diff options
author | Stephen Rothwell <sfr@canb.auug.org.au> | 2009-10-12 14:00:02 +1100 |
---|---|---|
committer | Stephen Rothwell <sfr@canb.auug.org.au> | 2009-10-12 14:00:02 +1100 |
commit | 02eba89fc8df61906fbec44c191af6040fbef367 (patch) | |
tree | 30ee306aadb4ac8e57afe4a8d82fc420b348d766 | |
parent | 7a5afd9d09fbbe8d3aa678095891278cf48a7313 (diff) | |
parent | cf71f3cf1697ab9ad9064635e23fecd57af2c327 (diff) |
Merge remote branch 'fsnotify/for-next'
-rw-r--r-- | Documentation/feature-removal-schedule.txt | 7 | ||||
-rw-r--r-- | fs/compat.c | 5 | ||||
-rw-r--r-- | fs/exec.c | 4 | ||||
-rw-r--r-- | fs/nfsd/vfs.c | 4 | ||||
-rw-r--r-- | fs/notify/dnotify/dnotify.c | 8 | ||||
-rw-r--r-- | fs/notify/fsnotify.c | 16 | ||||
-rw-r--r-- | fs/notify/group.c | 50 | ||||
-rw-r--r-- | fs/notify/inode_mark.c | 24 | ||||
-rw-r--r-- | fs/notify/inotify/inotify_fsnotify.c | 59 | ||||
-rw-r--r-- | fs/notify/inotify/inotify_user.c | 219 | ||||
-rw-r--r-- | fs/notify/notification.c | 161 | ||||
-rw-r--r-- | fs/open.c | 2 | ||||
-rw-r--r-- | fs/read_write.c | 8 | ||||
-rw-r--r-- | include/linux/fsnotify.h | 44 | ||||
-rw-r--r-- | include/linux/fsnotify_backend.h | 30 | ||||
-rw-r--r-- | init/Kconfig | 10 | ||||
-rw-r--r-- | kernel/Makefile | 5 | ||||
-rw-r--r-- | kernel/audit.c | 1 | ||||
-rw-r--r-- | kernel/audit.h | 26 | ||||
-rw-r--r-- | kernel/audit_tree.c | 223 | ||||
-rw-r--r-- | kernel/audit_watch.c | 294 | ||||
-rw-r--r-- | kernel/auditfilter.c | 39 | ||||
-rw-r--r-- | kernel/auditsc.c | 9 |
23 files changed, 726 insertions, 522 deletions
diff --git a/Documentation/feature-removal-schedule.txt b/Documentation/feature-removal-schedule.txt index 3c6ec37ef040..6198068815dc 100644 --- a/Documentation/feature-removal-schedule.txt +++ b/Documentation/feature-removal-schedule.txt @@ -409,6 +409,13 @@ When: 2.6.33 Why: Should be implemented in userspace, policy daemon. Who: Johannes Berg <johannes@sipsolutions.net> +--------------------------- + +What: CONFIG_INOTIFY +When: 2.6.33 +Why: No known users, fsnotify more generic and more easily maintained. +Who: Eric Paris <eparis@redhat.com> + ---------------------------- What: lock_policy_rwsem_* and unlock_policy_rwsem_* will not be diff --git a/fs/compat.c b/fs/compat.c index d576b552e8e2..441ef9135179 100644 --- a/fs/compat.c +++ b/fs/compat.c @@ -1170,11 +1170,10 @@ out: if (iov != iovstack) kfree(iov); if ((ret + (type == READ)) > 0) { - struct dentry *dentry = file->f_path.dentry; if (type == READ) - fsnotify_access(dentry); + fsnotify_access(file); else - fsnotify_modify(dentry); + fsnotify_modify(file); } return ret; } diff --git a/fs/exec.c b/fs/exec.c index d49be6bc1793..3253f4a61278 100644 --- a/fs/exec.c +++ b/fs/exec.c @@ -130,7 +130,7 @@ SYSCALL_DEFINE1(uselib, const char __user *, library) if (file->f_path.mnt->mnt_flags & MNT_NOEXEC) goto exit; - fsnotify_open(file->f_path.dentry); + fsnotify_open(file); error = -ENOEXEC; if(file->f_op) { @@ -665,7 +665,7 @@ struct file *open_exec(const char *name) if (file->f_path.mnt->mnt_flags & MNT_NOEXEC) goto exit; - fsnotify_open(file->f_path.dentry); + fsnotify_open(file); err = deny_write_access(file); if (err) diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c index 638573968dcf..098c4b4dfa1f 100644 --- a/fs/nfsd/vfs.c +++ b/fs/nfsd/vfs.c @@ -970,7 +970,7 @@ nfsd_vfs_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, nfsdstats.io_read += host_err; *count = host_err; err = 0; - fsnotify_access(file->f_path.dentry); + fsnotify_access(file); } else err = nfserrno(host_err); out: @@ -1081,7 +1081,7 @@ nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, goto out_nfserr; *cnt = host_err; nfsdstats.io_write += host_err; - fsnotify_modify(file->f_path.dentry); + fsnotify_modify(file); /* clear setuid/setgid flag after write */ if (inode->i_mode & (S_ISUID | S_ISGID)) diff --git a/fs/notify/dnotify/dnotify.c b/fs/notify/dnotify/dnotify.c index 828a889be909..8c568a09138e 100644 --- a/fs/notify/dnotify/dnotify.c +++ b/fs/notify/dnotify/dnotify.c @@ -132,7 +132,8 @@ static int dnotify_handle_event(struct fsnotify_group *group, * userspace notification for that pair. */ static bool dnotify_should_send_event(struct fsnotify_group *group, - struct inode *inode, __u32 mask) + struct inode *inode, __u32 mask, + void *data, int data_type) { struct fsnotify_mark_entry *entry; bool send; @@ -361,7 +362,7 @@ int fcntl_dirnotify(int fd, struct file *filp, unsigned long arg) dnentry = container_of(entry, struct dnotify_mark_entry, fsn_entry); spin_lock(&entry->lock); } else { - fsnotify_add_mark(new_entry, dnotify_group, inode); + fsnotify_add_mark(new_entry, dnotify_group, inode, 0); spin_lock(&new_entry->lock); entry = new_entry; dnentry = new_dnentry; @@ -431,8 +432,7 @@ static int __init dnotify_init(void) dnotify_struct_cache = KMEM_CACHE(dnotify_struct, SLAB_PANIC); dnotify_mark_entry_cache = KMEM_CACHE(dnotify_mark_entry, SLAB_PANIC); - dnotify_group = fsnotify_obtain_group(DNOTIFY_GROUP_NUM, - 0, &dnotify_fsnotify_ops); + dnotify_group = fsnotify_obtain_group(0, &dnotify_fsnotify_ops); if (IS_ERR(dnotify_group)) panic("unable to allocate fsnotify group for dnotify\n"); return 0; diff --git a/fs/notify/fsnotify.c b/fs/notify/fsnotify.c index 037e878e03fc..d4b02729598e 100644 --- a/fs/notify/fsnotify.c +++ b/fs/notify/fsnotify.c @@ -77,13 +77,16 @@ void __fsnotify_update_child_dentry_flags(struct inode *inode) } /* Notify this dentry's parent about a child's events. */ -void __fsnotify_parent(struct dentry *dentry, __u32 mask) +void __fsnotify_parent(struct file *file, struct dentry *dentry, __u32 mask) { struct dentry *parent; struct inode *p_inode; bool send = false; bool should_update_children = false; + if (file) + dentry = file->f_path.dentry; + if (!(dentry->d_flags & DCACHE_FSNOTIFY_PARENT_WATCHED)) return; @@ -114,8 +117,12 @@ void __fsnotify_parent(struct dentry *dentry, __u32 mask) * specifies these are events which came from a child. */ mask |= FS_EVENT_ON_CHILD; - fsnotify(p_inode, mask, dentry->d_inode, FSNOTIFY_EVENT_INODE, - dentry->d_name.name, 0); + if (file) + fsnotify(p_inode, mask, file, FSNOTIFY_EVENT_FILE, + dentry->d_name.name, 0); + else + fsnotify(p_inode, mask, dentry->d_inode, FSNOTIFY_EVENT_INODE, + dentry->d_name.name, 0); dput(parent); } @@ -156,7 +163,8 @@ void fsnotify(struct inode *to_tell, __u32 mask, void *data, int data_is, const idx = srcu_read_lock(&fsnotify_grp_srcu); list_for_each_entry_rcu(group, &fsnotify_groups, group_list) { if (test_mask & group->mask) { - if (!group->ops->should_send_event(group, to_tell, mask)) + if (!group->ops->should_send_event(group, to_tell, mask, + data, data_is)) continue; if (!event) { event = fsnotify_create_event(to_tell, mask, data, diff --git a/fs/notify/group.c b/fs/notify/group.c index 0e1677144bc5..62fb8961a57b 100644 --- a/fs/notify/group.c +++ b/fs/notify/group.c @@ -78,15 +78,6 @@ void fsnotify_recalc_group_mask(struct fsnotify_group *group) } /* - * Take a reference to a group so things found under the fsnotify_grp_mutex - * can't get freed under us - */ -static void fsnotify_get_group(struct fsnotify_group *group) -{ - atomic_inc(&group->refcnt); -} - -/* * Final freeing of a group */ void fsnotify_final_destroy_group(struct fsnotify_group *group) @@ -171,50 +162,23 @@ void fsnotify_put_group(struct fsnotify_group *group) } /* - * Simply run the fsnotify_groups list and find a group which matches - * the given parameters. If a group is found we take a reference to that - * group. - */ -static struct fsnotify_group *fsnotify_find_group(unsigned int group_num, __u32 mask, - const struct fsnotify_ops *ops) -{ - struct fsnotify_group *group_iter; - struct fsnotify_group *group = NULL; - - BUG_ON(!mutex_is_locked(&fsnotify_grp_mutex)); - - list_for_each_entry_rcu(group_iter, &fsnotify_groups, group_list) { - if (group_iter->group_num == group_num) { - if ((group_iter->mask == mask) && - (group_iter->ops == ops)) { - fsnotify_get_group(group_iter); - group = group_iter; - } else - group = ERR_PTR(-EEXIST); - } - } - return group; -} - -/* * Either finds an existing group which matches the group_num, mask, and ops or * creates a new group and adds it to the global group list. In either case we * take a reference for the group returned. */ -struct fsnotify_group *fsnotify_obtain_group(unsigned int group_num, __u32 mask, +struct fsnotify_group *fsnotify_obtain_group(__u32 mask, const struct fsnotify_ops *ops) { - struct fsnotify_group *group, *tgroup; + struct fsnotify_group *group; /* very low use, simpler locking if we just always alloc */ - group = kmalloc(sizeof(struct fsnotify_group), GFP_KERNEL); + group = kzalloc(sizeof(struct fsnotify_group), GFP_KERNEL); if (!group) return ERR_PTR(-ENOMEM); atomic_set(&group->refcnt, 1); group->on_group_list = 0; - group->group_num = group_num; group->mask = mask; mutex_init(&group->notification_mutex); @@ -230,14 +194,6 @@ struct fsnotify_group *fsnotify_obtain_group(unsigned int group_num, __u32 mask, group->ops = ops; mutex_lock(&fsnotify_grp_mutex); - tgroup = fsnotify_find_group(group_num, mask, ops); - if (tgroup) { - /* group already exists */ - mutex_unlock(&fsnotify_grp_mutex); - /* destroy the new one we made */ - fsnotify_put_group(group); - return tgroup; - } /* group not found, add a new one */ list_add_rcu(&group->group_list, &fsnotify_groups); diff --git a/fs/notify/inode_mark.c b/fs/notify/inode_mark.c index c8a07c65482b..574c5c620db3 100644 --- a/fs/notify/inode_mark.c +++ b/fs/notify/inode_mark.c @@ -283,12 +283,20 @@ struct fsnotify_mark_entry *fsnotify_find_mark_entry(struct fsnotify_group *grou return NULL; } +void fsnotify_duplicate_mark(struct fsnotify_mark_entry *new, struct fsnotify_mark_entry *old) +{ + assert_spin_locked(&old->lock); + new->inode = old->inode; + new->group = old->group; + new->mask = old->mask; + new->free_mark = old->free_mark; +} + /* * Nothing fancy, just initialize lists and locks and counters. */ void fsnotify_init_mark(struct fsnotify_mark_entry *entry, void (*free_mark)(struct fsnotify_mark_entry *entry)) - { spin_lock_init(&entry->lock); atomic_set(&entry->refcnt, 1); @@ -305,9 +313,10 @@ void fsnotify_init_mark(struct fsnotify_mark_entry *entry, * event types should be delivered to which group and for which inodes. */ int fsnotify_add_mark(struct fsnotify_mark_entry *entry, - struct fsnotify_group *group, struct inode *inode) + struct fsnotify_group *group, struct inode *inode, + int allow_dups) { - struct fsnotify_mark_entry *lentry; + struct fsnotify_mark_entry *lentry = NULL; int ret = 0; inode = igrab(inode); @@ -324,11 +333,12 @@ int fsnotify_add_mark(struct fsnotify_mark_entry *entry, spin_lock(&group->mark_lock); spin_lock(&inode->i_lock); - entry->group = group; - entry->inode = inode; - - lentry = fsnotify_find_mark_entry(group, inode); + if (!allow_dups) + lentry = fsnotify_find_mark_entry(group, inode); if (!lentry) { + entry->group = group; + entry->inode = inode; + hlist_add_head(&entry->i_list, &inode->i_fsnotify_mark_entries); list_add(&entry->g_list, &group->mark_entries); diff --git a/fs/notify/inotify/inotify_fsnotify.c b/fs/notify/inotify/inotify_fsnotify.c index c9ee67b442e1..027e07bf30ce 100644 --- a/fs/notify/inotify/inotify_fsnotify.c +++ b/fs/notify/inotify/inotify_fsnotify.c @@ -31,6 +31,60 @@ #include "inotify.h" +/* + * Check if 2 events contain the same information. We do not compare private data + * but at this moment that isn't a problem for any know fsnotify listeners. + */ +static bool event_compare(struct fsnotify_event *old, struct fsnotify_event *new) +{ + if ((old->mask == new->mask) && + (old->to_tell == new->to_tell) && + (old->data_type == new->data_type) && + (old->name_len == new->name_len)) { + switch (old->data_type) { + case (FSNOTIFY_EVENT_INODE): + /* remember, after old was put on the wait_q we aren't + * allowed to look at the inode any more, only thing + * left to check was if the file_name is the same */ + if (old->name_len && + !strcmp(old->file_name, new->file_name)) + return true; + break; + case (FSNOTIFY_EVENT_PATH): + if ((old->path.mnt == new->path.mnt) && + (old->path.dentry == new->path.dentry)) + return true; + break; + case (FSNOTIFY_EVENT_NONE): + if (old->mask & FS_Q_OVERFLOW) + return true; + else if (old->mask & FS_IN_IGNORED) + return false; + return true; + }; + } + return false; +} + +static int inotify_merge(struct list_head *list, struct fsnotify_event *event) +{ + struct fsnotify_event_holder *last_holder; + struct fsnotify_event *last_event; + int ret = 0; + + /* and the list better be locked by something too */ + spin_lock(&event->lock); + + last_holder = list_entry(list->prev, struct fsnotify_event_holder, event_list); + last_event = last_holder->event; + if (event_compare(last_event, event)) + ret = -EEXIST; + + spin_unlock(&event->lock); + + return ret; +} + static int inotify_handle_event(struct fsnotify_group *group, struct fsnotify_event *event) { struct fsnotify_mark_entry *entry; @@ -61,7 +115,7 @@ static int inotify_handle_event(struct fsnotify_group *group, struct fsnotify_ev fsn_event_priv->group = group; event_priv->wd = wd; - ret = fsnotify_add_notify_event(group, event, fsn_event_priv); + ret = fsnotify_add_notify_event(group, event, fsn_event_priv, inotify_merge); if (ret) { inotify_free_event_priv(fsn_event_priv); /* EEXIST says we tail matched, EOVERFLOW isn't something @@ -85,7 +139,8 @@ static void inotify_freeing_mark(struct fsnotify_mark_entry *entry, struct fsnot inotify_ignored_and_remove_idr(entry, group); } -static bool inotify_should_send_event(struct fsnotify_group *group, struct inode *inode, __u32 mask) +static bool inotify_should_send_event(struct fsnotify_group *group, struct inode *inode, + __u32 mask, void *data, int data_type) { struct fsnotify_mark_entry *entry; bool send; diff --git a/fs/notify/inotify/inotify_user.c b/fs/notify/inotify/inotify_user.c index dcd2040d330c..43930437a086 100644 --- a/fs/notify/inotify/inotify_user.c +++ b/fs/notify/inotify/inotify_user.c @@ -55,12 +55,6 @@ int inotify_max_user_watches __read_mostly; static struct kmem_cache *inotify_inode_mark_cachep __read_mostly; struct kmem_cache *event_priv_cachep __read_mostly; -/* - * When inotify registers a new group it increments this and uses that - * value as an offset to set the fsnotify group "name" and priority. - */ -static atomic_t inotify_grp_num; - #ifdef CONFIG_SYSCTL #include <linux/sysctl.h> @@ -367,6 +361,77 @@ static int inotify_find_inode(const char __user *dirname, struct path *path, uns return error; } +static int inotify_add_to_idr(struct idr *idr, spinlock_t *idr_lock, + int last_wd, + struct inotify_inode_mark_entry *ientry) +{ + int ret; + + do { + if (unlikely(!idr_pre_get(idr, GFP_KERNEL))) + return -ENOMEM; + + spin_lock(idr_lock); + ret = idr_get_new_above(idr, ientry, last_wd + 1, + &ientry->wd); + /* we added the mark to the idr, take a reference */ + if (!ret) + fsnotify_get_mark(&ientry->fsn_entry); + spin_unlock(idr_lock); + } while (ret == -EAGAIN); + + return ret; +} + +static struct inotify_inode_mark_entry *inotify_idr_find_locked(struct fsnotify_group *group, + int wd) +{ + struct idr *idr = &group->inotify_data.idr; + spinlock_t *idr_lock = &group->inotify_data.idr_lock; + struct inotify_inode_mark_entry *ientry; + + assert_spin_locked(idr_lock); + + ientry = idr_find(idr, wd); + if (ientry) { + struct fsnotify_mark_entry *fsn_entry = &ientry->fsn_entry; + + fsnotify_get_mark(fsn_entry); + /* One ref for being in the idr, one ref we just took */ + BUG_ON(atomic_read(&fsn_entry->refcnt) < 2); + } + + return ientry; +} + +static struct inotify_inode_mark_entry *inotify_idr_find(struct fsnotify_group *group, + int wd) +{ + struct inotify_inode_mark_entry *ientry; + spinlock_t *idr_lock = &group->inotify_data.idr_lock; + + spin_lock(idr_lock); + ientry = inotify_idr_find_locked(group, wd); + spin_unlock(idr_lock); + + return ientry; +} + +static void do_inotify_remove_from_idr(struct fsnotify_group *group, + struct inotify_inode_mark_entry *ientry) +{ + struct idr *idr = &group->inotify_data.idr; + spinlock_t *idr_lock = &group->inotify_data.idr_lock; + int wd = ientry->wd; + + assert_spin_locked(idr_lock); + + idr_remove(idr, wd); + + /* removed from the idr, drop that ref */ + fsnotify_put_mark(&ientry->fsn_entry); +} + /* * Remove the mark from the idr (if present) and drop the reference * on the mark because it was in the idr. @@ -374,42 +439,72 @@ static int inotify_find_inode(const char __user *dirname, struct path *path, uns static void inotify_remove_from_idr(struct fsnotify_group *group, struct inotify_inode_mark_entry *ientry) { - struct idr *idr; - struct fsnotify_mark_entry *entry; - struct inotify_inode_mark_entry *found_ientry; + spinlock_t *idr_lock = &group->inotify_data.idr_lock; + struct inotify_inode_mark_entry *found_ientry = NULL; int wd; - spin_lock(&group->inotify_data.idr_lock); - idr = &group->inotify_data.idr; + spin_lock(idr_lock); wd = ientry->wd; - if (wd == -1) + /* + * does this ientry think it is in the idr? we shouldn't get called + * if it wasn't.... + */ + if (wd == -1) { + printk(KERN_WARNING "%s: ientry=%p ientry->wd=%d ientry->group=%p" + " ientry->inode=%p\n", __func__, ientry, ientry->wd, + ientry->fsn_entry.group, ientry->fsn_entry.inode); + WARN_ON(1); goto out; + } - entry = idr_find(&group->inotify_data.idr, wd); - if (unlikely(!entry)) + /* Lets look in the idr to see if we find it */ + found_ientry = inotify_idr_find_locked(group, wd); + if (unlikely(!found_ientry)) { + printk(KERN_WARNING "%s: ientry=%p ientry->wd=%d ientry->group=%p" + " ientry->inode=%p\n", __func__, ientry, ientry->wd, + ientry->fsn_entry.group, ientry->fsn_entry.inode); + WARN_ON(1); goto out; + } - found_ientry = container_of(entry, struct inotify_inode_mark_entry, fsn_entry); + /* + * We found an entry in the idr at the right wd, but it's + * not the entry we were told to remove. eparis seriously + * fucked up somewhere. + */ if (unlikely(found_ientry != ientry)) { - /* We found an entry in the idr with the right wd, but it's - * not the entry we were told to remove. eparis seriously - * fucked up somewhere. */ WARN_ON(1); - ientry->wd = -1; + printk(KERN_WARNING "%s: ientry=%p ientry->wd=%d ientry->group=%p " + "entry->inode=%p found_ientry=%p found_ientry->wd=%d " + "found_ientry->group=%p found_ientry->inode=%p\n", + __func__, ientry, ientry->wd, ientry->fsn_entry.group, + ientry->fsn_entry.inode, found_ientry, found_ientry->wd, + found_ientry->fsn_entry.group, + found_ientry->fsn_entry.inode); goto out; } - /* One ref for being in the idr, one ref held by the caller */ - BUG_ON(atomic_read(&entry->refcnt) < 2); - - idr_remove(idr, wd); - ientry->wd = -1; + /* + * One ref for being in the idr + * one ref held by the caller trying to kill us + * one ref grabbed by inotify_idr_find + */ + if (unlikely(atomic_read(&ientry->fsn_entry.refcnt) < 3)) { + printk(KERN_WARNING "%s: ientry=%p ientry->wd=%d ientry->group=%p" + " ientry->inode=%p\n", __func__, ientry, ientry->wd, + ientry->fsn_entry.group, ientry->fsn_entry.inode); + /* we can't really recover with bad ref cnting.. */ + BUG(); + } - /* removed from the idr, drop that ref */ - fsnotify_put_mark(entry); + do_inotify_remove_from_idr(group, ientry); out: - spin_unlock(&group->inotify_data.idr_lock); + /* match the ref taken by inotify_idr_find_locked() */ + if (found_ientry) + fsnotify_put_mark(&found_ientry->fsn_entry); + ientry->wd = -1; + spin_unlock(idr_lock); } /* @@ -441,7 +536,7 @@ void inotify_ignored_and_remove_idr(struct fsnotify_mark_entry *entry, fsn_event_priv->group = group; event_priv->wd = ientry->wd; - ret = fsnotify_add_notify_event(group, ignored_event, fsn_event_priv); + ret = fsnotify_add_notify_event(group, ignored_event, fsn_event_priv, NULL); if (ret) inotify_free_event_priv(fsn_event_priv); @@ -459,7 +554,9 @@ skip_send_ignore: /* ding dong the mark is dead */ static void inotify_free_mark(struct fsnotify_mark_entry *entry) { - struct inotify_inode_mark_entry *ientry = (struct inotify_inode_mark_entry *)entry; + struct inotify_inode_mark_entry *ientry; + + ientry = container_of(entry, struct inotify_inode_mark_entry, fsn_entry); kmem_cache_free(inotify_inode_mark_cachep, ientry); } @@ -534,6 +631,8 @@ static int inotify_new_watch(struct fsnotify_group *group, struct inotify_inode_mark_entry *tmp_ientry; __u32 mask; int ret; + struct idr *idr = &group->inotify_data.idr; + spinlock_t *idr_lock = &group->inotify_data.idr_lock; /* don't allow invalid bits: we don't want flags set */ mask = inotify_arg_to_mask(arg); @@ -551,28 +650,14 @@ static int inotify_new_watch(struct fsnotify_group *group, ret = -ENOSPC; if (atomic_read(&group->inotify_data.user->inotify_watches) >= inotify_max_user_watches) goto out_err; -retry: - ret = -ENOMEM; - if (unlikely(!idr_pre_get(&group->inotify_data.idr, GFP_KERNEL))) - goto out_err; - spin_lock(&group->inotify_data.idr_lock); - ret = idr_get_new_above(&group->inotify_data.idr, &tmp_ientry->fsn_entry, - group->inotify_data.last_wd, - &tmp_ientry->wd); - spin_unlock(&group->inotify_data.idr_lock); - if (ret) { - /* idr was out of memory allocate and try again */ - if (ret == -EAGAIN) - goto retry; + ret = inotify_add_to_idr(idr, idr_lock, group->inotify_data.last_wd, + tmp_ientry); + if (ret) goto out_err; - } - - /* we put the mark on the idr, take a reference */ - fsnotify_get_mark(&tmp_ientry->fsn_entry); /* we are on the idr, now get on the inode */ - ret = fsnotify_add_mark(&tmp_ientry->fsn_entry, group, inode); + ret = fsnotify_add_mark(&tmp_ientry->fsn_entry, group, inode, 0); if (ret) { /* we failed to get on the inode, get off the idr */ inotify_remove_from_idr(group, tmp_ientry); @@ -588,16 +673,13 @@ retry: /* return the watch descriptor for this new entry */ ret = tmp_ientry->wd; - /* match the ref from fsnotify_init_markentry() */ - fsnotify_put_mark(&tmp_ientry->fsn_entry); - /* if this mark added a new event update the group mask */ if (mask & ~group->mask) fsnotify_recalc_group_mask(group); out_err: - if (ret < 0) - kmem_cache_free(inotify_inode_mark_cachep, tmp_ientry); + /* match the ref from fsnotify_init_markentry() */ + fsnotify_put_mark(&tmp_ientry->fsn_entry); return ret; } @@ -626,11 +708,8 @@ retry: static struct fsnotify_group *inotify_new_group(struct user_struct *user, unsigned int max_events) { struct fsnotify_group *group; - unsigned int grp_num; - /* fsnotify_obtain_group took a reference to group, we put this when we kill the file in the end */ - grp_num = (INOTIFY_GROUP_NUM - atomic_inc_return(&inotify_grp_num)); - group = fsnotify_obtain_group(grp_num, 0, &inotify_fsnotify_ops); + group = fsnotify_obtain_group(0, &inotify_fsnotify_ops); if (IS_ERR(group)) return group; @@ -638,7 +717,7 @@ static struct fsnotify_group *inotify_new_group(struct user_struct *user, unsign spin_lock_init(&group->inotify_data.idr_lock); idr_init(&group->inotify_data.idr); - group->inotify_data.last_wd = 1; + group->inotify_data.last_wd = 0; group->inotify_data.user = user; group->inotify_data.fa = NULL; @@ -760,7 +839,7 @@ fput_and_out: SYSCALL_DEFINE2(inotify_rm_watch, int, fd, __s32, wd) { struct fsnotify_group *group; - struct fsnotify_mark_entry *entry; + struct inotify_inode_mark_entry *ientry; struct file *filp; int ret = 0, fput_needed; @@ -769,25 +848,23 @@ SYSCALL_DEFINE2(inotify_rm_watch, int, fd, __s32, wd) return -EBADF; /* verify that this is indeed an inotify instance */ - if (unlikely(filp->f_op != &inotify_fops)) { - ret = -EINVAL; + ret = -EINVAL; + if (unlikely(filp->f_op != &inotify_fops)) goto out; - } group = filp->private_data; - spin_lock(&group->inotify_data.idr_lock); - entry = idr_find(&group->inotify_data.idr, wd); - if (unlikely(!entry)) { - spin_unlock(&group->inotify_data.idr_lock); - ret = -EINVAL; + ret = -EINVAL; + ientry = inotify_idr_find(group, wd); + if (unlikely(!ientry)) goto out; - } - fsnotify_get_mark(entry); - spin_unlock(&group->inotify_data.idr_lock); - fsnotify_destroy_mark_by_entry(entry); - fsnotify_put_mark(entry); + ret = 0; + + fsnotify_destroy_mark_by_entry(&ientry->fsn_entry); + + /* match ref taken by inotify_idr_find */ + fsnotify_put_mark(&ientry->fsn_entry); out: fput_light(filp, fput_needed); diff --git a/fs/notify/notification.c b/fs/notify/notification.c index 3816d5750dd5..dafd0b7687b8 100644 --- a/fs/notify/notification.c +++ b/fs/notify/notification.c @@ -56,7 +56,7 @@ static struct kmem_cache *fsnotify_event_holder_cachep; * it is needed. It's refcnt is set 1 at kernel init time and will never * get set to 0 so it will never get 'freed' */ -static struct fsnotify_event q_overflow_event; +static struct fsnotify_event *q_overflow_event; static atomic_t fsnotify_sync_cookie = ATOMIC_INIT(0); /** @@ -104,7 +104,8 @@ struct fsnotify_event_holder *fsnotify_alloc_event_holder(void) void fsnotify_destroy_event_holder(struct fsnotify_event_holder *holder) { - kmem_cache_free(fsnotify_event_holder_cachep, holder); + if (holder) + kmem_cache_free(fsnotify_event_holder_cachep, holder); } /* @@ -129,53 +130,17 @@ struct fsnotify_event_private_data *fsnotify_remove_priv_from_event(struct fsnot } /* - * Check if 2 events contain the same information. We do not compare private data - * but at this moment that isn't a problem for any know fsnotify listeners. - */ -static bool event_compare(struct fsnotify_event *old, struct fsnotify_event *new) -{ - if ((old->mask == new->mask) && - (old->to_tell == new->to_tell) && - (old->data_type == new->data_type) && - (old->name_len == new->name_len)) { - switch (old->data_type) { - case (FSNOTIFY_EVENT_INODE): - /* remember, after old was put on the wait_q we aren't - * allowed to look at the inode any more, only thing - * left to check was if the file_name is the same */ - if (old->name_len && - !strcmp(old->file_name, new->file_name)) - return true; - break; - case (FSNOTIFY_EVENT_PATH): - if ((old->path.mnt == new->path.mnt) && - (old->path.dentry == new->path.dentry)) - return true; - break; - case (FSNOTIFY_EVENT_NONE): - if (old->mask & FS_Q_OVERFLOW) - return true; - else if (old->mask & FS_IN_IGNORED) - return false; - return false; - }; - } - return false; -} - -/* * Add an event to the group notification queue. The group can later pull this * event off the queue to deal with. If the event is successfully added to the * group's notification queue, a reference is taken on event. */ int fsnotify_add_notify_event(struct fsnotify_group *group, struct fsnotify_event *event, - struct fsnotify_event_private_data *priv) + struct fsnotify_event_private_data *priv, + int (*merge)(struct list_head *, struct fsnotify_event *)) { struct fsnotify_event_holder *holder = NULL; struct list_head *list = &group->notification_list; - struct fsnotify_event_holder *last_holder; - struct fsnotify_event *last_event; - int ret = 0; + int rc = 0; /* * There is one fsnotify_event_holder embedded inside each fsnotify_event. @@ -195,12 +160,24 @@ alloc_holder: mutex_lock(&group->notification_mutex); if (group->q_len >= group->max_events) { - event = &q_overflow_event; - ret = -EOVERFLOW; + event = q_overflow_event; + rc = -EOVERFLOW; /* sorry, no private data on the overflow event */ priv = NULL; } + if (!list_empty(list) && merge) { + int ret; + + ret = merge(list, event); + if (ret) { + mutex_unlock(&group->notification_mutex); + if (holder != &event->holder) + fsnotify_destroy_event_holder(holder); + return ret; + } + } + spin_lock(&event->lock); if (list_empty(&event->holder.event_list)) { @@ -215,18 +192,6 @@ alloc_holder: goto alloc_holder; } - if (!list_empty(list)) { - last_holder = list_entry(list->prev, struct fsnotify_event_holder, event_list); - last_event = last_holder->event; - if (event_compare(last_event, event)) { - spin_unlock(&event->lock); - mutex_unlock(&group->notification_mutex); - if (holder != &event->holder) - fsnotify_destroy_event_holder(holder); - return -EEXIST; - } - } - group->q_len++; holder->event = event; @@ -238,7 +203,7 @@ alloc_holder: mutex_unlock(&group->notification_mutex); wake_up(&group->notification_waitq); - return ret; + return rc; } /* @@ -314,25 +279,77 @@ void fsnotify_flush_notify(struct fsnotify_group *group) static void initialize_event(struct fsnotify_event *event) { - event->holder.event = NULL; INIT_LIST_HEAD(&event->holder.event_list); atomic_set(&event->refcnt, 1); spin_lock_init(&event->lock); - event->path.dentry = NULL; - event->path.mnt = NULL; - event->inode = NULL; - event->data_type = FSNOTIFY_EVENT_NONE; - INIT_LIST_HEAD(&event->private_data_list); +} + +/* + * Caller damn well better be holding whatever mutex is protecting the + * old_holder->event_list and the new_event must be a clean event which + * cannot be found anywhere else in the kernel. + */ +int fsnotify_replace_event(struct fsnotify_event_holder *old_holder, + struct fsnotify_event *new_event) +{ + struct fsnotify_event *old_event = old_holder->event; + struct fsnotify_event_holder *new_holder = &new_event->holder; - event->to_tell = NULL; + enum event_spinlock_class { + SPINLOCK_OLD, + SPINLOCK_NEW, + }; - event->file_name = NULL; - event->name_len = 0; + /* + * if the new_event's embedded holder is in use someone + * screwed up and didn't give us a clean new event. + */ + BUG_ON(!list_empty(&new_holder->event_list)); + + spin_lock_nested(&old_event->lock, SPINLOCK_OLD); + spin_lock_nested(&new_event->lock, SPINLOCK_NEW); + + new_holder->event = new_event; + list_replace_init(&old_holder->event_list, &new_holder->event_list); + + spin_unlock(&new_event->lock); + spin_unlock(&old_event->lock); + + /* event == holder means we are referenced through the in event holder */ + if (old_holder != &old_event->holder) + fsnotify_destroy_event_holder(old_holder); + + fsnotify_get_event(new_event); /* on the list take reference */ + fsnotify_put_event(old_event); /* off the list, drop reference */ + + return 0; +} + +struct fsnotify_event *fsnotify_clone_event(struct fsnotify_event *old_event) +{ + struct fsnotify_event *event; - event->sync_cookie = 0; + event = kmem_cache_alloc(fsnotify_event_cachep, GFP_KERNEL); + if (!event) + return NULL; + + memcpy(event, old_event, sizeof(*event)); + initialize_event(event); + + if (event->name_len) { + event->file_name = kstrdup(old_event->file_name, GFP_KERNEL); + if (!event->file_name) { + kmem_cache_free(fsnotify_event_cachep, event); + return NULL; + } + } + if (event->data_type == FSNOTIFY_EVENT_PATH) + path_get(&event->path); + + return event; } /* @@ -353,7 +370,7 @@ struct fsnotify_event *fsnotify_create_event(struct inode *to_tell, __u32 mask, { struct fsnotify_event *event; - event = kmem_cache_alloc(fsnotify_event_cachep, gfp); + event = kmem_cache_zalloc(fsnotify_event_cachep, gfp); if (!event) return NULL; @@ -370,6 +387,7 @@ struct fsnotify_event *fsnotify_create_event(struct inode *to_tell, __u32 mask, event->sync_cookie = cookie; event->to_tell = to_tell; + event->data_type = data_type; switch (data_type) { case FSNOTIFY_EVENT_FILE: { @@ -386,12 +404,10 @@ struct fsnotify_event *fsnotify_create_event(struct inode *to_tell, __u32 mask, event->path.dentry = path->dentry; event->path.mnt = path->mnt; path_get(&event->path); - event->data_type = FSNOTIFY_EVENT_PATH; break; } case FSNOTIFY_EVENT_INODE: event->inode = data; - event->data_type = FSNOTIFY_EVENT_INODE; break; case FSNOTIFY_EVENT_NONE: event->inode = NULL; @@ -412,8 +428,11 @@ __init int fsnotify_notification_init(void) fsnotify_event_cachep = KMEM_CACHE(fsnotify_event, SLAB_PANIC); fsnotify_event_holder_cachep = KMEM_CACHE(fsnotify_event_holder, SLAB_PANIC); - initialize_event(&q_overflow_event); - q_overflow_event.mask = FS_Q_OVERFLOW; + q_overflow_event = fsnotify_create_event(NULL, FS_Q_OVERFLOW, NULL, + FSNOTIFY_EVENT_NONE, NULL, 0, + GFP_KERNEL); + if (!q_overflow_event) + panic("unable to allocate fsnotify q_overflow_event\n"); return 0; } diff --git a/fs/open.c b/fs/open.c index 4f01e06227c6..c064e7be15e5 100644 --- a/fs/open.c +++ b/fs/open.c @@ -1041,7 +1041,7 @@ long do_sys_open(int dfd, const char __user *filename, int flags, int mode) put_unused_fd(fd); fd = PTR_ERR(f); } else { - fsnotify_open(f->f_path.dentry); + fsnotify_open(f); fd_install(fd, f); } } diff --git a/fs/read_write.c b/fs/read_write.c index 3ac28987f22a..0ddf1626e202 100644 --- a/fs/read_write.c +++ b/fs/read_write.c @@ -293,7 +293,7 @@ ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos) else ret = do_sync_read(file, buf, count, pos); if (ret > 0) { - fsnotify_access(file->f_path.dentry); + fsnotify_access(file); add_rchar(current, ret); } inc_syscr(current); @@ -348,7 +348,7 @@ ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_ else ret = do_sync_write(file, buf, count, pos); if (ret > 0) { - fsnotify_modify(file->f_path.dentry); + fsnotify_modify(file); add_wchar(current, ret); } inc_syscw(current); @@ -656,9 +656,9 @@ out: kfree(iov); if ((ret + (type == READ)) > 0) { if (type == READ) - fsnotify_access(file->f_path.dentry); + fsnotify_access(file); else - fsnotify_modify(file->f_path.dentry); + fsnotify_modify(file); } return ret; } diff --git a/include/linux/fsnotify.h b/include/linux/fsnotify.h index 936f9aa8bb97..c0de28b217c9 100644 --- a/include/linux/fsnotify.h +++ b/include/linux/fsnotify.h @@ -29,9 +29,14 @@ static inline void fsnotify_d_instantiate(struct dentry *entry, } /* Notify this dentry's parent about a child's events. */ -static inline void fsnotify_parent(struct dentry *dentry, __u32 mask) +static inline void fsnotify_parent(struct file *file, struct dentry *dentry, __u32 mask) { - __fsnotify_parent(dentry, mask); + BUG_ON(file && dentry); + + if (file) + dentry = file->f_path.dentry; + + __fsnotify_parent(file, dentry, mask); inotify_dentry_parent_queue_event(dentry, mask, 0, dentry->d_name.name); } @@ -124,7 +129,7 @@ static inline void fsnotify_nameremove(struct dentry *dentry, int isdir) if (isdir) mask |= FS_IN_ISDIR; - fsnotify_parent(dentry, mask); + fsnotify_parent(NULL, dentry, mask); } /* @@ -183,9 +188,9 @@ static inline void fsnotify_mkdir(struct inode *inode, struct dentry *dentry) /* * fsnotify_access - file was read */ -static inline void fsnotify_access(struct dentry *dentry) +static inline void fsnotify_access(struct file *file) { - struct inode *inode = dentry->d_inode; + struct inode *inode = file->f_path.dentry->d_inode; __u32 mask = FS_ACCESS; if (S_ISDIR(inode->i_mode)) @@ -193,16 +198,16 @@ static inline void fsnotify_access(struct dentry *dentry) inotify_inode_queue_event(inode, mask, 0, NULL, NULL); - fsnotify_parent(dentry, mask); - fsnotify(inode, mask, inode, FSNOTIFY_EVENT_INODE, NULL, 0); + fsnotify_parent(file, NULL, mask); + fsnotify(inode, mask, file, FSNOTIFY_EVENT_FILE, NULL, 0); } /* * fsnotify_modify - file was modified */ -static inline void fsnotify_modify(struct dentry *dentry) +static inline void fsnotify_modify(struct file *file) { - struct inode *inode = dentry->d_inode; + struct inode *inode = file->f_path.dentry->d_inode; __u32 mask = FS_MODIFY; if (S_ISDIR(inode->i_mode)) @@ -210,16 +215,16 @@ static inline void fsnotify_modify(struct dentry *dentry) inotify_inode_queue_event(inode, mask, 0, NULL, NULL); - fsnotify_parent(dentry, mask); - fsnotify(inode, mask, inode, FSNOTIFY_EVENT_INODE, NULL, 0); + fsnotify_parent(file, NULL, mask); + fsnotify(inode, mask, file, FSNOTIFY_EVENT_FILE, NULL, 0); } /* * fsnotify_open - file was opened */ -static inline void fsnotify_open(struct dentry *dentry) +static inline void fsnotify_open(struct file *file) { - struct inode *inode = dentry->d_inode; + struct inode *inode = file->f_path.dentry->d_inode; __u32 mask = FS_OPEN; if (S_ISDIR(inode->i_mode)) @@ -227,8 +232,8 @@ static inline void fsnotify_open(struct dentry *dentry) inotify_inode_queue_event(inode, mask, 0, NULL, NULL); - fsnotify_parent(dentry, mask); - fsnotify(inode, mask, inode, FSNOTIFY_EVENT_INODE, NULL, 0); + fsnotify_parent(file, NULL, mask); + fsnotify(inode, mask, file, FSNOTIFY_EVENT_FILE, NULL, 0); } /* @@ -236,8 +241,7 @@ static inline void fsnotify_open(struct dentry *dentry) */ static inline void fsnotify_close(struct file *file) { - struct dentry *dentry = file->f_path.dentry; - struct inode *inode = dentry->d_inode; + struct inode *inode = file->f_path.dentry->d_inode; fmode_t mode = file->f_mode; __u32 mask = (mode & FMODE_WRITE) ? FS_CLOSE_WRITE : FS_CLOSE_NOWRITE; @@ -246,7 +250,7 @@ static inline void fsnotify_close(struct file *file) inotify_inode_queue_event(inode, mask, 0, NULL, NULL); - fsnotify_parent(dentry, mask); + fsnotify_parent(file, NULL, mask); fsnotify(inode, mask, file, FSNOTIFY_EVENT_FILE, NULL, 0); } @@ -263,7 +267,7 @@ static inline void fsnotify_xattr(struct dentry *dentry) inotify_inode_queue_event(inode, mask, 0, NULL, NULL); - fsnotify_parent(dentry, mask); + fsnotify_parent(NULL, dentry, mask); fsnotify(inode, mask, inode, FSNOTIFY_EVENT_INODE, NULL, 0); } @@ -299,7 +303,7 @@ static inline void fsnotify_change(struct dentry *dentry, unsigned int ia_valid) mask |= FS_IN_ISDIR; inotify_inode_queue_event(inode, mask, 0, NULL, NULL); - fsnotify_parent(dentry, mask); + fsnotify_parent(NULL, dentry, mask); fsnotify(inode, mask, inode, FSNOTIFY_EVENT_INODE, NULL, 0); } } diff --git a/include/linux/fsnotify_backend.h b/include/linux/fsnotify_backend.h index 4d6f47b51189..57e503d017c8 100644 --- a/include/linux/fsnotify_backend.h +++ b/include/linux/fsnotify_backend.h @@ -58,9 +58,7 @@ FS_MOVED_FROM | FS_MOVED_TO | FS_CREATE |\ FS_DELETE) -/* listeners that hard code group numbers near the top */ -#define DNOTIFY_GROUP_NUM UINT_MAX -#define INOTIFY_GROUP_NUM (DNOTIFY_GROUP_NUM-1) +#define FS_MOVE (FS_MOVED_FROM | FS_MOVED_TO) struct fsnotify_group; struct fsnotify_event; @@ -80,7 +78,8 @@ struct fsnotify_event_private_data; * valid group and inode to use to clean up. */ struct fsnotify_ops { - bool (*should_send_event)(struct fsnotify_group *group, struct inode *inode, __u32 mask); + bool (*should_send_event)(struct fsnotify_group *group, struct inode *inode, + __u32 mask, void *data, int data_type); int (*handle_event)(struct fsnotify_group *group, struct fsnotify_event *event); void (*free_group_priv)(struct fsnotify_group *group); void (*freeing_mark)(struct fsnotify_mark_entry *entry, struct fsnotify_group *group); @@ -119,7 +118,6 @@ struct fsnotify_group { * closed. */ atomic_t refcnt; /* things with interest in this group */ - unsigned int group_num; /* simply prevents accidental group collision */ const struct fsnotify_ops *ops; /* how this group handles things */ @@ -254,7 +252,7 @@ struct fsnotify_mark_entry { /* main fsnotify call to send events */ extern void fsnotify(struct inode *to_tell, __u32 mask, void *data, int data_is, const char *name, u32 cookie); -extern void __fsnotify_parent(struct dentry *dentry, __u32 mask); +extern void __fsnotify_parent(struct file *file, struct dentry *dentry, __u32 mask); extern void __fsnotify_inode_delete(struct inode *inode); extern u32 fsnotify_get_cookie(void); @@ -307,8 +305,7 @@ static inline void __fsnotify_d_instantiate(struct dentry *dentry, struct inode /* must call when a group changes its ->mask */ extern void fsnotify_recalc_global_mask(void); /* get a reference to an existing or create a new group */ -extern struct fsnotify_group *fsnotify_obtain_group(unsigned int group_num, - __u32 mask, +extern struct fsnotify_group *fsnotify_obtain_group(__u32 mask, const struct fsnotify_ops *ops); /* run all marks associated with this group and update group->mask */ extern void fsnotify_recalc_group_mask(struct fsnotify_group *group); @@ -323,8 +320,10 @@ extern struct fsnotify_event_private_data *fsnotify_remove_priv_from_event(struc struct fsnotify_event *event); /* attach the event to the group notification queue */ -extern int fsnotify_add_notify_event(struct fsnotify_group *group, struct fsnotify_event *event, - struct fsnotify_event_private_data *priv); +extern int fsnotify_add_notify_event(struct fsnotify_group *group, + struct fsnotify_event *event, + struct fsnotify_event_private_data *priv, + int (*merge)(struct list_head *, struct fsnotify_event *)); /* true if the group notification queue is empty */ extern bool fsnotify_notify_queue_is_empty(struct fsnotify_group *group); /* return, but do not dequeue the first event on the notification queue */ @@ -339,8 +338,10 @@ extern void fsnotify_recalc_inode_mask(struct inode *inode); extern void fsnotify_init_mark(struct fsnotify_mark_entry *entry, void (*free_mark)(struct fsnotify_mark_entry *entry)); /* find (and take a reference) to a mark associated with group and inode */ extern struct fsnotify_mark_entry *fsnotify_find_mark_entry(struct fsnotify_group *group, struct inode *inode); +/* copy the values from old into new */ +extern void fsnotify_duplicate_mark(struct fsnotify_mark_entry *new, struct fsnotify_mark_entry *old); /* attach the mark to both the group and the inode */ -extern int fsnotify_add_mark(struct fsnotify_mark_entry *entry, struct fsnotify_group *group, struct inode *inode); +extern int fsnotify_add_mark(struct fsnotify_mark_entry *entry, struct fsnotify_group *group, struct inode *inode, int allow_dups); /* given a mark, flag it to be freed when all references are dropped */ extern void fsnotify_destroy_mark_by_entry(struct fsnotify_mark_entry *entry); /* run all the marks in a group, and flag them to be freed */ @@ -354,13 +355,18 @@ extern struct fsnotify_event *fsnotify_create_event(struct inode *to_tell, __u32 void *data, int data_is, const char *name, u32 cookie, gfp_t gfp); +/* fanotify likes to change events after they are on lists... */ +extern struct fsnotify_event *fsnotify_clone_event(struct fsnotify_event *old_event); +extern int fsnotify_replace_event(struct fsnotify_event_holder *old_holder, + struct fsnotify_event *new_event); + #else static inline void fsnotify(struct inode *to_tell, __u32 mask, void *data, int data_is, const char *name, u32 cookie) {} -static inline void __fsnotify_parent(struct dentry *dentry, __u32 mask) +static inline void __fsnotify_parent(struct file *file, struct dentry *dentry, __u32 mask) {} static inline void __fsnotify_inode_delete(struct inode *inode) diff --git a/init/Kconfig b/init/Kconfig index 2e6e59f3e1b8..08928803a328 100644 --- a/init/Kconfig +++ b/init/Kconfig @@ -302,13 +302,17 @@ config AUDITSYSCALL help Enable low-overhead system-call auditing infrastructure that can be used independently or with another kernel subsystem, - such as SELinux. To use audit's filesystem watch feature, please - ensure that INOTIFY is configured. + such as SELinux. + +config AUDIT_WATCH + def_bool y + depends on AUDITSYSCALL + select FSNOTIFY config AUDIT_TREE def_bool y depends on AUDITSYSCALL - select INOTIFY + select FSNOTIFY menu "RCU Subsystem" diff --git a/kernel/Makefile b/kernel/Makefile index b8d4cd8ac0b9..7556faf509c1 100644 --- a/kernel/Makefile +++ b/kernel/Makefile @@ -68,10 +68,11 @@ obj-$(CONFIG_IKCONFIG) += configs.o obj-$(CONFIG_RESOURCE_COUNTERS) += res_counter.o obj-$(CONFIG_STOP_MACHINE) += stop_machine.o obj-$(CONFIG_KPROBES_SANITY_TEST) += test_kprobes.o -obj-$(CONFIG_AUDIT) += audit.o auditfilter.o audit_watch.o +obj-$(CONFIG_AUDIT) += audit.o auditfilter.o obj-$(CONFIG_AUDITSYSCALL) += auditsc.o -obj-$(CONFIG_GCOV_KERNEL) += gcov/ +obj-$(CONFIG_AUDIT_WATCH) += audit_watch.o obj-$(CONFIG_AUDIT_TREE) += audit_tree.o +obj-$(CONFIG_GCOV_KERNEL) += gcov/ obj-$(CONFIG_KPROBES) += kprobes.o obj-$(CONFIG_KGDB) += kgdb.o obj-$(CONFIG_DETECT_SOFTLOCKUP) += softlockup.o diff --git a/kernel/audit.c b/kernel/audit.c index 5feed232be9d..057a3a4dbfa5 100644 --- a/kernel/audit.c +++ b/kernel/audit.c @@ -55,7 +55,6 @@ #include <net/netlink.h> #include <linux/skbuff.h> #include <linux/netlink.h> -#include <linux/inotify.h> #include <linux/freezer.h> #include <linux/tty.h> diff --git a/kernel/audit.h b/kernel/audit.h index 208687be4f30..f7206db4e13d 100644 --- a/kernel/audit.h +++ b/kernel/audit.h @@ -103,21 +103,27 @@ extern struct mutex audit_filter_mutex; extern void audit_free_rule_rcu(struct rcu_head *); extern struct list_head audit_filter_list[]; +extern struct audit_entry *audit_dupe_rule(struct audit_krule *old); + /* audit watch functions */ -extern unsigned long audit_watch_inode(struct audit_watch *watch); -extern dev_t audit_watch_dev(struct audit_watch *watch); +#ifdef CONFIG_AUDIT_WATCH extern void audit_put_watch(struct audit_watch *watch); extern void audit_get_watch(struct audit_watch *watch); extern int audit_to_watch(struct audit_krule *krule, char *path, int len, u32 op); -extern int audit_add_watch(struct audit_krule *krule); -extern void audit_remove_watch(struct audit_watch *watch); -extern void audit_remove_watch_rule(struct audit_krule *krule, struct list_head *list); -extern void audit_inotify_unregister(struct list_head *in_list); +extern int audit_add_watch(struct audit_krule *krule, struct list_head **list); +extern void audit_remove_watch_rule(struct audit_krule *krule); extern char *audit_watch_path(struct audit_watch *watch); -extern struct list_head *audit_watch_rules(struct audit_watch *watch); - -extern struct audit_entry *audit_dupe_rule(struct audit_krule *old, - struct audit_watch *watch); +extern int audit_watch_compare(struct audit_watch *watch, unsigned long ino, dev_t dev); +#else +#define audit_put_watch(w) {} +#define audit_get_watch(w) {} +#define audit_to_watch(k, p, l, o) (-EINVAL) +#define audit_add_watch(k, l) (-EINVAL) +#define audit_remove_watch_rule(k) BUG() +#define audit_watch_path(w) "" +#define audit_watch_compare(w, i, d) 0 + +#endif /* CONFIG_AUDIT_WATCH */ #ifdef CONFIG_AUDIT_TREE extern struct audit_chunk *audit_tree_lookup(const struct inode *); diff --git a/kernel/audit_tree.c b/kernel/audit_tree.c index 2451dc6f3282..d9aab73d90e2 100644 --- a/kernel/audit_tree.c +++ b/kernel/audit_tree.c @@ -1,5 +1,5 @@ #include "audit.h" -#include <linux/inotify.h> +#include <linux/fsnotify_backend.h> #include <linux/namei.h> #include <linux/mount.h> #include <linux/kthread.h> @@ -21,7 +21,7 @@ struct audit_tree { struct audit_chunk { struct list_head hash; - struct inotify_watch watch; + struct fsnotify_mark_entry mark; struct list_head trees; /* with root here */ int dead; int count; @@ -67,7 +67,7 @@ static LIST_HEAD(prune_list); * that makes a difference. Some. */ -static struct inotify_handle *rtree_ih; +static struct fsnotify_group *audit_tree_group; static struct audit_tree *alloc_tree(const char *s) { @@ -110,29 +110,6 @@ const char *audit_tree_path(struct audit_tree *tree) return tree->pathname; } -static struct audit_chunk *alloc_chunk(int count) -{ - struct audit_chunk *chunk; - size_t size; - int i; - - size = offsetof(struct audit_chunk, owners) + count * sizeof(struct node); - chunk = kzalloc(size, GFP_KERNEL); - if (!chunk) - return NULL; - - INIT_LIST_HEAD(&chunk->hash); - INIT_LIST_HEAD(&chunk->trees); - chunk->count = count; - atomic_long_set(&chunk->refs, 1); - for (i = 0; i < count; i++) { - INIT_LIST_HEAD(&chunk->owners[i].list); - chunk->owners[i].index = i; - } - inotify_init_watch(&chunk->watch); - return chunk; -} - static void free_chunk(struct audit_chunk *chunk) { int i; @@ -156,6 +133,35 @@ static void __put_chunk(struct rcu_head *rcu) audit_put_chunk(chunk); } +static void audit_tree_destroy_watch(struct fsnotify_mark_entry *entry) +{ + struct audit_chunk *chunk = container_of(entry, struct audit_chunk, mark); + call_rcu(&chunk->head, __put_chunk); +} + +static struct audit_chunk *alloc_chunk(int count) +{ + struct audit_chunk *chunk; + size_t size; + int i; + + size = offsetof(struct audit_chunk, owners) + count * sizeof(struct node); + chunk = kzalloc(size, GFP_KERNEL); + if (!chunk) + return NULL; + + INIT_LIST_HEAD(&chunk->hash); + INIT_LIST_HEAD(&chunk->trees); + chunk->count = count; + atomic_long_set(&chunk->refs, 1); + for (i = 0; i < count; i++) { + INIT_LIST_HEAD(&chunk->owners[i].list); + chunk->owners[i].index = i; + } + fsnotify_init_mark(&chunk->mark, audit_tree_destroy_watch); + return chunk; +} + enum {HASH_SIZE = 128}; static struct list_head chunk_hash_heads[HASH_SIZE]; static __cacheline_aligned_in_smp DEFINE_SPINLOCK(hash_lock); @@ -166,10 +172,15 @@ static inline struct list_head *chunk_hash(const struct inode *inode) return chunk_hash_heads + n % HASH_SIZE; } -/* hash_lock is held by caller */ +/* hash_lock & entry->lock is held by caller */ static void insert_hash(struct audit_chunk *chunk) { - struct list_head *list = chunk_hash(chunk->watch.inode); + struct fsnotify_mark_entry *entry = &chunk->mark; + struct list_head *list; + + if (!entry->inode) + return; + list = chunk_hash(entry->inode); list_add_rcu(&chunk->hash, list); } @@ -180,7 +191,8 @@ struct audit_chunk *audit_tree_lookup(const struct inode *inode) struct audit_chunk *p; list_for_each_entry_rcu(p, list, hash) { - if (p->watch.inode == inode) { + /* mark.inode may have gone NULL, but who cares? */ + if (p->mark.inode == inode) { atomic_long_inc(&p->refs); return p; } @@ -209,38 +221,19 @@ static struct audit_chunk *find_chunk(struct node *p) static void untag_chunk(struct node *p) { struct audit_chunk *chunk = find_chunk(p); + struct fsnotify_mark_entry *entry = &chunk->mark; struct audit_chunk *new; struct audit_tree *owner; int size = chunk->count - 1; int i, j; - if (!pin_inotify_watch(&chunk->watch)) { - /* - * Filesystem is shutting down; all watches are getting - * evicted, just take it off the node list for this - * tree and let the eviction logics take care of the - * rest. - */ - owner = p->owner; - if (owner->root == chunk) { - list_del_init(&owner->same_root); - owner->root = NULL; - } - list_del_init(&p->list); - p->owner = NULL; - put_tree(owner); - return; - } + fsnotify_get_mark(entry); spin_unlock(&hash_lock); - /* - * pin_inotify_watch() succeeded, so the watch won't go away - * from under us. - */ - mutex_lock(&chunk->watch.inode->inotify_mutex); - if (chunk->dead) { - mutex_unlock(&chunk->watch.inode->inotify_mutex); + spin_lock(&entry->lock); + if (chunk->dead || !entry->inode) { + spin_unlock(&entry->lock); goto out; } @@ -255,16 +248,17 @@ static void untag_chunk(struct node *p) list_del_init(&p->list); list_del_rcu(&chunk->hash); spin_unlock(&hash_lock); - inotify_evict_watch(&chunk->watch); - mutex_unlock(&chunk->watch.inode->inotify_mutex); - put_inotify_watch(&chunk->watch); + spin_unlock(&entry->lock); + fsnotify_destroy_mark_by_entry(entry); + fsnotify_put_mark(entry); goto out; } new = alloc_chunk(size); if (!new) goto Fallback; - if (inotify_clone_watch(&chunk->watch, &new->watch) < 0) { + fsnotify_duplicate_mark(&new->mark, entry); + if (fsnotify_add_mark(&new->mark, new->mark.group, new->mark.inode, 1)) { free_chunk(new); goto Fallback; } @@ -297,9 +291,9 @@ static void untag_chunk(struct node *p) list_for_each_entry(owner, &new->trees, same_root) owner->root = new; spin_unlock(&hash_lock); - inotify_evict_watch(&chunk->watch); - mutex_unlock(&chunk->watch.inode->inotify_mutex); - put_inotify_watch(&chunk->watch); + spin_unlock(&entry->lock); + fsnotify_destroy_mark_by_entry(entry); + fsnotify_put_mark(entry); goto out; Fallback: @@ -313,31 +307,33 @@ Fallback: p->owner = NULL; put_tree(owner); spin_unlock(&hash_lock); - mutex_unlock(&chunk->watch.inode->inotify_mutex); + spin_unlock(&entry->lock); out: - unpin_inotify_watch(&chunk->watch); + fsnotify_put_mark(entry); spin_lock(&hash_lock); } static int create_chunk(struct inode *inode, struct audit_tree *tree) { + struct fsnotify_mark_entry *entry; struct audit_chunk *chunk = alloc_chunk(1); if (!chunk) return -ENOMEM; - if (inotify_add_watch(rtree_ih, &chunk->watch, inode, IN_IGNORED | IN_DELETE_SELF) < 0) { + entry = &chunk->mark; + if (fsnotify_add_mark(entry, audit_tree_group, inode, 0)) { free_chunk(chunk); return -ENOSPC; } - mutex_lock(&inode->inotify_mutex); + spin_lock(&entry->lock); spin_lock(&hash_lock); if (tree->goner) { spin_unlock(&hash_lock); chunk->dead = 1; - inotify_evict_watch(&chunk->watch); - mutex_unlock(&inode->inotify_mutex); - put_inotify_watch(&chunk->watch); + spin_unlock(&entry->lock); + fsnotify_destroy_mark_by_entry(entry); + fsnotify_put_mark(entry); return 0; } chunk->owners[0].index = (1U << 31); @@ -350,30 +346,33 @@ static int create_chunk(struct inode *inode, struct audit_tree *tree) } insert_hash(chunk); spin_unlock(&hash_lock); - mutex_unlock(&inode->inotify_mutex); + spin_unlock(&entry->lock); return 0; } /* the first tagged inode becomes root of tree */ static int tag_chunk(struct inode *inode, struct audit_tree *tree) { - struct inotify_watch *watch; + struct fsnotify_mark_entry *old_entry, *chunk_entry; struct audit_tree *owner; struct audit_chunk *chunk, *old; struct node *p; int n; - if (inotify_find_watch(rtree_ih, inode, &watch) < 0) + spin_lock(&inode->i_lock); + old_entry = fsnotify_find_mark_entry(audit_tree_group, inode); + spin_unlock(&inode->i_lock); + if (!old_entry) return create_chunk(inode, tree); - old = container_of(watch, struct audit_chunk, watch); + old = container_of(old_entry, struct audit_chunk, mark); /* are we already there? */ spin_lock(&hash_lock); for (n = 0; n < old->count; n++) { if (old->owners[n].owner == tree) { spin_unlock(&hash_lock); - put_inotify_watch(watch); + fsnotify_put_mark(old_entry); return 0; } } @@ -382,22 +381,40 @@ static int tag_chunk(struct inode *inode, struct audit_tree *tree) chunk = alloc_chunk(old->count + 1); if (!chunk) return -ENOMEM; + chunk_entry = &chunk->mark; - mutex_lock(&inode->inotify_mutex); - if (inotify_clone_watch(&old->watch, &chunk->watch) < 0) { - mutex_unlock(&inode->inotify_mutex); - put_inotify_watch(&old->watch); + spin_lock(&old_entry->lock); + if (!old_entry->inode) { + /* old_entry is being shot, lets just lie */ + spin_unlock(&old_entry->lock); + fsnotify_put_mark(old_entry); free_chunk(chunk); + return -ENOENT; + } + + fsnotify_duplicate_mark(chunk_entry, old_entry); + if (fsnotify_add_mark(chunk_entry, chunk_entry->group, chunk_entry->inode, 1)) { + spin_unlock(&old_entry->lock); + free_chunk(chunk); + fsnotify_put_mark(old_entry); return -ENOSPC; } + + /* even though we hold old_entry->lock, this is safe since chunk_entry->lock could NEVER have been grabbed before */ + spin_lock(&chunk_entry->lock); spin_lock(&hash_lock); + + /* we now hold old_entry->lock, chunk_entry->lock, and hash_lock */ if (tree->goner) { spin_unlock(&hash_lock); chunk->dead = 1; - inotify_evict_watch(&chunk->watch); - mutex_unlock(&inode->inotify_mutex); - put_inotify_watch(&old->watch); - put_inotify_watch(&chunk->watch); + spin_unlock(&chunk_entry->lock); + spin_unlock(&old_entry->lock); + + fsnotify_destroy_mark_by_entry(chunk_entry); + + fsnotify_put_mark(chunk_entry); + fsnotify_put_mark(old_entry); return 0; } list_replace_init(&old->trees, &chunk->trees); @@ -423,9 +440,10 @@ static int tag_chunk(struct inode *inode, struct audit_tree *tree) list_add(&tree->same_root, &chunk->trees); } spin_unlock(&hash_lock); - inotify_evict_watch(&old->watch); - mutex_unlock(&inode->inotify_mutex); - put_inotify_watch(&old->watch); + spin_unlock(&chunk_entry->lock); + spin_unlock(&old_entry->lock); + fsnotify_destroy_mark_by_entry(old_entry); + fsnotify_put_mark(old_entry); return 0; } @@ -578,7 +596,8 @@ void audit_trim_trees(void) spin_lock(&hash_lock); list_for_each_entry(node, &tree->chunks, list) { struct audit_chunk *chunk = find_chunk(node); - struct inode *inode = chunk->watch.inode; + /* this could be NULL if the watch is dieing else where... */ + struct inode *inode = chunk->mark.inode; struct vfsmount *mnt; node->index |= 1U<<31; list_for_each_entry(mnt, &list, mnt_list) { @@ -925,34 +944,40 @@ static void evict_chunk(struct audit_chunk *chunk) mutex_unlock(&audit_filter_mutex); } -static void handle_event(struct inotify_watch *watch, u32 wd, u32 mask, - u32 cookie, const char *dname, struct inode *inode) +static int audit_tree_handle_event(struct fsnotify_group *group, struct fsnotify_event *event) { - struct audit_chunk *chunk = container_of(watch, struct audit_chunk, watch); + BUG(); + return -EOPNOTSUPP; +} - if (mask & IN_IGNORED) { - evict_chunk(chunk); - put_inotify_watch(watch); - } +static void audit_tree_freeing_mark(struct fsnotify_mark_entry *entry, struct fsnotify_group *group) +{ + struct audit_chunk *chunk = container_of(entry, struct audit_chunk, mark); + + evict_chunk(chunk); + fsnotify_put_mark(entry); } -static void destroy_watch(struct inotify_watch *watch) +static bool audit_tree_send_event(struct fsnotify_group *group, struct inode *inode, + __u32 mask, void *data, int data_type) { - struct audit_chunk *chunk = container_of(watch, struct audit_chunk, watch); - call_rcu(&chunk->head, __put_chunk); + return 0; } -static const struct inotify_operations rtree_inotify_ops = { - .handle_event = handle_event, - .destroy_watch = destroy_watch, +static const struct fsnotify_ops audit_tree_ops = { + .handle_event = audit_tree_handle_event, + .should_send_event = audit_tree_send_event, + .free_group_priv = NULL, + .free_event_priv = NULL, + .freeing_mark = audit_tree_freeing_mark, }; static int __init audit_tree_init(void) { int i; - rtree_ih = inotify_init(&rtree_inotify_ops); - if (IS_ERR(rtree_ih)) + audit_tree_group = fsnotify_obtain_group(0, &audit_tree_ops); + if (IS_ERR(audit_tree_group)) audit_panic("cannot initialize inotify handle for rectree watches"); for (i = 0; i < HASH_SIZE; i++) diff --git a/kernel/audit_watch.c b/kernel/audit_watch.c index cc7e87936cbc..10ec0e21d0ec 100644 --- a/kernel/audit_watch.c +++ b/kernel/audit_watch.c @@ -24,17 +24,17 @@ #include <linux/kthread.h> #include <linux/mutex.h> #include <linux/fs.h> +#include <linux/fsnotify_backend.h> #include <linux/namei.h> #include <linux/netlink.h> #include <linux/sched.h> -#include <linux/inotify.h> #include <linux/security.h> #include "audit.h" /* * Reference counting: * - * audit_parent: lifetime is from audit_init_parent() to receipt of an IN_IGNORED + * audit_parent: lifetime is from audit_init_parent() to receipt of an FS_IGNORED * event. Each audit_watch holds a reference to its associated parent. * * audit_watch: if added to lists, lifetime is from audit_init_watch() to @@ -50,40 +50,64 @@ struct audit_watch { unsigned long ino; /* associated inode number */ struct audit_parent *parent; /* associated parent */ struct list_head wlist; /* entry in parent->watches list */ - struct list_head rules; /* associated rules */ + struct list_head