[PATCH] hash: fix maybe-uninitialized warnings on build

[PATCH] hash: fix maybe-uninitialized warnings on build

[Bruce Richardson]

Following recent changes to the test-meson-build.sh script[1] to use more
build-types warnings are seen in the hash library GFNI code about values
possibly being uninitialized.

In function ‘_mm512_mask_permutexvar_epi8’,
    inlined from ‘__rte_thash_gfni’ at ../lib/hash/rte_thash_x86_gfni.h:150:6:
/usr/lib/gcc/x86_64-linux-gnu/15/include/avx512vbmiintrin.h:97:20: error: ‘tuple_bytes_2’ may be used uninitialized [-Werror=maybe-uninitialized]

These all appear to be a false positive, but fix these by explicitly
zero-initializing the offending values.

[1] 177acaf5cf75 ("devtools: test different build types")

Fixes: 4fd8c4cb0de1 ("hash: add new Toeplitz hash implementation")
Cc: stable@dpdk.org

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
 lib/hash/rte_thash_x86_gfni.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/lib/hash/rte_thash_x86_gfni.h b/lib/hash/rte_thash_x86_gfni.h
index 99226b7592..69973503fd 100644
--- a/lib/hash/rte_thash_x86_gfni.h
+++ b/lib/hash/rte_thash_x86_gfni.h
@@ -89,9 +89,9 @@ __rte_thash_gfni(const uint64_t *mtrx, const uint8_t *tuple,
 	const __m512i shift_8 = _mm512_set1_epi8(8);
 	__m512i xor_acc = _mm512_setzero_si512();
 	__m512i perm_bytes = _mm512_setzero_si512();
-	__m512i vals, matrixes, tuple_bytes_2;
+	__m512i vals, matrixes, tuple_bytes_2 = _mm512_setzero_si512();
 	__m512i tuple_bytes = _mm512_setzero_si512();
-	__mmask64 load_mask, permute_mask_2;
+	__mmask64 load_mask, permute_mask_2 = 0;
 	__mmask64 permute_mask = 0;
 	int chunk_len = 0, i = 0;
 	uint8_t mtrx_msk;
-- 
2.51.0

Re: [PATCH] hash: fix maybe-uninitialized warnings on build

[Medvedkin, Vladimir]

Acked-by: Vladimir Medvedkin <vladimir.medvedkin@intel.com>

On 2/12/2026 4:27 PM, Bruce Richardson wrote:
> Following recent changes to the test-meson-build.sh script[1] to use more
> build-types warnings are seen in the hash library GFNI code about values
> possibly being uninitialized.
>
> In function ‘_mm512_mask_permutexvar_epi8’,
>      inlined from ‘__rte_thash_gfni’ at ../lib/hash/rte_thash_x86_gfni.h:150:6:
> /usr/lib/gcc/x86_64-linux-gnu/15/include/avx512vbmiintrin.h:97:20: error: ‘tuple_bytes_2’ may be used uninitialized [-Werror=maybe-uninitialized]
>
> These all appear to be a false positive, but fix these by explicitly
> zero-initializing the offending values.
>
> [1] 177acaf5cf75 ("devtools: test different build types")
>
> Fixes: 4fd8c4cb0de1 ("hash: add new Toeplitz hash implementation")
> Cc: stable@dpdk.org
>
> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> ---
>   lib/hash/rte_thash_x86_gfni.h | 4 ++--
>   1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/lib/hash/rte_thash_x86_gfni.h b/lib/hash/rte_thash_x86_gfni.h
> index 99226b7592..69973503fd 100644
> --- a/lib/hash/rte_thash_x86_gfni.h
> +++ b/lib/hash/rte_thash_x86_gfni.h
> @@ -89,9 +89,9 @@ __rte_thash_gfni(const uint64_t *mtrx, const uint8_t *tuple,
>   	const __m512i shift_8 = _mm512_set1_epi8(8);
>   	__m512i xor_acc = _mm512_setzero_si512();
>   	__m512i perm_bytes = _mm512_setzero_si512();
> -	__m512i vals, matrixes, tuple_bytes_2;
> +	__m512i vals, matrixes, tuple_bytes_2 = _mm512_setzero_si512();
>   	__m512i tuple_bytes = _mm512_setzero_si512();
> -	__mmask64 load_mask, permute_mask_2;
> +	__mmask64 load_mask, permute_mask_2 = 0;
>   	__mmask64 permute_mask = 0;
>   	int chunk_len = 0, i = 0;
>   	uint8_t mtrx_msk;

