2 * linux/net/sunrpc/xprtsock.c
4 * Client-side transport implementation for sockets.
6 * TCP callback races fixes (C) 1998 Red Hat
7 * TCP send fixes (C) 1998 Red Hat
8 * TCP NFS related read + write fixes
9 * (C) 1999 Dave Airlie, University of Limerick, Ireland <airlied@linux.ie>
11 * Rewrite of larges part of the code in order to stabilize TCP stuff.
12 * Fix behaviour when socket buffer is full.
13 * (C) 1999 Trond Myklebust <trond.myklebust@fys.uio.no>
15 * IP socket transport implementation, (C) 2005 Chuck Lever <cel@netapp.com>
17 * IPv6 support contributed by Gilles Quillard, Bull Open Source, 2005.
18 * <gilles.quillard@bull.net>
21 #include <linux/types.h>
22 #include <linux/string.h>
23 #include <linux/slab.h>
24 #include <linux/module.h>
25 #include <linux/capability.h>
26 #include <linux/pagemap.h>
27 #include <linux/errno.h>
28 #include <linux/socket.h>
30 #include <linux/net.h>
33 #include <linux/udp.h>
34 #include <linux/tcp.h>
35 #include <linux/sunrpc/clnt.h>
36 #include <linux/sunrpc/addr.h>
37 #include <linux/sunrpc/sched.h>
38 #include <linux/sunrpc/svcsock.h>
39 #include <linux/sunrpc/xprtsock.h>
40 #include <linux/file.h>
41 #ifdef CONFIG_SUNRPC_BACKCHANNEL
42 #include <linux/sunrpc/bc_xprt.h>
46 #include <net/checksum.h>
50 #include <trace/events/sunrpc.h>
54 static void xs_close(struct rpc_xprt
*xprt
);
59 static unsigned int xprt_udp_slot_table_entries
= RPC_DEF_SLOT_TABLE
;
60 static unsigned int xprt_tcp_slot_table_entries
= RPC_MIN_SLOT_TABLE
;
61 static unsigned int xprt_max_tcp_slot_table_entries
= RPC_MAX_SLOT_TABLE
;
63 static unsigned int xprt_min_resvport
= RPC_DEF_MIN_RESVPORT
;
64 static unsigned int xprt_max_resvport
= RPC_DEF_MAX_RESVPORT
;
66 #define XS_TCP_LINGER_TO (15U * HZ)
67 static unsigned int xs_tcp_fin_timeout __read_mostly
= XS_TCP_LINGER_TO
;
70 * We can register our own files under /proc/sys/sunrpc by
71 * calling register_sysctl_table() again. The files in that
72 * directory become the union of all files registered there.
74 * We simply need to make sure that we don't collide with
75 * someone else's file names!
80 static unsigned int min_slot_table_size
= RPC_MIN_SLOT_TABLE
;
81 static unsigned int max_slot_table_size
= RPC_MAX_SLOT_TABLE
;
82 static unsigned int max_tcp_slot_table_limit
= RPC_MAX_SLOT_TABLE_LIMIT
;
83 static unsigned int xprt_min_resvport_limit
= RPC_MIN_RESVPORT
;
84 static unsigned int xprt_max_resvport_limit
= RPC_MAX_RESVPORT
;
86 static struct ctl_table_header
*sunrpc_table_header
;
89 * FIXME: changing the UDP slot table size should also resize the UDP
90 * socket buffers for existing UDP transports
92 static struct ctl_table xs_tunables_table
[] = {
94 .procname
= "udp_slot_table_entries",
95 .data
= &xprt_udp_slot_table_entries
,
96 .maxlen
= sizeof(unsigned int),
98 .proc_handler
= proc_dointvec_minmax
,
99 .extra1
= &min_slot_table_size
,
100 .extra2
= &max_slot_table_size
103 .procname
= "tcp_slot_table_entries",
104 .data
= &xprt_tcp_slot_table_entries
,
105 .maxlen
= sizeof(unsigned int),
107 .proc_handler
= proc_dointvec_minmax
,
108 .extra1
= &min_slot_table_size
,
109 .extra2
= &max_slot_table_size
112 .procname
= "tcp_max_slot_table_entries",
113 .data
= &xprt_max_tcp_slot_table_entries
,
114 .maxlen
= sizeof(unsigned int),
116 .proc_handler
= proc_dointvec_minmax
,
117 .extra1
= &min_slot_table_size
,
118 .extra2
= &max_tcp_slot_table_limit
121 .procname
= "min_resvport",
122 .data
= &xprt_min_resvport
,
123 .maxlen
= sizeof(unsigned int),
125 .proc_handler
= proc_dointvec_minmax
,
126 .extra1
= &xprt_min_resvport_limit
,
127 .extra2
= &xprt_max_resvport_limit
130 .procname
= "max_resvport",
131 .data
= &xprt_max_resvport
,
132 .maxlen
= sizeof(unsigned int),
134 .proc_handler
= proc_dointvec_minmax
,
135 .extra1
= &xprt_min_resvport_limit
,
136 .extra2
= &xprt_max_resvport_limit
139 .procname
= "tcp_fin_timeout",
140 .data
= &xs_tcp_fin_timeout
,
141 .maxlen
= sizeof(xs_tcp_fin_timeout
),
143 .proc_handler
= proc_dointvec_jiffies
,
148 static struct ctl_table sunrpc_table
[] = {
150 .procname
= "sunrpc",
152 .child
= xs_tunables_table
160 * Wait duration for a reply from the RPC portmapper.
162 #define XS_BIND_TO (60U * HZ)
165 * Delay if a UDP socket connect error occurs. This is most likely some
166 * kind of resource problem on the local host.
168 #define XS_UDP_REEST_TO (2U * HZ)
171 * The reestablish timeout allows clients to delay for a bit before attempting
172 * to reconnect to a server that just dropped our connection.
174 * We implement an exponential backoff when trying to reestablish a TCP
175 * transport connection with the server. Some servers like to drop a TCP
176 * connection when they are overworked, so we start with a short timeout and
177 * increase over time if the server is down or not responding.
179 #define XS_TCP_INIT_REEST_TO (3U * HZ)
180 #define XS_TCP_MAX_REEST_TO (5U * 60 * HZ)
183 * TCP idle timeout; client drops the transport socket if it is idle
184 * for this long. Note that we also timeout UDP sockets to prevent
185 * holding port numbers when there is no RPC traffic.
187 #define XS_IDLE_DISC_TO (5U * 60 * HZ)
190 # undef RPC_DEBUG_DATA
191 # define RPCDBG_FACILITY RPCDBG_TRANS
194 #ifdef RPC_DEBUG_DATA
195 static void xs_pktdump(char *msg
, u32
*packet
, unsigned int count
)
197 u8
*buf
= (u8
*) packet
;
200 dprintk("RPC: %s\n", msg
);
201 for (j
= 0; j
< count
&& j
< 128; j
+= 4) {
205 dprintk("0x%04x ", j
);
207 dprintk("%02x%02x%02x%02x ",
208 buf
[j
], buf
[j
+1], buf
[j
+2], buf
[j
+3]);
213 static inline void xs_pktdump(char *msg
, u32
*packet
, unsigned int count
)
220 struct rpc_xprt xprt
;
225 struct socket
* sock
;
229 * State of TCP reply receive
238 unsigned long tcp_copied
,
242 * Connection of transports
244 struct delayed_work connect_worker
;
245 struct sockaddr_storage srcaddr
;
246 unsigned short srcport
;
249 * UDP socket buffer size parameters
255 * Saved socket callback addresses
257 void (*old_data_ready
)(struct sock
*, int);
258 void (*old_state_change
)(struct sock
*);
259 void (*old_write_space
)(struct sock
*);
263 * TCP receive state flags
265 #define TCP_RCV_LAST_FRAG (1UL << 0)
266 #define TCP_RCV_COPY_FRAGHDR (1UL << 1)
267 #define TCP_RCV_COPY_XID (1UL << 2)
268 #define TCP_RCV_COPY_DATA (1UL << 3)
269 #define TCP_RCV_READ_CALLDIR (1UL << 4)
270 #define TCP_RCV_COPY_CALLDIR (1UL << 5)
275 #define TCP_RPC_REPLY (1UL << 6)
277 static inline struct sockaddr
*xs_addr(struct rpc_xprt
*xprt
)
279 return (struct sockaddr
*) &xprt
->addr
;
282 static inline struct sockaddr_un
*xs_addr_un(struct rpc_xprt
*xprt
)
284 return (struct sockaddr_un
*) &xprt
->addr
;
287 static inline struct sockaddr_in
*xs_addr_in(struct rpc_xprt
*xprt
)
289 return (struct sockaddr_in
*) &xprt
->addr
;
292 static inline struct sockaddr_in6
*xs_addr_in6(struct rpc_xprt
*xprt
)
294 return (struct sockaddr_in6
*) &xprt
->addr
;
297 static void xs_format_common_peer_addresses(struct rpc_xprt
*xprt
)
299 struct sockaddr
*sap
= xs_addr(xprt
);
300 struct sockaddr_in6
*sin6
;
301 struct sockaddr_in
*sin
;
302 struct sockaddr_un
*sun
;
305 switch (sap
->sa_family
) {
307 sun
= xs_addr_un(xprt
);
308 strlcpy(buf
, sun
->sun_path
, sizeof(buf
));
309 xprt
->address_strings
[RPC_DISPLAY_ADDR
] =
310 kstrdup(buf
, GFP_KERNEL
);
313 (void)rpc_ntop(sap
, buf
, sizeof(buf
));
314 xprt
->address_strings
[RPC_DISPLAY_ADDR
] =
315 kstrdup(buf
, GFP_KERNEL
);
316 sin
= xs_addr_in(xprt
);
317 snprintf(buf
, sizeof(buf
), "%08x", ntohl(sin
->sin_addr
.s_addr
));
320 (void)rpc_ntop(sap
, buf
, sizeof(buf
));
321 xprt
->address_strings
[RPC_DISPLAY_ADDR
] =
322 kstrdup(buf
, GFP_KERNEL
);
323 sin6
= xs_addr_in6(xprt
);
324 snprintf(buf
, sizeof(buf
), "%pi6", &sin6
->sin6_addr
);
330 xprt
->address_strings
[RPC_DISPLAY_HEX_ADDR
] = kstrdup(buf
, GFP_KERNEL
);
333 static void xs_format_common_peer_ports(struct rpc_xprt
*xprt
)
335 struct sockaddr
*sap
= xs_addr(xprt
);
338 snprintf(buf
, sizeof(buf
), "%u", rpc_get_port(sap
));
339 xprt
->address_strings
[RPC_DISPLAY_PORT
] = kstrdup(buf
, GFP_KERNEL
);
341 snprintf(buf
, sizeof(buf
), "%4hx", rpc_get_port(sap
));
342 xprt
->address_strings
[RPC_DISPLAY_HEX_PORT
] = kstrdup(buf
, GFP_KERNEL
);
345 static void xs_format_peer_addresses(struct rpc_xprt
*xprt
,
346 const char *protocol
,
349 xprt
->address_strings
[RPC_DISPLAY_PROTO
] = protocol
;
350 xprt
->address_strings
[RPC_DISPLAY_NETID
] = netid
;
351 xs_format_common_peer_addresses(xprt
);
352 xs_format_common_peer_ports(xprt
);
355 static void xs_update_peer_port(struct rpc_xprt
*xprt
)
357 kfree(xprt
->address_strings
[RPC_DISPLAY_HEX_PORT
]);
358 kfree(xprt
->address_strings
[RPC_DISPLAY_PORT
]);
360 xs_format_common_peer_ports(xprt
);
363 static void xs_free_peer_addresses(struct rpc_xprt
*xprt
)
367 for (i
= 0; i
< RPC_DISPLAY_MAX
; i
++)
369 case RPC_DISPLAY_PROTO
:
370 case RPC_DISPLAY_NETID
:
373 kfree(xprt
->address_strings
[i
]);
377 #define XS_SENDMSG_FLAGS (MSG_DONTWAIT | MSG_NOSIGNAL)
379 static int xs_send_kvec(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
, struct kvec
*vec
, unsigned int base
, int more
)
381 struct msghdr msg
= {
383 .msg_namelen
= addrlen
,
384 .msg_flags
= XS_SENDMSG_FLAGS
| (more
? MSG_MORE
: 0),
387 .iov_base
= vec
->iov_base
+ base
,
388 .iov_len
= vec
->iov_len
- base
,
391 if (iov
.iov_len
!= 0)
392 return kernel_sendmsg(sock
, &msg
, &iov
, 1, iov
.iov_len
);
393 return kernel_sendmsg(sock
, &msg
, NULL
, 0, 0);
396 static int xs_send_pagedata(struct socket
*sock
, struct xdr_buf
*xdr
, unsigned int base
, int more
)
399 unsigned int remainder
;
402 remainder
= xdr
->page_len
- base
;
403 base
+= xdr
->page_base
;
404 ppage
= xdr
->pages
+ (base
>> PAGE_SHIFT
);
407 unsigned int len
= min_t(unsigned int, PAGE_SIZE
- base
, remainder
);
408 int flags
= XS_SENDMSG_FLAGS
;
411 if (remainder
!= 0 || more
)
413 err
= sock
->ops
->sendpage(sock
, *ppage
, base
, len
, flags
);
414 if (remainder
== 0 || err
!= len
)
428 * xs_sendpages - write pages directly to a socket
429 * @sock: socket to send on
430 * @addr: UDP only -- address of destination
431 * @addrlen: UDP only -- length of destination address
432 * @xdr: buffer containing this request
433 * @base: starting position in the buffer
436 static int xs_sendpages(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
, struct xdr_buf
*xdr
, unsigned int base
)
438 unsigned int remainder
= xdr
->len
- base
;
444 clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
);
450 if (base
< xdr
->head
[0].iov_len
|| addr
!= NULL
) {
451 unsigned int len
= xdr
->head
[0].iov_len
- base
;
453 err
= xs_send_kvec(sock
, addr
, addrlen
, &xdr
->head
[0], base
, remainder
!= 0);
454 if (remainder
== 0 || err
!= len
)
459 base
-= xdr
->head
[0].iov_len
;
461 if (base
< xdr
->page_len
) {
462 unsigned int len
= xdr
->page_len
- base
;
464 err
= xs_send_pagedata(sock
, xdr
, base
, remainder
!= 0);
465 if (remainder
== 0 || err
!= len
)
470 base
-= xdr
->page_len
;
472 if (base
>= xdr
->tail
[0].iov_len
)
474 err
= xs_send_kvec(sock
, NULL
, 0, &xdr
->tail
[0], base
, 0);
483 static void xs_nospace_callback(struct rpc_task
*task
)
485 struct sock_xprt
*transport
= container_of(task
->tk_rqstp
->rq_xprt
, struct sock_xprt
, xprt
);
487 transport
->inet
->sk_write_pending
--;
488 clear_bit(SOCK_ASYNC_NOSPACE
, &transport
->sock
->flags
);
492 * xs_nospace - place task on wait queue if transmit was incomplete
493 * @task: task to put to sleep
496 static int xs_nospace(struct rpc_task
*task
)
498 struct rpc_rqst
*req
= task
->tk_rqstp
;
499 struct rpc_xprt
*xprt
= req
->rq_xprt
;
500 struct sock_xprt
*transport
= container_of(xprt
, struct sock_xprt
, xprt
);
503 dprintk("RPC: %5u xmit incomplete (%u left of %u)\n",
504 task
->tk_pid
, req
->rq_slen
- req
->rq_bytes_sent
,
507 /* Protect against races with write_space */
508 spin_lock_bh(&xprt
->transport_lock
);
510 /* Don't race with disconnect */
511 if (xprt_connected(xprt
)) {
512 if (test_bit(SOCK_ASYNC_NOSPACE
, &transport
->sock
->flags
)) {
514 * Notify TCP that we're limited by the application
517 set_bit(SOCK_NOSPACE
, &transport
->sock
->flags
);
518 transport
->inet
->sk_write_pending
++;
519 /* ...and wait for more buffer space */
520 xprt_wait_for_buffer_space(task
, xs_nospace_callback
);
523 clear_bit(SOCK_ASYNC_NOSPACE
, &transport
->sock
->flags
);
527 spin_unlock_bh(&xprt
->transport_lock
);
532 * Construct a stream transport record marker in @buf.
