1 /*-
2  * Copyright (c) 1990 The Regents of the University of California.
3  * Copyright (c) 2008 The DragonFly Project.
4  * All rights reserved.
5  *
6  * This code is derived from software contributed to Berkeley by
7  * William Jolitz.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. All advertising materials mentioning features or use of this software
18  *    must display the following acknowledgement:
19  *	This product includes software developed by the University of
20  *	California, Berkeley and its contributors.
21  * 4. Neither the name of the University nor the names of its contributors
22  *    may be used to endorse or promote products derived from this software
23  *    without specific prior written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35  * SUCH DAMAGE.
36  *
37  *	from: @(#)autoconf.c	7.1 (Berkeley) 5/9/91
38  * $FreeBSD: src/sys/i386/i386/autoconf.c,v 1.146.2.2 2001/06/07 06:05:58 dd Exp $
39  */
40 
41 /*
42  * Setup the system to run on the current machine.
43  *
44  * Configure() is called at boot time and initializes the vba
45  * device tables and the memory controller monitoring.  Available
46  * devices are determined (from possibilities mentioned in ioconf.c),
47  * and the drivers are initialized.
48  */
49 #include "opt_bootp.h"
50 #include "opt_ffs.h"
51 #include "opt_cd9660.h"
52 #include "opt_nfs.h"
53 #include "opt_nfsroot.h"
54 #include "opt_rootdevname.h"
55 
56 #include "use_isa.h"
57 
58 #include <sys/param.h>
59 #include <sys/systm.h>
60 #include <sys/bootmaj.h>
61 #include <sys/bus.h>
62 #include <sys/buf.h>
63 #include <sys/conf.h>
64 #include <sys/diskslice.h>
65 #include <sys/reboot.h>
66 #include <sys/kernel.h>
67 #include <sys/malloc.h>
68 #include <sys/mount.h>
69 #include <sys/cons.h>
70 #include <sys/thread.h>
71 #include <sys/device.h>
72 #include <sys/machintr.h>
73 
74 #include <vm/vm_kern.h>
75 #include <vm/vm_extern.h>
76 #include <vm/vm_pager.h>
77 
78 #if 0
79 #include <machine/pcb.h>
80 #include <machine/pcb_ext.h>
81 #endif
82 #include <machine/smp.h>
83 #include <machine/globaldata.h>
84 #include <machine/md_var.h>
85 
86 #if NISA > 0
87 #include <bus/isa/isavar.h>
88 
89 device_t isa_bus_device = NULL;
90 #endif
91 
92 static void cpu_startup (void *);
93 static void configure_first (void *);
94 static void configure (void *);
95 static void configure_final (void *);
96 
97 #if defined(FFS) && defined(FFS_ROOT)
98 static void	setroot (void);
99 #endif
100 
101 #if defined(NFS) && defined(NFS_ROOT)
102 #if !defined(BOOTP_NFSROOT)
103 static void	pxe_setup_nfsdiskless(void);
104 #endif
105 #endif
106 
107 SYSINIT(cpu, SI_BOOT2_START_CPU, SI_ORDER_FIRST, cpu_startup, NULL);
108 SYSINIT(configure1, SI_SUB_CONFIGURE, SI_ORDER_FIRST, configure_first, NULL);
109 /* SI_ORDER_SECOND is hookable */
110 SYSINIT(configure2, SI_SUB_CONFIGURE, SI_ORDER_THIRD, configure, NULL);
111 /* SI_ORDER_MIDDLE is hookable */
112 SYSINIT(configure3, SI_SUB_CONFIGURE, SI_ORDER_ANY, configure_final, NULL);
113 
114 cdev_t	rootdev = NULL;
115 cdev_t	dumpdev = NULL;
116 
117 /*
118  * nfsroot.iosize may be set in loader.conf, 32768 is recommended to
119  * be able to max-out a GigE link if the server supports it.  Many servers
120  * do not so the default is 8192.
121  *
122  * nfsroot.rahead defaults to something reasonable, can be overridden.
