target-arm: fix LDMIA bug on page boundary
PATCH 1/1 target-arm: fix LDMIA bug on page boundary
When consecutive memory locations are on page boundary, a base register may be loaded before page fault occurs. After page fault handling, it losts the memory location information. To solve this problem, loading a base register has to put back.
--- target-arm/translate.c | 18 +++++++++++++----- 1 files changed, 13 insertions(+), 5 deletions(-)
target-arm/translate.c +13 -5
@@ -7984,11 +7984,16 @@ static int disas_thumb2_insn(CPUState *env, DisasContext *s, uint16_t insn_hw1)continue ;if (insn & (1 <<20 )) {/* Load. */ - tmp =gen_ld32 (addr,IS_USER (s)); -if (i ==15 ) { -gen_bx (s, tmp); - }else { -store_reg (s, i, tmp); +if (i == rn) { + tmp2 =gen_ld32 (addr,IS_USER (s)); + } +else { + tmp =gen_ld32 (addr,IS_USER (s)); +if (i ==15 ) { +gen_bx (s, tmp); + }else { +store_reg (s, i, tmp); + } } }else {/* Store. */
@@ -7997,6 +8002,9 @@ static int disas_thumb2_insn(CPUState *env, DisasContext *s, uint16_t insn_hw1) }tcg_gen_addi_i32 (addr, addr,4 ); } +if ((insn & (1 <<20 )) && (insn & (1 << rn))) { +store_reg (s, rn, tmp2); + }if (insn & (1 <<21 )) {/* Base register writeback. */ if (insn & (1 <<24 )) {
Signature · 1 line
1.7.4.msysgit.0
2011/4/24 오유연 <[email protected]>:
Quoted 3 lines
When consecutive memory locations are on page boundary, a base register may be loaded before page fault occurs. After page fault handling, it losts the memory location information. To solve this problem, loading a base register has to put back.
Thanks for finding this. I agree the fix is required, but I think I'd prefer it if the thumb code for this case handled it in the same way the disas_arm_insn() code does:
loaded_base = 0;
TCGV_UNUSED(loaded_var);
[...]
tmp = gen_ld32(addr, IS_USER(s));
if (i == 15) {
gen_bx(s, tmp);
} else if (i == rn) {
loaded_var = tmp;
loaded_base = 1;
} else {
store_reg(s, i, tmp);
}
[...]
if (loaded_base) {
store_reg(s, rn, loaded_var);}
-- PMM