linux-kernel - Re: [PATCH 06/15] genhd: Add return code to device_add

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20160817084823.GB19326@al.usersys.redhat.com>
Date:	Wed, 17 Aug 2016 16:48:23 +0800
From:	Fam Zheng <famz@...hat.com>
To:	Cornelia Huck <cornelia.huck@...ibm.com>
Cc:	linux-kernel@...r.kernel.org, Jens Axboe <axboe@...nel.dk>,
	linux-block@...r.kernel.org,
	Sergey Senozhatsky <sergey.senozhatsky.work@...il.com>,
	"Michael S. Tsirkin" <mst@...hat.com>,
	Michael Ellerman <mpe@...erman.id.au>,
	virtualization@...ts.linux-foundation.org,
	linux-nvme@...ts.infradead.org, "Ed L. Cashin" <ed.cashin@....org>,
	Keith Busch <keith.busch@...el.com>,
	Minchan Kim <minchan@...nel.org>,
	Paul Mackerras <paulus@...ba.org>,
	Benjamin Herrenschmidt <benh@...nel.crashing.org>,
	linux-mtd@...ts.infradead.org,
	Brian Norris <computersforpeace@...il.com>,
	linuxppc-dev@...ts.ozlabs.org,
	David Woodhouse <dwmw2@...radead.org>,
	Nitin Gupta <ngupta@...are.org>
Subject: Re: [PATCH 06/15] genhd: Add return code to device_add_disk

On Wed, 08/17 10:49, Cornelia Huck wrote:
> On Wed, 17 Aug 2016 15:15:06 +0800
> Fam Zheng <famz@...hat.com> wrote:
> 
> > @@ -613,10 +614,8 @@ void device_add_disk(struct device *parent, struct gendisk *disk)
> >  	disk->flags |= GENHD_FL_UP;
> > 
> >  	retval = blk_alloc_devt(&disk->part0, &devt);
> > -	if (retval) {
> > -		WARN_ON(1);
> > -		return;
> > -	}
> > +	if (retval)
> > +		goto fail;
> >  	disk_to_dev(disk)->devt = devt;
> > 
> >  	/* ->major and ->first_minor aren't supposed to be
> > @@ -625,16 +624,26 @@ void device_add_disk(struct device *parent, struct gendisk *disk)
> >  	disk->major = MAJOR(devt);
> >  	disk->first_minor = MINOR(devt);
> > 
> > -	disk_alloc_events(disk);
> > +	retval = disk_alloc_events(disk);
> > +	if (retval)
> > +		goto fail;
> > 
> >  	/* Register BDI before referencing it from bdev */
> >  	bdi = &disk->queue->backing_dev_info;
> > -	bdi_register_owner(bdi, disk_to_dev(disk));
> > +	retval = bdi_register_owner(bdi, disk_to_dev(disk));
> > +	if (retval)
> > +		goto fail;
> > 
> > -	blk_register_region(disk_devt(disk), disk->minors, NULL,
> > -			    exact_match, exact_lock, disk);
> > -	register_disk(parent, disk);
> > -	blk_register_queue(disk);
> > +	retval = blk_register_region(disk_devt(disk), disk->minors, NULL,
> > +				     exact_match, exact_lock, disk);
> > +	if (retval)
> > +		goto fail;
> > +	retval = register_disk(parent, disk);
> > +	if (retval)
> > +		goto fail;
> > +	retval = blk_register_queue(disk);
> > +	if (retval)
> > +		goto fail;
> > 
> >  	/*
> >  	 * Take an extra ref on queue which will be put on disk_release()
> > @@ -644,10 +653,20 @@ void device_add_disk(struct device *parent, struct gendisk *disk)
> > 
> >  	retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj,
> >  				   "bdi");
> > +	if (retval)
> > +		goto fail;
> > +
> > +	retval = disk_add_events(disk);
> > +	if (retval)
> > +		goto fail;
> > +
> > +	retval = blk_integrity_add(disk);
> > +	if (retval)
> > +		goto fail;
> > +	return 0;
> > +fail:
> >  	WARN_ON(retval);
> > -
> > -	disk_add_events(disk);
> > -	blk_integrity_add(disk);
> > +	return retval;
> >  }
> 
> Noticed this when trying to figure out whether the error handling in
> virtio_blk was correct:
> 
> Shouldn't you try to cleanup/rewind so that any structures are in a
> sane state after failure? The caller doesn't know where device_add_disk
> failed, and calling del_gendisk unconditionally like virtio_blk does is
> probably not the right thing to do (at the very least, I don't think
> unregistering a device that has not been registered is likely to work).
> 

Yes, I think all callers need to be reviewed before device_add_disk do the
clean up on error. For this patchset I wanted to keep the change small.

Fam