fs: provide rcu-walk aware permission i_ops

Signed-off-by: Nick Piggin <[e-mail]>

Documentation/filesystems/Locking

···
 	void * (*follow_link) (struct dentry *, struct nameidata *);
 	void (*put_link) (struct dentry *, struct nameidata *, void *);
 	void (*truncate) (struct inode *);
-	int (*permission) (struct inode *, int, struct nameidata *);
-	int (*check_acl)(struct inode *, int);
+	int (*permission) (struct inode *, int, unsigned int);
+	int (*check_acl)(struct inode *, int, unsigned int);
 	int (*setattr) (struct dentry *, struct iattr *);
 	int (*getattr) (struct vfsmount *, struct dentry *, struct kstat *);
 	int (*setxattr) (struct dentry *, const char *,const void *,size_t,int);
···
 put_link:	no
 truncate:	yes		(see below)
 setattr:	yes
-permission:	no
+permission:	no (may not block if called in rcu-walk mode)
 check_acl:	no
 getattr:	no
 setxattr:	yes

Documentation/filesystems/path-lookup.txt

···
 
 The cases where rcu-walk cannot continue are:
 * NULL dentry (ie. any uncached path element)
-* parent with d_inode->i_op->permission or ACLs
 * Following links
 
-In future patches, permission checks become rcu-walk aware. It may be possible
-eventually to make following links rcu-walk aware.
+It may be possible eventually to make following links rcu-walk aware.
 
 Uncached path elements will always require dropping to ref-walk mode, at the
 very least because i_mutex needs to be grabbed, and objects allocated.
···
 scalability of path resolution.
 
 
+Interesting statistics
+======================
+
+The following table gives rcu lookup statistics for a few simple workloads
+(2s12c24t Westmere, debian non-graphical system). Ungraceful are attempts to
+drop rcu that fail due to d_seq failure and requiring the entire path lookup
+again. Other cases are successful rcu-drops that are required before the final
+element, nodentry for missing dentry, revalidate for filesystem revalidate
+routine requiring rcu drop, permission for permission check requiring drop,
+and link for symlink traversal requiring drop.
+
+     rcu-lookups     restart  nodentry          link  revalidate  permission
+bootup     47121           0      4624          1010       10283        7852
+dbench  25386793           0   6778659(26.7%)     55         549        1156
+kbuild   2696672          10     64442(2.3%)  108764(4.0%)     1        1590
+git diff   39605           0        28             2           0         106
+vfstest 24185492        4945    708725(2.9%) 1076136(4.4%)     0        2651
+
+What this shows is that failed rcu-walk lookups, ie. ones that are restarted
+entirely with ref-walk, are quite rare. Even the "vfstest" case which
+specifically has concurrent renames/mkdir/rmdir/ creat/unlink/etc to excercise
+such races is not showing a huge amount of restarts.
+
+Dropping from rcu-walk to ref-walk mean that we have encountered a dentry where
+the reference count needs to be taken for some reason. This is either because
+we have reached the target of the path walk, or because we have encountered a
+condition that can't be resolved in rcu-walk mode.  Ideally, we drop rcu-walk
+only when we have reached the target dentry, so the other statistics show where
+this does not happen.
+
+Note that a graceful drop from rcu-walk mode due to something such as the
+dentry not existing (which can be common) is not necessarily a failure of
+rcu-walk scheme, because some elements of the path may have been walked in
+rcu-walk mode. The further we get from common path elements (such as cwd or
+root), the less contended the dentry is likely to be. The closer we are to
+common path elements, the more likely they will exist in dentry cache.
+
+
 Papers and other documentation on dcache locking
 ================================================
 
 1. Scaling dcache with RCU (http://linuxjournal.com/article.php?sid=7124).
 
 2. http://lse.sourceforge.net/locking/dcache/dcache.html
+
+

Documentation/filesystems/porting

···
 may now be called in rcu-walk mode (nd->flags & LOOKUP_RCU). -ECHILD should be
 returned if the filesystem cannot handle rcu-walk. See
 Documentation/filesystems/vfs.txt for more details.
+
+	permission and check_acl are inode permission checks that are called
+on many or all directory inodes on the way down a path walk (to check for
+exec permission). These must now be rcu-walk aware (flags & IPERM_RCU). See
+Documentation/filesystems/vfs.txt for more details.

