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"
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"
31 static inline bool needs_byteswap(const CPUPPCState *env)
33 #if defined(TARGET_WORDS_BIGENDIAN)
40 /*****************************************************************************/
41 /* Memory load and stores */
43 static inline target_ulong addr_add(CPUPPCState *env, target_ulong addr,
46 #if defined(TARGET_PPC64)
47 if (!msr_is_64bit(env, env->msr)) {
48 return (uint32_t)(addr + arg);
56 void helper_lmw(CPUPPCState *env, target_ulong addr, uint32_t reg)
58 for (; reg < 32; reg++) {
59 if (needs_byteswap(env)) {
60 env->gpr[reg] = bswap32(cpu_ldl_data_ra(env, addr, GETPC()));
62 env->gpr[reg] = cpu_ldl_data_ra(env, addr, GETPC());
64 addr = addr_add(env, addr, 4);
68 void helper_stmw(CPUPPCState *env, target_ulong addr, uint32_t reg)
70 for (; reg < 32; reg++) {
71 if (needs_byteswap(env)) {
72 cpu_stl_data_ra(env, addr, bswap32((uint32_t)env->gpr[reg]),
75 cpu_stl_data_ra(env, addr, (uint32_t)env->gpr[reg], GETPC());
77 addr = addr_add(env, addr, 4);
81 static void do_lsw(CPUPPCState *env, target_ulong addr, uint32_t nb,
82 uint32_t reg, uintptr_t raddr)
86 for (; nb > 3; nb -= 4) {
87 env->gpr[reg] = cpu_ldl_data_ra(env, addr, raddr);
89 addr = addr_add(env, addr, 4);
91 if (unlikely(nb > 0)) {
93 for (sh = 24; nb > 0; nb--, sh -= 8) {
94 env->gpr[reg] |= cpu_ldub_data_ra(env, addr, raddr) << sh;
95 addr = addr_add(env, addr, 1);
100 void helper_lsw(CPUPPCState *env, target_ulong addr, uint32_t nb, uint32_t reg)
102 do_lsw(env, addr, nb, reg, GETPC());
105 /* PPC32 specification says we must generate an exception if
106 * rA is in the range of registers to be loaded.
107 * In an other hand, IBM says this is valid, but rA won't be loaded.
108 * For now, I'll follow the spec...
110 void helper_lswx(CPUPPCState *env, target_ulong addr, uint32_t reg,
111 uint32_t ra, uint32_t rb)
113 if (likely(xer_bc != 0)) {
114 int num_used_regs = (xer_bc + 3) / 4;
115 if (unlikely((ra != 0 && lsw_reg_in_range(reg, num_used_regs, ra)) ||
116 lsw_reg_in_range(reg, num_used_regs, rb))) {
117 raise_exception_err_ra(env, POWERPC_EXCP_PROGRAM,
119 POWERPC_EXCP_INVAL_LSWX, GETPC());
121 do_lsw(env, addr, xer_bc, reg, GETPC());
126 void helper_stsw(CPUPPCState *env, target_ulong addr, uint32_t nb,
131 for (; nb > 3; nb -= 4) {
132 cpu_stl_data_ra(env, addr, env->gpr[reg], GETPC());
133 reg = (reg + 1) % 32;
134 addr = addr_add(env, addr, 4);
136 if (unlikely(nb > 0)) {
137 for (sh = 24; nb > 0; nb--, sh -= 8) {
138 cpu_stb_data_ra(env, addr, (env->gpr[reg] >> sh) & 0xFF, GETPC());
139 addr = addr_add(env, addr, 1);
144 void helper_dcbz(CPUPPCState *env, target_ulong addr, uint32_t opcode)
146 target_ulong mask, dcbz_size = env->dcache_line_size;
150 #if defined(TARGET_PPC64)
151 /* Check for dcbz vs dcbzl on 970 */
152 if (env->excp_model == POWERPC_EXCP_970 &&
153 !(opcode & 0x00200000) && ((env->spr[SPR_970_HID5] >> 7) & 0x3) == 1) {
159 mask = ~(dcbz_size - 1);
162 /* Check reservation */
163 if ((env->reserve_addr & mask) == (addr & mask)) {
164 env->reserve_addr = (target_ulong)-1ULL;
167 /* Try fast path translate */
168 haddr = tlb_vaddr_to_host(env, addr, MMU_DATA_STORE, env->dmmu_idx);
170 memset(haddr, 0, dcbz_size);
173 for (i = 0; i < dcbz_size; i += 8) {
174 cpu_stq_data_ra(env, addr + i, 0, GETPC());
179 void helper_icbi(CPUPPCState *env, target_ulong addr)
181 addr &= ~(env->dcache_line_size - 1);
182 /* Invalidate one cache line :
183 * PowerPC specification says this is to be treated like a load
184 * (not a fetch) by the MMU. To be sure it will be so,
185 * do the load "by hand".
