Your basic concept makes sense, but you missed the fact that the "12 V" line of a car will sometimes have significant voltage spikes on it. Any cicuit connected directly to this power needs to be able to withstand 50 at least for short periods.
When the pump is on, even a short spike will apply high voltage to the FET gate, which will blow out the oxide instantly. Overvoltaging the FET S-D isn't good either.
Use a 60 V FET, and do something to clamp the gate voltage to a safe level.
Added:
I forgot to mention this earlier since the main issue was about nasty voltage spikes on the vehicle power line. No, 1N4007 is a bad choice for the diodes. In this case, I'd use Schottkys rated for 20 or 30 V. Those are cheap and readily available. Since the current will only run in them for a short time as the stored inductive energy is dissipated, you can use the peak current rating instead of the continuous current rating of the diodes. The diodes need to be able to handle peak current of whatever the motor current is.
We can make this work well for you.
But attention must be given to the changes required.
You are trying to use the transistor in an unnecessary mode in order to attempt to eliminate a very common design requirement which simply needs to be faced.
ie you do not "need" a higher Vgs FEt - you just need to drive an otherwise suitable FET correctly by limiting Vgsmax to a voltage which is enough to drive the FET correctly worst case -but still less or much less than Vgs_absmax.
In addition, you need to specify switching speed requirements and address them explicitly as part of your design.
You ask (in a comment) "If we are using the PMOS for fast switching applications (~10μs),..." -> if you are using timings of the order of 10 uS (and you do not give any indication of switching speeds in your question) then the passive 30k R1 resistor is going to utterly mangle your switching waveforms. Without looking at the datasheet you can use a rule of thumb gate capacitance of 1 nF. That with 30k Rgs for turn off gives a time constant of 30 microseconds. Zener response times are far from your greatest problem.
Vgsmax of the MOSFET only needs to be safely greater than the maximum Vgs you need for full enhancement of the MOSFET under the worst case conditions of interest.
Very few MOSFETS need more than 12V Vgs to be driven as hard as is possible, so a Vgsmax rating of 20V is very acceptable.
By making R1 = R2 you reduce Vgsmax to about 12.5V.
Looks at data sheet ...
VGs of 10V (actually -10V) is the maximum you need to fully drive the FET worst case.
Gate capacitance appears to be comfortable under 1 nF.
Better - use the gate charge curve at top right of page 6 to see what gate current you will need to remove the gate charge at the Vgs you wish to work at.
Note that they only specify Vgs up to 10V on this graph, and that gate charge is substantially higher at high Vgs levels.
"At a guesstimate", using 1k for R1 and R2 would probably achieve switching that is in the order of fast enough, and Vgsmax can probably be less than 10V allowing even smaller R1 - so lower RC time constant and less charge to remove as well.
It's not a marvellous GET - Rdson is 125 milliOhm typical at 10V Vgs, ~= 4A, 25 C. That mans you would get about 4 x 125 = 500 milliOhms Rds at 4A. That may be acceptable, but much better is possible at modest cost.
To design this driver possible we need to know:
Maximum switched current.
Switch timing waveforms - on or off for how long?, what rise & fall times needed and why?
What is the load? (resistive, inductive, heated filament, ...?)
Anything else important that will change the question when we know about it.
Note that FET gate voltages want to be kept well away from Vgs_abs_max. The gate to channel interface is an oxide layer whose thickness is measurable in "atoms thick" and can be broken down by a whiff of overvoltage if unprotected. Driving with say 10V abs max and then providing a say 12V or 15V reverse biased zener connected gate to source and physically close to the FET. With a load with any inductance, adding this zener may transform reliability. This is because Millar coupling from Drain to gate can otherwise wreak overvoltage havoc on the gate oxide. (Ask me how I know :-) ).
Best Answer
Yes, you will exceed the absolute maximum gate voltage rating if you do that. The MOSFET may not die until 30V or 50V- on the other hand you should stay well away from the absolute maximum ratings for reliability and to consider any transients that may occur on the supply line.
You may be able to add a resistor and a zener diode to limit the gate voltage, for example you might use a 10V Zener. That will tend to slow the switching, which may or may not be an issue depending on what you are doing.
For example:
simulate this circuit – Schematic created using CircuitLab
Edit: Okay, you've added some stuff about 100kHz distinguishing this from a simple static switching situation.
You can hang a driver off the 30V supply, with a regulated (negative regulator) voltage if you like. Here is a very simple driver that does level shifting. You can keep the Zener diode in there on the gate to prevent any possibility of exceeding the maximum.
Or just use a negative regulator and a gate driver chip with a level shifter (or fast logic-output optoisolator) to the input. One advantage most gate drivers offer (along with fast, high current drive) is that they include some type of undervoltage lockout, which could save your MOSFET in brownout conditions.