Index: linux-2.6.25.1/fs/Kconfig =================================================================== --- linux-2.6.25.1.orig/fs/Kconfig +++ linux-2.6.25.1/fs/Kconfig @@ -1449,6 +1449,9 @@ config VXFS_FS To compile this as a module, choose M here: the module will be called freevxfs. If unsure, say N. +config MINI_FO + tristate "Mini fanout overlay filesystem" + config MINIX_FS tristate "Minix file system support" depends on BLOCK Index: linux-2.6.25.1/fs/Makefile =================================================================== --- linux-2.6.25.1.orig/fs/Makefile +++ linux-2.6.25.1/fs/Makefile @@ -77,6 +77,7 @@ obj-$(CONFIG_SQUASHFS) += squashfs/ obj-y += ramfs/ obj-$(CONFIG_HUGETLBFS) += hugetlbfs/ obj-$(CONFIG_CODA_FS) += coda/ +obj-$(CONFIG_MINI_FO) += mini_fo/ obj-$(CONFIG_MINIX_FS) += minix/ obj-$(CONFIG_FAT_FS) += fat/ obj-$(CONFIG_MSDOS_FS) += msdos/ Index: linux-2.6.25.1/fs/mini_fo/aux.c =================================================================== --- /dev/null +++ linux-2.6.25.1/fs/mini_fo/aux.c @@ -0,0 +1,577 @@ +/* + * Copyright (c) 1997-2003 Erez Zadok + * Copyright (c) 2001-2003 Stony Brook University + * + * For specific licensing information, see the COPYING file distributed with + * this package, or get one from ftp://ftp.filesystems.org/pub/fist/COPYING. + * + * This Copyright notice must be kept intact and distributed with all + * fistgen sources INCLUDING sources generated by fistgen. + */ +/* + * Copyright (C) 2004, 2005 Markus Klotzbuecher + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ +/* + * $Id$ + */ + +#ifdef HAVE_CONFIG_H +# include +#endif + +#include "fist.h" +#include "mini_fo.h" + +/* check if file exists in storage */ +int exists_in_storage(dentry_t *dentry) +{ + check_mini_fo_dentry(dentry); + if(dtost(dentry) == MODIFIED || dtost(dentry) == CREATED || dtost(dentry) == DEL_REWRITTEN) + return 1; + return 0; +} + +/* check if dentry is in an existing state */ +int is_mini_fo_existant(dentry_t *dentry) +{ + check_mini_fo_dentry(dentry); + + if(dtost(dentry) == DELETED || dtost(dentry) == NON_EXISTANT) + return 0; + else + return 1; +} + +/* + * This function will create a negative storage dentry for + * dentry, what is required for many create like options. + * It will create the storage structure if necessary. + */ +int get_neg_sto_dentry(dentry_t *dentry) +{ + int err = 0; + unsigned int len; + const unsigned char *name; + + if(!dentry || + !dtopd(dentry) || + !(dtost(dentry) == UNMODIFIED || + dtost(dentry) == NON_EXISTANT || + dtost(dentry) == DELETED)) { + printk(KERN_CRIT "mini_fo: get_neg_sto_dentry: invalid dentry passed.\n"); + err = -1; + goto out; + } + /* Have we got a neg. dentry already? */ + if(dtohd2(dentry)) { + err = 0; + goto out; + } + if(dtost(dentry->d_parent) == UNMODIFIED) { + /* build sto struct */ + err = build_sto_structure(dentry->d_parent->d_parent, dentry->d_parent); + if(err || + dtost(dentry->d_parent) != MODIFIED) { + printk(KERN_CRIT "mini_fo: get_neg_sto_dentry: ERROR building sto structure.\n"); + err = -1; + goto out; + } + } + + len = dentry->d_name.len; + name = dentry->d_name.name; + + dtohd2(dentry) = + lookup_one_len(name, dtohd2(dentry->d_parent), len); + + out: + return err; +} + +int check_mini_fo_dentry(dentry_t *dentry) +{ + ASSERT(dentry != NULL); + ASSERT(dtopd(dentry) != NULL); + ASSERT((dtohd(dentry) != NULL) || (dtohd2(dentry) != NULL)); + +/* if(dtost(dentry) == MODIFIED) { */ +/* ASSERT(dentry->d_inode != NULL); */ +/* ASSERT(dtohd(dentry) != NULL); */ +/* ASSERT(dtohd(dentry)->d_inode != NULL); */ +/* ASSERT(dtohd2(dentry) != NULL); */ +/* ASSERT(dtohd2(dentry)->d_inode != NULL); */ +/* } */ +/* else if(dtost(dentry) == UNMODIFIED) { */ +/* ASSERT(dentry->d_inode != NULL); */ +/* ASSERT( */ +/* } */ + return 0; +} + +int check_mini_fo_file(file_t *file) +{ + ASSERT(file != NULL); + ASSERT(ftopd(file) != NULL); + ASSERT(file->f_dentry != NULL); + + /* violent checking, check depending of state and type + * if(S_ISDIR(file->f_dentry->d_inode->i_mode)) {} + */ + ASSERT((ftohf(file) != NULL) || (ftohf2(file) != NULL)); + return 0; +} + +int check_mini_fo_inode(inode_t *inode) +{ + ASSERT(inode != NULL); + ASSERT(itopd(inode) != NULL); + ASSERT((itohi(inode) != NULL) || (itohi2(inode) != NULL)); + return 0; +} + +/* + * will walk a base path as provided by get_mini_fo_bpath and return + * the (hopefully ;-) ) positive dentry of the renamed base dir. + * + * This does some work of path_init. + */ +dentry_t *bpath_walk(super_block_t *sb, char *bpath) +{ + int err; + struct vfsmount *mnt; + struct nameidata nd; + + /* be paranoid */ + if(!bpath || bpath[0] != '/') { + printk(KERN_CRIT "mini_fo: bpath_walk: Invalid string.\n"); + return NULL; + } + if(!sb || !stopd(sb)) { + printk(KERN_CRIT "mini_fo: bpath_walk: Invalid sb.\n"); + return NULL; + } + + /* fix this: how do I reach this lock? + * read_lock(¤t->fs->lock); */ + mnt = mntget(stopd(sb)->hidden_mnt); + /* read_unlock(¤t->fs->lock); */ + + err = vfs_path_lookup(mnt->mnt_root, mnt, bpath+1, 0, &nd); + + /* validate */ + if (err || !nd.dentry || !nd.dentry->d_inode) { + printk(KERN_CRIT "mini_fo: bpath_walk: path_walk failed.\n"); + return NULL; + } + return nd.dentry; +} + + +/* returns the full path of the basefile incl. its name */ +int get_mini_fo_bpath(dentry_t *dentry, char **bpath, int *bpath_len) +{ + char *buf_walker; + int len = 0; + dentry_t *sky_walker; + + if(!dentry || !dtohd(dentry)) { + printk(KERN_CRIT "mini_fo: get_mini_fo_bpath: invalid dentry passed.\n"); + return -1; + } + sky_walker = dtohd(dentry); + + do { + len += sky_walker->d_name.len + 1 ; /* 1 for '/' */ + sky_walker = sky_walker->d_parent; + } while(sky_walker != stopd(dentry->d_inode->i_sb)->base_dir_dentry); + + /* 1 to oil the loop */ + *bpath = (char*) kmalloc(len + 1, GFP_KERNEL); + if(!*bpath) { + printk(KERN_CRIT "mini_fo: get_mini_fo_bpath: out of mem.\n"); + return -1; + } + buf_walker = *bpath+len; /* put it on last char */ + *buf_walker = '\n'; + sky_walker = dtohd(dentry); + + do { + buf_walker -= sky_walker->d_name.len; + strncpy(buf_walker, + sky_walker->d_name.name, + sky_walker->d_name.len); + *(--buf_walker) = '/'; + sky_walker = sky_walker->d_parent; + } while(sky_walker != stopd(dentry->d_inode->i_sb)->base_dir_dentry); + + /* bpath_len doesn't count newline! */ + *bpath_len = len; + return 0; +} + +int mini_fo_cp_cont(dentry_t *tgt_dentry, struct vfsmount *tgt_mnt, + dentry_t *src_dentry, struct vfsmount *src_mnt) +{ + void *buf; + mm_segment_t old_fs; + file_t *tgt_file; + file_t *src_file; + int bytes, len, tmp, err; + err = 0; + + if(!(tgt_dentry->d_inode && src_dentry->d_inode)) { + printk(KERN_CRIT "mini_fo_cp_cont: ERROR, neg. dentry passed.\n"); + err = -EINVAL; + goto out; + } + + dget(tgt_dentry); + dget(src_dentry); + mntget(tgt_mnt); + mntget(src_mnt); + + /* open file write only */ + tgt_file = dentry_open(tgt_dentry, tgt_mnt, 0x1); + if(!tgt_file || IS_ERR(tgt_file)) { + printk(KERN_CRIT "mini_fo_cp_cont: ERROR opening target file.\n"); + err = PTR_ERR(tgt_file); + goto out_err; + } + + /* open file read only */ + src_file = dentry_open(src_dentry, src_mnt, 0x0); + if(!src_file || IS_ERR(src_file)) { + printk(KERN_CRIT "mini_fo_cp_cont: ERROR opening source file.\n"); + err = PTR_ERR(src_file); + + /* close target file */ + fput(tgt_file); + goto out_err; + } + + /* check if the filesystem(s) support read respective write */ + if(!src_file->f_op->read || !tgt_file->f_op->write) { + printk(KERN_CRIT "mini_fo_cp_cont: ERROR, no fs read or write support.\n"); + err = -EPERM; + goto out_close; + } + + /* allocate a page for transfering the data */ + buf = (void *) __get_free_page(GFP_KERNEL); + if(!buf) { + printk(KERN_CRIT "mini_fo_cp_cont: ERROR, out of kernel mem.\n"); + goto out_err; + } + + tgt_file->f_pos = 0; + src_file->f_pos = 0; + + old_fs = get_fs(); + set_fs(KERNEL_DS); + + /* Doing this I assume that a read operation will return a full + * buffer while there is still data to read, and a less than + * full buffer when all data has been read. + */ + bytes = len = PAGE_SIZE; + while(bytes == len) { + bytes = src_file->f_op->read(src_file, buf, len, + &src_file->f_pos); + tmp = tgt_file->f_op->write(tgt_file, buf, bytes, + &tgt_file->f_pos); + if(tmp != bytes) { + printk(KERN_CRIT "mini_fo_cp_cont: ERROR writing.\n"); + goto out_close_unset; + } + } + + free_page((unsigned long) buf); + set_fs(old_fs); + fput(tgt_file); + fput(src_file); + goto out; + + out_close_unset: + free_page((unsigned long) buf); + set_fs(old_fs); + + out_close: + fput(tgt_file); + fput(src_file); + + out_err: + dput(tgt_dentry); + dput(src_dentry); + + /* mk: not sure if this need to be done */ + mntput(tgt_mnt); + mntput(src_mnt); + + out: + return err; +} + +/* mk: + * ndl (no-duplicate list) stuff + * This is used in mini_fo_readdir, to save the storage directory contents + * and later when reading base, match them against the list in order + * to avoid duplicates. + */ + +/* add a file specified by name and len to the ndl + * Return values: 0 on success, <0 on failure. + */ +int ndl_add_entry(struct readdir_data *rd, const char *name, int len) +{ + struct ndl_entry *tmp_entry; + + tmp_entry = (struct ndl_entry *) + kmalloc(sizeof(struct ndl_entry), GFP_KERNEL); + if(!tmp_entry) { + printk(KERN_CRIT "mini_fo: ndl_add_entry: out of mem.\n"); + return -ENOMEM; + } + tmp_entry->name = (char*) kmalloc(len, GFP_KERNEL); + if(!tmp_entry->name) { + printk(KERN_CRIT "mini_fo: ndl_add_entry: out of mem.\n"); + return -ENOMEM; + } + strncpy(tmp_entry->name, name, len); + tmp_entry->len = len; + + list_add(&tmp_entry->list, &rd->ndl_list); + rd->ndl_size++; + return 0; +} + +/* delete all list entries and free memory */ +void ndl_put_list(struct readdir_data *rd) +{ + struct list_head *tmp; + struct ndl_entry *tmp_entry; + + if(rd->ndl_size <= 0) + return; + while(!list_empty(&rd->ndl_list)) { + tmp = rd->ndl_list.next; + list_del(tmp); + tmp_entry = list_entry(tmp, struct ndl_entry, list); + kfree(tmp_entry->name); + kfree(tmp_entry); + } + rd->ndl_size = 0; +} + +/* Check if a file specified by name and len is in the ndl + * Return value: 0 if not in list, 1 if file is found in ndl. + */ +int ndl_check_entry(struct readdir_data *rd, const char *name, int len) +{ + struct list_head *tmp; + struct ndl_entry *tmp_entry; + + if(rd->ndl_size <= 0) + return 0; + + list_for_each(tmp, &rd->ndl_list) { + tmp_entry = list_entry(tmp, struct ndl_entry, list); + if(tmp_entry->len != len) + continue; + if(!strncmp(tmp_entry->name, name, len)) + return 1; + } + return 0; +} + +/* mk: + * Recursive function to create corresponding directorys in the storage fs. + * The function will build the storage directorys up to dentry. + */ +int build_sto_structure(dentry_t *dir, dentry_t *dentry) +{ + int err; + dentry_t *hidden_sto_dentry; + dentry_t *hidden_sto_dir_dentry; + + if(dentry->d_parent != dir) { + printk(KERN_CRIT "mini_fo: build_sto_structure: invalid parameter or meta data corruption [1].\n"); + return 1; + } + + if(dtost(dir) != MODIFIED) { + err = build_sto_structure(dir->d_parent, dentry->d_parent); + if(err) + return err; + } + + /* ok, coming back again. */ + check_mini_fo_dentry(dentry); + hidden_sto_dentry = dtohd2(dentry); + + if(!hidden_sto_dentry) { + /* + * This is the case after creating the first + * hidden_sto_dentry. + * After one negative storage_dentry, all pointers to + * hidden_storage dentries are set to NULL. We need to + * create the negative dentry before we create the storage + * file. + */ + unsigned int len; + const unsigned char *name; + len = dtohd(dentry)->d_name.len; + name = dtohd(dentry)->d_name.name; + hidden_sto_dentry = lookup_one_len(name, dtohd2(dir), len); + dtohd2(dentry) = hidden_sto_dentry; + } + + /* was: hidden_sto_dir_dentry = lock_parent(hidden_sto_dentry); */ + hidden_sto_dir_dentry = dget(hidden_sto_dentry->d_parent); +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + down(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + /* lets be safe */ + if(dtohd2(dir) != hidden_sto_dir_dentry) { + printk(KERN_CRIT "mini_fo: build_sto_structure: invalid parameter or meta data corruption [2].\n"); + return 1; + } + + /* check for errors in lock_parent */ + err = PTR_ERR(hidden_sto_dir_dentry); + if(IS_ERR(hidden_sto_dir_dentry)) { + printk(KERN_CRIT "mini_fo: build_sto_structure: lock_parent failed.\n"); + return err; + } + + err = vfs_mkdir(hidden_sto_dir_dentry->d_inode, + hidden_sto_dentry, + dir->d_inode->i_mode); + + if(err) { + printk(KERN_CRIT "mini_fo: build_sto_structure: failed to create storage dir [1].\n"); + /* was: unlock_dir(dir); */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&dir->d_inode->i_mutex); +#else + up(&dir->d_inode->i_sem); +#endif + dput(dir); + return err; + } + + /* everything ok! */ + if(!dtohd2(dentry)->d_inode) { + printk(KERN_CRIT "mini_fo: build_sto_structure: failed to create storage dir [2].\n"); + /* was: unlock_dir(dir); */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&dir->d_inode->i_mutex); +#else + up(&dir->d_inode->i_sem); +#endif + dput(dir); + return 1; + } + + /* interpose the new inode and set new state */ + itohi2(dentry->d_inode) = igrab(dtohd2(dentry)->d_inode); + dtopd(dentry)->state = MODIFIED; + + /* initalize the wol list */ + itopd(dentry->d_inode)->deleted_list_size = -1; + itopd(dentry->d_inode)->renamed_list_size = -1; + meta_build_lists(dentry); + + fist_copy_attr_all(dentry->d_inode, itohi2(dentry->d_inode)); + fist_copy_attr_timesizes(dir->d_inode, + hidden_sto_dir_dentry->d_inode); + dir->d_inode->i_nlink++; + /* was: unlock_dir(hidden_sto_dir_dentry); */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + up(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + dput(hidden_sto_dir_dentry); + return 0; +} + + +#if 0 /* unused */ + +/* + * Read "len" bytes from "filename" into "buf". + * "buf" is in kernel space. + */ +int +mini_fo_read_file(const char *filename, void *buf, int len) +{ + file_t *filp; + mm_segment_t oldfs; + int bytes; + /* Chroot? Maybe NULL isn't right here */ + filp = filp_open(filename, O_RDONLY, 0); + if (!filp || IS_ERR(filp)) { + printk("mini_fo_read_file err %d\n", (int) PTR_ERR(filp)); + return -1; /* or do something else */ + } + + if (!filp->f_op->read) + return -2; /* file(system) doesn't allow reads */ + + /* now read len bytes from offset 0 */ + filp->f_pos = 0; /* start offset */ + oldfs = get_fs(); + set_fs(KERNEL_DS); + bytes = filp->f_op->read(filp, buf, len, &filp->f_pos); + set_fs(oldfs); + + /* close the file */ + fput(filp); + + return bytes; +} + + + +/* + * Write "len" bytes from "buf" to "filename" + * "buf" is in kernel space. + */ +int +mini_fo_write_file(const char *filename, void *buf, int len) +{ + file_t *filp; + mm_segment_t oldfs; + int bytes; + /* Chroot? Maybe NULL isn't right here */ + filp = filp_open(filename, O_RDWR|O_CREAT, 0640); + if (!filp || IS_ERR(filp)) { + printk("mini_fo_write_file err %d\n", (int) PTR_ERR(filp)); + return -1; /* or do something else */ + } + + if (!filp->f_op->write) + return -2; /* file(system) doesn't allow writes */ + + /* now write len bytes from offset 0 */ + filp->f_pos = 0; /* start offset */ + oldfs = get_fs(); + set_fs(KERNEL_DS); + bytes = filp->f_op->write(filp, buf, len, &filp->f_pos); + set_fs(oldfs); + + /* close the file */ + fput(filp); + + return bytes; +} + +#endif /* unused */ + Index: linux-2.6.25.1/fs/mini_fo/ChangeLog =================================================================== --- /dev/null +++ linux-2.6.25.1/fs/mini_fo/ChangeLog @@ -0,0 +1,281 @@ +2006-01-24 Markus Klotzbuecher + + * Add tons of ugly ifdefs to Ed L. Cashin's mutex patch to + retain backwards compatibility. + +2006-01-24 Ed L. Cashin + + * Support for the new mutex infrastructure + (7892f2f48d165a34b0b8130c8a195dfd807b8cb6) + +2005-10-15 Markus Klotzbuecher + + * Bugfix for a serious memory leak in mini_fo_follow_link. + +2005-09-21 Markus Klotzbuecher + + * new release 0.6.1 + + * fix of a compiler warning due to changes in 2.6.13 + +2005-09-21 Klaus Wenninger + + * file.c: readdir: fix for a bug that caused directory entries + to show up twice when using storage filesystems such as + minixfs or pramfs. + +2005-06-30 Eric Lammerts + + * fix for an oops when overwriting a binary thats beeing + executed. + +2005-06-09 + + * Renamed overlay to mini_fo-overlay. + + * Added mini_fo-merge script to allow merging of storage and base + after making modifications. + +2005-05-22 root + + * Added overlay script that allows to easily mount mini_fo ontop + of a given base directory + +2005-05-10 + + * inode.c: xattr functions return -EOPNOSUPP instead of + -ENOSUPP, what confuses "ls -l" + + * Changed license from LGPL to GPL. + +2005-05-08 root + + * Makefile: clean it up and added make install and make + uninstall. + +2005-05-06 + + * merged devel branch back to main. [v0-6-0-pre3] + + * removed unused files print.c and fist_ioctl. [devel-0-0-18] + + * ioctl: removed fist_ioctl stuff, that is not needed for + now. + +2005-05-03 + + * file.c: simplified mini_fo_open and mini_fo_setattr using + new state changing functions. [devel-0-0-17] + + * inode.c: Fixed getattr state bug (see below) in 2.4 function + mini_fo_inode revalidate. + + * inode.c: found an other bug in mini_fo_getattr. States are not + reliable in this function, as a file can be opened, unlinked and + the getattr function called. This results in a deleted dentry + with an inode. Fix is to ignore states and simply use the inode + available. + +2005-04-29 + + * file.c: Bugfix and cleanup in fasync and fsync. [devel-0-0-16] + + * file.c: do not use mini_fo_lock so the generic version is + used (I guess). + + * inode.c: getattr, never call getattr on lower files, as this + will cause the inum to change. + + * inode.c: rename_reg_file renamed to rename_nondir, as it + doesn't matter as long it't not a dir. Removed all + rename_dev_file etc. + + * tagged as devel-0-0-15 + + * inode.c: added support for chosing support for extended + attrs at compile time by XATTR define in mini_fo.h . + + * inode.c: fixed mini_fo_getattr to use mini_fo inode and not + lower again, what avoids inode number changes that confused + rm again. This is the proper solution. + +2005-04-24 + + * all files: updated Copyright notive to 2005. [devel-0-0-14] + + * inode.c: fixed mini_fo_getattr to not change the inode + number, even if lower files change. + + * super.c: fixed a bug that caused deleted base file to show + up suddenly after some time, or after creating a special + file. The problem was that after some time or after special + file creating sync_sb_inodes is called by the vfs, that + called our mini_fo_put_inode. There was (wrongly) called + __meta_put_lists, that nuked the lists, although the inode + was going to continue its life. Moving __meta_put_lists to + mini_fo_clear_inode, where an inode is really destroyed, + solved the problem. + + +2005-04-23 + + * state.c, aux.c: more cleaning up and + simplifications. [devel-0-0-13] + + * inode.c: implemented mini_fo_getattr, that was required for + 2.6 because inode_revalidate has been remove there, and the + old "du" bug returned. + + +2005-04-20 + + * aux.c: get_neg_sto_dentry(): allow to be called for dentries + in state UNMODIFIED, NON_EXISTANT _and_ DELETED. + +2005-04-19 + + * Fixed a bug under 2.6 that caused files deleted via mini_fo + not to be deleted properly and therefore the fs filled up + untill no memory was left. [devel-0-0-12] + + * Added basic hard link support. This means that creating + hardlinks will work, but existing ones will be treated as + individual files. [devel-0-0-11] + +2005-04-17 + + * Bugfixes + +2005-04-13 root + + * Added file state.c for the state transition + functions. Doesn't work very well yet, though... + +2005-04-12 + + * Porting to 2.6 started, which is easier than expected, also + due to Olivier previous work. + +2005-04-08 + + * Fixed the bug that caused du to return invalid sizes of + directory trees. The problem was that + mini_fo_inode_revalidate didn't always copy the attributes + from the base inode properly. + +2005-04-01 Markus Klotzbuecher + + * Merged devel branch back to main trunk and updated the + RELEASE notes. This will be 0-6-0-pre1. + +2005-03-31 Markus Klotzbuecher + + * Fixed some bugs in rename_reg_file, that only showed up in + the kernel compile test. Kernel compiles cleanly ontop of + mini_fo, now also make mrproper etc. work. Seems pretty stable. + +2005-03-28 Markus Klotzbuecher + + * Many, many directory renaming bugfixes and a lot of other + cleanup. Dir renaming seems to work relatively stable. + +2005-03-22 Markus Klotzbuecher + + * Finished implementing lightweight directory renaming. Some + basic testing indicates it works fine. + Next is to implement testcases for the testsuite and confirm + everything is really working ok. + +2005-03-18 Markus Klotzbuecher + + * Finished implementing meta.c stuff required for directory + renaming. + +2005-03-17 Markus Klotzbuecher + + * Fixed all compile warnings + an extremly old bug that + somehow crept in while reworking the wol stuff to the META + system. Turning on -Werror again... :-) + + * Fixed some bugs in the new rename_reg_file function. + + * Rewrote mini_fo rename and split it into several + subfunctions, that handle the different types + seperately. Rewrote the regular file function aswell, as it + was implemented somewhat inefficient. + +2005-03-16 Markus Klotzbuecher + + * Implemented new META subsystem, removed old WOL stuff in favor + if it. + + * After some basic testing everything seems ok... + +2005-03-11 Markus Klotzbuecher + + * Renaming a non regular file caused trouble because I always + tried to copy the contents. Now I only do this for regular + files. mini_fo_rename still isn't implemented properly, renaming + of device files, symlinks etc. results in a empty regular file + instead of the proper type. + + * Directory renaming suddenly works! What a surprise! I guess + this is because renaming is implemented as making a copy and + removing the original. Still this might not work + everywhere... + +2005-03-09 Markus Klotzbuecher + + * Bugfix, when a mini_fo directory that exists in storage + (state: MODIFIED, CREATED and DEL_REWRITTEN) is deleted, a + possibly existing WOL file contained in it needs to be + deleted too. + + * Starting cleanup: defined state names in order to get rid of + the state numbers. + +2005-03-08 Markus Klotzbuecher + + * Makefile fix, fist_ioctl was built against wrong sources if ARCH=um + + * Fixed a bug in dentry.c, mini_fo_d_hash. In state 4 = + DEL_REWRITTEN the hash was calculated from the base dentry, + which was wrong and and caused assertions in + __mini_fo_hidden_dentry to fail. + +2005-02-21 + + * Implemented directory deleting (inode.c) + + * main.c: made mini_fo_parse_options a little more robust. + +2004-12-22 + + * Makefile cleanup and uml stuff, removed unneccessary files + + * Created a new and hopefully more informative README + + * CHANGELOG: created a new CHANGELOG and added old entries reversely + + +2004-10-24 Gleb Natapov + + * Fix: owner and group where not correctly copied from base to + storage. + + +2004-10-05 Gleb Natapov + + * Implementation of fsync, fasync and lock mini_fo functions. + + +2004-09-29 Bob Lee + + * Fix of a serious pointer bug + + +2004-09-28 Gleb Natapov + + * Implementation of mini_fo_mknod and mini_fo_rename, support + for device files. + Index: linux-2.6.25.1/fs/mini_fo/dentry.c =================================================================== --- /dev/null +++ linux-2.6.25.1/fs/mini_fo/dentry.c @@ -0,0 +1,244 @@ +/* + * Copyright (c) 1997-2003 Erez Zadok + * Copyright (c) 2001-2003 Stony Brook University + * + * For specific licensing information, see the COPYING file distributed with + * this package, or get one from ftp://ftp.filesystems.org/pub/fist/COPYING. + * + * This Copyright notice must be kept intact and distributed with all + * fistgen sources INCLUDING sources generated by fistgen. + */ +/* + * Copyright (C) 2004, 2005 Markus Klotzbuecher + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ + +/* + * $Id$ + */ + +#ifdef HAVE_CONFIG_H +# include +#endif + +#include "fist.h" +#include "mini_fo.h" + +/* + * THIS IS A BOOLEAN FUNCTION: returns 1 if valid, 0 otherwise. + */ +STATIC int +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) +mini_fo_d_revalidate(dentry_t *dentry, struct nameidata *nd) +#else +mini_fo_d_revalidate(dentry_t *dentry, int flags) +#endif +{ + int err1 = 1; /* valid = 1, invalid = 0 */ + int err2 = 1; + dentry_t *hidden_dentry; + dentry_t *hidden_sto_dentry; + + + check_mini_fo_dentry(dentry); + + hidden_dentry = dtohd(dentry); + hidden_sto_dentry = dtohd2(dentry); + + if(hidden_dentry && + hidden_dentry->d_op && + hidden_dentry->d_op->d_revalidate) { +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) + err1 = hidden_dentry->d_op->d_revalidate(hidden_dentry, nd); +#else + err1 = hidden_dentry->d_op->d_revalidate(hidden_dentry, flags); +#endif + } + if(hidden_sto_dentry && + hidden_sto_dentry->d_op && + hidden_sto_dentry->d_op->d_revalidate) { +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) + err2 = hidden_sto_dentry->d_op->d_revalidate(hidden_sto_dentry, + nd); +#else + err2 = hidden_sto_dentry->d_op->d_revalidate(hidden_sto_dentry, + flags); +#endif + } + + /* mk: if one of the lower level dentries are valid, + * the mini_fo dentry is too. + */ + return (err1 || err2); +} + + +STATIC int +mini_fo_d_hash(dentry_t *dentry, qstr_t *name) +{ + int err = 0; + dentry_t *hidden_dentry; + dentry_t *hidden_sto_dentry; + + /* hidden_dentry = mini_fo_hidden_dentry(dentry); + * hidden_sto_dentry = mini_fo_hidden_sto_dentry(dentry); */ + + /* state 1, 3, 4, 5: build the hash for the storage dentry */ + if((dtopd(dentry)->state == MODIFIED) || + (dtopd(dentry)->state == CREATED) || + (dtopd(dentry)->state == DEL_REWRITTEN) || + (dtopd(dentry)->state == DELETED)) { + hidden_sto_dentry = dtohd2(dentry); + if(hidden_sto_dentry && + hidden_sto_dentry->d_op && + hidden_sto_dentry->d_op->d_hash) { + err = hidden_sto_dentry->d_op->d_hash(hidden_sto_dentry, name); + } + goto out; + } + /* state 2: build the hash for the base dentry */ + if(dtopd(dentry)->state == UNMODIFIED) { + hidden_dentry = dtohd(dentry); + if(hidden_dentry && + hidden_dentry->d_op && + hidden_dentry->d_op->d_hash) { + err = hidden_dentry->d_op->d_hash(hidden_dentry, name); + } + goto out; + } + /* state 6: build hash for the dentry that exists */ + if(dtopd(dentry)->state == NON_EXISTANT) { + hidden_sto_dentry = dtohd2(dentry); + if(hidden_sto_dentry && + hidden_sto_dentry->d_op && + hidden_sto_dentry->d_op->d_hash) { + err = hidden_sto_dentry->d_op->d_hash(hidden_sto_dentry, name); + goto out; + } + hidden_dentry = dtohd(dentry); + if(hidden_dentry && + hidden_dentry->d_op && + hidden_dentry->d_op->d_hash) { + err = hidden_dentry->d_op->d_hash(hidden_dentry, name); + goto out; + } + } + + printk(KERN_CRIT "mini_fo: d_hash: invalid state detected.