Merge with 2.3.19.

1 files changed, 88 insertions, 363 deletions

diff --git a/fs/fat/file.c b/fs/fat/file.c
index 5d964da2b..19b91b74f 100644
--- a/fs/fat/file.c
+++ b/fs/fat/file.c
@@ -57,7 +57,7 @@ struct inode_operations fat_file_inode_operations = {
 	NULL,			/* rename */
 	NULL,			/* readlink */
 	NULL,			/* follow_link */
-	fat_bmap,		/* get_block */
+	fat_get_block,		/* get_block */
 	block_read_full_page,	/* readpage */
 	NULL,			/* writepage */
 	NULL,			/* flushpage */
@@ -67,392 +67,117 @@ struct inode_operations fat_file_inode_operations = {
 	NULL			/* revalidate */
 };
 
-/* #Specification: msdos / special devices / mmap	
-	Mmapping does work because a special mmap is provide in that case.
-	Note that it is much less efficient than the generic_file_mmap normally
-	used since it allocate extra buffer. generic_file_mmap is used for
-	normal device (512 bytes hardware sectors).
-*/
-static struct file_operations fat_file_operations_1024 = {
-	NULL,			/* lseek - default */
-	fat_file_read,		/* read */
-	fat_file_write,		/* write */
-	NULL,			/* readdir - bad */
-	NULL,			/* select v2.0.x/poll v2.1.x - default */
-	NULL,			/* ioctl - default */
-	fat_mmap,		/* mmap */
-	NULL,			/* no special open is needed */
-	NULL,			/* flush */
-	NULL,			/* release */
-	file_fsync		/* fsync */
-};
-
-/* #Specification: msdos / special devices / swap file
-	Swap file can't work on special devices with a large sector
-	size (1024 bytes hard sector). Those devices have a weird
-	MS-DOS filesystem layout. Generally a single hardware sector
-	may contain 2 unrelated logical sector. This mean that there is
-	no easy way to do a mapping between disk sector of a file and virtual
-	memory. So swap file is difficult (not available right now)
-	on those devices. Off course, Ext2 does not have this problem.
-*/
-struct inode_operations fat_file_inode_operations_1024 = {
-	&fat_file_operations_1024,	/* default file operations */
-	NULL,			/* create */
-	NULL,			/* lookup */
-	NULL,			/* link */
-	NULL,			/* unlink */
-	NULL,			/* symlink */
-	NULL,			/* mkdir */
-	NULL,			/* rmdir */
-	NULL,			/* mknod */
-	NULL,			/* rename */
-	NULL,			/* readlink */
-	NULL,			/* follow_link */
-	NULL,			/* get_block */
-	block_read_full_page,	/* readpage */
-	NULL,			/* writepage */
-	NULL,			/* flushpage */
-	fat_truncate,		/* truncate */
-	NULL,			/* permission */
-	NULL,			/* smap */
-	NULL			/* revalidate */
-};
-
-static struct file_operations fat_file_operations_readpage = {
-	NULL,			/* lseek - default */
-	fat_file_read,		/* read */
-	fat_file_write,		/* write */
-	NULL,			/* readdir - bad */
-	NULL,			/* select v2.0.x/poll v2.1.x - default */
-	NULL,			/* ioctl - default */
-	generic_file_mmap,	/* mmap */
-	NULL,			/* no special open is needed */
-	NULL,			/* flush */
-	NULL,			/* release */
-	file_fsync		/* fsync */
-};
-
-struct inode_operations fat_file_inode_operations_readpage = {
-	&fat_file_operations_readpage,	/* default file operations */
-	NULL,			/* create */
-	NULL,			/* lookup */
-	NULL,			/* link */
-	NULL,			/* unlink */
-	NULL,			/* symlink */
-	NULL,			/* mkdir */
-	NULL,			/* rmdir */
-	NULL,			/* mknod */
-	NULL,			/* rename */
-	NULL,			/* readlink */
-	NULL,			/* follow_link */
-	NULL,			/* get_block */
-	fat_readpage,		/* readpage */
-	NULL,			/* writepage */
-	NULL,			/* flushpage */
-	fat_truncate,		/* truncate */
-	NULL,			/* permission */
-	NULL,			/* smap */
-	NULL			/* revalidate */
-};
-
-#define MSDOS_PREFETCH	32
-struct fat_pre {
-	int file_sector;/* Next sector to read in the prefetch table */
-			/* This is relative to the file, not the disk */
-	struct buffer_head *bhlist[MSDOS_PREFETCH];	/* All buffers needed */
-	int nblist;	/* Number of buffers in bhlist */
-	int nolist;	/* index in bhlist */
-};
-/*
-	Order the prefetch of more sectors.
