diff --git a/sys/amd64/amd64/elf_machdep.c b/sys/amd64/amd64/elf_machdep.c
index a3fb1184bb05..a7cef2a8d92a 100644
--- a/sys/amd64/amd64/elf_machdep.c
+++ b/sys/amd64/amd64/elf_machdep.c
@@ -1,426 +1,432 @@
/*-
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright 1996-1998 John D. Polstra.
* 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 ``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 <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/kernel.h>
#include <sys/systm.h>
#include <sys/exec.h>
#include <sys/imgact.h>
#include <sys/linker.h>
#include <sys/proc.h>
#include <sys/sysent.h>
#include <sys/imgact_elf.h>
#include <sys/syscall.h>
#include <sys/signalvar.h>
#include <sys/vnode.h>
#include <vm/vm.h>
#include <vm/pmap.h>
#include <vm/vm_param.h>
#include <machine/elf.h>
#include <machine/fpu.h>
#include <machine/md_var.h>
struct sysentvec elf64_freebsd_sysvec_la48 = {
.sv_size = SYS_MAXSYSCALL,
.sv_table = sysent,
.sv_transtrap = NULL,
.sv_fixup = __elfN(freebsd_fixup),
.sv_sendsig = sendsig,
.sv_sigcode = sigcode,
.sv_szsigcode = &szsigcode,
.sv_name = "FreeBSD ELF64",
.sv_coredump = __elfN(coredump),
+ .sv_elf_core_osabi = ELFOSABI_FREEBSD,
+ .sv_elf_core_abi_vendor = FREEBSD_ABI_VENDOR,
+ .sv_elf_core_prepare_notes = __elfN(prepare_notes),
.sv_imgact_try = NULL,
.sv_minsigstksz = MINSIGSTKSZ,
.sv_minuser = VM_MIN_ADDRESS,
.sv_maxuser = VM_MAXUSER_ADDRESS_LA48,
.sv_usrstack = USRSTACK_LA48,
.sv_psstrings = PS_STRINGS_LA48,
.sv_stackprot = VM_PROT_ALL,
.sv_copyout_auxargs = __elfN(freebsd_copyout_auxargs),
.sv_copyout_strings = exec_copyout_strings,
.sv_setregs = exec_setregs,
.sv_fixlimit = NULL,
.sv_maxssiz = NULL,
.sv_flags = SV_ABI_FREEBSD | SV_ASLR | SV_LP64 | SV_SHP |
SV_TIMEKEEP | SV_RNG_SEED_VER,
.sv_set_syscall_retval = cpu_set_syscall_retval,
.sv_fetch_syscall_args = cpu_fetch_syscall_args,
.sv_syscallnames = syscallnames,
.sv_shared_page_base = SHAREDPAGE_LA48,
.sv_shared_page_len = PAGE_SIZE,
.sv_schedtail = NULL,
.sv_thread_detach = NULL,
.sv_trap = NULL,
.sv_stackgap = elf64_stackgap,
};
struct sysentvec elf64_freebsd_sysvec_la57 = {
.sv_size = SYS_MAXSYSCALL,
.sv_table = sysent,
.sv_transtrap = NULL,
.sv_fixup = __elfN(freebsd_fixup),
.sv_sendsig = sendsig,
.sv_sigcode = sigcode,
.sv_szsigcode = &szsigcode,
.sv_name = "FreeBSD ELF64",
.sv_coredump = __elfN(coredump),
+ .sv_elf_core_osabi = ELFOSABI_FREEBSD,
+ .sv_elf_core_abi_vendor = FREEBSD_ABI_VENDOR,
+ .sv_elf_core_prepare_notes = __elfN(prepare_notes),
.sv_imgact_try = NULL,
.sv_minsigstksz = MINSIGSTKSZ,
.sv_minuser = VM_MIN_ADDRESS,
.sv_maxuser = VM_MAXUSER_ADDRESS_LA57,
.sv_usrstack = USRSTACK_LA57,
.sv_psstrings = PS_STRINGS_LA57,
.sv_stackprot = VM_PROT_ALL,
.sv_copyout_auxargs = __elfN(freebsd_copyout_auxargs),
.sv_copyout_strings = exec_copyout_strings,
.sv_setregs = exec_setregs,
.sv_fixlimit = NULL,
.sv_maxssiz = NULL,
.sv_flags = SV_ABI_FREEBSD | SV_ASLR | SV_LP64 | SV_SHP |
SV_TIMEKEEP | SV_RNG_SEED_VER,
.sv_set_syscall_retval = cpu_set_syscall_retval,
.sv_fetch_syscall_args = cpu_fetch_syscall_args,
.sv_syscallnames = syscallnames,
.sv_shared_page_base = SHAREDPAGE_LA57,
.sv_shared_page_len = PAGE_SIZE,
.sv_schedtail = NULL,
.sv_thread_detach = NULL,
.sv_trap = NULL,
.sv_stackgap = elf64_stackgap,
};
static void
amd64_init_sysvecs(void *arg)
{
amd64_lower_shared_page(&elf64_freebsd_sysvec_la48);
if (la57) {
exec_sysvec_init(&elf64_freebsd_sysvec_la57);
exec_sysvec_init_secondary(&elf64_freebsd_sysvec_la57,
&elf64_freebsd_sysvec_la48);
} else {
exec_sysvec_init(&elf64_freebsd_sysvec_la48);
}
}
SYSINIT(elf64_sysvec, SI_SUB_EXEC, SI_ORDER_ANY, amd64_init_sysvecs, NULL);
void
amd64_lower_shared_page(struct sysentvec *sv)
{
if (hw_lower_amd64_sharedpage != 0) {
sv->sv_maxuser -= PAGE_SIZE;
sv->sv_shared_page_base -= PAGE_SIZE;
sv->sv_usrstack -= PAGE_SIZE;
sv->sv_psstrings -= PAGE_SIZE;
}
}
static boolean_t
freebsd_brand_info_la57_img_compat(struct image_params *imgp,
int32_t *osrel __unused, uint32_t *fctl0)
{
if ((imgp->proc->p_md.md_flags & P_MD_LA57) != 0)
return (TRUE);
if (fctl0 == NULL || (*fctl0 & NT_FREEBSD_FCTL_LA48) != 0)
return (FALSE);
if ((imgp->proc->p_md.md_flags & P_MD_LA48) != 0)
return (FALSE);
return (TRUE);
}
static Elf64_Brandinfo freebsd_brand_info_la48 = {
.brand = ELFOSABI_FREEBSD,
.machine = EM_X86_64,
.compat_3_brand = "FreeBSD",
.emul_path = NULL,
.interp_path = "/libexec/ld-elf.so.1",
.sysvec = &elf64_freebsd_sysvec_la48,
.interp_newpath = NULL,
.brand_note = &elf64_freebsd_brandnote,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE,
};
static Elf64_Brandinfo freebsd_brand_info_la57 = {
.brand = ELFOSABI_FREEBSD,
.machine = EM_X86_64,
.compat_3_brand = "FreeBSD",
.emul_path = NULL,
.interp_path = "/libexec/ld-elf.so.1",
.sysvec = &elf64_freebsd_sysvec_la57,
.interp_newpath = NULL,
.brand_note = &elf64_freebsd_brandnote,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE,
.header_supported = freebsd_brand_info_la57_img_compat,
};
static void
sysinit_register_elf64_brand_entries(void *arg __unused)
{
/*
* _57 must go first so it can either claim the image or hand
* it to _48.
*/
if (la57)
elf64_insert_brand_entry(&freebsd_brand_info_la57);
elf64_insert_brand_entry(&freebsd_brand_info_la48);
}
SYSINIT(elf64, SI_SUB_EXEC, SI_ORDER_FIRST,
sysinit_register_elf64_brand_entries, NULL);
static Elf64_Brandinfo freebsd_brand_oinfo = {
.brand = ELFOSABI_FREEBSD,
.machine = EM_X86_64,
.compat_3_brand = "FreeBSD",
.emul_path = NULL,
.interp_path = "/usr/libexec/ld-elf.so.1",
.sysvec = &elf64_freebsd_sysvec_la48,
.interp_newpath = NULL,
.brand_note = &elf64_freebsd_brandnote,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
};
SYSINIT(oelf64, SI_SUB_EXEC, SI_ORDER_ANY,
(sysinit_cfunc_t)elf64_insert_brand_entry, &freebsd_brand_oinfo);
static Elf64_Brandinfo kfreebsd_brand_info = {
.brand = ELFOSABI_FREEBSD,
.machine = EM_X86_64,
.compat_3_brand = "FreeBSD",
.emul_path = NULL,
.interp_path = "/lib/ld-kfreebsd-x86-64.so.1",
.sysvec = &elf64_freebsd_sysvec_la48,
.interp_newpath = NULL,
.brand_note = &elf64_kfreebsd_brandnote,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE_MANDATORY
};
SYSINIT(kelf64, SI_SUB_EXEC, SI_ORDER_ANY,
(sysinit_cfunc_t)elf64_insert_brand_entry, &kfreebsd_brand_info);
void
elf64_dump_thread(struct thread *td, void *dst, size_t *off)
{
void *buf;
size_t len;
len = 0;
if (use_xsave) {
if (dst != NULL) {
fpugetregs(td);
len += elf64_populate_note(NT_X86_XSTATE,
get_pcb_user_save_td(td), dst,
cpu_max_ext_state_size, &buf);
*(uint64_t *)((char *)buf + X86_XSTATE_XCR0_OFFSET) =
xsave_mask;
} else
len += elf64_populate_note(NT_X86_XSTATE, NULL, NULL,
cpu_max_ext_state_size, NULL);
}
*off = len;
}
bool
elf_is_ifunc_reloc(Elf_Size r_info)
{
return (ELF_R_TYPE(r_info) == R_X86_64_IRELATIVE);
}
/* Process one elf relocation with addend. */
static int
elf_reloc_internal(linker_file_t lf, Elf_Addr relocbase, const void *data,
int type, bool late_ifunc, elf_lookup_fn lookup)
{
Elf64_Addr *where, val;
Elf32_Addr *where32, val32;
Elf_Addr addr;
Elf_Addr addend;
Elf_Size rtype, symidx;
const Elf_Rel *rel;
const Elf_Rela *rela;
int error;
switch (type) {
case ELF_RELOC_REL:
rel = (const Elf_Rel *)data;
where = (Elf_Addr *) (relocbase + rel->r_offset);
rtype = ELF_R_TYPE(rel->r_info);
symidx = ELF_R_SYM(rel->r_info);
/* Addend is 32 bit on 32 bit relocs */
switch (rtype) {
case R_X86_64_PC32:
case R_X86_64_32S:
case R_X86_64_PLT32:
addend = *(Elf32_Addr *)where;
break;
default:
addend = *where;
break;
}
break;
case ELF_RELOC_RELA:
rela = (const Elf_Rela *)data;
where = (Elf_Addr *) (relocbase + rela->r_offset);
addend = rela->r_addend;
rtype = ELF_R_TYPE(rela->r_info);
symidx = ELF_R_SYM(rela->r_info);
break;
default:
panic("unknown reloc type %d\n", type);
}
if (late_ifunc) {
KASSERT(type == ELF_RELOC_RELA,
("Only RELA ifunc relocations are supported"));
if (rtype != R_X86_64_IRELATIVE)
return (0);
}
switch (rtype) {
case R_X86_64_NONE: /* none */
break;
case R_X86_64_64: /* S + A */
error = lookup(lf, symidx, 1, &addr);
val = addr + addend;
if (error != 0)
return (-1);
if (*where != val)
*where = val;
break;
case R_X86_64_PC32: /* S + A - P */
case R_X86_64_PLT32: /* L + A - P, L is PLT location for
the symbol, which we treat as S */
error = lookup(lf, symidx, 1, &addr);
where32 = (Elf32_Addr *)where;
val32 = (Elf32_Addr)(addr + addend - (Elf_Addr)where);
if (error != 0)
return (-1);
if (*where32 != val32)
*where32 = val32;
break;
case R_X86_64_32S: /* S + A sign extend */
error = lookup(lf, symidx, 1, &addr);
val32 = (Elf32_Addr)(addr + addend);
where32 = (Elf32_Addr *)where;
if (error != 0)
return (-1);
if (*where32 != val32)
*where32 = val32;
break;
case R_X86_64_COPY: /* none */
/*
* There shouldn't be copy relocations in kernel
* objects.
*/
printf("kldload: unexpected R_COPY relocation, "
"symbol index %ld\n", symidx);
return (-1);
case R_X86_64_GLOB_DAT: /* S */
case R_X86_64_JMP_SLOT: /* XXX need addend + offset */
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
if (*where != addr)
*where = addr;
break;
case R_X86_64_RELATIVE: /* B + A */
addr = elf_relocaddr(lf, relocbase + addend);
val = addr;
if (*where != val)
*where = val;
break;
case R_X86_64_IRELATIVE:
addr = relocbase + addend;
val = ((Elf64_Addr (*)(void))addr)();
if (*where != val)
*where = val;
break;
default:
printf("kldload: unexpected relocation type %ld, "
"symbol index %ld\n", rtype, symidx);
return (-1);
}
return (0);
}
int
elf_reloc(linker_file_t lf, Elf_Addr relocbase, const void *data, int type,
elf_lookup_fn lookup)
{
return (elf_reloc_internal(lf, relocbase, data, type, false, lookup));
}
int
elf_reloc_local(linker_file_t lf, Elf_Addr relocbase, const void *data,
int type, elf_lookup_fn lookup)
{
return (elf_reloc_internal(lf, relocbase, data, type, false, lookup));
}
int
elf_reloc_late(linker_file_t lf, Elf_Addr relocbase, const void *data,
int type, elf_lookup_fn lookup)
{
return (elf_reloc_internal(lf, relocbase, data, type, true, lookup));
}
int
elf_cpu_load_file(linker_file_t lf __unused)
{
return (0);
}
int
elf_cpu_unload_file(linker_file_t lf __unused)
{
return (0);
}
int
elf_cpu_parse_dynamic(caddr_t loadbase __unused, Elf_Dyn *dynamic __unused)
{
return (0);
}
diff --git a/sys/amd64/cloudabi32/cloudabi32_sysvec.c b/sys/amd64/cloudabi32/cloudabi32_sysvec.c
index 5f04c35a5c72..164f87e90e91 100644
--- a/sys/amd64/cloudabi32/cloudabi32_sysvec.c
+++ b/sys/amd64/cloudabi32/cloudabi32_sysvec.c
@@ -1,233 +1,236 @@
/*-
* Copyright (c) 2015-2016 Nuxi, https://nuxi.nl/
*
* 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 <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/imgact.h>
#include <sys/kernel.h>
#include <sys/proc.h>
#include <sys/sysent.h>
#include <vm/vm.h>
#include <vm/pmap.h>
#include <machine/frame.h>
#include <machine/pcb.h>
#include <machine/vmparam.h>
#include <compat/cloudabi/cloudabi_util.h>
#include <compat/cloudabi32/cloudabi32_syscall.h>
#include <compat/cloudabi32/cloudabi32_util.h>
#include <compat/ia32/ia32_signal.h>
#include <compat/ia32/ia32_util.h>
extern const char *cloudabi32_syscallnames[];
extern struct sysent cloudabi32_sysent[];
extern unsigned long ia32_maxssiz;
static int
cloudabi32_fixup_tcb(uintptr_t *stack_base, struct image_params *imgp)
{
int error;
uint32_t args[2];
/* Place auxiliary vector and TCB on the stack. */
error = cloudabi32_fixup(stack_base, imgp);
if (error != 0)
return (error);
/*
* On i386, the TCB is referred to by %gs:0. Reuse the empty
* space normally used by the return address (args[0]) to store
* a single element array, containing a pointer to the TCB. %gs
* base will point to this.
*
* Also let the first argument of the entry point (args[1])
* refer to the auxiliary vector, which is stored right after
* the TCB.
*/
args[0] = *stack_base;
args[1] = *stack_base +
roundup(sizeof(cloudabi32_tcb_t), sizeof(register_t));
*stack_base -= roundup2(sizeof(args), sizeof(register_t));
return (copyout(args, (void *)*stack_base, sizeof(args)));
}
static void
cloudabi32_proc_setregs(struct thread *td, struct image_params *imgp,
uintptr_t stack)
{
ia32_setregs(td, imgp, stack);
(void)cpu_set_user_tls(td, TO_PTR(stack));
}
static int
cloudabi32_fetch_syscall_args(struct thread *td)
{
struct trapframe *frame;
struct syscall_args *sa;
int error;
frame = td->td_frame;
sa = &td->td_sa;
/* Obtain system call number. */
sa->code = frame->tf_rax;
if (sa->code >= CLOUDABI32_SYS_MAXSYSCALL)
return (ENOSYS);
sa->callp = &cloudabi32_sysent[sa->code];
/*
* Fetch system call arguments.
*
* The vDSO has already made sure that the arguments are
* eight-byte aligned. Pointers and size_t parameters are
* zero-extended. This makes it possible to copy in the
* arguments directly. As long as the call doesn't use 32-bit
* data structures, we can just invoke the same system call
* implementation used by 64-bit processes.
*/
error = copyin((void *)frame->tf_rcx, sa->args,
sa->callp->sy_narg * sizeof(sa->args[0]));
if (error != 0)
return (error);
/* Default system call return values. */
td->td_retval[0] = 0;
td->td_retval[1] = 0;
return (0);
}
static void
cloudabi32_set_syscall_retval(struct thread *td, int error)
{
struct trapframe *frame = td->td_frame;
switch (error) {
case 0:
/*
* System call succeeded.
*
* Simply copy out the 64-bit return values into the
* same buffer provided for system call arguments. The
* vDSO will copy them to the right spot, truncating
* pointers and size_t values to 32 bits.
*/
frame->tf_rax = copyout(td->td_retval, (void *)frame->tf_rcx,
sizeof(td->td_retval)) == 0 ? 0 : CLOUDABI_EFAULT;
break;
case ERESTART:
/* Restart system call. */
frame->tf_rip -= frame->tf_err;
frame->tf_r10 = frame->tf_rcx;
set_pcb_flags(td->td_pcb, PCB_FULL_IRET);
break;
case EJUSTRETURN:
break;
default:
/* System call returned an error. */
frame->tf_rax = cloudabi_convert_errno(error);
break;
}
}
static void
cloudabi32_schedtail(struct thread *td)
{
struct trapframe *frame = td->td_frame;
register_t retval[2];
/* Return values for processes returning from fork. */
if ((td->td_pflags & TDP_FORKING) != 0) {
retval[0] = CLOUDABI_PROCESS_CHILD;
retval[1] = td->td_tid;
copyout(retval, (void *)frame->tf_rcx, sizeof(retval));
}
}
int
cloudabi32_thread_setregs(struct thread *td,
const cloudabi32_threadattr_t *attr, uint32_t tcb)
{
stack_t stack;
uint32_t args[3];
void *frameptr;
int error;
/* Perform standard register initialization. */
stack.ss_sp = TO_PTR(attr->stack);
stack.ss_size = attr->stack_len - sizeof(args);
cpu_set_upcall(td, TO_PTR(attr->entry_point), NULL, &stack);
/*
* Copy the arguments for the thread entry point onto the stack
* (args[1] and args[2]). Similar to process startup, use the
* otherwise unused return address (args[0]) for TLS.
*/
args[0] = tcb;
args[1] = td->td_tid;
args[2] = attr->argument;
frameptr = (void *)td->td_frame->tf_rsp;
error = copyout(args, frameptr, sizeof(args));
if (error != 0)
return (error);
return (cpu_set_user_tls(td, frameptr));
}
static struct sysentvec cloudabi32_elf_sysvec = {
.sv_size = CLOUDABI32_SYS_MAXSYSCALL,
.sv_table = cloudabi32_sysent,
.sv_fixup = cloudabi32_fixup_tcb,
.sv_name = "CloudABI 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_minuser = FREEBSD32_MINUSER,
.sv_maxuser = FREEBSD32_MAXUSER,
.sv_stackprot = VM_PROT_READ | VM_PROT_WRITE,
.sv_copyout_strings = cloudabi32_copyout_strings,
.sv_setregs = cloudabi32_proc_setregs,
.sv_fixlimit = ia32_fixlimit,
.sv_maxssiz = &ia32_maxssiz,
.sv_flags = SV_ABI_CLOUDABI | SV_CAPSICUM | SV_IA32 | SV_ILP32,
.sv_set_syscall_retval = cloudabi32_set_syscall_retval,
.sv_fetch_syscall_args = cloudabi32_fetch_syscall_args,
.sv_syscallnames = cloudabi32_syscallnames,
.sv_schedtail = cloudabi32_schedtail,
};
INIT_SYSENTVEC(elf_sysvec, &cloudabi32_elf_sysvec);
Elf32_Brandinfo cloudabi32_brand = {
.brand = ELFOSABI_CLOUDABI,
.machine = EM_386,
.sysvec = &cloudabi32_elf_sysvec,
.flags = BI_BRAND_ONLY_STATIC,
};
diff --git a/sys/amd64/cloudabi64/cloudabi64_sysvec.c b/sys/amd64/cloudabi64/cloudabi64_sysvec.c
index 34ad0aedfe8b..d3893902b08e 100644
--- a/sys/amd64/cloudabi64/cloudabi64_sysvec.c
+++ b/sys/amd64/cloudabi64/cloudabi64_sysvec.c
@@ -1,220 +1,223 @@
/*-
* Copyright (c) 2015 Nuxi, https://nuxi.nl/
*
* 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 <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/imgact.h>
#include <sys/kernel.h>
#include <sys/proc.h>
#include <sys/sysent.h>
#include <vm/vm.h>
#include <vm/pmap.h>
#include <machine/frame.h>
#include <machine/pcb.h>
#include <machine/vmparam.h>
#include <compat/cloudabi/cloudabi_util.h>
#include <compat/cloudabi64/cloudabi64_syscall.h>
#include <compat/cloudabi64/cloudabi64_util.h>
extern const char *cloudabi64_syscallnames[];
extern struct sysent cloudabi64_sysent[];
static int
cloudabi64_fixup_tcb(uintptr_t *stack_base, struct image_params *imgp)
{
int error;
register_t tcbptr;
/* Place auxiliary vector and TCB on the stack. */
error = cloudabi64_fixup(stack_base, imgp);
if (error != 0)
return (error);
/*
* On x86-64, the TCB is referred to by %fs:0. Take some space
* from the top of the stack to store a single element array,
* containing a pointer to the TCB. %fs base will point to this.
*/
tcbptr = (register_t)*stack_base;
*stack_base -= sizeof(tcbptr);
return (copyout(&tcbptr, (void *)*stack_base, sizeof(tcbptr)));
}
static void
cloudabi64_proc_setregs(struct thread *td, struct image_params *imgp,
uintptr_t stack)
{
struct trapframe *regs;
exec_setregs(td, imgp, stack);
/*
* The stack now contains a pointer to the TCB, the TCB itself,
* and the auxiliary vector. Let %rdx point to the auxiliary
* vector, and set %fs base to the address of the TCB.
*/
regs = td->td_frame;
regs->tf_rdi = stack + sizeof(register_t) +
roundup(sizeof(cloudabi64_tcb_t), sizeof(register_t));
(void)cpu_set_user_tls(td, TO_PTR(stack));
}
static int
cloudabi64_fetch_syscall_args(struct thread *td)
{
struct trapframe *frame;
struct syscall_args *sa;
frame = td->td_frame;
sa = &td->td_sa;
/* Obtain system call number. */
sa->code = frame->tf_rax;
if (sa->code >= CLOUDABI64_SYS_MAXSYSCALL)
return (ENOSYS);
sa->callp = &cloudabi64_sysent[sa->code];
/* Fetch system call arguments. */
sa->args[0] = frame->tf_rdi;
sa->args[1] = frame->tf_rsi;
sa->args[2] = frame->tf_rdx;
sa->args[3] = frame->tf_rcx; /* Actually %r10. */
sa->args[4] = frame->tf_r8;
sa->args[5] = frame->tf_r9;
/* Default system call return values. */
td->td_retval[0] = 0;
td->td_retval[1] = frame->tf_rdx;
return (0);
}
static void
cloudabi64_set_syscall_retval(struct thread *td, int error)
{
struct trapframe *frame = td->td_frame;
switch (error) {
case 0:
/* System call succeeded. */
frame->tf_rax = td->td_retval[0];
frame->tf_rdx = td->td_retval[1];
frame->tf_rflags &= ~PSL_C;
break;
case ERESTART:
/* Restart system call. */
frame->tf_rip -= frame->tf_err;
frame->tf_r10 = frame->tf_rcx;
set_pcb_flags(td->td_pcb, PCB_FULL_IRET);
break;
case EJUSTRETURN:
break;
default:
/* System call returned an error. */
frame->tf_rax = cloudabi_convert_errno(error);
frame->tf_rflags |= PSL_C;
break;
}
}
static void
cloudabi64_schedtail(struct thread *td)
{
struct trapframe *frame = td->td_frame;
/* Initial register values for processes returning from fork. */
frame->tf_rax = CLOUDABI_PROCESS_CHILD;
frame->tf_rdx = td->td_tid;
}
int
cloudabi64_thread_setregs(struct thread *td,
const cloudabi64_threadattr_t *attr, uint64_t tcb)
{
struct trapframe *frame;
stack_t stack;
uint64_t tcbptr;
int error;
/*
* On x86-64, the TCB is referred to by %fs:0. Take some space
* from the top of the stack to store a single element array,
* containing a pointer to the TCB. %fs base will point to this.
*/
tcbptr = rounddown(attr->stack + attr->stack_len - sizeof(tcbptr),
_Alignof(tcbptr));
error = copyout(&tcb, (void *)tcbptr, sizeof(tcb));
if (error != 0)
return (error);
/* Perform standard register initialization. */
stack.ss_sp = TO_PTR(attr->stack);
stack.ss_size = tcbptr - attr->stack;
cpu_set_upcall(td, TO_PTR(attr->entry_point), NULL, &stack);
/*
* Pass in the thread ID of the new thread and the argument
* pointer provided by the parent thread in as arguments to the
* entry point.
*/
frame = td->td_frame;
frame->tf_rdi = td->td_tid;
frame->tf_rsi = attr->argument;
return (cpu_set_user_tls(td, TO_PTR(tcbptr)));
}
static struct sysentvec cloudabi64_elf_sysvec = {
.sv_size = CLOUDABI64_SYS_MAXSYSCALL,
.sv_table = cloudabi64_sysent,
.sv_fixup = cloudabi64_fixup_tcb,
.sv_name = "CloudABI ELF64",
.sv_coredump = elf64_coredump,
+ .sv_elf_core_osabi = ELFOSABI_FREEBSD,
+ .sv_elf_core_abi_vendor = FREEBSD_ABI_VENDOR,
+ .sv_elf_core_prepare_notes = elf64_prepare_notes,
.sv_minuser = VM_MIN_ADDRESS,
/* Keep top page reserved to work around AMD Ryzen stability issues. */
.sv_maxuser = VM_MAXUSER_ADDRESS - PAGE_SIZE,
.sv_stackprot = VM_PROT_READ | VM_PROT_WRITE,
.sv_copyout_strings = cloudabi64_copyout_strings,
.sv_setregs = cloudabi64_proc_setregs,
.sv_flags = SV_ABI_CLOUDABI | SV_CAPSICUM | SV_LP64,
.sv_set_syscall_retval = cloudabi64_set_syscall_retval,
.sv_fetch_syscall_args = cloudabi64_fetch_syscall_args,
.sv_syscallnames = cloudabi64_syscallnames,
.sv_schedtail = cloudabi64_schedtail,
};
INIT_SYSENTVEC(elf_sysvec, &cloudabi64_elf_sysvec);
Elf64_Brandinfo cloudabi64_brand = {
.brand = ELFOSABI_CLOUDABI,
.machine = EM_X86_64,
.sysvec = &cloudabi64_elf_sysvec,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_ONLY_STATIC,
};
diff --git a/sys/amd64/linux/linux_sysvec.c b/sys/amd64/linux/linux_sysvec.c
index 32db8399ad42..655214b10047 100644
--- a/sys/amd64/linux/linux_sysvec.c
+++ b/sys/amd64/linux/linux_sysvec.c
@@ -1,942 +1,945 @@
/*-
* Copyright (c) 2013 Dmitry Chagin
* 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 <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#define __ELF_WORD_SIZE 64
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/exec.h>
#include <sys/fcntl.h>
#include <sys/imgact.h>
#include <sys/imgact_elf.h>
#include <sys/kernel.h>
#include <sys/ktr.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/module.h>
#include <sys/mutex.h>
#include <sys/proc.h>
#include <sys/resourcevar.h>
#include <sys/signalvar.h>
#include <sys/syscallsubr.h>
#include <sys/sysctl.h>
#include <sys/sysent.h>
#include <sys/sysproto.h>
#include <sys/vnode.h>
#include <sys/eventhandler.h>
#include <vm/vm.h>
#include <vm/pmap.h>
#include <vm/vm_extern.h>
#include <vm/vm_map.h>
#include <vm/vm_object.h>
#include <vm/vm_page.h>
#include <vm/vm_param.h>
#include <machine/cpu.h>
#include <machine/md_var.h>
#include <machine/pcb.h>
#include <machine/specialreg.h>
#include <machine/trap.h>
#include <amd64/linux/linux.h>
#include <amd64/linux/linux_proto.h>
#include <compat/linux/linux_emul.h>
#include <compat/linux/linux_ioctl.h>
#include <compat/linux/linux_mib.h>
#include <compat/linux/linux_misc.h>
#include <compat/linux/linux_signal.h>
#include <compat/linux/linux_sysproto.h>
#include <compat/linux/linux_util.h>
#include <compat/linux/linux_vdso.h>
MODULE_VERSION(linux64, 1);
static int linux_szsigcode;
static vm_object_t linux_shared_page_obj;
static char *linux_shared_page_mapping;
extern char _binary_linux_locore_o_start;
extern char _binary_linux_locore_o_end;
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(void *param);
static void linux_vdso_deinstall(void *param);
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 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)<nitems(_bsd_to_linux_trapcode)? \
_bsd_to_linux_trapcode[(code)]: \
LINUX_T_UNKNOWN)
LINUX_VDSO_SYM_INTPTR(linux_rt_sigcode);
LINUX_VDSO_SYM_CHAR(linux_platform);
/*
* If FreeBSD & Linux have a difference of opinion about what a trap
* means, deal with it here.
*
* MPSAFE
*/
static int
linux_translate_traps(int signal, int trap_code)
{
if (signal != SIGBUS)
return (signal);
switch (trap_code) {
case T_PROTFLT:
case T_TSSFLT:
case T_DOUBLEFLT:
case T_PAGEFLT:
return (SIGSEGV);
default:
return (signal);
}
}
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->args[0] = frame->tf_rdi;
sa->args[1] = frame->tf_rsi;
sa->args[2] = frame->tf_rdx;
sa->args[3] = frame->tf_rcx;
sa->args[4] = frame->tf_r8;
sa->args[5] = frame->tf_r9;
sa->code = frame->tf_rax;
if (sa->code >= p->p_sysent->sv_size)
/* nosys */
sa->callp = &p->p_sysent->sv_table[p->p_sysent->sv_size - 1];
else
sa->callp = &p->p_sysent->sv_table[sa->code];
td->td_retval[0] = 0;
return (0);
}
static void
linux_set_syscall_retval(struct thread *td, int error)
{
struct trapframe *frame;
frame = td->td_frame;
switch (error) {
case 0:
frame->tf_rax = td->td_retval[0];
frame->tf_r10 = frame->tf_rcx;
break;
case ERESTART:
/*
* Reconstruct pc, we know that 'syscall' is 2 bytes,
* lcall $X,y is 7 bytes, int 0x80 is 2 bytes.
* We saved this in tf_err.
*
*/
frame->tf_rip -= frame->tf_err;
frame->tf_r10 = frame->tf_rcx;
break;
case EJUSTRETURN:
break;
default:
frame->tf_rax = bsd_to_linux_errno(error);
frame->tf_r10 = frame->tf_rcx;
break;
}
/*
* Differently from FreeBSD native ABI, on Linux only %rcx
* and %r11 values are not preserved across the syscall.
* Require full context restore to get all registers except
* those two restored at return to usermode.
*
* XXX: Would be great to be able to avoid PCB_FULL_IRET
* for the error == 0 case.
*/
set_pcb_flags(td->td_pcb, PCB_FULL_IRET);
}
static 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,
imgp->proc->p_sysent->sv_shared_page_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;
/* Calculate string base and vector table pointers. */
if (imgp->execpath != NULL && imgp->auxargs != NULL)
execpath_len = strlen(imgp->execpath) + 1;
else
execpath_len = 0;
p = imgp->proc;
arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings;
destp = (uintptr_t)arginfo;
if (execpath_len != 0) {
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.
*/
#define RFLAG_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0)
if (!RFLAG_SECURE(rflags & ~PSL_RF, regs->tf_rflags & ~PSL_RF)) {
printf("linux_rt_sigreturn: rflags = 0x%lx\n", 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.
*/
#define CS_SECURE(cs) (ISPL(cs) == SEL_UPL)
if (!CS_SECURE(context->sc_cs)) {
printf("linux_rt_sigreturn: cs = 0x%x\n", 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);
/* Translate the signal. */
sig = bsd_to_linux_signal(sig);
/* Save user context. */
bzero(&sf, sizeof(sf));
bsd_to_linux_sigset(mask, &sf.sf_sc.uc_sigmask);
bsd_to_linux_sigset(mask, &sf.sf_sc.uc_mcontext.sc_mask);
sf.sf_sc.uc_stack.ss_sp = PTROUT(td->td_sigstk.ss_sp);
sf.sf_sc.uc_stack.ss_size = td->td_sigstk.ss_size;
sf.sf_sc.uc_stack.ss_flags = (td->td_pflags & TDP_ALTSTACK)
? ((oonstack) ? LINUX_SS_ONSTACK : 0) : LINUX_SS_DISABLE;
sf.sf_sc.uc_mcontext.sc_rdi = regs->tf_rdi;
sf.sf_sc.uc_mcontext.sc_rsi = regs->tf_rsi;
sf.sf_sc.uc_mcontext.sc_rdx = regs->tf_rdx;
sf.sf_sc.uc_mcontext.sc_rbp = regs->tf_rbp;
sf.sf_sc.uc_mcontext.sc_rbx = regs->tf_rbx;
sf.sf_sc.uc_mcontext.sc_rcx = regs->tf_rcx;
sf.sf_sc.uc_mcontext.sc_rax = regs->tf_rax;
sf.sf_sc.uc_mcontext.sc_rip = regs->tf_rip;
sf.sf_sc.uc_mcontext.sc_rsp = regs->tf_rsp;
sf.sf_sc.uc_mcontext.sc_r8 = regs->tf_r8;
sf.sf_sc.uc_mcontext.sc_r9 = regs->tf_r9;
sf.sf_sc.uc_mcontext.sc_r10 = regs->tf_r10;
sf.sf_sc.uc_mcontext.sc_r11 = regs->tf_r11;
sf.sf_sc.uc_mcontext.sc_r12 = regs->tf_r12;
sf.sf_sc.uc_mcontext.sc_r13 = regs->tf_r13;
sf.sf_sc.uc_mcontext.sc_r14 = regs->tf_r14;
sf.sf_sc.uc_mcontext.sc_r15 = regs->tf_r15;
sf.sf_sc.uc_mcontext.sc_cs = regs->tf_cs;
sf.sf_sc.uc_mcontext.sc_rflags = regs->tf_rflags;
sf.sf_sc.uc_mcontext.sc_err = regs->tf_err;
sf.sf_sc.uc_mcontext.sc_trapno = bsd_to_linux_trapcode(code);
sf.sf_sc.uc_mcontext.sc_cr2 = (register_t)ksi->ksi_addr;
/* Allocate space for the signal handler context. */
if ((td->td_pflags & TDP_ALTSTACK) != 0 && !oonstack &&
SIGISMEMBER(psp->ps_sigonstack, sig)) {
sp = (caddr_t)td->td_sigstk.ss_sp + td->td_sigstk.ss_size -
sizeof(struct l_rt_sigframe);
} else
sp = (caddr_t)regs->tf_rsp - sizeof(struct l_rt_sigframe) - 128;
/* Align to 16 bytes. */
sfp = (struct l_rt_sigframe *)((unsigned long)sp & ~0xFul);
/* Build the argument list for the signal handler. */
regs->tf_rdi = sig; /* arg 1 in %rdi */
regs->tf_rax = 0;
regs->tf_rsi = (register_t)&sfp->sf_si; /* arg 2 in %rsi */
regs->tf_rdx = (register_t)&sfp->sf_sc; /* arg 3 in %rdx */
/* Fill in POSIX parts. */
siginfo_to_lsiginfo(&ksi->ksi_info, &sf.sf_si, sig);
sf.sf_handler = catcher;
mtx_unlock(&psp->ps_mtx);
PROC_UNLOCK(p);
/* Copy the sigframe out to the user's stack. */
if (copyout(&sf, sfp, sizeof(*sfp)) != 0) {
PROC_LOCK(p);
sigexit(td, SIGILL);
}
regs->tf_rsp = (long)sfp;
regs->tf_rip = linux_rt_sigcode;
regs->tf_rflags &= ~(PSL_T | PSL_D);
regs->tf_cs = _ucodesel;
set_pcb_flags(td->td_pcb, PCB_FULL_IRET);
PROC_LOCK(p);
mtx_lock(&psp->ps_mtx);
}
#define LINUX_VSYSCALL_START (-10UL << 20)
#define LINUX_VSYSCALL_SZ 1024
const unsigned long linux_vsyscall_vector[] = {
LINUX_SYS_gettimeofday,
LINUX_SYS_linux_time,
LINUX_SYS_linux_getcpu,
};
static int
linux_vsyscall(struct thread *td)
{
struct trapframe *frame;
uint64_t retqaddr;
int code, traced;
int error;
frame = td->td_frame;
/* Check %rip for vsyscall area. */
if (__predict_true(frame->tf_rip < LINUX_VSYSCALL_START))
return (EINVAL);
if ((frame->tf_rip & (LINUX_VSYSCALL_SZ - 1)) != 0)
return (EINVAL);
code = (frame->tf_rip - LINUX_VSYSCALL_START) / LINUX_VSYSCALL_SZ;
if (code >= nitems(linux_vsyscall_vector))
return (EINVAL);
/*
* vsyscall called as callq *(%rax), so we must
* use return address from %rsp and also fixup %rsp.
*/
error = copyin((void *)frame->tf_rsp, &retqaddr, sizeof(retqaddr));
if (error)
return (error);
frame->tf_rip = retqaddr;
frame->tf_rax = linux_vsyscall_vector[code];
frame->tf_rsp += 8;
traced = (frame->tf_flags & PSL_T);
amd64_syscall(td, traced);
return (0);
}
struct sysentvec elf_linux_sysvec = {
.sv_size = LINUX_SYS_MAXSYSCALL,
.sv_table = linux_sysent,
.sv_transtrap = linux_translate_traps,
.sv_fixup = linux_fixup_elf,
.sv_sendsig = linux_rt_sendsig,
.sv_sigcode = &_binary_linux_locore_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 = FREEBSD_ABI_VENDOR,
+ .sv_elf_core_prepare_notes = elf64_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 = USRSTACK_LA48,
.sv_psstrings = PS_STRINGS_LA48,
.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_set_syscall_retval = linux_set_syscall_retval,
.sv_fetch_syscall_args = linux_fetch_syscall_args,
.sv_syscallnames = NULL,
.sv_shared_page_base = 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,
.sv_onexit = linux_on_exit,
.sv_ontdexit = linux_thread_dtor,
.sv_setid_allowed = &linux_setid_allowed_query,
};
static void
linux_vdso_install(void *param)
{
amd64_lower_shared_page(&elf_linux_sysvec);
linux_szsigcode = (&_binary_linux_locore_o_end -
&_binary_linux_locore_o_start);
if (linux_szsigcode > elf_linux_sysvec.sv_shared_page_len)
panic("Linux invalid vdso size\n");
__elfN(linux_vdso_fixup)(&elf_linux_sysvec);
linux_shared_page_obj = __elfN(linux_shared_page_init)
(&linux_shared_page_mapping);
__elfN(linux_vdso_reloc)(&elf_linux_sysvec);
bcopy(elf_linux_sysvec.sv_sigcode, linux_shared_page_mapping,
linux_szsigcode);
elf_linux_sysvec.sv_shared_page_obj = linux_shared_page_obj;
}
SYSINIT(elf_linux_vdso_init, SI_SUB_EXEC, SI_ORDER_ANY,
linux_vdso_install, NULL);
static void
linux_vdso_deinstall(void *param)
{
__elfN(linux_shared_page_fini)(linux_shared_page_obj,
linux_shared_page_mapping);
}
SYSUNINIT(elf_linux_vdso_uninit, SI_SUB_EXEC, SI_ORDER_FIRST,
linux_vdso_deinstall, NULL);
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
};
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 ELF exec handler removed\n");
} else
printf("Could not deinstall ELF interpreter entry\n");
break;
default:
return (EOPNOTSUPP);
}
return (error);
}
static moduledata_t linux64_elf_mod = {
"linux64elf",
linux64_elf_modevent,
0
};
DECLARE_MODULE_TIED(linux64elf, linux64_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY);
MODULE_DEPEND(linux64elf, linux_common, 1, 1, 1);
FEATURE(linux64, "Linux 64bit support");
diff --git a/sys/amd64/linux32/linux32_sysvec.c b/sys/amd64/linux32/linux32_sysvec.c
index 8e3e728161ce..740e8a54a4da 100644
--- a/sys/amd64/linux32/linux32_sysvec.c
+++ b/sys/amd64/linux32/linux32_sysvec.c
@@ -1,1112 +1,1115 @@
/*-
* 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 <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#ifndef COMPAT_FREEBSD32
#error "Unable to compile Linux-emulator due to missing COMPAT_FREEBSD32 option!"
