Re: [dpdk-dev] [PATCH v2] mbuf:using sanity checks do not panic on null mbuf

[Olivier Matz <olivier.matz@6wind.com>]

On Tue, Jan 09, 2018 at 08:29:28AM -0600, Keith Wiles wrote:
> The rte_pktmbuf_free() allows for NULL mbuf pointer, but
> when sanity check is enabled it will panic with null pointer.
> 
> Signed-off-by: Keith Wiles <keith.wiles@intel.com>
> ---
>  lib/librte_mbuf/rte_mbuf.c | 10 ++++++++--
>  test/test/test_mbuf.c      |  4 +---
>  2 files changed, 9 insertions(+), 5 deletions(-)
> 
> diff --git a/lib/librte_mbuf/rte_mbuf.c b/lib/librte_mbuf/rte_mbuf.c
> index 7543662f7..621679c92 100644
> --- a/lib/librte_mbuf/rte_mbuf.c
> +++ b/lib/librte_mbuf/rte_mbuf.c
> @@ -205,8 +205,9 @@ rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header)
>  	const struct rte_mbuf *m_seg;
>  	unsigned int nb_segs;
>  
> -	if (m == NULL)
> -		rte_panic("mbuf is NULL\n");
> +	/* Calling with NULL is valid in the API */
> +	if (!m)
> +		return;
>  
>  	/* generic checks */
>  	if (m->pool == NULL)
> @@ -243,6 +244,11 @@ rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len)
>  
>  	__rte_mbuf_sanity_check(m, 1);
>  
> +	if (!m) {
> +		fprintf(stderr, "MBUF pointer is NULL\n");
> +		return;
> +	}
> +
>  	fprintf(f, "dump mbuf at %p, iova=%"PRIx64", buf_len=%u\n",
>  	       m, (uint64_t)m->buf_iova, (unsigned)m->buf_len);
>  	fprintf(f, "  pkt_len=%"PRIu32", ol_flags=%"PRIx64", nb_segs=%u, "
> diff --git a/test/test/test_mbuf.c b/test/test/test_mbuf.c
> index 9e82a20be..146eaf0e5 100644
> --- a/test/test/test_mbuf.c
> +++ b/test/test/test_mbuf.c
> @@ -864,10 +864,8 @@ test_failing_mbuf_sanity_check(struct rte_mempool *pktmbuf_pool)
>  
>  	printf("Now checking for error conditions\n");
>  
> -	if (verify_mbuf_check_panics(NULL)) {
> -		printf("Error with NULL mbuf test\n");
> +	if (verify_mbuf_check_panics(NULL) != -1)
>  		return -1;
> -	}
>  
>  	badbuf = *buf;
>  	badbuf.pool = NULL;
> -- 
> 2.14.1
> 

The problem is a panic when rte_pktmbuf_free(NULL) when mbuf debug is enabled,
right?

A NULL mbuf is only valid in case of a free because it is convenient, but for
most (all ?) other mbuf functions, the mbuf must not be NULL.

What about this patch instead:

  --- a/lib/librte_mbuf/rte_mbuf.h
  +++ b/lib/librte_mbuf/rte_mbuf.h
  @@ -1413,13 +1413,14 @@ rte_pktmbuf_free_seg(struct rte_mbuf *m)
    * segment is added back into its original mempool.
    *
    * @param m
  - *   The packet mbuf to be freed.
  + *   The packet mbuf to be freed. If NULL, the function does nothing.
    */
   static inline void rte_pktmbuf_free(struct rte_mbuf *m)
   {
          struct rte_mbuf *m_next;
  
  -       __rte_mbuf_sanity_check(m, 1);
  +       if (m != NULL)
  +               __rte_mbuf_sanity_check(m, 1);
  
          while (m != NULL) {
                  m_next = m->next;