 -0400333 ext4_unlock_group(sb, block_group);
334
Ritesh Harjanib45f1892021-04-29 16:13:44 +0530335 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
336 percpu_counter_inc(&sbi->s_freeinodes_counter);
Dmitry Monakhovd17413c2010-05-16 07:00:00 -0400337 if (sbi->s_log_groups_per_flex) {
Suraj Jitindar Singh7c990722020-02-18 19:08:51 -0800338 struct flex_groups *fg;
Dmitry Monakhovd17413c2010-05-16 07:00:00 -0400339
Suraj Jitindar Singh7c990722020-02-18 19:08:51 -0800340 fg = sbi_array_rcu_deref(sbi, s_flex_groups,
341 ext4_flex_group(sbi, block_group));
342 atomic_inc(&fg->free_inodes);
Dmitry Monakhovd17413c2010-05-16 07:00:00 -0400343 if (is_directory)
Suraj Jitindar Singh7c990722020-02-18 19:08:51 -0800344 atomic_dec(&fg->used_dirs);
Dmitry Monakhovd17413c2010-05-16 07:00:00 -0400345 }
346 BUFFER_TRACE(bh2, "call ext4_handle_dirty_metadata");
347 fatal = ext4_handle_dirty_metadata(handle, NULL, bh2);
348out:
349 if (cleared) {
350 BUFFER_TRACE(bitmap_bh, "call ext4_handle_dirty_metadata");
351 err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh);
352 if (!fatal)
353 fatal = err;
Darrick J. Wong87a39382013-08-28 18:32:58 -0400354 } else {
Dmitry Monakhovd17413c2010-05-16 07:00:00 -0400355 ext4_error(sb, "bit already cleared for inode %lu", ino);
Wang Shilongdb79e6d2018-05-12 11:39:40 -0400356 ext4_mark_group_bitmap_corrupted(sb, block_group,
357 EXT4_GROUP_INFO_IBITMAP_CORRUPT);
Darrick J. Wong87a39382013-08-28 18:32:58 -0400358 }
Dmitry Monakhovd17413c2010-05-16 07:00:00 -0400359
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700360error_return:
361 brelse(bitmap_bh);
Mingming Cao617ba132006-10-11 01:20:53 -0700362 ext4_std_error(sb, fatal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700363}
364
Theodore Ts'oa4912122009-03-12 12:18:34 -0400365struct orlov_stats {
Theodore Ts'o90ba9832013-03-11 23:39:59 -0400366 __u64 free_clusters;
Theodore Ts'oa4912122009-03-12 12:18:34 -0400367 __u32 free_inodes;
Theodore Ts'oa4912122009-03-12 12:18:34 -0400368 __u32 used_dirs;
369};
370
371/*
372 * Helper function for Orlov's allocator; returns critical information
373 * for a particular block group or flex_bg. If flex_size is 1, then g
374 * is a block group number; otherwise it is flex_bg number.
375 */
Theodore Ts'o1f109d52010-10-27 21:30:14 -0400376static void get_orlov_stats(struct super_block *sb, ext4_group_t g,
377 int flex_size, struct orlov_stats *stats)
Theodore Ts'oa4912122009-03-12 12:18:34 -0400378{
379 struct ext4_group_desc *desc;
Theodore Ts'oa4912122009-03-12 12:18:34 -0400380
Theodore Ts'o7d39db12009-03-04 19:31:53 -0500381 if (flex_size > 1) {
Suraj Jitindar Singh7c990722020-02-18 19:08:51 -0800382 struct flex_groups *fg = sbi_array_rcu_deref(EXT4_SB(sb),
383 s_flex_groups, g);
384 stats->free_inodes = atomic_read(&fg->free_inodes);
385 stats->free_clusters = atomic64_read(&fg->free_clusters);
386 stats->used_dirs = atomic_read(&fg->used_dirs);
Theodore Ts'o7d39db12009-03-04 19:31:53 -0500387 return;
388 }
Theodore Ts'oa4912122009-03-12 12:18:34 -0400389
Theodore Ts'o7d39db12009-03-04 19:31:53 -0500390 desc = ext4_get_group_desc(sb, g, NULL);
391 if (desc) {
392 stats->free_inodes = ext4_free_inodes_count(sb, desc);
Theodore Ts'o021b65b2011-09-09 19:08:51 -0400393 stats->free_clusters = ext4_free_group_clusters(sb, desc);
Theodore Ts'o7d39db12009-03-04 19:31:53 -0500394 stats->used_dirs = ext4_used_dirs_count(sb, desc);
395 } else {
396 stats->free_inodes = 0;
Theodore Ts'o24aaa8e2011-09-09 18:58:51 -0400397 stats->free_clusters = 0;
Theodore Ts'o7d39db12009-03-04 19:31:53 -0500398 stats->used_dirs = 0;
Theodore Ts'oa4912122009-03-12 12:18:34 -0400399 }
400}
401
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700402/*
403 * Orlov's allocator for directories.
404 *
405 * We always try to spread first-level directories.
406 *
Pan Dongc89849c2021-05-25 15:36:56 +0800407 * If there are blockgroups with both free inodes and free clusters counts
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700408 * not worse than average we return one with smallest directory count.
409 * Otherwise we simply return a random group.
410 *
411 * For the rest rules look so:
412 *
413 * It's OK to put directory into a group unless
414 * it has too many directories already (max_dirs) or
415 * it has too few free inodes left (min_inodes) or
Pan Dongc89849c2021-05-25 15:36:56 +0800416 * it has too few free clusters left (min_clusters) or
Benoit Boissinot1cc8dcf2008-04-21 22:45:55 +0000417 * Parent's group is preferred, if it doesn't satisfy these
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700418 * conditions we search cyclically through the rest. If none
419 * of the groups look good we just look for a group with more
420 * free inodes than average (starting at parent's group).
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700421 */
422
Avantika Mathur2aa9fc42008-01-28 23:58:27 -0500423static int find_group_orlov(struct super_block *sb, struct inode *parent,
Al Virodcca3fe2011-07-26 02:48:06 -0400424 ext4_group_t *group, umode_t mode,
Theodore Ts'of157a4a2009-06-13 11:09:42 -0400425 const struct qstr *qstr)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700426{
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500427 ext4_group_t parent_group = EXT4_I(parent)->i_block_group;
Mingming Cao617ba132006-10-11 01:20:53 -0700428 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'o8df96752009-05-01 08:50:38 -0400429 ext4_group_t real_ngroups = ext4_get_groups_count(sb);
Mingming Cao617ba132006-10-11 01:20:53 -0700430 int inodes_per_group = EXT4_INODES_PER_GROUP(sb);
Theodore Ts'o14c83c92011-12-28 20:25:13 -0500431 unsigned int freei, avefreei, grp_free;
Pan Dongc89849c2021-05-25 15:36:56 +0800432 ext4_fsblk_t freec, avefreec;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700433 unsigned int ndirs;
Theodore Ts'oa4912122009-03-12 12:18:34 -0400434 int max_dirs, min_inodes;
Theodore Ts'o24aaa8e2011-09-09 18:58:51 -0400435 ext4_grpblk_t min_clusters;
Theodore Ts'o8df96752009-05-01 08:50:38 -0400436 ext4_group_t i, grp, g, ngroups;
Mingming Cao617ba132006-10-11 01:20:53 -0700437 struct ext4_group_desc *desc;
Theodore Ts'oa4912122009-03-12 12:18:34 -0400438 struct orlov_stats stats;
439 int flex_size = ext4_flex_bg_size(sbi);
Theodore Ts'of157a4a2009-06-13 11:09:42 -0400440 struct dx_hash_info hinfo;
Theodore Ts'oa4912122009-03-12 12:18:34 -0400441
Theodore Ts'o8df96752009-05-01 08:50:38 -0400442 ngroups = real_ngroups;
Theodore Ts'oa4912122009-03-12 12:18:34 -0400443 if (flex_size > 1) {
Theodore Ts'o8df96752009-05-01 08:50:38 -0400444 ngroups = (real_ngroups + flex_size - 1) >>
Theodore Ts'oa4912122009-03-12 12:18:34 -0400445 sbi->s_log_groups_per_flex;
446 parent_group >>= sbi->s_log_groups_per_flex;
447 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700448
449 freei = percpu_counter_read_positive(&sbi->s_freeinodes_counter);
450 avefreei = freei / ngroups;
Pan Dongc89849c2021-05-25 15:36:56 +0800451 freec = percpu_counter_read_positive(&sbi->s_freeclusters_counter);
452 avefreec = freec;
Theodore Ts'o24aaa8e2011-09-09 18:58:51 -0400453 do_div(avefreec, ngroups);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700454 ndirs = percpu_counter_read_positive(&sbi->s_dirs_counter);
455
Theodore Ts'oa4912122009-03-12 12:18:34 -0400456 if (S_ISDIR(mode) &&
David Howells2b0143b2015-03-17 22:25:59 +0000457 ((parent == d_inode(sb->s_root)) ||
Dmitry Monakhov12e9b892010-05-16 22:00:00 -0400458 (ext4_test_inode_flag(parent, EXT4_INODE_TOPDIR)))) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700459 int best_ndir = inodes_per_group;
Avantika Mathur2aa9fc42008-01-28 23:58:27 -0500460 int ret = -1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700461
Theodore Ts'of157a4a2009-06-13 11:09:42 -0400462 if (qstr) {
463 hinfo.hash_version = DX_HASH_HALF_MD4;
464 hinfo.seed = sbi->s_hash_seed;
Gabriel Krisman Bertazib886ee32019-04-25 14:12:08 -0400465 ext4fs_dirhash(parent, qstr->name, qstr->len, &hinfo);
Jason A. Donenfeld8b3ccbc2022-10-05 16:43:38 +0200466 parent_group = hinfo.hash % ngroups;
Theodore Ts'of157a4a2009-06-13 11:09:42 -0400467 } else
Jason A. Donenfeld8032bf12022-10-09 20:44:02 -0600468 parent_group = get_random_u32_below(ngroups);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700469 for (i = 0; i < ngroups; i++) {
Theodore Ts'oa4912122009-03-12 12:18:34 -0400470 g = (parent_group + i) % ngroups;
471 get_orlov_stats(sb, g, flex_size, &stats);
472 if (!stats.free_inodes)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700473 continue;
Theodore Ts'oa4912122009-03-12 12:18:34 -0400474 if (stats.used_dirs >= best_ndir)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700475 continue;
Theodore Ts'oa4912122009-03-12 12:18:34 -0400476 if (stats.free_inodes < avefreei)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700477 continue;
Theodore Ts'o24aaa8e2011-09-09 18:58:51 -0400478 if (stats.free_clusters < avefreec)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700479 continue;
Theodore Ts'oa4912122009-03-12 12:18:34 -0400480 grp = g;
Avantika Mathur2aa9fc42008-01-28 23:58:27 -0500481 ret = 0;
Theodore Ts'oa4912122009-03-12 12:18:34 -0400482 best_ndir = stats.used_dirs;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700483 }
Theodore Ts'oa4912122009-03-12 12:18:34 -0400484 if (ret)
485 goto fallback;
486 found_flex_bg:
487 if (flex_size == 1) {
488 *group = grp;
489 return 0;
490 }
491
492 /*
493 * We pack inodes at the beginning of the flexgroup's
494 * inode tables. Block allocation decisions will do
495 * something similar, although regular files will
496 * start at 2nd block group of the flexgroup. See
497 * ext4_ext_find_goal() and ext4_find_near().
