Blame view
fs/dlm/user.c
24.5 KB
2522fe45a
|
1 |
// SPDX-License-Identifier: GPL-2.0-only |
597d0cae0
|
2 |
/* |
7fe2b3190
|
3 |
* Copyright (C) 2006-2010 Red Hat, Inc. All rights reserved. |
597d0cae0
|
4 5 6 7 8 |
*/ #include <linux/miscdevice.h> #include <linux/init.h> #include <linux/wait.h> |
597d0cae0
|
9 10 11 12 13 14 15 |
#include <linux/file.h> #include <linux/fs.h> #include <linux/poll.h> #include <linux/signal.h> #include <linux/spinlock.h> #include <linux/dlm.h> #include <linux/dlm_device.h> |
5a0e3ad6a
|
16 |
#include <linux/slab.h> |
174cd4b1e
|
17 |
#include <linux/sched/signal.h> |
597d0cae0
|
18 19 20 21 22 |
#include "dlm_internal.h" #include "lockspace.h" #include "lock.h" #include "lvb_table.h" |
84c6e8cd3
|
23 |
#include "user.h" |
8304d6f24
|
24 |
#include "ast.h" |
3595c5593
|
25 |
#include "config.h" |
597d0cae0
|
26 |
|
0fe410d3f
|
27 |
static const char name_prefix[] = "dlm"; |
00977a59b
|
28 |
static const struct file_operations device_fops; |
dc68c7ed3
|
29 30 |
static atomic_t dlm_monitor_opened; static int dlm_monitor_unused = 1; |
597d0cae0
|
31 32 33 34 35 36 |
#ifdef CONFIG_COMPAT struct dlm_lock_params32 { __u8 mode; __u8 namelen; |
d7db923ea
|
37 38 |
__u16 unused; __u32 flags; |
597d0cae0
|
39 40 |
__u32 lkid; __u32 parent; |
d7db923ea
|
41 42 |
__u64 xid; __u64 timeout; |
597d0cae0
|
43 44 45 46 47 |
__u32 castparam; __u32 castaddr; __u32 bastparam; __u32 bastaddr; __u32 lksb; |
597d0cae0
|
48 49 50 51 52 53 54 55 56 57 58 59 60 |
char lvb[DLM_USER_LVB_LEN]; char name[0]; }; struct dlm_write_request32 { __u32 version[3]; __u8 cmd; __u8 is64bit; __u8 unused[2]; union { struct dlm_lock_params32 lock; struct dlm_lspace_params lspace; |
72c2be776
|
61 |
struct dlm_purge_params purge; |
597d0cae0
|
62 63 64 65 66 67 68 69 70 71 72 |
} i; }; struct dlm_lksb32 { __u32 sb_status; __u32 sb_lkid; __u8 sb_flags; __u32 sb_lvbptr; }; struct dlm_lock_result32 { |
d7db923ea
|
73 |
__u32 version[3]; |
597d0cae0
|
74 75 76 77 78 79 80 81 82 83 84 85 |
__u32 length; __u32 user_astaddr; __u32 user_astparam; __u32 user_lksb; struct dlm_lksb32 lksb; __u8 bast_mode; __u8 unused[3]; /* Offsets may be zero if no data is present */ __u32 lvb_offset; }; static void compat_input(struct dlm_write_request *kb, |
2a79289e8
|
86 |
struct dlm_write_request32 *kb32, |
1fecb1c4b
|
87 |
int namelen) |
597d0cae0
|
88 89 90 91 92 93 94 95 96 97 98 |
{ kb->version[0] = kb32->version[0]; kb->version[1] = kb32->version[1]; kb->version[2] = kb32->version[2]; kb->cmd = kb32->cmd; kb->is64bit = kb32->is64bit; if (kb->cmd == DLM_USER_CREATE_LOCKSPACE || kb->cmd == DLM_USER_REMOVE_LOCKSPACE) { kb->i.lspace.flags = kb32->i.lspace.flags; kb->i.lspace.minor = kb32->i.lspace.minor; |
1fecb1c4b
|
99 |
memcpy(kb->i.lspace.name, kb32->i.lspace.name, namelen); |
72c2be776
|
100 101 102 |
} else if (kb->cmd == DLM_USER_PURGE) { kb->i.purge.nodeid = kb32->i.purge.nodeid; kb->i.purge.pid = kb32->i.purge.pid; |
597d0cae0
|
103 104 105 106 107 108 |
} else { kb->i.lock.mode = kb32->i.lock.mode; kb->i.lock.namelen = kb32->i.lock.namelen; kb->i.lock.flags = kb32->i.lock.flags; kb->i.lock.lkid = kb32->i.lock.lkid; kb->i.lock.parent = kb32->i.lock.parent; |
d7db923ea
|
109 110 |
kb->i.lock.xid = kb32->i.lock.xid; kb->i.lock.timeout = kb32->i.lock.timeout; |
597d0cae0
|
111 112 113 114 115 116 |
kb->i.lock.castparam = (void *)(long)kb32->i.lock.castparam; kb->i.lock.castaddr = (void *)(long)kb32->i.lock.castaddr; kb->i.lock.bastparam = (void *)(long)kb32->i.lock.bastparam; kb->i.lock.bastaddr = (void *)(long)kb32->i.lock.bastaddr; kb->i.lock.lksb = (void *)(long)kb32->i.lock.lksb; memcpy(kb->i.lock.lvb, kb32->i.lock.lvb, DLM_USER_LVB_LEN); |
1fecb1c4b
|
117 |
memcpy(kb->i.lock.name, kb32->i.lock.name, namelen); |
597d0cae0
|
118 119 120 121 122 123 |
} } static void compat_output(struct dlm_lock_result *res, struct dlm_lock_result32 *res32) { |
8286d6b14
|
