Re: [PATCH v3 2/3] pci-assign: Use PCI-2.3-based shared legacy interrupts

[Jan Kiszka <jan.kiszka@siemens.com>]

On 2012-03-07 19:42, Jan Kiszka wrote:
> On 2012-03-07 19:32, Alex Williamson wrote:
>> On Wed, 2012-03-07 at 19:18 +0100, Jan Kiszka wrote:
>>> Enable the new KVM feature that allows legacy interrupt sharing for
>>> PCI-2.3-compliant devices. This requires to synchronize any guest
>>> change of the INTx mask bit to the kernel.
>>>
>>> The feature is controlled by the property 'share_intx' and is off by
>>> default for now.
>>>
>>> Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
>>> ---
>>>
>>> Changes in v3:
>>>  - fixed nit :)
>>>
>>>  hw/device-assignment.c |   24 ++++++++++++++++++++++++
>>>  hw/device-assignment.h |   10 ++++++----
>>>  qemu-kvm.c             |    9 +++++++++
>>>  qemu-kvm.h             |    2 ++
>>>  4 files changed, 41 insertions(+), 4 deletions(-)
>>>
>>> diff --git a/hw/device-assignment.c b/hw/device-assignment.c
>>> index a5f1abb..e2a8479 100644
>>> --- a/hw/device-assignment.c
>>> +++ b/hw/device-assignment.c
>>> @@ -782,6 +782,13 @@ static int assign_device(AssignedDevice *dev)
>>>                  "cause host memory corruption if the device issues DMA write "
>>>                  "requests!\n");
>>>      }
>>> +    if (dev->features & ASSIGNED_DEVICE_SHARE_INTX_MASK) {
>>> +        assigned_dev_data.flags |= KVM_DEV_ASSIGN_PCI_2_3;
>>> +
>>> +        /* hide host-side INTx masking from the guest */
>>> +        dev->emulate_config_read[PCI_COMMAND + 1] |=
>>> +            PCI_COMMAND_INTX_DISABLE >> 8;
>>> +    }
>>>  
>>>      r = kvm_assign_pci_device(kvm_state, &assigned_dev_data);
>>>      if (r < 0) {
>>> @@ -1121,10 +1128,25 @@ static void assigned_dev_pci_write_config(PCIDevice *pci_dev, uint32_t address,
>>>                                            uint32_t val, int len)
>>>  {
>>>      AssignedDevice *assigned_dev = DO_UPCAST(AssignedDevice, dev, pci_dev);
>>> +    uint16_t old_cmd = pci_get_word(pci_dev->config + PCI_COMMAND);
>>>      uint32_t emulate_mask, full_emulation_mask;
>>> +    int ret;
>>>  
>>>      pci_default_write_config(pci_dev, address, val, len);
>>>  
>>> +    if (range_covers_byte(address, len, PCI_COMMAND + 1)) {
>>
>> Hmm, now that I've acked this... shouldn't we have a feature check here
>> to avoid the ioctl when share_intx=false or when the host kernel doesn't
>> support this?  Thanks,
> 
> Hmm, I thought I handled this, somehow. But it's not there. v4 follows...

To be precise: We only need to worry about false positive error messages
in the absence of KVM_CAP_PCI_2_3. Devices not supporting INTx sharing
will either refuse to work due to IRQ conflicts, or we will map INTx
masking on IRQ line masking.

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux