mISDN: hfcpci: Fix use-after-free bug in hfcpci_Timer

Message ID 20221013125729.105652-1-duoming@zju.edu.cn
State New
Headers
Series mISDN: hfcpci: Fix use-after-free bug in hfcpci_Timer |

Commit Message

Duoming Zhou Oct. 13, 2022, 12:57 p.m. UTC
  If the timer handler hfcpci_Timer() is running, the
del_timer(&hc->hw.timer) in release_io_hfcpci() could
not stop it. As a result, the use-after-free bug will
happen. The process is shown below:

    (cleanup routine)          |        (timer handler)
release_card()                 | hfcpci_Timer()
  release_io_hfcpci            |
    del_timer(&hc->hw.timer)   |
  ...                          |  ...
  kfree(hc) //[1]FREE          |
                               |   hc->hw.timer.expires //[2]USE

The hfc_pci is deallocated in position [1] and used in
position [2].

Fix by changing del_timer() in release_io_hfcpci() to
del_timer_sync(), which makes sure the hfcpci_Timer()
have finished before the hfc_pci is deallocated.

Fixes: 1700fe1a10dc ("Add mISDN HFC PCI driver")
Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
---
 drivers/isdn/hardware/mISDN/hfcpci.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
  

Comments

Greg KH Oct. 13, 2022, 1:20 p.m. UTC | #1
On Thu, Oct 13, 2022 at 08:57:29PM +0800, Duoming Zhou wrote:
> If the timer handler hfcpci_Timer() is running, the
> del_timer(&hc->hw.timer) in release_io_hfcpci() could
> not stop it. As a result, the use-after-free bug will
> happen. The process is shown below:
> 
>     (cleanup routine)          |        (timer handler)
> release_card()                 | hfcpci_Timer()
>   release_io_hfcpci            |
>     del_timer(&hc->hw.timer)   |
>   ...                          |  ...
>   kfree(hc) //[1]FREE          |
>                                |   hc->hw.timer.expires //[2]USE
> 
> The hfc_pci is deallocated in position [1] and used in
> position [2].
> 
> Fix by changing del_timer() in release_io_hfcpci() to
> del_timer_sync(), which makes sure the hfcpci_Timer()
> have finished before the hfc_pci is deallocated.
> 
> Fixes: 1700fe1a10dc ("Add mISDN HFC PCI driver")
> Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
> ---
>  drivers/isdn/hardware/mISDN/hfcpci.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/isdn/hardware/mISDN/hfcpci.c b/drivers/isdn/hardware/mISDN/hfcpci.c
> index af17459c1a5..5cf37fe7de2 100644
> --- a/drivers/isdn/hardware/mISDN/hfcpci.c
> +++ b/drivers/isdn/hardware/mISDN/hfcpci.c
> @@ -157,7 +157,7 @@ release_io_hfcpci(struct hfc_pci *hc)
>  {
>  	/* disable memory mapped ports + busmaster */
>  	pci_write_config_word(hc->pdev, PCI_COMMAND, 0);
> -	del_timer(&hc->hw.timer);
> +	del_timer_sync(&hc->hw.timer);

Nice, how did you test that this will work properly?  Do you have this
hardware for testing?  How was this issue found and verified that this
is the correct resolution?

thanks,

greg k-h
  
Leon Romanovsky Oct. 14, 2022, 6:07 a.m. UTC | #2
On Thu, Oct 13, 2022 at 03:20:05PM +0200, Greg KH wrote:
> On Thu, Oct 13, 2022 at 08:57:29PM +0800, Duoming Zhou wrote:
> > If the timer handler hfcpci_Timer() is running, the
> > del_timer(&hc->hw.timer) in release_io_hfcpci() could
> > not stop it. As a result, the use-after-free bug will
> > happen. The process is shown below:
> > 
> >     (cleanup routine)          |        (timer handler)
> > release_card()                 | hfcpci_Timer()
> >   release_io_hfcpci            |
> >     del_timer(&hc->hw.timer)   |
> >   ...                          |  ...
> >   kfree(hc) //[1]FREE          |
> >                                |   hc->hw.timer.expires //[2]USE
> > 
> > The hfc_pci is deallocated in position [1] and used in
> > position [2].
> > 
> > Fix by changing del_timer() in release_io_hfcpci() to
> > del_timer_sync(), which makes sure the hfcpci_Timer()
> > have finished before the hfc_pci is deallocated.
> > 
> > Fixes: 1700fe1a10dc ("Add mISDN HFC PCI driver")
> > Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
> > ---
> >  drivers/isdn/hardware/mISDN/hfcpci.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/drivers/isdn/hardware/mISDN/hfcpci.c b/drivers/isdn/hardware/mISDN/hfcpci.c
> > index af17459c1a5..5cf37fe7de2 100644
> > --- a/drivers/isdn/hardware/mISDN/hfcpci.c
> > +++ b/drivers/isdn/hardware/mISDN/hfcpci.c
> > @@ -157,7 +157,7 @@ release_io_hfcpci(struct hfc_pci *hc)
> >  {
> >  	/* disable memory mapped ports + busmaster */
> >  	pci_write_config_word(hc->pdev, PCI_COMMAND, 0);
> > -	del_timer(&hc->hw.timer);
> > +	del_timer_sync(&hc->hw.timer);
> 
> Nice, how did you test that this will work properly?  Do you have this
> hardware for testing?  How was this issue found and verified that this
> is the correct resolution?

