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:
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:
15 * The above copyright notice and this permission notice shall be included in
16 * all copies or substantial portions of the Software.
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
27 #ifndef __XEN_BLKIF__BACKEND__COMMON_H__
28 #define __XEN_BLKIF__BACKEND__COMMON_H__
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>
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>
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>
48 extern unsigned int xen_blkif_max_ring_order;
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.
53 #define MAX_INDIRECT_SEGMENTS 256
56 * Xen use 4K pages. The guest may use different page size (4K or 64K)
57 * Number of Xen pages per segment
59 #define XEN_PAGES_PER_SEGMENT (PAGE_SIZE / XEN_PAGE_SIZE)
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)
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)
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 {
78 /* i386 protocol version */
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__));
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) */
94 } __attribute__((__packed__));
96 struct blkif_x86_32_request_other {
99 uint64_t id; /* private guest value, echoed in resp */
100 } __attribute__((__packed__));
102 struct blkif_x86_32_request_indirect {
104 uint16_t nr_segments;
106 blkif_sector_t sector_number;
109 grant_ref_t indirect_grefs[BLKIF_MAX_INDIRECT_PAGES_PER_REQUEST];
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.
117 uint64_t _pad2; /* make it 64 byte aligned */
118 } __attribute__((__packed__));
120 struct blkif_x86_32_request {
121 uint8_t operation; /* BLKIF_OP_??? */
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;
128 } __attribute__((__packed__));
130 /* x86_64 protocol version */
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 */
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__));
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 */
146 blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */
148 } __attribute__((__packed__));
150 struct blkif_x86_64_request_other {
153 uint32_t _pad3; /* offsetof(blkif_..,u.discard.id)==8 */
154 uint64_t id; /* private guest value, echoed in resp */
155 } __attribute__((__packed__));
157 struct blkif_x86_64_request_indirect {
159 uint16_t nr_segments;
160 uint32_t _pad1; /* offsetof(blkif_..,u.indirect.id)==8 */
162 blkif_sector_t sector_number;
165 grant_ref_t indirect_grefs[BLKIF_MAX_INDIRECT_PAGES_PER_REQUEST];
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.
173 uint32_t _pad3; /* make it 64 byte aligned */
174 } __attribute__((__packed__));
176 struct blkif_x86_64_request {
177 uint8_t operation; /* BLKIF_OP_??? */
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;
184 } __attribute__((__packed__));
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);
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;
200 enum blkif_protocol {
201 BLKIF_PROTOCOL_NATIVE = 1,
202 BLKIF_PROTOCOL_X86_32 = 2,
203 BLKIF_PROTOCOL_X86_64 = 3,
207 * Default protocol if the frontend doesn't specify one.
210 # define BLKIF_PROTOCOL_DEFAULT BLKIF_PROTOCOL_X86_32
212 # define BLKIF_PROTOCOL_DEFAULT BLKIF_PROTOCOL_NATIVE
216 /* What the domain refers to this vbd as. */
218 /* Non-zero -> read-only */
219 unsigned char readonly;
222 /* phys device that this vbd maps to. */
224 struct block_device *bdev;
225 /* Cached size parameter. */
227 unsigned int flush_support:1;
228 unsigned int discard_secure:1;
229 unsigned int feature_gnt_persistent:1;
230 unsigned int overflow_max_grants:1;
235 /* Number of available flags */
236 #define PERSISTENT_GNT_FLAGS_SIZE 2
237 /* This persistent grant is currently in use */
238 #define PERSISTENT_GNT_ACTIVE 0
240 * This persistent grant has been used, this flag is set when we remove the
241 * PERSISTENT_GNT_ACTIVE, to know that this grant has been used recently.
