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 | |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 24 | #define list_entry_next(pos, member) \ |
| 25 | list_entry(pos->member.next, typeof(*pos), member) |
| 26 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 27 | /* |
| 28 | * Ceph uses the messenger to exchange ceph_msg messages with other |
| 29 | * hosts in the system. The messenger provides ordered and reliable |
| 30 | * delivery. We tolerate TCP disconnects by reconnecting (with |
| 31 | * exponential backoff) in the case of a fault (disconnection, bad |
| 32 | * crc, protocol error). Acks allow sent messages to be discarded by |
| 33 | * the sender. |
| 34 | */ |
| 35 | |
Alex Elder | bc18f4b | 2012-06-20 21:53:53 -0500 | [diff] [blame] | 36 | /* |
| 37 | * We track the state of the socket on a given connection using |
| 38 | * values defined below. The transition to a new socket state is |
| 39 | * handled by a function which verifies we aren't coming from an |
| 40 | * unexpected state. |
| 41 | * |
| 42 | * -------- |
| 43 | * | NEW* | transient initial state |
| 44 | * -------- |
| 45 | * | con_sock_state_init() |
| 46 | * v |
| 47 | * ---------- |
| 48 | * | CLOSED | initialized, but no socket (and no |
| 49 | * ---------- TCP connection) |
| 50 | * ^ \ |
| 51 | * | \ con_sock_state_connecting() |
| 52 | * | ---------------------- |
| 53 | * | \ |
| 54 | * + con_sock_state_closed() \ |
Sage Weil | fbb85a4 | 2012-06-27 12:31:02 -0700 | [diff] [blame] | 55 | * |+--------------------------- \ |
| 56 | * | \ \ \ |
| 57 | * | ----------- \ \ |
| 58 | * | | CLOSING | socket event; \ \ |
| 59 | * | ----------- await close \ \ |
| 60 | * | ^ \ | |
| 61 | * | | \ | |
| 62 | * | + con_sock_state_closing() \ | |
| 63 | * | / \ | | |
| 64 | * | / --------------- | | |
| 65 | * | / \ v v |
Alex Elder | bc18f4b | 2012-06-20 21:53:53 -0500 | [diff] [blame] | 66 | * | / -------------- |
| 67 | * | / -----------------| CONNECTING | socket created, TCP |
| 68 | * | | / -------------- connect initiated |
| 69 | * | | | con_sock_state_connected() |
| 70 | * | | v |
| 71 | * ------------- |
| 72 | * | CONNECTED | TCP connection established |
| 73 | * ------------- |
| 74 | * |
| 75 | * State values for ceph_connection->sock_state; NEW is assumed to be 0. |
| 76 | */ |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 77 | |
| 78 | #define CON_SOCK_STATE_NEW 0 /* -> CLOSED */ |
| 79 | #define CON_SOCK_STATE_CLOSED 1 /* -> CONNECTING */ |
| 80 | #define CON_SOCK_STATE_CONNECTING 2 /* -> CONNECTED or -> CLOSING */ |
| 81 | #define CON_SOCK_STATE_CONNECTED 3 /* -> CLOSING or -> CLOSED */ |
| 82 | #define CON_SOCK_STATE_CLOSING 4 /* -> CLOSED */ |
| 83 | |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 84 | /* |
| 85 | * connection states |
| 86 | */ |
| 87 | #define CON_STATE_CLOSED 1 /* -> PREOPEN */ |
| 88 | #define CON_STATE_PREOPEN 2 /* -> CONNECTING, CLOSED */ |
| 89 | #define CON_STATE_CONNECTING 3 /* -> NEGOTIATING, CLOSED */ |
| 90 | #define CON_STATE_NEGOTIATING 4 /* -> OPEN, CLOSED */ |
| 91 | #define CON_STATE_OPEN 5 /* -> STANDBY, CLOSED */ |
| 92 | #define CON_STATE_STANDBY 6 /* -> PREOPEN, CLOSED */ |
| 93 | |
Sage Weil | 4a86169 | 2012-07-20 17:29:55 -0700 | [diff] [blame] | 94 | /* |
| 95 | * ceph_connection flag bits |
| 96 | */ |
| 97 | #define CON_FLAG_LOSSYTX 0 /* we can close channel or drop |
| 98 | * messages on errors */ |
| 99 | #define CON_FLAG_KEEPALIVE_PENDING 1 /* we need to send a keepalive */ |
| 100 | #define CON_FLAG_WRITE_PENDING 2 /* we have data ready to send */ |
| 101 | #define CON_FLAG_SOCK_CLOSED 3 /* socket state changed to closed */ |
| 102 | #define CON_FLAG_BACKOFF 4 /* need to retry queuing delayed work */ |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 103 | |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 104 | static bool con_flag_valid(unsigned long con_flag) |
| 105 | { |
| 106 | switch (con_flag) { |
| 107 | case CON_FLAG_LOSSYTX: |
| 108 | case CON_FLAG_KEEPALIVE_PENDING: |
| 109 | case CON_FLAG_WRITE_PENDING: |
| 110 | case CON_FLAG_SOCK_CLOSED: |
| 111 | case CON_FLAG_BACKOFF: |
| 112 | return true; |
| 113 | default: |
| 114 | return false; |
| 115 | } |
| 116 | } |
| 117 | |
| 118 | static void con_flag_clear(struct ceph_connection *con, unsigned long con_flag) |
| 119 | { |
| 120 | BUG_ON(!con_flag_valid(con_flag)); |
| 121 | |
| 122 | clear_bit(con_flag, &con->flags); |
| 123 | } |
| 124 | |
| 125 | static void con_flag_set(struct ceph_connection *con, unsigned long con_flag) |
| 126 | { |
| 127 | BUG_ON(!con_flag_valid(con_flag)); |
| 128 | |
| 129 | set_bit(con_flag, &con->flags); |
| 130 | } |
| 131 | |
| 132 | static bool con_flag_test(struct ceph_connection *con, unsigned long con_flag) |
| 133 | { |
| 134 | BUG_ON(!con_flag_valid(con_flag)); |
| 135 | |
| 136 | return test_bit(con_flag, &con->flags); |
| 137 | } |
| 138 | |
| 139 | static bool con_flag_test_and_clear(struct ceph_connection *con, |
| 140 | unsigned long con_flag) |
| 141 | { |
| 142 | BUG_ON(!con_flag_valid(con_flag)); |
| 143 | |
| 144 | return test_and_clear_bit(con_flag, &con->flags); |
| 145 | } |
| 146 | |
| 147 | static bool con_flag_test_and_set(struct ceph_connection *con, |
| 148 | unsigned long con_flag) |
| 149 | { |
| 150 | BUG_ON(!con_flag_valid(con_flag)); |
| 151 | |
| 152 | return test_and_set_bit(con_flag, &con->flags); |
| 153 | } |
| 154 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 155 | /* static tag bytes (protocol control messages) */ |
| 156 | static char tag_msg = CEPH_MSGR_TAG_MSG; |
| 157 | static char tag_ack = CEPH_MSGR_TAG_ACK; |
| 158 | static char tag_keepalive = CEPH_MSGR_TAG_KEEPALIVE; |
| 159 | |
Sage Weil | a6a5349 | 2010-04-13 14:07:07 -0700 | [diff] [blame] | 160 | #ifdef CONFIG_LOCKDEP |
| 161 | static struct lock_class_key socket_class; |
| 162 | #endif |
| 163 | |
Alex Elder | 84495f4 | 2012-02-15 07:43:55 -0600 | [diff] [blame] | 164 | /* |
| 165 | * When skipping (ignoring) a block of input we read it into a "skip |
| 166 | * buffer," which is this many bytes in size. |
| 167 | */ |
| 168 | #define SKIP_BUF_SIZE 1024 |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 169 | |
| 170 | static void queue_con(struct ceph_connection *con); |
| 171 | static void con_work(struct work_struct *); |
Alex Elder | 9320926 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 172 | static void con_fault(struct ceph_connection *con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 173 | |
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 | * Nicely render a sockaddr as a string. An array of formatted |
| 176 | * strings is used, to approximate reentrancy. |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 177 | */ |
Alex Elder | f64a931 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 178 | #define ADDR_STR_COUNT_LOG 5 /* log2(# address strings in array) */ |
| 179 | #define ADDR_STR_COUNT (1 << ADDR_STR_COUNT_LOG) |
| 180 | #define ADDR_STR_COUNT_MASK (ADDR_STR_COUNT - 1) |
| 181 | #define MAX_ADDR_STR_LEN 64 /* 54 is enough */ |
| 182 | |
| 183 | static char addr_str[ADDR_STR_COUNT][MAX_ADDR_STR_LEN]; |
| 184 | static atomic_t addr_str_seq = ATOMIC_INIT(0); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 185 | |
Alex Elder | 5766651 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 186 | static struct page *zero_page; /* used in certain error cases */ |
Alex Elder | 5766651 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 187 | |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 188 | const char *ceph_pr_addr(const struct sockaddr_storage *ss) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 189 | { |
| 190 | int i; |
| 191 | char *s; |
Alex Elder | 99f0f3b | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 192 | struct sockaddr_in *in4 = (struct sockaddr_in *) ss; |
| 193 | struct sockaddr_in6 *in6 = (struct sockaddr_in6 *) ss; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 194 | |
Alex Elder | f64a931 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 195 | i = atomic_inc_return(&addr_str_seq) & ADDR_STR_COUNT_MASK; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 196 | s = addr_str[i]; |
| 197 | |
| 198 | switch (ss->ss_family) { |
| 199 | case AF_INET: |
Alex Elder | bd40614 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 200 | snprintf(s, MAX_ADDR_STR_LEN, "%pI4:%hu", &in4->sin_addr, |
| 201 | ntohs(in4->sin_port)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 202 | break; |
| 203 | |
| 204 | case AF_INET6: |
Alex Elder | bd40614 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 205 | snprintf(s, MAX_ADDR_STR_LEN, "[%pI6c]:%hu", &in6->sin6_addr, |
| 206 | ntohs(in6->sin6_port)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 207 | break; |
| 208 | |
| 209 | default: |
Alex Elder | d3002b9 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 210 | snprintf(s, MAX_ADDR_STR_LEN, "(unknown sockaddr family %hu)", |
| 211 | ss->ss_family); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 212 | } |
| 213 | |
| 214 | return s; |
| 215 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 216 | EXPORT_SYMBOL(ceph_pr_addr); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 217 | |
Sage Weil | 63f2d21 | 2009-11-03 15:17:56 -0800 | [diff] [blame] | 218 | static void encode_my_addr(struct ceph_messenger *msgr) |
| 219 | { |
| 220 | memcpy(&msgr->my_enc_addr, &msgr->inst.addr, sizeof(msgr->my_enc_addr)); |
| 221 | ceph_encode_addr(&msgr->my_enc_addr); |
| 222 | } |
| 223 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 224 | /* |
| 225 | * work queue for all reading and writing to/from the socket. |
| 226 | */ |
Alex Elder | e0f43c9 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 227 | static struct workqueue_struct *ceph_msgr_wq; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 228 | |
Alex Elder | 1541716 | 2013-02-19 12:25:56 -0600 | [diff] [blame] | 229 | static void _ceph_msgr_exit(void) |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 230 | { |
Alex Elder | d3002b9 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 231 | if (ceph_msgr_wq) { |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 232 | destroy_workqueue(ceph_msgr_wq); |
Alex Elder | d3002b9 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 233 | ceph_msgr_wq = NULL; |
| 234 | } |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 235 | |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 236 | BUG_ON(zero_page == NULL); |
| 237 | kunmap(zero_page); |
| 238 | page_cache_release(zero_page); |
| 239 | zero_page = NULL; |
| 240 | } |
| 241 | |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 242 | int ceph_msgr_init(void) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 243 | { |
Alex Elder | 5766651 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 244 | BUG_ON(zero_page != NULL); |
| 245 | zero_page = ZERO_PAGE(0); |
| 246 | page_cache_get(zero_page); |
| 247 | |
Tejun Heo | f363e45 | 2011-01-03 14:49:46 +0100 | [diff] [blame] | 248 | ceph_msgr_wq = alloc_workqueue("ceph-msgr", WQ_NON_REENTRANT, 0); |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 249 | if (ceph_msgr_wq) |
| 250 | return 0; |
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 | pr_err("msgr_init failed to create workqueue\n"); |
| 253 | _ceph_msgr_exit(); |
Alex Elder | 5766651 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 254 | |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 255 | return -ENOMEM; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 256 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 257 | EXPORT_SYMBOL(ceph_msgr_init); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 258 | |
| 259 | void ceph_msgr_exit(void) |
| 260 | { |
Alex Elder | 5766651 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 261 | BUG_ON(ceph_msgr_wq == NULL); |
Alex Elder | 5766651 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 262 | |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 263 | _ceph_msgr_exit(); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 264 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 265 | EXPORT_SYMBOL(ceph_msgr_exit); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 266 | |
Yehuda Sadeh | cd84db6 | 2010-06-11 16:58:48 -0700 | [diff] [blame] | 267 | void ceph_msgr_flush(void) |
Sage Weil | a922d38 | 2010-05-29 09:41:23 -0700 | [diff] [blame] | 268 | { |
| 269 | flush_workqueue(ceph_msgr_wq); |
| 270 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 271 | EXPORT_SYMBOL(ceph_msgr_flush); |
Sage Weil | a922d38 | 2010-05-29 09:41:23 -0700 | [diff] [blame] | 272 | |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 273 | /* Connection socket state transition functions */ |
| 274 | |
| 275 | static void con_sock_state_init(struct ceph_connection *con) |
| 276 | { |
| 277 | int old_state; |
| 278 | |
| 279 | old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CLOSED); |
| 280 | if (WARN_ON(old_state != CON_SOCK_STATE_NEW)) |
| 281 | printk("%s: unexpected old state %d\n", __func__, old_state); |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 282 | dout("%s con %p sock %d -> %d\n", __func__, con, old_state, |
| 283 | CON_SOCK_STATE_CLOSED); |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 284 | } |
| 285 | |
| 286 | static void con_sock_state_connecting(struct ceph_connection *con) |
| 287 | { |
| 288 | int old_state; |
| 289 | |
| 290 | old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CONNECTING); |
| 291 | if (WARN_ON(old_state != CON_SOCK_STATE_CLOSED)) |
| 292 | printk("%s: unexpected old state %d\n", __func__, old_state); |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 293 | dout("%s con %p sock %d -> %d\n", __func__, con, old_state, |
| 294 | CON_SOCK_STATE_CONNECTING); |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 295 | } |
| 296 | |
| 297 | static void con_sock_state_connected(struct ceph_connection *con) |
| 298 | { |
| 299 | int old_state; |
| 300 | |
| 301 | old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CONNECTED); |
| 302 | if (WARN_ON(old_state != CON_SOCK_STATE_CONNECTING)) |
| 303 | printk("%s: unexpected old state %d\n", __func__, old_state); |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 304 | dout("%s con %p sock %d -> %d\n", __func__, con, old_state, |
| 305 | CON_SOCK_STATE_CONNECTED); |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 306 | } |
| 307 | |
| 308 | static void con_sock_state_closing(struct ceph_connection *con) |
| 309 | { |
| 310 | int old_state; |
| 311 | |
| 312 | old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CLOSING); |
| 313 | if (WARN_ON(old_state != CON_SOCK_STATE_CONNECTING && |
| 314 | old_state != CON_SOCK_STATE_CONNECTED && |
| 315 | old_state != CON_SOCK_STATE_CLOSING)) |
| 316 | printk("%s: unexpected old state %d\n", __func__, old_state); |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 317 | dout("%s con %p sock %d -> %d\n", __func__, con, old_state, |
| 318 | CON_SOCK_STATE_CLOSING); |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 319 | } |
| 320 | |
| 321 | static void con_sock_state_closed(struct ceph_connection *con) |
| 322 | { |
| 323 | int old_state; |
| 324 | |
| 325 | old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CLOSED); |
| 326 | if (WARN_ON(old_state != CON_SOCK_STATE_CONNECTED && |
Sage Weil | fbb85a4 | 2012-06-27 12:31:02 -0700 | [diff] [blame] | 327 | old_state != CON_SOCK_STATE_CLOSING && |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 328 | old_state != CON_SOCK_STATE_CONNECTING && |
| 329 | old_state != CON_SOCK_STATE_CLOSED)) |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 330 | printk("%s: unexpected old state %d\n", __func__, old_state); |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 331 | dout("%s con %p sock %d -> %d\n", __func__, con, old_state, |
| 332 | CON_SOCK_STATE_CLOSED); |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 333 | } |
Sage Weil | a922d38 | 2010-05-29 09:41:23 -0700 | [diff] [blame] | 334 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 335 | /* |
| 336 | * socket callback functions |
| 337 | */ |
| 338 | |
| 339 | /* data available on socket, or listen socket received a connect */ |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 340 | static void ceph_sock_data_ready(struct sock *sk, int count_unused) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 341 | { |
Alex Elder | bd40614 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 342 | struct ceph_connection *con = sk->sk_user_data; |
Guanjun He | a2a3258 | 2012-07-08 19:50:33 -0700 | [diff] [blame] | 343 | if (atomic_read(&con->msgr->stopping)) { |
| 344 | return; |
| 345 | } |
Alex Elder | bd40614 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 346 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 347 | if (sk->sk_state != TCP_CLOSE_WAIT) { |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 348 | dout("%s on %p state = %lu, queueing work\n", __func__, |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 349 | con, con->state); |
| 350 | queue_con(con); |
| 351 | } |
| 352 | } |
| 353 | |
| 354 | /* socket has buffer space for writing */ |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 355 | static void ceph_sock_write_space(struct sock *sk) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 356 | { |
Alex Elder | d3002b9 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 357 | struct ceph_connection *con = sk->sk_user_data; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 358 | |
Jim Schutt | 182fac2 | 2012-02-29 08:30:58 -0700 | [diff] [blame] | 359 | /* only queue to workqueue if there is data we want to write, |
| 360 | * and there is sufficient space in the socket buffer to accept |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 361 | * more data. clear SOCK_NOSPACE so that ceph_sock_write_space() |
Jim Schutt | 182fac2 | 2012-02-29 08:30:58 -0700 | [diff] [blame] | 362 | * doesn't get called again until try_write() fills the socket |
| 363 | * buffer. See net/ipv4/tcp_input.c:tcp_check_space() |
| 364 | * and net/core/stream.c:sk_stream_write_space(). |
| 365 | */ |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 366 | if (con_flag_test(con, CON_FLAG_WRITE_PENDING)) { |
Jim Schutt | 182fac2 | 2012-02-29 08:30:58 -0700 | [diff] [blame] | 367 | if (sk_stream_wspace(sk) >= sk_stream_min_wspace(sk)) { |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 368 | dout("%s %p queueing write work\n", __func__, con); |
Jim Schutt | 182fac2 | 2012-02-29 08:30:58 -0700 | [diff] [blame] | 369 | clear_bit(SOCK_NOSPACE, &sk->sk_socket->flags); |
| 370 | queue_con(con); |
| 371 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 372 | } else { |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 373 | dout("%s %p nothing to write\n", __func__, con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 374 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 375 | } |
| 376 | |
| 377 | /* socket's state has changed */ |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 378 | static void ceph_sock_state_change(struct sock *sk) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 379 | { |
Alex Elder | bd40614 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 380 | struct ceph_connection *con = sk->sk_user_data; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 381 | |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 382 | dout("%s %p state = %lu sk_state = %u\n", __func__, |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 383 | con, con->state, sk->sk_state); |
| 384 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 385 | switch (sk->sk_state) { |
| 386 | case TCP_CLOSE: |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 387 | dout("%s TCP_CLOSE\n", __func__); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 388 | case TCP_CLOSE_WAIT: |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 389 | dout("%s TCP_CLOSE_WAIT\n", __func__); |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 390 | con_sock_state_closing(con); |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 391 | con_flag_set(con, CON_FLAG_SOCK_CLOSED); |
Alex Elder | d65c9e0 | 2012-06-20 21:53:53 -0500 | [diff] [blame] | 392 | queue_con(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 393 | break; |
| 394 | case TCP_ESTABLISHED: |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 395 | dout("%s TCP_ESTABLISHED\n", __func__); |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 396 | con_sock_state_connected(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 397 | queue_con(con); |
| 398 | break; |
Alex Elder | d3002b9 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 399 | default: /* Everything else is uninteresting */ |
| 400 | break; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 401 | } |
| 402 | } |
| 403 | |
| 404 | /* |
| 405 | * set up socket callbacks |
| 406 | */ |
| 407 | static void set_sock_callbacks(struct socket *sock, |
| 408 | struct ceph_connection *con) |
| 409 | { |
| 410 | struct sock *sk = sock->sk; |
Alex Elder | bd40614 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 411 | sk->sk_user_data = con; |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 412 | sk->sk_data_ready = ceph_sock_data_ready; |
| 413 | sk->sk_write_space = ceph_sock_write_space; |
| 414 | sk->sk_state_change = ceph_sock_state_change; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 415 | } |
| 416 | |
| 417 | |
| 418 | /* |
| 419 | * socket helpers |
| 420 | */ |
| 421 | |
| 422 | /* |
| 423 | * initiate connection to a remote socket. |
| 424 | */ |
Alex Elder | 41617d0 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 425 | static int ceph_tcp_connect(struct ceph_connection *con) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 426 | { |
Sage Weil | f91d347 | 2010-07-01 15:18:31 -0700 | [diff] [blame] | 427 | struct sockaddr_storage *paddr = &con->peer_addr.in_addr; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 428 | struct socket *sock; |
| 429 | int ret; |
| 430 | |
| 431 | BUG_ON(con->sock); |
Sage Weil | f91d347 | 2010-07-01 15:18:31 -0700 | [diff] [blame] | 432 | ret = sock_create_kern(con->peer_addr.in_addr.ss_family, SOCK_STREAM, |
| 433 | IPPROTO_TCP, &sock); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 434 | if (ret) |
Alex Elder | 41617d0 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 435 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 436 | sock->sk->sk_allocation = GFP_NOFS; |
| 437 | |
Sage Weil | a6a5349 | 2010-04-13 14:07:07 -0700 | [diff] [blame] | 438 | #ifdef CONFIG_LOCKDEP |
| 439 | lockdep_set_class(&sock->sk->sk_lock, &socket_class); |
| 440 | #endif |
| 441 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 442 | set_sock_callbacks(sock, con); |
