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 strings to be used to create device node properties. 68 * (vdc uses the capitalised versions of these properties as they are 64-bit) 69 */ 70 #define VDC_NBLOCKS_PROP_NAME "Nblocks" 71 #define VDC_SIZE_PROP_NAME "Size" 72 73 /* 74 * Definitions of MD nodes/properties. 75 */ 76 #define VDC_MD_CHAN_NAME "channel-endpoint" 77 #define VDC_MD_VDEV_NAME "virtual-device" 78 #define VDC_MD_PORT_NAME "virtual-device-port" 79 #define VDC_MD_DISK_NAME "disk" 80 #define VDC_MD_CFG_HDL "cfg-handle" 81 #define VDC_MD_TIMEOUT "vdc-timeout" 82 #define VDC_MD_ID "id" 83 84 /* 85 * Definition of actions to be carried out when processing the sequence ID 86 * of a message received from the vDisk server. The function verifying the 87 * sequence number checks the 'seq_num_xxx' fields in the soft state and 88 * returns whether the message should be processed (VDC_SEQ_NUM_TODO) or 89 * whether it was it was previously processed (VDC_SEQ_NUM_SKIP). 90 */ 91 #define VDC_SEQ_NUM_INVALID -1 /* Error */ 92 #define VDC_SEQ_NUM_SKIP 0 /* Request already processed */ 93 #define VDC_SEQ_NUM_TODO 1 /* Request needs processing */ 94 95 /* 96 * Macros to get UNIT and PART number 97 */ 98 #define VDCUNIT_SHIFT 3 99 #define VDCPART_MASK 7 100 101 #define VDCUNIT(dev) (getminor((dev)) >> VDCUNIT_SHIFT) 102 #define VDCPART(dev) (getminor((dev)) & VDCPART_MASK) 103 104 /* 105 * Scheme to store the instance number and the slice number in the minor number. 106 * (NOTE: Uses the same format and definitions as the sd(7D) driver) 107 */ 108 #define VD_MAKE_DEV(instance, minor) ((instance << VDCUNIT_SHIFT) | minor) 109 110 /* 111 * variables controlling how long to wait before timing out and how many 112 * retries to attempt before giving up when communicating with vds. 113 * 114 * These values need to be sufficiently large so that a guest can survive 115 * the reboot of the service domain. 116 */ 117 #define VDC_RETRIES 10 118 119 #define VDC_USEC_TIMEOUT_MIN (30 * MICROSEC) /* 30 sec */ 120 121 /* 122 * This macro returns the number of Hz that the vdc driver should wait before 123 * a timeout is triggered. The 'timeout' parameter specifiecs the wait 124 * time in Hz. The 'mul' parameter allows for a multiplier to be 125 * specified allowing for a backoff to be implemented (e.g. using the 126 * retry number as a multiplier) where the wait time will get longer if 127 * there is no response on the previous retry. 128 */ 129 #define VD_GET_TIMEOUT_HZ(timeout, mul) \ 130 (ddi_get_lbolt() + ((timeout) * MAX(1, (mul)))) 131 132 /* 133 * Macros to manipulate Descriptor Ring variables in the soft state 134 * structure. 135 */ 136 #define VDC_GET_NEXT_REQ_ID(vdc) ((vdc)->req_id++) 137 138 #define VDC_GET_DRING_ENTRY_PTR(vdc, idx) \ 139 (vd_dring_entry_t *)(uintptr_t)((vdc)->dring_mem_info.vaddr + \ 140 (idx * (vdc)->dring_entry_size)) 141 142 #define VDC_MARK_DRING_ENTRY_FREE(vdc, idx) \ 143 { \ 144 vd_dring_entry_t *dep = NULL; \ 145 ASSERT(vdc != NULL); \ 146 ASSERT(idx < vdc->dring_len); \ 147 ASSERT(vdc->dring_mem_info.vaddr != NULL); \ 148 dep = (vd_dring_entry_t *)(uintptr_t) \ 149 (vdc->dring_mem_info.vaddr + \ 150 (idx * vdc->dring_entry_size)); \ 151 ASSERT(dep != NULL); \ 152 dep->hdr.dstate = VIO_DESC_FREE; \ 153 } 154 155 /* Initialise the Session ID and Sequence Num in the DRing msg */ 156 #define VDC_INIT_DRING_DATA_MSG_IDS(dmsg, vdc) \ 157 ASSERT(vdc != NULL); \ 158 dmsg.tag.vio_sid = vdc->session_id; \ 159 dmsg.seq_num = vdc->seq_num; 160 161 /* 162 * The states that the read thread can be in. 163 */ 164 typedef enum vdc_rd_state { 165 VDC_READ_IDLE, /* idling - conn is not up */ 166 VDC_READ_WAITING, /* waiting for data */ 167 VDC_READ_PENDING, /* pending data avail for read */ 168 VDC_READ_RESET /* channel was reset - stop reads */ 169 } vdc_rd_state_t; 170 171 /* 172 * The states that the vdc-vds connection can be in. 173 */ 174 typedef enum vdc_state { 175 VDC_STATE_INIT, /* device is initialized */ 176 VDC_STATE_INIT_WAITING, /* waiting for ldc connection */ 177 VDC_STATE_NEGOTIATE, /* doing handshake negotiation */ 178 VDC_STATE_HANDLE_PENDING, /* handle requests in backup dring */ 179 VDC_STATE_RUNNING, /* running and accepting requests */ 180 VDC_STATE_DETACH, /* detaching */ 181 VDC_STATE_RESETTING /* resetting connection with vds */ 182 } vdc_state_t; 183 184 /* 185 * The states that the vdc instance can be in. 186 */ 187 typedef enum vdc_lc_state { 188 VDC_LC_ATTACHING, /* driver is attaching */ 189 VDC_LC_ONLINE, /* driver is attached and online */ 190 VDC_LC_DETACHING /* driver is detaching */ 191 } vdc_lc_state_t; 192 193 /* 194 * Local Descriptor Ring entry 195 * 196 * vdc creates a Local (private) descriptor ring the same size as the 197 * public descriptor ring it exports to vds. 198 */ 199 200 typedef enum { 201 VIO_read_dir, /* read data from server */ 202 VIO_write_dir, /* write data to server */ 203 VIO_both_dir /* transfer both in and out in same buffer */ 204 } vio_desc_direction_t; 205 206 typedef enum { 207 CB_STRATEGY, /* non-blocking strategy call */ 208 CB_SYNC /* synchronous operation */ 209 } vio_cb_type_t; 210 211 typedef struct vdc_local_desc { 212 boolean_t is_free; /* local state - inuse or not */ 213 214 int operation; /* VD_OP_xxx to be performed */ 215 caddr_t addr; /* addr passed in by consumer */ 216 int slice; 217 diskaddr_t offset; /* disk offset */ 218 size_t nbytes; 219 vio_cb_type_t cb_type; /* operation type blk/nonblk */ 220 void *cb_arg; /* buf passed to strategy() */ 221 vio_desc_direction_t dir; /* direction of transfer */ 222 223 caddr_t align_addr; /* used if addr non-aligned */ 224 ldc_mem_handle_t desc_mhdl; /* Mem handle of buf */ 225 vd_dring_entry_t *dep; /* public Dring Entry Pointer */ 226 227 } vdc_local_desc_t; 228 229 /* 230 * I/O queue used by failfast 231 */ 232 typedef struct vdc_io { 233 struct vdc_io *vio_next; /* next pending I/O in the queue */ 234 struct buf *vio_buf; /* buf for CB_STRATEGY I/O */ 235 clock_t vio_qtime; /* time the I/O was queued */ 236 } vdc_io_t; 237 238 /* 239 * Per vDisk server channel states 240 */ 241 #define VDC_LDC_INIT 0x0001 242 #define VDC_LDC_CB 0x0002 243 #define VDC_LDC_OPEN 0x0004 244 #define VDC_LDC (VDC_LDC_INIT | VDC_LDC_CB | VDC_LDC_OPEN) 245 246 /* 247 * vDisk server information 248 */ 249 typedef struct vdc_server { 250 struct vdc_server *next; /* Next server */ 251 struct vdc *vdcp; /* Ptr to vdc struct */ 252 uint64_t id; /* Server port id */ 253 uint64_t state; /* Server state */ 254 uint64_t ldc_id; /* Server LDC id */ 255 ldc_handle_t