\n"); + + out: + return err; +} + + +STATIC int +mini_fo_d_compare(dentry_t *dentry, qstr_t *a, qstr_t *b) +{ + int err; + dentry_t *hidden_dentry=NULL; + + /* hidden_dentry = mini_fo_hidden_dentry(dentry); */ + if(dtohd2(dentry)) + hidden_dentry = dtohd2(dentry); + else if(dtohd(dentry)) + hidden_dentry = dtohd(dentry); + + if (hidden_dentry && hidden_dentry->d_op && hidden_dentry->d_op->d_compare) { + err = hidden_dentry->d_op->d_compare(hidden_dentry, a, b); + } else { + err = ((a->len != b->len) || memcmp(a->name, b->name, b->len)); + } + + return err; +} + + +int +mini_fo_d_delete(dentry_t *dentry) +{ + dentry_t *hidden_dentry; + dentry_t *hidden_sto_dentry; + int err = 0; + + /* this could be a negative dentry, so check first */ + if (!dtopd(dentry)) { + printk(KERN_CRIT "mini_fo_d_delete: negative dentry passed.\n"); + goto out; + } + hidden_dentry = dtohd(dentry); + hidden_sto_dentry = dtohd2(dentry); + + if(hidden_dentry) { + if(hidden_dentry->d_op && + hidden_dentry->d_op->d_delete) { + err = hidden_dentry->d_op->d_delete(hidden_dentry); + } + } + if(hidden_sto_dentry) { + if(hidden_sto_dentry->d_op && + hidden_sto_dentry->d_op->d_delete) { + err = hidden_sto_dentry->d_op->d_delete(hidden_sto_dentry); + } + } + + out: + return err; +} + + +void +mini_fo_d_release(dentry_t *dentry) +{ + dentry_t *hidden_dentry; + dentry_t *hidden_sto_dentry; + + + /* this could be a negative dentry, so check first */ + if (!dtopd(dentry)) { + printk(KERN_CRIT "mini_fo_d_release: no private data.\n"); + goto out; + } + hidden_dentry = dtohd(dentry); + hidden_sto_dentry = dtohd2(dentry); + + if(hidden_dentry) { + /* decrement hidden dentry's counter and free its inode */ + dput(hidden_dentry); + } + if(hidden_sto_dentry) { + /* decrement hidden dentry's counter and free its inode */ + dput(hidden_sto_dentry); + } + + /* free private data (mini_fo_dentry_info) here */ + kfree(dtopd(dentry)); + __dtopd(dentry) = NULL; /* just to be safe */ + out: + return; +} + + +/* + * we don't really need mini_fo_d_iput, because dentry_iput will call iput() if + * mini_fo_d_iput is not defined. We left this implemented for ease of + * tracing/debugging. + */ +void +mini_fo_d_iput(dentry_t *dentry, inode_t *inode) +{ + iput(inode); +} + + +struct dentry_operations mini_fo_dops = { + d_revalidate: mini_fo_d_revalidate, + d_hash: mini_fo_d_hash, + d_compare: mini_fo_d_compare, + d_release: mini_fo_d_release, + d_delete: mini_fo_d_delete, + d_iput: mini_fo_d_iput, +}; Index: linux-2.6.25.1/fs/mini_fo/file.c =================================================================== --- /dev/null +++ linux-2.6.25.1/fs/mini_fo/file.c @@ -0,0 +1,713 @@ +/* + * Copyright (c) 1997-2003 Erez Zadok + * Copyright (c) 2001-2003 Stony Brook University + * + * For specific licensing information, see the COPYING file distributed with + * this package, or get one from ftp://ftp.filesystems.org/pub/fist/COPYING. + * + * This Copyright notice must be kept intact and distributed with all + * fistgen sources INCLUDING sources generated by fistgen. + */ +/* + * Copyright (C) 2004, 2005 Markus Klotzbuecher + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ + +/* + * $Id$ + */ + +#ifdef HAVE_CONFIG_H +# include +#endif + +#include "fist.h" +#include "mini_fo.h" +#define ROUND_UP(x) (((x)+sizeof(long)-1) & ~(sizeof(long)-1)) + +/******************* + * File Operations * + *******************/ + +STATIC loff_t +mini_fo_llseek(file_t *file, loff_t offset, int origin) +{ + loff_t err; + file_t *hidden_file = NULL; + + if(S_ISDIR(file->f_dentry->d_inode->i_mode)) { + /* Check if trying to llseek from a directory */ + err = -EISDIR; + goto out; + } + if (ftopd(file) != NULL) { + if(ftohf2(file)) { + hidden_file = ftohf2(file); + } else { + hidden_file = ftohf(file); + } + } + + /* always set hidden position to this one */ + hidden_file->f_pos = file->f_pos; + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) + memcpy(&(hidden_file->f_ra), + &(file->f_ra), + sizeof(struct file_ra_state)); +#else + if (file->f_reada) { /* update readahead information if needed */ + hidden_file->f_reada = file->f_reada; + hidden_file->f_ramax = file->f_ramax; + hidden_file->f_raend = file->f_raend; + hidden_file->f_ralen = file->f_ralen; + hidden_file->f_rawin = file->f_rawin; + } +#endif + if (hidden_file->f_op && hidden_file->f_op->llseek) + err = hidden_file->f_op->llseek(hidden_file, offset, origin); + else + err = generic_file_llseek(hidden_file, offset, origin); + + if (err < 0) + goto out; + + if (err != file->f_pos) { + file->f_pos = err; + // ION maybe this? + // file->f_pos = hidden_file->f_pos; +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0) + file->f_reada = 0; +#endif + file->f_version++; + } + + out: + return err; +} + + +/* mk: fanout capable */ +STATIC ssize_t +mini_fo_read(file_t *file, char *buf, size_t count, loff_t *ppos) +{ + int err = -EINVAL; + file_t *hidden_file = NULL; + loff_t pos = *ppos; + + if(S_ISDIR(file->f_dentry->d_inode->i_mode)) { + /* Check if trying to read from a directory */ + /* printk(KERN_CRIT "mini_fo_read: ERROR: trying to read data from a directory.\n"); */ + err = -EISDIR; + goto out; + } + + if (ftopd(file) != NULL) { + if(ftohf2(file)) { + hidden_file = ftohf2(file); + } else { + hidden_file = ftohf(file); + } + } + + if (!hidden_file->f_op || !hidden_file->f_op->read) + goto out; + + err = hidden_file->f_op->read(hidden_file, buf, count, &pos); + *ppos = pos; + + if (err >= 0) { + /* atime should also be updated for reads of size zero or more */ + fist_copy_attr_atime(file->f_dentry->d_inode, + hidden_file->f_dentry->d_inode); + } + +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0) + /* + * MAJOR HACK + * because pread() does not have any way to tell us that it is + * our caller, then we don't know for sure if we have to update + * the file positions. This hack relies on read() having passed us + * the "real" pointer of its struct file's f_pos field. + */ + if (ppos == &file->f_pos) + hidden_file->f_pos = *ppos = pos; + if (hidden_file->f_reada) { /* update readahead information if needed */ + file->f_reada = hidden_file->f_reada; + file->f_ramax = hidden_file->f_ramax; + file->f_raend = hidden_file->f_raend; + file->f_ralen = hidden_file->f_ralen; + file->f_rawin = hidden_file->f_rawin; + } +#else + memcpy(&(file->f_ra),&(hidden_file->f_ra),sizeof(struct file_ra_state)); +#endif + + out: + return err; +} + + +/* this mini_fo_write() does not modify data pages! */ +STATIC ssize_t +mini_fo_write(file_t *file, const char *buf, size_t count, loff_t *ppos) +{ + int err = -EINVAL; + file_t *hidden_file = NULL; + inode_t *inode; + inode_t *hidden_inode; + loff_t pos = *ppos; + + /* mk: fan out: */ + if (ftopd(file) != NULL) { + if(ftohf2(file)) { + hidden_file = ftohf2(file); + } else { + /* This is bad! We have no storage file to write to. This + * should never happen because if a file is opened for + * writing, a copy should have been made earlier. + */ + printk(KERN_CRIT "mini_fo: write : ERROR, no storage file to write.\n"); + err = -EINVAL; + goto out; + } + } + + inode = file->f_dentry->d_inode; + hidden_inode = itohi2(inode); + if(!hidden_inode) { + printk(KERN_CRIT "mini_fo: write: no sto inode found, not good.\n"); + goto out; + } + + if (!hidden_file->f_op || !hidden_file->f_op->write) + goto out; + + /* adjust for append -- seek to the end of the file */ + if (file->f_flags & O_APPEND) + pos = inode->i_size; + + err = hidden_file->f_op->write(hidden_file, buf, count, &pos); + + /* + * copy ctime and mtime from lower layer attributes + * atime is unchanged for both layers + */ + if (err >= 0) + fist_copy_attr_times(inode, hidden_inode); + + *ppos = pos; +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0) + /* + * XXX: MAJOR HACK + * + * because pwrite() does not have any way to tell us that it is + * our caller, then we don't know for sure if we have to update + * the file positions. This hack relies on write() having passed us + * the "real" pointer of its struct file's f_pos field. + */ + if (ppos == &file->f_pos) + hidden_file->f_pos = *ppos = pos; +#endif + /* update this inode's size */ + if (pos > inode->i_size) + inode->i_size = pos; + + out: + return err; +} + +/* Global variable to hold a file_t pointer. + * This serves to allow mini_fo_filldir function to know which file is + * beeing read, which is required for two reasons: + * + * - be able to call wol functions in order to avoid listing deleted + * base files. + * - if we're reading a directory which is in state 1, we need to + * maintain a list (in mini_fo_filldir) of which files allready + * have been copied to userspace,to detect files existing in base + * and storage and not list them twice. + */ +filldir_t mini_fo_filldir_orig; +file_t *mini_fo_filldir_file; + +/* mainly copied from fs/readdir.c */ +STATIC int +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,18) +mini_fo_filldir(void * __buf, const char * name, int namlen, loff_t offset, + u64 ino, unsigned int d_type) +#else +mini_fo_filldir(void * __buf, const char * name, int namlen, loff_t offset, + ino_t ino, unsigned int d_type) +#endif +{ + struct getdents_callback * buf = (struct getdents_callback *) __buf; + file_t* file = mini_fo_filldir_file; + + /* In theses states we filter meta files in storage (WOL) */ + if(file && (dtopd(file->f_dentry)->state == MODIFIED || + dtopd(file->f_dentry)->state == CREATED || + dtopd(file->f_dentry)->state == DEL_REWRITTEN)) { + + int tmp = strlen(META_FILENAME); + if(tmp == namlen) { + if(!strncmp(name, META_FILENAME, namlen)) + return 0; + } + } + + /* check if we are merging the contents of storage and base */ + if(file && dtopd(file->f_dentry)->state == MODIFIED) { + /* check if we are still reading storage contents, if + * yes, we just save the name of the file for duplicate + * checking later. */ + + if(!ftopd(file)->rd.sto_done) { + /* put file into ndl list */ + if(ndl_add_entry(&ftopd(file)->rd, name, namlen)) + printk(KERN_CRIT "mini_fo_filldir: Error adding to ndl.\n"); + } else { + /* check if file has been deleted */ + if(meta_check_d_entry(file->f_dentry, name, namlen)) + return 0; + + /* do duplicate checking */ + if(ndl_check_entry(&ftopd(file)->rd, name, namlen)) + return 0; + } + } + + return mini_fo_filldir_orig(buf, name, namlen, offset, ino, d_type); +} + + +STATIC int +mini_fo_readdir(file_t *file, void *dirent, filldir_t filldir) +{ + int err = 0;/* mk: ??? -ENOTDIR; */ + file_t *hidden_file = NULL; + file_t *hidden_sto_file = NULL; + inode_t *inode; + struct getdents_callback *buf; + int oldcount; + +#if defined(FIST_FILTER_NAME) || defined(FIST_FILTER_SCA) + struct mini_fo_getdents_callback buf; +#endif /* FIST_FILTER_NAME || FIST_FILTER_SCA */ + + buf = (struct getdents_callback *) dirent; + oldcount = buf->count; + inode = file->f_dentry->d_inode; + mini_fo_filldir_file = file; + mini_fo_filldir_orig = filldir; + + ftopd(file)->rd.sto_done = 0; + do { + if (ftopd(file) != NULL) { + if(ftohf2(file)) { + hidden_sto_file = ftohf2(file); + err = vfs_readdir(hidden_sto_file, mini_fo_filldir, dirent); + file->f_pos = hidden_sto_file->f_pos; + if (err > 0) + fist_copy_attr_atime(inode, hidden_sto_file->f_dentry->d_inode); + /* not finshed yet, we'll be called again */ + if (buf->count != oldcount) + break; + } + + ftopd(file)->rd.sto_done = 1; + + if(ftohf(file)) { + hidden_file = ftohf(file); + err = vfs_readdir(hidden_file, mini_fo_filldir, dirent); + file->f_pos = hidden_file->f_pos; + if (err > 0) + fist_copy_attr_atime(inode, hidden_file->f_dentry->d_inode); + } + + } + } while (0); + + /* mk: + * we need to check if all the directory data has been copied to userspace, + * or if we will be called again by userspace to complete the operation. + */ + if(buf->count == oldcount) { + ndl_put_list(&ftopd(file)->rd); + } + + /* unset this, safe */ + mini_fo_filldir_file = NULL; + return err; +} + + +STATIC unsigned int +mini_fo_poll(file_t *file, poll_table *wait) +{ + unsigned int mask = DEFAULT_POLLMASK; + file_t *hidden_file = NULL; + + if (ftopd(file) != NULL) { + if(ftohf2(file)) { + hidden_file = ftohf2(file); + } else { + hidden_file = ftohf(file); + } + } + + if (!hidden_file->f_op || !hidden_file->f_op->poll) + goto out; + + mask = hidden_file->f_op->poll(hidden_file, wait); + + out: + return mask; +} + +/* FIST-LITE special version of mmap */ +STATIC int +mini_fo_mmap(file_t *file, vm_area_t *vma) +{ + int err = 0; + file_t *hidden_file = NULL; + + /* fanout capability */ + if (ftopd(file) != NULL) { + if(ftohf2(file)) { + hidden_file = ftohf2(file); + } else { + hidden_file = ftohf(file); + } + } + + ASSERT(hidden_file != NULL); + ASSERT(hidden_file->f_op != NULL); + ASSERT(hidden_file->f_op->mmap != NULL); + + vma->vm_file = hidden_file; + err = hidden_file->f_op->mmap(hidden_file, vma); + get_file(hidden_file); /* make sure it doesn't get freed on us */ + fput(file); /* no need to keep extra ref on ours */ + + return err; +} + + + +STATIC int +mini_fo_open(inode_t *inode, file_t *file) +{ + int err = 0; + int hidden_flags; + file_t *hidden_file = NULL; + dentry_t *hidden_dentry = NULL; + + /* fanout stuff */ + file_t *hidden_sto_file = NULL; + dentry_t *hidden_sto_dentry = NULL; + + __ftopd(file) = + kmalloc(sizeof(struct mini_fo_file_info), GFP_KERNEL); + if (!ftopd(file)) { + err = -ENOMEM; + goto out; + } + + /* init the readdir_helper structure */ + INIT_LIST_HEAD(&ftopd(file)->rd.ndl_list); + ftopd(file)->rd.ndl_size = 0; + + /* In certain paths this could stay uninitalized and cause trouble */ + ftohf(file) = NULL; + ftohf2(file) = NULL; + hidden_flags = file->f_flags; + + /* create storage files? */ + if(dtost(file->f_dentry) == UNMODIFIED) { + if(!IS_WRITE_FLAG(file->f_flags)) { + hidden_dentry = dtohd(file->f_dentry); + dget(hidden_dentry); + /* dentry_open will decrement mnt refcnt if err. + * otherwise fput() will do an mntput() for us upon file close. */ + mntget(stopd(inode->i_sb)->hidden_mnt); + hidden_file = dentry_open(hidden_dentry, + stopd(inode->i_sb)->hidden_mnt, + hidden_flags); + if (IS_ERR(hidden_file)) { + err = PTR_ERR(hidden_file); + dput(hidden_dentry); + goto out; + } + ftohf(file) = hidden_file; /* link two files */ + goto out; + } + else { + if(S_ISDIR(file->f_dentry->d_inode->i_mode)) { + err = dir_unmod_to_mod(file->f_dentry); + } else + err = nondir_unmod_to_mod(file->f_dentry, 1); + + if (err) { + printk("mini_fo_open: ERROR creating storage file.\n"); + goto out; + } + } + } + hidden_sto_dentry = dtohd2(file->f_dentry); + dget(hidden_sto_dentry); + + if(dtopd(file->f_dentry)->state == MODIFIED) { + /* Directorys are special, interpose on both lower level files */ + if(S_ISDIR(itohi(inode)->i_mode)) { + /* check for invalid file types of lower level files */ + if(!(S_ISDIR(itohi(inode)->i_mode) && S_ISDIR(itohi2(inode)->i_mode))) { + printk(KERN_CRIT "mini_fo_open: meta data corruption detected.\n"); + dput(hidden_sto_dentry); + err = -EINVAL; + goto out; + } + + /* lower level directorys are ok, open the base file */ + hidden_dentry = dtohd(file->f_dentry); + dget(hidden_dentry); + + mntget(stopd(inode->i_sb)->hidden_mnt); + hidden_file = dentry_open(hidden_dentry, + stopd(inode->i_sb)->hidden_mnt, + hidden_flags); + if (IS_ERR(hidden_file)) { + err = PTR_ERR(hidden_file); + dput(hidden_dentry); + dput(hidden_sto_dentry); + goto out; + } + ftohf(file) = hidden_file; /* link the two files */ + } + } + + if(!exists_in_storage(file->f_dentry)) { + printk(KERN_CRIT "mini_fo_open: invalid file state detected.\n"); + err = -EINVAL; + dput(hidden_sto_dentry); + + /* If the base file has been opened, we need to close it here */ + if(ftohf(file)) { + if (hidden_file->f_op && hidden_file->f_op->flush) +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,18) + hidden_file->f_op->flush(hidden_file, NULL); +#else + hidden_file->f_op->flush(hidden_file); +#endif + dput(hidden_dentry); + } + goto out; + } + + /* ok, now we can safely open the storage file */ + mntget(stopd(inode->i_sb)->hidden_mnt2); + hidden_sto_file = dentry_open(hidden_sto_dentry, + stopd(inode->i_sb)->hidden_mnt2, + hidden_flags); + + /* dentry_open dputs the dentry if it fails */ + if (IS_ERR(hidden_sto_file)) { + err = PTR_ERR(hidden_sto_file); + /* close base file if open */ + if(ftohf(file)) { + if (hidden_file->f_op && hidden_file->f_op->flush) +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,18) + hidden_file->f_op->flush(hidden_file, NULL); +#else + hidden_file->f_op->flush(hidden_file); +#endif + dput(hidden_dentry); + } + goto out; + } + ftohf2(file) = hidden_sto_file; /* link storage file */ + + out: + if (err < 0 && ftopd(file)) { + kfree(ftopd(file)); + } + return err; +} + +STATIC int +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,18) +mini_fo_flush(file_t *file, fl_owner_t id) +#else +mini_fo_flush(file_t *file) +#endif +{ + int err1 = 0; /* assume ok (see open.c:close_fp) */ + int err2 = 0; + file_t *hidden_file = NULL; + + check_mini_fo_file(file); + + /* mk: we don't do any state checking here, as its not worth the time. + * Just flush the lower level files if they exist. + */ + if(ftopd(file) != NULL) { + if(ftohf(file) != NULL) { + hidden_file = ftohf(file); + if (hidden_file->f_op && hidden_file->f_op->flush) +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,18) + err1 = hidden_file->f_op->flush(hidden_file, id); +#else + err1 = hidden_file->f_op->flush(hidden_file); +#endif + } + if(ftohf2(file) != NULL) { + hidden_file = ftohf2(file); + if (hidden_file->f_op && hidden_file->f_op->flush) +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,18) + err2 = hidden_file->f_op->flush(hidden_file, id); +#else + err2 = hidden_file->f_op->flush(hidden_file); +#endif + } + } + return (err1 | err2); +} + + +STATIC int +mini_fo_release(inode_t *inode, file_t *file) +{ + int err = 0; + file_t *hidden_file = NULL; + + if (ftopd(file) != NULL) { + if(ftohf(file)) { + hidden_file = ftohf(file); + fput(hidden_file); + } + if(ftohf2(file)) { + hidden_file = ftohf2(file); + fput(hidden_file); + } + kfree(ftopd(file)); + } + return err; +} + +STATIC int +mini_fo_fsync(file_t *file, dentry_t *dentry, int datasync) +{ + int err1 = 0; + int err2 = 0; + file_t *hidden_file = NULL; + dentry_t *hidden_dentry; + + check_mini_fo_file(file); + + if ((hidden_file = ftohf(file)) != NULL) { + hidden_dentry = dtohd(dentry); + if (hidden_file->f_op && hidden_file->f_op->fsync) { +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&hidden_dentry->d_inode->i_mutex); +#else + down(&hidden_dentry->d_inode->i_sem); +#endif + err1 = hidden_file->f_op->fsync(hidden_file, hidden_dentry, datasync); +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&hidden_dentry->d_inode->i_mutex); +#else + up(&hidden_dentry->d_inode->i_sem); +#endif + } + } + + if ((hidden_file = ftohf2(file)) != NULL) { + hidden_dentry = dtohd2(dentry); + if (hidden_file->f_op && hidden_file->f_op->fsync) { +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&hidden_dentry->d_inode->i_mutex); +#else + down(&hidden_dentry->d_inode->i_sem); +#endif + err2 = hidden_file->f_op->fsync(hidden_file, hidden_dentry, datasync); +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&hidden_dentry->d_inode->i_mutex); +#else + up(&hidden_dentry->d_inode->i_sem); +#endif + } + } + else + goto err; + +err: + return (err1 || err2); +} + + +STATIC int +mini_fo_fasync(int fd, file_t *file, int flag) +{ + int err1 = 0; + int err2 = 0; + + file_t *hidden_file = NULL; + + check_mini_fo_file(file); + + if((hidden_file = ftohf(file)) != NULL) { + err1 = hidden_file->f_op->fasync(fd, hidden_file, flag); + } + if((hidden_file = ftohf2(file)) != NULL) { + err2 = hidden_file->f_op->fasync(fd, hidden_file, flag); + } + + return (err1 || err2); +} + + + +struct file_operations mini_fo_dir_fops = + { + read: generic_read_dir, + write: mini_fo_write, + readdir: mini_fo_readdir, + poll: mini_fo_poll, + /* ioctl: mini_fo_ioctl, */ + mmap: mini_fo_mmap, + open: mini_fo_open, + flush: mini_fo_flush, + release: mini_fo_release, + fsync: mini_fo_fsync, + fasync: mini_fo_fasync, + /* not needed lock: mini_fo_lock, */ + /* not needed: readv */ + /* not needed: writev */ + /* not implemented: sendpage */ + /* not implemented: get_unmapped_area */ + }; + +struct file_operations mini_fo_main_fops = + { + llseek: mini_fo_llseek, + read: mini_fo_read, + write: mini_fo_write, + readdir: mini_fo_readdir, + poll: mini_fo_poll, + /* ioctl: mini_fo_ioctl, */ + mmap: mini_fo_mmap, + open: mini_fo_open, + flush: mini_fo_flush, + release: mini_fo_release, + fsync: mini_fo_fsync, + fasync: mini_fo_fasync, + /* not needed: lock: mini_fo_lock, */ + /* not needed: readv */ + /* not needed: writev */ + /* not implemented: sendpage */ + /* not implemented: get_unmapped_area */ + }; Index: linux-2.6.25.1/fs/mini_fo/fist.h =================================================================== --- /dev/null +++ linux-2.6.25.1/fs/mini_fo/fist.h @@ -0,0 +1,252 @@ +/* + * Copyright (c) 1997-2003 Erez Zadok + * Copyright (c) 2001-2003 Stony Brook University + * + * For specific licensing information, see the COPYING file distributed with + * this package, or get one from ftp://ftp.filesystems.org/pub/fist/COPYING. + * + * This Copyright notice must be kept intact and distributed with all + * fistgen sources INCLUDING sources generated by fistgen. + */ +/* + * Copyright (C) 2004, 2005 Markus Klotzbuecher + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ + + +/* + * $Id$ + */ + +#ifndef __FIST_H_ +#define __FIST_H_ + +/* + * KERNEL ONLY CODE: + */ +#ifdef __KERNEL__ +#include +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,19) +#include +#else +#include +#endif +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0) +#ifdef CONFIG_MODVERSIONS +# define MODVERSIONS +# include +#endif /* CONFIG_MODVERSIONS */ +#endif /* KERNEL_VERSION < 2.6.0 */ +#include +#include +#include +#include +#include +#include +#include +#include +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0) +#include +#else +#include +#include +#include +#include +#include +#include +#include +#include +#endif +#include +#include +#include +#include +#include +#include +#include +#include + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,20) +#include +#endif + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) +#include +#endif + +#include + +#include +/* #include */ +#include +#include + +/* + * MACROS: + */ + +/* those mapped to ATTR_* were copied from linux/fs.h */ +#define FA_MODE ATTR_MODE +#define FA_UID ATTR_UID +#define FA_GID ATTR_GID +#define FA_SIZE ATTR_SIZE +#define FA_ATIME ATTR_ATIME +#define FA_MTIME ATTR_MTIME +#define FA_CTIME ATTR_CTIME +#define FA_ATIME_SET ATTR_ATIME_SET +#define FA_MTIME_SET ATTR_MTIME_SET +#define FA_FORCE ATTR_FORCE +#define FA_ATTR_FLAGS ATTR_ATTR_FLAG + +/* must be greater than all other ATTR_* flags! */ +#define FA_NLINK 2048 +#define FA_BLKSIZE 4096 +#define FA_BLOCKS 8192 +#define FA_TIMES (FA_ATIME|FA_MTIME|FA_CTIME) +#define FA_ALL 0 + +/* macros to manage changes between kernels */ +#define INODE_DATA(i) (&(i)->i_data) + +#define MIN(x,y) ((x < y) ? (x) : (y)) +#define MAX(x,y) ((x > y) ? (x) : (y)) +#define MAXPATHLEN PATH_MAX + +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,5) +# define lookup_one_len(a,b,c) lookup_one(a,b) +#endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2,4,5) */ + +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,8) +# define generic_file_llseek default_llseek +#endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2,4,8) */ + +#ifndef SEEK_SET +# define SEEK_SET 0 +#endif /* not SEEK_SET */ + +#ifndef SEEK_CUR +# define SEEK_CUR 1 +#endif /* not SEEK_CUR */ + +#ifndef SEEK_END +# define SEEK_END 2 +#endif /* not SEEK_END */ + +#ifndef DEFAULT_POLLMASK +# define DEFAULT_POLLMASK (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM) +#endif /* not DEFAULT_POLLMASK */ + +/* XXX: fix this so fistgen generates kfree() code directly */ +#define kfree_s(a,b) kfree(a) + +/* + * TYPEDEFS: + */ +typedef struct dentry dentry_t; +typedef struct file file_t; +typedef struct inode inode_t; +typedef inode_t vnode_t; +typedef struct page page_t; +typedef struct qstr qstr_t; +typedef struct super_block super_block_t; +typedef super_block_t vfs_t; +typedef struct vm_area_struct vm_area_t; + + +/* + * EXTERNALS: + */ + +#define FPPF(str,page) printk("PPF %s 0x%x/%d: Lck:%d Err:%d Ref:%d Upd:%d Other::%d:%d:%d:%d:\n", \ + str, \ + (int) page, \ + (int) page->index, \ + (PageLocked(page) ? 