| 443 | |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 444 | dout("connect %s\n", ceph_pr_addr(&con->peer_addr.in_addr)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 445 | |
Sage Weil | 89a86be | 2012-06-09 14:19:21 -0700 | [diff] [blame] | 446 | con_sock_state_connecting(con); |
Sage Weil | f91d347 | 2010-07-01 15:18:31 -0700 | [diff] [blame] | 447 | ret = sock->ops->connect(sock, (struct sockaddr *)paddr, sizeof(*paddr), |
| 448 | O_NONBLOCK); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 449 | if (ret == -EINPROGRESS) { |
| 450 | dout("connect %s EINPROGRESS sk_state = %u\n", |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 451 | ceph_pr_addr(&con->peer_addr.in_addr), |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 452 | sock->sk->sk_state); |
Alex Elder | a5bc3129a | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 453 | } else if (ret < 0) { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 454 | pr_err("connect %s error %d\n", |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 455 | ceph_pr_addr(&con->peer_addr.in_addr), ret); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 456 | sock_release(sock); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 457 | con->error_msg = "connect error"; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 458 | |
Alex Elder | 41617d0 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 459 | return ret; |
Alex Elder | a5bc3129a | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 460 | } |
| 461 | con->sock = sock; |
Alex Elder | 41617d0 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 462 | return 0; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 463 | } |
| 464 | |
| 465 | static int ceph_tcp_recvmsg(struct socket *sock, void *buf, size_t len) |
| 466 | { |
| 467 | struct kvec iov = {buf, len}; |
| 468 | struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL }; |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 469 | int r; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 470 | |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 471 | r = kernel_recvmsg(sock, &msg, &iov, 1, len, msg.msg_flags); |
| 472 | if (r == -EAGAIN) |
| 473 | r = 0; |
| 474 | return r; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 475 | } |
| 476 | |
Alex Elder | afb3d90 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 477 | static int ceph_tcp_recvpage(struct socket *sock, struct page *page, |
| 478 | int page_offset, size_t length) |
| 479 | { |
| 480 | void *kaddr; |
| 481 | int ret; |
| 482 | |
| 483 | BUG_ON(page_offset + length > PAGE_SIZE); |
| 484 | |
| 485 | kaddr = kmap(page); |
| 486 | BUG_ON(!kaddr); |
| 487 | ret = ceph_tcp_recvmsg(sock, kaddr + page_offset, length); |
| 488 | kunmap(page); |
| 489 | |
| 490 | return ret; |
| 491 | } |
| 492 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 493 | /* |
| 494 | * write something. @more is true if caller will be sending more data |
| 495 | * shortly. |
| 496 | */ |
| 497 | static int ceph_tcp_sendmsg(struct socket *sock, struct kvec *iov, |
| 498 | size_t kvlen, size_t len, int more) |
| 499 | { |
| 500 | struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL }; |
Sage Weil | 42961d2 | 2011-01-25 08:19:34 -0800 | [diff] [blame] | 501 | int r; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 502 | |
| 503 | if (more) |
| 504 | msg.msg_flags |= MSG_MORE; |
| 505 | else |
| 506 | msg.msg_flags |= MSG_EOR; /* superfluous, but what the hell */ |
| 507 | |
Sage Weil | 42961d2 | 2011-01-25 08:19:34 -0800 | [diff] [blame] | 508 | r = kernel_sendmsg(sock, &msg, iov, kvlen, len); |
| 509 | if (r == -EAGAIN) |
| 510 | r = 0; |
| 511 | return r; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 512 | } |
| 513 | |
Alex Elder | 3173913 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 514 | static int ceph_tcp_sendpage(struct socket *sock, struct page *page, |
Alex Elder | e1dcb12 | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 515 | int offset, size_t size, bool more) |
Alex Elder | 3173913 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 516 | { |
| 517 | int flags = MSG_DONTWAIT | MSG_NOSIGNAL | (more ? MSG_MORE : MSG_EOR); |
| 518 | int ret; |
| 519 | |
| 520 | ret = kernel_sendpage(sock, page, offset, size, flags); |
| 521 | if (ret == -EAGAIN) |
| 522 | ret = 0; |
| 523 | |
| 524 | return ret; |
| 525 | } |
| 526 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 527 | |
| 528 | /* |
| 529 | * Shutdown/close the socket for the given connection. |
| 530 | */ |
| 531 | static int con_close_socket(struct ceph_connection *con) |
| 532 | { |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 533 | int rc = 0; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 534 | |
| 535 | dout("con_close_socket on %p sock %p\n", con, con->sock); |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 536 | if (con->sock) { |
| 537 | rc = con->sock->ops->shutdown(con->sock, SHUT_RDWR); |
| 538 | sock_release(con->sock); |
| 539 | con->sock = NULL; |
| 540 | } |
Alex Elder | 456ea46 | 2012-06-20 21:53:53 -0500 | [diff] [blame] | 541 | |
| 542 | /* |
Sage Weil | 4a86169 | 2012-07-20 17:29:55 -0700 | [diff] [blame] | 543 | * Forcibly clear the SOCK_CLOSED flag. It gets set |
Alex Elder | 456ea46 | 2012-06-20 21:53:53 -0500 | [diff] [blame] | 544 | * independent of the connection mutex, and we could have |
| 545 | * received a socket close event before we had the chance to |
| 546 | * shut the socket down. |
| 547 | */ |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 548 | con_flag_clear(con, CON_FLAG_SOCK_CLOSED); |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 549 | |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 550 | con_sock_state_closed(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 551 | return rc; |
| 552 | } |
| 553 | |
| 554 | /* |
| 555 | * Reset a connection. Discard all incoming and outgoing messages |
| 556 | * and clear *_seq state. |
| 557 | */ |
| 558 | static void ceph_msg_remove(struct ceph_msg *msg) |
| 559 | { |
| 560 | list_del_init(&msg->list_head); |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 561 | BUG_ON(msg->con == NULL); |
Sage Weil | 36eb71a | 2012-06-21 12:47:08 -0700 | [diff] [blame] | 562 | msg->con->ops->put(msg->con); |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 563 | msg->con = NULL; |
| 564 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 565 | ceph_msg_put(msg); |
| 566 | } |
| 567 | static void ceph_msg_remove_list(struct list_head *head) |
| 568 | { |
| 569 | while (!list_empty(head)) { |
| 570 | struct ceph_msg *msg = list_first_entry(head, struct ceph_msg, |
| 571 | list_head); |
| 572 | ceph_msg_remove(msg); |
| 573 | } |
| 574 | } |
| 575 | |
| 576 | static void reset_connection(struct ceph_connection *con) |
| 577 | { |
| 578 | /* reset connection, out_queue, msg_ and connect_seq */ |
| 579 | /* discard existing out_queue and msg_seq */ |
Sage Weil | 0fa6ebc | 2012-12-27 20:27:04 -0600 | [diff] [blame] | 580 | dout("reset_connection %p\n", con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 581 | ceph_msg_remove_list(&con->out_queue); |
| 582 | ceph_msg_remove_list(&con->out_sent); |
| 583 | |
Sage Weil | cf3e5c4 | 2009-12-11 09:48:05 -0800 | [diff] [blame] | 584 | if (con->in_msg) { |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 585 | BUG_ON(con->in_msg->con != con); |
| 586 | con->in_msg->con = NULL; |
Sage Weil | cf3e5c4 | 2009-12-11 09:48:05 -0800 | [diff] [blame] | 587 | ceph_msg_put(con->in_msg); |
| 588 | con->in_msg = NULL; |
Sage Weil | 36eb71a | 2012-06-21 12:47:08 -0700 | [diff] [blame] | 589 | con->ops->put(con); |
Sage Weil | cf3e5c4 | 2009-12-11 09:48:05 -0800 | [diff] [blame] | 590 | } |
| 591 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 592 | con->connect_seq = 0; |
| 593 | con->out_seq = 0; |
Sage Weil | c86a293 | 2009-12-14 14:04:30 -0800 | [diff] [blame] | 594 | if (con->out_msg) { |
| 595 | ceph_msg_put(con->out_msg); |
| 596 | con->out_msg = NULL; |
| 597 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 598 | con->in_seq = 0; |
Sage Weil | 0e0d5e0 | 2010-04-02 16:07:19 -0700 | [diff] [blame] | 599 | con->in_seq_acked = 0; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 600 | } |
| 601 | |
| 602 | /* |
| 603 | * mark a peer down. drop any open connections. |
| 604 | */ |
| 605 | void ceph_con_close(struct ceph_connection *con) |
| 606 | { |
Sage Weil | 8c50c81 | 2012-07-30 16:24:37 -0700 | [diff] [blame] | 607 | mutex_lock(&con->mutex); |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 608 | dout("con_close %p peer %s\n", con, |
| 609 | ceph_pr_addr(&con->peer_addr.in_addr)); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 610 | con->state = CON_STATE_CLOSED; |
Alex Elder | a5988c4 | 2012-05-29 11:04:58 -0500 | [diff] [blame] | 611 | |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 612 | con_flag_clear(con, CON_FLAG_LOSSYTX); /* so we retry next connect */ |
| 613 | con_flag_clear(con, CON_FLAG_KEEPALIVE_PENDING); |
| 614 | con_flag_clear(con, CON_FLAG_WRITE_PENDING); |
| 615 | con_flag_clear(con, CON_FLAG_BACKOFF); |
Alex Elder | a5988c4 | 2012-05-29 11:04:58 -0500 | [diff] [blame] | 616 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 617 | reset_connection(con); |
Sage Weil | 6f2bc3f | 2010-04-02 16:16:34 -0700 | [diff] [blame] | 618 | con->peer_global_seq = 0; |
Sage Weil | 91e45ce3 | 2010-02-15 12:05:09 -0800 | [diff] [blame] | 619 | cancel_delayed_work(&con->work); |
Sage Weil | ee76e07 | 2012-07-20 16:45:49 -0700 | [diff] [blame] | 620 | con_close_socket(con); |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 621 | mutex_unlock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 622 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 623 | EXPORT_SYMBOL(ceph_con_close); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 624 | |
| 625 | /* |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 626 | * Reopen a closed connection, with a new peer address. |
| 627 | */ |
Sage Weil | b7a9e5d | 2012-06-27 12:24:08 -0700 | [diff] [blame] | 628 | void ceph_con_open(struct ceph_connection *con, |
| 629 | __u8 entity_type, __u64 entity_num, |
| 630 | struct ceph_entity_addr *addr) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 631 | { |
Sage Weil | 5469155 | 2012-07-30 16:21:40 -0700 | [diff] [blame] | 632 | mutex_lock(&con->mutex); |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 633 | 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] | 634 | |
Alex Elder | 122070a | 2012-12-26 10:43:57 -0600 | [diff] [blame] | 635 | WARN_ON(con->state != CON_STATE_CLOSED); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 636 | con->state = CON_STATE_PREOPEN; |
Alex Elder | a5988c4 | 2012-05-29 11:04:58 -0500 | [diff] [blame] | 637 | |
Sage Weil | b7a9e5d | 2012-06-27 12:24:08 -0700 | [diff] [blame] | 638 | con->peer_name.type = (__u8) entity_type; |
| 639 | con->peer_name.num = cpu_to_le64(entity_num); |
| 640 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 641 | memcpy(&con->peer_addr, addr, sizeof(*addr)); |
Sage Weil | 03c677e | 2009-11-20 15:14:15 -0800 | [diff] [blame] | 642 | con->delay = 0; /* reset backoff memory */ |
Sage Weil | 5469155 | 2012-07-30 16:21:40 -0700 | [diff] [blame] | 643 | mutex_unlock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 644 | queue_con(con); |
| 645 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 646 | EXPORT_SYMBOL(ceph_con_open); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 647 | |
| 648 | /* |
Sage Weil | 87b315a | 2010-03-22 14:51:18 -0700 | [diff] [blame] | 649 | * return true if this connection ever successfully opened |
| 650 | */ |
| 651 | bool ceph_con_opened(struct ceph_connection *con) |
| 652 | { |
| 653 | return con->connect_seq > 0; |
| 654 | } |
| 655 | |
| 656 | /* |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 657 | * initialize a new connection. |
| 658 | */ |
Alex Elder | 1bfd89f | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 659 | void ceph_con_init(struct ceph_connection *con, void *private, |
| 660 | const struct ceph_connection_operations *ops, |
Sage Weil | b7a9e5d | 2012-06-27 12:24:08 -0700 | [diff] [blame] | 661 | struct ceph_messenger *msgr) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 662 | { |
| 663 | dout("con_init %p\n", con); |
| 664 | memset(con, 0, sizeof(*con)); |
Alex Elder | 1bfd89f | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 665 | con->private = private; |
| 666 | con->ops = ops; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 667 | con->msgr = msgr; |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 668 | |
| 669 | con_sock_state_init(con); |
| 670 | |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 671 | mutex_init(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 672 | INIT_LIST_HEAD(&con->out_queue); |
| 673 | INIT_LIST_HEAD(&con->out_sent); |
| 674 | INIT_DELAYED_WORK(&con->work, con_work); |
Alex Elder | a5988c4 | 2012-05-29 11:04:58 -0500 | [diff] [blame] | 675 | |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 676 | con->state = CON_STATE_CLOSED; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 677 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 678 | EXPORT_SYMBOL(ceph_con_init); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 679 | |
| 680 | |
| 681 | /* |
| 682 | * We maintain a global counter to order connection attempts. Get |
| 683 | * a unique seq greater than @gt. |
| 684 | */ |
| 685 | static u32 get_global_seq(struct ceph_messenger *msgr, u32 gt) |
| 686 | { |
| 687 | u32 ret; |
| 688 | |
| 689 | spin_lock(&msgr->global_seq_lock); |
| 690 | if (msgr->global_seq < gt) |
| 691 | msgr->global_seq = gt; |
| 692 | ret = ++msgr->global_seq; |
| 693 | spin_unlock(&msgr->global_seq_lock); |
| 694 | return ret; |
| 695 | } |
| 696 | |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 697 | static void con_out_kvec_reset(struct ceph_connection *con) |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 698 | { |
| 699 | con->out_kvec_left = 0; |
| 700 | con->out_kvec_bytes = 0; |
| 701 | con->out_kvec_cur = &con->out_kvec[0]; |
| 702 | } |
| 703 | |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 704 | static void con_out_kvec_add(struct ceph_connection *con, |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 705 | size_t size, void *data) |
| 706 | { |
| 707 | int index; |
| 708 | |
| 709 | index = con->out_kvec_left; |
| 710 | BUG_ON(index >= ARRAY_SIZE(con->out_kvec)); |
| 711 | |
| 712 | con->out_kvec[index].iov_len = size; |
| 713 | con->out_kvec[index].iov_base = data; |
| 714 | con->out_kvec_left++; |
| 715 | con->out_kvec_bytes += size; |
| 716 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 717 | |
Alex Elder | df6ad1f | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 718 | #ifdef CONFIG_BLOCK |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 719 | |
| 720 | /* |
| 721 | * For a bio data item, a piece is whatever remains of the next |
| 722 | * entry in the current bio iovec, or the first entry in the next |
| 723 | * bio in the list. |
| 724 | */ |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 725 | static void ceph_msg_data_bio_cursor_init(struct ceph_msg_data *data, |
| 726 | size_t length) |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 727 | { |
| 728 | struct ceph_msg_data_cursor *cursor = &data->cursor; |
| 729 | struct bio *bio; |
| 730 | |
| 731 | BUG_ON(data->type != CEPH_MSG_DATA_BIO); |
| 732 | |
| 733 | bio = data->bio; |
| 734 | BUG_ON(!bio); |
| 735 | BUG_ON(!bio->bi_vcnt); |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 736 | |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 737 | cursor->resid = length; |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 738 | cursor->bio = bio; |
| 739 | cursor->vector_index = 0; |
| 740 | cursor->vector_offset = 0; |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 741 | cursor->last_piece = length <= bio->bi_io_vec[0].bv_len; |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 742 | } |
| 743 | |
| 744 | static struct page *ceph_msg_data_bio_next(struct ceph_msg_data *data, |
| 745 | size_t *page_offset, |
| 746 | size_t *length) |
| 747 | { |
| 748 | struct ceph_msg_data_cursor *cursor = &data->cursor; |
| 749 | struct bio *bio; |
| 750 | struct bio_vec *bio_vec; |
| 751 | unsigned int index; |
| 752 | |
| 753 | BUG_ON(data->type != CEPH_MSG_DATA_BIO); |
| 754 | |
| 755 | bio = cursor->bio; |
| 756 | BUG_ON(!bio); |
| 757 | |
| 758 | index = cursor->vector_index; |
| 759 | BUG_ON(index >= (unsigned int) bio->bi_vcnt); |
| 760 | |
| 761 | bio_vec = &bio->bi_io_vec[index]; |
| 762 | BUG_ON(cursor->vector_offset >= bio_vec->bv_len); |
| 763 | *page_offset = (size_t) (bio_vec->bv_offset + cursor->vector_offset); |
| 764 | BUG_ON(*page_offset >= PAGE_SIZE); |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 765 | if (cursor->last_piece) /* pagelist offset is always 0 */ |
| 766 | *length = cursor->resid; |
| 767 | else |
| 768 | *length = (size_t) (bio_vec->bv_len - cursor->vector_offset); |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 769 | BUG_ON(*length > PAGE_SIZE); |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 770 | BUG_ON(*length > cursor->resid); |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 771 | |
| 772 | return bio_vec->bv_page; |
| 773 | } |
| 774 | |
| 775 | static bool ceph_msg_data_bio_advance(struct ceph_msg_data *data, size_t bytes) |
| 776 | { |
| 777 | struct ceph_msg_data_cursor *cursor = &data->cursor; |
| 778 | struct bio *bio; |
| 779 | struct bio_vec *bio_vec; |
| 780 | unsigned int index; |
| 781 | |
| 782 | BUG_ON(data->type != CEPH_MSG_DATA_BIO); |
| 783 | |
| 784 | bio = cursor->bio; |
| 785 | BUG_ON(!bio); |
| 786 | |
| 787 | index = cursor->vector_index; |
| 788 | BUG_ON(index >= (unsigned int) bio->bi_vcnt); |
| 789 | bio_vec = &bio->bi_io_vec[index]; |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 790 | |
| 791 | /* Advance the cursor offset */ |
| 792 | |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 793 | BUG_ON(cursor->resid < bytes); |
| 794 | cursor->resid -= bytes; |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 795 | cursor->vector_offset += bytes; |
| 796 | if (cursor->vector_offset < bio_vec->bv_len) |
| 797 | return false; /* more bytes to process in this segment */ |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 798 | BUG_ON(cursor->vector_offset != bio_vec->bv_len); |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 799 | |
| 800 | /* Move on to the next segment, and possibly the next bio */ |
| 801 | |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 802 | if (++index == (unsigned int) bio->bi_vcnt) { |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 803 | bio = bio->bi_next; |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 804 | index = 0; |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 805 | } |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 806 | cursor->bio = bio; |
| 807 | cursor->vector_index = index; |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 808 | cursor->vector_offset = 0; |
| 809 | |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 810 | if (!cursor->last_piece) { |
| 811 | BUG_ON(!cursor->resid); |
| 812 | BUG_ON(!bio); |
| 813 | /* A short read is OK, so use <= rather than == */ |
| 814 | if (cursor->resid <= bio->bi_io_vec[index].bv_len) |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 815 | cursor->last_piece = true; |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 816 | } |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 817 | |
| 818 | return true; |
| 819 | } |
Alex Elder | df6ad1f | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 820 | #endif |
| 821 | |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 822 | /* |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 823 | * For a page array, a piece comes from the first page in the array |
| 824 | * that has not already been fully consumed. |
| 825 | */ |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 826 | static void ceph_msg_data_pages_cursor_init(struct ceph_msg_data *data, |
| 827 | size_t length) |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 828 | { |
| 829 | struct ceph_msg_data_cursor *cursor = &data->cursor; |
| 830 | int page_count; |
| 831 | |
| 832 | BUG_ON(data->type != CEPH_MSG_DATA_PAGES); |
| 833 | |
| 834 | BUG_ON(!data->pages); |
| 835 | BUG_ON(!data->length); |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 836 | BUG_ON(length != data->length); |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 837 | |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 838 | cursor->resid = length; |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 839 | page_count = calc_pages_for(data->alignment, (u64)data->length); |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 840 | cursor->page_offset = data->alignment & ~PAGE_MASK; |
| 841 | cursor->page_index = 0; |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 842 | BUG_ON(page_count > (int) USHRT_MAX); |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 843 | cursor->page_count = (unsigned short) page_count; |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 844 | cursor->last_piece = length <= PAGE_SIZE; |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 845 | } |
| 846 | |
| 847 | static struct page *ceph_msg_data_pages_next(struct ceph_msg_data *data, |
| 848 | size_t *page_offset, |
| 849 | size_t *length) |
| 850 | { |
| 851 | struct ceph_msg_data_cursor *cursor = &data->cursor; |
| 852 | |
| 853 | BUG_ON(data->type != CEPH_MSG_DATA_PAGES); |
| 854 | |
| 855 | BUG_ON(cursor->page_index >= cursor->page_count); |
| 856 | BUG_ON(cursor->page_offset >= PAGE_SIZE); |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 857 | |
| 858 | *page_offset = cursor->page_offset; |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 859 | if (cursor->last_piece) |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 860 | *length = cursor->resid; |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 861 | else |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 862 | *length = PAGE_SIZE - *page_offset; |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 863 | |
| 864 | return data->pages[cursor->page_index]; |
| 865 | } |
| 866 | |
| 867 | static bool ceph_msg_data_pages_advance(struct ceph_msg_data *data, |
| 868 | size_t bytes) |
| 869 | { |
| 870 | struct ceph_msg_data_cursor *cursor = &data->cursor; |
| 871 | |
| 872 | BUG_ON(data->type != CEPH_MSG_DATA_PAGES); |
| 873 | |
| 874 | BUG_ON(cursor->page_offset + bytes > PAGE_SIZE); |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 875 | |
| 876 | /* Advance the cursor page offset */ |
| 877 | |
| 878 | cursor->resid -= bytes; |
| 879 | cursor->page_offset += bytes; |
| 880 | if (!bytes || cursor->page_offset & ~PAGE_MASK) |
