diff --git a/sys/amd64/linux/linux_machdep.c b/sys/amd64/linux/linux_machdep.c index 5de27a7ee7d3..dc6b6140f60c 100644 --- a/sys/amd64/linux/linux_machdep.c +++ b/sys/amd64/linux/linux_machdep.c @@ -1,331 +1,332 @@ /*- * Copyright (c) 2004 Tim J. Robbins * Copyright (c) 2002 Doug Rabson * Copyright (c) 2000 Marcel Moolenaar * All rights reserved. * Copyright (c) 2013 Dmitry Chagin * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer * in this position and unchanged. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. The name of the author may not be used to endorse or promote products * derived from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include +#include #include #include #include #include #include int linux_execve(struct thread *td, struct linux_execve_args *args) { struct image_args eargs; char *path; int error; LINUX_CTR(execve); if (!LUSECONVPATH(td)) { error = exec_copyin_args(&eargs, args->path, UIO_USERSPACE, args->argp, args->envp); } else { LCONVPATHEXIST(td, args->path, &path); error = exec_copyin_args(&eargs, path, UIO_SYSSPACE, args->argp, args->envp); LFREEPATH(path); } if (error == 0) error = linux_common_execve(td, &eargs); AUDIT_SYSCALL_EXIT(error == EJUSTRETURN ? 0 : error, td); return (error); } int linux_set_upcall(struct thread *td, register_t stack) { if (stack) td->td_frame->tf_rsp = stack; /* * The newly created Linux thread returns * to the user space by the same path that a parent does. */ td->td_frame->tf_rax = 0; return (0); } int linux_mmap2(struct thread *td, struct linux_mmap2_args *args) { return (linux_mmap_common(td, args->addr, args->len, args->prot, args->flags, args->fd, args->pgoff)); } int linux_mprotect(struct thread *td, struct linux_mprotect_args *uap) { return (linux_mprotect_common(td, uap->addr, uap->len, uap->prot)); } int linux_madvise(struct thread *td, struct linux_madvise_args *uap) { return (linux_madvise_common(td, uap->addr, uap->len, uap->behav)); } int linux_iopl(struct thread *td, struct linux_iopl_args *args) { int error; LINUX_CTR(iopl); if (args->level > 3) return (EINVAL); if ((error = priv_check(td, PRIV_IO)) != 0) return (error); if ((error = securelevel_gt(td->td_ucred, 0)) != 0) return (error); td->td_frame->tf_rflags = (td->td_frame->tf_rflags & ~PSL_IOPL) | (args->level * (PSL_IOPL / 3)); return (0); } int linux_pause(struct thread *td, struct linux_pause_args *args) { struct proc *p = td->td_proc; sigset_t sigmask; LINUX_CTR(pause); PROC_LOCK(p); sigmask = td->td_sigmask; PROC_UNLOCK(p); return (kern_sigsuspend(td, sigmask)); } int linux_arch_prctl(struct thread *td, struct linux_arch_prctl_args *args) { unsigned long long cet[3]; struct pcb *pcb; int error; pcb = td->td_pcb; LINUX_CTR2(arch_prctl, "0x%x, %p", args->code, args->addr); switch (args->code) { case LINUX_ARCH_SET_GS: if (args->addr < VM_MAXUSER_ADDRESS) { update_pcb_bases(pcb); pcb->pcb_gsbase = args->addr; td->td_frame->tf_gs = _ugssel; error = 0; } else error = EPERM; break; case LINUX_ARCH_SET_FS: if (args->addr < VM_MAXUSER_ADDRESS) { update_pcb_bases(pcb); pcb->pcb_fsbase = args->addr; td->td_frame->tf_fs = _ufssel; error = 0; } else error = EPERM; break; case LINUX_ARCH_GET_FS: error = copyout(&pcb->pcb_fsbase, PTRIN(args->addr), sizeof(args->addr)); break; case LINUX_ARCH_GET_GS: error = copyout(&pcb->pcb_gsbase, PTRIN(args->addr), sizeof(args->addr)); break; case LINUX_ARCH_CET_STATUS: memset(cet, 0, sizeof(cet)); error = copyout(&cet, PTRIN(args->addr), sizeof(cet)); break; default: linux_msg(td, "unsupported arch_prctl code %#x", args->code); error = EINVAL; } return (error); } int linux_set_cloned_tls(struct thread *td, void *desc) { struct pcb *pcb; if ((uint64_t)desc >= VM_MAXUSER_ADDRESS) return (EPERM); pcb = td->td_pcb; update_pcb_bases(pcb); pcb->pcb_fsbase = (register_t)desc; td->td_frame->tf_fs = _ufssel; return (0); } int futex_xchgl_nosmap(int oparg, uint32_t *uaddr, int *oldval); int futex_xchgl_smap(int oparg, uint32_t *uaddr, int *oldval); DEFINE_IFUNC(, int, futex_xchgl, (int, uint32_t *, int *)) { return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ? futex_xchgl_smap : futex_xchgl_nosmap); } int futex_addl_nosmap(int oparg, uint32_t *uaddr, int *oldval); int futex_addl_smap(int oparg, uint32_t *uaddr, int *oldval); DEFINE_IFUNC(, int, futex_addl, (int, uint32_t *, int *)) { return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ? futex_addl_smap : futex_addl_nosmap); } int futex_orl_nosmap(int oparg, uint32_t *uaddr, int *oldval); int futex_orl_smap(int oparg, uint32_t *uaddr, int *oldval); DEFINE_IFUNC(, int, futex_orl, (int, uint32_t *, int *)) { return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ? futex_orl_smap : futex_orl_nosmap); } int futex_andl_nosmap(int oparg, uint32_t *uaddr, int *oldval); int futex_andl_smap(int oparg, uint32_t *uaddr, int *oldval); DEFINE_IFUNC(, int, futex_andl, (int, uint32_t *, int *)) { return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ? futex_andl_smap : futex_andl_nosmap); } int futex_xorl_nosmap(int oparg, uint32_t *uaddr, int *oldval); int futex_xorl_smap(int oparg, uint32_t *uaddr, int *oldval); DEFINE_IFUNC(, int, futex_xorl, (int, uint32_t *, int *)) { return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ? futex_xorl_smap : futex_xorl_nosmap); } void bsd_to_linux_regset(struct reg *b_reg, struct linux_pt_regset *l_regset) { l_regset->r15 = b_reg->r_r15; l_regset->r14 = b_reg->r_r14; l_regset->r13 = b_reg->r_r13; l_regset->r12 = b_reg->r_r12; l_regset->rbp = b_reg->r_rbp; l_regset->rbx = b_reg->r_rbx; l_regset->r11 = b_reg->r_r11; l_regset->r10 = b_reg->r_r10; l_regset->r9 = b_reg->r_r9; l_regset->r8 = b_reg->r_r8; l_regset->rax = b_reg->r_rax; l_regset->rcx = b_reg->r_rcx; l_regset->rdx = b_reg->r_rdx; l_regset->rsi = b_reg->r_rsi; l_regset->rdi = b_reg->r_rdi; l_regset->orig_rax = b_reg->r_rax; l_regset->rip = b_reg->r_rip; l_regset->cs = b_reg->r_cs; l_regset->eflags = b_reg->r_rflags; l_regset->rsp = b_reg->r_rsp; l_regset->ss = b_reg->r_ss; l_regset->fs_base = 0; l_regset->gs_base = 0; l_regset->ds = b_reg->r_ds; l_regset->es = b_reg->r_es; l_regset->fs = b_reg->r_fs; l_regset->gs = b_reg->r_gs; } diff --git a/sys/amd64/linux/linux_sysvec.c b/sys/amd64/linux/linux_sysvec.c index f9ab19a40ef2..b008b4b49c1c 100644 --- a/sys/amd64/linux/linux_sysvec.c +++ b/sys/amd64/linux/linux_sysvec.c @@ -1,1066 +1,1067 @@ /*- * Copyright (c) 2004 Tim J. Robbins * Copyright (c) 2003 Peter Wemm * Copyright (c) 2002 Doug Rabson * Copyright (c) 1998-1999 Andrew Gallatin * Copyright (c) 1994-1996 Søren Schmidt * All rights reserved. * Copyright (c) 2013, 2021 Dmitry Chagin * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer * in this position and unchanged. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. The name of the author may not be used to endorse or promote products * derived from this software without specific prior written permission * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #define __ELF_WORD_SIZE 64 #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include +#include #include #include #include #include #include #include #include MODULE_VERSION(linux64, 1); #define LINUX_VDSOPAGE_SIZE PAGE_SIZE * 2 #define LINUX_VDSOPAGE_LA48 (VM_MAXUSER_ADDRESS_LA48 - \ LINUX_VDSOPAGE_SIZE) #define LINUX_SHAREDPAGE_LA48 (LINUX_VDSOPAGE_LA48 - PAGE_SIZE) /* * PAGE_SIZE - the size * of the native SHAREDPAGE */ #define LINUX_USRSTACK_LA48 LINUX_SHAREDPAGE_LA48 #define LINUX_PS_STRINGS_LA48 (LINUX_USRSTACK_LA48 - \ sizeof(struct ps_strings)) static int linux_szsigcode; static vm_object_t linux_vdso_obj; static char *linux_vdso_mapping; extern char _binary_linux_vdso_so_o_start; extern char _binary_linux_vdso_so_o_end; static vm_offset_t linux_vdso_base; extern struct sysent linux_sysent[LINUX_SYS_MAXSYSCALL]; SET_DECLARE(linux_ioctl_handler_set, struct linux_ioctl_handler); static int linux_copyout_strings(struct image_params *imgp, uintptr_t *stack_base); static int linux_fixup_elf(uintptr_t *stack_base, struct image_params *iparams); static bool linux_trans_osrel(const Elf_Note *note, int32_t *osrel); static void linux_vdso_install(const void *param); static void linux_vdso_deinstall(const void *param); static void linux_vdso_reloc(char *mapping, Elf_Addr offset); static void linux_set_syscall_retval(struct thread *td, int error); static int linux_fetch_syscall_args(struct thread *td); static void linux_exec_setregs(struct thread *td, struct image_params *imgp, uintptr_t stack); static void linux_exec_sysvec_init(void *param); static int linux_on_exec_vmspace(struct proc *p, struct image_params *imgp); static void linux_set_fork_retval(struct thread *td); static int linux_vsyscall(struct thread *td); #define LINUX_T_UNKNOWN 255 static int _bsd_to_linux_trapcode[] = { LINUX_T_UNKNOWN, /* 0 */ 6, /* 1 T_PRIVINFLT */ LINUX_T_UNKNOWN, /* 2 */ 3, /* 3 T_BPTFLT */ LINUX_T_UNKNOWN, /* 4 */ LINUX_T_UNKNOWN, /* 5 */ 16, /* 6 T_ARITHTRAP */ 254, /* 7 T_ASTFLT */ LINUX_T_UNKNOWN, /* 8 */ 13, /* 9 T_PROTFLT */ 1, /* 10 T_TRCTRAP */ LINUX_T_UNKNOWN, /* 11 */ 14, /* 12 T_PAGEFLT */ LINUX_T_UNKNOWN, /* 13 */ 17, /* 14 T_ALIGNFLT */ LINUX_T_UNKNOWN, /* 15 */ LINUX_T_UNKNOWN, /* 16 */ LINUX_T_UNKNOWN, /* 17 */ 0, /* 18 T_DIVIDE */ 2, /* 19 T_NMI */ 4, /* 20 T_OFLOW */ 5, /* 21 T_BOUND */ 7, /* 22 T_DNA */ 8, /* 23 T_DOUBLEFLT */ 9, /* 24 T_FPOPFLT */ 10, /* 25 T_TSSFLT */ 11, /* 26 T_SEGNPFLT */ 12, /* 27 T_STKFLT */ 18, /* 28 T_MCHK */ 19, /* 29 T_XMMFLT */ 15 /* 30 T_RESERVED */ }; #define bsd_to_linux_trapcode(code) \ ((code)td_proc; frame = td->td_frame; sa = &td->td_sa; sa->args[0] = frame->tf_rdi; sa->args[1] = frame->tf_rsi; sa->args[2] = frame->tf_rdx; sa->args[3] = frame->tf_rcx; sa->args[4] = frame->tf_r8; sa->args[5] = frame->tf_r9; sa->code = frame->tf_rax; if (sa->code >= p->p_sysent->sv_size) /* nosys */ sa->callp = &p->p_sysent->sv_table[p->p_sysent->sv_size - 1]; else sa->callp = &p->p_sysent->sv_table[sa->code]; td->td_retval[0] = 0; return (0); } static void linux_set_syscall_retval(struct thread *td, int error) { struct trapframe *frame; frame = td->td_frame; switch (error) { case 0: frame->tf_rax = td->td_retval[0]; frame->tf_r10 = frame->tf_rcx; break; case ERESTART: /* * Reconstruct pc, we know that 'syscall' is 2 bytes, * lcall $X,y is 7 bytes, int 0x80 is 2 bytes. * We saved this in tf_err. * */ frame->tf_rip -= frame->tf_err; frame->tf_r10 = frame->tf_rcx; break; case EJUSTRETURN: break; default: frame->tf_rax = bsd_to_linux_errno(error); frame->tf_r10 = frame->tf_rcx; break; } /* * Differently from FreeBSD native ABI, on Linux only %rcx * and %r11 values are not preserved across the syscall. * Require full context restore to get all registers except * those two restored at return to usermode. * * XXX: Would be great to be able to avoid PCB_FULL_IRET * for the error == 0 case. */ set_pcb_flags(td->td_pcb, PCB_FULL_IRET); } static void linux_set_fork_retval(struct thread *td) { struct trapframe *frame = td->td_frame; frame->tf_rax = 0; } static int linux_copyout_auxargs(struct image_params *imgp, uintptr_t base) { Elf_Auxargs *args; Elf_Auxinfo *argarray, *pos; struct proc *p; int error, issetugid; p = imgp->proc; args = (Elf64_Auxargs *)imgp->auxargs; argarray = pos = malloc(LINUX_AT_COUNT * sizeof(*pos), M_TEMP, M_WAITOK | M_ZERO); issetugid = p->p_flag & P_SUGID ? 1 : 0; AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO_EHDR, linux_vdso_base); AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, cpu_feature); AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz); AUXARGS_ENTRY(pos, LINUX_AT_CLKTCK, stclohz); AUXARGS_ENTRY(pos, AT_PHDR, args->phdr); AUXARGS_ENTRY(pos, AT_PHENT, args->phent); AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum); AUXARGS_ENTRY(pos, AT_BASE, args->base); AUXARGS_ENTRY(pos, AT_FLAGS, args->flags); AUXARGS_ENTRY(pos, AT_ENTRY, args->entry); AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_ucred->cr_ruid); AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid); AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_ucred->cr_rgid); AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid); AUXARGS_ENTRY(pos, LINUX_AT_SECURE, issetugid); AUXARGS_ENTRY_PTR(pos, LINUX_AT_RANDOM, imgp->canary); AUXARGS_ENTRY(pos, LINUX_AT_HWCAP2, 0); if (imgp->execpathp != 0) AUXARGS_ENTRY_PTR(pos, LINUX_AT_EXECFN, imgp->execpathp); if (args->execfd != -1) AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd); AUXARGS_ENTRY(pos, LINUX_AT_PLATFORM, PTROUT(linux_platform)); AUXARGS_ENTRY(pos, AT_NULL, 0); free(imgp->auxargs, M_TEMP); imgp->auxargs = NULL; KASSERT(pos - argarray <= LINUX_AT_COUNT, ("Too many auxargs")); error = copyout(argarray, (void *)base, sizeof(*argarray) * LINUX_AT_COUNT); free(argarray, M_TEMP); return (error); } static int linux_fixup_elf(uintptr_t *stack_base, struct image_params *imgp) { Elf_Addr *base; base = (Elf64_Addr *)*stack_base; base--; if (suword(base, (uint64_t)imgp->args->argc) == -1) return (EFAULT); *stack_base = (uintptr_t)base; return (0); } /* * Copy strings out to the new process address space, constructing new arg * and env vector tables. Return a pointer to the base so that it can be used * as the initial stack pointer. */ static int linux_copyout_strings(struct image_params *imgp, uintptr_t *stack_base) { int argc, envc, error; char **vectp; char *stringp; uintptr_t destp, ustringp; struct ps_strings *arginfo; char canary[LINUX_AT_RANDOM_LEN]; size_t execpath_len; struct proc *p; p = imgp->proc; arginfo = (struct ps_strings *)PROC_PS_STRINGS(p); destp = (uintptr_t)arginfo; if (imgp->execpath != NULL && imgp->auxargs != NULL) { execpath_len = strlen(imgp->execpath) + 1; destp -= execpath_len; destp = rounddown2(destp, sizeof(void *)); imgp->execpathp = (void *)destp; error = copyout(imgp->execpath, imgp->execpathp, execpath_len); if (error != 0) return (error); } /* Prepare the canary for SSP. */ arc4rand(canary, sizeof(canary), 0); destp -= roundup(sizeof(canary), sizeof(void *)); imgp->canary = (void *)destp; error = copyout(canary, imgp->canary, sizeof(canary)); if (error != 0) return (error); /* Allocate room for the argument and environment strings. */ destp -= ARG_MAX - imgp->args->stringspace; destp = rounddown2(destp, sizeof(void *)); ustringp = destp; if (imgp->auxargs) { /* * Allocate room on the stack for the ELF auxargs * array. It has LINUX_AT_COUNT entries. */ destp -= LINUX_AT_COUNT * sizeof(Elf64_Auxinfo); destp = rounddown2(destp, sizeof(void *)); } vectp = (char **)destp; /* * Allocate room for the argv[] and env vectors including the * terminating NULL pointers. */ vectp -= imgp->args->argc + 1 + imgp->args->envc + 1; /* * Starting with 2.24, glibc depends on a 16-byte stack alignment. * One "long argc" will be prepended later. */ vectp = (char **)((((uintptr_t)vectp + 8) & ~0xF) - 8); /* vectp also becomes our initial stack base. */ *stack_base = (uintptr_t)vectp; stringp = imgp->args->begin_argv; argc = imgp->args->argc; envc = imgp->args->envc; /* Copy out strings - arguments and environment. */ error = copyout(stringp, (void *)ustringp, ARG_MAX - imgp->args->stringspace); if (error != 0) return (error); /* Fill in "ps_strings" struct for ps, w, etc. */ if (suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp) != 0 || suword(&arginfo->ps_nargvstr, argc) != 0) return (EFAULT); /* Fill in argument portion of vector table. */ for (; argc > 0; --argc) { if (suword(vectp++, ustringp) != 0) return (EFAULT); while (*stringp++ != 0) ustringp++; ustringp++; } /* A null vector table pointer separates the argp's from the envp's. */ if (suword(vectp++, 0) != 0) return (EFAULT); if (suword(&arginfo->ps_envstr, (long)(intptr_t)vectp) != 0 || suword(&arginfo->ps_nenvstr, envc) != 0) return (EFAULT); /* Fill in environment portion of vector table. */ for (; envc > 0; --envc) { if (suword(vectp++, ustringp) != 0) return (EFAULT); while (*stringp++ != 0) ustringp++; ustringp++; } /* The end of the vector table is a null pointer. */ if (suword(vectp, 0) != 0) return (EFAULT); if (imgp->auxargs) { vectp++; error = imgp->sysent->sv_copyout_auxargs(imgp, (uintptr_t)vectp); if (error != 0) return (error); } return (0); } /* * Reset registers to default values on exec. */ static void linux_exec_setregs(struct thread *td, struct image_params *imgp, uintptr_t stack) { struct trapframe *regs; struct pcb *pcb; register_t saved_rflags; regs = td->td_frame; pcb = td->td_pcb; if (td->td_proc->p_md.md_ldt != NULL) user_ldt_free(td); pcb->pcb_fsbase = 0; pcb->pcb_gsbase = 0; clear_pcb_flags(pcb, PCB_32BIT); pcb->pcb_initial_fpucw = __LINUX_NPXCW__; set_pcb_flags(pcb, PCB_FULL_IRET); saved_rflags = regs->tf_rflags & PSL_T; bzero((char *)regs, sizeof(struct trapframe)); regs->tf_rip = imgp->entry_addr; regs->tf_rsp = stack; regs->tf_rflags = PSL_USER | saved_rflags; regs->tf_ss = _udatasel; regs->tf_cs = _ucodesel; regs->tf_ds = _udatasel; regs->tf_es = _udatasel; regs->tf_fs = _ufssel; regs->tf_gs = _ugssel; regs->tf_flags = TF_HASSEGS; x86_clear_dbregs(pcb); /* * Drop the FP state if we hold it, so that the process gets a * clean FP state if it uses the FPU again. */ fpstate_drop(td); } /* * Copied from amd64/amd64/machdep.c * * XXX fpu state need? don't think so */ int linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args) { struct proc *p; struct l_ucontext uc; struct l_sigcontext *context; struct trapframe *regs; unsigned long rflags; int error; ksiginfo_t ksi; regs = td->td_frame; error = copyin((void *)regs->tf_rbx, &uc, sizeof(uc)); if (error != 0) return (error); p = td->td_proc; context = &uc.uc_mcontext; rflags = context->sc_rflags; /* * Don't allow users to change privileged or reserved flags. */ /* * XXX do allow users to change the privileged flag PSL_RF. * The cpu sets PSL_RF in tf_rflags for faults. Debuggers * should sometimes set it there too. tf_rflags is kept in * the signal context during signal handling and there is no * other place to remember it, so the PSL_RF bit may be * corrupted by the signal handler without us knowing. * Corruption of the PSL_RF bit at worst causes one more or * one less debugger trap, so allowing it is fairly harmless. */ if (!EFL_SECURE(rflags & ~PSL_RF, regs->tf_rflags & ~PSL_RF)) { uprintf("pid %d comm %s linux mangled rflags %#lx\n", p->p_pid, p->p_comm, rflags); return (EINVAL); } /* * Don't allow users to load a valid privileged %cs. Let the * hardware check for invalid selectors, excess privilege in * other selectors, invalid %eip's and invalid %esp's. */ if (!CS_SECURE(context->sc_cs)) { uprintf("pid %d comm %s linux mangled cs %#x\n", p->p_pid, p->p_comm, context->sc_cs); ksiginfo_init_trap(&ksi); ksi.ksi_signo = SIGBUS; ksi.ksi_code = BUS_OBJERR; ksi.ksi_trapno = T_PROTFLT; ksi.ksi_addr = (void *)regs->tf_rip; trapsignal(td, &ksi); return (EINVAL); } PROC_LOCK(p); linux_to_bsd_sigset(&uc.uc_sigmask, &td->td_sigmask); SIG_CANTMASK(td->td_sigmask); signotify(td); PROC_UNLOCK(p); regs->tf_rdi = context->sc_rdi; regs->tf_rsi = context->sc_rsi; regs->tf_rdx = context->sc_rdx; regs->tf_rbp = context->sc_rbp; regs->tf_rbx = context->sc_rbx; regs->tf_rcx = context->sc_rcx; regs->tf_rax = context->sc_rax; regs->tf_rip = context->sc_rip; regs->tf_rsp = context->sc_rsp; regs->tf_r8 = context->sc_r8; regs->tf_r9 = context->sc_r9; regs->tf_r10 = context->sc_r10; regs->tf_r11 = context->sc_r11; regs->tf_r12 = context->sc_r12; regs->tf_r13 = context->sc_r13; regs->tf_r14 = context->sc_r14; regs->tf_r15 = context->sc_r15; regs->tf_cs = context->sc_cs; regs->tf_err = context->sc_err; regs->tf_rflags = rflags; set_pcb_flags(td->td_pcb, PCB_FULL_IRET); return (EJUSTRETURN); } /* * copied from amd64/amd64/machdep.c * * Send an interrupt to process. */ static void linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) { struct l_rt_sigframe sf, *sfp; struct proc *p; struct thread *td; struct sigacts *psp; caddr_t sp; struct trapframe *regs; int sig, code; int oonstack; td = curthread; p = td->td_proc; PROC_LOCK_ASSERT(p, MA_OWNED); sig = ksi->ksi_signo; psp = p->p_sigacts; code = ksi->ksi_code; mtx_assert(&psp->ps_mtx, MA_OWNED); regs = td->td_frame; oonstack = sigonstack(regs->tf_rsp); LINUX_CTR4(rt_sendsig, "%p, %d, %p, %u", catcher, sig, mask, code); /* Save user context. */ bzero(&sf, sizeof(sf)); bsd_to_linux_sigset(mask, &sf.sf_sc.uc_sigmask); bsd_to_linux_sigset(mask, &sf.sf_sc.uc_mcontext.sc_mask); sf.sf_sc.uc_stack.ss_sp = PTROUT(td->td_sigstk.ss_sp); sf.sf_sc.uc_stack.ss_size = td->td_sigstk.ss_size; sf.sf_sc.uc_stack.ss_flags = (td->td_pflags & TDP_ALTSTACK) ? ((oonstack) ? LINUX_SS_ONSTACK : 0) : LINUX_SS_DISABLE; sf.sf_sc.uc_mcontext.sc_rdi = regs->tf_rdi; sf.sf_sc.uc_mcontext.sc_rsi = regs->tf_rsi; sf.sf_sc.uc_mcontext.sc_rdx = regs->tf_rdx; sf.sf_sc.uc_mcontext.sc_rbp = regs->tf_rbp; sf.sf_sc.uc_mcontext.sc_rbx = regs->tf_rbx; sf.sf_sc.uc_mcontext.sc_rcx = regs->tf_rcx; sf.sf_sc.uc_mcontext.sc_rax = regs->tf_rax; sf.sf_sc.uc_mcontext.sc_rip = regs->tf_rip; sf.sf_sc.uc_mcontext.sc_rsp = regs->tf_rsp; sf.sf_sc.uc_mcontext.sc_r8 = regs->tf_r8; sf.sf_sc.uc_mcontext.sc_r9 = regs->tf_r9; sf.sf_sc.uc_mcontext.sc_r10 = regs->tf_r10; sf.sf_sc.uc_mcontext.sc_r11 = regs->tf_r11; sf.sf_sc.uc_mcontext.sc_r12 = regs->tf_r12; sf.sf_sc.uc_mcontext.sc_r13 = regs->tf_r13; sf.sf_sc.uc_mcontext.sc_r14 = regs->tf_r14; sf.sf_sc.uc_mcontext.sc_r15 = regs->tf_r15; sf.sf_sc.uc_mcontext.sc_cs = regs->tf_cs; sf.sf_sc.uc_mcontext.sc_rflags = regs->tf_rflags; sf.sf_sc.uc_mcontext.sc_err = regs->tf_err; sf.sf_sc.uc_mcontext.sc_trapno = bsd_to_linux_trapcode(code); sf.sf_sc.uc_mcontext.sc_cr2 = (register_t)ksi->ksi_addr; /* Allocate space for the signal handler context. */ if ((td->td_pflags & TDP_ALTSTACK) != 0 && !oonstack && SIGISMEMBER(psp->ps_sigonstack, sig)) { sp = (caddr_t)td->td_sigstk.ss_sp + td->td_sigstk.ss_size - sizeof(struct l_rt_sigframe); } else sp = (caddr_t)regs->tf_rsp - sizeof(struct l_rt_sigframe) - 128; /* Align to 16 bytes. */ sfp = (struct l_rt_sigframe *)((unsigned long)sp & ~0xFul); /* Translate the signal. */ sig = bsd_to_linux_signal(sig); /* Build the argument list for the signal handler. */ regs->tf_rdi = sig; /* arg 1 in %rdi */ regs->tf_rax = 0; regs->tf_rsi = (register_t)&sfp->sf_si; /* arg 2 in %rsi */ regs->tf_rdx = (register_t)&sfp->sf_sc; /* arg 3 in %rdx */ /* Fill in POSIX parts. */ siginfo_to_lsiginfo(&ksi->ksi_info, &sf.sf_si, sig); sf.sf_handler = catcher; mtx_unlock(&psp->ps_mtx); PROC_UNLOCK(p); /* Copy the sigframe out to the user's stack. */ if (copyout(&sf, sfp, sizeof(*sfp)) != 0) { PROC_LOCK(p); sigexit(td, SIGILL); } regs->tf_rsp = (long)sfp; regs->tf_rip = linux_rt_sigcode; regs->tf_rflags &= ~(PSL_T | PSL_D); regs->tf_cs = _ucodesel; set_pcb_flags(td->td_pcb, PCB_FULL_IRET); PROC_LOCK(p); mtx_lock(&psp->ps_mtx); } #define LINUX_VSYSCALL_START (-10UL << 20) #define LINUX_VSYSCALL_SZ 1024 const unsigned long linux_vsyscall_vector[] = { LINUX_SYS_gettimeofday, LINUX_SYS_linux_time, LINUX_SYS_linux_getcpu, }; static int linux_vsyscall(struct thread *td) { struct trapframe *frame; uint64_t retqaddr; int code, traced; int error; frame = td->td_frame; /* Check %rip for vsyscall area. */ if (__predict_true(frame->tf_rip < LINUX_VSYSCALL_START)) return (EINVAL); if ((frame->tf_rip & (LINUX_VSYSCALL_SZ - 1)) != 0) return (EINVAL); code = (frame->tf_rip - LINUX_VSYSCALL_START) / LINUX_VSYSCALL_SZ; if (code >= nitems(linux_vsyscall_vector)) return (EINVAL); /* * vsyscall called as callq *(%rax), so we must * use return address from %rsp and also fixup %rsp. */ error = copyin((void *)frame->tf_rsp, &retqaddr, sizeof(retqaddr)); if (error) return (error); frame->tf_rip = retqaddr; frame->tf_rax = linux_vsyscall_vector[code]; frame->tf_rsp += 8; traced = (frame->tf_flags & PSL_T); amd64_syscall(td, traced); return (0); } struct sysentvec elf_linux_sysvec = { .sv_size = LINUX_SYS_MAXSYSCALL, .sv_table = linux_sysent, .sv_transtrap = linux_translate_traps, .sv_fixup = linux_fixup_elf, .sv_sendsig = linux_rt_sendsig, .sv_sigcode = &_binary_linux_vdso_so_o_start, .sv_szsigcode = &linux_szsigcode, .sv_name = "Linux ELF64", .sv_coredump = elf64_coredump, .sv_elf_core_osabi = ELFOSABI_NONE, .sv_elf_core_abi_vendor = LINUX_ABI_VENDOR, .sv_elf_core_prepare_notes = linux64_prepare_notes, .sv_imgact_try = linux_exec_imgact_try, .sv_minsigstksz = LINUX_MINSIGSTKSZ, .sv_minuser = VM_MIN_ADDRESS, .sv_maxuser = VM_MAXUSER_ADDRESS_LA48, .sv_usrstack = LINUX_USRSTACK_LA48, .sv_psstrings = LINUX_PS_STRINGS_LA48, .sv_psstringssz = sizeof(struct ps_strings), .sv_stackprot = VM_PROT_ALL, .sv_copyout_auxargs = linux_copyout_auxargs, .sv_copyout_strings = linux_copyout_strings, .sv_setregs = linux_exec_setregs, .sv_fixlimit = NULL, .sv_maxssiz = NULL, .sv_flags = SV_ABI_LINUX | SV_LP64 | SV_SHP | SV_SIG_DISCIGN | SV_SIG_WAITNDQ | SV_TIMEKEEP, .sv_set_syscall_retval = linux_set_syscall_retval, .sv_fetch_syscall_args = linux_fetch_syscall_args, .sv_syscallnames = NULL, .sv_shared_page_base = LINUX_SHAREDPAGE_LA48, .sv_shared_page_len = PAGE_SIZE, .sv_schedtail = linux_schedtail, .sv_thread_detach = linux_thread_detach, .sv_trap = linux_vsyscall, .sv_onexec = linux_on_exec_vmspace, .sv_onexit = linux_on_exit, .sv_ontdexit = linux_thread_dtor, .sv_setid_allowed = &linux_setid_allowed_query, .sv_set_fork_retval = linux_set_fork_retval, }; static int linux_on_exec_vmspace(struct proc *p, struct image_params *imgp) { int error; error = linux_map_vdso(p, linux_vdso_obj, linux_vdso_base, LINUX_VDSOPAGE_SIZE, imgp); if (error == 0) linux_on_exec(p, imgp); return (error); } /* * linux_vdso_install() and linux_exec_sysvec_init() must be called * after exec_sysvec_init() which is SI_SUB_EXEC (SI_ORDER_ANY). */ static void linux_exec_sysvec_init(void *param) { l_uintptr_t *ktimekeep_base, *ktsc_selector; struct sysentvec *sv; ptrdiff_t tkoff; sv = param; amd64_lower_shared_page(sv); /* Fill timekeep_base */ exec_sysvec_init(sv); tkoff = kern_timekeep_base - linux_vdso_base; ktimekeep_base = (l_uintptr_t *)(linux_vdso_mapping + tkoff); *ktimekeep_base = sv->sv_timekeep_base; tkoff = kern_tsc_selector - linux_vdso_base; ktsc_selector = (l_uintptr_t *)(linux_vdso_mapping + tkoff); *ktsc_selector = linux_vdso_tsc_selector_idx(); if (bootverbose) printf("Linux x86-64 vDSO tsc_selector: %lu\n", *ktsc_selector); } SYSINIT(elf_linux_exec_sysvec_init, SI_SUB_EXEC + 1, SI_ORDER_ANY, linux_exec_sysvec_init, &elf_linux_sysvec); static void linux_vdso_install(const void *param) { char *vdso_start = &_binary_linux_vdso_so_o_start; char *vdso_end = &_binary_linux_vdso_so_o_end; linux_szsigcode = vdso_end - vdso_start; MPASS(linux_szsigcode <= LINUX_VDSOPAGE_SIZE); linux_vdso_base = LINUX_VDSOPAGE_LA48; if (hw_lower_amd64_sharedpage != 0) linux_vdso_base -= PAGE_SIZE; __elfN(linux_vdso_fixup)(vdso_start, linux_vdso_base); linux_vdso_obj = __elfN(linux_shared_page_init) (&linux_vdso_mapping, LINUX_VDSOPAGE_SIZE); bcopy(vdso_start, linux_vdso_mapping, linux_szsigcode); linux_vdso_reloc(linux_vdso_mapping, linux_vdso_base); } SYSINIT(elf_linux_vdso_init, SI_SUB_EXEC + 1, SI_ORDER_FIRST, linux_vdso_install, NULL); static void linux_vdso_deinstall(const void *param) { __elfN(linux_shared_page_fini)(linux_vdso_obj, linux_vdso_mapping, LINUX_VDSOPAGE_SIZE); } SYSUNINIT(elf_linux_vdso_uninit, SI_SUB_EXEC, SI_ORDER_FIRST, linux_vdso_deinstall, NULL); static void linux_vdso_reloc(char *mapping, Elf_Addr offset) { const Elf_Ehdr *ehdr; const Elf_Shdr *shdr; Elf64_Addr *where, val; Elf_Size rtype, symidx; const Elf_Rela *rela; Elf_Addr addr, addend; int relacnt; int i, j; MPASS(offset != 0); relacnt = 0; ehdr = (const Elf_Ehdr *)mapping; shdr = (const Elf_Shdr *)(mapping + ehdr->e_shoff); for (i = 0; i < ehdr->e_shnum; i++) { switch (shdr[i].sh_type) { case SHT_REL: printf("Linux x86_64 vDSO: unexpected Rel section\n"); break; case SHT_RELA: rela = (const Elf_Rela *)(mapping + shdr[i].sh_offset); relacnt = shdr[i].sh_size / sizeof(*rela); } } for (j = 0; j < relacnt; j++, rela++) { where = (Elf_Addr *)(mapping + rela->r_offset); addend = rela->r_addend; rtype = ELF_R_TYPE(rela->r_info); symidx = ELF_R_SYM(rela->r_info); switch (rtype) { case R_X86_64_NONE: /* none */ break; case R_X86_64_RELATIVE: /* B + A */ addr = (Elf_Addr)(offset + addend); val = addr; if (*where != val) *where = val; break; case R_X86_64_IRELATIVE: printf("Linux x86_64 vDSO: unexpected ifunc relocation, " "symbol index %ld\n", symidx); break; default: printf("Linux x86_64 vDSO: unexpected relocation type %ld, " "symbol index %ld\n", rtype, symidx); } } } static char GNULINUX_ABI_VENDOR[] = "GNU"; static int GNULINUX_ABI_DESC = 0; static bool linux_trans_osrel(const Elf_Note *note, int32_t *osrel) { const Elf32_Word *desc; uintptr_t p; p = (uintptr_t)(note + 1); p += roundup2(note->n_namesz, sizeof(Elf32_Addr)); desc = (const Elf32_Word *)p; if (desc[0] != GNULINUX_ABI_DESC) return (false); /* * For Linux we encode osrel using the Linux convention of * (version << 16) | (major << 8) | (minor) * See macro in linux_mib.h */ *osrel = LINUX_KERNVER(desc[1], desc[2], desc[3]); return (true); } static Elf_Brandnote linux64_brandnote = { .hdr.n_namesz = sizeof(GNULINUX_ABI_VENDOR), .hdr.n_descsz = 16, .hdr.n_type = 1, .vendor = GNULINUX_ABI_VENDOR, .flags = BN_TRANSLATE_OSREL, .trans_osrel = linux_trans_osrel }; static Elf64_Brandinfo linux_glibc2brand = { .brand = ELFOSABI_LINUX, .machine = EM_X86_64, .compat_3_brand = "Linux", .emul_path = linux_emul_path, .interp_path = "/lib64/ld-linux-x86-64.so.2", .sysvec = &elf_linux_sysvec, .interp_newpath = NULL, .brand_note = &linux64_brandnote, .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE }; static Elf64_Brandinfo linux_glibc2brandshort = { .brand = ELFOSABI_LINUX, .machine = EM_X86_64, .compat_3_brand = "Linux", .emul_path = linux_emul_path, .interp_path = "/lib64/ld-linux.so.2", .sysvec = &elf_linux_sysvec, .interp_newpath = NULL, .brand_note = &linux64_brandnote, .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE }; static Elf64_Brandinfo linux_muslbrand = { .brand = ELFOSABI_LINUX, .machine = EM_X86_64, .compat_3_brand = "Linux", .emul_path = linux_emul_path, .interp_path = "/lib/ld-musl-x86_64.so.1", .sysvec = &elf_linux_sysvec, .interp_newpath = NULL, .brand_note = &linux64_brandnote, .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE | LINUX_BI_FUTEX_REQUEUE }; Elf64_Brandinfo *linux_brandlist[] = { &linux_glibc2brand, &linux_glibc2brandshort, &linux_muslbrand, NULL }; static int linux64_elf_modevent(module_t mod, int type, void *data) { Elf64_Brandinfo **brandinfo; int error; struct linux_ioctl_handler **lihp; error = 0; switch(type) { case MOD_LOAD: for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; ++brandinfo) if (elf64_insert_brand_entry(*brandinfo) < 0) error = EINVAL; if (error == 0) { SET_FOREACH(lihp, linux_ioctl_handler_set) linux_ioctl_register_handler(*lihp); stclohz = (stathz ? stathz : hz); if (bootverbose) printf("Linux x86-64 ELF exec handler installed\n"); } else printf("cannot insert Linux x86-64 ELF brand handler\n"); break; case MOD_UNLOAD: for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; ++brandinfo) if (elf64_brand_inuse(*brandinfo)) error = EBUSY; if (error == 0) { for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; ++brandinfo) if (elf64_remove_brand_entry(*brandinfo) < 0) error = EINVAL; } if (error == 0) { SET_FOREACH(lihp, linux_ioctl_handler_set) linux_ioctl_unregister_handler(*lihp); if (bootverbose) printf("Linux x86_64 ELF exec handler removed\n"); } else printf("Could not deinstall Linux x86_64 ELF interpreter entry\n"); break; default: return (EOPNOTSUPP); } return (error); } static moduledata_t linux64_elf_mod = { "linux64elf", linux64_elf_modevent, 0 }; DECLARE_MODULE_TIED(linux64elf, linux64_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY); MODULE_DEPEND(linux64elf, linux_common, 1, 1, 1); FEATURE(linux64, "Linux 64bit support"); diff --git a/sys/amd64/linux32/linux32_machdep.c b/sys/amd64/linux32/linux32_machdep.c index 926c69c9ebac..5801fac7f126 100644 --- a/sys/amd64/linux32/linux32_machdep.c +++ b/sys/amd64/linux32/linux32_machdep.c @@ -1,745 +1,746 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 2004 Tim J. Robbins * Copyright (c) 2002 Doug Rabson * Copyright (c) 2000 Marcel Moolenaar * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer * in this position and unchanged. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. The name of the author may not be used to endorse or promote products * derived from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include "opt_compat.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include +#include #include #include #include #include #include static void bsd_to_linux_rusage(struct rusage *ru, struct l_rusage *lru); struct l_old_select_argv { l_int nfds; l_uintptr_t readfds; l_uintptr_t writefds; l_uintptr_t exceptfds; l_uintptr_t timeout; } __packed; static void bsd_to_linux_rusage(struct rusage *ru, struct l_rusage *lru) { lru->ru_utime.tv_sec = ru->ru_utime.tv_sec; lru->ru_utime.tv_usec = ru->ru_utime.tv_usec; lru->ru_stime.tv_sec = ru->ru_stime.tv_sec; lru->ru_stime.tv_usec = ru->ru_stime.tv_usec; lru->ru_maxrss = ru->ru_maxrss; lru->ru_ixrss = ru->ru_ixrss; lru->ru_idrss = ru->ru_idrss; lru->ru_isrss = ru->ru_isrss; lru->ru_minflt = ru->ru_minflt; lru->ru_majflt = ru->ru_majflt; lru->ru_nswap = ru->ru_nswap; lru->ru_inblock = ru->ru_inblock; lru->ru_oublock = ru->ru_oublock; lru->ru_msgsnd = ru->ru_msgsnd; lru->ru_msgrcv = ru->ru_msgrcv; lru->ru_nsignals = ru->ru_nsignals; lru->ru_nvcsw = ru->ru_nvcsw; lru->ru_nivcsw = ru->ru_nivcsw; } int linux_copyout_rusage(struct rusage *ru, void *uaddr) { struct l_rusage lru; bsd_to_linux_rusage(ru, &lru); return (copyout(&lru, uaddr, sizeof(struct l_rusage))); } int linux_execve(struct thread *td, struct linux_execve_args *args) { struct image_args eargs; char *path; int error; LCONVPATHEXIST(td, args->path, &path); error = freebsd32_exec_copyin_args(&eargs, path, UIO_SYSSPACE, args->argp, args->envp); free(path, M_TEMP); if (error == 0) error = linux_common_execve(td, &eargs); AUDIT_SYSCALL_EXIT(error == EJUSTRETURN ? 0 : error, td); return (error); } CTASSERT(sizeof(struct l_iovec32) == 8); int linux32_copyinuio(struct l_iovec32 *iovp, l_ulong iovcnt, struct uio **uiop) { struct l_iovec32 iov32; struct iovec *iov; struct uio *uio; uint32_t iovlen; int error, i; *uiop = NULL; if (iovcnt > UIO_MAXIOV) return (EINVAL); iovlen = iovcnt * sizeof(struct iovec); uio = malloc(iovlen + sizeof(*uio), M_IOV, M_WAITOK); iov = (struct iovec *)(uio + 1); for (i = 0; i < iovcnt; i++) { error = copyin(&iovp[i], &iov32, sizeof(struct l_iovec32)); if (error) { free(uio, M_IOV); return (error); } iov[i].iov_base = PTRIN(iov32.iov_base); iov[i].iov_len = iov32.iov_len; } uio->uio_iov = iov; uio->uio_iovcnt = iovcnt; uio->uio_segflg = UIO_USERSPACE; uio->uio_offset = -1; uio->uio_resid = 0; for (i = 0; i < iovcnt; i++) { if (iov->iov_len > INT_MAX - uio->uio_resid) { free(uio, M_IOV); return (EINVAL); } uio->uio_resid += iov->iov_len; iov++; } *uiop = uio; return (0); } int linux32_copyiniov(struct l_iovec32 *iovp32, l_ulong iovcnt, struct iovec **iovp, int error) { struct l_iovec32 iov32; struct iovec *iov; uint32_t iovlen; int i; *iovp = NULL; if (iovcnt > UIO_MAXIOV) return (error); iovlen = iovcnt * sizeof(struct iovec); iov = malloc(iovlen, M_IOV, M_WAITOK); for (i = 0; i < iovcnt; i++) { error = copyin(&iovp32[i], &iov32, sizeof(struct l_iovec32)); if (error) { free(iov, M_IOV); return (error); } iov[i].iov_base = PTRIN(iov32.iov_base); iov[i].iov_len = iov32.iov_len; } *iovp = iov; return(0); } int linux_readv(struct thread *td, struct linux_readv_args *uap) { struct uio *auio; int error; error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio); if (error) return (error); error = kern_readv(td, uap->fd, auio); free(auio, M_IOV); return (error); } int linux_writev(struct thread *td, struct linux_writev_args *uap) { struct uio *auio; int error; error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio); if (error) return (error); error = kern_writev(td, uap->fd, auio); free(auio, M_IOV); return (error); } struct l_ipc_kludge { l_uintptr_t msgp; l_long msgtyp; } __packed; int linux_ipc(struct thread *td, struct linux_ipc_args *args) { switch (args->what & 0xFFFF) { case LINUX_SEMOP: { struct linux_semop_args a; a.semid = args->arg1; a.tsops = PTRIN(args->ptr); a.nsops = args->arg2; return (linux_semop(td, &a)); } case LINUX_SEMGET: { struct linux_semget_args a; a.key = args->arg1; a.nsems = args->arg2; a.semflg = args->arg3; return (linux_semget(td, &a)); } case LINUX_SEMCTL: { struct linux_semctl_args a; int error; a.semid = args->arg1; a.semnum = args->arg2; a.cmd = args->arg3; error = copyin(PTRIN(args->ptr), &a.arg, sizeof(a.arg)); if (error) return (error); return (linux_semctl(td, &a)); } case LINUX_MSGSND: { struct linux_msgsnd_args a; a.msqid = args->arg1; a.msgp = PTRIN(args->ptr); a.msgsz = args->arg2; a.msgflg = args->arg3; return (linux_msgsnd(td, &a)); } case LINUX_MSGRCV: { struct linux_msgrcv_args a; a.msqid = args->arg1; a.msgsz = args->arg2; a.msgflg = args->arg3; if ((args->what >> 16) == 0) { struct l_ipc_kludge tmp; int error; if (args->ptr == 0) return (EINVAL); error = copyin(PTRIN(args->ptr), &tmp, sizeof(tmp)); if (error) return (error); a.msgp = PTRIN(tmp.msgp); a.msgtyp = tmp.msgtyp; } else { a.msgp = PTRIN(args->ptr); a.msgtyp = args->arg5; } return (linux_msgrcv(td, &a)); } case LINUX_MSGGET: { struct linux_msgget_args a; a.key = args->arg1; a.msgflg = args->arg2; return (linux_msgget(td, &a)); } case LINUX_MSGCTL: { struct linux_msgctl_args a; a.msqid = args->arg1; a.cmd = args->arg2; a.buf = PTRIN(args->ptr); return (linux_msgctl(td, &a)); } case LINUX_SHMAT: { struct linux_shmat_args a; l_uintptr_t addr; int error; a.shmid = args->arg1; a.shmaddr = PTRIN(args->ptr); a.shmflg = args->arg2; error = linux_shmat(td, &a); if (error != 0) return (error); addr = td->td_retval[0]; error = copyout(&addr, PTRIN(args->arg3), sizeof(addr)); td->td_retval[0] = 0; return (error); } case LINUX_SHMDT: { struct linux_shmdt_args a; a.shmaddr = PTRIN(args->ptr); return (linux_shmdt(td, &a)); } case LINUX_SHMGET: { struct linux_shmget_args a; a.key = args->arg1; a.size = args->arg2; a.shmflg = args->arg3; return (linux_shmget(td, &a)); } case LINUX_SHMCTL: { struct linux_shmctl_args a; a.shmid = args->arg1; a.cmd = args->arg2; a.buf = PTRIN(args->ptr); return (linux_shmctl(td, &a)); } default: break; } return (EINVAL); } int linux_old_select(struct thread *td, struct linux_old_select_args *args) { struct l_old_select_argv linux_args; struct linux_select_args newsel; int error; error = copyin(args->ptr, &linux_args, sizeof(linux_args)); if (error) return (error); newsel.nfds = linux_args.nfds; newsel.readfds = PTRIN(linux_args.readfds); newsel.writefds = PTRIN(linux_args.writefds); newsel.exceptfds = PTRIN(linux_args.exceptfds); newsel.timeout = PTRIN(linux_args.timeout); return (linux_select(td, &newsel)); } int linux_set_cloned_tls(struct thread *td, void *desc) { struct l_user_desc info; struct pcb *pcb; int error; error = copyin(desc, &info, sizeof(struct l_user_desc)); if (error) { linux_msg(td, "set_cloned_tls copyin info failed!"); } else { /* We might copy out the entry_number as GUGS32_SEL. */ info.entry_number = GUGS32_SEL; error = copyout(&info, desc, sizeof(struct l_user_desc)); if (error) linux_msg(td, "set_cloned_tls copyout info failed!"); pcb = td->td_pcb; update_pcb_bases(pcb); pcb->pcb_gsbase = (register_t)info.base_addr; td->td_frame->tf_gs = GSEL(GUGS32_SEL, SEL_UPL); } return (error); } int linux_set_upcall(struct thread *td, register_t stack) { if (stack) td->td_frame->tf_rsp = stack; /* * The newly created Linux thread returns * to the user space by the same path that a parent do. */ td->td_frame->tf_rax = 0; return (0); } int linux_mmap2(struct thread *td, struct linux_mmap2_args *args) { return (linux_mmap_common(td, PTROUT(args->addr), args->len, args->prot, args->flags, args->fd, (uint64_t)(uint32_t)args->pgoff * PAGE_SIZE)); } int linux_mmap(struct thread *td, struct linux_mmap_args *args) { int error; struct l_mmap_argv linux_args; error = copyin(args->ptr, &linux_args, sizeof(linux_args)); if (error) return (error); return (linux_mmap_common(td, linux_args.addr, linux_args.len, linux_args.prot, linux_args.flags, linux_args.fd, (uint32_t)linux_args.pgoff)); } int linux_mprotect(struct thread *td, struct linux_mprotect_args *uap) { return (linux_mprotect_common(td, PTROUT(uap->addr), uap->len, uap->prot)); } int linux_madvise(struct thread *td, struct linux_madvise_args *uap) { return (linux_madvise_common(td, PTROUT(uap->addr), uap->len, uap->behav)); } int linux_iopl(struct thread *td, struct linux_iopl_args *args) { int error; if (args->level < 0 || args->level > 3) return (EINVAL); if ((error = priv_check(td, PRIV_IO)) != 0) return (error); if ((error = securelevel_gt(td->td_ucred, 0)) != 0) return (error); td->td_frame->tf_rflags = (td->td_frame->tf_rflags & ~PSL_IOPL) | (args->level * (PSL_IOPL / 3)); return (0); } int linux_sigaction(struct thread *td, struct linux_sigaction_args *args) { l_osigaction_t osa; l_sigaction_t act, oact; int error; if (args->nsa != NULL) { error = copyin(args->nsa, &osa, sizeof(l_osigaction_t)); if (error) return (error); act.lsa_handler = osa.lsa_handler; act.lsa_flags = osa.lsa_flags; act.lsa_restorer = osa.lsa_restorer; LINUX_SIGEMPTYSET(act.lsa_mask); act.lsa_mask.__mask = osa.lsa_mask; } error = linux_do_sigaction(td, args->sig, args->nsa ? &act : NULL, args->osa ? &oact : NULL); if (args->osa != NULL && !error) { osa.lsa_handler = oact.lsa_handler; osa.lsa_flags = oact.lsa_flags; osa.lsa_restorer = oact.lsa_restorer; osa.lsa_mask = oact.lsa_mask.__mask; error = copyout(&osa, args->osa, sizeof(l_osigaction_t)); } return (error); } /* * Linux has two extra args, restart and oldmask. We don't use these, * but it seems that "restart" is actually a context pointer that * enables the signal to happen with a different register set. */ int linux_sigsuspend(struct thread *td, struct linux_sigsuspend_args *args) { sigset_t sigmask; l_sigset_t mask; LINUX_SIGEMPTYSET(mask); mask.__mask = args->mask; linux_to_bsd_sigset(&mask, &sigmask); return (kern_sigsuspend(td, sigmask)); } int linux_pause(struct thread *td, struct linux_pause_args *args) { struct proc *p = td->td_proc; sigset_t sigmask; PROC_LOCK(p); sigmask = td->td_sigmask; PROC_UNLOCK(p); return (kern_sigsuspend(td, sigmask)); } int linux_gettimeofday(struct thread *td, struct linux_gettimeofday_args *uap) { struct timeval atv; l_timeval atv32; struct timezone rtz; int error = 0; if (uap->tp) { microtime(&atv); atv32.tv_sec = atv.tv_sec; atv32.tv_usec = atv.tv_usec; error = copyout(&atv32, uap->tp, sizeof(atv32)); } if (error == 0 && uap->tzp != NULL) { rtz.tz_minuteswest = 0; rtz.tz_dsttime = 0; error = copyout(&rtz, uap->tzp, sizeof(rtz)); } return (error); } int linux_settimeofday(struct thread *td, struct linux_settimeofday_args *uap) { l_timeval atv32; struct timeval atv, *tvp; struct timezone atz, *tzp; int error; if (uap->tp) { error = copyin(uap->tp, &atv32, sizeof(atv32)); if (error) return (error); atv.tv_sec = atv32.tv_sec; atv.tv_usec = atv32.tv_usec; tvp = &atv; } else tvp = NULL; if (uap->tzp) { error = copyin(uap->tzp, &atz, sizeof(atz)); if (error) return (error); tzp = &atz; } else tzp = NULL; return (kern_settimeofday(td, tvp, tzp)); } int linux_getrusage(struct thread *td, struct linux_getrusage_args *uap) { struct rusage s; int error; error = kern_getrusage(td, uap->who, &s); if (error != 0) return (error); if (uap->rusage != NULL) error = linux_copyout_rusage(&s, uap->rusage); return (error); } int linux_set_thread_area(struct thread *td, struct linux_set_thread_area_args *args) { struct l_user_desc info; struct pcb *pcb; int error; error = copyin(args->desc, &info, sizeof(struct l_user_desc)); if (error) return (error); /* * Semantics of Linux version: every thread in the system has array * of three TLS descriptors. 