ldc_handle; /* Server LDC handle */ 256 ldc_status_t ldc_state; /* Server LDC state */ 257 uint64_t ctimeout; /* conn tmout (secs) */ 258 } vdc_server_t; 259 260 /* 261 * vdc soft state structure 262 */ 263 typedef struct vdc { 264 265 kmutex_t lock; /* protects next 2 sections of vars */ 266 kcondvar_t running_cv; /* signal when upper layers can send */ 267 kcondvar_t initwait_cv; /* signal when ldc conn is up */ 268 kcondvar_t dring_free_cv; /* signal when desc is avail */ 269 kcondvar_t membind_cv; /* signal when mem can be bound */ 270 boolean_t self_reset; 271 272 int initialized; /* keeps track of what's init'ed */ 273 vdc_lc_state_t lifecycle; /* Current state of the vdc instance */ 274 275 int hshake_cnt; /* number of failed handshakes */ 276 uint8_t open[OTYPCNT]; /* mask of opened slices */ 277 uint8_t open_excl; /* mask of exclusively opened slices */ 278 ulong_t open_lyr[V_NUMPAR]; /* number of layered opens */ 279 int dkio_flush_pending; /* # outstanding DKIO flushes */ 280 int validate_pending; /* # outstanding validate request */ 281 vd_disk_label_t vdisk_label; /* label type of device/disk imported */ 282 struct vtoc *vtoc; /* structure to store VTOC data */ 283 struct dk_geom *geom; /* structure to store geometry data */ 284 vd_slice_t slice[V_NUMPAR]; /* logical partitions */ 285 286 kthread_t *msg_proc_thr; /* main msg processing thread */ 287 288 kmutex_t read_lock; /* lock to protect read */ 289 kcondvar_t read_cv; /* cv to wait for READ events */ 290 vdc_rd_state_t read_state; /* current read state */ 291 292 uint32_t sync_op_cnt; /* num of active sync operations */ 293 boolean_t sync_op_pending; /* sync operation is pending */ 294 boolean_t sync_op_blocked; /* blocked waiting to do sync op */ 295 uint32_t sync_op_status; /* status of sync operation */ 296 kcondvar_t sync_pending_cv; /* cv wait for sync op to finish */ 297 kcondvar_t sync_blocked_cv; /* cv wait for other syncs to finish */ 298 299 uint64_t session_id; /* common ID sent with all messages */ 300 uint64_t seq_num; /* most recent sequence num generated */ 301 uint64_t seq_num_reply; /* Last seq num ACK/NACK'ed by vds */ 302 uint64_t req_id; /* Most recent Request ID generated */ 303 uint64_t req_id_proc; /* Last request ID processed by vdc */ 304 vdc_state_t state; /* Current disk client-server state */ 305 306 dev_info_t *dip; /* device info pointer */ 307 int instance; /* driver instance number */ 308 309 vio_ver_t ver; /* version number agreed with server */ 310 vd_disk_type_t vdisk_type; /* type of device/disk being imported */ 311 uint32_t vdisk_media; /* physical media type of vDisk */ 312 uint64_t vdisk_size; /* device size in blocks */ 313 uint64_t max_xfer_sz; /* maximum block size of a descriptor */ 314 uint64_t block_size; /* device block size used */ 315 uint64_t operations; /* bitmask of ops. server supports */ 316 struct dk_cinfo *cinfo; /* structure to store DKIOCINFO data */ 317 struct dk_minfo *minfo; /* structure for DKIOCGMEDIAINFO data */ 318 ddi_devid_t devid; /* device id */ 319 boolean_t ctimeout_reached; /* connection timeout has expired */ 320 321 /* 322 * The ownership fields are protected by the lock mutex. The 323 * ownership_lock mutex is used to serialize ownership operations; 324 * it should be acquired before the lock mutex. 