Documentation/filesystems/vfs.txt

···
         void * (*follow_link) (struct dentry *, struct nameidata *);
         void (*put_link) (struct dentry *, struct nameidata *, void *);
 	void (*truncate) (struct inode *);
-	int (*permission) (struct inode *, int, struct nameidata *);
+	int (*permission) (struct inode *, int, unsigned int);
+	int (*check_acl)(struct inode *, int, unsigned int);
 	int (*setattr) (struct dentry *, struct iattr *);
 	int (*getattr) (struct vfsmount *mnt, struct dentry *, struct kstat *);
 	int (*setxattr) (struct dentry *, const char *,const void *,size_t,int);
···
 
   permission: called by the VFS to check for access rights on a POSIX-like
   	filesystem.
+
+	May be called in rcu-walk mode (flags & IPERM_RCU). If in rcu-walk
+	mode, the filesystem must check the permission without blocking or
+	storing to the inode.
+
+	If a situation is encountered that rcu-walk cannot handle, return
+	-ECHILD and it will be called again in ref-walk mode.
 
   setattr: called by the VFS to set attributes for a file. This method
   	is called by chmod(2) and related system calls.

drivers/staging/smbfs/file.c

···
  * privileges, so we need our own check for this.
  */
 static int
-smb_file_permission(struct inode *inode, int mask)
+smb_file_permission(struct inode *inode, int mask, unsigned int flags)
 {
 	int mode = inode->i_mode;
 	int error = 0;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
 	VERBOSE("mode=%x, mask=%x\n", mode, mask);
 

fs/9p/acl.c

···
 	return acl;
 }
 
-int v9fs_check_acl(struct inode *inode, int mask)
+int v9fs_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
 	struct posix_acl *acl;
 	struct v9fs_session_info *v9ses;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
 	v9ses = v9fs_inode2v9ses(inode);
 	if ((v9ses->flags & V9FS_ACCESS_MASK) != V9FS_ACCESS_CLIENT) {

fs/9p/acl.h

···
 
 #ifdef CONFIG_9P_FS_POSIX_ACL
 extern int v9fs_get_acl(struct inode *, struct p9_fid *);
-extern int v9fs_check_acl(struct inode *inode, int mask);
+extern int v9fs_check_acl(struct inode *inode, int mask, unsigned int flags);
 extern int v9fs_acl_chmod(struct dentry *);
 extern int v9fs_set_create_acl(struct dentry *,
 			       struct posix_acl *, struct posix_acl *);

fs/afs/internal.h

···
 extern void afs_cache_permit(struct afs_vnode *, struct key *, long);
 extern void afs_zap_permits(struct rcu_head *);
 extern struct key *afs_request_key(struct afs_cell *);
-extern int afs_permission(struct inode *, int);
+extern int afs_permission(struct inode *, int, unsigned int);
 
 /*
  * server.c

fs/afs/security.c

···
  * - AFS ACLs are attached to directories only, and a file is controlled by its
  *   parent directory's ACL
  */
-int afs_permission(struct inode *inode, int mask)
+int afs_permission(struct inode *inode, int mask, unsigned int flags)
 {
 	struct afs_vnode *vnode = AFS_FS_I(inode);
 	afs_access_t uninitialized_var(access);
 	struct key *key;
 	int ret;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
 	_enter("{{%x:%u},%lx},%x,",
 	       vnode->fid.vid, vnode->fid.vnode, vnode->flags, mask);
···
 	}
 
 	key_put(key);
-	ret = generic_permission(inode, mask, NULL);
+	ret = generic_permission(inode, mask, flags, NULL);
 	_leave(" = %d", ret);
 	return ret;
 

fs/bad_inode.c

···
 	return -EIO;
 }
 
-static int bad_inode_permission(struct inode *inode, int mask)
+static int bad_inode_permission(struct inode *inode, int mask, unsigned int flags)
 {
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	return -EIO;
 }
 

fs/btrfs/acl.c

···
 	return ret;
 }
 
-int btrfs_check_acl(struct inode *inode, int mask)
+int btrfs_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
 	struct posix_acl *acl;
 	int error = -EAGAIN;
 
-	acl = btrfs_get_acl(inode, ACL_TYPE_ACCESS);
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
+	acl = btrfs_get_acl(inode, ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))
 		return PTR_ERR(acl);
 	if (acl) {

fs/btrfs/ctree.h

···
 
 /* acl.c */
 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
-int btrfs_check_acl(struct inode *inode, int mask);
+int btrfs_check_acl(struct inode *inode, int mask, unsigned int flags);
 #else
 #define btrfs_check_acl NULL
 #endif

fs/btrfs/inode.c

···
 	return __set_page_dirty_nobuffers(page);
 }
 
-static int btrfs_permission(struct inode *inode, int mask)
+static int btrfs_permission(struct inode *inode, int mask, unsigned int flags)
 {
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	if ((BTRFS_I(inode)->flags & BTRFS_INODE_READONLY) && (mask & MAY_WRITE))
 		return -EACCES;
-	return generic_permission(inode, mask, btrfs_check_acl);
+	return generic_permission(inode, mask, flags, btrfs_check_acl);
 }
 
 static const struct inode_operations btrfs_dir_inode_operations = {

fs/ceph/inode.c

···
  * Check inode permissions.  We verify we have a valid value for
  * the AUTH cap, then call the generic handler.
  */
-int ceph_permission(struct inode *inode, int mask)
+int ceph_permission(struct inode *inode, int mask, unsigned int flags)
 {
-	int err = ceph_do_getattr(inode, CEPH_CAP_AUTH_SHARED);
+	int err;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
+	err = ceph_do_getattr(inode, CEPH_CAP_AUTH_SHARED);
 
