DPDK usage discussions
 help / color / mirror / Atom feed
* [dpdk-users] Buffer leak and RSS performance problem
@ 2017-07-01 12:01 lilongfei1110
  0 siblings, 0 replies; only message in thread
From: lilongfei1110 @ 2017-07-01 12:01 UTC (permalink / raw)
  To: users

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset="UTF-8", Size: 1149 bytes --]

    Hi,
    Thanks for your reading.
    I encounter two problems when I use l2fwd and l3fwd. Here are the problems:
    P1. In l2fwd source code,I change these values,and it do not receive packets after receiving about 1K packets.
#define NB_MBUF   81920
#define RTE_TEST_RX_DESC_DEFAULT 4032
    Is it relevant with the i40e driver code below?
" rxq->rx_tail = (uint16_t)(rxq->rx_tail & (rxq->nb_rx_desc - 1));"

    P2. I use l3fwd to process about 35Gps IP packet and packet length is 64B,the process drop many packets. Then I use 4 rx_queues and 4 lcores, but it's performance  improves a little. The same result happens when I use 6 queues and 6 lcores. So how to location where the problem is. Is there any suggestions?
    My environment:
OS:Redhat 6.5 X86_64
CPU:Intel Xeon E5-2650 v3@2.3GHz
NIC:Intel Corporation Device [8086:1580]
DPDK: 16.11


    Thanks & Best Regards.\x16º&’+\x1e•ëýö	šŠW(™&­&éuÛM·×]´×´^¶êç=«a’+\x1e•ëýö	šŠW(™\x17œz+Þuúèšf¢–Wô\x7f ¢ˆ%yÊ&™¨¥•ý\x1fãh(¢	^r‰¶Óß9Û^xÙ¼¥Ù(®\x03è²×âÇ\b­„DŒLø÷]4\x13`@\x15ú+ºÇ«±Ú]’ŠàI«u&évÓ^öÓmõ×í6Ó@„I4^qè¯yÖò™¨¥•ý\x1fãh(¢	^r‰°ŠØR13âu½´îÊ<ã¿tÛ¾å~\rŸ¢»¬z»\x1d¥Ù(®\x04š·MIº]´×½uÛ}uÓ½4<4Ã(ƒ\x12Š	Ú¶êÞ

^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2017-07-01 12:01 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-07-01 12:01 [dpdk-users] Buffer leak and RSS performance problem lilongfei1110

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).