On Fri, Oct 08, 2021 at 08:31:29AM +0100, Colin Ian King wrote:
> On 08/10/2021 06:58, Dan Carpenter wrote:
> > On Fri, Oct 08, 2021 at 01:15:58AM +0100, Colin King wrote:
> > > From: Colin Ian King <colin.king@xxxxxxxxxxxxx>
> > >
> > > There is an error return path where the error return is being
> > > assigned to err rather than count and the error exit path does
> > > not return -EAGAIN as expected. Fix this by setting the error
> > > return to variable count as this is the value that is returned
> > > at the end of the function.
> > >
> > > Addresses-Coverity: ("Unused value")
> > > Fixes: 00c4da27a421 ("carl9170: firmware parser and debugfs code")
> > > Signed-off-by: Colin Ian King <colin.king@xxxxxxxxxxxxx>
> > > ---
> > > drivers/net/wireless/ath/carl9170/debug.c | 2 +-
> > > 1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/net/wireless/ath/carl9170/debug.c b/drivers/net/wireless/ath/carl9170/debug.c
> > > index bb40889d7c72..f163c6bdac8f 100644
> > > --- a/drivers/net/wireless/ath/carl9170/debug.c
> > > +++ b/drivers/net/wireless/ath/carl9170/debug.c
> > > @@ -628,7 +628,7 @@ static ssize_t carl9170_debugfs_bug_write(struct ar9170 *ar, const char *buf,
> > > case 'R':
> > > if (!IS_STARTED(ar)) {
> > > - err = -EAGAIN;
> > > + count = -EAGAIN;
> > > goto out;
> >
> > This is ugly. The bug wouldn't have happened with a direct return, it's
> > only the goto out which causes it. Better to replace all the error
> > paths with direct returns. There are two other direct returns so it's
> > not like a new thing...
>
> Yep, I agree it was ugly, I was trying to keep to the coding style and
> reduce the patch delta size. I can do a V2 if the maintainers deem it's a
> cleaner solution.
>
> >
> > Goto out on the success path is fine here, though.
>
> Yep. I believe that a goto to one exit return point (may possibly?) make the
> code smaller rather than a sprinkling of returns in a function, so I'm never
> sure if this is a win or not with these kind of cases.
>
My default position is to hate goto out labels, but today I'm thinking
maybe they kind of make sense for read/write "return count;" functions...
This is debugfs code and debugfs code is always the worst code in any
driver...
The other style anti-patten here is to have two variables which store
error codes. Maybe it would be better to do something like below. Or
your patch is also find... Let's just go with that. It's fine.
regards,
dan carpenter
diff --git a/drivers/net/wireless/ath/carl9170/debug.c b/drivers/net/wireless/ath/carl9170/debug.c
index bb40889d7c72..8ccd0d3bea4c 100644
--- a/drivers/net/wireless/ath/carl9170/debug.c
+++ b/drivers/net/wireless/ath/carl9170/debug.c
@@ -616,7 +616,7 @@ DEBUGFS_DECLARE_RW_FILE(hw_ioread32, CARL9170_DEBUG_RING_SIZE * 40);
static ssize_t carl9170_debugfs_bug_write(struct ar9170 *ar, const char *buf,
size_t count)
{
- int err;
+ int err = 0;
if (count < 1)
return -EINVAL;
@@ -638,7 +638,7 @@ static ssize_t carl9170_debugfs_bug_write(struct ar9170 *ar, const char *buf,
case 'M':
err = carl9170_mac_reset(ar);
if (err < 0)
- count = err;
+ goto out;
goto out;
@@ -646,7 +646,7 @@ static ssize_t carl9170_debugfs_bug_write(struct ar9170 *ar, const char *buf,
err = carl9170_set_channel(ar, ar->hw->conf.chandef.chan,
cfg80211_get_chandef_type(&ar->hw->conf.chandef));
if (err < 0)
- count = err;
+ goto out;
goto out;
@@ -657,6 +657,8 @@ static ssize_t carl9170_debugfs_bug_write(struct ar9170 *ar, const char *buf,
carl9170_restart(ar, CARL9170_RR_USER_REQUEST);
out:
+ if (err < 0)
+ return err;
return count;
}
