diff --git a/sys/amd64/linux/linux.h b/sys/amd64/linux/linux.h index 920ca98ce01b..08cf3d4b6bbd 100644 --- a/sys/amd64/linux/linux.h +++ b/sys/amd64/linux/linux.h @@ -1,472 +1,471 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 1994-1996 Søren Schmidt * 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 _AMD64_LINUX_H_ #define _AMD64_LINUX_H_ #include #include #include #define LINUX_LEGACY_SYSCALLS #define LINUX_DTRACE linuxulator /* * Provide a separate set of types for the Linux types. */ typedef int32_t l_int; typedef int64_t l_long; typedef int16_t l_short; typedef uint32_t l_uint; typedef uint64_t l_ulong; typedef uint16_t l_ushort; typedef l_ulong l_uintptr_t; typedef l_long l_clock_t; typedef l_int l_daddr_t; typedef l_ulong l_dev_t; typedef l_uint l_gid_t; typedef l_ushort l_gid16_t; typedef l_uint l_uid_t; typedef l_ushort l_uid16_t; typedef l_ulong l_ino_t; typedef l_int l_key_t; typedef l_long l_loff_t; typedef l_uint l_mode_t; typedef l_long l_off_t; typedef l_int l_pid_t; typedef l_ulong l_size_t; typedef l_long l_ssize_t; typedef l_long l_suseconds_t; typedef l_long l_time_t; typedef l_int l_timer_t; typedef l_int l_mqd_t; typedef l_size_t l_socklen_t; typedef l_ulong l_fd_mask; typedef struct { l_int val[2]; } l_fsid_t; typedef struct { l_time_t tv_sec; l_suseconds_t tv_usec; } l_timeval; #define l_fd_set fd_set /* * Miscellaneous */ #define LINUX_AT_COUNT 20 /* Count of used aux entry types. */ struct l___sysctl_args { l_uintptr_t name; l_int nlen; l_uintptr_t oldval; l_uintptr_t oldlenp; l_uintptr_t newval; l_size_t newlen; l_ulong __spare[4]; }; /* Resource limits */ #define LINUX_RLIMIT_CPU 0 #define LINUX_RLIMIT_FSIZE 1 #define LINUX_RLIMIT_DATA 2 #define LINUX_RLIMIT_STACK 3 #define LINUX_RLIMIT_CORE 4 #define LINUX_RLIMIT_RSS 5 #define LINUX_RLIMIT_NPROC 6 #define LINUX_RLIMIT_NOFILE 7 #define LINUX_RLIMIT_MEMLOCK 8 #define LINUX_RLIMIT_AS 9 /* Address space limit */ #define LINUX_RLIM_NLIMITS 10 struct l_rlimit { l_ulong rlim_cur; l_ulong rlim_max; }; /* * stat family of syscalls */ struct l_timespec { l_time_t tv_sec; l_long tv_nsec; }; struct l_newstat { l_dev_t st_dev; l_ino_t st_ino; l_ulong st_nlink; l_uint st_mode; l_uid_t st_uid; l_gid_t st_gid; l_uint __st_pad1; l_dev_t st_rdev; l_off_t st_size; l_long st_blksize; l_long st_blocks; struct l_timespec st_atim; struct l_timespec st_mtim; struct l_timespec st_ctim; l_long __unused1; l_long __unused2; l_long __unused3; }; /* sigaction flags */ #define LINUX_SA_NOCLDSTOP 0x00000001 #define LINUX_SA_NOCLDWAIT 0x00000002 #define LINUX_SA_SIGINFO 0x00000004 #define LINUX_SA_RESTORER 0x04000000 #define LINUX_SA_ONSTACK 0x08000000 #define LINUX_SA_RESTART 0x10000000 #define LINUX_SA_INTERRUPT 0x20000000 #define LINUX_SA_NOMASK 0x40000000 #define LINUX_SA_ONESHOT 0x80000000 /* sigprocmask actions */ #define LINUX_SIG_BLOCK 0 #define LINUX_SIG_UNBLOCK 1 #define LINUX_SIG_SETMASK 2 /* sigaltstack */ #define LINUX_MINSIGSTKSZ 2048 typedef void (*l_handler_t)(l_int); typedef struct { l_handler_t lsa_handler; l_ulong lsa_flags; l_uintptr_t lsa_restorer; l_sigset_t lsa_mask; } l_sigaction_t; typedef struct { l_uintptr_t ss_sp; l_int ss_flags; l_size_t ss_size; } l_stack_t; struct l_fpstate { u_int16_t cwd; u_int16_t swd; u_int16_t twd; u_int16_t fop; u_int64_t rip; u_int64_t rdp; u_int32_t mxcsr; u_int32_t mxcsr_mask; u_int32_t st_space[32]; u_int32_t xmm_space[64]; u_int32_t reserved2[24]; }; struct l_sigcontext { l_ulong sc_r8; l_ulong sc_r9; l_ulong sc_r10; l_ulong sc_r11; l_ulong sc_r12; l_ulong sc_r13; l_ulong sc_r14; l_ulong sc_r15; l_ulong sc_rdi; l_ulong sc_rsi; l_ulong sc_rbp; l_ulong sc_rbx; l_ulong sc_rdx; l_ulong sc_rax; l_ulong sc_rcx; l_ulong sc_rsp; l_ulong sc_rip; l_ulong sc_rflags; l_ushort sc_cs; l_ushort sc_gs; l_ushort sc_fs; l_ushort sc___pad0; l_ulong sc_err; l_ulong sc_trapno; l_sigset_t sc_mask; l_ulong sc_cr2; struct l_fpstate *sc_fpstate; l_ulong sc_reserved1[8]; }; struct l_ucontext { l_ulong uc_flags; l_uintptr_t uc_link; l_stack_t uc_stack; struct l_sigcontext uc_mcontext; l_sigset_t uc_sigmask; }; #define LINUX_SI_PREAMBLE_SIZE (4 * sizeof(int)) #define LINUX_SI_MAX_SIZE 128 #define LINUX_SI_PAD_SIZE ((LINUX_SI_MAX_SIZE - \ LINUX_SI_PREAMBLE_SIZE) / sizeof(l_int)) typedef union l_sigval { l_int sival_int; l_uintptr_t sival_ptr; } l_sigval_t; typedef struct l_siginfo { l_int lsi_signo; l_int lsi_errno; l_int lsi_code; union { l_int _pad[LINUX_SI_PAD_SIZE]; struct { l_pid_t _pid; l_uid_t _uid; } _kill; struct { l_timer_t _tid; l_int _overrun; char _pad[sizeof(l_uid_t) - sizeof(int)]; union l_sigval _sigval; l_uint _sys_private; } _timer; struct { l_pid_t _pid; /* sender's pid */ l_uid_t _uid; /* sender's uid */ union l_sigval _sigval; } _rt; struct { l_pid_t _pid; /* which child */ l_uid_t _uid; /* sender's uid */ l_int _status; /* exit code */ l_clock_t _utime; l_clock_t _stime; } _sigchld; struct { l_uintptr_t _addr; /* Faulting insn/memory ref. */ } _sigfault; struct { l_long _band; /* POLL_IN,POLL_OUT,POLL_MSG */ l_int _fd; } _sigpoll; } _sifields; } l_siginfo_t; #define lsi_pid _sifields._kill._pid #define lsi_uid _sifields._kill._uid #define lsi_tid _sifields._timer._tid #define lsi_overrun _sifields._timer._overrun #define lsi_sys_private _sifields._timer._sys_private #define lsi_status _sifields._sigchld._status #define lsi_utime _sifields._sigchld._utime #define lsi_stime _sifields._sigchld._stime #define lsi_value _sifields._rt._sigval #define lsi_int _sifields._rt._sigval.sival_int #define lsi_ptr _sifields._rt._sigval.sival_ptr #define lsi_addr _sifields._sigfault._addr #define lsi_band _sifields._sigpoll._band #define lsi_fd _sifields._sigpoll._fd /* * We make the stack look like Linux expects it when calling a signal * handler, but use the BSD way of calling the handler and sigreturn(). * This means that we need to pass the pointer to the handler too. * It is appended to the frame to not interfere with the rest of it. */ struct l_rt_sigframe { struct l_ucontext sf_sc; struct l_siginfo sf_si; - l_handler_t sf_handler; }; /* * mount flags */ #define LINUX_MS_RDONLY 0x0001 #define