Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1 | #include <linux/ceph/ceph_debug.h> |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2 | |
| 3 | #include <linux/crc32c.h> |
| 4 | #include <linux/ctype.h> |
| 5 | #include <linux/highmem.h> |
| 6 | #include <linux/inet.h> |
| 7 | #include <linux/kthread.h> |
| 8 | #include <linux/net.h> |
Tejun Heo | 5a0e3ad | 2010-03-24 17:04:11 +0900 | [diff] [blame] | 9 | #include <linux/slab.h> |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 10 | #include <linux/socket.h> |
| 11 | #include <linux/string.h> |
Alex Elder | 3ebc21f | 2013-01-31 16:02:01 -0600 | [diff] [blame] | 12 | #ifdef CONFIG_BLOCK |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 13 | #include <linux/bio.h> |
Alex Elder | 3ebc21f | 2013-01-31 16:02:01 -0600 | [diff] [blame] | 14 | #endif /* CONFIG_BLOCK */ |
Noah Watkins | ee3b56f | 2011-09-23 11:48:42 -0700 | [diff] [blame] | 15 | #include <linux/dns_resolver.h> |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 16 | #include <net/tcp.h> |
| 17 | |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 18 | #include <linux/ceph/libceph.h> |
| 19 | #include <linux/ceph/messenger.h> |
| 20 | #include <linux/ceph/decode.h> |
| 21 | #include <linux/ceph/pagelist.h> |
Paul Gortmaker | bc3b2d7 | 2011-07-15 11:47:34 -0400 | [diff] [blame] | 22 | #include <linux/export.h> |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 23 | |
| 24 | /* |
| 25 | * Ceph uses the messenger to exchange ceph_msg messages with other |
| 26 | * hosts in the system. The messenger provides ordered and reliable |
| 27 | * delivery. We tolerate TCP disconnects by reconnecting (with |
| 28 | * exponential backoff) in the case of a fault (disconnection, bad |
| 29 | * crc, protocol error). Acks allow sent messages to be discarded by |
| 30 | * the sender. |
| 31 | */ |
| 32 | |
Alex Elder | bc18f4b | 2012-06-20 21:53:53 -0500 | [diff] [blame] | 33 | /* |
| 34 | * We track the state of the socket on a given connection using |
| 35 | * values defined below. The transition to a new socket state is |
| 36 | * handled by a function which verifies we aren't coming from an |
| 37 | * unexpected state. |
| 38 | * |
| 39 | * -------- |
| 40 | * | NEW* | transient initial state |
| 41 | * -------- |
| 42 | * | con_sock_state_init() |
| 43 | * v |
| 44 | * ---------- |
| 45 | * | CLOSED | initialized, but no socket (and no |
| 46 | * ---------- TCP connection) |
| 47 | * ^ \ |
| 48 | * | \ con_sock_state_connecting() |
| 49 | * | ---------------------- |
| 50 | * | \ |
| 51 | * + con_sock_state_closed() \ |
Sage Weil | fbb85a4 | 2012-06-27 12:31:02 -0700 | [diff] [blame] | 52 | * |+--------------------------- \ |
| 53 | * | \ \ \ |
| 54 | * | ----------- \ \ |
| 55 | * | | CLOSING | socket event; \ \ |
| 56 | * | ----------- await close \ \ |
| 57 | * | ^ \ | |
| 58 | * | | \ | |
| 59 | * | + con_sock_state_closing() \ | |
| 60 | * | / \ | | |
| 61 | * | / --------------- | | |
| 62 | * | / \ v v |
Alex Elder | bc18f4b | 2012-06-20 21:53:53 -0500 | [diff] [blame] | 63 | * | / -------------- |
| 64 | * | / -----------------| CONNECTING | socket created, TCP |
| 65 | * | | / -------------- connect initiated |
| 66 | * | | | con_sock_state_connected() |
| 67 | * | | v |
| 68 | * ------------- |
| 69 | * | CONNECTED | TCP connection established |
| 70 | * ------------- |
| 71 | * |
| 72 | * State values for ceph_connection->sock_state; NEW is assumed to be 0. |
| 73 | */ |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 74 | |
| 75 | #define CON_SOCK_STATE_NEW 0 /* -> CLOSED */ |
| 76 | #define CON_SOCK_STATE_CLOSED 1 /* -> CONNECTING */ |
| 77 | #define CON_SOCK_STATE_CONNECTING 2 /* -> CONNECTED or -> CLOSING */ |
| 78 | #define CON_SOCK_STATE_CONNECTED 3 /* -> CLOSING or -> CLOSED */ |
| 79 | #define CON_SOCK_STATE_CLOSING 4 /* -> CLOSED */ |
| 80 | |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 81 | /* |
| 82 | * connection states |
| 83 | */ |
| 84 | #define CON_STATE_CLOSED 1 /* -> PREOPEN */ |
| 85 | #define CON_STATE_PREOPEN 2 /* -> CONNECTING, CLOSED */ |
| 86 | #define CON_STATE_CONNECTING 3 /* -> NEGOTIATING, CLOSED */ |
| 87 | #define CON_STATE_NEGOTIATING 4 /* -> OPEN, CLOSED */ |
| 88 | #define CON_STATE_OPEN 5 /* -> STANDBY, CLOSED */ |
| 89 | #define CON_STATE_STANDBY 6 /* -> PREOPEN, CLOSED */ |
| 90 | |
Sage Weil | 4a86169 | 2012-07-20 17:29:55 -0700 | [diff] [blame] | 91 | /* |
| 92 | * ceph_connection flag bits |
| 93 | */ |
| 94 | #define CON_FLAG_LOSSYTX 0 /* we can close channel or drop |
| 95 | * messages on errors */ |
| 96 | #define CON_FLAG_KEEPALIVE_PENDING 1 /* we need to send a keepalive */ |
| 97 | #define CON_FLAG_WRITE_PENDING 2 /* we have data ready to send */ |
| 98 | #define CON_FLAG_SOCK_CLOSED 3 /* socket state changed to closed */ |
| 99 | #define CON_FLAG_BACKOFF 4 /* need to retry queuing delayed work */ |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 100 | |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 101 | static bool con_flag_valid(unsigned long con_flag) |
| 102 | { |
| 103 | switch (con_flag) { |
| 104 | case CON_FLAG_LOSSYTX: |
| 105 | case CON_FLAG_KEEPALIVE_PENDING: |
| 106 | case CON_FLAG_WRITE_PENDING: |
| 107 | case CON_FLAG_SOCK_CLOSED: |
| 108 | case CON_FLAG_BACKOFF: |
| 109 | return true; |
| 110 | default: |
| 111 | return false; |
| 112 | } |
| 113 | } |
| 114 | |
| 115 | static void con_flag_clear(struct ceph_connection *con, unsigned long con_flag) |
| 116 | { |
| 117 | BUG_ON(!con_flag_valid(con_flag)); |
| 118 | |
| 119 | clear_bit(con_flag, &con->flags); |
| 120 | } |
| 121 | |
| 122 | static void con_flag_set(struct ceph_connection *con, unsigned long con_flag) |
| 123 | { |
| 124 | BUG_ON(!con_flag_valid(con_flag)); |
| 125 | |
| 126 | set_bit(con_flag, &con->flags); |
| 127 | } |
| 128 | |
| 129 | static bool con_flag_test(struct ceph_connection *con, unsigned long con_flag) |
| 130 | { |
| 131 | BUG_ON(!con_flag_valid(con_flag)); |
| 132 | |
| 133 | return test_bit(con_flag, &con->flags); |
| 134 | } |
| 135 | |
| 136 | static bool con_flag_test_and_clear(struct ceph_connection *con, |
| 137 | unsigned long con_flag) |
| 138 | { |
| 139 | BUG_ON(!con_flag_valid(con_flag)); |
| 140 | |
| 141 | return test_and_clear_bit(con_flag, &con->flags); |
| 142 | } |
| 143 | |
| 144 | static bool con_flag_test_and_set(struct ceph_connection *con, |
| 145 | unsigned long con_flag) |
| 146 | { |
| 147 | BUG_ON(!con_flag_valid(con_flag)); |
| 148 | |
| 149 | return test_and_set_bit(con_flag, &con->flags); |
| 150 | } |
| 151 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 152 | /* static tag bytes (protocol control messages) */ |
| 153 | static char tag_msg = CEPH_MSGR_TAG_MSG; |
| 154 | static char tag_ack = CEPH_MSGR_TAG_ACK; |
| 155 | static char tag_keepalive = CEPH_MSGR_TAG_KEEPALIVE; |
| 156 | |
Sage Weil | a6a5349 | 2010-04-13 14:07:07 -0700 | [diff] [blame] | 157 | #ifdef CONFIG_LOCKDEP |
| 158 | static struct lock_class_key socket_class; |
| 159 | #endif |
| 160 | |
Alex Elder | 84495f4 | 2012-02-15 07:43:55 -0600 | [diff] [blame] | 161 | /* |
| 162 | * When skipping (ignoring) a block of input we read it into a "skip |
| 163 | * buffer," which is this many bytes in size. |
| 164 | */ |
| 165 | #define SKIP_BUF_SIZE 1024 |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 166 | |
| 167 | static void queue_con(struct ceph_connection *con); |
| 168 | static void con_work(struct work_struct *); |
Alex Elder | 9320926 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 169 | static void con_fault(struct ceph_connection *con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 170 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 171 | /* |
Alex Elder | f64a931 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 172 | * Nicely render a sockaddr as a string. An array of formatted |
| 173 | * strings is used, to approximate reentrancy. |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 174 | */ |
Alex Elder | f64a931 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 175 | #define ADDR_STR_COUNT_LOG 5 /* log2(# address strings in array) */ |
| 176 | #define ADDR_STR_COUNT (1 << ADDR_STR_COUNT_LOG) |
| 177 | #define ADDR_STR_COUNT_MASK (ADDR_STR_COUNT - 1) |
| 178 | #define MAX_ADDR_STR_LEN 64 /* 54 is enough */ |
| 179 | |
| 180 | static char addr_str[ADDR_STR_COUNT][MAX_ADDR_STR_LEN]; |
| 181 | static atomic_t addr_str_seq = ATOMIC_INIT(0); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 182 | |
Alex Elder | 5766651 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 183 | static struct page *zero_page; /* used in certain error cases */ |
Alex Elder | 5766651 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 184 | |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 185 | const char *ceph_pr_addr(const struct sockaddr_storage *ss) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 186 | { |
| 187 | int i; |
| 188 | char *s; |
Alex Elder | 99f0f3b | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 189 | struct sockaddr_in *in4 = (struct sockaddr_in *) ss; |
| 190 | struct sockaddr_in6 *in6 = (struct sockaddr_in6 *) ss; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 191 | |
Alex Elder | f64a931 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 192 | i = atomic_inc_return(&addr_str_seq) & ADDR_STR_COUNT_MASK; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 193 | s = addr_str[i]; |
| 194 | |
| 195 | switch (ss->ss_family) { |
| 196 | case AF_INET: |
Alex Elder | bd40614 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 197 | snprintf(s, MAX_ADDR_STR_LEN, "%pI4:%hu", &in4->sin_addr, |
| 198 | ntohs(in4->sin_port)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 199 | break; |
| 200 | |
| 201 | case AF_INET6: |
Alex Elder | bd40614 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 202 | snprintf(s, MAX_ADDR_STR_LEN, "[%pI6c]:%hu", &in6->sin6_addr, |
| 203 | ntohs(in6->sin6_port)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 204 | break; |
| 205 | |
| 206 | default: |
Alex Elder | d3002b9 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 207 | snprintf(s, MAX_ADDR_STR_LEN, "(unknown sockaddr family %hu)", |
| 208 | ss->ss_family); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 209 | } |
| 210 | |
| 211 | return s; |
| 212 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 213 | EXPORT_SYMBOL(ceph_pr_addr); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 214 | |
Sage Weil | 63f2d21 | 2009-11-03 15:17:56 -0800 | [diff] [blame] | 215 | static void encode_my_addr(struct ceph_messenger *msgr) |
| 216 | { |
| 217 | memcpy(&msgr->my_enc_addr, &msgr->inst.addr, sizeof(msgr->my_enc_addr)); |
| 218 | ceph_encode_addr(&msgr->my_enc_addr); |
| 219 | } |
| 220 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 221 | /* |
| 222 | * work queue for all reading and writing to/from the socket. |
| 223 | */ |
Alex Elder | e0f43c9 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 224 | static struct workqueue_struct *ceph_msgr_wq; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 225 | |
Alex Elder | 1541716 | 2013-02-19 12:25:56 -0600 | [diff] [blame] | 226 | static void _ceph_msgr_exit(void) |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 227 | { |
Alex Elder | d3002b9 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 228 | if (ceph_msgr_wq) { |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 229 | destroy_workqueue(ceph_msgr_wq); |
Alex Elder | d3002b9 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 230 | ceph_msgr_wq = NULL; |
| 231 | } |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 232 | |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 233 | BUG_ON(zero_page == NULL); |
| 234 | kunmap(zero_page); |
| 235 | page_cache_release(zero_page); |
| 236 | zero_page = NULL; |
| 237 | } |
| 238 | |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 239 | int ceph_msgr_init(void) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 240 | { |
Alex Elder | 5766651 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 241 | BUG_ON(zero_page != NULL); |
| 242 | zero_page = ZERO_PAGE(0); |
| 243 | page_cache_get(zero_page); |
| 244 | |
Tejun Heo | f363e45fd1 | 2011-01-03 14:49:46 +0100 | [diff] [blame] | 245 | ceph_msgr_wq = alloc_workqueue("ceph-msgr", WQ_NON_REENTRANT, 0); |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 246 | if (ceph_msgr_wq) |
| 247 | return 0; |
Alex Elder | 5766651 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 248 | |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 249 | pr_err("msgr_init failed to create workqueue\n"); |
| 250 | _ceph_msgr_exit(); |
Alex Elder | 5766651 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 251 | |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 252 | return -ENOMEM; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 253 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 254 | EXPORT_SYMBOL(ceph_msgr_init); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 255 | |
| 256 | void ceph_msgr_exit(void) |
| 257 | { |
Alex Elder | 5766651 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 258 | BUG_ON(ceph_msgr_wq == NULL); |
Alex Elder | 5766651 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 259 | |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 260 | _ceph_msgr_exit(); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 261 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 262 | EXPORT_SYMBOL(ceph_msgr_exit); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 263 | |
Yehuda Sadeh | cd84db6 | 2010-06-11 16:58:48 -0700 | [diff] [blame] | 264 | void ceph_msgr_flush(void) |
Sage Weil | a922d38 | 2010-05-29 09:41:23 -0700 | [diff] [blame] | 265 | { |
| 266 | flush_workqueue(ceph_msgr_wq); |
| 267 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 268 | EXPORT_SYMBOL(ceph_msgr_flush); |
Sage Weil | a922d38 | 2010-05-29 09:41:23 -0700 | [diff] [blame] | 269 | |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 270 | /* Connection socket state transition functions */ |
| 271 | |
| 272 | static void con_sock_state_init(struct ceph_connection *con) |
| 273 | { |
| 274 | int old_state; |
| 275 | |
| 276 | old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CLOSED); |
| 277 | if (WARN_ON(old_state != CON_SOCK_STATE_NEW)) |
| 278 | printk("%s: unexpected old state %d\n", __func__, old_state); |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 279 | dout("%s con %p sock %d -> %d\n", __func__, con, old_state, |
| 280 | CON_SOCK_STATE_CLOSED); |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 281 | } |
| 282 | |
| 283 | static void con_sock_state_connecting(struct ceph_connection *con) |
| 284 | { |
| 285 | int old_state; |
| 286 | |
| 287 | old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CONNECTING); |
| 288 | if (WARN_ON(old_state != CON_SOCK_STATE_CLOSED)) |
| 289 | printk("%s: unexpected old state %d\n", __func__, old_state); |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 290 | dout("%s con %p sock %d -> %d\n", __func__, con, old_state, |
| 291 | CON_SOCK_STATE_CONNECTING); |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 292 | } |
| 293 | |
| 294 | static void con_sock_state_connected(struct ceph_connection *con) |
| 295 | { |
| 296 | int old_state; |
| 297 | |
| 298 | old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CONNECTED); |
| 299 | if (WARN_ON(old_state != CON_SOCK_STATE_CONNECTING)) |
| 300 | printk("%s: unexpected old state %d\n", __func__, old_state); |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 301 | dout("%s con %p sock %d -> %d\n", __func__, con, old_state, |
| 302 | CON_SOCK_STATE_CONNECTED); |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 303 | } |
| 304 | |
| 305 | static void con_sock_state_closing(struct ceph_connection *con) |
| 306 | { |
| 307 | int old_state; |
| 308 | |
| 309 | old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CLOSING); |
| 310 | if (WARN_ON(old_state != CON_SOCK_STATE_CONNECTING && |
| 311 | old_state != CON_SOCK_STATE_CONNECTED && |
| 312 | old_state != CON_SOCK_STATE_CLOSING)) |
| 313 | printk("%s: unexpected old state %d\n", __func__, old_state); |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 314 | dout("%s con %p sock %d -> %d\n", __func__, con, old_state, |
| 315 | CON_SOCK_STATE_CLOSING); |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 316 | } |
| 317 | |
| 318 | static void con_sock_state_closed(struct ceph_connection *con) |
| 319 | { |
| 320 | int old_state; |
| 321 | |
| 322 | old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CLOSED); |
| 323 | if (WARN_ON(old_state != CON_SOCK_STATE_CONNECTED && |
Sage Weil | fbb85a4 | 2012-06-27 12:31:02 -0700 | [diff] [blame] | 324 | old_state != CON_SOCK_STATE_CLOSING && |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 325 | old_state != CON_SOCK_STATE_CONNECTING && |
| 326 | old_state != CON_SOCK_STATE_CLOSED)) |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 327 | printk("%s: unexpected old state %d\n", __func__, old_state); |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 328 | dout("%s con %p sock %d -> %d\n", __func__, con, old_state, |
| 329 | CON_SOCK_STATE_CLOSED); |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 330 | } |
Sage Weil | a922d38 | 2010-05-29 09:41:23 -0700 | [diff] [blame] | 331 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 332 | /* |
| 333 | * socket callback functions |
| 334 | */ |
| 335 | |
| 336 | /* data available on socket, or listen socket received a connect */ |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 337 | static void ceph_sock_data_ready(struct sock *sk, int count_unused) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 338 | { |
Alex Elder | bd40614 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 339 | struct ceph_connection *con = sk->sk_user_data; |
Guanjun He | a2a3258 | 2012-07-08 19:50:33 -0700 | [diff] [blame] | 340 | if (atomic_read(&con->msgr->stopping)) { |
| 341 | return; |
| 342 | } |
Alex Elder | bd40614 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 343 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 344 | if (sk->sk_state != TCP_CLOSE_WAIT) { |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 345 | dout("%s on %p state = %lu, queueing work\n", __func__, |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 346 | con, con->state); |
| 347 | queue_con(con); |
| 348 | } |
| 349 | } |
| 350 | |
| 351 | /* socket has buffer space for writing */ |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 352 | static void ceph_sock_write_space(struct sock *sk) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 353 | { |
Alex Elder | d3002b9 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 354 | struct ceph_connection *con = sk->sk_user_data; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 355 | |
Jim Schutt | 182fac2 | 2012-02-29 08:30:58 -0700 | [diff] [blame] | 356 | /* only queue to workqueue if there is data we want to write, |
| 357 | * and there is sufficient space in the socket buffer to accept |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 358 | * more data. clear SOCK_NOSPACE so that ceph_sock_write_space() |
Jim Schutt | 182fac2 | 2012-02-29 08:30:58 -0700 | [diff] [blame] | 359 | * doesn't get called again until try_write() fills the socket |
| 360 | * buffer. See net/ipv4/tcp_input.c:tcp_check_space() |
| 361 | * and net/core/stream.c:sk_stream_write_space(). |
| 362 | */ |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 363 | if (con_flag_test(con, CON_FLAG_WRITE_PENDING)) { |
Jim Schutt | 182fac2 | 2012-02-29 08:30:58 -0700 | [diff] [blame] | 364 | if (sk_stream_wspace(sk) >= sk_stream_min_wspace(sk)) { |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 365 | dout("%s %p queueing write work\n", __func__, con); |
Jim Schutt | 182fac2 | 2012-02-29 08:30:58 -0700 | [diff] [blame] | 366 | clear_bit(SOCK_NOSPACE, &sk->sk_socket->flags); |
| 367 | queue_con(con); |
| 368 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 369 | } else { |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 370 | dout("%s %p nothing to write\n", __func__, con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 371 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 372 | } |
