xref: /netbsd-src/sys/compat/linux/arch/i386/linux_machdep.c (revision 7cc2f76925f078d01ddc9e640a98f4ccfc9f8c3b)
1 /*	$NetBSD: linux_machdep.c,v 1.54 2000/12/12 15:11:56 fvdl Exp $	*/
2 
3 /*-
4  * Copyright (c) 1995, 2000 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Frank van der Linden.
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  * 3. All advertising materials mentioning features or use of this software
19  *    must display the following acknowledgement:
20  *	This product includes software developed by the NetBSD
21  *	Foundation, Inc. and its contributors.
22  * 4. Neither the name of The NetBSD Foundation nor the names of its
23  *    contributors may be used to endorse or promote products derived
24  *    from this software without specific prior written permission.
25  *
26  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
27  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
28  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
29  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
30  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36  * POSSIBILITY OF SUCH DAMAGE.
37  */
38 
39 #if defined(KERNEL) && !defined(_LKM)
40 #include "opt_vm86.h"
41 #include "opt_user_ldt.h"
42 #endif
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/signalvar.h>
47 #include <sys/kernel.h>
48 #include <sys/map.h>
49 #include <sys/proc.h>
50 #include <sys/user.h>
51 #include <sys/buf.h>
52 #include <sys/reboot.h>
53 #include <sys/conf.h>
54 #include <sys/exec.h>
55 #include <sys/file.h>
56 #include <sys/callout.h>
57 #include <sys/malloc.h>
58 #include <sys/mbuf.h>
59 #include <sys/msgbuf.h>
60 #include <sys/mount.h>
61 #include <sys/vnode.h>
62 #include <sys/device.h>
63 #include <sys/syscallargs.h>
64 #include <sys/filedesc.h>
65 #include <sys/exec_elf.h>
66 #include <sys/disklabel.h>
67 #include <miscfs/specfs/specdev.h>
68 
69 #include <compat/linux/common/linux_types.h>
70 #include <compat/linux/common/linux_signal.h>
71 #include <compat/linux/common/linux_util.h>
72 #include <compat/linux/common/linux_ioctl.h>
73 #include <compat/linux/common/linux_hdio.h>
74 #include <compat/linux/common/linux_exec.h>
75 #include <compat/linux/common/linux_machdep.h>
76 
77 #include <compat/linux/linux_syscallargs.h>
78 
79 #include <machine/cpu.h>
80 #include <machine/cpufunc.h>
81 #include <machine/psl.h>
82 #include <machine/reg.h>
83 #include <machine/segments.h>
84 #include <machine/specialreg.h>
85 #include <machine/sysarch.h>
86 #include <machine/vm86.h>
87 #include <machine/vmparam.h>
88 
89 /*
90  * To see whether wscons is configured (for virtual console ioctl calls).
91  */
92 #if defined(_KERNEL) && !defined(_LKM)
93 #include "wsdisplay.h"
94 #endif
95 #if (NWSDISPLAY > 0)
96 #include <sys/ioctl.h>
97 #include <dev/wscons/wsconsio.h>
98 #include <dev/wscons/wsdisplay_usl_io.h>
99 #if defined(_KERNEL) && !defined(_LKM)
100 #include "opt_xserver.h"
101 #endif
102 #endif
103 
104 #ifdef USER_LDT
105 #include <machine/cpu.h>
106 int linux_read_ldt __P((struct proc *, struct linux_sys_modify_ldt_args *,
107     register_t *));
108 int linux_write_ldt __P((struct proc *, struct linux_sys_modify_ldt_args *,
109     register_t *));
110 #endif
111 
112 static struct biosdisk_info *fd2biosinfo __P((struct proc *, struct file *));
113 extern struct disklist *i386_alldisks;
114 extern const char *findblkname __P((int));
115 
116 /*
117  * Deal with some i386-specific things in the Linux emulation code.
118  */
119 
120 void
121 linux_setregs(p, epp, stack)
122 	struct proc *p;
123 	struct exec_package *epp;
124 	u_long stack;
125 {
126 	struct pcb *pcb = &p->p_addr->u_pcb;
127 
128 	setregs(p, epp, stack);
129 	pcb->pcb_savefpu.sv_env.en_cw = __Linux_NPXCW__;
130 }
131 
132 /*
133  * Send an interrupt to process.
134  *
135  * Stack is set up to allow sigcode stored
136  * in u. to call routine, followed by kcall
137  * to sigreturn routine below.  After sigreturn
138  * resets the signal mask, the stack, and the
139  * frame pointer, it returns to the user
140  * specified pc, psl.