| 881 | return false; /* more bytes to process in the current page */ |
| 882 | |
| 883 | /* Move on to the next page */ |
| 884 | |
| 885 | BUG_ON(cursor->page_index >= cursor->page_count); |
| 886 | cursor->page_offset = 0; |
| 887 | cursor->page_index++; |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 888 | cursor->last_piece = cursor->resid <= PAGE_SIZE; |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 889 | |
| 890 | return true; |
| 891 | } |
| 892 | |
| 893 | /* |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 894 | * For a pagelist, a piece is whatever remains to be consumed in the |
| 895 | * first page in the list, or the front of the next page. |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 896 | */ |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 897 | static void ceph_msg_data_pagelist_cursor_init(struct ceph_msg_data *data, |
| 898 | size_t length) |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 899 | { |
| 900 | struct ceph_msg_data_cursor *cursor = &data->cursor; |
| 901 | struct ceph_pagelist *pagelist; |
| 902 | struct page *page; |
| 903 | |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 904 | BUG_ON(data->type != CEPH_MSG_DATA_PAGELIST); |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 905 | |
| 906 | pagelist = data->pagelist; |
| 907 | BUG_ON(!pagelist); |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 908 | BUG_ON(length != pagelist->length); |
| 909 | |
| 910 | if (!length) |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 911 | return; /* pagelist can be assigned but empty */ |
| 912 | |
| 913 | BUG_ON(list_empty(&pagelist->head)); |
| 914 | page = list_first_entry(&pagelist->head, struct page, lru); |
| 915 | |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 916 | cursor->resid = length; |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 917 | cursor->page = page; |
| 918 | cursor->offset = 0; |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 919 | cursor->last_piece = length <= PAGE_SIZE; |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 920 | } |
| 921 | |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 922 | static struct page *ceph_msg_data_pagelist_next(struct ceph_msg_data *data, |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 923 | size_t *page_offset, |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 924 | size_t *length) |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 925 | { |
| 926 | struct ceph_msg_data_cursor *cursor = &data->cursor; |
| 927 | struct ceph_pagelist *pagelist; |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 928 | |
| 929 | BUG_ON(data->type != CEPH_MSG_DATA_PAGELIST); |
| 930 | |
| 931 | pagelist = data->pagelist; |
| 932 | BUG_ON(!pagelist); |
| 933 | |
| 934 | BUG_ON(!cursor->page); |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 935 | BUG_ON(cursor->offset + cursor->resid != pagelist->length); |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 936 | |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 937 | *page_offset = cursor->offset & ~PAGE_MASK; |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 938 | if (cursor->last_piece) /* pagelist offset is always 0 */ |
| 939 | *length = cursor->resid; |
| 940 | else |
| 941 | *length = PAGE_SIZE - *page_offset; |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 942 | |
| 943 | return data->cursor.page; |
| 944 | } |
| 945 | |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 946 | static bool ceph_msg_data_pagelist_advance(struct ceph_msg_data *data, |
| 947 | size_t bytes) |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 948 | { |
| 949 | struct ceph_msg_data_cursor *cursor = &data->cursor; |
| 950 | struct ceph_pagelist *pagelist; |
| 951 | |
| 952 | BUG_ON(data->type != CEPH_MSG_DATA_PAGELIST); |
| 953 | |
| 954 | pagelist = data->pagelist; |
| 955 | BUG_ON(!pagelist); |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 956 | |
| 957 | BUG_ON(cursor->offset + cursor->resid != pagelist->length); |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 958 | BUG_ON((cursor->offset & ~PAGE_MASK) + bytes > PAGE_SIZE); |
| 959 | |
| 960 | /* Advance the cursor offset */ |
| 961 | |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 962 | cursor->resid -= bytes; |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 963 | cursor->offset += bytes; |
| 964 | /* pagelist offset is always 0 */ |
| 965 | if (!bytes || cursor->offset & ~PAGE_MASK) |
| 966 | return false; /* more bytes to process in the current page */ |
| 967 | |
| 968 | /* Move on to the next page */ |
| 969 | |
| 970 | BUG_ON(list_is_last(&cursor->page->lru, &pagelist->head)); |
| 971 | cursor->page = list_entry_next(cursor->page, lru); |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 972 | cursor->last_piece = cursor->resid <= PAGE_SIZE; |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 973 | |
| 974 | return true; |
| 975 | } |
| 976 | |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 977 | /* |
| 978 | * Message data is handled (sent or received) in pieces, where each |
| 979 | * piece resides on a single page. The network layer might not |
| 980 | * consume an entire piece at once. A data item's cursor keeps |
| 981 | * track of which piece is next to process and how much remains to |
| 982 | * be processed in that piece. It also tracks whether the current |
| 983 | * piece is the last one in the data item. |
| 984 | */ |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 985 | static void ceph_msg_data_cursor_init(struct ceph_msg_data *data, |
| 986 | size_t length) |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 987 | { |
| 988 | switch (data->type) { |
| 989 | case CEPH_MSG_DATA_PAGELIST: |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 990 | ceph_msg_data_pagelist_cursor_init(data, length); |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 991 | break; |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 992 | case CEPH_MSG_DATA_PAGES: |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 993 | ceph_msg_data_pages_cursor_init(data, length); |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 994 | break; |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 995 | #ifdef CONFIG_BLOCK |
| 996 | case CEPH_MSG_DATA_BIO: |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 997 | ceph_msg_data_bio_cursor_init(data, length); |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 998 | break; |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 999 | #endif /* CONFIG_BLOCK */ |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1000 | case CEPH_MSG_DATA_NONE: |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1001 | default: |
| 1002 | /* BUG(); */ |
| 1003 | break; |
| 1004 | } |
| 1005 | } |
| 1006 | |
| 1007 | /* |
| 1008 | * Return the page containing the next piece to process for a given |
| 1009 | * data item, and supply the page offset and length of that piece. |
| 1010 | * Indicate whether this is the last piece in this data item. |
| 1011 | */ |
| 1012 | static struct page *ceph_msg_data_next(struct ceph_msg_data *data, |
| 1013 | size_t *page_offset, |
| 1014 | size_t *length, |
| 1015 | bool *last_piece) |
| 1016 | { |
| 1017 | struct page *page; |
| 1018 | |
| 1019 | switch (data->type) { |
| 1020 | case CEPH_MSG_DATA_PAGELIST: |
| 1021 | page = ceph_msg_data_pagelist_next(data, page_offset, length); |
| 1022 | break; |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 1023 | case CEPH_MSG_DATA_PAGES: |
| 1024 | page = ceph_msg_data_pages_next(data, page_offset, length); |
| 1025 | break; |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1026 | #ifdef CONFIG_BLOCK |
| 1027 | case CEPH_MSG_DATA_BIO: |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1028 | page = ceph_msg_data_bio_next(data, page_offset, length); |
| 1029 | break; |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1030 | #endif /* CONFIG_BLOCK */ |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1031 | case CEPH_MSG_DATA_NONE: |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1032 | default: |
| 1033 | page = NULL; |
| 1034 | break; |
| 1035 | } |
| 1036 | BUG_ON(!page); |
| 1037 | BUG_ON(*page_offset + *length > PAGE_SIZE); |
| 1038 | BUG_ON(!*length); |
| 1039 | if (last_piece) |
| 1040 | *last_piece = data->cursor.last_piece; |
| 1041 | |
| 1042 | return page; |
| 1043 | } |
| 1044 | |
| 1045 | /* |
| 1046 | * Returns true if the result moves the cursor on to the next piece |
| 1047 | * of the data item. |
| 1048 | */ |
| 1049 | static bool ceph_msg_data_advance(struct ceph_msg_data *data, size_t bytes) |
| 1050 | { |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 1051 | struct ceph_msg_data_cursor *cursor = &data->cursor; |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1052 | bool new_piece; |
| 1053 | |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 1054 | BUG_ON(bytes > cursor->resid); |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1055 | switch (data->type) { |
| 1056 | case CEPH_MSG_DATA_PAGELIST: |
| 1057 | new_piece = ceph_msg_data_pagelist_advance(data, bytes); |
| 1058 | break; |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 1059 | case CEPH_MSG_DATA_PAGES: |
| 1060 | new_piece = ceph_msg_data_pages_advance(data, bytes); |
| 1061 | break; |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1062 | #ifdef CONFIG_BLOCK |
| 1063 | case CEPH_MSG_DATA_BIO: |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1064 | new_piece = ceph_msg_data_bio_advance(data, bytes); |
| 1065 | break; |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1066 | #endif /* CONFIG_BLOCK */ |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1067 | case CEPH_MSG_DATA_NONE: |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1068 | default: |
| 1069 | BUG(); |
| 1070 | break; |
| 1071 | } |
| 1072 | |
| 1073 | return new_piece; |
| 1074 | } |
| 1075 | |
Alex Elder | 7862505 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1076 | static void prepare_message_data(struct ceph_msg *msg, |
| 1077 | struct ceph_msg_pos *msg_pos) |
Alex Elder | 739c905 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1078 | { |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 1079 | size_t data_len; |
| 1080 | |
Alex Elder | 739c905 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1081 | BUG_ON(!msg); |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 1082 | |
| 1083 | data_len = le32_to_cpu(msg->hdr.data_len); |
| 1084 | BUG_ON(!data_len); |
Alex Elder | 739c905 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1085 | |
| 1086 | /* initialize page iterator */ |
Alex Elder | bae6acd | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1087 | msg_pos->page = 0; |
Alex Elder | 97fb1c7 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 1088 | if (ceph_msg_has_pages(msg)) |
Alex Elder | f9e1577 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 1089 | msg_pos->page_pos = msg->p.alignment; |
Alex Elder | 739c905 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1090 | else |
Alex Elder | bae6acd | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1091 | msg_pos->page_pos = 0; |
Alex Elder | bae6acd | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1092 | msg_pos->data_pos = 0; |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1093 | |
Alex Elder | 7fe1e5e | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1094 | /* Initialize data cursors */ |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1095 | |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1096 | #ifdef CONFIG_BLOCK |
| 1097 | if (ceph_msg_has_bio(msg)) |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 1098 | ceph_msg_data_cursor_init(&msg->b, data_len); |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1099 | #endif /* CONFIG_BLOCK */ |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 1100 | if (ceph_msg_has_pages(msg)) |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 1101 | ceph_msg_data_cursor_init(&msg->p, data_len); |
Alex Elder | 7fe1e5e | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1102 | if (ceph_msg_has_pagelist(msg)) |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 1103 | ceph_msg_data_cursor_init(&msg->l, data_len); |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1104 | |
Alex Elder | bae6acd | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1105 | msg_pos->did_page_crc = false; |
Alex Elder | 739c905 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1106 | } |
| 1107 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1108 | /* |
| 1109 | * Prepare footer for currently outgoing message, and finish things |
| 1110 | * off. Assumes out_kvec* are already valid.. we just add on to the end. |
| 1111 | */ |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1112 | static void prepare_write_message_footer(struct ceph_connection *con) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1113 | { |
| 1114 | struct ceph_msg *m = con->out_msg; |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1115 | int v = con->out_kvec_left; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1116 | |
Alex Elder | fd154f3 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1117 | m->footer.flags |= CEPH_MSG_FOOTER_COMPLETE; |
| 1118 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1119 | dout("prepare_write_message_footer %p\n", con); |
| 1120 | con->out_kvec_is_msg = true; |
| 1121 | con->out_kvec[v].iov_base = &m->footer; |
| 1122 | con->out_kvec[v].iov_len = sizeof(m->footer); |
| 1123 | con->out_kvec_bytes += sizeof(m->footer); |
| 1124 | con->out_kvec_left++; |
| 1125 | con->out_more = m->more_to_follow; |
Sage Weil | c86a293 | 2009-12-14 14:04:30 -0800 | [diff] [blame] | 1126 | con->out_msg_done = true; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1127 | } |
| 1128 | |
| 1129 | /* |
| 1130 | * Prepare headers for the next outgoing message. |
| 1131 | */ |
| 1132 | static void prepare_write_message(struct ceph_connection *con) |
| 1133 | { |
| 1134 | struct ceph_msg *m; |
Alex Elder | a9a0c51 | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 1135 | u32 crc; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1136 | |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1137 | con_out_kvec_reset(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1138 | con->out_kvec_is_msg = true; |
Sage Weil | c86a293 | 2009-12-14 14:04:30 -0800 | [diff] [blame] | 1139 | con->out_msg_done = false; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1140 | |
| 1141 | /* Sneak an ack in there first? If we can get it into the same |
| 1142 | * TCP packet that's a good thing. */ |
| 1143 | if (con->in_seq > con->in_seq_acked) { |
| 1144 | con->in_seq_acked = con->in_seq; |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1145 | con_out_kvec_add(con, sizeof (tag_ack), &tag_ack); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1146 | con->out_temp_ack = cpu_to_le64(con->in_seq_acked); |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1147 | con_out_kvec_add(con, sizeof (con->out_temp_ack), |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1148 | &con->out_temp_ack); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1149 | } |
| 1150 | |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 1151 | BUG_ON(list_empty(&con->out_queue)); |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1152 | m = list_first_entry(&con->out_queue, struct ceph_msg, list_head); |
Sage Weil | c86a293 | 2009-12-14 14:04:30 -0800 | [diff] [blame] | 1153 | con->out_msg = m; |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 1154 | BUG_ON(m->con != con); |
Sage Weil | 4cf9d54 | 2011-07-26 11:27:24 -0700 | [diff] [blame] | 1155 | |
| 1156 | /* put message on sent list */ |
| 1157 | ceph_msg_get(m); |
| 1158 | list_move_tail(&m->list_head, &con->out_sent); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1159 | |
Sage Weil | e84346b | 2010-05-11 21:20:38 -0700 | [diff] [blame] | 1160 | /* |
| 1161 | * only assign outgoing seq # if we haven't sent this message |
| 1162 | * yet. if it is requeued, resend with it's original seq. |
| 1163 | */ |
| 1164 | if (m->needs_out_seq) { |
| 1165 | m->hdr.seq = cpu_to_le64(++con->out_seq); |
| 1166 | m->needs_out_seq = false; |
| 1167 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1168 | |
Alex Elder | 4a73ef2 | 2013-03-07 15:38:26 -0600 | [diff] [blame] | 1169 | 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] | 1170 | m, con->out_seq, le16_to_cpu(m->hdr.type), |
| 1171 | le32_to_cpu(m->hdr.front_len), le32_to_cpu(m->hdr.middle_len), |
Alex Elder | f9e1577 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 1172 | le32_to_cpu(m->hdr.data_len), m->p.length); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1173 | BUG_ON(le32_to_cpu(m->hdr.front_len) != m->front.iov_len); |
| 1174 | |
| 1175 | /* tag + hdr + front + middle */ |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1176 | con_out_kvec_add(con, sizeof (tag_msg), &tag_msg); |
| 1177 | con_out_kvec_add(con, sizeof (m->hdr), &m->hdr); |
| 1178 | 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] | 1179 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1180 | if (m->middle) |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1181 | con_out_kvec_add(con, m->middle->vec.iov_len, |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1182 | m->middle->vec.iov_base); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1183 | |
| 1184 | /* fill in crc (except data pages), footer */ |
Alex Elder | a9a0c51 | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 1185 | crc = crc32c(0, &m->hdr, offsetof(struct ceph_msg_header, crc)); |
| 1186 | con->out_msg->hdr.crc = cpu_to_le32(crc); |
Alex Elder | fd154f3 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1187 | con->out_msg->footer.flags = 0; |
Alex Elder | a9a0c51 | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 1188 | |
| 1189 | crc = crc32c(0, m->front.iov_base, m->front.iov_len); |
| 1190 | con->out_msg->footer.front_crc = cpu_to_le32(crc); |
| 1191 | if (m->middle) { |
| 1192 | crc = crc32c(0, m->middle->vec.iov_base, |
| 1193 | m->middle->vec.iov_len); |
| 1194 | con->out_msg->footer.middle_crc = cpu_to_le32(crc); |
| 1195 | } else |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1196 | con->out_msg->footer.middle_crc = 0; |
Alex Elder | 739c905 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1197 | dout("%s front_crc %u middle_crc %u\n", __func__, |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1198 | le32_to_cpu(con->out_msg->footer.front_crc), |
| 1199 | le32_to_cpu(con->out_msg->footer.middle_crc)); |
| 1200 | |
| 1201 | /* is there a data payload? */ |
Alex Elder | 739c905 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1202 | con->out_msg->footer.data_crc = 0; |
Alex Elder | 7862505 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1203 | if (m->hdr.data_len) { |
| 1204 | prepare_message_data(con->out_msg, &con->out_msg_pos); |
| 1205 | con->out_more = 1; /* data + footer will follow */ |
| 1206 | } else { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1207 | /* no, queue up footer too and be done */ |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1208 | prepare_write_message_footer(con); |
Alex Elder | 7862505 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1209 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1210 | |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 1211 | con_flag_set(con, CON_FLAG_WRITE_PENDING); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1212 | } |
| 1213 | |
| 1214 | /* |
| 1215 | * Prepare an ack. |
| 1216 | */ |
| 1217 | static void prepare_write_ack(struct ceph_connection *con) |
| 1218 | { |
| 1219 | dout("prepare_write_ack %p %llu -> %llu\n", con, |
| 1220 | con->in_seq_acked, con->in_seq); |
| 1221 | con->in_seq_acked = con->in_seq; |
| 1222 | |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1223 | con_out_kvec_reset(con); |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1224 | |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1225 | con_out_kvec_add(con, sizeof (tag_ack), &tag_ack); |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1226 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1227 | con->out_temp_ack = cpu_to_le64(con->in_seq_acked); |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1228 | con_out_kvec_add(con, sizeof (con->out_temp_ack), |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1229 | &con->out_temp_ack); |
| 1230 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1231 | con->out_more = 1; /* more will follow.. eventually.. */ |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 1232 | con_flag_set(con, CON_FLAG_WRITE_PENDING); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1233 | } |
| 1234 | |
| 1235 | /* |
Sage Weil | 3a23083 | 2013-03-25 08:47:40 -0700 | [diff] [blame] | 1236 | * Prepare to share the seq during handshake |
| 1237 | */ |
| 1238 | static void prepare_write_seq(struct ceph_connection *con) |
| 1239 | { |
| 1240 | dout("prepare_write_seq %p %llu -> %llu\n", con, |
| 1241 | con->in_seq_acked, con->in_seq); |
| 1242 | con->in_seq_acked = con->in_seq; |
| 1243 | |
| 1244 | con_out_kvec_reset(con); |
| 1245 | |
| 1246 | con->out_temp_ack = cpu_to_le64(con->in_seq_acked); |
| 1247 | con_out_kvec_add(con, sizeof (con->out_temp_ack), |
| 1248 | &con->out_temp_ack); |
| 1249 | |
| 1250 | con_flag_set(con, CON_FLAG_WRITE_PENDING); |
| 1251 | } |
| 1252 | |
| 1253 | /* |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1254 | * Prepare to write keepalive byte. |
| 1255 | */ |
| 1256 | static void prepare_write_keepalive(struct ceph_connection *con) |
| 1257 | { |
| 1258 | dout("prepare_write_keepalive %p\n", con); |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1259 | con_out_kvec_reset(con); |
| 1260 | con_out_kvec_add(con, sizeof (tag_keepalive), &tag_keepalive); |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 1261 | con_flag_set(con, CON_FLAG_WRITE_PENDING); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1262 | } |
| 1263 | |
| 1264 | /* |
| 1265 | * Connection negotiation. |
| 1266 | */ |
| 1267 | |
Alex Elder | dac1e71 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1268 | static struct ceph_auth_handshake *get_connect_authorizer(struct ceph_connection *con, |
| 1269 | int *auth_proto) |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1270 | { |
Alex Elder | a3530df | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1271 | struct ceph_auth_handshake *auth; |
Alex Elder | b1c6b98 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 1272 | |
| 1273 | if (!con->ops->get_authorizer) { |
| 1274 | con->out_connect.authorizer_protocol = CEPH_AUTH_UNKNOWN; |
| 1275 | con->out_connect.authorizer_len = 0; |
Alex Elder | 729796b | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1276 | return NULL; |
