On Tue, May 04, 2004 at 09:57:48AM -0400, Alan Stern wrote: > On Mon, 3 May 2004, Matthew Dharm wrote: > > > On Mon, May 03, 2004 at 06:16:01PM -0700, Pete Zaitcev wrote: > > > Stuart responed to the suggestion above with the attached patch. > > > I changed comments a little. > > > > > > I think I'll have to accept it for Red Hat. Matt, Alan - what's the > > > community's take? > > > > It looks pretty good to me. I could argue about the benefits/drawbacks of > > taking/releasing the lock twice in the invoke_transport() function... but > > this seems like a reasonable approach. > > > > Alan, what do you think? > > I also think it looks good. Getting/releasing the lock twice doesn't > bother me because that only happens when a command fails and we > auto-sense. > > We need to do _something_ in 2.6 to accomplish more or less the same > result. One possibility, not quite so invasive as this (but perhaps also > not as effective?) is to have usbfs refuse to send URBs to a device unless > either none of the interfaces are bound or at least one of them is bound > to usbfs itself.
If we just refused access unless bound, then could utilities like lsusb
work?
Matt
--
Matthew Dharm Home: [EMAIL PROTECTED]
Maintainer, Linux USB Mass Storage Driver
Ye gods! I have feet??!
-- Dust Puppy
User Friendly, 12/4/1997
pgp00000.pgp
Description: PGP signature
