1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 22 /* 23 * Copyright 2008 Sun Microsystems, Inc. All rights reserved. 24 * Use is subject to license terms. 25 */ 26 27 #ifndef _VDC_H 28 #define _VDC_H 29 30 #pragma ident "%Z%%M% %I% %E% SMI" 31 32 /* 33 * Virtual disk client implementation definitions 34 */ 35 36 #include <sys/sysmacros.h> 37 #include <sys/note.h> 38 39 #include <sys/ldc.h> 40 #include <sys/vio_mailbox.h> 41 #include <sys/vdsk_mailbox.h> 42 #include <sys/vdsk_common.h> 43 44 #ifdef __cplusplus 45 extern "C" { 46 #endif 47 48 #define VDC_DRIVER_NAME "vdc" 49 50 /* 51 * Bit-field values to indicate if parts of the vdc driver are initialised. 52 */ 53 #define VDC_SOFT_STATE 0x0001 54 #define VDC_LOCKS 0x0002 55 #define VDC_MINOR 0x0004 56 #define VDC_THREAD 0x0008 57 #define VDC_DRING_INIT 0x0010 /* The DRing was created */ 58 #define VDC_DRING_BOUND 0x0020 /* The DRing was bound to an LDC channel */ 59 #define VDC_DRING_LOCAL 0x0040 /* The local private DRing was allocated */ 60 #define VDC_DRING_ENTRY 0x0080 /* At least one DRing entry was initialised */ 61 #define VDC_DRING (VDC_DRING_INIT | VDC_DRING_BOUND | \ 62 VDC_DRING_LOCAL | VDC_DRING_ENTRY) 63 #define VDC_HANDSHAKE 0x0100 /* Indicates if a handshake is in progress */ 64 #define VDC_HANDSHAKE_STOP 0x0200 /* stop further handshakes */ 65 66 /* 67 * Definitions of MD nodes/properties. 68 */ 69 #define VDC_MD_CHAN_NAME "channel-endpoint" 70 #define VDC_MD_VDEV_NAME "virtual-device" 71 #define VDC_MD_PORT_NAME "virtual-device-port" 72 #define VDC_MD_DISK_NAME "disk" 73 #define VDC_MD_CFG_HDL "cfg-handle" 74 #define VDC_MD_TIMEOUT "vdc-timeout" 75 #define VDC_MD_ID "id" 76 77 /* 78 * Definition of actions to be carried out when processing the sequence ID 79 * of a message received from the vDisk server. The function verifying the 80 * sequence number checks the 'seq_num_xxx' fields in the soft state and 81 * returns whether the message should be processed (VDC_SEQ_NUM_TODO) or 82 * whether it was it was previously processed (VDC_SEQ_NUM_SKIP). 83 */ 84 #define VDC_SEQ_NUM_INVALID -1 /* Error */ 85 #define VDC_SEQ_NUM_SKIP 0 /* Request already processed */ 86 #define VDC_SEQ_NUM_TODO 1 /* Request needs processing */ 87 88 /* 89 * Macros to get UNIT and PART number 90 */ 91 #define VDCUNIT_SHIFT 3 92 #define VDCPART_MASK 7 93 94 #define VDCUNIT(dev) (getminor((dev)) >> VDCUNIT_SHIFT) 95 #define VDCPART(dev) (getminor((dev)) & VDCPART_MASK) 96 97 /* 98 * Scheme to store the instance number and the slice number in the minor number. 99 * (NOTE: Uses the same format and definitions as the sd(7D) driver) 100 */ 101 #define VD_MAKE_DEV(instance, minor) ((instance << VDCUNIT_SHIFT) | minor) 102 103 /* 104 * variables controlling how long to wait before timing out and how many 105 * retries to attempt before giving up when communicating with vds. 106 * 107 * These values need to be sufficiently large so that a guest can survive 108 * the reboot of the service domain. 109 */ 110 #define VDC_RETRIES 10 111 112 #define VDC_USEC_TIMEOUT_MIN (30 * MICROSEC) /* 30 sec */ 113 114 /* 115 * This macro returns the number of Hz that the vdc driver should wait before 116 * a timeout is triggered. The 'timeout' parameter specifiecs the wait 117 * time in Hz. The 'mul' parameter allows for a multiplier to be 118 * specified allowing for a backoff to be implemented (e.g. using the 119 * retry number as a multiplier) where the wait time will get longer if 120 * there is no response on the previous retry. 