Merge tag 'for-hubcap-v4.9-readahead' of git://github.com/martinbrandenburg/linux

orangefs: integrate readahead cache changes from out-of-tree

The readahead cache has long been present as a compile time option to
OrangeFS. As it had not been tested in some time, it was disabled by
default. Recently, Walt Ligon started work reviving it, which eventually
culminated in the commit below to OrangeFS SVN.

r12519 | walt | 2016-07-13 14:32:42 -0400 (Wed, 13 Jul 2016) | 1 line
reintegrating readahead cache code

This cache is implemented almost entirely in userspace. There are a few
parameters exposed via sysfs, and the cache must be flushed after an
inode is released.

8 files changed, 160 insertions, 40 deletions

diff --git a/include/linux/ceph/ceph_fs.h b/include/linux/ceph/ceph_fs.h
index dfce616002ad..7868d602c0a0 100644
--- a/include/linux/ceph/ceph_fs.h
+++ b/include/linux/ceph/ceph_fs.h
@@ -34,9 +34,9 @@
 #define CEPH_MAX_MON   31
 
 /*
- * ceph_file_layout - describe data layout for a file/inode
+ * legacy ceph_file_layoute
  */
-struct ceph_file_layout {
+struct ceph_file_layout_legacy {
 	/* file -> object mapping */
 	__le32 fl_stripe_unit;     /* stripe unit, in bytes.  must be multiple
 				      of page size. */
@@ -53,33 +53,27 @@ struct ceph_file_layout {
 	__le32 fl_pg_pool;      /* namespace, crush ruleset, rep level */
 } __attribute__ ((packed));
 
-#define ceph_file_layout_su(l) ((__s32)le32_to_cpu((l).fl_stripe_unit))
-#define ceph_file_layout_stripe_count(l) \
-	((__s32)le32_to_cpu((l).fl_stripe_count))
-#define ceph_file_layout_object_size(l) ((__s32)le32_to_cpu((l).fl_object_size))
-#define ceph_file_layout_cas_hash(l) ((__s32)le32_to_cpu((l).fl_cas_hash))
-#define ceph_file_layout_object_su(l) \
-	((__s32)le32_to_cpu((l).fl_object_stripe_unit))
-#define ceph_file_layout_pg_pool(l) \
-	((__s32)le32_to_cpu((l).fl_pg_pool))
-
-static inline unsigned ceph_file_layout_stripe_width(struct ceph_file_layout *l)
-{
-	return le32_to_cpu(l->fl_stripe_unit) *
-		le32_to_cpu(l->fl_stripe_count);
-}
-
-/* "period" == bytes before i start on a new set of objects */
-static inline unsigned ceph_file_layout_period(struct ceph_file_layout *l)
-{
-	return le32_to_cpu(l->fl_object_size) *
-		le32_to_cpu(l->fl_stripe_count);
-}
+struct ceph_string;
+/*
+ * ceph_file_layout - describe data layout for a file/inode
+ */
+struct ceph_file_layout {
+	/* file -> object mapping */
+	u32 stripe_unit;   /* stripe unit, in bytes */
+	u32 stripe_count;  /* over this many objects */
+	u32 object_size;   /* until objects are this big */
+	s64 pool_id;        /* rados pool id */
+	struct ceph_string __rcu *pool_ns; /* rados pool namespace */
+};
+
+extern int ceph_file_layout_is_valid(const struct ceph_file_layout *layout);
+extern void ceph_file_layout_from_legacy(struct ceph_file_layout *fl,
+				struct ceph_file_layout_legacy *legacy);
+extern void ceph_file_layout_to_legacy(struct ceph_file_layout *fl,
+				struct ceph_file_layout_legacy *legacy);
 
 #define CEPH_MIN_STRIPE_UNIT 65536
 
-int ceph_file_layout_is_valid(const struct ceph_file_layout *layout);
-
 struct ceph_dir_layout {
 	__u8   dl_dir_hash;   /* see ceph_hash.h for ids */
 	__u8   dl_unused1;
@@ -127,6 +121,7 @@ struct ceph_dir_layout {
 