 	if (!err)
-		err = generic_permission(inode, mask, NULL);
+		err = generic_permission(inode, mask, flags, NULL);
 	return err;
 }
 

fs/ceph/super.h

···
 extern void ceph_queue_writeback(struct inode *inode);
 
 extern int ceph_do_getattr(struct inode *inode, int mask);
-extern int ceph_permission(struct inode *inode, int mask);
+extern int ceph_permission(struct inode *inode, int mask, unsigned int flags);
 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
 extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
 			struct kstat *stat);

fs/cifs/cifsfs.c

···
 	return 0;
 }
 
-static int cifs_permission(struct inode *inode, int mask)
+static int cifs_permission(struct inode *inode, int mask, unsigned int flags)
 {
 	struct cifs_sb_info *cifs_sb;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
 	cifs_sb = CIFS_SB(inode->i_sb);
 
···
 		on the client (above and beyond ACL on servers) for
 		servers which do not support setting and viewing mode bits,
 		so allowing client to check permissions is useful */
-		return generic_permission(inode, mask, NULL);
+		return generic_permission(inode, mask, flags, NULL);
 }
 
 static struct kmem_cache *cifs_inode_cachep;

fs/coda/dir.c

···
 }
 
 
-int coda_permission(struct inode *inode, int mask)
+int coda_permission(struct inode *inode, int mask, unsigned int flags)
 {
 	int error;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
 	mask &= MAY_READ | MAY_WRITE | MAY_EXEC;
  

fs/coda/pioctl.c

···
 #include <linux/coda_psdev.h>
 
 /* pioctl ops */
-static int coda_ioctl_permission(struct inode *inode, int mask);
+static int coda_ioctl_permission(struct inode *inode, int mask, unsigned int flags);
 static long coda_pioctl(struct file *filp, unsigned int cmd,
 			unsigned long user_data);
 
···
 };
 
 /* the coda pioctl inode ops */
-static int coda_ioctl_permission(struct inode *inode, int mask)
+static int coda_ioctl_permission(struct inode *inode, int mask, unsigned int flags)
 {
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 	return (mask & MAY_EXEC) ? -EACCES : 0;
 }
 

fs/ecryptfs/inode.c

···
 }
 
 static int
-ecryptfs_permission(struct inode *inode, int mask)
+ecryptfs_permission(struct inode *inode, int mask, unsigned int flags)
 {
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 	return inode_permission(ecryptfs_inode_to_lower(inode), mask);
 }
 

fs/ext2/acl.c

···
 }
 
 int
-ext2_check_acl(struct inode *inode, int mask)
+ext2_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
-	struct posix_acl *acl = ext2_get_acl(inode, ACL_TYPE_ACCESS);
+	struct posix_acl *acl;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
+	acl = ext2_get_acl(inode, ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))
 		return PTR_ERR(acl);
 	if (acl) {

fs/ext2/acl.h

···
 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 
 /* acl.c */
-extern int ext2_check_acl (struct inode *, int);
+extern int ext2_check_acl (struct inode *, int, unsigned int);
 extern int ext2_acl_chmod (struct inode *);
 extern int ext2_init_acl (struct inode *, struct inode *);
 

fs/ext3/acl.c

···
 }
 
 int
-ext3_check_acl(struct inode *inode, int mask)
+ext3_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
-	struct posix_acl *acl = ext3_get_acl(inode, ACL_TYPE_ACCESS);
+	struct posix_acl *acl;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
+	acl = ext3_get_acl(inode, ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))
 		return PTR_ERR(acl);
 	if (acl) {

fs/ext3/acl.h

···
 #ifdef CONFIG_EXT3_FS_POSIX_ACL
 
 /* acl.c */
-extern int ext3_check_acl (struct inode *, int);
+extern int ext3_check_acl (struct inode *, int, unsigned int);
 extern int ext3_acl_chmod (struct inode *);
 extern int ext3_init_acl (handle_t *, struct inode *, struct inode *);
 

fs/ext4/acl.c

···
 }
 
 int
-ext4_check_acl(struct inode *inode, int mask)
+ext4_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
-	struct posix_acl *acl = ext4_get_acl(inode, ACL_TYPE_ACCESS);
+	struct posix_acl *acl;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
+	acl = ext4_get_acl(inode, ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))
 		return PTR_ERR(acl);
 	if (acl) {

fs/ext4/acl.h

···
 #ifdef CONFIG_EXT4_FS_POSIX_ACL
 
 /* acl.c */
-extern int ext4_check_acl(struct inode *, int);
+extern int ext4_check_acl(struct inode *, int, unsigned int);
 extern int ext4_acl_chmod(struct inode *);
 extern int ext4_init_acl(handle_t *, struct inode *, struct inode *);
 

fs/fuse/dir.c

···
  * access request is sent.  Execute permission is still checked
  * locally based on file mode.
  */
-static int fuse_permission(struct inode *inode, int mask)
+static int fuse_permission(struct inode *inode, int mask, unsigned int flags)
 {
 	struct fuse_conn *fc = get_fuse_conn(inode);
 	bool refreshed = false;
 	int err = 0;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
 	if (!fuse_allow_task(fc, current))
 		return -EACCES;
···
 	}
 
 	if (fc->flags & FUSE_DEFAULT_PERMISSIONS) {
-		err = generic_permission(inode, mask, NULL);
+		err = generic_permission(inode, mask, flags, NULL);
 
