qemu-devel
[Top][All Lists]
Advanced

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

Re: [Qemu-devel] [PATCH] ppc/xics: fix ICP reset path


From: Cédric Le Goater
Subject: Re: [Qemu-devel] [PATCH] ppc/xics: fix ICP reset path
Date: Mon, 23 Jul 2018 13:36:36 +0200
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101 Thunderbird/52.9.1

On 07/12/2018 12:01 PM, Greg Kurz wrote:
> Recent cleanup in commit a028dd423ee6 dropped the ICPStateClass::reset
> handler. It is now up to child ICP classes to call the DeviceClass::reset
> handler of the parent class, thanks to device_class_set_parent_reset().
> This is a better object programming pattern,

I think that the device_class_set_parent* routines just obfuscate a little
more the object programming approach of QEMU. 

> but unfortunately it causes QEMU to crash during CPU hotplug:

I guess I did not try that :/ Would it be complex to add a spapr unit test 
for CPU hotplug ? It would be useful.

> (qemu) device_add host-spapr-cpu-core,id=core1,core-id=1
> Segmentation fault (core dumped)
> 
> When the hotplug path tries to reset the ICP device, we end up calling:
> 
> static void icp_kvm_reset(DeviceState *dev)
> {
>     ICPStateClass *icpc = ICP_GET_CLASS(dev);
> 
>     icpc->parent_reset(dev);
> 
> but icpc->parent_reset is NULL... This happens because icp_kvm_class_init()
> calls:
> 
>     device_class_set_parent_reset(dc, icp_kvm_reset,
>                                   &icpc->parent_reset);
> 
> but dc->reset, ie, DeviceClass::reset for the TYPE_ICP type, is
> itself NULL.>
> This patch hence sets DeviceClass::reset for the TYPE_ICP type to
> point to icp_reset(). It then registers a reset handler that calls
> DeviceClass::reset. If the ICP subtype has configured its own reset
> handler with device_class_set_parent_reset(), this ensures it will
> be called first and it can then call ICPStateClass::parent_reset
> safely. This fixes the reset path for the TYPE_KVM_ICP type, which
> is the only subtype that defines its own reset function.

So, now, isn't ICP reset called twice on KVM ? 

Anyway, we can sort that out if necessary.

Thanks,

C.
 
> Reported-by: Satheesh Rajendran <address@hidden>
> Suggested-by: David Gibson <address@hidden>
> Fixes: a028dd423ee6dfd091a8c63028240832bf10f671
> Signed-off-by: Greg Kurz <address@hidden>
> ---
>  hw/intc/xics.c |   14 +++++++++++---
>  1 file changed, 11 insertions(+), 3 deletions(-)
> 
> diff --git a/hw/intc/xics.c b/hw/intc/xics.c
> index b9f1a3c97214..c90c893228dc 100644
> --- a/hw/intc/xics.c
> +++ b/hw/intc/xics.c
> @@ -291,7 +291,7 @@ static const VMStateDescription vmstate_icp_server = {
>      },
>  };
>  
> -static void icp_reset(void *dev)
> +static void icp_reset(DeviceState *dev)
>  {
>      ICPState *icp = ICP(dev);
>  
> @@ -303,6 +303,13 @@ static void icp_reset(void *dev)
>      qemu_set_irq(icp->output, 0);
>  }
>  
> +static void icp_reset_handler(void *dev)
> +{
> +    DeviceClass *dc = DEVICE_GET_CLASS(dev);
> +
> +    dc->reset(dev);
> +}
> +
>  static void icp_realize(DeviceState *dev, Error **errp)
>  {
>      ICPState *icp = ICP(dev);
> @@ -345,7 +352,7 @@ static void icp_realize(DeviceState *dev, Error **errp)
>          return;
>      }
>  
> -    qemu_register_reset(icp_reset, dev);
> +    qemu_register_reset(icp_reset_handler, dev);
>      vmstate_register(NULL, icp->cs->cpu_index, &vmstate_icp_server, icp);
>  }
>  
> @@ -354,7 +361,7 @@ static void icp_unrealize(DeviceState *dev, Error **errp)
>      ICPState *icp = ICP(dev);
>  
>      vmstate_unregister(NULL, &vmstate_icp_server, icp);
> -    qemu_unregister_reset(icp_reset, dev);
> +    qemu_unregister_reset(icp_reset_handler, dev);
>  }
>  
>  static void icp_class_init(ObjectClass *klass, void *data)
> @@ -363,6 +370,7 @@ static void icp_class_init(ObjectClass *klass, void *data)
>  
>      dc->realize = icp_realize;
>      dc->unrealize = icp_unrealize;
> +    dc->reset = icp_reset;
>  }
>  
>  static const TypeInfo icp_info = {
> 




reply via email to

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