David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 1 | // SPDX-License-Identifier: GPL-2.0-or-later |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 2 | /* RxRPC recvmsg() implementation |
| 3 | * |
| 4 | * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved. |
| 5 | * Written by David Howells (dhowells@redhat.com) |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 6 | */ |
| 7 | |
| 8 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
| 9 | |
| 10 | #include <linux/net.h> |
| 11 | #include <linux/skbuff.h> |
| 12 | #include <linux/export.h> |
| 13 | #include <linux/sched/signal.h> |
| 14 | |
| 15 | #include <net/sock.h> |
| 16 | #include <net/af_rxrpc.h> |
| 17 | #include "ar-internal.h" |
| 18 | |
| 19 | /* |
| 20 | * Post a call for attention by the socket or kernel service. Further |
| 21 | * notifications are suppressed by putting recvmsg_link on a dummy queue. |
| 22 | */ |
| 23 | void rxrpc_notify_socket(struct rxrpc_call *call) |
| 24 | { |
| 25 | struct rxrpc_sock *rx; |
| 26 | struct sock *sk; |
| 27 | |
| 28 | _enter("%d", call->debug_id); |
| 29 | |
| 30 | if (!list_empty(&call->recvmsg_link)) |
| 31 | return; |
| 32 | |
| 33 | rcu_read_lock(); |
| 34 | |
| 35 | rx = rcu_dereference(call->socket); |
| 36 | sk = &rx->sk; |
| 37 | if (rx && sk->sk_state < RXRPC_CLOSE) { |
| 38 | if (call->notify_rx) { |
| 39 | spin_lock_bh(&call->notify_lock); |
| 40 | call->notify_rx(sk, call, call->user_call_ID); |
| 41 | spin_unlock_bh(&call->notify_lock); |
| 42 | } else { |
| 43 | write_lock_bh(&rx->recvmsg_lock); |
| 44 | if (list_empty(&call->recvmsg_link)) { |
| 45 | rxrpc_get_call(call, rxrpc_call_got); |
| 46 | list_add_tail(&call->recvmsg_link, &rx->recvmsg_q); |
| 47 | } |
| 48 | write_unlock_bh(&rx->recvmsg_lock); |
| 49 | |
| 50 | if (!sock_flag(sk, SOCK_DEAD)) { |
| 51 | _debug("call %ps", sk->sk_data_ready); |
| 52 | sk->sk_data_ready(sk); |
| 53 | } |
| 54 | } |
| 55 | } |
| 56 | |
| 57 | rcu_read_unlock(); |
| 58 | _leave(""); |
| 59 | } |
| 60 | |
| 61 | /* |
| 62 | * Pass a call terminating message to userspace. |
| 63 | */ |
| 64 | static int rxrpc_recvmsg_term(struct rxrpc_call *call, struct msghdr *msg) |
| 65 | { |
| 66 | u32 tmp = 0; |
| 67 | int ret; |
| 68 | |
| 69 | switch (call->completion) { |
| 70 | case RXRPC_CALL_SUCCEEDED: |
| 71 | ret = 0; |
| 72 | if (rxrpc_is_service_call(call)) |
| 73 | ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ACK, 0, &tmp); |
| 74 | break; |
| 75 | case RXRPC_CALL_REMOTELY_ABORTED: |
| 76 | tmp = call->abort_code; |
| 77 | ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp); |
| 78 | break; |
| 79 | case RXRPC_CALL_LOCALLY_ABORTED: |
| 80 | tmp = call->abort_code; |
| 81 | ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp); |
| 82 | break; |
| 83 | case RXRPC_CALL_NETWORK_ERROR: |
| 84 | tmp = -call->error; |
| 85 | ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NET_ERROR, 4, &tmp); |
| 86 | break; |
| 87 | case RXRPC_CALL_LOCAL_ERROR: |
| 88 | tmp = -call->error; |
| 89 | ret = put_cmsg(msg, SOL_RXRPC, RXRPC_LOCAL_ERROR, 4, &tmp); |
| 90 | break; |
| 91 | default: |
| 92 | pr_err("Invalid terminal call state %u\n", call->state); |
| 93 | BUG(); |
| 94 | break; |
| 95 | } |
| 96 | |
| 97 | trace_rxrpc_recvmsg(call, rxrpc_recvmsg_terminal, call->rx_hard_ack, |
| 98 | call->rx_pkt_offset, call->rx_pkt_len, ret); |
| 99 | return ret; |
| 100 | } |
| 101 | |
| 102 | /* |
| 103 | * Pass back notification of a new call. The call is added to the |
| 104 | * to-be-accepted list. This means that the next call to be accepted might not |
| 105 | * be the last call seen awaiting acceptance, but unless we leave this on the |
| 106 | * front of the queue and block all other messages until someone gives us a |
