kern - Convert aio from zalloc to objcache
[dragonfly.git] / sys / kern / vfs_conf.c
blob0730ccca765c1e067ef4aee45224b42d4d32a2f3
1 /*-
2 * Copyright (c) 1999 Michael Smith
3 * All rights reserved.
4 * Copyright (c) 1999 Poul-Henning Kamp
5 * All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * SUCH DAMAGE.
28 * $FreeBSD: src/sys/kern/vfs_conf.c,v 1.49.2.5 2003/01/07 11:56:53 joerg Exp $
29 * $DragonFly: src/sys/kern/vfs_conf.c,v 1.34 2008/05/24 19:08:28 dillon Exp $
33 * Locate and mount the root filesystem.
35 * The root filesystem is detailed in the kernel environment variable
36 * vfs.root.mountfrom, which is expected to be in the general format
38 * <vfsname>:[<path>]
39 * vfsname := the name of a VFS known to the kernel and capable
40 * of being mounted as root
41 * path := disk device name or other data used by the filesystem
42 * to locate its physical store
46 #include "opt_rootdevname.h"
48 #include <sys/param.h>
49 #include <sys/kernel.h>
50 #include <sys/systm.h>
51 #include <sys/proc.h>
52 #include <sys/vnode.h>
53 #include <sys/mount.h>
54 #include <sys/malloc.h>
55 #include <sys/reboot.h>
56 #include <sys/diskslice.h>
57 #include <sys/conf.h>
58 #include <sys/cons.h>
59 #include <sys/device.h>
60 #include <sys/disk.h>
61 #include <sys/namecache.h>
62 #include <sys/paths.h>
63 #include <sys/thread2.h>
64 #include <sys/nlookup.h>
65 #include <sys/devfs.h>
66 #include <sys/sysctl.h>
68 #include "opt_ddb.h"
69 #ifdef DDB
70 #include <ddb/ddb.h>
71 #endif
73 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure");
75 #define ROOTNAME "root_device"
77 struct vnode *rootvnode;
78 struct nchandle rootnch;
80 /*
81 * The root specifiers we will try if RB_CDROM is specified. Note that
82 * with DEVFS we do not use the compatibility slice's whole-disk 'c'
83 * partition. Instead we just use the whole disk, e.g. cd0 or cd0s0.
85 static char *cdrom_rootdevnames[] = {
86 "cd9660:cd0", /* SCSI (including AHCI and SILI) */
87 "cd9660:acd0", /* NATA */
88 "cd9660:cd1", /* SCSI (including AHCI and SILI) */
89 "cd9660:acd1", /* NATA */
90 "cd9660:cd8", /* USB */
91 "cd9660:cd9", /* USB */
92 NULL
95 int vfs_mountroot_devfs(void);
96 static void vfs_mountroot(void *junk);
97 static int vfs_mountroot_try(const char *mountfrom);
98 static int vfs_mountroot_ask(void);
99 static int getline(char *cp, int limit);
101 /* legacy find-root code */
102 char *rootdevnames[2] = {NULL, NULL};
103 static int setrootbyname(char *name);
105 SYSINIT(mountroot, SI_SUB_MOUNT_ROOT, SI_ORDER_SECOND, vfs_mountroot, NULL);
108 * Find and mount the root filesystem
110 static void
111 vfs_mountroot(void *junk)
113 cdev_t save_rootdev = rootdev;
114 int i;
115 int dummy;
118 * Make sure all disk devices created so far have also been probed,
119 * and also make sure that the newly created device nodes for
120 * probed disks are ready, too.
122 * Messages can fly around here so get good synchronization
123 * coverage.
125 * XXX - Delay an additional 2 seconds to help drivers which pickup
126 * devices asynchronously and are not caught by CAM's initial
127 * probe.
129 sync_devs();
130 tsleep(&dummy, 0, "syncer", hz*2);
134 * The root filesystem information is compiled in, and we are
135 * booted with instructions to use it.
137 #ifdef ROOTDEVNAME
138 if ((boothowto & RB_DFLTROOT) &&
139 !vfs_mountroot_try(ROOTDEVNAME))
140 return;
141 #endif
143 * We are booted with instructions to prompt for the root filesystem,
144 * or to use the compiled-in default when it doesn't exist.
146 if (boothowto & (RB_DFLTROOT | RB_ASKNAME)) {
147 if (!vfs_mountroot_ask())
148 return;
152 * We've been given the generic "use CDROM as root" flag. This is
153 * necessary because one media may be used in many different
154 * devices, so we need to search for them.
156 if (boothowto & RB_CDROM) {
157 for (i = 0; cdrom_rootdevnames[i] != NULL; i++) {
158 if (!vfs_mountroot_try(cdrom_rootdevnames[i]))
159 return;
164 * Try to use the value read by the loader from /etc/fstab, or
165 * supplied via some other means. This is the preferred
166 * mechanism.
