patrick/zephyr
1
0
Fork 0
Code Issues Pull requests Actions 8 Packages Projects Releases Wiki Activity
zephyr/include/toolchain/gcc.h
Andy Ross b69d0da82d arch/x86_64: New architecture added 
This patch adds a x86_64 architecture and qemu_x86_64 board to Zephyr.
Only the basic architecture support needed to run 64 bit code is
added; no drivers are added, though a low-level console exists and is
wired to printk().

The support is built on top of a "X86 underkernel" layer, which can be
built in isolation as a unit test on a Linux host.

Limitations:

+ Right now the SDK lacks an x86_64 toolchain.  The build will fall
  back to a host toolchain if it finds no cross compiler defined,
  which is tested to work on gcc 8.2.1 right now.

+ No x87/SSE/AVX usage is allowed.  This is a stronger limitation than
  other architectures where the instructions work from one thread even
  if the context switch code doesn't support it.  We are passing
  -no-sse to prevent gcc from automatically generating SSE
  instructions for non-floating-point purposes, which has the side
  effect of changing the ABI.  Future work to handle the FPU registers
  will need to be combined with an "application" ABI distinct from the
  kernel one (or just to require USERSPACE).

+ Paging is enabled (it has to be in long mode), but is a 1:1 mapping
  of all memory.  No MMU/USERSPACE support yet.

+ We are building with -mno-red-zone for stack size reasons, but this
  is a valuable optimization.  Enabling it requires automatic stack
  switching, which requires a TSS, which means it has to happen after
  MMU support.

+ The OS runs in 64 bit mode, but for compatibility reasons is
  compiled to the 32 bit "X32" ABI.  So while the full 64 bit
  registers and instruction set are available, C pointers are 32 bits
  long and Zephyr is constrained to run in the bottom 4G of memory.

Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
2019-01-11 15:18:52 -05:00

379 lines
11 KiB
C
Raw Blame History

/*
* Copyright (c) 2010-2014,2017 Wind River Systems, Inc.
*
* SPDX-License-Identifier: Apache-2.0
*/
#ifndef ZEPHYR_INCLUDE_TOOLCHAIN_GCC_H_
#define ZEPHYR_INCLUDE_TOOLCHAIN_GCC_H_
/**
* @file
* @brief GCC toolchain abstraction
*
* Macros to abstract compiler capabilities for GCC toolchain.
*/
/* C++11 has static_assert built in */
#ifdef __cplusplus
#define BUILD_ASSERT(EXPR) static_assert(EXPR, "")
#define BUILD_ASSERT_MSG(EXPR, MSG) static_assert(EXPR, MSG)
/*
* GCC 4.6 and higher have the C11 _Static_assert built in, and its
* output is easier to understand than the common BUILD_ASSERT macros.
*/
#elif (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6)) || \
(__STDC_VERSION__) >= 201100
#define BUILD_ASSERT(EXPR) _Static_assert(EXPR, "")
#define BUILD_ASSERT_MSG(EXPR, MSG) _Static_assert(EXPR, MSG)
#endif
#include <toolchain/common.h>
#include <stdbool.h>
#define ALIAS_OF(of) __attribute__((alias(#of)))
#define FUNC_ALIAS(real_func, new_alias, return_type) \
return_type new_alias() ALIAS_OF(real_func)
#if defined(CONFIG_ARCH_POSIX)
/*let's not segfault if this were to happen for some reason*/
#define CODE_UNREACHABLE \
{\
posix_print_error_and_exit("CODE_UNREACHABLE reached from %s:%d\n",\
__FILE__, __LINE__);\
__builtin_unreachable(); \
}
#else
#define CODE_UNREACHABLE __builtin_unreachable()
#endif
#define FUNC_NORETURN __attribute__((__noreturn__))
/* The GNU assembler for Cortex-M3 uses # for immediate values, not
* comments, so the @nobits# trick does not work.
*/
#if defined(CONFIG_ARM)
#define _NODATA_SECTION(segment) __attribute__((section(#segment)))
#else
#define _NODATA_SECTION(segment) \
__attribute__((section(#segment ",\"wa\",@nobits#")))
#endif
/* Unaligned access */
#define UNALIGNED_GET(p) \
__extension__ ({ \
struct __attribute__((__packed__)) { \
__typeof__(*(p)) __v; \
} *__p = (__typeof__(__p)) (p); \
__p->__v; \
})
#if __GNUC__ >= 7 && defined(CONFIG_ARM)
/* Version of UNALIGNED_PUT() which issues a compiler_barrier() after
* the store. It is required to workaround an apparent optimization
* bug in GCC for ARM Cortex-M3 and higher targets, when multiple
* byte, half-word and word stores (strb, strh, str instructions),
* which support unaligned access, can be coalesced into store double
* (strd) instruction, which doesn't support unaligned access (the
* compilers in question do this optimization ignoring __packed__
* attribute).
*/
#define UNALIGNED_PUT(v, p) \
do { \
struct __attribute__((__packed__)) { \
__typeof__(*p) __v; \
} *__p = (__typeof__(__p)) (p); \
__p->__v = (v); \
compiler_barrier(); \
} while (false)
#else
#define UNALIGNED_PUT(v, p) \
do { \
struct __attribute__((__packed__)) { \
__typeof__(*p) __v; \
} *__p = (__typeof__(__p)) (p); \
__p->__v = (v); \
} while (false)
#endif
/* Double indirection to ensure section names are expanded before
* stringification
*/
#define __GENERIC_SECTION(segment) __attribute__((section(STRINGIFY(segment))))
#define _GENERIC_SECTION(segment) __GENERIC_SECTION(segment)
#define ___in_section(a, b, c) \
__attribute__((section("." _STRINGIFY(a) \
"." _STRINGIFY(b) \
"." _STRINGIFY(c))))
#define __in_section(a, b, c) ___in_section(a, b, c)
#define __in_section_unique(seg) ___in_section(seg, __FILE__, __COUNTER__)
#ifdef CONFIG_APPLICATION_MEMORY
#define __kernel __in_section_unique(kernel)
#define __kernel_noinit __in_section_unique(kernel_noinit)
#define __kernel_bss __in_section_unique(kernel_bss)
#else
#define __kernel
#define __kernel_noinit __noinit
#define __kernel_bss
#endif
#ifndef __packed
#define __packed __attribute__((__packed__))
#endif
#ifndef __aligned
#define __aligned(x) __attribute__((__aligned__(x)))
#endif
#define __may_alias __attribute__((__may_alias__))
#ifndef __printf_like
#define __printf_like(f, a) __attribute__((format (printf, f, a)))
#endif
#define __used __attribute__((__used__))
#define __deprecated __attribute__((deprecated))
#define ARG_UNUSED(x) (void)(x)
#define likely(x) __builtin_expect((bool)!!(x), true)
#define unlikely(x) __builtin_expect((bool)!!(x), false)
#define popcount(x) __builtin_popcount(x)
#ifndef __weak
#define __weak __attribute__((__weak__))
#endif
#define __unused __attribute__((__unused__))
/* Be *very* careful with this, you cannot filter out with -wno-deprecated,
* which has implications for -Werror
*/
#define __DEPRECATED_MACRO _Pragma("GCC warning \"Macro is deprecated\"")
/* These macros allow having ARM asm functions callable from thumb */
#if defined(_ASMLANGUAGE) && !defined(_LINKER)
#ifdef CONFIG_ARM
#if defined(CONFIG_ISA_THUMB)
#define FUNC_CODE() \
.code 16; \
.thumb_func;
#define FUNC_INSTR(a) \
BX pc; \
NOP; \
.code 32; \
A##a:
#elif defined(CONFIG_ISA_THUMB2)
#define FUNC_CODE() .thumb;
#define FUNC_INSTR(a)
#elif defined(CONFIG_ISA_ARM)
#define FUNC_CODE() .code 32;
#define FUNC_INSTR(a)
#else
#error unknown instruction set
#endif /* ISA */
#else
#define FUNC_CODE()
#define FUNC_INSTR(a)
