mkfs_ext2: fixes for huge filesystems
authorDenys Vlasenko <vda.linux@googlemail.com>
Sun, 18 Oct 2009 16:05:27 +0000 (18:05 +0200)
committerDenys Vlasenko <vda.linux@googlemail.com>
Sun, 18 Oct 2009 16:05:27 +0000 (18:05 +0200)
function                                             old     new   delta
mkfs_ext2_main                                      2145    2617    +472
div_roundup                                           15      35     +20
packed_usage                                       26792   26776     -16

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
include/usage.h
util-linux/mkfs_ext2.c

index bdd9ae33b4c4fd269ceb4150041a905e4a0c8058..80111e5ef0f54ca88cb5195fdc687690d97947f1 100644 (file)
        "[-L LABEL] " \
        "[-n] " \
        /* "[-M last-mounted-directory] [-S] [-T filesystem-type] " */ \
-       "DEVICE [BLK_COUNT]"
+       "DEVICE [KBYTES]"
 #define mkfs_ext2_full_usage "\n" \
      "\n       -b BLK_SIZE     Block size in bytes" \
 /*   "\n       -c              Check for bad blocks before creating" */ \
index a990668df0dea84ac33df851eccdc6527e15ab8a..56203c789bf58e3675ac9d6c13dee4b9eb09537e 100644 (file)
@@ -70,7 +70,11 @@ static unsigned int_log2(unsigned arg)
 // taken from mkfs_minix.c. libbb candidate?
 static unsigned div_roundup(uint32_t size, uint32_t n)
 {
-       return (size + n-1) / n;
+       // Overflow-resistant
+       uint32_t res = size / n;
+       if (res * n != size)
+               res++;
+       return res;
 }
 
 static void allocate(uint8_t *bitmap, uint32_t blocksize, uint32_t start, uint32_t end)
@@ -161,6 +165,7 @@ int mkfs_ext2_main(int argc UNUSED_PARAM, char **argv)
        unsigned i, pos, n;
        unsigned bs, blocksize, blocksize_log2;
        unsigned nreserved = 5;
+       unsigned long long nblocks_ull;
        uint32_t nblocks;
        uint32_t ngroups;
        unsigned bytes_per_inode;
@@ -179,23 +184,13 @@ int mkfs_ext2_main(int argc UNUSED_PARAM, char **argv)
        struct ext2_dir *dir;
        uint8_t *buf;
 
-       bs = EXT2_MIN_BLOCK_SIZE;
        opt_complementary = "-1:b+:m+:i+";
        opts = getopt32(argv, "cl:b:f:i:I:J:G:N:m:o:g:L:M:O:r:E:T:U:jnqvFS",
                NULL, &bs, NULL, &bytes_per_inode, NULL, NULL, NULL, NULL,
                &nreserved, NULL, NULL, &label, NULL, NULL, NULL, NULL, NULL, NULL);
        argv += optind; // argv[0] -- device
 
-       // block size minimax, block size is a multiple of minimum
-       blocksize = bs;
-       if (blocksize < EXT2_MIN_BLOCK_SIZE
-        || blocksize > EXT2_MAX_BLOCK_SIZE
-        || (blocksize & (blocksize - 1)) // not power of 2
-       ) {
-               bb_error_msg_and_die("-%c is bad", 'b');
-       }
-
-       // reserved blocks count
+       // reserved blocks percentage
        if (nreserved > 50)
                bb_error_msg_and_die("-%c is bad", 'm');
 
@@ -210,6 +205,42 @@ int mkfs_ext2_main(int argc UNUSED_PARAM, char **argv)
        if (find_mount_point(argv[0], 0))
                bb_error_msg_and_die("can't format mounted filesystem");
 
+       // open the device, get size in kbytes
+       xmove_fd(xopen3(argv[0], O_WRONLY | O_CREAT, 0666), fd);
+       if (argv[1]) {
+               nblocks_ull = xatoull(argv[1]);
+       } else {
+               nblocks_ull = (uoff_t)xlseek(fd, 0, SEEK_END) / 1024;
+       }
+
+       // block size is a multiple of 1024
+       blocksize = 1024;
+       if (nblocks_ull >= 512*1024) // mke2fs 1.41.9 compat
+               blocksize = 4096;
+       if (EXT2_MAX_BLOCK_SIZE > 4096) {
+               // nblocks_ull >> 22 == size in 4gigabyte chunks.
+               // if it is >= 16k gigs, blocksize must be increased.
+               // Try "mke2fs -F image_std $((16 * 1024*1024*1024))"
+               while ((nblocks_ull >> 22) >= blocksize)
+                       blocksize *= 2;
+       }
+       if (opts & OPT_b)
+               blocksize = bs;
+       if (blocksize < EXT2_MIN_BLOCK_SIZE
+        || blocksize > EXT2_MAX_BLOCK_SIZE
+        || (blocksize & (blocksize - 1)) // not power of 2
+       ) {
+               bb_error_msg_and_die("blocksize %u is bad", blocksize);
+       }
+       blocksize_log2 = int_log2(blocksize);
+       nblocks_ull >>= (blocksize_log2 - EXT2_MIN_BLOCK_LOG_SIZE);
+       // nblocks: the total number of blocks in the filesystem
+       nblocks = nblocks_ull;
+       if (nblocks != nblocks_ull)
+               bb_error_msg_and_die("block count doesn't fit in 32 bits");
+       if (nblocks < 8)
+               bb_error_msg_and_die("need >= 8 blocks");
+
        // TODO: 5?/5 WE MUST NOT DEPEND ON WHETHER DEVICE IS /dev/zero 'ed OR NOT
        // TODO: 3/5 refuse if mounted
        // TODO: 4/5 compat options
@@ -226,7 +257,6 @@ int mkfs_ext2_main(int argc UNUSED_PARAM, char **argv)
        sb->s_magic = EXT2_SUPER_MAGIC;
        sb->s_inode_size = sizeof(*inode);
        sb->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
-       blocksize_log2 = int_log2(blocksize);
        sb->s_log_block_size = sb->s_log_frag_size = blocksize_log2 - EXT2_MIN_BLOCK_LOG_SIZE;
        // first 1024 bytes of the device are for boot record. If block size is 1024 bytes, then
        // the first block available for data is 1, otherwise 0
@@ -261,18 +291,8 @@ int mkfs_ext2_main(int argc UNUSED_PARAM, char **argv)
         */
        sb->s_max_mnt_count += sb->s_uuid[ARRAY_SIZE(sb->s_uuid)-1] % EXT2_DFL_MAX_MNT_COUNT;
 
-       // open the device, get number of blocks
-       xmove_fd(xopen3(argv[0], O_WRONLY | O_CREAT, 0666), fd);
-       if (argv[1]) {
-               nblocks = xatou(argv[1]);
-       } else {
-               nblocks = (uoff_t)xlseek(fd, 0, SEEK_END) >> blocksize_log2;
-       }
        sb->s_blocks_count = nblocks;
 
-       // nblocks is the total number of blocks in the filesystem
-       if (nblocks < 8)
-               bb_error_msg_and_die("nblocks");
        // reserve blocks for superuser
        sb->s_r_blocks_count = ((uint64_t) nblocks * nreserved) / 100;