jcs.org
qemu with hax to log dma reads & writes
jcs.org/2018/11/12/vfio
1/*
2 * Copyright (C) 2016-2019 Red Hat, Inc.
3 * Copyright (C) 2005 Anthony Liguori <anthony@codemonkey.ws>
4 *
5 * Network Block Device
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; under version 2 of the License.
10 *
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License
17 * along with this program; if not, see <http://www.gnu.org/licenses/>.
18 */
19
20#ifndef NBD_H
21#define NBD_H
22
23#include "qapi/qapi-types-block.h"
24#include "io/channel-socket.h"
25#include "crypto/tlscreds.h"
26#include "qapi/error.h"
27
28/* Handshake phase structs - this struct is passed on the wire */
29
30struct NBDOption {
31 uint64_t magic; /* NBD_OPTS_MAGIC */
32 uint32_t option; /* NBD_OPT_* */
33 uint32_t length;
34} QEMU_PACKED;
35typedef struct NBDOption NBDOption;
36
37struct NBDOptionReply {
38 uint64_t magic; /* NBD_REP_MAGIC */
39 uint32_t option; /* NBD_OPT_* */
40 uint32_t type; /* NBD_REP_* */
41 uint32_t length;
42} QEMU_PACKED;
43typedef struct NBDOptionReply NBDOptionReply;
44
45typedef struct NBDOptionReplyMetaContext {
46 NBDOptionReply h; /* h.type = NBD_REP_META_CONTEXT, h.length > 4 */
47 uint32_t context_id;
48 /* meta context name follows */
49} QEMU_PACKED NBDOptionReplyMetaContext;
50
51/* Transmission phase structs
52 *
53 * Note: these are _NOT_ the same as the network representation of an NBD
54 * request and reply!
55 */
56struct NBDRequest {
57 uint64_t handle;
58 uint64_t from;
59 uint32_t len;
60 uint16_t flags; /* NBD_CMD_FLAG_* */
61 uint16_t type; /* NBD_CMD_* */
62};
63typedef struct NBDRequest NBDRequest;
64
65typedef struct NBDSimpleReply {
66 uint32_t magic; /* NBD_SIMPLE_REPLY_MAGIC */
67 uint32_t error;
68 uint64_t handle;
69} QEMU_PACKED NBDSimpleReply;
70
71/* Header of all structured replies */
72typedef struct NBDStructuredReplyChunk {
73 uint32_t magic; /* NBD_STRUCTURED_REPLY_MAGIC */
74 uint16_t flags; /* combination of NBD_REPLY_FLAG_* */
75 uint16_t type; /* NBD_REPLY_TYPE_* */
76 uint64_t handle; /* request handle */
77 uint32_t length; /* length of payload */
78} QEMU_PACKED NBDStructuredReplyChunk;
79
80typedef union NBDReply {
81 NBDSimpleReply simple;
82 NBDStructuredReplyChunk structured;
83 struct {
84 /* @magic and @handle fields have the same offset and size both in
85 * simple reply and structured reply chunk, so let them be accessible
86 * without ".simple." or ".structured." specification
87 */
88 uint32_t magic;
89 uint32_t _skip;
90 uint64_t handle;
91 } QEMU_PACKED;
92} NBDReply;
93
94/* Header of chunk for NBD_REPLY_TYPE_OFFSET_DATA */
95typedef struct NBDStructuredReadData {
96 NBDStructuredReplyChunk h; /* h.length >= 9 */
97 uint64_t offset;
98 /* At least one byte of data payload follows, calculated from h.length */
