blob: e50110877bd835ddc9876294534ae4f81be5054c [file] [log] [blame]
/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
* vim:expandtab:shiftwidth=8:tabstop=8:
*
* Copyright (C) 2000 Stelias Computing, Inc.
* Copyright (C) 2000 Red Hat, Inc.
* Copyright (C) 2000 Mountain View Data, Inc.
*
* Extended Attribute Support
* Copyright (C) 2001 Shirish H. Phatak, Tacit Networks, Inc.
*
* This file is part of InterMezzo, http://www.inter-mezzo.org.
*
* InterMezzo is free software; you can redistribute it and/or
* modify it under the terms of version 2 of the GNU General Public
* License as published by the Free Software Foundation.
*
* InterMezzo is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with InterMezzo; if not, write to the Free Software
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
*
*/
#include <stdarg.h>
#include <asm/bitops.h>
#include <asm/uaccess.h>
#include <asm/system.h>
#include <linux/errno.h>
#include <linux/fs.h>
#include <linux/ext2_fs.h>
#include <linux/slab.h>
#include <linux/vmalloc.h>
#include <linux/sched.h>
#include <linux/stat.h>
#include <linux/string.h>
#include <linux/locks.h>
#include <linux/blkdev.h>
#include <linux/init.h>
#define __NO_VERSION__
#include <linux/module.h>
#include <linux/fsfilter.h>
#include <linux/intermezzo_fs.h>
int filter_print_entry = 0;
int filter_debug = 0xfffffff;
/*
* The function in this file are responsible for setting up the
* correct methods layered file systems like InterMezzo and snapfs
*/
static struct filter_fs filter_oppar[FILTER_FS_TYPES];
/* get to the upper methods (intermezzo, snapfs) */
inline struct super_operations *filter_c2usops(struct filter_fs *cache)
{
return &cache->o_fops.filter_sops;
}
inline struct inode_operations *filter_c2udiops(struct filter_fs *cache)
{
return &cache->o_fops.filter_dir_iops;
}
inline struct inode_operations *filter_c2ufiops(struct filter_fs *cache)
{
return &cache->o_fops.filter_file_iops;
}
inline struct inode_operations *filter_c2usiops(struct filter_fs *cache)
{
return &cache->o_fops.filter_sym_iops;
}
inline struct file_operations *filter_c2udfops(struct filter_fs *cache)
{
return &cache->o_fops.filter_dir_fops;
}
inline struct file_operations *filter_c2uffops(struct filter_fs *cache)
{
return &cache->o_fops.filter_file_fops;
}
inline struct file_operations *filter_c2usfops(struct filter_fs *cache)
{
return &cache->o_fops.filter_sym_fops;
}
inline struct dentry_operations *filter_c2udops(struct filter_fs *cache)
{
return &cache->o_fops.filter_dentry_ops;
}
/* get to the cache (lower) methods */
inline struct super_operations *filter_c2csops(struct filter_fs *cache)
{
return cache->o_caops.cache_sops;
}
inline struct inode_operations *filter_c2cdiops(struct filter_fs *cache)
{
return cache->o_caops.cache_dir_iops;
}
inline struct inode_operations *filter_c2cfiops(struct filter_fs *cache)
{
return cache->o_caops.cache_file_iops;
}
inline struct inode_operations *filter_c2csiops(struct filter_fs *cache)
{
return cache->o_caops.cache_sym_iops;
}
inline struct file_operations *filter_c2cdfops(struct filter_fs *cache)
{
return cache->o_caops.cache_dir_fops;
}
inline struct file_operations *filter_c2cffops(struct filter_fs *cache)
{
return cache->o_caops.cache_file_fops;
}
inline struct file_operations *filter_c2csfops(struct filter_fs *cache)
