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 | |
Ilya Dryomov | 2b3e0c9 | 2013-12-24 21:19:24 +0200 | [diff] [blame] | 18 | #include <linux/ceph/ceph_features.h> |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 19 | #include <linux/ceph/libceph.h> |
| 20 | #include <linux/ceph/messenger.h> |
| 21 | #include <linux/ceph/decode.h> |
| 22 | #include <linux/ceph/pagelist.h> |
Paul Gortmaker | bc3b2d7 | 2011-07-15 11:47:34 -0400 | [diff] [blame] | 23 | #include <linux/export.h> |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 24 | |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 25 | #define list_entry_next(pos, member) \ |
| 26 | list_entry(pos->member.next, typeof(*pos), member) |
| 27 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 28 | /* |
| 29 | * Ceph uses the messenger to exchange ceph_msg messages with other |
| 30 | * hosts in the system. The messenger provides ordered and reliable |
| 31 | * delivery. We tolerate TCP disconnects by reconnecting (with |
| 32 | * exponential backoff) in the case of a fault (disconnection, bad |
| 33 | * crc, protocol error). Acks allow sent messages to be discarded by |
| 34 | * the sender. |
| 35 | */ |
| 36 | |
Alex Elder | bc18f4b | 2012-06-20 21:53:53 -0500 | [diff] [blame] | 37 | /* |
| 38 | * We track the state of the socket on a given connection using |
| 39 | * values defined below. The transition to a new socket state is |
| 40 | * handled by a function which verifies we aren't coming from an |
| 41 | * unexpected state. |
| 42 | * |
| 43 | * -------- |
| 44 | * | NEW* | transient initial state |
| 45 | * -------- |
| 46 | * | con_sock_state_init() |
| 47 | * v |
| 48 | * ---------- |
| 49 | * | CLOSED | initialized, but no socket (and no |
| 50 | * ---------- TCP connection) |
| 51 | * ^ \ |
| 52 | * | \ con_sock_state_connecting() |
| 53 | * | ---------------------- |
| 54 | * | \ |
| 55 | * + con_sock_state_closed() \ |
Sage Weil | fbb85a4 | 2012-06-27 12:31:02 -0700 | [diff] [blame] | 56 | * |+--------------------------- \ |
| 57 | * | \ \ \ |
| 58 | * | ----------- \ \ |
| 59 | * | | CLOSING | socket event; \ \ |
| 60 | * | ----------- await close \ \ |
| 61 | * | ^ \ | |
| 62 | * | | \ | |
| 63 | * | + con_sock_state_closing() \ | |
| 64 | * | / \ | | |
| 65 | * | / --------------- | | |
| 66 | * | / \ v v |
Alex Elder | bc18f4b | 2012-06-20 21:53:53 -0500 | [diff] [blame] | 67 | * | / -------------- |
| 68 | * | / -----------------| CONNECTING | socket created, TCP |
| 69 | * | | / -------------- connect initiated |
| 70 | * | | | con_sock_state_connected() |
| 71 | * | | v |
| 72 | * ------------- |
| 73 | * | CONNECTED | TCP connection established |
| 74 | * ------------- |
| 75 | * |
| 76 | * State values for ceph_connection->sock_state; NEW is assumed to be 0. |
| 77 | */ |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 78 | |
| 79 | #define CON_SOCK_STATE_NEW 0 /* -> CLOSED */ |
| 80 | #define CON_SOCK_STATE_CLOSED 1 /* -> CONNECTING */ |
| 81 | #define CON_SOCK_STATE_CONNECTING 2 /* -> CONNECTED or -> CLOSING */ |
| 82 | #define CON_SOCK_STATE_CONNECTED 3 /* -> CLOSING or -> CLOSED */ |
| 83 | #define CON_SOCK_STATE_CLOSING 4 /* -> CLOSED */ |
| 84 | |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 85 | /* |
| 86 | * connection states |
| 87 | */ |
| 88 | #define CON_STATE_CLOSED 1 /* -> PREOPEN */ |
| 89 | #define CON_STATE_PREOPEN 2 /* -> CONNECTING, CLOSED */ |
| 90 | #define CON_STATE_CONNECTING 3 /* -> NEGOTIATING, CLOSED */ |
| 91 | #define CON_STATE_NEGOTIATING 4 /* -> OPEN, CLOSED */ |
| 92 | #define CON_STATE_OPEN 5 /* -> STANDBY, CLOSED */ |
| 93 | #define CON_STATE_STANDBY 6 /* -> PREOPEN, CLOSED */ |
| 94 | |
Sage Weil | 4a86169 | 2012-07-20 17:29:55 -0700 | [diff] [blame] | 95 | /* |
| 96 | * ceph_connection flag bits |
| 97 | */ |
| 98 | #define CON_FLAG_LOSSYTX 0 /* we can close channel or drop |
| 99 | * messages on errors */ |
| 100 | #define CON_FLAG_KEEPALIVE_PENDING 1 /* we need to send a keepalive */ |
| 101 | #define CON_FLAG_WRITE_PENDING 2 /* we have data ready to send */ |
| 102 | #define CON_FLAG_SOCK_CLOSED 3 /* socket state changed to closed */ |
| 103 | #define CON_FLAG_BACKOFF 4 /* need to retry queuing delayed work */ |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 104 | |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 105 | static bool con_flag_valid(unsigned long con_flag) |
| 106 | { |
| 107 | switch (con_flag) { |
| 108 | case CON_FLAG_LOSSYTX: |
| 109 | case CON_FLAG_KEEPALIVE_PENDING: |
| 110 | case CON_FLAG_WRITE_PENDING: |
| 111 | case CON_FLAG_SOCK_CLOSED: |
| 112 | case CON_FLAG_BACKOFF: |
| 113 | return true; |
| 114 | default: |
| 115 | return false; |
| 116 | } |
| 117 | } |
| 118 | |
| 119 | static void con_flag_clear(struct ceph_connection *con, unsigned long con_flag) |
| 120 | { |
| 121 | BUG_ON(!con_flag_valid(con_flag)); |
| 122 | |
| 123 | clear_bit(con_flag, &con->flags); |
| 124 | } |
| 125 | |
| 126 | static void con_flag_set(struct ceph_connection *con, unsigned long con_flag) |
| 127 | { |
| 128 | BUG_ON(!con_flag_valid(con_flag)); |
| 129 | |
| 130 | set_bit(con_flag, &con->flags); |
| 131 | } |
| 132 | |
| 133 | static bool con_flag_test(struct ceph_connection *con, unsigned long con_flag) |
| 134 | { |
| 135 | BUG_ON(!con_flag_valid(con_flag)); |
| 136 | |
| 137 | return test_bit(con_flag, &con->flags); |
| 138 | } |
| 139 | |
| 140 | static bool con_flag_test_and_clear(struct ceph_connection *con, |
| 141 | unsigned long con_flag) |
| 142 | { |
| 143 | BUG_ON(!con_flag_valid(con_flag)); |
| 144 | |
| 145 | return test_and_clear_bit(con_flag, &con->flags); |
| 146 | } |
| 147 | |
| 148 | static bool con_flag_test_and_set(struct ceph_connection *con, |
| 149 | unsigned long con_flag) |
| 150 | { |
| 151 | BUG_ON(!con_flag_valid(con_flag)); |
| 152 | |
| 153 | return test_and_set_bit(con_flag, &con->flags); |
| 154 | } |
| 155 | |
Alex Elder | e3d5d63 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 156 | /* Slab caches for frequently-allocated structures */ |
| 157 | |
| 158 | static struct kmem_cache *ceph_msg_cache; |
Alex Elder | 81b36be | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 159 | static struct kmem_cache *ceph_msg_data_cache; |
Alex Elder | e3d5d63 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 160 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 161 | /* static tag bytes (protocol control messages) */ |
| 162 | static char tag_msg = CEPH_MSGR_TAG_MSG; |
| 163 | static char tag_ack = CEPH_MSGR_TAG_ACK; |
| 164 | static char tag_keepalive = CEPH_MSGR_TAG_KEEPALIVE; |
| 165 | |
Sage Weil | a6a5349 | 2010-04-13 14:07:07 -0700 | [diff] [blame] | 166 | #ifdef CONFIG_LOCKDEP |
| 167 | static struct lock_class_key socket_class; |
| 168 | #endif |
| 169 | |
Alex Elder | 84495f4 | 2012-02-15 07:43:55 -0600 | [diff] [blame] | 170 | /* |
| 171 | * When skipping (ignoring) a block of input we read it into a "skip |
| 172 | * buffer," which is this many bytes in size. |
| 173 | */ |
| 174 | #define SKIP_BUF_SIZE 1024 |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 175 | |
| 176 | static void queue_con(struct ceph_connection *con); |
| 177 | static void con_work(struct work_struct *); |
Alex Elder | 9320926 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 178 | static void con_fault(struct ceph_connection *con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 179 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 180 | /* |
Alex Elder | f64a931 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 181 | * Nicely render a sockaddr as a string. An array of formatted |
| 182 | * strings is used, to approximate reentrancy. |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 183 | */ |
Alex Elder | f64a931 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 184 | #define ADDR_STR_COUNT_LOG 5 /* log2(# address strings in array) */ |
| 185 | #define ADDR_STR_COUNT (1 << ADDR_STR_COUNT_LOG) |
| 186 | #define ADDR_STR_COUNT_MASK (ADDR_STR_COUNT - 1) |
| 187 | #define MAX_ADDR_STR_LEN 64 /* 54 is enough */ |
| 188 | |
| 189 | static char addr_str[ADDR_STR_COUNT][MAX_ADDR_STR_LEN]; |
| 190 | static atomic_t addr_str_seq = ATOMIC_INIT(0); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 191 | |
Alex Elder | 5766651 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 192 | static struct page *zero_page; /* used in certain error cases */ |
Alex Elder | 5766651 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 193 | |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 194 | const char *ceph_pr_addr(const struct sockaddr_storage *ss) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 195 | { |
| 196 | int i; |
| 197 | char *s; |
Alex Elder | 99f0f3b | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 198 | struct sockaddr_in *in4 = (struct sockaddr_in *) ss; |
| 199 | struct sockaddr_in6 *in6 = (struct sockaddr_in6 *) ss; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 200 | |
Alex Elder | f64a931 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 201 | i = atomic_inc_return(&addr_str_seq) & ADDR_STR_COUNT_MASK; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 202 | s = addr_str[i]; |
| 203 | |
| 204 | switch (ss->ss_family) { |
| 205 | case AF_INET: |
Alex Elder | bd40614 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 206 | snprintf(s, MAX_ADDR_STR_LEN, "%pI4:%hu", &in4->sin_addr, |
| 207 | ntohs(in4->sin_port)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 208 | break; |
| 209 | |
| 210 | case AF_INET6: |
Alex Elder | bd40614 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 211 | snprintf(s, MAX_ADDR_STR_LEN, "[%pI6c]:%hu", &in6->sin6_addr, |
| 212 | ntohs(in6->sin6_port)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 213 | break; |
| 214 | |
| 215 | default: |
Alex Elder | d3002b9 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 216 | snprintf(s, MAX_ADDR_STR_LEN, "(unknown sockaddr family %hu)", |
| 217 | ss->ss_family); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 218 | } |
| 219 | |
| 220 | return s; |
| 221 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 222 | EXPORT_SYMBOL(ceph_pr_addr); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 223 | |
Sage Weil | 63f2d21 | 2009-11-03 15:17:56 -0800 | [diff] [blame] | 224 | static void encode_my_addr(struct ceph_messenger *msgr) |
| 225 | { |
| 226 | memcpy(&msgr->my_enc_addr, &msgr->inst.addr, sizeof(msgr->my_enc_addr)); |
| 227 | ceph_encode_addr(&msgr->my_enc_addr); |
| 228 | } |
| 229 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 230 | /* |
| 231 | * work queue for all reading and writing to/from the socket. |
| 232 | */ |
Alex Elder | e0f43c9 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 233 | static struct workqueue_struct *ceph_msgr_wq; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 234 | |
Alex Elder | e3d5d63 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 235 | static int ceph_msgr_slab_init(void) |
| 236 | { |
| 237 | BUG_ON(ceph_msg_cache); |
| 238 | ceph_msg_cache = kmem_cache_create("ceph_msg", |
| 239 | sizeof (struct ceph_msg), |
| 240 | __alignof__(struct ceph_msg), 0, NULL); |
Alex Elder | 81b36be | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 241 | |
| 242 | if (!ceph_msg_cache) |
| 243 | return -ENOMEM; |
| 244 | |
| 245 | BUG_ON(ceph_msg_data_cache); |
| 246 | ceph_msg_data_cache = kmem_cache_create("ceph_msg_data", |
| 247 | sizeof (struct ceph_msg_data), |
| 248 | __alignof__(struct ceph_msg_data), |
| 249 | 0, NULL); |
| 250 | if (ceph_msg_data_cache) |
| 251 | return 0; |
| 252 | |
| 253 | kmem_cache_destroy(ceph_msg_cache); |
| 254 | ceph_msg_cache = NULL; |
| 255 | |
| 256 | return -ENOMEM; |
Alex Elder | e3d5d63 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 257 | } |
| 258 | |
| 259 | static void ceph_msgr_slab_exit(void) |
| 260 | { |
Alex Elder | 81b36be | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 261 | BUG_ON(!ceph_msg_data_cache); |
| 262 | kmem_cache_destroy(ceph_msg_data_cache); |
| 263 | ceph_msg_data_cache = NULL; |
| 264 | |
Alex Elder | e3d5d63 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 265 | BUG_ON(!ceph_msg_cache); |
| 266 | kmem_cache_destroy(ceph_msg_cache); |
| 267 | ceph_msg_cache = NULL; |
| 268 | } |
| 269 | |
Alex Elder | 1541716 | 2013-02-19 12:25:56 -0600 | [diff] [blame] | 270 | static void _ceph_msgr_exit(void) |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 271 | { |
Alex Elder | d3002b9 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 272 | if (ceph_msgr_wq) { |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 273 | destroy_workqueue(ceph_msgr_wq); |
Alex Elder | d3002b9 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 274 | ceph_msgr_wq = NULL; |
| 275 | } |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 276 | |
Alex Elder | e3d5d63 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 277 | ceph_msgr_slab_exit(); |
| 278 | |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 279 | BUG_ON(zero_page == NULL); |
| 280 | kunmap(zero_page); |
| 281 | page_cache_release(zero_page); |
| 282 | zero_page = NULL; |
| 283 | } |
| 284 | |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 285 | int ceph_msgr_init(void) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 286 | { |
Alex Elder | 5766651 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 287 | BUG_ON(zero_page != NULL); |
| 288 | zero_page = ZERO_PAGE(0); |
| 289 | page_cache_get(zero_page); |
| 290 | |
Alex Elder | e3d5d63 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 291 | if (ceph_msgr_slab_init()) |
| 292 | return -ENOMEM; |
| 293 | |
Tejun Heo | 4d1829a | 2013-07-30 08:40:27 -0400 | [diff] [blame] | 294 | ceph_msgr_wq = alloc_workqueue("ceph-msgr", 0, 0); |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 295 | if (ceph_msgr_wq) |
| 296 | return 0; |
Alex Elder | 5766651 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 297 | |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 298 | pr_err("msgr_init failed to create workqueue\n"); |
| 299 | _ceph_msgr_exit(); |
Alex Elder | 5766651 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 300 | |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 301 | return -ENOMEM; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 302 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 303 | EXPORT_SYMBOL(ceph_msgr_init); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 304 | |
| 305 | void ceph_msgr_exit(void) |
| 306 | { |
Alex Elder | 5766651 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 307 | BUG_ON(ceph_msgr_wq == NULL); |
Alex Elder | 5766651 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 308 | |
Alex Elder | 6173d1f | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 309 | _ceph_msgr_exit(); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 310 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 311 | EXPORT_SYMBOL(ceph_msgr_exit); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 312 | |
Yehuda Sadeh | cd84db6 | 2010-06-11 16:58:48 -0700 | [diff] [blame] | 313 | void ceph_msgr_flush(void) |
Sage Weil | a922d38 | 2010-05-29 09:41:23 -0700 | [diff] [blame] | 314 | { |
| 315 | flush_workqueue(ceph_msgr_wq); |
| 316 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 317 | EXPORT_SYMBOL(ceph_msgr_flush); |
Sage Weil | a922d38 | 2010-05-29 09:41:23 -0700 | [diff] [blame] | 318 | |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 319 | /* Connection socket state transition functions */ |
| 320 | |
| 321 | static void con_sock_state_init(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_NEW)) |
| 327 | printk("%s: unexpected old state %d\n", __func__, old_state); |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 328 | dout("%s con %p sock %d -> %d\n", __func__, con, old_state, |
| 329 | CON_SOCK_STATE_CLOSED); |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 330 | } |
| 331 | |
| 332 | static void con_sock_state_connecting(struct ceph_connection *con) |
| 333 | { |
| 334 | int old_state; |
| 335 | |
| 336 | old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CONNECTING); |
| 337 | if (WARN_ON(old_state != CON_SOCK_STATE_CLOSED)) |
| 338 | printk("%s: unexpected old state %d\n", __func__, old_state); |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 339 | dout("%s con %p sock %d -> %d\n", __func__, con, old_state, |
| 340 | CON_SOCK_STATE_CONNECTING); |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 341 | } |
| 342 | |
| 343 | static void con_sock_state_connected(struct ceph_connection *con) |
| 344 | { |
| 345 | int old_state; |
| 346 | |
| 347 | old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CONNECTED); |
| 348 | if (WARN_ON(old_state != CON_SOCK_STATE_CONNECTING)) |
| 349 | printk("%s: unexpected old state %d\n", __func__, old_state); |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 350 | dout("%s con %p sock %d -> %d\n", __func__, con, old_state, |
| 351 | CON_SOCK_STATE_CONNECTED); |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 352 | } |
| 353 | |
| 354 | static void con_sock_state_closing(struct ceph_connection *con) |
| 355 | { |
| 356 | int old_state; |
| 357 | |
| 358 | old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CLOSING); |
| 359 | if (WARN_ON(old_state != CON_SOCK_STATE_CONNECTING && |
| 360 | old_state != CON_SOCK_STATE_CONNECTED && |
| 361 | old_state != CON_SOCK_STATE_CLOSING)) |
| 362 | printk("%s: unexpected old state %d\n", __func__, old_state); |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 363 | dout("%s con %p sock %d -> %d\n", __func__, con, old_state, |
| 364 | CON_SOCK_STATE_CLOSING); |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 365 | } |
| 366 | |
| 367 | static void con_sock_state_closed(struct ceph_connection *con) |
| 368 | { |
| 369 | int old_state; |
| 370 | |
| 371 | old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CLOSED); |
| 372 | if (WARN_ON(old_state != CON_SOCK_STATE_CONNECTED && |
Sage Weil | fbb85a4 | 2012-06-27 12:31:02 -0700 | [diff] [blame] | 373 | old_state != CON_SOCK_STATE_CLOSING && |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 374 | old_state != CON_SOCK_STATE_CONNECTING && |
| 375 | old_state != CON_SOCK_STATE_CLOSED)) |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 376 | printk("%s: unexpected old state %d\n", __func__, old_state); |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 377 | dout("%s con %p sock %d -> %d\n", __func__, con, old_state, |
| 378 | CON_SOCK_STATE_CLOSED); |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 379 | } |
Sage Weil | a922d38 | 2010-05-29 09:41:23 -0700 | [diff] [blame] | 380 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 381 | /* |
| 382 | * socket callback functions |
| 383 | */ |
| 384 | |
| 385 | /* data available on socket, or listen socket received a connect */ |
David S. Miller | 676d236 | 2014-04-11 16:15:36 -0400 | [diff] [blame] | 386 | static void ceph_sock_data_ready(struct sock *sk) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 387 | { |
Alex Elder | bd40614 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 388 | struct ceph_connection *con = sk->sk_user_data; |
Guanjun He | a2a3258 | 2012-07-08 19:50:33 -0700 | [diff] [blame] | 389 | if (atomic_read(&con->msgr->stopping)) { |
| 390 | return; |
| 391 | } |
Alex Elder | bd40614 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 392 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 393 | if (sk->sk_state != TCP_CLOSE_WAIT) { |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 394 | dout("%s on %p state = %lu, queueing work\n", __func__, |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 395 | con, con->state); |
| 396 | queue_con(con); |
| 397 | } |
| 398 | } |
| 399 | |
| 400 | /* socket has buffer space for writing */ |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 401 | static void ceph_sock_write_space(struct sock *sk) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 402 | { |
Alex Elder | d3002b9 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 403 | struct ceph_connection *con = sk->sk_user_data; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 404 | |
Jim Schutt | 182fac2 | 2012-02-29 08:30:58 -0700 | [diff] [blame] | 405 | /* only queue to workqueue if there is data we want to write, |
| 406 | * and there is sufficient space in the socket buffer to accept |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 407 | * more data. clear SOCK_NOSPACE so that ceph_sock_write_space() |
Jim Schutt | 182fac2 | 2012-02-29 08:30:58 -0700 | [diff] [blame] | 408 | * doesn't get called again until try_write() fills the socket |
| 409 | * buffer. See net/ipv4/tcp_input.c:tcp_check_space() |
| 410 | * and net/core/stream.c:sk_stream_write_space(). |
| 411 | */ |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 412 | if (con_flag_test(con, CON_FLAG_WRITE_PENDING)) { |
Eric Dumazet | 64dc613 | 2013-07-22 20:26:31 -0700 | [diff] [blame] | 413 | if (sk_stream_is_writeable(sk)) { |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 414 | dout("%s %p queueing write work\n", __func__, con); |
Jim Schutt | 182fac2 | 2012-02-29 08:30:58 -0700 | [diff] [blame] | 415 | clear_bit(SOCK_NOSPACE, &sk->sk_socket->flags); |
| 416 | queue_con(con); |
| 417 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 418 | } else { |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 419 | dout("%s %p nothing to write\n", __func__, con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 420 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 421 | } |
| 422 | |
| 423 | /* socket's state has changed */ |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 424 | static void ceph_sock_state_change(struct sock *sk) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 425 | { |
Alex Elder | bd40614 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 426 | struct ceph_connection *con = sk->sk_user_data; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 427 | |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 428 | dout("%s %p state = %lu sk_state = %u\n", __func__, |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 429 | con, con->state, sk->sk_state); |
| 430 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 431 | switch (sk->sk_state) { |
| 432 | case TCP_CLOSE: |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 433 | dout("%s TCP_CLOSE\n", __func__); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 434 | case TCP_CLOSE_WAIT: |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 435 | dout("%s TCP_CLOSE_WAIT\n", __func__); |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 436 | con_sock_state_closing(con); |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 437 | con_flag_set(con, CON_FLAG_SOCK_CLOSED); |
Alex Elder | d65c9e0 | 2012-06-20 21:53:53 -0500 | [diff] [blame] | 438 | queue_con(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 439 | break; |
| 440 | case TCP_ESTABLISHED: |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 441 | dout("%s TCP_ESTABLISHED\n", __func__); |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 442 | con_sock_state_connected(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 443 | queue_con(con); |
| 444 | break; |
Alex Elder | d3002b9 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 445 | default: /* Everything else is uninteresting */ |
| 446 | break; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 447 | } |
