2 * Copyright (c) 1990 The Regents of the University of California.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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.
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
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 $
35 * $DragonFly: src/sys/platform/pc32/i386/sys_machdep.c,v 1.32 2008/01/06 16:55:53 swildner Exp $
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/sysproto.h>
42 #include <sys/malloc.h>
43 #include <sys/thread.h>
46 #include <sys/thread.h>
47 #include <sys/memrange.h>
52 #include <vm/vm_map.h>
53 #include <vm/vm_extern.h>
57 #include <machine/cpu.h>
58 #include <machine/pcb_ext.h> /* pcb.h included by sys/user.h */
59 #include <machine/sysarch.h>
60 #include <machine/smp.h>
61 #include <machine/globaldata.h> /* mdcpu */
63 #include <vm/vm_kern.h> /* for kernel_map */
64 #include <sys/thread2.h>
67 #define LD_PER_PAGE 512
68 #define NEW_MAX_LD(num) ((num + LD_PER_PAGE) & ~(LD_PER_PAGE-1))
69 #define SIZE_FROM_LARGEST_LD(num) (NEW_MAX_LD(num) << 3)
73 static int ki386_get_ldt(struct lwp *, char *, int *);
74 static int ki386_set_ldt(struct lwp *, char *, int *);
75 static int ki386_get_ioperm(struct lwp *, char *);
76 static int ki386_set_ioperm(struct lwp *, char *);
77 static int check_descs(union descriptor *, int);
78 int i386_extend_pcb(struct lwp *);
81 * sysarch_args(int op, char *params)
85 sys_sysarch(struct sysarch_args *uap)
87 struct lwp *lp = curthread->td_lwp;
92 error = ki386_get_ldt(lp, uap->parms, &uap->sysmsg_result);
95 error = ki386_set_ldt(lp, uap->parms, &uap->sysmsg_result);
98 error = ki386_get_ioperm(lp, uap->parms);
100 case I386_SET_IOPERM:
101 error = ki386_set_ioperm(lp, uap->parms);
104 error = vm86_sysarch(lp, uap->parms);
114 i386_extend_pcb(struct lwp *lp)
119 struct soft_segment_descriptor ssd = {
120 0, /* segment base address (overwritten) */
121 ctob(IOPAGES + 1) - 1, /* length */
122 SDT_SYS386TSS, /* segment type */
123 0, /* priority level */
124 1, /* descriptor present */
126 0, /* default 32 size */
130 ext = (struct pcb_ext *)kmem_alloc(&kernel_map, ctob(IOPAGES+1));
133 bzero(ext, sizeof(struct pcb_ext));
134 ext->ext_tss.tss_esp0 = (unsigned)((char *)lp->lwp_thread->td_pcb - 16);
135 ext->ext_tss.tss_ss0 = GSEL(GDATA_SEL, SEL_KPL);
137 * The last byte of the i/o map must be followed by an 0xff byte.
138 * We arbitrarily allocate 16 bytes here, to keep the starting
139 * address on a doubleword boundary.
141 offset = PAGE_SIZE - 16;
142 ext->ext_tss.tss_ioopt =
143 (offset - ((unsigned)&ext->ext_tss - (unsigned)ext)) << 16;
144 ext->ext_iomap = (caddr_t)ext + offset;
145 ext->ext_vm86.vm86_intmap = (caddr_t)ext + offset - 32;
147 addr = (u_long *)ext->ext_vm86.vm86_intmap;
148 for (i = 0; i < (ctob(IOPAGES) + 32 + 16) / sizeof(u_long); i++)
151 ssd.ssd_base = (unsigned)&ext->ext_tss;
152 ssd.ssd_limit -= ((unsigned)&ext->ext_tss - (unsigned)ext);
153 ssdtosd(&ssd, &ext->ext_tssd);
156 * Put the new TSS where the switch code can find it. Do
157 * a forced switch to ourself to activate it.