#endif /* !CONFIG_ARM */
#endif /* _ASMLANGUAGE && !_LINKER */
/*
* These macros are used to declare assembly language symbols that need
* to be typed properly(func or data) to be visible to the OMF tool.
* So that the build tool could mark them as an entry point to be linked
* correctly. This is an elfism. Use #if 0 for a.out.
*/
#if defined(_ASMLANGUAGE) && !defined(_LINKER)
#if defined(CONFIG_ARM) || defined(CONFIG_NIOS2) || defined(CONFIG_RISCV32) \
|| defined(CONFIG_XTENSA)
#define GTEXT(sym) .global sym; .type sym, %function
#define GDATA(sym) .global sym; .type sym, %object
#define WTEXT(sym) .weak sym; .type sym, %function
#define WDATA(sym) .weak sym; .type sym, %object
#elif defined(CONFIG_ARC)
/*
* Need to use assembly macros because ';' is interpreted as the start of
* a single line comment in the ARC assembler.
*/
.macro glbl_text symbol
.globl \symbol
.type \symbol, %function
.endm
.macro glbl_data symbol
.globl \symbol
.type \symbol, %object
.endm
.macro weak_data symbol
.weak \symbol
.type \symbol, %object
.endm
#define GTEXT(sym) glbl_text sym
#define GDATA(sym) glbl_data sym
#define WDATA(sym) weak_data sym
#else /* !CONFIG_ARM && !CONFIG_ARC */
#define GTEXT(sym) .globl sym; .type sym, @function
#define GDATA(sym) .globl sym; .type sym, @object
#endif
/*
* These macros specify the section in which a given function or variable
* resides.
*
* - SECTION_FUNC allows only one function to reside in a sub-section
* - SECTION_SUBSEC_FUNC allows multiple functions to reside in a sub-section
* This ensures that garbage collection only discards the section
* if all functions in the sub-section are not referenced.
*/
#if defined(CONFIG_ARC)
/*
* Need to use assembly macros because ';' is interpreted as the start of
* a single line comment in the ARC assembler.
*
* Also, '\()' is needed in the .section directive of these macros for
* correct substitution of the 'section' variable.
*/
.macro section_var section, symbol
.section .\section\().\symbol
\symbol :
.endm
.macro section_func section, symbol
.section .\section\().\symbol, "ax"
FUNC_CODE()
PERFOPT_ALIGN
\symbol :
FUNC_INSTR(\symbol)
.endm
.macro section_subsec_func section, subsection, symbol
.section .\section\().\subsection, "ax"
PERFOPT_ALIGN
\symbol :
.endm
#define SECTION_VAR(sect, sym) section_var sect, sym
#define SECTION_FUNC(sect, sym) section_func sect, sym
#define SECTION_SUBSEC_FUNC(sect, subsec, sym) \
section_subsec_func sect, subsec, sym
#else /* !CONFIG_ARC */
#define SECTION_VAR(sect, sym) .section .sect.##sym; sym :
#define SECTION_FUNC(sect, sym) \
.section .sect.sym, "ax"; \
FUNC_CODE() \
PERFOPT_ALIGN; sym : \
FUNC_INSTR(sym)
#define SECTION_SUBSEC_FUNC(sect, subsec, sym) \
.section .sect.subsec, "ax"; PERFOPT_ALIGN; sym :
#endif /* CONFIG_ARC */
#endif /* _ASMLANGUAGE && !_LINKER */
#if defined(CONFIG_ARM) && defined(_ASMLANGUAGE)
#if defined(CONFIG_ISA_THUMB2)
/* '.syntax unified' is a gcc-ism used in thumb-2 asm files */
#define _ASM_FILE_PROLOGUE .text; .syntax unified; .thumb
#elif defined(CONFIG_ISA_THUMB)
#define _ASM_FILE_PROLOGUE .text; .code 16
#else
#define _ASM_FILE_PROLOGUE .text; .code 32
#endif
#endif
/*
* These macros generate absolute symbols for GCC
*/
/* create an extern reference to the absolute symbol */
#define GEN_OFFSET_EXTERN(name) extern const char name[]
#define GEN_ABS_SYM_BEGIN(name) \
EXTERN_C void name(void); \
void name(void) \
{
#define GEN_ABS_SYM_END }
#if defined(CONFIG_ARM)
/*
* GNU/ARM backend does not have a proper operand modifier which does not
* produces prefix # followed by value, such as %0 for PowerPC, Intel, and
* MIPS. The workaround performed here is using %B0 which converts
* the value to ~(value). Thus "n"(~(value)) is set in operand constraint
* to output (value) in the ARM specific GEN_OFFSET macro.
*/
#define GEN_ABSOLUTE_SYM(name, value) \
__asm__(".globl\t" #name "\n\t.equ\t" #name \
",%B0" \
"\n\t.type\t" #name ",%%object" : : "n"(~(value)))
#elif defined(CONFIG_X86) || defined(CONFIG_ARC)
#define GEN_ABSOLUTE_SYM(name, value) \
__asm__(".globl\t" #name "\n\t.equ\t" #name \
",%c0" \
"\n\t.type\t" #name ",@object" : : "n"(value))
#elif defined(CONFIG_X86_64)
#define GEN_ABSOLUTE_SYM(name, value) \
__asm__(".globl\t" #name "\n\t.equ\t" #name \
",%0" \
"\n\t.type\t" #name ",@object" : : "n"(value))
#elif defined(CONFIG_NIOS2) || defined(CONFIG_RISCV32) || defined(CONFIG_XTENSA)
/* No special prefixes necessary for constants in this arch AFAICT */
#define GEN_ABSOLUTE_SYM(name, value) \
__asm__(".globl\t" #name "\n\t.equ\t" #name \
",%0" \
"\n\t.type\t" #name ",%%object" : : "n"(value))
#elif defined(CONFIG_ARCH_POSIX)
#define GEN_ABSOLUTE_SYM(name, value) \
__asm__(".globl\t" #name "\n\t.equ\t" #name \
",%c0" \
"\n\t.type\t" #name ",@object" : : "n"(value))
#else
#error processor architecture not supported
#endif
#define compiler_barrier() do { \
__asm__ __volatile__ ("" ::: "memory"); \
} while (false)
#endif /* ZEPHYR_INCLUDE_TOOLCHAIN_GCC_H_ */
Reference in a new issue View git blame Copy permalink