xref: /freebsd/sys/amd64/linux/linux_machdep.c (revision 401ab69cff8fa2320a9f8ea4baa114a6da6c952b)
1 /*-
2  * Copyright (c) 2004 Tim J. Robbins
3  * Copyright (c) 2002 Doug Rabson
4  * Copyright (c) 2000 Marcel Moolenaar
5  * All rights reserved.
6  * Copyright (c) 2013 Dmitry Chagin <dchagin@FreeBSD.org>
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer
13  *    in this position and unchanged.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. The name of the author may not be used to endorse or promote products
18  *    derived from this software without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
21  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
22  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
23  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
24  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
25  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
26  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
27  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
28  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
29  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 #include <sys/cdefs.h>
33 #include <sys/param.h>
34 #include <sys/systm.h>
35 #include <sys/ktr.h>
36 #include <sys/lock.h>
37 #include <sys/malloc.h>
38 #include <sys/mman.h>
39 #include <sys/mutex.h>
40 #include <sys/priv.h>
41 #include <sys/proc.h>
42 #include <sys/ptrace.h>
43 #include <sys/syscallsubr.h>
44 
45 #include <security/mac/mac_framework.h>
46 
47 #include <ufs/ufs/extattr.h>
48 #include <ufs/ufs/quota.h>
49 #include <ufs/ufs/ufsmount.h>
50 
51 #include <machine/frame.h>
52 #include <machine/md_var.h>
53 #include <machine/pcb.h>
54 #include <machine/psl.h>
55 #include <machine/segments.h>
56 #include <machine/specialreg.h>
57 
58 #include <vm/pmap.h>
59 #include <vm/vm.h>
60 #include <vm/vm_param.h>
61 #include <vm/vm_extern.h>
62 #include <vm/vm_kern.h>
63 #include <vm/vm_map.h>
64 
65 #include <x86/ifunc.h>
66 #include <x86/reg.h>
67 #include <x86/sysarch.h>
68 
69 #include <amd64/linux/linux.h>
70 #include <amd64/linux/linux_proto.h>
71 #include <compat/linux/linux_fork.h>
72 #include <compat/linux/linux_misc.h>
73 #include <compat/linux/linux_mmap.h>
74 #include <compat/linux/linux_util.h>
75 
76 #define	LINUX_ARCH_AMD64		0xc000003e
77 
78 int
79 linux_set_upcall(struct thread *td, register_t stack)
80 {
81 
82 	if (stack)
83 		td->td_frame->tf_rsp = stack;
84 
85 	/*
86 	 * The newly created Linux thread returns
87 	 * to the user space by the same path that a parent does.
88 	 */
89 	td->td_frame->tf_rax = 0;
90 	return (0);
91 }
92 
93 int
94 linux_mmap2(struct thread *td, struct linux_mmap2_args *args)
95 {
96 
97 	return (linux_mmap_common(td, args->addr, args->len, args->prot,
98 		args->flags, args->fd, args->pgoff));
99 }
100 
101 int
102 linux_mprotect(struct thread *td, struct linux_mprotect_args *uap)
103 {
104 
105 	return (linux_mprotect_common(td, uap->addr, uap->len, uap->prot));
106 }
107 
108 int
109 linux_madvise(struct thread *td, struct linux_madvise_args *uap)
110 {
111 
112 	return (linux_madvise_common(td, uap->addr, uap->len, uap->behav));
113 }
114 
115 int
116 linux_iopl(struct thread *td, struct linux_iopl_args *args)
117 {
118 	int error;
119 
120 	LINUX_CTR(iopl);
121 
122 	if (args->level > 3)
123 		return (EINVAL);
124 	if ((error = priv_check(td, PRIV_IO)) != 0)
125 		return (error);
126 	if ((error = securelevel_gt(td->td_ucred, 0)) != 0)
127 		return (error);
128 	td->td_frame->tf_rflags = (td->td_frame->tf_rflags & ~PSL_IOPL) |
129 	    (args->level * (PSL_IOPL / 3));
