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.
44 #include <sys/diskslice.h>
50 #include <sys/kern_syscall.h>
52 static int hammer2_ioctl_version_get(hammer2_inode_t
*ip
, void *data
);
53 static int hammer2_ioctl_recluster(hammer2_inode_t
*ip
, void *data
);
54 static int hammer2_ioctl_remote_scan(hammer2_inode_t
*ip
, void *data
);
55 static int hammer2_ioctl_remote_add(hammer2_inode_t
*ip
, void *data
);
56 static int hammer2_ioctl_remote_del(hammer2_inode_t
*ip
, void *data
);
57 static int hammer2_ioctl_remote_rep(hammer2_inode_t
*ip
, void *data
);
58 static int hammer2_ioctl_socket_get(hammer2_inode_t
*ip
, void *data
);
59 static int hammer2_ioctl_socket_set(hammer2_inode_t
*ip
, void *data
);
60 static int hammer2_ioctl_pfs_get(hammer2_inode_t
*ip
, void *data
);
61 static int hammer2_ioctl_pfs_lookup(hammer2_inode_t
*ip
, void *data
);
62 static int hammer2_ioctl_pfs_create(hammer2_inode_t
*ip
, void *data
);
63 static int hammer2_ioctl_pfs_snapshot(hammer2_inode_t
*ip
, void *data
);
64 static int hammer2_ioctl_pfs_delete(hammer2_inode_t
*ip
, void *data
);
65 static int hammer2_ioctl_inode_get(hammer2_inode_t
*ip
, void *data
);
66 static int hammer2_ioctl_inode_set(hammer2_inode_t
*ip
, void *data
);
67 static int hammer2_ioctl_debug_dump(hammer2_inode_t
*ip
, u_int flags
);
68 static int hammer2_ioctl_emerg_mode(hammer2_inode_t
*ip
, u_int mode
);
69 static int hammer2_ioctl_growfs(hammer2_inode_t
*ip
, void *data
,
71 //static int hammer2_ioctl_inode_comp_set(hammer2_inode_t *ip, void *data);
72 //static int hammer2_ioctl_inode_comp_rec_set(hammer2_inode_t *ip, void *data);
73 //static int hammer2_ioctl_inode_comp_rec_set2(hammer2_inode_t *ip, void *data);
74 static int hammer2_ioctl_bulkfree_scan(hammer2_inode_t
*ip
, void *data
);
75 static int hammer2_ioctl_destroy(hammer2_inode_t
*ip
, void *data
);
76 static int hammer2_ioctl_volume_list(hammer2_inode_t
*ip
, void *data
);
79 hammer2_ioctl(hammer2_inode_t
*ip
, u_long com
, void *data
, int fflag
,
85 * Standard root cred checks, will be selectively ignored below
86 * for ioctls that do not require root creds.
88 error
= caps_priv_check(cred
, SYSCAP_NOVFS_IOCTL
);
91 case HAMMER2IOC_VERSION_GET
:
92 error
= hammer2_ioctl_version_get(ip
, data
);
94 case HAMMER2IOC_RECLUSTER
:
96 error
= hammer2_ioctl_recluster(ip
, data
);
98 case HAMMER2IOC_REMOTE_SCAN
:
100 error
= hammer2_ioctl_remote_scan(ip
, data
);
102 case HAMMER2IOC_REMOTE_ADD
:
104 error
= hammer2_ioctl_remote_add(ip
, data
);
106 case HAMMER2IOC_REMOTE_DEL
:
108 error
= hammer2_ioctl_remote_del(ip
, data
);
110 case HAMMER2IOC_REMOTE_REP
:
112 error
= hammer2_ioctl_remote_rep(ip
, data
);
114 case HAMMER2IOC_SOCKET_GET
:
116 error
= hammer2_ioctl_socket_get(ip
, data
);
118 case HAMMER2IOC_SOCKET_SET
:
120 error
= hammer2_ioctl_socket_set(ip
, data
);
122 case HAMMER2IOC_PFS_GET
:
124 error
= hammer2_ioctl_pfs_get(ip
, data
);
126 case HAMMER2IOC_PFS_LOOKUP
:
128 error
= hammer2_ioctl_pfs_lookup(ip
, data
);
130 case HAMMER2IOC_PFS_CREATE
:
132 error
= hammer2_ioctl_pfs_create(ip
, data
);
134 case HAMMER2IOC_PFS_DELETE
:
136 error
= hammer2_ioctl_pfs_delete(ip
, data
);
138 case HAMMER2IOC_PFS_SNAPSHOT
:
140 error
= hammer2_ioctl_pfs_snapshot(ip
, data
);
142 case HAMMER2IOC_INODE_GET
:
143 error
= hammer2_ioctl_inode_get(ip
, data
);
145 case HAMMER2IOC_INODE_SET
:
147 error
= hammer2_ioctl_inode_set(ip
, data
);
149 case HAMMER2IOC_BULKFREE_SCAN
:
150 error
= hammer2_ioctl_bulkfree_scan(ip
, data
);
152 case HAMMER2IOC_BULKFREE_ASYNC
:
153 error
= hammer2_ioctl_bulkfree_scan(ip
, NULL
);
155 case HAMMER2IOC_DESTROY
:
157 error
= hammer2_ioctl_destroy(ip
, data
);
159 case HAMMER2IOC_DEBUG_DUMP
:
160 error
= hammer2_ioctl_debug_dump(ip
, *(u_int
*)data
);
162 case HAMMER2IOC_EMERG_MODE
:
164 error
= hammer2_ioctl_emerg_mode(ip
, *(u_int
*)data
);
166 case HAMMER2IOC_GROWFS
:
168 error
= hammer2_ioctl_growfs(ip
, data
, cred
);
170 case HAMMER2IOC_VOLUME_LIST
:
172 error
= hammer2_ioctl_volume_list(ip
, data
);
183 * Retrieve version and basic info
186 hammer2_ioctl_version_get(hammer2_inode_t
*ip