498 */
499 grp *= flex_size;
500 for (i = 0; i < flex_size; i++) {
Theodore Ts'o8df96752009-05-01 08:50:38 -0400501 if (grp+i >= real_ngroups)
Theodore Ts'oa4912122009-03-12 12:18:34 -0400502 break;
503 desc = ext4_get_group_desc(sb, grp+i, NULL);
504 if (desc && ext4_free_inodes_count(sb, desc)) {
505 *group = grp+i;
506 return 0;
507 }
508 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700509 goto fallback;
510 }
511
Jan Kara613c5a82022-09-08 11:21:26 +0200512 max_dirs = ndirs / ngroups + inodes_per_group*flex_size / 16;
Theodore Ts'oa4912122009-03-12 12:18:34 -0400513 min_inodes = avefreei - inodes_per_group*flex_size / 4;
514 if (min_inodes < 1)
515 min_inodes = 1;
Theodore Ts'o24aaa8e2011-09-09 18:58:51 -0400516 min_clusters = avefreec - EXT4_CLUSTERS_PER_GROUP(sb)*flex_size / 4;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700517
Theodore Ts'oa4912122009-03-12 12:18:34 -0400518 /*
519 * Start looking in the flex group where we last allocated an
520 * inode for this parent directory
521 */
522 if (EXT4_I(parent)->i_last_alloc_group != ~0) {
523 parent_group = EXT4_I(parent)->i_last_alloc_group;
524 if (flex_size > 1)
525 parent_group >>= sbi->s_log_groups_per_flex;
526 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700527
528 for (i = 0; i < ngroups; i++) {
Theodore Ts'oa4912122009-03-12 12:18:34 -0400529 grp = (parent_group + i) % ngroups;
530 get_orlov_stats(sb, grp, flex_size, &stats);
531 if (stats.used_dirs >= max_dirs)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700532 continue;
Theodore Ts'oa4912122009-03-12 12:18:34 -0400533 if (stats.free_inodes < min_inodes)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700534 continue;
Theodore Ts'o24aaa8e2011-09-09 18:58:51 -0400535 if (stats.free_clusters < min_clusters)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700536 continue;
Theodore Ts'oa4912122009-03-12 12:18:34 -0400537 goto found_flex_bg;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700538 }
539
540fallback:
Theodore Ts'o8df96752009-05-01 08:50:38 -0400541 ngroups = real_ngroups;
Theodore Ts'oa4912122009-03-12 12:18:34 -0400542 avefreei = freei / ngroups;
Theodore Ts'ob5451f72009-04-22 21:00:36 -0400543fallback_retry:
Theodore Ts'oa4912122009-03-12 12:18:34 -0400544 parent_group = EXT4_I(parent)->i_block_group;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700545 for (i = 0; i < ngroups; i++) {
Theodore Ts'oa4912122009-03-12 12:18:34 -0400546 grp = (parent_group + i) % ngroups;
547 desc = ext4_get_group_desc(sb, grp, NULL);
Dan Carpenterbb3d1322012-05-28 14:16:57 -0400548 if (desc) {
549 grp_free = ext4_free_inodes_count(sb, desc);
550 if (grp_free && grp_free >= avefreei) {
551 *group = grp;
552 return 0;
553 }
Theodore Ts'oa4912122009-03-12 12:18:34 -0400554 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700555 }
556
557 if (avefreei) {
558 /*
559 * The free-inodes counter is approximate, and for really small
560 * filesystems the above test can fail to find any blockgroups
561 */
562 avefreei = 0;
Theodore Ts'ob5451f72009-04-22 21:00:36 -0400563 goto fallback_retry;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700564 }
565
566 return -1;
567}
568
Avantika Mathur2aa9fc42008-01-28 23:58:27 -0500569static int find_group_other(struct super_block *sb, struct inode *parent,
Al Virodcca3fe2011-07-26 02:48:06 -0400570 ext4_group_t *group, umode_t mode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700571{
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500572 ext4_group_t parent_group = EXT4_I(parent)->i_block_group;
Theodore Ts'o8df96752009-05-01 08:50:38 -0400573 ext4_group_t i, last, ngroups = ext4_get_groups_count(sb);
Mingming Cao617ba132006-10-11 01:20:53 -0700574 struct ext4_group_desc *desc;
Theodore Ts'oa4912122009-03-12 12:18:34 -0400575 int flex_size = ext4_flex_bg_size(EXT4_SB(sb));
576
577 /*
578 * Try to place the inode is the same flex group as its
579 * parent. If we can't find space, use the Orlov algorithm to
580 * find another flex group, and store that information in the
581 * parent directory's inode information so that use that flex
582 * group for future allocations.
583 */
584 if (flex_size > 1) {
585 int retry = 0;
586
587 try_again:
588 parent_group &= ~(flex_size-1);
589 last = parent_group + flex_size;
590 if (last > ngroups)
591 last = ngroups;
592 for (i = parent_group; i < last; i++) {
593 desc = ext4_get_group_desc(sb, i, NULL);
594 if (desc && ext4_free_inodes_count(sb, desc)) {
595 *group = i;
596 return 0;
597 }
598 }
599 if (!retry && EXT4_I(parent)->i_last_alloc_group != ~0) {
600 retry = 1;
601 parent_group = EXT4_I(parent)->i_last_alloc_group;
602 goto try_again;
603 }
604 /*
605 * If this didn't work, use the Orlov search algorithm
606 * to find a new flex group; we pass in the mode to
607 * avoid the topdir algorithms.
608 */
609 *group = parent_group + flex_size;
610 if (*group > ngroups)
611 *group = 0;
Peter Huewe7dc57612011-02-21 21:01:42 -0500612 return find_group_orlov(sb, parent, group, mode, NULL);
Theodore Ts'oa4912122009-03-12 12:18:34 -0400613 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700614
615 /*
616 * Try to place the inode in its parent directory
617 */
Avantika Mathur2aa9fc42008-01-28 23:58:27 -0500618 *group = parent_group;
619 desc = ext4_get_group_desc(sb, *group, NULL);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500620 if (desc && ext4_free_inodes_count(sb, desc) &&
Theodore Ts'o021b65b2011-09-09 19:08:51 -0400621 ext4_free_group_clusters(sb, desc))
Avantika Mathur2aa9fc42008-01-28 23:58:27 -0500622 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700623
624 /*
625 * We're going to place this inode in a different blockgroup from its
626 * parent. We want to cause files in a common directory to all land in
627 * the same blockgroup. But we want files which are in a different
628 * directory which shares a blockgroup with our parent to land in a
629 * different blockgroup.
630 *
631 * So add our directory's i_ino into the starting point for the hash.
632 */
Avantika Mathur2aa9fc42008-01-28 23:58:27 -0500633 *group = (*group + parent->i_ino) % ngroups;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700634
635 /*
636 * Use a quadratic hash to find a group with a free inode and some free
637 * blocks.
638 */
639 for (i = 1; i < ngroups; i <<= 1) {
Avantika Mathur2aa9fc42008-01-28 23:58:27 -0500640 *group += i;
641 if (*group >= ngroups)
642 *group -= ngroups;
643 desc = ext4_get_group_desc(sb, *group, NULL);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500644 if (desc && ext4_free_inodes_count(sb, desc) &&
Theodore Ts'o021b65b2011-09-09 19:08:51 -0400645 ext4_free_group_clusters(sb, desc))
Avantika Mathur2aa9fc42008-01-28 23:58:27 -0500646 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700647 }
648
649 /*
650 * That failed: try linear search for a free inode, even if that group
651 * has no free blocks.
652 */
Avantika Mathur2aa9fc42008-01-28 23:58:27 -0500653 *group = parent_group;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700654 for (i = 0; i < ngroups; i++) {
Avantika Mathur2aa9fc42008-01-28 23:58:27 -0500655 if (++*group >= ngroups)
656 *group = 0;
657 desc = ext4_get_group_desc(sb, *group, NULL);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500658 if (desc && ext4_free_inodes_count(sb, desc))
Avantika Mathur2aa9fc42008-01-28 23:58:27 -0500659 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700660 }
661
662 return -1;
663}
664
665/*
Theodore Ts'o19883bd2013-08-16 22:06:55 -0400666 * In no journal mode, if an inode has recently been deleted, we want
667 * to avoid reusing it until we're reasonably sure the inode table
668 * block has been written back to disk. (Yes, these values are
669 * somewhat arbitrary...)
