Hi Ian
I have the same problem as you described. But I
think that the problem is not based in the USB
stack but in the kernel itself.
Have you observed that when you mount a large USB
partition the ammount of physical memory
decreases, and the more that you use the FS, the
less memory that you get until the system
completely hangs, that is, without swapping,
doesn\'t matter that I have 256 MB in RAM, once
the physical memory is exhuasted, the system hangs.
I will be trying a kernel patch to improve the
handling of the systems memory, one called
pre-emtive patch, but I only can do it on week
ends, so we have to wait and see.
In the while, does anybody have any good idea ?,
may be i am following the wrong clue.
Saludos
JARCH
Quoting Ian David Flintoft <[EMAIL PROTECTED]>:
>
> Hello,
>
> I\'ve just got a Lacie Pocket Drive
> USB 2.0 40 Gb HDD.
> At the moment I\'ve tried using the HDD
> only with USB1.1
> controllers.
>
> The disk is basically working however
> when I use it with
> an ext2 filesystem mounted asynchronously
> it locks the
> machine hard until the transfer is
> finished for data
> transfers greater than a few 100 Mb\'s. If
> I mount
> the partition with the sync option then
> it seems to
> work reliably but at a very low rate -
> which isn\'t much
> fun with a 40 GB HDD.
>
> The problem gets worse the larger the
> amount of data
> written. With about 312 MB on a system
> with 256MB of RAM
> a large portion is written to buffers
> quickly and then
> once it starts to commit to the disk
> there are a few
> minor latency problems but nothing too
> serious.
>
> With a large copy (e.g. 2 Gb) once the
> buffers start to
> be written to disk the system very
> quickly locks up
> completely until the transfer is finished
> and the
> effective data rate goes through the
> floor ( < 14 kb/s).
> Although I\'m no expert on kernels it
> \"feels\" like some
> kind of deadlock/race, maybe in the ext2
> fs or block layer -
> possibly due to the very low rate
> available over USB1.1
> for such large transfers breaking some
> assumptions.
> Nothing in the logs.
>
> Using a vfat filesystem the problem seems
> to be better,
> certainly with moderate size transfer I
> get a reasonable
> rate and no significant latency issues. I
> haven\'t had time
> to try a 2Gb copy yet. Unfortunately vfat
> doesn\'t support
> symbolic links and screws around with
> filename cases so
> it is not well suited for keeping a
> debain archive on, for
> example.
>
> I have seen something similar before with
> old parallel port
> zip disk on early 2.4 kernels using
> certain transfer modes
> which would also lock the system until
> finished and take about
> an hour to write 100 Mb.
>
> I haven\'t had time to play with other
> kernels yet so maybe
> that is worth a try. I\'ll also try the
> verbose debugging
> options in the USB system.
>
> Anyone got any ideas or seen anything
> similar?
>
> Details of system and performance
> observed below.
> Similar things happen on two other
> machines -
> a dual AMD AthlonMP, 512Mb using ohci and
> a
> Pentium II, 512Mb with uhci.
>
>
> Performance
> -----------
>
> File system on sda1 made with
>
> mke2fs /dev/sda1
>
> or
>
> mkfs.vfat -F 32 /dev/sda1
>
> Write 312MB of files (cp -a), check
> md5sums of files
> (md5sum -c ./md5sum.txt) and then delete
> all files (rm -rf).
>
> fs opts
>
-------------------------------------------------------------------------
> ext2 sync write 312MB 36m30s
> 142kB/s OK
> delete 312MB 0m38s -
> OK
>
> vfat write 312MB 7m25s
> 701kB/s OK
> md5sum 7m23s
> 704kB/s OK
> delete 312MB 0m05s -
> OK
>
> ext2 async write 312MB 7m14s
> 719kB/s OKish minor latency issues
> md5sum 7m01s
> 741kB/s OK
> delete 312MB 0m04s -
> OK
>
> ext2 async write 2100MB 65m to write
> 65Mb 16 kb/s
> Gave up (= reset button)
> after 65Mb before giving up.
> e2fsck fixed a few broken
> inodes etc on reboot.
> md5sums on what was written
> (after e2fsck) are OK
> Appears to be working but
> very very slowly and blocks all opther
> processes?