 		/* If permission is denied, try to refresh file
 		   attributes.  This is also needed, because the root
···
 		if (err == -EACCES && !refreshed) {
 			err = fuse_do_getattr(inode, NULL, NULL);
 			if (!err)
-				err = generic_permission(inode, mask, NULL);
+				err = generic_permission(inode, mask,
+							flags, NULL);
 		}
 
 		/* Note: the opposite of the above test does not

fs/generic_acl.c

···
 }
 
 int
-generic_check_acl(struct inode *inode, int mask)
+generic_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
-	struct posix_acl *acl = get_cached_acl(inode, ACL_TYPE_ACCESS);
+	struct posix_acl *acl;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
+	acl = get_cached_acl(inode, ACL_TYPE_ACCESS);
 	if (acl) {
 		int error = posix_acl_permission(inode, acl, mask);
 		posix_acl_release(acl);

fs/gfs2/acl.c

···
  * Returns: errno
  */
 
-int gfs2_check_acl(struct inode *inode, int mask)
+int gfs2_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
 	struct posix_acl *acl;
 	int error;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
 	acl = gfs2_acl_get(GFS2_I(inode), ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))

fs/gfs2/acl.h

···
 #define GFS2_POSIX_ACL_DEFAULT		"posix_acl_default"
 #define GFS2_ACL_MAX_ENTRIES		25
 
-extern int gfs2_check_acl(struct inode *inode, int mask);
+extern int gfs2_check_acl(struct inode *inode, int mask, unsigned int);
 extern int gfs2_acl_create(struct gfs2_inode *dip, struct inode *inode);
 extern int gfs2_acl_chmod(struct gfs2_inode *ip, struct iattr *attr);
 extern const struct xattr_handler gfs2_xattr_system_handler;

fs/gfs2/file.c

···
 	    !capable(CAP_LINUX_IMMUTABLE))
 		goto out;
 	if (!IS_IMMUTABLE(inode)) {
-		error = gfs2_permission(inode, MAY_WRITE);
+		error = gfs2_permission(inode, MAY_WRITE, 0);
 		if (error)
 			goto out;
 	}

fs/gfs2/inode.c

···
 	}
 
 	if (!is_root) {
-		error = gfs2_permission(dir, MAY_EXEC);
+		error = gfs2_permission(dir, MAY_EXEC, 0);
 		if (error)
 			goto out;
 	}
···
 {
 	int error;
 
-	error = gfs2_permission(&dip->i_inode, MAY_WRITE | MAY_EXEC);
+	error = gfs2_permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, 0);
 	if (error)
 		return error;
 

fs/gfs2/inode.h

···
 extern struct inode *gfs2_createi(struct gfs2_holder *ghs,
 				  const struct qstr *name,
 				  unsigned int mode, dev_t dev);
-extern int gfs2_permission(struct inode *inode, int mask);
+extern int gfs2_permission(struct inode *inode, int mask, unsigned int flags);
 extern int gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr);
 extern struct inode *gfs2_lookup_simple(struct inode *dip, const char *name);
 extern void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf);

fs/gfs2/ops_inode.c

···
 	if (error)
 		goto out_child;
 
-	error = gfs2_permission(dir, MAY_WRITE | MAY_EXEC);
+	error = gfs2_permission(dir, MAY_WRITE | MAY_EXEC, 0);
 	if (error)
 		goto out_gunlock;
 
···
 	if (IS_APPEND(&dip->i_inode))
 		return -EPERM;
 
-	error = gfs2_permission(&dip->i_inode, MAY_WRITE | MAY_EXEC);
+	error = gfs2_permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, 0);
 	if (error)
 		return error;
 
···
 			}
 		}
 	} else {
-		error = gfs2_permission(ndir, MAY_WRITE | MAY_EXEC);
+		error = gfs2_permission(ndir, MAY_WRITE | MAY_EXEC, 0);
 		if (error)
 			goto out_gunlock;
 
···
 	/* Check out the dir to be renamed */
 
 	if (dir_rename) {
-		error = gfs2_permission(odentry->d_inode, MAY_WRITE);
+		error = gfs2_permission(odentry->d_inode, MAY_WRITE, 0);
 		if (error)
 			goto out_gunlock;
 	}
···
  * Returns: errno
  */
 
