All of lore.kernel.org
 help / color / mirror / Atom feed
From: "Edgar E. Iglesias" <edgar.iglesias@gmail.com>
To: Peter Maydell <peter.maydell@linaro.org>
Cc: "Rob Herring" <rob.herring@linaro.org>,
	"Peter Crosthwaite" <peter.crosthwaite@xilinx.com>,
	"Fabian Aggeler" <aggelerf@ethz.ch>,
	"QEMU Developers" <qemu-devel@nongnu.org>,
	"Alexander Graf" <agraf@suse.de>,
	"Greg Bellows" <greg.bellows@linaro.org>,
	"Paolo Bonzini" <pbonzini@redhat.com>,
	"Alex Bennée" <alex.bennee@linaro.org>,
	"Christoffer Dall" <christoffer.dall@linaro.org>,
	"Richard Henderson" <rth@twiddle.net>
Subject: Re: [Qemu-devel] [PATCH v6 07/10] target-arm: A64: Emulate the HVC insn
Date: Fri, 26 Sep 2014 08:20:57 +1000	[thread overview]
Message-ID: <20140925222057.GS16081@toto> (raw)
In-Reply-To: <CAFEAcA9YSd0CcMFT9ocBXoKhvFxgmZv6RgxEMSedSkCvv2Khkw@mail.gmail.com>

On Thu, Sep 25, 2014 at 07:39:32PM +0100, Peter Maydell wrote:
> On 13 September 2014 05:29, Edgar E. Iglesias <edgar.iglesias@gmail.com> wrote:
> 
> > --- a/target-arm/helper.c
> > +++ b/target-arm/helper.c
> > @@ -3652,7 +3652,33 @@ void switch_mode(CPUARMState *env, int mode)
> >   */
> >  unsigned int arm_excp_target_el(CPUState *cs, unsigned int excp_idx)
> >  {
> > -    return 1;
> > +    CPUARMState *env = cs->env_ptr;
> 
> Everywhere else that we start with a CPUState* and need
> a CPUARMState* we do it with:
> 
>     ARMCPU *cpu = ARM_CPU(cs);
>     CPUARMState *env = &cpu->env;
> 
> I think we should be consistent.
> 
> > +    unsigned int cur_el = arm_current_pl(env);
> > +    unsigned int target_el = 1;
> > +    bool route_to_el2 = false;
> > +    /* FIXME: Use actual secure state.  */
> > +    bool secure = false;
> > +
> > +    if (!env->aarch64) {
> > +        /* TODO: Add EL2 and 3 exception handling for AArch32.  */
> > +        return 1;
> > +    }
> > +
> > +    if (!secure
> > +        && arm_feature(env, ARM_FEATURE_EL2)
> > +        && cur_el < 2
> > +        && (env->cp15.hcr_el2 & HCR_TGE)) {
> > +        route_to_el2 = true;
> > +    }
> 
> HCR.TGE isn't actually relevant for some exceptions
> (eg SMC), and the HVC handling you have below
> effectively ends up ignoring the route_to_el2
> information. That suggests to me that you should put
> this code in a default: case in the switch below.

I don't really test nor support TGE so I'll just drop the TGE part.

