NFSv4: Ensure client submounts when following a referral

Set up mountpoint when hitting a referral on moved error by getting
fs_locations.

Signed-off-by: Manoj Naik <manoj@almaden.ibm.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>

2 files changed, 269 insertions, 3 deletions

diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c
index ebdab885c475..0d8302e59d69 100644
--- a/fs/nfs/inode.c
+++ b/fs/nfs/inode.c
@@ -36,6 +36,8 @@
 #include <linux/mount.h>
 #include <linux/nfs_idmap.h>
 #include <linux/vfs.h>
+#include <linux/inet.h>
+#include <linux/nfs_xdr.h>
 
 #include <asm/system.h>
 #include <asm/uaccess.h>
@@ -1714,6 +1716,10 @@ struct nfs_clone_mount {
 	const struct dentry *dentry;
 	struct nfs_fh *fh;
 	struct nfs_fattr *fattr;
+	char *hostname;
+	char *mnt_path;
+	struct sockaddr_in *addr;
+	rpc_authflavor_t authflavor;
 };
 
 static struct super_block *nfs_clone_generic_sb(struct nfs_clone_mount *data,
@@ -1724,17 +1730,19 @@ static struct super_block *nfs_clone_generic_sb(struct nfs_clone_mount *data,
 	struct nfs_server *parent = NFS_SB(data->sb);
 	struct super_block *sb = ERR_PTR(-EINVAL);
 	void *err = ERR_PTR(-ENOMEM);
+	char *hostname;
 	int len;
 
 	server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
 	if (server == NULL)
 		goto out_err;
 	memcpy(server, parent, sizeof(*server));
-	len = strlen(parent->hostname) + 1;
+	hostname = (data->hostname != NULL) ? data->hostname : parent->hostname;
+	len = strlen(hostname) + 1;
 	server->hostname = kmalloc(len, GFP_KERNEL);
 	if (server->hostname == NULL)
 		goto free_server;
-	memcpy(server->hostname, parent->hostname, len);
+	memcpy(server->hostname, hostname, len);
 	if (rpciod_up() != 0)
 		goto free_hostname;
 
@@ -2458,7 +2466,8 @@ static inline void unregister_nfs4fs(void)
 	nfs_unregister_sysctl();
 }
 #else
-#define nfs4_clone_client(a,b) ERR_PTR(-EINVAL)
+#define nfs4_fill_sb(a,b)	ERR_PTR(-EINVAL)
+#define nfs4_fill_super(a,b)	ERR_PTR(-EINVAL)
 #define nfs4_init_once(nfsi) \
 	do { } while (0)
 #define register_nfs4fs() (0)
@@ -2521,6 +2530,261 @@ out:
 	return mnt;
 }
 
