Loading ...
Sorry, an error occurred while loading the content.

Fixing USB HDD Xfer @ 6MB/s (10 times slower than USB2)

Expand Messages
  • Frenchy
    Do we have a solution to fix the extra slow USB transfers to external HDD s? If you plug your USB2 disk on the front usb port, it should get mounted as
    Message 1 of 5 , Jul 27, 2005
    • 0 Attachment
      Do we have a solution to fix the extra slow USB transfers to
      external HDD's?

      If you plug your USB2 disk on the front usb port, it should get
      mounted as /dev/sda

      Type:
      /sbin/hdparm -t /dev/sda

      Answer...:
      Timing buffered disk reads: 64 MB in 10.08 seconds = 6.35 MB/sec

      This may have something to do with the USB device driver or murasaki
      fonctionality. USB in LS is still new to me.

      There is not much that hdparm V3.9 can probe about USB disks because
      they are connected as SCSI. Hdparm can not use IDE signaling accross
      a USB connection to access disk registers ..:
      hdparm -I /dev/sda
      operation not supported on SCSI disks

      If we do the math, 6MB/s = 48Mb/s ie 10 times less than USB2. The
      interface must not be setup correctly and makes the LS USB2
      interface sub-standard. The disk itself puts out about 30MB/s so the
      bottle neck is with the interface.

      As a lead, I can only guess that if the USB driver is polling the
      USB chip instead of using IRQ we'll get this range of performance.
      Like I say I don't know a thing about USB implementation on LS!
    • Frenchy
      I guess we know why USB disk Xfers are extra slow... In /var/log/messages (see below) you can see that usbcore is missing. From what I understand, usbcore is
      Message 2 of 5 , Jul 27, 2005
      • 0 Attachment
        I guess we know why USB disk Xfers are extra slow...
        In /var/log/messages (see below) you can see that "usbcore" is
        missing.
        From what I understand, usbcore is basically a kernel driver for
        murasaki USB engine.
        If we can find a copy "usbcore" file that will allow USB disk
        transfers to use kernel hooks.


        murasaki.usb[1964]: usb device is added
        murasaki.usb[1964]: vendor:0x922 product:0x8 Dclass:0x0
        murasaki.usb[1964]: MATCH(printer) -> match_flags:0x380 vendor:0x0

        murasaki.usb[1964]: Dependence(usbcore) was found
        murasaki.usb[1964]: Loading printer

        murasaki.usb[1964]: Loading usbcore
        modprobe: modprobe: Can't locate module usbcore

        Where can we get a copy "usbcore" compiled for LS?
        -sam.

        --- In LinkStation_General@yahoogroups.com, "Frenchy "
        <glasseron@y...> wrote:
        > Do we have a solution to fix the extra slow USB transfers to
        > external HDD's?
        >
        > If you plug your USB2 disk on the front usb port, it should get
        > mounted as /dev/sda
        >
        > Type:
        > /sbin/hdparm -t /dev/sda
        >
        > Answer...:
        > Timing buffered disk reads: 64 MB in 10.08 seconds = 6.35 MB/sec
        >
        > This may have something to do with the USB device driver or
        murasaki
        > fonctionality. USB in LS is still new to me.
        >
        > There is not much that hdparm V3.9 can probe about USB disks
        because
        > they are connected as SCSI. Hdparm can not use IDE signaling
        accross
        > a USB connection to access disk registers ..:
        > hdparm -I /dev/sda
        > operation not supported on SCSI disks
        >
        > If we do the math, 6MB/s = 48Mb/s ie 10 times less than USB2. The
        > interface must not be setup correctly and makes the LS USB2
        > interface sub-standard. The disk itself puts out about 30MB/s so
        the
        > bottle neck is with the interface.
        >
        > As a lead, I can only guess that if the USB driver is polling the
        > USB chip instead of using IRQ we'll get this range of performance.
        > Like I say I don't know a thing about USB implementation on LS!
      • irvine_stock
        I thought the USB performance was the PRINTF in the driver. Not easy to fix without changing the kernel. Do a search and I think you will find it. Rich
        Message 3 of 5 , Jul 27, 2005
        • 0 Attachment
          I thought the USB performance was the PRINTF in the driver.
          Not easy to fix without changing the kernel.
          Do a search and I think you will find it.
          Rich
          --- In LinkStation_General@yahoogroups.com, "Frenchy "
          <glasseron@y...> wrote:
          > I guess we know why USB disk Xfers are extra slow...
          > In /var/log/messages (see below) you can see that "usbcore" is
          > missing.
          > From what I understand, usbcore is basically a kernel driver for
          > murasaki USB engine.
          > If we can find a copy "usbcore" file that will allow USB disk
          > transfers to use kernel hooks.
          >
          >
          > murasaki.usb[1964]: usb device is added
          > murasaki.usb[1964]: vendor:0x922 product:0x8 Dclass:0x0
          > murasaki.usb[1964]: MATCH(printer) -> match_flags:0x380 vendor:0x0
          >
          > murasaki.usb[1964]: Dependence(usbcore) was found
          > murasaki.usb[1964]: Loading printer
          >
          > murasaki.usb[1964]: Loading usbcore
          > modprobe: modprobe: Can't locate module usbcore
          >
          > Where can we get a copy "usbcore" compiled for LS?
          > -sam.
          >
          > --- In LinkStation_General@yahoogroups.com, "Frenchy "
          > <glasseron@y...> wrote:
          > > Do we have a solution to fix the extra slow USB transfers to
          > > external HDD's?
          > >
          > > If you plug your USB2 disk on the front usb port, it should get
          > > mounted as /dev/sda
          > >
          > > Type:
          > > /sbin/hdparm -t /dev/sda
          > >
          > > Answer...:
          > > Timing buffered disk reads: 64 MB in 10.08 seconds = 6.35 MB/sec
          > >
          > > This may have something to do with the USB device driver or
          > murasaki
          > > fonctionality. USB in LS is still new to me.
          > >
          > > There is not much that hdparm V3.9 can probe about USB disks
          > because
          > > they are connected as SCSI. Hdparm can not use IDE signaling
          > accross
          > > a USB connection to access disk registers ..:
          > > hdparm -I /dev/sda
          > > operation not supported on SCSI disks
          > >
          > > If we do the math, 6MB/s = 48Mb/s ie 10 times less than USB2. The
          > > interface must not be setup correctly and makes the LS USB2
          > > interface sub-standard. The disk itself puts out about 30MB/s so
          > the
          > > bottle neck is with the interface.
          > >
          > > As a lead, I can only guess that if the USB driver is polling the
          > > USB chip instead of using IRQ we'll get this range of performance.
          > > Like I say I don't know a thing about USB implementation on LS!
        • James Stewart
          On the PPC processor based units, they left some very high overhead debug statements compiled into the USB driver. Maybe they did with the MIPS based units as
          Message 4 of 5 , Jul 29, 2005
          • 0 Attachment
            On the PPC processor based units, they left some very high overhead
            debug statements compiled into the USB driver. Maybe they did with
            the MIPS based units as well. To fix it you have to recompile and
            reinstall the kernel. There are only a few people who have attempted
            this over on www.kurobox.com with about half the people having success
            and the other half destroying their Kuroboxes. There may be some more
            recent developments however, I'm a little out-of-date on the status of
            things.