123  */
124 static int nfsroot_iosize = 8192;
125 TUNABLE_INT("nfsroot.iosize", &nfsroot_iosize);
126 static int nfsroot_rahead = 4;
127 TUNABLE_INT("nfsroot.rahead", &nfsroot_rahead);
128 
129 /*
130  *
131  */
132 static void
133 cpu_startup(void *dummy)
134 {
135 	vm_offset_t buffer_sva;
136 	vm_offset_t buffer_eva;
137 	vm_offset_t pager_sva;
138 	vm_offset_t pager_eva;
139 
140 	kprintf("%s", version);
141 	kprintf("real memory = %ju (%juK bytes)\n",
142 	    (uintmax_t)ptoa(Maxmem), (uintmax_t)(ptoa(Maxmem) / 1024));
143 
144 	if (nbuf == 0) {
145 		int factor = 4 * BKVASIZE / 1024;
146 		int kbytes = Maxmem * (PAGE_SIZE / 1024);
147 
148 		nbuf = 50;
149 		if (kbytes > 4096)
150 			nbuf += min((kbytes - 4096) / factor, 65536 / factor);
151 		if (kbytes > 65536)
152 			nbuf += (kbytes - 65536) * 2 / (factor * 5);
153 		if (maxbcache && nbuf > maxbcache / BKVASIZE)
154 			nbuf = maxbcache / BKVASIZE;
155 	}
156 	if (nbuf > (virtual_end - virtual_start) / (BKVASIZE * 2)) {
157 		nbuf = (virtual_end - virtual_start) / (BKVASIZE * 2);
158 		kprintf("Warning: nbufs capped at %ld\n", nbuf);
159 	}
160 
161 	nswbuf = lmax(lmin(nbuf / 4, 256), 16);
162 #ifdef NSWBUF_MIN
163 	if (nswbuf < NSWBUF_MIN)
164 		nswbuf = NSWBUF_MIN;
165 #endif
166 
167 	/*
168 	 * Allocate memory for the buffer cache
169 	 */
170 	buf = (void *)kmem_alloc(&kernel_map, nbuf * sizeof(struct buf));
171 	swbuf = (void *)kmem_alloc(&kernel_map, nswbuf * sizeof(struct buf));
172 
173 
174 #ifdef DIRECTIO
175         ffs_rawread_setup();
176 #endif
177 	kmem_suballoc(&kernel_map, &clean_map, &clean_sva, &clean_eva,
178 		      (nbuf*BKVASIZE*2) + (nswbuf*MAXPHYS) + pager_map_size);
179 	kmem_suballoc(&clean_map, &buffer_map, &buffer_sva, &buffer_eva,
180 		      (nbuf*BKVASIZE*2));
181 	buffer_map.system_map = 1;
182 	kmem_suballoc(&clean_map, &pager_map, &pager_sva, &pager_eva,
183 		      (nswbuf*MAXPHYS) + pager_map_size);
184 	pager_map.system_map = 1;
185 	kprintf("avail memory = %lu (%luK bytes)\n", ptoa(vmstats.v_free_count),
186 		ptoa(vmstats.v_free_count) / 1024);
187 	mp_start();
188 	mp_announce();
189 	cpu_setregs();
190 }
191 
192 /*
193  * Determine i/o configuration for a machine.
194  */
195 static void
196 configure_first(void *dummy)
197 {
198 }
199 
200 static void
201 configure(void *dummy)
202 {
203         /*
204 	 * Final interrupt support acviation, then enable hardware interrupts.
205 	 */
206 	MachIntrABI.finalize();
207 	cpu_enable_intr();
208 
209 	/*
210 	 * This will configure all devices, generally starting with the
211 	 * nexus (i386/i386/nexus.c).  The nexus ISA code explicitly
212 	 * dummies up the attach in order to delay legacy initialization
213 	 * until after all other busses/subsystems have had a chance
214 	 * at those resources.
215 	 */
216 	root_bus_configure();
217 
218 #if NISA > 0
219 	/*
220 	 * Explicitly probe and attach ISA last.  The isa bus saves
221 	 * it's device node at attach time for us here.
222 	 */
223 	if (isa_bus_device)
224 		isa_probe_children(isa_bus_device);
225 #endif
226 
227 	/*
228 	 * Allow lowering of the ipl to the lowest kernel level if we
229 	 * panic (or call tsleep() before clearing `cold').  No level is
230 	 * completely safe (since a panic may occur in a critical region
231 	 * at splhigh()), but we want at least bio interrupts to work.
232 	 */
233 	safepri = TDPRI_KERN_USER;
234 }
235 
236 static void
237 configure_final(void *dummy)
238 {
239 	cninit_finish();
240 
241 	if (bootverbose)
242 		kprintf("Device configuration finished.\n");
243 }
244 
245 #ifdef BOOTP
246 void bootpc_init(void);
247 #endif
248 /*
249  * Do legacy root filesystem discovery.
