please help backporting some patches to stable release 21.11.8

[David Marchand]

Hello,

On Sat, Aug 31, 2024 at 6:42 PM Stephen Hemminger
<stephen at networkplumber.org> wrote:
>
> If I build 21.11 with current Gcc there are lot of warnings.
>
>
> ~/DPDK/21.11 $ ninja -C build
> ninja: Entering directory `build'
> [162/2951] Compiling C object lib/librte_hash.a.p/hash_rte_thash.c.o
> In file included from /usr/lib/gcc/x86_64-linux-gnu/14/include/immintrin.h:77,
>                  from /usr/lib/gcc/x86_64-linux-gnu/14/include/x86intrin.h:32,
>                  from ../lib/eal/x86/include/rte_vect.h:31,
>                  from ../lib/eal/x86/include/rte_memcpy.h:17,
>                  from ../lib/mempool/rte_mempool.h:50,
>                  from ../lib/mbuf/rte_mbuf.h:38,
>                  from ../lib/net/rte_ip.h:32,
>                  from ../lib/hash/rte_thash.h:25,
>                  from ../lib/hash/rte_thash.c:7:
> In function ‘_mm512_mask_permutexvar_epi8’,
>     inlined from ‘__rte_thash_gfni’ at ../lib/hash/rte_thash_x86_gfni.h:141:17,
>     inlined from ‘rte_thash_gfni’ at ../lib/hash/rte_thash_x86_gfni.h:181:20,
>     inlined from ‘rte_thash_adjust_tuple’ at ../lib/hash/rte_thash.c:785:11:
> /usr/lib/gcc/x86_64-linux-gnu/14/include/avx512vbmiintrin.h:97:20: warning: ‘tuple_bytes’ may be used uninitialized [-Wmaybe-uninitialized]
>    97 |   return (__m512i) __builtin_ia32_permvarqi512_mask ((__v64qi) __B,
>       |                    ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>    98 |                                                      (__v64qi) __A,
>       |                                                      ~~~~~~~~~~~~~~
>    99 |                                                      (__v64qi) __W,
>       |                                                      ~~~~~~~~~~~~~~
>   100 |                                                      (__mmask64) __M);
>       |                                                      ~~~~~~~~~~~~~~~~
> In file included from ../lib/hash/rte_thash_gfni.h:16,
>                  from ../lib/hash/rte_thash.h:27:
> ../lib/hash/rte_thash_x86_gfni.h: In function ‘rte_thash_adjust_tuple’:
> ../lib/hash/rte_thash_x86_gfni.h:90:33: note: ‘tuple_bytes’ was declared here
>    90 |         __m512i vals, matrixes, tuple_bytes, tuple_bytes_2;
>       |                                 ^~~~~~~~~~~
> In function ‘_mm512_mask_permutexvar_epi8’,
>     inlined from ‘__rte_thash_gfni’ at ../lib/hash/rte_thash_x86_gfni.h:141:17,
>     inlined from ‘rte_thash_gfni’ at ../lib/hash/rte_thash_x86_gfni.h:181:20,
>     inlined from ‘rte_thash_adjust_tuple’ at ../lib/hash/rte_thash.c:785:11:
> /usr/lib/gcc/x86_64-linux-gnu/14/include/avx512vbmiintrin.h:97:20: warning: ‘permute_mask’ may be used uninitialized [-Wmaybe-uninitialized]
>    97 |   return (__m512i) __builtin_ia32_permvarqi512_mask ((__v64qi) __B,
>       |                    ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>    98 |                                                      (__v64qi) __A,
>       |                                                      ~~~~~~~~~~~~~~
>    99 |                                                      (__v64qi) __W,
>       |                                                      ~~~~~~~~~~~~~~
>   100 |                                                      (__mmask64) __M);
>       |                                                      ~~~~~~~~~~~~~~~~
> ../lib/hash/rte_thash_x86_gfni.h: In function ‘rte_thash_adjust_tuple’:
> ../lib/hash/rte_thash_x86_gfni.h:91:30: note: ‘permute_mask’ was declared here
>    91 |         __mmask64 load_mask, permute_mask, permute_mask_2;
>       |                              ^~~~~~~~~~~~

Those above should be fixed in v21.11.4 with 47951ef1dc21 ("hash: fix
GFNI implementation build with GCC 12").
I quickly checked other warnings and I see issues already fixed in
dpdk-stable/21.11.

Please test with latest HEAD of the 21.11 branch.


-- 
David Marchand