Index: head/sys/mips/include/atomic.h =================================================================== --- head/sys/mips/include/atomic.h (revision 327096) +++ head/sys/mips/include/atomic.h (revision 327097) @@ -1,776 +1,758 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 1998 Doug Rabson * 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. * * from: src/sys/alpha/include/atomic.h,v 1.21.2.3 2005/10/06 18:12:05 jhb * $FreeBSD$ */ #ifndef _MACHINE_ATOMIC_H_ #define _MACHINE_ATOMIC_H_ #ifndef _SYS_CDEFS_H_ #error this file needs sys/cdefs.h as a prerequisite #endif #include /* * Note: All the 64-bit atomic operations are only atomic when running * in 64-bit mode. It is assumed that code compiled for n32 and n64 * fits into this definition and no further safeties are needed. * * It is also assumed that the add, subtract and other arithmetic is * done on numbers not pointers. The special rules for n32 pointers * do not have atomic operations defined for them, but generally shouldn't * need atomic operations. */ #ifndef __MIPS_PLATFORM_SYNC_NOPS #define __MIPS_PLATFORM_SYNC_NOPS "" #endif static __inline void mips_sync(void) { __asm __volatile (".set noreorder\n" "\tsync\n" __MIPS_PLATFORM_SYNC_NOPS ".set reorder\n" : : : "memory"); } #define mb() mips_sync() #define wmb() mips_sync() #define rmb() mips_sync() /* * Various simple arithmetic on memory which is atomic in the presence * of interrupts and SMP safe. */ void atomic_set_8(__volatile uint8_t *, uint8_t); void atomic_clear_8(__volatile uint8_t *, uint8_t); void atomic_add_8(__volatile uint8_t *, uint8_t); void atomic_subtract_8(__volatile uint8_t *, uint8_t); void atomic_set_16(__volatile uint16_t *, uint16_t); void atomic_clear_16(__volatile uint16_t *, uint16_t); void atomic_add_16(__volatile uint16_t *, uint16_t); void atomic_subtract_16(__volatile uint16_t *, uint16_t); static __inline void atomic_set_32(__volatile uint32_t *p, uint32_t v) { uint32_t temp; __asm __volatile ( "1:\tll %0, %3\n\t" /* load old value */ "or %0, %2, %0\n\t" /* calculate new value */ "sc %0, %1\n\t" /* attempt to store */ "beqz %0, 1b\n\t" /* spin if failed */ : "=&r" (temp), "=m" (*p) : "r" (v), "m" (*p) : "memory"); } static __inline void atomic_clear_32(__volatile uint32_t *p, uint32_t v) { uint32_t temp; v = ~v; __asm __volatile ( "1:\tll %0, %3\n\t" /* load old value */ "and %0, %2, %0\n\t" /* calculate new value */ "sc %0, %1\n\t" /* attempt to store */ "beqz %0, 1b\n\t" /* spin if failed */ : "=&r" (temp), "=m" (*p) : "r" (v), "m" (*p) : "memory"); } static __inline void atomic_add_32(__volatile uint32_t *p, uint32_t v) { uint32_t temp; __asm __volatile ( "1:\tll %0, %3\n\t" /* load old value */ "addu %0, %2, %0\n\t" /* calculate new value */ "sc %0, %1\n\t" /* attempt to store */ "beqz %0, 1b\n\t" /* spin if failed */ : "=&r" (temp), "=m" (*p) : "r" (v), "m" (*p) : "memory"); } static __inline void atomic_subtract_32(__volatile uint32_t *p, uint32_t v) { uint32_t temp; __asm __volatile ( "1:\tll %0, %3\n\t" /* load old value */ "subu %0, %2\n\t" /* calculate new value */ "sc %0, %1\n\t" /* attempt to store */ "beqz %0, 1b\n\t" /* spin if failed */ : "=&r" (temp), "=m" (*p) : "r" (v), "m" (*p) : "memory"); } static __inline uint32_t atomic_readandclear_32(__volatile uint32_t *addr) { uint32_t result,temp; __asm __volatile ( "1:\tll %0,%3\n\t" /* load current value, asserting lock */ "li %1,0\n\t" /* value to store */ "sc %1,%2\n\t" /* attempt to store */ "beqz %1, 1b\n\t" /* if the store failed, spin */ : "=&r"(result), "=&r"(temp), "=m" (*addr) : "m" (*addr) : "memory"); return result; } static __inline uint32_t atomic_readandset_32(__volatile uint32_t *addr, uint32_t value) { uint32_t result,temp; __asm __volatile ( "1:\tll %0,%3\n\t" /* load current value, asserting lock */ "or %1,$0,%4\n\t" "sc %1,%2\n\t" /* attempt to store */ "beqz %1, 1b\n\t" /* if the store failed, spin */ : "=&r"(result), "=&r"(temp), "=m" (*addr) : "m" (*addr), "r" (value) : "memory"); return result; } #if defined(__mips_n64) || defined(__mips_n32) static __inline void atomic_set_64(__volatile uint64_t *p, uint64_t v) { uint64_t temp; __asm __volatile ( "1:\n\t" "lld %0, %3\n\t" /* load old value */ "or %0, %2, %0\n\t" /* calculate new value */ "scd %0, %1\n\t" /* attempt to store */ "beqz %0, 1b\n\t" /* spin if failed */ : "=&r" (temp), "=m" (*p) : "r" (v), "m" (*p) : "memory"); } static __inline void atomic_clear_64(__volatile uint64_t *p, uint64_t v) { uint64_t temp; v = ~v; __asm __volatile ( "1:\n\t" "lld %0, %3\n\t" /* load old value */ "and %0, %2, %0\n\t" /* calculate new value */ "scd %0, %1\n\t" /* attempt to store */ "beqz %0, 1b\n\t" /* spin if failed */ : "=&r" (temp), "=m" (*p) : "r" (v), "m" (*p) : "memory"); } static __inline void atomic_add_64(__volatile uint64_t *p, uint64_t v) { uint64_t temp; __asm __volatile ( "1:\n\t" "lld %0, %3\n\t" /* load old value */ "daddu %0, %2, %0\n\t" /* calculate new value */ "scd %0, %1\n\t" /* attempt to store */ "beqz %0, 1b\n\t" /* spin if failed */ : "=&r" (temp), "=m" (*p) : "r" (v), "m" (*p) : "memory"); } static __inline void atomic_subtract_64(__volatile uint64_t *p, uint64_t v) { uint64_t temp; __asm __volatile ( "1:\n\t" "lld %0, %3\n\t" /* load old value */ "dsubu %0, %2\n\t" /* calculate new value */ "scd %0, %1\n\t" /* attempt to store */ "beqz %0, 1b\n\t" /* spin if failed */ : "=&r" (temp), "=m" (*p) : "r" (v), "m" (*p) : "memory"); } static __inline uint64_t atomic_readandclear_64(__volatile uint64_t *addr) { uint64_t result,temp; __asm __volatile ( "1:\n\t" "lld %0, %3\n\t" /* load old value */ "li %1, 0\n\t" /* value to store */ "scd %1, %2\n\t" /* attempt to store */ "beqz %1, 1b\n\t" /* if the store failed, spin */ : "=&r"(result), "=&r"(temp), "=m" (*addr) : "m" (*addr) : "memory"); return result; } static __inline uint64_t atomic_readandset_64(__volatile uint64_t *addr, uint64_t value) { uint64_t result,temp; __asm __volatile ( "1:\n\t" "lld %0,%3\n\t" /* Load old value*/ "or %1,$0,%4\n\t" "scd %1,%2\n\t" /* attempt to store */ "beqz %1, 