xref: /freebsd/sys/amd64/linux/linux_machdep.c (revision 0c785f06020f3b02e34c97eb27fecd3af8eb2a7b)
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/param.h>
33 #include <sys/systm.h>
34 #include <sys/ktr.h>
35 #include <sys/lock.h>
36 #include <sys/malloc.h>
37 #include <sys/mman.h>
38 #include <sys/mutex.h>
39 #include <sys/priv.h>
40 #include <sys/proc.h>
41 #include <sys/ptrace.h>
42 #include <sys/syscallsubr.h>
43 
44 #include <security/mac/mac_framework.h>
45 
46 #include <ufs/ufs/extattr.h>
47 #include <ufs/ufs/quota.h>
48 #include <ufs/ufs/ufsmount.h>
49 
50 #include <machine/frame.h>
51 #include <machine/md_var.h>
52 #include <machine/pcb.h>
53 #include <machine/psl.h>
54 #include <machine/segments.h>
55 #include <machine/specialreg.h>
56 
57 #include <vm/pmap.h>
58 #include <vm/vm.h>
59 #include <vm/vm_param.h>
60 #include <vm/vm_extern.h>
61 #include <vm/vm_kern.h>
62 #include <vm/vm_map.h>
63 
64 #include <x86/ifunc.h>
65 #include <x86/reg.h>
66 #include <x86/sysarch.h>
67 
68 #include <amd64/linux/linux.h>
69 #include <amd64/linux/linux_proto.h>
70 #include <compat/linux/linux_fork.h>
71 #include <compat/linux/linux_misc.h>
72 #include <compat/linux/linux_mmap.h>
73 #include <compat/linux/linux_util.h>
74 
75 #define	LINUX_ARCH_AMD64		0xc000003e
76 
77 int
78 linux_set_upcall(struct thread *td, register_t stack)
79 {
80 
81 	if (stack)
82 		td->td_frame->tf_rsp = stack;
83 
84 	/*
85 	 * The newly created Linux thread returns
86 	 * to the user space by the same path that a parent does.
87 	 */
88 	td->td_frame->tf_rax = 0;
89 	return (0);
90 }
91 
92 int
93 linux_mmap2(struct thread *td, struct linux_mmap2_args *args)
94 {
95 
96 	return (linux_mmap_common(td, args->addr, args->len, args->prot,
97 		args->flags, args->fd, args->pgoff));
98 }
99 
100 int
101 linux_mprotect(struct thread *td, struct linux_mprotect_args *uap)
102 {
103 
104 	return (linux_mprotect_common(td, uap->addr, uap->len, uap->prot));
105 }
106 
107 int
108 linux_madvise(struct thread *td, struct linux_madvise_args *uap)
109 {
110 
111 	return (linux_madvise_common(td, uap->addr, uap->len, uap->behav));
112 }
113 
114 int
115 linux_iopl(struct thread *td, struct linux_iopl_args *args)
116 {
117 	int error;
118 
119 	LINUX_CTR(iopl);
120 
121 	if (args->level > 3)
122 		return (EINVAL);
123 	if ((error = priv_check(td, PRIV_IO)) != 0)
124 		return (error);
125 	if ((error = securelevel_gt(td->td_ucred, 0)) != 0)
126 		return (error);
127 	td->td_frame->tf_rflags = (td->td_frame->tf_rflags & ~PSL_IOPL) |
128 	    (args->level * (PSL_IOPL / 3));
129 
130 	return (0);
131 }
132 
133 int
134 linux_pause(struct thread *td, struct linux_pause_args *args)
135 {
136 	struct proc *p = td->td_proc;
137 	sigset_t sigmask;
138 
139 	LINUX_CTR(pause);
140 
141 	PROC_LOCK(p);
142 	sigmask = td->td_sigmask;
143 	PROC_UNLOCK(p);
144 	return (kern_sigsuspend(td, sigmask));
145 }
146 
147 int
148 linux_arch_prctl(struct thread *td, struct linux_arch_prctl_args *args)
149 {
150 	unsigned long long cet[3];
151 	struct pcb *pcb;
152 	int error;
153 