160 lp->lwp_thread->td_pcb->pcb_ext = ext;
161 lp->lwp_thread->td_switch(lp->lwp_thread);
168 ki386_set_ioperm(struct lwp *lp, char *args)
171 struct i386_ioperm_args ua;
174 if ((error = copyin(args, &ua, sizeof(struct i386_ioperm_args))) != 0)
177 if ((error = priv_check_cred(lp->lwp_proc->p_ucred, PRIV_ROOT, 0)) != 0)
183 * While this is restricted to root, we should probably figure out
184 * whether any other driver is using this i/o address, as so not to
185 * cause confusion. This probably requires a global 'usage registry'.
188 if (lp->lwp_thread->td_pcb->pcb_ext == 0)
189 if ((error = i386_extend_pcb(lp)) != 0)
191 iomap = (char *)lp->lwp_thread->td_pcb->pcb_ext->ext_iomap;
193 if (ua.start + ua.length > IOPAGES * PAGE_SIZE * NBBY)
196 for (i = ua.start; i < ua.start + ua.length; i++) {
198 iomap[i >> 3] &= ~(1 << (i & 7));
200 iomap[i >> 3] |= (1 << (i & 7));
206 ki386_get_ioperm(struct lwp *lp, char *args)
209 struct i386_ioperm_args ua;
212 if ((error = copyin(args, &ua, sizeof(struct i386_ioperm_args))) != 0)
214 if (ua.start >= IOPAGES * PAGE_SIZE * NBBY)
217 if (lp->lwp_thread->td_pcb->pcb_ext == 0) {
222 iomap = (char *)lp->lwp_thread->td_pcb->pcb_ext->ext_iomap;
225 state = (iomap[i >> 3] >> (i & 7)) & 1;
229 for (i = ua.start + 1; i < IOPAGES * PAGE_SIZE * NBBY; i++) {
230 if (state != ((iomap[i >> 3] >> (i & 7)) & 1))
236 error = copyout(&ua, args, sizeof(struct i386_ioperm_args));
241 * Update the TLS entries for the process. Used by assembly, do not staticize.
243 * Must be called from a critical section (else an interrupt thread preemption
244 * may cause %gs to fault). Normally called from the low level swtch.s code.
251 struct thread *td = curthread;
254 int off = GTLS_START + mycpu->gd_cpuid * NGDT;
256 const int off = GTLS_START;
258 for (i = 0; i < NGTLS; ++i)
259 gdt[off + i].sd = td->td_tls.tls[i];
265 set_user_ldt_cpusync(struct lwkt_cpusync *cmd)
267 set_user_ldt(cmd->cs_data);
272 * Update the GDT entry pointing to the LDT to point to the LDT of the
273 * current process. Used by assembly, do not staticize.
275 * Must be called from a critical section (else an interrupt thread preemption
276 * may cause %gs to fault). Normally called from the low level swtch.s code.
279 set_user_ldt(struct pcb *pcb)
281 struct pcb_ldt *pcb_ldt;
283 if (pcb != curthread->td_pcb)
286 pcb_ldt = pcb->pcb_ldt;
288 gdt[mycpu->gd_cpuid * NGDT + GUSERLDT_SEL].sd = pcb_ldt->ldt_sd;
290 gdt[GUSERLDT_SEL].sd = pcb_ldt->ldt_sd;
292 lldt(GSEL(GUSERLDT_SEL, SEL_KPL));
293 mdcpu->gd_currentldt = GSEL(GUSERLDT_SEL, SEL_KPL);
297 user_ldt_alloc(struct pcb *pcb, int len)
299 struct pcb_ldt *pcb_ldt, *new_ldt;
301 MALLOC(new_ldt, struct pcb_ldt *, sizeof(struct pcb_ldt),
302 M_SUBPROC, M_WAITOK);
304 new_ldt->ldt_len = len = NEW_MAX_LD(len);
305 new_ldt->ldt_base = (caddr_t)kmem_alloc(&kernel_map,
306 len * sizeof(union descriptor));
307 if (new_ldt->ldt_base == NULL) {
308 FREE(new_ldt, M_SUBPROC);
