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  *
119  */
120 static void
121 cpu_startup(void *dummy)
122 {
123 	vm_offset_t buffer_sva;
124 	vm_offset_t buffer_eva;
125 	vm_offset_t pager_sva;
126 	vm_offset_t pager_eva;
127 
128 	kprintf("%s", version);
129 	kprintf("real memory = %ju (%juK bytes)\n",
130 	    (uintmax_t)ptoa(Maxmem), (uintmax_t)(ptoa(Maxmem) / 1024));
131 
132 	if (nbuf == 0) {
133 		int factor = 4 * BKVASIZE / 1024;
134 		int kbytes = Maxmem * (PAGE_SIZE / 1024);
135 
136 		nbuf = 50;
137 		if (kbytes > 4096)
138 			nbuf += min((kbytes - 4096) / factor, 65536 / factor);
139 		if (kbytes > 65536)
140 			nbuf += (kbytes - 65536) * 2 / (factor * 5);
141 		if (maxbcache && nbuf > maxbcache / BKVASIZE)
142 			nbuf = maxbcache / BKVASIZE;
143 	}
144 	if (nbuf > (virtual_end - virtual_start) / (BKVASIZE * 2)) {
145 		nbuf = (virtual_end - virtual_start) / (BKVASIZE * 2);
146 		kprintf("Warning: nbufs capped at %ld\n", nbuf);
147 	}
148 
149 	nswbuf = lmax(lmin(nbuf / 4, 256), 16);
150 #ifdef NSWBUF_MIN
151 	if (nswbuf < NSWBUF_MIN)
152 		nswbuf = NSWBUF_MIN;
153 #endif
154 
155 	/*
156 	 * Allocate memory for the buffer cache
157 	 */
158 	buf = (void *)kmem_alloc(&kernel_map, nbuf * sizeof(struct buf));
159 	swbuf = (void *)kmem_alloc(&kernel_map, nswbuf * sizeof(struct buf));
160 
161 
162 #ifdef DIRECTIO
163         ffs_rawread_setup();
164 #endif
165 	kmem_suballoc(&kernel_map, &clean_map, &clean_sva, &clean_eva,
166 		      (nbuf*BKVASIZE*2) + (nswbuf*MAXPHYS) + pager_map_size);
167 	kmem_suballoc(&clean_map, &buffer_map, &buffer_sva, &buffer_eva,
168 		      (nbuf*BKVASIZE*2));
169 	buffer_map.system_map = 1;
170 	kmem_suballoc(&clean_map, &pager_map, &pager_sva, &pager_eva,
171 		      (nswbuf*MAXPHYS) + pager_map_size);
172 	pager_map.system_map = 1;
173 #if defined(USERCONFIG)
174         userconfig();
175 	cninit();               /* the preferred console may have changed */
176 #endif
177 	kprintf("avail memory = %lu (%luK bytes)\n", ptoa(vmstats.v_free_count),
178 		ptoa(vmstats.v_free_count) / 1024);
179 	mp_start();
180 	mp_announce();
181 	cpu_setregs();
182 }
183 
184 /*
185  * Determine i/o configuration for a machine.
186  */
187 static void
188 configure_first(void *dummy)
189 {
190 }
191 
192 static void
193 configure(void *dummy)
194 {
195         /*
196 	 * Final interrupt support acviation, then enable hardware interrupts.
197 	 */
198 	MachIntrABI.finalize();
199 	cpu_enable_intr();
200 
201 	/*
202 	 * This will configure all devices, generally starting with the
203 	 * nexus (i386/i386/nexus.c).  The nexus ISA code explicitly
204 	 * dummies up the attach in order to delay legacy initialization
205 	 * until after all other busses/subsystems have had a chance
206 	 * at those resources.
207 	 */
208 	root_bus_configure();
209 
210 #if NISA > 0
211 	/*
212 	 * Explicitly probe and attach ISA last.  The isa bus saves
213 	 * it's device node at attach time for us here.
214 	 */
215 	if (isa_bus_device)
216 		isa_probe_children(isa_bus_device);
217 #endif
218 
219 	/*
220 	 * Allow lowering of the ipl to the lowest kernel level if we
221 	 * panic (or call tsleep() before clearing `cold').  No level is
222 	 * completely safe (since a panic may occur in a critical region
223 	 * at splhigh()), but we want at least bio interrupts to work.
