lustre/kernel_patches/patches/ext3-reserve-inode-space-2.4.24.patch

   1 Index: linux-2.4.24/fs/ext3/ialloc.c
   2 ===================================================================
   3 --- linux-2.4.24.orig/fs/ext3/ialloc.c  2004-08-02 22:13:13.000000000 +0400
   4 +++ linux-2.4.24/fs/ext3/ialloc.c       2004-08-04 01:10:30.000000000 +0400
   5 @@ -41,6 +41,24 @@
   6   * when a file system is mounted (see ext3_read_super).
   7   */
   8
   9 +/*
  10 + * this is very simple policy: files with O_INRESERVE goes to last group;
  11 + * files with no O_INRESERVE goes to all groups, but last. probably we'll
  12 + * specify group for O_INRESERVE files later -bzzz */
  13 +static inline int ext3_group_allowed(struct super_block *sb, int mode, int group)
  14 +{
  15 +       if (!test_opt(sb, INRESERVE) || EXT3_SB(sb)->s_groups_count == 1)
  16 +               return 1;
  17 +
  18 +       if (mode & EXT3_S_INRESERVE) {
  19 +               if (group != EXT3_SB(sb)->s_groups_count - 1)
  20 +                       return 0;
  21 +       } else {
  22 +               if (group == EXT3_SB(sb)->s_groups_count -1 )
  23 +                       return 0;
  24 +       }
  25 +       return 1;
  26 +}
  27
  28  /*
  29   * Read the inode allocation bitmap for a given block_group, reading
  30 @@ -337,20 +355,33 @@
  31         return !ext3_test_bit(nr, bh2jh(bh)->b_committed_data);
  32  }
  33
  34 -int ext3_find_group_dir(const struct inode *dir,
  35 +int ext3_find_group_dir(const struct inode *dir, int mode,
  36                                 struct ext3_group_desc **gdp,
  37                                 struct buffer_head **bh)
  38  {
  39         struct super_block *sb = dir->i_sb;
  40         struct ext3_super_block *es;
  41         struct ext3_group_desc *tmp;
  42 -       int i = 0, j, avefreei;
  43 +       int i = 0, j, ifree, avefreei;
  44
  45         es = EXT3_SB(sb)->s_es;
  46 -       avefreei = le32_to_cpu(es->s_free_inodes_count) /
  47 -                       sb->u.ext3_sb.s_groups_count;
  48 +       ifree = le32_to_cpu(es->s_free_inodes_count);
  49 +       j = sb->u.ext3_sb.s_groups_count;
  50 +       if (test_opt(sb, INRESERVE) && EXT3_SB(sb)->s_groups_count > 1) {
  51 +               tmp = ext3_get_group_desc(sb, EXT3_SB(sb)->s_groups_count-1, NULL);
  52 +               if (mode & EXT3_S_INRESERVE) {
  53 +                       ifree = le16_to_cpu(tmp->bg_free_blocks_count);
  54 +                       j = 1;
  55 +               } else {
  56 +                       ifree -= le16_to_cpu(tmp->bg_free_blocks_count);
  57 +                       j--;
  58 +               }
  59 +       }
  60 +       avefreei = ifree / j;
  61         for (j = 0; j < sb->u.ext3_sb.s_groups_count; j++) {
  62                 struct buffer_head *temp_buffer;
  63 +               if (!ext3_group_allowed(sb, mode, j))
  64 +                       continue;
  65                 tmp = ext3_get_group_desc(sb, j, &temp_buffer);
  66                 if (tmp && le16_to_cpu(tmp->bg_free_inodes_count) &&
  67                         le16_to_cpu(tmp->bg_free_inodes_count) >= avefreei) {
  68 @@ -366,7 +397,7 @@
  69         return i;
  70  }
  71
  72 -int ext3_find_group_other(const struct inode *dir,
  73 +int ext3_find_group_other(const struct inode *dir, int mode,
  74                                 struct ext3_group_desc **gdp,
  75                                 struct buffer_head **bh)
  76  {
  77 @@ -379,7 +410,8 @@
  78          */
  79         i = dir->u.ext3_i.i_block_group;
  80         tmp = ext3_get_group_desc(sb, i, bh);
  81 -       if (tmp && le16_to_cpu(tmp->bg_free_inodes_count))
  82 +       if (tmp && le16_to_cpu(tmp->bg_free_inodes_count)