670 */
Theodore Ts'oa17a9d92020-04-13 22:30:52 -0400671#define RECENTCY_MIN 60
Andreas Dilgerb5f51572017-08-31 11:09:45 -0400672#define RECENTCY_DIRTY 300
Theodore Ts'o19883bd2013-08-16 22:06:55 -0400673
674static int recently_deleted(struct super_block *sb, ext4_group_t group, int ino)
675{
676 struct ext4_group_desc *gdp;
677 struct ext4_inode *raw_inode;
678 struct buffer_head *bh;
Andreas Dilgerb5f51572017-08-31 11:09:45 -0400679 int inodes_per_block = EXT4_SB(sb)->s_inodes_per_block;
680 int offset, ret = 0;
681 int recentcy = RECENTCY_MIN;
682 u32 dtime, now;
Theodore Ts'o19883bd2013-08-16 22:06:55 -0400683
684 gdp = ext4_get_group_desc(sb, group, NULL);
685 if (unlikely(!gdp))
686 return 0;
687
Jan Kara4f9d9562017-08-24 11:52:21 -0400688 bh = sb_find_get_block(sb, ext4_inode_table(sb, gdp) +
Theodore Ts'o19883bd2013-08-16 22:06:55 -0400689 (ino / inodes_per_block));
Jan Kara4f9d9562017-08-24 11:52:21 -0400690 if (!bh || !buffer_uptodate(bh))
Theodore Ts'o19883bd2013-08-16 22:06:55 -0400691 /*
692 * If the block is not in the buffer cache, then it
693 * must have been written out.
694 */
695 goto out;
696
697 offset = (ino % inodes_per_block) * EXT4_INODE_SIZE(sb);
698 raw_inode = (struct ext4_inode *) (bh->b_data + offset);
Andreas Dilgerb5f51572017-08-31 11:09:45 -0400699
700 /* i_dtime is only 32 bits on disk, but we only care about relative
701 * times in the range of a few minutes (i.e. long enough to sync a
702 * recently-deleted inode to disk), so using the low 32 bits of the
703 * clock (a 68 year range) is enough, see time_before32() */
Theodore Ts'o19883bd2013-08-16 22:06:55 -0400704 dtime = le32_to_cpu(raw_inode->i_dtime);
Andreas Dilgerb5f51572017-08-31 11:09:45 -0400705 now = ktime_get_real_seconds();
Theodore Ts'o19883bd2013-08-16 22:06:55 -0400706 if (buffer_dirty(bh))
707 recentcy += RECENTCY_DIRTY;
708
Andreas Dilgerb5f51572017-08-31 11:09:45 -0400709 if (dtime && time_before32(dtime, now) &&
710 time_before32(now, dtime + recentcy))
Theodore Ts'o19883bd2013-08-16 22:06:55 -0400711 ret = 1;
712out:
713 brelse(bh);
714 return ret;
715}
716
Wang Shilong901ed072017-08-24 12:56:35 -0400717static int find_inode_bit(struct super_block *sb, ext4_group_t group,
718 struct buffer_head *bitmap, unsigned long *ino)
719{
Jan Karad05466b2020-03-18 13:13:17 +0100720 bool check_recently_deleted = EXT4_SB(sb)->s_journal == NULL;
721 unsigned long recently_deleted_ino = EXT4_INODES_PER_GROUP(sb);
722
Wang Shilong901ed072017-08-24 12:56:35 -0400723next:
724 *ino = ext4_find_next_zero_bit((unsigned long *)
725 bitmap->b_data,
726 EXT4_INODES_PER_GROUP(sb), *ino);
727 if (*ino >= EXT4_INODES_PER_GROUP(sb))
Jan Karad05466b2020-03-18 13:13:17 +0100728 goto not_found;
Wang Shilong901ed072017-08-24 12:56:35 -0400729
Jan Karad05466b2020-03-18 13:13:17 +0100730 if (check_recently_deleted && recently_deleted(sb, group, *ino)) {
731 recently_deleted_ino = *ino;
Wang Shilong901ed072017-08-24 12:56:35 -0400732 *ino = *ino + 1;
733 if (*ino < EXT4_INODES_PER_GROUP(sb))
734 goto next;
Jan Karad05466b2020-03-18 13:13:17 +0100735 goto not_found;
Wang Shilong901ed072017-08-24 12:56:35 -0400736 }
Jan Karad05466b2020-03-18 13:13:17 +0100737 return 1;
738not_found:
739 if (recently_deleted_ino >= EXT4_INODES_PER_GROUP(sb))
740 return 0;
741 /*
742 * Not reusing recently deleted inodes is mostly a preference. We don't
743 * want to report ENOSPC or skew allocation patterns because of that.
744 * So return even recently deleted inode if we could find better in the
745 * given range.
746 */
747 *ino = recently_deleted_ino;
Wang Shilong901ed072017-08-24 12:56:35 -0400748 return 1;
749}
750
Harshad Shirwadkar8016e292020-10-15 13:37:59 -0700751int ext4_mark_inode_used(struct super_block *sb, int ino)
752{
753 unsigned long max_ino = le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count);
754 struct buffer_head *inode_bitmap_bh = NULL, *group_desc_bh = NULL;
755 struct ext4_group_desc *gdp;
756 ext4_group_t group;
757 int bit;
758 int err = -EFSCORRUPTED;
759
760 if (ino < EXT4_FIRST_INO(sb) || ino > max_ino)
761 goto out;
762
763 group = (ino - 1) / EXT4_INODES_PER_GROUP(sb);
764 bit = (ino - 1) % EXT4_INODES_PER_GROUP(sb);
765 inode_bitmap_bh = ext4_read_inode_bitmap(sb, group);
766 if (IS_ERR(inode_bitmap_bh))
767 return PTR_ERR(inode_bitmap_bh);
768
769 if (ext4_test_bit(bit, inode_bitmap_bh->b_data)) {
770 err = 0;
771 goto out;
772 }
773
774 gdp = ext4_get_group_desc(sb, group, &group_desc_bh);
775 if (!gdp || !group_desc_bh) {
776 err = -EINVAL;
777 goto out;
778 }
779
780 ext4_set_bit(bit, inode_bitmap_bh->b_data);
781
782 BUFFER_TRACE(inode_bitmap_bh, "call ext4_handle_dirty_metadata");
783 err = ext4_handle_dirty_metadata(NULL, NULL, inode_bitmap_bh);
784 if (err) {
785 ext4_std_error(sb, err);
786 goto out;
787 }
788 err = sync_dirty_buffer(inode_bitmap_bh);
789 if (err) {
790 ext4_std_error(sb, err);
791 goto out;
792 }
793
794 /* We may have to initialize the block bitmap if it isn't already */
795 if (ext4_has_group_desc_csum(sb) &&
796 gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
797 struct buffer_head *block_bitmap_bh;
798
799 block_bitmap_bh = ext4_read_block_bitmap(sb, group);
800 if (IS_ERR(block_bitmap_bh)) {
801 err = PTR_ERR(block_bitmap_bh);
802 goto out;
803 }
804
805 BUFFER_TRACE(block_bitmap_bh, "dirty block bitmap");
806 err = ext4_handle_dirty_metadata(NULL, NULL, block_bitmap_bh);
807 sync_dirty_buffer(block_bitmap_bh);
808
809 /* recheck and clear flag under lock if we still need to */
810 ext4_lock_group(sb, group);
811 if (ext4_has_group_desc_csum(sb) &&
812 (gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT))) {
813 gdp->bg_flags &= cpu_to_le16(~EXT4_BG_BLOCK_UNINIT);
814 ext4_free_group_clusters_set(sb, gdp,
815 ext4_free_clusters_after_init(sb, group, gdp));
Kemeng Shi1df9bde2023-02-22 04:30:27 +0800816 ext4_block_bitmap_csum_set(sb, gdp, block_bitmap_bh);
Harshad Shirwadkar8016e292020-10-15 13:37:59 -0700817 ext4_group_desc_csum_set(sb, group, gdp);
818 }
819 ext4_unlock_group(sb, group);
820 brelse(block_bitmap_bh);
821
822 if (err) {
823 ext4_std_error(sb, err);
824 goto out;
825 }
826 }
827
828 /* Update the relevant bg descriptor fields */
829 if (ext4_has_group_desc_csum(sb)) {
830 int free;
831
832 ext4_lock_group(sb, group); /* while we modify the bg desc */
833 free = EXT4_INODES_PER_GROUP(sb) -
834 ext4_itable_unused_count(sb, gdp);
835 if (gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_UNINIT)) {
836 gdp->bg_flags &= cpu_to_le16(~EXT4_BG_INODE_UNINIT);
837 free = 0;
838 }
839
840 /*
841 * Check the relative inode number against the last used
842 * relative inode number in this group. if it is greater
843 * we need to update the bg_itable_unused count
844 */
845 if (bit >= free)
846 ext4_itable_unused_set(sb, gdp,
847 (EXT4_INODES_PER_GROUP(sb) - bit - 1));
848 } else {
849 ext4_lock_group(sb, group);
850 }
851
852 ext4_free_inodes_set(sb, gdp, ext4_free_inodes_count(sb, gdp) - 1);
853 if (ext4_has_group_desc_csum(sb)) {
Kemeng Shi4fd873c2023-02-22 04:30:25 +0800854 ext4_inode_bitmap_csum_set(sb, gdp, inode_bitmap_bh,
Harshad Shirwadkar8016e292020-10-15 13:37:59 -0700855 EXT4_INODES_PER_GROUP(sb) / 8);
856 ext4_group_desc_csum_set(sb, group, gdp);
857 }
858
859 ext4_unlock_group(sb, group);
860 err = ext4_handle_dirty_metadata(NULL, NULL, group_desc_bh);
861 sync_dirty_buffer(group_desc_bh);
862out:
863 return err;
864}
865
Eric Biggers177cc0e2020-09-16 21:11:25 -0700866static int ext4_xattr_credits_for_new_inode(struct inode *dir, mode_t mode,
867 bool encrypt)
868{
869 struct super_block *sb = dir->i_sb;
870 int nblocks = 0;
871#ifdef CONFIG_EXT4_FS_POSIX_ACL
Christian Braunercac2f8b2022-09-22 17:17:00 +0200872 struct posix_acl *p = get_inode_acl(dir, ACL_TYPE_DEFAULT);
Eric Biggers177cc0e2020-09-16 21:11:25 -0700873
874 if (IS_ERR(p))
875 return PTR_ERR(p);
876 if (p) {
877 int acl_size = p->a_count * sizeof(ext4_acl_entry);
878
879 nblocks += (S_ISDIR(mode) ? 2 : 1) *
880 __ext4_xattr_set_credits(sb, NULL /* inode */,
881 NULL /* block_bh */, acl_size,
882 true /* is_create */);
883 posix_acl_release(p);
884 }
885#endif
886
887#ifdef CONFIG_SECURITY
888 {
889 int num_security_xattrs = 1;
890
891#ifdef CONFIG_INTEGRITY
892 num_security_xattrs++;
893#endif
894 /*
895 * We assume that security xattrs are never more than 1k.