130 
131 	return (0);
132 }
133 
134 int
135 linux_pause(struct thread *td, struct linux_pause_args *args)
136 {
137 	struct proc *p = td->td_proc;
138 	sigset_t sigmask;
139 
140 	LINUX_CTR(pause);
141 
142 	PROC_LOCK(p);
143 	sigmask = td->td_sigmask;
144 	PROC_UNLOCK(p);
145 	return (kern_sigsuspend(td, sigmask));
146 }
147 
148 int
149 linux_arch_prctl(struct thread *td, struct linux_arch_prctl_args *args)
150 {
151 	unsigned long long cet[3];
152 	struct pcb *pcb;
153 	int error;
154 
155 	pcb = td->td_pcb;
156 	LINUX_CTR2(arch_prctl, "0x%x, %p", args->code, args->addr);
157 
158 	switch (args->code) {
159 	case LINUX_ARCH_SET_GS:
160 		if (args->addr < VM_MAXUSER_ADDRESS) {
161 			update_pcb_bases(pcb);
162 			pcb->pcb_gsbase = args->addr;
163 			td->td_frame->tf_gs = _ugssel;
164 			error = 0;
165 		} else
166 			error = EPERM;
167 		break;
168 	case LINUX_ARCH_SET_FS:
169 		if (args->addr < VM_MAXUSER_ADDRESS) {
170 			update_pcb_bases(pcb);
171 			pcb->pcb_fsbase = args->addr;
172 			td->td_frame->tf_fs = _ufssel;
173 			error = 0;
174 		} else
175 			error = EPERM;
176 		break;
177 	case LINUX_ARCH_GET_FS:
178 		error = copyout(&pcb->pcb_fsbase, PTRIN(args->addr),
179 		    sizeof(args->addr));
180 		break;
181 	case LINUX_ARCH_GET_GS:
182 		error = copyout(&pcb->pcb_gsbase, PTRIN(args->addr),
183 		    sizeof(args->addr));
184 		break;
185 	case LINUX_ARCH_CET_STATUS:
186 		memset(cet, 0, sizeof(cet));
187 		error = copyout(&cet, PTRIN(args->addr), sizeof(cet));
188 		break;
189 	default:
190 		linux_msg(td, "unsupported arch_prctl code %#x", args->code);
191 		error = EINVAL;
192 	}
193 	return (error);
194 }
195 
196 int
197 linux_set_cloned_tls(struct thread *td, void *desc)
198 {
199 	struct pcb *pcb;
200 
201 	if ((uint64_t)desc >= VM_MAXUSER_ADDRESS)
202 		return (EPERM);
203 
204 	pcb = td->td_pcb;
205 	update_pcb_bases(pcb);
206 	pcb->pcb_fsbase = (register_t)desc;
207 	td->td_frame->tf_fs = _ufssel;
208 
209 	return (0);
210 }
211 
212 int futex_xchgl_nosmap(int oparg, uint32_t *uaddr, int *oldval);
213 int futex_xchgl_smap(int oparg, uint32_t *uaddr, int *oldval);
214 DEFINE_IFUNC(, int, futex_xchgl, (int, uint32_t *, int *))
215 {
216 
217 	return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ?
218 	    futex_xchgl_smap : futex_xchgl_nosmap);
219 }
220 
221 int futex_addl_nosmap(int oparg, uint32_t *uaddr, int *oldval);
222 int futex_addl_smap(int oparg, uint32_t *uaddr, int *oldval);
223 DEFINE_IFUNC(, int, futex_addl, (int, uint32_t *, int *))
224 {
225 
226 	return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ?
227 	    futex_addl_smap : futex_addl_nosmap);
228 }
229 
230 int futex_orl_nosmap(int oparg, uint32_t *uaddr, int *oldval);
231 int futex_orl_smap(int oparg, uint32_t *uaddr, int *oldval);
232 DEFINE_IFUNC(, int, futex_orl, (int, uint32_t *, int *))
233 {
234 
235 	return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ?
236 	    futex_orl_smap : futex_orl_nosmap);
237 }
238 
239 int futex_andl_nosmap(int oparg, uint32_t *uaddr, int *oldval);
240 int futex_andl_smap(int oparg, uint32_t *uaddr, int *oldval);
241 DEFINE_IFUNC(, int, futex_andl, (int, uint32_t *, int *))
242 {
243 
244 	return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ?
245 	    futex_andl_smap : futex_andl_nosmap);
246 }
247 
248 int futex_xorl_nosmap(int oparg, uint32_t *uaddr, int *oldval);
249 int futex_xorl_smap(int oparg, uint32_t *uaddr, int *oldval);
250 DEFINE_IFUNC(, int, futex_xorl, (int, uint32_t *, int *))
251 {
252 
253 	return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ?
