#include <linux/wait.h> /* wait_queue_head_t, etc */
#include <linux/spinlock.h> /* spinlock_t, etc */
-#include <linux/atomic.h> /* atomic_t, etc */
+#include <linux/atomic.h> /* atomic_t, etc */
+#include <linux/kref.h> /* struct kref */
#include <linux/workqueue.h> /* struct work_struct */
#include <rdma/rdma_cm.h> /* RDMA connection api */
* Interface Adapter -- one per transport instance
*/
struct rpcrdma_ia {
- struct ib_device *ri_device;
struct rdma_cm_id *ri_id;
struct ib_pd *ri_pd;
- struct completion ri_done;
- struct completion ri_remove_done;
int ri_async_rc;
unsigned int ri_max_segs;
unsigned int ri_max_frwr_depth;
- unsigned int ri_max_inline_write;
- unsigned int ri_max_inline_read;
unsigned int ri_max_send_sges;
bool ri_implicit_roundup;
enum ib_mr_type ri_mrtype;
unsigned long ri_flags;
+ struct completion ri_done;
+ struct completion ri_remove_done;
};
enum {
struct rpcrdma_ep {
unsigned int rep_send_count;
unsigned int rep_send_batch;
+ unsigned int rep_max_inline_send;
+ unsigned int rep_max_inline_recv;
int rep_connected;
struct ib_qp_init_attr rep_attr;
wait_queue_head_t rep_connect_wait;
struct rpcrdma_connect_private rep_cm_private;
struct rdma_conn_param rep_remote_cma;
+ unsigned int rep_max_requests; /* set by /proc */
+ unsigned int rep_inline_send; /* negotiated */
+ unsigned int rep_inline_recv; /* negotiated */
int rep_receive_count;
};
/* Pre-allocate extra Work Requests for handling backward receives
* and sends. This is a fixed value because the Work Queues are
- * allocated when the forward channel is set up.
+ * allocated when the forward channel is set up, long before the
+ * backchannel is provisioned. This value is two times
+ * NFS4_DEF_CB_SLOT_TABLE_SIZE.
*/
#if defined(CONFIG_SUNRPC_BACKCHANNEL)
-#define RPCRDMA_BACKWARD_WRS (8)
+#define RPCRDMA_BACKWARD_WRS (32)
#else
-#define RPCRDMA_BACKWARD_WRS (0)
+#define RPCRDMA_BACKWARD_WRS (0)
#endif
/* Registered buffer -- registered kmalloc'd memory for RDMA SEND/RECV
- *
- * The below structure appears at the front of a large region of kmalloc'd
- * memory, which always starts on a good alignment boundary.
*/
struct rpcrdma_regbuf {
struct ib_sge rg_iov;
struct ib_device *rg_device;
enum dma_data_direction rg_direction;
- __be32 rg_base[0] __attribute__ ((aligned(256)));
+ void *rg_data;
};
-static inline u64
-rdmab_addr(struct rpcrdma_regbuf *rb)
+static inline u64 rdmab_addr(struct rpcrdma_regbuf *rb)
{
return rb->rg_iov.addr;
}
-static inline u32
-rdmab_length(struct rpcrdma_regbuf *rb)
+static inline u32 rdmab_length(struct rpcrdma_regbuf *rb)
{
return rb->rg_iov.length;
}
-static inline u32
-rdmab_lkey(struct rpcrdma_regbuf *rb)
+static inline u32 rdmab_lkey(struct rpcrdma_regbuf *rb)
{
return rb->rg_iov.lkey;
}
-static inline struct ib_device *
-rdmab_device(struct rpcrdma_regbuf *rb)
+static inline struct ib_device *rdmab_device(struct rpcrdma_regbuf *rb)
{
return rb->rg_device;
}
+static inline void *rdmab_data(const struct rpcrdma_regbuf *rb)
+{
+ return rb->rg_data;
+}
+
#define RPCRDMA_DEF_GFP (GFP_NOIO | __GFP_NOWARN)
/* To ensure a transport can always make forward progress,
bool rr_temp;
struct rpcrdma_regbuf *rr_rdmabuf;
struct rpcrdma_xprt *rr_rxprt;
- struct work_struct rr_work;
+ struct rpc_rqst *rr_rqst;
struct xdr_buf rr_hdrbuf;
struct xdr_stream rr_stream;
- struct rpc_rqst *rr_rqst;
struct list_head rr_list;
struct ib_recv_wr rr_recv_wr;
};
struct rpcrdma_sendctx {
struct ib_send_wr sc_wr;
struct ib_cqe sc_cqe;
+ struct ib_device *sc_device;
struct rpcrdma_xprt *sc_xprt;
struct rpcrdma_req *sc_req;
unsigned int sc_unmap_count;
struct ib_sge sc_sges[];
};
-/* Limit the number of SGEs that can be unmapped during one
- * Send completion. This caps the amount of work a single
- * completion can do before returning to the provider.