, void *data
)
188 hammer2_ioc_version_t
*version
= data
;
191 hmp
= ip
->pmp
->pfs_hmps
[0];
193 version
->version
= hmp
->voldata
.version
;
195 version
->version
= -1;
201 hammer2_ioctl_recluster(hammer2_inode_t
*ip
, void *data
)
203 hammer2_ioc_recluster_t
*recl
= data
;
204 struct vnode
*vproot
;
206 hammer2_cluster_t
*cluster
;
209 fp
= holdfp(curthread
, recl
->fd
, -1);
211 error
= VFS_ROOT(ip
->pmp
->mp
, &vproot
);
213 cluster
= &ip
->pmp
->iroot
->cluster
;
214 kprintf("reconnect to cluster: nc=%d focus=%p\n",
215 cluster
->nchains
, cluster
->focus
);
216 if (cluster
->nchains
!= 1 || cluster
->focus
== NULL
) {
217 kprintf("not a local device mount\n");
220 hammer2_cluster_reconnect(cluster
->focus
->hmp
,
234 * Retrieve information about a remote
237 hammer2_ioctl_remote_scan(hammer2_inode_t
*ip
, void *data
)
240 hammer2_ioc_remote_t
*remote
= data
;
241 int copyid
= remote
->copyid
;
243 hmp
= ip
->pmp
->pfs_hmps
[0];
247 if (copyid
< 0 || copyid
>= HAMMER2_COPYID_COUNT
)
250 hammer2_voldata_lock(hmp
);
251 remote
->copy1
= hmp
->voldata
.copyinfo
[copyid
];
252 hammer2_voldata_unlock(hmp
);
255 * Adjust nextid (GET only)
257 while (++copyid
< HAMMER2_COPYID_COUNT
&&
258 hmp
->voldata
.copyinfo
[copyid
].copyid
== 0) {
261 if (copyid
== HAMMER2_COPYID_COUNT
)
264 remote
->nextid
= copyid
;
270 * Add new remote entry
273 hammer2_ioctl_remote_add(hammer2_inode_t
*ip
, void *data
)
275 hammer2_ioc_remote_t
*remote
= data
;
276 hammer2_pfs_t
*pmp
= ip
->pmp
;
278 int copyid
= remote
->copyid
;
281 hmp
= pmp
->pfs_hmps
[0];
284 if (copyid
>= HAMMER2_COPYID_COUNT
)
287 hammer2_voldata_lock(hmp
);
289 for (copyid
= 1; copyid
< HAMMER2_COPYID_COUNT
; ++copyid
) {
290 if (hmp
->voldata
.copyinfo
[copyid
].copyid
== 0)
293 if (copyid
== HAMMER2_COPYID_COUNT
) {
298 hammer2_voldata_modify(hmp
);
299 remote
->copy1
.copyid
= copyid
;
300 hmp
->voldata
.copyinfo
[copyid
] = remote
->copy1
;
301 hammer2_volconf_update(hmp
, copyid
);
303 hammer2_voldata_unlock(hmp
);
308 * Delete existing remote entry
311 hammer2_ioctl_remote_del(hammer2_inode_t
*ip
, void *data
)
313 hammer2_ioc_remote_t
*remote
= data
;
314 hammer2_pfs_t
*pmp
= ip
->pmp
;
316 int copyid
= remote
->copyid
;
319 hmp
= pmp
->pfs_hmps
[0];
322 if (copyid
>= HAMMER2_COPYID_COUNT
)
324 remote
->copy1
.path
[sizeof(remote
->copy1
.path
) - 1] = 0;
325 hammer2_voldata_lock(hmp
);
327 for (copyid
= 1; copyid
< HAMMER2_COPYID_COUNT
; ++copyid
) {
328 if (hmp
->voldata
.copyinfo
[copyid
].copyid
== 0)
330 if (strcmp(remote
->copy1
.path
,
331 hmp
->voldata
.copyinfo
[copyid
].path
) == 0) {
335 if (copyid
== HAMMER2_COPYID_COUNT
) {
340 hammer2_voldata_modify(hmp
);
341 hmp
->voldata
.copyinfo
[copyid
].copyid
= 0;
342 hammer2_volconf_update(hmp
, copyid
);
344 hammer2_voldata_unlock(hmp
);
349 * Replace existing remote entry
352 hammer2_ioctl_remote_rep(hammer2_inode_t
*ip
, void *data
)
354 hammer2_ioc_remote_t
*remote
= data
;
356 int copyid
= remote
->copyid
;
358 hmp
= ip
->pmp
->pfs_hmps
[0];
361 if (copyid
< 0 || copyid
>= HAMMER2_COPYID_COUNT
)
364 hammer2_voldata_lock(hmp
);
365 hammer2_voldata_modify(hmp
);
366 /*hammer2_volconf_update(hmp, copyid);*/
367 hammer2_voldata_unlock(hmp
);
373 * Retrieve communications socket
376 hammer2_ioctl_socket_get(hammer2_inode_t
*ip
, void *data
)
382 * Set communications socket for connection
385 hammer2_ioctl_socket_set(hammer2_inode_t
*ip
, void *data
)
387 hammer2_ioc_remote_t
*remote
= data
;
389 int copyid
= remote
->copyid
;
391 hmp
= ip
->pmp
->pfs_hmps
[0];
394 if (copyid
< 0 || copyid
>= HAMMER2_COPYID_COUNT
)
397 hammer2_voldata_lock(hmp
);
398 hammer2_voldata_unlock(hmp
);
405 * Used to scan and retrieve PFS information. PFS's are directories under
408 * To scan PFSs pass name_key=0. The function will scan for the next
409 * PFS and set all fields, as well as set name_next to the next key.
410 * When no PFSs remain, name_next is set to (hammer2_key_t)-1.
412 * To retrieve a particular PFS by key, specify the key but note that
413 * the ioctl will return the lowest key >= specified_key, so the caller
414 * must verify the key.
