Re: [PATCH v2] misc: alcor_pci: fix null-ptr-deref when there is no PCI bridge

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Wed, May 12, 2021 at 09:24:55AM -0700, Tong Zhang wrote:
> On Tue, May 11, 2021 at 11:24 PM Greg Kroah-Hartman
> <gregkh@linuxfoundation.org> wrote:
> >
> > On Tue, May 11, 2021 at 05:29:38PM -0400, Tong Zhang wrote:
> > > Device might be attached to root complex directly. In this case,
> > > bus->self(bridge) will 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>
> > > Co-Developed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> > > ---
> > > v2: check before calling alcor_pci_find_cap_offset()
> > >
> > >  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..175c6b06f7aa 100644
> > > --- a/drivers/misc/cardreader/alcor_pci.c
> > > +++ b/drivers/misc/cardreader/alcor_pci.c
> > > @@ -139,6 +139,9 @@ static void alcor_pci_init_check_aspm(struct alcor_pci_priv *priv)
> > >       u32 val32;
> > >
> > >       priv->pdev_cap_off    = alcor_pci_find_cap_offset(priv, priv->pdev);
> > > +
> > > +     if (!priv->parent_pdev)
> > > +             return;
> >
> > That feels wrong, you just prevented all of the remaining logic in this
> > call to not be set up.  Did you test this and did the driver and device
> > still work properly if it hits this check?
> >
> > thanks,
> >
> > greg k-h
> 
> Sorry, probably I misunderstood your previous email. Please correct me
> if I am wrong.
> What I did here is to disable ASPM completely if it is attached to the
> root complex, which is OK since ASPM is optional and we cannot really
> do ASPM on the root complex.
> Also, alcor_pci_init_check_aspm() is responsible for checking the
> device and its parent(bridge) aspm capability offset.
> This function will set priv->parent_cap_off and priv->pdev_cap_off.
> Those two capability offset will be used in alcor_pci_aspm_ctrl() to
> determine whether the PCI link+device supports aspm or not.
> In our case the pdev_cap_off remains 0 when alcor_pci_aspm_ctrl() is
> called and it simply returns.
> So I think it can still work.

Ok, that makes more sense.

Can you document that better and add a comment here, and properly handle
the whitespace and resubmit?

thanks,

greg k-h