Patchwork [v2,16/21] q35: smbus: Remove PCI_STATUS_SIG_SYSTEM_ERROR and PCI_STATUS_DETECTED_PARITY from w1cmask

login
register
mail settings
Submitter Jason Baron
Date Oct. 9, 2012, 3:30 a.m.
Message ID <092c0acd50dc7d66ec0d46277e2fec207c41ffdf.1349749915.git.jbaron@redhat.com>
Download mbox | patch
Permalink /patch/190195/
State New
Headers show

Comments

Jason Baron - Oct. 9, 2012, 3:30 a.m.
From: Jan Kiszka <jan.kiszka@siemens.com>

Both bits are added to the write-1-to-clear mask by default. As the
smbus device does not allow writes at all, we have to remove it from
that mask, also to avoid triggering a runtime assertion.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Jason Baron <jbaron@redhat.com>
---
 hw/q35_smbus.c |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)
Paolo Bonzini - Oct. 9, 2012, 7:54 a.m.
Il 09/10/2012 05:30, Jason Baron ha scritto:
> From: Jan Kiszka <jan.kiszka@siemens.com>
> 
> Both bits are added to the write-1-to-clear mask by default. As the
> smbus device does not allow writes at all, we have to remove it from
> that mask, also to avoid triggering a runtime assertion.
> 
> Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
> Signed-off-by: Jason Baron <jbaron@redhat.com>
> ---
>  hw/q35_smbus.c |    4 ++++
>  1 files changed, 4 insertions(+), 0 deletions(-)
> 
> diff --git a/hw/q35_smbus.c b/hw/q35_smbus.c
> index 5efbe6c..dd64aa2 100644
> --- a/hw/q35_smbus.c
> +++ b/hw/q35_smbus.c
> @@ -97,6 +97,10 @@ static int ich9_smbus_initfn(PCIDevice *d)
>  
>      pci_set_word(d->wmask + PCI_STATUS,
>                      PCI_STATUS_SIG_SYSTEM_ERROR | PCI_STATUS_DETECTED_PARITY);
> +    pci_set_word(d->w1cmask + PCI_STATUS,
> +                  pci_get_word(d->w1cmask + PCI_STATUS) &
> +                  ~(PCI_STATUS_SIG_SYSTEM_ERROR | PCI_STATUS_DETECTED_PARITY));
> +
>      /* TODO? D31IP.SMIP in chipset configuration space */
>      pci_config_set_interrupt_pin(d->config, 0x01); /* interrupt pin 1 */
>  
> 

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Michael S. Tsirkin - Oct. 11, 2012, 2:53 p.m.
On Mon, Oct 08, 2012 at 11:30:36PM -0400, Jason Baron wrote:
> From: Jan Kiszka <jan.kiszka@siemens.com>
> 
> Both bits are added to the write-1-to-clear mask by default. As the
> smbus device does not allow writes at all, we have to remove it from
> that mask, also to avoid triggering a runtime assertion.
> 
> Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
> Signed-off-by: Jason Baron <jbaron@redhat.com>

Yes but a bugger question for me is why are these writeable
while spec says they are W1C?
Could you please add a code comment explaining that?

> ---
>  hw/q35_smbus.c |    4 ++++
>  1 files changed, 4 insertions(+), 0 deletions(-)
> 
> diff --git a/hw/q35_smbus.c b/hw/q35_smbus.c
> index 5efbe6c..dd64aa2 100644
> --- a/hw/q35_smbus.c
> +++ b/hw/q35_smbus.c
> @@ -97,6 +97,10 @@ static int ich9_smbus_initfn(PCIDevice *d)
>  
>      pci_set_word(d->wmask + PCI_STATUS,
>                      PCI_STATUS_SIG_SYSTEM_ERROR | PCI_STATUS_DETECTED_PARITY);
> +    pci_set_word(d->w1cmask + PCI_STATUS,
> +                  pci_get_word(d->w1cmask + PCI_STATUS) &
> +                  ~(PCI_STATUS_SIG_SYSTEM_ERROR | PCI_STATUS_DETECTED_PARITY));
> +
>      /* TODO? D31IP.SMIP in chipset configuration space */
>      pci_config_set_interrupt_pin(d->config, 0x01); /* interrupt pin 1 */
>  
> -- 
> 1.7.1
Jason Baron - Oct. 19, 2012, 3:13 p.m.
On Thu, Oct 11, 2012 at 04:53:24PM +0200, Michael S. Tsirkin wrote:
> On Mon, Oct 08, 2012 at 11:30:36PM -0400, Jason Baron wrote:
> > From: Jan Kiszka <jan.kiszka@siemens.com>
> > 
> > Both bits are added to the write-1-to-clear mask by default. As the
> > smbus device does not allow writes at all, we have to remove it from
> > that mask, also to avoid triggering a runtime assertion.
> > 
> > Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
> > Signed-off-by: Jason Baron <jbaron@redhat.com>
> 
> Yes but a bugger question for me is why are these writeable
> while spec says they are W1C?
> Could you please add a code comment explaining that?
> 