141  */
142 
143 void
144 linux_sendsig(catcher, sig, mask, code)
145 	sig_t catcher;
146 	int sig;
147 	sigset_t *mask;
148 	u_long code;
149 {
150 	struct proc *p = curproc;
151 	struct trapframe *tf;
152 	struct linux_sigframe *fp, frame;
153 	struct sigacts *psp = p->p_sigacts;
154 
155 	tf = p->p_md.md_regs;
156 
157 	/* Allocate space for the signal handler context. */
158 	/* XXX Linux doesn't support the signal stack. */
159 	fp = (struct linux_sigframe *)tf->tf_esp;
160 	fp--;
161 
162 	/* Build stack frame for signal trampoline. */
163 	frame.sf_handler = catcher;
164 	frame.sf_sig = native_to_linux_sig[sig];
165 
166 	/* Save register context. */
167 #ifdef VM86
168 	if (tf->tf_eflags & PSL_VM) {
169 		frame.sf_sc.sc_gs = tf->tf_vm86_gs;
170 		frame.sf_sc.sc_fs = tf->tf_vm86_fs;
171 		frame.sf_sc.sc_es = tf->tf_vm86_es;
172 		frame.sf_sc.sc_ds = tf->tf_vm86_ds;
173 		frame.sf_sc.sc_eflags = get_vflags(p);
174 	} else
175 #endif
176 	{
177 		__asm("movl %%gs,%w0" : "=r" (frame.sf_sc.sc_gs));
178 		__asm("movl %%fs,%w0" : "=r" (frame.sf_sc.sc_fs));
179 		frame.sf_sc.sc_es = tf->tf_es;
180 		frame.sf_sc.sc_ds = tf->tf_ds;
181 		frame.sf_sc.sc_eflags = tf->tf_eflags;
182 	}
183 	frame.sf_sc.sc_edi = tf->tf_edi;
184 	frame.sf_sc.sc_esi = tf->tf_esi;
185 	frame.sf_sc.sc_ebp = tf->tf_ebp;
186 	frame.sf_sc.sc_ebx = tf->tf_ebx;
187 	frame.sf_sc.sc_edx = tf->tf_edx;
188 	frame.sf_sc.sc_ecx = tf->tf_ecx;
189 	frame.sf_sc.sc_eax = tf->tf_eax;
190 	frame.sf_sc.sc_eip = tf->tf_eip;
191 	frame.sf_sc.sc_cs = tf->tf_cs;
192 	frame.sf_sc.sc_esp_at_signal = tf->tf_esp;
193 	frame.sf_sc.sc_ss = tf->tf_ss;
194 	frame.sf_sc.sc_err = tf->tf_err;
195 	frame.sf_sc.sc_trapno = tf->tf_trapno;
196 
197 	/* Save signal stack. */
198 	/* XXX Linux doesn't support the signal stack. */
199 
200 	/* Save signal mask. */
201 	native_to_linux_old_sigset(mask, &frame.sf_sc.sc_mask);
202 
203 	if (copyout(&frame, fp, sizeof(frame)) != 0) {
204 		/*
205 		 * Process has trashed its stack; give it an illegal
206 		 * instruction to halt it in its tracks.
207 		 */
208 		sigexit(p, SIGILL);
209 		/* NOTREACHED */
210 	}
211 
212 	/*
213 	 * Build context to run handler in.
214 	 */
215 	tf->tf_es = GSEL(GUDATA_SEL, SEL_UPL);
216 	tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL);
217 	tf->tf_eip = (int)psp->ps_sigcode;
218 	tf->tf_cs = GSEL(GUCODE_SEL, SEL_UPL);
219 	tf->tf_eflags &= ~(PSL_T|PSL_VM|PSL_AC);
220 	tf->tf_esp = (int)fp;
221 	tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL);
222 
223 	/* Remember that we're now on the signal stack. */
224 	/* XXX Linux doesn't support the signal stack. */
225 }
226 
227 /*
228  * System call to cleanup state after a signal
229  * has been taken.  Reset signal mask and
230  * stack state from context left by sendsig (above).
231  * Return to previous pc and psl as specified by
232  * context left by sendsig. Check carefully to
233  * make sure that the user has not modified the
234  * psl to gain improper privileges or to cause
235  * a machine fault.
236  */
237 int
238 linux_sys_rt_sigreturn(p, v, retval)
239 	struct proc *p;
240 	void *v;
241 	register_t *retval;
242 {
243 	/* XXX XAX write me */
244 	return(ENOSYS);
245 }
246 
247 int
248 linux_sys_sigreturn(p, v, retval)
249 	struct proc *p;
250 	void *v;
251 	register_t *retval;
252 {
253 	struct linux_sys_sigreturn_args /* {
254 		syscallarg(struct linux_sigcontext *) scp;
255 	} */ *uap = v;
256 	struct linux_sigcontext *scp, context;
257 	struct trapframe *tf;
258 	sigset_t mask;
259 
260 	/*
261 	 * The trampoline code hands us the context.
