2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 2023 Tomohiro Kusumi <tkusumi@netbsd.org>
5 * Copyright (c) 2011-2023 The DragonFly Project. All rights reserved.
7 * This code is derived from software contributed to The DragonFly Project
8 * by Matthew Dillon <dillon@dragonflybsd.org>
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in
18 * the documentation and/or other materials provided with the
20 * 3. Neither the name of The DragonFly Project nor the names of its
21 * contributors may be used to endorse or promote products derived
22 * from this software without specific, prior written permission.
24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
25 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
26 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
27 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
28 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
29 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
30 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
31 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
32 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
33 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
34 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
40 * WARNING! The ioctl functions which manipulate the connection state need
41 * to be able to run without deadlock on the volume's chain lock.
42 * Most of these functions use a separate lock.
48 #include <sys/kern_syscall.h>
50 static int hammer2_ioctl_version_get(hammer2_inode_t
*ip
, void *data
);
51 static int hammer2_ioctl_recluster(hammer2_inode_t
*ip
, void *data
);
52 static int hammer2_ioctl_remote_scan(hammer2_inode_t
*ip
, void *data
);
53 static int hammer2_ioctl_remote_add(hammer2_inode_t
*ip
, void *data
);
54 static int hammer2_ioctl_remote_del(hammer2_inode_t
*ip
, void *data
);
55 static int hammer2_ioctl_remote_rep(hammer2_inode_t
*ip
, void *data
);
56 static int hammer2_ioctl_socket_get(hammer2_inode_t
*ip
, void *data
);
57 static int hammer2_ioctl_socket_set(hammer2_inode_t
*ip
, void *data
);
58 static int hammer2_ioctl_pfs_get(hammer2_inode_t
*ip
, void *data
);
59 static int hammer2_ioctl_pfs_lookup(hammer2_inode_t
*ip
, void *data
);
60 static int hammer2_ioctl_pfs_create(hammer2_inode_t
*ip
, void *data
);
61 static int hammer2_ioctl_pfs_snapshot(hammer2_inode_t
*ip
, void *data
);
62 static int hammer2_ioctl_pfs_delete(hammer2_inode_t
*ip
, void *data
);
63 static int hammer2_ioctl_inode_get(hammer2_inode_t
*ip
, void *data
);
64 static int hammer2_ioctl_inode_set(hammer2_inode_t
*ip
, void *data
);
65 static int hammer2_ioctl_debug_dump(hammer2_inode_t
*ip
, u_int flags
);
66 static int hammer2_ioctl_emerg_mode(hammer2_inode_t
*ip
, u_int mode
);
67 static int hammer2_ioctl_growfs(hammer2_inode_t
*ip
, void *data
,
69 //static int hammer2_ioctl_inode_comp_set(hammer2_inode_t *ip, void *data);
70 //static int hammer2_ioctl_inode_comp_rec_set(hammer2_inode_t *ip, void *data);
71 //static int hammer2_ioctl_inode_comp_rec_set2(hammer2_inode_t *ip, void *data);
72 static int hammer2_ioctl_bulkfree_scan(hammer2_inode_t
*ip
, void *data
);
73 static int hammer2_ioctl_destroy(hammer2_inode_t
*ip
, void *data
);
74 static int hammer2_ioctl_volume_list(hammer2_inode_t
*ip
, void *data
);
77 hammer2_ioctl(hammer2_inode_t
*ip
, u_long com
, void *data
, int fflag
,
83 * Standard root cred checks, will be selectively ignored below
84 * for ioctls that do not require root creds.
86 error
= priv_check_cred(cred
, PRIV_HAMMER_IOCTL
, 0);
89 case HAMMER2IOC_VERSION_GET
:
90 error
= hammer2_ioctl_version_get(ip
, data
);
92 case HAMMER2IOC_RECLUSTER
:
94 error
= hammer2_ioctl_recluster(ip
, data
);
96 case HAMMER2IOC_REMOTE_SCAN
:
98 error
= hammer2_ioctl_remote_scan(ip
, data
);
100 case HAMMER2IOC_REMOTE_ADD
:
102 error
= hammer2_ioctl_remote_add(ip
, data
);
104 case HAMMER2IOC_REMOTE_DEL
:
106 error
= hammer2_ioctl_remote_del(ip
, data
);
108 case HAMMER2IOC_REMOTE_REP
:
110 error
= hammer2_ioctl_remote_rep(ip
, data
);
112 case HAMMER2IOC_SOCKET_GET
:
114 error
= hammer2_ioctl_socket_get(ip
, data
);
116 case HAMMER2IOC_SOCKET_SET
:
118 error
= hammer2_ioctl_socket_set(ip
, data
);
120 case HAMMER2IOC_PFS_GET
:
122 error
= hammer2_ioctl_pfs_get(ip
, data
);
124 case HAMMER2IOC_PFS_LOOKUP
:
126 error
= hammer2_ioctl_pfs_lookup(ip
, data
);
128 case HAMMER2IOC_PFS_CREATE
:
130 error
= hammer2_ioctl_pfs_create(ip
, data
);
132 case HAMMER2IOC_PFS_DELETE
:
134 error
= hammer2_ioctl_pfs_delete(ip
, data
);
136 case HAMMER2IOC_PFS_SNAPSHOT
:
138 error
= hammer2_ioctl_pfs_snapshot(ip
, data
);
140 case HAMMER2IOC_INODE_GET
:
141 error
= hammer2_ioctl_inode_get(ip