1st is GLIBC TLS, 2nd is WINE, 3rd unknown. * This syscall loads one of the selected TLS decriptors with a value * and also loads GDT descriptors 6, 7 and 8 with the content of * the per-thread descriptors. * * Semantics of FreeBSD version: I think we can ignore that Linux has * three per-thread descriptors and use just the first one. * The tls_array[] is used only in [gs]et_thread_area() syscalls and * for loading the GDT descriptors. We use just one GDT descriptor * for TLS, so we will load just one. * * XXX: This doesn't work when a user space process tries to use more * than one TLS segment. Comment in the Linux source says wine might * do this. */ /* * GLIBC reads current %gs and call set_thread_area() with it. * We should let GUDATA_SEL and GUGS32_SEL proceed as well because * we use these segments. */ switch (info.entry_number) { case GUGS32_SEL: case GUDATA_SEL: case 6: case -1: info.entry_number = GUGS32_SEL; break; default: return (EINVAL); } /* * We have to copy out the GDT entry we use. * * XXX: What if a user space program does not check the return value * and tries to use 6, 7 or 8? */ error = copyout(&info, args->desc, sizeof(struct l_user_desc)); if (error) return (error); pcb = td->td_pcb; update_pcb_bases(pcb); pcb->pcb_gsbase = (register_t)info.base_addr; update_gdt_gsbase(td, info.base_addr); return (0); } void bsd_to_linux_regset32(struct reg32 *b_reg, struct linux_pt_regset32 *l_regset) { l_regset->ebx = b_reg->r_ebx; l_regset->ecx = b_reg->r_ecx; l_regset->edx = b_reg->r_edx; l_regset->esi = b_reg->r_esi; l_regset->edi = b_reg->r_edi; l_regset->ebp = b_reg->r_ebp; l_regset->eax = b_reg->r_eax; l_regset->ds = b_reg->r_ds; l_regset->es = b_reg->r_es; l_regset->fs = b_reg->r_fs; l_regset->gs = b_reg->r_gs; l_regset->orig_eax = b_reg->r_eax; l_regset->eip = b_reg->r_eip; l_regset->cs = b_reg->r_cs; l_regset->eflags = b_reg->r_eflags; l_regset->esp = b_reg->r_esp; l_regset->ss = b_reg->r_ss; } int futex_xchgl_nosmap(int oparg, uint32_t *uaddr, int *oldval); int futex_xchgl_smap(int oparg, uint32_t *uaddr, int *oldval); DEFINE_IFUNC(, int, futex_xchgl, (int, uint32_t *, int *)) { return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ? futex_xchgl_smap : futex_xchgl_nosmap); } int futex_addl_nosmap(int oparg, uint32_t *uaddr, int *oldval); int futex_addl_smap(int oparg, uint32_t *uaddr, int *oldval); DEFINE_IFUNC(, int, futex_addl, (int, uint32_t *, int *)) { return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ? futex_addl_smap : futex_addl_nosmap); } int futex_orl_nosmap(int oparg, uint32_t *uaddr, int *oldval); int futex_orl_smap(int oparg, uint32_t *uaddr, int *oldval); DEFINE_IFUNC(, int, futex_orl, (int, uint32_t *, int *)) { return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ? futex_orl_smap : futex_orl_nosmap); } int futex_andl_nosmap(int oparg, uint32_t *uaddr, int *oldval); int futex_andl_smap(int oparg, uint32_t *uaddr, int *oldval); DEFINE_IFUNC(, int, futex_andl, (int, uint32_t *, int *)) { return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ? futex_andl_smap : futex_andl_nosmap); } int futex_xorl_nosmap(int oparg, uint32_t *uaddr, int *oldval); int futex_xorl_smap(int oparg, uint32_t *uaddr, int *oldval); DEFINE_IFUNC(, int, futex_xorl, (int, uint32_t *, int *)) { return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ? futex_xorl_smap : futex_xorl_nosmap); } diff --git a/sys/amd64/linux32/linux32_sysvec.c b/sys/amd64/linux32/linux32_sysvec.c index 42b40ee3094a..7905b8587192 100644 --- a/sys/amd64/linux32/linux32_sysvec.c +++ b/sys/amd64/linux32/linux32_sysvec.c @@ -1,1226 +1,1227 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 2004 Tim J. Robbins * Copyright (c) 2003 Peter Wemm * Copyright (c) 2002 Doug Rabson * Copyright (c) 1998-1999 Andrew Gallatin * Copyright (c) 1994-1996 Søren Schmidt * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer * in this position and unchanged. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. The name of the author may not be used to endorse or promote products * derived from this software without specific prior written permission * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include "opt_compat.h" #include __FBSDID("$FreeBSD$"); #ifndef COMPAT_FREEBSD32 #error "Unable to compile Linux-emulator due to missing COMPAT_FREEBSD32 option!" #endif #define __ELF_WORD_SIZE 32 #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include +#include #include #include #include #include #include #include MODULE_VERSION(linux, 1); #define LINUX32_MAXUSER ((1ul << 32) - PAGE_SIZE) #define LINUX32_VDSOPAGE_SIZE PAGE_SIZE * 2 #define LINUX32_VDSOPAGE (LINUX32_MAXUSER - LINUX32_VDSOPAGE_SIZE) #define LINUX32_SHAREDPAGE (LINUX32_VDSOPAGE - PAGE_SIZE) /* * PAGE_SIZE - the size * of the native SHAREDPAGE */ #define LINUX32_USRSTACK LINUX32_SHAREDPAGE static int linux_szsigcode; static vm_object_t linux_vdso_obj; static char *linux_vdso_mapping; extern char _binary_linux32_vdso_so_o_start; extern char _binary_linux32_vdso_so_o_end; static vm_offset_t linux_vdso_base; extern struct sysent linux32_sysent[LINUX32_SYS_MAXSYSCALL]; SET_DECLARE(linux_ioctl_handler_set, struct linux_ioctl_handler); static int linux_fixup_elf(uintptr_t *stack_base, struct image_params *iparams); static int linux_copyout_strings(struct image_params *imgp, uintptr_t *stack_base); static void linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask); static void linux_exec_setregs(struct thread *td, struct image_params *imgp, uintptr_t stack); static void linux_exec_sysvec_init(void *param); static int linux_on_exec_vmspace(struct proc *p, struct image_params *imgp); static void linux32_fixlimit(struct rlimit *rl, int which); static bool linux32_trans_osrel(const Elf_Note *note, int32_t *osrel); static void linux_vdso_install(const void *param); static void linux_vdso_deinstall(const void *param); static void linux_vdso_reloc(char *mapping, Elf_Addr offset); static void linux32_set_fork_retval(struct thread *td); static void linux32_set_syscall_retval(struct thread *td, int error); #define LINUX_T_UNKNOWN 255 static int _bsd_to_linux_trapcode[] = { LINUX_T_UNKNOWN, /* 0 */ 6, /* 1 T_PRIVINFLT */ LINUX_T_UNKNOWN, /* 2 */ 3, /* 3 T_BPTFLT */ LINUX_T_UNKNOWN, /* 4 */ LINUX_T_UNKNOWN, /* 5 */ 16, /* 6 T_ARITHTRAP */ 254, /* 7 T_ASTFLT */ LINUX_T_UNKNOWN, /* 8 */ 13, /* 9 T_PROTFLT */ 1, /* 10 T_TRCTRAP */ LINUX_T_UNKNOWN, /* 11 */ 14, /* 12 T_PAGEFLT */ LINUX_T_UNKNOWN, /* 13 */ 17, /* 14 T_ALIGNFLT */ LINUX_T_UNKNOWN, /* 15 */ LINUX_T_UNKNOWN, /* 16 */ LINUX_T_UNKNOWN, /* 17 */ 0, /* 18 T_DIVIDE */ 2, /* 19 T_NMI */ 4, /* 20 T_OFLOW */ 5, /* 21 T_BOUND */ 7, /* 22 T_DNA */ 8, /* 23 T_DOUBLEFLT */ 9, /* 24 T_FPOPFLT */ 10, /* 25 T_TSSFLT */ 11, /* 26 T_SEGNPFLT */ 12, /* 27 T_STKFLT */ 18, /* 28 T_MCHK */ 19, /* 29 T_XMMFLT */ 15 /* 30 T_RESERVED */ }; #define bsd_to_linux_trapcode(code) \ ((code)auxargs; argarray = pos = malloc(LINUX_AT_COUNT * sizeof(*pos), M_TEMP, M_WAITOK | M_ZERO); issetugid = imgp->proc->p_flag & P_SUGID ? 1 : 0; AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO, __kernel_vsyscall); AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO_EHDR, linux_vdso_base); AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, cpu_feature); AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz); /* * Do not export AT_CLKTCK when emulating Linux kernel prior to 2.4.0, * as it has appeared in the 2.4.0-rc7 first time. * Being exported, AT_CLKTCK is returned by sysconf(_SC_CLK_TCK), * glibc falls back to the hard-coded CLK_TCK value when aux entry * is not present. * Also see linux_times() implementation. */ if (linux_kernver(curthread) >= LINUX_KERNVER_2004000) AUXARGS_ENTRY(pos, LINUX_AT_CLKTCK, stclohz); AUXARGS_ENTRY(pos, AT_PHDR, args->phdr); AUXARGS_ENTRY(pos, AT_PHENT, args->phent); AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum); AUXARGS_ENTRY(pos, AT_BASE, args->base); AUXARGS_ENTRY(pos, AT_FLAGS, args->flags); AUXARGS_ENTRY(pos, AT_ENTRY, args->entry); AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_ucred->cr_ruid); AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid); AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_ucred->cr_rgid); AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid); AUXARGS_ENTRY(pos, LINUX_AT_SECURE, issetugid); AUXARGS_ENTRY(pos, LINUX_AT_RANDOM, PTROUT(imgp->canary)); AUXARGS_ENTRY(pos, LINUX_AT_HWCAP2, 0); if (imgp->execpathp != 0) AUXARGS_ENTRY(pos, LINUX_AT_EXECFN, PTROUT(imgp->execpathp)); if (args->execfd != -1) AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd); AUXARGS_ENTRY(pos, LINUX_AT_PLATFORM, PTROUT(linux_platform)); AUXARGS_ENTRY(pos, AT_NULL, 0); free(imgp->auxargs, M_TEMP); imgp->auxargs = NULL; KASSERT(pos - argarray <= LINUX_AT_COUNT, ("Too many auxargs")); error = copyout(argarray, (void *)base, sizeof(*argarray) * LINUX_AT_COUNT); free(argarray, M_TEMP); return (error); } static int linux_fixup_elf(uintptr_t *stack_base, struct image_params *imgp) { Elf32_Addr *base; base = (Elf32_Addr *)*stack_base; base--; if (suword32(base, (uint32_t)imgp->args->argc) == -1) return (EFAULT); *stack_base = (uintptr_t)base; return (0); } static void linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) { struct thread *td = curthread; struct proc *p = td->td_proc; struct sigacts *psp; struct trapframe *regs; struct l_rt_sigframe *fp, frame; int oonstack; int sig; int code; sig = ksi->ksi_signo; code = ksi->ksi_code; PROC_LOCK_ASSERT(p, MA_OWNED); psp = p->p_sigacts; mtx_assert(&psp->ps_mtx, MA_OWNED); regs = td->td_frame; oonstack = sigonstack(regs->tf_rsp); /* Allocate space for the signal handler context. */ if ((td->td_pflags & TDP_ALTSTACK) && !oonstack && SIGISMEMBER(psp->ps_sigonstack, sig)) { fp = (struct l_rt_sigframe *)((uintptr_t)td->td_sigstk.ss_sp + td->td_sigstk.ss_size - sizeof(struct l_rt_sigframe)); } else fp = (struct l_rt_sigframe *)regs->tf_rsp - 1; mtx_unlock(&psp->ps_mtx); /* Build the argument list for the signal handler. */ sig = bsd_to_linux_signal(sig); bzero(&frame, sizeof(frame)); frame.sf_handler = PTROUT(catcher); frame.sf_sig = sig; frame.sf_siginfo = PTROUT(&fp->sf_si); frame.sf_ucontext = PTROUT(&fp->sf_sc); /* Fill in POSIX parts. */ siginfo_to_lsiginfo(&ksi->ksi_info, &frame.sf_si, sig); /* * Build the signal context to be used by sigreturn and libgcc unwind. */ frame.sf_sc.uc_flags = 0; /* XXX ??? */ frame.sf_sc.uc_link = 0; /* XXX ??? */ frame.sf_sc.uc_stack.ss_sp = PTROUT(td->td_sigstk.ss_sp); frame.sf_sc.uc_stack.ss_size = td->td_sigstk.ss_size; frame.sf_sc.uc_stack.ss_flags = (td->td_pflags & TDP_ALTSTACK) ? ((oonstack) ? LINUX_SS_ONSTACK : 0) : LINUX_SS_DISABLE; PROC_UNLOCK(p); bsd_to_linux_sigset(mask, &frame.sf_sc.uc_sigmask); frame.sf_sc.uc_mcontext.sc_mask = frame.sf_sc.uc_sigmask.__mask; frame.sf_sc.uc_mcontext.sc_edi = regs->tf_rdi; frame.sf_sc.uc_mcontext.sc_esi = regs->tf_rsi; frame.sf_sc.uc_mcontext.sc_ebp = regs->tf_rbp; frame.sf_sc.uc_mcontext.sc_ebx = regs->tf_rbx; frame.sf_sc.uc_mcontext.sc_esp = regs->tf_rsp; frame.sf_sc.uc_mcontext.sc_edx = regs->tf_rdx; frame.sf_sc.uc_mcontext.sc_ecx = regs->tf_rcx; frame.sf_sc.uc_mcontext.sc_eax = regs->tf_rax; frame.sf_sc.uc_mcontext.sc_eip = regs->tf_rip; frame.sf_sc.uc_mcontext.sc_cs = regs->tf_cs; frame.sf_sc.uc_mcontext.sc_gs = regs->tf_gs; frame.sf_sc.uc_mcontext.sc_fs = regs->tf_fs; frame.sf_sc.uc_mcontext.sc_es = regs->tf_es; frame.sf_sc.uc_mcontext.sc_ds = regs->tf_ds; frame.sf_sc.uc_mcontext.sc_eflags = regs->tf_rflags; frame.sf_sc.uc_mcontext.sc_esp_at_signal = regs->tf_rsp; frame.sf_sc.uc_mcontext.sc_ss = regs->tf_ss; frame.sf_sc.uc_mcontext.sc_err = regs->tf_err; frame.sf_sc.uc_mcontext.sc_cr2 = (u_int32_t)(uintptr_t)ksi->ksi_addr; frame.sf_sc.uc_mcontext.sc_trapno = bsd_to_linux_trapcode(code); if (copyout(&frame, fp, sizeof(frame)) != 0) { /* * Process has trashed its stack; give it an illegal * instruction to halt it in its tracks. */ PROC_LOCK(p); sigexit(td, SIGILL); } /* Build context to run handler in. */ regs->tf_rsp = PTROUT(fp); regs->tf_rip = __kernel_rt_sigreturn; regs->tf_rflags &= ~(PSL_T | PSL_D); regs->tf_cs = _ucode32sel; regs->tf_ss = _udatasel; regs->tf_ds = _udatasel; regs->tf_es = _udatasel; regs->tf_fs = _ufssel; regs->tf_gs = _ugssel; regs->tf_flags = TF_HASSEGS; set_pcb_flags(td->td_pcb, PCB_FULL_IRET); PROC_LOCK(p); mtx_lock(&psp->ps_mtx); } /* * Send an interrupt to process. * * Stack is set up to allow sigcode stored * in u. to call routine, followed by kcall * to sigreturn routine below. After sigreturn * resets the signal mask, the stack, and the * frame pointer, it returns to the user * specified pc, psl. */ static void linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) { struct thread *td = curthread; struct proc *p = td->td_proc; struct sigacts *psp; struct trapframe *regs; struct l_sigframe *fp, frame; l_sigset_t lmask; int oonstack; int sig, code; sig = ksi->ksi_signo; code = ksi->ksi_code; PROC_LOCK_ASSERT(p, MA_OWNED); psp = p->p_sigacts; mtx_assert(&psp->ps_mtx, MA_OWNED); if (SIGISMEMBER(psp->ps_siginfo, sig)) { /* Signal handler installed with SA_SIGINFO. */ linux_rt_sendsig(catcher, ksi, mask); return; } regs = td->td_frame; oonstack = sigonstack(regs->tf_rsp); /* Allocate space for the signal handler context. */ if ((td->td_pflags & TDP_ALTSTACK) && !oonstack && SIGISMEMBER(psp->ps_sigonstack, sig)) { fp = (struct l_sigframe *)((uintptr_t)td->td_sigstk.ss_sp + td->td_sigstk.ss_size - sizeof(struct l_sigframe)); } else fp = (struct l_sigframe *)regs->tf_rsp - 1; mtx_unlock(&psp->ps_mtx); PROC_UNLOCK(p); /* Build the argument list for the signal handler. */ sig = bsd_to_linux_signal(sig); bzero(&frame, sizeof(frame)); frame.sf_handler = PTROUT(catcher); frame.sf_sig = sig; bsd_to_linux_sigset(mask, &lmask); /* Build the signal context to be used by sigreturn. */ frame.sf_sc.sc_mask = lmask.__mask; frame.sf_sc.sc_gs = regs->tf_gs; frame.sf_sc.sc_fs = regs->tf_fs; frame.sf_sc.sc_es = regs->tf_es; frame.sf_sc.sc_ds = regs->tf_ds; frame.sf_sc.sc_edi = regs->tf_rdi; frame.sf_sc.sc_esi = regs->tf_rsi; frame.sf_sc.sc_ebp = regs->tf_rbp; frame.sf_sc.sc_ebx = regs->tf_rbx; frame.sf_sc.sc_esp = regs->tf_rsp; frame.sf_sc.sc_edx = regs->tf_rdx; frame.sf_sc.sc_ecx = regs->tf_rcx; frame.sf_sc.sc_eax = regs->tf_rax; frame.sf_sc.sc_eip = regs->tf_rip; frame.sf_sc.sc_cs = regs->tf_cs; frame.sf_sc.sc_eflags = regs->tf_rflags; frame.sf_sc.sc_esp_at_signal = regs->tf_rsp; frame.sf_sc.sc_ss = regs->tf_ss; frame.sf_sc.sc_err = regs->tf_err; frame.sf_sc.sc_cr2 = (u_int32_t)(uintptr_t)ksi->ksi_addr; frame.sf_sc.sc_trapno = bsd_to_linux_trapcode(code); frame.sf_extramask[0] = lmask.__mask; if (copyout(&frame, fp, sizeof(frame)) != 0) { /* * Process has trashed its stack; give it an illegal * instruction to halt it in its tracks. */ PROC_LOCK(p); sigexit(td, SIGILL); } /* Build context to run handler in. */ regs->tf_rsp = PTROUT(fp); regs->tf_rip = __kernel_sigreturn; regs->tf_rflags &= ~(PSL_T | PSL_D); regs->tf_cs = _ucode32sel; regs->tf_ss = _udatasel; regs->tf_ds = _udatasel; regs->tf_es = _udatasel; regs->tf_fs = _ufssel; regs->tf_gs = _ugssel; regs->tf_flags = TF_HASSEGS; set_pcb_flags(td->td_pcb, PCB_FULL_IRET); PROC_LOCK(p); mtx_lock(&psp->ps_mtx); } /* * System call to cleanup state after a signal * has been taken. Reset signal mask and * stack state from context left by sendsig (above). * Return to previous pc and psl as specified by * context left by sendsig. Check carefully to * make sure that the user has not modified the * psl to gain improper privileges or to cause * a machine fault. */ int linux_sigreturn(struct thread *td, struct linux_sigreturn_args *args) { struct l_sigframe frame; struct trapframe *regs; sigset_t bmask; l_sigset_t lmask; int eflags; ksiginfo_t ksi; regs = td->td_frame; /* * The trampoline code hands us the sigframe. * It is unsafe to keep track of it ourselves, in the event that a * program jumps out of a signal handler. */ if (copyin(args->sfp, &frame, sizeof(frame)) != 0) return (EFAULT); /* Check for security violations. */ eflags = frame.sf_sc.sc_eflags; if (!EFL_SECURE(eflags, regs->tf_rflags)) return(EINVAL); /* * Don't allow users to load a valid privileged %cs. Let the * hardware check for invalid selectors, excess privilege in * other selectors, invalid %eip's and invalid %esp's. */ if (!CS_SECURE(frame.sf_sc.sc_cs)) { ksiginfo_init_trap(&ksi); ksi.ksi_signo = SIGBUS; ksi.ksi_code = BUS_OBJERR; ksi.ksi_trapno = T_PROTFLT; ksi.ksi_addr = (void *)regs->tf_rip; trapsignal(td, &ksi); return(EINVAL); } lmask.__mask = frame.sf_sc.sc_mask; lmask.__mask = frame.sf_extramask[0]; linux_to_bsd_sigset(&lmask, &bmask); kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0); /* Restore signal context. */ regs->tf_rdi = frame.sf_sc.sc_edi; regs->tf_rsi = frame.sf_sc.sc_esi; regs->tf_rbp = frame.sf_sc.sc_ebp; regs->tf_rbx = frame.sf_sc.sc_ebx; regs->tf_rdx = frame.sf_sc.sc_edx; regs->tf_rcx = frame.sf_sc.sc_ecx; regs->tf_rax = frame.sf_sc.sc_eax; regs->tf_rip = frame.sf_sc.sc_eip; regs->tf_cs = frame.sf_sc.sc_cs; regs->tf_ds = frame.sf_sc.sc_ds; regs->tf_es = frame.sf_sc.sc_es; regs->tf_fs = frame.sf_sc.sc_fs; regs->tf_gs = frame.sf_sc.sc_gs; regs->tf_rflags = eflags; regs->tf_rsp = frame.sf_sc.sc_esp_at_signal; regs->tf_ss = frame.sf_sc.sc_ss; set_pcb_flags(td->td_pcb, PCB_FULL_IRET); return (EJUSTRETURN); } /* * System call to cleanup state after a signal * has been taken. Reset signal mask and * stack state from context left by rt_sendsig (above). * Return to previous pc and psl as specified by * context left by sendsig. Check carefully to * make sure that the user has not modified the * psl to gain improper privileges or to cause * a machine fault. */ int linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args) { struct l_ucontext uc; struct l_sigcontext *context; sigset_t bmask; l_stack_t *lss; stack_t ss; struct trapframe *regs; int eflags; ksiginfo_t ksi; regs = td->td_frame; /* * The trampoline code hands us the ucontext. * It is unsafe to keep track of it ourselves, in the event that a * program jumps out of a signal handler. */ if (copyin(args->ucp, &uc, sizeof(uc)) != 0) return (EFAULT); context = &uc.uc_mcontext; /* Check for security violations. */ eflags = context->sc_eflags; if (!EFL_SECURE(eflags, regs->tf_rflags)) return(EINVAL); /* * Don't allow users to load a valid privileged %cs. Let the * hardware check for invalid selectors, excess privilege in * other selectors, invalid %eip's and invalid %esp's. */ if (!CS_SECURE(context->sc_cs)) { ksiginfo_init_trap(&ksi); ksi.ksi_signo = SIGBUS; ksi.ksi_code = BUS_OBJERR; ksi.ksi_trapno = T_PROTFLT; ksi.ksi_addr = (void *)regs->tf_rip; trapsignal(td, &ksi); return(EINVAL); } linux_to_bsd_sigset(&uc.uc_sigmask, &bmask); kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0); /* * Restore signal context */ regs->tf_gs = context->sc_gs; regs->tf_fs = context->sc_fs; regs->tf_es = context->sc_es; regs->tf_ds = context->sc_ds; regs->tf_rdi = context->sc_edi; regs->tf_rsi = context->sc_esi; regs->tf_rbp = context->sc_ebp; regs->tf_rbx = context->sc_ebx; regs->tf_rdx = context->sc_edx; regs->tf_rcx = context->sc_ecx; regs->tf_rax = context->sc_eax; regs->tf_rip = context->sc_eip; regs->tf_cs = context->sc_cs; regs->tf_rflags = eflags; regs->tf_rsp = context->sc_esp_at_signal; regs->tf_ss = context->sc_ss; set_pcb_flags(td->td_pcb, PCB_FULL_IRET); /* * call sigaltstack & ignore results.. */ lss = &uc.uc_stack; ss.ss_sp = PTRIN(lss->ss_sp); ss.ss_size = lss->ss_size; ss.ss_flags = linux_to_bsd_sigaltstack(lss->ss_flags); (void)kern_sigaltstack(td, &ss, NULL); return (EJUSTRETURN); } static int linux32_fetch_syscall_args(struct thread *td) { struct proc *p; struct trapframe *frame; struct syscall_args *sa; p = td->td_proc; frame = td->td_frame; sa = &td->td_sa; sa->args[0] = frame->tf_rbx; sa->args[1] = frame->tf_rcx; sa->args[2] = frame->tf_rdx; sa->args[3] = frame->tf_rsi; sa->args[4] = frame->tf_rdi; sa->args[5] = frame->tf_rbp; /* Unconfirmed */ sa->code = frame->tf_rax; if (sa->code >= p->p_sysent->sv_size) /* nosys */ sa->callp = &p->p_sysent->sv_table[p->p_sysent->sv_size - 1]; else sa->callp = &p->p_sysent->sv_table[sa->code]; td->td_retval[0] = 0; td->td_retval[1] = frame->tf_rdx; return (0); } static void linux32_set_syscall_retval(struct thread *td, int error) { struct trapframe *frame = td->td_frame; cpu_set_syscall_retval(td, error); if (__predict_false(error != 0)) { if (error != ERESTART && error != EJUSTRETURN) frame->tf_rax = bsd_to_linux_errno(error); } } static void linux32_set_fork_retval(struct thread *td) { struct trapframe *frame = td->td_frame; frame->tf_rax = 0; } /* * Clear registers on exec * XXX copied from ia32_signal.c. */ static void linux_exec_setregs(struct thread *td, struct image_params *imgp, uintptr_t stack) { struct trapframe *regs = td->td_frame; struct pcb *pcb = td->td_pcb; register_t saved_rflags; regs = td->td_frame; pcb = td->td_pcb; if (td->td_proc->p_md.md_ldt != NULL) user_ldt_free(td); critical_enter(); wrmsr(MSR_FSBASE, 0); wrmsr(MSR_KGSBASE, 0); /* User value while we're in the kernel */ pcb->pcb_fsbase = 0; pcb->pcb_gsbase = 0; critical_exit(); pcb->pcb_initial_fpucw = __LINUX_NPXCW__; saved_rflags = regs->tf_rflags & PSL_T; bzero((char *)regs, sizeof(struct trapframe)); regs->tf_rip = imgp->entry_addr; regs->tf_rsp = stack; regs->tf_rflags = PSL_USER | saved_rflags; regs->tf_gs = _ugssel; regs->tf_fs = _ufssel; regs->tf_es = _udatasel; regs->tf_ds = _udatasel; regs->tf_ss = _udatasel; regs->tf_flags = TF_HASSEGS; regs->tf_cs = _ucode32sel; regs->tf_rbx = (register_t)imgp->ps_strings; x86_clear_dbregs(pcb); fpstate_drop(td); /* Do full restore on return so that we can change to a different %cs */ set_pcb_flags(pcb, PCB_32BIT | PCB_FULL_IRET); } /* * XXX copied from ia32_sysvec.c. */ static int linux_copyout_strings(struct image_params *imgp, uintptr_t *stack_base) { int argc, envc, error; u_int32_t *vectp; char *stringp; uintptr_t destp, ustringp; struct linux32_ps_strings *arginfo; char canary[LINUX_AT_RANDOM_LEN]; size_t execpath_len; arginfo = (struct linux32_ps_strings *)LINUX32_PS_STRINGS; destp = (uintptr_t)arginfo; if (imgp->execpath != NULL && imgp->auxargs != NULL) { execpath_len = strlen(imgp->execpath) + 1; destp -= execpath_len; destp = rounddown2(destp, sizeof(uint32_t)); imgp->execpathp = (void *)destp; error = copyout(imgp->execpath, imgp->execpathp, execpath_len); if (error != 0) return (error); } /* Prepare the canary for SSP. */ arc4rand(canary, sizeof(canary), 0); destp -= roundup(sizeof(canary), sizeof(uint32_t)); imgp->canary = (void *)destp; error = copyout(canary, imgp->canary, sizeof(canary)); if (error != 0) return (error); /* Allocate room for the argument and environment strings. */ destp -= ARG_MAX - imgp->args->stringspace; destp = rounddown2(destp, sizeof(uint32_t)); ustringp = destp; if (imgp->auxargs) { /* * Allocate room on the stack for the ELF auxargs * array. It has LINUX_AT_COUNT entries. */ destp -= LINUX_AT_COUNT * sizeof(Elf32_Auxinfo); destp = rounddown2(destp, sizeof(uint32_t)); } vectp = (uint32_t *)destp; /* * Allocate room for the argv[] and env vectors including the * terminating NULL pointers. */ vectp -= imgp->args->argc + 1 + imgp->args->envc + 1; /* vectp also becomes our initial stack base. */ *stack_base = (uintptr_t)vectp; stringp = imgp->args->begin_argv; argc = imgp->args->argc; envc = imgp->args->envc; /* Copy out strings - arguments and environment. */ error = copyout(stringp, (void *)ustringp, ARG_MAX - imgp->args->stringspace); if (error != 0) return (error); /* Fill in "ps_strings" struct for ps, w, etc. */ if (suword32(&arginfo->ps_argvstr, (uint32_t)(intptr_t)vectp) != 0 || suword32(&arginfo->ps_nargvstr, argc) != 0) return (EFAULT); /* Fill in argument portion of vector table. */ for (; argc > 0; --argc) { if (suword32(vectp++, ustringp) != 0) return (EFAULT); while (*stringp++ != 0) ustringp++; ustringp++; } /* A null vector table pointer separates the argp's from the envp's. */ if (suword32(vectp++, 0) != 0) return (EFAULT); if (suword32(&arginfo->ps_envstr, (uint32_t)(intptr_t)vectp) != 0 || suword32(&arginfo->ps_nenvstr, envc) != 0) return (EFAULT); /* Fill in environment portion of vector table. */ for (; envc > 0; --envc) { if (suword32(vectp++, ustringp) != 0) return (EFAULT); while (*stringp++ != 0) ustringp++; ustringp++; } /* The end of the vector table is a null pointer. */ if (suword32(vectp, 0) != 0) return (EFAULT); if (imgp->auxargs) { vectp++; error = imgp->sysent->sv_copyout_auxargs(imgp, (uintptr_t)vectp); if (error != 0) return (error); } return (0); } static SYSCTL_NODE(_compat, OID_AUTO, linux32, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, "32-bit Linux emulation"); static u_long linux32_maxdsiz = LINUX32_MAXDSIZ; SYSCTL_ULONG(_compat_linux32, OID_AUTO, maxdsiz, CTLFLAG_RW, &linux32_maxdsiz, 0, ""); static u_long linux32_maxssiz = LINUX32_MAXSSIZ; SYSCTL_ULONG(_compat_linux32, OID_AUTO, maxssiz, CTLFLAG_RW, &linux32_maxssiz, 0, ""); static u_long linux32_maxvmem = LINUX32_MAXVMEM; SYSCTL_ULONG(_compat_linux32, OID_AUTO, maxvmem, CTLFLAG_RW, &linux32_maxvmem, 0, ""); static void linux32_fixlimit(struct rlimit *rl, int which) { switch (which) { case RLIMIT_DATA: if (linux32_maxdsiz != 0) { if (rl->rlim_cur > linux32_maxdsiz) rl->rlim_cur = linux32_maxdsiz; if (rl->rlim_max > linux32_maxdsiz) rl->rlim_max = linux32_maxdsiz; } break; case RLIMIT_STACK: if (linux32_maxssiz != 0) { if (rl->rlim_cur > linux32_maxssiz) rl->rlim_cur = linux32_maxssiz; if (rl->rlim_max > linux32_maxssiz) rl->rlim_max = linux32_maxssiz; } break; case RLIMIT_VMEM: if (linux32_maxvmem != 0) { if (rl->rlim_cur > linux32_maxvmem) rl->rlim_cur = linux32_maxvmem; if (rl->rlim_max > linux32_maxvmem) rl->rlim_max = linux32_maxvmem; } break; } } struct sysentvec elf_linux_sysvec = { .sv_size = LINUX32_SYS_MAXSYSCALL, .sv_table = linux32_sysent, .sv_transtrap = linux_translate_traps, .sv_fixup = linux_fixup_elf, .sv_sendsig = linux_sendsig, .sv_sigcode = &_binary_linux32_vdso_so_o_start, .sv_szsigcode = &linux_szsigcode, .sv_name = "Linux ELF32", .sv_coredump = elf32_coredump, .sv_elf_core_osabi = ELFOSABI_NONE, .sv_elf_core_abi_vendor = LINUX_ABI_VENDOR, .sv_elf_core_prepare_notes = linux32_prepare_notes, .sv_imgact_try = linux_exec_imgact_try, .sv_minsigstksz = LINUX_MINSIGSTKSZ, .sv_minuser = VM_MIN_ADDRESS, .sv_maxuser = LINUX32_MAXUSER, .sv_usrstack = LINUX32_USRSTACK, .sv_psstrings = LINUX32_PS_STRINGS, .sv_psstringssz = sizeof(struct linux32_ps_strings), .sv_stackprot = VM_PROT_ALL, .sv_copyout_auxargs = linux_copyout_auxargs, .sv_copyout_strings = linux_copyout_strings, .sv_setregs = linux_exec_setregs, .sv_fixlimit = linux32_fixlimit, .sv_maxssiz = &linux32_maxssiz, .sv_flags = SV_ABI_LINUX | SV_ILP32 | SV_IA32 | SV_SHP | SV_SIG_DISCIGN | SV_SIG_WAITNDQ | SV_TIMEKEEP, .sv_set_syscall_retval = linux32_set_syscall_retval, .sv_fetch_syscall_args = linux32_fetch_syscall_args, .sv_syscallnames = NULL, .sv_shared_page_base = LINUX32_SHAREDPAGE, .sv_shared_page_len = PAGE_SIZE, .sv_schedtail = linux_schedtail, .sv_thread_detach = linux_thread_detach, .sv_trap = NULL, .sv_onexec = linux_on_exec_vmspace, .sv_onexit = linux_on_exit, .sv_ontdexit = linux_thread_dtor, .sv_setid_allowed = &linux_setid_allowed_query, .sv_set_fork_retval = linux32_set_fork_retval, }; static int linux_on_exec_vmspace(struct proc *p, struct image_params *imgp) { int error; error = linux_map_vdso(p, linux_vdso_obj, linux_vdso_base, LINUX32_VDSOPAGE_SIZE, imgp); if (error == 0) linux_on_exec(p, imgp); return (error); } /* * linux_vdso_install() and linux_exec_sysvec_init() must be called * after exec_sysvec_init() which is SI_SUB_EXEC (SI_ORDER_ANY). */ static void linux_exec_sysvec_init(void *param) { l_uintptr_t *ktimekeep_base, *ktsc_selector; struct sysentvec *sv; ptrdiff_t tkoff; sv = param; /* Fill timekeep_base */ exec_sysvec_init(sv); tkoff = kern_timekeep_base - linux_vdso_base; ktimekeep_base = (l_uintptr_t *)(linux_vdso_mapping + tkoff); *ktimekeep_base = sv->sv_timekeep_base; tkoff = kern_tsc_selector - linux_vdso_base; ktsc_selector = (l_uintptr_t *)(linux_vdso_mapping + tkoff); *ktsc_selector = linux_vdso_tsc_selector_idx(); if (bootverbose) printf("Linux i386 vDSO tsc_selector: %u\n", *ktsc_selector); } SYSINIT(elf_linux_exec_sysvec_init, SI_SUB_EXEC + 1, SI_ORDER_ANY, linux_exec_sysvec_init, &elf_linux_sysvec); static void linux_vdso_install(const void *param) { char *vdso_start = &_binary_linux32_vdso_so_o_start; char *vdso_end = &_binary_linux32_vdso_so_o_end; linux_szsigcode = vdso_end - vdso_start; MPASS(linux_szsigcode <= LINUX32_VDSOPAGE_SIZE); linux_vdso_base = LINUX32_VDSOPAGE; __elfN(linux_vdso_fixup)(vdso_start, linux_vdso_base); linux_vdso_obj = __elfN(linux_shared_page_init) (&linux_vdso_mapping, LINUX32_VDSOPAGE_SIZE); bcopy(vdso_start, linux_vdso_mapping, linux_szsigcode); linux_vdso_reloc(linux_vdso_mapping, linux_vdso_base); } SYSINIT(elf_linux_vdso_init, SI_SUB_EXEC + 1, SI_ORDER_FIRST, linux_vdso_install, NULL); static void linux_vdso_deinstall(const void *param) { __elfN(linux_shared_page_fini)(linux_vdso_obj, linux_vdso_mapping, LINUX32_VDSOPAGE_SIZE); } SYSUNINIT(elf_linux_vdso_uninit, SI_SUB_EXEC, SI_ORDER_FIRST, linux_vdso_deinstall, NULL); static void linux_vdso_reloc(char *mapping, Elf_Addr offset) { const Elf_Shdr *shdr; const Elf_Rel *rel; const Elf_Ehdr *ehdr; Elf32_Addr *where; Elf_Size rtype, symidx; Elf32_Addr addr, addend; int i, relcnt; MPASS(offset != 0); relcnt = 0; ehdr = (const Elf_Ehdr *)mapping; shdr = (const Elf_Shdr *)(mapping + ehdr->e_shoff); for (i = 0; i < ehdr->e_shnum; i++) { switch (shdr[i].sh_type) { case SHT_REL: rel = (const Elf_Rel *)(mapping + shdr[i].sh_offset); relcnt = shdr[i].sh_size / sizeof(*rel); break; case SHT_RELA: printf("Linux i386 vDSO: unexpected Rela section\n"); break; } } for (i = 0; i < relcnt; i++, rel++) { where = (Elf32_Addr *)(mapping + rel->r_offset); addend = *where; rtype = ELF_R_TYPE(rel->r_info); symidx = ELF_R_SYM(rel->r_info); switch (rtype) { case R_386_NONE: /* none */ break; case R_386_RELATIVE: /* B + A */ addr = (Elf32_Addr)PTROUT(offset + addend); if (*where != addr) *where = addr; break; case R_386_IRELATIVE: printf("Linux i386 vDSO: unexpected ifunc relocation, " "symbol index %ld\n", (intmax_t)symidx); break; default: printf("Linux i386 vDSO: unexpected relocation type %ld, " "symbol index %ld\n", (intmax_t)rtype, (intmax_t)symidx); } } } static char GNU_ABI_VENDOR[] = "GNU"; static int GNULINUX_ABI_DESC = 0; static bool linux32_trans_osrel(const Elf_Note *note, int32_t *osrel) { const Elf32_Word *desc; uintptr_t p; p = (uintptr_t)(note + 1); p += roundup2(note->n_namesz, sizeof(Elf32_Addr)); desc = (const Elf32_Word *)p; if (desc[0] != GNULINUX_ABI_DESC) return (false); /* * For Linux we encode osrel using the Linux convention of * (version << 16) | (major << 8) | (minor) * See macro in linux_mib.h */ *osrel = LINUX_KERNVER(desc[1], desc[2], desc[3]); return (true); } static Elf_Brandnote linux32_brandnote = { .hdr.n_namesz = sizeof(GNU_ABI_VENDOR), .hdr.n_descsz = 16, /* XXX at least 16 */ .hdr.n_type = 1, .vendor = GNU_ABI_VENDOR, .flags = BN_TRANSLATE_OSREL, .trans_osrel = linux32_trans_osrel }; static Elf32_Brandinfo linux_brand = { .brand = ELFOSABI_LINUX, .machine = EM_386, .compat_3_brand = "Linux", .emul_path = linux_emul_path, .interp_path = "/lib/ld-linux.so.1", .sysvec = &elf_linux_sysvec, .interp_newpath = NULL, .brand_note = &linux32_brandnote, .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE }; static Elf32_Brandinfo linux_glibc2brand = { .brand = ELFOSABI_LINUX, .machine = EM_386, .compat_3_brand = "Linux", .emul_path = linux_emul_path, .interp_path = "/lib/ld-linux.so.2", .sysvec = &elf_linux_sysvec, .interp_newpath = NULL, .brand_note = &linux32_brandnote, .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE }; static Elf32_Brandinfo linux_muslbrand = { .brand = ELFOSABI_LINUX, .machine = EM_386, .compat_3_brand = "Linux", .emul_path = linux_emul_path, .interp_path = "/lib/ld-musl-i386.so.1", .sysvec = &elf_linux_sysvec, .interp_newpath = NULL, .brand_note = &linux32_brandnote, .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE | LINUX_BI_FUTEX_REQUEUE }; Elf32_Brandinfo *linux_brandlist[] = { &linux_brand, &linux_glibc2brand, &linux_muslbrand, NULL }; static int linux_elf_modevent(module_t mod, int type, void *data) { Elf32_Brandinfo **brandinfo; int error; struct linux_ioctl_handler **lihp; error = 0; switch(type) { case MOD_LOAD: for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; ++brandinfo) if (elf32_insert_brand_entry(*brandinfo) < 0) error = EINVAL; if (error == 0) { SET_FOREACH(lihp, linux_ioctl_handler_set) linux32_ioctl_register_handler(*lihp); stclohz = (stathz ? stathz : hz); if (bootverbose) printf("Linux i386 ELF exec handler installed\n"); } else printf("cannot insert Linux i386 ELF brand handler\n"); break; case MOD_UNLOAD: for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; ++brandinfo) if (elf32_brand_inuse(*brandinfo)) error = EBUSY; if (error == 0) { for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; ++brandinfo) if (elf32_remove_brand_entry(*brandinfo) < 0) error = EINVAL; } if (error == 0) { SET_FOREACH(lihp, linux_ioctl_handler_set) linux32_ioctl_unregister_handler(*lihp); if (bootverbose) printf("Linux i386 ELF exec handler removed\n"); } else printf("Could not deinstall Linux i386 ELF interpreter entry\n"); break; default: return (EOPNOTSUPP); } return (error); } static moduledata_t linux_elf_mod = { "linuxelf", linux_elf_modevent, 0 }; DECLARE_MODULE_TIED(linuxelf, linux_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY); MODULE_DEPEND(linuxelf, linux_common, 1, 1, 1); FEATURE(linux, "Linux 32bit support"); diff --git a/sys/arm64/linux/linux_machdep.c b/sys/arm64/linux/linux_machdep.c index eff084eb9f96..055e9374424c 100644 --- a/sys/arm64/linux/linux_machdep.c +++ b/sys/arm64/linux/linux_machdep.c @@ -1,148 +1,149 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 2018 Turing Robotic Industries Inc. * Copyright (c) 2000 Marcel Moolenaar * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $FreeBSD$ */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include +#include #include #include #include /* DTrace init */ LIN_SDT_PROVIDER_DECLARE(LINUX_DTRACE); /* DTrace probes */ LIN_SDT_PROBE_DEFINE0(machdep, linux_mmap2, todo); /* * LINUXTODO: deduplicate; linux_execve is common across archs, except that on * amd64 compat linuxulator it calls freebsd32_exec_copyin_args. */ int linux_execve(struct thread *td, struct linux_execve_args *uap) { struct image_args eargs; char *path; int error; if (!LUSECONVPATH(td)) { error = exec_copyin_args(&eargs, uap->path, UIO_USERSPACE, uap->argp, uap->envp); } else { LCONVPATHEXIST(td, uap->path, &path); error = exec_copyin_args(&eargs, path, UIO_SYSSPACE, uap->argp, uap->envp); LFREEPATH(path); } if (error == 0) error = linux_common_execve(td, &eargs); AUDIT_SYSCALL_EXIT(error == EJUSTRETURN ? 0 : error, td); return (error); } int linux_set_upcall(struct thread *td, register_t stack) { if (stack) td->td_frame->tf_sp = stack; /* * The newly created Linux thread returns * to the user space by the same path that a parent does. */ td->td_frame->tf_x[0] = 0; return (0); } /* LINUXTODO: deduplicate arm64 linux_mmap2 */ int linux_mmap2(struct thread *td, struct linux_mmap2_args *uap) { LIN_SDT_PROBE0(machdep, linux_mmap2, todo); return (linux_mmap_common(td, PTROUT(uap->addr), uap->len, uap->prot, uap->flags, uap->fd, uap->pgoff)); } int linux_mprotect(struct thread *td, struct linux_mprotect_args *uap) { return (linux_mprotect_common(td, PTROUT(uap->addr), uap->len, uap->prot)); } int linux_madvise(struct thread *td, struct linux_madvise_args *uap) { return (linux_madvise_common(td, PTROUT(uap->addr), uap->len, uap->behav)); } int linux_set_cloned_tls(struct thread *td, void *desc) { if ((uint64_t)desc >= VM_MAXUSER_ADDRESS) return (EPERM); return (cpu_set_user_tls(td, desc)); } void bsd_to_linux_regset(struct reg *b_reg, struct linux_pt_regset *l_regset) { KASSERT(sizeof(l_regset->x) == sizeof(b_reg->x) + sizeof(l_ulong), ("%s: size mismatch\n", __func__)); memcpy(l_regset->x, b_reg->x, sizeof(b_reg->x)); l_regset->x[30] = b_reg->lr; l_regset->sp = b_reg->sp; l_regset->pc = b_reg->elr; l_regset->cpsr = b_reg->spsr; } diff --git a/sys/arm64/linux/linux_sysvec.c b/sys/arm64/linux/linux_sysvec.c index 63438cb62ec4..dfbc12ee4b42 100644 --- a/sys/arm64/linux/linux_sysvec.c +++ b/sys/arm64/linux/linux_sysvec.c @@ -1,768 +1,769 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 1994-1996 Søren Schmidt * Copyright (c) 2018 Turing Robotic Industries Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include +#include #include #include #include #include #include #include #include #ifdef VFP #include #endif MODULE_VERSION(linux64elf, 1); #define LINUX_VDSOPAGE_SIZE PAGE_SIZE * 2 #define LINUX_VDSOPAGE (VM_MAXUSER_ADDRESS - \ LINUX_VDSOPAGE_SIZE) #define LINUX_SHAREDPAGE (LINUX_VDSOPAGE - PAGE_SIZE) /* * PAGE_SIZE - the size * of the native SHAREDPAGE */ #define LINUX_USRSTACK LINUX_SHAREDPAGE #define LINUX_PS_STRINGS (LINUX_USRSTACK - \ sizeof(struct ps_strings)) static int linux_szsigcode; static vm_object_t linux_vdso_obj; static char *linux_vdso_mapping; extern char _binary_linux_vdso_so_o_start; extern char _binary_linux_vdso_so_o_end; static vm_offset_t linux_vdso_base; extern struct sysent linux_sysent[LINUX_SYS_MAXSYSCALL]; SET_DECLARE(linux_ioctl_handler_set, struct linux_ioctl_handler); static int linux_copyout_strings(struct image_params *imgp, uintptr_t *stack_base); static int linux_elf_fixup(uintptr_t *stack_base, struct image_params *iparams); static bool linux_trans_osrel(const Elf_Note *note, int32_t *osrel); static void linux_vdso_install(const void *param); static void linux_vdso_deinstall(const void *param); static void linux_vdso_reloc(char *mapping, Elf_Addr offset); static void linux_set_syscall_retval(struct thread *td, int error); static int linux_fetch_syscall_args(struct thread *td); static void linux_exec_setregs(struct thread *td, struct image_params *imgp, uintptr_t stack); static void linux_exec_sysvec_init(void *param); static int linux_on_exec_vmspace(struct proc *p, struct image_params *imgp); /* DTrace init */ LIN_SDT_PROVIDER_DECLARE(LINUX_DTRACE); /* DTrace probes */ LIN_SDT_PROBE_DEFINE2(sysvec, linux_translate_traps, todo, "int", "int"); LIN_SDT_PROBE_DEFINE0(sysvec, linux_exec_setregs, todo); LIN_SDT_PROBE_DEFINE0(sysvec, linux_copyout_auxargs, todo); LIN_SDT_PROBE_DEFINE0(sysvec, linux_elf_fixup, todo); LIN_SDT_PROBE_DEFINE0(sysvec, linux_rt_sigreturn, todo); LIN_SDT_PROBE_DEFINE0(sysvec, linux_rt_sendsig, todo); LIN_SDT_PROBE_DEFINE0(sysvec, linux_vdso_install, todo); LIN_SDT_PROBE_DEFINE0(sysvec, linux_vdso_deinstall, todo); LINUX_VDSO_SYM_CHAR(linux_platform); LINUX_VDSO_SYM_INTPTR(kern_timekeep_base); LINUX_VDSO_SYM_INTPTR(__kernel_rt_sigreturn); /* LINUXTODO: do we have traps to translate? */ static int linux_translate_traps(int signal, int trap_code) { LIN_SDT_PROBE2(sysvec, linux_translate_traps, todo, signal, trap_code); return (signal); } static int linux_fetch_syscall_args(struct thread *td) { struct proc *p; struct syscall_args *sa; register_t *ap; p = td->td_proc; ap = td->td_frame->tf_x; sa = &td->td_sa; sa->code = td->td_frame->tf_x[8]; /* LINUXTODO: generic syscall? */ if (sa->code >= p->p_sysent->sv_size) sa->callp = &p->p_sysent->sv_table[0]; else sa->callp = &p->p_sysent->sv_table[sa->code]; if (sa->callp->sy_narg > MAXARGS) panic("ARM64TODO: Could we have more than %d args?", MAXARGS); memcpy(sa->args, ap, MAXARGS * sizeof(register_t)); td->td_retval[0] = 0; return (0); } static void linux_set_syscall_retval(struct thread *td, int error) { td->td_retval[1] = td->td_frame->tf_x[1]; cpu_set_syscall_retval(td, error); if (__predict_false(error != 0)) { if (error != ERESTART && error != EJUSTRETURN) td->td_frame->tf_x[0] = bsd_to_linux_errno(error); } } static int linux_copyout_auxargs(struct image_params *imgp, uintptr_t base) { Elf_Auxargs *args; Elf_Auxinfo *argarray, *pos; struct proc *p; int error, issetugid; LIN_SDT_PROBE0(sysvec, linux_copyout_auxargs, todo); p = imgp->proc; args = (Elf64_Auxargs *)imgp->auxargs; argarray = pos = malloc(LINUX_AT_COUNT * sizeof(*pos), M_TEMP, M_WAITOK | M_ZERO); issetugid = p->p_flag & P_SUGID ? 1 : 0; AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO_EHDR, linux_vdso_base); AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, *imgp->sysent->sv_hwcap); AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz); AUXARGS_ENTRY(pos, LINUX_AT_CLKTCK, stclohz); AUXARGS_ENTRY(pos, AT_PHDR, args->phdr); AUXARGS_ENTRY(pos, AT_PHENT, args->phent); AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum); AUXARGS_ENTRY(pos, AT_BASE, args->base); AUXARGS_ENTRY(pos, AT_FLAGS, args->flags); AUXARGS_ENTRY(pos, AT_ENTRY, args->entry); AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_ucred->cr_ruid); AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid); AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_ucred->cr_rgid); AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid); AUXARGS_ENTRY(pos, LINUX_AT_SECURE, issetugid); AUXARGS_ENTRY_PTR(pos, LINUX_AT_RANDOM, imgp->canary); AUXARGS_ENTRY(pos, LINUX_AT_HWCAP2, *imgp->sysent->sv_hwcap2); if (imgp->execpathp != 0) AUXARGS_ENTRY_PTR(pos, LINUX_AT_EXECFN, imgp->execpathp); if (args->execfd != -1) AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd); AUXARGS_ENTRY(pos, LINUX_AT_PLATFORM, PTROUT(linux_platform)); AUXARGS_ENTRY(pos, AT_NULL, 0); free(imgp->auxargs, M_TEMP); imgp->auxargs = NULL; KASSERT(pos - argarray <= LINUX_AT_COUNT, ("Too many auxargs")); error = copyout(argarray, (void *)base, sizeof(*argarray) * LINUX_AT_COUNT); free(argarray, M_TEMP); return (error); } static int linux_elf_fixup(uintptr_t *stack_base, struct image_params *imgp) { LIN_SDT_PROBE0(sysvec, linux_elf_fixup, todo); return (0); } /* * Copy strings out to the new process address space, constructing new arg * and env vector tables. Return a pointer to the base so that it can be used * as the initial stack pointer. * LINUXTODO: deduplicate against other linuxulator archs */ static int linux_copyout_strings(struct image_params *imgp, uintptr_t *stack_base) { char **vectp; char *stringp; uintptr_t destp, ustringp; struct ps_strings *arginfo; char canary[LINUX_AT_RANDOM_LEN]; size_t execpath_len; struct proc *p; int argc, envc, error; p = imgp->proc; arginfo = (struct ps_strings *)PROC_PS_STRINGS(p); destp = (uintptr_t)arginfo; if (imgp->execpath != NULL && imgp->auxargs != NULL) { execpath_len = strlen(imgp->execpath) + 1; destp -= execpath_len; destp = rounddown2(destp, sizeof(void *)); imgp->execpathp = (void *)destp; error = copyout(imgp->execpath, imgp->execpathp, execpath_len); if (error != 0) return (error); } /* Prepare the canary for SSP. */ arc4rand(canary, sizeof(canary), 0); destp -= roundup(sizeof(canary), sizeof(void *)); imgp->canary = (void *)destp; error = copyout(canary, imgp->canary, sizeof(canary)); if (error != 0) return (error); /* Allocate room for the argument and environment strings. */ destp -= ARG_MAX - imgp->args->stringspace; destp = rounddown2(destp, sizeof(void *)); ustringp = destp; if (imgp->auxargs) { /* * Allocate room on the stack for the ELF auxargs * array. It has up to LINUX_AT_COUNT entries. */ destp -= LINUX_AT_COUNT * sizeof(Elf64_Auxinfo); destp = rounddown2(destp, sizeof(void *)); } vectp = (char **)destp; /* * Allocate room for argc and the argv[] and env vectors including the * terminating NULL pointers. */ vectp -= 1 + imgp->args->argc + 1 + imgp->args->envc + 1; vectp = (char **)STACKALIGN(vectp); /* vectp also becomes our initial stack base. */ *stack_base = (uintptr_t)vectp; stringp = imgp->args->begin_argv; argc = imgp->args->argc; envc = imgp->args->envc; /* Copy out strings - arguments and environment. */ error = copyout(stringp, (void *)ustringp, ARG_MAX - imgp->args->stringspace); if (error != 0) return (error); /* Fill in "ps_strings" struct for ps, w, etc. */ if (suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp) != 0 || suword(&arginfo->ps_nargvstr, argc) != 0) return (EFAULT); if (suword(vectp++, argc) != 0) return (EFAULT); /* Fill in argument portion of vector table. */ for (; argc > 0; --argc) { if (suword(vectp++, ustringp) != 0) return (EFAULT); while (*stringp++ != 0) ustringp++; ustringp++; } /* A null vector table pointer separates the argp's from the envp's. */ if (suword(vectp++, 0) != 0) return (EFAULT); if (suword(&arginfo->ps_envstr, (long)(intptr_t)vectp) != 0 || suword(&arginfo->ps_nenvstr, envc) != 0) return (EFAULT); /* Fill in environment portion of vector table. */ for (; envc > 0; --envc) { if (suword(vectp++, ustringp) != 0) return (EFAULT); while (*stringp++ != 0) ustringp++; ustringp++; } /* The end of the vector table is a null pointer. */ if (suword(vectp, 0) != 0) return (EFAULT); if (imgp->auxargs) { vectp++; error = imgp->sysent->sv_copyout_auxargs(imgp, (uintptr_t)vectp); if (error != 0) return (error); } return (0); } /* * Reset registers to default values on exec. */ static void linux_exec_setregs(struct thread *td, struct image_params *imgp, uintptr_t stack) { struct trapframe *regs = td->td_frame; struct pcb *pcb = td->td_pcb; /* LINUXTODO: validate */ LIN_SDT_PROBE0(sysvec, linux_exec_setregs, todo); memset(regs, 0, sizeof(*regs)); /* glibc start.S registers function pointer in x0 with atexit. */ regs->tf_sp = stack; #if 0 /* LINUXTODO: See if this is used. */ regs->tf_lr = imgp->entry_addr; #else regs->tf_lr = 0xffffffffffffffff; #endif regs->tf_elr = imgp->entry_addr; pcb->pcb_tpidr_el0 = 0; pcb->pcb_tpidrro_el0 = 0; WRITE_SPECIALREG(tpidrro_el0, 0); WRITE_SPECIALREG(tpidr_el0, 0); #ifdef VFP vfp_reset_state(td, pcb); #endif /* * Clear debug register state. It is not applicable to the new process. */ bzero(&pcb->pcb_dbg_regs, sizeof(pcb->pcb_dbg_regs)); } int linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args) { struct l_sigframe frame; struct trapframe *tf; int error; tf = td->td_frame; if (copyin((void *)tf->tf_sp, &frame, sizeof(frame))) return (EFAULT); error = set_mcontext(td, &frame.sf_uc.uc_mcontext); if (error != 0) return (error); /* Restore signal mask. */ kern_sigprocmask(td, SIG_SETMASK, &frame.sf_uc.uc_sigmask, NULL, 0); return (EJUSTRETURN); } static void linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) { struct thread *td; struct proc *p; struct trapframe *tf; struct l_sigframe *fp, frame; struct sigacts *psp; int onstack, sig; uint32_t spsr; td = curthread; p = td->td_proc; PROC_LOCK_ASSERT(p, MA_OWNED); sig = ksi->ksi_signo; psp = p->p_sigacts; mtx_assert(&psp->ps_mtx, MA_OWNED); tf = td->td_frame; onstack = sigonstack(tf->tf_sp); CTR4(KTR_SIG, "sendsig: td=%p (%s) catcher=%p sig=%d", td, p->p_comm, catcher, sig); /* Allocate and validate space for the signal handler context. */ if ((td->td_pflags & TDP_ALTSTACK) != 0 && !onstack && SIGISMEMBER(psp->ps_sigonstack, sig)) { fp = (struct l_sigframe *)((uintptr_t)td->td_sigstk.ss_sp + td->td_sigstk.ss_size); #if defined(COMPAT_43) td->td_sigstk.ss_flags |= SS_ONSTACK; #endif } else { fp = (struct l_sigframe *)td->td_frame->tf_sp; } /* Make room, keeping the stack aligned */ fp--; fp = (struct l_sigframe *)STACKALIGN(fp); /* Fill in the frame to copy out */ bzero(&frame, sizeof(frame)); get_mcontext(td, &frame.sf_uc.uc_mcontext, 0); spsr = frame.sf_uc.uc_mcontext.mc_gpregs.gp_spsr; /* Translate the signal. */ sig = bsd_to_linux_signal(sig); siginfo_to_lsiginfo(&ksi->ksi_info, &frame.sf_si, sig); frame.sf_uc.uc_sigmask = *mask; frame.sf_uc.uc_stack = td->td_sigstk; frame.sf_uc.uc_stack.ss_flags = (td->td_pflags & TDP_ALTSTACK) != 0 ? (onstack ? SS_ONSTACK : 0) : SS_DISABLE; mtx_unlock(&psp->ps_mtx); PROC_UNLOCK(td->td_proc); /* Copy the sigframe out to the user's stack. */ if (copyout(&frame, fp, sizeof(*fp)) != 0) { /* Process has trashed its stack. Kill it. */ CTR2(KTR_SIG, "sendsig: sigexit td=%p fp=%p", td, fp); PROC_LOCK(p); sigexit(td, SIGILL); } tf->tf_x[0]= sig; tf->tf_x[1] = (register_t)&fp->sf_si; tf->tf_x[2] = (register_t)&fp->sf_uc; tf->tf_elr = (register_t)catcher; tf->tf_sp = (register_t)fp; tf->tf_lr = (register_t)__kernel_rt_sigreturn; CTR3(KTR_SIG, "sendsig: return td=%p pc=%#x sp=%#x", td, tf->tf_elr, tf->tf_sp); PROC_LOCK(p); mtx_lock(&psp->ps_mtx); } struct sysentvec elf_linux_sysvec = { .sv_size = LINUX_SYS_MAXSYSCALL, .sv_table = linux_sysent, .sv_transtrap = linux_translate_traps, .sv_fixup = linux_elf_fixup, .sv_sendsig = linux_rt_sendsig, .sv_sigcode = &_binary_linux_vdso_so_o_start, .sv_szsigcode = &linux_szsigcode, .sv_name = "Linux ELF64", .sv_coredump = elf64_coredump, .sv_elf_core_osabi = ELFOSABI_NONE, .sv_elf_core_abi_vendor = LINUX_ABI_VENDOR, .sv_elf_core_prepare_notes = linux64_prepare_notes, .sv_imgact_try = linux_exec_imgact_try, .sv_minsigstksz = LINUX_MINSIGSTKSZ, .sv_minuser = VM_MIN_ADDRESS, .sv_maxuser = VM_MAXUSER_ADDRESS, .sv_usrstack = LINUX_USRSTACK, .sv_psstrings = LINUX_PS_STRINGS, .sv_psstringssz = sizeof(struct ps_strings), .sv_stackprot = VM_PROT_READ | VM_PROT_WRITE, .sv_copyout_auxargs = linux_copyout_auxargs, .sv_copyout_strings = linux_copyout_strings, .sv_setregs = linux_exec_setregs, .sv_fixlimit = NULL, .sv_maxssiz = NULL, .sv_flags = SV_ABI_LINUX | SV_LP64 | SV_SHP | SV_SIG_DISCIGN | SV_SIG_WAITNDQ | SV_TIMEKEEP, .sv_set_syscall_retval = linux_set_syscall_retval, .sv_fetch_syscall_args = linux_fetch_syscall_args, .sv_syscallnames = NULL, .sv_shared_page_base = LINUX_SHAREDPAGE, .sv_shared_page_len = PAGE_SIZE, .sv_schedtail = linux_schedtail, .sv_thread_detach = linux_thread_detach, .sv_trap = NULL, .sv_hwcap = &elf_hwcap, .sv_hwcap2 = &elf_hwcap2, .sv_onexec = linux_on_exec_vmspace, .sv_onexit = linux_on_exit, .sv_ontdexit = linux_thread_dtor, .sv_setid_allowed = &linux_setid_allowed_query, }; static int linux_on_exec_vmspace(struct proc *p, struct image_params *imgp) { int error; error = linux_map_vdso(p, linux_vdso_obj, linux_vdso_base, LINUX_VDSOPAGE_SIZE, imgp); if (error == 0) linux_on_exec(p, imgp); return (error); } /* * linux_vdso_install() and linux_exec_sysvec_init() must be called * after exec_sysvec_init() which is SI_SUB_EXEC (SI_ORDER_ANY). */ static void linux_exec_sysvec_init(void *param) { l_uintptr_t *ktimekeep_base; struct sysentvec *sv; ptrdiff_t tkoff; sv = param; /* Fill timekeep_base */ exec_sysvec_init(sv); tkoff = kern_timekeep_base - linux_vdso_base; ktimekeep_base = (l_uintptr_t *)(linux_vdso_mapping + tkoff); *ktimekeep_base = sv->sv_timekeep_base; } SYSINIT(elf_linux_exec_sysvec_init, SI_SUB_EXEC + 1, SI_ORDER_ANY, linux_exec_sysvec_init, &elf_linux_sysvec); static void linux_vdso_install(const void *param) { char *vdso_start = &_binary_linux_vdso_so_o_start; char *vdso_end = &_binary_linux_vdso_so_o_end; linux_szsigcode = vdso_end - vdso_start; MPASS(linux_szsigcode <= LINUX_VDSOPAGE_SIZE); linux_vdso_base = LINUX_VDSOPAGE; __elfN(linux_vdso_fixup)(vdso_start, linux_vdso_base); linux_vdso_obj = __elfN(linux_shared_page_init) (&linux_vdso_mapping, LINUX_VDSOPAGE_SIZE); bcopy(vdso_start, linux_vdso_mapping, linux_szsigcode); linux_vdso_reloc(linux_vdso_mapping, linux_vdso_base); } SYSINIT(elf_linux_vdso_init, SI_SUB_EXEC + 1, SI_ORDER_FIRST, linux_vdso_install, NULL); static void linux_vdso_deinstall(const void *param) { __elfN(linux_shared_page_fini)(linux_vdso_obj, linux_vdso_mapping, LINUX_VDSOPAGE_SIZE); } SYSUNINIT(elf_linux_vdso_uninit, SI_SUB_EXEC, SI_ORDER_FIRST, linux_vdso_deinstall, NULL); static void linux_vdso_reloc(char *mapping, Elf_Addr offset) { Elf_Size rtype, symidx; const Elf_Rela *rela; const Elf_Shdr *shdr; const Elf_Ehdr *ehdr; Elf_Addr *where; Elf_Addr addr, addend; int i, relacnt; MPASS(offset != 0); relacnt = 0; ehdr = (const Elf_Ehdr *)mapping; shdr = (const Elf_Shdr *)(mapping + ehdr->e_shoff); for (i = 0; i < ehdr->e_shnum; i++) { switch (shdr[i].sh_type) { case SHT_REL: printf("Linux Aarch64 vDSO: unexpected Rel section\n"); break; case SHT_RELA: rela = (const Elf_Rela *)(mapping + shdr[i].sh_offset); relacnt = shdr[i].sh_size / sizeof(*rela); } } for (i = 0; i < relacnt; i++, rela++) { where = (Elf_Addr *)(mapping + rela->r_offset); addend = rela->r_addend; rtype = ELF_R_TYPE(rela->r_info); symidx = ELF_R_SYM(rela->r_info); switch (rtype) { case R_AARCH64_NONE: /* none */ break; case R_AARCH64_RELATIVE: /* B + A */ addr = (Elf_Addr)(mapping + addend); if (*where != addr) *where = addr; break; default: printf("Linux Aarch64 vDSO: unexpected relocation type %ld, " "symbol index %ld\n", rtype, symidx); } } } static char GNU_ABI_VENDOR[] = "GNU"; static int GNU_ABI_LINUX = 0; /* LINUXTODO: deduplicate */ static bool linux_trans_osrel(const Elf_Note *note, int32_t *osrel) { const Elf32_Word *desc; uintptr_t p; p = (uintptr_t)(note + 1); p += roundup2(note->n_namesz, sizeof(Elf32_Addr)); desc = (const Elf32_Word *)p; if (desc[0] != GNU_ABI_LINUX) return (false); *osrel = LINUX_KERNVER(desc[1], desc[2], desc[3]); return (true); } static Elf_Brandnote linux64_brandnote = { .hdr.n_namesz = sizeof(GNU_ABI_VENDOR), .hdr.n_descsz = 16, .hdr.n_type = 1, .vendor = GNU_ABI_VENDOR, .flags = BN_TRANSLATE_OSREL, .trans_osrel = linux_trans_osrel }; static Elf64_Brandinfo linux_glibc2brand = { .brand = ELFOSABI_LINUX, .machine = EM_AARCH64, .compat_3_brand = "Linux", .emul_path = linux_emul_path, .interp_path = "/lib64/ld-linux-x86-64.so.2", .sysvec = &elf_linux_sysvec, .interp_newpath = NULL, .brand_note = &linux64_brandnote, .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE }; Elf64_Brandinfo *linux_brandlist[] = { &linux_glibc2brand, NULL }; static int linux64_elf_modevent(module_t mod, int type, void *data) { Elf64_Brandinfo **brandinfo; struct linux_ioctl_handler**lihp; int error; error = 0; switch(type) { case MOD_LOAD: for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; ++brandinfo) if (elf64_insert_brand_entry(*brandinfo) < 0) error = EINVAL; if (error == 0) { SET_FOREACH(lihp, linux_ioctl_handler_set) linux_ioctl_register_handler(*lihp); stclohz = (stathz ? stathz : hz); if (bootverbose) printf("Linux arm64 ELF exec handler installed\n"); } break; case MOD_UNLOAD: for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; ++brandinfo) if (elf64_brand_inuse(*brandinfo)) error = EBUSY; if (error == 0) { for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; ++brandinfo) if (elf64_remove_brand_entry(*brandinfo) < 0) error = EINVAL; } if (error == 0) { SET_FOREACH(lihp, linux_ioctl_handler_set) linux_ioctl_unregister_handler(*lihp); if (bootverbose) printf("Linux arm64 ELF exec handler removed\n"); } else printf("Could not deinstall Linux arm64 ELF interpreter entry\n"); break; default: return (EOPNOTSUPP); } return (error); } static moduledata_t linux64_elf_mod = { "linux64elf", linux64_elf_modevent, 0 }; DECLARE_MODULE_TIED(linux64elf, linux64_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY); MODULE_DEPEND(linux64elf, linux_common, 1, 1, 1); FEATURE(linux64, "AArch64 Linux 64bit support"); diff --git a/sys/compat/linux/linux.h b/sys/compat/linux/linux.h index 0b405b692796..f73a0484cf55 100644 --- a/sys/compat/linux/linux.h +++ b/sys/compat/linux/linux.h @@ -1,262 +1,264 @@ /*- * Copyright (c) 2015 Dmitry Chagin * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $FreeBSD$ */ #ifndef _LINUX_MI_H_ #define _LINUX_MI_H_ #include /* * Private Brandinfo flags */ #define LINUX_BI_FUTEX_REQUEUE 0x01000000 /* * poll() */ #define LINUX_POLLIN 0x0001 #define LINUX_POLLPRI 0x0002 #define LINUX_POLLOUT 0x0004 #define LINUX_POLLERR 0x0008 #define LINUX_POLLHUP 0x0010 #define LINUX_POLLNVAL 0x0020 #define LINUX_POLLRDNORM 0x0040 #define LINUX_POLLRDBAND 0x0080 #define LINUX_POLLWRNORM 0x0100 #define LINUX_POLLWRBAND 0x0200 #define LINUX_POLLMSG 0x0400 #define LINUX_POLLREMOVE 0x1000 #define LINUX_POLLRDHUP 0x2000 #define LINUX_IFHWADDRLEN 6 #define LINUX_IFNAMSIZ 16 struct l_sockaddr { unsigned short sa_family; char sa_data[14]; }; #define LINUX_ARPHRD_ETHER 1 #define LINUX_ARPHRD_LOOPBACK 772 /* * Supported address families */ #define LINUX_AF_UNSPEC 0 #define LINUX_AF_UNIX 1 #define LINUX_AF_INET 2 #define LINUX_AF_AX25 3 #define LINUX_AF_IPX 4 #define LINUX_AF_APPLETALK 5 #define LINUX_AF_INET6 10 #define LINUX_AF_NETLINK 16 #define LINUX_NETLINK_ROUTE 0 #define LINUX_NETLINK_SOCK_DIAG 4 #define LINUX_NETLINK_NFLOG 5 #define LINUX_NETLINK_SELINUX 7 #define LINUX_NETLINK_AUDIT 9 #define LINUX_NETLINK_FIB_LOOKUP 10 #define LINUX_NETLINK_NETFILTER 12 #define LINUX_NETLINK_KOBJECT_UEVENT 15 /* * net device flags */ #define LINUX_IFF_UP 0x0001 #define LINUX_IFF_BROADCAST 0x0002 #define LINUX_IFF_DEBUG 0x0004 #define LINUX_IFF_LOOPBACK 0x0008 #define LINUX_IFF_POINTOPOINT 0x0010 #define LINUX_IFF_NOTRAILERS 0x0020 #define LINUX_IFF_RUNNING 0x0040 #define LINUX_IFF_NOARP 0x0080 #define LINUX_IFF_PROMISC 0x0100 #define LINUX_IFF_ALLMULTI 0x0200 #define LINUX_IFF_MASTER 0x0400 #define LINUX_IFF_SLAVE 0x0800 #define LINUX_IFF_MULTICAST 0x1000 #define LINUX_IFF_PORTSEL 0x2000 #define LINUX_IFF_AUTOMEDIA 0x4000 #define LINUX_IFF_DYNAMIC 0x8000 /* sigaltstack */ #define LINUX_SS_ONSTACK 1 #define LINUX_SS_DISABLE 2 int linux_to_bsd_sigaltstack(int lsa); int bsd_to_linux_sigaltstack(int bsa); /* sigset */ typedef struct { uint64_t __mask; } l_sigset_t; /* primitives to manipulate sigset_t */ #define LINUX_SIGEMPTYSET(set) (set).__mask = 0 #define LINUX_SIGISMEMBER(set, sig) (1UL & ((set).__mask >> _SIG_IDX(sig))) #define LINUX_SIGADDSET(set, sig) (set).__mask |= 1UL << _SIG_IDX(sig) void linux_to_bsd_sigset(l_sigset_t *, sigset_t *); void bsd_to_linux_sigset(sigset_t *, l_sigset_t *); /* signaling */ #define LINUX_SIGHUP 1 #define LINUX_SIGINT 2 #define LINUX_SIGQUIT 3 #define LINUX_SIGILL 4 #define LINUX_SIGTRAP 5 #define LINUX_SIGABRT 6 #define LINUX_SIGIOT LINUX_SIGABRT #define LINUX_SIGBUS 7 #define LINUX_SIGFPE 8 #define LINUX_SIGKILL 9 #define LINUX_SIGUSR1 10 #define LINUX_SIGSEGV 11 #define LINUX_SIGUSR2 12 #define LINUX_SIGPIPE 13 #define LINUX_SIGALRM 14 #define LINUX_SIGTERM 15 #define LINUX_SIGSTKFLT 16 #define LINUX_SIGCHLD 17 #define LINUX_SIGCONT 18 #define LINUX_SIGSTOP 19 #define LINUX_SIGTSTP 20 #define LINUX_SIGTTIN 21 #define LINUX_SIGTTOU 22 #define LINUX_SIGURG 23 #define LINUX_SIGXCPU 24 #define LINUX_SIGXFSZ 25 #define LINUX_SIGVTALRM 26 #define LINUX_SIGPROF 27 #define LINUX_SIGWINCH 28 #define LINUX_SIGIO 29 #define LINUX_SIGPOLL LINUX_SIGIO #define LINUX_SIGPWR 30 #define LINUX_SIGSYS 31 #define LINUX_SIGTBLSZ 31 #define LINUX_SIGRTMIN 32 #define LINUX_SIGRTMAX 64 #define LINUX_SIG_VALID(sig) ((sig) <= LINUX_SIGRTMAX && (sig) > 0) int linux_to_bsd_signal(int sig); int bsd_to_linux_signal(int sig); void linux_dev_shm_create(void); void linux_dev_shm_destroy(void); /* * mask=0 is not sensible for this application, so it will be taken to mean * a mask equivalent to the value. Otherwise, (word & mask) == value maps to * (word & ~mask) | value in a bitfield for the platform we're converting to. */ struct bsd_to_linux_bitmap { int bsd_mask; int bsd_value; int linux_mask; int linux_value; }; int bsd_to_linux_bits_(int value, struct bsd_to_linux_bitmap *bitmap, size_t mapcnt, int no_value); int linux_to_bsd_bits_(int value, struct bsd_to_linux_bitmap *bitmap, size_t mapcnt, int no_value); /* * These functions are used for simplification of BSD <-> Linux bit conversions. * Given `value`, a bit field, these functions will walk the given bitmap table * and set the appropriate bits for the target platform. If any bits were * successfully converted, then the return value is the equivalent of value * represented with the bit values appropriate for the target platform. * Otherwise, the value supplied as `no_value` is returned. */ #define bsd_to_linux_bits(_val, _bmap, _noval) \ bsd_to_linux_bits_((_val), (_bmap), nitems((_bmap)), (_noval)) #define linux_to_bsd_bits(_val, _bmap, _noval) \ linux_to_bsd_bits_((_val), (_bmap), nitems((_bmap)), (_noval)) /* * Easy mapping helpers. BITMAP_EASY_LINUX represents a single bit to be * translated, and the FreeBSD and Linux values are supplied. BITMAP_1t1_LINUX * is the extreme version of this, where not only is it a single bit, but the * name of the macro used to represent the Linux version of a bit literally has * LINUX_ prepended to the normal name. */ #define BITMAP_EASY_LINUX(_name, _linux_name) \ { \ .bsd_value = (_name), \ .linux_value = (_linux_name), \ } #define BITMAP_1t1_LINUX(_name) BITMAP_EASY_LINUX(_name, LINUX_##_name) int bsd_to_linux_errno(int error); void linux_check_errtbl(void); #define STATX_BASIC_STATS 0x07ff #define STATX_BTIME 0x0800 #define STATX_ALL 0x0fff #define STATX_ATTR_COMPRESSED 0x0004 #define STATX_ATTR_IMMUTABLE 0x0010 #define STATX_ATTR_APPEND 0x0020 #define STATX_ATTR_NODUMP 0x0040 #define STATX_ATTR_ENCRYPTED 0x0800 #define STATX_ATTR_AUTOMOUNT 0x1000 struct l_statx_timestamp { int64_t tv_sec; int32_t tv_nsec; int32_t __spare0; }; struct l_statx { uint32_t stx_mask; uint32_t stx_blksize; uint64_t stx_attributes; uint32_t stx_nlink; uint32_t stx_uid; uint32_t stx_gid; uint16_t stx_mode; uint16_t __spare0[1]; uint64_t stx_ino; uint64_t stx_size; uint64_t stx_blocks; uint64_t stx_attributes_mask; struct l_statx_timestamp stx_atime; struct l_statx_timestamp stx_btime; struct l_statx_timestamp stx_ctime; struct l_statx_timestamp stx_mtime; uint32_t stx_rdev_major; uint32_t stx_rdev_minor; uint32_t stx_dev_major; uint32_t stx_dev_minor; uint64_t stx_mnt_id; uint64_t __spare2[13]; }; +#define lower_32_bits(n) ((uint32_t)((n) & 0xffffffff)) + #endif /* _LINUX_MI_H_ */ diff --git a/sys/compat/linux/linux_emul.c b/sys/compat/linux/linux_emul.c index fbfe9080cac7..86688d50c8a8 100644 --- a/sys/compat/linux/linux_emul.c +++ b/sys/compat/linux/linux_emul.c @@ -1,383 +1,383 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 1994-1996 Søren Schmidt * Copyright (c) 2006 Roman Divacky * All rights reserved. * Copyright (c) 2013 Dmitry Chagin * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #if BYTE_ORDER == LITTLE_ENDIAN #define SHELLMAGIC 0x2123 /* #! */ #else #define SHELLMAGIC 0x2321 #endif /* * This returns reference to the thread emuldata entry (if found) * * Hold PROC_LOCK when referencing emuldata from other threads. */ struct linux_emuldata * em_find(struct thread *td) { struct linux_emuldata *em; em = td->td_emuldata; return (em); } /* * This returns reference to the proc pemuldata entry (if found) * * Hold PROC_LOCK when referencing proc pemuldata from other threads. * Hold LINUX_PEM_LOCK wher referencing pemuldata members. */ struct linux_pemuldata * pem_find(struct proc *p) { struct linux_pemuldata *pem; pem = p->p_emuldata; return (pem); } /* * Linux apps generally expect the soft open file limit to be set * to 1024, often iterating over all the file descriptors up to that * limit instead of using closefrom(2). Give them what they want, * unless there already is a resource limit in place. */ static void linux_set_default_openfiles(struct thread *td, struct proc *p) { struct rlimit rlim; int error; if (linux_default_openfiles < 0) return; PROC_LOCK(p); lim_rlimit_proc(p, RLIMIT_NOFILE, &rlim); PROC_UNLOCK(p); if (rlim.rlim_cur != rlim.rlim_max || rlim.rlim_cur <= linux_default_openfiles) return; rlim.rlim_cur = linux_default_openfiles; error = kern_proc_setrlimit(td, p, RLIMIT_NOFILE, &rlim); KASSERT(error == 0, ("kern_proc_setrlimit failed")); } /* * The default stack size limit in Linux is 8MB. */ static void linux_set_default_stacksize(struct thread *td, struct proc *p) { struct rlimit rlim; int error; if (linux_default_stacksize < 0) return; PROC_LOCK(p); lim_rlimit_proc(p, RLIMIT_STACK, &rlim); PROC_UNLOCK(p); if (rlim.rlim_cur != rlim.rlim_max || rlim.rlim_cur <= linux_default_stacksize) return; rlim.rlim_cur = linux_default_stacksize; error = kern_proc_setrlimit(td, p, RLIMIT_STACK, &rlim); KASSERT(error == 0, ("kern_proc_setrlimit failed")); } void -linux_proc_init(struct thread *td, struct thread *newtd, int flags) +linux_proc_init(struct thread *td, struct thread *newtd, bool init_thread) { struct linux_emuldata *em; struct linux_pemuldata *pem; struct proc *p; if (newtd != NULL) { p = newtd->td_proc; /* non-exec call */ em = malloc(sizeof(*em), M_TEMP, M_WAITOK | M_ZERO); - if (flags & LINUX_CLONE_THREAD) { + if (init_thread) { LINUX_CTR1(proc_init, "thread newtd(%d)", newtd->td_tid); em->em_tid = newtd->td_tid; } else { LINUX_CTR1(proc_init, "fork newtd(%d)", p->p_pid); em->em_tid = p->p_pid; pem = malloc(sizeof(*pem), M_LINUX, M_WAITOK | M_ZERO); sx_init(&pem->pem_sx, "lpemlk"); p->p_emuldata = pem; } newtd->td_emuldata = em; linux_set_default_openfiles(td, p); linux_set_default_stacksize(td, p); } else { p = td->td_proc; /* exec */ LINUX_CTR1(proc_init, "exec newtd(%d)", p->p_pid); /* lookup the old one */ em = em_find(td); KASSERT(em != NULL, ("proc_init: emuldata not found in exec case.\n")); em->em_tid = p->p_pid; em->flags = 0; em->robust_futexes = NULL; em->child_clear_tid = NULL; em->child_set_tid = NULL; pem = pem_find(p); KASSERT(pem != NULL, ("proc_exit: proc emuldata not found.