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>
59 #include <sys/mplock2.h>
60 #include <sys/msgport2.h>
61 #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_AUTOSPAN
|
91 KDMSG_IOCOMF_AUTOCIRC
,
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
->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 are switching on the first message's command. The
203 * actual message command within the transaction may be
204 * different (if streaming within a transaction).
206 if (msg
->state
== NULL
) {
207 kdmsg_msg_reply(msg
, DMSG_ERR_NOSUPP
);
211 switch(msg
->state
->rxcmd
& DMSGF_CMDSWMASK
) {
214 disk_blk_open(dp
, msg
);
217 disk_blk_read(dp
, msg
);
220 disk_blk_write(dp
, msg
);
223 disk_blk_flush(dp
, msg
);
225 case DMSG_BLK_FREEBLKS
:
226 disk_blk_freeblks(dp
, msg
);
229 if ((msg
->any
.head
.cmd
& DMSGF_REPLY
) == 0) {
230 if (msg
->any
.head
.cmd
& DMSGF_DELETE
)
231 kdmsg_msg_reply(msg
, DMSG_ERR_NOSUPP
);
233 kdmsg_msg_result(msg
, DMSG_ERR_NOSUPP
);
242 disk_blk_open(struct disk
*dp
, kdmsg_msg_t
*msg
)
244 struct dios_open
*openst
;
245 int error
= DMSG_ERR_NOSUPP
;
248 openst
= msg
->state
->any
.any
;
249 if ((msg
->any
.head
.cmd
& DMSGF_CMDSWMASK
) == DMSG_BLK_OPEN
) {
250 if (openst
== NULL
) {
251 openst
= kmalloc(sizeof(*openst
), M_DEVBUF
,
253 msg
->state
->any
.any
= openst
;
256 if (msg
->any
.blk_open
.modes
& DMSG_BLKOPEN_RD
)
258 if (msg
->any
.blk_open
.modes
& DMSG_BLKOPEN_WR
)
260 error
= dev_dopen(dp
->d_rawdev
, fflags
, S_IFCHR
, proc0
.p_ucred
);
264 if (msg
->any
.blk_open
.modes
& DMSG_BLKOPEN_RD
)
266 if (msg
->any
.blk_open
.modes
& DMSG_BLKOPEN_WR
)
270 if ((msg
->any
.head
.cmd
& DMSGF_CMDSWMASK
) == DMSG_BLK_CLOSE
&&
273 if ((msg
->any
.blk_open
.modes
& DMSG_BLKOPEN_RD
) &&
277 if ((msg
->any
.blk_open
.modes
& DMSG_BLKOPEN_WR
) &&
281 error
= dev_dclose(dp
->d_rawdev
, fflags
, S_IFCHR
);
285 if (msg
->any
.blk_open
.modes
& DMSG_BLKOPEN_RD
)
287 if (msg
->any
.blk_open
.modes
& DMSG_BLKOPEN_WR
)
291 if (msg
->any
.head
.cmd
& DMSGF_DELETE
) {
293 while (openst
->openrd
&& openst
->openwr
) {
296 dev_dclose(dp
->d_rawdev
, FREAD
|FWRITE
, S_IFCHR
);
298 while (openst
->openrd
) {
300 dev_dclose(dp
->d_rawdev
, FREAD
, S_IFCHR
);
302 while (openst
->openwr
) {
304 dev_dclose(dp
->d_rawdev
, FWRITE
, S_IFCHR
);
306 kfree(openst
, M_DEVBUF
);
307 msg
->state
->any
.any
= NULL
;
309 kdmsg_msg_reply(msg
, error
);
311 kdmsg_msg_result(msg
, error
);
317 disk_blk_read(struct disk
*dp
, kdmsg_msg_t
*msg
)
319 struct dios_io
*iost
;
322 int error
= DMSG_ERR_NOSUPP
;
326 * Only DMSG_BLK_READ commands imply read ops.