1b\n\t" /* if the store failed, spin */ : "=&r"(result), "=&r"(temp), "=m" (*addr) : "m" (*addr), "r" (value) : "memory"); return result; } #endif #define ATOMIC_ACQ_REL(NAME, WIDTH) \ static __inline void \ atomic_##NAME##_acq_##WIDTH(__volatile uint##WIDTH##_t *p, uint##WIDTH##_t v)\ { \ atomic_##NAME##_##WIDTH(p, v); \ mips_sync(); \ } \ \ static __inline void \ atomic_##NAME##_rel_##WIDTH(__volatile uint##WIDTH##_t *p, uint##WIDTH##_t v)\ { \ mips_sync(); \ atomic_##NAME##_##WIDTH(p, v); \ } /* Variants of simple arithmetic with memory barriers. */ ATOMIC_ACQ_REL(set, 8) ATOMIC_ACQ_REL(clear, 8) ATOMIC_ACQ_REL(add, 8) ATOMIC_ACQ_REL(subtract, 8) ATOMIC_ACQ_REL(set, 16) ATOMIC_ACQ_REL(clear, 16) ATOMIC_ACQ_REL(add, 16) ATOMIC_ACQ_REL(subtract, 16) ATOMIC_ACQ_REL(set, 32) ATOMIC_ACQ_REL(clear, 32) ATOMIC_ACQ_REL(add, 32) ATOMIC_ACQ_REL(subtract, 32) #if defined(__mips_n64) || defined(__mips_n32) ATOMIC_ACQ_REL(set, 64) ATOMIC_ACQ_REL(clear, 64) ATOMIC_ACQ_REL(add, 64) ATOMIC_ACQ_REL(subtract, 64) #endif #undef ATOMIC_ACQ_REL /* * We assume that a = b will do atomic loads and stores. */ #define ATOMIC_STORE_LOAD(WIDTH) \ static __inline uint##WIDTH##_t \ atomic_load_acq_##WIDTH(__volatile uint##WIDTH##_t *p) \ { \ uint##WIDTH##_t v; \ \ v = *p; \ mips_sync(); \ return (v); \ } \ \ static __inline void \ atomic_store_rel_##WIDTH(__volatile uint##WIDTH##_t *p, uint##WIDTH##_t v)\ { \ mips_sync(); \ *p = v; \ } ATOMIC_STORE_LOAD(32) ATOMIC_STORE_LOAD(64) -#if !defined(__mips_n64) && !defined(__mips_n32) -void atomic_store_64(__volatile uint64_t *, uint64_t); -uint64_t atomic_load_64(__volatile uint64_t *); -#elif defined (__mips_n32) -static __inline void -atomic_store_64(__volatile uint64_t *p, uint64_t v) -{ - *p = v; -} - -static __inline uint64_t -atomic_load_64(__volatile uint64_t *p) -{ - return (*p); -} -/* #else atomic_common.h definitions of atomic_load/store_64 are used */ -#endif - #undef ATOMIC_STORE_LOAD /* * Atomically compare the value stored at *p with cmpval and if the * two values are equal, update the value of *p with newval. Returns * zero if the compare failed, nonzero otherwise. */ static __inline uint32_t atomic_cmpset_32(__volatile uint32_t *p, uint32_t cmpval, uint32_t newval) { uint32_t ret; __asm __volatile ( "1:\tll %0, %4\n\t" /* load old value */ "bne %0, %2, 2f\n\t" /* compare */ "move %0, %3\n\t" /* value to store */ "sc %0, %1\n\t" /* attempt to store */ "beqz %0, 1b\n\t" /* if it failed, spin */ "j 3f\n\t" "2:\n\t" "li %0, 0\n\t" "3:\n" : "=&r" (ret), "=m" (*p) : "r" (cmpval), "r" (newval), "m" (*p) : "memory"); return ret; } /* * Atomically compare the value stored at *p with cmpval and if the * two values are equal, update the value of *p with newval. Returns * zero if the compare failed, nonzero otherwise. */ static __inline uint32_t atomic_cmpset_acq_32(__volatile uint32_t *p, uint32_t cmpval, uint32_t newval) { int retval; retval = atomic_cmpset_32(p, cmpval, newval); mips_sync(); return (retval); } static __inline uint32_t atomic_cmpset_rel_32(__volatile uint32_t *p, uint32_t cmpval, uint32_t newval) { mips_sync(); return (atomic_cmpset_32(p, cmpval, newval)); } static __inline uint32_t atomic_fcmpset_32(__volatile uint32_t *p, uint32_t *cmpval, uint32_t newval) { uint32_t ret; __asm __volatile ( "1:\n\t" "ll %0, %1\n\t" /* load old value */ "bne %0, %4, 2f\n\t" /* compare */ "move %0, %3\n\t" /* value to store */ "sc %0, %1\n\t" /* attempt to store */ "beqz %0, 1b\n\t" /* if it failed, spin */ "j 3f\n\t" "2:\n\t" "sw %0, %2\n\t" /* save old value */ "li %0, 0\n\t" "3:\n" : "=&r" (ret), "+m" (*p), "=m" (*cmpval) : "r" (newval), "r" (*cmpval) : "memory"); return ret; } static __inline uint32_t atomic_fcmpset_acq_32(__volatile uint32_t *p, uint32_t *cmpval, uint32_t newval) { int retval; retval = atomic_fcmpset_32(p, cmpval, newval); mips_sync(); return (retval); } static __inline uint32_t atomic_fcmpset_rel_32(__volatile uint32_t *p, uint32_t *cmpval, uint32_t newval) { mips_sync(); return (atomic_fcmpset_32(p, cmpval, newval)); } /* * Atomically add the value of v to the integer pointed to by p and return * the previous value of *p. */ static __inline uint32_t atomic_fetchadd_32(__volatile uint32_t *p, uint32_t v) { uint32_t value, temp; __asm __volatile ( "1:\tll %0, %1\n\t" /* load old value */ "addu %2, %3, %0\n\t" /* calculate new value */ "sc %2, %1\n\t" /* attempt to store */ "beqz %2, 1b\n\t" /* spin if failed */ : "=&r" (value), "=m" (*p), "=&r" (temp) : "r" (v), "m" (*p)); return (value); } #if defined(__mips_n64) || defined(__mips_n32) /* * Atomically compare the value stored at *p with cmpval and if the * two values are equal, update the value of *p with newval. Returns * zero if the compare failed, nonzero otherwise. */ static __inline uint64_t atomic_cmpset_64(__volatile uint64_t *p, uint64_t cmpval, uint64_t newval) { uint64_t ret; __asm __volatile ( "1:\n\t" "lld %0, %4\n\t" /* load old value */ "bne %0, %2, 2f\n\t" /* compare */ "move %0, %3\n\t" /* value to store */ "scd %0, %1\n\t" /* attempt to store */ "beqz %0, 1b\n\t" /* if it failed, spin */ "j 3f\n\t" "2:\n\t" "li %0, 0\n\t" "3:\n" : "=&r" (ret), "=m" (*p) : "r" (cmpval), "r" (newval), "m" (*p) : "memory"); return ret; } /* * Atomically compare the value stored at *p with cmpval and if the * two values are equal, update the value of *p with newval. Returns * zero if the compare failed, nonzero otherwise. */ static __inline uint64_t atomic_cmpset_acq_64(__volatile uint64_t *p, uint64_t cmpval, uint64_t newval) { int retval; retval = atomic_cmpset_64(p, cmpval, newval); mips_sync(); return (retval); } static __inline uint64_t atomic_cmpset_rel_64(__volatile uint64_t *p, uint64_t cmpval, uint64_t newval) { mips_sync(); return (atomic_cmpset_64(p, cmpval, newval)); } static __inline uint32_t atomic_fcmpset_64(__volatile uint64_t *p, uint64_t *cmpval, uint64_t newval) { uint32_t ret; __asm __volatile ( "1:\n\t" "lld %0, %1\n\t" /* load old value */ "bne %0, %4, 2f\n\t" /* compare */ "move %0, %3\n\t" /* value to store */ "scd %0, %1\n\t" /* attempt to store */ "beqz %0, 1b\n\t" /* if it failed, spin */ "j 3f\n\t" "2:\n\t" "sd %0, %2\n\t" /* save old value */ "li %0, 0\n\t" "3:\n" : "=&r" (ret), "+m" (*p), "=m" (*cmpval) : "r" (newval), "r" (*cmpval) : "memory"); return ret; } static __inline uint64_t atomic_fcmpset_acq_64(__volatile uint64_t *p, uint64_t *cmpval, uint64_t newval) { int retval; retval = atomic_fcmpset_64(p, cmpval, newval); mips_sync(); return (retval); } static __inline uint64_t atomic_fcmpset_rel_64(__volatile uint64_t *p, uint64_t *cmpval, uint64_t newval) { mips_sync(); return (atomic_fcmpset_64(p, cmpval, newval)); } /* * Atomically add the value of v to the integer pointed to by p and return * the previous value of *p. */ static __inline uint64_t atomic_fetchadd_64(__volatile uint64_t *p, uint64_t v) { uint64_t value, temp; __asm __volatile ( "1:\n\t" "lld %0, %1\n\t" /* load old value */ "daddu %2, %3, %0\n\t" /* calculate new value */ "scd %2, %1\n\t" /* attempt to store */ "beqz %2, 1b\n\t" /* spin if failed */ : "=&r" (value), "=m" (*p), "=&r" (temp) : "r" (v), "m" (*p)); return (value); } #endif static __inline void atomic_thread_fence_acq(void) { mips_sync(); } static __inline void atomic_thread_fence_rel(void) { mips_sync(); } static __inline void atomic_thread_fence_acq_rel(void) { mips_sync(); } static __inline void atomic_thread_fence_seq_cst(void) { mips_sync(); } /* Operations on chars. */ #define atomic_set_char atomic_set_8 #define atomic_set_acq_char atomic_set_acq_8 #define atomic_set_rel_char atomic_set_rel_8 #define atomic_clear_char atomic_clear_8 #define atomic_clear_acq_char atomic_clear_acq_8 #define atomic_clear_rel_char atomic_clear_rel_8 #define atomic_add_char atomic_add_8 #define atomic_add_acq_char atomic_add_acq_8 #define atomic_add_rel_char atomic_add_rel_8 #define atomic_subtract_char atomic_subtract_8 #define atomic_subtract_acq_char atomic_subtract_acq_8 #define atomic_subtract_rel_char atomic_subtract_rel_8 /* Operations on shorts. */ #define atomic_set_short atomic_set_16 #define atomic_set_acq_short atomic_set_acq_16 #define atomic_set_rel_short atomic_set_rel_16 #define atomic_clear_short atomic_clear_16 #define atomic_clear_acq_short atomic_clear_acq_16 #define atomic_clear_rel_short atomic_clear_rel_16 #define atomic_add_short atomic_add_16 #define atomic_add_acq_short atomic_add_acq_16 #define atomic_add_rel_short atomic_add_rel_16 #define atomic_subtract_short atomic_subtract_16 #define atomic_subtract_acq_short atomic_subtract_acq_16 #define atomic_subtract_rel_short atomic_subtract_rel_16 /* Operations on ints. */ #define atomic_set_int atomic_set_32 #define atomic_set_acq_int atomic_set_acq_32 #define atomic_set_rel_int atomic_set_rel_32 #define atomic_clear_int atomic_clear_32 #define atomic_clear_acq_int atomic_clear_acq_32 #define atomic_clear_rel_int atomic_clear_rel_32 #define atomic_add_int atomic_add_32 #define atomic_add_acq_int atomic_add_acq_32 #define atomic_add_rel_int atomic_add_rel_32 #define atomic_subtract_int atomic_subtract_32 #define atomic_subtract_acq_int atomic_subtract_acq_32 #define atomic_subtract_rel_int atomic_subtract_rel_32 #define atomic_cmpset_int atomic_cmpset_32 #define atomic_cmpset_acq_int atomic_cmpset_acq_32 #define atomic_cmpset_rel_int atomic_cmpset_rel_32 #define atomic_fcmpset_int atomic_fcmpset_32 #define atomic_fcmpset_acq_int atomic_fcmpset_acq_32 #define atomic_fcmpset_rel_int atomic_fcmpset_rel_32 #define atomic_load_acq_int atomic_load_acq_32 #define atomic_store_rel_int atomic_store_rel_32 #define atomic_readandclear_int atomic_readandclear_32 #define atomic_readandset_int atomic_readandset_32 #define atomic_fetchadd_int atomic_fetchadd_32 /* * I think the following is right, even for n32. For n32 the pointers * are still 32-bits, so we need to operate on them as 32-bit quantities, * even though they are sign extended in operation. For longs, there's * no question because they are always 32-bits. */ #ifdef __mips_n64 /* Operations on longs. */ #define atomic_set_long atomic_set_64 #define atomic_set_acq_long atomic_set_acq_64 #define atomic_set_rel_long atomic_set_rel_64 #define atomic_clear_long atomic_clear_64 #define atomic_clear_acq_long atomic_clear_acq_64 #define atomic_clear_rel_long atomic_clear_rel_64 #define atomic_add_long atomic_add_64 #define atomic_add_acq_long atomic_add_acq_64 #define atomic_add_rel_long atomic_add_rel_64 #define atomic_subtract_long atomic_subtract_64 #define atomic_subtract_acq_long atomic_subtract_acq_64 #define atomic_subtract_rel_long atomic_subtract_rel_64 #define atomic_cmpset_long atomic_cmpset_64 #define atomic_cmpset_acq_long atomic_cmpset_acq_64 #define atomic_cmpset_rel_long atomic_cmpset_rel_64 #define atomic_fcmpset_long atomic_fcmpset_64 #define atomic_fcmpset_acq_long atomic_fcmpset_acq_64 #define atomic_fcmpset_rel_long atomic_fcmpset_rel_64 #define atomic_load_acq_long atomic_load_acq_64 #define atomic_store_rel_long atomic_store_rel_64 #define atomic_fetchadd_long atomic_fetchadd_64 #define atomic_readandclear_long atomic_readandclear_64 #else /* !__mips_n64 */ /* Operations on longs. */ #define atomic_set_long(p, v) \ atomic_set_32((volatile u_int *)(p), (u_int)(v)) #define atomic_set_acq_long(p, v) \ atomic_set_acq_32((volatile u_int *)(p), (u_int)(v)) #define atomic_set_rel_long(p, v) \ atomic_set_rel_32((volatile u_int *)(p), (u_int)(v)) #define atomic_clear_long(p, v) \ atomic_clear_32((volatile u_int *)(p), (u_int)(v)) #define atomic_clear_acq_long(p, v) \ atomic_clear_acq_32((volatile u_int *)(p), (u_int)(v)) #define atomic_clear_rel_long(p, v) \ atomic_clear_rel_32((volatile u_int *)(p), (u_int)(v)) #define atomic_add_long(p, v) \ atomic_add_32((volatile u_int *)(p), (u_int)(v)) #define atomic_add_acq_long(p, v) \ atomic_add_32((volatile u_int *)(p), (u_int)(v)) #define atomic_add_rel_long(p, v) \ atomic_add_32((volatile u_int *)(p), (u_int)(v)) #define atomic_subtract_long(p, v) \ atomic_subtract_32((volatile u_int *)(p), (u_int)(v)) #define atomic_subtract_acq_long(p, v) \ atomic_subtract_acq_32((volatile