diff options
Diffstat (limited to 'common/atomic/i386/atomic.s')
-rw-r--r-- | common/atomic/i386/atomic.s | 720 |
1 files changed, 720 insertions, 0 deletions
diff --git a/common/atomic/i386/atomic.s b/common/atomic/i386/atomic.s new file mode 100644 index 000000000000..4fa525ba20af --- /dev/null +++ b/common/atomic/i386/atomic.s @@ -0,0 +1,720 @@ +/* + * CDDL HEADER START + * + * The contents of this file are subject to the terms of the + * Common Development and Distribution License (the "License"). + * You may not use this file except in compliance with the License. + * + * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE + * or http://www.opensolaris.org/os/licensing. + * See the License for the specific language governing permissions + * and limitations under the License. + * + * When distributing Covered Code, include this CDDL HEADER in each + * file and include the License file at usr/src/OPENSOLARIS.LICENSE. + * If applicable, add the following below this CDDL HEADER, with the + * fields enclosed by brackets "[]" replaced with your own identifying + * information: Portions Copyright [yyyy] [name of copyright owner] + * + * CDDL HEADER END + */ + +/* + * Copyright 2010 Sun Microsystems, Inc. All rights reserved. + * Use is subject to license terms. + */ + + .file "atomic.s" + +#include <sys/asm_linkage.h> + +#if defined(_KERNEL) + /* + * Legacy kernel interfaces; they will go away (eventually). + */ + ANSI_PRAGMA_WEAK2(cas8,atomic_cas_8,function) + ANSI_PRAGMA_WEAK2(cas32,atomic_cas_32,function) + ANSI_PRAGMA_WEAK2(cas64,atomic_cas_64,function) + ANSI_PRAGMA_WEAK2(caslong,atomic_cas_ulong,function) + ANSI_PRAGMA_WEAK2(casptr,atomic_cas_ptr,function) + ANSI_PRAGMA_WEAK2(atomic_and_long,atomic_and_ulong,function) + ANSI_PRAGMA_WEAK2(atomic_or_long,atomic_or_ulong,function) +#endif + + ENTRY(atomic_inc_8) + ALTENTRY(atomic_inc_uchar) + movl 4(%esp), %eax + lock + incb (%eax) + ret + SET_SIZE(atomic_inc_uchar) + SET_SIZE(atomic_inc_8) + + ENTRY(atomic_inc_16) + ALTENTRY(atomic_inc_ushort) + movl 4(%esp), %eax + lock + incw (%eax) + ret + SET_SIZE(atomic_inc_ushort) + SET_SIZE(atomic_inc_16) + + ENTRY(atomic_inc_32) + ALTENTRY(atomic_inc_uint) + ALTENTRY(atomic_inc_ulong) + movl 4(%esp), %eax + lock + incl (%eax) + ret + SET_SIZE(atomic_inc_ulong) + SET_SIZE(atomic_inc_uint) + SET_SIZE(atomic_inc_32) + + ENTRY(atomic_inc_8_nv) + ALTENTRY(atomic_inc_uchar_nv) + movl 4(%esp), %edx / %edx = target address + xorl %eax, %eax / clear upper bits of %eax + incb %al / %al = 1 + lock + xaddb %al, (%edx) / %al = old value, inc (%edx) + incb %al / return new value + ret + SET_SIZE(atomic_inc_uchar_nv) + SET_SIZE(atomic_inc_8_nv) + + ENTRY(atomic_inc_16_nv) + ALTENTRY(atomic_inc_ushort_nv) + movl 4(%esp), %edx / %edx = target address + xorl %eax, %eax / clear upper bits of %eax + incw %ax / %ax = 1 + lock + xaddw %ax, (%edx) / %ax = old value, inc (%edx) + incw %ax / return new value + ret + SET_SIZE(atomic_inc_ushort_nv) + SET_SIZE(atomic_inc_16_nv) + + ENTRY(atomic_inc_32_nv) + ALTENTRY(atomic_inc_uint_nv) + ALTENTRY(atomic_inc_ulong_nv) + movl 4(%esp), %edx / %edx = target address + xorl %eax, %eax / %eax = 0 + incl %eax / %eax = 1 + lock + xaddl %eax, (%edx) / %eax = old value, inc (%edx) + incl %eax / return new value + ret + SET_SIZE(atomic_inc_ulong_nv) + SET_SIZE(atomic_inc_uint_nv) + SET_SIZE(atomic_inc_32_nv) + + /* + * NOTE: If atomic_inc_64 and atomic_inc_64_nv are ever + * separated, you need to also edit the libc i386 platform + * specific mapfile and remove the NODYNSORT attribute + * from atomic_inc_64_nv. + */ + ENTRY(atomic_inc_64) + ALTENTRY(atomic_inc_64_nv) + pushl %edi + pushl %ebx + movl 12(%esp), %edi / %edi = target address + movl (%edi), %eax + movl 4(%edi), %edx / %edx:%eax = old value +1: + xorl %ebx, %ebx + xorl %ecx, %ecx + incl %ebx / %ecx:%ebx = 1 + addl %eax, %ebx + adcl %edx, %ecx / add in the carry from inc + lock + cmpxchg8b (%edi) / try to stick it in + jne 1b + movl %ebx, %eax + movl %ecx, %edx / return new value + popl %ebx + popl %edi + ret + SET_SIZE(atomic_inc_64_nv) + SET_SIZE(atomic_inc_64) + + ENTRY(atomic_dec_8) + ALTENTRY(atomic_dec_uchar) + movl 4(%esp), %eax + lock + decb (%eax) + ret + SET_SIZE(atomic_dec_uchar) + SET_SIZE(atomic_dec_8) + + ENTRY(atomic_dec_16) + ALTENTRY(atomic_dec_ushort) + movl 4(%esp), %eax + lock + decw (%eax) + ret + SET_SIZE(atomic_dec_ushort) + SET_SIZE(atomic_dec_16) + + ENTRY(atomic_dec_32) + ALTENTRY(atomic_dec_uint) + ALTENTRY(atomic_dec_ulong) + movl 4(%esp), %eax + lock + decl (%eax) + ret + SET_SIZE(atomic_dec_ulong) + SET_SIZE(atomic_dec_uint) + SET_SIZE(atomic_dec_32) + + ENTRY(atomic_dec_8_nv) + ALTENTRY(atomic_dec_uchar_nv) + movl 4(%esp), %edx / %edx = target address + xorl %eax, %eax / zero upper bits of %eax + decb %al / %al = -1 + lock + xaddb %al, (%edx) / %al = old value, dec (%edx) + decb %al / return new value + ret + SET_SIZE(atomic_dec_uchar_nv) + SET_SIZE(atomic_dec_8_nv) + + ENTRY(atomic_dec_16_nv) + ALTENTRY(atomic_dec_ushort_nv) + movl 4(%esp), %edx / %edx = target address + xorl %eax, %eax / zero upper bits of %eax + decw %ax / %ax = -1 + lock + xaddw %ax, (%edx) / %ax = old value, dec (%edx) + decw %ax / return new value + ret + SET_SIZE(atomic_dec_ushort_nv) + SET_SIZE(atomic_dec_16_nv) + + ENTRY(atomic_dec_32_nv) + ALTENTRY(atomic_dec_uint_nv) + ALTENTRY(atomic_dec_ulong_nv) + movl 4(%esp), %edx / %edx = target address + xorl %eax, %eax / %eax = 0 + decl %eax / %eax = -1 + lock + xaddl %eax, (%edx) / %eax = old value, dec (%edx) + decl %eax / return new value + ret + SET_SIZE(atomic_dec_ulong_nv) + SET_SIZE(atomic_dec_uint_nv) + SET_SIZE(atomic_dec_32_nv) + + /* + * NOTE: If atomic_dec_64 and atomic_dec_64_nv