99} QEMU_PACKED NBDStructuredReadData;
100
101/* Complete chunk for NBD_REPLY_TYPE_OFFSET_HOLE */
102typedef struct NBDStructuredReadHole {
103 NBDStructuredReplyChunk h; /* h.length == 12 */
104 uint64_t offset;
105 uint32_t length;
106} QEMU_PACKED NBDStructuredReadHole;
107
108/* Header of all NBD_REPLY_TYPE_ERROR* errors */
109typedef struct NBDStructuredError {
110 NBDStructuredReplyChunk h; /* h.length >= 6 */
111 uint32_t error;
112 uint16_t message_length;
113} QEMU_PACKED NBDStructuredError;
114
115/* Header of NBD_REPLY_TYPE_BLOCK_STATUS */
116typedef struct NBDStructuredMeta {
117 NBDStructuredReplyChunk h; /* h.length >= 12 (at least one extent) */
118 uint32_t context_id;
119 /* extents follows */
120} QEMU_PACKED NBDStructuredMeta;
121
122/* Extent chunk for NBD_REPLY_TYPE_BLOCK_STATUS */
123typedef struct NBDExtent {
124 uint32_t length;
125 uint32_t flags; /* NBD_STATE_* */
126} QEMU_PACKED NBDExtent;
127
128/* Transmission (export) flags: sent from server to client during handshake,
129 but describe what will happen during transmission */
130enum {
131 NBD_FLAG_HAS_FLAGS_BIT = 0, /* Flags are there */
132 NBD_FLAG_READ_ONLY_BIT = 1, /* Device is read-only */
133 NBD_FLAG_SEND_FLUSH_BIT = 2, /* Send FLUSH */
134 NBD_FLAG_SEND_FUA_BIT = 3, /* Send FUA (Force Unit Access) */
135 NBD_FLAG_ROTATIONAL_BIT = 4, /* Use elevator algorithm -
136 rotational media */
137 NBD_FLAG_SEND_TRIM_BIT = 5, /* Send TRIM (discard) */
138 NBD_FLAG_SEND_WRITE_ZEROES_BIT = 6, /* Send WRITE_ZEROES */
139 NBD_FLAG_SEND_DF_BIT = 7, /* Send DF (Do not Fragment) */
140 NBD_FLAG_CAN_MULTI_CONN_BIT = 8, /* Multi-client cache consistent */
141 NBD_FLAG_SEND_RESIZE_BIT = 9, /* Send resize */
142 NBD_FLAG_SEND_CACHE_BIT = 10, /* Send CACHE (prefetch) */
143 NBD_FLAG_SEND_FAST_ZERO_BIT = 11, /* FAST_ZERO flag for WRITE_ZEROES */
144};
145
146#define NBD_FLAG_HAS_FLAGS (1 << NBD_FLAG_HAS_FLAGS_BIT)
147#define NBD_FLAG_READ_ONLY (1 << NBD_FLAG_READ_ONLY_BIT)
148#define NBD_FLAG_SEND_FLUSH (1 << NBD_FLAG_SEND_FLUSH_BIT)
149#define NBD_FLAG_SEND_FUA (1 << NBD_FLAG_SEND_FUA_BIT)
150#define NBD_FLAG_ROTATIONAL (1 << NBD_FLAG_ROTATIONAL_BIT)
151#define NBD_FLAG_SEND_TRIM (1 << NBD_FLAG_SEND_TRIM_BIT)
152#define NBD_FLAG_SEND_WRITE_ZEROES (1 << NBD_FLAG_SEND_WRITE_ZEROES_BIT)
153#define NBD_FLAG_SEND_DF (1 << NBD_FLAG_SEND_DF_BIT)
154#define NBD_FLAG_CAN_MULTI_CONN (1 << NBD_FLAG_CAN_MULTI_CONN_BIT)
155#define NBD_FLAG_SEND_RESIZE (1 << NBD_FLAG_SEND_RESIZE_BIT)
156#define NBD_FLAG_SEND_CACHE (1 << NBD_FLAG_SEND_CACHE_BIT)
157#define NBD_FLAG_SEND_FAST_ZERO (1 << NBD_FLAG_SEND_FAST_ZERO_BIT)
158
159/* New-style handshake (global) flags, sent from server to client, and
160 control what will happen during handshake phase. */
161#define NBD_FLAG_FIXED_NEWSTYLE (1 << 0) /* Fixed newstyle protocol. */
