tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: vwakeup: neg numoutput



On Mon, Jun 30, 2008 at 09:10:09PM +0200, Manuel Bouyer wrote:
> On Mon, Jun 30, 2008 at 10:27:07AM -0700, Bill Stouder-Studenmund wrote:
> > > > Would this be a good candidate for the atomic update operations? That 
> > > > way 
> > > > you don't need a lock.
> > > 
> > > Could be, but then it needs to be updated atomically everywhere.
> > 
> > Right.
> > 
> > But if you can't take a lock, then you need to do something. :-) And 
> > either cscope or eid (probably eid) can help you change the whole tree at 
> > once.
> 
> I don't know either, but I can do it with grep :)

Heh. I recommend learning mkid tools. eid is basically grep funneled into 
your favorite editor.

> I don't know how to translate this kind of construct then:
>                 mutex_enter(&vp->v_interlock);
>               while (vp->v_numoutput > 0)
>                       cv_wait(&vp->v_cv, &vp->v_interlock);
>               mutex_exit(&vp->v_interlock);

I think that's ok.

The only place that this code construct means you need to do something 
special is when you decriment (assuming we can never overflow 
v_numoutput). When decirmenting (which wasn't what started this as I 
recall), you need to also hold the interlock. Or, you need to notice when 
you decrimented to 0 then lock the interlock then cv_broadcast().

> they're used in several places. There are also places where it seems it's
> expected to have vp->v_numoutput stable for several lines of code:
> vfs_subr2.c:
>         mutex_enter(&vp->v_interlock);
>               while (vp->v_numoutput != 0)
>                       cv_wait(&vp->v_cv, &vp->v_interlock);
>       dirty = !LIST_EMPTY(&vp->v_dirtyblkhd);
>       mutex_exit(&vp->v_interlock);
> 
> there's also genfs_do_putpages().
> I'm not sure it's doable to change this to atomic ops.

I haven't looked at it, so I can't say.

> Last, I suspect ccd.c also has locking issues with v_numoutput and locking.
> 
> Would using a software interrupt to run the xbdback handler work to take
> v_interlock mutex ?

Off the cuff, probably.

Take care,

Bill

Attachment: pgpzFVC5AsORv.pgp
Description: PGP signature



Home | Main Index | Thread Index | Old Index