| 107 | * user_ID for it, there's not a lot we can do. |
| 108 | */ |
| 109 | static int rxrpc_recvmsg_new_call(struct rxrpc_sock *rx, |
| 110 | struct rxrpc_call *call, |
| 111 | struct msghdr *msg, int flags) |
| 112 | { |
| 113 | int tmp = 0, ret; |
| 114 | |
| 115 | ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NEW_CALL, 0, &tmp); |
| 116 | |
| 117 | if (ret == 0 && !(flags & MSG_PEEK)) { |
| 118 | _debug("to be accepted"); |
| 119 | write_lock_bh(&rx->recvmsg_lock); |
| 120 | list_del_init(&call->recvmsg_link); |
| 121 | write_unlock_bh(&rx->recvmsg_lock); |
| 122 | |
| 123 | rxrpc_get_call(call, rxrpc_call_got); |
| 124 | write_lock(&rx->call_lock); |
| 125 | list_add_tail(&call->accept_link, &rx->to_be_accepted); |
| 126 | write_unlock(&rx->call_lock); |
| 127 | } |
| 128 | |
| 129 | trace_rxrpc_recvmsg(call, rxrpc_recvmsg_to_be_accepted, 1, 0, 0, ret); |
| 130 | return ret; |
| 131 | } |
| 132 | |
| 133 | /* |
| 134 | * End the packet reception phase. |
| 135 | */ |
| 136 | static void rxrpc_end_rx_phase(struct rxrpc_call *call, rxrpc_serial_t serial) |
| 137 | { |
| 138 | _enter("%d,%s", call->debug_id, rxrpc_call_states[call->state]); |
| 139 | |
| 140 | trace_rxrpc_receive(call, rxrpc_receive_end, 0, call->rx_top); |
| 141 | ASSERTCMP(call->rx_hard_ack, ==, call->rx_top); |
| 142 | |
| 143 | if (call->state == RXRPC_CALL_CLIENT_RECV_REPLY) { |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 144 | rxrpc_propose_ACK(call, RXRPC_ACK_IDLE, serial, false, true, |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 145 | rxrpc_propose_ack_terminal_ack); |
| 146 | //rxrpc_send_ack_packet(call, false, NULL); |
| 147 | } |
| 148 | |
| 149 | write_lock_bh(&call->state_lock); |
| 150 | |
| 151 | switch (call->state) { |
| 152 | case RXRPC_CALL_CLIENT_RECV_REPLY: |
| 153 | __rxrpc_call_completed(call); |
| 154 | write_unlock_bh(&call->state_lock); |
| 155 | break; |
| 156 | |
| 157 | case RXRPC_CALL_SERVER_RECV_REQUEST: |
| 158 | call->tx_phase = true; |
| 159 | call->state = RXRPC_CALL_SERVER_ACK_REQUEST; |
| 160 | call->expect_req_by = jiffies + MAX_JIFFY_OFFSET; |
| 161 | write_unlock_bh(&call->state_lock); |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 162 | rxrpc_propose_ACK(call, RXRPC_ACK_DELAY, serial, false, true, |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 163 | rxrpc_propose_ack_processing_op); |
| 164 | break; |
| 165 | default: |
| 166 | write_unlock_bh(&call->state_lock); |
| 167 | break; |
| 168 | } |
| 169 | } |
| 170 | |
| 171 | /* |
| 172 | * Discard a packet we've used up and advance the Rx window by one. |
| 173 | */ |
| 174 | static void rxrpc_rotate_rx_window(struct rxrpc_call *call) |
| 175 | { |
| 176 | struct rxrpc_skb_priv *sp; |
| 177 | struct sk_buff *skb; |
| 178 | rxrpc_serial_t serial; |
| 179 | rxrpc_seq_t hard_ack, top; |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 180 | bool last = false; |
| 181 | u8 subpacket; |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 182 | int ix; |
| 183 | |
| 184 | _enter("%d", call->debug_id); |
| 185 | |
| 186 | hard_ack = call->rx_hard_ack; |
| 187 | top = smp_load_acquire(&call->rx_top); |
| 188 | ASSERT(before(hard_ack, top)); |
| 189 | |
| 190 | hard_ack++; |
| 191 | ix = hard_ack & RXRPC_RXTX_BUFF_MASK; |
| 192 | skb = call->rxtx_buffer[ix]; |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 193 | rxrpc_see_skb(skb, rxrpc_skb_rotated); |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 194 | sp = rxrpc_skb(skb); |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 195 | |
| 196 | subpacket = call->rxtx_annotations[ix] & RXRPC_RX_ANNO_SUBPACKET; |
| 197 | serial = sp->hdr.serial + subpacket; |
| 198 | |
| 199 | if (subpacket == sp->nr_subpackets - 1 && |