, data
);
143 case HAMMER2IOC_INODE_SET
:
145 error
= hammer2_ioctl_inode_set(ip
, data
);
147 case HAMMER2IOC_BULKFREE_SCAN
:
148 error
= hammer2_ioctl_bulkfree_scan(ip
, data
);
150 case HAMMER2IOC_BULKFREE_ASYNC
:
151 error
= hammer2_ioctl_bulkfree_scan(ip
, NULL
);
153 case HAMMER2IOC_DESTROY
:
155 error
= hammer2_ioctl_destroy(ip
, data
);
157 case HAMMER2IOC_DEBUG_DUMP
:
158 error
= hammer2_ioctl_debug_dump(ip
, *(u_int
*)data
);
160 case HAMMER2IOC_EMERG_MODE
:
162 error
= hammer2_ioctl_emerg_mode(ip
, *(u_int
*)data
);
164 case HAMMER2IOC_GROWFS
:
166 error
= hammer2_ioctl_growfs(ip
, data
, cred
);
168 case HAMMER2IOC_VOLUME_LIST
:
170 error
= hammer2_ioctl_volume_list(ip
, data
);
180 * Retrieve version and basic info
183 hammer2_ioctl_version_get(hammer2_inode_t
*ip
, void *data
)
185 hammer2_ioc_version_t
*version
= data
;
188 hmp
= ip
->pmp
->pfs_hmps
[0];
190 version
->version
= hmp
->voldata
.version
;
192 version
->version
= -1;
197 hammer2_ioctl_recluster(hammer2_inode_t
*ip
, void *data
)
199 hammer2_ioc_recluster_t
*recl
= data
;
200 struct vnode
*vproot
;
202 hammer2_cluster_t
*cluster
;
205 fp
= holdfp(curthread
, recl
->fd
, -1);
207 error
= VFS_ROOT(ip
->pmp
->mp
, &vproot
);
209 cluster
= &ip
->pmp
->iroot
->cluster
;
210 kprintf("reconnect to cluster: nc=%d focus=%p\n",
211 cluster
->nchains
, cluster
->focus
);
212 if (cluster
->nchains
!= 1 || cluster
->focus
== NULL
) {
213 kprintf("not a local device mount\n");
216 hammer2_cluster_reconnect(cluster
->focus
->hmp
,
230 * Retrieve information about a remote
233 hammer2_ioctl_remote_scan(hammer2_inode_t
*ip
, void *data
)
236 hammer2_ioc_remote_t
*remote
= data
;
237 int copyid
= remote
->copyid
;
239 hmp
= ip
->pmp
->pfs_hmps
[0];
243 if (copyid
< 0 || copyid
>= HAMMER2_COPYID_COUNT
)
246 hammer2_voldata_lock(hmp
);
247 remote
->copy1
= hmp
->voldata
.copyinfo
[copyid
];
248 hammer2_voldata_unlock(hmp
);
251 * Adjust nextid (GET only)
253 while (++copyid
< HAMMER2_COPYID_COUNT
&&
254 hmp
->voldata
.copyinfo
[copyid
].copyid
== 0) {
257 if (copyid
== HAMMER2_COPYID_COUNT
)
260 remote
->nextid
= copyid
;
266 * Add new remote entry
269 hammer2_ioctl_remote_add(hammer2_inode_t
*ip
, void *data
)
271 hammer2_ioc_remote_t
*remote
= data
;
272 hammer2_pfs_t
*pmp
= ip
->pmp
;
274 int copyid
= remote
->copyid
;
277 hmp
= pmp
->pfs_hmps
[0];
280 if (copyid
>= HAMMER2_COPYID_COUNT
)
283 hammer2_voldata_lock(hmp
);
285 for (copyid
= 1; copyid
< HAMMER2_COPYID_COUNT
; ++copyid
) {
286 if (hmp
->voldata
.copyinfo
[copyid
].copyid
== 0)
289 if (copyid
== HAMMER2_COPYID_COUNT
) {
294 hammer2_voldata_modify(hmp
);
295 remote
->copy1
.copyid
= copyid
;
296 hmp
->voldata
.copyinfo
[copyid
] = remote
->copy1
;
297 hammer2_volconf_update(hmp
, copyid
);
299 hammer2_voldata_unlock(hmp
);
304 * Delete existing remote entry
307 hammer2_ioctl_remote_del(hammer2_inode_t
*ip
, void *data
)
309 hammer2_ioc_remote_t
*remote
= data
;
310 hammer2_pfs_t
*pmp
= ip
->pmp
;
312 int copyid
= remote
->copyid
;
315 hmp
= pmp
->pfs_hmps
[0];
318 if (copyid
>= HAMMER2_COPYID_COUNT
)
320 remote
->copy1
.path
[sizeof(remote
->copy1
.path
) - 1] = 0;
321 hammer2_voldata_lock(hmp
);
323 for (copyid
= 1; copyid
< HAMMER2_COPYID_COUNT
; ++copyid
) {
324 if (hmp
->voldata
.copyinfo
[copyid
].copyid
== 0)
326 if (strcmp(remote
->copy1
.path
,
327 hmp
->voldata
.copyinfo
[copyid
].path
) == 0) {
331 if (copyid
== HAMMER2_COPYID_COUNT
) {
336 hammer2_voldata_modify(hmp
);
337 hmp
->voldata
.copyinfo
[copyid
].copyid
= 0;
338 hammer2_volconf_update(hmp
, copyid
);
340 hammer2_voldata_unlock(hmp
);
345 * Replace existing remote entry
348 hammer2_ioctl_remote_rep(hammer2_inode_t
*ip
, void *data
)
350 hammer2_ioc_remote_t
*remote
= data
;
352 int copyid
= remote
->copyid
;
354 hmp
= ip
->pmp
->pfs_hmps
[0];
357 if (copyid
< 0 || copyid
>= HAMMER2_COPYID_COUNT
)
360 hammer2_voldata_lock(hmp
);
361 hammer2_voldata_modify(hmp
);
362 /*hammer2_volconf_update(hmp, copyid);*/
363 hammer2_voldata_unlock(hmp
);
369 * Retrieve communications socket
372 hammer2_ioctl_socket_get(hammer2_inode_t
*ip
, void *data
)
378 * Set communications socket for connection
381 hammer2_ioctl_socket_set(hammer2_inode_t
*ip
, void *data
)
383 hammer2_ioc_remote_t
*remote
= data
;
385 int copyid
= remote
->copyid
;
387 hmp
= ip
->pmp
->pfs_hmps
[0];
390 if (copyid
< 0 || copyid
>= HAMMER2_COPYID_COUNT
)
393 hammer2_voldata_lock(hmp
);
394 hammer2_voldata_unlock(hmp
);
400 * Used to scan and retrieve PFS information. PFS's are directories under
403 * To scan PFSs pass name_key=0. The function will scan for the next
404 * PFS and set all fields, as well as set name_next to the next key.
