staging: erofs: add namei functions
authorGao Xiang <gaoxiang25@huawei.com>
Thu, 26 Jul 2018 12:21:50 +0000 (20:21 +0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Fri, 27 Jul 2018 15:22:00 +0000 (17:22 +0200)
This commit adds functions that transfer names to inodes.

Signed-off-by: Miao Xie <miaoxie@huawei.com>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Gao Xiang <gaoxiang25@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/staging/erofs/namei.c [new file with mode: 0644]

diff --git a/drivers/staging/erofs/namei.c b/drivers/staging/erofs/namei.c
new file mode 100644 (file)
index 0000000..39db643
--- /dev/null
@@ -0,0 +1,243 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * linux/drivers/staging/erofs/namei.c
+ *
+ * Copyright (C) 2017-2018 HUAWEI, Inc.
+ *             http://www.huawei.com/
+ * Created by Gao Xiang <gaoxiang25@huawei.com>
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file COPYING in the main directory of the Linux
+ * distribution for more details.
+ */
+#include "internal.h"
+
+/* based on the value of qn->len is accurate */
+static inline int dirnamecmp(struct qstr *qn,
+       struct qstr *qd, unsigned *matched)
+{
+       unsigned i = *matched, len = min(qn->len, qd->len);
+loop:
+       if (unlikely(i >= len)) {
+               *matched = i;
+               if (qn->len < qd->len) {
+                       /*
+                        * actually (qn->len == qd->len)
+                        * when qd->name[i] == '\0'
+                        */
+                       return qd->name[i] == '\0' ? 0 : -1;
+               }
+               return (qn->len > qd->len);
+       }
+
+       if (qn->name[i] != qd->name[i]) {
+               *matched = i;
+               return qn->name[i] > qd->name[i] ? 1 : -1;
+       }
+
+       ++i;
+       goto loop;
+}
+
+static struct erofs_dirent *find_target_dirent(
+       struct qstr *name,
+       u8 *data, int maxsize)
+{
+       unsigned ndirents, head, back;
+       unsigned startprfx, endprfx;
+       struct erofs_dirent *const de = (struct erofs_dirent *)data;
+
+       /* make sure that maxsize is valid */
+       BUG_ON(maxsize < sizeof(struct erofs_dirent));
+
+       ndirents = le16_to_cpu(de->nameoff) / sizeof(*de);
+
+       /* corrupted dir (may be unnecessary...) */
+       BUG_ON(!ndirents);
+
+       head = 0;
+       back = ndirents - 1;
+       startprfx = endprfx = 0;
+
+       while (head <= back) {
+               unsigned mid = head + (back - head) / 2;
+               unsigned nameoff = le16_to_cpu(de[mid].nameoff);
+               unsigned matched = min(startprfx, endprfx);
+
+               struct qstr dname = QSTR_INIT(data + nameoff,
+                       unlikely(mid >= ndirents - 1) ?
+                               maxsize - nameoff :
+                               le16_to_cpu(de[mid + 1].nameoff) - nameoff);
+
+               /* string comparison without already matched prefix */
+               int ret = dirnamecmp(name, &dname, &matched);
+
+               if (unlikely(!ret))
+                       return de + mid;
+               else if (ret > 0) {
+                       head = mid + 1;
+                       startprfx = matched;
+               } else if (unlikely(mid < 1))   /* fix "mid" overflow */
+                       break;
+               else {
+                       back = mid - 1;
+                       endprfx = matched;
+               }
+       }
+
+       return ERR_PTR(-ENOENT);
+}
+
+static struct page *find_target_block_classic(
+       struct inode *dir,
+       struct qstr *name, int *_diff)
+{
+       unsigned startprfx, endprfx;
+       unsigned head, back;
+       struct address_space *const mapping = dir->i_mapping;
+       struct page *candidate = ERR_PTR(-ENOENT);
+
+       startprfx = endprfx = 0;
+       head = 0;
+       back = inode_datablocks(dir) - 1;
+
+       while (head <= back) {
+               unsigned mid = head + (back - head) / 2;
+               struct page *page = read_mapping_page(mapping, mid, NULL);
+
+               if (IS_ERR(page)) {
+exact_out:
+                       if (!IS_ERR(candidate)) /* valid candidate */
+                               put_page(candidate);
+                       return page;
+               } else {
+                       int diff;
+                       unsigned ndirents, matched;
+                       struct qstr dname;
+                       struct erofs_dirent *de = kmap_atomic(page);
+                       unsigned nameoff = le16_to_cpu(de->nameoff);
+
+                       ndirents = nameoff / sizeof(*de);
+
+                       /* corrupted dir (should have one entry at least) */
+                       BUG_ON(!ndirents || nameoff > PAGE_SIZE);
+
+                       matched = min(startprfx, endprfx);
+
+                       dname.name = (u8 *)de + nameoff;
+                       dname.len = ndirents == 1 ?