262 	 * It is unsafe to keep track of it ourselves, in the event that a
263 	 * program jumps out of a signal handler.
264 	 */
265 	scp = SCARG(uap, scp);
266 	if (copyin((caddr_t)scp, &context, sizeof(*scp)) != 0)
267 		return (EFAULT);
268 
269 	/* Restore register context. */
270 	tf = p->p_md.md_regs;
271 #ifdef VM86
272 	if (context.sc_eflags & PSL_VM) {
273 		tf->tf_vm86_gs = context.sc_gs;
274 		tf->tf_vm86_fs = context.sc_fs;
275 		tf->tf_vm86_es = context.sc_es;
276 		tf->tf_vm86_ds = context.sc_ds;
277 		set_vflags(p, context.sc_eflags);
278 	} else
279 #endif
280 	{
281 		/*
282 		 * Check for security violations.  If we're returning to
283 		 * protected mode, the CPU will validate the segment registers
284 		 * automatically and generate a trap on violations.  We handle
285 		 * the trap, rather than doing all of the checking here.
286 		 */
287 		if (((context.sc_eflags ^ tf->tf_eflags) & PSL_USERSTATIC) != 0 ||
288 		    !USERMODE(context.sc_cs, context.sc_eflags))
289 			return (EINVAL);
290 
291 		/* %fs and %gs were restored by the trampoline. */
292 		tf->tf_es = context.sc_es;
293 		tf->tf_ds = context.sc_ds;
294 		tf->tf_eflags = context.sc_eflags;
295 	}
296 	tf->tf_edi = context.sc_edi;
297 	tf->tf_esi = context.sc_esi;
298 	tf->tf_ebp = context.sc_ebp;
299 	tf->tf_ebx = context.sc_ebx;
300 	tf->tf_edx = context.sc_edx;
301 	tf->tf_ecx = context.sc_ecx;
302 	tf->tf_eax = context.sc_eax;
303 	tf->tf_eip = context.sc_eip;
304 	tf->tf_cs = context.sc_cs;
305 	tf->tf_esp = context.sc_esp_at_signal;
306 	tf->tf_ss = context.sc_ss;
307 
308 	/* Restore signal stack. */
309 	p->p_sigacts->ps_sigstk.ss_flags &= ~SS_ONSTACK;
310 
311 	/* Restore signal mask. */
312 	linux_old_to_native_sigset(&context.sc_mask, &mask);
313 	(void) sigprocmask1(p, SIG_SETMASK, &mask, 0);
314 
315 	return (EJUSTRETURN);
316 }
317 
318 #ifdef USER_LDT
319 
320 int
321 linux_read_ldt(p, uap, retval)
322 	struct proc *p;
323 	struct linux_sys_modify_ldt_args /* {
324 		syscallarg(int) func;
325 		syscallarg(void *) ptr;
326 		syscallarg(size_t) bytecount;
327 	} */ *uap;
328 	register_t *retval;
329 {
330 	struct i386_get_ldt_args gl;
331 	int error;
332 	caddr_t sg;
333 	char *parms;
334 
335 	sg = stackgap_init(p->p_emul);
336 
337 	gl.start = 0;
338 	gl.desc = SCARG(uap, ptr);
339 	gl.num = SCARG(uap, bytecount) / sizeof(union descriptor);
340 
341 	parms = stackgap_alloc(&sg, sizeof(gl));
342 
343 	if ((error = copyout(&gl, parms, sizeof(gl))) != 0)
344 		return (error);
345 
346 	if ((error = i386_get_ldt(p, parms, retval)) != 0)
347 		return (error);
348 
349 	*retval *= sizeof(union descriptor);
350 	return (0);
351 }
352 
353 struct linux_ldt_info {
354 	u_int entry_number;
355 	u_long base_addr;
356 	u_int limit;
357 	u_int seg_32bit:1;
358 	u_int contents:2;
359 	u_int read_exec_only:1;
360 	u_int limit_in_pages:1;
361 	u_int seg_not_present:1;
362 };
363 
364 int