405 * When no PFSs remain, name_next is set to (hammer2_key_t)-1.
407 * To retrieve a particular PFS by key, specify the key but note that
408 * the ioctl will return the lowest key >= specified_key, so the caller
409 * must verify the key.
411 * To retrieve the PFS associated with the file descriptor, pass
412 * name_key set to (hammer2_key_t)-1.
415 hammer2_ioctl_pfs_get(hammer2_inode_t
*ip
, void *data
)
417 const hammer2_inode_data_t
*ripdata
;
419 hammer2_ioc_pfs_t
*pfs
;
420 hammer2_chain_t
*parent
;
421 hammer2_chain_t
*chain
;
422 hammer2_key_t key_next
;
423 hammer2_key_t save_key
;
426 hmp
= ip
->pmp
->pfs_hmps
[0];
431 save_key
= pfs
->name_key
;
437 if (save_key
== (hammer2_key_t
)-1) {
438 hammer2_inode_lock(ip
->pmp
->iroot
, 0);
440 chain
= hammer2_inode_chain(ip
->pmp
->iroot
, 0,
441 HAMMER2_RESOLVE_ALWAYS
|
442 HAMMER2_RESOLVE_SHARED
);
444 hammer2_inode_lock(hmp
->spmp
->iroot
, 0);
445 parent
= hammer2_inode_chain(hmp
->spmp
->iroot
, 0,
446 HAMMER2_RESOLVE_ALWAYS
|
447 HAMMER2_RESOLVE_SHARED
);
448 chain
= hammer2_chain_lookup(&parent
, &key_next
,
449 pfs
->name_key
, HAMMER2_KEY_MAX
,
451 HAMMER2_LOOKUP_SHARED
);
458 if (chain
->bref
.type
== HAMMER2_BREF_TYPE_INODE
)
460 if (parent
== NULL
) {
461 hammer2_chain_unlock(chain
);
462 hammer2_chain_drop(chain
);
466 chain
= hammer2_chain_next(&parent
, chain
, &key_next
,
467 key_next
, HAMMER2_KEY_MAX
,
469 HAMMER2_LOOKUP_SHARED
);
471 error
= hammer2_error_to_errno(error
);
474 * Load the data being returned by the ioctl.
476 if (chain
&& chain
->error
== 0) {
477 ripdata
= &chain
->data
->ipdata
;
478 pfs
->name_key
= ripdata
->meta
.name_key
;
479 pfs
->pfs_type
= ripdata
->meta
.pfs_type
;
480 pfs
->pfs_subtype
= ripdata
->meta
.pfs_subtype
;
481 pfs
->pfs_clid
= ripdata
->meta
.pfs_clid
;
482 pfs
->pfs_fsid
= ripdata
->meta
.pfs_fsid
;
483 KKASSERT(ripdata
->meta
.name_len
< sizeof(pfs
->name
));
484 bcopy(ripdata
->filename
, pfs
->name
, ripdata
->meta
.name_len
);
485 pfs
->name
[ripdata
->meta
.name_len
] = 0;
486 ripdata
= NULL
; /* safety */
489 * Calculate name_next, if any. We are only accessing
490 * chain->bref so we can ignore chain->error (if the key
491 * is used later it will error then).
493 if (parent
== NULL
) {
494 pfs
->name_next
= (hammer2_key_t
)-1;
496 chain
= hammer2_chain_next(&parent
, chain
, &key_next
,
497 key_next
, HAMMER2_KEY_MAX
,
499 HAMMER2_LOOKUP_SHARED
);
501 pfs
->name_next
= chain
->bref
.key
;
503 pfs
->name_next
= (hammer2_key_t
)-1;
506 pfs
->name_next
= (hammer2_key_t
)-1;
514 hammer2_chain_unlock(chain
);
515 hammer2_chain_drop(chain
);
518 hammer2_chain_unlock(parent
);
519 hammer2_chain_drop(parent
);
521 if (save_key
== (hammer2_key_t
)-1) {
522 hammer2_inode_unlock(ip
->pmp
->iroot
);
524 hammer2_inode_unlock(hmp
->spmp
->iroot
);
531 * Find a specific PFS by name
534 hammer2_ioctl_pfs_lookup(hammer2_inode_t
*ip
, void *data
)
536 const hammer2_inode_data_t
*ripdata
;
538 hammer2_ioc_pfs_t
*pfs
;
539 hammer2_chain_t
*parent
;
540 hammer2_chain_t
*chain
;
541 hammer2_key_t key_next
;
546 hmp
= ip
->pmp
->pfs_hmps
[0];
553 hammer2_inode_lock(hmp
->spmp
->iroot
, HAMMER2_RESOLVE_SHARED
);
554 parent
= hammer2_inode_chain(hmp
->spmp
->iroot
, 0,
555 HAMMER2_RESOLVE_ALWAYS
|
556 HAMMER2_RESOLVE_SHARED
);
558 pfs
->name
[sizeof(pfs
->name
) - 1] = 0;
559 len
= strlen(pfs
->name
);
560 lhc
= hammer2_dirhash(pfs
->name
, len
);
562 chain
= hammer2_chain_lookup(&parent
, &key_next
,
563 lhc
, lhc
+ HAMMER2_DIRHASH_LOMASK
,
564 &error
, HAMMER2_LOOKUP_SHARED
);
566 if (hammer2_chain_dirent_test(chain
, pfs
->name
, len
))
568 chain
= hammer2_chain_next(&parent
, chain
, &key_next
,
570 lhc
+ HAMMER2_DIRHASH_LOMASK
,
571 &error
, HAMMER2_LOOKUP_SHARED
);
573 error
= hammer2_error_to_errno(error
);
576 * Load the data being returned by the ioctl.