-int gfs2_permission(struct inode *inode, int mask)
+int gfs2_permission(struct inode *inode, int mask, unsigned int flags)
 {
-	struct gfs2_inode *ip = GFS2_I(inode);
+	struct gfs2_inode *ip;
 	struct gfs2_holder i_gh;
 	int error;
 	int unlock = 0;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
+	ip = GFS2_I(inode);
 	if (gfs2_glock_is_locked_by_me(ip->i_gl) == NULL) {
 		error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY, &i_gh);
 		if (error)
···
 	if ((mask & MAY_WRITE) && IS_IMMUTABLE(inode))
 		error = -EACCES;
 	else
-		error = generic_permission(inode, mask, gfs2_check_acl);
+		error = generic_permission(inode, mask, flags, gfs2_check_acl);
 	if (unlock)
 		gfs2_glock_dq_uninit(&i_gh);
 

fs/hostfs/hostfs_kern.c

···
 	return err;
 }
 
-int hostfs_permission(struct inode *ino, int desired)
+int hostfs_permission(struct inode *ino, int desired, unsigned int flags)
 {
 	char *name;
 	int r = 0, w = 0, x = 0, err;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
 	if (desired & MAY_READ) r = 1;
 	if (desired & MAY_WRITE) w = 1;
···
 		err = access_file(name, r, w, x);
 	__putname(name);
 	if (!err)
-		err = generic_permission(ino, desired, NULL);
+		err = generic_permission(ino, desired, flags, NULL);
 	return err;
 }
 

fs/hpfs/namei.c

···
 			unlock_kernel();
 			return -ENOSPC;
 		}
-		if (generic_permission(inode, MAY_WRITE, NULL) ||
+		if (generic_permission(inode, MAY_WRITE, 0, NULL) ||
 		    !S_ISREG(inode->i_mode) ||
 		    get_write_access(inode)) {
 			d_rehash(dentry);

fs/jffs2/acl.c

···
 	return rc;
 }
 
-int jffs2_check_acl(struct inode *inode, int mask)
+int jffs2_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
 	struct posix_acl *acl;
 	int rc;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
 	acl = jffs2_get_acl(inode, ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))

fs/jffs2/acl.h

···
 
 #ifdef CONFIG_JFFS2_FS_POSIX_ACL
 
-extern int jffs2_check_acl(struct inode *, int);
+extern int jffs2_check_acl(struct inode *, int, unsigned int);
 extern int jffs2_acl_chmod(struct inode *);
 extern int jffs2_init_acl_pre(struct inode *, struct inode *, int *);
 extern int jffs2_init_acl_post(struct inode *);

fs/jfs/acl.c

···
 	return rc;
 }
 
-int jfs_check_acl(struct inode *inode, int mask)
+int jfs_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
-	struct posix_acl *acl = jfs_get_acl(inode, ACL_TYPE_ACCESS);
+	struct posix_acl *acl;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
+	acl = jfs_get_acl(inode, ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))
 		return PTR_ERR(acl);
 	if (acl) {

fs/jfs/jfs_acl.h

···
 
 #ifdef CONFIG_JFS_POSIX_ACL
 
-int jfs_check_acl(struct inode *, int);
+int jfs_check_acl(struct inode *, int, unsigned int flags);
 int jfs_init_acl(tid_t, struct inode *, struct inode *);
 int jfs_acl_chmod(struct inode *inode);
 

fs/logfs/dir.c

···
 	return __logfs_create(dir, dentry, inode, target, destlen);
 }
 
-static int logfs_permission(struct inode *inode, int mask)
+static int logfs_permission(struct inode *inode, int mask, unsigned int flags)
 {
-	return generic_permission(inode, mask, NULL);
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+	return generic_permission(inode, mask, flags, NULL);
 }
 
 static int logfs_link(struct dentry *old_dentry, struct inode *dir,

fs/namei.c

···
 /*
  * This does basic POSIX ACL permission checking
  */
-static inline int __acl_permission_check(struct inode *inode, int mask,
-		int (*check_acl)(struct inode *inode, int mask), int rcu)
+static int acl_permission_check(struct inode *inode, int mask, unsigned int flags,
+		int (*check_acl)(struct inode *inode, int mask, unsigned int flags))
 {
 	umode_t			mode = inode->i_mode;
 
···
 		mode >>= 6;
 	else {
 		if (IS_POSIXACL(inode) && (mode & S_IRWXG) && check_acl) {
-			if (rcu) {
-				return -ECHILD;
-			} else {
-				int error = check_acl(inode, mask);
-				if (error != -EAGAIN)
-					return error;
-			}
+			int error = check_acl(inode, mask, flags);
+			if (error != -EAGAIN)
+				return error;
 		}
 
 		if (in_group_p(inode->i_gid))
···
 	return -EACCES;
 }
 
-static inline int acl_permission_check(struct inode *inode, int mask,
-		int (*check_acl)(struct inode *inode, int mask))
-{
-	return __acl_permission_check(inode, mask, check_acl, 0);
-}
-
 /**
- * generic_permission  -  check for access rights on a Posix-like filesystem
+ * generic_permission -  check for access rights on a Posix-like filesystem
  * @inode:	inode to check access rights for
  * @mask:	right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC)
  * @check_acl:	optional callback to check for Posix ACLs
+ * @flags	IPERM_FLAG_ flags.
  *
  * Used to check for read/write/execute permissions on a file.
  * We use "fsuid" for this, letting us set arbitrary permissions
  * for filesystem access without changing the "normal" uids which
- * are used for other things..
+ * are used for other things.
+ *
+ * generic_permission is rcu-walk aware. It returns -ECHILD in case an rcu-walk
+ * request cannot be satisfied (eg. requires blocking or too much complexity).
+ * It would then be called again in ref-walk mode.
  */
-int generic_permission(struct inode *inode, int mask,
-		int (*check_acl)(struct inode *inode, int mask))
+int generic_permission(struct inode *inode, int mask, unsigned int flags,
+	int (*check_acl)(struct inode *inode, int mask, unsigned int flags))
 {
 	int ret;
 
