[dpdk-dev] [PATCH] ixgbe: Make vector stores unaligned

[dpdk-dev] [PATCH] ixgbe: Make vector stores unaligned

[Bruce Richardson]

When writing to the mbuf array for receiving packets, do not assume
16-byte alignment by using aligned stores. If the pointers are only
8-byte aligned, the program will crash due to incorrect alignment.
Changing "store" to "storeu" fixes this.

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

diff --git a/lib/librte_pmd_ixgbe/ixgbe_rxtx_vec.c b/lib/librte_pmd_ixgbe/ixgbe_rxtx_vec.c
index 826e085..bafb215 100644
--- a/lib/librte_pmd_ixgbe/ixgbe_rxtx_vec.c
+++ b/lib/librte_pmd_ixgbe/ixgbe_rxtx_vec.c
@@ -256,7 +256,7 @@ ixgbe_recv_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,
 		descs[3] = _mm_loadu_si128((__m128i *)(rxdp + 3));
 
 		/* B.2 copy 2 mbuf point into rx_pkts  */
-		_mm_store_si128((__m128i *)&rx_pkts[pos], mbp1);
+		_mm_storeu_si128((__m128i *)&rx_pkts[pos], mbp1);
 
 		/* B.1 load 1 mbuf point */
 		mbp2 = _mm_loadu_si128((__m128i *)&sw_ring[pos+2]);
@@ -267,7 +267,7 @@ ixgbe_recv_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,
 		descs[0] = _mm_loadu_si128((__m128i *)(rxdp));
 
 		/* B.2 copy 2 mbuf point into rx_pkts  */
-		_mm_store_si128((__m128i *)&rx_pkts[pos+2], mbp2);
+		_mm_storeu_si128((__m128i *)&rx_pkts[pos+2], mbp2);
 
 		/* avoid compiler reorder optimization */
 		rte_compiler_barrier();
-- 
1.9.3

Re: [dpdk-dev] [PATCH] ixgbe: Make vector stores unaligned

[Thomas Monjalon]

> When writing to the mbuf array for receiving packets, do not assume
> 16-byte alignment by using aligned stores. If the pointers are only
> 8-byte aligned, the program will crash due to incorrect alignment.
> Changing "store" to "storeu" fixes this.
> 
> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>

Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>

Applied for version 1.7.1.

Thanks
-- 
Thomas