Implementations of some libroot functions for x86_64.

These are the functions required by the kernel. These are all full
implementations except for system_time(), which will be implemented later.

headers/posix/arch/x86_64/arch_setjmp.h

@@ -1,12 +1,11 @@
 /*
- * Copyright 2005-2010 Haiku Inc. All Rights Reserved.
+ * Copyright 2005-2012 Haiku Inc. All Rights Reserved.
  * Distributed under the terms of the MIT License.
  */
 #ifndef _ARCH_SETJMP_H_
 #define _ARCH_SETJMP_H_
 
 
-/* TODO: A jmp_buf size of 12 might not be large enough. Increase to a large size if needed. */
-typedef int __jmp_buf[12];
+typedef int __jmp_buf[8];
 
 #endif	/* _ARCH_SETJMP_H_ */

src/system/libroot/os/arch/x86_64/atomic.S

@@ -0,0 +1,128 @@
+/*
+ * Copyright 2012, Alex Smith, alex@alex-smith.me.uk.
+ * Distributed under the terms of the MIT License.
+ */
+
+
+#include <asm_defs.h>
+
+
+.text
+
+/* int32 atomic_set(vint32 *value, int32 newValue) */
+FUNCTION(atomic_set):
+	movl		%esi, %eax
+	lock
+	xchgl		%eax, (%rdi)
+	ret
+FUNCTION_END(atomic_set)
+
+/* int32 atomic_test_and_set(vint32 *value, int32 newValue, int32 testAgainst) */
+FUNCTION(atomic_test_and_set):
+	movl		%edx, %eax
+	lock
+	cmpxchgl	%esi, (%rdi)
+	ret
+FUNCTION_END(atomic_test_and_set)
+
+/* int32 atomic_add(vint32 *value, int32 addValue) */
+FUNCTION(atomic_add):
+	movl		%esi, %eax
+	lock
+	xaddl		%eax, (%rdi)
+	ret
+FUNCTION_END(atomic_add)
+
+/* int32 atomic_and(vint32 *value, int32 andValue) */
+FUNCTION(atomic_and):
+	movl		(%rdi), %eax
+1:	movl		%eax, %edx
+	movl		%eax, %ecx
+	andl		%esi, %edx
+	lock
+	cmpxchgl	%edx, (%rdi)
+	jnz			1b
+	movl		%ecx, %eax
+	ret
+FUNCTION_END(atomic_and)
+
+/* int32 atomic_or(vint32 *value, int32 orValue) */
+FUNCTION(atomic_or):
+	movl		(%rdi), %eax
+1:	movl		%eax, %edx
+	movl		%eax, %ecx
+	orl			%esi, %edx
+	lock
+	cmpxchgl	%edx, (%rdi)
+	jnz			1b
+	movl		%ecx, %eax
+	ret
+FUNCTION_END(atomic_or)
+
+/* int32 atomic_get(vint32 *value) */
+FUNCTION(atomic_get):
+	movl		(%rdi), %eax
+1:	lock
+	cmpxchgl	%eax, (%rdi)
+	jnz			1b
+	ret
+FUNCTION_END(atomic_get)
+
+/* int64 atomic_set64(vint64 *value, int64 newValue) */
+FUNCTION(atomic_set64):
+	movq		%rsi, %rax
+	lock
+	xchgq		%rax, (%rdi)
+	ret
+FUNCTION_END(atomic_set64)
+
+/* int64 atomic_test_and_set64(vint64 *value, int64 newValue, int64 testAgainst) */
+FUNCTION(atomic_test_and_set64):
+	movq		%rdx, %rax
+	lock
+	cmpxchgq	%rsi, (%rdi)
+	ret
+FUNCTION_END(atomic_test_and_set64)
+
+/* int64 atomic_add64(vint64 *value, int64 addValue) */
+FUNCTION(atomic_add64):
+	movq		%rsi, %rax
+	lock
+	xaddq		%rax, (%rdi)
+	ret
+FUNCTION_END(atomic_add64)
+
+/* int64 atomic_and64(vint64 *value, int64 andValue) */
+FUNCTION(atomic_and64):
+	movq		(%rdi), %rax
+1:	movq		%rax, %rdx
+	movq		%rax, %rcx
+	andq		%rsi, %rdx
+	lock
+	cmpxchgq	%rdx, (%rdi)
+	jnz			1b
+	movq		%rcx, %rax
+	ret
+FUNCTION_END(atomic_and64)
+
+/* int64 atomic_or64(vint64 *value, int64 orValue) */
+FUNCTION(atomic_or64):
+	movq		(%rdi), %rax
+1:	movq		%rax, %rdx
+	movq		%rax, %rcx
+	orq			%rsi, %rdx
+	lock
+	cmpxchgq	%rdx, (%rdi)
+	jnz			1b
+	movq		%rcx, %rax
+	ret
+FUNCTION_END(atomic_or64)
+
+/* int64 atomic_get64(vint64 *value) */
+FUNCTION(atomic_get64):
+	movq		(%rdi), %rax
+1:	lock
+	cmpxchgq	%rax, (%rdi)
+	jnz			1b
+	ret
+FUNCTION_END(atomic_get64)

