2 * Copyright (c) 2003 Peter Wemm.
3 * Copyright (c) 1990 The Regents of the University of California.
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * 4. Neither the name of the University nor the names of its contributors
15 * may be used to endorse or promote products derived from this software
16 * without specific prior written permission.
18 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * from: @(#)sys_machdep.c 5.5 (Berkeley) 1/19/91
33 #include <sys/cdefs.h>
34 __FBSDID("$FreeBSD$");
36 #include "opt_capsicum.h"
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/capability.h>
41 #include <sys/kernel.h>
43 #include <sys/malloc.h>
44 #include <sys/mutex.h>
47 #include <sys/sysproto.h>
52 #include <vm/vm_kern.h> /* for kernel_map */
53 #include <vm/vm_extern.h>
55 #include <machine/frame.h>
56 #include <machine/md_var.h>
57 #include <machine/pcb.h>
58 #include <machine/specialreg.h>
59 #include <machine/sysarch.h>
60 #include <machine/tss.h>
61 #include <machine/vmparam.h>
63 #include <security/audit/audit.h>
67 int max_ldt_segment = 1024;
68 SYSCTL_INT(_machdep, OID_AUTO, max_ldt_segment, CTLFLAG_RDTUN,
70 "Maximum number of allowed LDT segments in the single address space");
73 max_ldt_segment_init(void *arg __unused)
76 TUNABLE_INT_FETCH("machdep.max_ldt_segment", &max_ldt_segment);
77 if (max_ldt_segment <= 0)
79 if (max_ldt_segment > MAX_LD)
80 max_ldt_segment = MAX_LD;
82 SYSINIT(maxldt, SI_SUB_VM_CONF, SI_ORDER_ANY, max_ldt_segment_init, NULL);
86 static void set_user_ldt_rv(struct vmspace *vmsp);
89 static void user_ldt_derefl(struct proc_ldt *pldt);
91 #ifndef _SYS_SYSPROTO_H_
99 sysarch_ldt(struct thread *td, struct sysarch_args *uap, int uap_space)
101 struct i386_ldt_args *largs, la;
102 struct user_segment_descriptor *lp;
106 * XXXKIB check that the BSM generation code knows to encode
109 AUDIT_ARG_CMD(uap->op);
110 if (uap_space == UIO_USERSPACE) {
111 error = copyin(uap->parms, &la, sizeof(struct i386_ldt_args));
116 largs = (struct i386_ldt_args *)uap->parms;
120 error = amd64_get_ldt(td, largs);
123 if (largs->descs != NULL && largs->num > max_ldt_segment)
125 set_pcb_flags(td->td_pcb, PCB_FULL_IRET);
126 if (largs->descs != NULL) {
127 lp = malloc(largs->num * sizeof(struct
128 user_segment_descriptor), M_TEMP, M_WAITOK);
129 error = copyin(largs->descs, lp, largs->num *
130 sizeof(struct user_segment_descriptor));
132 error = amd64_set_ldt(td, largs, lp);
135 error = amd64_set_ldt(td, largs, NULL);
143 update_gdt_gsbase(struct thread *td, uint32_t base)
145 struct user_segment_descriptor *sd;
149 set_pcb_flags(td->td_pcb, PCB_FULL_IRET);
151 sd = PCPU_GET(gs32p);
152 sd->sd_lobase = base & 0xffffff;
153 sd->sd_hibase = (base >> 24) & 0xff;
158 update_gdt_fsbase(struct thread *td, uint32_t base)
160 struct user_segment_descriptor *sd;
164 set_pcb_flags(td->td_pcb, PCB_FULL_IRET);
166 sd = PCPU_GET(fs32p);
167 sd->sd_lobase = base & 0xffffff;
168 sd->sd_hibase = (base >> 24) & 0xff;
175 register struct sysarch_args *uap;
178 struct pcb *pcb = curthread->td_pcb;
181 struct i386_ioperm_args iargs;
183 #ifdef CAPABILITY_MODE
185 * When adding new operations, add a new case statement here to
186 * explicitly indicate whether or not the operation is safe to
187 * perform in capability mode.