168 if (!vfs_mountroot_try(kgetenv("vfs.root.mountfrom")))
169 return;
172 * If a vfs set rootdev, try it (XXX VINUM HACK!)
174 if (save_rootdev != NULL) {
175 rootdev = save_rootdev;
176 if (!vfs_mountroot_try(""))
177 return;
181 * Try values that may have been computed by the machine-dependant
182 * legacy code.
184 if (rootdevnames[0] && !vfs_mountroot_try(rootdevnames[0]))
185 return;
186 if (rootdevnames[1] && !vfs_mountroot_try(rootdevnames[1]))
187 return;
190 * If we have a compiled-in default, and haven't already tried it, try
191 * it now.
193 #ifdef ROOTDEVNAME
194 if (!(boothowto & RB_DFLTROOT))
195 if (!vfs_mountroot_try(ROOTDEVNAME))
196 return;
197 #endif
200 * Everything so far has failed, prompt on the console if we haven't
201 * already tried that.
203 if (!(boothowto & (RB_DFLTROOT | RB_ASKNAME)) && !vfs_mountroot_ask())
204 return;
205 panic("Root mount failed, startup aborted.");
210 vfs_mountroot_devfs(void)
212 struct vnode *vp;
213 struct nchandle nch;
214 struct nlookupdata nd;
215 struct mount *mp;
216 struct vfsconf *vfsp;
217 int error;
218 struct ucred *cred = proc0.p_ucred;
221 * Lookup the requested path and extract the nch and vnode.
223 error = nlookup_init_raw(&nd,
224 "/dev", UIO_SYSSPACE, NLC_FOLLOW,
225 cred, &rootnch);
227 if (error == 0) {
228 devfs_debug(DEVFS_DEBUG_DEBUG, "vfs_mountroot_devfs: nlookup_init is ok...\n");
229 if ((error = nlookup(&nd)) == 0) {
230 devfs_debug(DEVFS_DEBUG_DEBUG, "vfs_mountroot_devfs: nlookup is ok...\n");
231 if (nd.nl_nch.ncp->nc_vp == NULL) {
232 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: nlookup: simply not found\n");
233 error = ENOENT;
237 if (error) {
238 nlookup_done(&nd);
239 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: nlookup failed, error: %d\n", error);
240 return (error);
244 * Extract the locked+refd ncp and cleanup the nd structure
246 nch = nd.nl_nch;
247 cache_zero(&nd.nl_nch);
248 nlookup_done(&nd);
251 * now we have the locked ref'd nch and unreferenced vnode.
253 vp = nch.ncp->nc_vp;
254 if ((error = vget(vp, LK_EXCLUSIVE)) != 0) {
255 cache_put(&nch);
256 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vget failed\n");
257 return (error);
259 cache_unlock(&nch);
261 if ((error = vinvalbuf(vp, V_SAVE, 0, 0)) != 0) {
262 cache_drop(&nch);
263 vput(vp);
264 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vinvalbuf failed\n");
265 return (error);
267 if (vp->v_type != VDIR) {
268 cache_drop(&nch);
269 vput(vp);
270 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vp is not VDIR\n");
271 return (ENOTDIR);
274 vfsp = vfsconf_find_by_name("devfs");
275 vsetflags(vp, VMOUNT);
278 * Allocate and initialize the filesystem.
280 mp = kmalloc(sizeof(struct mount), M_MOUNT, M_ZERO|M_WAITOK);
281 mount_init(mp);
282 vfs_busy(mp, LK_NOWAIT);
283 mp->mnt_op = vfsp->vfc_vfsops;
284 mp->mnt_vfc = vfsp;
285 vfsp->vfc_refcount++;
286 mp->mnt_stat.f_type = vfsp->vfc_typenum;
287 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
288 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
289 mp->mnt_stat.f_owner = cred->cr_uid;
290 vn_unlock(vp);
293 * Mount the filesystem.
295 error = VFS_MOUNT(mp, "/dev", NULL, cred);
297 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
300 * Put the new filesystem on the mount list after root. The mount
301 * point gets its own mnt_ncmountpt (unless the VFS already set one
302 * up) which represents the root of the mount. The lookup code
303 * detects the mount point going forward and checks the root of
304 * the mount going backwards.
306 * It is not necessary to invalidate or purge the vnode underneath
307 * because elements under the mount will be given their own glue
308 * namecache record.
