[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH v4] ppc: Fix signal delivery in ppc-user and ppc
From: |
David Gibson |
Subject: |
Re: [Qemu-devel] [PATCH v4] ppc: Fix signal delivery in ppc-user and ppc64-user |
Date: |
Mon, 12 Sep 2016 11:56:40 +1000 |
User-agent: |
Mutt/1.7.0 (2016-08-17) |
On Wed, Aug 03, 2016 at 10:38:51PM +1000, Benjamin Herrenschmidt wrote:
> There were a number of bugs in the implementation:
>
> - The structure alignment was wrong for 64-bit.
>
> - Also 64-bit only does RT signals.
>
> - On 64-bit, we need to put a pointer to the (aligned) vector registers
> in the frame and use it for restoring
>
> - We had endian bugs when saving/restoring vector registers
>
> - My recent fixes for exception NIP broke sigreturn in user mode
> causing us to resume one instruction too far.
>
> - Add VSR second halves
>
> Signed-off-by: Benjamin Herrenschmidt <address@hidden>
Applied to ppc-for-2.8. thanks.
> ---
> v2. Fix endian bugs too
> Fix bad PC on sigreturn
>
> v3. Add missing VSX second halves
> Tested with ppc32, ppc64be and ppc64le, verified reading
> and writing VSX and VMX registers from a signal handler
> and observing the result in the main program. Compared
> successfully to running in actual hardware.
>
> v4. Improve comments about alignment.
>
> linux-user/main.c | 2 +-
> linux-user/ppc/syscall_nr.h | 2 +
> linux-user/signal.c | 124
> +++++++++++++++++++++++++++++++-------------
> 3 files changed, 90 insertions(+), 38 deletions(-)
>
> diff --git a/linux-user/main.c b/linux-user/main.c
> index 1d149dc..24f34e6 100644
> --- a/linux-user/main.c
> +++ b/linux-user/main.c
> @@ -2003,12 +2003,12 @@ void cpu_loop(CPUPPCState *env)
> if (ret == -TARGET_ERESTARTSYS) {
> break;
> }
> - env->nip += 4;
> if (ret == (target_ulong)(-TARGET_QEMU_ESIGRETURN)) {
> /* Returning from a successful sigreturn syscall.
> Avoid corrupting register state. */
> break;
> }
> + env->nip += 4;
> if (ret > (target_ulong)(-515)) {
> env->crf[0] |= 0x1;
> ret = -ret;
> diff --git a/linux-user/ppc/syscall_nr.h b/linux-user/ppc/syscall_nr.h
> index 46ed8a6..afa3654 100644
> --- a/linux-user/ppc/syscall_nr.h
> +++ b/linux-user/ppc/syscall_nr.h
> @@ -120,7 +120,9 @@
> #define TARGET_NR_sysinfo 116
> #define TARGET_NR_ipc 117
> #define TARGET_NR_fsync 118
> +#if !defined(TARGET_PPC64)
> #define TARGET_NR_sigreturn 119
> +#endif
> #define TARGET_NR_clone 120
> #define TARGET_NR_setdomainname 121
> #define TARGET_NR_uname 122
> diff --git a/linux-user/signal.c b/linux-user/signal.c
> index 9a4d894..55d9046 100644
> --- a/linux-user/signal.c
> +++ b/linux-user/signal.c
> @@ -4408,7 +4408,12 @@ struct target_mcontext {
> target_ulong mc_gregs[48];
> /* Includes fpscr. */
> uint64_t mc_fregs[33];
> +#if defined(TARGET_PPC64)
> + /* Pointer to the vector regs */
> + target_ulong v_regs;
> +#else
> target_ulong mc_pad[2];
> +#endif
> /* We need to handle Altivec and SPE at the same time, which no
> kernel needs to do. Fortunately, the kernel defines this bit to
> be Altivec-register-large all the time, rather than trying to
> @@ -4418,15 +4423,30 @@ struct target_mcontext {
> uint32_t spe[33];
> /* Altivec vector registers. The packing of VSCR and VRSAVE
> varies depending on whether we're PPC64 or not: PPC64 splits
> - them apart; PPC32 stuffs them together. */
> + them apart; PPC32 stuffs them together.
> + We also need to account for the VSX registers on PPC64
> + */
> #if defined(TARGET_PPC64)
> -#define QEMU_NVRREG 34
> +#define QEMU_NVRREG (34 + 16)
> + /* On ppc64, this mcontext structure is naturally *unaligned*,
> + * or rather it is aligned on a 8 bytes boundary but not on
> + * a 16 bytes one. This pad fixes it up. This is also why the
> + * vector regs are referenced by the v_regs pointer above so
> + * any amount of padding can be added here
> + */
> + target_ulong pad;
> #else
> + /* On ppc32, we are already aligned to 16 bytes */
> #define QEMU_NVRREG 33
> #endif
> - ppc_avr_t altivec[QEMU_NVRREG];
> + /* We cannot use ppc_avr_t here as we do *not* want the implied
> + * 16-bytes alignment that would result from it. This would have
> + * the effect of making the whole struct target_mcontext aligned
> + * which breaks the layout of struct target_ucontext on ppc64.