src/system/libroot/os/arch/x86_64/byteorder.S

@@ -0,0 +1,54 @@
+/*
+ * Copyright 2012, Alex Smith, alex@alex-smith.me.uk.
+ * Distributed under the terms of the MIT License.
+ */
+
+
+#include <asm_defs.h>
+
+
+/* uint16 __swap_int16(uint16 value) */
+FUNCTION(__swap_int16):
+	movl	%edi, %eax
+	bswap	%eax
+	shr		$16, %eax
+	ret
+FUNCTION_END(__swap_int16)
+
+/* uint32 __swap_int32(uint32 value) */
+FUNCTION(__swap_int32):
+	movl	%edi, %eax
+	bswap	%eax
+	ret
+FUNCTION_END(__swap_int32)
+
+/* uint64 __swap_int64(uint64 value) */
+FUNCTION(__swap_int64):
+	movq	%rdi, %rax
+	bswap	%rax
+	ret
+FUNCTION_END(__swap_int64)
+
+/* float __swap_float(float value) */
+FUNCTION(__swap_float):
+	sub		$8, %rsp
+	movss	%xmm0, (%rsp)
+	movl	(%rsp), %eax
+	bswap	%eax
+	movl	%eax, (%rsp)
+	movss	(%rsp), %xmm0
+	add		$8, %rsp
+	ret
+FUNCTION_END(__swap_float)
+
+/* double __swap_double(double value) */
+FUNCTION(__swap_double):
+	sub		$8, %rsp
+	movsd	%xmm0, (%rsp)
+	movq	(%rsp), %rax
+	bswap	%rax
+	movq	%rax, (%rsp)
+	movsd	(%rsp), %xmm0
+	add		$8, %rsp
+	ret
+FUNCTION_END(__swap_double)

src/system/libroot/os/arch/x86_64/system_time.c

@@ -0,0 +1,4 @@
+/*
+	Just a dummy to avoid a special case in the build system. system_time()
+	is implemented in system_time_asm.S.
+*/