365 linux_write_ldt(p, uap, retval)
366 	struct proc *p;
367 	struct linux_sys_modify_ldt_args /* {
368 		syscallarg(int) func;
369 		syscallarg(void *) ptr;
370 		syscallarg(size_t) bytecount;
371 	} */ *uap;
372 	register_t *retval;
373 {
374 	struct linux_ldt_info ldt_info;
375 	struct segment_descriptor sd;
376 	struct i386_set_ldt_args sl;
377 	int error;
378 	caddr_t sg;
379 	char *parms;
380 
381 	if (SCARG(uap, bytecount) != sizeof(ldt_info))
382 		return (EINVAL);
383 	if ((error = copyin(SCARG(uap, ptr), &ldt_info, sizeof(ldt_info))) != 0)
384 		return error;
385 	if (ldt_info.contents == 3)
386 		return (EINVAL);
387 
388 	sg = stackgap_init(p->p_emul);
389 
390 	sd.sd_lobase = ldt_info.base_addr & 0xffffff;
391 	sd.sd_hibase = (ldt_info.base_addr >> 24) & 0xff;
392 	sd.sd_lolimit = ldt_info.limit & 0xffff;
393 	sd.sd_hilimit = (ldt_info.limit >> 16) & 0xf;
394 	sd.sd_type =
395 	    16 | (ldt_info.contents << 2) | (!ldt_info.read_exec_only << 1);
396 	sd.sd_dpl = SEL_UPL;
397 	sd.sd_p = !ldt_info.seg_not_present;
398 	sd.sd_def32 = ldt_info.seg_32bit;
399 	sd.sd_gran = ldt_info.limit_in_pages;
400 
401 	sl.start = ldt_info.entry_number;
402 	sl.desc = stackgap_alloc(&sg, sizeof(sd));
403 	sl.num = 1;
404 
405 #if 0
406 	printf("linux_write_ldt: idx=%d, base=%x, limit=%x\n",
407 	    ldt_info.entry_number, ldt_info.base_addr, ldt_info.limit);
408 #endif
409 
410 	parms = stackgap_alloc(&sg, sizeof(sl));
411 
412 	if ((error = copyout(&sd, sl.desc, sizeof(sd))) != 0)
413 		return (error);
414 	if ((error = copyout(&sl, parms, sizeof(sl))) != 0)
415 		return (error);
416 
417 	if ((error = i386_set_ldt(p, parms, retval)) != 0)
418 		return (error);
419 
420 	*retval = 0;
421 	return (0);
422 }
423 
424 #endif /* USER_LDT */
425 
426 int
427 linux_sys_modify_ldt(p, v, retval)
428 	struct proc *p;
429 	void *v;
430 	register_t *retval;
431 {
432 	struct linux_sys_modify_ldt_args /* {
433 		syscallarg(int) func;
434 		syscallarg(void *) ptr;
435 		syscallarg(size_t) bytecount;
436 	} */ *uap = v;
437 
438 	switch (SCARG(uap, func)) {
439 #ifdef USER_LDT
440 	case 0:
441 		return (linux_read_ldt(p, uap, retval));
442 
443 	case 1:
444 		return (linux_write_ldt(p, uap, retval));
445 #endif /* USER_LDT */
446 
447 	default:
448 		return (ENOSYS);
449 	}
450 }
451 
452 /*
453  * XXX Pathetic hack to make svgalib work. This will fake the major
454  * device number of an opened VT so that svgalib likes it. grmbl.
455  * Should probably do it 'wrong the right way' and use a mapping
456  * array for all major device numbers, and map linux_mknod too.
457  */
458 dev_t
459 linux_fakedev(dev)
460 	dev_t dev;
461 {
462 #if (NWSDISPLAY > 0)
463 	if (major(dev) == NETBSD_WSCONS_MAJOR)
464 		return makedev(LINUX_CONS_MAJOR, (minor(dev) + 1));
465 #endif
466 	return dev;
467 }
468 
469 #if (NWSDISPLAY > 0)
470 /*
471  * That's not complete, but enough to get an X server running.