> + */
> + uint64_t altivec[QEMU_NVRREG][2];
> #undef QEMU_NVRREG
> - } mc_vregs __attribute__((__aligned__(16)));
> + } mc_vregs;
> };
>
> /* See arch/powerpc/include/asm/sigcontext.h. */
> @@ -4580,6 +4600,16 @@ static target_ulong get_sigframe(struct
> target_sigaction *ka,
> return (oldsp - frame_size) & ~0xFUL;
> }
>
> +#if ((defined(TARGET_WORDS_BIGENDIAN) && defined(HOST_WORDS_BIGENDIAN)) || \
> + (!defined(HOST_WORDS_BIGENDIAN) && !defined(TARGET_WORDS_BIGENDIAN)))
> +#define PPC_VEC_HI 0
> +#define PPC_VEC_LO 1
> +#else
> +#define PPC_VEC_HI 1
> +#define PPC_VEC_LO 0
> +#endif
> +
> +
> static void save_user_regs(CPUPPCState *env, struct target_mcontext *frame)
> {
> target_ulong msr = env->msr;
> @@ -4606,18 +4636,33 @@ static void save_user_regs(CPUPPCState *env, struct
> target_mcontext *frame)
>
> /* Save Altivec registers if necessary. */
> if (env->insns_flags & PPC_ALTIVEC) {
> + uint32_t *vrsave;
> for (i = 0; i < ARRAY_SIZE(env->avr); i++) {
> ppc_avr_t *avr = &env->avr[i];
> - ppc_avr_t *vreg = &frame->mc_vregs.altivec[i];
> + ppc_avr_t *vreg = (ppc_avr_t *)&frame->mc_vregs.altivec[i];
>
> - __put_user(avr->u64[0], &vreg->u64[0]);
> - __put_user(avr->u64[1], &vreg->u64[1]);
> + __put_user(avr->u64[PPC_VEC_HI], &vreg->u64[0]);
> + __put_user(avr->u64[PPC_VEC_LO], &vreg->u64[1]);
> }
> /* Set MSR_VR in the saved MSR value to indicate that
> frame->mc_vregs contains valid data. */
> msr |= MSR_VR;
> - __put_user((uint32_t)env->spr[SPR_VRSAVE],
> - &frame->mc_vregs.altivec[32].u32[3]);
> +#if defined(TARGET_PPC64)
> + vrsave = (uint32_t *)&frame->mc_vregs.altivec[33];
> + /* 64-bit needs to put a pointer to the vectors in the frame */
> + __put_user(h2g(frame->mc_vregs.altivec), &frame->v_regs);
> +#else
> + vrsave = (uint32_t *)&frame->mc_vregs.altivec[32];
> +#endif
> + __put_user((uint32_t)env->spr[SPR_VRSAVE], vrsave);
> + }
> +
> + /* Save VSX second halves */
> + if (env->insns_flags2 & PPC2_VSX) {
> + uint64_t *vsregs = (uint64_t *)&frame->mc_vregs.altivec[34];
> + for (i = 0; i < ARRAY_SIZE(env->vsr); i++) {
> + __put_user(env->vsr[i], &vsregs[i]);
> + }
> }
>
> /* Save floating point registers. */
> @@ -4697,17 +4742,39 @@ static void restore_user_regs(CPUPPCState *env,
>
> /* Restore Altivec registers if necessary. */
> if (env->insns_flags & PPC_ALTIVEC) {
> + ppc_avr_t *v_regs;
> + uint32_t *vrsave;
> +#if defined(TARGET_PPC64)
> + uint64_t v_addr;
> + /* 64-bit needs to recover the pointer to the vectors from the frame
> */
> + __get_user(v_addr, &frame->v_regs);
> + v_regs = g2h(v_addr);
> +#else
> + v_regs = (ppc_avr_t *)frame->mc_vregs.altivec;
> +#endif
> for (i = 0; i < ARRAY_SIZE(env->avr); i++) {
> ppc_avr_t *avr = &env->avr[i];
> - ppc_avr_t *vreg = &frame->mc_vregs.altivec[i];
> + ppc_avr_t *vreg = &v_regs[i];
>
> - __get_user(avr->u64[0], &vreg->u64[0]);
> - __get_user(avr->u64[1], &vreg->u64[1]);
> + __get_user(avr->u64[PPC_VEC_HI], &vreg->u64[0]);
> + __get_user(avr->u64[PPC_VEC_LO], &vreg->u64[1]);
> }
> /* Set MSR_VEC in the saved MSR value to indicate that
> frame->mc_vregs contains valid data. */
> - __get_user(env->spr[SPR_VRSAVE],
> - (target_ulong *)(&frame->mc_vregs.altivec[32].u32[3]));
> +#if defined(TARGET_PPC64)
> + vrsave = (uint32_t *)&v_regs[33];
> +#else
> + vrsave = (uint32_t *)&v_regs[32];
> +#endif
> + __get_user(env->spr[SPR_VRSAVE], vrsave);
> + }
> +
> + /* Restore VSX second halves */
> + if (env->insns_flags2 & PPC2_VSX) {