are ever + * separated, it is important to edit the libc i386 platform + * specific mapfile and remove the NODYNSORT attribute + * from atomic_dec_64_nv. + */ + ENTRY(atomic_dec_64) + ALTENTRY(atomic_dec_64_nv) + pushl %edi + pushl %ebx + movl 12(%esp), %edi / %edi = target address + movl (%edi), %eax + movl 4(%edi), %edx / %edx:%eax = old value +1: + xorl %ebx, %ebx + xorl %ecx, %ecx + not %ecx + not %ebx / %ecx:%ebx = -1 + addl %eax, %ebx + adcl %edx, %ecx / add in the carry from inc + lock + cmpxchg8b (%edi) / try to stick it in + jne 1b + movl %ebx, %eax + movl %ecx, %edx / return new value + popl %ebx + popl %edi + ret + SET_SIZE(atomic_dec_64_nv) + SET_SIZE(atomic_dec_64) + + ENTRY(atomic_add_8) + ALTENTRY(atomic_add_char) + movl 4(%esp), %eax + movl 8(%esp), %ecx + lock + addb %cl, (%eax) + ret + SET_SIZE(atomic_add_char) + SET_SIZE(atomic_add_8) + + ENTRY(atomic_add_16) + ALTENTRY(atomic_add_short) + movl 4(%esp), %eax + movl 8(%esp), %ecx + lock + addw %cx, (%eax) + ret + SET_SIZE(atomic_add_short) + SET_SIZE(atomic_add_16) + + ENTRY(atomic_add_32) + ALTENTRY(atomic_add_int) + ALTENTRY(atomic_add_ptr) + ALTENTRY(atomic_add_long) + movl 4(%esp), %eax + movl 8(%esp), %ecx + lock + addl %ecx, (%eax) + ret + SET_SIZE(atomic_add_long) + SET_SIZE(atomic_add_ptr) + SET_SIZE(atomic_add_int) + SET_SIZE(atomic_add_32) + + ENTRY(atomic_or_8) + ALTENTRY(atomic_or_uchar) + movl 4(%esp), %eax + movb 8(%esp), %cl + lock + orb %cl, (%eax) + ret + SET_SIZE(atomic_or_uchar) + SET_SIZE(atomic_or_8) + + ENTRY(atomic_or_16) + ALTENTRY(atomic_or_ushort) + movl 4(%esp), %eax + movw 8(%esp), %cx + lock + orw %cx, (%eax) + ret + SET_SIZE(atomic_or_ushort) + SET_SIZE(atomic_or_16) + + ENTRY(atomic_or_32) + ALTENTRY(atomic_or_uint) + ALTENTRY(atomic_or_ulong) + movl 4(%esp), %eax + movl 8(%esp), %ecx + lock + orl %ecx, (%eax) + ret + SET_SIZE(atomic_or_ulong) + SET_SIZE(atomic_or_uint) + SET_SIZE(atomic_or_32) + + ENTRY(atomic_and_8) + ALTENTRY(atomic_and_uchar) + movl 4(%esp), %eax + movb 8(%esp), %cl + lock + andb %cl, (%eax) + ret + SET_SIZE(atomic_and_uchar) + SET_SIZE(atomic_and_8) + + ENTRY(atomic_and_16) + ALTENTRY(atomic_and_ushort) + movl 4(%esp), %eax + movw 8(%esp), %cx + lock + andw %cx, (%eax) + ret + SET_SIZE(atomic_and_ushort) + SET_SIZE(atomic_and_16) + + ENTRY(atomic_and_32) + ALTENTRY(atomic_and_uint) + ALTENTRY(atomic_and_ulong) + movl 4(%esp), %eax + movl 8(%esp), %ecx + lock + andl %ecx, (%eax) + ret + SET_SIZE(atomic_and_ulong) + SET_SIZE(atomic_and_uint) + SET_SIZE(atomic_and_32) + + ENTRY(atomic_add_8_nv) + ALTENTRY(atomic_add_char_nv) + movl 4(%esp), %edx / %edx = target address + movb 8(%esp), %cl / %cl = delta + movzbl %cl, %eax / %al = delta, zero extended + lock + xaddb %cl, (%edx) / %cl = old value, (%edx) = sum + addb %cl, %al / return old value plus delta + ret + SET_SIZE(atomic_add_char_nv) + SET_SIZE(atomic_add_8_nv) + + ENTRY(atomic_add_16_nv) + ALTENTRY(atomic_add_short_nv) + movl 4(%esp), %edx / %edx = target address + movw 8(%esp), %cx / %cx = delta + movzwl %cx, %eax / %ax = delta, zero extended + lock + xaddw %cx, (%edx) / %cx = old value, (%edx) = sum + addw %cx, %ax / return old value plus delta + ret + SET_SIZE(atomic_add_short_nv) + SET_SIZE(atomic_add_16_nv) + + ENTRY(atomic_add_32_nv) + ALTENTRY(atomic_add_int_nv) + ALTENTRY(atomic_add_ptr_nv) + ALTENTRY(atomic_add_long_nv) + movl 4(%esp), %edx / %edx = target address + movl 8(%esp), %eax / %eax = delta + movl %eax, %ecx / %ecx = delta + lock + xaddl %eax, (%edx) / %eax = old value, (%edx) = sum + addl %ecx, %eax / return old value plus delta + ret + SET_SIZE(atomic_add_long_nv) + SET_SIZE(atomic_add_ptr_nv) + SET_SIZE(atomic_add_int_nv) + SET_SIZE(atomic_add_32_nv) + + /* + * NOTE: If atomic_add_64 and atomic_add_64_nv are ever + * separated, it is important to edit the libc i386 platform + * specific mapfile and remove the NODYNSORT attribute + * from atomic_add_64_nv. + */ + ENTRY(atomic_add_64) + ALTENTRY(atomic_add_64_nv) + pushl %edi + pushl %ebx + movl 12(%esp), %edi / %edi = target address + movl (%edi), %eax + movl 4(%edi), %edx / %edx:%eax = old value +1: + movl 16(%esp), %ebx + movl 20(%esp), %ecx / %ecx:%ebx = delta + addl %eax, %ebx + adcl %edx, %ecx / %ecx:%ebx = new value + lock + cmpxchg8b (%edi) / try to stick it in + jne 1b + movl %ebx, %eax + movl %ecx, %edx / return new value + popl %ebx + popl %edi + ret + SET_SIZE(atomic_add_64_nv) + SET_SIZE(atomic_add_64) + + ENTRY(atomic_or_8_nv) + ALTENTRY(atomic_or_uchar_nv) + movl 4(%esp), %edx / %edx = target address + movb (%edx), %al / %al = old value +1: + movl 8(%esp), %ecx / %ecx = delta + orb %al, %cl / %cl = new value + lock + cmpxchgb %cl, (%edx) / try to stick it in + jne 1b + movzbl %cl, %eax / return new value + ret + SET_SIZE(atomic_or_uchar_nv) + SET_SIZE(atomic_or_8_nv) + + ENTRY(atomic_or_16_nv) + ALTENTRY(atomic_or_ushort_nv) + movl 4(%esp), %edx / %edx = target address + movw (%edx), %ax / %ax = old value +1: + movl 8(%esp), %ecx / %ecx = delta + orw %ax, %cx / %cx = new value + lock + cmpxchgw %cx, (%edx) / try to stick it in + jne 1b + movzwl %cx, %eax / return new value + ret + SET_SIZE(atomic_or_ushort_nv) + SET_SIZE(atomic_or_16_nv) + + ENTRY(atomic_or_32_nv) + ALTENTRY(atomic_or_uint_nv) + ALTENTRY(atomic_or_ulong_nv) + movl 4(%esp), %edx / %edx = target address + movl (%edx), %eax / %eax = old value +1: + movl 8(%esp), %ecx / %ecx = delta + orl %eax, %ecx / %ecx = new value + lock + cmpxchgl %ecx, (%edx) / try to stick it in + jne 1b + movl %ecx, %eax / return new value + ret + SET_SIZE(atomic_or_ulong_nv) + SET_SIZE(atomic_or_uint_nv) + SET_SIZE(atomic_or_32_nv) + + /* + * NOTE: If atomic_or_64 and atomic_or_64_nv are ever + * separated, it is important to edit the libc i386 platform + * specific mapfile and remove the NODYNSORT attribute + * from atomic_or_64_nv. + */ + ENTRY(atomic_or_64) + ALTENTRY(atomic_or_64_nv) + pushl %edi + pushl %ebx + movl 12(%esp), %edi / %edi = target address + movl (%edi), %eax + movl 4(%edi), %edx / %edx:%eax = old value +1: + movl 16(%esp), %ebx + movl 20(%esp), %ecx / %ecx:%ebx = delta + orl %eax, %ebx + orl %edx, %ecx / %ecx:%ebx = new value + lock + cmpxchg8b (%edi) / try to stick it in + jne 1b + movl %ebx, %eax + movl %ecx, %edx / return new value + popl %ebx + popl %edi + ret + SET_SIZE(atomic_or_64_nv) + SET_SIZE(atomic_or_64) + + ENTRY(atomic_and_8_nv) + ALTENTRY(atomic_and_uchar_nv) + movl 4(%esp), %edx / %edx = target address + movb (%edx), %al / %al = old value +1: + movl 8(%esp), %ecx / %ecx = delta + andb %al, %cl / %cl = new value + lock + cmpxchgb %cl, (%edx) / try to stick it in + jne 1b + movzbl %cl, %eax / return new value + ret + SET_SIZE(atomic_and_uchar_nv) + SET_SIZE(atomic_and_8_nv) + + ENTRY(atomic_and_16_nv) + ALTENTRY(atomic_and_ushort_nv) + movl 4(%esp), %edx / %edx = target address + movw (%edx), %ax / %ax = old value +1: + movl 8(%esp), %ecx / %ecx = delta + andw %ax, %cx / %cx = new value + lock + cmpxchgw %cx, (%edx) / try to stick it in + jne 1b + movzwl %cx, %eax / return new value + ret + SET_SIZE(atomic_and_ushort_nv) + SET_SIZE(atomic_and_16_nv) + + ENTRY(atomic_and_32_nv) + ALTENTRY(atomic_and_uint_nv) + ALTENTRY(atomic_and_ulong_nv) + movl 4(%esp), %edx / %edx = target address + movl (%edx), %eax / %eax = old value +1: + movl 8(%esp), %ecx / %ecx = delta + andl %eax, %ecx / %ecx = new value + lock + cmpxchgl %ecx, (%edx) / try to stick it in + jne 1b + movl %ecx, %eax / return new value + ret + SET_SIZE(atomic_and_ulong_nv) + SET_SIZE(atomic_and_uint_nv) + SET_SIZE(atomic_and_32_nv) + + /* + * NOTE: If atomic_and_64 and atomic_and_64_nv are ever + * separated, it is important to edit the libc i386 platform + * specific mapfile and remove the NODYNSORT attribute + * from atomic_and_64_nv. + */ + ENTRY(atomic_and_64) + ALTENTRY(atomic_and_64_nv) + pushl %edi + pushl %ebx + movl 12(%esp), %edi / %edi = target address + movl (%edi), %eax + movl 4(%edi), %edx / %edx:%eax = old value +1: + movl 16(%esp), %ebx + movl 20(%esp), %ecx / %ecx:%ebx = delta + andl %eax, %ebx + andl %edx, %ecx / %ecx:%ebx = new value + lock + cmpxchg8b (%edi) / try to stick it in + jne 1b + movl %ebx, %eax + movl %ecx, %edx / return new value + popl %ebx + popl %edi + ret + SET_SIZE(atomic_and_64_nv) + SET_SIZE(atomic_and_64) + + ENTRY(atomic_cas_8) + ALTENTRY(atomic_cas_uchar) + movl 4(%esp), %edx + movzbl 8(%esp), %eax + movb 12(%esp), %cl + lock + cmpxchgb %cl, (%edx) + ret + SET_SIZE(atomic_cas_uchar) + SET_SIZE(atomic_cas_8) + + ENTRY(atomic_cas_16) + ALTENTRY(atomic_cas_ushort) + movl 