-*/
-static void fat_prefetch (
-	struct inode *inode,
-	struct fat_pre *pre,
-	int nb)		/* How many must we prefetch at once */
-{
-	struct super_block *sb = inode->i_sb;
-	struct buffer_head *bhreq[MSDOS_PREFETCH];	/* Buffers not */
-							/* already read */
-	int nbreq = 0;			/* Number of buffers in bhreq */
-	int i;
-	for (i=0; i<nb; i++){
-		int sector = fat_smap(inode,pre->file_sector);
-		if (sector != 0){
-			struct buffer_head *bh;
-			PRINTK (("fsector2 %d -> %d\n",pre->file_sector-1,sector));
-			pre->file_sector++;
-			bh = fat_getblk(sb, sector);
-			if (bh == NULL)	break;
-			pre->bhlist[pre->nblist++] = bh;
-			if (!fat_is_uptodate(sb,bh))
-				bhreq[nbreq++] = bh;
-		}else{
-			break;
-		}
-	}
-	if (nbreq > 0) fat_ll_rw_block (sb,READ,nbreq,bhreq);
-	for (i=pre->nblist; i<MSDOS_PREFETCH; i++) pre->bhlist[i] = NULL;
-}
-
-/*
-	Read a file into user space
-*/
-static ssize_t fat_file_read_text(
+ssize_t fat_file_read(
 	struct file *filp,
 	char *buf,
 	size_t count,
 	loff_t *ppos)
 {
 	struct inode *inode = filp->f_dentry->d_inode;
-	struct super_block *sb = inode->i_sb;
-	char *start = buf;
-	char *end   = buf + count;
-	int i;
-	int left_in_file;
-	struct fat_pre pre;
-		
+	return MSDOS_SB(inode->i_sb)->cvf_format
+			->cvf_file_read(filp,buf,count,ppos);
+}
 
-	if (!inode) {
-		printk("fat_file_read: inode = NULL\n");
-		return -EINVAL;
+
+int fat_get_block(struct inode *inode, long iblock, struct buffer_head *bh_result, int create) {
+	unsigned long phys;
+	phys = fat_bmap(inode, iblock);
+	if (phys) {
+		bh_result->b_dev = inode->i_dev;
+		bh_result->b_blocknr = phys;
+		bh_result->b_state |= (1UL << BH_Mapped);
+		return 0;
 	}
-	/* S_ISLNK allows for UMSDOS. Should never happen for normal MSDOS */
-	if (!S_ISREG(inode->i_mode) && !S_ISLNK(inode->i_mode)) {
-		printk("fat_file_read: mode = %07o\n",inode->i_mode);
-		return -EINVAL;
+	if (!create)
+		return 0;
+	if (iblock<<9 != MSDOS_I(inode)->i_realsize) {
+		BUG();
+		return -EIO;
 	}
-	if (*ppos >= inode->i_size || count == 0) return 0;
-	/*
-		Tell the buffer cache which block we expect to read in advance
-		Since we are limited with the stack, we preread only MSDOS_PREFETCH
-		because we have to keep the result into the local
-		arrays pre.bhlist and bhreq.
-		
-		Each time we process one block in bhlist, we replace
-		it by a new prefetch block if needed.
-	*/
-	PRINTK (("#### ino %ld pos %ld size %ld count %d\n",inode->i_ino,*ppos,inode->i_size,count));
-	{
-		/*
-			We must prefetch complete block, so we must
-			take in account the offset in the first block.
