kernfs: use 64bit inos if ino_t is 64bit

Each kernfs_node is identified with a 64bit ID.  The low 32bit is
exposed as ino and the high gen.  While this already allows using inos
as keys by looking up with wildcard generation number of 0, it's
adding unnecessary complications for 64bit ino archs which can
directly use kernfs_node IDs as inos to uniquely identify each cgroup
instance.

This patch exposes IDs directly as inos on 64bit ino archs.  The
conversion is mostly straight-forward.

* 32bit ino archs behave the same as before.  64bit ino archs now use
  the whole 64bit ID as ino and the generation number is fixed at 1.

* 64bit inos still use the same idr allocator which gurantees that the
  lower 32bits identify the current live instance uniquely and the
  high 32bits are incremented whenever the low bits wrap.  As the
  upper 32bits are no longer used as gen and we don't wanna start ino
  allocation with 33rd bit set, the initial value for highbits
  allocation is changed to 0 on 64bit ino archs.

* blktrace exposes two 32bit numbers - (INO,GEN) pair - to identify
  the issuing cgroup.  Userland builds FILEID_INO32_GEN fids from
  these numbers to look up the cgroups.  To remain compatible with the
  behavior, always output (LOW32,HIGH32) which will be constructed
  back to the original 64bit ID by __kernfs_fh_to_dentry().

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>

fs/kernfs/dir.c

@@ -532,7 +532,7 @@ void kernfs_put(struct kernfs_node *kn)
 		kmem_cache_free(kernfs_iattrs_cache, kn->iattr);
 	}
 	spin_lock(&kernfs_idr_lock);
-	idr_remove(&root->ino_idr, kernfs_ino(kn));
+	idr_remove(&root->ino_idr, (u32)kernfs_ino(kn));
 	spin_unlock(&kernfs_idr_lock);
 	kmem_cache_free(kernfs_node_cache, kn);
 
@@ -617,7 +617,7 @@ static struct kernfs_node *__kernfs_new_node(struct kernfs_root *root,
 					     unsigned flags)
 {
 	struct kernfs_node *kn;
-	u32 gen;
+	u32 id_highbits;
 	int ret;
 
 	name = kstrdup_const(name, GFP_KERNEL);
@@ -631,16 +631,16 @@ static struct kernfs_node *__kernfs_new_node(struct kernfs_root *root,
 	idr_preload(GFP_KERNEL);
 	spin_lock(&kernfs_idr_lock);
 	ret = idr_alloc_cyclic(&root->ino_idr, kn, 1, 0, GFP_ATOMIC);
-	if (ret >= 0 && ret < root->last_ino)
-		root->next_generation++;
-	gen = root->next_generation;
-	root->last_ino = ret;
+	if (ret >= 0 && ret < root->last_id_lowbits)
+		root->id_highbits++;
+	id_highbits = root->id_highbits;
+	root->last_id_lowbits = ret;
 	spin_unlock(&kernfs_idr_lock);
 	idr_preload_end();
 	if (ret < 0)
 		goto err_out2;
 
-	kn->id = (u64)gen << 32 | ret;
+	kn->id = (u64)id_highbits << 32 | ret;
 
 	atomic_set(&kn->count, 1);
 	atomic_set(&kn->active, KN_DEACTIVATED_BIAS);
@@ -671,7 +671,7 @@ static struct kernfs_node *__kernfs_new_node(struct kernfs_root *root,
 	return kn;
 
  err_out3:
-	idr_remove(&root->ino_idr, kernfs_ino(kn));
+	idr_remove(&root->ino_idr, (u32)kernfs_ino(kn));
  err_out2:
 	kmem_cache_free(kernfs_node_cache, kn);
  err_out1:
@@ -715,13 +715,19 @@ struct kernfs_node *kernfs_find_and_get_node_by_id(struct kernfs_root *root,
 
 	spin_lock(&kernfs_idr_lock);
 
-	kn = idr_find(&root->ino_idr, ino);
+	kn = idr_find(&root->ino_idr, (u32)ino);
 	if (!kn)
 		goto err_unlock;
 
-	/* 0 matches all generations */
-	if (unlikely(gen && kernfs_gen(kn) != gen))
-		goto err_unlock;
+	if (sizeof(ino_t) >= sizeof(u64)) {
+		/* we looked up with the low 32bits, compare the whole */
+		if (kernfs_ino(kn) != ino)
+			goto err_unlock;
+	} else {
+		/* 0 matches all generations */
+		if (unlikely(gen && kernfs_gen(kn) != gen))
+			goto err_unlock;
+	}
 
