Message-Id: <1206629746-4298-23-git-send-email-Artem.Bityutskiy@nokia.com>
Date:	Thu, 27 Mar 2008 16:55:42 +0200
From:	Artem Bityutskiy <Artem.Bityutskiy@...ia.com>
To:	LKML <linux-kernel@...r.kernel.org>
Cc:	Adrian Hunter <ext-adrian.hunter@...ia.com>,
	Artem Bityutskiy <Artem.Bityutskiy@...ia.com>
Subject: [RFC PATCH 22/26] UBIFS: add extended attribute support

Extended attributes are implemented as separate inodes. This makes
it very easy to implement them and to re-use nearly all the existing
code. This might be not the fastest implementation, though. ACL support
is not implemented.

Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@...ia.com>
Signed-off-by: Adrian Hunter <ext-adrian.hunter@...ia.com>
---
 fs/ubifs/xattr.c |  587 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 files changed, 587 insertions(+), 0 deletions(-)

diff --git a/fs/ubifs/xattr.c b/fs/ubifs/xattr.c
new file mode 100644
index 0000000..85b1088
--- /dev/null
+++ b/fs/ubifs/xattr.c
@@ -0,0 +1,587 @@
+/*
+ * This file is part of UBIFS.
+ *
+ * Copyright (C) 2006-2008 Nokia Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc., 51
+ * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ *
+ * Authors: Artem Bityutskiy (Битюцкий Артём)
+ *          Adrian Hunter
+ */
+
+/*
+ * This file implements UBIFS extended attributes support.
+ *
+ * Extended attributes are implemented as regular inodes with attached data,
+ * which limits extended attribute size to UBIFS block size (4KiB). Names of
+ * extended attributes are described by extended attribute entries (xentries),
+ * which are almost identical to directory entries, but have different key type.
+ *
+ * In other words, the situation with extended attributes is very similar to
+ * directories. Indeed, any inode (but of course not xattr inodes) may have a
+ * number of associated xentries, just like directory inodes have associated
+ * directory entries. Extended attribute entries store the name of the extended
+ * attribute, the host inode number, and the extended attribute inode number.
+ * Similarly, direntries store the name, the parent and the target inode
+ * numbers. Thus, most of the common UBIFS mechanisms may be re-used for
+ * extended attributes.
+ *
+ * The number of extended attributes is not limited, but there is Linux
+ * limitation on the maximum possible size of the list of all extended
+ * attributes associated with an inode (%XATTR_LIST_MAX), so UBIFS makes sure
+ * the sum of all extended attribute names of the inode does not exceed that
+ * limit.
+ *
+ * Extended attributes are synchronous, which means they are written to the
+ * flash media synchronously and there is no write-back for extended attribute
+ * inodes. The extended attribute values are not stored in compressed form on
+ * the media.
+ *
+ * Since extended attributes are represented by regular inodes, they are cached
+ * in the VFS inode cache. The xentries are cached in the LNC cache (see
+ * tnc.c).
+ *
+ * ACL support is not implemented.
+ */
+
+#include <linux/xattr.h>
+#include <linux/posix_acl_xattr.h>
+#include "ubifs.h"
+
+/* How many bytes an extended attribute adds to the host inode */
+#define CALC_XATTR_BYTES(data_len) ALIGN(UBIFS_INO_NODE_SZ + (data_len) + 1, 8)
+
+/*
+ * Extended attribute type constants.
+ *
+ * USER_XATTR: user extended attribute ("user.*")
+ * TRUSTED_XATTR: trusted extended attribute ("trusted.*)
+ * SECURITY_XATTR: security extended attribute ("security.*")
+ */
+enum {
+	USER_XATTR,
+	TRUSTED_XATTR,
+	SECURITY_XATTR,
+};
+
+static struct inode_operations none_inode_operations;
+static struct address_space_operations none_address_operations;
+static struct file_operations none_file_operations;
+
+/**
+ * create_xattr - create an extended attribute.
+ * @c: UBIFS file-system description object
+ * @host: host inode
+ * @nm: extended attribute name
+ * @value: extended attribute value
+ * @size: size of extended attribute value
+ *
+ * This is a helper function which creates an extended attribute of name @nm
+ * and value @value for inode @host. The host inode is also updated on flash
+ * because the ctime and extended attribute accounting data changes. This
+ * function returns zero in case of success and a negative error code in case
+ * of failure.
