[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [RFC PATCH 7/7] linux-user: Reorg cpu_signal_handler
From: |
Philippe Mathieu-Daudé |
Subject: |
Re: [RFC PATCH 7/7] linux-user: Reorg cpu_signal_handler |
Date: |
Thu, 16 Sep 2021 10:51:51 +0200 |
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101 Thunderbird/78.11.0 |
Hi Richard,
On 9/14/21 12:05 AM, Richard Henderson wrote:
> Split out two functions into linux-user/host/arch/host-signal.h.
> Since linux-user requires a linux host, drop all of the BSD and
> Solaris ifdefs. These should be recreated under bsd-user/ when
> the current blanks there are filled.
>
> Fold the remnants of handle_cpu_signal into host_signal_handler.
>
> Drop the call to cc->tcg_ops->tlb_fill. This was simply an indirect
> method to raise SIGSEGV; it is far easier to pass the host siginfo_t
> along to the guest. This fixes all of the guest cpu_loop code that
> currently fails to properly fill in SEGV_MAPERR vs SEGV_ACCERR.
>
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
> include/exec/exec-all.h | 12 -
> linux-user/host/aarch64/host-signal.h | 73 +++
> linux-user/host/alpha/host-signal.h | 41 ++
> linux-user/host/arm/host-signal.h | 30 ++
> linux-user/host/i386/host-signal.h | 24 +
> linux-user/host/mips/host-signal.h | 61 +++
> linux-user/host/ppc/host-signal.h | 24 +
> linux-user/host/ppc64/host-signal.h | 1 +
> linux-user/host/riscv32/host-signal.h | 57 +++
> linux-user/host/riscv64/host-signal.h | 1 +
> linux-user/host/s390/host-signal.h | 92 ++++
> linux-user/host/s390x/host-signal.h | 1 +
> linux-user/host/sparc/host-signal.h | 53 ++
> linux-user/host/sparc64/host-signal.h | 1 +
> linux-user/host/x86_64/host-signal.h | 24 +
> accel/tcg/user-exec.c | 712 --------------------------
> linux-user/signal.c | 93 ++--
> 17 files changed, 543 insertions(+), 757 deletions(-)
> create mode 100644 linux-user/host/aarch64/host-signal.h
> create mode 100644 linux-user/host/alpha/host-signal.h
> create mode 100644 linux-user/host/arm/host-signal.h
> create mode 100644 linux-user/host/i386/host-signal.h
> create mode 100644 linux-user/host/mips/host-signal.h
> create mode 100644 linux-user/host/ppc/host-signal.h
> create mode 100644 linux-user/host/ppc64/host-signal.h
> create mode 100644 linux-user/host/riscv32/host-signal.h
> create mode 100644 linux-user/host/riscv64/host-signal.h
> create mode 100644 linux-user/host/s390/host-signal.h
> create mode 100644 linux-user/host/s390x/host-signal.h
> create mode 100644 linux-user/host/sparc/host-signal.h
> create mode 100644 linux-user/host/sparc64/host-signal.h
> create mode 100644 linux-user/host/x86_64/host-signal.h
> diff --git a/linux-user/signal.c b/linux-user/signal.c
> index 73c0f9066b..509dad7850 100644
> --- a/linux-user/signal.c
> +++ b/linux-user/signal.c
> -static void host_signal_handler(int host_signum, siginfo_t *info,
> - void *puc)
> +static void host_signal_handler(int host_sig, siginfo_t *info, void *puc)
> {
> CPUArchState *env = thread_cpu->env_ptr;
> CPUState *cpu = env_cpu(env);
> TaskState *ts = cpu->opaque;
> -
> - int sig;
> + bool sync_sig = false;
> target_siginfo_t tinfo;
> ucontext_t *uc = puc;
> struct emulated_sigtable *k;
> + uintptr_t pc = 0;
> + int guest_sig;
>
> - /* the CPU emulator uses some host signals to detect exceptions,
> - we forward to it some signals */
> - if ((host_signum == SIGSEGV || host_signum == SIGBUS)
> - && info->si_code > 0) {
> - if (cpu_signal_handler(host_signum, info, puc))
> + /*
> + * Non-spoofed SIGSEGV and SIGBUS are synchronous, and need special
> + * handling wrt signal blocking and unwinding. SIGSEGV may need to
> + * remove write-protection and restart the instruction.
> + */
> + if ((host_sig == SIGSEGV || host_sig == SIGBUS) && info->si_code > 0) {
> + pc = adjust_signal_pc(host_signal_pc(uc));
> + if (host_sig == SIGSEGV &&
> + info->si_code == SEGV_ACCERR &&
> + host_sigsegv_write(info, uc) &&
> + handle_sigsegv_accerr_write(cpu, &uc->uc_sigmask, pc,
> + (uintptr_t)info->si_addr)) {
> return;
> + }
> + sync_sig = true;
> + } else {
> + rewind_if_in_safe_syscall(puc);
> +
> + /*
> + * Block host signals until target signal handler entered.