896 * In practice they are under 128 bytes.
897 */
898 nblocks += num_security_xattrs *
899 __ext4_xattr_set_credits(sb, NULL /* inode */,
900 NULL /* block_bh */, 1024,
901 true /* is_create */);
902 }
903#endif
904 if (encrypt)
905 nblocks += __ext4_xattr_set_credits(sb,
906 NULL /* inode */,
907 NULL /* block_bh */,
908 FSCRYPT_SET_CONTEXT_MAX_SIZE,
909 true /* is_create */);
910 return nblocks;
911}
912
Theodore Ts'o19883bd2013-08-16 22:06:55 -0400913/*
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700914 * There are two policies for allocating an inode. If the new inode is
915 * a directory, then a forward search is made for a block group with both
916 * free space and a low directory-to-inode ratio; if that fails, then of
917 * the groups with above-average free space, that group with the fewest
918 * directories already is chosen.
919 *
920 * For other inodes, search forward from the parent directory's block
921 * group to find a free inode.
922 */
Christian Braunerf2d40142023-01-13 12:49:25 +0100923struct inode *__ext4_new_inode(struct mnt_idmap *idmap,
Christian Brauner14f3db52021-01-21 14:19:57 +0100924 handle_t *handle, struct inode *dir,
Theodore Ts'o11395752013-02-09 16:27:09 -0500925 umode_t mode, const struct qstr *qstr,
Tahsin Erdogan1b917ed2017-06-21 21:21:39 -0400926 __u32 goal, uid_t *owner, __u32 i_flags,
927 int handle_type, unsigned int line_no,
928 int nblocks)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700929{
930 struct super_block *sb;
Aneesh Kumar K.V3300bed2009-01-03 22:33:39 -0500931 struct buffer_head *inode_bitmap_bh = NULL;
932 struct buffer_head *group_desc_bh;
Theodore Ts'o8df96752009-05-01 08:50:38 -0400933 ext4_group_t ngroups, group = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700934 unsigned long ino = 0;
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400935 struct inode *inode;
936 struct ext4_group_desc *gdp = NULL;
Mingming Cao617ba132006-10-11 01:20:53 -0700937 struct ext4_inode_info *ei;
938 struct ext4_sb_info *sbi;
Jan Karaa7cdade2015-06-29 16:22:54 +0200939 int ret2, err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700940 struct inode *ret;
Avantika Mathur2aa9fc42008-01-28 23:58:27 -0500941 ext4_group_t i;
Jose R. Santos772cb7c2008-07-11 19:27:31 -0400942 ext4_group_t flex_group;
Harshad Shirwadkar8016e292020-10-15 13:37:59 -0700943 struct ext4_group_info *grp = NULL;
Eric Biggers02ce5312020-09-16 21:11:26 -0700944 bool encrypt = false;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700945
946 /* Cannot create files in a deleted directory */
947 if (!dir || !dir->i_nlink)
948 return ERR_PTR(-EPERM);
949
Tahsin Erdoganaf652072017-07-06 00:01:59 -0400950 sb = dir->i_sb;
951 sbi = EXT4_SB(sb);
952
Jan Karaeb8ab442023-06-16 18:50:49 +0200953 if (unlikely(ext4_forced_shutdown(sb)))
Theodore Ts'o0db1ff22017-02-05 01:28:48 -0500954 return ERR_PTR(-EIO);
955
Theodore Ts'o8df96752009-05-01 08:50:38 -0400956 ngroups = ext4_get_groups_count(sb);
Theodore Ts'o9bffad12009-06-17 11:48:11 -0400957 trace_ext4_request_inode(dir, mode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700958 inode = new_inode(sb);
959 if (!inode)
960 return ERR_PTR(-ENOMEM);
Mingming Cao617ba132006-10-11 01:20:53 -0700961 ei = EXT4_I(inode);
Jose R. Santos772cb7c2008-07-11 19:27:31 -0400962
Jan Karaeb9cc7e2013-04-19 13:38:14 -0400963 /*
Adam Buchbinderb8a074632016-03-09 23:49:05 -0500964 * Initialize owners and quota early so that we don't have to account
Jan Karaeb9cc7e2013-04-19 13:38:14 -0400965 * for quota initialization worst case in standard inode creating
966 * transaction
967 */
968 if (owner) {
969 inode->i_mode = mode;
970 i_uid_write(inode, owner[0]);
971 i_gid_write(inode, owner[1]);
972 } else if (test_opt(sb, GRPID)) {
973 inode->i_mode = mode;
Christian Braunerc14329d2023-01-13 12:49:31 +0100974 inode_fsuid_set(inode, idmap);
Jan Karaeb9cc7e2013-04-19 13:38:14 -0400975 inode->i_gid = dir->i_gid;
976 } else
Christian Braunerf2d40142023-01-13 12:49:25 +0100977 inode_init_owner(idmap, inode, dir, mode);
Li Xi040cb372016-01-08 16:01:21 -0500978
Kaho Ng0b7b7772016-09-05 23:11:58 -0400979 if (ext4_has_feature_project(sb) &&
Li Xi040cb372016-01-08 16:01:21 -0500980 ext4_test_inode_flag(dir, EXT4_INODE_PROJINHERIT))
981 ei->i_projid = EXT4_I(dir)->i_projid;
982 else
983 ei->i_projid = make_kprojid(&init_user_ns, EXT4_DEF_PROJID);
984
Eric Biggers02ce5312020-09-16 21:11:26 -0700985 if (!(i_flags & EXT4_EA_INODE_FL)) {
986 err = fscrypt_prepare_new_inode(dir, inode, &encrypt);
987 if (err)
988 goto out;
989 }
990
Jan Karaa7cdade2015-06-29 16:22:54 +0200991 err = dquot_initialize(inode);
992 if (err)
993 goto out;
Jan Karaeb9cc7e2013-04-19 13:38:14 -0400994
Eric Biggers02ce5312020-09-16 21:11:26 -0700995 if (!handle && sbi->s_journal && !(i_flags & EXT4_EA_INODE_FL)) {
996 ret2 = ext4_xattr_credits_for_new_inode(dir, mode, encrypt);
997 if (ret2 < 0) {
998 err = ret2;
999 goto out;
1000 }
1001 nblocks += ret2;
1002 }
1003
Andreas Dilger11013912009-06-13 11:45:35 -04001004 if (!goal)
1005 goal = sbi->s_inode_goal;
1006
Johann Lombardie6462862009-07-05 23:45:11 -04001007 if (goal && goal <= le32_to_cpu(sbi->s_es->s_inodes_count)) {
Andreas Dilger11013912009-06-13 11:45:35 -04001008 group = (goal - 1) / EXT4_INODES_PER_GROUP(sb);
1009 ino = (goal - 1) % EXT4_INODES_PER_GROUP(sb);
1010 ret2 = 0;
1011 goto got_group;
1012 }
1013
Lukas Czerner4113c4c2011-10-08 14:34:47 -04001014 if (S_ISDIR(mode))
1015 ret2 = find_group_orlov(sb, dir, &group, mode, qstr);
1016 else
Theodore Ts'oa4912122009-03-12 12:18:34 -04001017 ret2 = find_group_other(sb, dir, &group, mode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001018
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001019got_group:
Theodore Ts'oa4912122009-03-12 12:18:34 -04001020 EXT4_I(dir)->i_last_alloc_group = group;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001021 err = -ENOSPC;
Avantika Mathur2aa9fc42008-01-28 23:58:27 -05001022 if (ret2 == -1)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001023 goto out;
1024
Theodore Ts'o119c0d42012-02-06 20:12:03 -05001025 /*
1026 * Normally we will only go through one pass of this loop,
1027 * unless we get unlucky and it turns out the group we selected
1028 * had its last inode grabbed by someone else.