416 * To retrieve the PFS associated with the file descriptor, pass
417 * name_key set to (hammer2_key_t)-1.
420 hammer2_ioctl_pfs_get(hammer2_inode_t
*ip
, void *data
)
422 const hammer2_inode_data_t
*ripdata
;
424 hammer2_ioc_pfs_t
*pfs
;
425 hammer2_chain_t
*parent
;
426 hammer2_chain_t
*chain
;
427 hammer2_key_t key_next
;
428 hammer2_key_t save_key
;
431 hmp
= ip
->pmp
->pfs_hmps
[0];
436 save_key
= pfs
->name_key
;
442 if (save_key
== (hammer2_key_t
)-1) {
443 hammer2_inode_lock(ip
->pmp
->iroot
, 0);
445 chain
= hammer2_inode_chain(ip
->pmp
->iroot
, 0,
446 HAMMER2_RESOLVE_ALWAYS
|
447 HAMMER2_RESOLVE_SHARED
);
449 hammer2_inode_lock(hmp
->spmp
->iroot
, 0);
450 parent
= hammer2_inode_chain(hmp
->spmp
->iroot
, 0,
451 HAMMER2_RESOLVE_ALWAYS
|
452 HAMMER2_RESOLVE_SHARED
);
453 chain
= hammer2_chain_lookup(&parent
, &key_next
,
454 pfs
->name_key
, HAMMER2_KEY_MAX
,
456 HAMMER2_LOOKUP_SHARED
);
463 if (chain
->bref
.type
== HAMMER2_BREF_TYPE_INODE
)
465 if (parent
== NULL
) {
466 hammer2_chain_unlock(chain
);
467 hammer2_chain_drop(chain
);
471 chain
= hammer2_chain_next(&parent
, chain
, &key_next
,
472 key_next
, HAMMER2_KEY_MAX
,
474 HAMMER2_LOOKUP_SHARED
);
476 error
= hammer2_error_to_errno(error
);
479 * Load the data being returned by the ioctl.
481 if (chain
&& chain
->error
== 0) {
482 ripdata
= &chain
->data
->ipdata
;
483 pfs
->name_key
= ripdata
->meta
.name_key
;
484 pfs
->pfs_type
= ripdata
->meta
.pfs_type
;
485 pfs
->pfs_subtype
= ripdata
->meta
.pfs_subtype
;
486 pfs
->pfs_clid
= ripdata
->meta
.pfs_clid
;
487 pfs
->pfs_fsid
= ripdata
->meta
.pfs_fsid
;
488 KKASSERT(ripdata
->meta
.name_len
< sizeof(pfs
->name
));
489 bcopy(ripdata
->filename
, pfs
->name
, ripdata
->meta
.name_len
);
490 pfs
->name
[ripdata
->meta
.name_len
] = 0;
491 ripdata
= NULL
; /* safety */
494 * Calculate name_next, if any. We are only accessing
495 * chain->bref so we can ignore chain->error (if the key
496 * is used later it will error then).
498 if (parent
== NULL
) {
499 pfs
->name_next
= (hammer2_key_t
)-1;
501 chain
= hammer2_chain_next(&parent
, chain
, &key_next
,
502 key_next
, HAMMER2_KEY_MAX
,
504 HAMMER2_LOOKUP_SHARED
);
506 pfs
->name_next
= chain
->bref
.key
;
508 pfs
->name_next
= (hammer2_key_t
)-1;
511 pfs
->name_next
= (hammer2_key_t
)-1;
519 hammer2_chain_unlock(chain
);
520 hammer2_chain_drop(chain
);
523 hammer2_chain_unlock(parent
);
524 hammer2_chain_drop(parent
);
526 if (save_key
== (hammer2_key_t
)-1) {
527 hammer2_inode_unlock(ip
->pmp
->iroot
);
529 hammer2_inode_unlock(hmp
->spmp
->iroot
);
536 * Find a specific PFS by name
539 hammer2_ioctl_pfs_lookup(hammer2_inode_t
*ip
, void *data
)
541 const hammer2_inode_data_t
*ripdata
;
543 hammer2_ioc_pfs_t
*pfs
;
544 hammer2_chain_t
*parent
;
545 hammer2_chain_t
*chain
;
546 hammer2_key_t key_next
;
551 hmp
= ip
->pmp
->pfs_hmps
[0];
558 hammer2_inode_lock(hmp
->spmp
->iroot
, HAMMER2_RESOLVE_SHARED
);
559 parent
= hammer2_inode_chain(hmp
->spmp
->iroot
, 0,
560 HAMMER2_RESOLVE_ALWAYS
|
561 HAMMER2_RESOLVE_SHARED
);
563 pfs
->name
[sizeof(pfs
->name
) - 1] = 0;
564 len
= strlen(pfs
->name
);
565 lhc
= hammer2_dirhash(pfs
->name
, len
);
567 chain
= hammer2_chain_lookup(&parent
, &key_next
,
568 lhc
, lhc
+ HAMMER2_DIRHASH_LOMASK
,
569 &error
, HAMMER2_LOOKUP_SHARED
);
571 if (hammer2_chain_dirent_test(chain
, pfs
->name
, len
))
573 chain
= hammer2_chain_next(&parent
, chain
, &key_next
,
575 lhc
+ HAMMER2_DIRHASH_LOMASK
,
576 &error
, HAMMER2_LOOKUP_SHARED
);
578 error
= hammer2_error_to_errno(error
);
581 * Load the data being returned by the ioctl.
