First of all DO'NT, "little boys that play with fire wet their bed". Second, if you have some morbid desire to screw around with proper design methodology, the method is:
in the low level interrupt at entry
disable all interrupts
do the necessary pushes
reenable the interrupts
in the high level interrupt:
1) did it interrupt the low level ?
if no, go to 3)
2) replace the return address from the high level interrupt with the address of the first POP before the RETI in the low lvel interrupt
3) continue the high level interrupt

CAVEATS:
some processors 'protect' the first instruction in an interrupt routine, if so good, if not a high level interrupt happening between entry of the low level int and the interrupt disable, would misfire. If the low level interrupt uses pushes and pops anywhere but entry and exit, it will misfire.

YOU END UP WITH UNMAINTAINEABLE CODE.
Eriks Law: "it does not matter if code works if it is maintainable"