xref: /netbsd-src/sys/compat/linux/arch/amd64/linux_machdep.c (revision df0caa2637da0538ecdf6b878c4d08e684b43d8f)
1 /*	$NetBSD: linux_machdep.c,v 1.9 2005/06/24 22:57:05 manu Exp $ */
2 
3 /*-
4  * Copyright (c) 2005 Emmanuel Dreyfus, all rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
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  * 3. All advertising materials mentioning features or use of this software
15  *    must display the following acknowledgement:
16  *	This product includes software developed by Emmanuel Dreyfus
17  * 4. The name of the author may not be used to endorse or promote
18  *    products derived from this software without specific prior written
19  *    permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE THE AUTHOR AND CONTRIBUTORS ``AS IS''
22  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
23  * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
24  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS
25  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
28  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
29  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
30  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31  * POSSIBILITY OF SUCH DAMAGE.
32  */
33 
34 #include <sys/cdefs.h>
35 
36 __KERNEL_RCSID(0, "$NetBSD: linux_machdep.c,v 1.9 2005/06/24 22:57:05 manu Exp $");
37 
38 #include <sys/param.h>
39 #include <sys/types.h>
40 #include <sys/systm.h>
41 #include <sys/signal.h>
42 #include <sys/exec.h>
43 #include <sys/proc.h>
44 #include <sys/ptrace.h> /* for process_read_fpregs() */
45 #include <sys/user.h>
46 #include <sys/ucontext.h>
47 
48 #include <machine/reg.h>
49 #include <machine/pcb.h>
50 #include <machine/fpu.h>
51 #include <machine/mcontext.h>
52 #include <machine/specialreg.h>
53 #include <machine/vmparam.h>
54 
55 #include <compat/linux/common/linux_signal.h>
56 #include <compat/linux/common/linux_errno.h>
57 #include <compat/linux/common/linux_exec.h>
58 #include <compat/linux/common/linux_ioctl.h>
59 #include <compat/linux/common/linux_prctl.h>
60 #include <compat/linux/common/linux_machdep.h>
61 #include <compat/linux/linux_syscall.h>
62 #include <compat/linux/linux_syscallargs.h>
63 
64 static void linux_buildcontext(struct lwp *, void *, void *);
65 
66 void
67 linux_setregs(l, epp, stack)
68         struct lwp *l;
69 	struct exec_package *epp;
70 	u_long stack;
71 {
72 	struct pcb *pcb = &l->l_addr->u_pcb;
73 	struct trapframe *tf;
74 
75 	/* If we were using the FPU, forget about it. */
76 	if (l->l_addr->u_pcb.pcb_fpcpu != NULL)
77 		fpusave_lwp(l, 0);
78 
79 	l->l_md.md_flags &= ~MDP_USEDFPU;
80 	pcb->pcb_flags = 0;
81 	pcb->pcb_savefpu.fp_fxsave.fx_fcw = __NetBSD_NPXCW__;
82 	pcb->pcb_savefpu.fp_fxsave.fx_mxcsr = __INITIAL_MXCSR__;
83 	pcb->pcb_savefpu.fp_fxsave.fx_mxcsr_mask = __INITIAL_MXCSR_MASK__;
84 	pcb->pcb_fs = 0;
85 	pcb->pcb_gs = 0;
86 
87 	l->l_proc->p_flag &= ~P_32;
88 
89 	tf = l->l_md.md_regs;