LINUX_MS_NOSUID 0x0002 #define LINUX_MS_NODEV 0x0004 #define LINUX_MS_NOEXEC 0x0008 #define LINUX_MS_REMOUNT 0x0020 /* * SystemV IPC defines */ #define LINUX_IPC_RMID 0 #define LINUX_IPC_SET 1 #define LINUX_IPC_STAT 2 #define LINUX_IPC_INFO 3 #define LINUX_SHM_LOCK 11 #define LINUX_SHM_UNLOCK 12 #define LINUX_SHM_STAT 13 #define LINUX_SHM_INFO 14 #define LINUX_SHM_RDONLY 0x1000 #define LINUX_SHM_RND 0x2000 #define LINUX_SHM_REMAP 0x4000 /* semctl commands */ #define LINUX_GETPID 11 #define LINUX_GETVAL 12 #define LINUX_GETALL 13 #define LINUX_GETNCNT 14 #define LINUX_GETZCNT 15 #define LINUX_SETVAL 16 #define LINUX_SETALL 17 #define LINUX_SEM_STAT 18 #define LINUX_SEM_INFO 19 union l_semun { l_int val; l_uintptr_t buf; l_uintptr_t array; l_uintptr_t __buf; l_uintptr_t __pad; }; struct l_ifmap { l_ulong mem_start; l_ulong mem_end; l_ushort base_addr; u_char irq; u_char dma; u_char port; /* 3 bytes spare */ }; struct l_ifreq { union { char ifrn_name[LINUX_IFNAMSIZ]; } ifr_ifrn; union { struct l_sockaddr ifru_addr; struct l_sockaddr ifru_dstaddr; struct l_sockaddr ifru_broadaddr; struct l_sockaddr ifru_netmask; struct l_sockaddr ifru_hwaddr; l_short ifru_flags[1]; l_int ifru_ivalue; l_int ifru_mtu; struct l_ifmap ifru_map; char ifru_slave[LINUX_IFNAMSIZ]; l_uintptr_t ifru_data; } ifr_ifru; }; #define ifr_name ifr_ifrn.ifrn_name /* Interface name */ #define ifr_hwaddr ifr_ifru.ifru_hwaddr /* MAC address */ #define ifr_ifindex ifr_ifru.ifru_ivalue /* Interface index */ struct l_ifconf { int ifc_len; union { l_uintptr_t ifcu_buf; l_uintptr_t ifcu_req; } ifc_ifcu; }; #define ifc_buf ifc_ifcu.ifcu_buf #define ifc_req ifc_ifcu.ifcu_req #define LINUX_ARCH_SET_GS 0x1001 #define LINUX_ARCH_SET_FS 0x1002 #define LINUX_ARCH_GET_FS 0x1003 #define LINUX_ARCH_GET_GS 0x1004 #define LINUX_ARCH_CET_STATUS 0x3001 #define linux_copyout_rusage(r, u) copyout(r, u, sizeof(*r)) /* robust futexes */ struct linux_robust_list { l_uintptr_t next; }; struct linux_robust_list_head { struct linux_robust_list list; l_long futex_offset; l_uintptr_t pending_list; }; /* This corresponds to 'struct user_regs_struct' in Linux. */ struct linux_pt_regset { l_ulong r15; l_ulong r14; l_ulong r13; l_ulong r12; l_ulong rbp; l_ulong rbx; l_ulong r11; l_ulong r10; l_ulong r9; l_ulong r8; l_ulong rax; l_ulong rcx; l_ulong rdx; l_ulong rsi; l_ulong rdi; l_ulong orig_rax; l_ulong rip; l_ulong cs; l_ulong eflags; l_ulong rsp; l_ulong ss; l_ulong fs_base; l_ulong gs_base; l_ulong ds; l_ulong es; l_ulong fs; l_ulong gs; }; struct reg; struct syscall_info; void bsd_to_linux_regset(const struct reg *b_reg, struct linux_pt_regset *l_regset); void linux_to_bsd_regset(struct reg *b_reg, const struct linux_pt_regset *l_regset); void linux_ptrace_get_syscall_info_machdep(const struct reg *reg, struct syscall_info *si); int linux_ptrace_getregs_machdep(struct thread *td, pid_t pid, struct linux_pt_regset *l_regset); #endif /* !_AMD64_LINUX_H_ */ diff --git a/sys/amd64/linux/linux_genassym.c b/sys/amd64/linux/linux_genassym.c index 0edb6a043531..73febbe6ef40 100644 --- a/sys/amd64/linux/linux_genassym.c +++ b/sys/amd64/linux/linux_genassym.c @@ -1,15 +1,14 @@ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include -ASSYM(LINUX_RT_SIGF_HANDLER, offsetof(struct l_rt_sigframe, sf_handler)); ASSYM(LINUX_RT_SIGF_UC, offsetof(struct l_rt_sigframe, sf_sc)); ASSYM(LINUX_RT_SIGF_SC, offsetof(struct l_ucontext, uc_mcontext)); ASSYM(LINUX_VERSION_CODE, LINUX_VERSION_CODE); ASSYM(LINUX_SC_RSP, offsetof(struct l_sigcontext, sc_rsp)); diff --git a/sys/amd64/linux/linux_locore.asm b/sys/amd64/linux/linux_locore.asm index dd482cb87cc8..8b6833d7352f 100644 --- a/sys/amd64/linux/linux_locore.asm +++ b/sys/amd64/linux/linux_locore.asm @@ -1,82 +1,73 @@ /* $FreeBSD$ */ #include "linux_assym.h" /* system definitions */ #include /* miscellaneous asm macros */ #include /* system call numbers */ .data .globl linux_platform linux_platform: .asciz "x86_64" - .text -/* - * To avoid excess stack frame the signal trampoline code emulates - * the 'call' instruction. - */ + ENTRY(linux_rt_sigcode) - movq %rsp, %rbx /* preserve sigframe */ - call .getip -.getip: - popq %rax - add $.startrtsigcode-.getip, %rax /* ret address */ - pushq %rax - jmp *LINUX_RT_SIGF_HANDLER(%rbx) + movq %rsp, %rbx /* rt_sigframe for rt_sigreturn */ + call *%rcx /* call signal handler */ .startrtsigcode: - movq $LINUX_SYS_linux_rt_sigreturn,%rax /* linux_rt_sigreturn() */ - syscall /* enter kernel with args */ + movq $LINUX_SYS_linux_rt_sigreturn, %rax + syscall hlt .endrtsigcode: 0: jmp 0b #if 0 .section .note.Linux, "a",@note .long 2f - 1f /* namesz */ .balign 4 .long 4f - 3f /* descsz */ .long 0 1: .asciz "Linux" 2: .balign 4 3: .long LINUX_VERSION_CODE 4: .balign 4 .previous #endif .section .eh_frame,"a",@progbits .LSTARTFRAMEDLSI0: .long .LENDCIEDLSI0-.LSTARTCIEDLSI0 .LSTARTCIEDLSI0: .long 0 /* CIE ID */ .byte 1 /* Version number */ .string "zR" /* NULL-terminated * augmentation string */ .uleb128 1 /* Code alignment factor */ .sleb128 -4 /* Data alignment factor */ .byte 8 /* Return address register column */ .uleb128 1 /* Augmentation value length */ .byte 0x1b /* DW_EH_PE_pcrel|DW_EH_PE_sdata4. */ .byte 0x0c /* DW_CFA_def_cfa */ .uleb128 4 .uleb128 4 .byte 0x88 /* DW_CFA_offset, column 0x8 */ .uleb128 1 .align 4 .LENDCIEDLSI0: .long .LENDFDEDLSI0-.LSTARTFDEDLSI0 /* Length FDE */ .LSTARTFDEDLSI0: .long .LSTARTFDEDLSI0-.LSTARTFRAMEDLSI0 /* CIE pointer */ .long .startrtsigcode-. /* PC-relative start address */ .long .endrtsigcode-.startrtsigcode .uleb128 0 .align 4 .LENDFDEDLSI0: .previous diff --git a/sys/amd64/linux/linux_sysvec.c b/sys/amd64/linux/linux_sysvec.c index 1ea274da8935..ea91dcae5d49 100644 --- a/sys/amd64/linux/linux_sysvec.c +++ b/sys/amd64/linux/linux_sysvec.c @@ -1,1077 +1,1077 @@ /*- * 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; sa->original_code = sa->code; 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; } else sp = (caddr_t)regs->tf_rsp - 128; sp -= sizeof(struct l_rt_sigframe); /* 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 */ + regs->tf_rcx = (register_t)catcher; /* 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) { uprintf("pid %d comm %s has trashed its stack, killing\n", p->p_pid, p->p_comm); 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); tkoff = kern_cpu_selector - linux_vdso_base; ktsc_selector = (l_uintptr_t *)(linux_vdso_mapping + tkoff); *ktsc_selector = linux_vdso_cpu_selector_idx(); if (bootverbose) printf("Linux x86-64 vDSO cpu_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/linux.h b/sys/amd64/linux32/linux.h index 1273ff485dd9..ba7aeac517e1 100644 --- a/sys/amd64/linux32/linux.h +++ b/sys/amd64/linux32/linux.h @@ -1,642 +1,638 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 2004 Tim J. Robbins * Copyright (c) 2001 Doug Rabson * 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. * * $FreeBSD$ */ #ifndef _AMD64_LINUX_H_ #define _AMD64_LINUX_H_ #include #include #include #define LINUX_LEGACY_SYSCALLS #define LINUX_DTRACE linuxulator32 #define LINUX32_MAXDSIZ (512 * 1024 * 1024) /* 512MB */ #define LINUX32_MAXSSIZ (64 * 1024 * 1024) /* 64MB */ #define LINUX32_MAXVMEM 0 /* Unlimited */ /* * Provide a separate set of types for the Linux types. */ typedef int l_int; typedef int32_t l_long; typedef int64_t l_longlong; typedef short l_short; typedef unsigned int l_uint; typedef uint32_t l_ulong; typedef uint64_t l_ulonglong; typedef unsigned short l_ushort; typedef l_ulong l_uintptr_t; typedef l_long l_clock_t; typedef l_int l_daddr_t; typedef l_ushort l_dev_t; typedef l_uint l_gid_t; typedef l_ushort l_gid16_t; typedef l_ulong l_ino_t; typedef l_int l_key_t; typedef l_longlong l_loff_t; typedef l_ushort l_mode_t; typedef l_long l_off_t; typedef l_int l_pid_t; typedef l_uint l_size_t; typedef l_long l_suseconds_t; typedef l_long l_time_t; typedef l_longlong l_time64_t; typedef l_uint l_uid_t; typedef l_ushort l_uid16_t; typedef l_int l_timer_t; typedef l_int l_mqd_t; typedef l_ulong l_fd_mask; typedef struct { l_int val[2]; } l_fsid_t; typedef struct { l_time_t tv_sec; l_suseconds_t tv_usec; } l_timeval; #define l_fd_set fd_set /* * Miscellaneous */ #define LINUX_AT_COUNT 21 /* Count of used aux entry types. * Keep this synchronized with * linux_fixup_elf() code. */ struct l___sysctl_args { l_uintptr_t name; l_int nlen; l_uintptr_t oldval; l_uintptr_t oldlenp; l_uintptr_t newval; l_size_t newlen; l_ulong __spare[4]; }; /* Resource limits */ #define LINUX_RLIMIT_CPU 0 #define LINUX_RLIMIT_FSIZE 1 #define LINUX_RLIMIT_DATA 2 #define LINUX_RLIMIT_STACK 3 #define LINUX_RLIMIT_CORE 4 #define LINUX_RLIMIT_RSS 5 #define LINUX_RLIMIT_NPROC 6 #define LINUX_RLIMIT_NOFILE 7 #define LINUX_RLIMIT_MEMLOCK 8 #define LINUX_RLIMIT_AS 9 /* Address space limit */ #define LINUX_RLIM_NLIMITS 10 struct l_rlimit { l_ulong rlim_cur; l_ulong rlim_max; }; struct l_rusage { l_timeval ru_utime; l_timeval ru_stime; l_long ru_maxrss; l_long ru_ixrss; l_long ru_idrss; l_long ru_isrss; l_long ru_minflt; l_long ru_majflt; l_long ru_nswap; l_long ru_inblock; l_long ru_oublock; l_long ru_msgsnd; l_long ru_msgrcv; l_long ru_nsignals; l_long ru_nvcsw; l_long ru_nivcsw; }; struct l_mmap_argv { l_uintptr_t addr; l_size_t len; l_int prot; l_int flags; l_int fd; l_ulong pgoff; }; /* * stat family of syscalls */ struct l_timespec { l_time_t tv_sec; l_long tv_nsec; }; /* __kernel_timespec */ struct l_timespec64 { l_time64_t tv_sec; l_longlong tv_nsec; }; struct l_newstat { l_ushort st_dev; l_ushort __pad1; l_ulong st_ino; l_ushort st_mode; l_ushort st_nlink; l_ushort st_uid; l_ushort st_gid; l_ushort st_rdev; l_ushort __pad2; l_ulong st_size; l_ulong st_blksize; l_ulong st_blocks; struct l_timespec st_atim; struct l_timespec st_mtim; struct l_timespec st_ctim; l_ulong __unused4; l_ulong __unused5; }; struct l_stat { l_ushort st_dev; l_ulong st_ino; l_ushort st_mode; l_ushort st_nlink; l_ushort st_uid; l_ushort st_gid; l_ushort st_rdev; l_long st_size; struct l_timespec st_atim; struct l_timespec st_mtim; struct l_timespec st_ctim; l_long st_blksize; l_long st_blocks; l_ulong st_flags; l_ulong st_gen; }; struct l_stat64 { l_ushort st_dev; u_char __pad0[10]; l_ulong __st_ino; l_uint st_mode; l_uint st_nlink; l_ulong st_uid; l_ulong st_gid; l_ushort st_rdev; u_char __pad3[10]; l_longlong st_size; l_ulong st_blksize; l_ulong st_blocks; l_ulong __pad4; struct l_timespec st_atim; struct l_timespec st_mtim; struct l_timespec st_ctim; l_ulonglong st_ino; } __packed; struct l_statfs64 { l_int f_type; l_int f_bsize; uint64_t f_blocks; uint64_t f_bfree; uint64_t f_bavail; uint64_t f_files; uint64_t f_ffree; l_fsid_t f_fsid; l_int f_namelen; l_int f_frsize; l_int f_flags; l_int f_spare[4]; } __packed; /* sigaction flags */ #define LINUX_SA_NOCLDSTOP 0x00000001 #define LINUX_SA_NOCLDWAIT 0x00000002 #define LINUX_SA_SIGINFO 0x00000004 #define LINUX_SA_RESTORER 0x04000000 #define LINUX_SA_ONSTACK 0x08000000 #define LINUX_SA_RESTART 0x10000000 #define LINUX_SA_INTERRUPT 0x20000000 #define LINUX_SA_NOMASK 0x40000000 #define LINUX_SA_ONESHOT 0x80000000 /* sigprocmask actions */ #define LINUX_SIG_BLOCK 0 #define LINUX_SIG_UNBLOCK 1 #define LINUX_SIG_SETMASK 2 /* sigaltstack */ #define LINUX_MINSIGSTKSZ 2048 typedef l_uintptr_t l_handler_t; typedef l_ulong l_osigset_t; typedef struct { l_handler_t lsa_handler; l_osigset_t lsa_mask; l_ulong lsa_flags; l_uintptr_t lsa_restorer; } l_osigaction_t; typedef struct { l_handler_t lsa_handler; l_ulong lsa_flags; l_uintptr_t lsa_restorer; l_sigset_t lsa_mask; } __packed l_sigaction_t; typedef struct { l_uintptr_t ss_sp; l_int ss_flags; l_size_t ss_size; } l_stack_t; /* The Linux sigcontext, pretty much a standard 386 trapframe. */ struct l_sigcontext { l_uint sc_gs; l_uint sc_fs; l_uint sc_es; l_uint sc_ds; l_uint sc_edi; l_uint sc_esi; l_uint sc_ebp; l_uint sc_esp; l_uint sc_ebx; l_uint sc_edx; l_uint sc_ecx; l_uint sc_eax; l_uint sc_trapno; l_uint sc_err; l_uint sc_eip; l_uint sc_cs; l_uint sc_eflags; l_uint sc_esp_at_signal; l_uint sc_ss; l_uint sc_387; l_uint sc_mask; l_uint sc_cr2; }; struct l_ucontext { l_ulong uc_flags; l_uintptr_t uc_link; l_stack_t uc_stack; struct l_sigcontext uc_mcontext; l_sigset_t