Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20210426220728.1230340-1-ztong0001@gmail.com> <YJlE+Z2VKhamVWaw@kroah.com>
In-Reply-To: <YJlE+Z2VKhamVWaw@kroah.com>
From:   Tong Zhang <ztong0001@gmail.com>
Date:   Mon, 10 May 2021 15:20:02 -0700
Message-ID: <CAA5qM4CAdb_Aaay1_gjy_AC48Doxtga0f69E1P36=8vscsR9Eg@mail.gmail.com>
Subject: Re: [PATCH] misc: alcor_pci: fix null-ptr-deref when there is no PCI bridge
To:     Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc:     Arnd Bergmann <arnd@arndb.de>,
        open list <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Mon, May 10, 2021 at 7:36 AM Greg Kroah-Hartman
<gregkh@linuxfoundation.org> wrote:
>
> On Mon, Apr 26, 2021 at 06:07:27PM -0400, Tong Zhang wrote:
> > the PCI bridge might be NULL, so we'd better check before use it
> >
> > [    1.246492] BUG: kernel NULL pointer dereference, address: 00000000000000c0
> > [    1.248731] RIP: 0010:pci_read_config_byte+0x5/0x40
> > [    1.253998] Call Trace:
> > [    1.254131]  ? alcor_pci_find_cap_offset.isra.0+0x3a/0x100 [alcor_pci]
> > [    1.254476]  alcor_pci_probe+0x169/0x2d5 [alcor_pci]
> >
> > Signed-off-by: Tong Zhang <ztong0001@gmail.com>
> > ---
> >  drivers/misc/cardreader/alcor_pci.c | 3 +++
> >  1 file changed, 3 insertions(+)
> >
> > diff --git a/drivers/misc/cardreader/alcor_pci.c b/drivers/misc/cardreader/alcor_pci.c
> > index cd402c89189e..1c33453fd5c7 100644
> > --- a/drivers/misc/cardreader/alcor_pci.c
> > +++ b/drivers/misc/cardreader/alcor_pci.c
> > @@ -102,6 +102,9 @@ static int alcor_pci_find_cap_offset(struct alcor_pci_priv *priv,
> >       u8 val8;
> >       u32 val32;
> >
> > +     if (!pci)
> > +             return 0;
> > +
> >       where = ALCOR_CAP_START_OFFSET;
> >       pci_read_config_byte(pci, where, &val8);
> >       if (!val8)
> > --
> > 2.25.1
> >
>
> I do not understand, how can pci ever be NULL?  There is only 1 way this

Hi Greg,
I think the problem is with
    priv->parent_pdev = pdev->bus->self
where bus->self can be NULL. when bus->self is NULL, calling
    alcor_pci_find_cap_offset(priv, priv->parent_pdev) is equivalent
to alcor_pci_find_cap_offset(priv, NULL)

- Tong

> function can be called, and it's through the alcor_pci_probe() call,
> which should have always set up the parent and pci pointers that get
> passed to this function.
>
> How can that not happen?  If it can happen, then something earlier than
> this should be fixed instead of papering over the root problem here.
>
> How did you duplicate the crash you list above?
>
> thanks,
>
> greg k-h