472  */
473 #define NR_KEYS 128
474 static u_short plain_map[NR_KEYS] = {
475 	0x0200,	0x001b,	0x0031,	0x0032,	0x0033,	0x0034,	0x0035,	0x0036,
476 	0x0037,	0x0038,	0x0039,	0x0030,	0x002d,	0x003d,	0x007f,	0x0009,
477 	0x0b71,	0x0b77,	0x0b65,	0x0b72,	0x0b74,	0x0b79,	0x0b75,	0x0b69,
478 	0x0b6f,	0x0b70,	0x005b,	0x005d,	0x0201,	0x0702,	0x0b61,	0x0b73,
479 	0x0b64,	0x0b66,	0x0b67,	0x0b68,	0x0b6a,	0x0b6b,	0x0b6c,	0x003b,
480 	0x0027,	0x0060,	0x0700,	0x005c,	0x0b7a,	0x0b78,	0x0b63,	0x0b76,
481 	0x0b62,	0x0b6e,	0x0b6d,	0x002c,	0x002e,	0x002f,	0x0700,	0x030c,
482 	0x0703,	0x0020,	0x0207,	0x0100,	0x0101,	0x0102,	0x0103,	0x0104,
483 	0x0105,	0x0106,	0x0107,	0x0108,	0x0109,	0x0208,	0x0209,	0x0307,
484 	0x0308,	0x0309,	0x030b,	0x0304,	0x0305,	0x0306,	0x030a,	0x0301,
485 	0x0302,	0x0303,	0x0300,	0x0310,	0x0206,	0x0200,	0x003c,	0x010a,
486 	0x010b,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,
487 	0x030e,	0x0702,	0x030d,	0x001c,	0x0701,	0x0205,	0x0114,	0x0603,
488 	0x0118,	0x0601,	0x0602,	0x0117,	0x0600,	0x0119,	0x0115,	0x0116,
489 	0x011a,	0x010c,	0x010d,	0x011b,	0x011c,	0x0110,	0x0311,	0x011d,
490 	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,
491 }, shift_map[NR_KEYS] = {
492 	0x0200,	0x001b,	0x0021,	0x0040,	0x0023,	0x0024,	0x0025,	0x005e,
493 	0x0026,	0x002a,	0x0028,	0x0029,	0x005f,	0x002b,	0x007f,	0x0009,
494 	0x0b51,	0x0b57,	0x0b45,	0x0b52,	0x0b54,	0x0b59,	0x0b55,	0x0b49,
495 	0x0b4f,	0x0b50,	0x007b,	0x007d,	0x0201,	0x0702,	0x0b41,	0x0b53,
496 	0x0b44,	0x0b46,	0x0b47,	0x0b48,	0x0b4a,	0x0b4b,	0x0b4c,	0x003a,
497 	0x0022,	0x007e,	0x0700,	0x007c,	0x0b5a,	0x0b58,	0x0b43,	0x0b56,
498 	0x0b42,	0x0b4e,	0x0b4d,	0x003c,	0x003e,	0x003f,	0x0700,	0x030c,
499 	0x0703,	0x0020,	0x0207,	0x010a,	0x010b,	0x010c,	0x010d,	0x010e,
500 	0x010f,	0x0110,	0x0111,	0x0112,	0x0113,	0x0213,	0x0203,	0x0307,
501 	0x0308,	0x0309,	0x030b,	0x0304,	0x0305,	0x0306,	0x030a,	0x0301,
502 	0x0302,	0x0303,	0x0300,	0x0310,	0x0206,	0x0200,	0x003e,	0x010a,
503 	0x010b,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,
504 	0x030e,	0x0702,	0x030d,	0x0200,	0x0701,	0x0205,	0x0114,	0x0603,
505 	0x020b,	0x0601,	0x0602,	0x0117,	0x0600,	0x020a,	0x0115,	0x0116,
506 	0x011a,	0x010c,	0x010d,	0x011b,	0x011c,	0x0110,	0x0311,	0x011d,
507 	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,
508 }, altgr_map[NR_KEYS] = {
509 	0x0200,	0x0200,	0x0200,	0x0040,	0x0200,	0x0024,	0x0200,	0x0200,
510 	0x007b,	0x005b,	0x005d,	0x007d,	0x005c,	0x0200,	0x0200,	0x0200,
511 	0x0b71,	0x0b77,	0x0918,	0x0b72,	0x0b74,	0x0b79,	0x0b75,	0x0b69,
512 	0x0b6f,	0x0b70,	0x0200,	0x007e,	0x0201,	0x0702,	0x0914,	0x0b73,
513 	0x0917,	0x0919,	0x0b67,	0x0b68,	0x0b6a,	0x0b6b,	0x0b6c,	0x0200,
514 	0x0200,	0x0200,	0x0700,	0x0200,	0x0b7a,	0x0b78,	0x0916,	0x0b76,
515 	0x0915,	0x0b6e,	0x0b6d,	0x0200,	0x0200,	0x0200,	0x0700,	0x030c,
516 	0x0703,	0x0200,	0x0207,	0x050c,	0x050d,	0x050e,	0x050f,	0x0510,
517 	0x0511,	0x0512,	0x0513,	0x0514,	0x0515,	0x0208,	0x0202,	0x0911,
518 	0x0912,	0x0913,	0x030b,	0x090e,	0x090f,	0x0910,	0x030a,	0x090b,
519 	0x090c,	0x090d,	0x090a,	0x0310,	0x0206,	0x0200,	0x007c,	0x0516,
520 	0x0517,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,
521 	0x030e,	0x0702,	0x030d,	0x0200,	0x0701,	0x0205,	0x0114,	0x0603,
522 	0x0118,	0x0601,	0x0602,	0x0117,	0x0600,	0x0119,	0x0115,	0x0116,