#endif
#define __ELF_WORD_SIZE 32
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/exec.h>
#include <sys/fcntl.h>
#include <sys/imgact.h>
#include <sys/imgact_elf.h>
#include <sys/kernel.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/module.h>
#include <sys/mutex.h>
#include <sys/proc.h>
#include <sys/resourcevar.h>
#include <sys/signalvar.h>
#include <sys/syscallsubr.h>
#include <sys/sysctl.h>
#include <sys/sysent.h>
#include <sys/sysproto.h>
#include <sys/vnode.h>
#include <sys/eventhandler.h>
#include <vm/vm.h>
#include <vm/pmap.h>
#include <vm/vm_extern.h>
#include <vm/vm_map.h>
#include <vm/vm_object.h>
#include <vm/vm_page.h>
#include <vm/vm_param.h>
#include <machine/cpu.h>
#include <machine/md_var.h>
#include <machine/pcb.h>
#include <machine/specialreg.h>
#include <machine/trap.h>
#include <amd64/linux32/linux.h>
#include <amd64/linux32/linux32_proto.h>
#include <compat/linux/linux_emul.h>
#include <compat/linux/linux_ioctl.h>
#include <compat/linux/linux_mib.h>
#include <compat/linux/linux_misc.h>
#include <compat/linux/linux_signal.h>
#include <compat/linux/linux_util.h>
#include <compat/linux/linux_vdso.h>
MODULE_VERSION(linux, 1);
#define LINUX32_MAXUSER ((1ul << 32) - PAGE_SIZE)
#define LINUX32_SHAREDPAGE (LINUX32_MAXUSER - PAGE_SIZE)
#define LINUX32_USRSTACK LINUX32_SHAREDPAGE
static int linux_szsigcode;
static vm_object_t linux_shared_page_obj;
static char *linux_shared_page_mapping;
extern char _binary_linux32_locore_o_start;
extern char _binary_linux32_locore_o_end;
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 linux32_fixlimit(struct rlimit *rl, int which);
static bool linux32_trans_osrel(const Elf_Note *note, int32_t *osrel);
static void linux_vdso_install(void *param);
static void linux_vdso_deinstall(void *param);
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)<nitems(_bsd_to_linux_trapcode)? \
_bsd_to_linux_trapcode[(code)]: \
LINUX_T_UNKNOWN)
struct linux32_ps_strings {
u_int32_t ps_argvstr; /* first of 0 or more argument strings */
u_int ps_nargvstr; /* the number of argument strings */
u_int32_t ps_envstr; /* first of 0 or more environment strings */
u_int ps_nenvstr; /* the number of environment strings */
};
#define LINUX32_PS_STRINGS (LINUX32_USRSTACK - \
sizeof(struct linux32_ps_strings))
LINUX_VDSO_SYM_INTPTR(linux32_sigcode);
LINUX_VDSO_SYM_INTPTR(linux32_rt_sigcode);
LINUX_VDSO_SYM_INTPTR(linux32_vsyscall);
LINUX_VDSO_SYM_CHAR(linux_platform);
/*
* If FreeBSD & Linux have a difference of opinion about what a trap
* means, deal with it here.
*
* MPSAFE
*/
static int
linux_translate_traps(int signal, int trap_code)
{
if (signal != SIGBUS)
return (signal);
switch (trap_code) {
case T_PROTFLT:
case T_TSSFLT:
case T_DOUBLEFLT:
case T_PAGEFLT:
return (SIGSEGV);
default:
return (signal);
}
}
static int
linux_copyout_auxargs(struct image_params *imgp, uintptr_t base)
{
Elf32_Auxargs *args;
Elf32_Auxinfo *argarray, *pos;
int error, issetugid;
args = (Elf32_Auxargs *)imgp->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, linux32_vsyscall);
AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO_EHDR,
imgp->proc->p_sysent->sv_shared_page_base);
AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, cpu_feature);
AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz);
/*
* Do not export AT_CLKTCK when emulating Linux kernel prior to 2.4.0,
* as it has appeared in the 2.4.0-rc7 first time.
* Being exported, AT_CLKTCK is returned by sysconf(_SC_CLK_TCK),
* glibc falls back to the hard-coded CLK_TCK value when aux entry
* is not present.
* Also see linux_times() implementation.
*/
if (linux_kernver(curthread) >= LINUX_KERNVER_2004000)
AUXARGS_ENTRY(pos, LINUX_AT_CLKTCK, stclohz);
AUXARGS_ENTRY(pos, AT_PHDR, args->phdr);
AUXARGS_ENTRY(pos, AT_PHENT, args->phent);
AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum);
AUXARGS_ENTRY(pos, AT_BASE, args->base);
AUXARGS_ENTRY(pos, AT_FLAGS, args->flags);
AUXARGS_ENTRY(pos, AT_ENTRY, args->entry);
AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_ucred->cr_ruid);
AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid);
AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_ucred->cr_rgid);
AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid);
AUXARGS_ENTRY(pos, LINUX_AT_SECURE, issetugid);
AUXARGS_ENTRY(pos, LINUX_AT_RANDOM, PTROUT(imgp->canary));
AUXARGS_ENTRY(pos, LINUX_AT_HWCAP2, 0);
if (imgp->execpathp != 0)
AUXARGS_ENTRY(pos, LINUX_AT_EXECFN, PTROUT(imgp->execpathp));
if (args->execfd != -1)
AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd);
AUXARGS_ENTRY(pos, LINUX_AT_PLATFORM, PTROUT(linux_platform));
AUXARGS_ENTRY(pos, AT_NULL, 0);
free(imgp->auxargs, M_TEMP);
imgp->auxargs = NULL;
KASSERT(pos - argarray <= LINUX_AT_COUNT, ("Too many auxargs"));
error = copyout(argarray, (void *)base,
sizeof(*argarray) * LINUX_AT_COUNT);
free(argarray, M_TEMP);
return (error);
}
static int
linux_fixup_elf(uintptr_t *stack_base, struct image_params *imgp)
{
Elf32_Addr *base;
base = (Elf32_Addr *)*stack_base;
base--;
if (suword32(base, (uint32_t)imgp->args->argc) == -1)
return (EFAULT);
*stack_base = (uintptr_t)base;
return (0);
}
static void
linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
{
struct thread *td = curthread;
struct proc *p = td->td_proc;
struct sigacts *psp;
struct trapframe *regs;
struct l_rt_sigframe *fp, frame;
int oonstack;
int sig;
int code;
sig = ksi->ksi_signo;
code = ksi->ksi_code;
PROC_LOCK_ASSERT(p, MA_OWNED);
psp = p->p_sigacts;
mtx_assert(&psp->ps_mtx, MA_OWNED);
regs = td->td_frame;
oonstack = sigonstack(regs->tf_rsp);
/* Allocate space for the signal handler context. */
if ((td->td_pflags & TDP_ALTSTACK) && !oonstack &&
SIGISMEMBER(psp->ps_sigonstack, sig)) {
fp = (struct l_rt_sigframe *)((uintptr_t)td->td_sigstk.ss_sp +
td->td_sigstk.ss_size - sizeof(struct l_rt_sigframe));
} else
fp = (struct l_rt_sigframe *)regs->tf_rsp - 1;
mtx_unlock(&psp->ps_mtx);
/* Build the argument list for the signal handler. */
sig = bsd_to_linux_signal(sig);
bzero(&frame, sizeof(frame));
frame.sf_handler = PTROUT(catcher);
frame.sf_sig = sig;
frame.sf_siginfo = PTROUT(&fp->sf_si);
frame.sf_ucontext = PTROUT(&fp->sf_sc);
/* Fill in POSIX parts. */
siginfo_to_lsiginfo(&ksi->ksi_info, &frame.sf_si, sig);
/*
* Build the signal context to be used by sigreturn and libgcc unwind.
*/
frame.sf_sc.uc_flags = 0; /* XXX ??? */
frame.sf_sc.uc_link = 0; /* XXX ??? */
frame.sf_sc.uc_stack.ss_sp = PTROUT(td->td_sigstk.ss_sp);
frame.sf_sc.uc_stack.ss_size = td->td_sigstk.ss_size;
frame.sf_sc.uc_stack.ss_flags = (td->td_pflags & TDP_ALTSTACK)
? ((oonstack) ? LINUX_SS_ONSTACK : 0) : LINUX_SS_DISABLE;
PROC_UNLOCK(p);
bsd_to_linux_sigset(mask, &frame.sf_sc.uc_sigmask);
frame.sf_sc.uc_mcontext.sc_mask = frame.sf_sc.uc_sigmask.__mask;
frame.sf_sc.uc_mcontext.sc_edi = regs->tf_rdi;
frame.sf_sc.uc_mcontext.sc_esi = regs->tf_rsi;
frame.sf_sc.uc_mcontext.sc_ebp = regs->tf_rbp;
frame.sf_sc.uc_mcontext.sc_ebx = regs->tf_rbx;
frame.sf_sc.uc_mcontext.sc_esp = regs->tf_rsp;
frame.sf_sc.uc_mcontext.sc_edx = regs->tf_rdx;
frame.sf_sc.uc_mcontext.sc_ecx = regs->tf_rcx;
frame.sf_sc.uc_mcontext.sc_eax = regs->tf_rax;
frame.sf_sc.uc_mcontext.sc_eip = regs->tf_rip;
frame.sf_sc.uc_mcontext.sc_cs = regs->tf_cs;
frame.sf_sc.uc_mcontext.sc_gs = regs->tf_gs;
frame.sf_sc.uc_mcontext.sc_fs = regs->tf_fs;
frame.sf_sc.uc_mcontext.sc_es = regs->tf_es;
frame.sf_sc.uc_mcontext.sc_ds = regs->tf_ds;
frame.sf_sc.uc_mcontext.sc_eflags = regs->tf_rflags;
frame.sf_sc.uc_mcontext.sc_esp_at_signal = regs->tf_rsp;
frame.sf_sc.uc_mcontext.sc_ss = regs->tf_ss;
frame.sf_sc.uc_mcontext.sc_err = regs->tf_err;
frame.sf_sc.uc_mcontext.sc_cr2 = (u_int32_t)(uintptr_t)ksi->ksi_addr;
frame.sf_sc.uc_mcontext.sc_trapno = bsd_to_linux_trapcode(code);
if (copyout(&frame, fp, sizeof(frame)) != 0) {
/*
* Process has trashed its stack; give it an illegal
* instruction to halt it in its tracks.
*/
PROC_LOCK(p);
sigexit(td, SIGILL);
}
/* Build context to run handler in. */
regs->tf_rsp = PTROUT(fp);
regs->tf_rip = linux32_rt_sigcode;
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 = linux32_sigcode;
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. */
#define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0)
eflags = frame.sf_sc.sc_eflags;
if (!EFLAGS_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.
*/
#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_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. */
#define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0)
eflags = context->sc_eflags;
if (!EFLAGS_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.
*/
#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_rip;
trapsignal(td, &ksi);
return(EINVAL);
}
linux_to_bsd_sigset(&uc.uc_sigmask, &bmask);
kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0);
/*
* Restore signal context
*/
regs->tf_gs = context->sc_gs;
regs->tf_fs = context->sc_fs;
regs->tf_es = context->sc_es;
regs->tf_ds = context->sc_ds;
regs->tf_rdi = context->sc_edi;
regs->tf_rsi = context->sc_esi;
regs->tf_rbp = context->sc_ebp;
regs->tf_rbx = context->sc_ebx;
regs->tf_rdx = context->sc_edx;
regs->tf_rcx = context->sc_ecx;
regs->tf_rax = context->sc_eax;
regs->tf_rip = context->sc_eip;
regs->tf_cs = context->sc_cs;
regs->tf_rflags = eflags;
regs->tf_rsp = context->sc_esp_at_signal;
regs->tf_ss = context->sc_ss;
set_pcb_flags(td->td_pcb, PCB_FULL_IRET);
/*
* call sigaltstack & ignore results..
*/
lss = &uc.uc_stack;
ss.ss_sp = PTRIN(lss->ss_sp);
ss.ss_size = lss->ss_size;
ss.ss_flags = linux_to_bsd_sigaltstack(lss->ss_flags);
(void)kern_sigaltstack(td, &ss, NULL);
return (EJUSTRETURN);
}
static int
linux32_fetch_syscall_args(struct thread *td)
{
struct proc *p;
struct trapframe *frame;
struct syscall_args *sa;
p = td->td_proc;
frame = td->td_frame;
sa = &td->td_sa;
sa->args[0] = frame->tf_rbx;
sa->args[1] = frame->tf_rcx;
sa->args[2] = frame->tf_rdx;
sa->args[3] = frame->tf_rsi;
sa->args[4] = frame->tf_rdi;
sa->args[5] = frame->tf_rbp; /* Unconfirmed */
sa->code = frame->tf_rax;
if (sa->code >= p->p_sysent->sv_size)
/* nosys */
sa->callp = &p->p_sysent->sv_table[p->p_sysent->sv_size - 1];
else
sa->callp = &p->p_sysent->sv_table[sa->code];
td->td_retval[0] = 0;
td->td_retval[1] = frame->tf_rdx;
return (0);
}
static void
linux32_set_syscall_retval(struct thread *td, int error)
{
struct trapframe *frame = td->td_frame;
cpu_set_syscall_retval(td, error);
if (__predict_false(error != 0)) {
if (error != ERESTART && error != EJUSTRETURN)
frame->tf_rax = bsd_to_linux_errno(error);
}
}
/*
* 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;
/* Calculate string base and vector table pointers. */
if (imgp->execpath != NULL && imgp->auxargs != NULL)
execpath_len = strlen(imgp->execpath) + 1;
else
execpath_len = 0;
arginfo = (struct linux32_ps_strings *)LINUX32_PS_STRINGS;
destp = (uintptr_t)arginfo;
if (execpath_len != 0) {
destp -= execpath_len;
destp = rounddown2(destp, sizeof(uint32_t));
imgp->execpathp = (void *)destp;
error = copyout(imgp->execpath, imgp->execpathp, execpath_len);
if (error != 0)
return (error);
}
/* Prepare the canary for SSP. */
arc4rand(canary, sizeof(canary), 0);
destp -= roundup(sizeof(canary), sizeof(uint32_t));
imgp->canary = (void *)destp;
error = copyout(canary, imgp->canary, sizeof(canary));
if (error != 0)
return (error);
/* Allocate room for the argument and environment strings. */
destp -= ARG_MAX - imgp->args->stringspace;
destp = rounddown2(destp, sizeof(uint32_t));
ustringp = destp;
if (imgp->auxargs) {
/*
* Allocate room on the stack for the ELF auxargs
* array. It has LINUX_AT_COUNT entries.
*/
destp -= LINUX_AT_COUNT * sizeof(Elf32_Auxinfo);
destp = rounddown2(destp, sizeof(uint32_t));
}
vectp = (uint32_t *)destp;
/*
* Allocate room for the argv[] and env vectors including the
* terminating NULL pointers.
*/
vectp -= imgp->args->argc + 1 + imgp->args->envc + 1;
/* vectp also becomes our initial stack base. */
*stack_base = (uintptr_t)vectp;
stringp = imgp->args->begin_argv;
argc = imgp->args->argc;
envc = imgp->args->envc;
/* Copy out strings - arguments and environment. */
error = copyout(stringp, (void *)ustringp,
ARG_MAX - imgp->args->stringspace);
if (error != 0)
return (error);
/* Fill in "ps_strings" struct for ps, w, etc. */
if (suword32(&arginfo->ps_argvstr, (uint32_t)(intptr_t)vectp) != 0 ||
suword32(&arginfo->ps_nargvstr, argc) != 0)
return (EFAULT);
/* Fill in argument portion of vector table. */
for (; argc > 0; --argc) {
if (suword32(vectp++, ustringp) != 0)
return (EFAULT);
while (*stringp++ != 0)
ustringp++;
ustringp++;
}
/* A null vector table pointer separates the argp's from the envp's. */
if (suword32(vectp++, 0) != 0)
return (EFAULT);
if (suword32(&arginfo->ps_envstr, (uint32_t)(intptr_t)vectp) != 0 ||
suword32(&arginfo->ps_nenvstr, envc) != 0)
return (EFAULT);
/* Fill in environment portion of vector table. */
for (; envc > 0; --envc) {
if (suword32(vectp++, ustringp) != 0)
return (EFAULT);
while (*stringp++ != 0)
ustringp++;
ustringp++;
}
/* The end of the vector table is a null pointer. */
if (suword32(vectp, 0) != 0)
return (EFAULT);
if (imgp->auxargs) {
vectp++;
error = imgp->sysent->sv_copyout_auxargs(imgp,
(uintptr_t)vectp);
if (error != 0)
return (error);
}
return (0);
}
static SYSCTL_NODE(_compat, OID_AUTO, linux32, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
"32-bit Linux emulation");
static u_long linux32_maxdsiz = LINUX32_MAXDSIZ;
SYSCTL_ULONG(_compat_linux32, OID_AUTO, maxdsiz, CTLFLAG_RW,
&linux32_maxdsiz, 0, "");
static u_long linux32_maxssiz = LINUX32_MAXSSIZ;
SYSCTL_ULONG(_compat_linux32, OID_AUTO, maxssiz, CTLFLAG_RW,
&linux32_maxssiz, 0, "");
static u_long linux32_maxvmem = LINUX32_MAXVMEM;
SYSCTL_ULONG(_compat_linux32, OID_AUTO, maxvmem, CTLFLAG_RW,
&linux32_maxvmem, 0, "");
static void
linux32_fixlimit(struct rlimit *rl, int which)
{
switch (which) {
case RLIMIT_DATA:
if (linux32_maxdsiz != 0) {
if (rl->rlim_cur > linux32_maxdsiz)
rl->rlim_cur = linux32_maxdsiz;
if (rl->rlim_max > linux32_maxdsiz)
rl->rlim_max = linux32_maxdsiz;
}
break;
case RLIMIT_STACK:
if (linux32_maxssiz != 0) {
if (rl->rlim_cur > linux32_maxssiz)
rl->rlim_cur = linux32_maxssiz;
if (rl->rlim_max > linux32_maxssiz)
rl->rlim_max = linux32_maxssiz;
}
break;
case RLIMIT_VMEM:
if (linux32_maxvmem != 0) {
if (rl->rlim_cur > linux32_maxvmem)
rl->rlim_cur = linux32_maxvmem;
if (rl->rlim_max > linux32_maxvmem)
rl->rlim_max = linux32_maxvmem;
}
break;
}
}
struct sysentvec elf_linux_sysvec = {
.sv_size = LINUX32_SYS_MAXSYSCALL,
.sv_table = linux32_sysent,
.sv_transtrap = linux_translate_traps,
.sv_fixup = linux_fixup_elf,
.sv_sendsig = linux_sendsig,
.sv_sigcode = &_binary_linux32_locore_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 = LINUX32_MAXUSER,
.sv_usrstack = LINUX32_USRSTACK,
.sv_psstrings = 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_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,
.sv_onexit = linux_on_exit,
.sv_ontdexit = linux_thread_dtor,
.sv_setid_allowed = &linux_setid_allowed_query,
};
static void
linux_vdso_install(void *param)
{
linux_szsigcode = (&_binary_linux32_locore_o_end -
&_binary_linux32_locore_o_start);
if (linux_szsigcode > elf_linux_sysvec.sv_shared_page_len)
panic("Linux invalid vdso size\n");
__elfN(linux_vdso_fixup)(&elf_linux_sysvec);
linux_shared_page_obj = __elfN(linux_shared_page_init)
(&linux_shared_page_mapping);
__elfN(linux_vdso_reloc)(&elf_linux_sysvec);
bcopy(elf_linux_sysvec.sv_sigcode, linux_shared_page_mapping,
linux_szsigcode);
elf_linux_sysvec.sv_shared_page_obj = linux_shared_page_obj;
}
SYSINIT(elf_linux_vdso_init, SI_SUB_EXEC, SI_ORDER_ANY,
linux_vdso_install, NULL);
static void
linux_vdso_deinstall(void *param)
{
__elfN(linux_shared_page_fini)(linux_shared_page_obj,
linux_shared_page_mapping);
}
SYSUNINIT(elf_linux_vdso_uninit, SI_SUB_EXEC, SI_ORDER_FIRST,
linux_vdso_deinstall, NULL);
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
};
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 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)
linux32_ioctl_unregister_handler(*lihp);
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);
MODULE_DEPEND(linuxelf, linux_common, 1, 1, 1);
FEATURE(linux, "Linux 32bit support");
diff --git a/sys/arm/arm/elf_machdep.c b/sys/arm/arm/elf_machdep.c
index d32dbb5baf90..c67cd225a1af 100644
--- a/sys/arm/arm/elf_machdep.c
+++ b/sys/arm/arm/elf_machdep.c
@@ -1,336 +1,339 @@
/*-
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright 1996-1998 John D. Polstra.
* 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 ``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 <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/kernel.h>
#include <sys/systm.h>
#include <sys/exec.h>
#include <sys/imgact.h>
#include <sys/linker.h>
#include <sys/sysent.h>
#include <sys/imgact_elf.h>
#include <sys/proc.h>
#include <sys/syscall.h>
#include <sys/signalvar.h>
#include <sys/vnode.h>
#include <vm/vm.h>
#include <vm/pmap.h>
#include <vm/vm_param.h>
#include <machine/elf.h>
#include <machine/md_var.h>
#include <machine/stack.h>
#ifdef VFP
#include <machine/vfp.h>
#endif
#include "opt_ddb.h" /* for OPT_DDB */
#include "opt_global.h" /* for OPT_KDTRACE_HOOKS */
#include "opt_stack.h" /* for OPT_STACK */
static boolean_t elf32_arm_abi_supported(struct image_params *, int32_t *,
uint32_t *);
u_long elf_hwcap;
u_long elf_hwcap2;
struct sysentvec elf32_freebsd_sysvec = {
.sv_size = SYS_MAXSYSCALL,
.sv_table = sysent,
.sv_transtrap = NULL,
.sv_fixup = __elfN(freebsd_fixup),
.sv_sendsig = sendsig,
.sv_sigcode = sigcode,
.sv_szsigcode = &szsigcode,
.sv_name = "FreeBSD ELF32",
.sv_coredump = __elfN(coredump),
+ .sv_elf_core_osabi = ELFOSABI_FREEBSD,
+ .sv_elf_core_abi_vendor = FREEBSD_ABI_VENDOR,
+ .sv_elf_core_prepare_notes = __elfN(prepare_notes),
.sv_imgact_try = NULL,
.sv_minsigstksz = MINSIGSTKSZ,
.sv_minuser = VM_MIN_ADDRESS,
.sv_maxuser = VM_MAXUSER_ADDRESS,
.sv_usrstack = USRSTACK,
.sv_psstrings = PS_STRINGS,
.sv_stackprot = VM_PROT_ALL,
.sv_copyout_auxargs = __elfN(freebsd_copyout_auxargs),
.sv_copyout_strings = exec_copyout_strings,
.sv_setregs = exec_setregs,
.sv_fixlimit = NULL,
.sv_maxssiz = NULL,
.sv_flags =
SV_ASLR | SV_SHP | SV_TIMEKEEP | SV_RNG_SEED_VER |
SV_ABI_FREEBSD | SV_ILP32,
.sv_set_syscall_retval = cpu_set_syscall_retval,
.sv_fetch_syscall_args = cpu_fetch_syscall_args,
.sv_syscallnames = syscallnames,
.sv_shared_page_base = SHAREDPAGE,
.sv_shared_page_len = PAGE_SIZE,
.sv_schedtail = NULL,
.sv_thread_detach = NULL,
.sv_trap = NULL,
.sv_hwcap = &elf_hwcap,
.sv_hwcap2 = &elf_hwcap2,
};
INIT_SYSENTVEC(elf32_sysvec, &elf32_freebsd_sysvec);
static Elf32_Brandinfo freebsd_brand_info = {
.brand = ELFOSABI_FREEBSD,
.machine = EM_ARM,
.compat_3_brand = "FreeBSD",
.emul_path = NULL,
.interp_path = "/libexec/ld-elf.so.1",
.sysvec = &elf32_freebsd_sysvec,
.interp_newpath = NULL,
.brand_note = &elf32_freebsd_brandnote,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE,
.header_supported= elf32_arm_abi_supported,
};
SYSINIT(elf32, SI_SUB_EXEC, SI_ORDER_FIRST,
(sysinit_cfunc_t) elf32_insert_brand_entry,
&freebsd_brand_info);
static boolean_t
elf32_arm_abi_supported(struct image_params *imgp, int32_t *osrel __unused,
uint32_t *fctl0 __unused)
{
const Elf_Ehdr *hdr = (const Elf_Ehdr *)imgp->image_header;
/*
* When configured for EABI, FreeBSD supports EABI vesions 4 and 5.
*/
if (EF_ARM_EABI_VERSION(hdr->e_flags) < EF_ARM_EABI_FREEBSD_MIN) {
if (bootverbose)
uprintf("Attempting to execute non EABI binary (rev %d) image %s",
EF_ARM_EABI_VERSION(hdr->e_flags), imgp->args->fname);
return (FALSE);
}
return (TRUE);
}
void
elf32_dump_thread(struct thread *td, void *dst, size_t *off)
{
#ifdef VFP
mcontext_vfp_t vfp;
if (dst != NULL) {
get_vfpcontext(td, &vfp);
*off = elf32_populate_note(NT_ARM_VFP, &vfp, dst, sizeof(vfp),
NULL);
} else
*off = elf32_populate_note(NT_ARM_VFP, NULL, NULL, sizeof(vfp),
NULL);
#endif
}
bool
elf_is_ifunc_reloc(Elf_Size r_info __unused)
{
return (false);
}
/*
* It is possible for the compiler to emit relocations for unaligned data.
* We handle this situation with these inlines.
*/
#define RELOC_ALIGNED_P(x) \
(((uintptr_t)(x) & (sizeof(void *) - 1)) == 0)
static __inline Elf_Addr
load_ptr(Elf_Addr *where)
{
Elf_Addr res;
if (RELOC_ALIGNED_P(where))
return *where;
memcpy(&res, where, sizeof(res));
return (res);
}
static __inline void
store_ptr(Elf_Addr *where, Elf_Addr val)
{
if (RELOC_ALIGNED_P(where))
*where = val;
else
memcpy(where, &val, sizeof(val));
}
#undef RELOC_ALIGNED_P
/* Process one elf relocation with addend. */
static int
elf_reloc_internal(linker_file_t lf, Elf_Addr relocbase, const void *data,
int type, int local, elf_lookup_fn lookup)
{
Elf_Addr *where;
Elf_Addr addr;
Elf_Addr addend;
Elf_Word rtype, symidx;
const Elf_Rel *rel;
const Elf_Rela *rela;
int error;
switch (type) {
case ELF_RELOC_REL:
rel = (const Elf_Rel *)data;
where = (Elf_Addr *) (relocbase + rel->r_offset);
addend = load_ptr(where);
rtype = ELF_R_TYPE(rel->r_info);
symidx = ELF_R_SYM(rel->r_info);
break;
case ELF_RELOC_RELA:
rela = (const Elf_Rela *)data;
where = (Elf_Addr *) (relocbase + rela->r_offset);
addend = rela->r_addend;
rtype = ELF_R_TYPE(rela->r_info);
symidx = ELF_R_SYM(rela->r_info);
break;
default:
panic("unknown reloc type %d\n", type);
}
if (local) {
if (rtype == R_ARM_RELATIVE) { /* A + B */
addr = elf_relocaddr(lf, relocbase + addend);
if (load_ptr(where) != addr)
store_ptr(where, addr);
}
return (0);
}
switch (rtype) {
case R_ARM_NONE: /* none */
break;
case R_ARM_ABS32:
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
store_ptr(where, addr + load_ptr(where));
break;
case R_ARM_COPY: /* none */
/*
* There shouldn't be copy relocations in kernel
* objects.
*/
printf("kldload: unexpected R_COPY relocation, "
"symbol index %d\n", symidx);
return (-1);
break;
case R_ARM_JUMP_SLOT:
error = lookup(lf, symidx, 1, &addr);
if (error == 0) {
store_ptr(where, addr);
return (0);
}
return (-1);
case R_ARM_RELATIVE:
break;
default:
printf("kldload: unexpected relocation type %d, "
"symbol index %d\n", rtype, symidx);
return (-1);
}
return(0);
}
int
elf_reloc(linker_file_t lf, Elf_Addr relocbase, const void *data, int type,
elf_lookup_fn lookup)
{
return (elf_reloc_internal(lf, relocbase, data, type, 0, lookup));
}
int
elf_reloc_local(linker_file_t lf, Elf_Addr relocbase, const void *data,
int type, elf_lookup_fn lookup)
{
return (elf_reloc_internal(lf, relocbase, data, type, 1, lookup));
}
int
elf_cpu_load_file(linker_file_t lf)
{
/*
* The pmap code does not do an icache sync upon establishing executable
* mappings in the kernel pmap. It's an optimization based on the fact
* that kernel memory allocations always have EXECUTABLE protection even
* when the memory isn't going to hold executable code. The only time
* kernel memory holding instructions does need a sync is after loading
* a kernel module, and that's when this function gets called.
*
* This syncs data and instruction caches after loading a module. We
* don't worry about the kernel itself (lf->id is 1) as locore.S did
* that on entry. Even if data cache maintenance was done by IO code,
* the relocation fixup process creates dirty cache entries that we must
* write back before doing icache sync. The instruction cache sync also
* invalidates the branch predictor cache on platforms that have one.
*/
if (lf->id == 1)
return (0);
dcache_wb_pou((vm_offset_t)lf->address, (vm_size_t)lf->size);
icache_inv_all();
#if defined(DDB) || defined(KDTRACE_HOOKS) || defined(STACK)
/*
* Inform the stack(9) code of the new module, so it can acquire its
* per-module unwind data.
*/
unwind_module_loaded(lf);
#endif
return (0);
}
int
elf_cpu_parse_dynamic(caddr_t loadbase __unused, Elf_Dyn *dynamic __unused)
{
return (0);
}
int
elf_cpu_unload_file(linker_file_t lf)
{
#if defined(DDB) || defined(KDTRACE_HOOKS) || defined(STACK)
/* Inform the stack(9) code that this module is gone. */
unwind_module_unloaded(lf);
#endif
return (0);
}
diff --git a/sys/arm/cloudabi32/cloudabi32_sysvec.c b/sys/arm/cloudabi32/cloudabi32_sysvec.c
index c6dfe0661b7c..a8c5da47d265 100644
--- a/sys/arm/cloudabi32/cloudabi32_sysvec.c
+++ b/sys/arm/cloudabi32/cloudabi32_sysvec.c
@@ -1,195 +1,198 @@
/*-
* Copyright (c) 2015-2016 Nuxi, https://nuxi.nl/
*
* 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 <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/imgact.h>
#include <sys/kernel.h>
#include <sys/proc.h>
#include <sys/sysent.h>
#include <vm/vm.h>
#include <vm/pmap.h>
#include <machine/frame.h>
#include <machine/pcb.h>
#include <machine/vmparam.h>
#include <compat/cloudabi/cloudabi_util.h>
#include <compat/cloudabi32/cloudabi32_syscall.h>
#include <compat/cloudabi32/cloudabi32_util.h>
extern const char *cloudabi32_syscallnames[];
extern struct sysent cloudabi32_sysent[];
static void
cloudabi32_proc_setregs(struct thread *td, struct image_params *imgp,
uintptr_t stack)
{
struct trapframe *regs;
exec_setregs(td, imgp, stack);
/*
* The stack now contains a pointer to the TCB and the auxiliary
* vector. Let r0 point to the auxiliary vector, and set
* tpidrurw to the TCB.
*/
regs = td->td_frame;
regs->tf_r0 =
stack + roundup(sizeof(cloudabi32_tcb_t), sizeof(register_t));
(void)cpu_set_user_tls(td, TO_PTR(stack));
}
static int
cloudabi32_fetch_syscall_args(struct thread *td)
{
struct trapframe *frame;
struct syscall_args *sa;
int error;
frame = td->td_frame;
sa = &td->td_sa;
/* Obtain system call number. */
sa->code = frame->tf_r12;
if (sa->code >= CLOUDABI32_SYS_MAXSYSCALL)
return (ENOSYS);
sa->callp = &cloudabi32_sysent[sa->code];
/* Fetch system call arguments from registers and the stack. */
sa->args[0] = frame->tf_r0;
sa->args[1] = frame->tf_r1;
sa->args[2] = frame->tf_r2;
sa->args[3] = frame->tf_r3;
if (sa->callp->sy_narg > 4) {
error = copyin((void *)td->td_frame->tf_usr_sp, &sa->args[4],
(sa->callp->sy_narg - 4) * sizeof(register_t));
if (error != 0)
return (error);
}
/* Default system call return values. */
td->td_retval[0] = 0;
td->td_retval[1] = frame->tf_r1;
return (0);
}
static void
cloudabi32_set_syscall_retval(struct thread *td, int error)
{
struct trapframe *frame = td->td_frame;
switch (error) {
case 0:
/* System call succeeded. */
frame->tf_r0 = td->td_retval[0];
frame->tf_r1 = td->td_retval[1];
frame->tf_spsr &= ~PSR_C;
break;
case ERESTART:
/* Restart system call. */
frame->tf_pc -= 4;
break;
case EJUSTRETURN:
break;
default:
/* System call returned an error. */
frame->tf_r0 = cloudabi_convert_errno(error);
frame->tf_spsr |= PSR_C;
break;
}
}
static void
cloudabi32_schedtail(struct thread *td)
{
struct trapframe *frame = td->td_frame;
/*
* Initial register values for processes returning from fork.
* Make sure that we only set these values when forking, not
* when creating a new thread.
*/
if ((td->td_pflags & TDP_FORKING) != 0) {
frame->tf_r0 = CLOUDABI_PROCESS_CHILD;
frame->tf_r1 = td->td_tid;
}
}
int
cloudabi32_thread_setregs(struct thread *td,
const cloudabi32_threadattr_t *attr, uint32_t tcb)
{
struct trapframe *frame;
stack_t stack;
/* Perform standard register initialization. */
stack.ss_sp = TO_PTR(attr->stack);
stack.ss_size = attr->stack_len;
cpu_set_upcall(td, TO_PTR(attr->entry_point), NULL, &stack);
/*
* Pass in the thread ID of the new thread and the argument
* pointer provided by the parent thread in as arguments to the
* entry point.
*/
frame = td->td_frame;
frame->tf_r0 = td->td_tid;
frame->tf_r1 = attr->argument;
/* Set up TLS. */
return (cpu_set_user_tls(td, TO_PTR(tcb)));
}
static struct sysentvec cloudabi32_elf_sysvec = {
.sv_size = CLOUDABI32_SYS_MAXSYSCALL,
.sv_table = cloudabi32_sysent,
.sv_fixup = cloudabi32_fixup,
.sv_name = "CloudABI 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_minuser = VM_MIN_ADDRESS,
.sv_maxuser = VM_MAXUSER_ADDRESS,
.sv_stackprot = VM_PROT_READ | VM_PROT_WRITE,
.sv_copyout_strings = cloudabi32_copyout_strings,
.sv_setregs = cloudabi32_proc_setregs,
.sv_flags = SV_ABI_CLOUDABI | SV_CAPSICUM | SV_ILP32,
.sv_set_syscall_retval = cloudabi32_set_syscall_retval,
.sv_fetch_syscall_args = cloudabi32_fetch_syscall_args,
.sv_syscallnames = cloudabi32_syscallnames,
.sv_schedtail = cloudabi32_schedtail,
};
INIT_SYSENTVEC(elf_sysvec, &cloudabi32_elf_sysvec);
Elf32_Brandinfo cloudabi32_brand = {
.brand = ELFOSABI_CLOUDABI,
.machine = EM_ARM,
.sysvec = &cloudabi32_elf_sysvec,
.flags = BI_BRAND_ONLY_STATIC,
};
diff --git a/sys/arm64/arm64/elf32_machdep.c b/sys/arm64/arm64/elf32_machdep.c
index b9669616e1dd..7792374f865b 100644
--- a/sys/arm64/arm64/elf32_machdep.c
+++ b/sys/arm64/arm64/elf32_machdep.c
@@ -1,273 +1,276 @@
/*-
* Copyright (c) 2014, 2015 The FreeBSD Foundation.
* Copyright (c) 2014, 2017 Andrew Turner.
* Copyright (c) 2018 Olivier Houchard
* All rights reserved.
*
* This software was developed by Andrew Turner under
* sponsorship from the FreeBSD Foundation.
*
* Portions of this software were developed by Konstantin Belousov
* under sponsorship from the FreeBSD Foundation.
*
* 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 <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#define __ELF_WORD_SIZE 32
#include <sys/param.h>
#include <sys/kernel.h>
#include <sys/systm.h>
#include <sys/exec.h>
#include <sys/imgact.h>
#include <sys/linker.h>
#include <sys/proc.h>
#include <sys/sysent.h>
#include <sys/imgact_elf.h>
#include <sys/syscall.h>
#include <sys/signalvar.h>
#include <sys/vnode.h>
#include <machine/elf.h>
#ifdef VFP
#include <machine/vfp.h>
#endif
#include <compat/freebsd32/freebsd32_util.h>
#define FREEBSD32_MINUSER 0x00001000
#define FREEBSD32_MAXUSER ((1ul << 32) - PAGE_SIZE)
#define FREEBSD32_SHAREDPAGE (FREEBSD32_MAXUSER - PAGE_SIZE)
#define FREEBSD32_USRSTACK FREEBSD32_SHAREDPAGE
extern const char *freebsd32_syscallnames[];
extern char aarch32_sigcode[];
extern int sz_aarch32_sigcode;
static int freebsd32_fetch_syscall_args(struct thread *td);
static void freebsd32_setregs(struct thread *td, struct image_params *imgp,
u_long stack);
static void freebsd32_set_syscall_retval(struct thread *, int);
static boolean_t elf32_arm_abi_supported(struct image_params *, int32_t *,
uint32_t *);
extern void freebsd32_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask);
static struct sysentvec elf32_freebsd_sysvec = {
.sv_size = SYS_MAXSYSCALL,
.sv_table = freebsd32_sysent,
.sv_transtrap = NULL,
.sv_fixup = elf32_freebsd_fixup,
.sv_sendsig = freebsd32_sendsig,
.sv_sigcode = aarch32_sigcode,
.sv_szsigcode = &sz_aarch32_sigcode,
.sv_name = "FreeBSD 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 = NULL,
.sv_minsigstksz = MINSIGSTKSZ,
.sv_minuser = FREEBSD32_MINUSER,
.sv_maxuser = FREEBSD32_MAXUSER,
.sv_usrstack = FREEBSD32_USRSTACK,
.sv_psstrings = FREEBSD32_PS_STRINGS,
.sv_stackprot = VM_PROT_READ | VM_PROT_WRITE,
.sv_copyout_auxargs = elf32_freebsd_copyout_auxargs,
.sv_copyout_strings = freebsd32_copyout_strings,
.sv_setregs = freebsd32_setregs,
.sv_fixlimit = NULL, // XXX
.sv_maxssiz = NULL,
.sv_flags = SV_ABI_FREEBSD | SV_ILP32 | SV_SHP | SV_TIMEKEEP |
SV_RNG_SEED_VER,
.sv_set_syscall_retval = freebsd32_set_syscall_retval,
.sv_fetch_syscall_args = freebsd32_fetch_syscall_args,
.sv_syscallnames = freebsd32_syscallnames,
.sv_shared_page_base = FREEBSD32_SHAREDPAGE,
.sv_shared_page_len = PAGE_SIZE,
.sv_schedtail = NULL,
.sv_thread_detach = NULL,
.sv_trap = NULL,
};
INIT_SYSENTVEC(elf32_sysvec, &elf32_freebsd_sysvec);
static Elf32_Brandinfo freebsd32_brand_info = {
.brand = ELFOSABI_FREEBSD,
.machine = EM_ARM,
.compat_3_brand = "FreeBSD",
.emul_path = NULL,
.interp_path = "/libexec/ld-elf.so.1",
.sysvec = &elf32_freebsd_sysvec,
.interp_newpath = "/libexec/ld-elf32.so.1",
.brand_note = &elf32_freebsd_brandnote,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE,
.header_supported= elf32_arm_abi_supported,
};
SYSINIT(elf32, SI_SUB_EXEC, SI_ORDER_FIRST,
(sysinit_cfunc_t)elf32_insert_brand_entry, &freebsd32_brand_info);
static boolean_t
elf32_arm_abi_supported(struct image_params *imgp, int32_t *osrel __unused,
uint32_t *fctl0 __unused)
{
const Elf32_Ehdr *hdr;
/* Check if we support AArch32 */
if (ID_AA64PFR0_EL0_VAL(READ_SPECIALREG(id_aa64pfr0_el1)) !=
ID_AA64PFR0_EL0_64_32)
return (FALSE);
#define EF_ARM_EABI_VERSION(x) (((x) & EF_ARM_EABIMASK) >> 24)
#define EF_ARM_EABI_FREEBSD_MIN 4
hdr = (const Elf32_Ehdr *)imgp->image_header;
if (EF_ARM_EABI_VERSION(hdr->e_flags) < EF_ARM_EABI_FREEBSD_MIN) {
if (bootverbose)
uprintf("Attempting to execute non EABI binary "
"(rev %d) image %s",
EF_ARM_EABI_VERSION(hdr->e_flags),
imgp->args->fname);
return (FALSE);
}
return (TRUE);
}
static int
freebsd32_fetch_syscall_args(struct thread *td)
{
struct proc *p;
register_t *ap;
struct syscall_args *sa;
int error, i, nap, narg;
unsigned int args[4];
nap = 4;
p = td->td_proc;
ap = td->td_frame->tf_x;
sa = &td->td_sa;
/* r7 is the syscall id */
sa->code = td->td_frame->tf_x[7];
if (sa->code == SYS_syscall) {
sa->code = *ap++;
nap--;
} else if (sa->code == SYS___syscall) {
sa->code = ap[1];
nap -= 2;
ap += 2;
}
if (sa->code >= p->p_sysent->sv_size)
sa->callp = &p->p_sysent->sv_table[0];
else
sa->callp = &p->p_sysent->sv_table[sa->code];
narg = sa->callp->sy_narg;
for (i = 0; i < nap; i++)
sa->args[i] = ap[i];
if (narg > nap) {
if (narg - nap > nitems(args))
panic("Too many system call arguiments");
error = copyin((void *)td->td_frame->tf_x[13], args,
(narg - nap) * sizeof(int));
for (i = 0; i < (narg - nap); i++)
sa->args[i + nap] = args[i];
}
td->td_retval[0] = 0;
td->td_retval[1] = 0;
return (0);
}
static void
freebsd32_set_syscall_retval(struct thread *td, int error)
{
struct trapframe *frame;
frame = td->td_frame;
switch (error) {
case 0:
frame->tf_x[0] = td->td_retval[0];
frame->tf_x[1] = td->td_retval[1];
frame->tf_spsr &= ~PSR_C;
break;
case ERESTART:
/*
* Reconstruct the pc to point at the swi.
*/
if ((frame->tf_spsr & PSR_T) != 0)
frame->tf_elr -= 2; //THUMB_INSN_SIZE;
else
frame->tf_elr -= 4; //INSN_SIZE;
break;
case EJUSTRETURN:
/* nothing to do */
break;
default:
frame->tf_x[0] = error;
frame->tf_spsr |= PSR_C;
break;
}
}
static void
freebsd32_setregs(struct thread *td, struct image_params *imgp,
uintptr_t stack)
{
struct trapframe *tf = td->td_frame;
struct pcb *pcb = td->td_pcb;
memset(tf, 0, sizeof(struct trapframe));
/*
* We need to set x0 for init as it doesn't call
* cpu_set_syscall_retval to copy the value. We also
* need to set td_retval for the cases where we do.