uc_sigmask; } __packed; #define LINUX_SI_MAX_SIZE 128 #define LINUX_SI_PAD_SIZE ((LINUX_SI_MAX_SIZE/sizeof(l_int)) - 3) typedef union l_sigval { l_int sival_int; l_uintptr_t sival_ptr; } l_sigval_t; typedef struct l_siginfo { l_int lsi_signo; l_int lsi_errno; l_int lsi_code; union { l_int _pad[LINUX_SI_PAD_SIZE]; struct { l_pid_t _pid; l_uid_t _uid; } _kill; struct { l_timer_t _tid; l_int _overrun; char _pad[sizeof(l_uid_t) - sizeof(l_int)]; l_sigval_t _sigval; l_int _sys_private; } _timer; struct { l_pid_t _pid; /* sender's pid */ l_uid_t _uid; /* sender's uid */ l_sigval_t _sigval; } _rt; struct { l_pid_t _pid; /* which child */ l_uid_t _uid; /* sender's uid */ l_int _status; /* exit code */ l_clock_t _utime; l_clock_t _stime; } _sigchld; struct { l_uintptr_t _addr; /* Faulting insn/memory ref. */ } _sigfault; struct { l_long _band; /* POLL_IN,POLL_OUT,POLL_MSG */ l_int _fd; } _sigpoll; } _sifields; } l_siginfo_t; #define lsi_pid _sifields._kill._pid #define lsi_uid _sifields._kill._uid #define lsi_tid _sifields._timer._tid #define lsi_overrun _sifields._timer._overrun #define lsi_sys_private _sifields._timer._sys_private #define lsi_status _sifields._sigchld._status #define lsi_utime _sifields._sigchld._utime #define lsi_stime _sifields._sigchld._stime #define lsi_value _sifields._rt._sigval #define lsi_int _sifields._rt._sigval.sival_int #define lsi_ptr _sifields._rt._sigval.sival_ptr #define lsi_addr _sifields._sigfault._addr #define lsi_band _sifields._sigpoll._band #define lsi_fd _sifields._sigpoll._fd struct l_fpreg { u_int16_t significand[4]; u_int16_t exponent; }; struct l_fpxreg { u_int16_t significand[4]; u_int16_t exponent; u_int16_t padding[3]; }; struct l_xmmreg { u_int32_t element[4]; }; struct l_fpstate { /* Regular FPU environment */ u_int32_t cw; u_int32_t sw; u_int32_t tag; u_int32_t ipoff; u_int32_t cssel; u_int32_t dataoff; u_int32_t datasel; struct l_fpreg _st[8]; u_int16_t status; u_int16_t magic; /* 0xffff = regular FPU data */ /* FXSR FPU environment */ u_int32_t _fxsr_env[6]; /* env is ignored. */ u_int32_t mxcsr; u_int32_t reserved; struct l_fpxreg _fxsr_st[8]; /* reg data is ignored. */ struct l_xmmreg _xmm[8]; u_int32_t padding[56]; }; /* * We make the stack look like Linux expects it when calling a signal * handler, but use the BSD way of calling the handler and sigreturn(). - * This means that we need to pass the pointer to the handler too. - * It is appended to the frame to not interfere with the rest of it. */ struct l_sigframe { l_int sf_sig; struct l_sigcontext sf_sc; struct l_fpstate sf_fpstate; l_uint sf_extramask[1]; - l_handler_t sf_handler; }; struct l_rt_sigframe { l_int sf_sig; l_uintptr_t sf_siginfo; l_uintptr_t sf_ucontext; l_siginfo_t sf_si; struct l_ucontext sf_sc; - l_handler_t sf_handler; } __packed; /* * arch specific open/fcntl flags */ #define LINUX_F_GETLK64 12 #define LINUX_F_SETLK64 13 #define LINUX_F_SETLKW64 14 union l_semun { l_int val; l_uintptr_t buf; l_uintptr_t array; l_uintptr_t __buf; l_uintptr_t __pad; }; struct l_ifmap { l_ulong mem_start; l_ulong mem_end; l_ushort base_addr; u_char irq; u_char dma; u_char port; /* 3 bytes spare */ }; struct l_ifreq { union { char ifrn_name[LINUX_IFNAMSIZ]; } ifr_ifrn; union { struct l_sockaddr ifru_addr; struct l_sockaddr ifru_dstaddr; struct l_sockaddr ifru_broadaddr; struct l_sockaddr ifru_netmask; struct l_sockaddr ifru_hwaddr; l_short ifru_flags[1]; l_int ifru_ivalue; l_int ifru_mtu; struct l_ifmap ifru_map; char ifru_slave[LINUX_IFNAMSIZ]; l_uintptr_t ifru_data; } ifr_ifru; }; #define ifr_name ifr_ifrn.ifrn_name /* Interface name */ #define ifr_hwaddr ifr_ifru.ifru_hwaddr /* MAC address */ #define ifr_ifindex ifr_ifru.ifru_ivalue /* Interface index */ struct l_ifconf { int ifc_len; union { l_uintptr_t ifcu_buf; l_uintptr_t ifcu_req; } ifc_ifcu; }; #define ifc_buf ifc_ifcu.ifcu_buf #define ifc_req ifc_ifcu.ifcu_req struct l_user_desc { l_uint entry_number; l_uint 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; }; #define LINUX_LOWERWORD 0x0000ffff /* * Macros which does the same thing as those in Linux include/asm-um/ldt-i386.h. * These convert Linux user space descriptor to machine one. */ #define LINUX_LDT_entry_a(info) \ ((((info)->base_addr & LINUX_LOWERWORD) << 16) | \ ((info)->limit & LINUX_LOWERWORD)) #define LINUX_ENTRY_B_READ_EXEC_ONLY 9 #define LINUX_ENTRY_B_CONTENTS 10 #define LINUX_ENTRY_B_SEG_NOT_PRESENT 15 #define LINUX_ENTRY_B_BASE_ADDR 16 #define LINUX_ENTRY_B_USEABLE 20 #define LINUX_ENTRY_B_SEG32BIT 22 #define LINUX_ENTRY_B_LIMIT 23 #define LINUX_LDT_entry_b(info) \ (((info)->base_addr & 0xff000000) | \ ((info)->limit & 0xf0000) | \ ((info)->contents << LINUX_ENTRY_B_CONTENTS) | \ (((info)->seg_not_present == 0) << LINUX_ENTRY_B_SEG_NOT_PRESENT) | \ (((info)->base_addr & 0x00ff0000) >> LINUX_ENTRY_B_BASE_ADDR) | \ (((info)->read_exec_only == 0) << LINUX_ENTRY_B_READ_EXEC_ONLY) | \ ((info)->seg_32bit << LINUX_ENTRY_B_SEG32BIT) | \ ((info)->useable << LINUX_ENTRY_B_USEABLE) | \ ((info)->limit_in_pages << LINUX_ENTRY_B_LIMIT) | 0x7000) #define LINUX_LDT_empty(info) \ ((info)->base_addr == 0 && \ (info)->limit == 0 && \ (info)->contents == 0 && \ (info)->seg_not_present == 1 && \ (info)->read_exec_only == 1 && \ (info)->seg_32bit == 0 && \ (info)->limit_in_pages == 0 && \ (info)->useable == 0) /* * Macros for converting segments. * They do the same as those in arch/i386/kernel/process.c in Linux. */ #define LINUX_GET_BASE(desc) \ ((((desc)->a >> 16) & LINUX_LOWERWORD) | \ (((desc)->b << 16) & 0x00ff0000) | \ ((desc)->b & 0xff000000)) #define LINUX_GET_LIMIT(desc) \ (((desc)->a & LINUX_LOWERWORD) | \ ((desc)->b & 0xf0000)) #define LINUX_GET_32BIT(desc) \ (((desc)->b >> LINUX_ENTRY_B_SEG32BIT) & 1) #define LINUX_GET_CONTENTS(desc) \ (((desc)->b >> LINUX_ENTRY_B_CONTENTS) & 3) #define LINUX_GET_WRITABLE(desc) \ (((desc)->b >> LINUX_ENTRY_B_READ_EXEC_ONLY) & 1) #define LINUX_GET_LIMIT_PAGES(desc) \ (((desc)->b >> LINUX_ENTRY_B_LIMIT) & 1) #define LINUX_GET_PRESENT(desc) \ (((desc)->b >> LINUX_ENTRY_B_SEG_NOT_PRESENT) & 1) #define LINUX_GET_USEABLE(desc) \ (((desc)->b >> LINUX_ENTRY_B_USEABLE) & 