| 448 | } |
| 449 | |
| 450 | /* |
| 451 | * set up socket callbacks |
| 452 | */ |
| 453 | static void set_sock_callbacks(struct socket *sock, |
| 454 | struct ceph_connection *con) |
| 455 | { |
| 456 | struct sock *sk = sock->sk; |
Alex Elder | bd40614 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 457 | sk->sk_user_data = con; |
Alex Elder | 327800b | 2012-05-22 11:41:43 -0500 | [diff] [blame] | 458 | sk->sk_data_ready = ceph_sock_data_ready; |
| 459 | sk->sk_write_space = ceph_sock_write_space; |
| 460 | sk->sk_state_change = ceph_sock_state_change; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 461 | } |
| 462 | |
| 463 | |
| 464 | /* |
| 465 | * socket helpers |
| 466 | */ |
| 467 | |
| 468 | /* |
| 469 | * initiate connection to a remote socket. |
| 470 | */ |
Alex Elder | 41617d0 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 471 | static int ceph_tcp_connect(struct ceph_connection *con) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 472 | { |
Sage Weil | f91d347 | 2010-07-01 15:18:31 -0700 | [diff] [blame] | 473 | struct sockaddr_storage *paddr = &con->peer_addr.in_addr; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 474 | struct socket *sock; |
| 475 | int ret; |
| 476 | |
| 477 | BUG_ON(con->sock); |
Sage Weil | f91d347 | 2010-07-01 15:18:31 -0700 | [diff] [blame] | 478 | ret = sock_create_kern(con->peer_addr.in_addr.ss_family, SOCK_STREAM, |
| 479 | IPPROTO_TCP, &sock); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 480 | if (ret) |
Alex Elder | 41617d0 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 481 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 482 | sock->sk->sk_allocation = GFP_NOFS; |
| 483 | |
Sage Weil | a6a5349 | 2010-04-13 14:07:07 -0700 | [diff] [blame] | 484 | #ifdef CONFIG_LOCKDEP |
| 485 | lockdep_set_class(&sock->sk->sk_lock, &socket_class); |
| 486 | #endif |
| 487 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 488 | set_sock_callbacks(sock, con); |
| 489 | |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 490 | dout("connect %s\n", ceph_pr_addr(&con->peer_addr.in_addr)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 491 | |
Sage Weil | 89a86be | 2012-06-09 14:19:21 -0700 | [diff] [blame] | 492 | con_sock_state_connecting(con); |
Sage Weil | f91d347 | 2010-07-01 15:18:31 -0700 | [diff] [blame] | 493 | ret = sock->ops->connect(sock, (struct sockaddr *)paddr, sizeof(*paddr), |
| 494 | O_NONBLOCK); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 495 | if (ret == -EINPROGRESS) { |
| 496 | dout("connect %s EINPROGRESS sk_state = %u\n", |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 497 | ceph_pr_addr(&con->peer_addr.in_addr), |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 498 | sock->sk->sk_state); |
Alex Elder | a5bc312 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 499 | } else if (ret < 0) { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 500 | pr_err("connect %s error %d\n", |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 501 | ceph_pr_addr(&con->peer_addr.in_addr), ret); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 502 | sock_release(sock); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 503 | con->error_msg = "connect error"; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 504 | |
Alex Elder | 41617d0 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 505 | return ret; |
Alex Elder | a5bc312 | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 506 | } |
| 507 | con->sock = sock; |
Alex Elder | 41617d0 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 508 | return 0; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 509 | } |
| 510 | |
| 511 | static int ceph_tcp_recvmsg(struct socket *sock, void *buf, size_t len) |
| 512 | { |
| 513 | struct kvec iov = {buf, len}; |
| 514 | struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL }; |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 515 | int r; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 516 | |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 517 | r = kernel_recvmsg(sock, &msg, &iov, 1, len, msg.msg_flags); |
| 518 | if (r == -EAGAIN) |
| 519 | r = 0; |
| 520 | return r; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 521 | } |
| 522 | |
Alex Elder | afb3d90 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 523 | static int ceph_tcp_recvpage(struct socket *sock, struct page *page, |
| 524 | int page_offset, size_t length) |
| 525 | { |
| 526 | void *kaddr; |
| 527 | int ret; |
| 528 | |
| 529 | BUG_ON(page_offset + length > PAGE_SIZE); |
| 530 | |
| 531 | kaddr = kmap(page); |
| 532 | BUG_ON(!kaddr); |
| 533 | ret = ceph_tcp_recvmsg(sock, kaddr + page_offset, length); |
| 534 | kunmap(page); |
| 535 | |
| 536 | return ret; |
| 537 | } |
| 538 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 539 | /* |
| 540 | * write something. @more is true if caller will be sending more data |
| 541 | * shortly. |
| 542 | */ |
| 543 | static int ceph_tcp_sendmsg(struct socket *sock, struct kvec *iov, |
| 544 | size_t kvlen, size_t len, int more) |
| 545 | { |
| 546 | struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL }; |
Sage Weil | 42961d2 | 2011-01-25 08:19:34 -0800 | [diff] [blame] | 547 | int r; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 548 | |
| 549 | if (more) |
| 550 | msg.msg_flags |= MSG_MORE; |
| 551 | else |
| 552 | msg.msg_flags |= MSG_EOR; /* superfluous, but what the hell */ |
| 553 | |
Sage Weil | 42961d2 | 2011-01-25 08:19:34 -0800 | [diff] [blame] | 554 | r = kernel_sendmsg(sock, &msg, iov, kvlen, len); |
| 555 | if (r == -EAGAIN) |
| 556 | r = 0; |
| 557 | return r; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 558 | } |
| 559 | |
Chunwei Chen | 178eda2 | 2014-04-23 12:35:09 +0800 | [diff] [blame] | 560 | static int __ceph_tcp_sendpage(struct socket *sock, struct page *page, |
Alex Elder | e1dcb12 | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 561 | int offset, size_t size, bool more) |
Alex Elder | 3173913 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 562 | { |
| 563 | int flags = MSG_DONTWAIT | MSG_NOSIGNAL | (more ? MSG_MORE : MSG_EOR); |
| 564 | int ret; |
| 565 | |
| 566 | ret = kernel_sendpage(sock, page, offset, size, flags); |
| 567 | if (ret == -EAGAIN) |
| 568 | ret = 0; |
| 569 | |
| 570 | return ret; |
| 571 | } |
| 572 | |
Chunwei Chen | 178eda2 | 2014-04-23 12:35:09 +0800 | [diff] [blame] | 573 | static int ceph_tcp_sendpage(struct socket *sock, struct page *page, |
| 574 | int offset, size_t size, bool more) |
| 575 | { |
| 576 | int ret; |
| 577 | struct kvec iov; |
| 578 | |
| 579 | /* sendpage cannot properly handle pages with page_count == 0, |
| 580 | * we need to fallback to sendmsg if that's the case */ |
| 581 | if (page_count(page) >= 1) |
| 582 | return __ceph_tcp_sendpage(sock, page, offset, size, more); |
| 583 | |
| 584 | iov.iov_base = kmap(page) + offset; |
| 585 | iov.iov_len = size; |
| 586 | ret = ceph_tcp_sendmsg(sock, &iov, 1, size, more); |
| 587 | kunmap(page); |
| 588 | |
| 589 | return ret; |
| 590 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 591 | |
| 592 | /* |
| 593 | * Shutdown/close the socket for the given connection. |
| 594 | */ |
| 595 | static int con_close_socket(struct ceph_connection *con) |
| 596 | { |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 597 | int rc = 0; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 598 | |
| 599 | dout("con_close_socket on %p sock %p\n", con, con->sock); |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 600 | if (con->sock) { |
| 601 | rc = con->sock->ops->shutdown(con->sock, SHUT_RDWR); |
| 602 | sock_release(con->sock); |
| 603 | con->sock = NULL; |
| 604 | } |
Alex Elder | 456ea46 | 2012-06-20 21:53:53 -0500 | [diff] [blame] | 605 | |
| 606 | /* |
Sage Weil | 4a86169 | 2012-07-20 17:29:55 -0700 | [diff] [blame] | 607 | * Forcibly clear the SOCK_CLOSED flag. It gets set |
Alex Elder | 456ea46 | 2012-06-20 21:53:53 -0500 | [diff] [blame] | 608 | * independent of the connection mutex, and we could have |
| 609 | * received a socket close event before we had the chance to |
| 610 | * shut the socket down. |
| 611 | */ |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 612 | con_flag_clear(con, CON_FLAG_SOCK_CLOSED); |
Sage Weil | 8007b8d | 2012-07-30 18:16:16 -0700 | [diff] [blame] | 613 | |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 614 | con_sock_state_closed(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 615 | return rc; |
| 616 | } |
| 617 | |
| 618 | /* |
| 619 | * Reset a connection. Discard all incoming and outgoing messages |
| 620 | * and clear *_seq state. |
| 621 | */ |
| 622 | static void ceph_msg_remove(struct ceph_msg *msg) |
| 623 | { |
| 624 | list_del_init(&msg->list_head); |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 625 | BUG_ON(msg->con == NULL); |
Sage Weil | 36eb71a | 2012-06-21 12:47:08 -0700 | [diff] [blame] | 626 | msg->con->ops->put(msg->con); |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 627 | msg->con = NULL; |
| 628 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 629 | ceph_msg_put(msg); |
| 630 | } |
| 631 | static void ceph_msg_remove_list(struct list_head *head) |
| 632 | { |
| 633 | while (!list_empty(head)) { |
| 634 | struct ceph_msg *msg = list_first_entry(head, struct ceph_msg, |
| 635 | list_head); |
| 636 | ceph_msg_remove(msg); |
| 637 | } |
| 638 | } |
| 639 | |
| 640 | static void reset_connection(struct ceph_connection *con) |
| 641 | { |
| 642 | /* reset connection, out_queue, msg_ and connect_seq */ |
| 643 | /* discard existing out_queue and msg_seq */ |
Sage Weil | 0fa6ebc | 2012-12-27 20:27:04 -0600 | [diff] [blame] | 644 | dout("reset_connection %p\n", con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 645 | ceph_msg_remove_list(&con->out_queue); |
| 646 | ceph_msg_remove_list(&con->out_sent); |
| 647 | |
Sage Weil | cf3e5c4 | 2009-12-11 09:48:05 -0800 | [diff] [blame] | 648 | if (con->in_msg) { |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 649 | BUG_ON(con->in_msg->con != con); |
| 650 | con->in_msg->con = NULL; |
Sage Weil | cf3e5c4 | 2009-12-11 09:48:05 -0800 | [diff] [blame] | 651 | ceph_msg_put(con->in_msg); |
| 652 | con->in_msg = NULL; |
Sage Weil | 36eb71a | 2012-06-21 12:47:08 -0700 | [diff] [blame] | 653 | con->ops->put(con); |
Sage Weil | cf3e5c4 | 2009-12-11 09:48:05 -0800 | [diff] [blame] | 654 | } |
| 655 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 656 | con->connect_seq = 0; |
| 657 | con->out_seq = 0; |
Sage Weil | c86a293 | 2009-12-14 14:04:30 -0800 | [diff] [blame] | 658 | if (con->out_msg) { |
| 659 | ceph_msg_put(con->out_msg); |
| 660 | con->out_msg = NULL; |
| 661 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 662 | con->in_seq = 0; |
Sage Weil | 0e0d5e0 | 2010-04-02 16:07:19 -0700 | [diff] [blame] | 663 | con->in_seq_acked = 0; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 664 | } |
| 665 | |
| 666 | /* |
| 667 | * mark a peer down. drop any open connections. |
| 668 | */ |
| 669 | void ceph_con_close(struct ceph_connection *con) |
| 670 | { |
Sage Weil | 8c50c81 | 2012-07-30 16:24:37 -0700 | [diff] [blame] | 671 | mutex_lock(&con->mutex); |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 672 | dout("con_close %p peer %s\n", con, |
| 673 | ceph_pr_addr(&con->peer_addr.in_addr)); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 674 | con->state = CON_STATE_CLOSED; |
Alex Elder | a5988c4 | 2012-05-29 11:04:58 -0500 | [diff] [blame] | 675 | |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 676 | con_flag_clear(con, CON_FLAG_LOSSYTX); /* so we retry next connect */ |
| 677 | con_flag_clear(con, CON_FLAG_KEEPALIVE_PENDING); |
| 678 | con_flag_clear(con, CON_FLAG_WRITE_PENDING); |
| 679 | con_flag_clear(con, CON_FLAG_BACKOFF); |
Alex Elder | a5988c4 | 2012-05-29 11:04:58 -0500 | [diff] [blame] | 680 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 681 | reset_connection(con); |
Sage Weil | 6f2bc3f | 2010-04-02 16:16:34 -0700 | [diff] [blame] | 682 | con->peer_global_seq = 0; |
Sage Weil | 91e45ce3 | 2010-02-15 12:05:09 -0800 | [diff] [blame] | 683 | cancel_delayed_work(&con->work); |
Sage Weil | ee76e07 | 2012-07-20 16:45:49 -0700 | [diff] [blame] | 684 | con_close_socket(con); |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 685 | mutex_unlock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 686 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 687 | EXPORT_SYMBOL(ceph_con_close); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 688 | |
| 689 | /* |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 690 | * Reopen a closed connection, with a new peer address. |
| 691 | */ |
Sage Weil | b7a9e5d | 2012-06-27 12:24:08 -0700 | [diff] [blame] | 692 | void ceph_con_open(struct ceph_connection *con, |
| 693 | __u8 entity_type, __u64 entity_num, |
| 694 | struct ceph_entity_addr *addr) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 695 | { |
Sage Weil | 5469155 | 2012-07-30 16:21:40 -0700 | [diff] [blame] | 696 | mutex_lock(&con->mutex); |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 697 | 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] | 698 | |
Alex Elder | 122070a | 2012-12-26 10:43:57 -0600 | [diff] [blame] | 699 | WARN_ON(con->state != CON_STATE_CLOSED); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 700 | con->state = CON_STATE_PREOPEN; |
Alex Elder | a5988c4 | 2012-05-29 11:04:58 -0500 | [diff] [blame] | 701 | |
Sage Weil | b7a9e5d | 2012-06-27 12:24:08 -0700 | [diff] [blame] | 702 | con->peer_name.type = (__u8) entity_type; |
| 703 | con->peer_name.num = cpu_to_le64(entity_num); |
| 704 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 705 | memcpy(&con->peer_addr, addr, sizeof(*addr)); |
Sage Weil | 03c677e | 2009-11-20 15:14:15 -0800 | [diff] [blame] | 706 | con->delay = 0; /* reset backoff memory */ |
Sage Weil | 5469155 | 2012-07-30 16:21:40 -0700 | [diff] [blame] | 707 | mutex_unlock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 708 | queue_con(con); |
| 709 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 710 | EXPORT_SYMBOL(ceph_con_open); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 711 | |
| 712 | /* |
Sage Weil | 87b315a | 2010-03-22 14:51:18 -0700 | [diff] [blame] | 713 | * return true if this connection ever successfully opened |
| 714 | */ |
| 715 | bool ceph_con_opened(struct ceph_connection *con) |
| 716 | { |
| 717 | return con->connect_seq > 0; |
| 718 | } |
| 719 | |
| 720 | /* |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 721 | * initialize a new connection. |
| 722 | */ |
Alex Elder | 1bfd89f | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 723 | void ceph_con_init(struct ceph_connection *con, void *private, |
| 724 | const struct ceph_connection_operations *ops, |
Sage Weil | b7a9e5d | 2012-06-27 12:24:08 -0700 | [diff] [blame] | 725 | struct ceph_messenger *msgr) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 726 | { |
| 727 | dout("con_init %p\n", con); |
| 728 | memset(con, 0, sizeof(*con)); |
Alex Elder | 1bfd89f | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 729 | con->private = private; |
| 730 | con->ops = ops; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 731 | con->msgr = msgr; |
Alex Elder | ce2c890 | 2012-05-22 22:15:49 -0500 | [diff] [blame] | 732 | |
| 733 | con_sock_state_init(con); |
| 734 | |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 735 | mutex_init(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 736 | INIT_LIST_HEAD(&con->out_queue); |
| 737 | INIT_LIST_HEAD(&con->out_sent); |
| 738 | INIT_DELAYED_WORK(&con->work, con_work); |
Alex Elder | a5988c4 | 2012-05-29 11:04:58 -0500 | [diff] [blame] | 739 | |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 740 | con->state = CON_STATE_CLOSED; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 741 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 742 | EXPORT_SYMBOL(ceph_con_init); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 743 | |
| 744 | |
| 745 | /* |
| 746 | * We maintain a global counter to order connection attempts. Get |
| 747 | * a unique seq greater than @gt. |
| 748 | */ |
| 749 | static u32 get_global_seq(struct ceph_messenger *msgr, u32 gt) |
| 750 | { |
| 751 | u32 ret; |
| 752 | |
| 753 | spin_lock(&msgr->global_seq_lock); |
| 754 | if (msgr->global_seq < gt) |
| 755 | msgr->global_seq = gt; |
| 756 | ret = ++msgr->global_seq; |
| 757 | spin_unlock(&msgr->global_seq_lock); |
| 758 | return ret; |
| 759 | } |
| 760 | |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 761 | static void con_out_kvec_reset(struct ceph_connection *con) |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 762 | { |
| 763 | con->out_kvec_left = 0; |
| 764 | con->out_kvec_bytes = 0; |
| 765 | con->out_kvec_cur = &con->out_kvec[0]; |
| 766 | } |
| 767 | |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 768 | static void con_out_kvec_add(struct ceph_connection *con, |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 769 | size_t size, void *data) |
| 770 | { |
| 771 | int index; |
| 772 | |
| 773 | index = con->out_kvec_left; |
| 774 | BUG_ON(index >= ARRAY_SIZE(con->out_kvec)); |
| 775 | |
| 776 | con->out_kvec[index].iov_len = size; |
| 777 | con->out_kvec[index].iov_base = data; |
| 778 | con->out_kvec_left++; |
| 779 | con->out_kvec_bytes += size; |
| 780 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 781 | |
Alex Elder | df6ad1f | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 782 | #ifdef CONFIG_BLOCK |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 783 | |
| 784 | /* |
| 785 | * For a bio data item, a piece is whatever remains of the next |
| 786 | * entry in the current bio iovec, or the first entry in the next |
| 787 | * bio in the list. |
| 788 | */ |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 789 | static void ceph_msg_data_bio_cursor_init(struct ceph_msg_data_cursor *cursor, |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 790 | size_t length) |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 791 | { |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 792 | struct ceph_msg_data *data = cursor->data; |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 793 | struct bio *bio; |
| 794 | |
| 795 | BUG_ON(data->type != CEPH_MSG_DATA_BIO); |
| 796 | |
| 797 | bio = data->bio; |
| 798 | BUG_ON(!bio); |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 799 | |
Alex Elder | ca8b3a6 | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 800 | cursor->resid = min(length, data->bio_length); |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 801 | cursor->bio = bio; |
Kent Overstreet | f38a518 | 2013-08-07 14:30:24 -0700 | [diff] [blame] | 802 | cursor->bvec_iter = bio->bi_iter; |
| 803 | cursor->last_piece = |
| 804 | cursor->resid <= bio_iter_len(bio, cursor->bvec_iter); |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 805 | } |
| 806 | |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 807 | static struct page *ceph_msg_data_bio_next(struct ceph_msg_data_cursor *cursor, |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 808 | size_t *page_offset, |
| 809 | size_t *length) |
| 810 | { |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 811 | struct ceph_msg_data *data = cursor->data; |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 812 | struct bio *bio; |
Kent Overstreet | f38a518 | 2013-08-07 14:30:24 -0700 | [diff] [blame] | 813 | struct bio_vec bio_vec; |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 814 | |
| 815 | BUG_ON(data->type != CEPH_MSG_DATA_BIO); |
| 816 | |
| 817 | bio = cursor->bio; |
| 818 | BUG_ON(!bio); |
| 819 | |
Kent Overstreet | f38a518 | 2013-08-07 14:30:24 -0700 | [diff] [blame] | 820 | bio_vec = bio_iter_iovec(bio, cursor->bvec_iter); |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 821 | |
Kent Overstreet | f38a518 | 2013-08-07 14:30:24 -0700 | [diff] [blame] | 822 | *page_offset = (size_t) bio_vec.bv_offset; |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 823 | BUG_ON(*page_offset >= PAGE_SIZE); |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 824 | if (cursor->last_piece) /* pagelist offset is always 0 */ |
| 825 | *length = cursor->resid; |
| 826 | else |
Kent Overstreet | f38a518 | 2013-08-07 14:30:24 -0700 | [diff] [blame] | 827 | *length = (size_t) bio_vec.bv_len; |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 828 | BUG_ON(*length > cursor->resid); |
Alex Elder | 5df521b | 2013-03-30 15:09:59 -0500 | [diff] [blame] | 829 | BUG_ON(*page_offset + *length > PAGE_SIZE); |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 830 | |
Kent Overstreet | f38a518 | 2013-08-07 14:30:24 -0700 | [diff] [blame] | 831 | return bio_vec.bv_page; |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 832 | } |
| 833 | |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 834 | static bool ceph_msg_data_bio_advance(struct ceph_msg_data_cursor *cursor, |
| 835 | size_t bytes) |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 836 | { |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 837 | struct bio *bio; |
Kent Overstreet | f38a518 | 2013-08-07 14:30:24 -0700 | [diff] [blame] | 838 | struct bio_vec bio_vec; |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 839 | |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 840 | BUG_ON(cursor->data->type != CEPH_MSG_DATA_BIO); |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 841 | |
| 842 | bio = cursor->bio; |
| 843 | BUG_ON(!bio); |
| 844 | |
Kent Overstreet | f38a518 | 2013-08-07 14:30:24 -0700 | [diff] [blame] | 845 | bio_vec = bio_iter_iovec(bio, cursor->bvec_iter); |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 846 | |
| 847 | /* Advance the cursor offset */ |
| 848 | |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 849 | BUG_ON(cursor->resid < bytes); |
| 850 | cursor->resid -= bytes; |
Kent Overstreet | f38a518 | 2013-08-07 14:30:24 -0700 | [diff] [blame] | 851 | |
| 852 | bio_advance_iter(bio, &cursor->bvec_iter, bytes); |
| 853 | |
| 854 | if (bytes < bio_vec.bv_len) |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 855 | return false; /* more bytes to process in this segment */ |
| 856 | |
| 857 | /* Move on to the next segment, and possibly the next bio */ |
| 858 | |
Kent Overstreet | f38a518 | 2013-08-07 14:30:24 -0700 | [diff] [blame] | 859 | if (!cursor->bvec_iter.bi_size) { |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 860 | bio = bio->bi_next; |
Ilya Dryomov | 0ec1d15 | 2014-02-05 15:19:55 +0200 | [diff] [blame] | 861 | cursor->bio = bio; |
| 862 | if (bio) |
| 863 | cursor->bvec_iter = bio->bi_iter; |
| 864 | else |
| 865 | memset(&cursor->bvec_iter, 0, |
| 866 | sizeof(cursor->bvec_iter)); |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 867 | } |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 868 | |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 869 | if (!cursor->last_piece) { |
| 870 | BUG_ON(!cursor->resid); |
| 871 | BUG_ON(!bio); |
| 872 | /* A short read is OK, so use <= rather than == */ |
Kent Overstreet | f38a518 | 2013-08-07 14:30:24 -0700 | [diff] [blame] | 873 | if (cursor->resid <= bio_iter_len(bio, cursor->bvec_iter)) |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 874 | cursor->last_piece = true; |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 875 | } |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 876 | |
