Home | History | Annotate | Line # | Download | only in amd64
linux_machdep.c revision 1.44
      1 /*	$NetBSD: linux_machdep.c,v 1.44 2014/01/04 00:10:03 dsl 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 1.44 2014/01/04 00:10:03 dsl 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/ucontext.h>
     46 #include <sys/conf.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 #include <machine/cpufunc.h>
     55 
     56 /*
     57  * To see whether wscons is configured (for virtual console ioctl calls).
     58  */
     59 #if defined(_KERNEL_OPT)
     60 #include "wsdisplay.h"
     61 #endif
     62 #if (NWSDISPLAY > 0)
     63 #include <dev/wscons/wsconsio.h>
     64 #include <dev/wscons/wsdisplay_usl_io.h>
     65 #endif
     66 
     67 
     68 #include <compat/linux/common/linux_signal.h>
     69 #include <compat/linux/common/linux_errno.h>
     70 #include <compat/linux/common/linux_exec.h>
     71 #include <compat/linux/common/linux_ioctl.h>
     72 #include <compat/linux/common/linux_prctl.h>
     73 #include <compat/linux/common/linux_machdep.h>
     74 #include <compat/linux/common/linux_ipc.h>
     75 #include <compat/linux/common/linux_sem.h>
     76 #include <compat/linux/linux_syscall.h>
     77 #include <compat/linux/linux_syscallargs.h>
     78 
     79 static void linux_buildcontext(struct lwp *, void *, void *);
     80 
     81 void
     82 linux_setregs(struct lwp *l, struct exec_package *epp, vaddr_t stack)
     83 {
     84 	struct pcb *pcb = lwp_getpcb(l);
     85 	struct trapframe *tf;
     86 
     87 	/* If we were using the FPU, forget about it. */
     88 	if (pcb->pcb_fpcpu != NULL)
     89 		fpusave_lwp(l, 0);
     90 
     91 	l->l_md.md_flags &= ~MDL_USEDFPU;
     92 	pcb->pcb_flags = 0;
     93 	pcb->pcb_savefpu.fp_fxsave.fx_fcw = __NetBSD_NPXCW__;
     94 	pcb->pcb_savefpu.fp_fxsave.fx_mxcsr = __INITIAL_MXCSR__;
     95 	pcb->pcb_savefpu.fp_fxsave.fx_mxcsr_mask = __INITIAL_MXCSR_MASK__;
     96 
     97 	l->l_proc->p_flag &= ~PK_32;
     98 
     99 	tf = l->l_md.md_regs;
    100 	tf->tf_rax = 0;
    101 	tf->tf_rbx = 0;
    102 	tf->tf_rcx = epp->ep_entry;
    103 	tf->tf_rdx = 0;
    104 	tf->tf_rsi = 0;
    105 	tf->tf_rdi = 0;
    106 	tf->tf_rbp = 0;
    107 	tf->tf_rsp = stack;
    108 	tf->tf_r8 = 0;
    109 	tf->tf_r9 = 0;
    110 	tf->tf_r10 = 0;
    111 	tf->tf_r11 = 0;
    112 	tf->tf_r12 = 0;
    113 	tf->tf_r13 = 0;
    114 	tf->tf_r14 = 0;
    115 	tf->tf_r15 = 0;
    116 	tf->tf_rip = epp->ep_entry;
    117 	tf->tf_rflags = PSL_USERSET;
    118 	tf->tf_cs = GSEL(GUCODE_SEL, SEL_UPL);
    119 	tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL);
    120 	tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL);
    121 	tf->tf_es = 0;
