2019-12-15 18:13:48

by Aditya Pakki

Subject: [PATCH] staging: kpc2000: replace assertion with recovery code

In kpc_dma_transfer, if either priv or ldev is NULL, crashing the
process is excessive and is not needed. Instead of asserting, by
passing the error upstream, the error can be handled.

Signed-off-by: Aditya Pakki <[email protected]>
---
drivers/staging/kpc2000/kpc_dma/fileops.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/staging/kpc2000/kpc_dma/fileops.c b/drivers/staging/kpc2000/kpc_dma/fileops.c
index cb52bd9a6d2f..1c4633267cc1 100644
--- a/drivers/staging/kpc2000/kpc_dma/fileops.c
+++ b/drivers/staging/kpc2000/kpc_dma/fileops.c
@@ -49,9 +49,11 @@ static int kpc_dma_transfer(struct dev_private_data *priv,
u64 dma_addr;
u64 user_ctl;

- BUG_ON(priv == NULL);
+ if (!priv)
+ return -EINVAL;
ldev = priv->ldev;
- BUG_ON(ldev == NULL);
+ if (!ldev)
+ return -EINVAL;

acd = kzalloc(sizeof(*acd), GFP_KERNEL);
if (!acd) {
--
2.20.1

2019-12-15 18:31:13

by Greg KH

Subject: Re: [PATCH] staging: kpc2000: replace assertion with recovery code

On Sun, Dec 15, 2019 at 12:12:37PM -0600, Aditya Pakki wrote:
> In kpc_dma_transfer, if either priv or ldev is NULL, crashing the
> process is excessive and is not needed. Instead of asserting, by
> passing the error upstream, the error can be handled.
>
> Signed-off-by: Aditya Pakki <[email protected]>
> ---
> drivers/staging/kpc2000/kpc_dma/fileops.c | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/staging/kpc2000/kpc_dma/fileops.c b/drivers/staging/kpc2000/kpc_dma/fileops.c
> index cb52bd9a6d2f..1c4633267cc1 100644
> --- a/drivers/staging/kpc2000/kpc_dma/fileops.c
> +++ b/drivers/staging/kpc2000/kpc_dma/fileops.c
> @@ -49,9 +49,11 @@ static int kpc_dma_transfer(struct dev_private_data *priv,
> u64 dma_addr;
> u64 user_ctl;
>
> - BUG_ON(priv == NULL);
> + if (!priv)
> + return -EINVAL;

How can prive ever be NULL here? Can you track back to all callers to
verify this? If so, just remove the check.

thanks,

greg k-h

2020-01-03 08:52:08

by Dan Carpenter

Subject: Re: [PATCH] staging: kpc2000: replace assertion with recovery code

On Sun, Dec 15, 2019 at 07:28:14PM +0100, Greg Kroah-Hartman wrote:
> On Sun, Dec 15, 2019 at 12:12:37PM -0600, Aditya Pakki wrote:
> > In kpc_dma_transfer, if either priv or ldev is NULL, crashing the
> > process is excessive and is not needed. Instead of asserting, by
> > passing the error upstream, the error can be handled.
> >
> > Signed-off-by: Aditya Pakki <[email protected]>
> > ---
> > drivers/staging/kpc2000/kpc_dma/fileops.c | 6 ++++--
> > 1 file changed, 4 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/staging/kpc2000/kpc_dma/fileops.c b/drivers/staging/kpc2000/kpc_dma/fileops.c
> > index cb52bd9a6d2f..1c4633267cc1 100644
> > --- a/drivers/staging/kpc2000/kpc_dma/fileops.c
> > +++ b/drivers/staging/kpc2000/kpc_dma/fileops.c
> > @@ -49,9 +49,11 @@ static int kpc_dma_transfer(struct dev_private_data *priv,
> > u64 dma_addr;
> > u64 user_ctl;
> >
> > - BUG_ON(priv == NULL);
> > + if (!priv)
> > + return -EINVAL;
>
> How can prive ever be NULL here? Can you track back to all callers to
> verify this? If so, just remove the check.
>

Smatch says that "priv" can't be NULL.

Also if you have a NULL dereference the kernel prints a nice stack
trace. Normally just doing the NULL dereference and Oopsing is better
than doing a BUG_ON(). The one exception would be when the Oops leads
to filesystem corruption.

regards,
dan carpenter