Dave,
Please take a look at
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=...
Can we include this simple patch in Fedora 7 and Rawhide please? Without this, events arrive in the wrong order meaning that udev and hal may get confused. Thanks.
David
On Fri, Jun 15, 2007 at 09:24:09AM -0400, David Zeuthen wrote:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=...
Can we include this simple patch in Fedora 7 and Rawhide please? Without this, events arrive in the wrong order meaning that udev and hal may get confused. Thanks.
Are you sure this is the diff you want? The changelog comment doesn't really imply anything regarding ordering.
rawhide is a no-brainer, it'll get picked up on a rebase soon.
Dave
On Fri, 2007-06-15 at 14:59 -0400, Dave Jones wrote:
On Fri, Jun 15, 2007 at 09:24:09AM -0400, David Zeuthen wrote:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=...
Can we include this simple patch in Fedora 7 and Rawhide please? Without this, events arrive in the wrong order meaning that udev and hal may get confused. Thanks.
Are you sure this is the diff you want? The changelog comment doesn't really imply anything regarding ordering.
Yea, I'm 100% positive - the problem is that uevents for "struct class_device" don't include the PHYSDEV* variables. This makes it impossible for udev to reorder the events, e.g. wait for the device that PHYSDEVPATH that the class device specifies. As a result you sometimes get to process the event for an input device *before* the physical device (e.g. USB interface) have been processed. SCSI generic devices are affected in a similar way.
David
On Fri, Jun 15, 2007 at 03:22:10PM -0400, David Zeuthen wrote:
On Fri, 2007-06-15 at 14:59 -0400, Dave Jones wrote:
On Fri, Jun 15, 2007 at 09:24:09AM -0400, David Zeuthen wrote:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=...
Can we include this simple patch in Fedora 7 and Rawhide please? Without this, events arrive in the wrong order meaning that udev and hal may get confused. Thanks.
Are you sure this is the diff you want? The changelog comment doesn't really imply anything regarding ordering.
Yea, I'm 100% positive - the problem is that uevents for "struct class_device" don't include the PHYSDEV* variables. This makes it impossible for udev to reorder the events, e.g. wait for the device that PHYSDEVPATH that the class device specifies. As a result you sometimes get to process the event for an input device *before* the physical device (e.g. USB interface) have been processed. SCSI generic devices are affected in a similar way.
ok, kernel-2_6_21-1_3222_fc8 is building which should have it tomorrow. I've no objection to including it (as long as it doesn't have additional dependancies). Chuck? anyone else?
Dave
kernel@lists.fedoraproject.org
-
Dave Jones -
David Zeuthen