linux_machdep.c revision 1.43.14.1 1 /* $NetBSD: linux_machdep.c,v 1.43.14.1 2009/05/13 17:18:56 jym Exp $ */
2
3 /*-
4 * Copyright (c) 1998 The NetBSD Foundation, Inc.
5 * All rights reserved.
6 *
7 * This code is derived from software contributed to The NetBSD Foundation
8 * by Eric Haszlakiewicz.
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 * Based on sys/arch/i386/i386/linux_machdep.c:
32 * linux_machdep.c,v 1.42 1998/09/11 12:50:06 mycroft Exp
33 * written by Frank van der Linden
34 *
35 */
36
37 #include <sys/cdefs.h>
38 __KERNEL_RCSID(0, "$NetBSD: linux_machdep.c,v 1.43.14.1 2009/05/13 17:18:56 jym Exp $");
39
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/signalvar.h>
43 #include <sys/kernel.h>
44 #include <sys/proc.h>
45 #include <sys/user.h>
46 #include <sys/buf.h>
47 #include <sys/reboot.h>
48 #include <sys/conf.h>
49 #include <sys/exec.h>
50 #include <sys/file.h>
51 #include <sys/callout.h>
52 #include <sys/malloc.h>
53 #include <sys/mbuf.h>
54 #include <sys/msgbuf.h>
55 #include <sys/mount.h>
56 #include <sys/vnode.h>
57 #include <sys/device.h>
58 #include <sys/syscallargs.h>
59 #include <sys/filedesc.h>
60 #include <sys/exec_elf.h>
61 #include <sys/ioctl.h>
62 #include <sys/kauth.h>
63
64 #include <uvm/uvm_extern.h>
65
66 #include <compat/linux/common/linux_types.h>
67 #include <compat/linux/common/linux_signal.h>
68 #include <compat/linux/common/linux_siginfo.h>
69 #include <compat/linux/common/linux_util.h>
70 #include <compat/linux/common/linux_ioctl.h>
71 #include <compat/linux/common/linux_exec.h>
72 #include <compat/linux/common/linux_machdep.h>
73 #include <compat/linux/common/linux_emuldata.h>
74
75 #include <compat/linux/linux_syscallargs.h>
76
77 #include <machine/alpha.h>
78 #include <machine/reg.h>
79
80 #if defined(_KERNEL_OPT)
81 #include "wsdisplay.h"
82 #endif
83 #if (NWSDISPLAY >0)
84 #include <dev/wscons/wsdisplay_usl_io.h>
85 #endif
86 #ifdef DEBUG
87 #include <machine/sigdebug.h>
88 #endif
89
90 /*
91 * Deal with some alpha-specific things in the Linux emulation code.
92 */
93
94 void
95 linux_setregs(struct lwp *l, struct exec_package *epp, u_long stack)
96 {
97 #ifdef DEBUG
98 struct trapframe *tfp = l->l_md.md_tf;
99 #endif
100
101 setregs(l, epp, stack);
102 #ifdef DEBUG
103 /*
104 * Linux has registers set to zero on entry; for DEBUG kernels
105 * the alpha setregs() fills registers with 0xbabefacedeadbeef.
106 */
107 memset(tfp->tf_regs, 0, FRAME_SIZE * sizeof tfp->tf_regs[0]);
108 #endif
109 }
110
111 void
112 setup_linux_rt_sigframe(struct trapframe *tf, int sig, const sigset_t *mask)
113 {
114 struct lwp *l = curlwp;
115 struct proc *p = l->l_proc;
116 struct linux_rt_sigframe *sfp, sigframe;
117 int onstack, error;
118 int fsize, rndfsize;
119 extern char linux_rt_sigcode[], linux_rt_esigcode[];
120
121 /* Do we need to jump onto the signal stack? */
122 onstack = (l->l_sigstk.ss_flags & (SS_DISABLE | SS_ONSTACK)) == 0 &&
123 (SIGACTION(p, sig).sa_flags & SA_ONSTACK) != 0;
124
125 /* Allocate space for the signal handler context. */
126 fsize = sizeof(struct linux_rt_sigframe);
127 rndfsize = ((fsize + 15) / 16) * 16;
128
129 if (onstack)
130 sfp = (struct linux_rt_sigframe *)
131 ((char *)l->l_sigstk.ss_sp + l->l_sigstk.ss_size);
132 else
133 sfp = (struct linux_rt_sigframe *)(alpha_pal_rdusp());
134 sfp = (struct linux_rt_sigframe *)((char *)sfp - rndfsize);
135
136 #ifdef DEBUG
137 if ((sigdebug & SDB_KSTACK) && (p->p_pid == sigpid))
138 printf("linux_sendsig(%d): sig %d ssp %p usp %p\n", p->p_pid,
139 sig, &onstack, sfp);
140 #endif /* DEBUG */
141
142 /*
143 * Build the signal context to be used by sigreturn.