90 	tf->tf_rax = 0;
91 	tf->tf_rbx = 0;
92 	tf->tf_rcx = epp->ep_entry;
93 	tf->tf_rdx = 0;
94 	tf->tf_rsi = 0;
95 	tf->tf_rdi = 0;
96 	tf->tf_rbp = 0;
97 	tf->tf_rsp = stack;
98 	tf->tf_r8 = 0;
99 	tf->tf_r9 = 0;
100 	tf->tf_r10 = 0;
101 	tf->tf_r11 = 0;
102 	tf->tf_r12 = 0;
103 	tf->tf_r13 = 0;
104 	tf->tf_r14 = 0;
105 	tf->tf_r15 = 0;
106 	tf->tf_rip = epp->ep_entry;
107 	tf->tf_rflags = PSL_USERSET;
108 	tf->tf_cs = GSEL(GUCODE_SEL, SEL_UPL);
109 	tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL);
110 	tf->tf_ds = 0;
111 	tf->tf_es = 0;
112 	tf->tf_fs = 0;
113 	tf->tf_gs = 0;
114 
115 	return;
116 }
117 
118 void
119 linux_sendsig(ksi, mask)
120 	const ksiginfo_t *ksi;
121 	const sigset_t *mask;
122 {
123 	struct lwp *l = curlwp;
124 	struct proc *p = l->l_proc;
125 	struct sigacts *ps = p->p_sigacts;
126 	int onstack;
127 	int sig = ksi->ksi_signo;
128 	struct linux_rt_sigframe *sfp, sigframe;
129 	struct linux__fpstate *fpsp, fpstate;
130 	struct fpreg fpregs;
131 	struct trapframe *tf = l->l_md.md_regs;
132 	sig_t catcher = SIGACTION(p, sig).sa_handler;
133 	linux_sigset_t lmask;
134 	char *sp;
135 	int error;
136 
137 	/* Do we need to jump onto the signal stack? */
138 	onstack =
139 	    (p->p_sigctx.ps_sigstk.ss_flags & (SS_DISABLE | SS_ONSTACK)) == 0 &&
140 	    (SIGACTION(p, sig).sa_flags & SA_ONSTACK) != 0;
141 
142 	/* Allocate space for the signal handler context. */
143 	if (onstack)
144 		sp = ((caddr_t)p->p_sigctx.ps_sigstk.ss_sp +
145 		    p->p_sigctx.ps_sigstk.ss_size);
146 	else
147 		sp = (caddr_t)tf->tf_rsp - 128;
148 
149 
150 	/*
151 	 * Save FPU state, if any
152 	 */
153 	if (l->l_md.md_flags & MDP_USEDFPU) {
154 		sp = (char *)
155 		    (((long)sp - sizeof(struct linux__fpstate)) & ~0xfUL);
156 		fpsp = (struct linux__fpstate *)sp;
157 
158 		(void)process_read_fpregs(l, &fpregs);
159 		bzero(&fpstate, sizeof(fpstate));
160 
161 		fpstate.cwd = fpregs.fp_fcw;
162 		fpstate.swd = fpregs.fp_fsw;
163 		fpstate.twd = fpregs.fp_ftw;
164 		fpstate.fop = fpregs.fp_fop;
165 		fpstate.rip = fpregs.fp_rip;
166 		fpstate.rdp = fpregs.fp_rdp;
167 		fpstate.mxcsr = fpregs.fp_mxcsr;
168 		fpstate.mxcsr_mask = fpregs.fp_mxcsr_mask;
169 		memcpy(&fpstate.st_space, &fpregs.fp_st,
170 		    sizeof(fpstate.st_space));
171 		memcpy(&fpstate.xmm_space, &fpregs.fp_xmm,
172 		    sizeof(fpstate.xmm_space));
173 
174 		if ((error = copyout(&fpstate, fpsp, sizeof(fpstate))) != 0) {
175 			sigexit(l, SIGILL);
176 			return;
177 		}
178 	} else {
179 		fpsp = NULL;
180 	}
181 
182 	/*
183 	 * Populate the rt_sigframe
184 	 */
185 	sp = (char *)
186 	    ((((long)sp - sizeof(struct linux_rt_sigframe)) & ~0xfUL) - 8);
187 	sfp = (struct linux_rt_sigframe *)sp;
188 
189 	bzero(&sigframe, sizeof(sigframe));
190 	if (ps->sa_sigdesc[sig].sd_vers != 0)
191 		sigframe.pretcode =
192 		    (char *)(u_long)ps->sa_sigdesc[sig].sd_tramp;
193 	else
194 		sigframe.pretcode = NULL;
195 
196 	/*