124 |
memset(res32, 0, sizeof(*res32)); |
d7db923ea
|
125 126 127 |
res32->version[0] = res->version[0]; res32->version[1] = res->version[1]; res32->version[2] = res->version[2]; |
597d0cae0
|
128 129 130 131 132 133 134 135 136 137 138 139 140 141 |
res32->user_astaddr = (__u32)(long)res->user_astaddr; res32->user_astparam = (__u32)(long)res->user_astparam; res32->user_lksb = (__u32)(long)res->user_lksb; res32->bast_mode = res->bast_mode; res32->lvb_offset = res->lvb_offset; res32->length = res->length; res32->lksb.sb_status = res->lksb.sb_status; res32->lksb.sb_flags = res->lksb.sb_flags; res32->lksb.sb_lkid = res->lksb.sb_lkid; res32->lksb.sb_lvbptr = (__u32)(long)res->lksb.sb_lvbptr; } #endif |
84d8cd69a
|
142 143 144 145 146 147 148 149 150 151 |
/* Figure out if this lock is at the end of its life and no longer available for the application to use. The lkb still exists until the final ast is read. A lock becomes EOL in three situations: 1. a noqueue request fails with EAGAIN 2. an unlock completes with EUNLOCK 3. a cancel of a waiting request completes with ECANCEL/EDEADLK An EOL lock needs to be removed from the process's list of locks. And we can't allow any new operation on an EOL lock. This is not related to the lifetime of the lkb struct which is managed entirely by refcount. */ |
8304d6f24
|
152 |
static int lkb_is_endoflife(int mode, int status) |
84d8cd69a
|
153 |
{ |
8304d6f24
|
154 |
switch (status) { |
84d8cd69a
|
155 156 157 158 |
case -DLM_EUNLOCK: return 1; case -DLM_ECANCEL: case -ETIMEDOUT: |
8b4021fa4
|
159 |
case -EDEADLK: |
84d8cd69a
|
160 |
case -EAGAIN: |
8304d6f24
|
161 |
if (mode == DLM_LOCK_IV) |
84d8cd69a
|
162 163 164 165 166 |
return 1; break; } return 0; } |
ef0c2bb05
|
167 168 |
/* we could possibly check if the cancel of an orphan has resulted in the lkb being removed and then remove that lkb from the orphans list and free it */ |
597d0cae0
|
169 |
|
8304d6f24
|
170 171 |
void dlm_user_add_ast(struct dlm_lkb *lkb, uint32_t flags, int mode, int status, uint32_t sbflags, uint64_t seq) |
597d0cae0
|
172 173 174 175 |
{ struct dlm_ls *ls; struct dlm_user_args *ua; struct dlm_user_proc *proc; |
8304d6f24
|
176 |
int rv; |
597d0cae0
|
177 |
|
ef0c2bb05
|
178 |
if (lkb->lkb_flags & (DLM_IFL_ORPHAN | DLM_IFL_DEAD)) |
597d0cae0
|
179 |
return; |
597d0cae0
|
180 181 182 183 184 185 |
ls = lkb->lkb_resource->res_ls; mutex_lock(&ls->ls_clear_proc_locks); /* If ORPHAN/DEAD flag is set, it means the process is dead so an ast can't be delivered. For ORPHAN's, dlm_clear_proc_locks() freed |
ef0c2bb05
|
186 187 188 |
lkb->ua so we can't try to use it. This second check is necessary for cases where a completion ast is received for an operation that began before clear_proc_locks did its cancel/unlock. */ |
597d0cae0
|
189 |
|
ef0c2bb05
|
190 |
if (lkb->lkb_flags & (DLM_IFL_ORPHAN | DLM_IFL_DEAD)) |
597d0cae0
|
191 |
goto out; |
597d0cae0
|
192 |
|
d292c0cc4
|
193 194 |
DLM_ASSERT(lkb->lkb_ua, dlm_print_lkb(lkb);); ua = lkb->lkb_ua; |
597d0cae0
|
195 |
proc = ua->proc; |
8304d6f24
|
196 |
if ((flags & DLM_CB_BAST) && ua->bastaddr == NULL) |
597d0cae0
|
197 |
goto out; |
8304d6f24
|
198 199 |
if ((flags & DLM_CB_CAST) && lkb_is_endoflife(mode, status)) lkb->lkb_flags |= DLM_IFL_ENDOFLIFE; |
597d0cae0
|
200 |
spin_lock(&proc->asts_spin); |
ef0c2bb05
|
201 |
|
8304d6f24
|
202 203 204 205 206 |
rv = dlm_add_lkb_callback(lkb, flags, mode, status, sbflags, seq); if (rv < 0) { spin_unlock(&proc->asts_spin); goto out; } |
ef0c2bb05
|
207 |
|
23e8e1aaa
|
208 |
if (list_empty(&lkb->lkb_cb_list)) { |
597d0cae0
|
209 |
kref_get(&lkb->lkb_ref); |
23e8e1aaa
|
210 |
list_add_tail(&lkb->lkb_cb_list, &proc->asts); |
597d0cae0
|
211 212 |
wake_up_interruptible(&proc->wait); } |
597d0cae0
|
213 |
spin_unlock(&proc->asts_spin); |
34e22bed1
|
214 |
|
8304d6f24
|
215 216 |
if (lkb->lkb_flags & DLM_IFL_ENDOFLIFE) { /* N.B. spin_lock locks_spin, not asts_spin */ |
