Linux-NVDIMM Archive on lore.kernel.org
 help / color / Atom feed
From: Dan Carpenter <dan.carpenter@oracle.com>
To: Dan Williams <dan.j.williams@intel.com>
Cc: linux-nvdimm <linux-nvdimm@lists.01.org>
Subject: Re: [bug report] libnvdimm, nvdimm: dimm driver and base libnvdimm device-driver infrastructure
Date: Wed, 5 Feb 2020 22:08:45 +0300
Message-ID: <20200205190845.GD24804@kadam> (raw)
In-Reply-To: <CAPcyv4itFypOmv38Oo=DRWk_1Y3PFhPpYPDzxShmZVY9ZsTNLA@mail.gmail.com>

On Wed, Feb 05, 2020 at 10:23:00AM -0800, Dan Williams wrote:
> > > >    506          if (device_add(dev) != 0) {
> > > >    507                  dev_err(dev, "%s: failed\n", __func__);
> > > >    508                  put_device(dev);
> > > >                         ^^^^^^^^^^^^^^^
> > > >    509          }
> > > >    510          put_device(dev);
> > > >                 ^^^^^^^^^^^^^^
> > > >    511          if (dev->parent)
> > > >    512                  put_device(dev->parent);
> > > >    513  }
> > > >
> > > > We call get_device() from __nd_device_register(), I guess.  It seems
> > > > buggy to call put device twice on error.
> > >
> > > The registration path does:
> > >
> > >         get_device(dev);
> > >
> > >         async_schedule_dev_domain(nd_async_device_register, dev,
> > >                                   &nd_async_domain);
> > >
> > > ...and device_add() does its own get_device().
> >
> > device_add() does its own put_device() at the end so it's a net zero.
> >
> 
> It does it's own, yes, but the put_device() after device_add() failure
> is there to drop the reference taken by device_initialize().
> Otherwise, device_add() has always documented:
> 
>  * NOTE: _Never_ directly free @dev after calling this function, even
>  * if it returned an error! Always use put_device() to give up your
>  * reference instead.
> 
> ...so what am I missing?

The "never call kfree" is hopefully straight forward because the kobject
needs to do its own cleanup.

__nvdimm_create() allocates the dev.
nd_device_register() calls device_initialize() which call kobject_init()
   so the refcount is 1.
__nd_device_register() call get_device() so the refcount is now two.
nd_async_device_register() decrements the refcount once on success.

But if device_add() fails then it decrements it twice.  Now the refcount
is zero so we call nvdimm_release().  This leads to a use after free on
the next line:

	put_device(dev);
	if (dev->parent)

There is a trick here because depending on the debug options it
might free immediately or it might call nvdimm_release() after 4
seconds.  See kobject_release() for details.

Either way if device_add() fails we return back to __nvdimm_create()
and return the zero reference count "nvdimm" pointer, which is going
to be a problem.

regards,
dan carpenter
_______________________________________________
Linux-nvdimm mailing list -- linux-nvdimm@lists.01.org
To unsubscribe send an email to linux-nvdimm-leave@lists.01.org

  reply index

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-02-05 12:38 Dan Carpenter
2020-02-05 17:47 ` Dan Williams
2020-02-05 18:10   ` Dan Carpenter
2020-02-05 18:23     ` Dan Williams
2020-02-05 19:08       ` Dan Carpenter [this message]
2020-02-05 19:16         ` Dan Williams
2020-02-05 19:28           ` Dan Carpenter
2020-02-05 20:04             ` Dan Williams
2020-02-05 20:18               ` Dan Carpenter

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20200205190845.GD24804@kadam \
    --to=dan.carpenter@oracle.com \
    --cc=dan.j.williams@intel.com \
    --cc=linux-nvdimm@lists.01.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Linux-NVDIMM Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-nvdimm/0 linux-nvdimm/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-nvdimm linux-nvdimm/ https://lore.kernel.org/linux-nvdimm \
		linux-nvdimm@lists.01.org
	public-inbox-index linux-nvdimm

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.01.lists.linux-nvdimm


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git