    122 	cpu_fsgs_zero(l);
    123 
    124 	return;
    125 }
    126 
    127 void
    128 linux_sendsig(const ksiginfo_t *ksi, const sigset_t *mask)
    129 {
    130 	struct lwp *l = curlwp;
    131 	struct proc *p = l->l_proc;
    132 	struct pcb *pcb = lwp_getpcb(l);
    133 	struct sigacts *ps = p->p_sigacts;
    134 	int onstack, error;
    135 	int sig = ksi->ksi_signo;
    136 	struct linux_rt_sigframe *sfp, sigframe;
    137 	struct linux__fpstate *fpsp, fpstate;
    138 	struct fpreg fpregs;
    139 	struct trapframe *tf = l->l_md.md_regs;
    140 	sig_t catcher = SIGACTION(p, sig).sa_handler;
    141 	linux_sigset_t lmask;
    142 	char *sp;
    143 
    144 	/* Do we need to jump onto the signal stack? */
    145 	onstack =
    146 	    (l->l_sigstk.ss_flags & (SS_DISABLE | SS_ONSTACK)) == 0 &&
    147 	    (SIGACTION(p, sig).sa_flags & SA_ONSTACK) != 0;
    148 
    149 	/* Allocate space for the signal handler context. */
    150 	if (onstack)
    151 		sp = ((char *)l->l_sigstk.ss_sp +
    152 		    l->l_sigstk.ss_size);
    153 	else
    154 		sp = (char *)tf->tf_rsp - 128;
    155 
    156 	/*
    157 	 * Save FPU state, if any
    158 	 */
    159 	if (l->l_md.md_flags & MDL_USEDFPU) {
    160 		sp = (char *)
    161 		    (((long)sp - sizeof(struct linux__fpstate)) & ~0xfUL);
    162 		fpsp = (struct linux__fpstate *)sp;
    163 	} else
    164 		fpsp = NULL;
    165 
    166 	/*
    167 	 * Populate the rt_sigframe
    168 	 */
    169 	sp = (char *)
    170 	    ((((long)sp - sizeof(struct linux_rt_sigframe)) & ~0xfUL) - 8);
    171 	sfp = (struct linux_rt_sigframe *)sp;
    172 
    173 	memset(&sigframe, 0, sizeof(sigframe));
    174 	if (ps->sa_sigdesc[sig].sd_vers != 0)
    175 		sigframe.pretcode =
    176 		    (char *)(u_long)ps->sa_sigdesc[sig].sd_tramp;
    177 	else
    178 		sigframe.pretcode = NULL;
    179 
    180 	/*
    181 	 * The user context
    182 	 */
    183 	sigframe.uc.luc_flags = 0;
    184 	sigframe.uc.luc_link = NULL;
    185 
    186 	/* This is used regardless of SA_ONSTACK in Linux */
    187 	sigframe.uc.luc_stack.ss_sp = l->l_sigstk.ss_sp;
    188 	sigframe.uc.luc_stack.ss_size = l->l_sigstk.ss_size;
    189 	sigframe.uc.luc_stack.ss_flags = 0;
    190 	if (l->l_sigstk.ss_flags & SS_ONSTACK)
    191 		sigframe.uc.luc_stack.ss_flags |= LINUX_SS_ONSTACK;
    192 	if (l->l_sigstk.ss_flags & SS_DISABLE)
    193 		sigframe.uc.luc_stack.ss_flags |= LINUX_SS_DISABLE;
    194 
    195 	sigframe.uc.luc_mcontext.r8 = tf->tf_r8;
    196 	sigframe.uc.luc_mcontext.r9 = tf->tf_r9;
    197 	sigframe.uc.luc_mcontext.r10 = tf->tf_r10;
    198 	sigframe.uc.luc_mcontext.r11 = tf->tf_r11;
    199 	sigframe.uc.luc_mcontext.r12 = tf->tf_r12;
