[CIFS] DFS support: provide shrinkable mounts

Signed-off-by: Igor Mammedov <niallain@gmail.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>

5 files changed, 387 insertions, 1 deletions

diff --git a/fs/cifs/Makefile b/fs/cifs/Makefile
index 09898b8dc69b..6ba43fb346fb 100644
--- a/fs/cifs/Makefile
+++ b/fs/cifs/Makefile
@@ -10,4 +10,4 @@ cifs-y := cifsfs.o cifssmb.o cifs_debug.o connect.o dir.o file.o inode.o \
 
 cifs-$(CONFIG_CIFS_UPCALL) += cifs_spnego.o
 
-cifs-$(CONFIG_CIFS_DFS_UPCALL) += dns_resolve.o
+cifs-$(CONFIG_CIFS_DFS_UPCALL) += dns_resolve.o cifs_dfs_ref.o
diff --git a/fs/cifs/cifs_dfs_ref.c b/fs/cifs/cifs_dfs_ref.c
new file mode 100644
index 000000000000..15e31f8435ba
--- /dev/null
+++ b/fs/cifs/cifs_dfs_ref.c
@@ -0,0 +1,375 @@
+/*
+ *   Contains the CIFS DFS referral mounting routines used for handling
+ *   traversal via DFS junction point
+ *
+ *   Copyright (c) 2007 Igor Mammedov
+ *   Author(s): Igor Mammedov (niallain@gmail.com)
+ *
+ *   This program is free software; you can redistribute it and/or
+ *   modify it under the terms of the GNU General Public License
+ *   as published by the Free Software Foundation; either version
+ *   2 of the License, or (at your option) any later version.
+ */
+
+#include <linux/dcache.h>
+#include <linux/mount.h>
+#include <linux/namei.h>
+#include <linux/vfs.h>
+#include <linux/fs.h>
+#include "cifsglob.h"
+#include "cifsproto.h"
+#include "cifsfs.h"
+#include "dns_resolve.h"
+#include "cifs_debug.h"
+
+LIST_HEAD(cifs_dfs_automount_list);
+
+/*
+ * DFS functions
+*/
+
+void dfs_shrink_umount_helper(struct vfsmount *vfsmnt)
+{
+	mark_mounts_for_expiry(&cifs_dfs_automount_list);
+	mark_mounts_for_expiry(&cifs_dfs_automount_list);
+	shrink_submounts(vfsmnt, &cifs_dfs_automount_list);
+}
+
+/**
+ * cifs_get_share_name	-	extracts share name from UNC
+ * @node_name:	pointer to UNC string
+ *
+ * Extracts sharename form full UNC.
+ * i.e. strips from UNC trailing path that is not part of share
+ * name and fixup missing '\' in the begining of DFS node refferal
+ * if neccessary.
+ * Returns pointer to share name on success or NULL on error.
+ * Caller is responsible for freeing returned string.
+ */
+static char *cifs_get_share_name(const char *node_name)
+{
+	int len;
+	char *UNC;
+	char *pSep;
+
+	len = strlen(node_name);
+	UNC = kmalloc(len+2 /*for term null and additional \ if it's missed */,
+			 GFP_KERNEL);
+	if (!UNC)
+		return NULL;
+
+	/* get share name and server name */
+	if (node_name[1] != '\\') {
+		UNC[0] = '\\';
+		strncpy(UNC+1, node_name, len);
+		len++;
+		UNC[len] = 0;
+	} else {
+		strncpy(UNC, node_name, len);
+		UNC[len] = 0;
+	}
+
+	/* find server name end */
+	pSep = memchr(UNC+2, '\\', len-2);
+	if (!pSep) {
+		cERROR(1, ("%s: no server name end in node name: %s",
+			__FUNCTION__, node_name));
+		kfree(UNC);
+		return NULL;
+	}
+
+	/* find sharename end */
+	pSep++;
+	pSep = memchr(UNC+(pSep-UNC), '\\', len-(pSep-UNC));
+	if (!pSep) {
+		cERROR(1, ("%s:2 cant find share name in node name: %s",
+			__FUNCTION__, node_name));
+		kfree(UNC);
+		return NULL;
+	}
+	/* trim path up to sharename end
+	 *          * now we have share name in UNC */
+	*pSep = 0;
+
+	return UNC;
+}
+
+
+/**
+ * compose_mount_options	-	creates mount options for refferral
+ * @sb_mountdata:	parent/root DFS mount options (template)
+ * @ref_unc:		refferral server UNC
+ * @devname:		pointer for saving device name
+ *
+ * creates mount options for submount based on template options sb_mountdata