ce5246b97
|
217 |
spin_lock(&proc->locks_spin); |
ef0c2bb05
|
218 219 220 221 |
if (!list_empty(&lkb->lkb_ownqueue)) { list_del_init(&lkb->lkb_ownqueue); dlm_put_lkb(lkb); } |
ce5246b97
|
222 |
spin_unlock(&proc->locks_spin); |
34e22bed1
|
223 |
} |
597d0cae0
|
224 225 226 227 228 229 230 231 232 |
out: mutex_unlock(&ls->ls_clear_proc_locks); } static int device_user_lock(struct dlm_user_proc *proc, struct dlm_lock_params *params) { struct dlm_ls *ls; struct dlm_user_args *ua; |
2ab4bd8ea
|
233 |
uint32_t lkid; |
597d0cae0
|
234 235 236 237 238 239 240 241 242 243 |
int error = -ENOMEM; ls = dlm_find_lockspace_local(proc->lockspace); if (!ls) return -ENOENT; if (!params->castaddr || !params->lksb) { error = -EINVAL; goto out; } |
573c24c4a
|
244 |
ua = kzalloc(sizeof(struct dlm_user_args), GFP_NOFS); |
597d0cae0
|
245 246 247 248 249 250 251 252 |
if (!ua) goto out; ua->proc = proc; ua->user_lksb = params->lksb; ua->castparam = params->castparam; ua->castaddr = params->castaddr; ua->bastparam = params->bastparam; ua->bastaddr = params->bastaddr; |
d7db923ea
|
253 |
ua->xid = params->xid; |
597d0cae0
|
254 |
|
2ab4bd8ea
|
255 |
if (params->flags & DLM_LKF_CONVERT) { |
597d0cae0
|
256 257 |
error = dlm_user_convert(ls, ua, params->mode, params->flags, |
d7db923ea
|
258 259 |
params->lkid, params->lvb, (unsigned long) params->timeout); |
2ab4bd8ea
|
260 261 262 263 264 265 266 267 268 |
} else if (params->flags & DLM_LKF_ORPHAN) { error = dlm_user_adopt_orphan(ls, ua, params->mode, params->flags, params->name, params->namelen, (unsigned long) params->timeout, &lkid); if (!error) error = lkid; } else { |
597d0cae0
|
269 270 271 |
error = dlm_user_request(ls, ua, params->mode, params->flags, params->name, params->namelen, |
d7db923ea
|
272 |
(unsigned long) params->timeout); |
597d0cae0
|
273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 |
if (!error) error = ua->lksb.sb_lkid; } out: dlm_put_lockspace(ls); return error; } static int device_user_unlock(struct dlm_user_proc *proc, struct dlm_lock_params *params) { struct dlm_ls *ls; struct dlm_user_args *ua; int error = -ENOMEM; ls = dlm_find_lockspace_local(proc->lockspace); if (!ls) return -ENOENT; |
573c24c4a
|
291 |
ua = kzalloc(sizeof(struct dlm_user_args), GFP_NOFS); |
597d0cae0
|
292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 |
if (!ua) goto out; ua->proc = proc; ua->user_lksb = params->lksb; ua->castparam = params->castparam; ua->castaddr = params->castaddr; if (params->flags & DLM_LKF_CANCEL) error = dlm_user_cancel(ls, ua, params->flags, params->lkid); else error = dlm_user_unlock(ls, ua, params->flags, params->lkid, params->lvb); out: dlm_put_lockspace(ls); return error; } |
8b4021fa4
|
308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 |
static int device_user_deadlock(struct dlm_user_proc *proc, struct dlm_lock_params *params) { struct dlm_ls *ls; int error; ls = dlm_find_lockspace_local(proc->lockspace); if (!ls) return -ENOENT; error = dlm_user_deadlock(ls, params->flags, params->lkid); dlm_put_lockspace(ls); return error; } |
0f8e0d9a3
|
323 |
static int dlm_device_register(struct dlm_ls *ls, char *name) |
254da030d
|
324 325 |
{ int error, len; |
0f8e0d9a3
|
326 327 328 329 |
/* The device is already registered. This happens when the lockspace is created multiple times from userspace. */ if (ls->ls_device.name) return 0; |
254da030d
|
330 331 |
error = -ENOMEM; len = strlen(name) + strlen(name_prefix) + 2; |
573c24c4a
|
332 |
ls->ls_device.name = kzalloc(len, GFP_NOFS); |
254da030d
|
333 334 335 336 337 338 339 340 341 342 343 |
if (!ls->ls_device.name) goto fail; snprintf((char *)ls->ls_device.name, len, "%s_%s", name_prefix, name); ls->ls_device.fops = &device_fops; ls->ls_device.minor = MISC_DYNAMIC_MINOR; error = misc_register(&ls->ls_device); if (error) { kfree(ls->ls_device.name); |
55acdd926
|
344 345 346 347 |