    200 	sigframe.uc.luc_mcontext.r13 = tf->tf_r13;
    201 	sigframe.uc.luc_mcontext.r14 = tf->tf_r14;
    202 	sigframe.uc.luc_mcontext.r15 = tf->tf_r15;
    203 	sigframe.uc.luc_mcontext.rdi = tf->tf_rdi;
    204 	sigframe.uc.luc_mcontext.rsi = tf->tf_rsi;
    205 	sigframe.uc.luc_mcontext.rbp = tf->tf_rbp;
    206 	sigframe.uc.luc_mcontext.rbx = tf->tf_rbx;
    207 	sigframe.uc.luc_mcontext.rdx = tf->tf_rdx;
    208 	sigframe.uc.luc_mcontext.rax = tf->tf_rax;
    209 	sigframe.uc.luc_mcontext.rcx = tf->tf_rcx;
    210 	sigframe.uc.luc_mcontext.rsp = tf->tf_rsp;
    211 	sigframe.uc.luc_mcontext.rip = tf->tf_rip;
    212 	sigframe.uc.luc_mcontext.eflags = tf->tf_rflags;
    213 	sigframe.uc.luc_mcontext.cs = tf->tf_cs;
    214 	sigframe.uc.luc_mcontext.gs = tf->tf_gs;
    215 	sigframe.uc.luc_mcontext.fs = tf->tf_fs;
    216 	sigframe.uc.luc_mcontext.err = tf->tf_err;
    217 	sigframe.uc.luc_mcontext.trapno = tf->tf_trapno;
    218 	native_to_linux_sigset(&lmask, mask);
    219 	sigframe.uc.luc_mcontext.oldmask = lmask.sig[0];
    220 	sigframe.uc.luc_mcontext.cr2 = (long)pcb->pcb_onfault;
    221 	sigframe.uc.luc_mcontext.fpstate = fpsp;
    222 	native_to_linux_sigset(&sigframe.uc.luc_sigmask, mask);
    223 	native_to_linux_siginfo(&sigframe.info, &ksi->ksi_info);
    224 	sendsig_reset(l, sig);
    225 	mutex_exit(p->p_lock);
    226 	error = 0;
    227 
    228 	/*
    229 	 * Save FPU state, if any
    230 	 */
    231 	if (fpsp != NULL) {
    232 		size_t fp_size = sizeof fpregs;
    233 		(void)process_read_fpregs(l, &fpregs, &fp_size);
    234 		memset(&fpstate, 0, sizeof(fpstate));
    235 		fpstate.cwd = fpregs.fp_fcw;
    236 		fpstate.swd = fpregs.fp_fsw;
    237 		fpstate.twd = fpregs.fp_ftw;
    238 		fpstate.fop = fpregs.fp_fop;
    239 		fpstate.rip = fpregs.fp_rip;
    240 		fpstate.rdp = fpregs.fp_rdp;
    241 		fpstate.mxcsr = fpregs.fp_mxcsr;
    242 		fpstate.mxcsr_mask = fpregs.fp_mxcsr_mask;
    243 		memcpy(&fpstate.st_space, &fpregs.fp_st,
    244 		    sizeof(fpstate.st_space));
    245 		memcpy(&fpstate.xmm_space, &fpregs.fp_xmm,
    246 		    sizeof(fpstate.xmm_space));
    247 		error = copyout(&fpstate, fpsp, sizeof(fpstate));
    248 	}
    249 
    250 	if (error == 0)
    251 		error = copyout(&sigframe, sp, sizeof(sigframe));
    252 
    253 	mutex_enter(p->p_lock);
    254 
    255 	if (error != 0) {
    256 		sigexit(l, SIGILL);
    257 		return;
    258 	}
    259 
    260 	linux_buildcontext(l, catcher, sp);
    261 	tf->tf_rdi = sigframe.info.lsi_signo;
    262 	tf->tf_rax = 0;
    263 	tf->tf_rsi = (long)&sfp->info;
    264 	tf->tf_rdx = (long)&sfp->uc;
    265 
    266 	/*
    267 	 * Remember we use signal stack
    268 	 */
    269 	if (onstack)