224 	 */
225 	safepri = TDPRI_KERN_USER;
226 }
227 
228 static void
229 configure_final(void *dummy)
230 {
231 	cninit_finish();
232 
233 	if (bootverbose)
234 		kprintf("Device configuration finished.\n");
235 }
236 
237 #ifdef BOOTP
238 void bootpc_init(void);
239 #endif
240 /*
241  * Do legacy root filesystem discovery.
242  */
243 void
244 cpu_rootconf(void)
245 {
246 #ifdef BOOTP
247         bootpc_init();
248 #endif
249 #if defined(NFS) && defined(NFS_ROOT)
250 #if !defined(BOOTP_NFSROOT)
251 	pxe_setup_nfsdiskless();
252 	if (nfs_diskless_valid)
253 #endif
254 		rootdevnames[0] = "nfs:";
255 #endif
256 #if defined(FFS) && defined(FFS_ROOT)
257         if (!rootdevnames[0])
258                 setroot();
259 #endif
260 }
261 SYSINIT(cpu_rootconf, SI_SUB_ROOT_CONF, SI_ORDER_FIRST, cpu_rootconf, NULL)
262 
263 u_long	bootdev = 0;		/* not a cdev_t - encoding is different */
264 
265 #if defined(FFS) && defined(FFS_ROOT)
266 
267 /*
268  * The boot code uses old block device major numbers to pass bootdev to
269  * us.  We have to translate these to character device majors because
270  * we don't have block devices any more.
271  */
272 static int
273 boot_translate_majdev(int bmajor)
274 {
275 	static int conv[] = { BOOTMAJOR_CONVARY };
276 
277 	if (bmajor >= 0 && bmajor < NELEM(conv))
278 		return(conv[bmajor]);
279 	return(-1);
280 }
281 
282 /*
283  * Attempt to find the device from which we were booted.
284  * If we can do so, and not instructed not to do so,
285  * set rootdevs[] and rootdevnames[] to correspond to the
286  * boot device(s).
287  *
288  * This code survives in order to allow the system to be
289  * booted from legacy environments that do not correctly
290  * populate the kernel environment. There are significant
291  * restrictions on the bootability of the system in this
292  * situation; it can only be mounting root from a 'da'
293  * 'wd' or 'fd' device, and the root filesystem must be ufs.
294  */
295 static void
296 setroot(void)
297 {
298 	int majdev, mindev, unit, slice, part;
299 	cdev_t newrootdev, dev;
300 	char partname[2];
301 	char *sname;
302 
303 	if ((bootdev & B_MAGICMASK) != B_DEVMAGIC) {
304 		kprintf("no B_DEVMAGIC (bootdev=%#lx)\n", bootdev);
305 		return;
306 	}
307 	majdev = boot_translate_majdev(B_TYPE(bootdev));
308 	if (bootverbose) {
309 		kprintf("bootdev: %08lx type=%ld unit=%ld "
310 			"slice=%ld part=%ld major=%d\n",
311 			bootdev, B_TYPE(bootdev), B_UNIT(bootdev),
312 			B_SLICE(bootdev), B_PARTITION(bootdev), majdev);
313 	}
314 	dev = udev2dev(makeudev(majdev, 0), 0);
315 	if (!dev_is_good(dev))
316 		return;
317 	unit = B_UNIT(bootdev);
318 	slice = B_SLICE(bootdev);
319 	if (slice == WHOLE_DISK_SLICE)
320 		slice = COMPATIBILITY_SLICE;
321 	if (slice < 0 || slice >= MAX_SLICES) {
322 		kprintf("bad slice\n");
323 		return;
324 	}
325 
326 	part = B_PARTITION(bootdev);
327 	mindev = dkmakeminor(unit, slice, part);
328 	newrootdev = udev2dev(makeudev(majdev, mindev), 0);
329 	if (!dev_is_good(newrootdev))
330 		return;
331 	sname = dsname(newrootdev, unit, slice, part, partname);
332 	rootdevnames[0] = kmalloc(strlen(sname) + 6, M_DEVBUF, M_WAITOK);
333 	ksprintf(rootdevnames[0], "ufs:%s%s", sname, partname);
334 
335 	/*
336 	 * For properly dangerously dedicated disks (ones with a historical
337 	 * bogus partition table), the boot blocks will give slice = 4, but
338 	 * the kernel will only provide the compatibility slice since it
339 	 * knows that slice 4 is not a real slice.  Arrange to try mounting
340 	 * the compatibility slice as root if mounting the slice passed by
341 	 * the boot blocks fails.  This handles the dangerously dedicated
342 	 * case and perhaps others.