{
return cache->o_caops.cache_sym_fops;
}
inline struct dentry_operations *filter_c2cdops(struct filter_fs *cache)
{
return cache->o_caops.cache_dentry_ops;
}
void filter_setup_journal_ops(struct filter_fs *ops, char *cache_type)
{
if ( strlen(cache_type) == strlen("ext2") &&
memcmp(cache_type, "ext2", strlen("ext2")) == 0 ) {
#if CONFIG_EXT2_FS
ops->o_trops = &presto_ext2_journal_ops;
#else
ops->o_trops = NULL;
#endif
FDEBUG(D_SUPER, "ops at %p\n", ops);
}
if ( strlen(cache_type) == strlen("ext3") &&
memcmp(cache_type, "ext3", strlen("ext3")) == 0 ) {
#if defined(CONFIG_EXT3_FS) || defined (CONFIG_EXT3_FS_MODULE)
ops->o_trops = &presto_ext3_journal_ops;
#else
ops->o_trops = NULL;
#endif
FDEBUG(D_SUPER, "ops at %p\n", ops);
}
if ( strlen(cache_type) == strlen("tmpfs") &&
memcmp(cache_type, "tmpfs", strlen("tmpfs")) == 0 ) {
#if defined(CONFIG_TMPFS)
ops->o_trops = &presto_tmpfs_journal_ops;
#else
ops->o_trops = NULL;
#endif
FDEBUG(D_SUPER, "ops at %p\n", ops);
}
if ( strlen(cache_type) == strlen("reiserfs") &&
memcmp(cache_type, "reiserfs", strlen("reiserfs")) == 0 ) {
#if 0
/* #if defined(CONFIG_REISERFS_FS) || defined(CONFIG_REISERFS_FS_MODULE) */
ops->o_trops = &presto_reiserfs_journal_ops;
#else
ops->o_trops = NULL;
#endif
FDEBUG(D_SUPER, "ops at %p\n", ops);
}
if ( strlen(cache_type) == strlen("xfs") &&
memcmp(cache_type, "xfs", strlen("xfs")) == 0 ) {
#if 0
/*#if defined(CONFIG_XFS_FS) || defined (CONFIG_XFS_FS_MODULE) */
ops->o_trops = &presto_xfs_journal_ops;
#else
ops->o_trops = NULL;
#endif
FDEBUG(D_SUPER, "ops at %p\n", ops);
}
if ( strlen(cache_type) == strlen("obdfs") &&
memcmp(cache_type, "obdfs", strlen("obdfs")) == 0 ) {
#if defined(CONFIG_OBDFS_FS) || defined (CONFIG_OBDFS_FS_MODULE)
ops->o_trops = presto_obdfs_journal_ops;
#else
ops->o_trops = NULL;
#endif
FDEBUG(D_SUPER, "ops at %p\n", ops);
}
}
/* find the cache for this FS */
struct filter_fs *filter_get_filter_fs(const char *cache_type)
{
struct filter_fs *ops = NULL;
FENTRY;
if ( strlen(cache_type) == strlen("ext2") &&
memcmp(cache_type, "ext2", strlen("ext2")) == 0 ) {
ops = &filter_oppar[FILTER_FS_EXT2];
FDEBUG(D_SUPER, "ops at %p\n", ops);
}
if ( strlen(cache_type) == strlen("xfs") &&
memcmp(cache_type, "xfs", strlen("xfs")) == 0 ) {
ops = &filter_oppar[FILTER_FS_XFS];
FDEBUG(D_SUPER, "ops at %p\n", ops);
}
if ( strlen(cache_type) == strlen("ext3") &&
memcmp(cache_type, "ext3", strlen("ext3")) == 0 ) {
ops = &filter_oppar[FILTER_FS_EXT3];
FDEBUG(D_SUPER, "ops at %p\n", ops);
}
if ( strlen(cache_type) == strlen("tmpfs") &&
memcmp(cache_type, "tmpfs", strlen("tmpfs")) == 0 ) {
ops = &filter_oppar[FILTER_FS_TMPFS];
FDEBUG(D_SUPER, "ops at %p\n", ops);
}
if ( strlen(cache_type) == strlen("reiserfs") &&
memcmp(cache_type, "reiserfs", strlen("reiserfs")) == 0 ) {
ops = &filter_oppar[FILTER_FS_REISERFS];
FDEBUG(D_SUPER, "ops at %p\n", ops);
}
if ( strlen(cache_type) == strlen("obdfs") &&
memcmp(cache_type, "obdfs", strlen("obdfs")) == 0 ) {
ops = &filter_oppar[FILTER_FS_OBDFS];
FDEBUG(D_SUPER, "ops at %p\n", ops);
}
if (ops == NULL) {
CERROR("prepare to die: unrecognized cache type for Filter\n");
}
FEXIT;
return ops;
}
/*
* Frobnicate the InterMezzo operations
* this establishes the link between the InterMezzo file system
* and the underlying file system used for the cache.