| 877 | return true; |
| 878 | } |
Alex Elder | ea96571 | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 879 | #endif /* CONFIG_BLOCK */ |
Alex Elder | df6ad1f | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 880 | |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 881 | /* |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 882 | * For a page array, a piece comes from the first page in the array |
| 883 | * that has not already been fully consumed. |
| 884 | */ |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 885 | static void ceph_msg_data_pages_cursor_init(struct ceph_msg_data_cursor *cursor, |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 886 | size_t length) |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 887 | { |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 888 | struct ceph_msg_data *data = cursor->data; |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 889 | int page_count; |
| 890 | |
| 891 | BUG_ON(data->type != CEPH_MSG_DATA_PAGES); |
| 892 | |
| 893 | BUG_ON(!data->pages); |
| 894 | BUG_ON(!data->length); |
| 895 | |
Alex Elder | ca8b3a6 | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 896 | cursor->resid = min(length, data->length); |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 897 | page_count = calc_pages_for(data->alignment, (u64)data->length); |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 898 | cursor->page_offset = data->alignment & ~PAGE_MASK; |
| 899 | cursor->page_index = 0; |
Alex Elder | 56fc565 | 2013-03-30 23:46:55 -0500 | [diff] [blame] | 900 | BUG_ON(page_count > (int)USHRT_MAX); |
| 901 | cursor->page_count = (unsigned short)page_count; |
| 902 | BUG_ON(length > SIZE_MAX - cursor->page_offset); |
| 903 | cursor->last_piece = (size_t)cursor->page_offset + length <= PAGE_SIZE; |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 904 | } |
| 905 | |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 906 | static struct page * |
| 907 | ceph_msg_data_pages_next(struct ceph_msg_data_cursor *cursor, |
| 908 | size_t *page_offset, size_t *length) |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 909 | { |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 910 | struct ceph_msg_data *data = cursor->data; |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 911 | |
| 912 | BUG_ON(data->type != CEPH_MSG_DATA_PAGES); |
| 913 | |
| 914 | BUG_ON(cursor->page_index >= cursor->page_count); |
| 915 | BUG_ON(cursor->page_offset >= PAGE_SIZE); |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 916 | |
| 917 | *page_offset = cursor->page_offset; |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 918 | if (cursor->last_piece) |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 919 | *length = cursor->resid; |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 920 | else |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 921 | *length = PAGE_SIZE - *page_offset; |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 922 | |
| 923 | return data->pages[cursor->page_index]; |
| 924 | } |
| 925 | |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 926 | static bool ceph_msg_data_pages_advance(struct ceph_msg_data_cursor *cursor, |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 927 | size_t bytes) |
| 928 | { |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 929 | BUG_ON(cursor->data->type != CEPH_MSG_DATA_PAGES); |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 930 | |
| 931 | BUG_ON(cursor->page_offset + bytes > PAGE_SIZE); |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 932 | |
| 933 | /* Advance the cursor page offset */ |
| 934 | |
| 935 | cursor->resid -= bytes; |
Alex Elder | 5df521b | 2013-03-30 15:09:59 -0500 | [diff] [blame] | 936 | cursor->page_offset = (cursor->page_offset + bytes) & ~PAGE_MASK; |
| 937 | if (!bytes || cursor->page_offset) |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 938 | return false; /* more bytes to process in the current page */ |
| 939 | |
Yan, Zheng | d90deda | 2014-03-23 06:50:39 +0800 | [diff] [blame] | 940 | if (!cursor->resid) |
| 941 | return false; /* no more data */ |
| 942 | |
Alex Elder | 5df521b | 2013-03-30 15:09:59 -0500 | [diff] [blame] | 943 | /* Move on to the next page; offset is already at 0 */ |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 944 | |
| 945 | BUG_ON(cursor->page_index >= cursor->page_count); |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 946 | cursor->page_index++; |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 947 | cursor->last_piece = cursor->resid <= PAGE_SIZE; |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 948 | |
| 949 | return true; |
| 950 | } |
| 951 | |
| 952 | /* |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 953 | * For a pagelist, a piece is whatever remains to be consumed in the |
| 954 | * first page in the list, or the front of the next page. |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 955 | */ |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 956 | static void |
| 957 | ceph_msg_data_pagelist_cursor_init(struct ceph_msg_data_cursor *cursor, |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 958 | size_t length) |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 959 | { |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 960 | struct ceph_msg_data *data = cursor->data; |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 961 | struct ceph_pagelist *pagelist; |
| 962 | struct page *page; |
| 963 | |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 964 | BUG_ON(data->type != CEPH_MSG_DATA_PAGELIST); |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 965 | |
| 966 | pagelist = data->pagelist; |
| 967 | BUG_ON(!pagelist); |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 968 | |
| 969 | if (!length) |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 970 | return; /* pagelist can be assigned but empty */ |
| 971 | |
| 972 | BUG_ON(list_empty(&pagelist->head)); |
| 973 | page = list_first_entry(&pagelist->head, struct page, lru); |
| 974 | |
Alex Elder | ca8b3a6 | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 975 | cursor->resid = min(length, pagelist->length); |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 976 | cursor->page = page; |
| 977 | cursor->offset = 0; |
Alex Elder | a51b272e | 2013-04-19 15:34:49 -0500 | [diff] [blame] | 978 | cursor->last_piece = cursor->resid <= PAGE_SIZE; |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 979 | } |
| 980 | |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 981 | static struct page * |
| 982 | ceph_msg_data_pagelist_next(struct ceph_msg_data_cursor *cursor, |
| 983 | size_t *page_offset, size_t *length) |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 984 | { |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 985 | struct ceph_msg_data *data = cursor->data; |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 986 | struct ceph_pagelist *pagelist; |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 987 | |
| 988 | BUG_ON(data->type != CEPH_MSG_DATA_PAGELIST); |
| 989 | |
| 990 | pagelist = data->pagelist; |
| 991 | BUG_ON(!pagelist); |
| 992 | |
| 993 | BUG_ON(!cursor->page); |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 994 | BUG_ON(cursor->offset + cursor->resid != pagelist->length); |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 995 | |
Alex Elder | 5df521b | 2013-03-30 15:09:59 -0500 | [diff] [blame] | 996 | /* offset of first page in pagelist is always 0 */ |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 997 | *page_offset = cursor->offset & ~PAGE_MASK; |
Alex Elder | 5df521b | 2013-03-30 15:09:59 -0500 | [diff] [blame] | 998 | if (cursor->last_piece) |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 999 | *length = cursor->resid; |
| 1000 | else |
| 1001 | *length = PAGE_SIZE - *page_offset; |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1002 | |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1003 | return cursor->page; |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1004 | } |
| 1005 | |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1006 | static bool ceph_msg_data_pagelist_advance(struct ceph_msg_data_cursor *cursor, |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1007 | size_t bytes) |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1008 | { |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1009 | struct ceph_msg_data *data = cursor->data; |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1010 | struct ceph_pagelist *pagelist; |
| 1011 | |
| 1012 | BUG_ON(data->type != CEPH_MSG_DATA_PAGELIST); |
| 1013 | |
| 1014 | pagelist = data->pagelist; |
| 1015 | BUG_ON(!pagelist); |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 1016 | |
| 1017 | BUG_ON(cursor->offset + cursor->resid != pagelist->length); |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1018 | BUG_ON((cursor->offset & ~PAGE_MASK) + bytes > PAGE_SIZE); |
| 1019 | |
| 1020 | /* Advance the cursor offset */ |
| 1021 | |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 1022 | cursor->resid -= bytes; |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1023 | cursor->offset += bytes; |
Alex Elder | 5df521b | 2013-03-30 15:09:59 -0500 | [diff] [blame] | 1024 | /* offset of first page in pagelist is always 0 */ |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1025 | if (!bytes || cursor->offset & ~PAGE_MASK) |
| 1026 | return false; /* more bytes to process in the current page */ |
| 1027 | |
Yan, Zheng | d90deda | 2014-03-23 06:50:39 +0800 | [diff] [blame] | 1028 | if (!cursor->resid) |
| 1029 | return false; /* no more data */ |
| 1030 | |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1031 | /* Move on to the next page */ |
| 1032 | |
| 1033 | BUG_ON(list_is_last(&cursor->page->lru, &pagelist->head)); |
| 1034 | cursor->page = list_entry_next(cursor->page, lru); |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 1035 | cursor->last_piece = cursor->resid <= PAGE_SIZE; |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1036 | |
| 1037 | return true; |
| 1038 | } |
| 1039 | |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1040 | /* |
| 1041 | * Message data is handled (sent or received) in pieces, where each |
| 1042 | * piece resides on a single page. The network layer might not |
| 1043 | * consume an entire piece at once. A data item's cursor keeps |
| 1044 | * track of which piece is next to process and how much remains to |
| 1045 | * be processed in that piece. It also tracks whether the current |
| 1046 | * piece is the last one in the data item. |
| 1047 | */ |
Alex Elder | ca8b3a6 | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 1048 | static void __ceph_msg_data_cursor_init(struct ceph_msg_data_cursor *cursor) |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1049 | { |
Alex Elder | ca8b3a6 | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 1050 | size_t length = cursor->total_resid; |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1051 | |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1052 | switch (cursor->data->type) { |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1053 | case CEPH_MSG_DATA_PAGELIST: |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1054 | ceph_msg_data_pagelist_cursor_init(cursor, length); |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1055 | break; |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 1056 | case CEPH_MSG_DATA_PAGES: |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1057 | ceph_msg_data_pages_cursor_init(cursor, length); |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 1058 | break; |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1059 | #ifdef CONFIG_BLOCK |
| 1060 | case CEPH_MSG_DATA_BIO: |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1061 | ceph_msg_data_bio_cursor_init(cursor, length); |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1062 | break; |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1063 | #endif /* CONFIG_BLOCK */ |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1064 | case CEPH_MSG_DATA_NONE: |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1065 | default: |
| 1066 | /* BUG(); */ |
| 1067 | break; |
| 1068 | } |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1069 | cursor->need_crc = true; |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1070 | } |
| 1071 | |
Alex Elder | ca8b3a6 | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 1072 | static void ceph_msg_data_cursor_init(struct ceph_msg *msg, size_t length) |
| 1073 | { |
| 1074 | struct ceph_msg_data_cursor *cursor = &msg->cursor; |
| 1075 | struct ceph_msg_data *data; |
| 1076 | |
| 1077 | BUG_ON(!length); |
| 1078 | BUG_ON(length > msg->data_length); |
| 1079 | BUG_ON(list_empty(&msg->data)); |
| 1080 | |
Alex Elder | ca8b3a6 | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 1081 | cursor->data_head = &msg->data; |
| 1082 | cursor->total_resid = length; |
| 1083 | data = list_first_entry(&msg->data, struct ceph_msg_data, links); |
| 1084 | cursor->data = data; |
| 1085 | |
| 1086 | __ceph_msg_data_cursor_init(cursor); |
| 1087 | } |
| 1088 | |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1089 | /* |
| 1090 | * Return the page containing the next piece to process for a given |
| 1091 | * data item, and supply the page offset and length of that piece. |
| 1092 | * Indicate whether this is the last piece in this data item. |
| 1093 | */ |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1094 | static struct page *ceph_msg_data_next(struct ceph_msg_data_cursor *cursor, |
| 1095 | size_t *page_offset, size_t *length, |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1096 | bool *last_piece) |
| 1097 | { |
| 1098 | struct page *page; |
| 1099 | |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1100 | switch (cursor->data->type) { |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1101 | case CEPH_MSG_DATA_PAGELIST: |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1102 | page = ceph_msg_data_pagelist_next(cursor, page_offset, length); |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1103 | break; |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 1104 | case CEPH_MSG_DATA_PAGES: |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1105 | page = ceph_msg_data_pages_next(cursor, page_offset, length); |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 1106 | break; |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1107 | #ifdef CONFIG_BLOCK |
| 1108 | case CEPH_MSG_DATA_BIO: |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1109 | page = ceph_msg_data_bio_next(cursor, page_offset, length); |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1110 | break; |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1111 | #endif /* CONFIG_BLOCK */ |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1112 | case CEPH_MSG_DATA_NONE: |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1113 | default: |
| 1114 | page = NULL; |
| 1115 | break; |
| 1116 | } |
| 1117 | BUG_ON(!page); |
| 1118 | BUG_ON(*page_offset + *length > PAGE_SIZE); |
| 1119 | BUG_ON(!*length); |
| 1120 | if (last_piece) |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1121 | *last_piece = cursor->last_piece; |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1122 | |
| 1123 | return page; |
| 1124 | } |
| 1125 | |
| 1126 | /* |
| 1127 | * Returns true if the result moves the cursor on to the next piece |
| 1128 | * of the data item. |
| 1129 | */ |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1130 | static bool ceph_msg_data_advance(struct ceph_msg_data_cursor *cursor, |
| 1131 | size_t bytes) |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1132 | { |
| 1133 | bool new_piece; |
| 1134 | |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 1135 | BUG_ON(bytes > cursor->resid); |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1136 | switch (cursor->data->type) { |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1137 | case CEPH_MSG_DATA_PAGELIST: |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1138 | new_piece = ceph_msg_data_pagelist_advance(cursor, bytes); |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1139 | break; |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 1140 | case CEPH_MSG_DATA_PAGES: |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1141 | new_piece = ceph_msg_data_pages_advance(cursor, bytes); |
Alex Elder | e766d7b | 2013-03-07 15:38:28 -0600 | [diff] [blame] | 1142 | break; |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1143 | #ifdef CONFIG_BLOCK |
| 1144 | case CEPH_MSG_DATA_BIO: |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1145 | new_piece = ceph_msg_data_bio_advance(cursor, bytes); |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1146 | break; |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1147 | #endif /* CONFIG_BLOCK */ |
Alex Elder | 6aaa451 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1148 | case CEPH_MSG_DATA_NONE: |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1149 | default: |
| 1150 | BUG(); |
| 1151 | break; |
| 1152 | } |
Alex Elder | ca8b3a6 | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 1153 | cursor->total_resid -= bytes; |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1154 | |
Alex Elder | ca8b3a6 | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 1155 | if (!cursor->resid && cursor->total_resid) { |
| 1156 | WARN_ON(!cursor->last_piece); |
| 1157 | BUG_ON(list_is_last(&cursor->data->links, cursor->data_head)); |
| 1158 | cursor->data = list_entry_next(cursor->data, links); |
| 1159 | __ceph_msg_data_cursor_init(cursor); |
Alex Elder | a51b272e | 2013-04-19 15:34:49 -0500 | [diff] [blame] | 1160 | new_piece = true; |
Alex Elder | ca8b3a6 | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 1161 | } |
Alex Elder | a51b272e | 2013-04-19 15:34:49 -0500 | [diff] [blame] | 1162 | cursor->need_crc = new_piece; |
Alex Elder | ca8b3a6 | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 1163 | |
Alex Elder | dd236fc | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1164 | return new_piece; |
| 1165 | } |
| 1166 | |
Alex Elder | 98fa5dd | 2013-04-02 12:09:50 -0500 | [diff] [blame] | 1167 | static void prepare_message_data(struct ceph_msg *msg, u32 data_len) |
Alex Elder | 739c905 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1168 | { |
Alex Elder | 739c905 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1169 | BUG_ON(!msg); |
Alex Elder | 25aff7c | 2013-03-11 23:34:22 -0500 | [diff] [blame] | 1170 | BUG_ON(!data_len); |
Alex Elder | 739c905 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1171 | |
Alex Elder | 4c59b4a | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 1172 | /* Initialize data cursor */ |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1173 | |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1174 | ceph_msg_data_cursor_init(msg, (size_t)data_len); |
Alex Elder | 739c905 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1175 | } |
| 1176 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1177 | /* |
| 1178 | * Prepare footer for currently outgoing message, and finish things |
| 1179 | * off. Assumes out_kvec* are already valid.. we just add on to the end. |
| 1180 | */ |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1181 | static void prepare_write_message_footer(struct ceph_connection *con) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1182 | { |
| 1183 | struct ceph_msg *m = con->out_msg; |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1184 | int v = con->out_kvec_left; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1185 | |
Alex Elder | fd154f3 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1186 | m->footer.flags |= CEPH_MSG_FOOTER_COMPLETE; |
| 1187 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1188 | dout("prepare_write_message_footer %p\n", con); |
| 1189 | con->out_kvec_is_msg = true; |
| 1190 | con->out_kvec[v].iov_base = &m->footer; |
| 1191 | con->out_kvec[v].iov_len = sizeof(m->footer); |
| 1192 | con->out_kvec_bytes += sizeof(m->footer); |
| 1193 | con->out_kvec_left++; |
| 1194 | con->out_more = m->more_to_follow; |
Sage Weil | c86a293 | 2009-12-14 14:04:30 -0800 | [diff] [blame] | 1195 | con->out_msg_done = true; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1196 | } |
| 1197 | |
| 1198 | /* |
| 1199 | * Prepare headers for the next outgoing message. |
| 1200 | */ |
| 1201 | static void prepare_write_message(struct ceph_connection *con) |
| 1202 | { |
| 1203 | struct ceph_msg *m; |
Alex Elder | a9a0c51 | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 1204 | u32 crc; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1205 | |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1206 | con_out_kvec_reset(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1207 | con->out_kvec_is_msg = true; |
Sage Weil | c86a293 | 2009-12-14 14:04:30 -0800 | [diff] [blame] | 1208 | con->out_msg_done = false; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1209 | |
| 1210 | /* Sneak an ack in there first? If we can get it into the same |
| 1211 | * TCP packet that's a good thing. */ |
| 1212 | if (con->in_seq > con->in_seq_acked) { |
| 1213 | con->in_seq_acked = con->in_seq; |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1214 | con_out_kvec_add(con, sizeof (tag_ack), &tag_ack); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1215 | con->out_temp_ack = cpu_to_le64(con->in_seq_acked); |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1216 | con_out_kvec_add(con, sizeof (con->out_temp_ack), |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1217 | &con->out_temp_ack); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1218 | } |
| 1219 | |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 1220 | BUG_ON(list_empty(&con->out_queue)); |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1221 | m = list_first_entry(&con->out_queue, struct ceph_msg, list_head); |
Sage Weil | c86a293 | 2009-12-14 14:04:30 -0800 | [diff] [blame] | 1222 | con->out_msg = m; |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 1223 | BUG_ON(m->con != con); |
Sage Weil | 4cf9d54 | 2011-07-26 11:27:24 -0700 | [diff] [blame] | 1224 | |
| 1225 | /* put message on sent list */ |
| 1226 | ceph_msg_get(m); |
| 1227 | list_move_tail(&m->list_head, &con->out_sent); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1228 | |
Sage Weil | e84346b | 2010-05-11 21:20:38 -0700 | [diff] [blame] | 1229 | /* |
| 1230 | * only assign outgoing seq # if we haven't sent this message |
| 1231 | * yet. if it is requeued, resend with it's original seq. |
| 1232 | */ |
| 1233 | if (m->needs_out_seq) { |
| 1234 | m->hdr.seq = cpu_to_le64(++con->out_seq); |
| 1235 | m->needs_out_seq = false; |
| 1236 | } |
Alex Elder | 98fa5dd | 2013-04-02 12:09:50 -0500 | [diff] [blame] | 1237 | WARN_ON(m->data_length != le32_to_cpu(m->hdr.data_len)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1238 | |
Alex Elder | 98fa5dd | 2013-04-02 12:09:50 -0500 | [diff] [blame] | 1239 | dout("prepare_write_message %p seq %lld type %d len %d+%d+%zd\n", |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1240 | m, con->out_seq, le16_to_cpu(m->hdr.type), |
| 1241 | le32_to_cpu(m->hdr.front_len), le32_to_cpu(m->hdr.middle_len), |