> 
> > +    target_el = MAX(cur_el, route_to_el2 ? 2 : 1);
> > +
> > +    switch (excp_idx) {
> > +    case EXCP_HVC:
> > +        target_el = MAX(target_el, 2);
> > +        break;
> > +    }
> > +    return target_el;
> >  }
> >
> >  static void v7m_push(CPUARMState *env, uint32_t val)
> > diff --git a/target-arm/helper.h b/target-arm/helper.h
> > index 1d7003b..75fc1b3 100644
> > --- a/target-arm/helper.h
> > +++ b/target-arm/helper.h
> > @@ -50,6 +50,7 @@ DEF_HELPER_2(exception_internal, void, env, i32)
> >  DEF_HELPER_3(exception_with_syndrome, void, env, i32, i32)
> >  DEF_HELPER_1(wfi, void, env)
> >  DEF_HELPER_1(wfe, void, env)
> > +DEF_HELPER_1(pre_hvc, void, env)
> >
> >  DEF_HELPER_3(cpsr_write, void, env, i32, i32)
> >  DEF_HELPER_1(cpsr_read, i32, env)
> > diff --git a/target-arm/internals.h b/target-arm/internals.h
> > index 64751a0..afcc4e0 100644
> > --- a/target-arm/internals.h
> > +++ b/target-arm/internals.h
> > @@ -53,6 +53,7 @@ static const char * const excnames[] = {
> >      [EXCP_EXCEPTION_EXIT] = "QEMU v7M exception exit",
> >      [EXCP_KERNEL_TRAP] = "QEMU intercept of kernel commpage",
> >      [EXCP_STREX] = "QEMU intercept of STREX",
> > +    [EXCP_HVC] = "Hypervisor Call",
> >  };
> >
> >  static inline void arm_log_exception(int idx)
> > @@ -215,6 +216,11 @@ static inline uint32_t syn_aa64_svc(uint32_t imm16)
> >      return (EC_AA64_SVC << ARM_EL_EC_SHIFT) | ARM_EL_IL | (imm16 & 0xffff);
> >  }
> >
> > +static inline uint32_t syn_aa64_hvc(uint32_t imm16)
> > +{
> > +    return (EC_AA64_HVC << ARM_EL_EC_SHIFT) | ARM_EL_IL | (imm16 & 0xffff);
> > +}
> > +
> >  static inline uint32_t syn_aa32_svc(uint32_t imm16, bool is_thumb)
> >  {
> >      return (EC_AA32_SVC << ARM_EL_EC_SHIFT) | (imm16 & 0xffff)
> > diff --git a/target-arm/op_helper.c b/target-arm/op_helper.c
> > index b956216..8441c27 100644
> > --- a/target-arm/op_helper.c
> > +++ b/target-arm/op_helper.c
> > @@ -374,6 +374,38 @@ void HELPER(clear_pstate_ss)(CPUARMState *env)
> >      env->pstate &= ~PSTATE_SS;
> >  }
> >
> > +void HELPER(pre_hvc)(CPUARMState *env)
> > +{
> > +    int cur_el = arm_current_pl(env);
> > +    /* FIXME: Use actual secure state.  */
> > +    bool secure = false;
> > +    bool udef;
> 
> The standard spelling is "UNDEF" (except in the constant name
> "EXCP_UDEF", which we should probably change, though I won't
> actually do that now since it will cause a bunch of pointless
> conflicts for you. Another tempting rename is EXCP_SWI and
> DISAS_SWI to EXCP_SVC and DISAS_SVC, though in that case the
> old names have the virtue of having been right once.)
> 
> > +
> > +    /* We've already checked that EL2 exists at translation time.
> > +     * EL3.HCE has priority over EL2.HCD.
> > +     */
> > +    if (arm_feature(env, ARM_FEATURE_EL3)) {
> > +        udef = !(env->cp15.scr_el3 & SCR_HCE);
> > +    } else {
> > +        udef = env->cp15.hcr_el2 & HCR_HCD;
> > +    }
> > +
> > +    /* In ARMv7 and ARMv8/AArch32, HVC is udef in secure state.
> > +     * For ARMv8/AArch64, HVC is allowed in EL3.
> > +     * Note that we've already trapped HVC from EL0 at translation
> > +     * time.
> > +     */
> > +    if (secure && (!is_a64(env) || cur_el == 1)) {
> > +        udef = true;
> > +    }
> > +
> > +    if (udef) {
> > +        /* UDEFs trap on the HVC, roll back to the PC to the HVC insn.  */
> 
> "with the PC pointing to the HVC insn" ?  At any rate, nothing
> in this function is rolling anything back.

This is a hang-over from v5 prior to the singlestepping support
when the code was rolling back the pc.

HVCs that lead to undef are taken with the prefered exception
return pointing to the HVC, i.e before advancing the PC.

> 
> > +        env->exception.syndrome = syn_uncategorized();
> > +        raise_exception(env, EXCP_UDEF);
> > +    }
> > +}
> > +
> >  void HELPER(exception_return)(CPUARMState *env)
> >  {
> >      int cur_el = arm_current_pl(env);
> > diff --git a/target-arm/translate-a64.c b/target-arm/translate-a64.c
> > index 8e66b6c..8eaa85f 100644
> > --- a/target-arm/translate-a64.c
> > +++ b/target-arm/translate-a64.c
> > @@ -1473,20 +1473,32 @@ static void disas_exc(DisasContext *s, uint32_t insn)
> >
> >      switch (opc) {
> >      case 0:
> > -        /* SVC, HVC, SMC; since we don't support the Virtualization
> > -         * or TrustZone extensions these all UNDEF except SVC.
> > -         */
> > -        if (op2_ll != 1) {
> > -            unallocated_encoding(s);
> > -            break;
> > -        }
> >          /* For SVC, HVC and SMC we advance the single-step state
> >           * machine before taking the exception. This is architecturally
> >           * mandated, to ensure that single-stepping a system call
> >           * instruction works properly.
> >           */
> > -        gen_ss_advance(s);
> > -        gen_exception_insn(s, 0, EXCP_SWI, syn_aa64_svc(imm16));
> > +        switch (op2_ll) {
> > +        case 1:
> > +            gen_ss_advance(s);
> > +            gen_exception_insn(s, 0, EXCP_SWI, syn_aa64_svc(imm16));
> > +            break;
> > +        case 2:
> > +            if (!arm_dc_feature(s, ARM_FEATURE_EL2) || s->current_pl == 0) {
> > +                unallocated_encoding(s);
> > +                break;
> > +            }
> > +            /* The pre HVC helper handles cases when HVC gets trapped
> > +             * as an undefined insn by runtime configuration.  */
> 
> "*/" on a line of its own, please.
> 
> > +            gen_a64_set_pc_im(s->pc - 4);
> > +            gen_helper_pre_hvc(cpu_env);
> > +            gen_ss_advance(s);
> > +            gen_exception_insn(s, 0, EXCP_HVC, syn_aa64_hvc(imm16));
> > +            break;
> > +        default:
> > +            unallocated_encoding(s);
> > +            break;
> > +        }
> >          break;
> >      case 1:
> >          if (op2_ll != 0) {
> > --
> > 1.9.1
> >
> 
> thanks
> -- PMM