154 	pcb = td->td_pcb;
155 	LINUX_CTR2(arch_prctl, "0x%x, %p", args->code, args->addr);
156 
157 	switch (args->code) {
158 	case LINUX_ARCH_SET_GS:
159 		if (args->addr < VM_MAXUSER_ADDRESS) {
160 			update_pcb_bases(pcb);
161 			pcb->pcb_gsbase = args->addr;
162 			td->td_frame->tf_gs = _ugssel;
163 			error = 0;
164 		} else
165 			error = EPERM;
166 		break;
167 	case LINUX_ARCH_SET_FS:
168 		if (args->addr < VM_MAXUSER_ADDRESS) {
169 			update_pcb_bases(pcb);
170 			pcb->pcb_fsbase = args->addr;
171 			td->td_frame->tf_fs = _ufssel;
172 			error = 0;
173 		} else
174 			error = EPERM;
175 		break;
176 	case LINUX_ARCH_GET_FS:
177 		error = copyout(&pcb->pcb_fsbase, PTRIN(args->addr),
178 		    sizeof(args->addr));
179 		break;
180 	case LINUX_ARCH_GET_GS:
181 		error = copyout(&pcb->pcb_gsbase, PTRIN(args->addr),
182 		    sizeof(args->addr));
183 		break;
184 	case LINUX_ARCH_CET_STATUS:
185 		memset(cet, 0, sizeof(cet));
186 		error = copyout(&cet, PTRIN(args->addr), sizeof(cet));
187 		break;
188 	default:
189 		linux_msg(td, "unsupported arch_prctl code %#x", args->code);
190 		error = EINVAL;
191 	}
192 	return (error);
193 }
194 
195 int
196 linux_set_cloned_tls(struct thread *td, void *desc)
197 {
198 	struct pcb *pcb;
199 
200 	if ((uint64_t)desc >= VM_MAXUSER_ADDRESS)
201 		return (EPERM);
202 
203 	pcb = td->td_pcb;
204 	update_pcb_bases(pcb);
205 	pcb->pcb_fsbase = (register_t)desc;
206 	td->td_frame->tf_fs = _ufssel;
207 
208 	return (0);
209 }
210 
211 int futex_xchgl_nosmap(int oparg, uint32_t *uaddr, int *oldval);
212 int futex_xchgl_smap(int oparg, uint32_t *uaddr, int *oldval);
213 DEFINE_IFUNC(, int, futex_xchgl, (int, uint32_t *, int *))
214 {
215 
216 	return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ?
217 	    futex_xchgl_smap : futex_xchgl_nosmap);
218 }
219 
220 int futex_addl_nosmap(int oparg, uint32_t *uaddr, int *oldval);
221 int futex_addl_smap(int oparg, uint32_t *uaddr, int *oldval);
222 DEFINE_IFUNC(, int, futex_addl, (int, uint32_t *, int *))
223 {
224 
225 	return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ?
226 	    futex_addl_smap : futex_addl_nosmap);
227 }
228 
229 int futex_orl_nosmap(int oparg, uint32_t *uaddr, int *oldval);
230 int futex_orl_smap(int oparg, uint32_t *uaddr, int *oldval);
231 DEFINE_IFUNC(, int, futex_orl, (int, uint32_t *, int *))
232 {
233 
234 	return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ?
235 	    futex_orl_smap : futex_orl_nosmap);
236 }
237 
238 int futex_andl_nosmap(int oparg, uint32_t *uaddr, int *oldval);
239 int futex_andl_smap(int oparg, uint32_t *uaddr, int *oldval);
240 DEFINE_IFUNC(, int, futex_andl, (int, uint32_t *, int *))
241 {
242 
243 	return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ?
244 	    futex_andl_smap : futex_andl_nosmap);
245 }
246 
247 int futex_xorl_nosmap(int oparg, uint32_t *uaddr, int *oldval);
248 int futex_xorl_smap(int oparg, uint32_t *uaddr, int *oldval);
249 DEFINE_IFUNC(, int, futex_xorl, (int, uint32_t *, int *))
250 {
251 
252 	return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ?