  83 +                       && ext3_group_allowed(sb, mode, i))
  84                 *gdp = tmp;
  85         else {
  86                 /*
  87 @@ -390,6 +422,8 @@
  88                         i += j;
  89                         if (i >= sb->u.ext3_sb.s_groups_count)
  90                                 i -= sb->u.ext3_sb.s_groups_count;
  91 +                       if (!ext3_group_allowed(sb, mode, i))
  92 +                               continue;
  93                         tmp = ext3_get_group_desc (sb, i, bh);
  94                         if (tmp && le16_to_cpu(tmp->bg_free_inodes_count)) {
  95                                 *gdp = tmp;
  96 @@ -405,6 +439,8 @@
  97                 for (j = 2; j < sb->u.ext3_sb.s_groups_count; j++) {
  98                         if (++i >= sb->u.ext3_sb.s_groups_count)
  99                                 i = 0;
 100 +                       if (!ext3_group_allowed(sb, mode, i))
 101 +                               continue;
 102                         tmp = ext3_get_group_desc (sb, i, bh);
 103                         if (tmp && le16_to_cpu(tmp->bg_free_inodes_count)) {
 104                                 *gdp = tmp;
 105 @@ -421,8 +457,8 @@
 106                                 struct buffer_head **bh)
 107  {
 108         if (S_ISDIR(mode))
 109 -               return ext3_find_group_dir(dir, gdp, bh);
 110 -       return ext3_find_group_other(dir, gdp, bh);
 111 +               return ext3_find_group_dir(dir, mode, gdp, bh);
 112 +       return ext3_find_group_other(dir, mode, gdp, bh);
 113  }
 114
 115  static int ext3_find_usable_inode(struct super_block *sb,
 116 @@ -549,6 +585,8 @@
 117         for (i = i + 1; i != k; i++) {
 118                 if (i >= sb->u.ext3_sb.s_groups_count)
 119                         i = 0;
 120 +               if (!ext3_group_allowed(sb, mode, i))
 121 +                       continue;
 122                 tmp = ext3_get_group_desc(sb, i, &bh2);
 123                 if (le16_to_cpu(tmp->bg_free_inodes_count) == 0)
 124                         continue;
 125 @@ -588,6 +626,7 @@
 126         if (buffer_jbd(bh) && bh2jh(bh)->b_committed_data)
 127                 J_ASSERT_BH(bh, !ext3_test_bit(j, bh2jh(bh)->b_committed_data));
 128
 129 +       J_ASSERT(ext3_group_allowed(sb, mode, i));
 130         j += i * EXT3_INODES_PER_GROUP(sb) + 1;
 131         if (j < EXT3_FIRST_INO(sb) || j > le32_to_cpu(es->s_inodes_count)) {
 132                 ext3_error (sb, "ext3_new_inode",
 133 @@ -628,7 +667,7 @@
 134                         mode |= S_ISGID;
 135         } else
 136                 inode->i_gid = current->fsgid;
 137 -       inode->i_mode = mode;
 138 +       inode->i_mode = mode & ~EXT3_S_INRESERVE;
 139
 140         inode->i_ino = j;
 141         /* This is the optimal IO size (for stat), not the fs block size */
 142 Index: linux-2.4.24/fs/ext3/super.c
 143 ===================================================================
 144 --- linux-2.4.24.orig/fs/ext3/super.c   2004-08-02 22:13:12.000000000 +0400
 145 +++ linux-2.4.24/fs/ext3/super.c        2004-08-04 00:10:07.000000000 +0400
 146 @@ -723,6 +723,10 @@
 147                                 return 0;
 148                         *resgid = v;
 149                 }
 150 +               else if (!strcmp (this_char, "inrsv")) {
 151 +                       set_opt (*mount_options, INRESERVE);
 152 +                       J_ASSERT((EXT3_S_INRESERVE & S_IALLUGO) == 0);
 153 +               }
 154                 else if (!strcmp (this_char, "resuid")) {
 155                         unsigned long v;
 156                         if (want_numeric(value, "resuid", &v))