 /* client <-> mds */
 #define CEPH_MSG_MDS_MAP                21
+#define CEPH_MSG_FS_MAP_USER            103
 
 #define CEPH_MSG_CLIENT_SESSION         22
 #define CEPH_MSG_CLIENT_RECONNECT       23
@@ -399,7 +394,7 @@ union ceph_mds_request_args {
 		__le32 flags;
 	} __attribute__ ((packed)) setxattr;
 	struct {
-		struct ceph_file_layout layout;
+		struct ceph_file_layout_legacy layout;
 	} __attribute__ ((packed)) setlayout;
 	struct {
 		__u8 rule; /* currently fcntl or flock */
@@ -478,7 +473,7 @@ struct ceph_mds_reply_inode {
 	__le64 version;                /* inode version */
 	__le64 xattr_version;          /* version for xattr blob */
 	struct ceph_mds_reply_cap cap; /* caps issued for this inode */
-	struct ceph_file_layout layout;
+	struct ceph_file_layout_legacy layout;
 	struct ceph_timespec ctime, mtime, atime;
 	__le32 time_warp_seq;
 	__le64 size, max_size, truncate_size;
@@ -531,7 +526,7 @@ struct ceph_filelock {
 #define CEPH_FILE_MODE_WR         2
 #define CEPH_FILE_MODE_RDWR       3  /* RD | WR */
 #define CEPH_FILE_MODE_LAZY       4  /* lazy io */
-#define CEPH_FILE_MODE_NUM        8  /* bc these are bit fields.. mostly */
+#define CEPH_FILE_MODE_BITS       4
 
 int ceph_flags_to_mode(int flags);
 
@@ -673,7 +668,7 @@ struct ceph_mds_caps {
 	__le64 size, max_size, truncate_size;
 	__le32 truncate_seq;
 	struct ceph_timespec mtime, atime, ctime;
-	struct ceph_file_layout layout;
+	struct ceph_file_layout_legacy layout;
 	__le32 time_warp_seq;
 } __attribute__ ((packed));
 
diff --git a/include/linux/ceph/decode.h b/include/linux/ceph/decode.h
index 19e9932f3e77..f990f2cc907a 100644
--- a/include/linux/ceph/decode.h
+++ b/include/linux/ceph/decode.h
@@ -3,6 +3,7 @@
 
 #include <linux/err.h>
 #include <linux/bug.h>
+#include <linux/slab.h>
 #include <linux/time.h>
 #include <asm/unaligned.h>
 
@@ -217,6 +218,60 @@ static inline void ceph_encode_string(void **p, void *end,
 	*p += len;
 }
 
