From mboxrd@z Thu Jan 1 00:00:00 1970 From: Keir Fraser Subject: Re: Dom0 crashing when built with c/s 881 (Was: blktap2: need more than 3 values to unpack) Date: Thu, 04 Jun 2009 08:34:07 +0100 Message-ID: References: <45141205-3b4c-4258-9547-29acccc0be66@default> Mime-Version: 1.0 Content-Type: text/plain; charset="US-ASCII" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <45141205-3b4c-4258-9547-29acccc0be66@default> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Dan Magenheimer , Isaku Yamahata Cc: "Xen-Devel (E-mail)" List-Id: xen-devel@lists.xenproject.org If memory corruption is the issue, c/s 881 could be innocently triggering a latent bug. -- Keir On 04/06/2009 05:49, "Dan Magenheimer" wrote: > I'm quitting for the night now, but I see that > the string "nous" is followed by a "b" in the > immediately following word. Put this together > and you have "nousb" (read as turn off USB). > > Given that the "Driver 'xxx' needs updating" > are for 'usbfs', 'hub', and 'usb', and given > that usb is very related to hotplug, the string > "nousb" where a pair of function pointers should > be is suspicious (though it may also be a complete > coincidence). > > I'll try turning off CONFIG_HOTPLUG_PCI tomorrow. > > Dan > >> -----Original Message----- >> From: Dan Magenheimer >> Sent: Wednesday, June 03, 2009 7:09 PM >> To: Keir Fraser; Isaku Yamahata >> Cc: Xen-Devel (E-mail) >> Subject: RE: Dom0 crashing when built with c/s 881 (Was: [Xen-devel] >> blktap2: need more than 3 values to unpack) >> >> >>> If that doesn't help, I can go add some printk's where that >>> output is happening and see exactly what forces the failure. >> >> Here's what I am seeing. Looks like the drv->bus structure >> is getting overwritten with ASCII. >> >> Driver 'usbfs' needs updating - please use bus_type methods >> drv->bus->probe = 73756f6e (THAT's "nous" I think) >> drv->probe = d1069ab0 >> drv->bus->remove = 00000062 >> drv->remove = d1069960 >> drv->bus->shutdown = 00000000 >> drv->shutdown = 00000000 >> usbcore: registered new driver usbfs >> Driver 'hub' needs updating - please use bus_type methods >> drv->bus->probe = 73756f6e >> drv->probe = d1069ab0 >> drv->bus->remove = 00000062 >> drv->remove = d1069960 >> drv->bus->shutdown = 00000000 >> drv->shutdown = 00000000 >> usbcore: registered new driver hub >> Driver 'usb' needs updating - please use bus_type methods >> drv->bus->probe = 73756f6e >> drv->probe = d1069570 >> >>> -----Original Message----- >>> From: Dan Magenheimer >>> Sent: Wednesday, June 03, 2009 4:12 PM >>> To: Keir Fraser; Isaku Yamahata >>> Cc: Xen-Devel (E-mail) >>> Subject: RE: Dom0 crashing when built with c/s 881 (Was: [Xen-devel] >>> blktap2: need more than 3 values to unpack) >>> >>> >>>> Actually I missed that the dependent code is only built if >>>> CONFIG_PCI_IOMULTI && CONFIG_HOTPLUG_PCI. You've probably got >>>> one or both >>>> disabled and hence 882 doesn't break without 881 in your case. >>> >>> In case it helps, with 889 and 881 reverted, I am using the >>> default config which results in CONFIG_PCI_IOMULTI=y and >>> CONFIG_HOTPLUG_PCI=m. >>> >>>> I can't explain that from reading the code. It looks like a >>>> no-op to me >>>> unless the new function bus_register_notifier() is called. >>> >>> Looking at my boot output on failure, there are those lines: >>> >>> "Driver 'xxx' needs updating - please use bus_type methods" >>> >>> which is printed in linux/drivers/base/driver.c:driver_register() >>> if a complex set of circumstances are true. So clearly >>> something is different. >>> >>> If that doesn't help, I can go add some printk's where that >>> output is happening and see exactly what forces the failure. >>> >>> Dan >>> >>> >>>> -----Original Message----- >>>> From: Keir Fraser [mailto:keir.fraser@eu.citrix.com] >>>> Sent: Wednesday, June 03, 2009 3:28 PM >>>> To: Dan Magenheimer; Isaku Yamahata >>>> Cc: Xen-Devel (E-mail) >>>> Subject: Re: Dom0 crashing when built with c/s 881 (Was: >> [Xen-devel] >>>> blktap2: need more than 3 values to unpack) >>>> >>>> >>>> On 03/06/2009 22:17, "Dan Magenheimer" >>>> wrote: >>>> >>>>>> 1. It only adds extra functionality which by default would >>>>>> not do anything. >>>>>> Just reading the patch I can see it's harmless by itself >>>> as it merely >>>>>> introduces a new notifier chain which noone registers on. >>>>> >>>>> I won't pretend to understand the code, but I did >>>>> reproduce the problem with 881 by itself and with >>>>> 889 with 881 reverted. >>>> >>>> I can't explain that from reading the code. It looks like a >>>> no-op to me >>>> unless the new function bus_register_notifier() is called. >>>> >>>>>> 2. Changeset 882 depends on 881. If you revert just 881 then >>>>>> the tree will not build. >>>>> >>>>> This doesn't appear to be the case for me, though >>>>> I'm certainly willing to believe there's something >>>>> wrong with my process: >>>> >>>> Actually I missed that the dependent code is only built if >>>> CONFIG_PCI_IOMULTI && CONFIG_HOTPLUG_PCI. You've probably got >>>> one or both >>>> disabled and hence 882 doesn't break without 881 in your case. >>>> >>>> -- Keir >>>> >>>> >>>>