src/system/libroot/os/arch/x86_64/system_time_asm.S

@@ -0,0 +1,149 @@
+/*
+ * Copyright 2009, Ingo Weinhold, ingo_weinhold@gmx.de.
+ * Copyright 2012, Alex Smith, alex@alex-smith.me.uk.
+ * Distributed under the terms of the MIT License.
+ *
+ * Copyright 2001, Travis Geiselbrecht. All rights reserved.
+ * Distributed under the terms of the NewOS License.
+ */
+
+#include <asm_defs.h>
+
+FUNCTION(__x86_setup_system_time):
+	ret
+FUNCTION_END(__x86_setup_system_time)
+
+// TODO: Implement these.
+
+/* int64 system_time(); */
+FUNCTION(system_time):
+	ud2a
+	ret
+FUNCTION_END(system_time)
+
+
+/* int64 system_time_nsecs(); */
+FUNCTION(system_time_nsecs):
+	ud2a
+	ret
+FUNCTION_END(system_time_nsecs)
+
+#if 0
+/* saves the conversion factor needed for system_time */
+.lcomm cv_factor				4
+.lcomm cv_factor_nsecs			4
+.lcomm cv_factor_nsecs_shift	1
+
+
+.text
+
+
+FUNCTION(__x86_setup_system_time):
+	movl	4(%esp), %eax
+	movl	%eax, cv_factor
+	movl	8(%esp), %eax
+	movl	%eax, cv_factor_nsecs
+	movb	12(%esp), %al
+	movb	%al, cv_factor_nsecs_shift
+	ret
+FUNCTION_END(__x86_setup_system_time)
+
+
+/* int64 system_time(); */
+FUNCTION(system_time):
+	pushl	%ebx
+	pushl	%ecx
+	movl	cv_factor, %ebx
+
+	/* load 64-bit factor into %eax (low), %edx (high) */
+	rdtsc		/* time in %edx,%eax */
+
+	movl	%edx, %ecx	/* save high half */
+	mull	%ebx 		/* truncate %eax, but keep %edx */
+	movl	%ecx, %eax
+	movl	%edx, %ecx	/* save high half of low */
+	mull	%ebx /*, %eax*/
+	/* now compute  [%edx, %eax] + [%ecx], propagating carry */
+	subl	%ebx, %ebx	/* need zero to propagate carry */
+	addl	%ecx, %eax
+	adc		%ebx, %edx
+	popl	%ecx
+	popl	%ebx
+	ret
+FUNCTION_END(system_time)
+
+
+/* int64 system_time_nsecs(); */
+FUNCTION(system_time_nsecs):
+	testb	$0, cv_factor_nsecs_shift
+	jne		1f
+
+	/* same algorithm as system_time(), just with a different factor */
+
+	pushl	%ebx
+	pushl	%ecx
+	movl	cv_factor_nsecs, %ebx
+
+	/* load 64-bit factor into %eax (low), %edx (high) */
+	rdtsc		/* time in %edx,%eax */
+
+	movl	%edx, %ecx	/* save high half */
+	mull	%ebx 		/* truncate %eax, but keep %edx */
+	movl	%ecx, %eax
+	movl	%edx, %ecx	/* save high half of low */
+	mull	%ebx /*, %eax*/
+	/* now compute  [%edx, %eax] + [%ecx], propagating carry */
+	subl	%ebx, %ebx	/* need zero to propagate carry */
+	addl	%ecx, %eax
+	adc		%ebx, %edx
+	popl	%ecx
+	popl	%ebx
+	ret
+
+1:
+	/* TSC frequency is less than 1 GHz -- we shift everything up 16 bit */
+
+	pushl	%ebx
+	pushl	%ecx
+	pushl	%esi
+	movl	cv_factor_nsecs, %ebx
+
+	/* load 64-bit factor into %eax (low), %edx (high) */
+	rdtsc		/* time in %edx,%eax */
+
+	/* save high half */
+	movl	%edx, %ecx
+
+	/* multiply low half by conversion factor */
+	mull	%ebx
+
+	/* save result */
+	movl	%eax, %esi	/* low half -> %esi */
+	movl	%ecx, %eax
+	movl	%edx, %ecx	/* high half -> %ecx */
+
+	/* multiply high half by conversion factor */
+	mull	%ebx
+
+	/* now compute  [%edx, %eax] + [%ecx], propagating carry */
+	xorl	%ebx, %ebx	/* need zero to propagate carry */
+	addl	%ecx, %eax
+	adc		%ebx, %edx
+
+	/* shift the result left 16 bit */
+	shl		$16, %edx
+	movl	%eax, %ebx
+	shr		$16, %ebx
+	orw		%bx, %dx
+	shl		$16, %eax
+
+	/* add the high 16 bit of the low half of the low product */
+	shr		$16, %esi
+	orw		%si, %ax
+
+	popl	%esi
+	popl	%ecx
+	popl	%ebx
+	ret
+FUNCTION_END(system_time_nsecs)
+#endif

