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 2006 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_LDC 0x0010 58 #define VDC_LDC_INIT 0x0020 59 #define VDC_LDC_CB 0x0040 60 #define VDC_LDC_OPEN 0x0080 61 #define VDC_DRING_INIT 0x0100 /* The DRing was created */ 62 #define VDC_DRING_BOUND 0x0200 /* The DRing was bound to an LDC channel */ 63 #define VDC_DRING_LOCAL 0x0400 /* The local private DRing was allocated */ 64 #define VDC_DRING_ENTRY 0x0800 /* At least one DRing entry was initialised */ 65 #define VDC_DRING (VDC_DRING_INIT | VDC_DRING_BOUND | \ 66 VDC_DRING_LOCAL | VDC_DRING_ENTRY) 67 #define VDC_HANDSHAKE 0x1000 /* Indicates if a handshake is in progress */ 68 #define VDC_HANDSHAKE_STOP 0x2000 /* stop further handshakes */ 69 70 /* 71 * Definitions of strings to be used to create device node properties. 72 * (vdc uses the capitalised versions of these properties as they are 64-bit) 73 */ 74 #define VDC_NBLOCKS_PROP_NAME "Nblocks" 75 #define VDC_SIZE_PROP_NAME "Size" 76 77 /* 78 * Definitions of MD nodes/properties. 79 */ 80 #define VDC_MD_CHAN_NAME "channel-endpoint" 81 #define VDC_MD_VDEV_NAME "virtual-device" 82 #define VDC_MD_DISK_NAME "disk" 83 #define VDC_MD_CFG_HDL "cfg-handle" 84 #define VDC_ID_PROP "id" 85 86 /* 87 * Definition of actions to be carried out when processing the sequence ID 88 * of a message received from the vDisk server. The function verifying the 89 * sequence number checks the 'seq_num_xxx' fields in the soft state and 90 * returns whether the message should be processed (VDC_SEQ_NUM_TODO) or 91 * whether it was it was previously processed (VDC_SEQ_NUM_SKIP). 92 */ 93 #define VDC_SEQ_NUM_INVALID -1 /* Error */ 94 #define VDC_SEQ_NUM_SKIP 0 /* Request already processed */ 95 #define VDC_SEQ_NUM_TODO 1 /* Request needs processing */ 96 97 /* 98 * Scheme to store the instance number and the slice number in the minor number. 99 * (Uses the same format and definitions as the sd(7D) driver) 100 */ 101 #define VD_MAKE_DEV(instance, minor) ((instance << SDUNIT_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 *)((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 >= 0) && (idx < vdc->dring_len)); \ 140 ASSERT(vdc->dring_mem_info.vaddr != NULL); \ 141 dep = (vd_dring_entry_t *)(vdc->dring_mem_info.vaddr + \ 142 (idx * vdc->dring_entry_size)); \ 143 ASSERT(dep != NULL); \ 144 dep->hdr.dstate = VIO_DESC_FREE; \ 145 } 146 147 /* Initialise the Session ID and Sequence Num in the DRing msg */ 148 #define VDC_INIT_DRING_DATA_MSG_IDS(dmsg, vdc) \ 149 ASSERT(vdc != NULL); \ 150 dmsg.tag.vio_sid = vdc->session_id; \ 151 dmsg.seq_num = vdc->seq_num; 152 153 /* 154 * The states that the read thread can be in. 155 */ 156 typedef enum vdc_rd_state { 157 VDC_READ_IDLE, /* idling - conn is not up */ 158 VDC_READ_WAITING, /* waiting for data */ 159 VDC_READ_PENDING, /* pending data avail for read */ 160 VDC_READ_RESET /* channel was reset - stop reads */ 161 } vdc_rd_state_t; 162 163 /* 164 * The states that the vdc-vds connection can be in. 165 */ 166 typedef enum vdc_state { 167 VDC_STATE_INIT, /* device is initialized */ 168 VDC_STATE_INIT_WAITING, /* waiting