Re: splinter work on generic fs notification framework

On Tue, Jul 12, 2005 at 02:18:20PM -0500, Timothy R. Chavez wrote: > Hi, > > I've CCed some folks that I thought might be interested in this update and may > not be on linux-audit. Unfortunatly linux-audit is a closed list, so this post will probably bounce :(

> Based on recent discussion on LKML and some IRC conversations, a generic file > system notification framework is being written to merge common functionality > between the "auditfs" component of the audit framework and Inotify. This > framework will be mostly comprised of the watch logic implemented by the > "auditfs" component (I_AUDIT, hash table, update hook, helper functions, etc) > and the fsnotify hooks written by Robert Love. User's of this framework will > pass their own callback functions in when adding a watch via this framework > which will effectively be stored on the watch and called from the watch hook > that discovers it. > > We introduce fs/watch.c and include/linux/watch.h fswatch.h perhaps?

> diff --exclude=.git -Nurp audit-2.6/include/linux/watch.h audit-2.6.git-fsnotify/include/linux/watch.h > --- audit-2.6/include/linux/watch.h 1969-12-31 18:00:00.000000000 -0600 > +++ audit-2.6.git-fsnotify/include/linux/watch.h 2005-07-11 01:33:45.000000000 -0500 > @@ -0,0 +1,57 @@ > +#ifndef _WATCH_H > +#define _WATCH_H > + > +struct watch { > + atomic_t w_count; Use a struct kref instead of rolling your own here.

> + struct hlist_node w_node; /* per-directory list */ > + struct hlist_node w_master; /* Master watch list */ > + struct hlist_node w_watched; /* Watches on inode */ > + dev_t w_dev; /* Superblock device */ > + __u32 w_perms; /* Permissions filtering */ > + char *w_name; /* Watch beneath parent */ > + char *w_path; /* Insertion path */ > + char *w_filterkey; /* An arbitrary filtering key */ > + void (*w_func); /* Callback function */ > +}; > + > +struct watch_inode_data { > + int count; same here. > + struct watch_inode_data *next_hash; /* Watch data hash table */ formatting issue. > + struct inode *inode; /* Inode to which it belongs */ > + struct hlist_head watches; /* List of watches on inode */ > + struct hlist_head watchlist; /* Watches for children */ > +}; > + > +/* User */ > +extern int watch_add(struct watch *, struct inode *); > +extern int watch_rem(struct watch *, struct inode *); > +extern struct watch *watch_alloc(void); > +extern struct watch *watch_get(struct watch *); > +extern void watch_free(struct watch *); > +extern void watch_put(struct watch *); Why a difference? The last watch_put should free the structure, right?

> + > +/* Internals */ > + > +extern int watch_init(void); > +extern void watch_inode_free(struct inode *); > +extern void watch_update(struct dentry *, int); > + > +/* Hooks */ > +#ifdef CONFIG_FSWATCH > +extern const char *watch_notify_oldname_init(const char *); > +extern void watch_notify_oldname_free(const char *); > +extern void watch_notify_move(struct inode *, struct inode *, const char *, const char *, int); > +extern void watch_notify_unlink(struct dentry *, struct inode *); > +extern void watch_notify_rmdir(struct dentry *, struct inode *, struct inode *); > +extern void watch_notify_create(struct inode *, const char *); > +extern void watch_notify_mkdir(struct inode *, const char *); > +extern void watch_notify_access(struct dentry *dentry); > +extern void watch_notify_modify(struct dentry *dentry); > +extern void watch_notify_open(struct dentry *dentry, int mask); > +extern void watch_notify_close(struct file *file); > +extern void watch_notify_xattr(struct dentry *dentry); > +extern void watch_notify_change(struct dentry *dentry, unsigned int ia_valid); > +#endif /* CONFIG_FSWATCH */ No static inlines if !CONFIG_FSWATCH?

thanks, greg k-h