Alex Elder | b1c6b98 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 1277 | } |
| 1278 | |
| 1279 | /* Can't hold the mutex while getting authorizer */ |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 1280 | mutex_unlock(&con->mutex); |
Alex Elder | dac1e71 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1281 | auth = con->ops->get_authorizer(con, auth_proto, con->auth_retry); |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 1282 | mutex_lock(&con->mutex); |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1283 | |
Alex Elder | a3530df | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1284 | if (IS_ERR(auth)) |
Alex Elder | 729796b | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1285 | return auth; |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 1286 | if (con->state != CON_STATE_NEGOTIATING) |
Alex Elder | 729796b | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1287 | return ERR_PTR(-EAGAIN); |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 1288 | |
Alex Elder | 8f43fb5 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1289 | con->auth_reply_buf = auth->authorizer_reply_buf; |
| 1290 | con->auth_reply_buf_len = auth->authorizer_reply_buf_len; |
Alex Elder | 729796b | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1291 | return auth; |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1292 | } |
| 1293 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1294 | /* |
| 1295 | * We connected to a peer and are saying hello. |
| 1296 | */ |
Alex Elder | e825a66 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 1297 | static void prepare_write_banner(struct ceph_connection *con) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1298 | { |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1299 | con_out_kvec_add(con, strlen(CEPH_BANNER), CEPH_BANNER); |
| 1300 | con_out_kvec_add(con, sizeof (con->msgr->my_enc_addr), |
Alex Elder | e825a66 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 1301 | &con->msgr->my_enc_addr); |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1302 | |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1303 | con->out_more = 0; |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 1304 | con_flag_set(con, CON_FLAG_WRITE_PENDING); |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1305 | } |
| 1306 | |
Alex Elder | e825a66 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 1307 | static int prepare_write_connect(struct ceph_connection *con) |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1308 | { |
Eric Dumazet | 95c9617 | 2012-04-15 05:58:06 +0000 | [diff] [blame] | 1309 | unsigned int global_seq = get_global_seq(con->msgr, 0); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1310 | int proto; |
Alex Elder | dac1e71 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1311 | int auth_proto; |
Alex Elder | 729796b | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1312 | struct ceph_auth_handshake *auth; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1313 | |
| 1314 | switch (con->peer_name.type) { |
| 1315 | case CEPH_ENTITY_TYPE_MON: |
| 1316 | proto = CEPH_MONC_PROTOCOL; |
| 1317 | break; |
| 1318 | case CEPH_ENTITY_TYPE_OSD: |
| 1319 | proto = CEPH_OSDC_PROTOCOL; |
| 1320 | break; |
| 1321 | case CEPH_ENTITY_TYPE_MDS: |
| 1322 | proto = CEPH_MDSC_PROTOCOL; |
| 1323 | break; |
| 1324 | default: |
| 1325 | BUG(); |
| 1326 | } |
| 1327 | |
| 1328 | dout("prepare_write_connect %p cseq=%d gseq=%d proto=%d\n", con, |
| 1329 | con->connect_seq, global_seq, proto); |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1330 | |
Alex Elder | e825a66 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 1331 | con->out_connect.features = cpu_to_le64(con->msgr->supported_features); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1332 | con->out_connect.host_type = cpu_to_le32(CEPH_ENTITY_TYPE_CLIENT); |
| 1333 | con->out_connect.connect_seq = cpu_to_le32(con->connect_seq); |
| 1334 | con->out_connect.global_seq = cpu_to_le32(global_seq); |
| 1335 | con->out_connect.protocol_version = cpu_to_le32(proto); |
| 1336 | con->out_connect.flags = 0; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1337 | |
Alex Elder | dac1e71 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1338 | auth_proto = CEPH_AUTH_UNKNOWN; |
| 1339 | auth = get_connect_authorizer(con, &auth_proto); |
Alex Elder | 729796b | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1340 | if (IS_ERR(auth)) |
| 1341 | return PTR_ERR(auth); |
Alex Elder | 3da5477 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1342 | |
Alex Elder | dac1e71 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1343 | con->out_connect.authorizer_protocol = cpu_to_le32(auth_proto); |
Alex Elder | 3da5477 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1344 | con->out_connect.authorizer_len = auth ? |
| 1345 | cpu_to_le32(auth->authorizer_buf_len) : 0; |
| 1346 | |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1347 | con_out_kvec_add(con, sizeof (con->out_connect), |
Alex Elder | 3da5477 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1348 | &con->out_connect); |
| 1349 | if (auth && auth->authorizer_buf_len) |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1350 | con_out_kvec_add(con, auth->authorizer_buf_len, |
Alex Elder | 3da5477 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1351 | auth->authorizer_buf); |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1352 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1353 | con->out_more = 0; |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 1354 | con_flag_set(con, CON_FLAG_WRITE_PENDING); |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1355 | |
Alex Elder | e10c758 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 1356 | return 0; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1357 | } |
| 1358 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1359 | /* |
| 1360 | * write as much of pending kvecs to the socket as we can. |
| 1361 | * 1 -> done |
| 1362 | * 0 -> socket full, but more to do |
| 1363 | * <0 -> error |
| 1364 | */ |
| 1365 | static int write_partial_kvec(struct ceph_connection *con) |
| 1366 | { |
| 1367 | int ret; |
| 1368 | |
| 1369 | dout("write_partial_kvec %p %d left\n", con, con->out_kvec_bytes); |
| 1370 | while (con->out_kvec_bytes > 0) { |
| 1371 | ret = ceph_tcp_sendmsg(con->sock, con->out_kvec_cur, |
| 1372 | con->out_kvec_left, con->out_kvec_bytes, |
| 1373 | con->out_more); |
| 1374 | if (ret <= 0) |
| 1375 | goto out; |
| 1376 | con->out_kvec_bytes -= ret; |
| 1377 | if (con->out_kvec_bytes == 0) |
| 1378 | break; /* done */ |
Alex Elder | f42299e | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 1379 | |
| 1380 | /* account for full iov entries consumed */ |
| 1381 | while (ret >= con->out_kvec_cur->iov_len) { |
| 1382 | BUG_ON(!con->out_kvec_left); |
| 1383 | ret -= con->out_kvec_cur->iov_len; |
| 1384 | con->out_kvec_cur++; |
| 1385 | con->out_kvec_left--; |
| 1386 | } |
| 1387 | /* and for a partially-consumed entry */ |
| 1388 | if (ret) { |
| 1389 | con->out_kvec_cur->iov_len -= ret; |
| 1390 | con->out_kvec_cur->iov_base += ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1391 | } |
| 1392 | } |
| 1393 | con->out_kvec_left = 0; |
| 1394 | con->out_kvec_is_msg = false; |
| 1395 | ret = 1; |
| 1396 | out: |
| 1397 | dout("write_partial_kvec %p %d left in %d kvecs ret = %d\n", con, |
| 1398 | con->out_kvec_bytes, con->out_kvec_left, ret); |
| 1399 | return ret; /* done! */ |
| 1400 | } |
| 1401 | |
Alex Elder | 84ca8fc | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1402 | static void out_msg_pos_next(struct ceph_connection *con, struct page *page, |
Alex Elder | 9d2a06c | 2013-03-08 13:35:36 -0600 | [diff] [blame] | 1403 | size_t len, size_t sent) |
Alex Elder | 84ca8fc | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1404 | { |
| 1405 | struct ceph_msg *msg = con->out_msg; |
Alex Elder | bae6acd | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1406 | struct ceph_msg_pos *msg_pos = &con->out_msg_pos; |
Alex Elder | 7fe1e5e | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1407 | bool need_crc = false; |
Alex Elder | 84ca8fc | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1408 | |
| 1409 | BUG_ON(!msg); |
| 1410 | BUG_ON(!sent); |
| 1411 | |
Alex Elder | bae6acd | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1412 | msg_pos->data_pos += sent; |
| 1413 | msg_pos->page_pos += sent; |
Alex Elder | 9d2a06c | 2013-03-08 13:35:36 -0600 | [diff] [blame] | 1414 | if (ceph_msg_has_pages(msg)) |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 1415 | need_crc = ceph_msg_data_advance(&msg->p, sent); |
Alex Elder | 7fe1e5e | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1416 | else if (ceph_msg_has_pagelist(msg)) |
| 1417 | need_crc = ceph_msg_data_advance(&msg->l, sent); |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1418 | #ifdef CONFIG_BLOCK |
| 1419 | else if (ceph_msg_has_bio(msg)) |
| 1420 | need_crc = ceph_msg_data_advance(&msg->b, sent); |
| 1421 | #endif /* CONFIG_BLOCK */ |
Alex Elder | 7fe1e5e | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1422 | BUG_ON(need_crc && sent != len); |
| 1423 | |
Alex Elder | 5821bd8 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1424 | if (sent < len) |
| 1425 | return; |
| 1426 | |
| 1427 | BUG_ON(sent != len); |
Alex Elder | bae6acd | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1428 | msg_pos->page_pos = 0; |
| 1429 | msg_pos->page++; |
| 1430 | msg_pos->did_page_crc = false; |
Alex Elder | 84ca8fc | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1431 | } |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1432 | |
Alex Elder | e788182 | 2013-03-08 18:51:04 -0600 | [diff] [blame] | 1433 | static void in_msg_pos_next(struct ceph_connection *con, size_t len, |
| 1434 | size_t received) |
| 1435 | { |
| 1436 | struct ceph_msg *msg = con->in_msg; |
Alex Elder | bae6acd | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1437 | struct ceph_msg_pos *msg_pos = &con->in_msg_pos; |
Alex Elder | e788182 | 2013-03-08 18:51:04 -0600 | [diff] [blame] | 1438 | |
| 1439 | BUG_ON(!msg); |
| 1440 | BUG_ON(!received); |
| 1441 | |
Alex Elder | bae6acd | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1442 | msg_pos->data_pos += received; |
| 1443 | msg_pos->page_pos += received; |
Alex Elder | 878efab | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 1444 | if (ceph_msg_has_pages(msg)) |
| 1445 | (void) ceph_msg_data_advance(&msg->p, received); |
Alex Elder | 463207a | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 1446 | #ifdef CONFIG_BLOCK |
Alex Elder | 878efab | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 1447 | else if (ceph_msg_has_bio(msg)) |
Alex Elder | 463207a | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 1448 | (void) ceph_msg_data_advance(&msg->b, received); |
| 1449 | #endif /* CONFIG_BLOCK */ |
Alex Elder | e788182 | 2013-03-08 18:51:04 -0600 | [diff] [blame] | 1450 | if (received < len) |
| 1451 | return; |
| 1452 | |
| 1453 | BUG_ON(received != len); |
Alex Elder | bae6acd | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1454 | msg_pos->page_pos = 0; |
| 1455 | msg_pos->page++; |
Alex Elder | e788182 | 2013-03-08 18:51:04 -0600 | [diff] [blame] | 1456 | } |
| 1457 | |
Alex Elder | 35b6280 | 2013-03-08 20:59:00 -0600 | [diff] [blame] | 1458 | static u32 ceph_crc32c_page(u32 crc, struct page *page, |
| 1459 | unsigned int page_offset, |
| 1460 | unsigned int length) |
| 1461 | { |
| 1462 | char *kaddr; |
| 1463 | |
| 1464 | kaddr = kmap(page); |
| 1465 | BUG_ON(kaddr == NULL); |
| 1466 | crc = crc32c(crc, kaddr + page_offset, length); |
| 1467 | kunmap(page); |
| 1468 | |
| 1469 | return crc; |
| 1470 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1471 | /* |
| 1472 | * Write as much message data payload as we can. If we finish, queue |
| 1473 | * up the footer. |
| 1474 | * 1 -> done, footer is now queued in out_kvec[]. |
| 1475 | * 0 -> socket full, but more to do |
| 1476 | * <0 -> error |
| 1477 | */ |
Alex Elder | 34d2d20 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 1478 | static int write_partial_message_data(struct ceph_connection *con) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1479 | { |
| 1480 | struct ceph_msg *msg = con->out_msg; |
Alex Elder | bae6acd | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1481 | struct ceph_msg_pos *msg_pos = &con->out_msg_pos; |
Eric Dumazet | 95c9617 | 2012-04-15 05:58:06 +0000 | [diff] [blame] | 1482 | unsigned int data_len = le32_to_cpu(msg->hdr.data_len); |
Alex Elder | 37675b0 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 1483 | bool do_datacrc = !con->msgr->nocrc; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1484 | int ret; |
| 1485 | |
Alex Elder | 34d2d20 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 1486 | dout("%s %p msg %p page %d offset %d\n", __func__, |
Alex Elder | bae6acd | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1487 | con, msg, msg_pos->page, msg_pos->page_pos); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1488 | |
Alex Elder | 5821bd8 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1489 | /* |
| 1490 | * Iterate through each page that contains data to be |
| 1491 | * written, and send as much as possible for each. |
| 1492 | * |
| 1493 | * If we are calculating the data crc (the default), we will |
| 1494 | * need to map the page. If we have no pages, they have |
| 1495 | * been revoked, so use the zero page. |
| 1496 | */ |
Alex Elder | bae6acd | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1497 | while (data_len > msg_pos->data_pos) { |
Alex Elder | 8a166d0 | 2013-03-08 13:35:36 -0600 | [diff] [blame] | 1498 | struct page *page; |
Alex Elder | e387d52 | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1499 | size_t page_offset; |
| 1500 | size_t length; |
Alex Elder | 8a166d0 | 2013-03-08 13:35:36 -0600 | [diff] [blame] | 1501 | bool last_piece; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1502 | |
Alex Elder | 9d2a06c | 2013-03-08 13:35:36 -0600 | [diff] [blame] | 1503 | if (ceph_msg_has_pages(msg)) { |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 1504 | page = ceph_msg_data_next(&msg->p, &page_offset, |
| 1505 | &length, &last_piece); |
Alex Elder | 97fb1c7 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 1506 | } else if (ceph_msg_has_pagelist(msg)) { |
Alex Elder | 7fe1e5e | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1507 | page = ceph_msg_data_next(&msg->l, &page_offset, |
| 1508 | &length, &last_piece); |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1509 | #ifdef CONFIG_BLOCK |
Alex Elder | 97fb1c7 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 1510 | } else if (ceph_msg_has_bio(msg)) { |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1511 | page = ceph_msg_data_next(&msg->b, &page_offset, |
| 1512 | &length, &last_piece); |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 1513 | #endif |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1514 | } else { |
Alex Elder | 61fcdc9 | 2013-03-11 23:34:22 -0500 | [diff] [blame^] | 1515 | WARN(1, "con %p data_len %u but no outbound data\n", |
| 1516 | con, data_len); |
| 1517 | ret = -EINVAL; |
| 1518 | goto out; |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1519 | } |
Alex Elder | bae6acd | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1520 | if (do_datacrc && !msg_pos->did_page_crc) { |
Alex Elder | 5821bd8 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1521 | u32 crc = le32_to_cpu(msg->footer.data_crc); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1522 | |
Alex Elder | 35b6280 | 2013-03-08 20:59:00 -0600 | [diff] [blame] | 1523 | crc = ceph_crc32c_page(crc, page, page_offset, length); |
Alex Elder | 84ca8fc | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1524 | msg->footer.data_crc = cpu_to_le32(crc); |
Alex Elder | bae6acd | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1525 | msg_pos->did_page_crc = true; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1526 | } |
Alex Elder | e387d52 | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1527 | ret = ceph_tcp_sendpage(con->sock, page, page_offset, |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1528 | length, last_piece); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1529 | if (ret <= 0) |
| 1530 | goto out; |
| 1531 | |
Alex Elder | 9d2a06c | 2013-03-08 13:35:36 -0600 | [diff] [blame] | 1532 | out_msg_pos_next(con, page, length, (size_t) ret); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1533 | } |
| 1534 | |
Alex Elder | 34d2d20 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 1535 | dout("%s %p msg %p done\n", __func__, con, msg); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1536 | |
| 1537 | /* prepare and queue up footer, too */ |
Alex Elder | 37675b0 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 1538 | if (!do_datacrc) |
Alex Elder | 84ca8fc | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1539 | msg->footer.flags |= CEPH_MSG_FOOTER_NOCRC; |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1540 | con_out_kvec_reset(con); |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1541 | prepare_write_message_footer(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1542 | ret = 1; |
| 1543 | out: |
| 1544 | return ret; |
| 1545 | } |
| 1546 | |
| 1547 | /* |
| 1548 | * write some zeros |
| 1549 | */ |
| 1550 | static int write_partial_skip(struct ceph_connection *con) |
| 1551 | { |
| 1552 | int ret; |
| 1553 | |
| 1554 | while (con->out_skip > 0) { |
Alex Elder | 3173913 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 1555 | size_t size = min(con->out_skip, (int) PAGE_CACHE_SIZE); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1556 | |
Alex Elder | e1dcb12 | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1557 | ret = ceph_tcp_sendpage(con->sock, zero_page, 0, size, true); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1558 | if (ret <= 0) |
| 1559 | goto out; |
| 1560 | con->out_skip -= ret; |
| 1561 | } |
| 1562 | ret = 1; |
| 1563 | out: |
| 1564 | return ret; |
| 1565 | } |
| 1566 | |
| 1567 | /* |
| 1568 | * Prepare to read connection handshake, or an ack. |
| 1569 | */ |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1570 | static void prepare_read_banner(struct ceph_connection *con) |
| 1571 | { |
| 1572 | dout("prepare_read_banner %p\n", con); |
| 1573 | con->in_base_pos = 0; |
| 1574 | } |
| 1575 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1576 | static void prepare_read_connect(struct ceph_connection *con) |
| 1577 | { |
| 1578 | dout("prepare_read_connect %p\n", con); |
| 1579 | con->in_base_pos = 0; |
| 1580 | } |
| 1581 | |
| 1582 | static void prepare_read_ack(struct ceph_connection *con) |
| 1583 | { |
| 1584 | dout("prepare_read_ack %p\n", con); |
| 1585 | con->in_base_pos = 0; |
| 1586 | } |
| 1587 | |
Sage Weil | 3a23083 | 2013-03-25 08:47:40 -0700 | [diff] [blame] | 1588 | static void prepare_read_seq(struct ceph_connection *con) |
| 1589 | { |
| 1590 | dout("prepare_read_seq %p\n", con); |
| 1591 | con->in_base_pos = 0; |
| 1592 | con->in_tag = CEPH_MSGR_TAG_SEQ; |
| 1593 | } |
| 1594 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1595 | static void prepare_read_tag(struct ceph_connection *con) |
| 1596 | { |
| 1597 | dout("prepare_read_tag %p\n", con); |
| 1598 | con->in_base_pos = 0; |
| 1599 | con->in_tag = CEPH_MSGR_TAG_READY; |
| 1600 | } |
| 1601 | |
| 1602 | /* |
| 1603 | * Prepare to read a message. |
| 1604 | */ |
| 1605 | static int prepare_read_message(struct ceph_connection *con) |
| 1606 | { |
| 1607 | dout("prepare_read_message %p\n", con); |
| 1608 | BUG_ON(con->in_msg != NULL); |
| 1609 | con->in_base_pos = 0; |
| 1610 | con->in_front_crc = con->in_middle_crc = con->in_data_crc = 0; |
| 1611 | return 0; |
| 1612 | } |
| 1613 | |
| 1614 | |
| 1615 | static int read_partial(struct ceph_connection *con, |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1616 | int end, int size, void *object) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1617 | { |
Alex Elder | e6cee71 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1618 | while (con->in_base_pos < end) { |
| 1619 | int left = end - con->in_base_pos; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1620 | int have = size - left; |
| 1621 | int ret = ceph_tcp_recvmsg(con->sock, object + have, left); |
| 1622 | if (ret <= 0) |
| 1623 | return ret; |
| 1624 | con->in_base_pos += ret; |
| 1625 | } |
| 1626 | return 1; |
| 1627 | } |
| 1628 | |
| 1629 | |
| 1630 | /* |
| 1631 | * Read all or part of the connect-side handshake on a new connection |
| 1632 | */ |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1633 | static int read_partial_banner(struct ceph_connection *con) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1634 | { |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1635 | int size; |
| 1636 | int end; |
| 1637 | int ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1638 | |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1639 | 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] | 1640 | |
| 1641 | /* peer's banner */ |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1642 | size = strlen(CEPH_BANNER); |
| 1643 | end = size; |
| 1644 | ret = read_partial(con, end, size, con->in_banner); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1645 | if (ret <= 0) |
| 1646 | goto out; |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1647 | |
| 1648 | size = sizeof (con->actual_peer_addr); |
| 1649 | end += size; |
| 1650 | ret = read_partial(con, end, size, &con->actual_peer_addr); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1651 | if (ret <= 0) |
| 1652 | goto out; |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1653 | |
| 1654 | size = sizeof (con->peer_addr_for_me); |
| 1655 | end += size; |
| 1656 | ret = read_partial(con, end, size, &con->peer_addr_for_me); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1657 | if (ret <= 0) |
| 1658 | goto out; |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1659 | |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1660 | out: |
| 1661 | return ret; |
| 1662 | } |
| 1663 | |
| 1664 | static int read_partial_connect(struct ceph_connection *con) |
| 1665 | { |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1666 | int size; |
| 1667 | int end; |
| 1668 | int ret; |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1669 | |
| 1670 | dout("read_partial_connect %p at %d\n", con, con->in_base_pos); |
| 1671 | |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1672 | size = sizeof (con->in_reply); |
| 1673 | end = size; |
| 1674 | ret = read_partial(con, end, size, &con->in_reply); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1675 | if (ret <= 0) |