328 iost
= msg
->state
->any
.any
;
329 if ((msg
->any
.head
.cmd
& DMSGF_CMDSWMASK
) == DMSG_BLK_READ
) {
330 if (msg
->any
.blk_read
.bytes
< DEV_BSIZE
||
331 msg
->any
.blk_read
.bytes
> MAXPHYS
) {
332 error
= DMSG_ERR_PARAM
;
336 iost
= kmalloc(sizeof(*iost
), M_DEVBUF
,
338 msg
->state
->any
.any
= iost
;
341 bp
= geteblk(msg
->any
.blk_read
.bytes
);
343 bp
->b_cmd
= BUF_CMD_READ
;
344 bp
->b_bcount
= msg
->any
.blk_read
.bytes
;
345 bp
->b_resid
= bp
->b_bcount
;
346 bio
->bio_offset
= msg
->any
.blk_read
.offset
;
347 bio
->bio_caller_info1
.ptr
= msg
->state
;
348 bio
->bio_done
= diskiodone
;
349 /* kdmsg_state_hold(msg->state); */
351 atomic_add_int(&iost
->count
, 1);
352 if (msg
->any
.head
.cmd
& DMSGF_DELETE
)
355 dev_dstrategy(dp
->d_rawdev
, bio
);
359 if (msg
->any
.head
.cmd
& DMSGF_DELETE
) {
360 if (iost
&& iost
->count
== 0) {
361 kfree(iost
, M_DEVBUF
);
362 msg
->state
->any
.any
= NULL
;
364 kdmsg_msg_reply(msg
, error
);
366 kdmsg_msg_result(msg
, error
);
373 disk_blk_write(struct disk
*dp
, kdmsg_msg_t
*msg
)
375 struct dios_io
*iost
;
378 int error
= DMSG_ERR_NOSUPP
;
382 * Only DMSG_BLK_WRITE commands imply read ops.
384 iost
= msg
->state
->any
.any
;
385 if ((msg
->any
.head
.cmd
& DMSGF_CMDSWMASK
) == DMSG_BLK_WRITE
) {
386 if (msg
->any
.blk_write
.bytes
< DEV_BSIZE
||
387 msg
->any
.blk_write
.bytes
> MAXPHYS
) {
388 error
= DMSG_ERR_PARAM
;
392 iost
= kmalloc(sizeof(*iost
), M_DEVBUF
,
394 msg
->state
->any
.any
= iost
;
398 * Issue WRITE. Short data implies zeros. Try to optimize
399 * the buffer cache buffer for the case where we can just
400 * use the message's data pointer.
403 if (msg
->aux_size
>= msg
->any
.blk_write
.bytes
)
406 bp
= geteblk(msg
->any
.blk_write
.bytes
);
408 bp
->b_cmd
= BUF_CMD_WRITE
;
409 bp
->b_bcount
= msg
->any
.blk_write
.bytes
;
410 bp
->b_resid
= bp
->b_bcount
;
411 if (msg
->aux_size
>= msg
->any
.blk_write
.bytes
) {
412 bp
->b_data
= msg
->aux_data
;
414 bcopy(msg
->aux_data
, bp
->b_data
, msg
->aux_size
);
415 bzero(bp
->b_data
+ msg
->aux_size
,
416 msg
->any
.blk_write
.bytes
- msg
->aux_size
);
418 bio
->bio_offset
= msg
->any
.blk_write
.offset
;
419 bio
->bio_caller_info1
.ptr
= msg
->state
;
420 bio
->bio_done
= diskiodone
;
421 /* kdmsg_state_hold(msg->state); */
423 atomic_add_int(&iost
->count
, 1);
424 if (msg
->any
.head
.cmd
& DMSGF_DELETE
)
427 dev_dstrategy(dp
->d_rawdev
, bio
);
431 if (msg
->any
.head
.cmd
& DMSGF_DELETE
) {
432 if (iost
&& iost
->count
== 0) {
433 kfree(iost
, M_DEVBUF
);
434 msg
->state
->any
.any
= NULL
;
436 kdmsg_msg_reply(msg
, error
);
438 kdmsg_msg_result(msg
, error
);
445 disk_blk_flush(struct disk
*dp
, kdmsg_msg_t
*msg
)
447 struct dios_io
*iost
;
450 int error
= DMSG_ERR_NOSUPP
;
454 * Only DMSG_BLK_FLUSH commands imply read ops.
456 iost
= msg
->state
->any
.any
;
457 if ((msg
->any
.head
.cmd
& DMSGF_CMDSWMASK
) == DMSG_BLK_FLUSH
) {
459 iost
= kmalloc(sizeof(*iost
), M_DEVBUF
,
461 msg
->state
->any
.any
= iost
;
466 bp
->b_cmd
= BUF_CMD_FLUSH
;
467 bp
->b_bcount
= msg
->any
.blk_flush
.bytes
;
469 bio
->bio_offset
= msg
->any
.blk_flush
.offset
;
470 bio
->bio_caller_info1
.ptr
= msg
->state
;
471 bio
->bio_done
= diskiodone
;
472 /* kdmsg_state_hold(msg->state); */
474 atomic_add_int(&iost
->count
, 1);
475 if (msg
->any
.head
.cmd
& DMSGF_DELETE
)
478 dev_dstrategy(dp
->d_rawdev
, bio
);
481 if (msg
->any
.head
.cmd
& DMSGF_DELETE
) {
482 if (iost
&& iost
->count
== 0) {
483 kfree(iost
, M_DEVBUF
);
484 msg
->state
->any
.any
= NULL
;
486 kdmsg_msg_reply(msg
, error
);
488 kdmsg_msg_result(msg
, error
);
495 disk_blk_freeblks(struct disk
*dp
, kdmsg_msg_t
*msg
)
497 struct dios_io
*iost
;
500 int error
= DMSG_ERR_NOSUPP
;
504 * Only DMSG_BLK_FREEBLKS commands imply read ops.