523 	0x011a,	0x010c,	0x010d,	0x011b,	0x011c,	0x0110,	0x0311,	0x011d,
524 	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,
525 }, ctrl_map[NR_KEYS] = {
526 	0x0200,	0x0200,	0x0200,	0x0000,	0x001b,	0x001c,	0x001d,	0x001e,
527 	0x001f,	0x007f,	0x0200,	0x0200,	0x001f,	0x0200,	0x0008,	0x0200,
528 	0x0011,	0x0017,	0x0005,	0x0012,	0x0014,	0x0019,	0x0015,	0x0009,
529 	0x000f,	0x0010,	0x001b,	0x001d,	0x0201,	0x0702,	0x0001,	0x0013,
530 	0x0004,	0x0006,	0x0007,	0x0008,	0x000a,	0x000b,	0x000c,	0x0200,
531 	0x0007,	0x0000,	0x0700,	0x001c,	0x001a,	0x0018,	0x0003,	0x0016,
532 	0x0002,	0x000e,	0x000d,	0x0200,	0x020e,	0x007f,	0x0700,	0x030c,
533 	0x0703,	0x0000,	0x0207,	0x0100,	0x0101,	0x0102,	0x0103,	0x0104,
534 	0x0105,	0x0106,	0x0107,	0x0108,	0x0109,	0x0208,	0x0204,	0x0307,
535 	0x0308,	0x0309,	0x030b,	0x0304,	0x0305,	0x0306,	0x030a,	0x0301,
536 	0x0302,	0x0303,	0x0300,	0x0310,	0x0206,	0x0200,	0x0200,	0x010a,
537 	0x010b,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,
538 	0x030e,	0x0702,	0x030d,	0x001c,	0x0701,	0x0205,	0x0114,	0x0603,
539 	0x0118,	0x0601,	0x0602,	0x0117,	0x0600,	0x0119,	0x0115,	0x0116,
540 	0x011a,	0x010c,	0x010d,	0x011b,	0x011c,	0x0110,	0x0311,	0x011d,
541 	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,	0x0200,
542 };
543 
544 u_short *linux_keytabs[] = {
545 	plain_map, shift_map, altgr_map, altgr_map, ctrl_map
546 };
547 #endif
548 
549 static struct biosdisk_info *
550 fd2biosinfo(p, fp)
551 	struct proc *p;
552 	struct file *fp;
553 {
554 	struct vnode *vp;
555 	const char *blkname;
556 	char diskname[16];
557 	int i;
558 	struct nativedisk_info *nip;
559 	struct disklist *dl = i386_alldisks;
560 
561 	if (fp->f_type != DTYPE_VNODE)
562 		return NULL;
563 	vp = (struct vnode *)fp->f_data;
564 
565 	if (vp->v_type != VBLK)
566 		return NULL;
567 
568 	blkname = findblkname(major(vp->v_rdev));
569 	snprintf(diskname, sizeof diskname, "%s%u", blkname,
570 	    DISKUNIT(vp->v_rdev));
571 
572 	for (i = 0; i < dl->dl_nnativedisks; i++) {
573 		nip = &dl->dl_nativedisks[i];
574 		if (strcmp(diskname, nip->ni_devname))
575 			continue;
576 		if (nip->ni_nmatches != 0)
577 			return &dl->dl_biosdisks[nip->ni_biosmatches[0]];
578 	}
579 
580 	return NULL;
581 }
582 
583 
584 /*
585  * We come here in a last attempt to satisfy a Linux ioctl() call
586  */
587 int
588 linux_machdepioctl(p, v, retval)
589 	struct proc *p;
590 	void *v;
591 	register_t *retval;
592 {
593 	struct linux_sys_ioctl_args /* {
594 		syscallarg(int) fd;
595 		syscallarg(u_long) com;
596 		syscallarg(caddr_t) data;
597 	} */ *uap = v;
598 	struct sys_ioctl_args bia;
599 	u_long com;
600 	int error, error1;
601 #if (NWSDISPLAY > 0)
602 	struct vt_mode lvt;
603 	caddr_t bvtp, sg;
604 	struct kbentry kbe;
605 #endif
606 	struct linux_hd_geometry hdg;
607 	struct linux_hd_big_geometry hdg_big;
608 	struct biosdisk_info *bip;
609 	struct filedesc *fdp;
610 	struct file *fp;
611 	int fd;
612 	struct disklabel label, *labp;
613 	struct partinfo partp;
614 	int (*ioctlf) __P((struct file *, u_long, caddr_t, struct proc *));
615 	u_long start, biostotal, realtotal;
616 	u_char heads, sectors;
617 	u_int cylinders;
618 
619 	fd = SCARG(uap, fd);
620 	SCARG(&bia, fd) = fd;
621 	SCARG(&bia, data) = SCARG(uap, data);
622 	com = SCARG(uap, com);
623 
624 	fdp = p->p_fd;
625 
