Opening a UDP Socket in RISC-V Assembly

The following is an aggregation of a Twitter thread I posted on April 14th, 2022.

Ever wondered how to open a UDP socket in @risc_v assembly? Wonder no more!


li    a0, 2
li    a1, 2
li    a2, 0
li    a7, 198
ecall

Let’s walk through it! 👇🏼🧵


The first thing to understand is that we are just a “normal” program running in user space. We don’t have special privileges in the system, and opening a socket is a privileged operation. In order to accomplish this, we’ll need to “ask” the kernel to do something on our behalf.


Programs in user space can talk to the kernel via system calls. @risc_v provides the ecall instruction to communicate from a lower privilege level to a higher one. In this case, we are communicating from “user mode” to “supervisor mode”. Read more here.


Issuing an ecall instruction causes a “precise trap”. The kernel registers “trap handlers” to respond to different events that cause traps. In this case, we need to tell the kernel what we need it to do for us.


The kernel supports different syscall “calling conventions” for different architectures. These can be found in the Linux man pages, but here is a summary:

22-05-14_1


We can see for @risc_v that the system call (syscall) number needs to be present in argument register 7 (a7). Walking backwards up our sequence of instructions, you can see we are using the load immediate pseudo-instruction (li) to load 198 into a7.


So how do we know to use 198? We need to look in the kernel! Syscall numbers are defined in various locations for different architectures; for @risc_v we can follow include directives from unistd.h:


If you don’t want to look in kernel source to find the syscall number for a given architecture, you can use this wonderful page from @haerwu that lists syscall numbers for all architectures: link.


Okay, now that we know that 198 corresponds to the socket syscall, we need to determine what arguments it requires to inform the type of socket (UDP) we want to open. Back to the man pages we go!

22-05-14_2


So we need to provide domain, type, and protocol. Luckily the man page also defines the options available to us, but we need to map those to the correct integer values. Specifically we want:

  • Domain: AF_INET
  • Type: SOCK_DGRAM
  • Protocol: 0 (default)

22-05-14_3


AF_INET and SOCK_DGRAM are both defined in socket.h in glibc:

Conveniently, both are defined with the integer 2.


Issuing a syscall is not so different from calling a function in your program as it requires passing arguments according to a calling convention. We follow the @risc_v psABI calling convention and pass our arguments in the a0, a1, and a2 registers.


Lastly, we need to issue our ecall command. This will cause the hart to jump to the address defined in the Supervisor Trap Vector register (stvec), which is where the kernel has placed the logic to handle different types of traps.


The kernel will handle our request and return either the file descriptor for the socket, or -1 indicating a failed operation. That’s all for today though! If folks enjoy threads like this, I’ll continue posting as I work on my @risc_v assembly implementation of @quicwg :)