2 * Copyright (c) 2013 Dmitry Chagin
3 * Copyright (c) 2004 Tim J. Robbins
4 * Copyright (c) 2003 Peter Wemm
5 * Copyright (c) 2002 Doug Rabson
6 * Copyright (c) 1998-1999 Andrew Gallatin
7 * Copyright (c) 1994-1996 Søren Schmidt
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer
15 * in this position and unchanged.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. The name of the author may not be used to endorse or promote products
20 * derived from this software without specific prior written permission
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.
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
37 #define __ELF_WORD_SIZE 64
39 #include <sys/param.h>
40 #include <sys/systm.h>
42 #include <sys/fcntl.h>
43 #include <sys/imgact.h>
44 #include <sys/imgact_elf.h>
45 #include <sys/kernel.h>
48 #include <sys/malloc.h>
49 #include <sys/module.h>
50 #include <sys/mutex.h>
52 #include <sys/resourcevar.h>
53 #include <sys/signalvar.h>
54 #include <sys/syscallsubr.h>
55 #include <sys/sysctl.h>
56 #include <sys/sysent.h>
57 #include <sys/sysproto.h>
58 #include <sys/vnode.h>
59 #include <sys/eventhandler.h>
63 #include <vm/vm_extern.h>
64 #include <vm/vm_map.h>
65 #include <vm/vm_object.h>
66 #include <vm/vm_page.h>
67 #include <vm/vm_param.h>
69 #include <machine/cpu.h>
70 #include <machine/md_var.h>
71 #include <machine/pcb.h>
72 #include <machine/specialreg.h>
73 #include <machine/trap.h>
75 #include <amd64/linux/linux.h>
76 #include <amd64/linux/linux_proto.h>
77 #include <compat/linux/linux_emul.h>
78 #include <compat/linux/linux_ioctl.h>
79 #include <compat/linux/linux_mib.h>
80 #include <compat/linux/linux_misc.h>
81 #include <compat/linux/linux_signal.h>
82 #include <compat/linux/linux_sysproto.h>
83 #include <compat/linux/linux_util.h>
84 #include <compat/linux/linux_vdso.h>
86 MODULE_VERSION(linux64, 1);
88 const char *linux_kplatform;
89 static int linux_szsigcode;
90 static vm_object_t linux_shared_page_obj;
91 static char *linux_shared_page_mapping;
92 extern char _binary_linux_locore_o_start;
93 extern char _binary_linux_locore_o_end;
95 extern struct sysent linux_sysent[LINUX_SYS_MAXSYSCALL];
97 SET_DECLARE(linux_ioctl_handler_set, struct linux_ioctl_handler);
99 static int linux_copyout_strings(struct image_params *imgp,
100 register_t **stack_base);
101 static int linux_fixup_elf(register_t **stack_base,
102 struct image_params *iparams);
103 static bool linux_trans_osrel(const Elf_Note *note, int32_t *osrel);
104 static void linux_vdso_install(void *param);
105 static void linux_vdso_deinstall(void *param);
106 static void linux_set_syscall_retval(struct thread *td, int error);
107 static int linux_fetch_syscall_args(struct thread *td);
108 static void linux_exec_setregs(struct thread *td, struct image_params *imgp,
110 static int linux_vsyscall(struct thread *td);
112 #define LINUX_T_UNKNOWN 255
113 static int _bsd_to_linux_trapcode[] = {
114 LINUX_T_UNKNOWN, /* 0 */
115 6, /* 1 T_PRIVINFLT */
116 LINUX_T_UNKNOWN, /* 2 */
118 LINUX_T_UNKNOWN, /* 4 */
119 LINUX_T_UNKNOWN, /* 5 */
120 16, /* 6 T_ARITHTRAP */
121 254, /* 7 T_ASTFLT */
122 LINUX_T_UNKNOWN, /* 8 */
123 13, /* 9 T_PROTFLT */
124 1, /* 10 T_TRCTRAP */
125 LINUX_T_UNKNOWN, /* 11 */
126 14, /* 12 T_PAGEFLT */
127 LINUX_T_UNKNOWN, /* 13 */
128 17, /* 14 T_ALIGNFLT */
129 LINUX_T_UNKNOWN, /* 15 */
130 LINUX_T_UNKNOWN, /* 16 */
131 LINUX_T_UNKNOWN, /* 17 */
137 8, /* 23 T_DOUBLEFLT */
138 9, /* 24 T_FPOPFLT */
139 10, /* 25 T_TSSFLT */
140 11, /* 26 T_SEGNPFLT */
141 12, /* 27 T_STKFLT */
143 19, /* 29 T_XMMFLT */
144 15 /* 30 T_RESERVED */
146 #define bsd_to_linux_trapcode(code) \
147 ((code)<nitems(_bsd_to_linux_trapcode)? \
148 _bsd_to_linux_trapcode[(code)]: \
151 LINUX_VDSO_SYM_INTPTR(linux_rt_sigcode);
152 LINUX_VDSO_SYM_CHAR(linux_platform);
155 * If FreeBSD & Linux have a difference of opinion about what a trap
156 * means, deal with it here.
