tech-kern archive

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

Re: very bad behavior on overquota writes

On Wed, Dec 05, 2012 at 07:45:42AM -0800, Chuck Silvers wrote:
> don't you need a "mutex_enter(vp->v_interlock)" before calling VOP_PUTPAGES()?
> (did you test it with LOCKDEBUG?)  otherwise it looks ok to me.

Yes the mutex_enter() is needed, I found it later (I didn't remember
we didn't have DIAGNOSTIC in GENERIC in the release branch).

> in the places were we skip the list walk because there aren't supposed to be
> any pages with offsets larger than some threshold, it would be nice if we 
> could
> have a DEBUG check that would verify there actually aren't any pages with
> offsets that are larger than that.

I'll try to look at adding this.

> I should also mention that this kind of workaround shouldn't be necessary
> anymore once the yamt-pagecache branch is merged... there, the uvm_object list
> of pages is replaced by a tree so that finding a few pages is fast no matter
> how many other pages of that object are present.  I don't know when that
> will be ready though.

Good to know, thanks !

Manuel Bouyer <>
     NetBSD: 26 ans d'experience feront toujours la difference

Home | Main Index | Thread Index | Old Index