[ldv-project] [PATCH] lag_conf: Added pointer check

[Simon Horman]

On Tue, Nov 15, 2022 at 11:56:37AM +0300, Denis Arefev wrote:
> [You don't often get email from arefev at swemel.ru. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]
> 
> Return value of a function 'kmalloc_array' is dereferenced at lag_conf.c:347
> without checking for null, but it is usually checked for this function.
> 
> Found by Linux Verification Center (linuxtesting.org) with SVACE.
> 
> Signed-off-by: Denis Arefev <arefev at swemel.ru>

Hi Denis,

thanks for highlighting this problem.

> ---
>  drivers/net/ethernet/netronome/nfp/flower/lag_conf.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/net/ethernet/netronome/nfp/flower/lag_conf.c b/drivers/net/ethernet/netronome/nfp/flower/lag_conf.c
> index 63907aeb3884..95ba6e92197d 100644
> --- a/drivers/net/ethernet/netronome/nfp/flower/lag_conf.c
> +++ b/drivers/net/ethernet/netronome/nfp/flower/lag_conf.c
> @@ -276,7 +276,7 @@ static void nfp_fl_lag_do_work(struct work_struct *work)
> 
>         mutex_lock(&lag->lock);
>         list_for_each_entry_safe(entry, storage, &lag->group_list, list) {
> -               struct net_device *iter_netdev, **acti_netdevs;
> +               struct net_device *iter_netdev, **acti_netdevs = NULL;

I don't think it's necessary to set acti_netdevs here as
it is always set before use by the call to kmalloc_array().

>                 struct nfp_flower_repr_priv *repr_priv;
>                 int active_count = 0, slaves = 0;
>                 struct nfp_repr *repr;
> @@ -308,6 +308,8 @@ static void nfp_fl_lag_do_work(struct work_struct *work)
> 
>                 acti_netdevs = kmalloc_array(entry->slave_cnt,
>                                              sizeof(*acti_netdevs), GFP_KERNEL);
> +               if (!acti_netdevs)
> +                break;

The indentation here doesn't look right.


Regarding the problem at hand, yes, I agree that it seems
that kmalloc_array() should be checked. But I am concerned that
simply break'ing here may lead to a bad state. And I'd like to ask
for some time to examine this more closely.

>                 /* Include sanity check in the loop. It may be that a bond has
>                  * changed between processing the last notification and the
> --
> 2.25.1
>