From: "Gábor LENCSE" <lencse@hit.bme.hu>
To: "users@dpdk.org" <users@dpdk.org>
Subject: Re: dpdk-testpmd fails with a Mellanox ConnectX-4 Lx NIC
Date: Mon, 24 Feb 2025 22:32:40 +0100 [thread overview]
Message-ID: <e9a97143-ed9e-407b-98be-80057da18cda@hit.bme.hu> (raw)
In-Reply-To: <20250224233720.04fb1bae@sovereign>
[-- Attachment #1: Type: text/plain, Size: 2592 bytes --]
Hi Dmitry,
2025. 02. 24. 21:37 keltezéssel, Dmitry Kozlyuk írta:
> Mellanox NICs need mlx5_core kernel driver even when used via DPDK.
Indeed, it works! Thank you very much! I would have never thought of that.
However, it seems to be rather slow and it loses frames.
With my siitperf ( https://github.com/lencsegabor/siitperf ) measurement
program I can achieve 7.1Mfps per direction using bidirectional traffic
with 64-byte frame size when I use X540. And the bottleneck is surely
the X540, as the numbers are the same when I use fixed IP addresses and
port numbers and when any or both of them varies. With X710, I can
achieve more that 8Mfps using RFC 4814 pseudorandom port numbers and
more than 10Mfps using fixed frames (using bidirectional traffic with
64-byte frame size).
However the first step of the binary search of the throughput test using
8Mfps rate lasted more than 120s (instead of 60s).
What is worse, the binary search counts down to 0 due to the loss of a
small amount of packets.
Here is the current output:
--------------------------------------------------- 2025-02-24
22:26:18.165766809 Iteration no. 1-8
--------------------------------------------------- Testing rate: 31250
fps. EAL: Detected 16 lcore(s) EAL: Detected 2 NUMA nodes EAL: Detected
shared linkage of DPDK EAL: Multi-process socket
/var/run/dpdk/rte/mp_socket EAL: Selected IOVA mode 'VA' EAL: Probing
VFIO support... EAL: VFIO support initialized EAL: Probe PCI driver:
mlx5_pci (15b3:1015) device: 0000:05:00.0 (socket 0) EAL: Probe PCI
driver: mlx5_pci (15b3:1015) device: 0000:05:00.1 (socket 0) EAL: Probe
PCI driver: net_mlx4 (15b3:1007) device: 0000:82:00.0 (socket 1) EAL: No
legacy callbacks, legacy socket not created Info: Left port and Left
Sender CPU core belong to the same NUMA node: 0 Info: Right port and
Right Receiver CPU core belong to the same NUMA node: 0 Info: Right port
and Right Sender CPU core belong to the same NUMA node: 0 Info: Left
port and Left Receiver CPU core belong to the same NUMA node: 0 Info:
Testing initiated at 2025-02-24 22:26:19 Info: Forward sender's sending
took 59.9999682437 seconds. Forward frames sent: 1875000 Info: Reverse
sender's sending took 59.9999682300 seconds. Reverse frames sent:
1875000 Reverse frames received: 1874963 Forward frames received:
1874958 Info: Test finished. Forward: 1874958 frames were received from
the required 1875000 frames Reverse: 1874963 frames were received from
the required 1875000 frames TEST FAILED
Do you have any idea, what is happening here?
Best regards,
Gábor
[-- Attachment #2: Type: text/html, Size: 3815 bytes --]
prev parent reply other threads:[~2025-02-24 21:33 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-02-24 20:19 Gábor LENCSE
2025-02-24 20:37 ` Dmitry Kozlyuk
2025-02-24 21:32 ` Gábor LENCSE [this message]
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=e9a97143-ed9e-407b-98be-80057da18cda@hit.bme.hu \
--to=lencse@hit.bme.hu \
--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).