omfs: add inode routines
Bob Copeland [Sat, 26 Jul 2008 02:45:15 +0000 (19:45 -0700)]
Add basic superblock and inode handling routines for OMFS

Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs/omfs/inode.c [new file with mode: 0644]

diff --git a/fs/omfs/inode.c b/fs/omfs/inode.c
new file mode 100644 (file)
index 0000000..d865f55
--- /dev/null
@@ -0,0 +1,553 @@
+/*
+ * Optimized MPEG FS - inode and super operations.
+ * Copyright (C) 2006 Bob Copeland <me@bobcopeland.com>
+ * Released under GPL v2.
+ */
+#include <linux/version.h>
+#include <linux/module.h>
+#include <linux/sched.h>
+#include <linux/fs.h>
+#include <linux/vfs.h>
+#include <linux/parser.h>
+#include <linux/buffer_head.h>
+#include <linux/vmalloc.h>
+#include <linux/crc-itu-t.h>
+#include "omfs.h"
+
+MODULE_AUTHOR("Bob Copeland <me@bobcopeland.com>");
+MODULE_DESCRIPTION("OMFS (ReplayTV/Karma) Filesystem for Linux");
+MODULE_LICENSE("GPL");
+
+struct inode *omfs_new_inode(struct inode *dir, int mode)
+{
+       struct inode *inode;
+       u64 new_block;
+       int err;
+       int len;
+       struct omfs_sb_info *sbi = OMFS_SB(dir->i_sb);
+
+       inode = new_inode(dir->i_sb);
+       if (!inode)
+               return ERR_PTR(-ENOMEM);
+
+       err = omfs_allocate_range(dir->i_sb, sbi->s_mirrors, sbi->s_mirrors,
+                       &new_block, &len);
+       if (err)
+               goto fail;
+
+       inode->i_ino = new_block;
+       inode->i_mode = mode;
+       inode->i_uid = current->fsuid;
+       inode->i_gid = current->fsgid;
+       inode->i_blocks = 0;
+       inode->i_mapping->a_ops = &omfs_aops;
+
+       inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
+       switch (mode & S_IFMT) {
+       case S_IFDIR:
+               inode->i_op = &omfs_dir_inops;
+               inode->i_fop = &omfs_dir_operations;
+               inode->i_size = sbi->s_sys_blocksize;
+               inc_nlink(inode);
+               break;
+       case S_IFREG:
+               inode->i_op = &omfs_file_inops;
+               inode->i_fop = &omfs_file_operations;
+               inode->i_size = 0;
+               break;
+       }
+
+       insert_inode_hash(inode);
+       mark_inode_dirty(inode);
+       return inode;
+fail:
+       make_bad_inode(inode);
+       iput(inode);
+       return ERR_PTR(err);
+}
+
+/*
+ * Update the header checksums for a dirty inode based on its contents.
+ * Caller is expected to hold the buffer head underlying oi and mark it
+ * dirty.