| 1676 | goto out; |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1677 | |
| 1678 | size = le32_to_cpu(con->in_reply.authorizer_len); |
| 1679 | end += size; |
| 1680 | ret = read_partial(con, end, size, con->auth_reply_buf); |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1681 | if (ret <= 0) |
| 1682 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1683 | |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1684 | dout("read_partial_connect %p tag %d, con_seq = %u, g_seq = %u\n", |
| 1685 | con, (int)con->in_reply.tag, |
| 1686 | le32_to_cpu(con->in_reply.connect_seq), |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1687 | le32_to_cpu(con->in_reply.global_seq)); |
| 1688 | out: |
| 1689 | return ret; |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1690 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1691 | } |
| 1692 | |
| 1693 | /* |
| 1694 | * Verify the hello banner looks okay. |
| 1695 | */ |
| 1696 | static int verify_hello(struct ceph_connection *con) |
| 1697 | { |
| 1698 | if (memcmp(con->in_banner, CEPH_BANNER, strlen(CEPH_BANNER))) { |
Sage Weil | 13e38c8 | 2009-10-09 16:36:34 -0700 | [diff] [blame] | 1699 | pr_err("connect to %s got bad banner\n", |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1700 | ceph_pr_addr(&con->peer_addr.in_addr)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1701 | con->error_msg = "protocol error, bad banner"; |
| 1702 | return -1; |
| 1703 | } |
| 1704 | return 0; |
| 1705 | } |
| 1706 | |
| 1707 | static bool addr_is_blank(struct sockaddr_storage *ss) |
| 1708 | { |
| 1709 | switch (ss->ss_family) { |
| 1710 | case AF_INET: |
| 1711 | return ((struct sockaddr_in *)ss)->sin_addr.s_addr == 0; |
| 1712 | case AF_INET6: |
| 1713 | return |
| 1714 | ((struct sockaddr_in6 *)ss)->sin6_addr.s6_addr32[0] == 0 && |
| 1715 | ((struct sockaddr_in6 *)ss)->sin6_addr.s6_addr32[1] == 0 && |
| 1716 | ((struct sockaddr_in6 *)ss)->sin6_addr.s6_addr32[2] == 0 && |
| 1717 | ((struct sockaddr_in6 *)ss)->sin6_addr.s6_addr32[3] == 0; |
| 1718 | } |
| 1719 | return false; |
| 1720 | } |
| 1721 | |
| 1722 | static int addr_port(struct sockaddr_storage *ss) |
| 1723 | { |
| 1724 | switch (ss->ss_family) { |
| 1725 | case AF_INET: |
Sage Weil | f28bcfb | 2009-11-04 11:46:35 -0800 | [diff] [blame] | 1726 | return ntohs(((struct sockaddr_in *)ss)->sin_port); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1727 | case AF_INET6: |
Sage Weil | f28bcfb | 2009-11-04 11:46:35 -0800 | [diff] [blame] | 1728 | return ntohs(((struct sockaddr_in6 *)ss)->sin6_port); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1729 | } |
| 1730 | return 0; |
| 1731 | } |
| 1732 | |
| 1733 | static void addr_set_port(struct sockaddr_storage *ss, int p) |
| 1734 | { |
| 1735 | switch (ss->ss_family) { |
| 1736 | case AF_INET: |
| 1737 | ((struct sockaddr_in *)ss)->sin_port = htons(p); |
Sage Weil | a2a7960 | 2011-05-12 15:34:24 -0700 | [diff] [blame] | 1738 | break; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1739 | case AF_INET6: |
| 1740 | ((struct sockaddr_in6 *)ss)->sin6_port = htons(p); |
Sage Weil | a2a7960 | 2011-05-12 15:34:24 -0700 | [diff] [blame] | 1741 | break; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1742 | } |
| 1743 | } |
| 1744 | |
| 1745 | /* |
Noah Watkins | ee3b56f | 2011-09-23 11:48:42 -0700 | [diff] [blame] | 1746 | * Unlike other *_pton function semantics, zero indicates success. |
| 1747 | */ |
| 1748 | static int ceph_pton(const char *str, size_t len, struct sockaddr_storage *ss, |
| 1749 | char delim, const char **ipend) |
| 1750 | { |
Alex Elder | 99f0f3b | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 1751 | struct sockaddr_in *in4 = (struct sockaddr_in *) ss; |
| 1752 | struct sockaddr_in6 *in6 = (struct sockaddr_in6 *) ss; |
Noah Watkins | ee3b56f | 2011-09-23 11:48:42 -0700 | [diff] [blame] | 1753 | |
| 1754 | memset(ss, 0, sizeof(*ss)); |
| 1755 | |
| 1756 | if (in4_pton(str, len, (u8 *)&in4->sin_addr.s_addr, delim, ipend)) { |
| 1757 | ss->ss_family = AF_INET; |
| 1758 | return 0; |
| 1759 | } |
| 1760 | |
| 1761 | if (in6_pton(str, len, (u8 *)&in6->sin6_addr.s6_addr, delim, ipend)) { |
| 1762 | ss->ss_family = AF_INET6; |
| 1763 | return 0; |
| 1764 | } |
| 1765 | |
| 1766 | return -EINVAL; |
| 1767 | } |
| 1768 | |
| 1769 | /* |
| 1770 | * Extract hostname string and resolve using kernel DNS facility. |
| 1771 | */ |
| 1772 | #ifdef CONFIG_CEPH_LIB_USE_DNS_RESOLVER |
| 1773 | static int ceph_dns_resolve_name(const char *name, size_t namelen, |
| 1774 | struct sockaddr_storage *ss, char delim, const char **ipend) |
| 1775 | { |
| 1776 | const char *end, *delim_p; |
| 1777 | char *colon_p, *ip_addr = NULL; |
| 1778 | int ip_len, ret; |
| 1779 | |
| 1780 | /* |
| 1781 | * The end of the hostname occurs immediately preceding the delimiter or |
| 1782 | * the port marker (':') where the delimiter takes precedence. |
| 1783 | */ |
| 1784 | delim_p = memchr(name, delim, namelen); |
| 1785 | colon_p = memchr(name, ':', namelen); |
| 1786 | |
| 1787 | if (delim_p && colon_p) |
| 1788 | end = delim_p < colon_p ? delim_p : colon_p; |
| 1789 | else if (!delim_p && colon_p) |
| 1790 | end = colon_p; |
| 1791 | else { |
| 1792 | end = delim_p; |
| 1793 | if (!end) /* case: hostname:/ */ |
| 1794 | end = name + namelen; |
| 1795 | } |
| 1796 | |
| 1797 | if (end <= name) |
| 1798 | return -EINVAL; |
| 1799 | |
| 1800 | /* do dns_resolve upcall */ |
| 1801 | ip_len = dns_query(NULL, name, end - name, NULL, &ip_addr, NULL); |
| 1802 | if (ip_len > 0) |
| 1803 | ret = ceph_pton(ip_addr, ip_len, ss, -1, NULL); |
| 1804 | else |
| 1805 | ret = -ESRCH; |
| 1806 | |
| 1807 | kfree(ip_addr); |
| 1808 | |
| 1809 | *ipend = end; |
| 1810 | |
| 1811 | pr_info("resolve '%.*s' (ret=%d): %s\n", (int)(end - name), name, |
| 1812 | ret, ret ? "failed" : ceph_pr_addr(ss)); |
| 1813 | |
| 1814 | return ret; |
| 1815 | } |
| 1816 | #else |
| 1817 | static inline int ceph_dns_resolve_name(const char *name, size_t namelen, |
| 1818 | struct sockaddr_storage *ss, char delim, const char **ipend) |
| 1819 | { |
| 1820 | return -EINVAL; |
| 1821 | } |
| 1822 | #endif |
| 1823 | |
| 1824 | /* |
| 1825 | * Parse a server name (IP or hostname). If a valid IP address is not found |
| 1826 | * then try to extract a hostname to resolve using userspace DNS upcall. |
| 1827 | */ |
| 1828 | static int ceph_parse_server_name(const char *name, size_t namelen, |
| 1829 | struct sockaddr_storage *ss, char delim, const char **ipend) |
| 1830 | { |
| 1831 | int ret; |
| 1832 | |
| 1833 | ret = ceph_pton(name, namelen, ss, delim, ipend); |
| 1834 | if (ret) |
| 1835 | ret = ceph_dns_resolve_name(name, namelen, ss, delim, ipend); |
| 1836 | |
| 1837 | return ret; |
| 1838 | } |
| 1839 | |
| 1840 | /* |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1841 | * Parse an ip[:port] list into an addr array. Use the default |
| 1842 | * monitor port if a port isn't specified. |
| 1843 | */ |
| 1844 | int ceph_parse_ips(const char *c, const char *end, |
| 1845 | struct ceph_entity_addr *addr, |
| 1846 | int max_count, int *count) |
| 1847 | { |
Noah Watkins | ee3b56f | 2011-09-23 11:48:42 -0700 | [diff] [blame] | 1848 | int i, ret = -EINVAL; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1849 | const char *p = c; |
| 1850 | |
| 1851 | dout("parse_ips on '%.*s'\n", (int)(end-c), c); |
| 1852 | for (i = 0; i < max_count; i++) { |
| 1853 | const char *ipend; |
| 1854 | struct sockaddr_storage *ss = &addr[i].in_addr; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1855 | int port; |
Sage Weil | 39139f6 | 2010-07-08 09:54:52 -0700 | [diff] [blame] | 1856 | char delim = ','; |
| 1857 | |
| 1858 | if (*p == '[') { |
| 1859 | delim = ']'; |
| 1860 | p++; |
| 1861 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1862 | |
Noah Watkins | ee3b56f | 2011-09-23 11:48:42 -0700 | [diff] [blame] | 1863 | ret = ceph_parse_server_name(p, end - p, ss, delim, &ipend); |
| 1864 | if (ret) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1865 | goto bad; |
Noah Watkins | ee3b56f | 2011-09-23 11:48:42 -0700 | [diff] [blame] | 1866 | ret = -EINVAL; |
| 1867 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1868 | p = ipend; |
| 1869 | |
Sage Weil | 39139f6 | 2010-07-08 09:54:52 -0700 | [diff] [blame] | 1870 | if (delim == ']') { |
| 1871 | if (*p != ']') { |
| 1872 | dout("missing matching ']'\n"); |
| 1873 | goto bad; |
| 1874 | } |
| 1875 | p++; |
| 1876 | } |
| 1877 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1878 | /* port? */ |
| 1879 | if (p < end && *p == ':') { |
| 1880 | port = 0; |
| 1881 | p++; |
| 1882 | while (p < end && *p >= '0' && *p <= '9') { |
| 1883 | port = (port * 10) + (*p - '0'); |
| 1884 | p++; |
| 1885 | } |
| 1886 | if (port > 65535 || port == 0) |
| 1887 | goto bad; |
| 1888 | } else { |
| 1889 | port = CEPH_MON_PORT; |
| 1890 | } |
| 1891 | |
| 1892 | addr_set_port(ss, port); |
| 1893 | |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1894 | dout("parse_ips got %s\n", ceph_pr_addr(ss)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1895 | |
| 1896 | if (p == end) |
| 1897 | break; |
| 1898 | if (*p != ',') |
| 1899 | goto bad; |
| 1900 | p++; |
| 1901 | } |
| 1902 | |
| 1903 | if (p != end) |
| 1904 | goto bad; |
| 1905 | |
| 1906 | if (count) |
| 1907 | *count = i + 1; |
| 1908 | return 0; |
| 1909 | |
| 1910 | bad: |
Sage Weil | 39139f6 | 2010-07-08 09:54:52 -0700 | [diff] [blame] | 1911 | pr_err("parse_ips bad ip '%.*s'\n", (int)(end - c), c); |
Noah Watkins | ee3b56f | 2011-09-23 11:48:42 -0700 | [diff] [blame] | 1912 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1913 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1914 | EXPORT_SYMBOL(ceph_parse_ips); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1915 | |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1916 | static int process_banner(struct ceph_connection *con) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1917 | { |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1918 | dout("process_banner on %p\n", con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1919 | |
| 1920 | if (verify_hello(con) < 0) |
| 1921 | return -1; |
| 1922 | |
Sage Weil | 63f2d21 | 2009-11-03 15:17:56 -0800 | [diff] [blame] | 1923 | ceph_decode_addr(&con->actual_peer_addr); |
| 1924 | ceph_decode_addr(&con->peer_addr_for_me); |
| 1925 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1926 | /* |
| 1927 | * Make sure the other end is who we wanted. note that the other |
| 1928 | * end may not yet know their ip address, so if it's 0.0.0.0, give |
| 1929 | * them the benefit of the doubt. |
| 1930 | */ |
Sage Weil | 103e2d3 | 2010-01-07 16:12:36 -0800 | [diff] [blame] | 1931 | if (memcmp(&con->peer_addr, &con->actual_peer_addr, |
| 1932 | sizeof(con->peer_addr)) != 0 && |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1933 | !(addr_is_blank(&con->actual_peer_addr.in_addr) && |
| 1934 | con->actual_peer_addr.nonce == con->peer_addr.nonce)) { |
Yehuda Sadeh | cd84db6 | 2010-06-11 16:58:48 -0700 | [diff] [blame] | 1935 | pr_warning("wrong peer, want %s/%d, got %s/%d\n", |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1936 | ceph_pr_addr(&con->peer_addr.in_addr), |
Yehuda Sadeh | cd84db6 | 2010-06-11 16:58:48 -0700 | [diff] [blame] | 1937 | (int)le32_to_cpu(con->peer_addr.nonce), |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1938 | ceph_pr_addr(&con->actual_peer_addr.in_addr), |
Yehuda Sadeh | cd84db6 | 2010-06-11 16:58:48 -0700 | [diff] [blame] | 1939 | (int)le32_to_cpu(con->actual_peer_addr.nonce)); |
Sage Weil | 58bb3b3 | 2009-12-23 12:12:31 -0800 | [diff] [blame] | 1940 | con->error_msg = "wrong peer at address"; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1941 | return -1; |
| 1942 | } |
| 1943 | |
| 1944 | /* |
| 1945 | * did we learn our address? |
| 1946 | */ |
| 1947 | if (addr_is_blank(&con->msgr->inst.addr.in_addr)) { |
| 1948 | int port = addr_port(&con->msgr->inst.addr.in_addr); |
| 1949 | |
| 1950 | memcpy(&con->msgr->inst.addr.in_addr, |
| 1951 | &con->peer_addr_for_me.in_addr, |
| 1952 | sizeof(con->peer_addr_for_me.in_addr)); |
| 1953 | addr_set_port(&con->msgr->inst.addr.in_addr, port); |
Sage Weil | 63f2d21 | 2009-11-03 15:17:56 -0800 | [diff] [blame] | 1954 | encode_my_addr(con->msgr); |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1955 | dout("process_banner learned my addr is %s\n", |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1956 | ceph_pr_addr(&con->msgr->inst.addr.in_addr)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1957 | } |
| 1958 | |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1959 | return 0; |
| 1960 | } |
| 1961 | |
| 1962 | static int process_connect(struct ceph_connection *con) |
| 1963 | { |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1964 | u64 sup_feat = con->msgr->supported_features; |
| 1965 | u64 req_feat = con->msgr->required_features; |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 1966 | u64 server_feat = le64_to_cpu(con->in_reply.features); |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 1967 | int ret; |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 1968 | |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1969 | dout("process_connect on %p tag %d\n", con, (int)con->in_tag); |
| 1970 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1971 | switch (con->in_reply.tag) { |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 1972 | case CEPH_MSGR_TAG_FEATURES: |
| 1973 | pr_err("%s%lld %s feature set mismatch," |
| 1974 | " my %llx < server's %llx, missing %llx\n", |
| 1975 | ENTITY_NAME(con->peer_name), |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1976 | ceph_pr_addr(&con->peer_addr.in_addr), |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 1977 | sup_feat, server_feat, server_feat & ~sup_feat); |
| 1978 | con->error_msg = "missing required protocol features"; |
Sage Weil | 0fa6ebc | 2012-12-27 20:27:04 -0600 | [diff] [blame] | 1979 | reset_connection(con); |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 1980 | return -1; |
| 1981 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1982 | case CEPH_MSGR_TAG_BADPROTOVER: |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1983 | pr_err("%s%lld %s protocol version mismatch," |
| 1984 | " my %d != server's %d\n", |
| 1985 | ENTITY_NAME(con->peer_name), |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1986 | ceph_pr_addr(&con->peer_addr.in_addr), |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1987 | le32_to_cpu(con->out_connect.protocol_version), |
| 1988 | le32_to_cpu(con->in_reply.protocol_version)); |
| 1989 | con->error_msg = "protocol version mismatch"; |
Sage Weil | 0fa6ebc | 2012-12-27 20:27:04 -0600 | [diff] [blame] | 1990 | reset_connection(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1991 | return -1; |
| 1992 | |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1993 | case CEPH_MSGR_TAG_BADAUTHORIZER: |
| 1994 | con->auth_retry++; |
| 1995 | dout("process_connect %p got BADAUTHORIZER attempt %d\n", con, |
| 1996 | con->auth_retry); |
| 1997 | if (con->auth_retry == 2) { |
| 1998 | con->error_msg = "connect authorization failure"; |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1999 | return -1; |
| 2000 | } |
Jim Schutt | 6d4221b | 2012-08-10 10:37:38 -0700 | [diff] [blame] | 2001 | con_out_kvec_reset(con); |
Alex Elder | e825a66 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 2002 | ret = prepare_write_connect(con); |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 2003 | if (ret < 0) |
| 2004 | return ret; |
Sage Weil | 63733a0 | 2010-03-15 15:47:22 -0700 | [diff] [blame] | 2005 | prepare_read_connect(con); |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 2006 | break; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2007 | |
| 2008 | case CEPH_MSGR_TAG_RESETSESSION: |
| 2009 | /* |
| 2010 | * If we connected with a large connect_seq but the peer |
| 2011 | * has no record of a session with us (no connection, or |
| 2012 | * connect_seq == 0), they will send RESETSESION to indicate |
| 2013 | * that they must have reset their session, and may have |
| 2014 | * dropped messages. |
| 2015 | */ |
| 2016 | dout("process_connect got RESET peer seq %u\n", |
Sage Weil | 5bdca4e | 2012-07-10 11:53:34 -0700 | [diff] [blame] | 2017 | le32_to_cpu(con->in_reply.connect_seq)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2018 | pr_err("%s%lld %s connection reset\n", |
| 2019 | ENTITY_NAME(con->peer_name), |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2020 | ceph_pr_addr(&con->peer_addr.in_addr)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2021 | reset_connection(con); |
Jim Schutt | 6d4221b | 2012-08-10 10:37:38 -0700 | [diff] [blame] | 2022 | con_out_kvec_reset(con); |
Alex Elder | 5a0f8fd | 2012-05-16 21:51:59 -0500 | [diff] [blame] | 2023 | ret = prepare_write_connect(con); |
| 2024 | if (ret < 0) |
| 2025 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2026 | prepare_read_connect(con); |
| 2027 | |
| 2028 | /* Tell ceph about it. */ |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 2029 | mutex_unlock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2030 | pr_info("reset on %s%lld\n", ENTITY_NAME(con->peer_name)); |
| 2031 | if (con->ops->peer_reset) |
| 2032 | con->ops->peer_reset(con); |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 2033 | mutex_lock(&con->mutex); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2034 | if (con->state != CON_STATE_NEGOTIATING) |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 2035 | return -EAGAIN; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2036 | break; |
| 2037 | |
| 2038 | case CEPH_MSGR_TAG_RETRY_SESSION: |
| 2039 | /* |
| 2040 | * If we sent a smaller connect_seq than the peer has, try |
| 2041 | * again with a larger value. |
| 2042 | */ |
Sage Weil | 5bdca4e | 2012-07-10 11:53:34 -0700 | [diff] [blame] | 2043 | dout("process_connect got RETRY_SESSION my seq %u, peer %u\n", |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2044 | le32_to_cpu(con->out_connect.connect_seq), |
Sage Weil | 5bdca4e | 2012-07-10 11:53:34 -0700 | [diff] [blame] | 2045 | le32_to_cpu(con->in_reply.connect_seq)); |
| 2046 | con->connect_seq = le32_to_cpu(con->in_reply.connect_seq); |
Jim Schutt | 6d4221b | 2012-08-10 10:37:38 -0700 | [diff] [blame] | 2047 | con_out_kvec_reset(con); |
Alex Elder | 5a0f8fd | 2012-05-16 21:51:59 -0500 | [diff] [blame] | 2048 | ret = prepare_write_connect(con); |
| 2049 | if (ret < 0) |
| 2050 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2051 | prepare_read_connect(con); |
| 2052 | break; |
| 2053 | |
| 2054 | case CEPH_MSGR_TAG_RETRY_GLOBAL: |
| 2055 | /* |
| 2056 | * If we sent a smaller global_seq than the peer has, try |
| 2057 | * again with a larger value. |
| 2058 | */ |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 2059 | dout("process_connect got RETRY_GLOBAL my %u peer_gseq %u\n", |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2060 | con->peer_global_seq, |
Sage Weil | 5bdca4e | 2012-07-10 11:53:34 -0700 | [diff] [blame] | 2061 | le32_to_cpu(con->in_reply.global_seq)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2062 | get_global_seq(con->msgr, |
Sage Weil | 5bdca4e | 2012-07-10 11:53:34 -0700 | [diff] [blame] | 2063 | le32_to_cpu(con->in_reply.global_seq)); |
Jim Schutt | 6d4221b | 2012-08-10 10:37:38 -0700 | [diff] [blame] | 2064 | con_out_kvec_reset(con); |
Alex Elder | 5a0f8fd | 2012-05-16 21:51:59 -0500 | [diff] [blame] | 2065 | ret = prepare_write_connect(con); |
| 2066 | if (ret < 0) |
| 2067 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2068 | prepare_read_connect(con); |
| 2069 | break; |
| 2070 | |
Sage Weil | 3a23083 | 2013-03-25 08:47:40 -0700 | [diff] [blame] | 2071 | case CEPH_MSGR_TAG_SEQ: |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2072 | case CEPH_MSGR_TAG_READY: |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 2073 | if (req_feat & ~server_feat) { |
| 2074 | pr_err("%s%lld %s protocol feature mismatch," |
| 2075 | " my required %llx > server's %llx, need %llx\n", |
| 2076 | ENTITY_NAME(con->peer_name), |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2077 | ceph_pr_addr(&con->peer_addr.in_addr), |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 2078 | req_feat, server_feat, req_feat & ~server_feat); |
| 2079 | con->error_msg = "missing required protocol features"; |
Sage Weil | 0fa6ebc | 2012-12-27 20:27:04 -0600 | [diff] [blame] | 2080 | reset_connection(con); |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 2081 | return -1; |
| 2082 | } |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2083 | |
Alex Elder | 122070a | 2012-12-26 10:43:57 -0600 | [diff] [blame] | 2084 | WARN_ON(con->state != CON_STATE_NEGOTIATING); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2085 | con->state = CON_STATE_OPEN; |
Sage Weil | 20e55c4 | 2013-03-25 09:30:13 -0700 | [diff] [blame] | 2086 | con->auth_retry = 0; /* we authenticated; clear flag */ |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2087 | con->peer_global_seq = le32_to_cpu(con->in_reply.global_seq); |
| 2088 | con->connect_seq++; |
Sage Weil | aba558e | 2010-05-12 15:23:30 -0700 | [diff] [blame] | 2089 | con->peer_features = server_feat; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2090 | dout("process_connect got READY gseq %d cseq %d (%d)\n", |
| 2091 | con->peer_global_seq, |
| 2092 | le32_to_cpu(con->in_reply.connect_seq), |
| 2093 | con->connect_seq); |
| 2094 | WARN_ON(con->connect_seq != |
| 2095 | le32_to_cpu(con->in_reply.connect_seq)); |
Sage Weil | 92ac41d | 2009-12-14 14:56:56 -0800 | [diff] [blame] | 2096 | |
| 2097 | if (con->in_reply.flags & CEPH_MSG_CONNECT_LOSSY) |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2098 | con_flag_set(con, CON_FLAG_LOSSYTX); |
Sage Weil | 92ac41d | 2009-12-14 14:56:56 -0800 | [diff] [blame] | 2099 | |
Sage Weil | 85effe1 | 2012-07-30 16:22:05 -0700 | [diff] [blame] | 2100 | con->delay = 0; /* reset backoff memory */ |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2101 | |
Sage Weil | 3a23083 | 2013-03-25 08:47:40 -0700 | [diff] [blame] | 2102 | if (con->in_reply.tag == CEPH_MSGR_TAG_SEQ) { |
| 2103 | prepare_write_seq(con); |
| 2104 | prepare_read_seq(con); |
| 2105 | } else { |
| 2106 | prepare_read_tag(con); |
| 2107 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2108 | break; |
| 2109 | |
| 2110 | case CEPH_MSGR_TAG_WAIT: |
| 2111 | /* |
| 2112 | * If there is a connection race (we are opening |
| 2113 | * connections to each other), one of us may just have |
| 2114 | * to WAIT. This shouldn't happen if we are the |
| 2115 | * client. |
| 2116 | */ |
Sage Weil | 0417788 | 2011-05-12 15:33:17 -0700 | [diff] [blame] | 2117 | pr_err("process_connect got WAIT as client\n"); |
