baremetal aarch64: create C version of multicore.S as well

Attempted to do the same for arm, but it failed.
This commit is contained in:
Ciro Santilli 六四事件 法轮功
2019-08-21 00:00:00 +00:00
parent 6f88fa17bc
commit f28191a735
21 changed files with 336 additions and 89 deletions

View File

@@ -15470,15 +15470,23 @@ Semihosting is implemented both on some real devices and on simulators such as Q
It is documented at: https://developer.arm.com/docs/100863/latest/introduction
For example, the following code makes QEMU exit:
For example, all the following code make QEMU exit:
....
./run --arch arm --baremetal baremetal/arch/arm/semihost_exit.S
./run --arch arm --baremetal baremetal/arch/arm/no_bootloader/semihost_exit.S
./run --arch aarch64 --baremetal baremetal/arch/aarch64/semihost_exit.S
./run --arch aarch64 --baremetal baremetal/arch/aarch64/no_bootloader/semihost_exit.S
....
Source: link:baremetal/arch/arm/no_bootloader/semihost_exit.S[]
Sources:
That program program contains the code:
* link:baremetal/arch/arm/semihost_exit.S[]
* link:baremetal/arch/arm/no_bootloader/semihost_exit.S[]
* link:baremetal/arch/aarch64/semihost_exit.S[]
* link:baremetal/arch/aarch64/no_bootloader/semihost_exit.S[]
That `arm` program program contains the code:
....
mov r0, #0x18
@@ -16171,17 +16179,26 @@ See the example at: xref:arm-svc-instruction[xrefstyle=full]
==== ARM multicore
Examples:
....
./run --arch aarch64 --baremetal baremetal/arch/aarch64/multicore.S --cpus 2
./run --arch aarch64 --baremetal baremetal/arch/aarch64/multicore.S --cpus 2 --emulator gem5
./run --arch arm --baremetal baremetal/arch/aarch64/multicore.S --cpus 2
./run --arch arm --baremetal baremetal/arch/aarch64/multicore.S --cpus 2 --emulator gem5
./run --arch aarch64 --baremetal baremetal/arch/aarch64/no_bootloader/multicore_asm.S --cpus 2
./run --arch aarch64 --baremetal baremetal/arch/aarch64/no_bootloader/multicore_asm.S --cpus 2 --emulator gem5
./run --arch aarch64 --baremetal baremetal/arch/aarch64/multicore.c --cpus 2
./run --arch aarch64 --baremetal baremetal/arch/aarch64/multicore.c --cpus 2 --emulator gem5
./run --arch arm --baremetal baremetal/arch/arm/no_bootloader/multicore_asm.S --cpus 2
./run --arch arm --baremetal baremetal/arch/arm/no_bootloader/multicore_asm.S --cpus 2 --emulator gem5
# TODO not working, hangs.
# ./run --arch arm --baremetal baremetal/arch/arm/multicore.c --cpus 2
./run --arch arm --baremetal baremetal/arch/arm/multicore.c --cpus 2 --emulator gem5
....
Sources:
* link:baremetal/arch/aarch64/multicore.S[]
* link:baremetal/arch/arm/multicore.S[]
* link:baremetal/arch/aarch64/no_bootloader/multicore_asm.S[]
* link:baremetal/arch/aarch64/multicore.c[]
* link:baremetal/arch/arm/no_bootloader/multicore_asm.S[]
* link:baremetal/arch/arm/multicore.c[]
CPU 0 of this program enters a spinlock loop: it repeatedly checks if a given memory address is 1.
@@ -16190,7 +16207,7 @@ So, we need CPU 1 to come to the rescue and set that memory address to 1, otherw
Don't believe me? Then try:
....
./run --arch aarch64 --baremetal baremetal/arch/aarch64/multicore.S --cpus 1
./run --arch aarch64 --baremetal baremetal/arch/aarch64/multicore.c --cpus 1
....
and watch it hang forever.
@@ -16198,7 +16215,7 @@ and watch it hang forever.
Note that if you try the same thing on gem5:
....
./run --arch aarch64 --baremetal baremetal/arch/aarch64/multicore.S --cpus 1 --emulator gem5
./run --arch aarch64 --baremetal baremetal/arch/aarch64/multicore.c --cpus 1 --emulator gem5
....
then the gem5 actually exits with <<gem5-simulate-limit-reached>> as opposed to the expected:

