| Message ID | 20210128224141.638790-21-richard.henderson@linaro.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | target-arm: Implement ARMv8.5-MemTag, user mode | expand |
On Thu, 28 Jan 2021 at 22:42, Richard Henderson <richard.henderson@linaro.org> wrote: > > The real kernel collects _TIF_MTE_ASYNC_FAULT into the current thread's > state on any kernel entry (interrupt, exception etc), and then delivers > the signal in advance of resuming the thread. > > This means that while the signal won't be delivered immediately, it will > not be delayed forever -- at minimum it will be delivered after the next > clock interrupt. > > We don't have a clock interrupt in linux-user, so we issue a cpu_kick > to signal a return to the main loop at the end of the current TB. > > Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> thanks -- PMM
diff --git a/linux-user/aarch64/target_signal.h b/linux-user/aarch64/target_signal.h index 777fb667fe..18013e1b23 100644 --- a/linux-user/aarch64/target_signal.h +++ b/linux-user/aarch64/target_signal.h @@ -21,6 +21,7 @@ typedef struct target_sigaltstack { #include "../generic/signal.h" +#define TARGET_SEGV_MTEAERR 8 /* Asynchronous ARM MTE error */ #define TARGET_SEGV_MTESERR 9 /* Synchronous ARM MTE exception */ #define TARGET_ARCH_HAS_SETUP_FRAME diff --git a/linux-user/aarch64/cpu_loop.c b/linux-user/aarch64/cpu_loop.c index b6a2e65593..7c42f65706 100644 --- a/linux-user/aarch64/cpu_loop.c +++ b/linux-user/aarch64/cpu_loop.c @@ -164,6 +164,17 @@ void cpu_loop(CPUARMState *env) EXCP_DUMP(env, "qemu: unhandled CPU exception 0x%x - aborting\n", trapnr); abort(); } + + /* Check for MTE asynchronous faults */ + if (unlikely(env->cp15.tfsr_el[0])) { + env->cp15.tfsr_el[0] = 0; + info.si_signo = TARGET_SIGSEGV; + info.si_errno = 0; + info._sifields._sigfault._addr = 0; + info.si_code = TARGET_SEGV_MTEAERR; + queue_signal(env, info.si_signo, QEMU_SI_FAULT, &info); + } + process_pending_signals(env); /* Exception return on AArch64 always clears the exclusive monitor, * so any return to running guest code implies this. diff --git a/target/arm/mte_helper.c b/target/arm/mte_helper.c index 153bd1e9df..d55f8d1e1e 100644 --- a/target/arm/mte_helper.c +++ b/target/arm/mte_helper.c @@ -565,6 +565,16 @@ static void mte_check_fail(CPUARMState *env, uint32_t desc, select = 0; } env->cp15.tfsr_el[el] |= 1 << select; +#ifdef CONFIG_USER_ONLY + /* + * Stand in for a timer irq, setting _TIF_MTE_ASYNC_FAULT, + * which then sends a SIGSEGV when the thread is next scheduled. + * This cpu will return to the main loop at the end of the TB, + * which is rather sooner than "normal". But the alternative + * is waiting until the next syscall. + */ + qemu_cpu_kick(env_cpu(env)); +#endif break; default:
The real kernel collects _TIF_MTE_ASYNC_FAULT into the current thread's state on any kernel entry (interrupt, exception etc), and then delivers the signal in advance of resuming the thread. This means that while the signal won't be delivered immediately, it will not be delayed forever -- at minimum it will be delivered after the next clock interrupt. We don't have a clock interrupt in linux-user, so we issue a cpu_kick to signal a return to the main loop at the end of the current TB. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- linux-user/aarch64/target_signal.h | 1 + linux-user/aarch64/cpu_loop.c | 11 +++++++++++ target/arm/mte_helper.c | 10 ++++++++++ 3 files changed, 22 insertions(+)