Merge remote-tracking branch 'remotes/gkurz/tags/for-upstream' into staging
[qemu/ar7.git] / target-ppc / mem_helper.c
blob1ab8a6eab4ee2e56789ca8ddfb8ec2a68bd7326a
1 /*
2 * PowerPC memory access emulation helpers for QEMU.
4 * Copyright (c) 2003-2007 Jocelyn Mayer
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2 of the License, or (at your option) any later version.
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with this library; if not, see <http://www.gnu.org/licenses/>.
19 #include "qemu/osdep.h"
20 #include "cpu.h"
21 #include "exec/exec-all.h"
22 #include "qemu/host-utils.h"
23 #include "exec/helper-proto.h"
25 #include "helper_regs.h"
26 #include "exec/cpu_ldst.h"
28 //#define DEBUG_OP
30 static inline bool needs_byteswap(const CPUPPCState *env)
32 #if defined(TARGET_WORDS_BIGENDIAN)
33 return msr_le;
34 #else
35 return !msr_le;
36 #endif
39 /*****************************************************************************/
40 /* Memory load and stores */
42 static inline target_ulong addr_add(CPUPPCState *env, target_ulong addr,
43 target_long arg)
45 #if defined(TARGET_PPC64)
46 if (!msr_is_64bit(env, env->msr)) {
47 return (uint32_t)(addr + arg);
48 } else
49 #endif
51 return addr + arg;
55 void helper_lmw(CPUPPCState *env, target_ulong addr, uint32_t reg)
57 for (; reg < 32; reg++) {
58 if (needs_byteswap(env)) {
59 env->gpr[reg] = bswap32(cpu_ldl_data_ra(env, addr, GETPC()));
60 } else {
61 env->gpr[reg] = cpu_ldl_data_ra(env, addr, GETPC());
63 addr = addr_add(env, addr, 4);
67 void helper_stmw(CPUPPCState *env, target_ulong addr, uint32_t reg)
69 for (; reg < 32; reg++) {
70 if (needs_byteswap(env)) {
71 cpu_stl_data_ra(env, addr, bswap32((uint32_t)env->gpr[reg]),
72 GETPC());
73 } else {
74 cpu_stl_data_ra(env, addr, (uint32_t)env->gpr[reg], GETPC());
76 addr = addr_add(env, addr, 4);
80 static void do_lsw(CPUPPCState *env, target_ulong addr, uint32_t nb,
81 uint32_t reg, uintptr_t raddr)
83 int sh;
85 for (; nb > 3; nb -= 4) {
86 env->gpr[reg] = cpu_ldl_data_ra(env, addr, raddr);
87 reg = (reg + 1) % 32;
88 addr = addr_add(env, addr, 4);
90 if (unlikely(nb > 0)) {
91 env->gpr[reg] = 0;
92 for (sh = 24; nb > 0; nb--, sh -= 8) {
93 env->gpr[reg] |= cpu_ldub_data_ra(env, addr, raddr) << sh;
94 addr = addr_add(env, addr, 1);
99 void helper_lsw(CPUPPCState *env, target_ulong addr, uint32_t nb, uint32_t reg)
101 do_lsw(env, addr, nb, reg, GETPC());
104 /* PPC32 specification says we must generate an exception if
105 * rA is in the range of registers to be loaded.
106 * In an other hand, IBM says this is valid, but rA won't be loaded.
107 * For now, I'll follow the spec...
109 void helper_lswx(CPUPPCState *env, target_ulong addr, uint32_t reg,
110 uint32_t ra, uint32_t rb)
112 if (likely(xer_bc != 0)) {
113 int num_used_regs = (xer_bc + 3) / 4;
114 if (unlikely((ra != 0 && lsw_reg_in_range(reg, num_used_regs, ra)) ||
115 lsw_reg_in_range(reg, num_used_regs, rb))) {
116 raise_exception_err_ra(env, POWERPC_EXCP_PROGRAM,
117 POWERPC_EXCP_INVAL |
118 POWERPC_EXCP_INVAL_LSWX, GETPC());
119 } else {
120 do_lsw(env, addr, xer_bc, reg, GETPC());
125 void helper_stsw(CPUPPCState *env, target_ulong addr, uint32_t nb,
126 uint32_t reg)
128 int sh;
130 for (; nb > 3; nb -= 4) {
131 cpu_stl_data_ra(env, addr, env->gpr[reg], GETPC());
132 reg = (reg + 1) % 32;
133 addr = addr_add(env, addr, 4);
135 if (unlikely(nb > 0)) {
136 for (sh = 24; nb > 0; nb--, sh -= 8) {
137 cpu_stb_data_ra(env, addr, (env->gpr[reg] >> sh) & 0xFF, GETPC());
138 addr = addr_add(env, addr, 1);
143 void helper_dcbz(CPUPPCState *env, target_ulong addr, uint32_t opcode)
145 target_ulong mask, dcbz_size = env->dcache_line_size;
146 uint32_t i;
147 void *haddr;
149 #if defined(TARGET_PPC64)
150 /* Check for dcbz vs dcbzl on 970 */
151 if (env->excp_model == POWERPC_EXCP_970 &&
152 !(opcode & 0x00200000) && ((env->spr[SPR_970_HID5] >> 7) & 0x3) == 1) {
153 dcbz_size = 32;
155 #endif
157 /* Align address */
158 mask = ~(dcbz_size - 1);
159 addr &= mask;
161 /* Check reservation */
162 if ((env->reserve_addr & mask) == (addr & mask)) {
163 env->reserve_addr = (target_ulong)-1ULL;
166 /* Try fast path translate */
167 haddr = tlb_vaddr_to_host(env, addr, MMU_DATA_STORE, env->dmmu_idx);
168 if (haddr) {
169 memset(haddr, 0, dcbz_size);
170 } else {
171 /* Slow path */
172 for (i = 0; i < dcbz_size; i += 8) {
173 cpu_stq_data_ra(env, addr + i, 0, GETPC());
178 void helper_icbi(CPUPPCState *env, target_ulong addr)
180 addr &= ~(env->dcache_line_size - 1);
181 /* Invalidate one cache line :
182 * PowerPC specification says this is to be treated like a load
183 * (not a fetch) by the MMU. To be sure it will be so,
184 * do the load "by hand".
