On Mon, Sep 29, 2008 at 12:54 AM, Stefan Richter <stefanr@xxxxxxxxxxxxxxxxx> wrote: > the path_id script was apparently written with the assumption that the > hierarchy of IEEE1394 device representation begins with fw-host devices and > ends with children (node entry devices) or grandchildren (unit directory > devices) of them. > > If there are more levels of descendant devices than that, i.e. if there are > children of unit directory devices, then path_id may get stuck with 100% CPU > utilization. > > This is not a problem with current mainline drivers. But there is a new > driver to be added soon which adds input devices which would naturally be > children of unit directory devices. (It's the firedtv driver for FireDTV > DVB devices which feature a remote control. See > http://ieee1394.wiki.kernel.org/index.php/Out-of-tree_Kernel_Drivers and > http://user.in-berlin.de/~s5r6/linux1394/firedtv/.) > > I will work around this path_id bug by making the input devices children of > the fw-host devices. The drawback is that some useful information from the > unit directory device and node entry device (like model, vendor, and unique > ID) will be unavailable that way. Hmm, would be nice to avoid that, and have a proper sysfs hierarchy, otherwise we will end up with silly workarounds like we do today for that other huge storage subsystem with the strange sysfs layout of disconnected and duplicated nodes. :). I guess, we better don't make make such decisions because of a broken path_id. That script is a complete nightmare anyway. We should fix it instead of working around it in the kernel. > Strangely, path_id does not hang due to the dvb child and dvb/* > grandchildren of the node entry device. > > (I tested with Gentoo's udev-124-r1.) > > I hope that I won't need that workaround when I port firedtv from the > ieee1394 stack to the firewire stack. The firewire stack does not have > fw-host devices (only node entry devices and unit directory devices, with > different names and attributes than in the ieee1394 stack), hence > path_id's ieee1394 heuristics will not be triggered if the firewire stack is > running. Thanks, Kay -- To unsubscribe from this list: send the line "unsubscribe linux-hotplug" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