 	/*
 	 * Do the basic POSIX ACL permission checks.
 	 */
-	ret = acl_permission_check(inode, mask, check_acl);
+	ret = acl_permission_check(inode, mask, flags, check_acl);
 	if (ret != -EACCES)
 		return ret;
 
···
 	}
 
 	if (inode->i_op->permission)
-		retval = inode->i_op->permission(inode, mask);
+		retval = inode->i_op->permission(inode, mask, 0);
 	else
-		retval = generic_permission(inode, mask, inode->i_op->check_acl);
+		retval = generic_permission(inode, mask, 0,
+				inode->i_op->check_acl);
 
 	if (retval)
 		return retval;
···
  * short-cut DAC fails, then call ->permission() to do more
  * complete permission check.
  */
-static inline int __exec_permission(struct inode *inode, int rcu)
+static inline int exec_permission(struct inode *inode, unsigned int flags)
 {
 	int ret;
 
 	if (inode->i_op->permission) {
-		if (rcu)
-			return -ECHILD;
-		ret = inode->i_op->permission(inode, MAY_EXEC);
-		if (!ret)
-			goto ok;
-		return ret;
+		ret = inode->i_op->permission(inode, MAY_EXEC, flags);
+	} else {
+		ret = acl_permission_check(inode, MAY_EXEC, flags,
+				inode->i_op->check_acl);
 	}
-	ret = __acl_permission_check(inode, MAY_EXEC, inode->i_op->check_acl, rcu);
-	if (!ret)
+	if (likely(!ret))
 		goto ok;
-	if (rcu && ret == -ECHILD)
+	if (ret == -ECHILD)
 		return ret;
 
 	if (capable(CAP_DAC_OVERRIDE) || capable(CAP_DAC_READ_SEARCH))
···
 
 	return ret;
 ok:
-	return security_inode_exec_permission(inode, rcu);
-}
-
-static int exec_permission(struct inode *inode)
-{
-	return __exec_permission(inode, 0);
-}
-
-static int exec_permission_rcu(struct inode *inode)
-{
-	return __exec_permission(inode, 1);
+	return security_inode_exec_permission(inode, flags);
 }
 
 static __always_inline void set_root(struct nameidata *nd)
···
 
 		nd->flags |= LOOKUP_CONTINUE;
 		if (nd->flags & LOOKUP_RCU) {
-			err = exec_permission_rcu(nd->inode);
+			err = exec_permission(nd->inode, IPERM_FLAG_RCU);
 			if (err == -ECHILD) {
 				if (nameidata_drop_rcu(nd))
 					return -ECHILD;
···
 			}
 		} else {
 exec_again:
-			err = exec_permission(nd->inode);
+			err = exec_permission(nd->inode, 0);
 		}
  		if (err)
 			break;
···
 	struct dentry *dentry;
 	int err;
 
-	err = exec_permission(inode);
+	err = exec_permission(inode, 0);
 	if (err)
 		return ERR_PTR(err);
 

fs/nfs/dir.c

···
 	return nfs_do_access(inode, cred, nfs_open_permission_mask(openflags));
 }
 
-int nfs_permission(struct inode *inode, int mask)
+int nfs_permission(struct inode *inode, int mask, unsigned int flags)
 {
 	struct rpc_cred *cred;
 	int res = 0;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
 	nfs_inc_stats(inode, NFSIOS_VFSACCESS);
 
···
 out_notsup:
 	res = nfs_revalidate_inode(NFS_SERVER(inode), inode);
 	if (res == 0)
-		res = generic_permission(inode, mask, NULL);
+		res = generic_permission(inode, mask, flags, NULL);
 	goto out;
 }
 

fs/nilfs2/inode.c

···
 	return err;
 }
 
-int nilfs_permission(struct inode *inode, int mask)
+int nilfs_permission(struct inode *inode, int mask, unsigned int flags)
 {
-	struct nilfs_root *root = NILFS_I(inode)->i_root;
+	struct nilfs_root *root;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
+	root = NILFS_I(inode)->i_root;
 	if ((mask & MAY_WRITE) && root &&
 	    root->cno != NILFS_CPTREE_CURRENT_CNO)
 		return -EROFS; /* snapshot is not writable */
 