-		*/
-		int count_max = (*ppos & (SECTOR_SIZE-1)) + count;
-		int   to_reada;	/* How many block to read all at once */
-		pre.file_sector = *ppos >> SECTOR_BITS;
-		to_reada = count_max / SECTOR_SIZE;
-		if (count_max & (SECTOR_SIZE-1)) to_reada++;
-		if (filp->f_reada || !MSDOS_I(inode)->i_binary){
-			/* Doing a read ahead on ASCII file make sure we always */
-			/* read enough, since we don't know how many blocks */
-			/* we really need */
-			int ahead = read_ahead[MAJOR(inode->i_dev)];
-			PRINTK (("to_reada %d ahead %d\n",to_reada,ahead));
-			if (ahead == 0) ahead = 8;
-			to_reada += ahead;
-		}
-		if (to_reada > MSDOS_PREFETCH) to_reada = MSDOS_PREFETCH;
-		pre.nblist = 0;
-		fat_prefetch (inode,&pre,to_reada);
+	if (!(iblock % MSDOS_SB(inode->i_sb)->cluster_size)) {
+		if (fat_add_cluster(inode))
+			return -ENOSPC;
 	}
-	pre.nolist = 0;
-	PRINTK (("count %d ahead %d nblist %d\n",count,read_ahead[MAJOR(inode->i_dev)],pre.nblist));
-	while ((left_in_file = inode->i_size - *ppos) > 0
-		&& buf < end){
-		struct buffer_head *bh = pre.bhlist[pre.nolist];
-		char *data;
-		int size,offset;
-		if (bh == NULL) break;
-		pre.bhlist[pre.nolist] = NULL;
-		pre.nolist++;
-		if (pre.nolist == MSDOS_PREFETCH/2){
-			memcpy (pre.bhlist,pre.bhlist+MSDOS_PREFETCH/2
-				,(MSDOS_PREFETCH/2)*sizeof(pre.bhlist[0]));
-			pre.nblist -= MSDOS_PREFETCH/2;
-			fat_prefetch (inode,&pre,MSDOS_PREFETCH/2);
-			pre.nolist = 0;
-		}
-		PRINTK (("file_read pos %ld nblist %d %d %d\n",*ppos,pre.nblist,pre.fetched,count));
-		wait_on_buffer(bh);
-		if (!fat_is_uptodate(sb,bh)){
-			/* read error  ? */
-			fat_brelse (sb, bh);
-			break;
-		}
-		offset = *ppos & (SECTOR_SIZE-1);
-		data = bh->b_data + offset;
-		size = MIN(SECTOR_SIZE-offset,left_in_file);
-		if (MSDOS_I(inode)->i_binary) {
-			size = MIN(size,end-buf);
-			copy_to_user(buf,data,size);
-			buf += size;
-			*ppos += size;
-		}else{
-			for (; size && buf < end; size--) {
-				char ch = *data++;
-				++*ppos;
-				if (ch == 26){
-					*ppos = inode->i_size;
-					break;
-				}else if (ch != '\r'){
-					put_user(ch,buf++);
-				}
+	MSDOS_I(inode)->i_realsize+=SECTOR_SIZE;
+	phys=fat_bmap(inode, iblock);
+	if (!phys)
+		BUG();
+	bh_result->b_dev = inode->i_dev;
+	bh_result->b_blocknr = phys;
+	bh_result->b_state |= (1UL << BH_Mapped);
+	bh_result->b_state |= (1UL << BH_New);
+	return 0;
+}
+
+static int fat_write_partial_page(struct file *file, struct page *page, unsigned long offset, unsigned long bytes, const char * buf)
+{
+	struct dentry *dentry = file->f_dentry;
+	struct inode *inode = dentry->d_inode;
+	struct page *new_page, **hash;
+	unsigned long pgpos;
+	unsigned long page_cache = 0;
+	long status;
+
+	pgpos = inode->i_size & PAGE_CACHE_MASK;
+	while (pgpos < page->offset) {
+		hash = page_hash(inode, pgpos);
+repeat_find:	new_page = __find_lock_page(inode, pgpos, hash);
+		if (!new_page) {
+			if (!page_cache) {
+				page_cache = page_cache_alloc();
+				if (page_cache)
+					goto repeat_find;