161 linux_translate_traps(int signal, int trap_code)
164 if (signal != SIGBUS)
178 linux_fetch_syscall_args(struct thread *td)
181 struct trapframe *frame;
182 struct syscall_args *sa;
185 frame = td->td_frame;
188 sa->args[0] = frame->tf_rdi;
189 sa->args[1] = frame->tf_rsi;
190 sa->args[2] = frame->tf_rdx;
191 sa->args[3] = frame->tf_rcx;
192 sa->args[4] = frame->tf_r8;
193 sa->args[5] = frame->tf_r9;
194 sa->code = frame->tf_rax;
196 if (sa->code >= p->p_sysent->sv_size)
198 sa->callp = &p->p_sysent->sv_table[p->p_sysent->sv_size - 1];
200 sa->callp = &p->p_sysent->sv_table[sa->code];
201 sa->narg = sa->callp->sy_narg;
203 td->td_retval[0] = 0;
208 linux_set_syscall_retval(struct thread *td, int error)
210 struct trapframe *frame = td->td_frame;
213 * On Linux only %rcx and %r11 values are not preserved across
214 * the syscall. So, do not clobber %rdx and %r10.
216 td->td_retval[1] = frame->tf_rdx;
217 if (error != EJUSTRETURN)
218 frame->tf_r10 = frame->tf_rcx;
220 cpu_set_syscall_retval(td, error);
222 /* Restore all registers. */
223 set_pcb_flags(td->td_pcb, PCB_FULL_IRET);
227 linux_copyout_auxargs(struct image_params *imgp, u_long *base)
230 Elf_Auxinfo *argarray, *pos;
233 int error, issetugid;
236 args = (Elf64_Auxargs *)imgp->auxargs;
237 argarray = pos = malloc(LINUX_AT_COUNT * sizeof(*pos), M_TEMP,
240 issetugid = p->p_flag & P_SUGID ? 1 : 0;
241 AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO_EHDR,
242 imgp->proc->p_sysent->sv_shared_page_base);
243 AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, cpu_feature);
244 AUXARGS_ENTRY(pos, LINUX_AT_CLKTCK, stclohz);
245 AUXARGS_ENTRY(pos, AT_PHDR, args->phdr);
246 AUXARGS_ENTRY(pos, AT_PHENT, args->phent);
247 AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum);
248 AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz);
249 AUXARGS_ENTRY(pos, AT_BASE, args->base);
250 AUXARGS_ENTRY(pos, AT_FLAGS, args->flags);
251 AUXARGS_ENTRY(pos, AT_ENTRY, args->entry);
252 AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_ucred->cr_ruid);
253 AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid);
254 AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_ucred->cr_rgid);
255 AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid);
256 AUXARGS_ENTRY(pos, LINUX_AT_SECURE, issetugid);
257 AUXARGS_ENTRY(pos, LINUX_AT_PLATFORM, PTROUT(linux_platform));
258 AUXARGS_ENTRY(pos, LINUX_AT_RANDOM, imgp->canary);
259 if (imgp->execpathp != 0)
260 AUXARGS_ENTRY(pos, LINUX_AT_EXECFN, imgp->execpathp);
261 if (args->execfd != -1)
262 AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd);
263 AUXARGS_ENTRY(pos, AT_NULL, 0);
265 free(imgp->auxargs, M_TEMP);
266 imgp->auxargs = NULL;
267 KASSERT(pos - argarray <= LINUX_AT_COUNT, ("Too many auxargs"));
269 auxlen = sizeof(*argarray) * (pos - argarray);
271 error = copyout(argarray, (void *)*base, auxlen);
272 free(argarray, M_TEMP);
277 linux_fixup_elf(register_t **stack_base, struct image_params *imgp)
281 base = (Elf64_Addr *)*stack_base;
283 if (suword(base, (uint64_t)imgp->args->argc) == -1)
286 *stack_base = (register_t *)base;
291 * Copy strings out to the new process address space, constructing new arg
292 * and env vector tables. Return a pointer to the base so that it can be used
293 * as the initial stack pointer.