    270 		l->l_sigstk.ss_flags |= SS_ONSTACK;
    271 	return;
    272 }
    273 
    274 int
    275 linux_sys_modify_ldt(struct lwp *l, const struct linux_sys_modify_ldt_args *v, register_t *retval)
    276 {
    277 	printf("linux_sys_modify_ldt\n");
    278 	return 0;
    279 }
    280 
    281 int
    282 linux_sys_iopl(struct lwp *l, const struct linux_sys_iopl_args *v, register_t *retval)
    283 {
    284 	return 0;
    285 }
    286 
    287 int
    288 linux_sys_ioperm(struct lwp *l, const struct linux_sys_ioperm_args *v, register_t *retval)
    289 {
    290 	return 0;
    291 }
    292 
    293 dev_t
    294 linux_fakedev(dev_t dev, int raw)
    295 {
    296 
    297        extern const struct cdevsw ptc_cdevsw, pts_cdevsw;
    298        const struct cdevsw *cd = cdevsw_lookup(dev);
    299 
    300        if (raw) {
    301 #if (NWSDISPLAY > 0)
    302 	       extern const struct cdevsw wsdisplay_cdevsw;
    303 	       if (cd == &wsdisplay_cdevsw)
    304 		       return makedev(LINUX_CONS_MAJOR, (minor(dev) + 1));
    305 #endif
    306        }
    307 
    308        if (cd == &ptc_cdevsw)
    309 	       return makedev(LINUX_PTC_MAJOR, minor(dev));
    310        if (cd == &pts_cdevsw)
    311 	       return makedev(LINUX_PTS_MAJOR, minor(dev));
    312 
    313 	return ((minor(dev) & 0xff) | ((major(dev) & 0xfff) << 8)
    314 	    | (((unsigned long long int) (minor(dev) & ~0xff)) << 12)
    315 	    | (((unsigned long long int) (major(dev) & ~0xfff)) << 32));
    316 }
    317 
    318 int
    319 linux_machdepioctl(struct lwp *l, const struct linux_sys_ioctl_args *v, register_t *retval)
    320 {
    321 	return 0;
    322 }
    323 
    324 int
    325 linux_sys_rt_sigreturn(struct lwp *l, const void *v, register_t *retval)
    326 {
    327 	struct linux_ucontext *luctx;
    328 	struct trapframe *tf = l->l_md.md_regs;
    329 	struct linux_sigcontext *lsigctx;
    330 	struct linux__fpstate fpstate;
    331 	struct linux_rt_sigframe frame, *fp;
    332 	ucontext_t uctx;
    333 	mcontext_t *mctx;
    334 	struct fxsave64 *fxarea;
    335 	int error;
    336 
    337 	fp = (struct linux_rt_sigframe *)(tf->tf_rsp - 8);
    338 	if ((error = copyin(fp, &frame, sizeof(frame))) != 0) {
    339 		mutex_enter(l->l_proc->p_lock);
    340 		sigexit(l, SIGILL);
    341 		return error;
    342 	}
    343 	luctx = &frame.uc;
    344 	lsigctx = &luctx->luc_mcontext;
    345 
    346 	memset(&uctx, 0, sizeof(uctx));
    347 	mctx = (mcontext_t *)&uctx.uc_mcontext;
    348 	fxarea = (struct fxsave64 *)&mctx->__fpregs;
    349 
    350 	/*
    351 	 * Set the flags. Linux always have CPU, stack and signal state,
    352 	 * FPU is optional. uc_flags is not used to tell what we have.