186 cpu_ldl_data_ra(env, addr, GETPC());
189 /* XXX: to be tested */
190 target_ulong helper_lscbx(CPUPPCState *env, target_ulong addr, uint32_t reg,
191 uint32_t ra, uint32_t rb)
193 int i, c, d;
195 d = 24;
196 for (i = 0; i < xer_bc; i++) {
197 c = cpu_ldub_data_ra(env, addr, GETPC());
198 addr = addr_add(env, addr, 1);
199 /* ra (if not 0) and rb are never modified */
200 if (likely(reg != rb && (ra == 0 || reg != ra))) {
201 env->gpr[reg] = (env->gpr[reg] & ~(0xFF << d)) | (c << d);
203 if (unlikely(c == xer_cmp)) {
204 break;
206 if (likely(d != 0)) {
207 d -= 8;
208 } else {
209 d = 24;
210 reg++;
211 reg = reg & 0x1F;
214 return i;
217 /*****************************************************************************/
218 /* Altivec extension helpers */
219 #if defined(HOST_WORDS_BIGENDIAN)
220 #define HI_IDX 0
221 #define LO_IDX 1
222 #else
223 #define HI_IDX 1
224 #define LO_IDX 0
225 #endif
227 /* We use msr_le to determine index ordering in a vector. However,
228 byteswapping is not simply controlled by msr_le. We also need to take
229 into account endianness of the target. This is done for the little-endian
230 PPC64 user-mode target. */
232 #define LVE(name, access, swap, element) \
233 void helper_##name(CPUPPCState *env, ppc_avr_t *r, \
234 target_ulong addr) \
236 size_t n_elems = ARRAY_SIZE(r->element); \
237 int adjust = HI_IDX*(n_elems - 1); \
238 int sh = sizeof(r->element[0]) >> 1; \
239 int index = (addr & 0xf) >> sh; \
240 if (msr_le) { \
241 index = n_elems - index - 1; \
244 if (needs_byteswap(env)) { \
245 r->element[LO_IDX ? index : (adjust - index)] = \
246 swap(access(env, addr, GETPC())); \
247 } else { \
248 r->element[LO_IDX ? index : (adjust - index)] = \
249 access(env, addr, GETPC()); \
252 #define I(x) (x)
253 LVE(lvebx, cpu_ldub_data_ra, I, u8)
254 LVE(lvehx, cpu_lduw_data_ra, bswap16, u16)
255 LVE(lvewx, cpu_ldl_data_ra, bswap32, u32)
256 #undef I
257 #undef LVE
259 #define STVE(name, access, swap, element) \
260 void helper_##name(CPUPPCState *env, ppc_avr_t *r, \
261 target_ulong addr) \
263 size_t n_elems = ARRAY_SIZE(r->element); \
264 int adjust = HI_IDX * (n_elems - 1); \
265 int sh = sizeof(r->element[0]) >> 1; \
266 int index = (addr & 0xf) >> sh; \
267 if (msr_le) { \
268 index = n_elems - index - 1; \
271 if (needs_byteswap(env)) { \
272 access(env, addr, swap(r->element[LO_IDX ? index : \
273 (adjust - index)]), \
274 GETPC()); \
275 } else { \
276 access(env, addr, r->element[LO_IDX ? index : \
277 (adjust - index)], GETPC()); \
280 #define I(x) (x)
281 STVE(stvebx, cpu_stb_data_ra, I, u8)
282 STVE(stvehx, cpu_stw_data_ra, bswap16, u16)
283 STVE(stvewx, cpu_stl_data_ra, bswap32, u32)
284 #undef I
285 #undef LVE
287 #undef HI_IDX
288 #undef LO_IDX
290 void helper_tbegin(CPUPPCState *env)
292 /* As a degenerate implementation, always fail tbegin. The reason
293 * given is "Nesting overflow". The "persistent" bit is set,
294 * providing a hint to the error handler to not retry. The TFIAR
295 * captures the address of the failure, which is this tbegin
296 * instruction. Instruction execution will continue with the
297 * next instruction in memory, which is precisely what we want.
300 env->spr[SPR_TEXASR] =
301 (1ULL << TEXASR_FAILURE_PERSISTENT) |
302 (1ULL << TEXASR_NESTING_OVERFLOW) |
303 (msr_hv << TEXASR_PRIVILEGE_HV) |
304 (msr_pr << TEXASR_PRIVILEGE_PR) |
305 (1ULL << TEXASR_FAILURE_SUMMARY) |
306 (1ULL << TEXASR_TFIAR_EXACT);
307 env->spr[SPR_TFIAR] = env->nip | (msr_hv << 1) | msr_pr;
308 env->spr[SPR_TFHAR] = env->nip + 4;
309 env->crf[0] = 0xB; /* 0b1010 = transaction failure */