ceph: new mount option that specifies fscache uniquifier

Current ceph uses FSID as primary index key of fscache data. This
allows ceph to retain cached data across remount. But this causes
problem (kernel opps, fscache does not support sharing data) when
a filesystem get mounted several times (with fscache enabled, with
different mount options).

The fix is adding a new mount option, which specifies uniquifier
for fscache.
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Acked-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>

fs/ceph/cache.c

@@ -35,18 +35,34 @@ struct fscache_netfs ceph_cache_netfs = {
 	.version	= 0,
 };
 
+static DEFINE_MUTEX(ceph_fscache_lock);
+static LIST_HEAD(ceph_fscache_list);
+
+struct ceph_fscache_entry {
+	struct list_head list;
+	struct fscache_cookie *fscache;
+	struct ceph_fsid fsid;
+	size_t uniq_len;
+	char uniquifier[0];
+};
+
 static uint16_t ceph_fscache_session_get_key(const void *cookie_netfs_data,
 					     void *buffer, uint16_t maxbuf)
 {
 	const struct ceph_fs_client* fsc = cookie_netfs_data;
-	uint16_t klen;
+	const char *fscache_uniq = fsc->mount_options->fscache_uniq;
+	uint16_t fsid_len, uniq_len;
 
-	klen = sizeof(fsc->client->fsid);
-	if (klen > maxbuf)
+	fsid_len = sizeof(fsc->client->fsid);
+	uniq_len = fscache_uniq ? strlen(fscache_uniq) : 0;
+	if (fsid_len + uniq_len > maxbuf)
 		return 0;
 
-	memcpy(buffer, &fsc->client->fsid, klen);
-	return klen;
+	memcpy(buffer, &fsc->client->fsid, fsid_len);
+	if (uniq_len)
+		memcpy(buffer + fsid_len, fscache_uniq, uniq_len);
+
+	return fsid_len + uniq_len;
 }
 
 static const struct fscache_cookie_def ceph_fscache_fsid_object_def = {
@@ -67,13 +83,54 @@ void ceph_fscache_unregister(void)
 
 int ceph_fscache_register_fs(struct ceph_fs_client* fsc)
 {
+	const struct ceph_fsid *fsid = &fsc->client->fsid;
+	const char *fscache_uniq = fsc->mount_options->fscache_uniq;
+	size_t uniq_len = fscache_uniq ? strlen(fscache_uniq) : 0;
+	struct ceph_fscache_entry *ent;
+	int err = 0;
+
+	mutex_lock(&ceph_fscache_lock);
+	list_for_each_entry(ent, &ceph_fscache_list, list) {
+		if (memcmp(&ent->fsid, fsid, sizeof(*fsid)))
+			continue;
+		if (ent->uniq_len != uniq_len)
+			continue;
+		if (uniq_len && memcmp(ent->uniquifier, fscache_uniq, uniq_len))
+			continue;
+
+		pr_err("fscache cookie already registered for fsid %pU\n", fsid);
+		pr_err("  use fsc=%%s mount option to specify a uniquifier\n");
+		err = -EBUSY;
+		goto out_unlock;
+	}
+
+	ent = kzalloc(sizeof(*ent) + uniq_len, GFP_KERNEL);
+	if (!ent) {
+		err = -ENOMEM;
+		goto out_unlock;
+	}
+
 	fsc->fscache = fscache_acquire_cookie(ceph_cache_netfs.primary_index,
 					      &ceph_fscache_fsid_object_def,
 					      fsc, true);
-	if (!fsc->fscache)
-		pr_err("Unable to register fsid: %p fscache cookie\n", fsc);
 
-	return 0;
+	if (fsc->fscache) {
+		memcpy(&ent->fsid, fsid, sizeof(*fsid));
+		if (uniq_len > 0) {
+			memcpy(&ent->uniquifier, fscache_uniq, uniq_len);
+			ent->uniq_len = uniq_len;
+		}
+		ent->fscache = fsc->fscache;
+		list_add_tail(&ent->list, &ceph_fscache_list);
+	} else {
+		kfree(ent);
+		pr_err("unable to register fscache cookie for fsid %pU\n",
+		       fsid);
+		/* all other fs ignore this error */
+	}
+out_unlock:
+	mutex_unlock(&ceph_fscache_lock);
+	return err;
 }
 
 static uint16_t ceph_fscache_inode_get_key(const void *cookie_netfs_data,
@@ -349,7 +406,24 @@ void ceph_invalidate_fscache_page(struct inode* inode, struct page *page)
 
 void ceph_fscache_unregister_fs(struct ceph_fs_client* fsc)
 {
-	fscache_relinquish_cookie(fsc->fscache, 0);
+	if (fscache_cookie_valid(fsc->fscache)) {
+		struct ceph_fscache_entry *ent;
+		bool found = false;
+
+		mutex_lock(&ceph_fscache_lock);
+		list_for_each_entry(ent, &ceph_fscache_list, list) {
+			if (ent->fscache == fsc->fscache) {
+				list_del(&ent->list);
+				kfree(ent);
+				found = true;
+				break;
+			}
+		}
+		WARN_ON_ONCE(!found);
+		mutex_unlock(&ceph_fscache_lock);
+
+		__fscache_relinquish_cookie(fsc->fscache, 0);
+	}
 	fsc->fscache = NULL;
 }
 