> + * We can't block SIGSEGV or SIGBUS while we're executing
> + * guest code in case the guest code provokes one in the
> + * window between now and it getting out to the main loop.
> + * Signals will be unblocked again in process_pending_signals().
> + *
> + * WARNING: we cannot use sigfillset() here because the uc_sigmask
> + * field is a kernel sigset_t, which is much smaller than the
> + * libc sigset_t which sigfillset() operates on. Using sigfillset()
> + * would write 0xff bytes off the end of the structure and trash
> + * data on the struct.
> + * We can't use sizeof(uc->uc_sigmask) either, because the libc
> + * headers define the struct field with the wrong (too large) type.
> + */
> + memset(&uc->uc_sigmask, 0xff, SIGSET_T_SIZE);
> + sigdelset(&uc->uc_sigmask, SIGSEGV);
> + sigdelset(&uc->uc_sigmask, SIGBUS);
> }
>
> /* get target signal number */
> - sig = host_to_target_signal(host_signum);
> - if (sig < 1 || sig > TARGET_NSIG)
> + guest_sig = host_to_target_signal(host_sig);
> + if (guest_sig < 1 || guest_sig > TARGET_NSIG) {
> return;
> - trace_user_host_signal(env, host_signum, sig);
> -
> - rewind_if_in_safe_syscall(puc);
> + }
> + trace_user_host_signal(env, host_sig, guest_sig);
>
> host_to_target_siginfo_noswap(&tinfo, info);
> - k = &ts->sigtab[sig - 1];
> + k = &ts->sigtab[guest_sig - 1];
> k->info = tinfo;
> - k->pending = sig;
> + k->pending = guest_sig;
> ts->signal_pending = 1;
>
> - /* Block host signals until target signal handler entered. We
> - * can't block SIGSEGV or SIGBUS while we're executing guest
> - * code in case the guest code provokes one in the window between
> - * now and it getting out to the main loop. Signals will be
> - * unblocked again in process_pending_signals().
> - *
> - * WARNING: we cannot use sigfillset() here because the uc_sigmask
> - * field is a kernel sigset_t, which is much smaller than the
> - * libc sigset_t which sigfillset() operates on. Using sigfillset()
> - * would write 0xff bytes off the end of the structure and trash
> - * data on the struct.
> - * We can't use sizeof(uc->uc_sigmask) either, because the libc
> - * headers define the struct field with the wrong (too large) type.
> + /*
> + * For synchronous signals, unwind the cpu state to the faulting
> + * insn and then exit back to the main loop so that the signal
> + * is delivered immediately.
> */
> - memset(&uc->uc_sigmask, 0xff, SIGSET_T_SIZE);
> - sigdelset(&uc->uc_sigmask, SIGSEGV);
> - sigdelset(&uc->uc_sigmask, SIGBUS);
> + if (sync_sig) {
> + clear_helper_retaddr();
> + sigprocmask(SIG_SETMASK, &uc->uc_sigmask, NULL);
> + cpu->exception_index = EXCP_INTERRUPT;
> + cpu_loop_exit_restore(cpu, pc);
> + }
>
> - /* interrupt the virtual CPU as soon as possible */
> + /*
> + * Interrupt the virtual CPU as soon as possible, but for now
> + * return to continue with the current TB.
> + */
> cpu_exit(thread_cpu);
> }
Is it possible to split this patch per architectures,
doing for the first arch:
#if ARCH1
new host_signal_handler() {}
#else
old unmodified host_signal_handler() {}
#endif
Then for the second:
#if ARCH1 || ARCH2
new host_signal_handler() {}
#else
old unmodified host_signal_handler() {}
#endif
Last patch being cleaning the transition:
- #if ARCH1 || ARCH2 || ...
new host_signal_handler() {}
- #else
-
- old unmodified host_signal_handler() {}
-
- #endif
?
- [RFC PATCH 3/7] accel/tcg: Split out handle_sigsegv_accerr_write, (continued)
- [RFC PATCH 3/7] accel/tcg: Split out handle_sigsegv_accerr_write, Richard Henderson, 2021/09/13
- [RFC PATCH 1/7] include/exec: Move cpu_signal_handler declaration, Richard Henderson, 2021/09/13
- [RFC PATCH 4/7] accel/tcg: Move clear_helper_retaddr to cpu loop, Richard Henderson, 2021/09/13
- [RFC PATCH 5/7] accel/tcg: Fold cpu_exit_tb_from_sighandler into caller, Richard Henderson, 2021/09/13
- [RFC PATCH 7/7] linux-user: Reorg cpu_signal_handler, Richard Henderson, 2021/09/13
- Re: [RFC PATCH 0/7] linux-user: Streamline handling of SIGSEGV/SIGBUS, Richard Henderson, 2021/09/13