According to the standard Operating System Concepts classification, which…

System calls need a way to hand arguments from user code to the kernel, and operating-system theory identifies three general techniques for doing this:

• Pass the parameters directly in CPU registers.

• Store the parameters in a block (table) in memory, and pass only the address of that block in a register.

• Push the parameters onto the stack, to be popped off by the operating system.

Register-only passing is simple but is limited by the number of registers available, so it cannot scale to calls with many arguments. To avoid that limit, Linux (along with Solaris) uses the block/table technique: the calling program packs its parameters into a memory block and loads just that block's address into a register before trapping into the kernel, so the kernel can read an arbitrary number of arguments through this indirection.

The real x86-64 Linux syscall ABI does load the first several arguments directly into specific CPU registers (such as RDI, RSI, and RDX) -- that is a hardware/ABI-level implementation detail. The standard operating-systems textbook treatment (Silberschatz/Galvin) that this question is drawn from separately classifies parameter-passing into the three techniques above and explicitly attributes the block/table scheme to Linux (and Solaris). Among the four techniques offered here, it is that block/table scheme -- not the register-ABI detail -- which the question and its answer key are testing, which is why "By using block or table" is the technique that matches.