343 	 */
344 	if (slice == COMPATIBILITY_SLICE)
345 		return;
346 	slice = COMPATIBILITY_SLICE;
347 	sname = dsname(newrootdev, unit, slice, part, partname);
348 	rootdevnames[1] = kmalloc(strlen(sname) + 6, M_DEVBUF, M_WAITOK);
349 	ksprintf(rootdevnames[1], "ufs:%s%s", sname, partname);
350 }
351 #endif
352 
353 #if defined(NFS) && defined(NFS_ROOT)
354 #if !defined(BOOTP_NFSROOT)
355 
356 #include <sys/socket.h>
357 #include <net/if.h>
358 #include <net/if_dl.h>
359 #include <net/if_types.h>
360 #include <net/if_var.h>
361 #include <net/ethernet.h>
362 #include <netinet/in.h>
363 #include <vfs/nfs/rpcv2.h>
364 #include <vfs/nfs/nfsproto.h>
365 #include <vfs/nfs/nfs.h>
366 #include <vfs/nfs/nfsdiskless.h>
367 
368 extern struct nfs_diskless	nfs_diskless;
369 
370 /*
371  * Convert a kenv variable to a sockaddr.  If the kenv variable does not
372  * exist the sockaddr will remain zerod out (callers typically just check
373  * sin_len).  A network address of 0.0.0.0 is equivalent to failure.
374  */
375 static int
376 inaddr_to_sockaddr(char *ev, struct sockaddr_in *sa)
377 {
378 	u_int32_t	a[4];
379 	char		*cp;
380 
381 	bzero(sa, sizeof(*sa));
382 
383 	if ((cp = kgetenv(ev)) == NULL)
384 		return(1);
385 	if (ksscanf(cp, "%d.%d.%d.%d", &a[0], &a[1], &a[2], &a[3]) != 4)
386 		return(1);
387 	if (a[0] == 0 && a[1] == 0 && a[2] == 0 && a[3] == 0)
388 		return(1);
389 	/* XXX is this ordering correct? */
390 	sa->sin_addr.s_addr = (a[3] << 24) + (a[2] << 16) + (a[1] << 8) + a[0];
391 	sa->sin_len = sizeof(*sa);
392 	sa->sin_family = AF_INET;
393 	return(0);
394 }
395 
396 static int
397 hwaddr_to_sockaddr(char *ev, struct sockaddr_dl *sa)
398 {
399 	char		*cp;
400 	u_int32_t	a[6];
401 
402 	bzero(sa, sizeof(*sa));
403 	sa->sdl_len = sizeof(*sa);
404 	sa->sdl_family = AF_LINK;
405 	sa->sdl_type = IFT_ETHER;
406 	sa->sdl_alen = ETHER_ADDR_LEN;
407 	if ((cp = kgetenv(ev)) == NULL)
408 		return(1);
409 	if (ksscanf(cp, "%x:%x:%x:%x:%x:%x", &a[0], &a[1], &a[2], &a[3], &a[4], &a[5]) != 6)
410 		return(1);
411 	sa->sdl_data[0] = a[0];
412 	sa->sdl_data[1] = a[1];
413 	sa->sdl_data[2] = a[2];
414 	sa->sdl_data[3] = a[3];
415 	sa->sdl_data[4] = a[4];
416 	sa->sdl_data[5] = a[5];
417 	return(0);
418 }
419 
420 static int
421 decode_nfshandle(char *ev, u_char *fh)
422 {
423 	u_char	*cp;
424 	int	len, val;
425 
426 	if (((cp = kgetenv(ev)) == NULL) || (strlen(cp) < 2) || (*cp != 'X'))
427 		return(0);
428 	len = 0;
429 	cp++;
430 	for (;;) {
431 		if (*cp == 'X')
432 			return(len);
433 		if ((ksscanf(cp, "%2x", &val) != 1) || (val > 0xff))
434 			return(0);
435 		*(fh++) = val;
436 		len++;
437 		cp += 2;
438 		if (len > NFSX_V2FH)
439 		    return(0);
440 	}
441 }
442 
443 /*
444  * Populate the essential fields in the nfsv3_diskless structure.
