X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=fs%2Focfs2%2Fnamei.c;h=476684190461edcb13b9692fdcd4b4b74d4f9553;hb=97bf2856c6014879bd04983a3e9dfcdac1e7fe85;hp=f6b77ff1d2bf414fb395c14dd7f8ee1147acb7f2;hpb=76828883507a47dae78837ab5dec5a5b4513c667;p=linux-2.6.git diff --git a/fs/ocfs2/namei.c b/fs/ocfs2/namei.c index f6b77ff1d..476684190 100644 --- a/fs/ocfs2/namei.c +++ b/fs/ocfs2/namei.c @@ -40,6 +40,7 @@ #include #include #include +#include #define MLOG_MASK_PREFIX ML_NAMEI #include @@ -56,6 +57,7 @@ #include "journal.h" #include "namei.h" #include "suballoc.h" +#include "super.h" #include "symlink.h" #include "sysfile.h" #include "uptodate.h" @@ -74,12 +76,12 @@ static int inline ocfs2_search_dirblock(struct buffer_head *bh, unsigned long offset, struct ocfs2_dir_entry **res_dir); -static int ocfs2_delete_entry(struct ocfs2_journal_handle *handle, +static int ocfs2_delete_entry(handle_t *handle, struct inode *dir, struct ocfs2_dir_entry *de_del, struct buffer_head *bh); -static int __ocfs2_add_entry(struct ocfs2_journal_handle *handle, +static int __ocfs2_add_entry(handle_t *handle, struct inode *dir, const char *name, int namelen, struct inode *inode, u64 blkno, @@ -92,43 +94,37 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb, dev_t dev, struct buffer_head **new_fe_bh, struct buffer_head *parent_fe_bh, - struct ocfs2_journal_handle *handle, + handle_t *handle, struct inode **ret_inode, struct ocfs2_alloc_context *inode_ac); static int ocfs2_fill_new_dir(struct ocfs2_super *osb, - struct ocfs2_journal_handle *handle, + handle_t *handle, struct inode *parent, struct inode *inode, struct buffer_head *fe_bh, struct ocfs2_alloc_context *data_ac); -static int ocfs2_double_lock(struct ocfs2_super *osb, - struct ocfs2_journal_handle *handle, - struct buffer_head **bh1, - struct inode *inode1, - struct buffer_head **bh2, - struct inode *inode2); - static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb, - struct ocfs2_journal_handle *handle, + struct inode **ret_orphan_dir, struct inode *inode, char *name, struct buffer_head **de_bh); static int ocfs2_orphan_add(struct ocfs2_super *osb, - struct ocfs2_journal_handle *handle, + handle_t *handle, struct inode *inode, struct ocfs2_dinode *fe, char *name, - struct buffer_head *de_bh); + struct buffer_head *de_bh, + struct inode *orphan_dir_inode); static int ocfs2_create_symlink_data(struct ocfs2_super *osb, - struct ocfs2_journal_handle *handle, + handle_t *handle, struct inode *inode, const char *symname); -static inline int ocfs2_add_entry(struct ocfs2_journal_handle *handle, +static inline int ocfs2_add_entry(handle_t *handle, struct dentry *dentry, struct inode *inode, u64 blkno, struct buffer_head *parent_fe_bh, @@ -161,10 +157,10 @@ static struct dentry *ocfs2_lookup(struct inode *dir, struct dentry *dentry, goto bail; } - mlog(0, "find name %.*s in directory %"MLFu64"\n", dentry->d_name.len, - dentry->d_name.name, OCFS2_I(dir)->ip_blkno); + mlog(0, "find name %.*s in directory %llu\n", dentry->d_name.len, + dentry->d_name.name, (unsigned long long)OCFS2_I(dir)->ip_blkno); - status = ocfs2_meta_lock(dir, NULL, NULL, 0); + status = ocfs2_meta_lock(dir, NULL, 0); if (status < 0) { if (status != -ENOENT) mlog_errno(status); @@ -178,9 +174,10 @@ static struct dentry *ocfs2_lookup(struct inode *dir, struct dentry *dentry, if (status < 0) goto bail_add; - inode = ocfs2_iget(OCFS2_SB(dir->i_sb), blkno); + inode = ocfs2_iget(OCFS2_SB(dir->i_sb), blkno, 0); if (IS_ERR(inode)) { - mlog(ML_ERROR, "Unable to create inode %"MLFu64"\n", blkno); + mlog(ML_ERROR, "Unable to create inode %llu\n", + (unsigned long long)blkno); ret = ERR_PTR(-EACCES); goto bail_unlock; } @@ -197,10 +194,32 @@ static struct dentry *ocfs2_lookup(struct inode *dir, struct dentry *dentry, spin_unlock(&oi->ip_lock); bail_add: - dentry->d_op = &ocfs2_dentry_ops; ret = d_splice_alias(inode, dentry); + if (inode) { + /* + * If d_splice_alias() finds a DCACHE_DISCONNECTED + * dentry, it will d_move() it on top of ourse. The + * return value will indicate this however, so in + * those cases, we switch them around for the locking + * code. + * + * NOTE: This dentry already has ->d_op set from + * ocfs2_get_parent() and ocfs2_get_dentry() + */ + if (ret) + dentry = ret; + + status = ocfs2_dentry_attach_lock(dentry, inode, + OCFS2_I(dir)->ip_blkno); + if (status) { + mlog_errno(status); + ret = ERR_PTR(status); + goto bail_unlock; + } + } + bail_unlock: /* Don't drop the cluster lock until *after* the d_add -- * unlink on another node will message us to remove that @@ -218,7 +237,7 @@ bail: } static int ocfs2_fill_new_dir(struct ocfs2_super *osb, - struct ocfs2_journal_handle *handle, + handle_t *handle, struct inode *parent, struct inode *inode, struct buffer_head *fe_bh, @@ -293,7 +312,7 @@ static int ocfs2_mknod(struct inode *dir, { int status = 0; struct buffer_head *parent_fe_bh = NULL; - struct ocfs2_journal_handle *handle = NULL; + handle_t *handle = NULL; struct ocfs2_super *osb; struct ocfs2_dinode *dirfe; struct buffer_head *new_fe_bh = NULL; @@ -309,24 +328,15 @@ static int ocfs2_mknod(struct inode *dir, /* get our super block */ osb = OCFS2_SB(dir->i_sb); - if (S_ISDIR(mode) && (dir->i_nlink >= OCFS2_LINK_MAX)) { - mlog(ML_ERROR, "inode %"MLFu64" has i_nlink of %u\n", - OCFS2_I(dir)->ip_blkno, dir->i_nlink); - status = -EMLINK; - goto leave; - } - - handle = ocfs2_alloc_handle(osb); - if (handle == NULL) { - status = -ENOMEM; - mlog_errno(status); - goto leave; - } - - status = ocfs2_meta_lock(dir, handle, &parent_fe_bh, 1); + status = ocfs2_meta_lock(dir, &parent_fe_bh, 1); if (status < 0) { if (status != -ENOENT) mlog_errno(status); + return status; + } + + if (S_ISDIR(mode) && (dir->i_nlink >= OCFS2_LINK_MAX)) { + status = -EMLINK; goto leave; } @@ -352,7 +362,7 @@ static int ocfs2_mknod(struct inode *dir, } /* reserve an inode spot */ - status = ocfs2_reserve_new_inode(osb, handle, &inode_ac); + status = ocfs2_reserve_new_inode(osb, &inode_ac); if (status < 0) { if (status != -ENOSPC) mlog_errno(status); @@ -362,7 +372,7 @@ static int ocfs2_mknod(struct inode *dir, /* are we making a directory? If so, reserve a cluster for his * 1st extent. */ if (S_ISDIR(mode)) { - status = ocfs2_reserve_clusters(osb, handle, 1, &data_ac); + status = ocfs2_reserve_clusters(osb, 1, &data_ac); if (status < 0) { if (status != -ENOSPC) mlog_errno(status); @@ -370,7 +380,7 @@ static int ocfs2_mknod(struct inode *dir, } } - handle = ocfs2_start_trans(osb, handle, OCFS2_MKNOD_CREDITS); + handle = ocfs2_start_trans(osb, OCFS2_MKNOD_CREDITS); if (IS_ERR(handle)) { status = PTR_ERR(handle); handle = NULL; @@ -407,7 +417,7 @@ static int ocfs2_mknod(struct inode *dir, mlog_errno(status); goto leave; } - dir->i_nlink++; + inc_nlink(dir); } status = ocfs2_add_entry(handle, dentry, inode, @@ -418,13 +428,22 @@ static int ocfs2_mknod(struct inode *dir, goto leave; } + status = ocfs2_dentry_attach_lock(dentry, inode, + OCFS2_I(dir)->ip_blkno); + if (status) { + mlog_errno(status); + goto leave; + } + insert_inode_hash(inode); dentry->d_op = &ocfs2_dentry_ops; d_instantiate(dentry, inode); status = 0; leave: if (handle) - ocfs2_commit_trans(handle); + ocfs2_commit_trans(osb, handle); + + ocfs2_meta_unlock(dir, 1); if (status == -ENOSPC) mlog(0, "Disk is full\n"); @@ -458,7 +477,7 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb, dev_t dev, struct buffer_head **new_fe_bh, struct buffer_head *parent_fe_bh, - struct ocfs2_journal_handle *handle, + handle_t *handle, struct inode **ret_inode, struct ocfs2_alloc_context *inode_ac) { @@ -468,6 +487,9 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb, u64 fe_blkno = 0; u16 suballoc_bit; struct inode *inode = NULL; + uid_t uid; + gid_t gid; + tag_t tag; mlog_entry("(0x%p, 0x%p, %d, %lu, '%.*s')\n", dir, dentry, mode, (unsigned long)dev, dentry->d_name.len, @@ -527,13 +549,19 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb, fe->i_blkno = cpu_to_le64(fe_blkno); fe->i_suballoc_bit = cpu_to_le16(suballoc_bit); fe->i_suballoc_slot = cpu_to_le16(osb->slot_num); - fe->i_uid = cpu_to_le32(current->fsuid); + + tag = dx_current_fstag(osb->sb); + uid = current->fsuid; if (dir->i_mode & S_ISGID) { - fe->i_gid = cpu_to_le32(dir->i_gid); + gid = dir->i_gid; if (S_ISDIR(mode)) mode |= S_ISGID; } else - fe->i_gid = cpu_to_le32(current->fsgid); + gid = current->fsgid; + + fe->i_uid = cpu_to_le32(TAGINO_UID(DX_TAG(inode), uid, tag)); + fe->i_gid = cpu_to_le32(TAGINO_GID(DX_TAG(inode), gid, tag)); + inode->i_tag = tag; fe->i_mode = cpu_to_le16(mode); if (S_ISCHR(mode) || S_ISBLK(mode)) fe->id1.dev1.i_rdev = cpu_to_le64(huge_encode_dev(dev)); @@ -562,16 +590,18 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb, if (ocfs2_populate_inode(inode, fe, 1) < 0) { mlog(ML_ERROR, "populate inode failed! bh->b_blocknr=%llu, " - "i_blkno=%"MLFu64", i_ino=%lu\n", + "i_blkno=%llu, i_ino=%lu\n", (unsigned long long) (*new_fe_bh)->b_blocknr, - fe->i_blkno, inode->i_ino); + (unsigned long long)fe->i_blkno, inode->i_ino); BUG(); } ocfs2_inode_set_new(osb, inode); - status = ocfs2_create_new_inode_locks(inode); - if (status < 0) - mlog_errno(status); + if (!ocfs2_mount_local(osb)) { + status = ocfs2_create_new_inode_locks(inode); + if (status < 0) + mlog_errno(status); + } status = 0; /* error in ocfs2_create_new_inode_locks is not * critical */ @@ -624,7 +654,7 @@ static int ocfs2_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry) { - struct ocfs2_journal_handle *handle = NULL; + handle_t *handle; struct inode *inode = old_dentry->d_inode; int err; struct buffer_head *fe_bh = NULL; @@ -637,71 +667,63 @@ static int ocfs2_link(struct dentry *old_dentry, old_dentry->d_name.len, old_dentry->d_name.name, dentry->d_name.len, dentry->d_name.name); - if (S_ISDIR(inode->i_mode)) { - err = -EPERM; - goto bail; - } - - if (inode->i_nlink >= OCFS2_LINK_MAX) { - err = -EMLINK; - goto bail; - } - - handle = ocfs2_alloc_handle(osb); - if (handle == NULL) { - err = -ENOMEM; - goto bail; - } + if (S_ISDIR(inode->i_mode)) + return -EPERM; - err = ocfs2_meta_lock(dir, handle, &parent_fe_bh, 1); + err = ocfs2_meta_lock(dir, &parent_fe_bh, 1); if (err < 0) { if (err != -ENOENT) mlog_errno(err); - goto bail; + return err; + } + + if (!dir->i_nlink) { + err = -ENOENT; + goto out; } err = ocfs2_check_dir_for_entry(dir, dentry->d_name.name, dentry->d_name.len); if (err) - goto bail; + goto out; err = ocfs2_prepare_dir_for_insert(osb, dir, parent_fe_bh, dentry->d_name.name, dentry->d_name.len, &de_bh); if (err < 0) { mlog_errno(err); - goto bail; + goto out; } - err = ocfs2_meta_lock(inode, handle, &fe_bh, 1); + err = ocfs2_meta_lock(inode, &fe_bh, 1); if (err < 0) { if (err != -ENOENT) mlog_errno(err); - goto bail; + goto out; } fe = (struct ocfs2_dinode *) fe_bh->b_data; if (le16_to_cpu(fe->i_links_count) >= OCFS2_LINK_MAX) { err = -EMLINK; - goto bail; + goto out_unlock_inode; } - handle = ocfs2_start_trans(osb, handle, OCFS2_LINK_CREDITS); + handle = ocfs2_start_trans(osb, OCFS2_LINK_CREDITS); if (IS_ERR(handle)) { err = PTR_ERR(handle); handle = NULL; mlog_errno(err); - goto bail; + goto out_unlock_inode; } err = ocfs2_journal_access(handle, inode, fe_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (err < 0) { mlog_errno(err); - goto bail; + goto out_commit; } - inode->i_nlink++; + inc_nlink(inode); inode->i_ctime = CURRENT_TIME; fe->i_links_count = cpu_to_le16(inode->i_nlink); fe->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec); @@ -710,9 +732,9 @@ static int ocfs2_link(struct dentry *old_dentry, err = ocfs2_journal_dirty(handle, fe_bh); if (err < 0) { le16_add_cpu(&fe->i_links_count, -1); - inode->i_nlink--; + drop_nlink(inode); mlog_errno(err); - goto bail; + goto out_commit; } err = ocfs2_add_entry(handle, dentry, inode, @@ -720,17 +742,29 @@ static int ocfs2_link(struct dentry *old_dentry, parent_fe_bh, de_bh); if (err) { le16_add_cpu(&fe->i_links_count, -1); - inode->i_nlink--; + drop_nlink(inode); mlog_errno(err); - goto bail; + goto out_commit; + } + + err = ocfs2_dentry_attach_lock(dentry, inode, OCFS2_I(dir)->ip_blkno); + if (err) { + mlog_errno(err); + goto out_commit; } atomic_inc(&inode->i_count); dentry->d_op = &ocfs2_dentry_ops; d_instantiate(dentry, inode); -bail: - if (handle) - ocfs2_commit_trans(handle); + +out_commit: + ocfs2_commit_trans(osb, handle); +out_unlock_inode: + ocfs2_meta_unlock(inode, 1); + +out: + ocfs2_meta_unlock(dir, 1); + if (de_bh) brelse(de_bh); if (fe_bh) @@ -743,18 +777,49 @@ bail: return err; } +/* + * Takes and drops an exclusive lock on the given dentry. This will + * force other nodes to drop it. + */ +static int ocfs2_remote_dentry_delete(struct dentry *dentry) +{ + int ret; + + ret = ocfs2_dentry_lock(dentry, 1); + if (ret) + mlog_errno(ret); + else + ocfs2_dentry_unlock(dentry, 1); + + return ret; +} + +static inline int inode_is_unlinkable(struct inode *inode) +{ + if (S_ISDIR(inode->i_mode)) { + if (inode->i_nlink == 2) + return 1; + return 0; + } + + if (inode->i_nlink == 1) + return 1; + return 0; +} + static int ocfs2_unlink(struct inode *dir, struct dentry *dentry) { int status; - unsigned int saved_nlink = 0; + int child_locked = 0; struct inode *inode = dentry->d_inode; + struct inode *orphan_dir = NULL; struct ocfs2_super *osb = OCFS2_SB(dir->i_sb); u64 blkno; struct ocfs2_dinode *fe = NULL; struct buffer_head *fe_bh = NULL; struct buffer_head *parent_node_bh = NULL; - struct ocfs2_journal_handle *handle = NULL; + handle_t *handle = NULL; struct ocfs2_dir_entry *dirent = NULL; struct buffer_head *dirent_bh = NULL; char orphan_name[OCFS2_ORPHAN_NAMELEN + 1]; @@ -765,26 +830,18 @@ static int ocfs2_unlink(struct inode *dir, BUG_ON(dentry->d_parent->d_inode != dir); - mlog(0, "ino = %"MLFu64"\n", OCFS2_I(inode)->ip_blkno); + mlog(0, "ino = %llu\n", (unsigned long long)OCFS2_I(inode)->ip_blkno); if (inode == osb->root_inode) { mlog(0, "Cannot delete the root directory\n"); - status = -EPERM; - goto leave; - } - - handle = ocfs2_alloc_handle(osb); - if (handle == NULL) { - status = -ENOMEM; - mlog_errno(status); - goto leave; + return -EPERM; } - status = ocfs2_meta_lock(dir, handle, &parent_node_bh, 1); + status = ocfs2_meta_lock(dir, &parent_node_bh, 1); if (status < 0) { if (status != -ENOENT) mlog_errno(status); - goto leave; + return