578 if (chain
&& chain
->error
== 0) {
579 KKASSERT(chain
->bref
.type
== HAMMER2_BREF_TYPE_INODE
);
580 ripdata
= &chain
->data
->ipdata
;
581 pfs
->name_key
= ripdata
->meta
.name_key
;
582 pfs
->pfs_type
= ripdata
->meta
.pfs_type
;
583 pfs
->pfs_subtype
= ripdata
->meta
.pfs_subtype
;
584 pfs
->pfs_clid
= ripdata
->meta
.pfs_clid
;
585 pfs
->pfs_fsid
= ripdata
->meta
.pfs_fsid
;
588 hammer2_chain_unlock(chain
);
589 hammer2_chain_drop(chain
);
590 } else if (error
== 0) {
594 hammer2_chain_unlock(parent
);
595 hammer2_chain_drop(parent
);
597 hammer2_inode_unlock(hmp
->spmp
->iroot
);
603 * Create a new PFS under the super-root
606 hammer2_ioctl_pfs_create(hammer2_inode_t
*ip
, void *data
)
608 hammer2_inode_data_t
*nipdata
;
609 hammer2_chain_t
*nchain
;
611 hammer2_dev_t
*force_local
;
612 hammer2_ioc_pfs_t
*pfs
;
613 hammer2_inode_t
*nip
;
617 hmp
= ip
->pmp
->pfs_hmps
[0]; /* XXX */
624 if (pfs
->name
[0] == 0)
626 pfs
->name
[sizeof(pfs
->name
) - 1] = 0; /* ensure 0-termination */
628 if (hammer2_ioctl_pfs_lookup(ip
, pfs
) == 0)
631 hammer2_trans_init(hmp
->spmp
, HAMMER2_TRANS_ISFLUSH
);
632 mtid
= hammer2_trans_sub(hmp
->spmp
);
633 nip
= hammer2_inode_create_pfs(hmp
->spmp
, pfs
->name
, strlen(pfs
->name
),
636 atomic_set_int(&nip
->flags
, HAMMER2_INODE_NOSIDEQ
);
637 hammer2_inode_modify(nip
);
638 nchain
= hammer2_inode_chain(nip
, 0, HAMMER2_RESOLVE_ALWAYS
);
639 error
= hammer2_chain_modify(nchain
, mtid
, 0, 0);
640 KKASSERT(error
== 0);
641 nipdata
= &nchain
->data
->ipdata
;
643 nip
->meta
.pfs_type
= pfs
->pfs_type
;
644 nip
->meta
.pfs_subtype
= pfs
->pfs_subtype
;
645 nip
->meta
.pfs_clid
= pfs
->pfs_clid
;
646 nip
->meta
.pfs_fsid
= pfs
->pfs_fsid
;
647 nip
->meta
.op_flags
|= HAMMER2_OPFLAG_PFSROOT
;
650 * Set default compression and check algorithm. This
651 * can be changed later.
653 * Do not allow compression on PFS's with the special name
654 * "boot", the boot loader can't decompress (yet).
656 nip
->meta
.comp_algo
=
657 HAMMER2_ENC_ALGO(HAMMER2_COMP_NEWFS_DEFAULT
);
658 nip
->meta
.check_algo
=
659 HAMMER2_ENC_ALGO( HAMMER2_CHECK_XXHASH64
);
661 if (strcasecmp(pfs
->name
, "boot") == 0) {
662 nip
->meta
.comp_algo
=
663 HAMMER2_ENC_ALGO(HAMMER2_COMP_AUTOZERO
);
667 * Super-root isn't mounted, fsync it
669 hammer2_chain_unlock(nchain
);
670 hammer2_inode_ref(nip
);
671 hammer2_inode_unlock(nip
);
672 hammer2_inode_chain_sync(nip
);
673 hammer2_inode_chain_flush(nip
, HAMMER2_XOP_INODE_STOP
|
675 hammer2_inode_drop(nip
);
679 * We still have a ref on the chain, relock and associate
680 * with an appropriate PFS.
682 force_local
= (hmp
->hflags
& HMNT2_LOCAL
) ? hmp
: NULL
;
684 hammer2_chain_lock(nchain
, HAMMER2_RESOLVE_ALWAYS
);
685 nipdata
= &nchain
->data
->ipdata
;
686 kprintf("ADD LOCAL PFS (IOCTL): %s\n", nipdata
->filename
);
687 hammer2_pfsalloc(nchain
, nipdata
, force_local
);
689 hammer2_chain_unlock(nchain
);
690 hammer2_chain_drop(nchain
);
692 hammer2_trans_done(hmp
->spmp
, HAMMER2_TRANS_ISFLUSH
|
693 HAMMER2_TRANS_SIDEQ
);
699 * Destroy an existing PFS under the super-root
702 hammer2_ioctl_pfs_delete(hammer2_inode_t
*ip
, void *data
)
704 hammer2_ioc_pfs_t
*pfs
= data
;
708 hammer2_xop_unlink_t
*xop
;
709 hammer2_inode_t
*dip
;
710 hammer2_inode_t
*iroot
;
715 * The PFS should be probed, so we should be able to
716 * locate it. We only delete the PFS from the
717 * specific H2 block device (hmp), not all of
718 * them. We must remove the PFS from the cluster
719 * before we can destroy it.