 	/*
 	 * ACTIVATED is protected with kernfs_mutex but it was clear when
@@ -949,7 +955,17 @@ struct kernfs_root *kernfs_create_root(struct kernfs_syscall_ops *scops,
 
 	idr_init(&root->ino_idr);
 	INIT_LIST_HEAD(&root->supers);
-	root->next_generation = 1;
+
+	/*
+	 * On 64bit ino setups, id is ino.  On 32bit, low 32bits are ino.
+	 * High bits generation.  The starting value for both ino and
+	 * genenration is 1.  Initialize upper 32bit allocation
+	 * accordingly.
+	 */
+	if (sizeof(ino_t) >= sizeof(u64))
+		root->id_highbits = 0;
+	else
+		root->id_highbits = 1;
 
 	kn = __kernfs_new_node(root, NULL, "", S_IFDIR | S_IRUGO | S_IXUGO,
 			       GLOBAL_ROOT_UID, GLOBAL_ROOT_GID,

fs/kernfs/mount.c

@@ -87,9 +87,10 @@ static struct dentry *__kernfs_fh_to_dentry(struct super_block *sb,
 	case FILEID_INO32_GEN:
 	case FILEID_INO32_GEN_PARENT:
 		/*
-		 * blk_log_action() exposes (ino,gen) pair without type and
-		 * userland can call us with generic fid constructed from
-		 * them.  Combine it back to ID.  See blk_log_action().
+		 * blk_log_action() exposes "LOW32,HIGH32" pair without
+		 * type and userland can call us with generic fid
+		 * constructed from them.  Combine it back to ID.  See
+		 * blk_log_action().
 		 */
 		id = ((u64)fid->i32.gen << 32) | fid->i32.ino;
 		break;

include/linux/kernfs.h

@@ -141,8 +141,8 @@ struct kernfs_node {
 	void			*priv;
 
 	/*
-	 * 64bit unique ID.  Lower 32bits carry the inode number and lower
-	 * generation.
+	 * 64bit unique ID.  On 64bit ino setups, id is the ino.  On 32bit,
+	 * the low 32bits are ino and upper generation.
 	 */
 	u64			id;
 
@@ -177,8 +177,8 @@ struct kernfs_root {
 
 	/* private fields, do not use outside kernfs proper */
 	struct idr		ino_idr;
-	u32			last_ino;
-	u32			next_generation;
+	u32			last_id_lowbits;
+	u32			id_highbits;
 	struct kernfs_syscall_ops *syscall_ops;
 
 	/* list of kernfs_super_info of this root, protected by kernfs_mutex */
@@ -284,12 +284,20 @@ static inline enum kernfs_node_type kernfs_type(struct kernfs_node *kn)
 
 static inline ino_t kernfs_id_ino(u64 id)
 {
-	return (u32)id;
+	/* id is ino if ino_t is 64bit; otherwise, low 32bits */
+	if (sizeof(ino_t) >= sizeof(u64))
+		return id;
+	else
+		return (u32)id;
 }
 
 static inline u32 kernfs_id_gen(u64 id)
 {
-	return id >> 32;
+	/* gen is fixed at 1 if ino_t is 64bit; otherwise, high 32bits */
+	if (sizeof(ino_t) >= sizeof(u64))
+		return 1;
+	else
+		return id >> 32;
 }
 
 static inline ino_t kernfs_ino(struct kernfs_node *kn)

kernel/trace/blktrace.c

@@ -1261,12 +1261,25 @@ static void blk_log_action(struct trace_iterator *iter, const char *act,
 			trace_seq_printf(&iter->seq, "%3d,%-3d %s %2s %3s ",
 				 MAJOR(t->device), MINOR(t->device),
 				 blkcg_name_buf, act, rwbs);
-		} else
+		} else {
+			/*
+			 * The cgid portion used to be "INO,GEN".  Userland
+			 * builds a FILEID_INO32_GEN fid out of them and
+			 * opens the cgroup using open_by_handle_at(2).
+			 * While 32bit ino setups are still the same, 64bit
+			 * ones now use the 64bit ino as the whole ID and
+			 * no longer use generation.
+			 *
+			 * Regarldess of the content, always output
+			 * "LOW32,HIGH32" so that FILEID_INO32_GEN fid can
+			 * be mapped back to @id on both 64 and 32bit ino
+			 * setups.  See __kernfs_fh_to_dentry().
+			 */
 			trace_seq_printf(&iter->seq,
-				 "%3d,%-3d %lx,%-x %2s %3s ",
+				 "%3d,%-3d %llx,%-llx %2s %3s ",
 				 MAJOR(t->device), MINOR(t->device),
-				 kernfs_id_ino(id), kernfs_id_gen(id),
-				 act, rwbs);
+				 id & U32_MAX, id >> 32, act, rwbs);
+		}
 	} else
 		trace_seq_printf(&iter->seq, "%3d,%-3d %2s %3s ",
 				 MAJOR(t->device), MINOR(t->device), act, rwbs);