| Message ID | 20250504205714.3432096-11-richard.henderson@linaro.org |
|---|---|
| State | New |
| Headers | show |
| Series | accel/tcg: Fix cross-page pointer wrapping issue | expand |
On 4/5/25 22:57, Richard Henderson wrote: > Use the existing wrap_address function. > > Cc: qemu-s390x@nongnu.org > Signed-off-by: Richard Henderson <richard.henderson@linaro.org> > --- > target/s390x/cpu.c | 9 +++++++++ > 1 file changed, 9 insertions(+) > > diff --git a/target/s390x/cpu.c b/target/s390x/cpu.c > index 9c1158ebcc..f05ce317da 100644 > --- a/target/s390x/cpu.c > +++ b/target/s390x/cpu.c > @@ -347,6 +347,14 @@ static TCGTBCPUState s390x_get_tb_cpu_state(CPUState *cs) > }; > } > > +#ifndef CONFIG_USER_ONLY > +static vaddr s390_pointer_wrap(CPUState *cs, int mmu_idx, > + vaddr result, vaddr base) > +{ > + return wrap_address(cpu_env(cs), result); > +} > +#endif > + > static const TCGCPUOps s390_tcg_ops = { > .mttcg_supported = true, > .precise_smc = true, > @@ -367,6 +375,7 @@ static const TCGCPUOps s390_tcg_ops = { > .record_sigbus = s390_cpu_record_sigbus, > #else > .tlb_fill = s390_cpu_tlb_fill, > + .pointer_wrap = s390_pointer_wrap, As future cleanup, we might now remove the wrap_address() calls in target/s390x/tcg/crypto_helper.c, and target/s390x/tcg/vec_helper.c, is that correct? Also some uses in target/s390x/tcg/mem_helper.c. Meanwhile for this patch: Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
On 5/5/25 07:41, Philippe Mathieu-Daudé wrote: > On 4/5/25 22:57, Richard Henderson wrote: >> Use the existing wrap_address function. >> >> Cc: qemu-s390x@nongnu.org >> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> >> --- >> target/s390x/cpu.c | 9 +++++++++ >> 1 file changed, 9 insertions(+) >> >> diff --git a/target/s390x/cpu.c b/target/s390x/cpu.c >> index 9c1158ebcc..f05ce317da 100644 >> --- a/target/s390x/cpu.c >> +++ b/target/s390x/cpu.c >> @@ -347,6 +347,14 @@ static TCGTBCPUState s390x_get_tb_cpu_state(CPUState *cs) >> }; >> } >> +#ifndef CONFIG_USER_ONLY >> +static vaddr s390_pointer_wrap(CPUState *cs, int mmu_idx, >> + vaddr result, vaddr base) >> +{ >> + return wrap_address(cpu_env(cs), result); >> +} >> +#endif >> + >> static const TCGCPUOps s390_tcg_ops = { >> .mttcg_supported = true, >> .precise_smc = true, >> @@ -367,6 +375,7 @@ static const TCGCPUOps s390_tcg_ops = { >> .record_sigbus = s390_cpu_record_sigbus, >> #else >> .tlb_fill = s390_cpu_tlb_fill, >> + .pointer_wrap = s390_pointer_wrap, > > As future cleanup, we might now remove the wrap_address() calls > in target/s390x/tcg/crypto_helper.c, and target/s390x/tcg/vec_helper.c, > is that correct? > > Also some uses in target/s390x/tcg/mem_helper.c. No, not correct. The new pointer_wrap hook is only used for unaligned accesses that cross page boundaries. It does not apply to a sequence of individual accesses like we have in the s390x helpers. r~
diff --git a/target/s390x/cpu.c b/target/s390x/cpu.c index 9c1158ebcc..f05ce317da 100644 --- a/target/s390x/cpu.c +++ b/target/s390x/cpu.c @@ -347,6 +347,14 @@ static TCGTBCPUState s390x_get_tb_cpu_state(CPUState *cs) }; } +#ifndef CONFIG_USER_ONLY +static vaddr s390_pointer_wrap(CPUState *cs, int mmu_idx, + vaddr result, vaddr base) +{ + return wrap_address(cpu_env(cs), result); +} +#endif + static const TCGCPUOps s390_tcg_ops = { .mttcg_supported = true, .precise_smc = true, @@ -367,6 +375,7 @@ static const TCGCPUOps s390_tcg_ops = { .record_sigbus = s390_cpu_record_sigbus, #else .tlb_fill = s390_cpu_tlb_fill, + .pointer_wrap = s390_pointer_wrap, .cpu_exec_interrupt = s390_cpu_exec_interrupt, .cpu_exec_halt = s390_cpu_has_work, .cpu_exec_reset = cpu_reset,
Use the existing wrap_address function. Cc: qemu-s390x@nongnu.org Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/s390x/cpu.c | 9 +++++++++ 1 file changed, 9 insertions(+)