| 373 | |
| 374 | /* socket's state has changed */ |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 375 | static void ceph_sock_state_change(struct sock *sk) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 376 | { |
Alex Elder | bd40614 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 377 | struct ceph_connection *con = sk->sk_user_data; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 378 | |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 379 | dout("%s %p state = %lu sk_state = %u\n", __func__, |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 380 | con, con->state, sk->sk_state); |
| 381 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 382 | switch (sk->sk_state) { |
| 383 | case TCP_CLOSE: |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 384 | dout("%s TCP_CLOSE\n", __func__); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 385 | case TCP_CLOSE_WAIT: |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 386 | dout("%s TCP_CLOSE_WAIT\n", __func__); |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 387 | con_sock_state_closing(con); |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 388 | con_flag_set(con, CON_FLAG_SOCK_CLOSED); |
Alex Elder | d65c9e0 | 2012-06-20 21:53:53 -0500 | [diff] [blame] | 389 | queue_con(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 390 | break; |
| 391 | case TCP_ESTABLISHED: |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 392 | dout("%s TCP_ESTABLISHED\n", __func__); |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 393 | con_sock_state_connected(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 394 | queue_con(con); |
| 395 | break; |
Alex Elder | d3002b9 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 396 | default: /* Everything else is uninteresting */ |
| 397 | break; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 398 | } |
| 399 | } |
| 400 | |
| 401 | /* |
| 402 | * set up socket callbacks |
| 403 | */ |
| 404 | static void set_sock_callbacks(struct socket *sock, |
| 405 | struct ceph_connection *con) |
| 406 | { |
| 407 | struct sock *sk = sock->sk; |
Alex Elder | bd40614 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 408 | sk->sk_user_data = con; |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 409 | sk->sk_data_ready = ceph_sock_data_ready; |
| 410 | sk->sk_write_space = ceph_sock_write_space; |
| 411 | sk->sk_state_change = ceph_sock_state_change; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 412 | } |
| 413 | |
| 414 | |
| 415 | /* |
| 416 | * socket helpers |
| 417 | */ |
| 418 | |
| 419 | /* |
| 420 | * initiate connection to a remote socket. |
| 421 | */ |
Alex Elder | 41617d0 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 422 | static int ceph_tcp_connect(struct ceph_connection *con) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 423 | { |
Sage Weil | f91d347 | 2010-07-01 15:18:31 -0700 | [diff] [blame] | 424 | struct sockaddr_storage *paddr = &con->peer_addr.in_addr; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 425 | struct socket *sock; |
| 426 | int ret; |
| 427 | |
| 428 | BUG_ON(con->sock); |
Sage Weil | f91d347 | 2010-07-01 15:18:31 -0700 | [diff] [blame] | 429 | ret = sock_create_kern(con->peer_addr.in_addr.ss_family, SOCK_STREAM, |
| 430 | IPPROTO_TCP, &sock); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 431 | if (ret) |
Alex Elder | 41617d0 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 432 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 433 | sock->sk->sk_allocation = GFP_NOFS; |
| 434 | |
Sage Weil | a6a5349 | 2010-04-13 14:07:07 -0700 | [diff] [blame] | 435 | #ifdef CONFIG_LOCKDEP |
| 436 | lockdep_set_class(&sock->sk->sk_lock, &socket_class); |
| 437 | #endif |
| 438 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 439 | set_sock_callbacks(sock, con); |
| 440 | |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 441 | dout("connect %s\n", ceph_pr_addr(&con->peer_addr.in_addr)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 442 | |
Sage Weil | 89a86be | 2012-06-09 14:19:21 -0700 | [diff] [blame] | 443 | con_sock_state_connecting(con); |
Sage Weil | f91d347 | 2010-07-01 15:18:31 -0700 | [diff] [blame] | 444 | ret = sock->ops->connect(sock, (struct sockaddr *)paddr, sizeof(*paddr), |
| 445 | O_NONBLOCK); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 446 | if (ret == -EINPROGRESS) { |
| 447 | dout("connect %s EINPROGRESS sk_state = %u\n", |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 448 | ceph_pr_addr(&con->peer_addr.in_addr), |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 449 | sock->sk->sk_state); |
Alex Elder | a5bc312 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 450 | } else if (ret < 0) { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 451 | pr_err("connect %s error %d\n", |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 452 | ceph_pr_addr(&con->peer_addr.in_addr), ret); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 453 | sock_release(sock); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 454 | con->error_msg = "connect error"; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 455 | |
Alex Elder | 41617d0 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 456 | return ret; |
Alex Elder | a5bc312 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 457 | } |
| 458 | con->sock = sock; |
Alex Elder | 41617d0 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 459 | return 0; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 460 | } |
| 461 | |
| 462 | static int ceph_tcp_recvmsg(struct socket *sock, void *buf, size_t len) |
| 463 | { |
| 464 | struct kvec iov = {buf, len}; |
| 465 | struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL }; |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 466 | int r; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 467 | |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 468 | r = kernel_recvmsg(sock, &msg, &iov, 1, len, msg.msg_flags); |
| 469 | if (r == -EAGAIN) |
| 470 | r = 0; |
| 471 | return r; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 472 | } |
| 473 | |
| 474 | /* |
| 475 | * write something. @more is true if caller will be sending more data |
| 476 | * shortly. |
| 477 | */ |
| 478 | static int ceph_tcp_sendmsg(struct socket *sock, struct kvec *iov, |
| 479 | size_t kvlen, size_t len, int more) |
| 480 | { |
| 481 | struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL }; |
Sage Weil | 42961d2 | 2011-01-25 08:19:34 -0800 | [diff] [blame] | 482 | int r; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 483 | |
| 484 | if (more) |
| 485 | msg.msg_flags |= MSG_MORE; |
| 486 | else |
| 487 | msg.msg_flags |= MSG_EOR; /* superfluous, but what the hell */ |
| 488 | |
Sage Weil | 42961d2 | 2011-01-25 08:19:34 -0800 | [diff] [blame] | 489 | r = kernel_sendmsg(sock, &msg, iov, kvlen, len); |
| 490 | if (r == -EAGAIN) |
| 491 | r = 0; |
| 492 | return r; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 493 | } |
| 494 | |
Alex Elder | 3173913 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 495 | static int ceph_tcp_sendpage(struct socket *sock, struct page *page, |
Alex Elder | e1dcb12 | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 496 | int offset, size_t size, bool more) |
Alex Elder | 3173913 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 497 | { |
| 498 | int flags = MSG_DONTWAIT | MSG_NOSIGNAL | (more ? MSG_MORE : MSG_EOR); |
| 499 | int ret; |
| 500 | |
| 501 | ret = kernel_sendpage(sock, page, offset, size, flags); |
| 502 | if (ret == -EAGAIN) |
| 503 | ret = 0; |
| 504 | |
| 505 | return ret; |
| 506 | } |
| 507 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 508 | |
| 509 | /* |
| 510 | * Shutdown/close the socket for the given connection. |
| 511 | */ |
| 512 | static int con_close_socket(struct ceph_connection *con) |
| 513 | { |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 514 | int rc = 0; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 515 | |
| 516 | dout("con_close_socket on %p sock %p\n", con, con->sock); |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 517 | if (con->sock) { |
| 518 | rc = con->sock->ops->shutdown(con->sock, SHUT_RDWR); |
| 519 | sock_release(con->sock); |
| 520 | con->sock = NULL; |
| 521 | } |
Alex Elder | 456ea46 | 2012-06-20 21:53:53 -0500 | [diff] [blame] | 522 | |
| 523 | /* |
Sage Weil | 4a86169 | 2012-07-20 17:29:55 -0700 | [diff] [blame] | 524 | * Forcibly clear the SOCK_CLOSED flag. It gets set |
Alex Elder | 456ea46 | 2012-06-20 21:53:53 -0500 | [diff] [blame] | 525 | * independent of the connection mutex, and we could have |
| 526 | * received a socket close event before we had the chance to |
| 527 | * shut the socket down. |
| 528 | */ |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 529 | con_flag_clear(con, CON_FLAG_SOCK_CLOSED); |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 530 | |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 531 | con_sock_state_closed(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 532 | return rc; |
| 533 | } |
| 534 | |
| 535 | /* |
| 536 | * Reset a connection. Discard all incoming and outgoing messages |
| 537 | * and clear *_seq state. |
| 538 | */ |
| 539 | static void ceph_msg_remove(struct ceph_msg *msg) |
| 540 | { |
| 541 | list_del_init(&msg->list_head); |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 542 | BUG_ON(msg->con == NULL); |
Sage Weil | 36eb71a | 2012-06-21 12:47:08 -0700 | [diff] [blame] | 543 | msg->con->ops->put(msg->con); |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 544 | msg->con = NULL; |
| 545 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 546 | ceph_msg_put(msg); |
| 547 | } |
| 548 | static void ceph_msg_remove_list(struct list_head *head) |
| 549 | { |
| 550 | while (!list_empty(head)) { |
| 551 | struct ceph_msg *msg = list_first_entry(head, struct ceph_msg, |
| 552 | list_head); |
| 553 | ceph_msg_remove(msg); |
| 554 | } |
| 555 | } |
| 556 | |
| 557 | static void reset_connection(struct ceph_connection *con) |
| 558 | { |
| 559 | /* reset connection, out_queue, msg_ and connect_seq */ |
| 560 | /* discard existing out_queue and msg_seq */ |
Sage Weil | 0fa6ebc | 2012-12-27 20:27:04 -0600 | [diff] [blame] | 561 | dout("reset_connection %p\n", con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 562 | ceph_msg_remove_list(&con->out_queue); |
| 563 | ceph_msg_remove_list(&con->out_sent); |
| 564 | |
Sage Weil | cf3e5c4 | 2009-12-11 09:48:05 -0800 | [diff] [blame] | 565 | if (con->in_msg) { |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 566 | BUG_ON(con->in_msg->con != con); |
| 567 | con->in_msg->con = NULL; |
Sage Weil | cf3e5c4 | 2009-12-11 09:48:05 -0800 | [diff] [blame] | 568 | ceph_msg_put(con->in_msg); |
| 569 | con->in_msg = NULL; |
Sage Weil | 36eb71a | 2012-06-21 12:47:08 -0700 | [diff] [blame] | 570 | con->ops->put(con); |
Sage Weil | cf3e5c4 | 2009-12-11 09:48:05 -0800 | [diff] [blame] | 571 | } |
| 572 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 573 | con->connect_seq = 0; |
| 574 | con->out_seq = 0; |
Sage Weil | c86a293 | 2009-12-14 14:04:30 -0800 | [diff] [blame] | 575 | if (con->out_msg) { |
| 576 | ceph_msg_put(con->out_msg); |
| 577 | con->out_msg = NULL; |
| 578 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 579 | con->in_seq = 0; |
Sage Weil | 0e0d5e0 | 2010-04-02 16:07:19 -0700 | [diff] [blame] | 580 | con->in_seq_acked = 0; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 581 | } |
| 582 | |
| 583 | /* |
| 584 | * mark a peer down. drop any open connections. |
| 585 | */ |
| 586 | void ceph_con_close(struct ceph_connection *con) |
| 587 | { |
Sage Weil | 8c50c81 | 2012-07-30 16:24:37 -0700 | [diff] [blame] | 588 | mutex_lock(&con->mutex); |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 589 | dout("con_close %p peer %s\n", con, |
| 590 | ceph_pr_addr(&con->peer_addr.in_addr)); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 591 | con->state = CON_STATE_CLOSED; |
Alex Elder | a5988c4 | 2012-05-29 11:04:58 -0500 | [diff] [blame] | 592 | |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 593 | con_flag_clear(con, CON_FLAG_LOSSYTX); /* so we retry next connect */ |
| 594 | con_flag_clear(con, CON_FLAG_KEEPALIVE_PENDING); |
| 595 | con_flag_clear(con, CON_FLAG_WRITE_PENDING); |
| 596 | con_flag_clear(con, CON_FLAG_BACKOFF); |
Alex Elder | a5988c4 | 2012-05-29 11:04:58 -0500 | [diff] [blame] | 597 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 598 | reset_connection(con); |
Sage Weil | 6f2bc3f | 2010-04-02 16:16:34 -0700 | [diff] [blame] | 599 | con->peer_global_seq = 0; |
Sage Weil | 91e45ce3 | 2010-02-15 12:05:09 -0800 | [diff] [blame] | 600 | cancel_delayed_work(&con->work); |
Sage Weil | ee76e07 | 2012-07-20 16:45:49 -0700 | [diff] [blame] | 601 | con_close_socket(con); |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 602 | mutex_unlock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 603 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 604 | EXPORT_SYMBOL(ceph_con_close); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 605 | |
| 606 | /* |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 607 | * Reopen a closed connection, with a new peer address. |
| 608 | */ |
Sage Weil | b7a9e5d | 2012-06-27 12:24:08 -0700 | [diff] [blame] | 609 | void ceph_con_open(struct ceph_connection *con, |
| 610 | __u8 entity_type, __u64 entity_num, |
| 611 | struct ceph_entity_addr *addr) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 612 | { |
Sage Weil | 5469155 | 2012-07-30 16:21:40 -0700 | [diff] [blame] | 613 | mutex_lock(&con->mutex); |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 614 | dout("con_open %p %s\n", con, ceph_pr_addr(&addr->in_addr)); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 615 | |
Alex Elder | 122070a | 2012-12-26 10:43:57 -0600 | [diff] [blame] | 616 | WARN_ON(con->state != CON_STATE_CLOSED); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 617 | con->state = CON_STATE_PREOPEN; |
Alex Elder | a5988c4 | 2012-05-29 11:04:58 -0500 | [diff] [blame] | 618 | |
Sage Weil | b7a9e5d | 2012-06-27 12:24:08 -0700 | [diff] [blame] | 619 | con->peer_name.type = (__u8) entity_type; |
| 620 | con->peer_name.num = cpu_to_le64(entity_num); |
| 621 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 622 | memcpy(&con->peer_addr, addr, sizeof(*addr)); |
Sage Weil | 03c677e | 2009-11-20 15:14:15 -0800 | [diff] [blame] | 623 | con->delay = 0; /* reset backoff memory */ |
Sage Weil | 5469155 | 2012-07-30 16:21:40 -0700 | [diff] [blame] | 624 | mutex_unlock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 625 | queue_con(con); |
| 626 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 627 | EXPORT_SYMBOL(ceph_con_open); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 628 | |
| 629 | /* |
Sage Weil | 87b315a | 2010-03-22 14:51:18 -0700 | [diff] [blame] | 630 | * return true if this connection ever successfully opened |
| 631 | */ |
| 632 | bool ceph_con_opened(struct ceph_connection *con) |
| 633 | { |
| 634 | return con->connect_seq > 0; |
| 635 | } |
| 636 | |
| 637 | /* |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 638 | * initialize a new connection. |
| 639 | */ |
Alex Elder | 1bfd89f | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 640 | void ceph_con_init(struct ceph_connection *con, void *private, |
| 641 | const struct ceph_connection_operations *ops, |
Sage Weil | b7a9e5d | 2012-06-27 12:24:08 -0700 | [diff] [blame] | 642 | struct ceph_messenger *msgr) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 643 | { |
| 644 | dout("con_init %p\n", con); |
| 645 | memset(con, 0, sizeof(*con)); |
Alex Elder | 1bfd89f | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 646 | con->private = private; |
| 647 | con->ops = ops; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 648 | con->msgr = msgr; |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 649 | |
| 650 | con_sock_state_init(con); |
| 651 | |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 652 | mutex_init(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 653 | INIT_LIST_HEAD(&con->out_queue); |
| 654 | INIT_LIST_HEAD(&con->out_sent); |
| 655 | INIT_DELAYED_WORK(&con->work, con_work); |
Alex Elder | a5988c4 | 2012-05-29 11:04:58 -0500 | [diff] [blame] | 656 | |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 657 | con->state = CON_STATE_CLOSED; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 658 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 659 | EXPORT_SYMBOL(ceph_con_init); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 660 | |
| 661 | |
| 662 | /* |
| 663 | * We maintain a global counter to order connection attempts. Get |
| 664 | * a unique seq greater than @gt. |
| 665 | */ |
| 666 | static u32 get_global_seq(struct ceph_messenger *msgr, u32 gt) |
| 667 | { |
| 668 | u32 ret; |
| 669 | |
| 670 | spin_lock(&msgr->global_seq_lock); |
| 671 | if (msgr->global_seq < gt) |
| 672 | msgr->global_seq = gt; |
| 673 | ret = ++msgr->global_seq; |
| 674 | spin_unlock(&msgr->global_seq_lock); |
| 675 | return ret; |
| 676 | } |
| 677 | |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 678 | static void con_out_kvec_reset(struct ceph_connection *con) |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 679 | { |
| 680 | con->out_kvec_left = 0; |
| 681 | con->out_kvec_bytes = 0; |
| 682 | con->out_kvec_cur = &con->out_kvec[0]; |
| 683 | } |
| 684 | |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 685 | static void con_out_kvec_add(struct ceph_connection *con, |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 686 | size_t size, void *data) |
| 687 | { |
| 688 | int index; |
| 689 | |
| 690 | index = con->out_kvec_left; |
| 691 | BUG_ON(index >= ARRAY_SIZE(con->out_kvec)); |
| 692 | |
| 693 | con->out_kvec[index].iov_len = size; |
| 694 | con->out_kvec[index].iov_base = data; |
| 695 | con->out_kvec_left++; |
| 696 | con->out_kvec_bytes += size; |
| 697 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 698 | |
Alex Elder | df6ad1f | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 699 | #ifdef CONFIG_BLOCK |
Alex Elder | 07c09b7 | 2013-02-15 22:10:17 -0600 | [diff] [blame] | 700 | static void init_bio_iter(struct bio *bio, struct bio **bio_iter, |
| 701 | unsigned int *bio_seg) |
Alex Elder | df6ad1f | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 702 | { |
| 703 | if (!bio) { |
Alex Elder | 07c09b7 | 2013-02-15 22:10:17 -0600 | [diff] [blame] | 704 | *bio_iter = NULL; |
| 705 | *bio_seg = 0; |
Alex Elder | df6ad1f | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 706 | return; |
| 707 | } |
Alex Elder | 07c09b7 | 2013-02-15 22:10:17 -0600 | [diff] [blame] | 708 | *bio_iter = bio; |
| 709 | *bio_seg = (unsigned int) bio->bi_idx; |
Alex Elder | df6ad1f | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 710 | } |
| 711 | |
Alex Elder | 07c09b7 | 2013-02-15 22:10:17 -0600 | [diff] [blame] | 712 | static void iter_bio_next(struct bio **bio_iter, unsigned int *seg) |
Alex Elder | df6ad1f | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 713 | { |
| 714 | if (*bio_iter == NULL) |
| 715 | return; |
| 716 | |
| 717 | BUG_ON(*seg >= (*bio_iter)->bi_vcnt); |
| 718 | |
| 719 | (*seg)++; |
| 720 | if (*seg == (*bio_iter)->bi_vcnt) |
| 721 | init_bio_iter((*bio_iter)->bi_next, bio_iter, seg); |
| 722 | } |
| 723 | #endif |
| 724 | |
Alex Elder | 739c905 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 725 | static void prepare_write_message_data(struct ceph_connection *con) |
| 726 | { |
| 727 | struct ceph_msg *msg = con->out_msg; |
| 728 | |
| 729 | BUG_ON(!msg); |
| 730 | BUG_ON(!msg->hdr.data_len); |
| 731 | |
| 732 | /* initialize page iterator */ |
| 733 | con->out_msg_pos.page = 0; |
| 734 | if (msg->pages) |
| 735 | con->out_msg_pos.page_pos = msg->page_alignment; |
| 736 | else |
| 737 | con->out_msg_pos.page_pos = 0; |
Alex Elder | 572c588 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 738 | #ifdef CONFIG_BLOCK |
