On Tue, Oct 27, 2009 at 06:50:30PM -0400, eschvoca wrote: > On Tue, Oct 27, 2009 at 12:10 PM, Paul E. McKenney > <paulmck@xxxxxxxxxxxxxxxxxx> wrote: > > On Tue, Oct 27, 2009 at 10:24:22AM -0400, eschvoca@xxxxxxxxx wrote: > >> Hi Paul, > >> On Tue, Oct 27, 2009 at 1:00 AM, Paul E. McKenney < > >> paulmck@xxxxxxxxxxxxxxxxxx> wrote: > >> > >> > On Mon, Oct 26, 2009 at 09:33:33PM -0700, Arjan van de Ven wrote: > >> > > On Mon, 26 Oct 2009 22:31:16 -0400 > >> > > eschvoca <eschvoca@xxxxxxxxx> wrote: > >> > > > >> > > > Hi Oliver, > >> > > > > >> > > > (Adding cc: Arjan) > >> > > > > >> > > > On Wed, Oct 14, 2009 at 10:54 PM, eschvoca <eschvoca@xxxxxxxxx> wrote: > >> > > > > Hi Oliver, > >> > > > > > >> > > > > It ends up that the problem happens sporadically -- some reboots the > >> > > > > keyboard and bluetooth adapter don't work while other reboots it > >> > > > > does. I tried variations of reboot vs suspend vs shutdown vs unplug > >> > > > > power and couldn't find a repeatable pattern. Perhaps the problem > >> > > > > is related to Ubuntu trying to speed up the boot process which > >> > > > > doesn't wait long enough for the USB devices to respond. My mouse > >> > > > > always works while my USB bluetooth dongle and keyboard (through a > >> > > > > GASIA PS2toUSB Adapter) do not. > >> > > > > > >> > > > > I've been looking for a guide on how to do a "git bisect" on vanilla > >> > > > > kernels on ubuntu but haven't found anything that works well. I > >> > > > > compiled a vanilla kernel but when I went to install the debs I was > >> > > > > warned that there already was a 2.6.28 kernel there and I would > >> > > > > overwrite it. I would like to do a "git bisect" on the vanilla > >> > > > > kernel and be able to install them without conflicts. Currently > >> > > > > the problem is showing up in about 25% of reboots. > >> > > > > > >> > > > > Please direct. > >> > > > > > >> > > > > Thanks. > >> > > > > >> > > > After 16 bisects I tracked down the problem with my USB keyboard > >> > > > adapter and USB bluetooth dongle not initializing on boot-up to this > >> > > > commit: > >> > > > >> > > my fear is the following: something ELSE in the system is missing a > >> > > synchronize_rcu(), and this case happened to save it by virtue of being > >> > > spurious. > >> > > > >> > > Paul: do we call synchronize_rcu() before we go to userspace for the > >> > > first time somewhere? > >> > > >> > In many cases, yes. But I don't know of a specific synchronize_rcu() > >> > that is guaranteed to be executed before the first userspace execution > >> > on all kernel configurations for all architectures. > >> > > >> > I recommend putting a printk() in synchronize_rcu() to check up on this > >> > on the system in question. (Or a WARN_ON_ONCE() if log space is limited.) > >> > >> If you are recommending I put in a prink() then I will need a patch. Also > >> tell me what information you want me to collect. I'm just an avid Linux > >> fan/user that would really like my keyboard/USB devices to work on boot-up. > > > > Fair enough. Could you please send me your .config? The output of > > "grep RCU .config" will suffice. I also need to know what version of > > the kernel you are using -- looks like 2.6.28, but please let me know. > > > > I'm using a vanilla kernel repository to compile at commit > b79e83bdd961ec9b862191c0df51aaeb3cb85664 (according to the Makefile it > is 2.6.29-rc7). > > $> grep RCU .config > # RCU Subsystem > CONFIG_TREE_RCU=y > # CONFIG_TREE_PREEMPT_RCU is not set > # CONFIG_RCU_TRACE is not set > CONFIG_RCU_FANOUT=32 > # CONFIG_RCU_FANOUT_EXACT is not set > # CONFIG_TREE_RCU_TRACE is not set > # CONFIG_RCU_TORTURE_TEST is not set > # CONFIG_RCU_CPU_STALL_DETECTOR is not set Got it, thank you! Could you please give the following (untested) patch a go? Thanx, Paul diff --git a/kernel/rcupdate.c b/kernel/rcupdate.c index cae8a05..ce855b2 100644 --- a/kernel/rcupdate.c +++ b/kernel/rcupdate.c @@ -86,6 +86,7 @@ void synchronize_rcu(void) if (rcu_blocking_is_gp()) return; +WARN_ONCE(1, "First blocking call to synchronize_rcu()"); init_completion(&rcu.completion); /* Will wake me after RCU finished. */ call_rcu(&rcu.head, wakeme_after_rcu); -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
