2 * Copyright (c) 2002, Jeffrey Roberson <jeff@freebsd.org>
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice unmodified, this list of conditions, and the following
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
18 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 #include <sys/cdefs.h>
28 __FBSDID("$FreeBSD$");
30 #include <sys/param.h>
31 #include <sys/systm.h>
32 #include <sys/kernel.h>
33 #include <sys/kthread.h>
35 #include <sys/mount.h>
36 #include <sys/mutex.h>
37 #include <sys/namei.h>
39 #include <sys/vnode.h>
41 #include <sys/malloc.h>
42 #include <sys/unistd.h>
43 #include <sys/fcntl.h>
44 #include <sys/eventhandler.h>
46 #include <security/mac/mac_framework.h>
48 /* Async. Logging Queue */
50 int aq_entmax; /* Max entries */
51 int aq_entlen; /* Entry length */
52 char *aq_entbuf; /* Buffer for stored entries */
53 int aq_flags; /* Queue flags */
54 struct mtx aq_mtx; /* Queue lock */
55 struct vnode *aq_vp; /* Open vnode handle */
56 struct ucred *aq_cred; /* Credentials of the opening thread */
57 struct ale *aq_first; /* First ent */
58 struct ale *aq_entfree; /* First free ent */
59 struct ale *aq_entvalid; /* First ent valid for writing */
60 LIST_ENTRY(alq) aq_act; /* List of active queues */
61 LIST_ENTRY(alq) aq_link; /* List of all queues */
64 #define AQ_WANTED 0x0001 /* Wakeup sleeper when io is done */
65 #define AQ_ACTIVE 0x0002 /* on the active list */
66 #define AQ_FLUSHING 0x0004 /* doing IO */
67 #define AQ_SHUTDOWN 0x0008 /* Queue no longer valid */
69 #define ALQ_LOCK(alq) mtx_lock_spin(&(alq)->aq_mtx)
70 #define ALQ_UNLOCK(alq) mtx_unlock_spin(&(alq)->aq_mtx)
72 static MALLOC_DEFINE(M_ALD, "ALD", "ALD");
75 * The ald_mtx protects the ald_queues list and the ald_active list.
77 static struct mtx ald_mtx;
78 static LIST_HEAD(, alq) ald_queues;
79 static LIST_HEAD(, alq) ald_active;
80 static int ald_shutingdown = 0;
81 struct thread *ald_thread;
82 static struct proc *ald_proc;
84 #define ALD_LOCK() mtx_lock(&ald_mtx)
85 #define ALD_UNLOCK() mtx_unlock(&ald_mtx)
87 /* Daemon functions */
88 static int ald_add(struct alq *);
89 static int ald_rem(struct alq *);
90 static void ald_startup(void *);
91 static void ald_daemon(void);
92 static void ald_shutdown(void *, int);
93 static void ald_activate(struct alq *);
94 static void ald_deactivate(struct alq *);
96 /* Internal queue functions */
97 static void alq_shutdown(struct alq *);
98 static int alq_doio(struct alq *);
102 * Add a new queue to the global list. Fail if we're shutting down.
105 ald_add(struct alq *alq)
112 if (ald_shutingdown) {
116 LIST_INSERT_HEAD(&ald_queues, alq, aq_link);
123 * Remove a queue from the global list unless we're shutting down. If so,
124 * the ald will take care of cleaning up it's resources.
127 ald_rem(struct alq *alq)
134 if (ald_shutingdown) {
138 LIST_REMOVE(alq, aq_link);
145 * Put a queue on the active list. This will schedule it for writing.
148 ald_activate(struct alq *alq)
150 LIST_INSERT_HEAD(&ald_active, alq, aq_act);
155 ald_deactivate(struct alq *alq)
157 LIST_REMOVE(alq, aq_act);
158 alq->aq_flags &= ~AQ_ACTIVE;
162 ald_startup(void *unused)
164 mtx_init(&ald_mtx, "ALDmtx", NULL, MTX_DEF|MTX_QUIET);
165 LIST_INIT(&ald_queues);
166 LIST_INIT(&ald_active);
175 ald_thread = FIRST_THREAD_IN_PROC(ald_proc);
177 EVENTHANDLER_REGISTER(shutdown_pre_sync, ald_shutdown, NULL,
183 while ((alq = LIST_FIRST(&ald_active)) == NULL)
184 msleep(&ald_active, &ald_mtx, PWAIT, "aldslp", 0);
189 needwakeup = alq_doio(alq);
198 ald_shutdown(void *arg, int howto)
205 while ((alq = LIST_FIRST(&ald_queues)) != NULL) {
206 LIST_REMOVE(alq, aq_link);
215 alq_shutdown(struct alq *alq)
219 /* Stop any new writers. */
220 alq->aq_flags |= AQ_SHUTDOWN;
223 while (alq->aq_flags & (AQ_FLUSHING|AQ_ACTIVE)) {
224 alq->aq_flags |= AQ_WANTED;
225 msleep_spin(alq, &alq->aq_mtx, "aldclose", 0);
229 vn_close(alq->aq_vp, FWRITE, alq->aq_cred,
231 crfree(alq->aq_cred);
235 * Flush all pending data to disk. This operation will block.