296 linux_copyout_strings(struct image_params *imgp, register_t **stack_base)
298 int argc, envc, error;
300 char *stringp, *destp;
301 struct ps_strings *arginfo;
302 char canary[LINUX_AT_RANDOM_LEN];
306 /* Calculate string base and vector table pointers. */
307 if (imgp->execpath != NULL && imgp->auxargs != NULL)
308 execpath_len = strlen(imgp->execpath) + 1;
313 arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings;
314 destp = (caddr_t)arginfo - SPARE_USRSPACE -
315 roundup(sizeof(canary), sizeof(char *)) -
316 roundup(execpath_len, sizeof(char *)) -
317 roundup(ARG_MAX - imgp->args->stringspace, sizeof(char *));
319 if (execpath_len != 0) {
320 imgp->execpathp = (uintptr_t)arginfo - execpath_len;
321 error = copyout(imgp->execpath, (void *)imgp->execpathp,
327 /* Prepare the canary for SSP. */
328 arc4rand(canary, sizeof(canary), 0);
329 imgp->canary = (uintptr_t)arginfo -
330 roundup(execpath_len, sizeof(char *)) -
331 roundup(sizeof(canary), sizeof(char *));
332 error = copyout(canary, (void *)imgp->canary, sizeof(canary));
336 vectp = (char **)destp;
339 * Starting with 2.24, glibc depends on a 16-byte stack alignment.
340 * One "long argc" will be prepended later.
342 vectp = (char **)((((uintptr_t)vectp + 8) & ~0xF) - 8);
345 error = imgp->sysent->sv_copyout_auxargs(imgp,
352 * Allocate room for the argv[] and env vectors including the
353 * terminating NULL pointers.
355 vectp -= imgp->args->argc + 1 + imgp->args->envc + 1;
357 /* vectp also becomes our initial stack base. */
358 *stack_base = (register_t *)vectp;
360 stringp = imgp->args->begin_argv;
361 argc = imgp->args->argc;
362 envc = imgp->args->envc;
364 /* Copy out strings - arguments and environment. */
365 error = copyout(stringp, destp, ARG_MAX - imgp->args->stringspace);
369 /* Fill in "ps_strings" struct for ps, w, etc. */
370 if (suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp) != 0 ||
371 suword(&arginfo->ps_nargvstr, argc) != 0)
374 /* Fill in argument portion of vector table. */
375 for (; argc > 0; --argc) {
376 if (suword(vectp++, (long)(intptr_t)destp) != 0)
378 while (*stringp++ != 0)
383 /* A null vector table pointer separates the argp's from the envp's. */
384 if (suword(vectp++, 0) != 0)
387 if (suword(&arginfo->ps_envstr, (long)(intptr_t)vectp) != 0 ||
388 suword(&arginfo->ps_nenvstr, envc) != 0)
391 /* Fill in environment portion of vector table. */
392 for (; envc > 0; --envc) {
393 if (suword(vectp++, (long)(intptr_t)destp) != 0)
395 while (*stringp++ != 0)
400 /* The end of the vector table is a null pointer. */
401 if (suword(vectp, 0) != 0)
408 * Reset registers to default values on exec.
411 linux_exec_setregs(struct thread *td, struct image_params *imgp, u_long stack)
413 struct trapframe *regs;
415 register_t saved_rflags;
420 if (td->td_proc->p_md.md_ldt != NULL)
425 clear_pcb_flags(pcb, PCB_32BIT);
426 pcb->pcb_initial_fpucw = __LINUX_NPXCW__;
427 set_pcb_flags(pcb, PCB_FULL_IRET);
429 saved_rflags = regs->tf_rflags & PSL_T;
430 bzero((char *)regs, sizeof(struct trapframe));
431 regs->tf_rip = imgp->entry_addr;
432 regs->tf_rsp = stack;
433 regs->tf_rflags = PSL_USER | saved_rflags;
434 regs->tf_ss = _udatasel;
435 regs->tf_cs = _ucodesel;
436 regs->tf_ds = _udatasel;
437 regs->tf_es = _udatasel;
438 regs->tf_fs = _ufssel;
439 regs->tf_gs = _ugssel;
440 regs->tf_flags = TF_HASSEGS;
443 * Reset the hardware debug registers if they were in use.