*/
void filter_setup_super_ops(struct filter_fs *cache, struct super_operations *cache_sops, struct super_operations *filter_sops)
{
/* Get ptr to the shared struct snapfs_ops structure. */
struct filter_ops *props = &cache->o_fops;
/* Get ptr to the shared struct cache_ops structure. */
struct cache_ops *caops = &cache->o_caops;
FENTRY;
if ( cache->o_flags & FILTER_DID_SUPER_OPS ) {
FEXIT;
return;
}
cache->o_flags |= FILTER_DID_SUPER_OPS;
/* Set the cache superblock operations to point to the
superblock operations of the underlying file system. */
caops->cache_sops = cache_sops;
/*
* Copy the cache (real fs) superblock ops to the "filter"
* superblock ops as defaults. Some will be changed below
*/
memcpy(&props->filter_sops, cache_sops, sizeof(*cache_sops));
/* 'put_super' unconditionally is that of filter */
if (filter_sops->put_super) {
props->filter_sops.put_super = filter_sops->put_super;
}
if (cache_sops->read_inode) {
props->filter_sops.read_inode = filter_sops->read_inode;
FDEBUG(D_INODE, "setting filter_read_inode, cache_ops %p, cache %p, ri at %p\n",
cache, cache, props->filter_sops.read_inode);
}
if (cache_sops->remount_fs)
props->filter_sops.remount_fs = filter_sops->remount_fs;
FEXIT;
}
void filter_setup_dir_ops(struct filter_fs *cache, struct inode *inode, struct inode_operations *filter_iops, struct file_operations *filter_fops)
{
struct inode_operations *cache_filter_iops;
struct inode_operations *cache_iops = inode->i_op;
struct file_operations *cache_fops = inode->i_fop;
FENTRY;
if ( cache->o_flags & FILTER_DID_DIR_OPS ) {
FEXIT;
return;
}
cache->o_flags |= FILTER_DID_DIR_OPS;
/* former ops become cache_ops */
cache->o_caops.cache_dir_iops = cache_iops;
cache->o_caops.cache_dir_fops = cache_fops;
FDEBUG(D_SUPER, "filter at %p, cache iops %p, iops %p\n",
cache, cache_iops, filter_c2udiops(cache));
/* setup our dir iops: copy and modify */
memcpy(filter_c2udiops(cache), cache_iops, sizeof(*cache_iops));
/* abbreviate */
cache_filter_iops = filter_c2udiops(cache);
/* methods that filter if cache filesystem has these ops */
if (cache_iops->lookup && filter_iops->lookup)
cache_filter_iops->lookup = filter_iops->lookup;
if (cache_iops->create && filter_iops->create)
cache_filter_iops->create = filter_iops->create;
if (cache_iops->link && filter_iops->link)
cache_filter_iops->link = filter_iops->link;
if (cache_iops->unlink && filter_iops->unlink)
cache_filter_iops->unlink = filter_iops->unlink;
if (cache_iops->mkdir && filter_iops->mkdir)
cache_filter_iops->mkdir = filter_iops->mkdir;
if (cache_iops->rmdir && filter_iops->rmdir)
cache_filter_iops->rmdir = filter_iops->rmdir;
if (cache_iops->symlink && filter_iops->symlink)
cache_filter_iops->symlink = filter_iops->symlink;
if (cache_iops->rename && filter_iops->rename)
cache_filter_iops->rename = filter_iops->rename;
if (cache_iops->mknod && filter_iops->mknod)
cache_filter_iops->mknod = filter_iops->mknod;
if (cache_iops->permission && filter_iops->permission)
cache_filter_iops->permission = filter_iops->permission;
if (cache_iops->getattr)
cache_filter_iops->getattr = filter_iops->getattr;
/* Some filesystems do not use a setattr method of their own
instead relying on inode_setattr/write_inode. We still need to
journal these so we make setattr an unconditional operation.