Alex Elder | 98fa5dd | 2013-04-02 12:09:50 -0500 | [diff] [blame] | 1242 | m->data_length); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1243 | BUG_ON(le32_to_cpu(m->hdr.front_len) != m->front.iov_len); |
| 1244 | |
| 1245 | /* tag + hdr + front + middle */ |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1246 | con_out_kvec_add(con, sizeof (tag_msg), &tag_msg); |
| 1247 | con_out_kvec_add(con, sizeof (m->hdr), &m->hdr); |
| 1248 | 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] | 1249 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1250 | if (m->middle) |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1251 | con_out_kvec_add(con, m->middle->vec.iov_len, |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1252 | m->middle->vec.iov_base); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1253 | |
| 1254 | /* fill in crc (except data pages), footer */ |
Alex Elder | a9a0c51 | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 1255 | crc = crc32c(0, &m->hdr, offsetof(struct ceph_msg_header, crc)); |
| 1256 | con->out_msg->hdr.crc = cpu_to_le32(crc); |
Alex Elder | fd154f3 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1257 | con->out_msg->footer.flags = 0; |
Alex Elder | a9a0c51 | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 1258 | |
| 1259 | crc = crc32c(0, m->front.iov_base, m->front.iov_len); |
| 1260 | con->out_msg->footer.front_crc = cpu_to_le32(crc); |
| 1261 | if (m->middle) { |
| 1262 | crc = crc32c(0, m->middle->vec.iov_base, |
| 1263 | m->middle->vec.iov_len); |
| 1264 | con->out_msg->footer.middle_crc = cpu_to_le32(crc); |
| 1265 | } else |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1266 | con->out_msg->footer.middle_crc = 0; |
Alex Elder | 739c905 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1267 | dout("%s front_crc %u middle_crc %u\n", __func__, |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1268 | le32_to_cpu(con->out_msg->footer.front_crc), |
| 1269 | le32_to_cpu(con->out_msg->footer.middle_crc)); |
| 1270 | |
| 1271 | /* is there a data payload? */ |
Alex Elder | 739c905 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1272 | con->out_msg->footer.data_crc = 0; |
Alex Elder | 98fa5dd | 2013-04-02 12:09:50 -0500 | [diff] [blame] | 1273 | if (m->data_length) { |
| 1274 | prepare_message_data(con->out_msg, m->data_length); |
Alex Elder | 7862505 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1275 | con->out_more = 1; /* data + footer will follow */ |
| 1276 | } else { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1277 | /* no, queue up footer too and be done */ |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1278 | prepare_write_message_footer(con); |
Alex Elder | 7862505 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1279 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1280 | |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 1281 | con_flag_set(con, CON_FLAG_WRITE_PENDING); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1282 | } |
| 1283 | |
| 1284 | /* |
| 1285 | * Prepare an ack. |
| 1286 | */ |
| 1287 | static void prepare_write_ack(struct ceph_connection *con) |
| 1288 | { |
| 1289 | dout("prepare_write_ack %p %llu -> %llu\n", con, |
| 1290 | con->in_seq_acked, con->in_seq); |
| 1291 | con->in_seq_acked = con->in_seq; |
| 1292 | |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1293 | con_out_kvec_reset(con); |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1294 | |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1295 | con_out_kvec_add(con, sizeof (tag_ack), &tag_ack); |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1296 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1297 | con->out_temp_ack = cpu_to_le64(con->in_seq_acked); |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1298 | con_out_kvec_add(con, sizeof (con->out_temp_ack), |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1299 | &con->out_temp_ack); |
| 1300 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1301 | con->out_more = 1; /* more will follow.. eventually.. */ |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 1302 | con_flag_set(con, CON_FLAG_WRITE_PENDING); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1303 | } |
| 1304 | |
| 1305 | /* |
Sage Weil | 3a23083 | 2013-03-25 08:47:40 -0700 | [diff] [blame] | 1306 | * Prepare to share the seq during handshake |
| 1307 | */ |
| 1308 | static void prepare_write_seq(struct ceph_connection *con) |
| 1309 | { |
| 1310 | dout("prepare_write_seq %p %llu -> %llu\n", con, |
| 1311 | con->in_seq_acked, con->in_seq); |
| 1312 | con->in_seq_acked = con->in_seq; |
| 1313 | |
| 1314 | con_out_kvec_reset(con); |
| 1315 | |
| 1316 | con->out_temp_ack = cpu_to_le64(con->in_seq_acked); |
| 1317 | con_out_kvec_add(con, sizeof (con->out_temp_ack), |
| 1318 | &con->out_temp_ack); |
| 1319 | |
| 1320 | con_flag_set(con, CON_FLAG_WRITE_PENDING); |
| 1321 | } |
| 1322 | |
| 1323 | /* |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1324 | * Prepare to write keepalive byte. |
| 1325 | */ |
| 1326 | static void prepare_write_keepalive(struct ceph_connection *con) |
| 1327 | { |
| 1328 | dout("prepare_write_keepalive %p\n", con); |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1329 | con_out_kvec_reset(con); |
| 1330 | con_out_kvec_add(con, sizeof (tag_keepalive), &tag_keepalive); |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 1331 | con_flag_set(con, CON_FLAG_WRITE_PENDING); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1332 | } |
| 1333 | |
| 1334 | /* |
| 1335 | * Connection negotiation. |
| 1336 | */ |
| 1337 | |
Alex Elder | dac1e71 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1338 | static struct ceph_auth_handshake *get_connect_authorizer(struct ceph_connection *con, |
| 1339 | int *auth_proto) |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1340 | { |
Alex Elder | a3530df | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1341 | struct ceph_auth_handshake *auth; |
Alex Elder | b1c6b98 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 1342 | |
| 1343 | if (!con->ops->get_authorizer) { |
| 1344 | con->out_connect.authorizer_protocol = CEPH_AUTH_UNKNOWN; |
| 1345 | con->out_connect.authorizer_len = 0; |
Alex Elder | 729796b | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1346 | return NULL; |
Alex Elder | b1c6b98 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 1347 | } |
| 1348 | |
| 1349 | /* Can't hold the mutex while getting authorizer */ |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 1350 | mutex_unlock(&con->mutex); |
Alex Elder | dac1e71 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1351 | auth = con->ops->get_authorizer(con, auth_proto, con->auth_retry); |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 1352 | mutex_lock(&con->mutex); |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1353 | |
Alex Elder | a3530df | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1354 | if (IS_ERR(auth)) |
Alex Elder | 729796b | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1355 | return auth; |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 1356 | if (con->state != CON_STATE_NEGOTIATING) |
Alex Elder | 729796b | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1357 | return ERR_PTR(-EAGAIN); |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 1358 | |
Alex Elder | 8f43fb5 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1359 | con->auth_reply_buf = auth->authorizer_reply_buf; |
| 1360 | con->auth_reply_buf_len = auth->authorizer_reply_buf_len; |
Alex Elder | 729796b | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1361 | return auth; |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1362 | } |
| 1363 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1364 | /* |
| 1365 | * We connected to a peer and are saying hello. |
| 1366 | */ |
Alex Elder | e825a66 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 1367 | static void prepare_write_banner(struct ceph_connection *con) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1368 | { |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1369 | con_out_kvec_add(con, strlen(CEPH_BANNER), CEPH_BANNER); |
| 1370 | con_out_kvec_add(con, sizeof (con->msgr->my_enc_addr), |
Alex Elder | e825a66 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 1371 | &con->msgr->my_enc_addr); |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1372 | |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1373 | con->out_more = 0; |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 1374 | con_flag_set(con, CON_FLAG_WRITE_PENDING); |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1375 | } |
| 1376 | |
Alex Elder | e825a66 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 1377 | static int prepare_write_connect(struct ceph_connection *con) |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1378 | { |
Eric Dumazet | 95c9617 | 2012-04-15 05:58:06 +0000 | [diff] [blame] | 1379 | unsigned int global_seq = get_global_seq(con->msgr, 0); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1380 | int proto; |
Alex Elder | dac1e71 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1381 | int auth_proto; |
Alex Elder | 729796b | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1382 | struct ceph_auth_handshake *auth; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1383 | |
| 1384 | switch (con->peer_name.type) { |
| 1385 | case CEPH_ENTITY_TYPE_MON: |
| 1386 | proto = CEPH_MONC_PROTOCOL; |
| 1387 | break; |
| 1388 | case CEPH_ENTITY_TYPE_OSD: |
| 1389 | proto = CEPH_OSDC_PROTOCOL; |
| 1390 | break; |
| 1391 | case CEPH_ENTITY_TYPE_MDS: |
| 1392 | proto = CEPH_MDSC_PROTOCOL; |
| 1393 | break; |
| 1394 | default: |
| 1395 | BUG(); |
| 1396 | } |
| 1397 | |
| 1398 | dout("prepare_write_connect %p cseq=%d gseq=%d proto=%d\n", con, |
| 1399 | con->connect_seq, global_seq, proto); |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1400 | |
Alex Elder | e825a66 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 1401 | con->out_connect.features = cpu_to_le64(con->msgr->supported_features); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1402 | con->out_connect.host_type = cpu_to_le32(CEPH_ENTITY_TYPE_CLIENT); |
| 1403 | con->out_connect.connect_seq = cpu_to_le32(con->connect_seq); |
| 1404 | con->out_connect.global_seq = cpu_to_le32(global_seq); |
| 1405 | con->out_connect.protocol_version = cpu_to_le32(proto); |
| 1406 | con->out_connect.flags = 0; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1407 | |
Alex Elder | dac1e71 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1408 | auth_proto = CEPH_AUTH_UNKNOWN; |
| 1409 | auth = get_connect_authorizer(con, &auth_proto); |
Alex Elder | 729796b | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1410 | if (IS_ERR(auth)) |
| 1411 | return PTR_ERR(auth); |
Alex Elder | 3da5477 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1412 | |
Alex Elder | dac1e71 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1413 | con->out_connect.authorizer_protocol = cpu_to_le32(auth_proto); |
Alex Elder | 3da5477 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1414 | con->out_connect.authorizer_len = auth ? |
| 1415 | cpu_to_le32(auth->authorizer_buf_len) : 0; |
| 1416 | |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1417 | con_out_kvec_add(con, sizeof (con->out_connect), |
Alex Elder | 3da5477 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1418 | &con->out_connect); |
| 1419 | if (auth && auth->authorizer_buf_len) |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1420 | con_out_kvec_add(con, auth->authorizer_buf_len, |
Alex Elder | 3da5477 | 2012-05-16 15:16:39 -0500 | [diff] [blame] | 1421 | auth->authorizer_buf); |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1422 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1423 | con->out_more = 0; |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 1424 | con_flag_set(con, CON_FLAG_WRITE_PENDING); |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1425 | |
Alex Elder | e10c758 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 1426 | return 0; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1427 | } |
| 1428 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1429 | /* |
| 1430 | * write as much of pending kvecs to the socket as we can. |
| 1431 | * 1 -> done |
| 1432 | * 0 -> socket full, but more to do |
| 1433 | * <0 -> error |
| 1434 | */ |
| 1435 | static int write_partial_kvec(struct ceph_connection *con) |
| 1436 | { |
| 1437 | int ret; |
| 1438 | |
| 1439 | dout("write_partial_kvec %p %d left\n", con, con->out_kvec_bytes); |
| 1440 | while (con->out_kvec_bytes > 0) { |
| 1441 | ret = ceph_tcp_sendmsg(con->sock, con->out_kvec_cur, |
| 1442 | con->out_kvec_left, con->out_kvec_bytes, |
| 1443 | con->out_more); |
| 1444 | if (ret <= 0) |
| 1445 | goto out; |
| 1446 | con->out_kvec_bytes -= ret; |
| 1447 | if (con->out_kvec_bytes == 0) |
| 1448 | break; /* done */ |
Alex Elder | f42299e | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 1449 | |
| 1450 | /* account for full iov entries consumed */ |
| 1451 | while (ret >= con->out_kvec_cur->iov_len) { |
| 1452 | BUG_ON(!con->out_kvec_left); |
| 1453 | ret -= con->out_kvec_cur->iov_len; |
| 1454 | con->out_kvec_cur++; |
| 1455 | con->out_kvec_left--; |
| 1456 | } |
| 1457 | /* and for a partially-consumed entry */ |
| 1458 | if (ret) { |
| 1459 | con->out_kvec_cur->iov_len -= ret; |
| 1460 | con->out_kvec_cur->iov_base += ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1461 | } |
| 1462 | } |
| 1463 | con->out_kvec_left = 0; |
| 1464 | con->out_kvec_is_msg = false; |
| 1465 | ret = 1; |
| 1466 | out: |
| 1467 | dout("write_partial_kvec %p %d left in %d kvecs ret = %d\n", con, |
| 1468 | con->out_kvec_bytes, con->out_kvec_left, ret); |
| 1469 | return ret; /* done! */ |
| 1470 | } |
| 1471 | |
Alex Elder | 35b6280 | 2013-03-08 20:59:00 -0600 | [diff] [blame] | 1472 | static u32 ceph_crc32c_page(u32 crc, struct page *page, |
| 1473 | unsigned int page_offset, |
| 1474 | unsigned int length) |
| 1475 | { |
| 1476 | char *kaddr; |
| 1477 | |
| 1478 | kaddr = kmap(page); |
| 1479 | BUG_ON(kaddr == NULL); |
| 1480 | crc = crc32c(crc, kaddr + page_offset, length); |
| 1481 | kunmap(page); |
| 1482 | |
| 1483 | return crc; |
| 1484 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1485 | /* |
| 1486 | * Write as much message data payload as we can. If we finish, queue |
| 1487 | * up the footer. |
| 1488 | * 1 -> done, footer is now queued in out_kvec[]. |
| 1489 | * 0 -> socket full, but more to do |
| 1490 | * <0 -> error |
| 1491 | */ |
Alex Elder | 34d2d20 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 1492 | static int write_partial_message_data(struct ceph_connection *con) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1493 | { |
| 1494 | struct ceph_msg *msg = con->out_msg; |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1495 | struct ceph_msg_data_cursor *cursor = &msg->cursor; |
Alex Elder | 37675b0 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 1496 | bool do_datacrc = !con->msgr->nocrc; |
Alex Elder | f5db90b | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 1497 | u32 crc; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1498 | |
Alex Elder | 859a35d | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 1499 | dout("%s %p msg %p\n", __func__, con, msg); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1500 | |
Alex Elder | 5240d9f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1501 | if (list_empty(&msg->data)) |
Alex Elder | 4c59b4a | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 1502 | return -EINVAL; |
| 1503 | |
Alex Elder | 5821bd8 | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1504 | /* |
| 1505 | * Iterate through each page that contains data to be |
| 1506 | * written, and send as much as possible for each. |
| 1507 | * |
| 1508 | * If we are calculating the data crc (the default), we will |
| 1509 | * need to map the page. If we have no pages, they have |
| 1510 | * been revoked, so use the zero page. |
| 1511 | */ |
Alex Elder | f5db90b | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 1512 | crc = do_datacrc ? le32_to_cpu(msg->footer.data_crc) : 0; |
Alex Elder | 643c68a | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 1513 | while (cursor->resid) { |
Alex Elder | 8a166d0 | 2013-03-08 13:35:36 -0600 | [diff] [blame] | 1514 | struct page *page; |
Alex Elder | e387d52 | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1515 | size_t page_offset; |
| 1516 | size_t length; |
Alex Elder | 8a166d0 | 2013-03-08 13:35:36 -0600 | [diff] [blame] | 1517 | bool last_piece; |
Alex Elder | 8ea299b | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 1518 | bool need_crc; |
Alex Elder | f5db90b | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 1519 | int ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1520 | |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1521 | page = ceph_msg_data_next(&msg->cursor, &page_offset, &length, |
Alex Elder | 4c59b4a | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 1522 | &last_piece); |
Alex Elder | e387d52 | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1523 | ret = ceph_tcp_sendpage(con->sock, page, page_offset, |
Alex Elder | fe38a2b | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 1524 | length, last_piece); |
Alex Elder | f5db90b | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 1525 | if (ret <= 0) { |
| 1526 | if (do_datacrc) |
| 1527 | msg->footer.data_crc = cpu_to_le32(crc); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1528 | |
Alex Elder | f5db90b | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 1529 | return ret; |
| 1530 | } |
Alex Elder | 143334f | 2013-03-29 11:44:10 -0500 | [diff] [blame] | 1531 | if (do_datacrc && cursor->need_crc) |
| 1532 | crc = ceph_crc32c_page(crc, page, page_offset, length); |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 1533 | need_crc = ceph_msg_data_advance(&msg->cursor, (size_t)ret); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1534 | } |
| 1535 | |
Alex Elder | 34d2d20 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 1536 | dout("%s %p msg %p done\n", __func__, con, msg); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1537 | |
| 1538 | /* prepare and queue up footer, too */ |
Alex Elder | f5db90b | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 1539 | if (do_datacrc) |
| 1540 | msg->footer.data_crc = cpu_to_le32(crc); |
| 1541 | else |
Alex Elder | 84ca8fc | 2012-06-11 14:57:13 -0500 | [diff] [blame] | 1542 | msg->footer.flags |= CEPH_MSG_FOOTER_NOCRC; |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 1543 | con_out_kvec_reset(con); |
Alex Elder | 859eb79 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 1544 | prepare_write_message_footer(con); |
Alex Elder | f5db90b | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 1545 | |
| 1546 | return 1; /* must return > 0 to indicate success */ |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1547 | } |
| 1548 | |
| 1549 | /* |
| 1550 | * write some zeros |
| 1551 | */ |
| 1552 | static int write_partial_skip(struct ceph_connection *con) |
| 1553 | { |
| 1554 | int ret; |
| 1555 | |
| 1556 | while (con->out_skip > 0) { |
Alex Elder | 3173913 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 1557 | size_t size = min(con->out_skip, (int) PAGE_CACHE_SIZE); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1558 | |
Alex Elder | e1dcb12 | 2013-03-06 23:39:38 -0600 | [diff] [blame] | 1559 | ret = ceph_tcp_sendpage(con->sock, zero_page, 0, size, true); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1560 | if (ret <= 0) |
| 1561 | goto out; |
| 1562 | con->out_skip -= ret; |
| 1563 | } |
| 1564 | ret = 1; |
| 1565 | out: |
| 1566 | return ret; |
| 1567 | } |
| 1568 | |
| 1569 | /* |
| 1570 | * Prepare to read connection handshake, or an ack. |
| 1571 | */ |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1572 | static void prepare_read_banner(struct ceph_connection *con) |
| 1573 | { |
| 1574 | dout("prepare_read_banner %p\n", con); |
| 1575 | con->in_base_pos = 0; |
| 1576 | } |
| 1577 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1578 | static void prepare_read_connect(struct ceph_connection *con) |
| 1579 | { |
| 1580 | dout("prepare_read_connect %p\n", con); |
| 1581 | con->in_base_pos = 0; |
| 1582 | } |
| 1583 | |
| 1584 | static void prepare_read_ack(struct ceph_connection *con) |
| 1585 | { |
| 1586 | dout("prepare_read_ack %p\n", con); |
| 1587 | con->in_base_pos = 0; |
| 1588 | } |
| 1589 | |
Sage Weil | 3a23083 | 2013-03-25 08:47:40 -0700 | [diff] [blame] | 1590 | static void prepare_read_seq(struct ceph_connection *con) |
| 1591 | { |
| 1592 | dout("prepare_read_seq %p\n", con); |
| 1593 | con->in_base_pos = 0; |
| 1594 | con->in_tag = CEPH_MSGR_TAG_SEQ; |
| 1595 | } |
| 1596 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1597 | static void prepare_read_tag(struct ceph_connection *con) |
| 1598 | { |
| 1599 | dout("prepare_read_tag %p\n", con); |
| 1600 | con->in_base_pos = 0; |
| 1601 | con->in_tag = CEPH_MSGR_TAG_READY; |
| 1602 | } |
| 1603 | |
| 1604 | /* |
| 1605 | * Prepare to read a message. |
| 1606 | */ |
| 1607 | static int prepare_read_message(struct ceph_connection *con) |
| 1608 | { |
| 1609 | dout("prepare_read_message %p\n", con); |
| 1610 | BUG_ON(con->in_msg != NULL); |
| 1611 | con->in_base_pos = 0; |
| 1612 | con->in_front_crc = con->in_middle_crc = con->in_data_crc = 0; |
| 1613 | return 0; |
| 1614 | } |
| 1615 | |
| 1616 | |
| 1617 | static int read_partial(struct ceph_connection *con, |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1618 | int end, int size, void *object) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1619 | { |
Alex Elder | e6cee71 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1620 | while (con->in_base_pos < end) { |
| 1621 | int left = end - con->in_base_pos; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1622 | int have = size - left; |
| 1623 | int ret = ceph_tcp_recvmsg(con->sock, object + have, left); |
| 1624 | if (ret <= 0) |
| 1625 | return ret; |
| 1626 | con->in_base_pos += ret; |
| 1627 | } |
| 1628 | return 1; |
| 1629 | } |
| 1630 | |
| 1631 | |
| 1632 | /* |
| 1633 | * Read all or part of the connect-side handshake on a new connection |
| 1634 | */ |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1635 | static int read_partial_banner(struct ceph_connection *con) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1636 | { |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1637 | int size; |
| 1638 | int end; |
| 1639 | int ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1640 | |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1641 | 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] | 1642 | |
| 1643 | /* peer's banner */ |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1644 | size = strlen(CEPH_BANNER); |
| 1645 | end = size; |
| 1646 | ret = read_partial(con, end, size, con->in_banner); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1647 | if (ret <= 0) |
| 1648 | goto out; |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1649 | |
| 1650 | size = sizeof (con->actual_peer_addr); |
| 1651 | end += size; |
| 1652 | ret = read_partial(con, end, size, &con->actual_peer_addr); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1653 | if (ret <= 0) |
