It is typical to use a 32-bit depth-buffer:
GL_DEPTH_COMPONENT32F //internal format
GL_DEPTH_COMPONENT //depth only
GL_FLOAT //float(32-bit depth)
If you also wish to utilize the stencil buffer, it is typical to dedicate 8 of those 32-bits to the stencil:
GL_DEPTH24_STENCIL8 //internal format
GL_DEPTH_STENCIL //depth + stencil
GL_UNSIGNED_INT_24_8 //ushort(24-bit depth) + uchar(stencil)
If you require 32-bit precision for depth, but also require the stencil:
GL_DEPTH32F_STENCIL8 //internal format
GL_DEPTH_STENCIL //depth + stencil
GL_FLOAT_32_UNSIGNED_INT_24_8_REV //float(32-bit depth) + ushort(spare) + uchar(stencil)
Refer to the last chart under "Description".
Stencil diagram:
