"No new RX buf available" error in sample Echo and HttpServer applications when running on FRDM_N947 boards

[ekwus]

Describe the bug

We've been struggling to get the NXP FRDM_N947 boards to run an http server with any reliability.

We've switched to the recent 4.2.00-rc1 tag as nothing works without the pull requests #91312 and #90181

Regardless we end up with RX buffer issues when payloads are greater than 131 bytes.

Regression

• This is a regression.

Steps to reproduce

To recreate the issue we simply build the unmodified Echo or HttpServer sample apps and run them on a board (only statis IP address config has changed to our network).

For the HttpServer, simply loading the sample app page starts with the uptime and network statistics being updated every seconds until it eventually freezes anywhere from 20 seconds to 4 mins. Once it stops, the console logging is printing out that it failed to allocate an Rx buffer.

For Echo, again using the latest codebase with only the IP address changed. We connect to the listening 4242 port with a C# client that is configurable to send different payload sizes in a repeated loop with controllable delay between the sends.

It generally fails with "No new RX buf available" after a few hundred sends on a single connection when the payload is greater than 131 bytes. Anything above 131 bytes fails reasonably quickly and anything below tends to work until I stop testing.

I've tried setting the TCP Window values to 500, 1000, 1500 and played with the network buffer size but those settings don't help any.

