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