1029 */
Andreas Dilger11013912009-06-13 11:45:35 -04001030 for (i = 0; i < ngroups; i++, ino = 0) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001031 err = -EIO;
1032
Aneesh Kumar K.V3300bed2009-01-03 22:33:39 -05001033 gdp = ext4_get_group_desc(sb, group, &group_desc_bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001034 if (!gdp)
Jan Karaeb9cc7e2013-04-19 13:38:14 -04001035 goto out;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001036
Yongqiang Yangf2a09af2012-09-23 23:16:03 -04001037 /*
1038 * Check free inodes count before loading bitmap.
1039 */
Wang Shilong2fe435d82017-08-24 11:58:18 -04001040 if (ext4_free_inodes_count(sb, gdp) == 0)
1041 goto next_group;
Yongqiang Yangf2a09af2012-09-23 23:16:03 -04001042
Harshad Shirwadkar8016e292020-10-15 13:37:59 -07001043 if (!(sbi->s_mount_state & EXT4_FC_REPLAY)) {
1044 grp = ext4_get_group_info(sb, group);
1045 /*
1046 * Skip groups with already-known suspicious inode
1047 * tables
1048 */
Theodore Ts'o5354b2a2023-04-29 00:06:28 -04001049 if (!grp || EXT4_MB_GRP_IBITMAP_CORRUPT(grp))
Harshad Shirwadkar8016e292020-10-15 13:37:59 -07001050 goto next_group;
1051 }
Darrick J. Wong87a39382013-08-28 18:32:58 -04001052
Aneesh Kumar K.V3300bed2009-01-03 22:33:39 -05001053 brelse(inode_bitmap_bh);
1054 inode_bitmap_bh = ext4_read_inode_bitmap(sb, group);
Darrick J. Wong87a39382013-08-28 18:32:58 -04001055 /* Skip groups with suspicious inode tables */
Harshad Shirwadkar8016e292020-10-15 13:37:59 -07001056 if (((!(sbi->s_mount_state & EXT4_FC_REPLAY))
1057 && EXT4_MB_GRP_IBITMAP_CORRUPT(grp)) ||
Darrick J. Wong9008a582015-10-17 21:33:24 -04001058 IS_ERR(inode_bitmap_bh)) {
1059 inode_bitmap_bh = NULL;
Wang Shilong2fe435d82017-08-24 11:58:18 -04001060 goto next_group;
Darrick J. Wong87a39382013-08-28 18:32:58 -04001061 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001062
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001063repeat_in_this_group:
Wang Shilong901ed072017-08-24 12:56:35 -04001064 ret2 = find_inode_bit(sb, group, inode_bitmap_bh, &ino);
1065 if (!ret2)
Theodore Ts'oa34eb502013-07-26 15:15:46 -04001066 goto next_group;
Wang Shilong901ed072017-08-24 12:56:35 -04001067
1068 if (group == 0 && (ino + 1) < EXT4_FIRST_INO(sb)) {
Theodore Ts'o119c0d42012-02-06 20:12:03 -05001069 ext4_error(sb, "reserved inode found cleared - "
1070 "inode=%lu", ino + 1);
Wang Shilong206f6d52018-05-12 12:15:21 -04001071 ext4_mark_group_bitmap_corrupted(sb, group,
1072 EXT4_GROUP_INFO_IBITMAP_CORRUPT);
Wang Shilong2fe435d82017-08-24 11:58:18 -04001073 goto next_group;
Theodore Ts'o119c0d42012-02-06 20:12:03 -05001074 }
Wang Shilong901ed072017-08-24 12:56:35 -04001075
Harshad Shirwadkar8016e292020-10-15 13:37:59 -07001076 if ((!(sbi->s_mount_state & EXT4_FC_REPLAY)) && !handle) {
Theodore Ts'o11395752013-02-09 16:27:09 -05001077 BUG_ON(nblocks <= 0);
changfengnan5f3e2402022-10-08 20:05:18 +08001078 handle = __ext4_journal_start_sb(NULL, dir->i_sb,
1079 line_no, handle_type, nblocks, 0,
yangerkuna70fd5a2019-12-13 09:49:00 +08001080 ext4_trans_default_revoke_credits(sb));
Theodore Ts'o11395752013-02-09 16:27:09 -05001081 if (IS_ERR(handle)) {
1082 err = PTR_ERR(handle);
Jan Karaeb9cc7e2013-04-19 13:38:14 -04001083 ext4_std_error(sb, err);
1084 goto out;
Theodore Ts'o11395752013-02-09 16:27:09 -05001085 }
1086 }
Eric Sandeenffb53872012-10-28 22:24:57 -04001087 BUFFER_TRACE(inode_bitmap_bh, "get_write_access");
Jan Kara188c2992021-08-16 11:57:04 +02001088 err = ext4_journal_get_write_access(handle, sb, inode_bitmap_bh,
1089 EXT4_JTR_NONE);
Jan Karaeb9cc7e2013-04-19 13:38:14 -04001090 if (err) {
1091 ext4_std_error(sb, err);
1092 goto out;
1093 }
Theodore Ts'o119c0d42012-02-06 20:12:03 -05001094 ext4_lock_group(sb, group);
1095 ret2 = ext4_test_and_set_bit(ino, inode_bitmap_bh->b_data);
Wang Shilong901ed072017-08-24 12:56:35 -04001096 if (ret2) {
1097 /* Someone already took the bit. Repeat the search
1098 * with lock held.
1099 */
1100 ret2 = find_inode_bit(sb, group, inode_bitmap_bh, &ino);
1101 if (ret2) {
1102 ext4_set_bit(ino, inode_bitmap_bh->b_data);
1103 ret2 = 0;
1104 } else {
1105 ret2 = 1; /* we didn't grab the inode */
1106 }
1107 }
Theodore Ts'o119c0d42012-02-06 20:12:03 -05001108 ext4_unlock_group(sb, group);
1109 ino++; /* the inode bitmap is zero-based */
1110 if (!ret2)
1111 goto got; /* we grabbed the inode! */
Wang Shilong901ed072017-08-24 12:56:35 -04001112
Theodore Ts'o119c0d42012-02-06 20:12:03 -05001113 if (ino < EXT4_INODES_PER_GROUP(sb))
1114 goto repeat_in_this_group;
Theodore Ts'oa34eb502013-07-26 15:15:46 -04001115next_group:
1116 if (++group == ngroups)
1117 group = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001118 }
1119 err = -ENOSPC;
1120 goto out;
1121
1122got:
Eric Sandeenffb53872012-10-28 22:24:57 -04001123 BUFFER_TRACE(inode_bitmap_bh, "call ext4_handle_dirty_metadata");
1124 err = ext4_handle_dirty_metadata(handle, NULL, inode_bitmap_bh);
Jan Karaeb9cc7e2013-04-19 13:38:14 -04001125 if (err) {
1126 ext4_std_error(sb, err);
1127 goto out;
1128 }
Eric Sandeenffb53872012-10-28 22:24:57 -04001129
Theodore Ts'o61c219f2014-07-05 16:28:35 -04001130 BUFFER_TRACE(group_desc_bh, "get_write_access");
Jan Kara188c2992021-08-16 11:57:04 +02001131 err = ext4_journal_get_write_access(handle, sb, group_desc_bh,
1132 EXT4_JTR_NONE);
Theodore Ts'o61c219f2014-07-05 16:28:35 -04001133 if (err) {
1134 ext4_std_error(sb, err);
1135 goto out;
1136 }
1137
Andreas Dilger717d50e2007-10-16 18:38:25 -04001138 /* We may have to initialize the block bitmap if it isn't already */
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04001139 if (ext4_has_group_desc_csum(sb) &&
Andreas Dilger717d50e2007-10-16 18:38:25 -04001140 gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
Aneesh Kumar K.V3300bed2009-01-03 22:33:39 -05001141 struct buffer_head *block_bitmap_bh;
Andreas Dilger717d50e2007-10-16 18:38:25 -04001142
Aneesh Kumar K.V3300bed2009-01-03 22:33:39 -05001143 block_bitmap_bh = ext4_read_block_bitmap(sb, group);
Darrick J. Wong9008a582015-10-17 21:33:24 -04001144 if (IS_ERR(block_bitmap_bh)) {
1145 err = PTR_ERR(block_bitmap_bh);
Jan Kara599a9b72014-10-30 10:53:16 -04001146 goto out;
1147 }
Aneesh Kumar K.V3300bed2009-01-03 22:33:39 -05001148 BUFFER_TRACE(block_bitmap_bh, "get block bitmap access");
Jan Kara188c2992021-08-16 11:57:04 +02001149 err = ext4_journal_get_write_access(handle, sb, block_bitmap_bh,
1150 EXT4_JTR_NONE);
Andreas Dilger717d50e2007-10-16 18:38:25 -04001151 if (err) {
Aneesh Kumar K.V3300bed2009-01-03 22:33:39 -05001152 brelse(block_bitmap_bh);
Jan Karaeb9cc7e2013-04-19 13:38:14 -04001153 ext4_std_error(sb, err);
1154 goto out;
Andreas Dilger717d50e2007-10-16 18:38:25 -04001155 }
1156
Theodore Ts'ofd034a82011-09-09 18:42:51 -04001157 BUFFER_TRACE(block_bitmap_bh, "dirty block bitmap");
1158 err = ext4_handle_dirty_metadata(handle, NULL, block_bitmap_bh);
Theodore Ts'ofd034a82011-09-09 18:42:51 -04001159
Andreas Dilger717d50e2007-10-16 18:38:25 -04001160 /* recheck and clear flag under lock if we still need to */
Theodore Ts'ofd034a82011-09-09 18:42:51 -04001161 ext4_lock_group(sb, group);
Theodore Ts'o88446182018-06-14 00:58:00 -04001162 if (ext4_has_group_desc_csum(sb) &&
1163 (gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT))) {
Aneesh Kumar K.V3300bed2009-01-03 22:33:39 -05001164 gdp->bg_flags &= cpu_to_le16(~EXT4_BG_BLOCK_UNINIT);
Theodore Ts'o021b65b2011-09-09 19:08:51 -04001165 ext4_free_group_clusters_set(sb, gdp,
Theodore Ts'ocff1dfd72011-09-09 19:12:51 -04001166 ext4_free_clusters_after_init(sb, group, gdp));
Kemeng Shi1df9bde2023-02-22 04:30:27 +08001167 ext4_block_bitmap_csum_set(sb, gdp, block_bitmap_bh);
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04001168 ext4_group_desc_csum_set(sb, group, gdp);
Andreas Dilger717d50e2007-10-16 18:38:25 -04001169 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04001170 ext4_unlock_group(sb, group);
Theodore Ts'oaeb1e5d62012-11-29 21:21:22 -05001171 brelse(block_bitmap_bh);
