xref: /openbmc/qemu/hw/9pfs/9p.h (revision 9f6df01d)
1 #ifndef QEMU_9P_H
2 #define QEMU_9P_H
3 
4 #include <dirent.h>
5 #include <utime.h>
6 #include <sys/resource.h>
7 #include "fsdev/file-op-9p.h"
8 #include "fsdev/9p-iov-marshal.h"
9 #include "qemu/thread.h"
10 #include "qemu/coroutine.h"
11 #include "qemu/qht.h"
12 
13 enum {
14     P9_TLERROR = 6,
15     P9_RLERROR,
16     P9_TSTATFS = 8,
17     P9_RSTATFS,
18     P9_TLOPEN = 12,
19     P9_RLOPEN,
20     P9_TLCREATE = 14,
21     P9_RLCREATE,
22     P9_TSYMLINK = 16,
23     P9_RSYMLINK,
24     P9_TMKNOD = 18,
25     P9_RMKNOD,
26     P9_TRENAME = 20,
27     P9_RRENAME,
28     P9_TREADLINK = 22,
29     P9_RREADLINK,
30     P9_TGETATTR = 24,
31     P9_RGETATTR,
32     P9_TSETATTR = 26,
33     P9_RSETATTR,
34     P9_TXATTRWALK = 30,
35     P9_RXATTRWALK,
36     P9_TXATTRCREATE = 32,
37     P9_RXATTRCREATE,
38     P9_TREADDIR = 40,
39     P9_RREADDIR,
40     P9_TFSYNC = 50,
41     P9_RFSYNC,
42     P9_TLOCK = 52,
43     P9_RLOCK,
44     P9_TGETLOCK = 54,
45     P9_RGETLOCK,
46     P9_TLINK = 70,
47     P9_RLINK,
48     P9_TMKDIR = 72,
49     P9_RMKDIR,
50     P9_TRENAMEAT = 74,
51     P9_RRENAMEAT,
52     P9_TUNLINKAT = 76,
53     P9_RUNLINKAT,
54     P9_TVERSION = 100,
55     P9_RVERSION,
56     P9_TAUTH = 102,
57     P9_RAUTH,
58     P9_TATTACH = 104,
59     P9_RATTACH,
60     P9_TERROR = 106,
61     P9_RERROR,
62     P9_TFLUSH = 108,
63     P9_RFLUSH,
64     P9_TWALK = 110,
65     P9_RWALK,
66     P9_TOPEN = 112,
67     P9_ROPEN,
68     P9_TCREATE = 114,
69     P9_RCREATE,
70     P9_TREAD = 116,
71     P9_RREAD,
72     P9_TWRITE = 118,
73     P9_RWRITE,
74     P9_TCLUNK = 120,
75     P9_RCLUNK,
76     P9_TREMOVE = 122,
77     P9_RREMOVE,
78     P9_TSTAT = 124,
79     P9_RSTAT,
80     P9_TWSTAT = 126,
81     P9_RWSTAT,
82 };
83 
84 
85 /* qid.types */
86 enum {
87     P9_QTDIR = 0x80,
88     P9_QTAPPEND = 0x40,
89     P9_QTEXCL = 0x20,
90     P9_QTMOUNT = 0x10,
91     P9_QTAUTH = 0x08,
92     P9_QTTMP = 0x04,
93     P9_QTSYMLINK = 0x02,
94     P9_QTLINK = 0x01,
95     P9_QTFILE = 0x00,
96 };
97 
98 typedef enum P9ProtoVersion {
99     V9FS_PROTO_2000U = 0x01,
100     V9FS_PROTO_2000L = 0x02,
101 } P9ProtoVersion;
102 
103 /**
104  * @brief Minimum message size supported by this 9pfs server.
105  *
106  * A client establishes a session by sending a Tversion request along with a
107  * 'msize' parameter which suggests the server a maximum message size ever to be
108  * used for communication (for both requests and replies) between client and
109  * server during that session. If client suggests a 'msize' smaller than this
110  * value then session is denied by server with an error response.