- *
- * Setting this to zero disables Send completion batching.
- */
-enum {
- RPCRDMA_MAX_SEND_BATCH = 7,
-};
-
/*
* struct rpcrdma_mr - external memory region metadata
*
* An external memory region is any buffer or page that is registered
* on the fly (ie, not pre-registered).
- *
- * Each rpcrdma_buffer has a list of free MWs anchored in rb_mrs. During
- * call_allocate, rpcrdma_buffer_get() assigns one to each segment in
- * an rpcrdma_req. Then rpcrdma_register_external() grabs these to keep
- * track of registration metadata while each RPC is pending.
- * rpcrdma_deregister_external() uses this metadata to unmap and
- * release these resources when an RPC is complete.
*/
-enum rpcrdma_frwr_state {
- FRWR_IS_INVALID, /* ready to be used */
- FRWR_IS_VALID, /* in use */
- FRWR_FLUSHED_FR, /* flushed FASTREG WR */
- FRWR_FLUSHED_LI, /* flushed LOCALINV WR */
-};
-
+struct rpcrdma_req;
struct rpcrdma_frwr {
struct ib_mr *fr_mr;
struct ib_cqe fr_cqe;
- enum rpcrdma_frwr_state fr_state;
struct completion fr_linv_done;
+ struct rpcrdma_req *fr_req;
union {
struct ib_reg_wr fr_regwr;
struct ib_send_wr fr_invwr;
struct rpcrdma_req {
struct list_head rl_list;
struct rpc_rqst rl_slot;
- struct rpcrdma_buffer *rl_buffer;
struct rpcrdma_rep *rl_reply;
struct xdr_stream rl_stream;
struct xdr_buf rl_hdrbuf;
struct rpcrdma_regbuf *rl_recvbuf; /* rq_rcv_buf */
struct list_head rl_all;
- unsigned long rl_flags;
+ struct kref rl_kref;
struct list_head rl_registered; /* registered segments */
struct rpcrdma_mr_seg rl_segments[RPCRDMA_MAX_SEGS];
};
-/* rl_flags */
-enum {
- RPCRDMA_REQ_F_PENDING = 0,
- RPCRDMA_REQ_F_TX_RESOURCES,
-};
-
static inline struct rpcrdma_req *
rpcr_to_rdmar(const struct rpc_rqst *rqst)
{
struct list_head rb_recv_bufs;
struct list_head rb_allreqs;
- unsigned long rb_flags;
u32 rb_max_requests;
u32 rb_credits; /* most recent credit grant */
u32 rb_bc_srv_max_requests;
u32 rb_bc_max_requests;
- struct workqueue_struct *rb_completion_wq;
struct delayed_work rb_refresh_worker;
};
-/* rb_flags */
-enum {
- RPCRDMA_BUF_F_EMPTY_SCQ = 0,
-};
-
-/*
- * Internal structure for transport instance creation. This
- * exists primarily for modularity.
