drivers/block/xen-blkback/common.h

   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/pgalloc.h>
  40 #include <asm/hypervisor.h>
  41 #include <xen/grant_table.h>
  42 #include <xen/page.h>
  43 #include <xen/xenbus.h>
  44 #include <xen/interface/io/ring.h>
  45 #include <xen/interface/io/blkif.h>
  46 #include <xen/interface/io/protocols.h>
  47
  48 extern unsigned int xen_blkif_max_ring_order;
  49 extern unsigned int xenblk_max_queues;
  50 /*
  51  * This is the maximum number of segments that would be allowed in indirect
  52  * requests. This value will also be passed to the frontend.
  53  */
  54 #define MAX_INDIRECT_SEGMENTS 256
  55
  56 /*
  57  * Xen use 4K pages. The guest may use different page size (4K or 64K)
  58  * Number of Xen pages per segment
  59  */
  60 #define XEN_PAGES_PER_SEGMENT   (PAGE_SIZE / XEN_PAGE_SIZE)
  61
  62 #define XEN_PAGES_PER_INDIRECT_FRAME \
  63         (XEN_PAGE_SIZE/sizeof(struct blkif_request_segment))
  64 #define SEGS_PER_INDIRECT_FRAME \
  65         (XEN_PAGES_PER_INDIRECT_FRAME / XEN_PAGES_PER_SEGMENT)
  66
  67 #define MAX_INDIRECT_PAGES \
  68         ((MAX_INDIRECT_SEGMENTS + SEGS_PER_INDIRECT_FRAME - 1)/SEGS_PER_INDIRECT_FRAME)
  69 #define INDIRECT_PAGES(_segs) DIV_ROUND_UP(_segs, XEN_PAGES_PER_INDIRECT_FRAME)
  70
  71 /* Not a real protocol.  Used to generate ring structs which contain
  72  * the elements common to all protocols only.  This way we get a
  73  * compiler-checkable way to use common struct elements, so we can
  74  * avoid using switch(protocol) in a number of places.  */
  75 struct blkif_common_request {
  76         char dummy;
  77 };
  78 struct blkif_common_response {
  79         char dummy;
  80 };
  81
  82 struct blkif_x86_32_request_rw {
  83         uint8_t        nr_segments;  /* number of segments                   */
  84         blkif_vdev_t   handle;       /* only for read/write requests         */
  85         uint64_t       id;           /* private guest value, echoed in resp  */
  86         blkif_sector_t sector_number;/* start sector idx on disk (r/w only)  */
  87         struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
  88 } __attribute__((__packed__));
  89
  90 struct blkif_x86_32_request_discard {
  91         uint8_t        flag;         /* BLKIF_DISCARD_SECURE or zero         */
  92         blkif_vdev_t   _pad1;        /* was "handle" for read/write requests */
  93         uint64_t       id;           /* private guest value, echoed in resp  */
  94         blkif_sector_t sector_number;/* start sector idx on disk (r/w only)  */
  95         uint64_t       nr_sectors;
  96 } __attribute__((__packed__));
  97
  98 struct blkif_x86_32_request_other {
  99         uint8_t        _pad1;
 100         blkif_vdev_t   _pad2;
 101         uint64_t       id;           /* private guest value, echoed in resp  */
 102 } __attribute__((__packed__));
 103
 104 struct blkif_x86_32_request_indirect {
 105         uint8_t        indirect_op;
 106         uint16_t       nr_segments;
 107         uint64_t       id;
 108         blkif_sector_t sector_number;
 109         blkif_vdev_t   handle;
 110         uint16_t       _pad1;
 111         grant_ref_t    indirect_grefs[BLKIF_MAX_INDIRECT_PAGES_PER_REQUEST];
 112         /*
 113          * The maximum number of indirect segments (and pages) that will
 114          * be used is determined by MAX_INDIRECT_SEGMENTS, this value
 115          * is also exported to the guest (via xenstore
 116          * feature-max-indirect-segments entry), so the frontend knows how
 117          * many indirect segments the backend supports.
