/*
* linux/fs/hfs/inode.c
*
* Copyright (C) 1995-1997 Paul H. Hargrove
* This file may be distributed under the terms of the GNU General Public License.
*
* This file contains inode-related functions which do not depend on
* which scheme is being used to represent forks.
*
* Based on the minix file system code, (C) 1991, 1992 by Linus Torvalds
*
* "XXX" in a comment is a note to myself to consider changing something.
*
* In function preconditions the term "valid" applied to a pointer to
* a structure means that the pointer is non-NULL and the structure it
* points to has all fields initialized to consistent values.
*/
#include "hfs.h"
#include <linux/hfs_fs_sb.h>
#include <linux/hfs_fs_i.h>
#include <linux/hfs_fs.h>
#include <linux/smp_lock.h>
/*================ Variable-like macros ================*/
#define HFS_VALID_MODE_BITS (S_IFREG | S_IFDIR | S_IRWXUGO)
/*================ File-local functions ================*/
/*
* init_file_inode()
*
* Given an HFS catalog entry initialize an inode for a file.
*/
static void init_file_inode(struct inode *inode, hfs_u8 fork)
{
struct hfs_fork *fk;
struct hfs_cat_entry *entry = HFS_I(inode)->entry;
if (fork == HFS_FK_DATA) {
inode->i_mode = S_IRWXUGO | S_IFREG;
} else {
inode->i_mode = S_IRUGO | S_IWUGO | S_IFREG;
}
if (fork == HFS_FK_DATA) {
#if 0 /* XXX: disable crlf translations for now */
hfs_u32 type = hfs_get_nl(entry->info.file.finfo.fdType);
HFS_I(inode)->convert =
((HFS_SB(inode->i_sb)->s_conv == 't') ||
((HFS_SB(inode->i_sb)->s_conv == 'a') &&
((type == htonl(0x54455854)) || /* "TEXT" */
(type == htonl(0x7474726f))))); /* "ttro" */
#else
HFS_I(inode)->convert = 0;
#endif
fk = &entry->u.file.data_fork;
} else {
fk = &entry->u.file.rsrc_fork;
HFS_I(inode)->convert = 0;
}
HFS_I(inode)->fork = fk;
inode->i_size = fk->lsize;
inode->i_blocks = fk->psize;
inode->i_nlink = 1;
}
/*================ Global functions ================*/
/*
* hfs_put_inode()
*
* This is the put_inode() entry in the super_operations for HFS
* filesystems. The purpose is to perform any filesystem-dependent
* cleanup necessary when the use-count of an inode falls to zero.
*/
void hfs_put_inode(struct inode * inode)
{
struct hfs_cat_entry *entry = HFS_I(inode)->entry;
lock_kernel();
hfs_cat_put(entry);
if (atomic_read(&inode->i_count) == 1) {
struct hfs_hdr_layout *tmp = HFS_I(inode)->layout;
if (tmp) {
HFS_I(inode)->layout = NULL;
HFS_DELETE(tmp);
}
}
unlock_kernel();
}
/*
* hfs_notify_change()
*
* Based very closely on fs/msdos/inode.c by Werner Almesberger
*
* This is the notify_change() field in the super_operations structure
* for HFS file systems. The purpose is to take that changes made to
* an inode and apply then in a filesystem-dependent manner. In this
* case the process has a few of tasks to do:
* 1) prevent changes to the i_uid and i_gid fields.
* 2) map file permissions to the closest allowable permissions
* 3) Since multiple Linux files can share the same on-disk inode under
* HFS (for instance the data and resource forks of a file) a change
* to permissions must be applied to all other in-core inodes which
* correspond to the same HFS file.
*/
enum {HFS_NORM, HFS_HDR, HFS_CAP};
static int __hfs_notify_change(struct dentry *dentry, struct iattr * attr, int kind)
{
struct inode *inode = dentry->d_inode;
struct hfs_cat_entry *entry = HFS_I(inode)->entry;
struct dentry **de = entry->sys_entry;
struct hfs_sb_info *hsb = HFS_SB(inode->i_sb);
int error, i;
error = inode_change_ok(inode, attr); /* basic permission checks */
if (error) {
/* Let netatalk's afpd think chmod() always succeeds */
if (hsb->s_afpd &&
(attr->ia_valid == (ATTR_MODE | ATTR_CTIME))) {
return 0;
} else {
return error;
}
}
/* no uig/gid changes and limit which mode bits can be set */
if (((attr->ia_valid & ATTR_UID) &&
(attr->ia_uid != hsb->s_uid)) ||
((attr->ia_valid & ATTR_GID) &&
(attr->ia_gid != hsb->s_gid)) ||
((attr->ia_valid & ATTR_MODE) &&
(((entry->type == HFS_CDR_DIR) &&
(attr->ia_mode != inode->i_mode))||
(attr->ia_mode & ~HFS_VALID_MODE_BITS)))) {
return hsb->s_quiet ? 0 : error;
}
if (entry->type == HFS_CDR_DIR) {
attr->ia_valid &= ~ATTR_MODE;
} else if (attr->ia_valid & ATTR_MODE) {
/* Only the 'w' bits can ever change and only all together. */
if (attr->ia_mode & S_IWUSR) {
attr->ia_mode = inode->i_mode | S_IWUGO;
} else {
attr->ia_mode = inode->i_mode & ~S_IWUGO;
}
attr->ia_mode &= ~hsb->s_umask;
}
/*
* Normal files handle size change in normal way.