506 iost
= msg
->state
->any
.any
;
507 if ((msg
->any
.head
.cmd
& DMSGF_CMDSWMASK
) == DMSG_BLK_FREEBLKS
) {
509 iost
= kmalloc(sizeof(*iost
), M_DEVBUF
,
511 msg
->state
->any
.any
= iost
;
516 bp
->b_cmd
= BUF_CMD_FREEBLKS
;
517 bp
->b_bcount
= msg
->any
.blk_freeblks
.bytes
;
519 bio
->bio_offset
= msg
->any
.blk_freeblks
.offset
;
520 bio
->bio_caller_info1
.ptr
= msg
->state
;
521 bio
->bio_done
= diskiodone
;
522 /* kdmsg_state_hold(msg->state); */
524 atomic_add_int(&iost
->count
, 1);
525 if (msg
->any
.head
.cmd
& DMSGF_DELETE
)
528 dev_dstrategy(dp
->d_rawdev
, bio
);
531 if (msg
->any
.head
.cmd
& DMSGF_DELETE
) {
532 if (iost
&& iost
->count
== 0) {
533 kfree(iost
, M_DEVBUF
);
534 msg
->state
->any
.any
= NULL
;
536 kdmsg_msg_reply(msg
, error
);
538 kdmsg_msg_result(msg
, error
);
545 diskiodone(struct bio
*bio
)
547 struct buf
*bp
= bio
->bio_buf
;
548 kdmsg_state_t
*state
= bio
->bio_caller_info1
.ptr
;
550 struct dios_io
*iost
= state
->any
.any
;
557 cmd
= DMSG_LNK_ERROR
;
563 cmd
= DMSG_LNK_ERROR
;
565 bytes
= bp
->b_bcount
;
568 if (bp
->b_flags
& B_ERROR
) {
576 case BUF_CMD_FREEBLKS
:
577 if (bp
->b_flags
& B_ERROR
)
583 panic("diskiodone: Unknown bio cmd = %d\n",
584 bio
->bio_buf
->b_cmd
);
585 error
= 0; /* avoid compiler warning */
586 break; /* NOT REACHED */
590 * Convert error to DMSG_ERR_* code.
596 * Convert LNK_ERROR or BLK_ERROR if non-zero resid. READS will
597 * have already converted cmd to BLK_ERROR and set up data to return.
599 if (resid
&& cmd
== DMSG_LNK_ERROR
)
600 cmd
= DMSG_BLK_ERROR
;
601 /* XXX txcmd is delayed so this won't work for streaming */
602 if ((state
->txcmd
& DMSGF_CREATE
) == 0) /* assume serialized */
605 if (atomic_fetchadd_int(&iost
->count
, -1) == 1)
608 atomic_add_int(&iost
->count
, -1);
613 * Allocate a basic or extended reply. Be careful not to populate
614 * extended header fields unless we allocated an extended reply.
616 rmsg
= kdmsg_msg_alloc_state(state
, cmd
, NULL
, 0);
618 rmsg
->aux_data
= kmalloc(bytes
, state
->iocom
->mmsg
, M_INTWAIT
);
619 rmsg
->aux_size
= bytes
;
620 rmsg
->flags
|= KDMSG_FLAG_AUXALLOC
;
621 bcopy(data
, rmsg
->aux_data
, bytes
);
623 rmsg
->any
.blk_error
.head
.error
= error
;
624 if ((cmd
& DMSGF_BASECMDMASK
) == DMSG_BLK_ERROR
)
625 rmsg
->any
.blk_error
.resid
= resid
;
626 bio
->bio_caller_info1
.ptr
= NULL
;
627 /* kdmsg_state_drop(state); */
628 kdmsg_msg_write(rmsg
);
629 if (bp
->b_flags
& B_PAGING
) {
630 relpbuf(bio
->bio_buf
, NULL
);
632 bp
->b_flags
|= B_INVAL
| B_AGE
;