+				status = -ENOMEM;
+				goto out;
 			}
+			new_page = page_cache_entry(page_cache);
+			if (add_to_page_cache_unique(new_page,inode,pgpos,hash))
+				goto repeat_find;
+			page_cache = 0;
 		}
-		fat_brelse(sb, bh);
+		status = block_write_cont_page(file, new_page, PAGE_SIZE, 0, NULL);
+		UnlockPage(new_page);
+		page_cache_release(new_page);
+		if (status < 0)
+			goto out;
+		pgpos = MSDOS_I(inode)->i_realsize & PAGE_CACHE_MASK;
 	}
-	PRINTK (("--- %d -> %d\n",count,(int)(buf-start)));
-	for (i=0; i<pre.nblist; i++)
-		fat_brelse (sb, pre.bhlist[i]);
-	if (start == buf)
-		return -EIO;
-	if (!IS_RDONLY(inode))
-		inode->i_atime = CURRENT_TIME;
-	filp->f_reada = 1;	/* Will be reset if a lseek is done */
-	return buf-start;
+	status = block_write_cont_page(file, page, offset, bytes, buf);
+out:
+	if (page_cache)
+		page_cache_free(page_cache);
+	return status;
 }
 
-ssize_t fat_file_read(
+ssize_t fat_file_write(
 	struct file *filp,
-	char *buf,
+	const char *buf,
 	size_t count,
 	loff_t *ppos)
 {
 	struct inode *inode = filp->f_dentry->d_inode;
-	if (MSDOS_SB(inode->i_sb)->cvf_format &&
-	    MSDOS_SB(inode->i_sb)->cvf_format->cvf_file_read)
-		return MSDOS_SB(inode->i_sb)->cvf_format
-			->cvf_file_read(filp,buf,count,ppos);
-
-	/*
-	 * MS-DOS filesystems with a blocksize > 512 may have blocks
-	 * spread over several hardware sectors (unaligned), which
-	 * is not something the generic routines can (or would want
-	 * to) handle).
-	 */
-	if (!MSDOS_I(inode)->i_binary || inode->i_sb->s_blocksize > 512)
-		return fat_file_read_text(filp, buf, count, ppos);
-	return generic_file_read(filp, buf, count, ppos);
+	struct super_block *sb = inode->i_sb;
+	return MSDOS_SB(sb)->cvf_format
+			->cvf_file_write(filp,buf,count,ppos);
 }
-/*
-	Write to a file either from user space
-*/
-ssize_t fat_file_write(
+
+ssize_t default_fat_file_write(
 	struct file *filp,
 	const char *buf,
 	size_t count,
 	loff_t *ppos)
 {
 	struct inode *inode = filp->f_dentry->d_inode;
-	struct super_block *sb = inode->i_sb;
-	int sector,offset,size,left,written;
-	int error,carry;
-	const char *start;
-	char *to,ch;
-	struct buffer_head *bh;
-	int binary_mode = MSDOS_I(inode)->i_binary;
-
-	PRINTK(("fat_file_write: dentry=%p, inode=%p, ino=%ld\n",
-		filp->f_dentry, inode, inode->i_ino));
-	if (!inode) {
-		printk("fat_file_write: inode = NULL\n");
-		return -EINVAL;
-	}
-        if (MSDOS_SB(sb)->cvf_format &&
-	    MSDOS_SB(sb)->cvf_format->cvf_file_write)
-		return MSDOS_SB(sb)->cvf_format
-			->cvf_file_write(filp,buf,count,ppos);
-
-	/* S_ISLNK allows for UMSDOS. Should never happen for normal MSDOS */
-	if (!S_ISREG(inode->i_mode) && !S_ISLNK(inode->i_mode)) {
-		printk("fat_file_write: mode = %07o\n",inode->i_mode);
-		return -EINVAL;
-	}
-	/* system files may be immutable */
-	if (IS_IMMUTABLE(inode))
-		return -EPERM;
-/*
- * ok, append may not work when many processes are writing at the same time
- * but so what. That way leads to madness anyway.
