Gap between two consecutive order statistics under arbitrary distribution.

Question

Consider an arbitrary distribution $\mathcal{D}$ supported on $[a,b]$ with density function $\phi(x)\in[\gamma, \Gamma]$ where $\Gamma\geq \gamma>0$. M i.i.d samples $\{d_i\}_{i=1}^M$ are drawn from $\mathcal{D}$. Denote $\{d_{k_j}\}_{j=1}^M$ be the ordered sequence: $d_{k_1}\leq d_{k_2}\leq\ldots\leq d_{k_M}$. I wonder whether we have some gap bound on $\max_{j\in [M]}|d_{k_j}-d_{k_{j-1}}|$? Specifically, I wonder what is the bound on $Pr(\max_{j\in [M]}|d_{k_j}-d_{k_{j-1}}|\geq \epsilon)$. Thanks!