Alex Elder | abdaa6a | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 739 | if (msg->bio) |
Alex Elder | 572c588 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 740 | init_bio_iter(msg->bio, &msg->bio_iter, &msg->bio_seg); |
| 741 | #endif |
Alex Elder | 739c905 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 742 | con->out_msg_pos.data_pos = 0; |
| 743 | con->out_msg_pos.did_page_crc = false; |
| 744 | con->out_more = 1; /* data + footer will follow */ |
| 745 | } |
| 746 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 747 | /* |
| 748 | * Prepare footer for currently outgoing message, and finish things |
| 749 | * off. Assumes out_kvec* are already valid.. we just add on to the end. |
| 750 | */ |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 751 | static void prepare_write_message_footer(struct ceph_connection *con) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 752 | { |
| 753 | struct ceph_msg *m = con->out_msg; |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 754 | int v = con->out_kvec_left; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 755 | |
Alex Elder | fd154f3 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 756 | m->footer.flags |= CEPH_MSG_FOOTER_COMPLETE; |
| 757 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 758 | dout("prepare_write_message_footer %p\n", con); |
| 759 | con->out_kvec_is_msg = true; |
| 760 | con->out_kvec[v].iov_base = &m->footer; |
| 761 | con->out_kvec[v].iov_len = sizeof(m->footer); |
| 762 | con->out_kvec_bytes += sizeof(m->footer); |
| 763 | con->out_kvec_left++; |
| 764 | con->out_more = m->more_to_follow; |
Sage Weil | c86a293 | 2009-12-14 14:04:30 -0800 | [diff] [blame] | 765 | con->out_msg_done = true; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 766 | } |
| 767 | |
| 768 | /* |
| 769 | * Prepare headers for the next outgoing message. |
| 770 | */ |
| 771 | static void prepare_write_message(struct ceph_connection *con) |
| 772 | { |
| 773 | struct ceph_msg *m; |
Alex Elder | a9a0c51 | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 774 | u32 crc; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 775 | |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 776 | con_out_kvec_reset(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 777 | con->out_kvec_is_msg = true; |
Sage Weil | c86a293 | 2009-12-14 14:04:30 -0800 | [diff] [blame] | 778 | con->out_msg_done = false; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 779 | |
| 780 | /* Sneak an ack in there first? If we can get it into the same |
| 781 | * TCP packet that's a good thing. */ |
| 782 | if (con->in_seq > con->in_seq_acked) { |
| 783 | con->in_seq_acked = con->in_seq; |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 784 | con_out_kvec_add(con, sizeof (tag_ack), &tag_ack); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 785 | con->out_temp_ack = cpu_to_le64(con->in_seq_acked); |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 786 | con_out_kvec_add(con, sizeof (con->out_temp_ack), |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 787 | &con->out_temp_ack); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 788 | } |
| 789 | |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 790 | BUG_ON(list_empty(&con->out_queue)); |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 791 | m = list_first_entry(&con->out_queue, struct ceph_msg, list_head); |
Sage Weil | c86a293 | 2009-12-14 14:04:30 -0800 | [diff] [blame] | 792 | con->out_msg = m; |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 793 | BUG_ON(m->con != con); |
Sage Weil | 4cf9d54 | 2011-07-26 11:27:24 -0700 | [diff] [blame] | 794 | |
| 795 | /* put message on sent list */ |
| 796 | ceph_msg_get(m); |
| 797 | list_move_tail(&m->list_head, &con->out_sent); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 798 | |
Sage Weil | e84346b | 2010-05-11 21:20:38 -0700 | [diff] [blame] | 799 | /* |
| 800 | * only assign outgoing seq # if we haven't sent this message |
| 801 | * yet. if it is requeued, resend with it's original seq. |
| 802 | */ |
| 803 | if (m->needs_out_seq) { |
| 804 | m->hdr.seq = cpu_to_le64(++con->out_seq); |
| 805 | m->needs_out_seq = false; |
| 806 | } |
Yan, Zheng | b132cf4 | 2012-06-06 19:35:55 -0500 | [diff] [blame] | 807 | #ifdef CONFIG_BLOCK |
| 808 | else |
| 809 | m->bio_iter = NULL; |
| 810 | #endif |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 811 | |
Alex Elder | 4a73ef2 | 2013-03-07 15:38:26 -0600 | [diff] [blame] | 812 | dout("prepare_write_message %p seq %lld type %d len %d+%d+%d (%zd)\n", |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 813 | m, con->out_seq, le16_to_cpu(m->hdr.type), |
| 814 | le32_to_cpu(m->hdr.front_len), le32_to_cpu(m->hdr.middle_len), |
Alex Elder | 4a73ef2 | 2013-03-07 15:38:26 -0600 | [diff] [blame] | 815 | le32_to_cpu(m->hdr.data_len), m->length); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 816 | BUG_ON(le32_to_cpu(m->hdr.front_len) != m->front.iov_len); |
| 817 | |
| 818 | /* tag + hdr + front + middle */ |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 819 | con_out_kvec_add(con, sizeof (tag_msg), &tag_msg); |
| 820 | con_out_kvec_add(con, sizeof (m->hdr), &m->hdr); |
| 821 | con_out_kvec_add(con, m->front.iov_len, m->front.iov_base); |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 822 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 823 | if (m->middle) |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 824 | con_out_kvec_add(con, m->middle->vec.iov_len, |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 825 | m->middle->vec.iov_base); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 826 | |
| 827 | /* fill in crc (except data pages), footer */ |
Alex Elder | a9a0c51 | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 828 | crc = crc32c(0, &m->hdr, offsetof(struct ceph_msg_header, crc)); |
| 829 | con->out_msg->hdr.crc = cpu_to_le32(crc); |
Alex Elder | fd154f3 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 830 | con->out_msg->footer.flags = 0; |
Alex Elder | a9a0c51 | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 831 | |
| 832 | crc = crc32c(0, m->front.iov_base, m->front.iov_len); |
| 833 | con->out_msg->footer.front_crc = cpu_to_le32(crc); |
| 834 | if (m->middle) { |
| 835 | crc = crc32c(0, m->middle->vec.iov_base, |
| 836 | m->middle->vec.iov_len); |
| 837 | con->out_msg->footer.middle_crc = cpu_to_le32(crc); |
| 838 | } else |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 839 | con->out_msg->footer.middle_crc = 0; |
Alex Elder | 739c905 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 840 | dout("%s front_crc %u middle_crc %u\n", __func__, |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 841 | le32_to_cpu(con->out_msg->footer.front_crc), |
| 842 | le32_to_cpu(con->out_msg->footer.middle_crc)); |
| 843 | |
| 844 | /* is there a data payload? */ |
Alex Elder | 739c905 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 845 | con->out_msg->footer.data_crc = 0; |
| 846 | if (m->hdr.data_len) |
| 847 | prepare_write_message_data(con); |
| 848 | else |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 849 | /* no, queue up footer too and be done */ |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 850 | prepare_write_message_footer(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 851 | |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 852 | con_flag_set(con, CON_FLAG_WRITE_PENDING); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 853 | } |
| 854 | |
| 855 | /* |
| 856 | * Prepare an ack. |
| 857 | */ |
| 858 | static void prepare_write_ack(struct ceph_connection *con) |
| 859 | { |
| 860 | dout("prepare_write_ack %p %llu -> %llu\n", con, |
| 861 | con->in_seq_acked, con->in_seq); |
| 862 | con->in_seq_acked = con->in_seq; |
| 863 | |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 864 | con_out_kvec_reset(con); |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 865 | |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 866 | con_out_kvec_add(con, sizeof (tag_ack), &tag_ack); |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 867 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 868 | con->out_temp_ack = cpu_to_le64(con->in_seq_acked); |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 869 | con_out_kvec_add(con, sizeof (con->out_temp_ack), |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 870 | &con->out_temp_ack); |
| 871 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 872 | con->out_more = 1; /* more will follow.. eventually.. */ |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 873 | con_flag_set(con, CON_FLAG_WRITE_PENDING); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 874 | } |
| 875 | |
| 876 | /* |
| 877 | * Prepare to write keepalive byte. |
| 878 | */ |
| 879 | static void prepare_write_keepalive(struct ceph_connection *con) |
| 880 | { |
| 881 | dout("prepare_write_keepalive %p\n", con); |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 882 | con_out_kvec_reset(con); |
| 883 | con_out_kvec_add(con, sizeof (tag_keepalive), &tag_keepalive); |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 884 | con_flag_set(con, CON_FLAG_WRITE_PENDING); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 885 | } |
| 886 | |
| 887 | /* |
| 888 | * Connection negotiation. |
| 889 | */ |
| 890 | |
Alex Elder | dac1e71 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 891 | static struct ceph_auth_handshake *get_connect_authorizer(struct ceph_connection *con, |
| 892 | int *auth_proto) |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 893 | { |
Alex Elder | a3530df | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 894 | struct ceph_auth_handshake *auth; |
Alex Elder | b1c6b98 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 895 | |
| 896 | if (!con->ops->get_authorizer) { |
| 897 | con->out_connect.authorizer_protocol = CEPH_AUTH_UNKNOWN; |
| 898 | con->out_connect.authorizer_len = 0; |
Alex Elder | 729796b | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 899 | return NULL; |
Alex Elder | b1c6b98 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 900 | } |
| 901 | |
| 902 | /* Can't hold the mutex while getting authorizer */ |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 903 | mutex_unlock(&con->mutex); |
Alex Elder | dac1e71 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 904 | auth = con->ops->get_authorizer(con, auth_proto, con->auth_retry); |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 905 | mutex_lock(&con->mutex); |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 906 | |
Alex Elder | a3530df | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 907 | if (IS_ERR(auth)) |
Alex Elder | 729796b | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 908 | return auth; |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 909 | if (con->state != CON_STATE_NEGOTIATING) |
Alex Elder | 729796b | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 910 | return ERR_PTR(-EAGAIN); |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 911 | |
Alex Elder | 8f43fb5 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 912 | con->auth_reply_buf = auth->authorizer_reply_buf; |
| 913 | con->auth_reply_buf_len = auth->authorizer_reply_buf_len; |
Alex Elder | 729796b | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 914 | return auth; |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 915 | } |
| 916 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 917 | /* |
| 918 | * We connected to a peer and are saying hello. |
| 919 | */ |
Alex Elder | e825a66 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 920 | static void prepare_write_banner(struct ceph_connection *con) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 921 | { |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 922 | con_out_kvec_add(con, strlen(CEPH_BANNER), CEPH_BANNER); |
| 923 | con_out_kvec_add(con, sizeof (con->msgr->my_enc_addr), |
Alex Elder | e825a66 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 924 | &con->msgr->my_enc_addr); |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 925 | |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 926 | con->out_more = 0; |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 927 | con_flag_set(con, CON_FLAG_WRITE_PENDING); |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 928 | } |
| 929 | |
Alex Elder | e825a66 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 930 | static int prepare_write_connect(struct ceph_connection *con) |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 931 | { |
Eric Dumazet | 95c9617 | 2012-04-15 05:58:06 +0000 | [diff] [blame] | 932 | unsigned int global_seq = get_global_seq(con->msgr, 0); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 933 | int proto; |
Alex Elder | dac1e71 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 934 | int auth_proto; |
Alex Elder | 729796b | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 935 | struct ceph_auth_handshake *auth; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 936 | |
| 937 | switch (con->peer_name.type) { |
| 938 | case CEPH_ENTITY_TYPE_MON: |
| 939 | proto = CEPH_MONC_PROTOCOL; |
| 940 | break; |
| 941 | case CEPH_ENTITY_TYPE_OSD: |
| 942 | proto = CEPH_OSDC_PROTOCOL; |
| 943 | break; |
| 944 | case CEPH_ENTITY_TYPE_MDS: |
| 945 | proto = CEPH_MDSC_PROTOCOL; |
| 946 | break; |
| 947 | default: |
| 948 | BUG(); |
| 949 | } |
| 950 | |
| 951 | dout("prepare_write_connect %p cseq=%d gseq=%d proto=%d\n", con, |
| 952 | con->connect_seq, global_seq, proto); |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 953 | |
Alex Elder | e825a66 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 954 | con->out_connect.features = cpu_to_le64(con->msgr->supported_features); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 955 | con->out_connect.host_type = cpu_to_le32(CEPH_ENTITY_TYPE_CLIENT); |
| 956 | con->out_connect.connect_seq = cpu_to_le32(con->connect_seq); |
| 957 | con->out_connect.global_seq = cpu_to_le32(global_seq); |
| 958 | con->out_connect.protocol_version = cpu_to_le32(proto); |
| 959 | con->out_connect.flags = 0; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 960 | |
Alex Elder | dac1e71 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 961 | auth_proto = CEPH_AUTH_UNKNOWN; |
| 962 | auth = get_connect_authorizer(con, &auth_proto); |
Alex Elder | 729796b | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 963 | if (IS_ERR(auth)) |
| 964 | return PTR_ERR(auth); |
Alex Elder | 3da5477 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 965 | |
Alex Elder | dac1e71 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 966 | con->out_connect.authorizer_protocol = cpu_to_le32(auth_proto); |
Alex Elder | 3da5477 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 967 | con->out_connect.authorizer_len = auth ? |
| 968 | cpu_to_le32(auth->authorizer_buf_len) : 0; |
| 969 | |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 970 | con_out_kvec_add(con, sizeof (con->out_connect), |
Alex Elder | 3da5477 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 971 | &con->out_connect); |
| 972 | if (auth && auth->authorizer_buf_len) |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 973 | con_out_kvec_add(con, auth->authorizer_buf_len, |
Alex Elder | 3da5477 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 974 | auth->authorizer_buf); |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 975 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 976 | con->out_more = 0; |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 977 | con_flag_set(con, CON_FLAG_WRITE_PENDING); |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 978 | |
Alex Elder | e10c758 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 979 | return 0; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 980 | } |
| 981 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 982 | /* |
| 983 | * write as much of pending kvecs to the socket as we can. |
| 984 | * 1 -> done |
| 985 | * 0 -> socket full, but more to do |
| 986 | * <0 -> error |
| 987 | */ |
| 988 | static int write_partial_kvec(struct ceph_connection *con) |
| 989 | { |
| 990 | int ret; |
| 991 | |
| 992 | dout("write_partial_kvec %p %d left\n", con, con->out_kvec_bytes); |
| 993 | while (con->out_kvec_bytes > 0) { |
| 994 | ret = ceph_tcp_sendmsg(con->sock, con->out_kvec_cur, |
| 995 | con->out_kvec_left, con->out_kvec_bytes, |
| 996 | con->out_more); |
| 997 | if (ret <= 0) |
| 998 | goto out; |
| 999 | con->out_kvec_bytes -= ret; |
| 1000 | if (con->out_kvec_bytes == 0) |
| 1001 | break; /* done */ |
Alex Elder | f42299e | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 1002 | |
| 1003 | /* account for full iov entries consumed */ |
| 1004 | while (ret >= con->out_kvec_cur->iov_len) { |
| 1005 | BUG_ON(!con->out_kvec_left); |
| 1006 | ret -= con->out_kvec_cur->iov_len; |
| 1007 | con->out_kvec_cur++; |
| 1008 | con->out_kvec_left--; |
| 1009 | } |
| 1010 | /* and for a partially-consumed entry */ |
| 1011 | if (ret) { |
| 1012 | con->out_kvec_cur->iov_len -= ret; |
| 1013 | con->out_kvec_cur->iov_base += ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1014 | } |
| 1015 | } |
| 1016 | con->out_kvec_left = 0; |
| 1017 | con->out_kvec_is_msg = false; |
| 1018 | ret = 1; |
| 1019 | out: |
| 1020 | dout("write_partial_kvec %p %d left in %d kvecs ret = %d\n", con, |
| 1021 | con->out_kvec_bytes, con->out_kvec_left, ret); |
| 1022 | return ret; /* done! */ |
| 1023 | } |
| 1024 | |
Alex Elder | 84ca8fc | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1025 | static void out_msg_pos_next(struct ceph_connection *con, struct page *page, |
| 1026 | size_t len, size_t sent, bool in_trail) |
| 1027 | { |
| 1028 | struct ceph_msg *msg = con->out_msg; |
| 1029 | |
| 1030 | BUG_ON(!msg); |
| 1031 | BUG_ON(!sent); |
| 1032 | |
| 1033 | con->out_msg_pos.data_pos += sent; |
| 1034 | con->out_msg_pos.page_pos += sent; |
Alex Elder | 5821bd8 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1035 | if (sent < len) |
| 1036 | return; |
| 1037 | |
| 1038 | BUG_ON(sent != len); |
| 1039 | con->out_msg_pos.page_pos = 0; |
| 1040 | con->out_msg_pos.page++; |
| 1041 | con->out_msg_pos.did_page_crc = false; |
| 1042 | if (in_trail) |
Alex Elder | 35c7bfb | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1043 | list_rotate_left(&msg->trail->head); |
Alex Elder | 5821bd8 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1044 | else if (msg->pagelist) |
Alex Elder | 35c7bfb | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1045 | list_rotate_left(&msg->pagelist->head); |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1046 | #ifdef CONFIG_BLOCK |
Alex Elder | 5821bd8 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1047 | else if (msg->bio) |
| 1048 | iter_bio_next(&msg->bio_iter, &msg->bio_seg); |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1049 | #endif |
Alex Elder | 84ca8fc | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1050 | } |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1051 | |
Alex Elder | e788182 | 2013-03-08 18:51:04 -0600 | [diff] [blame] | 1052 | static void in_msg_pos_next(struct ceph_connection *con, size_t len, |
| 1053 | size_t received) |
| 1054 | { |
| 1055 | struct ceph_msg *msg = con->in_msg; |
| 1056 | |
| 1057 | BUG_ON(!msg); |
| 1058 | BUG_ON(!received); |
| 1059 | |
| 1060 | con->in_msg_pos.data_pos += received; |
| 1061 | con->in_msg_pos.page_pos += received; |
| 1062 | if (received < len) |
| 1063 | return; |
| 1064 | |
| 1065 | BUG_ON(received != len); |
| 1066 | con->in_msg_pos.page_pos = 0; |
| 1067 | con->in_msg_pos.page++; |
| 1068 | #ifdef CONFIG_BLOCK |
| 1069 | if (msg->bio) |
| 1070 | iter_bio_next(&msg->bio_iter, &msg->bio_seg); |
| 1071 | #endif /* CONFIG_BLOCK */ |
| 1072 | } |
| 1073 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1074 | /* |
| 1075 | * Write as much message data payload as we can. If we finish, queue |
| 1076 | * up the footer. |
| 1077 | * 1 -> done, footer is now queued in out_kvec[]. |
| 1078 | * 0 -> socket full, but more to do |
| 1079 | * <0 -> error |
| 1080 | */ |