*/
tf->tf_x[0] = stack;
/* SP_usr is mapped to x13 */
tf->tf_x[13] = stack;
/* LR_usr is mapped to x14 */
tf->tf_x[14] = imgp->entry_addr;
tf->tf_elr = imgp->entry_addr;
tf->tf_spsr = PSR_M_32;
#ifdef VFP
vfp_reset_state(td, pcb);
#endif
/*
* Clear debug register state. It is not applicable to the new process.
*/
bzero(&pcb->pcb_dbg_regs, sizeof(pcb->pcb_dbg_regs));
}
void
elf32_dump_thread(struct thread *td, void *dst, size_t *off)
{
/* XXX: VFP */
}
diff --git a/sys/arm64/arm64/elf_machdep.c b/sys/arm64/arm64/elf_machdep.c
index cd12d40c37df..82ecdf894911 100644
--- a/sys/arm64/arm64/elf_machdep.c
+++ b/sys/arm64/arm64/elf_machdep.c
@@ -1,285 +1,288 @@
/*-
* Copyright (c) 2014, 2015 The FreeBSD Foundation.
* Copyright (c) 2014 Andrew Turner.
* All rights reserved.
*
* This software was developed by Andrew Turner under
* sponsorship from the FreeBSD Foundation.
*
* Portions of this software were developed by Konstantin Belousov
* under sponsorship from the FreeBSD Foundation.
*
* 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 <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/kernel.h>
#include <sys/systm.h>
#include <sys/exec.h>
#include <sys/imgact.h>
#include <sys/linker.h>
#include <sys/proc.h>
#include <sys/sysent.h>
#include <sys/imgact_elf.h>
#include <sys/syscall.h>
#include <sys/signalvar.h>
#include <sys/vnode.h>
#include <vm/vm.h>
#include <vm/vm_param.h>
#include <machine/elf.h>
#include <machine/md_var.h>
#include "linker_if.h"
u_long __read_frequently elf_hwcap;
u_long __read_frequently elf_hwcap2;
static struct sysentvec elf64_freebsd_sysvec = {
.sv_size = SYS_MAXSYSCALL,
.sv_table = sysent,
.sv_transtrap = NULL,
.sv_fixup = __elfN(freebsd_fixup),
.sv_sendsig = sendsig,
.sv_sigcode = sigcode,
.sv_szsigcode = &szsigcode,
.sv_name = "FreeBSD ELF64",
.sv_coredump = __elfN(coredump),
+ .sv_elf_core_osabi = ELFOSABI_FREEBSD,
+ .sv_elf_core_abi_vendor = FREEBSD_ABI_VENDOR,
+ .sv_elf_core_prepare_notes = __elfN(prepare_notes),
.sv_imgact_try = NULL,
.sv_minsigstksz = MINSIGSTKSZ,
.sv_minuser = VM_MIN_ADDRESS,
.sv_maxuser = VM_MAXUSER_ADDRESS,
.sv_usrstack = USRSTACK,
.sv_psstrings = PS_STRINGS,
.sv_stackprot = VM_PROT_READ | VM_PROT_WRITE,
.sv_copyout_auxargs = __elfN(freebsd_copyout_auxargs),
.sv_copyout_strings = exec_copyout_strings,
.sv_setregs = exec_setregs,
.sv_fixlimit = NULL,
.sv_maxssiz = NULL,
.sv_flags = SV_SHP | SV_TIMEKEEP | SV_ABI_FREEBSD | SV_LP64 |
SV_ASLR | SV_RNG_SEED_VER,
.sv_set_syscall_retval = cpu_set_syscall_retval,
.sv_fetch_syscall_args = cpu_fetch_syscall_args,
.sv_syscallnames = syscallnames,
.sv_shared_page_base = SHAREDPAGE,
.sv_shared_page_len = PAGE_SIZE,
.sv_schedtail = NULL,
.sv_thread_detach = NULL,
.sv_trap = NULL,
.sv_hwcap = &elf_hwcap,
.sv_hwcap2 = &elf_hwcap2,
};
INIT_SYSENTVEC(elf64_sysvec, &elf64_freebsd_sysvec);
static Elf64_Brandinfo freebsd_brand_info = {
.brand = ELFOSABI_FREEBSD,
.machine = EM_AARCH64,
.compat_3_brand = "FreeBSD",
.emul_path = NULL,
.interp_path = "/libexec/ld-elf.so.1",
.sysvec = &elf64_freebsd_sysvec,
.interp_newpath = NULL,
.brand_note = &elf64_freebsd_brandnote,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
};
SYSINIT(elf64, SI_SUB_EXEC, SI_ORDER_FIRST,
(sysinit_cfunc_t)elf64_insert_brand_entry, &freebsd_brand_info);
void
elf64_dump_thread(struct thread *td __unused, void *dst __unused,
size_t *off __unused)
{
}
bool
elf_is_ifunc_reloc(Elf_Size r_info __unused)
{
return (ELF_R_TYPE(r_info) == R_AARCH64_IRELATIVE);
}
static int
reloc_instr_imm(Elf32_Addr *where, Elf_Addr val, u_int msb, u_int lsb)
{
/* Check bounds: upper bits must be all ones or all zeros. */
if ((uint64_t)((int64_t)val >> (msb + 1)) + 1 > 1)
return (-1);
val >>= lsb;
val &= (1 << (msb - lsb + 1)) - 1;
*where |= (Elf32_Addr)val;
return (0);
}
/*
* Process a relocation. Support for some static relocations is required
* in order for the -zifunc-noplt optimization to work.
*/
static int
elf_reloc_internal(linker_file_t lf, Elf_Addr relocbase, const void *data,
int type, int flags, elf_lookup_fn lookup)
{
#define ARM64_ELF_RELOC_LOCAL (1 << 0)
#define ARM64_ELF_RELOC_LATE_IFUNC (1 << 1)
Elf_Addr *where, addr, addend, val;
Elf_Word rtype, symidx;
const Elf_Rel *rel;
const Elf_Rela *rela;
int error;
switch (type) {
case ELF_RELOC_REL:
rel = (const Elf_Rel *)data;
where = (Elf_Addr *) (relocbase + rel->r_offset);
addend = *where;
rtype = ELF_R_TYPE(rel->r_info);
symidx = ELF_R_SYM(rel->r_info);
break;
case ELF_RELOC_RELA:
rela = (const Elf_Rela *)data;
where = (Elf_Addr *) (relocbase + rela->r_offset);
addend = rela->r_addend;
rtype = ELF_R_TYPE(rela->r_info);
symidx = ELF_R_SYM(rela->r_info);
break;
default:
panic("unknown reloc type %d\n", type);
}
if ((flags & ARM64_ELF_RELOC_LATE_IFUNC) != 0) {
KASSERT(type == ELF_RELOC_RELA,
("Only RELA ifunc relocations are supported"));
if (rtype != R_AARCH64_IRELATIVE)
return (0);
}
if ((flags & ARM64_ELF_RELOC_LOCAL) != 0) {
if (rtype == R_AARCH64_RELATIVE)
*where = elf_relocaddr(lf, relocbase + addend);
return (0);
}
error = 0;
switch (rtype) {
case R_AARCH64_NONE:
case R_AARCH64_RELATIVE:
break;
case R_AARCH64_TSTBR14:
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
error = reloc_instr_imm((Elf32_Addr *)where,
addr + addend - (Elf_Addr)where, 15, 2);
break;
case R_AARCH64_CONDBR19:
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
error = reloc_instr_imm((Elf32_Addr *)where,
addr + addend - (Elf_Addr)where, 20, 2);
break;
case R_AARCH64_JUMP26:
case R_AARCH64_CALL26:
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
error = reloc_instr_imm((Elf32_Addr *)where,
addr + addend - (Elf_Addr)where, 27, 2);
break;
case R_AARCH64_ABS64:
case R_AARCH64_GLOB_DAT:
case R_AARCH64_JUMP_SLOT:
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
*where = addr + addend;
break;
case R_AARCH64_IRELATIVE:
addr = relocbase + addend;
val = ((Elf64_Addr (*)(void))addr)();
if (*where != val)
*where = val;
break;
default:
printf("kldload: unexpected relocation type %d, "
"symbol index %d\n", rtype, symidx);
return (-1);
}
return (error);
}
int
elf_reloc_local(linker_file_t lf, Elf_Addr relocbase, const void *data,
int type, elf_lookup_fn lookup)
{
return (elf_reloc_internal(lf, relocbase, data, type,
ARM64_ELF_RELOC_LOCAL, lookup));
}
/* Process one elf relocation with addend. */
int
elf_reloc(linker_file_t lf, Elf_Addr relocbase, const void *data, int type,
elf_lookup_fn lookup)
{
return (elf_reloc_internal(lf, relocbase, data, type, 0, lookup));
}
int
elf_reloc_late(linker_file_t lf, Elf_Addr relocbase, const void *data,
int type, elf_lookup_fn lookup)
{
return (elf_reloc_internal(lf, relocbase, data, type,
ARM64_ELF_RELOC_LATE_IFUNC, lookup));
}
int
elf_cpu_load_file(linker_file_t lf)
{
if (lf->id != 1)
cpu_icache_sync_range((vm_offset_t)lf->address, lf->size);
return (0);
}
int
elf_cpu_unload_file(linker_file_t lf __unused)
{
return (0);
}
int
elf_cpu_parse_dynamic(caddr_t loadbase __unused, Elf_Dyn *dynamic __unused)
{
return (0);
}
diff --git a/sys/arm64/cloudabi32/cloudabi32_sysvec.c b/sys/arm64/cloudabi32/cloudabi32_sysvec.c
index eb8c51742dc7..889393560ede 100644
--- a/sys/arm64/cloudabi32/cloudabi32_sysvec.c
+++ b/sys/arm64/cloudabi32/cloudabi32_sysvec.c
@@ -1,203 +1,206 @@
/*-
* Copyright (c) 2015-2017 Nuxi, https://nuxi.nl/
*
* 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 <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/imgact.h>
#include <sys/kernel.h>
#include <sys/proc.h>
#include <sys/sysent.h>
#include <vm/vm.h>
#include <vm/pmap.h>
#include <machine/frame.h>
#include <machine/pcb.h>
#include <machine/vmparam.h>
#include <compat/cloudabi/cloudabi_util.h>
#include <compat/cloudabi32/cloudabi32_syscall.h>
#include <compat/cloudabi32/cloudabi32_util.h>
extern const char *cloudabi32_syscallnames[];
extern struct sysent cloudabi32_sysent[];
static void
cloudabi32_proc_setregs(struct thread *td, struct image_params *imgp,
uintptr_t stack)
{
struct trapframe *regs;
regs = td->td_frame;
memset(regs, 0, sizeof(*regs));
regs->tf_x[0] =
stack + roundup(sizeof(cloudabi32_tcb_t), sizeof(register_t));
regs->tf_x[13] = STACKALIGN(stack);
regs->tf_elr = imgp->entry_addr;
regs->tf_spsr |= PSR_AARCH32;
(void)cpu_set_user_tls(td, TO_PTR(stack));
}
static int
cloudabi32_fetch_syscall_args(struct thread *td)
{
struct trapframe *frame;
struct syscall_args *sa;
int error;
frame = td->td_frame;
sa = &td->td_sa;
/* Obtain system call number. */
sa->code = frame->tf_x[0];
if (sa->code >= CLOUDABI32_SYS_MAXSYSCALL)
return (ENOSYS);
sa->callp = &cloudabi32_sysent[sa->code];
/*
* Fetch system call arguments.
*
* The vDSO has already made sure that the arguments are
* eight-byte aligned. Pointers and size_t parameters are
* zero-extended. This makes it possible to copy in the
* arguments directly. As long as the call doesn't use 32-bit
* data structures, we can just invoke the same system call
* implementation used by 64-bit processes.
*/
error = copyin((void *)frame->tf_x[2], sa->args,
sa->callp->sy_narg * sizeof(sa->args[0]));
if (error != 0)
return (error);
/* Default system call return values. */
td->td_retval[0] = 0;
td->td_retval[1] = 0;
return (0);
}
static void
cloudabi32_set_syscall_retval(struct thread *td, int error)
{
struct trapframe *frame = td->td_frame;
switch (error) {
case 0:
/*
* System call succeeded.
*
* Simply copy out the 64-bit return values into the
* same buffer provided for system call arguments. The
* vDSO will copy them to the right spot, truncating
* pointers and size_t values to 32 bits.
*/
if (copyout(td->td_retval, (void *)frame->tf_x[2],
sizeof(td->td_retval)) == 0) {
frame->tf_x[0] = 0;
frame->tf_spsr &= ~PSR_C;
} else {
frame->tf_x[0] = CLOUDABI_EFAULT;
frame->tf_spsr |= PSR_C;
}
break;
case ERESTART:
/* Restart system call. */
frame->tf_elr -= 4;
break;
case EJUSTRETURN:
break;
default:
/* System call returned an error. */
frame->tf_x[0] = cloudabi_convert_errno(error);
frame->tf_spsr |= PSR_C;
break;
}
}
static void
cloudabi32_schedtail(struct thread *td)
{
struct trapframe *frame = td->td_frame;
register_t retval[2];
/* Return values for processes returning from fork. */
if ((td->td_pflags & TDP_FORKING) != 0) {
retval[0] = CLOUDABI_PROCESS_CHILD;
retval[1] = td->td_tid;
copyout(retval, (void *)frame->tf_x[2], sizeof(retval));
}
frame->tf_spsr |= PSR_AARCH32;
}
int
cloudabi32_thread_setregs(struct thread *td,
const cloudabi32_threadattr_t *attr, uint32_t tcb)
{
struct trapframe *frame;
/*
* Pass in the thread ID of the new thread and the argument
* pointer provided by the parent thread in as arguments to the
* entry point.
*/
frame = td->td_frame;
memset(frame, 0, sizeof(*frame));
frame->tf_x[0] = td->td_tid;
frame->tf_x[1] = attr->argument;
frame->tf_x[13] = STACKALIGN(attr->stack + attr->stack_len);
frame->tf_elr = attr->entry_point;
/* Set up TLS. */
return (cpu_set_user_tls(td, TO_PTR(tcb)));
}
static struct sysentvec cloudabi32_elf_sysvec = {
.sv_size = CLOUDABI32_SYS_MAXSYSCALL,
.sv_table = cloudabi32_sysent,
.sv_fixup = cloudabi32_fixup,
.sv_name = "CloudABI 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_minuser = VM_MIN_ADDRESS,
.sv_maxuser = (uintmax_t)1 << 32,
.sv_stackprot = VM_PROT_READ | VM_PROT_WRITE,
.sv_copyout_strings = cloudabi32_copyout_strings,
.sv_setregs = cloudabi32_proc_setregs,
.sv_flags = SV_ABI_CLOUDABI | SV_CAPSICUM | SV_ILP32,
.sv_set_syscall_retval = cloudabi32_set_syscall_retval,
.sv_fetch_syscall_args = cloudabi32_fetch_syscall_args,
.sv_syscallnames = cloudabi32_syscallnames,
.sv_schedtail = cloudabi32_schedtail,
};
INIT_SYSENTVEC(elf_sysvec, &cloudabi32_elf_sysvec);
Elf32_Brandinfo cloudabi32_brand = {
.brand = ELFOSABI_CLOUDABI,
.machine = EM_ARM,
.sysvec = &cloudabi32_elf_sysvec,
.flags = BI_BRAND_ONLY_STATIC,
};
diff --git a/sys/arm64/cloudabi64/cloudabi64_sysvec.c b/sys/arm64/cloudabi64/cloudabi64_sysvec.c
index 92deb356a112..bdbd828b7b62 100644
--- a/sys/arm64/cloudabi64/cloudabi64_sysvec.c
+++ b/sys/arm64/cloudabi64/cloudabi64_sysvec.c
@@ -1,187 +1,190 @@
/*-
* Copyright (c) 2015 Nuxi, https://nuxi.nl/
*
* 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 <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/imgact.h>
#include <sys/kernel.h>
#include <sys/proc.h>
#include <sys/sysent.h>
#include <vm/vm.h>
#include <vm/pmap.h>
#include <machine/frame.h>
#include <machine/pcb.h>
#include <machine/vmparam.h>
#include <compat/cloudabi/cloudabi_util.h>
#include <compat/cloudabi64/cloudabi64_syscall.h>
#include <compat/cloudabi64/cloudabi64_util.h>
extern const char *cloudabi64_syscallnames[];
extern struct sysent cloudabi64_sysent[];
static void
cloudabi64_proc_setregs(struct thread *td, struct image_params *imgp,
uintptr_t stack)
{
struct trapframe *regs;
exec_setregs(td, imgp, stack);
/*
* The stack now contains a pointer to the TCB and the auxiliary
* vector. Let x0 point to the auxiliary vector, and set
* tpidr_el0 to the TCB.
*/
regs = td->td_frame;
regs->tf_x[0] =
stack + roundup(sizeof(cloudabi64_tcb_t), sizeof(register_t));
(void)cpu_set_user_tls(td, TO_PTR(stack));
}
static int
cloudabi64_fetch_syscall_args(struct thread *td)
{
struct trapframe *frame;
struct syscall_args *sa;
int i;
frame = td->td_frame;
sa = &td->td_sa;
/* Obtain system call number. */
sa->code = frame->tf_x[8];
if (sa->code >= CLOUDABI64_SYS_MAXSYSCALL)
return (ENOSYS);
sa->callp = &cloudabi64_sysent[sa->code];
/* Fetch system call arguments. */
for (i = 0; i < MAXARGS; i++)
sa->args[i] = frame->tf_x[i];
/* Default system call return values. */
td->td_retval[0] = 0;
td->td_retval[1] = frame->tf_x[1];
return (0);
}
static void
cloudabi64_set_syscall_retval(struct thread *td, int error)
{
struct trapframe *frame = td->td_frame;
switch (error) {
case 0:
/* System call succeeded. */
frame->tf_x[0] = td->td_retval[0];
frame->tf_x[1] = td->td_retval[1];
frame->tf_spsr &= ~PSR_C;
break;
case ERESTART:
/* Restart system call. */
frame->tf_elr -= 4;
break;
case EJUSTRETURN:
break;
default:
/* System call returned an error. */
frame->tf_x[0] = cloudabi_convert_errno(error);
frame->tf_spsr |= PSR_C;
break;
}
}
static void
cloudabi64_schedtail(struct thread *td)
{
struct trapframe *frame = td->td_frame;
/*
* Initial register values for processes returning from fork.
* Make sure that we only set these values when forking, not
* when creating a new thread.
*/
if ((td->td_pflags & TDP_FORKING) != 0) {
frame->tf_x[0] = CLOUDABI_PROCESS_CHILD;
frame->tf_x[1] = td->td_tid;
}
}
int
cloudabi64_thread_setregs(struct thread *td,
const cloudabi64_threadattr_t *attr, uint64_t tcb)
{
struct trapframe *frame;
stack_t stack;
/* Perform standard register initialization. */
stack.ss_sp = TO_PTR(attr->stack);
stack.ss_size = attr->stack_len;
cpu_set_upcall(td, TO_PTR(attr->entry_point), NULL, &stack);
/*
* Pass in the thread ID of the new thread and the argument
* pointer provided by the parent thread in as arguments to the
* entry point.
*/
frame = td->td_frame;
frame->tf_x[0] = td->td_tid;
frame->tf_x[1] = attr->argument;
/* Set up TLS. */
return (cpu_set_user_tls(td, TO_PTR(tcb)));
}
static struct sysentvec cloudabi64_elf_sysvec = {
.sv_size = CLOUDABI64_SYS_MAXSYSCALL,
.sv_table = cloudabi64_sysent,
.sv_fixup = cloudabi64_fixup,
.sv_name = "CloudABI ELF64",
.sv_coredump = elf64_coredump,
+ .sv_elf_core_osabi = ELFOSABI_FREEBSD,
+ .sv_elf_core_abi_vendor = FREEBSD_ABI_VENDOR,
+ .sv_elf_core_prepare_notes = elf64_prepare_notes,
.sv_minuser = VM_MIN_ADDRESS,
.sv_maxuser = VM_MAXUSER_ADDRESS,
.sv_stackprot = VM_PROT_READ | VM_PROT_WRITE,
.sv_copyout_strings = cloudabi64_copyout_strings,
.sv_setregs = cloudabi64_proc_setregs,
.sv_flags = SV_ABI_CLOUDABI | SV_CAPSICUM | SV_LP64,
.sv_set_syscall_retval = cloudabi64_set_syscall_retval,
.sv_fetch_syscall_args = cloudabi64_fetch_syscall_args,
.sv_syscallnames = cloudabi64_syscallnames,
.sv_schedtail = cloudabi64_schedtail,
};
INIT_SYSENTVEC(elf_sysvec, &cloudabi64_elf_sysvec);
Elf64_Brandinfo cloudabi64_brand = {
.brand = ELFOSABI_CLOUDABI,
.machine = EM_AARCH64,
.sysvec = &cloudabi64_elf_sysvec,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_ONLY_STATIC,
};
diff --git a/sys/arm64/linux/linux_sysvec.c b/sys/arm64/linux/linux_sysvec.c
index afcca9f7e948..33bc303b5ada 100644
--- a/sys/arm64/linux/linux_sysvec.c
+++ b/sys/arm64/linux/linux_sysvec.c
@@ -1,586 +1,589 @@
/*-
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright (c) 1994-1996 Søren Schmidt
* Copyright (c) 2018 Turing Robotic Industries Inc.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/cdefs.h>
#include <sys/elf.h>
#include <sys/exec.h>
#include <sys/imgact.h>
#include <sys/imgact_elf.h>
#include <sys/kernel.h>
#include <sys/lock.h>
#include <sys/module.h>
#include <sys/mutex.h>
#include <sys/proc.h>
#include <sys/signalvar.h>
#include <sys/sysctl.h>
#include <sys/sysent.h>
#include <vm/vm_param.h>
#include <arm64/linux/linux.h>
#include <arm64/linux/linux_proto.h>
#include <compat/linux/linux_dtrace.h>
#include <compat/linux/linux_emul.h>
#include <compat/linux/linux_ioctl.h>
#include <compat/linux/linux_mib.h>
#include <compat/linux/linux_misc.h>
#include <compat/linux/linux_util.h>
#include <compat/linux/linux_vdso.h>
#include <machine/md_var.h>
#ifdef VFP
#include <machine/vfp.h>
#endif
MODULE_VERSION(linux64elf, 1);
static int linux_szsigcode;
static vm_object_t linux_shared_page_obj;
static char *linux_shared_page_mapping;
extern char _binary_linux_locore_o_start;
extern char _binary_linux_locore_o_end;
extern struct sysent linux_sysent[LINUX_SYS_MAXSYSCALL];
SET_DECLARE(linux_ioctl_handler_set, struct linux_ioctl_handler);
static int linux_copyout_strings(struct image_params *imgp,
uintptr_t *stack_base);
static int linux_elf_fixup(uintptr_t *stack_base,
struct image_params *iparams);
static bool linux_trans_osrel(const Elf_Note *note, int32_t *osrel);
static void linux_vdso_install(const void *param);
static void linux_vdso_deinstall(const void *param);
static void linux_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 int linux_vsyscall(struct thread *td);
/* DTrace init */
LIN_SDT_PROVIDER_DECLARE(LINUX_DTRACE);
/* DTrace probes */
LIN_SDT_PROBE_DEFINE2(sysvec, linux_translate_traps, todo, "int", "int");
LIN_SDT_PROBE_DEFINE0(sysvec, linux_exec_setregs, todo);
LIN_SDT_PROBE_DEFINE0(sysvec, linux_copyout_auxargs, todo);
LIN_SDT_PROBE_DEFINE0(sysvec, linux_elf_fixup, todo);
LIN_SDT_PROBE_DEFINE0(sysvec, linux_rt_sigreturn, todo);
LIN_SDT_PROBE_DEFINE0(sysvec, linux_rt_sendsig, todo);
LIN_SDT_PROBE_DEFINE0(sysvec, linux_vsyscall, todo);
LIN_SDT_PROBE_DEFINE0(sysvec, linux_vdso_install, todo);
LIN_SDT_PROBE_DEFINE0(sysvec, linux_vdso_deinstall, todo);
/* LINUXTODO: do we have traps to translate? */
static int
linux_translate_traps(int signal, int trap_code)
{
LIN_SDT_PROBE2(sysvec, linux_translate_traps, todo, signal, trap_code);
return (signal);
}
LINUX_VDSO_SYM_CHAR(linux_platform);
static int
linux_fetch_syscall_args(struct thread *td)
{
struct proc *p;
struct syscall_args *sa;
register_t *ap;
p = td->td_proc;
ap = td->td_frame->tf_x;
sa = &td->td_sa;
sa->code = td->td_frame->tf_x[8];
/* LINUXTODO: generic syscall? */
if (sa->code >= p->p_sysent->sv_size)
sa->callp = &p->p_sysent->sv_table[0];
else
sa->callp = &p->p_sysent->sv_table[sa->code];
if (sa->callp->sy_narg > MAXARGS)
panic("ARM64TODO: Could we have more than %d args?", MAXARGS);
memcpy(sa->args, ap, MAXARGS * sizeof(register_t));
td->td_retval[0] = 0;
return (0);
}
static void
linux_set_syscall_retval(struct thread *td, int error)
{
td->td_retval[1] = td->td_frame->tf_x[1];
cpu_set_syscall_retval(td, error);
if (__predict_false(error != 0)) {
if (error != ERESTART && error != EJUSTRETURN)
td->td_frame->tf_x[0] = bsd_to_linux_errno(error);
}
}
static int
linux_copyout_auxargs(struct image_params *imgp, uintptr_t base)
{
Elf_Auxargs *args;
Elf_Auxinfo *argarray, *pos;
struct proc *p;
int error, issetugid;
LIN_SDT_PROBE0(sysvec, linux_copyout_auxargs, todo);
p = imgp->proc;
args = (Elf64_Auxargs *)imgp->auxargs;
argarray = pos = malloc(LINUX_AT_COUNT * sizeof(*pos), M_TEMP,
M_WAITOK | M_ZERO);
issetugid = p->p_flag & P_SUGID ? 1 : 0;
AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO_EHDR,
imgp->proc->p_sysent->sv_shared_page_base);
AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, *imgp->sysent->sv_hwcap);
AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz);
AUXARGS_ENTRY(pos, LINUX_AT_CLKTCK, stclohz);
AUXARGS_ENTRY(pos, AT_PHDR, args->phdr);
AUXARGS_ENTRY(pos, AT_PHENT, args->phent);
AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum);
AUXARGS_ENTRY(pos, AT_BASE, args->base);
AUXARGS_ENTRY(pos, AT_FLAGS, args->flags);
AUXARGS_ENTRY(pos, AT_ENTRY, args->entry);
AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_ucred->cr_ruid);
AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid);
AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_ucred->cr_rgid);
AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid);
AUXARGS_ENTRY(pos, LINUX_AT_SECURE, issetugid);
AUXARGS_ENTRY_PTR(pos, LINUX_AT_RANDOM, imgp->canary);
AUXARGS_ENTRY(pos, LINUX_AT_HWCAP2, *imgp->sysent->sv_hwcap2);
if (imgp->execpathp != 0)
AUXARGS_ENTRY_PTR(pos, LINUX_AT_EXECFN, imgp->execpathp);
if (args->execfd != -1)
AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd);
AUXARGS_ENTRY(pos, LINUX_AT_PLATFORM, PTROUT(linux_platform));
AUXARGS_ENTRY(pos, AT_NULL, 0);
free(imgp->auxargs, M_TEMP);
imgp->auxargs = NULL;
KASSERT(pos - argarray <= LINUX_AT_COUNT, ("Too many auxargs"));
error = copyout(argarray, (void *)base,
sizeof(*argarray) * LINUX_AT_COUNT);
free(argarray, M_TEMP);
return (error);
}
static int
linux_elf_fixup(uintptr_t *stack_base, struct image_params *imgp)
{
LIN_SDT_PROBE0(sysvec, linux_elf_fixup, todo);
return (0);
}
/*
* Copy strings out to the new process address space, constructing new arg
* and env vector tables. Return a pointer to the base so that it can be used
* as the initial stack pointer.
* LINUXTODO: deduplicate against other linuxulator archs
*/
static int
linux_copyout_strings(struct image_params *imgp, uintptr_t *stack_base)
{
char **vectp;
char *stringp;
uintptr_t destp, ustringp;
struct ps_strings *arginfo;
char canary[LINUX_AT_RANDOM_LEN];
size_t execpath_len;
struct proc *p;
int argc, envc, error;
/* Calculate string base and vector table pointers. */
if (imgp->execpath != NULL && imgp->auxargs != NULL)
execpath_len = strlen(imgp->execpath) + 1;
else
execpath_len = 0;
p = imgp->proc;
arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings;
destp = (uintptr_t)arginfo;
if (execpath_len != 0) {
destp -= execpath_len;
destp = rounddown2(destp, sizeof(void *));
imgp->execpathp = (void *)destp;
error = copyout(imgp->execpath, imgp->execpathp, execpath_len);
if (error != 0)
return (error);
}
/* Prepare the canary for SSP. */
arc4rand(canary, sizeof(canary), 0);
destp -= roundup(sizeof(canary), sizeof(void *));
imgp->canary = (void *)destp;
error = copyout(canary, imgp->canary, sizeof(canary));
if (error != 0)
return (error);
/* Allocate room for the argument and environment strings. */
destp -= ARG_MAX - imgp->args->stringspace;
destp = rounddown2(destp, sizeof(void *));
ustringp = destp;
if (imgp->auxargs) {
/*
* Allocate room on the stack for the ELF auxargs
* array. It has up to LINUX_AT_COUNT entries.
*/
destp -= LINUX_AT_COUNT * sizeof(Elf64_Auxinfo);
destp = rounddown2(destp, sizeof(void *));
}
vectp = (char **)destp;
/*
* Allocate room for argc and the argv[] and env vectors including the
* terminating NULL pointers.
*/
vectp -= 1 + imgp->args->argc + 1 + imgp->args->envc + 1;
vectp = (char **)STACKALIGN(vectp);
/* vectp also becomes our initial stack base. */
*stack_base = (uintptr_t)vectp;
stringp = imgp->args->begin_argv;
argc = imgp->args->argc;
envc = imgp->args->envc;
/* Copy out strings - arguments and environment. */
error = copyout(stringp, (void *)ustringp,
ARG_MAX - imgp->args->stringspace);
if (error != 0)
return (error);
/* Fill in "ps_strings" struct for ps, w, etc. */
if (suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp) != 0 ||
suword(&arginfo->ps_nargvstr, argc) != 0)
return (EFAULT);
if (suword(vectp++, argc) != 0)
return (EFAULT);
/* Fill in argument portion of vector table. */
for (; argc > 0; --argc) {
if (suword(vectp++, ustringp) != 0)
return (EFAULT);
while (*stringp++ != 0)
ustringp++;
ustringp++;
}
/* A null vector table pointer separates the argp's from the envp's. */
if (suword(vectp++, 0) != 0)
return (EFAULT);
if (suword(&arginfo->ps_envstr, (long)(intptr_t)vectp) != 0 ||
suword(&arginfo->ps_nenvstr, envc) != 0)
return (EFAULT);
/* Fill in environment portion of vector table. */
for (; envc > 0; --envc) {
if (suword(vectp++, ustringp) != 0)
return (EFAULT);
while (*stringp++ != 0)
ustringp++;
ustringp++;
}
/* The end of the vector table is a null pointer. */
if (suword(vectp, 0) != 0)
return (EFAULT);
if (imgp->auxargs) {
vectp++;
error = imgp->sysent->sv_copyout_auxargs(imgp,
(uintptr_t)vectp);
if (error != 0)
return (error);
}
return (0);
}
/*
* Reset registers to default values on exec.
*/
static void
linux_exec_setregs(struct thread *td, struct image_params *imgp,
uintptr_t stack)
{
struct trapframe *regs = td->td_frame;
struct pcb *pcb = td->td_pcb;
/* LINUXTODO: validate */
LIN_SDT_PROBE0(sysvec, linux_exec_setregs, todo);
memset(regs, 0, sizeof(*regs));
/* glibc start.S registers function pointer in x0 with atexit. */
regs->tf_sp = stack;
#if 0 /* LINUXTODO: See if this is used. */
regs->tf_lr = imgp->entry_addr;
#else
regs->tf_lr = 0xffffffffffffffff;
#endif
regs->tf_elr = imgp->entry_addr;
pcb->pcb_tpidr_el0 = 0;
pcb->pcb_tpidrro_el0 = 0;
WRITE_SPECIALREG(tpidrro_el0, 0);
WRITE_SPECIALREG(tpidr_el0, 0);
#ifdef VFP
vfp_reset_state(td, pcb);
#endif
/*
* Clear debug register state. It is not applicable to the new process.
*/
bzero(&pcb->pcb_dbg_regs, sizeof(pcb->pcb_dbg_regs));
}
int
linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args)
{
/* LINUXTODO: implement */
LIN_SDT_PROBE0(sysvec, linux_rt_sigreturn, todo);
return (EDOOFUS);
}
static void
linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
{
/* LINUXTODO: implement */
LIN_SDT_PROBE0(sysvec, linux_rt_sendsig, todo);
}
static int
linux_vsyscall(struct thread *td)
{
/* LINUXTODO: implement */
LIN_SDT_PROBE0(sysvec, linux_vsyscall, todo);
return (EDOOFUS);
}
struct sysentvec elf_linux_sysvec = {
.sv_size = LINUX_SYS_MAXSYSCALL,
.sv_table = linux_sysent,
.sv_transtrap = linux_translate_traps,
.sv_fixup = linux_elf_fixup,
.sv_sendsig = linux_rt_sendsig,
.sv_sigcode = &_binary_linux_locore_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 = FREEBSD_ABI_VENDOR,
+ .sv_elf_core_prepare_notes = elf64_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 = USRSTACK,
.sv_psstrings = PS_STRINGS, /* XXX */
.sv_stackprot = VM_PROT_READ | VM_PROT_WRITE,
.sv_copyout_auxargs = linux_copyout_auxargs,
.sv_copyout_strings = linux_copyout_strings,
.sv_setregs = linux_exec_setregs,
.sv_fixlimit = NULL,
.sv_maxssiz = NULL,
.sv_flags = SV_ABI_LINUX | SV_LP64 | SV_SHP | SV_SIG_DISCIGN |
SV_SIG_WAITNDQ,
.sv_set_syscall_retval = linux_set_syscall_retval,
.sv_fetch_syscall_args = linux_fetch_syscall_args,
.sv_syscallnames = NULL,
.sv_shared_page_base = SHAREDPAGE,
.sv_shared_page_len = PAGE_SIZE,
.sv_schedtail = linux_schedtail,
.sv_thread_detach = linux_thread_detach,
.sv_trap = linux_vsyscall,
.sv_hwcap = &elf_hwcap,
.sv_hwcap2 = &elf_hwcap2,
.sv_onexec = linux_on_exec,
.sv_onexit = linux_on_exit,
.sv_ontdexit = linux_thread_dtor,
.sv_setid_allowed = &linux_setid_allowed_query,
};
static void
linux_vdso_install(const void *param)
{
linux_szsigcode = (&_binary_linux_locore_o_end -
&_binary_linux_locore_o_start);
if (linux_szsigcode > elf_linux_sysvec.sv_shared_page_len)
panic("invalid Linux VDSO size\n");
__elfN(linux_vdso_fixup)(&elf_linux_sysvec);
linux_shared_page_obj = __elfN(linux_shared_page_init)
(&linux_shared_page_mapping);
__elfN(linux_vdso_reloc)(&elf_linux_sysvec);
memcpy(linux_shared_page_mapping, elf_linux_sysvec.sv_sigcode,
linux_szsigcode);
elf_linux_sysvec.sv_shared_page_obj = linux_shared_page_obj;
}
SYSINIT(elf_linux_vdso_init, SI_SUB_EXEC, SI_ORDER_ANY,
linux_vdso_install, NULL);
static void
linux_vdso_deinstall(const void *param)
{
LIN_SDT_PROBE0(sysvec, linux_vdso_deinstall, todo);
__elfN(linux_shared_page_fini)(linux_shared_page_obj,
linux_shared_page_mapping);
}
SYSUNINIT(elf_linux_vdso_uninit, SI_SUB_EXEC, SI_ORDER_FIRST,
linux_vdso_deinstall, NULL);
static char GNU_ABI_VENDOR[] = "GNU";
static int GNU_ABI_LINUX = 0;
/* LINUXTODO: deduplicate */
static bool
linux_trans_osrel(const Elf_Note *note, int32_t *osrel)
{
const Elf32_Word *desc;
uintptr_t p;
p = (uintptr_t)(note + 1);
p += roundup2(note->n_namesz, sizeof(Elf32_Addr));
desc = (const Elf32_Word *)p;
if (desc[0] != GNU_ABI_LINUX)
return (false);
*osrel = LINUX_KERNVER(desc[1], desc[2], desc[3]);
return (true);
}
static Elf_Brandnote linux64_brandnote = {
.hdr.n_namesz = sizeof(GNU_ABI_VENDOR),
.hdr.n_descsz = 16,
.hdr.n_type = 1,
.vendor = GNU_ABI_VENDOR,
.flags = BN_TRANSLATE_OSREL,
.trans_osrel = linux_trans_osrel
};
static Elf64_Brandinfo linux_glibc2brand = {
.brand = ELFOSABI_LINUX,
.machine = EM_AARCH64,
.compat_3_brand = "Linux",
.emul_path = linux_emul_path,
.interp_path = "/lib64/ld-linux-x86-64.so.2",
.sysvec = &elf_linux_sysvec,
.interp_newpath = NULL,
.brand_note = &linux64_brandnote,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
};
Elf64_Brandinfo *linux_brandlist[] = {
&linux_glibc2brand,
NULL
};
static int
linux64_elf_modevent(module_t mod, int type, void *data)
{
Elf64_Brandinfo **brandinfo;
struct linux_ioctl_handler**lihp;
int error;
error = 0;
switch(type) {
case MOD_LOAD:
for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL;
++brandinfo)
if (elf64_insert_brand_entry(*brandinfo) < 0)
error = EINVAL;
if (error == 0) {
SET_FOREACH(lihp, linux_ioctl_handler_set)
linux_ioctl_register_handler(*lihp);
stclohz = (stathz ? stathz : hz);
if (bootverbose)
printf("Linux arm64 ELF exec handler installed\n");
}
break;
case MOD_UNLOAD:
for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL;
++brandinfo)
if (elf64_brand_inuse(*brandinfo))
error = EBUSY;
if (error == 0) {
for (brandinfo = &linux_brandlist[0];
*brandinfo != NULL; ++brandinfo)
if (elf64_remove_brand_entry(*brandinfo) < 0)
error = EINVAL;
}
if (error == 0) {
SET_FOREACH(lihp, linux_ioctl_handler_set)
linux_ioctl_unregister_handler(*lihp);
if (bootverbose)
printf("Linux ELF exec handler removed\n");
} else
printf("Could not deinstall ELF interpreter entry\n");
break;
default:
return (EOPNOTSUPP);
}
return (error);
}
static moduledata_t linux64_elf_mod = {
"linux64elf",
linux64_elf_modevent,
0
};
DECLARE_MODULE_TIED(linux64elf, linux64_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY);
MODULE_DEPEND(linux64elf, linux_common, 1, 1, 1);
FEATURE(linux64, "AArch64 Linux 64bit support");
diff --git a/sys/compat/ia32/ia32_sysvec.c b/sys/compat/ia32/ia32_sysvec.c
index f5a2c1200fc6..2b019e3b0323 100644
--- a/sys/compat/ia32/ia32_sysvec.c
+++ b/sys/compat/ia32/ia32_sysvec.c
@@ -1,233 +1,236 @@
/*-
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright (c) 2002 Doug Rabson
* Copyright (c) 2003 Peter Wemm
* 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 <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#define __ELF_WORD_SIZE 32
#include <sys/param.h>
#include <sys/exec.h>
#include <sys/fcntl.h>
#include <sys/imgact.h>
#include <sys/kernel.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/mutex.h>
#include <sys/mman.h>
#include <sys/namei.h>
#include <sys/proc.h>
#include <sys/procfs.h>
#include <sys/resourcevar.h>
#include <sys/systm.h>
#include <sys/signalvar.h>
#include <sys/stat.h>
#include <sys/sx.h>
#include <sys/syscall.h>
#include <sys/sysctl.h>
#include <sys/sysent.h>
#include <sys/vnode.h>
#include <sys/imgact_elf.h>
#include <vm/vm.h>
#include <vm/vm_kern.h>
#include <vm/vm_param.h>
#include <vm/pmap.h>
#include <vm/vm_map.h>
#include <vm/vm_object.h>
#include <vm/vm_extern.h>
#include <compat/freebsd32/freebsd32_signal.h>
#include <compat/freebsd32/freebsd32_util.h>
#include <compat/freebsd32/freebsd32_proto.h>
#include <compat/freebsd32/freebsd32_syscall.h>
#include <compat/ia32/ia32_signal.h>
#include <machine/frame.h>
#include <machine/md_var.h>
#include <machine/pcb.h>
#include <machine/cpufunc.h>
CTASSERT(sizeof(struct ia32_mcontext) == 640);
CTASSERT(sizeof(struct ia32_ucontext) == 704);
CTASSERT(sizeof(struct ia32_sigframe) == 800);
CTASSERT(sizeof(struct siginfo32) == 64);
#ifdef COMPAT_FREEBSD4
CTASSERT(sizeof(struct ia32_mcontext4) == 260);
CTASSERT(sizeof(struct ia32_ucontext4) == 324);
CTASSERT(sizeof(struct ia32_sigframe4) == 408);
#endif
extern const char *freebsd32_syscallnames[];
static SYSCTL_NODE(_compat, OID_AUTO, ia32, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
"ia32 mode");
static u_long ia32_maxdsiz = IA32_MAXDSIZ;
SYSCTL_ULONG(_compat_ia32, OID_AUTO, maxdsiz, CTLFLAG_RWTUN, &ia32_maxdsiz, 0, "");
u_long ia32_maxssiz = IA32_MAXSSIZ;
SYSCTL_ULONG(_compat_ia32, OID_AUTO, maxssiz, CTLFLAG_RWTUN, &ia32_maxssiz, 0, "");
static u_long ia32_maxvmem = IA32_MAXVMEM;
SYSCTL_ULONG(_compat_ia32, OID_AUTO, maxvmem, CTLFLAG_RWTUN, &ia32_maxvmem, 0, "");
struct sysentvec ia32_freebsd_sysvec = {
.sv_size = FREEBSD32_SYS_MAXSYSCALL,
.sv_table = freebsd32_sysent,
.sv_transtrap = NULL,
.sv_fixup = elf32_freebsd_fixup,
.sv_sendsig = ia32_sendsig,
.sv_sigcode = ia32_sigcode,
.sv_szsigcode = &sz_ia32_sigcode,
.sv_name = "FreeBSD 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 = NULL,
.sv_minsigstksz = MINSIGSTKSZ,
.sv_minuser = FREEBSD32_MINUSER,
.sv_maxuser = FREEBSD32_MAXUSER,
.sv_usrstack = FREEBSD32_USRSTACK,
.sv_psstrings = FREEBSD32_PS_STRINGS,
.sv_stackprot = VM_PROT_ALL,
.sv_copyout_auxargs = elf32_freebsd_copyout_auxargs,
.sv_copyout_strings = freebsd32_copyout_strings,
.sv_setregs = ia32_setregs,
.sv_fixlimit = ia32_fixlimit,
.sv_maxssiz = &ia32_maxssiz,
.sv_flags = SV_ABI_FREEBSD | SV_ASLR | SV_IA32 | SV_ILP32 |
SV_SHP | SV_TIMEKEEP | SV_RNG_SEED_VER,
.sv_set_syscall_retval = ia32_set_syscall_retval,
.sv_fetch_syscall_args = ia32_fetch_syscall_args,
.sv_syscallnames = freebsd32_syscallnames,
.sv_shared_page_base = FREEBSD32_SHAREDPAGE,
.sv_shared_page_len = PAGE_SIZE,
.sv_schedtail = NULL,
.sv_thread_detach = NULL,
.sv_trap = NULL,
.sv_stackgap = elf32_stackgap,
};
INIT_SYSENTVEC(elf_ia32_sysvec, &ia32_freebsd_sysvec);
static Elf32_Brandinfo ia32_brand_info = {
.brand = ELFOSABI_FREEBSD,
.machine = EM_386,
.compat_3_brand = "FreeBSD",
.emul_path = NULL,
.interp_path = "/libexec/ld-elf.so.1",
.sysvec = &ia32_freebsd_sysvec,
.interp_newpath = "/libexec/ld-elf32.so.1",
.brand_note = &elf32_freebsd_brandnote,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
};
SYSINIT(ia32, SI_SUB_EXEC, SI_ORDER_MIDDLE,
(sysinit_cfunc_t) elf32_insert_brand_entry,
&ia32_brand_info);
static Elf32_Brandinfo ia32_brand_oinfo = {
.brand = ELFOSABI_FREEBSD,
.machine = EM_386,
.compat_3_brand = "FreeBSD",
.emul_path = NULL,
.interp_path = "/usr/libexec/ld-elf.so.1",
.sysvec = &ia32_freebsd_sysvec,
.interp_newpath = "/libexec/ld-elf32.so.1",
.brand_note = &elf32_freebsd_brandnote,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
};
SYSINIT(oia32, SI_SUB_EXEC, SI_ORDER_ANY,
(sysinit_cfunc_t) elf32_insert_brand_entry,
&ia32_brand_oinfo);
static Elf32_Brandinfo kia32_brand_info = {
.brand = ELFOSABI_FREEBSD,
.machine = EM_386,
.compat_3_brand = "FreeBSD",
.emul_path = NULL,
.interp_path = "/lib/ld.so.1",
.sysvec = &ia32_freebsd_sysvec,
.brand_note = &elf32_kfreebsd_brandnote,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE_MANDATORY
};
SYSINIT(kia32, SI_SUB_EXEC, SI_ORDER_ANY,
(sysinit_cfunc_t) elf32_insert_brand_entry,
&kia32_brand_info);
void
elf32_dump_thread(struct thread *td, void *dst, size_t *off)
{
void *buf;
size_t len;
len = 0;
if (use_xsave) {
if (dst != NULL) {
fpugetregs(td);
len += elf32_populate_note(NT_X86_XSTATE,
get_pcb_user_save_td(td), dst,
cpu_max_ext_state_size, &buf);
*(uint64_t *)((char *)buf + X86_XSTATE_XCR0_OFFSET) =
xsave_mask;
} else
len += elf32_populate_note(NT_X86_XSTATE, NULL, NULL,
cpu_max_ext_state_size, NULL);
}
*off = len;
}
void
ia32_fixlimit(struct rlimit *rl, int which)
{
switch (which) {
case RLIMIT_DATA:
if (ia32_maxdsiz != 0) {
if (rl->rlim_cur > ia32_maxdsiz)
rl->rlim_cur = ia32_maxdsiz;
if (rl->rlim_max > ia32_maxdsiz)
rl->rlim_max = ia32_maxdsiz;
}
break;
case RLIMIT_STACK:
if (ia32_maxssiz != 0) {
if (rl->rlim_cur > ia32_maxssiz)
rl->rlim_cur = ia32_maxssiz;
if (rl->rlim_max > ia32_maxssiz)
rl->rlim_max = ia32_maxssiz;
}
break;
case RLIMIT_VMEM:
if (ia32_maxvmem != 0) {
if (rl->rlim_cur > ia32_maxvmem)
rl->rlim_cur = ia32_maxvmem;
if (rl->rlim_max > ia32_maxvmem)
rl->rlim_max = ia32_maxvmem;
}
break;
}
}
diff --git a/sys/i386/i386/elf_machdep.c b/sys/i386/i386/elf_machdep.c
index 0d870748967e..cb7b5617715c 100644
--- a/sys/i386/i386/elf_machdep.c
+++ b/sys/i386/i386/elf_machdep.c
@@ -1,302 +1,305 @@
/*-
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright 1996-1998 John D. Polstra.
* 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 ``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 <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_cpu.h"
#include <sys/param.h>
#include <sys/kernel.h>
#include <sys/systm.h>
#include <sys/exec.h>
#include <sys/imgact.h>
#include <sys/linker.h>
#include <sys/proc.h>
#include <sys/sysent.h>
#include <sys/imgact_elf.h>
#include <sys/syscall.h>
#include <sys/signalvar.h>
#include <sys/vnode.h>
#include <vm/vm.h>
#include <vm/pmap.h>
#include <vm/vm_param.h>
#include <machine/elf.h>
#include <machine/md_var.h>
#include <machine/npx.h>
struct sysentvec elf32_freebsd_sysvec = {
.sv_size = SYS_MAXSYSCALL,
.sv_table = sysent,
.sv_transtrap = NULL,
.sv_fixup = __elfN(freebsd_fixup),
.sv_sendsig = sendsig,
.sv_sigcode = sigcode,
.sv_szsigcode = &szsigcode,
.sv_name = "FreeBSD ELF32",
.sv_coredump = __elfN(coredump),
+ .sv_elf_core_osabi = ELFOSABI_FREEBSD,
+ .sv_elf_core_abi_vendor = FREEBSD_ABI_VENDOR,
+ .sv_elf_core_prepare_notes = __elfN(prepare_notes),
.sv_imgact_try = NULL,
.sv_minsigstksz = MINSIGSTKSZ,
.sv_minuser = VM_MIN_ADDRESS,
.sv_maxuser = VM_MAXUSER_ADDRESS,
.sv_usrstack = USRSTACK,
.sv_psstrings = PS_STRINGS,
.sv_stackprot = VM_PROT_ALL,
.sv_copyout_auxargs = __elfN(freebsd_copyout_auxargs),
.sv_copyout_strings = exec_copyout_strings,
.sv_setregs = exec_setregs,
.sv_fixlimit = NULL,
.sv_maxssiz = NULL,
.sv_flags = SV_ABI_FREEBSD | SV_ASLR | SV_IA32 | SV_ILP32 |
SV_SHP | SV_TIMEKEEP | SV_RNG_SEED_VER,
.sv_set_syscall_retval = cpu_set_syscall_retval,
.sv_fetch_syscall_args = cpu_fetch_syscall_args,
.sv_syscallnames = syscallnames,
.sv_shared_page_base = SHAREDPAGE,
.sv_shared_page_len = PAGE_SIZE,
.sv_schedtail = NULL,
.sv_thread_detach = NULL,
.sv_trap = NULL,
};
INIT_SYSENTVEC(elf32_sysvec, &elf32_freebsd_sysvec);
static Elf32_Brandinfo freebsd_brand_info = {
.brand = ELFOSABI_FREEBSD,
.machine = EM_386,
.compat_3_brand = "FreeBSD",
.emul_path = NULL,
.interp_path = "/libexec/ld-elf.so.1",
.sysvec = &elf32_freebsd_sysvec,
.interp_newpath = NULL,
.brand_note = &elf32_freebsd_brandnote,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
};
SYSINIT(elf32, SI_SUB_EXEC, SI_ORDER_FIRST,
(sysinit_cfunc_t) elf32_insert_brand_entry,
&freebsd_brand_info);
static Elf32_Brandinfo freebsd_brand_oinfo = {
.brand = ELFOSABI_FREEBSD,
.machine = EM_386,
.compat_3_brand = "FreeBSD",
.emul_path = NULL,
.interp_path = "/usr/libexec/ld-elf.so.1",
.sysvec = &elf32_freebsd_sysvec,
.interp_newpath = NULL,
.brand_note = &elf32_freebsd_brandnote,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
};
SYSINIT(oelf32, SI_SUB_EXEC, SI_ORDER_ANY,
(sysinit_cfunc_t) elf32_insert_brand_entry,
&freebsd_brand_oinfo);
static Elf32_Brandinfo kfreebsd_brand_info = {
.brand = ELFOSABI_FREEBSD,
.machine = EM_386,
.compat_3_brand = "FreeBSD",
.emul_path = NULL,
.interp_path = "/lib/ld.so.1",
.sysvec = &elf32_freebsd_sysvec,
.interp_newpath = NULL,
.brand_note = &elf32_kfreebsd_brandnote,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE_MANDATORY
};
SYSINIT(kelf32, SI_SUB_EXEC, SI_ORDER_ANY,
(sysinit_cfunc_t) elf32_insert_brand_entry,
&kfreebsd_brand_info);
void
elf32_dump_thread(struct thread *td, void *dst, size_t *off)
{
void *buf;
size_t len;
len = 0;
if (use_xsave) {
if (dst != NULL) {
npxgetregs(td);
len += elf32_populate_note(NT_X86_XSTATE,
get_pcb_user_save_td(td), dst,
cpu_max_ext_state_size, &buf);
*(uint64_t *)((char *)buf + X86_XSTATE_XCR0_OFFSET) =
xsave_mask;
} else
len += elf32_populate_note(NT_X86_XSTATE, NULL, NULL,
cpu_max_ext_state_size, NULL);
}
*off = len;
}
bool
elf_is_ifunc_reloc(Elf_Size r_info)
{
return (ELF_R_TYPE(r_info) == R_386_IRELATIVE);
}
#define ERI_LOCAL 0x0001
/* Process one elf relocation with addend. */
static int
elf_reloc_internal(linker_file_t lf, Elf_Addr relocbase, const void *data,
int type, elf_lookup_fn lookup, int flags)
{
Elf_Addr *where;
Elf_Addr addr;
Elf_Addr addend;
Elf_Word rtype, symidx;
const Elf_Rel *rel;
const Elf_Rela *rela;
int error;
switch (type) {
case ELF_RELOC_REL:
rel = (const Elf_Rel *)data;
where = (Elf_Addr *) (relocbase + rel->r_offset);
addend = *where;
rtype = ELF_R_TYPE(rel->r_info);
symidx = ELF_R_SYM(rel->r_info);
break;
case ELF_RELOC_RELA:
rela = (const Elf_Rela *)data;
where = (Elf_Addr *) (relocbase + rela->r_offset);
addend = rela->r_addend;
rtype = ELF_R_TYPE(rela->r_info);
symidx = ELF_R_SYM(rela->r_info);
break;
default:
panic("unknown reloc type %d\n", type);
}
if ((flags & ERI_LOCAL) != 0) {
if (rtype == R_386_RELATIVE) { /* A + B */
addr = elf_relocaddr(lf, relocbase + addend);
if (*where != addr)
*where = addr;
}
return (0);
}
switch (rtype) {
case R_386_NONE: /* none */
break;
case R_386_32: /* S + A */
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
addr += addend;
if (*where != addr)
*where = addr;
break;
case R_386_PC32: /* S + A - P */
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
addr += addend - (Elf_Addr)where;
if (*where != addr)
*where = addr;
break;
case R_386_COPY: /* none */
/*
* There shouldn't be copy relocations in kernel
* objects.
*/
printf("kldload: unexpected R_COPY relocation, "
"symbol index %d\n", symidx);
return (-1);
break;
case R_386_GLOB_DAT: /* S */
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
if (*where != addr)
*where = addr;
break;
case R_386_RELATIVE:
break;
case R_386_IRELATIVE:
addr = relocbase + addend;
addr = ((Elf_Addr (*)(void))addr)();
if (*where != addr)
*where = addr;
break;
default:
printf("kldload: unexpected relocation type %d, "
"symbol index %d\n", rtype, symidx);
return (-1);
}
return(0);
}
int
elf_reloc(linker_file_t lf, Elf_Addr relocbase, const void *data, int type,
elf_lookup_fn lookup)
{
return (elf_reloc_internal(lf, relocbase, data, type, lookup, 0));
}
int
elf_reloc_local(linker_file_t lf, Elf_Addr relocbase, const void *data,
int type, elf_lookup_fn lookup)
{
return (elf_reloc_internal(lf, relocbase, data, type, lookup,
ERI_LOCAL));
}
int
elf_cpu_load_file(linker_file_t lf __unused)
{
return (0);
}
int
elf_cpu_unload_file(linker_file_t lf __unused)
{
return (0);
}
int
elf_cpu_parse_dynamic(caddr_t loadbase __unused, Elf_Dyn *dynamic __unused)
{
return (0);
}
diff --git a/sys/i386/linux/linux_sysvec.c b/sys/i386/linux/linux_sysvec.c
index 9cc15bbfd5c3..adb70fded6dc 100644
--- a/sys/i386/linux/linux_sysvec.c
+++ b/sys/i386/linux/linux_sysvec.c
@@ -1,1060 +1,1063 @@
/*-
* 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 <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/exec.h>
#include <sys/fcntl.h>
#include <sys/imgact.h>
#include <sys/imgact_aout.h>
#include <sys/imgact_elf.h>
#include <sys/kernel.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/module.h>
#include <sys/mutex.h>
#include <sys/proc.h>
#include <sys/signalvar.h>
#include <sys/syscallsubr.h>
#include <sys/sysctl.h>
#include <sys/sysent.h>
#include <sys/sysproto.h>
#include <sys/vnode.h>
#include <vm/vm.h>
#include <vm/pmap.h>
#include <vm/vm_extern.h>
#include <vm/vm_map.h>
#include <vm/vm_object.h>
#include <vm/vm_page.h>
#include <vm/vm_param.h>
#include <machine/cpu.h>
#include <machine/cputypes.h>
#include <machine/md_var.h>
#include <machine/pcb.h>
#include <machine/trap.h>
#include <i386/linux/linux.h>
#include <i386/linux/linux_proto.h>
#include <compat/linux/linux_emul.h>
#include <compat/linux/linux_ioctl.h>
#include <compat/linux/linux_mib.h>
#include <compat/linux/linux_misc.h>
#include <compat/linux/linux_signal.h>
#include <compat/linux/linux_util.h>
#include <compat/linux/linux_vdso.h>
MODULE_VERSION(linux, 1);
#define LINUX_PS_STRINGS (LINUX_USRSTACK - sizeof(struct ps_strings))
static int linux_szsigcode;
static vm_object_t linux_shared_page_obj;
static char *linux_shared_page_mapping;
extern char _binary_linux_locore_o_start;
extern char _binary_linux_locore_o_end;
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 int linux_copyout_strings(struct image_params *imgp,
uintptr_t *stack_base);
static bool linux_trans_osrel(const Elf_Note *note, int32_t *osrel);
static void linux_vdso_install(void *param);
static void linux_vdso_deinstall(void *param);
#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)<nitems(_bsd_to_linux_trapcode)? \
_bsd_to_linux_trapcode[(code)]: \
LINUX_T_UNKNOWN)
LINUX_VDSO_SYM_CHAR(linux_platform);
LINUX_VDSO_SYM_INTPTR(linux_sigcode);
LINUX_VDSO_SYM_INTPTR(linux_rt_sigcode);
LINUX_VDSO_SYM_INTPTR(linux_vsyscall);
/*
* If FreeBSD & Linux have a difference of opinion about what a trap
* means, deal with it here.
*
* MPSAFE
*/
static int
linux_translate_traps(int signal, int trap_code)
{
if (signal != SIGBUS)
return (signal);
switch (trap_code) {
case T_PROTFLT:
case T_TSSFLT:
case T_DOUBLEFLT:
case T_PAGEFLT:
return (SIGSEGV);
default:
return (signal);
}
}
static int
linux_fixup(uintptr_t *stack_base, struct image_params *imgp)
{
register_t *base, *argv, *envp;
base = (register_t *)*stack_base;
argv = base;
envp = base + (imgp->args->argc + 1);
base--;
suword(base, (intptr_t)envp);
base--;
suword(base, (intptr_t)argv);
base--;
suword(base, imgp->args->argc);
*stack_base = (uintptr_t)base;
return (0);
}
static int
linux_copyout_auxargs(struct image_params *imgp, uintptr_t base)
{
struct proc *p;
Elf32_Auxargs *args;
Elf32_Auxinfo *argarray, *pos;
struct ps_strings *arginfo;
int error, issetugid;
p = imgp->proc;
issetugid = imgp->proc->p_flag & P_SUGID ? 1 : 0;
arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings;
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,
imgp->proc->p_sysent->sv_shared_page_base);
AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO, linux_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;
/* Calculate string base and vector table pointers. */
p = imgp->proc;
if (imgp->execpath != NULL && imgp->auxargs != NULL)
execpath_len = strlen(imgp->execpath) + 1;
else
execpath_len = 0;
arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings;
destp = (uintptr_t)arginfo;
if (execpath_len != 0) {
destp -= execpath_len;
destp = rounddown2(destp, sizeof(void *));
imgp->execpathp = (void *)destp;
error = copyout(imgp->execpath, imgp->execpathp, execpath_len);
if (error != 0)
return (error);
}
/* Prepare the canary for SSP. */
arc4rand(canary, sizeof(canary), 0);
destp -= roundup(sizeof(canary), sizeof(void *));
imgp->canary = (void *)destp;
error = copyout(canary, imgp->canary, sizeof(canary));
if (error != 0)
return (error);
/* Allocate room for the argument and environment strings. */
destp -= ARG_MAX - imgp->args->stringspace;
destp = rounddown2(destp, sizeof(void *));
ustringp = destp;
if (imgp->auxargs) {
/*
* Allocate room on the stack for the ELF auxargs
* array. It has LINUX_AT_COUNT entries.
*/
destp -= LINUX_AT_COUNT * sizeof(Elf32_Auxinfo);
destp = rounddown2(destp, sizeof(void *));
}
vectp = (char **)destp;
/*
* Allocate room for the argv[] and env vectors including the
* terminating NULL pointers.
*/
vectp -= imgp->args->argc + 1 + imgp->args->envc + 1;
/* vectp also becomes our initial stack base. */
*stack_base = (uintptr_t)vectp;
stringp = imgp->args->begin_argv;
argc = imgp->args->argc;
envc = imgp->args->envc;
/* Copy out strings - arguments and environment. */
error = copyout(stringp, (void *)ustringp,
ARG_MAX - imgp->args->stringspace);
if (error != 0)
return (error);
/* Fill in "ps_strings" struct for ps, w, etc. */
if (suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp) != 0 ||
suword(&arginfo->ps_nargvstr, argc) != 0)
return (EFAULT);
/* Fill in argument portion of vector table. */
for (; argc > 0; --argc) {
if (suword(vectp++, ustringp) != 0)
return (EFAULT);
while (*stringp++ != 0)
ustringp++;
ustringp++;
}
/* A null vector table pointer separates the argp's from the envp's. */
if (suword(vectp++, 0) != 0)
return (EFAULT);
if (suword(&arginfo->ps_envstr, (long)(intptr_t)vectp) != 0 ||
suword(&arginfo->ps_nenvstr, envc) != 0)
return (EFAULT);
/* Fill in environment portion of vector table. */
for (; envc > 0; --envc) {
if (suword(vectp++, ustringp) != 0)
return (EFAULT);
while (*stringp++ != 0)
ustringp++;
ustringp++;
}
/* The end of the vector table is a null pointer. */
if (suword(vectp, 0) != 0)
return (EFAULT);
if (imgp->auxargs) {
vectp++;
error = imgp->sysent->sv_copyout_auxargs(imgp,
(uintptr_t)vectp);
if (error != 0)
return (error);
}
return (0);
}
static void
linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
{
struct thread *td = curthread;
struct proc *p = td->td_proc;
struct sigacts *psp;
struct trapframe *regs;
struct l_rt_sigframe *fp, frame;
int sig, code;
int oonstack;
sig = ksi->ksi_signo;
code = ksi->ksi_code;
PROC_LOCK_ASSERT(p, MA_OWNED);
psp = p->p_sigacts;
mtx_assert(&psp->ps_mtx, MA_OWNED);
regs = td->td_frame;
oonstack = sigonstack(regs->tf_esp);
/* Allocate space for the signal handler context. */
if ((td->td_pflags & TDP_ALTSTACK) && !oonstack &&
SIGISMEMBER(psp->ps_sigonstack, sig)) {
fp = (struct l_rt_sigframe *)((uintptr_t)td->td_sigstk.ss_sp +
td->td_sigstk.ss_size - sizeof(struct l_rt_sigframe));
} else
fp = (struct l_rt_sigframe *)regs->tf_esp - 1;
mtx_unlock(&psp->ps_mtx);
/* Build the argument list for the signal handler. */
sig = bsd_to_linux_signal(sig);
bzero(&frame, sizeof(frame));
frame.sf_handler = catcher;
frame.sf_sig = sig;
frame.sf_siginfo = &fp->sf_si;
frame.sf_ucontext = &fp->sf_sc;
/* Fill in POSIX parts. */
siginfo_to_lsiginfo(&ksi->ksi_info, &frame.sf_si, sig);
/* Build the signal context to be used by sigreturn. */
frame.sf_sc.uc_flags = 0; /* XXX ??? */
frame.sf_sc.uc_link = NULL; /* XXX ??? */
frame.sf_sc.uc_stack.ss_sp = 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 = linux_rt_sigcode;
regs->tf_eflags &= ~(PSL_T | PSL_VM | PSL_D);
regs->tf_cs = _ucodesel;
regs->tf_ds = _udatasel;
regs->tf_es = _udatasel;
regs->tf_fs = _udatasel;
regs->tf_ss = _udatasel;
PROC_LOCK(p);
mtx_lock(&psp->ps_mtx);
}
/*
* Send an interrupt to process.
*
* Stack is set up to allow sigcode stored
* in u. to call routine, followed by kcall
* to sigreturn routine below. After sigreturn
* resets the signal mask, the stack, and the
* frame pointer, it returns to the user
* specified pc, psl.
*/
static void
linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
{
struct thread *td = curthread;
struct proc *p = td->td_proc;
struct sigacts *psp;
struct trapframe *regs;
struct l_sigframe *fp, frame;
l_sigset_t lmask;
int sig, code;
int oonstack;
PROC_LOCK_ASSERT(p, MA_OWNED);
psp = p->p_sigacts;
sig = ksi->ksi_signo;
code = ksi->ksi_code;
mtx_assert(&psp->ps_mtx, MA_OWNED);
if (SIGISMEMBER(psp->ps_siginfo, sig)) {
/* Signal handler installed with SA_SIGINFO. */
linux_rt_sendsig(catcher, ksi, mask);
return;
}
regs = td->td_frame;
oonstack = sigonstack(regs->tf_esp);
/* Allocate space for the signal handler context. */
if ((td->td_pflags & TDP_ALTSTACK) && !oonstack &&
SIGISMEMBER(psp->ps_sigonstack, sig)) {
fp = (struct l_sigframe *)((uintptr_t)td->td_sigstk.ss_sp +
td->td_sigstk.ss_size - sizeof(struct l_sigframe));
} else
fp = (struct l_sigframe *)regs->tf_esp - 1;
mtx_unlock(&psp->ps_mtx);
PROC_UNLOCK(p);
/* Build the argument list for the signal handler. */
sig = bsd_to_linux_signal(sig);
bzero(&frame, sizeof(frame));
frame.sf_handler = catcher;
frame.sf_sig = sig;
bsd_to_linux_sigset(mask, &lmask);
/* Build the signal context to be used by sigreturn. */
frame.sf_sc.sc_mask = lmask.__mask;
frame.sf_sc.sc_gs = rgs();
frame.sf_sc.sc_fs = regs->tf_fs;
frame.sf_sc.sc_es = regs->tf_es;
frame.sf_sc.sc_ds = regs->tf_ds;
frame.sf_sc.sc_edi = regs->tf_edi;
frame.sf_sc.sc_esi = regs->tf_esi;
frame.sf_sc.sc_ebp = regs->tf_ebp;
frame.sf_sc.sc_ebx = regs->tf_ebx;
frame.sf_sc.sc_esp = regs->tf_esp;
frame.sf_sc.sc_edx = regs->tf_edx;
frame.sf_sc.sc_ecx = regs->tf_ecx;
frame.sf_sc.sc_eax = regs->tf_eax;
frame.sf_sc.sc_eip = regs->tf_eip;
frame.sf_sc.sc_cs = regs->tf_cs;
frame.sf_sc.sc_eflags = regs->tf_eflags;
frame.sf_sc.sc_esp_at_signal = regs->tf_esp;
frame.sf_sc.sc_ss = regs->tf_ss;
frame.sf_sc.sc_err = regs->tf_err;
frame.sf_sc.sc_cr2 = (register_t)ksi->ksi_addr;
frame.sf_sc.sc_trapno = bsd_to_linux_trapcode(ksi->ksi_trapno);
frame.sf_extramask[0] = lmask.__mask;
if (copyout(&frame, fp, sizeof(frame)) != 0) {
/*
* Process has trashed its stack; give it an illegal
* instruction to halt it in its tracks.
*/
PROC_LOCK(p);
sigexit(td, SIGILL);
}
/* Build context to run handler in. */
regs->tf_esp = (int)fp;
regs->tf_eip = linux_sigcode;
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 = lss->ss_sp;
ss.ss_size = lss->ss_size;
ss.ss_flags = linux_to_bsd_sigaltstack(lss->ss_flags);
(void)kern_sigaltstack(td, &ss, NULL);
return (EJUSTRETURN);
}
static int
linux_fetch_syscall_args(struct thread *td)
{
struct proc *p;
struct trapframe *frame;
struct syscall_args *sa;
p = td->td_proc;
frame = td->td_frame;
sa = &td->td_sa;
sa->code = frame->tf_eax;
sa->args[0] = frame->tf_ebx;
sa->args[1] = frame->tf_ecx;
sa->args[2] = frame->tf_edx;
sa->args[3] = frame->tf_esi;
sa->args[4] = frame->tf_edi;
sa->args[5] = frame->tf_ebp; /* Unconfirmed */
if (sa->code >= p->p_sysent->sv_size)
/* nosys */
sa->callp = &p->p_sysent->sv_table[p->p_sysent->sv_size - 1];
else
sa->callp = &p->p_sysent->sv_table[sa->code];
td->td_retval[0] = 0;
td->td_retval[1] = frame->tf_edx;
return (0);
}
static void
linux_set_syscall_retval(struct thread *td, int error)
{
struct trapframe *frame = td->td_frame;
cpu_set_syscall_retval(td, error);
if (__predict_false(error != 0)) {
if (error != ERESTART && error != EJUSTRETURN)
frame->tf_eax = bsd_to_linux_errno(error);
}
}
/*
* 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_locore_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_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,
.sv_onexit = linux_on_exit,
.sv_ontdexit = linux_thread_dtor,
.sv_setid_allowed = &linux_setid_allowed_query,
};
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_locore_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_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_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,
.sv_onexit = linux_on_exit,
.sv_ontdexit = linux_thread_dtor,
.sv_setid_allowed = &linux_setid_allowed_query,
};
static void
linux_vdso_install(void *param)
{
linux_szsigcode = (&_binary_linux_locore_o_end -
&_binary_linux_locore_o_start);
if (linux_szsigcode > elf_linux_sysvec.sv_shared_page_len)
panic("Linux invalid vdso size\n");
__elfN(linux_vdso_fixup)(&elf_linux_sysvec);
linux_shared_page_obj = __elfN(linux_shared_page_init)
(&linux_shared_page_mapping);
__elfN(linux_vdso_reloc)(&elf_linux_sysvec);
bcopy(elf_linux_sysvec.sv_sigcode, linux_shared_page_mapping,
linux_szsigcode);
elf_linux_sysvec.sv_shared_page_obj = linux_shared_page_obj;
}
SYSINIT(elf_linux_vdso_init, SI_SUB_EXEC, SI_ORDER_ANY,
linux_vdso_install, NULL);
static void
linux_vdso_deinstall(void *param)
{
__elfN(linux_shared_page_fini)(linux_shared_page_obj,
linux_shared_page_mapping);
}
SYSUNINIT(elf_linux_vdso_uninit, SI_SUB_EXEC, SI_ORDER_FIRST,
linux_vdso_deinstall, NULL);
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
};
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);
LIST_INIT(&futex_list);
mtx_init(&futex_mtx, "ftllk", NULL, MTX_DEF);
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);
mtx_destroy(&futex_mtx);
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");
diff --git a/sys/kern/imgact_elf.c b/sys/kern/imgact_elf.c
index 4a649664c917..8c56660b1fe4 100644
--- a/sys/kern/imgact_elf.c
+++ b/sys/kern/imgact_elf.c
@@ -1,2697 +1,2697 @@
/*-
* SPDX-License-Identifier: BSD-3-Clause
*
* Copyright (c) 2017 Dell EMC
* Copyright (c) 2000-2001, 2003 David O'Brien
* Copyright (c) 1995-1996 Søren Schmidt
* Copyright (c) 1996 Peter Wemm
* 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 <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_capsicum.h"
#include <sys/param.h>
#include <sys/capsicum.h>
#include <sys/compressor.h>
#include <sys/exec.h>
#include <sys/fcntl.h>
#include <sys/imgact.h>
#include <sys/imgact_elf.h>
#include <sys/jail.h>
#include <sys/kernel.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/mount.h>
#include <sys/mman.h>
#include <sys/namei.h>
#include <sys/proc.h>
#include <sys/procfs.h>
#include <sys/ptrace.h>
#include <sys/racct.h>
#include <sys/resourcevar.h>
#include <sys/rwlock.h>
#include <sys/sbuf.h>
#include <sys/sf_buf.h>
#include <sys/smp.h>
#include <sys/systm.h>
#include <sys/signalvar.h>
#include <sys/stat.h>
#include <sys/sx.h>
#include <sys/syscall.h>
#include <sys/sysctl.h>
#include <sys/sysent.h>
#include <sys/vnode.h>
#include <sys/syslog.h>
#include <sys/eventhandler.h>
#include <sys/user.h>
#include <vm/vm.h>
#include <vm/vm_kern.h>
#include <vm/vm_param.h>
#include <vm/pmap.h>
#include <vm/vm_map.h>
#include <vm/vm_object.h>
#include <vm/vm_extern.h>
#include <machine/elf.h>
#include <machine/md_var.h>
#define ELF_NOTE_ROUNDSIZE 4
#define OLD_EI_BRAND 8
static int __elfN(check_header)(const Elf_Ehdr *hdr);
static Elf_Brandinfo *__elfN(get_brandinfo)(struct image_params *imgp,
const char *interp, int32_t *osrel, uint32_t *fctl0);
static int __elfN(load_file)(struct proc *p, const char *file, u_long *addr,
u_long *entry);
static int __elfN(load_section)(struct image_params *imgp, vm_ooffset_t offset,
caddr_t vmaddr, size_t memsz, size_t filsz, vm_prot_t prot);
static int __CONCAT(exec_, __elfN(imgact))(struct image_params *imgp);
static bool __elfN(freebsd_trans_osrel)(const Elf_Note *note,
int32_t *osrel);
static bool kfreebsd_trans_osrel(const Elf_Note *note, int32_t *osrel);
static boolean_t __elfN(check_note)(struct image_params *imgp,
Elf_Brandnote *checknote, int32_t *osrel, boolean_t *has_fctl0,
uint32_t *fctl0);
static vm_prot_t __elfN(trans_prot)(Elf_Word);
static Elf_Word __elfN(untrans_prot)(vm_prot_t);
SYSCTL_NODE(_kern, OID_AUTO, __CONCAT(elf, __ELF_WORD_SIZE),
CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
"");
int __elfN(fallback_brand) = -1;
SYSCTL_INT(__CONCAT(_kern_elf, __ELF_WORD_SIZE), OID_AUTO,
fallback_brand, CTLFLAG_RWTUN, &__elfN(fallback_brand), 0,
__XSTRING(__CONCAT(ELF, __ELF_WORD_SIZE)) " brand of last resort");
static int elf_legacy_coredump = 0;
SYSCTL_INT(_debug, OID_AUTO, __elfN(legacy_coredump), CTLFLAG_RW,
&elf_legacy_coredump, 0,
"include all and only RW pages in core dumps");
int __elfN(nxstack) =
#if defined(__amd64__) || defined(__powerpc64__) /* both 64 and 32 bit */ || \
(defined(__arm__) && __ARM_ARCH >= 7) || defined(__aarch64__) || \
defined(__riscv)
1;
#else
0;
#endif
SYSCTL_INT(__CONCAT(_kern_elf, __ELF_WORD_SIZE), OID_AUTO,
nxstack, CTLFLAG_RW, &__elfN(nxstack), 0,
__XSTRING(__CONCAT(ELF, __ELF_WORD_SIZE)) ": enable non-executable stack");
#if __ELF_WORD_SIZE == 32 && (defined(__amd64__) || defined(__i386__))
int i386_read_exec = 0;
SYSCTL_INT(_kern_elf32, OID_AUTO, read_exec, CTLFLAG_RW, &i386_read_exec, 0,
"enable execution from readable segments");
#endif
static u_long __elfN(pie_base) = ET_DYN_LOAD_ADDR;
static int
sysctl_pie_base(SYSCTL_HANDLER_ARGS)
{
u_long val;
int error;
val = __elfN(pie_base);
error = sysctl_handle_long(oidp, &val, 0, req);
if (error != 0 || req->newptr == NULL)
return (error);
if ((val & PAGE_MASK) != 0)
return (EINVAL);
__elfN(pie_base) = val;
return (0);
}
SYSCTL_PROC(__CONCAT(_kern_elf, __ELF_WORD_SIZE), OID_AUTO, pie_base,
CTLTYPE_ULONG | CTLFLAG_MPSAFE | CTLFLAG_RW, NULL, 0,
sysctl_pie_base, "LU",
"PIE load base without randomization");
SYSCTL_NODE(__CONCAT(_kern_elf, __ELF_WORD_SIZE), OID_AUTO, aslr,
CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
"");
#define ASLR_NODE_OID __CONCAT(__CONCAT(_kern_elf, __ELF_WORD_SIZE), _aslr)
static int __elfN(aslr_enabled) = 0;
SYSCTL_INT(ASLR_NODE_OID, OID_AUTO, enable, CTLFLAG_RWTUN,
&__elfN(aslr_enabled), 0,
__XSTRING(__CONCAT(ELF, __ELF_WORD_SIZE))
": enable address map randomization");
static int __elfN(pie_aslr_enabled) = 0;
SYSCTL_INT(ASLR_NODE_OID, OID_AUTO, pie_enable, CTLFLAG_RWTUN,
&__elfN(pie_aslr_enabled), 0,
__XSTRING(__CONCAT(ELF, __ELF_WORD_SIZE))
": enable address map randomization for PIE binaries");
static int __elfN(aslr_honor_sbrk) = 1;
SYSCTL_INT(ASLR_NODE_OID, OID_AUTO, honor_sbrk, CTLFLAG_RW,
&__elfN(aslr_honor_sbrk), 0,
__XSTRING(__CONCAT(ELF, __ELF_WORD_SIZE)) ": assume sbrk is used");
static int __elfN(aslr_stack_gap) = 3;
SYSCTL_INT(ASLR_NODE_OID, OID_AUTO, stack_gap, CTLFLAG_RW,
&__elfN(aslr_stack_gap), 0,
__XSTRING(__CONCAT(ELF, __ELF_WORD_SIZE))
": maximum percentage of main stack to waste on a random gap");
static int __elfN(sigfastblock) = 1;
SYSCTL_INT(__CONCAT(_kern_elf, __ELF_WORD_SIZE), OID_AUTO, sigfastblock,
CTLFLAG_RWTUN, &__elfN(sigfastblock), 0,
"enable sigfastblock for new processes");
static bool __elfN(allow_wx) = true;
SYSCTL_BOOL(__CONCAT(_kern_elf, __ELF_WORD_SIZE), OID_AUTO, allow_wx,
CTLFLAG_RWTUN, &__elfN(allow_wx), 0,
"Allow pages to be mapped simultaneously writable and executable");
static Elf_Brandinfo *elf_brand_list[MAX_BRANDS];
#define aligned(a, t) (rounddown2((u_long)(a), sizeof(t)) == (u_long)(a))
-static const char FREEBSD_ABI_VENDOR[] = "FreeBSD";
-
Elf_Brandnote __elfN(freebsd_brandnote) = {
.hdr.n_namesz = sizeof(FREEBSD_ABI_VENDOR),
.hdr.n_descsz = sizeof(int32_t),
.hdr.n_type = NT_FREEBSD_ABI_TAG,
.vendor = FREEBSD_ABI_VENDOR,
.flags = BN_TRANSLATE_OSREL,
.trans_osrel = __elfN(freebsd_trans_osrel)
};
static bool
__elfN(freebsd_trans_osrel)(const Elf_Note *note, int32_t *osrel)
{
uintptr_t p;
p = (uintptr_t)(note + 1);
p += roundup2(note->n_namesz, ELF_NOTE_ROUNDSIZE);
*osrel = *(const int32_t *)(p);
return (true);
}
static const char GNU_ABI_VENDOR[] = "GNU";
static int GNU_KFREEBSD_ABI_DESC = 3;
Elf_Brandnote __elfN(kfreebsd_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 = kfreebsd_trans_osrel
};
static bool
kfreebsd_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, ELF_NOTE_ROUNDSIZE);
desc = (const Elf32_Word *)p;
if (desc[0] != GNU_KFREEBSD_ABI_DESC)
return (false);
/*
* Debian GNU/kFreeBSD embed the earliest compatible kernel version
* (__FreeBSD_version: <major><two digit minor>Rxx) in the LSB way.
*/
*osrel = desc[1] * 100000 + desc[2] * 1000 + desc[3];
return (true);
}
int
__elfN(insert_brand_entry)(Elf_Brandinfo *entry)
{
int i;
for (i = 0; i < MAX_BRANDS; i++) {
if (elf_brand_list[i] == NULL) {
elf_brand_list[i] = entry;
break;
}
}
if (i == MAX_BRANDS) {
printf("WARNING: %s: could not insert brandinfo entry: %p\n",
__func__, entry);
return (-1);
}
return (0);
}
int
__elfN(remove_brand_entry)(Elf_Brandinfo *entry)
{
int i;
for (i = 0; i < MAX_BRANDS; i++) {
if (elf_brand_list[i] == entry) {
elf_brand_list[i] = NULL;
break;
}
}
if (i == MAX_BRANDS)
return (-1);
return (0);
}
int
__elfN(brand_inuse)(Elf_Brandinfo *entry)
{
struct proc *p;
int rval = FALSE;
sx_slock(&allproc_lock);
FOREACH_PROC_IN_SYSTEM(p) {
if (p->p_sysent == entry->sysvec) {
rval = TRUE;
break;
}
}
sx_sunlock(&allproc_lock);
return (rval);
}
static Elf_Brandinfo *
__elfN(get_brandinfo)(struct image_params *imgp, const char *interp,
int32_t *osrel, uint32_t *fctl0)
{
const Elf_Ehdr *hdr = (const Elf_Ehdr *)imgp->image_header;
Elf_Brandinfo *bi, *bi_m;
boolean_t ret, has_fctl0;
int i, interp_name_len;
interp_name_len = interp != NULL ? strlen(interp) + 1 : 0;
/*
* We support four types of branding -- (1) the ELF EI_OSABI field
* that SCO added to the ELF spec, (2) FreeBSD 3.x's traditional string
* branding w/in the ELF header, (3) path of the `interp_path'
* field, and (4) the ".note.ABI-tag" ELF section.
*/
/* Look for an ".note.ABI-tag" ELF section */
bi_m = NULL;
for (i = 0; i < MAX_BRANDS; i++) {
bi = elf_brand_list[i];
if (bi == NULL)
continue;
if (interp != NULL && (bi->flags & BI_BRAND_ONLY_STATIC) != 0)
continue;
if (hdr->e_machine == bi->machine && (bi->flags &
(BI_BRAND_NOTE|BI_BRAND_NOTE_MANDATORY)) != 0) {
has_fctl0 = false;
*fctl0 = 0;
*osrel = 0;
ret = __elfN(check_note)(imgp, bi->brand_note, osrel,
&has_fctl0, fctl0);
/* Give brand a chance to veto check_note's guess */
if (ret && bi->header_supported) {
ret = bi->header_supported(imgp, osrel,
has_fctl0 ? fctl0 : NULL);
}
/*
* If note checker claimed the binary, but the
* interpreter path in the image does not
* match default one for the brand, try to
* search for other brands with the same
* interpreter. Either there is better brand
* with the right interpreter, or, failing
* this, we return first brand which accepted
* our note and, optionally, header.
*/
if (ret && bi_m == NULL && interp != NULL &&
(bi->interp_path == NULL ||
(strlen(bi->interp_path) + 1 != interp_name_len ||
strncmp(interp, bi->interp_path, interp_name_len)
!= 0))) {
bi_m = bi;
ret = 0;
}
if (ret)
return (bi);
}
}
if (bi_m != NULL)
return (bi_m);
/* If the executable has a brand, search for it in the brand list. */
for (i = 0; i < MAX_BRANDS; i++) {
bi = elf_brand_list[i];
if (bi == NULL || (bi->flags & BI_BRAND_NOTE_MANDATORY) != 0 ||
(interp != NULL && (bi->flags & BI_BRAND_ONLY_STATIC) != 0))
continue;
if (hdr->e_machine == bi->machine &&
(hdr->e_ident[EI_OSABI] == bi->brand ||
(bi->compat_3_brand != NULL &&
strcmp((const char *)&hdr->e_ident[OLD_EI_BRAND],
bi->compat_3_brand) == 0))) {
/* Looks good, but give brand a chance to veto */
if (bi->header_supported == NULL ||
bi->header_supported(imgp, NULL, NULL)) {
/*
* Again, prefer strictly matching
* interpreter path.