238 alq_doio(struct alq *alq)
244 struct iovec aiov[2];
256 alstart = ale = alq->aq_entvalid;
257 alq->aq_entvalid = NULL;
259 bzero(&aiov, sizeof(aiov));
260 bzero(&auio, sizeof(auio));
263 if (aiov[iov].iov_base == NULL)
264 aiov[iov].iov_base = ale->ae_data;
265 aiov[iov].iov_len += alq->aq_entlen;
266 totlen += alq->aq_entlen;
267 /* Check to see if we're wrapping the buffer */
268 if (ale->ae_data + alq->aq_entlen != ale->ae_next->ae_data)
270 ale->ae_flags &= ~AE_VALID;
272 } while (ale->ae_flags & AE_VALID);
274 alq->aq_flags |= AQ_FLUSHING;
277 if (iov == 2 || aiov[iov].iov_base == NULL)
280 auio.uio_iov = &aiov[0];
282 auio.uio_segflg = UIO_SYSSPACE;
283 auio.uio_rw = UIO_WRITE;
284 auio.uio_iovcnt = iov + 1;
285 auio.uio_resid = totlen;
289 * Do all of the junk required to write now.
291 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
292 vn_start_write(vp, &mp, V_WAIT);
293 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
295 * XXX: VOP_WRITE error checks are ignored.
298 if (mac_vnode_check_write(alq->aq_cred, NOCRED, vp) == 0)
300 VOP_WRITE(vp, &auio, IO_UNIT | IO_APPEND, alq->aq_cred);
302 vn_finished_write(mp);
303 VFS_UNLOCK_GIANT(vfslocked);
306 alq->aq_flags &= ~AQ_FLUSHING;
308 if (alq->aq_entfree == NULL)
309 alq->aq_entfree = alstart;
311 if (alq->aq_flags & AQ_WANTED) {
312 alq->aq_flags &= ~AQ_WANTED;
319 static struct kproc_desc ald_kp = {
325 SYSINIT(aldthread, SI_SUB_KTHREAD_IDLE, SI_ORDER_ANY, kproc_start, &ald_kp);
326 SYSINIT(ald, SI_SUB_LOCK, SI_ORDER_ANY, ald_startup, NULL);
329 /* User visible queue functions */
332 * Create the queue data structure, allocate the buffer, and open the file.
335 alq_open(struct alq **alqp, const char *file, struct ucred *cred, int cmode,
351 NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE, file, td);
352 flags = FWRITE | O_NOFOLLOW | O_CREAT;
354 error = vn_open_cred(&nd, &flags, cmode, 0, cred, NULL);
358 vfslocked = NDHASGIANT(&nd);
359 NDFREE(&nd, NDF_ONLY_PNBUF);
360 /* We just unlock so we hold a reference */
361 VOP_UNLOCK(nd.ni_vp, 0);
362 VFS_UNLOCK_GIANT(vfslocked);
364 alq = malloc(sizeof(*alq), M_ALD, M_WAITOK|M_ZERO);
365 alq->aq_entbuf = malloc(count * size, M_ALD, M_WAITOK|M_ZERO);
366 alq->aq_first = malloc(sizeof(*ale) * count, M_ALD, M_WAITOK|M_ZERO);
367 alq->aq_vp = nd.ni_vp;
368 alq->aq_cred = crhold(cred);
369 alq->aq_entmax = count;
370 alq->aq_entlen = size;
371 alq->aq_entfree = alq->aq_first;
373 mtx_init(&alq->aq_mtx, "ALD Queue", NULL, MTX_SPIN|MTX_QUIET);
375 bufp = alq->aq_entbuf;
379 /* Match up entries with buffers */
380 for (i = 0; i < count; i++) {
389 alp->ae_next = alq->aq_first;
391 if ((error = ald_add(alq)) != 0)
399 * Copy a new entry into the queue. If the operation would block either
400 * wait or return an error depending on the value of waitok.
403 alq_write(struct alq *alq, void *data, int waitok)
407 if ((ale = alq_get(alq, waitok)) == NULL)
408 return (EWOULDBLOCK);
410 bcopy(data, ale->ae_data, alq->aq_entlen);
417 alq_get(struct alq *alq, int waitok)
426 /* Loop until we get an entry or we're shutting down */
427 while ((alq->aq_flags & AQ_SHUTDOWN) == 0 &&
428 (ale = alq->aq_entfree) == NULL &&
429 (waitok & ALQ_WAITOK)) {
430 alq->aq_flags |= AQ_WANTED;
431 msleep_spin(alq, &alq->aq_mtx, "alqget", 0);
436 if ((aln->ae_flags & AE_VALID) == 0)
437 alq->aq_entfree = aln;
439 alq->aq_entfree = NULL;
448 alq_post(struct alq *alq, struct ale *ale)
452 ale->ae_flags |= AE_VALID;
454 if (alq->aq_entvalid == NULL)
455 alq->aq_entvalid = ale;
457 if ((alq->aq_flags & AQ_ACTIVE) == 0) {
458 alq->aq_flags |= AQ_ACTIVE;
472 alq_flush(struct alq *alq)
478 if (alq->aq_flags & AQ_ACTIVE) {
481 needwakeup = alq_doio(alq);
491 * Flush remaining data, close the file and free all resources.
494 alq_close(struct alq *alq)
497 * If we're already shuting down someone else will flush and close
500 if (ald_rem(alq) != 0)
504 * Drain all pending IO.
508 mtx_destroy(&alq->aq_mtx);
509 free(alq->aq_first, M_ALD);
510 free(alq->aq_entbuf, M_ALD);