From: Chao Gao <chao.gao@intel.com>

According to VT-d spec Interrupt Remapping and Interrupt Posting ->
Interrupt Remapping -> Interrupt Request Formats On Intel 64
Platforms, fields of MSI data register have changed. This patch
avoids wrongly regarding a remappable format interrupt request as
an interrupt binded with an event channel.

Signed-off-by: Chao Gao <chao.gao@intel.com>
Signed-off-by: Lan Tianyu <tianyu.lan@intel.com>
---
 hw/pci/msi.c         | 5 +++--
 hw/pci/msix.c        | 4 +++-
 hw/xen/xen_pt_msi.c  | 2 +-
 include/hw/xen/xen.h | 2 +-
 xen-hvm-stub.c       | 2 +-
 xen-hvm.c            | 7 ++++++-
 6 files changed, 15 insertions(+), 7 deletions(-)

diff --git a/hw/pci/msi.c b/hw/pci/msi.c
index a87b227..199cb47 100644
--- a/hw/pci/msi.c
+++ b/hw/pci/msi.c
@@ -289,7 +289,7 @@ void msi_reset(PCIDevice *dev)
 static bool msi_is_masked(const PCIDevice *dev, unsigned int vector)
 {
     uint16_t flags = pci_get_word(dev->config + msi_flags_off(dev));
-    uint32_t mask, data;
+    uint32_t mask, data, addr_lo;
     bool msi64bit = flags & PCI_MSI_FLAGS_64BIT;
     assert(vector < PCI_MSI_VECTORS_MAX);
 
@@ -298,7 +298,8 @@ static bool msi_is_masked(const PCIDevice *dev, unsigned int vector)
     }
 
     data = pci_get_word(dev->config + msi_data_off(dev, msi64bit));
-    if (xen_is_pirq_msi(data)) {
+    addr_lo = pci_get_long(dev->config + msi_address_lo_off(dev));
+    if (xen_is_pirq_msi(data, addr_lo)) {
         return false;
     }
 
diff --git a/hw/pci/msix.c b/hw/pci/msix.c
index bb54e8b..efe2982 100644
--- a/hw/pci/msix.c
+++ b/hw/pci/msix.c
@@ -82,9 +82,11 @@ static bool msix_vector_masked(PCIDevice *dev, unsigned int vector, bool fmask)
 {
     unsigned offset = vector * PCI_MSIX_ENTRY_SIZE;
     uint8_t *data = &dev->msix_table[offset + PCI_MSIX_ENTRY_DATA];
+    uint8_t *addr_lo = &dev->msix_table[offset + PCI_MSIX_ENTRY_LOWER_ADDR];
     /* MSIs on Xen can be remapped into pirqs. In those cases, masking
      * and unmasking go through the PV evtchn path. */
-    if (xen_enabled() && xen_is_pirq_msi(pci_get_long(data))) {
+    if (xen_enabled() && xen_is_pirq_msi(pci_get_long(data),
+                                         pci_get_long(addr_lo))) {
         return false;
     }
     return fmask || dev->msix_table[offset + PCI_MSIX_ENTRY_VECTOR_CTRL] &
diff --git a/hw/xen/xen_pt_msi.c b/hw/xen/xen_pt_msi.c
index 5fab95e..45a9e9f 100644
--- a/hw/xen/xen_pt_msi.c
+++ b/hw/xen/xen_pt_msi.c
@@ -114,7 +114,7 @@ static int msi_msix_setup(XenPCIPassthroughState *s,
 
     assert((!is_msix && msix_entry == 0) || is_msix);
 
