Fuse-wise, you need to make sure the primary one has Full-Swing Crystal Oscillator enabled (not the Low-Power one you typically find enabled by default when you burn an arduino bootloader via the IDE), the ClockOut fuse bit is enabled, then PORTB0 (aka digital pin 8 on an arduino) has to go into XTAL1 on the 2nd one and that one needs its fuses set to External Oscillator. Likewise, digital pin 8 on the first processor can't be used for I/O.
For the 328P, example fuse settings...
Master (with the crystal across XTAL1/XTAL2 and 22pF ceramic caps between each XTAL line and GND)-
Extended fuse: 0x05 (BOD enabled)
High fuse: 0xDE (Self-Programming enabled, Reset Vector=Bootloader)
Low fuse: 0x97 (CLKO, SUT=01, CKSEL=Full Swing Oscillator)
**note that the exact fuses may differ depending on what kind of crystal/resonator you have:
A quartz crystal with BOD enabled, as we have it here, should have SUT=01 with CKSEL bits = 0111, yielding a low fuse of 0x97
A ceramic resonator with BOD enabled should have SUT=10 and CKSEL bits = 0110, which would yield a low-fuse of 0xA6.
Slave processor (with master's PORTB0 pin going into slave's XTAL1, no caps required and XTAL2 suddenly becomes PORTB7, a usable I/O pin)-
Extended fuse: 0x05
High fuse: 0xDE
Low fuse: 0xC0 (SUT=00, CKSEL=External Oscillator)
Note : This advice comes from page 28 of the ATmega48A/48PA/88A/88PA/168A/168PA/328/328P datasheet;
Low Power Crystal Oscillator
Pins XTAL1 and XTAL2 are input and output, respectively, of an inverting amplifier which can be configured for use as an On-chip Oscillator, as shown in Figure 8-2 on page 29. Either a quartz or ceramic resonator may be used.
This Crystal Oscillator is a low power oscillator, with reduced voltage swing on the XTAL2 output. It gives the lowest power consumption, but is not capable of driving other clock inputs, and may be more susceptible to noise in noisy environments. In these cases, refer to the "Full Swing Crystal Oscillator" on page 30.
... There is some confusion though, because the Clock Output Buffer section implies any clock may be used:
Clock Output Buffer
The device can output the system clock on the CLKO pin. To enable the output, the CKOUT Fuse has to be programmed. This mode is suitable when the chip clock is used to drive other circuits on the system. The clock also will be output during reset, and the normal operation of I/O pin will be overridden when the fuse is programmed. Any clock source, including the internal RC Oscillator, can be selected when the clock is output on CLKO. If the System Clock Prescalar is used, it is the divided system clock that is output.