534 static inline void xs_encode_stream_record_marker(struct xdr_buf
*buf
)
536 u32 reclen
= buf
->len
- sizeof(rpc_fraghdr
);
537 rpc_fraghdr
*base
= buf
->head
[0].iov_base
;
538 *base
= cpu_to_be32(RPC_LAST_STREAM_FRAGMENT
| reclen
);
542 * xs_local_send_request - write an RPC request to an AF_LOCAL socket
543 * @task: RPC task that manages the state of an RPC request
546 * 0: The request has been sent
547 * EAGAIN: The socket was blocked, please call again later to
548 * complete the request
549 * ENOTCONN: Caller needs to invoke connect logic then call again
550 * other: Some other error occured, the request was not sent
552 static int xs_local_send_request(struct rpc_task
*task
)
554 struct rpc_rqst
*req
= task
->tk_rqstp
;
555 struct rpc_xprt
*xprt
= req
->rq_xprt
;
556 struct sock_xprt
*transport
=
557 container_of(xprt
, struct sock_xprt
, xprt
);
558 struct xdr_buf
*xdr
= &req
->rq_snd_buf
;
561 xs_encode_stream_record_marker(&req
->rq_snd_buf
);
563 xs_pktdump("packet data:",
564 req
->rq_svec
->iov_base
, req
->rq_svec
->iov_len
);
566 status
= xs_sendpages(transport
->sock
, NULL
, 0,
567 xdr
, req
->rq_bytes_sent
);
568 dprintk("RPC: %s(%u) = %d\n",
569 __func__
, xdr
->len
- req
->rq_bytes_sent
, status
);
570 if (likely(status
>= 0)) {
571 req
->rq_bytes_sent
+= status
;
572 req
->rq_xmit_bytes_sent
+= status
;
573 if (likely(req
->rq_bytes_sent
>= req
->rq_slen
)) {
574 req
->rq_bytes_sent
= 0;
582 status
= xs_nospace(task
);
585 dprintk("RPC: sendmsg returned unrecognized error %d\n",
596 * xs_udp_send_request - write an RPC request to a UDP socket
597 * @task: address of RPC task that manages the state of an RPC request
600 * 0: The request has been sent
601 * EAGAIN: The socket was blocked, please call again later to
602 * complete the request
603 * ENOTCONN: Caller needs to invoke connect logic then call again
604 * other: Some other error occurred, the request was not sent
606 static int xs_udp_send_request(struct rpc_task
*task
)
608 struct rpc_rqst
*req
= task
->tk_rqstp
;
609 struct rpc_xprt
*xprt
= req
->rq_xprt
;
610 struct sock_xprt
*transport
= container_of(xprt
, struct sock_xprt
, xprt
);
611 struct xdr_buf
*xdr
= &req
->rq_snd_buf
;
614 xs_pktdump("packet data:",
615 req
->rq_svec
->iov_base
,
616 req
->rq_svec
->iov_len
);
618 if (!xprt_bound(xprt
))
620 status
= xs_sendpages(transport
->sock
,
625 dprintk("RPC: xs_udp_send_request(%u) = %d\n",
626 xdr
->len
- req
->rq_bytes_sent
, status
);
629 req
->rq_xmit_bytes_sent
+= status
;
630 if (status
>= req
->rq_slen
)
632 /* Still some bytes left; set up for a retry later. */
639 /* Should we call xs_close() here? */
642 status
= xs_nospace(task
);
645 dprintk("RPC: sendmsg returned unrecognized error %d\n",
650 /* When the server has died, an ICMP port unreachable message
651 * prompts ECONNREFUSED. */
652 clear_bit(SOCK_ASYNC_NOSPACE
, &transport
->sock
->flags
);
659 * xs_tcp_shutdown - gracefully shut down a TCP socket
662 * Initiates a graceful shutdown of the TCP socket by calling the
663 * equivalent of shutdown(SHUT_WR);
665 static void xs_tcp_shutdown(struct rpc_xprt
*xprt
)
667 struct sock_xprt
*transport
= container_of(xprt
, struct sock_xprt
, xprt
);
668 struct socket
*sock
= transport
->sock
;
671 kernel_sock_shutdown(sock
, SHUT_WR
);
672 trace_rpc_socket_shutdown(xprt
, sock
);
677 * xs_tcp_send_request - write an RPC request to a TCP socket
678 * @task: address of RPC task that manages the state of an RPC request
681 * 0: The request has been sent
682 * EAGAIN: The socket was blocked, please call again later to
683 * complete the request
684 * ENOTCONN: Caller needs to invoke connect logic then call again
685 * other: Some other error occurred, the request was not sent
687 * XXX: In the case of soft timeouts, should we eventually give up
688 * if sendmsg is not able to make progress?
690 static int xs_tcp_send_request(struct rpc_task
*task
)
692 struct rpc_rqst
*req
= task
->tk_rqstp
;
693 struct rpc_xprt
*xprt
= req
->rq_xprt
;
694 struct sock_xprt
*transport
= container_of(xprt
, struct sock_xprt
, xprt
);
695 struct xdr_buf
*xdr
= &req
->rq_snd_buf
;
698 xs_encode_stream_record_marker(&req
->rq_snd_buf
);
700 xs_pktdump("packet data:",
701 req
->rq_svec
->iov_base
,
702 req
->rq_svec
->iov_len
);
704 /* Continue transmitting the packet/record. We must be careful
705 * to cope with writespace callbacks arriving _after_ we have
706 * called sendmsg(). */
708 status
= xs_sendpages(transport
->sock
,
709 NULL
, 0, xdr
, req
->rq_bytes_sent
);
711 dprintk("RPC: xs_tcp_send_request(%u) = %d\n",
712 xdr
->len
- req
->rq_bytes_sent
, status
);
714 if (unlikely(status
< 0))
717 /* If we've sent the entire packet, immediately
718 * reset the count of bytes sent. */
719 req
->rq_bytes_sent
+= status
;
720 req
->rq_xmit_bytes_sent
+= status
;
721 if (likely(req
->rq_bytes_sent
>= req
->rq_slen
)) {
722 req
->rq_bytes_sent
= 0;
735 /* Should we call xs_close() here? */
738 status
= xs_nospace(task
);
741 dprintk("RPC: sendmsg returned unrecognized error %d\n",
744 xs_tcp_shutdown(xprt
);
748 clear_bit(SOCK_ASYNC_NOSPACE
, &transport
->sock
->flags
);
755 * xs_tcp_release_xprt - clean up after a tcp transmission
759 * This cleans up if an error causes us to abort the transmission of a request.
760 * In this case, the socket may need to be reset in order to avoid confusing
763 static void xs_tcp_release_xprt(struct rpc_xprt
*xprt
, struct rpc_task
*task
)
765 struct rpc_rqst
*req
;
767 if (task
!= xprt
->snd_task
)
771 req
= task
->tk_rqstp
;
774 if (req
->rq_bytes_sent
== 0)
776 if (req
->rq_bytes_sent
== req
->rq_snd_buf
.len
)
778 set_bit(XPRT_CLOSE_WAIT
, &xprt
->state
);
780 xprt_release_xprt(xprt
, task
);
783 static void xs_save_old_callbacks(struct sock_xprt
*transport
, struct sock
*sk
)
785 transport
->old_data_ready
= sk
->sk_data_ready
;
786 transport
->old_state_change
= sk
->sk_state_change
;
787 transport
->old_write_space
= sk
->sk_write_space
;
790 static void xs_restore_old_callbacks(struct sock_xprt
*transport
, struct sock
*sk
)
792 sk
->sk_data_ready
= transport
->old_data_ready
;
793 sk
->sk_state_change
= transport
->old_state_change
;
794 sk
->sk_write_space
= transport
->old_write_space
;
797 static void xs_reset_transport(struct sock_xprt
*transport
)
799 struct socket
*sock
= transport
->sock
;
800 struct sock
*sk
= transport
->inet
;
805 transport
->srcport
= 0;
807 write_lock_bh(&sk
->sk_callback_lock
);
808 transport
->inet
= NULL
;
809 transport
->sock
= NULL
;
811 sk
->sk_user_data
= NULL
;
813 xs_restore_old_callbacks(transport
, sk
);
814 write_unlock_bh(&sk
->sk_callback_lock
);
818 trace_rpc_socket_close(&transport
->xprt
, sock
);
823 * xs_close - close a socket
826 * This is used when all requests are complete; ie, no DRC state remains
827 * on the server we want to save.
829 * The caller _must_ be holding XPRT_LOCKED in order to avoid issues with
830 * xs_reset_transport() zeroing the socket from underneath a writer.