583 if (chain
&& chain
->error
== 0) {
584 KKASSERT(chain
->bref
.type
== HAMMER2_BREF_TYPE_INODE
);
585 ripdata
= &chain
->data
->ipdata
;
586 pfs
->name_key
= ripdata
->meta
.name_key
;
587 pfs
->pfs_type
= ripdata
->meta
.pfs_type
;
588 pfs
->pfs_subtype
= ripdata
->meta
.pfs_subtype
;
589 pfs
->pfs_clid
= ripdata
->meta
.pfs_clid
;
590 pfs
->pfs_fsid
= ripdata
->meta
.pfs_fsid
;
593 hammer2_chain_unlock(chain
);
594 hammer2_chain_drop(chain
);
595 } else if (error
== 0) {
599 hammer2_chain_unlock(parent
);
600 hammer2_chain_drop(parent
);
602 hammer2_inode_unlock(hmp
->spmp
->iroot
);
608 * Create a new PFS under the super-root
611 hammer2_ioctl_pfs_create(hammer2_inode_t
*ip
, void *data
)
613 hammer2_inode_data_t
*nipdata
;
614 hammer2_chain_t
*nchain
;
616 hammer2_dev_t
*force_local
;
617 hammer2_ioc_pfs_t
*pfs
;
618 hammer2_inode_t
*nip
;
622 hmp
= ip
->pmp
->pfs_hmps
[0]; /* XXX */
629 if (pfs
->name
[0] == 0)
631 pfs
->name
[sizeof(pfs
->name
) - 1] = 0; /* ensure 0-termination */
633 if (hammer2_ioctl_pfs_lookup(ip
, pfs
) == 0)
636 hammer2_trans_init(hmp
->spmp
, HAMMER2_TRANS_ISFLUSH
);
637 mtid
= hammer2_trans_sub(hmp
->spmp
);
638 nip
= hammer2_inode_create_pfs(hmp
->spmp
, pfs
->name
, strlen(pfs
->name
),
641 atomic_set_int(&nip
->flags
, HAMMER2_INODE_NOSIDEQ
);
642 hammer2_inode_modify(nip
);
643 nchain
= hammer2_inode_chain(nip
, 0, HAMMER2_RESOLVE_ALWAYS
);
644 error
= hammer2_chain_modify(nchain
, mtid
, 0, 0);
645 KKASSERT(error
== 0);
646 nipdata
= &nchain
->data
->ipdata
;
648 nip
->meta
.pfs_type
= pfs
->pfs_type
;
649 nip
->meta
.pfs_subtype
= pfs
->pfs_subtype
;
650 nip
->meta
.pfs_clid
= pfs
->pfs_clid
;
651 nip
->meta
.pfs_fsid
= pfs
->pfs_fsid
;
652 nip
->meta
.op_flags
|= HAMMER2_OPFLAG_PFSROOT
;
655 * Set default compression and check algorithm. This
656 * can be changed later.
658 * Do not allow compression on PFS's with the special name
659 * "boot", the boot loader can't decompress (yet).
661 nip
->meta
.comp_algo
=
662 HAMMER2_ENC_ALGO(HAMMER2_COMP_DEFAULT
);
663 nip
->meta
.check_algo
=
664 HAMMER2_ENC_ALGO(HAMMER2_CHECK_DEFAULT
);
666 if (strcasecmp(pfs
->name
, "boot") == 0) {
667 nip
->meta
.comp_algo
=
668 HAMMER2_ENC_ALGO(HAMMER2_COMP_AUTOZERO
);
672 * Super-root isn't mounted, fsync it
674 hammer2_chain_unlock(nchain
);
675 hammer2_inode_ref(nip
);
676 hammer2_inode_unlock(nip
);
677 hammer2_inode_chain_sync(nip
);
678 hammer2_inode_chain_flush(nip
, HAMMER2_XOP_INODE_STOP
|
680 hammer2_inode_drop(nip
);
684 * We still have a ref on the chain, relock and associate
685 * with an appropriate PFS.
687 force_local
= (hmp
->hflags
& HMNT2_LOCAL
) ? hmp
: NULL
;
689 hammer2_chain_lock(nchain
, HAMMER2_RESOLVE_ALWAYS
);
690 nipdata
= &nchain
->data
->ipdata
;
691 kprintf("ADD LOCAL PFS (IOCTL): %s\n", nipdata
->filename
);
692 hammer2_pfsalloc(nchain
, nipdata
, force_local
);
694 hammer2_chain_unlock(nchain
);
695 hammer2_chain_drop(nchain
);
697 hammer2_trans_done(hmp
->spmp
, HAMMER2_TRANS_ISFLUSH
|
698 HAMMER2_TRANS_SIDEQ
);
704 * Destroy an existing PFS under the super-root
707 hammer2_ioctl_pfs_delete(hammer2_inode_t
*ip
, void *data
)
709 hammer2_ioc_pfs_t
*pfs
= data
;
713 hammer2_xop_unlink_t
*xop
;
714 hammer2_inode_t
*dip
;
719 * The PFS should be probed, so we should be able to
720 * locate it. We only delete the PFS from the
721 * specific H2 block device (hmp), not all of
722 * them. We must remove the PFS from the cluster
723 * before we can destroy it.