+ * and replacing unc,ip,prefixpath options with ones we've got form ref_unc.
+ *
+ * Returns: pointer to new mount options or ERR_PTR.
+ * Caller is responcible for freeing retunrned value if it is not error.
+ */
+char *compose_mount_options(const char *sb_mountdata, const char *ref_unc,
+				char **devname)
+{
+	int rc;
+	char *mountdata;
+	int md_len;
+	char *tkn_e;
+	char *srvIP = NULL;
+	char sep = ',';
+	int off, noff;
+
+	if (sb_mountdata == NULL)
+		return ERR_PTR(-EINVAL);
+
+	*devname = cifs_get_share_name(ref_unc);
+	rc = dns_resolve_server_name_to_ip(*devname, &srvIP);
+	if (rc != 0) {
+		cERROR(1, ("%s: Failed to resolve server part of %s to IP",
+			  __FUNCTION__, *devname));
+		mountdata = ERR_PTR(rc);
+		goto compose_mount_options_out;
+	}
+	md_len = strlen(sb_mountdata) + strlen(srvIP) + strlen(ref_unc) + 3;
+	mountdata = kzalloc(md_len+1, GFP_KERNEL);
+	if (mountdata == NULL) {
+		mountdata = ERR_PTR(-ENOMEM);
+		goto compose_mount_options_out;
+	}
+
+	/* copy all options except of unc,ip,prefixpath */
+	off = 0;
+	if (strncmp(sb_mountdata, "sep=", 4) == 0) {
+			sep = sb_mountdata[4];
+			strncpy(mountdata, sb_mountdata, 5);
+			off += 5;
+	}
+	while ((tkn_e = strchr(sb_mountdata+off, sep))) {
+		noff = (tkn_e - (sb_mountdata+off)) + 1;
+		if (strnicmp(sb_mountdata+off, "unc=", 4) == 0) {
+			off += noff;
+			continue;
+		}
+		if (strnicmp(sb_mountdata+off, "ip=", 3) == 0) {
+			off += noff;
+			continue;
+		}
+		if (strnicmp(sb_mountdata+off, "prefixpath=", 3) == 0) {
+			off += noff;
+			continue;
+		}
+		strncat(mountdata, sb_mountdata+off, noff);
+		off += noff;
+	}
+	strcat(mountdata, sb_mountdata+off);
+	mountdata[md_len] = '\0';
+
+	/* copy new IP and ref share name */
+	strcat(mountdata, ",ip=");
+	strcat(mountdata, srvIP);
+	strcat(mountdata, ",unc=");
+	strcat(mountdata, *devname);
+
+	/* find & copy prefixpath */
+	tkn_e = strchr(ref_unc+2, '\\');
+	if (tkn_e) {
+		tkn_e = strchr(tkn_e+1, '\\');
+		if (tkn_e) {
+			strcat(mountdata, ",prefixpath=");
+			strcat(mountdata, tkn_e);
+		}
+	}
+
+	/*cFYI(1,("%s: parent mountdata: %s", __FUNCTION__,sb_mountdata));*/
+	/*cFYI(1, ("%s: submount mountdata: %s", __FUNCTION__, mountdata ));*/
+
+compose_mount_options_out:
+	kfree(srvIP);
+	return mountdata;
+}
+
+
+struct vfsmount *cifs_dfs_do_refmount(const struct vfsmount *mnt_parent,
+		struct dentry *dentry, char *ref_unc)
+{
+	struct cifs_sb_info *cifs_sb;
+	struct vfsmount *mnt;
+	char *mountdata;
+	char *devname;
+
+	cifs_sb = CIFS_SB(dentry->d_inode->i_sb);
+	mountdata = compose_mount_options(cifs_sb->mountdata,
+						ref_unc, &devname);
+
+	if (IS_ERR(mountdata))
+		return (struct vfsmount *)mountdata;
+
+	mnt = vfs_kern_mount(&cifs_fs_type, 0, devname, mountdata);
+	kfree(mountdata);
+	kfree(devname);
+	return mnt;
+
+}
+
+static char *build_full_dfs_path_from_dentry(struct dentry *dentry)
+{
+	char *full_path = NULL;
+	char *search_path;
+	char *tmp_path;
+	size_t l_max_len;
+	struct cifs_sb_info *cifs_sb;
+
+	if (dentry->d_inode == NULL)
+		return NULL;
+
+	cifs_sb = CIFS_SB(dentry->d_inode->i_sb);
+
+	if (cifs_sb->tcon == NULL)
+		return NULL;
+
+	search_path = build_path_from_dentry(dentry);
+	if (search_path == NULL)
+		return NULL;
+
+	if (cifs_sb->tcon->Flags & SMB_SHARE_IS_IN_DFS) {
+		/* we should use full path name to correct working with DFS */
+		l_max_len = strnlen(cifs_sb->tcon->treeName, MAX_TREE_SIZE+1) +
+					strnlen(search_path, MAX_PATHCONF) + 1;
+		tmp_path = kmalloc(l_max_len, GFP_KERNEL);
+		if (tmp_path == NULL) {
+			kfree(search_path);
+			return NULL;
+		}
