From: <martin_curran-gray@keysight.com>
To: <users@dpdk.org>
Cc: <martin_curran-gray@keysight.com>
Subject: [dpdk-users] X710/i40e and vector pmd rx 2.2.0 in an rx only mode
Date: Tue, 2 Aug 2016 09:53:23 +0000 [thread overview]
Message-ID: <22C95CA62CBADB498D32A348F0F073BC11A1320E@wcosexch02k.cos.is.keysight.com> (raw)
Hi,
So I found my issue with vector mode not being enabled was due to a linking issue in my build environment, it was picking up the stub that returned -1 on the check of conditions, even though I had enabled vector mode in the config before I built the environment.
However now I have vector mode working, I never seem to free any mbufs, so my app works great for about a second, then of course everything coming in gets dropped.
Is there something different I have to do to get my mbufs free'd if the vector rx function is being used rather than the bulk_alloc rx function?
I can run the l2fwd function fine, and it is using vector mode, but
L2fwd output
Initializing port 0... PMD: i40e_dev_rx_queue_setup(): Rx Burst Bulk Alloc Preconditions are satisfied. Rx Burst Bulk Alloc function will be used on port=0, queue=0.
PMD: i40e_set_tx_function_flag(): Vector tx can be enabled on this txq.
PMD: i40e_set_tx_function(): Vector tx finally be used.
PMD: i40e_set_rx_function(): Vector rx enabled, please make sure RX burst size no less than 4 (port=0).
done:
Port 0, MAC address: 68:05:CA:27:F1:BC
My output
PMD: i40e_dev_rx_queue_setup(): Rx Burst Bulk Alloc Preconditions are satisfied. Rx Burst Bulk Alloc function will be used on port=0, queue=0.
PMD: i40e_dev_rx_queue_setup(): Rx Burst Bulk Alloc Preconditions are satisfied. Rx Burst Bulk Alloc function will be used on port=0, queue=1.
PMD: i40e_dev_rx_queue_setup(): Rx Burst Bulk Alloc Preconditions are satisfied. Rx Burst Bulk Alloc function will be used on port=0, queue=2.
PMD: i40e_dev_rx_queue_setup(): Rx Burst Bulk Alloc Preconditions are satisfied. Rx Burst Bulk Alloc function will be used on port=0, queue=3.
PMD: i40e_set_tx_function(): Xmit tx finally be used.
PMD: i40e_pf_config_rss(): Max of contiguous 4 PF queues are configured
PMD: i40e_set_rx_function(): Vector rx enabled, please make sure RX burst size no less than 4 (port=0).
So the code in i40e_rxtx.c seems to imply I should be either using vector or bulk
dev->rx_pkt_burst = i40e_recv_pkts_vec;
or
dev->rx_pkt_burst = i40e_recv_pkts_bulk_alloc;
I seem to be going through this function twice? But so does l2fwd...
however
l2fwd I think has the transmit function free up the buffers after they are sent?
I use explicit calls to rte_mempool_put_bulk to free up buffers I've finished with?
I can see a report in the release notes for 16.04.0 about an issue with ixgbe and releasing mbufs, did/does the i40e suffer from similar problems?
Any suggestions gratefully received
thanks
Martin
martin _ curran - gray @ keysight . com
(remove the spaces)
Martin Curran-Gray
HW/FPGA/SW Engineer
Keysight Technologies UK Ltd
reply other threads:[~2016-08-02 9:53 UTC|newest]
Thread overview: [no followups] expand[flat|nested] mbox.gz Atom feed
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=22C95CA62CBADB498D32A348F0F073BC11A1320E@wcosexch02k.cos.is.keysight.com \
--to=martin_curran-gray@keysight.com \
--cc=users@dpdk.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).