1 /* 2 * This program is free software; you can redistribute it and/or 3 * modify it under the terms of the GNU General Public License version 2 4 * as published by the Free Software Foundation; or, when distributed 5 * separately from the Linux kernel or incorporated into other 6 * software packages, subject to the following license: 7 * 8 * Permission is hereby granted, free of charge, to any person obtaining a copy 9 * of this source file (the "Software"), to deal in the Software without 10 * restriction, including without limitation the rights to use, copy, modify, 11 * merge, publish, distribute, sublicense, and/or sell copies of the Software, 12 * and to permit persons to whom the Software is furnished to do so, subject to 13 * the following conditions: 14 * 15 * The above copyright notice and this permission notice shall be included in 16 * all copies or substantial portions of the Software. 17 * 18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 19 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 20 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 21 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 22 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING 23 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS 24 * IN THE SOFTWARE. 25 */ 26 27 #ifndef __XEN_BLKIF__BACKEND__COMMON_H__ 28 #define __XEN_BLKIF__BACKEND__COMMON_H__ 29 30 #include <linux/module.h> 31 #include <linux/interrupt.h> 32 #include <linux/slab.h> 33 #include <linux/blkdev.h> 34 #include <linux/vmalloc.h> 35 #include <linux/wait.h> 36 #include <linux/io.h> 37 #include <linux/rbtree.h> 38 #include <asm/setup.h> 39 #include <asm/hypervisor.h> 40 #include <xen/grant_table.h> 41 #include <xen/page.h> 42 #include <xen/xenbus.h> 43 #include <xen/interface/io/ring.h> 44 #include <xen/interface/io/blkif.h> 45 #include <xen/interface/io/protocols.h> 46 47 extern unsigned int xen_blkif_max_ring_order; 48 extern unsigned int xenblk_max_queues; 49 /* 50 * This is the maximum number of segments that would be allowed in indirect 51 * requests. This value will also be passed to the frontend. 52 */ 53 #define MAX_INDIRECT_SEGMENTS 256 54 55 /* 56 * Xen use 4K pages. The guest may use different page size (4K or 64K) 57 * Number of Xen pages per segment 58 */ 59 #define XEN_PAGES_PER_SEGMENT (PAGE_SIZE / XEN_PAGE_SIZE) 60 61 #define XEN_PAGES_PER_INDIRECT_FRAME \ 62 (XEN_PAGE_SIZE/sizeof(struct blkif_request_segment)) 63 #define SEGS_PER_INDIRECT_FRAME \ 64 (XEN_PAGES_PER_INDIRECT_FRAME / XEN_PAGES_PER_SEGMENT) 65 66 #define MAX_INDIRECT_PAGES \ 67 ((MAX_INDIRECT_SEGMENTS + SEGS_PER_INDIRECT_FRAME - 1)/SEGS_PER_INDIRECT_FRAME) 68 #define INDIRECT_PAGES(_segs) DIV_ROUND_UP(_segs, XEN_PAGES_PER_INDIRECT_FRAME) 69 70 /* Not a real protocol. Used to generate ring structs which contain 71 * the elements common to all protocols only. This way we get a 72 * compiler-checkable way to use common struct elements, so we can 73 * avoid using switch(protocol) in a number of places. */ 74 struct blkif_common_request { 75 char dummy; 76 }; 77 78 /* i386 protocol version */ 79 80 struct blkif_x86_32_request_rw { 81 uint8_t nr_segments; /* number of segments */ 82 blkif_vdev_t handle; /* only for read/write requests */ 83 uint64_t id; /* private guest value, echoed in resp */ 84 blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */ 85 struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 86 } __attribute__((__packed__)); 87 88 struct blkif_x86_32_request_discard { 89 uint8_t flag; /* BLKIF_DISCARD_SECURE or zero */ 90 blkif_vdev_t _pad1; /* was "handle" for read/write requests */ 91 uint64_t id; /* private guest value, echoed in resp */ 92 blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */ 93 uint64_t nr_sectors; 94 } __attribute__((__packed__)); 95 96 struct blkif_x86_32_request_other { 97 uint8_t _pad1; 98 blkif_vdev_t _pad2; 99 uint64_t id; /* private guest value, echoed in resp */ 100 } __attribute__((__packed__)); 101 102 struct blkif_x86_32_request_indirect { 103 uint8_t indirect_op; 104 uint16_t nr_segments; 105 uint64_t id; 106 blkif_sector_t sector_number; 107 blkif_vdev_t handle; 108 uint16_t _pad1; 109 grant_ref_t indirect_grefs[BLKIF_MAX_INDIRECT_PAGES_PER_REQUEST]; 110 /* 111 * The maximum number of indirect segments (and pages) that will 112 * be used is determined by MAX_INDIRECT_SEGMENTS, this value 113 * is also exported to the guest (via xenstore 114 * feature-max-indirect-segments entry), so the frontend knows how 115 * many indirect segments the backend supports. 116 */ 117 uint64_t _pad2; /* make it 64 byte aligned */ 118 } __attribute__((__packed__)); 119 120 struct blkif_x86_32_request { 121 uint8_t operation; /* BLKIF_OP_??? */ 122 union { 123 struct blkif_x86_32_request_rw rw; 124 struct blkif_x86_32_request_discard discard; 125 struct blkif_x86_32_request_other other; 126 struct blkif_x86_32_request_indirect indirect; 127 } u; 128 } __attribute__((__packed__)); 129 130 /* x86_64 protocol version */ 131 132 struct blkif_x86_64_request_rw { 133 uint8_t nr_segments; /* number of segments */ 134 blkif_vdev_t handle; /* only for read/write requests */ 135 uint32_t _pad1; /* offsetof(blkif_reqest..,u.rw.id)==8 */ 136 uint64_t id; 137 blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */ 138 struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 139 } __attribute__((__packed__)); 140 141 struct blkif_x86_64_request_discard { 142 uint8_t flag; /* BLKIF_DISCARD_SECURE or zero */ 143 blkif_vdev_t _pad1; /* was "handle" for read/write requests */ 144 uint32_t _pad2; /* offsetof(blkif_..,u.discard.id)==8 */ 145 uint64_t id; 146 blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */ 147 uint64_t nr_sectors; 148 } __attribute__((__packed__)); 149 150 struct blkif_x86_64_request_other { 151 uint8_t _pad1; 152 blkif_vdev_t _pad2; 153 uint32_t _pad3; /* offsetof(blkif_..,u.discard.id)==8 */ 154 uint64_t id; /* private guest value, echoed in resp */ 155 } __attribute__((__packed__)); 156 157 struct blkif_x86_64_request_indirect { 158 uint8_t indirect_op; 159 uint16_t nr_segments; 160 uint32_t _pad1; /* offsetof(blkif_..,u.indirect.id)==8 */ 161 uint64_t id; 162 blkif_sector_t sector_number; 163 blkif_vdev_t handle; 164 uint16_t _pad2; 165 grant_ref_t indirect_grefs[BLKIF_MAX_INDIRECT_PAGES_PER_REQUEST]; 166 /* 167 * The maximum number of indirect segments (and pages) that will 168 * be used is determined by MAX_INDIRECT_SEGMENTS, this value 169 * is also exported to the guest (via xenstore 170 * feature-max-indirect-segments entry), so the frontend knows how 171 * many indirect segments the backend supports. 