253 	    futex_xorl_smap : futex_xorl_nosmap);
254 }
255 
256 void
257 bsd_to_linux_regset(const struct reg *b_reg, struct linux_pt_regset *l_regset)
258 {
259 
260 	l_regset->r15 = b_reg->r_r15;
261 	l_regset->r14 = b_reg->r_r14;
262 	l_regset->r13 = b_reg->r_r13;
263 	l_regset->r12 = b_reg->r_r12;
264 	l_regset->rbp = b_reg->r_rbp;
265 	l_regset->rbx = b_reg->r_rbx;
266 	l_regset->r11 = b_reg->r_r11;
267 	l_regset->r10 = b_reg->r_r10;
268 	l_regset->r9 = b_reg->r_r9;
269 	l_regset->r8 = b_reg->r_r8;
270 	l_regset->rax = b_reg->r_rax;
271 	l_regset->rcx = b_reg->r_rcx;
272 	l_regset->rdx = b_reg->r_rdx;
273 	l_regset->rsi = b_reg->r_rsi;
274 	l_regset->rdi = b_reg->r_rdi;
275 	l_regset->orig_rax = b_reg->r_rax;
276 	l_regset->rip = b_reg->r_rip;
277 	l_regset->cs = b_reg->r_cs;
278 	l_regset->eflags = b_reg->r_rflags;
279 	l_regset->rsp = b_reg->r_rsp;
280 	l_regset->ss = b_reg->r_ss;
281 	l_regset->fs_base = 0;
282 	l_regset->gs_base = 0;
283 	l_regset->ds = b_reg->r_ds;
284 	l_regset->es = b_reg->r_es;
285 	l_regset->fs = b_reg->r_fs;
286 	l_regset->gs = b_reg->r_gs;
287 }
288 
289 void
290 linux_to_bsd_regset(struct reg *b_reg, const struct linux_pt_regset *l_regset)
291 {
292 
293 	b_reg->r_r15 = l_regset->r15;
294 	b_reg->r_r14 = l_regset->r14;
295 	b_reg->r_r13 = l_regset->r13;
296 	b_reg->r_r12 = l_regset->r12;
297 	b_reg->r_rbp = l_regset->rbp;
298 	b_reg->r_rbx = l_regset->rbx;
299 	b_reg->r_r11 = l_regset->r11;
300 	b_reg->r_r10 = l_regset->r10;
301 	b_reg->r_r9 = l_regset->r9;
302 	b_reg->r_r8 = l_regset->r8;
303 	b_reg->r_rax = l_regset->rax;
304 	b_reg->r_rcx = l_regset->rcx;
305 	b_reg->r_rdx = l_regset->rdx;
306 	b_reg->r_rsi = l_regset->rsi;
307 	b_reg->r_rdi = l_regset->rdi;
308 	b_reg->r_rax = l_regset->orig_rax;
309 	b_reg->r_rip = l_regset->rip;
310 	b_reg->r_cs = l_regset->cs;
311 	b_reg->r_rflags = l_regset->eflags;
312 	b_reg->r_rsp = l_regset->rsp;
313 	b_reg->r_ss = l_regset->ss;
314 	b_reg->r_ds = l_regset->ds;
315 	b_reg->r_es = l_regset->es;
316 	b_reg->r_fs = l_regset->fs;
317 	b_reg->r_gs = l_regset->gs;
318 }
319 
320 void
321 linux_ptrace_get_syscall_info_machdep(const struct reg *reg,
322     struct syscall_info *si)
323 {
324 
325 	si->arch = LINUX_ARCH_AMD64;
326 	si->instruction_pointer = reg->r_rip;
327 	si->stack_pointer = reg->r_rsp;
328 }
329 
330 int
331 linux_ptrace_getregs_machdep(struct thread *td, pid_t pid,
332     struct linux_pt_regset *l_regset)
333 {
334 	struct ptrace_lwpinfo lwpinfo;
335 	struct pcb *pcb;
336 	int error;
337 
338 	pcb = td->td_pcb;
339 	if (td == curthread)
340 		update_pcb_bases(pcb);
341 
342 	l_regset->fs_base = pcb->pcb_fsbase;
343 	l_regset->gs_base = pcb->pcb_gsbase;
344 
345 	error = kern_ptrace(td, PT_LWPINFO, pid, &lwpinfo, sizeof(lwpinfo));
346 	if (error != 0) {
347 		linux_msg(td, "PT_LWPINFO failed with error %d", error);
348 		return (error);
349 	}
350 	if ((lwpinfo.pl_flags & (PL_FLAG_SCE | PL_FLAG_SCX)) != 0) {
351 		/*
352 		 * In Linux, the syscall number - passed to the syscall
353 		 * as rax - is preserved in orig_rax; rax gets overwritten
354 		 * with syscall return value.