-    if (xen_is_pirq_msi(data)) {
+    if (xen_is_pirq_msi(data, addr)) {
         *ppirq = msi_ext_dest_id(addr >> 32) | msi_dest_id(addr);
         if (!*ppirq) {
             /* this probably identifies an misconfiguration of the guest,
diff --git a/include/hw/xen/xen.h b/include/hw/xen/xen.h
index 09c2ce5..af759bc 100644
--- a/include/hw/xen/xen.h
+++ b/include/hw/xen/xen.h
@@ -33,7 +33,7 @@ int xen_pci_slot_get_pirq(PCIDevice *pci_dev, int irq_num);
 void xen_piix3_set_irq(void *opaque, int irq_num, int level);
 void xen_piix_pci_write_config_client(uint32_t address, uint32_t val, int len);
 void xen_hvm_inject_msi(uint64_t addr, uint32_t data);
-int xen_is_pirq_msi(uint32_t msi_data);
+int xen_is_pirq_msi(uint32_t msi_data, uint32_t msi_addr_lo);
 
 qemu_irq *xen_interrupt_controller_init(void);
 
diff --git a/xen-hvm-stub.c b/xen-hvm-stub.c
index c500325..dae421c 100644
--- a/xen-hvm-stub.c
+++ b/xen-hvm-stub.c
@@ -31,7 +31,7 @@ void xen_hvm_inject_msi(uint64_t addr, uint32_t data)
 {
 }
 
-int xen_is_pirq_msi(uint32_t msi_data)
+int xen_is_pirq_msi(uint32_t msi_data, uint32_t msi_addr_lo)
 {
     return 0;
 }
diff --git a/xen-hvm.c b/xen-hvm.c
index 5043beb..db29121 100644
--- a/xen-hvm.c
+++ b/xen-hvm.c
@@ -146,8 +146,13 @@ void xen_piix_pci_write_config_client(uint32_t address, uint32_t val, int len)
     }
 }
 
-int xen_is_pirq_msi(uint32_t msi_data)
+int xen_is_pirq_msi(uint32_t msi_data, uint32_t msi_addr_lo)
 {
+    /* If msi address is configurate to remapping format, the msi will not
+     * remapped into a pirq.
+     */
+    if (msi_addr_lo & MSI_ADDR_IF_MASK)
+        return 0;
     /* If vector is 0, the msi is remapped into a pirq, passed as
      * dest_id.
      */
-- 
1.8.3.1