* Oddballs are served here.
*/
if (attr->ia_valid & ATTR_SIZE) {
if (kind == HFS_CAP) {
inode->i_size = attr->ia_size;
if (inode->i_size > HFS_FORK_MAX)
inode->i_size = HFS_FORK_MAX;
mark_inode_dirty(inode);
attr->ia_valid &= ~ATTR_SIZE;
} else if (kind == HFS_HDR) {
hdr_truncate(inode, attr->ia_size);
attr->ia_valid &= ~ATTR_SIZE;
}
}
error = inode_setattr(inode, attr);
if (error)
return error;
/* We wouldn't want to mess with the sizes of the other fork */
attr->ia_valid &= ~ATTR_SIZE;
/* We must change all in-core inodes corresponding to this file. */
for (i = 0; i < 4; ++i) {
if (de[i] && (de[i] != dentry)) {
inode_setattr(de[i]->d_inode, attr);
}
}
/* Change the catalog entry if needed */
if (attr->ia_valid & ATTR_MTIME) {
entry->modify_date = hfs_u_to_mtime(inode->i_mtime);
hfs_cat_mark_dirty(entry);
}
if (attr->ia_valid & ATTR_MODE) {
hfs_u8 new_flags;
if (inode->i_mode & S_IWUSR) {
new_flags = entry->u.file.flags & ~HFS_FIL_LOCK;
} else {
new_flags = entry->u.file.flags | HFS_FIL_LOCK;
}
if (new_flags != entry->u.file.flags) {
entry->u.file.flags = new_flags;
hfs_cat_mark_dirty(entry);
}
}
/* size changes handled in hfs_extent_adj() */
return 0;
}
int hfs_notify_change(struct dentry *dentry, struct iattr * attr)
{
return __hfs_notify_change(dentry, attr, HFS_NORM);
}
int hfs_notify_change_cap(struct dentry *dentry, struct iattr * attr)
{
return __hfs_notify_change(dentry, attr, HFS_CAP);
}
int hfs_notify_change_hdr(struct dentry *dentry, struct iattr * attr)
{
return __hfs_notify_change(dentry, attr, HFS_HDR);
}
static int hfs_writepage(struct page *page)
{
return block_write_full_page(page,hfs_get_block);
}
static int hfs_readpage(struct file *file, struct page *page)
{
return block_read_full_page(page,hfs_get_block);
}
static int hfs_prepare_write(struct file *file, struct page *page, unsigned from, unsigned to)
{
return cont_prepare_write(page,from,to,hfs_get_block,
&page->mapping->host->u.hfs_i.mmu_private);
}
static int hfs_bmap(struct address_space *mapping, long block)
{
return generic_block_bmap(mapping,block,hfs_get_block);
}
struct address_space_operations hfs_aops = {
readpage: hfs_readpage,
writepage: hfs_writepage,
sync_page: block_sync_page,
prepare_write: hfs_prepare_write,
commit_write: generic_commit_write,
bmap: hfs_bmap
};
/*
* __hfs_iget()
*
* Given the MDB for a HFS filesystem, a 'key' and an 'entry' in
* the catalog B-tree and the 'type' of the desired file return the
* inode for that file/directory or NULL. Note that 'type' indicates
* whether we want the actual file or directory, or the corresponding
* metadata (AppleDouble header file or CAP metadata file).
*
* In an ideal world we could call iget() and would not need this
* function. However, since there is no way to even know the inode
* number until we've found the file/directory in the catalog B-tree
* that simply won't happen.
*
* The main idea here is to look in the catalog B-tree to get the
* vital info about the file or directory (including the file id which
* becomes the inode number) and then to call iget() and return the
* inode if it is complete. If it is not then we use the catalog
* entry to fill in the missing info, by calling the appropriate
* 'fillin' function. Note that these fillin functions are
* essentially hfs_*_read_inode() functions, but since there is no way
* to pass the catalog entry through iget() to such a read_inode()
* function, we have to call them after iget() returns an incomplete
* inode to us. This is pretty much the same problem faced in the NFS
* code, and pretty much the same solution. The SMB filesystem deals
* with this in a different way: by using the address of the
* kmalloc()'d space which holds the data as the inode number.