Andreas Dilger717d50e2007-10-16 18:38:25 -04001172
Jan Karaeb9cc7e2013-04-19 13:38:14 -04001173 if (err) {
1174 ext4_std_error(sb, err);
1175 goto out;
1176 }
Andreas Dilger717d50e2007-10-16 18:38:25 -04001177 }
Theodore Ts'o119c0d42012-02-06 20:12:03 -05001178
Theodore Ts'o119c0d42012-02-06 20:12:03 -05001179 /* Update the relevant bg descriptor fields */
Darrick J. Wong41a246d2012-04-29 18:33:10 -04001180 if (ext4_has_group_desc_csum(sb)) {
Theodore Ts'o119c0d42012-02-06 20:12:03 -05001181 int free;
Harshad Shirwadkar8016e292020-10-15 13:37:59 -07001182 struct ext4_group_info *grp = NULL;
Theodore Ts'o119c0d42012-02-06 20:12:03 -05001183
Harshad Shirwadkar8016e292020-10-15 13:37:59 -07001184 if (!(sbi->s_mount_state & EXT4_FC_REPLAY)) {
1185 grp = ext4_get_group_info(sb, group);
Theodore Ts'o5354b2a2023-04-29 00:06:28 -04001186 if (!grp) {
1187 err = -EFSCORRUPTED;
1188 goto out;
1189 }
Harshad Shirwadkar8016e292020-10-15 13:37:59 -07001190 down_read(&grp->alloc_sem); /*
1191 * protect vs itable
1192 * lazyinit
1193 */
1194 }
Theodore Ts'o119c0d42012-02-06 20:12:03 -05001195 ext4_lock_group(sb, group); /* while we modify the bg desc */
1196 free = EXT4_INODES_PER_GROUP(sb) -
1197 ext4_itable_unused_count(sb, gdp);
1198 if (gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_UNINIT)) {
1199 gdp->bg_flags &= cpu_to_le16(~EXT4_BG_INODE_UNINIT);
1200 free = 0;
1201 }
1202 /*
1203 * Check the relative inode number against the last used
1204 * relative inode number in this group. if it is greater
1205 * we need to update the bg_itable_unused count
1206 */
1207 if (ino > free)
1208 ext4_itable_unused_set(sb, gdp,
1209 (EXT4_INODES_PER_GROUP(sb) - ino));
Harshad Shirwadkar8016e292020-10-15 13:37:59 -07001210 if (!(sbi->s_mount_state & EXT4_FC_REPLAY))
1211 up_read(&grp->alloc_sem);
Tao Ma6f2e9f02012-05-28 18:20:59 -04001212 } else {
1213 ext4_lock_group(sb, group);
Theodore Ts'o119c0d42012-02-06 20:12:03 -05001214 }
Tao Ma6f2e9f02012-05-28 18:20:59 -04001215
Theodore Ts'o119c0d42012-02-06 20:12:03 -05001216 ext4_free_inodes_set(sb, gdp, ext4_free_inodes_count(sb, gdp) - 1);
1217 if (S_ISDIR(mode)) {
1218 ext4_used_dirs_set(sb, gdp, ext4_used_dirs_count(sb, gdp) + 1);
1219 if (sbi->s_log_groups_per_flex) {
1220 ext4_group_t f = ext4_flex_group(sbi, group);
1221
Suraj Jitindar Singh7c990722020-02-18 19:08:51 -08001222 atomic_inc(&sbi_array_rcu_deref(sbi, s_flex_groups,
1223 f)->used_dirs);
Theodore Ts'o119c0d42012-02-06 20:12:03 -05001224 }
1225 }
Darrick J. Wong41a246d2012-04-29 18:33:10 -04001226 if (ext4_has_group_desc_csum(sb)) {
Kemeng Shi4fd873c2023-02-22 04:30:25 +08001227 ext4_inode_bitmap_csum_set(sb, gdp, inode_bitmap_bh,
Darrick J. Wong41a246d2012-04-29 18:33:10 -04001228 EXT4_INODES_PER_GROUP(sb) / 8);
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04001229 ext4_group_desc_csum_set(sb, group, gdp);
Theodore Ts'o119c0d42012-02-06 20:12:03 -05001230 }
Tao Ma6f2e9f02012-05-28 18:20:59 -04001231 ext4_unlock_group(sb, group);
Theodore Ts'o119c0d42012-02-06 20:12:03 -05001232
Aneesh Kumar K.V3300bed2009-01-03 22:33:39 -05001233 BUFFER_TRACE(group_desc_bh, "call ext4_handle_dirty_metadata");
1234 err = ext4_handle_dirty_metadata(handle, NULL, group_desc_bh);
Jan Karaeb9cc7e2013-04-19 13:38:14 -04001235 if (err) {
1236 ext4_std_error(sb, err);
1237 goto out;
1238 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001239
1240 percpu_counter_dec(&sbi->s_freeinodes_counter);
1241 if (S_ISDIR(mode))
1242 percpu_counter_inc(&sbi->s_dirs_counter);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001243
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001244 if (sbi->s_log_groups_per_flex) {
1245 flex_group = ext4_flex_group(sbi, group);
Suraj Jitindar Singh7c990722020-02-18 19:08:51 -08001246 atomic_dec(&sbi_array_rcu_deref(sbi, s_flex_groups,
1247 flex_group)->free_inodes);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001248 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001249
Andreas Dilger717d50e2007-10-16 18:38:25 -04001250 inode->i_ino = ino + group * EXT4_INODES_PER_GROUP(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001251 /* This is the optimal IO size (for stat), not the fs block size */
1252 inode->i_blocks = 0;
Jeff Laytonb898ab22023-10-04 14:52:20 -04001253 simple_inode_init_ts(inode);
1254 ei->i_crtime = inode_get_mtime(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001255
1256 memset(ei->i_data, 0, sizeof(ei->i_data));
1257 ei->i_dir_start_lookup = 0;
1258 ei->i_disksize = 0;
1259
Eryu Guan4af83502011-10-31 18:21:29 -04001260 /* Don't inherit extent flag from directory, amongst others. */
Duane Griffin2dc6b0d2009-02-15 18:09:20 -05001261 ei->i_flags =
1262 ext4_mask_flags(mode, EXT4_I(dir)->i_flags & EXT4_FL_INHERITED);
Tahsin Erdogan1b917ed2017-06-21 21:21:39 -04001263 ei->i_flags |= i_flags;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001264 ei->i_file_acl = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001265 ei->i_dtime = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001266 ei->i_block_group = group;
Theodore Ts'oa4912122009-03-12 12:18:34 -04001267 ei->i_last_alloc_group = ~0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001268
Ira Weiny043546e2020-05-28 07:59:59 -07001269 ext4_set_inode_flags(inode, true);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001270 if (IS_DIRSYNC(inode))
Frank Mayhar03901312009-01-07 00:06:22 -05001271 ext4_handle_sync(handle);
Al Viro6b38e842008-12-30 02:03:31 -05001272 if (insert_inode_locked(inode) < 0) {
Jan Karaacd6ad82011-12-18 17:37:02 -05001273 /*
1274 * Likely a bitmap corruption causing inode to be allocated
1275 * twice.
1276 */
1277 err = -EIO;
Jan Karaeb9cc7e2013-04-19 13:38:14 -04001278 ext4_error(sb, "failed to insert inode %lu: doubly allocated?",
1279 inode->i_ino);
Wang Shilong206f6d52018-05-12 12:15:21 -04001280 ext4_mark_group_bitmap_corrupted(sb, group,
1281 EXT4_GROUP_INFO_IBITMAP_CORRUPT);
Jan Karaeb9cc7e2013-04-19 13:38:14 -04001282 goto out;
Al Viro6b38e842008-12-30 02:03:31 -05001283 }
Jason A. Donenfelda251c172022-10-05 17:43:22 +02001284 inode->i_generation = get_random_u32();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001285
Darrick J. Wong814525f2012-04-29 18:31:10 -04001286 /* Precompute checksum seed for inode metadata */
Dmitry Monakhov9aa5d32b2014-10-13 03:36:16 -04001287 if (ext4_has_metadata_csum(sb)) {
Darrick J. Wong814525f2012-04-29 18:31:10 -04001288 __u32 csum;
Darrick J. Wong814525f2012-04-29 18:31:10 -04001289 __le32 inum = cpu_to_le32(inode->i_ino);
1290 __le32 gen = cpu_to_le32(inode->i_generation);
1291 csum = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&inum,
1292 sizeof(inum));
1293 ei->i_csum_seed = ext4_chksum(sbi, csum, (__u8 *)&gen,
1294 sizeof(gen));
1295 }
1296
Theodore Ts'o353eb832011-01-10 12:18:25 -05001297 ext4_clear_state_flags(ei); /* Only relevant on 32-bit archs */
Theodore Ts'o19f5fb72010-01-24 14:34:07 -05001298 ext4_set_inode_state(inode, EXT4_STATE_NEW);
Kalpak Shahef7f3832007-07-18 09:15:20 -04001299
Jun Piao49598e02018-01-11 13:17:49 -05001300 ei->i_extra_isize = sbi->s_want_extra_isize;
Tao Maf08225d2012-12-10 14:06:03 -05001301 ei->i_inline_off = 0;
Theodore Ts'o4811d992021-04-12 17:19:00 -04001302 if (ext4_has_feature_inline_data(sb) &&
1303 (!(ei->i_flags & EXT4_DAX_FL) || S_ISDIR(mode)))
Tao Maf08225d2012-12-10 14:06:03 -05001304 ext4_set_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001305 ret = inode;
Christoph Hellwig63936dd2010-03-03 09:05:01 -05001306 err = dquot_alloc_inode(inode);
1307 if (err)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001308 goto fail_drop;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001309
Eric Biggersaa1dca32017-05-02 00:49:54 -04001310 /*
1311 * Since the encryption xattr will always be unique, create it first so
1312 * that it's less likely to end up in an external xattr block and
1313 * prevent its deduplication.