- *
- * This data should be set with mount options
- */
-struct rpcrdma_create_data_internal {
- unsigned int max_requests; /* max requests (slots) in flight */
- unsigned int rsize; /* mount rsize - max read hdr+data */
- unsigned int wsize; /* mount wsize - max write hdr+data */
- unsigned int inline_rsize; /* max non-rdma read data payload */
- unsigned int inline_wsize; /* max non-rdma write data payload */
-};
-
/*
* Statistics for RPCRDMA
*/
struct rpcrdma_ia rx_ia;
struct rpcrdma_ep rx_ep;
struct rpcrdma_buffer rx_buf;
- struct rpcrdma_create_data_internal rx_data;
struct delayed_work rx_connect_worker;
+ struct rpc_timeout rx_timeout;
struct rpcrdma_stats rx_stats;
};
#define rpcx_to_rdmax(x) container_of(x, struct rpcrdma_xprt, rx_xprt)
-#define rpcx_to_rdmad(x) (rpcx_to_rdmax(x)->rx_data)
static inline const char *
rpcrdma_addrstr(const struct rpcrdma_xprt *r_xprt)
/*
* Endpoint calls - xprtrdma/verbs.c
*/
-int rpcrdma_ep_create(struct rpcrdma_ep *, struct rpcrdma_ia *,
- struct rpcrdma_create_data_internal *);
-void rpcrdma_ep_destroy(struct rpcrdma_ep *, struct rpcrdma_ia *);
+int rpcrdma_ep_create(struct rpcrdma_xprt *r_xprt);
+void rpcrdma_ep_destroy(struct rpcrdma_xprt *r_xprt);
int rpcrdma_ep_connect(struct rpcrdma_ep *, struct rpcrdma_ia *);
void rpcrdma_ep_disconnect(struct rpcrdma_ep *, struct rpcrdma_ia *);
/*
* Buffer calls - xprtrdma/verbs.c
*/
-struct rpcrdma_req *rpcrdma_create_req(struct rpcrdma_xprt *);
+struct rpcrdma_req *rpcrdma_req_create(struct rpcrdma_xprt *r_xprt, size_t size,
+ gfp_t flags);
void rpcrdma_req_destroy(struct rpcrdma_req *req);
int rpcrdma_buffer_create(struct rpcrdma_xprt *);
void rpcrdma_buffer_destroy(struct rpcrdma_buffer *);
-struct rpcrdma_sendctx *rpcrdma_sendctx_get_locked(struct rpcrdma_buffer *buf);
+struct rpcrdma_sendctx *rpcrdma_sendctx_get_locked(struct rpcrdma_xprt *r_xprt);
struct rpcrdma_mr *rpcrdma_mr_get(struct rpcrdma_xprt *r_xprt);
void rpcrdma_mr_put(struct rpcrdma_mr *mr);
}
struct rpcrdma_req *rpcrdma_buffer_get(struct rpcrdma_buffer *);
-void rpcrdma_buffer_put(struct rpcrdma_req *);
+void rpcrdma_buffer_put(struct rpcrdma_buffer *buffers,
+ struct rpcrdma_req *req);
void rpcrdma_recv_buffer_put(struct rpcrdma_rep *);
-struct rpcrdma_regbuf *rpcrdma_alloc_regbuf(size_t, enum dma_data_direction,
- gfp_t);
-bool __rpcrdma_dma_map_regbuf(struct rpcrdma_ia *, struct rpcrdma_regbuf *);
-void rpcrdma_free_regbuf(struct rpcrdma_regbuf *);
+bool rpcrdma_regbuf_realloc(struct rpcrdma_regbuf *rb, size_t size,
+ gfp_t flags);
+bool __rpcrdma_regbuf_dma_map(struct rpcrdma_xprt *r_xprt,
+ struct rpcrdma_regbuf *rb);
-static inline bool
-rpcrdma_regbuf_is_mapped(struct rpcrdma_regbuf *rb)
+/**
+ * rpcrdma_regbuf_is_mapped - check if buffer is DMA mapped
+ *
+ * Returns true if the buffer is now mapped to rb->rg_device.