| 1081 | static int write_partial_msg_pages(struct ceph_connection *con) |
| 1082 | { |
| 1083 | struct ceph_msg *msg = con->out_msg; |
Eric Dumazet | 95c9617 | 2012-04-15 05:58:06 +0000 | [diff] [blame] | 1084 | unsigned int data_len = le32_to_cpu(msg->hdr.data_len); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1085 | size_t len; |
Alex Elder | 37675b0 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 1086 | bool do_datacrc = !con->msgr->nocrc; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1087 | int ret; |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1088 | int total_max_write; |
Alex Elder | 84ca8fc | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1089 | bool in_trail = false; |
Alex Elder | 5821bd8 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1090 | const size_t trail_len = (msg->trail ? msg->trail->length : 0); |
| 1091 | const size_t trail_off = data_len - trail_len; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1092 | |
Alex Elder | 4a73ef2 | 2013-03-07 15:38:26 -0600 | [diff] [blame] | 1093 | dout("write_partial_msg_pages %p msg %p page %d offset %d\n", |
| 1094 | con, msg, con->out_msg_pos.page, con->out_msg_pos.page_pos); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1095 | |
Alex Elder | 5821bd8 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1096 | /* |
| 1097 | * Iterate through each page that contains data to be |
| 1098 | * written, and send as much as possible for each. |
| 1099 | * |
| 1100 | * If we are calculating the data crc (the default), we will |
| 1101 | * need to map the page. If we have no pages, they have |
| 1102 | * been revoked, so use the zero page. |
| 1103 | */ |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1104 | while (data_len > con->out_msg_pos.data_pos) { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1105 | struct page *page = NULL; |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1106 | int max_write = PAGE_SIZE; |
Alex Elder | 9bd1966 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 1107 | int bio_offset = 0; |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1108 | |
Alex Elder | 5821bd8 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1109 | in_trail = in_trail || con->out_msg_pos.data_pos >= trail_off; |
| 1110 | if (!in_trail) |
| 1111 | total_max_write = trail_off - con->out_msg_pos.data_pos; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1112 | |
Alex Elder | 5821bd8 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1113 | if (in_trail) { |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1114 | total_max_write = data_len - con->out_msg_pos.data_pos; |
| 1115 | |
| 1116 | page = list_first_entry(&msg->trail->head, |
| 1117 | struct page, lru); |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1118 | } else if (msg->pages) { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1119 | page = msg->pages[con->out_msg_pos.page]; |
Sage Weil | 58bb3b3 | 2009-12-23 12:12:31 -0800 | [diff] [blame] | 1120 | } else if (msg->pagelist) { |
| 1121 | page = list_first_entry(&msg->pagelist->head, |
| 1122 | struct page, lru); |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1123 | #ifdef CONFIG_BLOCK |
| 1124 | } else if (msg->bio) { |
| 1125 | struct bio_vec *bv; |
| 1126 | |
| 1127 | bv = bio_iovec_idx(msg->bio_iter, msg->bio_seg); |
| 1128 | page = bv->bv_page; |
Alex Elder | 9bd1966 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 1129 | bio_offset = bv->bv_offset; |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1130 | max_write = bv->bv_len; |
| 1131 | #endif |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1132 | } else { |
Alex Elder | 5766651 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 1133 | page = zero_page; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1134 | } |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1135 | len = min_t(int, max_write - con->out_msg_pos.page_pos, |
| 1136 | total_max_write); |
| 1137 | |
Alex Elder | 37675b0 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 1138 | if (do_datacrc && !con->out_msg_pos.did_page_crc) { |
Alex Elder | 9bd1966 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 1139 | void *base; |
Alex Elder | 5821bd8 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1140 | u32 crc = le32_to_cpu(msg->footer.data_crc); |
Alex Elder | 8d63e31 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 1141 | char *kaddr; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1142 | |
Alex Elder | 8d63e31 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 1143 | kaddr = kmap(page); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1144 | BUG_ON(kaddr == NULL); |
Alex Elder | 9bd1966 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 1145 | base = kaddr + con->out_msg_pos.page_pos + bio_offset; |
Alex Elder | 5821bd8 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1146 | crc = crc32c(crc, base, len); |
Alex Elder | 5ce765a | 2012-09-21 17:59:58 -0500 | [diff] [blame] | 1147 | kunmap(page); |
Alex Elder | 84ca8fc | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1148 | msg->footer.data_crc = cpu_to_le32(crc); |
Alex Elder | bca064d | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 1149 | con->out_msg_pos.did_page_crc = true; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1150 | } |
Alex Elder | e36b13c | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 1151 | ret = ceph_tcp_sendpage(con->sock, page, |
Alex Elder | 9bd1966 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 1152 | con->out_msg_pos.page_pos + bio_offset, |
Alex Elder | e1dcb12 | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1153 | len, true); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1154 | if (ret <= 0) |
| 1155 | goto out; |
| 1156 | |
Alex Elder | 84ca8fc | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1157 | out_msg_pos_next(con, page, len, (size_t) ret, in_trail); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1158 | } |
| 1159 | |
| 1160 | dout("write_partial_msg_pages %p msg %p done\n", con, msg); |
| 1161 | |
| 1162 | /* prepare and queue up footer, too */ |
Alex Elder | 37675b0 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 1163 | if (!do_datacrc) |
Alex Elder | 84ca8fc | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1164 | msg->footer.flags |= CEPH_MSG_FOOTER_NOCRC; |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1165 | con_out_kvec_reset(con); |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1166 | prepare_write_message_footer(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1167 | ret = 1; |
| 1168 | out: |
| 1169 | return ret; |
| 1170 | } |
| 1171 | |
| 1172 | /* |
| 1173 | * write some zeros |
| 1174 | */ |
| 1175 | static int write_partial_skip(struct ceph_connection *con) |
| 1176 | { |
| 1177 | int ret; |
| 1178 | |
| 1179 | while (con->out_skip > 0) { |
Alex Elder | 3173913 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 1180 | size_t size = min(con->out_skip, (int) PAGE_CACHE_SIZE); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1181 | |
Alex Elder | e1dcb12 | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1182 | ret = ceph_tcp_sendpage(con->sock, zero_page, 0, size, true); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1183 | if (ret <= 0) |
| 1184 | goto out; |
| 1185 | con->out_skip -= ret; |
| 1186 | } |
| 1187 | ret = 1; |
| 1188 | out: |
| 1189 | return ret; |
| 1190 | } |
| 1191 | |
| 1192 | /* |
| 1193 | * Prepare to read connection handshake, or an ack. |
| 1194 | */ |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1195 | static void prepare_read_banner(struct ceph_connection *con) |
| 1196 | { |
| 1197 | dout("prepare_read_banner %p\n", con); |
| 1198 | con->in_base_pos = 0; |
| 1199 | } |
| 1200 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1201 | static void prepare_read_connect(struct ceph_connection *con) |
| 1202 | { |
| 1203 | dout("prepare_read_connect %p\n", con); |
| 1204 | con->in_base_pos = 0; |
| 1205 | } |
| 1206 | |
| 1207 | static void prepare_read_ack(struct ceph_connection *con) |
| 1208 | { |
| 1209 | dout("prepare_read_ack %p\n", con); |
| 1210 | con->in_base_pos = 0; |
| 1211 | } |
| 1212 | |
| 1213 | static void prepare_read_tag(struct ceph_connection *con) |
| 1214 | { |
| 1215 | dout("prepare_read_tag %p\n", con); |
| 1216 | con->in_base_pos = 0; |
| 1217 | con->in_tag = CEPH_MSGR_TAG_READY; |
| 1218 | } |
| 1219 | |
| 1220 | /* |
| 1221 | * Prepare to read a message. |
| 1222 | */ |
| 1223 | static int prepare_read_message(struct ceph_connection *con) |
| 1224 | { |
| 1225 | dout("prepare_read_message %p\n", con); |
| 1226 | BUG_ON(con->in_msg != NULL); |
| 1227 | con->in_base_pos = 0; |
| 1228 | con->in_front_crc = con->in_middle_crc = con->in_data_crc = 0; |
| 1229 | return 0; |
| 1230 | } |
| 1231 | |
| 1232 | |
| 1233 | static int read_partial(struct ceph_connection *con, |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1234 | int end, int size, void *object) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1235 | { |
Alex Elder | e6cee71 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1236 | while (con->in_base_pos < end) { |
| 1237 | int left = end - con->in_base_pos; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1238 | int have = size - left; |
| 1239 | int ret = ceph_tcp_recvmsg(con->sock, object + have, left); |
| 1240 | if (ret <= 0) |
| 1241 | return ret; |
| 1242 | con->in_base_pos += ret; |
| 1243 | } |
| 1244 | return 1; |
| 1245 | } |
| 1246 | |
| 1247 | |
| 1248 | /* |
| 1249 | * Read all or part of the connect-side handshake on a new connection |
| 1250 | */ |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1251 | static int read_partial_banner(struct ceph_connection *con) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1252 | { |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1253 | int size; |
| 1254 | int end; |
| 1255 | int ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1256 | |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1257 | dout("read_partial_banner %p at %d\n", con, con->in_base_pos); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1258 | |
| 1259 | /* peer's banner */ |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1260 | size = strlen(CEPH_BANNER); |
| 1261 | end = size; |
| 1262 | ret = read_partial(con, end, size, con->in_banner); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1263 | if (ret <= 0) |
| 1264 | goto out; |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1265 | |
| 1266 | size = sizeof (con->actual_peer_addr); |
| 1267 | end += size; |
| 1268 | ret = read_partial(con, end, size, &con->actual_peer_addr); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1269 | if (ret <= 0) |
| 1270 | goto out; |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1271 | |
| 1272 | size = sizeof (con->peer_addr_for_me); |
| 1273 | end += size; |
| 1274 | ret = read_partial(con, end, size, &con->peer_addr_for_me); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1275 | if (ret <= 0) |
| 1276 | goto out; |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1277 | |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1278 | out: |
| 1279 | return ret; |
| 1280 | } |
| 1281 | |
| 1282 | static int read_partial_connect(struct ceph_connection *con) |
| 1283 | { |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1284 | int size; |
| 1285 | int end; |
| 1286 | int ret; |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1287 | |
| 1288 | dout("read_partial_connect %p at %d\n", con, con->in_base_pos); |
| 1289 | |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1290 | size = sizeof (con->in_reply); |
| 1291 | end = size; |
| 1292 | ret = read_partial(con, end, size, &con->in_reply); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1293 | if (ret <= 0) |
| 1294 | goto out; |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1295 | |
| 1296 | size = le32_to_cpu(con->in_reply.authorizer_len); |
| 1297 | end += size; |
| 1298 | ret = read_partial(con, end, size, con->auth_reply_buf); |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1299 | if (ret <= 0) |
| 1300 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1301 | |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1302 | dout("read_partial_connect %p tag %d, con_seq = %u, g_seq = %u\n", |
| 1303 | con, (int)con->in_reply.tag, |
| 1304 | le32_to_cpu(con->in_reply.connect_seq), |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1305 | le32_to_cpu(con->in_reply.global_seq)); |
| 1306 | out: |
| 1307 | return ret; |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1308 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1309 | } |
| 1310 | |
| 1311 | /* |
| 1312 | * Verify the hello banner looks okay. |
| 1313 | */ |
| 1314 | static int verify_hello(struct ceph_connection *con) |
| 1315 | { |
| 1316 | if (memcmp(con->in_banner, CEPH_BANNER, strlen(CEPH_BANNER))) { |
Sage Weil | 13e38c8 | 2009-10-09 16:36:34 -0700 | [diff] [blame] | 1317 | pr_err("connect to %s got bad banner\n", |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1318 | ceph_pr_addr(&con->peer_addr.in_addr)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1319 | con->error_msg = "protocol error, bad banner"; |
| 1320 | return -1; |
| 1321 | } |
| 1322 | return 0; |
| 1323 | } |
| 1324 | |
| 1325 | static bool addr_is_blank(struct sockaddr_storage *ss) |
| 1326 | { |
| 1327 | switch (ss->ss_family) { |
| 1328 | case AF_INET: |
| 1329 | return ((struct sockaddr_in *)ss)->sin_addr.s_addr == 0; |
| 1330 | case AF_INET6: |
| 1331 | return |
| 1332 | ((struct sockaddr_in6 *)ss)->sin6_addr.s6_addr32[0] == 0 && |
| 1333 | ((struct sockaddr_in6 *)ss)->sin6_addr.s6_addr32[1] == 0 && |
| 1334 | ((struct sockaddr_in6 *)ss)->sin6_addr.s6_addr32[2] == 0 && |
| 1335 | ((struct sockaddr_in6 *)ss)->sin6_addr.s6_addr32[3] == 0; |
| 1336 | } |
| 1337 | return false; |
| 1338 | } |
| 1339 | |
| 1340 | static int addr_port(struct sockaddr_storage *ss) |
| 1341 | { |
| 1342 | switch (ss->ss_family) { |
| 1343 | case AF_INET: |
Sage Weil | f28bcfb | 2009-11-04 11:46:35 -0800 | [diff] [blame] | 1344 | return ntohs(((struct sockaddr_in *)ss)->sin_port); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1345 | case AF_INET6: |
Sage Weil | f28bcfb | 2009-11-04 11:46:35 -0800 | [diff] [blame] | 1346 | return ntohs(((struct sockaddr_in6 *)ss)->sin6_port); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1347 | } |
| 1348 | return 0; |
| 1349 | } |
| 1350 | |
| 1351 | static void addr_set_port(struct sockaddr_storage *ss, int p) |
| 1352 | { |
| 1353 | switch (ss->ss_family) { |
| 1354 | case AF_INET: |
| 1355 | ((struct sockaddr_in *)ss)->sin_port = htons(p); |
Sage Weil | a2a7960 | 2011-05-12 15:34:24 -0700 | [diff] [blame] | 1356 | break; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1357 | case AF_INET6: |
| 1358 | ((struct sockaddr_in6 *)ss)->sin6_port = htons(p); |
Sage Weil | a2a7960 | 2011-05-12 15:34:24 -0700 | [diff] [blame] | 1359 | break; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1360 | } |
| 1361 | } |
| 1362 | |
| 1363 | /* |
Noah Watkins | ee3b56f | 2011-09-23 11:48:42 -0700 | [diff] [blame] | 1364 | * Unlike other *_pton function semantics, zero indicates success. |
| 1365 | */ |
| 1366 | static int ceph_pton(const char *str, size_t len, struct sockaddr_storage *ss, |
| 1367 | char delim, const char **ipend) |
| 1368 | { |
Alex Elder | 99f0f3b | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 1369 | struct sockaddr_in *in4 = (struct sockaddr_in *) ss; |
| 1370 | struct sockaddr_in6 *in6 = (struct sockaddr_in6 *) ss; |
Noah Watkins | ee3b56f | 2011-09-23 11:48:42 -0700 | [diff] [blame] | 1371 | |
| 1372 | memset(ss, 0, sizeof(*ss)); |
| 1373 | |
| 1374 | if (in4_pton(str, len, (u8 *)&in4->sin_addr.s_addr, delim, ipend)) { |
| 1375 | ss->ss_family = AF_INET; |
| 1376 | return 0; |
| 1377 | } |
| 1378 | |
| 1379 | if (in6_pton(str, len, (u8 *)&in6->sin6_addr.s6_addr, delim, ipend)) { |
| 1380 | ss->ss_family = AF_INET6; |
| 1381 | return 0; |
| 1382 | } |
| 1383 | |
| 1384 | return -EINVAL; |
| 1385 | } |
| 1386 | |
| 1387 | /* |
| 1388 | * Extract hostname string and resolve using kernel DNS facility. |
| 1389 | */ |
| 1390 | #ifdef CONFIG_CEPH_LIB_USE_DNS_RESOLVER |
| 1391 | static int ceph_dns_resolve_name(const char *name, size_t namelen, |
| 1392 | struct sockaddr_storage *ss, char delim, const char **ipend) |
| 1393 | { |
| 1394 | const char *end, *delim_p; |
| 1395 | char *colon_p, *ip_addr = NULL; |
| 1396 | int ip_len, ret; |
| 1397 | |
| 1398 | /* |
| 1399 | * The end of the hostname occurs immediately preceding the delimiter or |
| 1400 | * the port marker (':') where the delimiter takes precedence. |
| 1401 | */ |
| 1402 | delim_p = memchr(name, delim, namelen); |
| 1403 | colon_p = memchr(name, ':', namelen); |
| 1404 | |
| 1405 | if (delim_p && colon_p) |
| 1406 | end = delim_p < colon_p ? delim_p : colon_p; |
| 1407 | else if (!delim_p && colon_p) |
| 1408 | end = colon_p; |
| 1409 | else { |
| 1410 | end = delim_p; |
| 1411 | if (!end) /* case: hostname:/ */ |
| 1412 | end = name + namelen; |
| 1413 | } |
| 1414 | |
| 1415 | if (end <= name) |
| 1416 | return -EINVAL; |
| 1417 | |
| 1418 | /* do dns_resolve upcall */ |
| 1419 | ip_len = dns_query(NULL, name, end - name, NULL, &ip_addr, NULL); |
| 1420 | if (ip_len > 0) |
| 1421 | ret = ceph_pton(ip_addr, ip_len, ss, -1, NULL); |
| 1422 | else |
| 1423 | ret = -ESRCH; |
| 1424 | |
| 1425 | kfree(ip_addr); |
| 1426 | |
| 1427 | *ipend = end; |
| 1428 | |
| 1429 | pr_info("resolve '%.*s' (ret=%d): %s\n", (int)(end - name), name, |
| 1430 | ret, ret ? "failed" : ceph_pr_addr(ss)); |
| 1431 | |
| 1432 | return ret; |
| 1433 | } |
| 1434 | #else |
| 1435 | static inline int ceph_dns_resolve_name(const char *name, size_t namelen, |
| 1436 | struct sockaddr_storage *ss, char delim, const char **ipend) |
| 1437 | { |
| 1438 | return -EINVAL; |
| 1439 | } |
| 1440 | #endif |
| 1441 | |
| 1442 | /* |
| 1443 | * Parse a server name (IP or hostname). If a valid IP address is not found |
| 1444 | * then try to extract a hostname to resolve using userspace DNS upcall. |
| 1445 | */ |
| 1446 | static int ceph_parse_server_name(const char *name, size_t namelen, |
| 1447 | struct sockaddr_storage *ss, char delim, const char **ipend) |
| 1448 | { |
| 1449 | int ret; |
| 1450 | |
| 1451 | ret = ceph_pton(name, namelen, ss, delim, ipend); |
| 1452 | if (ret) |
| 1453 | ret = ceph_dns_resolve_name(name, namelen, ss, delim, ipend); |
| 1454 | |
| 1455 | return ret; |
| 1456 | } |
| 1457 | |
| 1458 | /* |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1459 | * Parse an ip[:port] list into an addr array. Use the default |
| 1460 | * monitor port if a port isn't specified. |
| 1461 | */ |
| 1462 | int ceph_parse_ips(const char *c, const char *end, |
| 1463 | struct ceph_entity_addr *addr, |
| 1464 | int max_count, int *count) |
| 1465 | { |
Noah Watkins | ee3b56f | 2011-09-23 11:48:42 -0700 | [diff] [blame] | 1466 | int i, ret = -EINVAL; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1467 | const char *p = c; |
| 1468 | |
| 1469 | dout("parse_ips on '%.*s'\n", (int)(end-c), c); |
| 1470 | for (i = 0; i < max_count; i++) { |
| 1471 | const char *ipend; |
| 1472 | struct sockaddr_storage *ss = &addr[i].in_addr; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1473 | int port; |
Sage Weil | 39139f6 | 2010-07-08 09:54:52 -0700 | [diff] [blame] | 1474 | char delim = ','; |
| 1475 | |
| 1476 | if (*p == '[') { |
| 1477 | delim = ']'; |
| 1478 | p++; |
| 1479 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1480 | |
Noah Watkins | ee3b56f | 2011-09-23 11:48:42 -0700 | [diff] [blame] | 1481 | ret = ceph_parse_server_name(p, end - p, ss, delim, &ipend); |
| 1482 | if (ret) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1483 | goto bad; |
Noah Watkins | ee3b56f | 2011-09-23 11:48:42 -0700 | [diff] [blame] | 1484 | ret = -EINVAL; |
| 1485 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1486 | p = ipend; |
| 1487 | |
Sage Weil | 39139f6 | 2010-07-08 09:54:52 -0700 | [diff] [blame] | 1488 | if (delim == ']') { |
| 1489 | if (*p != ']') { |