445  *
446  * The loader is expected to export the following environment variables:
447  *
448  * boot.netif.ip		IP address on boot interface
449  * boot.netif.netmask		netmask on boot interface
450  * boot.netif.gateway		default gateway (optional)
451  * boot.netif.hwaddr		hardware address of boot interface
452  * boot.netif.name		name of boot interface (instead of hw addr)
453  * boot.nfsroot.server		IP address of root filesystem server
454  * boot.nfsroot.path		path of the root filesystem on server
455  * boot.nfsroot.nfshandle	NFS handle for root filesystem on server
456  */
457 static void
458 pxe_setup_nfsdiskless(void)
459 {
460 	struct nfs_diskless	*nd = &nfs_diskless;
461 	struct ifnet		*ifp;
462 	struct ifaddr		*ifa;
463 	struct sockaddr_dl	*sdl, ourdl;
464 	struct sockaddr_in	myaddr, netmask;
465 	char			*cp;
466 
467 	/* set up interface */
468 	if (inaddr_to_sockaddr("boot.netif.ip", &myaddr))
469 		return;
470 	if (inaddr_to_sockaddr("boot.netif.netmask", &netmask)) {
471 		kprintf("PXE: no netmask\n");
472 		return;
473 	}
474 	bcopy(&myaddr, &nd->myif.ifra_addr, sizeof(myaddr));
475 	bcopy(&myaddr, &nd->myif.ifra_broadaddr, sizeof(myaddr));
476 	((struct sockaddr_in *) &nd->myif.ifra_broadaddr)->sin_addr.s_addr =
477 		myaddr.sin_addr.s_addr | ~ netmask.sin_addr.s_addr;
478 	bcopy(&netmask, &nd->myif.ifra_mask, sizeof(netmask));
479 
480 	if ((cp = kgetenv("boot.netif.name")) != NULL) {
481 		TAILQ_FOREACH(ifp, &ifnet, if_link) {
482 			if (strcmp(cp, ifp->if_xname) == 0)
483 				break;
484 		}
485 		if (ifp)
486 			goto match_done;
487 		kprintf("PXE: cannot find interface %s\n", cp);
488 		return;
489 	}
490 
491 	if (hwaddr_to_sockaddr("boot.netif.hwaddr", &ourdl)) {
492 		kprintf("PXE: no hardware address\n");
493 		return;
494 	}
495 	ifa = NULL;
496 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
497 		struct ifaddr_container *ifac;
498 
499 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
500 			ifa = ifac->ifa;
501 
502 			if ((ifa->ifa_addr->sa_family == AF_LINK) &&
503 			    (sdl = ((struct sockaddr_dl *)ifa->ifa_addr))) {
504 				if ((sdl->sdl_type == ourdl.sdl_type) &&
505 				    (sdl->sdl_alen == ourdl.sdl_alen) &&
506 				    !bcmp(sdl->sdl_data + sdl->sdl_nlen,
507 					  ourdl.sdl_data + ourdl.sdl_nlen,
508 					  sdl->sdl_alen))
509 				    goto match_done;
510 			}
511 		}
512 	}
513 	kprintf("PXE: no interface\n");
514 	return;	/* no matching interface */
515 match_done:
516 	strlcpy(nd->myif.ifra_name, ifp->if_xname, sizeof(nd->myif.ifra_name));
517 
518 	/* set up gateway */
519 	inaddr_to_sockaddr("boot.netif.gateway", &nd->mygateway);
520 
521 	/* XXX set up swap? */
522 
523 	/* set up root mount */
524 	nd->root_args.rsize = 8192;		/* XXX tunable? */
525 	nd->root_args.wsize = 8192;
526 	nd->root_args.sotype = SOCK_STREAM;
527 	nd->root_args.flags = NFSMNT_WSIZE | NFSMNT_RSIZE | NFSMNT_RESVPORT;
528 	if (inaddr_to_sockaddr("boot.nfsroot.server", &nd->root_saddr)) {
529 		kprintf("PXE: no server\n");
530 		return;
531 	}
532 	nd->root_saddr.sin_port = htons(NFS_PORT);
533 
534 	/*
535 	 * A tftp-only loader may pass NFS path information without a
536 	 * root handle.  Generate a warning but continue configuring.
537 	 */
538 	if (decode_nfshandle("boot.nfsroot.nfshandle", &nd->root_fh[0]) == 0) {
539 		kprintf("PXE: Warning, no NFS handle passed from loader\n");
540 	}
541 	if ((cp = kgetenv("boot.nfsroot.path")) != NULL)
542 		strncpy(nd->root_hostnam, cp, MNAMELEN - 1);
543 
544 	nfs_diskless_valid = 1;
545 }
546 
547 #endif
548 #endif
549