 118          */
 119         uint64_t       _pad2;        /* make it 64 byte aligned */
 120 } __attribute__((__packed__));
 121
 122 struct blkif_x86_32_request {
 123         uint8_t        operation;    /* BLKIF_OP_???                         */
 124         union {
 125                 struct blkif_x86_32_request_rw rw;
 126                 struct blkif_x86_32_request_discard discard;
 127                 struct blkif_x86_32_request_other other;
 128                 struct blkif_x86_32_request_indirect indirect;
 129         } u;
 130 } __attribute__((__packed__));
 131
 132 /* i386 protocol version */
 133 #pragma pack(push, 4)
 134 struct blkif_x86_32_response {
 135         uint64_t        id;              /* copied from request */
 136         uint8_t         operation;       /* copied from request */
 137         int16_t         status;          /* BLKIF_RSP_???       */
 138 };
 139 #pragma pack(pop)
 140 /* x86_64 protocol version */
 141
 142 struct blkif_x86_64_request_rw {
 143         uint8_t        nr_segments;  /* number of segments                   */
 144         blkif_vdev_t   handle;       /* only for read/write requests         */
 145         uint32_t       _pad1;        /* offsetof(blkif_reqest..,u.rw.id)==8  */
 146         uint64_t       id;
 147         blkif_sector_t sector_number;/* start sector idx on disk (r/w only)  */
 148         struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
 149 } __attribute__((__packed__));
 150
 151 struct blkif_x86_64_request_discard {
 152         uint8_t        flag;         /* BLKIF_DISCARD_SECURE or zero         */
 153         blkif_vdev_t   _pad1;        /* was "handle" for read/write requests */
 154         uint32_t       _pad2;        /* offsetof(blkif_..,u.discard.id)==8   */
 155         uint64_t       id;
 156         blkif_sector_t sector_number;/* start sector idx on disk (r/w only)  */
 157         uint64_t       nr_sectors;
 158 } __attribute__((__packed__));
 159
 160 struct blkif_x86_64_request_other {
 161         uint8_t        _pad1;
 162         blkif_vdev_t   _pad2;
 163         uint32_t       _pad3;        /* offsetof(blkif_..,u.discard.id)==8   */
 164         uint64_t       id;           /* private guest value, echoed in resp  */
 165 } __attribute__((__packed__));
 166
 167 struct blkif_x86_64_request_indirect {
 168         uint8_t        indirect_op;
 169         uint16_t       nr_segments;
 170         uint32_t       _pad1;        /* offsetof(blkif_..,u.indirect.id)==8   */
 171         uint64_t       id;
 172         blkif_sector_t sector_number;
 173         blkif_vdev_t   handle;
 174         uint16_t       _pad2;
 175         grant_ref_t    indirect_grefs[BLKIF_MAX_INDIRECT_PAGES_PER_REQUEST];
 176         /*
 177          * The maximum number of indirect segments (and pages) that will
 178          * be used is determined by MAX_INDIRECT_SEGMENTS, this value
 179          * is also exported to the guest (via xenstore
 180          * feature-max-indirect-segments entry), so the frontend knows how
 181          * many indirect segments the backend supports.
 182          */
 183         uint32_t       _pad3;        /* make it 64 byte aligned */
 184 } __attribute__((__packed__));
 185
 186 struct blkif_x86_64_request {
 187         uint8_t        operation;    /* BLKIF_OP_???                         */
 188         union {
 189                 struct blkif_x86_64_request_rw rw;
 190                 struct blkif_x86_64_request_discard discard;
 191                 struct blkif_x86_64_request_other other;
 192                 struct blkif_x86_64_request_indirect indirect;
 193         } u;
 194 } __attribute__((__packed__));
 195
 196 struct blkif_x86_64_response {
 197         uint64_t       __attribute__((__aligned__(8))) id;
 198         uint8_t         operation;       /* copied from request */
 199         int16_t         status;          /* BLKIF_RSP_???       */
 200 };
 201
 202 DEFINE_RING_TYPES(blkif_common, struct blkif_common_request,
 203                   struct blkif_common_response);
 204 DEFINE_RING_TYPES(blkif_x86_32, struct blkif_x86_32_request,
 205                   struct blkif_x86_32_response);
 206 DEFINE_RING_TYPES(blkif_x86_64, struct blkif_x86_64_request,
 207                   struct blkif_x86_64_response);
 208
 209 union blkif_back_rings {
 210         struct blkif_back_ring        native;
 211         struct blkif_common_back_ring common;
 212         struct blkif_x86_32_back_ring x86_32;
 213         struct blkif_x86_64_back_ring x86_64;
 214 };
 215
 216 enum blkif_protocol {
 217         BLKIF_PROTOCOL_NATIVE = 1,
 218         BLKIF_PROTOCOL_X86_32 = 2,
 219         BLKIF_PROTOCOL_X86_64 = 3,
 220 };
 221
 222 /*
 223  * Default protocol if the frontend doesn't specify one.
