Re: [PATCH 2 of 4] Introduce i386 fibril scheduling

[Zach Brown <zach.brown@oracle.com>]

> let me clarify this: i very much like your AIO patchset in general, in
> the sense that it 'completes' the AIO implementation: finally  
> everything
> can be done via it, greatly increasing its utility and hopefully its
> penetration. This is the most important step, by far.

We violently agree on this :).

> what i dont really like /the particular/ concept above - the
> introduction of 'fibrils' as a hard distinction of kernel threads.  
> They
> are /almost/ kernel threads, but still by being different they create
> alot of duplication and miss out on a good deal of features that  
> kernel
> threads have naturally.

I might quibble with some of the details, but I understand your  
fundamental concern.  I do.  I don't get up each morning *thrilled*  
by the idea of having to update lockdep, sysrq-t, etc, to understand  
these fibril things :).  The current fibril switch isn't nearly as  
clever as the lock-free task scheduling switch.  It'd be nice if we  
didn't have to do that work to optimize the hell out of it, sure.

> It kind of hurts to say this because i'm usually quite concept-happy -
> one can easily get addicted to the introduction of new core kernel
> concepts :-)

:)

> so my suggestions center around the notion of extending kernel threads
> to support the features you find important in fibrils:
>
>> would it be hard to redo your AIO patches based on a pool of plain
>> simple kernel threads?

It'd certainly be doable to throw together a credible attempt to  
service "asys" system call submission with full-on kernel threads.   
That seems like reasonable due diligence to me.  If full-on threads  
are almost as cheap, great.  If fibrils are so much cheaper that they  
seem to warrant investing in, great.

I am concerned about the change in behaviour if we fall back to full  
kernel threads, though.  I really, really, want aio syscalls to  
behave just like sync ones.

Would your strategy be to update the syscall implementations to share  
data in task_struct so that there isn't as significant a change in  
behaviour?  (sharing current->ioprio, instead if just inheriting it,  
for example.).  We'd be betting that there would be few of these and  
that they'd be pretty reasonable to share?

- z