I just did a bit of testing on the audio input of my Canon EOS70d. This may be useful to others, to save some trouble. An introduction: I'm retired from a career at Hewlett-Packard/Agilent Technologies, where for several years I designed audio analysis instruments. Comments below are based on a few measurements on an HP dynamic signal analyzer, and analysis of the recorded audio track from a video taken by my 70d.
First thing to note is that the 70d audio input is stereo; two independent channels are recorded. Seems to have at least 50dB channel separation...probably somewhat more. Second, the inputs are leveled by an AGC (automatic gain control} system that appears to apply the same gain to both channels, based on the greatest input levels. The AGC attack time (time to respond to a sudden increase) is short, in the 5-10 millisecond range. The AGC release rate appears to be a little more than 10dB per second. I don't know if it's linear (constant dB/second). What this means is that if you are recording speech, a pause will allow the background noise to come up, perhaps significantly.
What started my interest in using an external mic is that I've needed to make some recordings to be put on the web, of someone speaking. The mic in the camera didn't give us enough "presence," and I wanted to have a better quality soundtrack. My initial efforts involved using a separate audio recorder and a pro-quality microphone, placed close to the narrator. I found that editing the video to put the recorded audio in place of the camera's audio track was taking me longer than I wanted, so I looked into connecting a mic to the camera. We got an inexpensive camera-mount directional mic, and it worked better than the camera's built-in mic but the noise level was way higher than we wanted. At first I blamed the mic, but I've discovered that the camera's external audio input has quite a bit of noise (electronic, "white" noise roughly 50dB below full scale if the AGC isn't activated), not camera mechanical noise, but with autofocus that can be a problem too). Our inexpensive mic, at about 5 feet from the narrator, with its high-output ("+20dB") setting, wasn't putting out enough to activate the 70d's AGC, and the resulting video had weak audio with roughly a 20dB signal-to-noise ratio...not very good at all! Getting that mic (with it's short cord) closer to the narrator didn't look promising (physically), so I investigated using our good mics through a good mic preamplifier. At first, I was thinking "mic level input" and was a bit afraid to try higher (line level) input. What I've found is that signal level peaks about -20dBVrms are about right...kind of a low line level, but quite a bit more than a typical microphone without a preamp. The AGC starts to activate at about -50dBVrms input, so you have a fairly wide range of input levels that work well with the camera, though levels higher than typical un-amplified mic levels are best.
The biggest input I tested was 0dBVrms; the result was quite distorted but didn't do any damage that I can tell. Distortion noted from -15dBVrms on up. In a 600-ohm pro-audio system, 0dBVrms translates to +2.2dBm; you'd want to run a mixer output with peaks around -18dBm into the camera. And of course, all that assumes that all 70d's perform the same, and I have no way to know if that's true.
Main take-away is that a low line-level should be OK...do some of your own testing to see where you start to get distortion!
Happy videoing! Cheers, Tom