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

[Jens Freimann]

On Sun, Dec 16, 2012 at 04:30:21PM +0100, Andreas Färber wrote:
> 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. :)

Ok

> >      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?

Yes, together with the ipl device this fixme is resolved.

> >      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.)

Ok, I'll keep that in mind and send a followup patch.

Thanks!
Jens

> 
> >  }
> >  
> >  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
> 

-- 
Mit freundlichen Grüßen / Kind regards 

Jens Freimann 

-- 
IBM Linux Technology Center / Boeblingen lab
IBM Systems &Technology Group, Systems Software Development
-------------------------------------------------------------
IBM Deutschland
Schoenaicher Str 220
71032 Boeblingen
Phone: +49-7031-16 x1122
E-Mail: address@hidden
-------------------------------------------------------------
IBM Deutschland Research & Development GmbH
Vorsitzende des Aufsichtsrats: Martina Koederitz
Geschäftsführung: Dirk Wittkopp
Sitz der Gesellschaft: Böblingen
Registergericht: Amtsgericht Stuttgart, HRB 243294