/* this has to be set to NULL * to avoid a double-free in dlm_device_deregister */ ls->ls_device.name = NULL; |
254da030d
|
348 349 350 351 |
} fail: return error; } |
0f8e0d9a3
|
352 353 |
int dlm_device_deregister(struct dlm_ls *ls) { |
0f8e0d9a3
|
354 355 356 357 358 |
/* The device is not registered. This happens when the lockspace was never used from userspace, or when device_create_lockspace() calls dlm_release_lockspace() after the register fails. */ if (!ls->ls_device.name) return 0; |
f368ed608
|
359 360 361 |
misc_deregister(&ls->ls_device); kfree(ls->ls_device.name); return 0; |
0f8e0d9a3
|
362 |
} |
72c2be776
|
363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 |
static int device_user_purge(struct dlm_user_proc *proc, struct dlm_purge_params *params) { struct dlm_ls *ls; int error; ls = dlm_find_lockspace_local(proc->lockspace); if (!ls) return -ENOENT; error = dlm_user_purge(ls, proc, params->nodeid, params->pid); dlm_put_lockspace(ls); return error; } |
597d0cae0
|
378 379 380 381 |
static int device_create_lockspace(struct dlm_lspace_params *params) { dlm_lockspace_t *lockspace; struct dlm_ls *ls; |
254da030d
|
382 |
int error; |
597d0cae0
|
383 384 385 |
if (!capable(CAP_SYS_ADMIN)) return -EPERM; |
3595c5593
|
386 |
error = dlm_new_lockspace(params->name, dlm_config.ci_cluster_name, params->flags, |
60f98d183
|
387 388 |
DLM_USER_LVB_LEN, NULL, NULL, NULL, &lockspace); |
597d0cae0
|
389 390 391 392 393 394 |
if (error) return error; ls = dlm_find_lockspace_local(lockspace); if (!ls) return -ENOENT; |
0f8e0d9a3
|
395 |
error = dlm_device_register(ls, params->name); |
597d0cae0
|
396 |
dlm_put_lockspace(ls); |
597d0cae0
|
397 |
|
254da030d
|
398 399 400 401 |
if (error) dlm_release_lockspace(lockspace, 0); else error = ls->ls_device.minor; |
597d0cae0
|
402 403 404 405 406 407 408 |
return error; } static int device_remove_lockspace(struct dlm_lspace_params *params) { dlm_lockspace_t *lockspace; struct dlm_ls *ls; |
c6e6f0ba8
|
409 |
int error, force = 0; |
597d0cae0
|
410 411 412 413 414 415 416 |
if (!capable(CAP_SYS_ADMIN)) return -EPERM; ls = dlm_find_lockspace_device(params->minor); if (!ls) return -ENOENT; |
c6e6f0ba8
|
417 418 |
if (params->flags & DLM_USER_LSFLG_FORCEFREE) force = 2; |
597d0cae0
|
419 |
lockspace = ls->ls_local_handle; |
0f8e0d9a3
|
420 |
dlm_put_lockspace(ls); |
597d0cae0
|
421 |
|
0f8e0d9a3
|
422 423 424 425 426 427 |
/* The final dlm_release_lockspace waits for references to go to zero, so all processes will need to close their device for the ls before the release will proceed. release also calls the device_deregister above. Converting a positive return value from release to zero means that userspace won't know when its release was the final one, but it shouldn't need to know. */ |
597d0cae0
|
428 |
|
c6e6f0ba8
|
429 |
error = dlm_release_lockspace(lockspace, force); |
0f8e0d9a3
|
430 431 |
if (error > 0) error = 0; |
597d0cae0
|
432 433 434 435 436 437 438 439 440 441 442 443 444 445 |
return error; } /* Check the user's version matches ours */ static int check_version(struct dlm_write_request *req) { if (req->version[0] != DLM_DEVICE_VERSION_MAJOR || (req->version[0] == DLM_DEVICE_VERSION_MAJOR && req->version[1] > DLM_DEVICE_VERSION_MINOR)) { printk(KERN_DEBUG "dlm: process %s (%d) version mismatch " "user (%d.%d.%d) kernel (%d.%d.%d) ", current->comm, |
ba25f9dcc
|
446 |
task_pid_nr(current), |
597d0cae0
|
447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 |
req->version[0], req->version[1], req->version[2], DLM_DEVICE_VERSION_MAJOR, DLM_DEVICE_VERSION_MINOR, DLM_DEVICE_VERSION_PATCH); return -EINVAL; } return 0; } /* * device_write * * device_user_lock * dlm_user_request -> request_lock * dlm_user_convert -> convert_lock * * device_user_unlock * dlm_user_unlock -> unlock_lock * dlm_user_cancel -> cancel_lock * * device_create_lockspace * dlm_new_lockspace * * device_remove_lockspace * dlm_release_lockspace */ /* a write to a lockspace device is a lock or unlock request, a write to the control device is to create/remove a lockspace */ static ssize_t device_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos) { struct dlm_user_proc *proc = file->private_data; struct dlm_write_request *kbuf; |