    353 	 */
    354 	uctx.uc_flags = (_UC_SIGMASK|_UC_CPU|_UC_STACK|_UC_CLRSTACK);
    355 	if (lsigctx->fpstate != NULL)
    356 		uctx.uc_flags |= _UC_FPU;
    357 	uctx.uc_link = NULL;
    358 
    359 	/*
    360 	 * Signal set
    361 	 */
    362 	linux_to_native_sigset(&uctx.uc_sigmask, &luctx->luc_sigmask);
    363 
    364 	/*
    365 	 * CPU state
    366 	 */
    367 	mctx->__gregs[_REG_R8] = lsigctx->r8;
    368 	mctx->__gregs[_REG_R9] = lsigctx->r9;
    369 	mctx->__gregs[_REG_R10] = lsigctx->r10;
    370 	mctx->__gregs[_REG_R11] = lsigctx->r11;
    371 	mctx->__gregs[_REG_R12] = lsigctx->r12;
    372 	mctx->__gregs[_REG_R13] = lsigctx->r13;
    373 	mctx->__gregs[_REG_R14] = lsigctx->r14;
    374 	mctx->__gregs[_REG_R15] = lsigctx->r15;
    375 	mctx->__gregs[_REG_RDI] = lsigctx->rdi;
    376 	mctx->__gregs[_REG_RSI] = lsigctx->rsi;
    377 	mctx->__gregs[_REG_RBP] = lsigctx->rbp;
    378 	mctx->__gregs[_REG_RBX] = lsigctx->rbx;
    379 	mctx->__gregs[_REG_RAX] = lsigctx->rax;
    380 	mctx->__gregs[_REG_RDX] = lsigctx->rdx;
    381 	mctx->__gregs[_REG_RCX] = lsigctx->rcx;
    382 	mctx->__gregs[_REG_RIP] = lsigctx->rip;
    383 	mctx->__gregs[_REG_RFLAGS] = lsigctx->eflags;
    384 	mctx->__gregs[_REG_CS] = lsigctx->cs;
    385 	mctx->__gregs[_REG_GS] = lsigctx->gs;
    386 	mctx->__gregs[_REG_FS] = lsigctx->fs;
    387 	mctx->__gregs[_REG_ERR] = lsigctx->err;
    388 	mctx->__gregs[_REG_TRAPNO] = lsigctx->trapno;
    389 	mctx->__gregs[_REG_ES] = tf->tf_es;
    390 	mctx->__gregs[_REG_DS] = tf->tf_ds;
    391 	mctx->__gregs[_REG_RSP] = lsigctx->rsp; /* XXX */
    392 	mctx->__gregs[_REG_SS] = tf->tf_ss;
    393 
    394 	/*
    395 	 * FPU state
    396 	 */
    397 	if (lsigctx->fpstate != NULL) {
    398 		error = copyin(lsigctx->fpstate, &fpstate, sizeof(fpstate));
    399 		if (error != 0) {
    400 			mutex_enter(l->l_proc->p_lock);
    401 			sigexit(l, SIGILL);
    402 			return error;
    403 		}
    404 
    405 		fxarea->fx_fcw = fpstate.cwd;
    406 		fxarea->fx_fsw = fpstate.swd;
    407 		fxarea->fx_ftw = fpstate.twd;
    408 		fxarea->fx_fop = fpstate.fop;
    409 		fxarea->fx_rip = fpstate.rip;
    410 		fxarea->fx_rdp = fpstate.rdp;
    411 		fxarea->fx_mxcsr = fpstate.mxcsr;
    412 		fxarea->fx_mxcsr_mask = fpstate.mxcsr_mask;
    413 		memcpy(&fxarea->fx_st, &fpstate.st_space,
    414 		    sizeof(fxarea->fx_st));
    415 		memcpy(&fxarea->fx_xmm, &fpstate.xmm_space,
    416 		    sizeof(fxarea->fx_xmm));
    417 	}
    418 
    419 	/*
    420 	 * And the stack
    421 	 */
    422 	uctx.uc_stack.ss_flags = 0;
    423 	if (luctx->luc_stack.ss_flags & LINUX_SS_ONSTACK)
    424 		uctx.uc_stack.ss_flags |= SS_ONSTACK;
    425 
    426 	if (luctx->luc_stack.ss_flags & LINUX_SS_DISABLE)
    427 		uctx.uc_stack.ss_flags |= SS_DISABLE;
    428 
    429 	uctx.uc_stack.ss_sp = luctx->luc_stack.ss_sp;
    430 	uctx.uc_stack.ss_size = luctx->luc_stack.ss_size;
    431 
    432 	/*
    433 	 * And let setucontext deal with that.