121 */ 122 #define VD_GET_TIMEOUT_HZ(timeout, mul) \ 123 (ddi_get_lbolt() + ((timeout) * MAX(1, (mul)))) 124 125 /* 126 * Macros to manipulate Descriptor Ring variables in the soft state 127 * structure. 128 */ 129 #define VDC_GET_NEXT_REQ_ID(vdc) ((vdc)->req_id++) 130 131 #define VDC_GET_DRING_ENTRY_PTR(vdc, idx) \ 132 (vd_dring_entry_t *)(uintptr_t)((vdc)->dring_mem_info.vaddr + \ 133 (idx * (vdc)->dring_entry_size)) 134 135 #define VDC_MARK_DRING_ENTRY_FREE(vdc, idx) \ 136 { \ 137 vd_dring_entry_t *dep = NULL; \ 138 ASSERT(vdc != NULL); \ 139 ASSERT(idx < vdc->dring_len); \ 140 ASSERT(vdc->dring_mem_info.vaddr != NULL); \ 141 dep = (vd_dring_entry_t *)(uintptr_t) \ 142 (vdc->dring_mem_info.vaddr + \ 143 (idx * vdc->dring_entry_size)); \ 144 ASSERT(dep != NULL); \ 145 dep->hdr.dstate = VIO_DESC_FREE; \ 146 } 147 148 /* Initialise the Session ID and Sequence Num in the DRing msg */ 149 #define VDC_INIT_DRING_DATA_MSG_IDS(dmsg, vdc) \ 150 ASSERT(vdc != NULL); \ 151 dmsg.tag.vio_sid = vdc->session_id; \ 152 dmsg.seq_num = vdc->seq_num; 153 154 /* 155 * The states that the read thread can be in. 156 */ 157 typedef enum vdc_rd_state { 158 VDC_READ_IDLE, /* idling - conn is not up */ 159 VDC_READ_WAITING, /* waiting for data */ 160 VDC_READ_PENDING, /* pending data avail for read */ 161 VDC_READ_RESET /* channel was reset - stop reads */ 162 } vdc_rd_state_t; 163 164 /* 165 * The states that the vdc-vds connection can be in. 166 */ 167 typedef enum vdc_state { 168 VDC_STATE_INIT, /* device is initialized */ 169 VDC_STATE_INIT_WAITING, /* waiting for ldc connection */ 170 VDC_STATE_NEGOTIATE, /* doing handshake negotiation */ 171 VDC_STATE_HANDLE_PENDING, /* handle requests in backup dring */ 172 VDC_STATE_RUNNING, /* running and accepting requests */ 173 VDC_STATE_DETACH, /* detaching */ 174 VDC_STATE_RESETTING /* resetting connection with vds */ 175 } vdc_state_t; 176 177 /* 178 * The states that the vdc instance can be in. 179 */ 180 typedef enum vdc_lc_state { 181 VDC_LC_ATTACHING, /* driver is attaching */ 182 VDC_LC_ONLINE, /* driver is attached and online */ 183 VDC_LC_DETACHING /* driver is detaching */ 184 } vdc_lc_state_t; 185 186 /* 187 * Local Descriptor Ring entry 188 * 189 * vdc creates a Local (private) descriptor ring the same size as the 190 * public descriptor ring it exports to vds. 191 */ 192 193 typedef enum { 194 VIO_read_dir, /* read data from server */ 195 VIO_write_dir, /* write data to server */ 196 VIO_both_dir /* transfer both in and out in same buffer */ 197 } vio_desc_direction_t; 198 199 typedef enum { 200 CB_STRATEGY, /* non-blocking strategy call */ 201 CB_SYNC /* synchronous operation */ 202 } vio_cb_type_t; 203 204 typedef struct vdc_local_desc { 205 boolean_t is_free; /* local state - inuse or not */ 206 207 int operation; /* VD_OP_xxx to be performed */ 208 caddr_t addr; /* addr passed in by consumer */ 209 int slice; 210 diskaddr_t offset; /* disk offset */ 211 size_t nbytes; 212 vio_cb_type_t cb_type; /* operation type blk/nonblk */ 213 void *cb_arg; /* buf passed to strategy() */ 214 vio_desc_direction_t dir; /* direction of transfer */ 215 216 caddr_t align_addr; /* used if addr non-aligned */ 217 ldc_mem_handle_t desc_mhdl; /* Mem handle of buf */ 218 vd_dring_entry_t *dep; /* public Dring Entry Pointer */ 219 220 } vdc_local_desc_t; 221 222 /* 223 * I/O queue used by failfast 224 */ 225 typedef struct vdc_io { 226 struct vdc_io *vio_next; /* next pending I/O in the queue */ 227 struct buf *vio_buf; /* buf for CB_STRATEGY I/O */ 228 clock_t vio_qtime; /* time the I/O was queued */ 229 } vdc_io_t; 