721 hmp
= ip
->pmp
->pfs_hmps
[0];
725 pfs
->name
[sizeof(pfs
->name
) - 1] = 0; /* ensure termination */
727 lockmgr(&hammer2_mntlk
, LK_EXCLUSIVE
);
729 TAILQ_FOREACH(pmp
, &hammer2_pfslist
, mntentry
) {
730 for (i
= 0; i
< HAMMER2_MAXCLUSTER
; ++i
) {
731 if (pmp
->pfs_hmps
[i
] != hmp
)
733 if (pmp
->pfs_names
[i
] &&
734 strcmp(pmp
->pfs_names
[i
], pfs
->name
) == 0) {
738 if (i
!= HAMMER2_MAXCLUSTER
)
743 lockmgr(&hammer2_mntlk
, LK_RELEASE
);
747 lockmgr(&hammer2_mntlk
, LK_RELEASE
);
752 * Ok, we found the pmp and we have the index. Permanently remove
753 * the PFS from the cluster
756 kprintf("FOUND PFS %s CLINDEX %d\n", pfs
->name
, i
);
757 hammer2_pfsdealloc(pmp
, i
, 1);
759 lockmgr(&hammer2_mntlk
, LK_RELEASE
);
762 * Now destroy the PFS under its device using the per-device
767 hammer2_trans_init(spmp
, 0);
768 hammer2_inode_lock(dip
, 0);
770 xop
= hammer2_xop_alloc(dip
, HAMMER2_XOP_MODIFYING
);
771 hammer2_xop_setname(&xop
->head
, pfs
->name
, strlen(pfs
->name
));
773 xop
->dopermanent
= H2DOPERM_PERMANENT
| H2DOPERM_FORCE
;
774 hammer2_xop_start(&xop
->head
, &hammer2_unlink_desc
);
776 error
= hammer2_xop_collect(&xop
->head
, 0);
778 hammer2_inode_unlock(dip
);
782 ip
= hammer2_inode_get(dip
->pmp
, &xop
->head
, -1, -1);
783 hammer2_xop_retire(&xop
->head
, HAMMER2_XOPMASK_VOP
);
785 hammer2_inode_unlink_finisher(ip
, NULL
);
786 hammer2_inode_unlock(ip
);
789 hammer2_xop_retire(&xop
->head
, HAMMER2_XOPMASK_VOP
);
792 hammer2_xop_retire(&xop
->head
, HAMMER2_XOPMASK_VOP
);
794 hammer2_trans_done(spmp
, HAMMER2_TRANS_SIDEQ
);
796 return (hammer2_error_to_errno(error
));
800 hammer2_ioctl_pfs_snapshot(hammer2_inode_t
*ip
, void *data
)
802 hammer2_ioc_pfs_t
*pfs
= data
;
805 hammer2_chain_t
*chain
;
806 hammer2_inode_t
*nip
;
815 if (pfs
->name
[0] == 0)
817 if (pfs
->name
[sizeof(pfs
->name
)-1] != 0)
823 hmp
= pmp
->pfs_hmps
[0];
827 lockmgr(&hmp
->bulklk
, LK_EXCLUSIVE
);
830 * NOSYNC is for debugging. We skip the filesystem sync and use
831 * a normal transaction (which is less likely to stall). used for
832 * testing filesystem consistency.
834 * In normal mode we sync the filesystem and use a flush transaction.
836 if (pfs
->pfs_flags
& HAMMER2_PFSFLAGS_NOSYNC
) {
837 hammer2_trans_init(pmp
, 0);
839 hammer2_vfs_sync(pmp
->mp
, MNT_WAIT
);
840 hammer2_trans_init(pmp
, HAMMER2_TRANS_ISFLUSH
);
842 mtid
= hammer2_trans_sub(pmp
);
843 hammer2_inode_lock(ip
, 0);
844 hammer2_inode_modify(ip
);
845 ip
->meta
.pfs_lsnap_tid
= mtid
;
847 /* XXX cluster it! */
848 chain
= hammer2_inode_chain(ip
, 0, HAMMER2_RESOLVE_ALWAYS
);
850 name_len
= strlen(pfs
->name
);
851 lhc
= hammer2_dirhash(pfs
->name
, name_len
);
859 * Create the snapshot directory under the super-root
861 * Set PFS type, generate a unique filesystem id, and generate
862 * a cluster id. Use the same clid when snapshotting a PFS root,
863 * which theoretically allows the snapshot to be used as part of
864 * the same cluster (perhaps as a cache).
866 * Note that pfs_lsnap_tid must be set in the snapshot as well,
867 * ensuring that any nocrc/nocomp file data modifications force
870 * Copy the (flushed) blockref array. Theoretically we could use
871 * chain_duplicate() but it becomes difficult to disentangle
872 * the shared core so for now just brute-force it.
874 hammer2_chain_unlock(chain
);
875 nip
= hammer2_inode_create_pfs(hmp
->spmp
, pfs
->name
, name_len
, &error
);
876 hammer2_chain_lock(chain
, HAMMER2_RESOLVE_ALWAYS
);
879 hammer2_dev_t
*force_local
;
880 hammer2_chain_t
*nchain
;
881 hammer2_inode_data_t
*wipdata
;
882 hammer2_tid_t starting_inum
;
884 atomic_set_int(&nip
->flags
, HAMMER2_INODE_NOSIDEQ
);
885 hammer2_inode_modify(nip
);
886 nchain
= hammer2_inode_chain(nip
, 0, HAMMER2_RESOLVE_ALWAYS
);
887 error
= hammer2_chain_modify(nchain
, mtid
, 0, 0);
888 KKASSERT(error
== 0);
889 wipdata
= &nchain
->data
->ipdata
;
891 starting_inum
= ip
->pmp
->inode_tid
+ 1;
892 nip
->meta
.pfs_inum
= starting_inum
;
893 nip
->meta
.pfs_type
= HAMMER2_PFSTYPE_MASTER
;
894 nip
->meta
.pfs_subtype
= HAMMER2_PFSSUBTYPE_SNAPSHOT
;
895 nip
->meta
.op_flags
|= HAMMER2_OPFLAG_PFSROOT
;
896 nip
->meta
.pfs_lsnap_tid
= mtid
;
897 nchain
->bref
.embed
.stats
= chain
->bref
.embed
.stats
;
899 kern_uuidgen(&nip
->meta
.pfs_fsid
, 1);
903 * Give the snapshot its own private cluster id. As a
904 * snapshot no further synchronization with the original
905 * cluster will be done.