| 1490 | dout("missing matching ']'\n"); |
| 1491 | goto bad; |
| 1492 | } |
| 1493 | p++; |
| 1494 | } |
| 1495 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1496 | /* port? */ |
| 1497 | if (p < end && *p == ':') { |
| 1498 | port = 0; |
| 1499 | p++; |
| 1500 | while (p < end && *p >= '0' && *p <= '9') { |
| 1501 | port = (port * 10) + (*p - '0'); |
| 1502 | p++; |
| 1503 | } |
| 1504 | if (port > 65535 || port == 0) |
| 1505 | goto bad; |
| 1506 | } else { |
| 1507 | port = CEPH_MON_PORT; |
| 1508 | } |
| 1509 | |
| 1510 | addr_set_port(ss, port); |
| 1511 | |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1512 | dout("parse_ips got %s\n", ceph_pr_addr(ss)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1513 | |
| 1514 | if (p == end) |
| 1515 | break; |
| 1516 | if (*p != ',') |
| 1517 | goto bad; |
| 1518 | p++; |
| 1519 | } |
| 1520 | |
| 1521 | if (p != end) |
| 1522 | goto bad; |
| 1523 | |
| 1524 | if (count) |
| 1525 | *count = i + 1; |
| 1526 | return 0; |
| 1527 | |
| 1528 | bad: |
Sage Weil | 39139f6 | 2010-07-08 09:54:52 -0700 | [diff] [blame] | 1529 | pr_err("parse_ips bad ip '%.*s'\n", (int)(end - c), c); |
Noah Watkins | ee3b56f | 2011-09-23 11:48:42 -0700 | [diff] [blame] | 1530 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1531 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1532 | EXPORT_SYMBOL(ceph_parse_ips); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1533 | |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1534 | static int process_banner(struct ceph_connection *con) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1535 | { |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1536 | dout("process_banner on %p\n", con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1537 | |
| 1538 | if (verify_hello(con) < 0) |
| 1539 | return -1; |
| 1540 | |
Sage Weil | 63f2d21 | 2009-11-03 15:17:56 -0800 | [diff] [blame] | 1541 | ceph_decode_addr(&con->actual_peer_addr); |
| 1542 | ceph_decode_addr(&con->peer_addr_for_me); |
| 1543 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1544 | /* |
| 1545 | * Make sure the other end is who we wanted. note that the other |
| 1546 | * end may not yet know their ip address, so if it's 0.0.0.0, give |
| 1547 | * them the benefit of the doubt. |
| 1548 | */ |
Sage Weil | 103e2d3 | 2010-01-07 16:12:36 -0800 | [diff] [blame] | 1549 | if (memcmp(&con->peer_addr, &con->actual_peer_addr, |
| 1550 | sizeof(con->peer_addr)) != 0 && |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1551 | !(addr_is_blank(&con->actual_peer_addr.in_addr) && |
| 1552 | con->actual_peer_addr.nonce == con->peer_addr.nonce)) { |
Yehuda Sadeh | cd84db6 | 2010-06-11 16:58:48 -0700 | [diff] [blame] | 1553 | pr_warning("wrong peer, want %s/%d, got %s/%d\n", |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1554 | ceph_pr_addr(&con->peer_addr.in_addr), |
Yehuda Sadeh | cd84db6 | 2010-06-11 16:58:48 -0700 | [diff] [blame] | 1555 | (int)le32_to_cpu(con->peer_addr.nonce), |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1556 | ceph_pr_addr(&con->actual_peer_addr.in_addr), |
Yehuda Sadeh | cd84db6 | 2010-06-11 16:58:48 -0700 | [diff] [blame] | 1557 | (int)le32_to_cpu(con->actual_peer_addr.nonce)); |
Sage Weil | 58bb3b3 | 2009-12-23 12:12:31 -0800 | [diff] [blame] | 1558 | con->error_msg = "wrong peer at address"; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1559 | return -1; |
| 1560 | } |
| 1561 | |
| 1562 | /* |
| 1563 | * did we learn our address? |
| 1564 | */ |
| 1565 | if (addr_is_blank(&con->msgr->inst.addr.in_addr)) { |
| 1566 | int port = addr_port(&con->msgr->inst.addr.in_addr); |
| 1567 | |
| 1568 | memcpy(&con->msgr->inst.addr.in_addr, |
| 1569 | &con->peer_addr_for_me.in_addr, |
| 1570 | sizeof(con->peer_addr_for_me.in_addr)); |
| 1571 | addr_set_port(&con->msgr->inst.addr.in_addr, port); |
Sage Weil | 63f2d21 | 2009-11-03 15:17:56 -0800 | [diff] [blame] | 1572 | encode_my_addr(con->msgr); |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1573 | dout("process_banner learned my addr is %s\n", |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1574 | ceph_pr_addr(&con->msgr->inst.addr.in_addr)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1575 | } |
| 1576 | |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1577 | return 0; |
| 1578 | } |
| 1579 | |
| 1580 | static int process_connect(struct ceph_connection *con) |
| 1581 | { |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1582 | u64 sup_feat = con->msgr->supported_features; |
| 1583 | u64 req_feat = con->msgr->required_features; |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 1584 | u64 server_feat = le64_to_cpu(con->in_reply.features); |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 1585 | int ret; |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 1586 | |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1587 | dout("process_connect on %p tag %d\n", con, (int)con->in_tag); |
| 1588 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1589 | switch (con->in_reply.tag) { |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 1590 | case CEPH_MSGR_TAG_FEATURES: |
| 1591 | pr_err("%s%lld %s feature set mismatch," |
| 1592 | " my %llx < server's %llx, missing %llx\n", |
| 1593 | ENTITY_NAME(con->peer_name), |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1594 | ceph_pr_addr(&con->peer_addr.in_addr), |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 1595 | sup_feat, server_feat, server_feat & ~sup_feat); |
| 1596 | con->error_msg = "missing required protocol features"; |
Sage Weil | 0fa6ebc | 2012-12-27 20:27:04 -0600 | [diff] [blame] | 1597 | reset_connection(con); |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 1598 | return -1; |
| 1599 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1600 | case CEPH_MSGR_TAG_BADPROTOVER: |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1601 | pr_err("%s%lld %s protocol version mismatch," |
| 1602 | " my %d != server's %d\n", |
| 1603 | ENTITY_NAME(con->peer_name), |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1604 | ceph_pr_addr(&con->peer_addr.in_addr), |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1605 | le32_to_cpu(con->out_connect.protocol_version), |
| 1606 | le32_to_cpu(con->in_reply.protocol_version)); |
| 1607 | con->error_msg = "protocol version mismatch"; |
Sage Weil | 0fa6ebc | 2012-12-27 20:27:04 -0600 | [diff] [blame] | 1608 | reset_connection(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1609 | return -1; |
| 1610 | |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1611 | case CEPH_MSGR_TAG_BADAUTHORIZER: |
| 1612 | con->auth_retry++; |
| 1613 | dout("process_connect %p got BADAUTHORIZER attempt %d\n", con, |
| 1614 | con->auth_retry); |
| 1615 | if (con->auth_retry == 2) { |
| 1616 | con->error_msg = "connect authorization failure"; |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1617 | return -1; |
| 1618 | } |
| 1619 | con->auth_retry = 1; |
Jim Schutt | 6d4221b | 2012-08-10 10:37:38 -0700 | [diff] [blame] | 1620 | con_out_kvec_reset(con); |
Alex Elder | e825a66 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 1621 | ret = prepare_write_connect(con); |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 1622 | if (ret < 0) |
| 1623 | return ret; |
Sage Weil | 63733a0 | 2010-03-15 15:47:22 -0700 | [diff] [blame] | 1624 | prepare_read_connect(con); |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1625 | break; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1626 | |
| 1627 | case CEPH_MSGR_TAG_RESETSESSION: |
| 1628 | /* |
| 1629 | * If we connected with a large connect_seq but the peer |
| 1630 | * has no record of a session with us (no connection, or |
| 1631 | * connect_seq == 0), they will send RESETSESION to indicate |
| 1632 | * that they must have reset their session, and may have |
| 1633 | * dropped messages. |
| 1634 | */ |
| 1635 | dout("process_connect got RESET peer seq %u\n", |
Sage Weil | 5bdca4e | 2012-07-10 11:53:34 -0700 | [diff] [blame] | 1636 | le32_to_cpu(con->in_reply.connect_seq)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1637 | pr_err("%s%lld %s connection reset\n", |
| 1638 | ENTITY_NAME(con->peer_name), |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1639 | ceph_pr_addr(&con->peer_addr.in_addr)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1640 | reset_connection(con); |
Jim Schutt | 6d4221b | 2012-08-10 10:37:38 -0700 | [diff] [blame] | 1641 | con_out_kvec_reset(con); |
Alex Elder | 5a0f8fd | 2012-05-16 21:51:59 -0500 | [diff] [blame] | 1642 | ret = prepare_write_connect(con); |
| 1643 | if (ret < 0) |
| 1644 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1645 | prepare_read_connect(con); |
| 1646 | |
| 1647 | /* Tell ceph about it. */ |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 1648 | mutex_unlock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1649 | pr_info("reset on %s%lld\n", ENTITY_NAME(con->peer_name)); |
| 1650 | if (con->ops->peer_reset) |
| 1651 | con->ops->peer_reset(con); |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 1652 | mutex_lock(&con->mutex); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 1653 | if (con->state != CON_STATE_NEGOTIATING) |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 1654 | return -EAGAIN; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1655 | break; |
| 1656 | |
| 1657 | case CEPH_MSGR_TAG_RETRY_SESSION: |
| 1658 | /* |
| 1659 | * If we sent a smaller connect_seq than the peer has, try |
| 1660 | * again with a larger value. |
| 1661 | */ |
Sage Weil | 5bdca4e | 2012-07-10 11:53:34 -0700 | [diff] [blame] | 1662 | dout("process_connect got RETRY_SESSION my seq %u, peer %u\n", |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1663 | le32_to_cpu(con->out_connect.connect_seq), |
Sage Weil | 5bdca4e | 2012-07-10 11:53:34 -0700 | [diff] [blame] | 1664 | le32_to_cpu(con->in_reply.connect_seq)); |
| 1665 | con->connect_seq = le32_to_cpu(con->in_reply.connect_seq); |
Jim Schutt | 6d4221b | 2012-08-10 10:37:38 -0700 | [diff] [blame] | 1666 | con_out_kvec_reset(con); |
Alex Elder | 5a0f8fd | 2012-05-16 21:51:59 -0500 | [diff] [blame] | 1667 | ret = prepare_write_connect(con); |
| 1668 | if (ret < 0) |
| 1669 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1670 | prepare_read_connect(con); |
| 1671 | break; |
| 1672 | |
| 1673 | case CEPH_MSGR_TAG_RETRY_GLOBAL: |
| 1674 | /* |
| 1675 | * If we sent a smaller global_seq than the peer has, try |
| 1676 | * again with a larger value. |
| 1677 | */ |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1678 | dout("process_connect got RETRY_GLOBAL my %u peer_gseq %u\n", |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1679 | con->peer_global_seq, |
Sage Weil | 5bdca4e | 2012-07-10 11:53:34 -0700 | [diff] [blame] | 1680 | le32_to_cpu(con->in_reply.global_seq)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1681 | get_global_seq(con->msgr, |
Sage Weil | 5bdca4e | 2012-07-10 11:53:34 -0700 | [diff] [blame] | 1682 | le32_to_cpu(con->in_reply.global_seq)); |
Jim Schutt | 6d4221b | 2012-08-10 10:37:38 -0700 | [diff] [blame] | 1683 | con_out_kvec_reset(con); |
Alex Elder | 5a0f8fd | 2012-05-16 21:51:59 -0500 | [diff] [blame] | 1684 | ret = prepare_write_connect(con); |
| 1685 | if (ret < 0) |
| 1686 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1687 | prepare_read_connect(con); |
| 1688 | break; |
| 1689 | |
| 1690 | case CEPH_MSGR_TAG_READY: |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 1691 | if (req_feat & ~server_feat) { |
| 1692 | pr_err("%s%lld %s protocol feature mismatch," |
| 1693 | " my required %llx > server's %llx, need %llx\n", |
| 1694 | ENTITY_NAME(con->peer_name), |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1695 | ceph_pr_addr(&con->peer_addr.in_addr), |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 1696 | req_feat, server_feat, req_feat & ~server_feat); |
| 1697 | con->error_msg = "missing required protocol features"; |
Sage Weil | 0fa6ebc | 2012-12-27 20:27:04 -0600 | [diff] [blame] | 1698 | reset_connection(con); |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 1699 | return -1; |
| 1700 | } |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 1701 | |
Alex Elder | 122070a | 2012-12-26 10:43:57 -0600 | [diff] [blame] | 1702 | WARN_ON(con->state != CON_STATE_NEGOTIATING); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 1703 | con->state = CON_STATE_OPEN; |
| 1704 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1705 | con->peer_global_seq = le32_to_cpu(con->in_reply.global_seq); |
| 1706 | con->connect_seq++; |
Sage Weil | aba558e | 2010-05-12 15:23:30 -0700 | [diff] [blame] | 1707 | con->peer_features = server_feat; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1708 | dout("process_connect got READY gseq %d cseq %d (%d)\n", |
| 1709 | con->peer_global_seq, |
| 1710 | le32_to_cpu(con->in_reply.connect_seq), |
| 1711 | con->connect_seq); |
| 1712 | WARN_ON(con->connect_seq != |
| 1713 | le32_to_cpu(con->in_reply.connect_seq)); |
Sage Weil | 92ac41d | 2009-12-14 14:56:56 -0800 | [diff] [blame] | 1714 | |
| 1715 | if (con->in_reply.flags & CEPH_MSG_CONNECT_LOSSY) |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 1716 | con_flag_set(con, CON_FLAG_LOSSYTX); |
Sage Weil | 92ac41d | 2009-12-14 14:56:56 -0800 | [diff] [blame] | 1717 | |
Sage Weil | 85effe1 | 2012-07-30 16:22:05 -0700 | [diff] [blame] | 1718 | con->delay = 0; /* reset backoff memory */ |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1719 | |
| 1720 | prepare_read_tag(con); |
| 1721 | break; |
| 1722 | |
| 1723 | case CEPH_MSGR_TAG_WAIT: |
| 1724 | /* |
| 1725 | * If there is a connection race (we are opening |
| 1726 | * connections to each other), one of us may just have |
| 1727 | * to WAIT. This shouldn't happen if we are the |
| 1728 | * client. |
| 1729 | */ |
Sage Weil | 0417788 | 2011-05-12 15:33:17 -0700 | [diff] [blame] | 1730 | pr_err("process_connect got WAIT as client\n"); |
| 1731 | con->error_msg = "protocol error, got WAIT as client"; |
| 1732 | return -1; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1733 | |
| 1734 | default: |
| 1735 | pr_err("connect protocol error, will retry\n"); |
| 1736 | con->error_msg = "protocol error, garbage tag during connect"; |
| 1737 | return -1; |
| 1738 | } |
| 1739 | return 0; |
| 1740 | } |
| 1741 | |
| 1742 | |
| 1743 | /* |
| 1744 | * read (part of) an ack |
| 1745 | */ |
| 1746 | static int read_partial_ack(struct ceph_connection *con) |
| 1747 | { |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1748 | int size = sizeof (con->in_temp_ack); |
| 1749 | int end = size; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1750 | |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1751 | return read_partial(con, end, size, &con->in_temp_ack); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1752 | } |
| 1753 | |
| 1754 | |
| 1755 | /* |
| 1756 | * We can finally discard anything that's been acked. |
| 1757 | */ |
| 1758 | static void process_ack(struct ceph_connection *con) |
| 1759 | { |
| 1760 | struct ceph_msg *m; |
| 1761 | u64 ack = le64_to_cpu(con->in_temp_ack); |
| 1762 | u64 seq; |
| 1763 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1764 | while (!list_empty(&con->out_sent)) { |
| 1765 | m = list_first_entry(&con->out_sent, struct ceph_msg, |
| 1766 | list_head); |
| 1767 | seq = le64_to_cpu(m->hdr.seq); |
| 1768 | if (seq > ack) |
| 1769 | break; |
| 1770 | dout("got ack for seq %llu type %d at %p\n", seq, |
| 1771 | le16_to_cpu(m->hdr.type), m); |
Sage Weil | 4cf9d54 | 2011-07-26 11:27:24 -0700 | [diff] [blame] | 1772 | m->ack_stamp = jiffies; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1773 | ceph_msg_remove(m); |
| 1774 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1775 | prepare_read_tag(con); |
| 1776 | } |
| 1777 | |
| 1778 | |
| 1779 | |
| 1780 | |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 1781 | static int read_partial_message_section(struct ceph_connection *con, |
Sage Weil | 213c99e | 2010-08-03 10:25:11 -0700 | [diff] [blame] | 1782 | struct kvec *section, |
| 1783 | unsigned int sec_len, u32 *crc) |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 1784 | { |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1785 | int ret, left; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1786 | |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 1787 | BUG_ON(!section); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1788 | |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 1789 | while (section->iov_len < sec_len) { |
| 1790 | BUG_ON(section->iov_base == NULL); |
| 1791 | left = sec_len - section->iov_len; |
| 1792 | ret = ceph_tcp_recvmsg(con->sock, (char *)section->iov_base + |
| 1793 | section->iov_len, left); |
| 1794 | if (ret <= 0) |
| 1795 | return ret; |
| 1796 | section->iov_len += ret; |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 1797 | } |
Alex Elder | fe3ad59 | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 1798 | if (section->iov_len == sec_len) |
| 1799 | *crc = crc32c(0, section->iov_base, section->iov_len); |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 1800 | |
| 1801 | return 1; |
| 1802 | } |
| 1803 | |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 1804 | static int ceph_con_in_msg_alloc(struct ceph_connection *con, int *skip); |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1805 | |
| 1806 | static int read_partial_message_pages(struct ceph_connection *con, |
| 1807 | struct page **pages, |
Eric Dumazet | 95c9617 | 2012-04-15 05:58:06 +0000 | [diff] [blame] | 1808 | unsigned int data_len, bool do_datacrc) |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1809 | { |
Alex Elder | e788182 | 2013-03-08 18:51:04 -0600 | [diff] [blame] | 1810 | struct page *page; |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1811 | void *p; |
| 1812 | int ret; |
| 1813 | int left; |
| 1814 | |
| 1815 | left = min((int)(data_len - con->in_msg_pos.data_pos), |
| 1816 | (int)(PAGE_SIZE - con->in_msg_pos.page_pos)); |
| 1817 | /* (page) data */ |
| 1818 | BUG_ON(pages == NULL); |
Alex Elder | e788182 | 2013-03-08 18:51:04 -0600 | [diff] [blame] | 1819 | page = pages[con->in_msg_pos.page]; |
| 1820 | p = kmap(page); |
| 1821 | ret = ceph_tcp_recvmsg(con->sock, p + con->in_msg_pos.page_pos, left); |
Alex Elder | bca064d | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 1822 | if (ret > 0 && do_datacrc) |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1823 | con->in_data_crc = |
| 1824 | crc32c(con->in_data_crc, |
| 1825 | p + con->in_msg_pos.page_pos, ret); |
Alex Elder | e788182 | 2013-03-08 18:51:04 -0600 | [diff] [blame] | 1826 | kunmap(page); |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1827 | if (ret <= 0) |
| 1828 | return ret; |
Alex Elder | e788182 | 2013-03-08 18:51:04 -0600 | [diff] [blame] | 1829 | |
| 1830 | in_msg_pos_next(con, left, ret); |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1831 | |
| 1832 | return ret; |
| 1833 | } |
| 1834 | |
| 1835 | #ifdef CONFIG_BLOCK |
| 1836 | static int read_partial_message_bio(struct ceph_connection *con, |
Eric Dumazet | 95c9617 | 2012-04-15 05:58:06 +0000 | [diff] [blame] | 1837 | unsigned int data_len, bool do_datacrc) |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1838 | { |
Alex Elder | b3d56fa | 2013-03-08 18:51:03 -0600 | [diff] [blame] | 1839 | struct ceph_msg *msg = con->in_msg; |
| 1840 | struct bio_vec *bv; |
Alex Elder | e788182 | 2013-03-08 18:51:04 -0600 | [diff] [blame] | 1841 | struct page *page; |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1842 | void *p; |
| 1843 | int ret, left; |
| 1844 | |
Alex Elder | b3d56fa | 2013-03-08 18:51:03 -0600 | [diff] [blame] | 1845 | BUG_ON(!msg); |
| 1846 | BUG_ON(!msg->bio_iter); |
| 1847 | bv = bio_iovec_idx(msg->bio_iter, msg->bio_seg); |
Alex Elder | e788182 | 2013-03-08 18:51:04 -0600 | [diff] [blame] | 1848 | |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1849 | left = min((int)(data_len - con->in_msg_pos.data_pos), |
| 1850 | (int)(bv->bv_len - con->in_msg_pos.page_pos)); |
| 1851 | |