status; } status = ocfs2_find_files_on_disk(dentry->d_name.name, @@ -799,18 +856,19 @@ static int ocfs2_unlink(struct inode *dir, if (OCFS2_I(inode)->ip_blkno != blkno) { status = -ENOENT; - mlog(0, "ip_blkno (%"MLFu64") != dirent blkno (%"MLFu64") " - "ip_flags = %x\n", OCFS2_I(inode)->ip_blkno, blkno, - OCFS2_I(inode)->ip_flags); + mlog(0, "ip_blkno %llu != dirent blkno %llu ip_flags = %x\n", + (unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)blkno, OCFS2_I(inode)->ip_flags); goto leave; } - status = ocfs2_meta_lock(inode, handle, &fe_bh, 1); + status = ocfs2_meta_lock(inode, &fe_bh, 1); if (status < 0) { if (status != -ENOENT) mlog_errno(status); goto leave; } + child_locked = 1; if (S_ISDIR(inode->i_mode)) { if (!ocfs2_empty_dir(inode)) { @@ -822,18 +880,7 @@ static int ocfs2_unlink(struct inode *dir, } } - /* There are still a few steps left until we can consider the - * unlink to have succeeded. Save off nlink here before - * modification so we can set it back in case we hit an issue - * before commit. */ - saved_nlink = inode->i_nlink; - if (S_ISDIR(inode->i_mode)) - inode->i_nlink = 0; - else - inode->i_nlink--; - - status = ocfs2_request_unlink_vote(inode, dentry, - (unsigned int) inode->i_nlink); + status = ocfs2_remote_dentry_delete(dentry); if (status < 0) { /* This vote should succeed under all normal * circumstances. */ @@ -841,8 +888,8 @@ static int ocfs2_unlink(struct inode *dir, goto leave; } - if (!inode->i_nlink) { - status = ocfs2_prepare_orphan_dir(osb, handle, inode, + if (inode_is_unlinkable(inode)) { + status = ocfs2_prepare_orphan_dir(osb, &orphan_dir, inode, orphan_name, &orphan_entry_bh); if (status < 0) { @@ -851,7 +898,7 @@ static int ocfs2_unlink(struct inode *dir, } } - handle = ocfs2_start_trans(osb, handle, OCFS2_UNLINK_CREDITS); + handle = ocfs2_start_trans(osb, OCFS2_UNLINK_CREDITS); if (IS_ERR(handle)) { status = PTR_ERR(handle); handle = NULL; @@ -868,9 +915,9 @@ static int ocfs2_unlink(struct inode *dir, fe = (struct ocfs2_dinode *) fe_bh->b_data; - if (!inode->i_nlink) { + if (inode_is_unlinkable(inode)) { status = ocfs2_orphan_add(osb, handle, inode, fe, orphan_name, - orphan_entry_bh); + orphan_entry_bh, orphan_dir); if (status < 0) { mlog_errno(status); goto leave; @@ -884,10 +931,10 @@ static int ocfs2_unlink(struct inode *dir, goto leave; } - /* We can set nlink on the dinode now. clear the saved version - * so that it doesn't get set later. */ + if (S_ISDIR(inode->i_mode)) + drop_nlink(inode); + drop_nlink(inode); fe->i_links_count = cpu_to_le16(inode->i_nlink); - saved_nlink = 0; status = ocfs2_journal_dirty(handle, fe_bh); if (status < 0) { @@ -895,22 +942,32 @@ static int ocfs2_unlink(struct inode *dir, goto leave; } - if (S_ISDIR(inode->i_mode)) { - dir->i_nlink--; - status = ocfs2_mark_inode_dirty(handle, dir, - parent_node_bh); - if (status < 0) { - mlog_errno(status); - dir->i_nlink++; - } + dir->i_ctime = dir->i_mtime = CURRENT_TIME; + if (S_ISDIR(inode->i_mode)) + drop_nlink(dir); + + status = ocfs2_mark_inode_dirty(handle, dir, parent_node_bh); + if (status < 0) { + mlog_errno(status); + if (S_ISDIR(inode->i_mode)) + inc_nlink(dir); } leave: - if (status < 0 && saved_nlink) - inode->i_nlink = saved_nlink; - if (handle) - ocfs2_commit_trans(handle); + ocfs2_commit_trans(osb, handle); + + if (child_locked) + ocfs2_meta_unlock(inode, 1); + + ocfs2_meta_unlock(dir, 1); + + if (orphan_dir) { + /* This was locked for us in ocfs2_prepare_orphan_dir() */ + ocfs2_meta_unlock(orphan_dir, 1); + mutex_unlock(&orphan_dir->i_mutex); + iput(orphan_dir); + } if (fe_bh) brelse(fe_bh); @@ -934,7 +991,6 @@ leave: * if they have the same id, then the 1st one is the only one locked. */ static int ocfs2_double_lock(struct ocfs2_super *osb, - struct ocfs2_journal_handle *handle, struct buffer_head **bh1, struct inode *inode1, struct buffer_head **bh2, @@ -946,10 +1002,9 @@ static int ocfs2_double_lock(struct ocfs2_super *osb, struct buffer_head **tmpbh; struct inode *tmpinode; - mlog_entry("(inode1 = %"MLFu64", inode2 = %"MLFu64")\n", - oi1->ip_blkno, oi2->ip_blkno); - - BUG_ON(!handle); + mlog_entry("(inode1 = %llu, inode2 = %llu)\n", + (unsigned long long)oi1->ip_blkno, + (unsigned long long)oi2->ip_blkno); if (*bh1) *bh1 = NULL; @@ -970,25 +1025,41 @@ static int ocfs2_double_lock(struct ocfs2_super *osb, inode1 = tmpinode; } /* lock id2 */ - status = ocfs2_meta_lock(inode2, handle, bh2, 1); + status = ocfs2_meta_lock(inode2, bh2, 1); if (status < 0) { if (status != -ENOENT) mlog_errno(status); goto bail; } } + /* lock id1 */ - status = ocfs2_meta_lock(inode1, handle, bh1, 1); + status = ocfs2_meta_lock(inode1, bh1, 1); if (status < 0) { + /* + * An error return must mean that no cluster locks + * were held on function exit. + */ + if (oi1->ip_blkno != oi2->ip_blkno) + ocfs2_meta_unlock(inode2, 1); + if (status != -ENOENT) mlog_errno(status); - goto bail; } + bail: mlog_exit(status); return status; } +static void ocfs2_double_unlock(struct inode *inode1, struct inode *inode2) +{ + ocfs2_meta_unlock(inode1, 1); + + if (inode1 != inode2) + ocfs2_meta_unlock(inode2, 1); +} + #define PARENT_INO(buffer) \ ((struct ocfs2_dir_entry *) \ ((char *)buffer + \ @@ -999,17 +1070,20 @@ static int ocfs2_rename(struct inode *old_dir, struct inode *new_dir, struct dentry *new_dentry) { - int status = 0, rename_lock = 0; + int status = 0, rename_lock = 0, parents_locked = 0; + int old_child_locked = 0, new_child_locked = 0; struct inode *old_inode = old_dentry->d_inode; struct inode *new_inode = new_dentry->d_inode; + struct inode *orphan_dir = NULL; struct ocfs2_dinode *newfe = NULL; char orphan_name[OCFS2_ORPHAN_NAMELEN + 1]; struct buffer_head *orphan_entry_bh = NULL; struct buffer_head *newfe_bh = NULL; + struct buffer_head *old_inode_bh = NULL; struct buffer_head *insert_entry_bh = NULL; struct ocfs2_super *osb = NULL; u64 newfe_blkno; - struct ocfs2_journal_handle *handle = NULL; + handle_t *handle = NULL; struct buffer_head *old_dir_bh = NULL; struct buffer_head *new_dir_bh = NULL; struct ocfs2_dir_entry *old_de = NULL, *new_de = NULL; // dirent for old_dentry @@ -1017,8 +1091,7 @@ static int ocfs2_rename(struct inode *old_dir, struct buffer_head *new_de_bh = NULL, *old_de_bh = NULL; // bhs for above struct buffer_head *old_inode_de_bh = NULL; // if old_dentry is a dir, // this is the 1st dirent bh - nlink_t old_dir_nlink = old_dir->i_nlink, new_dir_nlink = new_dir->i_nlink; - unsigned int links_count; + nlink_t old_dir_nlink = old_dir->i_nlink; /* At some point it might be nice to break this function up a * bit. */ @@ -1035,14 +1108,6 @@ static int ocfs2_rename(struct inode *old_dir, BUG(); } - if (atomic_read(&old_dentry->d_count) > 2) { - shrink_dcache_parent(old_dentry); - if (atomic_read(&old_dentry->d_count) > 2) { - status = -EBUSY; - goto bail; - } - } - /* Assume a directory heirarchy thusly: * a/b/c * a/d @@ -1063,21 +1128,14 @@ static int ocfs2_rename(struct inode *old_dir, rename_lock = 1; } - handle = ocfs2_alloc_handle(osb); - if (handle == NULL) { - status = -ENOMEM; - mlog_errno(status); - goto bail; - } - /* if old and new are the same, this'll just do one lock. */ - status = ocfs2_double_lock(osb, handle, - &old_dir_bh, old_dir, - &new_dir_bh, new_dir); + status = ocfs2_double_lock(osb, &old_dir_bh, old_dir, + &new_dir_bh, new_dir); if (status < 0) { mlog_errno(status); goto bail; } + parents_locked = 1; /* make sure both dirs have bhs * get an extra ref on old_dir_bh if old==new */ @@ -1092,23 +1150,26 @@ static int ocfs2_rename(struct inode *old_dir, } } - if (S_ISDIR(old_inode->i_mode)) { - /* Directories actually require metadata updates to - * the directory info so we can't get away with not - * doing node locking on it. */ - status = ocfs2_meta_lock(old_inode, handle, NULL, 1); - if (status < 0) { - if (status != -ENOENT) - mlog_errno(status); - goto bail; - } - - status = ocfs2_request_rename_vote(old_inode, old_dentry); - if (status < 0) { + /* + * Aside from allowing a meta data update, the locking here + * also ensures that the vote thread on other nodes won't have + * to concurrently downconvert the inode and the dentry locks. + */ + status = ocfs2_meta_lock(old_inode, &old_inode_bh, 1); + if (status < 0) { + if (status != -ENOENT) mlog_errno(status); - goto bail; - } + goto bail; + } + old_child_locked = 1; + + status = ocfs2_remote_dentry_delete(old_dentry); + if (status < 0) { + mlog_errno(status); + goto bail; + } + if (S_ISDIR(old_inode->i_mode)) { status = -EIO; old_inode_de_bh = ocfs2_bread(old_inode, 0, &status, 0); if (!old_inode_de_bh) @@ -1122,14 +1183,6 @@ static int ocfs2_rename(struct inode *old_dir, if (!new_inode && new_dir!=old_dir && new_dir->i_nlink >= OCFS2_LINK_MAX) goto bail; - } else { - /* Ah, the simple case - we're a file so just send a - * message. */ - status = ocfs2_request_rename_vote(old_inode, old_dentry); - if (status < 0) { - mlog_errno(status); - goto bail; - } } status = -ENOENT; @@ -1187,27 +1240,22 @@ static int ocfs2_rename(struct inode *old_dir, if (OCFS2_I(new_inode)->ip_blkno != newfe_blkno) { status = -EACCES; - mlog(0, "Inode blkno (%"MLFu64") and dir (%"MLFu64") " - "disagree. ip_flags = %x\n", - OCFS2_I(new_inode)->ip_blkno, newfe_blkno, + mlog(0, "Inode %llu and dir %llu disagree. flags = %x\n", + (unsigned long long)OCFS2_I(new_inode)->ip_blkno, + (unsigned long long)newfe_blkno, OCFS2_I(new_inode)->ip_flags); goto bail; } - status = ocfs2_meta_lock(new_inode, handle, &newfe_bh, 1); + status = ocfs2_meta_lock(new_inode, &newfe_bh, 1); if (status < 