1 /*-
2 * SPDX-License-Identifier: BSD-4-Clause
3 *
4 * Copyright (c) 1982, 1986 The Regents of the University of California.
5 * Copyright (c) 1989, 1990 William Jolitz
6 * Copyright (c) 1994 John Dyson
7 * All rights reserved.
8 *
9 * This code is derived from software contributed to Berkeley by
10 * the Systems Programming Group of the University of Utah Computer
11 * Science Department, and William Jolitz.
12 *
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
15 * are met:
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
21 * 3. All advertising materials mentioning features or use of this software
22 * must display the following acknowledgement:
23 * This product includes software developed by the University of
24 * California, Berkeley and its contributors.
25 * 4. Neither the name of the University nor the names of its contributors
26 * may be used to endorse or promote products derived from this software
27 * without specific prior written permission.
28 *
29 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
30 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
31 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
32 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
33 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
34 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
35 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
36 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
37 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
38 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
39 * SUCH DAMAGE.
40 *
41 * from: @(#)vm_machdep.c 7.3 (Berkeley) 5/13/91
42 * Utah $Hdr: vm_machdep.c 1.16.1.1 89/06/23$
43 */
44
45 #include <sys/cdefs.h>
46 __FBSDID("$FreeBSD$");
47
48 #include "opt_isa.h"
49 #include "opt_cpu.h"
50
51 #include <sys/param.h>
52 #include <sys/systm.h>
53 #include <sys/bio.h>
54 #include <sys/buf.h>
55 #include <sys/kernel.h>
56 #include <sys/ktr.h>
57 #include <sys/lock.h>
58 #include <sys/malloc.h>
59 #include <sys/mbuf.h>
60 #include <sys/mutex.h>
61 #include <sys/priv.h>
62 #include <sys/proc.h>
63 #include <sys/procctl.h>
64 #include <sys/smp.h>
65 #include <sys/sysctl.h>
66 #include <sys/sysent.h>
67 #include <sys/unistd.h>
68 #include <sys/vnode.h>
69 #include <sys/vmmeter.h>
70 #include <sys/wait.h>
71
72 #include <machine/cpu.h>
73 #include <machine/md_var.h>
74 #include <machine/pcb.h>
75 #include <machine/smp.h>
76 #include <machine/specialreg.h>
77 #include <machine/tss.h>
78
79 #include <vm/vm.h>
80 #include <vm/vm_extern.h>
81 #include <vm/vm_kern.h>
82 #include <vm/vm_page.h>
83 #include <vm/vm_map.h>
84 #include <vm/vm_param.h>
85
86 _Static_assert(OFFSETOF_MONITORBUF == offsetof(struct pcpu, pc_monitorbuf),
87 "OFFSETOF_MONITORBUF does not correspond with offset of pc_monitorbuf.");
88
89 void
set_top_of_stack_td(struct thread * td)90 set_top_of_stack_td(struct thread *td)
91 {
92 td->td_md.md_stack_base = td->td_kstack +
93 td->td_kstack_pages * PAGE_SIZE;
94 }
95
96 struct savefpu *
get_pcb_user_save_td(struct thread * td)97 get_pcb_user_save_td(struct thread *td)
98 {
99 KASSERT(((vm_offset_t)td->td_md.md_usr_fpu_save %
100 XSAVE_AREA_ALIGN) == 0,
101 ("Unaligned pcb_user_save area ptr %p td %p",
102 td->td_md.md_usr_fpu_save, td));
103 return (td->td_md.md_usr_fpu_save);
104 }
105
106 struct pcb *
get_pcb_td(struct thread * td)107 get_pcb_td(struct thread *td)
108 {
109
110 return (&td->td_md.md_pcb);
111 }
112
113 struct savefpu *
get_pcb_user_save_pcb(struct pcb * pcb)114 get_pcb_user_save_pcb(struct pcb *pcb)
115 {
116 struct thread *td;
117
118 td = __containerof(pcb, struct thread, td_md.md_pcb);
119 return (get_pcb_user_save_td(td));
120 }
121
122 void *
alloc_fpusave(int flags)123 alloc_fpusave(int flags)
124 {
125 void *res;
126 struct savefpu_ymm *sf;
127
128 res = malloc(cpu_max_ext_state_size, M_DEVBUF, flags);
129 if (use_xsave) {
130 sf = (struct savefpu_ymm *)res;
131 bzero(&sf->sv_xstate.sx_hd, sizeof(sf->sv_xstate.sx_hd));
132 sf->sv_xstate.sx_hd.xstate_bv = xsave_mask;
133 }
134 return (res);
135 }
136
137 /*
138 * Common code shared between cpu_fork() and cpu_copy_thread() for
139 * initializing a thread.
