2 * Copyright (c) 2012 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
17 * 3. Neither the name of The DragonFly Project nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific, prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 #include <sys/param.h>
35 #include <sys/systm.h>
36 #include <sys/kernel.h>
38 #include <sys/sysctl.h>
41 #include <sys/disklabel.h>
42 #include <sys/disklabel32.h>
43 #include <sys/disklabel64.h>
44 #include <sys/diskslice.h>
45 #include <sys/diskmbr.h>
47 #include <sys/malloc.h>
48 #include <sys/device.h>
49 #include <sys/devfs.h>
50 #include <sys/thread.h>
51 #include <sys/queue.h>
58 #include <sys/mplock2.h>
59 #include <sys/msgport2.h>
60 #include <sys/thread2.h>
73 static MALLOC_DEFINE(M_DMSG_DISK, "dmsg_disk", "disk dmsg");
75 static int disk_iocom_reconnect(struct disk *dp, struct file *fp);
76 static int disk_rcvdmsg(kdmsg_msg_t *msg);
78 static void disk_blk_open(struct disk *dp, kdmsg_msg_t *msg);
79 static void disk_blk_read(struct disk *dp, kdmsg_msg_t *msg);
80 static void disk_blk_write(struct disk *dp, kdmsg_msg_t *msg);
81 static void disk_blk_flush(struct disk *dp, kdmsg_msg_t *msg);
82 static void disk_blk_freeblks(struct disk *dp, kdmsg_msg_t *msg);
83 static void diskiodone(struct bio *bio);
86 disk_iocom_init(struct disk *dp)
88 kdmsg_iocom_init(&dp->d_iocom, dp,
89 KDMSG_IOCOMF_AUTOCONN |
90 KDMSG_IOCOMF_AUTORXSPAN |
91 KDMSG_IOCOMF_AUTOTXSPAN,
92 M_DMSG_DISK, disk_rcvdmsg);
96 disk_iocom_update(struct disk *dp)
101 disk_iocom_uninit(struct disk *dp)
103 kdmsg_iocom_uninit(&dp->d_iocom);
107 disk_iocom_ioctl(struct disk *dp, int cmd, void *data)
110 struct disk_ioc_recluster *recl;
116 fp = holdfp(curproc->p_fd, recl->fd, -1);
118 error = disk_iocom_reconnect(dp, fp);
132 disk_iocom_reconnect(struct disk *dp, struct file *fp)
136 ksnprintf(devname, sizeof(devname), "%s%d",
137 dev_dname(dp->d_rawdev), dkunit(dp->d_rawdev));
139 kdmsg_iocom_reconnect(&dp->d_iocom, fp, devname);
141 dp->d_iocom.auto_lnk_conn.pfs_type = DMSG_PFSTYPE_SERVER;
142 dp->d_iocom.auto_lnk_conn.proto_version = DMSG_SPAN_PROTO_1;
143 dp->d_iocom.auto_lnk_conn.peer_type = DMSG_PEER_BLOCK;
144 dp->d_iocom.auto_lnk_conn.peer_mask = 1LLU << DMSG_PEER_BLOCK;
145 dp->d_iocom.auto_lnk_conn.pfs_mask = (uint64_t)-1;
146 ksnprintf(dp->d_iocom.auto_lnk_conn.cl_label,
147 sizeof(dp->d_iocom.auto_lnk_conn.cl_label),
148 "%s/%s", hostname, devname);
149 if (dp->d_info.d_serialno) {
150 ksnprintf(dp->d_iocom.auto_lnk_conn.fs_label,
151 sizeof(dp->d_iocom.auto_lnk_conn.fs_label),
152 "%s", dp->d_info.d_serialno);
155 dp->d_iocom.auto_lnk_span.pfs_type = DMSG_PFSTYPE_SERVER;
156 dp->d_iocom.auto_lnk_span.proto_version = DMSG_SPAN_PROTO_1;
157 dp->d_iocom.auto_lnk_span.peer_type = DMSG_PEER_BLOCK;
158 dp->d_iocom.auto_lnk_span.media.block.bytes =
159 dp->d_info.d_media_size;
160 dp->d_iocom.auto_lnk_span.media.block.blksize =
161 dp->d_info.d_media_blksize;
162 ksnprintf(dp->d_iocom.auto_lnk_span.cl_label,
163 sizeof(dp->d_iocom.auto_lnk_span.cl_label),
164 "%s/%s", hostname, devname);
165 if (dp->d_info.d_serialno) {
166 ksnprintf(dp->d_iocom.auto_lnk_span.fs_label,
167 sizeof(dp->d_iocom.auto_lnk_span.fs_label),
168 "%s", dp->d_info.d_serialno);
171 kdmsg_iocom_autoinitiate(&dp->d_iocom, NULL);
177 disk_rcvdmsg(kdmsg_msg_t *msg)
179 struct disk *dp = msg->state->iocom->handle;
182 * Handle debug messages (these might not be in transactions)
184 switch(msg->any.head.cmd & DMSGF_CMDSWMASK) {
187 * Execute shell command (not supported atm)
189 kdmsg_msg_reply(msg, DMSG_ERR_NOSUPP);
191 case DMSG_DBG_SHELL | DMSGF_REPLY:
193 msg->aux_data[msg->aux_size - 1] = 0;
194 kprintf("diskiocom: DEBUGMSG: %s\n", msg->aux_data);
200 * All remaining messages must be in a transaction.
