proc->thread stage 4: rework the VFS and DEVICE subsystems to take thread
[dragonfly.git] / sys / i386 / i386 / sys_machdep.c
CommitLineData
984263bc
MD
1/*-
2 * Copyright (c) 1990 The Regents of the University of California.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 * from: @(#)sys_machdep.c 5.5 (Berkeley) 1/19/91
34 * $FreeBSD: src/sys/i386/i386/sys_machdep.c,v 1.47.2.3 2002/10/07 17:20:00 jhb Exp $
dadab5e9 35 * $DragonFly: src/sys/i386/i386/Attic/sys_machdep.c,v 1.5 2003/06/25 03:55:53 dillon Exp $
984263bc
MD
36 *
37 */
38
39#include "opt_user_ldt.h"
40
41#include <sys/param.h>
42#include <sys/systm.h>
43#include <sys/sysproto.h>
44#include <sys/malloc.h>
45#include <sys/proc.h>
46
47#include <vm/vm.h>
48#include <sys/lock.h>
49#include <vm/pmap.h>
50#include <vm/vm_map.h>
51#include <vm/vm_extern.h>
52
53#include <sys/user.h>
54
55#include <machine/cpu.h>
56#include <machine/ipl.h>
57#include <machine/pcb_ext.h> /* pcb.h included by sys/user.h */
58#include <machine/sysarch.h>
59#ifdef SMP
60#include <machine/smp.h>
61#endif
62
63#include <vm/vm_kern.h> /* for kernel_map */
64
65#define MAX_LD 8192
66#define LD_PER_PAGE 512
67#define NEW_MAX_LD(num) ((num + LD_PER_PAGE) & ~(LD_PER_PAGE-1))
68#define SIZE_FROM_LARGEST_LD(num) (NEW_MAX_LD(num) << 3)
69
70
71
72#ifdef USER_LDT
73static int i386_get_ldt __P((struct proc *, char *));
74static int i386_set_ldt __P((struct proc *, char *));
75#endif
76static int i386_get_ioperm __P((struct proc *, char *));
77static int i386_set_ioperm __P((struct proc *, char *));
78int i386_extend_pcb __P((struct proc *));
79
41c20dac
MD
80/*
81 * sysarch_args(int op, char *params)
82 */
984263bc
MD
83
84int
41c20dac 85sysarch(struct sysarch_args *uap)
984263bc 86{
41c20dac 87 struct proc *p = curproc;
984263bc
MD
88 int error = 0;
89
90 switch(uap->op) {
91#ifdef USER_LDT
92 case I386_GET_LDT:
93 error = i386_get_ldt(p, uap->parms);
94 break;
95
96 case I386_SET_LDT:
97 error = i386_set_ldt(p, uap->parms);
98 break;
99#endif
100 case I386_GET_IOPERM:
101 error = i386_get_ioperm(p, uap->parms);
102 break;
103 case I386_SET_IOPERM:
104 error = i386_set_ioperm(p, uap->parms);
105 break;
106 case I386_VM86:
107 error = vm86_sysarch(p, uap->parms);
108 break;
109 default:
110 error = EOPNOTSUPP;
111 break;
112 }
113 return (error);
114}
115
116int
117i386_extend_pcb(struct proc *p)
118{
119 int i, offset;
120 u_long *addr;
121 struct pcb_ext *ext;
122 struct soft_segment_descriptor ssd = {
123 0, /* segment base address (overwritten) */
124 ctob(IOPAGES + 1) - 1, /* length */
125 SDT_SYS386TSS, /* segment type */
126 0, /* priority level */
127 1, /* descriptor present */
128 0, 0,
129 0, /* default 32 size */
130 0 /* granularity */
131 };
132
133 ext = (struct pcb_ext *)kmem_alloc(kernel_map, ctob(IOPAGES+1));
134 if (ext == 0)
135 return (ENOMEM);
b7c628e4 136 p->p_thread->td_pcb->pcb_ext = ext;
984263bc 137 bzero(ext, sizeof(struct pcb_ext));
b7c628e4 138 ext->ext_tss.tss_esp0 = (unsigned)((char *)p->p_thread->td_pcb - 16);
984263bc
MD
139 ext->ext_tss.tss_ss0 = GSEL(GDATA_SEL, SEL_KPL);
140 /*
141 * The last byte of the i/o map must be followed by an 0xff byte.