rules; /* anchor for krule->rlist */ }; struct audit_parent { - struct list_head ilist; /* entry in inotify registration list */ - struct list_head watches; /* associated watches */ - struct inotify_watch wdata; /* inotify watch data */ - unsigned flags; /* status flags */ + struct list_head watches; /* anchor for audit_watch->wlist */ + struct fsnotify_mark_entry mark; /* fsnotify mark on the inode */ }; -/* Inotify handle. */ -struct inotify_handle *audit_ih; +/* fsnotify handle. */ +struct fsnotify_group *audit_watch_group; -/* - * audit_parent status flags: - * - * AUDIT_PARENT_INVALID - set anytime rules/watches are auto-removed due to - * a filesystem event to ensure we're adding audit watches to a valid parent. - * Technically not needed for IN_DELETE_SELF or IN_UNMOUNT events, as we cannot - * receive them while we have nameidata, but must be used for IN_MOVE_SELF which - * we can receive while holding nameidata. - */ -#define AUDIT_PARENT_INVALID 0x001 +/* fsnotify events we care about. */ +#define AUDIT_FS_WATCH (FS_MOVE | FS_CREATE | FS_DELETE | FS_DELETE_SELF |\ + FS_MOVE_SELF | FS_EVENT_ON_CHILD) -/* Inotify events we care about. */ -#define AUDIT_IN_WATCH IN_MOVE|IN_CREATE|IN_DELETE|IN_DELETE_SELF|IN_MOVE_SELF +static void audit_free_parent(struct audit_parent *parent) +{ + WARN_ON(!list_empty(&parent->watches)); + kfree(parent); +} -static void audit_free_parent(struct inotify_watch *i_watch) +static void audit_watch_free_mark(struct fsnotify_mark_entry *entry) { struct audit_parent *parent; - parent = container_of(i_watch, struct audit_parent, wdata); - WARN_ON(!list_empty(&parent->watches)); - kfree(parent); + parent = container_of(entry, struct audit_parent, mark); + audit_free_parent(parent); +} + +static void audit_get_parent(struct audit_parent *parent) +{ + if (likely(parent)) + fsnotify_get_mark(&parent->mark); +} + +static void audit_put_parent(struct audit_parent *parent) +{ + if (likely(parent)) + fsnotify_put_mark(&parent->mark); +} + +/* + * Find and return the audit_parent on the given inode. If found a reference + * is taken on this parent. + */ +static inline struct audit_parent *audit_find_parent(struct inode *inode) +{ + struct audit_parent *parent = NULL; + struct fsnotify_mark_entry *entry; + + spin_lock(&inode->i_lock); + entry = fsnotify_find_mark_entry(audit_watch_group, inode); + spin_unlock(&inode->i_lock); + + if (entry) + parent = container_of(entry, struct audit_parent, mark); + + return parent; } void audit_get_watch(struct audit_watch *watch) @@ -104,7 +128,7 @@ void audit_put_watch(struct audit_watch *watch) void audit_remove_watch(struct audit_watch *watch) { list_del(&watch->wlist); - put_inotify_watch(&watch->parent->wdata); + audit_put_parent(watch->parent); watch->parent = NULL; audit_put_watch(watch); /* match initial get */ } @@ -114,42 +138,32 @@ char *audit_watch_path(struct audit_watch *watch) return watch->path; } -struct list_head *audit_watch_rules(struct audit_watch *watch) -{ - return &watch->rules; -} - -unsigned long audit_watch_inode(struct audit_watch *watch) -{ - return watch->ino; -} - -dev_t audit_watch_dev(struct audit_watch *watch) +int audit_watch_compare(struct audit_watch *watch, unsigned long ino, dev_t dev) { - return watch->dev; + return (watch->ino != (unsigned long)-1) && + (watch->ino == ino) && + (watch->dev == dev); } /* Initialize a parent watch entry. */ static struct audit_parent *audit_init_parent(struct nameidata *ndp) { + struct inode *inode = ndp->path.dentry->d_inode; struct audit_parent *parent; - s32 wd; + int ret; parent = kzalloc(sizeof(*parent), GFP_KERNEL); if (unlikely(!parent)) return ERR_PTR(-ENOMEM); INIT_LIST_HEAD(&parent->watches); - parent->flags = 0; - - inotify_init_watch(&parent->wdata); - /* grab a ref so inotify watch hangs around until we take audit_filter_mutex */ - get_inotify_watch(&parent->wdata); - wd = inotify_add_watch(audit_ih, &parent->wdata, - ndp->path.dentry->d_inode, AUDIT_IN_WATCH); - if (wd < 0) { - audit_free_parent(&parent->wdata); - return ERR_PTR(wd); + + fsnotify_init_mark(&parent->mark, audit_watch_free_mark); + parent->mark.mask = AUDIT_FS_WATCH; + ret = fsnotify_add_mark(&parent->mark, audit_watch_group, inode, 0); + if (ret < 0) { + audit_free_parent(parent); + return ERR_PTR(ret); } return parent; @@ -178,7 +192,7 @@ int audit_to_watch(struct audit_krule *krule, char *path, int len, u32 op) { struct audit_watch *watch; - if (!audit_ih) + if (!audit_watch_group) return -EOPNOTSUPP; if (path[0] != '/' || path[len-1] == '/' || @@ -216,7 +230,7 @@ static struct audit_watch *audit_dupe_watch(struct audit_watch *old) new->dev = old->dev; new->ino = old->ino; - get_inotify_watch(&old->parent->wdata); + audit_get_parent(old->parent); new->parent = old->parent; out: @@ -250,15 +264,19 @@ static void audit_update_watch(struct audit_parent *parent, struct audit_entry *oentry, *nentry; mutex_lock(&audit_filter_mutex); + /* Run all of the watches on this parent looking for the one that + * matches the given dname */ list_for_each_entry_safe(owatch, nextw, &parent->watches, wlist) { if (audit_compare_dname_path(dname, owatch->path, NULL)) continue; /* If the update involves invalidating rules, do the inode-based * filtering now, so we don't omit records. */ - if (invalidating && current->audit_context) + if (invalidating && !audit_dummy_context()) audit_filter_inodes(current, current->audit_context); + /* updating ino will likely change which audit_hash_list we + * are on so we need a new watch for the new list */ nwatch = audit_dupe_watch(owatch); if (IS_ERR(nwatch)) { mutex_unlock(&audit_filter_mutex); @@ -274,12 +292,21 @@ static void audit_update_watch(struct audit_parent *parent, list_del(&oentry->rule.rlist); list_del_rcu(&oentry->list); - nentry = audit_dupe_rule(&oentry->rule, nwatch); + nentry = audit_dupe_rule(&oentry->rule); if (IS_ERR(nentry)) { list_del(&oentry->rule.list); audit_panic("error updating watch, removing"); } else { int h = audit_hash_ino((u32)ino); + + /* + * nentry->rule.watch == oentry->rule.watch so + * we must drop that reference and set it to our + * new watch. + */ + audit_put_watch(nentry->rule.watch); + audit_get_watch(nwatch); + nentry->rule.watch = nwatch; list_add(&nentry->rule.rlist, &nwatch->rules); list_add_rcu(&nentry->list, &audit_inode_hash[h]); list_replace(&oentry->rule.list, @@ -311,7 +338,6 @@ static void audit_remove_parent_watches(struct audit_parent *parent) struct audit_entry *e; mutex_lock(&audit_filter_mutex); - parent->flags |= AUDIT_PARENT_INVALID; list_for_each_entry_safe(w, nextw, &parent->watches, wlist) { list_for_each_entry_safe(r, nextr, &w->rules, rlist) { e = container_of(r, struct audit_entry, rule); @@ -324,20 +350,8 @@ static void audit_remove_parent_watches(struct audit_parent *parent) audit_remove_watch(w); } mutex_unlock(&audit_filter_mutex); -} - -/* Unregister inotify watches for parents on in_list. - * Generates an IN_IGNORED event. */ -void audit_inotify_unregister(struct list_head *in_list) -{ - struct audit_parent *p, *n; - list_for_each_entry_safe(p, n, in_list, ilist) { - list_del(&p->ilist); - inotify_rm_watch(audit_ih, &p->wdata); - /* the unpin matching the pin in audit_do_del_rule() */ - unpin_inotify_watch(&p->wdata); - } + fsnotify_destroy_mark_by_entry(&parent->mark); } /* Get path information necessary for adding watches. */ @@ -388,7 +402,7 @@ static void audit_put_nd(struct nameidata *ndp, struct nameidata *ndw) } } -/* Associate the given rule with an existing parent inotify_watch. +/* Associate the given rule with an existing parent. * Caller must hold audit_filter_mutex. */ static void audit_add_to_parent(struct audit_krule *krule, struct audit_parent *parent) @@ -396,6 +410,8 @@ static void audit_add_to_parent(struct audit_krule *krule, struct audit_watch *w, *watch = krule->watch; int watch_found = 0; + BUG_ON(!mutex_is_locked(&audit_filter_mutex)); + list_for_each_entry(w, &parent->watches, wlist) { if (strcmp(watch->path, w->path)) continue; @@ -412,7 +428,7 @@ static void audit_add_to_parent(struct audit_krule *krule, } if (!watch_found) { - get_inotify_watch(&parent->wdata); + audit_get_parent(parent); watch->parent = parent; list_add(&watch->wlist, &parent->watches); @@ -422,13 +438,12 @@ static void audit_add_to_parent(struct audit_krule *krule, /* Find a matching watch entry, or add this one. * Caller must hold audit_filter_mutex. */ -int audit_add_watch(struct audit_krule *krule) +int audit_add_watch(struct audit_krule *krule, struct list_head **list) { struct audit_watch *watch = krule->watch; - struct inotify_watch *i_watch; struct audit_parent *parent; struct nameidata *ndp = NULL, *ndw = NULL; - int ret = 0; + int h, ret = 0; mutex_unlock(&audit_filter_mutex); @@ -440,47 +455,38 @@ int audit_add_watch(struct audit_krule *krule) goto error; } + mutex_lock(&audit_filter_mutex); + /* update watch filter fields */ if (ndw) { watch->dev = ndw->path.dentry->d_inode->i_sb->s_dev; watch->ino = ndw->path.dentry->d_inode->i_ino; } - /* The audit_filter_mutex must not be held during inotify calls because - * we hold it during inotify event callback processing. If an existing - * inotify watch is found, inotify_find_watch() grabs a reference before - * returning. - */ - if (inotify_find_watch(audit_ih, ndp->path.dentry->d_inode, - &i_watch) < 0) { + /* either find an old parent or attach a new one */ + parent = audit_find_parent(ndp->path.dentry->d_inode); + if (!parent) { parent = audit_init_parent(ndp); if (IS_ERR(parent)) { - /* caller expects mutex locked */ - mutex_lock(&audit_filter_mutex); ret = PTR_ERR(parent); goto error; } - } else - parent = container_of(i_watch, struct audit_parent, wdata); - - mutex_lock(&audit_filter_mutex); + } - /* parent