197 	 * The user context
198 	 */
199 	sigframe.uc.luc_flags = 0;
200 	sigframe.uc.luc_link = NULL;
201 
202 	/* This is used regardless of SA_ONSTACK in Linux */
203 	sigframe.uc.luc_stack.ss_sp = p->p_sigctx.ps_sigstk.ss_sp;
204 	sigframe.uc.luc_stack.ss_size = p->p_sigctx.ps_sigstk.ss_size;
205 	sigframe.uc.luc_stack.ss_flags = 0;
206 	if (p->p_sigctx.ps_sigstk.ss_flags & SS_ONSTACK)
207 		sigframe.uc.luc_stack.ss_flags |= LINUX_SS_ONSTACK;
208 	if (p->p_sigctx.ps_sigstk.ss_flags & SS_DISABLE)
209 		sigframe.uc.luc_stack.ss_flags |= LINUX_SS_DISABLE;
210 
211 	sigframe.uc.luc_mcontext.r8 = tf->tf_r8;
212 	sigframe.uc.luc_mcontext.r9 = tf->tf_r9;
213 	sigframe.uc.luc_mcontext.r10 = tf->tf_r10;
214 	sigframe.uc.luc_mcontext.r11 = tf->tf_r11;
215 	sigframe.uc.luc_mcontext.r12 = tf->tf_r12;
216 	sigframe.uc.luc_mcontext.r13 = tf->tf_r13;
217 	sigframe.uc.luc_mcontext.r14 = tf->tf_r14;
218 	sigframe.uc.luc_mcontext.r15 = tf->tf_r15;
219 	sigframe.uc.luc_mcontext.rdi = tf->tf_rdi;
220 	sigframe.uc.luc_mcontext.rsi = tf->tf_rsi;
221 	sigframe.uc.luc_mcontext.rbp = tf->tf_rbp;
222 	sigframe.uc.luc_mcontext.rbx = tf->tf_rbx;
223 	sigframe.uc.luc_mcontext.rdx = tf->tf_rdx;
224 	sigframe.uc.luc_mcontext.rcx = tf->tf_rcx;
225 	sigframe.uc.luc_mcontext.rsp = tf->tf_rsp;
226 	sigframe.uc.luc_mcontext.rip = tf->tf_rip;
227 	sigframe.uc.luc_mcontext.eflags = tf->tf_rflags;
228 	sigframe.uc.luc_mcontext.cs = tf->tf_cs;
229 	sigframe.uc.luc_mcontext.gs = tf->tf_gs;
230 	sigframe.uc.luc_mcontext.fs = tf->tf_fs;
231 	sigframe.uc.luc_mcontext.err = tf->tf_err;
232 	sigframe.uc.luc_mcontext.trapno = tf->tf_trapno;
233 	native_to_linux_sigset(&lmask, mask);
234 	sigframe.uc.luc_mcontext.oldmask = lmask.sig[0];
235 	sigframe.uc.luc_mcontext.cr2 = (long)l->l_addr->u_pcb.pcb_onfault;
236 	sigframe.uc.luc_mcontext.fpstate = fpsp;
237 	native_to_linux_sigset(&sigframe.uc.luc_sigmask, mask);
238 
239 	/*
240 	 * the siginfo structure
241 	 */
242 	sigframe.info.lsi_signo = native_to_linux_signo[sig];
243 	sigframe.info.lsi_errno = native_to_linux_errno[ksi->ksi_errno];
244 	sigframe.info.lsi_code = ksi->ksi_code;
245 
246 	/* XXX This is a rought conversion, taken from i386 code */
247 	switch (sigframe.info.lsi_signo) {
248 	case LINUX_SIGILL:
249 	case LINUX_SIGFPE:
250 	case LINUX_SIGSEGV:
251 	case LINUX_SIGBUS:
252 	case LINUX_SIGTRAP:
253 		sigframe.info._sifields._sigfault._addr = ksi->ksi_addr;
254 		break;
255 	case LINUX_SIGCHLD:
256 		sigframe.info._sifields._sigchld._pid = ksi->ksi_pid;
257 		sigframe.info._sifields._sigchld._uid = ksi->ksi_uid;
258 		sigframe.info._sifields._sigchld._utime = ksi->ksi_utime;
259 		sigframe.info._sifields._sigchld._stime = ksi->ksi_stime;
260 
261 		if (WCOREDUMP(ksi->ksi_status)) {
262 			sigframe.info.lsi_code = LINUX_CLD_DUMPED;
263 			sigframe.info._sifields._sigchld._status =
264 			    _WSTATUS(ksi->ksi_status);
265 		} else if (_WSTATUS(ksi->ksi_status)) {