444 * They won't have any meaning for the newly exec'd process.
446 if (pcb->pcb_flags & PCB_DBREGS) {
455 * Clear the debug registers on the running
456 * CPU, otherwise they will end up affecting
457 * the next process we switch to.
461 clear_pcb_flags(pcb, PCB_DBREGS);
465 * Drop the FP state if we hold it, so that the process gets a
466 * clean FP state if it uses the FPU again.
472 * Copied from amd64/amd64/machdep.c
474 * XXX fpu state need? don't think so
477 linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args)
480 struct l_ucontext uc;
481 struct l_sigcontext *context;
482 struct trapframe *regs;
483 unsigned long rflags;
488 error = copyin((void *)regs->tf_rbx, &uc, sizeof(uc));
493 context = &uc.uc_mcontext;
494 rflags = context->sc_rflags;
497 * Don't allow users to change privileged or reserved flags.
500 * XXX do allow users to change the privileged flag PSL_RF.
501 * The cpu sets PSL_RF in tf_rflags for faults. Debuggers
502 * should sometimes set it there too. tf_rflags is kept in
503 * the signal context during signal handling and there is no
504 * other place to remember it, so the PSL_RF bit may be
505 * corrupted by the signal handler without us knowing.
506 * Corruption of the PSL_RF bit at worst causes one more or
507 * one less debugger trap, so allowing it is fairly harmless.
510 #define RFLAG_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0)
511 if (!RFLAG_SECURE(rflags & ~PSL_RF, regs->tf_rflags & ~PSL_RF)) {
512 printf("linux_rt_sigreturn: rflags = 0x%lx\n", rflags);
517 * Don't allow users to load a valid privileged %cs. Let the
518 * hardware check for invalid selectors, excess privilege in
519 * other selectors, invalid %eip's and invalid %esp's.
521 #define CS_SECURE(cs) (ISPL(cs) == SEL_UPL)
522 if (!CS_SECURE(context->sc_cs)) {
523 printf("linux_rt_sigreturn: cs = 0x%x\n", context->sc_cs);
524 ksiginfo_init_trap(&ksi);
525 ksi.ksi_signo = SIGBUS;
526 ksi.ksi_code = BUS_OBJERR;
527 ksi.ksi_trapno = T_PROTFLT;
528 ksi.ksi_addr = (void *)regs->tf_rip;
529 trapsignal(td, &ksi);
534 linux_to_bsd_sigset(&uc.uc_sigmask, &td->td_sigmask);
535 SIG_CANTMASK(td->td_sigmask);
539 regs->tf_rdi = context->sc_rdi;
540 regs->tf_rsi = context->sc_rsi;
541 regs->tf_rdx = context->sc_rdx;
542 regs->tf_rbp = context->sc_rbp;
543 regs->tf_rbx = context->sc_rbx;
544 regs->tf_rcx = context->sc_rcx;
545 regs->tf_rax = context->sc_rax;
546 regs->tf_rip = context->sc_rip;
547 regs->tf_rsp = context->sc_rsp;
548 regs->tf_r8 = context->sc_r8;
549 regs->tf_r9 = context->sc_r9;
550 regs->tf_r10 = context->sc_r10;
551 regs->tf_r11 = context->sc_r11;
552 regs->tf_r12 = context->sc_r12;
553 regs->tf_r13 = context->sc_r13;
554 regs->tf_r14 = context->sc_r14;
555 regs->tf_r15 = context->sc_r15;
556 regs->tf_cs = context->sc_cs;
557 regs->tf_err = context->sc_err;
558 regs->tf_rflags = rflags;
560 set_pcb_flags(td->td_pcb, PCB_FULL_IRET);
561 return (EJUSTRETURN);
565 * copied from amd64/amd64/machdep.c
567 * Send an interrupt to process.