Alex Elder | e788182 | 2013-03-08 18:51:04 -0600 | [diff] [blame] | 1852 | page = bv->bv_page; |
| 1853 | p = kmap(page) + bv->bv_offset; |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1854 | |
Alex Elder | e788182 | 2013-03-08 18:51:04 -0600 | [diff] [blame] | 1855 | ret = ceph_tcp_recvmsg(con->sock, p + con->in_msg_pos.page_pos, left); |
Alex Elder | bca064d | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 1856 | if (ret > 0 && do_datacrc) |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1857 | con->in_data_crc = |
| 1858 | crc32c(con->in_data_crc, |
| 1859 | p + con->in_msg_pos.page_pos, ret); |
Alex Elder | e788182 | 2013-03-08 18:51:04 -0600 | [diff] [blame] | 1860 | kunmap(page); |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1861 | if (ret <= 0) |
| 1862 | return ret; |
Alex Elder | e788182 | 2013-03-08 18:51:04 -0600 | [diff] [blame] | 1863 | |
| 1864 | in_msg_pos_next(con, left, ret); |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1865 | |
| 1866 | return ret; |
| 1867 | } |
| 1868 | #endif |
| 1869 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1870 | /* |
| 1871 | * read (part of) a message. |
| 1872 | */ |
| 1873 | static int read_partial_message(struct ceph_connection *con) |
| 1874 | { |
| 1875 | struct ceph_msg *m = con->in_msg; |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1876 | int size; |
| 1877 | int end; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1878 | int ret; |
Eric Dumazet | 95c9617 | 2012-04-15 05:58:06 +0000 | [diff] [blame] | 1879 | unsigned int front_len, middle_len, data_len; |
Alex Elder | 37675b0 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 1880 | bool do_datacrc = !con->msgr->nocrc; |
Sage Weil | ae18756 | 2010-04-22 07:47:01 -0700 | [diff] [blame] | 1881 | u64 seq; |
Alex Elder | fe3ad59 | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 1882 | u32 crc; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1883 | |
| 1884 | dout("read_partial_message con %p msg %p\n", con, m); |
| 1885 | |
| 1886 | /* header */ |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1887 | size = sizeof (con->in_hdr); |
| 1888 | end = size; |
| 1889 | ret = read_partial(con, end, size, &con->in_hdr); |
Alex Elder | 57dac9d | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1890 | if (ret <= 0) |
| 1891 | return ret; |
Alex Elder | fe3ad59 | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 1892 | |
| 1893 | crc = crc32c(0, &con->in_hdr, offsetof(struct ceph_msg_header, crc)); |
| 1894 | if (cpu_to_le32(crc) != con->in_hdr.crc) { |
| 1895 | pr_err("read_partial_message bad hdr " |
| 1896 | " crc %u != expected %u\n", |
| 1897 | crc, con->in_hdr.crc); |
| 1898 | return -EBADMSG; |
| 1899 | } |
| 1900 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1901 | front_len = le32_to_cpu(con->in_hdr.front_len); |
| 1902 | if (front_len > CEPH_MSG_MAX_FRONT_LEN) |
| 1903 | return -EIO; |
| 1904 | middle_len = le32_to_cpu(con->in_hdr.middle_len); |
Alex Elder | 7b11ba37 | 2013-03-08 18:51:03 -0600 | [diff] [blame] | 1905 | if (middle_len > CEPH_MSG_MAX_MIDDLE_LEN) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1906 | return -EIO; |
| 1907 | data_len = le32_to_cpu(con->in_hdr.data_len); |
| 1908 | if (data_len > CEPH_MSG_MAX_DATA_LEN) |
| 1909 | return -EIO; |
| 1910 | |
Sage Weil | ae18756 | 2010-04-22 07:47:01 -0700 | [diff] [blame] | 1911 | /* verify seq# */ |
| 1912 | seq = le64_to_cpu(con->in_hdr.seq); |
| 1913 | if ((s64)seq - (s64)con->in_seq < 1) { |
Sage Weil | df9f86f | 2010-11-01 15:49:23 -0700 | [diff] [blame] | 1914 | pr_info("skipping %s%lld %s seq %lld expected %lld\n", |
Sage Weil | ae18756 | 2010-04-22 07:47:01 -0700 | [diff] [blame] | 1915 | ENTITY_NAME(con->peer_name), |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1916 | ceph_pr_addr(&con->peer_addr.in_addr), |
Sage Weil | ae18756 | 2010-04-22 07:47:01 -0700 | [diff] [blame] | 1917 | seq, con->in_seq + 1); |
| 1918 | con->in_base_pos = -front_len - middle_len - data_len - |
| 1919 | sizeof(m->footer); |
| 1920 | con->in_tag = CEPH_MSGR_TAG_READY; |
Sage Weil | ae18756 | 2010-04-22 07:47:01 -0700 | [diff] [blame] | 1921 | return 0; |
| 1922 | } else if ((s64)seq - (s64)con->in_seq > 1) { |
| 1923 | pr_err("read_partial_message bad seq %lld expected %lld\n", |
| 1924 | seq, con->in_seq + 1); |
| 1925 | con->error_msg = "bad message sequence # for incoming message"; |
| 1926 | return -EBADMSG; |
| 1927 | } |
| 1928 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1929 | /* allocate message? */ |
| 1930 | if (!con->in_msg) { |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 1931 | int skip = 0; |
| 1932 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1933 | dout("got hdr type %d front %d data %d\n", con->in_hdr.type, |
Alex Elder | 6ebc8b3 | 2013-03-08 18:51:03 -0600 | [diff] [blame] | 1934 | front_len, data_len); |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 1935 | ret = ceph_con_in_msg_alloc(con, &skip); |
| 1936 | if (ret < 0) |
| 1937 | return ret; |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 1938 | if (skip) { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1939 | /* skip this message */ |
Sage Weil | a79832f | 2010-04-01 16:06:19 -0700 | [diff] [blame] | 1940 | dout("alloc_msg said skip message\n"); |
Sage Weil | ae32be3 | 2010-06-13 10:30:19 -0700 | [diff] [blame] | 1941 | BUG_ON(con->in_msg); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1942 | con->in_base_pos = -front_len - middle_len - data_len - |
| 1943 | sizeof(m->footer); |
| 1944 | con->in_tag = CEPH_MSGR_TAG_READY; |
Sage Weil | 684be25 | 2010-04-21 20:45:59 -0700 | [diff] [blame] | 1945 | con->in_seq++; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1946 | return 0; |
| 1947 | } |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 1948 | |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 1949 | BUG_ON(!con->in_msg); |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 1950 | BUG_ON(con->in_msg->con != con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1951 | m = con->in_msg; |
| 1952 | m->front.iov_len = 0; /* haven't read it yet */ |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 1953 | if (m->middle) |
| 1954 | m->middle->vec.iov_len = 0; |
Yehuda Sadeh | 9d7f0f1 | 2010-01-11 10:32:02 -0800 | [diff] [blame] | 1955 | |
| 1956 | con->in_msg_pos.page = 0; |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1957 | if (m->pages) |
Sage Weil | c5c6b19 | 2010-11-09 12:40:00 -0800 | [diff] [blame] | 1958 | con->in_msg_pos.page_pos = m->page_alignment; |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1959 | else |
| 1960 | con->in_msg_pos.page_pos = 0; |
Yehuda Sadeh | 9d7f0f1 | 2010-01-11 10:32:02 -0800 | [diff] [blame] | 1961 | con->in_msg_pos.data_pos = 0; |
Sage Weil | a410702 | 2012-07-30 16:20:25 -0700 | [diff] [blame] | 1962 | |
| 1963 | #ifdef CONFIG_BLOCK |
| 1964 | if (m->bio) |
| 1965 | init_bio_iter(m->bio, &m->bio_iter, &m->bio_seg); |
| 1966 | #endif |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1967 | } |
| 1968 | |
| 1969 | /* front */ |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 1970 | ret = read_partial_message_section(con, &m->front, front_len, |
| 1971 | &con->in_front_crc); |
| 1972 | if (ret <= 0) |
| 1973 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1974 | |
| 1975 | /* middle */ |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 1976 | if (m->middle) { |
Sage Weil | 213c99e | 2010-08-03 10:25:11 -0700 | [diff] [blame] | 1977 | ret = read_partial_message_section(con, &m->middle->vec, |
| 1978 | middle_len, |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 1979 | &con->in_middle_crc); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1980 | if (ret <= 0) |
| 1981 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1982 | } |
| 1983 | |
| 1984 | /* (page) data */ |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1985 | while (con->in_msg_pos.data_pos < data_len) { |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1986 | if (m->pages) { |
| 1987 | ret = read_partial_message_pages(con, m->pages, |
Alex Elder | bca064d | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 1988 | data_len, do_datacrc); |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1989 | if (ret <= 0) |
| 1990 | return ret; |
| 1991 | #ifdef CONFIG_BLOCK |
| 1992 | } else if (m->bio) { |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1993 | ret = read_partial_message_bio(con, |
Alex Elder | bca064d | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 1994 | data_len, do_datacrc); |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1995 | if (ret <= 0) |
| 1996 | return ret; |
| 1997 | #endif |
| 1998 | } else { |
| 1999 | BUG_ON(1); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2000 | } |
| 2001 | } |
| 2002 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2003 | /* footer */ |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 2004 | size = sizeof (m->footer); |
| 2005 | end += size; |
| 2006 | ret = read_partial(con, end, size, &m->footer); |
Alex Elder | 57dac9d | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 2007 | if (ret <= 0) |
| 2008 | return ret; |
| 2009 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2010 | dout("read_partial_message got msg %p %d (%u) + %d (%u) + %d (%u)\n", |
| 2011 | m, front_len, m->footer.front_crc, middle_len, |
| 2012 | m->footer.middle_crc, data_len, m->footer.data_crc); |
| 2013 | |
| 2014 | /* crc ok? */ |
| 2015 | if (con->in_front_crc != le32_to_cpu(m->footer.front_crc)) { |
| 2016 | pr_err("read_partial_message %p front crc %u != exp. %u\n", |
| 2017 | m, con->in_front_crc, m->footer.front_crc); |
| 2018 | return -EBADMSG; |
| 2019 | } |
| 2020 | if (con->in_middle_crc != le32_to_cpu(m->footer.middle_crc)) { |
| 2021 | pr_err("read_partial_message %p middle crc %u != exp %u\n", |
| 2022 | m, con->in_middle_crc, m->footer.middle_crc); |
| 2023 | return -EBADMSG; |
| 2024 | } |
Alex Elder | bca064d | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 2025 | if (do_datacrc && |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2026 | (m->footer.flags & CEPH_MSG_FOOTER_NOCRC) == 0 && |
| 2027 | con->in_data_crc != le32_to_cpu(m->footer.data_crc)) { |
| 2028 | pr_err("read_partial_message %p data crc %u != exp. %u\n", m, |
| 2029 | con->in_data_crc, le32_to_cpu(m->footer.data_crc)); |
| 2030 | return -EBADMSG; |
| 2031 | } |
| 2032 | |
| 2033 | return 1; /* done! */ |
| 2034 | } |
| 2035 | |
| 2036 | /* |
| 2037 | * Process message. This happens in the worker thread. The callback should |
| 2038 | * be careful not to do anything that waits on other incoming messages or it |
| 2039 | * may deadlock. |
| 2040 | */ |
| 2041 | static void process_message(struct ceph_connection *con) |
| 2042 | { |
Sage Weil | 5e095e8 | 2009-12-14 14:30:34 -0800 | [diff] [blame] | 2043 | struct ceph_msg *msg; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2044 | |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2045 | BUG_ON(con->in_msg->con != con); |
| 2046 | con->in_msg->con = NULL; |
Sage Weil | 5e095e8 | 2009-12-14 14:30:34 -0800 | [diff] [blame] | 2047 | msg = con->in_msg; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2048 | con->in_msg = NULL; |
Sage Weil | 36eb71a | 2012-06-21 12:47:08 -0700 | [diff] [blame] | 2049 | con->ops->put(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2050 | |
| 2051 | /* if first message, set peer_name */ |
| 2052 | if (con->peer_name.type == 0) |
Sage Weil | dbad185 | 2010-03-25 15:45:38 -0700 | [diff] [blame] | 2053 | con->peer_name = msg->hdr.src; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2054 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2055 | con->in_seq++; |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 2056 | mutex_unlock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2057 | |
| 2058 | dout("===== %p %llu from %s%lld %d=%s len %d+%d (%u %u %u) =====\n", |
| 2059 | msg, le64_to_cpu(msg->hdr.seq), |
Sage Weil | dbad185 | 2010-03-25 15:45:38 -0700 | [diff] [blame] | 2060 | ENTITY_NAME(msg->hdr.src), |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2061 | le16_to_cpu(msg->hdr.type), |
| 2062 | ceph_msg_type_name(le16_to_cpu(msg->hdr.type)), |
| 2063 | le32_to_cpu(msg->hdr.front_len), |
| 2064 | le32_to_cpu(msg->hdr.data_len), |
| 2065 | con->in_front_crc, con->in_middle_crc, con->in_data_crc); |
| 2066 | con->ops->dispatch(con, msg); |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 2067 | |
| 2068 | mutex_lock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2069 | } |
| 2070 | |
| 2071 | |
| 2072 | /* |
| 2073 | * Write something to the socket. Called in a worker thread when the |
| 2074 | * socket appears to be writeable and we have something ready to send. |
| 2075 | */ |
| 2076 | static int try_write(struct ceph_connection *con) |
| 2077 | { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2078 | int ret = 1; |
| 2079 | |
Sage Weil | d59315c | 2012-06-21 12:49:23 -0700 | [diff] [blame] | 2080 | dout("try_write start %p state %lu\n", con, con->state); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2081 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2082 | more: |
| 2083 | dout("try_write out_kvec_bytes %d\n", con->out_kvec_bytes); |
| 2084 | |
| 2085 | /* open the socket first? */ |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2086 | if (con->state == CON_STATE_PREOPEN) { |
| 2087 | BUG_ON(con->sock); |
| 2088 | con->state = CON_STATE_CONNECTING; |
Alex Elder | a5988c4 | 2012-05-29 11:04:58 -0500 | [diff] [blame] | 2089 | |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 2090 | con_out_kvec_reset(con); |
Alex Elder | e825a66 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 2091 | prepare_write_banner(con); |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 2092 | prepare_read_banner(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2093 | |
Sage Weil | cf3e5c4 | 2009-12-11 09:48:05 -0800 | [diff] [blame] | 2094 | BUG_ON(con->in_msg); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2095 | con->in_tag = CEPH_MSGR_TAG_READY; |
| 2096 | dout("try_write initiating connect on %p new state %lu\n", |
| 2097 | con, con->state); |
Alex Elder | 41617d0 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 2098 | ret = ceph_tcp_connect(con); |
| 2099 | if (ret < 0) { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2100 | con->error_msg = "connect error"; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2101 | goto out; |
| 2102 | } |
| 2103 | } |
| 2104 | |
| 2105 | more_kvec: |
| 2106 | /* kvec data queued? */ |
| 2107 | if (con->out_skip) { |
| 2108 | ret = write_partial_skip(con); |
| 2109 | if (ret <= 0) |
Sage Weil | 42961d2 | 2011-01-25 08:19:34 -0800 | [diff] [blame] | 2110 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2111 | } |
| 2112 | if (con->out_kvec_left) { |
| 2113 | ret = write_partial_kvec(con); |
| 2114 | if (ret <= 0) |
Sage Weil | 42961d2 | 2011-01-25 08:19:34 -0800 | [diff] [blame] | 2115 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2116 | } |
| 2117 | |
| 2118 | /* msg pages? */ |
| 2119 | if (con->out_msg) { |
Sage Weil | c86a293 | 2009-12-14 14:04:30 -0800 | [diff] [blame] | 2120 | if (con->out_msg_done) { |
| 2121 | ceph_msg_put(con->out_msg); |
| 2122 | con->out_msg = NULL; /* we're done with this one */ |
| 2123 | goto do_next; |
| 2124 | } |
| 2125 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2126 | ret = write_partial_msg_pages(con); |
| 2127 | if (ret == 1) |
| 2128 | goto more_kvec; /* we need to send the footer, too! */ |
| 2129 | if (ret == 0) |
Sage Weil | 42961d2 | 2011-01-25 08:19:34 -0800 | [diff] [blame] | 2130 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2131 | if (ret < 0) { |
| 2132 | dout("try_write write_partial_msg_pages err %d\n", |
| 2133 | ret); |
Sage Weil | 42961d2 | 2011-01-25 08:19:34 -0800 | [diff] [blame] | 2134 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2135 | } |
| 2136 | } |
| 2137 | |
Sage Weil | c86a293 | 2009-12-14 14:04:30 -0800 | [diff] [blame] | 2138 | do_next: |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2139 | if (con->state == CON_STATE_OPEN) { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2140 | /* is anything else pending? */ |
| 2141 | if (!list_empty(&con->out_queue)) { |
| 2142 | prepare_write_message(con); |
| 2143 | goto more; |
| 2144 | } |
| 2145 | if (con->in_seq > con->in_seq_acked) { |
| 2146 | prepare_write_ack(con); |
| 2147 | goto more; |
| 2148 | } |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2149 | if (con_flag_test_and_clear(con, CON_FLAG_KEEPALIVE_PENDING)) { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2150 | prepare_write_keepalive(con); |
| 2151 | goto more; |
| 2152 | } |
| 2153 | } |
| 2154 | |
| 2155 | /* Nothing to do! */ |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2156 | con_flag_clear(con, CON_FLAG_WRITE_PENDING); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2157 | dout("try_write nothing else to write.\n"); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2158 | ret = 0; |
| 2159 | out: |
Sage Weil | 42961d2 | 2011-01-25 08:19:34 -0800 | [diff] [blame] | 2160 | dout("try_write done on %p ret %d\n", con, ret); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2161 | return ret; |
| 2162 | } |
| 2163 | |
| 2164 | |
| 2165 | |
| 2166 | /* |
| 2167 | * Read what we can from the socket. |
| 2168 | */ |
| 2169 | static int try_read(struct ceph_connection *con) |
| 2170 | { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2171 | int ret = -1; |
| 2172 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2173 | more: |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2174 | dout("try_read start on %p state %lu\n", con, con->state); |
| 2175 | if (con->state != CON_STATE_CONNECTING && |
| 2176 | con->state != CON_STATE_NEGOTIATING && |
| 2177 | con->state != CON_STATE_OPEN) |
| 2178 | return 0; |
| 2179 | |
| 2180 | BUG_ON(!con->sock); |
| 2181 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2182 | dout("try_read tag %d in_base_pos %d\n", (int)con->in_tag, |
| 2183 | con->in_base_pos); |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 2184 | |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2185 | if (con->state == CON_STATE_CONNECTING) { |
Alex Elder | 7593af9 | 2012-05-24 11:55:03 -0500 | [diff] [blame] | 2186 | dout("try_read connecting\n"); |
| 2187 | ret = read_partial_banner(con); |
| 2188 | if (ret <= 0) |
Alex Elder | ab166d5 | 2012-05-31 11:37:29 -0500 | [diff] [blame] | 2189 | goto out; |
Alex Elder | 7593af9 | 2012-05-24 11:55:03 -0500 | [diff] [blame] | 2190 | ret = process_banner(con); |
| 2191 | if (ret < 0) |
| 2192 | goto out; |
| 2193 | |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2194 | con->state = CON_STATE_NEGOTIATING; |
Alex Elder | 7593af9 | 2012-05-24 11:55:03 -0500 | [diff] [blame] | 2195 | |
Jim Schutt | 6d4221b | 2012-08-10 10:37:38 -0700 | [diff] [blame] | 2196 | /* |
| 2197 | * Received banner is good, exchange connection info. |
| 2198 | * Do not reset out_kvec, as sending our banner raced |
| 2199 | * with receiving peer banner after connect completed. |
| 2200 | */ |
Alex Elder | 7593af9 | 2012-05-24 11:55:03 -0500 | [diff] [blame] | 2201 | ret = prepare_write_connect(con); |
| 2202 | if (ret < 0) |
| 2203 | goto out; |
| 2204 | prepare_read_connect(con); |
| 2205 | |
| 2206 | /* Send connection info before awaiting response */ |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 2207 | goto out; |
| 2208 | } |
| 2209 | |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2210 | if (con->state == CON_STATE_NEGOTIATING) { |
Alex Elder | 7593af9 | 2012-05-24 11:55:03 -0500 | [diff] [blame] | 2211 | dout("try_read negotiating\n"); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2212 | ret = read_partial_connect(con); |
| 2213 | if (ret <= 0) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2214 | goto out; |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2215 | ret = process_connect(con); |
| 2216 | if (ret < 0) |
| 2217 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2218 | goto more; |
| 2219 | } |
| 2220 | |