\n")); pem->persona = 0; } } void linux_on_exit(struct proc *p) { struct linux_pemuldata *pem; struct thread *td = curthread; MPASS(SV_CURPROC_ABI() == SV_ABI_LINUX); LINUX_CTR3(proc_exit, "thread(%d) proc(%d) p %p", td->td_tid, p->p_pid, p); pem = pem_find(p); if (pem == NULL) return; (p->p_sysent->sv_thread_detach)(td); p->p_emuldata = NULL; sx_destroy(&pem->pem_sx); free(pem, M_LINUX); } /* * If a Linux binary is exec'ing something, try this image activator * first. We override standard shell script execution in order to * be able to modify the interpreter path. We only do this if a Linux * binary is doing the exec, so we do not create an EXEC module for it. */ int linux_exec_imgact_try(struct image_params *imgp) { const char *head = (const char *)imgp->image_header; char *rpath; int error = -1; /* * The interpreter for shell scripts run from a Linux binary needs * to be located in /compat/linux if possible in order to recursively * maintain Linux path emulation. */ if (((const short *)head)[0] == SHELLMAGIC) { /* * Run our normal shell image activator. If it succeeds attempt * to use the alternate path for the interpreter. If an * alternate path is found, use our stringspace to store it. */ if ((error = exec_shell_imgact(imgp)) == 0) { linux_emul_convpath(FIRST_THREAD_IN_PROC(imgp->proc), imgp->interpreter_name, UIO_SYSSPACE, &rpath, 0, AT_FDCWD); if (rpath != NULL) imgp->args->fname_buf = imgp->interpreter_name = rpath; } } return (error); } int linux_common_execve(struct thread *td, struct image_args *eargs) { struct linux_pemuldata *pem; struct vmspace *oldvmspace; struct linux_emuldata *em; struct proc *p; int error; p = td->td_proc; error = pre_execve(td, &oldvmspace); if (error != 0) return (error); error = kern_execve(td, eargs, NULL, oldvmspace); post_execve(td, error, oldvmspace); if (error != EJUSTRETURN) return (error); /* * In a case of transition from Linux binary execing to * FreeBSD binary we destroy Linux emuldata thread & proc entries. */ if (SV_CURPROC_ABI() != SV_ABI_LINUX) { PROC_LOCK(p); em = em_find(td); KASSERT(em != NULL, ("proc_exec: thread emuldata not found.\n")); td->td_emuldata = NULL; pem = pem_find(p); KASSERT(pem != NULL, ("proc_exec: proc pemuldata not found.\n")); p->p_emuldata = NULL; PROC_UNLOCK(p); free(em, M_TEMP); free(pem, M_LINUX); } return (EJUSTRETURN); } void linux_on_exec(struct proc *p, struct image_params *imgp) { struct thread *td; struct thread *othertd; #if defined(__amd64__) struct linux_pemuldata *pem; #endif td = curthread; MPASS((imgp->sysent->sv_flags & SV_ABI_MASK) == SV_ABI_LINUX); /* * When execing to Linux binary, we create Linux emuldata * thread entry. */ if (SV_PROC_ABI(p) == SV_ABI_LINUX) { /* * Process already was under Linuxolator * before exec. Update emuldata to reflect * single-threaded cleaned state after exec. */ - linux_proc_init(td, NULL, 0); + linux_proc_init(td, NULL, false); } else { /* * We are switching the process to Linux emulator. */ - linux_proc_init(td, td, 0); + linux_proc_init(td, td, false); /* * Create a transient td_emuldata for all suspended * threads, so that p->p_sysent->sv_thread_detach() == * linux_thread_detach() can find expected but unused * emuldata. */ FOREACH_THREAD_IN_PROC(td->td_proc, othertd) { if (othertd == td) continue; - linux_proc_init(td, othertd, LINUX_CLONE_THREAD); + linux_proc_init(td, othertd, true); } } #if defined(__amd64__) /* * An IA32 executable which has executable stack will have the * READ_IMPLIES_EXEC personality flag set automatically. */ if (SV_PROC_FLAG(td->td_proc, SV_ILP32) && imgp->stack_prot & VM_PROT_EXECUTE) { pem = pem_find(p); pem->persona |= LINUX_READ_IMPLIES_EXEC; } #endif } void linux_thread_dtor(struct thread *td) { struct linux_emuldata *em; em = em_find(td); if (em == NULL) return; td->td_emuldata = NULL; LINUX_CTR1(thread_dtor, "thread(%d)", em->em_tid); free(em, M_TEMP); } void linux_schedtail(struct thread *td) { struct linux_emuldata *em; struct proc *p; int error = 0; int *child_set_tid; p = td->td_proc; em = em_find(td); KASSERT(em != NULL, ("linux_schedtail: thread emuldata not found.\n")); child_set_tid = em->child_set_tid; if (child_set_tid != NULL) { error = copyout(&em->em_tid, child_set_tid, sizeof(em->em_tid)); LINUX_CTR4(schedtail, "thread(%d) %p stored %d error %d", td->td_tid, child_set_tid, em->em_tid, error); } else LINUX_CTR1(schedtail, "thread(%d)", em->em_tid); } diff --git a/sys/compat/linux/linux_emul.h b/sys/compat/linux/linux_emul.h index ed5332b2c96c..70646cc93847 100644 --- a/sys/compat/linux/linux_emul.h +++ b/sys/compat/linux/linux_emul.h @@ -1,88 +1,87 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 2006 Roman Divacky * All rights reserved. * Copyright (c) 2013 Dmitry Chagin * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $FreeBSD$ */ #ifndef _LINUX_EMUL_H_ #define _LINUX_EMUL_H_ struct image_params; struct note_info_list; /* Linux core notes are labeled "CORE" */ #define LINUX_ABI_VENDOR "CORE" /* * modeled after similar structure in NetBSD * this will be extended as we need more functionality */ struct linux_emuldata { int *child_set_tid; /* in clone(): Child's TID to set on clone */ int *child_clear_tid;/* in clone(): Child's TID to clear on exit */ int flags; /* thread emuldata flags */ int em_tid; /* thread id */ struct linux_robust_list_head *robust_futexes; }; struct linux_emuldata *em_find(struct thread *); int linux_exec_imgact_try(struct image_params *); -void linux_proc_init(struct thread *, struct thread *, int); +void linux_proc_init(struct thread *, struct thread *, bool); void linux_on_exit(struct proc *); void linux_schedtail(struct thread *); void linux_on_exec(struct proc *, struct image_params *); void linux_thread_dtor(struct thread *); -void linux_thread_detach(struct thread *); int linux_common_execve(struct thread *, struct image_args *); void linux32_prepare_notes(struct thread *, struct note_info_list *, size_t *); void linux64_prepare_notes(struct thread *, struct note_info_list *, size_t *); /* process emuldata flags */ #define LINUX_XDEPR_REQUEUEOP 0x00000001 /* uses deprecated futex REQUEUE op*/ #define LINUX_XUNSUP_EPOLL 0x00000002 /* unsupported epoll events */ #define LINUX_XUNSUP_FUTEXPIOP 0x00000004 /* uses unsupported pi futex */ struct linux_pemuldata { uint32_t flags; /* process emuldata flags */ struct sx pem_sx; /* lock for this struct */ uint32_t persona; /* process execution domain */ uint32_t ptrace_flags; /* used by ptrace(2) */ }; #define LINUX_PEM_XLOCK(p) sx_xlock(&(p)->pem_sx) #define LINUX_PEM_XUNLOCK(p) sx_xunlock(&(p)->pem_sx) #define LINUX_PEM_SLOCK(p) sx_slock(&(p)->pem_sx) #define LINUX_PEM_SUNLOCK(p) sx_sunlock(&(p)->pem_sx) struct linux_pemuldata *pem_find(struct proc *); #endif /* !_LINUX_EMUL_H_ */ diff --git a/sys/compat/linux/linux_fork.c b/sys/compat/linux/linux_fork.c index 955e9e512bd1..97f5b7d89de4 100644 --- a/sys/compat/linux/linux_fork.c +++ b/sys/compat/linux/linux_fork.c @@ -1,446 +1,457 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 2004 Tim J. Robbins * Copyright (c) 2002 Doug Rabson * Copyright (c) 2000 Marcel Moolenaar * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer * in this position and unchanged. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include "opt_compat.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef COMPAT_LINUX32 #include #include #else #include #include #endif +#include #include +#include #include #include #include #ifdef LINUX_LEGACY_SYSCALLS int linux_fork(struct thread *td, struct linux_fork_args *args) { struct fork_req fr; int error; struct proc *p2; struct thread *td2; bzero(&fr, sizeof(fr)); fr.fr_flags = RFFDG | RFPROC | RFSTOPPED; fr.fr_procp = &p2; if ((error = fork1(td, &fr)) != 0) return (error); td2 = FIRST_THREAD_IN_PROC(p2); - linux_proc_init(td, td2, 0); + linux_proc_init(td, td2, false); td->td_retval[0] = p2->p_pid; /* * Make this runnable after we are finished with it. */ thread_lock(td2); TD_SET_CAN_RUN(td2); sched_add(td2, SRQ_BORING); return (0); } int linux_vfork(struct thread *td, struct linux_vfork_args *args) { struct fork_req fr; int error; struct proc *p2; struct thread *td2; bzero(&fr, sizeof(fr)); fr.fr_flags = RFFDG | RFPROC | RFMEM | RFPPWAIT | RFSTOPPED; fr.fr_procp = &p2; if ((error = fork1(td, &fr)) != 0) return (error); td2 = FIRST_THREAD_IN_PROC(p2); - linux_proc_init(td, td2, 0); + linux_proc_init(td, td2, false); td->td_retval[0] = p2->p_pid; /* * Make this runnable after we are finished with it. */ thread_lock(td2); TD_SET_CAN_RUN(td2); sched_add(td2, SRQ_BORING); return (0); } #endif static int -linux_clone_proc(struct thread *td, struct linux_clone_args *args) +linux_clone_proc(struct thread *td, struct l_clone_args *args) { struct fork_req fr; int error, ff = RFPROC | RFSTOPPED, f2; struct proc *p2; struct thread *td2; int exit_signal; struct linux_emuldata *em; f2 = 0; - exit_signal = args->flags & 0x000000ff; - if (LINUX_SIG_VALID(exit_signal)) { - exit_signal = linux_to_bsd_signal(exit_signal); - } else if (exit_signal != 0) + if (LINUX_SIG_VALID(args->exit_signal)) { + exit_signal = linux_to_bsd_signal(args->exit_signal); + } else if (args->exit_signal != 0) return (EINVAL); + else + exit_signal = 0; if (args->flags & LINUX_CLONE_VM) ff |= RFMEM; if (args->flags & LINUX_CLONE_SIGHAND) ff |= RFSIGSHARE; if (args->flags & LINUX_CLONE_FILES) { if (!(args->flags & LINUX_CLONE_FS)) f2 |= FR2_SHARE_PATHS; } else { ff |= RFFDG; if (args->flags & LINUX_CLONE_FS) f2 |= FR2_SHARE_PATHS; } if (args->flags & LINUX_CLONE_PARENT_SETTID) - if (args->parent_tidptr == NULL) + if (args->parent_tid == NULL) return (EINVAL); if (args->flags & LINUX_CLONE_VFORK) ff |= RFPPWAIT; bzero(&fr, sizeof(fr)); fr.fr_flags = ff; fr.fr_flags2 = f2; fr.fr_procp = &p2; error = fork1(td, &fr); if (error) return (error); td2 = FIRST_THREAD_IN_PROC(p2); /* create the emuldata */ - linux_proc_init(td, td2, args->flags); + linux_proc_init(td, td2, false); em = em_find(td2); KASSERT(em != NULL, ("clone_proc: emuldata not found.\n")); if (args->flags & LINUX_CLONE_CHILD_SETTID) - em->child_set_tid = args->child_tidptr; + em->child_set_tid = args->child_tid; else em->child_set_tid = NULL; if (args->flags & LINUX_CLONE_CHILD_CLEARTID) - em->child_clear_tid = args->child_tidptr; + em->child_clear_tid = args->child_tid; else em->child_clear_tid = NULL; if (args->flags & LINUX_CLONE_PARENT_SETTID) { - error = copyout(&p2->p_pid, args->parent_tidptr, + error = copyout(&p2->p_pid, args->parent_tid, sizeof(p2->p_pid)); if (error) linux_msg(td, "copyout p_pid failed!"); } PROC_LOCK(p2); p2->p_sigparent = exit_signal; PROC_UNLOCK(p2); /* * In a case of stack = NULL, we are supposed to COW calling process * stack. This is what normal fork() does, so we just keep tf_rsp arg * intact. */ linux_set_upcall(td2, args->stack); if (args->flags & LINUX_CLONE_SETTLS) linux_set_cloned_tls(td2, PTRIN(args->tls)); /* * If CLONE_PARENT is set, then the parent of the new process will be * the same as that of the calling process. */ if (args->flags & LINUX_CLONE_PARENT) { sx_xlock(&proctree_lock); PROC_LOCK(p2); proc_reparent(p2, td->td_proc->p_pptr, true); PROC_UNLOCK(p2); sx_xunlock(&proctree_lock); } /* * Make this runnable after we are finished with it. */ thread_lock(td2); TD_SET_CAN_RUN(td2); sched_add(td2, SRQ_BORING); td->td_retval[0] = p2->p_pid; return (0); } static int -linux_clone_thread(struct thread *td, struct linux_clone_args *args) +linux_clone_thread(struct thread *td, struct l_clone_args *args) { struct linux_emuldata *em; struct thread *newtd; struct proc *p; int error; LINUX_CTR4(clone_thread, "thread(%d) flags %x ptid %p ctid %p", td->td_tid, (unsigned)args->flags, - args->parent_tidptr, args->child_tidptr); + args->parent_tid, args->child_tid); if ((args->flags & LINUX_CLONE_PARENT) != 0) return (EINVAL); if (args->flags & LINUX_CLONE_PARENT_SETTID) - if (args->parent_tidptr == NULL) + if (args->parent_tid == NULL) return (EINVAL); /* Threads should be created with own stack */ if (PTRIN(args->stack) == NULL) return (EINVAL); p = td->td_proc; #ifdef RACCT if (racct_enable) { PROC_LOCK(p); error = racct_add(p, RACCT_NTHR, 1); PROC_UNLOCK(p); if (error != 0) return (EPROCLIM); } #endif /* Initialize our td */ error = kern_thr_alloc(p, 0, &newtd); if (error) goto fail; cpu_copy_thread(newtd, td); bzero(&newtd->td_startzero, __rangeof(struct thread, td_startzero, td_endzero)); bcopy(&td->td_startcopy, &newtd->td_startcopy, __rangeof(struct thread, td_startcopy, td_endcopy)); newtd->td_proc = p; thread_cow_get(newtd, td); /* create the emuldata */ - linux_proc_init(td, newtd, args->flags); + linux_proc_init(td, newtd, true); em = em_find(newtd); KASSERT(em != NULL, ("clone_thread: emuldata not found.\n")); if (args->flags & LINUX_CLONE_SETTLS) linux_set_cloned_tls(newtd, PTRIN(args->tls)); if (args->flags & LINUX_CLONE_CHILD_SETTID) - em->child_set_tid = args->child_tidptr; + em->child_set_tid = args->child_tid; else em->child_set_tid = NULL; if (args->flags & LINUX_CLONE_CHILD_CLEARTID) - em->child_clear_tid = args->child_tidptr; + em->child_clear_tid = args->child_tid; else em->child_clear_tid = NULL; cpu_thread_clean(newtd); linux_set_upcall(newtd, args->stack); PROC_LOCK(p); p->p_flag |= P_HADTHREADS; thread_link(newtd, p); bcopy(p->p_comm, newtd->td_name, sizeof(newtd->td_name)); thread_lock(td); /* let the scheduler know about these things. */ sched_fork_thread(td, newtd); thread_unlock(td); if (P_SHOULDSTOP(p)) newtd->td_flags |= TDF_ASTPENDING | TDF_NEEDSUSPCHK; if (p->p_ptevents & PTRACE_LWP) newtd->td_dbgflags |= TDB_BORN; PROC_UNLOCK(p); tidhash_add(newtd); LINUX_CTR2(clone_thread, "thread(%d) successful clone to %d", td->td_tid, newtd->td_tid); if (args->flags & LINUX_CLONE_PARENT_SETTID) { - error = copyout(&newtd->td_tid, args->parent_tidptr, + error = copyout(&newtd->td_tid, args->parent_tid, sizeof(newtd->td_tid)); if (error) linux_msg(td, "clone_thread: copyout td_tid failed!"); } /* * Make this runnable after we are finished with it. */ thread_lock(newtd); TD_SET_CAN_RUN(newtd); sched_add(newtd, SRQ_BORING); td->td_retval[0] = newtd->td_tid; return (0); fail: #ifdef RACCT if (racct_enable) { PROC_LOCK(p); racct_sub(p, RACCT_NTHR, 1); PROC_UNLOCK(p); } #endif return (error); } int linux_clone(struct thread *td, struct linux_clone_args *args) { + struct l_clone_args ca = { + .flags = (lower_32_bits(args->flags) & ~LINUX_CSIGNAL), + .child_tid = args->child_tidptr, + .parent_tid = args->parent_tidptr, + .exit_signal = (lower_32_bits(args->flags) & LINUX_CSIGNAL), + .stack = args->stack, + .tls = args->tls, + }; if (args->flags & LINUX_CLONE_THREAD) - return (linux_clone_thread(td, args)); + return (linux_clone_thread(td, &ca)); else - return (linux_clone_proc(td, args)); + return (linux_clone_proc(td, &ca)); } int linux_exit(struct thread *td, struct linux_exit_args *args) { struct linux_emuldata *em; em = em_find(td); KASSERT(em != NULL, ("exit: emuldata not found.\n")); LINUX_CTR2(exit, "thread(%d) (%d)", em->em_tid, args->rval); linux_thread_detach(td); /* * XXX. When the last two threads of a process * exit via pthread_exit() try thr_exit() first. */ kern_thr_exit(td); exit1(td, args->rval, 0); /* NOTREACHED */ } int linux_set_tid_address(struct thread *td, struct linux_set_tid_address_args *args) { struct linux_emuldata *em; em = em_find(td); KASSERT(em != NULL, ("set_tid_address: emuldata not found.\n")); em->child_clear_tid = args->tidptr; td->td_retval[0] = em->em_tid; LINUX_CTR3(set_tid_address, "tidptr(%d) %p, returns %d", em->em_tid, args->tidptr, td->td_retval[0]); return (0); } void linux_thread_detach(struct thread *td) { struct linux_emuldata *em; int *child_clear_tid; int error; em = em_find(td); KASSERT(em != NULL, ("thread_detach: emuldata not found.\n")); LINUX_CTR1(thread_detach, "thread(%d)", em->em_tid); release_futexes(td, em); child_clear_tid = em->child_clear_tid; if (child_clear_tid != NULL) { LINUX_CTR2(thread_detach, "thread(%d) %p", em->em_tid, child_clear_tid); error = suword32(child_clear_tid, 0); if (error != 0) return; error = futex_wake(td, child_clear_tid, 1, false); /* * this cannot happen at the moment and if this happens it * probably means there is a user space bug */ if (error != 0) linux_msg(td, "futex stuff in thread_detach failed."); } /* * Do not rely on the robust list which is maintained by userspace, * cleanup remaining pi (if any) after release_futexes anyway. */ umtx_thread_exit(td); } diff --git a/sys/compat/linux/linux_fork.h b/sys/compat/linux/linux_fork.h new file mode 100644 index 000000000000..04dfb8ac8a70 --- /dev/null +++ b/sys/compat/linux/linux_fork.h @@ -0,0 +1,92 @@ +/*- + * Copyright (c) 2021 Dmitry Chagin + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + * + * $FreeBSD$ + */ + +#ifndef _LINUX_FORK_H_ +#define _LINUX_FORK_H_ + +#define LINUX_CLONE_VM 0x00000100 +#define LINUX_CLONE_FS 0x00000200 +#define LINUX_CLONE_FILES 0x00000400 +#define LINUX_CLONE_SIGHAND 0x00000800 +#define LINUX_CLONE_PID 0x00001000 /* No longer exist in Linux */ +#define LINUX_CLONE_PTRACE 0x00002000 +#define LINUX_CLONE_VFORK 0x00004000 +#define LINUX_CLONE_PARENT 0x00008000 +#define LINUX_CLONE_THREAD 0x00010000 +#define LINUX_CLONE_NEWNS 0x00020000 /* New mount NS */ +#define LINUX_CLONE_SYSVSEM 0x00040000 +#define LINUX_CLONE_SETTLS 0x00080000 +#define LINUX_CLONE_PARENT_SETTID 0x00100000 +#define LINUX_CLONE_CHILD_CLEARTID 0x00200000 +#define LINUX_CLONE_DETACHED 0x00400000 /* Unused */ +#define LINUX_CLONE_UNTRACED 0x00800000 +#define LINUX_CLONE_CHILD_SETTID 0x01000000 +#define LINUX_CLONE_NEWCGROUP 0x02000000 /* New cgroup NS */ +#define LINUX_CLONE_NEWUTS 0x04000000 +#define LINUX_CLONE_NEWIPC 0x08000000 +#define LINUX_CLONE_NEWUSER 0x10000000 +#define LINUX_CLONE_NEWPID 0x20000000 +#define LINUX_CLONE_NEWNET 0x40000000 +#define LINUX_CLONE_IO 0x80000000 + +#define LINUX_CSIGNAL 0x000000ff + +/* + * User-space clone3 args layout. + */ +struct l_user_clone_args { + uint64_t flags; + uint64_t pidfd; + uint64_t child_tid; + uint64_t parent_tid; + uint64_t exit_signal; + uint64_t stack; + uint64_t stack_size; + uint64_t tls; + uint64_t set_tid; + uint64_t set_tid_size; + uint64_t cgroup; +}; + +/* + * Kernel clone3 args layout. + */ +struct l_clone_args { + uint64_t flags; + l_int *child_tid; + l_int *parent_tid; + l_int exit_signal; + l_ulong stack; + l_ulong stack_size; + l_ulong tls; +}; + +int linux_set_upcall(struct thread *, register_t); +int linux_set_cloned_tls(struct thread *, void *); +void linux_thread_detach(struct thread *); + +#endif /* _LINUX_FORK_H_ */ diff --git a/sys/compat/linux/linux_misc.h b/sys/compat/linux/linux_misc.h index dc17ed430014..80f6b8a58e81 100644 --- a/sys/compat/linux/linux_misc.h +++ b/sys/compat/linux/linux_misc.h @@ -1,187 +1,160 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 2006 Roman Divacky * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $FreeBSD$ */ #ifndef _LINUX_MISC_H_ #define _LINUX_MISC_H_ #include /* bits per mask */ #define LINUX_NFDBITS sizeof(l_fd_mask) * 8 /* * Miscellaneous */ #define LINUX_NAME_MAX 255 #define LINUX_MAX_UTSNAME 65 #define LINUX_CTL_MAXNAME 10 /* defines for prctl */ #define LINUX_PR_SET_PDEATHSIG 1 /* Second arg is a signal. */ #define LINUX_PR_GET_PDEATHSIG 2 /* * Second arg is a ptr to return the * signal. */ #define LINUX_PR_GET_DUMPABLE 3 #define LINUX_PR_SET_DUMPABLE 4 #define LINUX_PR_GET_KEEPCAPS 7 /* Get drop capabilities on setuid */ #define LINUX_PR_SET_KEEPCAPS 8 /* Set drop capabilities on setuid */ #define LINUX_PR_SET_NAME 15 /* Set process name. */ #define LINUX_PR_GET_NAME 16 /* Get process name. */ #define LINUX_PR_GET_SECCOMP 21 #define LINUX_PR_SET_SECCOMP 22 #define LINUX_PR_CAPBSET_READ 23 #define LINUX_PR_SET_NO_NEW_PRIVS 38 #define LINUX_PR_SET_PTRACER 1499557217 #define LINUX_MAX_COMM_LEN 16 /* Maximum length of the process name. */ /* For GET/SET DUMPABLE */ #define LINUX_SUID_DUMP_DISABLE 0 /* Don't coredump setuid processes. */ #define LINUX_SUID_DUMP_USER 1 /* Dump as user of process. */ #define LINUX_SUID_DUMP_ROOT 2 /* Dump as root. */ #define LINUX_MREMAP_MAYMOVE 1 #define LINUX_MREMAP_FIXED 2 #define LINUX_PATH_MAX 4096 /* * Non-standard aux entry types used in Linux ELF binaries. */ #define LINUX_AT_PLATFORM 15 /* String identifying CPU */ #define LINUX_AT_HWCAP 16 /* CPU capabilities */ #define LINUX_AT_CLKTCK 17 /* frequency at which times() increments */ #define LINUX_AT_SECURE 23 /* secure mode boolean */ #define LINUX_AT_BASE_PLATFORM 24 /* string identifying real platform, may * differ from AT_PLATFORM. */ #define LINUX_AT_RANDOM 25 /* address of random bytes */ #define LINUX_AT_HWCAP2 26 /* CPU capabilities, second part */ #define LINUX_AT_EXECFN 31 /* filename of program */ #define LINUX_AT_SYSINFO 32 /* vsyscall */ #define LINUX_AT_SYSINFO_EHDR 33 /* vdso header */ #define LINUX_AT_RANDOM_LEN 16 /* size of random bytes */ /* Linux sets the i387 to extended precision. */ #if defined(__i386__) || defined(__amd64__) #define __LINUX_NPXCW__ 0x37f #endif -#define LINUX_CLONE_VM 0x00000100 -#define LINUX_CLONE_FS 0x00000200 -#define LINUX_CLONE_FILES 0x00000400 -#define LINUX_CLONE_SIGHAND 0x00000800 -#define LINUX_CLONE_PID 0x00001000 /* No longer exist in Linux */ -#define LINUX_CLONE_PTRACE 0x00002000 -#define LINUX_CLONE_VFORK 0x00004000 -#define LINUX_CLONE_PARENT 0x00008000 -#define LINUX_CLONE_THREAD 0x00010000 -#define LINUX_CLONE_NEWNS 0x00020000 /* New mount NS */ -#define LINUX_CLONE_SYSVSEM 0x00040000 -#define LINUX_CLONE_SETTLS 0x00080000 -#define LINUX_CLONE_PARENT_SETTID 0x00100000 -#define LINUX_CLONE_CHILD_CLEARTID 0x00200000 -#define LINUX_CLONE_DETACHED 0x00400000 /* Unused */ -#define LINUX_CLONE_UNTRACED 0x00800000 -#define LINUX_CLONE_CHILD_SETTID 0x01000000 -#define LINUX_CLONE_NEWCGROUP 0x02000000 /* New cgroup NS */ -#define LINUX_CLONE_NEWUTS 0x04000000 -#define LINUX_CLONE_NEWIPC 0x08000000 -#define LINUX_CLONE_NEWUSER 0x10000000 -#define LINUX_CLONE_NEWPID 0x20000000 -#define LINUX_CLONE_NEWNET 0x40000000 -#define LINUX_CLONE_IO 0x80000000 - /* Scheduling policies */ #define LINUX_SCHED_OTHER 0 #define LINUX_SCHED_FIFO 1 #define LINUX_SCHED_RR 2 #define LINUX_MAX_RT_PRIO 100 struct l_new_utsname { char sysname[LINUX_MAX_UTSNAME]; char nodename[LINUX_MAX_UTSNAME]; char release[LINUX_MAX_UTSNAME]; char version[LINUX_MAX_UTSNAME]; char machine[LINUX_MAX_UTSNAME]; char domainname[LINUX_MAX_UTSNAME]; }; #define LINUX_UTIME_NOW 0x3FFFFFFF #define LINUX_UTIME_OMIT 0x3FFFFFFE extern int stclohz; #define LINUX_WNOHANG 0x00000001 #define LINUX_WUNTRACED 0x00000002 #define LINUX_WSTOPPED LINUX_WUNTRACED #define LINUX_WEXITED 0x00000004 #define LINUX_WCONTINUED 0x00000008 #define LINUX_WNOWAIT 0x01000000 #define __WNOTHREAD 0x20000000 #define __WALL 0x40000000 #define __WCLONE 0x80000000 /* Linux waitid idtype */ #define LINUX_P_ALL 0 #define LINUX_P_PID 1 #define LINUX_P_PGID 2 #define LINUX_RLIMIT_LOCKS 10 #define LINUX_RLIMIT_SIGPENDING 11 #define LINUX_RLIMIT_MSGQUEUE 12 #define LINUX_RLIMIT_NICE 13 #define LINUX_RLIMIT_RTPRIO 14 #define LINUX_RLIMIT_RTTIME 15 #define LINUX_RLIM_INFINITY (~0UL) /* Linux getrandom flags */ #define LINUX_GRND_NONBLOCK 0x0001 #define LINUX_GRND_RANDOM 0x0002 /* Linux syslog flags */ #define LINUX_SYSLOG_ACTION_READ_ALL 3 #if defined(__amd64__) && !defined(COMPAT_LINUX32) int linux_ptrace_status(struct thread *td, int pid, int status); #endif void linux_to_bsd_waitopts(int options, int *bsdopts); -int linux_set_upcall(struct thread *td, register_t stack); -int linux_set_cloned_tls(struct thread *td, void *desc); struct thread *linux_tdfind(struct thread *, lwpid_t, pid_t); #endif /* _LINUX_MISC_H_ */ diff --git a/sys/i386/linux/linux_machdep.c b/sys/i386/linux/linux_machdep.c index 1a51b7954063..c81bf1b4e2e5 100644 --- a/sys/i386/linux/linux_machdep.c +++ b/sys/i386/linux/linux_machdep.c @@ -1,695 +1,696 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 2000 Marcel Moolenaar * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include +#include #include #include #include #include #include #include /* needed for pcb definition in linux_set_thread_area */ #include "opt_posix.h" struct l_descriptor { l_uint entry_number; l_ulong base_addr; l_uint limit; l_uint seg_32bit:1; l_uint contents:2; l_uint read_exec_only:1; l_uint limit_in_pages:1; l_uint seg_not_present:1; l_uint useable:1; }; struct l_old_select_argv { l_int nfds; l_fd_set *readfds; l_fd_set *writefds; l_fd_set *exceptfds; struct l_timeval *timeout; }; int linux_execve(struct thread *td, struct linux_execve_args *args) { struct image_args eargs; char *newpath; int error; if (!LUSECONVPATH(td)) { error = exec_copyin_args(&eargs, args->path, UIO_USERSPACE, args->argp, args->envp); } else { LCONVPATHEXIST(td, args->path, &newpath); error = exec_copyin_args(&eargs, newpath, UIO_SYSSPACE, args->argp, args->envp); LFREEPATH(newpath); } if (error == 0) error = linux_common_execve(td, &eargs); AUDIT_SYSCALL_EXIT(error == EJUSTRETURN ? 