u_int *)(p), (u_int)(v)) #define atomic_subtract_rel_long(p, v) \ atomic_subtract_rel_32((volatile u_int *)(p), (u_int)(v)) #define atomic_cmpset_long(p, cmpval, newval) \ atomic_cmpset_32((volatile u_int *)(p), (u_int)(cmpval), \ (u_int)(newval)) #define atomic_cmpset_acq_long(p, cmpval, newval) \ atomic_cmpset_acq_32((volatile u_int *)(p), (u_int)(cmpval), \ (u_int)(newval)) #define atomic_cmpset_rel_long(p, cmpval, newval) \ atomic_cmpset_rel_32((volatile u_int *)(p), (u_int)(cmpval), \ (u_int)(newval)) #define atomic_fcmpset_long(p, cmpval, newval) \ atomic_fcmpset_32((volatile u_int *)(p), (u_int *)(cmpval), \ (u_int)(newval)) #define atomic_fcmpset_acq_long(p, cmpval, newval) \ atomic_fcmpset_acq_32((volatile u_int *)(p), (u_int *)(cmpval), \ (u_int)(newval)) #define atomic_fcmpset_rel_long(p, cmpval, newval) \ atomic_fcmpset_rel_32((volatile u_int *)(p), (u_int *)(cmpval), \ (u_int)(newval)) #define atomic_load_acq_long(p) \ (u_long)atomic_load_acq_32((volatile u_int *)(p)) #define atomic_store_rel_long(p, v) \ atomic_store_rel_32((volatile u_int *)(p), (u_int)(v)) #define atomic_fetchadd_long(p, v) \ atomic_fetchadd_32((volatile u_int *)(p), (u_int)(v)) #define atomic_readandclear_long(p) \ atomic_readandclear_32((volatile u_int *)(p)) #endif /* __mips_n64 */ /* Operations on pointers. */ #define atomic_set_ptr atomic_set_long #define atomic_set_acq_ptr atomic_set_acq_long #define atomic_set_rel_ptr atomic_set_rel_long #define atomic_clear_ptr atomic_clear_long #define atomic_clear_acq_ptr atomic_clear_acq_long #define atomic_clear_rel_ptr atomic_clear_rel_long #define atomic_add_ptr atomic_add_long #define atomic_add_acq_ptr atomic_add_acq_long #define atomic_add_rel_ptr atomic_add_rel_long #define atomic_subtract_ptr atomic_subtract_long #define atomic_subtract_acq_ptr atomic_subtract_acq_long #define atomic_subtract_rel_ptr atomic_subtract_rel_long #define atomic_cmpset_ptr atomic_cmpset_long #define atomic_cmpset_acq_ptr atomic_cmpset_acq_long #define atomic_cmpset_rel_ptr atomic_cmpset_rel_long #define atomic_fcmpset_ptr atomic_fcmpset_long #define atomic_fcmpset_acq_ptr atomic_fcmpset_acq_long #define atomic_fcmpset_rel_ptr atomic_fcmpset_rel_long #define atomic_load_acq_ptr atomic_load_acq_long #define atomic_store_rel_ptr atomic_store_rel_long #define atomic_readandclear_ptr atomic_readandclear_long #endif /* ! _MACHINE_ATOMIC_H_ */ Index: head/sys/mips/mips/db_interface.c =================================================================== --- head/sys/mips/mips/db_interface.c (revision 327096) +++ head/sys/mips/mips/db_interface.c (revision 327097) @@ -1,350 +1,350 @@ /* $OpenBSD: db_machdep.c,v 1.2 1998/09/15 10:50:13 pefo Exp $ */ /*- * SPDX-License-Identifier: BSD-4-Clause * * Copyright (c) 1998 Per Fogelstrom, Opsycon AB * * 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. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed under OpenBSD by * Per Fogelstrom, Opsycon AB, Sweden. * 4. 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. * * JNPR: db_interface.c,v 1.6.2.1 2007/08/29 12:24:49 girish */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static db_varfcn_t db_frame; #define DB_OFFSET(x) (db_expr_t *)offsetof(struct trapframe, x) struct db_variable db_regs[] = { { "at", DB_OFFSET(ast), db_frame }, { "v0", DB_OFFSET(v0), db_frame }, { "v1", DB_OFFSET(v1), db_frame }, { "a0", DB_OFFSET(a0), db_frame }, { "a1", DB_OFFSET(a1), db_frame }, { "a2", DB_OFFSET(a2), db_frame }, { "a3", DB_OFFSET(a3), db_frame }, #if defined(__mips_n32) || defined(__mips_n64) { "a4", DB_OFFSET(a4), db_frame }, { "a5", DB_OFFSET(a5), db_frame }, { "a6", DB_OFFSET(a6), db_frame }, { "a7", DB_OFFSET(a7), db_frame }, { "t0", DB_OFFSET(t0), db_frame }, { "t1", DB_OFFSET(t1), db_frame }, { "t2", DB_OFFSET(t2), db_frame }, { "t3", DB_OFFSET(t3), db_frame }, #else { "t0", DB_OFFSET(t0), db_frame }, { "t1", DB_OFFSET(t1), db_frame }, { "t2", DB_OFFSET(t2), db_frame }, { "t3", DB_OFFSET(t3), db_frame }, { "t4", DB_OFFSET(t4), db_frame }, { "t5", DB_OFFSET(t5), db_frame }, { "t6", DB_OFFSET(t6), db_frame }, { "t7", DB_OFFSET(t7), db_frame }, #endif { "s0", DB_OFFSET(s0), db_frame }, { "s1", DB_OFFSET(s1), db_frame }, { "s2", DB_OFFSET(s2), db_frame }, { "s3", DB_OFFSET(s3), db_frame }, { "s4", DB_OFFSET(s4), db_frame }, { "s5", DB_OFFSET(s5), db_frame }, { "s6", DB_OFFSET(s6), db_frame }, { "s7", DB_OFFSET(s7), db_frame }, { "t8", DB_OFFSET(t8), db_frame }, { "t9", DB_OFFSET(t9), db_frame }, { "k0", DB_OFFSET(k0), db_frame }, { "k1", DB_OFFSET(k1), db_frame }, { "gp", DB_OFFSET(gp), db_frame }, { "sp", DB_OFFSET(sp), db_frame }, { "s8", DB_OFFSET(s8), db_frame }, { "ra", DB_OFFSET(ra), db_frame }, { "sr", DB_OFFSET(sr), db_frame }, { "lo", DB_OFFSET(mullo), db_frame }, { "hi", DB_OFFSET(mulhi), db_frame }, { "bad", DB_OFFSET(badvaddr), db_frame }, { "cs", DB_OFFSET(cause), db_frame }, { "pc", DB_OFFSET(pc), db_frame }, }; struct db_variable *db_eregs = db_regs + nitems(db_regs); int (*do_db_log_stack_trace_cmd)(char *); static int db_frame(struct db_variable *vp, db_expr_t *valuep, int op) { register_t *reg; if (kdb_frame == NULL) return (0); reg = (register_t *)((uintptr_t)kdb_frame + (size_t)(intptr_t)vp->valuep); if (op == DB_VAR_GET) *valuep = *reg; else *reg = *valuep; return (1); } int db_read_bytes(vm_offset_t addr, size_t size, char *data) { jmp_buf jb; void *prev_jb; int ret; prev_jb = kdb_jmpbuf(jb); ret = setjmp(jb); if (ret == 0) { /* * 'addr' could be a memory-mapped I/O address. Try to * do atomic load/store in unit of size requested. + * size == 8 is only atomic on 64bit or n32 kernel. */ if ((size == 2 || size == 4 || size == 8) && ((addr & (size -1)) == 0) && (((vm_offset_t)data & (size -1)) == 0)) { switch (size) { case 2: *(uint16_t *)data = *(uint16_t *)addr; break; case 4: *(uint32_t *)data = *(uint32_t *)addr; break; case 8: - *(uint64_t *)data = atomic_load_64( - (void *)addr); + *(uint64_t *)data = *(uint64_t *)addr; break; } } else { char *src; src = (char *)addr; while (size-- > 0) *data++ = *src++; } } (void)kdb_jmpbuf(prev_jb); return (ret); } int db_write_bytes(vm_offset_t addr, size_t size, char *data) { int ret; jmp_buf jb; void *prev_jb; prev_jb = kdb_jmpbuf(jb); ret = setjmp(jb); if (ret == 0) { /* * 'addr' could be a memory-mapped I/O address. Try to * do atomic load/store in unit of size requested. + * size == 8 is only atomic on 64bit or n32 kernel. */ if ((size == 2 || size == 4 || size == 8) && ((addr & (size -1)) == 0) && (((vm_offset_t)data & (size -1)) == 0)) { switch (size) { case 2: *(uint16_t *)addr = *(uint16_t *)data; break; case 4: *(uint32_t *)addr = *(uint32_t *)data; break; case 8: - atomic_store_64((uint64_t *)addr, - *(uint64_t *)data); + *(uint64_t *)addr = *(uint64_t *)data; break; } } else { char *dst; size_t len = size; dst = (char *)addr; while (len-- > 0) *dst++ = *data++; } mips_icache_sync_range((db_addr_t) addr, size); mips_dcache_wbinv_range((db_addr_t) addr, size); } (void)kdb_jmpbuf(prev_jb); return (ret); } /* * To do a single step ddb needs to know the next address * that we will get to. It means that we need to find out * both the address for a branch taken and for not taken, NOT! :-) * MipsEmulateBranch will do the job to find out _exactly_ which * address we will end up at so the 'dual bp' method is not * requiered. */ db_addr_t next_instr_address(db_addr_t pc, boolean_t bd) { db_addr_t next; next = (db_addr_t)MipsEmulateBranch(kdb_frame, pc, 0, 0); return (next); } /* * Decode instruction and figure out type. */ int db_inst_type(int ins) { InstFmt inst; int ityp = 0; inst.word = ins; switch ((int)inst.JType.op) { case OP_SPECIAL: switch ((int)inst.RType.func) { case OP_JR: ityp = IT_BRANCH; break; case OP_JALR: case OP_SYSCALL: ityp = IT_CALL; break; } break; case OP_BCOND: switch ((int)inst.IType.rt) { case OP_BLTZ: case OP_BLTZL: case OP_BGEZ: case OP_BGEZL: ityp = IT_BRANCH; break; case OP_BLTZAL: case OP_BLTZALL: case OP_BGEZAL: case OP_BGEZALL: ityp = IT_CALL; break; } break; case OP_JAL: ityp = IT_CALL; break; case OP_J: case OP_BEQ: case OP_BEQL: case OP_BNE: case OP_BNEL: case OP_BLEZ: case OP_BLEZL: case OP_BGTZ: case OP_BGTZL: ityp = IT_BRANCH; break; case OP_COP1: switch (inst.RType.rs) { case OP_BCx: case OP_BCy: ityp = IT_BRANCH; break; } break; case OP_LB: case OP_LH: case OP_LW: case OP_LD: case OP_LBU: case OP_LHU: case OP_LWU: case OP_LWC1: ityp = IT_LOAD; break; case OP_SB: case OP_SH: case OP_SW: case OP_SD: case OP_SWC1: ityp = IT_STORE; break; } return (ityp); } /* * Return the next pc if the given branch is taken. * MachEmulateBranch() runs analysis for branch delay slot. */ db_addr_t branch_taken(int inst, db_addr_t pc) { db_addr_t ra; register_t fpucsr; /* TBD: when is fsr set */ fpucsr = (curthread) ? curthread->td_pcb->pcb_regs.fsr : 0; ra = (db_addr_t)MipsEmulateBranch(kdb_frame, pc, fpucsr, 0); return (ra); } Index: head/sys/mips/mips/support.S =================================================================== --- head/sys/mips/mips/support.S (revision 327096) +++ head/sys/mips/mips/support.S (revision 327097) @@ -1,1099 +1,1032 @@ /* $OpenBSD: locore.S,v 1.18 1998/09/15 10:58:53 pefo Exp $ */ /*- * Copyright (c) 1992, 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * Digital Equipment Corporation and Ralph Campbell. * * 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. * 4. 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. * * Copyright (C) 1989 Digital Equipment Corporation. * Permission to use, copy, modify, and distribute this software and * its documentation for any purpose and without fee is hereby granted, * provided that the above copyright notice appears in all copies. * Digital Equipment Corporation makes no representations about the * suitability of this software for any purpose. It is provided "as is" * without express or implied warranty. * * from: Header: /sprite/src/kernel/mach/ds3100.md/RCS/loMem.s, * v 1.1 89/07/11 17:55:04 nelson Exp SPRITE (DECWRL) * from: Header: /sprite/src/kernel/mach/ds3100.md/RCS/machAsm.s, * v 9.2 90/01/29 18:00:39 shirriff Exp SPRITE (DECWRL) * from: Header: /sprite/src/kernel/vm/ds3100.md/vmPmaxAsm.s, * v 1.1 89/07/10 14:27:41 nelson Exp SPRITE (DECWRL) * * from: @(#)locore.s 8.5 (Berkeley) 1/4/94 * JNPR: support.S,v 1.5.2.2 2007/08/29 10:03:49 girish * $FreeBSD$ */ /* * Copyright (c) 1997 Jonathan Stone (hereinafter referred to as the author) * 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. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by Jonathan R. Stone for * the NetBSD Project. * 4. 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. */ /* * Contains assembly language support routines. */ #include "opt_ddb.h" #include #include #include #include #include #include #include "assym.s" .set noreorder # Noreorder is default style! /* * Primitives */ .text /* * See if access to addr with a len type instruction causes a machine check. * len is length of access (1=byte, 2=short, 4=int) * * badaddr(addr, len) * char *addr; * int len; */ LEAF(badaddr) PTR_LA v0, baderr GET_CPU_PCPU(v1) PTR_L v1, PC_CURPCB(v1) bne a1, 1, 2f PTR_S v0, U_PCB_ONFAULT(v1) b 5f lbu v0, (a0) 2: bne a1, 2, 4f nop b 5f lhu v0, (a0) 4: lw v0, (a0) 5: PTR_S zero, U_PCB_ONFAULT(v1) j ra move v0, zero # made it w/o errors baderr: j ra li v0, 1 # trap sends us here END(badaddr) /* * int copystr(void *kfaddr, void *kdaddr, size_t maxlen, size_t *lencopied) * Copy a NIL-terminated string, at most maxlen characters long. Return the * number of characters copied (including the NIL) in *lencopied. If the * string is too long, return ENAMETOOLONG; else return 0. */ LEAF(copystr) move t0, a2 beq a2, zero, 4f 1: lbu v0, 0(a0) PTR_SUBU a2, a2, 1 beq v0, zero, 2f sb v0, 0(a1) # each byte until NIL PTR_ADDU a0, a0, 1 bne a2, zero, 1b # less than maxlen PTR_ADDU a1, a1, 1 4: li v0, ENAMETOOLONG # run out of space 2: beq a3, zero, 3f # return num. of copied bytes PTR_SUBU a2, t0, a2 # if the 4th arg was non-NULL PTR_S a2, 0(a3) 3: j ra # v0 is 0 or ENAMETOOLONG nop END(copystr) /* * Copy a null terminated string from the user address