We measure the sum of the neutrino particle masses using the three-dimensional galaxy power spectrum of the SDSS-III Baryon Oscillation Spectroscopic Survey (BOSS) Data Release 9 (DR9) CMASS galaxy sample. Combined with the cosmic microwave background (CMB), supernova (SN) and additional baryonic acoustic oscillation (BAO) data, we find upper 95 percent confidence limits of the neutrino mass $Sigma m_{ u}<0.340$ eV within a flat $Lambda$CDM background, and $Sigma m_{ u}<0.821$ eV, assuming a more general background cosmological model. The number of neutrino species is measured to be $N_{rm eff}=4.308pm0.794$ and $N_{rm eff}=4.032^{+0.870}_{-0.894}$ for these two cases respectively. We study and quantify the effect of several factors on the neutrino measurements, including the galaxy power spectrum bias model, the effect of redshift-space distortion, the cutoff scale of the power spectrum, and the choice of additional data. The impact of neutrinos with unknown masses on other cosmological parameter measurements is investigated. The fractional matter density and the Hubble parameter are measured to be $Omega_M=0.2796pm0.0097$, $H_0=69.72^{+0.90}_{-0.91}$ km/s/Mpc (flat $Lambda$CDM) and $Omega_M=0.2798^{+0.0132}_{-0.0136}$, $H_0=73.78^{+3.16}_{-3.17}$ km/s/Mpc (more general background model). Based on a Chevallier-Polarski-Linder (CPL) parametrisation of the equation-of-state $w$ of dark energy, we find that $w=-1$ is consistent with observations, even allowing for neutrinos. Similarly, the curvature Omega_K and the running of the spectral index $alpha_s$ are both consistent with zero. The tensor-to-scaler ratio is constrained down to $r<0.198$ (95 percent CL, flat $Lambda$ CDM) and $r<0.440$ (95 percent CL, more general background model).