Alex Elder | 122070a | 2012-12-26 10:43:57 -0600 | [diff] [blame] | 2221 | WARN_ON(con->state != CON_STATE_OPEN); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2222 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2223 | if (con->in_base_pos < 0) { |
| 2224 | /* |
| 2225 | * skipping + discarding content. |
| 2226 | * |
| 2227 | * FIXME: there must be a better way to do this! |
| 2228 | */ |
Alex Elder | 84495f4 | 2012-02-15 07:43:55 -0600 | [diff] [blame] | 2229 | static char buf[SKIP_BUF_SIZE]; |
| 2230 | int skip = min((int) sizeof (buf), -con->in_base_pos); |
| 2231 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2232 | dout("skipping %d / %d bytes\n", skip, -con->in_base_pos); |
| 2233 | ret = ceph_tcp_recvmsg(con->sock, buf, skip); |
| 2234 | if (ret <= 0) |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2235 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2236 | con->in_base_pos += ret; |
| 2237 | if (con->in_base_pos) |
| 2238 | goto more; |
| 2239 | } |
| 2240 | if (con->in_tag == CEPH_MSGR_TAG_READY) { |
| 2241 | /* |
| 2242 | * what's next? |
| 2243 | */ |
| 2244 | ret = ceph_tcp_recvmsg(con->sock, &con->in_tag, 1); |
| 2245 | if (ret <= 0) |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2246 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2247 | dout("try_read got tag %d\n", (int)con->in_tag); |
| 2248 | switch (con->in_tag) { |
| 2249 | case CEPH_MSGR_TAG_MSG: |
| 2250 | prepare_read_message(con); |
| 2251 | break; |
| 2252 | case CEPH_MSGR_TAG_ACK: |
| 2253 | prepare_read_ack(con); |
| 2254 | break; |
| 2255 | case CEPH_MSGR_TAG_CLOSE: |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2256 | con_close_socket(con); |
| 2257 | con->state = CON_STATE_CLOSED; |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2258 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2259 | default: |
| 2260 | goto bad_tag; |
| 2261 | } |
| 2262 | } |
| 2263 | if (con->in_tag == CEPH_MSGR_TAG_MSG) { |
| 2264 | ret = read_partial_message(con); |
| 2265 | if (ret <= 0) { |
| 2266 | switch (ret) { |
| 2267 | case -EBADMSG: |
| 2268 | con->error_msg = "bad crc"; |
| 2269 | ret = -EIO; |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2270 | break; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2271 | case -EIO: |
| 2272 | con->error_msg = "io error"; |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2273 | break; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2274 | } |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2275 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2276 | } |
| 2277 | if (con->in_tag == CEPH_MSGR_TAG_READY) |
| 2278 | goto more; |
| 2279 | process_message(con); |
Sage Weil | 7b862e0 | 2012-07-30 18:16:56 -0700 | [diff] [blame] | 2280 | if (con->state == CON_STATE_OPEN) |
| 2281 | prepare_read_tag(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2282 | goto more; |
| 2283 | } |
| 2284 | if (con->in_tag == CEPH_MSGR_TAG_ACK) { |
| 2285 | ret = read_partial_ack(con); |
| 2286 | if (ret <= 0) |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2287 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2288 | process_ack(con); |
| 2289 | goto more; |
| 2290 | } |
| 2291 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2292 | out: |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2293 | dout("try_read done on %p ret %d\n", con, ret); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2294 | return ret; |
| 2295 | |
| 2296 | bad_tag: |
| 2297 | pr_err("try_read bad con->in_tag = %d\n", (int)con->in_tag); |
| 2298 | con->error_msg = "protocol error, garbage tag"; |
| 2299 | ret = -1; |
| 2300 | goto out; |
| 2301 | } |
| 2302 | |
| 2303 | |
| 2304 | /* |
Alex Elder | 802c6d9 | 2012-10-08 20:37:30 -0700 | [diff] [blame] | 2305 | * Atomically queue work on a connection after the specified delay. |
| 2306 | * Bump @con reference to avoid races with connection teardown. |
| 2307 | * Returns 0 if work was queued, or an error code otherwise. |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2308 | */ |
Alex Elder | 802c6d9 | 2012-10-08 20:37:30 -0700 | [diff] [blame] | 2309 | static int queue_con_delay(struct ceph_connection *con, unsigned long delay) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2310 | { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2311 | if (!con->ops->get(con)) { |
Alex Elder | 802c6d9 | 2012-10-08 20:37:30 -0700 | [diff] [blame] | 2312 | dout("%s %p ref count 0\n", __func__, con); |
| 2313 | |
| 2314 | return -ENOENT; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2315 | } |
| 2316 | |
Alex Elder | 802c6d9 | 2012-10-08 20:37:30 -0700 | [diff] [blame] | 2317 | if (!queue_delayed_work(ceph_msgr_wq, &con->work, delay)) { |
| 2318 | dout("%s %p - already queued\n", __func__, con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2319 | con->ops->put(con); |
Alex Elder | 802c6d9 | 2012-10-08 20:37:30 -0700 | [diff] [blame] | 2320 | |
| 2321 | return -EBUSY; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2322 | } |
Alex Elder | 802c6d9 | 2012-10-08 20:37:30 -0700 | [diff] [blame] | 2323 | |
| 2324 | dout("%s %p %lu\n", __func__, con, delay); |
| 2325 | |
| 2326 | return 0; |
| 2327 | } |
| 2328 | |
| 2329 | static void queue_con(struct ceph_connection *con) |
| 2330 | { |
| 2331 | (void) queue_con_delay(con, 0); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2332 | } |
| 2333 | |
Alex Elder | 7bb21d68 | 2012-12-07 19:50:07 -0600 | [diff] [blame] | 2334 | static bool con_sock_closed(struct ceph_connection *con) |
| 2335 | { |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2336 | if (!con_flag_test_and_clear(con, CON_FLAG_SOCK_CLOSED)) |
Alex Elder | 7bb21d68 | 2012-12-07 19:50:07 -0600 | [diff] [blame] | 2337 | return false; |
| 2338 | |
| 2339 | #define CASE(x) \ |
| 2340 | case CON_STATE_ ## x: \ |
| 2341 | con->error_msg = "socket closed (con state " #x ")"; \ |
| 2342 | break; |
| 2343 | |
| 2344 | switch (con->state) { |
| 2345 | CASE(CLOSED); |
| 2346 | CASE(PREOPEN); |
| 2347 | CASE(CONNECTING); |
| 2348 | CASE(NEGOTIATING); |
| 2349 | CASE(OPEN); |
| 2350 | CASE(STANDBY); |
| 2351 | default: |
| 2352 | pr_warning("%s con %p unrecognized state %lu\n", |
| 2353 | __func__, con, con->state); |
| 2354 | con->error_msg = "unrecognized con state"; |
| 2355 | BUG(); |
| 2356 | break; |
| 2357 | } |
| 2358 | #undef CASE |
| 2359 | |
| 2360 | return true; |
| 2361 | } |
| 2362 | |
Alex Elder | f20a39f | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2363 | static bool con_backoff(struct ceph_connection *con) |
| 2364 | { |
| 2365 | int ret; |
| 2366 | |
| 2367 | if (!con_flag_test_and_clear(con, CON_FLAG_BACKOFF)) |
| 2368 | return false; |
| 2369 | |
| 2370 | ret = queue_con_delay(con, round_jiffies_relative(con->delay)); |
| 2371 | if (ret) { |
| 2372 | dout("%s: con %p FAILED to back off %lu\n", __func__, |
| 2373 | con, con->delay); |
| 2374 | BUG_ON(ret == -ENOENT); |
| 2375 | con_flag_set(con, CON_FLAG_BACKOFF); |
| 2376 | } |
| 2377 | |
| 2378 | return true; |
| 2379 | } |
| 2380 | |
Alex Elder | 9320926 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2381 | /* Finish fault handling; con->mutex must *not* be held here */ |
| 2382 | |
| 2383 | static void con_fault_finish(struct ceph_connection *con) |
| 2384 | { |
| 2385 | /* |
| 2386 | * in case we faulted due to authentication, invalidate our |
| 2387 | * current tickets so that we can get new ones. |
| 2388 | */ |
| 2389 | if (con->auth_retry && con->ops->invalidate_authorizer) { |
| 2390 | dout("calling invalidate_authorizer()\n"); |
| 2391 | con->ops->invalidate_authorizer(con); |
| 2392 | } |
| 2393 | |
| 2394 | if (con->ops->fault) |
| 2395 | con->ops->fault(con); |
| 2396 | } |
| 2397 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2398 | /* |
| 2399 | * Do some work on a connection. Drop a connection ref when we're done. |
| 2400 | */ |
| 2401 | static void con_work(struct work_struct *work) |
| 2402 | { |
| 2403 | struct ceph_connection *con = container_of(work, struct ceph_connection, |
| 2404 | work.work); |
Alex Elder | 4965941 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2405 | bool fault; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2406 | |
Sage Weil | 9dd4658 | 2010-04-28 13:51:50 -0700 | [diff] [blame] | 2407 | mutex_lock(&con->mutex); |
Alex Elder | 4965941 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2408 | while (true) { |
| 2409 | int ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2410 | |
Alex Elder | 4965941 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2411 | if ((fault = con_sock_closed(con))) { |
| 2412 | dout("%s: con %p SOCK_CLOSED\n", __func__, con); |
| 2413 | break; |
| 2414 | } |
| 2415 | if (con_backoff(con)) { |
| 2416 | dout("%s: con %p BACKOFF\n", __func__, con); |
| 2417 | break; |
| 2418 | } |
| 2419 | if (con->state == CON_STATE_STANDBY) { |
| 2420 | dout("%s: con %p STANDBY\n", __func__, con); |
| 2421 | break; |
| 2422 | } |
| 2423 | if (con->state == CON_STATE_CLOSED) { |
| 2424 | dout("%s: con %p CLOSED\n", __func__, con); |
| 2425 | BUG_ON(con->sock); |
| 2426 | break; |
| 2427 | } |
| 2428 | if (con->state == CON_STATE_PREOPEN) { |
| 2429 | dout("%s: con %p PREOPEN\n", __func__, con); |
| 2430 | BUG_ON(con->sock); |
| 2431 | } |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 2432 | |
Alex Elder | 4965941 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2433 | ret = try_read(con); |
| 2434 | if (ret < 0) { |
| 2435 | if (ret == -EAGAIN) |
| 2436 | continue; |
| 2437 | con->error_msg = "socket error on read"; |
| 2438 | fault = true; |
| 2439 | break; |
| 2440 | } |
| 2441 | |
| 2442 | ret = try_write(con); |
| 2443 | if (ret < 0) { |
| 2444 | if (ret == -EAGAIN) |
| 2445 | continue; |
| 2446 | con->error_msg = "socket error on write"; |
| 2447 | fault = true; |
| 2448 | } |
| 2449 | |
| 2450 | break; /* If we make it to here, we're done */ |
Sage Weil | 3a140a0 | 2012-07-30 16:24:21 -0700 | [diff] [blame] | 2451 | } |
Alex Elder | b6e7b6a | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2452 | if (fault) |
| 2453 | con_fault(con); |
Sage Weil | 9dd4658 | 2010-04-28 13:51:50 -0700 | [diff] [blame] | 2454 | mutex_unlock(&con->mutex); |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 2455 | |
Alex Elder | b6e7b6a | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2456 | if (fault) |
| 2457 | con_fault_finish(con); |
| 2458 | |
| 2459 | con->ops->put(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2460 | } |
| 2461 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2462 | /* |
| 2463 | * Generic error/fault handler. A retry mechanism is used with |
| 2464 | * exponential backoff |
| 2465 | */ |
Alex Elder | 9320926 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2466 | static void con_fault(struct ceph_connection *con) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2467 | { |
Alex Elder | 2836298 | 2012-12-14 16:47:41 -0600 | [diff] [blame] | 2468 | pr_warning("%s%lld %s %s\n", ENTITY_NAME(con->peer_name), |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2469 | ceph_pr_addr(&con->peer_addr.in_addr), con->error_msg); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2470 | dout("fault %p state %lu to peer %s\n", |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2471 | con, con->state, ceph_pr_addr(&con->peer_addr.in_addr)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2472 | |
Alex Elder | 122070a | 2012-12-26 10:43:57 -0600 | [diff] [blame] | 2473 | WARN_ON(con->state != CON_STATE_CONNECTING && |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2474 | con->state != CON_STATE_NEGOTIATING && |
| 2475 | con->state != CON_STATE_OPEN); |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 2476 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2477 | con_close_socket(con); |
Sage Weil | 5e095e8 | 2009-12-14 14:30:34 -0800 | [diff] [blame] | 2478 | |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2479 | if (con_flag_test(con, CON_FLAG_LOSSYTX)) { |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2480 | dout("fault on LOSSYTX channel, marking CLOSED\n"); |
| 2481 | con->state = CON_STATE_CLOSED; |
Alex Elder | 9320926 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2482 | return; |
Sage Weil | 3b5ede0 | 2012-07-20 15:22:53 -0700 | [diff] [blame] | 2483 | } |
| 2484 | |
Sage Weil | 5e095e8 | 2009-12-14 14:30:34 -0800 | [diff] [blame] | 2485 | if (con->in_msg) { |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2486 | BUG_ON(con->in_msg->con != con); |
| 2487 | con->in_msg->con = NULL; |
Sage Weil | 5e095e8 | 2009-12-14 14:30:34 -0800 | [diff] [blame] | 2488 | ceph_msg_put(con->in_msg); |
| 2489 | con->in_msg = NULL; |
Sage Weil | 36eb71a | 2012-06-21 12:47:08 -0700 | [diff] [blame] | 2490 | con->ops->put(con); |
Sage Weil | 5e095e8 | 2009-12-14 14:30:34 -0800 | [diff] [blame] | 2491 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2492 | |
Sage Weil | e80a52d | 2010-02-25 12:40:45 -0800 | [diff] [blame] | 2493 | /* Requeue anything that hasn't been acked */ |
| 2494 | list_splice_init(&con->out_sent, &con->out_queue); |
Sage Weil | 9bd2e6f | 2010-02-02 16:21:06 -0800 | [diff] [blame] | 2495 | |
Sage Weil | e76661d | 2011-03-03 10:10:15 -0800 | [diff] [blame] | 2496 | /* If there are no messages queued or keepalive pending, place |
| 2497 | * the connection in a STANDBY state */ |
| 2498 | if (list_empty(&con->out_queue) && |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2499 | !con_flag_test(con, CON_FLAG_KEEPALIVE_PENDING)) { |
Sage Weil | e00de34 | 2011-03-04 12:25:05 -0800 | [diff] [blame] | 2500 | dout("fault %p setting STANDBY clearing WRITE_PENDING\n", con); |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2501 | con_flag_clear(con, CON_FLAG_WRITE_PENDING); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2502 | con->state = CON_STATE_STANDBY; |
Sage Weil | e80a52d | 2010-02-25 12:40:45 -0800 | [diff] [blame] | 2503 | } else { |
| 2504 | /* retry after a delay. */ |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2505 | con->state = CON_STATE_PREOPEN; |
Sage Weil | e80a52d | 2010-02-25 12:40:45 -0800 | [diff] [blame] | 2506 | if (con->delay == 0) |
| 2507 | con->delay = BASE_DELAY_INTERVAL; |
| 2508 | else if (con->delay < MAX_DELAY_INTERVAL) |
| 2509 | con->delay *= 2; |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2510 | con_flag_set(con, CON_FLAG_BACKOFF); |
Alex Elder | 8618e30 | 2012-10-08 20:37:30 -0700 | [diff] [blame] | 2511 | queue_con(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2512 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2513 | } |
| 2514 | |
| 2515 | |
| 2516 | |
| 2517 | /* |
Alex Elder | 15d9882 | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 2518 | * initialize a new messenger instance |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2519 | */ |
Alex Elder | 15d9882 | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 2520 | void ceph_messenger_init(struct ceph_messenger *msgr, |
| 2521 | struct ceph_entity_addr *myaddr, |
| 2522 | u32 supported_features, |
| 2523 | u32 required_features, |
| 2524 | bool nocrc) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2525 | { |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2526 | msgr->supported_features = supported_features; |
| 2527 | msgr->required_features = required_features; |
| 2528 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2529 | spin_lock_init(&msgr->global_seq_lock); |
| 2530 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2531 | if (myaddr) |
| 2532 | msgr->inst.addr = *myaddr; |
| 2533 | |
| 2534 | /* select a random nonce */ |
Sage Weil | ac8839d | 2010-01-27 14:28:10 -0800 | [diff] [blame] | 2535 | msgr->inst.addr.type = 0; |
Sage Weil | 103e2d3 | 2010-01-07 16:12:36 -0800 | [diff] [blame] | 2536 | get_random_bytes(&msgr->inst.addr.nonce, sizeof(msgr->inst.addr.nonce)); |
Sage Weil | 63f2d21 | 2009-11-03 15:17:56 -0800 | [diff] [blame] | 2537 | encode_my_addr(msgr); |
Alex Elder | 15d9882 | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 2538 | msgr->nocrc = nocrc; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2539 | |
Guanjun He | a2a3258 | 2012-07-08 19:50:33 -0700 | [diff] [blame] | 2540 | atomic_set(&msgr->stopping, 0); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2541 | |
Alex Elder | 15d9882 | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 2542 | dout("%s %p\n", __func__, msgr); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2543 | } |
Alex Elder | 15d9882 | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 2544 | EXPORT_SYMBOL(ceph_messenger_init); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2545 | |
Sage Weil | e00de34 | 2011-03-04 12:25:05 -0800 | [diff] [blame] | 2546 | static void clear_standby(struct ceph_connection *con) |
| 2547 | { |
| 2548 | /* come back from STANDBY? */ |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2549 | if (con->state == CON_STATE_STANDBY) { |
Sage Weil | e00de34 | 2011-03-04 12:25:05 -0800 | [diff] [blame] | 2550 | dout("clear_standby %p and ++connect_seq\n", con); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2551 | con->state = CON_STATE_PREOPEN; |
Sage Weil | e00de34 | 2011-03-04 12:25:05 -0800 | [diff] [blame] | 2552 | con->connect_seq++; |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2553 | WARN_ON(con_flag_test(con, CON_FLAG_WRITE_PENDING)); |
| 2554 | WARN_ON(con_flag_test(con, CON_FLAG_KEEPALIVE_PENDING)); |
Sage Weil | e00de34 | 2011-03-04 12:25:05 -0800 | [diff] [blame] | 2555 | } |
| 2556 | } |
| 2557 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2558 | /* |
| 2559 | * Queue up an outgoing message on the given connection. |
| 2560 | */ |
| 2561 | void ceph_con_send(struct ceph_connection *con, struct ceph_msg *msg) |
| 2562 | { |
Sage Weil | a59b55a | 2012-07-20 15:34:04 -0700 | [diff] [blame] | 2563 | /* set src+dst */ |
| 2564 | msg->hdr.src = con->msgr->inst.name; |
| 2565 | BUG_ON(msg->front.iov_len != le32_to_cpu(msg->hdr.front_len)); |
| 2566 | msg->needs_out_seq = true; |
| 2567 | |
| 2568 | mutex_lock(&con->mutex); |
| 2569 | |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2570 | if (con->state == CON_STATE_CLOSED) { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2571 | dout("con_send %p closed, dropping %p\n", con, msg); |
| 2572 | ceph_msg_put(msg); |
Sage Weil | a59b55a | 2012-07-20 15:34:04 -0700 | [diff] [blame] | 2573 | mutex_unlock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2574 | return; |
| 2575 | } |
| 2576 | |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2577 | BUG_ON(msg->con != NULL); |
Sage Weil | 36eb71a | 2012-06-21 12:47:08 -0700 | [diff] [blame] | 2578 | msg->con = con->ops->get(con); |
Alex Elder | 92ce034 | 2012-06-04 14:43:33 -0500 | [diff] [blame] | 2579 | BUG_ON(msg->con == NULL); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2580 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2581 | BUG_ON(!list_empty(&msg->list_head)); |
| 2582 | list_add_tail(&msg->list_head, &con->out_queue); |
| 2583 | dout("----- %p to %s%lld %d=%s len %d+%d+%d -----\n", msg, |
| 2584 | ENTITY_NAME(con->peer_name), le16_to_cpu(msg->hdr.type), |
| 2585 | ceph_msg_type_name(le16_to_cpu(msg->hdr.type)), |
| 2586 | le32_to_cpu(msg->hdr.front_len), |
| 2587 | le32_to_cpu(msg->hdr.middle_len), |
| 2588 | le32_to_cpu(msg->hdr.data_len)); |
Sage Weil | 0065093 | 2012-07-20 15:33:04 -0700 | [diff] [blame] | 2589 | |
| 2590 | clear_standby(con); |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 2591 | mutex_unlock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2592 | |
| 2593 | /* if there wasn't anything waiting to send before, queue |
| 2594 | * new work */ |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2595 | if (con_flag_test_and_set(con, CON_FLAG_WRITE_PENDING) == 0) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2596 | queue_con(con); |
| 2597 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2598 | EXPORT_SYMBOL(ceph_con_send); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2599 | |
| 2600 | /* |
| 2601 | * Revoke a message that was previously queued for send |
| 2602 | */ |