0) { if (status != -ENOENT) mlog_errno(status); goto bail; } + new_child_locked = 1; - if (S_ISDIR(new_inode->i_mode)) - links_count = 0; - else - links_count = (unsigned int) (new_inode->i_nlink - 1); - - status = ocfs2_request_unlink_vote(new_inode, new_dentry, - links_count); + status = ocfs2_remote_dentry_delete(new_dentry); if (status < 0) { mlog_errno(status); goto bail; @@ -1215,13 +1263,13 @@ static int ocfs2_rename(struct inode *old_dir, newfe = (struct ocfs2_dinode *) newfe_bh->b_data; - mlog(0, "aha rename over existing... new_de=%p " - "new_blkno=%"MLFu64" newfebh=%p bhblocknr=%llu\n", - new_de, newfe_blkno, newfe_bh, newfe_bh ? + mlog(0, "aha rename over existing... new_de=%p new_blkno=%llu " + "newfebh=%p bhblocknr=%llu\n", new_de, + (unsigned long long)newfe_blkno, newfe_bh, newfe_bh ? (unsigned long long)newfe_bh->b_blocknr : 0ULL); if (S_ISDIR(new_inode->i_mode) || (new_inode->i_nlink == 1)) { - status = ocfs2_prepare_orphan_dir(osb, handle, + status = ocfs2_prepare_orphan_dir(osb, &orphan_dir, new_inode, orphan_name, &orphan_entry_bh); @@ -1249,7 +1297,7 @@ static int ocfs2_rename(struct inode *old_dir, } } - handle = ocfs2_start_trans(osb, handle, OCFS2_RENAME_CREDITS); + handle = ocfs2_start_trans(osb, OCFS2_RENAME_CREDITS); if (IS_ERR(handle)) { status = PTR_ERR(handle); handle = NULL; @@ -1276,7 +1324,7 @@ static int ocfs2_rename(struct inode *old_dir, (newfe->i_links_count == cpu_to_le16(1))){ status = ocfs2_orphan_add(osb, handle, new_inode, newfe, orphan_name, - orphan_entry_bh); + orphan_entry_bh, orphan_dir); if (status < 0) { mlog_errno(status); goto bail; @@ -1318,6 +1366,7 @@ static int ocfs2_rename(struct inode *old_dir, old_inode->i_ctime = CURRENT_TIME; mark_inode_dirty(old_inode); + ocfs2_mark_inode_dirty(handle, old_inode, old_inode_bh); /* now that the name has been added to new_dir, remove the old name */ status = ocfs2_delete_entry(handle, old_dir, old_de, old_de_bh); @@ -1342,40 +1391,34 @@ static int ocfs2_rename(struct inode *old_dir, if (new_inode) { new_inode->i_nlink--; } else { - new_dir->i_nlink++; + inc_nlink(new_dir); mark_inode_dirty(new_dir); } } mark_inode_dirty(old_dir); - if (new_inode) + ocfs2_mark_inode_dirty(handle, old_dir, old_dir_bh); + if (new_inode) { mark_inode_dirty(new_inode); + ocfs2_mark_inode_dirty(handle, new_inode, newfe_bh); + } - if (old_dir != new_dir) - if (new_dir_nlink != new_dir->i_nlink) { - if (!new_dir_bh) { - mlog(ML_ERROR, "need to change nlink for new " - "dir %"MLFu64" from %d to %d but bh is " - "NULL\n", OCFS2_I(new_dir)->ip_blkno, - (int)new_dir_nlink, new_dir->i_nlink); - } else { - struct ocfs2_dinode *fe; - status = ocfs2_journal_access(handle, - new_dir, - new_dir_bh, - OCFS2_JOURNAL_ACCESS_WRITE); - fe = (struct ocfs2_dinode *) new_dir_bh->b_data; - fe->i_links_count = cpu_to_le16(new_dir->i_nlink); - status = ocfs2_journal_dirty(handle, new_dir_bh); - } - } + if (old_dir != new_dir) { + /* Keep the same times on both directories.*/ + new_dir->i_ctime = new_dir->i_mtime = old_dir->i_ctime; + + /* + * This will also pick up the i_nlink change from the + * block above. + */ + ocfs2_mark_inode_dirty(handle, new_dir, new_dir_bh); + } if (old_dir_nlink != old_dir->i_nlink) { if (!old_dir_bh) { mlog(ML_ERROR, "need to change nlink for old dir " - "%"MLFu64" from %d to %d but bh is NULL!\n", - OCFS2_I(old_dir)->ip_blkno, - (int)old_dir_nlink, - old_dir->i_nlink); + "%llu from %d to %d but bh is NULL!\n", + (unsigned long long)OCFS2_I(old_dir)->ip_blkno, + (int)old_dir_nlink, old_dir->i_nlink); } else { struct ocfs2_dinode *fe; status = ocfs2_journal_access(handle, old_dir, @@ -1387,13 +1430,30 @@ static int ocfs2_rename(struct inode *old_dir, } } + ocfs2_dentry_move(old_dentry, new_dentry, old_dir, new_dir); status = 0; bail: if (rename_lock) ocfs2_rename_unlock(osb); if (handle) - ocfs2_commit_trans(handle); + ocfs2_commit_trans(osb, handle); + + if (parents_locked) + ocfs2_double_unlock(old_dir, new_dir); + + if (old_child_locked) + ocfs2_meta_unlock(old_inode, 1); + + if (new_child_locked) + ocfs2_meta_unlock(new_inode, 1); + + if (orphan_dir) { + /* This was locked for us in ocfs2_prepare_orphan_dir() */ + ocfs2_meta_unlock(orphan_dir, 1); + mutex_unlock(&orphan_dir->i_mutex); + iput(orphan_dir); + } if (new_inode) sync_mapping_buffers(old_inode->i_mapping); @@ -1402,6 +1462,8 @@ bail: iput(new_inode); if (newfe_bh) brelse(newfe_bh); + if (old_inode_bh) + brelse(old_inode_bh); if (old_dir_bh) brelse(old_dir_bh); if (new_dir_bh) @@ -1427,7 +1489,7 @@ bail: * data, including the null terminator. */ static int ocfs2_create_symlink_data(struct ocfs2_super *osb, - struct ocfs2_journal_handle *handle, + handle_t *handle, struct inode *inode, const char *symname) { @@ -1443,8 +1505,9 @@ static int