0 : error, td); return (error); } struct l_ipc_kludge { struct l_msgbuf *msgp; l_long msgtyp; }; int linux_ipc(struct thread *td, struct linux_ipc_args *args) { switch (args->what & 0xFFFF) { case LINUX_SEMOP: { struct linux_semop_args a; a.semid = args->arg1; a.tsops = PTRIN(args->ptr); a.nsops = args->arg2; return (linux_semop(td, &a)); } case LINUX_SEMGET: { struct linux_semget_args a; a.key = args->arg1; a.nsems = args->arg2; a.semflg = args->arg3; return (linux_semget(td, &a)); } case LINUX_SEMCTL: { struct linux_semctl_args a; int error; a.semid = args->arg1; a.semnum = args->arg2; a.cmd = args->arg3; error = copyin(PTRIN(args->ptr), &a.arg, sizeof(a.arg)); if (error) return (error); return (linux_semctl(td, &a)); } case LINUX_MSGSND: { struct linux_msgsnd_args a; a.msqid = args->arg1; a.msgp = PTRIN(args->ptr); a.msgsz = args->arg2; a.msgflg = args->arg3; return (linux_msgsnd(td, &a)); } case LINUX_MSGRCV: { struct linux_msgrcv_args a; a.msqid = args->arg1; a.msgsz = args->arg2; a.msgflg = args->arg3; if ((args->what >> 16) == 0) { struct l_ipc_kludge tmp; int error; if (args->ptr == 0) return (EINVAL); error = copyin(PTRIN(args->ptr), &tmp, sizeof(tmp)); if (error) return (error); a.msgp = PTRIN(tmp.msgp); a.msgtyp = tmp.msgtyp; } else { a.msgp = PTRIN(args->ptr); a.msgtyp = args->arg5; } return (linux_msgrcv(td, &a)); } case LINUX_MSGGET: { struct linux_msgget_args a; a.key = args->arg1; a.msgflg = args->arg2; return (linux_msgget(td, &a)); } case LINUX_MSGCTL: { struct linux_msgctl_args a; a.msqid = args->arg1; a.cmd = args->arg2; a.buf = PTRIN(args->ptr); return (linux_msgctl(td, &a)); } case LINUX_SHMAT: { struct linux_shmat_args a; l_uintptr_t addr; int error; a.shmid = args->arg1; a.shmaddr = PTRIN(args->ptr); a.shmflg = args->arg2; error = linux_shmat(td, &a); if (error != 0) return (error); addr = td->td_retval[0]; error = copyout(&addr, PTRIN(args->arg3), sizeof(addr)); td->td_retval[0] = 0; return (error); } case LINUX_SHMDT: { struct linux_shmdt_args a; a.shmaddr = PTRIN(args->ptr); return (linux_shmdt(td, &a)); } case LINUX_SHMGET: { struct linux_shmget_args a; a.key = args->arg1; a.size = args->arg2; a.shmflg = args->arg3; return (linux_shmget(td, &a)); } case LINUX_SHMCTL: { struct linux_shmctl_args a; a.shmid = args->arg1; a.cmd = args->arg2; a.buf = PTRIN(args->ptr); return (linux_shmctl(td, &a)); } default: break; } return (EINVAL); } int linux_old_select(struct thread *td, struct linux_old_select_args *args) { struct l_old_select_argv linux_args; struct linux_select_args newsel; int error; error = copyin(args->ptr, &linux_args, sizeof(linux_args)); if (error) return (error); newsel.nfds = linux_args.nfds; newsel.readfds = linux_args.readfds; newsel.writefds = linux_args.writefds; newsel.exceptfds = linux_args.exceptfds; newsel.timeout = linux_args.timeout; return (linux_select(td, &newsel)); } int linux_set_cloned_tls(struct thread *td, void *desc) { struct segment_descriptor sd; struct l_user_desc info; int idx, error; int a[2]; error = copyin(desc, &info, sizeof(struct l_user_desc)); if (error) { linux_msg(td, "set_cloned_tls copyin failed!"); } else { idx = info.entry_number; /* * looks like we're getting the idx we returned * in the set_thread_area() syscall */ if (idx != 6 && idx != 3) { linux_msg(td, "set_cloned_tls resetting idx!"); idx = 3; } /* this doesnt happen in practice */ if (idx == 6) { /* we might copy out the entry_number as 3 */ info.entry_number = 3; error = copyout(&info, desc, sizeof(struct l_user_desc)); if (error) linux_msg(td, "set_cloned_tls copyout failed!"); } a[0] = LINUX_LDT_entry_a(&info); a[1] = LINUX_LDT_entry_b(&info); memcpy(&sd, &a, sizeof(a)); /* set %gs */ td->td_pcb->pcb_gsd = sd; td->td_pcb->pcb_gs = GSEL(GUGS_SEL, SEL_UPL); } return (error); } int linux_set_upcall(struct thread *td, register_t stack) { if (stack) td->td_frame->tf_esp = stack; /* * The newly created Linux thread returns * to the user space by the same path that a parent do. */ td->td_frame->tf_eax = 0; return (0); } int linux_mmap2(struct thread *td, struct linux_mmap2_args *args) { return (linux_mmap_common(td, args->addr, args->len, args->prot, args->flags, args->fd, (uint64_t)(uint32_t)args->pgoff * PAGE_SIZE)); } int linux_mmap(struct thread *td, struct linux_mmap_args *args) { int error; struct l_mmap_argv linux_args; error = copyin(args->ptr, &linux_args, sizeof(linux_args)); if (error) return (error); return (linux_mmap_common(td, linux_args.addr, linux_args.len, linux_args.prot, linux_args.flags, linux_args.fd, (uint32_t)linux_args.pgoff)); } int linux_mprotect(struct thread *td, struct linux_mprotect_args *uap) { return (linux_mprotect_common(td, PTROUT(uap->addr), uap->len, uap->prot)); } int linux_madvise(struct thread *td, struct linux_madvise_args *uap) { return (linux_madvise_common(td, PTROUT(uap->addr), uap->len, uap->behav)); } int linux_ioperm(struct thread *td, struct linux_ioperm_args *args) { int error; struct i386_ioperm_args iia; iia.start = args->start; iia.length = args->length; iia.enable = args->enable; error = i386_set_ioperm(td, &iia); return (error); } int linux_iopl(struct thread *td, struct linux_iopl_args *args) { int error; if (args->level < 0 || args->level > 3) return (EINVAL); if ((error = priv_check(td, PRIV_IO)) != 0) return (error); if ((error = securelevel_gt(td->td_ucred, 0)) != 0) return (error); td->td_frame->tf_eflags = (td->td_frame->tf_eflags & ~PSL_IOPL) | (args->level * (PSL_IOPL / 3)); return (0); } int linux_modify_ldt(struct thread *td, struct linux_modify_ldt_args *uap) { int error; struct i386_ldt_args ldt; struct l_descriptor ld; union descriptor desc; int size, written; switch (uap->func) { case 0x00: /* read_ldt */ ldt.start = 0; ldt.descs = uap->ptr; ldt.num = uap->bytecount / sizeof(union descriptor); error = i386_get_ldt(td, &ldt); td->td_retval[0] *= sizeof(union descriptor); break; case 0x02: /* read_default_ldt = 0 */ size = 5*sizeof(struct l_desc_struct); if (size > uap->bytecount) size = uap->bytecount; for (written = error = 0; written < size && error == 0; written++) error = subyte((char *)uap->ptr + written, 0); td->td_retval[0] = written; break; case 0x01: /* write_ldt */ case 0x11: /* write_ldt */ if (uap->bytecount != sizeof(ld)) return (EINVAL); error = copyin(uap->ptr, &ld, sizeof(ld)); if (error) return (error); ldt.start = ld.entry_number; ldt.descs = &desc; ldt.num = 1; desc.sd.sd_lolimit = (ld.limit & 0x0000ffff); desc.sd.sd_hilimit = (ld.limit & 0x000f0000) >> 16; desc.sd.sd_lobase = (ld.base_addr & 0x00ffffff); desc.sd.sd_hibase = (ld.base_addr & 0xff000000) >> 24; desc.sd.sd_type = SDT_MEMRO | ((ld.read_exec_only ^ 1) << 1) | (ld.contents << 2); desc.sd.sd_dpl = 3; desc.sd.sd_p = (ld.seg_not_present ^ 1); desc.sd.sd_xx = 0; desc.sd.sd_def32 = ld.seg_32bit; desc.sd.sd_gran = ld.limit_in_pages; error = i386_set_ldt(td, &ldt, &desc); break; default: error = ENOSYS; break; } if (error == EOPNOTSUPP) { linux_msg(td, "modify_ldt needs kernel option USER_LDT"); error = ENOSYS; } return (error); } int linux_sigaction(struct thread *td, struct linux_sigaction_args *args) { l_osigaction_t osa; l_sigaction_t act, oact; int error; if (args->nsa != NULL) { error = copyin(args->nsa, &osa, sizeof(l_osigaction_t)); if (error) return (error); act.lsa_handler = osa.lsa_handler; act.lsa_flags = osa.lsa_flags; act.lsa_restorer = osa.lsa_restorer; LINUX_SIGEMPTYSET(act.lsa_mask); act.lsa_mask.__mask = osa.lsa_mask; } error = linux_do_sigaction(td, args->sig, args->nsa ? &act : NULL, args->osa ? &oact : NULL); if (args->osa != NULL && !error) { osa.lsa_handler = oact.lsa_handler; osa.lsa_flags = oact.lsa_flags; osa.lsa_restorer = oact.lsa_restorer; osa.lsa_mask = oact.lsa_mask.__mask; error = copyout(&osa, args->osa, sizeof(l_osigaction_t)); } return (error); } /* * Linux has two extra args, restart and oldmask. We dont use these, * but it seems that "restart" is actually a context pointer that * enables the signal to happen with a different register set. */ int linux_sigsuspend(struct thread *td, struct linux_sigsuspend_args *args) { sigset_t sigmask; l_sigset_t mask; LINUX_SIGEMPTYSET(mask); mask.__mask = args->mask; linux_to_bsd_sigset(&mask, &sigmask); return (kern_sigsuspend(td, sigmask)); } int linux_pause(struct thread *td, struct linux_pause_args *args) { struct proc *p = td->td_proc; sigset_t sigmask; PROC_LOCK(p); sigmask = td->td_sigmask; PROC_UNLOCK(p); return (kern_sigsuspend(td, sigmask)); } int linux_set_thread_area(struct thread *td, struct linux_set_thread_area_args *args) { struct l_user_desc info; int error; int idx; int a[2]; struct segment_descriptor sd; error = copyin(args->desc, &info, sizeof(struct l_user_desc)); if (error) return (error); idx = info.entry_number; /* * Semantics of Linux version: every thread in the system has array of * 3 tls descriptors. 1st is GLIBC TLS, 2nd is WINE, 3rd unknown. This * syscall loads one of the selected tls decriptors with a value and * also loads GDT descriptors 6, 7 and 8 with the content of the * per-thread descriptors. * * Semantics of FreeBSD version: I think we can ignore that Linux has 3 * per-thread descriptors and use just the 1st one. The tls_array[] * is used only in set/get-thread_area() syscalls and for loading the * GDT descriptors. In FreeBSD we use just one GDT descriptor for TLS * so we will load just one. * * XXX: this doesn't work when a user space process tries to use more * than 1 TLS segment. Comment in the Linux sources says wine might do * this. */ /* * we support just GLIBC TLS now * we should let 3 proceed as well because we use this segment so * if code does two subsequent calls it should succeed */ if (idx != 6 && idx != -1 && idx != 3) return (EINVAL); /* * we have to copy out the GDT entry we use * FreeBSD uses GDT entry #3 for storing %gs so load that * * XXX: what if a user space program doesn't check this value and tries * to use 6, 7 or 8? */ idx = info.entry_number = 3; error = copyout(&info, args->desc, sizeof(struct l_user_desc)); if (error) return (error); if (LINUX_LDT_empty(&info)) { a[0] = 0; a[1] = 0; } else { a[0] = LINUX_LDT_entry_a(&info); a[1] = LINUX_LDT_entry_b(&info); } memcpy(&sd, &a, sizeof(a)); /* this is taken from i386 version of cpu_set_user_tls() */ critical_enter(); /* set %gs */ td->td_pcb->pcb_gsd = sd; PCPU_GET(fsgs_gdt)[1] = sd; load_gs(GSEL(GUGS_SEL, SEL_UPL)); critical_exit(); return (0); } int linux_get_thread_area(struct thread *td, struct linux_get_thread_area_args *args) { struct l_user_desc info; int error; int idx; struct l_desc_struct desc; struct segment_descriptor sd; error = copyin(args->desc, &info, sizeof(struct l_user_desc)); if (error) return (error); idx = info.entry_number; /* XXX: I am not sure if we want 3 to be allowed too. */ if (idx != 6 && idx != 3) return (EINVAL); idx = 3; memset(&info, 0, sizeof(info)); sd = PCPU_GET(fsgs_gdt)[1]; memcpy(&desc, &sd, sizeof(desc)); info.entry_number = idx; info.base_addr = LINUX_GET_BASE(&desc); info.limit = LINUX_GET_LIMIT(&desc); info.seg_32bit = LINUX_GET_32BIT(&desc); info.contents = LINUX_GET_CONTENTS(&desc); info.read_exec_only = !LINUX_GET_WRITABLE(&desc); info.limit_in_pages = LINUX_GET_LIMIT_PAGES(&desc); info.seg_not_present = !LINUX_GET_PRESENT(&desc); info.useable = LINUX_GET_USEABLE(&desc); error = copyout(&info, args->desc, sizeof(struct l_user_desc)); if (error) return (EFAULT); return (0); } /* XXX: this wont work with module - convert it */ int linux_mq_open(struct thread *td, struct linux_mq_open_args *args) { #ifdef P1003_1B_MQUEUE return (sys_kmq_open(td, (struct kmq_open_args *)args)); #else return (ENOSYS); #endif } int linux_mq_unlink(struct thread *td, struct linux_mq_unlink_args *args) { #ifdef P1003_1B_MQUEUE return (sys_kmq_unlink(td, (struct kmq_unlink_args *)args)); #else return (ENOSYS); #endif } int linux_mq_timedsend(struct thread *td, struct linux_mq_timedsend_args *args) { #ifdef P1003_1B_MQUEUE return (sys_kmq_timedsend(td, (struct kmq_timedsend_args *)args)); #else return (ENOSYS); #endif } int linux_mq_timedreceive(struct thread *td, struct linux_mq_timedreceive_args *args) { #ifdef P1003_1B_MQUEUE return (sys_kmq_timedreceive(td, (struct kmq_timedreceive_args *)args)); #else return (ENOSYS); #endif } int linux_mq_notify(struct thread *td, struct linux_mq_notify_args *args) { #ifdef P1003_1B_MQUEUE return (sys_kmq_notify(td, (struct kmq_notify_args *)args)); #else return (ENOSYS); #endif } int linux_mq_getsetattr(struct thread *td, struct linux_mq_getsetattr_args *args) { #ifdef P1003_1B_MQUEUE return (sys_kmq_setattr(td, (struct kmq_setattr_args *)args)); #else return (ENOSYS); #endif } diff --git a/sys/i386/linux/linux_sysvec.c b/sys/i386/linux/linux_sysvec.c index 6805a29204c5..8968a491ce95 100644 --- a/sys/i386/linux/linux_sysvec.c +++ b/sys/i386/linux/linux_sysvec.c @@ -1,1179 +1,1180 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 1994-1996 Søren Schmidt * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include +#include #include #include #include #include #include #include MODULE_VERSION(linux, 1); #define LINUX_VDSOPAGE_SIZE PAGE_SIZE * 2 #define LINUX_VDSOPAGE (VM_MAXUSER_ADDRESS - LINUX_VDSOPAGE_SIZE) #define LINUX_SHAREDPAGE (LINUX_VDSOPAGE - PAGE_SIZE) /* * PAGE_SIZE - the size * of the native SHAREDPAGE */ #define LINUX_USRSTACK LINUX_SHAREDPAGE #define LINUX_PS_STRINGS (LINUX_USRSTACK - sizeof(struct ps_strings)) static int linux_szsigcode; static vm_object_t linux_vdso_obj; static char *linux_vdso_mapping; extern char _binary_linux_vdso_so_o_start; extern char _binary_linux_vdso_so_o_end; static vm_offset_t linux_vdso_base; extern struct sysent linux_sysent[LINUX_SYS_MAXSYSCALL]; SET_DECLARE(linux_ioctl_handler_set, struct linux_ioctl_handler); static int linux_fixup(uintptr_t *stack_base, struct image_params *iparams); static int linux_fixup_elf(uintptr_t *stack_base, struct image_params *iparams); static void linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask); static void linux_exec_setregs(struct thread *td, struct image_params *imgp, uintptr_t stack); static void linux_exec_sysvec_init(void *param); static int linux_on_exec_vmspace(struct proc *p, struct image_params *imgp); static int linux_copyout_strings(struct image_params *imgp, uintptr_t *stack_base); static void linux_set_fork_retval(struct thread *td); static bool linux_trans_osrel(const Elf_Note *note, int32_t *osrel); static void linux_vdso_install(const void *param); static void linux_vdso_deinstall(const void *param); static void linux_vdso_reloc(char *mapping, Elf_Addr offset); #define LINUX_T_UNKNOWN 255 static int _bsd_to_linux_trapcode[] = { LINUX_T_UNKNOWN, /* 0 */ 6, /* 1 T_PRIVINFLT */ LINUX_T_UNKNOWN, /* 2 */ 3, /* 3 T_BPTFLT */ LINUX_T_UNKNOWN, /* 4 */ LINUX_T_UNKNOWN, /* 5 */ 16, /* 6 T_ARITHTRAP */ 254, /* 7 T_ASTFLT */ LINUX_T_UNKNOWN, /* 8 */ 13, /* 9 T_PROTFLT */ 1, /* 10 T_TRCTRAP */ LINUX_T_UNKNOWN, /* 11 */ 14, /* 12 T_PAGEFLT */ LINUX_T_UNKNOWN, /* 13 */ 17, /* 14 T_ALIGNFLT */ LINUX_T_UNKNOWN, /* 15 */ LINUX_T_UNKNOWN, /* 16 */ LINUX_T_UNKNOWN, /* 17 */ 0, /* 18 T_DIVIDE */ 2, /* 19 T_NMI */ 4, /* 20 T_OFLOW */ 5, /* 21 T_BOUND */ 7, /* 22 T_DNA */ 8, /* 23 T_DOUBLEFLT */ 9, /* 24 T_FPOPFLT */ 10, /* 25 T_TSSFLT */ 11, /* 26 T_SEGNPFLT */ 12, /* 27 T_STKFLT */ 18, /* 28 T_MCHK */ 19, /* 29 T_XMMFLT */ 15 /* 30 T_RESERVED */ }; #define bsd_to_linux_trapcode(code) \ ((code)args->argc + 1); base--; suword(base, (intptr_t)envp); base--; suword(base, (intptr_t)argv); base--; suword(base, imgp->args->argc); *stack_base = (uintptr_t)base; return (0); } static int linux_copyout_auxargs(struct image_params *imgp, uintptr_t base) { struct proc *p; Elf32_Auxargs *args; Elf32_Auxinfo *argarray, *pos; struct ps_strings *arginfo; int error, issetugid; p = imgp->proc; issetugid = imgp->proc->p_flag & P_SUGID ? 1 : 0; arginfo = (struct ps_strings *)PROC_PS_STRINGS(p); args = (Elf32_Auxargs *)imgp->auxargs; argarray = pos = malloc(LINUX_AT_COUNT * sizeof(*pos), M_TEMP, M_WAITOK | M_ZERO); AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO_EHDR, linux_vdso_base); AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO, __kernel_vsyscall); AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, cpu_feature); /* * Do not export AT_CLKTCK when emulating Linux kernel prior to 2.4.0, * as it has appeared in the 2.4.0-rc7 first time. * Being exported, AT_CLKTCK is returned by sysconf(_SC_CLK_TCK), * glibc falls back to the hard-coded CLK_TCK value when aux entry * is not present. * Also see linux_times() implementation. */ if (linux_kernver(curthread) >= LINUX_KERNVER_2004000) AUXARGS_ENTRY(pos, LINUX_AT_CLKTCK, stclohz); AUXARGS_ENTRY(pos, AT_PHDR, args->phdr); AUXARGS_ENTRY(pos, AT_PHENT, args->phent); AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum); AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz); AUXARGS_ENTRY(pos, AT_FLAGS, args->flags); AUXARGS_ENTRY(pos, AT_ENTRY, args->entry); AUXARGS_ENTRY(pos, AT_BASE, args->base); AUXARGS_ENTRY(pos, LINUX_AT_SECURE, issetugid); AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_ucred->cr_ruid); AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid); AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_ucred->cr_rgid); AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid); AUXARGS_ENTRY(pos, LINUX_AT_PLATFORM, PTROUT(linux_platform)); AUXARGS_ENTRY_PTR(pos, LINUX_AT_RANDOM, imgp->canary); if (imgp->execpathp != 0) AUXARGS_ENTRY_PTR(pos, LINUX_AT_EXECFN, imgp->execpathp); if (args->execfd != -1) AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd); AUXARGS_ENTRY(pos, AT_NULL, 0); free(imgp->auxargs, M_TEMP); imgp->auxargs = NULL; KASSERT(pos - argarray <= LINUX_AT_COUNT, ("Too many auxargs")); error = copyout(argarray, (void *)base, sizeof(*argarray) * LINUX_AT_COUNT); free(argarray, M_TEMP); return (error); } static int linux_fixup_elf(uintptr_t *stack_base, struct image_params *imgp) { register_t *base; base = (register_t *)*stack_base; base--; if (suword(base, (register_t)imgp->args->argc) == -1) return (EFAULT); *stack_base = (uintptr_t)base; return (0); } /* * Copied from kern/kern_exec.c */ static int linux_copyout_strings(struct image_params *imgp, uintptr_t *stack_base) { int argc, envc, error; char **vectp; char *stringp; uintptr_t destp, ustringp; struct ps_strings *arginfo; char canary[LINUX_AT_RANDOM_LEN]; size_t execpath_len; struct proc *p; p = imgp->proc; arginfo = (struct ps_strings *)PROC_PS_STRINGS(p); destp = (uintptr_t)arginfo; if (imgp->execpath != NULL && imgp->auxargs != NULL) { execpath_len = strlen(imgp->execpath) + 1; destp -= execpath_len; destp = rounddown2(destp, sizeof(void *)); imgp->execpathp = (void *)destp; error = copyout(imgp->execpath, imgp->execpathp, execpath_len); if (error != 0) return (error); } /* Prepare the canary for SSP. */ arc4rand(canary, sizeof(canary), 0); destp -= roundup(sizeof(canary), sizeof(void *)); imgp->canary = (void *)destp; error = copyout(canary, imgp->canary, sizeof(canary)); if (error != 0) return (error); /* Allocate room for the argument and environment strings. */ destp -= ARG_MAX - imgp->args->stringspace; destp = rounddown2(destp, sizeof(void *)); ustringp = destp; if (imgp->auxargs) { /* * Allocate room on the stack for the ELF auxargs * array. It has LINUX_AT_COUNT entries. */ destp -= LINUX_AT_COUNT * sizeof(Elf32_Auxinfo); destp = rounddown2(destp, sizeof(void *)); } vectp = (char **)destp; /* * Allocate room for the argv[] and env vectors including the * terminating NULL pointers. */ vectp -= imgp->args->argc + 1 + imgp->args->envc + 1; /* vectp also becomes our initial stack base. */ *stack_base = (uintptr_t)vectp; stringp = imgp->args->begin_argv; argc = imgp->args->argc; envc = imgp->args->envc; /* Copy out strings - arguments and environment. */ error = copyout(stringp, (void *)ustringp, ARG_MAX - imgp->args->stringspace); if (error != 0) return (error); /* Fill in "ps_strings" struct for ps, w, etc. */ if (suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp) != 0 || suword(&arginfo->ps_nargvstr, argc) != 0) return (EFAULT); /* Fill in argument portion of vector table. */ for (; argc > 0; --argc) { if (suword(vectp++, ustringp) != 0) return (EFAULT); while (*stringp++ != 0) ustringp++; ustringp++; } /* A null vector table pointer separates the argp's from the envp's. */ if (suword(vectp++, 0) != 0) return (EFAULT); if (suword(&arginfo->ps_envstr, (long)(intptr_t)vectp) != 0 || suword(&arginfo->ps_nenvstr, envc) != 0) return (EFAULT); /* Fill in environment portion of vector table. */ for (; envc > 0; --envc) { if (suword(vectp++, ustringp) != 0) return (EFAULT); while (*stringp++ != 0) ustringp++; ustringp++; } /* The end of the vector table is a null pointer. */ if (suword(vectp, 0) != 0) return (EFAULT); if (imgp->auxargs) { vectp++; error = imgp->sysent->sv_copyout_auxargs(imgp, (uintptr_t)vectp); if (error != 0) return (error); } return (0); } static void linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) { struct thread *td = curthread; struct proc *p = td->td_proc; struct sigacts *psp; struct trapframe *regs; struct l_rt_sigframe *fp, frame; int sig, code; int oonstack; sig = ksi->ksi_signo; code = ksi->ksi_code; PROC_LOCK_ASSERT(p, MA_OWNED); psp = p->p_sigacts; mtx_assert(&psp->ps_mtx, MA_OWNED); regs = td->td_frame; oonstack = sigonstack(regs->tf_esp); /* Allocate space for the signal handler context. */ if ((td->td_pflags & TDP_ALTSTACK) && !oonstack && SIGISMEMBER(psp->ps_sigonstack, sig)) { fp = (struct l_rt_sigframe *)((uintptr_t)td->td_sigstk.ss_sp + td->td_sigstk.ss_size - sizeof(struct l_rt_sigframe)); } else fp = (struct l_rt_sigframe *)regs->tf_esp - 1; mtx_unlock(&psp->ps_mtx); /* Build the argument list for the signal handler. */ sig = bsd_to_linux_signal(sig); bzero(&frame, sizeof(frame)); frame.sf_handler = catcher; frame.sf_sig = sig; frame.sf_siginfo = &fp->sf_si; frame.sf_ucontext = &fp->sf_sc; /* Fill in POSIX parts. */ siginfo_to_lsiginfo(&ksi->ksi_info, &frame.sf_si, sig); /* Build the signal context to be used by sigreturn. */ frame.sf_sc.uc_flags = 0; /* XXX ??? */ frame.sf_sc.uc_link = NULL; /* XXX ??? */ frame.sf_sc.uc_stack.ss_sp = PTROUT(td->td_sigstk.ss_sp); frame.sf_sc.uc_stack.ss_size = td->td_sigstk.ss_size; frame.sf_sc.uc_stack.ss_flags = (td->td_pflags & TDP_ALTSTACK) ? ((oonstack) ? LINUX_SS_ONSTACK : 0) : LINUX_SS_DISABLE; PROC_UNLOCK(p); bsd_to_linux_sigset(mask, &frame.sf_sc.uc_sigmask); frame.sf_sc.uc_mcontext.sc_mask = frame.sf_sc.uc_sigmask.