310 if (!error) {
311 if (mp->mnt_ncmountpt.ncp == NULL) {
313 * allocate, then unlock, but leave the ref intact
315 cache_allocroot(&mp->mnt_ncmountpt, mp, NULL);
316 cache_unlock(&mp->mnt_ncmountpt);
318 mp->mnt_ncmounton = nch; /* inherits ref */
319 nch.ncp->nc_flag |= NCF_ISMOUNTPT;
321 /* XXX get the root of the fs and cache_setvp(mnt_ncmountpt...) */
322 vclrflags(vp, VMOUNT);
323 mountlist_insert(mp, MNTINS_LAST);
324 vn_unlock(vp);
325 //checkdirs(&mp->mnt_ncmounton, &mp->mnt_ncmountpt);
326 error = vfs_allocate_syncvnode(mp);
327 if (error) {
328 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vfs_allocate_syncvnode failed\n");
330 vfs_unbusy(mp);
331 error = VFS_START(mp, 0);
332 vrele(vp);
333 } else {
334 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops);
335 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops);
336 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops);
337 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops);
338 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops);
339 vclrflags(vp, VMOUNT);
340 mp->mnt_vfc->vfc_refcount--;
341 vfs_unbusy(mp);
342 kfree(mp, M_MOUNT);
343 cache_drop(&nch);
344 vput(vp);
345 devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: mount failed\n");
348 devfs_debug(DEVFS_DEBUG_DEBUG, "rootmount_devfs done with error: %d\n", error);
349 return (error);
354 * Mount (mountfrom) as the root filesystem.
356 static int
357 vfs_mountroot_try(const char *mountfrom)
359 struct mount *mp, *mp2;
360 char *vfsname, *devname;
361 int error;
362 char patt[32];
363 int mountfromlen, len;
364 const char *cp, *ep;
365 char *mf;
367 vfsname = NULL;
368 devname = NULL;
369 mp = NULL;
370 mp2 = NULL;
371 error = EINVAL;
373 if (mountfrom == NULL)
374 return(error); /* don't complain */
376 crit_enter();
377 kprintf("Mounting root from %s\n", mountfrom);
378 crit_exit();
380 mountfromlen = strlen(mountfrom);
381 cp = mountfrom;
382 /* parse vfs name and devname */
383 vfsname = kmalloc(MFSNAMELEN, M_MOUNT, M_WAITOK);
384 devname = kmalloc(MNAMELEN, M_MOUNT, M_WAITOK);
385 mf = kmalloc(MFSNAMELEN+MNAMELEN, M_MOUNT, M_WAITOK);
386 for(;;) {
387 for (ep = cp; (*ep != 0) && (*ep != ';'); ep++);
388 len = ep - cp;
389 bzero(vfsname, MFSNAMELEN);
390 bzero(devname, MNAMELEN);
391 bzero(mf, MFSNAMELEN+MNAMELEN);
392 strncpy(mf, cp, MFSNAMELEN+MNAMELEN);
394 vfsname[0] = devname[0] = 0;
395 ksprintf(patt, "%%%d[a-z0-9]:%%%ds", MFSNAMELEN, MNAMELEN);
396 if (ksscanf(mf, patt, vfsname, devname) < 1)
397 goto end;
399 /* allocate a root mount */
400 error = vfs_rootmountalloc(vfsname,
401 devname[0] != 0 ? devname : ROOTNAME, &mp);
402 if (error != 0) {
403 kprintf("Can't allocate root mount for filesystem '%s': %d\n",
404 vfsname, error);
405 goto end;
407 mp->mnt_flag |= MNT_ROOTFS;
409 /* do our best to set rootdev */
410 if ((strcmp(vfsname, "hammer") != 0) && (devname[0] != 0) &&
411 setrootbyname(devname))
412 kprintf("setrootbyname failed\n");
414 /* If the root device is a type "memory disk", mount RW */
415 if (rootdev != NULL && dev_is_good(rootdev) &&
416 (dev_dflags(rootdev) & D_MEMDISK)) {
417 mp->mnt_flag &= ~MNT_RDONLY;
420 error = VFS_MOUNT(mp, NULL, NULL, proc0.p_ucred);
422 if (!error)
423 break;
424 end:
425 if(*ep == 0)
426 break;
427 cp = ep + 1;
430 if (vfsname != NULL)
431 kfree(vfsname, M_MOUNT);
432 if (devname != NULL)
433 kfree(devname, M_MOUNT);
434 if (mf != NULL)
435 kfree(mf, M_MOUNT);
436 if (error == 0) {
437 /* register with list of mounted filesystems */
438 mountlist_insert(mp, MNTINS_FIRST);
440 /* sanity check system clock against root fs timestamp */
441 inittodr(mp->mnt_time);
442 vfs_unbusy(mp);
443 if (mp->mnt_syncer == NULL) {
444 error = vfs_allocate_syncvnode(mp);
445 if (error)
446 kprintf("Warning: no syncer vp for root!\n");
447 error = 0;
449 } else {
450 if (mp != NULL) {
451 vfs_unbusy(mp);
452 kfree(mp, M_MOUNT);
454 kprintf("Root mount failed: %d\n", error);
456 return(error);
460 static void
461 vfs_mountroot_ask_callback(cdev_t dev, void *arg __unused)
463 if (dev_is_good(dev) && (dev_dflags(dev) & D_DISK))
464 kprintf(" \"%s\" ", dev->si_name);
469 * Spin prompting on the console for a suitable root filesystem
471 static int
472 vfs_mountroot_ask(void)
474 char name[128];
475 int llimit = 100;
477 kprintf("\nManual root filesystem specification:\n");
478 kprintf(" <fstype>:<device> Specify root (e.g. ufs:da0s1a)\n");
479 kprintf(" ? List valid disk boot devices\n");
480 kprintf(" panic Just panic\n");
481 kprintf(" abort Abort manual input\n");
482 while (llimit--) {
483 kprintf("\nmountroot> ");
485 if (getline(name, 128) < 0)
486 break;
487 if (name[0] == 0) {
489 } else if (name[0] == '?') {
490 kprintf("Possibly valid devices for root FS:\n");
491 //enumerate all disk devices
492 devfs_scan_callback(vfs_mountroot_ask_callback, NULL);
493 kprintf("\n");
494 continue;
495 } else if (strcmp(name, "panic") == 0) {
496 panic("panic from console");
497 } else if (strcmp(name, "abort") == 0) {
498 break;
499 } else if (vfs_mountroot_try(name) == 0) {
500 return(0);
503 return(1);
507 static int
508 getline(char *cp, int limit)
510 char *lp;
511 int c;
513 lp = cp;
514 for (;;) {
515 c = cngetc();
517 switch (c) {
518 case -1:
519 return(-1);
520 case '\n':
521 case '\r':
522 kprintf("\n");
523 *lp++ = '\0';
524 return(0);
525 case '\b':
526 case '\177':
527 if (lp > cp) {
528 kprintf("\b \b");
529 lp--;
530 } else {
531 kprintf("%c", 7);
533 continue;
534 case '#':
535 kprintf("#");
536 lp--;
537 if (lp < cp)
538 lp = cp;
539 continue;
540 case '@':
541 case 'u' & 037:
542 lp = cp;
543 kprintf("%c", '\n');
544 continue;
545 default:
546 if (lp - cp >= limit - 1) {
547 kprintf("%c", 7);
548 } else {
549 kprintf("%c", c);
550 *lp++ = c;
552 continue;
558 * Convert a given name to the cdev_t of the disk-like device
559 * it refers to.
561 struct kdbn_info {
562 const char *name;
563 int nlen;
564 int minor;
565 cdev_t dev;
569 cdev_t
570 kgetdiskbyname(const char *name)
572 char *cp;
573 cdev_t rdev;
576 * Get the base name of the device
578 if (strncmp(name, __SYS_PATH_DEV, sizeof(__SYS_PATH_DEV) - 1) == 0)
579 name += sizeof(__SYS_PATH_DEV) - 1;
580 cp = __DECONST(char *, name);
583 * Locate the device
585 kprintf("tryroot %s\n", name);
586 rdev = devfs_find_device_by_name(name);
587 if (rdev == NULL) {
588 kprintf("no disk named '%s'\n", name);
591 * FOUND DEVICE
593 return(rdev);
597 * Set rootdev to match (name), given that we expect it to
598 * refer to a disk-like device.
600 static int
601 setrootbyname(char *name)
603 cdev_t diskdev;
605 diskdev = kgetdiskbyname(name);
606 if (diskdev != NULL) {
607 rootdev = diskdev;
608 return (0);
610 /* set to NULL if kgetdiskbyname() fails so that if the first rootdev is
611 * found by fails to mount and the second one isn't found, mountroot_try
612 * doesn't try again with the first one
614 rootdev = NULL;
615 return (1);
618 #ifdef DDB
619 DB_SHOW_COMMAND(disk, db_getdiskbyname)
621 cdev_t dev;
623 if (modif[0] == '\0') {
624 db_error("usage: show disk/devicename");
625 return;
627 dev = kgetdiskbyname(modif);
628 if (dev != NULL)
629 db_printf("cdev_t = %p\n", dev);
630 else
631 db_printf("No disk device matched.\n");
633 #endif
635 static int
636 vfs_sysctl_real_root(SYSCTL_HANDLER_ARGS)
638 char *real_root;
639 size_t len;
640 int error;
642 real_root = kgetenv("vfs.root.realroot");
644 if (real_root == NULL)
645 real_root = "";
647 len = strlen(real_root) + 1;
649 error = sysctl_handle_string(oidp, real_root, len, req);
651 return error;
654 SYSCTL_PROC(_vfs, OID_AUTO, real_root,
655 CTLTYPE_STRING | CTLFLAG_RD, 0, 0, vfs_sysctl_real_root,
656 "A", "Real root mount string");