230 231 /* 232 * Per vDisk server channel states 233 */ 234 #define VDC_LDC_INIT 0x0001 235 #define VDC_LDC_CB 0x0002 236 #define VDC_LDC_OPEN 0x0004 237 #define VDC_LDC (VDC_LDC_INIT | VDC_LDC_CB | VDC_LDC_OPEN) 238 239 /* 240 * vDisk server information 241 */ 242 typedef struct vdc_server { 243 struct vdc_server *next; /* Next server */ 244 struct vdc *vdcp; /* Ptr to vdc struct */ 245 uint64_t id; /* Server port id */ 246 uint64_t state; /* Server state */ 247 uint64_t ldc_id; /* Server LDC id */ 248 ldc_handle_t ldc_handle; /* Server LDC handle */ 249 ldc_status_t ldc_state; /* Server LDC state */ 250 uint64_t ctimeout; /* conn tmout (secs) */ 251 } vdc_server_t; 252 253 /* 254 * vdc soft state structure 255 */ 256 typedef struct vdc { 257 258 kmutex_t lock; /* protects next 2 sections of vars */ 259 kcondvar_t running_cv; /* signal when upper layers can send */ 260 kcondvar_t initwait_cv; /* signal when ldc conn is up */ 261 kcondvar_t dring_free_cv; /* signal when desc is avail */ 262 kcondvar_t membind_cv; /* signal when mem can be bound */ 263 boolean_t self_reset; 264 265 int initialized; /* keeps track of what's init'ed */ 266 vdc_lc_state_t lifecycle; /* Current state of the vdc instance */ 267 268 int hshake_cnt; /* number of failed handshakes */ 269 uint8_t open[OTYPCNT]; /* mask of opened slices */ 270 uint8_t open_excl; /* mask of exclusively opened slices */ 271 ulong_t open_lyr[V_NUMPAR]; /* number of layered opens */ 272 int dkio_flush_pending; /* # outstanding DKIO flushes */ 273 int validate_pending; /* # outstanding validate request */ 274 vd_disk_label_t vdisk_label; /* label type of device/disk imported */ 275 struct vtoc *vtoc; /* structure to store VTOC data */ 276 struct dk_geom *geom; /* structure to store geometry data */ 277 vd_slice_t slice[V_NUMPAR]; /* logical partitions */ 278 279 kthread_t *msg_proc_thr; /* main msg processing thread */ 280 281 kmutex_t read_lock; /* lock to protect read */ 282 kcondvar_t read_cv; /* cv to wait for READ events */ 283 vdc_rd_state_t read_state; /* current read state */ 284 285 uint32_t sync_op_cnt; /* num of active sync operations */ 286 boolean_t sync_op_pending; /* sync operation is pending */ 287 boolean_t sync_op_blocked; /* blocked waiting to do sync op */ 288 uint32_t sync_op_status; /* status of sync operation */ 289 kcondvar_t sync_pending_cv; /* cv wait for sync op to finish */ 290 kcondvar_t sync_blocked_cv; /* cv wait for other syncs to finish */ 291 292 uint64_t session_id; /* common ID sent with all messages */ 293 uint64_t seq_num; /* most recent sequence num generated */ 294 uint64_t seq_num_reply; /* Last seq num ACK/NACK'ed by vds */ 295 uint64_t req_id; /* Most recent Request ID generated */ 296 uint64_t req_id_proc; /* Last request ID processed by vdc */ 297 vdc_state_t state; /* Current disk client-server state */ 298 299 dev_info_t *dip; /* device info pointer */ 300 int instance; /* driver instance number */ 301 302 vio_ver_t ver; /* version number agreed with server */ 303 vd_disk_type_t vdisk_type; /* type of device/disk being imported */ 304 uint32_t vdisk_media; /* physical media type of vDisk */ 305 uint64_t vdisk_size; /* device size in blocks */ 306 uint64_t max_xfer_sz; /* maximum block size of a descriptor */ 307 uint64_t block_size; /* device block size used */ 308 uint64_t operations; /* bitmask of ops. server supports */ 309 struct dk_cinfo *cinfo; /* structure to store DKIOCINFO data */ 310 struct dk_minfo *minfo; /* structure for DKIOCGMEDIAINFO data */ 311 ddi_devid_t devid; /* device id */ 312 boolean_t ctimeout_reached; /* connection timeout has expired */ 313 314 /* 315 * The ownership fields are protected by the lock mutex. The 316 * ownership_lock mutex is used to serialize ownership operations; 317 * it should be acquired before the lock mutex. 