111  */
112 #define P9_MIN_MSIZE    4096
113 
114 #define P9_NOTAG    UINT16_MAX
115 #define P9_NOFID    UINT32_MAX
116 #define P9_MAXWELEM 16
117 
118 #define FID_REFERENCED          0x1
119 #define FID_NON_RECLAIMABLE     0x2
120 static inline char *rpath(FsContext *ctx, const char *path)
121 {
122     return g_strdup_printf("%s/%s", ctx->fs_root, path);
123 }
124 
125 /*
126  * ample room for Twrite/Rread header
127  * size[4] Tread/Twrite tag[2] fid[4] offset[8] count[4]
128  */
129 #define P9_IOHDRSZ 24
130 
131 typedef struct V9fsPDU V9fsPDU;
132 typedef struct V9fsState V9fsState;
133 typedef struct V9fsTransport V9fsTransport;
134 
135 typedef struct {
136     uint32_t size_le;
137     uint8_t id;
138     uint16_t tag_le;
139 } QEMU_PACKED P9MsgHeader;
140 /* According to the specification, 9p messages start with a 7-byte header.
141  * Since most of the code uses this header size in literal form, we must be
142  * sure this is indeed the case.
143  */
144 QEMU_BUILD_BUG_ON(sizeof(P9MsgHeader) != 7);
145 
146 struct V9fsPDU
147 {
148     uint32_t size;
149     uint16_t tag;
150     uint8_t id;
151     uint8_t cancelled;
152     CoQueue complete;
153     V9fsState *s;
154     QLIST_ENTRY(V9fsPDU) next;
155     uint32_t idx;
156 };
157 
158 
159 /* FIXME
160  * 1) change user needs to set groups and stuff
161  */
162 
163 #define MAX_REQ         128
164 #define MAX_TAG_LEN     32
165 
166 #define BUG_ON(cond) assert(!(cond))
167 
168 typedef struct V9fsFidState V9fsFidState;
169 
170 enum {
171     P9_FID_NONE = 0,
172     P9_FID_FILE,
173     P9_FID_DIR,
174     P9_FID_XATTR,
175 };
176 
177 typedef struct V9fsConf
178 {
179     /* tag name for the device */
180     char *tag;
181     char *fsdev_id;
182 } V9fsConf;
183 
184 /* 9p2000.L xattr flags (matches Linux values) */
185 #define P9_XATTR_CREATE 1
186 #define P9_XATTR_REPLACE 2
187 
188 typedef struct V9fsXattr
189 {
190     uint64_t copied_len;
191     uint64_t len;
192     void *value;
193     V9fsString name;
194     int flags;
195     bool xattrwalk_fid;
196 } V9fsXattr;
197 
198 typedef struct V9fsDir {
199     DIR *stream;
200     P9ProtoVersion proto_version;
201     /* readdir mutex type used for 9P2000.u protocol variant */
202     CoMutex readdir_mutex_u;
203     /* readdir mutex type used for 9P2000.L protocol variant */
204     QemuMutex readdir_mutex_L;
205 } V9fsDir;
206 
207 static inline void v9fs_readdir_lock(V9fsDir *dir)
208 {
209     if (dir->proto_version == V9FS_PROTO_2000U) {
210         qemu_co_mutex_lock(&dir->readdir_mutex_u);
211     } else {
212         qemu_mutex_lock(&dir->readdir_mutex_L);
213     }
214 }
215 
216 static inline void v9fs_readdir_unlock(V9fsDir *dir)
217 {
218     if (dir->proto_version == V9FS_PROTO_2000U) {
219         qemu_co_mutex_unlock(&dir->readdir_mutex_u);
220     } else {
221         qemu_mutex_unlock(&dir->readdir_mutex_L);
222     }
223 }
224 
225 static inline void v9fs_readdir_init(P9ProtoVersion proto_version, V9fsDir *dir)
226 {
227     dir->proto_version = proto_version;
228     if (proto_version == V9FS_PROTO_2000U) {
229         qemu_co_mutex_init(&dir->readdir_mutex_u);
230     } else {
231         qemu_mutex_init(&dir->readdir_mutex_L);
232     }
233 }
234 
235 /**
236  * Type for 9p fs drivers' (a.k.a. 9p backends) result of readdir requests,
237  * which is a chained list of directory entries.
238  */
239 typedef struct V9fsDirEnt {
240     /* mandatory (must not be NULL) information for all readdir requests */
241     struct dirent *dent;
242     /*
243      * optional (may be NULL): A full stat of each directory entry is just
244      * done if explicitly told to fs driver.
245      */
246     struct stat *st;
247     /*
248      * instead of an array, directory entries are always returned as
249      * chained list, that's because the amount of entries retrieved by fs
250      * drivers is dependent on the individual entries' name (since response
251      * messages are size limited), so the final amount cannot be estimated
252      * before hand
253      */
254     struct V9fsDirEnt *next;
255 } V9fsDirEnt;
256 
257 /*
258  * Filled by fs driver on open and other
259  * calls.