for ldc connection */ 169 VDC_STATE_NEGOTIATE, /* doing handshake negotiation */ 170 VDC_STATE_HANDLE_PENDING, /* handle requests in backup dring */ 171 VDC_STATE_RUNNING, /* running and accepting requests */ 172 VDC_STATE_DETACH, /* detaching */ 173 VDC_STATE_RESETTING /* resetting connection with vds */ 174 } vdc_state_t; 175 176 /* 177 * The states that the vdc instance can be in. 178 */ 179 typedef enum vdc_lc_state { 180 VDC_LC_ATTACHING, /* driver is attaching */ 181 VDC_LC_ONLINE, /* driver is attached and online */ 182 VDC_LC_DETACHING /* driver is detaching */ 183 } vdc_lc_state_t; 184 185 /* 186 * Local Descriptor Ring entry 187 * 188 * vdc creates a Local (private) descriptor ring the same size as the 189 * public descriptor ring it exports to vds. 190 */ 191 192 typedef enum { 193 VIO_read_dir, /* read data from server */ 194 VIO_write_dir, /* write data to server */ 195 VIO_both_dir /* transfer both in and out in same buffer */ 196 } vio_desc_direction_t; 197 198 typedef enum { 199 CB_STRATEGY, /* non-blocking strategy call */ 200 CB_SYNC /* synchronous operation */ 201 } vio_cb_type_t; 202 203 typedef struct vdc_local_desc { 204 boolean_t is_free; /* local state - inuse or not */ 205 206 int operation; /* VD_OP_xxx to be performed */ 207 caddr_t addr; /* addr passed in by consumer */ 208 int slice; 209 diskaddr_t offset; /* disk offset */ 210 size_t nbytes; 211 vio_cb_type_t cb_type; /* operation type blk/nonblk */ 212 void *cb_arg; /* buf passed to strategy() */ 213 vio_desc_direction_t dir; /* direction of transfer */ 214 215 caddr_t align_addr; /* used if addr non-aligned */ 216 ldc_mem_handle_t desc_mhdl; /* Mem handle of buf */ 217 vd_dring_entry_t *dep; /* public Dring Entry Pointer */ 218 219 } vdc_local_desc_t; 220 221 /* 222 * vdc soft state structure 223 */ 224 typedef struct vdc { 225 226 kmutex_t lock; /* protects next 2 sections of vars */ 227 kcondvar_t running_cv; /* signal when upper layers can send */ 228 kcondvar_t initwait_cv; /* signal when ldc conn is up */ 229 kcondvar_t dring_free_cv; /* signal when desc is avail */ 230 kcondvar_t membind_cv; /* signal when mem can be bound */ 231 boolean_t self_reset; 232 233 int initialized; /* keeps track of what's init'ed */ 234 vdc_lc_state_t lifecycle; /* Current state of the vdc instance */ 235 236 int hshake_cnt; /* number of failed handshakes */ 237 int open_count; /* count of outstanding opens */ 238 int dkio_flush_pending; /* # outstanding DKIO flushes */ 239 240 kthread_t *msg_proc_thr; /* main msg processing thread */ 241 242 kmutex_t read_lock; /* lock to protect read */ 243 kcondvar_t read_cv; /* cv to wait for READ events */ 244 vdc_rd_state_t read_state; /* current read state */ 245 246 uint32_t sync_op_cnt; /* num of active sync operations */ 247 boolean_t sync_op_pending; /* sync operation is pending */ 248 boolean_t sync_op_blocked; /* blocked waiting to do sync op */ 249 uint32_t sync_op_status; /* status of sync operation */ 250 kcondvar_t sync_pending_cv; /* cv wait for sync op to finish */ 251 kcondvar_t sync_blocked_cv; /* cv wait for other syncs to finish */ 252 253 uint64_t