+ */
+static int create_xattr(struct ubifs_info *c, struct inode *host,
+			const struct qstr *nm, const void *value, int size)
+{
+	struct ubifs_inode *ui, *host_ui = ubifs_inode(host);
+	struct ubifs_budget_req req = { .new_ino = 1, .new_dent = 1,
+					.new_ino_d = size };
+	struct inode *inode;
+	int err;
+
+	/*
+	 * Linux limits the maximum size of the extended attribute names list
+	 * to %XATTR_LIST_MAX. This means we should not allow creating more*
+	 * extended attributes if the name list becomes larger. This limitation
+	 * is artificial for UBIFS, though.
+	 */
+	if (host_ui->xattr_names + host_ui->xattr_cnt +
+					nm->len + 1 > XATTR_LIST_MAX)
+		return -ENOSPC;
+
+	err = ubifs_budget_inode_op(c, host, &req);
+	if (err)
+		return err;
+
+	inode = ubifs_new_inode(c, host, S_IFREG | S_IRWXUGO);
+	if (IS_ERR(inode)) {
+		err = PTR_ERR(inode);
+		goto out_budg;
+	}
+
+	/* Re-define all operations to be "nothing" */
+	inode->i_mapping->a_ops = &none_address_operations;
+	inode->i_op = &none_inode_operations;
+	inode->i_fop = &none_file_operations;
+
+	inode->i_flags |= S_SYNC | S_NOATIME | S_NOCMTIME | S_NOQUOTA;
+	ui = ubifs_inode(inode);
+	ui->xattr = 1;
+	ui->data = kmalloc(size, GFP_KERNEL);
+	if (!ui->data) {
+		err = -ENOMEM;
+		goto out_inode;
+	}
+
+	memcpy(ui->data, value, size);
+	host->i_ctime = CURRENT_TIME_SEC;
+	host_ui->xattr_cnt += 1;
+	spin_lock(&host->i_lock);
+	host_ui->xattr_size += CALC_DENT_SIZE(nm->len);
+	host_ui->xattr_size += CALC_XATTR_BYTES(size);
+	spin_unlock(&host->i_lock);
+	host_ui->xattr_names += nm->len;
+
+	/*
+	 * We do not use i_size_write() because nobody can race with us as we
+	 * are holding host @host->i_mutex - every xattr operation for this
+	 * inode is serialized by it.
+	 */
+	inode->i_size = size;
+	ui->data_len = size;
+
+	/*
+	 * Note, it is important that 'ubifs_jrn_update()' writes the @host
+	 * inode last, so when it gets synchronized and the write-buffer is
+	 * flushed, the extended attribute is flushed as well.
+	 */
+	err = ubifs_jrn_update(c, host, nm, inode, 0, IS_DIRSYNC(host), 1);
+	if (err)
+		goto out_cancel;
+
+	ubifs_release_ino_clean(c, host, &req);
+	insert_inode_hash(inode);
+	iput(inode);
+	return 0;
+
+out_cancel:
+	host_ui->xattr_cnt -= 1;
+	spin_lock(&host->i_lock);
+	host_ui->xattr_size -= CALC_DENT_SIZE(nm->len);
+	host_ui->xattr_size -= CALC_XATTR_BYTES(size);
+	spin_unlock(&host->i_lock);
+out_inode:
+	make_bad_inode(inode);
+	iput(inode);
+out_budg:
+	ubifs_cancel_ino_op(c, host, &req);
+	return err;
+}
+
+/**
+ * change_xattr - change an extended attribute.
+ * @c: UBIFS file-system description object
+ * @host: host inode
+ * @inode: extended attribute inode
+ * @value: extended attribute value
+ * @size: size of extended attribute value
+ *
+ * This helper function changes the value of extended attribute @inode with new
+ * data from @value. Returns zero in case of success and a negative error code
+ * in case of failure.
+ */
+static int change_xattr(struct ubifs_info *c, struct inode *host,
+			struct inode *inode, const void *value, int size)
+{
+	struct ubifs_inode *host_ui = ubifs_inode(host);
+	struct ubifs_inode *ui = ubifs_inode(inode);
+	struct ubifs_budget_req req = { .dirtied_ino = 1,
+					.dirtied_ino_d = ui->data_len };
+	int err;
+
+	ubifs_assert(ui->data_len == inode->i_size);
+
+	err = ubifs_budget_inode_op(c, host, &req);
+	if (err)
+		return err;
+
+	host->i_ctime = CURRENT_TIME_SEC;
+	spin_lock(&host->i_lock);
+	host_ui->xattr_size -= CALC_XATTR_BYTES(ui->data_len);
+	host_ui->xattr_size += CALC_XATTR_BYTES(size);
+	spin_unlock(&host->i_lock);
+
+	kfree(ui->data);
+	ui->data = kmalloc(size, GFP_KERNEL);
+	if (!ui->data) {
+		err = -ENOMEM;
+		goto out_budg;
+	}
+
+	memcpy(ui->data, value, size);
+	inode->i_size = size;
+	ui->data_len = size;
+
+	/*
+	 * It is important to write the host inode after the xattr inode
+	 * because if the host inode gets synchronized, then the extended
+	 * attribute inode gets synchronized, because it goes before the host
+	 * inode in the write-buffer.
