On 21.08.2019 13:38, Konstantin Khorenko wrote:
> As there are still possible cases when we expect code to work without
> statistic inited, it's theoretically possible some day we again get an
> update request for non-inited stat struct, and corrupt stack-protector
> cookie and who knows what alse.
> 
> So add a sanity check to prevent it.
> 
> https://jira.sw.ru/browse/PSBM-96128
> 
> Signed-off-by: Konstantin Khorenko <[email protected]>

Reviewed-by: Kirill Tkhai <[email protected]>

> ---
>  fs/fuse/kio/pcs/fuse_stat.c | 11 +++++++++++
>  1 file changed, 11 insertions(+)
> 
> diff --git a/fs/fuse/kio/pcs/fuse_stat.c b/fs/fuse/kio/pcs/fuse_stat.c
> index b57be607f5e1..c91bfe8b3bf7 100644
> --- a/fs/fuse/kio/pcs/fuse_stat.c
> +++ b/fs/fuse/kio/pcs/fuse_stat.c
> @@ -31,6 +31,12 @@ static u64 lat_oreder_list[] = {
>  
>  static inline void fuse_val_stat_update(struct fuse_val_stat *s, u64 val)
>  {
> +     /*
> +      * Sanity check. In case s == NULL, stack protector cookie is
> +      * corrupted, quite difficult to debug.
> +      */
> +     BUG_ON(!s);
> +
>       preempt_disable();
>       if (!__this_cpu_read(s->events)) {
>               __this_cpu_write(s->val_min, val);
> @@ -48,6 +54,11 @@ void fuse_latency_update(struct fuse_lat_stat *s, u64 val)
>  {
>       int i;
>  
> +     /*
> +      * Sanity check. In case s == NULL, stack protector cookie is
> +      * corrupted, quite difficult to debug.
> +      */
> +     BUG_ON(!s);
>       this_cpu_inc(s->count);
>       for (i = LAT_ORDER1; i <= LAT_ORDER5; i++) {
>               if (likely(val <= lat_oreder_list[i])) {
> 

_______________________________________________
Devel mailing list
[email protected]
https://lists.openvz.org/mailman/listinfo/devel

Reply via email to