140 */
141 static void
copy_thread(struct thread * td1,struct thread * td2)142 copy_thread(struct thread *td1, struct thread *td2)
143 {
144 struct pcb *pcb2;
145
146 pcb2 = td2->td_pcb;
147
148 /* Ensure that td1's pcb is up to date for user threads. */
149 if ((td2->td_pflags & TDP_KTHREAD) == 0) {
150 MPASS(td1 == curthread);
151 fpuexit(td1);
152 update_pcb_bases(td1->td_pcb);
153 }
154
155 /* Copy td1's pcb */
156 bcopy(td1->td_pcb, pcb2, sizeof(*pcb2));
157
158 /* Properly initialize pcb_save */
159 pcb2->pcb_save = get_pcb_user_save_pcb(pcb2);
160
161 /* Kernel threads start with clean FPU and segment bases. */
162 if ((td2->td_pflags & TDP_KTHREAD) != 0) {
163 pcb2->pcb_fsbase = 0;
164 pcb2->pcb_gsbase = 0;
165 clear_pcb_flags(pcb2, PCB_FPUINITDONE | PCB_USERFPUINITDONE |
166 PCB_KERNFPU | PCB_KERNFPU_THR);
167 } else {
168 MPASS((pcb2->pcb_flags & (PCB_KERNFPU | PCB_KERNFPU_THR)) == 0);
169 bcopy(get_pcb_user_save_td(td1), get_pcb_user_save_pcb(pcb2),
170 cpu_max_ext_state_size);
171 }
172
173 /*
174 * Set registers for trampoline to user mode. Leave space for the
175 * return address on stack. These are the kernel mode register values.
176 */
177 pcb2->pcb_r12 = (register_t)fork_return; /* fork_trampoline argument */
178 pcb2->pcb_rbp = 0;
179 pcb2->pcb_rsp = (register_t)td2->td_frame - sizeof(void *);
180 pcb2->pcb_rbx = (register_t)td2; /* fork_trampoline argument */
181 pcb2->pcb_rip = (register_t)fork_trampoline;
182 /*-
183 * pcb2->pcb_dr*: cloned above.
184 * pcb2->pcb_savefpu: cloned above.
185 * pcb2->pcb_flags: cloned above.
186 * pcb2->pcb_onfault: cloned above (always NULL here?).
187 * pcb2->pcb_[fg]sbase: cloned above
188 */
189
190 pcb2->pcb_tssp = NULL;
191
192 /* Setup to release spin count in fork_exit(). */
193 td2->td_md.md_spinlock_count = 1;
194 td2->td_md.md_saved_flags = PSL_KERNEL | PSL_I;
195 pmap_thread_init_invl_gen(td2);
196 }
197
198 /*
199 * Finish a fork operation, with process p2 nearly set up.
200 * Copy and update the pcb, set up the stack so that the child
201 * ready to run and return to user mode.