260  */
261 union V9fsFidOpenState {
262     int fd;
263     V9fsDir dir;
264     V9fsXattr xattr;
265     /*
266      * private pointer for fs drivers, that
267      * have its own internal representation of
268      * open files.
269      */
270     void *private;
271 };
272 
273 struct V9fsFidState
274 {
275     int fid_type;
276     int32_t fid;
277     V9fsPath path;
278     V9fsFidOpenState fs;
279     V9fsFidOpenState fs_reclaim;
280     int flags;
281     int open_flags;
282     uid_t uid;
283     int ref;
284     int clunked;
285     V9fsFidState *next;
286     V9fsFidState *rclm_lst;
287 };
288 
289 typedef enum AffixType_t {
290     AffixType_Prefix,
291     AffixType_Suffix, /* A.k.a. postfix. */
292 } AffixType_t;
293 
294 /**
295  * @brief Unique affix of variable length.
296  *
297  * An affix is (currently) either a suffix or a prefix, which is either
298  * going to be prepended (prefix) or appended (suffix) with some other
299  * number for the goal to generate unique numbers. Accordingly the
300  * suffixes (or prefixes) we generate @b must all have the mathematical
301  * property of being suffix-free (or prefix-free in case of prefixes)
302  * so that no matter what number we concatenate the affix with, that we
303  * always reliably get unique numbers as result after concatenation.
304  */
305 typedef struct VariLenAffix {
306     AffixType_t type; /* Whether this affix is a suffix or a prefix. */
307     uint64_t value; /* Actual numerical value of this affix. */
308     /*
309      * Lenght of the affix, that is how many (of the lowest) bits of @c value
310      * must be used for appending/prepending this affix to its final resulting,
311      * unique number.
312      */
313     int bits;
314 } VariLenAffix;
315 
316 /* See qid_inode_prefix_hash_bits(). */
317 typedef struct {
318     dev_t dev; /* FS device on host. */
319     /*
320      * How many (high) bits of the original inode number shall be used for
321      * hashing.
322      */
323     int prefix_bits;
324 } QpdEntry;
325 
326 /* QID path prefix entry, see stat_to_qid */
327 typedef struct {
328     dev_t dev;
329     uint16_t ino_prefix;
330     uint32_t qp_affix_index;
331     VariLenAffix qp_affix;
332 } QppEntry;
333 
334 /* QID path full entry, as above */
335 typedef struct {
336     dev_t dev;
337     ino_t ino;
338     uint64_t path;
339 } QpfEntry;
340 
341 struct V9fsState
342 {
343     QLIST_HEAD(, V9fsPDU) free_list;
344     QLIST_HEAD(, V9fsPDU) active_list;
345     V9fsFidState *fid_list;
346     FileOperations *ops;
347     FsContext ctx;
348     char *tag;
349     P9ProtoVersion proto_version;
350     int32_t msize;
351     V9fsPDU pdus[MAX_REQ];
352     const V9fsTransport *transport;
353     /*
354      * lock ensuring atomic path update
355      * on rename.
