RE: [PATCH] xen-blkback: allow module to be cleanly unloaded

[Durrant, Paul]

> -----Original Message-----
> From: Jan Beulich <jbeulich@xxxxxxxx>
> Sent: 29 November 2019 11:56
> To: Durrant, Paul <pdurrant@xxxxxxxxxx>
> Cc: xen-devel@xxxxxxxxxxxxxxxxxxxx; linux-block@xxxxxxxxxxxxxxx; linux-
> kernel@xxxxxxxxxxxxxxx; Roger Pau Monnà <roger.pau@xxxxxxxxxx>; Jens Axboe
> <axboe@xxxxxxxxx>; Konrad Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx>
> Subject: Re: [PATCH] xen-blkback: allow module to be cleanly unloaded
> 
> On 29.11.2019 12:31, Paul Durrant wrote:
> > --- a/drivers/block/xen-blkback/xenbus.c
> > +++ b/drivers/block/xen-blkback/xenbus.c
> > @@ -173,6 +173,8 @@ static struct xen_blkif *xen_blkif_alloc(domid_t
> domid)
> >  	init_completion(&blkif->drain_complete);
> >  	INIT_WORK(&blkif->free_work, xen_blkif_deferred_free);
> >
> > +	__module_get(THIS_MODULE);
> > +
> >  	return blkif;
> >  }
> >
> > @@ -320,6 +322,8 @@ static void xen_blkif_free(struct xen_blkif *blkif)
> >
> >  	/* Make sure everything is drained before shutting down */
> >  	kmem_cache_free(xen_blkif_cachep, blkif);
> > +
> > +	module_put(THIS_MODULE);
> >  }
> 
> I realize there are various example of this in the tree, but
> isn't this a flawed approach? __module_get() (nor even
> try_module_get()) will prevent an unload attempt ahead of it
> getting invoked, while execution is already in this module's
> .text section.

Good point. That does appear to be a race.

> I think the xenbus driver should do this
> before calling ->probe(), in case of its failure, and after
> a successful call to ->remove().
> 

That does sound better. I'll see if I can pick up other occurrences (certainly netback) and fix.

  Paul