+ */
+static void omfs_update_checksums(struct omfs_inode *oi)
+{
+       int xor, i, ofs = 0, count;
+       u16 crc = 0;
+       unsigned char *ptr = (unsigned char *) oi;
+
+       count = be32_to_cpu(oi->i_head.h_body_size);
+       ofs = sizeof(struct omfs_header);
+
+       crc = crc_itu_t(crc, ptr + ofs, count);
+       oi->i_head.h_crc = cpu_to_be16(crc);
+
+       xor = ptr[0];
+       for (i = 1; i < OMFS_XOR_COUNT; i++)
+               xor ^= ptr[i];
+
+       oi->i_head.h_check_xor = xor;
+}
+
+static int omfs_write_inode(struct inode *inode, int wait)
+{
+       struct omfs_inode *oi;
+       struct omfs_sb_info *sbi = OMFS_SB(inode->i_sb);
+       struct buffer_head *bh, *bh2;
+       unsigned int block;
+       u64 ctime;
+       int i;
+       int ret = -EIO;
+       int sync_failed = 0;
+
+       /* get current inode since we may have written sibling ptrs etc. */
+       block = clus_to_blk(sbi, inode->i_ino);
+       bh = sb_bread(inode->i_sb, block);
+       if (!bh)
+               goto out;
+
+       oi = (struct omfs_inode *) bh->b_data;
+
+       oi->i_head.h_self = cpu_to_be64(inode->i_ino);
+       if (S_ISDIR(inode->i_mode))
+               oi->i_type = OMFS_DIR;
+       else if (S_ISREG(inode->i_mode))
+               oi->i_type = OMFS_FILE;
+       else {
+               printk(KERN_WARNING "omfs: unknown file type: %d\n",
+                       inode->i_mode);
+               goto out_brelse;
+       }
+
+       oi->i_head.h_body_size = cpu_to_be32(sbi->s_sys_blocksize -
+               sizeof(struct omfs_header));
+       oi->i_head.h_version = 1;
+       oi->i_head.h_type = OMFS_INODE_NORMAL;
+       oi->i_head.h_magic = OMFS_IMAGIC;
+       oi->i_size = cpu_to_be64(inode->i_size);
+
+       ctime = inode->i_ctime.tv_sec * 1000LL +
+               ((inode->i_ctime.tv_nsec + 999)/1000);
+       oi->i_ctime = cpu_to_be64(ctime);
+
+       omfs_update_checksums(oi);
+
+       mark_buffer_dirty(bh);
+       if (wait) {
+               sync_dirty_buffer(bh);
+               if (buffer_req(bh) && !buffer_uptodate(bh))
+                       sync_failed = 1;
+       }
+
+       /* if mirroring writes, copy to next fsblock */
+       for (i = 1; i < sbi->s_mirrors; i++) {
+               bh2 = sb_bread(inode->i_sb, block + i *
+                       (sbi->s_blocksize / sbi->s_sys_blocksize));
+               if (!bh2)
+                       goto out_brelse;
+
+               memcpy(bh2->b_data, bh->b_data, bh->b_size);
+               mark_buffer_dirty(bh2);
+               if (wait) {
+                       sync_dirty_buffer(bh2);
+                       if (buffer_req(bh2) && !buffer_uptodate(bh2))
+                               sync_failed = 1;
+               }
+               brelse(bh2);
+       }
+       ret = (sync_failed) ? -EIO : 0;
+out_brelse:
+       brelse(bh);
+out:
+       return ret;
+}
+
+int omfs_sync_inode(struct inode *inode)
+{
+       return omfs_write_inode(inode, 1);
+}
+
+/*
+ * called when an entry is deleted, need to clear the bits in the
+ * bitmaps.
+ */
+static void omfs_delete_inode(struct inode *inode)
+{
+       truncate_inode_pages(&inode->i_data, 0);
+
+       if (S_ISREG(inode->i_mode)) {
+               inode->i_size = 0;
+               omfs_shrink_inode(inode);
+       }
+