725 hmp
= ip
->pmp
->pfs_hmps
[0];
729 pfs
->name
[sizeof(pfs
->name
) - 1] = 0; /* ensure termination */
731 lockmgr(&hammer2_mntlk
, LK_EXCLUSIVE
);
733 TAILQ_FOREACH(pmp
, &hammer2_pfslist
, mntentry
) {
734 for (i
= 0; i
< HAMMER2_MAXCLUSTER
; ++i
) {
735 if (pmp
->pfs_hmps
[i
] != hmp
)
737 if (pmp
->pfs_names
[i
] &&
738 strcmp(pmp
->pfs_names
[i
], pfs
->name
) == 0) {
742 if (i
!= HAMMER2_MAXCLUSTER
)
747 lockmgr(&hammer2_mntlk
, LK_RELEASE
);
751 lockmgr(&hammer2_mntlk
, LK_RELEASE
);
756 * Ok, we found the pmp and we have the index. Permanently remove
757 * the PFS from the cluster
759 kprintf("FOUND PFS %s CLINDEX %d\n", pfs
->name
, i
);
760 hammer2_pfsdealloc(pmp
, i
, 1);
762 lockmgr(&hammer2_mntlk
, LK_RELEASE
);
765 * Now destroy the PFS under its device using the per-device
770 hammer2_trans_init(spmp
, 0);
771 hammer2_inode_lock(dip
, 0);
773 xop
= hammer2_xop_alloc(dip
, HAMMER2_XOP_MODIFYING
);
774 hammer2_xop_setname(&xop
->head
, pfs
->name
, strlen(pfs
->name
));
776 xop
->dopermanent
= H2DOPERM_PERMANENT
| H2DOPERM_FORCE
;
777 hammer2_xop_start(&xop
->head
, &hammer2_unlink_desc
);
779 error
= hammer2_xop_collect(&xop
->head
, 0);
781 hammer2_inode_unlock(dip
);
785 ip
= hammer2_inode_get(dip
->pmp
, &xop
->head
, -1, -1);
786 hammer2_xop_retire(&xop
->head
, HAMMER2_XOPMASK_VOP
);
788 hammer2_inode_unlink_finisher(ip
, NULL
);
789 hammer2_inode_unlock(ip
);
792 hammer2_xop_retire(&xop
->head
, HAMMER2_XOPMASK_VOP
);
795 hammer2_xop_retire(&xop
->head
, HAMMER2_XOPMASK_VOP
);
797 hammer2_trans_done(spmp
, HAMMER2_TRANS_SIDEQ
);
799 return (hammer2_error_to_errno(error
));
803 hammer2_ioctl_pfs_snapshot(hammer2_inode_t
*ip
, void *data
)
805 hammer2_ioc_pfs_t
*pfs
= data
;
808 hammer2_chain_t
*chain
;
809 hammer2_inode_t
*nip
;
817 if (pfs
->name
[0] == 0)
819 if (pfs
->name
[sizeof(pfs
->name
)-1] != 0)
825 hmp
= pmp
->pfs_hmps
[0];
829 lockmgr(&hmp
->bulklk
, LK_EXCLUSIVE
);
832 * NOSYNC is for debugging. We skip the filesystem sync and use
833 * a normal transaction (which is less likely to stall). used for
834 * testing filesystem consistency.
836 * In normal mode we sync the filesystem and use a flush transaction.
838 if (pfs
->pfs_flags
& HAMMER2_PFSFLAGS_NOSYNC
) {
839 hammer2_trans_init(pmp
, 0);
841 hammer2_vfs_sync(pmp
->mp
, MNT_WAIT
);
842 hammer2_trans_init(pmp
, HAMMER2_TRANS_ISFLUSH
);
844 mtid
= hammer2_trans_sub(pmp
);
845 hammer2_inode_lock(ip
, 0);
846 hammer2_inode_modify(ip
);
847 ip
->meta
.pfs_lsnap_tid
= mtid
;
849 /* XXX cluster it! */
850 chain
= hammer2_inode_chain(ip
, 0, HAMMER2_RESOLVE_ALWAYS
);
852 name_len
= strlen(pfs
->name
);
856 * Create the snapshot directory under the super-root
858 * Set PFS type, generate a unique filesystem id, and generate
859 * a cluster id. Use the same clid when snapshotting a PFS root,
860 * which theoretically allows the snapshot to be used as part of
861 * the same cluster (perhaps as a cache).
863 * Note that pfs_lsnap_tid must be set in the snapshot as well,
864 * ensuring that any nocrc/nocomp file data modifications force
867 * Copy the (flushed) blockref array. Theoretically we could use
868 * chain_duplicate() but it becomes difficult to disentangle
869 * the shared core so for now just brute-force it.
871 hammer2_chain_unlock(chain
);
872 nip
= hammer2_inode_create_pfs(hmp
->spmp
, pfs
->name
, name_len
, &error
);
873 hammer2_chain_lock(chain
, HAMMER2_RESOLVE_ALWAYS
);
876 hammer2_dev_t
*force_local
;
877 hammer2_chain_t
*nchain
;
878 hammer2_inode_data_t
*wipdata
;
879 hammer2_tid_t starting_inum
;
881 atomic_set_int(&nip
->flags
, HAMMER2_INODE_NOSIDEQ
);
882 hammer2_inode_modify(nip
);
883 nchain
= hammer2_inode_chain(nip
, 0, HAMMER2_RESOLVE_ALWAYS
);
884 error
= hammer2_chain_modify(nchain
, mtid
, 0, 0);
885 KKASSERT(error
== 0);
886 wipdata
= &nchain
->data
->ipdata
;
888 starting_inum
= ip
->pmp
->inode_tid
+ 1;
889 nip
->meta
.pfs_inum
= starting_inum
;
890 nip
->meta
.pfs_type
= HAMMER2_PFSTYPE_MASTER
;
891 nip
->meta
.pfs_subtype
= HAMMER2_PFSSUBTYPE_SNAPSHOT
;
892 nip
->meta
.op_flags
|= HAMMER2_OPFLAG_PFSROOT
;
893 nip
->meta
.pfs_lsnap_tid
= mtid
;
894 nchain
->bref
.embed
.stats
= chain
->bref
.embed
.stats
;
896 uuid_create(&nip
->meta
.pfs_fsid
, NULL
);
900 * Give the snapshot its own private cluster id. As a
901 * snapshot no further synchronization with the original
902 * cluster will be done.
