Earl has a good point. Using the Kinetis to generate the mater clocks will add a lot of jitter on your audio and this impacts the audio quality.
If you're targeting an USB application with audio, a 2 oscillator approach will be the best. One for the MCU input to generate USB clock based on PLL (8 or 4 Mhz work pretty well) and the 12.288Mhz for the MCLK on the I2S.
The SAI module on Kinetis is capable to use just one BCLK and one LRCLK to feed Tx and RX. This is called synchronous mode.
You can take a look to the TWR-AUDIO-SGTL on for an example on how to connect the clocks. Here is the link to the schematics:
http://tinyurl.com/9hg7vra
In this case, the codec is the master, providing MCLK, BCLK and LRCLK. Here you'll find a sample code running on a TWR-K60D100M that might help you:
https://community.freescale.com/thread/81904
You also can feed MCLK to Kinetis and use the internal dividers to generate the I2S clocks, in this case, Kinetis will be the master and you avoid the excessive jitter becuase a dedicated oscillator is used as reference.