+       omfs_clear_range(inode->i_sb, inode->i_ino, 2);
+       clear_inode(inode);
+}
+
+struct inode *omfs_iget(struct super_block *sb, ino_t ino)
+{
+       struct omfs_sb_info *sbi = OMFS_SB(sb);
+       struct omfs_inode *oi;
+       struct buffer_head *bh;
+       unsigned int block;
+       u64 ctime;
+       unsigned long nsecs;
+       struct inode *inode;
+
+       inode = iget_locked(sb, ino);
+       if (!inode)
+               return ERR_PTR(-ENOMEM);
+       if (!(inode->i_state & I_NEW))
+               return inode;
+
+       block = clus_to_blk(sbi, ino);
+       bh = sb_bread(inode->i_sb, block);
+       if (!bh)
+               goto iget_failed;
+
+       oi = (struct omfs_inode *)bh->b_data;
+
+       /* check self */
+       if (ino != be64_to_cpu(oi->i_head.h_self))
+               goto fail_bh;
+
+       inode->i_uid = sbi->s_uid;
+       inode->i_gid = sbi->s_gid;
+
+       ctime = be64_to_cpu(oi->i_ctime);
+       nsecs = do_div(ctime, 1000) * 1000L;
+
+       inode->i_atime.tv_sec = ctime;
+       inode->i_mtime.tv_sec = ctime;
+       inode->i_ctime.tv_sec = ctime;
+       inode->i_atime.tv_nsec = nsecs;
+       inode->i_mtime.tv_nsec = nsecs;
+       inode->i_ctime.tv_nsec = nsecs;
+
+       inode->i_mapping->a_ops = &omfs_aops;
+
+       switch (oi->i_type) {
+       case OMFS_DIR:
+               inode->i_mode = S_IFDIR | (S_IRWXUGO & ~sbi->s_dmask);
+               inode->i_op = &omfs_dir_inops;
+               inode->i_fop = &omfs_dir_operations;
+               inode->i_size = be32_to_cpu(oi->i_head.h_body_size) +
+                       sizeof(struct omfs_header);
+               inc_nlink(inode);
+               break;
+       case OMFS_FILE:
+               inode->i_mode = S_IFREG | (S_IRWXUGO & ~sbi->s_fmask);
+               inode->i_fop = &omfs_file_operations;
+               inode->i_size = be64_to_cpu(oi->i_size);
+               break;
+       }
+       brelse(bh);
+       unlock_new_inode(inode);
+       return inode;
+fail_bh:
+       brelse(bh);
+iget_failed:
+       iget_failed(inode);
+       return ERR_PTR(-EIO);
+}
+
+static void omfs_put_super(struct super_block *sb)
+{
+       struct omfs_sb_info *sbi = OMFS_SB(sb);
+       kfree(sbi->s_imap);
+       kfree(sbi);
+       sb->s_fs_info = NULL;
+}
+
+static int omfs_statfs(struct dentry *dentry, struct kstatfs *buf)
+{
+       struct super_block *s = dentry->d_sb;
+       struct omfs_sb_info *sbi = OMFS_SB(s);
+       buf->f_type = OMFS_MAGIC;
+       buf->f_bsize = sbi->s_blocksize;
+       buf->f_blocks = sbi->s_num_blocks;
+       buf->f_files = sbi->s_num_blocks;
+       buf->f_namelen = OMFS_NAMELEN;
+
+       buf->f_bfree = buf->f_bavail = buf->f_ffree =
+               omfs_count_free(s);
+       return 0;
+}
+
+static struct super_operations omfs_sops = {
+       .write_inode    = omfs_write_inode,
+       .delete_inode   = omfs_delete_inode,
+       .put_super      = omfs_put_super,
+       .statfs         = omfs_statfs,
+       .show_options   = generic_show_options,
+};
+
+/*
+ * For Rio Karma, there is an on-disk free bitmap whose location is
+ * stored in the root block.  For ReplayTV, there is no such free bitmap
+ * so we have to walk the tree.  Both inodes and file data are allocated
+ * from the same map.  This array can be big (300k) so we allocate
+ * in units of the blocksize.