189 if (IN_CAPABILITY_MODE(td)) {
193 case I386_GET_IOPERM:
194 case I386_GET_FSBASE:
195 case I386_SET_FSBASE:
196 case I386_GET_GSBASE:
197 case I386_SET_GSBASE:
198 case AMD64_GET_FSBASE:
199 case AMD64_SET_FSBASE:
200 case AMD64_GET_GSBASE:
201 case AMD64_SET_GSBASE:
204 case I386_SET_IOPERM:
211 if (uap->op == I386_GET_LDT || uap->op == I386_SET_LDT)
212 return (sysarch_ldt(td, uap, UIO_USERSPACE));
214 * XXXKIB check that the BSM generation code knows to encode
217 AUDIT_ARG_CMD(uap->op);
219 case I386_GET_IOPERM:
220 case I386_SET_IOPERM:
221 if ((error = copyin(uap->parms, &iargs,
222 sizeof(struct i386_ioperm_args))) != 0)
230 case I386_GET_IOPERM:
231 error = amd64_get_ioperm(td, &iargs);
233 error = copyout(&iargs, uap->parms,
234 sizeof(struct i386_ioperm_args));
236 case I386_SET_IOPERM:
237 error = amd64_set_ioperm(td, &iargs);
239 case I386_GET_FSBASE:
240 i386base = pcb->pcb_fsbase;
241 error = copyout(&i386base, uap->parms, sizeof(i386base));
243 case I386_SET_FSBASE:
244 error = copyin(uap->parms, &i386base, sizeof(i386base));
246 pcb->pcb_fsbase = i386base;
247 td->td_frame->tf_fs = _ufssel;
248 update_gdt_fsbase(td, i386base);
251 case I386_GET_GSBASE:
252 i386base = pcb->pcb_gsbase;
253 error = copyout(&i386base, uap->parms, sizeof(i386base));
255 case I386_SET_GSBASE:
256 error = copyin(uap->parms, &i386base, sizeof(i386base));
258 pcb->pcb_gsbase = i386base;
259 td->td_frame->tf_gs = _ugssel;
260 update_gdt_gsbase(td, i386base);
263 case AMD64_GET_FSBASE:
264 error = copyout(&pcb->pcb_fsbase, uap->parms, sizeof(pcb->pcb_fsbase));
267 case AMD64_SET_FSBASE:
268 error = copyin(uap->parms, &a64base, sizeof(a64base));
270 if (a64base < VM_MAXUSER_ADDRESS) {
271 pcb->pcb_fsbase = a64base;
272 set_pcb_flags(pcb, PCB_FULL_IRET);
273 td->td_frame->tf_fs = _ufssel;
279 case AMD64_GET_GSBASE:
280 error = copyout(&pcb->pcb_gsbase, uap->parms, sizeof(pcb->pcb_gsbase));
283 case AMD64_SET_GSBASE:
284 error = copyin(uap->parms, &a64base, sizeof(a64base));
286 if (a64base < VM_MAXUSER_ADDRESS) {
287 pcb->pcb_gsbase = a64base;
288 set_pcb_flags(pcb, PCB_FULL_IRET);
289 td->td_frame->tf_gs = _ugssel;
303 amd64_set_ioperm(td, uap)
305 struct i386_ioperm_args *uap;
309 struct amd64tss *tssp;
310 struct system_segment_descriptor *tss_sd;
314 if ((error = priv_check(td, PRIV_IO)) != 0)
316 if ((error = securelevel_gt(td->td_ucred, 0)) != 0)
318 if (uap->start + uap->length > IOPAGES * PAGE_SIZE * NBBY)
323 * While this is restricted to root, we should probably figure out
324 * whether any other driver is using this i/o address, as so not to
325 * cause confusion. This probably requires a global 'usage registry'.
328 if (pcb->pcb_tssp == NULL) {
329 tssp = (struct amd64tss *)kmem_alloc(kernel_map,
333 iomap = (char *)&tssp[1];
334 addr = (u_long *)iomap;
335 for (i = 0; i < (ctob(IOPAGES) + 1) / sizeof(u_long); i++)
338 /* Takes care of tss_rsp0. */
339 memcpy(tssp, &common_tss[PCPU_GET(cpuid)],
340 sizeof(struct amd64tss));
341 tssp->tss_iobase = sizeof(*tssp);
342 pcb->pcb_tssp = tssp;
343 tss_sd = PCPU_GET(tss);
344 tss_sd->sd_lobase = (u_long)tssp & 0xffffff;
