[PATCH v4 3/5] fcntl: add F_{SET/GET}_RW_HINT_EX

[Kanchan Joshi]

This is similar to existing F_{SET/GET}_RW_HINT but more
generic/extensible.

F_SET/GET_RW_HINT_EX take a pointer to a struct rw_hint_ex as argument:

struct rw_hint_ex {
        __u8    type;
        __u8    pad[7];
        __u64   val;
};

With F_SET_RW_HINT_EX, the user passes the hint type and its value.
Hint type can be either life hint (TYPE_RW_LIFE_HINT) or placement hint
(TYPE_RW_PLACEMENT_HINT). The interface allows to add more hint types in
future.

Valid values for life hints are same as enforced by existing
fcntl(F_SET_RW_HINT).
Valid values for placement hints are between 0 to 127, both inclusive.

The inode retains either the life hint or the placement hint, whichever
is set later. The set hint type and its value can be queried by
F_GET_RW_HINT_EX.

The i_write_hint field of the inode is a 1-byte field. Use the most
significant bit as the hint type. This bit is set for placement hint.
For life hint, this bit remains zero.

Signed-off-by: Kanchan Joshi <joshi.k at samsung.com>
Signed-off-by: Nitesh Shetty <nj.shetty at samsung.com>
---
 fs/fcntl.c                 | 67 ++++++++++++++++++++++++++++++++++++++
 include/linux/rw_hint.h    | 13 ++++++++
 include/uapi/linux/fcntl.h | 14 ++++++++
 3 files changed, 94 insertions(+)

diff --git a/fs/fcntl.c b/fs/fcntl.c
index 46cf08f67278..d82fd4142104 100644
--- a/fs/fcntl.c
+++ b/fs/fcntl.c
@@ -291,6 +291,14 @@ static bool rw_life_hint_valid(u64 hint)
 	}
 }
 
+static inline bool rw_placement_hint_valid(u64 val)
+{
+	if (val <= MAX_PLACEMENT_HINT_VAL)
+		return true;
+
+	return false;
+}
+
 static long fcntl_get_rw_life_hint(struct file *file, unsigned int cmd,
 			      unsigned long arg)
 {
@@ -327,6 +335,59 @@ static long fcntl_set_rw_life_hint(struct file *file, unsigned int cmd,
 	return 0;
 }
 
+static long fcntl_get_rw_hint_ex(struct file *file, unsigned int cmd,
+			      unsigned long arg)
+{
+	struct rw_hint_ex __user *rw_hint_ex_p = (void __user *)arg;
+	struct rw_hint_ex rwh = {};
+	struct inode *inode = file_inode(file);
+	u8 hint = READ_ONCE(inode->i_write_hint);
+
+	rwh.type = WRITE_HINT_TYPE(hint);
+	rwh.val = WRITE_HINT_VAL(hint);
+
+	if (copy_to_user(rw_hint_ex_p, &rwh, sizeof(rwh)))
+		return -EFAULT;
+
+	return 0;
+}
+
+static long fcntl_set_rw_hint_ex(struct file *file, unsigned int cmd,
+			      unsigned long arg)
+{
+	struct rw_hint_ex __user *rw_hint_ex_p = (void __user *)arg;
+	struct rw_hint_ex rwh;
+	struct inode *inode = file_inode(file);
+	u64 hint;
+	int i;
+
+	if (copy_from_user(&rwh, rw_hint_ex_p, sizeof(rwh)))
+		return -EFAULT;
+	for (i = 0; i < ARRAY_SIZE(rwh.pad); i++)
+		if (rwh.pad[i])
+			return -EINVAL;
+	switch (rwh.type) {
+	case TYPE_RW_LIFE_HINT:
+		if (!rw_life_hint_valid(rwh.val))
+			return -EINVAL;
+		hint = rwh.val;
+		break;
+	case TYPE_RW_PLACEMENT_HINT:
+		if (!rw_placement_hint_valid(rwh.val))
+			return -EINVAL;
+		hint = PLACEMENT_HINT_TYPE | rwh.val;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	WRITE_ONCE(inode->i_write_hint, hint);
+	if (file->f_mapping->host != inode)
+		WRITE_ONCE(file->f_mapping->host->i_write_hint, hint);
+
+	return 0;
+}
+
 /* Is the file descriptor a dup of the file? */
 static long f_dupfd_query(int fd, struct file *filp)
 {
@@ -454,6 +515,12 @@ static long do_fcntl(int fd, unsigned int cmd, unsigned long arg,
 	case F_SET_RW_HINT:
 		err = fcntl_set_rw_life_hint(filp, cmd, arg);
 		break;
+	case F_GET_RW_HINT_EX:
+		err = fcntl_get_rw_hint_ex(filp, cmd, arg);
+		break;
+	case F_SET_RW_HINT_EX:
+		err = fcntl_set_rw_hint_ex(filp, cmd, arg);
+		break;
 	default:
 		break;
 	}
diff --git a/include/linux/rw_hint.h b/include/linux/rw_hint.h
index e17fd9fa65d4..611c51d23d96 100644
--- a/include/linux/rw_hint.h
+++ b/include/linux/rw_hint.h
@@ -16,4 +16,17 @@ enum rw_life_hint {
 	WRITE_LIFE_EXTREME	= RWH_WRITE_LIFE_EXTREME,
 };
 
