Home | History | Annotate | Line # | Download | only in amd64 
      1 /*	$NetBSD: linux_syscall.c,v 1.33 2023/10/05 19:41:03 ad Exp $ */
      2 
      3 /*-
      4  * Copyright (c) 1998, 2000 The NetBSD Foundation, Inc.
      5  * All rights reserved.
      6  *
      7  * This code is derived from software contributed to The NetBSD Foundation
      8  * by Charles M. Hannum.
      9  *
     10  * Redistribution and use in source and binary forms, with or without
     11  * modification, are permitted provided that the following conditions
     12  * are met:
     13  * 1. Redistributions of source code must retain the above copyright
     14  *    notice, this list of conditions and the following disclaimer.
     15  * 2. Redistributions in binary form must reproduce the above copyright
     16  *    notice, this list of conditions and the following disclaimer in the
     17  *    documentation and/or other materials provided with the distribution.
     18  *
     19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     29  * POSSIBILITY OF SUCH DAMAGE.
     30  */
     31 
     32 #include <sys/cdefs.h>
     33 __KERNEL_RCSID(0, "$NetBSD: linux_syscall.c,v 1.33 2023/10/05 19:41:03 ad Exp $");
     34 
     35 #if defined(_KERNEL_OPT)
     36 #include "opt_compat_linux.h"
     37 #endif
     38 
     39 #include <sys/param.h>
     40 #include <sys/systm.h>
     41 #include <sys/proc.h>
     42 #include <sys/signal.h>
     43 #include <sys/syscall.h>
     44 #include <sys/syscallvar.h>
     45 
     46 #include <machine/cpu.h>
     47 #include <machine/psl.h>
     48 #include <machine/userret.h>
     49 
     50 #include <compat/linux/linux_syscall.h>
     51 #include <compat/linux/common/linux_types.h>
     52 #include <compat/linux/common/linux_errno.h>
     53 #include <compat/linux/common/linux_signal.h>
     54 #include <compat/linux/common/linux_siginfo.h>
     55 #include <compat/linux/arch/amd64/linux_siginfo.h>
     56 #include <compat/linux/arch/amd64/linux_machdep.h>
     57 
     58 void linux_syscall_intern(struct proc *);
     59 static void linux_syscall(struct trapframe *);
     60 
     61 void
     62 linux_syscall_intern(struct proc *p)
     63 {
     64 
     65 	p->p_md.md_syscall = linux_syscall;
     66 }
     67 
     68 /*
     69  * syscall(frame):
     70  *	System call request from POSIX system call gate interface to kernel.
     71  * Like trap(), argument is call by reference.
     72  */
     73 static void
     74 linux_syscall(struct trapframe *frame)
     75 {
     76 	const struct sysent *callp;
     77 	struct proc *p;
     78 	struct lwp *l;
     79 	int error;
     80 	register_t code, rval[2];
     81 	#define args (&frame->tf_rdi)
     82 
     83 	l = curlwp;
     84 	p = l->l_proc;
     85 
     86 	code = frame->tf_rax;
     87 
     88 	callp = p->p_emul->e_sysent;
     89 
     90 	code &= (LINUX_SYS_NSYSENT - 1);
     91 	callp += code;
     92 
     93 	/*
     94 	 * Linux system calls have a maximum of 6 arguments, they are
     95 	 * already adjacent in the syscall trapframe.
     96 	 */
     97 
     98 	if (__predict_false(p->p_trace_enabled || KDTRACE_ENTRY(callp->sy_entry))
     99 	    && (error = trace_enter(code, callp, args)) != 0)
    100 		goto out;
    101 
    102 	rval[0] = 0;
    103 	rval[1] = 0;
    104 	error = sy_call(callp, l, args, rval);
    105 out:
    106 	switch (error) {
    107 	case 0:
    108 		frame->tf_rax = rval[0];
    109 		break;
    110 	case ERESTART:
    111 		/*
    112 		 * The offset to adjust the PC by depends on whether we entered
    113 		 * the kernel through the trap or call gate.  We pushed the
    114 		 * size of the instruction into tf_err on entry.
    115 		 */
    116 		frame->tf_rip -= frame->tf_err;
    117 		break;
    118 	case EJUSTRETURN:
    119 		/* nothing to do */
    120 		break;
    121 	default:
    122 		error = native_to_linux_errno[error];
    123 		frame->tf_rax = error;
    124 		break;
    125 	}
    126 
    127 	if (__predict_false(p->p_trace_enabled || KDTRACE_ENTRY(callp->sy_return)))
    128 		trace_exit(code, callp, args, rval, error);
    129 
    130 	userret(l);
    131 }
    132