View File

@@ -0,0 +1,45 @@
/* https://cirosantilli.com/linux-kernel-module-cheat#arm-multicore
*
* Beware: things will blow up if the stack for CPU0 grow too much and
* reaches that of CPU1. This is why it is so hard to do multithreading
* without an OS that manages paging.
*/
#include <lkmc.h>
uint64_t spinlock = 0;
__asm__(
".text\n"
".global lkmc_cpu_not_0\n"
"lkmc_cpu_not_0:\n"
/* Put all CPUs except CPU1 to sleep. */
" cmp x0, 1\n"
" bne .Lsleep_forever\n"
/* Prepare the stack for CPU1. This is what we need
* this assembly function for. */
" ldr x0, =(stack_top - 0x1000)\n"
" mov sp, x0\n"
" bl main_cpu1\n"
".Lsleep_forever:\n"
" wfe\n"
" b .Lsleep_forever\n"
);
static void main_cpu1(void) {
spinlock = 1;
lkmc_arm_aarch64_dmb(sy);
lkmc_arm_aarch64_sev();
while (1) {
lkmc_arm_aarch64_wfe();
}
}
int main(void) {
#if !LKMC_GEM5
lkmc_aarch64_psci_cpu_on(1, (uint64_t)main_cpu1, 0);
#endif
while (!spinlock) {
lkmc_arm_aarch64_wfe();
}
}

View File

@@ -0,0 +1,8 @@
/* Test _exit. */
#include <lkmc.h>
.global lkmc_start
lkmc_start:
mov x0, 0
bl _exit

View File

@@ -1,11 +1,15 @@
/* https://cirosantilli.com/linux-kernel-module-cheat#arm-multicore */
/* https://cirosantilli.com/linux-kernel-module-cheat#arm-multicore
*
* This has to be in no_bootloader
*/
#include <lkmc.h>
LKMC_PROLOGUE
.global lkmc_start
lkmc_start:
/* Reset spinlock. */
mov x0, 0
ldr x1, =spinlock
ldr x1, =.Lspinlock
str x0, [x1]
/* Read cpu id into x1.
@@ -14,11 +18,11 @@ LKMC_PROLOGUE
*/
mrs x1, mpidr_el1
ands x1, x1, 3
beq cpu0_only
beq .Lcpu0_only
.Lcpu1_only:
/* Only CPU 1 reaches this point and sets the spinlock. */
mov x0, 1
ldr x1, =spinlock
ldr x1, =.Lspinlock
str x0, [x1]
/* Ensure that CPU 0 sees the write right now.
* Optional, but could save some useless CPU 1 loops.
@@ -34,7 +38,7 @@ LKMC_PROLOGUE
*/
wfe
b .Lcpu1_sleep_forever
cpu0_only:
.Lcpu0_only:
/* Only CPU 0 reaches this point. */
#if !LKMC_GEM5
@@ -61,11 +65,14 @@ cpu0_only:
hvc 0
#endif
spinlock_start:
ldr x0, spinlock
.Lspinlock_start:
ldr x0, .Lspinlock
/* Hint CPU 0 to enter low power mode. */
wfe
cbz x0, spinlock_start
LKMC_EPILOGUE
spinlock:
cbz x0, .Lspinlock_start
mov x0, 0
bl _exit
.Lspinlock:
.skip 8

View File

@@ -1,15 +1,19 @@
/* https://cirosantilli.com/linux-kernel-module-cheat#semihosting */
/* https://cirosantilli.com/linux-kernel-module-cheat#semihosting
*
* Since our stack pointer is not setup, we justa allocate a memory
* region to contain the semihosting arguments, which must be in memory.
*/
.global lkmc_start
lkmc_start:
mov x1, 0x26
movk x1, 2, lsl 16
ldr x2, =semihost_args
ldr x2, =.Lsemihost_args
str x1, [x2, 0]
mov x0, 0
str x0, [x2, 8]
mov x1, x2
mov w0, 0x18
hlt 0xf000
semihost_args:
.Lsemihost_args:
.skip 16

View File

@@ -1,3 +1,5 @@
/* https://cirosantilli.com/linux-kernel-module-cheat#semihosting */
.global main
main:
/* 0x20026 == ADP_Stopped_ApplicationExit */

View File

@@ -71,7 +71,7 @@ int main(void) {
printf("&after_svc %p\n", &&after_svc);
assert(myvar == 0);
/* Max 16-bits. */
lkmc_svc(0xABCD);
lkmc_arm_aarch64_svc(0xABCD);
after_svc:
assert(myvar == 1);
return 0;

View File

@@ -113,7 +113,7 @@ int main(void) {
enable_irq();
}
while (1) {
lkmc_wfi();
lkmc_arm_aarch64_wfi();
}
return 0;
}

View File

@@ -0,0 +1,37 @@
/* https://cirosantilli.com/linux-kernel-module-cheat#arm-multicore */
#include <lkmc.h>
uint64_t spinlock = 0;
__asm__(
".syntax unified\n"
".text\n"
".global lkmc_cpu_not_0\n"
"lkmc_cpu_not_0:\n"
" cmp r0, 1\n"
" bne .Lsleep_forever\n"
" ldr sp, =(stack_top - 0x1000)\n"
" bl main_cpu1\n"
".Lsleep_forever:\n"
" wfe\n"
" b .Lsleep_forever\n"
);
static void main_cpu1(void) {
spinlock = 1;
lkmc_arm_aarch64_dmb(sy);
lkmc_arm_aarch64_sev();
while (1) {
lkmc_arm_aarch64_wfe();
}
}
int main(void) {
#if !LKMC_GEM5
lkmc_arm_psci_cpu_on(1, (uint32_t)main_cpu1, 0);
#endif
while (!spinlock) {
lkmc_arm_aarch64_wfe();
}
}

View File

@@ -2,24 +2,25 @@
#include <lkmc.h>
LKMC_PROLOGUE
.global lkmc_start
lkmc_start:
mov r0, 0
ldr r1, =spinlock
ldr r1, =.Lspinlock
str r0, [r1]
/* Get CPU ID. */
mrc p15, 0, r1, c0, c0, 5
ands r1, r1, 3
beq cpu0_only
beq .Lcpu0_only
.Lcpu1_only:
mov r0, 1
ldr r1, =spinlock
ldr r1, =.Lspinlock
str r0, [r1]
dmb sy
sev
.Lcpu1_sleep_forever:
wfe
b .Lcpu1_sleep_forever
cpu0_only:
.Lcpu0_only:
#if !LKMC_GEM5
/* PSCI CPU_ON. */
ldr r0, =0x84000003
@@ -28,11 +29,12 @@ cpu0_only:
mov r3, 0
hvc 0
#endif
spinlock_start:
ldr r0, spinlock
.Lspinlock_start:
ldr r0, .Lspinlock
wfe
cmp r0, 0
beq spinlock_start
LKMC_EPILOGUE
spinlock:
beq .Lspinlock_start
mov r0, 0
bl _exit
.Lspinlock:
.skip 4

View File

@@ -2,6 +2,11 @@
.global lkmc_start
lkmc_start:
/* Make all CPUs except CPU0 sleep by default. */
mrs x0, mpidr_el1
ands x0, x0, 3
bne lkmc_cpu_not_0
/* Load the vector table. */
ldr x0, =lkmc_vector_table
msr vbar_el1, x0
@@ -41,3 +46,8 @@ LKMC_WEAK(lkmc_vector_trap_handler)
bl abort
lkmc_vector_trap_handler_error_message:
.asciz "error: unexpected interrupt"
/* Default action for CPUs besides the first one: sleep forever. */
LKMC_WEAK(lkmc_cpu_not_0)
wfe
b lkmc_cpu_not_0