+	 */
+	err = ubifs_jrn_write_2_inodes(c, inode, host, IS_DIRSYNC(host));
+	if (err)
+		goto out_cancel;
+
+	ubifs_release_ino_clean(c, host, &req);
+	return 0;
+
+out_cancel:
+	spin_lock(&host->i_lock);
+	host_ui->xattr_size -= CALC_XATTR_BYTES(size);
+	host_ui->xattr_size += CALC_XATTR_BYTES(ui->data_len);
+	spin_unlock(&host->i_lock);
+	make_bad_inode(inode);
+out_budg:
+	ubifs_cancel_ino_op(c, host, &req);
+	return err;
+}
+
+/**
+ * check_namespace - check extended attribute name-space.
+ * @nm: extended attribute name
+ *
+ * This function makes sure the extended attribute name belongs to one of the
+ * supported extended attribute name-spaces. Returns name-space index in case
+ * of success and a negative error code in case of failure.
+ */
+static int check_namespace(const struct qstr *nm)
+{
+	int type;
+
+	if (nm->len > UBIFS_MAX_NLEN)
+		return -ENAMETOOLONG;
+
+	if (!strncmp(nm->name, XATTR_TRUSTED_PREFIX,
+		     XATTR_TRUSTED_PREFIX_LEN)) {
+		if (nm->name[sizeof(XATTR_TRUSTED_PREFIX) - 1] == '\0')
+			return -EINVAL;
+		type = TRUSTED_XATTR;
+	} else if (!strncmp(nm->name, XATTR_USER_PREFIX,
+				      XATTR_USER_PREFIX_LEN)) {
+		if (nm->name[XATTR_USER_PREFIX_LEN] == '\0')
+			return -EINVAL;
+		type = USER_XATTR;
+	} else if (!strncmp(nm->name, XATTR_SECURITY_PREFIX,
+				     XATTR_SECURITY_PREFIX_LEN)) {
+		if (nm->name[sizeof(XATTR_SECURITY_PREFIX) - 1] == '\0')
+			return -EINVAL;
+		type = SECURITY_XATTR;
+	} else
+		return -EOPNOTSUPP;
+
+	return type;
+}
+
+int ubifs_setxattr(struct dentry *dentry, const char *name,
+		   const void *value, size_t size, int flags)
+{
+	struct inode *inode, *host = dentry->d_inode;
+	struct ubifs_info *c = host->i_sb->s_fs_info;
+	struct qstr nm = { .name = name, .len = strlen(name) };
+	struct ubifs_dent_node *xent;
+	union ubifs_key key;
+	int err, type;
+
+	dbg_gen("xattr '%s', host ino %lu ('%.*s'), size %d", name,
+		host->i_ino, dentry->d_name.len, dentry->d_name.name, size);
+	ubifs_assert(mutex_is_locked(&host->i_mutex));
+	ubifs_assert(ubifs_inode(host)->xattr_cnt >= 0);
+	ubifs_assert(ubifs_inode(host)->xattr_size >= 0);
+	ubifs_assert(ubifs_inode(host)->xattr_msize >= 0);
+	ubifs_assert(ubifs_inode(host)->xattr_names >= 0);
+
+	if (size > UBIFS_MAX_INO_DATA)
+		return -ERANGE;
+
+	type = check_namespace(&nm);
+	if (type < 0)
+		return type;
+
+	xent = kmalloc(UBIFS_MAX_XENT_NODE_SZ, GFP_NOFS);
+	if (!xent)
+		return -ENOMEM;
+
+	/*
+	 * The extended attribute entries are stored in LNC, so multiple
+	 * look-ups do not involve reading the flash.