904 if (chain
->flags
& HAMMER2_CHAIN_PFSBOUNDARY
)
905 nip
->meta
.pfs_clid
= opfs_clid
;
907 uuid_create(&nip
->meta
.pfs_clid
, NULL
);
909 uuid_create(&nip
->meta
.pfs_clid
, NULL
);
910 nchain
->bref
.flags
|= HAMMER2_BREF_FLAG_PFSROOT
;
912 /* XXX hack blockset copy */
913 /* XXX doesn't work with real cluster */
914 wipdata
->meta
= nip
->meta
;
915 hammer2_spin_ex(&pmp
->blockset_spin
);
916 wipdata
->u
.blockset
= pmp
->pfs_iroot_blocksets
[0];
917 hammer2_spin_unex(&pmp
->blockset_spin
);
919 KKASSERT(wipdata
== &nchain
->data
->ipdata
);
921 hammer2_chain_unlock(nchain
);
922 hammer2_inode_ref(nip
);
923 hammer2_inode_unlock(nip
);
924 hammer2_inode_chain_sync(nip
);
925 hammer2_inode_chain_flush(nip
, HAMMER2_XOP_INODE_STOP
|
927 /* XXX | HAMMER2_XOP_VOLHDR */
928 hammer2_inode_drop(nip
);
931 force_local
= (hmp
->hflags
& HMNT2_LOCAL
) ? hmp
: NULL
;
933 hammer2_chain_lock(nchain
, HAMMER2_RESOLVE_ALWAYS
);
934 wipdata
= &nchain
->data
->ipdata
;
935 kprintf("SNAPSHOT LOCAL PFS (IOCTL): %s\n", wipdata
->filename
);
936 hammer2_pfsalloc(nchain
, wipdata
, force_local
);
937 nchain
->pmp
->inode_tid
= starting_inum
;
939 hammer2_chain_unlock(nchain
);
940 hammer2_chain_drop(nchain
);
943 hammer2_chain_unlock(chain
);
944 hammer2_chain_drop(chain
);
946 hammer2_inode_unlock(ip
);
947 if (pfs
->pfs_flags
& HAMMER2_PFSFLAGS_NOSYNC
) {
948 hammer2_trans_done(pmp
, 0);
950 hammer2_trans_done(pmp
, HAMMER2_TRANS_ISFLUSH
|
951 HAMMER2_TRANS_SIDEQ
);
954 lockmgr(&hmp
->bulklk
, LK_RELEASE
);
956 return (hammer2_error_to_errno(error
));
960 * Retrieve the raw inode structure, non-inclusive of node-specific data.
963 hammer2_ioctl_inode_get(hammer2_inode_t
*ip
, void *data
)
965 hammer2_ioc_inode_t
*ino
= data
;
967 hammer2_inode_lock(ip
, HAMMER2_RESOLVE_SHARED
);
968 ino
->data_count
= hammer2_inode_data_count(ip
);
969 ino
->inode_count
= hammer2_inode_inode_count(ip
);
971 bzero(&ino
->ip_data
, sizeof(ino
->ip_data
));
972 ino
->ip_data
.meta
= ip
->meta
;
973 hammer2_inode_unlock(ip
);
979 * Set various parameters in an inode which cannot be set through
980 * normal filesystem VNOPS.
983 hammer2_ioctl_inode_set(hammer2_inode_t
*ip
, void *data
)
985 hammer2_ioc_inode_t
*ino
= data
;
987 hammer2_trans_init(ip
->pmp
, 0);
988 hammer2_inode_lock(ip
, 0);
990 if ((ino
->flags
& HAMMER2IOC_INODE_FLAG_CHECK
) &&
991 ip
->meta
.check_algo
!= ino
->ip_data
.meta
.check_algo
) {
992 hammer2_inode_modify(ip
);
993 ip
->meta
.check_algo
= ino
->ip_data
.meta
.check_algo
;
995 if ((ino
->flags
& HAMMER2IOC_INODE_FLAG_COMP
) &&
996 ip
->meta
.comp_algo
!= ino
->ip_data
.meta
.comp_algo
) {
997 hammer2_inode_modify(ip
);
998 ip
->meta
.comp_algo
= ino
->ip_data
.meta
.comp_algo
;
1001 /* Ignore these flags for now...*/
1002 if ((ino
->flags
& HAMMER2IOC_INODE_FLAG_IQUOTA
) &&
1003 ip
->meta
.inode_quota
!= ino
->ip_data
.meta
.inode_quota
) {
1004 hammer2_inode_modify(ip
);
1005 ip
->meta
.inode_quota
= ino
->ip_data
.meta
.inode_quota
;
1007 if ((ino
->flags
& HAMMER2IOC_INODE_FLAG_DQUOTA
) &&
1008 ip
->meta
.data_quota
!= ino
->ip_data
.meta
.data_quota
) {
1009 hammer2_inode_modify(ip
);
1010 ip
->meta
.data_quota
= ino
->ip_data
.meta
.data_quota
;
1012 if ((ino
->flags
& HAMMER2IOC_INODE_FLAG_COPIES
) &&
1013 ip
->meta
.ncopies
!= ino
->ip_data
.meta
.ncopies
) {
1014 hammer2_inode_modify(ip
);
1015 ip
->meta
.ncopies
= ino
->ip_data
.meta
.ncopies
;
1017 hammer2_inode_unlock(ip
);
1018 hammer2_trans_done(ip
->pmp
, HAMMER2_TRANS_SIDEQ
);
1026 hammer2_ioctl_debug_dump(hammer2_inode_t
*ip
, u_int flags
)
1028 hammer2_chain_t
*chain
;
1032 for (i
= 0; i
< ip
->cluster
.nchains
; ++i
) {
1033 chain
= ip
->cluster
.array
[i
].chain
;
1036 kprintf("cluster #%d\n", i
);
1037 hammer2_dump_chain(chain
, 0, 0, &count
, 'i', flags
);
1044 * Turn on or off emergency mode on a filesystem.
1047 hammer2_ioctl_emerg_mode(hammer2_inode_t
*ip
, u_int mode
)
1055 kprintf("hammer2: WARNING: Emergency mode enabled\n");
1056 atomic_set_int(&pmp
->flags
, HAMMER2_PMPF_EMERG
);
1058 kprintf("hammer2: WARNING: Emergency mode disabled\n");
1059 atomic_clear_int(&pmp
->flags
, HAMMER2_PMPF_EMERG
);
1061 for (i
= 0; i
< HAMMER2_MAXCLUSTER
; ++i
) {
1062 hmp
= pmp
->pfs_hmps
[i
];
1066 atomic_set_int(&hmp
->hflags
, HMNT2_EMERG
);
1068 atomic_clear_int(&hmp
->hflags
, HMNT2_EMERG
);
1074 * Do a bulkfree scan on media related to the PFS. This routine will
1075 * flush all PFSs associated with the media before doing the bulkfree
1078 * This version can only run on non-clustered media. A new ioctl or a
1079 * temporary mount of @LOCAL will be needed to run on clustered media.