1314 */
1315 if (encrypt) {
Eric Biggers02ce5312020-09-16 21:11:26 -07001316 err = fscrypt_set_context(inode, handle);
Eric Biggersaa1dca32017-05-02 00:49:54 -04001317 if (err)
1318 goto fail_free_drop;
1319 }
1320
Tahsin Erdogan1b917ed2017-06-21 21:21:39 -04001321 if (!(ei->i_flags & EXT4_EA_INODE_FL)) {
1322 err = ext4_init_acl(handle, inode, dir);
1323 if (err)
1324 goto fail_free_drop;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001325
Tahsin Erdoganad47f9532017-07-06 00:00:59 -04001326 err = ext4_init_security(handle, inode, dir, qstr);
1327 if (err)
1328 goto fail_free_drop;
1329 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001330
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04001331 if (ext4_has_feature_extents(sb)) {
Eric Sandeene4079a12008-07-11 19:27:31 -04001332 /* set extent flag only for directory, file and normal symlink*/
Aneesh Kumar K.Ve65187e2008-04-29 08:11:12 -04001333 if (S_ISDIR(mode) || S_ISREG(mode) || S_ISLNK(mode)) {
Dmitry Monakhov12e9b892010-05-16 22:00:00 -04001334 ext4_set_inode_flag(inode, EXT4_INODE_EXTENTS);
Aneesh Kumar K.V42bf0382008-02-25 16:38:03 -05001335 ext4_ext_tree_init(handle, inode);
Aneesh Kumar K.V42bf0382008-02-25 16:38:03 -05001336 }
Alex Tomasa86c6182006-10-11 01:21:03 -07001337 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001338
Luis Henriques (SUSE)2d4d6bd2024-05-27 17:14:47 +01001339 ext4_update_inode_fsync_trans(handle, inode, 1);
Theodore Ts'o688f8692011-03-16 17:16:31 -04001340
Aneesh Kumar K.V8753e882008-04-29 22:00:36 -04001341 err = ext4_mark_inode_dirty(handle, inode);
1342 if (err) {
1343 ext4_std_error(sb, err);
1344 goto fail_free_drop;
1345 }
1346
Mingming Cao617ba132006-10-11 01:20:53 -07001347 ext4_debug("allocating inode %lu\n", inode->i_ino);
Theodore Ts'o9bffad12009-06-17 11:48:11 -04001348 trace_ext4_allocate_inode(inode, dir, mode);
Aneesh Kumar K.V3300bed2009-01-03 22:33:39 -05001349 brelse(inode_bitmap_bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001350 return ret;
1351
1352fail_free_drop:
Christoph Hellwig63936dd2010-03-03 09:05:01 -05001353 dquot_free_inode(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001354fail_drop:
Miklos Szeredi6d6b77f2011-10-28 14:13:28 +02001355 clear_nlink(inode);
Al Viro6b38e842008-12-30 02:03:31 -05001356 unlock_new_inode(inode);
Jan Karaeb9cc7e2013-04-19 13:38:14 -04001357out:
1358 dquot_drop(inode);
1359 inode->i_flags |= S_NOQUOTA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001360 iput(inode);
Aneesh Kumar K.V3300bed2009-01-03 22:33:39 -05001361 brelse(inode_bitmap_bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001362 return ERR_PTR(err);
1363}
1364
1365/* Verify that we are loading a valid orphan from disk */
Mingming Cao617ba132006-10-11 01:20:53 -07001366struct inode *ext4_orphan_get(struct super_block *sb, unsigned long ino)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001367{
Mingming Cao617ba132006-10-11 01:20:53 -07001368 unsigned long max_ino = le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count);
Avantika Mathurfd2d4292008-01-28 23:58:27 -05001369 ext4_group_t block_group;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001370 int bit;
Theodore Ts'o7827a7f2016-04-30 00:49:54 -04001371 struct buffer_head *bitmap_bh = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001372 struct inode *inode = NULL;
Theodore Ts'o7827a7f2016-04-30 00:49:54 -04001373 int err = -EFSCORRUPTED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001374
Theodore Ts'o7827a7f2016-04-30 00:49:54 -04001375 if (ino < EXT4_FIRST_INO(sb) || ino > max_ino)
1376 goto bad_orphan;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001377
Mingming Cao617ba132006-10-11 01:20:53 -07001378 block_group = (ino - 1) / EXT4_INODES_PER_GROUP(sb);
1379 bit = (ino - 1) % EXT4_INODES_PER_GROUP(sb);
Eric Sandeene29d1cd2008-08-02 21:21:02 -04001380 bitmap_bh = ext4_read_inode_bitmap(sb, block_group);
Wang Shilong206f6d52018-05-12 12:15:21 -04001381 if (IS_ERR(bitmap_bh))
Al Viroe884bce2018-10-10 16:41:40 -04001382 return ERR_CAST(bitmap_bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001383
1384 /* Having the inode bit set should be a 100% indicator that this
1385 * is a valid orphan (no e2fsck run on fs). Orphans also include
1386 * inodes that were being truncated, so we can't check i_nlink==0.
1387 */
David Howells1d1fe1e2008-02-07 00:15:37 -08001388 if (!ext4_test_bit(bit, bitmap_bh->b_data))
1389 goto bad_orphan;
1390
Theodore Ts'o8a363972018-12-19 12:29:13 -05001391 inode = ext4_iget(sb, ino, EXT4_IGET_NORMAL);
Theodore Ts'o7827a7f2016-04-30 00:49:54 -04001392 if (IS_ERR(inode)) {
1393 err = PTR_ERR(inode);
Theodore Ts'o54d3adb2020-03-28 19:33:43 -04001394 ext4_error_err(sb, -err,
1395 "couldn't read orphan inode %lu (err %d)",
1396 ino, err);
Xiyu Yang3bbd0ef2020-04-23 13:09:27 +08001397 brelse(bitmap_bh);
Theodore Ts'o7827a7f2016-04-30 00:49:54 -04001398 return inode;
1399 }
David Howells1d1fe1e2008-02-07 00:15:37 -08001400
Duane Griffin91ef4ca2008-07-11 19:27:31 -04001401 /*
Theodore Ts'oc9eb13a2016-04-30 00:48:54 -04001402 * If the orphans has i_nlinks > 0 then it should be able to
1403 * be truncated, otherwise it won't be removed from the orphan
1404 * list during processing and an infinite loop will result.
1405 * Similarly, it must not be a bad inode.
