_set_sqe_cmd_op: Simplify the code - direct variable access

[bsbernd]

Members of the union can be accessed directly using the other
union member off is not needed here.

[ming1]

When I write ublksrv, the header isn't updated yet, even though SQE128 is supported. It is just fine to use the current way, at least it still can build successfully against old header.

But your patch is fine, and I will apply it after automake patches are merged,
and liburing2.2 is required explicitly by that patchset.

[bsbernd]

Sure! I didn't know about the history and didn't check liburing versions. I'm just going through the code to understand how it works and how I can use it for fuse.

[ming1]

BTW, just curious how you use it for fuse? I understand fuse is userspace FS, and look it needn't one block device.

[bsbernd]

I'm not too much interested in the block device part right now, I was just looking how the IORING_OP_URING_CMD works. It is an interesting idea to send an SQE from userspace and to let it wait until the kernel needs it - I guess I can do exactly the same for fuse.
I'm also interested in zero-copy, but looks like this is not implemented in ubd yet. Existing splice in fuse doesn't work perfectly for us (network file system) - it just ends up as a bit faster memory copy.