 224  */
 225 #ifdef CONFIG_X86
 226 #  define BLKIF_PROTOCOL_DEFAULT BLKIF_PROTOCOL_X86_32
 227 #else
 228 #  define BLKIF_PROTOCOL_DEFAULT BLKIF_PROTOCOL_NATIVE
 229 #endif
 230
 231 struct xen_vbd {
 232         /* What the domain refers to this vbd as. */
 233         blkif_vdev_t            handle;
 234         /* Non-zero -> read-only */
 235         unsigned char           readonly;
 236         /* VDISK_xxx */
 237         unsigned char           type;
 238         /* phys device that this vbd maps to. */
 239         u32                     pdevice;
 240         struct block_device     *bdev;
 241         /* Cached size parameter. */
 242         sector_t                size;
 243         unsigned int            flush_support:1;
 244         unsigned int            discard_secure:1;
 245         unsigned int            feature_gnt_persistent:1;
 246         unsigned int            overflow_max_grants:1;
 247 };
 248
 249 struct backend_info;
 250
 251 /* Number of available flags */
 252 #define PERSISTENT_GNT_FLAGS_SIZE       2
 253 /* This persistent grant is currently in use */
 254 #define PERSISTENT_GNT_ACTIVE           0
 255 /*
 256  * This persistent grant has been used, this flag is set when we remove the
 257  * PERSISTENT_GNT_ACTIVE, to know that this grant has been used recently.
 258  */
 259 #define PERSISTENT_GNT_WAS_ACTIVE       1
 260
 261 /* Number of requests that we can fit in a ring */
 262 #define XEN_BLKIF_REQS_PER_PAGE         32
 263
 264 struct persistent_gnt {
 265         struct page *page;
 266         grant_ref_t gnt;
 267         grant_handle_t handle;
 268         DECLARE_BITMAP(flags, PERSISTENT_GNT_FLAGS_SIZE);
 269         struct rb_node node;
 270         struct list_head remove_node;
 271 };
 272
 273 /* Per-ring information. */
 274 struct xen_blkif_ring {
 275         /* Physical parameters of the comms window. */
 276         unsigned int            irq;
 277         union blkif_back_rings  blk_rings;
 278         void                    *blk_ring;
 279         /* Private fields. */
 280         spinlock_t              blk_ring_lock;
 281
 282         wait_queue_head_t       wq;
 283         atomic_t                inflight;
 284         bool                    active;
 285         /* One thread per blkif ring. */
 286         struct task_struct      *xenblkd;
 287         unsigned int            waiting_reqs;
 288
 289         /* List of all 'pending_req' available */
 290         struct list_head        pending_free;
 291         /* And its spinlock. */
 292         spinlock_t              pending_free_lock;
 293         wait_queue_head_t       pending_free_wq;
 294
 295         /* Tree to store persistent grants. */
 296         spinlock_t              pers_gnts_lock;
 297         struct rb_root          persistent_gnts;
 298         unsigned int            persistent_gnt_c;
 299         atomic_t                persistent_gnt_in_use;
 300         unsigned long           next_lru;
 301
 302         /* Statistics. */
 303         unsigned long           st_print;
 304         unsigned long long      st_rd_req;
 305         unsigned long long      st_wr_req;
 306         unsigned long long      st_oo_req;
 307         unsigned long long      st_f_req;
 308         unsigned long long      st_ds_req;
 309         unsigned long long      st_rd_sect;
 310         unsigned long long      st_wr_sect;
 311
 312         /* Used by the kworker that offload work from the persistent purge. */
 313         struct list_head        persistent_purge_list;
 314         struct work_struct      persistent_purge_work;
 315
 316         /* Buffer of free pages to map grant refs. */
 317         spinlock_t              free_pages_lock;
 318         int                     free_pages_num;
 319         struct list_head        free_pages;
 320
 321         struct work_struct      free_work;
 322         /* Thread shutdown wait queue. */
 323         wait_queue_head_t       shutdown_wq;
 324         struct xen_blkif        *blkif;
 325 };
 326
 327 struct xen_blkif {
 328         /* Unique identifier for this interface. */
 329         domid_t                 domid;
 330         unsigned int            handle;
 331         /* Comms information. */
 332         enum blkif_protocol     blk_protocol;
 333         /* The VBD attached to this interface. */
 334         struct xen_vbd          vbd;
 335         /* Back pointer to the backend_info. */
 336         struct backend_info     *be;
 337         atomic_t                refcnt;
 338         /* for barrier (drain) requests */
 339         struct completion       drain_complete;
 340         atomic_t                drain;
 341
 342         struct work_struct      free_work;
 343         unsigned int            nr_ring_pages;
 344         /* All rings for this device. */
 345         struct xen_blkif_ring   *rings;
 346         unsigned int            nr_rings;
 347 };
 348
 349 struct seg_buf {
 350         unsigned long offset;
 351         unsigned int nsec;
 352 };
 353
 354 struct grant_page {
 355         struct page             *page;
 356         struct persistent_gnt   *persistent_gnt;
 357         grant_handle_t          handle;
 358         grant_ref_t             gref;
 359 };
 360
 361 /*
 362  * Each outstanding request that we've passed to the lower device layers has a
 363  * 'pending_req' allocated to it. Each buffer_head that completes decrements
 364  * the pendcnt towards zero. When it hits zero, the specified domain has a
 365  * response queued for it, with the saved 'id' passed back.