202 */
203 void
cpu_fork(struct thread * td1,struct proc * p2,struct thread * td2,int flags)204 cpu_fork(struct thread *td1, struct proc *p2, struct thread *td2, int flags)
205 {
206 struct proc *p1;
207 struct pcb *pcb2;
208 struct mdproc *mdp1, *mdp2;
209 struct proc_ldt *pldt;
210
211 p1 = td1->td_proc;
212 if ((flags & RFPROC) == 0) {
213 if ((flags & RFMEM) == 0) {
214 /* unshare user LDT */
215 mdp1 = &p1->p_md;
216 mtx_lock(&dt_lock);
217 if ((pldt = mdp1->md_ldt) != NULL &&
218 pldt->ldt_refcnt > 1 &&
219 user_ldt_alloc(p1, 1) == NULL)
220 panic("could not copy LDT");
221 mtx_unlock(&dt_lock);
222 }
223 return;
224 }
225
226 /* Point the stack and pcb to the actual location */
227 set_top_of_stack_td(td2);
228 td2->td_pcb = pcb2 = get_pcb_td(td2);
229
230 copy_thread(td1, td2);
231
232 /* Reset debug registers in the new process */
233 x86_clear_dbregs(pcb2);
234
235 /* Point mdproc and then copy over p1's contents */
236 mdp2 = &p2->p_md;
237 bcopy(&p1->p_md, mdp2, sizeof(*mdp2));
238
239 /*
240 * Copy the trap frame for the return to user mode as if from a
241 * syscall. This copies most of the user mode register values.
242 */
243 td2->td_frame = (struct trapframe *)td2->td_md.md_stack_base - 1;
244 bcopy(td1->td_frame, td2->td_frame, sizeof(struct trapframe));
245
246 td2->td_frame->tf_rax = 0; /* Child returns zero */
247 td2->td_frame->tf_rflags &= ~PSL_C; /* success */
248 td2->td_frame->tf_rdx = 1;
249
250 /*
251 * If the parent process has the trap bit set (i.e. a debugger
252 * had single stepped the process to the system call), we need
253 * to clear the trap flag from the new frame.
254 */
255 td2->td_frame->tf_rflags &= ~PSL_T;
256
257 /* As on i386, do not copy io permission bitmap. */
258 pcb2->pcb_tssp = NULL;
259
260 /* New segment registers. */
261 set_pcb_flags_raw(pcb2, PCB_FULL_IRET);
262
263 /* Copy the LDT, if necessary. */
264 mdp1 = &td1->td_proc->p_md;
265 mdp2 = &p2->p_md;
266 if (mdp1->md_ldt == NULL) {
267 mdp2->md_ldt = NULL;
268 return;
269 }
270 mtx_lock(&dt_lock);
271 if (mdp1->md_ldt != NULL) {
272 if (flags & RFMEM) {
273 mdp1->md_ldt->ldt_refcnt++;
274 mdp2->md_ldt = mdp1->md_ldt;
275 bcopy(&mdp1->md_ldt_sd, &mdp2->md_ldt_sd, sizeof(struct
276 system_segment_descriptor));
277 } else {
278 mdp2->md_ldt = NULL;
279 mdp2->md_ldt = user_ldt_alloc(p2, 0);
280 if (mdp2->md_ldt == NULL)
281 panic("could not copy LDT");
282 amd64_set_ldt_data(td2, 0, max_ldt_segment,
283 (struct user_segment_descriptor *)
284 mdp1->md_ldt->ldt_base);
285 }
286 } else
287 mdp2->md_ldt = NULL;
288 mtx_unlock(&dt_lock);
289
290 /*
291 * Now, cpu_switch() can schedule the new process.
292 * pcb_rsp is loaded pointing to the cpu_switch() stack frame
293 * containing the return address when exiting cpu_switch.
294 * This will normally be to fork_trampoline(), which will have
295 * %rbx loaded with the new proc's pointer. fork_trampoline()
296 * will set up a stack to call fork_return(p, frame); to complete
297 * the return to user-mode.
298 */
299 }
300
301 /*
302 * Intercept the return address from a freshly forked process that has NOT
303 * been scheduled yet.
