We make a detailed analysis of some charmless self-tagging B decays, e.g., ${\mathit{B}}_{\mathit{u}}^{+}$\ensuremath{\rightarrow}${\mathit{K}}^{+}$K${\mathrm{\ifmmode\bar\else\textasciimacron\fi{}}}^{0}$, ${\mathit{K}}^{\mathrm{*}+}$${\mathrm{\ensuremath{\pi}}}^{0}$, and ${\mathit{B}}_{\mathit{d}}^{0}$\ensuremath{\rightarrow}${\mathit{K}}^{+}$${\mathrm{\ensuremath{\pi}}}^{\mathrm{\ensuremath{-}}}$, which are of great experimental interest for probing CP violation in the decay amplitude. A few neutral B decays into CP eigenstates such as ${\mathit{B}}_{\mathit{d}}^{0}$\ensuremath{\rightarrow}${\mathrm{\ensuremath{\pi}}}^{+}$${\mathrm{\ensuremath{\pi}}}^{\mathrm{\ensuremath{-}}}$ are also studied on the \ensuremath{\Upsilon}(4S) resonance to distinguish CP violation induced by penguin loops from that via ${\mathit{B}}_{\mathit{d}}^{0\mathrm{\ensuremath{-}}}$B${\mathrm{\ifmmode\bar\else\textasciimacron\fi{}}}_{\mathit{d}}^{0}$ mixing. We present a clear factorization description of partial decay rates and CP asymmetries for B\ensuremath{\rightarrow}PP, PV, and VV processes, and illustrate the rescattering effects of final-state particles on CP violation in those tree-penguin interfering channels. Our numerical estimates show that it is possible to uncover the penguin-diagram-induced CP violation in a handful of self-tagging modes if about ${10}^{8}$--${10}^{9}$ B events are accumulated.