1082 hammer2_ioctl_bulkfree_scan(hammer2_inode_t
*ip
, void *data
)
1084 hammer2_ioc_bulkfree_t
*bfi
= data
;
1087 hammer2_chain_t
*vchain
;
1094 hmp
= pmp
->pfs_hmps
[0];
1101 * Bulkfree has to be serialized to guarantee at least one sync
1102 * inbetween bulkfrees.
1104 error
= lockmgr(&hmp
->bflock
, LK_EXCLUSIVE
| LK_PCATCH
);
1109 * Sync all mounts related to the media
1111 lockmgr(&hammer2_mntlk
, LK_EXCLUSIVE
);
1112 TAILQ_FOREACH(pmp
, &hammer2_pfslist
, mntentry
) {
1116 for (i
= 0; i
< HAMMER2_MAXCLUSTER
; ++i
) {
1117 if (pmp
->pfs_hmps
[i
] != hmp
)
1119 etmp
= hammer2_vfs_sync_pmp(pmp
, MNT_WAIT
);
1120 if (etmp
&& (error
== 0 || error
== ENOSPC
))
1125 lockmgr(&hammer2_mntlk
, LK_RELEASE
);
1127 if (error
&& error
!= ENOSPC
)
1131 * If we have an ENOSPC error we have to bulkfree on the live
1132 * topology. Otherwise we can bulkfree on a snapshot.
1135 kprintf("hammer2: WARNING! Bulkfree forced to use live "
1136 "topology due to ENOSPC\n");
1137 vchain
= &hmp
->vchain
;
1138 hammer2_chain_ref(vchain
);
1141 vchain
= hammer2_chain_bulksnap(hmp
);
1146 * Normal bulkfree operations do not require a transaction because
1147 * they operate on a snapshot, and so can run concurrently with
1148 * any operation except another bulkfree.
1150 * If we are running bulkfree on the live topology we have to be
1151 * in a FLUSH transaction.
1154 hammer2_trans_init(hmp
->spmp
, HAMMER2_TRANS_ISFLUSH
);
1157 hammer2_thr_freeze(&hmp
->bfthr
);
1158 error
= hammer2_bulkfree_pass(hmp
, vchain
, bfi
);
1159 hammer2_thr_unfreeze(&hmp
->bfthr
);
1162 hammer2_chain_bulkdrop(vchain
);
1164 hammer2_chain_drop(vchain
);
1165 hammer2_trans_done(hmp
->spmp
, HAMMER2_TRANS_ISFLUSH
|
1166 HAMMER2_TRANS_SIDEQ
);
1168 error
= hammer2_error_to_errno(error
);
1171 lockmgr(&hmp
->bflock
, LK_RELEASE
);
1176 * Unconditionally delete meta-data in a hammer2 filesystem
1179 hammer2_ioctl_destroy(hammer2_inode_t
*ip
, void *data
)
1181 hammer2_ioc_destroy_t
*iocd
= data
;
1182 hammer2_pfs_t
*pmp
= ip
->pmp
;
1191 case HAMMER2_DELETE_FILE
:
1193 * Destroy a bad directory entry by name. Caller must
1194 * pass the directory as fd.
1197 hammer2_xop_unlink_t
*xop
;
1199 if (iocd
->path
[sizeof(iocd
->path
)-1]) {
1203 if (ip
->meta
.type
!= HAMMER2_OBJTYPE_DIRECTORY
) {
1207 hammer2_pfs_memory_wait(pmp
);
1208 hammer2_trans_init(pmp
, 0);
1209 hammer2_inode_lock(ip
, 0);
1211 xop
= hammer2_xop_alloc(ip
, HAMMER2_XOP_MODIFYING
);
1212 hammer2_xop_setname(&xop
->head
, iocd
->path
, strlen(iocd
->path
));
1214 xop
->dopermanent
= H2DOPERM_PERMANENT
|
1217 hammer2_xop_start(&xop
->head
, &hammer2_unlink_desc
);
1219 error
= hammer2_xop_collect(&xop
->head
, 0);
1220 error
= hammer2_error_to_errno(error
);
1221 hammer2_inode_unlock(ip
);
1222 hammer2_xop_retire(&xop
->head
, HAMMER2_XOPMASK_VOP
);
1223 hammer2_trans_done(pmp
, HAMMER2_TRANS_SIDEQ
);
1226 case HAMMER2_DELETE_INUM
:
1228 * Destroy a bad inode by inode number.
1231 hammer2_xop_lookup_t
*xop
;
1233 if (iocd
->inum
< 1) {
1237 hammer2_pfs_memory_wait(pmp
);
1238 hammer2_trans_init(pmp
, 0);
1240 xop
= hammer2_xop_alloc(pmp
->iroot
, HAMMER2_XOP_MODIFYING
);
1241 xop
->lhc
= iocd
->inum
;
1242 hammer2_xop_start(&xop
->head
, &hammer2_delete_desc
);
1243 error
= hammer2_xop_collect(&xop
->head
, 0);
1244 error
= hammer2_error_to_errno(error
);
1245 hammer2_xop_retire(&xop
->head
, HAMMER2_XOPMASK_VOP
);
1246 hammer2_trans_done(pmp
, HAMMER2_TRANS_SIDEQ
);
1257 * Grow a filesystem into its partition size
1260 hammer2_ioctl_growfs(hammer2_inode_t
*ip
, void *data
, struct ucred
*cred
)
1262 hammer2_ioc_growfs_t
*grow
= data
;
1264 hammer2_off_t size
, delta
;
1266 struct partinfo part
;
1272 hmp
= ip
->pmp
->pfs_hmps
[0];
1274 if (hmp
->nvolumes
> 1) {
1275 kprintf("hammer2: growfs currently unsupported "
1276 "with multiple volumes\n");
1279 KKASSERT(hmp
->total_size
== hmp
->voldata
.volu_size
);
1282 * Extract from disklabel
1284 if (ioctl(hmp
->devvp
->fs
->fd
, DIOCGPART
, &part
) == 0) {
1285 size
= part
.media_size
;
1286 kprintf("hammer2: growfs partition-auto to %016jx\n",
1288 } else if (fstat(hmp
->devvp
->fs
->fd
, &st
) == 0) {
1290 kprintf("hammer2: growfs fstat-auto to %016jx\n",
1297 * Expand to devvp size unless specified.