254 	    futex_xorl_smap : futex_xorl_nosmap);
255 }
256 
257 void
258 bsd_to_linux_regset(const struct reg *b_reg, struct linux_pt_regset *l_regset)
259 {
260 
261 	l_regset->r15 = b_reg->r_r15;
262 	l_regset->r14 = b_reg->r_r14;
263 	l_regset->r13 = b_reg->r_r13;
264 	l_regset->r12 = b_reg->r_r12;
265 	l_regset->rbp = b_reg->r_rbp;
266 	l_regset->rbx = b_reg->r_rbx;
267 	l_regset->r11 = b_reg->r_r11;
268 	l_regset->r10 = b_reg->r_r10;
269 	l_regset->r9 = b_reg->r_r9;
270 	l_regset->r8 = b_reg->r_r8;
271 	l_regset->rax = b_reg->r_rax;
272 	l_regset->rcx = b_reg->r_rcx;
273 	l_regset->rdx = b_reg->r_rdx;
274 	l_regset->rsi = b_reg->r_rsi;
275 	l_regset->rdi = b_reg->r_rdi;
276 	l_regset->orig_rax = b_reg->r_rax;
277 	l_regset->rip = b_reg->r_rip;
278 	l_regset->cs = b_reg->r_cs;
279 	l_regset->eflags = b_reg->r_rflags;
280 	l_regset->rsp = b_reg->r_rsp;
281 	l_regset->ss = b_reg->r_ss;
282 	l_regset->fs_base = 0;
283 	l_regset->gs_base = 0;
284 	l_regset->ds = b_reg->r_ds;
285 	l_regset->es = b_reg->r_es;
286 	l_regset->fs = b_reg->r_fs;
287 	l_regset->gs = b_reg->r_gs;
288 }
289 
290 void
291 linux_to_bsd_regset(struct reg *b_reg, const struct linux_pt_regset *l_regset)
292 {
293 
294 	b_reg->r_r15 = l_regset->r15;
295 	b_reg->r_r14 = l_regset->r14;
296 	b_reg->r_r13 = l_regset->r13;
297 	b_reg->r_r12 = l_regset->r12;
298 	b_reg->r_rbp = l_regset->rbp;
299 	b_reg->r_rbx = l_regset->rbx;
300 	b_reg->r_r11 = l_regset->r11;
301 	b_reg->r_r10 = l_regset->r10;
302 	b_reg->r_r9 = l_regset->r9;
303 	b_reg->r_r8 = l_regset->r8;
304 	b_reg->r_rax = l_regset->rax;
305 	b_reg->r_rcx = l_regset->rcx;
306 	b_reg->r_rdx = l_regset->rdx;
307 	b_reg->r_rsi = l_regset->rsi;
308 	b_reg->r_rdi = l_regset->rdi;
309 	b_reg->r_rax = l_regset->orig_rax;
310 	b_reg->r_rip = l_regset->rip;
311 	b_reg->r_cs = l_regset->cs;
312 	b_reg->r_rflags = l_regset->eflags;
313 	b_reg->r_rsp = l_regset->rsp;
314 	b_reg->r_ss = l_regset->ss;
315 	b_reg->r_ds = l_regset->ds;
316 	b_reg->r_es = l_regset->es;
317 	b_reg->r_fs = l_regset->fs;
318 	b_reg->r_gs = l_regset->gs;
319 }
320 
321 void
322 linux_ptrace_get_syscall_info_machdep(const struct reg *reg,
323     struct syscall_info *si)
324 {
325 
326 	si->arch = LINUX_ARCH_AMD64;
327 	si->instruction_pointer = reg->r_rip;
328 	si->stack_pointer = reg->r_rsp;
329 }
330 
331 int
332 linux_ptrace_getregs_machdep(struct thread *td, pid_t pid,
333     struct linux_pt_regset *l_regset)
334 {
335 	struct ptrace_lwpinfo lwpinfo;
336 	struct pcb *pcb;
337 	int error;
338 
339 	pcb = td->td_pcb;
340 	if (td == curthread)
341 		update_pcb_bases(pcb);
342 
343 	l_regset->fs_base = pcb->pcb_fsbase;
344 	l_regset->gs_base = pcb->pcb_gsbase;
345 
346 	error = kern_ptrace(td, PT_LWPINFO, pid, &lwpinfo, sizeof(lwpinfo));
347 	if (error != 0) {
348 		linux_msg(td, "PT_LWPINFO failed with error %d", error);
349 		return (error);
350 	}
351 	if ((lwpinfo.pl_flags & (PL_FLAG_SCE | PL_FLAG_SCX)) != 0) {
352 		/*
353 		 * In Linux, the syscall number - passed to the syscall
354 		 * as rax - is preserved in orig_rax; rax gets overwritten
355 		 * with syscall return value.