+	 */
+	xent_key_init(c, &key, host->i_ino, &nm);
+	err = ubifs_tnc_lookup_nm(c, &key, xent, &nm);
+	if (err) {
+		if (err != -ENOENT)
+			goto out_free;
+
+		if (flags & XATTR_REPLACE)
+			/* We are asked not to create the xattr */
+			err = -ENODATA;
+		else
+			err = create_xattr(c, host, &nm, value, size);
+		goto out_free;
+	}
+
+	if (flags & XATTR_CREATE) {
+		/* We are asked not to replace the xattr */
+		err = -EEXIST;
+		goto out_free;
+	}
+
+	inode = ubifs_iget(c->vfs_sb, le64_to_cpu(xent->inum));
+	if (IS_ERR(inode)) {
+		ubifs_err("dead extended attribute node entry");
+		ubifs_ro_mode(c);
+		err = PTR_ERR(inode);
+		goto out_free;
+	}
+
+	err = change_xattr(c, host, inode, value, size);
+	iput(inode);
+
+out_free:
+	kfree(xent);
+	return err;
+}
+
+ssize_t ubifs_getxattr(struct dentry *dentry, const char *name, void *buf,
+		       size_t size)
+{
+	struct inode *inode, *host = dentry->d_inode;
+	struct ubifs_info *c = host->i_sb->s_fs_info;
+	struct qstr nm = { .name = name, .len = strlen(name) };
+	struct ubifs_inode *ui;
+	struct ubifs_dent_node *xent;
+	union ubifs_key key;
+	int err;
+
+	dbg_gen("xattr '%s', ino %lu ('%.*s'), buf size %d", name,
+		host->i_ino, dentry->d_name.len, dentry->d_name.name, size);
+	ubifs_assert(ubifs_inode(host)->xattr_cnt >= 0);
+	ubifs_assert(ubifs_inode(host)->xattr_size >= 0);
+	ubifs_assert(ubifs_inode(host)->xattr_msize >= 0);
+	ubifs_assert(ubifs_inode(host)->xattr_names >= 0);
+
+	err = check_namespace(&nm);
+	if (err < 0)
+		return err;
+
+	xent = kmalloc(UBIFS_MAX_XENT_NODE_SZ, GFP_NOFS);
+	if (!xent)
+		return -ENOMEM;
+
+	mutex_lock(&host->i_mutex);
+	xent_key_init(c, &key, host->i_ino, &nm);
+	err = ubifs_tnc_lookup_nm(c, &key, xent, &nm);
+	if (err) {
+		if (err == -ENOENT)
+			err = -ENODATA;
+		goto out_unlock;
+	}
+
+	inode = ubifs_iget(c->vfs_sb, le64_to_cpu(xent->inum));
+	if (IS_ERR(inode)) {
+		ubifs_err("dead extended attribute node entry");
+		ubifs_ro_mode(c);
+		err = PTR_ERR(inode);
+		goto out_unlock;
+	}
+
+	ui = ubifs_inode(inode);
+	ubifs_assert(inode->i_size == ui->data_len);
+	ubifs_assert(ubifs_inode(host)->xattr_size > ui->data_len);
+
+	if (buf) {
+		/* If @buf is %NULL we are supposed to return the length */
+		if (ui->data_len > size) {
+			dbg_err("buffer size %d, xattr len %d",
+				size, ui->data_len);
+			err = -ERANGE;
+			goto out_iput;
+		}
+
+		memcpy(buf, ui->data, ui->data_len);
+	}
+	err = ui->data_len;
+
+out_iput:
+	iput(inode);
+out_unlock:
+	mutex_unlock(&host->i_mutex);
+	kfree(xent);
+	return err;
+}
+
+ssize_t ubifs_listxattr(struct dentry *dentry, char *buffer, size_t size)
+{
+	struct inode *host = dentry->d_inode;
+	struct ubifs_info *c = host->i_sb->s_fs_info;
+	struct ubifs_inode *host_ui = ubifs_inode(host);
+	union ubifs_key key;
+	struct ubifs_dent_node *xent, *pxent = NULL;
+	int err, len, written = 0;
+	struct qstr nm = { .name = NULL };
+
+	dbg_gen("ino %lu ('%.*s'), buffer size %zd", host->i_ino,
+		dentry->d_name.len, dentry->d_name.name, size);
+	ubifs_assert(host_ui->xattr_cnt >= 0);
+	ubifs_assert(host_ui->xattr_size >= 0);
+	ubifs_assert(host_ui->xattr_msize >= 0);
+	ubifs_assert(host_ui->xattr_names >= 0);
+
+	len = host_ui->xattr_names + host_ui->xattr_cnt;
+	if (!buffer)
+		/*
+		 * We should return the minimum buffer size which will fit a
+		 * null-terminated list of all the extended attribute names.