| 200 | sp->rx_flags & RXRPC_SKB_INCL_LAST) |
| 201 | last = true; |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 202 | |
| 203 | call->rxtx_buffer[ix] = NULL; |
| 204 | call->rxtx_annotations[ix] = 0; |
| 205 | /* Barrier against rxrpc_input_data(). */ |
| 206 | smp_store_release(&call->rx_hard_ack, hard_ack); |
| 207 | |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 208 | rxrpc_free_skb(skb, rxrpc_skb_freed); |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 209 | |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 210 | trace_rxrpc_receive(call, rxrpc_receive_rotate, serial, hard_ack); |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 211 | if (last) { |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 212 | rxrpc_end_rx_phase(call, serial); |
| 213 | } else { |
| 214 | /* Check to see if there's an ACK that needs sending. */ |
| 215 | if (after_eq(hard_ack, call->ackr_consumed + 2) || |
| 216 | after_eq(top, call->ackr_seen + 2) || |
| 217 | (hard_ack == top && after(hard_ack, call->ackr_consumed))) |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 218 | rxrpc_propose_ACK(call, RXRPC_ACK_DELAY, serial, |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 219 | true, true, |
| 220 | rxrpc_propose_ack_rotate_rx); |
| 221 | if (call->ackr_reason && call->ackr_reason != RXRPC_ACK_DELAY) |
| 222 | rxrpc_send_ack_packet(call, false, NULL); |
| 223 | } |
| 224 | } |
| 225 | |
| 226 | /* |
| 227 | * Decrypt and verify a (sub)packet. The packet's length may be changed due to |
| 228 | * padding, but if this is the case, the packet length will be resident in the |
| 229 | * socket buffer. Note that we can't modify the master skb info as the skb may |
| 230 | * be the home to multiple subpackets. |
| 231 | */ |
| 232 | static int rxrpc_verify_packet(struct rxrpc_call *call, struct sk_buff *skb, |
| 233 | u8 annotation, |
| 234 | unsigned int offset, unsigned int len) |
| 235 | { |
| 236 | struct rxrpc_skb_priv *sp = rxrpc_skb(skb); |
| 237 | rxrpc_seq_t seq = sp->hdr.seq; |
| 238 | u16 cksum = sp->hdr.cksum; |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 239 | u8 subpacket = annotation & RXRPC_RX_ANNO_SUBPACKET; |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 240 | |
| 241 | _enter(""); |
| 242 | |
| 243 | /* For all but the head jumbo subpacket, the security checksum is in a |
| 244 | * jumbo header immediately prior to the data. |
| 245 | */ |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 246 | if (subpacket > 0) { |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 247 | __be16 tmp; |
| 248 | if (skb_copy_bits(skb, offset - 2, &tmp, 2) < 0) |
| 249 | BUG(); |
| 250 | cksum = ntohs(tmp); |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 251 | seq += subpacket; |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 252 | } |
| 253 | |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 254 | return call->security->verify_packet(call, skb, offset, len, |
| 255 | seq, cksum); |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 256 | } |
| 257 | |
| 258 | /* |
| 259 | * Locate the data within a packet. This is complicated by: |
| 260 | * |
| 261 | * (1) An skb may contain a jumbo packet - so we have to find the appropriate |
| 262 | * subpacket. |
| 263 | * |
| 264 | * (2) The (sub)packets may be encrypted and, if so, the encrypted portion |
| 265 | * contains an extra header which includes the true length of the data, |
| 266 | * excluding any encrypted padding. |
| 267 | */ |
| 268 | static int rxrpc_locate_data(struct rxrpc_call *call, struct sk_buff *skb, |
| 269 | u8 *_annotation, |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 270 | unsigned int *_offset, unsigned int *_len, |
