Re: [dpdk-dev] [PATCH v2] eal: fix use-after-free issue on thread creation

[Olivier Matz]

Hi Jianfeng,

On Wed, May 02, 2018 at 10:17:50AM +0000, Jianfeng Tan wrote:
> After below commit, we encounter some strange issue:
>   1) Dead lock as described here:
>      http://dpdk.org/ml/archives/dev/2018-April/099806.html
>   2) SIGSEGV issue when starting a testpmd in VM.
> 
> Considering below commit changes to use dynamic memory instead of
> stack for memory barrier, we doubt it's caused by use-after-free.
> 
> Fixes: 3d09a6e26d8b ("eal: fix threads block on barrier")
> 
> Reported-by: Maxime Coquelin <maxime.coque...@redhat.com>
> Reported-by: Lei Yao <lei.a....@intel.com>
> Suggested-by: Stephen Hemminger <step...@networkplumber.org>
> Signed-off-by: Jianfeng Tan <jianfeng....@intel.com>
> ---
>  v1->v2:
>  - Destroy barrier if failure happens.
>  lib/librte_eal/common/eal_common_thread.c | 15 ++++++++++++---
>  1 file changed, 12 insertions(+), 3 deletions(-)
> 
> diff --git a/lib/librte_eal/common/eal_common_thread.c 
> b/lib/librte_eal/common/eal_common_thread.c
> index de69452..5f0c61f 100644
> --- a/lib/librte_eal/common/eal_common_thread.c
> +++ b/lib/librte_eal/common/eal_common_thread.c
> @@ -149,11 +149,16 @@ struct rte_thread_ctrl_params {
>  
>  static void *rte_thread_init(void *arg)
>  {
> +     int ret;
>       struct rte_thread_ctrl_params *params = arg;
>       void *(*start_routine)(void *) = params->start_routine;
>       void *routine_arg = params->arg;
>  
> -     pthread_barrier_wait(&params->configured);
> +     ret = pthread_barrier_wait(&params->configured);
> +     if (ret == PTHREAD_BARRIER_SERIAL_THREAD) {
> +             pthread_barrier_destroy(&params->configured);
> +             free(params);
> +     }
>  
>       return start_routine(routine_arg);
>  }
> @@ -204,12 +209,16 @@ rte_ctrl_thread_create(pthread_t *thread, const char 
> *name,
>       if (ret < 0)
>               goto fail;
>  
> -     pthread_barrier_wait(&params->configured);
> -     free(params);
> +     ret = pthread_barrier_wait(&params->configured);
> +     if (ret == PTHREAD_BARRIER_SERIAL_THREAD) {
> +             pthread_barrier_destroy(&params->configured);
> +             free(params);
> +     }
>  
>       return 0;
>  
>  fail:
> +     pthread_barrier_destroy(&params->configured);

I think we should have the same code than above in the fail case:

        ret = pthread_barrier_wait(&params->configured);
        if (ret == PTHREAD_BARRIER_SERIAL_THREAD) {
                pthread_barrier_destroy(&params->configured);
                free(params);
        }

Else, the child will wait forever on the barrier on failure.

This can be tested with this standalone program:
https://www.droids-corp.org/~zer0/hidden/ctrl_thread.c

gcc -W -Wall -Werror -Wextra -pthread ctrl_thread.c
./a.out -> fail

gcc -W -Wall -Werror -Wextra -pthread -DFIX ctrl_thread.c
./a.out -> ok



>       pthread_cancel(*thread);
>       pthread_join(*thread, NULL);
>       free(params);
> -- 
> 2.7.4
>