-- 
Regards,
Vladimir

Re: [PATCH] hash: fix maybe-uninitialized warnings on build

[David Marchand]

On Thu, 12 Feb 2026 at 17:27, Bruce Richardson
<bruce.richardson@intel.com> wrote:
>
> Following recent changes to the test-meson-build.sh script[1] to use more
> build-types warnings are seen in the hash library GFNI code about values
> possibly being uninitialized.
>
> In function ‘_mm512_mask_permutexvar_epi8’,
>     inlined from ‘__rte_thash_gfni’ at ../lib/hash/rte_thash_x86_gfni.h:150:6:
> /usr/lib/gcc/x86_64-linux-gnu/15/include/avx512vbmiintrin.h:97:20: error: ‘tuple_bytes_2’ may be used uninitialized [-Werror=maybe-uninitialized]
>
> These all appear to be a false positive, but fix these by explicitly
> zero-initializing the offending values.
>
> [1] 177acaf5cf75 ("devtools: test different build types")
>
> Fixes: 4fd8c4cb0de1 ("hash: add new Toeplitz hash implementation")
> Cc: stable@dpdk.org
>
> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>

Too bad Intel CI did not catch it.

Acked-by: David Marchand <david.marchand@redhat.com>


-- 
David Marchand

Re: [PATCH] hash: fix maybe-uninitialized warnings on build

[Bruce Richardson]

On Thu, Feb 12, 2026 at 05:54:22PM +0100, David Marchand wrote:
> On Thu, 12 Feb 2026 at 17:27, Bruce Richardson
> <bruce.richardson@intel.com> wrote:
> >
> > Following recent changes to the test-meson-build.sh script[1] to use more
> > build-types warnings are seen in the hash library GFNI code about values
> > possibly being uninitialized.
> >
> > In function ‘_mm512_mask_permutexvar_epi8’,
> >     inlined from ‘__rte_thash_gfni’ at ../lib/hash/rte_thash_x86_gfni.h:150:6:
> > /usr/lib/gcc/x86_64-linux-gnu/15/include/avx512vbmiintrin.h:97:20: error: ‘tuple_bytes_2’ may be used uninitialized [-Werror=maybe-uninitialized]
> >
> > These all appear to be a false positive, but fix these by explicitly
> > zero-initializing the offending values.
> >
> > [1] 177acaf5cf75 ("devtools: test different build types")
> >
> > Fixes: 4fd8c4cb0de1 ("hash: add new Toeplitz hash implementation")
> > Cc: stable@dpdk.org
> >
> > Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> 
> Too bad Intel CI did not catch it.
> 
> Acked-by: David Marchand <david.marchand@redhat.com>
> 
I don't think any of the CIs use the test-meson-build scripts, do they?
It's more a convenience for maintainers.

Re: [PATCH] hash: fix maybe-uninitialized warnings on build

[Thomas Monjalon]

12/02/2026 17:54, David Marchand:
> On Thu, 12 Feb 2026 at 17:27, Bruce Richardson
> <bruce.richardson@intel.com> wrote:
> >
> > Following recent changes to the test-meson-build.sh script[1] to use more
> > build-types warnings are seen in the hash library GFNI code about values
> > possibly being uninitialized.
> >
> > In function ‘_mm512_mask_permutexvar_epi8’,
> >     inlined from ‘__rte_thash_gfni’ at ../lib/hash/rte_thash_x86_gfni.h:150:6:
> > /usr/lib/gcc/x86_64-linux-gnu/15/include/avx512vbmiintrin.h:97:20: error: ‘tuple_bytes_2’ may be used uninitialized [-Werror=maybe-uninitialized]
> >
> > These all appear to be a false positive, but fix these by explicitly
> > zero-initializing the offending values.
> >
> > [1] 177acaf5cf75 ("devtools: test different build types")
> >
> > Fixes: 4fd8c4cb0de1 ("hash: add new Toeplitz hash implementation")
> > Cc: stable@dpdk.org
> >
> > Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> 
> Too bad Intel CI did not catch it.
> 
> Acked-by: David Marchand <david.marchand@redhat.com>

Applied, thanks.