Figure below shows my bandpass filter's topology. Its Q factor can be adjusted by the two resistors R1 an R2. Resonant frequency is  sqrt(Gm1*Gm2/C1C2).  But after investigating the return ratio by breaking the loop at p, the phase margin is nearly 0 degree at moderate frequencies. Then i tried transient analysis with HSPICE,  it seems stable from the results.

  I am puzzled with this issue., and eager for any constructive suggestions .