+ */
+static int omfs_get_imap(struct super_block *sb)
+{
+       int bitmap_size;
+       int array_size;
+       int count;
+       struct omfs_sb_info *sbi = OMFS_SB(sb);
+       struct buffer_head *bh;
+       unsigned long **ptr;
+       sector_t block;
+
+       bitmap_size = DIV_ROUND_UP(sbi->s_num_blocks, 8);
+       array_size = DIV_ROUND_UP(bitmap_size, sb->s_blocksize);
+
+       if (sbi->s_bitmap_ino == ~0ULL)
+               goto out;
+
+       sbi->s_imap_size = array_size;
+       sbi->s_imap = kzalloc(array_size * sizeof(unsigned long *), GFP_KERNEL);
+       if (!sbi->s_imap)
+               goto nomem;
+
+       block = clus_to_blk(sbi, sbi->s_bitmap_ino);
+       ptr = sbi->s_imap;
+       for (count = bitmap_size; count > 0; count -= sb->s_blocksize) {
+               bh = sb_bread(sb, block++);
+               if (!bh)
+                       goto nomem_free;
+               *ptr = kmalloc(sb->s_blocksize, GFP_KERNEL);
+               if (!*ptr) {
+                       brelse(bh);
+                       goto nomem_free;
+               }
+               memcpy(*ptr, bh->b_data, sb->s_blocksize);
+               if (count < sb->s_blocksize)
+                       memset((void *)*ptr + count, 0xff,
+                               sb->s_blocksize - count);
+               brelse(bh);
+               ptr++;
+       }
+out:
+       return 0;
+
+nomem_free:
+       for (count = 0; count < array_size; count++)
+               kfree(sbi->s_imap[count]);
+
+       kfree(sbi->s_imap);
+nomem:
+       sbi->s_imap = NULL;
+       sbi->s_imap_size = 0;
+       return -ENOMEM;
+}
+
+enum {
+       Opt_uid, Opt_gid, Opt_umask, Opt_dmask, Opt_fmask
+};
+
+static match_table_t tokens = {
+       {Opt_uid, "uid=%u"},
+       {Opt_gid, "gid=%u"},
+       {Opt_umask, "umask=%o"},
+       {Opt_dmask, "dmask=%o"},
+       {Opt_fmask, "fmask=%o"},
+};
+
+static int parse_options(char *options, struct omfs_sb_info *sbi)
+{
+       char *p;
+       substring_t args[MAX_OPT_ARGS];
+       int option;
+
+       if (!options)
+               return 1;
+
+       while ((p = strsep(&options, ",")) != NULL) {
+               int token;
+               if (!*p)
+                       continue;
+
+               token = match_token(p, tokens, args);
+               switch (token) {
+               case Opt_uid:
+                       if (match_int(&args[0], &option))
+                               return 0;
+                       sbi->s_uid = option;
+                       break;
+               case Opt_gid:
+                       if (match_int(&args[0], &option))
+                               return 0;
+                       sbi->s_gid = option;
+                       break;
+               case Opt_umask:
+                       if (match_octal(&args[0], &option))
+                               return 0;
+                       sbi->s_fmask = sbi->s_dmask = option;
+                       break;
+               case Opt_dmask:
+                       if (match_octal(&args[0], &option))
+                               return 0;
+                       sbi->s_dmask = option;
+                       break;
+               case Opt_fmask:
+                       if (match_octal(&args[0], &option))
+                               return 0;
+                       sbi->s_fmask = option;
+                       break;
+               default:
+                       return 0;
+               }
+       }
+       return 1;
+}
+
+static int omfs_fill_super(struct super_block *sb, void *data, int silent)
+{
+       struct buffer_head *bh, *bh2;
+       struct omfs_super_block *omfs_sb;
+       struct omfs_root_block *omfs_rb;
+       struct omfs_sb_info *sbi;
+       struct inode *root;
+       sector_t start;
+       int ret = -EINVAL;
+
+       save_mount_options(sb, (char *) data);
+
+       sbi = kzalloc(sizeof(struct omfs_sb_info), GFP_KERNEL);
+       if (!sbi)
+               return -ENOMEM;
+
+       sb->s_fs_info = sbi;
+
+       sbi->s_uid = current->uid;
+       sbi->s_gid = current->gid;
+       sbi->s_dmask = sbi->s_fmask = current->fs->umask;
+
+       if (!parse_options((char *) data, sbi))
+               goto end;
+
+       sb->s_maxbytes = 0xffffffff;
+
+       sb_set_blocksize(sb, 0x200);
+
+       bh = sb_bread(sb, 0);
+       if (!bh)
+               goto end;
+
+       omfs_sb = (struct omfs_super_block *)bh->b_data;
+
+       if (omfs_sb->s_magic != cpu_to_be32(OMFS_MAGIC)) {
+               if (!silent)
+                       printk(KERN_ERR "omfs: Invalid superblock (%x)\n",
+                                  omfs_sb->s_magic);
+               goto out_brelse_bh;
+       }
+       sb->s_magic = OMFS_MAGIC;
+
+       sbi->s_num_blocks = be64_to_cpu(omfs_sb->s_num_blocks);
+       sbi->s_blocksize = be32_to_cpu(omfs_sb->s_blocksize);
+       sbi->s_mirrors = be32_to_cpu(omfs_sb->s_mirrors);
+       sbi->s_root_ino = be64_to_cpu(omfs_sb->s_root_block);
+       sbi->s_sys_blocksize = be32_to_cpu(omfs_sb->s_sys_blocksize);
+       mutex_init(&sbi->s_bitmap_lock);
+
+       if (sbi->s_sys_blocksize > PAGE_SIZE) {
+               printk(KERN_ERR "omfs: sysblock size (%d) is out of range\n",
+                       sbi->s_sys_blocksize);
+               goto out_brelse_bh;
+       }
+
+       if (sbi->s_blocksize < sbi->s_sys_blocksize ||
+           sbi->s_blocksize > OMFS_MAX_BLOCK_SIZE) {
+               printk(KERN_ERR "omfs: block size (%d) is out of range\n",
+                       sbi->s_blocksize);
+               goto out_brelse_bh;
+       }
+
+       /*
+        * Use sys_blocksize as the fs block since it is smaller than a
+        * page while the fs blocksize can be larger.