318 */ 319 kmutex_t ownership_lock; /* serialize ownership ops */ 320 int ownership; /* ownership status flags */ 321 kthread_t *ownership_thread; /* ownership thread */ 322 kcondvar_t ownership_cv; /* cv for ownership update */ 323 324 /* 325 * The failfast fields are protected by the lock mutex. 326 */ 327 kthread_t *failfast_thread; /* failfast thread */ 328 clock_t failfast_interval; /* interval in microsecs */ 329 kcondvar_t failfast_cv; /* cv for failfast update */ 330 kcondvar_t failfast_io_cv; /* cv wait for I/O to finish */ 331 vdc_io_t *failfast_io_queue; /* failfast io queue */ 332 333 /* 334 * kstats used to store I/O statistics consumed by iostat(1M). 335 * These are protected by the lock mutex. 336 */ 337 kstat_t *io_stats; 338 kstat_t *err_stats; 339 340 ldc_dring_handle_t dring_hdl; /* dring handle */ 341 ldc_mem_info_t dring_mem_info; /* dring information */ 342 uint_t dring_curr_idx; /* current index */ 343 uint32_t dring_len; /* dring length */ 344 uint32_t dring_max_cookies; /* dring max cookies */ 345 uint32_t dring_cookie_count; /* num cookies */ 346 uint32_t dring_entry_size; /* descriptor size */ 347 ldc_mem_cookie_t *dring_cookie; /* dring cookies */ 348 uint64_t dring_ident; /* dring ident */ 349 350 uint64_t threads_pending; /* num of threads */ 351 352 vdc_local_desc_t *local_dring; /* local dring */ 353 vdc_local_desc_t *local_dring_backup; /* local dring backup */ 354 int local_dring_backup_tail; /* backup dring tail */ 355 int local_dring_backup_len; /* backup dring len */ 356 357 int num_servers; /* no. of servers */ 358 vdc_server_t *server_list; /* vdisk server list */ 359 vdc_server_t *curr_server; /* curr vdisk server */ 360 } vdc_t; 361 362 /* 363 * Ownership status flags 364 */ 365 #define VDC_OWNERSHIP_NONE 0x00 /* no ownership wanted */ 366 #define VDC_OWNERSHIP_WANTED 0x01 /* ownership is wanted */ 367 #define VDC_OWNERSHIP_GRANTED 0x02 /* ownership has been granted */ 368 #define VDC_OWNERSHIP_RESET 0x04 /* ownership has been reset */ 369 370 /* 371 * Reservation conflict panic message 372 */ 373 #define VDC_RESV_CONFLICT_FMT_STR "Reservation Conflict\nDisk: " 374 #define VDC_RESV_CONFLICT_FMT_LEN (sizeof (VDC_RESV_CONFLICT_FMT_STR)) 375 376 /* 377 * Debugging macros 378 */ 379 #ifdef DEBUG 380 extern int vdc_msglevel; 381 extern uint64_t vdc_matchinst; 382 383 #define DMSG(_vdc, err_level, format, ...) \ 384 do { \ 385 if (vdc_msglevel > err_level && \ 386 (vdc_matchinst & (1ull << (_vdc)->instance))) \ 387 cmn_err(CE_CONT, "?[%d,t@%p] %s: "format, \ 388 (_vdc)->instance, (void *)curthread, \ 389 __func__, __VA_ARGS__); \ 390 _NOTE(CONSTANTCONDITION) \ 391 } while (0); 392 393 #define DMSGX(err_level, format, ...) \ 394 do { \ 395 if (vdc_msglevel > err_level) \ 396 cmn_err(CE_CONT, "?%s: "format, __func__, __VA_ARGS__);\ 397 _NOTE(CONSTANTCONDITION) \ 398 } while (0); 399 400 #define VDC_DUMP_DRING_MSG(dmsgp) \ 401 DMSGX(0, "sq:%lu start:%d end:%d ident:%lu\n", \ 402 dmsgp->seq_num, dmsgp->start_idx, \ 403 dmsgp->end_idx, dmsgp->dring_ident); 404 405 #else /* !DEBUG */ 406 #define DMSG(err_level, ...) 407 #define DMSGX(err_level, format, ...) 408 #define VDC_DUMP_DRING_MSG(dmsgp) 409 410 #endif /* !DEBUG */ 411 412 #ifdef __cplusplus 413 } 414 #endif 415 416 #endif /* _VDC_H */ 417