Anomalous USB I/O errors when copying from one large external drive to another

funkapus · 01-27-2022, 07:17 AM

Running Linux 5.4.0 under Ubuntu 20.04.3 LTS.

I've got an issue that's driving me crazy. I recently picked up a few 12 TB external USB drives (WD MyBook 12 TB, repartitioned ext4 via gparted), with the intent of migrating from 4 TB external USB drives I'm currently using. Under most circumstances, the 12 TB drives work fine; but under some, I get I/O errors from them. More specifically:

Copying the contents of a nearly-full 4 TB drive to a new 12 TB drive: no problems
Comparing over 3 TB of data copied onto a new 12 TB drive back to the old 4 TB drive from which it came: no problems.
Copying all of the over 3 TB of data on a new 12 TB drive to /dev/null, just as a quick read test: no problems.
Doing those three steps above with a different new 12 TB drive: no problems.
Copying from one of the new 12 TB drives to another of the new 12 TB drives: I/O errors out the wazoo from the source drive in the copy (the drive being read).

Again, just to make clear: the new 12 TB drives do not generate errors when I copy to them from a smaller drive, or when I read from them in a comparison with the original smaller drive, or when I read from them to copy to the bitbucket. But if I copy from one of the new drives onto another one of them, I get I/O errors. The errors only occur when I'm using two of the new 12 TB drives at the same time (reading from one, writing to the other); and the errors are generated by reading the source drive. In any other other circumstance, they work just fine.

The errors associated with the source 12 TB drive being read from look like:

Code:

Jan 25 00:00:25 stax kernel: [190374.338268] usb 2-6.2: Disable of device-initiated U2 failed.
Jan 25 00:00:25 stax kernel: [190374.418582] usb 2-6.2: reset SuperSpeed Gen 1 USB device number 4 using xhci_hcd
Jan 25 00:00:25 stax kernel: [190374.441059] sd 4:0:0:0: [sdb] tag#0 FAILED Result: hostbyte=DID_ERROR driverbyte=DRIVER_OK
Jan 25 00:00:25 stax kernel: [190374.441066] sd 4:0:0:0: [sdb] tag#0 CDB: Read(16) 88 00 00 00 00 01 5e 4e d7 00 00 00 02 00 00 00
Jan 25 00:00:25 stax kernel: [190374.441073] blk_update_request: I/O error, dev sdb, sector 5877192448 op 0x0:(READ) flags 0x80700 phys_seg 32 prio class 0

I get a non-stop stream of these errors until I ctrl-c the copy. I've googled these error messages, and the usual consensus is that these are associated with a failing drive. But ignoring the fact that they're brand new (since new drives do sometimes come broken), badblocks returns no problems with either of the new 12 TB drives, and they work perfectly fine in every way so long as I'm not trying to copy from one of the new drives to another of the new drives. In particular, they work fine when copying to/from smaller drives I already had.

In all cases, the drives used are attached to powered USB hubs, which are then attached to physical ports on the computer. I've tried putting the two 12 TB drives on the same hub, as well as on different hubs that are themselves connected to physical USB ports associated with different USB controllers on the motherboard.

This is so bizarre to me, and I'm at a loss on how to proceed. Any advice would be very helpful.

smallpond · 01-27-2022, 08:22 AM

Code:

Disable of device-initiated U2 failed

U2 is something to do with power-save mode. I think U1 and U2 are sleep/wakeup or something. Check power management setting and try turning off save or suspend to the USB drive.

suramya · 01-29-2022, 12:53 PM

Could be that the hub doesn't have the capacity to support both drives under heavy load or is underpowered and looses power when both drives start pulling power. Try connecting them directly to USB ports on the motherboard and run the test again.

funkapus · 02-01-2022, 04:09 PM

Hi folks. Thanks for your replies!

Quote:

Originally Posted by smallpond

Code:

Disable of device-initiated U2 failed

U2 is something to do with power-save mode. I think U1 and U2 are sleep/wakeup or something. Check power management setting and try turning off save or suspend to the USB drive.

smartctl indicates that APM is disabled. I'm not sure how one goes about turning off suspend. I did some googling and came across the /etc/default/grub edits described here. Is this what you mean?

Quote:

Originally Posted by suramya

Could be that the hub doesn't have the capacity to support both drives under heavy load or is underpowered and looses power when both drives start pulling power. Try connecting them directly to USB ports on the motherboard and run the test again.

They were on different hubs that are on different controllers (one USB3, one USB2) during some of the failure attempts; but I'll give this a try as soon as a current badblocks re-run completes. I also tried using Y-connectors to put more power capacity onto the cables to the drives, but that didn't solve the problem . . .