162#define NBD_FLAG_NO_ZEROES (1 << 1) /* End handshake without zeroes. */
163
164/* New-style client flags, sent from client to server to control what happens
165 during handshake phase. */
166#define NBD_FLAG_C_FIXED_NEWSTYLE (1 << 0) /* Fixed newstyle protocol. */
167#define NBD_FLAG_C_NO_ZEROES (1 << 1) /* End handshake without zeroes. */
168
169/* Option requests. */
170#define NBD_OPT_EXPORT_NAME (1)
171#define NBD_OPT_ABORT (2)
172#define NBD_OPT_LIST (3)
173/* #define NBD_OPT_PEEK_EXPORT (4) not in use */
174#define NBD_OPT_STARTTLS (5)
175#define NBD_OPT_INFO (6)
176#define NBD_OPT_GO (7)
177#define NBD_OPT_STRUCTURED_REPLY (8)
178#define NBD_OPT_LIST_META_CONTEXT (9)
179#define NBD_OPT_SET_META_CONTEXT (10)
180
181/* Option reply types. */
182#define NBD_REP_ERR(value) ((UINT32_C(1) << 31) | (value))
183
184#define NBD_REP_ACK (1) /* Data sending finished. */
185#define NBD_REP_SERVER (2) /* Export description. */
186#define NBD_REP_INFO (3) /* NBD_OPT_INFO/GO. */
187#define NBD_REP_META_CONTEXT (4) /* NBD_OPT_{LIST,SET}_META_CONTEXT */
188
189#define NBD_REP_ERR_UNSUP NBD_REP_ERR(1) /* Unknown option */
190#define NBD_REP_ERR_POLICY NBD_REP_ERR(2) /* Server denied */
191#define NBD_REP_ERR_INVALID NBD_REP_ERR(3) /* Invalid length */
192#define NBD_REP_ERR_PLATFORM NBD_REP_ERR(4) /* Not compiled in */
193#define NBD_REP_ERR_TLS_REQD NBD_REP_ERR(5) /* TLS required */
194#define NBD_REP_ERR_UNKNOWN NBD_REP_ERR(6) /* Export unknown */
195#define NBD_REP_ERR_SHUTDOWN NBD_REP_ERR(7) /* Server shutting down */
196#define NBD_REP_ERR_BLOCK_SIZE_REQD NBD_REP_ERR(8) /* Need INFO_BLOCK_SIZE */
197
198/* Info types, used during NBD_REP_INFO */
199#define NBD_INFO_EXPORT 0
200#define NBD_INFO_NAME 1
201#define NBD_INFO_DESCRIPTION 2
202#define NBD_INFO_BLOCK_SIZE 3
203
204/* Request flags, sent from client to server during transmission phase */
205#define NBD_CMD_FLAG_FUA (1 << 0) /* 'force unit access' during write */
206#define NBD_CMD_FLAG_NO_HOLE (1 << 1) /* don't punch hole on zero run */
207#define NBD_CMD_FLAG_DF (1 << 2) /* don't fragment structured read */
208#define NBD_CMD_FLAG_REQ_ONE (1 << 3) /* only one extent in BLOCK_STATUS
209 * reply chunk */
210#define NBD_CMD_FLAG_FAST_ZERO (1 << 4) /* fail if WRITE_ZEROES is not fast */
211
212/* Supported request types */
213enum {
214 NBD_CMD_READ = 0,
215 NBD_CMD_WRITE = 1,
216 NBD_CMD_DISC = 2,
217 NBD_CMD_FLUSH = 3,
218 NBD_CMD_TRIM = 4,
219 NBD_CMD_CACHE = 5,
220 NBD_CMD_WRITE_ZEROES = 6,
221 NBD_CMD_BLOCK_STATUS = 7,
222};
223
224#define NBD_DEFAULT_PORT 10809
225
226/* Maximum size of a single READ/WRITE data buffer */
227#define NBD_MAX_BUFFER_SIZE (32 * 1024 * 1024)
228
229/*
230 * Maximum size of a protocol string (export name, meta context name,
231 * etc.). Use malloc rather than stack allocation for storage of a
232 * string.
