Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1222338 > unrolled thread

Re: [PATCH] usb: gadget: amd5536udc: fix NULL pointer dereference

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: [PATCH] usb: gadget: amd5536udc: fix NULL pointer dereference David Cohen <david.a.cohen@linux.intel.com> - 2015-09-10 20:10 +0200
    Re: [PATCH] usb: gadget: amd5536udc: fix NULL pointer dereference Sudip Mukherjee <sudipm.mukherjee@gmail.com> - 2015-09-11 12:10 +0200
      Re: [PATCH] usb: gadget: amd5536udc: fix NULL pointer dereference Felipe Balbi <balbi@ti.com> - 2015-09-11 15:30 +0200
        Re: [PATCH] usb: gadget: amd5536udc: fix NULL pointer dereference Sudip Mukherjee <sudipm.mukherjee@gmail.com> - 2015-09-11 16:30 +0200
          Re: [PATCH] usb: gadget: amd5536udc: fix NULL pointer dereference Felipe Balbi <balbi@ti.com> - 2015-09-11 17:10 +0200

#1222338 — Re: [PATCH] usb: gadget: amd5536udc: fix NULL pointer dereference

Hi Sudip,

On Fri, Sep 04, 2015 at 05:12:23PM +0530, Sudip Mukherjee wrote:
> We were checking if dev->regs is NULL but it was done after
> dereferencing it. Lets reset the controller and iounmap dev->regs only
> if it is not NULL.
> free_irq() does not need dev->regs, so unmaping it before freeing the
> irq should not matter.
> 
> Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
> ---
>  drivers/usb/gadget/udc/amd5536udc.c | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/usb/gadget/udc/amd5536udc.c b/drivers/usb/gadget/udc/amd5536udc.c
> index fdacddb..26066d3 100644
> --- a/drivers/usb/gadget/udc/amd5536udc.c
> +++ b/drivers/usb/gadget/udc/amd5536udc.c
> @@ -3135,11 +3135,12 @@ static void udc_pci_remove(struct pci_dev *pdev)
>  	}
>  
>  	/* reset controller */
> -	writel(AMD_BIT(UDC_DEVCFG_SOFTRESET), &dev->regs->cfg);
> +	if (dev->regs) {
> +		writel(AMD_BIT(UDC_DEVCFG_SOFTRESET), &dev->regs->cfg);
> +		iounmap(dev->regs);

I'm not familiar with the driver, but you're iounmap'ing before freeing
irq. Looks fishy to me.

Br, David

> +	}
>  	if (dev->irq_registered)
>  		free_irq(pdev->irq, dev);
> -	if (dev->regs)
> -		iounmap(dev->regs);
>  	if (dev->mem_region)
>  		release_mem_region(pci_resource_start(pdev, 0),
>  				pci_resource_len(pdev, 0));
> -- 
> 1.9.1
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-usb" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [next] | [standalone]

#1222676

On Thu, Sep 10, 2015 at 11:03:34AM -0700, David Cohen wrote:
> Hi Sudip,
> 
> On Fri, Sep 04, 2015 at 05:12:23PM +0530, Sudip Mukherjee wrote:
> > We were checking if dev->regs is NULL but it was done after
> > dereferencing it. Lets reset the controller and iounmap dev->regs only
> > if it is not NULL.
> > free_irq() does not need dev->regs, so unmaping it before freeing the
> > irq should not matter.
> > 
> > Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
> > ---
> >  drivers/usb/gadget/udc/amd5536udc.c | 7 ++++---
> >  1 file changed, 4 insertions(+), 3 deletions(-)
> > 
> > diff --git a/drivers/usb/gadget/udc/amd5536udc.c b/drivers/usb/gadget/udc/amd5536udc.c
> > index fdacddb..26066d3 100644
> > --- a/drivers/usb/gadget/udc/amd5536udc.c
> > +++ b/drivers/usb/gadget/udc/amd5536udc.c
> > @@ -3135,11 +3135,12 @@ static void udc_pci_remove(struct pci_dev *pdev)
<snip>
> 
> I'm not familiar with the driver, but you're iounmap'ing before freeing
> irq. Looks fishy to me.
Well, I thought you will be able to give me some idea about how fix it. :)
Then I guess we should be on the safe side and what about the following:


diff --git a/drivers/usb/gadget/udc/amd5536udc.c b/drivers/usb/gadget/udc/amd5536udc.c
index fdacddb..82f36f6 100644
--- a/drivers/usb/gadget/udc/amd5536udc.c
+++ b/drivers/usb/gadget/udc/amd5536udc.c
@@ -3134,8 +3134,9 @@ static void udc_pci_remove(struct pci_dev *pdev)
 		pci_pool_destroy(dev->stp_requests);
 	}
 
-	/* reset controller */
-	writel(AMD_BIT(UDC_DEVCFG_SOFTRESET), &dev->regs->cfg);
+	if (dev->regs)
+		/* reset controller */
+		writel(AMD_BIT(UDC_DEVCFG_SOFTRESET), &dev->regs->cfg);
 	if (dev->irq_registered)
 		free_irq(pdev->irq, dev);
 	if (dev->regs)

And just for my information: for a device what might happen if I iounmap
before I free the irq? One thing I can think of is that after iounmap
just at that moment one interrupt comes and the driver tries to access
the io memory while servicing the irq.

regards
sudip
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [next] | [standalone]

#1222839

[Multipart message — attachments visible in raw view] — view raw

Hi,

On Fri, Sep 11, 2015 at 03:32:56PM +0530, Sudip Mukherjee wrote:
> On Thu, Sep 10, 2015 at 11:03:34AM -0700, David Cohen wrote:
> > Hi Sudip,
> > 
> > On Fri, Sep 04, 2015 at 05:12:23PM +0530, Sudip Mukherjee wrote:
> > > We were checking if dev->regs is NULL but it was done after
> > > dereferencing it. Lets reset the controller and iounmap dev->regs only
> > > if it is not NULL.
> > > free_irq() does not need dev->regs, so unmaping it before freeing the
> > > irq should not matter.
> > > 
> > > Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
> > > ---
> > >  drivers/usb/gadget/udc/amd5536udc.c | 7 ++++---
> > >  1 file changed, 4 insertions(+), 3 deletions(-)
> > > 
> > > diff --git a/drivers/usb/gadget/udc/amd5536udc.c b/drivers/usb/gadget/udc/amd5536udc.c
> > > index fdacddb..26066d3 100644
> > > --- a/drivers/usb/gadget/udc/amd5536udc.c
> > > +++ b/drivers/usb/gadget/udc/amd5536udc.c
> > > @@ -3135,11 +3135,12 @@ static void udc_pci_remove(struct pci_dev *pdev)
> <snip>
> > 
> > I'm not familiar with the driver, but you're iounmap'ing before freeing
> > irq. Looks fishy to me.
> Well, I thought you will be able to give me some idea about how fix it. :)
> Then I guess we should be on the safe side and what about the following:
> 
> 
> diff --git a/drivers/usb/gadget/udc/amd5536udc.c b/drivers/usb/gadget/udc/amd5536udc.c
> index fdacddb..82f36f6 100644
> --- a/drivers/usb/gadget/udc/amd5536udc.c
> +++ b/drivers/usb/gadget/udc/amd5536udc.c
> @@ -3134,8 +3134,9 @@ static void udc_pci_remove(struct pci_dev *pdev)
>  		pci_pool_destroy(dev->stp_requests);
>  	}
>  
> -	/* reset controller */
> -	writel(AMD_BIT(UDC_DEVCFG_SOFTRESET), &dev->regs->cfg);
> +	if (dev->regs)
> +		/* reset controller */
> +		writel(AMD_BIT(UDC_DEVCFG_SOFTRESET), &dev->regs->cfg);

no, the check is pointless. Most of these are. Just look at your probe()
and you'll see that if dev->virt_addr is NULL (meaning ioremap_nocache()
failed) you exit from probe() with error. The driver doesn't probe at
all. So you can be sure that by remove, dev->regs is valid.