+/*
+ * version and length starting block encoders/decoders
+ */
+
+/* current code version (u8) + compat code version (u8) + len of struct (u32) */
+#define CEPH_ENCODING_START_BLK_LEN 6
+
+/**
+ * ceph_start_encoding - start encoding block
+ * @struct_v: current (code) version of the encoding
+ * @struct_compat: oldest code version that can decode it
+ * @struct_len: length of struct encoding
+ */
+static inline void ceph_start_encoding(void **p, u8 struct_v, u8 struct_compat,
+				       u32 struct_len)
+{
+	ceph_encode_8(p, struct_v);
+	ceph_encode_8(p, struct_compat);
+	ceph_encode_32(p, struct_len);
+}
+
+/**
+ * ceph_start_decoding - start decoding block
+ * @v: current version of the encoding that the code supports
+ * @name: name of the struct (free-form)
+ * @struct_v: out param for the encoding version
+ * @struct_len: out param for the length of struct encoding
+ *
+ * Validates the length of struct encoding, so unsafe ceph_decode_*
+ * variants can be used for decoding.
+ */
+static inline int ceph_start_decoding(void **p, void *end, u8 v,
+				      const char *name, u8 *struct_v,
+				      u32 *struct_len)
+{
+	u8 struct_compat;
+
+	ceph_decode_need(p, end, CEPH_ENCODING_START_BLK_LEN, bad);
+	*struct_v = ceph_decode_8(p);
+	struct_compat = ceph_decode_8(p);
+	if (v < struct_compat) {
+		pr_warn("got struct_v %d struct_compat %d > %d of %s\n",
+			*struct_v, struct_compat, v, name);
+		return -EINVAL;
+	}
+
+	*struct_len = ceph_decode_32(p);
+	ceph_decode_need(p, end, *struct_len, bad);
+	return 0;
+
+bad:
+	return -ERANGE;
+}
+
 #define ceph_encode_need(p, end, n, bad)			\
 	do {							\
 		if (!likely(ceph_has_room(p, end, n)))		\
diff --git a/include/linux/ceph/libceph.h b/include/linux/ceph/libceph.h
index 690985daad1c..83fc1fff7061 100644
--- a/include/linux/ceph/libceph.h
+++ b/include/linux/ceph/libceph.h
@@ -21,6 +21,7 @@
 #include <linux/ceph/mon_client.h>
 #include <linux/ceph/osd_client.h>
 #include <linux/ceph/ceph_fs.h>
+#include <linux/ceph/string_table.h>
 
 /*
  * mount options
@@ -214,8 +215,9 @@ static void erase_##name(struct rb_root *root, type *t)			\
 }
 
 #define DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)		\
+extern type __lookup_##name##_key;					\
 static type *lookup_##name(struct rb_root *root,			\
-			   typeof(((type *)0)->keyfld) key)		\
+			   typeof(__lookup_##name##_key.keyfld) key)	\
 {									\
 	struct rb_node *n = root->rb_node;				\
 									\
diff --git a/include/linux/ceph/mon_client.h b/include/linux/ceph/mon_client.h
index e2a92df08b47..24d704d1ea5c 100644
--- a/include/linux/ceph/mon_client.h
+++ b/include/linux/ceph/mon_client.h
@@ -95,7 +95,7 @@ struct ceph_mon_client {
 		struct ceph_mon_subscribe_item item;
 		bool want;
 		u32 have; /* epoch */
-	} subs[3];
+	} subs[4];
 	int fs_cluster_id; /* "mdsmap.<id>" sub */
 
 #ifdef CONFIG_DEBUG_FS
@@ -111,9 +111,10 @@ extern int ceph_monc_init(struct ceph_mon_client *monc, struct ceph_client *cl);
 extern void ceph_monc_stop(struct ceph_mon_client *monc);
 
 enum {
-	CEPH_SUB_MDSMAP = 0,
-	CEPH_SUB_MONMAP,
+	CEPH_SUB_MONMAP = 0,
 	CEPH_SUB_OSDMAP,
+	CEPH_SUB_FSMAP,
+	CEPH_SUB_MDSMAP,
 };
 
 extern const char *ceph_sub_str[];
diff --git a/include/linux/ceph/msgpool.h b/include/linux/ceph/msgpool.h
index 4b0d38960726..ddd0d48d0384 100644
--- a/include/linux/ceph/msgpool.h
+++ b/include/linux/ceph/msgpool.h
@@ -2,7 +2,6 @@
 #define _FS_CEPH_MSGPOOL
 
 #include <linux/mempool.h>
-#include <linux/ceph/messenger.h>
 
 /*
  * we use memory pools for preallocating messages we may receive, to
diff --git a/include/linux/ceph/osd_client.h b/include/linux/ceph/osd_client.h
index 1b3b6e155392..858932304260 100644
--- a/include/linux/ceph/osd_client.h
+++ b/include/linux/ceph/osd_client.h
@@ -9,6 +9,7 @@
 #include <linux/ceph/types.h>
 #include <linux/ceph/osdmap.h>
 #include <linux/ceph/messenger.h>
+#include <linux/ceph/msgpool.h>
 #include <linux/ceph/auth.h>
 #include <linux/ceph/pagelist.h>
 