907 if (chain
->flags
& HAMMER2_CHAIN_PFSBOUNDARY
)
908 nip
->meta
.pfs_clid
= opfs_clid
;
910 kern_uuidgen(&nip
->meta
.pfs_clid
, 1);
912 kern_uuidgen(&nip
->meta
.pfs_clid
, 1);
913 nchain
->bref
.flags
|= HAMMER2_BREF_FLAG_PFSROOT
;
915 /* XXX hack blockset copy */
916 /* XXX doesn't work with real cluster */
917 wipdata
->meta
= nip
->meta
;
918 hammer2_spin_ex(&pmp
->inum_spin
);
919 wipdata
->u
.blockset
= pmp
->pfs_iroot_blocksets
[0];
920 hammer2_spin_unex(&pmp
->inum_spin
);
922 KKASSERT(wipdata
== &nchain
->data
->ipdata
);
924 hammer2_chain_unlock(nchain
);
925 hammer2_inode_ref(nip
);
926 hammer2_inode_unlock(nip
);
927 hammer2_inode_chain_sync(nip
);
928 hammer2_inode_chain_flush(nip
, HAMMER2_XOP_INODE_STOP
|
930 /* XXX | HAMMER2_XOP_VOLHDR */
931 hammer2_inode_drop(nip
);
934 force_local
= (hmp
->hflags
& HMNT2_LOCAL
) ? hmp
: NULL
;
936 hammer2_chain_lock(nchain
, HAMMER2_RESOLVE_ALWAYS
);
937 wipdata
= &nchain
->data
->ipdata
;
938 kprintf("SNAPSHOT LOCAL PFS (IOCTL): %s\n", wipdata
->filename
);
939 hammer2_pfsalloc(nchain
, wipdata
, force_local
);
940 nchain
->pmp
->inode_tid
= starting_inum
;
942 hammer2_chain_unlock(nchain
);
943 hammer2_chain_drop(nchain
);
946 hammer2_chain_unlock(chain
);
947 hammer2_chain_drop(chain
);
949 hammer2_inode_unlock(ip
);
950 if (pfs
->pfs_flags
& HAMMER2_PFSFLAGS_NOSYNC
) {
951 hammer2_trans_done(pmp
, 0);
953 hammer2_trans_done(pmp
, HAMMER2_TRANS_ISFLUSH
|
954 HAMMER2_TRANS_SIDEQ
);
957 lockmgr(&hmp
->bulklk
, LK_RELEASE
);
959 return (hammer2_error_to_errno(error
));
963 * Retrieve the raw inode structure, non-inclusive of node-specific data.
966 hammer2_ioctl_inode_get(hammer2_inode_t
*ip
, void *data
)
968 hammer2_ioc_inode_t
*ino
= data
;
970 hammer2_inode_lock(ip
, HAMMER2_RESOLVE_SHARED
);
971 ino
->data_count
= hammer2_inode_data_count(ip
);
972 ino
->inode_count
= hammer2_inode_inode_count(ip
);
974 bzero(&ino
->ip_data
, sizeof(ino
->ip_data
));
975 ino
->ip_data
.meta
= ip
->meta
;
976 hammer2_inode_unlock(ip
);
982 * Set various parameters in an inode which cannot be set through
983 * normal filesystem VNOPS.
986 hammer2_ioctl_inode_set(hammer2_inode_t
*ip
, void *data
)
988 hammer2_ioc_inode_t
*ino
= data
;
990 hammer2_trans_init(ip
->pmp
, 0);
991 hammer2_inode_lock(ip
, 0);
993 if ((ino
->flags
& HAMMER2IOC_INODE_FLAG_CHECK
) &&
994 ip
->meta
.check_algo
!= ino
->ip_data
.meta
.check_algo
) {
995 hammer2_inode_modify(ip
);
996 ip
->meta
.check_algo
= ino
->ip_data
.meta
.check_algo
;
998 if ((ino
->flags
& HAMMER2IOC_INODE_FLAG_COMP
) &&
999 ip
->meta
.comp_algo
!= ino
->ip_data
.meta
.comp_algo
) {
1000 hammer2_inode_modify(ip
);
1001 ip
->meta
.comp_algo
= ino
->ip_data
.meta
.comp_algo
;
1004 /* Ignore these flags for now...*/
1005 if ((ino
->flags
& HAMMER2IOC_INODE_FLAG_IQUOTA
) &&
1006 ip
->meta
.inode_quota
!= ino
->ip_data
.meta
.inode_quota
) {
1007 hammer2_inode_modify(ip
);
1008 ip
->meta
.inode_quota
= ino
->ip_data
.meta
.inode_quota
;
1010 if ((ino
->flags
& HAMMER2IOC_INODE_FLAG_DQUOTA
) &&
1011 ip
->meta
.data_quota
!= ino
->ip_data
.meta
.data_quota
) {
1012 hammer2_inode_modify(ip
);
1013 ip
->meta
.data_quota
= ino
->ip_data
.meta
.data_quota
;
1015 if ((ino
->flags
& HAMMER2IOC_INODE_FLAG_COPIES
) &&
1016 ip
->meta
.ncopies
!= ino
->ip_data
.meta
.ncopies
) {
1017 hammer2_inode_modify(ip
);
1018 ip
->meta
.ncopies
= ino
->ip_data
.meta
.ncopies
;
1020 hammer2_inode_unlock(ip
);
1021 hammer2_trans_done(ip
->pmp
, HAMMER2_TRANS_SIDEQ
);
1028 hammer2_ioctl_debug_dump(hammer2_inode_t
*ip
, u_int flags
)
1030 hammer2_chain_t
*chain
;
1034 for (i
= 0; i
< ip
->cluster
.nchains
; ++i
) {
1035 chain
= ip
->cluster
.array
[i
].chain
;
1038 kprintf("cluster #%d\n", i
);
1039 hammer2_dump_chain(chain
, 0, 0, &count
, 'i', flags
);
1045 * Turn on or off emergency mode on a filesystem.