*
* XXX: Both this function and NFS's corresponding nfs_fhget() would
* benefit from a way to pass an additional (void *) through iget() to
* the VFS read_inode() function.
*
* this will hfs_cat_put() the entry if it fails.
*/
struct inode *hfs_iget(struct hfs_cat_entry *entry, ino_t type,
struct dentry *dentry)
{
struct dentry **sys_entry;
struct super_block *sb;
struct inode *inode;
if (!entry) {
return NULL;
}
/* If there are several processes all calling __iget() for
the same inode then they will all get the same one back.
The first one to return from __iget() will notice that the
i_mode field of the inode is blank and KNOW that it is
the first to return. Therefore, it will set the appropriate
'sys_entry' field in the entry and initialize the inode.
All the initialization must be done without sleeping,
or else other processes could end up using a partially
initialized inode. */
sb = entry->mdb->sys_mdb;
sys_entry = &entry->sys_entry[HFS_ITYPE_TO_INT(type)];
if (!(inode = iget(sb, ntohl(entry->cnid) | type))) {
hfs_cat_put(entry);
return NULL;
}
if (inode->i_dev != sb->s_dev) {
iput(inode); /* automatically does an hfs_cat_put */
inode = NULL;
} else if (!inode->i_mode || (*sys_entry == NULL)) {
/* Initialize the inode */
struct hfs_sb_info *hsb = HFS_SB(sb);
inode->i_rdev = 0;
inode->i_ctime = inode->i_atime = inode->i_mtime =
hfs_m_to_utime(entry->modify_date);
inode->i_blksize = HFS_SECTOR_SIZE;
inode->i_uid = hsb->s_uid;
inode->i_gid = hsb->s_gid;
memset(HFS_I(inode), 0, sizeof(struct hfs_inode_info));
HFS_I(inode)->magic = HFS_INO_MAGIC;
HFS_I(inode)->entry = entry;
HFS_I(inode)->tz_secondswest = hfs_to_utc(0);
hsb->s_ifill(inode, type, hsb->s_version);
if (!hsb->s_afpd && (entry->type == HFS_CDR_FIL) &&
(entry->u.file.flags & HFS_FIL_LOCK)) {
inode->i_mode &= ~S_IWUGO;
}
inode->i_mode &= ~hsb->s_umask;
if (!inode->i_mode) {
iput(inode); /* does an hfs_cat_put */
inode = NULL;
} else
*sys_entry = dentry; /* cache dentry */
}
return inode;
}
/*================ Scheme-specific functions ================*/
/*
* hfs_cap_ifill()
*
* This function serves the same purpose as a read_inode() function does
* in other filesystems. It is called by __hfs_iget() to fill in
* the missing fields of an uninitialized inode under the CAP scheme.
*/
void hfs_cap_ifill(struct inode * inode, ino_t type, const int version)
{
struct hfs_cat_entry *entry = HFS_I(inode)->entry;
HFS_I(inode)->d_drop_op = hfs_cap_drop_dentry;
if (type == HFS_CAP_FNDR) {
inode->i_size = sizeof(struct hfs_cap_info);
inode->i_blocks = 0;
inode->i_nlink = 1;
inode->i_mode = S_IRUGO | S_IWUGO | S_IFREG;
inode->i_op = &hfs_cap_info_inode_operations;
inode->i_fop = &hfs_cap_info_operations;
} else if (entry->type == HFS_CDR_FIL) {
init_file_inode(inode, (type == HFS_CAP_DATA) ?
HFS_FK_DATA : HFS_FK_RSRC);
inode->i_op = &hfs_file_inode_operations;
inode->i_fop = &hfs_file_operations;
inode->i_mapping->a_ops = &hfs_aops;
inode->u.hfs_i.mmu_private = inode->i_size;
} else { /* Directory */
struct hfs_dir *hdir = &entry->u.dir;
inode->i_blocks = 0;
inode->i_size = hdir->files + hdir->dirs + 5;
HFS_I(inode)->dir_size = 1;
if (type == HFS_CAP_NDIR) {
inode->i_mode = S_IRWXUGO | S_IFDIR;
inode->i_nlink = hdir->dirs + 4;
inode->i_op = &hfs_cap_ndir_inode_operations;
inode->i_fop = &hfs_cap_dir_operations;
HFS_I(inode)->file_type = HFS_CAP_NORM;
} else if (type == HFS_CAP_FDIR) {
inode->i_mode = S_IRUGO | S_IXUGO | S_IFDIR;
inode->i_nlink = 2;
inode->i_op = &hfs_cap_fdir_inode_operations;
inode->i_fop = &hfs_cap_dir_operations;
HFS_I(inode)->file_type = HFS_CAP_FNDR;
} else if (type == HFS_CAP_RDIR) {
inode->i_mode = S_IRUGO | S_IXUGO | S_IFDIR;
inode->i_nlink = 2;
inode->i_op = &hfs_cap_rdir_inode_operations;
inode->i_fop = &hfs_cap_dir_operations;
HFS_I(inode)->file_type = HFS_CAP_RSRC;
}
}
}
/*
* hfs_dbl_ifill()
*
* This function serves the same purpose as a read_inode() function does
* in other filesystems. It is called by __hfs_iget() to fill in
* the missing fields of an uninitialized inode under the AppleDouble
* scheme.