-	return generic_permission(inode, mask, NULL);
+	return generic_permission(inode, mask, flags, NULL);
 }
 
 int nilfs_load_inode_block(struct nilfs_sb_info *sbi, struct inode *inode,

fs/nilfs2/nilfs.h

···
 extern void nilfs_truncate(struct inode *);
 extern void nilfs_evict_inode(struct inode *);
 extern int nilfs_setattr(struct dentry *, struct iattr *);
-int nilfs_permission(struct inode *inode, int mask);
+int nilfs_permission(struct inode *inode, int mask, unsigned int flags);
 extern int nilfs_load_inode_block(struct nilfs_sb_info *, struct inode *,
 				  struct buffer_head **);
 extern int nilfs_inode_dirty(struct inode *);

fs/ocfs2/acl.c

···
 	return ret;
 }
 
-int ocfs2_check_acl(struct inode *inode, int mask)
+int ocfs2_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
-	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	struct ocfs2_super *osb;
 	struct buffer_head *di_bh = NULL;
 	struct posix_acl *acl;
 	int ret = -EAGAIN;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
+	osb = OCFS2_SB(inode->i_sb);
 	if (!(osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL))
 		return ret;
 

fs/ocfs2/acl.h

···
 	__le32 e_id;
 };
 
-extern int ocfs2_check_acl(struct inode *, int);
+extern int ocfs2_check_acl(struct inode *, int, unsigned int);
 extern int ocfs2_acl_chmod(struct inode *);
 extern int ocfs2_init_acl(handle_t *, struct inode *, struct inode *,
 			  struct buffer_head *, struct buffer_head *,

fs/ocfs2/file.c

···
 	return err;
 }
 
-int ocfs2_permission(struct inode *inode, int mask)
+int ocfs2_permission(struct inode *inode, int mask, unsigned int flags)
 {
 	int ret;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
 	mlog_entry_void();
 
···
 		goto out;
 	}
 
-	ret = generic_permission(inode, mask, ocfs2_check_acl);
+	ret = generic_permission(inode, mask, flags, ocfs2_check_acl);
 
 	ocfs2_inode_unlock(inode, 0);
 out:

fs/ocfs2/file.h

···
 int ocfs2_setattr(struct dentry *dentry, struct iattr *attr);
 int ocfs2_getattr(struct vfsmount *mnt, struct dentry *dentry,
 		  struct kstat *stat);
-int ocfs2_permission(struct inode *inode, int mask);
+int ocfs2_permission(struct inode *inode, int mask, unsigned int flags);
 
 int ocfs2_should_update_atime(struct inode *inode,
 			      struct vfsmount *vfsmnt);

fs/proc/base.c

···
  * /proc/pid/fd needs a special permission handler so that a process can still
  * access /proc/self/fd after it has executed a setuid().
  */
-static int proc_fd_permission(struct inode *inode, int mask)
+static int proc_fd_permission(struct inode *inode, int mask, unsigned int flags)
 {
 	int rv;
 
-	rv = generic_permission(inode, mask, NULL);
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+	rv = generic_permission(inode, mask, flags, NULL);
 	if (rv == 0)
 		return 0;
 	if (task_pid(current) == proc_pid(inode))

fs/proc/proc_sysctl.c

···
 	return ret;
 }
 
-static int proc_sys_permission(struct inode *inode, int mask)
+static int proc_sys_permission(struct inode *inode, int mask,unsigned int flags)
 {
 	/*
 	 * sysctl entries that are not writeable,
···
 	struct ctl_table_header *head;
 	struct ctl_table *table;
 	int error;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
 	/* Executable files are not allowed under /proc/sys/ */
 	if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode))

fs/reiserfs/xattr.c

···
 	return err;
 }
 
-static int reiserfs_check_acl(struct inode *inode, int mask)
+static int reiserfs_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
 	struct posix_acl *acl;
 	int error = -EAGAIN; /* do regular unix permission checks by default */
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
 	acl = reiserfs_get_acl(inode, ACL_TYPE_ACCESS);
 
···
 	return 0;
 }
 
-int reiserfs_permission(struct inode *inode, int mask)
+int reiserfs_permission(struct inode *inode, int mask, unsigned int flags)
 {
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 	/*
 	 * We don't do permission checks on the internal objects.
 	 * Permissions are determined by the "owning" object.
···
 	 * Stat data v1 doesn't support ACLs.
 	 */
 	if (get_inode_sd_version(inode) != STAT_DATA_V1)
-		return generic_permission(inode, mask, reiserfs_check_acl);
+		return generic_permission(inode, mask, flags,
+					reiserfs_check_acl);
 #endif
-	return generic_permission(inode, mask, NULL);
+	return generic_permission(inode, mask, flags, NULL);
 }
 
 static int xattr_hide_revalidate(struct dentry *dentry, struct nameidata *nd)

fs/sysfs/inode.c

···
 		return -ENOENT;
 }
 
-int sysfs_permission(struct inode *inode, int mask)
+int sysfs_permission(struct inode *inode, int mask, unsigned int flags)
 {
-	struct sysfs_dirent *sd = inode->i_private;
+	struct sysfs_dirent *sd;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
+	sd = inode->i_private;
 
 	mutex_lock(&sysfs_mutex);
 	sysfs_refresh_inode(sd, inode);
 	mutex_unlock(&sysfs_mutex);
 
