If there is a bus consisting of PC--hub--device_A and another device_B is plugged in into the hub, then the PC may briefly loose contact with device_A. This is troublesome if a filesystem was mounted device_A or the device was otherwise in use. This only happens with _some_ devices as device_B, but presumably with all 6-port hubs. (There is only one known 6-port PHY chip, hence all 6-port hubs are very similar WRT to PHY functionality.) Affected are both the ieee1394 and the firewire driver stack. Windows XP and Mac OS X are known to handle this without connection loss, i.e. without danger to mounted filesystems. More data reported here: http://marc.info/?l=linux1394-user&m=121875239304371
Proposed patches: [1/3] ieee1394: regression in 2.6.25: updates should happen before probes http://lkml.org/lkml/2008/8/16/85 [2/3] ieee1394: don't drop nodes during bus reset series http://lkml.org/lkml/2008/8/16/86 [3/3] ieee1394: sbp2: let nodemgr retry node updates during bus reset series http://lkml.org/lkml/2008/8/16/88 They work for me but need to be confirmed by the original reporter.
Fix confirmed by the original reporter (Damien Benoist). To do: - Get the ieee1394 fixes upstream and into stable. - Implement a similar fix in the new firewire stack.
ieee1394 fixed upstream (2.6.27-rc in August), firewire not yet done
firewire fix posted: http://lkml.org/lkml/2009/1/17/156
firewire fixed in 2.6.29-rc4