144 */
145 memset(&sigframe.uc, 0, sizeof(struct linux_ucontext));
146 sigframe.uc.uc_mcontext.sc_onstack = onstack;
147
148 /* Setup potentially partial signal mask in sc_mask. */
149 /* But get all of it in uc_sigmask */
150 native_to_linux_old_sigset(&sigframe.uc.uc_mcontext.sc_mask, mask);
151 native_to_linux_sigset(&sigframe.uc.uc_sigmask, mask);
152
153 sigframe.uc.uc_mcontext.sc_pc = tf->tf_regs[FRAME_PC];
154 sigframe.uc.uc_mcontext.sc_ps = ALPHA_PSL_USERMODE;
155 frametoreg(tf, (struct reg *)sigframe.uc.uc_mcontext.sc_regs);
156 sigframe.uc.uc_mcontext.sc_regs[R_SP] = alpha_pal_rdusp();
157
158 alpha_enable_fp(l, 1);
159 sigframe.uc.uc_mcontext.sc_fpcr = alpha_read_fpcr();
160 sigframe.uc.uc_mcontext.sc_fp_control = alpha_read_fp_c(l);
161 alpha_pal_wrfen(0);
162
163 sigframe.uc.uc_mcontext.sc_traparg_a0 = tf->tf_regs[FRAME_A0];
164 sigframe.uc.uc_mcontext.sc_traparg_a1 = tf->tf_regs[FRAME_A1];
165 sigframe.uc.uc_mcontext.sc_traparg_a2 = tf->tf_regs[FRAME_A2];
166
167 /*
168 * XXX XAX Create bogus siginfo data. This can't really
169 * XXX be fixed until NetBSD has realtime signals.
170 * XXX Or we do the emuldata thing.
171 * XXX -erh
172 */
173 memset(&sigframe.info, 0, sizeof(struct linux_siginfo));
174 sigframe.info.lsi_signo = sig;
175 sigframe.info.lsi_code = LINUX_SI_USER;
176 sigframe.info.lsi_pid = p->p_pid;
177 sigframe.info.lsi_uid = kauth_cred_geteuid(l->l_cred); /* Use real uid here? */
178
179 sendsig_reset(l, sig);
180 mutex_exit(p->p_lock);
181 error = copyout((void *)&sigframe, (void *)sfp, fsize);
182 mutex_enter(p->p_lock);
183
184 if (error != 0) {
185 #ifdef DEBUG
186 if ((sigdebug & SDB_KSTACK) && p->p_pid == sigpid)
187 printf("sendsig(%d): copyout failed on sig %d\n",
188 p->p_pid, sig);
189 #endif
190 /*
191 * Process has trashed its stack; give it an illegal
192 * instruction to halt it in its tracks.
