container_of() only returns NULL if the passed pointer is NULL _and_
the embedded element is the first element of the structure. Even if that
is the case, testing against it is misleading and possibly dangerous
because the position of the embedded element may change. Explicitly
check if the parameter is NULL and bail out if so instead of checking
the result of container_of().
Signed-off-by: Guenter Roeck <[email protected]>
---
RFC:
The NULL check in the poll function is likely unnecessary. Interestingly,
there is no NULL check in the ioctl function, even though there is a
similar container_of() in that function. However, I do not feel
comfortable enough to change the functionality of this code and drop
the check entirely.
drivers/misc/xilinx_sdfec.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/drivers/misc/xilinx_sdfec.c b/drivers/misc/xilinx_sdfec.c
index 23c8448a9c3b..0a3721d31dea 100644
--- a/drivers/misc/xilinx_sdfec.c
+++ b/drivers/misc/xilinx_sdfec.c
@@ -1011,11 +1011,11 @@ static __poll_t xsdfec_poll(struct file *file, poll_table *wait)
__poll_t mask = 0;
struct xsdfec_dev *xsdfec;
- xsdfec = container_of(file->private_data, struct xsdfec_dev, miscdev);
-
- if (!xsdfec)
+ if (!file->private_data)
return EPOLLNVAL | EPOLLHUP;
+ xsdfec = container_of(file->private_data, struct xsdfec_dev, miscdev);
+
poll_wait(file, &xsdfec->waitq, wait);
/* XSDFEC ISR detected an error */
--
2.25.1
On Thu, May 20, 2021 at 10:01:50AM -0700, Guenter Roeck wrote:
> container_of() only returns NULL if the passed pointer is NULL _and_
> the embedded element is the first element of the structure. Even if that
> is the case, testing against it is misleading and possibly dangerous
> because the position of the embedded element may change. Explicitly
> check if the parameter is NULL and bail out if so instead of checking
> the result of container_of().
>
> Signed-off-by: Guenter Roeck <[email protected]>
> ---
> RFC:
>
> The NULL check in the poll function is likely unnecessary. Interestingly,
> there is no NULL check in the ioctl function, even though there is a
> similar container_of() in that function. However, I do not feel
> comfortable enough to change the functionality of this code and drop
> the check entirely.
>
> drivers/misc/xilinx_sdfec.c | 6 +++---
> 1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/misc/xilinx_sdfec.c b/drivers/misc/xilinx_sdfec.c
> index 23c8448a9c3b..0a3721d31dea 100644
> --- a/drivers/misc/xilinx_sdfec.c
> +++ b/drivers/misc/xilinx_sdfec.c
> @@ -1011,11 +1011,11 @@ static __poll_t xsdfec_poll(struct file *file, poll_table *wait)
> __poll_t mask = 0;
> struct xsdfec_dev *xsdfec;
>
> - xsdfec = container_of(file->private_data, struct xsdfec_dev, miscdev);
> -
> - if (!xsdfec)
> + if (!file->private_data)
> return EPOLLNVAL | EPOLLHUP;
>
> + xsdfec = container_of(file->private_data, struct xsdfec_dev, miscdev);
> +
> poll_wait(file, &xsdfec->waitq, wait);
>
> /* XSDFEC ISR detected an error */
> --
> 2.25.1
>
It should be safe not to check this for NULL as the misc device sets the
pointer in the open, and removes it in release. poll or ioctl can not
be called if release has already happened.
So feel free to drop the check here, xsdfec_dev_ioctl() looks correct.
thanks,
greg k-h
> -----Original Message-----
> From: Greg Kroah-Hartman <[email protected]>
> Sent: Friday 21 May 2021 19:40
> To: Guenter Roeck <[email protected]>
> Cc: Derek Kiernan <[email protected]>; Dragan Cvetic <[email protected]>; Arnd Bergmann <[email protected]>; Michal Simek
> <[email protected]>; [email protected]; [email protected]
> Subject: Re: [RFC PATCH] misc: xilinx-sdfec: Check if file->private_data is NULL
>
> On Thu, May 20, 2021 at 10:01:50AM -0700, Guenter Roeck wrote:
> > container_of() only returns NULL if the passed pointer is NULL _and_
> > the embedded element is the first element of the structure. Even if that
> > is the case, testing against it is misleading and possibly dangerous
> > because the position of the embedded element may change. Explicitly
> > check if the parameter is NULL and bail out if so instead of checking
> > the result of container_of().
> >
> > Signed-off-by: Guenter Roeck <[email protected]>
> > ---
> > RFC:
> >
> > The NULL check in the poll function is likely unnecessary. Interestingly,
> > there is no NULL check in the ioctl function, even though there is a
> > similar container_of() in that function. However, I do not feel
> > comfortable enough to change the functionality of this code and drop
> > the check entirely.
> >
> > drivers/misc/xilinx_sdfec.c | 6 +++---
> > 1 file changed, 3 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/misc/xilinx_sdfec.c b/drivers/misc/xilinx_sdfec.c
> > index 23c8448a9c3b..0a3721d31dea 100644
> > --- a/drivers/misc/xilinx_sdfec.c
> > +++ b/drivers/misc/xilinx_sdfec.c
> > @@ -1011,11 +1011,11 @@ static __poll_t xsdfec_poll(struct file *file, poll_table *wait)
> > __poll_t mask = 0;
> > struct xsdfec_dev *xsdfec;
> >
> > - xsdfec = container_of(file->private_data, struct xsdfec_dev, miscdev);
> > -
> > - if (!xsdfec)
> > + if (!file->private_data)
> > return EPOLLNVAL | EPOLLHUP;
> >
> > + xsdfec = container_of(file->private_data, struct xsdfec_dev, miscdev);
> > +
> > poll_wait(file, &xsdfec->waitq, wait);
> >
> > /* XSDFEC ISR detected an error */
> > --
> > 2.25.1
> >
>
>
> It should be safe not to check this for NULL as the misc device sets the
> pointer in the open, and removes it in release. poll or ioctl can not
> be called if release has already happened.
>
> So feel free to drop the check here, xsdfec_dev_ioctl() looks correct.
>
> thanks,
>
> greg k-h
Acked-by: Dragan Cvetic <[email protected]>