Alex Elder | 6740a84 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2603 | void ceph_msg_revoke(struct ceph_msg *msg) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2604 | { |
Alex Elder | 6740a84 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2605 | struct ceph_connection *con = msg->con; |
| 2606 | |
| 2607 | if (!con) |
| 2608 | return; /* Message not in our possession */ |
| 2609 | |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 2610 | mutex_lock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2611 | if (!list_empty(&msg->list_head)) { |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2612 | dout("%s %p msg %p - was on queue\n", __func__, con, msg); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2613 | list_del_init(&msg->list_head); |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2614 | BUG_ON(msg->con == NULL); |
Sage Weil | 36eb71a | 2012-06-21 12:47:08 -0700 | [diff] [blame] | 2615 | msg->con->ops->put(msg->con); |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2616 | msg->con = NULL; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2617 | msg->hdr.seq = 0; |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2618 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2619 | ceph_msg_put(msg); |
Sage Weil | ed98ada | 2010-07-05 12:15:14 -0700 | [diff] [blame] | 2620 | } |
| 2621 | if (con->out_msg == msg) { |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2622 | dout("%s %p msg %p - was sending\n", __func__, con, msg); |
Sage Weil | ed98ada | 2010-07-05 12:15:14 -0700 | [diff] [blame] | 2623 | con->out_msg = NULL; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2624 | if (con->out_kvec_is_msg) { |
| 2625 | con->out_skip = con->out_kvec_bytes; |
| 2626 | con->out_kvec_is_msg = false; |
| 2627 | } |
Sage Weil | ed98ada | 2010-07-05 12:15:14 -0700 | [diff] [blame] | 2628 | msg->hdr.seq = 0; |
Alex Elder | 92ce034 | 2012-06-04 14:43:33 -0500 | [diff] [blame] | 2629 | |
| 2630 | ceph_msg_put(msg); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2631 | } |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 2632 | mutex_unlock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2633 | } |
| 2634 | |
| 2635 | /* |
Yehuda Sadeh | 0d59ab8 | 2010-01-13 17:03:23 -0800 | [diff] [blame] | 2636 | * Revoke a message that we may be reading data into |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 2637 | */ |
Alex Elder | 8921d11 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2638 | void ceph_msg_revoke_incoming(struct ceph_msg *msg) |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 2639 | { |
Alex Elder | 8921d11 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2640 | struct ceph_connection *con; |
| 2641 | |
| 2642 | BUG_ON(msg == NULL); |
| 2643 | if (!msg->con) { |
| 2644 | dout("%s msg %p null con\n", __func__, msg); |
| 2645 | |
| 2646 | return; /* Message not in our possession */ |
| 2647 | } |
| 2648 | |
| 2649 | con = msg->con; |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 2650 | mutex_lock(&con->mutex); |
Alex Elder | 8921d11 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2651 | if (con->in_msg == msg) { |
Eric Dumazet | 95c9617 | 2012-04-15 05:58:06 +0000 | [diff] [blame] | 2652 | unsigned int front_len = le32_to_cpu(con->in_hdr.front_len); |
| 2653 | unsigned int middle_len = le32_to_cpu(con->in_hdr.middle_len); |
| 2654 | unsigned int data_len = le32_to_cpu(con->in_hdr.data_len); |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 2655 | |
| 2656 | /* skip rest of message */ |
Alex Elder | 8921d11 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2657 | dout("%s %p msg %p revoked\n", __func__, con, msg); |
| 2658 | con->in_base_pos = con->in_base_pos - |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 2659 | sizeof(struct ceph_msg_header) - |
Yehuda Sadeh | 0d59ab8 | 2010-01-13 17:03:23 -0800 | [diff] [blame] | 2660 | front_len - |
| 2661 | middle_len - |
| 2662 | data_len - |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 2663 | sizeof(struct ceph_msg_footer); |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 2664 | ceph_msg_put(con->in_msg); |
| 2665 | con->in_msg = NULL; |
| 2666 | con->in_tag = CEPH_MSGR_TAG_READY; |
Sage Weil | 684be25 | 2010-04-21 20:45:59 -0700 | [diff] [blame] | 2667 | con->in_seq++; |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 2668 | } else { |
Alex Elder | 8921d11 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2669 | dout("%s %p in_msg %p msg %p no-op\n", |
| 2670 | __func__, con, con->in_msg, msg); |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 2671 | } |
| 2672 | mutex_unlock(&con->mutex); |
| 2673 | } |
| 2674 | |
| 2675 | /* |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2676 | * Queue a keepalive byte to ensure the tcp connection is alive. |
| 2677 | */ |
| 2678 | void ceph_con_keepalive(struct ceph_connection *con) |
| 2679 | { |
Sage Weil | e00de34 | 2011-03-04 12:25:05 -0800 | [diff] [blame] | 2680 | dout("con_keepalive %p\n", con); |
Sage Weil | 0065093 | 2012-07-20 15:33:04 -0700 | [diff] [blame] | 2681 | mutex_lock(&con->mutex); |
Sage Weil | e00de34 | 2011-03-04 12:25:05 -0800 | [diff] [blame] | 2682 | clear_standby(con); |
Sage Weil | 0065093 | 2012-07-20 15:33:04 -0700 | [diff] [blame] | 2683 | mutex_unlock(&con->mutex); |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2684 | if (con_flag_test_and_set(con, CON_FLAG_KEEPALIVE_PENDING) == 0 && |
| 2685 | con_flag_test_and_set(con, CON_FLAG_WRITE_PENDING) == 0) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2686 | queue_con(con); |
| 2687 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2688 | EXPORT_SYMBOL(ceph_con_keepalive); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2689 | |
Alex Elder | 02afca6 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 2690 | void ceph_msg_data_set_pages(struct ceph_msg *msg, struct page **pages, |
Alex Elder | f1baeb2 | 2013-03-07 15:38:26 -0600 | [diff] [blame] | 2691 | size_t length, size_t alignment) |
Alex Elder | 02afca6 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 2692 | { |
Alex Elder | 07aa155 | 2013-03-04 18:29:06 -0600 | [diff] [blame^] | 2693 | BUG_ON(!pages); |
| 2694 | BUG_ON(!length); |
| 2695 | BUG_ON(msg->pages); |
| 2696 | BUG_ON(msg->length); |
Alex Elder | 02afca6 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 2697 | |
| 2698 | msg->pages = pages; |
Alex Elder | 4a73ef2 | 2013-03-07 15:38:26 -0600 | [diff] [blame] | 2699 | msg->length = length; |
Alex Elder | 02afca6 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 2700 | msg->page_alignment = alignment & ~PAGE_MASK; |
| 2701 | } |
| 2702 | EXPORT_SYMBOL(ceph_msg_data_set_pages); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2703 | |
Alex Elder | 27fa838 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 2704 | void ceph_msg_data_set_pagelist(struct ceph_msg *msg, |
| 2705 | struct ceph_pagelist *pagelist) |
| 2706 | { |
Alex Elder | 07aa155 | 2013-03-04 18:29:06 -0600 | [diff] [blame^] | 2707 | BUG_ON(!pagelist); |
| 2708 | BUG_ON(!pagelist->length); |
| 2709 | BUG_ON(msg->pagelist); |
Alex Elder | 27fa838 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 2710 | |
| 2711 | msg->pagelist = pagelist; |
| 2712 | } |
| 2713 | EXPORT_SYMBOL(ceph_msg_data_set_pagelist); |
| 2714 | |
| 2715 | void ceph_msg_data_set_bio(struct ceph_msg *msg, struct bio *bio) |
| 2716 | { |
Alex Elder | 07aa155 | 2013-03-04 18:29:06 -0600 | [diff] [blame^] | 2717 | BUG_ON(!bio); |
| 2718 | BUG_ON(msg->bio); |
Alex Elder | 27fa838 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 2719 | |
| 2720 | msg->bio = bio; |
| 2721 | } |
| 2722 | EXPORT_SYMBOL(ceph_msg_data_set_bio); |
| 2723 | |
| 2724 | void ceph_msg_data_set_trail(struct ceph_msg *msg, struct ceph_pagelist *trail) |
| 2725 | { |
Alex Elder | 07aa155 | 2013-03-04 18:29:06 -0600 | [diff] [blame^] | 2726 | BUG_ON(!trail); |
| 2727 | BUG_ON(!trail->length); |
| 2728 | BUG_ON(msg->trail); |
Alex Elder | 27fa838 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 2729 | |
| 2730 | msg->trail = trail; |
| 2731 | } |
| 2732 | EXPORT_SYMBOL(ceph_msg_data_set_trail); |
| 2733 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2734 | /* |
| 2735 | * construct a new message with given type, size |
| 2736 | * the new msg has a ref count of 1. |
| 2737 | */ |
Sage Weil | b61c276 | 2011-08-09 15:03:46 -0700 | [diff] [blame] | 2738 | struct ceph_msg *ceph_msg_new(int type, int front_len, gfp_t flags, |
| 2739 | bool can_fail) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2740 | { |
| 2741 | struct ceph_msg *m; |
| 2742 | |
Alex Elder | 9516e45 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 2743 | m = kzalloc(sizeof(*m), flags); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2744 | if (m == NULL) |
| 2745 | goto out; |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2746 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2747 | m->hdr.type = cpu_to_le16(type); |
Sage Weil | 45c6ceb | 2010-05-11 15:01:51 -0700 | [diff] [blame] | 2748 | m->hdr.priority = cpu_to_le16(CEPH_MSG_PRIO_DEFAULT); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2749 | m->hdr.front_len = cpu_to_le32(front_len); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2750 | |
Alex Elder | 9516e45 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 2751 | INIT_LIST_HEAD(&m->list_head); |
| 2752 | kref_init(&m->kref); |
Henry C Chang | ca20892 | 2011-05-03 02:29:56 +0000 | [diff] [blame] | 2753 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2754 | /* front */ |
Alex Elder | 9516e45 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 2755 | m->front_max = front_len; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2756 | if (front_len) { |
| 2757 | if (front_len > PAGE_CACHE_SIZE) { |
Yehuda Sadeh | 34d2376 | 2010-04-06 14:33:58 -0700 | [diff] [blame] | 2758 | m->front.iov_base = __vmalloc(front_len, flags, |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2759 | PAGE_KERNEL); |
| 2760 | m->front_is_vmalloc = true; |
| 2761 | } else { |
Yehuda Sadeh | 34d2376 | 2010-04-06 14:33:58 -0700 | [diff] [blame] | 2762 | m->front.iov_base = kmalloc(front_len, flags); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2763 | } |
| 2764 | if (m->front.iov_base == NULL) { |
Sage Weil | b61c276 | 2011-08-09 15:03:46 -0700 | [diff] [blame] | 2765 | dout("ceph_msg_new can't allocate %d bytes\n", |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2766 | front_len); |
| 2767 | goto out2; |
| 2768 | } |
| 2769 | } else { |
| 2770 | m->front.iov_base = NULL; |
| 2771 | } |
| 2772 | m->front.iov_len = front_len; |
| 2773 | |
Sage Weil | bb25766 | 2010-04-01 16:07:23 -0700 | [diff] [blame] | 2774 | dout("ceph_msg_new %p front %d\n", m, front_len); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2775 | return m; |
| 2776 | |
| 2777 | out2: |
| 2778 | ceph_msg_put(m); |
| 2779 | out: |
Sage Weil | b61c276 | 2011-08-09 15:03:46 -0700 | [diff] [blame] | 2780 | if (!can_fail) { |
| 2781 | pr_err("msg_new can't create type %d front %d\n", type, |
| 2782 | front_len); |
Sage Weil | f0ed1b7 | 2011-08-09 15:05:07 -0700 | [diff] [blame] | 2783 | WARN_ON(1); |
Sage Weil | b61c276 | 2011-08-09 15:03:46 -0700 | [diff] [blame] | 2784 | } else { |
| 2785 | dout("msg_new can't create type %d front %d\n", type, |
| 2786 | front_len); |
| 2787 | } |
Sage Weil | a79832f | 2010-04-01 16:06:19 -0700 | [diff] [blame] | 2788 | return NULL; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2789 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2790 | EXPORT_SYMBOL(ceph_msg_new); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2791 | |
| 2792 | /* |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2793 | * Allocate "middle" portion of a message, if it is needed and wasn't |
| 2794 | * allocated by alloc_msg. This allows us to read a small fixed-size |
| 2795 | * per-type header in the front and then gracefully fail (i.e., |
| 2796 | * propagate the error to the caller based on info in the front) when |
| 2797 | * the middle is too large. |
| 2798 | */ |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2799 | static int ceph_alloc_middle(struct ceph_connection *con, struct ceph_msg *msg) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2800 | { |
| 2801 | int type = le16_to_cpu(msg->hdr.type); |
| 2802 | int middle_len = le32_to_cpu(msg->hdr.middle_len); |
| 2803 | |
| 2804 | dout("alloc_middle %p type %d %s middle_len %d\n", msg, type, |
| 2805 | ceph_msg_type_name(type), middle_len); |
| 2806 | BUG_ON(!middle_len); |
| 2807 | BUG_ON(msg->middle); |
| 2808 | |
Sage Weil | b6c1d5b | 2009-12-07 12:17:17 -0800 | [diff] [blame] | 2809 | msg->middle = ceph_buffer_new(middle_len, GFP_NOFS); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2810 | if (!msg->middle) |
| 2811 | return -ENOMEM; |
| 2812 | return 0; |
| 2813 | } |
| 2814 | |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2815 | /* |
Alex Elder | 1c20f2d | 2012-06-04 14:43:32 -0500 | [diff] [blame] | 2816 | * Allocate a message for receiving an incoming message on a |
| 2817 | * connection, and save the result in con->in_msg. Uses the |
| 2818 | * connection's private alloc_msg op if available. |
| 2819 | * |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 2820 | * Returns 0 on success, or a negative error code. |
| 2821 | * |
| 2822 | * On success, if we set *skip = 1: |
| 2823 | * - the next message should be skipped and ignored. |
| 2824 | * - con->in_msg == NULL |
| 2825 | * or if we set *skip = 0: |
| 2826 | * - con->in_msg is non-null. |
| 2827 | * On error (ENOMEM, EAGAIN, ...), |
| 2828 | * - con->in_msg == NULL |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2829 | */ |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 2830 | static int ceph_con_in_msg_alloc(struct ceph_connection *con, int *skip) |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2831 | { |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 2832 | struct ceph_msg_header *hdr = &con->in_hdr; |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2833 | int middle_len = le32_to_cpu(hdr->middle_len); |
Alex Elder | 1d866d1 | 2013-03-01 18:00:14 -0600 | [diff] [blame] | 2834 | struct ceph_msg *msg; |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 2835 | int ret = 0; |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2836 | |
Alex Elder | 1c20f2d | 2012-06-04 14:43:32 -0500 | [diff] [blame] | 2837 | BUG_ON(con->in_msg != NULL); |
Alex Elder | 53ded49 | 2013-03-01 18:00:14 -0600 | [diff] [blame] | 2838 | BUG_ON(!con->ops->alloc_msg); |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2839 | |
Alex Elder | 53ded49 | 2013-03-01 18:00:14 -0600 | [diff] [blame] | 2840 | mutex_unlock(&con->mutex); |
| 2841 | msg = con->ops->alloc_msg(con, hdr, skip); |
| 2842 | mutex_lock(&con->mutex); |
| 2843 | if (con->state != CON_STATE_OPEN) { |
| 2844 | if (msg) |
Alex Elder | 1d866d1 | 2013-03-01 18:00:14 -0600 | [diff] [blame] | 2845 | ceph_msg_put(msg); |
Alex Elder | 53ded49 | 2013-03-01 18:00:14 -0600 | [diff] [blame] | 2846 | return -EAGAIN; |
| 2847 | } |
Alex Elder | 4137577 | 2013-03-05 09:25:10 -0600 | [diff] [blame] | 2848 | if (msg) { |
| 2849 | BUG_ON(*skip); |
| 2850 | con->in_msg = msg; |
Sage Weil | 36eb71a | 2012-06-21 12:47:08 -0700 | [diff] [blame] | 2851 | con->in_msg->con = con->ops->get(con); |
Alex Elder | 92ce034 | 2012-06-04 14:43:33 -0500 | [diff] [blame] | 2852 | BUG_ON(con->in_msg->con == NULL); |
Alex Elder | 4137577 | 2013-03-05 09:25:10 -0600 | [diff] [blame] | 2853 | } else { |
| 2854 | /* |
| 2855 | * Null message pointer means either we should skip |
| 2856 | * this message or we couldn't allocate memory. The |
| 2857 | * former is not an error. |
| 2858 | */ |
| 2859 | if (*skip) |
| 2860 | return 0; |
| 2861 | con->error_msg = "error allocating memory for incoming message"; |
| 2862 | |
Alex Elder | 53ded49 | 2013-03-01 18:00:14 -0600 | [diff] [blame] | 2863 | return -ENOMEM; |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2864 | } |
Alex Elder | 1c20f2d | 2012-06-04 14:43:32 -0500 | [diff] [blame] | 2865 | memcpy(&con->in_msg->hdr, &con->in_hdr, sizeof(con->in_hdr)); |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2866 | |
Alex Elder | 1c20f2d | 2012-06-04 14:43:32 -0500 | [diff] [blame] | 2867 | if (middle_len && !con->in_msg->middle) { |
| 2868 | ret = ceph_alloc_middle(con, con->in_msg); |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2869 | if (ret < 0) { |
Alex Elder | 1c20f2d | 2012-06-04 14:43:32 -0500 | [diff] [blame] | 2870 | ceph_msg_put(con->in_msg); |
| 2871 | con->in_msg = NULL; |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2872 | } |
| 2873 | } |
Yehuda Sadeh | 9d7f0f1 | 2010-01-11 10:32:02 -0800 | [diff] [blame] | 2874 | |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 2875 | return ret; |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2876 | } |
| 2877 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2878 | |
| 2879 | /* |
| 2880 | * Free a generically kmalloc'd message. |
| 2881 | */ |
| 2882 | void ceph_msg_kfree(struct ceph_msg *m) |
| 2883 | { |
| 2884 | dout("msg_kfree %p\n", m); |
| 2885 | if (m->front_is_vmalloc) |
| 2886 | vfree(m->front.iov_base); |
| 2887 | else |
| 2888 | kfree(m->front.iov_base); |
| 2889 | kfree(m); |
| 2890 | } |
| 2891 | |
| 2892 | /* |
| 2893 | * Drop a msg ref. Destroy as needed. |
| 2894 | */ |
Sage Weil | c2e552e | 2009-12-07 15:55:05 -0800 | [diff] [blame] | 2895 | void ceph_msg_last_put(struct kref *kref) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2896 | { |
Sage Weil | c2e552e | 2009-12-07 15:55:05 -0800 | [diff] [blame] | 2897 | struct ceph_msg *m = container_of(kref, struct ceph_msg, kref); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2898 | |
Sage Weil | c2e552e | 2009-12-07 15:55:05 -0800 | [diff] [blame] | 2899 | dout("ceph_msg_put last one on %p\n", m); |
| 2900 | WARN_ON(!list_empty(&m->list_head)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2901 | |
Sage Weil | c2e552e | 2009-12-07 15:55:05 -0800 | [diff] [blame] | 2902 | /* drop middle, data, if any */ |
| 2903 | if (m->middle) { |
| 2904 | ceph_buffer_put(m->middle); |
| 2905 | m->middle = NULL; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2906 | } |
Alex Elder | 4a73ef2 | 2013-03-07 15:38:26 -0600 | [diff] [blame] | 2907 | m->length = 0; |
Sage Weil | c2e552e | 2009-12-07 15:55:05 -0800 | [diff] [blame] | 2908 | m->pages = NULL; |
| 2909 | |
Sage Weil | 58bb3b3 | 2009-12-23 12:12:31 -0800 | [diff] [blame] | 2910 | if (m->pagelist) { |
| 2911 | ceph_pagelist_release(m->pagelist); |
| 2912 | kfree(m->pagelist); |
| 2913 | m->pagelist = NULL; |
| 2914 | } |
| 2915 | |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 2916 | m->trail = NULL; |
| 2917 | |
Sage Weil | c2e552e | 2009-12-07 15:55:05 -0800 | [diff] [blame] | 2918 | if (m->pool) |
| 2919 | ceph_msgpool_put(m->pool, m); |
| 2920 | else |
| 2921 | ceph_msg_kfree(m); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2922 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2923 | EXPORT_SYMBOL(ceph_msg_last_put); |
Sage Weil | 9ec7cab | 2009-12-14 15:13:47 -0800 | [diff] [blame] | 2924 | |
| 2925 | void ceph_msg_dump(struct ceph_msg *msg) |
| 2926 | { |
Alex Elder | 4a73ef2 | 2013-03-07 15:38:26 -0600 | [diff] [blame] | 2927 | pr_debug("msg_dump %p (front_max %d length %zd)\n", msg, |
| 2928 | msg->front_max, msg->length); |
Sage Weil | 9ec7cab | 2009-12-14 15:13:47 -0800 | [diff] [blame] | 2929 | print_hex_dump(KERN_DEBUG, "header: ", |
| 2930 | DUMP_PREFIX_OFFSET, 16, 1, |
| 2931 | &msg->hdr, sizeof(msg->hdr), true); |
| 2932 | print_hex_dump(KERN_DEBUG, " front: ", |
| 2933 | DUMP_PREFIX_OFFSET, 16, 1, |
| 2934 | msg->front.iov_base, msg->front.iov_len, true); |
| 2935 | if (msg->middle) |
| 2936 | print_hex_dump(KERN_DEBUG, "middle: ", |
| 2937 | DUMP_PREFIX_OFFSET, 16, 1, |
| 2938 | msg->middle->vec.iov_base, |
| 2939 | msg->middle->vec.iov_len, true); |
| 2940 | print_hex_dump(KERN_DEBUG, "footer: ", |
| 2941 | DUMP_PREFIX_OFFSET, 16, 1, |
| 2942 | &msg->footer, sizeof(msg->footer), true); |
| 2943 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2944 | EXPORT_SYMBOL(ceph_msg_dump); |