If you want to extract the DNA of the microrganisms living in the plant, you may simply use typical DNA extraction kit to extract all the DNA first. 

After doing NGS, you may use filter to filter out all the reads from plant genome. 

This can be done in several different ways. There is to my knowledge however not a specific kit to capture plant microorganisms so there would be quite extensive work with validating the methodology beforehand if you want to sequence only the bacteria and viruses.

If the aim is to only sequence bacteria and viruses filtration and centrifugation would enrich those parts of the metagenome. This can possibly be combined with rRNA depletion of the plant ribosomal RNA (unsure if there are kits available for this however).

MoBio (now owned by QiaGen) has a number of specialised kits for extraction of pure DNA/RNA from microbiome and could possibly be combined with an ultracentrifugation/filtration approach.

I would not recommend ultra deep sequencing followed by informatics filtration as a first resort since that solution normally goes by the paradigm with shit in à Shit out. The amount of contamination would most likely be staggering and what conclusions could be drawn would be severely hampered by the lack of control on the experiment.

Do you have any specific target microorganisms? If not, I believe it would be sufficient to simply extract the whole plant DNA with Phenol-Chloroform-CTAB method. Based on experience, I was able to extract DNA of asymptomatic oomycetes in my sample. From the DNA obtained, you may use universal primers for each group you want to sequence, then design a more specific primers for a nested PCR set up. 

In my case, I sent the samples for sequencing (the pair with the universal primer and with the specific primers). The universal primers is rather a hit-and-miss procedure, while the nested specific primers will give you a sequence of shorter length. If you want to obtain a whole marker for your samples, you may perform cloning experiments for your amplicons from derived from the universal primers.