On Thu, Apr 25, 2002 at 07:39:03AM -0700, David Brownell wrote: > This patch is the result of that discussion a short while back > to fix the "hub driver polls too quickly at high speed" bug. > > - redefines "interval" of usb_fill_int_urb() to be what > the endpoint descriptor returns, and transparently > does the log-to-linear conversion if it's high speed > > - (most of the patch by volume!!) moves declarations > forward so testing dev->speed there will compile. > > No driver changes were needed -- most drivers pass the > endpoint bInterval value already, and those that don't are > not dealing with high speed device quirks.
Applied, thanks. > p.s. Includes an unrelated one-liner: deletes export of > usb_inc_dev_use() so it now shows up in kernel doc. > Minor oversight in an earlier patch. > > p.p.s. Driver-visible API change in 2.5.10: if you use > fops, you MUST set num_minors ... listing in the > /proc/bus/usb/drivers file happens even if there's > no way the driver will be accessible. I think it'd be > better to fail the registration if num_minors is zero > for a driver using fops. Do you see any drivers in the tree that are doing this? I thought I fixed them all. If a driver doesn't specify num_minors, they do not get a minor, it's that simple. I think the author will figure that one out quick enough :) thanks, greg k-h _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