On Thu, May 18, 2017 at 01:33:00AM -0400, Lan Tianyu wrote:
> From: Chao Gao <chao.gao@intel.com>
> 
> According to VT-d spec Interrupt Remapping and Interrupt Posting ->
> Interrupt Remapping -> Interrupt Request Formats On Intel 64
> Platforms, fields of MSI data register have changed. This patch
> avoids wrongly regarding a remappable format interrupt request as
> an interrupt binded with an event channel.
> 
> Signed-off-by: Chao Gao <chao.gao@intel.com>
> Signed-off-by: Lan Tianyu <tianyu.lan@intel.com>
> ---
>  hw/pci/msi.c         | 5 +++--
>  hw/pci/msix.c        | 4 +++-
>  hw/xen/xen_pt_msi.c  | 2 +-
>  include/hw/xen/xen.h | 2 +-
>  xen-hvm-stub.c       | 2 +-
>  xen-hvm.c            | 7 ++++++-
>  6 files changed, 15 insertions(+), 7 deletions(-)
> 
> diff --git a/hw/pci/msi.c b/hw/pci/msi.c
> index a87b227..199cb47 100644
> --- a/hw/pci/msi.c
> +++ b/hw/pci/msi.c
> @@ -289,7 +289,7 @@ void msi_reset(PCIDevice *dev)
>  static bool msi_is_masked(const PCIDevice *dev, unsigned int vector)
>  {
>      uint16_t flags = pci_get_word(dev->config + msi_flags_off(dev));
> -    uint32_t mask, data;
> +    uint32_t mask, data, addr_lo;
>      bool msi64bit = flags & PCI_MSI_FLAGS_64BIT;
>      assert(vector < PCI_MSI_VECTORS_MAX);
>  
> @@ -298,7 +298,8 @@ static bool msi_is_masked(const PCIDevice *dev, unsigned int vector)
>      }
>  
>      data = pci_get_word(dev->config + msi_data_off(dev, msi64bit));
> -    if (xen_is_pirq_msi(data)) {
> +    addr_lo = pci_get_long(dev->config + msi_address_lo_off(dev));
> +    if (xen_is_pirq_msi(data, addr_lo)) {
>          return false;
>      }
>  
> diff --git a/hw/pci/msix.c b/hw/pci/msix.c
> index bb54e8b..efe2982 100644
> --- a/hw/pci/msix.c
> +++ b/hw/pci/msix.c
> @@ -82,9 +82,11 @@ static bool msix_vector_masked(PCIDevice *dev, unsigned int vector, bool fmask)
>  {
>      unsigned offset = vector * PCI_MSIX_ENTRY_SIZE;
>      uint8_t *data = &dev->msix_table[offset + PCI_MSIX_ENTRY_DATA];
> +    uint8_t *addr_lo = &dev->msix_table[offset + PCI_MSIX_ENTRY_LOWER_ADDR];
>      /* MSIs on Xen can be remapped into pirqs. In those cases, masking
>       * and unmasking go through the PV evtchn path. */
> -    if (xen_enabled() && xen_is_pirq_msi(pci_get_long(data))) {
> +    if (xen_enabled() && xen_is_pirq_msi(pci_get_long(data),
> +                                         pci_get_long(addr_lo))) {
>          return false;
>      }
>      return fmask || dev->msix_table[offset + PCI_MSIX_ENTRY_VECTOR_CTRL] &
> diff --git a/hw/xen/xen_pt_msi.c b/hw/xen/xen_pt_msi.c
> index 5fab95e..45a9e9f 100644
> --- a/hw/xen/xen_pt_msi.c
> +++ b/hw/xen/xen_pt_msi.c
> @@ -114,7 +114,7 @@ static int msi_msix_setup(XenPCIPassthroughState *s,
>  
>      assert((!is_msix && msix_entry == 0) || is_msix);
>  
> -    if (xen_is_pirq_msi(data)) {
> +    if (xen_is_pirq_msi(data, addr)) {
>          *ppirq = msi_ext_dest_id(addr >> 32) | msi_dest_id(addr);
>          if (!*ppirq) {
>              /* this probably identifies an misconfiguration of the guest,
> diff --git a/include/hw/xen/xen.h b/include/hw/xen/xen.h
> index 09c2ce5..af759bc 100644
> --- a/include/hw/xen/xen.h
> +++ b/include/hw/xen/xen.h
> @@ -33,7 +33,7 @@ int xen_pci_slot_get_pirq(PCIDevice *pci_dev, int irq_num);
>  void xen_piix3_set_irq(void *opaque, int irq_num, int level);
>  void xen_piix_pci_write_config_client(uint32_t address, uint32_t val, int len);
>  void xen_hvm_inject_msi(uint64_t addr, uint32_t data);
> -int xen_is_pirq_msi(uint32_t msi_data);
> +int xen_is_pirq_msi(uint32_t msi_data, uint32_t msi_addr_lo);

Maybe inverting the arguments would be better, so the arguments would be
the address first, then the data, like I think it is often the case.
What do you think?

>  
>  qemu_irq *xen_interrupt_controller_init(void);
>  
> diff --git a/xen-hvm-stub.c b/xen-hvm-stub.c
> index c500325..dae421c 100644
> --- a/xen-hvm-stub.c
> +++ b/xen-hvm-stub.c
> @@ -31,7 +31,7 @@ void xen_hvm_inject_msi(uint64_t addr, uint32_t data)
>  {
>  }
>  
> -int xen_is_pirq_msi(uint32_t msi_data)
> +int xen_is_pirq_msi(uint32_t msi_data, uint32_t msi_addr_lo)
>  {
>      return 0;
>  }
> diff --git a/xen-hvm.c b/xen-hvm.c
> index 5043beb..db29121 100644
> --- a/xen-hvm.c
> +++ b/xen-hvm.c
> @@ -146,8 +146,13 @@ void xen_piix_pci_write_config_client(uint32_t address, uint32_t val, int len)
>      }
>  }
>  
> -int xen_is_pirq_msi(uint32_t msi_data)
> +int xen_is_pirq_msi(uint32_t msi_data, uint32_t msi_addr_lo)
>  {
> +    /* If msi address is configurate to remapping format, the msi will not
> +     * remapped into a pirq.