626 	if ((u_int)fd >= fdp->fd_nfiles ||
627 	    (fp = fdp->fd_ofiles[fd]) == NULL ||
628 	    (fp->f_iflags & FIF_WANTCLOSE) != 0)
629 		return (EBADF);
630 
631 	switch (com) {
632 #if (NWSDISPLAY > 0)
633 	case LINUX_KDGKBMODE:
634 		com = KDGKBMODE;
635 		break;
636 	case LINUX_KDSKBMODE:
637 		com = KDSKBMODE;
638 		if ((unsigned)SCARG(uap, data) == LINUX_K_MEDIUMRAW)
639 			SCARG(&bia, data) = (caddr_t)K_RAW;
640 		break;
641 	case LINUX_KDMKTONE:
642 		com = KDMKTONE;
643 		break;
644 	case LINUX_KDSETMODE:
645 		com = KDSETMODE;
646 		break;
647 	case LINUX_KDENABIO:
648 		com = KDENABIO;
649 		break;
650 	case LINUX_KDDISABIO:
651 		com = KDDISABIO;
652 		break;
653 	case LINUX_KDGETLED:
654 		com = KDGETLED;
655 		break;
656 	case LINUX_KDSETLED:
657 		com = KDSETLED;
658 		break;
659 	case LINUX_VT_OPENQRY:
660 		com = VT_OPENQRY;
661 		break;
662 	case LINUX_VT_GETMODE:
663 		SCARG(&bia, com) = VT_GETMODE;
664 		if ((error = sys_ioctl(p, &bia, retval)))
665 			return error;
666 		if ((error = copyin(SCARG(uap, data), (caddr_t)&lvt,
667 		    sizeof (struct vt_mode))))
668 			return error;
669 		lvt.relsig = native_to_linux_sig[lvt.relsig];
670 		lvt.acqsig = native_to_linux_sig[lvt.acqsig];
671 		lvt.frsig = native_to_linux_sig[lvt.frsig];
672 		return copyout((caddr_t)&lvt, SCARG(uap, data),
673 		    sizeof (struct vt_mode));
674 	case LINUX_VT_SETMODE:
675 		com = VT_SETMODE;
676 		if ((error = copyin(SCARG(uap, data), (caddr_t)&lvt,
677 		    sizeof (struct vt_mode))))
678 			return error;
679 		lvt.relsig = linux_to_native_sig[lvt.relsig];
680 		lvt.acqsig = linux_to_native_sig[lvt.acqsig];
681 		lvt.frsig = linux_to_native_sig[lvt.frsig];
682 		sg = stackgap_init(p->p_emul);
683 		bvtp = stackgap_alloc(&sg, sizeof (struct vt_mode));
684 		if ((error = copyout(&lvt, bvtp, sizeof (struct vt_mode))))
685 			return error;
686 		SCARG(&bia, data) = bvtp;
687 		break;
688 	case LINUX_VT_DISALLOCATE:
689 		/* XXX should use WSDISPLAYIO_DELSCREEN */
690 		return 0;
691 	case LINUX_VT_RELDISP:
692 		com = VT_RELDISP;
693 		break;
694 	case LINUX_VT_ACTIVATE:
695 		com = VT_ACTIVATE;
696 		break;
697 	case LINUX_VT_WAITACTIVE:
698 		com = VT_WAITACTIVE;
699 		break;
700 	case LINUX_VT_GETSTATE:
701 		com = VT_GETSTATE;
702 		break;
703 	case LINUX_KDGKBTYPE:
704 		/* This is what Linux does. */
705 		return (subyte(SCARG(uap, data), KB_101));
706 	case LINUX_KDGKBENT:
707 		/*
708 		 * The Linux KDGKBENT ioctl is different from the
709 		 * SYSV original. So we handle it in machdep code.
710 		 * XXX We should use keyboard mapping information
711 		 * from wsdisplay, but this would be expensive.
712 		 */
713 		if ((error = copyin(SCARG(uap, data), &kbe,
714 				    sizeof(struct kbentry))))
715 			return (error);
716 		if (kbe.kb_table >= sizeof(linux_keytabs) / sizeof(u_short *)
717 		    || kbe.kb_index >= NR_KEYS)
718 			return (EINVAL);
719 		kbe.kb_value = linux_keytabs[kbe.kb_table][kbe.kb_index];
720 		return (copyout(&kbe, SCARG(uap, data),
721 				sizeof(struct kbentry)));
722 #endif
723 	case LINUX_HDIO_GETGEO:
724 	case LINUX_HDIO_GETGEO_BIG:
725 		/*
726 		 * Try to mimic Linux behaviour: return the BIOS geometry
727 		 * if possible (extending its # of cylinders if it's beyond
728 		 * the 1023 limit), fall back to the MI geometry (i.e.