1 : 0), \ + (PageError(page) ? 1 : 0), \ + (PageReferenced(page) ? 1 : 0), \ + (Page_Uptodate(page) ? 1 : 0), \ + (PageDecrAfter(page) ? 1 : 0), \ + (PageSlab(page) ? 1 : 0), \ + (PageSwapCache(page) ? 1 : 0), \ + (PageReserved(page) ? 1 : 0) \ + ) +#define EZKDBG printk("EZK %s:%d:%s\n",__FILE__,__LINE__,__FUNCTION__) +#if 0 +# define EZKDBG1 printk("EZK %s:%d\n",__FILE__,__LINE__) +#else +# define EZKDBG1 +#endif + +extern int fist_get_debug_value(void); +extern int fist_set_debug_value(int val); +#if 0 /* mini_fo doesn't need these */ +extern void fist_dprint_internal(int level, char *str,...); +extern void fist_print_dentry(char *str, const dentry_t *dentry); +extern void fist_print_inode(char *str, const inode_t *inode); +extern void fist_print_file(char *str, const file_t *file); +extern void fist_print_buffer_flags(char *str, struct buffer_head *buffer); +extern void fist_print_page_flags(char *str, page_t *page); +extern void fist_print_page_bytes(char *str, page_t *page); +extern void fist_print_pte_flags(char *str, const page_t *page); +extern void fist_checkinode(inode_t *inode, char *msg); +extern void fist_print_sb(char *str, const super_block_t *sb); + +/* ยง$% by mk: special debug functions */ +extern void fist_mk_print_dentry(char *str, const dentry_t *dentry); +extern void fist_mk_print_inode(char *str, const inode_t *inode); + +extern char *add_indent(void); +extern char *del_indent(void); +#endif/* mini_fo doesn't need these */ + + +#define STATIC +#define ASSERT(EX) \ +do { \ + if (!(EX)) { \ + printk(KERN_CRIT "ASSERTION FAILED: %s at %s:%d (%s)\n", #EX, \ + __FILE__, __LINE__, __FUNCTION__); \ + (*((char *)0))=0; \ + } \ +} while (0) +/* same ASSERT, but tell me who was the caller of the function */ +#define ASSERT2(EX) \ +do { \ + if (!(EX)) { \ + printk(KERN_CRIT "ASSERTION FAILED (caller): %s at %s:%d (%s)\n", #EX, \ + file, line, func); \ + (*((char *)0))=0; \ + } \ +} while (0) + +#if 0 /* mini_fo doesn't need these */ +#define dprintk(format, args...) printk(KERN_DEBUG format, ##args) +#define fist_dprint(level, str, args...) fist_dprint_internal(level, KERN_DEBUG str, ## args) +#define print_entry_location() fist_dprint(4, "%sIN: %s %s:%d\n", add_indent(), __FUNCTION__, __FILE__, __LINE__) +#define print_exit_location() fist_dprint(4, "%s OUT: %s %s:%d\n", del_indent(), __FUNCTION__, __FILE__, __LINE__) +#define print_exit_status(status) fist_dprint(4, "%s OUT: %s %s:%d, STATUS: %d\n", del_indent(), __FUNCTION__, __FILE__, __LINE__, status) +#define print_exit_pointer(status) \ +do { \ + if (IS_ERR(status)) \ + fist_dprint(4, "%s OUT: %s %s:%d, RESULT: %ld\n", del_indent(), __FUNCTION__, __FILE__, __LINE__, PTR_ERR(status)); \ + else \ + fist_dprint(4, "%s OUT: %s %s:%d, RESULT: 0x%x\n", del_indent(), __FUNCTION__, __FILE__, __LINE__, PTR_ERR(status)); \ +} while (0) +#endif/* mini_fo doesn't need these */ + +#endif /* __KERNEL__ */ + + +/* + * DEFINITIONS FOR USER AND KERNEL CODE: + * (Note: ioctl numbers 1--9 are reserved for fistgen, the rest + * are auto-generated automatically based on the user's .fist file.) + */ +# define FIST_IOCTL_GET_DEBUG_VALUE _IOR(0x15, 1, int) +# define FIST_IOCTL_SET_DEBUG_VALUE _IOW(0x15, 2, int) + +#endif /* not __FIST_H_ */ Index: linux-2.6.25.1/fs/mini_fo/inode.c =================================================================== --- /dev/null +++ linux-2.6.25.1/fs/mini_fo/inode.c @@ -0,0 +1,1564 @@ +/* + * Copyright (c) 1997-2003 Erez Zadok + * Copyright (c) 2001-2003 Stony Brook University + * + * For specific licensing information, see the COPYING file distributed with + * this package, or get one from ftp://ftp.filesystems.org/pub/fist/COPYING. + * + * This Copyright notice must be kept intact and distributed with all + * fistgen sources INCLUDING sources generated by fistgen. + */ +/* + * Copyright (C) 2004, 2005 Markus Klotzbuecher + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ + +/* + * $Id$ + */ + +#ifdef HAVE_CONFIG_H +# include +#endif + +#include "fist.h" +#include "mini_fo.h" + +STATIC int +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) +mini_fo_create(inode_t *dir, dentry_t *dentry, int mode, struct nameidata *nd) +#else +mini_fo_create(inode_t *dir, dentry_t *dentry, int mode) +#endif +{ + int err = 0; + + check_mini_fo_dentry(dentry); + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) + err = create_sto_reg_file(dentry, mode, nd); +#else + err = create_sto_reg_file(dentry, mode); +#endif + check_mini_fo_dentry(dentry); + return err; +} + + +STATIC dentry_t * +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) +mini_fo_lookup(inode_t *dir, dentry_t *dentry, struct nameidata* nd) +#else +mini_fo_lookup(inode_t *dir, dentry_t *dentry) +#endif +{ + int err = 0; + dentry_t *hidden_dir_dentry; + dentry_t *hidden_dentry = NULL; + + dentry_t *hidden_sto_dir_dentry; + dentry_t *hidden_sto_dentry = NULL; + + /* whiteout flag */ + int del_flag = 0; + char *bpath = NULL; + + const char *name; + unsigned int namelen; + + /* Don't allow lookups of META-files */ + namelen = strlen(META_FILENAME); + if(namelen == dentry->d_name.len) { + if(!strncmp(dentry->d_name.name, META_FILENAME, namelen)) { + err = -ENOENT; + goto out; + } + } + + hidden_dir_dentry = dtohd(dentry->d_parent); + hidden_sto_dir_dentry = dtohd2(dentry->d_parent); + + name = dentry->d_name.name; + namelen = dentry->d_name.len; + + /* must initialize dentry operations */ + dentry->d_op = &mini_fo_dops; + + /* setup the del_flag */ + del_flag = __meta_check_d_entry(dir, name, namelen); + bpath = __meta_check_r_entry(dir, name, namelen); + + /* perform the lookups of base and storage files: + * + * This caused some serious trouble, as a lookup_one_len passing + * a negative dentry oopses. Solution is to only do the lookup + * if the dentry is positive, else we set it to NULL + * More trouble, who said a *_dir_dentry can't be NULL? + */ + if(bpath) { + /* Cross-Interposing (C), yeah! */ + hidden_dentry = bpath_walk(dir->i_sb, bpath); + if(!hidden_dentry || !hidden_dentry->d_inode) { + printk(KERN_CRIT "mini_fo_lookup: bpath_walk failed.\n"); + err= -EINVAL; + goto out; + } + + /* this can be set up safely without fear of spaghetti + * interposing as it is only used for copying times */ + hidden_dir_dentry = hidden_dentry->d_parent; + kfree(bpath); + } + else if(hidden_dir_dentry && hidden_dir_dentry->d_inode) + hidden_dentry = + lookup_one_len(name, hidden_dir_dentry, namelen); + else + hidden_dentry = NULL; + + if(hidden_sto_dir_dentry && hidden_sto_dir_dentry->d_inode) + hidden_sto_dentry = + lookup_one_len(name, hidden_sto_dir_dentry, namelen); + else + hidden_sto_dentry = NULL; + + /* catch error in lookup */ + if (IS_ERR(hidden_dentry) || IS_ERR(hidden_sto_dentry)) { + /* mk: we need to call dput on the dentry, whose + * lookup_one_len operation failed, in order to avoid + * unmount trouble. + */ + if(IS_ERR(hidden_dentry)) { + printk(KERN_CRIT "mini_fo_lookup: ERR from base dentry, lookup failed.\n"); + err = PTR_ERR(hidden_dentry); + } else { + dput(hidden_dentry); + } + if(IS_ERR(hidden_sto_dentry)) { + printk(KERN_CRIT "mini_fo_lookup: ERR from storage dentry, lookup failed.\n"); + err = PTR_ERR(hidden_sto_dentry); + } else { + dput(hidden_sto_dentry); + } + goto out; + } + + /* allocate dentry private data */ + __dtopd(dentry) = (struct mini_fo_dentry_info *) + kmalloc(sizeof(struct mini_fo_dentry_info), GFP_KERNEL); + + if (!dtopd(dentry)) { + err = -ENOMEM; + goto out_dput; + } + + /* check for different states of the mini_fo file to be looked up. */ + + /* state 1, file has been modified */ + if(hidden_dentry && hidden_sto_dentry && + hidden_dentry->d_inode && hidden_sto_dentry->d_inode && !del_flag) { + + /* update parent directory's atime */ + fist_copy_attr_atime(dir, hidden_sto_dir_dentry->d_inode); + + dtopd(dentry)->state = MODIFIED; + dtohd(dentry) = hidden_dentry; + dtohd2(dentry) = hidden_sto_dentry; + + err = mini_fo_tri_interpose(hidden_dentry, + hidden_sto_dentry, + dentry, dir->i_sb, 1); + if (err) { + printk(KERN_CRIT "mini_fo_lookup: error interposing (state1).\n"); + goto out_free; + } + goto out; + } + /* state 2, file is unmodified */ + if(hidden_dentry && hidden_dentry->d_inode && !del_flag) { + + fist_copy_attr_atime(dir, hidden_dir_dentry->d_inode); + + dtopd(dentry)->state = UNMODIFIED; + dtohd(dentry) = hidden_dentry; + dtohd2(dentry) = hidden_sto_dentry; /* could be negative */ + + err = mini_fo_tri_interpose(hidden_dentry, + hidden_sto_dentry, + dentry, dir->i_sb, 1); + if (err) { + printk(KERN_CRIT "mini_fo_lookup: error interposing (state2).\n"); + goto out_free; + } + goto out; + } + /* state 3, file has been newly created */ + if(hidden_sto_dentry && hidden_sto_dentry->d_inode && !del_flag) { + + fist_copy_attr_atime(dir, hidden_sto_dir_dentry->d_inode); + dtopd(dentry)->state = CREATED; + dtohd(dentry) = hidden_dentry; /* could be negative */ + dtohd2(dentry) = hidden_sto_dentry; + + err = mini_fo_tri_interpose(hidden_dentry, + hidden_sto_dentry, + dentry, dir->i_sb, 1); + if (err) { + printk(KERN_CRIT "mini_fo_lookup: error interposing (state3).\n"); + goto out_free; + } + goto out; + } + + /* state 4, file has deleted and created again. */ + if(hidden_dentry && hidden_sto_dentry && + hidden_dentry->d_inode && + hidden_sto_dentry->d_inode && del_flag) { + + fist_copy_attr_atime(dir, hidden_sto_dir_dentry->d_inode); + dtopd(dentry)->state = DEL_REWRITTEN; + dtohd(dentry) = NULL; + dtohd2(dentry) = hidden_sto_dentry; + + err = mini_fo_tri_interpose(NULL, + hidden_sto_dentry, + dentry, dir->i_sb, 1); + if (err) { + printk(KERN_CRIT "mini_fo_lookup: error interposing (state4).\n"); + goto out_free; + } + /* We will never need this dentry again, as the file has been + * deleted from base */ + dput(hidden_dentry); + goto out; + } + /* state 5, file has been deleted in base */ + if(hidden_dentry && hidden_sto_dentry && + hidden_dentry->d_inode && + !hidden_sto_dentry->d_inode && del_flag) { + + /* check which parents atime we need for updating */ + if(hidden_sto_dir_dentry->d_inode) + fist_copy_attr_atime(dir, + hidden_sto_dir_dentry->d_inode); + else + fist_copy_attr_atime(dir, + hidden_dir_dentry->d_inode); + + dtopd(dentry)->state = DELETED; + dtohd(dentry) = NULL; + dtohd2(dentry) = hidden_sto_dentry; + + /* add negative dentry to dcache to speed up lookups */ + d_add(dentry, NULL); + dput(hidden_dentry); + goto out; + } + /* state 6, file does not exist */ + if(((hidden_dentry && !hidden_dentry->d_inode) || + (hidden_sto_dentry && !hidden_sto_dentry->d_inode)) && !del_flag) + { + /* check which parents atime we need for updating */ + if(hidden_sto_dir_dentry && hidden_sto_dir_dentry->d_inode) + fist_copy_attr_atime(dir, hidden_sto_dir_dentry->d_inode); + else + fist_copy_attr_atime(dir, hidden_dir_dentry->d_inode); + + dtopd(dentry)->state = NON_EXISTANT; + dtohd(dentry) = hidden_dentry; + dtohd2(dentry) = hidden_sto_dentry; + d_add(dentry, NULL); + goto out; + } + + /* if we get to here, were in an invalid state. bad. */ + printk(KERN_CRIT "mini_fo_lookup: ERROR, meta data corruption detected.\n"); + + /* end state checking */ + out_free: + d_drop(dentry); /* so that our bad dentry will get destroyed */ + kfree(dtopd(dentry)); + __dtopd(dentry) = NULL; /* be safe */ + + out_dput: + if(hidden_dentry) + dput(hidden_dentry); + if(hidden_sto_dentry) + dput(hidden_sto_dentry); /* drops usage count and marks for release */ + + out: + /* initalize wol if file exists and is directory */ + if(dentry->d_inode) { + if(S_ISDIR(dentry->d_inode->i_mode)) { + itopd(dentry->d_inode)->deleted_list_size = -1; + itopd(dentry->d_inode)->renamed_list_size = -1; + meta_build_lists(dentry); + } + } + return ERR_PTR(err); +} + + +STATIC int +mini_fo_link(dentry_t *old_dentry, inode_t *dir, dentry_t *new_dentry) +{ + int err; + dentry_t *hidden_old_dentry; + dentry_t *hidden_new_dentry; + dentry_t *hidden_dir_dentry; + + + check_mini_fo_dentry(old_dentry); + check_mini_fo_dentry(new_dentry); + check_mini_fo_inode(dir); + + /* no links to directorys and existing targets target allowed */ + if(S_ISDIR(old_dentry->d_inode->i_mode) || + is_mini_fo_existant(new_dentry)) { + err = -EPERM; + goto out; + } + + /* bring it directly from unmod to del_rew */ + if(dtost(old_dentry) == UNMODIFIED) { + err = nondir_unmod_to_mod(old_dentry, 1); + if(err) { + err = -EINVAL; + goto out; + } + err = meta_add_d_entry(old_dentry->d_parent, + old_dentry->d_name.name, + old_dentry->d_name.len); + if(err) { + err = -EINVAL; + goto out; + } + dput(dtohd(old_dentry)); + dtohd(old_dentry) = NULL; + dtost(old_dentry) = DEL_REWRITTEN; + } + + err = get_neg_sto_dentry(new_dentry); + if(err) { + err = -EINVAL; + goto out; + } + + hidden_old_dentry = dtohd2(old_dentry); + hidden_new_dentry = dtohd2(new_dentry); + + dget(hidden_old_dentry); + dget(hidden_new_dentry); + + /* was: hidden_dir_dentry = lock_parent(hidden_new_dentry); */ + hidden_dir_dentry = dget(hidden_new_dentry->d_parent); +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&hidden_dir_dentry->d_inode->i_mutex); +#else + down(&hidden_dir_dentry->d_inode->i_sem); +#endif + + err = vfs_link(hidden_old_dentry, + hidden_dir_dentry->d_inode, + hidden_new_dentry); + if (err || !hidden_new_dentry->d_inode) + goto out_lock; + + dtost(new_dentry) = CREATED; + err = mini_fo_tri_interpose(NULL, hidden_new_dentry, new_dentry, dir->i_sb, 0); + if (err) + goto out_lock; + + fist_copy_attr_timesizes(dir, hidden_new_dentry->d_inode); + /* propagate number of hard-links */ + old_dentry->d_inode->i_nlink = itohi2(old_dentry->d_inode)->i_nlink; + + out_lock: + /* was: unlock_dir(hidden_dir_dentry); */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&hidden_dir_dentry->d_inode->i_mutex); +#else + up(&hidden_dir_dentry->d_inode->i_sem); +#endif + dput(hidden_dir_dentry); + + dput(hidden_new_dentry); + dput(hidden_old_dentry); + if (!new_dentry->d_inode) + d_drop(new_dentry); + + out: + return err; +} + + +STATIC int +mini_fo_unlink(inode_t *dir, dentry_t *dentry) +{ + int err = 0; + + dget(dentry); + if(dtopd(dentry)->state == MODIFIED) { + err = nondir_mod_to_del(dentry); + goto out; + } + else if(dtopd(dentry)->state == UNMODIFIED) { + err = nondir_unmod_to_del(dentry); + goto out; + } + else if(dtopd(dentry)->state == CREATED) { + err = nondir_creat_to_del(dentry); + goto out; + } + else if(dtopd(dentry)->state == DEL_REWRITTEN) { + err = nondir_del_rew_to_del(dentry); + goto out; + } + + printk(KERN_CRIT "mini_fo_unlink: ERROR, invalid state detected.\n"); + + out: + fist_copy_attr_times(dir, itohi2(dentry->d_parent->d_inode)); + + if(!err) { + /* is this causing my pain? d_delete(dentry); */ + d_drop(dentry); + } + + dput(dentry); + return err; +} + + +STATIC int +mini_fo_symlink(inode_t *dir, dentry_t *dentry, const char *symname) +{ + int err=0; + dentry_t *hidden_sto_dentry; + dentry_t *hidden_sto_dir_dentry; +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) + umode_t mode; +#endif + + /* Fail if the symlink file exists */ + if(!(dtost(dentry) == DELETED || + dtost(dentry) == NON_EXISTANT)) { + err = -EEXIST; + goto out; + } + + err = get_neg_sto_dentry(dentry); + if(err) { + err = -EINVAL; + goto out; + } + hidden_sto_dentry = dtohd2(dentry); + + dget(hidden_sto_dentry); + /* was: hidden_sto_dir_dentry = lock_parent(hidden_sto_dentry); */ + hidden_sto_dir_dentry = dget(hidden_sto_dentry->d_parent); +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + down(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) + mode = S_IALLUGO; + err = vfs_symlink(hidden_sto_dir_dentry->d_inode, + hidden_sto_dentry, symname, mode); +#else + err = vfs_symlink(hidden_sto_dir_dentry->d_inode, + hidden_sto_dentry, + symname); +#endif + if (err || !hidden_sto_dentry->d_inode) + goto out_lock; + + if(dtost(dentry) == DELETED) { + dtost(dentry) = DEL_REWRITTEN; + err = mini_fo_tri_interpose(NULL, hidden_sto_dentry, dentry, dir->i_sb, 0); + if(err) + goto out_lock; + } else if(dtost(dentry) == NON_EXISTANT) { + dtost(dentry) = CREATED; + err = mini_fo_tri_interpose(dtohd(dentry), hidden_sto_dentry, dentry, dir->i_sb, 0); + if(err) + goto out_lock; + } + fist_copy_attr_timesizes(dir, hidden_sto_dir_dentry->d_inode); + + out_lock: + /* was: unlock_dir(hidden_sto_dir_dentry); */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + up(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + dput(hidden_sto_dir_dentry); + + dput(hidden_sto_dentry); + if (!dentry->d_inode) + d_drop(dentry); + out: + return err; +} + +STATIC int +mini_fo_mkdir(inode_t *dir, dentry_t *dentry, int mode) +{ + int err; + + err = create_sto_dir(dentry, mode); + + check_mini_fo_dentry(dentry); + + return err; +} + + +STATIC int +mini_fo_rmdir(inode_t *dir, dentry_t *dentry) +{ + int err = 0; + + dentry_t *hidden_sto_dentry; + dentry_t *hidden_sto_dir_dentry; + dentry_t *meta_dentry; + inode_t *hidden_sto_dir = NULL; + + check_mini_fo_dentry(dentry); + check_mini_fo_inode(dir); + + dget(dentry); + if(dtopd(dentry)->state == MODIFIED) { + /* XXX: disabled, because it does not bother to check files on + * the original filesystem - just a hack, but better than simply + * removing it without testing */ + err = -EINVAL; + goto out; + + hidden_sto_dir = itohi2(dir); + hidden_sto_dentry = dtohd2(dentry); + + /* was:hidden_sto_dir_dentry = lock_parent(hidden_sto_dentry); */ + hidden_sto_dir_dentry = dget(hidden_sto_dentry->d_parent); +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + down(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + + /* Delete an old WOL file contained in the storage dir */ + meta_dentry = lookup_one_len(META_FILENAME, + hidden_sto_dentry, + strlen(META_FILENAME)); + if(meta_dentry->d_inode) { + err = vfs_unlink(hidden_sto_dentry->d_inode, meta_dentry); + dput(meta_dentry); + if(!err) + d_delete(meta_dentry); + } + + err = vfs_rmdir(hidden_sto_dir, hidden_sto_dentry); + dput(hidden_sto_dentry); + if(!err) + d_delete(hidden_sto_dentry); + + /* propagate number of hard-links */ + dentry->d_inode->i_nlink = itohi2(dentry->d_inode)->i_nlink; + + dput(dtohd(dentry)); + + dtohd(dentry) = NULL; + dtopd(dentry)->state = DELETED; + + /* carefull with R files */ + if( __meta_is_r_entry(dir, + dentry->d_name.name, + dentry->d_name.len) == 1) { + err = meta_remove_r_entry(dentry->d_parent, + dentry->d_name.name, + dentry->d_name.len); + if(err) { + printk(KERN_CRIT "mini_fo: rmdir: meta_remove_r_entry failed.\n"); + goto out; + } + } + else { + /* ok, add deleted file to META */ + meta_add_d_entry(dentry->d_parent, + dentry->d_name.name, + dentry->d_name.len); + } + /* was: unlock_dir(hidden_sto_dir_dentry); */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + up(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + dput(hidden_sto_dir_dentry); + goto out; + } + else if(dtopd(dentry)->state == UNMODIFIED) { + /* XXX: simply adding it to the delete list here is fscking dangerous! + * as a temporary hack, i will disable rmdir on unmodified directories + * for now. + */ + err = -EINVAL; + goto out; + + err = get_neg_sto_dentry(dentry); + if(err) { + err = -EINVAL; + goto out; + } + + /* dput base dentry, this will relase the inode and free the + * dentry, as we will never need it again. */ + dput(dtohd(dentry)); + dtohd(dentry) = NULL; + dtopd(dentry)->state = DELETED; + + /* add deleted file to META-file */ + meta_add_d_entry(dentry->d_parent, + dentry->d_name.name, + dentry->d_name.len); + goto out; + } + else if(dtopd(dentry)->state == CREATED) { + hidden_sto_dir = itohi2(dir); + hidden_sto_dentry = dtohd2(dentry); + + /* was: hidden_sto_dir_dentry = lock_parent(hidden_sto_dentry);*/ + hidden_sto_dir_dentry = dget(hidden_sto_dentry->d_parent); + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + down(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + + /* Delete an old WOL file contained in the storage dir */ + meta_dentry = lookup_one_len(META_FILENAME, + hidden_sto_dentry, + strlen(META_FILENAME)); + if(meta_dentry->d_inode) { + /* is this necessary? dget(meta_dentry); */ + err = vfs_unlink(hidden_sto_dentry->d_inode, + meta_dentry); + dput(meta_dentry); + if(!err) + d_delete(meta_dentry); + } + + err = vfs_rmdir(hidden_sto_dir, hidden_sto_dentry); + dput(hidden_sto_dentry); + if(!err) + d_delete(hidden_sto_dentry); + + /* propagate number of hard-links */ + dentry->d_inode->i_nlink = itohi2(dentry->d_inode)->i_nlink; + dtopd(dentry)->state = NON_EXISTANT; + + /* was: unlock_dir(hidden_sto_dir_dentry); */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + up(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + dput(hidden_sto_dir_dentry); + + goto out; + } + else if(dtopd(dentry)->state == DEL_REWRITTEN) { + hidden_sto_dir = itohi2(dir); + hidden_sto_dentry = dtohd2(dentry); + + /* was: hidden_sto_dir_dentry = lock_parent(hidden_sto_dentry);*/ + hidden_sto_dir_dentry = dget(hidden_sto_dentry->d_parent); + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + down(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + + /* Delete an old WOL file contained in the storage dir */ + meta_dentry = lookup_one_len(META_FILENAME, + hidden_sto_dentry, + strlen(META_FILENAME)); + if(meta_dentry->d_inode) { + /* is this necessary? dget(meta_dentry); */ + err = vfs_unlink(hidden_sto_dentry->d_inode, + meta_dentry); + dput(meta_dentry); + if(!err) + d_delete(meta_dentry); + } + + err = vfs_rmdir(hidden_sto_dir, hidden_sto_dentry); + dput(hidden_sto_dentry); + if(!err) + d_delete(hidden_sto_dentry); + + /* propagate number of hard-links */ + dentry->d_inode->i_nlink = itohi2(dentry->d_inode)->i_nlink; + dtopd(dentry)->state = DELETED; + /* was: unlock_dir(hidden_sto_dir_dentry); */ + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + up(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + dput(hidden_sto_dir_dentry); + goto out; + } + + printk(KERN_CRIT "mini_fo_rmdir: ERROR, invalid state detected.\n"); + + out: + if(!err) { + d_drop(dentry); + } + + fist_copy_attr_times(dir, itohi2(dentry->d_parent->d_inode)); + dput(dentry); + + return err; +} + + +STATIC int +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) +mini_fo_mknod(inode_t *dir, dentry_t *dentry, int mode, dev_t dev) +#else +mini_fo_mknod(inode_t *dir, dentry_t *dentry, int mode, int dev) +#endif +{ + int err = 0; + + check_mini_fo_dentry(dentry); + + err = create_sto_nod(dentry, mode, dev); + if(err) { + printk(KERN_CRIT "mini_fo_mknod: creating sto nod failed.\n"); + err = -EINVAL; + } + + check_mini_fo_dentry(dentry); + return err; +} + + +STATIC int +mini_fo_rename(inode_t *old_dir, dentry_t *old_dentry, + inode_t *new_dir, dentry_t *new_dentry) +{ + /* dispatch */ + if(S_ISDIR(old_dentry->d_inode->i_mode)) + return rename_directory(old_dir, old_dentry, new_dir, new_dentry); + return rename_nondir(old_dir, old_dentry, new_dir, new_dentry); + +} + +int rename_directory(inode_t *old_dir, dentry_t *old_dentry, + inode_t *new_dir, dentry_t *new_dentry) +{ + int err, bpath_len; + char *bpath; + + dentry_t *hidden_old_dentry; + dentry_t *hidden_new_dentry; + dentry_t *hidden_old_dir_dentry; + dentry_t *hidden_new_dir_dentry; + + err = 0; + bpath = NULL; + bpath_len = 0; + + /* this is a test, chuck out if it works */ + if(!(dtopd(new_dentry)->state == DELETED || + dtopd(new_dentry)->state == NON_EXISTANT)) { + printk(KERN_CRIT "mini_fo: rename_directory: \ + uh, ah, new_dentry not negative.\n"); + /* return -1; */ + } + + /* state = UNMODIFIED */ + if(dtopd(old_dentry)->state == UNMODIFIED) { + err = dir_unmod_to_mod(old_dentry); + if (err) + goto out; + } + + /* state = MODIFIED */ + if(dtopd(old_dentry)->state == MODIFIED) { + bpath = meta_check_r_entry(old_dentry->d_parent, + old_dentry->d_name.name, + old_dentry->d_name.len); + if(bpath) { + err = meta_remove_r_entry(old_dentry->d_parent, + old_dentry->d_name.name, + old_dentry->d_name.len); + if(err) { + printk(KERN_CRIT "mini_fo: rename_directory:\ + meta_remove_r_entry \ + failed.