832 static void xs_close(struct rpc_xprt
*xprt
)
834 struct sock_xprt
*transport
= container_of(xprt
, struct sock_xprt
, xprt
);
836 dprintk("RPC: xs_close xprt %p\n", xprt
);
838 xs_reset_transport(transport
);
839 xprt
->reestablish_timeout
= 0;
841 smp_mb__before_clear_bit();
842 clear_bit(XPRT_CONNECTION_ABORT
, &xprt
->state
);
843 clear_bit(XPRT_CLOSE_WAIT
, &xprt
->state
);
844 clear_bit(XPRT_CLOSING
, &xprt
->state
);
845 smp_mb__after_clear_bit();
846 xprt_disconnect_done(xprt
);
849 static void xs_tcp_close(struct rpc_xprt
*xprt
)
851 if (test_and_clear_bit(XPRT_CONNECTION_CLOSE
, &xprt
->state
))
854 xs_tcp_shutdown(xprt
);
857 static void xs_local_destroy(struct rpc_xprt
*xprt
)
860 xs_free_peer_addresses(xprt
);
862 module_put(THIS_MODULE
);
866 * xs_destroy - prepare to shutdown a transport
867 * @xprt: doomed transport
870 static void xs_destroy(struct rpc_xprt
*xprt
)
872 struct sock_xprt
*transport
= container_of(xprt
, struct sock_xprt
, xprt
);
874 dprintk("RPC: xs_destroy xprt %p\n", xprt
);
876 cancel_delayed_work_sync(&transport
->connect_worker
);
878 xs_local_destroy(xprt
);
881 static inline struct rpc_xprt
*xprt_from_sock(struct sock
*sk
)
883 return (struct rpc_xprt
*) sk
->sk_user_data
;
886 static int xs_local_copy_to_xdr(struct xdr_buf
*xdr
, struct sk_buff
*skb
)
888 struct xdr_skb_reader desc
= {
890 .offset
= sizeof(rpc_fraghdr
),
891 .count
= skb
->len
- sizeof(rpc_fraghdr
),
894 if (xdr_partial_copy_from_skb(xdr
, 0, &desc
, xdr_skb_read_bits
) < 0)
902 * xs_local_data_ready - "data ready" callback for AF_LOCAL sockets
903 * @sk: socket with data to read
904 * @len: how much data to read
906 * Currently this assumes we can read the whole reply in a single gulp.
908 static void xs_local_data_ready(struct sock
*sk
, int len
)
910 struct rpc_task
*task
;
911 struct rpc_xprt
*xprt
;
912 struct rpc_rqst
*rovr
;
914 int err
, repsize
, copied
;
918 read_lock_bh(&sk
->sk_callback_lock
);
919 dprintk("RPC: %s...\n", __func__
);
920 xprt
= xprt_from_sock(sk
);
924 skb
= skb_recv_datagram(sk
, 0, 1, &err
);
928 repsize
= skb
->len
- sizeof(rpc_fraghdr
);
930 dprintk("RPC: impossible RPC reply size %d\n", repsize
);
934 /* Copy the XID from the skb... */
935 xp
= skb_header_pointer(skb
, sizeof(rpc_fraghdr
), sizeof(_xid
), &_xid
);
939 /* Look up and lock the request corresponding to the given XID */
940 spin_lock(&xprt
->transport_lock
);
941 rovr
= xprt_lookup_rqst(xprt
, *xp
);
944 task
= rovr
->rq_task
;
946 copied
= rovr
->rq_private_buf
.buflen
;
947 if (copied
> repsize
)
950 if (xs_local_copy_to_xdr(&rovr
->rq_private_buf
, skb
)) {
951 dprintk("RPC: sk_buff copy failed\n");
955 xprt_complete_rqst(task
, copied
);
958 spin_unlock(&xprt
->transport_lock
);
960 skb_free_datagram(sk
, skb
);
962 read_unlock_bh(&sk
->sk_callback_lock
);
966 * xs_udp_data_ready - "data ready" callback for UDP sockets
967 * @sk: socket with data to read
968 * @len: how much data to read
971 static void xs_udp_data_ready(struct sock
*sk
, int len
)
973 struct rpc_task
*task
;
974 struct rpc_xprt
*xprt
;
975 struct rpc_rqst
*rovr
;
977 int err
, repsize
, copied
;
981 read_lock_bh(&sk
->sk_callback_lock
);
982 dprintk("RPC: xs_udp_data_ready...\n");
983 if (!(xprt
= xprt_from_sock(sk
)))
986 if ((skb
= skb_recv_datagram(sk
, 0, 1, &err
)) == NULL
)
989 repsize
= skb
->len
- sizeof(struct udphdr
);
991 dprintk("RPC: impossible RPC reply size %d!\n", repsize
);
995 /* Copy the XID from the skb... */
996 xp
= skb_header_pointer(skb
, sizeof(struct udphdr
),
997 sizeof(_xid
), &_xid
);
1001 /* Look up and lock the request corresponding to the given XID */
1002 spin_lock(&xprt
->transport_lock
);
1003 rovr
= xprt_lookup_rqst(xprt
, *xp
);
1006 task
= rovr
->rq_task
;
1008 if ((copied
= rovr
->rq_private_buf
.buflen
) > repsize
)
1011 /* Suck it into the iovec, verify checksum if not done by hw. */
1012 if (csum_partial_copy_to_xdr(&rovr
->rq_private_buf
, skb
)) {
1013 UDPX_INC_STATS_BH(sk
, UDP_MIB_INERRORS
);
1017 UDPX_INC_STATS_BH(sk
, UDP_MIB_INDATAGRAMS
);
1019 xprt_adjust_cwnd(xprt
, task
, copied
);
1020 xprt_complete_rqst(task
, copied
);
1023 spin_unlock(&xprt
->transport_lock
);
1025 skb_free_datagram(sk
, skb
);
1027 read_unlock_bh(&sk
->sk_callback_lock
);
1031 * Helper function to force a TCP close if the server is sending
1032 * junk and/or it has put us in CLOSE_WAIT
1034 static void xs_tcp_force_close(struct rpc_xprt
*xprt
)
1036 set_bit(XPRT_CONNECTION_CLOSE
, &xprt
->state
);
1037 xprt_force_disconnect(xprt
);
1040 static inline void xs_tcp_read_fraghdr(struct rpc_xprt
*xprt
, struct xdr_skb_reader
*desc
)
1042 struct sock_xprt
*transport
= container_of(xprt
, struct sock_xprt
, xprt
);
1046 p
= ((char *) &transport
->tcp_fraghdr
) + transport
->tcp_offset
;
1047 len
= sizeof(transport
->tcp_fraghdr
) - transport
->tcp_offset
;
1048 used
= xdr_skb_read_bits(desc
, p
, len
);
1049 transport
->tcp_offset
+= used
;
1053 transport
->tcp_reclen
= ntohl(transport
->tcp_fraghdr
);
1054 if (transport
->tcp_reclen
& RPC_LAST_STREAM_FRAGMENT
)
1055 transport
->tcp_flags
|= TCP_RCV_LAST_FRAG
;
1057 transport
->tcp_flags
&= ~TCP_RCV_LAST_FRAG
;
1058 transport
->tcp_reclen
&= RPC_FRAGMENT_SIZE_MASK
;
1060 transport
->tcp_flags
&= ~TCP_RCV_COPY_FRAGHDR
;
1061 transport
->tcp_offset
= 0;
1063 /* Sanity check of the record length */
1064 if (unlikely(transport
->tcp_reclen
< 8)) {
1065 dprintk("RPC: invalid TCP record fragment length\n");
1066 xs_tcp_force_close(xprt
);
1069 dprintk("RPC: reading TCP record fragment of length %d\n",
1070 transport
->tcp_reclen
);
1073 static void xs_tcp_check_fraghdr(struct sock_xprt
*transport
)
1075 if (transport
->tcp_offset
== transport
->tcp_reclen
) {
1076 transport
->tcp_flags
|= TCP_RCV_COPY_FRAGHDR
;
1077 transport
->tcp_offset
= 0;
1078 if (transport
->tcp_flags
& TCP_RCV_LAST_FRAG
) {
1079 transport
->tcp_flags
&= ~TCP_RCV_COPY_DATA
;
1080 transport
->tcp_flags
|= TCP_RCV_COPY_XID
;
1081 transport
->tcp_copied
= 0;
1086 static inline void xs_tcp_read_xid(struct sock_xprt
*transport
, struct xdr_skb_reader
*desc
)
1091 len
= sizeof(transport
->tcp_xid
) - transport
->tcp_offset
;
1092 dprintk("RPC: reading XID (%Zu bytes)\n", len
);
1093 p
= ((char *) &transport
->tcp_xid
) + transport
->tcp_offset
;
1094 used
= xdr_skb_read_bits(desc
, p
, len
);
1095 transport
->tcp_offset
+= used
;
1098 transport
->tcp_flags
&= ~TCP_RCV_COPY_XID
;
1099 transport
->tcp_flags
|= TCP_RCV_READ_CALLDIR
;
1100 transport
->tcp_copied
= 4;
1101 dprintk("RPC: reading %s XID %08x\n",
1102 (transport
->tcp_flags
& TCP_RPC_REPLY
) ? "reply for"
1104 ntohl(transport
->tcp_xid
));
1105 xs_tcp_check_fraghdr(transport
);
1108 static inline void xs_tcp_read_calldir(struct sock_xprt
*transport
,
1109 struct xdr_skb_reader
*desc
)
1116 * We want transport->tcp_offset to be 8 at the end of this routine
1117 * (4 bytes for the xid and 4 bytes for the call/reply flag).
1118 * When this function is called for the first time,
1119 * transport->tcp_offset is 4 (after having already read the xid).
1121 offset
= transport
->tcp_offset
- sizeof(transport
->tcp_xid
);
1122 len
= sizeof(transport
->tcp_calldir
) - offset
;
1123 dprintk("RPC: reading CALL/REPLY flag (%Zu bytes)\n", len
);
1124 p
= ((char *) &transport
->tcp_calldir
) + offset
;
1125 used
= xdr_skb_read_bits(desc
, p
, len
);
1126 transport
->tcp_offset
+= used
;
1129 transport
->tcp_flags
&= ~TCP_RCV_READ_CALLDIR
;
1131 * We don't yet have the XDR buffer, so we will write the calldir
1132 * out after we get the buffer from the 'struct rpc_rqst'
1134 switch (ntohl(transport
->tcp_calldir
)) {
1136 transport
->tcp_flags
|= TCP_RCV_COPY_CALLDIR
;
1137 transport
->tcp_flags
|= TCP_RCV_COPY_DATA
;
1138 transport
->tcp_flags
|= TCP_RPC_REPLY
;
1141 transport
->tcp_flags
|= TCP_RCV_COPY_CALLDIR
;
1142 transport
->tcp_flags
|= TCP_RCV_COPY_DATA
;
1143 transport
->tcp_flags
&= ~TCP_RPC_REPLY
;
1146 dprintk("RPC: invalid request message type\n");
1147 xs_tcp_force_close(&transport
->xprt
);
1149 xs_tcp_check_fraghdr(transport
);
1152 static inline void xs_tcp_read_common(struct rpc_xprt
*xprt
,
1153 struct xdr_skb_reader
*desc
,
1154 struct rpc_rqst
*req
)
1156 struct sock_xprt
*transport
=
1157 container_of(xprt
, struct sock_xprt
, xprt
);
1158 struct xdr_buf
*rcvbuf
;
1162 rcvbuf
= &req
->rq_private_buf
;
1164 if (transport
->tcp_flags
& TCP_RCV_COPY_CALLDIR
) {
1166 * Save the RPC direction in the XDR buffer
1168 memcpy(rcvbuf
->head
[0].iov_base
+ transport
->tcp_copied
,
1169 &transport
->tcp_calldir
,
1170 sizeof(transport
->tcp_calldir
));
1171 transport
->tcp_copied
+= sizeof(transport
->tcp_calldir
);
1172 transport
->tcp_flags
&= ~TCP_RCV_COPY_CALLDIR
;
1176 if (len
> transport
->tcp_reclen
- transport
->tcp_offset
) {
1177 struct xdr_skb_reader my_desc
;
1179 len
= transport
->tcp_reclen
- transport
->tcp_offset
;
1180 memcpy(&my_desc
, desc
, sizeof(my_desc
));
1181 my_desc
.count
= len
;
1182 r
= xdr_partial_copy_from_skb(rcvbuf
, transport
->tcp_copied
,
1183 &my_desc
, xdr_skb_read_bits
);
1187 r
= xdr_partial_copy_from_skb(rcvbuf
, transport
->tcp_copied
,
1188 desc
, xdr_skb_read_bits
);
1191 transport
->tcp_copied
+= r
;
1192 transport
->tcp_offset
+= r
;
1195 /* Error when copying to the receive buffer,
1196 * usually because we weren't able to allocate
1197 * additional buffer pages. All we can do now
1198 * is turn off TCP_RCV_COPY_DATA, so the request
1199 * will not receive any additional updates,
1201 * Any remaining data from this record will
1204 transport
->tcp_flags
&= ~TCP_RCV_COPY_DATA
;
1205 dprintk("RPC: XID %08x truncated request\n",
1206 ntohl(transport
->tcp_xid
));
1207 dprintk("RPC: xprt = %p, tcp_copied = %lu, "
1208 "tcp_offset = %u, tcp_reclen = %u\n",
1209 xprt
, transport
->tcp_copied
,
1210 transport
->tcp_offset
, transport
->tcp_reclen
);
1214 dprintk("RPC: XID %08x read %Zd bytes\n",
1215 ntohl(transport
->tcp_xid
), r
);
1216 dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, "
1217 "tcp_reclen = %u\n", xprt
, transport
->tcp_copied
,
1218 transport
->tcp_offset
, transport
->tcp_reclen
);
1220 if (transport
->tcp_copied
== req
->rq_private_buf
.buflen
)
1221 transport
->tcp_flags
&= ~TCP_RCV_COPY_DATA
;
1222 else if (transport
->tcp_offset
== transport
->tcp_reclen
) {
1223 if (transport
->tcp_flags
& TCP_RCV_LAST_FRAG
)
1224 transport
->tcp_flags
&= ~TCP_RCV_COPY_DATA
;
1229 * Finds the request corresponding to the RPC xid and invokes the common
1230 * tcp read code to read the data.