*/
if (interp_name_len == 0 &&
bi->interp_path == NULL)
return (bi);
if (bi->interp_path != NULL &&
strlen(bi->interp_path) + 1 ==
interp_name_len && strncmp(interp,
bi->interp_path, interp_name_len) == 0)
return (bi);
if (bi_m == NULL)
bi_m = bi;
}
}
}
if (bi_m != NULL)
return (bi_m);
/* No known brand, see if the header is recognized by any brand */
for (i = 0; i < MAX_BRANDS; i++) {
bi = elf_brand_list[i];
if (bi == NULL || bi->flags & BI_BRAND_NOTE_MANDATORY ||
bi->header_supported == NULL)
continue;
if (hdr->e_machine == bi->machine) {
ret = bi->header_supported(imgp, NULL, NULL);
if (ret)
return (bi);
}
}
/* Lacking a known brand, search for a recognized interpreter. */
if (interp != NULL) {
for (i = 0; i < MAX_BRANDS; i++) {
bi = elf_brand_list[i];
if (bi == NULL || (bi->flags &
(BI_BRAND_NOTE_MANDATORY | BI_BRAND_ONLY_STATIC))
!= 0)
continue;
if (hdr->e_machine == bi->machine &&
bi->interp_path != NULL &&
/* ELF image p_filesz includes terminating zero */
strlen(bi->interp_path) + 1 == interp_name_len &&
strncmp(interp, bi->interp_path, interp_name_len)
== 0 && (bi->header_supported == NULL ||
bi->header_supported(imgp, NULL, NULL)))
return (bi);
}
}
/* Lacking a recognized interpreter, try the default brand */
for (i = 0; i < MAX_BRANDS; i++) {
bi = elf_brand_list[i];
if (bi == NULL || (bi->flags & BI_BRAND_NOTE_MANDATORY) != 0 ||
(interp != NULL && (bi->flags & BI_BRAND_ONLY_STATIC) != 0))
continue;
if (hdr->e_machine == bi->machine &&
__elfN(fallback_brand) == bi->brand &&
(bi->header_supported == NULL ||
bi->header_supported(imgp, NULL, NULL)))
return (bi);
}
return (NULL);
}
static bool
__elfN(phdr_in_zero_page)(const Elf_Ehdr *hdr)
{
return (hdr->e_phoff <= PAGE_SIZE &&
(u_int)hdr->e_phentsize * hdr->e_phnum <= PAGE_SIZE - hdr->e_phoff);
}
static int
__elfN(check_header)(const Elf_Ehdr *hdr)
{
Elf_Brandinfo *bi;
int i;
if (!IS_ELF(*hdr) ||
hdr->e_ident[EI_CLASS] != ELF_TARG_CLASS ||
hdr->e_ident[EI_DATA] != ELF_TARG_DATA ||
hdr->e_ident[EI_VERSION] != EV_CURRENT ||
hdr->e_phentsize != sizeof(Elf_Phdr) ||
hdr->e_version != ELF_TARG_VER)
return (ENOEXEC);
/*
* Make sure we have at least one brand for this machine.
*/
for (i = 0; i < MAX_BRANDS; i++) {
bi = elf_brand_list[i];
if (bi != NULL && bi->machine == hdr->e_machine)
break;
}
if (i == MAX_BRANDS)
return (ENOEXEC);
return (0);
}
static int
__elfN(map_partial)(vm_map_t map, vm_object_t object, vm_ooffset_t offset,
vm_offset_t start, vm_offset_t end, vm_prot_t prot)
{
struct sf_buf *sf;
int error;
vm_offset_t off;
/*
* Create the page if it doesn't exist yet. Ignore errors.
*/
vm_map_fixed(map, NULL, 0, trunc_page(start), round_page(end) -
trunc_page(start), VM_PROT_ALL, VM_PROT_ALL, MAP_CHECK_EXCL);
/*
* Find the page from the underlying object.
*/
if (object != NULL) {
sf = vm_imgact_map_page(object, offset);
if (sf == NULL)
return (KERN_FAILURE);
off = offset - trunc_page(offset);
error = copyout((caddr_t)sf_buf_kva(sf) + off, (caddr_t)start,
end - start);
vm_imgact_unmap_page(sf);
if (error != 0)
return (KERN_FAILURE);
}
return (KERN_SUCCESS);
}
static int
__elfN(map_insert)(struct image_params *imgp, vm_map_t map, vm_object_t object,
vm_ooffset_t offset, vm_offset_t start, vm_offset_t end, vm_prot_t prot,
int cow)
{
struct sf_buf *sf;
vm_offset_t off;
vm_size_t sz;
int error, locked, rv;
if (start != trunc_page(start)) {
rv = __elfN(map_partial)(map, object, offset, start,
round_page(start), prot);
if (rv != KERN_SUCCESS)
return (rv);
offset += round_page(start) - start;
start = round_page(start);
}
if (end != round_page(end)) {
rv = __elfN(map_partial)(map, object, offset +
trunc_page(end) - start, trunc_page(end), end, prot);
if (rv != KERN_SUCCESS)
return (rv);
end = trunc_page(end);
}
if (start >= end)
return (KERN_SUCCESS);
if ((offset & PAGE_MASK) != 0) {
/*
* The mapping is not page aligned. This means that we have
* to copy the data.
*/
rv = vm_map_fixed(map, NULL, 0, start, end - start,
prot | VM_PROT_WRITE, VM_PROT_ALL, MAP_CHECK_EXCL);
if (rv != KERN_SUCCESS)
return (rv);
if (object == NULL)
return (KERN_SUCCESS);
for (; start < end; start += sz) {
sf = vm_imgact_map_page(object, offset);
if (sf == NULL)
return (KERN_FAILURE);
off = offset - trunc_page(offset);
sz = end - start;
if (sz > PAGE_SIZE - off)
sz = PAGE_SIZE - off;
error = copyout((caddr_t)sf_buf_kva(sf) + off,
(caddr_t)start, sz);
vm_imgact_unmap_page(sf);
if (error != 0)
return (KERN_FAILURE);
offset += sz;
}
} else {
vm_object_reference(object);
rv = vm_map_fixed(map, object, offset, start, end - start,
prot, VM_PROT_ALL, cow | MAP_CHECK_EXCL |
(object != NULL ? MAP_VN_EXEC : 0));
if (rv != KERN_SUCCESS) {
locked = VOP_ISLOCKED(imgp->vp);
VOP_UNLOCK(imgp->vp);
vm_object_deallocate(object);
vn_lock(imgp->vp, locked | LK_RETRY);
return (rv);
} else if (object != NULL) {
MPASS(imgp->vp->v_object == object);
VOP_SET_TEXT_CHECKED(imgp->vp);
}
}
return (KERN_SUCCESS);
}
static int
__elfN(load_section)(struct image_params *imgp, vm_ooffset_t offset,
caddr_t vmaddr, size_t memsz, size_t filsz, vm_prot_t prot)
{
struct sf_buf *sf;
size_t map_len;
vm_map_t map;
vm_object_t object;
vm_offset_t map_addr;
int error, rv, cow;
size_t copy_len;
vm_ooffset_t file_addr;
/*
* It's necessary to fail if the filsz + offset taken from the
* header is greater than the actual file pager object's size.
* If we were to allow this, then the vm_map_find() below would
* walk right off the end of the file object and into the ether.
*
* While I'm here, might as well check for something else that
* is invalid: filsz cannot be greater than memsz.
*/
if ((filsz != 0 && (off_t)filsz + offset > imgp->attr->va_size) ||
filsz > memsz) {
uprintf("elf_load_section: truncated ELF file\n");
return (ENOEXEC);
}
object = imgp->object;
map = &imgp->proc->p_vmspace->vm_map;
map_addr = trunc_page((vm_offset_t)vmaddr);
file_addr = trunc_page(offset);
/*
* We have two choices. We can either clear the data in the last page
* of an oversized mapping, or we can start the anon mapping a page
* early and copy the initialized data into that first page. We
* choose the second.
*/
if (filsz == 0)
map_len = 0;
else if (memsz > filsz)
map_len = trunc_page(offset + filsz) - file_addr;
else
map_len = round_page(offset + filsz) - file_addr;
if (map_len != 0) {
/* cow flags: don't dump readonly sections in core */
cow = MAP_COPY_ON_WRITE | MAP_PREFAULT |
(prot & VM_PROT_WRITE ? 0 : MAP_DISABLE_COREDUMP);
rv = __elfN(map_insert)(imgp, map, object, file_addr,
map_addr, map_addr + map_len, prot, cow);
if (rv != KERN_SUCCESS)
return (EINVAL);
/* we can stop now if we've covered it all */
if (memsz == filsz)
return (0);
}
/*
* We have to get the remaining bit of the file into the first part
* of the oversized map segment. This is normally because the .data
* segment in the file is extended to provide bss. It's a neat idea
* to try and save a page, but it's a pain in the behind to implement.
*/
copy_len = filsz == 0 ? 0 : (offset + filsz) - trunc_page(offset +
filsz);
map_addr = trunc_page((vm_offset_t)vmaddr + filsz);
map_len = round_page((vm_offset_t)vmaddr + memsz) - map_addr;
/* This had damn well better be true! */
if (map_len != 0) {
rv = __elfN(map_insert)(imgp, map, NULL, 0, map_addr,
map_addr + map_len, prot, 0);
if (rv != KERN_SUCCESS)
return (EINVAL);
}
if (copy_len != 0) {
sf = vm_imgact_map_page(object, offset + filsz);
if (sf == NULL)
return (EIO);
/* send the page fragment to user space */
error = copyout((caddr_t)sf_buf_kva(sf), (caddr_t)map_addr,
copy_len);
vm_imgact_unmap_page(sf);
if (error != 0)
return (error);
}
/*
* Remove write access to the page if it was only granted by map_insert
* to allow copyout.
*/
if ((prot & VM_PROT_WRITE) == 0)
vm_map_protect(map, trunc_page(map_addr), round_page(map_addr +
map_len), prot, 0, VM_MAP_PROTECT_SET_PROT);
return (0);
}
static int
__elfN(load_sections)(struct image_params *imgp, const Elf_Ehdr *hdr,
const Elf_Phdr *phdr, u_long rbase, u_long *base_addrp)
{
vm_prot_t prot;
u_long base_addr;
bool first;
int error, i;
ASSERT_VOP_LOCKED(imgp->vp, __func__);
base_addr = 0;
first = true;
for (i = 0; i < hdr->e_phnum; i++) {
if (phdr[i].p_type != PT_LOAD || phdr[i].p_memsz == 0)
continue;
/* Loadable segment */
prot = __elfN(trans_prot)(phdr[i].p_flags);
error = __elfN(load_section)(imgp, phdr[i].p_offset,
(caddr_t)(uintptr_t)phdr[i].p_vaddr + rbase,
phdr[i].p_memsz, phdr[i].p_filesz, prot);
if (error != 0)
return (error);
/*
* Establish the base address if this is the first segment.
*/
if (first) {
base_addr = trunc_page(phdr[i].p_vaddr + rbase);
first = false;
}
}
if (base_addrp != NULL)
*base_addrp = base_addr;
return (0);
}
/*
* Load the file "file" into memory. It may be either a shared object
* or an executable.
*
* The "addr" reference parameter is in/out. On entry, it specifies
* the address where a shared object should be loaded. If the file is
* an executable, this value is ignored. On exit, "addr" specifies
* where the file was actually loaded.
*
* The "entry" reference parameter is out only. On exit, it specifies
* the entry point for the loaded file.
*/
static int
__elfN(load_file)(struct proc *p, const char *file, u_long *addr,
u_long *entry)
{
struct {
struct nameidata nd;
struct vattr attr;
struct image_params image_params;
} *tempdata;
const Elf_Ehdr *hdr = NULL;
const Elf_Phdr *phdr = NULL;
struct nameidata *nd;
struct vattr *attr;
struct image_params *imgp;
u_long rbase;
u_long base_addr = 0;
int error;
#ifdef CAPABILITY_MODE
/*
* XXXJA: This check can go away once we are sufficiently confident
* that the checks in namei() are correct.
*/
if (IN_CAPABILITY_MODE(curthread))
return (ECAPMODE);
#endif
tempdata = malloc(sizeof(*tempdata), M_TEMP, M_WAITOK | M_ZERO);
nd = &tempdata->nd;
attr = &tempdata->attr;
imgp = &tempdata->image_params;
/*
* Initialize part of the common data
*/
imgp->proc = p;
imgp->attr = attr;
NDINIT(nd, LOOKUP, ISOPEN | FOLLOW | LOCKSHARED | LOCKLEAF,
UIO_SYSSPACE, file, curthread);
if ((error = namei(nd)) != 0) {
nd->ni_vp = NULL;
goto fail;
}
NDFREE(nd, NDF_ONLY_PNBUF);
imgp->vp = nd->ni_vp;
/*
* Check permissions, modes, uid, etc on the file, and "open" it.
*/
error = exec_check_permissions(imgp);
if (error)
goto fail;
error = exec_map_first_page(imgp);
if (error)
goto fail;
imgp->object = nd->ni_vp->v_object;
hdr = (const Elf_Ehdr *)imgp->image_header;
if ((error = __elfN(check_header)(hdr)) != 0)
goto fail;
if (hdr->e_type == ET_DYN)
rbase = *addr;
else if (hdr->e_type == ET_EXEC)
rbase = 0;
else {
error = ENOEXEC;
goto fail;
}
/* Only support headers that fit within first page for now */
if (!__elfN(phdr_in_zero_page)(hdr)) {
error = ENOEXEC;
goto fail;
}
phdr = (const Elf_Phdr *)(imgp->image_header + hdr->e_phoff);
if (!aligned(phdr, Elf_Addr)) {
error = ENOEXEC;
goto fail;
}
error = __elfN(load_sections)(imgp, hdr, phdr, rbase, &base_addr);
if (error != 0)
goto fail;
*addr = base_addr;
*entry = (unsigned long)hdr->e_entry + rbase;
fail:
if (imgp->firstpage)
exec_unmap_first_page(imgp);
if (nd->ni_vp) {
if (imgp->textset)
VOP_UNSET_TEXT_CHECKED(nd->ni_vp);
vput(nd->ni_vp);
}
free(tempdata, M_TEMP);
return (error);
}
static u_long
__CONCAT(rnd_, __elfN(base))(vm_map_t map __unused, u_long minv, u_long maxv,
u_int align)
{
u_long rbase, res;
MPASS(vm_map_min(map) <= minv);
MPASS(maxv <= vm_map_max(map));
MPASS(minv < maxv);
MPASS(minv + align < maxv);
arc4rand(&rbase, sizeof(rbase), 0);
res = roundup(minv, (u_long)align) + rbase % (maxv - minv);
res &= ~((u_long)align - 1);
if (res >= maxv)
res -= align;
KASSERT(res >= minv,
("res %#lx < minv %#lx, maxv %#lx rbase %#lx",
res, minv, maxv, rbase));
KASSERT(res < maxv,
("res %#lx > maxv %#lx, minv %#lx rbase %#lx",
res, maxv, minv, rbase));
return (res);
}
static int
__elfN(enforce_limits)(struct image_params *imgp, const Elf_Ehdr *hdr,
const Elf_Phdr *phdr, u_long et_dyn_addr)
{
struct vmspace *vmspace;
const char *err_str;
u_long text_size, data_size, total_size, text_addr, data_addr;
u_long seg_size, seg_addr;
int i;
err_str = NULL;
text_size = data_size = total_size = text_addr = data_addr = 0;
for (i = 0; i < hdr->e_phnum; i++) {
if (phdr[i].p_type != PT_LOAD || phdr[i].p_memsz == 0)
continue;
seg_addr = trunc_page(phdr[i].p_vaddr + et_dyn_addr);
seg_size = round_page(phdr[i].p_memsz +
phdr[i].p_vaddr + et_dyn_addr - seg_addr);
/*
* Make the largest executable segment the official
* text segment and all others data.
*
* Note that obreak() assumes that data_addr + data_size == end
* of data load area, and the ELF file format expects segments
* to be sorted by address. If multiple data segments exist,
* the last one will be used.
*/
if ((phdr[i].p_flags & PF_X) != 0 && text_size < seg_size) {
text_size = seg_size;
text_addr = seg_addr;
} else {
data_size = seg_size;
data_addr = seg_addr;
}
total_size += seg_size;
}
if (data_addr == 0 && data_size == 0) {
data_addr = text_addr;
data_size = text_size;
}
/*
* Check limits. It should be safe to check the
* limits after loading the segments since we do
* not actually fault in all the segments pages.
*/
PROC_LOCK(imgp->proc);
if (data_size > lim_cur_proc(imgp->proc, RLIMIT_DATA))
err_str = "Data segment size exceeds process limit";
else if (text_size > maxtsiz)
err_str = "Text segment size exceeds system limit";
else if (total_size > lim_cur_proc(imgp->proc, RLIMIT_VMEM))
err_str = "Total segment size exceeds process limit";
else if (racct_set(imgp->proc, RACCT_DATA, data_size) != 0)
err_str = "Data segment size exceeds resource limit";
else if (racct_set(imgp->proc, RACCT_VMEM, total_size) != 0)
err_str = "Total segment size exceeds resource limit";
PROC_UNLOCK(imgp->proc);
if (err_str != NULL) {
uprintf("%s\n", err_str);
return (ENOMEM);
}
vmspace = imgp->proc->p_vmspace;
vmspace->vm_tsize = text_size >> PAGE_SHIFT;
vmspace->vm_taddr = (caddr_t)(uintptr_t)text_addr;
vmspace->vm_dsize = data_size >> PAGE_SHIFT;
vmspace->vm_daddr = (caddr_t)(uintptr_t)data_addr;
return (0);
}
static int
__elfN(get_interp)(struct image_params *imgp, const Elf_Phdr *phdr,
char **interpp, bool *free_interpp)
{
struct thread *td;
char *interp;
int error, interp_name_len;
KASSERT(phdr->p_type == PT_INTERP,
("%s: p_type %u != PT_INTERP", __func__, phdr->p_type));
ASSERT_VOP_LOCKED(imgp->vp, __func__);
td = curthread;
/* Path to interpreter */
if (phdr->p_filesz < 2 || phdr->p_filesz > MAXPATHLEN) {
uprintf("Invalid PT_INTERP\n");
return (ENOEXEC);
}
interp_name_len = phdr->p_filesz;
if (phdr->p_offset > PAGE_SIZE ||
interp_name_len > PAGE_SIZE - phdr->p_offset) {
/*
* The vnode lock might be needed by the pagedaemon to
* clean pages owned by the vnode. Do not allow sleep
* waiting for memory with the vnode locked, instead
* try non-sleepable allocation first, and if it
* fails, go to the slow path were we drop the lock
* and do M_WAITOK. A text reference prevents
* modifications to the vnode content.
*/
interp = malloc(interp_name_len + 1, M_TEMP, M_NOWAIT);
if (interp == NULL) {
VOP_UNLOCK(imgp->vp);
interp = malloc(interp_name_len + 1, M_TEMP, M_WAITOK);
vn_lock(imgp->vp, LK_SHARED | LK_RETRY);
}
error = vn_rdwr(UIO_READ, imgp->vp, interp,
interp_name_len, phdr->p_offset,
UIO_SYSSPACE, IO_NODELOCKED, td->td_ucred,
NOCRED, NULL, td);
if (error != 0) {
free(interp, M_TEMP);
uprintf("i/o error PT_INTERP %d\n", error);
return (error);
}
interp[interp_name_len] = '\0';
*interpp = interp;
*free_interpp = true;
return (0);
}
interp = __DECONST(char *, imgp->image_header) + phdr->p_offset;
if (interp[interp_name_len - 1] != '\0') {
uprintf("Invalid PT_INTERP\n");
return (ENOEXEC);
}
*interpp = interp;
*free_interpp = false;
return (0);
}
static int
__elfN(load_interp)(struct image_params *imgp, const Elf_Brandinfo *brand_info,
const char *interp, u_long *addr, u_long *entry)
{
char *path;
int error;
if (brand_info->emul_path != NULL &&
brand_info->emul_path[0] != '\0') {
path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
snprintf(path, MAXPATHLEN, "%s%s",
brand_info->emul_path, interp);
error = __elfN(load_file)(imgp->proc, path, addr, entry);
free(path, M_TEMP);
if (error == 0)
return (0);
}
if (brand_info->interp_newpath != NULL &&
(brand_info->interp_path == NULL ||
strcmp(interp, brand_info->interp_path) == 0)) {
error = __elfN(load_file)(imgp->proc,
brand_info->interp_newpath, addr, entry);
if (error == 0)
return (0);
}
error = __elfN(load_file)(imgp->proc, interp, addr, entry);
if (error == 0)
return (0);
uprintf("ELF interpreter %s not found, error %d\n", interp, error);
return (error);
}
/*
* Impossible et_dyn_addr initial value indicating that the real base
* must be calculated later with some randomization applied.
*/
#define ET_DYN_ADDR_RAND 1
static int
__CONCAT(exec_, __elfN(imgact))(struct image_params *imgp)
{
struct thread *td;
const Elf_Ehdr *hdr;
const Elf_Phdr *phdr;
Elf_Auxargs *elf_auxargs;
struct vmspace *vmspace;
vm_map_t map;
char *interp;
Elf_Brandinfo *brand_info;
struct sysentvec *sv;
u_long addr, baddr, et_dyn_addr, entry, proghdr;
u_long maxalign, mapsz, maxv, maxv1;
uint32_t fctl0;
int32_t osrel;
bool free_interp;
int error, i, n;
hdr = (const Elf_Ehdr *)imgp->image_header;
/*
* Do we have a valid ELF header ?
*
* Only allow ET_EXEC & ET_DYN here, reject ET_DYN later
* if particular brand doesn't support it.
*/
if (__elfN(check_header)(hdr) != 0 ||
(hdr->e_type != ET_EXEC && hdr->e_type != ET_DYN))
return (-1);
/*
* From here on down, we return an errno, not -1, as we've
* detected an ELF file.
*/
if (!__elfN(phdr_in_zero_page)(hdr)) {
uprintf("Program headers not in the first page\n");
return (ENOEXEC);
}
phdr = (const Elf_Phdr *)(imgp->image_header + hdr->e_phoff);
if (!aligned(phdr, Elf_Addr)) {
uprintf("Unaligned program headers\n");
return (ENOEXEC);
}
n = error = 0;
baddr = 0;
osrel = 0;
fctl0 = 0;
entry = proghdr = 0;
interp = NULL;
free_interp = false;
td = curthread;
maxalign = PAGE_SIZE;
mapsz = 0;
for (i = 0; i < hdr->e_phnum; i++) {
switch (phdr[i].p_type) {
case PT_LOAD:
if (n == 0)
baddr = phdr[i].p_vaddr;
if (phdr[i].p_align > maxalign)
maxalign = phdr[i].p_align;
mapsz += phdr[i].p_memsz;
n++;
/*
* If this segment contains the program headers,
* remember their virtual address for the AT_PHDR
* aux entry. Static binaries don't usually include
* a PT_PHDR entry.
*/
if (phdr[i].p_offset == 0 &&
hdr->e_phoff + hdr->e_phnum * hdr->e_phentsize
<= phdr[i].p_filesz)
proghdr = phdr[i].p_vaddr + hdr->e_phoff;
break;
case PT_INTERP:
/* Path to interpreter */
if (interp != NULL) {
uprintf("Multiple PT_INTERP headers\n");
error = ENOEXEC;
goto ret;
}
error = __elfN(get_interp)(imgp, &phdr[i], &interp,
&free_interp);
if (error != 0)
goto ret;
break;
case PT_GNU_STACK:
if (__elfN(nxstack))
imgp->stack_prot =
__elfN(trans_prot)(phdr[i].p_flags);
imgp->stack_sz = phdr[i].p_memsz;
break;
case PT_PHDR: /* Program header table info */
proghdr = phdr[i].p_vaddr;
break;
}
}
brand_info = __elfN(get_brandinfo)(imgp, interp, &osrel, &fctl0);
if (brand_info == NULL) {
uprintf("ELF binary type \"%u\" not known.\n",
hdr->e_ident[EI_OSABI]);
error = ENOEXEC;
goto ret;
}
sv = brand_info->sysvec;
et_dyn_addr = 0;
if (hdr->e_type == ET_DYN) {
if ((brand_info->flags & BI_CAN_EXEC_DYN) == 0) {
uprintf("Cannot execute shared object\n");
error = ENOEXEC;
goto ret;
}
/*
* Honour the base load address from the dso if it is
* non-zero for some reason.
*/
if (baddr == 0) {
if ((sv->sv_flags & SV_ASLR) == 0 ||
(fctl0 & NT_FREEBSD_FCTL_ASLR_DISABLE) != 0)
et_dyn_addr = __elfN(pie_base);
else if ((__elfN(pie_aslr_enabled) &&
(imgp->proc->p_flag2 & P2_ASLR_DISABLE) == 0) ||
(imgp->proc->p_flag2 & P2_ASLR_ENABLE) != 0)
et_dyn_addr = ET_DYN_ADDR_RAND;
else
et_dyn_addr = __elfN(pie_base);
}
}
/*
* Avoid a possible deadlock if the current address space is destroyed
* and that address space maps the locked vnode. In the common case,
* the locked vnode's v_usecount is decremented but remains greater
* than zero. Consequently, the vnode lock is not needed by vrele().
* However, in cases where the vnode lock is external, such as nullfs,
* v_usecount may become zero.
*
* The VV_TEXT flag prevents modifications to the executable while
* the vnode is unlocked.
*/
VOP_UNLOCK(imgp->vp);
/*
* Decide whether to enable randomization of user mappings.
* First, reset user preferences for the setid binaries.
* Then, account for the support of the randomization by the
* ABI, by user preferences, and make special treatment for
* PIE binaries.
*/
if (imgp->credential_setid) {
PROC_LOCK(imgp->proc);
imgp->proc->p_flag2 &= ~(P2_ASLR_ENABLE | P2_ASLR_DISABLE);
PROC_UNLOCK(imgp->proc);
}
if ((sv->sv_flags & SV_ASLR) == 0 ||
(imgp->proc->p_flag2 & P2_ASLR_DISABLE) != 0 ||
(fctl0 & NT_FREEBSD_FCTL_ASLR_DISABLE) != 0) {
KASSERT(et_dyn_addr != ET_DYN_ADDR_RAND,
("et_dyn_addr == RAND and !ASLR"));
} else if ((imgp->proc->p_flag2 & P2_ASLR_ENABLE) != 0 ||
(__elfN(aslr_enabled) && hdr->e_type == ET_EXEC) ||
et_dyn_addr == ET_DYN_ADDR_RAND) {
imgp->map_flags |= MAP_ASLR;
/*
* If user does not care about sbrk, utilize the bss
* grow region for mappings as well. We can select
* the base for the image anywere and still not suffer
* from the fragmentation.
*/
if (!__elfN(aslr_honor_sbrk) ||
(imgp->proc->p_flag2 & P2_ASLR_IGNSTART) != 0)
imgp->map_flags |= MAP_ASLR_IGNSTART;
}
if (!__elfN(allow_wx) && (fctl0 & NT_FREEBSD_FCTL_WXNEEDED) == 0)
imgp->map_flags |= MAP_WXORX;
error = exec_new_vmspace(imgp, sv);
vmspace = imgp->proc->p_vmspace;
map = &vmspace->vm_map;
imgp->proc->p_sysent = sv;
maxv = vm_map_max(map) - lim_max(td, RLIMIT_STACK);
if (et_dyn_addr == ET_DYN_ADDR_RAND) {
KASSERT((map->flags & MAP_ASLR) != 0,
("ET_DYN_ADDR_RAND but !MAP_ASLR"));
et_dyn_addr = __CONCAT(rnd_, __elfN(base))(map,
vm_map_min(map) + mapsz + lim_max(td, RLIMIT_DATA),
/* reserve half of the address space to interpreter */
maxv / 2, 1UL << flsl(maxalign));
}
vn_lock(imgp->vp, LK_SHARED | LK_RETRY);
if (error != 0)
goto ret;
error = __elfN(load_sections)(imgp, hdr, phdr, et_dyn_addr, NULL);
if (error != 0)
goto ret;
error = __elfN(enforce_limits)(imgp, hdr, phdr, et_dyn_addr);
if (error != 0)
goto ret;
entry = (u_long)hdr->e_entry + et_dyn_addr;
/*
* We load the dynamic linker where a userland call
* to mmap(0, ...) would put it. The rationale behind this
* calculation is that it leaves room for the heap to grow to
* its maximum allowed size.
*/
addr = round_page((vm_offset_t)vmspace->vm_daddr + lim_max(td,
RLIMIT_DATA));
if ((map->flags & MAP_ASLR) != 0) {
maxv1 = maxv / 2 + addr / 2;
MPASS(maxv1 >= addr); /* No overflow */
map->anon_loc = __CONCAT(rnd_, __elfN(base))(map, addr, maxv1,
(MAXPAGESIZES > 1 && pagesizes[1] != 0) ?
pagesizes[1] : pagesizes[0]);
} else {
map->anon_loc = addr;
}
imgp->entry_addr = entry;
if (interp != NULL) {
VOP_UNLOCK(imgp->vp);
if ((map->flags & MAP_ASLR) != 0) {
/* Assume that interpreter fits into 1/4 of AS */
maxv1 = maxv / 2 + addr / 2;
MPASS(maxv1 >= addr); /* No overflow */
addr = __CONCAT(rnd_, __elfN(base))(map, addr,
maxv1, PAGE_SIZE);
}
error = __elfN(load_interp)(imgp, brand_info, interp, &addr,
&imgp->entry_addr);
vn_lock(imgp->vp, LK_SHARED | LK_RETRY);
if (error != 0)
goto ret;
} else
addr = et_dyn_addr;
/*
* Construct auxargs table (used by the copyout_auxargs routine)
*/
elf_auxargs = malloc(sizeof(Elf_Auxargs), M_TEMP, M_NOWAIT);
if (elf_auxargs == NULL) {
VOP_UNLOCK(imgp->vp);
elf_auxargs = malloc(sizeof(Elf_Auxargs), M_TEMP, M_WAITOK);
vn_lock(imgp->vp, LK_SHARED | LK_RETRY);
}
elf_auxargs->execfd = -1;
elf_auxargs->phdr = proghdr + et_dyn_addr;
elf_auxargs->phent = hdr->e_phentsize;
elf_auxargs->phnum = hdr->e_phnum;
elf_auxargs->pagesz = PAGE_SIZE;
elf_auxargs->base = addr;
elf_auxargs->flags = 0;
elf_auxargs->entry = entry;
elf_auxargs->hdr_eflags = hdr->e_flags;
imgp->auxargs = elf_auxargs;
imgp->interpreted = 0;
imgp->reloc_base = addr;
imgp->proc->p_osrel = osrel;
imgp->proc->p_fctl0 = fctl0;
imgp->proc->p_elf_machine = hdr->e_machine;
imgp->proc->p_elf_flags = hdr->e_flags;
ret:
if (free_interp)
free(interp, M_TEMP);
return (error);
}
#define suword __CONCAT(suword, __ELF_WORD_SIZE)
int
__elfN(freebsd_copyout_auxargs)(struct image_params *imgp, uintptr_t base)
{
Elf_Auxargs *args = (Elf_Auxargs *)imgp->auxargs;
Elf_Auxinfo *argarray, *pos;
int error;
argarray = pos = malloc(AT_COUNT * sizeof(*pos), M_TEMP,
M_WAITOK | M_ZERO);
if (args->execfd != -1)
AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd);
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, AT_EHDRFLAGS, args->hdr_eflags);
if (imgp->execpathp != 0)
AUXARGS_ENTRY_PTR(pos, AT_EXECPATH, imgp->execpathp);
AUXARGS_ENTRY(pos, AT_OSRELDATE,
imgp->proc->p_ucred->cr_prison->pr_osreldate);
if (imgp->canary != 0) {
AUXARGS_ENTRY_PTR(pos, AT_CANARY, imgp->canary);
AUXARGS_ENTRY(pos, AT_CANARYLEN, imgp->canarylen);
}
AUXARGS_ENTRY(pos, AT_NCPUS, mp_ncpus);
if (imgp->pagesizes != 0) {
AUXARGS_ENTRY_PTR(pos, AT_PAGESIZES, imgp->pagesizes);
AUXARGS_ENTRY(pos, AT_PAGESIZESLEN, imgp->pagesizeslen);
}
if (imgp->sysent->sv_timekeep_base != 0) {
AUXARGS_ENTRY(pos, AT_TIMEKEEP,
imgp->sysent->sv_timekeep_base);
}
AUXARGS_ENTRY(pos, AT_STACKPROT, imgp->sysent->sv_shared_page_obj
!= NULL && imgp->stack_prot != 0 ? imgp->stack_prot :
imgp->sysent->sv_stackprot);
if (imgp->sysent->sv_hwcap != NULL)
AUXARGS_ENTRY(pos, AT_HWCAP, *imgp->sysent->sv_hwcap);
if (imgp->sysent->sv_hwcap2 != NULL)
AUXARGS_ENTRY(pos, AT_HWCAP2, *imgp->sysent->sv_hwcap2);
AUXARGS_ENTRY(pos, AT_BSDFLAGS, __elfN(sigfastblock) ?
ELF_BSDF_SIGFASTBLK : 0);
AUXARGS_ENTRY(pos, AT_ARGC, imgp->args->argc);
AUXARGS_ENTRY_PTR(pos, AT_ARGV, imgp->argv);
AUXARGS_ENTRY(pos, AT_ENVC, imgp->args->envc);
AUXARGS_ENTRY_PTR(pos, AT_ENVV, imgp->envv);
AUXARGS_ENTRY_PTR(pos, AT_PS_STRINGS, imgp->ps_strings);
if (imgp->sysent->sv_fxrng_gen_base != 0)
AUXARGS_ENTRY(pos, AT_FXRNG, imgp->sysent->sv_fxrng_gen_base);
AUXARGS_ENTRY(pos, AT_NULL, 0);
free(imgp->auxargs, M_TEMP);
imgp->auxargs = NULL;
KASSERT(pos - argarray <= AT_COUNT, ("Too many auxargs"));
error = copyout(argarray, (void *)base, sizeof(*argarray) * AT_COUNT);
free(argarray, M_TEMP);
return (error);
}
int
__elfN(freebsd_fixup)(uintptr_t *stack_base, struct image_params *imgp)
{
Elf_Addr *base;
base = (Elf_Addr *)*stack_base;
base--;
if (suword(base, imgp->args->argc) == -1)
return (EFAULT);
*stack_base = (uintptr_t)base;
return (0);
}
/*
* Code for generating ELF core dumps.
*/
typedef void (*segment_callback)(vm_map_entry_t, void *);
/* Closure for cb_put_phdr(). */
struct phdr_closure {
Elf_Phdr *phdr; /* Program header to fill in */
Elf_Off offset; /* Offset of segment in core file */
};
-typedef void (*outfunc_t)(void *, struct sbuf *, size_t *);
-
struct note_info {
int type; /* Note type. */
outfunc_t outfunc; /* Output function. */
void *outarg; /* Argument for the output function. */
size_t outsize; /* Output size. */
TAILQ_ENTRY(note_info) link; /* Link to the next note info. */
};
TAILQ_HEAD(note_info_list, note_info);
extern int compress_user_cores;
extern int compress_user_cores_level;
static void cb_put_phdr(vm_map_entry_t, void *);
static void cb_size_segment(vm_map_entry_t, void *);
static void each_dumpable_segment(struct thread *, segment_callback, void *,
int);
static int __elfN(corehdr)(struct coredump_params *, int, void *, size_t,
struct note_info_list *, size_t, int);
-static void __elfN(prepare_notes)(struct thread *, struct note_info_list *,
- size_t *);
-static void __elfN(putnote)(struct note_info *, struct sbuf *);
-static size_t register_note(struct note_info_list *, int, outfunc_t, void *);
+static void __elfN(putnote)(struct thread *td, struct note_info *, struct sbuf *);
static void __elfN(note_fpregset)(void *, struct sbuf *, size_t *);
static void __elfN(note_prpsinfo)(void *, struct sbuf *, size_t *);
static void __elfN(note_prstatus)(void *, struct sbuf *, size_t *);
static void __elfN(note_threadmd)(void *, struct sbuf *, size_t *);
static void __elfN(note_thrmisc)(void *, struct sbuf *, size_t *);
static void __elfN(note_ptlwpinfo)(void *, struct sbuf *, size_t *);
static void __elfN(note_procstat_auxv)(void *, struct sbuf *, size_t *);
static void __elfN(note_procstat_proc)(void *, struct sbuf *, size_t *);
static void __elfN(note_procstat_psstrings)(void *, struct sbuf *, size_t *);
static void note_procstat_files(void *, struct sbuf *, size_t *);
static void note_procstat_groups(void *, struct sbuf *, size_t *);
static void note_procstat_osrel(void *, struct sbuf *, size_t *);
static void note_procstat_rlimit(void *, struct sbuf *, size_t *);
static void note_procstat_umask(void *, struct sbuf *, size_t *);
static void note_procstat_vmmap(void *, struct sbuf *, size_t *);
static int
core_compressed_write(void *base, size_t len, off_t offset, void *arg)
{
return (core_write((struct coredump_params *)arg, base, len, offset,
UIO_SYSSPACE, NULL));
}
int
__elfN(coredump)(struct thread *td, struct vnode *vp, off_t limit, int flags)
{
struct ucred *cred = td->td_ucred;
int compm, error = 0;
struct sseg_closure seginfo;
struct note_info_list notelst;
struct coredump_params params;
struct note_info *ninfo;
void *hdr, *tmpbuf;
size_t hdrsize, notesz, coresize;
hdr = NULL;
tmpbuf = NULL;
TAILQ_INIT(¬elst);
/* Size the program segments. */
__elfN(size_segments)(td, &seginfo, flags);
/*
* Collect info about the core file header area.
*/
hdrsize = sizeof(Elf_Ehdr) + sizeof(Elf_Phdr) * (1 + seginfo.count);
if (seginfo.count + 1 >= PN_XNUM)
hdrsize += sizeof(Elf_Shdr);
- __elfN(prepare_notes)(td, ¬elst, ¬esz);
+ td->td_proc->p_sysent->sv_elf_core_prepare_notes(td, ¬elst, ¬esz);
coresize = round_page(hdrsize + notesz) + seginfo.size;
/* Set up core dump parameters. */
params.offset = 0;
params.active_cred = cred;
params.file_cred = NOCRED;
params.td = td;
params.vp = vp;
params.comp = NULL;
#ifdef RACCT
if (racct_enable) {
PROC_LOCK(td->td_proc);
error = racct_add(td->td_proc, RACCT_CORE, coresize);
PROC_UNLOCK(td->td_proc);
if (error != 0) {
error = EFAULT;
goto done;
}
}
#endif
if (coresize >= limit) {
error = EFAULT;
goto done;
}
/* Create a compression stream if necessary. */
compm = compress_user_cores;
if ((flags & (SVC_PT_COREDUMP | SVC_NOCOMPRESS)) == SVC_PT_COREDUMP &&
compm == 0)
compm = COMPRESS_GZIP;
if (compm != 0) {
params.comp = compressor_init(core_compressed_write,
compm, CORE_BUF_SIZE,
compress_user_cores_level, ¶ms);
if (params.comp == NULL) {
error = EFAULT;
goto done;
}
tmpbuf = malloc(CORE_BUF_SIZE, M_TEMP, M_WAITOK | M_ZERO);
}
/*
* Allocate memory for building the header, fill it up,
* and write it out following the notes.
*/
hdr = malloc(hdrsize, M_TEMP, M_WAITOK);
error = __elfN(corehdr)(¶ms, seginfo.count, hdr, hdrsize, ¬elst,
notesz, flags);
/* Write the contents of all of the writable segments. */
if (error == 0) {
Elf_Phdr *php;
off_t offset;
int i;
php = (Elf_Phdr *)((char *)hdr + sizeof(Elf_Ehdr)) + 1;
offset = round_page(hdrsize + notesz);
for (i = 0; i < seginfo.count; i++) {
error = core_output((char *)(uintptr_t)php->p_vaddr,
php->p_filesz, offset, ¶ms, tmpbuf);
if (error != 0)
break;
offset += php->p_filesz;
php++;
}
if (error == 0 && params.comp != NULL)
error = compressor_flush(params.comp);
}
if (error) {
log(LOG_WARNING,
"Failed to write core file for process %s (error %d)\n",
curproc->p_comm, error);
}
done:
free(tmpbuf, M_TEMP);
if (params.comp != NULL)
compressor_fini(params.comp);
while ((ninfo = TAILQ_FIRST(¬elst)) != NULL) {
TAILQ_REMOVE(¬elst, ninfo, link);
free(ninfo, M_TEMP);
}
if (hdr != NULL)
free(hdr, M_TEMP);
return (error);
}
/*
* A callback for each_dumpable_segment() to write out the segment's
* program header entry.
*/
static void
cb_put_phdr(vm_map_entry_t entry, void *closure)
{
struct phdr_closure *phc = (struct phdr_closure *)closure;
Elf_Phdr *phdr = phc->phdr;
phc->offset = round_page(phc->offset);
phdr->p_type = PT_LOAD;
phdr->p_offset = phc->offset;
phdr->p_vaddr = entry->start;
phdr->p_paddr = 0;
phdr->p_filesz = phdr->p_memsz = entry->end - entry->start;
phdr->p_align = PAGE_SIZE;
phdr->p_flags = __elfN(untrans_prot)(entry->protection);
phc->offset += phdr->p_filesz;
phc->phdr++;
}
/*
* A callback for each_dumpable_segment() to gather information about
* the number of segments and their total size.
*/
static void
cb_size_segment(vm_map_entry_t entry, void *closure)
{
struct sseg_closure *ssc = (struct sseg_closure *)closure;
ssc->count++;
ssc->size += entry->end - entry->start;
}
void
__elfN(size_segments)(struct thread *td, struct sseg_closure *seginfo,
int flags)
{
seginfo->count = 0;
seginfo->size = 0;
each_dumpable_segment(td, cb_size_segment, seginfo, flags);
}
/*
* For each writable segment in the process's memory map, call the given
* function with a pointer to the map entry and some arbitrary
* caller-supplied data.
*/
static void
each_dumpable_segment(struct thread *td, segment_callback func, void *closure,
int flags)
{
struct proc *p = td->td_proc;
vm_map_t map = &p->p_vmspace->vm_map;
vm_map_entry_t entry;
vm_object_t backing_object, object;
bool ignore_entry;
vm_map_lock_read(map);
VM_MAP_ENTRY_FOREACH(entry, map) {
/*
* Don't dump inaccessible mappings, deal with legacy
* coredump mode.
*
* Note that read-only segments related to the elf binary
* are marked MAP_ENTRY_NOCOREDUMP now so we no longer
* need to arbitrarily ignore such segments.