172 */ 173 uint32_t _pad3; /* make it 64 byte aligned */ 174 } __attribute__((__packed__)); 175 176 struct blkif_x86_64_request { 177 uint8_t operation; /* BLKIF_OP_??? */ 178 union { 179 struct blkif_x86_64_request_rw rw; 180 struct blkif_x86_64_request_discard discard; 181 struct blkif_x86_64_request_other other; 182 struct blkif_x86_64_request_indirect indirect; 183 } u; 184 } __attribute__((__packed__)); 185 186 DEFINE_RING_TYPES(blkif_common, struct blkif_common_request, 187 struct blkif_response); 188 DEFINE_RING_TYPES(blkif_x86_32, struct blkif_x86_32_request, 189 struct blkif_response __packed); 190 DEFINE_RING_TYPES(blkif_x86_64, struct blkif_x86_64_request, 191 struct blkif_response); 192 193 union blkif_back_rings { 194 struct blkif_back_ring native; 195 struct blkif_common_back_ring common; 196 struct blkif_x86_32_back_ring x86_32; 197 struct blkif_x86_64_back_ring x86_64; 198 }; 199 200 enum blkif_protocol { 201 BLKIF_PROTOCOL_NATIVE = 1, 202 BLKIF_PROTOCOL_X86_32 = 2, 203 BLKIF_PROTOCOL_X86_64 = 3, 204 }; 205 206 /* 207 * Default protocol if the frontend doesn't specify one. 208 */ 209 #ifdef CONFIG_X86 210 # define BLKIF_PROTOCOL_DEFAULT BLKIF_PROTOCOL_X86_32 211 #else 212 # define BLKIF_PROTOCOL_DEFAULT BLKIF_PROTOCOL_NATIVE 213 #endif 214 215 struct xen_vbd { 216 /* What the domain refers to this vbd as. */ 217 blkif_vdev_t handle; 218 /* Non-zero -> read-only */ 219 unsigned char readonly; 220 /* VDISK_xxx */ 221 unsigned char type; 222 /* phys device that this vbd maps to. */ 223 u32 pdevice; 224 struct block_device *bdev; 225 /* Cached size parameter. */ 226 sector_t size; 227 unsigned int flush_support:1; 228 unsigned int discard_secure:1; 229 /* Connect-time cached feature_persistent parameter value */ 230 unsigned int feature_gnt_persistent_parm:1; 231 /* Persistent grants feature negotiation result */ 232 unsigned int feature_gnt_persistent:1; 233 unsigned int overflow_max_grants:1; 234 }; 235 236 struct backend_info; 237 238 /* Number of requests that we can fit in a ring */ 239 #define XEN_BLKIF_REQS_PER_PAGE 32 240 241 struct persistent_gnt { 242 struct page *page; 243 grant_ref_t gnt; 244 grant_handle_t handle; 245 unsigned long last_used; 246 bool active; 247 struct rb_node node; 248 struct list_head remove_node; 249 }; 250 251 /* Per-ring information. */ 252 struct xen_blkif_ring { 253 /* Physical parameters of the comms window. */ 254 unsigned int irq; 255 union blkif_back_rings blk_rings; 256 void *blk_ring; 257 /* Private fields. */ 258 spinlock_t blk_ring_lock; 259 260 wait_queue_head_t wq; 261 atomic_t inflight; 262 bool active; 263 /* One thread per blkif ring. */ 264 struct task_struct *xenblkd; 265 unsigned int waiting_reqs; 266 267 /* List of all 'pending_req' available */ 268 struct list_head pending_free; 269 /* And its spinlock. */ 270 spinlock_t pending_free_lock; 271 wait_queue_head_t pending_free_wq; 272 273 /* Tree to store persistent grants. */ 274 struct rb_root persistent_gnts; 275 unsigned int persistent_gnt_c; 276 atomic_t persistent_gnt_in_use; 277 unsigned long next_lru; 278 279 /* Statistics. */ 280 unsigned long st_print; 281 unsigned long long st_rd_req; 282 unsigned long long st_wr_req; 283 unsigned long long st_oo_req; 284 unsigned long long st_f_req; 285 unsigned long long st_ds_req; 286 unsigned long long st_rd_sect; 287 unsigned long long st_wr_sect; 288 289 /* Used by the kworker that offload work from the persistent purge. */ 290 struct list_head persistent_purge_list; 291 struct work_struct persistent_purge_work; 292 293 /* Buffer of free pages to map grant refs. */ 294 struct gnttab_page_cache free_pages; 295 296 struct work_struct free_work; 297 /* Thread shutdown wait queue. */ 298 wait_queue_head_t shutdown_wq; 299 struct xen_blkif *blkif; 300 }; 301 302 struct xen_blkif { 303 /* Unique identifier for this interface. */ 304 domid_t domid; 305 unsigned int handle; 306 /* Comms information. */ 307 enum blkif_protocol blk_protocol; 308 /* The VBD attached to this interface. */ 309 struct xen_vbd vbd; 310 /* Back pointer to the backend_info. */ 311 struct backend_info *be; 312 atomic_t refcnt; 313 /* for barrier (drain) requests */ 314 struct completion drain_complete; 315 atomic_t drain; 316 317 struct work_struct free_work; 318 unsigned int nr_ring_pages; 319 bool multi_ref; 320 /* All rings for this device. */ 321 struct xen_blkif_ring *rings; 322 unsigned int nr_rings; 323 unsigned long buffer_squeeze_end; 324 }; 325 326 struct seg_buf { 327 unsigned long offset; 328 unsigned int nsec; 329 }; 330 331 struct grant_page { 332 struct page *page; 333 struct persistent_gnt *persistent_gnt; 334 grant_handle_t handle; 335 grant_ref_t gref; 336 }; 337 338 /* 339 * Each outstanding request that we've passed to the lower device layers has a 340 * 'pending_req' allocated to it. Each buffer_head that completes decrements 341 * the pendcnt towards zero. When it hits zero, the specified domain has a 342 * response queued for it, with the saved 'id' passed back. 343 */ 344 struct pending_req { 345 struct xen_blkif_ring *ring; 346 u64 id; 347 int nr_segs; 348 atomic_t pendcnt; 349 unsigned short operation; 350 int status; 351 struct list_head free_list; 352 struct grant_page *segments[MAX_INDIRECT_SEGMENTS]; 353 /* Indirect descriptors */ 354 struct grant_page *indirect_pages[MAX_INDIRECT_PAGES]; 355 struct seg_buf seg[MAX_INDIRECT_SEGMENTS]; 356 struct bio *biolist[MAX_INDIRECT_SEGMENTS]; 357 struct gnttab_unmap_grant_ref unmap[MAX_INDIRECT_SEGMENTS]; 358 struct page *unmap_pages[MAX_INDIRECT_SEGMENTS]; 359 struct gntab_unmap_queue_data gnttab_unmap_data; 360 }; 361 362 363 #define vbd_sz(_v) bdev_nr_sectors((_v)->bdev) 364 365 #define xen_blkif_get(_b) (atomic_inc(&(_b)->refcnt)) 366 #define xen_blkif_put(_b) \ 367 do { \ 368 if (atomic_dec_and_test(&(_b)->refcnt)) \ 369 schedule_work(&(_b)->free_work);\ 370 } while (0) 371 372 struct phys_req { 373 unsigned short dev; 374 blkif_sector_t nr_sects; 375 struct block_device *bdev; 376 blkif_sector_t sector_number; 377 }; 378 379 int xen_blkif_interface_init(void); 380 void xen_blkif_interface_fini(void); 381 382 int xen_blkif_xenbus_init(void); 383 void xen_blkif_xenbus_fini(void); 384 385 irqreturn_t xen_blkif_be_int(int irq, void *dev_id); 386 int xen_blkif_schedule(void *arg); 387 int xen_blkif_purge_persistent(void *arg); 388 void xen_blkbk_free_caches(struct xen_blkif_ring *ring); 389 390 int xen_blkbk_flush_diskcache(struct xenbus_transaction xbt, 391 struct backend_info *be, int state); 392 393 int xen_blkbk_barrier(struct xenbus_transaction xbt, 394 struct backend_info *be, int state); 395 struct xenbus_device *xen_blkbk_xenbus(struct backend_info *be); 396 void xen_blkbk_unmap_purged_grants(struct work_struct *work); 397 398 static inline void blkif_get_x86_32_req(struct blkif_request *dst, 399 struct blkif_x86_32_request *src) 400 { 401 int i, n = BLKIF_MAX_SEGMENTS_PER_REQUEST, j; 402 dst->operation = READ_ONCE(src->operation); 403 switch (dst->operation) { 404 case BLKIF_OP_READ: 405 case BLKIF_OP_WRITE: 406 case BLKIF_OP_WRITE_BARRIER: 407 case BLKIF_OP_FLUSH_DISKCACHE: 408 dst->u.rw.nr_segments = src->u.rw.nr_segments; 409 dst->u.rw.handle = src->u.rw.handle; 410 dst->u.rw.id = src->u.rw.id; 411 dst->u.rw.sector_number = src->u.rw.sector_number; 412 barrier(); 413 if (n > dst->u.rw.nr_segments) 414 n = dst->u.rw.nr_segments; 415 for (i = 0; i < n; i++) 416 dst->u.rw.seg[i] = src->u.rw.seg[i]; 417 break; 418 case BLKIF_OP_DISCARD: 419 dst->u.discard.flag = src->u.discard.flag; 420 dst->u.discard.id = src->u.discard.id; 421 dst->u.discard.sector_number = src->u.discard.sector_number; 422 dst->u.discard.nr_sectors = src->u.discard.nr_sectors; 423 break; 424 case BLKIF_OP_INDIRECT: 425 dst->u.indirect.indirect_op = src->u.indirect.indirect_op; 426 dst->u.indirect.nr_segments = src->u.indirect.nr_segments; 427 dst->u.indirect.handle = src->u.indirect.handle; 428 dst->u.indirect.id = src->u.indirect.id; 429 dst->u.indirect.sector_number = src->u.indirect.sector_number; 430 barrier(); 431 j = min(MAX_INDIRECT_PAGES, INDIRECT_PAGES(dst->u.indirect.nr_segments)); 432 for (i = 0; i < j; i++) 433 dst->u.indirect.indirect_grefs[i] = 434 src->u.indirect.indirect_grefs[i]; 435 break; 436 default: 437 /* 438 * Don't know how to translate this op. Only get the 439 * ID so failure can be reported to the frontend. 440 */ 441 dst->u.other.id = src->u.other.id; 442 break; 443 } 444 } 445 446 static inline void blkif_get_x86_64_req(struct blkif_request *dst, 447 struct blkif_x86_64_request *src) 448 { 449 int i, n = BLKIF_MAX_SEGMENTS_PER_REQUEST, j; 450 dst->operation = READ_ONCE(src->operation); 451 switch (dst->operation) { 452 case BLKIF_OP_READ: 453 case BLKIF_OP_WRITE: 454 case BLKIF_OP_WRITE_BARRIER: 455 case BLKIF_OP_FLUSH_DISKCACHE: 456 dst->u.rw.nr_segments = src->u.rw.nr_segments; 457 dst->u.rw.handle = src->u.rw.handle; 458 dst->u.rw.id = src->u.rw.id; 459 dst->u.rw.sector_number = src->u.rw.sector_number; 460 barrier(); 461 if (n > dst->u.rw.nr_segments) 462 n = dst->u.rw.nr_segments; 463 for (i = 0; i < n; i++) 464 dst->u.rw.seg[i] = src->u.rw.seg[i]; 465 break; 466 case BLKIF_OP_DISCARD: 467 dst->u.discard.flag = src->u.discard.flag; 468 dst->u.discard.id = src->u.discard.id; 469 dst->u.discard.sector_number = src->u.discard.sector_number; 470 dst->u.discard.nr_sectors = src->u.discard.nr_sectors; 471 break; 472 case BLKIF_OP_INDIRECT: 473 dst->u.indirect.indirect_op = src->u.indirect.indirect_op; 474 dst->u.indirect.nr_segments = src->u.indirect.nr_segments; 475 dst->u.indirect.handle = src->u.indirect.handle; 476 dst->u.indirect.id = src->u.indirect.id; 477 dst->u.indirect.sector_number = src->u.indirect.sector_number; 478 barrier(); 479 j = min(MAX_INDIRECT_PAGES, INDIRECT_PAGES(dst->u.indirect.nr_segments)); 480 for (i = 0; i < j; i++) 481 dst->u.indirect.indirect_grefs[i] = 482 src->u.indirect.indirect_grefs[i]; 483 break; 484 default: 485 /* 486 * Don't know how to translate this op. Only get the 487 * ID so failure can be reported to the frontend. 488 */ 489 dst->u.other.id = src->u.other.id; 490 break; 491 } 492 } 493 494 #endif /* __XEN_BLKIF__BACKEND__COMMON_H__ */ 495