Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20200404141827.26255-1-ap420073@gmail.com> <20200404155040.GC1476305@kroah.com>
In-Reply-To: <20200404155040.GC1476305@kroah.com>
From:   Taehee Yoo <ap420073@gmail.com>
Date:   Sun, 5 Apr 2020 01:35:16 +0900
Message-ID: <CAMArcTX6j5bCLmLOcD13W8NkXtMHO-uouOnhnRAi4i5xdU+B3Q@mail.gmail.com>
Subject: Re: [PATCH net v2 1/3] class: add class_has_file_ns() helper function
To:     Greg KH <gregkh@linuxfoundation.org>
Cc:     David Miller <davem@davemloft.net>,
        Jakub Kicinski <kuba@kernel.org>, rafael@kernel.org,
        j.vosburgh@gmail.com, vfalico@gmail.com,
        Andy Gospodarek <andy@greyhouse.net>,
        Netdev <netdev@vger.kernel.org>,
        LKML <linux-kernel@vger.kernel.org>, mitch.a.williams@intel.com
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Sun, 5 Apr 2020 at 00:50, Greg KH <gregkh@linuxfoundation.org> wrote:
>

Hi Greg,
Thank you for your review!

> On Sat, Apr 04, 2020 at 02:18:27PM +0000, Taehee Yoo wrote:
> > The new helper function is to check whether the class file is existing
> > or not. This function will be used by networking stack to
> > check "/sys/class/net/*" file.
> >
> > Reported-by: syzbot+830c6dbfc71edc4f0b8f@syzkaller.appspotmail.com
> > Fixes: b76cdba9cdb2 ("[PATCH] bonding: add sysfs functionality to bonding (large)")
> > Signed-off-by: Taehee Yoo <ap420073@gmail.com>
> > ---
> >
> > v1 -> v2:
> >  - Implement class_has_file_ns() instead of class_find_and_get_file_ns().
> >  - Change headline.
> >  - Add kernel documentation comment.
> >
> >  drivers/base/class.c         | 22 ++++++++++++++++++++++
> >  include/linux/device/class.h |  3 ++-
> >  2 files changed, 24 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/base/class.c b/drivers/base/class.c
> > index bcd410e6d70a..a2f2787f6aa7 100644
> > --- a/drivers/base/class.c
> > +++ b/drivers/base/class.c
> > @@ -105,6 +105,28 @@ void class_remove_file_ns(struct class *cls, const struct class_attribute *attr,
> >               sysfs_remove_file_ns(&cls->p->subsys.kobj, &attr->attr, ns);
> >  }
> >
> > +/**
> > + * class_has_file_ns - check whether file is existing or not
> > + * @cls: the compatibility class
> > + * @name: name to look for
> > + * @ns: the namespace tag to use
> > + */
> > +bool class_has_file_ns(struct class *cls, const char *name,
> > +                    const void *ns)
>
> Why would you use this?  And what happens if the file shows up, or goes
> away, instantly after this call is made?
>
> This feels very broken.
>

Ah, I missed considering other usescases.
If other users don't use locks, this function would return incorrect
information. The problem seems to become from that this function
calls kernfs_put().

Thanks a lot!
Taehee Yoo