| 2118 | con->error_msg = "protocol error, got WAIT as client"; |
| 2119 | return -1; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2120 | |
| 2121 | default: |
| 2122 | pr_err("connect protocol error, will retry\n"); |
| 2123 | con->error_msg = "protocol error, garbage tag during connect"; |
| 2124 | return -1; |
| 2125 | } |
| 2126 | return 0; |
| 2127 | } |
| 2128 | |
| 2129 | |
| 2130 | /* |
| 2131 | * read (part of) an ack |
| 2132 | */ |
| 2133 | static int read_partial_ack(struct ceph_connection *con) |
| 2134 | { |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 2135 | int size = sizeof (con->in_temp_ack); |
| 2136 | int end = size; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2137 | |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 2138 | return read_partial(con, end, size, &con->in_temp_ack); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2139 | } |
| 2140 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2141 | /* |
| 2142 | * We can finally discard anything that's been acked. |
| 2143 | */ |
| 2144 | static void process_ack(struct ceph_connection *con) |
| 2145 | { |
| 2146 | struct ceph_msg *m; |
| 2147 | u64 ack = le64_to_cpu(con->in_temp_ack); |
| 2148 | u64 seq; |
| 2149 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2150 | while (!list_empty(&con->out_sent)) { |
| 2151 | m = list_first_entry(&con->out_sent, struct ceph_msg, |
| 2152 | list_head); |
| 2153 | seq = le64_to_cpu(m->hdr.seq); |
| 2154 | if (seq > ack) |
| 2155 | break; |
| 2156 | dout("got ack for seq %llu type %d at %p\n", seq, |
| 2157 | le16_to_cpu(m->hdr.type), m); |
Sage Weil | 4cf9d54 | 2011-07-26 11:27:24 -0700 | [diff] [blame] | 2158 | m->ack_stamp = jiffies; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2159 | ceph_msg_remove(m); |
| 2160 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2161 | prepare_read_tag(con); |
| 2162 | } |
| 2163 | |
| 2164 | |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2165 | static int read_partial_message_section(struct ceph_connection *con, |
Sage Weil | 213c99e | 2010-08-03 10:25:11 -0700 | [diff] [blame] | 2166 | struct kvec *section, |
| 2167 | unsigned int sec_len, u32 *crc) |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2168 | { |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 2169 | int ret, left; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2170 | |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2171 | BUG_ON(!section); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2172 | |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2173 | while (section->iov_len < sec_len) { |
| 2174 | BUG_ON(section->iov_base == NULL); |
| 2175 | left = sec_len - section->iov_len; |
| 2176 | ret = ceph_tcp_recvmsg(con->sock, (char *)section->iov_base + |
| 2177 | section->iov_len, left); |
| 2178 | if (ret <= 0) |
| 2179 | return ret; |
| 2180 | section->iov_len += ret; |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2181 | } |
Alex Elder | fe3ad59 | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 2182 | if (section->iov_len == sec_len) |
| 2183 | *crc = crc32c(0, section->iov_base, section->iov_len); |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2184 | |
| 2185 | return 1; |
| 2186 | } |
| 2187 | |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 2188 | 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] | 2189 | |
| 2190 | static int read_partial_message_pages(struct ceph_connection *con, |
Eric Dumazet | 95c9617 | 2012-04-15 05:58:06 +0000 | [diff] [blame] | 2191 | unsigned int data_len, bool do_datacrc) |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 2192 | { |
Alex Elder | 28a89dd | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 2193 | struct ceph_msg *msg = con->in_msg; |
Alex Elder | e788182 | 2013-03-08 18:51:04 -0600 | [diff] [blame] | 2194 | struct page *page; |
Alex Elder | afb3d90 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2195 | size_t page_offset; |
| 2196 | size_t length; |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 2197 | int ret; |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 2198 | |
Alex Elder | 878efab | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 2199 | page = ceph_msg_data_next(&msg->p, &page_offset, &length, NULL); |
Alex Elder | afb3d90 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2200 | |
| 2201 | ret = ceph_tcp_recvpage(con->sock, page, page_offset, length); |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 2202 | if (ret <= 0) |
| 2203 | return ret; |
Alex Elder | e788182 | 2013-03-08 18:51:04 -0600 | [diff] [blame] | 2204 | |
Alex Elder | 35b6280 | 2013-03-08 20:59:00 -0600 | [diff] [blame] | 2205 | if (do_datacrc) |
| 2206 | con->in_data_crc = ceph_crc32c_page(con->in_data_crc, page, |
| 2207 | page_offset, ret); |
Alex Elder | afb3d90 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2208 | |
| 2209 | in_msg_pos_next(con, length, ret); |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 2210 | |
| 2211 | return ret; |
| 2212 | } |
| 2213 | |
| 2214 | #ifdef CONFIG_BLOCK |
| 2215 | static int read_partial_message_bio(struct ceph_connection *con, |
Eric Dumazet | 95c9617 | 2012-04-15 05:58:06 +0000 | [diff] [blame] | 2216 | unsigned int data_len, bool do_datacrc) |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 2217 | { |
Alex Elder | b3d56fa | 2013-03-08 18:51:03 -0600 | [diff] [blame] | 2218 | struct ceph_msg *msg = con->in_msg; |
Alex Elder | e788182 | 2013-03-08 18:51:04 -0600 | [diff] [blame] | 2219 | struct page *page; |
Alex Elder | afb3d90 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2220 | size_t page_offset; |
| 2221 | size_t length; |
Alex Elder | afb3d90 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2222 | int ret; |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 2223 | |
Alex Elder | b3d56fa | 2013-03-08 18:51:03 -0600 | [diff] [blame] | 2224 | BUG_ON(!msg); |
Alex Elder | 463207a | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 2225 | |
| 2226 | page = ceph_msg_data_next(&msg->b, &page_offset, &length, NULL); |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 2227 | |
Alex Elder | afb3d90 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2228 | ret = ceph_tcp_recvpage(con->sock, page, page_offset, length); |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 2229 | if (ret <= 0) |
| 2230 | return ret; |
Alex Elder | e788182 | 2013-03-08 18:51:04 -0600 | [diff] [blame] | 2231 | |
Alex Elder | 35b6280 | 2013-03-08 20:59:00 -0600 | [diff] [blame] | 2232 | if (do_datacrc) |
| 2233 | con->in_data_crc = ceph_crc32c_page(con->in_data_crc, page, |
| 2234 | page_offset, ret); |
Alex Elder | afb3d90 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2235 | |
| 2236 | in_msg_pos_next(con, length, ret); |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 2237 | |
| 2238 | return ret; |
| 2239 | } |
| 2240 | #endif |
| 2241 | |
Alex Elder | 34d2d20 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2242 | static int read_partial_msg_data(struct ceph_connection *con) |
| 2243 | { |
| 2244 | struct ceph_msg *msg = con->in_msg; |
| 2245 | struct ceph_msg_pos *msg_pos = &con->in_msg_pos; |
| 2246 | const bool do_datacrc = !con->msgr->nocrc; |
| 2247 | unsigned int data_len; |
| 2248 | int ret; |
| 2249 | |
| 2250 | BUG_ON(!msg); |
| 2251 | |
| 2252 | data_len = le32_to_cpu(con->in_hdr.data_len); |
| 2253 | while (msg_pos->data_pos < data_len) { |
Alex Elder | 97fb1c7 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 2254 | if (ceph_msg_has_pages(msg)) { |
Alex Elder | 28a89dd | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 2255 | ret = read_partial_message_pages(con, data_len, |
| 2256 | do_datacrc); |
Alex Elder | 34d2d20 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2257 | if (ret <= 0) |
| 2258 | return ret; |
| 2259 | #ifdef CONFIG_BLOCK |
Alex Elder | 97fb1c7 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 2260 | } else if (ceph_msg_has_bio(msg)) { |
Alex Elder | 34d2d20 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2261 | ret = read_partial_message_bio(con, |
| 2262 | data_len, do_datacrc); |
| 2263 | if (ret <= 0) |
| 2264 | return ret; |
| 2265 | #endif |
| 2266 | } else { |
| 2267 | BUG_ON(1); |
| 2268 | } |
| 2269 | } |
| 2270 | |
| 2271 | return 1; /* must return > 0 to indicate success */ |
| 2272 | } |
| 2273 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2274 | /* |
| 2275 | * read (part of) a message. |
| 2276 | */ |
| 2277 | static int read_partial_message(struct ceph_connection *con) |
| 2278 | { |
| 2279 | struct ceph_msg *m = con->in_msg; |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 2280 | int size; |
| 2281 | int end; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2282 | int ret; |
Eric Dumazet | 95c9617 | 2012-04-15 05:58:06 +0000 | [diff] [blame] | 2283 | unsigned int front_len, middle_len, data_len; |
Alex Elder | 37675b0 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 2284 | bool do_datacrc = !con->msgr->nocrc; |
Sage Weil | ae18756 | 2010-04-22 07:47:01 -0700 | [diff] [blame] | 2285 | u64 seq; |
Alex Elder | fe3ad59 | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 2286 | u32 crc; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2287 | |
| 2288 | dout("read_partial_message con %p msg %p\n", con, m); |
| 2289 | |
| 2290 | /* header */ |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 2291 | size = sizeof (con->in_hdr); |
| 2292 | end = size; |
| 2293 | ret = read_partial(con, end, size, &con->in_hdr); |
Alex Elder | 57dac9d | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 2294 | if (ret <= 0) |
| 2295 | return ret; |
Alex Elder | fe3ad59 | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 2296 | |
| 2297 | crc = crc32c(0, &con->in_hdr, offsetof(struct ceph_msg_header, crc)); |
| 2298 | if (cpu_to_le32(crc) != con->in_hdr.crc) { |
| 2299 | pr_err("read_partial_message bad hdr " |
| 2300 | " crc %u != expected %u\n", |
| 2301 | crc, con->in_hdr.crc); |
| 2302 | return -EBADMSG; |
| 2303 | } |
| 2304 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2305 | front_len = le32_to_cpu(con->in_hdr.front_len); |
| 2306 | if (front_len > CEPH_MSG_MAX_FRONT_LEN) |
| 2307 | return -EIO; |
| 2308 | middle_len = le32_to_cpu(con->in_hdr.middle_len); |
Alex Elder | 7b11ba3 | 2013-03-08 18:51:03 -0600 | [diff] [blame] | 2309 | if (middle_len > CEPH_MSG_MAX_MIDDLE_LEN) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2310 | return -EIO; |
| 2311 | data_len = le32_to_cpu(con->in_hdr.data_len); |
| 2312 | if (data_len > CEPH_MSG_MAX_DATA_LEN) |
| 2313 | return -EIO; |
| 2314 | |
Sage Weil | ae18756 | 2010-04-22 07:47:01 -0700 | [diff] [blame] | 2315 | /* verify seq# */ |
| 2316 | seq = le64_to_cpu(con->in_hdr.seq); |
| 2317 | if ((s64)seq - (s64)con->in_seq < 1) { |
Sage Weil | df9f86f | 2010-11-01 15:49:23 -0700 | [diff] [blame] | 2318 | pr_info("skipping %s%lld %s seq %lld expected %lld\n", |
Sage Weil | ae18756 | 2010-04-22 07:47:01 -0700 | [diff] [blame] | 2319 | ENTITY_NAME(con->peer_name), |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2320 | ceph_pr_addr(&con->peer_addr.in_addr), |
Sage Weil | ae18756 | 2010-04-22 07:47:01 -0700 | [diff] [blame] | 2321 | seq, con->in_seq + 1); |
| 2322 | con->in_base_pos = -front_len - middle_len - data_len - |
| 2323 | sizeof(m->footer); |
| 2324 | con->in_tag = CEPH_MSGR_TAG_READY; |
Sage Weil | ae18756 | 2010-04-22 07:47:01 -0700 | [diff] [blame] | 2325 | return 0; |
| 2326 | } else if ((s64)seq - (s64)con->in_seq > 1) { |
| 2327 | pr_err("read_partial_message bad seq %lld expected %lld\n", |
| 2328 | seq, con->in_seq + 1); |
| 2329 | con->error_msg = "bad message sequence # for incoming message"; |
| 2330 | return -EBADMSG; |
| 2331 | } |
| 2332 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2333 | /* allocate message? */ |
| 2334 | if (!con->in_msg) { |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 2335 | int skip = 0; |
| 2336 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2337 | 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] | 2338 | front_len, data_len); |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 2339 | ret = ceph_con_in_msg_alloc(con, &skip); |
| 2340 | if (ret < 0) |
| 2341 | return ret; |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2342 | if (skip) { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2343 | /* skip this message */ |
Sage Weil | a79832f | 2010-04-01 16:06:19 -0700 | [diff] [blame] | 2344 | dout("alloc_msg said skip message\n"); |
Sage Weil | ae32be3 | 2010-06-13 10:30:19 -0700 | [diff] [blame] | 2345 | BUG_ON(con->in_msg); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2346 | con->in_base_pos = -front_len - middle_len - data_len - |
| 2347 | sizeof(m->footer); |
| 2348 | con->in_tag = CEPH_MSGR_TAG_READY; |
Sage Weil | 684be25 | 2010-04-21 20:45:59 -0700 | [diff] [blame] | 2349 | con->in_seq++; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2350 | return 0; |
| 2351 | } |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2352 | |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 2353 | BUG_ON(!con->in_msg); |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2354 | BUG_ON(con->in_msg->con != con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2355 | m = con->in_msg; |
| 2356 | m->front.iov_len = 0; /* haven't read it yet */ |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2357 | if (m->middle) |
| 2358 | m->middle->vec.iov_len = 0; |
Yehuda Sadeh | 9d7f0f1 | 2010-01-11 10:32:02 -0800 | [diff] [blame] | 2359 | |
Alex Elder | 7862505 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 2360 | /* prepare for data payload, if any */ |
Sage Weil | a410702 | 2012-07-30 16:20:25 -0700 | [diff] [blame] | 2361 | |
Alex Elder | 7862505 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 2362 | if (data_len) |
| 2363 | prepare_message_data(con->in_msg, &con->in_msg_pos); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2364 | } |
| 2365 | |
| 2366 | /* front */ |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2367 | ret = read_partial_message_section(con, &m->front, front_len, |
| 2368 | &con->in_front_crc); |
| 2369 | if (ret <= 0) |
| 2370 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2371 | |
| 2372 | /* middle */ |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2373 | if (m->middle) { |
Sage Weil | 213c99e | 2010-08-03 10:25:11 -0700 | [diff] [blame] | 2374 | ret = read_partial_message_section(con, &m->middle->vec, |
| 2375 | middle_len, |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2376 | &con->in_middle_crc); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2377 | if (ret <= 0) |
| 2378 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2379 | } |
| 2380 | |
| 2381 | /* (page) data */ |
Alex Elder | 34d2d20 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2382 | if (data_len) { |
| 2383 | ret = read_partial_msg_data(con); |
| 2384 | if (ret <= 0) |
| 2385 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2386 | } |
| 2387 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2388 | /* footer */ |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 2389 | size = sizeof (m->footer); |
| 2390 | end += size; |
| 2391 | ret = read_partial(con, end, size, &m->footer); |
Alex Elder | 57dac9d | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 2392 | if (ret <= 0) |
| 2393 | return ret; |
| 2394 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2395 | dout("read_partial_message got msg %p %d (%u) + %d (%u) + %d (%u)\n", |
| 2396 | m, front_len, m->footer.front_crc, middle_len, |
| 2397 | m->footer.middle_crc, data_len, m->footer.data_crc); |
| 2398 | |
| 2399 | /* crc ok? */ |
| 2400 | if (con->in_front_crc != le32_to_cpu(m->footer.front_crc)) { |
| 2401 | pr_err("read_partial_message %p front crc %u != exp. %u\n", |
| 2402 | m, con->in_front_crc, m->footer.front_crc); |
| 2403 | return -EBADMSG; |
| 2404 | } |
| 2405 | if (con->in_middle_crc != le32_to_cpu(m->footer.middle_crc)) { |
| 2406 | pr_err("read_partial_message %p middle crc %u != exp %u\n", |
| 2407 | m, con->in_middle_crc, m->footer.middle_crc); |
| 2408 | return -EBADMSG; |
| 2409 | } |
Alex Elder | bca064d | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 2410 | if (do_datacrc && |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2411 | (m->footer.flags & CEPH_MSG_FOOTER_NOCRC) == 0 && |
| 2412 | con->in_data_crc != le32_to_cpu(m->footer.data_crc)) { |
| 2413 | pr_err("read_partial_message %p data crc %u != exp. %u\n", m, |
| 2414 | con->in_data_crc, le32_to_cpu(m->footer.data_crc)); |
| 2415 | return -EBADMSG; |
| 2416 | } |
| 2417 | |
| 2418 | return 1; /* done! */ |
| 2419 | } |
| 2420 | |
| 2421 | /* |
| 2422 | * Process message. This happens in the worker thread. The callback should |
| 2423 | * be careful not to do anything that waits on other incoming messages or it |
| 2424 | * may deadlock. |
| 2425 | */ |
| 2426 | static void process_message(struct ceph_connection *con) |
| 2427 | { |
Sage Weil | 5e095e8 | 2009-12-14 14:30:34 -0800 | [diff] [blame] | 2428 | struct ceph_msg *msg; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2429 | |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2430 | BUG_ON(con->in_msg->con != con); |
| 2431 | con->in_msg->con = NULL; |
Sage Weil | 5e095e8 | 2009-12-14 14:30:34 -0800 | [diff] [blame] | 2432 | msg = con->in_msg; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2433 | con->in_msg = NULL; |
Sage Weil | 36eb71a | 2012-06-21 12:47:08 -0700 | [diff] [blame] | 2434 | con->ops->put(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2435 | |
| 2436 | /* if first message, set peer_name */ |
| 2437 | if (con->peer_name.type == 0) |
Sage Weil | dbad185 | 2010-03-25 15:45:38 -0700 | [diff] [blame] | 2438 | con->peer_name = msg->hdr.src; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2439 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2440 | con->in_seq++; |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 2441 | mutex_unlock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2442 | |
| 2443 | dout("===== %p %llu from %s%lld %d=%s len %d+%d (%u %u %u) =====\n", |
| 2444 | msg, le64_to_cpu(msg->hdr.seq), |
Sage Weil | dbad185 | 2010-03-25 15:45:38 -0700 | [diff] [blame] | 2445 | ENTITY_NAME(msg->hdr.src), |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2446 | le16_to_cpu(msg->hdr.type), |
| 2447 | ceph_msg_type_name(le16_to_cpu(msg->hdr.type)), |
| 2448 | le32_to_cpu(msg->hdr.front_len), |
| 2449 | le32_to_cpu(msg->hdr.data_len), |
| 2450 | con->in_front_crc, con->in_middle_crc, con->in_data_crc); |
| 2451 | con->ops->dispatch(con, msg); |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 2452 | |
| 2453 | mutex_lock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2454 | } |
| 2455 | |
| 2456 | |
| 2457 | /* |
| 2458 | * Write something to the socket. Called in a worker thread when the |
| 2459 | * socket appears to be writeable and we have something ready to send. |
| 2460 | */ |
| 2461 | static int try_write(struct ceph_connection *con) |
| 2462 | { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2463 | int ret = 1; |
| 2464 | |
Sage Weil | d59315c | 2012-06-21 12:49:23 -0700 | [diff] [blame] | 2465 | dout("try_write start %p state %lu\n", con, con->state); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2466 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2467 | more: |
| 2468 | dout("try_write out_kvec_bytes %d\n", con->out_kvec_bytes); |
| 2469 | |
| 2470 | /* open the socket first? */ |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2471 | if (con->state == CON_STATE_PREOPEN) { |
| 2472 | BUG_ON(con->sock); |
| 2473 | con->state = CON_STATE_CONNECTING; |
Alex Elder | a5988c4 | 2012-05-29 11:04:58 -0500 | [diff] [blame] | 2474 | |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 2475 | con_out_kvec_reset(con); |
Alex Elder | e825a66 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 2476 | prepare_write_banner(con); |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 2477 | prepare_read_banner(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2478 | |
Sage Weil | cf3e5c4 | 2009-12-11 09:48:05 -0800 | [diff] [blame] | 2479 | BUG_ON(con->in_msg); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2480 | con->in_tag = CEPH_MSGR_TAG_READY; |
| 2481 | dout("try_write initiating connect on %p new state %lu\n", |
| 2482 | con, con->state); |
Alex Elder | 41617d0 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 2483 | ret = ceph_tcp_connect(con); |
| 2484 | if (ret < 0) { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2485 | con->error_msg = "connect error"; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2486 | goto out; |
| 2487 | } |
| 2488 | } |
| 2489 | |
| 2490 | more_kvec: |
| 2491 | /* kvec data queued? */ |
| 2492 | if (con->out_skip) { |
| 2493 | ret = write_partial_skip(con); |
| 2494 | if (ret <= 0) |
Sage Weil | 42961d2 | 2011-01-25 08:19:34 -0800 | [diff] [blame] | 2495 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2496 | } |
| 2497 | if (con->out_kvec_left) { |
| 2498 | ret = write_partial_kvec(con); |
| 2499 | if (ret <= 0) |
Sage Weil | 42961d2 | 2011-01-25 08:19:34 -0800 | [diff] [blame] | 2500 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2501 | } |
| 2502 | |
| 2503 | /* msg pages? */ |
| 2504 | if (con->out_msg) { |
Sage Weil | c86a293 | 2009-12-14 14:04:30 -0800 | [diff] [blame] | 2505 | if (con->out_msg_done) { |
| 2506 | ceph_msg_put(con->out_msg); |
| 2507 | con->out_msg = NULL; /* we're done with this one */ |
| 2508 | goto do_next; |
| 2509 | } |
| 2510 | |
Alex Elder | 34d2d20 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2511 | ret = write_partial_message_data(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2512 | if (ret == 1) |
| 2513 | goto more_kvec; /* we need to send the footer, too! */ |
| 2514 | if (ret == 0) |
Sage Weil | 42961d2 | 2011-01-25 08:19:34 -0800 | [diff] [blame] | 2515 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2516 | if (ret < 0) { |
Alex Elder | 34d2d20 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2517 | dout("try_write write_partial_message_data err %d\n", |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2518 | ret); |
Sage Weil | 42961d2 | 2011-01-25 08:19:34 -0800 | [diff] [blame] | 2519 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2520 | } |
| 2521 | } |
| 2522 | |
Sage Weil | c86a293 | 2009-12-14 14:04:30 -0800 | [diff] [blame] | 2523 | do_next: |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2524 | if (con->state == CON_STATE_OPEN) { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2525 | /* is anything else pending? */ |
| 2526 | if (!list_empty(&con->out_queue)) { |
| 2527 | prepare_write_message(con); |
| 2528 | goto more; |
| 2529 | } |
| 2530 | if (con->in_seq > con->in_seq_acked) { |