What do you think of: "If the MSI address is configured in remappable
format, the MSI will not be remapped into a pirq." ?

> +     */
> +    if (msi_addr_lo & MSI_ADDR_IF_MASK)
> +        return 0;
>      /* If vector is 0, the msi is remapped into a pirq, passed as
>       * dest_id.
>       */

Thanks,

-- 
Anthony PERARD

Hi Anthony:
	Thanks for your review.

On 2017年05月19日 19:57, Anthony PERARD wrote:
> On Thu, May 18, 2017 at 01:33:00AM -0400, Lan Tianyu wrote:
>> From: Chao Gao <chao.gao@intel.com>
>>
>> According to VT-d spec Interrupt Remapping and Interrupt Posting ->
>> Interrupt Remapping -> Interrupt Request Formats On Intel 64
>> Platforms, fields of MSI data register have changed. This patch
>> avoids wrongly regarding a remappable format interrupt request as
>> an interrupt binded with an event channel.
>>
>> Signed-off-by: Chao Gao <chao.gao@intel.com>
>> Signed-off-by: Lan Tianyu <tianyu.lan@intel.com>
>> ---
>>  hw/pci/msi.c         | 5 +++--
>>  hw/pci/msix.c        | 4 +++-
>>  hw/xen/xen_pt_msi.c  | 2 +-
>>  include/hw/xen/xen.h | 2 +-
>>  xen-hvm-stub.c       | 2 +-
>>  xen-hvm.c            | 7 ++++++-
>>  6 files changed, 15 insertions(+), 7 deletions(-)
>>
>> diff --git a/hw/pci/msi.c b/hw/pci/msi.c
>> index a87b227..199cb47 100644
>> --- a/hw/pci/msi.c
>> +++ b/hw/pci/msi.c
>> @@ -289,7 +289,7 @@ void msi_reset(PCIDevice *dev)
>>  static bool msi_is_masked(const PCIDevice *dev, unsigned int vector)
>>  {
>>      uint16_t flags = pci_get_word(dev->config + msi_flags_off(dev));
>> -    uint32_t mask, data;
>> +    uint32_t mask, data, addr_lo;
>>      bool msi64bit = flags & PCI_MSI_FLAGS_64BIT;
>>      assert(vector < PCI_MSI_VECTORS_MAX);
>>  
>> @@ -298,7 +298,8 @@ static bool msi_is_masked(const PCIDevice *dev, unsigned int vector)
>>      }
>>  
>>      data = pci_get_word(dev->config + msi_data_off(dev, msi64bit));
>> -    if (xen_is_pirq_msi(data)) {
>> +    addr_lo = pci_get_long(dev->config + msi_address_lo_off(dev));
>> +    if (xen_is_pirq_msi(data, addr_lo)) {
>>          return false;
>>      }
>>  
>> diff --git a/hw/pci/msix.c b/hw/pci/msix.c
>> index bb54e8b..efe2982 100644
>> --- a/hw/pci/msix.c
>> +++ b/hw/pci/msix.c
>> @@ -82,9 +82,11 @@ static bool msix_vector_masked(PCIDevice *dev, unsigned int vector, bool fmask)
>>  {
>>      unsigned offset = vector * PCI_MSIX_ENTRY_SIZE;
>>      uint8_t *data = &dev->msix_table[offset + PCI_MSIX_ENTRY_DATA];
>> +    uint8_t *addr_lo = &dev->msix_table[offset + PCI_MSIX_ENTRY_LOWER_ADDR];
>>      /* MSIs on Xen can be remapped into pirqs. In those cases, masking
>>       * and unmasking go through the PV evtchn path. */
>> -    if (xen_enabled() && xen_is_pirq_msi(pci_get_long(data))) {
>> +    if (xen_enabled() && xen_is_pirq_msi(pci_get_long(data),
>> +                                         pci_get_long(addr_lo))) {
>>          return false;
>>      }
>>      return fmask || dev->msix_table[offset + PCI_MSIX_ENTRY_VECTOR_CTRL] &
>> diff --git a/hw/xen/xen_pt_msi.c b/hw/xen/xen_pt_msi.c
>> index 5fab95e..45a9e9f 100644
>> --- a/hw/xen/xen_pt_msi.c
>> +++ b/hw/xen/xen_pt_msi.c
>> @@ -114,7 +114,7 @@ static int msi_msix_setup(XenPCIPassthroughState *s,
>>  
>>      assert((!is_msix && msix_entry == 0) || is_msix);
>>  
>> -    if (xen_is_pirq_msi(data)) {
>> +    if (xen_is_pirq_msi(data, addr)) {
>>          *ppirq = msi_ext_dest_id(addr >> 32) | msi_dest_id(addr);
>>          if (!*ppirq) {
>>              /* this probably identifies an misconfiguration of the guest,
>> diff --git a/include/hw/xen/xen.h b/include/hw/xen/xen.h
>> index 09c2ce5..af759bc 100644
>> --- a/include/hw/xen/xen.h
>> +++ b/include/hw/xen/xen.h
>> @@ -33,7 +33,7 @@ int xen_pci_slot_get_pirq(PCIDevice *pci_dev, int irq_num);
>>  void xen_piix3_set_irq(void *opaque, int irq_num, int level);
>>  void xen_piix_pci_write_config_client(uint32_t address, uint32_t val, int len);
>>  void xen_hvm_inject_msi(uint64_t addr, uint32_t data);
>> -int xen_is_pirq_msi(uint32_t msi_data);
>> +int xen_is_pirq_msi(uint32_t msi_data, uint32_t msi_addr_lo);
> 
> Maybe inverting the arguments would be better, so the arguments would be
> the address first, then the data, like I think it is often the case.
> What do you think?