142 * We arbitrarily allocate 16 bytes here, to keep the starting
143 * address on a doubleword boundary.
144 */
145 offset = PAGE_SIZE - 16;
146 ext->ext_tss.tss_ioopt =
147 (offset - ((unsigned)&ext->ext_tss - (unsigned)ext)) << 16;
148 ext->ext_iomap = (caddr_t)ext + offset;
149 ext->ext_vm86.vm86_intmap = (caddr_t)ext + offset - 32;
150
151 addr = (u_long *)ext->ext_vm86.vm86_intmap;
152 for (i = 0; i < (ctob(IOPAGES) + 32 + 16) / sizeof(u_long); i++)
153 *addr++ = ~0;
154
155 ssd.ssd_base = (unsigned)&ext->ext_tss;
156 ssd.ssd_limit -= ((unsigned)&ext->ext_tss - (unsigned)ext);
157 ssdtosd(&ssd, &ext->ext_tssd);
158
159 /* switch to the new TSS after syscall completes */
160 need_resched();
161
162 return 0;
163}
164
165static int
dadab5e9 166i386_set_ioperm(struct proc *p, char *args)
984263bc
MD
167{
168 int i, error;
169 struct i386_ioperm_args ua;
170 char *iomap;
171
172 if ((error = copyin(args, &ua, sizeof(struct i386_ioperm_args))) != 0)
173 return (error);
174
dadab5e9 175 if ((error = suser_cred(p->p_ucred, 0)) != 0)
984263bc
MD
176 return (error);
177 if (securelevel > 0)
178 return (EPERM);
179 /*
180 * XXX
181 * While this is restricted to root, we should probably figure out
182 * whether any other driver is using this i/o address, as so not to
183 * cause confusion. This probably requires a global 'usage registry'.
184 */
185
b7c628e4 186 if (p->p_thread->td_pcb->pcb_ext == 0)
984263bc
MD
187 if ((error = i386_extend_pcb(p)) != 0)
188 return (error);
b7c628e4 189 iomap = (char *)p->p_thread->td_pcb->pcb_ext->ext_iomap;
984263bc
MD
190
191 if (ua.start + ua.length > IOPAGES * PAGE_SIZE * NBBY)
192 return (EINVAL);
193
194 for (i = ua.start; i < ua.start + ua.length; i++) {
195 if (ua.enable)
196 iomap[i >> 3] &= ~(1 << (i & 7));
197 else
198 iomap[i >> 3] |= (1 << (i & 7));
199 }
200 return (error);
201}
202
203static int
204i386_get_ioperm(p, args)
205 struct proc *p;
206 char *args;
207{
208 int i, state, error;
209 struct i386_ioperm_args ua;
210 char *iomap;
211
212 if ((error = copyin(args, &ua, sizeof(struct i386_ioperm_args))) != 0)
213 return (error);
214 if (ua.start >= IOPAGES * PAGE_SIZE * NBBY)
215 return (EINVAL);
216
b7c628e4 217 if (p->p_thread->td_pcb->pcb_ext == 0) {
984263bc
MD
218 ua.length = 0;
219 goto done;
220 }
221
b7c628e4 222 iomap = (char *)p->p_thread->td_pcb->pcb_ext->ext_iomap;
984263bc
MD
223
224 i = ua.start;
225 state = (iomap[i >> 3] >> (i & 7)) & 1;
226 ua.enable = !state;
227 ua.length = 1;
228
229 for (i = ua.start + 1; i < IOPAGES * PAGE_SIZE * NBBY; i++) {
230 if (state != ((iomap[i >> 3] >> (i & 7)) & 1))
231 break;
232 ua.length++;
233 }
234
235done:
236 error = copyout(&ua, args, sizeof(struct i386_ioperm_args));
237 return (error);
238}
239
240#ifdef USER_LDT
241/*
242 * Update the GDT entry pointing to the LDT to point to the LDT of the
243 * current process. Do not staticize.
