Thanks, I'll get the info ASAP.
In the meantime, I have found some interesting discussion about UAS bugs in kernel 3.16: https://lists.debian.org/debian-kernel/2014/07/msg00318.html https://bbs.archlinux.org/viewtopic.php?id=183190 And another one in kernel 3.17: https://bbs.archlinux.org/viewtopic.php?id=189324 An here comes the output from my kern.log: Jun 22 15:12:02 D-NESTERKIN kernel: [ 1.377470] usb 3-2: new high-speed USB device number 2 using xhci_hcd Jun 22 15:12:02 D-NESTERKIN kernel: [ 1.508224] usb 3-2: New USB device found, idVendor=093b, idProduct=0023 Jun 22 15:12:02 D-NESTERKIN kernel: [ 1.508230] usb 3-2: New USB device strings: Mfr=85, Product=57, SerialNumber=44 Jun 22 15:12:02 D-NESTERKIN kernel: [ 1.508233] usb 3-2: Product: PLEXTOR USB Storage Adapter Jun 22 15:12:02 D-NESTERKIN kernel: [ 1.508235] usb 3-2: Manufacturer: PLEXTOR Jun 22 15:12:02 D-NESTERKIN kernel: [ 1.508238] usb 3-2: SerialNumber: ABC10CA1F554 Jun 22 15:12:02 D-NESTERKIN kernel: [ 1.510924] usb-storage 3-2:1.0: USB Mass Storage device detected Jun 22 15:12:02 D-NESTERKIN kernel: [ 1.511118] scsi6 : usb-storage 3-2:1.0 Jun 22 15:12:02 D-NESTERKIN kernel: [ 1.511270] usbcore: registered new interface driver usb-storage Looks somewhat similar for me. 2015-06-22 20:37 GMT+03:00 Thomas Schmitt <scdbac...@gmx.net>: > Hi, > > > I suppose it would be also helpful to compare not only kernel versions, > but > > also libs versions involved. > > Kernel modules for USB, (S)ATA, SCSI, maybe. > Userland libraries should hardly be involved, unless they > impose workload on USB. > > Last time i tried to understand SG_IO, i ended up at > the entrance of the i/o elevator. Dunno where the request > leaves that thing and hits the hardware controller driver. > > At some point of my current endeavor i will have to find > out why the new kernel shows such a miserable performance > with two concurrent DVD burns. I already read rumors that > it performs SG_IO synchronously in a single queue for all > devices. But i had thought that 2 times 4x DVD speed would > be no problem for a contemporary system. > My olde 2.6 on weaker hardware did it without problems. > > > Have a nice day :) > > Thomas > > -- С уважением, Нестеркин Дмитрий