570 linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
572 struct l_rt_sigframe sf, *sfp;
577 struct trapframe *regs;
583 PROC_LOCK_ASSERT(p, MA_OWNED);
584 sig = ksi->ksi_signo;
586 code = ksi->ksi_code;
587 mtx_assert(&psp->ps_mtx, MA_OWNED);
589 oonstack = sigonstack(regs->tf_rsp);
591 LINUX_CTR4(rt_sendsig, "%p, %d, %p, %u",
592 catcher, sig, mask, code);
594 /* Allocate space for the signal handler context. */
595 if ((td->td_pflags & TDP_ALTSTACK) != 0 && !oonstack &&
596 SIGISMEMBER(psp->ps_sigonstack, sig)) {
597 sp = (caddr_t)td->td_sigstk.ss_sp + td->td_sigstk.ss_size -
598 sizeof(struct l_rt_sigframe);
600 sp = (caddr_t)regs->tf_rsp - sizeof(struct l_rt_sigframe) - 128;
601 /* Align to 16 bytes. */
602 sfp = (struct l_rt_sigframe *)((unsigned long)sp & ~0xFul);
603 mtx_unlock(&psp->ps_mtx);
605 /* Translate the signal. */
606 sig = bsd_to_linux_signal(sig);
608 /* Save user context. */
609 bzero(&sf, sizeof(sf));
610 bsd_to_linux_sigset(mask, &sf.sf_sc.uc_sigmask);
611 bsd_to_linux_sigset(mask, &sf.sf_sc.uc_mcontext.sc_mask);
613 sf.sf_sc.uc_stack.ss_sp = PTROUT(td->td_sigstk.ss_sp);
614 sf.sf_sc.uc_stack.ss_size = td->td_sigstk.ss_size;
615 sf.sf_sc.uc_stack.ss_flags = (td->td_pflags & TDP_ALTSTACK)
616 ? ((oonstack) ? LINUX_SS_ONSTACK : 0) : LINUX_SS_DISABLE;
619 sf.sf_sc.uc_mcontext.sc_rdi = regs->tf_rdi;
620 sf.sf_sc.uc_mcontext.sc_rsi = regs->tf_rsi;
621 sf.sf_sc.uc_mcontext.sc_rdx = regs->tf_rdx;
622 sf.sf_sc.uc_mcontext.sc_rbp = regs->tf_rbp;
623 sf.sf_sc.uc_mcontext.sc_rbx = regs->tf_rbx;
624 sf.sf_sc.uc_mcontext.sc_rcx = regs->tf_rcx;
625 sf.sf_sc.uc_mcontext.sc_rax = regs->tf_rax;
626 sf.sf_sc.uc_mcontext.sc_rip = regs->tf_rip;
627 sf.sf_sc.uc_mcontext.sc_rsp = regs->tf_rsp;
628 sf.sf_sc.uc_mcontext.sc_r8 = regs->tf_r8;
629 sf.sf_sc.uc_mcontext.sc_r9 = regs->tf_r9;
630 sf.sf_sc.uc_mcontext.sc_r10 = regs->tf_r10;
631 sf.sf_sc.uc_mcontext.sc_r11 = regs->tf_r11;
632 sf.sf_sc.uc_mcontext.sc_r12 = regs->tf_r12;
633 sf.sf_sc.uc_mcontext.sc_r13 = regs->tf_r13;
634 sf.sf_sc.uc_mcontext.sc_r14 = regs->tf_r14;
635 sf.sf_sc.uc_mcontext.sc_r15 = regs->tf_r15;
636 sf.sf_sc.uc_mcontext.sc_cs = regs->tf_cs;
637 sf.sf_sc.uc_mcontext.sc_rflags = regs->tf_rflags;
638 sf.sf_sc.uc_mcontext.sc_err = regs->tf_err;
639 sf.sf_sc.uc_mcontext.sc_trapno = bsd_to_linux_trapcode(code);
640 sf.sf_sc.uc_mcontext.sc_cr2 = (register_t)ksi->ksi_addr;
642 /* Build the argument list for the signal handler. */
643 regs->tf_rdi = sig; /* arg 1 in %rdi */
645 regs->tf_rsi = (register_t)&sfp->sf_si; /* arg 2 in %rsi */
646 regs->tf_rdx = (register_t)&sfp->sf_sc; /* arg 3 in %rdx */
648 sf.sf_handler = catcher;
649 /* Fill in POSIX parts. */
650 ksiginfo_to_lsiginfo(ksi, &sf.sf_si, sig);
652 /* Copy the sigframe out to the user's stack. */
653 if (copyout(&sf, sfp, sizeof(*sfp)) != 0) {
658 regs->tf_rsp = (long)sfp;
659 regs->tf_rip = linux_rt_sigcode;
660 regs->tf_rflags &= ~(PSL_T | PSL_D);
661 regs->tf_cs = _ucodesel;
662 set_pcb_flags(td->td_pcb, PCB_FULL_IRET);
664 mtx_lock(&psp->ps_mtx);
667 #define LINUX_VSYSCALL_START (-10UL << 20)
668 #define LINUX_VSYSCALL_SZ 1024
670 const unsigned long linux_vsyscall_vector[] = {
671 LINUX_SYS_gettimeofday,
672 LINUX_SYS_linux_time,
673 /* getcpu not implemented */
677 linux_vsyscall(struct thread *td)
679 struct trapframe *frame;
684 frame = td->td_frame;
686 /* Check %rip for vsyscall area. */
687 if (__predict_true(frame->tf_rip < LINUX_VSYSCALL_START))
689 if ((frame->tf_rip & (LINUX_VSYSCALL_SZ - 1)) != 0)
691 code = (frame->tf_rip - LINUX_VSYSCALL_START) / LINUX_VSYSCALL_SZ;
692 if (code >= nitems(linux_vsyscall_vector))
696 * vsyscall called as callq *(%rax), so we must
697 * use return address from %rsp and also fixup %rsp.
