Is it possible to force USB3.0 Port into USB2 mode?

#1

So, long story short, I have a ROCK 5B with a 4TB NVMe SSD from Crucial and currently a 1TB SSD from Crucial as well attached via USB. When I attach the SSD using a USB3.0 Cable the drive tends to reset rather frequently(every couple hours) and sometimes fails to the point of the whole drive being removed from the kernels space. Its not like this drive is pulling major load, it should at no point have to do any continuous reads(full read load for longer than 1 minute) but is rather used as a media storage for a jellyfin server. I noticed that when I attach the drive to the system using a usb 2.0 cable it works flawlessly(no errors or resets for over a week).
Now, I have a second drive, thats an intenso SSD with 2TB that I would like to add to the system. Same game here, the drive works for a short time, eventually starts resetting, and will then at some point just drop out until physically disconnected and reconnected. However the Case for this drive uses USBA-A cables, and I just dont happen to have a non-USB3 cable lying around. I’m fairly certain the only reason that the first drive works with a 2.0 cable is power, as in requests from the drive are delivered at full speed and that peaks power for a short time, thus causing the device to reset, so limiting the speed of the drive down to USB-2.0 might be a workable solution.

I have also tried attaching both drives to the USB2 Ports on the rock5, but regardless of the cable they both start resetting during use. I suppose the usb2 port by design only delivers the 500mA its supposed to whereas the 3.0 ports have a higher limit regardless of whether they are being used in 2.0 or 3.0 mode.

Is there any way to force devices attached to usb3 to be used in usb2 mode?

#2

There are multiple ways to force USB 2.0 mode: isolate the SS pins (by plugging in the connector half-way through or use an USB 2 cable), disable USB 3 device nodes in the DeviceTree, disable the XHCI driver itself.
However, the bottom USB 3.0 port depends on XHCI so not idea if it’s possible to switch the driver into USB 2.0 mode.

But I’d just look into fixing the actual issue rather than working around it like this. You’re limiting the drives to 40-50 MB/s vs. 300-400 MB/s with USB 3.

What power supply are you using? Have you check dmesg logs?

#3

My 5B is attached to a 60w pd power supply from an Asus Notebook, patched in between a renkforce usb-c dock(since the rock 5 won’t negotiate with an Asus laptop pd brick). Sensors shows it is in 20v 2.88 A mode, so there should be plenty to give. And yes I did check dmesg logs. Basically when one of these drives is on a 3.0 port I get a lot of reset messages followed by re ack ok messages from the kernel. When reading or writing to a drive, if this occurs read/write will halt for about 2-3 seconds and then resume until the next controller reset comes. I can attest that neither the drives nor the cases are faulty, as the drives work perfectly fine when attached to a PC via either USB case or s-ata interface. There is enough rf static when using those drives on usb3 on a PC though to black out my Logitech unified receiver, but I guess that’s just due to the physical closeness of the ports. The only workaround I found so far is using a 2.0 cable. This makes the logical choice of possible faults that there is not enough current on the usb3 ports to handle a s-ata SSD, since throttling the drive to usb2 makes sure the band chips and controller never have to work at full load. I have not yet measured the exact power use between the SSD and the port, can’t find my USB power meter.

I also know that the usb3 ports themselves are fine, I had no issues whatsoever with other devices and even when putting one of the ssds into a 3.5" enclosure with its own external power supply. That isn’t a workable permanent solution simply because I cannot permanently fit 2 wall bricks.

#4

Ok, after doing some further reading in the thread about working pd power supplies I came onto a git page from some guy doing power and performance benchmarks who also had underpowering issues and linked to this forum post:

This shows the exact problem I get when attaching it on usb3 with a usb3 cable, and this also hints at underpower. The op solved this by adding external power to the drive which unfortunately is not a workable solution for me. However as my testing suggests, throttling the device cuts power usage(and I think the particular issue here might actually be spikes in power usage) enough for the device to become a reliable USB endpoint. As soon as I find my USB power meter I’ll check for that

#5

Where? Between SBC and USB consumers? Do you own a premium power meter able to report consumption spikes?

#6

Yes and no. It’s an esp32 based one with an external ADC and a 2khz sample logger. My hope is that the sampling rate is high enough to catch and sync it. And yes, the ideas is to measure between USB port and consumer