    434 	 */
    435 	mutex_enter(l->l_proc->p_lock);
    436 	error = setucontext(l, &uctx);
    437 	mutex_exit(l->l_proc->p_lock);
    438 	if (error)
    439 		return error;
    440 
    441 	return EJUSTRETURN;
    442 }
    443 
    444 int
    445 linux_sys_arch_prctl(struct lwp *l,
    446     const struct linux_sys_arch_prctl_args *uap, register_t *retval)
    447 {
    448 	/* {
    449 		syscallarg(int) code;
    450 		syscallarg(unsigned long) addr;
    451 	} */
    452 	void *addr = (void *)SCARG(uap, addr);
    453 
    454 	switch(SCARG(uap, code)) {
    455 	case LINUX_ARCH_SET_GS:
    456 		return x86_set_sdbase(addr, 'g', l, true);
    457 
    458 	case LINUX_ARCH_GET_GS:
    459 		return x86_get_sdbase(addr, 'g');
    460 
    461 	case LINUX_ARCH_SET_FS:
    462 		return x86_set_sdbase(addr, 'f', l, true);
    463 
    464 	case LINUX_ARCH_GET_FS:
    465 		return x86_get_sdbase(addr, 'f');
    466 
    467 	default:
    468 #ifdef DEBUG_LINUX
    469 		printf("linux_sys_arch_prctl: unexpected code %d\n",
    470 		    SCARG(uap, code));
    471 #endif
    472 		return EINVAL;
    473 	}
    474 	/* NOTREACHED */
    475 }
    476 
    477 const int linux_vsyscall_to_syscall[] = {
    478 	LINUX_SYS_gettimeofday,
    479 	LINUX_SYS_time,
    480 	LINUX_SYS_nosys,	/* nosys */
    481 	LINUX_SYS_nosys,	/* nosys */
    482 };
    483 
    484 int
    485 linux_usertrap(struct lwp *l, vaddr_t trapaddr, void *arg)
    486 {
    487 	struct trapframe *tf = arg;
    488 	uint64_t retaddr;
    489 	int vsyscallnr;
    490 
    491 	/*
    492 	 * Check for a vsyscall. %rip must be the fault address,
    493 	 * and the address must be in the Linux vsyscall area.
    494 	 * Also, vsyscalls are only done at 1024-byte boundaries.
    495 	 */
    496 
    497 	if (__predict_true(trapaddr < LINUX_VSYSCALL_START))
    498 		return 0;
    499 
    500 	if (trapaddr != tf->tf_rip)
    501 		return 0;
    502 
    503 	if ((tf->tf_rip & (LINUX_VSYSCALL_SIZE - 1)) != 0)
    504 		return 0;
    505 
    506 	vsyscallnr = (tf->tf_rip - LINUX_VSYSCALL_START) / LINUX_VSYSCALL_SIZE;
    507 
    508 	if (vsyscallnr > LINUX_VSYSCALL_MAXNR)
    509 		return 0;
    510 
    511 	/*
    512 	 * Get the return address from the top of the stack,
    513 	 * and fix up the return address.
    514 	 * This assumes the faulting instruction was callq *reg,
    515 	 * which is the only way that vsyscalls are ever entered.
    516 	 */
    517 	if (copyin((void *)tf->tf_rsp, &retaddr, sizeof retaddr) != 0)
    518 		return 0;
    519 	tf->tf_rip = retaddr;
    520 	tf->tf_rax = linux_vsyscall_to_syscall[vsyscallnr];
    521 	tf->tf_rsp += 8;	/* "pop" the return address */
    522 
    523 #if 0
    524 	printf("usertrap: rip %p rsp %p retaddr %p vsys %d sys %d\n",
    525 	    (void *)tf->tf_rip, (void *)tf->tf_rsp, (void *)retaddr,
    526 	    vsyscallnr, (int)tf->tf_rax);
    527 #endif
    528 
    529 	(*l->l_proc->p_md.md_syscall)(tf);
    530 
    531 	return 1;
    532 }
    533 
    534 static void
    535 linux_buildcontext(struct lwp *l, void *catcher, void *f)
    536 {
    537 	struct trapframe *tf = l->l_md.md_regs;
    538 
    539 	tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL);
    540 	tf->tf_rip = (u_int64_t)catcher;
    541 	tf->tf_cs = GSEL(GUCODE_SEL, SEL_UPL);
    542 	tf->tf_rflags &= ~PSL_CLEARSIG;
    543 	tf->tf_rsp = (u_int64_t)f;
    544 	tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL);
    545 }
    546