1) struct iovec; struct uio; struct l_iovec32 { uint32_t iov_base; l_size_t iov_len; }; int linux32_copyiniov(struct l_iovec32 *iovp32, l_ulong iovcnt, struct iovec **iovp, int error); int linux32_copyinuio(struct l_iovec32 *iovp, l_ulong iovcnt, struct uio **uiop); int linux_copyout_rusage(struct rusage *ru, void *uaddr); /* robust futexes */ struct linux_robust_list { l_uintptr_t next; }; struct linux_robust_list_head { struct linux_robust_list list; l_long futex_offset; l_uintptr_t pending_list; }; /* This corresponds to 'struct user_regs_struct32' in Linux. */ struct linux_pt_regset32 { l_uint ebx; l_uint ecx; l_uint edx; l_uint esi; l_uint edi; l_uint ebp; l_uint eax; l_uint ds; l_uint es; l_uint fs; l_uint gs; l_uint orig_eax; l_uint eip; l_uint cs; l_uint eflags; l_uint esp; l_uint ss; }; struct reg32; void bsd_to_linux_regset32(const struct reg32 *b_reg, struct linux_pt_regset32 *l_regset); extern bool linux32_emulate_i386; #endif /* !_AMD64_LINUX_H_ */ diff --git a/sys/amd64/linux32/linux32_genassym.c b/sys/amd64/linux32/linux32_genassym.c index bc94139c226d..ca618c01ffab 100644 --- a/sys/amd64/linux32/linux32_genassym.c +++ b/sys/amd64/linux32/linux32_genassym.c @@ -1,18 +1,16 @@ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include -ASSYM(LINUX_SIGF_HANDLER, offsetof(struct l_sigframe, sf_handler)); ASSYM(LINUX_SIGF_SC, offsetof(struct l_sigframe, sf_sc)); -ASSYM(LINUX_RT_SIGF_HANDLER, offsetof(struct l_rt_sigframe, sf_handler)); ASSYM(LINUX_RT_SIGF_UC, offsetof(struct l_rt_sigframe, sf_sc)); ASSYM(LINUX_RT_SIGF_SC, offsetof(struct l_ucontext, uc_mcontext)); ASSYM(LINUX_VERSION_CODE, LINUX_VERSION_CODE); ASSYM(LINUX_SC_ESP, offsetof(struct l_sigcontext, sc_esp)); diff --git a/sys/amd64/linux32/linux32_locore.asm b/sys/amd64/linux32/linux32_locore.asm index f96b3e730f9f..e0409969ee3b 100644 --- a/sys/amd64/linux32/linux32_locore.asm +++ b/sys/amd64/linux32/linux32_locore.asm @@ -1,156 +1,140 @@ /* $FreeBSD$ */ #include "linux32_assym.h" /* system definitions */ #include /* miscellaneous asm macros */ #include /* system call numbers */ .data .globl linux_platform linux_platform: .asciz "i686" .text .code32 -/* - * To avoid excess stack frame the signal trampoline code emulates - * the 'call' instruction. - */ ENTRY(__kernel_sigreturn) - movl %esp, %ebx /* preserve sigframe */ - call .getip0 -.getip0: - popl %eax - add $.startsigcode-.getip0, %eax /* ret address */ - push %eax - jmp *LINUX_SIGF_HANDLER(%ebx) + movl %esp, %ebx /* sigframe for sigreturn */ + call *%edi /* call signal handler */ .startsigcode: - popl %eax - movl $LINUX32_SYS_linux_sigreturn,%eax /* linux_sigreturn() */ - int $0x80 /* enter kernel with args */ + popl %eax /* gcc unwind code need this */ + movl $LINUX32_SYS_linux_sigreturn, %eax + int $0x80 .endsigcode: 0: jmp 0b ENTRY(__kernel_rt_sigreturn) - leal LINUX_RT_SIGF_UC(%esp),%ebx /* linux ucp */ - leal LINUX_RT_SIGF_SC(%ebx),%ecx /* linux sigcontext */ - movl %esp, %edi - call .getip1 -.getip1: - popl %eax - add $.startrtsigcode-.getip1, %eax /* ret address */ - push %eax - jmp *LINUX_RT_SIGF_HANDLER(%edi) + leal LINUX_RT_SIGF_UC(%esp), %ebx /* linux ucontext for rt_sigreturn */ + call *%edi /* call signal handler */ .startrtsigcode: - movl $LINUX32_SYS_linux_rt_sigreturn,%eax /* linux_rt_sigreturn() */ - int $0x80 /* enter kernel with args */ + movl $LINUX32_SYS_linux_rt_sigreturn, %eax + int $0x80 .endrtsigcode: 0: jmp 0b ENTRY(__kernel_vsyscall) .startvsyscall: int $0x80 ret .endvsyscall: #if 0 .section .note.Linux, "a",@note .long 2f - 1f /* namesz */ .balign 4 .long 4f - 3f /* descsz */ .long 0 1: .asciz "Linux" 2: .balign 4 3: .long LINUX_VERSION_CODE 4: .balign 4 .previous #endif #define do_cfa_expr(offset) \ .byte 0x0f; /* DW_CFA_def_cfa_expression */ \ .uleb128 11f-10f; /* length */ \ 10: .byte 0x74; /* DW_OP_breg4 */ \ .sleb128 offset; /* offset */ \ .byte 0x06; /* DW_OP_deref */ \ 11: /* CIE */ .section .eh_frame,"a",@progbits .LSTARTFRAMEDLSI1: .long .LENDCIEDLSI1-.LSTARTCIEDLSI1 .LSTARTCIEDLSI1: .long 0 /* CIE ID */ .byte 1 /* Version number */ .string "zRS" /* NULL-terminated * augmentation string */ .uleb128 1 /* Code alignment factor */ .sleb128 -4 /* Data alignment factor */ .byte 8 /* Return address * register column */ .uleb128 1 /* Augmentation value length */ .byte 0x1b /* DW_EH_PE_pcrel|DW_EH_PE_sdata4. */ .byte 0 /* DW_CFA_nop */ .align 4 .LENDCIEDLSI1: /* FDE */ .long .LENDFDEDLSI1-.LSTARTFDEDLSI1 /* Length FDE */ .LSTARTFDEDLSI1: .long .LSTARTFDEDLSI1-.LSTARTFRAMEDLSI1 /* CIE pointer */ .long .startsigcode-. /* PC-relative start address */ .long .endsigcode-.startsigcode .uleb128 0 /* Augmentation */ do_cfa_expr(LINUX_SIGF_SC-8) .align 4 .LENDFDEDLSI1: .long .LENDFDEDLSI2-.LSTARTFDEDLSI2 /* Length FDE */ .LSTARTFDEDLSI2: .long .LSTARTFDEDLSI2-.LSTARTFRAMEDLSI1 /* CIE pointer */ .long .startrtsigcode-. /* PC-relative start address */ .long .endrtsigcode-.startrtsigcode .uleb128 0 /* Augmentation */ do_cfa_expr(LINUX_RT_SIGF_SC-4+LINUX_SC_ESP) .align 4 .LENDFDEDLSI2: .previous .section .eh_frame,"a",@progbits .LSTARTFRAMEDLSI2: .long .LENDCIEDLSI2-.LSTARTCIEDLSI2 .LSTARTCIEDLSI2: .long 0 /* CIE ID */ .byte 1 /* Version number */ .string "zR" /* NULL-terminated * augmentation string */ .uleb128 1 /* Code alignment factor */ .sleb128 -4 /* Data alignment factor */ .byte 8 /* Return address register column */ .uleb128 1 /* Augmentation value length */ .byte 0x1b /* DW_EH_PE_pcrel|DW_EH_PE_sdata4. */ .byte 0x0c /* DW_CFA_def_cfa */ .uleb128 4 .uleb128 4 .byte 0x88 /* DW_CFA_offset, column 0x8 */ .uleb128 1 .align 4 .LENDCIEDLSI2: .long .LENDFDEDLSI3-.LSTARTFDEDLSI3 /* Length FDE */ .LSTARTFDEDLSI3: .long .LSTARTFDEDLSI3-.LSTARTFRAMEDLSI2 /* CIE pointer */ .long .startvsyscall-. /* PC-relative start address */ .long .endvsyscall-.startvsyscall .uleb128 0 .align 4 .LENDFDEDLSI3: .previous diff --git a/sys/amd64/linux32/linux32_sysvec.c b/sys/amd64/linux32/linux32_sysvec.c index ed430c8e579a..b2d0f2401bf9 100644 --- a/sys/amd64/linux32/linux32_sysvec.c +++ b/sys/amd64/linux32/linux32_sysvec.c @@ -1,1235 +1,1235 @@ /*- * 