Duane Griffin91ef4ca2008-07-11 19:27:31 -04001406 */
Theodore Ts'oc9eb13a2016-04-30 00:48:54 -04001407 if ((inode->i_nlink && !ext4_can_truncate(inode)) ||
1408 is_bad_inode(inode))
Duane Griffin91ef4ca2008-07-11 19:27:31 -04001409 goto bad_orphan;
1410
David Howells1d1fe1e2008-02-07 00:15:37 -08001411 if (NEXT_ORPHAN(inode) > max_ino)
1412 goto bad_orphan;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001413 brelse(bitmap_bh);
1414 return inode;
David Howells1d1fe1e2008-02-07 00:15:37 -08001415
David Howells1d1fe1e2008-02-07 00:15:37 -08001416bad_orphan:
Theodore Ts'o7827a7f2016-04-30 00:49:54 -04001417 ext4_error(sb, "bad orphan inode %lu", ino);
1418 if (bitmap_bh)
1419 printk(KERN_ERR "ext4_test_bit(bit=%d, block=%llu) = %d\n",
1420 bit, (unsigned long long)bitmap_bh->b_blocknr,
1421 ext4_test_bit(bit, bitmap_bh->b_data));
David Howells1d1fe1e2008-02-07 00:15:37 -08001422 if (inode) {
Theodore Ts'o7827a7f2016-04-30 00:49:54 -04001423 printk(KERN_ERR "is_bad_inode(inode)=%d\n",
David Howells1d1fe1e2008-02-07 00:15:37 -08001424 is_bad_inode(inode));
Theodore Ts'o7827a7f2016-04-30 00:49:54 -04001425 printk(KERN_ERR "NEXT_ORPHAN(inode)=%u\n",
David Howells1d1fe1e2008-02-07 00:15:37 -08001426 NEXT_ORPHAN(inode));
Theodore Ts'o7827a7f2016-04-30 00:49:54 -04001427 printk(KERN_ERR "max_ino=%lu\n", max_ino);
1428 printk(KERN_ERR "i_nlink=%u\n", inode->i_nlink);
David Howells1d1fe1e2008-02-07 00:15:37 -08001429 /* Avoid freeing blocks if we got a bad deleted inode */
1430 if (inode->i_nlink == 0)
1431 inode->i_blocks = 0;
1432 iput(inode);
1433 }
1434 brelse(bitmap_bh);
David Howells1d1fe1e2008-02-07 00:15:37 -08001435 return ERR_PTR(err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001436}
1437
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04001438unsigned long ext4_count_free_inodes(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001439{
1440 unsigned long desc_count;
Mingming Cao617ba132006-10-11 01:20:53 -07001441 struct ext4_group_desc *gdp;
Theodore Ts'o8df96752009-05-01 08:50:38 -04001442 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
Mingming Cao617ba132006-10-11 01:20:53 -07001443#ifdef EXT4FS_DEBUG
1444 struct ext4_super_block *es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001445 unsigned long bitmap_count, x;
1446 struct buffer_head *bitmap_bh = NULL;
1447
Mingming Cao617ba132006-10-11 01:20:53 -07001448 es = EXT4_SB(sb)->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001449 desc_count = 0;
1450 bitmap_count = 0;
1451 gdp = NULL;
Theodore Ts'o8df96752009-05-01 08:50:38 -04001452 for (i = 0; i < ngroups; i++) {
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04001453 gdp = ext4_get_group_desc(sb, i, NULL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001454 if (!gdp)
1455 continue;
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -05001456 desc_count += ext4_free_inodes_count(sb, gdp);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001457 brelse(bitmap_bh);
Eric Sandeene29d1cd2008-08-02 21:21:02 -04001458 bitmap_bh = ext4_read_inode_bitmap(sb, i);
Darrick J. Wong9008a582015-10-17 21:33:24 -04001459 if (IS_ERR(bitmap_bh)) {
1460 bitmap_bh = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001461 continue;
Darrick J. Wong9008a582015-10-17 21:33:24 -04001462 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001463
Theodore Ts'of6fb99c2012-06-30 19:14:57 -04001464 x = ext4_count_free(bitmap_bh->b_data,
1465 EXT4_INODES_PER_GROUP(sb) / 8);
Eric Sandeenc549a952008-01-28 23:58:27 -05001466 printk(KERN_DEBUG "group %lu: stored = %d, counted = %lu\n",
Peng Tao785b4b32009-07-27 21:44:40 -04001467 (unsigned long) i, ext4_free_inodes_count(sb, gdp), x);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001468 bitmap_count += x;
1469 }
1470 brelse(bitmap_bh);
Theodore Ts'o4776004f2008-09-08 23:00:52 -04001471 printk(KERN_DEBUG "ext4_count_free_inodes: "
1472 "stored = %u, computed = %lu, %lu\n",
1473 le32_to_cpu(es->s_free_inodes_count), desc_count, bitmap_count);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001474 return desc_count;
1475#else
1476 desc_count = 0;
Theodore Ts'o8df96752009-05-01 08:50:38 -04001477 for (i = 0; i < ngroups; i++) {
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04001478 gdp = ext4_get_group_desc(sb, i, NULL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001479 if (!gdp)
1480 continue;
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -05001481 desc_count += ext4_free_inodes_count(sb, gdp);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001482 cond_resched();
1483 }
1484 return desc_count;
1485#endif
1486}
1487
1488/* Called at mount-time, super-block is locked */
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04001489unsigned long ext4_count_dirs(struct super_block * sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001490{
1491 unsigned long count = 0;
Theodore Ts'o8df96752009-05-01 08:50:38 -04001492 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001493
Theodore Ts'o8df96752009-05-01 08:50:38 -04001494 for (i = 0; i < ngroups; i++) {
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04001495 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001496 if (!gdp)
1497 continue;
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -05001498 count += ext4_used_dirs_count(sb, gdp);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001499 }
1500 return count;
1501}
Lukas Czernerbfff6872010-10-27 21:30:05 -04001502
1503/*
1504 * Zeroes not yet zeroed inode table - just write zeroes through the whole
1505 * inode table. Must be called without any spinlock held. The only place
1506 * where it is called from on active part of filesystem is ext4lazyinit
1507 * thread, so we do not need any special locks, however we have to prevent
1508 * inode allocation from the current group, so we take alloc_sem lock, to
Theodore Ts'o119c0d42012-02-06 20:12:03 -05001509 * block ext4_new_inode() until we are finished.
Lukas Czernerbfff6872010-10-27 21:30:05 -04001510 */
H Hartley Sweetene0cbee32011-10-18 10:57:51 -04001511int ext4_init_inode_table(struct super_block *sb, ext4_group_t group,
Lukas Czernerbfff6872010-10-27 21:30:05 -04001512 int barrier)
1513{
1514 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1515 struct ext4_sb_info *sbi = EXT4_SB(sb);
1516 struct ext4_group_desc *gdp = NULL;
1517 struct buffer_head *group_desc_bh;
1518 handle_t *handle;
1519 ext4_fsblk_t blk;
1520 int num, ret = 0, used_blks = 0;
Zhang Yia149d2a2021-03-31 20:15:16 +08001521 unsigned long used_inos = 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04001522
Lukas Czernerbfff6872010-10-27 21:30:05 -04001523 gdp = ext4_get_group_desc(sb, group, &group_desc_bh);
Theodore Ts'o5354b2a2023-04-29 00:06:28 -04001524 if (!gdp || !grp)
Lukas Czernerbfff6872010-10-27 21:30:05 -04001525 goto out;
1526
1527 /*
1528 * We do not need to lock this, because we are the only one
1529 * handling this flag.
1530 */
1531 if (gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))
1532 goto out;
1533
Theodore Ts'o9924a922013-02-08 21:59:22 -05001534 handle = ext4_journal_start_sb(sb, EXT4_HT_MISC, 1);
Lukas Czernerbfff6872010-10-27 21:30:05 -04001535 if (IS_ERR(handle)) {
1536 ret = PTR_ERR(handle);
1537 goto out;
1538 }
1539
1540 down_write(&grp->alloc_sem);
1541 /*
1542 * If inode bitmap was already initialized there may be some
1543 * used inodes so we need to skip blocks with used inodes in
1544 * inode table.
1545 */
Zhang Yia149d2a2021-03-31 20:15:16 +08001546 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_UNINIT))) {
1547 used_inos = EXT4_INODES_PER_GROUP(sb) -
1548 ext4_itable_unused_count(sb, gdp);
1549 used_blks = DIV_ROUND_UP(used_inos, sbi->s_inodes_per_block);
Lukas Czernerbfff6872010-10-27 21:30:05 -04001550
Zhang Yia149d2a2021-03-31 20:15:16 +08001551 /* Bogus inode unused count? */
1552 if (used_blks < 0 || used_blks > sbi->s_itb_per_group) {
1553 ext4_error(sb, "Something is wrong with group %u: "
1554 "used itable blocks: %d; "
1555 "itable unused count: %u",
1556 group, used_blks,
1557 ext4_itable_unused_count(sb, gdp));
1558 ret = 1;
1559 goto err_out;
1560 }
1561
1562 used_inos += group * EXT4_INODES_PER_GROUP(sb);
1563 /*
1564 * Are there some uninitialized inodes in the inode table
1565 * before the first normal inode?
1566 */
1567 if ((used_blks != sbi->s_itb_per_group) &&
1568 (used_inos < EXT4_FIRST_INO(sb))) {
1569 ext4_error(sb, "Something is wrong with group %u: "
1570 "itable unused count: %u; "
1571 "itables initialized count: %ld",
1572 group, ext4_itable_unused_count(sb, gdp),
1573 used_inos);
1574 ret = 1;
1575 goto err_out;
1576 }
Lukas Czerner857ac882010-10-27 21:30:05 -04001577 }
1578
Lukas Czernerbfff6872010-10-27 21:30:05 -04001579 blk = ext4_inode_table(sb, gdp) + used_blks;
1580 num = sbi->s_itb_per_group - used_blks;
1581
1582 BUFFER_TRACE(group_desc_bh, "get_write_access");
Jan Kara188c2992021-08-16 11:57:04 +02001583 ret = ext4_journal_get_write_access(handle, sb, group_desc_bh,
1584 EXT4_JTR_NONE);
Lukas Czernerbfff6872010-10-27 21:30:05 -04001585 if (ret)
1586 goto err_out;
1587
Lukas Czernerbfff6872010-10-27 21:30:05 -04001588 /*
1589 * Skip zeroout if the inode table is full. But we set the ZEROED
1590 * flag anyway, because obviously, when it is full it does not need
1591 * further zeroing.
1592 */
1593 if (unlikely(num == 0))
1594 goto skip_zeroout;
1595
1596 ext4_debug("going to zero out inode table in group %d\n",
1597 group);
Theodore Ts'oa107e5a2010-10-27 23:44:47 -04001598 ret = sb_issue_zeroout(sb, blk, num, GFP_NOFS);
Lukas Czernerbfff6872010-10-27 21:30:05 -04001599 if (ret < 0)
1600 goto err_out;
Theodore Ts'oa107e5a2010-10-27 23:44:47 -04001601 if (barrier)
Christoph Hellwigc6bf3f02021-01-26 15:52:35 +01001602 blkdev_issue_flush(sb->s_bdev);
Lukas Czernerbfff6872010-10-27 21:30:05 -04001603
1604skip_zeroout:
1605 ext4_lock_group(sb, group);
1606 gdp->bg_flags |= cpu_to_le16(EXT4_BG_INODE_ZEROED);
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04001607 ext4_group_desc_csum_set(sb, group, gdp);
Lukas Czernerbfff6872010-10-27 21:30:05 -04001608 ext4_unlock_group(sb, group);
1609
1610 BUFFER_TRACE(group_desc_bh,
1611 "call ext4_handle_dirty_metadata");
1612 ret = ext4_handle_dirty_metadata(handle, NULL,
1613 group_desc_bh);
1614
1615err_out:
1616 up_write(&grp->alloc_sem);
1617 ext4_journal_stop(handle);
1618out:
1619 return ret;
1620}