202 * NOTE! We currently don't care if the transaction is just
203 * the span transaction (for disk probes) or if it is the
204 * BLK_OPEN transaction.
206 * NOTE! We are switching on the first message's command. The
207 * actual message command within the transaction may be
208 * different (if streaming within a transaction).
210 if (msg->state == &msg->state->iocom->state0) {
211 kdmsg_msg_reply(msg, DMSG_ERR_NOSUPP);
215 switch(msg->state->rxcmd & DMSGF_CMDSWMASK) {
217 disk_blk_open(dp, msg);
221 * not reached normally but leave in for completeness
223 disk_blk_read(dp, msg);
226 disk_blk_write(dp, msg);
229 disk_blk_flush(dp, msg);
231 case DMSG_BLK_FREEBLKS:
232 disk_blk_freeblks(dp, msg);
235 if ((msg->any.head.cmd & DMSGF_REPLY) == 0) {
236 if (msg->any.head.cmd & DMSGF_DELETE)
237 kdmsg_msg_reply(msg, DMSG_ERR_NOSUPP);
239 kdmsg_msg_result(msg, DMSG_ERR_NOSUPP);
248 disk_blk_open(struct disk *dp, kdmsg_msg_t *msg)
250 struct dios_open *openst;
251 int error = DMSG_ERR_NOSUPP;
254 openst = msg->state->any.any;
255 if ((msg->any.head.cmd & DMSGF_CMDSWMASK) == DMSG_BLK_OPEN) {
256 if (openst == NULL) {
257 openst = kmalloc(sizeof(*openst), M_DEVBUF,
259 msg->state->any.any = openst;
262 if (msg->any.blk_open.modes & DMSG_BLKOPEN_RD)
264 if (msg->any.blk_open.modes & DMSG_BLKOPEN_WR)
266 error = dev_dopen(dp->d_rawdev, fflags, S_IFCHR, proc0.p_ucred, NULL);
270 if (msg->any.blk_open.modes & DMSG_BLKOPEN_RD)
272 if (msg->any.blk_open.modes & DMSG_BLKOPEN_WR)
277 if ((msg->any.head.cmd & DMSGF_CMDSWMASK) == DMSG_BLK_CLOSE &&
280 if ((msg->any.blk_open.modes & DMSG_BLKOPEN_RD) &&
284 if ((msg->any.blk_open.modes & DMSG_BLKOPEN_WR) &&
288 error = dev_dclose(dp->d_rawdev, fflags, S_IFCHR, NULL);
292 if (msg->any.blk_open.modes & DMSG_BLKOPEN_RD)
294 if (msg->any.blk_open.modes & DMSG_BLKOPEN_WR)
299 if (msg->any.head.cmd & DMSGF_DELETE) {
301 while (openst->openrd && openst->openwr) {
304 dev_dclose(dp->d_rawdev, FREAD|FWRITE, S_IFCHR, NULL);
306 while (openst->openrd) {
308 dev_dclose(dp->d_rawdev, FREAD, S_IFCHR, NULL);
310 while (openst->openwr) {
312 dev_dclose(dp->d_rawdev, FWRITE, S_IFCHR, NULL);
314 kfree(openst, M_DEVBUF);
315 msg->state->any.any = NULL;
317 kdmsg_msg_reply(msg, error);
319 kdmsg_msg_result(msg, error);
325 disk_blk_read(struct disk *dp, kdmsg_msg_t *msg)
327 struct dios_io *iost;
330 int error = DMSG_ERR_NOSUPP;
334 * Only DMSG_BLK_READ commands imply read ops.