| 271 | bool *_last) |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 272 | { |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 273 | struct rxrpc_skb_priv *sp = rxrpc_skb(skb); |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 274 | unsigned int offset = sizeof(struct rxrpc_wire_header); |
| 275 | unsigned int len; |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 276 | bool last = false; |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 277 | int ret; |
| 278 | u8 annotation = *_annotation; |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 279 | u8 subpacket = annotation & RXRPC_RX_ANNO_SUBPACKET; |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 280 | |
| 281 | /* Locate the subpacket */ |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 282 | offset += subpacket * RXRPC_JUMBO_SUBPKTLEN; |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 283 | len = skb->len - offset; |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 284 | if (subpacket < sp->nr_subpackets - 1) |
| 285 | len = RXRPC_JUMBO_DATALEN; |
| 286 | else if (sp->rx_flags & RXRPC_SKB_INCL_LAST) |
| 287 | last = true; |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 288 | |
| 289 | if (!(annotation & RXRPC_RX_ANNO_VERIFIED)) { |
| 290 | ret = rxrpc_verify_packet(call, skb, annotation, offset, len); |
| 291 | if (ret < 0) |
| 292 | return ret; |
| 293 | *_annotation |= RXRPC_RX_ANNO_VERIFIED; |
| 294 | } |
| 295 | |
| 296 | *_offset = offset; |
| 297 | *_len = len; |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 298 | *_last = last; |
| 299 | call->security->locate_data(call, skb, _offset, _len); |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 300 | return 0; |
| 301 | } |
| 302 | |
| 303 | /* |
| 304 | * Deliver messages to a call. This keeps processing packets until the buffer |
| 305 | * is filled and we find either more DATA (returns 0) or the end of the DATA |
| 306 | * (returns 1). If more packets are required, it returns -EAGAIN. |
| 307 | */ |
| 308 | static int rxrpc_recvmsg_data(struct socket *sock, struct rxrpc_call *call, |
| 309 | struct msghdr *msg, struct iov_iter *iter, |
| 310 | size_t len, int flags, size_t *_offset) |
| 311 | { |
| 312 | struct rxrpc_skb_priv *sp; |
| 313 | struct sk_buff *skb; |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 314 | rxrpc_serial_t serial; |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 315 | rxrpc_seq_t hard_ack, top, seq; |
| 316 | size_t remain; |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 317 | bool rx_pkt_last; |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 318 | unsigned int rx_pkt_offset, rx_pkt_len; |
| 319 | int ix, copy, ret = -EAGAIN, ret2; |
| 320 | |
| 321 | if (test_and_clear_bit(RXRPC_CALL_RX_UNDERRUN, &call->flags) && |
| 322 | call->ackr_reason) |
| 323 | rxrpc_send_ack_packet(call, false, NULL); |
| 324 | |
| 325 | rx_pkt_offset = call->rx_pkt_offset; |
| 326 | rx_pkt_len = call->rx_pkt_len; |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 327 | rx_pkt_last = call->rx_pkt_last; |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 328 | |
| 329 | if (call->state >= RXRPC_CALL_SERVER_ACK_REQUEST) { |
| 330 | seq = call->rx_hard_ack; |
| 331 | ret = 1; |
| 332 | goto done; |
| 333 | } |
| 334 | |
| 335 | /* Barriers against rxrpc_input_data(). */ |
| 336 | hard_ack = call->rx_hard_ack; |
| 337 | seq = hard_ack + 1; |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 338 | |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 339 | while (top = smp_load_acquire(&call->rx_top), |
| 340 | before_eq(seq, top) |
| 341 | ) { |
| 342 | ix = seq & RXRPC_RXTX_BUFF_MASK; |
| 343 | skb = call->rxtx_buffer[ix]; |
| 344 | if (!skb) { |