355 		 */
356 		l_regset->orig_rax = lwpinfo.pl_syscall_code;
357 	}
358 
359 	return (0);
360 }
361 
362 #define	LINUX_URO(a,m) ((uintptr_t)a == offsetof(struct linux_pt_regset, m))
363 
364 int
365 linux_ptrace_peekuser(struct thread *td, pid_t pid, void *addr, void *data)
366 {
367 	struct linux_pt_regset reg;
368 	struct reg b_reg;
369 	uint64_t val;
370 	int error;
371 
372 	if ((uintptr_t)addr & (sizeof(data) -1) || (uintptr_t)addr < 0)
373 		return (EIO);
374 	if ((uintptr_t)addr >= sizeof(struct linux_pt_regset)) {
375 		LINUX_RATELIMIT_MSG_OPT1("PTRACE_PEEKUSER offset %ld "
376 		    "not implemented; returning EINVAL", (uintptr_t)addr);
377 		return (EINVAL);
378 	}
379 
380 	if (LINUX_URO(addr, fs_base))
381 		return (kern_ptrace(td, PT_GETFSBASE, pid, data, 0));
382 	if (LINUX_URO(addr, gs_base))
383 		return (kern_ptrace(td, PT_GETGSBASE, pid, data, 0));
384 	if ((error = kern_ptrace(td, PT_GETREGS, pid, &b_reg, 0)) != 0)
385 		return (error);
386 	bsd_to_linux_regset(&b_reg, &reg);
387 	val = *(&reg.r15 + ((uintptr_t)addr / sizeof(reg.r15)));
388 	return (copyout(&val, data, sizeof(val)));
389 }
390 
391 static inline bool
392 linux_invalid_selector(u_short val)
393 {
394 
395 	return (val != 0 && ISPL(val) != SEL_UPL);
396 }
397 
398 struct linux_segreg_off {
399 	uintptr_t	reg;
400 	bool		is0;
401 };
402 
403 const struct linux_segreg_off linux_segregs_off[] = {
404 	{
405 		.reg = offsetof(struct linux_pt_regset, gs),
406 		.is0 = true,
407 	},
408 	{
409 		.reg = offsetof(struct linux_pt_regset, fs),
410 		.is0 = true,
411 	},
412 	{
413 		.reg = offsetof(struct linux_pt_regset, ds),
414 		.is0 = true,
415 	},
416 	{
417 		.reg = offsetof(struct linux_pt_regset, es),
418 		.is0 = true,
419 	},
420 	{
421 		.reg = offsetof(struct linux_pt_regset, cs),
422 		.is0 = false,
423 	},
424 	{
425 		.reg = offsetof(struct linux_pt_regset, ss),
426 		.is0 = false,
427 	},
428 };
429 
430 int
431 linux_ptrace_pokeuser(struct thread *td, pid_t pid, void *addr, void *data)
432 {
433 	struct linux_pt_regset reg;
434 	struct reg b_reg, b_reg1;
435 	int error, i;
436 
437 	if ((uintptr_t)addr & (sizeof(data) -1) || (uintptr_t)addr < 0)
438 		return (EIO);
439 	if ((uintptr_t)addr >= sizeof(struct linux_pt_regset)) {
440 		LINUX_RATELIMIT_MSG_OPT1("PTRACE_POKEUSER offset %ld "
441 		    "not implemented; returning EINVAL", (uintptr_t)addr);
442 		return (EINVAL);
443 	}
444 
445 	if (LINUX_URO(addr, fs_base))
446 		return (kern_ptrace(td, PT_SETFSBASE, pid, data, 0));
447 	if (LINUX_URO(addr, gs_base))
448 		return (kern_ptrace(td, PT_SETGSBASE, pid, data, 0));
449 	for (i = 0; i < nitems(linux_segregs_off); i++) {
450 		if ((uintptr_t)addr == linux_segregs_off[i].reg) {
451 			if (linux_invalid_selector((uintptr_t)data))
452 				return (EIO);
453 			if (!linux_segregs_off[i].is0 && (uintptr_t)data == 0)
454 				return (EIO);
455 		}
456 	}
457 	if ((error = kern_ptrace(td, PT_GETREGS, pid, &b_reg, 0)) != 0)
458 		return (error);
459 	bsd_to_linux_regset(&b_reg, &reg);
460 	*(&reg.r15 + ((uintptr_t)addr / sizeof(reg.r15))) = (uint64_t)data;
461 	linux_to_bsd_regset(&b_reg1, &reg);
462 	b_reg1.r_err = b_reg.r_err;
463 	b_reg1.r_trapno = b_reg.r_trapno;
464 	return (kern_ptrace(td, PT_SETREGS, pid, &b_reg, 0));
465 }
466 #undef LINUX_URO
467