ocfs2_create_symlink_data(struct ocfs2_super *osb, * write i_size + 1 bytes. */ blocks = (bytes_left + sb->s_blocksize - 1) >> sb->s_blocksize_bits; - mlog_entry("i_blocks = %lu, i_size = %llu, blocks = %d\n", - inode->i_blocks, i_size_read(inode), blocks); + mlog_entry("i_blocks = %llu, i_size = %llu, blocks = %d\n", + (unsigned long long)inode->i_blocks, + i_size_read(inode), blocks); /* Sanity check -- make sure we're going to fit. */ if (bytes_left > @@ -1541,7 +1604,7 @@ static int ocfs2_symlink(struct inode *dir, struct buffer_head *parent_fe_bh = NULL; struct ocfs2_dinode *fe = NULL; struct ocfs2_dinode *dirfe; - struct ocfs2_journal_handle *handle = NULL; + handle_t *handle = NULL; struct ocfs2_alloc_context *inode_ac = NULL; struct ocfs2_alloc_context *data_ac = NULL; @@ -1555,19 +1618,12 @@ static int ocfs2_symlink(struct inode *dir, credits = ocfs2_calc_symlink_credits(sb); - handle = ocfs2_alloc_handle(osb); - if (handle == NULL) { - status = -ENOMEM; - mlog_errno(status); - goto bail; - } - /* lock the parent directory */ - status = ocfs2_meta_lock(dir, handle, &parent_fe_bh, 1); + status = ocfs2_meta_lock(dir, &parent_fe_bh, 1); if (status < 0) { if (status != -ENOENT) mlog_errno(status); - goto bail; + return status; } dirfe = (struct ocfs2_dinode *) parent_fe_bh->b_data; @@ -1590,7 +1646,7 @@ static int ocfs2_symlink(struct inode *dir, goto bail; } - status = ocfs2_reserve_new_inode(osb, handle, &inode_ac); + status = ocfs2_reserve_new_inode(osb, &inode_ac); if (status < 0) { if (status != -ENOSPC) mlog_errno(status); @@ -1599,7 +1655,7 @@ static int ocfs2_symlink(struct inode *dir, /* don't reserve bitmap space for fast symlinks. */ if (l > ocfs2_fast_symlink_chars(sb)) { - status = ocfs2_reserve_clusters(osb, handle, 1, &data_ac); + status = ocfs2_reserve_clusters(osb, 1, &data_ac); if (status < 0) { if (status != -ENOSPC) mlog_errno(status); @@ -1607,7 +1663,7 @@ static int ocfs2_symlink(struct inode *dir, } } - handle = ocfs2_start_trans(osb, handle, credits); + handle = ocfs2_start_trans(osb, credits); if (IS_ERR(handle)) { status = PTR_ERR(handle); handle = NULL; @@ -1634,9 +1690,9 @@ static int ocfs2_symlink(struct inode *dir, NULL); if (status < 0) { if (status != -ENOSPC && status != -EINTR) { - mlog(ML_ERROR, "Failed to extend file to " - "%"MLFu64"\n", - newsize); + mlog(ML_ERROR, + "Failed to extend file to %llu\n", + (unsigned long long)newsize); mlog_errno(status); status = -ENOSPC; } @@ -1674,12 +1730,21 @@ static int ocfs2_symlink(struct inode *dir, goto bail; } + status = ocfs2_dentry_attach_lock(dentry, inode, OCFS2_I(dir)->ip_blkno); + if (status) { + mlog_errno(status); + goto bail; + } + insert_inode_hash(inode); dentry->d_op = &ocfs2_dentry_ops; d_instantiate(dentry, inode); bail: if (handle) - ocfs2_commit_trans(handle); + ocfs2_commit_trans(osb, handle); + + ocfs2_meta_unlock(dir, 1); + if (new_fe_bh) brelse(new_fe_bh); if (parent_fe_bh) @@ -1716,10 +1781,11 @@ int ocfs2_check_dir_entry(struct inode * dir, error_msg = "directory entry across blocks"; if (error_msg != NULL) - mlog(ML_ERROR, "bad entry in directory #%"MLFu64": %s - " - "offset=%lu, inode=%"MLFu64", rec_len=%d, name_len=%d\n", - OCFS2_I(dir)->ip_blkno, error_msg, offset, - le64_to_cpu(de->inode), rlen, de->name_len); + mlog(ML_ERROR, "bad entry in directory #%llu: %s - " + "offset=%lu, inode=%llu, rec_len=%d, name_len=%d\n", + (unsigned long long)OCFS2_I(dir)->ip_blkno, error_msg, + offset, (unsigned long long)le64_to_cpu(de->inode), rlen, + de->name_len); return error_msg == NULL ? 1 : 0; } @@ -1729,7 +1795,7 @@ int ocfs2_check_dir_entry(struct inode * dir, * If you pass me insert_bh, I'll skip the search of the other dir * blocks and put the record in there. */ -static int __ocfs2_add_entry(struct ocfs2_journal_handle *handle, +static int __ocfs2_add_entry(handle_t *handle, struct inode *dir, const char *name, int namelen, struct inode *inode, u64 blkno, @@ -1769,6 +1835,13 @@ static int __ocfs2_add_entry(struct ocfs2_journal_handle *handle, (le16_to_cpu(de->rec_len) >= rec_len)) || (le16_to_cpu(de->rec_len) >= (OCFS2_DIR_REC_LEN(de->name_len) + rec_len))) { + dir->i_mtime = dir->i_ctime = CURRENT_TIME; + retval = ocfs2_mark_inode_dirty(handle, dir, parent_fe_bh); + if (retval < 0) { + mlog_errno(retval); + goto bail; + } + status = ocfs2_journal_access(handle, dir, insert_bh, OCFS2_JOURNAL_ACCESS_WRITE); /* By now the buffer is marked for journaling */ @@ -1791,7 +1864,6 @@ static int __ocfs2_add_entry(struct ocfs2_journal_handle *handle, de->name_len = namelen; memcpy(de->name, name, namelen); - dir->i_mtime = dir->i_ctime = CURRENT_TIME; dir->i_version++; status = ocfs2_journal_dirty(handle, insert_bh); retval = 0; @@ -1815,7 +1887,7 @@ bail: * ocfs2_delete_entry deletes