| 345 | trace_rxrpc_recvmsg(call, rxrpc_recvmsg_hole, seq, |
| 346 | rx_pkt_offset, rx_pkt_len, 0); |
| 347 | break; |
| 348 | } |
| 349 | smp_rmb(); |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 350 | rxrpc_see_skb(skb, rxrpc_skb_seen); |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 351 | sp = rxrpc_skb(skb); |
| 352 | |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 353 | if (!(flags & MSG_PEEK)) { |
| 354 | serial = sp->hdr.serial; |
| 355 | serial += call->rxtx_annotations[ix] & RXRPC_RX_ANNO_SUBPACKET; |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 356 | trace_rxrpc_receive(call, rxrpc_receive_front, |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 357 | serial, seq); |
| 358 | } |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 359 | |
| 360 | if (msg) |
| 361 | sock_recv_timestamp(msg, sock->sk, skb); |
| 362 | |
| 363 | if (rx_pkt_offset == 0) { |
| 364 | ret2 = rxrpc_locate_data(call, skb, |
| 365 | &call->rxtx_annotations[ix], |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 366 | &rx_pkt_offset, &rx_pkt_len, |
| 367 | &rx_pkt_last); |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 368 | trace_rxrpc_recvmsg(call, rxrpc_recvmsg_next, seq, |
| 369 | rx_pkt_offset, rx_pkt_len, ret2); |
| 370 | if (ret2 < 0) { |
| 371 | ret = ret2; |
| 372 | goto out; |
| 373 | } |
| 374 | } else { |
| 375 | trace_rxrpc_recvmsg(call, rxrpc_recvmsg_cont, seq, |
| 376 | rx_pkt_offset, rx_pkt_len, 0); |
| 377 | } |
| 378 | |
| 379 | /* We have to handle short, empty and used-up DATA packets. */ |
| 380 | remain = len - *_offset; |
| 381 | copy = rx_pkt_len; |
| 382 | if (copy > remain) |
| 383 | copy = remain; |
| 384 | if (copy > 0) { |
| 385 | ret2 = skb_copy_datagram_iter(skb, rx_pkt_offset, iter, |
| 386 | copy); |
| 387 | if (ret2 < 0) { |
| 388 | ret = ret2; |
| 389 | goto out; |
| 390 | } |
| 391 | |
| 392 | /* handle piecemeal consumption of data packets */ |
| 393 | rx_pkt_offset += copy; |
| 394 | rx_pkt_len -= copy; |
| 395 | *_offset += copy; |
| 396 | } |
| 397 | |
| 398 | if (rx_pkt_len > 0) { |
| 399 | trace_rxrpc_recvmsg(call, rxrpc_recvmsg_full, seq, |
| 400 | rx_pkt_offset, rx_pkt_len, 0); |
| 401 | ASSERTCMP(*_offset, ==, len); |
| 402 | ret = 0; |
| 403 | break; |
| 404 | } |
| 405 | |
| 406 | /* The whole packet has been transferred. */ |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 407 | if (!(flags & MSG_PEEK)) |
| 408 | rxrpc_rotate_rx_window(call); |
| 409 | rx_pkt_offset = 0; |
| 410 | rx_pkt_len = 0; |
| 411 | |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 412 | if (rx_pkt_last) { |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 413 | ASSERTCMP(seq, ==, READ_ONCE(call->rx_top)); |
| 414 | ret = 1; |
| 415 | goto out; |
| 416 | } |
| 417 | |
| 418 | seq++; |
| 419 | } |
| 420 | |
| 421 | out: |
| 422 | if (!(flags & MSG_PEEK)) { |
| 423 | call->rx_pkt_offset = rx_pkt_offset; |
| 424 | call->rx_pkt_len = rx_pkt_len; |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 425 | call->rx_pkt_last = rx_pkt_last; |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 426 | } |
| 427 | done: |
| 428 | trace_rxrpc_recvmsg(call, rxrpc_recvmsg_data_return, seq, |
| 429 | rx_pkt_offset, rx_pkt_len, ret); |
| 430 | if (ret == -EAGAIN) |
| 431 | set_bit(RXRPC_CALL_RX_UNDERRUN, &call->flags); |
| 432 | return ret; |
| 433 | } |
| 434 | |
| 435 | /* |
| 436 | * Receive a message from an RxRPC socket |
| 437 | * - we need to be careful about two or more threads calling recvmsg |
| 438 | * simultaneously |
| 439 | */ |
| 440 | int rxrpc_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, |
| 441 | int flags) |
| 442 | { |
| 443 | struct rxrpc_call *call; |
| 444 | struct rxrpc_sock *rx = rxrpc_sk(sock->sk); |