1049 hammer2_ioctl_emerg_mode(hammer2_inode_t
*ip
, u_int mode
)
1057 kprintf("hammer2: WARNING: Emergency mode enabled\n");
1058 atomic_set_int(&pmp
->flags
, HAMMER2_PMPF_EMERG
);
1060 kprintf("hammer2: WARNING: Emergency mode disabled\n");
1061 atomic_clear_int(&pmp
->flags
, HAMMER2_PMPF_EMERG
);
1063 for (i
= 0; i
< HAMMER2_MAXCLUSTER
; ++i
) {
1064 hmp
= pmp
->pfs_hmps
[i
];
1068 atomic_set_int(&hmp
->hflags
, HMNT2_EMERG
);
1070 atomic_clear_int(&hmp
->hflags
, HMNT2_EMERG
);
1077 * Do a bulkfree scan on media related to the PFS. This routine will
1078 * flush all PFSs associated with the media before doing the bulkfree
1081 * This version can only run on non-clustered media. A new ioctl or a
1082 * temporary mount of @LOCAL will be needed to run on clustered media.
1085 hammer2_ioctl_bulkfree_scan(hammer2_inode_t
*ip
, void *data
)
1087 hammer2_ioc_bulkfree_t
*bfi
= data
;
1090 hammer2_chain_t
*vchain
;
1097 hmp
= pmp
->pfs_hmps
[0];
1104 * Bulkfree has to be serialized to guarantee at least one sync
1105 * inbetween bulkfrees.
1107 error
= lockmgr(&hmp
->bflock
, LK_EXCLUSIVE
/*| LK_PCATCH*/);
1112 * Sync all mounts related to the media
1114 lockmgr(&hammer2_mntlk
, LK_EXCLUSIVE
);
1115 TAILQ_FOREACH(pmp
, &hammer2_pfslist
, mntentry
) {
1119 for (i
= 0; i
< HAMMER2_MAXCLUSTER
; ++i
) {
1120 if (pmp
->pfs_hmps
[i
] != hmp
)
1122 etmp
= hammer2_vfs_sync_pmp(pmp
, MNT_WAIT
);
1123 if (etmp
&& (error
== 0 || error
== ENOSPC
))
1128 lockmgr(&hammer2_mntlk
, LK_RELEASE
);
1130 if (error
&& error
!= ENOSPC
)
1134 * If we have an ENOSPC error we have to bulkfree on the live
1135 * topology. Otherwise we can bulkfree on a snapshot.
1138 kprintf("hammer2: WARNING! Bulkfree forced to use live "
1139 "topology due to ENOSPC\n");
1140 vchain
= &hmp
->vchain
;
1141 hammer2_chain_ref(vchain
);
1144 vchain
= hammer2_chain_bulksnap(hmp
);
1149 * Normal bulkfree operations do not require a transaction because
1150 * they operate on a snapshot, and so can run concurrently with
1151 * any operation except another bulkfree.
1153 * If we are running bulkfree on the live topology we have to be
1154 * in a FLUSH transaction.
1157 hammer2_trans_init(hmp
->spmp
, HAMMER2_TRANS_ISFLUSH
);
1160 hammer2_thr_freeze(&hmp
->bfthr
);
1161 error
= hammer2_bulkfree_pass(hmp
, vchain
, bfi
);
1162 hammer2_thr_unfreeze(&hmp
->bfthr
);
1165 hammer2_chain_bulkdrop(vchain
);
1167 hammer2_chain_drop(vchain
);
1168 hammer2_trans_done(hmp
->spmp
, HAMMER2_TRANS_ISFLUSH
|
1169 HAMMER2_TRANS_SIDEQ
);
1171 error
= hammer2_error_to_errno(error
);
1174 lockmgr(&hmp
->bflock
, LK_RELEASE
);
1180 * Unconditionally delete meta-data in a hammer2 filesystem
1184 hammer2_ioctl_destroy(hammer2_inode_t
*ip
, void *data
)
1186 hammer2_ioc_destroy_t
*iocd
= data
;
1187 hammer2_pfs_t
*pmp
= ip
->pmp
;
1196 case HAMMER2_DELETE_FILE
:
1198 * Destroy a bad directory entry by name. Caller must
1199 * pass the directory as fd.
1202 hammer2_xop_unlink_t
*xop
;
1204 if (iocd
->path
[sizeof(iocd
->path
)-1]) {
1208 if (ip
->meta
.type
!= HAMMER2_OBJTYPE_DIRECTORY
) {
1212 hammer2_pfs_memory_wait(pmp
);
1213 hammer2_trans_init(pmp
, 0);
1214 hammer2_inode_lock(ip
, 0);
1216 xop
= hammer2_xop_alloc(ip
, HAMMER2_XOP_MODIFYING
);
1217 hammer2_xop_setname(&xop
->head
, iocd
->path
, strlen(iocd
->path
));
1219 xop
->dopermanent
= H2DOPERM_PERMANENT
|
1222 hammer2_xop_start(&xop
->head
, &hammer2_unlink_desc
);
1224 error
= hammer2_xop_collect(&xop
->head
, 0);
1225 error
= hammer2_error_to_errno(error
);
1226 hammer2_inode_unlock(ip
);
1227 hammer2_xop_retire(&xop
->head
, HAMMER2_XOPMASK_VOP
);
1228 hammer2_trans_done(pmp
, HAMMER2_TRANS_SIDEQ
);
1231 case HAMMER2_DELETE_INUM
:
1233 * Destroy a bad inode by inode number.