View File

@@ -2,6 +2,11 @@
.global lkmc_start
lkmc_start:
/* Make all CPUs except CPU0 sleep by default. */
mrc p15, 0, r0, c0, c0, 5
ands r0, r0, 3
bne lkmc_cpu_not_0
/* Prepare the stack for main, mandatory for C code. */
ldr sp, =stack_top
@@ -28,3 +33,8 @@ lkmc_start:
/* If main returns, exit. */
bl exit
/* Default action for CPUs besides the first one: sleep forever. */
LKMC_WEAK(lkmc_cpu_not_0)
wfe
b lkmc_cpu_not_0

View File

@@ -3,14 +3,6 @@
#include <sys/stat.h>
#include <lkmc.h>
#include <lkmc/m5ops.h>
void lkmc_baremetal_on_exit_callback(int status, void *arg) {
(void)arg;
if (status != 0) {
printf("lkmc_exit_status_%d\n", status);
}
}
enum {
UART_FR_RXFE = 0x10,
@@ -19,44 +11,18 @@ enum {
#define UART_DR(baseaddr) (*(unsigned int *)(baseaddr))
#define UART_FR(baseaddr) (*(((unsigned int *)(baseaddr))+6))
void lkmc_baremetal_on_exit_callback(int status, void *arg) {
(void)arg;
if (status != 0) {
printf("lkmc_exit_status_%d\n", status);
}
}
int _close(int file) {
LKMC_UNUSED(file);
return -1;
}
void _exit(int status) {
LKMC_UNUSED(status);
#if LKMC_GEM5
LKMC_M5OPS_EXIT;
#else
#if defined(__arm__)
__asm__ __volatile__ (
"mov r0, #0x18\n"
"ldr r1, =#0x20026\n"
"svc 0x00123456\n"
:
:
: "r0", "r1"
);
#elif defined(__aarch64__)
/* TODO actually use the exit value here, just for fun. */
__asm__ __volatile__ (
"mov x1, #0x26\n" \
"movk x1, #2, lsl #16\n" \
"str x1, [sp,#0]\n" \
"mov x0, #0\n" \
"str x0, [sp,#8]\n" \
"mov x1, sp\n" \
"mov w0, #0x18\n" \
"hlt 0xf000\n"
:
:
: "x0", "x1"
);
#endif
#endif
}
int _fstat(int file, struct stat *st) {
LKMC_UNUSED(file);
st->st_mode = S_IFCHR;

View File

@@ -0,0 +1,35 @@
#include <lkmc.h>
/* This is implemented in assembly so that it does not use the stack,
* and thus can be called safely from programs without the bootloader.
* C signature:
*
* void _exit(int status)
*
* If only there was a GCC attribute to create such a function!
*/
.text
.global _exit
_exit:
#if LKMC_GEM5
LKMC_M5OPS_EXIT_ASM
#else
/* Use semihosting:
* https://github.com/cirosantilli/linux-kernel-module-cheat#semihosting */
#if defined(__arm__)
mov r0, #0x18
ldr r1, =#0x20026
svc 0x00123456
#elif defined(__aarch64__)
mov x1, 0x26
movk x1, 2, lsl 16
ldr x2, =.Lsemihost_args
str x1, [x2, 0]
str x0, [x2, 8]
mov x1, x2
mov w0, 0x18
hlt 0xf000
.Lsemihost_args:
.skip 16
#endif
#endif

View File

@@ -45,6 +45,14 @@ Build the baremetal examples with crosstool-NG.
self.env['baremetal_build_lib_dir'],
syscalls_basename_noext + self.env['obj_ext']
)
syscalls_asm_src = os.path.join(
self.env['baremetal_source_lib_dir'],
syscalls_basename_noext + '_asm' + self.env['asm_ext']
)
syscalls_asm_obj = os.path.join(
self.env['baremetal_build_lib_dir'],
syscalls_basename_noext + '_asm' + self.env['obj_ext']
)
cc_flags = [
'-I', self.env['root_dir'], LF,
'-O{}'.format(self.env['optimization_level']), LF,
@@ -93,6 +101,7 @@ Build the baremetal examples with crosstool-NG.
(bootloader_src, extra_obj_baremetal_bootloader),
(self.env['common_c'], extra_obj_lkmc_common),
(syscalls_src, syscalls_obj),
(syscalls_asm_src, syscalls_asm_obj),
]:
self._build_one(
in_path=in_path,
@@ -123,7 +132,7 @@ Build the baremetal examples with crosstool-NG.
self.env['baremetal_link_script'],
self.env['common_h']
],
'extra_objs': [syscalls_obj],
'extra_objs': [syscalls_obj, syscalls_asm_obj],
'extra_objs_baremetal_bootloader': [extra_obj_baremetal_bootloader],
'extra_objs_lkmc_common': [extra_obj_lkmc_common],
'in_path': in_path,