250  */
251 void
252 cpu_rootconf(void)
253 {
254 #ifdef BOOTP
255         bootpc_init();
256 #endif
257 #if defined(NFS) && defined(NFS_ROOT)
258 #if !defined(BOOTP_NFSROOT)
259 	pxe_setup_nfsdiskless();
260 	if (nfs_diskless_valid)
261 #endif
262 		rootdevnames[0] = "nfs:";
263 #endif
264 #if defined(FFS) && defined(FFS_ROOT)
265         if (!rootdevnames[0])
266                 setroot();
267 #endif
268 }
269 SYSINIT(cpu_rootconf, SI_SUB_ROOT_CONF, SI_ORDER_FIRST, cpu_rootconf, NULL);
270 
271 u_long	bootdev = 0;		/* not a cdev_t - encoding is different */
272 
273 #if defined(FFS) && defined(FFS_ROOT)
274 
275 /*
276  * The boot code uses old block device major numbers to pass bootdev to
277  * us.  We have to translate these to character device majors because
278  * we don't have block devices any more.
279  */
280 static int
281 boot_translate_majdev(int bmajor)
282 {
283 	static int conv[] = { BOOTMAJOR_CONVARY };
284 
285 	if (bmajor >= 0 && bmajor < NELEM(conv))
286 		return(conv[bmajor]);
287 	return(-1);
288 }
289 
290 /*
291  * Attempt to find the device from which we were booted.
292  * If we can do so, and not instructed not to do so,
293  * set rootdevs[] and rootdevnames[] to correspond to the
294  * boot device(s).
295  *
296  * This code survives in order to allow the system to be
297  * booted from legacy environments that do not correctly
298  * populate the kernel environment. There are significant
299  * restrictions on the bootability of the system in this
300  * situation; it can only be mounting root from a 'da'
301  * 'wd' or 'fd' device, and the root filesystem must be ufs.
302  */
303 static void
304 setroot(void)
305 {
306 	int majdev, mindev, unit, slice, part;
307 	cdev_t newrootdev, dev;
308 	char partname[2];
309 	char *sname;
310 
311 	if ((bootdev & B_MAGICMASK) != B_DEVMAGIC) {
312 		kprintf("no B_DEVMAGIC (bootdev=%#lx)\n", bootdev);
313 		return;
314 	}
315 	majdev = boot_translate_majdev(B_TYPE(bootdev));
316 	if (bootverbose) {
317 		kprintf("bootdev: %08lx type=%ld unit=%ld "
318 			"slice=%ld part=%ld major=%d\n",
319 			bootdev, B_TYPE(bootdev), B_UNIT(bootdev),
320 			B_SLICE(bootdev), B_PARTITION(bootdev), majdev);
321 	}
322 	dev = udev2dev(makeudev(majdev, 0), 0);
323 	if (!dev_is_good(dev))
324 		return;
325 	unit = B_UNIT(bootdev);
326 	slice = B_SLICE(bootdev);
327 	if (slice == WHOLE_DISK_SLICE)
328 		slice = COMPATIBILITY_SLICE;
329 	if (slice < 0 || slice >= MAX_SLICES) {
330 		kprintf("bad slice\n");
331 		return;
332 	}
333 
334 	part = B_PARTITION(bootdev);
335 	mindev = dkmakeminor(unit, slice, part);
336 	newrootdev = udev2dev(makeudev(majdev, mindev), 0);
337 	if (!dev_is_good(newrootdev))
338 		return;
339 	sname = dsname(newrootdev, unit, slice, part, partname);
340 	rootdevnames[0] = kmalloc(strlen(sname) + 6, M_DEVBUF, M_WAITOK);
341 	ksprintf(rootdevnames[0], "ufs:%s%s", sname, partname);
342 
343 	/*
344 	 * For properly dangerously dedicated disks (ones with a historical
345 	 * bogus partition table), the boot blocks will give slice = 4, but
346 	 * the kernel will only provide the compatibility slice since it
347 	 * knows that slice 4 is not a real slice.  Arrange to try mounting
348 	 * the compatibility slice as root if mounting the slice passed by
349 	 * the boot blocks fails.  This handles the dangerously dedicated
350 	 * case and perhaps others.
