On Wed, 15 Sep 2004, Nigel Metheringham wrote: > For a long time now I have had real problems with USB bulk transfer > stuff (anything that moves reasonable amounts of data to mass storage > devices and gnomad2 - an interface to a Creative Zen mp3 player, not > using a mass storage interface - has never worked right for me). > > Finally a comment on a mailing list lead me to suspect the gnome-pilot > daemon as the culprit. Further testing has shown that if something is > looping performing the following operations:- > * open /proc/bus/usb/devices > * stat the fh from above > * read fh > * close fh > * poll() on any filehandles (do not need to be USB related > - /dev/null works fine) > then USB operations frequently fail with "control timeout on ep0in" - as > has been reported on the linux-usb-users list:- > http://www.mail-archive.com/[EMAIL PROTECTED]/msg11221.html > > I've attached a little bit of perl which is the smallest bit of code I > have got that reproduces this effect (it does a deliberate stat() on the > filehandle - that appears to be needed. I did think about removing the > stat() from the gnome-pilot code as a workround, but the stat() is > implicit as part of the internals of fopen() apparently). Obviously you > need to do some USB data transfer operation whilst the perl code is > running. > > I have seen this problem previously with 2.4.x kernels, I currently see > it with 2.6.8 (Fedora 2 build - kernel-2.6.8-1.521 i686 arch) and have > in hindsight never seen bulk USB transfers work on a box where gnome- > pilot (or rather gpilotd) is running. > > Its going to be a real fun bug to pin down.... > > Nigel.
Not at all, we already know exactly what the problem is. It's a bug in the storage device -- the device can't handle a standard request (of the sort send by the /proc/bus/usb/devices driver when reading a device string for the Vendor, Product, or Serial number) while the device is in the middle of a mass storage data transfer. Devices are supposed to be able to do this; it's the equivalent of walking and chewing gum at the same time. But some can't, and when asked to they simply fail. So far nobody has come up with a really good solution. There's nothing illegal about making these other requests during a data transfer, and plenty of devices can handle them just fine. So we don't want to rule them out. But in the end maybe we'll have to. Alan Stern ------------------------------------------------------- This SF.Net email is sponsored by: thawte's Crypto Challenge Vl Crack the code and win a Sony DCRHC40 MiniDV Digital Handycam Camcorder. More prizes in the weekly Lunch Hour Challenge. Sign up NOW http://ad.doubleclick.net/clk;10740251;10262165;m _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