was moved before we took audit_filter_mutex */ - if (parent->flags & AUDIT_PARENT_INVALID) - ret = -ENOENT; - else - audit_add_to_parent(krule, parent); + audit_add_to_parent(krule, parent); - /* match get in audit_init_parent or inotify_find_watch */ - put_inotify_watch(&parent->wdata); + /* match get in audit_find_parent or audit_init_parent */ + audit_put_parent(parent); + h = audit_hash_ino((u32)watch->ino); + *list = &audit_inode_hash[h]; error: audit_put_nd(ndp, ndw); /* NULL args OK */ return ret; } -void audit_remove_watch_rule(struct audit_krule *krule, struct list_head *list) +void audit_remove_watch_rule(struct audit_krule *krule) { struct audit_watch *watch = krule->watch; struct audit_parent *parent = watch->parent; @@ -491,53 +497,91 @@ void audit_remove_watch_rule(struct audit_krule *krule, struct list_head *list) audit_remove_watch(watch); if (list_empty(&parent->watches)) { - /* Put parent on the inotify un-registration - * list. Grab a reference before releasing - * audit_filter_mutex, to be released in - * audit_inotify_unregister(). - * If filesystem is going away, just leave - * the sucker alone, eviction will take - * care of it. */ - if (pin_inotify_watch(&parent->wdata)) - list_add(&parent->ilist, list); + audit_get_parent(parent); + fsnotify_destroy_mark_by_entry(&parent->mark); + audit_put_parent(parent); } } } -/* Update watch data in audit rules based on inotify events. */ -static void audit_handle_ievent(struct inotify_watch *i_watch, u32 wd, u32 mask, - u32 cookie, const char *dname, struct inode *inode) +static bool audit_watch_should_send_event(struct fsnotify_group *group, struct inode *inode, + __u32 mask, void *data, int data_type) +{ + struct fsnotify_mark_entry *entry; + bool send; + + spin_lock(&inode->i_lock); + entry = fsnotify_find_mark_entry(group, inode); + spin_unlock(&inode->i_lock); + if (!entry) + return false; + + mask = (mask & ~FS_EVENT_ON_CHILD); + send = (entry->mask & mask); + + /* find took a reference */ + fsnotify_put_mark(entry); + + return send; +} + +/* Update watch data in audit rules based on fsnotify events. */ +static int audit_watch_handle_event(struct fsnotify_group *group, struct fsnotify_event *event) { + struct inode *inode; + __u32 mask = event->mask; + const char *dname = event->file_name; struct audit_parent *parent; - parent = container_of(i_watch, struct audit_parent, wdata); + BUG_ON(group != audit_watch_group); - if (mask & (IN_CREATE|IN_MOVED_TO) && inode) - audit_update_watch(parent, dname, inode->i_sb->s_dev, - inode->i_ino, 0); - else if (mask & (IN_DELETE|IN_MOVED_FROM)) + parent = audit_find_parent(event->to_tell); + if (unlikely(!parent)) + return 0; + + switch (event->data_type) { + case (FSNOTIFY_EVENT_PATH): + inode = event->path.dentry->d_inode; + break; + case (FSNOTIFY_EVENT_INODE): + inode = event->inode; + break; + default: + BUG(); + inode = NULL; + break; + }; + + if (mask & (FS_CREATE|FS_MOVED_TO) && inode) + audit_update_watch(parent, dname, inode->i_sb->s_dev, inode->i_ino, 0); + else if (mask & (FS_DELETE|FS_MOVED_FROM)) audit_update_watch(parent, dname, (dev_t)-1, (unsigned long)-1, 1); - /* inotify automatically removes the watch and sends IN_IGNORED */ - else if (mask & (IN_DELETE_SELF|IN_UNMOUNT)) + else if (mask & (FS_DELETE_SELF|FS_UNMOUNT|FS_MOVE_SELF)) audit_remove_parent_watches(parent); - /* inotify does not remove the watch, so remove it manually */ - else if(mask & IN_MOVE_SELF) { - audit_remove_parent_watches(parent); - inotify_remove_watch_locked(audit_ih, i_watch); - } else if (mask & IN_IGNORED) - put_inotify_watch(i_watch); + /* moved put_inotify_watch to freeing mark */ + + /* matched the ref taken by audit_find_parent */ + audit_put_parent(parent); + + return 0; } -static const struct inotify_operations audit_inotify_ops = { - .handle_event = audit_handle_ievent, - .destroy_watch = audit_free_parent, +static const struct fsnotify_ops audit_watch_fsnotify_ops = { + .should_send_event = audit_watch_should_send_event, + .handle_event = audit_watch_handle_event, + .free_group_priv = NULL, + .freeing_mark = NULL, + .free_event_priv = NULL, }; static int __init audit_watch_init(void) { - audit_ih = inotify_init(&audit_inotify_ops); - if (IS_ERR(audit_ih)) - audit_panic("cannot initialize inotify handle"); + audit_watch_group = fsnotify_obtain_group(AUDIT_FS_WATCH, + &audit_watch_fsnotify_ops); + if (IS_ERR(audit_watch_group)) { + audit_watch_group = NULL; + audit_panic("cannot create audit fsnotify group"); + } return 0; } -subsys_initcall(audit_watch_init); +device_initcall(audit_watch_init); diff --git a/kernel/auditfilter.c b/kernel/auditfilter.c index a70604047f3c..f5e4cae5ad82 100644 --- a/kernel/auditfilter.c +++ b/kernel/auditfilter.c @@ -70,6 +70,7 @@ static inline void audit_free_rule(struct audit_entry *e) { int i; struct audit_krule *erule = &e->rule; + /* some rules don't have associated watches */ if (erule->watch) audit_put_watch(erule->watch); @@ -745,8 +746,7 @@ static inline int audit_dupe_lsm_field(struct audit_field *df, * rule with the new rule in the filterlist, then free the old rule. * The rlist element is undefined; list manipulations are handled apart from * the initial copy. */ -struct audit_entry *audit_dupe_rule(struct audit_krule *old, - struct audit_watch *watch) +struct audit_entry *audit_dupe_rule(struct audit_krule *old) { u32 fcount = old->field_count; struct audit_entry *entry; @@ -768,8 +768,8 @@ struct audit_entry *audit_dupe_rule(struct audit_krule *old, new->prio = old->prio; new->buflen = old->buflen; new->inode_f = old->inode_f; - new->watch = NULL; new->field_count = old->field_count; + /* * note that we are OK with not refcounting here; audit_match_tree() * never dereferences tree and we can't get false positives there @@ -810,9 +810,9 @@ struct audit_entry *audit_dupe_rule(struct audit_krule *old, } } - if (watch) { - audit_get_watch(watch); - new->watch = watch; + if (old->watch) { + audit_get_watch(old->watch); + new->watch = old->watch; } return entry; @@ -865,7 +865,7 @@ static inline int audit_add_rule(struct audit_entry *entry) struct audit_watch *watch = entry->rule.watch; struct audit_tree *tree = entry->rule.tree; struct list_head *list; - int h, err; + int err; #ifdef CONFIG_AUDITSYSCALL int dont_count = 0; @@ -888,15 +888,11 @@ static inline int audit_add_rule(struct audit_entry *entry) if (watch) { /* audit_filter_mutex is dropped and re-taken during this call */ - err = audit_add_watch(&entry->rule); + err = audit_add_watch(&entry->rule, &list); if (err) { mutex_unlock(&audit_filter_mutex); goto error; } - /* entry->rule.watch may have changed during audit_add_watch() */ - watch = entry->rule.watch; - h = audit_hash_ino((u32)audit_watch_inode(watch)); - list = &audit_inode_hash[h]; } if (tree) { err = audit_add_tree_rule(&entry->rule); @@ -948,7 +944,6 @@ static inline int audit_del_rule(struct audit_entry *entry) struct audit_watch *watch = entry->rule.watch; struct audit_tree *tree = entry->rule.tree; struct list_head *list; - LIST_HEAD(inotify_list); int ret = 0; #ifdef CONFIG_AUDITSYSCALL int dont_count = 0; @@ -968,7 +963,7 @@ static inline int audit_del_rule(struct audit_entry *entry) } if (e->rule.watch) - audit_remove_watch_rule(&e->rule, &inotify_list); + audit_remove_watch_rule(&e->rule); if (e->rule.tree) audit_remove_tree_rule(&e->rule); @@ -986,9 +981,6 @@ static inline int audit_del_rule(struct audit_entry *entry) #endif mutex_unlock(&audit_filter_mutex); - if (!list_empty(&inotify_list)) - audit_inotify_unregister(&inotify_list); - out: if (watch) audit_put_watch(watch); /* match initial get */ @@ -1322,30 +1314,23 @@ static int update_lsm_rule(struct audit_krule *r) { struct audit_entry *entry = container_of(r, struct audit_entry, rule); struct audit_entry *nentry; - struct audit_watch *watch; - struct audit_tree *tree; int err = 0; if (!security_audit_rule_known(r)) return 0; - watch = r->watch; - tree = r->tree; - nentry = audit_dupe_rule(r, watch); + nentry = audit_dupe_rule(r); if (IS_ERR(nentry)) { /* save the first error encountered for the * return value */ err = PTR_ERR(nentry); audit_panic("error updating LSM filters"); - if (watch) + if (r->watch) list_del(&r->rlist); list_del_rcu(&entry->list); list_del(&r->list); } else { - if (watch) { - list_add(&nentry->rule.rlist, audit_watch_rules(watch)); - list_del(&r->rlist); - } else if (tree) + if (r->watch || r->tree) list_replace_init(&r->rlist, &nentry->rule.rlist); list_replace_rcu(&entry->list, &nentry->list); list_replace(&r->list, &nentry->rule.list); diff --git a/kernel/auditsc.c b/kernel/auditsc.c index 267e484f0198..ef0410c3c355 100644 --- a/kernel/auditsc.c +++ b/kernel/auditsc.c @@ -549,9 +549,8 @@ static int audit_filter_rules(struct task_struct *tsk, } break; case AUDIT_WATCH: - if (name && audit_watch_inode(rule->watch) != (unsigned long)-1) - result = (name->dev == audit_watch_dev(rule->watch) && - name->ino == audit_watch_inode(rule->watch)); + if (name) + result = audit_watch_compare(rule->watch, name->ino, name->dev); break; case AUDIT_DIR: if (ctx) @@ -1726,7 +1725,7 @@ static inline void handle_one(const struct inode *inode) struct audit_tree_refs *p; struct audit_chunk *chunk; int count; - if (likely(list_empty(&inode->inotify_watches))) + if (likely(hlist_empty(&inode->i_fsnotify_mark_entries))) return; context = current->audit_context; p = context->trees; @@ -1769,7 +1768,7 @@ retry: seq = read_seqbegin(&rename_lock); for(;;) { struct inode *inode = d->d_inode; - if (inode && unlikely(!list_empty(&inode->inotify_watches))) { + if (inode && unlikely(!hlist_empty(&inode->i_fsnotify_mark_entries))) { struct audit_chunk *chunk; chunk = audit_tree_lookup(inode); if (chunk) { |