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 /* Not a real protocol.  Used to generate ring structs which contain
54  * the elements common to all protocols only.  This way we get a
55  * compiler-checkable way to use common struct elements, so we can
56  * avoid using switch(protocol) in a number of places.  */
57 struct blkif_common_request {
58 	char dummy;
59 };
60 struct blkif_common_response {
61 	char dummy;
62 };
63 
64 struct blkif_x86_32_request_rw {
65 	uint8_t        nr_segments;  /* number of segments                   */
66 	blkif_vdev_t   handle;       /* only for read/write requests         */
67 	uint64_t       id;           /* private guest value, echoed in resp  */
68 	blkif_sector_t sector_number;/* start sector idx on disk (r/w only)  */
69 	struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
70 } __attribute__((__packed__));
71 
72 struct blkif_x86_32_request_discard {
73 	uint8_t        flag;         /* BLKIF_DISCARD_SECURE or zero         */
74 	blkif_vdev_t   _pad1;        /* was "handle" for read/write requests */
75 	uint64_t       id;           /* private guest value, echoed in resp  */
76 	blkif_sector_t sector_number;/* start sector idx on disk (r/w only)  */
77 	uint64_t       nr_sectors;
78 } __attribute__((__packed__));
79 
80 struct blkif_x86_32_request {
81 	uint8_t        operation;    /* BLKIF_OP_???                         */
82 	union {
83 		struct blkif_x86_32_request_rw rw;
84 		struct blkif_x86_32_request_discard discard;
85 	} u;
86 } __attribute__((__packed__));
87 
88 /* i386 protocol version */
89 #pragma pack(push, 4)
90 struct blkif_x86_32_response {
91 	uint64_t        id;              /* copied from request */
92 	uint8_t         operation;       /* copied from request */
93 	int16_t         status;          /* BLKIF_RSP_???       */
94 };
95 #pragma pack(pop)
96 /* x86_64 protocol version */
97 
98 struct blkif_x86_64_request_rw {
99 	uint8_t        nr_segments;  /* number of segments                   */
100 	blkif_vdev_t   handle;       /* only for read/write requests         */
101 	uint32_t       _pad1;        /* offsetof(blkif_reqest..,u.rw.id)==8  */
102 	uint64_t       id;
103 	blkif_sector_t sector_number;/* start sector idx on disk (r/w only)  */
104 	struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
105 } __attribute__((__packed__));
106 
107 struct blkif_x86_64_request_discard {
108 	uint8_t        flag;         /* BLKIF_DISCARD_SECURE or zero         */
109 	blkif_vdev_t   _pad1;        /* was "handle" for read/write requests */
110         uint32_t       _pad2;        /* offsetof(blkif_..,u.discard.id)==8   */
111 	uint64_t       id;
112 	blkif_sector_t sector_number;/* start sector idx on disk (r/w only)  */
113 	uint64_t       nr_sectors;
114 } __attribute__((__packed__));
115 
116 struct blkif_x86_64_request {
117 	uint8_t        operation;    /* BLKIF_OP_???                         */
118 	union {
119 		struct blkif_x86_64_request_rw rw;
120 		struct blkif_x86_64_request_discard discard;
121 	} u;
122 } __attribute__((__packed__));
123 
124 struct blkif_x86_64_response {
125 	uint64_t       __attribute__((__aligned__(8))) id;
126 	uint8_t         operation;       /* copied from request */
127 	int16_t         status;          /* BLKIF_RSP_???       */
128 };
129 
130 DEFINE_RING_TYPES(blkif_common, struct blkif_common_request,
131 		  struct blkif_common_response);
132 DEFINE_RING_TYPES(blkif_x86_32, struct blkif_x86_32_request,
133 		  struct blkif_x86_32_response);
134 DEFINE_RING_TYPES(blkif_x86_64, struct blkif_x86_64_request,
135 		  struct blkif_x86_64_response);
136 
137 union blkif_back_rings {
138 	struct blkif_back_ring        native;
139 	struct blkif_common_back_ring common;
140 	struct blkif_x86_32_back_ring x86_32;
141 	struct blkif_x86_64_back_ring x86_64;
142 };
143 
144 enum blkif_protocol {
145 	BLKIF_PROTOCOL_NATIVE = 1,
146 	BLKIF_PROTOCOL_X86_32 = 2,
147 	BLKIF_PROTOCOL_X86_64 = 3,
148 };
149 
150 struct xen_vbd {
151 	/* What the domain refers to this vbd as. */
152 	blkif_vdev_t		handle;
153 	/* Non-zero -> read-only */
154 	unsigned char		readonly;
155 	/* VDISK_xxx */
156 	unsigned char		type;
157 	/* phys device that this vbd maps to. */
158 	u32			pdevice;
159 	struct block_device	*bdev;
160 	/* Cached size parameter. */
161 	sector_t		size;
162 	unsigned int		flush_support:1;
163 	unsigned int		discard_secure:1;
164 	unsigned int		feature_gnt_persistent:1;
165 	unsigned int		overflow_max_grants:1;
166 };
167 