1232 static inline int xs_tcp_read_reply(struct rpc_xprt
*xprt
,
1233 struct xdr_skb_reader
*desc
)
1235 struct sock_xprt
*transport
=
1236 container_of(xprt
, struct sock_xprt
, xprt
);
1237 struct rpc_rqst
*req
;
1239 dprintk("RPC: read reply XID %08x\n", ntohl(transport
->tcp_xid
));
1241 /* Find and lock the request corresponding to this xid */
1242 spin_lock(&xprt
->transport_lock
);
1243 req
= xprt_lookup_rqst(xprt
, transport
->tcp_xid
);
1245 dprintk("RPC: XID %08x request not found!\n",
1246 ntohl(transport
->tcp_xid
));
1247 spin_unlock(&xprt
->transport_lock
);
1251 xs_tcp_read_common(xprt
, desc
, req
);
1253 if (!(transport
->tcp_flags
& TCP_RCV_COPY_DATA
))
1254 xprt_complete_rqst(req
->rq_task
, transport
->tcp_copied
);
1256 spin_unlock(&xprt
->transport_lock
);
1260 #if defined(CONFIG_SUNRPC_BACKCHANNEL)
1262 * Obtains an rpc_rqst previously allocated and invokes the common
1263 * tcp read code to read the data. The result is placed in the callback
1265 * If we're unable to obtain the rpc_rqst we schedule the closing of the
1266 * connection and return -1.
1268 static inline int xs_tcp_read_callback(struct rpc_xprt
*xprt
,
1269 struct xdr_skb_reader
*desc
)
1271 struct sock_xprt
*transport
=
1272 container_of(xprt
, struct sock_xprt
, xprt
);
1273 struct rpc_rqst
*req
;
1275 req
= xprt_alloc_bc_request(xprt
);
1277 printk(KERN_WARNING
"Callback slot table overflowed\n");
1278 xprt_force_disconnect(xprt
);
1282 req
->rq_xid
= transport
->tcp_xid
;
1283 dprintk("RPC: read callback XID %08x\n", ntohl(req
->rq_xid
));
1284 xs_tcp_read_common(xprt
, desc
, req
);
1286 if (!(transport
->tcp_flags
& TCP_RCV_COPY_DATA
)) {
1287 struct svc_serv
*bc_serv
= xprt
->bc_serv
;
1290 * Add callback request to callback list. The callback
1291 * service sleeps on the sv_cb_waitq waiting for new
1292 * requests. Wake it up after adding enqueing the
1295 dprintk("RPC: add callback request to list\n");
1296 spin_lock(&bc_serv
->sv_cb_lock
);
1297 list_add(&req
->rq_bc_list
, &bc_serv
->sv_cb_list
);
1298 spin_unlock(&bc_serv
->sv_cb_lock
);
1299 wake_up(&bc_serv
->sv_cb_waitq
);
1302 req
->rq_private_buf
.len
= transport
->tcp_copied
;
1307 static inline int _xs_tcp_read_data(struct rpc_xprt
*xprt
,
1308 struct xdr_skb_reader
*desc
)
1310 struct sock_xprt
*transport
=
1311 container_of(xprt
, struct sock_xprt
, xprt
);
1313 return (transport
->tcp_flags
& TCP_RPC_REPLY
) ?
1314 xs_tcp_read_reply(xprt
, desc
) :
1315 xs_tcp_read_callback(xprt
, desc
);
1318 static inline int _xs_tcp_read_data(struct rpc_xprt
*xprt
,
1319 struct xdr_skb_reader
*desc
)
1321 return xs_tcp_read_reply(xprt
, desc
);
1323 #endif /* CONFIG_SUNRPC_BACKCHANNEL */
1326 * Read data off the transport. This can be either an RPC_CALL or an
1327 * RPC_REPLY. Relay the processing to helper functions.
1329 static void xs_tcp_read_data(struct rpc_xprt
*xprt
,
1330 struct xdr_skb_reader
*desc
)
1332 struct sock_xprt
*transport
=
1333 container_of(xprt
, struct sock_xprt
, xprt
);
1335 if (_xs_tcp_read_data(xprt
, desc
) == 0)
1336 xs_tcp_check_fraghdr(transport
);
1339 * The transport_lock protects the request handling.
1340 * There's no need to hold it to update the tcp_flags.
1342 transport
->tcp_flags
&= ~TCP_RCV_COPY_DATA
;
1346 static inline void xs_tcp_read_discard(struct sock_xprt
*transport
, struct xdr_skb_reader
*desc
)
1350 len
= transport
->tcp_reclen
- transport
->tcp_offset
;
1351 if (len
> desc
->count
)
1354 desc
->offset
+= len
;
1355 transport
->tcp_offset
+= len
;
1356 dprintk("RPC: discarded %Zu bytes\n", len
);
1357 xs_tcp_check_fraghdr(transport
);
1360 static int xs_tcp_data_recv(read_descriptor_t
*rd_desc
, struct sk_buff
*skb
, unsigned int offset
, size_t len
)
1362 struct rpc_xprt
*xprt
= rd_desc
->arg
.data
;
1363 struct sock_xprt
*transport
= container_of(xprt
, struct sock_xprt
, xprt
);
1364 struct xdr_skb_reader desc
= {
1370 dprintk("RPC: xs_tcp_data_recv started\n");
1372 /* Read in a new fragment marker if necessary */
1373 /* Can we ever really expect to get completely empty fragments? */
1374 if (transport
->tcp_flags
& TCP_RCV_COPY_FRAGHDR
) {
1375 xs_tcp_read_fraghdr(xprt
, &desc
);
1378 /* Read in the xid if necessary */
1379 if (transport
->tcp_flags
& TCP_RCV_COPY_XID
) {
1380 xs_tcp_read_xid(transport
, &desc
);
1383 /* Read in the call/reply flag */
1384 if (transport
->tcp_flags
& TCP_RCV_READ_CALLDIR
) {
1385 xs_tcp_read_calldir(transport
, &desc
);
1388 /* Read in the request data */
1389 if (transport
->tcp_flags
& TCP_RCV_COPY_DATA
) {
1390 xs_tcp_read_data(xprt
, &desc
);
1393 /* Skip over any trailing bytes on short reads */
1394 xs_tcp_read_discard(transport
, &desc
);
1395 } while (desc
.count
);
1396 dprintk("RPC: xs_tcp_data_recv done\n");
1397 return len
- desc
.count
;
1401 * xs_tcp_data_ready - "data ready" callback for TCP sockets
1402 * @sk: socket with data to read
1403 * @bytes: how much data to read
1406 static void xs_tcp_data_ready(struct sock
*sk
, int bytes
)
1408 struct rpc_xprt
*xprt
;
1409 read_descriptor_t rd_desc
;
1412 dprintk("RPC: xs_tcp_data_ready...\n");
1414 read_lock_bh(&sk
->sk_callback_lock
);
1415 if (!(xprt
= xprt_from_sock(sk
)))
1417 /* Any data means we had a useful conversation, so
1418 * the we don't need to delay the next reconnect
1420 if (xprt
->reestablish_timeout
)
1421 xprt
->reestablish_timeout
= 0;
1423 /* We use rd_desc to pass struct xprt to xs_tcp_data_recv */
1424 rd_desc
.arg
.data
= xprt
;
1426 rd_desc
.count
= 65536;
1427 read
= tcp_read_sock(sk
, &rd_desc
, xs_tcp_data_recv
);
1430 read_unlock_bh(&sk
->sk_callback_lock
);
1434 * Do the equivalent of linger/linger2 handling for dealing with
1435 * broken servers that don't close the socket in a timely
1438 static void xs_tcp_schedule_linger_timeout(struct rpc_xprt
*xprt
,
1439 unsigned long timeout
)
1441 struct sock_xprt
*transport
;
1443 if (xprt_test_and_set_connecting(xprt
))
1445 set_bit(XPRT_CONNECTION_ABORT
, &xprt
->state
);
1446 transport
= container_of(xprt
, struct sock_xprt
, xprt
);
1447 queue_delayed_work(rpciod_workqueue
, &transport
->connect_worker
,
1451 static void xs_tcp_cancel_linger_timeout(struct rpc_xprt
*xprt
)
1453 struct sock_xprt
*transport
;
1455 transport
= container_of(xprt
, struct sock_xprt
, xprt
);
1457 if (!test_bit(XPRT_CONNECTION_ABORT
, &xprt
->state
) ||
1458 !cancel_delayed_work(&transport
->connect_worker
))
1460 clear_bit(XPRT_CONNECTION_ABORT
, &xprt
->state
);
1461 xprt_clear_connecting(xprt
);
1464 static void xs_sock_reset_connection_flags(struct rpc_xprt
*xprt
)
1466 smp_mb__before_clear_bit();
1467 clear_bit(XPRT_CONNECTION_ABORT
, &xprt
->state
);
1468 clear_bit(XPRT_CONNECTION_CLOSE
, &xprt
->state
);
1469 clear_bit(XPRT_CLOSE_WAIT
, &xprt
->state
);
1470 clear_bit(XPRT_CLOSING
, &xprt
->state
);
1471 smp_mb__after_clear_bit();
1474 static void xs_sock_mark_closed(struct rpc_xprt
*xprt
)
1476 xs_sock_reset_connection_flags(xprt
);
1477 /* Mark transport as closed and wake up all pending tasks */
1478 xprt_disconnect_done(xprt
);
1482 * xs_tcp_state_change - callback to handle TCP socket state changes
1483 * @sk: socket whose state has changed
1486 static void xs_tcp_state_change(struct sock
*sk
)
1488 struct rpc_xprt
*xprt
;
1490 read_lock_bh(&sk
->sk_callback_lock
);
1491 if (!(xprt
= xprt_from_sock(sk
)))
1493 dprintk("RPC: xs_tcp_state_change client %p...\n", xprt
);
1494 dprintk("RPC: state %x conn %d dead %d zapped %d sk_shutdown %d\n",
1495 sk
->sk_state
, xprt_connected(xprt
),
1496 sock_flag(sk
, SOCK_DEAD
),
1497 sock_flag(sk
, SOCK_ZAPPED
),
1500 trace_rpc_socket_state_change(xprt
, sk
->sk_socket
);
1501 switch (sk
->sk_state
) {
1502 case TCP_ESTABLISHED
:
1503 spin_lock(&xprt
->transport_lock
);
1504 if (!xprt_test_and_set_connected(xprt
)) {
1505 struct sock_xprt
*transport
= container_of(xprt
,
1506 struct sock_xprt
, xprt
);
1508 /* Reset TCP record info */
1509 transport
->tcp_offset
= 0;
1510 transport
->tcp_reclen
= 0;
1511 transport
->tcp_copied
= 0;
1512 transport
->tcp_flags
=
1513 TCP_RCV_COPY_FRAGHDR
| TCP_RCV_COPY_XID
;
1515 xprt_wake_pending_tasks(xprt
, -EAGAIN
);
1517 spin_unlock(&xprt
->transport_lock
);
1520 /* The client initiated a shutdown of the socket */
1521 xprt
->connect_cookie
++;
1522 xprt
->reestablish_timeout
= 0;
1523 set_bit(XPRT_CLOSING
, &xprt
->state
);
1524 smp_mb__before_clear_bit();
1525 clear_bit(XPRT_CONNECTED
, &xprt
->state
);
1526 clear_bit(XPRT_CLOSE_WAIT
, &xprt
->state
);
1527 smp_mb__after_clear_bit();
1528 xs_tcp_schedule_linger_timeout(xprt
, xs_tcp_fin_timeout
);
1530 case TCP_CLOSE_WAIT
:
1531 /* The server initiated a shutdown of the socket */
1532 xprt
->connect_cookie
++;
1533 clear_bit(XPRT_CONNECTED
, &xprt
->state
);
1534 xs_tcp_force_close(xprt
);
1537 * If the server closed down the connection, make sure that
1538 * we back off before reconnecting
1540 if (xprt
->reestablish_timeout
< XS_TCP_INIT_REEST_TO
)
1541 xprt
->reestablish_timeout
= XS_TCP_INIT_REEST_TO
;
1544 set_bit(XPRT_CLOSING
, &xprt
->state
);
1545 xs_tcp_schedule_linger_timeout(xprt
, xs_tcp_fin_timeout
);
1546 smp_mb__before_clear_bit();
1547 clear_bit(XPRT_CONNECTED
, &xprt
->state
);
1548 smp_mb__after_clear_bit();
1551 xs_tcp_cancel_linger_timeout(xprt
);
1552 xs_sock_mark_closed(xprt
);
1555 read_unlock_bh(&sk
->sk_callback_lock
);
1558 static void xs_write_space(struct sock
*sk
)
1560 struct socket
*sock
;
1561 struct rpc_xprt
*xprt
;
1563 if (unlikely(!(sock
= sk
->sk_socket
)))
1565 clear_bit(SOCK_NOSPACE
, &sock
->flags
);
1567 if (unlikely(!(xprt
= xprt_from_sock(sk
))))
1569 if (test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
) == 0)
1572 xprt_write_space(xprt
);
1576 * xs_udp_write_space - callback invoked when socket buffer space
1578 * @sk: socket whose state has changed
1580 * Called when more output buffer space is available for this socket.
