On Mon, Mar 6, 2023 at 7:52 AM Zhuobin Huang <zobin1...@gmail.com> wrote:
>
> Assign correct data length to each segments according
> to the given pkt_len and nb_pkt_segs, instead of
> using pkt_len as the data_len of every packet segment
>
> Fixes: a9c9e9698d5e ("app/test: allow to create packets of different sizes")
> Cc: cunming.li...@intel.com
>
> Signed-off-by: Zhuobin Huang <zobin1...@gmail.com>
Reviewed-by: David Marchand <david.march...@redhat.com>
Good catch.
I am curious to know how you noticed the issue since I see no in-tree
code asking for multiseg.
For the record, to check this patch, I compiled with
-DRTE_LIBRTE_MBUF_DEBUG and a small modification:
diff --git a/app/test/test_pmd_perf.c b/app/test/test_pmd_perf.c
index ff84d251ff..5c670ff165 100644
--- a/app/test/test_pmd_perf.c
+++ b/app/test/test_pmd_perf.c
@@ -209,7 +209,7 @@ init_traffic(struct rte_mempool *mp,
return generate_packet_burst(mp, pkts_burst, &pkt_eth_hdr,
0, &pkt_ipv4_hdr, 1,
&pkt_udp_hdr, burst_size,
- PACKET_BURST_GEN_PKT_LEN, 1);
+ PACKET_BURST_GEN_PKT_LEN, 2);
}
static int
Before the fix:
DPDK_TEST=pmd_perf_autotest ./build-gcc/app/test/dpdk-test --no-huge
-m 2048 --vdev=net_null0 --vdev=net_null1
...
Generate 4096 packets @socket 0
inject 2048 packet to port 0
PANIC in rte_mbuf_sanity_check():
bad pkt_len
Cc: ci@ in general, we should have more tests run with debug/check
options enabled.
--
David Marchand
