xref: /netbsd-src/sys/compat/linux/arch/i386/linux_machdep.c (revision 9573504567626934c7ee01c7dce0c4bb1dfe7403)
1 /*	$NetBSD: linux_machdep.c,v 1.23 1995/10/10 04:54:25 mycroft Exp $	*/
2 
3 /*
4  * Copyright (c) 1995 Frank van der Linden
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. All advertising materials mentioning features or use of this software
16  *    must display the following acknowledgement:
17  *      This product includes software developed for the NetBSD Project
18  *      by Frank van der Linden
19  * 4. The name of the author may not be used to endorse or promote products
20  *    derived from this software without specific prior written permission
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
23  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
24  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
25  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
26  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
27  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
31  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32  */
33 
34 #include <sys/param.h>
35 #include <sys/systm.h>
36 #include <sys/signalvar.h>
37 #include <sys/kernel.h>
38 #include <sys/map.h>
39 #include <sys/proc.h>
40 #include <sys/user.h>
41 #include <sys/buf.h>
42 #include <sys/reboot.h>
43 #include <sys/conf.h>
44 #include <sys/file.h>
45 #include <sys/callout.h>
46 #include <sys/malloc.h>
47 #include <sys/mbuf.h>
48 #include <sys/msgbuf.h>
49 #include <sys/mount.h>
50 #include <sys/vnode.h>
51 #include <sys/device.h>
52 #include <sys/sysctl.h>
53 #include <sys/syscallargs.h>
54 #include <sys/filedesc.h>
55 
56 #include <compat/linux/linux_types.h>
57 #include <compat/linux/linux_signal.h>
58 #include <compat/linux/linux_syscallargs.h>
59 #include <compat/linux/linux_util.h>
60 
61 #include <machine/cpu.h>
62 #include <machine/cpufunc.h>
63 #include <machine/psl.h>
64 #include <machine/reg.h>
65 #include <machine/segments.h>
66 #include <machine/specialreg.h>
67 #include <machine/sysarch.h>
68 #include <machine/linux_machdep.h>
69 
70 /*
71  * To see whether pcvt is configured (for virtual console ioctl calls).
72  */
73 #include "vt.h"
74 #if NVT > 0
75 #include <arch/i386/isa/pcvt/pcvt_ioctl.h>
76 #endif
77 
78 /*
79  * Deal with some i386-specific things in the Linux emulation code.
80  * This means just signals for now, will include stuff like
81  * I/O map permissions and V86 mode sometime.
82  */
83 
84 /*
85  * Send an interrupt to process.
86  *
87  * Stack is set up to allow sigcode stored
88  * in u. to call routine, followed by kcall
89  * to sigreturn routine below.  After sigreturn
90  * resets the signal mask, the stack, and the
91  * frame pointer, it returns to the user
92  * specified pc, psl.
93  */
94 
95 void
96 linux_sendsig(catcher, sig, mask, code)
97 	sig_t catcher;
98 	int sig, mask;
99 	u_long code;
100 {
101 	register struct proc *p = curproc;
102 	register struct trapframe *tf;
103 	struct linux_sigframe *fp, frame;
104 	struct sigacts *psp = p->p_sigacts;
105 	int oonstack;
106 	extern char linux_sigcode[], linux_esigcode[];
107 
108 	tf = p->p_md.md_regs;
109 	oonstack = psp->ps_sigstk.ss_flags & SS_ONSTACK;
110 
111 	/*
112 	 * Allocate space for the signal handler context.
113 	 */
114 	if ((psp->ps_flags & SAS_ALTSTACK) && !oonstack &&
115 	    (psp->ps_sigonstack & sigmask(sig))) {
116 		fp = (struct linux_sigframe *)(psp->ps_sigstk.ss_base +
117 		    psp->ps_sigstk.ss_size - sizeof(struct linux_sigframe));
118 		psp->ps_sigstk.ss_flags |= SS_ONSTACK;
119 	} else {
120 		fp = (struct linux_sigframe *)tf->tf_esp - 1;
121 	}
122 
123 	frame.sf_handler = catcher;
124 	frame.sf_sig = bsd_to_linux_sig[sig];
125 
126 	/*
127 	 * Build the signal context to be used by sigreturn.
