Using a copy from a VkBuffer instead of from a linear texture is definitely the right track. Trying to use a linear texture for staging is an exercise in frustration because you run into the irritating limitations that apply to linear textures (like limited format/mipmap support).

You shouldn't have to worry about implementation specifics though. Section 18.4 contains pseudocode describing how it expects your texture data to be laid out in your VkBuffer prior to a call to vkCmdCopyBufferToImage.

Assuming you're not using a block compressed format, the spec says:

rowLength = region->bufferRowLength;
if (rowLength == 0)
    rowLength = region->imageExtent.width;

imageHeight = region->bufferImageHeight;
if (imageHeight == 0)
    imageHeight = region->imageExtent.height;

elementSize = <element size of the format of the src/dstImage>;

address of (x,y,z) = region->bufferOffset + (((z * imageHeight) + y) * rowLength + x) * elementSize;

where x,y,z range from (0,0,0) to region->imageExtent.{width,height,depth}.

There is no need to worry about implementation specifics for staging texture data, that's the driver's problem/responsibility.