There are two issues with input voltage ripple: The input voltage being too high or low for correct operation, and high frequency noise on the input.
The first is pretty straight forward. You make sure that the nominal voltage ± the peak ripple is never above or below the guaranteed operating region of the SMPS. For example, if the SMPS is specified to work with 10 to 30 V input, then 20 V ±1 V is fine. 11 V ±1 V would be right on the edge.
Quickly changing input voltage is harder to know whether it is acceptable or not. The usual solution is to attenuate high frequency input with a big enough ceramic (low ESR) cap immediately on the input of the SMPS. If necessary, add an inductor in series before the cap and SMPS.
You definitely don't want significant input ripple anywhere near half the switching frequency or higher. That would almost guarantee instability. You can get some idea where input frequencies start to cause trouble by looking at the transient response time of the SMPS. That tells you effectively how fast the controller can respond. You don't want significant input voltage variations in that time. Again, the usual approach is to simply attenuate such input frequencies.
If the SMPS is being fed by an ideal constant voltage source, adding an inductor would not have any value?
Right. An ideal voltage source, is, well, the ideal.
Is there other EMI aspect regarding having bigger input voltage ripple?
Some input ripple will always make it to the output. That's what the line regulation spec is all about.