243 #define PERSISTENT_GNT_WAS_ACTIVE 1
245 /* Number of requests that we can fit in a ring */
246 #define XEN_BLKIF_REQS_PER_PAGE 32
248 struct persistent_gnt {
251 grant_handle_t handle;
252 DECLARE_BITMAP(flags, PERSISTENT_GNT_FLAGS_SIZE);
254 struct list_head remove_node;
258 /* Unique identifier for this interface. */
261 /* Physical parameters of the comms window. */
263 /* Comms information. */
264 enum blkif_protocol blk_protocol;
265 union blkif_back_rings blk_rings;
267 /* The VBD attached to this interface. */
269 /* Back pointer to the backend_info. */
270 struct backend_info *be;
271 /* Private fields. */
272 spinlock_t blk_ring_lock;
275 wait_queue_head_t wq;
276 /* for barrier (drain) requests */
277 struct completion drain_complete;
280 /* One thread per one blkif. */
281 struct task_struct *xenblkd;
282 unsigned int waiting_reqs;
284 /* tree to store persistent grants */
285 struct rb_root persistent_gnts;
286 unsigned int persistent_gnt_c;
287 atomic_t persistent_gnt_in_use;
288 unsigned long next_lru;
290 /* used by the kworker that offload work from the persistent purge */
291 struct list_head persistent_purge_list;
292 struct work_struct persistent_purge_work;
294 /* buffer of free pages to map grant refs */
295 spinlock_t free_pages_lock;
297 struct list_head free_pages;
299 /* List of all 'pending_req' available */
300 struct list_head pending_free;
301 /* And its spinlock. */
302 spinlock_t pending_free_lock;
303 wait_queue_head_t pending_free_wq;
306 unsigned long st_print;
307 unsigned long long st_rd_req;
308 unsigned long long st_wr_req;
309 unsigned long long st_oo_req;
310 unsigned long long st_f_req;
311 unsigned long long st_ds_req;
312 unsigned long long st_rd_sect;
313 unsigned long long st_wr_sect;
315 struct work_struct free_work;
316 /* Thread shutdown wait queue. */
317 wait_queue_head_t shutdown_wq;
318 unsigned int nr_ring_pages;
322 unsigned long offset;
328 struct persistent_gnt *persistent_gnt;
329 grant_handle_t handle;
334 * Each outstanding request that we've passed to the lower device layers has a
335 * 'pending_req' allocated to it. Each buffer_head that completes decrements
336 * the pendcnt towards zero. When it hits zero, the specified domain has a
337 * response queued for it, with the saved 'id' passed back.
340 struct xen_blkif *blkif;
344 unsigned short operation;
346 struct list_head free_list;
347 struct grant_page *segments[MAX_INDIRECT_SEGMENTS];
348 /* Indirect descriptors */
349 struct grant_page *indirect_pages[MAX_INDIRECT_PAGES];
350 struct seg_buf seg[MAX_INDIRECT_SEGMENTS];
351 struct bio *biolist[MAX_INDIRECT_SEGMENTS];
352 struct gnttab_unmap_grant_ref unmap[MAX_INDIRECT_SEGMENTS];
353 struct page *unmap_pages[MAX_INDIRECT_SEGMENTS];
354 struct gntab_unmap_queue_data gnttab_unmap_data;
358 #define vbd_sz(_v) ((_v)->bdev->bd_part ? \
359 (_v)->bdev->bd_part->nr_sects : \
360 get_capacity((_v)->bdev->bd_disk))
362 #define xen_blkif_get(_b) (atomic_inc(&(_b)->refcnt))
363 #define xen_blkif_put(_b) \
365 if (atomic_dec_and_test(&(_b)->refcnt)) \
366 schedule_work(&(_b)->free_work);\
371 blkif_sector_t nr_sects;
372 struct block_device *bdev;
373 blkif_sector_t sector_number;
375 int xen_blkif_interface_init(void);
377 int xen_blkif_xenbus_init(void);
379 irqreturn_t xen_blkif_be_int(int irq, void *dev_id);
380 int xen_blkif_schedule(void *arg);
381 int xen_blkif_purge_persistent(void *arg);
