Skip to content
View in the app

A better way to browse. Learn more.

Armbian Community Forums

A full-screen app on your home screen with push notifications, badges and more.

To install this app on iOS and iPadOS
  1. Tap the Share icon in Safari
  2. Scroll the menu and tap Add to Home Screen.
  3. Tap Add in the top-right corner.
To install this app on Android
  1. Tap the 3-dot menu (⋮) in the top-right corner of the browser.
  2. Tap Add to Home screen or Install app.
  3. Confirm by tapping Install.

xaja34

Members
  • Joined

  • Last visited

Everything posted by xaja34

  1. xaja34 posted a topic in Allwinner sunxi
    # Orange Pi 4 Pro (Allwinner A733) – Severe SSH typing delays across multiple Armbian distributions I'm experiencing a very unusual SSH responsiveness problem on three Orange Pi 4 Pro boards (Allwinner A733). All three boards exhibit intermittent delays of several seconds when typing in an interactive SSH terminal. Sometimes individual characters take approximately five seconds to appear. The problem occurs with multiple Armbian distributions and kernels, including: - Armbian Debian Trixie, kernel 6.6.98-vendor-sun60iw2 - Armbian Ubuntu Resolute, including newer edge kernels in the 7.3.x series The boards otherwise appear to operate normally. ## Symptoms - SSH typing is intermittently extremely sluggish, even when typing continuously. - Local console input is responsive. - Large network transfers work at normal speeds. - CPU performance appears normal. - The problem occurs using both OpenSSH and Dropbear. - The problem occurs using both Windows OpenSSH and Tabby. - Connecting from another Linux SBC produces the same hesitation. - My other six SBCs work normally with the same SSH clients and network. Because all three Orange Pi 4 Pro boards exhibit similar symptoms, I suspect a common software, driver, or board-specific issue rather than an isolated hardware failure. ## Ethernet configuration The interface is `end0`, using: - Driver: `st_gmac` - Platform driver: `dwmac-sunxi` - PHY: Motorcomm YT8531 Gigabit Ethernet - Link: 1000 Mb/s, full duplex Ethernet statistics show no CRC, FIFO, or DMA errors. ## Troubleshooting performed I have tried the following without resolving the problem: - Disabling Ethernet Energy Efficient Ethernet (EEE) - Disabling GRO and GSO - Disabling RX/TX interrupt coalescing - Testing with OpenSSH and Dropbear - Testing SSH from both Windows and Linux clients - Checking Ethernet error counters and CPU performance TCP statistics show some retransmissions and out-of-order packets, although these have not been conclusively linked to the delays. ## Very unusual observation This is perhaps the most interesting finding: **Running `watch -n 1 date` in a second SSH terminal almost completely eliminates the typing delays in the first SSH terminal.** As soon as I stop `watch`, the delays return. This behaviour is repeatable. However: - Running a continuous `sleep 1` loop does not help. - Continuously pinging the board from another SBC does not help. It appears that regularly generating output through a separate SSH connection somehow prevents the delays. ## Packet capture observations I captured SSH traffic using `tcpdump` both with and without `watch` running. Without `watch`, the capture showed gaps of several seconds in SSH traffic, including gaps of approximately 4–10 seconds. With `watch` running, the long interruptions disappeared. The captures suggest that the delays may involve TCP packet transmission or delivery rather than slow SSH processing after packets arrive. ## Questions 1. Has anyone experienced similar interactive SSH delays on the Orange Pi 4 Pro or another Allwinner A733 board? 2. Could this be related to the `st_gmac` / `dwmac-sunxi` Ethernet driver or the YT8531 PHY? 3. Is there a known issue involving small TCP packets, interrupt handling, or Ethernet power management on this platform? 4. Are there any recommended kernel patches or diagnostic tests? I have packet captures and additional diagnostic information available if useful. Any suggestions would be appreciated.
  2. Hi Guation, thanks for the 1GB RAM fix — it got my Cubie A5E (v1.2, 1GB) booting on Armbian Trixie, kernel 6.18.36-current-sunxi64. I'm seeing very low sysbench cpu scores (single-thread ~56 events/sec, 8-thread ~1246 events/sec) and tracked it down to two things: There's no cpufreq at all in this kernel — no /sys/devices/system/cpu/cpu*/cpufreq directory on any core, nothing in dmesg about cpufreq/opp/nvmem, and clk_summary shows no CPU-core PLL entry whatsoever (only pll-audio1). This matches linux-sunxi.org listing the A523 CPU clock/DVFS driver as still "work in progress" upstream, so it's likely a mainline gap rather than anything board-specific. In the forum thread about your 1GB fix, someone's boot log showed "Failed to set core voltage! Can't set CPU frequency" — which looks like a separate, firmware/U-Boot-level issue tied to the 1GB DRAM voltage parameters, possibly capping the CPU at a fallback clock before Linux even loads.