52
lkmc.c
View File

@@ -68,7 +68,30 @@ void lkmc_print_newline() {
printf("\n");
}
#if defined(__aarch64__)
#if defined(__arm__)
void lkmc_arm_psci_cpu_on(
uint32_t target_cpu,
uint32_t entry_point_address,
uint32_t context_id
) {
register int r0 __asm__ ("r0") = 0x84000003;
register int r1 __asm__ ("r1") = target_cpu;
register int r2 __asm__ ("r2") = entry_point_address;
register int r3 __asm__ ("r3") = context_id;
__asm__ __volatile__(
"hvc 0\n"
:
: "r" (r0),
"r" (r1),
"r" (r2),
"r" (r3)
:
);
}
#elif defined(__aarch64__)
#define LKMC_SYSREG_READ_WRITE(nbits, name) \
LKMC_CONCAT(LKMC_CONCAT(uint, nbits), _t) LKMC_CONCAT(LKMC_CONCAT(LKMC_SYSREG_SYMBOL_PREFIX, read_), name)(void) { \
LKMC_CONCAT(LKMC_CONCAT(uint, nbits), _t) name; \
@@ -83,4 +106,31 @@ void lkmc_print_newline() {
}
LKMC_SYSREG_OPS
#undef LKMC_SYSREG_READ_WRITE
uint64_t lkmc_aarch64_cpu_id() {
/* TODO: cores beyond 4th?
* Mnemonic: Main Processor ID Register
*/
return lkmc_sysreg_read_mpidr_el1() & 3;
}
void lkmc_aarch64_psci_cpu_on(
uint64_t target_cpu,
uint64_t entry_point_address,
uint64_t context_id
) {
register int w0 __asm__ ("w0") = 0xc4000003;
register int x1 __asm__ ("x1") = target_cpu;
register int x2 __asm__ ("x2") = entry_point_address;
register int x3 __asm__ ("x3") = context_id;
__asm__ __volatile__(
"hvc 0\n"
:
: "r" (w0),
"r" (x1),
"r" (x2),
"r" (x3)
:
);
}
#endif

7
lkmc.h
View File

@@ -1,4 +1,7 @@
/* https://cirosantilli.com/linux-kernel-module-cheat#lkmc-c */
/* https://cirosantilli.com/linux-kernel-module-cheat#lkmc-c
*
* This toplevel header includes all the lkmc/ *.h headers.
*/
#ifndef LKMC_H
#define LKMC_H
@@ -69,4 +72,6 @@ void lkmc_assert_memcmp(const void *s1, const void *s2, size_t n, uint32_t line)
#error
#endif
#include <lkmc/m5ops.h>
#endif