+		strncpy(tmp_path, cifs_sb->tcon->treeName, l_max_len);
+		strcat(tmp_path, search_path);
+		tmp_path[l_max_len-1] = 0;
+		full_path = tmp_path;
+		kfree(search_path);
+	} else {
+		full_path = search_path;
+	}
+	return full_path;
+}
+
+static int add_mount_helper(struct vfsmount *newmnt, struct nameidata *nd,
+				struct list_head *mntlist)
+{
+	/* stolen from afs code */
+	int err;
+
+	mntget(newmnt);
+	err = do_add_mount(newmnt, nd, nd->mnt->mnt_flags, mntlist);
+	switch (err) {
+	case 0:
+		dput(nd->dentry);
+		mntput(nd->mnt);
+		nd->mnt = newmnt;
+		nd->dentry = dget(newmnt->mnt_root);
+		break;
+	case -EBUSY:
+		/* someone else made a mount here whilst we were busy */
+		while (d_mountpoint(nd->dentry) &&
+		       follow_down(&nd->mnt, &nd->dentry))
+			;
+		err = 0;
+	default:
+		mntput(newmnt);
+		break;
+	}
+	return err;
+}
+
+void dump_referral(const struct dfs_info3_param *ref)
+{
+	cFYI(1, ("DFS: ref path: %s", ref->path_name));
+	cFYI(1, ("DFS: node path: %s", ref->node_name));
+	cFYI(1, ("DFS: fl: %hd, srv_type: %hd", ref->flags, ref->server_type));
+	cFYI(1, ("DFS: ref_flags: %hd, path_consumed: %hd", ref->ref_flag,
+				ref->PathConsumed));
+}
+
+
+static void*
+cifs_dfs_follow_mountpoint(struct dentry *dentry, struct nameidata *nd)
+{
+	struct dfs_info3_param *referrals = NULL;
+	unsigned int num_referrals = 0;
+	struct cifs_sb_info *cifs_sb;
+	struct cifsSesInfo *ses;
+	char *full_path = NULL;
+	int xid, i;
+	int rc = 0;
+	struct vfsmount *mnt = ERR_PTR(-ENOENT);
+
+	cFYI(1, ("in %s", __FUNCTION__));
+	BUG_ON(IS_ROOT(dentry));
+
+	xid = GetXid();
+
+	dput(nd->dentry);
+	nd->dentry = dget(dentry);
+
+	cifs_sb = CIFS_SB(dentry->d_inode->i_sb);
+	ses = cifs_sb->tcon->ses;
+
+	if (!ses) {
+		rc = -EINVAL;
+		goto out_err;
+	}
+
+	full_path = build_full_dfs_path_from_dentry(dentry);
+	if (full_path == NULL) {
+		rc = -ENOMEM;
+		goto out_err;
+	}
+
+	rc = get_dfs_path(xid, ses , full_path, cifs_sb->local_nls,
+		&num_referrals, &referrals,
+		cifs_sb->mnt_cifs_flags & CIFS_MOUNT_MAP_SPECIAL_CHR);
+
+	for (i = 0; i < num_referrals; i++) {
+		dump_referral(referrals+i);
+		/* connect to a storage node */
+		if (referrals[i].flags & DFSREF_STORAGE_SERVER) {
+			int len;
+			len = strlen(referrals[i].node_name);
+			if (len < 2) {
+				cERROR(1, ("%s: Net Address path too short: %s",
+					__FUNCTION__, referrals[i].node_name));
+				rc = -EINVAL;
+				goto out_err;
+			}
+			mnt = cifs_dfs_do_refmount(nd->mnt, nd->dentry,
+						referrals[i].node_name);
+			cFYI(1, ("%s: cifs_dfs_do_refmount:%s , mnt:%p",
+					 __FUNCTION__,
+					referrals[i].node_name, mnt));
+
+			/* complete mount procedure if we accured submount */
+			if (!IS_ERR(mnt))
+				break;
+		}
+	}
+
+	/* we need it cause for() above could exit without valid submount */
+	rc = PTR_ERR(mnt);
+	if (IS_ERR(mnt))
+		goto out_err;
+
+	nd->mnt->mnt_flags |= MNT_SHRINKABLE;
+	rc = add_mount_helper(mnt, nd, &cifs_dfs_automount_list);
+
+out:
+	FreeXid(xid);
+	free_dfs_info_array(referrals, num_referrals);
+	kfree(full_path);
+	cFYI(1, ("leaving %s" , __FUNCTION__));
+	return ERR_PTR(rc);
+out_err:
+	path_release(nd);
+	goto out;
+}
+
+struct inode_operations cifs_dfs_referral_inode_operations = {
+	.follow_link = cifs_dfs_follow_mountpoint,
+};
+
diff --git a/fs/cifs/cifsfs.c b/fs/cifs/cifsfs.c
index 93e107883a61..e9f4ec701092 100644
--- a/fs/cifs/cifsfs.c
+++ b/fs/cifs/cifsfs.c
@@ -471,6 +471,10 @@ static void cifs_umount_begin(struct vfsmount *vfsmnt, int flags)
 	struct cifs_sb_info *cifs_sb;
 	struct cifsTconInfo *tcon;
 