345 tss_sd->sd_hibase = ((u_long)tssp >> 24) & 0xfffffffffful;
346 tss_sd->sd_type = SDT_SYSTSS;
347 ltr(GSEL(GPROC0_SEL, SEL_KPL));
348 PCPU_SET(tssp, tssp);
351 iomap = (char *)&pcb->pcb_tssp[1];
352 for (i = uap->start; i < uap->start + uap->length; i++) {
354 iomap[i >> 3] &= ~(1 << (i & 7));
356 iomap[i >> 3] |= (1 << (i & 7));
362 amd64_get_ioperm(td, uap)
364 struct i386_ioperm_args *uap;
369 if (uap->start >= IOPAGES * PAGE_SIZE * NBBY)
371 if (td->td_pcb->pcb_tssp == NULL) {
376 iomap = (char *)&td->td_pcb->pcb_tssp[1];
379 state = (iomap[i >> 3] >> (i & 7)) & 1;
380 uap->enable = !state;
383 for (i = uap->start + 1; i < IOPAGES * PAGE_SIZE * NBBY; i++) {
384 if (state != ((iomap[i >> 3] >> (i & 7)) & 1))
394 * Update the GDT entry pointing to the LDT to point to the LDT of the
398 set_user_ldt(struct mdproc *mdp)
402 *PCPU_GET(ldt) = mdp->md_ldt_sd;
403 lldt(GSEL(GUSERLDT_SEL, SEL_KPL));
410 set_user_ldt_rv(struct vmspace *vmsp)
415 if (vmsp != td->td_proc->p_vmspace)
418 set_user_ldt(&td->td_proc->p_md);
424 user_ldt_alloc(struct proc *p, int force)
426 struct proc_ldt *pldt, *new_ldt;
428 struct soft_segment_descriptor sldt;
430 mtx_assert(&dt_lock, MA_OWNED);
432 if (!force && mdp->md_ldt != NULL)
433 return (mdp->md_ldt);
434 mtx_unlock(&dt_lock);
435 new_ldt = malloc(sizeof(struct proc_ldt), M_SUBPROC, M_WAITOK);
436 new_ldt->ldt_base = (caddr_t)kmem_alloc(kernel_map,
437 max_ldt_segment * sizeof(struct user_segment_descriptor));
438 if (new_ldt->ldt_base == NULL) {
439 FREE(new_ldt, M_SUBPROC);
443 new_ldt->ldt_refcnt = 1;
444 sldt.ssd_base = (uint64_t)new_ldt->ldt_base;
445 sldt.ssd_limit = max_ldt_segment *
446 sizeof(struct user_segment_descriptor) - 1;
447 sldt.ssd_type = SDT_SYSLDT;
448 sldt.ssd_dpl = SEL_KPL;
455 if (pldt != NULL && !force) {
456 kmem_free(kernel_map, (vm_offset_t)new_ldt->ldt_base,
457 max_ldt_segment * sizeof(struct user_segment_descriptor));
458 free(new_ldt, M_SUBPROC);
463 bcopy(pldt->ldt_base, new_ldt->ldt_base, max_ldt_segment *
464 sizeof(struct user_segment_descriptor));
465 user_ldt_derefl(pldt);
467 ssdtosyssd(&sldt, &p->p_md.md_ldt_sd);
468 atomic_store_rel_ptr((volatile uintptr_t *)&mdp->md_ldt,
473 return (mdp->md_ldt);
477 user_ldt_free(struct thread *td)
479 struct proc *p = td->td_proc;
480 struct mdproc *mdp = &p->p_md;
481 struct proc_ldt *pldt;
483 mtx_assert(&dt_lock, MA_OWNED);
484 if ((pldt = mdp->md_ldt) == NULL) {
485 mtx_unlock(&dt_lock);
490 bzero(&mdp->md_ldt_sd, sizeof(mdp->md_ldt_sd));
492 lldt(GSEL(GNULL_SEL, SEL_KPL));
493 user_ldt_deref(pldt);
497 user_ldt_derefl(struct proc_ldt *pldt)
500 if (--pldt->ldt_refcnt == 0) {
501 kmem_free(kernel_map, (vm_offset_t)pldt->ldt_base,
502 max_ldt_segment * sizeof(struct user_segment_descriptor));
503 free(pldt, M_SUBPROC);
508 user_ldt_deref(struct proc_ldt *pldt)
511 mtx_assert(&dt_lock, MA_OWNED);
512 user_ldt_derefl(pldt);
513 mtx_unlock(&dt_lock);
517 * Note for the authors of compat layers (linux, etc): copyout() in
518 * the function below is not a problem since it presents data in
519 * arch-specific format (i.e. i386-specific in this case), not in
520 * the OS-specific one.