- */
-	if (filp->f_flags & O_APPEND)
-		*ppos = inode->i_size;
-	if (count == 0)
-		return 0;
-	if (*ppos + count > 0x7FFFFFFFLL) {
-		count = 0x7FFFFFFFLL-*ppos;
-		if (!count)
-			return -EFBIG;
+	int retval;
+
+	retval = generic_file_write(filp, buf, count, ppos,
+					fat_write_partial_page);
+	if (retval > 0) {
+		inode->i_mtime = inode->i_ctime = CURRENT_TIME;
+		MSDOS_I(inode)->i_attrs |= ATTR_ARCH;
+		mark_inode_dirty(inode);
 	}
-
-	error = carry = 0;
-	for (start = buf; count || carry; count -= size) {
-		while (!(sector = fat_smap(inode,*ppos >> SECTOR_BITS)))
-			if ((error = fat_add_cluster(inode)) < 0) break;
-		if (error) {
-			fat_truncate(inode);
-			break;
-		}
-		offset = *ppos & (SECTOR_SIZE-1);
-		size = MIN(SECTOR_SIZE-offset,MAX(carry,count));
-		if (binary_mode
-			&& offset == 0
-			&& (size == SECTOR_SIZE
-				|| *ppos + size >= inode->i_size)){
-			/* No need to read the block first since we will */
-			/* completely overwrite it */
-			/* or at least write past the end of file */
-			if (!(bh = fat_getblk(sb,sector))){
-				error = -EIO;
-				break;
-			}
-		} else if (!(bh = fat_bread(sb,sector))) {
-			error = -EIO;
-			break;
-		}
-		if (binary_mode) {
-			copy_from_user(bh->b_data+offset,buf,written = size);
-			buf += size;
-		} else {
-			written = left = SECTOR_SIZE-offset;
-			to = (char *) bh->b_data+(*ppos & (SECTOR_SIZE-1));
-			if (carry) {
-				*to++ = '\n';
-				left--;
-				carry = 0;
-			}
-			for (size = 0; size < count && left; size++) {
-				get_user(ch, buf++);
-				if (ch == '\n') {
-					*to++ = '\r';
-					left--;
-				}
-				if (!left) carry = 1;
-				else {
-					*to++ = ch;
-					left--;
-				}
-			}
-			written -= left;
-		}
-		update_vm_cache(inode, *ppos, bh->b_data + (*ppos & (SECTOR_SIZE-1)), written);
-		*ppos += written;
-		if (*ppos > inode->i_size) {
-			inode->i_size = *ppos;
-			mark_inode_dirty(inode);
-		}
-		fat_set_uptodate(sb, bh, 1);
-		fat_mark_buffer_dirty(sb, bh, 0);
-		fat_brelse(sb, bh);
-	}
-	if (start == buf)
-		return error;
-	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
-	MSDOS_I(inode)->i_attrs |= ATTR_ARCH;
-	mark_inode_dirty(inode);
-	return buf-start;
+	return retval;
 }
 
 void fat_truncate(struct inode *inode)
@@ -460,14 +185,14 @@ void fat_truncate(struct inode *inode)
 	int cluster;
 
 	/* Why no return value?  Surely the disk could fail... */
+	if (IS_RDONLY (inode))
+		return /* -EPERM */;
 	if (IS_IMMUTABLE(inode))
 		return /* -EPERM */;
-	if(inode->i_sb->s_flags&MS_RDONLY) {
-		printk("FAT: fat_truncate called though fs is read-only, uhh...\n");
-		return /* -EROFS */;
-	}
 	cluster = SECTOR_SIZE*MSDOS_SB(inode->i_sb)->cluster_size;
+	MSDOS_I(inode)->i_realsize = ((inode->i_size-1) | (SECTOR_SIZE-1)) + 1;
 	(void) fat_free(inode,(inode->i_size+(cluster-1))/cluster);
 	MSDOS_I(inode)->i_attrs |= ATTR_ARCH;
+	inode->i_ctime = inode->i_mtime = CURRENT_TIME;
 	mark_inode_dirty(inode);
 }