| 445 | struct list_head *l; |
| 446 | size_t copied = 0; |
| 447 | long timeo; |
| 448 | int ret; |
| 449 | |
| 450 | DEFINE_WAIT(wait); |
| 451 | |
| 452 | trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_enter, 0, 0, 0, 0); |
| 453 | |
| 454 | if (flags & (MSG_OOB | MSG_TRUNC)) |
| 455 | return -EOPNOTSUPP; |
| 456 | |
| 457 | timeo = sock_rcvtimeo(&rx->sk, flags & MSG_DONTWAIT); |
| 458 | |
| 459 | try_again: |
| 460 | lock_sock(&rx->sk); |
| 461 | |
| 462 | /* Return immediately if a client socket has no outstanding calls */ |
| 463 | if (RB_EMPTY_ROOT(&rx->calls) && |
| 464 | list_empty(&rx->recvmsg_q) && |
| 465 | rx->sk.sk_state != RXRPC_SERVER_LISTENING) { |
| 466 | release_sock(&rx->sk); |
| 467 | return -ENODATA; |
| 468 | } |
| 469 | |
| 470 | if (list_empty(&rx->recvmsg_q)) { |
| 471 | ret = -EWOULDBLOCK; |
| 472 | if (timeo == 0) { |
| 473 | call = NULL; |
| 474 | goto error_no_call; |
| 475 | } |
| 476 | |
| 477 | release_sock(&rx->sk); |
| 478 | |
| 479 | /* Wait for something to happen */ |
| 480 | prepare_to_wait_exclusive(sk_sleep(&rx->sk), &wait, |
| 481 | TASK_INTERRUPTIBLE); |
| 482 | ret = sock_error(&rx->sk); |
| 483 | if (ret) |
| 484 | goto wait_error; |
| 485 | |
| 486 | if (list_empty(&rx->recvmsg_q)) { |
| 487 | if (signal_pending(current)) |
| 488 | goto wait_interrupted; |
| 489 | trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_wait, |
| 490 | 0, 0, 0, 0); |
| 491 | timeo = schedule_timeout(timeo); |
| 492 | } |
| 493 | finish_wait(sk_sleep(&rx->sk), &wait); |
| 494 | goto try_again; |
| 495 | } |
| 496 | |
| 497 | /* Find the next call and dequeue it if we're not just peeking. If we |
| 498 | * do dequeue it, that comes with a ref that we will need to release. |
| 499 | */ |
| 500 | write_lock_bh(&rx->recvmsg_lock); |
| 501 | l = rx->recvmsg_q.next; |
| 502 | call = list_entry(l, struct rxrpc_call, recvmsg_link); |
| 503 | if (!(flags & MSG_PEEK)) |
| 504 | list_del_init(&call->recvmsg_link); |
| 505 | else |
| 506 | rxrpc_get_call(call, rxrpc_call_got); |
| 507 | write_unlock_bh(&rx->recvmsg_lock); |
| 508 | |
| 509 | trace_rxrpc_recvmsg(call, rxrpc_recvmsg_dequeue, 0, 0, 0, 0); |
| 510 | |
| 511 | /* We're going to drop the socket lock, so we need to lock the call |
| 512 | * against interference by sendmsg. |
| 513 | */ |
| 514 | if (!mutex_trylock(&call->user_mutex)) { |
| 515 | ret = -EWOULDBLOCK; |
| 516 | if (flags & MSG_DONTWAIT) |
| 517 | goto error_requeue_call; |
| 518 | ret = -ERESTARTSYS; |
| 519 | if (mutex_lock_interruptible(&call->user_mutex) < 0) |
| 520 | goto error_requeue_call; |
| 521 | } |
| 522 | |
| 523 | release_sock(&rx->sk); |
| 524 | |
| 525 | if (test_bit(RXRPC_CALL_RELEASED, &call->flags)) |
| 526 | BUG(); |
| 527 | |
| 528 | if (test_bit(RXRPC_CALL_HAS_USERID, &call->flags)) { |
| 529 | if (flags & MSG_CMSG_COMPAT) { |
| 530 | unsigned int id32 = call->user_call_ID; |
| 531 | |
| 532 | ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID, |
| 533 | sizeof(unsigned int), &id32); |
| 534 | } else { |
| 535 | unsigned long idl = call->user_call_ID; |
| 536 | |
| 537 | ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID, |
| 538 | sizeof(unsigned long), &idl); |
| 539 | } |
| 540 | if (ret < 0) |
| 541 | goto error_unlock_call; |
| 542 | } |
| 543 | |
| 544 | if (msg->msg_name) { |
| 545 | struct sockaddr_rxrpc *srx = msg->msg_name; |
| 546 | size_t len = sizeof(call->peer->srx); |
| 547 | |
| 548 | memcpy(msg->msg_name, &call->peer->srx, len); |
| 549 | srx->srx_service = call->service_id; |
| 550 | msg->msg_namelen = len; |
| 551 | } |
| 552 | |
| 553 | switch (READ_ONCE(call->state)) { |
| 554 | case RXRPC_CALL_SERVER_ACCEPTING: |
| 555 | ret = rxrpc_recvmsg_new_call(rx, call, msg, flags); |