351 	 */
352 	if (slice == COMPATIBILITY_SLICE)
353 		return;
354 	slice = COMPATIBILITY_SLICE;
355 	sname = dsname(newrootdev, unit, slice, part, partname);
356 	rootdevnames[1] = kmalloc(strlen(sname) + 6, M_DEVBUF, M_WAITOK);
357 	ksprintf(rootdevnames[1], "ufs:%s%s", sname, partname);
358 }
359 #endif
360 
361 #if defined(NFS) && defined(NFS_ROOT)
362 #if !defined(BOOTP_NFSROOT)
363 
364 #include <sys/socket.h>
365 #include <net/if.h>
366 #include <net/if_dl.h>
367 #include <net/if_types.h>
368 #include <net/if_var.h>
369 #include <net/ethernet.h>
370 #include <netinet/in.h>
371 #include <vfs/nfs/rpcv2.h>
372 #include <vfs/nfs/nfsproto.h>
373 #include <vfs/nfs/nfs.h>
374 #include <vfs/nfs/nfsdiskless.h>
375 
376 extern struct nfs_diskless	nfs_diskless;
377 
378 /*
379  * Convert a kenv variable to a sockaddr.  If the kenv variable does not
380  * exist the sockaddr will remain zerod out (callers typically just check
381  * sin_len).  A network address of 0.0.0.0 is equivalent to failure.
382  */
383 static int
384 inaddr_to_sockaddr(char *ev, struct sockaddr_in *sa)
385 {
386 	u_int32_t	a[4];
387 	char		*cp;
388 
389 	bzero(sa, sizeof(*sa));
390 
391 	if ((cp = kgetenv(ev)) == NULL)
392 		return(1);
393 	if (ksscanf(cp, "%d.%d.%d.%d", &a[0], &a[1], &a[2], &a[3]) != 4)
394 		return(1);
395 	if (a[0] == 0 && a[1] == 0 && a[2] == 0 && a[3] == 0)
396 		return(1);
397 	/* XXX is this ordering correct? */
398 	sa->sin_addr.s_addr = (a[3] << 24) + (a[2] << 16) + (a[1] << 8) + a[0];
399 	sa->sin_len = sizeof(*sa);
400 	sa->sin_family = AF_INET;
401 	return(0);
402 }
403 
404 static int
405 hwaddr_to_sockaddr(char *ev, struct sockaddr_dl *sa)
406 {
407 	char		*cp;
408 	u_int32_t	a[6];
409 
410 	bzero(sa, sizeof(*sa));
411 	sa->sdl_len = sizeof(*sa);
412 	sa->sdl_family = AF_LINK;
413 	sa->sdl_type = IFT_ETHER;
414 	sa->sdl_alen = ETHER_ADDR_LEN;
415 	if ((cp = kgetenv(ev)) == NULL)
416 		return(1);
417 	if (ksscanf(cp, "%x:%x:%x:%x:%x:%x", &a[0], &a[1], &a[2], &a[3], &a[4], &a[5]) != 6)
418 		return(1);
419 	sa->sdl_data[0] = a[0];
420 	sa->sdl_data[1] = a[1];
421 	sa->sdl_data[2] = a[2];
422 	sa->sdl_data[3] = a[3];
423 	sa->sdl_data[4] = a[4];
424 	sa->sdl_data[5] = a[5];
425 	return(0);
426 }
427 
428 static int
429 decode_nfshandle(char *ev, u_char *fh)
430 {
431 	u_char	*cp;
432 	int	len, val;
433 
434 	if (((cp = kgetenv(ev)) == NULL) || (strlen(cp) < 2) || (*cp != 'X'))
435 		return(0);
436 	len = 0;
437 	cp++;
438 	for (;;) {
439 		if (*cp == 'X')
440 			return(len);
441 		if ((ksscanf(cp, "%2x", &val) != 1) || (val > 0xff))
442 			return(0);
443 		*(fh++) = val;
444 		len++;
445 		cp += 2;
446 		if (len > NFSX_V2FH)
447 		    return(0);
448 	}
449 }
450 
451 /*
452  * Populate the essential fields in the nfsv3_diskless structure.