\n"); + goto out; + } + err = meta_add_r_entry(new_dentry->d_parent, + bpath, + strlen(bpath), + new_dentry->d_name.name, + new_dentry->d_name.len); + kfree(bpath); + } + else {/* wol it */ + err = meta_add_d_entry(old_dentry->d_parent, + old_dentry->d_name.name, + old_dentry->d_name.len); + if (err) + goto out; + /* put it on rename list */ + err = get_mini_fo_bpath(old_dentry, + &bpath, + &bpath_len); + if (err) + goto out; + err = meta_add_r_entry(new_dentry->d_parent, + bpath, bpath_len, + new_dentry->d_name.name, + new_dentry->d_name.len); + if (err) + goto out; + } + /* no state change, MODIFIED stays MODIFIED */ + } + /* state = CREATED */ + if(dtopd(old_dentry)->state == CREATED || + dtopd(old_dentry)->state == DEL_REWRITTEN) { + if(dtohd(old_dentry)) + dput(dtohd(old_dentry)); + + if(dtopd(new_dentry)->state == DELETED) { + dtopd(old_dentry)->state = DEL_REWRITTEN; + dtohd(old_dentry) = NULL; + } + else if(dtopd(new_dentry)->state == NON_EXISTANT) { + dtopd(old_dentry)->state = CREATED; + /* steal new dentry's neg. base dentry */ + dtohd(old_dentry) = dtohd(new_dentry); + dtohd(new_dentry) = NULL; + } + } + if(dtopd(new_dentry)->state == UNMODIFIED || + dtopd(new_dentry)->state == NON_EXISTANT) { + err = get_neg_sto_dentry(new_dentry); + if(err) + goto out; + } + + /* now move sto file */ + hidden_old_dentry = dtohd2(old_dentry); + hidden_new_dentry = dtohd2(new_dentry); + + dget(hidden_old_dentry); + dget(hidden_new_dentry); + + hidden_old_dir_dentry = dget(hidden_old_dentry->d_parent); + hidden_new_dir_dentry = dget(hidden_new_dentry->d_parent); + double_lock(hidden_old_dir_dentry, hidden_new_dir_dentry); + + err = vfs_rename(hidden_old_dir_dentry->d_inode, hidden_old_dentry, + hidden_new_dir_dentry->d_inode, hidden_new_dentry); + if(err) + goto out_lock; + + fist_copy_attr_all(new_dir, hidden_new_dir_dentry->d_inode); + if (new_dir != old_dir) + fist_copy_attr_all(old_dir, + hidden_old_dir_dentry->d_inode); + + out_lock: + /* double_unlock will dput the new/old parent dentries + * whose refcnts were incremented via get_parent above. */ + double_unlock(hidden_old_dir_dentry, hidden_new_dir_dentry); + dput(hidden_new_dentry); + dput(hidden_old_dentry); + + out: + return err; +} + +int rename_nondir(inode_t *old_dir, dentry_t *old_dentry, + inode_t *new_dir, dentry_t *new_dentry) +{ + int err=0; + + check_mini_fo_dentry(old_dentry); + check_mini_fo_dentry(new_dentry); + check_mini_fo_inode(old_dir); + check_mini_fo_inode(new_dir); + + /* state: UNMODIFIED */ + if(dtost(old_dentry) == UNMODIFIED) { + err = nondir_unmod_to_mod(old_dentry, 1); + if(err) { + err = -EINVAL; + goto out; + } + } + + /* the easy states */ + if(exists_in_storage(old_dentry)) { + + dentry_t *hidden_old_dentry; + dentry_t *hidden_new_dentry; + dentry_t *hidden_old_dir_dentry; + dentry_t *hidden_new_dir_dentry; + + /* if old file is MODIFIED, add it to the deleted_list */ + if(dtopd(old_dentry)->state == MODIFIED) { + meta_add_d_entry(old_dentry->d_parent, + old_dentry->d_name.name, + old_dentry->d_name.len); + + dput(dtohd(old_dentry)); + } + /* if old file is CREATED, we only release the base dentry */ + if(dtopd(old_dentry)->state == CREATED) { + if(dtohd(old_dentry)) + dput(dtohd(old_dentry)); + } + + /* now setup the new states (depends on new_dentry state) */ + /* new dentry state = MODIFIED */ + if(dtopd(new_dentry)->state == MODIFIED) { + meta_add_d_entry(new_dentry->d_parent, + new_dentry->d_name.name, + new_dentry->d_name.len); + + /* new dentry will be d_put'ed later by the vfs + * so don't do it here + * dput(dtohd(new_dentry)); + */ + dtohd(old_dentry) = NULL; + dtopd(old_dentry)->state = DEL_REWRITTEN; + } + /* new dentry state = UNMODIFIED */ + else if(dtopd(new_dentry)->state == UNMODIFIED) { + if(get_neg_sto_dentry(new_dentry)) + return -EINVAL; + + meta_add_d_entry(new_dentry->d_parent, + new_dentry->d_name.name, + new_dentry->d_name.len); + + /* is this right??? */ + /*dput(dtohd(new_dentry));*/ + dtohd(old_dentry) = NULL; + dtopd(old_dentry)->state = DEL_REWRITTEN; + } + /* new dentry state = CREATED */ + else if(dtopd(new_dentry)->state == CREATED) { + /* we keep the neg. base dentry (if exists) */ + dtohd(old_dentry) = dtohd(new_dentry); + /* ...and set it to Null, or we'll get + * dcache.c:345 if it gets dput twice... */ + dtohd(new_dentry) = NULL; + dtopd(old_dentry)->state = CREATED; + } + /* new dentry state = NON_EXISTANT */ + else if(dtopd(new_dentry)->state == NON_EXISTANT) { + if(get_neg_sto_dentry(new_dentry)) + return -EINVAL; + + /* we keep the neg. base dentry (if exists) */ + dtohd(old_dentry) = dtohd(new_dentry); + /* ...and set it to Null, or we'll get + * Dr. dcache.c:345 if it gets dput twice... */ + dtohd(new_dentry) = NULL; + dtopd(old_dentry)->state = CREATED; + } + /* new dentry state = DEL_REWRITTEN or DELETED */ + else if(dtopd(new_dentry)->state == DEL_REWRITTEN || + dtopd(new_dentry)->state == DELETED) { + dtohd(old_dentry) = NULL; + dtopd(old_dentry)->state = DEL_REWRITTEN; + } + else { /* not possible, uhh, ahh */ + printk(KERN_CRIT + "mini_fo: rename_reg_file: invalid state detected [1].\n"); + return -1; + } + + /* now we definitely have a sto file */ + hidden_old_dentry = dtohd2(old_dentry); + hidden_new_dentry = dtohd2(new_dentry); + + dget(hidden_old_dentry); + dget(hidden_new_dentry); + + hidden_old_dir_dentry = dget(hidden_old_dentry->d_parent); + hidden_new_dir_dentry = dget(hidden_new_dentry->d_parent); + double_lock(hidden_old_dir_dentry, hidden_new_dir_dentry); + + err = vfs_rename(hidden_old_dir_dentry->d_inode, + hidden_old_dentry, + hidden_new_dir_dentry->d_inode, + hidden_new_dentry); + if(err) + goto out_lock; + + fist_copy_attr_all(new_dir, hidden_new_dir_dentry->d_inode); + if (new_dir != old_dir) + fist_copy_attr_all(old_dir, hidden_old_dir_dentry->d_inode); + + out_lock: + /* double_unlock will dput the new/old parent dentries + * whose refcnts were incremented via get_parent above. + */ + double_unlock(hidden_old_dir_dentry, hidden_new_dir_dentry); + dput(hidden_new_dentry); + dput(hidden_old_dentry); + out: + return err; + } + else { /* invalid state */ + printk(KERN_CRIT "mini_fo: rename_reg_file: ERROR: invalid state detected [2].\n"); + return -1; + } +} + + +STATIC int +mini_fo_readlink(dentry_t *dentry, char *buf, int bufsiz) +{ + int err=0; + dentry_t *hidden_dentry = NULL; + + if(dtohd2(dentry) && dtohd2(dentry)->d_inode) { + hidden_dentry = dtohd2(dentry); + } else if(dtohd(dentry) && dtohd(dentry)->d_inode) { + hidden_dentry = dtohd(dentry); + } else { + goto out; + } + + if (!hidden_dentry->d_inode->i_op || + !hidden_dentry->d_inode->i_op->readlink) { + err = -EINVAL; goto out; + } + + err = hidden_dentry->d_inode->i_op->readlink(hidden_dentry, + buf, + bufsiz); + if (err > 0) + fist_copy_attr_atime(dentry->d_inode, hidden_dentry->d_inode); + + out: + return err; +} + + +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,13) +static int mini_fo_follow_link(dentry_t *dentry, struct nameidata *nd) +#else +static void* mini_fo_follow_link(dentry_t *dentry, struct nameidata *nd) +#endif +{ + char *buf; + int len = PAGE_SIZE, err; + mm_segment_t old_fs; + + /* in 2.6 this is freed by mini_fo_put_link called by __do_follow_link */ + buf = kmalloc(len, GFP_KERNEL); + if (!buf) { + err = -ENOMEM; + goto out; + } + + /* read the symlink, and then we will follow it */ + old_fs = get_fs(); + set_fs(KERNEL_DS); + err = dentry->d_inode->i_op->readlink(dentry, buf, len); + set_fs(old_fs); + if (err < 0) { + kfree(buf); + buf = NULL; + goto out; + } + buf[err] = 0; +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) + nd_set_link(nd, buf); + err = 0; +#else + err = vfs_follow_link(nd, buf); +#endif + + out: +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0) + kfree(buf); +#endif + +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,13) + return err; +#else + return ERR_PTR(err); +#endif +} + +STATIC +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,13) +void mini_fo_put_link(struct dentry *dentry, struct nameidata *nd) +#else +void mini_fo_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) +#endif +{ + char *link; + link = nd_get_link(nd); + kfree(link); +} +#endif + +STATIC int +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) +mini_fo_permission(inode_t *inode, int mask, struct nameidata *nd) +#else +mini_fo_permission(inode_t *inode, int mask) +#endif +{ + inode_t *hidden_inode; + int mode; + int err; + + if(itohi2(inode)) { + hidden_inode = itohi2(inode); + } else { + hidden_inode = itohi(inode); + } + mode = inode->i_mode; + + /* not really needed, as permission handles everything: + * err = vfs_permission(inode, mask); + * if (err) + * goto out; + */ + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) + err = permission(hidden_inode, mask, nd); +#else + err = permission(hidden_inode, mask); +#endif + + /* out: */ + return err; +} + +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0) +STATIC int +mini_fo_inode_revalidate(dentry_t *dentry) +{ + int err = 0; + dentry_t *hidden_dentry; + inode_t *hidden_inode; + + ASSERT(dentry->d_inode); + ASSERT(itopd(dentry->d_inode)); + + if(itohi2(dentry->d_inode)) { + hidden_dentry = dtohd2(dentry); + hidden_inode = hidden_dentry->d_inode; + } else if(itohi(dentry->d_inode)) { + hidden_dentry = dtohd(dentry); + hidden_inode = hidden_dentry->d_inode; + } else { + printk(KERN_CRIT "mini_fo_inode_revalidate: ERROR, invalid state detected.\n"); + err = -ENOENT; + goto out; + } + if (hidden_inode && hidden_inode->i_op && hidden_inode->i_op->revalidate){ + err = hidden_inode->i_op->revalidate(hidden_dentry); + if (err) + goto out; + } + fist_copy_attr_all(dentry->d_inode, hidden_inode); + out: + return err; +} +#endif + +STATIC int +mini_fo_setattr(dentry_t *dentry, struct iattr *ia) +{ + int err = 0; + + check_mini_fo_dentry(dentry); + + if(!is_mini_fo_existant(dentry)) { + printk(KERN_CRIT "mini_fo_setattr: ERROR, invalid state detected [1].\n"); + goto out; + } + + if(dtost(dentry) == UNMODIFIED) { + if(!IS_COPY_FLAG(ia->ia_valid)) + goto out; /* we ignore these changes to base */ + + if(S_ISDIR(dentry->d_inode->i_mode)) { + err = dir_unmod_to_mod(dentry); + } else { + /* we copy contents if file is not beeing truncated */ + if(S_ISREG(dentry->d_inode->i_mode) && + !(ia->ia_size == 0 && (ia->ia_valid & ATTR_SIZE))) { + err = nondir_unmod_to_mod(dentry, 1); + } else + err = nondir_unmod_to_mod(dentry, 0); + } + if(err) { + err = -EINVAL; + printk(KERN_CRIT "mini_fo_setattr: ERROR changing states.\n"); + goto out; + } + } + if(!exists_in_storage(dentry)) { + printk(KERN_CRIT "mini_fo_setattr: ERROR, invalid state detected [2].\n"); + err = -EINVAL; + goto out; + } + ASSERT(dentry->d_inode); + ASSERT(dtohd2(dentry)); + ASSERT(itopd(dentry->d_inode)); + ASSERT(itohi2(dentry->d_inode)); + + err = notify_change(dtohd2(dentry), ia); + fist_copy_attr_all(dentry->d_inode, itohi2(dentry->d_inode)); + out: + return err; +} + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) +STATIC int +mini_fo_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) +{ + int err = 0; + dentry_t *hidden_dentry; + + ASSERT(dentry->d_inode); + ASSERT(itopd(dentry->d_inode)); + + if(itohi2(dentry->d_inode)) { + hidden_dentry = dtohd2(dentry); + } else if(itohi(dentry->d_inode)) { + hidden_dentry = dtohd(dentry); + } else { + printk(KERN_CRIT "mini_fo_getattr: ERROR, invalid state detected.\n"); + err = -ENOENT; + goto out; + } + fist_copy_attr_all(dentry->d_inode, hidden_dentry->d_inode); + + ASSERT(hidden_dentry); + ASSERT(hidden_dentry->d_inode); + ASSERT(hidden_dentry->d_inode->i_op); + + generic_fillattr(dentry->d_inode, stat); + if (!stat->blksize) { + struct super_block *s = hidden_dentry->d_inode->i_sb; + unsigned blocks; + blocks = (stat->size+s->s_blocksize-1) >> s->s_blocksize_bits; + stat->blocks = (s->s_blocksize / 512) * blocks; + stat->blksize = s->s_blocksize; + } + out: + return err; +} +#endif + +#if defined(XATTR) && (LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,20)) +#if 0 /* no xattr_alloc() and xattr_free() */ +/* This is lifted from fs/xattr.c */ +static void * +xattr_alloc(size_t size, size_t limit) +{ + void *ptr; + + if (size > limit) + return ERR_PTR(-E2BIG); + + if (!size) /* size request, no buffer is needed */ + return NULL; + else if (size <= PAGE_SIZE) + ptr = kmalloc((unsigned long) size, GFP_KERNEL); + else + ptr = vmalloc((unsigned long) size); + if (!ptr) + return ERR_PTR(-ENOMEM); + return ptr; +} + +static void +xattr_free(void *ptr, size_t size) +{ + if (!size) /* size request, no buffer was needed */ + return; + else if (size <= PAGE_SIZE) + kfree(ptr); + else + vfree(ptr); +} +#endif /* no xattr_alloc() and xattr_free() */ + +/* BKL held by caller. + * dentry->d_inode->i_sem down + */ +STATIC int +mini_fo_getxattr(struct dentry *dentry, const char *name, void *value, size_t size) { + struct dentry *hidden_dentry = NULL; + int err = -EOPNOTSUPP; + /* Define these anyway so we don't need as much ifdef'ed code. */ + char *encoded_name = NULL; + char *encoded_value = NULL; + + check_mini_fo_dentry(dentry); + + if(exists_in_storage(dentry)) + hidden_dentry = dtohd2(dentry); + else + hidden_dentry = dtohd(dentry); + + ASSERT(hidden_dentry); + ASSERT(hidden_dentry->d_inode); + ASSERT(hidden_dentry->d_inode->i_op); + + if (hidden_dentry->d_inode->i_op->getxattr) { + encoded_name = (char *)name; + encoded_value = (char *)value; + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&hidden_dentry->d_inode->i_mutex); +#else + down(&hidden_dentry->d_inode->i_sem); +#endif + /* lock_kernel() already done by caller. */ + err = hidden_dentry->d_inode->i_op->getxattr(hidden_dentry, encoded_name, encoded_value, size); + /* unlock_kernel() will be done by caller. */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&hidden_dentry->d_inode->i_mutex); +#else + up(&hidden_dentry->d_inode->i_sem); +#endif + } + return err; +} + +/* BKL held by caller. + * dentry->d_inode->i_sem down + */ +STATIC int +#if ((LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,21) \ + && LINUX_VERSION_CODE <= KERNEL_VERSION(2,4,23)) \ + || LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0)) +mini_fo_setxattr(struct dentry *dentry, const char *name, + const void *value, size_t size, int flags) +#else +mini_fo_setxattr(struct dentry *dentry, const char *name, + void *value, size_t size, int flags) +#endif + +{ + struct dentry *hidden_dentry = NULL; + int err = -EOPNOTSUPP; + + /* Define these anyway, so we don't have as much ifdef'ed code. */ + char *encoded_value = NULL; + char *encoded_name = NULL; + + check_mini_fo_dentry(dentry); + + if(exists_in_storage(dentry)) + hidden_dentry = dtohd2(dentry); + else + hidden_dentry = dtohd(dentry); + + ASSERT(hidden_dentry); + ASSERT(hidden_dentry->d_inode); + ASSERT(hidden_dentry->d_inode->i_op); + + if (hidden_dentry->d_inode->i_op->setxattr) { + encoded_name = (char *)name; + encoded_value = (char *)value; + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&hidden_dentry->d_inode->i_mutex); +#else + down(&hidden_dentry->d_inode->i_sem); +#endif + /* lock_kernel() already done by caller. */ + err = hidden_dentry->d_inode->i_op->setxattr(hidden_dentry, encoded_name, encoded_value, size, flags); + /* unlock_kernel() will be done by caller. */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&hidden_dentry->d_inode->i_mutex); +#else + up(&hidden_dentry->d_inode->i_sem); +#endif + } + return err; +} + +/* BKL held by caller. + * dentry->d_inode->i_sem down + */ +STATIC int +mini_fo_removexattr(struct dentry *dentry, const char *name) { + struct dentry *hidden_dentry = NULL; + int err = -EOPNOTSUPP; + char *encoded_name; + + check_mini_fo_dentry(dentry); + + if(exists_in_storage(dentry)) + hidden_dentry = dtohd2(dentry); + else + hidden_dentry = dtohd(dentry); + + ASSERT(hidden_dentry); + ASSERT(hidden_dentry->d_inode); + ASSERT(hidden_dentry->d_inode->i_op); + + if (hidden_dentry->d_inode->i_op->removexattr) { + encoded_name = (char *)name; + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&hidden_dentry->d_inode->i_mutex); +#else + down(&hidden_dentry->d_inode->i_sem); +#endif + /* lock_kernel() already done by caller. */ + err = hidden_dentry->d_inode->i_op->removexattr(hidden_dentry, encoded_name); + /* unlock_kernel() will be done by caller. */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&hidden_dentry->d_inode->i_mutex); +#else + up(&hidden_dentry->d_inode->i_sem); +#endif + } + return err; +} + +/* BKL held by caller. + * dentry->d_inode->i_sem down + */ +STATIC int +mini_fo_listxattr(struct dentry *dentry, char *list, size_t size) { + struct dentry *hidden_dentry = NULL; + int err = -EOPNOTSUPP; + char *encoded_list = NULL; + + check_mini_fo_dentry(dentry); + + if(exists_in_storage(dentry)) + hidden_dentry = dtohd2(dentry); + else + hidden_dentry = dtohd(dentry); + + ASSERT(hidden_dentry); + ASSERT(hidden_dentry->d_inode); + ASSERT(hidden_dentry->d_inode->i_op); + + if (hidden_dentry->d_inode->i_op->listxattr) { + encoded_list = list; + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&hidden_dentry->d_inode->i_mutex); +#else + down(&hidden_dentry->d_inode->i_sem); +#endif + /* lock_kernel() already done by caller. */ + err = hidden_dentry->d_inode->i_op->listxattr(hidden_dentry, encoded_list, size); + /* unlock_kernel() will be done by caller. */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&hidden_dentry->d_inode->i_mutex); +#else + up(&hidden_dentry->d_inode->i_sem); +#endif + } + return err; +} +# endif /* defined(XATTR) && (LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,20)) */ + +struct inode_operations mini_fo_symlink_iops = + { + readlink: mini_fo_readlink, + follow_link: mini_fo_follow_link, + /* mk: permission: mini_fo_permission, */ +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0) + revalidate: mini_fo_inode_revalidate, +#endif + setattr: mini_fo_setattr, +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) + getattr: mini_fo_getattr, + put_link: mini_fo_put_link, +#endif + +#if defined(XATTR) && (LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,20)) + setxattr: mini_fo_setxattr, + getxattr: mini_fo_getxattr, + listxattr: mini_fo_listxattr, + removexattr: mini_fo_removexattr +# endif /* defined(XATTR) && (LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,20)) */ + }; + +struct inode_operations mini_fo_dir_iops = + { + create: mini_fo_create, + lookup: mini_fo_lookup, + link: mini_fo_link, + unlink: mini_fo_unlink, + symlink: mini_fo_symlink, + mkdir: mini_fo_mkdir, + rmdir: mini_fo_rmdir, + mknod: mini_fo_mknod, + rename: mini_fo_rename, + /* no readlink/follow_link for non-symlinks */ + // off because we have setattr + // truncate: mini_fo_truncate, + /* mk:permission: mini_fo_permission, */ +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0) + revalidate: mini_fo_inode_revalidate, +#endif +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) + getattr: mini_fo_getattr, +#endif + setattr: mini_fo_setattr, +#if defined(XATTR) && (LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,20)) + setxattr: mini_fo_setxattr, + getxattr: mini_fo_getxattr, + listxattr: mini_fo_listxattr, + removexattr: mini_fo_removexattr +# endif /* XATTR && LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,20) */ + }; + +struct inode_operations mini_fo_main_iops = + { + /* permission: mini_fo_permission, */ +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0) + revalidate: mini_fo_inode_revalidate, +#endif + setattr: mini_fo_setattr, + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) + getattr: mini_fo_getattr, +#endif +#if defined(XATTR) && (LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,20)) + setxattr: mini_fo_setxattr, + getxattr: mini_fo_getxattr, + listxattr: mini_fo_listxattr, + removexattr: mini_fo_removexattr +# endif /* XATTR && LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,20) */ + }; Index: linux-2.6.25.1/fs/mini_fo/main.c =================================================================== --- /dev/null +++ linux-2.6.25.1/fs/mini_fo/main.c @@ -0,0 +1,423 @@ +/* + * Copyright (c) 1997-2003 Erez Zadok + * Copyright (c) 2001-2003 Stony Brook University + * + * For specific licensing information, see the COPYING file distributed with + * this package, or get one from ftp://ftp.filesystems.org/pub/fist/COPYING. + * + * This Copyright notice must be kept intact and distributed with all + * fistgen sources INCLUDING sources generated by fistgen. + */ +/* + * Copyright (C) 2004, 2005 Markus Klotzbuecher + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ + +/* + * $Id$ + */ + +#ifdef HAVE_CONFIG_H +# include +#endif + +#include "fist.h" +#include "mini_fo.h" +#include + +/* This definition must only appear after we include */ +#ifndef MODULE_LICENSE +# define MODULE_LICENSE(bison) +#endif /* not MODULE_LICENSE */ + +/* + * This is the mini_fo tri interpose function, which extends the + * functionality of the regular interpose by interposing a higher + * level inode on top of two lower level ones: the base filesystem + * inode and the storage filesystem inode. + * + * sb we pass is mini_fo's super_block + */ +int +mini_fo_tri_interpose(dentry_t *hidden_dentry, + dentry_t *hidden_sto_dentry, + dentry_t *dentry, super_block_t *sb, int flag) +{ + inode_t *hidden_inode = NULL; + inode_t *hidden_sto_inode = NULL; /* store corresponding storage inode */ + int err = 0; + inode_t *inode; + + /* Pointer to hidden_sto_inode if exists, else to hidden_inode. + * This is used to copy the attributes of the correct inode. */ + inode_t *master_inode; + + if(hidden_dentry) + hidden_inode = hidden_dentry->d_inode; + if(hidden_sto_dentry) + hidden_sto_inode = hidden_sto_dentry->d_inode; + + ASSERT(dentry->d_inode == NULL); + + /* mk: One of the inodes associated with the dentrys is likely to + * be NULL, so carefull: + */ + ASSERT((hidden_inode != NULL) || (hidden_sto_inode != NULL)); + + if(hidden_sto_inode) + master_inode = hidden_sto_inode; + else + master_inode = hidden_inode; + + /* + * We allocate our new inode below, by calling iget. + * iget will call our read_inode which will initialize some + * of the new inode's fields + */ + + /* + * original: inode = iget(sb, hidden_inode->i_ino); + */ + inode = iget(sb, iunique(sb, 25)); + if (!inode) { + err = -EACCES; /* should be impossible??? */ + goto out; + } + + /* + * interpose the inode if not already interposed + * this is possible if the inode is being reused + * XXX: what happens if we get_empty_inode() but there's another already? + * for now, ASSERT() that this can't happen; fix later. + */ + if (itohi(inode) != NULL) { + printk(KERN_CRIT "mini_fo_tri_interpose: itohi(inode) != NULL.\n"); + } + if (itohi2(inode) != NULL) { + printk(KERN_CRIT "mini_fo_tri_interpose: itohi2(inode) != NULL.\n"); + } + + /* mk: Carefull, igrab can't handle NULL inodes (ok, why should it?), so + * we need to check here: + */ + if(hidden_inode) + itohi(inode) = igrab(hidden_inode); + else + itohi(inode) = NULL; + + if(hidden_sto_inode) + itohi2(inode) = igrab(hidden_sto_inode); + else + itohi2(inode) = NULL; + + + /* Use different set of inode ops for symlinks & directories*/ + if (S_ISLNK(master_inode->i_mode)) + inode->i_op = &mini_fo_symlink_iops; + else if (S_ISDIR(master_inode->i_mode)) + inode->i_op = &mini_fo_dir_iops; + + /* Use different set of file ops for directories */ + if (S_ISDIR(master_inode->i_mode)) + inode->i_fop = &mini_fo_dir_fops; + + /* properly initialize special inodes */ + if (S_ISBLK(master_inode->i_mode) || S_ISCHR(master_inode->i_mode) || + S_ISFIFO(master_inode->i_mode) || S_ISSOCK(master_inode->i_mode)) { + init_special_inode(inode, master_inode->i_mode, master_inode->i_rdev); + } + + /* Fix our inode's address operations to that of the lower inode */ + if (inode->i_mapping->a_ops != master_inode->i_mapping->a_ops) { + inode->i_mapping->a_ops = master_inode->i_mapping->a_ops; + } + + /* only (our) lookup wants to do a d_add */ + if (flag) + d_add(dentry, inode); + else + d_instantiate(dentry, inode); + + ASSERT(dtopd(dentry) != NULL); + + /* all well, copy inode attributes */ + fist_copy_attr_all(inode, master_inode); + + out: + return err; +} + +/* parse mount options "base=" and "sto=" */ +dentry_t * +mini_fo_parse_options(super_block_t *sb, char *options) +{ + dentry_t *hidden_root = ERR_PTR(-EINVAL); + dentry_t *hidden_root2 = ERR_PTR(-EINVAL); + struct nameidata nd, nd2; + char *name, *tmp, *end; + int err = 0; + + /* We don't want to go off the end of our arguments later on. */ + for (end = options; *end; end++); + + while (options < end) { + tmp = options; + while (*tmp && *tmp != ',') + tmp++; + *tmp = '\0'; + if (!strncmp("base=", options, 5)) { + name = options + 5; + printk(KERN_INFO "mini_fo: using base directory: %s\n", name); + +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0) + if (path_init(name, LOOKUP_FOLLOW, &nd)) + err = path_walk(name, &nd); +#else + err = path_lookup(name, LOOKUP_FOLLOW, &nd); +#endif + if (err) { + printk(KERN_CRIT "mini_fo: error accessing hidden directory '%s'\n", name); + hidden_root = ERR_PTR(err); + goto out; + } + hidden_root = nd.dentry; + stopd(sb)->base_dir_dentry = nd.dentry; + stopd(sb)->hidden_mnt = nd.mnt; + + } else if(!strncmp("sto=", options, 4)) { + /* parse the storage dir */ + name = options + 4; + printk(KERN_INFO "mini_fo: using storage directory: %s\n", name); +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0) + if(path_init(name, LOOKUP_FOLLOW, &nd2)) + err = path_walk(name, &nd2); +#else + err = path_lookup(name, LOOKUP_FOLLOW, &nd2); +#endif + if(err) { + printk(KERN_CRIT "mini_fo: error accessing hidden storage directory '%s'\n", name); + + hidden_root2 = ERR_PTR(err); + goto out; + } + hidden_root2 = nd2.dentry; + stopd(sb)->storage_dir_dentry = nd2.dentry; + stopd(sb)->hidden_mnt2 = nd2.mnt; + stohs2(sb) = hidden_root2->d_sb; + + /* validate storage dir, this is done in + * mini_fo_read_super for the base directory. + */ + if (IS_ERR(hidden_root2)) { + printk(KERN_WARNING "mini_fo_parse_options: storage dentry lookup failed (err = %ld)\n", PTR_ERR(hidden_root2)); + goto out; + } + if (!hidden_root2->d_inode) { + printk(KERN_WARNING "mini_fo_parse_options: no storage dir to interpose on.