304 *
305 * This is needed to make kernel threads stay in kernel mode.
306 */
307 void
cpu_fork_kthread_handler(struct thread * td,void (* func)(void *),void * arg)308 cpu_fork_kthread_handler(struct thread *td, void (*func)(void *), void *arg)
309 {
310 /*
311 * Note that the trap frame follows the args, so the function
312 * is really called like this: func(arg, frame);
313 */
314 td->td_pcb->pcb_r12 = (long) func; /* function */
315 td->td_pcb->pcb_rbx = (long) arg; /* first arg */
316 }
317
318 void
cpu_exit(struct thread * td)319 cpu_exit(struct thread *td)
320 {
321
322 /*
323 * If this process has a custom LDT, release it.
324 */
325 if (td->td_proc->p_md.md_ldt != NULL)
326 user_ldt_free(td);
327 }
328
329 void
cpu_thread_exit(struct thread * td)330 cpu_thread_exit(struct thread *td)
331 {
332 struct pcb *pcb;
333
334 critical_enter();
335 if (td == PCPU_GET(fpcurthread))
336 fpudrop();
337 critical_exit();
338
339 pcb = td->td_pcb;
340
341 /* Disable any hardware breakpoints. */
342 if (pcb->pcb_flags & PCB_DBREGS) {
343 reset_dbregs();
344 clear_pcb_flags(pcb, PCB_DBREGS);
345 }
346 }
347
348 void
cpu_thread_clean(struct thread * td)349 cpu_thread_clean(struct thread *td)
350 {
351 struct pcb *pcb;
352
353 pcb = td->td_pcb;
354
355 /*
356 * Clean TSS/iomap
357 */
358 if (pcb->pcb_tssp != NULL) {
359 pmap_pti_remove_kva((vm_offset_t)pcb->pcb_tssp,
360 (vm_offset_t)pcb->pcb_tssp + ctob(IOPAGES + 1));
361 kmem_free((vm_offset_t)pcb->pcb_tssp, ctob(IOPAGES + 1));
362 pcb->pcb_tssp = NULL;
363 }
364 }
365
366 void
cpu_thread_swapin(struct thread * td)367 cpu_thread_swapin(struct thread *td)
368 {
369 }
370
371 void
cpu_thread_swapout(struct thread * td)372 cpu_thread_swapout(struct thread *td)
373 {
374 }
375
376 void
cpu_thread_alloc(struct thread * td)377 cpu_thread_alloc(struct thread *td)
378 {
379 struct pcb *pcb;
380 struct xstate_hdr *xhdr;
381
382 set_top_of_stack_td(td);
383 td->td_pcb = pcb = get_pcb_td(td);
384 td->td_frame = (struct trapframe *)td->td_md.md_stack_base - 1;
385 td->td_md.md_usr_fpu_save = fpu_save_area_alloc();
386 pcb->pcb_save = get_pcb_user_save_pcb(pcb);
387 if (use_xsave) {
388 xhdr = (struct xstate_hdr *)(pcb->pcb_save + 1);
389 bzero(xhdr, sizeof(*xhdr));
390 xhdr->xstate_bv = xsave_mask;
391 }
392 }
393
394 void
cpu_thread_free(struct thread * td)395 cpu_thread_free(struct thread *td)
396 {
397 cpu_thread_clean(td);
398
399 fpu_save_area_free(td->td_md.md_usr_fpu_save);
400 td->td_md.md_usr_fpu_save = NULL;
401 }
402
403 bool
cpu_exec_vmspace_reuse(struct proc * p,vm_map_t map)404 cpu_exec_vmspace_reuse(struct proc *p, vm_map_t map)
405 {
406
407 return (((curproc->p_md.md_flags & P_MD_KPTI) != 0) ==
408 (vm_map_pmap(map)->pm_ucr3 != PMAP_NO_CR3));
409 }
410
411 static void
cpu_procctl_kpti_ctl(struct proc * p,int val)412 cpu_procctl_kpti_ctl(struct proc *p, int val)
413 {
414
415 if (pti && val == PROC_KPTI_CTL_ENABLE_ON_EXEC)
416 p->p_md.md_flags |= P_MD_KPTI;
417 if (val == PROC_KPTI_CTL_DISABLE_ON_EXEC)
418 p->p_md.md_flags &= ~P_MD_KPTI;
419 }
420
421 static void
cpu_procctl_kpti_status(struct proc * p,int * val)422 cpu_procctl_kpti_status(struct proc *p, int *val)
423 {
424 *val = (p->p_md.md_flags & P_MD_KPTI) != 0 ?