a directory entry by merging it with the * previous entry */ -static int ocfs2_delete_entry(struct ocfs2_journal_handle *handle, +static int ocfs2_delete_entry(handle_t *handle, struct inode *dir, struct ocfs2_dir_entry *de_del, struct buffer_head *bh) @@ -1961,13 +2033,8 @@ restart: } num++; - /* XXX: questionable readahead stuff here */ bh = ocfs2_bread(dir, b++, &err, 1); bh_use[ra_max] = bh; -#if 0 // ??? - if (bh) - ll_rw_block(READ, 1, &bh); -#endif } } if ((bh = bh_use[ra_ptr++]) == NULL) @@ -1975,6 +2042,10 @@ restart: wait_on_buffer(bh); if (!buffer_uptodate(bh)) { /* read error, skip block & hope for the best */ + ocfs2_error(dir->i_sb, "reading directory %llu, " + "offset %lu\n", + (unsigned long long)OCFS2_I(dir)->ip_blkno, + block); brelse(bh); goto next; } @@ -2021,8 +2092,8 @@ static int ocfs2_blkno_stringify(u64 blkno, char *name) mlog_entry_void(); - namelen = snprintf(name, OCFS2_ORPHAN_NAMELEN + 1, "%016"MLFx64, - blkno); + namelen = snprintf(name, OCFS2_ORPHAN_NAMELEN + 1, "%016llx", + (long long)blkno); if (namelen <= 0) { if (namelen) status = namelen; @@ -2047,19 +2118,19 @@ bail: } static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb, - struct ocfs2_journal_handle *handle, + struct inode **ret_orphan_dir, struct inode *inode, char *name, struct buffer_head **de_bh) { - struct inode *orphan_dir_inode = NULL; + struct inode *orphan_dir_inode; struct buffer_head *orphan_dir_bh = NULL; int status = 0; status = ocfs2_blkno_stringify(OCFS2_I(inode)->ip_blkno, name); if (status < 0) { mlog_errno(status); - goto leave; + return status; } orphan_dir_inode = ocfs2_get_system_file_inode(osb, @@ -2068,11 +2139,12 @@ static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb, if (!orphan_dir_inode) { status = -ENOENT; mlog_errno(status); - goto leave; + return status; } - ocfs2_handle_add_inode(handle, orphan_dir_inode); - status = ocfs2_meta_lock(orphan_dir_inode, handle, &orphan_dir_bh, 1); + mutex_lock(&orphan_dir_inode->i_mutex); + + status = ocfs2_meta_lock(orphan_dir_inode, &orphan_dir_bh, 1); if (status < 0) { mlog_errno(status); goto leave; @@ -2082,13 +2154,19 @@ static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb, orphan_dir_bh, name, OCFS2_ORPHAN_NAMELEN, de_bh); if (status < 0) { + ocfs2_meta_unlock(orphan_dir_inode, 1); + mlog_errno(status); goto leave; } + *ret_orphan_dir = orphan_dir_inode; + leave: - if (orphan_dir_inode) + if (status) { + mutex_unlock(&orphan_dir_inode->i_mutex); iput(orphan_dir_inode); + } if (orphan_dir_bh) brelse(orphan_dir_bh); @@ -2098,28 +2176,19 @@ leave: } static int ocfs2_orphan_add(struct ocfs2_super *osb, - struct ocfs2_journal_handle *handle, + handle_t *handle, struct inode *inode, struct ocfs2_dinode *fe, char *name, - struct buffer_head *de_bh) + struct buffer_head *de_bh, + struct inode *orphan_dir_inode) { - struct inode *orphan_dir_inode = NULL; struct buffer_head *orphan_dir_bh = NULL; int status = 0; struct ocfs2_dinode *orphan_fe; mlog_entry("(inode->i_ino = %lu)\n", inode->i_ino); - orphan_dir_inode = ocfs2_get_system_file_inode(osb, - ORPHAN_DIR_SYSTEM_INODE, - osb->slot_num); - if (!orphan_dir_inode) { - status = -ENOENT; - mlog_errno(status); - goto leave; - } - status = ocfs2_read_block(osb, OCFS2_I(orphan_dir_inode)->ip_blkno, &orphan_dir_bh, OCFS2_BH_CACHED, @@ -2167,13 +2236,10 @@ static int ocfs2_orphan_add(struct ocfs2_super *osb, OCFS2_I(inode)->ip_orphaned_slot = osb->slot_num; spin_unlock(&OCFS2_I(inode)->ip_lock); - mlog(0, "Inode %"MLFu64" orphaned in slot %d\n", - OCFS2_I(inode)->ip_blkno, osb->slot_num); + mlog(0, "Inode %llu orphaned in slot %d\n", + (unsigned long long)OCFS2_I(inode)->ip_blkno, osb->slot_num); leave: - if (orphan_dir_inode) - iput(orphan_dir_inode); - if (orphan_dir_bh) brelse(orphan_dir_bh); @@ -2183,7 +2249,7 @@ leave: /* unlike orphan_add, we expect the orphan dir to already be locked here. */ int ocfs2_orphan_del(struct ocfs2_super *osb, - struct ocfs2_journal_handle *handle, + handle_t *handle, struct inode *orphan_dir_inode, struct inode *inode, struct buffer_head *orphan_dir_bh) @@ -2202,8 +2268,9 @@ int ocfs2_orphan_del(struct ocfs2_super *osb, goto leave; } - mlog(0, "removing '%s' from orphan dir %"MLFu64" (namelen=%d)\n", - name, OCFS2_I(orphan_dir_inode)->ip_blkno, OCFS2_ORPHAN_NAMELEN); + mlog(0, "removing '%s' from orphan dir %llu (namelen=%d)\n", + name, (unsigned long long)OCFS2_I(orphan_dir_inode)->ip_blkno, + OCFS2_ORPHAN_NAMELEN); /* find it's spot in the orphan directory */ target_de_bh = ocfs2_find_entry(name, OCFS2_ORPHAN_NAMELEN, @@ -2261,4 +2328,6 @@ struct inode_operations ocfs2_dir_iops = { .rename = ocfs2_rename, .setattr = ocfs2_setattr, .getattr = ocfs2_getattr, + .sync_flags = ocfs2_sync_flags, + .permission = ocfs2_permission, };