233 */
234#define NBD_MAX_STRING_SIZE 4096
235
236/* Two types of reply structures */
237#define NBD_SIMPLE_REPLY_MAGIC 0x67446698
238#define NBD_STRUCTURED_REPLY_MAGIC 0x668e33ef
239
240/* Structured reply flags */
241#define NBD_REPLY_FLAG_DONE (1 << 0) /* This reply-chunk is last */
242
243/* Structured reply types */
244#define NBD_REPLY_ERR(value) ((1 << 15) | (value))
245
246#define NBD_REPLY_TYPE_NONE 0
247#define NBD_REPLY_TYPE_OFFSET_DATA 1
248#define NBD_REPLY_TYPE_OFFSET_HOLE 2
249#define NBD_REPLY_TYPE_BLOCK_STATUS 5
250#define NBD_REPLY_TYPE_ERROR NBD_REPLY_ERR(1)
251#define NBD_REPLY_TYPE_ERROR_OFFSET NBD_REPLY_ERR(2)
252
253/* Extent flags for base:allocation in NBD_REPLY_TYPE_BLOCK_STATUS */
254#define NBD_STATE_HOLE (1 << 0)
255#define NBD_STATE_ZERO (1 << 1)
256
257/* Extent flags for qemu:dirty-bitmap in NBD_REPLY_TYPE_BLOCK_STATUS */
258#define NBD_STATE_DIRTY (1 << 0)
259
260static inline bool nbd_reply_type_is_error(int type)
261{
262 return type & (1 << 15);
263}
264
265/* NBD errors are based on errno numbers, so there is a 1:1 mapping,
266 * but only a limited set of errno values is specified in the protocol.
267 * Everything else is squashed to EINVAL.
268 */
269#define NBD_SUCCESS 0
270#define NBD_EPERM 1
271#define NBD_EIO 5
272#define NBD_ENOMEM 12
273#define NBD_EINVAL 22
274#define NBD_ENOSPC 28
275#define NBD_EOVERFLOW 75
276#define NBD_ENOTSUP 95
277#define NBD_ESHUTDOWN 108
278
279/* Details collected by NBD_OPT_EXPORT_NAME and NBD_OPT_GO */
280struct NBDExportInfo {
281 /* Set by client before nbd_receive_negotiate() */
282 bool request_sizes;
283 char *x_dirty_bitmap;
284
285 /* Set by client before nbd_receive_negotiate(), or by server results
286 * during nbd_receive_export_list() */
287 char *name; /* must be non-NULL */
288
289 /* In-out fields, set by client before nbd_receive_negotiate() and
290 * updated by server results during nbd_receive_negotiate() */
291 bool structured_reply;
292 bool base_allocation; /* base:allocation context for NBD_CMD_BLOCK_STATUS */
293
294 /* Set by server results during nbd_receive_negotiate() and
295 * nbd_receive_export_list() */
296 uint64_t size;
297 uint16_t flags;
298 uint32_t min_block;
299 uint32_t opt_block;
300 uint32_t max_block;
301
302 uint32_t context_id;
303
304 /* Set by server results during nbd_receive_export_list() */
305 char *description;
306 int n_contexts;
307 char **contexts;
308};
309typedef struct NBDExportInfo NBDExportInfo;
310
311int nbd_receive_negotiate(AioContext *aio_context, QIOChannel *ioc,
312 QCryptoTLSCreds *tlscreds,
313 const char *hostname, QIOChannel **outioc,
314 NBDExportInfo *info, Error **errp);
315void nbd_free_export_list(NBDExportInfo *info, int count);
316int nbd_receive_export_list(QIOChannel *ioc, QCryptoTLSCreds *tlscreds,
317 const char *hostname, NBDExportInfo **info,
318 Error **errp);
319int nbd_init(int fd, QIOChannelSocket *sioc, NBDExportInfo *info,
320 Error **errp);