597d0cae0
|
484 485 486 487 488 489 490 491 |
int error; #ifdef CONFIG_COMPAT if (count < sizeof(struct dlm_write_request32)) #else if (count < sizeof(struct dlm_write_request)) #endif return -EINVAL; |
d4b0bcf32
|
492 493 494 495 |
/* * can't compare against COMPAT/dlm_write_request32 because * we don't yet know if is64bit is zero */ |
2b75bc912
|
496 |
if (count > sizeof(struct dlm_write_request) + DLM_RESNAME_MAXLEN) |
2b75bc912
|
497 |
return -EINVAL; |
16e5c1fc3
|
498 |
kbuf = memdup_user_nul(buf, count); |
117aa41e8
|
499 |
if (IS_ERR(kbuf)) |
16e5c1fc3
|
500 |
return PTR_ERR(kbuf); |
597d0cae0
|
501 502 503 504 505 506 507 508 509 |
if (check_version(kbuf)) { error = -EBADE; goto out_free; } #ifdef CONFIG_COMPAT if (!kbuf->is64bit) { struct dlm_write_request32 *k32buf; |
1fecb1c4b
|
510 511 512 513 |
int namelen = 0; if (count > sizeof(struct dlm_write_request32)) namelen = count - sizeof(struct dlm_write_request32); |
597d0cae0
|
514 |
k32buf = (struct dlm_write_request32 *)kbuf; |
1fecb1c4b
|
515 516 517 |
/* add 1 after namelen so that the name string is terminated */ kbuf = kzalloc(sizeof(struct dlm_write_request) + namelen + 1, |
573c24c4a
|
518 |
GFP_NOFS); |
cb980d9a3
|
519 520 |
if (!kbuf) { kfree(k32buf); |
597d0cae0
|
521 |
return -ENOMEM; |
cb980d9a3
|
522 |
} |
597d0cae0
|
523 524 525 |
if (proc) set_bit(DLM_PROC_FLAGS_COMPAT, &proc->flags); |
1fecb1c4b
|
526 527 |
compat_input(kbuf, k32buf, namelen); |
597d0cae0
|
528 529 530 531 532 533 |
kfree(k32buf); } #endif /* do we really need this? can a write happen after a close? */ if ((kbuf->cmd == DLM_USER_LOCK || kbuf->cmd == DLM_USER_UNLOCK) && |
cb980d9a3
|
534 535 536 537 |
(proc && test_bit(DLM_PROC_FLAGS_CLOSING, &proc->flags))) { error = -EINVAL; goto out_free; } |
597d0cae0
|
538 |
|
597d0cae0
|
539 540 541 542 543 544 545 |
error = -EINVAL; switch (kbuf->cmd) { case DLM_USER_LOCK: if (!proc) { log_print("no locking on control device"); |
c6ca7bc91
|
546 |
goto out_free; |
597d0cae0
|
547 548 549 550 551 552 553 |
} error = device_user_lock(proc, &kbuf->i.lock); break; case DLM_USER_UNLOCK: if (!proc) { log_print("no locking on control device"); |
c6ca7bc91
|
554 |
goto out_free; |
597d0cae0
|
555 556 557 |
} error = device_user_unlock(proc, &kbuf->i.lock); break; |
8b4021fa4
|
558 559 560 |
case DLM_USER_DEADLOCK: if (!proc) { log_print("no locking on control device"); |
c6ca7bc91
|
561 |
goto out_free; |
8b4021fa4
|
562 563 564 |
} error = device_user_deadlock(proc, &kbuf->i.lock); break; |
597d0cae0
|
565 566 567 |
case DLM_USER_CREATE_LOCKSPACE: if (proc) { log_print("create/remove only on control device"); |
c6ca7bc91
|
568 |
goto out_free; |
597d0cae0
|
569 570 571 572 573 574 575 |
} error = device_create_lockspace(&kbuf->i.lspace); break; case DLM_USER_REMOVE_LOCKSPACE: if (proc) { log_print("create/remove only on control device"); |
c6ca7bc91
|
576 |
goto out_free; |
597d0cae0
|
577 578 579 |
} error = device_remove_lockspace(&kbuf->i.lspace); break; |
72c2be776
|
580 581 582 |
case DLM_USER_PURGE: if (!proc) { log_print("no locking on control device"); |
c6ca7bc91
|
583 |
goto out_free; |
72c2be776
|
584 585 586 |
} error = device_user_purge(proc, &kbuf->i.purge); break; |
597d0cae0
|
587 588 589 590 591 |
default: log_print("Unknown command passed to DLM device : %d ", kbuf->cmd); } |
597d0cae0
|
592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 |