View File

@@ -1,6 +1,8 @@
#ifndef LKMC_AARCH64_H
#define LKMC_AARCH64_H
#include <lkmc/arm_aarch64.h>
#define LKMC_ASSERT_EQ(reg, const) \
mov x0, reg; \
ldr x1, const; \
@@ -292,10 +294,7 @@ typedef struct {
} LkmcVectorExceptionFrame;
void lkmc_vector_trap_handler(LkmcVectorExceptionFrame *exception);
/* Misc assembly instructions. */
#define lkmc_svc(immediate) __asm__ __volatile__("svc " #immediate : : : )
#define lkmc_wfi() __asm__ __volatile__ ("wfi" : : : "memory")
void lkmc_cpu_not_0(uint64_t cpuid);
/* Sysreg read and write functions, e.g.:
*
@@ -310,16 +309,26 @@ void lkmc_vector_trap_handler(LkmcVectorExceptionFrame *exception);
void LKMC_CONCAT(LKMC_CONCAT(LKMC_SYSREG_SYMBOL_PREFIX, print_), name)(void);
#define LKMC_SYSREG_OPS \
LKMC_SYSREG_READ_WRITE(32, cntv_ctl_el0) \
LKMC_SYSREG_READ_WRITE(64, daif) \
LKMC_SYSREG_READ_WRITE(32, spsel) \
LKMC_SYSREG_READ_WRITE(64, cntfrq_el0) \
LKMC_SYSREG_READ_WRITE(64, cntv_cval_el0) \
LKMC_SYSREG_READ_WRITE(64, cntv_tval_el0) \
LKMC_SYSREG_READ_WRITE(64, cntvct_el0) \
LKMC_SYSREG_READ_WRITE(64, daif) \
LKMC_SYSREG_READ_WRITE(64, mpidr_el1) \
LKMC_SYSREG_READ_WRITE(64, sp_el1) \
LKMC_SYSREG_READ_WRITE(32, spsel) \
LKMC_SYSREG_READ_WRITE(64, vbar_el1)
LKMC_SYSREG_OPS
#undef LKMC_SYSREG_READ_WRITE
/* Determine what is the ID of the currently running CPU. */
uint64_t lkmc_aarch64_cpu_id();
void lkmc_aarch64_psci_cpu_on(
uint64_t target_cpu,
uint64_t entry_point_address,
uint64_t context_id
);
#endif
#endif

View File

@@ -1,8 +1,20 @@
#ifndef LKMC_ARM_H
#define LKMC_ARM_H
#include <lkmc/arm_aarch64.h>
#if defined(__ASSEMBLER__)
.syntax unified
#else
void lkmc_arm_psci_cpu_on(
uint32_t target_cpu,
uint32_t entry_point_address,
uint32_t context_id
);
#endif
#define LKMC_ASSERT_EQ(reg, const) \

12
lkmc/arm_aarch64.h Normal file
View File

@@ -0,0 +1,12 @@
#ifndef LKMC_ARM_AARCH64_H
#define LKMC_ARM_AARCH64_H
/* Stuff that is common between arm and aarch64. */
#define lkmc_arm_aarch64_wfe() __asm__ __volatile__ ("wfe" : : : )
#define lkmc_arm_aarch64_dmb(type) __asm__ __volatile__ ("dmb " #type : : : "memory")
#define lkmc_arm_aarch64_sev(immediate) __asm__ __volatile__("sev" : : : )
#define lkmc_arm_aarch64_hvc(immediate) __asm__ __volatile__("hvc " #immediate : : : )
#define lkmc_arm_aarch64_svc(immediate) __asm__ __volatile__("svc " #immediate : : : )
#define lkmc_arm_aarch64_wfi() __asm__ __volatile__ ("wfi" : : : )
#endif

View File

@@ -284,11 +284,17 @@ path_properties_tuples = (
'arm': (
{'allowed_archs': {'arm'}},
{
'multicore.S': {'test_run_args': {'cpus': 2}},
'multicore.c': {
# It is hard to get visibility into what is going on
# in that one due to the multicore business.
'skip_run_unclassified': True,
'test_run_args': {'cpus': 2}
},
'no_bootloader': (
{'extra_objs_disable_baremetal_bootloader': True},
{
'gem5_exit.S': {'requires_m5ops': True},
'multicore_asm.S': {'test_run_args': {'cpus': 2}},
'semihost_exit.S': {'requires_semihosting': True},
}
),
@@ -300,11 +306,12 @@ path_properties_tuples = (
'aarch64': (
{'allowed_archs': {'aarch64'}},
{
'multicore.S': {'test_run_args': {'cpus': 2}},
'multicore.c': {'test_run_args': {'cpus': 2}},
'no_bootloader': (
{'extra_objs_disable_baremetal_bootloader': True},
{
'gem5_exit.S': {'requires_m5ops': True},
'multicore_asm.S': {'test_run_args': {'cpus': 2}},
'semihost_exit.S': {'requires_semihosting': True},
'wfe_loop.S': {'more_than_1s': True},
}