356      */
357     CoRwlock rename_lock;
358     int32_t root_fid;
359     Error *migration_blocker;
360     V9fsConf fsconf;
361     V9fsQID root_qid;
362     dev_t dev_id;
363     struct qht qpd_table;
364     struct qht qpp_table;
365     struct qht qpf_table;
366     uint64_t qp_ndevices; /* Amount of entries in qpd_table. */
367     uint16_t qp_affix_next;
368     uint64_t qp_fullpath_next;
369 };
370 
371 /* 9p2000.L open flags */
372 #define P9_DOTL_RDONLY        00000000
373 #define P9_DOTL_WRONLY        00000001
374 #define P9_DOTL_RDWR          00000002
375 #define P9_DOTL_NOACCESS      00000003
376 #define P9_DOTL_CREATE        00000100
377 #define P9_DOTL_EXCL          00000200
378 #define P9_DOTL_NOCTTY        00000400
379 #define P9_DOTL_TRUNC         00001000
380 #define P9_DOTL_APPEND        00002000
381 #define P9_DOTL_NONBLOCK      00004000
382 #define P9_DOTL_DSYNC         00010000
383 #define P9_DOTL_FASYNC        00020000
384 #define P9_DOTL_DIRECT        00040000
385 #define P9_DOTL_LARGEFILE     00100000
386 #define P9_DOTL_DIRECTORY     00200000
387 #define P9_DOTL_NOFOLLOW      00400000
388 #define P9_DOTL_NOATIME       01000000
389 #define P9_DOTL_CLOEXEC       02000000
390 #define P9_DOTL_SYNC          04000000
391 
392 /* 9p2000.L at flags */
393 #define P9_DOTL_AT_REMOVEDIR         0x200
394 
395 /* 9P2000.L lock type */
396 #define P9_LOCK_TYPE_RDLCK 0
397 #define P9_LOCK_TYPE_WRLCK 1
398 #define P9_LOCK_TYPE_UNLCK 2
399 
400 #define P9_LOCK_SUCCESS 0
401 #define P9_LOCK_BLOCKED 1
402 #define P9_LOCK_ERROR 2
403 #define P9_LOCK_GRACE 3
404 
405 #define P9_LOCK_FLAGS_BLOCK 1
406 #define P9_LOCK_FLAGS_RECLAIM 2
407 
408 typedef struct V9fsFlock
409 {
410     uint8_t type;
411     uint32_t flags;
412     uint64_t start; /* absolute offset */
413     uint64_t length;
414     uint32_t proc_id;
415     V9fsString client_id;
416 } V9fsFlock;
417 
418 typedef struct V9fsGetlock
419 {
420     uint8_t type;
421     uint64_t start; /* absolute offset */
422     uint64_t length;
423     uint32_t proc_id;
424     V9fsString client_id;
425 } V9fsGetlock;
426 
427 extern int open_fd_hw;
428 extern int total_open_fd;
429 
430 static inline void v9fs_path_write_lock(V9fsState *s)
431 {
432     if (s->ctx.export_flags & V9FS_PATHNAME_FSCONTEXT) {
433         qemu_co_rwlock_wrlock(&s->rename_lock);
434     }
435 }
436 
437 static inline void v9fs_path_read_lock(V9fsState *s)
438 {
439     if (s->ctx.export_flags & V9FS_PATHNAME_FSCONTEXT) {
440         qemu_co_rwlock_rdlock(&s->rename_lock);
441     }
442 }
443 
444 static inline void v9fs_path_unlock(V9fsState *s)
445 {
446     if (s->ctx.export_flags & V9FS_PATHNAME_FSCONTEXT) {
447         qemu_co_rwlock_unlock(&s->rename_lock);
448     }
449 }
450 
451 static inline uint8_t v9fs_request_cancelled(V9fsPDU *pdu)
452 {
453     return pdu->cancelled;
454 }
455 
456 void coroutine_fn v9fs_reclaim_fd(V9fsPDU *pdu);
457 void v9fs_path_init(V9fsPath *path);
458 void v9fs_path_free(V9fsPath *path);
459 void v9fs_path_sprintf(V9fsPath *path, const char *fmt, ...);
460 void v9fs_path_copy(V9fsPath *dst, const V9fsPath *src);
461 size_t v9fs_readdir_response_size(V9fsString *name);
462 int v9fs_name_to_path(V9fsState *s, V9fsPath *dirpath,
463                       const char *name, V9fsPath *path);
464 int v9fs_device_realize_common(V9fsState *s, const V9fsTransport *t,
465                                Error **errp);
466 void v9fs_device_unrealize_common(V9fsState *s);
467 
468 V9fsPDU *pdu_alloc(V9fsState *s);
469 void pdu_free(V9fsPDU *pdu);
470 void pdu_submit(V9fsPDU *pdu, P9MsgHeader *hdr);
471 void v9fs_reset(V9fsState *s);
472 
473 struct V9fsTransport {
474     ssize_t     (*pdu_vmarshal)(V9fsPDU *pdu, size_t offset, const char *fmt,
475                                 va_list ap);
476     ssize_t     (*pdu_vunmarshal)(V9fsPDU *pdu, size_t offset, const char *fmt,
477                                   va_list ap);
478     void        (*init_in_iov_from_pdu)(V9fsPDU *pdu, struct iovec **piov,
479                                         unsigned int *pniov, size_t size);
480     void        (*init_out_iov_from_pdu)(V9fsPDU *pdu, struct iovec **piov,
481                                          unsigned int *pniov, size_t size);
482     void        (*push_and_notify)(V9fsPDU *pdu);
483 };
484 
485 #endif
486