193 */
194 sigexit(l, SIGILL);
195 /* NOTREACHED */
196 }
197
198 /* Pass pointers to siginfo and ucontext in the regs */
199 tf->tf_regs[FRAME_A1] = (unsigned long)&sfp->info;
200 tf->tf_regs[FRAME_A2] = (unsigned long)&sfp->uc;
201
202 /* Address of trampoline code. End up at this PC after mi_switch */
203 tf->tf_regs[FRAME_PC] =
204 (u_int64_t)(p->p_psstr - (linux_rt_esigcode - linux_rt_sigcode));
205
206 /* Adjust the stack */
207 alpha_pal_wrusp((unsigned long)sfp);
208
209 /* Remember that we're now on the signal stack. */
210 if (onstack)
211 l->l_sigstk.ss_flags |= SS_ONSTACK;
212 }
213
214 void setup_linux_sigframe(tf, sig, mask)
215 struct trapframe *tf;
216 int sig;
217 const sigset_t *mask;
218 {
219 struct lwp *l = curlwp;
220 struct proc *p = l->l_proc;
221 struct linux_sigframe *sfp, sigframe;
222 int onstack, error;
223 int fsize, rndfsize;
224 extern char linux_sigcode[], linux_esigcode[];
225
226 /* Do we need to jump onto the signal stack? */
227 onstack = (l->l_sigstk.ss_flags & (SS_DISABLE | SS_ONSTACK)) == 0 &&
228 (SIGACTION(p, sig).sa_flags & SA_ONSTACK) != 0;
229
230 /* Allocate space for the signal handler context. */
231 fsize = sizeof(struct linux_sigframe);
232 rndfsize = ((fsize + 15) / 16) * 16;
233
234 if (onstack)
235 sfp = (struct linux_sigframe *)
236 ((char *)l->l_sigstk.ss_sp + l->l_sigstk.ss_size);
237 else
238 sfp = (struct linux_sigframe *)(alpha_pal_rdusp());
239 sfp = (struct linux_sigframe *)((char *)sfp - rndfsize);
240
241 #ifdef DEBUG
242 if ((sigdebug & SDB_KSTACK) && (p->p_pid == sigpid))
243 printf("linux_sendsig(%d): sig %d ssp %p usp %p\n", p->p_pid,
244 sig, &onstack, sfp);
245 #endif /* DEBUG */
246
247 /*
248 * Build the signal context to be used by sigreturn.
249 */
250 memset(&sigframe.sf_sc, 0, sizeof(struct linux_sigcontext));
251 sigframe.sf_sc.sc_onstack = onstack;
252 native_to_linux_old_sigset(&sigframe.sf_sc.sc_mask, mask);
253 sigframe.sf_sc.sc_pc = tf->tf_regs[FRAME_PC];
254 sigframe.sf_sc.sc_ps = ALPHA_PSL_USERMODE;
255 frametoreg(tf, (struct reg *)sigframe.sf_sc.sc_regs);
256 sigframe.sf_sc.sc_regs[R_SP] = alpha_pal_rdusp();
257
258 if (l == fpcurlwp) {
259 alpha_pal_wrfen(1);
260 savefpstate(&l->l_addr->u_pcb.pcb_fp);
261 alpha_pal_wrfen(0);
262 sigframe.sf_sc.sc_fpcr = l->l_addr->u_pcb.pcb_fp.fpr_cr;
263 fpcurlwp = NULL;
264 }
265 /* XXX ownedfp ? etc...? */
266
267 sigframe.sf_sc.sc_traparg_a0 = tf->tf_regs[FRAME_A0];
268 sigframe.sf_sc.sc_traparg_a1 = tf->tf_regs[FRAME_A1];
269 sigframe.sf_sc.sc_traparg_a2 = tf->tf_regs[FRAME_A2];
270
271 sendsig_reset(l, sig);
272 mutex_exit(p->p_lock);
273 error = copyout((void *)&sigframe, (void *)sfp, fsize);
274 mutex_enter(p->p_lock);
275
276 if (error != 0) {
277 #ifdef DEBUG
278 if ((sigdebug & SDB_KSTACK) && p->p_pid == sigpid)
279 printf("sendsig(%d): copyout failed on sig %d\n",
280 p->p_pid, sig);
281 #endif
282 /*
283 * Process has trashed its stack; give it an illegal
284 * instruction to halt it in its tracks.
285 */
286 sigexit(l, SIGILL);
287 /* NOTREACHED */
288 }
289
290 /* Pass pointers to sigcontext in the regs */
291 tf->tf_regs[FRAME_A1] = 0;
292 tf->tf_regs[FRAME_A2] = (unsigned long)&sfp->sf_sc;
293
294 /* Address of trampoline code. End up at this PC after mi_switch */
295 tf->tf_regs[FRAME_PC] =
296 (u_int64_t)(p->p_psstr - (linux_esigcode - linux_sigcode));
297
298 /* Adjust the stack */
299 alpha_pal_wrusp((unsigned long)sfp);
300
301 /* Remember that we're now on the signal stack. */
302 if (onstack)
303 l->l_sigstk.ss_flags |= SS_ONSTACK;
304 }
305
306 /*
307 * Send an interrupt to process.