325 */ 326 kmutex_t ownership_lock; /* serialize ownership ops */ 327 int ownership; /* ownership status flags */ 328 kthread_t *ownership_thread; /* ownership thread */ 329 kcondvar_t ownership_cv; /* cv for ownership update */ 330 331 /* 332 * The failfast fields are protected by the lock mutex. 333 */ 334 kthread_t *failfast_thread; /* failfast thread */ 335 clock_t failfast_interval; /* interval in microsecs */ 336 kcondvar_t failfast_cv; /* cv for failfast update */ 337 kcondvar_t failfast_io_cv; /* cv wait for I/O to finish */ 338 vdc_io_t *failfast_io_queue; /* failfast io queue */ 339 340 /* 341 * kstats used to store I/O statistics consumed by iostat(1M). 342 * These are protected by the lock mutex. 343 */ 344 kstat_t *io_stats; 345 kstat_t *err_stats; 346 347 ldc_dring_handle_t dring_hdl; /* dring handle */ 348 ldc_mem_info_t dring_mem_info; /* dring information */ 349 uint_t dring_curr_idx; /* current index */ 350 uint32_t dring_len; /* dring length */ 351 uint32_t dring_max_cookies; /* dring max cookies */ 352 uint32_t dring_cookie_count; /* num cookies */ 353 uint32_t dring_entry_size; /* descriptor size */ 354 ldc_mem_cookie_t *dring_cookie; /* dring cookies */ 355 uint64_t dring_ident; /* dring ident */ 356 357 uint64_t threads_pending; /* num of threads */ 358 359 vdc_local_desc_t *local_dring; /* local dring */ 360 vdc_local_desc_t *local_dring_backup; /* local dring backup */ 361 int local_dring_backup_tail; /* backup dring tail */ 362 int local_dring_backup_len; /* backup dring len */ 363 364 int num_servers; /* no. of servers */ 365 vdc_server_t *server_list; /* vdisk server list */ 366 vdc_server_t *curr_server; /* curr vdisk server */ 367 } vdc_t; 368 369 /* 370 * Ownership status flags 371 */ 372 #define VDC_OWNERSHIP_NONE 0x00 /* no ownership wanted */ 373 #define VDC_OWNERSHIP_WANTED 0x01 /* ownership is wanted */ 374 #define VDC_OWNERSHIP_GRANTED 0x02 /* ownership has been granted */ 375 #define VDC_OWNERSHIP_RESET 0x04 /* ownership has been reset */ 376 377 /* 378 * Reservation conflict panic message 379 */ 380 #define VDC_RESV_CONFLICT_FMT_STR "Reservation Conflict\nDisk: " 381 #define VDC_RESV_CONFLICT_FMT_LEN (sizeof (VDC_RESV_CONFLICT_FMT_STR)) 382 383 /* 384 * Debugging macros 385 */ 386 #ifdef DEBUG 387 extern int vdc_msglevel; 388 extern uint64_t vdc_matchinst; 389 390 #define DMSG(_vdc, err_level, format, ...) \ 391 do { \ 392 if (vdc_msglevel > err_level && \ 393 (vdc_matchinst & (1ull << (_vdc)->instance))) \ 394 cmn_err(CE_CONT, "?[%d,t@%p] %s: "format, \ 395 (_vdc)->instance, (void *)curthread, \ 396 __func__, __VA_ARGS__); \ 397 _NOTE(CONSTANTCONDITION) \ 398 } while (0); 399 400 #define DMSGX(err_level, format, ...) \ 401 do { \ 402 if (vdc_msglevel > err_level) \ 403 cmn_err(CE_CONT, "?%s: "format, __func__, __VA_ARGS__);\ 404 _NOTE(CONSTANTCONDITION) \ 405 } while (0); 406 407 #define VDC_DUMP_DRING_MSG(dmsgp) \ 408 DMSGX(0, "sq:%lu start:%d end:%d ident:%lu\n", \ 409 dmsgp->seq_num, dmsgp->start_idx, \ 410 dmsgp->end_idx, dmsgp->dring_ident); 411 412 #else /* !DEBUG */ 413 #define DMSG(err_level, ...) 414 #define DMSGX(err_level, format, ...) 415 #define VDC_DUMP_DRING_MSG(dmsgp) 416 417 #endif /* !DEBUG */ 418 419 #ifdef __cplusplus 420 } 421 #endif 422 423 #endif /* _VDC_H */ 424