| 1654 | goto out; |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1655 | |
| 1656 | size = sizeof (con->peer_addr_for_me); |
| 1657 | end += size; |
| 1658 | ret = read_partial(con, end, size, &con->peer_addr_for_me); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1659 | if (ret <= 0) |
| 1660 | goto out; |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1661 | |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1662 | out: |
| 1663 | return ret; |
| 1664 | } |
| 1665 | |
| 1666 | static int read_partial_connect(struct ceph_connection *con) |
| 1667 | { |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1668 | int size; |
| 1669 | int end; |
| 1670 | int ret; |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1671 | |
| 1672 | dout("read_partial_connect %p at %d\n", con, con->in_base_pos); |
| 1673 | |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1674 | size = sizeof (con->in_reply); |
| 1675 | end = size; |
| 1676 | ret = read_partial(con, end, size, &con->in_reply); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1677 | if (ret <= 0) |
| 1678 | goto out; |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 1679 | |
| 1680 | size = le32_to_cpu(con->in_reply.authorizer_len); |
| 1681 | end += size; |
| 1682 | ret = read_partial(con, end, size, con->auth_reply_buf); |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1683 | if (ret <= 0) |
| 1684 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1685 | |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1686 | dout("read_partial_connect %p tag %d, con_seq = %u, g_seq = %u\n", |
| 1687 | con, (int)con->in_reply.tag, |
| 1688 | le32_to_cpu(con->in_reply.connect_seq), |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1689 | le32_to_cpu(con->in_reply.global_seq)); |
| 1690 | out: |
| 1691 | return ret; |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1692 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1693 | } |
| 1694 | |
| 1695 | /* |
| 1696 | * Verify the hello banner looks okay. |
| 1697 | */ |
| 1698 | static int verify_hello(struct ceph_connection *con) |
| 1699 | { |
| 1700 | if (memcmp(con->in_banner, CEPH_BANNER, strlen(CEPH_BANNER))) { |
Sage Weil | 13e38c8 | 2009-10-09 16:36:34 -0700 | [diff] [blame] | 1701 | pr_err("connect to %s got bad banner\n", |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1702 | ceph_pr_addr(&con->peer_addr.in_addr)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1703 | con->error_msg = "protocol error, bad banner"; |
| 1704 | return -1; |
| 1705 | } |
| 1706 | return 0; |
| 1707 | } |
| 1708 | |
| 1709 | static bool addr_is_blank(struct sockaddr_storage *ss) |
| 1710 | { |
| 1711 | switch (ss->ss_family) { |
| 1712 | case AF_INET: |
| 1713 | return ((struct sockaddr_in *)ss)->sin_addr.s_addr == 0; |
| 1714 | case AF_INET6: |
| 1715 | return |
| 1716 | ((struct sockaddr_in6 *)ss)->sin6_addr.s6_addr32[0] == 0 && |
| 1717 | ((struct sockaddr_in6 *)ss)->sin6_addr.s6_addr32[1] == 0 && |
| 1718 | ((struct sockaddr_in6 *)ss)->sin6_addr.s6_addr32[2] == 0 && |
| 1719 | ((struct sockaddr_in6 *)ss)->sin6_addr.s6_addr32[3] == 0; |
| 1720 | } |
| 1721 | return false; |
| 1722 | } |
| 1723 | |
| 1724 | static int addr_port(struct sockaddr_storage *ss) |
| 1725 | { |
| 1726 | switch (ss->ss_family) { |
| 1727 | case AF_INET: |
Sage Weil | f28bcfb | 2009-11-04 11:46:35 -0800 | [diff] [blame] | 1728 | return ntohs(((struct sockaddr_in *)ss)->sin_port); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1729 | case AF_INET6: |
Sage Weil | f28bcfb | 2009-11-04 11:46:35 -0800 | [diff] [blame] | 1730 | return ntohs(((struct sockaddr_in6 *)ss)->sin6_port); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1731 | } |
| 1732 | return 0; |
| 1733 | } |
| 1734 | |
| 1735 | static void addr_set_port(struct sockaddr_storage *ss, int p) |
| 1736 | { |
| 1737 | switch (ss->ss_family) { |
| 1738 | case AF_INET: |
| 1739 | ((struct sockaddr_in *)ss)->sin_port = htons(p); |
Sage Weil | a2a7960 | 2011-05-12 15:34:24 -0700 | [diff] [blame] | 1740 | break; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1741 | case AF_INET6: |
| 1742 | ((struct sockaddr_in6 *)ss)->sin6_port = htons(p); |
Sage Weil | a2a7960 | 2011-05-12 15:34:24 -0700 | [diff] [blame] | 1743 | break; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1744 | } |
| 1745 | } |
| 1746 | |
| 1747 | /* |
Noah Watkins | ee3b56f | 2011-09-23 11:48:42 -0700 | [diff] [blame] | 1748 | * Unlike other *_pton function semantics, zero indicates success. |
| 1749 | */ |
| 1750 | static int ceph_pton(const char *str, size_t len, struct sockaddr_storage *ss, |
| 1751 | char delim, const char **ipend) |
| 1752 | { |
Alex Elder | 99f0f3b | 2012-01-23 15:49:27 -0600 | [diff] [blame] | 1753 | struct sockaddr_in *in4 = (struct sockaddr_in *) ss; |
| 1754 | struct sockaddr_in6 *in6 = (struct sockaddr_in6 *) ss; |
Noah Watkins | ee3b56f | 2011-09-23 11:48:42 -0700 | [diff] [blame] | 1755 | |
| 1756 | memset(ss, 0, sizeof(*ss)); |
| 1757 | |
| 1758 | if (in4_pton(str, len, (u8 *)&in4->sin_addr.s_addr, delim, ipend)) { |
| 1759 | ss->ss_family = AF_INET; |
| 1760 | return 0; |
| 1761 | } |
| 1762 | |
| 1763 | if (in6_pton(str, len, (u8 *)&in6->sin6_addr.s6_addr, delim, ipend)) { |
| 1764 | ss->ss_family = AF_INET6; |
| 1765 | return 0; |
| 1766 | } |
| 1767 | |
| 1768 | return -EINVAL; |
| 1769 | } |
| 1770 | |
| 1771 | /* |
| 1772 | * Extract hostname string and resolve using kernel DNS facility. |
| 1773 | */ |
| 1774 | #ifdef CONFIG_CEPH_LIB_USE_DNS_RESOLVER |
| 1775 | static int ceph_dns_resolve_name(const char *name, size_t namelen, |
| 1776 | struct sockaddr_storage *ss, char delim, const char **ipend) |
| 1777 | { |
| 1778 | const char *end, *delim_p; |
| 1779 | char *colon_p, *ip_addr = NULL; |
| 1780 | int ip_len, ret; |
| 1781 | |
| 1782 | /* |
| 1783 | * The end of the hostname occurs immediately preceding the delimiter or |
| 1784 | * the port marker (':') where the delimiter takes precedence. |
| 1785 | */ |
| 1786 | delim_p = memchr(name, delim, namelen); |
| 1787 | colon_p = memchr(name, ':', namelen); |
| 1788 | |
| 1789 | if (delim_p && colon_p) |
| 1790 | end = delim_p < colon_p ? delim_p : colon_p; |
| 1791 | else if (!delim_p && colon_p) |
| 1792 | end = colon_p; |
| 1793 | else { |
| 1794 | end = delim_p; |
| 1795 | if (!end) /* case: hostname:/ */ |
| 1796 | end = name + namelen; |
| 1797 | } |
| 1798 | |
| 1799 | if (end <= name) |
| 1800 | return -EINVAL; |
| 1801 | |
| 1802 | /* do dns_resolve upcall */ |
| 1803 | ip_len = dns_query(NULL, name, end - name, NULL, &ip_addr, NULL); |
| 1804 | if (ip_len > 0) |
| 1805 | ret = ceph_pton(ip_addr, ip_len, ss, -1, NULL); |
| 1806 | else |
| 1807 | ret = -ESRCH; |
| 1808 | |
| 1809 | kfree(ip_addr); |
| 1810 | |
| 1811 | *ipend = end; |
| 1812 | |
| 1813 | pr_info("resolve '%.*s' (ret=%d): %s\n", (int)(end - name), name, |
| 1814 | ret, ret ? "failed" : ceph_pr_addr(ss)); |
| 1815 | |
| 1816 | return ret; |
| 1817 | } |
| 1818 | #else |
| 1819 | static inline int ceph_dns_resolve_name(const char *name, size_t namelen, |
| 1820 | struct sockaddr_storage *ss, char delim, const char **ipend) |
| 1821 | { |
| 1822 | return -EINVAL; |
| 1823 | } |
| 1824 | #endif |
| 1825 | |
| 1826 | /* |
| 1827 | * Parse a server name (IP or hostname). If a valid IP address is not found |
| 1828 | * then try to extract a hostname to resolve using userspace DNS upcall. |
| 1829 | */ |
| 1830 | static int ceph_parse_server_name(const char *name, size_t namelen, |
| 1831 | struct sockaddr_storage *ss, char delim, const char **ipend) |
| 1832 | { |
| 1833 | int ret; |
| 1834 | |
| 1835 | ret = ceph_pton(name, namelen, ss, delim, ipend); |
| 1836 | if (ret) |
| 1837 | ret = ceph_dns_resolve_name(name, namelen, ss, delim, ipend); |
| 1838 | |
| 1839 | return ret; |
| 1840 | } |
| 1841 | |
| 1842 | /* |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1843 | * Parse an ip[:port] list into an addr array. Use the default |
| 1844 | * monitor port if a port isn't specified. |
| 1845 | */ |
| 1846 | int ceph_parse_ips(const char *c, const char *end, |
| 1847 | struct ceph_entity_addr *addr, |
| 1848 | int max_count, int *count) |
| 1849 | { |
Noah Watkins | ee3b56f | 2011-09-23 11:48:42 -0700 | [diff] [blame] | 1850 | int i, ret = -EINVAL; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1851 | const char *p = c; |
| 1852 | |
| 1853 | dout("parse_ips on '%.*s'\n", (int)(end-c), c); |
| 1854 | for (i = 0; i < max_count; i++) { |
| 1855 | const char *ipend; |
| 1856 | struct sockaddr_storage *ss = &addr[i].in_addr; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1857 | int port; |
Sage Weil | 39139f6 | 2010-07-08 09:54:52 -0700 | [diff] [blame] | 1858 | char delim = ','; |
| 1859 | |
| 1860 | if (*p == '[') { |
| 1861 | delim = ']'; |
| 1862 | p++; |
| 1863 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1864 | |
Noah Watkins | ee3b56f | 2011-09-23 11:48:42 -0700 | [diff] [blame] | 1865 | ret = ceph_parse_server_name(p, end - p, ss, delim, &ipend); |
| 1866 | if (ret) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1867 | goto bad; |
Noah Watkins | ee3b56f | 2011-09-23 11:48:42 -0700 | [diff] [blame] | 1868 | ret = -EINVAL; |
| 1869 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1870 | p = ipend; |
| 1871 | |
Sage Weil | 39139f6 | 2010-07-08 09:54:52 -0700 | [diff] [blame] | 1872 | if (delim == ']') { |
| 1873 | if (*p != ']') { |
| 1874 | dout("missing matching ']'\n"); |
| 1875 | goto bad; |
| 1876 | } |
| 1877 | p++; |
| 1878 | } |
| 1879 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1880 | /* port? */ |
| 1881 | if (p < end && *p == ':') { |
| 1882 | port = 0; |
| 1883 | p++; |
| 1884 | while (p < end && *p >= '0' && *p <= '9') { |
| 1885 | port = (port * 10) + (*p - '0'); |
| 1886 | p++; |
| 1887 | } |
Ilya Dryomov | f48db1e | 2013-12-30 19:21:29 +0200 | [diff] [blame] | 1888 | if (port == 0) |
| 1889 | port = CEPH_MON_PORT; |
| 1890 | else if (port > 65535) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1891 | goto bad; |
| 1892 | } else { |
| 1893 | port = CEPH_MON_PORT; |
| 1894 | } |
| 1895 | |
| 1896 | addr_set_port(ss, port); |
| 1897 | |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1898 | dout("parse_ips got %s\n", ceph_pr_addr(ss)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1899 | |
| 1900 | if (p == end) |
| 1901 | break; |
| 1902 | if (*p != ',') |
| 1903 | goto bad; |
| 1904 | p++; |
| 1905 | } |
| 1906 | |
| 1907 | if (p != end) |
| 1908 | goto bad; |
| 1909 | |
| 1910 | if (count) |
| 1911 | *count = i + 1; |
| 1912 | return 0; |
| 1913 | |
| 1914 | bad: |
Sage Weil | 39139f6 | 2010-07-08 09:54:52 -0700 | [diff] [blame] | 1915 | pr_err("parse_ips bad ip '%.*s'\n", (int)(end - c), c); |
Noah Watkins | ee3b56f | 2011-09-23 11:48:42 -0700 | [diff] [blame] | 1916 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1917 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1918 | EXPORT_SYMBOL(ceph_parse_ips); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1919 | |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1920 | static int process_banner(struct ceph_connection *con) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1921 | { |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1922 | dout("process_banner on %p\n", con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1923 | |
| 1924 | if (verify_hello(con) < 0) |
| 1925 | return -1; |
| 1926 | |
Sage Weil | 63f2d21 | 2009-11-03 15:17:56 -0800 | [diff] [blame] | 1927 | ceph_decode_addr(&con->actual_peer_addr); |
| 1928 | ceph_decode_addr(&con->peer_addr_for_me); |
| 1929 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1930 | /* |
| 1931 | * Make sure the other end is who we wanted. note that the other |
| 1932 | * end may not yet know their ip address, so if it's 0.0.0.0, give |
| 1933 | * them the benefit of the doubt. |
| 1934 | */ |
Sage Weil | 103e2d3 | 2010-01-07 16:12:36 -0800 | [diff] [blame] | 1935 | if (memcmp(&con->peer_addr, &con->actual_peer_addr, |
| 1936 | sizeof(con->peer_addr)) != 0 && |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1937 | !(addr_is_blank(&con->actual_peer_addr.in_addr) && |
| 1938 | con->actual_peer_addr.nonce == con->peer_addr.nonce)) { |
Yehuda Sadeh | cd84db6 | 2010-06-11 16:58:48 -0700 | [diff] [blame] | 1939 | pr_warning("wrong peer, want %s/%d, got %s/%d\n", |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1940 | ceph_pr_addr(&con->peer_addr.in_addr), |
Yehuda Sadeh | cd84db6 | 2010-06-11 16:58:48 -0700 | [diff] [blame] | 1941 | (int)le32_to_cpu(con->peer_addr.nonce), |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1942 | ceph_pr_addr(&con->actual_peer_addr.in_addr), |
Yehuda Sadeh | cd84db6 | 2010-06-11 16:58:48 -0700 | [diff] [blame] | 1943 | (int)le32_to_cpu(con->actual_peer_addr.nonce)); |
Sage Weil | 58bb3b3 | 2009-12-23 12:12:31 -0800 | [diff] [blame] | 1944 | con->error_msg = "wrong peer at address"; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1945 | return -1; |
| 1946 | } |
| 1947 | |
| 1948 | /* |
| 1949 | * did we learn our address? |
| 1950 | */ |
| 1951 | if (addr_is_blank(&con->msgr->inst.addr.in_addr)) { |
| 1952 | int port = addr_port(&con->msgr->inst.addr.in_addr); |
| 1953 | |
| 1954 | memcpy(&con->msgr->inst.addr.in_addr, |
| 1955 | &con->peer_addr_for_me.in_addr, |
| 1956 | sizeof(con->peer_addr_for_me.in_addr)); |
| 1957 | addr_set_port(&con->msgr->inst.addr.in_addr, port); |
Sage Weil | 63f2d21 | 2009-11-03 15:17:56 -0800 | [diff] [blame] | 1958 | encode_my_addr(con->msgr); |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1959 | dout("process_banner learned my addr is %s\n", |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1960 | ceph_pr_addr(&con->msgr->inst.addr.in_addr)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1961 | } |
| 1962 | |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1963 | return 0; |
| 1964 | } |
| 1965 | |
| 1966 | static int process_connect(struct ceph_connection *con) |
| 1967 | { |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1968 | u64 sup_feat = con->msgr->supported_features; |
| 1969 | u64 req_feat = con->msgr->required_features; |
Ilya Dryomov | 2b3e0c9 | 2013-12-24 21:19:24 +0200 | [diff] [blame] | 1970 | u64 server_feat = ceph_sanitize_features( |
| 1971 | le64_to_cpu(con->in_reply.features)); |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 1972 | int ret; |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 1973 | |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 1974 | dout("process_connect on %p tag %d\n", con, (int)con->in_tag); |
| 1975 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1976 | switch (con->in_reply.tag) { |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 1977 | case CEPH_MSGR_TAG_FEATURES: |
| 1978 | pr_err("%s%lld %s feature set mismatch," |
| 1979 | " my %llx < server's %llx, missing %llx\n", |
| 1980 | ENTITY_NAME(con->peer_name), |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1981 | ceph_pr_addr(&con->peer_addr.in_addr), |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 1982 | sup_feat, server_feat, server_feat & ~sup_feat); |
| 1983 | con->error_msg = "missing required protocol features"; |
Sage Weil | 0fa6ebc | 2012-12-27 20:27:04 -0600 | [diff] [blame] | 1984 | reset_connection(con); |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 1985 | return -1; |
| 1986 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1987 | case CEPH_MSGR_TAG_BADPROTOVER: |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1988 | pr_err("%s%lld %s protocol version mismatch," |
| 1989 | " my %d != server's %d\n", |
| 1990 | ENTITY_NAME(con->peer_name), |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 1991 | ceph_pr_addr(&con->peer_addr.in_addr), |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1992 | le32_to_cpu(con->out_connect.protocol_version), |
| 1993 | le32_to_cpu(con->in_reply.protocol_version)); |
| 1994 | con->error_msg = "protocol version mismatch"; |
Sage Weil | 0fa6ebc | 2012-12-27 20:27:04 -0600 | [diff] [blame] | 1995 | reset_connection(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 1996 | return -1; |
| 1997 | |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 1998 | case CEPH_MSGR_TAG_BADAUTHORIZER: |
| 1999 | con->auth_retry++; |
| 2000 | dout("process_connect %p got BADAUTHORIZER attempt %d\n", con, |
| 2001 | con->auth_retry); |
| 2002 | if (con->auth_retry == 2) { |
| 2003 | con->error_msg = "connect authorization failure"; |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 2004 | return -1; |
| 2005 | } |
Jim Schutt | 6d4221b | 2012-08-10 10:37:38 -0700 | [diff] [blame] | 2006 | con_out_kvec_reset(con); |
Alex Elder | e825a66 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 2007 | ret = prepare_write_connect(con); |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 2008 | if (ret < 0) |
| 2009 | return ret; |
Sage Weil | 63733a0 | 2010-03-15 15:47:22 -0700 | [diff] [blame] | 2010 | prepare_read_connect(con); |
Sage Weil | 4e7a5dc | 2009-11-18 16:19:57 -0800 | [diff] [blame] | 2011 | break; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2012 | |
| 2013 | case CEPH_MSGR_TAG_RESETSESSION: |
| 2014 | /* |
| 2015 | * If we connected with a large connect_seq but the peer |
| 2016 | * has no record of a session with us (no connection, or |
| 2017 | * connect_seq == 0), they will send RESETSESION to indicate |
| 2018 | * that they must have reset their session, and may have |
| 2019 | * dropped messages. |
| 2020 | */ |
| 2021 | dout("process_connect got RESET peer seq %u\n", |
Sage Weil | 5bdca4e | 2012-07-10 11:53:34 -0700 | [diff] [blame] | 2022 | le32_to_cpu(con->in_reply.connect_seq)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2023 | pr_err("%s%lld %s connection reset\n", |
| 2024 | ENTITY_NAME(con->peer_name), |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2025 | ceph_pr_addr(&con->peer_addr.in_addr)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2026 | reset_connection(con); |
Jim Schutt | 6d4221b | 2012-08-10 10:37:38 -0700 | [diff] [blame] | 2027 | con_out_kvec_reset(con); |
Alex Elder | 5a0f8fd | 2012-05-16 21:51:59 -0500 | [diff] [blame] | 2028 | ret = prepare_write_connect(con); |
| 2029 | if (ret < 0) |
| 2030 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2031 | prepare_read_connect(con); |
| 2032 | |
| 2033 | /* Tell ceph about it. */ |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 2034 | mutex_unlock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2035 | pr_info("reset on %s%lld\n", ENTITY_NAME(con->peer_name)); |
| 2036 | if (con->ops->peer_reset) |
| 2037 | con->ops->peer_reset(con); |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 2038 | mutex_lock(&con->mutex); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2039 | if (con->state != CON_STATE_NEGOTIATING) |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 2040 | return -EAGAIN; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2041 | break; |
| 2042 | |
| 2043 | case CEPH_MSGR_TAG_RETRY_SESSION: |
| 2044 | /* |
| 2045 | * If we sent a smaller connect_seq than the peer has, try |
| 2046 | * again with a larger value. |
| 2047 | */ |
Sage Weil | 5bdca4e | 2012-07-10 11:53:34 -0700 | [diff] [blame] | 2048 | dout("process_connect got RETRY_SESSION my seq %u, peer %u\n", |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2049 | le32_to_cpu(con->out_connect.connect_seq), |
Sage Weil | 5bdca4e | 2012-07-10 11:53:34 -0700 | [diff] [blame] | 2050 | le32_to_cpu(con->in_reply.connect_seq)); |
| 2051 | con->connect_seq = le32_to_cpu(con->in_reply.connect_seq); |
Jim Schutt | 6d4221b | 2012-08-10 10:37:38 -0700 | [diff] [blame] | 2052 | con_out_kvec_reset(con); |
Alex Elder | 5a0f8fd | 2012-05-16 21:51:59 -0500 | [diff] [blame] | 2053 | ret = prepare_write_connect(con); |
| 2054 | if (ret < 0) |
| 2055 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2056 | prepare_read_connect(con); |
| 2057 | break; |
| 2058 | |
| 2059 | case CEPH_MSGR_TAG_RETRY_GLOBAL: |
| 2060 | /* |
| 2061 | * If we sent a smaller global_seq than the peer has, try |
| 2062 | * again with a larger value. |
| 2063 | */ |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 2064 | dout("process_connect got RETRY_GLOBAL my %u peer_gseq %u\n", |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2065 | con->peer_global_seq, |
Sage Weil | 5bdca4e | 2012-07-10 11:53:34 -0700 | [diff] [blame] | 2066 | le32_to_cpu(con->in_reply.global_seq)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2067 | get_global_seq(con->msgr, |
Sage Weil | 5bdca4e | 2012-07-10 11:53:34 -0700 | [diff] [blame] | 2068 | le32_to_cpu(con->in_reply.global_seq)); |
Jim Schutt | 6d4221b | 2012-08-10 10:37:38 -0700 | [diff] [blame] | 2069 | con_out_kvec_reset(con); |
Alex Elder | 5a0f8fd | 2012-05-16 21:51:59 -0500 | [diff] [blame] | 2070 | ret = prepare_write_connect(con); |
| 2071 | if (ret < 0) |
| 2072 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2073 | prepare_read_connect(con); |
| 2074 | break; |
| 2075 | |
Sage Weil | 3a23083 | 2013-03-25 08:47:40 -0700 | [diff] [blame] | 2076 | case CEPH_MSGR_TAG_SEQ: |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2077 | case CEPH_MSGR_TAG_READY: |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 2078 | if (req_feat & ~server_feat) { |
| 2079 | pr_err("%s%lld %s protocol feature mismatch," |
| 2080 | " my required %llx > server's %llx, need %llx\n", |
| 2081 | ENTITY_NAME(con->peer_name), |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2082 | ceph_pr_addr(&con->peer_addr.in_addr), |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 2083 | req_feat, server_feat, req_feat & ~server_feat); |
| 2084 | con->error_msg = "missing required protocol features"; |
Sage Weil | 0fa6ebc | 2012-12-27 20:27:04 -0600 | [diff] [blame] | 2085 | reset_connection(con); |
Sage Weil | 04a419f | 2009-12-23 09:30:21 -0800 | [diff] [blame] | 2086 | return -1; |
| 2087 | } |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2088 | |
Alex Elder | 122070a | 2012-12-26 10:43:57 -0600 | [diff] [blame] | 2089 | WARN_ON(con->state != CON_STATE_NEGOTIATING); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2090 | con->state = CON_STATE_OPEN; |
Sage Weil | 20e55c4 | 2013-03-25 09:30:13 -0700 | [diff] [blame] | 2091 | con->auth_retry = 0; /* we authenticated; clear flag */ |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2092 | con->peer_global_seq = le32_to_cpu(con->in_reply.global_seq); |
| 2093 | con->connect_seq++; |
Sage Weil | aba558e | 2010-05-12 15:23:30 -0700 | [diff] [blame] | 2094 | con->peer_features = server_feat; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2095 | dout("process_connect got READY gseq %d cseq %d (%d)\n", |
| 2096 | con->peer_global_seq, |
| 2097 | le32_to_cpu(con->in_reply.connect_seq), |
| 2098 | con->connect_seq); |
| 2099 | WARN_ON(con->connect_seq != |
| 2100 | le32_to_cpu(con->in_reply.connect_seq)); |
Sage Weil | 92ac41d | 2009-12-14 14:56:56 -0800 | [diff] [blame] | 2101 | |
| 2102 | if (con->in_reply.flags & CEPH_MSG_CONNECT_LOSSY) |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2103 | con_flag_set(con, CON_FLAG_LOSSYTX); |
Sage Weil | 92ac41d | 2009-12-14 14:56:56 -0800 | [diff] [blame] | 2104 | |
Sage Weil | 85effe1 | 2012-07-30 16:22:05 -0700 | [diff] [blame] | 2105 | con->delay = 0; /* reset backoff memory */ |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2106 | |