729 		 * the real geometry) if not found, by returning an
730 		 * error. See common/linux_hdio.c
731 		 */
732 		FILE_USE(fp);
733 		bip = fd2biosinfo(p, fp);
734 		ioctlf = fp->f_ops->fo_ioctl;
735 		error = ioctlf(fp, DIOCGDEFLABEL, (caddr_t)&label, p);
736 		error1 = ioctlf(fp, DIOCGPART, (caddr_t)&partp, p);
737 		FILE_UNUSE(fp, p);
738 		if (error != 0 && error1 != 0)
739 			return error1;
740 		labp = error != 0 ? &label : partp.disklab;
741 		start = error1 != 0 ? partp.part->p_offset : 0;
742 		if (bip != NULL && bip->bi_head != 0 && bip->bi_sec != 0
743 		    && bip->bi_cyl != 0) {
744 			heads = bip->bi_head;
745 			sectors = bip->bi_sec;
746 			cylinders = bip->bi_cyl;
747 			biostotal = heads * sectors * cylinders;
748 			realtotal = labp->d_ntracks * labp->d_nsectors *
749 			    labp->d_ncylinders;
750 			if (realtotal > biostotal)
751 				cylinders = realtotal / (heads * sectors);
752 		} else {
753 			heads = labp->d_ntracks;
754 			cylinders = labp->d_ncylinders;
755 			sectors = labp->d_nsectors;
756 		}
757 		if (com == LINUX_HDIO_GETGEO) {
758 			hdg.start = start;
759 			hdg.heads = heads;
760 			hdg.cylinders = cylinders;
761 			hdg.sectors = sectors;
762 			return copyout(&hdg, SCARG(uap, data), sizeof hdg);
763 		} else {
764 			hdg_big.start = start;
765 			hdg_big.heads = heads;
766 			hdg_big.cylinders = cylinders;
767 			hdg_big.sectors = sectors;
768 			return copyout(&hdg_big, SCARG(uap, data),
769 			    sizeof hdg_big);
770 		}
771 		return 0;
772 
773 	default:
774 		/*
775 		 * XXX just pass all for LKMs?
776 		 * XXX this means that device drivers specifically dealing
777 	 	 * XXX with Linux binaries will need to do copyin/copyout
778 		 * XXX handling themselves.
779 		 */
780 		if (com > LINUX_IOCTL_MIN_PASS &&
781 		    com < LINUX_IOCTL_MAX_PASS) {
782 			FILE_USE(fp);
783 			ioctlf = fp->f_ops->fo_ioctl;
784 			error = ioctlf(fp, com, SCARG(uap, data), p);
785 		}
786 		if (error == EINVAL)
787 			printf("linux_machdepioctl: invalid ioctl %08lx\n",
788 			    com);
789 		return error;
790 	}
791 	SCARG(&bia, com) = com;
792 	return sys_ioctl(p, &bia, retval);
793 }
794 
795 /*
796  * Set I/O permissions for a process. Just set the maximum level
797  * right away (ignoring the argument), otherwise we would have
798  * to rely on I/O permission maps, which are not implemented.
799  */
800 int
801 linux_sys_iopl(p, v, retval)
802 	struct proc *p;
803 	void *v;
804 	register_t *retval;
805 {
806 #if 0
807 	struct linux_sys_iopl_args /* {
808 		syscallarg(int) level;
809 	} */ *uap = v;
810 #endif
811 	struct trapframe *fp = p->p_md.md_regs;
812 
813 	if (suser(p->p_ucred, &p->p_acflag) != 0)
814 		return EPERM;
815 	fp->tf_eflags |= PSL_IOPL;
816 	*retval = 0;
817 	return 0;
818 }
819 
820 /*
821  * See above. If a root process tries to set access to an I/O port,
822  * just let it have the whole range.
823  */
824 int
825 linux_sys_ioperm(p, v, retval)
826 	struct proc *p;
827 	void *v;
828 	register_t *retval;
829 {
830 	struct linux_sys_ioperm_args /* {
831 		syscallarg(unsigned int) lo;
832 		syscallarg(unsigned int) hi;
833 		syscallarg(int) val;
834 	} */ *uap = v;
835 	struct trapframe *fp = p->p_md.md_regs;
836 
837 	if (suser(p->p_ucred, &p->p_acflag) != 0)
838 		return EPERM;
839 	if (SCARG(uap, val))
840 		fp->tf_eflags |= PSL_IOPL;
841 	*retval = 0;
842 	return 0;
843 }
844