| 2531 | prepare_write_ack(con); |
| 2532 | goto more; |
| 2533 | } |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2534 | if (con_flag_test_and_clear(con, CON_FLAG_KEEPALIVE_PENDING)) { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2535 | prepare_write_keepalive(con); |
| 2536 | goto more; |
| 2537 | } |
| 2538 | } |
| 2539 | |
| 2540 | /* Nothing to do! */ |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2541 | con_flag_clear(con, CON_FLAG_WRITE_PENDING); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2542 | dout("try_write nothing else to write.\n"); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2543 | ret = 0; |
| 2544 | out: |
Sage Weil | 42961d2 | 2011-01-25 08:19:34 -0800 | [diff] [blame] | 2545 | dout("try_write done on %p ret %d\n", con, ret); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2546 | return ret; |
| 2547 | } |
| 2548 | |
| 2549 | |
| 2550 | |
| 2551 | /* |
| 2552 | * Read what we can from the socket. |
| 2553 | */ |
| 2554 | static int try_read(struct ceph_connection *con) |
| 2555 | { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2556 | int ret = -1; |
| 2557 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2558 | more: |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2559 | dout("try_read start on %p state %lu\n", con, con->state); |
| 2560 | if (con->state != CON_STATE_CONNECTING && |
| 2561 | con->state != CON_STATE_NEGOTIATING && |
| 2562 | con->state != CON_STATE_OPEN) |
| 2563 | return 0; |
| 2564 | |
| 2565 | BUG_ON(!con->sock); |
| 2566 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2567 | dout("try_read tag %d in_base_pos %d\n", (int)con->in_tag, |
| 2568 | con->in_base_pos); |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 2569 | |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2570 | if (con->state == CON_STATE_CONNECTING) { |
Alex Elder | 7593af9 | 2012-05-24 11:55:03 -0500 | [diff] [blame] | 2571 | dout("try_read connecting\n"); |
| 2572 | ret = read_partial_banner(con); |
| 2573 | if (ret <= 0) |
Alex Elder | ab166d5 | 2012-05-31 11:37:29 -0500 | [diff] [blame] | 2574 | goto out; |
Alex Elder | 7593af9 | 2012-05-24 11:55:03 -0500 | [diff] [blame] | 2575 | ret = process_banner(con); |
| 2576 | if (ret < 0) |
| 2577 | goto out; |
| 2578 | |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2579 | con->state = CON_STATE_NEGOTIATING; |
Alex Elder | 7593af9 | 2012-05-24 11:55:03 -0500 | [diff] [blame] | 2580 | |
Jim Schutt | 6d4221b | 2012-08-10 10:37:38 -0700 | [diff] [blame] | 2581 | /* |
| 2582 | * Received banner is good, exchange connection info. |
| 2583 | * Do not reset out_kvec, as sending our banner raced |
| 2584 | * with receiving peer banner after connect completed. |
| 2585 | */ |
Alex Elder | 7593af9 | 2012-05-24 11:55:03 -0500 | [diff] [blame] | 2586 | ret = prepare_write_connect(con); |
| 2587 | if (ret < 0) |
| 2588 | goto out; |
| 2589 | prepare_read_connect(con); |
| 2590 | |
| 2591 | /* Send connection info before awaiting response */ |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 2592 | goto out; |
| 2593 | } |
| 2594 | |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2595 | if (con->state == CON_STATE_NEGOTIATING) { |
Alex Elder | 7593af9 | 2012-05-24 11:55:03 -0500 | [diff] [blame] | 2596 | dout("try_read negotiating\n"); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2597 | ret = read_partial_connect(con); |
| 2598 | if (ret <= 0) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2599 | goto out; |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2600 | ret = process_connect(con); |
| 2601 | if (ret < 0) |
| 2602 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2603 | goto more; |
| 2604 | } |
| 2605 | |
Alex Elder | 122070a | 2012-12-26 10:43:57 -0600 | [diff] [blame] | 2606 | WARN_ON(con->state != CON_STATE_OPEN); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2607 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2608 | if (con->in_base_pos < 0) { |
| 2609 | /* |
| 2610 | * skipping + discarding content. |
| 2611 | * |
| 2612 | * FIXME: there must be a better way to do this! |
| 2613 | */ |
Alex Elder | 84495f4 | 2012-02-15 07:43:55 -0600 | [diff] [blame] | 2614 | static char buf[SKIP_BUF_SIZE]; |
| 2615 | int skip = min((int) sizeof (buf), -con->in_base_pos); |
| 2616 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2617 | dout("skipping %d / %d bytes\n", skip, -con->in_base_pos); |
| 2618 | ret = ceph_tcp_recvmsg(con->sock, buf, skip); |
| 2619 | if (ret <= 0) |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2620 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2621 | con->in_base_pos += ret; |
| 2622 | if (con->in_base_pos) |
| 2623 | goto more; |
| 2624 | } |
| 2625 | if (con->in_tag == CEPH_MSGR_TAG_READY) { |
| 2626 | /* |
| 2627 | * what's next? |
| 2628 | */ |
| 2629 | ret = ceph_tcp_recvmsg(con->sock, &con->in_tag, 1); |
| 2630 | if (ret <= 0) |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2631 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2632 | dout("try_read got tag %d\n", (int)con->in_tag); |
| 2633 | switch (con->in_tag) { |
| 2634 | case CEPH_MSGR_TAG_MSG: |
| 2635 | prepare_read_message(con); |
| 2636 | break; |
| 2637 | case CEPH_MSGR_TAG_ACK: |
| 2638 | prepare_read_ack(con); |
| 2639 | break; |
| 2640 | case CEPH_MSGR_TAG_CLOSE: |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2641 | con_close_socket(con); |
| 2642 | con->state = CON_STATE_CLOSED; |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2643 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2644 | default: |
| 2645 | goto bad_tag; |
| 2646 | } |
| 2647 | } |
| 2648 | if (con->in_tag == CEPH_MSGR_TAG_MSG) { |
| 2649 | ret = read_partial_message(con); |
| 2650 | if (ret <= 0) { |
| 2651 | switch (ret) { |
| 2652 | case -EBADMSG: |
| 2653 | con->error_msg = "bad crc"; |
| 2654 | ret = -EIO; |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2655 | break; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2656 | case -EIO: |
| 2657 | con->error_msg = "io error"; |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2658 | break; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2659 | } |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2660 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2661 | } |
| 2662 | if (con->in_tag == CEPH_MSGR_TAG_READY) |
| 2663 | goto more; |
| 2664 | process_message(con); |
Sage Weil | 7b862e0 | 2012-07-30 18:16:56 -0700 | [diff] [blame] | 2665 | if (con->state == CON_STATE_OPEN) |
| 2666 | prepare_read_tag(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2667 | goto more; |
| 2668 | } |
Sage Weil | 3a23083 | 2013-03-25 08:47:40 -0700 | [diff] [blame] | 2669 | if (con->in_tag == CEPH_MSGR_TAG_ACK || |
| 2670 | con->in_tag == CEPH_MSGR_TAG_SEQ) { |
| 2671 | /* |
| 2672 | * the final handshake seq exchange is semantically |
| 2673 | * equivalent to an ACK |
| 2674 | */ |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2675 | ret = read_partial_ack(con); |
| 2676 | if (ret <= 0) |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2677 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2678 | process_ack(con); |
| 2679 | goto more; |
| 2680 | } |
| 2681 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2682 | out: |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2683 | dout("try_read done on %p ret %d\n", con, ret); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2684 | return ret; |
| 2685 | |
| 2686 | bad_tag: |
| 2687 | pr_err("try_read bad con->in_tag = %d\n", (int)con->in_tag); |
| 2688 | con->error_msg = "protocol error, garbage tag"; |
| 2689 | ret = -1; |
| 2690 | goto out; |
| 2691 | } |
| 2692 | |
| 2693 | |
| 2694 | /* |
Alex Elder | 802c6d9 | 2012-10-08 20:37:30 -0700 | [diff] [blame] | 2695 | * Atomically queue work on a connection after the specified delay. |
| 2696 | * Bump @con reference to avoid races with connection teardown. |
| 2697 | * Returns 0 if work was queued, or an error code otherwise. |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2698 | */ |
Alex Elder | 802c6d9 | 2012-10-08 20:37:30 -0700 | [diff] [blame] | 2699 | static int queue_con_delay(struct ceph_connection *con, unsigned long delay) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2700 | { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2701 | if (!con->ops->get(con)) { |
Alex Elder | 802c6d9 | 2012-10-08 20:37:30 -0700 | [diff] [blame] | 2702 | dout("%s %p ref count 0\n", __func__, con); |
| 2703 | |
| 2704 | return -ENOENT; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2705 | } |
| 2706 | |
Alex Elder | 802c6d9 | 2012-10-08 20:37:30 -0700 | [diff] [blame] | 2707 | if (!queue_delayed_work(ceph_msgr_wq, &con->work, delay)) { |
| 2708 | dout("%s %p - already queued\n", __func__, con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2709 | con->ops->put(con); |
Alex Elder | 802c6d9 | 2012-10-08 20:37:30 -0700 | [diff] [blame] | 2710 | |
| 2711 | return -EBUSY; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2712 | } |
Alex Elder | 802c6d9 | 2012-10-08 20:37:30 -0700 | [diff] [blame] | 2713 | |
| 2714 | dout("%s %p %lu\n", __func__, con, delay); |
| 2715 | |
| 2716 | return 0; |
| 2717 | } |
| 2718 | |
| 2719 | static void queue_con(struct ceph_connection *con) |
| 2720 | { |
| 2721 | (void) queue_con_delay(con, 0); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2722 | } |
| 2723 | |
Alex Elder | 7bb21d68 | 2012-12-07 19:50:07 -0600 | [diff] [blame] | 2724 | static bool con_sock_closed(struct ceph_connection *con) |
| 2725 | { |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2726 | if (!con_flag_test_and_clear(con, CON_FLAG_SOCK_CLOSED)) |
Alex Elder | 7bb21d68 | 2012-12-07 19:50:07 -0600 | [diff] [blame] | 2727 | return false; |
| 2728 | |
| 2729 | #define CASE(x) \ |
| 2730 | case CON_STATE_ ## x: \ |
| 2731 | con->error_msg = "socket closed (con state " #x ")"; \ |
| 2732 | break; |
| 2733 | |
| 2734 | switch (con->state) { |
| 2735 | CASE(CLOSED); |
| 2736 | CASE(PREOPEN); |
| 2737 | CASE(CONNECTING); |
| 2738 | CASE(NEGOTIATING); |
| 2739 | CASE(OPEN); |
| 2740 | CASE(STANDBY); |
| 2741 | default: |
| 2742 | pr_warning("%s con %p unrecognized state %lu\n", |
| 2743 | __func__, con, con->state); |
| 2744 | con->error_msg = "unrecognized con state"; |
| 2745 | BUG(); |
| 2746 | break; |
| 2747 | } |
| 2748 | #undef CASE |
| 2749 | |
| 2750 | return true; |
| 2751 | } |
| 2752 | |
Alex Elder | f20a39f | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2753 | static bool con_backoff(struct ceph_connection *con) |
| 2754 | { |
| 2755 | int ret; |
| 2756 | |
| 2757 | if (!con_flag_test_and_clear(con, CON_FLAG_BACKOFF)) |
| 2758 | return false; |
| 2759 | |
| 2760 | ret = queue_con_delay(con, round_jiffies_relative(con->delay)); |
| 2761 | if (ret) { |
| 2762 | dout("%s: con %p FAILED to back off %lu\n", __func__, |
| 2763 | con, con->delay); |
| 2764 | BUG_ON(ret == -ENOENT); |
| 2765 | con_flag_set(con, CON_FLAG_BACKOFF); |
| 2766 | } |
| 2767 | |
| 2768 | return true; |
| 2769 | } |
| 2770 | |
Alex Elder | 9320926 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2771 | /* Finish fault handling; con->mutex must *not* be held here */ |
| 2772 | |
| 2773 | static void con_fault_finish(struct ceph_connection *con) |
| 2774 | { |
| 2775 | /* |
| 2776 | * in case we faulted due to authentication, invalidate our |
| 2777 | * current tickets so that we can get new ones. |
| 2778 | */ |
| 2779 | if (con->auth_retry && con->ops->invalidate_authorizer) { |
| 2780 | dout("calling invalidate_authorizer()\n"); |
| 2781 | con->ops->invalidate_authorizer(con); |
| 2782 | } |
| 2783 | |
| 2784 | if (con->ops->fault) |
| 2785 | con->ops->fault(con); |
| 2786 | } |
| 2787 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2788 | /* |
| 2789 | * Do some work on a connection. Drop a connection ref when we're done. |
| 2790 | */ |
| 2791 | static void con_work(struct work_struct *work) |
| 2792 | { |
| 2793 | struct ceph_connection *con = container_of(work, struct ceph_connection, |
| 2794 | work.work); |
Alex Elder | 4965941 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2795 | bool fault; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2796 | |
Sage Weil | 9dd4658 | 2010-04-28 13:51:50 -0700 | [diff] [blame] | 2797 | mutex_lock(&con->mutex); |
Alex Elder | 4965941 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2798 | while (true) { |
| 2799 | int ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2800 | |
Alex Elder | 4965941 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2801 | if ((fault = con_sock_closed(con))) { |
| 2802 | dout("%s: con %p SOCK_CLOSED\n", __func__, con); |
| 2803 | break; |
| 2804 | } |
| 2805 | if (con_backoff(con)) { |
| 2806 | dout("%s: con %p BACKOFF\n", __func__, con); |
| 2807 | break; |
| 2808 | } |
| 2809 | if (con->state == CON_STATE_STANDBY) { |
| 2810 | dout("%s: con %p STANDBY\n", __func__, con); |
| 2811 | break; |
| 2812 | } |
| 2813 | if (con->state == CON_STATE_CLOSED) { |
| 2814 | dout("%s: con %p CLOSED\n", __func__, con); |
| 2815 | BUG_ON(con->sock); |
| 2816 | break; |
| 2817 | } |
| 2818 | if (con->state == CON_STATE_PREOPEN) { |
| 2819 | dout("%s: con %p PREOPEN\n", __func__, con); |
| 2820 | BUG_ON(con->sock); |
| 2821 | } |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 2822 | |
Alex Elder | 4965941 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2823 | ret = try_read(con); |
| 2824 | if (ret < 0) { |
| 2825 | if (ret == -EAGAIN) |
| 2826 | continue; |
| 2827 | con->error_msg = "socket error on read"; |
| 2828 | fault = true; |
| 2829 | break; |
| 2830 | } |
| 2831 | |
| 2832 | ret = try_write(con); |
| 2833 | if (ret < 0) { |
| 2834 | if (ret == -EAGAIN) |
| 2835 | continue; |
| 2836 | con->error_msg = "socket error on write"; |
| 2837 | fault = true; |
| 2838 | } |
| 2839 | |
| 2840 | break; /* If we make it to here, we're done */ |
Sage Weil | 3a140a0 | 2012-07-30 16:24:21 -0700 | [diff] [blame] | 2841 | } |
Alex Elder | b6e7b6a | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2842 | if (fault) |
| 2843 | con_fault(con); |
Sage Weil | 9dd4658 | 2010-04-28 13:51:50 -0700 | [diff] [blame] | 2844 | mutex_unlock(&con->mutex); |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 2845 | |
Alex Elder | b6e7b6a | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2846 | if (fault) |
| 2847 | con_fault_finish(con); |
| 2848 | |
| 2849 | con->ops->put(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2850 | } |
| 2851 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2852 | /* |
| 2853 | * Generic error/fault handler. A retry mechanism is used with |
| 2854 | * exponential backoff |
| 2855 | */ |
Alex Elder | 9320926 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2856 | static void con_fault(struct ceph_connection *con) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2857 | { |
Alex Elder | 2836298 | 2012-12-14 16:47:41 -0600 | [diff] [blame] | 2858 | pr_warning("%s%lld %s %s\n", ENTITY_NAME(con->peer_name), |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2859 | ceph_pr_addr(&con->peer_addr.in_addr), con->error_msg); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2860 | dout("fault %p state %lu to peer %s\n", |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2861 | con, con->state, ceph_pr_addr(&con->peer_addr.in_addr)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2862 | |
Alex Elder | 122070a | 2012-12-26 10:43:57 -0600 | [diff] [blame] | 2863 | WARN_ON(con->state != CON_STATE_CONNECTING && |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2864 | con->state != CON_STATE_NEGOTIATING && |
| 2865 | con->state != CON_STATE_OPEN); |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 2866 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2867 | con_close_socket(con); |
Sage Weil | 5e095e8 | 2009-12-14 14:30:34 -0800 | [diff] [blame] | 2868 | |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2869 | if (con_flag_test(con, CON_FLAG_LOSSYTX)) { |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2870 | dout("fault on LOSSYTX channel, marking CLOSED\n"); |
| 2871 | con->state = CON_STATE_CLOSED; |
Alex Elder | 9320926 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2872 | return; |
Sage Weil | 3b5ede0 | 2012-07-20 15:22:53 -0700 | [diff] [blame] | 2873 | } |
| 2874 | |
Sage Weil | 5e095e8 | 2009-12-14 14:30:34 -0800 | [diff] [blame] | 2875 | if (con->in_msg) { |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2876 | BUG_ON(con->in_msg->con != con); |
| 2877 | con->in_msg->con = NULL; |
Sage Weil | 5e095e8 | 2009-12-14 14:30:34 -0800 | [diff] [blame] | 2878 | ceph_msg_put(con->in_msg); |
| 2879 | con->in_msg = NULL; |
Sage Weil | 36eb71a | 2012-06-21 12:47:08 -0700 | [diff] [blame] | 2880 | con->ops->put(con); |
Sage Weil | 5e095e8 | 2009-12-14 14:30:34 -0800 | [diff] [blame] | 2881 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2882 | |
Sage Weil | e80a52d | 2010-02-25 12:40:45 -0800 | [diff] [blame] | 2883 | /* Requeue anything that hasn't been acked */ |
| 2884 | list_splice_init(&con->out_sent, &con->out_queue); |
Sage Weil | 9bd2e6f | 2010-02-02 16:21:06 -0800 | [diff] [blame] | 2885 | |
Sage Weil | e76661d | 2011-03-03 10:10:15 -0800 | [diff] [blame] | 2886 | /* If there are no messages queued or keepalive pending, place |
| 2887 | * the connection in a STANDBY state */ |
| 2888 | if (list_empty(&con->out_queue) && |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2889 | !con_flag_test(con, CON_FLAG_KEEPALIVE_PENDING)) { |
Sage Weil | e00de34 | 2011-03-04 12:25:05 -0800 | [diff] [blame] | 2890 | dout("fault %p setting STANDBY clearing WRITE_PENDING\n", con); |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2891 | con_flag_clear(con, CON_FLAG_WRITE_PENDING); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2892 | con->state = CON_STATE_STANDBY; |
Sage Weil | e80a52d | 2010-02-25 12:40:45 -0800 | [diff] [blame] | 2893 | } else { |
| 2894 | /* retry after a delay. */ |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2895 | con->state = CON_STATE_PREOPEN; |
Sage Weil | e80a52d | 2010-02-25 12:40:45 -0800 | [diff] [blame] | 2896 | if (con->delay == 0) |
| 2897 | con->delay = BASE_DELAY_INTERVAL; |
| 2898 | else if (con->delay < MAX_DELAY_INTERVAL) |
| 2899 | con->delay *= 2; |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2900 | con_flag_set(con, CON_FLAG_BACKOFF); |
Alex Elder | 8618e30 | 2012-10-08 20:37:30 -0700 | [diff] [blame] | 2901 | queue_con(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2902 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2903 | } |
| 2904 | |
| 2905 | |
| 2906 | |
| 2907 | /* |
Alex Elder | 15d9882 | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 2908 | * initialize a new messenger instance |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2909 | */ |
Alex Elder | 15d9882 | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 2910 | void ceph_messenger_init(struct ceph_messenger *msgr, |
| 2911 | struct ceph_entity_addr *myaddr, |
| 2912 | u32 supported_features, |
| 2913 | u32 required_features, |
| 2914 | bool nocrc) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2915 | { |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2916 | msgr->supported_features = supported_features; |
| 2917 | msgr->required_features = required_features; |
| 2918 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2919 | spin_lock_init(&msgr->global_seq_lock); |
| 2920 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2921 | if (myaddr) |
| 2922 | msgr->inst.addr = *myaddr; |
| 2923 | |
| 2924 | /* select a random nonce */ |
Sage Weil | ac8839d | 2010-01-27 14:28:10 -0800 | [diff] [blame] | 2925 | msgr->inst.addr.type = 0; |
Sage Weil | 103e2d3 | 2010-01-07 16:12:36 -0800 | [diff] [blame] | 2926 | get_random_bytes(&msgr->inst.addr.nonce, sizeof(msgr->inst.addr.nonce)); |
Sage Weil | 63f2d21 | 2009-11-03 15:17:56 -0800 | [diff] [blame] | 2927 | encode_my_addr(msgr); |
Alex Elder | 15d9882 | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 2928 | msgr->nocrc = nocrc; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2929 | |
Guanjun He | a2a3258 | 2012-07-08 19:50:33 -0700 | [diff] [blame] | 2930 | atomic_set(&msgr->stopping, 0); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2931 | |
Alex Elder | 15d9882 | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 2932 | dout("%s %p\n", __func__, msgr); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2933 | } |
Alex Elder | 15d9882 | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 2934 | EXPORT_SYMBOL(ceph_messenger_init); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2935 | |
Sage Weil | e00de34 | 2011-03-04 12:25:05 -0800 | [diff] [blame] | 2936 | static void clear_standby(struct ceph_connection *con) |
| 2937 | { |
| 2938 | /* come back from STANDBY? */ |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2939 | if (con->state == CON_STATE_STANDBY) { |
Sage Weil | e00de34 | 2011-03-04 12:25:05 -0800 | [diff] [blame] | 2940 | dout("clear_standby %p and ++connect_seq\n", con); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2941 | con->state = CON_STATE_PREOPEN; |
Sage Weil | e00de34 | 2011-03-04 12:25:05 -0800 | [diff] [blame] | 2942 | con->connect_seq++; |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2943 | WARN_ON(con_flag_test(con, CON_FLAG_WRITE_PENDING)); |
| 2944 | WARN_ON(con_flag_test(con, CON_FLAG_KEEPALIVE_PENDING)); |
Sage Weil | e00de34 | 2011-03-04 12:25:05 -0800 | [diff] [blame] | 2945 | } |
| 2946 | } |
| 2947 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2948 | /* |
| 2949 | * Queue up an outgoing message on the given connection. |
| 2950 | */ |
| 2951 | void ceph_con_send(struct ceph_connection *con, struct ceph_msg *msg) |
| 2952 | { |
Sage Weil | a59b55a | 2012-07-20 15:34:04 -0700 | [diff] [blame] | 2953 | /* set src+dst */ |
| 2954 | msg->hdr.src = con->msgr->inst.name; |
| 2955 | BUG_ON(msg->front.iov_len != le32_to_cpu(msg->hdr.front_len)); |
| 2956 | msg->needs_out_seq = true; |
| 2957 | |
| 2958 | mutex_lock(&con->mutex); |
| 2959 | |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2960 | if (con->state == CON_STATE_CLOSED) { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2961 | dout("con_send %p closed, dropping %p\n", con, msg); |
| 2962 | ceph_msg_put(msg); |
Sage Weil | a59b55a | 2012-07-20 15:34:04 -0700 | [diff] [blame] | 2963 | mutex_unlock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2964 | return; |
| 2965 | } |
| 2966 | |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2967 | BUG_ON(msg->con != NULL); |
Sage Weil | 36eb71a | 2012-06-21 12:47:08 -0700 | [diff] [blame] | 2968 | msg->con = con->ops->get(con); |
Alex Elder | 92ce034 | 2012-06-04 14:43:33 -0500 | [diff] [blame] | 2969 | BUG_ON(msg->con == NULL); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2970 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2971 | BUG_ON(!list_empty(&msg->list_head)); |