 366  */
 367 struct pending_req {
 368         struct xen_blkif_ring   *ring;
 369         u64                     id;
 370         int                     nr_segs;
 371         atomic_t                pendcnt;
 372         unsigned short          operation;
 373         int                     status;
 374         struct list_head        free_list;
 375         struct grant_page       *segments[MAX_INDIRECT_SEGMENTS];
 376         /* Indirect descriptors */
 377         struct grant_page       *indirect_pages[MAX_INDIRECT_PAGES];
 378         struct seg_buf          seg[MAX_INDIRECT_SEGMENTS];
 379         struct bio              *biolist[MAX_INDIRECT_SEGMENTS];
 380         struct gnttab_unmap_grant_ref unmap[MAX_INDIRECT_SEGMENTS];
 381         struct page                   *unmap_pages[MAX_INDIRECT_SEGMENTS];
 382         struct gntab_unmap_queue_data gnttab_unmap_data;
 383 };
 384
 385
 386 #define vbd_sz(_v)      ((_v)->bdev->bd_part ? \
 387                          (_v)->bdev->bd_part->nr_sects : \
 388                           get_capacity((_v)->bdev->bd_disk))
 389
 390 #define xen_blkif_get(_b) (atomic_inc(&(_b)->refcnt))
 391 #define xen_blkif_put(_b)                               \
 392         do {                                            \
 393                 if (atomic_dec_and_test(&(_b)->refcnt)) \
 394                         schedule_work(&(_b)->free_work);\
 395         } while (0)
 396
 397 struct phys_req {
 398         unsigned short          dev;
 399         blkif_sector_t          nr_sects;
 400         struct block_device     *bdev;
 401         blkif_sector_t          sector_number;
 402 };
 403 int xen_blkif_interface_init(void);
 404
 405 int xen_blkif_xenbus_init(void);
 406
 407 irqreturn_t xen_blkif_be_int(int irq, void *dev_id);
 408 int xen_blkif_schedule(void *arg);
 409 int xen_blkif_purge_persistent(void *arg);
 410 void xen_blkbk_free_caches(struct xen_blkif_ring *ring);
 411
 412 int xen_blkbk_flush_diskcache(struct xenbus_transaction xbt,
 413                               struct backend_info *be, int state);
 414
 415 int xen_blkbk_barrier(struct xenbus_transaction xbt,
 416                       struct backend_info *be, int state);
 417 struct xenbus_device *xen_blkbk_xenbus(struct backend_info *be);
 418 void xen_blkbk_unmap_purged_grants(struct work_struct *work);
 419
 420 static inline void blkif_get_x86_32_req(struct blkif_request *dst,
 421                                         struct blkif_x86_32_request *src)
 422 {
 423         int i, n = BLKIF_MAX_SEGMENTS_PER_REQUEST, j;
 424         dst->operation = READ_ONCE(src->operation);
 425         switch (dst->operation) {
 426         case BLKIF_OP_READ:
 427         case BLKIF_OP_WRITE:
 428         case BLKIF_OP_WRITE_BARRIER:
 429         case BLKIF_OP_FLUSH_DISKCACHE:
 430                 dst->u.rw.nr_segments = src->u.rw.nr_segments;
 431                 dst->u.rw.handle = src->u.rw.handle;
 432                 dst->u.rw.id = src->u.rw.id;
 433                 dst->u.rw.sector_number = src->u.rw.sector_number;
 434                 barrier();
 435                 if (n > dst->u.rw.nr_segments)
 436                         n = dst->u.rw.nr_segments;
 437                 for (i = 0; i < n; i++)
 438                         dst->u.rw.seg[i] = src->u.rw.seg[i];
 439                 break;
 440         case BLKIF_OP_DISCARD:
 441                 dst->u.discard.flag = src->u.discard.flag;
 442                 dst->u.discard.id = src->u.discard.id;
 443                 dst->u.discard.sector_number = src->u.discard.sector_number;