Sage Weil | 3a23083 | 2013-03-25 08:47:40 -0700 | [diff] [blame] | 2107 | if (con->in_reply.tag == CEPH_MSGR_TAG_SEQ) { |
| 2108 | prepare_write_seq(con); |
| 2109 | prepare_read_seq(con); |
| 2110 | } else { |
| 2111 | prepare_read_tag(con); |
| 2112 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2113 | break; |
| 2114 | |
| 2115 | case CEPH_MSGR_TAG_WAIT: |
| 2116 | /* |
| 2117 | * If there is a connection race (we are opening |
| 2118 | * connections to each other), one of us may just have |
| 2119 | * to WAIT. This shouldn't happen if we are the |
| 2120 | * client. |
| 2121 | */ |
Sage Weil | 0417788 | 2011-05-12 15:33:17 -0700 | [diff] [blame] | 2122 | pr_err("process_connect got WAIT as client\n"); |
| 2123 | con->error_msg = "protocol error, got WAIT as client"; |
| 2124 | return -1; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2125 | |
| 2126 | default: |
| 2127 | pr_err("connect protocol error, will retry\n"); |
| 2128 | con->error_msg = "protocol error, garbage tag during connect"; |
| 2129 | return -1; |
| 2130 | } |
| 2131 | return 0; |
| 2132 | } |
| 2133 | |
| 2134 | |
| 2135 | /* |
| 2136 | * read (part of) an ack |
| 2137 | */ |
| 2138 | static int read_partial_ack(struct ceph_connection *con) |
| 2139 | { |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 2140 | int size = sizeof (con->in_temp_ack); |
| 2141 | int end = size; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2142 | |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 2143 | return read_partial(con, end, size, &con->in_temp_ack); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2144 | } |
| 2145 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2146 | /* |
| 2147 | * We can finally discard anything that's been acked. |
| 2148 | */ |
| 2149 | static void process_ack(struct ceph_connection *con) |
| 2150 | { |
| 2151 | struct ceph_msg *m; |
| 2152 | u64 ack = le64_to_cpu(con->in_temp_ack); |
| 2153 | u64 seq; |
| 2154 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2155 | while (!list_empty(&con->out_sent)) { |
| 2156 | m = list_first_entry(&con->out_sent, struct ceph_msg, |
| 2157 | list_head); |
| 2158 | seq = le64_to_cpu(m->hdr.seq); |
| 2159 | if (seq > ack) |
| 2160 | break; |
| 2161 | dout("got ack for seq %llu type %d at %p\n", seq, |
| 2162 | le16_to_cpu(m->hdr.type), m); |
Sage Weil | 4cf9d54 | 2011-07-26 11:27:24 -0700 | [diff] [blame] | 2163 | m->ack_stamp = jiffies; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2164 | ceph_msg_remove(m); |
| 2165 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2166 | prepare_read_tag(con); |
| 2167 | } |
| 2168 | |
| 2169 | |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2170 | static int read_partial_message_section(struct ceph_connection *con, |
Sage Weil | 213c99e | 2010-08-03 10:25:11 -0700 | [diff] [blame] | 2171 | struct kvec *section, |
| 2172 | unsigned int sec_len, u32 *crc) |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2173 | { |
Yehuda Sadeh | 68b4476 | 2010-04-06 15:01:27 -0700 | [diff] [blame] | 2174 | int ret, left; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2175 | |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2176 | BUG_ON(!section); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2177 | |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2178 | while (section->iov_len < sec_len) { |
| 2179 | BUG_ON(section->iov_base == NULL); |
| 2180 | left = sec_len - section->iov_len; |
| 2181 | ret = ceph_tcp_recvmsg(con->sock, (char *)section->iov_base + |
| 2182 | section->iov_len, left); |
| 2183 | if (ret <= 0) |
| 2184 | return ret; |
| 2185 | section->iov_len += ret; |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2186 | } |
Alex Elder | fe3ad59 | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 2187 | if (section->iov_len == sec_len) |
| 2188 | *crc = crc32c(0, section->iov_base, section->iov_len); |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2189 | |
| 2190 | return 1; |
| 2191 | } |
| 2192 | |
Alex Elder | 34d2d20 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2193 | static int read_partial_msg_data(struct ceph_connection *con) |
| 2194 | { |
| 2195 | struct ceph_msg *msg = con->in_msg; |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 2196 | struct ceph_msg_data_cursor *cursor = &msg->cursor; |
Alex Elder | 34d2d20 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2197 | const bool do_datacrc = !con->msgr->nocrc; |
Alex Elder | 686be20 | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 2198 | struct page *page; |
| 2199 | size_t page_offset; |
| 2200 | size_t length; |
Alex Elder | f5db90b | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 2201 | u32 crc = 0; |
Alex Elder | 34d2d20 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2202 | int ret; |
| 2203 | |
| 2204 | BUG_ON(!msg); |
Alex Elder | 5240d9f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 2205 | if (list_empty(&msg->data)) |
Alex Elder | 4c59b4a | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 2206 | return -EIO; |
Alex Elder | 34d2d20 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2207 | |
Alex Elder | f5db90b | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 2208 | if (do_datacrc) |
| 2209 | crc = con->in_data_crc; |
Alex Elder | 643c68a | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 2210 | while (cursor->resid) { |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 2211 | page = ceph_msg_data_next(&msg->cursor, &page_offset, &length, |
Alex Elder | 4c59b4a | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 2212 | NULL); |
Alex Elder | 686be20 | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 2213 | ret = ceph_tcp_recvpage(con->sock, page, page_offset, length); |
Alex Elder | f5db90b | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 2214 | if (ret <= 0) { |
| 2215 | if (do_datacrc) |
| 2216 | con->in_data_crc = crc; |
| 2217 | |
Alex Elder | 686be20 | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 2218 | return ret; |
Alex Elder | f5db90b | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 2219 | } |
Alex Elder | 686be20 | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 2220 | |
| 2221 | if (do_datacrc) |
Alex Elder | f5db90b | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 2222 | crc = ceph_crc32c_page(crc, page, page_offset, ret); |
Alex Elder | 8ae4f4f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 2223 | (void) ceph_msg_data_advance(&msg->cursor, (size_t)ret); |
Alex Elder | 34d2d20 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2224 | } |
Alex Elder | f5db90b | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 2225 | if (do_datacrc) |
| 2226 | con->in_data_crc = crc; |
Alex Elder | 34d2d20 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2227 | |
| 2228 | return 1; /* must return > 0 to indicate success */ |
| 2229 | } |
| 2230 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2231 | /* |
| 2232 | * read (part of) a message. |
| 2233 | */ |
Alex Elder | 686be20 | 2013-03-11 23:34:23 -0500 | [diff] [blame] | 2234 | static int ceph_con_in_msg_alloc(struct ceph_connection *con, int *skip); |
| 2235 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2236 | static int read_partial_message(struct ceph_connection *con) |
| 2237 | { |
| 2238 | struct ceph_msg *m = con->in_msg; |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 2239 | int size; |
| 2240 | int end; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2241 | int ret; |
Eric Dumazet | 95c9617 | 2012-04-15 05:58:06 +0000 | [diff] [blame] | 2242 | unsigned int front_len, middle_len, data_len; |
Alex Elder | 37675b0 | 2012-03-07 11:40:08 -0600 | [diff] [blame] | 2243 | bool do_datacrc = !con->msgr->nocrc; |
Sage Weil | ae18756 | 2010-04-22 07:47:01 -0700 | [diff] [blame] | 2244 | u64 seq; |
Alex Elder | fe3ad59 | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 2245 | u32 crc; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2246 | |
| 2247 | dout("read_partial_message con %p msg %p\n", con, m); |
| 2248 | |
| 2249 | /* header */ |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 2250 | size = sizeof (con->in_hdr); |
| 2251 | end = size; |
| 2252 | ret = read_partial(con, end, size, &con->in_hdr); |
Alex Elder | 57dac9d | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 2253 | if (ret <= 0) |
| 2254 | return ret; |
Alex Elder | fe3ad59 | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 2255 | |
| 2256 | crc = crc32c(0, &con->in_hdr, offsetof(struct ceph_msg_header, crc)); |
| 2257 | if (cpu_to_le32(crc) != con->in_hdr.crc) { |
| 2258 | pr_err("read_partial_message bad hdr " |
| 2259 | " crc %u != expected %u\n", |
| 2260 | crc, con->in_hdr.crc); |
| 2261 | return -EBADMSG; |
| 2262 | } |
| 2263 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2264 | front_len = le32_to_cpu(con->in_hdr.front_len); |
| 2265 | if (front_len > CEPH_MSG_MAX_FRONT_LEN) |
| 2266 | return -EIO; |
| 2267 | middle_len = le32_to_cpu(con->in_hdr.middle_len); |
Alex Elder | 7b11ba3 | 2013-03-08 18:51:03 -0600 | [diff] [blame] | 2268 | if (middle_len > CEPH_MSG_MAX_MIDDLE_LEN) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2269 | return -EIO; |
| 2270 | data_len = le32_to_cpu(con->in_hdr.data_len); |
| 2271 | if (data_len > CEPH_MSG_MAX_DATA_LEN) |
| 2272 | return -EIO; |
| 2273 | |
Sage Weil | ae18756 | 2010-04-22 07:47:01 -0700 | [diff] [blame] | 2274 | /* verify seq# */ |
| 2275 | seq = le64_to_cpu(con->in_hdr.seq); |
| 2276 | if ((s64)seq - (s64)con->in_seq < 1) { |
Sage Weil | df9f86f | 2010-11-01 15:49:23 -0700 | [diff] [blame] | 2277 | pr_info("skipping %s%lld %s seq %lld expected %lld\n", |
Sage Weil | ae18756 | 2010-04-22 07:47:01 -0700 | [diff] [blame] | 2278 | ENTITY_NAME(con->peer_name), |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2279 | ceph_pr_addr(&con->peer_addr.in_addr), |
Sage Weil | ae18756 | 2010-04-22 07:47:01 -0700 | [diff] [blame] | 2280 | seq, con->in_seq + 1); |
| 2281 | con->in_base_pos = -front_len - middle_len - data_len - |
| 2282 | sizeof(m->footer); |
| 2283 | con->in_tag = CEPH_MSGR_TAG_READY; |
Sage Weil | ae18756 | 2010-04-22 07:47:01 -0700 | [diff] [blame] | 2284 | return 0; |
| 2285 | } else if ((s64)seq - (s64)con->in_seq > 1) { |
| 2286 | pr_err("read_partial_message bad seq %lld expected %lld\n", |
| 2287 | seq, con->in_seq + 1); |
| 2288 | con->error_msg = "bad message sequence # for incoming message"; |
| 2289 | return -EBADMSG; |
| 2290 | } |
| 2291 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2292 | /* allocate message? */ |
| 2293 | if (!con->in_msg) { |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 2294 | int skip = 0; |
| 2295 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2296 | 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] | 2297 | front_len, data_len); |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 2298 | ret = ceph_con_in_msg_alloc(con, &skip); |
| 2299 | if (ret < 0) |
| 2300 | return ret; |
Alex Elder | f759ebb | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 2301 | |
| 2302 | BUG_ON(!con->in_msg ^ skip); |
| 2303 | if (con->in_msg && data_len > con->in_msg->data_length) { |
| 2304 | pr_warning("%s skipping long message (%u > %zd)\n", |
| 2305 | __func__, data_len, con->in_msg->data_length); |
| 2306 | ceph_msg_put(con->in_msg); |
| 2307 | con->in_msg = NULL; |
| 2308 | skip = 1; |
| 2309 | } |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2310 | if (skip) { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2311 | /* skip this message */ |
Sage Weil | a79832f | 2010-04-01 16:06:19 -0700 | [diff] [blame] | 2312 | dout("alloc_msg said skip message\n"); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2313 | con->in_base_pos = -front_len - middle_len - data_len - |
| 2314 | sizeof(m->footer); |
| 2315 | con->in_tag = CEPH_MSGR_TAG_READY; |
Sage Weil | 684be25 | 2010-04-21 20:45:59 -0700 | [diff] [blame] | 2316 | con->in_seq++; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2317 | return 0; |
| 2318 | } |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2319 | |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 2320 | BUG_ON(!con->in_msg); |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2321 | BUG_ON(con->in_msg->con != con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2322 | m = con->in_msg; |
| 2323 | m->front.iov_len = 0; /* haven't read it yet */ |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2324 | if (m->middle) |
| 2325 | m->middle->vec.iov_len = 0; |
Yehuda Sadeh | 9d7f0f1 | 2010-01-11 10:32:02 -0800 | [diff] [blame] | 2326 | |
Alex Elder | 7862505 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 2327 | /* prepare for data payload, if any */ |
Sage Weil | a410702 | 2012-07-30 16:20:25 -0700 | [diff] [blame] | 2328 | |
Alex Elder | 7862505 | 2013-03-06 23:39:39 -0600 | [diff] [blame] | 2329 | if (data_len) |
Alex Elder | 98fa5dd | 2013-04-02 12:09:50 -0500 | [diff] [blame] | 2330 | prepare_message_data(con->in_msg, data_len); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2331 | } |
| 2332 | |
| 2333 | /* front */ |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2334 | ret = read_partial_message_section(con, &m->front, front_len, |
| 2335 | &con->in_front_crc); |
| 2336 | if (ret <= 0) |
| 2337 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2338 | |
| 2339 | /* middle */ |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2340 | if (m->middle) { |
Sage Weil | 213c99e | 2010-08-03 10:25:11 -0700 | [diff] [blame] | 2341 | ret = read_partial_message_section(con, &m->middle->vec, |
| 2342 | middle_len, |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 2343 | &con->in_middle_crc); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2344 | if (ret <= 0) |
| 2345 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2346 | } |
| 2347 | |
| 2348 | /* (page) data */ |
Alex Elder | 34d2d20 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2349 | if (data_len) { |
| 2350 | ret = read_partial_msg_data(con); |
| 2351 | if (ret <= 0) |
| 2352 | return ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2353 | } |
| 2354 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2355 | /* footer */ |
Alex Elder | fd51653 | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 2356 | size = sizeof (m->footer); |
| 2357 | end += size; |
| 2358 | ret = read_partial(con, end, size, &m->footer); |
Alex Elder | 57dac9d | 2012-05-10 10:29:50 -0500 | [diff] [blame] | 2359 | if (ret <= 0) |
| 2360 | return ret; |
| 2361 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2362 | dout("read_partial_message got msg %p %d (%u) + %d (%u) + %d (%u)\n", |
| 2363 | m, front_len, m->footer.front_crc, middle_len, |
| 2364 | m->footer.middle_crc, data_len, m->footer.data_crc); |
| 2365 | |
| 2366 | /* crc ok? */ |
| 2367 | if (con->in_front_crc != le32_to_cpu(m->footer.front_crc)) { |
| 2368 | pr_err("read_partial_message %p front crc %u != exp. %u\n", |
| 2369 | m, con->in_front_crc, m->footer.front_crc); |
| 2370 | return -EBADMSG; |
| 2371 | } |
| 2372 | if (con->in_middle_crc != le32_to_cpu(m->footer.middle_crc)) { |
| 2373 | pr_err("read_partial_message %p middle crc %u != exp %u\n", |
| 2374 | m, con->in_middle_crc, m->footer.middle_crc); |
| 2375 | return -EBADMSG; |
| 2376 | } |
Alex Elder | bca064d | 2012-02-15 07:43:54 -0600 | [diff] [blame] | 2377 | if (do_datacrc && |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2378 | (m->footer.flags & CEPH_MSG_FOOTER_NOCRC) == 0 && |
| 2379 | con->in_data_crc != le32_to_cpu(m->footer.data_crc)) { |
| 2380 | pr_err("read_partial_message %p data crc %u != exp. %u\n", m, |
| 2381 | con->in_data_crc, le32_to_cpu(m->footer.data_crc)); |
| 2382 | return -EBADMSG; |
| 2383 | } |
| 2384 | |
| 2385 | return 1; /* done! */ |
| 2386 | } |
| 2387 | |
| 2388 | /* |
| 2389 | * Process message. This happens in the worker thread. The callback should |
| 2390 | * be careful not to do anything that waits on other incoming messages or it |
| 2391 | * may deadlock. |
| 2392 | */ |
| 2393 | static void process_message(struct ceph_connection *con) |
| 2394 | { |
Sage Weil | 5e095e8 | 2009-12-14 14:30:34 -0800 | [diff] [blame] | 2395 | struct ceph_msg *msg; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2396 | |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2397 | BUG_ON(con->in_msg->con != con); |
| 2398 | con->in_msg->con = NULL; |
Sage Weil | 5e095e8 | 2009-12-14 14:30:34 -0800 | [diff] [blame] | 2399 | msg = con->in_msg; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2400 | con->in_msg = NULL; |
Sage Weil | 36eb71a | 2012-06-21 12:47:08 -0700 | [diff] [blame] | 2401 | con->ops->put(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2402 | |
| 2403 | /* if first message, set peer_name */ |
| 2404 | if (con->peer_name.type == 0) |
Sage Weil | dbad185 | 2010-03-25 15:45:38 -0700 | [diff] [blame] | 2405 | con->peer_name = msg->hdr.src; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2406 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2407 | con->in_seq++; |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 2408 | mutex_unlock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2409 | |
| 2410 | dout("===== %p %llu from %s%lld %d=%s len %d+%d (%u %u %u) =====\n", |
| 2411 | msg, le64_to_cpu(msg->hdr.seq), |
Sage Weil | dbad185 | 2010-03-25 15:45:38 -0700 | [diff] [blame] | 2412 | ENTITY_NAME(msg->hdr.src), |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2413 | le16_to_cpu(msg->hdr.type), |
| 2414 | ceph_msg_type_name(le16_to_cpu(msg->hdr.type)), |
| 2415 | le32_to_cpu(msg->hdr.front_len), |
| 2416 | le32_to_cpu(msg->hdr.data_len), |
| 2417 | con->in_front_crc, con->in_middle_crc, con->in_data_crc); |
| 2418 | con->ops->dispatch(con, msg); |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 2419 | |
| 2420 | mutex_lock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2421 | } |
| 2422 | |
| 2423 | |
| 2424 | /* |
| 2425 | * Write something to the socket. Called in a worker thread when the |
| 2426 | * socket appears to be writeable and we have something ready to send. |
| 2427 | */ |
| 2428 | static int try_write(struct ceph_connection *con) |
| 2429 | { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2430 | int ret = 1; |
| 2431 | |
Sage Weil | d59315c | 2012-06-21 12:49:23 -0700 | [diff] [blame] | 2432 | dout("try_write start %p state %lu\n", con, con->state); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2433 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2434 | more: |
| 2435 | dout("try_write out_kvec_bytes %d\n", con->out_kvec_bytes); |
| 2436 | |
| 2437 | /* open the socket first? */ |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2438 | if (con->state == CON_STATE_PREOPEN) { |
| 2439 | BUG_ON(con->sock); |
| 2440 | con->state = CON_STATE_CONNECTING; |
Alex Elder | a5988c4 | 2012-05-29 11:04:58 -0500 | [diff] [blame] | 2441 | |
Alex Elder | e220042 | 2012-05-23 14:35:23 -0500 | [diff] [blame] | 2442 | con_out_kvec_reset(con); |
Alex Elder | e825a66 | 2012-05-16 15:16:38 -0500 | [diff] [blame] | 2443 | prepare_write_banner(con); |
Sage Weil | eed0ef2 | 2009-11-10 14:34:36 -0800 | [diff] [blame] | 2444 | prepare_read_banner(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2445 | |
Sage Weil | cf3e5c4 | 2009-12-11 09:48:05 -0800 | [diff] [blame] | 2446 | BUG_ON(con->in_msg); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2447 | con->in_tag = CEPH_MSGR_TAG_READY; |
| 2448 | dout("try_write initiating connect on %p new state %lu\n", |
| 2449 | con, con->state); |
Alex Elder | 41617d0 | 2012-02-14 14:05:33 -0600 | [diff] [blame] | 2450 | ret = ceph_tcp_connect(con); |
| 2451 | if (ret < 0) { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2452 | con->error_msg = "connect error"; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2453 | goto out; |
| 2454 | } |
| 2455 | } |
| 2456 | |
| 2457 | more_kvec: |
| 2458 | /* kvec data queued? */ |
| 2459 | if (con->out_skip) { |
| 2460 | ret = write_partial_skip(con); |
| 2461 | if (ret <= 0) |
Sage Weil | 42961d2 | 2011-01-25 08:19:34 -0800 | [diff] [blame] | 2462 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2463 | } |
| 2464 | if (con->out_kvec_left) { |
| 2465 | ret = write_partial_kvec(con); |
| 2466 | if (ret <= 0) |
Sage Weil | 42961d2 | 2011-01-25 08:19:34 -0800 | [diff] [blame] | 2467 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2468 | } |
| 2469 | |
| 2470 | /* msg pages? */ |
| 2471 | if (con->out_msg) { |
Sage Weil | c86a293 | 2009-12-14 14:04:30 -0800 | [diff] [blame] | 2472 | if (con->out_msg_done) { |
| 2473 | ceph_msg_put(con->out_msg); |
| 2474 | con->out_msg = NULL; /* we're done with this one */ |
| 2475 | goto do_next; |
| 2476 | } |
| 2477 | |
Alex Elder | 34d2d20 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2478 | ret = write_partial_message_data(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2479 | if (ret == 1) |
| 2480 | goto more_kvec; /* we need to send the footer, too! */ |
| 2481 | if (ret == 0) |
Sage Weil | 42961d2 | 2011-01-25 08:19:34 -0800 | [diff] [blame] | 2482 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2483 | if (ret < 0) { |
Alex Elder | 34d2d20 | 2013-03-08 20:58:59 -0600 | [diff] [blame] | 2484 | dout("try_write write_partial_message_data err %d\n", |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2485 | ret); |
Sage Weil | 42961d2 | 2011-01-25 08:19:34 -0800 | [diff] [blame] | 2486 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2487 | } |
| 2488 | } |
| 2489 | |
Sage Weil | c86a293 | 2009-12-14 14:04:30 -0800 | [diff] [blame] | 2490 | do_next: |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2491 | if (con->state == CON_STATE_OPEN) { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2492 | /* is anything else pending? */ |
| 2493 | if (!list_empty(&con->out_queue)) { |
| 2494 | prepare_write_message(con); |
| 2495 | goto more; |
| 2496 | } |
| 2497 | if (con->in_seq > con->in_seq_acked) { |
| 2498 | prepare_write_ack(con); |
| 2499 | goto more; |
| 2500 | } |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2501 | if (con_flag_test_and_clear(con, CON_FLAG_KEEPALIVE_PENDING)) { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2502 | prepare_write_keepalive(con); |
| 2503 | goto more; |
| 2504 | } |
| 2505 | } |
| 2506 | |
| 2507 | /* Nothing to do! */ |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2508 | con_flag_clear(con, CON_FLAG_WRITE_PENDING); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2509 | dout("try_write nothing else to write.\n"); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2510 | ret = 0; |
| 2511 | out: |
Sage Weil | 42961d2 | 2011-01-25 08:19:34 -0800 | [diff] [blame] | 2512 | dout("try_write done on %p ret %d\n", con, ret); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2513 | return ret; |
| 2514 | } |
| 2515 | |
| 2516 | |
| 2517 | |
| 2518 | /* |
| 2519 | * Read what we can from the socket. |
| 2520 | */ |
| 2521 | static int try_read(struct ceph_connection *con) |
| 2522 | { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2523 | int ret = -1; |
| 2524 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2525 | more: |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2526 | dout("try_read start on %p state %lu\n", con, con->state); |
| 2527 | if (con->state != CON_STATE_CONNECTING && |
| 2528 | con->state != CON_STATE_NEGOTIATING && |
| 2529 | con->state != CON_STATE_OPEN) |