266 			sigframe.info.lsi_code = LINUX_CLD_KILLED;
267 			sigframe.info._sifields._sigchld._status =
268 			    _WSTATUS(ksi->ksi_status);
269 		} else {
270 			sigframe.info.lsi_code = LINUX_CLD_EXITED;
271 			sigframe.info._sifields._sigchld._status =
272 			    ((ksi->ksi_status & 0xff00U) >> 8);
273 		}
274 		break;
275 	case LINUX_SIGIO:
276 		sigframe.info._sifields._sigpoll._band = ksi->ksi_band;
277 		sigframe.info._sifields._sigpoll._fd = ksi->ksi_fd;
278 		break;
279 	default:
280 		sigframe.info._sifields._sigchld._pid = ksi->ksi_pid;
281 		sigframe.info._sifields._sigchld._uid = ksi->ksi_uid;
282 		if ((sigframe.info.lsi_signo == LINUX_SIGALRM) ||
283 		    (sigframe.info.lsi_signo >= LINUX_SIGRTMIN))
284 			sigframe.info._sifields._timer._sigval.sival_ptr =
285 			     ksi->ksi_sigval.sival_ptr;
286 		break;
287 	}
288 
289 	if ((error = copyout(&sigframe, sp, sizeof(sigframe))) != 0) {
290 		sigexit(l, SIGILL);
291 		return;
292 	}
293 
294 	linux_buildcontext(l, catcher, sp);
295 	tf->tf_rdi = sigframe.info.lsi_signo;
296 	tf->tf_rax = 0;
297 	tf->tf_rsi = (long)&sfp->info;
298 	tf->tf_rdx = (long)&sfp->uc;
299 
300 	/*
301 	 * Remember we use signal stack
302 	 */
303 	if (onstack)
304 		p->p_sigctx.ps_sigstk.ss_flags |= SS_ONSTACK;
305 	return;
306 }
307 
308 int
309 linux_sys_modify_ldt(l, v, retval)
310         struct lwp *l;
311         void *v;
312         register_t *retval;
313 {
314 	return 0;
315 }
316 
317 int
318 linux_sys_iopl(l, v, retval)
319         struct lwp *l;
320         void *v;
321         register_t *retval;
322 {
323 	return 0;
324 }
325 
326 int
327 linux_sys_ioperm(l, v, retval)
328         struct lwp *l;
329         void *v;
330         register_t *retval;
331 {
332 	return 0;
333 }
334 
335 dev_t
336 linux_fakedev(dev, raw)
337         dev_t dev;
338 	int raw;
339 {
340 	return 0;
341 }
342 
343 int
344 linux_machdepioctl(p, v, retval)
345         struct proc *p;
346         void *v;
347         register_t *retval;
348 {
349 	return 0;
350 }
351 
352 int
353 linux_sys_rt_sigreturn(l, v, retval)
354         struct lwp *l;
355         void *v;
356         register_t *retval;
357 {
358 	struct linux_ucontext *luctx;
359 	struct trapframe *tf = l->l_md.md_regs;
360 	struct linux_sigcontext *lsigctx;
361 	struct linux__fpstate fpstate;
362 	struct linux_rt_sigframe frame, *fp;
363 	ucontext_t uctx;
364 	mcontext_t *mctx;
365 	struct fxsave64 *fxsave;
366 	int error;
367 
368 	fp = (struct linux_rt_sigframe *)(tf->tf_rsp - 8);
369 	if ((error = copyin(fp, &frame, sizeof(frame))) != 0) {
370 		sigexit(l, SIGILL);
371 		return error;
372 	}
373 	luctx = &frame.uc;
374 	lsigctx = &luctx->luc_mcontext;
375 
376 	bzero(&uctx, sizeof(uctx));
377 	mctx = (mcontext_t *)&uctx.uc_mcontext;
378 	fxsave = (struct fxsave64 *)&mctx->__fpregs;
379 
380 	/*
381 	 * Set the flags. Linux always have CPU, stack and signal state,
382 	 * FPU is optional. uc_flags is not used to tell what we have.