308 *
309 * Stack is set up to allow sigcode stored
310 * in u. to call routine, followed by kcall
311 * to sigreturn routine below. After sigreturn
312 * resets the signal mask, the stack, and the
313 * frame pointer, it returns to the user
314 * specified pc, psl.
315 */
316 void
317 linux_sendsig(const ksiginfo_t *ksi, const sigset_t *mask)
318 {
319 struct lwp *l = curlwp;
320 struct proc *p = l->l_proc;
321 struct trapframe *tf = l->l_md.md_tf;
322 const int sig = ksi->ksi_signo;
323 sig_t catcher = SIGACTION(p, sig).sa_handler;
324 #ifdef notyet
325 struct linux_emuldata *edp;
326
327 /* Setup the signal frame (and part of the trapframe) */
328 /*OLD: if (p->p_sigacts->ps_siginfo & sigmask(sig))*/
329 /* XXX XAX this is broken now. need someplace to store what
330 XXX XAX kind of signal handler a signal has.*/
331 #if 0
332 edp = (struct linux_emuldata *)p->p_emuldata;
333 #else
334 edp = 0;
335 #endif
336 if (edp && sigismember(&edp->ps_siginfo, sig))
337 setup_linux_rt_sigframe(tf, sig, mask);
338 else
339 #endif /* notyet */
340 setup_linux_sigframe(tf, sig, mask);
341
342 /* Signal handler for trampoline code */
343 tf->tf_regs[FRAME_T12] = (u_int64_t)catcher;
344 tf->tf_regs[FRAME_A0] = native_to_linux_signo[sig];
345
346 /*
347 * Linux has a custom restorer option. To support it we would
348 * need to store an array of restorers and a sigcode block
349 * which knew to use it. Doesn't seem worth the trouble.
350 * -erh
351 */
352
353 #ifdef DEBUG
354 if (sigdebug & SDB_FOLLOW)
355 printf("sendsig(%d): pc %lx, catcher %lx\n", l->l_proc->p_pid,
356 tf->tf_regs[FRAME_PC], tf->tf_regs[FRAME_A3]);
357 if ((sigdebug & SDB_KSTACK) && l->l_proc->p_pid == sigpid)
358 printf("sendsig(%d): sig %d returns\n", l->l_proc->p_pid, sig);
359 #endif
360 }
361
362 /*
363 * System call to cleanup state after a signal
364 * has been taken. Reset signal mask and
365 * stack state from context left by sendsig (above).
366 * Return to previous pc as specified by context
367 * left by sendsig.
368 * Linux real-time signals use a different sigframe,
369 * but the sigcontext is the same.
370 */
371
372 int
373 linux_restore_sigcontext(struct lwp *l, struct linux_sigcontext context,
374 sigset_t *mask)
375 {
376 struct proc *p = l->l_proc;
377
378 /*
379 * Linux doesn't (yet) have alternate signal stacks.
380 * However, the OSF/1 sigcontext which they use has
381 * an onstack member. This could be needed in the future.
382 */
383 mutex_enter(p->p_lock);
384 if (context.sc_onstack & LINUX_SA_ONSTACK)
385 l->l_sigstk.ss_flags |= SS_ONSTACK;
386 else
387 l->l_sigstk.ss_flags &= ~SS_ONSTACK;
388
389 /* Reset the signal mask */
390 (void) sigprocmask1(l, SIG_SETMASK, mask, 0);
391 mutex_exit(p->p_lock);
392
393 /*
394 * Check for security violations.
395 * Linux doesn't allow any changes to the PSL.
396 */
397 if (context.sc_ps != ALPHA_PSL_USERMODE)
398 return(EINVAL);
399
400 l->l_md.md_tf->tf_regs[FRAME_PC] = context.sc_pc;
401 l->l_md.md_tf->tf_regs[FRAME_PS] = context.sc_ps;
402
403 regtoframe((struct reg *)context.sc_regs, l->l_md.md_tf);
404 alpha_pal_wrusp(context.sc_regs[R_SP]);
405
406 if (l == fpcurlwp)
407 fpcurlwp = NULL;
408
409 /* Restore fp regs and fpr_cr */
410 memcpy( &l->l_addr->u_pcb.pcb_fp, (struct fpreg *)context.sc_fpregs,
411 sizeof(struct fpreg));
412 /* XXX sc_ownedfp ? */
413 /* XXX sc_fp_control ? */
414
415 #ifdef DEBUG
416 if (sigdebug & SDB_FOLLOW)
417 printf("linux_rt_sigreturn(%d): returns\n", p->p_pid);
418 #endif
419 return (EJUSTRETURN);
420 }
421
422 int
423 linux_sys_rt_sigreturn(struct lwp *l, const struct linux_sys_rt_sigreturn_args *uap, register_t *retval)
424 {
425 /* {
426 syscallarg(struct linux_rt_sigframe *) sfp;
427 } */
428 struct linux_rt_sigframe *sfp, sigframe;
429 sigset_t mask;
430
431 /*
432 * The trampoline code hands us the context.
433 * It is unsafe to keep track of it ourselves, in the event that a
434 * program jumps out of a signal handler.
435 */
436
437 sfp = SCARG(uap, sfp);
438
439 if (ALIGN(sfp) != (u_int64_t)sfp)
440 return(EINVAL);
441
442 /*
443 * Fetch the frame structure.
444 */
445 if (copyin((void *)sfp, &sigframe,
446 sizeof(struct linux_rt_sigframe)) != 0)
447 return (EFAULT);
448
449 /* Grab the signal mask */
450 linux_to_native_sigset(&mask, &sigframe.uc.uc_sigmask);
451
452 return(linux_restore_sigcontext(l, sigframe.uc.uc_mcontext, &mask));
453 }
454
455
456 int
457 linux_sys_sigreturn(struct lwp *l, const struct linux_sys_sigreturn_args *uap, register_t *retval)
458 {
459 /* {
460 syscallarg(struct linux_sigframe *) sfp;
461 } */
462 struct linux_sigframe *sfp, frame;
463 sigset_t mask;
464
465 /*
466 * The trampoline code hands us the context.
467 * It is unsafe to keep track of it ourselves, in the event that a
468 * program jumps out of a signal handler.
469 */
470
471 sfp = SCARG(uap, sfp);
472 if (ALIGN(sfp) != (u_int64_t)sfp)
473 return(EINVAL);
474
475 /*
476 * Fetch the frame structure.
477 */
478 if (copyin((void *)sfp, &frame, sizeof(struct linux_sigframe)) != 0)
479 return(EFAULT);
480
481 /* Grab the signal mask. */
482 /* XXX use frame.extramask */
483 linux_old_to_native_sigset(&mask, frame.sf_sc.sc_mask);
484
485 return(linux_restore_sigcontext(l, frame.sf_sc, &mask));
486 }
487
488 /*
489 * We come here in a last attempt to satisfy a Linux ioctl() call
490 */
491 /* XXX XAX update this, add maps, etc... */
492 int
493 linux_machdepioctl(struct lwp *l, const struct linux_sys_ioctl_args *uap, register_t *retval)
494 {
495 /* {
496 syscallarg(int) fd;
497 syscallarg(u_long) com;
498 syscallarg(void *) data;
499 } */
500 struct sys_ioctl_args bia;
501 u_long com;
502
503 SCARG(&bia, fd) = SCARG(uap, fd);
504 SCARG(&bia, data) = SCARG(uap, data);
505 com = SCARG(uap, com);
506
507 switch (com) {
508 default:
509 printf("linux_machdepioctl: invalid ioctl %08lx\n", com);
510 return EINVAL;
511 }
512 SCARG(&bia, com) = com;
513 return sys_ioctl(l, &bia, retval);
514 }
515
516 /* XXX XAX fix this */
517 dev_t
518 linux_fakedev(dev_t dev, int raw)
519 {
520 return dev;
521 }
522
523 int
524 linux_usertrap(struct lwp *l, vaddr_t trapaddr, void *arg)
525 {
526 return 0;
527 }
528