311 new_ldt->ldt_refcnt = 1;
312 new_ldt->ldt_active = 0;
314 gdt_segs[GUSERLDT_SEL].ssd_base = (unsigned)new_ldt->ldt_base;
315 gdt_segs[GUSERLDT_SEL].ssd_limit = len * sizeof(union descriptor) - 1;
316 ssdtosd(&gdt_segs[GUSERLDT_SEL], &new_ldt->ldt_sd);
318 if ((pcb_ldt = pcb->pcb_ldt)) {
319 if (len > pcb_ldt->ldt_len)
320 len = pcb_ldt->ldt_len;
321 bcopy(pcb_ldt->ldt_base, new_ldt->ldt_base,
322 len * sizeof(union descriptor));
324 bcopy(ldt, new_ldt->ldt_base, sizeof(ldt));
330 user_ldt_free(struct pcb *pcb)
332 struct pcb_ldt *pcb_ldt = pcb->pcb_ldt;
338 if (pcb == curthread->td_pcb) {
340 mdcpu->gd_currentldt = _default_ldt;
345 if (--pcb_ldt->ldt_refcnt == 0) {
346 kmem_free(&kernel_map, (vm_offset_t)pcb_ldt->ldt_base,
347 pcb_ldt->ldt_len * sizeof(union descriptor));
348 FREE(pcb_ldt, M_SUBPROC);
353 ki386_get_ldt(struct lwp *lwp, char *args, int *res)
356 struct pcb *pcb = lwp->lwp_thread->td_pcb;
357 struct pcb_ldt *pcb_ldt = pcb->pcb_ldt;
358 unsigned int nldt, num;
359 union descriptor *lp;
360 struct i386_ldt_args ua, *uap = &ua;
362 if ((error = copyin(args, uap, sizeof(struct i386_ldt_args))) < 0)
366 kprintf("ki386_get_ldt: start=%d num=%d descs=%p\n",
367 uap->start, uap->num, (void *)uap->descs);
373 nldt = (unsigned int)pcb_ldt->ldt_len;
374 num = min(uap->num, nldt);
375 lp = &((union descriptor *)(pcb_ldt->ldt_base))[uap->start];
377 nldt = (unsigned int)(sizeof(ldt) / sizeof(ldt[0]));
378 num = min(uap->num, nldt);
379 lp = &ldt[uap->start];
383 * note: uap->(args), num, and nldt are unsigned. nldt and num
384 * are limited in scope, but uap->start can be anything.
386 if (uap->start > nldt || uap->start + num > nldt) {
391 error = copyout(lp, uap->descs, num * sizeof(union descriptor));
399 ki386_set_ldt(struct lwp *lp, char *args, int *res)
403 struct pcb *pcb = lp->lwp_thread->td_pcb;
404 struct pcb_ldt *pcb_ldt = pcb->pcb_ldt;
405 union descriptor *descs;
407 struct i386_ldt_args ua, *uap = &ua;
409 if ((error = copyin(args, uap, sizeof(struct i386_ldt_args))) < 0)
413 kprintf("ki386_set_ldt: start=%d num=%d descs=%p\n",
414 uap->start, uap->num, (void *)uap->descs);
417 /* verify range of descriptors to modify */
418 if ((uap->start < 0) || (uap->start >= MAX_LD) || (uap->num < 0) ||
423 largest_ld = uap->start + uap->num - 1;
424 if (largest_ld >= MAX_LD)
427 /* allocate user ldt */
428 if (!pcb_ldt || largest_ld >= pcb_ldt->ldt_len) {
429 struct pcb_ldt *new_ldt = user_ldt_alloc(pcb, largest_ld);
433 pcb_ldt->ldt_sd = new_ldt->ldt_sd;
434 kmem_free(&kernel_map, (vm_offset_t)pcb_ldt->ldt_base,
435 pcb_ldt->ldt_len * sizeof(union descriptor));
436 pcb_ldt->ldt_base = new_ldt->ldt_base;
437 pcb_ldt->ldt_len = new_ldt->ldt_len;
438 FREE(new_ldt, M_SUBPROC);
440 pcb->pcb_ldt = pcb_ldt = new_ldt;
443 * Since the LDT may be shared, we must signal other cpus to
444 * reload it. XXX we need to track which cpus might be
445 * using the shared ldt and only signal those.
