On 02-03-21, 13:06, Jie Deng wrote: > Yeah. Actually, the backend only needs "struct virtio_i2c_out_hdr out_hdr" > and "struct virtio_i2c_in_hdr in_hdr" for communication. So we only need to > keep > the first two in uapi and move "struct virtio_i2c_req" into the driver. > > But Jason wanted to include "struct virtio_i2c_req" in uapi. He explained in > this link > https://lists.linuxfoundation.org/pipermail/virtualization/2020-October/050222.html. > Do you agree with that explanation ? I am not sure I understood his reasoning well, but it doesn't make any sense to keep this in uapi header if this is never going to get transferred over the wire. Moreover, the struct virtio_i2c_req in spec is misleading to me and rather creates unnecessary confusion. There is no structure like this which ever get passed here, but rather there are multiple vq transactions which take place, one with just the out header, then one with buffer and finally one with in header. I am not sure what's the right way of documenting it or if this is a standard virtio world follows. -- viresh