XXX: we should probably check for write_inode. SHP
*/
/*if (cache_iops->setattr)*/
cache_filter_iops->setattr = filter_iops->setattr;
#ifdef CONFIG_FS_EXT_ATTR
/* For now we assume that posix acls are handled through extended
* attributes. If this is not the case, we must explicitly trap
* posix_set_acl. SHP
*/
if (cache_iops->set_ext_attr && filter_iops->set_ext_attr)
cache_filter_iops->set_ext_attr = filter_iops->set_ext_attr;
#endif
/* copy dir fops */
memcpy(filter_c2udfops(cache), cache_fops, sizeof(*cache_fops));
/* unconditional filtering operations */
filter_c2udfops(cache)->ioctl = filter_fops->ioctl;
FEXIT;
}
void filter_setup_file_ops(struct filter_fs *cache, struct inode *inode, struct inode_operations *filter_iops, struct file_operations *filter_fops)
{
struct inode_operations *pr_iops;
struct inode_operations *cache_iops = inode->i_op;
struct file_operations *cache_fops = inode->i_fop;
FENTRY;
if ( cache->o_flags & FILTER_DID_FILE_OPS ) {
FEXIT;
return;
}
cache->o_flags |= FILTER_DID_FILE_OPS;
/* steal the old ops */
/* former ops become cache_ops */
cache->o_caops.cache_file_iops = cache_iops;
cache->o_caops.cache_file_fops = cache_fops;
/* abbreviate */
pr_iops = filter_c2ufiops(cache);
/* setup our dir iops: copy and modify */
memcpy(pr_iops, cache_iops, sizeof(*cache_iops));
/* copy dir fops */
CERROR("*** cache file ops at %p\n", cache_fops);
memcpy(filter_c2uffops(cache), cache_fops, sizeof(*cache_fops));
/* assign */
/* See comments above in filter_setup_dir_ops. SHP */
/*if (cache_iops->setattr)*/
pr_iops->setattr = filter_iops->setattr;
if (cache_iops->getattr)
pr_iops->getattr = filter_iops->getattr;
/* XXX Should this be conditional rmr ? */
pr_iops->permission = filter_iops->permission;
#ifdef CONFIG_FS_EXT_ATTR
/* For now we assume that posix acls are handled through extended
* attributes. If this is not the case, we must explicitly trap and
* posix_set_acl
*/
if (cache_iops->set_ext_attr && filter_iops->set_ext_attr)
pr_iops->set_ext_attr = filter_iops->set_ext_attr;
#endif
/* unconditional filtering operations */
filter_c2uffops(cache)->open = filter_fops->open;
filter_c2uffops(cache)->release = filter_fops->release;
filter_c2uffops(cache)->write = filter_fops->write;
filter_c2uffops(cache)->ioctl = filter_fops->ioctl;
FEXIT;
}
/* XXX in 2.3 there are "fast" and "slow" symlink ops for ext2 XXX */
void filter_setup_symlink_ops(struct filter_fs *cache, struct inode *inode, struct inode_operations *filter_iops, struct file_operations *filter_fops)
{
struct inode_operations *pr_iops;
struct inode_operations *cache_iops = inode->i_op;
struct file_operations *cache_fops = inode->i_fop;
FENTRY;
if ( cache->o_flags & FILTER_DID_SYMLINK_OPS ) {
FEXIT;
return;
}
cache->o_flags |= FILTER_DID_SYMLINK_OPS;
/* steal the old ops */
cache->o_caops.cache_sym_iops = cache_iops;
cache->o_caops.cache_sym_fops = cache_fops;
/* abbreviate */
pr_iops = filter_c2usiops(cache);
/* setup our dir iops: copy and modify */
memcpy(pr_iops, cache_iops, sizeof(*cache_iops));
/* See comments above in filter_setup_dir_ops. SHP */
/* if (cache_iops->setattr) */
pr_iops->setattr = filter_iops->setattr;
if (cache_iops->getattr)
pr_iops->getattr = filter_iops->getattr;
/* assign */
/* copy fops - careful for symlinks they might be NULL */
if ( cache_fops ) {
memcpy(filter_c2usfops(cache), cache_fops, sizeof(*cache_fops));
}
FEXIT;
}
void filter_setup_dentry_ops(struct filter_fs *cache,
struct dentry_operations *cache_dop,
struct dentry_operations *filter_dop)
{
if ( cache->o_flags & FILTER_DID_DENTRY_OPS ) {
FEXIT;
return;
}
cache->o_flags |= FILTER_DID_DENTRY_OPS;
cache->o_caops.cache_dentry_ops = cache_dop;
memcpy(&cache->o_fops.filter_dentry_ops,
filter_dop, sizeof(*filter_dop));
if (cache_dop && cache_dop != filter_dop && cache_dop->d_revalidate){
CERROR("WARNING: filter overriding revalidation!\n");
}
return;
}