Interrupts and exceptions are both transparent, forced transfers of execution from a currently running program to a handler. They (misleadingly) are mutually registered in the IDT.
Interrupts -- occur randomly through hardware signals and deliberate INTs in code.
Exceptions -- occur as a result of error conditions in executing an instruction.

Interrupts can be divided into two groups:
1) Hardware interrupts -- only randomly occurring switches in entire kernel; thus, maskability matters. IF in EFLAGS masks these interrupts.
2) Software interrupts -- system calls from user land.

Exceptions can be divided into three groups:
3) Processor-detected program errors:
3a) Faults -- detected before %eip++, can be corrected.
3b) Traps -- detected after %eip++ (only difference from faults), can be corrected.
3c) Aborts -- can't be corrected.
4) Software-generated exceptions -- INTO, INT 3, BOUND.
5) Machine-check exceptions -- hardware errors.

We can handle interrupts and exceptions in two ways:
i) Task-gate descriptor -- hardware switches using TSS descriptors.
ii) Interrupt-gate descriptor -- clears IF flag, thus disabling maskable hardware interrupts.
iii) Trap-gate descriptor -- same as interrupt-gate, except doesn't clear IF.

These descriptors are customizable:
* Segment selector (of code segment) determines handler's permission level.
* Descriptor priviledge level (DPL) required to execute the handler. 0 for ring0 only, 3 for ring0-3 access.