qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] [PATCH 2/3] s390: Add CPU reset handler


From: Andreas Färber
Subject: Re: [Qemu-devel] [PATCH 2/3] s390: Add CPU reset handler
Date: Sun, 16 Dec 2012 16:30:21 +0100
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:17.0) Gecko/17.0 Thunderbird/17.0

Am 14.12.2012 17:46, schrieb Jens Freimann:
> Add a CPU reset handler to have all CPUs in a PoP compliant
> state.
> 
> Signed-off-by: Jens Freimann <address@hidden>

The logic looks okay now. Some comments inline.

> ---
> v2 -> v3: 
> * explain in comment which code sets cpu 0 to running during IPL
> 
> v1 -> v2:
> * move setting of control registers and psa to s390_cpu_reset
>   and call it from the new s390_machine_cpu_reset_cb()
>   This makes it more similar to how it is done on x86
> * in s390_cpu_reset() set env->halted state of cpu after
>   the memset. This is needed to keep our s390_cpu_running
>   counter in sync when s390_cpu_reset is called via the
>   qemu_devices_reset path
> * set env->halted state in s390_cpu_initfn to 1 to avoid
>   decrementing the cpu counter during first reset
> ---
>  target-s390x/cpu.c | 32 ++++++++++++++++++++++++++++++--
>  target-s390x/kvm.c |  9 ++++++++-
>  2 files changed, 38 insertions(+), 3 deletions(-)
> 
> diff --git a/target-s390x/cpu.c b/target-s390x/cpu.c
> index 619b202..75d4036 100644
> --- a/target-s390x/cpu.c
> +++ b/target-s390x/cpu.c
> @@ -4,6 +4,7 @@
>   * Copyright (c) 2009 Ulrich Hecht
>   * Copyright (c) 2011 Alexander Graf
>   * Copyright (c) 2012 SUSE LINUX Products GmbH
> + * Copyright (c) 2012 IBM Corp.
>   *
>   * This library is free software; you can redistribute it and/or
>   * modify it under the terms of the GNU Lesser General Public
> @@ -18,9 +19,13 @@
>   * You should have received a copy of the GNU Lesser General Public
>   * License along with this library; if not, see
>   * <http://www.gnu.org/licenses/lgpl-2.1.html>
> + * Contributions after 2012-12-11 are licensed under the terms of the
> + * GNU GPL, version 2 or (at your option) any later version.
> + *
>   */
>  
>  #include "cpu.h"
> +#include "hw/hw.h"
>  #include "qemu-common.h"
>  #include "qemu-timer.h"
>  
> @@ -37,12 +42,30 @@ static void s390_cpu_reset(CPUState *s)
>          log_cpu_state(env, 0);
>      }
>  
> -    scc->parent_reset(s);
> +    s390_del_running_cpu(env);
>  
> +    scc->parent_reset(s);

If this gets respun, a white line separating the parent reset from the
local reset would be nice. :)

>      memset(env, 0, offsetof(CPUS390XState, breakpoints));
> +
> +    /* architectured initial values for CR 0 and 14 */
> +    env->cregs[0] = 0xE0UL;
> +    env->cregs[14] = 0xC2000000UL;
> +    /* set to z/Architecture mode */
> +    env->psw.mask = 0x0000000180000000ULL;
> +    env->psa = 0;
> +    /* set halted to 1 to make sure we can add the cpu in
> +     * s390_ipl_cpu code, where env->halted is set back to 0
> +     * after incrementing the cpu counter */
> +    env->halted = 1;
>      /* FIXME: reset vector? */

Do the above added cregs/psw/psa reset values resolve this FIXME? Or
does that refer to something different?

>      tlb_flush(env, 1);
> -    s390_add_running_cpu(env);
> +}
> +
> +static void s390_cpu_machine_reset_cb(void *opaque)
> +{
> +    S390CPU *cpu = opaque;
> +
> +    cpu_reset(CPU(cpu));
>  }
>  
>  static void s390_cpu_initfn(Object *obj)
> @@ -66,7 +89,12 @@ static void s390_cpu_initfn(Object *obj)
>      env->cpu_num = cpu_num++;
>      env->ext_index = -1;
>  
> +    /* set env->halted state to 1 to avoid decrementing the running
> +     * cpu counter in s390_cpu_reset to a negative number at 
> +     * initial ipl */
> +    env->halted = 1;
>      cpu_reset(CPU(cpu));
> +    qemu_register_reset(s390_cpu_machine_reset_cb, cpu);

Since we register the reset handler in instance_init, we should
unregister it in a instance_finalize callback (uninitfn?). Since we do
not hot-unplug s390 CPUs yet to my knowledge, that could be done in a
follow-up.

(For x86 it it registered in the provisional realize function and not
unregistered lacking a matching unrealization mechanism today; elsewhere
reset registration is done in the machine.)

>  }
>  
>  static void s390_cpu_class_init(ObjectClass *oc, void *data)
> diff --git a/target-s390x/kvm.c b/target-s390x/kvm.c
> index 94de764..fda9f1f 100644
> --- a/target-s390x/kvm.c
> +++ b/target-s390x/kvm.c
> @@ -85,7 +85,14 @@ int kvm_arch_init_vcpu(CPUS390XState *env)
>  
>  void kvm_arch_reset_vcpu(CPUS390XState *env)
>  {

Note to Alex: In my upcoming KVM CPUState series, this argument type
changes to CPUState ...

> -    /* FIXME: add code to reset vcpu. */
> +   /* The initial reset call is needed here to reset in-kernel
> +    * vcpu data that we can't access directly from QEMU
> +    * (i.e. with older kernels which don't support sync_regs/ONE_REG).
> +    * Before this ioctl cpu_synchronize_state() is called in common kvm
> +    * code (kvm-all) */
> +    if (kvm_vcpu_ioctl(env, KVM_S390_INITIAL_RESET, NULL)) {

... but so does the argument for kvm_vcpu_ioctl(), so merging becomes a
trivial env -> cpu change.

> +        perror("Can't reset vcpu\n");
> +    }
>  }
>  
>  int kvm_arch_put_registers(CPUS390XState *env, int level)

Andreas

-- 
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg



reply via email to

[Prev in Thread] Current Thread [Next in Thread]