+/* Check if fs_root is valid */
+static inline char *nfs4_pathname_string(struct nfs4_pathname *pathname, char *buffer, ssize_t buflen)
+{
+	char *end = buffer + buflen;
+	int n;
+
+	*--end = '\0';
+	buflen--;
+
+	n = pathname->ncomponents;
+	while (--n >= 0) {
+		struct nfs4_string *component = &pathname->components[n];
+		buflen -= component->len + 1;
+		if (buflen < 0)
+			goto Elong;
+		end -= component->len;
+		memcpy(end, component->data, component->len);
+		*--end = '/';
+	}
+	return end;
+Elong:
+	return ERR_PTR(-ENAMETOOLONG);
+}
+
+/* Check if the string represents a "valid" IPv4 address */
+static inline int valid_ipaddr4(const char *buf)
+{
+	int rc, count, in[4];
+
+	rc = sscanf(buf, "%d.%d.%d.%d", &in[0], &in[1], &in[2], &in[3]);
+	if (rc != 4)
+		return -EINVAL;
+	for (count = 0; count < 4; count++) {
+		if (in[count] > 255)
+			return -EINVAL;
+	}
+	return 0;
+}
+
+static struct super_block *nfs4_referral_sb(struct nfs_server *server, struct nfs_clone_mount *data)
+{
+	struct super_block *sb = ERR_PTR(-ENOMEM);
+	int len;
+
+	len = strlen(data->mnt_path) + 1;
+	server->mnt_path = kmalloc(len, GFP_KERNEL);
+	if (server->mnt_path == NULL)
+		goto err;
+	memcpy(server->mnt_path, data->mnt_path, len);
+	memcpy(&server->addr, data->addr, sizeof(struct sockaddr_in));
+
+	sb = sget(&nfs4_fs_type, nfs4_compare_super, nfs_set_super, server);
+	if (IS_ERR(sb) || sb->s_root)
+		goto free_path;
+	return sb;
+free_path:
+	kfree(server->mnt_path);
+err:
+	server->mnt_path = NULL;
+	return sb;
+}
+
+static struct nfs_server *nfs4_referral_server(struct super_block *sb, struct nfs_clone_mount *data)
+{
+	struct nfs_server *server = NFS_SB(sb);
+	struct rpc_timeout timeparms;
+	int proto, timeo, retrans;
+	void *err;
+
+	proto = IPPROTO_TCP;
+	/* Since we are following a referral and there may be alternatives,
+	   set the timeouts and retries to low values */
+	timeo = 2;
+	retrans = 1;
+	nfs_init_timeout_values(&timeparms, proto, timeo, retrans);
+
+	server->client = nfs4_create_client(server, &timeparms, proto, data->authflavor);
+	if (IS_ERR((err = server->client)))
+		goto out_err;
+
+	sb->s_time_gran = 1;
+	sb->s_op = &nfs4_sops;
+	err = ERR_PTR(nfs_sb_init(sb, data->authflavor));
+	if (!IS_ERR(err))
+		return server;
+out_err:
+	return (struct nfs_server *)err;
+}
+
+static struct super_block *nfs_referral_nfs4_sb(struct file_system_type *fs_type,
+		int flags, const char *dev_name, void *raw_data)
+{
+	struct nfs_clone_mount *data = raw_data;
+	return nfs_clone_generic_sb(data, nfs4_referral_sb, nfs4_referral_server);
+}
+
+static struct file_system_type nfs_referral_nfs4_fs_type = {
+	.owner		= THIS_MODULE,
+	.name		= "nfs4",
+	.get_sb		= nfs_referral_nfs4_sb,
+	.kill_sb	= nfs4_kill_super,
+	.fs_flags	= FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
+};
+
+/**
+ * nfs_follow_referral - set up mountpoint when hitting a referral on moved error
+ * @mnt_parent - mountpoint of parent directory
+ * @dentry - parent directory
+ * @fspath - fs path returned in fs_locations
+ * @mntpath - mount path to new server
+ * @hostname - hostname of new server
+ * @addr - host addr of new server
+ *
+ */
+struct vfsmount *nfs_follow_referral(const struct vfsmount *mnt_parent,
+	     const struct dentry *dentry, struct nfs4_fs_locations *locations)
+{
+	struct vfsmount *mnt = ERR_PTR(-ENOENT);
+	struct nfs_clone_mount mountdata = {
+		.sb = mnt_parent->mnt_sb,
+		.dentry = dentry,
+		.authflavor = NFS_SB(mnt_parent->mnt_sb)->client->cl_auth->au_flavor,
+	};
+	char *page, *page2;
+	char *path, *fs_path;
+	char *devname;
+	int loc, s;
+
+	if (locations == NULL || locations->nlocations <= 0)
+		goto out;
+
+	dprintk("%s: referral at %s/%s\n", __FUNCTION__,
+		dentry->d_parent->d_name.name, dentry->d_name.name);
+