__mask; frame.sf_sc.uc_mcontext.sc_gs = rgs(); frame.sf_sc.uc_mcontext.sc_fs = regs->tf_fs; frame.sf_sc.uc_mcontext.sc_es = regs->tf_es; frame.sf_sc.uc_mcontext.sc_ds = regs->tf_ds; frame.sf_sc.uc_mcontext.sc_edi = regs->tf_edi; frame.sf_sc.uc_mcontext.sc_esi = regs->tf_esi; frame.sf_sc.uc_mcontext.sc_ebp = regs->tf_ebp; frame.sf_sc.uc_mcontext.sc_ebx = regs->tf_ebx; frame.sf_sc.uc_mcontext.sc_esp = regs->tf_esp; frame.sf_sc.uc_mcontext.sc_edx = regs->tf_edx; frame.sf_sc.uc_mcontext.sc_ecx = regs->tf_ecx; frame.sf_sc.uc_mcontext.sc_eax = regs->tf_eax; frame.sf_sc.uc_mcontext.sc_eip = regs->tf_eip; frame.sf_sc.uc_mcontext.sc_cs = regs->tf_cs; frame.sf_sc.uc_mcontext.sc_eflags = regs->tf_eflags; frame.sf_sc.uc_mcontext.sc_esp_at_signal = regs->tf_esp; frame.sf_sc.uc_mcontext.sc_ss = regs->tf_ss; frame.sf_sc.uc_mcontext.sc_err = regs->tf_err; frame.sf_sc.uc_mcontext.sc_cr2 = (register_t)ksi->ksi_addr; frame.sf_sc.uc_mcontext.sc_trapno = bsd_to_linux_trapcode(code); if (copyout(&frame, fp, sizeof(frame)) != 0) { /* * Process has trashed its stack; give it an illegal * instruction to halt it in its tracks. */ PROC_LOCK(p); sigexit(td, SIGILL); } /* Build context to run handler in. */ regs->tf_esp = (int)fp; regs->tf_eip = __kernel_rt_sigreturn; regs->tf_eflags &= ~(PSL_T | PSL_VM | PSL_D); regs->tf_cs = _ucodesel; regs->tf_ds = _udatasel; regs->tf_es = _udatasel; regs->tf_fs = _udatasel; regs->tf_ss = _udatasel; PROC_LOCK(p); mtx_lock(&psp->ps_mtx); } /* * Send an interrupt to process. * * Stack is set up to allow sigcode stored * in u. to call routine, followed by kcall * to sigreturn routine below. After sigreturn * resets the signal mask, the stack, and the * frame pointer, it returns to the user * specified pc, psl. */ static void linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) { struct thread *td = curthread; struct proc *p = td->td_proc; struct sigacts *psp; struct trapframe *regs; struct l_sigframe *fp, frame; l_sigset_t lmask; int sig, code; int oonstack; PROC_LOCK_ASSERT(p, MA_OWNED); psp = p->p_sigacts; sig = ksi->ksi_signo; code = ksi->ksi_code; mtx_assert(&psp->ps_mtx, MA_OWNED); if (SIGISMEMBER(psp->ps_siginfo, sig)) { /* Signal handler installed with SA_SIGINFO. */ linux_rt_sendsig(catcher, ksi, mask); return; } regs = td->td_frame; oonstack = sigonstack(regs->tf_esp); /* Allocate space for the signal handler context. */ if ((td->td_pflags & TDP_ALTSTACK) && !oonstack && SIGISMEMBER(psp->ps_sigonstack, sig)) { fp = (struct l_sigframe *)((uintptr_t)td->td_sigstk.ss_sp + td->td_sigstk.ss_size - sizeof(struct l_sigframe)); } else fp = (struct l_sigframe *)regs->tf_esp - 1; mtx_unlock(&psp->ps_mtx); PROC_UNLOCK(p); /* Build the argument list for the signal handler. */ sig = bsd_to_linux_signal(sig); bzero(&frame, sizeof(frame)); frame.sf_handler = catcher; frame.sf_sig = sig; bsd_to_linux_sigset(mask, &lmask); /* Build the signal context to be used by sigreturn. */ frame.sf_sc.sc_mask = lmask.__mask; frame.sf_sc.sc_gs = rgs(); frame.sf_sc.sc_fs = regs->tf_fs; frame.sf_sc.sc_es = regs->tf_es; frame.sf_sc.sc_ds = regs->tf_ds; frame.sf_sc.sc_edi = regs->tf_edi; frame.sf_sc.sc_esi = regs->tf_esi; frame.sf_sc.sc_ebp = regs->tf_ebp; frame.sf_sc.sc_ebx = regs->tf_ebx; frame.sf_sc.sc_esp = regs->tf_esp; frame.sf_sc.sc_edx = regs->tf_edx; frame.sf_sc.sc_ecx = regs->tf_ecx; frame.sf_sc.sc_eax = regs->tf_eax; frame.sf_sc.sc_eip = regs->tf_eip; frame.sf_sc.sc_cs = regs->tf_cs; frame.sf_sc.sc_eflags = regs->tf_eflags; frame.sf_sc.sc_esp_at_signal = regs->tf_esp; frame.sf_sc.sc_ss = regs->tf_ss; frame.sf_sc.sc_err = regs->tf_err; frame.sf_sc.sc_cr2 = (register_t)ksi->ksi_addr; frame.sf_sc.sc_trapno = bsd_to_linux_trapcode(ksi->ksi_trapno); frame.sf_extramask[0] = lmask.__mask; if (copyout(&frame, fp, sizeof(frame)) != 0) { /* * Process has trashed its stack; give it an illegal * instruction to halt it in its tracks. */ PROC_LOCK(p); sigexit(td, SIGILL); } /* Build context to run handler in. */ regs->tf_esp = (int)fp; regs->tf_eip = __kernel_sigreturn; regs->tf_eflags &= ~(PSL_T | PSL_VM | PSL_D); regs->tf_cs = _ucodesel; regs->tf_ds = _udatasel; regs->tf_es = _udatasel; regs->tf_fs = _udatasel; regs->tf_ss = _udatasel; PROC_LOCK(p); mtx_lock(&psp->ps_mtx); } /* * System call to cleanup state after a signal * has been taken. Reset signal mask and * stack state from context left by sendsig (above). * Return to previous pc and psl as specified by * context left by sendsig. Check carefully to * make sure that the user has not modified the * psl to gain improper privileges or to cause * a machine fault. */ int linux_sigreturn(struct thread *td, struct linux_sigreturn_args *args) { struct l_sigframe frame; struct trapframe *regs; l_sigset_t lmask; sigset_t bmask; int eflags; ksiginfo_t ksi; regs = td->td_frame; /* * The trampoline code hands us the sigframe. * It is unsafe to keep track of it ourselves, in the event that a * program jumps out of a signal handler. */ if (copyin(args->sfp, &frame, sizeof(frame)) != 0) return (EFAULT); /* Check for security violations. */ #define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0) eflags = frame.sf_sc.sc_eflags; if (!EFLAGS_SECURE(eflags, regs->tf_eflags)) return (EINVAL); /* * Don't allow users to load a valid privileged %cs. Let the * hardware check for invalid selectors, excess privilege in * other selectors, invalid %eip's and invalid %esp's. */ #define CS_SECURE(cs) (ISPL(cs) == SEL_UPL) if (!CS_SECURE(frame.sf_sc.sc_cs)) { ksiginfo_init_trap(&ksi); ksi.ksi_signo = SIGBUS; ksi.ksi_code = BUS_OBJERR; ksi.ksi_trapno = T_PROTFLT; ksi.ksi_addr = (void *)regs->tf_eip; trapsignal(td, &ksi); return (EINVAL); } lmask.__mask = frame.sf_sc.sc_mask; linux_to_bsd_sigset(&lmask, &bmask); kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0); /* Restore signal context. */ /* %gs was restored by the trampoline. */ regs->tf_fs = frame.sf_sc.sc_fs; regs->tf_es = frame.sf_sc.sc_es; regs->tf_ds = frame.sf_sc.sc_ds; regs->tf_edi = frame.sf_sc.sc_edi; regs->tf_esi = frame.sf_sc.sc_esi; regs->tf_ebp = frame.sf_sc.sc_ebp; regs->tf_ebx = frame.sf_sc.sc_ebx; regs->tf_edx = frame.sf_sc.sc_edx; regs->tf_ecx = frame.sf_sc.sc_ecx; regs->tf_eax = frame.sf_sc.sc_eax; regs->tf_eip = frame.sf_sc.sc_eip; regs->tf_cs = frame.sf_sc.sc_cs; regs->tf_eflags = eflags; regs->tf_esp = frame.sf_sc.sc_esp_at_signal; regs->tf_ss = frame.sf_sc.sc_ss; return (EJUSTRETURN); } /* * System call to cleanup state after a signal * has been taken. Reset signal mask and * stack state from context left by rt_sendsig (above). * Return to previous pc and psl as specified by * context left by sendsig. Check carefully to * make sure that the user has not modified the * psl to gain improper privileges or to cause * a machine fault. */ int linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args) { struct l_ucontext uc; struct l_sigcontext *context; sigset_t bmask; l_stack_t *lss; stack_t ss; struct trapframe *regs; int eflags; ksiginfo_t ksi; regs = td->td_frame; /* * The trampoline code hands us the ucontext. * It is unsafe to keep track of it ourselves, in the event that a * program jumps out of a signal handler. */ if (copyin(args->ucp, &uc, sizeof(uc)) != 0) return (EFAULT); context = &uc.uc_mcontext; /* Check for security violations. */ #define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0) eflags = context->sc_eflags; if (!EFLAGS_SECURE(eflags, regs->tf_eflags)) return (EINVAL); /* * Don't allow users to load a valid privileged %cs. Let the * hardware check for invalid selectors, excess privilege in * other selectors, invalid %eip's and invalid %esp's. */ #define CS_SECURE(cs) (ISPL(cs) == SEL_UPL) if (!CS_SECURE(context->sc_cs)) { ksiginfo_init_trap(&ksi); ksi.ksi_signo = SIGBUS; ksi.ksi_code = BUS_OBJERR; ksi.ksi_trapno = T_PROTFLT; ksi.ksi_addr = (void *)regs->tf_eip; trapsignal(td, &ksi); return (EINVAL); } linux_to_bsd_sigset(&uc.uc_sigmask, &bmask); kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0); /* Restore signal context. */ /* %gs was restored by the trampoline. */ regs->tf_fs = context->sc_fs; regs->tf_es = context->sc_es; regs->tf_ds = context->sc_ds; regs->tf_edi = context->sc_edi; regs->tf_esi = context->sc_esi; regs->tf_ebp = context->sc_ebp; regs->tf_ebx = context->sc_ebx; regs->tf_edx = context->sc_edx; regs->tf_ecx = context->sc_ecx; regs->tf_eax = context->sc_eax; regs->tf_eip = context->sc_eip; regs->tf_cs = context->sc_cs; regs->tf_eflags = eflags; regs->tf_esp = context->sc_esp_at_signal; regs->tf_ss = context->sc_ss; /* Call sigaltstack & ignore results. */ lss = &uc.uc_stack; ss.ss_sp = PTRIN(lss->ss_sp); ss.ss_size = lss->ss_size; ss.ss_flags = linux_to_bsd_sigaltstack(lss->ss_flags); (void)kern_sigaltstack(td, &ss, NULL); return (EJUSTRETURN); } static int linux_fetch_syscall_args(struct thread *td) { struct proc *p; struct trapframe *frame; struct syscall_args *sa; p = td->td_proc; frame = td->td_frame; sa = &td->td_sa; sa->code = frame->tf_eax; sa->args[0] = frame->tf_ebx; sa->args[1] = frame->tf_ecx; sa->args[2] = frame->tf_edx; sa->args[3] = frame->tf_esi; sa->args[4] = frame->tf_edi; sa->args[5] = frame->tf_ebp; /* Unconfirmed */ if (sa->code >= p->p_sysent->sv_size) /* nosys */ sa->callp = &p->p_sysent->sv_table[p->p_sysent->sv_size - 1]; else sa->callp = &p->p_sysent->sv_table[sa->code]; td->td_retval[0] = 0; td->td_retval[1] = frame->tf_edx; return (0); } static void linux_set_syscall_retval(struct thread *td, int error) { struct trapframe *frame = td->td_frame; cpu_set_syscall_retval(td, error); if (__predict_false(error != 0)) { if (error != ERESTART && error != EJUSTRETURN) frame->tf_eax = bsd_to_linux_errno(error); } } static void linux_set_fork_retval(struct thread *td) { struct trapframe *frame = td->td_frame; frame->tf_eax = 0; } /* * exec_setregs may initialize some registers differently than Linux * does, thus potentially confusing Linux binaries. If necessary, we * override the exec_setregs default(s) here. */ static void linux_exec_setregs(struct thread *td, struct image_params *imgp, uintptr_t stack) { struct pcb *pcb = td->td_pcb; exec_setregs(td, imgp, stack); /* Linux sets %gs to 0, we default to _udatasel. */ pcb->pcb_gs = 0; load_gs(0); pcb->pcb_initial_npxcw = __LINUX_NPXCW__; } struct sysentvec linux_sysvec = { .sv_size = LINUX_SYS_MAXSYSCALL, .sv_table = linux_sysent, .sv_transtrap = linux_translate_traps, .sv_fixup = linux_fixup, .sv_sendsig = linux_sendsig, .sv_sigcode = &_binary_linux_vdso_so_o_start, .sv_szsigcode = &linux_szsigcode, .sv_name = "Linux a.out", .sv_coredump = NULL, .sv_imgact_try = linux_exec_imgact_try, .sv_minsigstksz = LINUX_MINSIGSTKSZ, .sv_minuser = VM_MIN_ADDRESS, .sv_maxuser = VM_MAXUSER_ADDRESS, .sv_usrstack = LINUX_USRSTACK, .sv_psstrings = PS_STRINGS, .sv_psstringssz = sizeof(struct ps_strings), .sv_stackprot = VM_PROT_ALL, .sv_copyout_strings = exec_copyout_strings, .sv_setregs = linux_exec_setregs, .sv_fixlimit = NULL, .sv_maxssiz = NULL, .sv_flags = SV_ABI_LINUX | SV_AOUT | SV_IA32 | SV_ILP32 | SV_SIG_DISCIGN | SV_SIG_WAITNDQ, .sv_set_syscall_retval = linux_set_syscall_retval, .sv_fetch_syscall_args = linux_fetch_syscall_args, .sv_syscallnames = NULL, .sv_schedtail = linux_schedtail, .sv_thread_detach = linux_thread_detach, .sv_trap = NULL, .sv_onexec = linux_on_exec_vmspace, .sv_onexit = linux_on_exit, .sv_ontdexit = linux_thread_dtor, .sv_setid_allowed = &linux_setid_allowed_query, .sv_set_fork_retval = linux_set_fork_retval, }; INIT_SYSENTVEC(aout_sysvec, &linux_sysvec); struct sysentvec elf_linux_sysvec = { .sv_size = LINUX_SYS_MAXSYSCALL, .sv_table = linux_sysent, .sv_transtrap = linux_translate_traps, .sv_fixup = linux_fixup_elf, .sv_sendsig = linux_sendsig, .sv_sigcode = &_binary_linux_vdso_so_o_start, .sv_szsigcode = &linux_szsigcode, .sv_name = "Linux ELF32", .sv_coredump = elf32_coredump, .sv_elf_core_osabi = ELFOSABI_FREEBSD, .sv_elf_core_abi_vendor = FREEBSD_ABI_VENDOR, .sv_elf_core_prepare_notes = elf32_prepare_notes, .sv_imgact_try = linux_exec_imgact_try, .sv_minsigstksz = LINUX_MINSIGSTKSZ, .sv_minuser = VM_MIN_ADDRESS, .sv_maxuser = VM_MAXUSER_ADDRESS, .sv_usrstack = LINUX_USRSTACK, .sv_psstrings = LINUX_PS_STRINGS, .sv_psstringssz = sizeof(struct ps_strings), .sv_stackprot = VM_PROT_ALL, .sv_copyout_auxargs = linux_copyout_auxargs, .sv_copyout_strings = linux_copyout_strings, .sv_setregs = linux_exec_setregs, .sv_fixlimit = NULL, .sv_maxssiz = NULL, .sv_flags = SV_ABI_LINUX | SV_IA32 | SV_ILP32 | SV_SHP | SV_SIG_DISCIGN | SV_SIG_WAITNDQ | SV_TIMEKEEP, .sv_set_syscall_retval = linux_set_syscall_retval, .sv_fetch_syscall_args = linux_fetch_syscall_args, .sv_syscallnames = NULL, .sv_shared_page_base = LINUX_SHAREDPAGE, .sv_shared_page_len = PAGE_SIZE, .sv_schedtail = linux_schedtail, .sv_thread_detach = linux_thread_detach, .sv_trap = NULL, .sv_onexec = linux_on_exec_vmspace, .sv_onexit = linux_on_exit, .sv_ontdexit = linux_thread_dtor, .sv_setid_allowed = &linux_setid_allowed_query, .sv_set_fork_retval = linux_set_fork_retval, }; static int linux_on_exec_vmspace(struct proc *p, struct image_params *imgp) { int error = 0; if (SV_PROC_FLAG(p, SV_SHP) != 0) error = linux_map_vdso(p, linux_vdso_obj, linux_vdso_base, LINUX_VDSOPAGE_SIZE, imgp); if (error == 0) linux_on_exec(p, imgp); return (error); } /* * linux_vdso_install() and linux_exec_sysvec_init() must be called * after exec_sysvec_init() which is SI_SUB_EXEC (SI_ORDER_ANY). */ static void linux_exec_sysvec_init(void *param) { l_uintptr_t *ktimekeep_base, *ktsc_selector; struct sysentvec *sv; ptrdiff_t tkoff; sv = param; /* Fill timekeep_base */ exec_sysvec_init(sv); tkoff = kern_timekeep_base - linux_vdso_base; ktimekeep_base = (l_uintptr_t *)(linux_vdso_mapping + tkoff); *ktimekeep_base = sv->sv_timekeep_base; tkoff = kern_tsc_selector - linux_vdso_base; ktsc_selector = (l_uintptr_t *)(linux_vdso_mapping + tkoff); *ktsc_selector = linux_vdso_tsc_selector_idx(); if (bootverbose) printf("Linux i386 vDSO tsc_selector: %u\n", *ktsc_selector); } SYSINIT(elf_linux_exec_sysvec_init, SI_SUB_EXEC + 1, SI_ORDER_ANY, linux_exec_sysvec_init, &elf_linux_sysvec); static void linux_vdso_install(const void *param) { char *vdso_start = &_binary_linux_vdso_so_o_start; char *vdso_end = &_binary_linux_vdso_so_o_end; linux_szsigcode = vdso_end - vdso_start; MPASS(linux_szsigcode <= LINUX_VDSOPAGE_SIZE); linux_vdso_base = LINUX_VDSOPAGE; __elfN(linux_vdso_fixup)(vdso_start, linux_vdso_base); linux_vdso_obj = __elfN(linux_shared_page_init) (&linux_vdso_mapping, LINUX_VDSOPAGE_SIZE); bcopy(vdso_start, linux_vdso_mapping, linux_szsigcode); linux_vdso_reloc(linux_vdso_mapping, linux_vdso_base); } SYSINIT(elf_linux_vdso_init, SI_SUB_EXEC + 1, SI_ORDER_FIRST, linux_vdso_install, NULL); static void linux_vdso_deinstall(const void *param) { __elfN(linux_shared_page_fini)(linux_vdso_obj, linux_vdso_mapping, LINUX_VDSOPAGE_SIZE); } SYSUNINIT(elf_linux_vdso_uninit, SI_SUB_EXEC, SI_ORDER_FIRST, linux_vdso_deinstall, NULL); static void linux_vdso_reloc(char *mapping, Elf_Addr offset) { const Elf_Shdr *shdr; const Elf_Rel *rel; const Elf_Ehdr *ehdr; Elf_Addr *where; Elf_Size rtype, symidx; Elf_Addr addr, addend; int i, relcnt; MPASS(offset != 0); relcnt = 0; ehdr = (const Elf_Ehdr *)mapping; shdr = (const Elf_Shdr *)(mapping + ehdr->e_shoff); for (i = 0; i < ehdr->e_shnum; i++) { switch (shdr[i].sh_type) { case SHT_REL: rel = (const Elf_Rel *)(mapping + shdr[i].sh_offset); relcnt = shdr[i].sh_size / sizeof(*rel); break; case SHT_RELA: printf("Linux i386 vDSO: unexpected Rela section\n"); break; } } for (i = 0; i < relcnt; i++, rel++) { where = (Elf_Addr *)(mapping + rel->r_offset); addend = *where; rtype = ELF_R_TYPE(rel->r_info); symidx = ELF_R_SYM(rel->r_info); switch (rtype) { case R_386_NONE: /* none */ break; case R_386_RELATIVE: /* B + A */ addr = (Elf_Addr)PTROUT(offset + addend); if (*where != addr) *where = addr; break; case R_386_IRELATIVE: printf("Linux i386 vDSO: unexpected ifunc relocation, " "symbol index %d\n", symidx); break; default: printf("Linux i386 vDSO: unexpected relocation type %d, " "symbol index %d\n", rtype, symidx); } } } static char GNU_ABI_VENDOR[] = "GNU"; static int GNULINUX_ABI_DESC = 0; static bool linux_trans_osrel(const Elf_Note *note, int32_t *osrel) { const Elf32_Word *desc; uintptr_t p; p = (uintptr_t)(note + 1); p += roundup2(note->n_namesz, sizeof(Elf32_Addr)); desc = (const Elf32_Word *)p; if (desc[0] != GNULINUX_ABI_DESC) return (false); /* * For Linux we encode osrel using the Linux convention of * (version << 16) | (major << 8) | (minor) * See macro in linux_mib.h */ *osrel = LINUX_KERNVER(desc[1], desc[2], desc[3]); return (true); } static Elf_Brandnote linux_brandnote = { .hdr.n_namesz = sizeof(GNU_ABI_VENDOR), .hdr.n_descsz = 16, /* XXX at least 16 */ .hdr.n_type = 1, .vendor = GNU_ABI_VENDOR, .flags = BN_TRANSLATE_OSREL, .trans_osrel = linux_trans_osrel }; static Elf32_Brandinfo linux_brand = { .brand = ELFOSABI_LINUX, .machine = EM_386, .compat_3_brand = "Linux", .emul_path = linux_emul_path, .interp_path = "/lib/ld-linux.so.1", .sysvec = &elf_linux_sysvec, .interp_newpath = NULL, .brand_note = &linux_brandnote, .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE }; static Elf32_Brandinfo linux_glibc2brand = { .brand = ELFOSABI_LINUX, .machine = EM_386, .compat_3_brand = "Linux", .emul_path = linux_emul_path, .interp_path = "/lib/ld-linux.so.2", .sysvec = &elf_linux_sysvec, .interp_newpath = NULL, .brand_note = &linux_brandnote, .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE }; static Elf32_Brandinfo linux_muslbrand = { .brand = ELFOSABI_LINUX, .machine = EM_386, .compat_3_brand = "Linux", .emul_path = linux_emul_path, .interp_path = "/lib/ld-musl-i386.so.1", .sysvec = &elf_linux_sysvec, .interp_newpath = NULL, .brand_note = &linux_brandnote, .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE | LINUX_BI_FUTEX_REQUEUE }; Elf32_Brandinfo *linux_brandlist[] = { &linux_brand, &linux_glibc2brand, &linux_muslbrand, NULL }; static int linux_elf_modevent(module_t mod, int type, void *data) { Elf32_Brandinfo **brandinfo; int error; struct linux_ioctl_handler **lihp; error = 0; switch(type) { case MOD_LOAD: for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; ++brandinfo) if (elf32_insert_brand_entry(*brandinfo) < 0) error = EINVAL; if (error == 0) { SET_FOREACH(lihp, linux_ioctl_handler_set) linux_ioctl_register_handler(*lihp); linux_dev_shm_create(); linux_osd_jail_register(); stclohz = (stathz ? stathz : hz); if (bootverbose) printf("Linux ELF exec handler installed\n"); } else printf("cannot insert Linux ELF brand handler\n"); break; case MOD_UNLOAD: for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; ++brandinfo) if (elf32_brand_inuse(*brandinfo)) error = EBUSY; if (error == 0) { for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; ++brandinfo) if (elf32_remove_brand_entry(*brandinfo) < 0) error = EINVAL; } if (error == 0) { SET_FOREACH(lihp, linux_ioctl_handler_set) linux_ioctl_unregister_handler(*lihp); linux_dev_shm_destroy(); linux_osd_jail_deregister(); if (bootverbose) printf("Linux ELF exec handler removed\n"); } else printf("Could not deinstall ELF interpreter entry\n"); break; default: return (EOPNOTSUPP); } return (error); } static moduledata_t linux_elf_mod = { "linuxelf", linux_elf_modevent, 0 }; DECLARE_MODULE_TIED(linuxelf, linux_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY); FEATURE(linux, "Linux 32bit support");