out_free: kfree(kbuf); return error; } /* Every process that opens the lockspace device has its own "proc" structure hanging off the open file that's used to keep track of locks owned by the process and asts that need to be delivered to the process. */ static int device_open(struct inode *inode, struct file *file) { struct dlm_user_proc *proc; struct dlm_ls *ls; ls = dlm_find_lockspace_device(iminor(inode)); |
f9f2ed486
|
607 |
if (!ls) |
597d0cae0
|
608 |
return -ENOENT; |
573c24c4a
|
609 |
proc = kzalloc(sizeof(struct dlm_user_proc), GFP_NOFS); |
597d0cae0
|
610 611 612 613 614 615 616 617 |
if (!proc) { dlm_put_lockspace(ls); return -ENOMEM; } proc->lockspace = ls->ls_local_handle; INIT_LIST_HEAD(&proc->asts); INIT_LIST_HEAD(&proc->locks); |
a1bc86e6b
|
618 |
INIT_LIST_HEAD(&proc->unlocking); |
597d0cae0
|
619 620 621 622 623 624 625 626 627 628 629 630 |
spin_lock_init(&proc->asts_spin); spin_lock_init(&proc->locks_spin); init_waitqueue_head(&proc->wait); file->private_data = proc; return 0; } static int device_close(struct inode *inode, struct file *file) { struct dlm_user_proc *proc = file->private_data; struct dlm_ls *ls; |
597d0cae0
|
631 632 633 634 |
ls = dlm_find_lockspace_local(proc->lockspace); if (!ls) return -ENOENT; |
597d0cae0
|
635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 |
set_bit(DLM_PROC_FLAGS_CLOSING, &proc->flags); dlm_clear_proc_locks(ls, proc); /* at this point no more lkb's should exist for this lockspace, so there's no chance of dlm_user_add_ast() being called and looking for lkb->ua->proc */ kfree(proc); file->private_data = NULL; dlm_put_lockspace(ls); dlm_put_lockspace(ls); /* for the find in device_open() */ /* FIXME: AUTOFREE: if this ls is no longer used do device_remove_lockspace() */ |
597d0cae0
|
651 652 |
return 0; } |
8304d6f24
|
653 654 655 |
static int copy_result_to_user(struct dlm_user_args *ua, int compat, uint32_t flags, int mode, int copy_lvb, char __user *buf, size_t count) |
597d0cae0
|
656 657 658 659 660 661 662 663 664 665 666 |
{ #ifdef CONFIG_COMPAT struct dlm_lock_result32 result32; #endif struct dlm_lock_result result; void *resultptr; int error=0; int len; int struct_len; memset(&result, 0, sizeof(struct dlm_lock_result)); |
d7db923ea
|
667 668 669 |
result.version[0] = DLM_DEVICE_VERSION_MAJOR; result.version[1] = DLM_DEVICE_VERSION_MINOR; result.version[2] = DLM_DEVICE_VERSION_PATCH; |
9de30f3f7
|
670 |
memcpy(&result.lksb, &ua->lksb, offsetof(struct dlm_lksb, sb_lvbptr)); |
597d0cae0
|
671 672 673 674 675 676 677 |
result.user_lksb = ua->user_lksb; /* FIXME: dlm1 provides for the user's bastparam/addr to not be updated in a conversion unless the conversion is successful. See code in dlm_user_convert() for updating ua from ua_tmp. OpenVMS, though, notes that a new blocking AST address and parameter are set even if the conversion fails, so maybe we should just do that. */ |
8304d6f24
|
678 |
if (flags & DLM_CB_BAST) { |
597d0cae0
|
679 680 |
result.user_astaddr = ua->bastaddr; result.user_astparam = ua->bastparam; |
89d799d00
|
681 |
result.bast_mode = mode; |
597d0cae0
|
682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 |
} else { result.user_astaddr = ua->castaddr; result.user_astparam = ua->castparam; } #ifdef CONFIG_COMPAT if (compat) len = sizeof(struct dlm_lock_result32); else #endif len = sizeof(struct dlm_lock_result); struct_len = len; /* copy lvb to userspace if there is one, it's been updated, and the user buffer has space for it */ |
8304d6f24
|
697 |
if (copy_lvb && ua->lksb.sb_lvbptr && count >= len + DLM_USER_LVB_LEN) { |
597d0cae0
|
698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 |
if (copy_to_user(buf+len, ua->lksb.sb_lvbptr, DLM_USER_LVB_LEN)) { error = -EFAULT; goto out; } result.lvb_offset = len; len += DLM_USER_LVB_LEN; } result.length = len; resultptr = &result; #ifdef CONFIG_COMPAT if (compat) { compat_output(&result, &result32); resultptr = &result32; } #endif if (copy_to_user(buf, resultptr, struct_len)) error = -EFAULT; else error = len; out: return error; } |
d7db923ea
|
724 725 726 727 728 729 730 731 732 733 734 735 736 |
