PCI: rcar-gen2: Fix crash in resource_list_first_type()
diff mbox series

Message ID 20200804120430.9253-1-geert+renesas@glider.be
State Accepted
Delegated to: Lorenzo Pieralisi
Headers show
Series
  • PCI: rcar-gen2: Fix crash in resource_list_first_type()
Related show

Commit Message

Geert Uytterhoeven Aug. 4, 2020, 12:04 p.m. UTC
The conversion to modern host bridge probing made the driver allocate
its private data using devm_pci_alloc_host_bridge(), but forgot to
remove the old allocation.  Hence part of the driver initialization is
done using the new instance, while another part is done using the old
instance, leading to a crash due to uninitialized bridge DMA ranges:

    Unable to handle kernel NULL pointer dereference at virtual address 00000008
    pgd = (ptrval)
    [00000008] *pgd=00000000
    Internal error: Oops: 5 [#1] SMP ARM
    CPU: 0 PID: 1 Comm: swapper/0 Not tainted 5.8.0-rc1-shmobile-00035-g92d69cc6275845a7 #645
    Hardware name: Generic R-Car Gen2 (Flattened Device Tree)
    PC is at rcar_pci_probe+0x154/0x340
    LR is at _raw_spin_unlock_irqrestore+0x18/0x20

Fix this by dropping the old allocation.

Fixes: 92d69cc6275845a7 ("PCI: rcar-gen2: Convert to use modern host bridge probe functions")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
---
 drivers/pci/controller/pci-rcar-gen2.c | 4 ----
 1 file changed, 4 deletions(-)

Comments

Rob Herring Aug. 4, 2020, 3:13 p.m. UTC | #1
On Tue, Aug 4, 2020 at 6:04 AM Geert Uytterhoeven
<geert+renesas@glider.be> wrote:
>
> The conversion to modern host bridge probing made the driver allocate
> its private data using devm_pci_alloc_host_bridge(), but forgot to
> remove the old allocation.  Hence part of the driver initialization is
> done using the new instance, while another part is done using the old
> instance, leading to a crash due to uninitialized bridge DMA ranges:
>
>     Unable to handle kernel NULL pointer dereference at virtual address 00000008
>     pgd = (ptrval)
>     [00000008] *pgd=00000000
>     Internal error: Oops: 5 [#1] SMP ARM
>     CPU: 0 PID: 1 Comm: swapper/0 Not tainted 5.8.0-rc1-shmobile-00035-g92d69cc6275845a7 #645
>     Hardware name: Generic R-Car Gen2 (Flattened Device Tree)
>     PC is at rcar_pci_probe+0x154/0x340
>     LR is at _raw_spin_unlock_irqrestore+0x18/0x20
>
> Fix this by dropping the old allocation.
>
> Fixes: 92d69cc6275845a7 ("PCI: rcar-gen2: Convert to use modern host bridge probe functions")
> Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
> ---
>  drivers/pci/controller/pci-rcar-gen2.c | 4 ----
>  1 file changed, 4 deletions(-)

Acked-by: Rob Herring <robh@kernel.org>
Lorenzo Pieralisi Aug. 4, 2020, 4:13 p.m. UTC | #2
On Tue, Aug 04, 2020 at 02:04:30PM +0200, Geert Uytterhoeven wrote:
> The conversion to modern host bridge probing made the driver allocate
> its private data using devm_pci_alloc_host_bridge(), but forgot to
> remove the old allocation.  Hence part of the driver initialization is
> done using the new instance, while another part is done using the old
> instance, leading to a crash due to uninitialized bridge DMA ranges:
> 
>     Unable to handle kernel NULL pointer dereference at virtual address 00000008
>     pgd = (ptrval)
>     [00000008] *pgd=00000000
>     Internal error: Oops: 5 [#1] SMP ARM
>     CPU: 0 PID: 1 Comm: swapper/0 Not tainted 5.8.0-rc1-shmobile-00035-g92d69cc6275845a7 #645
>     Hardware name: Generic R-Car Gen2 (Flattened Device Tree)
>     PC is at rcar_pci_probe+0x154/0x340
>     LR is at _raw_spin_unlock_irqrestore+0x18/0x20
> 
> Fix this by dropping the old allocation.
> 
> Fixes: 92d69cc6275845a7 ("PCI: rcar-gen2: Convert to use modern host bridge probe functions")
> Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
> ---
>  drivers/pci/controller/pci-rcar-gen2.c | 4 ----
>  1 file changed, 4 deletions(-)

Squashed in the initial commit, pushed out on pci/misc.

Thanks,
Lorenzo

> diff --git a/drivers/pci/controller/pci-rcar-gen2.c b/drivers/pci/controller/pci-rcar-gen2.c
> index 046965d284a6d54e..c9530038ca9a53fc 100644
> --- a/drivers/pci/controller/pci-rcar-gen2.c
> +++ b/drivers/pci/controller/pci-rcar-gen2.c
> @@ -302,10 +302,6 @@ static int rcar_pci_probe(struct platform_device *pdev)
>  	if (mem_res->start & 0xFFFF)
>  		return -EINVAL;
>  
> -	priv = devm_kzalloc(dev, sizeof(struct rcar_pci_priv), GFP_KERNEL);
> -	if (!priv)
> -		return -ENOMEM;
> -
>  	priv->mem_res = *mem_res;
>  	priv->cfg_res = cfg_res;
>  
> -- 
> 2.17.1
>
Bjorn Helgaas Aug. 4, 2020, 4:22 p.m. UTC | #3
On Tue, Aug 04, 2020 at 05:13:25PM +0100, Lorenzo Pieralisi wrote:
> On Tue, Aug 04, 2020 at 02:04:30PM +0200, Geert Uytterhoeven wrote:
> > The conversion to modern host bridge probing made the driver allocate
> > its private data using devm_pci_alloc_host_bridge(), but forgot to
> > remove the old allocation.  Hence part of the driver initialization is
> > done using the new instance, while another part is done using the old
> > instance, leading to a crash due to uninitialized bridge DMA ranges:
> > 
> >     Unable to handle kernel NULL pointer dereference at virtual address 00000008
> >     pgd = (ptrval)
> >     [00000008] *pgd=00000000
> >     Internal error: Oops: 5 [#1] SMP ARM
> >     CPU: 0 PID: 1 Comm: swapper/0 Not tainted 5.8.0-rc1-shmobile-00035-g92d69cc6275845a7 #645
> >     Hardware name: Generic R-Car Gen2 (Flattened Device Tree)
> >     PC is at rcar_pci_probe+0x154/0x340
> >     LR is at _raw_spin_unlock_irqrestore+0x18/0x20
> > 
> > Fix this by dropping the old allocation.
> > 
> > Fixes: 92d69cc6275845a7 ("PCI: rcar-gen2: Convert to use modern host bridge probe functions")
> > Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
> > ---
> >  drivers/pci/controller/pci-rcar-gen2.c | 4 ----
> >  1 file changed, 4 deletions(-)
> 
> Squashed in the initial commit, pushed out on pci/misc.

I updated my 'next' branch with this.

Rob, are there any similar issues in other drivers that we should fix
before asking Linus to pull this?

> > diff --git a/drivers/pci/controller/pci-rcar-gen2.c b/drivers/pci/controller/pci-rcar-gen2.c
> > index 046965d284a6d54e..c9530038ca9a53fc 100644
> > --- a/drivers/pci/controller/pci-rcar-gen2.c
> > +++ b/drivers/pci/controller/pci-rcar-gen2.c
> > @@ -302,10 +302,6 @@ static int rcar_pci_probe(struct platform_device *pdev)
> >  	if (mem_res->start & 0xFFFF)
> >  		return -EINVAL;
> >  
> > -	priv = devm_kzalloc(dev, sizeof(struct rcar_pci_priv), GFP_KERNEL);
> > -	if (!priv)
> > -		return -ENOMEM;
> > -
> >  	priv->mem_res = *mem_res;
> >  	priv->cfg_res = cfg_res;
> >  
> > -- 
> > 2.17.1
> >
Rob Herring Aug. 4, 2020, 4:48 p.m. UTC | #4
On Tue, Aug 4, 2020 at 10:22 AM Bjorn Helgaas <helgaas@kernel.org> wrote:
>
> On Tue, Aug 04, 2020 at 05:13:25PM +0100, Lorenzo Pieralisi wrote:
> > On Tue, Aug 04, 2020 at 02:04:30PM +0200, Geert Uytterhoeven wrote:
> > > The conversion to modern host bridge probing made the driver allocate
> > > its private data using devm_pci_alloc_host_bridge(), but forgot to
> > > remove the old allocation.  Hence part of the driver initialization is
> > > done using the new instance, while another part is done using the old
> > > instance, leading to a crash due to uninitialized bridge DMA ranges:
> > >
> > >     Unable to handle kernel NULL pointer dereference at virtual address 00000008
> > >     pgd = (ptrval)
> > >     [00000008] *pgd=00000000
> > >     Internal error: Oops: 5 [#1] SMP ARM
> > >     CPU: 0 PID: 1 Comm: swapper/0 Not tainted 5.8.0-rc1-shmobile-00035-g92d69cc6275845a7 #645
> > >     Hardware name: Generic R-Car Gen2 (Flattened Device Tree)
> > >     PC is at rcar_pci_probe+0x154/0x340
> > >     LR is at _raw_spin_unlock_irqrestore+0x18/0x20
> > >
> > > Fix this by dropping the old allocation.
> > >
> > > Fixes: 92d69cc6275845a7 ("PCI: rcar-gen2: Convert to use modern host bridge probe functions")
> > > Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
> > > ---
> > >  drivers/pci/controller/pci-rcar-gen2.c | 4 ----
> > >  1 file changed, 4 deletions(-)
> >
> > Squashed in the initial commit, pushed out on pci/misc.
>
> I updated my 'next' branch with this.
>
> Rob, are there any similar issues in other drivers that we should fix
> before asking Linus to pull this?

I'd expect only different issues. :)

This commit is unique as it was using the old arm32 PCI functions and
the most complicated change of the lot.

Rob

Patch
diff mbox series

diff --git a/drivers/pci/controller/pci-rcar-gen2.c b/drivers/pci/controller/pci-rcar-gen2.c
index 046965d284a6d54e..c9530038ca9a53fc 100644
--- a/drivers/pci/controller/pci-rcar-gen2.c
+++ b/drivers/pci/controller/pci-rcar-gen2.c
@@ -302,10 +302,6 @@  static int rcar_pci_probe(struct platform_device *pdev)
 	if (mem_res->start & 0xFFFF)
 		return -EINVAL;
 
-	priv = devm_kzalloc(dev, sizeof(struct rcar_pci_priv), GFP_KERNEL);
-	if (!priv)
-		return -ENOMEM;
-
 	priv->mem_res = *mem_res;
 	priv->cfg_res = cfg_res;