> + uint64_t *vsregs = (uint64_t *)&frame->mc_vregs.altivec[34];
> + for (i = 0; i < ARRAY_SIZE(env->vsr); i++) {
> + __get_user(env->vsr[i], &vsregs[i]);
> + }
> }
>
> /* Restore floating point registers. */
> @@ -4738,6 +4805,7 @@ static void restore_user_regs(CPUPPCState *env,
> }
> }
>
> +#if !defined(TARGET_PPC64)
> static void setup_frame(int sig, struct target_sigaction *ka,
> target_sigset_t *set, CPUPPCState *env)
> {
> @@ -4745,9 +4813,6 @@ static void setup_frame(int sig, struct
> target_sigaction *ka,
> struct target_sigcontext *sc;
> target_ulong frame_addr, newsp;
> int err = 0;
> -#if defined(TARGET_PPC64)
> - struct image_info *image = ((TaskState *)thread_cpu->opaque)->info;
> -#endif
>
> frame_addr = get_sigframe(ka, env, sizeof(*frame));
> trace_user_setup_frame(env, frame_addr);
> @@ -4757,11 +4822,7 @@ static void setup_frame(int sig, struct
> target_sigaction *ka,
>
> __put_user(ka->_sa_handler, &sc->handler);
> __put_user(set->sig[0], &sc->oldmask);
> -#if TARGET_ABI_BITS == 64
> - __put_user(set->sig[0] >> 32, &sc->_unused[3]);
> -#else
> __put_user(set->sig[1], &sc->_unused[3]);
> -#endif
> __put_user(h2g(&frame->mctx), &sc->regs);
> __put_user(sig, &sc->signal);
>
> @@ -4790,22 +4851,7 @@ static void setup_frame(int sig, struct
> target_sigaction *ka,
> env->gpr[3] = sig;
> env->gpr[4] = frame_addr + offsetof(struct target_sigframe, sctx);
>
> -#if defined(TARGET_PPC64)
> - if (get_ppc64_abi(image) < 2) {
> - /* ELFv1 PPC64 function pointers are pointers to OPD entries. */
> - struct target_func_ptr *handler =
> - (struct target_func_ptr *)g2h(ka->_sa_handler);
> - env->nip = tswapl(handler->entry);
> - env->gpr[2] = tswapl(handler->toc);
> - } else {
> - /* ELFv2 PPC64 function pointers are entry points, but R12
> - * must also be set */
> - env->nip = tswapl((target_ulong) ka->_sa_handler);
> - env->gpr[12] = env->nip;
> - }
> -#else
> env->nip = (target_ulong) ka->_sa_handler;
> -#endif
>
> /* Signal handlers are entered in big-endian mode. */
> env->msr &= ~(1ull << MSR_LE);
> @@ -4817,6 +4863,7 @@ sigsegv:
> unlock_user_struct(frame, frame_addr, 1);
> force_sig(TARGET_SIGSEGV);
> }
> +#endif /* !defined(TARGET_PPC64) */
>
> static void setup_rt_frame(int sig, struct target_sigaction *ka,
> target_siginfo_t *info,
> @@ -4914,6 +4961,7 @@ sigsegv:
>
> }
>
> +#if !defined(TARGET_PPC64)
> long do_sigreturn(CPUPPCState *env)
> {
> struct target_sigcontext *sc = NULL;
> @@ -4950,6 +4998,7 @@ sigsegv:
> force_sig(TARGET_SIGSEGV);
> return 0;
> }
> +#endif /* !defined(TARGET_PPC64) */
>
> /* See arch/powerpc/kernel/signal_32.c. */
> static int do_setcontext(struct target_ucontext *ucp, CPUPPCState *env, int
> sig)
> @@ -5893,7 +5942,8 @@ static void handle_pending_signal(CPUArchState
> *cpu_env, int sig,
> #endif
> /* prepare the stack frame of the virtual CPU */
> #if defined(TARGET_ABI_MIPSN32) || defined(TARGET_ABI_MIPSN64) \
> - || defined(TARGET_OPENRISC) || defined(TARGET_TILEGX)
> + || defined(TARGET_OPENRISC) || defined(TARGET_TILEGX) \
> + || defined(TARGET_PPC64)
> /* These targets do not have traditional signals. */
> setup_rt_frame(sig, sa, &k->info, &target_old_set, cpu_env);
> #else
>
--
David Gibson | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_
| _way_ _around_!
http://www.ozlabs.org/~dgibson
signature.asc
Description: PGP signature
[Prev in Thread] |
Current Thread |
[Next in Thread] |
- Re: [Qemu-devel] [PATCH v4] ppc: Fix signal delivery in ppc-user and ppc64-user,
David Gibson <=