*/
void hfs_dbl_ifill(struct inode * inode, ino_t type, const int version)
{
struct hfs_cat_entry *entry = HFS_I(inode)->entry;
HFS_I(inode)->d_drop_op = hfs_dbl_drop_dentry;
if (type == HFS_DBL_HDR) {
if (entry->type == HFS_CDR_FIL) {
init_file_inode(inode, HFS_FK_RSRC);
inode->i_size += HFS_DBL_HDR_LEN;
HFS_I(inode)->default_layout = &hfs_dbl_fil_hdr_layout;
} else {
inode->i_size = HFS_DBL_HDR_LEN;
inode->i_mode = S_IRUGO | S_IWUGO | S_IFREG;
inode->i_nlink = 1;
HFS_I(inode)->default_layout = &hfs_dbl_dir_hdr_layout;
}
inode->i_op = &hfs_hdr_inode_operations;
inode->i_fop = &hfs_hdr_operations;
} else if (entry->type == HFS_CDR_FIL) {
init_file_inode(inode, HFS_FK_DATA);
inode->i_op = &hfs_file_inode_operations;
inode->i_fop = &hfs_file_operations;
inode->i_mapping->a_ops = &hfs_aops;
inode->u.hfs_i.mmu_private = inode->i_size;
} else { /* Directory */
struct hfs_dir *hdir = &entry->u.dir;
inode->i_blocks = 0;
inode->i_nlink = hdir->dirs + 2;
inode->i_size = 3 + 2 * (hdir->dirs + hdir->files);
inode->i_mode = S_IRWXUGO | S_IFDIR;
inode->i_op = &hfs_dbl_dir_inode_operations;
inode->i_fop = &hfs_dbl_dir_operations;
HFS_I(inode)->file_type = HFS_DBL_NORM;
HFS_I(inode)->dir_size = 2;
}
}
/*
* hfs_nat_ifill()
*
* This function serves the same purpose as a read_inode() function does
* in other filesystems. It is called by __hfs_iget() to fill in
* the missing fields of an uninitialized inode under the Netatalk
* scheme.
*/
void hfs_nat_ifill(struct inode * inode, ino_t type, const int version)
{
struct hfs_cat_entry *entry = HFS_I(inode)->entry;
HFS_I(inode)->d_drop_op = hfs_nat_drop_dentry;
if (type == HFS_NAT_HDR) {
if (entry->type == HFS_CDR_FIL) {
init_file_inode(inode, HFS_FK_RSRC);
inode->i_size += HFS_NAT_HDR_LEN;
} else {
inode->i_size = HFS_NAT_HDR_LEN;
inode->i_mode = S_IRUGO | S_IWUGO | S_IFREG;
inode->i_nlink = 1;
}
inode->i_op = &hfs_hdr_inode_operations;
inode->i_fop = &hfs_hdr_operations;
HFS_I(inode)->default_layout = (version == 2) ?
&hfs_nat2_hdr_layout : &hfs_nat_hdr_layout;
} else if (entry->type == HFS_CDR_FIL) {
init_file_inode(inode, HFS_FK_DATA);
inode->i_op = &hfs_file_inode_operations;
inode->i_fop = &hfs_file_operations;
inode->i_mapping->a_ops = &hfs_aops;
inode->u.hfs_i.mmu_private = inode->i_size;
} else { /* Directory */
struct hfs_dir *hdir = &entry->u.dir;
inode->i_blocks = 0;
inode->i_size = hdir->files + hdir->dirs + 4;
inode->i_mode = S_IRWXUGO | S_IFDIR;
HFS_I(inode)->dir_size = 1;
if (type == HFS_NAT_NDIR) {
inode->i_nlink = hdir->dirs + 3;
inode->i_op = &hfs_nat_ndir_inode_operations;
HFS_I(inode)->file_type = HFS_NAT_NORM;
} else if (type == HFS_NAT_HDIR) {
inode->i_nlink = 2;
inode->i_op = &hfs_nat_hdir_inode_operations;
HFS_I(inode)->file_type = HFS_NAT_HDR;
}
inode->i_fop = &hfs_nat_dir_operations;
}
}