Messages in this thread Patch in this message | | | Date | Sat, 16 Aug 2008 13:38:11 +0200 (CEST) | From | Stefan Richter <> | Subject | [patch 2/3] ieee1394: don't drop nodes during bus reset series |
| |
nodemgr_node_probe checked for generation increments too late and therefore prematurely reported nodes as "suspended".
Fixes http://bugzilla.kernel.org/show_bug.cgi?id=11349 for me.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de> --- drivers/ieee1394/nodemgr.c | 42 +++++++++++++++++++------------------ 1 file changed, 22 insertions(+), 20 deletions(-)
Index: linux/drivers/ieee1394/nodemgr.c =================================================================== --- linux.orig/drivers/ieee1394/nodemgr.c +++ linux/drivers/ieee1394/nodemgr.c @@ -1563,11 +1563,14 @@ struct probe_param { bool probe_now; }; -static int __nodemgr_node_probe(struct device *dev, void *data) +static int node_probe(struct device *dev, void *data) { struct probe_param *param = (struct probe_param *)data; struct node_entry *ne; + if (param->generation != get_hpsb_generation(param->hi->host)) + return -EAGAIN; + ne = container_of(dev, struct node_entry, node_dev); if (ne->needs_probe == param->probe_now) nodemgr_probe_ne(param->hi, ne, param->generation); @@ -1576,42 +1579,41 @@ static int __nodemgr_node_probe(struct d static void nodemgr_node_probe(struct host_info *hi, int generation) { - struct hpsb_host *host = hi->host; struct probe_param param; param.hi = hi; param.generation = generation; - /* Do some processing of the nodes we've probed. This pulls them + /* + * Do some processing of the nodes we've probed. This pulls them * into the sysfs layer if needed, and can result in processing of * unit-directories, or just updating the node and it's * unit-directories. * * Run updates before probes. Usually, updates are time-critical - * while probes are time-consuming. (Well, those probes need some - * improvement...) */ - + * while probes are time-consuming. + * + * Meanwhile, another bus reset may have happened. In this case we + * skip everything here and let the next bus scan handle it. + * Otherwise we may prematurely remove nodes which are still there. + */ param.probe_now = false; - class_for_each_device(&nodemgr_ne_class, ¶m, __nodemgr_node_probe); - param.probe_now = true; - class_for_each_device(&nodemgr_ne_class, ¶m, __nodemgr_node_probe); + if (class_for_each_device(&nodemgr_ne_class, ¶m, node_probe) != 0) + return; - /* If we had a bus reset while we were scanning the bus, it is - * possible that we did not probe all nodes. In that case, we - * skip the clean up for now, since we could remove nodes that - * were still on the bus. Another bus scan is pending which will - * do the clean up eventually. - * + param.probe_now = true; + if (class_for_each_device(&nodemgr_ne_class, ¶m, node_probe) != 0) + return; + /* * Now let's tell the bus to rescan our devices. This may seem * like overhead, but the driver-model core will only scan a * device for a driver when either the device is added, or when a * new driver is added. A bus reset is a good reason to rescan * devices that were there before. For example, an sbp2 device * may become available for login, if the host that held it was - * just removed. */ - - if (generation == get_hpsb_generation(host)) - if (bus_rescan_devices(&ieee1394_bus_type)) - HPSB_DEBUG("bus_rescan_devices had an error"); + * just removed. + */ + if (bus_rescan_devices(&ieee1394_bus_type) != 0) + HPSB_DEBUG("bus_rescan_devices had an error"); } static int nodemgr_send_resume_packet(struct hpsb_host *host) -- Stefan Richter -=====-==--- =--- =---- http://arcgraph.de/sr/
| |