383 	 */
384 	uctx.uc_flags = (_UC_SIGMASK|_UC_CPU|_UC_STACK|_UC_CLRSTACK);
385 	if (lsigctx->fpstate != NULL)
386 		uctx.uc_flags |= _UC_FPU;
387 	uctx.uc_link = NULL;
388 
389 	/*
390 	 * Signal set
391 	 */
392 	linux_to_native_sigset(&uctx.uc_sigmask, &luctx->luc_sigmask);
393 
394 	/*
395 	 * CPU state
396 	 */
397 	mctx->__gregs[_REG_R8] = lsigctx->r8;
398 	mctx->__gregs[_REG_R9] = lsigctx->r9;
399 	mctx->__gregs[_REG_R10] = lsigctx->r10;
400 	mctx->__gregs[_REG_R11] = lsigctx->r11;
401 	mctx->__gregs[_REG_R12] = lsigctx->r12;
402 	mctx->__gregs[_REG_R13] = lsigctx->r13;
403 	mctx->__gregs[_REG_R14] = lsigctx->r14;
404 	mctx->__gregs[_REG_R15] = lsigctx->r15;
405 	mctx->__gregs[_REG_RDI] = lsigctx->rdi;
406 	mctx->__gregs[_REG_RSI] = lsigctx->rsi;
407 	mctx->__gregs[_REG_RBP] = lsigctx->rbp;
408 	mctx->__gregs[_REG_RBX] = lsigctx->rbx;
409 	mctx->__gregs[_REG_RAX] = tf->tf_rax;
410 	mctx->__gregs[_REG_RDX] = lsigctx->rdx;
411 	mctx->__gregs[_REG_RCX] = lsigctx->rcx;
412 	mctx->__gregs[_REG_RIP] = lsigctx->rip;
413 	mctx->__gregs[_REG_RFL] = lsigctx->eflags;
414 	mctx->__gregs[_REG_CS] = lsigctx->cs;
415 	mctx->__gregs[_REG_GS] = lsigctx->gs;
416 	mctx->__gregs[_REG_FS] = lsigctx->fs;
417 	mctx->__gregs[_REG_ERR] = lsigctx->err;
418 	mctx->__gregs[_REG_TRAPNO] = lsigctx->trapno;
419 	mctx->__gregs[_REG_ES] = tf->tf_es;
420 	mctx->__gregs[_REG_DS] = tf->tf_ds;
421 	mctx->__gregs[_REG_URSP] = lsigctx->rsp; /* XXX */
422 	mctx->__gregs[_REG_SS] = tf->tf_ss;
423 
424 	/*
425 	 * FPU state
426 	 */
427 	if (lsigctx->fpstate != NULL) {
428 		error = copyin(lsigctx->fpstate, &fpstate, sizeof(fpstate));
429 		if (error != 0) {
430 			sigexit(l, SIGILL);
431 			return error;
432 		}
433 
434 		fxsave->fx_fcw = fpstate.cwd;
435 		fxsave->fx_fsw = fpstate.swd;
436 		fxsave->fx_ftw = fpstate.twd;
437 		fxsave->fx_fop = fpstate.fop;
438 		fxsave->fx_rip = fpstate.rip;
439 		fxsave->fx_rdp = fpstate.rdp;
440 		fxsave->fx_mxcsr = fpstate.mxcsr;
441 		fxsave->fx_mxcsr_mask = fpstate.mxcsr_mask;
442 		memcpy(&fxsave->fx_st, &fpstate.st_space,
443 		    sizeof(fxsave->fx_st));
444 		memcpy(&fxsave->fx_xmm, &fpstate.xmm_space,
445 		    sizeof(fxsave->fx_xmm));
446 	}
447 
448 	/*
449 	 * And the stack
450 	 */
451 	uctx.uc_stack.ss_flags = 0;
452 	if (luctx->luc_stack.ss_flags & LINUX_SS_ONSTACK);
453 		uctx.uc_stack.ss_flags = SS_ONSTACK;
454 
455 	if (luctx->luc_stack.ss_flags & LINUX_SS_DISABLE);
456 		uctx.uc_stack.ss_flags = SS_DISABLE;
457 
458 	uctx.uc_stack.ss_sp = luctx->luc_stack.ss_sp;
459 	uctx.uc_stack.ss_size = luctx->luc_stack.ss_size;
460 
461 	/*
462 	 * And let setucontext deal with that.
