A zero DN for all pixels? Was it for just one pixel?

See the link below for conversion to TOA reflectance.  For the computation of MNDWI and NDWI, either ERDAS, ARCGIS, or ENVI can easily do the calculations as long as you use the right equations.

http://landsat.usgs.gov/Landsat8_Using_Product.php

For Landsat 7 you have to calculate the radiation first. Therefore, you need values from min_max_pixel_value (say minP and maxP) and min_max_radiance (say minR and maxR) provided in the MTL.txt file that completes every Landsat product download.

Equation: RAD =  (maxR - minR)/(maxP - minP) * ((DN - minP) + minR); no Problem if DN = 0

Calculation of TOA is the second step. REF = (Pi * RAD * d^2)/(ESUN * cosTHETA) where d is the distance Earth-Sun in AU (e.g. in Chander and Markham, 2003), ESUN is provided in Chander et al. 2009 and theta is 90 - SUN_ELEVATION (MTL-file).

For Landsat 8 use the guide provided by Eric.

All in all no problems with DN = 0 technacally speaking. However, if the majority of your scenes contains zeros, something is clearly wrong!

I guess you got zero values on the NDWI and MNDWI results. In ERDAS 2014 from raster menu on radiometric group apply replace bad lines, then apply Landsat 7 reflectance and finally calculate the NDWI as shown in the attached file

You probably have bad lines on your scene. Be sure that your image was taken after 2003. As Mohamed Mortada El-Sharkawy said, use ERDAS to replace bad lines. Then follow the instructions given by Sacha Schneiderwind to convert your DN values into TOA reflectance.

But if your image was taken before 2003, that means there is something wrong with your data. Redownload the image to be sure of it.

Thanks to all of you. I've calculated TOA in ENVI 5.1 and got right results of NDWI and NDVI. Since the image is of a coastal area, so gap masking and cloud masking and then finally, TOA helps to get the right result, but definitely, I'll try replacing bad lines.as you've said. 

Thanks

Supriti