\n"); + goto out; + } + stohs2(sb) = hidden_root2->d_sb; + } else { + printk(KERN_WARNING "mini_fo: unrecognized option '%s'\n", options); + hidden_root = ERR_PTR(-EINVAL); + goto out; + } + options = tmp + 1; + } + + out: + if(IS_ERR(hidden_root2)) + return hidden_root2; + return hidden_root; +} + + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) +static int +#else +super_block_t * +#endif +mini_fo_read_super(super_block_t *sb, void *raw_data, int silent) +{ + dentry_t *hidden_root; + int err = 0; + + if (!raw_data) { + printk(KERN_WARNING "mini_fo_read_super: missing argument\n"); + err = -EINVAL; + goto out; + } + /* + * Allocate superblock private data + */ + __stopd(sb) = kmalloc(sizeof(struct mini_fo_sb_info), GFP_KERNEL); + if (!stopd(sb)) { + printk(KERN_WARNING "%s: out of memory\n", __FUNCTION__); + err = -ENOMEM; + goto out; + } + stohs(sb) = NULL; + + hidden_root = mini_fo_parse_options(sb, raw_data); + if (IS_ERR(hidden_root)) { + printk(KERN_WARNING "mini_fo_read_super: lookup_dentry failed (err = %ld)\n", PTR_ERR(hidden_root)); + err = PTR_ERR(hidden_root); + goto out_free; + } + if (!hidden_root->d_inode) { + printk(KERN_WARNING "mini_fo_read_super: no directory to interpose on\n"); + goto out_free; + } + stohs(sb) = hidden_root->d_sb; + + /* + * Linux 2.4.2-ac3 and beyond has code in + * mm/filemap.c:generic_file_write() that requires sb->s_maxbytes + * to be populated. If not set, all write()s under that sb will + * return 0. + * + * Linux 2.4.4+ automatically sets s_maxbytes to MAX_NON_LFS; + * the filesystem should override it only if it supports LFS. + */ + /* non-SCA code is good to go with LFS */ + sb->s_maxbytes = hidden_root->d_sb->s_maxbytes; + + sb->s_op = &mini_fo_sops; + /* + * we can't use d_alloc_root if we want to use + * our own interpose function unchanged, + * so we simply replicate *most* of the code in d_alloc_root here + */ +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0) + sb->s_root = d_alloc(NULL, &(const struct qstr) { "/", 1, 0 }); +#else + sb->s_root = d_alloc(NULL, &(const struct qstr){hash: 0, name: "/", len : 1}); +#endif + if (IS_ERR(sb->s_root)) { + printk(KERN_WARNING "mini_fo_read_super: d_alloc failed\n"); + err = -ENOMEM; + goto out_dput; + } + + sb->s_root->d_op = &mini_fo_dops; + sb->s_root->d_sb = sb; + sb->s_root->d_parent = sb->s_root; + + /* link the upper and lower dentries */ + __dtopd(sb->s_root) = (struct mini_fo_dentry_info *) + kmalloc(sizeof(struct mini_fo_dentry_info), GFP_KERNEL); + if (!dtopd(sb->s_root)) { + err = -ENOMEM; + goto out_dput2; + } + dtopd(sb->s_root)->state = MODIFIED; + dtohd(sb->s_root) = hidden_root; + + /* fanout relevant, interpose on storage root dentry too */ + dtohd2(sb->s_root) = stopd(sb)->storage_dir_dentry; + + /* ...and call tri-interpose to interpose root dir inodes + * if (mini_fo_interpose(hidden_root, sb->s_root, sb, 0)) + */ + if(mini_fo_tri_interpose(hidden_root, dtohd2(sb->s_root), sb->s_root, sb, 0)) + goto out_dput2; + + /* initalize the wol list */ + itopd(sb->s_root->d_inode)->deleted_list_size = -1; + itopd(sb->s_root->d_inode)->renamed_list_size = -1; + meta_build_lists(sb->s_root); + + goto out; + + out_dput2: + dput(sb->s_root); + out_dput: + dput(hidden_root); + dput(dtohd2(sb->s_root)); /* release the hidden_sto_dentry too */ + out_free: + kfree(stopd(sb)); + __stopd(sb) = NULL; + out: + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) + return err; +#else + if (err) { + return ERR_PTR(err); + } else { + return sb; + } +#endif +} + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,18) +static int mini_fo_get_sb(struct file_system_type *fs_type, + int flags, const char *dev_name, + void *raw_data, struct vfsmount *mnt) +{ + return get_sb_nodev(fs_type, flags, raw_data, mini_fo_read_super, mnt); +} +#else +static struct super_block *mini_fo_get_sb(struct file_system_type *fs_type, + int flags, const char *dev_name, + void *raw_data) +{ + return get_sb_nodev(fs_type, flags, raw_data, mini_fo_read_super); +} +#endif + +void mini_fo_kill_block_super(struct super_block *sb) +{ + generic_shutdown_super(sb); + /* + * XXX: BUG: Halcrow: Things get unstable sometime after this point: + * lib/rwsem-spinlock.c:127: spin_is_locked on uninitialized + * fs/fs-writeback.c:402: spin_lock(fs/super.c:a0381828) already + * locked by fs/fs-writeback.c/402 + * + * Apparently, someone's not releasing a lock on sb_lock... + */ +} + +static struct file_system_type mini_fo_fs_type = { + .owner = THIS_MODULE, + .name = "mini_fo", + .get_sb = mini_fo_get_sb, + .kill_sb = mini_fo_kill_block_super, + .fs_flags = 0, +}; + + +#else +static DECLARE_FSTYPE(mini_fo_fs_type, "mini_fo", mini_fo_read_super, 0); +#endif + +static int __init init_mini_fo_fs(void) +{ + printk("Registering mini_fo version $Id$\n"); + return register_filesystem(&mini_fo_fs_type); +} +static void __exit exit_mini_fo_fs(void) +{ + printk("Unregistering mini_fo version $Id$\n"); + unregister_filesystem(&mini_fo_fs_type); +} + +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0) +EXPORT_NO_SYMBOLS; +#endif + +MODULE_AUTHOR("Erez Zadok "); +MODULE_DESCRIPTION("FiST-generated mini_fo filesystem"); +MODULE_LICENSE("GPL"); + +/* MODULE_PARM(fist_debug_var, "i"); */ +/* MODULE_PARM_DESC(fist_debug_var, "Debug level"); */ + +module_init(init_mini_fo_fs) +module_exit(exit_mini_fo_fs) Index: linux-2.6.25.1/fs/mini_fo/Makefile =================================================================== --- /dev/null +++ linux-2.6.25.1/fs/mini_fo/Makefile @@ -0,0 +1,17 @@ +# +# Makefile for mini_fo 2.4 and 2.6 Linux kernels +# +# Copyright (C) 2004, 2005 Markus Klotzbuecher +# +# This program is free software; you can redistribute it and/or +# modify it under the terms of the GNU General Public License +# as published by the Free Software Foundation; either version +# 2 of the License, or (at your option) any later version. +# + +obj-$(CONFIG_MINI_FO) := mini_fo.o +mini_fo-objs := meta.o dentry.o file.o inode.o main.o super.o state.o aux.o + +# dependencies +${mini_fo-objs}: mini_fo.h fist.h + Index: linux-2.6.25.1/fs/mini_fo/meta.c =================================================================== --- /dev/null +++ linux-2.6.25.1/fs/mini_fo/meta.c @@ -0,0 +1,1000 @@ +/* + * Copyright (C) 2004, 2005 Markus Klotzbuecher + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ + +#ifdef HAVE_CONFIG_H +# include +#endif /* HAVE_CONFIG_H */ +#include "fist.h" +#include "mini_fo.h" + +int meta_build_lists(dentry_t *dentry) +{ + struct mini_fo_inode_info *inode_info; + + dentry_t *meta_dentry = 0; + file_t *meta_file = 0; + mm_segment_t old_fs; + void *buf; + + int bytes, len; + struct vfsmount *meta_mnt; + char *entry; + + inode_info = itopd(dentry->d_inode); + if(!(inode_info->deleted_list_size == -1 && + inode_info->renamed_list_size == -1)) { + printk(KERN_CRIT "mini_fo: meta_build_lists: \ + Error, list(s) not virgin.\n"); + return -1; + } + + /* init our meta lists */ + INIT_LIST_HEAD(&inode_info->deleted_list); + inode_info->deleted_list_size = 0; + + INIT_LIST_HEAD(&inode_info->renamed_list); + inode_info->renamed_list_size = 0; + + /* might there be a META-file? */ + if(dtohd2(dentry) && dtohd2(dentry)->d_inode) { + meta_dentry = lookup_one_len(META_FILENAME, + dtohd2(dentry), + strlen(META_FILENAME)); + if(!meta_dentry->d_inode) { + dput(meta_dentry); + goto out_ok; + } + /* $%& err, is this correct? */ + meta_mnt = stopd(dentry->d_inode->i_sb)->hidden_mnt2; + mntget(meta_mnt); + + + /* open META-file for reading */ + meta_file = dentry_open(meta_dentry, meta_mnt, 0x0); + if(!meta_file || IS_ERR(meta_file)) { + printk(KERN_CRIT "mini_fo: meta_build_lists: \ + ERROR opening META file.\n"); + goto out_err; + } + + /* check if fs supports reading */ + if(!meta_file->f_op->read) { + printk(KERN_CRIT "mini_fo: meta_build_lists: \ + ERROR, fs does not support reading.\n"); + goto out_err_close; + } + + /* allocate a page for transfering the data */ + buf = (void *) __get_free_page(GFP_KERNEL); + if(!buf) { + printk(KERN_CRIT "mini_fo: meta_build_lists: \ + ERROR, out of mem.\n"); + goto out_err_close; + } + meta_file->f_pos = 0; + old_fs = get_fs(); + set_fs(KERNEL_DS); + do { + char *c; + bytes = meta_file->f_op->read(meta_file, buf, PAGE_SIZE, &meta_file->f_pos); + if(bytes == PAGE_SIZE) { + /* trim a cut off filename and adjust f_pos to get it next time */ + for(c = (char*) buf+PAGE_SIZE; + *c != '\n'; + c--, bytes--, meta_file->f_pos--); + } + entry = (char *) buf; + while(entry < (char *) buf+bytes) { + + char *old_path; + char *dir_name; + int old_len, new_len; + + /* len without '\n'*/ + len = (int) (strchr(entry, '\n') - entry); + switch (*entry) { + case 'D': + /* format: "D filename" */ + meta_list_add_d_entry(dentry, + entry+2, + len-2); + break; + case 'R': + /* format: "R path/xy/dir newDir" */ + old_path = entry+2; + dir_name = strchr(old_path, ' ') + 1; + old_len = dir_name - old_path - 1; + new_len = ((int) entry) + len - ((int ) dir_name); + meta_list_add_r_entry(dentry, + old_path, + old_len, + dir_name, + new_len); + break; + default: + /* unknown entry type detected */ + break; + } + entry += len+1; + } + + } while(meta_file->f_pos < meta_dentry->d_inode->i_size); + + free_page((unsigned long) buf); + set_fs(old_fs); + fput(meta_file); + } + goto out_ok; + + out_err_close: + fput(meta_file); + out_err: + mntput(meta_mnt); + dput(meta_dentry); + return -1; + out_ok: + return 1; /* check this!!! inode_info->wol_size; */ +} + +/* cleanups up all lists and free's the mem by dentry */ +int meta_put_lists(dentry_t *dentry) +{ + if(!dentry || !dentry->d_inode) { + printk("mini_fo: meta_put_lists: invalid dentry passed.\n"); + return -1; + } + return __meta_put_lists(dentry->d_inode); +} + +/* cleanups up all lists and free's the mem by inode */ +int __meta_put_lists(inode_t *inode) +{ + int err = 0; + if(!inode || !itopd(inode)) { + printk("mini_fo: __meta_put_lists: invalid inode passed.\n"); + return -1; + } + err = __meta_put_d_list(inode); + err |= __meta_put_r_list(inode); + return err; +} + +int meta_sync_lists(dentry_t *dentry) +{ + int err = 0; + if(!dentry || !dentry->d_inode) { + printk("mini_fo: meta_sync_lists: \ + invalid dentry passed.\n"); + return -1; + } + err = meta_sync_d_list(dentry, 0); + err |= meta_sync_r_list(dentry, 1); + return err; +} + + +/* remove all D entries from the renamed list and free the mem */ +int __meta_put_d_list(inode_t *inode) +{ + struct list_head *tmp; + struct deleted_entry *del_entry; + struct mini_fo_inode_info *inode_info; + + if(!inode || !itopd(inode)) { + printk(KERN_CRIT "mini_fo: __meta_put_d_list: \ + invalid inode passed.\n"); + return -1; + } + inode_info = itopd(inode); + + /* nuke the DELETED-list */ + if(inode_info->deleted_list_size <= 0) + return 0; + + while(!list_empty(&inode_info->deleted_list)) { + tmp = inode_info->deleted_list.next; + list_del(tmp); + del_entry = list_entry(tmp, struct deleted_entry, list); + kfree(del_entry->name); + kfree(del_entry); + } + inode_info->deleted_list_size = 0; + + return 0; +} + +/* remove all R entries from the renamed list and free the mem */ +int __meta_put_r_list(inode_t *inode) +{ + struct list_head *tmp; + struct renamed_entry *ren_entry; + struct mini_fo_inode_info *inode_info; + + if(!inode || !itopd(inode)) { + printk(KERN_CRIT "mini_fo: meta_put_r_list: invalid inode.\n"); + return -1; + } + inode_info = itopd(inode); + + /* nuke the RENAMED-list */ + if(inode_info->renamed_list_size <= 0) + return 0; + + while(!list_empty(&inode_info->renamed_list)) { + tmp = inode_info->renamed_list.next; + list_del(tmp); + ren_entry = list_entry(tmp, struct renamed_entry, list); + kfree(ren_entry->new_name); + kfree(ren_entry->old_name); + kfree(ren_entry); + } + inode_info->renamed_list_size = 0; + + return 0; +} + +int meta_add_d_entry(dentry_t *dentry, const char *name, int len) +{ + int err = 0; + err = meta_list_add_d_entry(dentry, name, len); + err |= meta_write_d_entry(dentry,name,len); + return err; +} + +/* add a D entry to the deleted list */ +int meta_list_add_d_entry(dentry_t *dentry, const char *name, int len) +{ + struct deleted_entry *del_entry; + struct mini_fo_inode_info *inode_info; + + if(!dentry || !dentry->d_inode) { + printk(KERN_CRIT "mini_fo: meta_list_add_d_entry: \ + invalid dentry passed.\n"); + return -1; + } + inode_info = itopd(dentry->d_inode); + + if(inode_info->deleted_list_size < 0) + return -1; + + del_entry = (struct deleted_entry *) + kmalloc(sizeof(struct deleted_entry), GFP_KERNEL); + del_entry->name = (char*) kmalloc(len, GFP_KERNEL); + if(!del_entry || !del_entry->name) { + printk(KERN_CRIT "mini_fo: meta_list_add_d_entry: \ + out of mem.\n"); + kfree(del_entry->name); + kfree(del_entry); + return -ENOMEM; + } + + strncpy(del_entry->name, name, len); + del_entry->len = len; + + list_add(&del_entry->list, &inode_info->deleted_list); + inode_info->deleted_list_size++; + return 0; +} + +int meta_add_r_entry(dentry_t *dentry, + const char *old_name, int old_len, + const char *new_name, int new_len) +{ + int err = 0; + err = meta_list_add_r_entry(dentry, + old_name, old_len, + new_name, new_len); + err |= meta_write_r_entry(dentry, + old_name, old_len, + new_name, new_len); + return err; +} + +/* add a R entry to the renamed list */ +int meta_list_add_r_entry(dentry_t *dentry, + const char *old_name, int old_len, + const char *new_name, int new_len) +{ + struct renamed_entry *ren_entry; + struct mini_fo_inode_info *inode_info; + + if(!dentry || !dentry->d_inode) { + printk(KERN_CRIT "mini_fo: meta_list_add_r_entry: \ + invalid dentry passed.\n"); + return -1; + } + inode_info = itopd(dentry->d_inode); + + if(inode_info->renamed_list_size < 0) + return -1; + + ren_entry = (struct renamed_entry *) + kmalloc(sizeof(struct renamed_entry), GFP_KERNEL); + ren_entry->old_name = (char*) kmalloc(old_len, GFP_KERNEL); + ren_entry->new_name = (char*) kmalloc(new_len, GFP_KERNEL); + + if(!ren_entry || !ren_entry->old_name || !ren_entry->new_name) { + printk(KERN_CRIT "mini_fo: meta_list_add_r_entry: \ + out of mem.\n"); + kfree(ren_entry->new_name); + kfree(ren_entry->old_name); + kfree(ren_entry); + return -ENOMEM; + } + + strncpy(ren_entry->old_name, old_name, old_len); + ren_entry->old_len = old_len; + strncpy(ren_entry->new_name, new_name, new_len); + ren_entry->new_len = new_len; + + list_add(&ren_entry->list, &inode_info->renamed_list); + inode_info->renamed_list_size++; + return 0; +} + + +int meta_remove_r_entry(dentry_t *dentry, const char *name, int len) +{ + int err = 0; + if(!dentry || !dentry->d_inode) { + printk(KERN_CRIT + "mini_fo: meta_remove_r_entry: \ + invalid dentry passed.\n"); + return -1; + } + + err = meta_list_remove_r_entry(dentry, name, len); + err |= meta_sync_lists(dentry); + return err; +} + +int meta_list_remove_r_entry(dentry_t *dentry, const char *name, int len) +{ + if(!dentry || !dentry->d_inode) { + printk(KERN_CRIT + "mini_fo: meta_list_remove_r_entry: \ + invalid dentry passed.\n"); + return -1; + } + return __meta_list_remove_r_entry(dentry->d_inode, name, len); +} + +int __meta_list_remove_r_entry(inode_t *inode, const char *name, int len) +{ + struct list_head *tmp; + struct renamed_entry *ren_entry; + struct mini_fo_inode_info *inode_info; + + if(!inode || !itopd(inode)) + printk(KERN_CRIT + "mini_fo: __meta_list_remove_r_entry: \ + invalid inode passed.\n"); + inode_info = itopd(inode); + + if(inode_info->renamed_list_size < 0) + return -1; + if(inode_info->renamed_list_size == 0) + return 1; + + list_for_each(tmp, &inode_info->renamed_list) { + ren_entry = list_entry(tmp, struct renamed_entry, list); + if(ren_entry->new_len != len) + continue; + + if(!strncmp(ren_entry->new_name, name, len)) { + list_del(tmp); + kfree(ren_entry->new_name); + kfree(ren_entry->old_name); + kfree(ren_entry); + inode_info->renamed_list_size--; + return 0; + } + } + return 1; +} + + +/* append a single D entry to the meta file */ +int meta_write_d_entry(dentry_t *dentry, const char *name, int len) +{ + dentry_t *meta_dentry = 0; + file_t *meta_file = 0; + mm_segment_t old_fs; + + int bytes, err; + struct vfsmount *meta_mnt = 0; + char *buf; + + err = 0; + + if(itopd(dentry->d_inode)->deleted_list_size < 0) { + err = -1; + goto out; + } + + if(dtopd(dentry)->state == UNMODIFIED) { + err = build_sto_structure(dentry->d_parent, dentry); + if(err) { + printk(KERN_CRIT "mini_fo: meta_write_d_entry: \ + build_sto_structure failed.\n"); + goto out; + } + } + meta_dentry = lookup_one_len(META_FILENAME, + dtohd2(dentry), strlen (META_FILENAME)); + + /* We need to create a META-file */ + if(!meta_dentry->d_inode) { +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) + vfs_create(dtohd2(dentry)->d_inode, + meta_dentry, + S_IRUSR | S_IWUSR, + NULL); +#else + vfs_create(dtohd2(dentry)->d_inode, + meta_dentry, + S_IRUSR | S_IWUSR); +#endif + } + /* open META-file for writing */ + meta_file = dentry_open(meta_dentry, meta_mnt, 0x1); + if(!meta_file || IS_ERR(meta_file)) { + printk(KERN_CRIT "mini_fo: meta_write_d_entry: \ + ERROR opening meta file.\n"); + mntput(meta_mnt); /* $%& is this necessary? */ + dput(meta_dentry); + err = -1; + goto out; + } + + /* check if fs supports writing */ + if(!meta_file->f_op->write) { + printk(KERN_CRIT "mini_fo: meta_write_d_entry: \ + ERROR, fs does not support writing.\n"); + goto out_err_close; + } + + meta_file->f_pos = meta_dentry->d_inode->i_size; /* append */ + old_fs = get_fs(); + set_fs(KERNEL_DS); + + /* size: len for name, 1 for \n and 2 for "D " */ + buf = (char *) kmalloc(len+3, GFP_KERNEL); + if (!buf) { + printk(KERN_CRIT "mini_fo: meta_write_d_entry: \ + out of mem.\n"); + return -ENOMEM; + } + + buf[0] = 'D'; + buf[1] = ' '; + strncpy(buf+2, name, len); + buf[len+2] = '\n'; + bytes = meta_file->f_op->write(meta_file, buf, len+3, + &meta_file->f_pos); + if(bytes != len+3) { + printk(KERN_CRIT "mini_fo: meta_write_d_entry: \ + ERROR writing.\n"); + err = -1; + } + kfree(buf); + set_fs(old_fs); + + out_err_close: + fput(meta_file); + out: + return err; +} + +/* append a single R entry to the meta file */ +int meta_write_r_entry(dentry_t *dentry, + const char *old_name, int old_len, + const char *new_name, int new_len) +{ + dentry_t *meta_dentry = 0; + file_t *meta_file = 0; + mm_segment_t old_fs; + + int bytes, err, buf_len; + struct vfsmount *meta_mnt = 0; + char *buf; + + + err = 0; + + if(itopd(dentry->d_inode)->renamed_list_size < 0) { + err = -1; + goto out; + } + + /* build the storage structure? */ + if(dtopd(dentry)->state == UNMODIFIED) { + err = build_sto_structure(dentry->d_parent, dentry); + if(err) { + printk(KERN_CRIT "mini_fo: meta_write_r_entry: \ + build_sto_structure failed.\n"); + goto out; + } + } + meta_dentry = lookup_one_len(META_FILENAME, + dtohd2(dentry), + strlen (META_FILENAME)); + if(!meta_dentry->d_inode) { + /* We need to create a META-file */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) + vfs_create(dtohd2(dentry)->d_inode, + meta_dentry, S_IRUSR | S_IWUSR, NULL); +#else + vfs_create(dtohd2(dentry)->d_inode, + meta_dentry, S_IRUSR | S_IWUSR); +#endif + } + /* open META-file for writing */ + meta_file = dentry_open(meta_dentry, meta_mnt, 0x1); + if(!meta_file || IS_ERR(meta_file)) { + printk(KERN_CRIT "mini_fo: meta_write_r_entry: \ + ERROR opening meta file.\n"); + mntput(meta_mnt); + dput(meta_dentry); + err = -1; + goto out; + } + + /* check if fs supports writing */ + if(!meta_file->f_op->write) { + printk(KERN_CRIT "mini_fo: meta_write_r_entry: \ + ERROR, fs does not support writing.\n"); + goto out_err_close; + } + + meta_file->f_pos = meta_dentry->d_inode->i_size; /* append */ + old_fs = get_fs(); + set_fs(KERNEL_DS); + + /* size: 2 for "R ", old_len+new_len for names, 1 blank+1 \n */ + buf_len = old_len + new_len + 4; + buf = (char *) kmalloc(buf_len, GFP_KERNEL); + if (!buf) { + printk(KERN_CRIT "mini_fo: meta_write_r_entry: out of mem.\n"); + return -ENOMEM; + } + + buf[0] = 'R'; + buf[1] = ' '; + strncpy(buf + 2, old_name, old_len); + buf[old_len + 2] = ' '; + strncpy(buf + old_len + 3, new_name, new_len); + buf[buf_len -1] = '\n'; + bytes = meta_file->f_op->write(meta_file, buf, buf_len, &meta_file->f_pos); + if(bytes != buf_len) { + printk(KERN_CRIT "mini_fo: meta_write_r_entry: ERROR writing.\n"); + err = -1; + } + + kfree(buf); + set_fs(old_fs); + + out_err_close: + fput(meta_file); + out: + return err; +} + +/* sync D list to disk, append data if app_flag is 1 */ +/* check the meta_mnt, which seems not to be used (properly) */ + +int meta_sync_d_list(dentry_t *dentry, int app_flag) +{ + dentry_t *meta_dentry; + file_t *meta_file; + mm_segment_t old_fs; + + int bytes, err; + struct vfsmount *meta_mnt; + char *buf; + + struct list_head *tmp; + struct deleted_entry *del_entry; + struct mini_fo_inode_info *inode_info; + + err = 0; + meta_file=0; + meta_mnt=0; + + if(!dentry || !dentry->d_inode) { + printk(KERN_CRIT "mini_fo: meta_sync_d_list: \ + invalid inode passed.\n"); + err = -1; + goto out; + } + inode_info = itopd(dentry->d_inode); + + if(inode_info->deleted_list_size < 0) { + err = -1; + goto out; + } + + /* ok, there is something to sync */ + + /* build the storage structure? */ + if(!dtohd2(dentry) && !itohi2(dentry->d_inode)) { + err = build_sto_structure(dentry->d_parent, dentry); + if(err) { + printk(KERN_CRIT "mini_fo: meta_sync_d_list: \ + build_sto_structure failed.\n"); + goto out; + } + } + meta_dentry = lookup_one_len(META_FILENAME, + dtohd2(dentry), + strlen(META_FILENAME)); + if(!meta_dentry->d_inode) { + /* We need to create a META-file */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) + vfs_create(dtohd2(dentry)->d_inode, + meta_dentry, S_IRUSR | S_IWUSR, NULL); +#else + vfs_create(dtohd2(dentry)->d_inode, + meta_dentry, S_IRUSR | S_IWUSR); +#endif + app_flag = 0; + } + /* need we truncate the meta file? */ + if(!app_flag) { + struct iattr newattrs; + newattrs.ia_size = 0; + newattrs.ia_valid = ATTR_SIZE | ATTR_CTIME; + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&meta_dentry->d_inode->i_mutex); +#else + down(&meta_dentry->d_inode->i_sem); +#endif + err = notify_change(meta_dentry, &newattrs); + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&meta_dentry->d_inode->i_mutex); +#else + up(&meta_dentry->d_inode->i_sem); +#endif + + if(err || meta_dentry->d_inode->i_size != 0) { + printk(KERN_CRIT "mini_fo: meta_sync_d_list: \ + ERROR truncating meta file.\n"); + goto out_err_close; + } + } + + /* open META-file for writing */ + meta_file = dentry_open(meta_dentry, meta_mnt, 0x1); + if(!meta_file || IS_ERR(meta_file)) { + printk(KERN_CRIT "mini_fo: meta_sync_d_list: \ + ERROR opening meta file.\n"); + /* we don't mntget so we dont't mntput (for now) + * mntput(meta_mnt); + */ + dput(meta_dentry); + err = -1; + goto out; + } + + /* check if fs supports writing */ + if(!meta_file->f_op->write) { + printk(KERN_CRIT "mini_fo: meta_sync_d_list: \ + ERROR, fs does not support writing.\n"); + goto out_err_close; + } + + meta_file->f_pos = meta_dentry->d_inode->i_size; /* append */ + old_fs = get_fs(); + set_fs(KERNEL_DS); + + /* here we go... */ + list_for_each(tmp, &inode_info->deleted_list) { + del_entry = list_entry(tmp, struct deleted_entry, list); + + /* size: len for name, 1 for \n and 2 for "D " */ + buf = (char *) kmalloc(del_entry->len+3, GFP_KERNEL); + if (!buf) { + printk(KERN_CRIT "mini_fo: meta_sync_d_list: \ + out of mem.\n"); + return -ENOMEM; + } + + buf[0] = 'D'; + buf[1] = ' '; + strncpy(buf+2, del_entry->name, del_entry->len); + buf[del_entry->len+2] = '\n'; + bytes = meta_file->f_op->write(meta_file, buf, + del_entry->len+3, + &meta_file->f_pos); + if(bytes != del_entry->len+3) { + printk(KERN_CRIT "mini_fo: meta_sync_d_list: \ + ERROR writing.\n"); + err |= -1; + } + kfree(buf); + } + set_fs(old_fs); + + out_err_close: + fput(meta_file); + out: + return err; + +} + +int meta_sync_r_list(dentry_t *dentry, int app_flag) +{ + dentry_t *meta_dentry; + file_t *meta_file; + mm_segment_t old_fs; + + int bytes, err, buf_len; + struct vfsmount *meta_mnt; + char *buf; + + struct list_head *tmp; + struct renamed_entry *ren_entry; + struct mini_fo_inode_info *inode_info; + + err = 0; + meta_file=0; + meta_mnt=0; + + if(!dentry || !dentry->d_inode) { + printk(KERN_CRIT "mini_fo: meta_sync_r_list: \ + invalid dentry passed.\n"); + err = -1; + goto out; + } + inode_info = itopd(dentry->d_inode); + + if(inode_info->deleted_list_size < 0) { + err = -1; + goto out; + } + + /* ok, there is something to sync */ + + /* build the storage structure? */ + if(!dtohd2(dentry) && !itohi2(dentry->d_inode)) { + err = build_sto_structure(dentry->d_parent, dentry); + if(err) { + printk(KERN_CRIT "mini_fo: meta_sync_r_list: \ + build_sto_structure failed.\n"); + goto out; + } + } + meta_dentry = lookup_one_len(META_FILENAME, + dtohd2(dentry), + strlen(META_FILENAME)); + if(!meta_dentry->d_inode) { + /* We need to create a META-file */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) + vfs_create(dtohd2(dentry)->d_inode, + meta_dentry, S_IRUSR | S_IWUSR, NULL); +#else + vfs_create(dtohd2(dentry)->d_inode, + meta_dentry, S_IRUSR | S_IWUSR); +#endif + app_flag = 0; + } + /* need we truncate the meta file? */ + if(!app_flag) { + struct iattr newattrs; + newattrs.ia_size = 0; + newattrs.ia_valid = ATTR_SIZE | ATTR_CTIME; + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&meta_dentry->d_inode->i_mutex); +#else + down(&meta_dentry->d_inode->i_sem); +#endif + err = notify_change(meta_dentry, &newattrs); + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&meta_dentry->d_inode->i_mutex); +#else + up(&meta_dentry->d_inode->i_sem); +#endif + if(err || meta_dentry->d_inode->i_size != 0) { + printk(KERN_CRIT "mini_fo: meta_sync_r_list: \ + ERROR truncating meta file.