4(%esp), %edx + movzwl 8(%esp), %eax + movw 12(%esp), %cx + lock + cmpxchgw %cx, (%edx) + ret + SET_SIZE(atomic_cas_ushort) + SET_SIZE(atomic_cas_16) + + ENTRY(atomic_cas_32) + ALTENTRY(atomic_cas_uint) + ALTENTRY(atomic_cas_ulong) + ALTENTRY(atomic_cas_ptr) + movl 4(%esp), %edx + movl 8(%esp), %eax + movl 12(%esp), %ecx + lock + cmpxchgl %ecx, (%edx) + ret + SET_SIZE(atomic_cas_ptr) + SET_SIZE(atomic_cas_ulong) + SET_SIZE(atomic_cas_uint) + SET_SIZE(atomic_cas_32) + + ENTRY(atomic_cas_64) + pushl %ebx + pushl %esi + movl 12(%esp), %esi + movl 16(%esp), %eax + movl 20(%esp), %edx + movl 24(%esp), %ebx + movl 28(%esp), %ecx + lock + cmpxchg8b (%esi) + popl %esi + popl %ebx + ret + SET_SIZE(atomic_cas_64) + + ENTRY(atomic_swap_8) + ALTENTRY(atomic_swap_uchar) + movl 4(%esp), %edx + movzbl 8(%esp), %eax + lock + xchgb %al, (%edx) + ret + SET_SIZE(atomic_swap_uchar) + SET_SIZE(atomic_swap_8) + + ENTRY(atomic_swap_16) + ALTENTRY(atomic_swap_ushort) + movl 4(%esp), %edx + movzwl 8(%esp), %eax + lock + xchgw %ax, (%edx) + ret + SET_SIZE(atomic_swap_ushort) + SET_SIZE(atomic_swap_16) + + ENTRY(atomic_swap_32) + ALTENTRY(atomic_swap_uint) + ALTENTRY(atomic_swap_ptr) + ALTENTRY(atomic_swap_ulong) + movl 4(%esp), %edx + movl 8(%esp), %eax + lock + xchgl %eax, (%edx) + ret + SET_SIZE(atomic_swap_ulong) + SET_SIZE(atomic_swap_ptr) + SET_SIZE(atomic_swap_uint) + SET_SIZE(atomic_swap_32) + + ENTRY(atomic_swap_64) + pushl %esi + pushl %ebx + movl 12(%esp), %esi + movl 16(%esp), %ebx + movl 20(%esp), %ecx + movl (%esi), %eax + movl 4(%esi), %edx / %edx:%eax = old value +1: + lock + cmpxchg8b (%esi) + jne 1b + popl %ebx + popl %esi + ret + SET_SIZE(atomic_swap_64) + + ENTRY(atomic_set_long_excl) + movl 4(%esp), %edx / %edx = target address + movl 8(%esp), %ecx / %ecx = bit id + xorl %eax, %eax + lock + btsl %ecx, (%edx) + jnc 1f + decl %eax / return -1 +1: + ret + SET_SIZE(atomic_set_long_excl) + + ENTRY(atomic_clear_long_excl) + movl 4(%esp), %edx / %edx = target address + movl 8(%esp), %ecx / %ecx = bit id + xorl %eax, %eax + lock + btrl %ecx, (%edx) + jc 1f + decl %eax / return -1 +1: + ret + SET_SIZE(atomic_clear_long_excl) + +#if !defined(_KERNEL) + + /* + * NOTE: membar_enter, membar_exit, membar_producer, and + * membar_consumer are all identical routines. We define them + * separately, instead of using ALTENTRY definitions to alias them + * together, so that DTrace and debuggers will see a unique address + * for them, allowing more accurate tracing. + */ + + + ENTRY(membar_enter) + lock + xorl $0, (%esp) + ret + SET_SIZE(membar_enter) + + ENTRY(membar_exit) + lock + xorl $0, (%esp) + ret + SET_SIZE(membar_exit) + + ENTRY(membar_producer) + lock + xorl $0, (%esp) + ret + SET_SIZE(membar_producer) + + ENTRY(membar_consumer) + lock + xorl $0, (%esp) + ret + SET_SIZE(membar_consumer) + +#endif /* !_KERNEL */ |