*/
if ((flags & SVC_ALL) == 0) {
if (elf_legacy_coredump) {
if ((entry->protection & VM_PROT_RW) !=
VM_PROT_RW)
continue;
} else {
if ((entry->protection & VM_PROT_ALL) == 0)
continue;
}
}
/*
* Dont include memory segment in the coredump if
* MAP_NOCORE is set in mmap(2) or MADV_NOCORE in
* madvise(2). Do not dump submaps (i.e. parts of the
* kernel map).
*/
if ((entry->eflags & MAP_ENTRY_IS_SUB_MAP) != 0)
continue;
if ((entry->eflags & MAP_ENTRY_NOCOREDUMP) != 0 &&
(flags & SVC_ALL) == 0)
continue;
if ((object = entry->object.vm_object) == NULL)
continue;
/* Ignore memory-mapped devices and such things. */
VM_OBJECT_RLOCK(object);
while ((backing_object = object->backing_object) != NULL) {
VM_OBJECT_RLOCK(backing_object);
VM_OBJECT_RUNLOCK(object);
object = backing_object;
}
ignore_entry = (object->flags & OBJ_FICTITIOUS) != 0;
VM_OBJECT_RUNLOCK(object);
if (ignore_entry)
continue;
(*func)(entry, closure);
}
vm_map_unlock_read(map);
}
/*
* Write the core file header to the file, including padding up to
* the page boundary.
*/
static int
__elfN(corehdr)(struct coredump_params *p, int numsegs, void *hdr,
size_t hdrsize, struct note_info_list *notelst, size_t notesz,
int flags)
{
struct note_info *ninfo;
struct sbuf *sb;
int error;
/* Fill in the header. */
bzero(hdr, hdrsize);
__elfN(puthdr)(p->td, hdr, hdrsize, numsegs, notesz, flags);
sb = sbuf_new(NULL, NULL, CORE_BUF_SIZE, SBUF_FIXEDLEN);
sbuf_set_drain(sb, sbuf_drain_core_output, p);
sbuf_start_section(sb, NULL);
sbuf_bcat(sb, hdr, hdrsize);
TAILQ_FOREACH(ninfo, notelst, link)
- __elfN(putnote)(ninfo, sb);
+ __elfN(putnote)(p->td, ninfo, sb);
/* Align up to a page boundary for the program segments. */
sbuf_end_section(sb, -1, PAGE_SIZE, 0);
error = sbuf_finish(sb);
sbuf_delete(sb);
return (error);
}
-static void
+void
__elfN(prepare_notes)(struct thread *td, struct note_info_list *list,
size_t *sizep)
{
struct proc *p;
struct thread *thr;
size_t size;
p = td->td_proc;
size = 0;
- size += register_note(list, NT_PRPSINFO, __elfN(note_prpsinfo), p);
+ size += __elfN(register_note)(td, list, NT_PRPSINFO, __elfN(note_prpsinfo), p);
/*
* To have the debugger select the right thread (LWP) as the initial
* thread, we dump the state of the thread passed to us in td first.
* This is the thread that causes the core dump and thus likely to
* be the right thread one wants to have selected in the debugger.
*/
thr = td;
while (thr != NULL) {
- size += register_note(list, NT_PRSTATUS,
+ size += __elfN(register_note)(td, list, NT_PRSTATUS,
__elfN(note_prstatus), thr);
- size += register_note(list, NT_FPREGSET,
+ size += __elfN(register_note)(td, list, NT_FPREGSET,
__elfN(note_fpregset), thr);
- size += register_note(list, NT_THRMISC,
+ size += __elfN(register_note)(td, list, NT_THRMISC,
__elfN(note_thrmisc), thr);
- size += register_note(list, NT_PTLWPINFO,
+ size += __elfN(register_note)(td, list, NT_PTLWPINFO,
__elfN(note_ptlwpinfo), thr);
- size += register_note(list, -1,
+ size += __elfN(register_note)(td, list, -1,
__elfN(note_threadmd), thr);
thr = thr == td ? TAILQ_FIRST(&p->p_threads) :
TAILQ_NEXT(thr, td_plist);
if (thr == td)
thr = TAILQ_NEXT(thr, td_plist);
}
- size += register_note(list, NT_PROCSTAT_PROC,
+ size += __elfN(register_note)(td, list, NT_PROCSTAT_PROC,
__elfN(note_procstat_proc), p);
- size += register_note(list, NT_PROCSTAT_FILES,
+ size += __elfN(register_note)(td, list, NT_PROCSTAT_FILES,
note_procstat_files, p);
- size += register_note(list, NT_PROCSTAT_VMMAP,
+ size += __elfN(register_note)(td, list, NT_PROCSTAT_VMMAP,
note_procstat_vmmap, p);
- size += register_note(list, NT_PROCSTAT_GROUPS,
+ size += __elfN(register_note)(td, list, NT_PROCSTAT_GROUPS,
note_procstat_groups, p);
- size += register_note(list, NT_PROCSTAT_UMASK,
+ size += __elfN(register_note)(td, list, NT_PROCSTAT_UMASK,
note_procstat_umask, p);
- size += register_note(list, NT_PROCSTAT_RLIMIT,
+ size += __elfN(register_note)(td, list, NT_PROCSTAT_RLIMIT,
note_procstat_rlimit, p);
- size += register_note(list, NT_PROCSTAT_OSREL,
+ size += __elfN(register_note)(td, list, NT_PROCSTAT_OSREL,
note_procstat_osrel, p);
- size += register_note(list, NT_PROCSTAT_PSSTRINGS,
+ size += __elfN(register_note)(td, list, NT_PROCSTAT_PSSTRINGS,
__elfN(note_procstat_psstrings), p);
- size += register_note(list, NT_PROCSTAT_AUXV,
+ size += __elfN(register_note)(td, list, NT_PROCSTAT_AUXV,
__elfN(note_procstat_auxv), p);
*sizep = size;
}
void
__elfN(puthdr)(struct thread *td, void *hdr, size_t hdrsize, int numsegs,
size_t notesz, int flags)
{
Elf_Ehdr *ehdr;
Elf_Phdr *phdr;
Elf_Shdr *shdr;
struct phdr_closure phc;
ehdr = (Elf_Ehdr *)hdr;
ehdr->e_ident[EI_MAG0] = ELFMAG0;
ehdr->e_ident[EI_MAG1] = ELFMAG1;
ehdr->e_ident[EI_MAG2] = ELFMAG2;
ehdr->e_ident[EI_MAG3] = ELFMAG3;
ehdr->e_ident[EI_CLASS] = ELF_CLASS;
ehdr->e_ident[EI_DATA] = ELF_DATA;
ehdr->e_ident[EI_VERSION] = EV_CURRENT;
- ehdr->e_ident[EI_OSABI] = ELFOSABI_FREEBSD;
+ ehdr->e_ident[EI_OSABI] = td->td_proc->p_sysent->sv_elf_core_osabi;
ehdr->e_ident[EI_ABIVERSION] = 0;
ehdr->e_ident[EI_PAD] = 0;
ehdr->e_type = ET_CORE;
ehdr->e_machine = td->td_proc->p_elf_machine;
ehdr->e_version = EV_CURRENT;
ehdr->e_entry = 0;
ehdr->e_phoff = sizeof(Elf_Ehdr);
ehdr->e_flags = td->td_proc->p_elf_flags;
ehdr->e_ehsize = sizeof(Elf_Ehdr);
ehdr->e_phentsize = sizeof(Elf_Phdr);
ehdr->e_shentsize = sizeof(Elf_Shdr);
ehdr->e_shstrndx = SHN_UNDEF;
if (numsegs + 1 < PN_XNUM) {
ehdr->e_phnum = numsegs + 1;
ehdr->e_shnum = 0;
} else {
ehdr->e_phnum = PN_XNUM;
ehdr->e_shnum = 1;
ehdr->e_shoff = ehdr->e_phoff +
(numsegs + 1) * ehdr->e_phentsize;
KASSERT(ehdr->e_shoff == hdrsize - sizeof(Elf_Shdr),
("e_shoff: %zu, hdrsize - shdr: %zu",
(size_t)ehdr->e_shoff, hdrsize - sizeof(Elf_Shdr)));
shdr = (Elf_Shdr *)((char *)hdr + ehdr->e_shoff);
memset(shdr, 0, sizeof(*shdr));
/*
* A special first section is used to hold large segment and
* section counts. This was proposed by Sun Microsystems in
* Solaris and has been adopted by Linux; the standard ELF
* tools are already familiar with the technique.
*
* See table 7-7 of the Solaris "Linker and Libraries Guide"
* (or 12-7 depending on the version of the document) for more
* details.
*/
shdr->sh_type = SHT_NULL;
shdr->sh_size = ehdr->e_shnum;
shdr->sh_link = ehdr->e_shstrndx;
shdr->sh_info = numsegs + 1;
}
/*
* Fill in the program header entries.
*/
phdr = (Elf_Phdr *)((char *)hdr + ehdr->e_phoff);
/* The note segement. */
phdr->p_type = PT_NOTE;
phdr->p_offset = hdrsize;
phdr->p_vaddr = 0;
phdr->p_paddr = 0;
phdr->p_filesz = notesz;
phdr->p_memsz = 0;
phdr->p_flags = PF_R;
phdr->p_align = ELF_NOTE_ROUNDSIZE;
phdr++;
/* All the writable segments from the program. */
phc.phdr = phdr;
phc.offset = round_page(hdrsize + notesz);
each_dumpable_segment(td, cb_put_phdr, &phc, flags);
}
-static size_t
-register_note(struct note_info_list *list, int type, outfunc_t out, void *arg)
+size_t
+__elfN(register_note)(struct thread *td, struct note_info_list *list,
+ int type, outfunc_t out, void *arg)
{
+ const struct sysentvec *sv;
struct note_info *ninfo;
size_t size, notesize;
+ sv = td->td_proc->p_sysent;
size = 0;
out(arg, NULL, &size);
ninfo = malloc(sizeof(*ninfo), M_TEMP, M_ZERO | M_WAITOK);
ninfo->type = type;
ninfo->outfunc = out;
ninfo->outarg = arg;
ninfo->outsize = size;
TAILQ_INSERT_TAIL(list, ninfo, link);
if (type == -1)
return (size);
notesize = sizeof(Elf_Note) + /* note header */
- roundup2(sizeof(FREEBSD_ABI_VENDOR), ELF_NOTE_ROUNDSIZE) +
+ roundup2(strlen(sv->sv_elf_core_abi_vendor) + 1, ELF_NOTE_ROUNDSIZE) +
/* note name */
roundup2(size, ELF_NOTE_ROUNDSIZE); /* note description */
return (notesize);
}
static size_t
append_note_data(const void *src, void *dst, size_t len)
{
size_t padded_len;
padded_len = roundup2(len, ELF_NOTE_ROUNDSIZE);
if (dst != NULL) {
bcopy(src, dst, len);
bzero((char *)dst + len, padded_len - len);
}
return (padded_len);
}
size_t
__elfN(populate_note)(int type, void *src, void *dst, size_t size, void **descp)
{
Elf_Note *note;
char *buf;
size_t notesize;
buf = dst;
if (buf != NULL) {
note = (Elf_Note *)buf;
note->n_namesz = sizeof(FREEBSD_ABI_VENDOR);
note->n_descsz = size;
note->n_type = type;
buf += sizeof(*note);
buf += append_note_data(FREEBSD_ABI_VENDOR, buf,
sizeof(FREEBSD_ABI_VENDOR));
append_note_data(src, buf, size);
if (descp != NULL)
*descp = buf;
}
notesize = sizeof(Elf_Note) + /* note header */
roundup2(sizeof(FREEBSD_ABI_VENDOR), ELF_NOTE_ROUNDSIZE) +
/* note name */
roundup2(size, ELF_NOTE_ROUNDSIZE); /* note description */
return (notesize);
}
static void
-__elfN(putnote)(struct note_info *ninfo, struct sbuf *sb)
+__elfN(putnote)(struct thread *td, struct note_info *ninfo, struct sbuf *sb)
{
Elf_Note note;
+ const struct sysentvec *sv;
ssize_t old_len, sect_len;
size_t new_len, descsz, i;
if (ninfo->type == -1) {
ninfo->outfunc(ninfo->outarg, sb, &ninfo->outsize);
return;
}
- note.n_namesz = sizeof(FREEBSD_ABI_VENDOR);
+ sv = td->td_proc->p_sysent;
+
+ note.n_namesz = strlen(sv->sv_elf_core_abi_vendor) + 1;
note.n_descsz = ninfo->outsize;
note.n_type = ninfo->type;
sbuf_bcat(sb, ¬e, sizeof(note));
sbuf_start_section(sb, &old_len);
- sbuf_bcat(sb, FREEBSD_ABI_VENDOR, sizeof(FREEBSD_ABI_VENDOR));
+ sbuf_bcat(sb, sv->sv_elf_core_abi_vendor,
+ strlen(sv->sv_elf_core_abi_vendor) + 1);
sbuf_end_section(sb, old_len, ELF_NOTE_ROUNDSIZE, 0);
if (note.n_descsz == 0)
return;
sbuf_start_section(sb, &old_len);
ninfo->outfunc(ninfo->outarg, sb, &ninfo->outsize);
sect_len = sbuf_end_section(sb, old_len, ELF_NOTE_ROUNDSIZE, 0);
if (sect_len < 0)
return;
new_len = (size_t)sect_len;
descsz = roundup(note.n_descsz, ELF_NOTE_ROUNDSIZE);
if (new_len < descsz) {
/*
* It is expected that individual note emitters will correctly
* predict their expected output size and fill up to that size
* themselves, padding in a format-specific way if needed.
* However, in case they don't, just do it here with zeros.
*/
for (i = 0; i < descsz - new_len; i++)
sbuf_putc(sb, 0);
} else if (new_len > descsz) {
/*
* We can't always truncate sb -- we may have drained some
* of it already.
*/
KASSERT(new_len == descsz, ("%s: Note type %u changed as we "
"read it (%zu > %zu). Since it is longer than "
"expected, this coredump's notes are corrupt. THIS "
"IS A BUG in the note_procstat routine for type %u.\n",
__func__, (unsigned)note.n_type, new_len, descsz,
(unsigned)note.n_type));
}
}
/*
* Miscellaneous note out functions.
*/
#if defined(COMPAT_FREEBSD32) && __ELF_WORD_SIZE == 32
#include <compat/freebsd32/freebsd32.h>
#include <compat/freebsd32/freebsd32_signal.h>
typedef struct prstatus32 elf_prstatus_t;
typedef struct prpsinfo32 elf_prpsinfo_t;
typedef struct fpreg32 elf_prfpregset_t;
typedef struct fpreg32 elf_fpregset_t;
typedef struct reg32 elf_gregset_t;
typedef struct thrmisc32 elf_thrmisc_t;
#define ELF_KERN_PROC_MASK KERN_PROC_MASK32
typedef struct kinfo_proc32 elf_kinfo_proc_t;
typedef uint32_t elf_ps_strings_t;
#else
typedef prstatus_t elf_prstatus_t;
typedef prpsinfo_t elf_prpsinfo_t;
typedef prfpregset_t elf_prfpregset_t;
typedef prfpregset_t elf_fpregset_t;
typedef gregset_t elf_gregset_t;
typedef thrmisc_t elf_thrmisc_t;
#define ELF_KERN_PROC_MASK 0
typedef struct kinfo_proc elf_kinfo_proc_t;
typedef vm_offset_t elf_ps_strings_t;
#endif
static void
__elfN(note_prpsinfo)(void *arg, struct sbuf *sb, size_t *sizep)
{
struct sbuf sbarg;
size_t len;
char *cp, *end;
struct proc *p;
elf_prpsinfo_t *psinfo;
int error;
p = arg;
if (sb != NULL) {
KASSERT(*sizep == sizeof(*psinfo), ("invalid size"));
psinfo = malloc(sizeof(*psinfo), M_TEMP, M_ZERO | M_WAITOK);
psinfo->pr_version = PRPSINFO_VERSION;
psinfo->pr_psinfosz = sizeof(elf_prpsinfo_t);
strlcpy(psinfo->pr_fname, p->p_comm, sizeof(psinfo->pr_fname));
PROC_LOCK(p);
if (p->p_args != NULL) {
len = sizeof(psinfo->pr_psargs) - 1;
if (len > p->p_args->ar_length)
len = p->p_args->ar_length;
memcpy(psinfo->pr_psargs, p->p_args->ar_args, len);
PROC_UNLOCK(p);
error = 0;
} else {
_PHOLD(p);
PROC_UNLOCK(p);
sbuf_new(&sbarg, psinfo->pr_psargs,
sizeof(psinfo->pr_psargs), SBUF_FIXEDLEN);
error = proc_getargv(curthread, p, &sbarg);
PRELE(p);
if (sbuf_finish(&sbarg) == 0)
len = sbuf_len(&sbarg) - 1;
else
len = sizeof(psinfo->pr_psargs) - 1;
sbuf_delete(&sbarg);
}
if (error || len == 0)
strlcpy(psinfo->pr_psargs, p->p_comm,
sizeof(psinfo->pr_psargs));
else {
KASSERT(len < sizeof(psinfo->pr_psargs),
("len is too long: %zu vs %zu", len,
sizeof(psinfo->pr_psargs)));
cp = psinfo->pr_psargs;
end = cp + len - 1;
for (;;) {
cp = memchr(cp, '\0', end - cp);
if (cp == NULL)
break;
*cp = ' ';
}
}
psinfo->pr_pid = p->p_pid;
sbuf_bcat(sb, psinfo, sizeof(*psinfo));
free(psinfo, M_TEMP);
}
*sizep = sizeof(*psinfo);
}
static void
__elfN(note_prstatus)(void *arg, struct sbuf *sb, size_t *sizep)
{
struct thread *td;
elf_prstatus_t *status;
td = arg;
if (sb != NULL) {
KASSERT(*sizep == sizeof(*status), ("invalid size"));
status = malloc(sizeof(*status), M_TEMP, M_ZERO | M_WAITOK);
status->pr_version = PRSTATUS_VERSION;
status->pr_statussz = sizeof(elf_prstatus_t);
status->pr_gregsetsz = sizeof(elf_gregset_t);
status->pr_fpregsetsz = sizeof(elf_fpregset_t);
status->pr_osreldate = osreldate;
status->pr_cursig = td->td_proc->p_sig;
status->pr_pid = td->td_tid;
#if defined(COMPAT_FREEBSD32) && __ELF_WORD_SIZE == 32
fill_regs32(td, &status->pr_reg);
#else
fill_regs(td, &status->pr_reg);
#endif
sbuf_bcat(sb, status, sizeof(*status));
free(status, M_TEMP);
}
*sizep = sizeof(*status);
}
static void
__elfN(note_fpregset)(void *arg, struct sbuf *sb, size_t *sizep)
{
struct thread *td;
elf_prfpregset_t *fpregset;
td = arg;
if (sb != NULL) {
KASSERT(*sizep == sizeof(*fpregset), ("invalid size"));
fpregset = malloc(sizeof(*fpregset), M_TEMP, M_ZERO | M_WAITOK);
#if defined(COMPAT_FREEBSD32) && __ELF_WORD_SIZE == 32
fill_fpregs32(td, fpregset);
#else
fill_fpregs(td, fpregset);
#endif
sbuf_bcat(sb, fpregset, sizeof(*fpregset));
free(fpregset, M_TEMP);
}
*sizep = sizeof(*fpregset);
}
static void
__elfN(note_thrmisc)(void *arg, struct sbuf *sb, size_t *sizep)
{
struct thread *td;
elf_thrmisc_t thrmisc;
td = arg;
if (sb != NULL) {
KASSERT(*sizep == sizeof(thrmisc), ("invalid size"));
bzero(&thrmisc, sizeof(thrmisc));
strcpy(thrmisc.pr_tname, td->td_name);
sbuf_bcat(sb, &thrmisc, sizeof(thrmisc));
}
*sizep = sizeof(thrmisc);
}
static void
__elfN(note_ptlwpinfo)(void *arg, struct sbuf *sb, size_t *sizep)
{
struct thread *td;
size_t size;
int structsize;
#if defined(COMPAT_FREEBSD32) && __ELF_WORD_SIZE == 32
struct ptrace_lwpinfo32 pl;
#else
struct ptrace_lwpinfo pl;
#endif
td = arg;
size = sizeof(structsize) + sizeof(pl);
if (sb != NULL) {
KASSERT(*sizep == size, ("invalid size"));
structsize = sizeof(pl);
sbuf_bcat(sb, &structsize, sizeof(structsize));
bzero(&pl, sizeof(pl));
pl.pl_lwpid = td->td_tid;
pl.pl_event = PL_EVENT_NONE;
pl.pl_sigmask = td->td_sigmask;
pl.pl_siglist = td->td_siglist;
if (td->td_si.si_signo != 0) {
pl.pl_event = PL_EVENT_SIGNAL;
pl.pl_flags |= PL_FLAG_SI;
#if defined(COMPAT_FREEBSD32) && __ELF_WORD_SIZE == 32
siginfo_to_siginfo32(&td->td_si, &pl.pl_siginfo);
#else
pl.pl_siginfo = td->td_si;
#endif
}
strcpy(pl.pl_tdname, td->td_name);
/* XXX TODO: supply more information in struct ptrace_lwpinfo*/
sbuf_bcat(sb, &pl, sizeof(pl));
}
*sizep = size;
}
/*
* Allow for MD specific notes, as well as any MD
* specific preparations for writing MI notes.
*/
static void
__elfN(note_threadmd)(void *arg, struct sbuf *sb, size_t *sizep)
{
struct thread *td;
void *buf;
size_t size;
td = (struct thread *)arg;
size = *sizep;
if (size != 0 && sb != NULL)
buf = malloc(size, M_TEMP, M_ZERO | M_WAITOK);
else
buf = NULL;
size = 0;
__elfN(dump_thread)(td, buf, &size);
KASSERT(sb == NULL || *sizep == size, ("invalid size"));
if (size != 0 && sb != NULL)
sbuf_bcat(sb, buf, size);
free(buf, M_TEMP);
*sizep = size;
}
#ifdef KINFO_PROC_SIZE
CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
#endif
static void
__elfN(note_procstat_proc)(void *arg, struct sbuf *sb, size_t *sizep)
{
struct proc *p;
size_t size;
int structsize;
p = arg;
size = sizeof(structsize) + p->p_numthreads *
sizeof(elf_kinfo_proc_t);
if (sb != NULL) {
KASSERT(*sizep == size, ("invalid size"));
structsize = sizeof(elf_kinfo_proc_t);
sbuf_bcat(sb, &structsize, sizeof(structsize));
sx_slock(&proctree_lock);
PROC_LOCK(p);
kern_proc_out(p, sb, ELF_KERN_PROC_MASK);
sx_sunlock(&proctree_lock);
}
*sizep = size;
}
#ifdef KINFO_FILE_SIZE
CTASSERT(sizeof(struct kinfo_file) == KINFO_FILE_SIZE);
#endif
static void
note_procstat_files(void *arg, struct sbuf *sb, size_t *sizep)
{
struct proc *p;
size_t size, sect_sz, i;
ssize_t start_len, sect_len;
int structsize, filedesc_flags;
if (coredump_pack_fileinfo)
filedesc_flags = KERN_FILEDESC_PACK_KINFO;
else
filedesc_flags = 0;
p = arg;
structsize = sizeof(struct kinfo_file);
if (sb == NULL) {
size = 0;
sb = sbuf_new(NULL, NULL, 128, SBUF_FIXEDLEN);
sbuf_set_drain(sb, sbuf_count_drain, &size);
sbuf_bcat(sb, &structsize, sizeof(structsize));
PROC_LOCK(p);
kern_proc_filedesc_out(p, sb, -1, filedesc_flags);
sbuf_finish(sb);
sbuf_delete(sb);
*sizep = size;
} else {
sbuf_start_section(sb, &start_len);
sbuf_bcat(sb, &structsize, sizeof(structsize));
PROC_LOCK(p);
kern_proc_filedesc_out(p, sb, *sizep - sizeof(structsize),
filedesc_flags);
sect_len = sbuf_end_section(sb, start_len, 0, 0);
if (sect_len < 0)
return;
sect_sz = sect_len;
KASSERT(sect_sz <= *sizep,
("kern_proc_filedesc_out did not respect maxlen; "
"requested %zu, got %zu", *sizep - sizeof(structsize),
sect_sz - sizeof(structsize)));
for (i = 0; i < *sizep - sect_sz && sb->s_error == 0; i++)
sbuf_putc(sb, 0);
}
}
#ifdef KINFO_VMENTRY_SIZE
CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
#endif
static void
note_procstat_vmmap(void *arg, struct sbuf *sb, size_t *sizep)
{
struct proc *p;
size_t size;
int structsize, vmmap_flags;
if (coredump_pack_vmmapinfo)
vmmap_flags = KERN_VMMAP_PACK_KINFO;
else
vmmap_flags = 0;
p = arg;
structsize = sizeof(struct kinfo_vmentry);
if (sb == NULL) {
size = 0;
sb = sbuf_new(NULL, NULL, 128, SBUF_FIXEDLEN);
sbuf_set_drain(sb, sbuf_count_drain, &size);
sbuf_bcat(sb, &structsize, sizeof(structsize));
PROC_LOCK(p);
kern_proc_vmmap_out(p, sb, -1, vmmap_flags);
sbuf_finish(sb);
sbuf_delete(sb);
*sizep = size;
} else {
sbuf_bcat(sb, &structsize, sizeof(structsize));
PROC_LOCK(p);
kern_proc_vmmap_out(p, sb, *sizep - sizeof(structsize),
vmmap_flags);
}
}
static void
note_procstat_groups(void *arg, struct sbuf *sb, size_t *sizep)
{
struct proc *p;
size_t size;
int structsize;
p = arg;
size = sizeof(structsize) + p->p_ucred->cr_ngroups * sizeof(gid_t);
if (sb != NULL) {
KASSERT(*sizep == size, ("invalid size"));
structsize = sizeof(gid_t);
sbuf_bcat(sb, &structsize, sizeof(structsize));
sbuf_bcat(sb, p->p_ucred->cr_groups, p->p_ucred->cr_ngroups *
sizeof(gid_t));
}
*sizep = size;
}
static void
note_procstat_umask(void *arg, struct sbuf *sb, size_t *sizep)
{
struct proc *p;
size_t size;
int structsize;
p = arg;
size = sizeof(structsize) + sizeof(p->p_pd->pd_cmask);
if (sb != NULL) {
KASSERT(*sizep == size, ("invalid size"));
structsize = sizeof(p->p_pd->pd_cmask);
sbuf_bcat(sb, &structsize, sizeof(structsize));
sbuf_bcat(sb, &p->p_pd->pd_cmask, sizeof(p->p_pd->pd_cmask));
}
*sizep = size;
}
static void
note_procstat_rlimit(void *arg, struct sbuf *sb, size_t *sizep)
{
struct proc *p;
struct rlimit rlim[RLIM_NLIMITS];
size_t size;
int structsize, i;
p = arg;
size = sizeof(structsize) + sizeof(rlim);
if (sb != NULL) {
KASSERT(*sizep == size, ("invalid size"));
structsize = sizeof(rlim);
sbuf_bcat(sb, &structsize, sizeof(structsize));
PROC_LOCK(p);
for (i = 0; i < RLIM_NLIMITS; i++)
lim_rlimit_proc(p, i, &rlim[i]);
PROC_UNLOCK(p);
sbuf_bcat(sb, rlim, sizeof(rlim));
}
*sizep = size;
}
static void
note_procstat_osrel(void *arg, struct sbuf *sb, size_t *sizep)
{
struct proc *p;
size_t size;
int structsize;
p = arg;
size = sizeof(structsize) + sizeof(p->p_osrel);
if (sb != NULL) {
KASSERT(*sizep == size, ("invalid size"));
structsize = sizeof(p->p_osrel);
sbuf_bcat(sb, &structsize, sizeof(structsize));
sbuf_bcat(sb, &p->p_osrel, sizeof(p->p_osrel));
}
*sizep = size;
}
static void
__elfN(note_procstat_psstrings)(void *arg, struct sbuf *sb, size_t *sizep)
{
struct proc *p;
elf_ps_strings_t ps_strings;
size_t size;
int structsize;
p = arg;
size = sizeof(structsize) + sizeof(ps_strings);
if (sb != NULL) {
KASSERT(*sizep == size, ("invalid size"));
structsize = sizeof(ps_strings);
#if defined(COMPAT_FREEBSD32) && __ELF_WORD_SIZE == 32
ps_strings = PTROUT(p->p_sysent->sv_psstrings);
#else
ps_strings = p->p_sysent->sv_psstrings;
#endif
sbuf_bcat(sb, &structsize, sizeof(structsize));
sbuf_bcat(sb, &ps_strings, sizeof(ps_strings));
}
*sizep = size;
}
static void
__elfN(note_procstat_auxv)(void *arg, struct sbuf *sb, size_t *sizep)
{
struct proc *p;
size_t size;
int structsize;
p = arg;
if (sb == NULL) {
size = 0;
sb = sbuf_new(NULL, NULL, AT_COUNT * sizeof(Elf_Auxinfo),
SBUF_FIXEDLEN);
sbuf_set_drain(sb, sbuf_count_drain, &size);
sbuf_bcat(sb, &structsize, sizeof(structsize));
PHOLD(p);
proc_getauxv(curthread, p, sb);
PRELE(p);
sbuf_finish(sb);
sbuf_delete(sb);
*sizep = size;
} else {
structsize = sizeof(Elf_Auxinfo);
sbuf_bcat(sb, &structsize, sizeof(structsize));
PHOLD(p);
proc_getauxv(curthread, p, sb);
PRELE(p);
}
}
static boolean_t
__elfN(parse_notes)(struct image_params *imgp, Elf_Note *checknote,
const char *note_vendor, const Elf_Phdr *pnote,
boolean_t (*cb)(const Elf_Note *, void *, boolean_t *), void *cb_arg)
{
const Elf_Note *note, *note0, *note_end;
const char *note_name;
char *buf;
int i, error;
boolean_t res;
/* We need some limit, might as well use PAGE_SIZE. */
if (pnote == NULL || pnote->p_filesz > PAGE_SIZE)
return (FALSE);
ASSERT_VOP_LOCKED(imgp->vp, "parse_notes");
if (pnote->p_offset > PAGE_SIZE ||
pnote->p_filesz > PAGE_SIZE - pnote->p_offset) {
buf = malloc(pnote->p_filesz, M_TEMP, M_NOWAIT);
if (buf == NULL) {
VOP_UNLOCK(imgp->vp);
buf = malloc(pnote->p_filesz, M_TEMP, M_WAITOK);
vn_lock(imgp->vp, LK_SHARED | LK_RETRY);
}
error = vn_rdwr(UIO_READ, imgp->vp, buf, pnote->p_filesz,
pnote->p_offset, UIO_SYSSPACE, IO_NODELOCKED,
curthread->td_ucred, NOCRED, NULL, curthread);
if (error != 0) {
uprintf("i/o error PT_NOTE\n");
goto retf;
}
note = note0 = (const Elf_Note *)buf;
note_end = (const Elf_Note *)(buf + pnote->p_filesz);
} else {
note = note0 = (const Elf_Note *)(imgp->image_header +
pnote->p_offset);
note_end = (const Elf_Note *)(imgp->image_header +
pnote->p_offset + pnote->p_filesz);
buf = NULL;
}
for (i = 0; i < 100 && note >= note0 && note < note_end; i++) {
if (!aligned(note, Elf32_Addr) || (const char *)note_end -
(const char *)note < sizeof(Elf_Note)) {
goto retf;
}
if (note->n_namesz != checknote->n_namesz ||
note->n_descsz != checknote->n_descsz ||
note->n_type != checknote->n_type)
goto nextnote;
note_name = (const char *)(note + 1);
if (note_name + checknote->n_namesz >=
(const char *)note_end || strncmp(note_vendor,
note_name, checknote->n_namesz) != 0)
goto nextnote;
if (cb(note, cb_arg, &res))
goto ret;
nextnote:
note = (const Elf_Note *)((const char *)(note + 1) +
roundup2(note->n_namesz, ELF_NOTE_ROUNDSIZE) +
roundup2(note->n_descsz, ELF_NOTE_ROUNDSIZE));
}
retf:
res = FALSE;
ret:
free(buf, M_TEMP);
return (res);
}
struct brandnote_cb_arg {
Elf_Brandnote *brandnote;
int32_t *osrel;
};
static boolean_t
brandnote_cb(const Elf_Note *note, void *arg0, boolean_t *res)
{
struct brandnote_cb_arg *arg;
arg = arg0;
/*
* Fetch the osreldate for binary from the ELF OSABI-note if
* necessary.
*/
*res = (arg->brandnote->flags & BN_TRANSLATE_OSREL) != 0 &&
arg->brandnote->trans_osrel != NULL ?
arg->brandnote->trans_osrel(note, arg->osrel) : TRUE;
return (TRUE);
}
static Elf_Note fctl_note = {
.n_namesz = sizeof(FREEBSD_ABI_VENDOR),
.n_descsz = sizeof(uint32_t),
.n_type = NT_FREEBSD_FEATURE_CTL,
};
struct fctl_cb_arg {
boolean_t *has_fctl0;
uint32_t *fctl0;
};
static boolean_t
note_fctl_cb(const Elf_Note *note, void *arg0, boolean_t *res)
{
struct fctl_cb_arg *arg;
const Elf32_Word *desc;
uintptr_t p;
arg = arg0;
p = (uintptr_t)(note + 1);
p += roundup2(note->n_namesz, ELF_NOTE_ROUNDSIZE);
desc = (const Elf32_Word *)p;
*arg->has_fctl0 = TRUE;
*arg->fctl0 = desc[0];
*res = TRUE;
return (TRUE);
}
/*
* Try to find the appropriate ABI-note section for checknote, fetch
* the osreldate and feature control flags for binary from the ELF
* OSABI-note. Only the first page of the image is searched, the same
* as for headers.
*/
static boolean_t
__elfN(check_note)(struct image_params *imgp, Elf_Brandnote *brandnote,
int32_t *osrel, boolean_t *has_fctl0, uint32_t *fctl0)
{
const Elf_Phdr *phdr;
const Elf_Ehdr *hdr;
struct brandnote_cb_arg b_arg;
struct fctl_cb_arg f_arg;
int i, j;
hdr = (const Elf_Ehdr *)imgp->image_header;
phdr = (const Elf_Phdr *)(imgp->image_header + hdr->e_phoff);
b_arg.brandnote = brandnote;
b_arg.osrel = osrel;
f_arg.has_fctl0 = has_fctl0;
f_arg.fctl0 = fctl0;
for (i = 0; i < hdr->e_phnum; i++) {
if (phdr[i].p_type == PT_NOTE && __elfN(parse_notes)(imgp,
&brandnote->hdr, brandnote->vendor, &phdr[i], brandnote_cb,
&b_arg)) {
for (j = 0; j < hdr->e_phnum; j++) {
if (phdr[j].p_type == PT_NOTE &&
__elfN(parse_notes)(imgp, &fctl_note,
FREEBSD_ABI_VENDOR, &phdr[j],
note_fctl_cb, &f_arg))
break;
}
return (TRUE);
}
}
return (FALSE);
}
/*
* Tell kern_execve.c about it, with a little help from the linker.
*/
static struct execsw __elfN(execsw) = {
.ex_imgact = __CONCAT(exec_, __elfN(imgact)),
.ex_name = __XSTRING(__CONCAT(ELF, __ELF_WORD_SIZE))
};
EXEC_SET(__CONCAT(elf, __ELF_WORD_SIZE), __elfN(execsw));
static vm_prot_t
__elfN(trans_prot)(Elf_Word flags)
{
vm_prot_t prot;
prot = 0;
if (flags & PF_X)
prot |= VM_PROT_EXECUTE;
if (flags & PF_W)
prot |= VM_PROT_WRITE;
if (flags & PF_R)
prot |= VM_PROT_READ;
#if __ELF_WORD_SIZE == 32 && (defined(__amd64__) || defined(__i386__))
if (i386_read_exec && (flags & PF_R))
prot |= VM_PROT_EXECUTE;
#endif
return (prot);
}
static Elf_Word
__elfN(untrans_prot)(vm_prot_t prot)
{
Elf_Word flags;
flags = 0;
if (prot & VM_PROT_EXECUTE)
flags |= PF_X;
if (prot & VM_PROT_READ)
flags |= PF_R;
if (prot & VM_PROT_WRITE)
flags |= PF_W;
return (flags);
}
void
__elfN(stackgap)(struct image_params *imgp, uintptr_t *stack_base)
{
uintptr_t range, rbase, gap;
int pct;
pct = __elfN(aslr_stack_gap);
if (pct == 0)
return;
if (pct > 50)
pct = 50;
range = imgp->eff_stack_sz * pct / 100;
arc4rand(&rbase, sizeof(rbase), 0);
gap = rbase % range;
gap &= ~(sizeof(u_long) - 1);
*stack_base -= gap;
}
diff --git a/sys/mips/mips/elf_machdep.c b/sys/mips/mips/elf_machdep.c
index 9e27e5e7f2a4..9df6d1d36436 100644
--- a/sys/mips/mips/elf_machdep.c
+++ b/sys/mips/mips/elf_machdep.c
@@ -1,509 +1,511 @@
/*-
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright 1996-1998 John D. Polstra.
* 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 ``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.
*
* from: src/sys/i386/i386/elf_machdep.c,v 1.20 2004/08/11 02:35:05 marcel
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/kernel.h>
#include <sys/systm.h>
#include <sys/exec.h>
#include <sys/imgact.h>
#include <sys/linker.h>
#include <sys/sysent.h>
#include <sys/imgact_elf.h>
#include <sys/proc.h>
#include <sys/syscall.h>
#include <sys/signalvar.h>
#include <sys/vnode.h>
#include <vm/vm.h>
#include <vm/pmap.h>
#include <vm/vm_param.h>
#include <machine/elf.h>
#include <machine/md_var.h>
#include <machine/cache.h>
static struct sysentvec elf_freebsd_sysvec = {
.sv_size = SYS_MAXSYSCALL,
.sv_table = sysent,
.sv_transtrap = NULL,
.sv_fixup = __elfN(freebsd_fixup),
.sv_sendsig = sendsig,
.sv_sigcode = sigcode,
.sv_szsigcode = &szsigcode,
#ifdef __mips_n64
.sv_name = "FreeBSD ELF64",
#else
.sv_name = "FreeBSD ELF32",
#endif
.sv_coredump = __elfN(coredump),
+ .sv_elf_core_abi_vendor = FREEBSD_ABI_VENDOR,
+ .sv_elf_core_prepare_notes = __elfN(prepare_notes),
.sv_imgact_try = NULL,
.sv_minsigstksz = MINSIGSTKSZ,
.sv_minuser = VM_MIN_ADDRESS,
.sv_maxuser = VM_MAXUSER_ADDRESS,
.sv_usrstack = USRSTACK,
.sv_psstrings = PS_STRINGS,
.sv_stackprot = VM_PROT_ALL,
.sv_copyout_auxargs = __elfN(freebsd_copyout_auxargs),
.sv_copyout_strings = exec_copyout_strings,
.sv_setregs = exec_setregs,
.sv_fixlimit = NULL,
.sv_maxssiz = NULL,
.sv_flags = SV_ABI_FREEBSD | SV_ASLR | SV_RNG_SEED_VER |
#ifdef __mips_n64
SV_LP64,
#else
SV_ILP32,
#endif
.sv_set_syscall_retval = cpu_set_syscall_retval,
.sv_fetch_syscall_args = cpu_fetch_syscall_args,
.sv_syscallnames = syscallnames,
.sv_schedtail = NULL,
.sv_thread_detach = NULL,
.sv_trap = NULL,
};
static __ElfN(Brandinfo) freebsd_brand_info = {
.brand = ELFOSABI_FREEBSD,
.machine = EM_MIPS,
.compat_3_brand = "FreeBSD",
.emul_path = NULL,
.interp_path = "/libexec/ld-elf.so.1",
.sysvec = &elf_freebsd_sysvec,
.interp_newpath = NULL,
.brand_note = &__elfN(freebsd_brandnote),
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
};
SYSINIT(elf, SI_SUB_EXEC, SI_ORDER_ANY,
(sysinit_cfunc_t) __elfN(insert_brand_entry),
&freebsd_brand_info);
void
__elfN(dump_thread)(struct thread *td __unused, void *dst __unused,
size_t *off __unused)
{
}
/*
* The following MIPS relocation code for tracking multiple
* consecutive HI32/LO32 entries is because of the following:
*
* https://dmz-portal.mips.com/wiki/MIPS_relocation_types
*
* ===
*
* + R_MIPS_HI16
*
* An R_MIPS_HI16 must be followed eventually by an associated R_MIPS_LO16
* relocation record in the same SHT_REL section. The contents of the two
* fields to be relocated are combined to form a full 32-bit addend AHL.
* An R_MIPS_LO16 entry which does not immediately follow a R_MIPS_HI16 is
* combined with the most recent one encountered, i.e. multiple R_MIPS_LO16
* entries may be associated with a single R_MIPS_HI16. Use of these
* relocation types in a SHT_REL section is discouraged and may be
* forbidden to avoid this complication.
*
* A GNU extension allows multiple R_MIPS_HI16 records to share the same
* R_MIPS_LO16 relocation record(s). The association works like this within
* a single relocation section:
*
* + From the beginning of the section moving to the end of the section,
* until R_MIPS_LO16 is not found each found R_MIPS_HI16 relocation will
* be associated with the first R_MIPS_LO16.
*
* + Until another R_MIPS_HI16 record is found all found R_MIPS_LO16
* relocations found are associated with the last R_MIPS_HI16.
*
* ===
*
* This is so gcc can do dead code detection/removal without having to
* generate HI/LO pairs even if one of them would be deleted.
*
* So, the summary is:
*
* + A HI16 entry must occur before any LO16 entries;
* + Multiple consecutive HI16 RELA entries need to be buffered until the
* first LO16 RELA entry occurs - and then all HI16 RELA relocations use
* the offset in the LOW16 RELA for calculating their offsets;
* + The last HI16 RELA entry before a LO16 RELA entry is used (the AHL)
* for the first subsequent LO16 calculation;
* + If multiple consecutive LO16 RELA entries occur, only the first
* LO16 RELA entry triggers an update of buffered HI16 RELA entries;
* any subsequent LO16 RELA entry before another HI16 RELA entry will
* not cause any further updates to the HI16 RELA entries.
*
* Additionally, flush out any outstanding HI16 entries that don't have
* a LO16 entry in case some garbage entries are left in the file.
*/
struct mips_tmp_reloc;
struct mips_tmp_reloc {
struct mips_tmp_reloc *next;
Elf_Addr ahl;
Elf32_Addr *where_hi16;
};
static struct mips_tmp_reloc *ml = NULL;
/*
* Add a temporary relocation (ie, a HI16 reloc type.)