  reply	other threads:[~2014-09-25 22:26 UTC|newest]

Thread overview: 34+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2014-09-13  4:29 [Qemu-devel] [PATCH v6 00/10] target-arm: Parts of the AArch64 EL2/3 exception model Edgar E. Iglesias
2014-09-13  4:29 ` [Qemu-devel] [PATCH v6 01/10] target-arm: Add HCR_EL2 Edgar E. Iglesias
2014-09-13  4:29 ` [Qemu-devel] [PATCH v6 02/10] target-arm: Add SCR_EL3 Edgar E. Iglesias
2014-09-17 15:49   ` Greg Bellows
2014-09-25 18:15   ` Peter Maydell
2014-09-25 19:49     ` Greg Bellows
2014-09-25 19:53       ` Peter Maydell
2014-09-25 20:00         ` Greg Bellows
2014-09-25 22:12     ` Edgar E. Iglesias
2014-09-13  4:29 ` [Qemu-devel] [PATCH v6 03/10] target-arm: A64: Refactor aarch64_cpu_do_interrupt Edgar E. Iglesias
2014-09-13  4:29 ` [Qemu-devel] [PATCH v6 04/10] target-arm: Break out exception masking to a separate func Edgar E. Iglesias
2014-09-13  4:29 ` [Qemu-devel] [PATCH v6 05/10] target-arm: Don't take interrupts targeting lower ELs Edgar E. Iglesias
2014-09-13  4:29 ` [Qemu-devel] [PATCH v6 06/10] target-arm: A64: Correct updates to FAR and ESR on exceptions Edgar E. Iglesias
2014-09-13  4:29 ` [Qemu-devel] [PATCH v6 07/10] target-arm: A64: Emulate the HVC insn Edgar E. Iglesias
2014-09-17 21:47   ` Greg Bellows
2014-09-25 18:39   ` Peter Maydell
2014-09-25 22:20     ` Edgar E. Iglesias [this message]
2014-09-25 23:01       ` Peter Maydell
2014-09-25 23:06         ` Edgar E. Iglesias
2014-09-25 23:19           ` Peter Maydell
2014-09-13  4:29 ` [Qemu-devel] [PATCH v6 08/10] target-arm: A64: Emulate the SMC insn Edgar E. Iglesias
2014-09-17 22:43   ` Greg Bellows
2014-09-25 18:47   ` Peter Maydell
2014-09-25 22:55     ` Edgar E. Iglesias
2014-09-25 23:17       ` Peter Maydell
2014-09-25 23:31         ` Edgar E. Iglesias
2014-09-25 23:43           ` Peter Maydell
2014-09-25 23:45             ` Edgar E. Iglesias
2014-09-26  8:20             ` Edgar E. Iglesias
2014-09-13  4:29 ` [Qemu-devel] [PATCH v6 09/10] target-arm: Add IRQ and FIQ routing to EL2 and 3 Edgar E. Iglesias
2014-09-25 19:14   ` Peter Maydell
2014-09-13  4:29 ` [Qemu-devel] [PATCH v6 10/10] target-arm: Add support for VIRQ and VFIQ Edgar E. Iglesias
2014-09-25 19:36   ` Peter Maydell
2014-09-25 23:03     ` Edgar E. Iglesias

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20140925222057.GS16081@toto \
    --to=edgar.iglesias@gmail.com \
    --cc=aggelerf@ethz.ch \
    --cc=agraf@suse.de \
    --cc=alex.bennee@linaro.org \
    --cc=christoffer.dall@linaro.org \
    --cc=greg.bellows@linaro.org \
    --cc=pbonzini@redhat.com \
    --cc=peter.crosthwaite@xilinx.com \
    --cc=peter.maydell@linaro.org \
    --cc=qemu-devel@nongnu.org \
    --cc=rob.herring@linaro.org \
    --cc=rth@twiddle.net \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.