\n"); + goto out_err_close; + } + } + + /* open META-file for writing */ + meta_file = dentry_open(meta_dentry, meta_mnt, 0x1); + if(!meta_file || IS_ERR(meta_file)) { + printk(KERN_CRIT "mini_fo: meta_sync_r_list: \ + ERROR opening meta file.\n"); + /* we don't mntget so we dont't mntput (for now) + * mntput(meta_mnt); + */ + dput(meta_dentry); + err = -1; + goto out; + } + + /* check if fs supports writing */ + if(!meta_file->f_op->write) { + printk(KERN_CRIT "mini_fo: meta_sync_r_list: \ + ERROR, fs does not support writing.\n"); + goto out_err_close; + } + + meta_file->f_pos = meta_dentry->d_inode->i_size; /* append */ + old_fs = get_fs(); + set_fs(KERNEL_DS); + + /* here we go... */ + list_for_each(tmp, &inode_info->renamed_list) { + ren_entry = list_entry(tmp, struct renamed_entry, list); + /* size: + * 2 for "R ", old_len+new_len for names, 1 blank+1 \n */ + buf_len = ren_entry->old_len + ren_entry->new_len + 4; + buf = (char *) kmalloc(buf_len, GFP_KERNEL); + if (!buf) { + printk(KERN_CRIT "mini_fo: meta_sync_r_list: \ + out of mem.\n"); + return -ENOMEM; + } + buf[0] = 'R'; + buf[1] = ' '; + strncpy(buf + 2, ren_entry->old_name, ren_entry->old_len); + buf[ren_entry->old_len + 2] = ' '; + strncpy(buf + ren_entry->old_len + 3, + ren_entry->new_name, ren_entry->new_len); + buf[buf_len - 1] = '\n'; + bytes = meta_file->f_op->write(meta_file, buf, + buf_len, &meta_file->f_pos); + if(bytes != buf_len) { + printk(KERN_CRIT "mini_fo: meta_sync_r_list: \ + ERROR writing.\n"); + err |= -1; + } + kfree(buf); + } + set_fs(old_fs); + + out_err_close: + fput(meta_file); + out: + return err; +} + +int meta_check_d_entry(dentry_t *dentry, const char *name, int len) +{ + if(!dentry || !dentry->d_inode) + printk(KERN_CRIT "mini_fo: meta_check_d_dentry: \ + invalid dentry passed.\n"); + return __meta_check_d_entry(dentry->d_inode, name, len); +} + +int __meta_check_d_entry(inode_t *inode, const char *name, int len) +{ + struct list_head *tmp; + struct deleted_entry *del_entry; + struct mini_fo_inode_info *inode_info; + + if(!inode || !itopd(inode)) + printk(KERN_CRIT "mini_fo: __meta_check_d_dentry: \ + invalid inode passed.\n"); + + inode_info = itopd(inode); + + if(inode_info->deleted_list_size <= 0) + return 0; + + list_for_each(tmp, &inode_info->deleted_list) { + del_entry = list_entry(tmp, struct deleted_entry, list); + if(del_entry->len != len) + continue; + + if(!strncmp(del_entry->name, name, len)) + return 1; + } + return 0; +} + +/* + * check if file has been renamed and return path to orig. base dir. + * Implements no error return values so far, what of course sucks. + * String is null terminated.' + */ +char* meta_check_r_entry(dentry_t *dentry, const char *name, int len) +{ + if(!dentry || !dentry->d_inode) { + printk(KERN_CRIT "mini_fo: meta_check_r_dentry: \ + invalid dentry passed.\n"); + return NULL; + } + return __meta_check_r_entry(dentry->d_inode, name, len); +} + +char* __meta_check_r_entry(inode_t *inode, const char *name, int len) +{ + struct list_head *tmp; + struct renamed_entry *ren_entry; + struct mini_fo_inode_info *inode_info; + char *old_path; + + if(!inode || !itopd(inode)) { + printk(KERN_CRIT "mini_fo: meta_check_r_dentry: \ + invalid inode passed.\n"); + return NULL; + } + inode_info = itopd(inode); + + if(inode_info->renamed_list_size <= 0) + return NULL; + + list_for_each(tmp, &inode_info->renamed_list) { + ren_entry = list_entry(tmp, struct renamed_entry, list); + if(ren_entry->new_len != len) + continue; + + if(!strncmp(ren_entry->new_name, name, len)) { + old_path = (char *) + kmalloc(ren_entry->old_len+1, GFP_KERNEL); + strncpy(old_path, + ren_entry->old_name, + ren_entry->old_len); + old_path[ren_entry->old_len]='\0'; + return old_path; + } + } + return NULL; +} + +/* + * This version only checks if entry exists and return: + * 1 if exists, + * 0 if not, + * -1 if error. + */ +int meta_is_r_entry(dentry_t *dentry, const char *name, int len) +{ + if(!dentry || !dentry->d_inode) { + printk(KERN_CRIT "mini_fo: meta_check_r_dentry [2]: \ + invalid dentry passed.\n"); + return -1; + } + return __meta_is_r_entry(dentry->d_inode, name, len); +} + +int __meta_is_r_entry(inode_t *inode, const char *name, int len) +{ + struct list_head *tmp; + struct renamed_entry *ren_entry; + struct mini_fo_inode_info *inode_info; + + if(!inode || !itopd(inode)) { + printk(KERN_CRIT "mini_fo: meta_check_r_dentry [2]: \ + invalid inode passed.\n"); + return -1; + } + inode_info = itopd(inode); + + if(inode_info->renamed_list_size <= 0) + return -1; + + list_for_each(tmp, &inode_info->renamed_list) { + ren_entry = list_entry(tmp, struct renamed_entry, list); + if(ren_entry->new_len != len) + continue; + + if(!strncmp(ren_entry->new_name, name, len)) + return 1; + } + return 0; +} + Index: linux-2.6.25.1/fs/mini_fo/mini_fo.h =================================================================== --- /dev/null +++ linux-2.6.25.1/fs/mini_fo/mini_fo.h @@ -0,0 +1,510 @@ +/* + * Copyright (c) 1997-2003 Erez Zadok + * Copyright (c) 2001-2003 Stony Brook University + * + * For specific licensing information, see the COPYING file distributed with + * this package, or get one from ftp://ftp.filesystems.org/pub/fist/COPYING. + * + * This Copyright notice must be kept intact and distributed with all + * fistgen sources INCLUDING sources generated by fistgen. + */ +/* + * Copyright (C) 2004, 2005 Markus Klotzbuecher + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ + +/* + * $Id$ + */ + +#ifndef __MINI_FO_H_ +#define __MINI_FO_H_ + +#ifdef __KERNEL__ + +/* META stuff */ +#define META_FILENAME "META_dAfFgHE39ktF3HD2sr" + +/* use xattrs? */ +#define XATTR + +/* File attributes that when changed, result in a file beeing copied to storage */ +#define COPY_FLAGS ATTR_MODE | ATTR_UID | ATTR_GID | ATTR_SIZE + +/* + * mini_fo filestates + */ +#define MODIFIED 1 +#define UNMODIFIED 2 +#define CREATED 3 +#define DEL_REWRITTEN 4 +#define DELETED 5 +#define NON_EXISTANT 6 + +/* fist file systems superblock magic */ +# define MINI_FO_SUPER_MAGIC 0xf15f + +/* + * STRUCTURES: + */ + +/* mini_fo inode data in memory */ +struct mini_fo_inode_info { + inode_t *wii_inode; + inode_t *wii_inode2; /* pointer to storage inode */ + + /* META-data lists */ + /* deleted list, ex wol */ + struct list_head deleted_list; + int deleted_list_size; + + /* renamed list */ + struct list_head renamed_list; + int renamed_list_size; + + /* add other lists here ... */ +}; + +/* mini_fo dentry data in memory */ +struct mini_fo_dentry_info { + dentry_t *wdi_dentry; + dentry_t *wdi_dentry2; /* pointer to storage dentry */ + unsigned int state; /* state of the mini_fo dentry */ +}; + + +/* mini_fo super-block data in memory */ +struct mini_fo_sb_info { + super_block_t *wsi_sb, *wsi_sb2; /* mk: might point to the same sb */ + struct vfsmount *hidden_mnt, *hidden_mnt2; + dentry_t *base_dir_dentry; + dentry_t *storage_dir_dentry; + ; +}; + +/* readdir_data, readdir helper struct */ +struct readdir_data { + struct list_head ndl_list; /* linked list head ptr */ + int ndl_size; /* list size */ + int sto_done; /* flag to show that the storage dir entries have + * all been read an now follow base entries */ +}; + +/* file private data. */ +struct mini_fo_file_info { + struct file *wfi_file; + struct file *wfi_file2; /* pointer to storage file */ + struct readdir_data rd; +}; + +/* struct ndl_entry */ +struct ndl_entry { + struct list_head list; + char *name; + int len; +}; + +/******************************** + * META-data structures + ********************************/ + +/* deleted entry */ +struct deleted_entry { + struct list_head list; + char *name; + int len; +}; + +/* renamed entry */ +struct renamed_entry { + struct list_head list; + char *old_name; /* old directory with full path */ + int old_len; /* length of above string */ + char *new_name; /* new directory name */ + int new_len; /* length of above string */ +}; + +/* attr_change entry */ +struct attr_change_entry { + struct list_head list; + char *name; + int len; +}; + +/* link entry */ +struct link_entry { + struct list_head list; + int links_moved; + int inum_base; + int inum_sto; + char *weird_name; + int weird_name_len; +}; + + +/* Some other stuff required for mini_fo_filldir64, copied from + * fs/readdir.c + */ + +#define ROUND_UP64(x) (((x)+sizeof(u64)-1) & ~(sizeof(u64)-1)) +#define NAME_OFFSET(de) ((int) ((de)->d_name - (char *) (de))) + + +struct linux_dirent64 { + u64 d_ino; + s64 d_off; + unsigned short d_reclen; + unsigned char d_type; + char d_name[0]; +}; + + +struct getdents_callback64 { + struct linux_dirent64 * current_dir; + struct linux_dirent64 * previous; + int count; + int error; +}; + +struct linux_dirent { + unsigned long d_ino; + unsigned long d_off; + unsigned short d_reclen; + char d_name[1]; +}; + +struct getdents_callback { + struct linux_dirent * current_dir; + struct linux_dirent * previous; + int count; + int error; +}; + + +/* + * MACROS: + */ + +/* file TO private_data */ +# define ftopd(file) ((struct mini_fo_file_info *)((file)->private_data)) +# define __ftopd(file) ((file)->private_data) +/* file TO hidden_file */ +# define ftohf(file) ((ftopd(file))->wfi_file) +# define ftohf2(file) ((ftopd(file))->wfi_file2) + +/* inode TO private_data */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,18) +# define itopd(ino) ((struct mini_fo_inode_info *)(ino)->i_private) +# define __itopd(ino) ((ino)->i_private) +#else +# define itopd(ino) ((struct mini_fo_inode_info *)(ino)->u.generic_ip) +# define __itopd(ino) ((ino)->u.generic_ip) +#endif +/* inode TO hidden_inode */ +# define itohi(ino) (itopd(ino)->wii_inode) +# define itohi2(ino) (itopd(ino)->wii_inode2) + +/* superblock TO private_data */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) +# define stopd(super) ((struct mini_fo_sb_info *)(super)->s_fs_info) +# define __stopd(super) ((super)->s_fs_info) +#else +# define stopd(super) ((struct mini_fo_sb_info *)(super)->u.generic_sbp) +# define __stopd(super) ((super)->u.generic_sbp) +#endif + +/* unused? # define vfs2priv stopd */ +/* superblock TO hidden_superblock */ + +# define stohs(super) (stopd(super)->wsi_sb) +# define stohs2(super) (stopd(super)->wsi_sb2) + +/* dentry TO private_data */ +# define dtopd(dentry) ((struct mini_fo_dentry_info *)(dentry)->d_fsdata) +# define __dtopd(dentry) ((dentry)->d_fsdata) +/* dentry TO hidden_dentry */ +# define dtohd(dent) (dtopd(dent)->wdi_dentry) +# define dtohd2(dent) (dtopd(dent)->wdi_dentry2) + +/* dentry to state */ +# define dtost(dent) (dtopd(dent)->state) +# define sbt(sb) ((sb)->s_type->name) + +#define IS_WRITE_FLAG(flag) (flag & (O_RDWR | O_WRONLY | O_APPEND)) +#define IS_COPY_FLAG(flag) (flag & (COPY_FLAGS)) + +/* macros to simplify non-SCA code */ +# define MALLOC_PAGE_POINTERS(hidden_pages, num_hidden_pages) +# define MALLOC_PAGEDATA_POINTERS(hidden_pages_data, num_hidden_pages) +# define FREE_PAGE_POINTERS(hidden_pages, num) +# define FREE_PAGEDATA_POINTERS(hidden_pages_data, num) +# define FOR_EACH_PAGE +# define CURRENT_HIDDEN_PAGE hidden_page +# define CURRENT_HIDDEN_PAGEDATA hidden_page_data +# define CURRENT_HIDDEN_PAGEINDEX page->index + +/* + * EXTERNALS: + */ +extern struct file_operations mini_fo_main_fops; +extern struct file_operations mini_fo_dir_fops; +extern struct inode_operations mini_fo_main_iops; +extern struct inode_operations mini_fo_dir_iops; +extern struct inode_operations mini_fo_symlink_iops; +extern struct super_operations mini_fo_sops; +extern struct dentry_operations mini_fo_dops; +extern struct vm_operations_struct mini_fo_shared_vmops; +extern struct vm_operations_struct mini_fo_private_vmops; +extern struct address_space_operations mini_fo_aops; + +#if 0 /* unused by mini_fo */ +extern int mini_fo_interpose(dentry_t *hidden_dentry, dentry_t *this_dentry, super_block_t *sb, int flag); +#if defined(FIST_FILTER_DATA) || defined(FIST_FILTER_SCA) +extern page_t *mini_fo_get1page(file_t *file, int index); +extern int mini_fo_fill_zeros(file_t *file, page_t *page, unsigned from); +# endif /* FIST_FILTER_DATA || FIST_FILTER_SCA */ + + +# define mini_fo_hidden_dentry(d) __mini_fo_hidden_dentry(__FILE__,__FUNCTION__,__LINE__,(d)) +# define mini_fo_hidden_sto_dentry(d) __mini_fo_hidden_sto_dentry(__FILE__,__FUNCTION__,__LINE__,(d)) + +extern dentry_t *__mini_fo_hidden_dentry(char *file, char *func, int line, dentry_t *this_dentry); +extern dentry_t *__mini_fo_hidden_sto_dentry(char *file, char *func, int line, dentry_t *this_dentry); + +extern int mini_fo_read_file(const char *filename, void *buf, int len); +extern int mini_fo_write_file(const char *filename, void *buf, int len); +extern dentry_t *fist_lookup(dentry_t *dir, const char *name, vnode_t **out, uid_t uid, gid_t gid); +#endif /* unused by mini_fo */ + +/* state transition functions */ +extern int nondir_unmod_to_mod(dentry_t *dentry, int cp_flag); +extern int nondir_del_rew_to_del(dentry_t *dentry); +extern int nondir_creat_to_del(dentry_t *dentry); +extern int nondir_mod_to_del(dentry_t *dentry); +extern int nondir_unmod_to_del(dentry_t *dentry); + +extern int dir_unmod_to_mod(dentry_t *dentry); + +/* rename specials */ +extern int rename_directory(inode_t *old_dir, dentry_t *old_dentry, inode_t *new_dir, dentry_t *new_dentry); +extern int rename_nondir(inode_t *old_dir, dentry_t *old_dentry, inode_t *new_dir, dentry_t *new_dentry); + +/* misc stuff */ +extern int mini_fo_tri_interpose(dentry_t *hidden_dentry, + dentry_t *hidden_sto_dentry, + dentry_t *dentry, + super_block_t *sb, int flag); + +extern int mini_fo_cp_cont(dentry_t *tgt_dentry, struct vfsmount *tgt_mnt, + dentry_t *src_dentry, struct vfsmount *src_mnt); + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) +extern int mini_fo_create(inode_t *dir, dentry_t *dentry, int mode, struct nameidata *nd); + +extern int create_sto_nod(dentry_t *dentry, int mode, dev_t dev); +extern int create_sto_reg_file(dentry_t *dentry, int mode, struct nameidata *nd); +#else +extern int mini_fo_create(inode_t *dir, dentry_t *dentry, int mode); + +extern int create_sto_nod(dentry_t *dentry, int mode, int dev); +extern int create_sto_reg_file(dentry_t *dentry, int mode); +#endif + +extern int create_sto_dir(dentry_t *dentry, int mode); + +extern int exists_in_storage(dentry_t *dentry); +extern int is_mini_fo_existant(dentry_t *dentry); +extern int get_neg_sto_dentry(dentry_t *dentry); +extern int build_sto_structure(dentry_t *dir, dentry_t *dentry); +extern int get_mini_fo_bpath(dentry_t *dentry, char **bpath, int *bpath_len); +extern dentry_t *bpath_walk(super_block_t *sb, char *bpath); +extern int bpath_put(dentry_t *dentry); + +/* check_mini_fo types functions */ +extern int check_mini_fo_dentry(dentry_t *dentry); +extern int check_mini_fo_file(file_t *file); +extern int check_mini_fo_inode(inode_t *inode); + +/* General meta functions, can be called from outside of meta.c */ +extern int meta_build_lists(dentry_t *dentry); +extern int meta_put_lists(dentry_t *dentry); +extern int __meta_put_lists(inode_t *inode); + +extern int meta_add_d_entry(dentry_t *dentry, const char *name, int len); +extern int meta_add_r_entry(dentry_t *dentry, + const char *old_name, int old_len, + const char *new_name, int new_len); + +extern int meta_remove_r_entry(dentry_t *dentry, const char *name, int len); + +extern int meta_check_d_entry(dentry_t *dentry, const char *name, int len); +extern int __meta_check_d_entry(inode_t *inode, const char *name, int len); + +extern char* meta_check_r_entry(dentry_t *dentry, const char *name, int len); +extern char* __meta_check_r_entry(inode_t *inode, const char *name, int len); +extern int meta_is_r_entry(dentry_t *dentry, const char *name, int len); +extern int __meta_is_r_entry(inode_t *inode, const char *name, int len); + +/* Specific meta functions, should be called only inside meta.c */ +extern int __meta_put_d_list(inode_t *inode); +extern int __meta_put_r_list(inode_t *inode); + +extern int meta_list_add_d_entry(dentry_t *dentry, + const char *name, int len); +extern int meta_list_add_r_entry(dentry_t *dentry, + const char *old_name, int old_len, + const char *new_name, int new_len); + +extern int meta_list_remove_r_entry(dentry_t *dentry, + const char *name, int len); + +extern int __meta_list_remove_r_entry(inode_t *inode, + const char *name, int len); + +extern int meta_write_d_entry(dentry_t *dentry, const char *name, int len); +extern int meta_write_r_entry(dentry_t *dentry, + const char *old_name, int old_len, + const char *new_name, int new_len); + +extern int meta_sync_lists(dentry_t *dentry); +extern int meta_sync_d_list(dentry_t *dentry, int app_flag); +extern int meta_sync_r_list(dentry_t *dentry, int app_flag); + +/* ndl stuff */ +extern int ndl_add_entry(struct readdir_data *rd, const char *name, int len); +extern void ndl_put_list(struct readdir_data *rd); +extern int ndl_check_entry(struct readdir_data *rd, + const char *name, int len); + + +# define copy_inode_size(dst, src) \ + dst->i_size = src->i_size; \ + dst->i_blocks = src->i_blocks; + +static inline void +fist_copy_attr_atime(inode_t *dest, const inode_t *src) +{ + ASSERT(dest != NULL); + ASSERT(src != NULL); + dest->i_atime = src->i_atime; +} +static inline void +fist_copy_attr_times(inode_t *dest, const inode_t *src) +{ + ASSERT(dest != NULL); + ASSERT(src != NULL); + dest->i_atime = src->i_atime; + dest->i_mtime = src->i_mtime; + dest->i_ctime = src->i_ctime; +} +static inline void +fist_copy_attr_timesizes(inode_t *dest, const inode_t *src) +{ + ASSERT(dest != NULL); + ASSERT(src != NULL); + dest->i_atime = src->i_atime; + dest->i_mtime = src->i_mtime; + dest->i_ctime = src->i_ctime; + copy_inode_size(dest, src); +} +static inline void +fist_copy_attr_all(inode_t *dest, const inode_t *src) +{ + ASSERT(dest != NULL); + ASSERT(src != NULL); + dest->i_mode = src->i_mode; + dest->i_nlink = src->i_nlink; + dest->i_uid = src->i_uid; + dest->i_gid = src->i_gid; + dest->i_rdev = src->i_rdev; + dest->i_atime = src->i_atime; + dest->i_mtime = src->i_mtime; + dest->i_ctime = src->i_ctime; +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,19) + dest->i_blksize = src->i_blksize; +#endif +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,12) + dest->i_blkbits = src->i_blkbits; +# endif /* linux 2.4.12 and newer */ + copy_inode_size(dest, src); +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0) + dest->i_attr_flags = src->i_attr_flags; +#else + dest->i_flags = src->i_flags; +#endif +} + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) +/* copied from linux/fs.h */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) +static inline void double_lock(struct dentry *d1, struct dentry *d2) +{ + struct mutex *m1 = &d1->d_inode->i_mutex; + struct mutex *m2 = &d2->d_inode->i_mutex; + if (m1 != m2) { + if ((unsigned long) m1 < (unsigned long) m2) { + struct mutex *tmp = m2; + m2 = m1; m1 = tmp; + } + mutex_lock(m1); + } + mutex_lock(m2); +} + +static inline void double_unlock(struct dentry *d1, struct dentry *d2) +{ + struct mutex *m1 = &d1->d_inode->i_mutex; + struct mutex *m2 = &d2->d_inode->i_mutex; + mutex_unlock(m1); + if (m1 != m2) + mutex_unlock(m2); + dput(d1); + dput(d2); +} + +#else +static inline void double_down(struct semaphore *s1, struct semaphore *s2) +{ + if (s1 != s2) { + if ((unsigned long) s1 < (unsigned long) s2) { + struct semaphore *tmp = s2; + s2 = s1; s1 = tmp; + } + down(s1); + } + down(s2); +} + +static inline void double_up(struct semaphore *s1, struct semaphore *s2) +{ + up(s1); + if (s1 != s2) + up(s2); +} + +static inline void double_lock(struct dentry *d1, struct dentry *d2) +{ + double_down(&d1->d_inode->i_sem, &d2->d_inode->i_sem); +} + +static inline void double_unlock(struct dentry *d1, struct dentry *d2) +{ + double_up(&d1->d_inode->i_sem,&d2->d_inode->i_sem); + dput(d1); + dput(d2); +} +#endif /* if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) */ +#endif /* if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) */ +#endif /* __KERNEL__ */ + +/* + * Definitions for user and kernel code + */ + +/* ioctls */ + +#endif /* not __MINI_FO_H_ */ Index: linux-2.6.25.1/fs/mini_fo/mini_fo-merge =================================================================== --- /dev/null +++ linux-2.6.25.1/fs/mini_fo/mini_fo-merge @@ -0,0 +1,180 @@ +#!/bin/bash +# +# Copyright (C) 2005 Markus Klotzbuecher +# This program is free software; you can redistribute it and/or +# modify it under the terms of the GNU General Public License +# as published by the Free Software Foundation; either version +# 2 of the License, or (at your option) any later version. +# + +BASE= +STO= +HELP= +DRYRUN= +VERBOSE= +TMP="/tmp/" +META_NAME="META_dAfFgHE39ktF3HD2sr" +SKIP_DEL_LIST="skip-delete-list.mini_fo-merge" + +COMMAND= +exec_command() +{ + if [ x$DRYRUN == "xset" ]; then + echo " would run: $COMMAND" + elif ! [ x$DRYRUN == "xset" ]; then + if [ x$VERBOSE == "xset" ]; then + echo " running: $COMMAND" + fi + eval $COMMAND + fi +} + +usage() +{ +cat < -s +Version 0.1 + +This script merges the contents of a mini_fo storage file system back +to the base file system. + +!!! Warning: This will modify the base filesystem and can destroy data + if used wrongly. + +Options: + -b + the directory of the base file system. + + -s + the directory of the storage file system. + + -d dry run, will not change anything and print the commands that + would be executed. + + -t tmp dir for storing temporary file. default: $TMP + + -v show what operations are performed. + + -h displays this message. + +EOF +} + +# parse parameters +while getopts hdvt:b:s: OPTS + do + case $OPTS in + h) HELP="set";; + d) DRYRUN="set";; + v) VERBOSE="set";; + b) BASE="$OPTARG";; + s) STO="$OPTARG";; + t) TMP="$OPTARG";; + ?) usage + exit 1;; + esac +done + +if [ "x$HELP" == "xset" ]; then + usage + exit -1 +fi + +if ! [ -d "$BASE" ] || ! [ -d "$STO" ]; then + echo -e "$0:\n Error, -s and/or -b argument missing. type $0 -h for help." + exit -1; +fi + +# get full paths +pushd $STO; STO=`pwd`; popd +pushd $BASE; BASE=`pwd`; popd +TMP=${TMP%/} + + +cat< /dev/null +find . -name $META_NAME -type f -print0 | xargs -0 -e grep -e '^R ' | tr -s ':R' ' ' | while read ENTRY; do + echo "entry: $ENTRY" + META_FILE=`echo $ENTRY | cut -d ' ' -f 1` + OLD_B_DIR=`echo $ENTRY | cut -d ' ' -f 2 | sed -e 's/\///'` + NEW_NAME=`echo $ENTRY | cut -d ' ' -f 3` + NEW_B_DIR=`echo $META_FILE | sed -e "s/$META_NAME/$NEW_NAME/" | sed -e 's/^\.\///'` + echo "META_FILE: $META_FILE" + echo "OLD_B_DIR: $OLD_B_DIR" + echo "NEW_NAME: $NEW_NAME" + echo "NEW_B_DIR: $NEW_B_DIR" + + pushd $BASE &> /dev/null + # remove an existing dir in storage + COMMAND="rm -rf $NEW_B_DIR"; exec_command + COMMAND="cp -R $OLD_B_DIR $NEW_B_DIR"; exec_command + echo "" + popd &> /dev/null + + # remember this dir to exclude it from deleting later + echo $NEW_B_DIR >> $TMP/$SKIP_DEL_LIST +done + +# delete all whiteouted files from base +echo -e "\nDeleting whiteout'ed files from base file system..." +find . -name $META_NAME -type f -print0 | xargs -0 -e grep -e '^D ' | sed -e 's/:D//' | while read ENTRY; do + META_FILE=`echo $ENTRY | cut -d ' ' -f 1` + DEL_NAME=`echo $ENTRY | cut -d ' ' -f 2` + DEL_FILE=`echo $META_FILE | sed -e "s/$META_NAME/$DEL_NAME/" | sed -e 's/^\.\///'` + grep -x $DEL_FILE $TMP/$SKIP_DEL_LIST &> /dev/null + if [ $? -ne 0 ]; then + pushd $BASE &> /dev/null + COMMAND="rm -rf $DEL_FILE"; exec_command + popd &> /dev/null + else + echo " excluding: $DEL_FILE as in skip-del-list." + fi +done + +# create all dirs and update permissions +echo -e "\nSetting up directory structures in base file system..." +find . -type d | sed -e 's/^\.\///' | while read DIR; do + PERMS=`stat -c %a $DIR` + DIR_UID=`stat -c %u $DIR` + DIR_GID=`stat -c %g $DIR` + pushd $BASE &> /dev/null + if ! [ -d $DIR ]; then + COMMAND="mkdir -p $DIR"; exec_command + fi + COMMAND="chmod $PERMS $DIR"; exec_command + COMMAND="chown $DIR_UID:$DIR_GID $DIR"; exec_command + popd &> /dev/null +done + +# merge all non-directory files +echo -e "\nMerging all non-directory files...." +for i in b c p f l s; do + find . -type $i | sed -e 's/^\.\///' | grep -v "$META_NAME" | while read FILE; do + pushd $BASE #&> /dev/null + COMMAND="cp -df $STO/$FILE $BASE/$FILE"; exec_command + popd &> /dev/null + done +done +popd &> /dev/null + +#rm $TMP/$SKIP_DEL_LIST + +echo "Done!" Index: linux-2.6.25.1/fs/mini_fo/mini_fo-overlay =================================================================== --- /dev/null +++ linux-2.6.25.1/fs/mini_fo/mini_fo-overlay @@ -0,0 +1,130 @@ +#!