Example config changes

CONFIG_NET_BUF_DATA_SIZE=2048
CONFIG_NET_TCP_MAX_SEND_WINDOW_SIZE=1500
CONFIG_NET_TCP_MAX_RECV_WINDOW_SIZE=1500

Relevant log output

IPV6_V6ONLY option is on, turning it off.
Sharing same socket between IPv6 and IPv4
Single-threaded TCP echo server waits for a connection on port 4242...
[00:00:00.602,000] <err> phy_mii: Failed to reset PHY (0): -116
*** Booting Zephyr OS build 4bd1d39c67df ***
[00:00:00.603,000] <inf> net_config: Initializing network
[00:00:00.603,000] <inf> net_config: IPv4 address: 10.80.80.80
[00:00:00.703,000] <inf> net_config: IPv6 address: 2001:db8::1
[00:00:00.703,000] <inf> net_config: IPv6 address: 2001:db8::1
[00:00:00.703,000] <dbg> net_sock: zsock_socket_internal: (0x30001a50): socket: ctx=0x300020d0, fd=3
[00:00:44.359,000] <err> eth_nxp_enet_qos_mac: Rx descriptor does not have first descriptor flag, drop
[00:00:44.359,000] <err> eth_nxp_enet_qos_mac: Rx descriptor does not have first descriptor flag, drop
[00:00:44.359,000] <err> eth_nxp_enet_qos_mac: Rx descriptor does not have first descriptor flag, drop
...
...
[00:01:28.493,000] <dbg> net_sock: zsock_received_cb: (0x30001748): ctx=0x30002178, pkt=0x3000860c, st=0, user_data=0
[00:01:28.538,000] <dbg> net_sock: zsock_received_cb: (0x30001748): ctx=0x30002178, pkt=0x300085c8, st=0, user_data=0
[00:01:28.539,000] <dbg> net_sock: zsock_received_cb: (0x30001748): ctx=0x30002178, pkt=0x3000860c, st=0, user_data=0
[00:01:28.540,000] <dbg> net_sock: zsock_received_cb: (0x30001748): ctx=0x30002178, pkt=0x300085c8, st=0, user_data=0
[00:01:28.541,000] <dbg> net_sock: zsock_received_cb: (0x30001748): ctx=0x30002178, pkt=0x3000860c, st=0, user_data=0
[00:01:28.541,000] <dbg> net_sock: zsock_received_cb: (0x30001748): ctx=0x30002178, pkt=0x300085c8, st=0, user_data=0
[00:01:28.542,000] <dbg> net_sock: zsock_received_cb: (0x30001748): ctx=0x30002178, pkt=0x3000860c, st=0, user_data=0
[00:01:28.543,000] <dbg> net_sock: zsock_received_cb: (0x30001748): ctx=0x30002178, pkt=0x300085c8, st=0, user_data=0
[00:01:28.544,000] <dbg> net_sock: zsock_received_cb: (0x30001748): ctx=0x30002178, pkt=0x3000860c, st=0, user_data=0
[00:01:28.547,000] <dbg> net_sock: zsock_received_cb: (0x30001748): ctx=0x30002178, pkt=0x300085c8, st=0, user_data=0
[00:01:28.547,000] <dbg> net_sock: zsock_received_cb: (0x30001748): ctx=0x30002178, pkt=0x30008694, st=0, user_data=0
[00:01:28.548,000] <dbg> net_sock: zsock_received_cb: (0x30001748): ctx=0x30002178, pkt=0x300085c8, st=0, user_data=0
[00:01:28.549,000] <dbg> net_sock: zsock_received_cb: (0x30001748): ctx=0x30002178, pkt=0x30008694, st=0, user_data=0
[00:01:28.550,000] <dbg> net_sock: zsock_received_cb: (0x30001748): ctx=0x30002178, pkt=0x300085c8, st=0, user_data=0
[00:01:31.784,000] <err> eth_nxp_enet_qos_mac: No new RX buf available
[00:01:31.784,000] <err> eth_nxp_enet_qos_mac: Rx descriptor does not have first descriptor flag, drop
[00:01:31.784,000] <err> eth_nxp_enet_qos_mac: Rx descriptor does not have first descriptor flag, drop
[00:01:31.784,000] <err> eth_nxp_enet_qos_mac: Rx descriptor does not have first descriptor flag, drop
[00:01:31.884,000] <err> eth_nxp_enet_qos_mac: No new RX buf available
[00:01:31.884,000] <err> eth_nxp_enet_qos_mac: Rx descriptor does not have first descriptor flag, drop
[00:01:31.884,000] <err> eth_nxp_enet_qos_mac: Rx descriptor does not have first descriptor flag, drop
[00:01:31.884,000] <err> eth_nxp_enet_qos_mac: Rx descriptor does not have first descriptor flag, drop
[00:01:31.985,000] <err> eth_nxp_enet_qos_mac: No new RX buf available
[00:01:31.985,000] <err> eth_nxp_enet_qos_mac: Rx descriptor does not have first descriptor flag, drop
[00:01:31.985,000] <err> eth_nxp_enet_qos_mac: Rx descriptor does not have first descriptor flag, drop
[00:01:31.985,000] <err> eth_nxp_enet_qos_mac: Rx descriptor does not have first descriptor flag, drop
[00:01:32.085,000] <err> eth_nxp_enet_qos_mac: No new RX buf available
[00:01:32.085,000] <err> eth_nxp_enet_qos_mac: Rx descriptor does not have first descriptor flag, drop
[00:01:32.085,000] <err> eth_nxp_enet_qos_mac: Rx descriptor does not have first descriptor flag, drop
[00:01:32.085,000] <err> eth_nxp_enet_qos_mac: Rx descriptor does not have first descriptor flag, drop
[00:01:32.186,000] <err> eth_nxp_enet_qos_mac: No new RX buf available
[00:01:32.186,000] <err> eth_nxp_enet_qos_mac: Rx descriptor does not have first descriptor flag, drop
[00:01:32.186,000] <err> eth_nxp_enet_qos_mac: Rx descriptor does not have first descriptor flag, drop
[00:01:32.186,000] <err> eth_nxp_enet_qos_mac: Rx descriptor does not have first descriptor flag, drop
[00:01:32.287,000] <err> eth_nxp_enet_qos_mac: No new RX buf available
[00:01:32.288,000] <err> eth_nxp_enet_qos_mac: Rx descriptor does not have first descriptor flag, drop
[00:01:32.288,000] <err> eth_nxp_enet_qos_mac: Rx descriptor does not have first descriptor flag, drop
[00:01:32.288,000] <err> eth_nxp_enet_qos_mac: Rx descriptor does not have first descriptor flag, drop
[00:01:33.551,000] <err> eth_nxp_enet_qos_mac: No new RX buf available
[00:01:33.767,000] <err> eth_nxp_enet_qos_mac: No new RX buf available
---- Closed the serial port COM10 ----

Impact

Showstopper – Prevents release or major functionality; system unusable.