+		 */
+		return len;
+
+	if (len > size)
+		return -ERANGE;
+
+	lowest_xent_key(c, &key, host->i_ino);
+
+	mutex_lock(&host->i_mutex);
+	while (1) {
+		int type;
+
+		xent = ubifs_tnc_next_ent(c, &key, &nm);
+		if (unlikely(IS_ERR(xent))) {
+			err = PTR_ERR(xent);
+			break;
+		}
+
+		nm.name = xent->name;
+		nm.len = le16_to_cpu(xent->nlen);
+
+		type = check_namespace(&nm);
+		if (unlikely(type < 0)) {
+			err = type;
+			break;
+		}
+
+		/* Show trusted namespace only for "power" users */
+		if (type != TRUSTED_XATTR || capable(CAP_SYS_ADMIN)) {
+			memcpy(buffer + written, nm.name, nm.len + 1);
+			written += nm.len + 1;
+		}
+
+		kfree(pxent);
+		pxent = xent;
+		key_read(c, &xent->key, &key);
+	}
+	mutex_unlock(&host->i_mutex);
+
+	kfree(pxent);
+	if (err != -ENOENT) {
+		ubifs_err("cannot find next direntry, error %d", err);
+		return err;
+	}
+
+	ubifs_assert(written <= size);
+	return written;
+}
+
+static int remove_xattr(struct ubifs_info *c, struct inode *host,
+			struct inode *inode, const struct qstr *nm)
+{
+	struct ubifs_inode *host_ui = ubifs_inode(host);
+	struct ubifs_inode *ui = ubifs_inode(inode);
+	struct ubifs_budget_req req = { .dirtied_ino = 1, .mod_dent = 1 };
+	int err;
+
+	ubifs_assert(ui->data_len == inode->i_size);
+
+	err = ubifs_budget_inode_op(c, host, &req);
+	if (err)
+		return err;
+
+	host->i_ctime = CURRENT_TIME_SEC;
+	host_ui->xattr_cnt -= 1;
+	spin_lock(&host->i_lock);
+	host_ui->xattr_size -= CALC_XATTR_BYTES(ui->data_len);
+	spin_unlock(&host->i_lock);
+	host_ui->xattr_names -= nm->len;
+
+	err = ubifs_jrn_delete_xattr(c, host, inode, nm, IS_DIRSYNC(host));
+	if (err)
+		goto out_cancel;
+
+	ubifs_release_ino_clean(c, host, &req);
+	return 0;
+
+out_cancel:
+	ubifs_cancel_ino_op(c, host, &req);
+	host_ui->xattr_cnt += 1;
+	spin_lock(&host->i_lock);
+	host_ui->xattr_size += CALC_XATTR_BYTES(ui->data_len);
+	spin_unlock(&host->i_lock);
+	make_bad_inode(inode);
+	return err;
+}
+
+int ubifs_removexattr(struct dentry *dentry, const char *name)
+{
+	struct inode *inode, *host = dentry->d_inode;
+	struct ubifs_info *c = host->i_sb->s_fs_info;
+	struct qstr nm = { .name = name, .len = strlen(name) };
+	struct ubifs_dent_node *xent;
+	union ubifs_key key;
+	int err;
+
+	dbg_gen("xattr '%s', ino %lu ('%.*s')", name,
+		host->i_ino, dentry->d_name.len, dentry->d_name.name);
+	ubifs_assert(mutex_is_locked(&host->i_mutex));
+	ubifs_assert(ubifs_inode(host)->xattr_cnt >= 0);
+	ubifs_assert(ubifs_inode(host)->xattr_size >= 0);
+	ubifs_assert(ubifs_inode(host)->xattr_msize >= 0);
+	ubifs_assert(ubifs_inode(host)->xattr_names >= 0);
+
+	err = check_namespace(&nm);
+	if (err < 0)
+		return err;
+
+	xent = kmalloc(UBIFS_MAX_XENT_NODE_SZ, GFP_NOFS);
+	if (!xent)
+		return -ENOMEM;
+
+	xent_key_init(c, &key, host->i_ino, &nm);
+	err = ubifs_tnc_lookup_nm(c, &key, xent, &nm);
+	if (err) {
+		if (err == -ENOENT)
+			err = -ENODATA;
+		goto out_free;
+	}
+
+	inode = ubifs_iget(c->vfs_sb, le64_to_cpu(xent->inum));
+	if (IS_ERR(inode)) {
+		ubifs_err("dead extended attribute node entry");
+		ubifs_ro_mode(c);
+		err = PTR_ERR(inode);
+		goto out_free;
+	}
+
+	ubifs_assert(inode->i_nlink == 1);
+	inode->i_nlink = 0;
+	err = remove_xattr(c, host, inode, &nm);
+	if (err)
+		inode->i_nlink = 1;
+
+	/* If @i_nlink is 0, 'iput()' will delete the inode */
+	iput(inode);
+
+out_free:
+	kfree(xent);
+	return err;
+}
-- 
1.5.4.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists