I've used Figshare; it's free, gives you a DOI, and you can make a link from the data back to the article yourself.
What is presumably more useful is the ability to make a link from the article to your newly uploaded data. In general I do that by somehow citing the data's DOI within the article when submitting, but since your article has already been published you will need to contact the publisher, as they are the ones that must create that link.
Edit: just to expand on the latter:
I expect your article has a DOI - used in other papers when they cite it - which redirects to a webpage run by the publisher; that webpage is or has a link to the "Version of Record" of your article. It would be nice to also have a link to your source data. This second link has to be inserted by the publisher, since they control their webpage, but they first need to verify a couple of things:
is this the right data? Nobody ever accidentally clicks on the wrong file in the upload dialogue, of course... what if someone faked a screen name and then uploaded a version of the dataset maliciously modified so that the existing paper's analysis section now appears to be wrong or some form of misconduct?
is the data allowed to be shared? Have all the data's creators given permission? Study participants? Suitable licence?
Checking these at submission is straightforward: the Corresponding Author (submitter) provides the data (or link to it), and the Corresponding Author can sign off on any required permissions, ethical approval, etc. just as they do for the paper itself.
Making these checks after publication will be more problematic, and (personal opinion) I would be very wary of any process that allows this without direct communication with the original Corresponding Author.
(My concerns about validating data integrity are, of course, independent of which data-hosting platform is used!)