448 lwkt_cpusync_simple(-1, set_user_ldt_cpusync, pcb);
454 descs_size = uap->num * sizeof(union descriptor);
455 descs = (union descriptor *)kmem_alloc(&kernel_map, descs_size);
458 error = copyin(&uap->descs[0], descs, descs_size);
460 kmem_free(&kernel_map, (vm_offset_t)descs, descs_size);
463 /* Check descriptors for access violations */
464 error = check_descs(descs, uap->num);
466 kmem_free(&kernel_map, (vm_offset_t)descs, descs_size);
471 * Fill in the actual ldt entries. Since %fs or %gs might point to
472 * one of these entries a critical section is required to prevent an
473 * interrupt thread from preempting us, switch back, and faulting
474 * on the load of %fs due to a half-formed descriptor.
478 &((union descriptor *)(pcb_ldt->ldt_base))[uap->start],
479 uap->num * sizeof(union descriptor));
483 kmem_free(&kernel_map, (vm_offset_t)descs, descs_size);
488 check_descs(union descriptor *descs, int num)
492 /* Check descriptors for access violations */
493 for (i = 0; i < num; i++) {
494 union descriptor *dp;
497 switch (dp->sd.sd_type) {
498 case SDT_SYSNULL: /* system null */
501 case SDT_SYS286TSS: /* system 286 TSS available */
502 case SDT_SYSLDT: /* system local descriptor table */
503 case SDT_SYS286BSY: /* system 286 TSS busy */
504 case SDT_SYSTASKGT: /* system task gate */
505 case SDT_SYS286IGT: /* system 286 interrupt gate */
506 case SDT_SYS286TGT: /* system 286 trap gate */
507 case SDT_SYSNULL2: /* undefined by Intel */
508 case SDT_SYS386TSS: /* system 386 TSS available */
509 case SDT_SYSNULL3: /* undefined by Intel */
510 case SDT_SYS386BSY: /* system 386 TSS busy */
511 case SDT_SYSNULL4: /* undefined by Intel */
512 case SDT_SYS386IGT: /* system 386 interrupt gate */
513 case SDT_SYS386TGT: /* system 386 trap gate */
514 case SDT_SYS286CGT: /* system 286 call gate */
515 case SDT_SYS386CGT: /* system 386 call gate */
516 /* I can't think of any reason to allow a user proc
517 * to create a segment of these types. They are
522 /* memory segment types */
523 case SDT_MEMEC: /* memory execute only conforming */
524 case SDT_MEMEAC: /* memory execute only accessed conforming */
525 case SDT_MEMERC: /* memory execute read conforming */
526 case SDT_MEMERAC: /* memory execute read accessed conforming */
527 /* Must be "present" if executable and conforming. */
528 if (dp->sd.sd_p == 0)
531 case SDT_MEMRO: /* memory read only */
532 case SDT_MEMROA: /* memory read only accessed */
533 case SDT_MEMRW: /* memory read write */
534 case SDT_MEMRWA: /* memory read write accessed */
535 case SDT_MEMROD: /* memory read only expand dwn limit */
536 case SDT_MEMRODA: /* memory read only expand dwn lim accessed */
537 case SDT_MEMRWD: /* memory read write expand dwn limit */
538 case SDT_MEMRWDA: /* memory read write expand dwn lim acessed */
539 case SDT_MEME: /* memory execute only */
540 case SDT_MEMEA: /* memory execute only accessed */
541 case SDT_MEMER: /* memory execute read */
542 case SDT_MEMERA: /* memory execute read accessed */
549 /* Only user (ring-3) descriptors may be present. */
550 if ((dp->sd.sd_p != 0) && (dp->sd.sd_dpl != SEL_UPL))
557 * Called when /dev/io is opened
562 curthread->td_lwp->lwp_md.md_regs->tf_eflags |= PSL_IOPL;
567 * Called when /dev/io is closed
572 curthread->td_lwp->lwp_md.md_regs->tf_eflags &= ~PSL_IOPL;