Re: [dpdk-dev] [PATCH 2/3] crypto: fix pedentic compilation errors

[Neil Horman <nhorman@tuxdriver.com>]

On Wed, Nov 22, 2017 at 09:10:17AM +0100, Nelio Laranjeiro wrote:
>  /root/dpdk/x86_64-native-linuxapp-gcc/include/rte_crypto.h:126:28: error: ISO C forbids zero-size array ‘sym’ [-Werror=pedantic]
>    struct rte_crypto_sym_op sym[0];
>                             ^~~
> 
> Fixes: d2a4223c4c6d ("cryptodev: do not store pointer to op specific params")
> Cc: pablo.de.lara.guarch@intel.com
> Cc: stable@dpdk.org
> 
> Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
> ---
>  lib/librte_cryptodev/rte_crypto.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/lib/librte_cryptodev/rte_crypto.h b/lib/librte_cryptodev/rte_crypto.h
> index 3d672fe7d..dc6e91d1d 100644
> --- a/lib/librte_cryptodev/rte_crypto.h
> +++ b/lib/librte_cryptodev/rte_crypto.h
> @@ -123,7 +123,7 @@ struct rte_crypto_op {
>  
>  	RTE_STD_C11
>  	union {
> -		struct rte_crypto_sym_op sym[0];
> +		struct rte_crypto_sym_op *sym;
>  		/**< Symmetric operation parameters */
>  	}; /**< operation specific parameters */
>  };
> -- 
> 2.11.0
> 
> 
As Laura notes, this isn't the right solution.  In addition to adding a 64 bit
pointer, it I think also results in incorrect semantics.  That is to say, the
allocation path for this structure allocates the rte_crypto_op and additional
memory for the sym array contiguously, which the sym[0] syntax correctly
interprets to mean the storage for the array is inline with the structure.
Changing to a pointer means you are using the first elements of the array
storage as your pointer, which could be filled with any old value, leading to
corruption.

If you can't use zero length array semantics (which I assume you cant, as I
don't think clang supports that), a better soution might be to remove the sym
variable entirely, and replace it with a macro that access the sym array as an
offset from the start of the pointer.  That would seem to be an ABI change, but
if you went through that process you would wind up with the same sized struct,
which would be nice

Neil