mi_switch
— switch
to another thread context
The
mi_switch
()
function implements the machine-independent prelude to a thread context
switch. It is the single entry point for every context switch and is called
from only a few distinguished places in the kernel. The context switch is,
by necessity, always performed by the switched thread, even when the switch
is initiated from elsewhere; e.g. preemption requested via Inter-Processor
Interrupt (IPI).
The various major uses of
mi_switch
()
can be enumerated as follows:
- From within a function such as
sleepq_wait(9)
or
turnstile_wait
()
when the current thread voluntarily relinquishes the CPU to wait for some
resource or lock to become available.
- Involuntary preemption due to arrival of a higher-priority thread.
- At the tail end of
critical_exit(9),
if preemption was deferred due to the critical section.
- Within the TDA_SCHED AST handler, when rescheduling
before the return to usermode was requested. There are several reasons for
this, a notable one coming from
sched_clock
()
when the running thread has exceeded its time slice.
- In the signal handling code (see
issignal(9))
if a signal is delivered that causes a process to stop.
- In
thread_suspend_check
()
where a thread needs to stop execution due to the suspension state of the
process as a whole.
- In
kern_yield(9)
when a thread wants to voluntarily relinquish the processor.
The flags argument to
mi_switch
()
indicates the context switch type. One of the following must be passed:
In addition to the switch type, callers must specify the nature of
the switch by performing a bitwise OR with one of the
SW_VOL
or SW_INVOL
flags,
but not both. Respectively, these flags denote whether the context switch is
voluntary or involuntary on the part of the current thread. For an
involuntary context switch in which the running thread is being preempted,
the caller should also pass the SW_PREEMPT
flag.
Upon entry to
mi_switch
(),
the current thread must be holding its assigned thread lock. It may be
unlocked as part of the context switch. After they have been rescheduled and
execution resumes, threads will exit mi_switch
()
with their thread lock unlocked.
mi_switch
()
records the amount of time the current thread has been running before
handing control over to the scheduler, via
sched_switch
().
After selecting a new thread to run, the scheduler will call
cpu_switch
() to perform the low-level context
switch.
cpu_switch
()
is the machine-dependent function that performs the actual switch from the
running thread oldtd to the chosen thread
newtd.