+        */
+       sb_set_blocksize(sb, sbi->s_sys_blocksize);
+
+       /*
+        * ...and the difference goes into a shift.  sys_blocksize is always
+        * a power of two factor of blocksize.
+        */
+       sbi->s_block_shift = get_bitmask_order(sbi->s_blocksize) -
+               get_bitmask_order(sbi->s_sys_blocksize);
+
+       start = clus_to_blk(sbi, be64_to_cpu(omfs_sb->s_root_block));
+       bh2 = sb_bread(sb, start);
+       if (!bh2)
+               goto out_brelse_bh;
+
+       omfs_rb = (struct omfs_root_block *)bh2->b_data;
+
+       sbi->s_bitmap_ino = be64_to_cpu(omfs_rb->r_bitmap);
+       sbi->s_clustersize = be32_to_cpu(omfs_rb->r_clustersize);
+
+       if (sbi->s_num_blocks != be64_to_cpu(omfs_rb->r_num_blocks)) {
+               printk(KERN_ERR "omfs: block count discrepancy between "
+                       "super and root blocks (%llx, %llx)\n",
+                       sbi->s_num_blocks, be64_to_cpu(omfs_rb->r_num_blocks));
+               goto out_brelse_bh2;
+       }
+
+       ret = omfs_get_imap(sb);
+       if (ret)
+               goto out_brelse_bh2;
+
+       sb->s_op = &omfs_sops;
+
+       root = omfs_iget(sb, be64_to_cpu(omfs_rb->r_root_dir));
+       if (IS_ERR(root)) {
+               ret = PTR_ERR(root);
+               goto out_brelse_bh2;
+       }
+
+       sb->s_root = d_alloc_root(root);
+       if (!sb->s_root) {
+               iput(root);
+               goto out_brelse_bh2;
+       }
+       printk(KERN_DEBUG "omfs: Mounted volume %s\n", omfs_rb->r_name);
+
+       ret = 0;
+out_brelse_bh2:
+       brelse(bh2);
+out_brelse_bh:
+       brelse(bh);
+end:
+       return ret;
+}
+
+static int omfs_get_sb(struct file_system_type *fs_type,
+                       int flags, const char *dev_name,
+                       void *data, struct vfsmount *m)
+{
+       return get_sb_bdev(fs_type, flags, dev_name, data, omfs_fill_super, m);
+}
+
+static struct file_system_type omfs_fs_type = {
+       .owner = THIS_MODULE,
+       .name = "omfs",
+       .get_sb = omfs_get_sb,
+       .kill_sb = kill_block_super,
+       .fs_flags = FS_REQUIRES_DEV,
+};
+
+static int __init init_omfs_fs(void)
+{
+       return register_filesystem(&omfs_fs_type);
+}
+
+static void __exit exit_omfs_fs(void)
+{
+       unregister_filesystem(&omfs_fs_type);
+}
+
+module_init(init_omfs_fs);
+module_exit(exit_omfs_fs);