128 	 */
129 	frame.sf_sc.sc_mask   = mask;
130 #ifdef VM86
131 	if (tf->tf_eflags & PSL_VM) {
132 		frame.sf_sc.sc_gs = tf->tf_vm86_gs;
133 		frame.sf_sc.sc_fs = tf->tf_vm86_fs;
134 		frame.sf_sc.sc_es = tf->tf_vm86_es;
135 		frame.sf_sc.sc_ds = tf->tf_vm86_ds;
136 	} else
137 #endif
138 	{
139 		__asm("movl %%gs,%w0" : "=r" (frame.sf_sc.sc_gs));
140 		__asm("movl %%fs,%w0" : "=r" (frame.sf_sc.sc_fs));
141 		frame.sf_sc.sc_es = tf->tf_es;
142 		frame.sf_sc.sc_ds = tf->tf_ds;
143 	}
144 	frame.sf_sc.sc_edi    = tf->tf_edi;
145 	frame.sf_sc.sc_esi    = tf->tf_esi;
146 	frame.sf_sc.sc_ebp    = tf->tf_ebp;
147 	frame.sf_sc.sc_ebx    = tf->tf_ebx;
148 	frame.sf_sc.sc_edx    = tf->tf_edx;
149 	frame.sf_sc.sc_ecx    = tf->tf_ecx;
150 	frame.sf_sc.sc_eax    = tf->tf_eax;
151 	frame.sf_sc.sc_eip    = tf->tf_eip;
152 	frame.sf_sc.sc_cs     = tf->tf_cs;
153 	frame.sf_sc.sc_eflags = tf->tf_eflags;
154 	frame.sf_sc.sc_esp_at_signal = tf->tf_esp;
155 	frame.sf_sc.sc_ss     = tf->tf_ss;
156 	frame.sf_sc.sc_err    = tf->tf_err;
157 	frame.sf_sc.sc_trapno = tf->tf_trapno;
158 
159 	if (copyout(&frame, fp, sizeof(frame)) != 0) {
160 		/*
161 		 * Process has trashed its stack; give it an illegal
162 		 * instruction to halt it in its tracks.
163 		 */
164 		sigexit(p, SIGILL);
165 		/* NOTREACHED */
166 	}
167 
168 	/*
169 	 * Build context to run handler in.
170 	 */
171 	tf->tf_esp = (int)fp;
172 	tf->tf_eip = (int)(((char *)PS_STRINGS) -
173 	     (linux_esigcode - linux_sigcode));
174 #ifdef VM86
175 	tf->tf_eflags &= ~PSL_VM;
176 #endif
177 	tf->tf_cs = GSEL(GUCODE_SEL, SEL_UPL);
178 	tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL);
179 	tf->tf_es = GSEL(GUDATA_SEL, SEL_UPL);
180 	tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL);
181 }
182 
183 /*
184  * System call to cleanup state after a signal
185  * has been taken.  Reset signal mask and
186  * stack state from context left by sendsig (above).
187  * Return to previous pc and psl as specified by
188  * context left by sendsig. Check carefully to
189  * make sure that the user has not modified the
190  * psl to gain improper privileges or to cause
191  * a machine fault.
192  */
193 int
194 linux_sys_sigreturn(p, v, retval)
195 	struct proc *p;
196 	void *v;
197 	register_t *retval;
198 {
199 	struct linux_sys_sigreturn_args /* {
200 		syscallarg(struct linux_sigcontext *) scp;
201 	} */ *uap = v;
202 	struct linux_sigcontext *scp, context;
203 	register struct trapframe *tf;
204 
205 	tf = p->p_md.md_regs;
206 
207 	/*
208 	 * The trampoline code hands us the context.
209 	 * It is unsafe to keep track of it ourselves, in the event that a
210 	 * program jumps out of a signal handler.
211 	 */
212 	scp = SCARG(uap, scp);
213 	if (copyin((caddr_t)scp, &context, sizeof(*scp)) != 0)
214 		return (EFAULT);
215 
216 	/*
217 	 * Check for security violations.
