Texture and cudaMalloc3D?

Am I allowed to allocate memory using cudaMalloc3D, and then bind a texture to it?
If so, can I then access it using tex1Dfetch?
And finally, can I write to a location by passing in the pitchedPtr and extent, and writing as I normally would to memory allocated by cudaMalloc3D?
If so, am I correct in understanding that writes to texture memory complete at the end of a kernel invocation, meaning the next kernel invocation it will be available for use. (I don’t need to use the data I write to the texture bound memory in the same kernel call as the write)