| 2530 | return 0; |
| 2531 | |
| 2532 | BUG_ON(!con->sock); |
| 2533 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2534 | dout("try_read tag %d in_base_pos %d\n", (int)con->in_tag, |
| 2535 | con->in_base_pos); |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 2536 | |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2537 | if (con->state == CON_STATE_CONNECTING) { |
Alex Elder | 7593af9 | 2012-05-24 11:55:03 -0500 | [diff] [blame] | 2538 | dout("try_read connecting\n"); |
| 2539 | ret = read_partial_banner(con); |
| 2540 | if (ret <= 0) |
Alex Elder | ab166d5 | 2012-05-31 11:37:29 -0500 | [diff] [blame] | 2541 | goto out; |
Alex Elder | 7593af9 | 2012-05-24 11:55:03 -0500 | [diff] [blame] | 2542 | ret = process_banner(con); |
| 2543 | if (ret < 0) |
| 2544 | goto out; |
| 2545 | |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2546 | con->state = CON_STATE_NEGOTIATING; |
Alex Elder | 7593af9 | 2012-05-24 11:55:03 -0500 | [diff] [blame] | 2547 | |
Jim Schutt | 6d4221b | 2012-08-10 10:37:38 -0700 | [diff] [blame] | 2548 | /* |
| 2549 | * Received banner is good, exchange connection info. |
| 2550 | * Do not reset out_kvec, as sending our banner raced |
| 2551 | * with receiving peer banner after connect completed. |
| 2552 | */ |
Alex Elder | 7593af9 | 2012-05-24 11:55:03 -0500 | [diff] [blame] | 2553 | ret = prepare_write_connect(con); |
| 2554 | if (ret < 0) |
| 2555 | goto out; |
| 2556 | prepare_read_connect(con); |
| 2557 | |
| 2558 | /* Send connection info before awaiting response */ |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 2559 | goto out; |
| 2560 | } |
| 2561 | |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2562 | if (con->state == CON_STATE_NEGOTIATING) { |
Alex Elder | 7593af9 | 2012-05-24 11:55:03 -0500 | [diff] [blame] | 2563 | dout("try_read negotiating\n"); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2564 | ret = read_partial_connect(con); |
| 2565 | if (ret <= 0) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2566 | goto out; |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2567 | ret = process_connect(con); |
| 2568 | if (ret < 0) |
| 2569 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2570 | goto more; |
| 2571 | } |
| 2572 | |
Alex Elder | 122070a | 2012-12-26 10:43:57 -0600 | [diff] [blame] | 2573 | WARN_ON(con->state != CON_STATE_OPEN); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2574 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2575 | if (con->in_base_pos < 0) { |
| 2576 | /* |
| 2577 | * skipping + discarding content. |
| 2578 | * |
| 2579 | * FIXME: there must be a better way to do this! |
| 2580 | */ |
Alex Elder | 84495f4 | 2012-02-15 07:43:55 -0600 | [diff] [blame] | 2581 | static char buf[SKIP_BUF_SIZE]; |
| 2582 | int skip = min((int) sizeof (buf), -con->in_base_pos); |
| 2583 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2584 | dout("skipping %d / %d bytes\n", skip, -con->in_base_pos); |
| 2585 | ret = ceph_tcp_recvmsg(con->sock, buf, skip); |
| 2586 | if (ret <= 0) |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2587 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2588 | con->in_base_pos += ret; |
| 2589 | if (con->in_base_pos) |
| 2590 | goto more; |
| 2591 | } |
| 2592 | if (con->in_tag == CEPH_MSGR_TAG_READY) { |
| 2593 | /* |
| 2594 | * what's next? |
| 2595 | */ |
| 2596 | ret = ceph_tcp_recvmsg(con->sock, &con->in_tag, 1); |
| 2597 | if (ret <= 0) |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2598 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2599 | dout("try_read got tag %d\n", (int)con->in_tag); |
| 2600 | switch (con->in_tag) { |
| 2601 | case CEPH_MSGR_TAG_MSG: |
| 2602 | prepare_read_message(con); |
| 2603 | break; |
| 2604 | case CEPH_MSGR_TAG_ACK: |
| 2605 | prepare_read_ack(con); |
| 2606 | break; |
| 2607 | case CEPH_MSGR_TAG_CLOSE: |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2608 | con_close_socket(con); |
| 2609 | con->state = CON_STATE_CLOSED; |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2610 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2611 | default: |
| 2612 | goto bad_tag; |
| 2613 | } |
| 2614 | } |
| 2615 | if (con->in_tag == CEPH_MSGR_TAG_MSG) { |
| 2616 | ret = read_partial_message(con); |
| 2617 | if (ret <= 0) { |
| 2618 | switch (ret) { |
| 2619 | case -EBADMSG: |
| 2620 | con->error_msg = "bad crc"; |
| 2621 | ret = -EIO; |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2622 | break; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2623 | case -EIO: |
| 2624 | con->error_msg = "io error"; |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2625 | break; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2626 | } |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2627 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2628 | } |
| 2629 | if (con->in_tag == CEPH_MSGR_TAG_READY) |
| 2630 | goto more; |
| 2631 | process_message(con); |
Sage Weil | 7b862e0 | 2012-07-30 18:16:56 -0700 | [diff] [blame] | 2632 | if (con->state == CON_STATE_OPEN) |
| 2633 | prepare_read_tag(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2634 | goto more; |
| 2635 | } |
Sage Weil | 3a23083 | 2013-03-25 08:47:40 -0700 | [diff] [blame] | 2636 | if (con->in_tag == CEPH_MSGR_TAG_ACK || |
| 2637 | con->in_tag == CEPH_MSGR_TAG_SEQ) { |
| 2638 | /* |
| 2639 | * the final handshake seq exchange is semantically |
| 2640 | * equivalent to an ACK |
| 2641 | */ |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2642 | ret = read_partial_ack(con); |
| 2643 | if (ret <= 0) |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2644 | goto out; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2645 | process_ack(con); |
| 2646 | goto more; |
| 2647 | } |
| 2648 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2649 | out: |
Sage Weil | 98bdb0a | 2011-01-25 08:17:48 -0800 | [diff] [blame] | 2650 | dout("try_read done on %p ret %d\n", con, ret); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2651 | return ret; |
| 2652 | |
| 2653 | bad_tag: |
| 2654 | pr_err("try_read bad con->in_tag = %d\n", (int)con->in_tag); |
| 2655 | con->error_msg = "protocol error, garbage tag"; |
| 2656 | ret = -1; |
| 2657 | goto out; |
| 2658 | } |
| 2659 | |
| 2660 | |
| 2661 | /* |
Alex Elder | 802c6d9 | 2012-10-08 20:37:30 -0700 | [diff] [blame] | 2662 | * Atomically queue work on a connection after the specified delay. |
| 2663 | * Bump @con reference to avoid races with connection teardown. |
| 2664 | * Returns 0 if work was queued, or an error code otherwise. |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2665 | */ |
Alex Elder | 802c6d9 | 2012-10-08 20:37:30 -0700 | [diff] [blame] | 2666 | static int queue_con_delay(struct ceph_connection *con, unsigned long delay) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2667 | { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2668 | if (!con->ops->get(con)) { |
Alex Elder | 802c6d9 | 2012-10-08 20:37:30 -0700 | [diff] [blame] | 2669 | dout("%s %p ref count 0\n", __func__, con); |
| 2670 | |
| 2671 | return -ENOENT; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2672 | } |
| 2673 | |
Alex Elder | 802c6d9 | 2012-10-08 20:37:30 -0700 | [diff] [blame] | 2674 | if (!queue_delayed_work(ceph_msgr_wq, &con->work, delay)) { |
| 2675 | dout("%s %p - already queued\n", __func__, con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2676 | con->ops->put(con); |
Alex Elder | 802c6d9 | 2012-10-08 20:37:30 -0700 | [diff] [blame] | 2677 | |
| 2678 | return -EBUSY; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2679 | } |
Alex Elder | 802c6d9 | 2012-10-08 20:37:30 -0700 | [diff] [blame] | 2680 | |
| 2681 | dout("%s %p %lu\n", __func__, con, delay); |
| 2682 | |
| 2683 | return 0; |
| 2684 | } |
| 2685 | |
| 2686 | static void queue_con(struct ceph_connection *con) |
| 2687 | { |
| 2688 | (void) queue_con_delay(con, 0); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2689 | } |
| 2690 | |
Alex Elder | 7bb21d68 | 2012-12-07 19:50:07 -0600 | [diff] [blame] | 2691 | static bool con_sock_closed(struct ceph_connection *con) |
| 2692 | { |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2693 | if (!con_flag_test_and_clear(con, CON_FLAG_SOCK_CLOSED)) |
Alex Elder | 7bb21d68 | 2012-12-07 19:50:07 -0600 | [diff] [blame] | 2694 | return false; |
| 2695 | |
| 2696 | #define CASE(x) \ |
| 2697 | case CON_STATE_ ## x: \ |
| 2698 | con->error_msg = "socket closed (con state " #x ")"; \ |
| 2699 | break; |
| 2700 | |
| 2701 | switch (con->state) { |
| 2702 | CASE(CLOSED); |
| 2703 | CASE(PREOPEN); |
| 2704 | CASE(CONNECTING); |
| 2705 | CASE(NEGOTIATING); |
| 2706 | CASE(OPEN); |
| 2707 | CASE(STANDBY); |
| 2708 | default: |
| 2709 | pr_warning("%s con %p unrecognized state %lu\n", |
| 2710 | __func__, con, con->state); |
| 2711 | con->error_msg = "unrecognized con state"; |
| 2712 | BUG(); |
| 2713 | break; |
| 2714 | } |
| 2715 | #undef CASE |
| 2716 | |
| 2717 | return true; |
| 2718 | } |
| 2719 | |
Alex Elder | f20a39f | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2720 | static bool con_backoff(struct ceph_connection *con) |
| 2721 | { |
| 2722 | int ret; |
| 2723 | |
| 2724 | if (!con_flag_test_and_clear(con, CON_FLAG_BACKOFF)) |
| 2725 | return false; |
| 2726 | |
| 2727 | ret = queue_con_delay(con, round_jiffies_relative(con->delay)); |
| 2728 | if (ret) { |
| 2729 | dout("%s: con %p FAILED to back off %lu\n", __func__, |
| 2730 | con, con->delay); |
| 2731 | BUG_ON(ret == -ENOENT); |
| 2732 | con_flag_set(con, CON_FLAG_BACKOFF); |
| 2733 | } |
| 2734 | |
| 2735 | return true; |
| 2736 | } |
| 2737 | |
Alex Elder | 9320926 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2738 | /* Finish fault handling; con->mutex must *not* be held here */ |
| 2739 | |
| 2740 | static void con_fault_finish(struct ceph_connection *con) |
| 2741 | { |
| 2742 | /* |
| 2743 | * in case we faulted due to authentication, invalidate our |
| 2744 | * current tickets so that we can get new ones. |
| 2745 | */ |
| 2746 | if (con->auth_retry && con->ops->invalidate_authorizer) { |
| 2747 | dout("calling invalidate_authorizer()\n"); |
| 2748 | con->ops->invalidate_authorizer(con); |
| 2749 | } |
| 2750 | |
| 2751 | if (con->ops->fault) |
| 2752 | con->ops->fault(con); |
| 2753 | } |
| 2754 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2755 | /* |
| 2756 | * Do some work on a connection. Drop a connection ref when we're done. |
| 2757 | */ |
| 2758 | static void con_work(struct work_struct *work) |
| 2759 | { |
| 2760 | struct ceph_connection *con = container_of(work, struct ceph_connection, |
| 2761 | work.work); |
Alex Elder | 4965941 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2762 | bool fault; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2763 | |
Sage Weil | 9dd4658 | 2010-04-28 13:51:50 -0700 | [diff] [blame] | 2764 | mutex_lock(&con->mutex); |
Alex Elder | 4965941 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2765 | while (true) { |
| 2766 | int ret; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2767 | |
Alex Elder | 4965941 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2768 | if ((fault = con_sock_closed(con))) { |
| 2769 | dout("%s: con %p SOCK_CLOSED\n", __func__, con); |
| 2770 | break; |
| 2771 | } |
| 2772 | if (con_backoff(con)) { |
| 2773 | dout("%s: con %p BACKOFF\n", __func__, con); |
| 2774 | break; |
| 2775 | } |
| 2776 | if (con->state == CON_STATE_STANDBY) { |
| 2777 | dout("%s: con %p STANDBY\n", __func__, con); |
| 2778 | break; |
| 2779 | } |
| 2780 | if (con->state == CON_STATE_CLOSED) { |
| 2781 | dout("%s: con %p CLOSED\n", __func__, con); |
| 2782 | BUG_ON(con->sock); |
| 2783 | break; |
| 2784 | } |
| 2785 | if (con->state == CON_STATE_PREOPEN) { |
| 2786 | dout("%s: con %p PREOPEN\n", __func__, con); |
| 2787 | BUG_ON(con->sock); |
| 2788 | } |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 2789 | |
Alex Elder | 4965941 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2790 | ret = try_read(con); |
| 2791 | if (ret < 0) { |
| 2792 | if (ret == -EAGAIN) |
| 2793 | continue; |
| 2794 | con->error_msg = "socket error on read"; |
| 2795 | fault = true; |
| 2796 | break; |
| 2797 | } |
| 2798 | |
| 2799 | ret = try_write(con); |
| 2800 | if (ret < 0) { |
| 2801 | if (ret == -EAGAIN) |
| 2802 | continue; |
| 2803 | con->error_msg = "socket error on write"; |
| 2804 | fault = true; |
| 2805 | } |
| 2806 | |
| 2807 | break; /* If we make it to here, we're done */ |
Sage Weil | 3a140a0 | 2012-07-30 16:24:21 -0700 | [diff] [blame] | 2808 | } |
Alex Elder | b6e7b6a | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2809 | if (fault) |
| 2810 | con_fault(con); |
Sage Weil | 9dd4658 | 2010-04-28 13:51:50 -0700 | [diff] [blame] | 2811 | mutex_unlock(&con->mutex); |
Sage Weil | 0da5d70 | 2011-05-19 11:21:05 -0700 | [diff] [blame] | 2812 | |
Alex Elder | b6e7b6a | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2813 | if (fault) |
| 2814 | con_fault_finish(con); |
| 2815 | |
| 2816 | con->ops->put(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2817 | } |
| 2818 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2819 | /* |
| 2820 | * Generic error/fault handler. A retry mechanism is used with |
| 2821 | * exponential backoff |
| 2822 | */ |
Alex Elder | 9320926 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2823 | static void con_fault(struct ceph_connection *con) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2824 | { |
Alex Elder | 2836298 | 2012-12-14 16:47:41 -0600 | [diff] [blame] | 2825 | pr_warning("%s%lld %s %s\n", ENTITY_NAME(con->peer_name), |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2826 | ceph_pr_addr(&con->peer_addr.in_addr), con->error_msg); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2827 | dout("fault %p state %lu to peer %s\n", |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2828 | con, con->state, ceph_pr_addr(&con->peer_addr.in_addr)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2829 | |
Alex Elder | 122070a | 2012-12-26 10:43:57 -0600 | [diff] [blame] | 2830 | WARN_ON(con->state != CON_STATE_CONNECTING && |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2831 | con->state != CON_STATE_NEGOTIATING && |
| 2832 | con->state != CON_STATE_OPEN); |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 2833 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2834 | con_close_socket(con); |
Sage Weil | 5e095e8 | 2009-12-14 14:30:34 -0800 | [diff] [blame] | 2835 | |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2836 | if (con_flag_test(con, CON_FLAG_LOSSYTX)) { |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2837 | dout("fault on LOSSYTX channel, marking CLOSED\n"); |
| 2838 | con->state = CON_STATE_CLOSED; |
Alex Elder | 9320926 | 2013-02-19 12:25:57 -0600 | [diff] [blame] | 2839 | return; |
Sage Weil | 3b5ede0 | 2012-07-20 15:22:53 -0700 | [diff] [blame] | 2840 | } |
| 2841 | |
Sage Weil | 5e095e8 | 2009-12-14 14:30:34 -0800 | [diff] [blame] | 2842 | if (con->in_msg) { |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2843 | BUG_ON(con->in_msg->con != con); |
| 2844 | con->in_msg->con = NULL; |
Sage Weil | 5e095e8 | 2009-12-14 14:30:34 -0800 | [diff] [blame] | 2845 | ceph_msg_put(con->in_msg); |
| 2846 | con->in_msg = NULL; |
Sage Weil | 36eb71a | 2012-06-21 12:47:08 -0700 | [diff] [blame] | 2847 | con->ops->put(con); |
Sage Weil | 5e095e8 | 2009-12-14 14:30:34 -0800 | [diff] [blame] | 2848 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2849 | |
Sage Weil | e80a52d | 2010-02-25 12:40:45 -0800 | [diff] [blame] | 2850 | /* Requeue anything that hasn't been acked */ |
| 2851 | list_splice_init(&con->out_sent, &con->out_queue); |
Sage Weil | 9bd2e6f | 2010-02-02 16:21:06 -0800 | [diff] [blame] | 2852 | |
Sage Weil | e76661d | 2011-03-03 10:10:15 -0800 | [diff] [blame] | 2853 | /* If there are no messages queued or keepalive pending, place |
| 2854 | * the connection in a STANDBY state */ |
| 2855 | if (list_empty(&con->out_queue) && |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2856 | !con_flag_test(con, CON_FLAG_KEEPALIVE_PENDING)) { |
Sage Weil | e00de34 | 2011-03-04 12:25:05 -0800 | [diff] [blame] | 2857 | dout("fault %p setting STANDBY clearing WRITE_PENDING\n", con); |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2858 | con_flag_clear(con, CON_FLAG_WRITE_PENDING); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2859 | con->state = CON_STATE_STANDBY; |
Sage Weil | e80a52d | 2010-02-25 12:40:45 -0800 | [diff] [blame] | 2860 | } else { |
| 2861 | /* retry after a delay. */ |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2862 | con->state = CON_STATE_PREOPEN; |
Sage Weil | e80a52d | 2010-02-25 12:40:45 -0800 | [diff] [blame] | 2863 | if (con->delay == 0) |
| 2864 | con->delay = BASE_DELAY_INTERVAL; |
| 2865 | else if (con->delay < MAX_DELAY_INTERVAL) |
| 2866 | con->delay *= 2; |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2867 | con_flag_set(con, CON_FLAG_BACKOFF); |
Alex Elder | 8618e30 | 2012-10-08 20:37:30 -0700 | [diff] [blame] | 2868 | queue_con(con); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2869 | } |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2870 | } |
| 2871 | |
| 2872 | |
| 2873 | |
| 2874 | /* |
Alex Elder | 15d9882 | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 2875 | * initialize a new messenger instance |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2876 | */ |
Alex Elder | 15d9882 | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 2877 | void ceph_messenger_init(struct ceph_messenger *msgr, |
| 2878 | struct ceph_entity_addr *myaddr, |
Ilya Dryomov | 12b4629 | 2013-12-24 21:19:23 +0200 | [diff] [blame] | 2879 | u64 supported_features, |
| 2880 | u64 required_features, |
Alex Elder | 15d9882 | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 2881 | bool nocrc) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2882 | { |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2883 | msgr->supported_features = supported_features; |
| 2884 | msgr->required_features = required_features; |
| 2885 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2886 | spin_lock_init(&msgr->global_seq_lock); |
| 2887 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2888 | if (myaddr) |
| 2889 | msgr->inst.addr = *myaddr; |
| 2890 | |
| 2891 | /* select a random nonce */ |
Sage Weil | ac8839d | 2010-01-27 14:28:10 -0800 | [diff] [blame] | 2892 | msgr->inst.addr.type = 0; |
Sage Weil | 103e2d3 | 2010-01-07 16:12:36 -0800 | [diff] [blame] | 2893 | get_random_bytes(&msgr->inst.addr.nonce, sizeof(msgr->inst.addr.nonce)); |
Sage Weil | 63f2d21 | 2009-11-03 15:17:56 -0800 | [diff] [blame] | 2894 | encode_my_addr(msgr); |
Alex Elder | 15d9882 | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 2895 | msgr->nocrc = nocrc; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2896 | |
Guanjun He | a2a3258 | 2012-07-08 19:50:33 -0700 | [diff] [blame] | 2897 | atomic_set(&msgr->stopping, 0); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2898 | |
Alex Elder | 15d9882 | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 2899 | dout("%s %p\n", __func__, msgr); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2900 | } |
Alex Elder | 15d9882 | 2012-05-26 23:26:43 -0500 | [diff] [blame] | 2901 | EXPORT_SYMBOL(ceph_messenger_init); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2902 | |
Sage Weil | e00de34 | 2011-03-04 12:25:05 -0800 | [diff] [blame] | 2903 | static void clear_standby(struct ceph_connection *con) |
| 2904 | { |
| 2905 | /* come back from STANDBY? */ |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2906 | if (con->state == CON_STATE_STANDBY) { |
Sage Weil | e00de34 | 2011-03-04 12:25:05 -0800 | [diff] [blame] | 2907 | dout("clear_standby %p and ++connect_seq\n", con); |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2908 | con->state = CON_STATE_PREOPEN; |
Sage Weil | e00de34 | 2011-03-04 12:25:05 -0800 | [diff] [blame] | 2909 | con->connect_seq++; |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2910 | WARN_ON(con_flag_test(con, CON_FLAG_WRITE_PENDING)); |
| 2911 | WARN_ON(con_flag_test(con, CON_FLAG_KEEPALIVE_PENDING)); |
Sage Weil | e00de34 | 2011-03-04 12:25:05 -0800 | [diff] [blame] | 2912 | } |
| 2913 | } |
| 2914 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2915 | /* |
| 2916 | * Queue up an outgoing message on the given connection. |
| 2917 | */ |
| 2918 | void ceph_con_send(struct ceph_connection *con, struct ceph_msg *msg) |
| 2919 | { |
Sage Weil | a59b55a | 2012-07-20 15:34:04 -0700 | [diff] [blame] | 2920 | /* set src+dst */ |
| 2921 | msg->hdr.src = con->msgr->inst.name; |
| 2922 | BUG_ON(msg->front.iov_len != le32_to_cpu(msg->hdr.front_len)); |
| 2923 | msg->needs_out_seq = true; |
| 2924 | |
| 2925 | mutex_lock(&con->mutex); |
| 2926 | |
Sage Weil | 8dacc7d | 2012-07-20 17:24:40 -0700 | [diff] [blame] | 2927 | if (con->state == CON_STATE_CLOSED) { |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2928 | dout("con_send %p closed, dropping %p\n", con, msg); |
| 2929 | ceph_msg_put(msg); |
Sage Weil | a59b55a | 2012-07-20 15:34:04 -0700 | [diff] [blame] | 2930 | mutex_unlock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2931 | return; |
| 2932 | } |
| 2933 | |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2934 | BUG_ON(msg->con != NULL); |
Sage Weil | 36eb71a | 2012-06-21 12:47:08 -0700 | [diff] [blame] | 2935 | msg->con = con->ops->get(con); |
Alex Elder | 92ce034 | 2012-06-04 14:43:33 -0500 | [diff] [blame] | 2936 | BUG_ON(msg->con == NULL); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2937 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2938 | BUG_ON(!list_empty(&msg->list_head)); |
| 2939 | list_add_tail(&msg->list_head, &con->out_queue); |
| 2940 | dout("----- %p to %s%lld %d=%s len %d+%d+%d -----\n", msg, |
| 2941 | ENTITY_NAME(con->peer_name), le16_to_cpu(msg->hdr.type), |
| 2942 | ceph_msg_type_name(le16_to_cpu(msg->hdr.type)), |
| 2943 | le32_to_cpu(msg->hdr.front_len), |
| 2944 | le32_to_cpu(msg->hdr.middle_len), |
| 2945 | le32_to_cpu(msg->hdr.data_len)); |
Sage Weil | 0065093 | 2012-07-20 15:33:04 -0700 | [diff] [blame] | 2946 | |
| 2947 | clear_standby(con); |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 2948 | mutex_unlock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2949 | |
| 2950 | /* if there wasn't anything waiting to send before, queue |
| 2951 | * new work */ |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 2952 | if (con_flag_test_and_set(con, CON_FLAG_WRITE_PENDING) == 0) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2953 | queue_con(con); |
| 2954 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 2955 | EXPORT_SYMBOL(ceph_con_send); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2956 | |
| 2957 | /* |
| 2958 | * Revoke a message that was previously queued for send |
| 2959 | */ |
Alex Elder | 6740a84 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2960 | void ceph_msg_revoke(struct ceph_msg *msg) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2961 | { |
Alex Elder | 6740a84 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2962 | struct ceph_connection *con = msg->con; |
| 2963 | |
| 2964 | if (!con) |
| 2965 | return; /* Message not in our possession */ |
| 2966 | |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 2967 | mutex_lock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2968 | if (!list_empty(&msg->list_head)) { |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2969 | dout("%s %p msg %p - was on queue\n", __func__, con, msg); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2970 | list_del_init(&msg->list_head); |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2971 | BUG_ON(msg->con == NULL); |
Sage Weil | 36eb71a | 2012-06-21 12:47:08 -0700 | [diff] [blame] | 2972 | msg->con->ops->put(msg->con); |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2973 | msg->con = NULL; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2974 | msg->hdr.seq = 0; |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2975 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2976 | ceph_msg_put(msg); |
Sage Weil | ed98ada | 2010-07-05 12:15:14 -0700 | [diff] [blame] | 2977 | } |
| 2978 | if (con->out_msg == msg) { |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2979 | dout("%s %p msg %p - was sending\n", __func__, con, msg); |
Sage Weil | ed98ada | 2010-07-05 12:15:14 -0700 | [diff] [blame] | 2980 | con->out_msg = NULL; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2981 | if (con->out_kvec_is_msg) { |
| 2982 | con->out_skip = con->out_kvec_bytes; |
| 2983 | con->out_kvec_is_msg = false; |
| 2984 | } |
Sage Weil | ed98ada | 2010-07-05 12:15:14 -0700 | [diff] [blame] | 2985 | msg->hdr.seq = 0; |
Alex Elder | 92ce034 | 2012-06-04 14:43:33 -0500 | [diff] [blame] | 2986 | |
| 2987 | ceph_msg_put(msg); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2988 | } |
Sage Weil | ec30264 | 2009-12-22 10:43:42 -0800 | [diff] [blame] | 2989 | mutex_unlock(&con->mutex); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 2990 | } |
| 2991 | |
| 2992 | /* |
Yehuda Sadeh | 0d59ab8 | 2010-01-13 17:03:23 -0800 | [diff] [blame] | 2993 | * Revoke a message that we may be reading data into |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 2994 | */ |
Alex Elder | 8921d11 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2995 | void ceph_msg_revoke_incoming(struct ceph_msg *msg) |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 2996 | { |
Alex Elder | 8921d11 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 2997 | struct ceph_connection *con; |
| 2998 | |
| 2999 | BUG_ON(msg == NULL); |
| 3000 | if (!msg->con) { |
| 3001 | dout("%s msg %p null con\n", __func__, msg); |
| 3002 | |
| 3003 | return; /* Message not in our possession */ |
| 3004 | } |
| 3005 | |
| 3006 | con = msg->con; |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 3007 | mutex_lock(&con->mutex); |
Alex Elder | 8921d11 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 3008 | if (con->in_msg == msg) { |
Eric Dumazet | 95c9617 | 2012-04-15 05:58:06 +0000 | [diff] [blame] | 3009 | unsigned int front_len = le32_to_cpu(con->in_hdr.front_len); |
| 3010 | unsigned int middle_len = le32_to_cpu(con->in_hdr.middle_len); |
| 3011 | unsigned int data_len = le32_to_cpu(con->in_hdr.data_len); |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 3012 | |
| 3013 | /* skip rest of message */ |
Alex Elder | 8921d11 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 3014 | dout("%s %p msg %p revoked\n", __func__, con, msg); |
| 3015 | con->in_base_pos = con->in_base_pos - |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 3016 | sizeof(struct ceph_msg_header) - |
Yehuda Sadeh | 0d59ab8 | 2010-01-13 17:03:23 -0800 | [diff] [blame] | 3017 | front_len - |
| 3018 | middle_len - |
| 3019 | data_len - |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 3020 | sizeof(struct ceph_msg_footer); |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 3021 | ceph_msg_put(con->in_msg); |
| 3022 | con->in_msg = NULL; |
| 3023 | con->in_tag = CEPH_MSGR_TAG_READY; |
Sage Weil | 684be25 | 2010-04-21 20:45:59 -0700 | [diff] [blame] | 3024 | con->in_seq++; |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 3025 | } else { |
Alex Elder | 8921d11 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 3026 | dout("%s %p in_msg %p msg %p no-op\n", |
| 3027 | __func__, con, con->in_msg, msg); |
Sage Weil | 350b1c3 | 2009-12-22 10:45:45 -0800 | [diff] [blame] | 3028 | } |
| 3029 | mutex_unlock(&con->mutex); |
| 3030 | } |
| 3031 | |
| 3032 | /* |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3033 | * Queue a keepalive byte to ensure the tcp connection is alive. |
| 3034 | */ |
| 3035 | void ceph_con_keepalive(struct ceph_connection *con) |
| 3036 | { |
Sage Weil | e00de34 | 2011-03-04 12:25:05 -0800 | [diff] [blame] | 3037 | dout("con_keepalive %p\n", con); |
Sage Weil | 0065093 | 2012-07-20 15:33:04 -0700 | [diff] [blame] | 3038 | mutex_lock(&con->mutex); |
Sage Weil | e00de34 | 2011-03-04 12:25:05 -0800 | [diff] [blame] | 3039 | clear_standby(con); |
Sage Weil | 0065093 | 2012-07-20 15:33:04 -0700 | [diff] [blame] | 3040 | mutex_unlock(&con->mutex); |
Alex Elder | c9ffc77 | 2013-02-20 10:25:12 -0600 | [diff] [blame] | 3041 | if (con_flag_test_and_set(con, CON_FLAG_KEEPALIVE_PENDING) == 0 && |
| 3042 | con_flag_test_and_set(con, CON_FLAG_WRITE_PENDING) == 0) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3043 | queue_con(con); |
| 3044 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 3045 | EXPORT_SYMBOL(ceph_con_keepalive); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3046 | |
Alex Elder | 6644ed7 | 2013-03-11 23:34:24 -0500 | [diff] [blame] | 3047 | static struct ceph_msg_data *ceph_msg_data_create(enum ceph_msg_data_type type) |
Alex Elder | 4379450 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3048 | { |
Alex Elder | 6644ed7 | 2013-03-11 23:34:24 -0500 | [diff] [blame] | 3049 | struct ceph_msg_data *data; |
| 3050 | |
| 3051 | if (WARN_ON(!ceph_msg_data_type_valid(type))) |
| 3052 | return NULL; |
| 3053 | |
Alex Elder | 81b36be | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 3054 | data = kmem_cache_zalloc(ceph_msg_data_cache, GFP_NOFS); |
Alex Elder | 6644ed7 | 2013-03-11 23:34:24 -0500 | [diff] [blame] | 3055 | if (data) |
| 3056 | data->type = type; |
Alex Elder | 5240d9f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 3057 | INIT_LIST_HEAD(&data->links); |
Alex Elder | 6644ed7 | 2013-03-11 23:34:24 -0500 | [diff] [blame] | 3058 | |
| 3059 | return data; |
| 3060 | } |
| 3061 | |
| 3062 | static void ceph_msg_data_destroy(struct ceph_msg_data *data) |
| 3063 | { |
| 3064 | if (!data) |
| 3065 | return; |
| 3066 | |
Alex Elder | 5240d9f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 3067 | WARN_ON(!list_empty(&data->links)); |
Alex Elder | 6644ed7 | 2013-03-11 23:34:24 -0500 | [diff] [blame] | 3068 | if (data->type == CEPH_MSG_DATA_PAGELIST) { |
| 3069 | ceph_pagelist_release(data->pagelist); |
| 3070 | kfree(data->pagelist); |
| 3071 | } |
Alex Elder | 81b36be | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 3072 | kmem_cache_free(ceph_msg_data_cache, data); |
Alex Elder | 4379450 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3073 | } |
| 3074 | |
Alex Elder | 90af360 | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 3075 | void ceph_msg_data_add_pages(struct ceph_msg *msg, struct page **pages, |
Alex Elder | f1baeb2 | 2013-03-07 15:38:26 -0600 | [diff] [blame] | 3076 | size_t length, size_t alignment) |
Alex Elder | 02afca6 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 3077 | { |
Alex Elder | 6644ed7 | 2013-03-11 23:34:24 -0500 | [diff] [blame] | 3078 | struct ceph_msg_data *data; |
| 3079 | |
Alex Elder | 07aa155 | 2013-03-04 18:29:06 -0600 | [diff] [blame] | 3080 | BUG_ON(!pages); |
| 3081 | BUG_ON(!length); |
Alex Elder | 02afca6 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 3082 | |
Alex Elder | 6644ed7 | 2013-03-11 23:34:24 -0500 | [diff] [blame] | 3083 | data = ceph_msg_data_create(CEPH_MSG_DATA_PAGES); |
| 3084 | BUG_ON(!data); |
| 3085 | data->pages = pages; |
| 3086 | data->length = length; |
| 3087 | data->alignment = alignment & ~PAGE_MASK; |
| 3088 | |
Alex Elder | 5240d9f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 3089 | list_add_tail(&data->links, &msg->data); |
| 3090 | msg->data_length += length; |
Alex Elder | 02afca6 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 3091 | } |
Alex Elder | 90af360 | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 3092 | EXPORT_SYMBOL(ceph_msg_data_add_pages); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3093 | |
Alex Elder | 90af360 | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 3094 | void ceph_msg_data_add_pagelist(struct ceph_msg *msg, |
Alex Elder | 27fa838 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 3095 | struct ceph_pagelist *pagelist) |
| 3096 | { |
Alex Elder | 6644ed7 | 2013-03-11 23:34:24 -0500 | [diff] [blame] | 3097 | struct ceph_msg_data *data; |
| 3098 | |
Alex Elder | 07aa155 | 2013-03-04 18:29:06 -0600 | [diff] [blame] | 3099 | BUG_ON(!pagelist); |
| 3100 | BUG_ON(!pagelist->length); |
Alex Elder | 27fa838 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 3101 | |
Alex Elder | 6644ed7 | 2013-03-11 23:34:24 -0500 | [diff] [blame] | 3102 | data = ceph_msg_data_create(CEPH_MSG_DATA_PAGELIST); |
| 3103 | BUG_ON(!data); |
| 3104 | data->pagelist = pagelist; |
| 3105 | |
Alex Elder | 5240d9f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 3106 | list_add_tail(&data->links, &msg->data); |
| 3107 | msg->data_length += pagelist->length; |
Alex Elder | 27fa838 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 3108 | } |
Alex Elder | 90af360 | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 3109 | EXPORT_SYMBOL(ceph_msg_data_add_pagelist); |
Alex Elder | 27fa838 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 3110 | |
Alex Elder | ea96571 | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 3111 | #ifdef CONFIG_BLOCK |
Alex Elder | 90af360 | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 3112 | void ceph_msg_data_add_bio(struct ceph_msg *msg, struct bio *bio, |
Alex Elder | a193080 | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 3113 | size_t length) |
Alex Elder | 27fa838 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 3114 | { |
Alex Elder | 6644ed7 | 2013-03-11 23:34:24 -0500 | [diff] [blame] | 3115 | struct ceph_msg_data *data; |
Alex Elder | 27fa838 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 3116 | |
Alex Elder | 6644ed7 | 2013-03-11 23:34:24 -0500 | [diff] [blame] | 3117 | BUG_ON(!bio); |
Alex Elder | 6644ed7 | 2013-03-11 23:34:24 -0500 | [diff] [blame] | 3118 | |
| 3119 | data = ceph_msg_data_create(CEPH_MSG_DATA_BIO); |
| 3120 | BUG_ON(!data); |
| 3121 | data->bio = bio; |
Alex Elder | c851c49 | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 3122 | data->bio_length = length; |
Alex Elder | 6644ed7 | 2013-03-11 23:34:24 -0500 | [diff] [blame] | 3123 | |
Alex Elder | 5240d9f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 3124 | list_add_tail(&data->links, &msg->data); |
| 3125 | msg->data_length += length; |
Alex Elder | 27fa838 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 3126 | } |
Alex Elder | 90af360 | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 3127 | EXPORT_SYMBOL(ceph_msg_data_add_bio); |
Alex Elder | ea96571 | 2013-04-05 14:46:01 -0500 | [diff] [blame] | 3128 | #endif /* CONFIG_BLOCK */ |
Alex Elder | 27fa838 | 2013-02-14 12:16:43 -0600 | [diff] [blame] | 3129 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3130 | /* |
| 3131 | * construct a new message with given type, size |
| 3132 | * the new msg has a ref count of 1. |
| 3133 | */ |
Sage Weil | b61c276 | 2011-08-09 15:03:46 -0700 | [diff] [blame] | 3134 | struct ceph_msg *ceph_msg_new(int type, int front_len, gfp_t flags, |
| 3135 | bool can_fail) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3136 | { |
| 3137 | struct ceph_msg *m; |
| 3138 | |
Alex Elder | e3d5d63 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 3139 | m = kmem_cache_zalloc(ceph_msg_cache, flags); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3140 | if (m == NULL) |
| 3141 | goto out; |
Alex Elder | 38941f8 | 2012-06-01 14:56:43 -0500 | [diff] [blame] | 3142 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3143 | m->hdr.type = cpu_to_le16(type); |
Sage Weil | 45c6ceb | 2010-05-11 15:01:51 -0700 | [diff] [blame] | 3144 | m->hdr.priority = cpu_to_le16(CEPH_MSG_PRIO_DEFAULT); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3145 | m->hdr.front_len = cpu_to_le32(front_len); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3146 | |
Alex Elder | 9516e45 | 2013-03-01 18:00:16 -0600 | [diff] [blame] | 3147 | INIT_LIST_HEAD(&m->list_head); |
| 3148 | kref_init(&m->kref); |
Alex Elder | 5240d9f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 3149 | INIT_LIST_HEAD(&m->data); |
Henry C Chang | ca20892 | 2011-05-03 02:29:56 +0000 | [diff] [blame] | 3150 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3151 | /* front */ |
| 3152 | if (front_len) { |
Ilya Dryomov | eeb0bed | 2014-01-09 20:08:21 +0200 | [diff] [blame] | 3153 | m->front.iov_base = ceph_kvmalloc(front_len, flags); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3154 | if (m->front.iov_base == NULL) { |
Sage Weil | b61c276 | 2011-08-09 15:03:46 -0700 | [diff] [blame] | 3155 | dout("ceph_msg_new can't allocate %d bytes\n", |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3156 | front_len); |
| 3157 | goto out2; |
| 3158 | } |
| 3159 | } else { |
| 3160 | m->front.iov_base = NULL; |
| 3161 | } |
Ilya Dryomov | f2be82b | 2014-01-09 20:08:21 +0200 | [diff] [blame] | 3162 | m->front_alloc_len = m->front.iov_len = front_len; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3163 | |
Sage Weil | bb25766 | 2010-04-01 16:07:23 -0700 | [diff] [blame] | 3164 | dout("ceph_msg_new %p front %d\n", m, front_len); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3165 | return m; |
| 3166 | |
| 3167 | out2: |
| 3168 | ceph_msg_put(m); |
| 3169 | out: |
Sage Weil | b61c276 | 2011-08-09 15:03:46 -0700 | [diff] [blame] | 3170 | if (!can_fail) { |
| 3171 | pr_err("msg_new can't create type %d front %d\n", type, |
| 3172 | front_len); |
Sage Weil | f0ed1b7 | 2011-08-09 15:05:07 -0700 | [diff] [blame] | 3173 | WARN_ON(1); |
Sage Weil | b61c276 | 2011-08-09 15:03:46 -0700 | [diff] [blame] | 3174 | } else { |
| 3175 | dout("msg_new can't create type %d front %d\n", type, |
| 3176 | front_len); |
| 3177 | } |
Sage Weil | a79832f | 2010-04-01 16:06:19 -0700 | [diff] [blame] | 3178 | return NULL; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3179 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 3180 | EXPORT_SYMBOL(ceph_msg_new); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3181 | |
| 3182 | /* |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3183 | * Allocate "middle" portion of a message, if it is needed and wasn't |
| 3184 | * allocated by alloc_msg. This allows us to read a small fixed-size |
| 3185 | * per-type header in the front and then gracefully fail (i.e., |
| 3186 | * propagate the error to the caller based on info in the front) when |
| 3187 | * the middle is too large. |
| 3188 | */ |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3189 | 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] | 3190 | { |
| 3191 | int type = le16_to_cpu(msg->hdr.type); |
| 3192 | int middle_len = le32_to_cpu(msg->hdr.middle_len); |
| 3193 | |
| 3194 | dout("alloc_middle %p type %d %s middle_len %d\n", msg, type, |
| 3195 | ceph_msg_type_name(type), middle_len); |
| 3196 | BUG_ON(!middle_len); |
| 3197 | BUG_ON(msg->middle); |
| 3198 | |
Sage Weil | b6c1d5b | 2009-12-07 12:17:17 -0800 | [diff] [blame] | 3199 | msg->middle = ceph_buffer_new(middle_len, GFP_NOFS); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3200 | if (!msg->middle) |
| 3201 | return -ENOMEM; |
| 3202 | return 0; |
| 3203 | } |
| 3204 | |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3205 | /* |
Alex Elder | 1c20f2d | 2012-06-04 14:43:32 -0500 | [diff] [blame] | 3206 | * Allocate a message for receiving an incoming message on a |
| 3207 | * connection, and save the result in con->in_msg. Uses the |
| 3208 | * connection's private alloc_msg op if available. |
| 3209 | * |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 3210 | * Returns 0 on success, or a negative error code. |
| 3211 | * |
| 3212 | * On success, if we set *skip = 1: |
| 3213 | * - the next message should be skipped and ignored. |
| 3214 | * - con->in_msg == NULL |
| 3215 | * or if we set *skip = 0: |
| 3216 | * - con->in_msg is non-null. |
| 3217 | * On error (ENOMEM, EAGAIN, ...), |
| 3218 | * - con->in_msg == NULL |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3219 | */ |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 3220 | 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] | 3221 | { |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 3222 | struct ceph_msg_header *hdr = &con->in_hdr; |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3223 | int middle_len = le32_to_cpu(hdr->middle_len); |
Alex Elder | 1d866d1 | 2013-03-01 18:00:14 -0600 | [diff] [blame] | 3224 | struct ceph_msg *msg; |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 3225 | int ret = 0; |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3226 | |
Alex Elder | 1c20f2d | 2012-06-04 14:43:32 -0500 | [diff] [blame] | 3227 | BUG_ON(con->in_msg != NULL); |
Alex Elder | 53ded49 | 2013-03-01 18:00:14 -0600 | [diff] [blame] | 3228 | BUG_ON(!con->ops->alloc_msg); |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3229 | |
Alex Elder | 53ded49 | 2013-03-01 18:00:14 -0600 | [diff] [blame] | 3230 | mutex_unlock(&con->mutex); |
| 3231 | msg = con->ops->alloc_msg(con, hdr, skip); |
| 3232 | mutex_lock(&con->mutex); |
| 3233 | if (con->state != CON_STATE_OPEN) { |
| 3234 | if (msg) |
Alex Elder | 1d866d1 | 2013-03-01 18:00:14 -0600 | [diff] [blame] | 3235 | ceph_msg_put(msg); |
Alex Elder | 53ded49 | 2013-03-01 18:00:14 -0600 | [diff] [blame] | 3236 | return -EAGAIN; |
| 3237 | } |
Alex Elder | 4137577 | 2013-03-05 09:25:10 -0600 | [diff] [blame] | 3238 | if (msg) { |
| 3239 | BUG_ON(*skip); |
| 3240 | con->in_msg = msg; |
Sage Weil | 36eb71a | 2012-06-21 12:47:08 -0700 | [diff] [blame] | 3241 | con->in_msg->con = con->ops->get(con); |
Alex Elder | 92ce034 | 2012-06-04 14:43:33 -0500 | [diff] [blame] | 3242 | BUG_ON(con->in_msg->con == NULL); |
Alex Elder | 4137577 | 2013-03-05 09:25:10 -0600 | [diff] [blame] | 3243 | } else { |
| 3244 | /* |
| 3245 | * Null message pointer means either we should skip |
| 3246 | * this message or we couldn't allocate memory. The |
| 3247 | * former is not an error. |
| 3248 | */ |
| 3249 | if (*skip) |
| 3250 | return 0; |
| 3251 | con->error_msg = "error allocating memory for incoming message"; |
| 3252 | |
Alex Elder | 53ded49 | 2013-03-01 18:00:14 -0600 | [diff] [blame] | 3253 | return -ENOMEM; |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3254 | } |
Alex Elder | 1c20f2d | 2012-06-04 14:43:32 -0500 | [diff] [blame] | 3255 | memcpy(&con->in_msg->hdr, &con->in_hdr, sizeof(con->in_hdr)); |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3256 | |
Alex Elder | 1c20f2d | 2012-06-04 14:43:32 -0500 | [diff] [blame] | 3257 | if (middle_len && !con->in_msg->middle) { |
| 3258 | ret = ceph_alloc_middle(con, con->in_msg); |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3259 | if (ret < 0) { |
Alex Elder | 1c20f2d | 2012-06-04 14:43:32 -0500 | [diff] [blame] | 3260 | ceph_msg_put(con->in_msg); |
| 3261 | con->in_msg = NULL; |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3262 | } |
| 3263 | } |
Yehuda Sadeh | 9d7f0f1 | 2010-01-11 10:32:02 -0800 | [diff] [blame] | 3264 | |
Sage Weil | 4740a62 | 2012-07-30 18:19:30 -0700 | [diff] [blame] | 3265 | return ret; |
Yehuda Sadeh | 2450418 | 2010-01-08 13:58:34 -0800 | [diff] [blame] | 3266 | } |
| 3267 | |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3268 | |
| 3269 | /* |
| 3270 | * Free a generically kmalloc'd message. |
| 3271 | */ |
| 3272 | void ceph_msg_kfree(struct ceph_msg *m) |
| 3273 | { |
| 3274 | dout("msg_kfree %p\n", m); |
Ilya Dryomov | eeb0bed | 2014-01-09 20:08:21 +0200 | [diff] [blame] | 3275 | ceph_kvfree(m->front.iov_base); |
Alex Elder | e3d5d63 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 3276 | kmem_cache_free(ceph_msg_cache, m); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3277 | } |
| 3278 | |
| 3279 | /* |
| 3280 | * Drop a msg ref. Destroy as needed. |
| 3281 | */ |
Sage Weil | c2e552e | 2009-12-07 15:55:05 -0800 | [diff] [blame] | 3282 | void ceph_msg_last_put(struct kref *kref) |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3283 | { |
Sage Weil | c2e552e | 2009-12-07 15:55:05 -0800 | [diff] [blame] | 3284 | struct ceph_msg *m = container_of(kref, struct ceph_msg, kref); |
Alex Elder | 5240d9f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 3285 | LIST_HEAD(data); |
| 3286 | struct list_head *links; |
| 3287 | struct list_head *next; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3288 | |
Sage Weil | c2e552e | 2009-12-07 15:55:05 -0800 | [diff] [blame] | 3289 | dout("ceph_msg_put last one on %p\n", m); |
| 3290 | WARN_ON(!list_empty(&m->list_head)); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3291 | |
Sage Weil | c2e552e | 2009-12-07 15:55:05 -0800 | [diff] [blame] | 3292 | /* drop middle, data, if any */ |
| 3293 | if (m->middle) { |
| 3294 | ceph_buffer_put(m->middle); |
| 3295 | m->middle = NULL; |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3296 | } |
Alex Elder | 5240d9f | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 3297 | |
| 3298 | list_splice_init(&m->data, &data); |
| 3299 | list_for_each_safe(links, next, &data) { |
| 3300 | struct ceph_msg_data *data; |
| 3301 | |
| 3302 | data = list_entry(links, struct ceph_msg_data, links); |
| 3303 | list_del_init(links); |
| 3304 | ceph_msg_data_destroy(data); |
| 3305 | } |
Alex Elder | a193080 | 2013-03-14 14:09:06 -0500 | [diff] [blame] | 3306 | m->data_length = 0; |
Sage Weil | 58bb3b3 | 2009-12-23 12:12:31 -0800 | [diff] [blame] | 3307 | |
Sage Weil | c2e552e | 2009-12-07 15:55:05 -0800 | [diff] [blame] | 3308 | if (m->pool) |
| 3309 | ceph_msgpool_put(m->pool, m); |
| 3310 | else |
| 3311 | ceph_msg_kfree(m); |
Sage Weil | 31b8006 | 2009-10-06 11:31:13 -0700 | [diff] [blame] | 3312 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 3313 | EXPORT_SYMBOL(ceph_msg_last_put); |
Sage Weil | 9ec7cab | 2009-12-14 15:13:47 -0800 | [diff] [blame] | 3314 | |
| 3315 | void ceph_msg_dump(struct ceph_msg *msg) |
| 3316 | { |
Ilya Dryomov | 3cea4c3 | 2014-01-09 20:08:21 +0200 | [diff] [blame] | 3317 | pr_debug("msg_dump %p (front_alloc_len %d length %zd)\n", msg, |
| 3318 | msg->front_alloc_len, msg->data_length); |
Sage Weil | 9ec7cab | 2009-12-14 15:13:47 -0800 | [diff] [blame] | 3319 | print_hex_dump(KERN_DEBUG, "header: ", |
| 3320 | DUMP_PREFIX_OFFSET, 16, 1, |
| 3321 | &msg->hdr, sizeof(msg->hdr), true); |
| 3322 | print_hex_dump(KERN_DEBUG, " front: ", |
| 3323 | DUMP_PREFIX_OFFSET, 16, 1, |
| 3324 | msg->front.iov_base, msg->front.iov_len, true); |
| 3325 | if (msg->middle) |
| 3326 | print_hex_dump(KERN_DEBUG, "middle: ", |
| 3327 | DUMP_PREFIX_OFFSET, 16, 1, |
| 3328 | msg->middle->vec.iov_base, |
| 3329 | msg->middle->vec.iov_len, true); |
| 3330 | print_hex_dump(KERN_DEBUG, "footer: ", |
| 3331 | DUMP_PREFIX_OFFSET, 16, 1, |
| 3332 | &msg->footer, sizeof(msg->footer), true); |
| 3333 | } |
Yehuda Sadeh | 3d14c5d | 2010-04-06 15:14:15 -0700 | [diff] [blame] | 3334 | EXPORT_SYMBOL(ceph_msg_dump); |