/bin/bash +# +# Copyright (C) 2005 Markus Klotzbuecher +# This program is free software; you can redistribute it and/or +# modify it under the terms of the GNU General Public License +# as published by the Free Software Foundation; either version +# 2 of the License, or (at your option) any later version. +# + +HELP= +SUFF= +MNTP= +MNT_DIR="/mnt" +STO= +STO_DIR="/tmp" +BASE= + +usage() +{ +cat <" in $STO_DIR, +and mount point "mini_fo-" in $MNT_DIR. + +Options: + -s + add given suffix to storage directory and the mount + point. This is usefull for overlaying one base directory + several times and avoiding conflicts with storage directory + names and mount points. + + -d + change the directory in which the storage directory will be + created (default is currently "$STO_DIR". + + -m + use an alternative directory to create the mini_fo + mountpoint (default is currently "$MNT_DIR". + + -h displays this message. + +EOF +exit 1; +} + +while getopts hm:s:d: OPTS + do + case $OPTS in + s) SUFF="$OPTARG";; + d) STO_DIR="$OPTARG";; + m) MNT_DIR="$OPTARG";; + h) HELP="set";; + ?) usage + exit 1;; + esac +done +shift $(($OPTIND - 1)) + +BASE="$1" + +if [ "x$HELP" == "xset" ]; then + usage + exit -1 +fi + +# fix suffix +if [ "x$SUFF" != "x" ]; then + SUFF="-$SUFF" +fi + +# kill trailing slashes +MNT_DIR=${MNT_DIR%/} +STO_DIR=${STO_DIR%/} +BASE=${BASE%/} + + +if ! [ -d "$BASE" ]; then + echo "invalid base dir $BASE, run $0 -h for help." + exit -1 +fi + +# check opts +if ! [ -d "$MNT_DIR" ]; then + echo "invalid mount dir $MNT_DIR, run $0 -h for help." + exit -1 +fi + +if ! [ -d "$STO_DIR" ]; then + echo "invalid sto_dir_dir $STO_DIR, run $0 -h for help." + exit -1 +fi + +MNTP="$MNT_DIR/mini_fo-`basename $BASE`$SUFF" +STO="$STO_DIR/sto-`basename $BASE`$SUFF" + +# create the mount point if it doesn't exist +mkdir -p $MNTP +if [ $? -ne 0 ]; then + echo "Error, failed to create mount point $MNTP" +fi + +mkdir -p $STO +if [ $? -ne 0 ]; then + echo "Error, failed to create storage dir $STO" +fi + +# check if fs is already mounted +mount | grep mini_fo | grep $MNTP &> /dev/null +if [ $? -eq 0 ]; then + echo "Error, existing mini_fo mount at $MNTP." + exit -1 +fi + +mount | grep mini_fo | grep $STO &> /dev/null +if [ $? -eq 0 ]; then + echo "Error, $STO seems to be used already." + exit -1 +fi + +# mount +mount -t mini_fo -o base=$BASE,sto=$STO $BASE $MNTP + +if [ $? -ne 0 ]; then + echo "Error, mounting failed, maybe no permisson to mount?" +fi Index: linux-2.6.25.1/fs/mini_fo/mmap.c =================================================================== --- /dev/null +++ linux-2.6.25.1/fs/mini_fo/mmap.c @@ -0,0 +1,637 @@ +/* + * Copyright (c) 1997-2003 Erez Zadok + * Copyright (c) 2001-2003 Stony Brook University + * + * For specific licensing information, see the COPYING file distributed with + * this package, or get one from ftp://ftp.filesystems.org/pub/fist/COPYING. + * + * This Copyright notice must be kept intact and distributed with all + * fistgen sources INCLUDING sources generated by fistgen. + */ +/* + * Copyright (C) 2004, 2005 Markus Klotzbuecher + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ + +/* + * $Id$ + */ + +#ifdef HAVE_CONFIG_H +# include +#endif /* HAVE_CONFIG_H */ + +#include "fist.h" +#include "mini_fo.h" + + +#ifdef FIST_COUNT_WRITES +/* for counting writes in the middle vs. regular writes */ +unsigned long count_writes = 0, count_writes_middle = 0; +#endif /* FIST_COUNT_WRITES */ + +/* forward declaration of commit write and prepare write */ +STATIC int mini_fo_commit_write(file_t *file, page_t *page, unsigned from, unsigned to); +STATIC int mini_fo_prepare_write(file_t *file, page_t *page, unsigned from, unsigned to); + + +/* + * Function for handling creation of holes when lseek-ing past the + * end of the file and then writing some data. + */ +int +mini_fo_fill_zeros(file_t* file, page_t *page, unsigned from) +{ + int err = 0; + dentry_t *dentry = file->f_dentry; + inode_t *inode = dentry->d_inode; + page_t *tmp_page; + int index; + + print_entry_location(); + + for (index = inode->i_size >> PAGE_CACHE_SHIFT; index < page->index; index++) { + tmp_page = mini_fo_get1page(file, index); + if (IS_ERR(tmp_page)) { + err = PTR_ERR(tmp_page); + goto out; + } + + /* + * zero out rest of the contents of the page between the appropriate + * offsets. + */ + memset((char*)page_address(tmp_page) + (inode->i_size & ~PAGE_CACHE_MASK), 0, PAGE_CACHE_SIZE - (inode->i_size & ~PAGE_CACHE_MASK)); + + if (! (err = mini_fo_prepare_write(file, tmp_page, 0, PAGE_CACHE_SIZE))) + err = mini_fo_commit_write(file, tmp_page, 0, PAGE_CACHE_SIZE); + + page_cache_release(tmp_page); + if (err < 0) + goto out; + if (current->need_resched) + schedule(); + } + + /* zero out appropriate parts of last page */ + + /* + * if the encoding type is block, then adjust the 'from' (where the + * zeroing will start) offset appropriately + */ + from = from & (~(FIST_ENCODING_BLOCKSIZE - 1)); + + if ((from - (inode->i_size & ~PAGE_CACHE_MASK)) > 0) { + + memset((char*)page_address(page) + (inode->i_size & ~PAGE_CACHE_MASK), 0, from - (inode->i_size & ~PAGE_CACHE_MASK)); + if (! (err = mini_fo_prepare_write(file, page, 0, PAGE_CACHE_SIZE))) + err = mini_fo_commit_write(file, page, 0, PAGE_CACHE_SIZE); + + if (err < 0) + goto out; + if (current->need_resched) + schedule(); + } + + out: + print_exit_status(err); + return err; +} + + + +STATIC int +mini_fo_writepage(page_t *page) +{ + int err = -EIO; + inode_t *inode; + inode_t *hidden_inode; + page_t *hidden_page; + char *kaddr, *hidden_kaddr; + + print_entry_location(); + + inode = page->mapping->host; + hidden_inode = itohi(inode); + + /* + * writepage is called when shared mmap'ed files need to write + * their pages, while prepare/commit_write are called from the + * non-paged write() interface. (However, in 2.3 the two interfaces + * share the same cache, while in 2.2 they didn't.) + * + * So we pretty much have to duplicate much of what commit_write does. + */ + + /* find lower page (returns a locked page) */ + hidden_page = grab_cache_page(hidden_inode->i_mapping, page->index); + if (!hidden_page) + goto out; + + /* get page address, and encode it */ + kaddr = (char *) kmap(page); + hidden_kaddr = (char*) kmap(hidden_page); + mini_fo_encode_block(kaddr, hidden_kaddr, PAGE_CACHE_SIZE, inode, inode->i_sb, page->index); + /* if encode_block could fail, then return error */ + kunmap(page); + kunmap(hidden_page); + + /* call lower writepage (expects locked page) */ + err = hidden_inode->i_mapping->a_ops->writepage(hidden_page); + + /* + * update mtime and ctime of lower level file system + * mini_fo' mtime and ctime are updated by generic_file_write + */ + hidden_inode->i_mtime = hidden_inode->i_ctime = CURRENT_TIME; + +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,1) + UnlockPage(hidden_page); /* b/c grab_cache_page locked it */ +# endif /* kernel older than 2.4.1 */ + page_cache_release(hidden_page); /* b/c grab_cache_page increased refcnt */ + + if (err) + ClearPageUptodate(page); + else + SetPageUptodate(page); + out: +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,1) + UnlockPage(page); +# endif /* kernel 2.4.1 and newer */ + print_exit_status(err); + return err; +} + + +/* + * get one page from cache or lower f/s, return error otherwise. + * returns unlocked, up-to-date page (if ok), with increased refcnt. + */ +page_t * +mini_fo_get1page(file_t *file, int index) +{ + page_t *page; + dentry_t *dentry; + inode_t *inode; + struct address_space *mapping; + int err; + + print_entry_location(); + + dentry = file->f_dentry; /* CPW: Moved below print_entry_location */ + inode = dentry->d_inode; + mapping = inode->i_mapping; + + fist_dprint(8, "%s: read page index %d pid %d\n", __FUNCTION__, index, current->pid); + if (index < 0) { + printk("%s BUG: index=%d\n", __FUNCTION__, index); + page = ERR_PTR(-EIO); + goto out; + } + page = read_cache_page(mapping, + index, + (filler_t *) mapping->a_ops->readpage, + (void *) file); + if (IS_ERR(page)) + goto out; + wait_on_page(page); + if (!Page_Uptodate(page)) { + lock_page(page); + err = mapping->a_ops->readpage(file, page); + if (err) { + page = ERR_PTR(err); + goto out; + } + wait_on_page(page); + if (!Page_Uptodate(page)) { + page = ERR_PTR(-EIO); + goto out; + } + } + + out: + print_exit_pointer(page); + return page; +} + + +/* + * get one page from cache or lower f/s, return error otherwise. + * similar to get1page, but doesn't guarantee that it will return + * an unlocked page. + */ +page_t * +mini_fo_get1page_cached(file_t *file, int index) +{ + page_t *page; + dentry_t *dentry; + inode_t *inode; + struct address_space *mapping; + int err; + + print_entry_location(); + + dentry = file->f_dentry; /* CPW: Moved below print_entry_location */ + inode = dentry->d_inode; + mapping = inode->i_mapping; + + fist_dprint(8, "%s: read page index %d pid %d\n", __FUNCTION__, index, current->pid); + if (index < 0) { + printk("%s BUG: index=%d\n", __FUNCTION__, index); + page = ERR_PTR(-EIO); + goto out; + } + page = read_cache_page(mapping, + index, + (filler_t *) mapping->a_ops->readpage, + (void *) file); + if (IS_ERR(page)) + goto out; + + out: + print_exit_pointer(page); + return page; +} + + +/* + * readpage is called from generic_page_read and the fault handler. + * If your file system uses generic_page_read for the read op, it + * must implement readpage. + * + * Readpage expects a locked page, and must unlock it. + */ +STATIC int +mini_fo_do_readpage(file_t *file, page_t *page) +{ + int err = -EIO; + dentry_t *dentry; + file_t *hidden_file = NULL; + dentry_t *hidden_dentry; + inode_t *inode; + inode_t *hidden_inode; + char *page_data; + page_t *hidden_page; + char *hidden_page_data; + int real_size; + + print_entry_location(); + + dentry = file->f_dentry; /* CPW: Moved below print_entry_location */ + if (ftopd(file) != NULL) + hidden_file = ftohf(file); + hidden_dentry = dtohd(dentry); + inode = dentry->d_inode; + hidden_inode = itohi(inode); + + fist_dprint(7, "%s: requesting page %d from file %s\n", __FUNCTION__, page->index, dentry->d_name.name); + + MALLOC_PAGE_POINTERS(hidden_pages, num_hidden_pages); + MALLOC_PAGEDATA_POINTERS(hidden_pages_data, num_hidden_pages); + FOR_EACH_PAGE + CURRENT_HIDDEN_PAGE = NULL; + + /* find lower page (returns a locked page) */ + FOR_EACH_PAGE { + fist_dprint(8, "%s: Current page index = %d\n", __FUNCTION__, CURRENT_HIDDEN_PAGEINDEX); + CURRENT_HIDDEN_PAGE = read_cache_page(hidden_inode->i_mapping, + CURRENT_HIDDEN_PAGEINDEX, + (filler_t *) hidden_inode->i_mapping->a_ops->readpage, + (void *) hidden_file); + if (IS_ERR(CURRENT_HIDDEN_PAGE)) { + err = PTR_ERR(CURRENT_HIDDEN_PAGE); + CURRENT_HIDDEN_PAGE = NULL; + goto out_release; + } + } + + /* + * wait for the page data to show up + * (signaled by readpage as unlocking the page) + */ + FOR_EACH_PAGE { + wait_on_page(CURRENT_HIDDEN_PAGE); + if (!Page_Uptodate(CURRENT_HIDDEN_PAGE)) { + /* + * call readpage() again if we returned from wait_on_page with a + * page that's not up-to-date; that can happen when a partial + * page has a few buffers which are ok, but not the whole + * page. + */ + lock_page(CURRENT_HIDDEN_PAGE); + err = hidden_inode->i_mapping->a_ops->readpage(hidden_file, + CURRENT_HIDDEN_PAGE); + if (err) { + CURRENT_HIDDEN_PAGE = NULL; + goto out_release; + } + wait_on_page(CURRENT_HIDDEN_PAGE); + if (!Page_Uptodate(CURRENT_HIDDEN_PAGE)) { + err = -EIO; + goto out_release; + } + } + } + + /* map pages, get their addresses */ + page_data = (char *) kmap(page); + FOR_EACH_PAGE + CURRENT_HIDDEN_PAGEDATA = (char *) kmap(CURRENT_HIDDEN_PAGE); + + /* if decode_block could fail, then return error */ + err = 0; + real_size = hidden_inode->i_size - (page->index << PAGE_CACHE_SHIFT); + if (real_size <= 0) + memset(page_data, 0, PAGE_CACHE_SIZE); + else if (real_size < PAGE_CACHE_SIZE) { + mini_fo_decode_block(hidden_page_data, page_data, real_size, inode, inode->i_sb, page->index); + memset(page_data + real_size, 0, PAGE_CACHE_SIZE - real_size); + } else + mini_fo_decode_block(hidden_page_data, page_data, PAGE_CACHE_SIZE, inode, inode->i_sb, page->index); + + FOR_EACH_PAGE + kunmap(CURRENT_HIDDEN_PAGE); + kunmap(page); + + out_release: + FOR_EACH_PAGE + if (CURRENT_HIDDEN_PAGE) + page_cache_release(CURRENT_HIDDEN_PAGE); /* undo read_cache_page */ + + FREE_PAGE_POINTERS(hidden_pages, num_hidden_pages); + FREE_PAGEDATA_POINTERS(hidden_pages_data, num_hidden_pages); + + out: + if (err == 0) + SetPageUptodate(page); + else + ClearPageUptodate(page); + + print_exit_status(err); + return err; +} + + +STATIC int +mini_fo_readpage(file_t *file, page_t *page) +{ + int err; + print_entry_location(); + + err = mini_fo_do_readpage(file, page); + + /* + * we have to unlock our page, b/c we _might_ have gotten a locked page. + * but we no longer have to wakeup on our page here, b/c UnlockPage does + * it + */ + UnlockPage(page); + + print_exit_status(err); + return err; +} + + +STATIC int +mini_fo_prepare_write(file_t *file, page_t *page, unsigned from, unsigned to) +{ + int err = 0; + + print_entry_location(); + + /* + * we call kmap(page) only here, and do the kunmap + * and the actual downcalls, including unlockpage and uncache + * in commit_write. + */ + kmap(page); + + /* fast path for whole page writes */ + if (from == 0 && to == PAGE_CACHE_SIZE) + goto out; + /* read the page to "revalidate" our data */ + /* call the helper function which doesn't unlock the page */ + if (!Page_Uptodate(page)) + err = mini_fo_do_readpage(file, page); + + out: + print_exit_status(err); + return err; +} + + + +STATIC int +mini_fo_commit_write(file_t *file, page_t *page, unsigned from, unsigned to) +{ + int err = -ENOMEM; + inode_t *inode; + inode_t *hidden_inode; + page_t *hidden_page; + file_t *hidden_file = NULL; + loff_t pos; + unsigned bytes = to - from; + unsigned hidden_from, hidden_to, hidden_bytes; + + print_entry_location(); + + inode = page->mapping->host; /* CPW: Moved below print_entry_location */ + hidden_inode = itohi(inode); + + ASSERT(file != NULL); + /* + * here we have a kmapped page, with data from the user copied + * into it. we need to encode_block it, and then call the lower + * commit_write. We also need to simulate same behavior of + * generic_file_write, and call prepare_write on the lower f/s first. + */ +#ifdef FIST_COUNT_WRITES + count_writes++; +# endif /* FIST_COUNT_WRITES */ + + /* this is append and/or extend -- we can't have holes so fill them in */ + if (page->index > (hidden_inode->i_size >> PAGE_CACHE_SHIFT)) { + page_t *tmp_page; + int index; + for (index = hidden_inode->i_size >> PAGE_CACHE_SHIFT; index < page->index; index++) { + tmp_page = mini_fo_get1page(file, index); + if (IS_ERR(tmp_page)) { + err = PTR_ERR(tmp_page); + goto out; + } + /* zero out the contents of the page at the appropriate offsets */ + memset((char*)page_address(tmp_page) + (inode->i_size & ~PAGE_CACHE_MASK), 0, PAGE_CACHE_SIZE - (inode->i_size & ~PAGE_CACHE_MASK)); + if (!(err = mini_fo_prepare_write(file, tmp_page, 0, PAGE_CACHE_SIZE))) + err = mini_fo_commit_write(file, tmp_page, 0, PAGE_CACHE_SIZE); + page_cache_release(tmp_page); + if (err < 0) + goto out; + if (current->need_resched) + schedule(); + } + } + + if (ftopd(file) != NULL) + hidden_file = ftohf(file); + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&hidden_inode->i_mutex); +#else + down(&hidden_inode->i_sem); +#endif + /* find lower page (returns a locked page) */ + hidden_page = grab_cache_page(hidden_inode->i_mapping, page->index); + if (!hidden_page) + goto out; + +#if FIST_ENCODING_BLOCKSIZE > 1 +# error encoding_blocksize greater than 1 is not yet supported +# endif /* FIST_ENCODING_BLOCKSIZE > 1 */ + + hidden_from = from & (~(FIST_ENCODING_BLOCKSIZE - 1)); + hidden_to = ((to + FIST_ENCODING_BLOCKSIZE - 1) & (~(FIST_ENCODING_BLOCKSIZE - 1))); + if ((page->index << PAGE_CACHE_SHIFT) + to > hidden_inode->i_size) { + + /* + * if this call to commit_write had introduced holes and the code + * for handling holes was invoked, then the beginning of this page + * must be zeroed out + * zero out bytes from 'size_of_file%pagesize' to 'from'. + */ + if ((hidden_from - (inode->i_size & ~PAGE_CACHE_MASK)) > 0) + memset((char*)page_address(page) + (inode->i_size & ~PAGE_CACHE_MASK), 0, hidden_from - (inode->i_size & ~PAGE_CACHE_MASK)); + + } + hidden_bytes = hidden_to - hidden_from; + + /* call lower prepare_write */ + err = -EINVAL; + if (hidden_inode->i_mapping && + hidden_inode->i_mapping->a_ops && + hidden_inode->i_mapping->a_ops->prepare_write) + err = hidden_inode->i_mapping->a_ops->prepare_write(hidden_file, + hidden_page, + hidden_from, + hidden_to); + if (err) + /* don't leave locked pages behind, esp. on an ENOSPC */ + goto out_unlock; + + fist_dprint(8, "%s: encoding %d bytes\n", __FUNCTION__, hidden_bytes); + mini_fo_encode_block((char *) page_address(page) + hidden_from, (char*) page_address(hidden_page) + hidden_from, hidden_bytes, inode, inode->i_sb, page->index); + /* if encode_block could fail, then goto unlock and return error */ + + /* call lower commit_write */ + err = hidden_inode->i_mapping->a_ops->commit_write(hidden_file, + hidden_page, + hidden_from, + hidden_to); + + if (err < 0) + goto out_unlock; + + err = bytes; /* convert error to no. of bytes */ + + inode->i_blocks = hidden_inode->i_blocks; + /* we may have to update i_size */ + pos = (page->index << PAGE_CACHE_SHIFT) + to; + if (pos > inode->i_size) + inode->i_size = pos; + + /* + * update mtime and ctime of lower level file system + * mini_fo' mtime and ctime are updated by generic_file_write + */ + hidden_inode->i_mtime = hidden_inode->i_ctime = CURRENT_TIME; + + mark_inode_dirty_sync(inode); + + out_unlock: + UnlockPage(hidden_page); + page_cache_release(hidden_page); + kunmap(page); /* kmap was done in prepare_write */ + out: + /* we must set our page as up-to-date */ + if (err < 0) + ClearPageUptodate(page); + else + SetPageUptodate(page); + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&hidden_inode->i_mutex); +#else + up(&hidden_inode->i_sem); +#endif + print_exit_status(err); + return err; /* assume all is ok */ +} + + +STATIC int +mini_fo_bmap(struct address_space *mapping, long block) +{ + int err = 0; + inode_t *inode; + inode_t *hidden_inode; + + print_entry_location(); + + inode = (inode_t *) mapping->host; + hidden_inode = itohi(inode); + + if (hidden_inode->i_mapping->a_ops->bmap) + err = hidden_inode->i_mapping->a_ops->bmap(hidden_inode->i_mapping, block); + print_exit_location(); + return err; +} + + +/* + * This function is copied verbatim from mm/filemap.c. + * XXX: It should be simply moved to some header file instead -- bug Al about it! + */ +static inline int sync_page(struct page *page) +{ + struct address_space *mapping = page->mapping; + + if (mapping && mapping->a_ops && mapping->a_ops->sync_page) + return mapping->a_ops->sync_page(page); + return 0; +} + + +/* + * XXX: we may not need this function if not FIST_FILTER_DATA. + * FIXME: for FIST_FILTER_SCA, get all lower pages and sync them each. + */ +STATIC int +mini_fo_sync_page(page_t *page) +{ + int err = 0; + inode_t *inode; + inode_t *hidden_inode; + page_t *hidden_page; + + print_entry_location(); + + inode = page->mapping->host; /* CPW: Moved below print_entry_location */ + hidden_inode = itohi(inode); + + /* find lower page (returns a locked page) */ + hidden_page = grab_cache_page(hidden_inode->i_mapping, page->index); + if (!hidden_page) + goto out; + + err = sync_page(hidden_page); + + UnlockPage(hidden_page); /* b/c grab_cache_page locked it */ + page_cache_release(hidden_page); /* b/c grab_cache_page increased refcnt */ + + out: + print_exit_status(err); + return err; +} Index: linux-2.6.25.1/fs/mini_fo/README =================================================================== --- /dev/null +++ linux-2.6.25.1/fs/mini_fo/README @@ -0,0 +1,163 @@ +README for the mini_fo overlay file system +========================================= + + +WHAT IS MINI_FO? +---------------- + +mini_fo is a virtual kernel file system that can make read-only +file systems writable. This is done by redirecting modifying operations +to a writeable location called "storage directory", and leaving the +original data in the "base directory" untouched. When reading, the +file system merges the modifed and original data so that only the +newest versions will appear. This occurs transparently to the user, +who can access the data like on any other read-write file system. + +Base and storage directories may be located on the same or on +different partitions and may be of different file system types. While +the storage directory obviously needs to be writable, the base may or +may not be writable, what doesn't matter as it will no be modified +anyway. + + +WHAT IS GOOD FOR? +----------------- + +The primary purpose of the mini_fo file system is to allow easy +software updates to embedded systems, that often store their root +file system in a read-only flash file system, but there are many +more as for example sandboxing, or for allowing live-cds to +permanently store information. + + +BUILDING +-------- +This should be simple. Adjust the Makefile to point to the correct +kernel headers you want to build the module for. Then: + + # make + +should build "mini_fo.o" for a 2.4 kernel or "mini_fo.ko" for a 2.6 +kernel. + +If you are building the module for you current kernel, you can install +the module (as root): + + # make install + +or uninstall with + + # make uninstall + + +USING THE FILE SYSTEM +-------------------- + +the general mount syntax is: + + mount -t mini_fo -o base=,sto=\ + + +Example: + +You have mounted a cdrom to /mnt/cdrom and want to modifiy some files +on it: + +load the module (as root) + + # insmod mini_fo.o for a 2.4 kernel or + + # insmod mini_fo.ko for a 2.6 kernel + + +create a storage dir in tmp and a mountpoint for mini_fo: + + # mkdir /tmp/sto + # mkdir /mnt/mini_fo + +and mount the mini_fo file system: + + # mount -t mini_fo -o base=/mnt/cdrom,sto=/tmp/sto /mnt/cdrom /mnt/mini_fo + + +Now the data stored on the cd can be accessed via the mini_fo +mountpoint just like any read-write file system, files can be modified +and deleted, new ones can be created and so on. When done unmount the +file system: + + # unmount /mnt/mini_fo + +Note that if the file system is mounted again using the same storage +file system, of course it will appear in the modified state again. If +you remount it using an new empty storage directory, it will be +unmodified. Therefore by executing: + + # cd /tmp/sto + # rm -rf * + +you can nuke all the changes you made to the original file system. But + remember NEVER do this while the mini_fo file system is mounted! + + +Alternatively you can use the mini_fo-overlay bash script, that +simplifies managing mini_fo mounts. See TOOLS Section. + + +TOOLS +----- + +mini_fo-merge (experimental): + +This is a bash script that will merge changes contained in the storage +directory back to the base directory. This allows mini_fo to function +as a cache file system by overlaying a slow (network, ...) file system +and using a fast (ramdisk, ...) as storage. When done, changes can be +merged back to the (slow) base with mini_fo-merge. See "mini_fo-merge +-h" for details. + +It can be usefull for merging changes back after a successfull test +(patches, software updates...) + + +mini_fo-overlay: + +This bash script simplifies managing one or more mini_fo mounts. For +overlaying a directory called "basedir1", you can just call: + + # mini_fo-overlay basedir1 + +This will mount mini_fo with "basedir1" as base, "/tmp/sto-basedir1/" +as storage to "/mnt/mini_fo-basedir1/". It has more options though, +type "mini_fo-overlay -h" for details. + + +DOCUMENTATION, REPORTING BUGS, GETTING HELP +------------------------------------------- + +Please visit the mini_fo project page at: + +http://www.denx.de/twiki/bin/view/Know/MiniFOHome + + +WARNINGS +-------- + +Never modify the base or the storage directorys while the mini_fo +file system is mounted, or you might crash you system. Simply accessing +and reading should not cause any trouble. + +Exporting a mini_fo mount point via NFS has not been tested, and may +or may not work. + +Check the RELEASE_NOTES for details on bugs and features. + + + +Copyright (C) 2004, 2005 Markus Klotzbuecher + +This program is free software; you can redistribute it and/or +modify it under the terms of the GNU General Public License +as published by the Free Software Foundation; either version +2 of the License, or (at your option) any later version. + + Index: linux-2.6.25.1/fs/mini_fo/RELEASE_NOTES =================================================================== --- /dev/null +++ linux-2.6.25.1/fs/mini_fo/RELEASE_NOTES @@ -0,0 +1,111 @@ +Release: mini_fo-0.6.1 (v0-6-1) +Date: 21.09.2005 + + +Changes: +-------- +v0-6-1: + +- bugfixes (see ChangeLog) + +- two helper scripts "mini_fo_merge" and "mini_fo_overlay" (see + README for details). + +v0-6-0: + +- Support for 2.4 and 2.6 (see Makefile) + +- Partial hard link support (creating works as expected, but already + existing links in the base file system will be treated as if they + were individual files). + +- Various bugfixes and cleanups. + + +v0-6-0-pre1: + +- This is mini_fo-0-6-0-pre1! This release is a complete rewrite of + many vital mini_fo parts such as the old whiteout list code which + has been replaced by the new META subsystem. + +- Light weight directory renaming implemented. This means if a + directory is renamed via the mini_fo filesystem this will no longer + result in a complete copy in storage, instead only one empty + directory will be created. All base filed contained in the original + directory stay there until modified. + +- Special files (creating, renaming, deleting etc.) now working. + +- Many bugfixes and cleanup, mini_fo is now a lot more stable. + + +v0-5-10: + +- Final release of the 0-5-* versions. Next will be a complete rewrite + of many features. This release contains several bugfixes related to + directory renaming. + + +v0-5-10-pre6: + +- Lots of cleanup and several bugfixes related to directory deleting + +- Directory renaming suddenly works, what is most likely due to the + fact tha that "mv" is smart: if the classic rename doesn't work it + will assume that source and target file are on different fs and will + copy the directory and try to remove the source directory. Until + directory removing wasn't implemented, it would fail to do this and + rollback. + So, directory renaming works for now, but it doesn't yet do what you + would expect from a overlay fs, so use with care. + + +v0-5-10-pre5: + +- implemented directory deleting +- made parsing of mount options more stable +- New format of mount options! (See README) +- I can't reproduce the unknown panic with 2.4.25 anymore, so I'll + happily assume it never existed! + + +Implemented features: +--------------------- + +- creating hard links (see BUGS on already existing hard links) +- lightweight directory renaming +- renaming device files, pipes, sockets, etc. +- creating, renaming, deleting of special files +- deleting directorys +- general directory reading (simple "ls" ) +- creating files in existing directorys +- creating directorys +- renaming files. +- reading and writing files (involves opening) +- appending to files (creates copy in storage) +- deleting files +- llseek works too, what allows editors to work +- persistency (a deleted file stay deleted over remounts) +- use of symbolic links +- creating of device files + + +Not (yet) implemented features: +------------------------------- + +- full hard link support. + + + +BUGS: +----- + +Hard links in the base file system will be treated as individual +files, not as links to one inode. + +The main problem with hard links isn't allowing to create them, but +their pure existence. If you modify a base hard link, the changes made +will only show up on this link, the other link will remain in the +original state. I hope to fix this someday. Please note that this does +not effect the special hard links '.' and '..', that are handled +seperately by the lower fs. Index: linux-2.6.25.1/fs/mini_fo/state.c =================================================================== --- /dev/null +++ linux-2.6.25.1/fs/mini_fo/state.c @@ -0,0 +1,620 @@ +/* + * Copyright (C) 2005 Markus Klotzbuecher + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ + +#ifdef HAVE_CONFIG_H +# include +#endif /* HAVE_CONFIG_H */ + +#include "fist.h" +#include "mini_fo.h" + + +/* create the storage file, setup new states */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) +int create_sto_reg_file(dentry_t *dentry, int mode, struct nameidata *nd) +#else +int create_sto_reg_file(dentry_t *dentry, int mode) +#endif +{ + int err = 0; + inode_t *dir; + dentry_t *hidden_sto_dentry; + dentry_t *hidden_sto_dir_dentry; + + if(exists_in_storage(dentry)) { + printk(KERN_CRIT "mini_fo: create_sto_file: wrong type or state.