diff --git a/include/linux/ceph/osdmap.h b/include/linux/ceph/osdmap.h
index 9ccf4dbe55f8..9a9041784dcf 100644
--- a/include/linux/ceph/osdmap.h
+++ b/include/linux/ceph/osdmap.h
@@ -63,11 +63,13 @@ static inline bool ceph_can_shift_osds(struct ceph_pg_pool_info *pool)
 
 struct ceph_object_locator {
 	s64 pool;
+	struct ceph_string *pool_ns;
 };
 
 static inline void ceph_oloc_init(struct ceph_object_locator *oloc)
 {
 	oloc->pool = -1;
+	oloc->pool_ns = NULL;
 }
 
 static inline bool ceph_oloc_empty(const struct ceph_object_locator *oloc)
@@ -75,11 +77,9 @@ static inline bool ceph_oloc_empty(const struct ceph_object_locator *oloc)
 	return oloc->pool == -1;
 }
 
-static inline void ceph_oloc_copy(struct ceph_object_locator *dest,
-				  const struct ceph_object_locator *src)
-{
-	dest->pool = src->pool;
-}
+void ceph_oloc_copy(struct ceph_object_locator *dest,
+		    const struct ceph_object_locator *src);
+void ceph_oloc_destroy(struct ceph_object_locator *oloc);
 
 /*
  * Maximum supported by kernel client object name length
@@ -115,6 +115,11 @@ static inline void ceph_oid_init(struct ceph_object_id *oid)
 	oid->name_len = 0;
 }
 
+#define CEPH_OID_INIT_ONSTACK(oid)					\
+    ({ ceph_oid_init(&oid); oid; })
+#define CEPH_DEFINE_OID_ONSTACK(oid)					\
+	struct ceph_object_id oid = CEPH_OID_INIT_ONSTACK(oid)
+
 static inline bool ceph_oid_empty(const struct ceph_object_id *oid)
 {
 	return oid->name == oid->inline_name && !oid->name_len;
diff --git a/include/linux/ceph/string_table.h b/include/linux/ceph/string_table.h
new file mode 100644
index 000000000000..1b02c96daf75
--- /dev/null
+++ b/include/linux/ceph/string_table.h
@@ -0,0 +1,62 @@
+#ifndef _FS_CEPH_STRING_TABLE_H
+#define _FS_CEPH_STRING_TABLE_H
+
+#include <linux/types.h>
+#include <linux/kref.h>
+#include <linux/rbtree.h>
+#include <linux/rcupdate.h>
+
+struct ceph_string {
+	struct kref kref;
+	union {
+		struct rb_node node;
+		struct rcu_head rcu;
+	};
+	size_t len;
+	char str[];
+};
+
+extern void ceph_release_string(struct kref *ref);
+extern struct ceph_string *ceph_find_or_create_string(const char *str,
+						      size_t len);
+extern bool ceph_strings_empty(void);
+
+static inline struct ceph_string *ceph_get_string(struct ceph_string *str)
+{
+	kref_get(&str->kref);
+	return str;
+}
+
+static inline void ceph_put_string(struct ceph_string *str)
+{
+	if (!str)
+		return;
+	kref_put(&str->kref, ceph_release_string);
+}
+
+static inline int ceph_compare_string(struct ceph_string *cs,
+				      const char* str, size_t len)
+{
+	size_t cs_len = cs ? cs->len : 0;
+	if (cs_len != len)
+		return cs_len - len;
+	if (len == 0)
+		return 0;
+	return strncmp(cs->str, str, len);
+}
+
+#define ceph_try_get_string(x)					\
+({								\
+	struct ceph_string *___str;				\
+	rcu_read_lock();					\
+	for (;;) {						\
+		___str = rcu_dereference(x);			\
+		if (!___str ||					\
+		    kref_get_unless_zero(&___str->kref))	\
+			break;					\
+	}							\
+	rcu_read_unlock();					\
+	(___str);						\
+})
+
+#endif