[8/9] net: nb8800: avoid uninitialized variable warning

The nb8800_poll() function initializes the 'next' variable in the
loop looking for new input data. We know this will be called at
least once because 'budget' is a guaranteed to be a positive number
when we enter the function, but the compiler doesn't know that
and warns when the variable is used later:

drivers/net/ethernet/aurora/nb8800.c: In function 'nb8800_poll':
drivers/net/ethernet/aurora/nb8800.c:350:21: warning: 'next' may be used uninitialized in this function [-Wmaybe-uninitialized]

Changing the 'while() {}' loop to 'do {} while()' makes it obvious
to the compiler what is going on so it no longer warns.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
---
 drivers/net/ethernet/aurora/nb8800.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

Arnd Bergmann <arnd@arndb.de> writes:

> The nb8800_poll() function initializes the 'next' variable in the
> loop looking for new input data. We know this will be called at
> least once because 'budget' is a guaranteed to be a positive number
> when we enter the function, but the compiler doesn't know that
> and warns when the variable is used later:
>
> drivers/net/ethernet/aurora/nb8800.c: In function 'nb8800_poll':
> drivers/net/ethernet/aurora/nb8800.c:350:21: warning: 'next' may be used uninitialized in this function [-Wmaybe-uninitialized]

Which gcc version is this?  4.9 doesn't warn here, presumably because
it's clever enough to notice that the offending use of 'next' is under a
condition that can only be true if the first one was.  Of course fixing
the code so older compilers don't warn is a good idea.

> Changing the 'while() {}' loop to 'do {} while()' makes it obvious
> to the compiler what is going on so it no longer warns.
>
> Signed-off-by: Arnd Bergmann <arnd@arndb.de>

Acked-by: Mans Rullgard <mans@mansr.com>

> ---
>  drivers/net/ethernet/aurora/nb8800.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/net/ethernet/aurora/nb8800.c b/drivers/net/ethernet/aurora/nb8800.c
> index ecc4a334c507..f71ab2647a3b 100644
> --- a/drivers/net/ethernet/aurora/nb8800.c
> +++ b/drivers/net/ethernet/aurora/nb8800.c
> @@ -302,7 +302,7 @@ static int nb8800_poll(struct napi_struct *napi, int budget)
>  	nb8800_tx_done(dev);
>
>  again:
> -	while (work < budget) {
> +	do {
>  		struct nb8800_rx_buf *rxb;
>  		unsigned int len;
>
> @@ -330,7 +330,7 @@ again:
>  		rxd->report = 0;
>  		last = next;
>  		work++;
> -	}
> +	} while (work < budget);
>
>  	if (work) {
>  		priv->rx_descs[last].desc.config |= DESC_EOC;
> -- 
> 2.7.0
>

On Wednesday 27 January 2016 14:13:16 Måns Rullgård wrote:
> Arnd Bergmann <arnd@arndb.de> writes:
> 
> > The nb8800_poll() function initializes the 'next' variable in the
> > loop looking for new input data. We know this will be called at
> > least once because 'budget' is a guaranteed to be a positive number
> > when we enter the function, but the compiler doesn't know that
> > and warns when the variable is used later:
> >
> > drivers/net/ethernet/aurora/nb8800.c: In function 'nb8800_poll':
> > drivers/net/ethernet/aurora/nb8800.c:350:21: warning: 'next' may be used uninitialized in this function [-Wmaybe-uninitialized]
> 
> Which gcc version is this?  4.9 doesn't warn here, presumably because
> it's clever enough to notice that the offending use of 'next' is under a
> condition that can only be true if the first one was.  Of course fixing
> the code so older compilers don't warn is a good idea.

This was with gcc-5.2.1, but possibly in an unusual kernel configuration.
The only time I see it in my logs was with CONFIG_ARCH_IXP4XX=y, which
has its own mach/io.h and other headers that sometimes override generic
functions with a more elaborate version.

I have another patch for ixp4xx that simplifies its mach/io.h in order to
get rid of other false 'uninitialized use' warnings, but this one still
shows up with that other patch.

> > Changing the 'while() {}' loop to 'do {} while()' makes it obvious
> > to the compiler what is going on so it no longer warns.
> >
> > Signed-off-by: Arnd Bergmann <arnd@arndb.de>
> 
> Acked-by: Mans Rullgard <mans@mansr.com>

Thanks,

	Arnd