Very slow umass in 6.2-RC2

Kevin Oberman oberman at es.net
Mon Mar 5 19:18:22 UTC 2007


> Date: Thu, 22 Feb 2007 10:45:03 +0100 (CET)
> From: Oliver Fromme <olli at lurza.secnetix.de>
> Sender: owner-freebsd-stable at freebsd.org
> 
> Alexander Shikoff wrote:
>  > Oliver Fromme wrote:
>  > > Alexander Shikoff wrote:
>  > > > Wang Yi wrote:
>  > > > > Alexander Shikoff wrote:
>  > > > > > I have Apacer Flash:
>  > > > > > 
>  > > > > > umass0: vendor 0x1005 USB FLASH DRIVE, rev 2.00/1.00, addr 2
>  > > > > > da0 at umass-sim0 bus 0 target 0 lun 0
>  > > > > > da0: < USB FLASH DRIVE 34CH> Removable Direct Access SCSI-0 device
>  > > > > > da0: 40.000MB/s transfers
>  > > > > > da0: 3936MB (8060928 512 byte sectors: 255H 63S/T 501C)
>  > > > > > 
>  > > > > > Writing to this device is very slooooow.
>  > > > > > 
>  > > > > > Time taken to copy file of 1,4G is near 30 min.
>  > > > > 
>  > > > > Did you try it under Windows? How time does it spend?
>  > > > 
>  > > > You may laugh but I have no Windows-based boxes with USB 2.0... :)
>  > > > Whatever it seems that this Apacer flash drive under Windows and USB 1.0 
>  > > > is faster than under FreeBSD and USB 2.0...
>  > > > 
>  > > > Soon I will have a possibility to compare Apacer 4GB with Transcend 
>  > > > flash. I'll report detailed results here.
>  > > 
>  > > Are you sure that your USB device supports hi-speed?
>  > http://www.apacer.com/en/products/Handy_Steno_AH320_features.htm
> 
> And is it connected to your ehci device (not ohci/uhci)?
> (To find out you have to look at "dmesg" an "sysctl dev".)
> 
> E.g. I have this one:
> 
> umass0: vendor 0x090c Cn Memory, rev 2.00/11.00, addr 2
> da0 at umass-sim0 bus 0 target 0 lun 0
> da0: < Cn Memory 1100> Removable Direct Access SCSI-0 device
> da0: 40.000MB/s transfers
> da0: 1935MB (3963904 512 byte sectors: 255H 63S/T 246C)
> 
> dev.umass.0.%parent: uhub4
> dev.uhub.4.%parent: usb4
> 
> usb4: EHCI version 1.0
> usb4: <Intel 82801FB (ICH6) USB 2.0 controller> on ehci0
> uhub4: Intel EHCI root hub, class 9/0, rev 2.00/1.00, addr 1
> ehci0: <Intel 82801FB (ICH6) USB 2.0 controller> mem 0xb0040000-0xb00403ff irq 23 at device 29.7 on pci0
> 
> i.e. umass0 is connected to uhub4 which is a child of usb4
> which belongs to ehci0.
> 
> That stick has about 20 Mbit/s write and 30 Mbit/s read
> speed.  That's not terribly fast either, but definitely
> beyond the 12 Mbit/s limit of full-speed devices.  I can
> fill that 2 GB stick up in about 15 minutes.  The system
> CPU is 95% during that, so I assume it's the USB stick
> that's the limiting factor.
> 
> However, it would be interesting to test the performance
> with HPS' new USB code.

Well, I have learned a bit more about the performance I have been
seeing. Whether it is common to Alexander's problem, I can't say, but I
am suspicious.

First, my devices that were running so slow are formatted as FAT32. This
is most common for such devices, but I am not sure if that is the case
for all reported systems.

After batting against a wall for a time I tried running fsck_msdosfs on
the slice. After running it my transfer rate went from about .2 MB/sec
to about 17 MB/sec. It moved the drive from useless for files of any
size to at least reasonable to use. Oddly, fsck_msdosfs reported no
errors. Also, oddly, the writes before running fsck were all successful
and the drive worked fine. 

Under Windows, the drive was fine before the fsck, so something was
seriously upsetting FreeBSD that did not impact Windows.

In any case, my drive is happy again.

USB drives are often disconnected while mounted and so this may turn out
to be a very common issue. Since FAT drives are not "marked" as unclean
like FFS systems, this is easy to over-look.
-- 
R. Kevin Oberman, Network Engineer
Energy Sciences Network (ESnet)
Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab)
E-mail: oberman at es.net			Phone: +1 510 486-8634
Key fingerprint:059B 2DDF 031C 9BA3 14A4  EADA 927D EBB3 987B 3751
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 224 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-stable/attachments/20070305/d7dba29c/attachment.pgp


More information about the freebsd-stable mailing list