[v2] b43legacy: Add checking for null for ssb_get_devtypedata(dev)

  Since second call of ssb_get_devtypedata() may fail as well as the
first one, the NULL return value in 'wl' will be later dereferenced in
calls to b43legacy_one_core_attach() and schedule_work().

Instead of merely warning about this failure with
B43legacy_WARN_ON(), properly return with error to avoid any further
NULL pointer dereferences.

Found by Linux Verification Center (linuxtesting.org) with static
analysis tool SVACE.

Fixes: 75388acd0cd8 ("[B43LEGACY]: add mac80211-based driver for legacy BCM43xx devices")
Co-developed-by: Natalia Petrova <n.petrova@fintech.ru>
Signed-off-by: Nikita Zhandarovich <n.zhandarovich@fintech.ru>
---
v2: fix issues with overlooked null-ptr-dereferences pointed out by
Simon Horman <simon.horman@corigine.com>
Link: https://lore.kernel.org/all/Y+eb9mZntfe6rO3v@corigine.com/ 

 drivers/net/wireless/broadcom/b43legacy/main.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
  

On Tue, 18 Apr 2023 07:29:18 -0700
Nikita Zhandarovich <n.zhandarovich@fintech.ru> wrote:

> Since second call of ssb_get_devtypedata() may fail as well as the
> first one, the NULL return value in 'wl' will be later dereferenced in
> calls to b43legacy_one_core_attach() and schedule_work().

No, the second call can't fail, because b43legacy_wireless_init() will
always initialize it before returning 0.

> a/drivers/net/wireless/broadcom/b43legacy/main.c +++
> b/drivers/net/wireless/broadcom/b43legacy/main.c @@ -3857,7 +3857,11
> @@ static int b43legacy_probe(struct ssb_device *dev, if (err)
>  			goto out;
>  		wl = ssb_get_devtypedata(dev);
> -		B43legacy_WARN_ON(!wl);
> +		if (!wl) {
> +			B43legacy_WARN_ON(!wl);
> +			err = -ENODEV;
> +			goto out;

And the 'goto out' would be the wrong error recovery path, too.

> +		}
>  	}
>  	err = b43legacy_one_core_attach(dev, wl);
>  	if (err)


Nack.
Please drop this patch. The code is correct as-is.
  

On 4/18/23 09:29, Nikita Zhandarovich wrote:
> Since second call of ssb_get_devtypedata() may fail as well as the
> first one, the NULL return value in 'wl' will be later dereferenced in
> calls to b43legacy_one_core_attach() and schedule_work().
> 
> Instead of merely warning about this failure with
> B43legacy_WARN_ON(), properly return with error to avoid any further
> NULL pointer dereferences.
> 
> Found by Linux Verification Center (linuxtesting.org) with static
> analysis tool SVACE.
> 
> Fixes: 75388acd0cd8 ("[B43LEGACY]: add mac80211-based driver for legacy BCM43xx devices")
> Co-developed-by: Natalia Petrova <n.petrova@fintech.ru>
> Signed-off-by: Nikita Zhandarovich <n.zhandarovich@fintech.ru>
> ---
> v2: fix issues with overlooked null-ptr-dereferences pointed out by
> Simon Horman <simon.horman@corigine.com>
> Link: https://lore.kernel.org/all/Y+eb9mZntfe6rO3v@corigine.com/
> 
>   drivers/net/wireless/broadcom/b43legacy/main.c | 6 +++++-
>   1 file changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/net/wireless/broadcom/b43legacy/main.c b/drivers/net/wireless/broadcom/b43legacy/main.c
> index 760136638a95..5a706dd0b1a4 100644
> --- a/drivers/net/wireless/broadcom/b43legacy/main.c
> +++ b/drivers/net/wireless/broadcom/b43legacy/main.c
> @@ -3857,7 +3857,11 @@ static int b43legacy_probe(struct ssb_device *dev,
>   		if (err)
>   			goto out;
>   		wl = ssb_get_devtypedata(dev);
> -		B43legacy_WARN_ON(!wl);
> +		if (!wl) {
> +			B43legacy_WARN_ON(!wl);
> +			err = -ENODEV;
> +			goto out;
> +		}
>   	}
>   	err = b43legacy_one_core_attach(dev, wl);
>   	if (err)

I do not recall seeing v1. One additional nitpick: The latest convention would 
have the subject as "wifi: b43legacy:...". Kalle may be able to fix this on 
merging, but it not, a v3 might be required. Otherwise, the patch is good.

Reviewed-by: Larry Finger <Larry.Finger@lwfinger.net>

Thanks,

Larry
  

On Fri, 21 Apr 2023 17:14:18 -0500
Larry Finger <Larry.Finger@lwfinger.net> wrote:
> > (err) goto out;
> >   		wl = ssb_get_devtypedata(dev);
> > -		B43legacy_WARN_ON(!wl);
> > +		if (!wl) {
> > +			B43legacy_WARN_ON(!wl);
> > +			err = -ENODEV;
> > +			goto out;
> > +		}
> >   	}
> >   	err = b43legacy_one_core_attach(dev, wl);
> >   	if (err)  
> 
> I do not recall seeing v1. One additional nitpick: The latest
> convention would have the subject as "wifi: b43legacy:...". Kalle may
> be able to fix this on merging, but it not, a v3 might be required.
> Otherwise, the patch is good.
> 
> Reviewed-by: Larry Finger <Larry.Finger@lwfinger.net>

No, it's not good. It's wrong. I already replied to it.
wl can never be NULL here and the goto-out path is wrong (if there
was a chance for it to trigger).

Please drop this patch, Kalle.
  

Nikita Zhandarovich <n.zhandarovich@fintech.ru> wrote:

> Since second call of ssb_get_devtypedata() may fail as well as the
> first one, the NULL return value in 'wl' will be later dereferenced in
> calls to b43legacy_one_core_attach() and schedule_work().
> 
> Instead of merely warning about this failure with
> B43legacy_WARN_ON(), properly return with error to avoid any further
> NULL pointer dereferences.
> 
> Found by Linux Verification Center (linuxtesting.org) with static
> analysis tool SVACE.
> 
> Fixes: 75388acd0cd8 ("[B43LEGACY]: add mac80211-based driver for legacy BCM43xx devices")
> Co-developed-by: Natalia Petrova <n.petrova@fintech.ru>
> Signed-off-by: Nikita Zhandarovich <n.zhandarovich@fintech.ru>
> Reviewed-by: Larry Finger <Larry.Finger@lwfinger.net>

Dropped per Michael's request.

Patch set to Rejected.