356 		 */
357 		l_regset->orig_rax = lwpinfo.pl_syscall_code;
358 	}
359 
360 	return (0);
361 }
362 
363 #define	LINUX_URO(a,m) ((uintptr_t)a == offsetof(struct linux_pt_regset, m))
364 
365 int
366 linux_ptrace_peekuser(struct thread *td, pid_t pid, void *addr, void *data)
367 {
368 	struct linux_pt_regset reg;
369 	struct reg b_reg;
370 	uint64_t val;
371 	int error;
372 
373 	if ((uintptr_t)addr & (sizeof(data) -1) || (uintptr_t)addr < 0)
374 		return (EIO);
375 	if ((uintptr_t)addr >= sizeof(struct linux_pt_regset)) {
376 		LINUX_RATELIMIT_MSG_OPT1("PTRACE_PEEKUSER offset %ld "
377 		    "not implemented; returning EINVAL", (uintptr_t)addr);
378 		return (EINVAL);
379 	}
380 
381 	if (LINUX_URO(addr, fs_base))
382 		return (kern_ptrace(td, PT_GETFSBASE, pid, data, 0));
383 	if (LINUX_URO(addr, gs_base))
384 		return (kern_ptrace(td, PT_GETGSBASE, pid, data, 0));
385 	if ((error = kern_ptrace(td, PT_GETREGS, pid, &b_reg, 0)) != 0)
386 		return (error);
387 	bsd_to_linux_regset(&b_reg, &reg);
388 	val = *(&reg.r15 + ((uintptr_t)addr / sizeof(reg.r15)));
389 	return (copyout(&val, data, sizeof(val)));
390 }
391 
392 static inline bool
393 linux_invalid_selector(u_short val)
394 {
395 
396 	return (val != 0 && ISPL(val) != SEL_UPL);
397 }
398 
399 struct linux_segreg_off {
400 	uintptr_t	reg;
401 	bool		is0;
402 };
403 
404 const struct linux_segreg_off linux_segregs_off[] = {
405 	{
406 		.reg = offsetof(struct linux_pt_regset, gs),
407 		.is0 = true,
408 	},
409 	{
410 		.reg = offsetof(struct linux_pt_regset, fs),
411 		.is0 = true,
412 	},
413 	{
414 		.reg = offsetof(struct linux_pt_regset, ds),
415 		.is0 = true,
416 	},
417 	{
418 		.reg = offsetof(struct linux_pt_regset, es),
419 		.is0 = true,
420 	},
421 	{
422 		.reg = offsetof(struct linux_pt_regset, cs),
423 		.is0 = false,
424 	},
425 	{
426 		.reg = offsetof(struct linux_pt_regset, ss),
427 		.is0 = false,
428 	},
429 };
430 
431 int
432 linux_ptrace_pokeuser(struct thread *td, pid_t pid, void *addr, void *data)
433 {
434 	struct linux_pt_regset reg;
435 	struct reg b_reg, b_reg1;
436 	int error, i;
437 
438 	if ((uintptr_t)addr & (sizeof(data) -1) || (uintptr_t)addr < 0)
439 		return (EIO);
440 	if ((uintptr_t)addr >= sizeof(struct linux_pt_regset)) {
441 		LINUX_RATELIMIT_MSG_OPT1("PTRACE_POKEUSER offset %ld "
442 		    "not implemented; returning EINVAL", (uintptr_t)addr);
443 		return (EINVAL);
444 	}
445 
446 	if (LINUX_URO(addr, fs_base))
447 		return (kern_ptrace(td, PT_SETFSBASE, pid, data, 0));
448 	if (LINUX_URO(addr, gs_base))
449 		return (kern_ptrace(td, PT_SETGSBASE, pid, data, 0));
450 	for (i = 0; i < nitems(linux_segregs_off); i++) {
451 		if ((uintptr_t)addr == linux_segregs_off[i].reg) {
452 			if (linux_invalid_selector((uintptr_t)data))
453 				return (EIO);
454 			if (!linux_segregs_off[i].is0 && (uintptr_t)data == 0)
455 				return (EIO);
456 		}
457 	}
458 	if ((error = kern_ptrace(td, PT_GETREGS, pid, &b_reg, 0)) != 0)
459 		return (error);
460 	bsd_to_linux_regset(&b_reg, &reg);
461 	*(&reg.r15 + ((uintptr_t)addr / sizeof(reg.r15))) = (uint64_t)data;
462 	linux_to_bsd_regset(&b_reg1, &reg);
463 	b_reg1.r_err = b_reg.r_err;
464 	b_reg1.r_trapno = b_reg.r_trapno;
465 	return (kern_ptrace(td, PT_SETREGS, pid, &b_reg, 0));
466 }
467 #undef LINUX_URO
468