 157 Index: linux-2.4.24/fs/ext3/namei.c
 158 ===================================================================
 159 --- linux-2.4.24.orig/fs/ext3/namei.c   2004-08-02 22:13:13.000000000 +0400
 160 +++ linux-2.4.24/fs/ext3/namei.c        2004-08-03 03:21:09.000000000 +0400
 161 @@ -1971,6 +1971,36 @@
 162         return err;
 163  }
 164
 165 +static int ext3_create_it (struct inode * dir, struct dentry * dentry, int mode,
 166 +                               struct lookup_intent *it)
 167 +{
 168 +       handle_t *handle;
 169 +       struct inode * inode;
 170 +       int err;
 171 +
 172 +       handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS +
 173 +                                       EXT3_INDEX_EXTRA_TRANS_BLOCKS + 3);
 174 +       if (IS_ERR(handle)) {
 175 +               return PTR_ERR(handle);
 176 +       }
 177 +
 178 +       if (IS_SYNC(dir))
 179 +               handle->h_sync = 1;
 180 +
 181 +       if (it && it->it_flags & O_INRESERVE)
 182 +               mode |= EXT3_S_INRESERVE;
 183 +       inode = ext3_new_inode_wantedi (handle, dir, mode, dentry);
 184 +       err = PTR_ERR(inode);
 185 +       if (!IS_ERR(inode)) {
 186 +               inode->i_op = &ext3_file_inode_operations;
 187 +               inode->i_fop = &ext3_file_operations;
 188 +               inode->i_mapping->a_ops = &ext3_aops;
 189 +               err = ext3_add_nondir(handle, dentry, inode);
 190 +       }
 191 +       ext3_journal_stop(handle, dir);
 192 +       return err;
 193 +}
 194 +
 195  static int ext3_mknod (struct inode * dir, struct dentry *dentry,
 196                         int mode, int rdev)
 197  {
 198 @@ -2803,6 +2833,7 @@
 199   */
 200  struct inode_operations ext3_dir_inode_operations = {
 201         create:         ext3_create,            /* BKL held */
 202 +       create_it:      ext3_create_it,         /* BKL held */
 203         lookup:         ext3_lookup,            /* BKL held */
 204         lookup_raw:     ext3_lookup_raw,        /* BKL held */
 205         link:           ext3_link,              /* BKL held */
 206 Index: linux-2.4.24/include/asm-i386/fcntl.h
 207 ===================================================================
 208 --- linux-2.4.24.orig/include/asm-i386/fcntl.h  2001-09-18 00:16:30.000000000 +0400
 209 +++ linux-2.4.24/include/asm-i386/fcntl.h       2004-08-03 01:20:55.000000000 +0400
 210 @@ -20,6 +20,7 @@
 211  #define O_LARGEFILE    0100000
 212  #define O_DIRECTORY    0200000 /* must be a directory */
 213  #define O_NOFOLLOW     0400000 /* don't follow links */
 214 +#define O_INRESERVE    01000000 /* allocate inodes in reserved space */
 215
 216  #define F_DUPFD                0       /* dup */
 217  #define F_GETFD                1       /* get close_on_exec */
 218 Index: linux-2.4.24/include/linux/ext3_fs.h
 219 ===================================================================
 220 --- linux-2.4.24.orig/include/linux/ext3_fs.h   2004-08-02 22:13:12.000000000 +0400
 221 +++ linux-2.4.24/include/linux/ext3_fs.h        2004-08-03 01:03:39.000000000 +0400
 222 @@ -343,6 +343,7 @@
 223  #define EXT3_MOUNT_ASYNCDEL            0x20000 /* Delayed deletion */
 224  #define EXT3_MOUNT_EXTENTS             0x100000/* Extents support */
 225  #define EXT3_MOUNT_EXTDEBUG            0x200000/* Extents debug */
 226 +#define EXT3_MOUNT_INRESERVE           0x400000/* reserve one group for O_INRESERVE */
 227
 228  /* Compatibility, for having both ext2_fs.h and ext3_fs.h included at once */
 229  #ifndef _LINUX_EXT2_FS_H
 230 @@ -482,6 +483,8 @@
 231
 232  #define EXT3_GOOD_OLD_INODE_SIZE 128
 233
 234 +#define EXT3_S_INRESERVE       01000000
 235 +
 236  /*
 237   * Feature set definitions
 238   */