static int copy_version_to_user(char __user *buf, size_t count) { struct dlm_device_version ver; memset(&ver, 0, sizeof(struct dlm_device_version)); ver.version[0] = DLM_DEVICE_VERSION_MAJOR; ver.version[1] = DLM_DEVICE_VERSION_MINOR; ver.version[2] = DLM_DEVICE_VERSION_PATCH; if (copy_to_user(buf, &ver, sizeof(struct dlm_device_version))) return -EFAULT; return sizeof(struct dlm_device_version); } |
597d0cae0
|
737 738 739 740 741 742 743 |
/* a read returns a single ast described in a struct dlm_lock_result */ static ssize_t device_read(struct file *file, char __user *buf, size_t count, loff_t *ppos) { struct dlm_user_proc *proc = file->private_data; struct dlm_lkb *lkb; |
597d0cae0
|
744 |
DECLARE_WAITQUEUE(wait, current); |
8304d6f24
|
745 746 |
struct dlm_callback cb; int rv, resid, copy_lvb = 0; |
b96f46503
|
747 |
int old_mode, new_mode; |
597d0cae0
|
748 |
|
d7db923ea
|
749 |
if (count == sizeof(struct dlm_device_version)) { |
8304d6f24
|
750 751 |
rv = copy_version_to_user(buf, count); return rv; |
d7db923ea
|
752 753 754 755 756 757 |
} if (!proc) { log_print("non-version read from control device %zu", count); return -EINVAL; } |
597d0cae0
|
758 759 760 761 762 763 |
#ifdef CONFIG_COMPAT if (count < sizeof(struct dlm_lock_result32)) #else if (count < sizeof(struct dlm_lock_result)) #endif return -EINVAL; |
89d799d00
|
764 |
try_another: |
597d0cae0
|
765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 |
/* do we really need this? can a read happen after a close? */ if (test_bit(DLM_PROC_FLAGS_CLOSING, &proc->flags)) return -EINVAL; spin_lock(&proc->asts_spin); if (list_empty(&proc->asts)) { if (file->f_flags & O_NONBLOCK) { spin_unlock(&proc->asts_spin); return -EAGAIN; } add_wait_queue(&proc->wait, &wait); repeat: set_current_state(TASK_INTERRUPTIBLE); if (list_empty(&proc->asts) && !signal_pending(current)) { spin_unlock(&proc->asts_spin); schedule(); spin_lock(&proc->asts_spin); goto repeat; } set_current_state(TASK_RUNNING); remove_wait_queue(&proc->wait, &wait); if (signal_pending(current)) { spin_unlock(&proc->asts_spin); return -ERESTARTSYS; } } |
8304d6f24
|
794 |
/* if we empty lkb_callbacks, we don't want to unlock the spinlock |
23e8e1aaa
|
795 |
without removing lkb_cb_list; so empty lkb_cb_list is always |
8304d6f24
|
796 |
consistent with empty lkb_callbacks */ |
597d0cae0
|
797 |
|
23e8e1aaa
|
798 |
lkb = list_entry(proc->asts.next, struct dlm_lkb, lkb_cb_list); |
597d0cae0
|
799 |
|
b96f46503
|
800 801 |
/* rem_lkb_callback sets a new lkb_last_cast */ old_mode = lkb->lkb_last_cast.mode; |
8304d6f24
|
802 803 804 805 806 |
rv = dlm_rem_lkb_callback(lkb->lkb_resource->res_ls, lkb, &cb, &resid); if (rv < 0) { /* this shouldn't happen; lkb should have been removed from list when resid was zero */ log_print("dlm_rem_lkb_callback empty %x", lkb->lkb_id); |
23e8e1aaa
|
807 |
list_del_init(&lkb->lkb_cb_list); |
8304d6f24
|
808 809 810 811 |
spin_unlock(&proc->asts_spin); /* removes ref for proc->asts, may cause lkb to be freed */ dlm_put_lkb(lkb); goto try_another; |
89d799d00
|
812 |
} |
8304d6f24
|
813 |
if (!resid) |
23e8e1aaa
|
814 |
list_del_init(&lkb->lkb_cb_list); |
8304d6f24
|
815 |
spin_unlock(&proc->asts_spin); |
89d799d00
|
816 |
|
8304d6f24
|
817 818 819 820 821 |
if (cb.flags & DLM_CB_SKIP) { /* removes ref for proc->asts, may cause lkb to be freed */ if (!resid) dlm_put_lkb(lkb); goto try_another; |
89d799d00
|
822 |
} |
8304d6f24
|
823 |
if (cb.flags & DLM_CB_CAST) { |
8304d6f24
|
824 |
new_mode = cb.mode; |
597d0cae0
|
825 |
|
8304d6f24
|
826 827 828 |
if (!cb.sb_status && lkb->lkb_lksb->sb_lvbptr && dlm_lvb_operations[old_mode + 1][new_mode + 1]) copy_lvb = 1; |
89d799d00
|
829 |
|
8304d6f24
|
830 831 |
lkb->lkb_lksb->sb_status = cb.sb_status; lkb->lkb_lksb->sb_flags = cb.sb_flags; |
89d799d00
|
832 |
} |
597d0cae0
|
833 |
|
8304d6f24
|
834 835 836 |