244 */
245void
246set_user_ldt(struct pcb *pcb)
247{
248 struct pcb_ldt *pcb_ldt;
249
b7c628e4 250 if (pcb != curthread->td_pcb)
984263bc
MD
251 return;
252
253 pcb_ldt = pcb->pcb_ldt;
254#ifdef SMP
255 gdt[cpuid * NGDT + GUSERLDT_SEL].sd = pcb_ldt->ldt_sd;
256#else
257 gdt[GUSERLDT_SEL].sd = pcb_ldt->ldt_sd;
258#endif
259 lldt(GSEL(GUSERLDT_SEL, SEL_KPL));
260 currentldt = GSEL(GUSERLDT_SEL, SEL_KPL);
261}
262
263struct pcb_ldt *
264user_ldt_alloc(struct pcb *pcb, int len)
265{
266 struct pcb_ldt *pcb_ldt, *new_ldt;
267
268 MALLOC(new_ldt, struct pcb_ldt *, sizeof(struct pcb_ldt),
269 M_SUBPROC, M_WAITOK);
270 if (new_ldt == NULL)
271 return NULL;
272
273 new_ldt->ldt_len = len = NEW_MAX_LD(len);
274 new_ldt->ldt_base = (caddr_t)kmem_alloc(kernel_map,
275 len * sizeof(union descriptor));
276 if (new_ldt->ldt_base == NULL) {
277 FREE(new_ldt, M_SUBPROC);
278 return NULL;
279 }
280 new_ldt->ldt_refcnt = 1;
281 new_ldt->ldt_active = 0;
282
283 gdt_segs[GUSERLDT_SEL].ssd_base = (unsigned)new_ldt->ldt_base;
284 gdt_segs[GUSERLDT_SEL].ssd_limit = len * sizeof(union descriptor) - 1;
285 ssdtosd(&gdt_segs[GUSERLDT_SEL], &new_ldt->ldt_sd);
286
287 if ((pcb_ldt = pcb->pcb_ldt)) {
288 if (len > pcb_ldt->ldt_len)
289 len = pcb_ldt->ldt_len;
290 bcopy(pcb_ldt->ldt_base, new_ldt->ldt_base,
291 len * sizeof(union descriptor));
292 } else {
293 bcopy(ldt, new_ldt->ldt_base, sizeof(ldt));
294 }
295 return new_ldt;
296}
297
298void
299user_ldt_free(struct pcb *pcb)
300{
301 struct pcb_ldt *pcb_ldt = pcb->pcb_ldt;
302
303 if (pcb_ldt == NULL)
304 return;
305
b7c628e4 306 if (pcb == curthread->td_pcb) {
984263bc
MD
307 lldt(_default_ldt);
308 currentldt = _default_ldt;
309 }
310
311 if (--pcb_ldt->ldt_refcnt == 0) {
312 kmem_free(kernel_map, (vm_offset_t)pcb_ldt->ldt_base,
313 pcb_ldt->ldt_len * sizeof(union descriptor));
314 FREE(pcb_ldt, M_SUBPROC);
315 }
316 pcb->pcb_ldt = NULL;
317}
318
319static int
320i386_get_ldt(p, args)
321 struct proc *p;
322 char *args;
323{
324 int error = 0;
b7c628e4 325 struct pcb *pcb = p->p_thread->td_pcb;
984263bc
MD
326 struct pcb_ldt *pcb_ldt = pcb->pcb_ldt;
327 int nldt, num;
328 union descriptor *lp;
329 int s;
330 struct i386_ldt_args ua, *uap = &ua;
331
332 if ((error = copyin(args, uap, sizeof(struct i386_ldt_args))) < 0)
333 return(error);
334
335#ifdef DEBUG
336 printf("i386_get_ldt: start=%d num=%d descs=%p\n",
337 uap->start, uap->num, (void *)uap->descs);
338#endif
339
340 /* verify range of LDTs exist */
341 if ((uap->start < 0) || (uap->num <= 0))
342 return(EINVAL);
343
344 s = splhigh();
345
346 if (pcb_ldt) {
347 nldt = pcb_ldt->ldt_len;
348 num = min(uap->num, nldt);