425 PROC_KPTI_CTL_ENABLE_ON_EXEC:
426 PROC_KPTI_CTL_DISABLE_ON_EXEC;
427 if (vmspace_pmap(p->p_vmspace)->pm_ucr3 != PMAP_NO_CR3)
428 *val |= PROC_KPTI_STATUS_ACTIVE;
429 }
430
431 static int
cpu_procctl_la_ctl(struct proc * p,int val)432 cpu_procctl_la_ctl(struct proc *p, int val)
433 {
434 int error;
435
436 error = 0;
437 switch (val) {
438 case PROC_LA_CTL_LA48_ON_EXEC:
439 p->p_md.md_flags |= P_MD_LA48;
440 p->p_md.md_flags &= ~P_MD_LA57;
441 break;
442 case PROC_LA_CTL_LA57_ON_EXEC:
443 if (la57) {
444 p->p_md.md_flags &= ~P_MD_LA48;
445 p->p_md.md_flags |= P_MD_LA57;
446 } else {
447 error = ENOTSUP;
448 }
449 break;
450 case PROC_LA_CTL_DEFAULT_ON_EXEC:
451 p->p_md.md_flags &= ~(P_MD_LA48 | P_MD_LA57);
452 break;
453 }
454 return (error);
455 }
456
457 static void
cpu_procctl_la_status(struct proc * p,int * val)458 cpu_procctl_la_status(struct proc *p, int *val)
459 {
460 int res;
461
462 if ((p->p_md.md_flags & P_MD_LA48) != 0)
463 res = PROC_LA_CTL_LA48_ON_EXEC;
464 else if ((p->p_md.md_flags & P_MD_LA57) != 0)
465 res = PROC_LA_CTL_LA57_ON_EXEC;
466 else
467 res = PROC_LA_CTL_DEFAULT_ON_EXEC;
468 if (p->p_sysent->sv_maxuser == VM_MAXUSER_ADDRESS_LA48)
469 res |= PROC_LA_STATUS_LA48;
470 else
471 res |= PROC_LA_STATUS_LA57;
472 *val = res;
473 }
474
475 int
cpu_procctl(struct thread * td,int idtype,id_t id,int com,void * data)476 cpu_procctl(struct thread *td, int idtype, id_t id, int com, void *data)
477 {
478 struct proc *p;
479 int error, val;
480
481 switch (com) {
482 case PROC_KPTI_CTL:
483 case PROC_KPTI_STATUS:
484 case PROC_LA_CTL:
485 case PROC_LA_STATUS:
486 if (idtype != P_PID) {
487 error = EINVAL;
488 break;
489 }
490 if (com == PROC_KPTI_CTL) {
491 /* sad but true and not a joke */
492 error = priv_check(td, PRIV_IO);
493 if (error != 0)
494 break;
495 }
496 if (com == PROC_KPTI_CTL || com == PROC_LA_CTL) {
497 error = copyin(data, &val, sizeof(val));
498 if (error != 0)
499 break;
500 }
501 if (com == PROC_KPTI_CTL &&
502 val != PROC_KPTI_CTL_ENABLE_ON_EXEC &&
503 val != PROC_KPTI_CTL_DISABLE_ON_EXEC) {
504 error = EINVAL;
505 break;
506 }
507 if (com == PROC_LA_CTL &&
508 val != PROC_LA_CTL_LA48_ON_EXEC &&
509 val != PROC_LA_CTL_LA57_ON_EXEC &&
510 val != PROC_LA_CTL_DEFAULT_ON_EXEC) {
511 error = EINVAL;
512 break;
513 }
514 error = pget(id, PGET_CANSEE | PGET_NOTWEXIT | PGET_NOTID, &p);
515 if (error != 0)
516 break;
517 switch (com) {
518 case PROC_KPTI_CTL:
519 cpu_procctl_kpti_ctl(p, val);
520 break;
521 case PROC_KPTI_STATUS:
522 cpu_procctl_kpti_status(p, &val);
523 break;
524 case PROC_LA_CTL:
525 error = cpu_procctl_la_ctl(p, val);
526 break;
527 case PROC_LA_STATUS:
528 cpu_procctl_la_status(p, &val);