| 556 | break; |
| 557 | case RXRPC_CALL_CLIENT_RECV_REPLY: |
| 558 | case RXRPC_CALL_SERVER_RECV_REQUEST: |
| 559 | case RXRPC_CALL_SERVER_ACK_REQUEST: |
| 560 | ret = rxrpc_recvmsg_data(sock, call, msg, &msg->msg_iter, len, |
| 561 | flags, &copied); |
| 562 | if (ret == -EAGAIN) |
| 563 | ret = 0; |
| 564 | |
| 565 | if (after(call->rx_top, call->rx_hard_ack) && |
| 566 | call->rxtx_buffer[(call->rx_hard_ack + 1) & RXRPC_RXTX_BUFF_MASK]) |
| 567 | rxrpc_notify_socket(call); |
| 568 | break; |
| 569 | default: |
| 570 | ret = 0; |
| 571 | break; |
| 572 | } |
| 573 | |
| 574 | if (ret < 0) |
| 575 | goto error_unlock_call; |
| 576 | |
| 577 | if (call->state == RXRPC_CALL_COMPLETE) { |
| 578 | ret = rxrpc_recvmsg_term(call, msg); |
| 579 | if (ret < 0) |
| 580 | goto error_unlock_call; |
| 581 | if (!(flags & MSG_PEEK)) |
| 582 | rxrpc_release_call(rx, call); |
| 583 | msg->msg_flags |= MSG_EOR; |
| 584 | ret = 1; |
| 585 | } |
| 586 | |
| 587 | if (ret == 0) |
| 588 | msg->msg_flags |= MSG_MORE; |
| 589 | else |
| 590 | msg->msg_flags &= ~MSG_MORE; |
| 591 | ret = copied; |
| 592 | |
| 593 | error_unlock_call: |
| 594 | mutex_unlock(&call->user_mutex); |
| 595 | rxrpc_put_call(call, rxrpc_call_put); |
| 596 | trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, 0, 0, 0, ret); |
| 597 | return ret; |
| 598 | |
| 599 | error_requeue_call: |
| 600 | if (!(flags & MSG_PEEK)) { |
| 601 | write_lock_bh(&rx->recvmsg_lock); |
| 602 | list_add(&call->recvmsg_link, &rx->recvmsg_q); |
| 603 | write_unlock_bh(&rx->recvmsg_lock); |
| 604 | trace_rxrpc_recvmsg(call, rxrpc_recvmsg_requeue, 0, 0, 0, 0); |
| 605 | } else { |
| 606 | rxrpc_put_call(call, rxrpc_call_put); |
| 607 | } |
| 608 | error_no_call: |
| 609 | release_sock(&rx->sk); |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 610 | error_trace: |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 611 | trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, 0, 0, 0, ret); |
| 612 | return ret; |
| 613 | |
| 614 | wait_interrupted: |
| 615 | ret = sock_intr_errno(timeo); |
| 616 | wait_error: |
| 617 | finish_wait(sk_sleep(&rx->sk), &wait); |
| 618 | call = NULL; |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 619 | goto error_trace; |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 620 | } |
| 621 | |
| 622 | /** |
| 623 | * rxrpc_kernel_recv_data - Allow a kernel service to receive data/info |
| 624 | * @sock: The socket that the call exists on |
| 625 | * @call: The call to send data through |
| 626 | * @iter: The buffer to receive into |
| 627 | * @want_more: True if more data is expected to be read |
| 628 | * @_abort: Where the abort code is stored if -ECONNABORTED is returned |
| 629 | * @_service: Where to store the actual service ID (may be upgraded) |
| 630 | * |
| 631 | * Allow a kernel service to receive data and pick up information about the |
| 632 | * state of a call. Returns 0 if got what was asked for and there's more |
| 633 | * available, 1 if we got what was asked for and we're at the end of the data |
| 634 | * and -EAGAIN if we need more data. |
| 635 | * |
| 636 | * Note that we may return -EAGAIN to drain empty packets at the end of the |
| 637 | * data, even if we've already copied over the requested data. |
| 638 | * |
| 639 | * *_abort should also be initialised to 0. |
| 640 | */ |
| 641 | int rxrpc_kernel_recv_data(struct socket *sock, struct rxrpc_call *call, |
| 642 | struct iov_iter *iter, |
| 643 | bool want_more, u32 *_abort, u16 *_service) |
| 644 | { |
| 645 | size_t offset = 0; |
| 646 | int ret; |
| 647 | |
| 648 | _enter("{%d,%s},%zu,%d", |
| 649 | call->debug_id, rxrpc_call_states[call->state], |
| 650 | iov_iter_count(iter), want_more); |
| 651 | |