699 error = copyin((void *)frame->tf_rsp, &retqaddr, sizeof(retqaddr));
703 frame->tf_rip = retqaddr;
704 frame->tf_rax = linux_vsyscall_vector[code];
707 traced = (frame->tf_flags & PSL_T);
709 amd64_syscall(td, traced);
714 struct sysentvec elf_linux_sysvec = {
715 .sv_size = LINUX_SYS_MAXSYSCALL,
716 .sv_table = linux_sysent,
717 .sv_errsize = ELAST + 1,
718 .sv_errtbl = linux_errtbl,
719 .sv_transtrap = linux_translate_traps,
720 .sv_fixup = linux_fixup_elf,
721 .sv_sendsig = linux_rt_sendsig,
722 .sv_sigcode = &_binary_linux_locore_o_start,
723 .sv_szsigcode = &linux_szsigcode,
724 .sv_name = "Linux ELF64",
725 .sv_coredump = elf64_coredump,
726 .sv_imgact_try = linux_exec_imgact_try,
727 .sv_minsigstksz = LINUX_MINSIGSTKSZ,
728 .sv_minuser = VM_MIN_ADDRESS,
729 .sv_maxuser = VM_MAXUSER_ADDRESS,
730 .sv_usrstack = USRSTACK,
731 .sv_psstrings = PS_STRINGS,
732 .sv_stackprot = VM_PROT_ALL,
733 .sv_copyout_auxargs = linux_copyout_auxargs,
734 .sv_copyout_strings = linux_copyout_strings,
735 .sv_setregs = linux_exec_setregs,
738 .sv_flags = SV_ABI_LINUX | SV_LP64 | SV_SHP,
739 .sv_set_syscall_retval = linux_set_syscall_retval,
740 .sv_fetch_syscall_args = linux_fetch_syscall_args,
741 .sv_syscallnames = NULL,
742 .sv_shared_page_base = SHAREDPAGE,
743 .sv_shared_page_len = PAGE_SIZE,
744 .sv_schedtail = linux_schedtail,
745 .sv_thread_detach = linux_thread_detach,
746 .sv_trap = linux_vsyscall,
750 linux_vdso_install(void *param)
753 amd64_lower_shared_page(&elf_linux_sysvec);
755 linux_szsigcode = (&_binary_linux_locore_o_end -
756 &_binary_linux_locore_o_start);
758 if (linux_szsigcode > elf_linux_sysvec.sv_shared_page_len)
759 panic("Linux invalid vdso size\n");
761 __elfN(linux_vdso_fixup)(&elf_linux_sysvec);
763 linux_shared_page_obj = __elfN(linux_shared_page_init)
764 (&linux_shared_page_mapping);
766 __elfN(linux_vdso_reloc)(&elf_linux_sysvec);
768 bcopy(elf_linux_sysvec.sv_sigcode, linux_shared_page_mapping,
770 elf_linux_sysvec.sv_shared_page_obj = linux_shared_page_obj;
772 linux_kplatform = linux_shared_page_mapping +
773 (linux_platform - (caddr_t)elf_linux_sysvec.sv_shared_page_base);
775 SYSINIT(elf_linux_vdso_init, SI_SUB_EXEC, SI_ORDER_ANY,
776 linux_vdso_install, NULL);
779 linux_vdso_deinstall(void *param)
782 __elfN(linux_shared_page_fini)(linux_shared_page_obj);
784 SYSUNINIT(elf_linux_vdso_uninit, SI_SUB_EXEC, SI_ORDER_FIRST,
785 linux_vdso_deinstall, NULL);
787 static char GNULINUX_ABI_VENDOR[] = "GNU";
788 static int GNULINUX_ABI_DESC = 0;
791 linux_trans_osrel(const Elf_Note *note, int32_t *osrel)
793 const Elf32_Word *desc;
796 p = (uintptr_t)(note + 1);
797 p += roundup2(note->n_namesz, sizeof(Elf32_Addr));
799 desc = (const Elf32_Word *)p;
800 if (desc[0] != GNULINUX_ABI_DESC)