349 lp = &((union descriptor *)(pcb_ldt->ldt_base))[uap->start];
350 } else {
351 nldt = sizeof(ldt)/sizeof(ldt[0]);
352 num = min(uap->num, nldt);
353 lp = &ldt[uap->start];
354 }
355 if (uap->start + num > nldt) {
356 splx(s);
357 return(EINVAL);
358 }
359
360 error = copyout(lp, uap->descs, num * sizeof(union descriptor));
361 if (!error)
362 p->p_retval[0] = num;
363
364 splx(s);
365 return(error);
366}
367
368static int
369i386_set_ldt(p, args)
370 struct proc *p;
371 char *args;
372{
373 int error = 0, i, n;
374 int largest_ld;
b7c628e4 375 struct pcb *pcb = p->p_thread->td_pcb;
984263bc
MD
376 struct pcb_ldt *pcb_ldt = pcb->pcb_ldt;
377 union descriptor *descs;
378 int descs_size, s;
379 struct i386_ldt_args ua, *uap = &ua;
380
381 if ((error = copyin(args, uap, sizeof(struct i386_ldt_args))) < 0)
382 return(error);
383
384#ifdef DEBUG
385 printf("i386_set_ldt: start=%d num=%d descs=%p\n",
386 uap->start, uap->num, (void *)uap->descs);
387#endif
388
389 /* verify range of descriptors to modify */
390 if ((uap->start < 0) || (uap->start >= MAX_LD) || (uap->num < 0) ||
391 (uap->num > MAX_LD))
392 {
393 return(EINVAL);
394 }
395 largest_ld = uap->start + uap->num - 1;
396 if (largest_ld >= MAX_LD)
397 return(EINVAL);
398
399 /* allocate user ldt */
400 if (!pcb_ldt || largest_ld >= pcb_ldt->ldt_len) {
401 struct pcb_ldt *new_ldt = user_ldt_alloc(pcb, largest_ld);
402 if (new_ldt == NULL)
403 return ENOMEM;
404 if (pcb_ldt) {
405 pcb_ldt->ldt_sd = new_ldt->ldt_sd;
406 kmem_free(kernel_map, (vm_offset_t)pcb_ldt->ldt_base,
407 pcb_ldt->ldt_len * sizeof(union descriptor));
408 pcb_ldt->ldt_base = new_ldt->ldt_base;
409 pcb_ldt->ldt_len = new_ldt->ldt_len;
410 FREE(new_ldt, M_SUBPROC);
411 } else
412 pcb->pcb_ldt = pcb_ldt = new_ldt;
413#ifdef SMP
414 /* signal other cpus to reload ldt */
415 smp_rendezvous(NULL, (void (*)(void *))set_user_ldt, NULL, pcb);
416#else
417 set_user_ldt(pcb);
418#endif
419 }
420
421 descs_size = uap->num * sizeof(union descriptor);
422 descs = (union descriptor *)kmem_alloc(kernel_map, descs_size);
423 if (descs == NULL)
424 return (ENOMEM);
425 error = copyin(&uap->descs[0], descs, descs_size);
426 if (error) {
427 kmem_free(kernel_map, (vm_offset_t)descs, descs_size);
428 return (error);
429 }
430 /* Check descriptors for access violations */
431 for (i = 0, n = uap->start; i < uap->num; i++, n++) {
432 union descriptor *dp;
433 dp = &descs[i];
434
435 switch (dp->sd.sd_type) {
436 case SDT_SYSNULL: /* system null */
437 dp->sd.sd_p = 0;
438 break;
439 case SDT_SYS286TSS: /* system 286 TSS available */
440 case SDT_SYSLDT: /* system local descriptor table */