BTW, if probe fails, you have a TON of leaked resources!! You don't kfree()
dev, you don't pci_disable_device(), you don't release_mem_region(), you
don't iounmap() virt_addr, you don't free_irq().

Also the iounmap() call in remove is wrong. You ioremapped
dev->virt_addr but iounmap() dev->regs. Are you SURE that's ok ?

Man, what a mess! You gotta fix that up.

>  	if (dev->irq_registered)
>  		free_irq(pdev->irq, dev);
>  	if (dev->regs)
> 
> And just for my information: for a device what might happen if I iounmap
> before I free the irq? One thing I can think of is that after iounmap

what happens if an IRQ fires after you iounmap() but before you
free_irq() ?

> just at that moment one interrupt comes and the driver tries to access
> the io memory while servicing the irq.

there you go

-- 
balbi

[toc] | [prev] | [next] | [standalone]

#1222882

On Fri, Sep 11, 2015 at 08:28:34AM -0500, Felipe Balbi wrote:
> Hi,
> 
> On Fri, Sep 11, 2015 at 03:32:56PM +0530, Sudip Mukherjee wrote:
> > On Thu, Sep 10, 2015 at 11:03:34AM -0700, David Cohen wrote:
<snip>
> no, the check is pointless. Most of these are. Just look at your probe()
> and you'll see that if dev->virt_addr is NULL (meaning ioremap_nocache()
> failed) you exit from probe() with error. The driver doesn't probe at
> all. So you can be sure that by remove, dev->regs is valid.
> 
> BTW, if probe fails, you have a TON of leaked resources!! You don't kfree()
> dev, you don't pci_disable_device(), you don't release_mem_region(), you
> don't iounmap() virt_addr, you don't free_irq().
> 
> Also the iounmap() call in remove is wrong. You ioremapped
> dev->virt_addr but iounmap() dev->regs. Are you SURE that's ok ?
> 
> Man, what a mess! You gotta fix that up.
:(
Yes, total mess. I am on it.
BTW, while I am fixing it can i also include a patch which will
rearrange the functions and remove the forward declarations?

regards
sudip
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [next] | [standalone]

#1222904

[Multipart message — attachments visible in raw view] — view raw

On Fri, Sep 11, 2015 at 07:51:01PM +0530, Sudip Mukherjee wrote:
> On Fri, Sep 11, 2015 at 08:28:34AM -0500, Felipe Balbi wrote:
> > Hi,
> > 
> > On Fri, Sep 11, 2015 at 03:32:56PM +0530, Sudip Mukherjee wrote:
> > > On Thu, Sep 10, 2015 at 11:03:34AM -0700, David Cohen wrote:
> <snip>
> > no, the check is pointless. Most of these are. Just look at your probe()
> > and you'll see that if dev->virt_addr is NULL (meaning ioremap_nocache()
> > failed) you exit from probe() with error. The driver doesn't probe at
> > all. So you can be sure that by remove, dev->regs is valid.
> > 
> > BTW, if probe fails, you have a TON of leaked resources!! You don't kfree()
> > dev, you don't pci_disable_device(), you don't release_mem_region(), you
> > don't iounmap() virt_addr, you don't free_irq().
> > 
> > Also the iounmap() call in remove is wrong. You ioremapped
> > dev->virt_addr but iounmap() dev->regs. Are you SURE that's ok ?
> > 
> > Man, what a mess! You gotta fix that up.
> :(
> Yes, total mess. I am on it.
> BTW, while I am fixing it can i also include a patch which will
> rearrange the functions and remove the forward declarations?

sure, just make sure fixes and cleanups are separate. First all fixes,
then cleanups and reorderings.

cheers

-- 
balbi

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web