*/
static int
mips_tmp_reloc_add(Elf_Addr ahl, Elf32_Addr *where_hi16)
{
struct mips_tmp_reloc *r;
r = malloc(sizeof(struct mips_tmp_reloc), M_TEMP, M_NOWAIT);
if (r == NULL) {
printf("%s: failed to malloc\n", __func__);
return (0);
}
r->ahl = ahl;
r->where_hi16 = where_hi16;
r->next = ml;
ml = r;
return (1);
}
/*
* Flush the temporary relocation list.
*
* This should be done after a file is completely loaded
* so no stale relocations exist to confuse the next
* load.
*/
static void
mips_tmp_reloc_flush(void)
{
struct mips_tmp_reloc *r, *rn;
r = ml;
ml = NULL;
while (r != NULL) {
rn = r->next;
free(r, M_TEMP);
r = rn;
}
}
/*
* Get an entry from the reloc list; or NULL if we've run out.
*/
static struct mips_tmp_reloc *
mips_tmp_reloc_get(void)
{
struct mips_tmp_reloc *r;
r = ml;
if (r == NULL)
return (NULL);
ml = ml->next;
return (r);
}
/*
* Free a relocation entry.
*/
static void
mips_tmp_reloc_free(struct mips_tmp_reloc *r)
{
free(r, M_TEMP);
}
bool
elf_is_ifunc_reloc(Elf_Size r_info __unused)
{
return (false);
}
/* Process one elf relocation with addend. */
static int
elf_reloc_internal(linker_file_t lf, Elf_Addr relocbase, const void *data,
int type, int local, elf_lookup_fn lookup)
{
Elf32_Addr *where = (Elf32_Addr *)NULL;
Elf_Addr addr;
Elf_Addr addend = (Elf_Addr)0;
Elf_Word rtype = (Elf_Word)0, symidx;
struct mips_tmp_reloc *r;
const Elf_Rel *rel = NULL;
const Elf_Rela *rela = NULL;
int error;
/* Store the last seen ahl from a HI16 for LO16 processing */
static Elf_Addr last_ahl;
switch (type) {
case ELF_RELOC_REL:
rel = (const Elf_Rel *)data;
where = (Elf32_Addr *) (relocbase + rel->r_offset);
rtype = ELF_R_TYPE(rel->r_info);
symidx = ELF_R_SYM(rel->r_info);
switch (rtype) {
case R_MIPS_64:
addend = *(Elf64_Addr *)where;
break;
default:
addend = *where;
break;
}
break;
case ELF_RELOC_RELA:
rela = (const Elf_Rela *)data;
where = (Elf32_Addr *) (relocbase + rela->r_offset);
addend = rela->r_addend;
rtype = ELF_R_TYPE(rela->r_info);
symidx = ELF_R_SYM(rela->r_info);
break;
default:
panic("unknown reloc type %d\n", type);
}
switch (rtype) {
case R_MIPS_NONE: /* none */
break;
case R_MIPS_32: /* S + A */
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
addr += addend;
if (*where != addr)
*where = (Elf32_Addr)addr;
break;
case R_MIPS_26: /* ((A << 2) | (P & 0xf0000000) + S) >> 2 */
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
addend &= 0x03ffffff;
/*
* Addendum for .rela R_MIPS_26 is not shifted right
*/
if (rela == NULL)
addend <<= 2;
addr += ((Elf_Addr)where & 0xf0000000) | addend;
addr >>= 2;
*where &= ~0x03ffffff;
*where |= addr & 0x03ffffff;
break;
case R_MIPS_64: /* S + A */
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
addr += addend;
if (*(Elf64_Addr*)where != addr)
*(Elf64_Addr*)where = addr;
break;
/*
* Handle the two GNU extension cases:
*
* + Multiple HI16s followed by a LO16, and
* + A HI16 followed by multiple LO16s.
*
* The former is tricky - the HI16 relocations need
* to be buffered until a LO16 occurs, at which point
* each HI16 is replayed against the LO16 relocation entry
* (with the relevant overflow information.)
*
* The latter should be easy to handle - when the
* first LO16 is seen, write out and flush the
* HI16 buffer. Any subsequent LO16 entries will
* find a blank relocation buffer.
*
*/
case R_MIPS_HI16: /* ((AHL + S) - ((short)(AHL + S)) >> 16 */
if (rela != NULL) {
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
addr += addend;
*where &= 0xffff0000;
*where |= ((((long long) addr + 0x8000LL) >> 16) & 0xffff);
} else {
/*
* Add a temporary relocation to the list;
* will pop it off / free the list when
* we've found a suitable HI16.
*/
if (mips_tmp_reloc_add(addend << 16, where) == 0)
return (-1);
/*
* Track the last seen HI16 AHL for use by
* the first LO16 AHL calculation.
*
* The assumption is any intermediary deleted
* LO16's were optimised out, so the last
* HI16 before the LO16 is the "true" relocation
* entry to use for that LO16 write.
*/
last_ahl = addend << 16;
}
break;
case R_MIPS_LO16: /* AHL + S */
if (rela != NULL) {
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
addr += addend;
*where &= 0xffff0000;
*where |= addr & 0xffff;
} else {
Elf_Addr tmp_ahl;
Elf_Addr tmp_addend;
tmp_ahl = last_ahl + (int16_t) addend;
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
tmp_addend = addend & 0xffff0000;
/* Use the last seen ahl for calculating addend */
tmp_addend |= (uint16_t)(tmp_ahl + addr);
*where = tmp_addend;
/*
* This logic implements the "we saw multiple HI16
* before a LO16" assignment /and/ "we saw multiple
* LO16s".
*
* Multiple LO16s will be handled as a blank
* relocation list.
*
* Multple HI16's are iterated over here.
*/
while ((r = mips_tmp_reloc_get()) != NULL) {
Elf_Addr rahl;
/*
* We have the ahl from the HI16 entry, so
* offset it by the 16 bits of the low ahl.
*/
rahl = r->ahl;
rahl += (int16_t) addend;
tmp_addend = *(r->where_hi16);
tmp_addend &= 0xffff0000;
tmp_addend |= ((rahl + addr) -
(int16_t)(rahl + addr)) >> 16;
*(r->where_hi16) = tmp_addend;
mips_tmp_reloc_free(r);
}
}
break;
case R_MIPS_HIGHER: /* %higher(A+S) */
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
addr += addend;
*where &= 0xffff0000;
*where |= (((long long)addr + 0x80008000LL) >> 32) & 0xffff;
break;
case R_MIPS_HIGHEST: /* %highest(A+S) */
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
addr += addend;
*where &= 0xffff0000;
*where |= (((long long)addr + 0x800080008000LL) >> 48) & 0xffff;
break;
default:
printf("kldload: unexpected relocation type %d, "
"symbol index %d\n", rtype, symidx);
return (-1);
}
return (0);
}
int
elf_reloc(linker_file_t lf, Elf_Addr relocbase, const void *data, int type,
elf_lookup_fn lookup)
{
return (elf_reloc_internal(lf, relocbase, data, type, 0, lookup));
}
int
elf_reloc_local(linker_file_t lf, Elf_Addr relocbase, const void *data,
int type, elf_lookup_fn lookup)
{
return (elf_reloc_internal(lf, relocbase, data, type, 1, lookup));
}
int
elf_cpu_load_file(linker_file_t lf __unused)
{
/*
* Sync the I and D caches to make sure our relocations are visible.
*/
mips_icache_sync_all();
/* Flush outstanding relocations */
mips_tmp_reloc_flush();
return (0);
}
int
elf_cpu_unload_file(linker_file_t lf __unused)
{
return (0);
}
int
elf_cpu_parse_dynamic(caddr_t loadbase __unused, Elf_Dyn *dynamic __unused)
{
return (0);
}
diff --git a/sys/mips/mips/freebsd32_machdep.c b/sys/mips/mips/freebsd32_machdep.c
index 5d7fd4719efb..e30753e7a59c 100644
--- a/sys/mips/mips/freebsd32_machdep.c
+++ b/sys/mips/mips/freebsd32_machdep.c
@@ -1,485 +1,488 @@
/*-
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright (c) 2012 Juli Mallett <jmallett@FreeBSD.org>
* 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.
*
* $FreeBSD$
*/
/*
* Based on nwhitehorn's COMPAT_FREEBSD32 support code for PowerPC64.
*/
#define __ELF_WORD_SIZE 32
#include <sys/types.h>
#include <sys/param.h>
#include <sys/kernel.h>
#include <sys/systm.h>
#include <sys/sysent.h>
#include <sys/exec.h>
#include <sys/imgact.h>
#include <sys/ktr.h>
#include <sys/malloc.h>
#include <sys/proc.h>
#include <sys/namei.h>
#include <sys/fcntl.h>
#include <sys/sysent.h>
#include <sys/imgact_elf.h>
#include <sys/syscall.h>
#include <sys/syscallsubr.h>
#include <sys/sysproto.h>
#include <sys/signalvar.h>
#include <sys/vnode.h>
#include <sys/linker.h>
#include <vm/vm.h>
#include <vm/vm_param.h>
#include <machine/cpuinfo.h>
#include <machine/md_var.h>
#include <machine/reg.h>
#include <machine/sigframe.h>
#include <machine/sysarch.h>
#include <machine/tls.h>
#include <compat/freebsd32/freebsd32_signal.h>
#include <compat/freebsd32/freebsd32_util.h>
#include <compat/freebsd32/freebsd32_proto.h>
static int get_mcontext32(struct thread *, mcontext32_t *, int);
static int set_mcontext32(struct thread *, mcontext32_t *);
static void freebsd32_sendsig(sig_t, ksiginfo_t *, sigset_t *);
extern const char *freebsd32_syscallnames[];
struct sysentvec elf32_freebsd_sysvec = {
.sv_size = SYS_MAXSYSCALL,
.sv_table = freebsd32_sysent,
.sv_transtrap = NULL,
.sv_fixup = __elfN(freebsd_fixup),
.sv_sendsig = freebsd32_sendsig,
.sv_sigcode = sigcode32,
.sv_szsigcode = &szsigcode32,
.sv_name = "FreeBSD ELF32",
.sv_coredump = __elfN(coredump),
+ .sv_elf_core_osabi = ELFOSABI_FREEBSD,
+ .sv_elf_core_abi_vendor = FREEBSD_ABI_VENDOR,
+ .sv_elf_core_prepare_notes = __elfN(prepare_notes),
.sv_imgact_try = NULL,
.sv_minsigstksz = MINSIGSTKSZ,
.sv_minuser = VM_MIN_ADDRESS,
.sv_maxuser = ((vm_offset_t)0x80000000),
.sv_usrstack = FREEBSD32_USRSTACK,
.sv_psstrings = FREEBSD32_PS_STRINGS,
.sv_stackprot = VM_PROT_ALL,
.sv_copyout_auxargs = __elfN(freebsd_copyout_auxargs),
.sv_copyout_strings = freebsd32_copyout_strings,
.sv_setregs = exec_setregs,
.sv_fixlimit = NULL,
.sv_maxssiz = NULL,
.sv_flags = SV_ABI_FREEBSD | SV_ILP32 | SV_RNG_SEED_VER,
.sv_set_syscall_retval = cpu_set_syscall_retval,
.sv_fetch_syscall_args = cpu_fetch_syscall_args,
.sv_syscallnames = freebsd32_syscallnames,
.sv_schedtail = NULL,
.sv_thread_detach = NULL,
.sv_trap = NULL,
};
INIT_SYSENTVEC(elf32_sysvec, &elf32_freebsd_sysvec);
static Elf32_Brandinfo freebsd_brand_info = {
.brand = ELFOSABI_FREEBSD,
.machine = EM_MIPS,
.compat_3_brand = "FreeBSD",
.emul_path = NULL,
.interp_path = "/libexec/ld-elf.so.1",
.sysvec = &elf32_freebsd_sysvec,
.interp_newpath = "/libexec/ld-elf32.so.1",
.brand_note = &elf32_freebsd_brandnote,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
};
SYSINIT(elf32, SI_SUB_EXEC, SI_ORDER_FIRST,
(sysinit_cfunc_t) elf32_insert_brand_entry,
&freebsd_brand_info);
int
set_regs32(struct thread *td, struct reg32 *regs)
{
struct reg r;
unsigned i;
for (i = 0; i < NUMSAVEREGS; i++)
r.r_regs[i] = regs->r_regs[i];
return (set_regs(td, &r));
}
int
fill_regs32(struct thread *td, struct reg32 *regs)
{
struct reg r;
unsigned i;
int error;
error = fill_regs(td, &r);
if (error != 0)
return (error);
for (i = 0; i < NUMSAVEREGS; i++)
regs->r_regs[i] = r.r_regs[i];
return (0);
}
int
set_fpregs32(struct thread *td, struct fpreg32 *fpregs)
{
struct fpreg fp;
unsigned i;
for (i = 0; i < NUMFPREGS; i++)
fp.r_regs[i] = fpregs->r_regs[i];
return (set_fpregs(td, &fp));
}
int
fill_fpregs32(struct thread *td, struct fpreg32 *fpregs)
{
struct fpreg fp;
unsigned i;
int error;
error = fill_fpregs(td, &fp);
if (error != 0)
return (error);
for (i = 0; i < NUMFPREGS; i++)
fpregs->r_regs[i] = fp.r_regs[i];
return (0);
}
static int
get_mcontext32(struct thread *td, mcontext32_t *mcp, int flags)
{
mcontext_t mcp64;
unsigned i;
int error;
error = get_mcontext(td, &mcp64, flags);
if (error != 0)
return (error);
mcp->mc_onstack = mcp64.mc_onstack;
mcp->mc_pc = mcp64.mc_pc;
for (i = 0; i < 32; i++)
mcp->mc_regs[i] = mcp64.mc_regs[i];
mcp->sr = mcp64.sr;
mcp->mullo = mcp64.mullo;
mcp->mulhi = mcp64.mulhi;
mcp->mc_fpused = mcp64.mc_fpused;
for (i = 0; i < 33; i++)
mcp->mc_fpregs[i] = mcp64.mc_fpregs[i];
mcp->mc_fpc_eir = mcp64.mc_fpc_eir;
mcp->mc_tls = (int32_t)(intptr_t)mcp64.mc_tls;
return (0);
}
static int
set_mcontext32(struct thread *td, mcontext32_t *mcp)
{
mcontext_t mcp64;
unsigned i;
mcp64.mc_onstack = mcp->mc_onstack;
mcp64.mc_pc = mcp->mc_pc;
for (i = 0; i < 32; i++)
mcp64.mc_regs[i] = mcp->mc_regs[i];
mcp64.sr = mcp->sr;
mcp64.mullo = mcp->mullo;
mcp64.mulhi = mcp->mulhi;
mcp64.mc_fpused = mcp->mc_fpused;
for (i = 0; i < 33; i++)
mcp64.mc_fpregs[i] = mcp->mc_fpregs[i];
mcp64.mc_fpc_eir = mcp->mc_fpc_eir;
mcp64.mc_tls = (void *)(intptr_t)mcp->mc_tls;
return (set_mcontext(td, &mcp64));
}
int
freebsd32_sigreturn(struct thread *td, struct freebsd32_sigreturn_args *uap)
{
ucontext32_t uc;
int error;
CTR2(KTR_SIG, "sigreturn: td=%p ucp=%p", td, uap->sigcntxp);
if (copyin(uap->sigcntxp, &uc, sizeof(uc)) != 0) {
CTR1(KTR_SIG, "sigreturn: efault td=%p", td);
return (EFAULT);
}
error = set_mcontext32(td, &uc.uc_mcontext);
if (error != 0)
return (error);
kern_sigprocmask(td, SIG_SETMASK, &uc.uc_sigmask, NULL, 0);
#if 0
CTR3(KTR_SIG, "sigreturn: return td=%p pc=%#x sp=%#x",
td, uc.uc_mcontext.mc_srr0, uc.uc_mcontext.mc_gpr[1]);
#endif
return (EJUSTRETURN);
}
/*
* The first two fields of a ucontext_t are the signal mask and the machine
* context. The next field is uc_link; we want to avoid destroying the link
* when copying out contexts.
*/
#define UC32_COPY_SIZE offsetof(ucontext32_t, uc_link)
int
freebsd32_getcontext(struct thread *td, struct freebsd32_getcontext_args *uap)
{
ucontext32_t uc;
int ret;
if (uap->ucp == NULL)
ret = EINVAL;
else {
bzero(&uc, sizeof(uc));
get_mcontext32(td, &uc.uc_mcontext, GET_MC_CLEAR_RET);
PROC_LOCK(td->td_proc);
uc.uc_sigmask = td->td_sigmask;
PROC_UNLOCK(td->td_proc);
ret = copyout(&uc, uap->ucp, UC32_COPY_SIZE);
}
return (ret);
}
int
freebsd32_setcontext(struct thread *td, struct freebsd32_setcontext_args *uap)
{
ucontext32_t uc;
int ret;
if (uap->ucp == NULL)
ret = EINVAL;
else {
ret = copyin(uap->ucp, &uc, UC32_COPY_SIZE);
if (ret == 0) {
ret = set_mcontext32(td, &uc.uc_mcontext);
if (ret == 0) {
kern_sigprocmask(td, SIG_SETMASK,
&uc.uc_sigmask, NULL, 0);
}
}
}
return (ret == 0 ? EJUSTRETURN : ret);
}
int
freebsd32_swapcontext(struct thread *td, struct freebsd32_swapcontext_args *uap)
{
ucontext32_t uc;
int ret;
if (uap->oucp == NULL || uap->ucp == NULL)
ret = EINVAL;
else {
bzero(&uc, sizeof(uc));
get_mcontext32(td, &uc.uc_mcontext, GET_MC_CLEAR_RET);
PROC_LOCK(td->td_proc);
uc.uc_sigmask = td->td_sigmask;
PROC_UNLOCK(td->td_proc);
ret = copyout(&uc, uap->oucp, UC32_COPY_SIZE);
if (ret == 0) {
ret = copyin(uap->ucp, &uc, UC32_COPY_SIZE);
if (ret == 0) {
ret = set_mcontext32(td, &uc.uc_mcontext);
if (ret == 0) {
kern_sigprocmask(td, SIG_SETMASK,
&uc.uc_sigmask, NULL, 0);
}
}
}
}
return (ret == 0 ? EJUSTRETURN : ret);
}
#define UCONTEXT_MAGIC 0xACEDBADE
/*
* Send an interrupt to process.
*
* Stack is set up to allow sigcode stored
* at top 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
freebsd32_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
{
struct proc *p;
struct thread *td;
struct fpreg32 fpregs;
struct reg32 regs;
struct sigacts *psp;
struct sigframe32 sf, *sfp;
int sig;
int oonstack;
unsigned i;
td = curthread;
p = td->td_proc;
PROC_LOCK_ASSERT(p, MA_OWNED);
sig = ksi->ksi_signo;
psp = p->p_sigacts;
mtx_assert(&psp->ps_mtx, MA_OWNED);
fill_regs32(td, ®s);
oonstack = sigonstack(td->td_frame->sp);
/* save user context */
bzero(&sf, sizeof sf);
sf.sf_uc.uc_sigmask = *mask;
sf.sf_uc.uc_stack.ss_sp = (int32_t)(intptr_t)td->td_sigstk.ss_sp;
sf.sf_uc.uc_stack.ss_size = td->td_sigstk.ss_size;
sf.sf_uc.uc_stack.ss_flags = td->td_sigstk.ss_flags;
sf.sf_uc.uc_mcontext.mc_onstack = (oonstack) ? 1 : 0;
sf.sf_uc.uc_mcontext.mc_pc = regs.r_regs[PC];
sf.sf_uc.uc_mcontext.mullo = regs.r_regs[MULLO];
sf.sf_uc.uc_mcontext.mulhi = regs.r_regs[MULHI];
sf.sf_uc.uc_mcontext.mc_tls = (int32_t)(intptr_t)td->td_md.md_tls;
sf.sf_uc.uc_mcontext.mc_regs[0] = UCONTEXT_MAGIC; /* magic number */
for (i = 1; i < 32; i++)
sf.sf_uc.uc_mcontext.mc_regs[i] = regs.r_regs[i];
sf.sf_uc.uc_mcontext.mc_fpused = td->td_md.md_flags & MDTD_FPUSED;
if (sf.sf_uc.uc_mcontext.mc_fpused) {
/* if FPU has current state, save it first */
if (td == PCPU_GET(fpcurthread))
MipsSaveCurFPState(td);
fill_fpregs32(td, &fpregs);
for (i = 0; i < 33; i++)
sf.sf_uc.uc_mcontext.mc_fpregs[i] = fpregs.r_regs[i];
}
/* Allocate and validate space for the signal handler context. */
if ((td->td_pflags & TDP_ALTSTACK) != 0 && !oonstack &&
SIGISMEMBER(psp->ps_sigonstack, sig)) {
sfp = (struct sigframe32 *)(((uintptr_t)td->td_sigstk.ss_sp +
td->td_sigstk.ss_size - sizeof(struct sigframe32))
& ~(sizeof(__int64_t) - 1));
} else
sfp = (struct sigframe32 *)((vm_offset_t)(td->td_frame->sp -
sizeof(struct sigframe32)) & ~(sizeof(__int64_t) - 1));
/* Build the argument list for the signal handler. */
td->td_frame->a0 = sig;
td->td_frame->a2 = (register_t)(intptr_t)&sfp->sf_uc;
if (SIGISMEMBER(psp->ps_siginfo, sig)) {
/* Signal handler installed with SA_SIGINFO. */
td->td_frame->a1 = (register_t)(intptr_t)&sfp->sf_si;
/* sf.sf_ahu.sf_action = (__siginfohandler_t *)catcher; */
/* fill siginfo structure */
sf.sf_si.si_signo = sig;
sf.sf_si.si_code = ksi->ksi_code;
sf.sf_si.si_addr = td->td_frame->badvaddr;
} else {
/* Old FreeBSD-style arguments. */
td->td_frame->a1 = ksi->ksi_code;
td->td_frame->a3 = td->td_frame->badvaddr;
/* sf.sf_ahu.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(struct sigframe32)) != 0) {
/*
* Something is wrong with the stack pointer.
* ...Kill the process.
*/
PROC_LOCK(p);
sigexit(td, SIGILL);
}
td->td_frame->pc = (register_t)(intptr_t)catcher;
td->td_frame->t9 = (register_t)(intptr_t)catcher;
td->td_frame->sp = (register_t)(intptr_t)sfp;
/*
* Signal trampoline code is at base of user stack.
*/
td->td_frame->ra = (register_t)(intptr_t)FREEBSD32_PS_STRINGS - *(p->p_sysent->sv_szsigcode);
PROC_LOCK(p);
mtx_lock(&psp->ps_mtx);
}
int
freebsd32_sysarch(struct thread *td, struct freebsd32_sysarch_args *uap)
{
int error;
int32_t tlsbase;
switch (uap->op) {
case MIPS_SET_TLS:
td->td_md.md_tls = (void *)(intptr_t)uap->parms;
/*
* If there is an user local register implementation (ULRI)
* update it as well. Add the TLS and TCB offsets so the
* value in this register is adjusted like in the case of the
* rdhwr trap() instruction handler.
*/
if (cpuinfo.userlocal_reg == true) {
mips_wr_userlocal((unsigned long)(uap->parms +
td->td_proc->p_md.md_tls_tcb_offset));
}
return (0);
case MIPS_GET_TLS:
tlsbase = (int32_t)(intptr_t)td->td_md.md_tls;
error = copyout(&tlsbase, uap->parms, sizeof(tlsbase));
return (error);
default:
break;
}
return (EINVAL);
}
void
elf32_dump_thread(struct thread *td __unused, void *dst __unused,
size_t *off __unused)
{
}
diff --git a/sys/powerpc/powerpc/elf32_machdep.c b/sys/powerpc/powerpc/elf32_machdep.c
index 04150f02bb9c..2c4928f7d9df 100644
--- a/sys/powerpc/powerpc/elf32_machdep.c
+++ b/sys/powerpc/powerpc/elf32_machdep.c
@@ -1,464 +1,467 @@
/*-
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright 1996-1998 John D. Polstra.
* 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 ``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.
*
* $FreeBSD$
*/
#include <sys/param.h>
#include <sys/kernel.h>
#include <sys/systm.h>
#define __ELF_WORD_SIZE 32
#include <sys/exec.h>
#include <sys/imgact.h>
#include <sys/malloc.h>
#include <sys/proc.h>
#include <sys/namei.h>
#include <sys/fcntl.h>
#include <sys/sysent.h>
#include <sys/imgact_elf.h>
#include <sys/jail.h>
#include <sys/smp.h>
#include <sys/syscall.h>
#include <sys/sysctl.h>
#include <sys/signalvar.h>
#include <sys/vnode.h>
#include <sys/linker.h>
#include <vm/vm.h>
#include <vm/vm_param.h>
#include <machine/altivec.h>
#include <machine/cpu.h>
#include <machine/fpu.h>
#include <machine/elf.h>
#include <machine/reg.h>
#include <machine/md_var.h>
#include <powerpc/powerpc/elf_common.c>
#ifdef __powerpc64__
#include <compat/freebsd32/freebsd32_proto.h>
#include <compat/freebsd32/freebsd32_util.h>
extern const char *freebsd32_syscallnames[];
static void ppc32_fixlimit(struct rlimit *rl, int which);
static SYSCTL_NODE(_compat, OID_AUTO, ppc32, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
"32-bit mode");
#define PPC32_MAXDSIZ (1024*1024*1024)
static u_long ppc32_maxdsiz = PPC32_MAXDSIZ;
SYSCTL_ULONG(_compat_ppc32, OID_AUTO, maxdsiz, CTLFLAG_RWTUN, &ppc32_maxdsiz,
0, "");
#define PPC32_MAXSSIZ (64*1024*1024)
u_long ppc32_maxssiz = PPC32_MAXSSIZ;
SYSCTL_ULONG(_compat_ppc32, OID_AUTO, maxssiz, CTLFLAG_RWTUN, &ppc32_maxssiz,
0, "");
#else
static void ppc32_runtime_resolve(void);
#endif
struct sysentvec elf32_freebsd_sysvec = {
.sv_size = SYS_MAXSYSCALL,
#ifdef __powerpc64__
.sv_table = freebsd32_sysent,
#else
.sv_table = sysent,
#endif
.sv_transtrap = NULL,
.sv_fixup = __elfN(freebsd_fixup),
.sv_copyout_auxargs = __elfN(powerpc_copyout_auxargs),
.sv_sendsig = sendsig,
.sv_sigcode = sigcode32,
.sv_szsigcode = &szsigcode32,
.sv_name = "FreeBSD ELF32",
.sv_coredump = __elfN(coredump),
+ .sv_elf_core_osabi = ELFOSABI_FREEBSD,
+ .sv_elf_core_abi_vendor = FREEBSD_ABI_VENDOR,
+ .sv_elf_core_prepare_notes = __elfN(prepare_notes),
.sv_imgact_try = NULL,
.sv_minsigstksz = MINSIGSTKSZ,
.sv_minuser = VM_MIN_ADDRESS,
.sv_stackprot = VM_PROT_ALL,
#ifdef __powerpc64__
.sv_maxuser = VM_MAXUSER_ADDRESS32,
.sv_usrstack = FREEBSD32_USRSTACK,
.sv_psstrings = FREEBSD32_PS_STRINGS,
.sv_copyout_strings = freebsd32_copyout_strings,
.sv_setregs = ppc32_setregs,
.sv_syscallnames = freebsd32_syscallnames,
.sv_fixlimit = ppc32_fixlimit,
#else
.sv_maxuser = VM_MAXUSER_ADDRESS,
.sv_usrstack = USRSTACK,
.sv_psstrings = PS_STRINGS,
.sv_copyout_strings = exec_copyout_strings,
.sv_setregs = exec_setregs,
.sv_syscallnames = syscallnames,
.sv_fixlimit = NULL,
#endif
.sv_maxssiz = NULL,
.sv_flags = SV_ABI_FREEBSD | SV_ILP32 | SV_SHP | SV_ASLR |
SV_TIMEKEEP | SV_RNG_SEED_VER,
.sv_set_syscall_retval = cpu_set_syscall_retval,
.sv_fetch_syscall_args = cpu_fetch_syscall_args,
.sv_shared_page_base = FREEBSD32_SHAREDPAGE,
.sv_shared_page_len = PAGE_SIZE,
.sv_schedtail = NULL,
.sv_thread_detach = NULL,
.sv_trap = NULL,
.sv_hwcap = &cpu_features,
.sv_hwcap2 = &cpu_features2,
};
INIT_SYSENTVEC(elf32_sysvec, &elf32_freebsd_sysvec);
static Elf32_Brandinfo freebsd_brand_info = {
.brand = ELFOSABI_FREEBSD,
.machine = EM_PPC,
.compat_3_brand = "FreeBSD",
.emul_path = NULL,
.interp_path = "/libexec/ld-elf.so.1",
.sysvec = &elf32_freebsd_sysvec,
#ifdef __powerpc64__
.interp_newpath = "/libexec/ld-elf32.so.1",
#else
.interp_newpath = NULL,
#endif
.brand_note = &elf32_freebsd_brandnote,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
};
SYSINIT(elf32, SI_SUB_EXEC, SI_ORDER_FIRST,
(sysinit_cfunc_t) elf32_insert_brand_entry,
&freebsd_brand_info);
static Elf32_Brandinfo freebsd_brand_oinfo = {
.brand = ELFOSABI_FREEBSD,
.machine = EM_PPC,
.compat_3_brand = "FreeBSD",
.emul_path = NULL,
.interp_path = "/usr/libexec/ld-elf.so.1",
.sysvec = &elf32_freebsd_sysvec,
.interp_newpath = NULL,
.brand_note = &elf32_freebsd_brandnote,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
};
SYSINIT(oelf32, SI_SUB_EXEC, SI_ORDER_ANY,
(sysinit_cfunc_t) elf32_insert_brand_entry,
&freebsd_brand_oinfo);
void elf_reloc_self(Elf_Dyn *dynp, Elf_Addr relocbase);
void
elf32_dump_thread(struct thread *td, void *dst, size_t *off)
{
size_t len;
struct pcb *pcb;
uint64_t vshr[32];
uint64_t *vsr_dw1;
int vsr_idx;
len = 0;
pcb = td->td_pcb;
if (pcb->pcb_flags & PCB_VEC) {
save_vec_nodrop(td);
if (dst != NULL) {
len += elf32_populate_note(NT_PPC_VMX,
&pcb->pcb_vec, (char *)dst + len,
sizeof(pcb->pcb_vec), NULL);
} else
len += elf32_populate_note(NT_PPC_VMX, NULL, NULL,
sizeof(pcb->pcb_vec), NULL);
}
if (pcb->pcb_flags & PCB_VSX) {
save_fpu_nodrop(td);
if (dst != NULL) {
/*
* Doubleword 0 of VSR0-VSR31 overlap with FPR0-FPR31 and
* VSR32-VSR63 overlap with VR0-VR31, so we only copy
* the non-overlapping data, which is doubleword 1 of VSR0-VSR31.
*/
for (vsr_idx = 0; vsr_idx < nitems(vshr); vsr_idx++) {
vsr_dw1 = (uint64_t *)&pcb->pcb_fpu.fpr[vsr_idx].vsr[2];
vshr[vsr_idx] = *vsr_dw1;
}
len += elf32_populate_note(NT_PPC_VSX,
vshr, (char *)dst + len,
sizeof(vshr), NULL);
} else
len += elf32_populate_note(NT_PPC_VSX, NULL, NULL,
sizeof(vshr), NULL);
}
*off = len;
}
#ifndef __powerpc64__
bool
elf_is_ifunc_reloc(Elf_Size r_info)
{
return (ELF_R_TYPE(r_info) == R_PPC_IRELATIVE);
}
/* Process one elf relocation with addend. */
static int
elf_reloc_internal(linker_file_t lf, Elf_Addr relocbase, const void *data,
int type, int local, elf_lookup_fn lookup)
{
Elf_Addr *where;
Elf_Half *hwhere;
Elf_Addr addr;
Elf_Addr addend, val;
Elf_Word rtype, symidx;
const Elf_Rela *rela;
int error;
switch (type) {
case ELF_RELOC_REL:
panic("PPC only supports RELA relocations");
break;
case ELF_RELOC_RELA:
rela = (const Elf_Rela *)data;
where = (Elf_Addr *) ((uintptr_t)relocbase + rela->r_offset);
hwhere = (Elf_Half *) ((uintptr_t)relocbase + rela->r_offset);
addend = rela->r_addend;
rtype = ELF_R_TYPE(rela->r_info);
symidx = ELF_R_SYM(rela->r_info);
break;
default:
panic("elf_reloc: unknown relocation mode %d\n", type);
}
switch (rtype) {
case R_PPC_NONE:
break;
case R_PPC_ADDR32: /* word32 S + A */
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
*where = elf_relocaddr(lf, addr + addend);
break;
case R_PPC_ADDR16_LO: /* #lo(S) */
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
/*
* addend values are sometimes relative to sections
* (i.e. .rodata) in rela, where in reality they
* are relative to relocbase. Detect this condition.
*/
if (addr > relocbase && addr <= (relocbase + addend))
addr = relocbase;
addr = elf_relocaddr(lf, addr + addend);
*hwhere = addr & 0xffff;
break;
case R_PPC_ADDR16_HA: /* #ha(S) */
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
/*
* addend values are sometimes relative to sections
* (i.e. .rodata) in rela, where in reality they
* are relative to relocbase. Detect this condition.
*/
if (addr > relocbase && addr <= (relocbase + addend))
addr = relocbase;
addr = elf_relocaddr(lf, addr + addend);
*hwhere = ((addr >> 16) + ((addr & 0x8000) ? 1 : 0))
& 0xffff;
break;
case R_PPC_RELATIVE: /* word32 B + A */
*where = elf_relocaddr(lf, relocbase + addend);
break;
case R_PPC_JMP_SLOT: /* PLT jump slot entry */
/*
* We currently only support Secure-PLT jump slots.
* Given that we reject BSS-PLT modules during load, we
* don't need to check again.
* The method we are using here is equivilent to
* LD_BIND_NOW.
*/
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
*where = elf_relocaddr(lf, addr + addend);
break;
case R_PPC_IRELATIVE:
addr = relocbase + addend;
val = ((Elf32_Addr (*)(void))addr)();
if (*where != val)
*where = val;
break;
default:
printf("kldload: unexpected relocation type %d, "
"symbol index %d\n", (int)rtype, symidx);
return (-1);
}
return (0);
}
void
elf_reloc_self(Elf_Dyn *dynp, Elf_Addr relocbase)
{
Elf_Rela *rela = NULL, *relalim;
Elf_Addr relasz = 0;
Elf_Addr *where;
/*
* Extract the rela/relasz values from the dynamic section
*/
for (; dynp->d_tag != DT_NULL; dynp++) {
switch (dynp->d_tag) {
case DT_RELA:
rela = (Elf_Rela *)(relocbase+dynp->d_un.d_ptr);
break;
case DT_RELASZ:
relasz = dynp->d_un.d_val;
break;
}
}
/*
* Relocate these values
*/
relalim = (Elf_Rela *)((caddr_t)rela + relasz);
for (; rela < relalim; rela++) {
if (ELF_R_TYPE(rela->r_info) != R_PPC_RELATIVE)
continue;
where = (Elf_Addr *)(relocbase + rela->r_offset);
*where = (Elf_Addr)(relocbase + rela->r_addend);
}
}
int
elf_reloc(linker_file_t lf, Elf_Addr relocbase, const void *data, int type,
elf_lookup_fn lookup)
{
return (elf_reloc_internal(lf, relocbase, data, type, 0, lookup));
}
int
elf_reloc_local(linker_file_t lf, Elf_Addr relocbase, const void *data,
int type, elf_lookup_fn lookup)
{
return (elf_reloc_internal(lf, relocbase, data, type, 1, lookup));
}
int
elf_cpu_load_file(linker_file_t lf)
{
/* Only sync the cache for non-kernel modules */
if (lf->id != 1)
__syncicache(lf->address, lf->size);
return (0);
}
int
elf_cpu_unload_file(linker_file_t lf __unused)
{
return (0);
}
static void
ppc32_runtime_resolve()
{
/*
* Since we don't support lazy binding, panic immediately if anyone
* manages to call the runtime resolver.
*/
panic("kldload: Runtime resolver was called unexpectedly!");
}
int
elf_cpu_parse_dynamic(caddr_t loadbase, Elf_Dyn *dynamic)
{
Elf_Dyn *dp;
bool has_plt = false;
bool secure_plt = false;
Elf_Addr *got;
for (dp = dynamic; dp->d_tag != DT_NULL; dp++) {
switch (dp->d_tag) {
case DT_PPC_GOT:
secure_plt = true;
got = (Elf_Addr *)(loadbase + dp->d_un.d_ptr);
/* Install runtime resolver canary. */
got[1] = (Elf_Addr)ppc32_runtime_resolve;
got[2] = (Elf_Addr)0;
break;
case DT_PLTGOT:
has_plt = true;
break;
}
}
if (has_plt && !secure_plt) {
printf("kldload: BSS-PLT modules are not supported.\n");
return (-1);
}
return (0);
}
#endif
#ifdef __powerpc64__
static void
ppc32_fixlimit(struct rlimit *rl, int which)
{
switch (which) {
case RLIMIT_DATA:
if (ppc32_maxdsiz != 0) {
if (rl->rlim_cur > ppc32_maxdsiz)
rl->rlim_cur = ppc32_maxdsiz;
if (rl->rlim_max > ppc32_maxdsiz)
rl->rlim_max = ppc32_maxdsiz;
}
break;
case RLIMIT_STACK:
if (ppc32_maxssiz != 0) {
if (rl->rlim_cur > ppc32_maxssiz)
rl->rlim_cur = ppc32_maxssiz;
if (rl->rlim_max > ppc32_maxssiz)
rl->rlim_max = ppc32_maxssiz;
}
break;
}
}
#endif
diff --git a/sys/powerpc/powerpc/elf64_machdep.c b/sys/powerpc/powerpc/elf64_machdep.c
index 1bca857d9ca5..38f76c158793 100644
--- a/sys/powerpc/powerpc/elf64_machdep.c
+++ b/sys/powerpc/powerpc/elf64_machdep.c
@@ -1,449 +1,455 @@
/*-
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright 1996-1998 John D. Polstra.
* 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 ``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.
*
* $FreeBSD$
*/
#include <sys/param.h>
#include <sys/kernel.h>
#include <sys/systm.h>
#include <sys/exec.h>
#include <sys/imgact.h>
#include <sys/malloc.h>
#include <sys/proc.h>
#include <sys/namei.h>
#include <sys/fcntl.h>
#include <sys/sysent.h>
#include <sys/imgact_elf.h>
#include <sys/jail.h>
#include <sys/smp.h>
#include <sys/syscall.h>
#include <sys/signalvar.h>
#include <sys/vnode.h>
#include <sys/linker.h>
#include <vm/vm.h>
#include <vm/vm_param.h>
#include <machine/altivec.h>
#include <machine/cpu.h>
#include <machine/fpu.h>
#include <machine/elf.h>
#include <machine/md_var.h>
#include <powerpc/powerpc/elf_common.c>
static void exec_setregs_funcdesc(struct thread *td, struct image_params *imgp,
uintptr_t stack);
struct sysentvec elf64_freebsd_sysvec_v1 = {
.sv_size = SYS_MAXSYSCALL,
.sv_table = sysent,
.sv_transtrap = NULL,
.sv_fixup = __elfN(freebsd_fixup),
.sv_sendsig = sendsig,
.sv_sigcode = sigcode64,
.sv_szsigcode = &szsigcode64,
.sv_name = "FreeBSD ELF64",
.sv_coredump = __elfN(coredump),
+ .sv_elf_core_osabi = ELFOSABI_FREEBSD,
+ .sv_elf_core_abi_vendor = FREEBSD_ABI_VENDOR,
+ .sv_elf_core_prepare_notes = __elfN(prepare_notes),
.sv_imgact_try = NULL,
.sv_minsigstksz = MINSIGSTKSZ,
.sv_minuser = VM_MIN_ADDRESS,
.sv_maxuser = VM_MAXUSER_ADDRESS,
.sv_usrstack = USRSTACK,
.sv_psstrings = PS_STRINGS,
.sv_stackprot = VM_PROT_ALL,
.sv_copyout_auxargs = __elfN(powerpc_copyout_auxargs),
.sv_copyout_strings = exec_copyout_strings,
.sv_setregs = exec_setregs_funcdesc,
.sv_fixlimit = NULL,
.sv_maxssiz = NULL,
.sv_flags = SV_ABI_FREEBSD | SV_LP64 | SV_SHP | SV_ASLR |
SV_TIMEKEEP | SV_RNG_SEED_VER,
.sv_set_syscall_retval = cpu_set_syscall_retval,
.sv_fetch_syscall_args = cpu_fetch_syscall_args,
.sv_syscallnames = syscallnames,
.sv_shared_page_base = SHAREDPAGE,
.sv_shared_page_len = PAGE_SIZE,
.sv_schedtail = NULL,
.sv_thread_detach = NULL,
.sv_trap = NULL,
.sv_hwcap = &cpu_features,
.sv_hwcap2 = &cpu_features2,
};
struct sysentvec elf64_freebsd_sysvec_v2 = {
.sv_size = SYS_MAXSYSCALL,
.sv_table = sysent,
.sv_transtrap = NULL,
.sv_fixup = __elfN(freebsd_fixup),
.sv_sendsig = sendsig,
.sv_sigcode = sigcode64, /* Fixed up in ppc64_init_sysvecs(). */
.sv_szsigcode = &szsigcode64,
.sv_name = "FreeBSD ELF64 V2",
.sv_coredump = __elfN(coredump),
+ .sv_elf_core_osabi = ELFOSABI_FREEBSD,
+ .sv_elf_core_abi_vendor = FREEBSD_ABI_VENDOR,
+ .sv_elf_core_prepare_notes = __elfN(prepare_notes),
.sv_imgact_try = NULL,
.sv_minsigstksz = MINSIGSTKSZ,
.sv_minuser = VM_MIN_ADDRESS,
.sv_maxuser = VM_MAXUSER_ADDRESS,
.sv_usrstack = USRSTACK,
.sv_psstrings = PS_STRINGS,
.sv_stackprot = VM_PROT_ALL,
.sv_copyout_auxargs = __elfN(powerpc_copyout_auxargs),
.sv_copyout_strings = exec_copyout_strings,
.sv_setregs = exec_setregs,
.sv_fixlimit = NULL,
.sv_maxssiz = NULL,
.sv_flags = SV_ABI_FREEBSD | SV_LP64 | SV_SHP |
SV_TIMEKEEP | SV_RNG_SEED_VER,
.sv_set_syscall_retval = cpu_set_syscall_retval,
.sv_fetch_syscall_args = cpu_fetch_syscall_args,
.sv_syscallnames = syscallnames,
.sv_shared_page_base = SHAREDPAGE,
.sv_shared_page_len = PAGE_SIZE,
.sv_schedtail = NULL,
.sv_thread_detach = NULL,
.sv_trap = NULL,
.sv_hwcap = &cpu_features,
.sv_hwcap2 = &cpu_features2,
};
static boolean_t ppc64_elfv1_header_match(struct image_params *params,
int32_t *, uint32_t *);
static boolean_t ppc64_elfv2_header_match(struct image_params *params,
int32_t *, uint32_t *);
static Elf64_Brandinfo freebsd_brand_info_elfv1 = {
.brand = ELFOSABI_FREEBSD,
.machine = EM_PPC64,
.compat_3_brand = "FreeBSD",
.emul_path = NULL,
.interp_path = "/libexec/ld-elf.so.1",
.sysvec = &elf64_freebsd_sysvec_v1,
.interp_newpath = NULL,
.brand_note = &elf64_freebsd_brandnote,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE,
.header_supported = &ppc64_elfv1_header_match
};
SYSINIT(elf64v1, SI_SUB_EXEC, SI_ORDER_ANY,
(sysinit_cfunc_t) elf64_insert_brand_entry,
&freebsd_brand_info_elfv1);
static Elf64_Brandinfo freebsd_brand_info_elfv2 = {
.brand = ELFOSABI_FREEBSD,
.machine = EM_PPC64,
.compat_3_brand = "FreeBSD",
.emul_path = NULL,
.interp_path = "/libexec/ld-elf.so.1",
.sysvec = &elf64_freebsd_sysvec_v2,
.interp_newpath = NULL,
.brand_note = &elf64_freebsd_brandnote,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE,
.header_supported = &ppc64_elfv2_header_match
};
SYSINIT(elf64v2, SI_SUB_EXEC, SI_ORDER_ANY,
(sysinit_cfunc_t) elf64_insert_brand_entry,
&freebsd_brand_info_elfv2);
static Elf64_Brandinfo freebsd_brand_oinfo = {
.brand = ELFOSABI_FREEBSD,
.machine = EM_PPC64,
.compat_3_brand = "FreeBSD",
.emul_path = NULL,
.interp_path = "/usr/libexec/ld-elf.so.1",
.sysvec = &elf64_freebsd_sysvec_v1,
.interp_newpath = NULL,
.brand_note = &elf64_freebsd_brandnote,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE,
.header_supported = &ppc64_elfv1_header_match
};
SYSINIT(oelf64, SI_SUB_EXEC, SI_ORDER_ANY,
(sysinit_cfunc_t) elf64_insert_brand_entry,
&freebsd_brand_oinfo);
void elf_reloc_self(Elf_Dyn *dynp, Elf_Addr relocbase);
static void
ppc64_init_sysvecs(void *arg)
{
exec_sysvec_init(&elf64_freebsd_sysvec_v2);
exec_sysvec_init_secondary(&elf64_freebsd_sysvec_v2,
&elf64_freebsd_sysvec_v1);
/*
* Adjust elfv2 sigcode after elfv1 sysvec is initialized.