space into * the kernel address space. * * copyinstr(fromaddr, toaddr, maxlength, &lencopied) * caddr_t fromaddr; * caddr_t toaddr; * u_int maxlength; * u_int *lencopied; */ NESTED(copyinstr, CALLFRAME_SIZ, ra) PTR_SUBU sp, sp, CALLFRAME_SIZ .mask 0x80000000, (CALLFRAME_RA - CALLFRAME_SIZ) PTR_LA v0, copyerr blt a0, zero, _C_LABEL(copyerr) # make sure address is in user space REG_S ra, CALLFRAME_RA(sp) GET_CPU_PCPU(v1) PTR_L v1, PC_CURPCB(v1) jal _C_LABEL(copystr) PTR_S v0, U_PCB_ONFAULT(v1) REG_L ra, CALLFRAME_RA(sp) GET_CPU_PCPU(v1) PTR_L v1, PC_CURPCB(v1) PTR_S zero, U_PCB_ONFAULT(v1) j ra PTR_ADDU sp, sp, CALLFRAME_SIZ END(copyinstr) /* * Copy a null terminated string from the kernel address space into * the user address space. * * copyoutstr(fromaddr, toaddr, maxlength, &lencopied) * caddr_t fromaddr; * caddr_t toaddr; * u_int maxlength; * u_int *lencopied; */ NESTED(copyoutstr, CALLFRAME_SIZ, ra) PTR_SUBU sp, sp, CALLFRAME_SIZ .mask 0x80000000, (CALLFRAME_RA - CALLFRAME_SIZ) PTR_LA v0, copyerr blt a1, zero, _C_LABEL(copyerr) # make sure address is in user space REG_S ra, CALLFRAME_RA(sp) GET_CPU_PCPU(v1) PTR_L v1, PC_CURPCB(v1) jal _C_LABEL(copystr) PTR_S v0, U_PCB_ONFAULT(v1) REG_L ra, CALLFRAME_RA(sp) GET_CPU_PCPU(v1) PTR_L v1, PC_CURPCB(v1) PTR_S zero, U_PCB_ONFAULT(v1) j ra PTR_ADDU sp, sp, CALLFRAME_SIZ END(copyoutstr) /* * Copy specified amount of data from user space into the kernel * copyin(from, to, len) * caddr_t *from; (user source address) * caddr_t *to; (kernel destination address) * unsigned len; */ NESTED(copyin, CALLFRAME_SIZ, ra) PTR_SUBU sp, sp, CALLFRAME_SIZ .mask 0x80000000, (CALLFRAME_RA - CALLFRAME_SIZ) PTR_LA v0, copyerr blt a0, zero, _C_LABEL(copyerr) # make sure address is in user space REG_S ra, CALLFRAME_RA(sp) GET_CPU_PCPU(v1) PTR_L v1, PC_CURPCB(v1) jal _C_LABEL(bcopy) PTR_S v0, U_PCB_ONFAULT(v1) REG_L ra, CALLFRAME_RA(sp) GET_CPU_PCPU(v1) PTR_L v1, PC_CURPCB(v1) # bcopy modified v1, so reload PTR_S zero, U_PCB_ONFAULT(v1) PTR_ADDU sp, sp, CALLFRAME_SIZ j ra move v0, zero END(copyin) /* * Copy specified amount of data from kernel to the user space * copyout(from, to, len) * caddr_t *from; (kernel source address) * caddr_t *to; (user destination address) * unsigned len; */ NESTED(copyout, CALLFRAME_SIZ, ra) PTR_SUBU sp, sp, CALLFRAME_SIZ .mask 0x80000000, (CALLFRAME_RA - CALLFRAME_SIZ) PTR_LA v0, copyerr blt a1, zero, _C_LABEL(copyerr) # make sure address is in user space REG_S ra, CALLFRAME_RA(sp) GET_CPU_PCPU(v1) PTR_L v1, PC_CURPCB(v1) jal _C_LABEL(bcopy) PTR_S v0, U_PCB_ONFAULT(v1) REG_L ra, CALLFRAME_RA(sp) GET_CPU_PCPU(v1) PTR_L v1, PC_CURPCB(v1) # bcopy modified v1, so reload PTR_S zero, U_PCB_ONFAULT(v1) PTR_ADDU sp, sp, CALLFRAME_SIZ j ra move v0, zero END(copyout) LEAF(copyerr) REG_L ra, CALLFRAME_RA(sp) PTR_ADDU sp, sp, CALLFRAME_SIZ j ra li v0, EFAULT # return error END(copyerr) /* * {fu,su},{ibyte,isword,iword}, fetch or store a byte, short or word to * user text space. * {fu,su},{byte,sword,word}, fetch or store a byte, short or word to * user data space. */ #ifdef __mips_n64 LEAF(fuword64) XLEAF(fuword) PTR_LA v0, fswberr blt a0, zero, fswberr # make sure address is in user space nop GET_CPU_PCPU(v1) PTR_L v1, PC_CURPCB(v1) PTR_S v0, U_PCB_ONFAULT(v1) ld v0, 0(a0) # fetch word j ra PTR_S zero, U_PCB_ONFAULT(v1) END(fuword64) #endif LEAF(fuword32) #ifndef __mips_n64 XLEAF(fuword) #endif PTR_LA v0, fswberr blt a0, zero, fswberr # make sure address is in user space nop GET_CPU_PCPU(v1) PTR_L v1, PC_CURPCB(v1) PTR_S v0, U_PCB_ONFAULT(v1) lw v0, 0(a0) # fetch word j ra PTR_S zero, U_PCB_ONFAULT(v1) END(fuword32) LEAF(fusword) PTR_LA v0, fswberr blt a0, zero, fswberr # make sure address is in user space nop GET_CPU_PCPU(v1) PTR_L v1, PC_CURPCB(v1) PTR_S v0, U_PCB_ONFAULT(v1) lhu v0, 0(a0) # fetch short j ra PTR_S zero, U_PCB_ONFAULT(v1) END(fusword) LEAF(fubyte) PTR_LA v0, fswberr blt a0, zero, fswberr # make sure address is in user space nop GET_CPU_PCPU(v1) PTR_L v1, PC_CURPCB(v1) PTR_S v0, U_PCB_ONFAULT(v1) lbu v0, 0(a0) # fetch byte j ra PTR_S zero, U_PCB_ONFAULT(v1) END(fubyte) LEAF(suword32) #ifndef __mips_n64 XLEAF(suword) #endif PTR_LA v0, fswberr blt a0, zero, fswberr # make sure address is in user space nop GET_CPU_PCPU(v1) PTR_L v1, PC_CURPCB(v1) PTR_S v0, U_PCB_ONFAULT(v1) sw a1, 0(a0) # store word PTR_S zero, U_PCB_ONFAULT(v1) j ra move v0, zero END(suword32) #ifdef __mips_n64 LEAF(suword64) XLEAF(suword) PTR_LA v0, fswberr blt a0, zero, fswberr # make sure address is in user space nop GET_CPU_PCPU(v1) PTR_L v1, PC_CURPCB(v1) PTR_S v0, U_PCB_ONFAULT(v1) sd a1, 0(a0) # store word PTR_S zero, U_PCB_ONFAULT(v1) j ra move v0, zero END(suword64) #endif /* * casuword(9) * u_long casuword(u_long *p, u_long oldval, u_long newval) */ /* * casuword32(9) * uint32_t casuword(uint32_t *p, uint32_t oldval, * uint32_t newval) */ LEAF(casuword32) #ifndef __mips_n64 XLEAF(casuword) #endif PTR_LA v0, fswberr blt a0, zero, fswberr # make sure address is in user space nop GET_CPU_PCPU(v1) PTR_L v1, PC_CURPCB(v1) PTR_S v0, U_PCB_ONFAULT(v1) 1: move t0, a2 ll v0, 0(a0) bne a1, v0, 2f nop sc t0, 0(a0) # store word beqz t0, 1b nop j 3f nop 2: li v0, -1 3: PTR_S zero, U_PCB_ONFAULT(v1) jr ra nop END(casuword32) #ifdef __mips_n64 LEAF(casuword64) XLEAF(casuword) PTR_LA v0, fswberr blt a0, zero, fswberr # make sure address is in user space nop GET_CPU_PCPU(v1) PTR_L v1, PC_CURPCB(v1) PTR_S v0, U_PCB_ONFAULT(v1) 1: move t0, a2 lld v0, 0(a0) bne a1, v0, 2f nop scd t0, 0(a0) # store double word beqz t0, 1b nop j 3f nop 2: li v0, -1 3: PTR_S zero, U_PCB_ONFAULT(v1) jr ra nop END(casuword64) #endif /* * Will have to flush the instruction cache if byte merging is done in hardware. */ LEAF(susword) PTR_LA v0, fswberr blt a0, zero, fswberr # make sure address is in user space nop GET_CPU_PCPU(v1) PTR_L v1, PC_CURPCB(v1) PTR_S v0, U_PCB_ONFAULT(v1) sh a1, 0(a0) # store short PTR_S zero, U_PCB_ONFAULT(v1) j ra move v0, zero END(susword) LEAF(subyte) PTR_LA v0, fswberr blt a0, zero, fswberr # make sure address is in