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_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_rdi = PTROUT(catcher); 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_rdi = PTROUT(catcher); 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; sa->original_code = sa->code; 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 *)PROC_PS_STRINGS(imgp->proc); 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, ""); bool linux32_emulate_i386 = false; SYSCTL_BOOL(_compat_linux32, OID_AUTO, emulate_i386, CTLFLAG_RWTUN, &linux32_emulate_i386, 0, "Emulate the real i386"); 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); tkoff = kern_cpu_selector - linux_vdso_base; ktsc_selector = (l_uintptr_t *)(linux_vdso_mapping + tkoff); *ktsc_selector = linux_vdso_cpu_selector_idx(); if (bootverbose) printf("Linux i386 vDSO cpu_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/i386/linux/linux.h b/sys/i386/linux/linux.h index 7fed468239a8..c26294ddc178 100644 --- a/sys/i386/linux/linux.h +++ b/sys/i386/linux/linux.h @@ -1,571 +1,567 @@ /*- * 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. * $FreeBSD$ */ #ifndef _I386_LINUX_H_ #define _I386_LINUX_H_ #include #include #include #define LINUX_LEGACY_SYSCALLS #define LINUX_DTRACE linuxulator /* * Provide a separate set of types for the Linux types. */ typedef int l_int; typedef int32_t l_long; typedef int64_t l_longlong; typedef short l_short; typedef unsigned int l_uint; typedef uint32_t l_ulong; typedef uint64_t l_ulonglong; typedef unsigned short l_ushort; typedef char *l_caddr_t; typedef l_ulong l_uintptr_t; typedef l_long l_clock_t; typedef l_int l_daddr_t; typedef l_ushort l_dev_t; typedef l_uint l_gid_t; typedef l_ushort l_gid16_t; typedef l_ulong l_ino_t; typedef l_int l_key_t; typedef l_longlong l_loff_t; typedef l_ushort l_mode_t; typedef l_long l_off_t; typedef l_int l_pid_t; typedef l_uint l_size_t; typedef l_long l_suseconds_t; typedef l_long l_time_t; typedef l_longlong l_time64_t; typedef l_uint l_uid_t; typedef l_ushort l_uid16_t; typedef l_int l_timer_t; typedef l_int l_mqd_t; typedef l_ulong l_fd_mask; typedef struct { l_int val[2]; } l_fsid_t; typedef struct { l_time_t tv_sec; l_suseconds_t tv_usec; } l_timeval; #define l_fd_set fd_set /* * Miscellaneous */ #define LINUX_AT_COUNT 20 /* Count of used aux entry types. * Keep this synchronized with * linux_fixup_elf() code. */ struct l___sysctl_args { l_int *name; l_int nlen; void *oldval; l_size_t *oldlenp; void *newval; l_size_t newlen; l_ulong __spare[4]; }; /* Resource limits */ #define LINUX_RLIMIT_CPU 0 #define LINUX_RLIMIT_FSIZE 1 #define LINUX_RLIMIT_DATA 2 #define LINUX_RLIMIT_STACK 3 #define LINUX_RLIMIT_CORE 4 #define LINUX_RLIMIT_RSS 5 #define LINUX_RLIMIT_NPROC 6 #define LINUX_RLIMIT_NOFILE 7 #define LINUX_RLIMIT_MEMLOCK 8 #define LINUX_RLIMIT_AS 9 /* Address space limit */ #define LINUX_RLIM_NLIMITS 10 struct l_rlimit { l_ulong rlim_cur; l_ulong rlim_max; }; struct l_mmap_argv { l_uintptr_t addr; l_size_t len; l_int prot; l_int flags; l_int fd; l_off_t pgoff; }; /* * stat family of syscalls */ struct l_timespec { l_time_t tv_sec; l_long tv_nsec; }; /* __kernel_timespec */ struct l_timespec64 { l_time64_t tv_sec; l_longlong tv_nsec; }; struct l_newstat { l_ushort st_dev; l_ushort __pad1; l_ulong st_ino; l_ushort st_mode; l_ushort st_nlink; l_ushort st_uid; l_ushort st_gid; l_ushort st_rdev; l_ushort __pad2; l_ulong st_size; l_ulong st_blksize; l_ulong st_blocks; struct l_timespec st_atim; struct l_timespec st_mtim; struct l_timespec st_ctim; l_ulong __unused4; l_ulong __unused5; }; struct l_stat { l_ushort st_dev; l_ulong st_ino; l_ushort st_mode; l_ushort st_nlink; l_ushort st_uid; l_ushort st_gid; l_ushort st_rdev; l_long st_size; struct l_timespec st_atim; struct l_timespec st_mtim; struct l_timespec st_ctim; l_long st_blksize; l_long st_blocks; l_ulong st_flags; l_ulong st_gen; }; struct l_stat64 { l_ushort st_dev; u_char __pad0[10]; l_ulong __st_ino; l_uint st_mode; l_uint st_nlink; l_ulong st_uid; l_ulong st_gid; l_ushort st_rdev; u_char __pad3[10]; l_longlong st_size; l_ulong st_blksize; l_ulong st_blocks; l_ulong __pad4; struct l_timespec st_atim; struct l_timespec st_mtim; struct l_timespec st_ctim; l_ulonglong st_ino; }; struct l_statfs64 { l_int f_type; l_int f_bsize; uint64_t f_blocks; uint64_t f_bfree; uint64_t f_bavail; uint64_t f_files; uint64_t f_ffree; l_fsid_t f_fsid; l_int f_namelen; l_int f_frsize; l_int f_flags; l_int f_spare[4]; }; #define LINUX_NSIG_WORDS 2 /* sigaction flags */ #define LINUX_SA_NOCLDSTOP 0x00000001 #define LINUX_SA_NOCLDWAIT 0x00000002 #define LINUX_SA_SIGINFO 0x00000004 #define LINUX_SA_RESTORER 0x04000000 #define LINUX_SA_ONSTACK 0x08000000 #define LINUX_SA_RESTART 0x10000000 #define LINUX_SA_INTERRUPT 0x20000000 #define LINUX_SA_NOMASK 0x40000000 #define LINUX_SA_ONESHOT 0x80000000 /* sigprocmask actions */ #define LINUX_SIG_BLOCK 0 #define LINUX_SIG_UNBLOCK 1 #define LINUX_SIG_SETMASK 2 /* sigaltstack */ #define LINUX_MINSIGSTKSZ 2048 typedef void (*l_handler_t)(l_int); typedef l_ulong l_osigset_t; typedef struct { l_handler_t lsa_handler; l_osigset_t lsa_mask; l_ulong lsa_flags; void (*lsa_restorer)(void); } l_osigaction_t; typedef struct { l_handler_t lsa_handler; l_ulong lsa_flags; void (*lsa_restorer)(void); l_sigset_t lsa_mask; } l_sigaction_t; typedef struct { l_uintptr_t ss_sp; l_int ss_flags; l_size_t ss_size; } l_stack_t; /* The Linux sigcontext, pretty much a standard 386 trapframe. */ struct l_sigcontext { l_int sc_gs; l_int sc_fs; l_int sc_es; l_int sc_ds; l_int sc_edi; l_int sc_esi; l_int sc_ebp; l_int sc_esp; l_int sc_ebx; l_int sc_edx; l_int sc_ecx; l_int sc_eax; l_int sc_trapno; l_int sc_err; l_int sc_eip; l_int sc_cs; l_int sc_eflags; l_int sc_esp_at_signal; l_int sc_ss; l_int sc_387; l_int sc_mask; l_int sc_cr2; }; struct l_ucontext { l_ulong uc_flags; void *uc_link; l_stack_t uc_stack; struct l_sigcontext uc_mcontext; l_sigset_t uc_sigmask; }; #define LINUX_SI_MAX_SIZE 128 #define LINUX_SI_PAD_SIZE ((LINUX_SI_MAX_SIZE/sizeof(l_int)) - 3) typedef union l_sigval { l_int sival_int; l_uintptr_t