Spec has these bits as: 'R/WC'. So I think you are right. We should drop
the wmask setting. Also, I think the setting of w1cmask is inverted.
Shouldn't we be setting the bits in w1cmask, not clearing them?

Yamahata, any reason not to drop the setting of wmask here?

Thanks,

-Jason

> > ---
> >  hw/q35_smbus.c |    4 ++++
> >  1 files changed, 4 insertions(+), 0 deletions(-)
> > 
> > diff --git a/hw/q35_smbus.c b/hw/q35_smbus.c
> > index 5efbe6c..dd64aa2 100644
> > --- a/hw/q35_smbus.c
> > +++ b/hw/q35_smbus.c
> > @@ -97,6 +97,10 @@ static int ich9_smbus_initfn(PCIDevice *d)
> >  
> >      pci_set_word(d->wmask + PCI_STATUS,
> >                      PCI_STATUS_SIG_SYSTEM_ERROR | PCI_STATUS_DETECTED_PARITY);
> > +    pci_set_word(d->w1cmask + PCI_STATUS,
> > +                  pci_get_word(d->w1cmask + PCI_STATUS) &
> > +                  ~(PCI_STATUS_SIG_SYSTEM_ERROR | PCI_STATUS_DETECTED_PARITY));
> > +
> >      /* TODO? D31IP.SMIP in chipset configuration space */
> >      pci_config_set_interrupt_pin(d->config, 0x01); /* interrupt pin 1 */
> >  
> > -- 
> > 1.7.1
Isaku Yamahata - Oct. 19, 2012, 4:17 p.m.
On Fri, Oct 19, 2012 at 11:13:53AM -0400, Jason Baron wrote:
> On Thu, Oct 11, 2012 at 04:53:24PM +0200, Michael S. Tsirkin wrote:
> > On Mon, Oct 08, 2012 at 11:30:36PM -0400, Jason Baron wrote:
> > > From: Jan Kiszka <jan.kiszka@siemens.com>
> > > 
> > > Both bits are added to the write-1-to-clear mask by default. As the
> > > smbus device does not allow writes at all, we have to remove it from
> > > that mask, also to avoid triggering a runtime assertion.
> > > 
> > > Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
> > > Signed-off-by: Jason Baron <jbaron@redhat.com>
> > 
> > Yes but a bugger question for me is why are these writeable
> > while spec says they are W1C?
> > Could you please add a code comment explaining that?
> > 
> 
> Spec has these bits as: 'R/WC'. So I think you are right. We should drop
> the wmask setting. Also, I think the setting of w1cmask is inverted.
> Shouldn't we be setting the bits in w1cmask, not clearing them?
> 
> Yamahata, any reason not to drop the setting of wmask here?

No reason. Let's drop it.

Patch

diff --git a/hw/q35_smbus.c b/hw/q35_smbus.c
index 5efbe6c..dd64aa2 100644
--- a/hw/q35_smbus.c
+++ b/hw/q35_smbus.c
@@ -97,6 +97,10 @@  static int ich9_smbus_initfn(PCIDevice *d)
 
     pci_set_word(d->wmask + PCI_STATUS,
                     PCI_STATUS_SIG_SYSTEM_ERROR | PCI_STATUS_DETECTED_PARITY);
+    pci_set_word(d->w1cmask + PCI_STATUS,
+                  pci_get_word(d->w1cmask + PCI_STATUS) &
+                  ~(PCI_STATUS_SIG_SYSTEM_ERROR | PCI_STATUS_DETECTED_PARITY));
+
     /* TODO? D31IP.SMIP in chipset configuration space */
     pci_config_set_interrupt_pin(d->config, 0x01); /* interrupt pin 1 */