523 amd64_get_ldt(td, uap)
525 struct i386_ldt_args *uap;
528 struct proc_ldt *pldt;
530 struct user_segment_descriptor *lp;
533 printf("amd64_get_ldt: start=%d num=%d descs=%p\n",
534 uap->start, uap->num, (void *)uap->descs);
537 if ((pldt = td->td_proc->p_md.md_ldt) != NULL) {
538 lp = &((struct user_segment_descriptor *)(pldt->ldt_base))
540 num = min(uap->num, max_ldt_segment);
544 if ((uap->start > (unsigned int)max_ldt_segment) ||
545 ((unsigned int)num > (unsigned int)max_ldt_segment) ||
546 ((unsigned int)(uap->start + num) > (unsigned int)max_ldt_segment))
549 error = copyout(lp, uap->descs, num *
550 sizeof(struct user_segment_descriptor));
552 td->td_retval[0] = num;
558 amd64_set_ldt(td, uap, descs)
560 struct i386_ldt_args *uap;
561 struct user_segment_descriptor *descs;
565 struct mdproc *mdp = &td->td_proc->p_md;
566 struct proc_ldt *pldt;
567 struct user_segment_descriptor *dp;
571 printf("amd64_set_ldt: start=%d num=%d descs=%p\n",
572 uap->start, uap->num, (void *)uap->descs);
575 set_pcb_flags(td->td_pcb, PCB_FULL_IRET);
578 /* Free descriptors */
579 if (uap->start == 0 && uap->num == 0)
580 uap->num = max_ldt_segment;
583 if ((pldt = mdp->md_ldt) == NULL ||
584 uap->start >= max_ldt_segment)
586 largest_ld = uap->start + uap->num;
587 if (largest_ld > max_ldt_segment)
588 largest_ld = max_ldt_segment;
589 i = largest_ld - uap->start;
591 bzero(&((struct user_segment_descriptor *)(pldt->ldt_base))
592 [uap->start], sizeof(struct user_segment_descriptor) * i);
593 mtx_unlock(&dt_lock);
597 if (!(uap->start == LDT_AUTO_ALLOC && uap->num == 1)) {
598 /* verify range of descriptors to modify */
599 largest_ld = uap->start + uap->num;
600 if (uap->start >= max_ldt_segment ||
601 largest_ld > max_ldt_segment)
605 /* Check descriptors for access violations */
606 for (i = 0; i < uap->num; i++) {
609 switch (dp->sd_type) {
610 case SDT_SYSNULL: /* system null */
628 /* I can't think of any reason to allow a user proc
629 * to create a segment of these types. They are
635 /* memory segment types */
636 case SDT_MEMEC: /* memory execute only conforming */
637 case SDT_MEMEAC: /* memory execute only accessed conforming */
638 case SDT_MEMERC: /* memory execute read conforming */
639 case SDT_MEMERAC: /* memory execute read accessed conforming */
640 /* Must be "present" if executable and conforming. */
644 case SDT_MEMRO: /* memory read only */
645 case SDT_MEMROA: /* memory read only accessed */
646 case SDT_MEMRW: /* memory read write */
647 case SDT_MEMRWA: /* memory read write accessed */
648 case SDT_MEMROD: /* memory read only expand dwn limit */
649 case SDT_MEMRODA: /* memory read only expand dwn lim accessed */
650 case SDT_MEMRWD: /* memory read write expand dwn limit */
651 case SDT_MEMRWDA: /* memory read write expand dwn lim acessed */
652 case SDT_MEME: /* memory execute only */
653 case SDT_MEMEA: /* memory execute only accessed */
654 case SDT_MEMER: /* memory execute read */
655 case SDT_MEMERA: /* memory execute read accessed */
662 /* Only user (ring-3) descriptors may be present. */
663 if ((dp->sd_p != 0) && (dp->sd_dpl != SEL_UPL))
667 if (uap->start == LDT_AUTO_ALLOC && uap->num == 1) {
668 /* Allocate a free slot */
670 pldt = user_ldt_alloc(p, 0);
672 mtx_unlock(&dt_lock);
677 * start scanning a bit up to leave room for NVidia and
678 * Wine, which still user the "Blat" method of allocation.
681 dp = &((struct user_segment_descriptor *)(pldt->ldt_base))[i];
682 for (; i < max_ldt_segment; ++i, ++dp) {
683 if (dp->sd_type == SDT_SYSNULL)
686 if (i >= max_ldt_segment) {
687 mtx_unlock(&dt_lock);
691 error = amd64_set_ldt_data(td, i, 1, descs);
692 mtx_unlock(&dt_lock);
694 largest_ld = uap->start + uap->num;
695 if (largest_ld > max_ldt_segment)
698 if (user_ldt_alloc(p, 0) != NULL) {
699 error = amd64_set_ldt_data(td, uap->start, uap->num,
702 mtx_unlock(&dt_lock);
705 td->td_retval[0] = uap->start;
710 amd64_set_ldt_data(struct thread *td, int start, int num,
711 struct user_segment_descriptor *descs)
713 struct mdproc *mdp = &td->td_proc->p_md;
714 struct proc_ldt *pldt = mdp->md_ldt;
716 mtx_assert(&dt_lock, MA_OWNED);
720 &((struct user_segment_descriptor *)(pldt->ldt_base))[start],
721 num * sizeof(struct user_segment_descriptor));