1300 if (grow
->size
== 0) {
1302 } else if (grow
->size
> size
) {
1303 kprintf("hammer2: growfs size %016jx exceeds device size "
1305 (intmax_t)grow
->size
, (intmax_t)size
);
1310 * This is typically ~8MB alignment to avoid edge cases accessing
1311 * reserved blocks at the base of each 2GB zone.
1313 grow
->size
&= ~HAMMER2_VOLUME_ALIGNMASK64
;
1314 delta
= grow
->size
- hmp
->voldata
.volu_size
;
1317 * Maximum allowed size is 2^63
1319 if (grow
->size
> 0x7FFFFFFFFFFFFFFFLU
) {
1320 kprintf("hammer2: growfs failure, limit is 2^63 - 1 bytes\n");
1325 * We can't shrink a filesystem
1327 if (grow
->size
< hmp
->voldata
.volu_size
) {
1328 kprintf("hammer2: growfs failure, "
1329 "would shrink from %016jx to %016jx\n",
1330 (intmax_t)hmp
->voldata
.volu_size
,
1331 (intmax_t)grow
->size
);
1336 kprintf("hammer2: growfs - size did not change\n");
1341 * Clear any new volume header backups that we extend into.
1342 * Skip volume headers that are already part of the filesystem.
1344 for (i
= 0; i
< HAMMER2_NUM_VOLHDRS
; ++i
) {
1345 if (i
* HAMMER2_ZONE_BYTES64
< hmp
->voldata
.volu_size
)
1347 if (i
* HAMMER2_ZONE_BYTES64
>= grow
->size
)
1349 kprintf("hammer2: growfs - clear volhdr %d ", i
);
1350 error
= breadx(hmp
->devvp
, i
* HAMMER2_ZONE_BYTES64
,
1351 HAMMER2_VOLUME_BYTES
, &bp
);
1354 kprintf("I/O error %d\n", error
);
1357 bzero(bp
->b_data
, HAMMER2_VOLUME_BYTES
);
1360 kprintf("I/O error %d\n", error
);
1366 hammer2_trans_init(hmp
->spmp
, HAMMER2_TRANS_ISFLUSH
);
1367 mtid
= hammer2_trans_sub(hmp
->spmp
);
1369 kprintf("hammer2: growfs - expand by %016jx to %016jx mtid %016jx\n",
1370 (intmax_t)delta
, (intmax_t)grow
->size
, (intmax_t)mtid
);
1373 hammer2_voldata_lock(hmp
);
1374 hammer2_voldata_modify(hmp
);
1377 * NOTE: Just adjusting total_size for a single-volume filesystem
1378 * or for the last volume in a multi-volume filesystem, is
1379 * fine. But we can't grow any other partition in a multi-volume
1380 * filesystem. For now we just punt (at the top) on any
1381 * multi-volume filesystem.
1383 hmp
->voldata
.volu_size
= grow
->size
;
1384 hmp
->voldata
.total_size
+= delta
;
1385 hmp
->voldata
.allocator_size
+= delta
;
1386 hmp
->voldata
.allocator_free
+= delta
;
1387 hmp
->total_size
+= delta
;
1388 hmp
->volumes
[0].size
+= delta
; /* note: indexes first (only) volume */
1390 hammer2_voldata_unlock(hmp
);
1392 hammer2_trans_done(hmp
->spmp
, HAMMER2_TRANS_ISFLUSH
|
1393 HAMMER2_TRANS_SIDEQ
);
1397 * Flush the mess right here and now. We could just let the
1398 * filesystem syncer do it, but this was a sensitive operation
1399 * so don't take any chances.
1401 hammer2_vfs_sync(ip
->pmp
->mp
, MNT_WAIT
);
1408 * Get a list of volumes.
1411 hammer2_ioctl_volume_list(hammer2_inode_t
*ip
, void *data
)
1413 hammer2_ioc_volume_list_t
*vollist
= data
;
1414 hammer2_ioc_volume_t entry
;
1415 hammer2_volume_t
*vol
;
1418 int i
, error
= 0, cnt
= 0;
1421 hmp
= pmp
->pfs_hmps
[0];
1425 hammer2_voldata_lock(hmp
);
1426 for (i
= 0; i
< hmp
->nvolumes
; ++i
) {
1427 if (cnt
>= vollist
->nvolumes
)
1429 vol
= &hmp
->volumes
[i
];
1430 bzero(&entry
, sizeof(entry
));
1431 /* copy hammer2_volume_t fields */
1433 bcopy(vol
->dev
->path
, entry
.path
, sizeof(entry
.path
));
1434 entry
.offset
= vol
->offset
;
1435 entry
.size
= vol
->size
;
1436 error
= copyout(&entry
, &vollist
->volumes
[cnt
], sizeof(entry
));
1441 vollist
->nvolumes
= cnt
;
1442 vollist
->version
= hmp
->voldata
.version
;
1443 bcopy(pmp
->pfs_names
[0], vollist
->pfs_name
, sizeof(vollist
->pfs_name
));
1445 hammer2_voldata_unlock(hmp
);