src/system/libroot/posix/arch/x86_64/setjmp_internal.h

@@ -0,0 +1,20 @@
+/*
+ * Copyright 2012, Alex Smith, alex@alex-smith.me.uk.
+ * Distributed under the terms of the MIT License.
+ */
+#ifndef SETJMP_INTERNAL_H
+#define SETJMP_INTERNAL_H
+
+
+#define JMP_REGS_IP		0
+#define JMP_REGS_SP		8
+#define JMP_REGS_BP		16
+#define JMP_REGS_BX		24
+#define JMP_REGS_R12	32
+#define JMP_REGS_R13	40
+#define JMP_REGS_R14	48
+#define JMP_REGS_R15	56
+
+#include <asm_defs.h>
+
+#endif	/* SETJMP_INTERNAL_H */

src/system/libroot/posix/arch/x86_64/siglongjmp.S

@@ -0,0 +1,32 @@
+/*
+ * Copyright 2012, Alex Smith, alex@alex-smith.me.uk.
+ * Distributed under the terms of the MIT License.
+ */
+
+#include "setjmp_internal.h"
+
+
+/* int __siglongjmp(jmp_buf buffer, int value) */
+FUNCTION(siglongjmp):
+FUNCTION(longjmp):
+FUNCTION(_longjmp):
+	// Restore new callee-save registers.
+	movq	JMP_REGS_R15(%rdi), %r15
+	movq	JMP_REGS_R14(%rdi), %r14
+	movq	JMP_REGS_R13(%rdi), %r13
+	movq	JMP_REGS_R12(%rdi), %r12
+	movq	JMP_REGS_BX(%rdi), %rbx
+	movq	JMP_REGS_BP(%rdi), %rbp
+
+	// Restore new stack pointer and push return address.
+	movq	JMP_REGS_SP(%rdi), %rsp
+	movq	JMP_REGS_IP(%rdi), %rax
+	push	%rax
+
+	// __longjmp_return restores the signal mask and sets the return value.
+	call	__longjmp_return
+	ret
+FUNCTION_END(siglongjmp)
+
+
+#pragma weak longjmp=siglongjmp

src/system/libroot/posix/arch/x86_64/sigsetjmp.S

@@ -0,0 +1,39 @@
+/*
+ * Copyright 2012, Alex Smith, alex@alex-smith.me.uk.
+ * Distributed under the terms of the MIT License.
+ */
+
+#include "setjmp_internal.h"
+
+
+/* int sigsetjmp(jmp_buf buffer, int saveMask) */
+FUNCTION(__sigsetjmp):
+FUNCTION(sigsetjmp):
+	// Save instruction/stack pointers.
+	movq	(%rsp), %rax
+	movq	%rax, JMP_REGS_IP(%rdi)
+	leaq	8(%rsp), %rax
+	movq	%rax, JMP_REGS_SP(%rdi)
+
+	// Save callee-save registers.
+	movq	%rbp, JMP_REGS_BP(%rdi)
+	movq	%rbx, JMP_REGS_BX(%rdi)
+	movq	%r12, JMP_REGS_R12(%rdi)
+	movq	%r13, JMP_REGS_R13(%rdi)
+	movq	%r14, JMP_REGS_R14(%rdi)
+	movq	%r15, JMP_REGS_R15(%rdi)
+
+	// __setjmp_save_sigs handles signal mask and return value.
+	jmp		__setjmp_save_sigs
+FUNCTION_END(setjmp)
+
+
+/* int setjmp(jmp_buf buffer) */
+FUNCTION(setjmp):
+	// Jump to sigsetjmp with a zero saveMask.
+	xorl	%edi, %edi
+	jmp		sigsetjmp
+FUNCTION_END(setjmp)
+
+
+#pragma weak _setjmp=setjmp