804 * For Linux we encode osrel using the Linux convention of
805 * (version << 16) | (major << 8) | (minor)
806 * See macro in linux_mib.h
808 *osrel = LINUX_KERNVER(desc[1], desc[2], desc[3]);
813 static Elf_Brandnote linux64_brandnote = {
814 .hdr.n_namesz = sizeof(GNULINUX_ABI_VENDOR),
817 .vendor = GNULINUX_ABI_VENDOR,
818 .flags = BN_TRANSLATE_OSREL,
819 .trans_osrel = linux_trans_osrel
822 static Elf64_Brandinfo linux_glibc2brand = {
823 .brand = ELFOSABI_LINUX,
824 .machine = EM_X86_64,
825 .compat_3_brand = "Linux",
826 .emul_path = "/compat/linux",
827 .interp_path = "/lib64/ld-linux-x86-64.so.2",
828 .sysvec = &elf_linux_sysvec,
829 .interp_newpath = NULL,
830 .brand_note = &linux64_brandnote,
831 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
834 static Elf64_Brandinfo linux_glibc2brandshort = {
835 .brand = ELFOSABI_LINUX,
836 .machine = EM_X86_64,
837 .compat_3_brand = "Linux",
838 .emul_path = "/compat/linux",
839 .interp_path = "/lib64/ld-linux.so.2",
840 .sysvec = &elf_linux_sysvec,
841 .interp_newpath = NULL,
842 .brand_note = &linux64_brandnote,
843 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
846 static Elf64_Brandinfo linux_muslbrand = {
847 .brand = ELFOSABI_LINUX,
848 .machine = EM_X86_64,
849 .compat_3_brand = "Linux",
850 .emul_path = "/compat/linux",
851 .interp_path = "/lib/ld-musl-x86_64.so.1",
852 .sysvec = &elf_linux_sysvec,
853 .interp_newpath = NULL,
854 .brand_note = &linux64_brandnote,
855 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
858 Elf64_Brandinfo *linux_brandlist[] = {
860 &linux_glibc2brandshort,
866 linux64_elf_modevent(module_t mod, int type, void *data)
868 Elf64_Brandinfo **brandinfo;
870 struct linux_ioctl_handler **lihp;
876 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL;
878 if (elf64_insert_brand_entry(*brandinfo) < 0)
881 SET_FOREACH(lihp, linux_ioctl_handler_set)
882 linux_ioctl_register_handler(*lihp);
883 stclohz = (stathz ? stathz : hz);
885 printf("Linux x86-64 ELF exec handler installed\n");
887 printf("cannot insert Linux x86-64 ELF brand handler\n");
890 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL;
892 if (elf64_brand_inuse(*brandinfo))
895 for (brandinfo = &linux_brandlist[0];
896 *brandinfo != NULL; ++brandinfo)
897 if (elf64_remove_brand_entry(*brandinfo) < 0)
901 SET_FOREACH(lihp, linux_ioctl_handler_set)
902 linux_ioctl_unregister_handler(*lihp);
904 printf("Linux ELF exec handler removed\n");
906 printf("Could not deinstall ELF interpreter entry\n");
914 static moduledata_t linux64_elf_mod = {
916 linux64_elf_modevent,
920 DECLARE_MODULE_TIED(linux64elf, linux64_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY);
921 MODULE_DEPEND(linux64elf, linux_common, 1, 1, 1);
922 FEATURE(linux64, "Linux 64bit support");