Environment

• OS: Windows 11
• Workspace: [4.2.00+0-rc1]
• Zephyr SDK: 0.17.0
• Commit/Tag: 4.2.00-rc1

Additional Context

We are not modifying anything in the code or the hardware. clean code from 4.2.00-rc1 and default FRDM_N947 board from NXP.

Note: phy_mii: Failed to reset PHY (0): -116 doesn't look to be related but maybe the FRDM_N947 board DTS is missing setting the correct PHY driver, however packets are sending and receiving so looks like the chip (LAN8741) defaults work from reset.

[github-actions]

Hi @ekwus! We appreciate you submitting your first issue for our open-source project. 🌟

Even though I'm a bot, I can assure you that the whole community is genuinely grateful for your time and effort. 🤖💙

[ekwus]

Looks like this is related to the closed issues #89435 and #85172

[elby0]

Hello,

I am witnessing the same issue as well. Despite using the latest tag 4.2.0-rc1, I noticed the same issue with http_server example.

My Environment setup:

OS: Windows 11
Workspace: 4.2.0-rc1
Zephyr SDK: 0.17.0
Commit/Tag: 4.2.0-rc1

Steps to reproduce:

To replicate, simply used the http_server example (didn't modify a single line of code). Flash them on a board and starting running.

For HTTP_SERVER, I noticed the following error showing up immediately, sometimes after 5 minutes and sometimes after 20 minutes. The longest I have seen this error not showing up is 23 minutes.

This is what I observed while running the sample http_server, there are repeated RX failures under the moderate HTTP traffic.

<err> eth_nxp_enet_qos_mac: RX descriptor does not have first descriptor flag, drop
<err> eth_nxp_enet_qos_mac: could not allocate new RX pkt
<err> eth_nxp_enet_qos_mac: could not allocate new RX pkt
<err> eth_nxp_enet_qos_mac: could not allocate new RX pkt
<err> eth_nxp_enet_qos_mac: could not allocate new RX pkt
<err> eth_nxp_enet_qos_mac: could not allocate new RX pkt

Here, the packets that span multiple DMA descriptors are dropped, and eventually the driver exhausts Zephyr's Rx buffers.

[dleach02]

When this occurs is the networking non-functional from this point on?

[elby0]

@dleach02 Yes, you are right. The uptime and network statistics freezes. Also, the "Led On/Led Off" button doesn't control the LED anymore and the board's LED is left to the last active state.

Additional info: Tried with Static IP address and with DHCP enabled, both results in the same.

[Hoog-V]

I am witnessing the same bug as well. Both with zephyr commit 4.2.0-rc1 & 4.1.0. With 4.2.0 release it happens instantaneous (with single service enabled) and with the 4.1.0 release, it happens when combining multiple services like HTTP + MQTT. After the bug occurs, the whole stack freezes and networking becomes non-functional.

My Environment Setup:

OS: Mac OS Sequoia 15.5 (aarch64) (and also tested on Fedora 41 x86_64)
Workspace: 4.2.0-rc1
Zephyr-sdk: 0.17.0
Commit/Tag: 4.2.0-rc1
Board: FRDM-MCXN947

Steps to reproduce

To replicate build and flash http_server or mqtt (had it also occur in this sample) networking samples. Nothing has to be changed in the code to reproduce.

After a few minutes the same messages as above will show up in my debug console.