large current through power mosfet in LTspice

I use some mosfet and BJT for a simple power path management in circuit. When Vprimary is larger than zener(indicate primary battery is healthy), Vprimary is connected to Vsys through M1 and M2 to supply power. When Vprimary falls below zener voltage, Q2 and M3 is biased so that Vbackup is connected to Vsys through M3 to supply power. The problem is that the current throught the three mosfet seems very large, why?

I think maybe LTspice's modeling about mosfet body diode not correct, however, use one mosfet from TI, the result is almost the same...

update

Thank you. The large current is caused by M1&M2 and M3 both conduct, shorting Vprimary and Vbackup. Followed by your advice, M1&M2 no longer conduct during M3 conducting. However, there's still momentary large current which is caused by Vpri->Vbak or Vbak->Vpri power source switching. I think it is caused by zener diode and the finite voltage source impedance. For example, during Vpri rising from 0 to 12V, the 9.1V zener diode will conduct when Vpri reachs 9.1V, causing M1&M2 conducts and shorting Vpri and Vbak through M1&M2&M3. Because of the 0.5ohm source impedance, the voltage on zener may below 9.1V. The large current will disapear once Vpri rises much larger than 9.1V. The duration may decrease if source impedance is decreased from 0.5 to 0.1ohm.

• You'll ned to concoct some sort of dead-time if you want to avoid the spikes. This is something all switching circuits need to deal with. If you plot the inverse of V(vprimary), you'll get two trapezoidal waveforms that overlap at the center, where the voltage is half of the input. Given that your MOSes are like a push-pull, driving them from a single source will always yield spikes at the turn on/off, and vice-versa. – a concerned citizen Sep 28 '17 at 7:16

As you can see following the red line, there is a current path from $V_{pri}$ to ground when $V_{pri}$ is greater than 2 diode drops. So, M1 & M2 are turned on almost all the time.
You can size R1 & R5 so the $V_{GS}$ is below M1's threshold voltage. A starting point is making R5 larger, like a 100k, and R1 smaller.