529 break;
530 }
531 PROC_UNLOCK(p);
532 if (com == PROC_KPTI_STATUS || com == PROC_LA_STATUS)
533 error = copyout(&val, data, sizeof(val));
534 break;
535 default:
536 error = EINVAL;
537 break;
538 }
539 return (error);
540 }
541
542 void
cpu_set_syscall_retval(struct thread * td,int error)543 cpu_set_syscall_retval(struct thread *td, int error)
544 {
545 struct trapframe *frame;
546
547 frame = td->td_frame;
548 if (__predict_true(error == 0)) {
549 frame->tf_rax = td->td_retval[0];
550 frame->tf_rdx = td->td_retval[1];
551 frame->tf_rflags &= ~PSL_C;
552 return;
553 }
554
555 switch (error) {
556 case ERESTART:
557 /*
558 * Reconstruct pc, we know that 'syscall' is 2 bytes,
559 * lcall $X,y is 7 bytes, int 0x80 is 2 bytes.
560 * We saved this in tf_err.
561 * %r10 (which was holding the value of %rcx) is restored
562 * for the next iteration.
563 * %r10 restore is only required for freebsd/amd64 processes,
564 * but shall be innocent for any ia32 ABI.
565 *
566 * Require full context restore to get the arguments
567 * in the registers reloaded at return to usermode.
568 */
569 frame->tf_rip -= frame->tf_err;
570 frame->tf_r10 = frame->tf_rcx;
571 set_pcb_flags(td->td_pcb, PCB_FULL_IRET);
572 break;
573
574 case EJUSTRETURN:
575 break;
576
577 default:
578 frame->tf_rax = error;
579 frame->tf_rflags |= PSL_C;
580 break;
581 }
582 }
583
584 /*
585 * Initialize machine state, mostly pcb and trap frame for a new
586 * thread, about to return to userspace. Put enough state in the new
587 * thread's PCB to get it to go back to the fork_return(), which
588 * finalizes the thread state and handles peculiarities of the first
589 * return to userspace for the new thread.
590 */
591 void
cpu_copy_thread(struct thread * td,struct thread * td0)592 cpu_copy_thread(struct thread *td, struct thread *td0)
593 {
594 copy_thread(td0, td);
595
596 /*
597 * Copy user general-purpose registers.
598 *
599 * Some of these registers are rewritten by cpu_set_upcall()
600 * and linux_set_upcall().
601 */
602 bcopy(td0->td_frame, td->td_frame, sizeof(struct trapframe));
603
604 /* If the current thread has the trap bit set (i.e. a debugger had
605 * single stepped the process to the system call), we need to clear
606 * the trap flag from the new frame. Otherwise, the new thread will
607 * receive a (likely unexpected) SIGTRAP when it executes the first
608 * instruction after returning to userland.
609 */
610 td->td_frame->tf_rflags &= ~PSL_T;
611
612 set_pcb_flags_raw(td->td_pcb, PCB_FULL_IRET);
613 }
614
615 /*
616 * Set that machine state for performing an upcall that starts
617 * the entry function with the given argument.