187 cpu_ldl_data_ra(env, addr, GETPC());
190 /* XXX: to be tested */
191 target_ulong helper_lscbx(CPUPPCState *env, target_ulong addr, uint32_t reg,
192 uint32_t ra, uint32_t rb)
197 for (i = 0; i < xer_bc; i++) {
198 c = cpu_ldub_data_ra(env, addr, GETPC());
199 addr = addr_add(env, addr, 1);
200 /* ra (if not 0) and rb are never modified */
201 if (likely(reg != rb && (ra == 0 || reg != ra))) {
202 env->gpr[reg] = (env->gpr[reg] & ~(0xFF << d)) | (c << d);
204 if (unlikely(c == xer_cmp)) {
207 if (likely(d != 0)) {
218 /*****************************************************************************/
219 /* Altivec extension helpers */
220 #if defined(HOST_WORDS_BIGENDIAN)
228 /* We use msr_le to determine index ordering in a vector. However,
229 byteswapping is not simply controlled by msr_le. We also need to take
230 into account endianness of the target. This is done for the little-endian
231 PPC64 user-mode target. */
233 #define LVE(name, access, swap, element) \
234 void helper_##name(CPUPPCState *env, ppc_avr_t *r, \
237 size_t n_elems = ARRAY_SIZE(r->element); \
238 int adjust = HI_IDX*(n_elems - 1); \
239 int sh = sizeof(r->element[0]) >> 1; \
240 int index = (addr & 0xf) >> sh; \
242 index = n_elems - index - 1; \
245 if (needs_byteswap(env)) { \
246 r->element[LO_IDX ? index : (adjust - index)] = \
247 swap(access(env, addr, GETPC())); \
249 r->element[LO_IDX ? index : (adjust - index)] = \
250 access(env, addr, GETPC()); \
254 LVE(lvebx, cpu_ldub_data_ra, I, u8)
255 LVE(lvehx, cpu_lduw_data_ra, bswap16, u16)
256 LVE(lvewx, cpu_ldl_data_ra, bswap32, u32)
260 #define STVE(name, access, swap, element) \
261 void helper_##name(CPUPPCState *env, ppc_avr_t *r, \
264 size_t n_elems = ARRAY_SIZE(r->element); \
265 int adjust = HI_IDX * (n_elems - 1); \
266 int sh = sizeof(r->element[0]) >> 1; \
267 int index = (addr & 0xf) >> sh; \
269 index = n_elems - index - 1; \
272 if (needs_byteswap(env)) { \
273 access(env, addr, swap(r->element[LO_IDX ? index : \
274 (adjust - index)]), \
277 access(env, addr, r->element[LO_IDX ? index : \
278 (adjust - index)], GETPC()); \
282 STVE(stvebx, cpu_stb_data_ra, I, u8)
283 STVE(stvehx, cpu_stw_data_ra, bswap16, u16)
284 STVE(stvewx, cpu_stl_data_ra, bswap32, u32)
289 #define GET_NB(rb) ((rb >> 56) & 0xFF)
291 #define VSX_LXVL(name, lj) \
292 void helper_##name(CPUPPCState *env, target_ulong addr, \
293 target_ulong xt_num, target_ulong rb) \
297 uint64_t nb = GET_NB(rb); \
299 xt.s128 = int128_zero(); \
301 nb = (nb >= 16) ? 16 : nb; \
302 if (msr_le && !lj) { \
303 for (i = 16; i > 16 - nb; i--) { \
304 xt.VsrB(i - 1) = cpu_ldub_data_ra(env, addr, GETPC()); \
305 addr = addr_add(env, addr, 1); \
308 for (i = 0; i < nb; i++) { \
309 xt.VsrB(i) = cpu_ldub_data_ra(env, addr, GETPC()); \
310 addr = addr_add(env, addr, 1); \
314 putVSR(xt_num, &xt, env); \
321 #define VSX_STXVL(name, lj) \
322 void helper_##name(CPUPPCState *env, target_ulong addr, \
323 target_ulong xt_num, target_ulong rb) \
327 target_ulong nb = GET_NB(rb); \
332 getVSR(xt_num, &xt, env); \
333 nb = (nb >= 16) ? 16 : nb; \
334 if (msr_le && !lj) { \
335 for (i = 16; i > 16 - nb; i--) { \
336 cpu_stb_data_ra(env, addr, xt.VsrB(i - 1), GETPC()); \
337 addr = addr_add(env, addr, 1); \
340 for (i = 0; i < nb; i++) { \
341 cpu_stb_data_ra(env, addr, xt.VsrB(i), GETPC()); \
342 addr = addr_add(env, addr, 1); \
350 #endif /* TARGET_PPC64 */
355 void helper_tbegin(CPUPPCState *env)
357 /* As a degenerate implementation, always fail tbegin. The reason
358 * given is "Nesting overflow". The "persistent" bit is set,
359 * providing a hint to the error handler to not retry. The TFIAR
360 * captures the address of the failure, which is this tbegin
361 * instruction. Instruction execution will continue with the
362 * next instruction in memory, which is precisely what we want.
365 env->spr[SPR_TEXASR] =
366 (1ULL << TEXASR_FAILURE_PERSISTENT) |
367 (1ULL << TEXASR_NESTING_OVERFLOW) |
368 (msr_hv << TEXASR_PRIVILEGE_HV) |
369 (msr_pr << TEXASR_PRIVILEGE_PR) |
370 (1ULL << TEXASR_FAILURE_SUMMARY) |
371 (1ULL << TEXASR_TFIAR_EXACT);
372 env->spr[SPR_TFIAR] = env->nip | (msr_hv << 1) | msr_pr;
373 env->spr[SPR_TFHAR] = env->nip + 4;
374 env->crf[0] = 0xB; /* 0b1010 = transaction failure */