1581 * We try not to wake our writers until they can make "significant"
1582 * progress, otherwise we'll waste resources thrashing kernel_sendmsg
1583 * with a bunch of small requests.
1585 static void xs_udp_write_space(struct sock
*sk
)
1587 read_lock_bh(&sk
->sk_callback_lock
);
1589 /* from net/core/sock.c:sock_def_write_space */
1590 if (sock_writeable(sk
))
1593 read_unlock_bh(&sk
->sk_callback_lock
);
1597 * xs_tcp_write_space - callback invoked when socket buffer space
1599 * @sk: socket whose state has changed
1601 * Called when more output buffer space is available for this socket.
1602 * We try not to wake our writers until they can make "significant"
1603 * progress, otherwise we'll waste resources thrashing kernel_sendmsg
1604 * with a bunch of small requests.
1606 static void xs_tcp_write_space(struct sock
*sk
)
1608 read_lock_bh(&sk
->sk_callback_lock
);
1610 /* from net/core/stream.c:sk_stream_write_space */
1611 if (sk_stream_is_writeable(sk
))
1614 read_unlock_bh(&sk
->sk_callback_lock
);
1617 static void xs_udp_do_set_buffer_size(struct rpc_xprt
*xprt
)
1619 struct sock_xprt
*transport
= container_of(xprt
, struct sock_xprt
, xprt
);
1620 struct sock
*sk
= transport
->inet
;
1622 if (transport
->rcvsize
) {
1623 sk
->sk_userlocks
|= SOCK_RCVBUF_LOCK
;
1624 sk
->sk_rcvbuf
= transport
->rcvsize
* xprt
->max_reqs
* 2;
1626 if (transport
->sndsize
) {
1627 sk
->sk_userlocks
|= SOCK_SNDBUF_LOCK
;
1628 sk
->sk_sndbuf
= transport
->sndsize
* xprt
->max_reqs
* 2;
1629 sk
->sk_write_space(sk
);
1634 * xs_udp_set_buffer_size - set send and receive limits
1635 * @xprt: generic transport
1636 * @sndsize: requested size of send buffer, in bytes
1637 * @rcvsize: requested size of receive buffer, in bytes
1639 * Set socket send and receive buffer size limits.
1641 static void xs_udp_set_buffer_size(struct rpc_xprt
*xprt
, size_t sndsize
, size_t rcvsize
)
1643 struct sock_xprt
*transport
= container_of(xprt
, struct sock_xprt
, xprt
);
1645 transport
->sndsize
= 0;
1647 transport
->sndsize
= sndsize
+ 1024;
1648 transport
->rcvsize
= 0;
1650 transport
->rcvsize
= rcvsize
+ 1024;
1652 xs_udp_do_set_buffer_size(xprt
);
1656 * xs_udp_timer - called when a retransmit timeout occurs on a UDP transport
1657 * @task: task that timed out
1659 * Adjust the congestion window after a retransmit timeout has occurred.
1661 static void xs_udp_timer(struct rpc_xprt
*xprt
, struct rpc_task
*task
)
1663 xprt_adjust_cwnd(xprt
, task
, -ETIMEDOUT
);
1666 static unsigned short xs_get_random_port(void)
1668 unsigned short range
= xprt_max_resvport
- xprt_min_resvport
;
1669 unsigned short rand
= (unsigned short) net_random() % range
;
1670 return rand
+ xprt_min_resvport
;
1674 * xs_set_port - reset the port number in the remote endpoint address
1675 * @xprt: generic transport
1676 * @port: new port number
1679 static void xs_set_port(struct rpc_xprt
*xprt
, unsigned short port
)
1681 dprintk("RPC: setting port for xprt %p to %u\n", xprt
, port
);
1683 rpc_set_port(xs_addr(xprt
), port
);
1684 xs_update_peer_port(xprt
);
1687 static unsigned short xs_get_srcport(struct sock_xprt
*transport
)
1689 unsigned short port
= transport
->srcport
;
1691 if (port
== 0 && transport
->xprt
.resvport
)
1692 port
= xs_get_random_port();
1696 static unsigned short xs_next_srcport(struct sock_xprt
*transport
, unsigned short port
)
1698 if (transport
->srcport
!= 0)
1699 transport
->srcport
= 0;
1700 if (!transport
->xprt
.resvport
)
1702 if (port
<= xprt_min_resvport
|| port
> xprt_max_resvport
)
1703 return xprt_max_resvport
;
1706 static int xs_bind(struct sock_xprt
*transport
, struct socket
*sock
)
1708 struct sockaddr_storage myaddr
;
1710 unsigned short port
= xs_get_srcport(transport
);
1711 unsigned short last
;
1713 memcpy(&myaddr
, &transport
->srcaddr
, transport
->xprt
.addrlen
);
1715 rpc_set_port((struct sockaddr
*)&myaddr
, port
);
1716 err
= kernel_bind(sock
, (struct sockaddr
*)&myaddr
,
1717 transport
->xprt
.addrlen
);
1721 transport
->srcport
= port
;
1725 port
= xs_next_srcport(transport
, port
);
1728 } while (err
== -EADDRINUSE
&& nloop
!= 2);
1730 if (myaddr
.ss_family
== AF_INET
)
1731 dprintk("RPC: %s %pI4:%u: %s (%d)\n", __func__
,
1732 &((struct sockaddr_in
*)&myaddr
)->sin_addr
,
1733 port
, err
? "failed" : "ok", err
);
1735 dprintk("RPC: %s %pI6:%u: %s (%d)\n", __func__
,
1736 &((struct sockaddr_in6
*)&myaddr
)->sin6_addr
,
1737 port
, err
? "failed" : "ok", err
);
1742 * We don't support autobind on AF_LOCAL sockets
1744 static void xs_local_rpcbind(struct rpc_task
*task
)
1747 xprt_set_bound(rcu_dereference(task
->tk_client
->cl_xprt
));
1751 static void xs_local_set_port(struct rpc_xprt
*xprt
, unsigned short port
)
1755 #ifdef CONFIG_DEBUG_LOCK_ALLOC
1756 static struct lock_class_key xs_key
[2];
1757 static struct lock_class_key xs_slock_key
[2];
1759 static inline void xs_reclassify_socketu(struct socket
*sock
)
1761 struct sock
*sk
= sock
->sk
;
1763 sock_lock_init_class_and_name(sk
, "slock-AF_LOCAL-RPC",
1764 &xs_slock_key
[1], "sk_lock-AF_LOCAL-RPC", &xs_key
[1]);
1767 static inline void xs_reclassify_socket4(struct socket
*sock
)
1769 struct sock
*sk
= sock
->sk
;
1771 sock_lock_init_class_and_name(sk
, "slock-AF_INET-RPC",
1772 &xs_slock_key
[0], "sk_lock-AF_INET-RPC", &xs_key
[0]);
1775 static inline void xs_reclassify_socket6(struct socket
*sock
)
1777 struct sock
*sk
= sock
->sk
;
1779 sock_lock_init_class_and_name(sk
, "slock-AF_INET6-RPC",
1780 &xs_slock_key
[1], "sk_lock-AF_INET6-RPC", &xs_key
[1]);
1783 static inline void xs_reclassify_socket(int family
, struct socket
*sock
)
1785 WARN_ON_ONCE(sock_owned_by_user(sock
->sk
));
1786 if (sock_owned_by_user(sock
->sk
))
1791 xs_reclassify_socketu(sock
);
1794 xs_reclassify_socket4(sock
);
1797 xs_reclassify_socket6(sock
);
1802 static inline void xs_reclassify_socketu(struct socket
*sock
)
1806 static inline void xs_reclassify_socket4(struct socket
*sock
)
1810 static inline void xs_reclassify_socket6(struct socket
*sock
)
1814 static inline void xs_reclassify_socket(int family
, struct socket
*sock
)
1819 static struct socket
*xs_create_sock(struct rpc_xprt
*xprt
,
1820 struct sock_xprt
*transport
, int family
, int type
, int protocol
)
1822 struct socket
*sock
;
1825 err
= __sock_create(xprt
->xprt_net
, family
, type
, protocol
, &sock
, 1);
1827 dprintk("RPC: can't create %d transport socket (%d).\n",
1831 xs_reclassify_socket(family
, sock
);
1833 err
= xs_bind(transport
, sock
);
1841 return ERR_PTR(err
);
1844 static int xs_local_finish_connecting(struct rpc_xprt
*xprt
,
1845 struct socket
*sock
)
1847 struct sock_xprt
*transport
= container_of(xprt
, struct sock_xprt
,
1850 if (!transport
->inet
) {
1851 struct sock
*sk
= sock
->sk
;
1853 write_lock_bh(&sk
->sk_callback_lock
);
1855 xs_save_old_callbacks(transport
, sk
);
1857 sk
->sk_user_data
= xprt
;
1858 sk
->sk_data_ready
= xs_local_data_ready
;
1859 sk
->sk_write_space
= xs_udp_write_space
;
1860 sk
->sk_allocation
= GFP_ATOMIC
;
1862 xprt_clear_connected(xprt
);
1864 /* Reset to new socket */
1865 transport
->sock
= sock
;
1866 transport
->inet
= sk
;
1868 write_unlock_bh(&sk
->sk_callback_lock
);
1871 /* Tell the socket layer to start connecting... */
1872 xprt
->stat
.connect_count
++;
1873 xprt
->stat
.connect_start
= jiffies
;
1874 return kernel_connect(sock
, xs_addr(xprt
), xprt
->addrlen
, 0);
1878 * xs_local_setup_socket - create AF_LOCAL socket, connect to a local endpoint
1879 * @xprt: RPC transport to connect
1880 * @transport: socket transport to connect
1881 * @create_sock: function to create a socket of the correct type
1883 static int xs_local_setup_socket(struct sock_xprt
*transport
)
1885 struct rpc_xprt
*xprt
= &transport
->xprt
;
1886 struct socket
*sock
;
1889 current
->flags
|= PF_FSTRANS
;
1891 clear_bit(XPRT_CONNECTION_ABORT
, &xprt
->state
);
1892 status
= __sock_create(xprt
->xprt_net
, AF_LOCAL
,
1893 SOCK_STREAM
, 0, &sock
, 1);
1895 dprintk("RPC: can't create AF_LOCAL "
1896 "transport socket (%d).\n", -status
);
1899 xs_reclassify_socketu(sock
);
1901 dprintk("RPC: worker connecting xprt %p via AF_LOCAL to %s\n",
1902 xprt
, xprt
->address_strings
[RPC_DISPLAY_ADDR
]);
1904 status
= xs_local_finish_connecting(xprt
, sock
);
1905 trace_rpc_socket_connect(xprt
, sock
, status
);
1908 dprintk("RPC: xprt %p connected to %s\n",
1909 xprt
, xprt
->address_strings
[RPC_DISPLAY_ADDR
]);
1910 xprt_set_connected(xprt
);
1913 dprintk("RPC: xprt %p: socket %s does not exist\n",
1914 xprt
, xprt
->address_strings
[RPC_DISPLAY_ADDR
]);
1917 dprintk("RPC: xprt %p: connection refused for %s\n",
1918 xprt
, xprt
->address_strings
[RPC_DISPLAY_ADDR
]);
1921 printk(KERN_ERR
"%s: unhandled error (%d) connecting to %s\n",
1923 xprt
->address_strings
[RPC_DISPLAY_ADDR
]);
1927 xprt_clear_connecting(xprt
);
1928 xprt_wake_pending_tasks(xprt
, status
);
1929 current
->flags
&= ~PF_FSTRANS
;
1933 static void xs_local_connect(struct rpc_xprt
*xprt
, struct rpc_task
*task
)
1935 struct sock_xprt
*transport
= container_of(xprt
, struct sock_xprt
, xprt
);
1938 if (RPC_IS_ASYNC(task
)) {
1940 * We want the AF_LOCAL connect to be resolved in the
1941 * filesystem namespace of the process making the rpc
1942 * call. Thus we connect synchronously.
