Usbat2 April 2003

usbat2@lists.lysator.liu.se

3 participants
5 discussions

new development version
by Herald van der Breggen 28 Apr '03

28 Apr '03

There is a new unfinished (but though working) development version of the driver available. Check out usbat-02/src from cvs. This driver has less (or no) problems with kernel version. Just make sure your kernel source is available at /usr/src/linux and do cd ..../usbat-02/src make complete make This should work in (almost) all cases (I hope). The "make complete" statement copies needed files from your kernel source and modifies a few files so that the usbat2 driver is activated. Let me know if this does not work in your case. To use the driver do modprobe sd_mod insmod usb-storage.o There seem to be distributions that don't include sd_mod (Suse?), don't know what should be loaded instead... Things I would like work out in this version is to add functionality that allows to change the cf-card while the device is connected (sniffings in http://usbat2.sourceforge.net/sniff could be handy for that). Another thing is to clean up the code and use the documentation to make it more readable/understandable. When that work is done, I guess it will be a lot easier for the kernel source maintainers to include the driver in the kernel source... I will try to work on it so now and then. Any help is welcome! Herald -- Herald van der Breggen <herald(a)breggen.xs4all.nl>

1 0

Re: Kernel patch
by Herald van der Breggen 26 Apr '03

26 Apr '03

On Sat, 2003-04-26 at 05:32, Ron wrote: > Not sure if this list will allow attachments or not, but > I've attached what should be a patch for the kernel in > Mandrake 9.1 Great! I have put it on http://usbat2.sourceforge.net/download.html > It compiled with a make modules and I just > installed the usb-storge.o manually instead of doing a make > modules_install It gets the scsi_mod in modules.dep, but > unfortunately not the sd_mod so that still needs to be > modprobed in manually before usb-storage. Not sure if it's > because of the way I installed the module, or what. Works > any way, so that's cool. Cool! > The driver in CVS would not compile > no matter what I tried. > Interesting.... what error messages did you see? > Mandrake automatically made a /mnt/memory_card dir and added > a line in fstab. So, once the driver was loaded it's done. > Bonus! > Great!

2 1

Red Hat 9 Dane-Elec CF with Patch vs 2.4.20 Works + Question re Buffering
by Simon Clift 25 Apr '03

25 Apr '03

The patch against Linux 2.4.20 worked OK against Red Hat 9's 2.4.20 kernel. My DaneElec CF reader now works. Thanks to those who did the work. :-) A couple of observations. Judging by the speed at which things happen I wonder if the disk-drive device that gets mounted uses buffers. If so, that would explain the corruption that occurs if the device is unplugged. It would seem to me that a device like this should be set up so that operations are unbuffered and not marked as completed until the device is flushed... I'll have to look at the source code if I get some time... not that I know much about drivers in Linux :-( These devices seem to be everywhere, and the USBAT chip fairly common. Despite their cheapness it might be a good idea from a Linux PR perspective if they worked a little more smoothly. Best regards --- Simon -- Simon Clift E-mail: ssclift(a)math.uwaterloo.ca Ph.D. Student Scientific Computation Group University of Waterloo Waterloo, Ontario, Canada N2L 3G1

2 4

Re: Buffering and Automount, Bad User Habits Unplugging Stuff
by Simon Clift 25 Apr '03

25 Apr '03

Replying to two e-mails here... I thought it might be good to have this discussion fully posted on the list. On Fri, 2003-04-25 at 03:39, christophe vangeneugden wrote: > > It would seem to me that a device like this should be set up > > so that operations are unbuffered and not marked as completed > > until the device is flushed... > I used the automount daemon to unmount the device when it is not used for a > few seconds. This seems to work ok. That's a good idea. I hunted for some supermount/autofs info and it seems on Red Hat the automount/autofs route is the one to take. I found a how-to that got me started. http://freespace.sourceforge.net/guidod/howto/autofs.html The page suggests setting automount to umount after 3 to 6 seconds, reducing the likelihood of corruption. That is exactly what you are doing, if I understand correctly. Thanks for the tip! On Fri, 2003-04-25 at 09:55, Chris Niekel wrote: > On Thu, Apr 24, 2003 at 05:18:40PM -0400, Simon Clift wrote: > >If so, that would explain the corruption that occurs > > if the device is unplugged. > All device drivers in linux are buffered I think. Hmm... shame there isn't an "autoflush" option at the kernel level, like there is in Perl. Then at least we would have the write guaranteed before the operation was marked as completed. > The device should not be unpluggable while in use by the computer. > Otherwise you can always get corruption. It would be nice if it operated like a journaling file system, though, with a way of rolling back partial operations (like a database transaction). That might be as simple as creating a temporary file name before completing the operation, at which point the file name is renamed to a visible/valid name. Reads are less of a problem. If buffers were marked for immediate flush then it should be possible to end up with a valid file with a high degree of probability. Unfortunately with this kind of media users are always going to be ripping it out too soon. :-( Best regards --- Simon -- Simon Clift <ssclift(a)math.uwaterloo.ca> University of Waterloo, Department of Computer Science

1 0

Kernel patch
by Ron 25 Apr '03

25 Apr '03

Just wanted to mention that the kernel patch for 2.4.20 also seems to work with the 2.4.21mdk kernel in MDK 9.1. However, I couldn't figure out how to edit the patch file to work with the different kernel, so just edited the kernel source files directly and plugged in the changes listed in the patch. If some one wants to explain to me how to create a patch file I could probably make up one for the kernel in MDK 9.1. Happy to have my card reader working again, hopefully this driver will make it into the kernel source soon. Thanks to all those who put this puppy together. ron -- The software said install Windows2000 or better.. so I installed Linux. -salsashrk

2 1

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

Usbat2 April 2003