1236 hammer2_xop_lookup_t
*xop
;
1238 if (iocd
->inum
< 1) {
1242 hammer2_pfs_memory_wait(pmp
);
1243 hammer2_trans_init(pmp
, 0);
1245 xop
= hammer2_xop_alloc(pmp
->iroot
, HAMMER2_XOP_MODIFYING
);
1246 xop
->lhc
= iocd
->inum
;
1247 hammer2_xop_start(&xop
->head
, &hammer2_delete_desc
);
1248 error
= hammer2_xop_collect(&xop
->head
, 0);
1249 error
= hammer2_error_to_errno(error
);
1250 hammer2_xop_retire(&xop
->head
, HAMMER2_XOPMASK_VOP
);
1251 hammer2_trans_done(pmp
, HAMMER2_TRANS_SIDEQ
);
1262 * Grow a filesystem into its partition size
1265 hammer2_ioctl_growfs(hammer2_inode_t
*ip
, void *data
, struct ucred
*cred
)
1267 hammer2_ioc_growfs_t
*grow
= data
;
1269 hammer2_off_t size
, delta
;
1271 struct partinfo part
;
1272 struct vattr_lite va
;
1277 hmp
= ip
->pmp
->pfs_hmps
[0];
1279 if (hmp
->nvolumes
> 1) {
1280 kprintf("hammer2: growfs currently unsupported "
1281 "with multiple volumes\n");
1284 KKASSERT(hmp
->total_size
== hmp
->voldata
.volu_size
);
1287 * Extract from disklabel
1289 if (VOP_IOCTL(hmp
->devvp
, DIOCGPART
, (void *)&part
, 0, cred
, NULL
) == 0) {
1290 size
= part
.media_size
;
1291 kprintf("hammer2: growfs partition-auto to %016jx\n",
1293 } else if (VOP_GETATTR_LITE(hmp
->devvp
, &va
) == 0) {
1295 kprintf("hammer2: growfs fstat-auto to %016jx\n",
1302 * Expand to devvp size unless specified.
1305 if (grow
->size
== 0) {
1307 } else if (grow
->size
> size
) {
1308 kprintf("hammer2: growfs size %016jx exceeds device size "
1310 (intmax_t)grow
->size
, (intmax_t)size
);
1315 * This is typically ~8MB alignment to avoid edge cases accessing
1316 * reserved blocks at the base of each 2GB zone.
1318 grow
->size
&= ~HAMMER2_VOLUME_ALIGNMASK64
;
1319 delta
= grow
->size
- hmp
->voldata
.volu_size
;
1322 * Maximum allowed size is 2^63
1324 if (grow
->size
> 0x7FFFFFFFFFFFFFFFLU
) {
1325 kprintf("hammer2: growfs failure, limit is 2^63 - 1 bytes\n");
1330 * We can't shrink a filesystem
1332 if (grow
->size
< hmp
->voldata
.volu_size
) {
1333 kprintf("hammer2: growfs failure, "
1334 "would shrink from %016jx to %016jx\n",
1335 (intmax_t)hmp
->voldata
.volu_size
,
1336 (intmax_t)grow
->size
);
1341 kprintf("hammer2: growfs - size did not change\n");
1346 * Clear any new volume header backups that we extend into.
1347 * Skip volume headers that are already part of the filesystem.
1349 for (i
= 0; i
< HAMMER2_NUM_VOLHDRS
; ++i
) {
1350 if (i
* HAMMER2_ZONE_BYTES64
< hmp
->voldata
.volu_size
)
1352 if (i
* HAMMER2_ZONE_BYTES64
>= grow
->size
)
1354 kprintf("hammer2: growfs - clear volhdr %d ", i
);
1355 error
= bread(hmp
->devvp
, i
* HAMMER2_ZONE_BYTES64
,
1356 HAMMER2_VOLUME_BYTES
, &bp
);
1359 kprintf("I/O error %d\n", error
);
1362 bzero(bp
->b_data
, HAMMER2_VOLUME_BYTES
);
1365 kprintf("I/O error %d\n", error
);
1371 hammer2_trans_init(hmp
->spmp
, HAMMER2_TRANS_ISFLUSH
);
1372 mtid
= hammer2_trans_sub(hmp
->spmp
);
1374 kprintf("hammer2: growfs - expand by %016jx to %016jx mtid %016jx\n",
1375 (intmax_t)delta
, (intmax_t)grow
->size
, (intmax_t)mtid
);
1378 hammer2_voldata_lock(hmp
);
1379 hammer2_voldata_modify(hmp
);
1382 * NOTE: Just adjusting total_size for a single-volume filesystem
1383 * or for the last volume in a multi-volume filesystem, is
1384 * fine. But we can't grow any other partition in a multi-volume
1385 * filesystem. For now we just punt (at the top) on any
1386 * multi-volume filesystem.
1388 hmp
->voldata
.volu_size
= grow
->size
;
1389 hmp
->voldata
.total_size
+= delta
;
1390 hmp
->voldata
.allocator_size
+= delta
;
1391 hmp
->voldata
.allocator_free
+= delta
;
1392 hmp
->total_size
+= delta
;
1393 hmp
->volumes
[0].size
+= delta
; /* note: indexes first (only) volume */
1395 hammer2_voldata_unlock(hmp
);
1397 hammer2_trans_done(hmp
->spmp
, HAMMER2_TRANS_ISFLUSH
|
1398 HAMMER2_TRANS_SIDEQ
);
1402 * Flush the mess right here and now. We could just let the
1403 * filesystem syncer do it, but this was a sensitive operation
1404 * so don't take any chances.
1406 hammer2_vfs_sync(ip
->pmp
->mp
, MNT_WAIT
);
1412 * Get a list of volumes.
1415 hammer2_ioctl_volume_list(hammer2_inode_t
*ip
, void *data
)
1417 hammer2_ioc_volume_list_t
*vollist
= data
;
1418 hammer2_ioc_volume_t entry
;
1419 hammer2_volume_t
*vol
;
1422 int i
, error
= 0, cnt
= 0;
1425 hmp
= pmp
->pfs_hmps
[0];
1429 hammer2_voldata_lock(hmp
);
1430 for (i
= 0; i
< hmp
->nvolumes
; ++i
) {
1431 if (cnt
>= vollist
->nvolumes
)
1433 vol
= &hmp
->volumes
[i
];
1434 bzero(&entry
, sizeof(entry
));
1435 /* copy hammer2_volume_t fields */
1437 bcopy(vol
->dev
->path
, entry
.path
, sizeof(entry
.path
));
1438 entry
.offset
= vol
->offset
;
1439 entry
.size
= vol
->size
;
1440 error
= copyout(&entry
, &vollist
->volumes
[cnt
], sizeof(entry
));
1445 vollist
->nvolumes
= cnt
;
1446 vollist
->version
= hmp
->voldata
.version
;
1447 bcopy(pmp
->pfs_names
[0], vollist
->pfs_name
, sizeof(vollist
->pfs_name
));
1449 hammer2_voldata_unlock(hmp
);