1944 * If we want to support asynchronous AF_LOCAL calls,
1945 * we'll need to figure out how to pass a namespace to
1948 rpc_exit(task
, -ENOTCONN
);
1951 ret
= xs_local_setup_socket(transport
);
1952 if (ret
&& !RPC_IS_SOFTCONN(task
))
1953 msleep_interruptible(15000);
1956 #ifdef CONFIG_SUNRPC_SWAP
1957 static void xs_set_memalloc(struct rpc_xprt
*xprt
)
1959 struct sock_xprt
*transport
= container_of(xprt
, struct sock_xprt
,
1963 sk_set_memalloc(transport
->inet
);
1967 * xs_swapper - Tag this transport as being used for swap.
1968 * @xprt: transport to tag
1969 * @enable: enable/disable
1972 int xs_swapper(struct rpc_xprt
*xprt
, int enable
)
1974 struct sock_xprt
*transport
= container_of(xprt
, struct sock_xprt
,
1980 xs_set_memalloc(xprt
);
1981 } else if (xprt
->swapper
) {
1983 sk_clear_memalloc(transport
->inet
);
1988 EXPORT_SYMBOL_GPL(xs_swapper
);
1990 static void xs_set_memalloc(struct rpc_xprt
*xprt
)
1995 static void xs_udp_finish_connecting(struct rpc_xprt
*xprt
, struct socket
*sock
)
1997 struct sock_xprt
*transport
= container_of(xprt
, struct sock_xprt
, xprt
);
1999 if (!transport
->inet
) {
2000 struct sock
*sk
= sock
->sk
;
2002 write_lock_bh(&sk
->sk_callback_lock
);
2004 xs_save_old_callbacks(transport
, sk
);
2006 sk
->sk_user_data
= xprt
;
2007 sk
->sk_data_ready
= xs_udp_data_ready
;
2008 sk
->sk_write_space
= xs_udp_write_space
;
2009 sk
->sk_no_check
= UDP_CSUM_NORCV
;
2010 sk
->sk_allocation
= GFP_ATOMIC
;
2012 xprt_set_connected(xprt
);
2014 /* Reset to new socket */
2015 transport
->sock
= sock
;
2016 transport
->inet
= sk
;
2018 xs_set_memalloc(xprt
);
2020 write_unlock_bh(&sk
->sk_callback_lock
);
2022 xs_udp_do_set_buffer_size(xprt
);
2025 static void xs_udp_setup_socket(struct work_struct
*work
)
2027 struct sock_xprt
*transport
=
2028 container_of(work
, struct sock_xprt
, connect_worker
.work
);
2029 struct rpc_xprt
*xprt
= &transport
->xprt
;
2030 struct socket
*sock
= transport
->sock
;
2033 current
->flags
|= PF_FSTRANS
;
2035 /* Start by resetting any existing state */
2036 xs_reset_transport(transport
);
2037 sock
= xs_create_sock(xprt
, transport
,
2038 xs_addr(xprt
)->sa_family
, SOCK_DGRAM
, IPPROTO_UDP
);
2042 dprintk("RPC: worker connecting xprt %p via %s to "
2043 "%s (port %s)\n", xprt
,
2044 xprt
->address_strings
[RPC_DISPLAY_PROTO
],
2045 xprt
->address_strings
[RPC_DISPLAY_ADDR
],
2046 xprt
->address_strings
[RPC_DISPLAY_PORT
]);
2048 xs_udp_finish_connecting(xprt
, sock
);
2049 trace_rpc_socket_connect(xprt
, sock
, 0);
2052 xprt_clear_connecting(xprt
);
2053 xprt_wake_pending_tasks(xprt
, status
);
2054 current
->flags
&= ~PF_FSTRANS
;
2058 * We need to preserve the port number so the reply cache on the server can
2059 * find our cached RPC replies when we get around to reconnecting.
2061 static void xs_abort_connection(struct sock_xprt
*transport
)
2064 struct sockaddr any
;
2066 dprintk("RPC: disconnecting xprt %p to reuse port\n", transport
);
2069 * Disconnect the transport socket by doing a connect operation
2070 * with AF_UNSPEC. This should return immediately...
2072 memset(&any
, 0, sizeof(any
));
2073 any
.sa_family
= AF_UNSPEC
;
2074 result
= kernel_connect(transport
->sock
, &any
, sizeof(any
), 0);
2075 trace_rpc_socket_reset_connection(&transport
->xprt
,
2076 transport
->sock
, result
);
2078 xs_sock_reset_connection_flags(&transport
->xprt
);
2079 dprintk("RPC: AF_UNSPEC connect return code %d\n", result
);
2082 static void xs_tcp_reuse_connection(struct sock_xprt
*transport
)
2084 unsigned int state
= transport
->inet
->sk_state
;
2086 if (state
== TCP_CLOSE
&& transport
->sock
->state
== SS_UNCONNECTED
) {
2087 /* we don't need to abort the connection if the socket
2088 * hasn't undergone a shutdown
2090 if (transport
->inet
->sk_shutdown
== 0)
2092 dprintk("RPC: %s: TCP_CLOSEd and sk_shutdown set to %d\n",
2093 __func__
, transport
->inet
->sk_shutdown
);
2095 if ((1 << state
) & (TCPF_ESTABLISHED
|TCPF_SYN_SENT
)) {
2096 /* we don't need to abort the connection if the socket
2097 * hasn't undergone a shutdown
2099 if (transport
->inet
->sk_shutdown
== 0)
2101 dprintk("RPC: %s: ESTABLISHED/SYN_SENT "
2102 "sk_shutdown set to %d\n",
2103 __func__
, transport
->inet
->sk_shutdown
);
2105 xs_abort_connection(transport
);
2108 static int xs_tcp_finish_connecting(struct rpc_xprt
*xprt
, struct socket
*sock
)
2110 struct sock_xprt
*transport
= container_of(xprt
, struct sock_xprt
, xprt
);
2111 int ret
= -ENOTCONN
;
2113 if (!transport
->inet
) {
2114 struct sock
*sk
= sock
->sk
;
2116 write_lock_bh(&sk
->sk_callback_lock
);
2118 xs_save_old_callbacks(transport
, sk
);
2120 sk
->sk_user_data
= xprt
;
2121 sk
->sk_data_ready
= xs_tcp_data_ready
;
2122 sk
->sk_state_change
= xs_tcp_state_change
;
2123 sk
->sk_write_space
= xs_tcp_write_space
;
2124 sk
->sk_allocation
= GFP_ATOMIC
;
2126 /* socket options */
2127 sk
->sk_userlocks
|= SOCK_BINDPORT_LOCK
;
2128 sock_reset_flag(sk
, SOCK_LINGER
);
2129 tcp_sk(sk
)->linger2
= 0;
2130 tcp_sk(sk
)->nonagle
|= TCP_NAGLE_OFF
;
2132 xprt_clear_connected(xprt
);
2134 /* Reset to new socket */
2135 transport
->sock
= sock
;
2136 transport
->inet
= sk
;
2138 write_unlock_bh(&sk
->sk_callback_lock
);
2141 if (!xprt_bound(xprt
))
2144 xs_set_memalloc(xprt
);
2146 /* Tell the socket layer to start connecting... */
2147 xprt
->stat
.connect_count
++;
2148 xprt
->stat
.connect_start
= jiffies
;
2149 ret
= kernel_connect(sock
, xs_addr(xprt
), xprt
->addrlen
, O_NONBLOCK
);
2154 xprt
->connect_cookie
++;
2155 if (xprt
->reestablish_timeout
< XS_TCP_INIT_REEST_TO
)
2156 xprt
->reestablish_timeout
= XS_TCP_INIT_REEST_TO
;
2163 * xs_tcp_setup_socket - create a TCP socket and connect to a remote endpoint
2164 * @xprt: RPC transport to connect
2165 * @transport: socket transport to connect
2166 * @create_sock: function to create a socket of the correct type
2168 * Invoked by a work queue tasklet.
2170 static void xs_tcp_setup_socket(struct work_struct
*work
)
2172 struct sock_xprt
*transport
=
2173 container_of(work
, struct sock_xprt
, connect_worker
.work
);
2174 struct socket
*sock
= transport
->sock
;
2175 struct rpc_xprt
*xprt
= &transport
->xprt
;
2178 current
->flags
|= PF_FSTRANS
;
2181 clear_bit(XPRT_CONNECTION_ABORT
, &xprt
->state
);
2182 sock
= xs_create_sock(xprt
, transport
,
2183 xs_addr(xprt
)->sa_family
, SOCK_STREAM
, IPPROTO_TCP
);
2185 status
= PTR_ERR(sock
);
2191 abort_and_exit
= test_and_clear_bit(XPRT_CONNECTION_ABORT
,
2193 /* "close" the socket, preserving the local port */
2194 xs_tcp_reuse_connection(transport
);
2200 dprintk("RPC: worker connecting xprt %p via %s to "
2201 "%s (port %s)\n", xprt
,
2202 xprt
->address_strings
[RPC_DISPLAY_PROTO
],
2203 xprt
->address_strings
[RPC_DISPLAY_ADDR
],
2204 xprt
->address_strings
[RPC_DISPLAY_PORT
]);
2206 status
= xs_tcp_finish_connecting(xprt
, sock
);
2207 trace_rpc_socket_connect(xprt
, sock
, status
);
2208 dprintk("RPC: %p connect status %d connected %d sock state %d\n",
2209 xprt
, -status
, xprt_connected(xprt
),
2210 sock
->sk
->sk_state
);
2213 printk("%s: connect returned unhandled error %d\n",
2215 case -EADDRNOTAVAIL
:
2216 /* We're probably in TIME_WAIT. Get rid of existing socket,
2219 xs_tcp_force_close(xprt
);
2224 xprt_clear_connecting(xprt
);
2225 current
->flags
&= ~PF_FSTRANS
;
2228 /* Happens, for instance, if the user specified a link
2229 * local IPv6 address without a scope-id.
2234 /* retry with existing socket, after a delay */
2240 xprt_clear_connecting(xprt
);
2241 xprt_wake_pending_tasks(xprt
, status
);
2242 current
->flags
&= ~PF_FSTRANS
;
2246 * xs_connect - connect a socket to a remote endpoint
2247 * @xprt: pointer to transport structure
2248 * @task: address of RPC task that manages state of connect request
2250 * TCP: If the remote end dropped the connection, delay reconnecting.
2252 * UDP socket connects are synchronous, but we use a work queue anyway
2253 * to guarantee that even unprivileged user processes can set up a
2254 * socket on a privileged port.
2256 * If a UDP socket connect fails, the delay behavior here prevents
2257 * retry floods (hard mounts).
2259 static void xs_connect(struct rpc_xprt
*xprt
, struct rpc_task
*task
)
2261 struct sock_xprt
*transport
= container_of(xprt
, struct sock_xprt
, xprt
);
2263 if (transport
->sock
!= NULL
&& !RPC_IS_SOFTCONN(task
)) {
2264 dprintk("RPC: xs_connect delayed xprt %p for %lu "
2266 xprt
, xprt
->reestablish_timeout
/ HZ
);
2267 queue_delayed_work(rpciod_workqueue
,
2268 &transport
->connect_worker
,
2269 xprt
->reestablish_timeout
);
2270 xprt
->reestablish_timeout
<<= 1;
2271 if (xprt
->reestablish_timeout
< XS_TCP_INIT_REEST_TO
)
2272 xprt
->reestablish_timeout
= XS_TCP_INIT_REEST_TO
;
2273 if (xprt
->reestablish_timeout
> XS_TCP_MAX_REEST_TO
)
2274 xprt
->reestablish_timeout
= XS_TCP_MAX_REEST_TO
;
2276 dprintk("RPC: xs_connect scheduled xprt %p\n", xprt
);
2277 queue_delayed_work(rpciod_workqueue
,
2278 &transport
->connect_worker
, 0);
2283 * xs_local_print_stats - display AF_LOCAL socket-specifc stats
2284 * @xprt: rpc_xprt struct containing statistics
2288 static void xs_local_print_stats(struct rpc_xprt
*xprt
, struct seq_file
*seq
)
2292 if (xprt_connected(xprt
))
2293 idle_time
= (long)(jiffies
- xprt
->last_used
) / HZ
;
2295 seq_printf(seq
, "\txprt:\tlocal %lu %lu %lu %ld %lu %lu %lu "
2296 "%llu %llu %lu %llu %llu\n",
2297 xprt
->stat
.bind_count
,
2298 xprt
->stat
.connect_count
,
2299 xprt
->stat
.connect_time
,
2303 xprt
->stat
.bad_xids
,
2306 xprt
->stat
.max_slots
,
2307 xprt
->stat
.sending_u
,
2308 xprt
->stat
.pending_u
);
2312 * xs_udp_print_stats - display UDP socket-specifc stats
2313 * @xprt: rpc_xprt struct containing statistics
2317 static void xs_udp_print_stats(struct rpc_xprt
*xprt
, struct seq_file
*seq
)
2319 struct sock_xprt
*transport
= container_of(xprt
, struct sock_xprt
, xprt
);
2321 seq_printf(seq
, "\txprt:\tudp %u %lu %lu %lu %lu %llu %llu "
2324 xprt
->stat
.bind_count
,
2327 xprt
->stat
.bad_xids
,
2330 xprt
->stat
.max_slots
,
2331 xprt
->stat
.sending_u
,
2332 xprt
->stat
.pending_u
);
2336 * xs_tcp_print_stats - display TCP socket-specifc stats
2337 * @xprt: rpc_xprt struct containing statistics
2341 static void xs_tcp_print_stats(struct rpc_xprt
*xprt
, struct seq_file
*seq
)
2343 struct sock_xprt
*transport
= container_of(xprt
, struct sock_xprt
, xprt
);
2346 if (xprt_connected(xprt
))
2347 idle_time
= (long)(jiffies
- xprt
->last_used
) / HZ
;
2349 seq_printf(seq
, "\txprt:\ttcp %u %lu %lu %lu %ld %lu %lu %lu "
2350 "%llu %llu %lu %llu %llu\n",
2352 xprt
->stat
.bind_count
,
2353 xprt
->stat
.connect_count
,
2354 xprt
->stat
.connect_time
,
2358 xprt
->stat
.bad_xids
,
2361 xprt
->stat
.max_slots
,
2362 xprt
->stat
.sending_u
,
2363 xprt
->stat
.pending_u
);
2367 * Allocate a bunch of pages for a scratch buffer for the rpc code. The reason
2368 * we allocate pages instead doing a kmalloc like rpc_malloc is because we want
2369 * to use the server side send routines.