218 	 */
219 	if (((context.sc_eflags ^ tf->tf_eflags) & PSL_USERSTATIC) != 0 ||
220 	    !USERMODE(context.sc_cs, context.sc_eflags))
221 		return (EINVAL);
222 
223 	p->p_sigacts->ps_sigstk.ss_flags &= ~SS_ONSTACK;
224 	p->p_sigmask = context.sc_mask & ~sigcantmask;
225 
226 	/*
227 	 * Restore signal context.
228 	 */
229 #ifdef VM86
230 	if (context.sc_eflags & PSL_VM) {
231 		tf->tf_vm86_gs = context.sc_gs;
232 		tf->tf_vm86_fs = context.sc_fs;
233 		tf->tf_vm86_es = context.sc_es;
234 		tf->tf_vm86_ds = context.sc_ds;
235 	} else
236 #endif
237 	{
238 		/* %fs and %gs were restored by the trampoline. */
239 		tf->tf_es = context.sc_es;
240 		tf->tf_ds = context.sc_ds;
241 	}
242 	tf->tf_edi    = context.sc_edi;
243 	tf->tf_esi    = context.sc_esi;
244 	tf->tf_ebp    = context.sc_ebp;
245 	tf->tf_ebx    = context.sc_ebx;
246 	tf->tf_edx    = context.sc_edx;
247 	tf->tf_ecx    = context.sc_ecx;
248 	tf->tf_eax    = context.sc_eax;
249 	tf->tf_eip    = context.sc_eip;
250 	tf->tf_cs     = context.sc_cs;
251 	tf->tf_eflags = context.sc_eflags;
252 	tf->tf_esp    = context.sc_esp_at_signal;
253 	tf->tf_ss     = context.sc_ss;
254 
255 	return (EJUSTRETURN);
256 }
257 
258 #ifdef USER_LDT
259 
260 int
261 linux_read_ldt(p, uap, retval)
262 	struct proc *p;
263 	struct linux_sys_modify_ldt_args /* {
264 		syscallarg(int) func;
265 		syscallarg(void *) ptr;
266 		syscallarg(size_t) bytecount;
267 	} */ *uap;
268 	register_t *retval;
269 {
270 	struct i386_get_ldt_args gl;
271 	int error;
272 	caddr_t sg;
273 	char *parms;
274 
275 	sg = stackgap_init(p->p_emul);
276 
277 	gl.start = 0;
278 	gl.desc = SCARG(uap, ptr);
279 	gl.num = SCARG(uap, bytecount) / sizeof(union descriptor);
280 
281 	parms = stackgap_alloc(&sg, sizeof(gl));
282 
283 	if (error = copyout(&gl, parms, sizeof(gl)))
284 		return (error);
285 
286 	if (error = i386_get_ldt(p, parms, retval))
287 		return (error);
288 
289 	*retval *= sizeof(union descriptor);
290 	return (0);
291 }
292 
293 struct linux_ldt_info {
294 	u_int entry_number;
295 	u_long base_addr;
296 	u_int limit;
297 	u_int seg_32bit:1;
298 	u_int contents:2;
299 	u_int read_exec_only:1;
300 	u_int limit_in_pages:1;
301 	u_int seg_not_present:1;
302 };
303 
304 int
305 linux_write_ldt(p, uap, retval)
306 	struct proc *p;
307 	struct linux_sys_modify_ldt_args /* {
308 		syscallarg(int) func;
309 		syscallarg(void *) ptr;
310 		syscallarg(size_t) bytecount;
311 	} */ *uap;
312 	register_t *retval;
313 {
314 	struct linux_ldt_info ldt_info;
315 	struct segment_descriptor sd;
316 	struct i386_set_ldt_args sl;
317 	int error;
318 	caddr_t sg;
319 	char *parms;
320 
321 	if (SCARG(uap, bytecount) != sizeof(ldt_info))
322 		return (EINVAL);
323 	if (error = copyin(SCARG(uap, ptr), &ldt_info, sizeof(ldt_info)))
324 		return error;
325 	if (ldt_info.contents == 3)
326 		return (EINVAL);
327 
328 	sg = stackgap_init(p->p_emul);
329 
330 	sd.sd_lobase = ldt_info.base_addr & 0xffffff;
331 	sd.sd_hibase = (ldt_info.base_addr >> 24) & 0xff;