321int nbd_send_request(QIOChannel *ioc, NBDRequest *request);
322int coroutine_fn nbd_receive_reply(BlockDriverState *bs, QIOChannel *ioc,
323 NBDReply *reply, Error **errp);
324int nbd_client(int fd);
325int nbd_disconnect(int fd);
326int nbd_errno_to_system_errno(int err);
327
328typedef struct NBDExport NBDExport;
329typedef struct NBDClient NBDClient;
330
331NBDExport *nbd_export_new(BlockDriverState *bs, uint64_t dev_offset,
332 uint64_t size, const char *name, const char *desc,
333 const char *bitmap, bool readonly, bool shared,
334 void (*close)(NBDExport *), bool writethrough,
335 BlockBackend *on_eject_blk, Error **errp);
336void nbd_export_close(NBDExport *exp);
337void nbd_export_remove(NBDExport *exp, NbdServerRemoveMode mode, Error **errp);
338void nbd_export_get(NBDExport *exp);
339void nbd_export_put(NBDExport *exp);
340
341BlockBackend *nbd_export_get_blockdev(NBDExport *exp);
342
343AioContext *nbd_export_aio_context(NBDExport *exp);
344NBDExport *nbd_export_find(const char *name);
345void nbd_export_close_all(void);
346
347void nbd_client_new(QIOChannelSocket *sioc,
348 QCryptoTLSCreds *tlscreds,
349 const char *tlsauthz,
350 void (*close_fn)(NBDClient *, bool));
351void nbd_client_get(NBDClient *client);
352void nbd_client_put(NBDClient *client);
353
354void nbd_server_start(SocketAddress *addr, const char *tls_creds,
355 const char *tls_authz, Error **errp);
356void nbd_server_start_options(NbdServerOptions *arg, Error **errp);
357
358/* nbd_read
359 * Reads @size bytes from @ioc. Returns 0 on success.
360 */
361static inline int nbd_read(QIOChannel *ioc, void *buffer, size_t size,
362 const char *desc, Error **errp)
363{
364 ERRP_GUARD();
365 int ret = qio_channel_read_all(ioc, buffer, size, errp) < 0 ? -EIO : 0;
366
367 if (ret < 0) {
368 if (desc) {
369 error_prepend(errp, "Failed to read %s: ", desc);
370 }
371 return -1;
372 }
373
374 return 0;
375}
376
377#define DEF_NBD_READ_N(bits) \
378static inline int nbd_read##bits(QIOChannel *ioc, \
379 uint##bits##_t *val, \
380 const char *desc, Error **errp) \
381{ \
382 if (nbd_read(ioc, val, sizeof(*val), desc, errp) < 0) { \
383 return -1; \
384 } \
385 *val = be##bits##_to_cpu(*val); \
386 return 0; \
387}
388
389DEF_NBD_READ_N(16) /* Defines nbd_read16(). */
390DEF_NBD_READ_N(32) /* Defines nbd_read32(). */
391DEF_NBD_READ_N(64) /* Defines nbd_read64(). */
392
393#undef DEF_NBD_READ_N
394
395static inline bool nbd_reply_is_simple(NBDReply *reply)
396{
397 return reply->magic == NBD_SIMPLE_REPLY_MAGIC;
398}
399
400static inline bool nbd_reply_is_structured(NBDReply *reply)
401{
402 return reply->magic == NBD_STRUCTURED_REPLY_MAGIC;
403}
404
405const char *nbd_reply_type_lookup(uint16_t type);
406const char *nbd_opt_lookup(uint32_t opt);
407const char *nbd_rep_lookup(uint32_t rep);
408const char *nbd_info_lookup(uint16_t info);
409const char *nbd_cmd_lookup(uint16_t info);
410const char *nbd_err_lookup(int err);
411
412#endif