 444                 dst->u.discard.nr_sectors = src->u.discard.nr_sectors;
 445                 break;
 446         case BLKIF_OP_INDIRECT:
 447                 dst->u.indirect.indirect_op = src->u.indirect.indirect_op;
 448                 dst->u.indirect.nr_segments = src->u.indirect.nr_segments;
 449                 dst->u.indirect.handle = src->u.indirect.handle;
 450                 dst->u.indirect.id = src->u.indirect.id;
 451                 dst->u.indirect.sector_number = src->u.indirect.sector_number;
 452                 barrier();
 453                 j = min(MAX_INDIRECT_PAGES, INDIRECT_PAGES(dst->u.indirect.nr_segments));
 454                 for (i = 0; i < j; i++)
 455                         dst->u.indirect.indirect_grefs[i] =
 456                                 src->u.indirect.indirect_grefs[i];
 457                 break;
 458         default:
 459                 /*
 460                  * Don't know how to translate this op. Only get the
 461                  * ID so failure can be reported to the frontend.
 462                  */
 463                 dst->u.other.id = src->u.other.id;
 464                 break;
 465         }
 466 }
 467
 468 static inline void blkif_get_x86_64_req(struct blkif_request *dst,
 469                                         struct blkif_x86_64_request *src)
 470 {
 471         int i, n = BLKIF_MAX_SEGMENTS_PER_REQUEST, j;
 472         dst->operation = READ_ONCE(src->operation);
 473         switch (dst->operation) {
 474         case BLKIF_OP_READ:
 475         case BLKIF_OP_WRITE:
 476         case BLKIF_OP_WRITE_BARRIER:
 477         case BLKIF_OP_FLUSH_DISKCACHE:
 478                 dst->u.rw.nr_segments = src->u.rw.nr_segments;
 479                 dst->u.rw.handle = src->u.rw.handle;
 480                 dst->u.rw.id = src->u.rw.id;
 481                 dst->u.rw.sector_number = src->u.rw.sector_number;
 482                 barrier();
 483                 if (n > dst->u.rw.nr_segments)
 484                         n = dst->u.rw.nr_segments;
 485                 for (i = 0; i < n; i++)
 486                         dst->u.rw.seg[i] = src->u.rw.seg[i];
 487                 break;
 488         case BLKIF_OP_DISCARD:
 489                 dst->u.discard.flag = src->u.discard.flag;
 490                 dst->u.discard.id = src->u.discard.id;
 491                 dst->u.discard.sector_number = src->u.discard.sector_number;
 492                 dst->u.discard.nr_sectors = src->u.discard.nr_sectors;
 493                 break;
 494         case BLKIF_OP_INDIRECT:
 495                 dst->u.indirect.indirect_op = src->u.indirect.indirect_op;
 496                 dst->u.indirect.nr_segments = src->u.indirect.nr_segments;
 497                 dst->u.indirect.handle = src->u.indirect.handle;
 498                 dst->u.indirect.id = src->u.indirect.id;
 499                 dst->u.indirect.sector_number = src->u.indirect.sector_number;
 500                 barrier();
 501                 j = min(MAX_INDIRECT_PAGES, INDIRECT_PAGES(dst->u.indirect.nr_segments));
 502                 for (i = 0; i < j; i++)
 503                         dst->u.indirect.indirect_grefs[i] =
 504                                 src->u.indirect.indirect_grefs[i];
 505                 break;
 506         default:
 507                 /*
 508                  * Don't know how to translate this op. Only get the
 509                  * ID so failure can be reported to the frontend.
 510                  */
 511                 dst->u.other.id = src->u.other.id;
 512                 break;
 513         }
 514 }
 515
 516 #endif /* __XEN_BLKIF__BACKEND__COMMON_H__ */