Intel 80286, 80287 manual Task State Segment Descriptors

TASKS AND STATE TRANSITIONS

Each TSS consists of two parts, a static portion and a dynamic portion. The static entries are never changed by the 80286, while the dynamic entries are changed by each task switch out of this task. The static portions of this segment are the task LDT selector and the initial SS:SP stack pointer addresses for levels 0-2.

The modifiable or dynamic portion of the task state segment consists of all dynamically-variable and programmer-visible processor registers, including flags, segment registers, and the instruction pointer. It also includes the linkage word used to chain nested invocations of different tasks.

The link word provides a history of which tasks invoked others~ The link word is important for restart- ing an interrupted task when the interrupt has been serviced. Placing the back link in the TSS protects the identity of the interrupted task from changes by the interrupt task, since the TSS is not writable by the interrupt task. (In most systems only the operating system has sufficient privilege to create or use a writable data segment "alias" descriptor for the TSS.)

The stack pointer entries in the TSS for privilege levels 0-2 are static (i.e., never written during a privilege or task switch). They define the stack to use upon entry to that privilege level. These stack entries are initialized by the operating system when the task is created. If a privilege level is never used, no stack need be allocated for it.

When entering a more privileged level, the caller's stack pointer is saved on the stack of the new privilege level, not in the TSS. Leaving the privilege level requires popping the caller's return address and stack pointer off the current stack. The stack pointer at that time will be the same as the initial value loaded from the TSS upon entry to the privilege level.

There is only one stack active at any time, the one defined by the SS and SP registers. The only other stacks that may be non-empty are those at outer (less privileged) levels that called the current level. Stacks for inner levels must be empty, since outward (to numerically larger privilege levels) calls from

inner levels are not allowed. .

The location of the stack pointer for an outer privilege level will always be found at the start of the stack of the inner privilege level called by that level. That stack may be the initial stack for this privilege level or an outer level. Look at the start of the stack for this privilege level. The TSS contains the starting stack address for levels 0-2. If the RPL of the saved SS selector is the privilege level required, then the stack pointer has been found. Otherwise, go to the beginning of the stack defined by that value and look at the saved SS:SP value there.

8.2.1 Task State Segment Descriptors

A special descriptor is used for task state segments. This descriptor must be accessible at all times; therefore, it can appear only in the GDT. The access byte distinguishes TSS descriptors from data or code segment descriptors. When bits 0 through 4 of the access byte are 00001 or 00011, the descriptor is for a TSS.

The complete layout of a task state segment descriptor is shown in figure 8-2.

Like a data segment, the descriptor contains a base address and limit field. The limit must be at least 002BH (43) to contain the minimum amount of information required for a TSS. An invalid task excep- tion will occur if an attempt is made to switch to a task whose TSS descriptor limit is less than 43. The error code will identify the bad TSS.

8-3