+	/* Ensure fs path is a prefix of current dentry path */
+	page = (char *) __get_free_page(GFP_USER);
+	if (page == NULL)
+		goto out;
+	page2 = (char *) __get_free_page(GFP_USER);
+	if (page2 == NULL)
+		goto out;
+
+	path = nfs4_path(dentry, page, PAGE_SIZE);
+	if (IS_ERR(path))
+		goto out_free;
+
+	fs_path = nfs4_pathname_string(&locations->fs_path, page2, PAGE_SIZE);
+	if (IS_ERR(fs_path))
+		goto out_free;
+
+	if (strncmp(path, fs_path, strlen(fs_path)) != 0) {
+		dprintk("%s: path %s does not begin with fsroot %s\n", __FUNCTION__, path, fs_path);
+		goto out_free;
+	}
+
+	devname = nfs_devname(mnt_parent, dentry, page, PAGE_SIZE);
+	if (IS_ERR(devname)) {
+		mnt = (struct vfsmount *)devname;
+		goto out_free;
+	}
+
+	loc = 0;
+	while (loc < locations->nlocations && IS_ERR(mnt)) {
+		struct nfs4_fs_location *location = &locations->locations[loc];
+		char *mnt_path;
+
+		if (location == NULL || location->nservers <= 0 ||
+		    location->rootpath.ncomponents == 0) {
+			loc++;
+			continue;
+		}
+
+		mnt_path = nfs4_pathname_string(&location->rootpath, page2, PAGE_SIZE);
+		if (IS_ERR(mnt_path)) {
+			loc++;
+			continue;
+		}
+		mountdata.mnt_path = mnt_path;
+
+		s = 0;
+		while (s < location->nservers) {
+			struct sockaddr_in addr = {};
+
+			if (location->servers[s].len <= 0 ||
+			    valid_ipaddr4(location->servers[s].data) < 0) {
+				s++;
+				continue;
+			}
+
+			mountdata.hostname = location->servers[s].data;
+			addr.sin_addr.s_addr = in_aton(mountdata.hostname);
+			addr.sin_family = AF_INET;
+			addr.sin_port = htons(NFS_PORT);
+			mountdata.addr = &addr;
+
+			mnt = vfs_kern_mount(&nfs_referral_nfs4_fs_type, 0, devname, &mountdata);
+			if (!IS_ERR(mnt)) {
+				break;
+			}
+			s++;
+		}
+		loc++;
+	}
+
+out_free:
+	free_page((unsigned long)page);
+	free_page((unsigned long)page2);
+out:
+	dprintk("%s: done\n", __FUNCTION__);
+	return mnt;
+}
+
+/*
+ * nfs_do_refmount - handle crossing a referral on server
+ * @dentry - dentry of referral
+ * @nd - nameidata info
+ *
+ */
+struct vfsmount *nfs_do_refmount(const struct vfsmount *mnt_parent, struct dentry *dentry)
+{
+	struct vfsmount *mnt = ERR_PTR(-ENOENT);
+	struct dentry *parent;
+	struct nfs4_fs_locations *fs_locations = NULL;
+	struct page *page;
+	int err;
+
+	/* BUG_ON(IS_ROOT(dentry)); */
+	dprintk("%s: enter\n", __FUNCTION__);
+
+	page = alloc_page(GFP_KERNEL);
+	if (page == NULL)
+		goto out;
+
+	fs_locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
+	if (fs_locations == NULL)
+		goto out_free;
+
+	/* Get locations */
+	parent = dget_parent(dentry);
+	dprintk("%s: getting locations for %s/%s\n", __FUNCTION__, parent->d_name.name, dentry->d_name.name);
+	err = nfs4_proc_fs_locations(parent->d_inode, dentry, fs_locations, page);
+	dput(parent);
+	if (err != 0 || fs_locations->nlocations <= 0 ||
+	    fs_locations->fs_path.ncomponents <= 0)
+		goto out_free;
+
+	mnt = nfs_follow_referral(mnt_parent, dentry, fs_locations);
+out_free:
+	__free_page(page);
+	kfree(fs_locations);
+out:
+	dprintk("%s: done\n", __FUNCTION__);
+	return mnt;
+}
+
 extern int nfs_init_nfspagecache(void);
 extern void nfs_destroy_nfspagecache(void);
 extern int nfs_init_readpagecache(void);
diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h
index a34b3ee443f1..09271b10f9a8 100644
--- a/include/linux/nfs_fs.h
+++ b/include/linux/nfs_fs.h
@@ -317,6 +317,8 @@ extern struct vfsmount *nfs_do_submount(const struct vfsmount *mnt_parent,
 					const struct dentry *dentry,
 					struct nfs_fh *fh,
 					struct nfs_fattr *fattr);
+extern struct vfsmount *nfs_do_refmount(const struct vfsmount *mnt_parent,
+					struct dentry *dentry);
 
 /* linux/net/ipv4/ipconfig.c: trims ip addr off front of name, too. */
 extern u32 root_nfs_parse_addr(char *name); /*__init*/