+                               /* since the rest of the last page is 0 */
+                               EROFS_BLKSIZ - nameoff
+                               : le16_to_cpu(de[1].nameoff) - nameoff;
+
+                       /* string comparison without already matched prefix */
+                       diff = dirnamecmp(name, &dname, &matched);
+                       kunmap_atomic(de);
+
+                       if (unlikely(!diff)) {
+                               *_diff = 0;
+                               goto exact_out;
+                       } else if (diff > 0) {
+                               head = mid + 1;
+                               startprfx = matched;
+
+                               if (likely(!IS_ERR(candidate)))
+                                       put_page(candidate);
+                               candidate = page;
+                       } else {
+                               put_page(page);
+
+                               if (unlikely(mid < 1))  /* fix "mid" overflow */
+                                       break;
+
+                               back = mid - 1;
+                               endprfx = matched;
+                       }
+               }
+       }
+       *_diff = 1;
+       return candidate;
+}
+
+int erofs_namei(struct inode *dir,
+       struct qstr *name,
+       erofs_nid_t *nid, unsigned *d_type)
+{
+       int diff;
+       struct page *page;
+       u8 *data;
+       struct erofs_dirent *de;
+
+       if (unlikely(!dir->i_size))
+               return -ENOENT;
+
+       diff = 1;
+       page = find_target_block_classic(dir, name, &diff);
+
+       if (unlikely(IS_ERR(page)))
+               return PTR_ERR(page);
+
+       data = kmap_atomic(page);
+       /* the target page has been mapped */
+       de = likely(diff) ?
+               /* since the rest of the last page is 0 */
+               find_target_dirent(name, data, EROFS_BLKSIZ) :
+               (struct erofs_dirent *)data;
+
+       if (likely(!IS_ERR(de))) {
+               *nid = le64_to_cpu(de->nid);
+               *d_type = de->file_type;
+       }
+
+       kunmap_atomic(data);
+       put_page(page);
+
+       return IS_ERR(de) ? PTR_ERR(de) : 0;
+}
+
+/* NOTE: i_mutex is already held by vfs */
+static struct dentry *erofs_lookup(struct inode *dir,
+       struct dentry *dentry, unsigned int flags)
+{
+       int err;
+       erofs_nid_t nid;
+       unsigned d_type;
+       struct inode *inode;
+
+       DBG_BUGON(!d_really_is_negative(dentry));
+       /* dentry must be unhashed in lookup, no need to worry about */
+       DBG_BUGON(!d_unhashed(dentry));
+
+       /* file name exceeds fs limit */
+       if (unlikely(dentry->d_name.len > EROFS_NAME_LEN))
+               return ERR_PTR(-ENAMETOOLONG);
+
+       /* false uninitialized warnings on gcc 4.8.x */
+       err = erofs_namei(dir, &dentry->d_name, &nid, &d_type);
+
+       if (err == -ENOENT) {
+               /* negative dentry */
+               inode = NULL;
+               goto negative_out;
+       } else if (unlikely(err))
+               return ERR_PTR(err);
+
+       debugln("%s, %s (nid %llu) found, d_type %u", __func__,
+               dentry->d_name.name, nid, d_type);
+
+       inode = erofs_iget(dir->i_sb, nid, d_type == EROFS_FT_DIR);
+       if (IS_ERR(inode))
+               return ERR_CAST(inode);
+
+negative_out:
+       return d_splice_alias(inode, dentry);
+}
+
+const struct inode_operations erofs_dir_iops = {
+       .lookup = erofs_lookup,
+};
+
+const struct inode_operations erofs_dir_xattr_iops = {
+       .lookup = erofs_lookup,
+};
+