Re: Re: emulating no-senders notifications in L4?

[Espen Skoglund]

[Ondrej Hurt]
>>> Are there no cases where the kernel kills a task without telling
>>> any other task?

>> No.  If, say, an exception is not handled properly the kernel just
>> resumes it (or just removes it from the run queue, generating the
>> impression that the task is spinning in an endless loop).  An
>> example of such an exception is when a thread generates a page
>> fault but there is no valid pager associated with thread that can
>> handle the fault.  The kernel can not simply decide to kill a task.
>> The decission to kill a task must be taken at user-level.

> But who tells the user-level that a task is a candidate for
> slaughter ? The L4 user manuals says that shouldn't an exception be
> handled, the faulting thread would be blocked forever. Nothing more.

The user-level (exception handler) tells the user-level.  Of course,
if there's no valid exception handler installed it can be pretty hard
to detect such faulty tasks.  It can be done through other ways of
monitoring the task, though.  The faulting task can also be given a
new exception handler at a later point in time, enabling the exception
to be caught when that happens.

        eSk