453  *
454  * The loader is expected to export the following environment variables:
455  *
456  * boot.netif.ip		IP address on boot interface
457  * boot.netif.netmask		netmask on boot interface
458  * boot.netif.gateway		default gateway (optional)
459  * boot.netif.hwaddr		hardware address of boot interface
460  * boot.netif.name		name of boot interface (instead of hw addr)
461  * boot.nfsroot.server		IP address of root filesystem server
462  * boot.nfsroot.path		path of the root filesystem on server
463  * boot.nfsroot.nfshandle	NFS handle for root filesystem on server
464  */
465 static void
466 pxe_setup_nfsdiskless(void)
467 {
468 	struct nfs_diskless	*nd = &nfs_diskless;
469 	struct ifnet		*ifp;
470 	struct ifaddr		*ifa;
471 	struct sockaddr_dl	*sdl, ourdl;
472 	struct sockaddr_in	myaddr, netmask;
473 	char			*cp;
474 
475 	/* set up interface */
476 	if (inaddr_to_sockaddr("boot.netif.ip", &myaddr))
477 		return;
478 	if (inaddr_to_sockaddr("boot.netif.netmask", &netmask)) {
479 		kprintf("PXE: no netmask\n");
480 		return;
481 	}
482 	bcopy(&myaddr, &nd->myif.ifra_addr, sizeof(myaddr));
483 	bcopy(&myaddr, &nd->myif.ifra_broadaddr, sizeof(myaddr));
484 	((struct sockaddr_in *) &nd->myif.ifra_broadaddr)->sin_addr.s_addr =
485 		myaddr.sin_addr.s_addr | ~ netmask.sin_addr.s_addr;
486 	bcopy(&netmask, &nd->myif.ifra_mask, sizeof(netmask));
487 
488 	if ((cp = kgetenv("boot.netif.name")) != NULL) {
489 		ifnet_lock();
490 		ifp = ifunit(cp);
491 		if (ifp) {
492 			strlcpy(nd->myif.ifra_name, ifp->if_xname,
493 			    sizeof(nd->myif.ifra_name));
494 			ifnet_unlock();
495 			goto match_done;
496 		}
497 		ifnet_unlock();
498 		kprintf("PXE: cannot find interface %s\n", cp);
499 		return;
500 	}
501 
502 	if (hwaddr_to_sockaddr("boot.netif.hwaddr", &ourdl)) {
503 		kprintf("PXE: no hardware address\n");
504 		return;
505 	}
506 	ifa = NULL;
507 	ifnet_lock();
508 	TAILQ_FOREACH(ifp, &ifnetlist, if_link) {
509 		struct ifaddr_container *ifac;
510 
511 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
512 			ifa = ifac->ifa;
513 
514 			if ((ifa->ifa_addr->sa_family == AF_LINK) &&
515 			    (sdl = ((struct sockaddr_dl *)ifa->ifa_addr))) {
516 				if ((sdl->sdl_type == ourdl.sdl_type) &&
517 				    (sdl->sdl_alen == ourdl.sdl_alen) &&
518 				    !bcmp(sdl->sdl_data + sdl->sdl_nlen,
519 					  ourdl.sdl_data + ourdl.sdl_nlen,
520 					  sdl->sdl_alen)) {
521 					strlcpy(nd->myif.ifra_name,
522 					    ifp->if_xname,
523 					    sizeof(nd->myif.ifra_name));
524 					ifnet_unlock();
525 					goto match_done;
526 				}
527 			}
528 		}
529 	}
530 	ifnet_unlock();
531 	kprintf("PXE: no interface\n");
532 	return;	/* no matching interface */
533 match_done:
534 	/* set up gateway */
535 	inaddr_to_sockaddr("boot.netif.gateway", &nd->mygateway);
536 
537 	/* XXX set up swap? */
538 
539 	/* set up root mount */
540 	nd->root_args.rsize = nfsroot_iosize;
541 	nd->root_args.wsize = nfsroot_iosize;
542 	nd->root_args.sotype = SOCK_STREAM;
543 	nd->root_args.readahead = nfsroot_rahead;
544 	nd->root_args.flags = NFSMNT_WSIZE | NFSMNT_RSIZE | NFSMNT_RESVPORT |
545 			      NFSMNT_READAHEAD;
546 	if (inaddr_to_sockaddr("boot.nfsroot.server", &nd->root_saddr)) {
547 		kprintf("PXE: no server\n");
548 		return;
549 	}
550 	nd->root_saddr.sin_port = htons(NFS_PORT);
551 
552 	/*
553 	 * A tftp-only loader may pass NFS path information without a
554 	 * root handle.  Generate a warning but continue configuring.
555 	 */
556 	if (decode_nfshandle("boot.nfsroot.nfshandle", &nd->root_fh[0]) == 0) {
557 		kprintf("PXE: Warning, no NFS handle passed from loader\n");
558 	}
559 	if ((cp = kgetenv("boot.nfsroot.path")) != NULL)
560 		strncpy(nd->root_hostnam, cp, MNAMELEN - 1);
561 
562 	nfs_diskless_valid = 1;
563 }
564 
565 #endif
566 #endif
567