session_id; /* common ID sent with all messages */ 254 uint64_t seq_num; /* most recent sequence num generated */ 255 uint64_t seq_num_reply; /* Last seq num ACK/NACK'ed by vds */ 256 uint64_t req_id; /* Most recent Request ID generated */ 257 uint64_t req_id_proc; /* Last request ID processed by vdc */ 258 vdc_state_t state; /* Current disk client-server state */ 259 260 dev_info_t *dip; /* device info pointer */ 261 int instance; /* driver instance number */ 262 263 vio_ver_t ver; /* version number agreed with server */ 264 vd_disk_type_t vdisk_type; /* type of device/disk being imported */ 265 vd_disk_label_t vdisk_label; /* label type of device/disk imported */ 266 uint64_t vdisk_size; /* device size in blocks */ 267 uint64_t max_xfer_sz; /* maximum block size of a descriptor */ 268 uint64_t block_size; /* device block size used */ 269 struct dk_label *label; /* structure to store disk label */ 270 struct dk_cinfo *cinfo; /* structure to store DKIOCINFO data */ 271 struct dk_minfo *minfo; /* structure for DKIOCGMEDIAINFO data */ 272 struct vtoc *vtoc; /* structure to store VTOC data */ 273 ddi_devid_t devid; /* device id */ 274 275 ldc_mem_info_t dring_mem_info; /* dring information */ 276 uint_t dring_curr_idx; /* current index */ 277 uint32_t dring_len; /* dring length */ 278 uint32_t dring_max_cookies; /* dring max cookies */ 279 uint32_t dring_cookie_count; /* num cookies */ 280 uint32_t dring_entry_size; /* descriptor size */ 281 ldc_mem_cookie_t *dring_cookie; /* dring cookies */ 282 uint64_t dring_ident; /* dring ident */ 283 284 uint64_t threads_pending; /* num of threads */ 285 286 vdc_local_desc_t *local_dring; /* local dring */ 287 vdc_local_desc_t *local_dring_backup; /* local dring backup */ 288 int local_dring_backup_tail; /* backup dring tail */ 289 int local_dring_backup_len; /* backup dring len */ 290 291 uint64_t ldc_id; /* LDC channel id */ 292 ldc_status_t ldc_state; /* LDC channel state */ 293 ldc_handle_t ldc_handle; /* LDC handle */ 294 ldc_dring_handle_t ldc_dring_hdl; /* LDC dring handle */ 295 } vdc_t; 296 297 /* 298 * Debugging macros 299 */ 300 #ifdef DEBUG 301 extern int vdc_msglevel; 302 extern uint64_t vdc_matchinst; 303 304 #define DMSG(_vdc, err_level, format, ...) \ 305 do { \ 306 if (vdc_msglevel > err_level && \ 307 (vdc_matchinst & (1ull << (_vdc)->instance))) \ 308 cmn_err(CE_CONT, "?[%d,t@%p] %s: "format, \ 309 (_vdc)->instance, (void *)curthread, \ 310 __func__, __VA_ARGS__); \ 311 _NOTE(CONSTANTCONDITION) \ 312 } while (0); 313 314 #define DMSGX(err_level, format, ...) \ 315 do { \ 316 if (vdc_msglevel > err_level) \ 317 cmn_err(CE_CONT, "?%s: "format, __func__, __VA_ARGS__);\ 318 _NOTE(CONSTANTCONDITION) \ 319 } while (0); 320 321 #define VDC_DUMP_DRING_MSG(dmsgp) \ 322 DMSGX(0, "sq:%lu start:%d end:%d ident:%lu\n", \ 323 dmsgp->seq_num, dmsgp->start_idx, \ 324 dmsgp->end_idx, dmsgp->dring_ident); 325 326 #else /* !DEBUG */ 327 #define DMSG(err_level, ...) 328 #define DMSGX(err_level, format, ...) 329 #define VDC_DUMP_DRING_MSG(dmsgp) 330 331 #endif /* !DEBUG */ 332 333 #ifdef __cplusplus 334 } 335 #endif 336 337 #endif /* _VDC_H */ 338