382 void xen_blkbk_free_caches(struct xen_blkif *blkif);
384 int xen_blkbk_flush_diskcache(struct xenbus_transaction xbt,
385 struct backend_info *be, int state);
387 int xen_blkbk_barrier(struct xenbus_transaction xbt,
388 struct backend_info *be, int state);
389 struct xenbus_device *xen_blkbk_xenbus(struct backend_info *be);
390 void xen_blkbk_unmap_purged_grants(struct work_struct *work);
392 static inline void blkif_get_x86_32_req(struct blkif_request *dst,
393 struct blkif_x86_32_request *src)
395 int i, n = BLKIF_MAX_SEGMENTS_PER_REQUEST, j;
396 dst->operation = READ_ONCE(src->operation);
397 switch (dst->operation) {
400 case BLKIF_OP_WRITE_BARRIER:
401 case BLKIF_OP_FLUSH_DISKCACHE:
402 dst->u.rw.nr_segments = src->u.rw.nr_segments;
403 dst->u.rw.handle = src->u.rw.handle;
404 dst->u.rw.id = src->u.rw.id;
405 dst->u.rw.sector_number = src->u.rw.sector_number;
407 if (n > dst->u.rw.nr_segments)
408 n = dst->u.rw.nr_segments;
409 for (i = 0; i < n; i++)
410 dst->u.rw.seg[i] = src->u.rw.seg[i];
412 case BLKIF_OP_DISCARD:
413 dst->u.discard.flag = src->u.discard.flag;
414 dst->u.discard.id = src->u.discard.id;
415 dst->u.discard.sector_number = src->u.discard.sector_number;
416 dst->u.discard.nr_sectors = src->u.discard.nr_sectors;
418 case BLKIF_OP_INDIRECT:
419 dst->u.indirect.indirect_op = src->u.indirect.indirect_op;
420 dst->u.indirect.nr_segments = src->u.indirect.nr_segments;
421 dst->u.indirect.handle = src->u.indirect.handle;
422 dst->u.indirect.id = src->u.indirect.id;
423 dst->u.indirect.sector_number = src->u.indirect.sector_number;
425 j = min(MAX_INDIRECT_PAGES, INDIRECT_PAGES(dst->u.indirect.nr_segments));
426 for (i = 0; i < j; i++)
427 dst->u.indirect.indirect_grefs[i] =
428 src->u.indirect.indirect_grefs[i];
432 * Don't know how to translate this op. Only get the
433 * ID so failure can be reported to the frontend.
435 dst->u.other.id = src->u.other.id;
440 static inline void blkif_get_x86_64_req(struct blkif_request *dst,
441 struct blkif_x86_64_request *src)
443 int i, n = BLKIF_MAX_SEGMENTS_PER_REQUEST, j;
444 dst->operation = READ_ONCE(src->operation);
445 switch (dst->operation) {
448 case BLKIF_OP_WRITE_BARRIER:
449 case BLKIF_OP_FLUSH_DISKCACHE:
450 dst->u.rw.nr_segments = src->u.rw.nr_segments;
451 dst->u.rw.handle = src->u.rw.handle;
452 dst->u.rw.id = src->u.rw.id;
453 dst->u.rw.sector_number = src->u.rw.sector_number;
455 if (n > dst->u.rw.nr_segments)
456 n = dst->u.rw.nr_segments;
457 for (i = 0; i < n; i++)
458 dst->u.rw.seg[i] = src->u.rw.seg[i];
460 case BLKIF_OP_DISCARD:
461 dst->u.discard.flag = src->u.discard.flag;
462 dst->u.discard.id = src->u.discard.id;
463 dst->u.discard.sector_number = src->u.discard.sector_number;
464 dst->u.discard.nr_sectors = src->u.discard.nr_sectors;
466 case BLKIF_OP_INDIRECT:
467 dst->u.indirect.indirect_op = src->u.indirect.indirect_op;
468 dst->u.indirect.nr_segments = src->u.indirect.nr_segments;
469 dst->u.indirect.handle = src->u.indirect.handle;
470 dst->u.indirect.id = src->u.indirect.id;
471 dst->u.indirect.sector_number = src->u.indirect.sector_number;
473 j = min(MAX_INDIRECT_PAGES, INDIRECT_PAGES(dst->u.indirect.nr_segments));
474 for (i = 0; i < j; i++)
475 dst->u.indirect.indirect_grefs[i] =
476 src->u.indirect.indirect_grefs[i];
480 * Don't know how to translate this op. Only get the
481 * ID so failure can be reported to the frontend.
483 dst->u.other.id = src->u.other.id;
488 #endif /* __XEN_BLKIF__BACKEND__COMMON_H__ */