0

I am using an external USB 3.0 4TB HDD, formatted on Windows with NTFS. When connecting it to Debian 11 with the Intel 8 Series/C220 Series USB 3.0 chipset I have read and write speeds of about 126MB/s. When connecting it to Debian 11 with the Fresco Logic FL1100 USB 3.0 or the Renesas uPD720201 USB 3.0 chipset both read & write speeds drop by 60%, but still above USB 2.0 speed.

Only the Intel chipset is capable of using uas, the other chipsets work properly with usb-storage only. But when forcing usb-storage instead of uas on the Intel chipset the speed there stays at 126MB/s. So it is not uas what makes the difference.

But there is NO significant speed difference when formatting the same drive with ext4 instead: In all cases I reach 126MB/s both ways.

I can reproduce this difference with Ubuntu 20.04, too. ntfs-3g seems to be the culprit.

Is there some strange mount parameter for ntfs-3g to overcome this slow-down.

Help will be appreciated!

Frank-Michael Fischer
  • 11
  • 2
  • I'm 99% sure ntfs-3g is not the culprit. FUSE might be but it sounds very unlikely. – Artem S. Tashkinov Apr 05 '22 at 11:37
  • Even FUSE: What does FUSE "know" about the USB chipset? – Frank-Michael Fischer Apr 05 '22 at 11:50
  • Again, NTFS-3G cannot possibly be the culprit, while FUSE might be. If you used a little bit better suited for the desktop Linux distro, you could use the native NTFS3 driver (available in Linux 5.15 and above) which works a ton better and faster. https://www.usenix.org/system/files/conference/fast17/fast17-vangoor.pdf https://www.fsl.cs.stonybrook.edu/docs/fuse/fuse-tos19-a15-vangoor.pdf – Artem S. Tashkinov Apr 05 '22 at 12:17
  • I did not test with small transfers at all. This is the way I'v tested: dd status=progress bs=2M count=15869 if=/dev/zero of=/media/mifi/3815430NTFS/test.dat – Frank-Michael Fischer Apr 05 '22 at 12:38
  • Make sure you've presented all the data. Make sure you've run `sync` after all your commands because dirty buffers may completely skew your timings and give you a false picture. Also read do follow on this: https://unix.stackexchange.com/questions/480399/why-were-usb-stick-stall-problems-reported-in-2013-why-wasnt-this-problem-so – Artem S. Tashkinov Apr 05 '22 at 12:43
  • I know this, believe me. I'v also used bonnie++ with fhe "-f" switch: no change in behaviour. – Frank-Michael Fischer Apr 05 '22 at 12:47
  • I'm pretty sure you're hitting some FUSE limitation (**not** ntfs-3g's) but I'm not an expert in the kernel internals, so I have no clue. This could even be caused by your NTFS partition file fragmentation which manifests only under certain conditions. – Artem S. Tashkinov Apr 05 '22 at 12:49

1 Answers1

1

Solved, many thanks to Artem S. Tashkinov! Changing to kernel 5.15 and mounting with ntfs3 delivers full performance under all USB3 chipsets again. FUSE seems to be the problem. SOLVED!

Frank-Michael Fischer
  • 11
  • 2