\n"); + err = -EINVAL; + goto out; + } + err = get_neg_sto_dentry(dentry); + + if (err) { + printk(KERN_CRIT "mini_fo: create_sto_file: ERROR getting neg. sto dentry.\n"); + goto out; + } + + dir = dentry->d_parent->d_inode; + hidden_sto_dentry = dtohd2(dentry); + + /* lock parent */ + hidden_sto_dir_dentry = dget(hidden_sto_dentry->d_parent); + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + down(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + + err = PTR_ERR(hidden_sto_dir_dentry); + if (IS_ERR(hidden_sto_dir_dentry)) + goto out; + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) + err = vfs_create(hidden_sto_dir_dentry->d_inode, + hidden_sto_dentry, + mode, nd); +#else + err = vfs_create(hidden_sto_dir_dentry->d_inode, + hidden_sto_dentry, + mode); +#endif + if(err) { + printk(KERN_CRIT "mini_fo: create_sto_file: ERROR creating sto file.\n"); + goto out_lock; + } + + if(!dtohd2(dentry)->d_inode) { + printk(KERN_CRIT "mini_fo: create_sto_file: ERROR creating sto file [2].\n"); + err = -EINVAL; + goto out_lock; + } + + /* interpose the new inode */ + if(dtost(dentry) == DELETED) { + dtost(dentry) = DEL_REWRITTEN; + err = mini_fo_tri_interpose(NULL, hidden_sto_dentry, dentry, dir->i_sb, 0); + if(err) + goto out_lock; + } + else if(dtost(dentry) == NON_EXISTANT) { + dtost(dentry) = CREATED; + err = mini_fo_tri_interpose(dtohd(dentry), hidden_sto_dentry, dentry, dir->i_sb, 0); + if(err) + goto out_lock; + } + else if(dtost(dentry) == UNMODIFIED) { + dtost(dentry) = MODIFIED; + /* interpose on new inode */ + if(itohi2(dentry->d_inode) != NULL) { + printk(KERN_CRIT "mini_fo: create_sto_file: invalid inode detected.\n"); + err = -EINVAL; + goto out_lock; + } + itohi2(dentry->d_inode) = igrab(dtohd2(dentry)->d_inode); + } + fist_copy_attr_timesizes(dentry->d_parent->d_inode, + hidden_sto_dir_dentry->d_inode); + + out_lock: +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + up(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + dput(hidden_sto_dir_dentry); + out: + return err; +} + +/* create the sto dir, setup states */ +int create_sto_dir(dentry_t *dentry, int mode) +{ + int err = 0; + inode_t *dir; + dentry_t *hidden_sto_dentry; + dentry_t *hidden_sto_dir_dentry; + + /* had to take the "!S_ISDIR(mode))" check out, because it failed */ + if(exists_in_storage(dentry)) { + printk(KERN_CRIT "mini_fo: create_sto_dir: wrong type or state.\\ +n"); + err = -EINVAL; + goto out; + } + + err = get_neg_sto_dentry(dentry); + if(err) { + err = -EINVAL; + goto out; + } + + dir = dentry->d_parent->d_inode; + hidden_sto_dentry = dtohd2(dentry); + + /* was: hidden_sto_dir_dentry = lock_parent(hidden_sto_dentry); */ + hidden_sto_dir_dentry = dget(hidden_sto_dentry->d_parent); + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + down(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + + err = PTR_ERR(hidden_sto_dir_dentry); + if (IS_ERR(hidden_sto_dir_dentry)) + goto out; + + err = vfs_mkdir(hidden_sto_dir_dentry->d_inode, + hidden_sto_dentry, + mode); + if(err) { + printk(KERN_CRIT "mini_fo: create_sto_dir: ERROR creating sto dir.\n"); + goto out_lock; + } + + if(!dtohd2(dentry)->d_inode) { + printk(KERN_CRIT "mini_fo: create_sto_dir: ERROR creating sto dir [2].\n"); + err = -EINVAL; + goto out_lock; + } + + /* interpose the new inode */ + if(dtost(dentry) == DELETED) { + dtost(dentry) = DEL_REWRITTEN; + err = mini_fo_tri_interpose(NULL, hidden_sto_dentry, dentry, dir->i_sb, 0); + if(err) + goto out_lock; + } + else if(dtopd(dentry)->state == NON_EXISTANT) { + dtopd(dentry)->state = CREATED; + err = mini_fo_tri_interpose(dtohd(dentry), hidden_sto_dentry, dentry, dir->i_sb, 0); + if(err) + goto out_lock; + } + else if(dtopd(dentry)->state == UNMODIFIED) { + dtopd(dentry)->state = MODIFIED; + /* interpose on new inode */ + if(itohi2(dentry->d_inode) != NULL) { + printk(KERN_CRIT "mini_fo: create_sto_dir: ERROR, invalid inode detected.\n"); + err = -EINVAL; + goto out_lock; + } + itohi2(dentry->d_inode) = igrab(dtohd2(dentry)->d_inode); + } + + fist_copy_attr_timesizes(dir, hidden_sto_dir_dentry->d_inode); + + /* initalize the wol list */ + itopd(dentry->d_inode)->deleted_list_size = -1; + itopd(dentry->d_inode)->renamed_list_size = -1; + meta_build_lists(dentry); + + + out_lock: + /* was: unlock_dir(hidden_sto_dir_dentry); */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + up(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + dput(hidden_sto_dir_dentry); + out: + return err; +} + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) +int create_sto_nod(dentry_t *dentry, int mode, dev_t dev) +#else +int create_sto_nod(dentry_t *dentry, int mode, int dev) +#endif +{ + int err = 0; + inode_t *dir; + dentry_t *hidden_sto_dentry; + dentry_t *hidden_sto_dir_dentry; + + if(exists_in_storage(dentry)) { + err = -EEXIST; + goto out; + } + err = get_neg_sto_dentry(dentry); + + if (err) { + printk(KERN_CRIT "mini_fo: create_sto_nod: ERROR getting neg. sto dentry.\n"); + goto out; + } + + dir = dentry->d_parent->d_inode; + hidden_sto_dentry = dtohd2(dentry); + + /* lock parent */ + hidden_sto_dir_dentry = dget(hidden_sto_dentry->d_parent); + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + down(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + + err = PTR_ERR(hidden_sto_dir_dentry); + if (IS_ERR(hidden_sto_dir_dentry)) + goto out; + + err = vfs_mknod(hidden_sto_dir_dentry->d_inode, hidden_sto_dentry, mode, dev); + if(err) + goto out_lock; + + if(!dtohd2(dentry)->d_inode) { + printk(KERN_CRIT "mini_fo: create_sto_nod: creating storage inode failed [1].\n"); + err = -EINVAL; /* return something indicating failure */ + goto out_lock; + } + + /* interpose the new inode */ + if(dtost(dentry) == DELETED) { + dtost(dentry) = DEL_REWRITTEN; + err = mini_fo_tri_interpose(NULL, hidden_sto_dentry, dentry, dir->i_sb, 0); + if(err) + goto out_lock; + } + else if(dtost(dentry) == NON_EXISTANT) { + dtost(dentry) = CREATED; + err = mini_fo_tri_interpose(dtohd(dentry), hidden_sto_dentry, dentry, dir->i_sb, 0); + if(err) + goto out_lock; + } + else if(dtost(dentry) == UNMODIFIED) { + dtost(dentry) = MODIFIED; + /* interpose on new inode */ + if(itohi2(dentry->d_inode) != NULL) { + printk(KERN_CRIT "mini_fo: create_sto_nod: error, invalid inode detected.\n"); + err = -EINVAL; + goto out_lock; + } + itohi2(dentry->d_inode) = igrab(dtohd2(dentry)->d_inode); + } + + fist_copy_attr_timesizes(dir, hidden_sto_dir_dentry->d_inode); + + out_lock: +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + up(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + dput(hidden_sto_dir_dentry); + out: + return err; +} + + +/* unimplemented (and possibly not usefull): + + nondir-del_to_del_rew + nondir-non_exist_to_creat + + dir-unmod_to_del + dir-mod_to_del + dir-creat_to_del + dir-del_rew_to_del + dir-del_to_del_rew + dir-non_exist_to_creat +*/ + + +/* bring a file of any type from state UNMODIFIED to MODIFIED */ +int nondir_unmod_to_mod(dentry_t *dentry, int cp_flag) +{ + int err = 0; + struct vfsmount *tgt_mnt; + struct vfsmount *src_mnt; + dentry_t *tgt_dentry; + dentry_t *src_dentry; + dentry_t *hidden_sto_dentry; + dentry_t *hidden_sto_dir_dentry; + + check_mini_fo_dentry(dentry); + + if((dtost(dentry) != UNMODIFIED) || + S_ISDIR(dentry->d_inode->i_mode)) { + printk(KERN_CRIT "mini_fo: nondir_unmod_to_mod: \ + wrong type or state.\n"); + err = -1; + goto out; + } + err = get_neg_sto_dentry(dentry); + + if (err) { + printk(KERN_CRIT "mini_fo: nondir_unmod_to_mod: \ + ERROR getting neg. sto dentry.\n"); + goto out; + } + + /* create sto file */ + hidden_sto_dentry = dtohd2(dentry); + + /* lock parent */ + hidden_sto_dir_dentry = dget(hidden_sto_dentry->d_parent); + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + down(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + + err = PTR_ERR(hidden_sto_dir_dentry); + if (IS_ERR(hidden_sto_dir_dentry)) + goto out; + + /* handle different types of nondirs */ + if(S_ISCHR(dentry->d_inode->i_mode) || + S_ISBLK(dentry->d_inode->i_mode)) { + err = vfs_mknod(hidden_sto_dir_dentry->d_inode, + hidden_sto_dentry, + dtohd(dentry)->d_inode->i_mode, + dtohd(dentry)->d_inode->i_rdev); + } + + else if(S_ISREG(dentry->d_inode->i_mode)) { + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) + err = vfs_create(hidden_sto_dir_dentry->d_inode, + hidden_sto_dentry, + dtohd(dentry)->d_inode->i_mode, NULL); +#else + err = vfs_create(hidden_sto_dir_dentry->d_inode, + hidden_sto_dentry, + dtohd(dentry)->d_inode->i_mode); +#endif + } + if(err) { + printk(KERN_CRIT "mini_fo: nondir_unmod_to_mod: \ + ERROR creating sto file.\n"); + goto out_lock; + } + + /* interpose on new inode */ + if(itohi2(dentry->d_inode) != NULL) { + printk(KERN_CRIT "mini_fo: nondir_unmod_to_mod: \ + ERROR, invalid inode detected.\n"); + err = -EINVAL; + goto out_lock; + } + + itohi2(dentry->d_inode) = igrab(dtohd2(dentry)->d_inode); + + fist_copy_attr_timesizes(dentry->d_parent->d_inode, + hidden_sto_dir_dentry->d_inode); + dtost(dentry) = MODIFIED; + + /* copy contents if regular file and cp_flag = 1 */ + if((cp_flag == 1) && S_ISREG(dentry->d_inode->i_mode)) { + + /* unlock first */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + up(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + + dput(hidden_sto_dir_dentry); + + tgt_dentry = dtohd2(dentry); + tgt_mnt = stopd(dentry->d_inode->i_sb)->hidden_mnt2; + src_dentry = dtohd(dentry); + src_mnt = stopd(dentry->d_inode->i_sb)->hidden_mnt; + + err = mini_fo_cp_cont(tgt_dentry, tgt_mnt, + src_dentry, src_mnt); + if(err) { + printk(KERN_CRIT "mini_fo: nondir_unmod_to_mod: \ + ERROR copying contents.\n"); + } + goto out; + } + + out_lock: +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + up(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + dput(hidden_sto_dir_dentry); + out: + return err; +} + +/* this function is currently identical to nondir_creat_to_del */ +int nondir_del_rew_to_del(dentry_t *dentry) +{ + return nondir_creat_to_del(dentry); +} + +int nondir_creat_to_del(dentry_t *dentry) +{ + int err = 0; + + inode_t *hidden_sto_dir_inode; + dentry_t *hidden_sto_dir_dentry; + dentry_t *hidden_sto_dentry; + + check_mini_fo_dentry(dentry); + + /* for now this function serves for both state DEL_REWRITTEN and + * CREATED */ + if(!(dtost(dentry) == CREATED || (dtost(dentry) == DEL_REWRITTEN)) || + S_ISDIR(dentry->d_inode->i_mode)) { + printk(KERN_CRIT "mini_fo: nondir_mod_to_del/del_rew_to_del: \ + wrong type or state.\n"); + err = -1; + goto out; + } + + hidden_sto_dir_inode = itohi2(dentry->d_parent->d_inode); + hidden_sto_dentry = dtohd2(dentry); + + /* was: hidden_sto_dir_dentry = lock_parent(hidden_sto_dentry);*/ + hidden_sto_dir_dentry = dget(hidden_sto_dentry->d_parent); + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + down(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + + /* avoid destroying the hidden inode if the file is in use */ + dget(hidden_sto_dentry); + err = vfs_unlink(hidden_sto_dir_inode, hidden_sto_dentry); + dput(hidden_sto_dentry); + if(!err) + d_delete(hidden_sto_dentry); + + /* propagate number of hard-links */ + dentry->d_inode->i_nlink = itohi2(dentry->d_inode)->i_nlink; + + dtost(dentry) = NON_EXISTANT; + + /* was: unlock_dir(hidden_sto_dir_dentry); */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + up(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + dput(hidden_sto_dir_dentry); + + out: + return err; +} + +int nondir_mod_to_del(dentry_t *dentry) +{ + int err; + dentry_t *hidden_sto_dentry; + inode_t *hidden_sto_dir_inode; + dentry_t *hidden_sto_dir_dentry; + + check_mini_fo_dentry(dentry); + + if(dtost(dentry) != MODIFIED || + S_ISDIR(dentry->d_inode->i_mode)) { + printk(KERN_CRIT "mini_fo: nondir_mod_to_del: \ + wrong type or state.\n"); + err = -1; + goto out; + } + + hidden_sto_dir_inode = itohi2(dentry->d_parent->d_inode); + hidden_sto_dentry = dtohd2(dentry); + + /* was hidden_sto_dir_dentry = lock_parent(hidden_sto_dentry); */ + hidden_sto_dir_dentry = dget(hidden_sto_dentry->d_parent); + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_lock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + down(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + + /* avoid destroying the hidden inode if the file is in use */ + dget(hidden_sto_dentry); + err = vfs_unlink(hidden_sto_dir_inode, hidden_sto_dentry); + dput(hidden_sto_dentry); + if(!err) + d_delete(hidden_sto_dentry); + + /* propagate number of hard-links */ + dentry->d_inode->i_nlink = itohi2(dentry->d_inode)->i_nlink; + + /* dput base dentry, this will relase the inode and free the + * dentry, as we will never need it again. */ + dput(dtohd(dentry)); + dtohd(dentry) = NULL; + dtost(dentry) = DELETED; + + /* add deleted file to META-file */ + meta_add_d_entry(dentry->d_parent, + dentry->d_name.name, + dentry->d_name.len); + + /* was: unlock_dir(hidden_sto_dir_dentry); */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + mutex_unlock(&hidden_sto_dir_dentry->d_inode->i_mutex); +#else + up(&hidden_sto_dir_dentry->d_inode->i_sem); +#endif + dput(hidden_sto_dir_dentry); + + out: + return err; +} + +int nondir_unmod_to_del(dentry_t *dentry) +{ + int err = 0; + + check_mini_fo_dentry(dentry); + + if(dtost(dentry) != UNMODIFIED || + S_ISDIR(dentry->d_inode->i_mode)) { + printk(KERN_CRIT "mini_fo: nondir_unmod_to_del: \ + wrong type or state.\n"); + err = -1; + goto out; + } + + /* next we have to get a negative dentry for the storage file */ + err = get_neg_sto_dentry(dentry); + + if(err) + goto out; + + /* add deleted file to META lists */ + err = meta_add_d_entry(dentry->d_parent, + dentry->d_name.name, + dentry->d_name.len); + + if(err) + goto out; + + /* dput base dentry, this will relase the inode and free the + * dentry, as we will never need it again. */ + dput(dtohd(dentry)); + dtohd(dentry) = NULL; + dtost(dentry) = DELETED; + + out: + return err; +} + +/* bring a dir from state UNMODIFIED to MODIFIED */ +int dir_unmod_to_mod(dentry_t *dentry) +{ + int err; + + check_mini_fo_dentry(dentry); + + if(dtost(dentry) != UNMODIFIED || + !S_ISDIR(dentry->d_inode->i_mode)) { + printk(KERN_CRIT "mini_fo: dir_unmod_to_mod: \ + wrong type or state.\n"); + err = -1; + goto out; + } + + /* this creates our dir incl. sto. structure */ + err = build_sto_structure(dentry->d_parent, dentry); + if(err) { + printk(KERN_CRIT "mini_fo: dir_unmod_to_mod: \ + build_sto_structure failed.\n"); + goto out; + } + out: + return err; +} + Index: linux-2.6.25.1/fs/mini_fo/super.c =================================================================== --- /dev/null +++ linux-2.6.25.1/fs/mini_fo/super.c @@ -0,0 +1,281 @@ +/* + * Copyright (c) 1997-2003 Erez Zadok + * Copyright (c) 2001-2003 Stony Brook University + * + * For specific licensing information, see the COPYING file distributed with + * this package, or get one from ftp://ftp.filesystems.org/pub/fist/COPYING. + * + * This Copyright notice must be kept intact and distributed with all + * fistgen sources INCLUDING sources generated by fistgen. + */ +/* + * Copyright (C) 2004, 2005 Markus Klotzbuecher + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ + +/* + * $Id$ + */ + +#ifdef HAVE_CONFIG_H +# include +#endif + +#include "fist.h" +#include "mini_fo.h" + + +STATIC void +mini_fo_read_inode(inode_t *inode) +{ + static struct address_space_operations mini_fo_empty_aops; + + __itopd(inode) = kmalloc(sizeof(struct mini_fo_inode_info), GFP_KERNEL); + if (!itopd(inode)) { + printk("<0>%s:%s:%d: No kernel memory!\n", __FILE__, __FUNCTION__, __LINE__); + ASSERT(NULL); + } + itohi(inode) = NULL; + itohi2(inode) = NULL; + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) + inode->i_version++; +#else + inode->i_version = ++event; /* increment inode version */ +#endif + inode->i_op = &mini_fo_main_iops; + inode->i_fop = &mini_fo_main_fops; +#if 0 + /* + * XXX: To export a file system via NFS, it has to have the + * FS_REQUIRES_DEV flag, so turn it on. But should we inherit it from + * the lower file system, or can we allow our file system to be exported + * even if the lower one cannot be natively exported. + */ + inode->i_sb->s_type->fs_flags |= FS_REQUIRES_DEV; + /* + * OK, the above was a hack, which is now turned off because it may + * cause a panic/oops on some systems. The correct way to export a + * "nodev" filesystem is via using nfs-utils > 1.0 and the "fsid=" export + * parameter, which requires 2.4.20 or later. + */ +#endif + /* I don't think ->a_ops is ever allowed to be NULL */ + inode->i_mapping->a_ops = &mini_fo_empty_aops; +} + + +#if defined(FIST_DEBUG) || defined(FIST_FILTER_SCA) +/* + * No need to call write_inode() on the lower inode, as it + * will have been marked 'dirty' anyway. But we might need + * to write some of our own stuff to disk. + */ +STATIC void +mini_fo_write_inode(inode_t *inode, int sync) +{ + print_entry_location(); + print_exit_location(); +} +#endif /* defined(FIST_DEBUG) || defined(FIST_FILTER_SCA) */ + + +STATIC void +mini_fo_put_inode(inode_t *inode) +{ + /* + * This is really funky stuff: + * Basically, if i_count == 1, iput will then decrement it and this inode will be destroyed. + * It is currently holding a reference to the hidden inode. + * Therefore, it needs to release that reference by calling iput on the hidden inode. + * iput() _will_ do it for us (by calling our clear_inode), but _only_ if i_nlink == 0. + * The problem is, NFS keeps i_nlink == 1 for silly_rename'd files. + * So we must for our i_nlink to 0 here to trick iput() into calling our clear_inode. + */ + if (atomic_read(&inode->i_count) == 1) + inode->i_nlink = 0; +} + + +#if defined(FIST_DEBUG) || defined(FIST_FILTER_SCA) +/* + * we now define delete_inode, because there are two VFS paths that may + * destroy an inode: one of them calls clear inode before doing everything + * else that's needed, and the other is fine. This way we truncate the inode + * size (and its pages) and then clear our own inode, which will do an iput + * on our and the lower inode. + */ +STATIC void +mini_fo_delete_inode(inode_t *inode) +{ + print_entry_location(); + + fist_checkinode(inode, "mini_fo_delete_inode IN"); + inode->i_size = 0; /* every f/s seems to do that */ + clear_inode(inode); + + print_exit_location(); +} +#endif /* defined(FIST_DEBUG) || defined(FIST_FILTER_SCA) */ + + +/* final actions when unmounting a file system */ +STATIC void +mini_fo_put_super(super_block_t *sb) +{ + if (stopd(sb)) { + mntput(stopd(sb)->hidden_mnt); + mntput(stopd(sb)->hidden_mnt2); + + /* mk: no! dput(stopd(sb)->base_dir_dentry); + dput(stopd(sb)->storage_dir_dentry); */ + + kfree(stopd(sb)); + __stopd(sb) = NULL; + } +} + + +#ifdef NOT_NEEDED +/* + * This is called in do_umount before put_super. + * The superblock lock is not held yet. + * We probably do not need to define this or call write_super + * on the hidden_sb, because sync_supers() will get to hidden_sb + * sooner or later. But it is also called from file_fsync()... + */ +STATIC void +mini_fo_write_super(super_block_t *sb) +{ + return; +} +#endif /* NOT_NEEDED */ + + +STATIC int +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,18) +mini_fo_statfs(struct dentry *d, struct kstatfs *buf) +#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) +mini_fo_statfs(super_block_t *sb, struct kstatfs *buf) +#else +mini_fo_statfs(super_block_t *sb, struct statfs *buf) +#endif +{ + int err = 0; +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,18) + struct dentry *hidden_d; + + hidden_d = dtohd(d); + err = vfs_statfs(hidden_d, buf); +#else + super_block_t *hidden_sb; + + hidden_sb = stohs(sb); + err = vfs_statfs(hidden_sb, buf); +#endif + + return err; +} + + +/* + * XXX: not implemented. This is not allowed yet. + * Should we call this on the hidden_sb? Probably not. + */ +STATIC int +mini_fo_remount_fs(super_block_t *sb, int *flags, char *data) +{ + //printk(KERN_CRIT "mini_fo_remount_fs: WARNING, this function is umimplemented.\n"); + return -ENOSYS; +} + + +/* + * Called by iput() when the inode reference count reached zero + * and the inode is not hashed anywhere. Used to clear anything + * that needs to be, before the inode is completely destroyed and put + * on the inode free list. + */ +STATIC void +mini_fo_clear_inode(inode_t *inode) +{ + /* + * Decrement a reference to a hidden_inode, which was incremented + * by our read_inode when it was created initially. + */ + + /* release the wol_list */ + if(S_ISDIR(inode->i_mode)) { + __meta_put_lists(inode); + } + + /* mk: fan out fun */ + if(itohi(inode)) + iput(itohi(inode)); + if(itohi2(inode)) + iput(itohi2(inode)); + + // XXX: why this assertion fails? + // because it doesn't like us + // ASSERT((inode->i_state & I_DIRTY) == 0); + kfree(itopd(inode)); + __itopd(inode) = NULL; +} + + +/* + * Called in do_umount() if the MNT_FORCE flag was used and this + * function is defined. See comment in linux/fs/super.c:do_umount(). + * Used only in nfs, to kill any pending RPC tasks, so that subsequent + * code can actually succeed and won't leave tasks that need handling. + * + * PS. I wonder if this is somehow useful to undo damage that was + * left in the kernel after a user level file server (such as amd) + * dies. + */ +STATIC void +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,18) +mini_fo_umount_begin(struct vfsmount *mnt, int flags) +{ + struct vfsmount *hidden_mnt; + + hidden_mnt = stopd(mnt->mnt_sb)->hidden_mnt; + + if (hidden_mnt->mnt_sb->s_op->umount_begin) + hidden_mnt->mnt_sb->s_op->umount_begin(hidden_mnt, flags); + +} +#else +mini_fo_umount_begin(super_block_t *sb) +{ + super_block_t *hidden_sb; + + hidden_sb = stohs(sb); + + if (hidden_sb->s_op->umount_begin) + hidden_sb->s_op->umount_begin(hidden_sb); + +} +#endif + + +struct super_operations mini_fo_sops = +{ + read_inode: mini_fo_read_inode, +#if defined(FIST_DEBUG) || defined(FIST_FILTER_SCA) + write_inode: mini_fo_write_inode, +#endif /* defined(FIST_DEBUG) || defined(FIST_FILTER_SCA) */ + put_inode: mini_fo_put_inode, +#if defined(FIST_DEBUG) || defined(FIST_FILTER_SCA) + delete_inode: mini_fo_delete_inode, +#endif /* defined(FIST_DEBUG) || defined(FIST_FILTER_SCA) */ + put_super: mini_fo_put_super, + statfs: mini_fo_statfs, + remount_fs: mini_fo_remount_fs, + clear_inode: mini_fo_clear_inode, + umount_begin: mini_fo_umount_begin, +};