fs/ceph/super.c

@@ -121,6 +121,7 @@ enum {
 	/* int args above */
 	Opt_snapdirname,
 	Opt_mds_namespace,
+	Opt_fscache_uniq,
 	Opt_last_string,
 	/* string args above */
 	Opt_dirstat,
@@ -158,6 +159,7 @@ static match_table_t fsopt_tokens = {
 	/* int args above */
 	{Opt_snapdirname, "snapdirname=%s"},
 	{Opt_mds_namespace, "mds_namespace=%s"},
+	{Opt_fscache_uniq, "fsc=%s"},
 	/* string args above */
 	{Opt_dirstat, "dirstat"},
 	{Opt_nodirstat, "nodirstat"},
@@ -223,6 +225,14 @@ static int parse_fsopt_token(char *c, void *private)
 		if (!fsopt->mds_namespace)
 			return -ENOMEM;
 		break;
+	case Opt_fscache_uniq:
+		fsopt->fscache_uniq = kstrndup(argstr[0].from,
+					       argstr[0].to-argstr[0].from,
+					       GFP_KERNEL);
+		if (!fsopt->fscache_uniq)
+			return -ENOMEM;
+		fsopt->flags |= CEPH_MOUNT_OPT_FSCACHE;
+		break;
 		/* misc */
 	case Opt_wsize:
 		fsopt->wsize = intval;
@@ -317,6 +327,7 @@ static void destroy_mount_options(struct ceph_mount_options *args)
 	kfree(args->snapdir_name);
 	kfree(args->mds_namespace);
 	kfree(args->server_path);
+	kfree(args->fscache_uniq);
 	kfree(args);
 }
 
@@ -350,8 +361,10 @@ static int compare_mount_options(struct ceph_mount_options *new_fsopt,
 	ret = strcmp_null(fsopt1->mds_namespace, fsopt2->mds_namespace);
 	if (ret)
 		return ret;
-
 	ret = strcmp_null(fsopt1->server_path, fsopt2->server_path);
+	if (ret)
+		return ret;
+	ret = strcmp_null(fsopt1->fscache_uniq, fsopt2->fscache_uniq);
 	if (ret)
 		return ret;
 
@@ -475,8 +488,12 @@ static int ceph_show_options(struct seq_file *m, struct dentry *root)
 		seq_puts(m, ",noasyncreaddir");
 	if ((fsopt->flags & CEPH_MOUNT_OPT_DCACHE) == 0)
 		seq_puts(m, ",nodcache");
-	if (fsopt->flags & CEPH_MOUNT_OPT_FSCACHE)
+	if (fsopt->flags & CEPH_MOUNT_OPT_FSCACHE) {
+		if (fsopt->fscache_uniq)
+			seq_printf(m, ",fsc=%s", fsopt->fscache_uniq);
+		else
 			seq_puts(m, ",fsc");
+	}
 	if (fsopt->flags & CEPH_MOUNT_OPT_NOPOOLPERM)
 		seq_puts(m, ",nopoolperm");
 
@@ -597,18 +614,11 @@ static struct ceph_fs_client *create_fs_client(struct ceph_mount_options *fsopt,
 	if (!fsc->wb_pagevec_pool)
 		goto fail_trunc_wq;
 
-	/* setup fscache */
-	if ((fsopt->flags & CEPH_MOUNT_OPT_FSCACHE) &&
-	    (ceph_fscache_register_fs(fsc) != 0))
-		goto fail_fscache;
-
 	/* caps */
 	fsc->min_caps = fsopt->max_readdir;
 
 	return fsc;
 
-fail_fscache:
-	ceph_fscache_unregister_fs(fsc);
 fail_trunc_wq:
 	destroy_workqueue(fsc->trunc_wq);
 fail_pg_inv_wq:
@@ -626,8 +636,6 @@ static void destroy_fs_client(struct ceph_fs_client *fsc)
 {
 	dout("destroy_fs_client %p\n", fsc);
 
-	ceph_fscache_unregister_fs(fsc);
-
 	destroy_workqueue(fsc->wb_wq);
 	destroy_workqueue(fsc->pg_inv_wq);
 	destroy_workqueue(fsc->trunc_wq);
@@ -820,6 +828,13 @@ static struct dentry *ceph_real_mount(struct ceph_fs_client *fsc)
 		if (err < 0)
 			goto out;
 
+		/* setup fscache */
+		if (fsc->mount_options->flags & CEPH_MOUNT_OPT_FSCACHE) {
+			err = ceph_fscache_register_fs(fsc);
+			if (err < 0)
+				goto out;
+		}
+
 		if (!fsc->mount_options->server_path) {
 			path = "";
 			dout("mount opening path \\t\n");
@@ -1042,6 +1057,8 @@ static void ceph_kill_sb(struct super_block *s)
 	fsc->client->extra_mon_dispatch = NULL;
 	ceph_fs_debugfs_cleanup(fsc);
 
+	ceph_fscache_unregister_fs(fsc);
+
 	ceph_mdsc_destroy(fsc);
 
 	destroy_fs_client(fsc);

fs/ceph/super.h

@@ -73,6 +73,7 @@ struct ceph_mount_options {
 	char *snapdir_name;   /* default ".snap" */
 	char *mds_namespace;  /* default NULL */
 	char *server_path;    /* default  "/" */
+	char *fscache_uniq;   /* default NULL */
 };
 
 struct ceph_fs_client {