168 struct backend_info;
169 
170 
171 struct persistent_gnt {
172 	struct page *page;
173 	grant_ref_t gnt;
174 	grant_handle_t handle;
175 	uint64_t dev_bus_addr;
176 	struct rb_node node;
177 };
178 
179 struct xen_blkif {
180 	/* Unique identifier for this interface. */
181 	domid_t			domid;
182 	unsigned int		handle;
183 	/* Physical parameters of the comms window. */
184 	unsigned int		irq;
185 	/* Comms information. */
186 	enum blkif_protocol	blk_protocol;
187 	union blkif_back_rings	blk_rings;
188 	void			*blk_ring;
189 	/* The VBD attached to this interface. */
190 	struct xen_vbd		vbd;
191 	/* Back pointer to the backend_info. */
192 	struct backend_info	*be;
193 	/* Private fields. */
194 	spinlock_t		blk_ring_lock;
195 	atomic_t		refcnt;
196 
197 	wait_queue_head_t	wq;
198 	/* for barrier (drain) requests */
199 	struct completion	drain_complete;
200 	atomic_t		drain;
201 	/* One thread per one blkif. */
202 	struct task_struct	*xenblkd;
203 	unsigned int		waiting_reqs;
204 
205 	/* tree to store persistent grants */
206 	struct rb_root		persistent_gnts;
207 	unsigned int		persistent_gnt_c;
208 
209 	/* statistics */
210 	unsigned long		st_print;
211 	int			st_rd_req;
212 	int			st_wr_req;
213 	int			st_oo_req;
214 	int			st_f_req;
215 	int			st_ds_req;
216 	int			st_rd_sect;
217 	int			st_wr_sect;
218 
219 	wait_queue_head_t	waiting_to_free;
220 };
221 
222 
223 #define vbd_sz(_v)	((_v)->bdev->bd_part ? \
224 			 (_v)->bdev->bd_part->nr_sects : \
225 			  get_capacity((_v)->bdev->bd_disk))
226 
227 #define xen_blkif_get(_b) (atomic_inc(&(_b)->refcnt))
228 #define xen_blkif_put(_b)				\
229 	do {						\
230 		if (atomic_dec_and_test(&(_b)->refcnt))	\
231 			wake_up(&(_b)->waiting_to_free);\
232 	} while (0)
233 
234 struct phys_req {
235 	unsigned short		dev;
236 	blkif_sector_t		nr_sects;
237 	struct block_device	*bdev;
238 	blkif_sector_t		sector_number;
239 };
240 int xen_blkif_interface_init(void);
241 
242 int xen_blkif_xenbus_init(void);
243 
244 irqreturn_t xen_blkif_be_int(int irq, void *dev_id);
245 int xen_blkif_schedule(void *arg);
246 
247 int xen_blkbk_flush_diskcache(struct xenbus_transaction xbt,
248 			      struct backend_info *be, int state);
249 
250 int xen_blkbk_barrier(struct xenbus_transaction xbt,
251 		      struct backend_info *be, int state);
252 struct xenbus_device *xen_blkbk_xenbus(struct backend_info *be);
253 
254 static inline void blkif_get_x86_32_req(struct blkif_request *dst,
255 					struct blkif_x86_32_request *src)
256 {
257 	int i, n = BLKIF_MAX_SEGMENTS_PER_REQUEST;
258 	dst->operation = src->operation;
259 	switch (src->operation) {
260 	case BLKIF_OP_READ:
261 	case BLKIF_OP_WRITE:
262 	case BLKIF_OP_WRITE_BARRIER:
263 	case BLKIF_OP_FLUSH_DISKCACHE:
264 		dst->u.rw.nr_segments = src->u.rw.nr_segments;
265 		dst->u.rw.handle = src->u.rw.handle;
266 		dst->u.rw.id = src->u.rw.id;
267 		dst->u.rw.sector_number = src->u.rw.sector_number;
268 		barrier();
269 		if (n > dst->u.rw.nr_segments)
270 			n = dst->u.rw.nr_segments;
271 		for (i = 0; i < n; i++)
272 			dst->u.rw.seg[i] = src->u.rw.seg[i];
273 		break;
274 	case BLKIF_OP_DISCARD:
275 		dst->u.discard.flag = src->u.discard.flag;
276 		dst->u.discard.id = src->u.discard.id;
277 		dst->u.discard.sector_number = src->u.discard.sector_number;
278 		dst->u.discard.nr_sectors = src->u.discard.nr_sectors;
279 		break;
280 	default:
281 		break;
282 	}
283 }
284 
285 static inline void blkif_get_x86_64_req(struct blkif_request *dst,
286 					struct blkif_x86_64_request *src)
287 {
288 	int i, n = BLKIF_MAX_SEGMENTS_PER_REQUEST;
289 	dst->operation = src->operation;
290 	switch (src->operation) {
291 	case BLKIF_OP_READ:
292 	case BLKIF_OP_WRITE:
293 	case BLKIF_OP_WRITE_BARRIER:
294 	case BLKIF_OP_FLUSH_DISKCACHE:
295 		dst->u.rw.nr_segments = src->u.rw.nr_segments;
296 		dst->u.rw.handle = src->u.rw.handle;
297 		dst->u.rw.id = src->u.rw.id;
298 		dst->u.rw.sector_number = src->u.rw.sector_number;
299 		barrier();
300 		if (n > dst->u.rw.nr_segments)
301 			n = dst->u.rw.nr_segments;
302 		for (i = 0; i < n; i++)
303 			dst->u.rw.seg[i] = src->u.rw.seg[i];
304 		break;
305 	case BLKIF_OP_DISCARD:
306 		dst->u.discard.flag = src->u.discard.flag;
307 		dst->u.discard.id = src->u.discard.id;
308 		dst->u.discard.sector_number = src->u.discard.sector_number;
309 		dst->u.discard.nr_sectors = src->u.discard.nr_sectors;
310 		break;
311 	default:
312 		break;
313 	}
314 }
315 
316 #endif /* __XEN_BLKIF__BACKEND__COMMON_H__ */
317