RE: [PATCH v1 05/22] vfio/common: Extract out vfio_kvm_device_[add/del]_fd

[Duan, Zhenzhong]

>-----Original Message-----
>From: Eric Auger <eric.auger@redhat.com>
>Sent: Wednesday, September 20, 2023 7:49 PM
>Subject: Re: [PATCH v1 05/22] vfio/common: Extract out
>vfio_kvm_device_[add/del]_fd
>
>Hi Zhenzhong,
>
>On 8/30/23 12:37, Zhenzhong Duan wrote:
>> ...which will be used by both legacy and iommufd backend.
>I prefer genuine sentences in the commit msg. Also you explain what you
>do but not why.
>
>suggestion: Introduce two new helpers, vfio_kvm_device_[add/del]_fd
>which take as input a file descriptor which can be either a group fd or
>a cdev fd. This uses the new KVM_DEV_VFIO_FILE VFIO KVM device group,
>which aliases to the legacy KVM_DEV_VFIO_GROUP.
>
>vfio_kvm_device_add/del_group then call those new helpers.

Thanks, will update in v2.

>
>
>
>>
>> Signed-off-by: Yi Liu <yi.l.liu@intel.com>
>> Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
>> ---
>>  hw/vfio/common.c              | 44 +++++++++++++++++++++++------------
>>  include/hw/vfio/vfio-common.h |  3 +++
>>  2 files changed, 32 insertions(+), 15 deletions(-)
>>
>> diff --git a/hw/vfio/common.c b/hw/vfio/common.c
>> index 67150e4575..949ad6714a 100644
>> --- a/hw/vfio/common.c
>> +++ b/hw/vfio/common.c
>> @@ -1759,17 +1759,17 @@ void vfio_reset_handler(void *opaque)
>>      }
>>  }
>>
>> -static void vfio_kvm_device_add_group(VFIOGroup *group)
>> +int vfio_kvm_device_add_fd(int fd)
>>  {
>>  #ifdef CONFIG_KVM
>>      struct kvm_device_attr attr = {
>> -        .group = KVM_DEV_VFIO_GROUP,
>> -        .attr = KVM_DEV_VFIO_GROUP_ADD,
>> -        .addr = (uint64_t)(unsigned long)&group->fd,
>> +        .group = KVM_DEV_VFIO_FILE,
>> +        .attr = KVM_DEV_VFIO_FILE_ADD,
>> +        .addr = (uint64_t)(unsigned long)&fd,
>>      };
>>
>>      if (!kvm_enabled()) {
>> -        return;
>> +        return 0;
>>      }
>>
>>      if (vfio_kvm_device_fd < 0) {
>> @@ -1779,37 +1779,51 @@ static void
>vfio_kvm_device_add_group(VFIOGroup *group)
>>
>>          if (kvm_vm_ioctl(kvm_state, KVM_CREATE_DEVICE, &cd)) {
>>              error_report("Failed to create KVM VFIO device: %m");
>> -            return;
>> +            return -ENODEV;
>can't you return -errno?
Will fix.

>>          }
>>
>>          vfio_kvm_device_fd = cd.fd;
>>      }
>>
>>      if (ioctl(vfio_kvm_device_fd, KVM_SET_DEVICE_ATTR, &attr)) {
>> -        error_report("Failed to add group %d to KVM VFIO device: %m",
>> -                     group->groupid);
>> +        error_report("Failed to add fd %d to KVM VFIO device: %m",
>> +                     fd);
>> +        return -errno;
>>      }
>>  #endif
>> +    return 0;
>>  }
>>
>> -static void vfio_kvm_device_del_group(VFIOGroup *group)
>> +static void vfio_kvm_device_add_group(VFIOGroup *group)
>> +{
>> +    vfio_kvm_device_add_fd(group->fd);
>Since vfio_kvm_device_add_fd now returns an error value, it's a pity not
>to use it and propagate it. Also you could fill an errp with the error
>msg and use it in vfio_connect_container(). But this is a new error
>handling there.

What about having vfio_kvm_device_add_fd return void as
vfio_kvm_device_add_group. I just realize vfio_connect_container()
doesn't get any failure of vfio_kvm_device_add_group, propagating
err to vfio_connect_container() is just to print it out there which I have
done in vfio_kvm_device_add_fd.

>> +}
>> +
>> +int vfio_kvm_device_del_fd(int fd)
>not sure we want this to return an error. But if we do, I think it would
>be nicer to propagate the error up.

Same question as above.

>>  {
>>  #ifdef CONFIG_KVM
>>      struct kvm_device_attr attr = {
>> -        .group = KVM_DEV_VFIO_GROUP,
>> -        .attr = KVM_DEV_VFIO_GROUP_DEL,
>> -        .addr = (uint64_t)(unsigned long)&group->fd,
>> +        .group = KVM_DEV_VFIO_FILE,
>> +        .attr = KVM_DEV_VFIO_FILE_DEL,
>> +        .addr = (uint64_t)(unsigned long)&fd,
>>      };
>>
>>      if (vfio_kvm_device_fd < 0) {
>> -        return;
>> +        return -EINVAL;
>>      }
>>
>>      if (ioctl(vfio_kvm_device_fd, KVM_SET_DEVICE_ATTR, &attr)) {
>> -        error_report("Failed to remove group %d from KVM VFIO device: %m",
>> -                     group->groupid);
>> +        error_report("Failed to remove fd %d from KVM VFIO device: %m",
>> +                     fd);
>> +        return -EBADF;
>-errno?
Sure.

Thanks
Zhenzhong