user space nop GET_CPU_PCPU(v1) PTR_L v1, PC_CURPCB(v1) PTR_S v0, U_PCB_ONFAULT(v1) sb a1, 0(a0) # store byte PTR_S zero, U_PCB_ONFAULT(v1) j ra move v0, zero END(subyte) LEAF(fswberr) j ra li v0, -1 END(fswberr) /* * fuswintr and suswintr are just like fusword and susword except that if * the page is not in memory or would cause a trap, then we return an error. * The important thing is to prevent sleep() and switch(). */ LEAF(fuswintr) PTR_LA v0, fswintrberr blt a0, zero, fswintrberr # make sure address is in user space nop GET_CPU_PCPU(v1) PTR_L v1, PC_CURPCB(v1) PTR_S v0, U_PCB_ONFAULT(v1) lhu v0, 0(a0) # fetch short j ra PTR_S zero, U_PCB_ONFAULT(v1) END(fuswintr) LEAF(suswintr) PTR_LA v0, fswintrberr blt a0, zero, fswintrberr # make sure address is in user space nop GET_CPU_PCPU(v1) PTR_L v1, PC_CURPCB(v1) PTR_S v0, U_PCB_ONFAULT(v1) sh a1, 0(a0) # store short PTR_S zero, U_PCB_ONFAULT(v1) j ra move v0, zero END(suswintr) LEAF(fswintrberr) j ra li v0, -1 END(fswintrberr) /* * memset(void *s1, int c, int len) * NetBSD: memset.S,v 1.3 2001/10/16 15:40:53 uch Exp */ LEAF(memset) .set noreorder blt a2, 12, memsetsmallclr # small amount to clear? move v0, a0 # save s1 for result sll t1, a1, 8 # compute c << 8 in t1 or t1, t1, a1 # compute c << 8 | c in 11 sll t2, t1, 16 # shift that left 16 or t1, t2, t1 # or together PTR_SUBU t0, zero, a0 # compute # bytes to word align address and t0, t0, 3 beq t0, zero, 1f # skip if word aligned PTR_SUBU a2, a2, t0 # subtract from remaining count SWHI t1, 0(a0) # store 1, 2, or 3 bytes to align PTR_ADDU a0, a0, t0 1: and v1, a2, 3 # compute number of whole words left PTR_SUBU t0, a2, v1 PTR_SUBU a2, a2, t0 PTR_ADDU t0, t0, a0 # compute ending address 2: PTR_ADDU a0, a0, 4 # clear words bne a0, t0, 2b # unrolling loop does not help sw t1, -4(a0) # since we are limited by memory speed memsetsmallclr: ble a2, zero, 2f PTR_ADDU t0, a2, a0 # compute ending address 1: PTR_ADDU a0, a0, 1 # clear bytes bne a0, t0, 1b sb a1, -1(a0) 2: j ra nop .set reorder END(memset) /* * bzero(s1, n) */ LEAF(bzero) XLEAF(blkclr) .set noreorder blt a1, 12, smallclr # small amount to clear? PTR_SUBU a3, zero, a0 # compute # bytes to word align address and a3, a3, 3 beq a3, zero, 1f # skip if word aligned PTR_SUBU a1, a1, a3 # subtract from remaining count SWHI zero, 0(a0) # clear 1, 2, or 3 bytes to align PTR_ADDU a0, a0, a3 1: and v0, a1, 3 # compute number of words left PTR_SUBU a3, a1, v0 move a1, v0 PTR_ADDU a3, a3, a0 # compute ending address 2: PTR_ADDU a0, a0, 4 # clear words bne a0, a3, 2b # unrolling loop does not help sw zero, -4(a0) # since we are limited by memory speed smallclr: ble a1, zero, 2f PTR_ADDU a3, a1, a0 # compute ending address 1: PTR_ADDU a0, a0, 1 # clear bytes bne a0, a3, 1b sb zero, -1(a0) 2: j ra nop END(bzero) /* * bcmp(s1, s2, n) */ LEAF(bcmp) .set noreorder blt a2, 16, smallcmp # is it worth any trouble? xor v0, a0, a1 # compare low two bits of addresses and v0, v0, 3 PTR_SUBU a3, zero, a1 # compute # bytes to word align address bne v0, zero, unalignedcmp # not possible to align addresses and a3, a3, 3 beq a3, zero, 1f PTR_SUBU a2, a2, a3 # subtract from remaining count move v0, v1 # init v0,v1 so unmodified bytes match LWHI v0, 0(a0) # read 1, 2, or 3 bytes LWHI v1, 0(a1) PTR_ADDU a1, a1, a3 bne v0, v1, nomatch PTR_ADDU a0, a0, a3 1: and a3, a2, ~3 # compute number of whole words left PTR_SUBU a2, a2, a3 # which has to be >= (16-3) & ~3 PTR_ADDU a3, a3, a0 # compute ending address 2: lw v0, 0(a0) # compare words lw v1, 0(a1) PTR_ADDU a0, a0, 4 bne v0, v1, nomatch PTR_ADDU a1, a1, 4 bne a0, a3, 2b nop b smallcmp # finish remainder nop unalignedcmp: beq a3, zero, 2f PTR_SUBU a2, a2, a3 # subtract from remaining count PTR_ADDU a3, a3, a0 # compute ending address 1: lbu v0, 0(a0) # compare bytes until a1 word aligned lbu v1, 0(a1) PTR_ADDU a0, a0, 1 bne v0, v1, nomatch PTR_ADDU a1, a1, 1 bne a0, a3, 1b nop 2: and a3, a2, ~3 # compute number of whole words left PTR_SUBU a2, a2, a3 # which has to be >= (16-3) & ~3 PTR_ADDU a3, a3, a0 # compute ending address 3: LWHI v0, 0(a0) # compare words a0 unaligned, a1 aligned LWLO v0, 3(a0) lw v1, 0(a1) PTR_ADDU a0, a0, 4 bne v0, v1, nomatch PTR_ADDU a1, a1, 4 bne a0, a3, 3b nop smallcmp: ble a2, zero, match PTR_ADDU a3, a2, a0 # compute ending address 1: lbu v0, 0(a0) lbu v1, 0(a1) PTR_ADDU a0, a0, 1 bne v0, v1, nomatch PTR_ADDU a1, a1, 1 bne a0, a3, 1b nop match: j ra move v0, zero nomatch: j ra li v0, 1 END(bcmp) /* * bit = ffs(value) */ LEAF(ffs) .set noreorder beq a0, zero, 2f move v0, zero 1: and v1, a0, 1 # bit set? addu v0, v0, 1 beq v1, zero, 1b # no, continue srl a0, a0, 1 2: j ra nop END(ffs) /** * void * atomic_set_16(u_int16_t *a, u_int16_t b) * { * *a |= b; * } */ LEAF(atomic_set_16) .set noreorder srl a0, a0, 2 # round down address to be 32-bit aligned sll a0, a0, 2 andi a1, a1, 0xffff 1: ll t0, 0(a0) or t0, t0, a1 sc t0, 0(a0) beq t0, zero, 1b nop j ra nop END(atomic_set_16) /** * void * atomic_clear_16(u_int16_t *a, u_int16_t b) * { * *a &= ~b; * } */ LEAF(atomic_clear_16) .set noreorder srl a0, a0, 2 # round down address to be 32-bit aligned sll a0, a0, 2 nor a1, zero, a1 1: ll t0, 0(a0) move t1, t0 andi t1, t1, 0xffff # t1 has the original lower 16 bits and t1, t1, a1 # t1 has the new lower 16 bits srl t0, t0, 16 # preserve original top 16 bits sll t0, t0, 16 or t0, t0, t1 sc t0, 0(a0) beq t0, zero, 1b nop j ra nop END(atomic_clear_16) /** * void * atomic_subtract_16(uint16_t *a, uint16_t b) * { * *a -= b; * } */ LEAF(atomic_subtract_16) .set noreorder srl a0, a0, 2 # round down address to be 32-bit aligned sll a0, a0, 2 1: ll t0, 0(a0) move t1, t0 andi t1, t1, 0xffff # t1 has the original lower 16 bits subu t1, t1, a1 andi t1, t1, 0xffff # t1 has the new lower 16 bits srl t0, t0, 16 # preserve original top 16 bits sll t0, t0, 16 or t0, t0, t1 sc t0, 0(a0) beq t0, zero, 1b nop j ra nop END(atomic_subtract_16) /** * void * atomic_add_16(uint16_t *a, uint16_t b) * { * *a += b; * } */ LEAF(atomic_add_16) .set noreorder srl a0, a0, 2 # round down address to be 32-bit aligned sll a0, a0, 2 1: ll t0, 0(a0) move t1, t0 andi t1, t1, 0xffff # t1 has the original lower 16 bits addu t1, t1, a1 andi t1, t1, 0xffff # t1 has the new lower 16 bits srl t0, t0, 16 # preserve original top 16 bits sll t0, t0, 16 or t0, t0, t1 sc t0, 0(a0) beq t0, zero, 1b nop j ra nop END(atomic_add_16) /** * void * atomic_add_8(uint8_t *a, uint8_t b) * { * *a += b; * } */ LEAF(atomic_add_8) .set noreorder srl a0, a0, 2 # round down address to be 32-bit aligned sll a0, a0, 2 1: ll t0, 0(a0) move t1, t0 andi t1, t1, 0xff # t1 has the original lower 8 bits addu t1, t1, a1 andi t1, t1, 0xff # t1 has the new lower 8 bits srl t0, t0, 8 # preserve original top 24 bits sll t0, t0, 8 or t0, t0, t1 sc t0, 0(a0) beq t0, zero, 1b nop j ra nop END(atomic_add_8) /** * void * atomic_subtract_8(uint8_t *a, uint8_t b) * { * *a += b; * } */ LEAF(atomic_subtract_8) .set noreorder srl a0, a0, 2 # round down address to be 32-bit aligned sll a0, a0, 2 1: ll t0, 0(a0) move t1, t0 andi t1, t1, 0xff # t1 has the original lower 8 bits subu t1, t1, a1 andi t1, t1, 0xff # t1 has the new lower 8 bits srl t0, t0, 8 # preserve original top 24 bits sll t0, t0, 8 or t0, t0, t1 sc t0, 0(a0) beq t0, zero, 1b nop j ra nop END(atomic_subtract_8) -/* - * atomic 64-bit register read/write assembly language support routines. - */ - .set noreorder # Noreorder is default style! - -#if !defined(__mips_n64) && !defined(__mips_n32) - /* - * I don't know if these routines have the right number of - * NOPs in it for all processors. XXX - * - * Maybe it would be better to just leave this undefined in that case. - * - * XXX These routines are not safe in the case of a TLB miss on a1 or - * a0 unless the trapframe is 64-bit, which it just isn't with O32. - * If we take any exception, not just an interrupt, the upper - * 32-bits will be clobbered. Use only N32 and N64 kernels if you - * want to use 64-bit registers while interrupts are enabled or - * with memory operations. Since this isn't even using load-linked - * and store-conditional, perhaps it should just use two registers - * instead, as is right and good with the O32 ABI. - */ -LEAF(atomic_store_64) - mfc0 t1, MIPS_COP_0_STATUS - and t2, t1, ~MIPS_SR_INT_IE - mtc0 t2, MIPS_COP_0_STATUS - nop - nop - nop - nop - ld t0, (a1) - nop - nop - sd t0, (a0) - nop - nop - mtc0 t1,MIPS_COP_0_STATUS - nop - nop - nop - nop - j ra - nop -END(atomic_store_64) - -LEAF(atomic_load_64) - mfc0 t1, MIPS_COP_0_STATUS - and t2, t1, ~MIPS_SR_INT_IE - mtc0 t2, MIPS_COP_0_STATUS - nop - nop - nop - nop - ld t0, (a0) - nop - nop - sd t0, (a1) - nop - nop - mtc0 t1,MIPS_COP_0_STATUS - nop - nop - nop - nop - j ra - nop -END(atomic_load_64) -#endif #if defined(DDB) || defined(DEBUG) LEAF(kdbpeek) PTR_LA v1, ddberr and v0, a0, 3 # unaligned ? GET_CPU_PCPU(t1) PTR_L t1, PC_CURPCB(t1) bne v0, zero, 1f PTR_S v1, U_PCB_ONFAULT(t1) lw v0, (a0) jr ra PTR_S zero, U_PCB_ONFAULT(t1) 1: LWHI v0, 0(a0) LWLO v0, 3(a0) jr ra PTR_S zero, U_PCB_ONFAULT(t1) END(kdbpeek) LEAF(kdbpeekd) PTR_LA v1, ddberr and v0, a0, 3 # unaligned ? GET_CPU_PCPU(t1) PTR_L t1, PC_CURPCB(t1) bne v0, zero, 1f PTR_S v1, U_PCB_ONFAULT(t1) ld v0, (a0) jr ra PTR_S zero, U_PCB_ONFAULT(t1) 1: REG_LHI v0, 0(a0) REG_LLO v0, 7(a0) jr ra PTR_S zero, U_PCB_ONFAULT(t1) END(kdbpeekd) ddberr: jr ra nop #if defined(DDB) LEAF(kdbpoke) PTR_LA v1, ddberr and v0, a0, 3 # unaligned ? GET_CPU_PCPU(t1) PTR_L t1, PC_CURPCB(t1) bne v0, zero, 1f PTR_S v1, U_PCB_ONFAULT(t1) sw a1, (a0) jr ra PTR_S zero, U_PCB_ONFAULT(t1) 1: SWHI a1, 0(a0) SWLO a1, 3(a0) jr ra PTR_S zero, U_PCB_ONFAULT(t1) END(kdbpoke) .data .globl esym esym: .word 0 #endif /* DDB */ #endif /* DDB || DEBUG */ .text LEAF(breakpoint) break MIPS_BREAK_SOVER_VAL jr ra nop END(breakpoint) LEAF(setjmp) mfc0 v0, MIPS_COP_0_STATUS # Later the "real" spl value! REG_S s0, (SZREG * PCB_REG_S0)(a0) REG_S s1, (SZREG * PCB_REG_S1)(a0) REG_S s2, (SZREG * PCB_REG_S2)(a0) REG_S s3, (SZREG * PCB_REG_S3)(a0) REG_S s4, (SZREG * PCB_REG_S4)(a0) REG_S s5, (SZREG * PCB_REG_S5)(a0) REG_S s6, (SZREG * PCB_REG_S6)(a0) REG_S s7, (SZREG * PCB_REG_S7)(a0) REG_S s8, (SZREG * PCB_REG_S8)(a0) REG_S sp, (SZREG * PCB_REG_SP)(a0) REG_S ra, (SZREG * PCB_REG_RA)(a0) REG_S v0, (SZREG * PCB_REG_SR)(a0) jr ra li v0, 0 # setjmp return END(setjmp) LEAF(longjmp) REG_L v0, (SZREG * PCB_REG_SR)(a0) REG_L ra, (SZREG * PCB_REG_RA)(a0) REG_L s0, (SZREG * PCB_REG_S0)(a0) REG_L s1, (SZREG * PCB_REG_S1)(a0) REG_L s2, (SZREG * PCB_REG_S2)(a0) REG_L s3, (SZREG * PCB_REG_S3)(a0) REG_L s4, (SZREG * PCB_REG_S4)(a0) REG_L s5, (SZREG * PCB_REG_S5)(a0) REG_L s6, (SZREG * PCB_REG_S6)(a0) REG_L s7, (SZREG * PCB_REG_S7)(a0) REG_L s8, (SZREG * PCB_REG_S8)(a0) REG_L sp, (SZREG * PCB_REG_SP)(a0) mtc0 v0, MIPS_COP_0_STATUS # Later the "real" spl value! ITLBNOPFIX jr ra li v0, 1 # longjmp return END(longjmp) LEAF(mips3_ld) .set push .set noreorder .set mips64 #if defined(__mips_o32) mfc0 t0, MIPS_COP_0_STATUS # turn off interrupts and t1, t0, ~(MIPS_SR_INT_IE) mtc0 t1, MIPS_COP_0_STATUS COP0_SYNC nop nop nop ld v0, 0(a0) #if _BYTE_ORDER == _BIG_ENDIAN dsll v1, v0, 32 dsra v1, v1, 32 # low word in v1 dsra v0, v0, 32 # high word in v0 #else dsra v1, v0, 32 # high word in v1 dsll v0, v0, 32 dsra v0, v0, 32 # low word in v0 #endif mtc0 t0, MIPS_COP_0_STATUS # restore intr status. COP0_SYNC nop #else /* !__mips_o32 */ ld v0, 0(a0) #endif /* !__mips_o32 */ jr ra nop .set pop END(mips3_ld) LEAF(mips3_sd) .set push .set mips64 .set noreorder #if defined(__mips_o32) mfc0 t0, MIPS_COP_0_STATUS # turn off interrupts and t1, t0, ~(MIPS_SR_INT_IE) mtc0 t1, MIPS_COP_0_STATUS COP0_SYNC nop nop nop # NOTE: a1 is padding! #if _BYTE_ORDER == _BIG_ENDIAN dsll a2, a2, 32 # high word in a2 dsll a3, a3, 32 # low word in a3 dsrl a3, a3, 32 #else dsll a2, a2, 32 # low word in a2 dsrl a2, a2, 32 dsll a3, a3, 32 # high word in a3 #endif or a1, a2, a3 sd a1, 0(a0) mtc0 t0, MIPS_COP_0_STATUS # restore intr status. COP0_SYNC nop #else /* !__mips_o32 */ sd a1, 0(a0) #endif /* !__mips_o32 */ jr ra nop .set pop END(mips3_sd)