+#ifdef CONFIG_CIFS_DFS_UPCALL
+	dfs_shrink_umount_helper(vfsmnt);
+#endif /* CONFIG CIFS_DFS_UPCALL */
+
 	if (!(flags & MNT_FORCE))
 		return;
 	cifs_sb = CIFS_SB(vfsmnt->mnt_sb);
diff --git a/fs/cifs/cifsfs.h b/fs/cifs/cifsfs.h
index 2e68126d07eb..195b14de5567 100644
--- a/fs/cifs/cifsfs.h
+++ b/fs/cifs/cifsfs.h
@@ -61,6 +61,10 @@ extern int cifs_setattr(struct dentry *, struct iattr *);
 
 extern const struct inode_operations cifs_file_inode_ops;
 extern const struct inode_operations cifs_symlink_inode_ops;
+extern struct list_head cifs_dfs_automount_list;
+extern struct inode_operations cifs_dfs_referral_inode_operations;
+
+
 
 /* Functions related to files and directories */
 extern const struct file_operations cifs_file_ops;
diff --git a/fs/cifs/cifsproto.h b/fs/cifs/cifsproto.h
index 7093cb4b0212..aaaf748f6a26 100644
--- a/fs/cifs/cifsproto.h
+++ b/fs/cifs/cifsproto.h
@@ -102,6 +102,9 @@ extern int mode_to_acl(struct inode *inode, const char *path, __u64);
 extern int cifs_mount(struct super_block *, struct cifs_sb_info *, char *,
 			const char *);
 extern int cifs_umount(struct super_block *, struct cifs_sb_info *);
+#ifdef CONFIG_CIFS_DFS_UPCALL
+extern void dfs_shrink_umount_helper(struct vfsmount *vfsmnt);
+#endif
 void cifs_proc_init(void);
 void cifs_proc_clean(void);