>
> (time includes time to umount device and
> sync buffers)
>
>
>
> Hardware
> ========
>
> Dell Inspiron 4100
> Intel Pentium III 1GHz
> 256 Mb SDRAM
> (512 Mb swap)
>
>
> OS
> ==
>
> Debian GNU/Linux 3.0 + all security
> update to 4/11/02
> KDE 3.0.3 self install
>
> linux 2.4.20-rc1
> alsa 0.9r5
>
> hotplug 20020114
> usbview 1.0
> usbutils 0.9
> modutils 2.4.15
>
>
> Relevant Modules
> ================
>
> nls_iso8859-1 2848 0
> (autoclean)
> nls_cp437 4384 0
> (autoclean)
> vfat 9308 0
> (autoclean)
> fat 29240 0
> (autoclean) [vfat]
> sd_mod 9916 2
> (autoclean)
> vmnet 18464 3
> parport_pc 13892 0
> parport 24576 0
> [parport_pc]
> vmmon 19444 0
> (unused)
> ds 6464 2
> irtty 7200 2
> (autoclean)
> irda 138476 1
> (autoclean) [irtty]
> usb-storage 54208 1
> scsi_mod 51868 2 [sd_mod
> usb-storage]
> usb-uhci 21220 0
> (unused)
>
>
> USB Root Hub
> ============
>
> USB 1.1
>
> usbview
> -------
>
> USB UHCI Root Hub
> Serial Number: bf80
> Speed: 12Mb/s (full)
> Number of Ports: 2
> Bandwidth allocated: 0 / 900 (0%)
> Total number of interrupt requests: 0
> Total number of isochronous requests: 0
> USB Version: 1.00
> Device Class: 09(hub )
> Device Subclass: 00
> Device Protocol: 00
> Maximum Default Endpoint Size: 8
> Number of Configurations: 1
>
> Config Number: 1
> Number of Interfaces: 1
> Attributes: 40
> MaxPower Needed: 0mA
>
> Interface Number: 0
> Name: hub
> Alternate Number: 0
> Class: 09(hub )
> Sub Class: 0
> Protocol: 0
> Number of Endpoints: 1
>
> Endpoint Address: 81
> Direction: in
> Attribute: 3
> Type: Int.
> Max Packet Size: 8
> Interval: 255ms
>
> USB Device
> ==========
>
> Lacie Pocketdrive 40 Gb
>
> USB 2.0
>
>
> usbview
> -------
>
> LaCie PocketDrive USB2
> Manufacturer: LaCie
> Serial Number: 11100E00050EE6EC
> Speed: 12Mb/s (full)
> USB Version: 2.00
> Device Class: 00(>ifc )
> Device Subclass: 00
> Device Protocol: 00
> Maximum Default Endpoint Size: 64
> Number of Configurations: 1
> Vendor Id: 059f
> Product Id: 0211
> Revision Number: 11.06
>
> Config Number: 2
> Number of Interfaces: 1
> Attributes: c0
> MaxPower Needed: 98mA
>
> Interface Number: 0
> Name: usb-storage
> Alternate Number: 0
> Class: 08(stor.)
> Sub Class: 6
> Protocol: 50
> Number of Endpoints: 3
>
> Endpoint Address:
> 01
> Direction: out
> Attribute: 2
> Type: Bulk
> Max Packet Size:
> 64
> Interval: 0ms
>
> Endpoint Address:
> 82
> Direction: in
> Attribute: 2
> Type: Bulk
> Max Packet Size:
> 64
> Interval: 0ms
>
> Endpoint Address:
> 83
> Direction: in
> Attribute: 3
> Type: Int.
> Max Packet Size:
> 2
> Interval: 32ms
>
> Interval: 32ms
>
> Disk
> ====
>
> Disk /dev/sda: 255 heads, 63 sectors,
> 4864 cylinders
> Units = cylinders of 16065 * 512 bytes
>
> Device Boot Start End
> Blocks Id System
> /dev/sda1 1 1020
> 8193118+ 83 Linux
> /dev/sda2 1021 1785
> 6144862+ 83 Linux
> /dev/sda3 1786 4864
> 24732067+ 5 Extended
> /dev/sda5 1786 2298
> 4120641 c Win95 FAT32 (LBA)
> /dev/sda6 2299 2811
> 4120641 83 Linux
> /dev/sda7 2812 3324
> 4120641 83 Linux
> /dev/sda8 3325 3837
> 4120641 83 Linux
> /dev/sda9 3838 4350
> 4120641 83 Linux
> /dev/sda10 4351 4864
> 4128673+ 83 Linux
>
>
> --
>
> Dr Ian David Flintoft Email:
> [EMAIL PROTECTED]
> Applied Electromagnetics Group Tel:
> +44 (0) 1904 432391
> Department of Electronics Fax:
> +44 (0) 1904 433224
> University of York
> Heslington
> YORK, UK
> YO10 5DD
>
>
>
>
>
-------------------------------------------------------
> This sf.net email is sponsored by: See
> the NEW Palm
> Tungsten T handheld. Power & Color in a
> compact size!
>
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
> _______________________________________________
> [EMAIL PROTECTED]
> To unsubscribe, use the last form field
> at:
>
https://lists.sourceforge.net/lists/listinfo/linux-usb-users
>
-------------------------------------------------------
This sf.net email is sponsored by: See the NEW Palm
Tungsten T handheld. Power & Color in a compact size!
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-users
