Re: [qemu-s390x] [PATCH v1 for-2-12 06/15] s390x/flic: factor out injection of floating interrupts

[Christian Borntraeger]

On 12/12/2017 02:49 PM, Cornelia Huck wrote:
> On Mon, 11 Dec 2017 14:47:31 +0100
> David Hildenbrand <address@hidden> wrote:
> 
>> Let the flic device handle it internally. This will allow us to later
>> on store floating interrupts in the flic for the TCG case.
>>
>> This now also simplifies kvm.c. All that's left is the fallback
>> interface for floating interrupts, which is no triggered directly via
> 
> s/no/now/
> 
>> the flic in case anything goes wrong.
>>
>> Signed-off-by: David Hildenbrand <address@hidden>
>> ---
>>  hw/intc/s390_flic.c          | 31 ++++++++++++++++++++
>>  hw/intc/s390_flic_kvm.c      | 63 ++++++++++++++++++++++++++++++++++++----
>>  include/hw/s390x/s390_flic.h |  5 ++++
>>  target/s390x/cpu.h           |  7 ++++-
>>  target/s390x/interrupt.c     | 42 +++++++++++----------------
>>  target/s390x/kvm-stub.c      | 13 ---------
>>  target/s390x/kvm.c           | 68 
>> ++++----------------------------------------
>>  target/s390x/kvm_s390x.h     | 10 +------
>>  8 files changed, 123 insertions(+), 116 deletions(-)
>>
>> diff --git a/hw/intc/s390_flic.c b/hw/intc/s390_flic.c
>> index a78bdf1d90..8d521c415a 100644
>> --- a/hw/intc/s390_flic.c
>> +++ b/hw/intc/s390_flic.c
>> @@ -131,6 +131,34 @@ static int qemu_s390_inject_airq(S390FLICState *fs, 
>> uint8_t type,
>>      return 0;
>>  }
>>  
>> +static void qemu_s390_inject_service(S390FLICState *fs, uint32_t parm)
>> +{
>> +
>> +    S390CPU *dummy_cpu = s390_cpu_addr2state(0);
>> +
>> +    /* FIXME: don't inject into dummy CPU */
>> +    cpu_inject_service(dummy_cpu, parm);
>> +}
>> +
>> +static void qemu_s390_inject_io(S390FLICState *fs, uint16_t subchannel_id,
>> +                                uint16_t subchannel_nr, uint32_t 
>> io_int_parm,
>> +                                uint32_t io_int_word)
>> +{
>> +    S390CPU *dummy_cpu = s390_cpu_addr2state(0);
>> +
>> +    /* FIXME: don't inject into dummy CPU */
>> +    cpu_inject_io(dummy_cpu, subchannel_id, subchannel_nr, io_int_parm,
>> +                  io_int_word);
>> +}
>> +
>> +static void qemu_s390_inject_crw_mchk(S390FLICState *fs)
>> +{
>> +    S390CPU *dummy_cpu = s390_cpu_addr2state(0);
>> +
>> +    /* FIXME: don't inject into dummy CPU */
>> +    cpu_inject_crw_mchk(dummy_cpu);
>> +}
>> +
>>  static void qemu_s390_flic_reset(DeviceState *dev)
>>  {
>>      QEMUS390FLICState *flic = QEMU_S390_FLIC(dev);
>> @@ -172,6 +200,9 @@ static void qemu_s390_flic_class_init(ObjectClass *oc, 
>> void *data)
>>      fsc->clear_io_irq = qemu_s390_clear_io_flic;
>>      fsc->modify_ais_mode = qemu_s390_modify_ais_mode;
>>      fsc->inject_airq = qemu_s390_inject_airq;
>> +    fsc->inject_service = qemu_s390_inject_service;
>> +    fsc->inject_io = qemu_s390_inject_io;
>> +    fsc->inject_crw_mchk = qemu_s390_inject_crw_mchk;
>>  }
> 
> As you now have a callback for ->inject_io(), make
> qemu_s390_inject_airq() invoke it directly instead of going the detour
> through s390_io_interrupt()?
> 
>>  
>>  static Property s390_flic_common_properties[] = {
> 
> (...)
> 
> Generally looks sane.
> 
> One thing I noticed: You removed the caching of the flic (in the old
> kvm inject routine), and you generally do more qom invocations (first,
> to find the common flic; then, to translate to the qemu or kvm flic).
> Not sure if this might be a problem (probably not).

Is any of these calls on a potential fast path (e.g. guest without adapter
interrupts)? If yes, then QOM is a no-go since it is really slow.