* exec_sysvec_init_secondary() assumes secondary sysvecs use
* identical signal code, and skips allocating a second copy.
* Since the ELFv2 trampoline is a strict subset of the ELFv1 code,
* we can work around this by adjusting the base address. This also
* avoids two copies of the trampoline code being allocated!
*/
elf64_freebsd_sysvec_v2.sv_sigcode_base +=
(uintptr_t)sigcode64_elfv2 - (uintptr_t)&sigcode64;
elf64_freebsd_sysvec_v2.sv_szsigcode = &szsigcode64_elfv2;
}
SYSINIT(elf64_sysvec, SI_SUB_EXEC, SI_ORDER_ANY, ppc64_init_sysvecs, NULL);
static boolean_t
ppc64_elfv1_header_match(struct image_params *params, int32_t *osrel __unused,
uint32_t *fctl0 __unused)
{
const Elf64_Ehdr *hdr = (const Elf64_Ehdr *)params->image_header;
int abi = (hdr->e_flags & 3);
return (abi == 0 || abi == 1);
}
static boolean_t
ppc64_elfv2_header_match(struct image_params *params, int32_t *osrel __unused,
uint32_t *fctl0 __unused)
{
const Elf64_Ehdr *hdr = (const Elf64_Ehdr *)params->image_header;
int abi = (hdr->e_flags & 3);
return (abi == 2);
}
static void
exec_setregs_funcdesc(struct thread *td, struct image_params *imgp,
uintptr_t stack)
{
struct trapframe *tf;
register_t entry_desc[3];
tf = trapframe(td);
exec_setregs(td, imgp, stack);
/*
* For 64-bit ELFv1, we need to disentangle the function
* descriptor
*
* 0. entry point
* 1. TOC value (r2)
* 2. Environment pointer (r11)
*/
(void)copyin((void *)imgp->entry_addr, entry_desc,
sizeof(entry_desc));
tf->srr0 = entry_desc[0] + imgp->reloc_base;
tf->fixreg[2] = entry_desc[1] + imgp->reloc_base;
tf->fixreg[11] = entry_desc[2] + imgp->reloc_base;
}
void
elf64_dump_thread(struct thread *td, void *dst, size_t *off)
{
size_t len;
struct pcb *pcb;
uint64_t vshr[32];
uint64_t *vsr_dw1;
int vsr_idx;
len = 0;
pcb = td->td_pcb;
if (pcb->pcb_flags & PCB_VEC) {
save_vec_nodrop(td);
if (dst != NULL) {
len += elf64_populate_note(NT_PPC_VMX,
&pcb->pcb_vec, (char *)dst + len,
sizeof(pcb->pcb_vec), NULL);
} else
len += elf64_populate_note(NT_PPC_VMX, NULL, NULL,
sizeof(pcb->pcb_vec), NULL);
}
if (pcb->pcb_flags & PCB_VSX) {
save_fpu_nodrop(td);
if (dst != NULL) {
/*
* Doubleword 0 of VSR0-VSR31 overlap with FPR0-FPR31 and
* VSR32-VSR63 overlap with VR0-VR31, so we only copy
* the non-overlapping data, which is doubleword 1 of VSR0-VSR31.
*/
for (vsr_idx = 0; vsr_idx < nitems(vshr); vsr_idx++) {
vsr_dw1 = (uint64_t *)&pcb->pcb_fpu.fpr[vsr_idx].vsr[2];
vshr[vsr_idx] = *vsr_dw1;
}
len += elf64_populate_note(NT_PPC_VSX,
vshr, (char *)dst + len,
sizeof(vshr), NULL);
} else
len += elf64_populate_note(NT_PPC_VSX, NULL, NULL,
sizeof(vshr), NULL);
}
*off = len;
}
bool
elf_is_ifunc_reloc(Elf_Size r_info)
{
return (ELF_R_TYPE(r_info) == R_PPC_IRELATIVE);
}
/* Process one elf relocation with addend. */
static int
elf_reloc_internal(linker_file_t lf, Elf_Addr relocbase, const void *data,
int type, int local, elf_lookup_fn lookup)
{
Elf_Addr *where;
Elf_Addr addr;
Elf_Addr addend, val;
Elf_Word rtype, symidx;
const Elf_Rela *rela;
int error;
switch (type) {
case ELF_RELOC_REL:
panic("PPC only supports RELA relocations");
break;
case ELF_RELOC_RELA:
rela = (const Elf_Rela *)data;
where = (Elf_Addr *) (relocbase + rela->r_offset);
addend = rela->r_addend;
rtype = ELF_R_TYPE(rela->r_info);
symidx = ELF_R_SYM(rela->r_info);
break;
default:
panic("elf_reloc: unknown relocation mode %d\n", type);
}
switch (rtype) {
case R_PPC_NONE:
break;
case R_PPC64_ADDR64: /* doubleword64 S + A */
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
addr += addend;
*where = addr;
break;
case R_PPC_RELATIVE: /* doubleword64 B + A */
*where = elf_relocaddr(lf, relocbase + addend);
break;
case R_PPC_JMP_SLOT: /* function descriptor copy */
lookup(lf, symidx, 1, &addr);
#if !defined(_CALL_ELF) || _CALL_ELF == 1
memcpy(where, (Elf_Addr *)addr, 3*sizeof(Elf_Addr));
#else
*where = addr;
#endif
__asm __volatile("dcbst 0,%0; sync" :: "r"(where) : "memory");
break;
case R_PPC_IRELATIVE:
addr = relocbase + addend;
val = ((Elf64_Addr (*)(void))addr)();
if (*where != val)
*where = val;
break;
default:
printf("kldload: unexpected relocation type %d, "
"symbol index %d\n", (int)rtype, symidx);
return (-1);
}
return (0);
}
void
elf_reloc_self(Elf_Dyn *dynp, Elf_Addr relocbase)
{
Elf_Rela *rela = NULL, *relalim;
Elf_Addr relasz = 0;
Elf_Addr *where;
/*
* Extract the rela/relasz values from the dynamic section
*/
for (; dynp->d_tag != DT_NULL; dynp++) {
switch (dynp->d_tag) {
case DT_RELA:
rela = (Elf_Rela *)(relocbase+dynp->d_un.d_ptr);
break;
case DT_RELASZ:
relasz = dynp->d_un.d_val;
break;
}
}
/*
* Relocate these values
*/
relalim = (Elf_Rela *)((caddr_t)rela + relasz);
for (; rela < relalim; rela++) {
if (ELF_R_TYPE(rela->r_info) != R_PPC_RELATIVE)
continue;
where = (Elf_Addr *)(relocbase + rela->r_offset);
*where = (Elf_Addr)(relocbase + rela->r_addend);
}
}
int
elf_reloc(linker_file_t lf, Elf_Addr relocbase, const void *data, int type,
elf_lookup_fn lookup)
{
return (elf_reloc_internal(lf, relocbase, data, type, 0, lookup));
}
int
elf_reloc_local(linker_file_t lf, Elf_Addr relocbase, const void *data,
int type, elf_lookup_fn lookup)
{
return (elf_reloc_internal(lf, relocbase, data, type, 1, lookup));
}
int
elf_cpu_load_file(linker_file_t lf)
{
/* Only sync the cache for non-kernel modules */
if (lf->id != 1)
__syncicache(lf->address, lf->size);
return (0);
}
int
elf_cpu_unload_file(linker_file_t lf __unused)
{
return (0);
}
int
elf_cpu_parse_dynamic(caddr_t loadbase __unused, Elf_Dyn *dynamic __unused)
{
return (0);
}
diff --git a/sys/riscv/riscv/elf_machdep.c b/sys/riscv/riscv/elf_machdep.c
index 06d117128ef7..ee80e04e8c31 100644
--- a/sys/riscv/riscv/elf_machdep.c
+++ b/sys/riscv/riscv/elf_machdep.c
@@ -1,526 +1,529 @@
/*-
* Copyright 1996-1998 John D. Polstra.
* Copyright (c) 2015 Ruslan Bukin <br@bsdpad.com>
* Copyright (c) 2016 Yukishige Shibata <y-shibat@mtd.biglobe.ne.jp>
* All rights reserved.
*
* Portions of this software were developed by SRI International and the
* University of Cambridge Computer Laboratory under DARPA/AFRL contract
* FA8750-10-C-0237 ("CTSRD"), as part of the DARPA CRASH research programme.
*
* Portions of this software were developed by the University of Cambridge
* Computer Laboratory as part of the CTSRD Project, with support from the
* UK Higher Education Innovation Fund (HEIF).
*
* 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 <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/kernel.h>
#include <sys/systm.h>
#include <sys/exec.h>
#include <sys/imgact.h>
#include <sys/linker.h>
#include <sys/proc.h>
#include <sys/sysctl.h>
#include <sys/sysent.h>
#include <sys/imgact_elf.h>
#include <sys/syscall.h>
#include <sys/signalvar.h>
#include <sys/vnode.h>
#include <vm/vm.h>
#include <vm/pmap.h>
#include <vm/vm_param.h>
#include <machine/elf.h>
#include <machine/md_var.h>
static const char *riscv_machine_arch(struct proc *p);
u_long elf_hwcap;
struct sysentvec elf64_freebsd_sysvec = {
.sv_size = SYS_MAXSYSCALL,
.sv_table = sysent,
.sv_transtrap = NULL,
.sv_fixup = __elfN(freebsd_fixup),
.sv_sendsig = sendsig,
.sv_sigcode = sigcode,
.sv_szsigcode = &szsigcode,
.sv_name = "FreeBSD ELF64",
.sv_coredump = __elfN(coredump),
+ .sv_elf_core_osabi = ELFOSABI_FREEBSD,
+ .sv_elf_core_abi_vendor = FREEBSD_ABI_VENDOR,
+ .sv_elf_core_prepare_notes = __elfN(prepare_notes),
.sv_imgact_try = NULL,
.sv_minsigstksz = MINSIGSTKSZ,
.sv_minuser = VM_MIN_ADDRESS,
.sv_maxuser = VM_MAXUSER_ADDRESS,
.sv_usrstack = USRSTACK,
.sv_psstrings = PS_STRINGS,
.sv_stackprot = VM_PROT_READ | VM_PROT_WRITE,
.sv_copyout_auxargs = __elfN(freebsd_copyout_auxargs),
.sv_copyout_strings = exec_copyout_strings,
.sv_setregs = exec_setregs,
.sv_fixlimit = NULL,
.sv_maxssiz = NULL,
.sv_flags = SV_ABI_FREEBSD | SV_LP64 | SV_SHP | SV_ASLR |
SV_RNG_SEED_VER,
.sv_set_syscall_retval = cpu_set_syscall_retval,
.sv_fetch_syscall_args = cpu_fetch_syscall_args,
.sv_syscallnames = syscallnames,
.sv_shared_page_base = SHAREDPAGE,
.sv_shared_page_len = PAGE_SIZE,
.sv_schedtail = NULL,
.sv_thread_detach = NULL,
.sv_trap = NULL,
.sv_hwcap = &elf_hwcap,
.sv_machine_arch = riscv_machine_arch,
};
INIT_SYSENTVEC(elf64_sysvec, &elf64_freebsd_sysvec);
static const char *
riscv_machine_arch(struct proc *p)
{
if ((p->p_elf_flags & EF_RISCV_FLOAT_ABI_MASK) ==
EF_RISCV_FLOAT_ABI_SOFT)
return (MACHINE_ARCH "sf");
return (MACHINE_ARCH);
}
static Elf64_Brandinfo freebsd_brand_info = {
.brand = ELFOSABI_FREEBSD,
.machine = EM_RISCV,
.compat_3_brand = "FreeBSD",
.emul_path = NULL,
.interp_path = "/libexec/ld-elf.so.1",
.sysvec = &elf64_freebsd_sysvec,
.interp_newpath = NULL,
.brand_note = &elf64_freebsd_brandnote,
.flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
};
SYSINIT(elf64, SI_SUB_EXEC, SI_ORDER_FIRST,
(sysinit_cfunc_t)elf64_insert_brand_entry, &freebsd_brand_info);
static bool debug_kld;
SYSCTL_BOOL(_debug, OID_AUTO, kld_reloc, CTLFLAG_RW, &debug_kld, 0,
"Activate debug prints in elf_reloc_internal()");
struct type2str_ent {
int type;
const char *str;
};
void
elf64_dump_thread(struct thread *td, void *dst, size_t *off)
{
}
/*
* Following 4 functions are used to manupilate bits on 32bit interger value.
* FIXME: I implemetend for ease-to-understand rather than for well-optimized.
*/
static uint32_t
gen_bitmask(int msb, int lsb)
{
uint32_t mask;
if (msb == sizeof(mask) * 8 - 1)
mask = ~0;
else
mask = (1U << (msb + 1)) - 1;
if (lsb > 0)
mask &= ~((1U << lsb) - 1);
return (mask);
}
static uint32_t
extract_bits(uint32_t x, int msb, int lsb)
{
uint32_t mask;
mask = gen_bitmask(msb, lsb);
x &= mask;
x >>= lsb;
return (x);
}
static uint32_t
insert_bits(uint32_t d, uint32_t s, int msb, int lsb)
{
uint32_t mask;
mask = gen_bitmask(msb, lsb);
d &= ~mask;
s <<= lsb;
s &= mask;
return (d | s);
}
static uint32_t
insert_imm(uint32_t insn, uint32_t imm, int imm_msb, int imm_lsb,
int insn_lsb)
{
int insn_msb;
uint32_t v;
v = extract_bits(imm, imm_msb, imm_lsb);
insn_msb = (imm_msb - imm_lsb) + insn_lsb;
return (insert_bits(insn, v, insn_msb, insn_lsb));
}
/*
* The RISC-V ISA is designed so that all of immediate values are
* sign-extended.
* An immediate value is sometimes generated at runtime by adding
* 12bit sign integer and 20bit signed integer. This requests 20bit
* immediate value to be ajusted if the MSB of the 12bit immediate
* value is asserted (sign-extended value is treated as negative value).
*
* For example, 0x123800 can be calculated by adding upper 20 bit of
* 0x124000 and sign-extended 12bit immediate whose bit pattern is
* 0x800 as follows:
* 0x123800
* = 0x123000 + 0x800
* = (0x123000 + 0x1000) + (-0x1000 + 0x800)
* = (0x123000 + 0x1000) + (0xff...ff800)
* = 0x124000 + sign-extention(0x800)
*/
static uint32_t
calc_hi20_imm(uint32_t value)
{
/*
* There is the arithmetical hack that can remove conditional
* statement. But I implement it in straightforward way.
*/
if ((value & 0x800) != 0)
value += 0x1000;
return (value & ~0xfff);
}
static const struct type2str_ent t2s[] = {
{ R_RISCV_NONE, "R_RISCV_NONE" },
{ R_RISCV_64, "R_RISCV_64" },
{ R_RISCV_JUMP_SLOT, "R_RISCV_JUMP_SLOT" },
{ R_RISCV_RELATIVE, "R_RISCV_RELATIVE" },
{ R_RISCV_JAL, "R_RISCV_JAL" },
{ R_RISCV_CALL, "R_RISCV_CALL" },
{ R_RISCV_PCREL_HI20, "R_RISCV_PCREL_HI20" },
{ R_RISCV_PCREL_LO12_I, "R_RISCV_PCREL_LO12_I" },
{ R_RISCV_PCREL_LO12_S, "R_RISCV_PCREL_LO12_S" },
{ R_RISCV_HI20, "R_RISCV_HI20" },
{ R_RISCV_LO12_I, "R_RISCV_LO12_I" },
{ R_RISCV_LO12_S, "R_RISCV_LO12_S" },
};
static const char *
reloctype_to_str(int type)
{
int i;
for (i = 0; i < sizeof(t2s) / sizeof(t2s[0]); ++i) {
if (type == t2s[i].type)
return t2s[i].str;
}
return "*unknown*";
}
bool
elf_is_ifunc_reloc(Elf_Size r_info __unused)
{
return (false);
}
/*
* Currently kernel loadable module for RISCV is compiled with -fPIC option.
* (see also additional CFLAGS definition for RISCV in sys/conf/kmod.mk)
* Only R_RISCV_64, R_RISCV_JUMP_SLOT and RISCV_RELATIVE are emitted in
* the module. Other relocations will be processed when kernel loadable
* modules are built in non-PIC.
*
* FIXME: only RISCV64 is supported.
*/
static int
elf_reloc_internal(linker_file_t lf, Elf_Addr relocbase, const void *data,
int type, int local, elf_lookup_fn lookup)
{
Elf_Size rtype, symidx;
const Elf_Rela *rela;
Elf_Addr val, addr;
Elf64_Addr *where;
Elf_Addr addend;
uint32_t before32_1;
uint32_t before32;
uint64_t before64;
uint32_t *insn32p;
uint32_t imm20;
int error;
switch (type) {
case ELF_RELOC_RELA:
rela = (const Elf_Rela *)data;
where = (Elf_Addr *)(relocbase + rela->r_offset);
insn32p = (uint32_t *)where;
addend = rela->r_addend;
rtype = ELF_R_TYPE(rela->r_info);
symidx = ELF_R_SYM(rela->r_info);
break;
default:
printf("%s:%d unknown reloc type %d\n",
__FUNCTION__, __LINE__, type);
return (-1);
}
switch (rtype) {
case R_RISCV_NONE:
break;
case R_RISCV_64:
case R_RISCV_JUMP_SLOT:
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
val = addr;
before64 = *where;
if (*where != val)
*where = val;
if (debug_kld)
printf("%p %c %-24s %016lx -> %016lx\n", where,
(local ? 'l' : 'g'), reloctype_to_str(rtype),
before64, *where);
break;
case R_RISCV_RELATIVE:
before64 = *where;
*where = elf_relocaddr(lf, relocbase + addend);
if (debug_kld)
printf("%p %c %-24s %016lx -> %016lx\n", where,
(local ? 'l' : 'g'), reloctype_to_str(rtype),
before64, *where);
break;
case R_RISCV_JAL:
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
val = addr - (Elf_Addr)where;
if (val <= -(1UL << 20) || (1UL << 20) <= val) {
printf("kldload: huge offset against R_RISCV_JAL\n");
return (-1);
}
before32 = *insn32p;
*insn32p = insert_imm(*insn32p, val, 20, 20, 31);
*insn32p = insert_imm(*insn32p, val, 10, 1, 21);
*insn32p = insert_imm(*insn32p, val, 11, 11, 20);
*insn32p = insert_imm(*insn32p, val, 19, 12, 12);
if (debug_kld)
printf("%p %c %-24s %08x -> %08x\n", where,
(local ? 'l' : 'g'), reloctype_to_str(rtype),
before32, *insn32p);
break;
case R_RISCV_CALL:
/*
* R_RISCV_CALL relocates 8-byte region that consists
* of the sequence of AUIPC and JALR.
*/
/* Calculate and check the pc relative offset. */
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
val = addr - (Elf_Addr)where;
if (val <= -(1UL << 32) || (1UL << 32) <= val) {
printf("kldload: huge offset against R_RISCV_CALL\n");
return (-1);
}
/* Relocate AUIPC. */
before32 = insn32p[0];
imm20 = calc_hi20_imm(val);
insn32p[0] = insert_imm(insn32p[0], imm20, 31, 12, 12);
/* Relocate JALR. */
before32_1 = insn32p[1];
insn32p[1] = insert_imm(insn32p[1], val, 11, 0, 20);
if (debug_kld)
printf("%p %c %-24s %08x %08x -> %08x %08x\n", where,
(local ? 'l' : 'g'), reloctype_to_str(rtype),
before32, insn32p[0], before32_1, insn32p[1]);
break;
case R_RISCV_PCREL_HI20:
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
val = addr - (Elf_Addr)where;
insn32p = (uint32_t *)where;
before32 = *insn32p;
imm20 = calc_hi20_imm(val);
*insn32p = insert_imm(*insn32p, imm20, 31, 12, 12);
if (debug_kld)
printf("%p %c %-24s %08x -> %08x\n", where,
(local ? 'l' : 'g'), reloctype_to_str(rtype),
before32, *insn32p);
break;
case R_RISCV_PCREL_LO12_I:
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
val = addr - (Elf_Addr)where;
insn32p = (uint32_t *)where;
before32 = *insn32p;
*insn32p = insert_imm(*insn32p, addr, 11, 0, 20);
if (debug_kld)
printf("%p %c %-24s %08x -> %08x\n", where,
(local ? 'l' : 'g'), reloctype_to_str(rtype),
before32, *insn32p);
break;
case R_RISCV_PCREL_LO12_S:
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
val = addr - (Elf_Addr)where;
insn32p = (uint32_t *)where;
before32 = *insn32p;
*insn32p = insert_imm(*insn32p, addr, 11, 5, 25);
*insn32p = insert_imm(*insn32p, addr, 4, 0, 7);
if (debug_kld)
printf("%p %c %-24s %08x -> %08x\n", where,
(local ? 'l' : 'g'), reloctype_to_str(rtype),
before32, *insn32p);
break;
case R_RISCV_HI20:
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
val = addr;
insn32p = (uint32_t *)where;
before32 = *insn32p;
imm20 = calc_hi20_imm(val);
*insn32p = insert_imm(*insn32p, imm20, 31, 12, 12);
if (debug_kld)
printf("%p %c %-24s %08x -> %08x\n", where,
(local ? 'l' : 'g'), reloctype_to_str(rtype),
before32, *insn32p);
break;
case R_RISCV_LO12_I:
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
val = addr;
insn32p = (uint32_t *)where;
before32 = *insn32p;
*insn32p = insert_imm(*insn32p, addr, 11, 0, 20);
if (debug_kld)
printf("%p %c %-24s %08x -> %08x\n", where,
(local ? 'l' : 'g'), reloctype_to_str(rtype),
before32, *insn32p);
break;
case R_RISCV_LO12_S:
error = lookup(lf, symidx, 1, &addr);
if (error != 0)
return (-1);
val = addr;
insn32p = (uint32_t *)where;
before32 = *insn32p;
*insn32p = insert_imm(*insn32p, addr, 11, 5, 25);
*insn32p = insert_imm(*insn32p, addr, 4, 0, 7);
if (debug_kld)
printf("%p %c %-24s %08x -> %08x\n", where,
(local ? 'l' : 'g'), reloctype_to_str(rtype),
before32, *insn32p);
break;
default:
printf("kldload: unexpected relocation type %ld, "
"symbol index %ld\n", rtype, symidx);
return (-1);
}
return (0);
}
int
elf_reloc(linker_file_t lf, Elf_Addr relocbase, const void *data, int type,
elf_lookup_fn lookup)
{
return (elf_reloc_internal(lf, relocbase, data, type, 0, lookup));
}
int
elf_reloc_local(linker_file_t lf, Elf_Addr relocbase, const void *data,
int type, elf_lookup_fn lookup)
{
return (elf_reloc_internal(lf, relocbase, data, type, 1, lookup));
}
int
elf_cpu_load_file(linker_file_t lf __unused)
{
return (0);
}
int
elf_cpu_unload_file(linker_file_t lf __unused)
{
return (0);
}
int
elf_cpu_parse_dynamic(caddr_t loadbase __unused, Elf_Dyn *dynamic __unused)
{
return (0);
}
diff --git a/sys/sys/imgact_elf.h b/sys/sys/imgact_elf.h
index f0c112b2bcdc..44a860359a75 100644
--- a/sys/sys/imgact_elf.h
+++ b/sys/sys/imgact_elf.h
@@ -1,128 +1,137 @@
/*-
* SPDX-License-Identifier: BSD-3-Clause
*
* Copyright (c) 1995-1996 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 _SYS_IMGACT_ELF_H_
#define _SYS_IMGACT_ELF_H_
#include <machine/elf.h>
#ifdef _KERNEL
#define AUXARGS_ENTRY(pos, id, val) \
{(pos)->a_type = (id); (pos)->a_un.a_val = (val); (pos)++;}
#if (defined(__LP64__) && __ELF_WORD_SIZE == 32)
#define AUXARGS_ENTRY_PTR(pos, id, ptr) \
{(pos)->a_type = (id); (pos)->a_un.a_val = (uintptr_t)(ptr); (pos)++;}
#else
#define AUXARGS_ENTRY_PTR(pos, id, ptr) \
{(pos)->a_type = (id); (pos)->a_un.a_ptr = (ptr); (pos)++;}
#endif
struct image_params;
struct thread;
struct vnode;
+struct note_info_list;
+struct sbuf;
/*
* Structure used to pass information from the loader to the
* stack fixup routine.
*/
typedef struct {
Elf_Ssize execfd;
Elf_Size phdr;
Elf_Size phent;
Elf_Size phnum;
Elf_Size pagesz;
Elf_Size base;
Elf_Size flags;
Elf_Size entry;
Elf_Word hdr_eflags; /* e_flags field from ehdr */
} __ElfN(Auxargs);
typedef struct {
Elf_Note hdr;
const char * vendor;
int flags;
bool (*trans_osrel)(const Elf_Note *, int32_t *);
#define BN_CAN_FETCH_OSREL 0x0001 /* Deprecated. */
#define BN_TRANSLATE_OSREL 0x0002 /* Use trans_osrel to fetch osrel */
/* after checking the image ABI specification, if needed. */
} Elf_Brandnote;
typedef struct {
int brand;
int machine;
const char *compat_3_brand; /* pre Binutils 2.10 method (FBSD 3) */
const char *emul_path;
const char *interp_path;
struct sysentvec *sysvec;
const char *interp_newpath;
int flags;
Elf_Brandnote *brand_note;
boolean_t (*header_supported)(struct image_params *,
int32_t *, uint32_t *);
#define BI_CAN_EXEC_DYN 0x0001
#define BI_BRAND_NOTE 0x0002 /* May have note.ABI-tag section. */
#define BI_BRAND_NOTE_MANDATORY 0x0004 /* Must have note.ABI-tag section. */
#define BI_BRAND_ONLY_STATIC 0x0008 /* Match only interp-less binaries. */
} __ElfN(Brandinfo);
__ElfType(Auxargs);
__ElfType(Brandinfo);
-#define MAX_BRANDS 8
+#define MAX_BRANDS 8
+#define FREEBSD_ABI_VENDOR "FreeBSD"
+
+typedef void (*outfunc_t)(void *, struct sbuf *, size_t *);
/* Closure for __elfN(size_segments)(). */
struct sseg_closure {
int count; /* Count of writable segments. */
size_t size; /* Total size of all writable segments. */
};
int __elfN(brand_inuse)(Elf_Brandinfo *entry);
int __elfN(insert_brand_entry)(Elf_Brandinfo *entry);
int __elfN(remove_brand_entry)(Elf_Brandinfo *entry);
int __elfN(freebsd_fixup)(uintptr_t *, struct image_params *);
int __elfN(coredump)(struct thread *, struct vnode *, off_t, int);
size_t __elfN(populate_note)(int, void *, void *, size_t, void **);
void __elfN(stackgap)(struct image_params *, uintptr_t *);
int __elfN(freebsd_copyout_auxargs)(struct image_params *, uintptr_t);
void __elfN(puthdr)(struct thread *, void *, size_t, int, size_t, int);
+void __elfN(prepare_notes)(struct thread *, struct note_info_list *,
+ size_t *);
void __elfN(size_segments)(struct thread *, struct sseg_closure *, int);
+size_t __elfN(register_note)(struct thread *, struct note_info_list *,
+ int, outfunc_t, void *);
/* Machine specific function to dump per-thread information. */
void __elfN(dump_thread)(struct thread *, void *, size_t *);
extern int __elfN(fallback_brand);
extern Elf_Brandnote __elfN(freebsd_brandnote);
extern Elf_Brandnote __elfN(kfreebsd_brandnote);
#endif /* _KERNEL */
#endif /* !_SYS_IMGACT_ELF_H_ */
diff --git a/sys/sys/sysent.h b/sys/sys/sysent.h
index 8b0903f7dcc0..e0e5e351ae20 100644
--- a/sys/sys/sysent.h
+++ b/sys/sys/sysent.h
@@ -1,325 +1,330 @@
/*-
* SPDX-License-Identifier: BSD-3-Clause
*
* Copyright (c) 1982, 1988, 1991 The Regents of the University of California.
* 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.
* 3. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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 _SYS_SYSENT_H_
#define _SYS_SYSENT_H_
#include <bsm/audit.h>
struct rlimit;
struct sysent;
struct thread;
struct ksiginfo;
struct syscall_args;
enum systrace_probe_t {
SYSTRACE_ENTRY,
SYSTRACE_RETURN,
};
typedef int sy_call_t(struct thread *, void *);
typedef void (*systrace_probe_func_t)(struct syscall_args *,
enum systrace_probe_t, int);
typedef void (*systrace_args_func_t)(int, void *, uint64_t *, int *);
#ifdef _KERNEL
extern systrace_probe_func_t systrace_probe_func;
extern bool systrace_enabled;
#ifdef KDTRACE_HOOKS
#define SYSTRACE_ENABLED() (systrace_enabled)
#else
#define SYSTRACE_ENABLED() (0)
#endif
#endif /* _KERNEL */
struct sysent { /* system call table */
sy_call_t *sy_call; /* implementing function */
systrace_args_func_t sy_systrace_args_func;
/* optional argument conversion function. */
u_int8_t sy_narg; /* number of arguments */
u_int8_t sy_flags; /* General flags for system calls. */
au_event_t sy_auevent; /* audit event associated with syscall */
u_int32_t sy_entry; /* DTrace entry ID for systrace. */
u_int32_t sy_return; /* DTrace return ID for systrace. */
u_int32_t sy_thrcnt;
};
/*
* A system call is permitted in capability mode.
*/
#define SYF_CAPENABLED 0x00000001
#define SY_THR_FLAGMASK 0x7
#define SY_THR_STATIC 0x1
#define SY_THR_DRAINING 0x2
#define SY_THR_ABSENT 0x4
#define SY_THR_INCR 0x8
#ifdef KLD_MODULE
#define SY_THR_STATIC_KLD 0
#else
#define SY_THR_STATIC_KLD SY_THR_STATIC
#endif
struct image_params;
struct proc;
struct __sigset;
struct trapframe;
struct vnode;
+struct note_info_list;
struct sysentvec {
int sv_size; /* number of entries */
struct sysent *sv_table; /* pointer to sysent */
int (*sv_transtrap)(int, int);
/* translate trap-to-signal mapping */
int (*sv_fixup)(uintptr_t *, struct image_params *);
/* stack fixup function */
void (*sv_sendsig)(void (*)(int), struct ksiginfo *, struct __sigset *);
/* send signal */
char *sv_sigcode; /* start of sigtramp code */
int *sv_szsigcode; /* size of sigtramp code */
char *sv_name; /* name of binary type */
int (*sv_coredump)(struct thread *, struct vnode *, off_t, int);
/* function to dump core, or NULL */
+ int sv_elf_core_osabi;
+ const char *sv_elf_core_abi_vendor;
+ void (*sv_elf_core_prepare_notes)(struct thread *,
+ struct note_info_list *, size_t *);
int (*sv_imgact_try)(struct image_params *);
void (*sv_stackgap)(struct image_params *, uintptr_t *);
int (*sv_copyout_auxargs)(struct image_params *,
uintptr_t);
int sv_minsigstksz; /* minimum signal stack size */
vm_offset_t sv_minuser; /* VM_MIN_ADDRESS */
vm_offset_t sv_maxuser; /* VM_MAXUSER_ADDRESS */
vm_offset_t sv_usrstack; /* USRSTACK */
vm_offset_t sv_psstrings; /* PS_STRINGS */
int sv_stackprot; /* vm protection for stack */
int (*sv_copyout_strings)(struct image_params *,
uintptr_t *);
void (*sv_setregs)(struct thread *, struct image_params *,
uintptr_t);
void (*sv_fixlimit)(struct rlimit *, int);
u_long *sv_maxssiz;
u_int sv_flags;
void (*sv_set_syscall_retval)(struct thread *, int);
int (*sv_fetch_syscall_args)(struct thread *);
const char **sv_syscallnames;
vm_offset_t sv_timekeep_base;
vm_offset_t sv_shared_page_base;
vm_offset_t sv_shared_page_len;
vm_offset_t sv_sigcode_base;
void *sv_shared_page_obj;
void (*sv_schedtail)(struct thread *);
void (*sv_thread_detach)(struct thread *);
int (*sv_trap)(struct thread *);
u_long *sv_hwcap; /* Value passed in AT_HWCAP. */
u_long *sv_hwcap2; /* Value passed in AT_HWCAP2. */
const char *(*sv_machine_arch)(struct proc *);
vm_offset_t sv_fxrng_gen_base;
void (*sv_onexec)(struct proc *, struct image_params *);
void (*sv_onexit)(struct proc *);
void (*sv_ontdexit)(struct thread *td);
int (*sv_setid_allowed)(struct thread *td,
struct image_params *imgp);
};
#define SV_ILP32 0x000100 /* 32-bit executable. */
#define SV_LP64 0x000200 /* 64-bit executable. */
#define SV_IA32 0x004000 /* Intel 32-bit executable. */
#define SV_AOUT 0x008000 /* a.out executable. */
#define SV_SHP 0x010000 /* Shared page. */
#define SV_CAPSICUM 0x020000 /* Force cap_enter() on startup. */
#define SV_TIMEKEEP 0x040000 /* Shared page timehands. */
#define SV_ASLR 0x080000 /* ASLR allowed. */
#define SV_RNG_SEED_VER 0x100000 /* random(4) reseed generation. */
#define SV_SIG_DISCIGN 0x200000 /* Do not discard ignored signals */
#define SV_SIG_WAITNDQ 0x400000 /* Wait does not dequeue SIGCHLD */
#define SV_ABI_MASK 0xff
#define SV_PROC_FLAG(p, x) ((p)->p_sysent->sv_flags & (x))
#define SV_PROC_ABI(p) ((p)->p_sysent->sv_flags & SV_ABI_MASK)
#define SV_CURPROC_FLAG(x) SV_PROC_FLAG(curproc, x)
#define SV_CURPROC_ABI() SV_PROC_ABI(curproc)
/* same as ELFOSABI_XXX, to prevent header pollution */
#define SV_ABI_LINUX 3
#define SV_ABI_FREEBSD 9
#define SV_ABI_CLOUDABI 17
#define SV_ABI_UNDEF 255
/* sv_coredump flags */
#define SVC_PT_COREDUMP 0x00000001 /* dump requested by ptrace(2) */
#define SVC_NOCOMPRESS 0x00000002 /* disable compression. */
#define SVC_ALL 0x00000004 /* dump everything */
#ifdef _KERNEL
extern struct sysentvec aout_sysvec;
extern struct sysent sysent[];
extern const char *syscallnames[];
#define NO_SYSCALL (-1)
struct module;
struct syscall_module_data {
int (*chainevh)(struct module *, int, void *); /* next handler */
void *chainarg; /* arg for next event handler */
int *offset; /* offset into sysent */
struct sysent *new_sysent; /* new sysent */
struct sysent old_sysent; /* old sysent */
int flags; /* flags for syscall_register */
};
/* separate initialization vector so it can be used in a substructure */
#define SYSENT_INIT_VALS(_syscallname) { \
.sy_narg = (sizeof(struct _syscallname ## _args ) \
/ sizeof(register_t)), \
.sy_call = (sy_call_t *)&sys_##_syscallname, \
.sy_auevent = SYS_AUE_##_syscallname, \
.sy_systrace_args_func = NULL, \
.sy_entry = 0, \
.sy_return = 0, \
.sy_flags = 0, \
.sy_thrcnt = 0 \
}
#define MAKE_SYSENT(syscallname) \
static struct sysent syscallname##_sysent = SYSENT_INIT_VALS(syscallname);
#define MAKE_SYSENT_COMPAT(syscallname) \
static struct sysent syscallname##_sysent = { \
(sizeof(struct syscallname ## _args ) \
/ sizeof(register_t)), \
(sy_call_t *)& syscallname, \
SYS_AUE_##syscallname \
}
#define SYSCALL_MODULE(name, offset, new_sysent, evh, arg) \
static struct syscall_module_data name##_syscall_mod = { \
evh, arg, offset, new_sysent, { 0, NULL, AUE_NULL } \
}; \
\
static moduledata_t name##_mod = { \
"sys/" #name, \
syscall_module_handler, \
&name##_syscall_mod \
}; \
DECLARE_MODULE(name, name##_mod, SI_SUB_SYSCALLS, SI_ORDER_MIDDLE)
#define SYSCALL_MODULE_HELPER(syscallname) \
static int syscallname##_syscall = SYS_##syscallname; \
MAKE_SYSENT(syscallname); \
SYSCALL_MODULE(syscallname, \
& syscallname##_syscall, & syscallname##_sysent, \
NULL, NULL)
#define SYSCALL_MODULE_PRESENT(syscallname) \
(sysent[SYS_##syscallname].sy_call != (sy_call_t *)lkmnosys && \
sysent[SYS_##syscallname].sy_call != (sy_call_t *)lkmressys)
/*
* Syscall registration helpers with resource allocation handling.
*/
struct syscall_helper_data {
struct sysent new_sysent;
struct sysent old_sysent;
int syscall_no;
int registered;
};
#define SYSCALL_INIT_HELPER_F(syscallname, flags) { \
.new_sysent = { \
.sy_narg = (sizeof(struct syscallname ## _args ) \
/ sizeof(register_t)), \
.sy_call = (sy_call_t *)& sys_ ## syscallname, \
.sy_auevent = SYS_AUE_##syscallname, \
.sy_flags = (flags) \
}, \
.syscall_no = SYS_##syscallname \
}
#define SYSCALL_INIT_HELPER_COMPAT_F(syscallname, flags) { \
.new_sysent = { \
.sy_narg = (sizeof(struct syscallname ## _args ) \
/ sizeof(register_t)), \
.sy_call = (sy_call_t *)& syscallname, \
.sy_auevent = SYS_AUE_##syscallname, \
.sy_flags = (flags) \
}, \
.syscall_no = SYS_##syscallname \
}
#define SYSCALL_INIT_HELPER(syscallname) \
SYSCALL_INIT_HELPER_F(syscallname, 0)
#define SYSCALL_INIT_HELPER_COMPAT(syscallname) \
SYSCALL_INIT_HELPER_COMPAT_F(syscallname, 0)
#define SYSCALL_INIT_LAST { \
.syscall_no = NO_SYSCALL \
}
int syscall_module_handler(struct module *mod, int what, void *arg);
int syscall_helper_register(struct syscall_helper_data *sd, int flags);
int syscall_helper_unregister(struct syscall_helper_data *sd);
/* Implementation, exposed for COMPAT code */
int kern_syscall_register(struct sysent *sysents, int *offset,
struct sysent *new_sysent, struct sysent *old_sysent, int flags);
int kern_syscall_deregister(struct sysent *sysents, int offset,
const struct sysent *old_sysent);
int kern_syscall_module_handler(struct sysent *sysents,
struct module *mod, int what, void *arg);
int kern_syscall_helper_register(struct sysent *sysents,
struct syscall_helper_data *sd, int flags);
int kern_syscall_helper_unregister(struct sysent *sysents,
struct syscall_helper_data *sd);
struct proc;
const char *syscallname(struct proc *p, u_int code);
/* Special purpose system call functions. */
struct nosys_args;
int lkmnosys(struct thread *, struct nosys_args *);
int lkmressys(struct thread *, struct nosys_args *);
int syscall_thread_enter(struct thread *td, struct sysent *se);
void syscall_thread_exit(struct thread *td, struct sysent *se);
int shared_page_alloc(int size, int align);
int shared_page_fill(int size, int align, const void *data);
void shared_page_write(int base, int size, const void *data);
void exec_sysvec_init(void *param);
void exec_sysvec_init_secondary(struct sysentvec *sv, struct sysentvec *sv2);
void exec_inittk(void);
#define INIT_SYSENTVEC(name, sv) \
SYSINIT(name, SI_SUB_EXEC, SI_ORDER_ANY, \
(sysinit_cfunc_t)exec_sysvec_init, sv);
#endif /* _KERNEL */
#endif /* !_SYS_SYSENT_H_ */