There have been plans to return samples from the surface of Mars for about the past 46 years. There was a project initiated in 1998 to return samples by 2008, but it was cancelled in 2000 due to other Mars mission failures in 1999.
Currently the first part of a Mars Sample Return (MSR) mission is in serious planning, which is a sample caching missing to be launched in 2020. It is not an approved mission yet, but it is well on its way. That first step would collect rock cores and other samples and store them in a can, called the cache, on the rover for later retrieval. This is the closest we've gotten to a "real" MSR project since 2000.
The 2020 mission could be followed by another lander, this time with a rocket to launch the samples into Mars orbit. Then a return orbiter could pick up the samples in Mars orbit, bring them to Earth, and send an entry vehicle in with the samples to land. Those missions being further in the future are in much more nascent stages of planning.
Returning samples from Mars, while difficult and complicated, is certainly possible. One thing that has been relatively consistent over the years is that the sample collected and returned would be on the order of 500 grams of material.
One of the requirements on MSR is to not inadvertently release Martian samples into Earth's environment on entry or landing. The samples have to go through an isolated receiving facility on Earth first and be investigated there before they can be released into Earth's environment for more open use in other laboratories. As a result, the safest approach appears to be direct entry and landing, since that can be done with the simplest, most robust vehicle with large margins. It doesn't even need a parachute. Going into Earth orbit, being retrieved by another vehicle, and then safely entering and landing that vehicle adds more failure modes for inadvertent entry and breakup.
The whole point is to get the samples into Earth laboratories, so sooner or later they have to enter and land.