On Saturday 10 October 2020 14:40:34 CEST Greg Kroah-Hartman wrote:
> On Sat, Oct 10, 2020 at 02:22:13PM +0200, Jérôme Pouiller wrote:
> > On Friday 9 October 2020 20:52:47 CEST Kalle Valo wrote:
> > > Jerome Pouiller <Jerome.Pouiller@xxxxxxxxxx> writes:
> > >
> > > > From: Jérôme Pouiller <jerome.pouiller@xxxxxxxxxx>
> > > >
> > > > Smatch complains:
> > > >
> > > > drivers/staging/wfx/hif_rx.c:177 hif_scan_complete_indication() warn: potential NULL parameter dereference 'wvif'
> > > > drivers/staging/wfx/data_tx.c:576 wfx_flush() warn: potential NULL parameter dereference 'wvif'
> > > >
> > > > Indeed, if the vif id returned by the device does not exist anymore,
> > > > wdev_to_wvif() could return NULL.
> > > >
> > > > In add, the error is not handled uniformly in the code, sometime a
> > > > WARN() is displayed but code continue, sometime a dev_warn() is
> > > > displayed, sometime it is just not tested, ...
> > > >
> > > > This patch standardize that.
> > > >
> > > > Reported-by: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
> > > > Signed-off-by: Jérôme Pouiller <jerome.pouiller@xxxxxxxxxx>
> > > > ---
> > > > drivers/staging/wfx/data_tx.c | 5 ++++-
> > > > drivers/staging/wfx/hif_rx.c | 34 ++++++++++++++++++++++++----------
> > > > drivers/staging/wfx/sta.c | 4 ++++
> > > > 3 files changed, 32 insertions(+), 11 deletions(-)
> > > >
> > > > diff --git a/drivers/staging/wfx/data_tx.c b/drivers/staging/wfx/data_tx.c
> > > > index b4d5dd3d2d23..8db0be08daf8 100644
> > > > --- a/drivers/staging/wfx/data_tx.c
> > > > +++ b/drivers/staging/wfx/data_tx.c
> > > > @@ -431,7 +431,10 @@ static void wfx_skb_dtor(struct wfx_vif *wvif, struct sk_buff *skb)
> > > > sizeof(struct hif_req_tx) +
> > > > req->fc_offset;
> > > >
> > > > - WARN_ON(!wvif);
> > > > + if (!wvif) {
> > > > + pr_warn("%s: vif associated with the skb does not exist anymore\n", __func__);
> > > > + return;
> > > > + }
> > >
> > > I'm not really a fan of using function names in warning or error
> > > messages as it clutters the log. In debug messages I think they are ok.
> >
> > In the initial code, I used WARN() that far more clutters the log (I
> > have stated that a backtrace won't provide any useful information, so
> > pr_warn() was better suited).
> >
> > In add, in my mind, these warnings are debug messages. If they appears,
> > the user should probably report a bug.
> >
> > Finally, in this patch, I use the same message several times (ok, not
> > this particular one). So the function name is a way to differentiate
> > them.
>
> You should use dev_*() for these, that way you can properly determine
> the exact device as well.
Totally agree. I initially did that. However, the device is a field of
wvif which is NULL in this case.
I could have changed the code to get the real pointer to the device. But
I didn't want to clutter the code just for a debug message (and also
because I was a bit lazy).
--
Jérôme Pouiller