  3. I'm running Armbian off NVMe, and flashed the U-Boot SPI firmware (u-boot-sunxi-with-spl.bin → /dev/mtdblock0) so I no longer need an SD card inserted just to get it booting. That part works great. But now I'd like to be able to pop in an SD card occasionally to try other distros/images without disturbing my NVMe install — and it seems that's not currently possible: with an SD card inserted, the board still boots straight from NVMe and ignores the card entirely. Dug into it a bit over serial console: The default boot_targets is fel nvme0 mmc_auto usb0 pxe dhcp — NVMe is checked before the SD/eMMC. Radxa hit the same issue on the RK3588 boards and fixed it upstream (radxa/u-boot#25, "give SD higher boot priority than NVME"). Would be great to see the same fix applied here. I tried working around it manually (setenv boot_targets 'mmc0 mmc1 nvme0 usb0 pxe dhcp'), which does work for that boot — SD is picked up correctly when present. But saveenv fails with "Unable to use mmc 0:1... Failed (1)", so the change doesn't survive a reboot. Looks like this build uses CONFIG_ENV_IS_IN_FAT targeting mmc 0:1, but the Armbian SD image ships a single ext4 partition there (confirmed with fstype mmc 0:1 → ext4), so the environment can never actually be saved. Is this a known limitation, or is there a recommended way to get persistent SD-over-NVMe priority on this board right now?
  4. Worked for me also. Thanks Samixa
  5. Guation. Would you be willing to share the patches (or Git commits) used to build the 1 GB Cubie A5E images? I'd like to incorporate them into the current Armbian build framework instead of using a prebuilt image.
  6. Guation, the image has no update MTD flash option in armbian-install so I can't boot from a nvme.
  7. To follow up: I have 4 orange pi 3b boards - 3 version 1.1 and one version 2.1. The 6.12.1-edge-rockchip64 kernel didn't work on the 1.1 boards but worked on the 2.1 board Vendoe 6.1.75 worked on both.
  8. Thanks Werner. 6.10.11-edge-rockchip64 worked fine. I guess there is no way to go back to this version. Unfortunately I didn't save an image. Bll
  9. After I shutdown Libre Sweet Potato, the whole network is somehow brought down unless I remove prower from the sbc. Any thoughts. Thanks
  10. I've flashed the Armbian-unofficial_24.8.0 - trunk sweet potato bookwork current 6.6.44 to EMMC and it boots fine to EMMC. However when I insert a SD card, it still boots from EMMC. I've added "boot_targets=mmc1 mmc0" to Armbianext.txt with no luck Can anyone help? Thsnks
  11. Werner Looks like it's just trixie. built an image with Bookworm (bookworm_edge_6.8.7_minimal) and works with no issues so far Bill
  12. After first boot and initial setup and login, I can enable and start the openssh server with no problem. I'm not sure how generating a set of new keys would help. This happens with numerous images. Armbian_24.5.0-trunk.429_Orangepi5-plus_trixie_edge_6.8.6_minimal.img Armbian_24.5.0-trunk.429_Orangepi5-plus_trixie_vendor_6.1.43_minimal.img Among others. Legacy image with 5.10.160 kernel does not have this problem Regards Bill
  13. Here is the partial boot log Starting ssh.service - OpenBSD Secure Shell server... sshd: no hostkeys available -- exiting. ssh.service: Control process exited, code=exited, status=1/FAILURE ssh.service: Failed with result 'exit-code'. Failed to start ssh.service - OpenBSD Secure Shell server. ssh.service: Scheduled restart job, restart counter is at 1. Starting ssh.service - OpenBSD Secure Shell server... sshd: no hostkeys available -- exiting. Not sure what to do as keys have not been generated yet. Thanks, Bill
  14. Orange Pi 5 plus. It seems that in the 'edge' kernels, ssh is disabled by default. Since HDMI doesn't function this is a problem. I used the debugging UART and after setup I got the message "ssh.service is disabled" I enabled the service and all is well. Any reason why ssh would be disabled?

Account

Navigation

Search

Search

Configure browser push notifications

Chrome (Android)
  1. Tap the lock icon next to the address bar.
  2. Tap Permissions → Notifications.
  3. Adjust your preference.
Chrome (Desktop)
  1. Click the padlock icon in the address bar.
  2. Select Site settings.
  3. Find Notifications and adjust your preference.