2019-12-15 17:59:42

by Aditya Pakki

Subject: [PATCH] hdlcdrv: replace assertion with recovery code

In hdlcdrv_register, failure to register the driver causes a crash.
However, by returning the error to the caller in case ops is NULL
can avoid the crash. The patch fixes this issue.

Signed-off-by: Aditya Pakki <[email protected]>
---
drivers/net/hamradio/hdlcdrv.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/net/hamradio/hdlcdrv.c b/drivers/net/hamradio/hdlcdrv.c
index df495b5595f5..38e5d1e54800 100644
--- a/drivers/net/hamradio/hdlcdrv.c
+++ b/drivers/net/hamradio/hdlcdrv.c
@@ -687,7 +687,8 @@ struct net_device *hdlcdrv_register(const struct hdlcdrv_ops *ops,
struct hdlcdrv_state *s;
int err;

- BUG_ON(ops == NULL);
+ if (!ops)
+ return ERR_PTR(-EINVAL);

if (privsize < sizeof(struct hdlcdrv_state))
privsize = sizeof(struct hdlcdrv_state);
--
2.20.1

2019-12-15 18:51:23

by Markus Elfring

Subject: Re: [PATCH] hdlcdrv: replace assertion with recovery code

> However, by returning the error to the caller in case ops is NULL
> can avoid the crash.

I suggest to improve this change description by choosing also
a more imperative wording.


> The patch fixes this issue.
Please replace this sentence by the tag “Fixes”.

Regards,
Markus

2019-12-15 19:21:19

by Stephen Hemminger

Subject: Re: [PATCH] hdlcdrv: replace assertion with recovery code

On Sun, 15 Dec 2019 11:58:41 -0600
Aditya Pakki <[email protected]> wrote:

> In hdlcdrv_register, failure to register the driver causes a crash.
> However, by returning the error to the caller in case ops is NULL
> can avoid the crash. The patch fixes this issue.
>
> Signed-off-by: Aditya Pakki <[email protected]>
> ---
> drivers/net/hamradio/hdlcdrv.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/hamradio/hdlcdrv.c b/drivers/net/hamradio/hdlcdrv.c
> index df495b5595f5..38e5d1e54800 100644
> --- a/drivers/net/hamradio/hdlcdrv.c
> +++ b/drivers/net/hamradio/hdlcdrv.c
> @@ -687,7 +687,8 @@ struct net_device *hdlcdrv_register(const struct hdlcdrv_ops *ops,
> struct hdlcdrv_state *s;h
> int err;
>
> - BUG_ON(ops == NULL);
> + if (!ops)
> + return ERR_PTR(-EINVAL);
>
> if (privsize < sizeof(struct hdlcdrv_state))
> privsize = sizeof(struct hdlcdrv_state);

It is good to remove BUG_ON's but this is not a good way to fix it.
The original code was being over paranoid. There are only 3 places
this function is called in the current kernel and all pass a valid
pointer. Better just remove the BUG_ON all together; it is not
worth carrying bug checks for "some day somebody might add broken code".