| 2972 | list_add_tail(&msg->list_head, &con->out_queue); |
| 2973 | dout("----- %p to %s%lld %d=%s len %d+%d+%d -----\n", msg, |
| 2974 | ENTITY_NAME(con->peer_name), le16_to_cpu(msg->hdr.type), |
| 2975 | ceph_msg_type_name(le16_to_cpu(msg->hdr.type)), |
| 2976 | le32_to_cpu(msg->hdr.front_len), |
| 2977 | le32_to_cpu(msg->hdr.middle_len), |
| 2978 | le32_to_cpu(msg->hdr.data_len)); |
Sage Weil | 0065093 | 2012-07-20 15:33:04 -0700 | [diff] [blame] | 2979 | |
| 2980 | clear_standby(con); |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 2981 | mutex_unlock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2982 | |
| 2983 | /* if there wasn't anything waiting to send before, queue |
| 2984 | * new work */ |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2985 | if (con_flag_test_and_set(con, CON_FLAG_WRITE_PENDING) == 0) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2986 | queue_con(con); |
| 2987 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2988 | EXPORT_SYMBOL(ceph_con_send); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2989 | |
| 2990 | /* |
| 2991 | * Revoke a message that was previously queued for send |
| 2992 | */ |
Alex Elder | 6740a84 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2993 | void ceph_msg_revoke(struct ceph_msg *msg) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2994 | { |
Alex Elder | 6740a84 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2995 | struct ceph_connection *con = msg->con; |
| 2996 | |
| 2997 | if (!con) |
| 2998 | return; /* Message not in our possession */ |
| 2999 | |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 3000 | mutex_lock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3001 | if (!list_empty(&msg->list_head)) { |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 3002 | dout("%s %p msg %p - was on queue\n", __func__, con, msg); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3003 | list_del_init(&msg->list_head); |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 3004 | BUG_ON(msg->con == NULL); |
Sage Weil | 36eb71a | 2012-06-21 12:47:08 -0700 | [diff] [blame] | 3005 | msg->con->ops->put(msg->con); |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 3006 | msg->con = NULL; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3007 | msg->hdr.seq = 0; |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 3008 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3009 | ceph_msg_put(msg); |
Sage Weil | ed98ada | 2010-07-05 12:15:14 -0700 | [diff] [blame] | 3010 | } |
| 3011 | if (con->out_msg == msg) { |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 3012 | dout("%s %p msg %p - was sending\n", __func__, con, msg); |
Sage Weil | ed98ada | 2010-07-05 12:15:14 -0700 | [diff] [blame] | 3013 | con->out_msg = NULL; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3014 | if (con->out_kvec_is_msg) { |
| 3015 | con->out_skip = con->out_kvec_bytes; |
| 3016 | con->out_kvec_is_msg = false; |
| 3017 | } |
Sage Weil | ed98ada | 2010-07-05 12:15:14 -0700 | [diff] [blame] | 3018 | msg->hdr.seq = 0; |
Alex Elder | 92ce034 | 2012-06-04 14:43:33 -0500 | [diff] [blame] | 3019 | |
| 3020 | ceph_msg_put(msg); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3021 | } |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 3022 | mutex_unlock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3023 | } |
| 3024 | |
| 3025 | /* |
Yehuda Sadeh | 0d59ab8 | 2010-01-13 17:03:23 -0800 | [diff] [blame] | 3026 | * Revoke a message that we may be reading data into |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 3027 | */ |
Alex Elder | 8921d11 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 3028 | void ceph_msg_revoke_incoming(struct ceph_msg *msg) |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 3029 | { |
Alex Elder | 8921d11 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 3030 | struct ceph_connection *con; |
| 3031 | |
| 3032 | BUG_ON(msg == NULL); |
| 3033 | if (!msg->con) { |
| 3034 | dout("%s msg %p null con\n", __func__, msg); |
| 3035 | |
| 3036 | return; /* Message not in our possession */ |
| 3037 | } |
| 3038 | |
| 3039 | con = msg->con; |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 3040 | mutex_lock(&con->mutex); |
Alex Elder | 8921d11 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 3041 | if (con->in_msg == msg) { |
Eric Dumazet | 95c9617 | 2012-04-15 05:58:06 +0000 | [diff] [blame] | 3042 | unsigned int front_len = le32_to_cpu(con->in_hdr.front_len); |
| 3043 | unsigned int middle_len = le32_to_cpu(con->in_hdr.middle_len); |
| 3044 | unsigned int data_len = le32_to_cpu(con->in_hdr.data_len); |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 3045 | |
| 3046 | /* skip rest of message */ |
Alex Elder | 8921d11 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 3047 | dout("%s %p msg %p revoked\n", __func__, con, msg); |
| 3048 | con->in_base_pos = con->in_base_pos - |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 3049 | sizeof(struct ceph_msg_header) - |
Yehuda Sadeh | 0d59ab8 | 2010-01-13 17:03:23 -0800 | [diff] [blame] | 3050 | front_len - |
| 3051 | middle_len - |
| 3052 | data_len - |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 3053 | sizeof(struct ceph_msg_footer); |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 3054 | ceph_msg_put(con->in_msg); |
| 3055 | con->in_msg = NULL; |
| 3056 | con->in_tag = CEPH_MSGR_TAG_READY; |
Sage Weil | 684be25 | 2010-04-21 20:45:59 -0700 | [diff] [blame] | 3057 | con->in_seq++; |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 3058 | } else { |
Alex Elder | 8921d11 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 3059 | dout("%s %p in_msg %p msg %p no-op\n", |
| 3060 | __func__, con, con->in_msg, msg); |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 3061 | } |
| 3062 | mutex_unlock(&con->mutex); |
| 3063 | } |
| 3064 | |
| 3065 | /* |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3066 | * Queue a keepalive byte to ensure the tcp connection is alive. |
| 3067 | */ |
| 3068 | void ceph_con_keepalive(struct ceph_connection *con) |
| 3069 | { |
Sage Weil | e00de34 | 2011-03-04 12:25:05 -0800 | [diff] [blame] | 3070 | dout("con_keepalive %p\n", con); |
Sage Weil | 0065093 | 2012-07-20 15:33:04 -0700 | [diff] [blame] | 3071 | mutex_lock(&con->mutex); |
Sage Weil | e00de34 | 2011-03-04 12:25:05 -0800 | [diff] [blame] | 3072 | clear_standby(con); |
Sage Weil | 0065093 | 2012-07-20 15:33:04 -0700 | [diff] [blame] | 3073 | mutex_unlock(&con->mutex); |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 3074 | if (con_flag_test_and_set(con, CON_FLAG_KEEPALIVE_PENDING) == 0 && |
| 3075 | con_flag_test_and_set(con, CON_FLAG_WRITE_PENDING) == 0) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3076 | queue_con(con); |
| 3077 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 3078 | EXPORT_SYMBOL(ceph_con_keepalive); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3079 | |
Alex Elder | 4379450 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3080 | static void ceph_msg_data_init(struct ceph_msg_data *data) |
| 3081 | { |
| 3082 | data->type = CEPH_MSG_DATA_NONE; |
| 3083 | } |
| 3084 | |
Alex Elder | 02afca6 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 3085 | 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] | 3086 | size_t length, size_t alignment) |
Alex Elder | 02afca6 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 3087 | { |
Alex Elder | 07aa155 | 2013-03-04 18:29:06 -0600 | [diff] [blame] | 3088 | BUG_ON(!pages); |
| 3089 | BUG_ON(!length); |
Alex Elder | 4379450 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3090 | BUG_ON(msg->p.type != CEPH_MSG_DATA_NONE); |
Alex Elder | 02afca6 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 3091 | |
Alex Elder | 4379450 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3092 | msg->p.type = CEPH_MSG_DATA_PAGES; |
Alex Elder | f9e1577 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3093 | msg->p.pages = pages; |
| 3094 | msg->p.length = length; |
| 3095 | msg->p.alignment = alignment & ~PAGE_MASK; |
Alex Elder | 02afca6 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 3096 | } |
| 3097 | EXPORT_SYMBOL(ceph_msg_data_set_pages); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3098 | |
Alex Elder | 27fa838 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 3099 | void ceph_msg_data_set_pagelist(struct ceph_msg *msg, |
| 3100 | struct ceph_pagelist *pagelist) |
| 3101 | { |
Alex Elder | 07aa155 | 2013-03-04 18:29:06 -0600 | [diff] [blame] | 3102 | BUG_ON(!pagelist); |
| 3103 | BUG_ON(!pagelist->length); |
Alex Elder | 4379450 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3104 | BUG_ON(msg->l.type != CEPH_MSG_DATA_NONE); |
Alex Elder | 27fa838 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 3105 | |
Alex Elder | 4379450 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3106 | msg->l.type = CEPH_MSG_DATA_PAGELIST; |
Alex Elder | f9e1577 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3107 | msg->l.pagelist = pagelist; |
Alex Elder | 27fa838 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 3108 | } |
| 3109 | EXPORT_SYMBOL(ceph_msg_data_set_pagelist); |
| 3110 | |
| 3111 | void ceph_msg_data_set_bio(struct ceph_msg *msg, struct bio *bio) |
| 3112 | { |
Alex Elder | 07aa155 | 2013-03-04 18:29:06 -0600 | [diff] [blame] | 3113 | BUG_ON(!bio); |
Alex Elder | 4379450 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3114 | BUG_ON(msg->b.type != CEPH_MSG_DATA_NONE); |
Alex Elder | 27fa838 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 3115 | |
Alex Elder | 4379450 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3116 | msg->b.type = CEPH_MSG_DATA_BIO; |
Alex Elder | f9e1577 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3117 | msg->b.bio = bio; |
Alex Elder | 27fa838 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 3118 | } |
| 3119 | EXPORT_SYMBOL(ceph_msg_data_set_bio); |
| 3120 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3121 | /* |
| 3122 | * construct a new message with given type, size |
| 3123 | * the new msg has a ref count of 1. |
| 3124 | */ |
Sage Weil | b61c276 | 2011-08-09 15:03:46 -0700 | [diff] [blame] | 3125 | struct ceph_msg *ceph_msg_new(int type, int front_len, gfp_t flags, |
| 3126 | bool can_fail) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3127 | { |
| 3128 | struct ceph_msg *m; |
| 3129 | |
Alex Elder | 9516e45 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3130 | m = kzalloc(sizeof(*m), flags); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3131 | if (m == NULL) |
| 3132 | goto out; |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 3133 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3134 | m->hdr.type = cpu_to_le16(type); |
Sage Weil | 45c6ceb | 2010-05-11 15:01:51 -0700 | [diff] [blame] | 3135 | m->hdr.priority = cpu_to_le16(CEPH_MSG_PRIO_DEFAULT); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3136 | m->hdr.front_len = cpu_to_le32(front_len); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3137 | |
Alex Elder | 9516e45 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3138 | INIT_LIST_HEAD(&m->list_head); |
| 3139 | kref_init(&m->kref); |
Henry C Chang | ca20892 | 2011-05-03 02:29:56 +0000 | [diff] [blame] | 3140 | |
Alex Elder | 4379450 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3141 | ceph_msg_data_init(&m->p); |
| 3142 | ceph_msg_data_init(&m->l); |
| 3143 | ceph_msg_data_init(&m->b); |
Alex Elder | 4379450 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3144 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3145 | /* front */ |
Alex Elder | 9516e45 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3146 | m->front_max = front_len; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3147 | if (front_len) { |
| 3148 | if (front_len > PAGE_CACHE_SIZE) { |
Yehuda Sadeh | 34d2376 | 2010-04-06 14:33:58 -0700 | [diff] [blame] | 3149 | m->front.iov_base = __vmalloc(front_len, flags, |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3150 | PAGE_KERNEL); |
| 3151 | m->front_is_vmalloc = true; |
| 3152 | } else { |
Yehuda Sadeh | 34d2376 | 2010-04-06 14:33:58 -0700 | [diff] [blame] | 3153 | m->front.iov_base = kmalloc(front_len, flags); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3154 | } |
| 3155 | if (m->front.iov_base == NULL) { |
Sage Weil | b61c276 | 2011-08-09 15:03:46 -0700 | [diff] [blame] | 3156 | dout("ceph_msg_new can't allocate %d bytes\n", |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3157 | front_len); |
| 3158 | goto out2; |
| 3159 | } |
| 3160 | } else { |
| 3161 | m->front.iov_base = NULL; |
| 3162 | } |
| 3163 | m->front.iov_len = front_len; |
| 3164 | |
Sage Weil | bb25766 | 2010-04-01 16:07:23 -0700 | [diff] [blame] | 3165 | dout("ceph_msg_new %p front %d\n", m, front_len); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3166 | return m; |
| 3167 | |
| 3168 | out2: |
| 3169 | ceph_msg_put(m); |
| 3170 | out: |
Sage Weil | b61c276 | 2011-08-09 15:03:46 -0700 | [diff] [blame] | 3171 | if (!can_fail) { |
| 3172 | pr_err("msg_new can't create type %d front %d\n", type, |
| 3173 | front_len); |
Sage Weil | f0ed1b7 | 2011-08-09 15:05:07 -0700 | [diff] [blame] | 3174 | WARN_ON(1); |
Sage Weil | b61c276 | 2011-08-09 15:03:46 -0700 | [diff] [blame] | 3175 | } else { |
| 3176 | dout("msg_new can't create type %d front %d\n", type, |
| 3177 | front_len); |
| 3178 | } |
Sage Weil | a79832f | 2010-04-01 16:06:19 -0700 | [diff] [blame] | 3179 | return NULL; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3180 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 3181 | EXPORT_SYMBOL(ceph_msg_new); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3182 | |
| 3183 | /* |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3184 | * Allocate "middle" portion of a message, if it is needed and wasn't |
| 3185 | * allocated by alloc_msg. This allows us to read a small fixed-size |
| 3186 | * per-type header in the front and then gracefully fail (i.e., |
| 3187 | * propagate the error to the caller based on info in the front) when |
| 3188 | * the middle is too large. |
| 3189 | */ |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3190 | 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] | 3191 | { |
| 3192 | int type = le16_to_cpu(msg->hdr.type); |
| 3193 | int middle_len = le32_to_cpu(msg->hdr.middle_len); |
| 3194 | |
| 3195 | dout("alloc_middle %p type %d %s middle_len %d\n", msg, type, |
| 3196 | ceph_msg_type_name(type), middle_len); |
| 3197 | BUG_ON(!middle_len); |
| 3198 | BUG_ON(msg->middle); |
| 3199 | |
Sage Weil | b6c1d5b | 2009-12-07 12:17:17 -0800 | [diff] [blame] | 3200 | msg->middle = ceph_buffer_new(middle_len, GFP_NOFS); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3201 | if (!msg->middle) |
| 3202 | return -ENOMEM; |
| 3203 | return 0; |
| 3204 | } |
| 3205 | |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3206 | /* |
Alex Elder | 1c20f2d | 2012-06-04 14:43:32 -0500 | [diff] [blame] | 3207 | * Allocate a message for receiving an incoming message on a |
| 3208 | * connection, and save the result in con->in_msg. Uses the |
| 3209 | * connection's private alloc_msg op if available. |
| 3210 | * |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 3211 | * Returns 0 on success, or a negative error code. |
| 3212 | * |
| 3213 | * On success, if we set *skip = 1: |
| 3214 | * - the next message should be skipped and ignored. |
| 3215 | * - con->in_msg == NULL |
| 3216 | * or if we set *skip = 0: |
| 3217 | * - con->in_msg is non-null. |
| 3218 | * On error (ENOMEM, EAGAIN, ...), |
| 3219 | * - con->in_msg == NULL |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3220 | */ |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 3221 | 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] | 3222 | { |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 3223 | struct ceph_msg_header *hdr = &con->in_hdr; |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3224 | int middle_len = le32_to_cpu(hdr->middle_len); |
Alex Elder | 1d866d1 | 2013-03-01 18:00:14 -0600 | [diff] [blame] | 3225 | struct ceph_msg *msg; |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 3226 | int ret = 0; |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3227 | |
Alex Elder | 1c20f2d | 2012-06-04 14:43:32 -0500 | [diff] [blame] | 3228 | BUG_ON(con->in_msg != NULL); |
Alex Elder | 53ded49 | 2013-03-01 18:00:14 -0600 | [diff] [blame] | 3229 | BUG_ON(!con->ops->alloc_msg); |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3230 | |
Alex Elder | 53ded49 | 2013-03-01 18:00:14 -0600 | [diff] [blame] | 3231 | mutex_unlock(&con->mutex); |
| 3232 | msg = con->ops->alloc_msg(con, hdr, skip); |
| 3233 | mutex_lock(&con->mutex); |
| 3234 | if (con->state != CON_STATE_OPEN) { |
| 3235 | if (msg) |
Alex Elder | 1d866d1 | 2013-03-01 18:00:14 -0600 | [diff] [blame] | 3236 | ceph_msg_put(msg); |
Alex Elder | 53ded49 | 2013-03-01 18:00:14 -0600 | [diff] [blame] | 3237 | return -EAGAIN; |
| 3238 | } |
Alex Elder | 4137577 | 2013-03-05 09:25:10 -0600 | [diff] [blame] | 3239 | if (msg) { |
| 3240 | BUG_ON(*skip); |
| 3241 | con->in_msg = msg; |
Sage Weil | 36eb71a | 2012-06-21 12:47:08 -0700 | [diff] [blame] | 3242 | con->in_msg->con = con->ops->get(con); |
Alex Elder | 92ce034 | 2012-06-04 14:43:33 -0500 | [diff] [blame] | 3243 | BUG_ON(con->in_msg->con == NULL); |
Alex Elder | 4137577 | 2013-03-05 09:25:10 -0600 | [diff] [blame] | 3244 | } else { |
| 3245 | /* |
| 3246 | * Null message pointer means either we should skip |
| 3247 | * this message or we couldn't allocate memory. The |
| 3248 | * former is not an error. |
| 3249 | */ |
| 3250 | if (*skip) |
| 3251 | return 0; |
| 3252 | con->error_msg = "error allocating memory for incoming message"; |
| 3253 | |
Alex Elder | 53ded49 | 2013-03-01 18:00:14 -0600 | [diff] [blame] | 3254 | return -ENOMEM; |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3255 | } |
Alex Elder | 1c20f2d | 2012-06-04 14:43:32 -0500 | [diff] [blame] | 3256 | memcpy(&con->in_msg->hdr, &con->in_hdr, sizeof(con->in_hdr)); |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3257 | |
Alex Elder | 1c20f2d | 2012-06-04 14:43:32 -0500 | [diff] [blame] | 3258 | if (middle_len && !con->in_msg->middle) { |
| 3259 | ret = ceph_alloc_middle(con, con->in_msg); |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3260 | if (ret < 0) { |
Alex Elder | 1c20f2d | 2012-06-04 14:43:32 -0500 | [diff] [blame] | 3261 | ceph_msg_put(con->in_msg); |
| 3262 | con->in_msg = NULL; |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3263 | } |
| 3264 | } |
Yehuda Sadeh | 9d7f0f1 | 2010-01-11 10:32:02 -0800 | [diff] [blame] | 3265 | |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 3266 | return ret; |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3267 | } |
| 3268 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3269 | |
| 3270 | /* |
| 3271 | * Free a generically kmalloc'd message. |
| 3272 | */ |
| 3273 | void ceph_msg_kfree(struct ceph_msg *m) |
| 3274 | { |
| 3275 | dout("msg_kfree %p\n", m); |
| 3276 | if (m->front_is_vmalloc) |
| 3277 | vfree(m->front.iov_base); |
| 3278 | else |
| 3279 | kfree(m->front.iov_base); |
| 3280 | kfree(m); |
| 3281 | } |
| 3282 | |
| 3283 | /* |
| 3284 | * Drop a msg ref. Destroy as needed. |
| 3285 | */ |
Sage Weil | c2e552e | 2009-12-07 15:55:05 -0800 | [diff] [blame] | 3286 | void ceph_msg_last_put(struct kref *kref) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3287 | { |
Sage Weil | c2e552e | 2009-12-07 15:55:05 -0800 | [diff] [blame] | 3288 | struct ceph_msg *m = container_of(kref, struct ceph_msg, kref); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3289 | |
Sage Weil | c2e552e | 2009-12-07 15:55:05 -0800 | [diff] [blame] | 3290 | dout("ceph_msg_put last one on %p\n", m); |
| 3291 | WARN_ON(!list_empty(&m->list_head)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3292 | |
Sage Weil | c2e552e | 2009-12-07 15:55:05 -0800 | [diff] [blame] | 3293 | /* drop middle, data, if any */ |
| 3294 | if (m->middle) { |
| 3295 | ceph_buffer_put(m->middle); |
| 3296 | m->middle = NULL; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3297 | } |
Alex Elder | 97fb1c7 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3298 | if (ceph_msg_has_pages(m)) { |
Alex Elder | f9e1577 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3299 | m->p.length = 0; |
| 3300 | m->p.pages = NULL; |
Alex Elder | 888334f9 | 2013-03-25 11:54:30 -0500 | [diff] [blame] | 3301 | m->p.type = CEPH_OSD_DATA_TYPE_NONE; |
Alex Elder | 97fb1c7 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3302 | } |
Alex Elder | 97fb1c7 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3303 | if (ceph_msg_has_pagelist(m)) { |
Alex Elder | f9e1577 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3304 | ceph_pagelist_release(m->l.pagelist); |
| 3305 | kfree(m->l.pagelist); |
| 3306 | m->l.pagelist = NULL; |
Alex Elder | 888334f9 | 2013-03-25 11:54:30 -0500 | [diff] [blame] | 3307 | m->l.type = CEPH_OSD_DATA_TYPE_NONE; |
| 3308 | } |
| 3309 | if (ceph_msg_has_bio(m)) { |
| 3310 | m->b.bio = NULL; |
| 3311 | m->b.type = CEPH_OSD_DATA_TYPE_NONE; |
Sage Weil | 58bb3b3 | 2009-12-23 12:12:31 -0800 | [diff] [blame] | 3312 | } |
| 3313 | |
Sage Weil | c2e552e | 2009-12-07 15:55:05 -0800 | [diff] [blame] | 3314 | if (m->pool) |
| 3315 | ceph_msgpool_put(m->pool, m); |
| 3316 | else |
| 3317 | ceph_msg_kfree(m); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3318 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 3319 | EXPORT_SYMBOL(ceph_msg_last_put); |
Sage Weil | 9ec7cab | 2009-12-14 15:13:47 -0800 | [diff] [blame] | 3320 | |
| 3321 | void ceph_msg_dump(struct ceph_msg *msg) |
| 3322 | { |
Alex Elder | 4a73ef2 | 2013-03-07 15:38:26 -0600 | [diff] [blame] | 3323 | pr_debug("msg_dump %p (front_max %d length %zd)\n", msg, |
Alex Elder | f9e1577 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3324 | msg->front_max, msg->p.length); |
Sage Weil | 9ec7cab | 2009-12-14 15:13:47 -0800 | [diff] [blame] | 3325 | print_hex_dump(KERN_DEBUG, "header: ", |
| 3326 | DUMP_PREFIX_OFFSET, 16, 1, |
| 3327 | &msg->hdr, sizeof(msg->hdr), true); |
| 3328 | print_hex_dump(KERN_DEBUG, " front: ", |
| 3329 | DUMP_PREFIX_OFFSET, 16, 1, |
| 3330 | msg->front.iov_base, msg->front.iov_len, true); |
| 3331 | if (msg->middle) |
| 3332 | print_hex_dump(KERN_DEBUG, "middle: ", |
| 3333 | DUMP_PREFIX_OFFSET, 16, 1, |
| 3334 | msg->middle->vec.iov_base, |
| 3335 | msg->middle->vec.iov_len, true); |
| 3336 | print_hex_dump(KERN_DEBUG, "footer: ", |
| 3337 | DUMP_PREFIX_OFFSET, 16, 1, |
| 3338 | &msg->footer, sizeof(msg->footer), true); |
| 3339 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 3340 | EXPORT_SYMBOL(ceph_msg_dump); |