sival_ptr; } l_sigval_t; typedef struct l_siginfo { l_int lsi_signo; l_int lsi_errno; l_int lsi_code; union { l_int _pad[LINUX_SI_PAD_SIZE]; struct { l_pid_t _pid; l_uid_t _uid; } _kill; struct { l_timer_t _tid; l_int _overrun; char _pad[sizeof(l_uid_t) - sizeof(l_int)]; l_sigval_t _sigval; l_int _sys_private; } _timer; struct { l_pid_t _pid; /* sender's pid */ l_uid_t _uid; /* sender's uid */ l_sigval_t _sigval; } _rt; struct { l_pid_t _pid; /* which child */ l_uid_t _uid; /* sender's uid */ l_int _status; /* exit code */ l_clock_t _utime; l_clock_t _stime; } _sigchld; struct { l_uintptr_t _addr; /* Faulting insn/memory ref. */ } _sigfault; struct { l_long _band; /* POLL_IN,POLL_OUT,POLL_MSG */ l_int _fd; } _sigpoll; } _sifields; } l_siginfo_t; #define lsi_pid _sifields._kill._pid #define lsi_uid _sifields._kill._uid #define lsi_tid _sifields._timer._tid #define lsi_overrun _sifields._timer._overrun #define lsi_sys_private _sifields._timer._sys_private #define lsi_status _sifields._sigchld._status #define lsi_utime _sifields._sigchld._utime #define lsi_stime _sifields._sigchld._stime #define lsi_value _sifields._rt._sigval #define lsi_int _sifields._rt._sigval.sival_int #define lsi_ptr _sifields._rt._sigval.sival_ptr #define lsi_addr _sifields._sigfault._addr #define lsi_band _sifields._sigpoll._band #define lsi_fd _sifields._sigpoll._fd struct l_fpreg { u_int16_t significand[4]; u_int16_t exponent; }; struct l_fpxreg { u_int16_t significand[4]; u_int16_t exponent; u_int16_t padding[3]; }; struct l_xmmreg { u_int32_t element[4]; }; struct l_fpstate { /* Regular FPU environment */ u_int32_t cw; u_int32_t sw; u_int32_t tag; u_int32_t ipoff; u_int32_t cssel; u_int32_t dataoff; u_int32_t datasel; struct l_fpreg _st[8]; u_int16_t status; u_int16_t magic; /* 0xffff = regular FPU data */ /* FXSR FPU environment */ u_int32_t _fxsr_env[6]; /* env is ignored. */ u_int32_t mxcsr; u_int32_t reserved; struct l_fpxreg _fxsr_st[8]; /* reg data is ignored. */ struct l_xmmreg _xmm[8]; u_int32_t padding[56]; }; /* * We make the stack look like Linux expects it when calling a signal * handler, but use the BSD way of calling the handler and sigreturn(). - * This means that we need to pass the pointer to the handler too. - * It is appended to the frame to not interfere with the rest of it. */ struct l_sigframe { l_int sf_sig; struct l_sigcontext sf_sc; struct l_fpstate sf_fpstate; l_uint sf_extramask[LINUX_NSIG_WORDS-1]; - l_handler_t sf_handler; }; struct l_rt_sigframe { l_int sf_sig; l_siginfo_t *sf_siginfo; struct l_ucontext *sf_ucontext; l_siginfo_t sf_si; struct l_ucontext sf_sc; - l_handler_t sf_handler; }; extern struct sysentvec linux_sysvec; /* * arch specific open/fcntl flags */ #define LINUX_F_GETLK64 12 #define LINUX_F_SETLK64 13 #define LINUX_F_SETLKW64 14 union l_semun { l_int val; l_uintptr_t buf; l_ushort *array; l_uintptr_t __buf; l_uintptr_t __pad; }; struct l_ifmap { l_ulong mem_start; l_ulong mem_end; l_ushort base_addr; u_char irq; u_char dma; u_char port; }; struct l_ifreq { union { char ifrn_name[LINUX_IFNAMSIZ]; } ifr_ifrn; union { struct l_sockaddr ifru_addr; struct l_sockaddr ifru_dstaddr; struct l_sockaddr ifru_broadaddr; struct l_sockaddr ifru_netmask; struct l_sockaddr ifru_hwaddr; l_short ifru_flags[1]; l_int ifru_ivalue; l_int ifru_mtu; struct l_ifmap ifru_map; char ifru_slave[LINUX_IFNAMSIZ]; l_caddr_t ifru_data; } ifr_ifru; }; #define ifr_name ifr_ifrn.ifrn_name /* Interface name */ #define ifr_hwaddr ifr_ifru.ifru_hwaddr /* MAC address */ #define ifr_ifindex ifr_ifru.ifru_ivalue /* Interface index */ struct l_user_desc { l_uint entry_number; l_uint 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_desc_struct { unsigned long a, b; }; #define LINUX_LOWERWORD 0x0000ffff /* * Macros which does the same thing as those in Linux include/asm-um/ldt-i386.h. * These convert Linux user space descriptor to machine one. */ #define LINUX_LDT_entry_a(info) \ ((((info)->base_addr & LINUX_LOWERWORD) << 16) | \ ((info)->limit & LINUX_LOWERWORD)) #define LINUX_ENTRY_B_READ_EXEC_ONLY 9 #define LINUX_ENTRY_B_CONTENTS 10 #define LINUX_ENTRY_B_SEG_NOT_PRESENT 15 #define LINUX_ENTRY_B_BASE_ADDR 16 #define LINUX_ENTRY_B_USEABLE 20 #define LINUX_ENTRY_B_SEG32BIT 22 #define LINUX_ENTRY_B_LIMIT 23 #define LINUX_LDT_entry_b(info) \ (((info)->base_addr & 0xff000000) | \ ((info)->limit & 0xf0000) | \ ((info)->contents << LINUX_ENTRY_B_CONTENTS) | \ (((info)->seg_not_present == 0) << LINUX_ENTRY_B_SEG_NOT_PRESENT) | \ (((info)->base_addr & 0x00ff0000) >> LINUX_ENTRY_B_BASE_ADDR) | \ (((info)->read_exec_only == 0) << LINUX_ENTRY_B_READ_EXEC_ONLY) | \ ((info)->seg_32bit << LINUX_ENTRY_B_SEG32BIT) | \ ((info)->useable << LINUX_ENTRY_B_USEABLE) | \ ((info)->limit_in_pages << LINUX_ENTRY_B_LIMIT) | 0x7000) #define LINUX_LDT_empty(info) \ ((info)->base_addr == 0 && \ (info)->limit == 0 && \ (info)->contents == 0 && \ (info)->seg_not_present == 1 && \ (info)->read_exec_only == 1 && \ (info)->seg_32bit == 0 && \ (info)->limit_in_pages == 0 && \ (info)->useable == 0) /* * Macros for converting segments. * They do the same as those in arch/i386/kernel/process.c in Linux. */ #define LINUX_GET_BASE(desc) \ ((((desc)->a >> 16) & LINUX_LOWERWORD) | \ (((desc)->b << 16) & 0x00ff0000) | \ ((desc)->b & 0xff000000)) #define LINUX_GET_LIMIT(desc) \ (((desc)->a & LINUX_LOWERWORD) | \ ((desc)->b & 0xf0000)) #define LINUX_GET_32BIT(desc) \ (((desc)->b >> LINUX_ENTRY_B_SEG32BIT) & 1) #define LINUX_GET_CONTENTS(desc) \ (((desc)->b >> LINUX_ENTRY_B_CONTENTS) & 3) #define LINUX_GET_WRITABLE(desc) \ (((desc)->b >> LINUX_ENTRY_B_READ_EXEC_ONLY) & 1) #define LINUX_GET_LIMIT_PAGES(desc) \ (((desc)->b >> LINUX_ENTRY_B_LIMIT) & 1) #define LINUX_GET_PRESENT(desc) \ (((desc)->b >> LINUX_ENTRY_B_SEG_NOT_PRESENT) & 1) #define LINUX_GET_USEABLE(desc) \ (((desc)->b >> LINUX_ENTRY_B_USEABLE) & 1) #define linux_copyout_rusage(r, u) copyout(r, u, sizeof(*r)) /* robust futexes */ struct linux_robust_list { struct linux_robust_list *next; }; struct linux_robust_list_head { struct linux_robust_list list; l_long futex_offset; struct linux_robust_list *pending_list; }; #endif /* !