According to his previous response [1], the answer will be no. I'm not
super-excited that this unmaintained and old driver chosen as playground
for new tool.

[1] https://lore.kernel.org/all/17ad6913.ff8e0.1838933840d.Coremail.duoming@zju.edu.cn/#t

> 
> thanks,
> 
> greg k-h
  
Duoming Zhou Oct. 14, 2022, 6:58 a.m. UTC | #3
Hello,

On Fri, 14 Oct 2022 09:07:07 +0300 Leon Romanovsky wrote:

> On Thu, Oct 13, 2022 at 03:20:05PM +0200, Greg KH wrote:
> > On Thu, Oct 13, 2022 at 08:57:29PM +0800, Duoming Zhou wrote:
> > > If the timer handler hfcpci_Timer() is running, the
> > > del_timer(&hc->hw.timer) in release_io_hfcpci() could
> > > not stop it. As a result, the use-after-free bug will
> > > happen. The process is shown below:
> > > 
> > >     (cleanup routine)          |        (timer handler)
> > > release_card()                 | hfcpci_Timer()
> > >   release_io_hfcpci            |
> > >     del_timer(&hc->hw.timer)   |
> > >   ...                          |  ...
> > >   kfree(hc) //[1]FREE          |
> > >                                |   hc->hw.timer.expires //[2]USE
> > > 
> > > The hfc_pci is deallocated in position [1] and used in
> > > position [2].
> > > 
> > > Fix by changing del_timer() in release_io_hfcpci() to
> > > del_timer_sync(), which makes sure the hfcpci_Timer()
> > > have finished before the hfc_pci is deallocated.
> > > 
> > > Fixes: 1700fe1a10dc ("Add mISDN HFC PCI driver")
> > > Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
> > > ---
> > >  drivers/isdn/hardware/mISDN/hfcpci.c | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > 
> > > diff --git a/drivers/isdn/hardware/mISDN/hfcpci.c b/drivers/isdn/hardware/mISDN/hfcpci.c
> > > index af17459c1a5..5cf37fe7de2 100644
> > > --- a/drivers/isdn/hardware/mISDN/hfcpci.c
> > > +++ b/drivers/isdn/hardware/mISDN/hfcpci.c
> > > @@ -157,7 +157,7 @@ release_io_hfcpci(struct hfc_pci *hc)
> > >  {
> > >  	/* disable memory mapped ports + busmaster */
> > >  	pci_write_config_word(hc->pdev, PCI_COMMAND, 0);
> > > -	del_timer(&hc->hw.timer);
> > > +	del_timer_sync(&hc->hw.timer);
> > 
> > Nice, how did you test that this will work properly?  Do you have this
> > hardware for testing?  How was this issue found and verified that this
> > is the correct resolution?

I am trying to simulate the hardware to verified that this is the correct
resolution. I will give feedback in a few weeks later.

> According to his previous response [1], the answer will be no. I'm not
> super-excited that this unmaintained and old driver chosen as playground
> for new tool.
> 
> [1] https://lore.kernel.org/all/17ad6913.ff8e0.1838933840d.Coremail.duoming@zju.edu.cn/#t

Best regards,
Duoming Zhou
  

Patch

diff --git a/drivers/isdn/hardware/mISDN/hfcpci.c b/drivers/isdn/hardware/mISDN/hfcpci.c
index af17459c1a5..5cf37fe7de2 100644
--- a/drivers/isdn/hardware/mISDN/hfcpci.c
+++ b/drivers/isdn/hardware/mISDN/hfcpci.c
@@ -157,7 +157,7 @@  release_io_hfcpci(struct hfc_pci *hc)
 {
 	/* disable memory mapped ports + busmaster */
 	pci_write_config_word(hc->pdev, PCI_COMMAND, 0);
-	del_timer(&hc->hw.timer);
+	del_timer_sync(&hc->hw.timer);
 	dma_free_coherent(&hc->pdev->dev, 0x8000, hc->hw.fifos,
 			  hc->hw.dmahandle);
 	iounmap(hc->hw.pci_io);