2371 static void *bc_malloc(struct rpc_task
*task
, size_t size
)
2374 struct rpc_buffer
*buf
;
2376 WARN_ON_ONCE(size
> PAGE_SIZE
- sizeof(struct rpc_buffer
));
2377 if (size
> PAGE_SIZE
- sizeof(struct rpc_buffer
))
2380 page
= alloc_page(GFP_KERNEL
);
2384 buf
= page_address(page
);
2385 buf
->len
= PAGE_SIZE
;
2391 * Free the space allocated in the bc_alloc routine
2393 static void bc_free(void *buffer
)
2395 struct rpc_buffer
*buf
;
2400 buf
= container_of(buffer
, struct rpc_buffer
, data
);
2401 free_page((unsigned long)buf
);
2405 * Use the svc_sock to send the callback. Must be called with svsk->sk_mutex
2406 * held. Borrows heavily from svc_tcp_sendto and xs_tcp_send_request.
2408 static int bc_sendto(struct rpc_rqst
*req
)
2411 struct xdr_buf
*xbufp
= &req
->rq_snd_buf
;
2412 struct rpc_xprt
*xprt
= req
->rq_xprt
;
2413 struct sock_xprt
*transport
=
2414 container_of(xprt
, struct sock_xprt
, xprt
);
2415 struct socket
*sock
= transport
->sock
;
2416 unsigned long headoff
;
2417 unsigned long tailoff
;
2419 xs_encode_stream_record_marker(xbufp
);
2421 tailoff
= (unsigned long)xbufp
->tail
[0].iov_base
& ~PAGE_MASK
;
2422 headoff
= (unsigned long)xbufp
->head
[0].iov_base
& ~PAGE_MASK
;
2423 len
= svc_send_common(sock
, xbufp
,
2424 virt_to_page(xbufp
->head
[0].iov_base
), headoff
,
2425 xbufp
->tail
[0].iov_base
, tailoff
);
2427 if (len
!= xbufp
->len
) {
2428 printk(KERN_NOTICE
"Error sending entire callback!\n");
2436 * The send routine. Borrows from svc_send
2438 static int bc_send_request(struct rpc_task
*task
)
2440 struct rpc_rqst
*req
= task
->tk_rqstp
;
2441 struct svc_xprt
*xprt
;
2444 dprintk("sending request with xid: %08x\n", ntohl(req
->rq_xid
));
2446 * Get the server socket associated with this callback xprt
2448 xprt
= req
->rq_xprt
->bc_xprt
;
2451 * Grab the mutex to serialize data as the connection is shared
2452 * with the fore channel
2454 if (!mutex_trylock(&xprt
->xpt_mutex
)) {
2455 rpc_sleep_on(&xprt
->xpt_bc_pending
, task
, NULL
);
2456 if (!mutex_trylock(&xprt
->xpt_mutex
))
2458 rpc_wake_up_queued_task(&xprt
->xpt_bc_pending
, task
);
2460 if (test_bit(XPT_DEAD
, &xprt
->xpt_flags
))
2463 len
= bc_sendto(req
);
2464 mutex_unlock(&xprt
->xpt_mutex
);
2473 * The close routine. Since this is client initiated, we do nothing
2476 static void bc_close(struct rpc_xprt
*xprt
)
2481 * The xprt destroy routine. Again, because this connection is client
2482 * initiated, we do nothing
2485 static void bc_destroy(struct rpc_xprt
*xprt
)
2489 static struct rpc_xprt_ops xs_local_ops
= {
2490 .reserve_xprt
= xprt_reserve_xprt
,
2491 .release_xprt
= xs_tcp_release_xprt
,
2492 .alloc_slot
= xprt_alloc_slot
,
2493 .rpcbind
= xs_local_rpcbind
,
2494 .set_port
= xs_local_set_port
,
2495 .connect
= xs_local_connect
,
2496 .buf_alloc
= rpc_malloc
,
2497 .buf_free
= rpc_free
,
2498 .send_request
= xs_local_send_request
,
2499 .set_retrans_timeout
= xprt_set_retrans_timeout_def
,
2501 .destroy
= xs_local_destroy
,
2502 .print_stats
= xs_local_print_stats
,
2505 static struct rpc_xprt_ops xs_udp_ops
= {
2506 .set_buffer_size
= xs_udp_set_buffer_size
,
2507 .reserve_xprt
= xprt_reserve_xprt_cong
,
2508 .release_xprt
= xprt_release_xprt_cong
,
2509 .alloc_slot
= xprt_alloc_slot
,
2510 .rpcbind
= rpcb_getport_async
,
2511 .set_port
= xs_set_port
,
2512 .connect
= xs_connect
,
2513 .buf_alloc
= rpc_malloc
,
2514 .buf_free
= rpc_free
,
2515 .send_request
= xs_udp_send_request
,
2516 .set_retrans_timeout
= xprt_set_retrans_timeout_rtt
,
2517 .timer
= xs_udp_timer
,
2518 .release_request
= xprt_release_rqst_cong
,
2520 .destroy
= xs_destroy
,
2521 .print_stats
= xs_udp_print_stats
,
2524 static struct rpc_xprt_ops xs_tcp_ops
= {
2525 .reserve_xprt
= xprt_reserve_xprt
,
2526 .release_xprt
= xs_tcp_release_xprt
,
2527 .alloc_slot
= xprt_lock_and_alloc_slot
,
2528 .rpcbind
= rpcb_getport_async
,
2529 .set_port
= xs_set_port
,
2530 .connect
= xs_connect
,
2531 .buf_alloc
= rpc_malloc
,
2532 .buf_free
= rpc_free
,
2533 .send_request
= xs_tcp_send_request
,
2534 .set_retrans_timeout
= xprt_set_retrans_timeout_def
,
2535 .close
= xs_tcp_close
,
2536 .destroy
= xs_destroy
,
2537 .print_stats
= xs_tcp_print_stats
,
2541 * The rpc_xprt_ops for the server backchannel
2544 static struct rpc_xprt_ops bc_tcp_ops
= {
2545 .reserve_xprt
= xprt_reserve_xprt
,
2546 .release_xprt
= xprt_release_xprt
,
2547 .alloc_slot
= xprt_alloc_slot
,
2548 .buf_alloc
= bc_malloc
,
2549 .buf_free
= bc_free
,
2550 .send_request
= bc_send_request
,
2551 .set_retrans_timeout
= xprt_set_retrans_timeout_def
,
2553 .destroy
= bc_destroy
,
2554 .print_stats
= xs_tcp_print_stats
,
2557 static int xs_init_anyaddr(const int family
, struct sockaddr
*sap
)
2559 static const struct sockaddr_in sin
= {
2560 .sin_family
= AF_INET
,
2561 .sin_addr
.s_addr
= htonl(INADDR_ANY
),
2563 static const struct sockaddr_in6 sin6
= {
2564 .sin6_family
= AF_INET6
,
2565 .sin6_addr
= IN6ADDR_ANY_INIT
,
2572 memcpy(sap
, &sin
, sizeof(sin
));
2575 memcpy(sap
, &sin6
, sizeof(sin6
));
2578 dprintk("RPC: %s: Bad address family\n", __func__
);
2579 return -EAFNOSUPPORT
;
2584 static struct rpc_xprt
*xs_setup_xprt(struct xprt_create
*args
,
2585 unsigned int slot_table_size
,
2586 unsigned int max_slot_table_size
)
2588 struct rpc_xprt
*xprt
;
2589 struct sock_xprt
*new;
2591 if (args
->addrlen
> sizeof(xprt
->addr
)) {
2592 dprintk("RPC: xs_setup_xprt: address too large\n");
2593 return ERR_PTR(-EBADF
);
2596 xprt
= xprt_alloc(args
->net
, sizeof(*new), slot_table_size
,
2597 max_slot_table_size
);
2599 dprintk("RPC: xs_setup_xprt: couldn't allocate "
2601 return ERR_PTR(-ENOMEM
);
2604 new = container_of(xprt
, struct sock_xprt
, xprt
);
2605 memcpy(&xprt
->addr
, args
->dstaddr
, args
->addrlen
);
2606 xprt
->addrlen
= args
->addrlen
;
2608 memcpy(&new->srcaddr
, args
->srcaddr
, args
->addrlen
);
2611 err
= xs_init_anyaddr(args
->dstaddr
->sa_family
,
2612 (struct sockaddr
*)&new->srcaddr
);
2615 return ERR_PTR(err
);
2622 static const struct rpc_timeout xs_local_default_timeout
= {
2623 .to_initval
= 10 * HZ
,
2624 .to_maxval
= 10 * HZ
,
2629 * xs_setup_local - Set up transport to use an AF_LOCAL socket
2630 * @args: rpc transport creation arguments
2632 * AF_LOCAL is a "tpi_cots_ord" transport, just like TCP
2634 static struct rpc_xprt
*xs_setup_local(struct xprt_create
*args
)
2636 struct sockaddr_un
*sun
= (struct sockaddr_un
*)args
->dstaddr
;
2637 struct sock_xprt
*transport
;
2638 struct rpc_xprt
*xprt
;
2639 struct rpc_xprt
*ret
;
2641 xprt
= xs_setup_xprt(args
, xprt_tcp_slot_table_entries
,
2642 xprt_max_tcp_slot_table_entries
);
2645 transport
= container_of(xprt
, struct sock_xprt
, xprt
);
2648 xprt
->tsh_size
= sizeof(rpc_fraghdr
) / sizeof(u32
);
2649 xprt
->max_payload
= RPC_MAX_FRAGMENT_SIZE
;
2651 xprt
->bind_timeout
= XS_BIND_TO
;
2652 xprt
->reestablish_timeout
= XS_TCP_INIT_REEST_TO
;
2653 xprt
->idle_timeout
= XS_IDLE_DISC_TO
;
2655 xprt
->ops
= &xs_local_ops
;
2656 xprt
->timeout
= &xs_local_default_timeout
;
2658 switch (sun
->sun_family
) {
2660 if (sun
->sun_path
[0] != '/') {
2661 dprintk("RPC: bad AF_LOCAL address: %s\n",
2663 ret
= ERR_PTR(-EINVAL
);
2666 xprt_set_bound(xprt
);
2667 xs_format_peer_addresses(xprt
, "local", RPCBIND_NETID_LOCAL
);
2668 ret
= ERR_PTR(xs_local_setup_socket(transport
));
2673 ret
= ERR_PTR(-EAFNOSUPPORT
);
2677 dprintk("RPC: set up xprt to %s via AF_LOCAL\n",
2678 xprt
->address_strings
[RPC_DISPLAY_ADDR
]);
2680 if (try_module_get(THIS_MODULE
))
2682 ret
= ERR_PTR(-EINVAL
);
2688 static const struct rpc_timeout xs_udp_default_timeout
= {
2689 .to_initval
= 5 * HZ
,
2690 .to_maxval
= 30 * HZ
,
2691 .to_increment
= 5 * HZ
,
2696 * xs_setup_udp - Set up transport to use a UDP socket
2697 * @args: rpc transport creation arguments
2700 static struct rpc_xprt
*xs_setup_udp(struct xprt_create
*args
)
2702 struct sockaddr
*addr
= args
->dstaddr
;
2703 struct rpc_xprt
*xprt
;
2704 struct sock_xprt
*transport
;
2705 struct rpc_xprt
*ret
;
2707 xprt
= xs_setup_xprt(args
, xprt_udp_slot_table_entries
,
2708 xprt_udp_slot_table_entries
);
2711 transport
= container_of(xprt
, struct sock_xprt
, xprt
);
2713 xprt
->prot
= IPPROTO_UDP
;
2715 /* XXX: header size can vary due to auth type, IPv6, etc. */
2716 xprt
->max_payload
= (1U << 16) - (MAX_HEADER
<< 3);
2718 xprt
->bind_timeout
= XS_BIND_TO
;
2719 xprt
->reestablish_timeout
= XS_UDP_REEST_TO
;
2720 xprt
->idle_timeout
= XS_IDLE_DISC_TO
;
2722 xprt
->ops
= &xs_udp_ops
;
2724 xprt
->timeout
= &xs_udp_default_timeout
;
2726 switch (addr
->sa_family
) {
2728 if (((struct sockaddr_in
*)addr
)->sin_port
!= htons(0))
2729 xprt_set_bound(xprt
);
2731 INIT_DELAYED_WORK(&transport
->connect_worker
,
2732 xs_udp_setup_socket
);
2733 xs_format_peer_addresses(xprt
, "udp", RPCBIND_NETID_UDP
);