332 	sd.sd_lolimit = ldt_info.limit & 0xffff;
333 	sd.sd_hilimit = (ldt_info.limit >> 16) & 0xf;
334 	sd.sd_type =
335 	    16 | (ldt_info.contents << 2) | (!ldt_info.read_exec_only << 1);
336 	sd.sd_dpl = SEL_UPL;
337 	sd.sd_p = !ldt_info.seg_not_present;
338 	sd.sd_def32 = ldt_info.seg_32bit;
339 	sd.sd_gran = ldt_info.limit_in_pages;
340 
341 	sl.start = ldt_info.entry_number;
342 	sl.desc = stackgap_alloc(&sg, sizeof(sd));
343 	sl.num = 1;
344 
345 #if 0
346 	printf("linux_write_ldt: idx=%d, base=%x, limit=%x\n",
347 	    ldt_info.entry_number, ldt_info.base_addr, ldt_info.limit);
348 #endif
349 
350 	parms = stackgap_alloc(&sg, sizeof(sl));
351 
352 	if (error = copyout(&sd, sl.desc, sizeof(sd)))
353 		return (error);
354 	if (error = copyout(&sl, parms, sizeof(sl)))
355 		return (error);
356 
357 	if (error = i386_set_ldt(p, parms, retval))
358 		return (error);
359 
360 	*retval = 0;
361 	return (0);
362 }
363 
364 #endif /* USER_LDT */
365 
366 int
367 linux_sys_modify_ldt(p, v, retval)
368 	struct proc *p;
369 	void *v;
370 	register_t *retval;
371 {
372 	struct linux_sys_modify_ldt_args /* {
373 		syscallarg(int) func;
374 		syscallarg(void *) ptr;
375 		syscallarg(size_t) bytecount;
376 	} */ *uap = v;
377 
378 	switch (SCARG(uap, func)) {
379 #ifdef USER_LDT
380 	case 0:
381 		return (linux_read_ldt(p, uap, retval));
382 
383 	case 1:
384 		return (linux_write_ldt(p, uap, retval));
385 #endif /* USER_LDT */
386 
387 	default:
388 		return (ENOSYS);
389 	}
390 }
391 
392 /*
393  * XXX Pathetic hack to make svgalib work. This will fake the major
394  * device number of an opened VT so that svgalib likes it. grmbl.
395  * Should probably do it 'wrong the right way' and use a mapping
396  * array for all major device numbers, and map linux_mknod too.
397  */
398 dev_t
399 linux_fakedev(dev)
400 	dev_t dev;
401 {
402 
403 	if (major(dev) == NETBSD_CONS_MAJOR)
404 		return makedev(LINUX_CONS_MAJOR, (minor(dev) + 1));
405 	return dev;
406 }
407 
408 /*
409  * We come here in a last attempt to satisfy a Linux ioctl() call
410  */
411 int
412 linux_machdepioctl(p, v, retval)
413 	struct proc *p;
414 	void *v;
415 	register_t *retval;
416 {
417 	struct linux_sys_ioctl_args /* {
418 		syscallarg(int) fd;
419 		syscallarg(u_long) com;
420 		syscallarg(caddr_t) data;
421 	} */ *uap = v;
422 	struct sys_ioctl_args bia, tmparg;
423 	u_long com;
424 #if NVT > 0
425 	int error, mode;
426 	struct vt_mode lvt;
427 	caddr_t bvtp, sg;
428 	u_int fd;
429 	struct file *fp;
430 	struct filedesc *fdp;
431 #endif
432 
433 	SCARG(&bia, fd) = SCARG(uap, fd);
434 	SCARG(&bia, data) = SCARG(uap, data);
435 	com = SCARG(uap, com);
436 
437 	switch (com) {
438 #if NVT > 0
439 	case LINUX_KDGKBMODE:
440 		com = KDGKBMODE;
441 		break;
442 	case LINUX_KDSKBMODE:
443 		com = KDSKBMODE;
444 		if ((unsigned)SCARG(uap, data) == LINUX_K_MEDIUMRAW)
445 			SCARG(&bia, data) = (caddr_t)K_RAW;
446 		break;
447 	case LINUX_KDMKTONE:
448 		com = KDMKTONE;
449 		break;
450 	case LINUX_KDSETMODE:
451 		com = KDSETMODE;
452 		break;
453 	case LINUX_KDENABIO:
454 		com = KDENABIO;
455 		break;
456 	case LINUX_KDDISABIO:
457 		com = KDDISABIO;
458 		break;
459 	case LINUX_KDGETLED:
460 		com = KDGETLED;
461 		break;
462 	case LINUX_KDSETLED:
463 		com = KDSETLED;
464 		break;
465 	case LINUX_VT_OPENQRY:
466 		com = VT_OPENQRY;
467 		break;
468 	case LINUX_VT_GETMODE:
469 		SCARG(&bia, com) = VT_GETMODE;
470 		if ((error = sys_ioctl(p, &bia, retval)))
471 			return error;
472 		if ((error = copyin(SCARG(uap, data), (caddr_t)&lvt,
473 		    sizeof (struct vt_mode))))
474 			return error;
475 		lvt.relsig = bsd_to_linux_sig[lvt.relsig];
476 		lvt.acqsig = bsd_to_linux_sig[lvt.acqsig];
477 		lvt.frsig = bsd_to_linux_sig[lvt.frsig];
478 		return copyout((caddr_t)&lvt, SCARG(uap, data),
479 		    sizeof (struct vt_mode));
480 	case LINUX_VT_SETMODE:
481 		com = VT_SETMODE;
482 		if ((error = copyin(SCARG(uap, data), (caddr_t)&lvt,
483 		    sizeof (struct vt_mode))))
484 			return error;
485 		lvt.relsig = linux_to_bsd_sig[lvt.relsig];
486 		lvt.acqsig = linux_to_bsd_sig[lvt.acqsig];
487 		lvt.frsig = linux_to_bsd_sig[lvt.frsig];
488 		sg = stackgap_init(p->p_emul);
489 		bvtp = stackgap_alloc(&sg, sizeof (struct vt_mode));
490 		if ((error = copyout(&lvt, bvtp, sizeof (struct vt_mode))))
491 			return error;
492 		SCARG(&bia, data) = bvtp;
493 		break;
494 	case LINUX_VT_RELDISP:
495 		com = VT_RELDISP;
496 		break;
497 	case LINUX_VT_ACTIVATE:
498 		com = VT_ACTIVATE;
499 		break;
500 	case LINUX_VT_WAITACTIVE:
501 		com = VT_WAITACTIVE;
502 		break;
503 #endif
504 	default:
505 		return EINVAL;
506 	}
507 	SCARG(&bia, com) = com;
508 	return sys_ioctl(p, &bia, retval);
509 }
510 
511 /*
512  * Set I/O permissions for a process. Just set the maximum level
513  * right away (ignoring the argument), otherwise we would have
514  * to rely on I/O permission maps, which are not implemented.
515  */
516 int
517 linux_sys_iopl(p, v, retval)
518 	struct proc *p;
519 	void *v;
520 	register_t *retval;
521 {
522 	struct linux_sys_iopl_args /* {
523 		syscallarg(int) level;
524 	} */ *uap = v;
525 	struct trapframe *fp = p->p_md.md_regs;
526 
527 	if (suser(p->p_ucred, &p->p_acflag) != 0)
528 		return EPERM;
529 	fp->tf_eflags |= PSL_IOPL;
530 	*retval = 0;
531 	return 0;
532 }
533 
534 /*
535  * See above. If a root process tries to set access to an I/O port,
536  * just let it have the whole range.
537  */
538 int
539 linux_sys_ioperm(p, v, retval)
540 	struct proc *p;
541 	void *v;
542 	register_t *retval;
543 {
544 	struct linux_sys_ioperm_args /* {
545 		syscallarg(unsigned int) lo;
546 		syscallarg(unsigned int) hi;
547 		syscallarg(int) val;
548 	} */ *uap = v;
549 	struct trapframe *fp = p->p_md.md_regs;
550 
551 	if (suser(p->p_ucred, &p->p_acflag) != 0)
552 		return EPERM;
553 	if (SCARG(uap, val))
554 		fp->tf_eflags |= PSL_IOPL;
555 	*retval = 0;
556 	return 0;
557 }
558