rv = copy_result_to_user(lkb->lkb_ua, test_bit(DLM_PROC_FLAGS_COMPAT, &proc->flags), cb.flags, cb.mode, copy_lvb, buf, count); |
89d799d00
|
837 |
|
8304d6f24
|
838 839 |
/* removes ref for proc->asts, may cause lkb to be freed */ if (!resid) |
597d0cae0
|
840 |
dlm_put_lkb(lkb); |
8304d6f24
|
841 |
return rv; |
597d0cae0
|
842 |
} |
076ccb76e
|
843 |
static __poll_t device_poll(struct file *file, poll_table *wait) |
597d0cae0
|
844 845 846 847 848 849 850 851 |
{ struct dlm_user_proc *proc = file->private_data; poll_wait(file, &proc->wait, wait); spin_lock(&proc->asts_spin); if (!list_empty(&proc->asts)) { spin_unlock(&proc->asts_spin); |
a9a08845e
|
852 |
return EPOLLIN | EPOLLRDNORM; |
597d0cae0
|
853 854 855 856 |
} spin_unlock(&proc->asts_spin); return 0; } |
dc68c7ed3
|
857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 |
int dlm_user_daemon_available(void) { /* dlm_controld hasn't started (or, has started, but not properly populated configfs) */ if (!dlm_our_nodeid()) return 0; /* This is to deal with versions of dlm_controld that don't know about the monitor device. We assume that if the dlm_controld was started (above), but the monitor device was never opened, that it's an old version. dlm_controld should open the monitor device before populating configfs. */ if (dlm_monitor_unused) return 1; return atomic_read(&dlm_monitor_opened) ? 1 : 0; } |
597d0cae0
|
876 877 878 879 880 881 882 883 884 885 |
static int ctl_device_open(struct inode *inode, struct file *file) { file->private_data = NULL; return 0; } static int ctl_device_close(struct inode *inode, struct file *file) { return 0; } |
dc68c7ed3
|
886 887 888 889 890 891 892 893 894 895 896 897 898 |
static int monitor_device_open(struct inode *inode, struct file *file) { atomic_inc(&dlm_monitor_opened); dlm_monitor_unused = 0; return 0; } static int monitor_device_close(struct inode *inode, struct file *file) { if (atomic_dec_and_test(&dlm_monitor_opened)) dlm_stop_lockspaces(); return 0; } |
00977a59b
|
899 |
static const struct file_operations device_fops = { |
597d0cae0
|
900 901 902 903 904 905 |
.open = device_open, .release = device_close, .read = device_read, .write = device_write, .poll = device_poll, .owner = THIS_MODULE, |
6038f373a
|
906 |
.llseek = noop_llseek, |
597d0cae0
|
907 |
}; |
00977a59b
|
908 |
static const struct file_operations ctl_device_fops = { |
597d0cae0
|
909 910 |
.open = ctl_device_open, .release = ctl_device_close, |
d7db923ea
|
911 |
.read = device_read, |
597d0cae0
|
912 913 |
.write = device_write, .owner = THIS_MODULE, |
6038f373a
|
914 |
.llseek = noop_llseek, |
597d0cae0
|
915 |
}; |
0fe410d3f
|
916 917 918 919 920 |
static struct miscdevice ctl_device = { .name = "dlm-control", .fops = &ctl_device_fops, .minor = MISC_DYNAMIC_MINOR, }; |
dc68c7ed3
|
921 922 923 924 |
static const struct file_operations monitor_device_fops = { .open = monitor_device_open, .release = monitor_device_close, .owner = THIS_MODULE, |
6038f373a
|
925 |
.llseek = noop_llseek, |
dc68c7ed3
|
926 927 928 929 930 931 932 |
}; static struct miscdevice monitor_device = { .name = "dlm-monitor", .fops = &monitor_device_fops, .minor = MISC_DYNAMIC_MINOR, }; |
30727174b
|
933 |
int __init dlm_user_init(void) |
597d0cae0
|
934 935 |
{ int error; |
dc68c7ed3
|
936 |
atomic_set(&dlm_monitor_opened, 0); |
597d0cae0
|
937 |
error = misc_register(&ctl_device); |
dc68c7ed3
|
938 |
if (error) { |
597d0cae0
|
939 |
log_print("misc_register failed for control device"); |
dc68c7ed3
|
940 941 |
goto out; } |
597d0cae0
|
942 |
|
dc68c7ed3
|
943 944 945 946 947 948 |
error = misc_register(&monitor_device); if (error) { log_print("misc_register failed for monitor device"); misc_deregister(&ctl_device); } out: |
597d0cae0
|
949 950 951 952 953 954 |
return error; } void dlm_user_exit(void) { misc_deregister(&ctl_device); |
dc68c7ed3
|
955 |
misc_deregister(&monitor_device); |
597d0cae0
|
956 |
} |