-	return generic_permission(inode, mask, NULL);
+	return generic_permission(inode, mask, flags, NULL);
 }

fs/sysfs/sysfs.h

···
 struct inode *sysfs_get_inode(struct super_block *sb, struct sysfs_dirent *sd);
 void sysfs_evict_inode(struct inode *inode);
 int sysfs_sd_setattr(struct sysfs_dirent *sd, struct iattr *iattr);
-int sysfs_permission(struct inode *inode, int mask);
+int sysfs_permission(struct inode *inode, int mask, unsigned int flags);
 int sysfs_setattr(struct dentry *dentry, struct iattr *iattr);
 int sysfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat);
 int sysfs_setxattr(struct dentry *dentry, const char *name, const void *value,

fs/xfs/linux-2.6/xfs_acl.c

···
 }
 
 int
-xfs_check_acl(struct inode *inode, int mask)
+xfs_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
-	struct xfs_inode *ip = XFS_I(inode);
+	struct xfs_inode *ip;
 	struct posix_acl *acl;
 	int error = -EAGAIN;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
+	ip = XFS_I(inode);
 	trace_xfs_check_acl(ip);
 
 	/*

fs/xfs/xfs_acl.h

···
 #define SGI_ACL_DEFAULT_SIZE	(sizeof(SGI_ACL_DEFAULT)-1)
 
 #ifdef CONFIG_XFS_POSIX_ACL
-extern int xfs_check_acl(struct inode *inode, int mask);
+extern int xfs_check_acl(struct inode *inode, int mask, unsigned int flags);
 extern struct posix_acl *xfs_get_acl(struct inode *inode, int type);
 extern int xfs_inherit_acl(struct inode *inode, struct posix_acl *default_acl);
 extern int xfs_acl_chmod(struct inode *inode);

include/linux/coda_linux.h

···
 /* operations shared over more than one file */
 int coda_open(struct inode *i, struct file *f);
 int coda_release(struct inode *i, struct file *f);
-int coda_permission(struct inode *inode, int mask);
+int coda_permission(struct inode *inode, int mask, unsigned int flags);
 int coda_revalidate_inode(struct dentry *);
 int coda_getattr(struct vfsmount *, struct dentry *, struct kstat *);
 int coda_setattr(struct dentry *, struct iattr *);

include/linux/fs.h

···
 	int (*setlease)(struct file *, long, struct file_lock **);
 };
 
+#define IPERM_FLAG_RCU	0x0001
+
 struct inode_operations {
 	struct dentry * (*lookup) (struct inode *,struct dentry *, struct nameidata *);
 	void * (*follow_link) (struct dentry *, struct nameidata *);
-	int (*permission) (struct inode *, int);
-	int (*check_acl)(struct inode *, int);
+	int (*permission) (struct inode *, int, unsigned int);
+	int (*check_acl)(struct inode *, int, unsigned int);
 
 	int (*readlink) (struct dentry *, char __user *,int);
 	void (*put_link) (struct dentry *, struct nameidata *, void *);
···
 #endif
 extern int notify_change(struct dentry *, struct iattr *);
 extern int inode_permission(struct inode *, int);
-extern int generic_permission(struct inode *, int,
-		int (*check_acl)(struct inode *, int));
+extern int generic_permission(struct inode *, int, unsigned int,
+		int (*check_acl)(struct inode *, int, unsigned int));
 
 static inline bool execute_ok(struct inode *inode)
 {

include/linux/generic_acl.h

···
 
 int generic_acl_init(struct inode *, struct inode *);
 int generic_acl_chmod(struct inode *);
-int generic_check_acl(struct inode *inode, int mask);
+int generic_check_acl(struct inode *inode, int mask, unsigned int flags);
 
 #endif /* LINUX_GENERIC_ACL_H */

include/linux/nfs_fs.h

···
 extern int nfs_post_op_update_inode(struct inode *inode, struct nfs_fattr *fattr);
 extern int nfs_post_op_update_inode_force_wcc(struct inode *inode, struct nfs_fattr *fattr);
 extern int nfs_getattr(struct vfsmount *, struct dentry *, struct kstat *);
-extern int nfs_permission(struct inode *, int);
+extern int nfs_permission(struct inode *, int, unsigned int);
 extern int nfs_open(struct inode *, struct file *);
 extern int nfs_release(struct inode *, struct file *);
 extern int nfs_attribute_timeout(struct inode *inode);

include/linux/reiserfs_xattr.h

···
 int reiserfs_lookup_privroot(struct super_block *sb);
 int reiserfs_delete_xattrs(struct inode *inode);
 int reiserfs_chown_xattrs(struct inode *inode, struct iattr *attrs);
-int reiserfs_permission(struct inode *inode, int mask);
+int reiserfs_permission(struct inode *inode, int mask, unsigned int flags);
 
 #ifdef CONFIG_REISERFS_FS_XATTR
 #define has_xattr_dir(inode) (REISERFS_I(inode)->i_flags & i_has_xattr_dir)