That make sense. Will update.

> 
>>  
>>  qemu_irq *xen_interrupt_controller_init(void);
>>  
>> diff --git a/xen-hvm-stub.c b/xen-hvm-stub.c
>> index c500325..dae421c 100644
>> --- a/xen-hvm-stub.c
>> +++ b/xen-hvm-stub.c
>> @@ -31,7 +31,7 @@ void xen_hvm_inject_msi(uint64_t addr, uint32_t data)
>>  {
>>  }
>>  
>> -int xen_is_pirq_msi(uint32_t msi_data)
>> +int xen_is_pirq_msi(uint32_t msi_data, uint32_t msi_addr_lo)
>>  {
>>      return 0;
>>  }
>> diff --git a/xen-hvm.c b/xen-hvm.c
>> index 5043beb..db29121 100644
>> --- a/xen-hvm.c
>> +++ b/xen-hvm.c
>> @@ -146,8 +146,13 @@ void xen_piix_pci_write_config_client(uint32_t address, uint32_t val, int len)
>>      }
>>  }
>>  
>> -int xen_is_pirq_msi(uint32_t msi_data)
>> +int xen_is_pirq_msi(uint32_t msi_data, uint32_t msi_addr_lo)
>>  {
>> +    /* If msi address is configurate to remapping format, the msi will not
>> +     * remapped into a pirq.
> 
> What do you think of: "If the MSI address is configured in remappable
> format, the MSI will not be remapped into a pirq." ?

Will update.

> 
>> +     */
>> +    if (msi_addr_lo & MSI_ADDR_IF_MASK)
>> +        return 0;
>>      /* If vector is 0, the msi is remapped into a pirq, passed as
>>       * dest_id.
>>       */
> 
> Thanks,
> 


-- 
Best regards
Tianyu Lan