618 */
619 void
cpu_set_upcall(struct thread * td,void (* entry)(void *),void * arg,stack_t * stack)620 cpu_set_upcall(struct thread *td, void (*entry)(void *), void *arg,
621 stack_t *stack)
622 {
623
624 /*
625 * Do any extra cleaning that needs to be done.
626 * The thread may have optional components
627 * that are not present in a fresh thread.
628 * This may be a recycled thread so make it look
629 * as though it's newly allocated.
630 */
631 cpu_thread_clean(td);
632
633 #ifdef COMPAT_FREEBSD32
634 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
635 /*
636 * Set the trap frame to point at the beginning of the entry
637 * function.
638 */
639 td->td_frame->tf_rbp = 0;
640 td->td_frame->tf_rsp =
641 (((uintptr_t)stack->ss_sp + stack->ss_size - 4) & ~0x0f) - 4;
642 td->td_frame->tf_rip = (uintptr_t)entry;
643
644 /* Return address sentinel value to stop stack unwinding. */
645 suword32((void *)td->td_frame->tf_rsp, 0);
646
647 /* Pass the argument to the entry point. */
648 suword32((void *)(td->td_frame->tf_rsp + sizeof(int32_t)),
649 (uint32_t)(uintptr_t)arg);
650
651 return;
652 }
653 #endif
654
655 /*
656 * Set the trap frame to point at the beginning of the uts
657 * function.
658 */
659 td->td_frame->tf_rbp = 0;
660 td->td_frame->tf_rsp =
661 ((register_t)stack->ss_sp + stack->ss_size) & ~0x0f;
662 td->td_frame->tf_rsp -= 8;
663 td->td_frame->tf_rip = (register_t)entry;
664 td->td_frame->tf_ds = _udatasel;
665 td->td_frame->tf_es = _udatasel;
666 td->td_frame->tf_fs = _ufssel;
667 td->td_frame->tf_gs = _ugssel;
668 td->td_frame->tf_flags = TF_HASSEGS;
669
670 /* Return address sentinel value to stop stack unwinding. */
671 suword((void *)td->td_frame->tf_rsp, 0);
672
673 /* Pass the argument to the entry point. */
674 td->td_frame->tf_rdi = (register_t)arg;
675 }
676
677 int
cpu_set_user_tls(struct thread * td,void * tls_base)678 cpu_set_user_tls(struct thread *td, void *tls_base)
679 {
680 struct pcb *pcb;
681
682 if ((u_int64_t)tls_base >= VM_MAXUSER_ADDRESS)
683 return (EINVAL);
684
685 pcb = td->td_pcb;
686 set_pcb_flags(pcb, PCB_FULL_IRET);
687 #ifdef COMPAT_FREEBSD32
688 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
689 pcb->pcb_gsbase = (register_t)tls_base;
690 return (0);
691 }
692 #endif
693 pcb->pcb_fsbase = (register_t)tls_base;
694 return (0);
695 }
696
697 /*
698 * Software interrupt handler for queued VM system processing.
699 */
700 void
swi_vm(void * dummy)701 swi_vm(void *dummy)
702 {
703 if (busdma_swi_pending != 0)
704 busdma_swi();
705 }
706
707 /*
708 * Tell whether this address is in some physical memory region.
709 * Currently used by the kernel coredump code in order to avoid
710 * dumping the ``ISA memory hole'' which could cause indefinite hangs,
711 * or other unpredictable behaviour.
712 */
713
714 int
is_physical_memory(vm_paddr_t addr)715 is_physical_memory(vm_paddr_t addr)
716 {
717
718 #ifdef DEV_ISA
719 /* The ISA ``memory hole''. */
720 if (addr >= 0xa0000 && addr < 0x100000)
721 return 0;
722 #endif
723
724 /*
725 * stuff other tests for known memory-mapped devices (PCI?)
726 * here
727 */
728
729 return 1;
730 }
731