+#define WRITE_HINT_TYPE_BIT	BIT(7)
+#define WRITE_HINT_VAL_MASK	(WRITE_HINT_TYPE_BIT - 1)
+#define WRITE_HINT_TYPE(h)	(((h) & WRITE_HINT_TYPE_BIT) ? \
+				TYPE_RW_PLACEMENT_HINT : TYPE_RW_LIFE_HINT)
+#define WRITE_HINT_VAL(h)	((h) & WRITE_HINT_VAL_MASK)
+
+#define WRITE_PLACEMENT_HINT(h)	(((h) & WRITE_HINT_TYPE_BIT) ? \
+				 WRITE_HINT_VAL(h) : 0)
+#define WRITE_LIFE_HINT(h)	(((h) & WRITE_HINT_TYPE_BIT) ? \
+				 0 : WRITE_HINT_VAL(h))
+
+#define PLACEMENT_HINT_TYPE	WRITE_HINT_TYPE_BIT
+#define MAX_PLACEMENT_HINT_VAL	WRITE_HINT_VAL_MASK
 #endif /* _LINUX_RW_HINT_H */
diff --git a/include/uapi/linux/fcntl.h b/include/uapi/linux/fcntl.h
index c0bcc185fa48..dfe77fa86776 100644
--- a/include/uapi/linux/fcntl.h
+++ b/include/uapi/linux/fcntl.h
@@ -57,6 +57,8 @@
 #define F_SET_RW_HINT		(F_LINUX_SPECIFIC_BASE + 12)
 #define F_GET_FILE_RW_HINT	(F_LINUX_SPECIFIC_BASE + 13)
 #define F_SET_FILE_RW_HINT	(F_LINUX_SPECIFIC_BASE + 14)
+#define F_GET_RW_HINT_EX	(F_LINUX_SPECIFIC_BASE + 15)
+#define F_SET_RW_HINT_EX	(F_LINUX_SPECIFIC_BASE + 16)
 
 /*
  * Valid hint values for F_{GET,SET}_RW_HINT. 0 is "not set", or can be
@@ -76,6 +78,18 @@
  */
 #define RWF_WRITE_LIFE_NOT_SET	RWH_WRITE_LIFE_NOT_SET
 
+enum rw_hint_type {
+	TYPE_RW_LIFE_HINT,
+	TYPE_RW_PLACEMENT_HINT
+};
+
+/* Exchange information with F_{GET/SET}_RW_HINT fcntl */
+struct rw_hint_ex {
+	__u8	type;
+	__u8	pad[7];
+	__u64	val;
+};
+
 /*
  * Types of directory notifications that may be requested.
  */
-- 
2.25.1