_I386_LINUX_H_ */ diff --git a/sys/i386/linux/linux_genassym.c b/sys/i386/linux/linux_genassym.c index 9735110b0472..08aa0979d793 100644 --- a/sys/i386/linux/linux_genassym.c +++ b/sys/i386/linux/linux_genassym.c @@ -1,19 +1,17 @@ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include -ASSYM(LINUX_SIGF_HANDLER, offsetof(struct l_sigframe, sf_handler)); ASSYM(LINUX_SIGF_SC, offsetof(struct l_sigframe, sf_sc)); ASSYM(LINUX_SC_GS, offsetof(struct l_sigcontext, sc_gs)); ASSYM(LINUX_SC_EFLAGS, offsetof(struct l_sigcontext, sc_eflags)); -ASSYM(LINUX_RT_SIGF_HANDLER, offsetof(struct l_rt_sigframe, sf_handler)); ASSYM(LINUX_RT_SIGF_UC, offsetof(struct l_rt_sigframe, sf_sc)); ASSYM(LINUX_RT_SIGF_SC, offsetof(struct l_ucontext, uc_mcontext)); ASSYM(LINUX_SC_ESP, offsetof(struct l_sigcontext, sc_esp)); ASSYM(LINUX_VERSION_CODE, LINUX_VERSION_CODE); diff --git a/sys/i386/linux/linux_locore.asm b/sys/i386/linux/linux_locore.asm index 4c5246bd5725..3459fb5a56de 100644 --- a/sys/i386/linux/linux_locore.asm +++ b/sys/i386/linux/linux_locore.asm @@ -1,157 +1,141 @@ /* $FreeBSD$ */ #include "linux_assym.h" /* system definitions */ #include /* miscellaneous asm macros */ #include /* system call numbers */ #include "assym.inc" .data .globl linux_platform linux_platform: .asciz "i686" .text -/* - * To avoid excess stack frame the signal trampoline code emulates - * the 'call' instruction. - */ ENTRY(__kernel_sigreturn) - movl %esp, %ebx /* preserve sigframe */ - call .getip0 -.getip0: - popl %eax - add $.startsigcode-.getip0, %eax /* ret address */ - push %eax - jmp *LINUX_SIGF_HANDLER(%ebx) + movl %esp, %ebx /* sigframe for sigreturn */ + call *%edi /* call signal handler */ .startsigcode: popl %eax /* gcc unwind code need this */ - movl $LINUX_SYS_linux_sigreturn,%eax /* linux_sigreturn() */ - int $0x80 /* enter kernel with args */ + movl $LINUX_SYS_linux_sigreturn, %eax + int $0x80 .endsigcode: 0: jmp 0b ENTRY(__kernel_rt_sigreturn) - leal LINUX_RT_SIGF_UC(%esp),%ebx /* linux ucp */ - leal LINUX_RT_SIGF_SC(%ebx),%ecx /* linux sigcontext */ - movl %esp, %edi - call .getip1 -.getip1: - popl %eax - add $.startrtsigcode-.getip1, %eax /* ret address */ - push %eax - jmp *LINUX_RT_SIGF_HANDLER(%edi) + leal LINUX_RT_SIGF_UC(%esp), %ebx /* linux ucontext for rt_sigreturn */ + call *%edi /* call signal handler */ .startrtsigcode: - movl $LINUX_SYS_linux_rt_sigreturn,%eax /* linux_rt_sigreturn() */ - int $0x80 /* enter kernel with args */ + movl $LINUX_SYS_linux_rt_sigreturn, %eax + int $0x80 .endrtsigcode: 0: jmp 0b ENTRY(__kernel_vsyscall) .startvsyscall: int $0x80 ret .endvsyscall: #if 0 .section .note.Linux, "a",@note .long 2f - 1f /* namesz */ .balign 4 .long 4f - 3f /* descsz */ .long 0 1: .asciz "Linux" 2: .balign 4 3: .long LINUX_VERSION_CODE 4: .balign 4 .previous #endif #define do_cfa_expr(offset) \ .byte 0x0f; /* DW_CFA_def_cfa_expression */ \ .uleb128 11f-10f; /* length */ \ 10: .byte 0x74; /* DW_OP_breg4 */ \ .sleb128 offset; /* offset */ \ .byte 0x06; /* DW_OP_deref */ \ 11: /* CIE */ .section .eh_frame,"a",@progbits .LSTARTFRAMEDLSI1: .long .LENDCIEDLSI1-.LSTARTCIEDLSI1 .LSTARTCIEDLSI1: .long 0 /* CIE ID */ .byte 1 /* Version number */ .string "zRS" /* NULL-terminated * augmentation string */ .uleb128 1 /* Code alignment factor */ .sleb128 -4 /* Data alignment factor */ .byte 8 /* Return address * register column */ .uleb128 1 /* Augmentation value length */ .byte 0x1b /* DW_EH_PE_pcrel|DW_EH_PE_sdata4. */ .byte 0 /* DW_CFA_nop */ .align 4 .LENDCIEDLSI1: /* FDE */ .long .LENDFDEDLSI1-.LSTARTFDEDLSI1 /* Length FDE */ .LSTARTFDEDLSI1: .long .LSTARTFDEDLSI1-.LSTARTFRAMEDLSI1 /* CIE pointer */ .long .startsigcode-. /* PC-relative start address */ .long .endsigcode-.startsigcode .uleb128 0 /* Augmentation */ do_cfa_expr(LINUX_SIGF_SC-8) .align 4 .LENDFDEDLSI1: .long .LENDFDEDLSI2-.LSTARTFDEDLSI2 /* Length FDE */ .LSTARTFDEDLSI2: .long .LSTARTFDEDLSI2-.LSTARTFRAMEDLSI1 /* CIE pointer */ .long .startrtsigcode-. /* PC-relative start address */ .long .endrtsigcode-.startrtsigcode .uleb128 0 /* Augmentation */ do_cfa_expr(LINUX_RT_SIGF_SC-4+LINUX_SC_ESP) .align 4 .LENDFDEDLSI2: .previous .section .eh_frame,"a",@progbits .LSTARTFRAMEDLSI2: .long .LENDCIEDLSI2-.LSTARTCIEDLSI2 .LSTARTCIEDLSI2: .long 0 /* CIE ID */ .byte 1 /* Version number */ .string "zR" /* NULL-terminated * augmentation string */ .uleb128 1 /* Code alignment factor */ .sleb128 -4 /* Data alignment factor */ .byte 8 /* Return address register column */ .uleb128 1 /* Augmentation value length */ .byte 0x1b /* DW_EH_PE_pcrel|DW_EH_PE_sdata4. */ .byte 0x0c /* DW_CFA_def_cfa */ .uleb128 4 .uleb128 4 .byte 0x88 /* DW_CFA_offset, column 0x8 */ .uleb128 1 .align 4 .LENDCIEDLSI2: .long .LENDFDEDLSI3-.LSTARTFDEDLSI3 /* Length FDE */ .LSTARTFDEDLSI3: .long .LSTARTFDEDLSI3-.LSTARTFRAMEDLSI2 /* CIE pointer */ .long .startvsyscall-. /* PC-relative start address */ .long .endvsyscall-.startvsyscall .uleb128 0 .align 4 .LENDFDEDLSI3: .previous diff --git a/sys/i386/linux/linux_sysvec.c b/sys/i386/linux/linux_sysvec.c index 9f0647f73c0c..d37250d7552f 100644 --- a/sys/i386/linux/linux_sysvec.c +++ b/sys/i386/linux/linux_sysvec.c @@ -1,1180 +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) { Elf32_Auxargs *args; Elf32_Auxinfo *argarray, *pos; int error, issetugid; issetugid = imgp->proc->p_flag & P_SUGID ? 1 : 0; 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_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_edi = catcher; 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; int oonstack; PROC_LOCK_ASSERT(p, MA_OWNED); psp = p->p_sigacts; sig = ksi->ksi_signo; 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_edi = catcher; 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->original_code = sa->code; 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); tkoff = kern_cpu_selector - linux_vdso_base; ktsc_selector = (l_uintptr_t *)(linux_vdso_mapping + tkoff); *ktsc_selector = linux_vdso_cpu_selector_idx(); if (bootverbose) printf("Linux i386 vDSO cpu_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");