+ */
+static inline bool rpcrdma_regbuf_is_mapped(struct rpcrdma_regbuf *rb)
{
return rb->rg_device != NULL;
}
-static inline bool
-rpcrdma_dma_map_regbuf(struct rpcrdma_ia *ia, struct rpcrdma_regbuf *rb)
+/**
+ * rpcrdma_regbuf_dma_map - DMA-map a regbuf
+ * @r_xprt: controlling transport instance
+ * @rb: regbuf to be mapped
+ *
+ * Returns true if the buffer is currently DMA mapped.
+ */
+static inline bool rpcrdma_regbuf_dma_map(struct rpcrdma_xprt *r_xprt,
+ struct rpcrdma_regbuf *rb)
{
if (likely(rpcrdma_regbuf_is_mapped(rb)))
return true;
- return __rpcrdma_dma_map_regbuf(ia, rb);
+ return __rpcrdma_regbuf_dma_map(r_xprt, rb);
}
/*
/* Memory registration calls xprtrdma/frwr_ops.c
*/
-bool frwr_is_supported(struct rpcrdma_ia *);
-int frwr_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep,
- struct rpcrdma_create_data_internal *cdata);
+bool frwr_is_supported(struct ib_device *device);
+void frwr_reset(struct rpcrdma_req *req);
+int frwr_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep);
int frwr_init_mr(struct rpcrdma_ia *ia, struct rpcrdma_mr *mr);
void frwr_release_mr(struct rpcrdma_mr *mr);
size_t frwr_maxpages(struct rpcrdma_xprt *r_xprt);
struct rpcrdma_mr **mr);
int frwr_send(struct rpcrdma_ia *ia, struct rpcrdma_req *req);
void frwr_reminv(struct rpcrdma_rep *rep, struct list_head *mrs);
-void frwr_unmap_sync(struct rpcrdma_xprt *r_xprt,
- struct list_head *mrs);
+void frwr_unmap_sync(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req);
+void frwr_unmap_async(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req);
/*
* RPC/RDMA protocol calls - xprtrdma/rpc_rdma.c
struct rpcrdma_req *req, u32 hdrlen,
struct xdr_buf *xdr,
enum rpcrdma_chunktype rtype);
-void rpcrdma_unmap_sendctx(struct rpcrdma_sendctx *sc);
+void rpcrdma_sendctx_unmap(struct rpcrdma_sendctx *sc);
int rpcrdma_marshal_req(struct rpcrdma_xprt *r_xprt, struct rpc_rqst *rqst);
void rpcrdma_set_max_header_sizes(struct rpcrdma_xprt *);
void rpcrdma_complete_rqst(struct rpcrdma_rep *rep);
void rpcrdma_reply_handler(struct rpcrdma_rep *rep);
-void rpcrdma_release_rqst(struct rpcrdma_xprt *r_xprt,
- struct rpcrdma_req *req);
-void rpcrdma_deferred_completion(struct work_struct *work);
static inline void rpcrdma_set_xdrlen(struct xdr_buf *xdr, size_t len)
{
/* RPC/RDMA module init - xprtrdma/transport.c
*/
+extern unsigned int xprt_rdma_slot_table_entries;
extern unsigned int xprt_rdma_max_inline_read;
+extern unsigned int xprt_rdma_max_inline_write;
void xprt_rdma_format_addresses(struct rpc_xprt *xprt, struct sockaddr *sap);
void xprt_rdma_free_addresses(struct rpc_xprt *xprt);
void xprt_rdma_close(struct rpc_xprt *xprt);
#if defined(CONFIG_SUNRPC_BACKCHANNEL)
int xprt_rdma_bc_setup(struct rpc_xprt *, unsigned int);
size_t xprt_rdma_bc_maxpayload(struct rpc_xprt *);
+unsigned int xprt_rdma_bc_max_slots(struct rpc_xprt *);
int rpcrdma_bc_post_recv(struct rpcrdma_xprt *, unsigned int);
void rpcrdma_bc_receive_call(struct rpcrdma_xprt *, struct rpcrdma_rep *);
int xprt_rdma_bc_send_reply(struct rpc_rqst *rqst);