2736 if (((struct sockaddr_in6
*)addr
)->sin6_port
!= htons(0))
2737 xprt_set_bound(xprt
);
2739 INIT_DELAYED_WORK(&transport
->connect_worker
,
2740 xs_udp_setup_socket
);
2741 xs_format_peer_addresses(xprt
, "udp", RPCBIND_NETID_UDP6
);
2744 ret
= ERR_PTR(-EAFNOSUPPORT
);
2748 if (xprt_bound(xprt
))
2749 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
2750 xprt
->address_strings
[RPC_DISPLAY_ADDR
],
2751 xprt
->address_strings
[RPC_DISPLAY_PORT
],
2752 xprt
->address_strings
[RPC_DISPLAY_PROTO
]);
2754 dprintk("RPC: set up xprt to %s (autobind) via %s\n",
2755 xprt
->address_strings
[RPC_DISPLAY_ADDR
],
2756 xprt
->address_strings
[RPC_DISPLAY_PROTO
]);
2758 if (try_module_get(THIS_MODULE
))
2760 ret
= ERR_PTR(-EINVAL
);
2766 static const struct rpc_timeout xs_tcp_default_timeout
= {
2767 .to_initval
= 60 * HZ
,
2768 .to_maxval
= 60 * HZ
,
2773 * xs_setup_tcp - Set up transport to use a TCP socket
2774 * @args: rpc transport creation arguments
2777 static struct rpc_xprt
*xs_setup_tcp(struct xprt_create
*args
)
2779 struct sockaddr
*addr
= args
->dstaddr
;
2780 struct rpc_xprt
*xprt
;
2781 struct sock_xprt
*transport
;
2782 struct rpc_xprt
*ret
;
2783 unsigned int max_slot_table_size
= xprt_max_tcp_slot_table_entries
;
2785 if (args
->flags
& XPRT_CREATE_INFINITE_SLOTS
)
2786 max_slot_table_size
= RPC_MAX_SLOT_TABLE_LIMIT
;
2788 xprt
= xs_setup_xprt(args
, xprt_tcp_slot_table_entries
,
2789 max_slot_table_size
);
2792 transport
= container_of(xprt
, struct sock_xprt
, xprt
);
2794 xprt
->prot
= IPPROTO_TCP
;
2795 xprt
->tsh_size
= sizeof(rpc_fraghdr
) / sizeof(u32
);
2796 xprt
->max_payload
= RPC_MAX_FRAGMENT_SIZE
;
2798 xprt
->bind_timeout
= XS_BIND_TO
;
2799 xprt
->reestablish_timeout
= XS_TCP_INIT_REEST_TO
;
2800 xprt
->idle_timeout
= XS_IDLE_DISC_TO
;
2802 xprt
->ops
= &xs_tcp_ops
;
2803 xprt
->timeout
= &xs_tcp_default_timeout
;
2805 switch (addr
->sa_family
) {
2807 if (((struct sockaddr_in
*)addr
)->sin_port
!= htons(0))
2808 xprt_set_bound(xprt
);
2810 INIT_DELAYED_WORK(&transport
->connect_worker
,
2811 xs_tcp_setup_socket
);
2812 xs_format_peer_addresses(xprt
, "tcp", RPCBIND_NETID_TCP
);
2815 if (((struct sockaddr_in6
*)addr
)->sin6_port
!= htons(0))
2816 xprt_set_bound(xprt
);
2818 INIT_DELAYED_WORK(&transport
->connect_worker
,
2819 xs_tcp_setup_socket
);
2820 xs_format_peer_addresses(xprt
, "tcp", RPCBIND_NETID_TCP6
);
2823 ret
= ERR_PTR(-EAFNOSUPPORT
);
2827 if (xprt_bound(xprt
))
2828 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
2829 xprt
->address_strings
[RPC_DISPLAY_ADDR
],
2830 xprt
->address_strings
[RPC_DISPLAY_PORT
],
2831 xprt
->address_strings
[RPC_DISPLAY_PROTO
]);
2833 dprintk("RPC: set up xprt to %s (autobind) via %s\n",
2834 xprt
->address_strings
[RPC_DISPLAY_ADDR
],
2835 xprt
->address_strings
[RPC_DISPLAY_PROTO
]);
2838 if (try_module_get(THIS_MODULE
))
2840 ret
= ERR_PTR(-EINVAL
);
2847 * xs_setup_bc_tcp - Set up transport to use a TCP backchannel socket
2848 * @args: rpc transport creation arguments
2851 static struct rpc_xprt
*xs_setup_bc_tcp(struct xprt_create
*args
)
2853 struct sockaddr
*addr
= args
->dstaddr
;
2854 struct rpc_xprt
*xprt
;
2855 struct sock_xprt
*transport
;
2856 struct svc_sock
*bc_sock
;
2857 struct rpc_xprt
*ret
;
2859 if (args
->bc_xprt
->xpt_bc_xprt
) {
2861 * This server connection already has a backchannel
2862 * export; we can't create a new one, as we wouldn't be
2863 * able to match replies based on xid any more. So,
2864 * reuse the already-existing one:
2866 return args
->bc_xprt
->xpt_bc_xprt
;
2868 xprt
= xs_setup_xprt(args
, xprt_tcp_slot_table_entries
,
2869 xprt_tcp_slot_table_entries
);
2872 transport
= container_of(xprt
, struct sock_xprt
, xprt
);
2874 xprt
->prot
= IPPROTO_TCP
;
2875 xprt
->tsh_size
= sizeof(rpc_fraghdr
) / sizeof(u32
);
2876 xprt
->max_payload
= RPC_MAX_FRAGMENT_SIZE
;
2877 xprt
->timeout
= &xs_tcp_default_timeout
;
2880 xprt_set_bound(xprt
);
2881 xprt
->bind_timeout
= 0;
2882 xprt
->reestablish_timeout
= 0;
2883 xprt
->idle_timeout
= 0;
2885 xprt
->ops
= &bc_tcp_ops
;
2887 switch (addr
->sa_family
) {
2889 xs_format_peer_addresses(xprt
, "tcp",
2893 xs_format_peer_addresses(xprt
, "tcp",
2894 RPCBIND_NETID_TCP6
);
2897 ret
= ERR_PTR(-EAFNOSUPPORT
);
2901 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
2902 xprt
->address_strings
[RPC_DISPLAY_ADDR
],
2903 xprt
->address_strings
[RPC_DISPLAY_PORT
],
2904 xprt
->address_strings
[RPC_DISPLAY_PROTO
]);
2907 * Once we've associated a backchannel xprt with a connection,
2908 * we want to keep it around as long as long as the connection
2909 * lasts, in case we need to start using it for a backchannel
2910 * again; this reference won't be dropped until bc_xprt is
2914 args
->bc_xprt
->xpt_bc_xprt
= xprt
;
2915 xprt
->bc_xprt
= args
->bc_xprt
;
2916 bc_sock
= container_of(args
->bc_xprt
, struct svc_sock
, sk_xprt
);
2917 transport
->sock
= bc_sock
->sk_sock
;
2918 transport
->inet
= bc_sock
->sk_sk
;
2921 * Since we don't want connections for the backchannel, we set
2922 * the xprt status to connected
2924 xprt_set_connected(xprt
);
2927 if (try_module_get(THIS_MODULE
))
2930 ret
= ERR_PTR(-EINVAL
);
2936 static struct xprt_class xs_local_transport
= {
2937 .list
= LIST_HEAD_INIT(xs_local_transport
.list
),
2938 .name
= "named UNIX socket",
2939 .owner
= THIS_MODULE
,
2940 .ident
= XPRT_TRANSPORT_LOCAL
,
2941 .setup
= xs_setup_local
,
2944 static struct xprt_class xs_udp_transport
= {
2945 .list
= LIST_HEAD_INIT(xs_udp_transport
.list
),
2947 .owner
= THIS_MODULE
,
2948 .ident
= XPRT_TRANSPORT_UDP
,
2949 .setup
= xs_setup_udp
,
2952 static struct xprt_class xs_tcp_transport
= {
2953 .list
= LIST_HEAD_INIT(xs_tcp_transport
.list
),
2955 .owner
= THIS_MODULE
,
2956 .ident
= XPRT_TRANSPORT_TCP
,
2957 .setup
= xs_setup_tcp
,
2960 static struct xprt_class xs_bc_tcp_transport
= {
2961 .list
= LIST_HEAD_INIT(xs_bc_tcp_transport
.list
),
2962 .name
= "tcp NFSv4.1 backchannel",
2963 .owner
= THIS_MODULE
,
2964 .ident
= XPRT_TRANSPORT_BC_TCP
,
2965 .setup
= xs_setup_bc_tcp
,
2969 * init_socket_xprt - set up xprtsock's sysctls, register with RPC client
2972 int init_socket_xprt(void)
2975 if (!sunrpc_table_header
)
2976 sunrpc_table_header
= register_sysctl_table(sunrpc_table
);
2979 xprt_register_transport(&xs_local_transport
);
2980 xprt_register_transport(&xs_udp_transport
);
2981 xprt_register_transport(&xs_tcp_transport
);
2982 xprt_register_transport(&xs_bc_tcp_transport
);
2988 * cleanup_socket_xprt - remove xprtsock's sysctls, unregister
2991 void cleanup_socket_xprt(void)
2994 if (sunrpc_table_header
) {
2995 unregister_sysctl_table(sunrpc_table_header
);
2996 sunrpc_table_header
= NULL
;
3000 xprt_unregister_transport(&xs_local_transport
);
3001 xprt_unregister_transport(&xs_udp_transport
);
3002 xprt_unregister_transport(&xs_tcp_transport
);
3003 xprt_unregister_transport(&xs_bc_tcp_transport
);
3006 static int param_set_uint_minmax(const char *val
,
3007 const struct kernel_param
*kp
,
3008 unsigned int min
, unsigned int max
)
3015 ret
= strict_strtoul(val
, 0, &num
);
3016 if (ret
== -EINVAL
|| num
< min
|| num
> max
)
3018 *((unsigned int *)kp
->arg
) = num
;
3022 static int param_set_portnr(const char *val
, const struct kernel_param
*kp
)
3024 return param_set_uint_minmax(val
, kp
,
3029 static struct kernel_param_ops param_ops_portnr
= {
3030 .set
= param_set_portnr
,
3031 .get
= param_get_uint
,
3034 #define param_check_portnr(name, p) \
3035 __param_check(name, p, unsigned int);
3037 module_param_named(min_resvport
, xprt_min_resvport
, portnr
, 0644);
3038 module_param_named(max_resvport
, xprt_max_resvport
, portnr
, 0644);
3040 static int param_set_slot_table_size(const char *val
,
3041 const struct kernel_param
*kp
)
3043 return param_set_uint_minmax(val
, kp
,
3045 RPC_MAX_SLOT_TABLE
);
3048 static struct kernel_param_ops param_ops_slot_table_size
= {
3049 .set
= param_set_slot_table_size
,
3050 .get
= param_get_uint
,
3053 #define param_check_slot_table_size(name, p) \
3054 __param_check(name, p, unsigned int);
3056 static int param_set_max_slot_table_size(const char *val
,
3057 const struct kernel_param
*kp
)
3059 return param_set_uint_minmax(val
, kp
,
3061 RPC_MAX_SLOT_TABLE_LIMIT
);
3064 static struct kernel_param_ops param_ops_max_slot_table_size
= {
3065 .set
= param_set_max_slot_table_size
,
3066 .get
= param_get_uint
,
3069 #define param_check_max_slot_table_size(name, p) \
3070 __param_check(name, p, unsigned int);
3072 module_param_named(tcp_slot_table_entries
, xprt_tcp_slot_table_entries
,
3073 slot_table_size
, 0644);
3074 module_param_named(tcp_max_slot_table_entries
, xprt_max_tcp_slot_table_entries
,
3075 max_slot_table_size
, 0644);
3076 module_param_named(udp_slot_table_entries
, xprt_udp_slot_table_entries
,
3077 slot_table_size
, 0644);