441 case SDT_SYS286BSY: /* system 286 TSS busy */
442 case SDT_SYSTASKGT: /* system task gate */
443 case SDT_SYS286IGT: /* system 286 interrupt gate */
444 case SDT_SYS286TGT: /* system 286 trap gate */
445 case SDT_SYSNULL2: /* undefined by Intel */
446 case SDT_SYS386TSS: /* system 386 TSS available */
447 case SDT_SYSNULL3: /* undefined by Intel */
448 case SDT_SYS386BSY: /* system 386 TSS busy */
449 case SDT_SYSNULL4: /* undefined by Intel */
450 case SDT_SYS386IGT: /* system 386 interrupt gate */
451 case SDT_SYS386TGT: /* system 386 trap gate */
452 case SDT_SYS286CGT: /* system 286 call gate */
453 case SDT_SYS386CGT: /* system 386 call gate */
454 /* I can't think of any reason to allow a user proc
455 * to create a segment of these types. They are
456 * for OS use only.
457 */
458 kmem_free(kernel_map, (vm_offset_t)descs, descs_size);
459 return EACCES;
460
461 /* memory segment types */
462 case SDT_MEMEC: /* memory execute only conforming */
463 case SDT_MEMEAC: /* memory execute only accessed conforming */
464 case SDT_MEMERC: /* memory execute read conforming */
465 case SDT_MEMERAC: /* memory execute read accessed conforming */
466 /* Must be "present" if executable and conforming. */
467 if (dp->sd.sd_p == 0) {
468 kmem_free(kernel_map, (vm_offset_t)descs,
469 descs_size);
470 return (EACCES);
471 }
472 break;
473 case SDT_MEMRO: /* memory read only */
474 case SDT_MEMROA: /* memory read only accessed */
475 case SDT_MEMRW: /* memory read write */
476 case SDT_MEMRWA: /* memory read write accessed */
477 case SDT_MEMROD: /* memory read only expand dwn limit */
478 case SDT_MEMRODA: /* memory read only expand dwn lim accessed */
479 case SDT_MEMRWD: /* memory read write expand dwn limit */
480 case SDT_MEMRWDA: /* memory read write expand dwn lim acessed */
481 case SDT_MEME: /* memory execute only */
482 case SDT_MEMEA: /* memory execute only accessed */
483 case SDT_MEMER: /* memory execute read */
484 case SDT_MEMERA: /* memory execute read accessed */
485 break;
486 default:
487 kmem_free(kernel_map, (vm_offset_t)descs, descs_size);
488 return(EINVAL);
489 /*NOTREACHED*/
490 }
491
492 /* Only user (ring-3) descriptors may be present. */
493 if ((dp->sd.sd_p != 0) && (dp->sd.sd_dpl != SEL_UPL)) {
494 kmem_free(kernel_map, (vm_offset_t)descs, descs_size);
495 return (EACCES);
496 }
497 }
498
499 s = splhigh();
500
501 /* Fill in range */
502 bcopy(descs,
503 &((union descriptor *)(pcb_ldt->ldt_base))[uap->start],
504 uap->num * sizeof(union descriptor));
505 p->p_retval[0] = uap->start;
506
507 splx(s);
508 kmem_free(kernel_map, (vm_offset_t)descs, descs_size);
509 return (0);
510}
511#endif /* USER_LDT */