336 iost = msg->state->any.any;
337 if ((msg->any.head.cmd & DMSGF_CMDSWMASK) == DMSG_BLK_READ) {
338 if (msg->any.blk_read.bytes < DEV_BSIZE ||
339 msg->any.blk_read.bytes > MAXPHYS) {
340 error = DMSG_ERR_PARAM;
344 iost = kmalloc(sizeof(*iost), M_DEVBUF,
346 msg->state->any.any = iost;
349 bp = geteblk(msg->any.blk_read.bytes);
351 bp->b_cmd = BUF_CMD_READ;
352 bp->b_bcount = msg->any.blk_read.bytes;
353 bp->b_resid = bp->b_bcount;
354 bio->bio_offset = msg->any.blk_read.offset;
355 bio->bio_caller_info1.ptr = msg->state;
356 bio->bio_done = diskiodone;
357 /* kdmsg_state_hold(msg->state); */
359 atomic_add_int(&iost->count, 1);
360 if (msg->any.head.cmd & DMSGF_DELETE)
363 dev_dstrategy(dp->d_rawdev, bio);
367 if (msg->any.head.cmd & DMSGF_DELETE) {
368 if (iost && iost->count == 0) {
369 kfree(iost, M_DEVBUF);
370 msg->state->any.any = NULL;
372 kdmsg_msg_reply(msg, error);
374 kdmsg_msg_result(msg, error);
381 disk_blk_write(struct disk *dp, kdmsg_msg_t *msg)
383 struct dios_io *iost;
386 int error = DMSG_ERR_NOSUPP;
390 * Only DMSG_BLK_WRITE commands imply read ops.
392 iost = msg->state->any.any;
393 if ((msg->any.head.cmd & DMSGF_CMDSWMASK) == DMSG_BLK_WRITE) {
394 if (msg->any.blk_write.bytes < DEV_BSIZE ||
395 msg->any.blk_write.bytes > MAXPHYS) {
396 error = DMSG_ERR_PARAM;
400 iost = kmalloc(sizeof(*iost), M_DEVBUF,
402 msg->state->any.any = iost;
406 * Issue WRITE. Short data implies zeros. Try to optimize
407 * the buffer cache buffer for the case where we can just
408 * use the message's data pointer.
411 if (msg->aux_size >= msg->any.blk_write.bytes)
414 bp = geteblk(msg->any.blk_write.bytes);
416 bp->b_cmd = BUF_CMD_WRITE;
417 bp->b_bcount = msg->any.blk_write.bytes;
418 bp->b_resid = bp->b_bcount;
419 if (msg->aux_size >= msg->any.blk_write.bytes) {
420 bp->b_data = msg->aux_data;
421 kdmsg_detach_aux_data(msg, &iost->data);
423 bcopy(msg->aux_data, bp->b_data, msg->aux_size);
424 bzero(bp->b_data + msg->aux_size,
425 msg->any.blk_write.bytes - msg->aux_size);
426 bzero(&iost->data, sizeof(iost->data));
428 bio->bio_offset = msg->any.blk_write.offset;
429 bio->bio_caller_info1.ptr = msg->state;
430 bio->bio_done = diskiodone;
431 /* kdmsg_state_hold(msg->state); */
433 atomic_add_int(&iost->count, 1);
434 if (msg->any.head.cmd & DMSGF_DELETE)
437 dev_dstrategy(dp->d_rawdev, bio);
441 if (msg->any.head.cmd & DMSGF_DELETE) {
442 if (iost && iost->count == 0) {
443 kfree(iost, M_DEVBUF);
444 msg->state->any.any = NULL;
446 kdmsg_msg_reply(msg, error);
448 kdmsg_msg_result(msg, error);
455 disk_blk_flush(struct disk *dp, kdmsg_msg_t *msg)
457 struct dios_io *iost;
460 int error = DMSG_ERR_NOSUPP;
464 * Only DMSG_BLK_FLUSH commands imply read ops.