            --- In LinkStation_General@yahoogroups.com, "Frenchy "
            <glasseron@y...> wrote:
            > Do we have a solution to fix the extra slow USB transfers to
            > external HDD's?
            >
            > If you plug your USB2 disk on the front usb port, it should get
            > mounted as /dev/sda
            >
            > Type:
            > /sbin/hdparm -t /dev/sda
            >
            > Answer...:
            > Timing buffered disk reads: 64 MB in 10.08 seconds = 6.35 MB/sec
            >
            > This may have something to do with the USB device driver or murasaki
            > fonctionality. USB in LS is still new to me.
            >
            > There is not much that hdparm V3.9 can probe about USB disks because
            > they are connected as SCSI. Hdparm can not use IDE signaling accross
            > a USB connection to access disk registers ..:
            > hdparm -I /dev/sda
            > operation not supported on SCSI disks
            >
            > If we do the math, 6MB/s = 48Mb/s ie 10 times less than USB2. The
            > interface must not be setup correctly and makes the LS USB2
            > interface sub-standard. The disk itself puts out about 30MB/s so the
            > bottle neck is with the interface.
            >
            > As a lead, I can only guess that if the USB driver is polling the
            > USB chip instead of using IRQ we'll get this range of performance.
            > Like I say I don't know a thing about USB implementation on LS!
          • johnnypirish
            Gordo @ kurobox.com has recompiled the kernel using a loader.o module that loads a newly compiled usb driver without the printk call. This increases the speed
            Message 5 of 5 , Jul 29, 2005
            • 0 Attachment
              Gordo @ kurobox.com has recompiled the kernel using a loader.o
              module that loads a newly compiled usb driver without the printk
              call. This increases the speed about 10%.

              JP


              --- In LinkStation_General@yahoogroups.com, "James Stewart"
              <wartstew@y...> wrote:
              > On the PPC processor based units, they left some very high overhead
              > debug statements compiled into the USB driver. Maybe they did with
              > the MIPS based units as well. To fix it you have to recompile and
              > reinstall the kernel. There are only a few people who have
              attempted
              > this over on www.kurobox.com with about half the people having
              success
              > and the other half destroying their Kuroboxes. There may be some
              more
              > recent developments however, I'm a little out-of-date on the
              status of
              > things.
              >
              > --- In LinkStation_General@yahoogroups.com, "Frenchy "
              > <glasseron@y...> wrote:
              > > Do we have a solution to fix the extra slow USB transfers to
              > > external HDD's?
              > >
              > > If you plug your USB2 disk on the front usb port, it should get
              > > mounted as /dev/sda
              > >
              > > Type:
              > > /sbin/hdparm -t /dev/sda
              > >
              > > Answer...:
              > > Timing buffered disk reads: 64 MB in 10.08 seconds = 6.35
              MB/sec
              > >
              > > This may have something to do with the USB device driver or
              murasaki
              > > fonctionality. USB in LS is still new to me.
              > >
              > > There is not much that hdparm V3.9 can probe about USB disks
              because
              > > they are connected as SCSI. Hdparm can not use IDE signaling
              accross
              > > a USB connection to access disk registers ..:
              > > hdparm -I /dev/sda
              > > operation not supported on SCSI disks
              > >
              > > If we do the math, 6MB/s = 48Mb/s ie 10 times less than USB2.
              The
              > > interface must not be setup correctly and makes the LS USB2
              > > interface sub-standard. The disk itself puts out about 30MB/s so
              the
              > > bottle neck is with the interface.
              > >
              > > As a lead, I can only guess that if the USB driver is polling
              the
              > > USB chip instead of using IRQ we'll get this range of
              performance.
              > > Like I say I don't know a thing about USB implementation on LS!
            Your message has been successfully submitted and would be delivered to recipients shortly.