463 	 */
464 	return setucontext(l, &uctx);
465 }
466 
467 int
468 linux_sys_arch_prctl(l, v, retval)
469 	struct lwp *l;
470 	void *v;
471 	register_t *retval;
472 {
473 	struct linux_sys_arch_prctl_args /* {
474 		syscallarg(int) code;
475 		syscallarg(unsigned long) addr;
476 	} */ *uap = v;
477 	struct pcb *pcb = &l->l_addr->u_pcb;
478 	struct trapframe *tf = l->l_md.md_regs;
479 	int error;
480 	uint64_t taddr;
481 
482 	switch(SCARG(uap, code)) {
483 	case LINUX_ARCH_SET_GS:
484 		taddr = SCARG(uap, addr);
485 		if (taddr >= VM_MAXUSER_ADDRESS)
486 			return EINVAL;
487 		pcb->pcb_gs = taddr;
488 		pcb->pcb_flags |= PCB_GS64;
489 		if (l == curlwp)
490 			wrmsr(MSR_KERNELGSBASE, taddr);
491 		break;
492 
493 	case LINUX_ARCH_GET_GS:
494 		if (pcb->pcb_flags & PCB_GS64)
495 			taddr = pcb->pcb_gs;
496 		else {
497 			error = memseg_baseaddr(l, tf->tf_fs, NULL, 0, &taddr);
498 			if (error != 0)
499 				return error;
500 		}
501 		error = copyout(&taddr, (char *)SCARG(uap, addr), 8);
502 		if (error != 0)
503 			return error;
504 		break;
505 
506 	case LINUX_ARCH_SET_FS:
507 		taddr = SCARG(uap, addr);
508 		if (taddr >= VM_MAXUSER_ADDRESS)
509 			return EINVAL;
510 		pcb->pcb_fs = taddr;
511 		pcb->pcb_flags |= PCB_FS64;
512 		if (l == curlwp)
513 			wrmsr(MSR_FSBASE, taddr);
514 		break;
515 
516 	case LINUX_ARCH_GET_FS:
517 		if (pcb->pcb_flags & PCB_FS64)
518 			taddr = pcb->pcb_fs;
519 		else {
520 			error = memseg_baseaddr(l, tf->tf_fs, NULL, 0, &taddr);
521 			if (error != 0)
522 				return error;
523 		}
524 		error = copyout(&taddr, (char *)SCARG(uap, addr), 8);
525 		if (error != 0)
526 			return error;
527 		break;
528 
529 	default:
530 #ifdef DEBUG_LINUX
531 		printf("linux_sys_arch_prctl: unexpected code %d\n",
532 		    SCARG(uap, code));
533 #endif
534 		return EINVAL;
535 	}
536 
537 	return 0;
538 }
539 
540 const int linux_vsyscall_to_syscall[] = {
541 	LINUX_SYS_gettimeofday,
542 	LINUX_SYS_time,
543 	LINUX_SYS_nosys,
544 	LINUX_SYS_nosys,
545 };
546 
547 int
548 linux_usertrap(struct lwp *l, vaddr_t trapaddr, void *arg)
549 {
550 	struct trapframe *tf = arg;
551 	uint64_t retaddr;
552 	int vsyscallnr;
553 
554 	/*
555 	 * Check for a vsyscall. %rip must be the fault address,
556 	 * and the address must be in the Linux vsyscall area.
557 	 * Also, vsyscalls are only done at 1024-byte boundaries.
558 	 */
559 
560 	if (__predict_true(trapaddr < LINUX_VSYSCALL_START))
561 		return 0;
562 
563 	if (trapaddr != tf->tf_rip)
564 		return 0;
565 
566 	if ((tf->tf_rip & (LINUX_VSYSCALL_SIZE - 1)) != 0)
567 		return 0;
568 
569 	vsyscallnr = (tf->tf_rip - LINUX_VSYSCALL_START) / LINUX_VSYSCALL_SIZE;
570 
571 	if (vsyscallnr > LINUX_VSYSCALL_MAXNR)
572 		return 0;
573 
574 	/*
575 	 * Get the return address from the top of the stack,
576 	 * and fix up the return address.
577 	 * This assumes the faulting instruction was callq *reg,
578 	 * which is the only way that vsyscalls are ever entered.
579 	 */
580 	if (copyin((void *)tf->tf_rsp, &retaddr, sizeof retaddr) != 0)
581 		return 0;
582 	tf->tf_rip = retaddr;
583 	tf->tf_rax = linux_vsyscall_to_syscall[vsyscallnr];
584 	tf->tf_rsp += 8;	/* "pop" the return address */
585 
586 #if 0
587 	printf("usertrap: rip %p rsp %p retaddr %p vsys %d sys %d\n",
588 	    (void *)tf->tf_rip, (void *)tf->tf_rsp, (void *)retaddr,
589 	    vsyscallnr, (int)tf->tf_rax);
590 #endif
591 
592 	(*l->l_proc->p_md.md_syscall)(tf);
593 
594 	return 1;
595 }
596 
597 static void
598 linux_buildcontext(struct lwp *l, void *catcher, void *f)
599 {
600 	struct trapframe *tf = l->l_md.md_regs;
601 
602 	tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL);
603 	tf->tf_rip = (u_int64_t)catcher;
604 	tf->tf_cs = GSEL(GUCODE_SEL, SEL_UPL);
605 	tf->tf_rflags &= ~(PSL_T|PSL_VM|PSL_AC);
606 	tf->tf_rsp = (u_int64_t)f;
607 	tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL);
608 }
609