466 iost = msg->state->any.any;
467 if ((msg->any.head.cmd & DMSGF_CMDSWMASK) == DMSG_BLK_FLUSH) {
469 iost = kmalloc(sizeof(*iost), M_DEVBUF,
471 msg->state->any.any = iost;
476 bp->b_cmd = BUF_CMD_FLUSH;
477 bp->b_bcount = msg->any.blk_flush.bytes;
479 bio->bio_offset = msg->any.blk_flush.offset;
480 bio->bio_caller_info1.ptr = msg->state;
481 bio->bio_done = diskiodone;
482 /* kdmsg_state_hold(msg->state); */
484 atomic_add_int(&iost->count, 1);
485 if (msg->any.head.cmd & DMSGF_DELETE)
488 dev_dstrategy(dp->d_rawdev, bio);
491 if (msg->any.head.cmd & DMSGF_DELETE) {
492 if (iost && iost->count == 0) {
493 kfree(iost, M_DEVBUF);
494 msg->state->any.any = NULL;
496 kdmsg_msg_reply(msg, error);
498 kdmsg_msg_result(msg, error);
505 disk_blk_freeblks(struct disk *dp, kdmsg_msg_t *msg)
507 struct dios_io *iost;
510 int error = DMSG_ERR_NOSUPP;
514 * Only DMSG_BLK_FREEBLKS commands imply read ops.
516 iost = msg->state->any.any;
517 if ((msg->any.head.cmd & DMSGF_CMDSWMASK) == DMSG_BLK_FREEBLKS) {
519 iost = kmalloc(sizeof(*iost), M_DEVBUF,
521 msg->state->any.any = iost;
526 bp->b_cmd = BUF_CMD_FREEBLKS;
527 bp->b_bcount = msg->any.blk_freeblks.bytes;
529 bio->bio_offset = msg->any.blk_freeblks.offset;
530 bio->bio_caller_info1.ptr = msg->state;
531 bio->bio_done = diskiodone;
532 /* kdmsg_state_hold(msg->state); */
534 atomic_add_int(&iost->count, 1);
535 if (msg->any.head.cmd & DMSGF_DELETE)
538 dev_dstrategy(dp->d_rawdev, bio);
541 if (msg->any.head.cmd & DMSGF_DELETE) {
542 if (iost && iost->count == 0) {
543 kfree(iost, M_DEVBUF);
544 msg->state->any.any = NULL;
546 kdmsg_msg_reply(msg, error);
548 kdmsg_msg_result(msg, error);
555 diskiodone(struct bio *bio)
557 struct buf *bp = bio->bio_buf;
558 kdmsg_state_t *state = bio->bio_caller_info1.ptr;
560 struct dios_io *iost = state->any.any;
567 cmd = DMSG_LNK_ERROR;
573 cmd = DMSG_LNK_ERROR;
575 bytes = bp->b_bcount;
578 if (bp->b_flags & B_ERROR) {
584 kdmsg_free_aux_data(&iost->data);
587 case BUF_CMD_FREEBLKS:
588 if (bp->b_flags & B_ERROR)
594 panic("diskiodone: Unknown bio cmd = %d\n",
595 bio->bio_buf->b_cmd);
596 error = 0; /* avoid compiler warning */
597 break; /* NOT REACHED */
601 * Convert error to DMSG_ERR_* code.
607 * Convert LNK_ERROR or BLK_ERROR if non-zero resid. READS will
608 * have already converted cmd to BLK_ERROR and set up data to return.
610 if (resid && cmd == DMSG_LNK_ERROR)
611 cmd = DMSG_BLK_ERROR;
612 /* XXX txcmd is delayed so this won't work for streaming */
613 if ((state->txcmd & DMSGF_CREATE) == 0) /* assume serialized */
616 if (atomic_fetchadd_int(&iost->count, -1) == 1)
619 atomic_add_int(&iost->count, -1);
624 * Allocate a basic or extended reply. Be careful not to populate
625 * extended header fields unless we allocated an extended reply.
627 rmsg = kdmsg_msg_alloc(state, cmd, NULL, 0);
629 rmsg->aux_data = kmalloc(bytes, state->iocom->mmsg, M_INTWAIT);
630 rmsg->aux_size = bytes;
631 rmsg->flags |= KDMSG_FLAG_AUXALLOC;
632 bcopy(data, rmsg->aux_data, bytes);
634 rmsg->any.blk_error.head.error = error;
635 if ((cmd & DMSGF_BASECMDMASK) == DMSG_BLK_ERROR)
636 rmsg->any.blk_error.resid = resid;
637 bio->bio_caller_info1.ptr = NULL;
638 /* kdmsg_state_drop(state); */
639 kdmsg_msg_write(rmsg);
640 if (bp->b_flags & B_PAGING) {
641 relpbuf(bio->bio_buf, NULL);
643 bp->b_flags |= B_INVAL | B_AGE;