| 652 | ASSERTCMP(call->state, !=, RXRPC_CALL_SERVER_ACCEPTING); |
| 653 | |
| 654 | mutex_lock(&call->user_mutex); |
| 655 | |
| 656 | switch (READ_ONCE(call->state)) { |
| 657 | case RXRPC_CALL_CLIENT_RECV_REPLY: |
| 658 | case RXRPC_CALL_SERVER_RECV_REQUEST: |
| 659 | case RXRPC_CALL_SERVER_ACK_REQUEST: |
| 660 | ret = rxrpc_recvmsg_data(sock, call, NULL, iter, |
| 661 | iov_iter_count(iter), 0, |
| 662 | &offset); |
| 663 | if (ret < 0) |
| 664 | goto out; |
| 665 | |
| 666 | /* We can only reach here with a partially full buffer if we |
| 667 | * have reached the end of the data. We must otherwise have a |
| 668 | * full buffer or have been given -EAGAIN. |
| 669 | */ |
| 670 | if (ret == 1) { |
| 671 | if (iov_iter_count(iter) > 0) |
| 672 | goto short_data; |
| 673 | if (!want_more) |
| 674 | goto read_phase_complete; |
| 675 | ret = 0; |
| 676 | goto out; |
| 677 | } |
| 678 | |
| 679 | if (!want_more) |
| 680 | goto excess_data; |
| 681 | goto out; |
| 682 | |
| 683 | case RXRPC_CALL_COMPLETE: |
| 684 | goto call_complete; |
| 685 | |
| 686 | default: |
| 687 | ret = -EINPROGRESS; |
| 688 | goto out; |
| 689 | } |
| 690 | |
| 691 | read_phase_complete: |
| 692 | ret = 1; |
| 693 | out: |
| 694 | switch (call->ackr_reason) { |
| 695 | case RXRPC_ACK_IDLE: |
| 696 | break; |
| 697 | case RXRPC_ACK_DELAY: |
| 698 | if (ret != -EAGAIN) |
| 699 | break; |
| 700 | /* Fall through */ |
| 701 | default: |
| 702 | rxrpc_send_ack_packet(call, false, NULL); |
| 703 | } |
| 704 | |
| 705 | if (_service) |
| 706 | *_service = call->service_id; |
| 707 | mutex_unlock(&call->user_mutex); |
| 708 | _leave(" = %d [%zu,%d]", ret, iov_iter_count(iter), *_abort); |
| 709 | return ret; |
| 710 | |
| 711 | short_data: |
| 712 | trace_rxrpc_rx_eproto(call, 0, tracepoint_string("short_data")); |
| 713 | ret = -EBADMSG; |
| 714 | goto out; |
| 715 | excess_data: |
| 716 | trace_rxrpc_rx_eproto(call, 0, tracepoint_string("excess_data")); |
| 717 | ret = -EMSGSIZE; |
| 718 | goto out; |
| 719 | call_complete: |
| 720 | *_abort = call->abort_code; |
| 721 | ret = call->error; |
| 722 | if (call->completion == RXRPC_CALL_SUCCEEDED) { |
| 723 | ret = 1; |
| 724 | if (iov_iter_count(iter) > 0) |
| 725 | ret = -ECONNRESET; |
| 726 | } |
| 727 | goto out; |
| 728 | } |
| 729 | EXPORT_SYMBOL(rxrpc_kernel_recv_data); |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 730 | |
| 731 | /** |
| 732 | * rxrpc_kernel_get_reply_time - Get timestamp on first reply packet |
| 733 | * @sock: The socket that the call exists on |
| 734 | * @call: The call to query |
| 735 | * @_ts: Where to put the timestamp |
| 736 | * |
| 737 | * Retrieve the timestamp from the first DATA packet of the reply if it is |
| 738 | * in the ring. Returns true if successful, false if not. |
| 739 | */ |
| 740 | bool rxrpc_kernel_get_reply_time(struct socket *sock, struct rxrpc_call *call, |
| 741 | ktime_t *_ts) |
| 742 | { |
| 743 | struct sk_buff *skb; |
| 744 | rxrpc_seq_t hard_ack, top, seq; |
| 745 | bool success = false; |
| 746 | |
| 747 | mutex_lock(&call->user_mutex); |
| 748 | |
| 749 | if (READ_ONCE(call->state) != RXRPC_CALL_CLIENT_RECV_REPLY) |
| 750 | goto out; |
| 751 | |
| 752 | hard_ack = call->rx_hard_ack; |
| 753 | if (hard_ack != 0) |
| 754 | goto out; |
| 755 | |
| 756 | seq = hard_ack + 1; |
| 757 | top = smp_load_acquire(&call->rx_top); |
| 758 | if (after(seq, top)) |
| 759 | goto out; |
| 760 | |
| 761 | skb = call->rxtx_buffer[seq & RXRPC_RXTX_BUFF_MASK]; |
| 762 | if (!skb) |
| 763 | goto out; |
| 764 | |
| 765 | *_ts = skb_get_ktime(skb); |
| 766 | success = true; |
| 767 | |
| 768 | out: |
| 769 | mutex_unlock(&call->user_mutex); |
| 770 | return success; |
| 771 | } |
| 772 | EXPORT_SYMBOL(rxrpc_kernel_get_reply_time); |