John McDermott (U.S. Navy Employee)
2011-Mar-18 12:14 UTC
Fwd: [Xen-devel] Segfault When Building Snapshot xen-3d14ce7 kernel From linux/kernel/git/jeremy/xen.git/shortlog
Sorry, did not go to the list: Begin forwarded message:> From: "John McDermott (U.S. Navy Employee)" <john.mcdermott@nrl.navy.mil> > Date: March 18, 2011 7:56:58 AM EDT > To: Pasi Kärkkäinen <pasik@iki.fi> > Subject: Re: [Xen-devel] Segfault When Building Snapshot xen-3d14ce7 kernel From linux/kernel/git/jeremy/xen.git/shortlog > > Pasi, > > Thanks. I should have said something about that: > > I am building the hypervisor about 15 times a day on this box, doing development on the hypervisor, and have no problems doing that. > > Sincerely, > > John > > On Mar 18, 2011, at 7:52 AM, Pasi Kärkkäinen wrote: > >> On Fri, Mar 18, 2011 at 07:44:54AM -0400, John McDermott wrote: >>> Xen Developers, >>> >>> I downloaded snapshot xen-3d14ce7 (it is currently at the top of the list) yesterday and tried to build it overnight on my Fedora 13 box. It reported >>> >>> gcc: Internal error: Segmentation fault (program as) >>> >> >> Hello, >> >>> Make reports the error as being in [net/ipv4/ipip.mod.o] but it has "CC net/ipv4/ipip.mod.o" further up the list, above a couple of other files, as successful. The last file listed before the segfault is net/ipv4/netfilter/arptable_filter.mod.o. Sorry I have no further information; I ran it as make -j4. Before I run make clean and re-try as a single thread, to get more error information, is there anything of interest I should look at, to pass along, before I clean? I see there was an update to arp_notify a few commits below this version. >>> >> >> So it fails compiling Linux kernel. >> Are you sure your hardware is OK? Faulty memory? >> >> gcc shouldn''t segfault.. >> >> -- Pasi >> >>> Sincerely, >>> >>> John >>> ---- >>> What is the formal meaning of the one-line program >>> #include "/dev/tty" >>> >>> J.P. McDermott building 12 >>> Code 5542 mcdermott@itd.nrl.navy.mil >>> Naval Research Laboratory voice: +1 202.404.8301 >>> Washington, DC 20375, US fax: +1 202.404.7942 >>> >>> >>> >>> >>> >>> >>> >>> >>> _______________________________________________ >>> Xen-devel mailing list >>> Xen-devel@lists.xensource.com >>> http://lists.xensource.com/xen-devel > > ---- > What is the formal meaning of the one-line program > #include "/dev/tty" > > J.P. McDermott building 12 > Code 5542 mcdermott@itd.nrl.navy.mil > Naval Research Laboratory voice: +1 202.404.8301 > Washington, DC 20375, US fax: +1 202.404.7942 > > > > > > >---- What is the formal meaning of the one-line program #include "/dev/tty" J.P. McDermott building 12 Code 5542 mcdermott@itd.nrl.navy.mil Naval Research Laboratory voice: +1 202.404.8301 Washington, DC 20375, US fax: +1 202.404.7942 _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Ian Campbell
2011-Mar-18 13:28 UTC
Re: Fwd: [Xen-devel] Segfault When Building Snapshot xen-3d14ce7 kernel From linux/kernel/git/jeremy/xen.git/shortlog
On Fri, 2011-03-18 at 12:14 +0000, John McDermott wrote:> Sorry, did not go to the list: > > Begin forwarded message: > > > From: "John McDermott (U.S. Navy Employee)" <john.mcdermott@nrl.navy.mil> > > Date: March 18, 2011 7:56:58 AM EDT > > To: Pasi Kärkkäinen <pasik@iki.fi> > > Subject: Re: [Xen-devel] Segfault When Building Snapshot xen-3d14ce7 kernel From linux/kernel/git/jeremy/xen.git/shortlog > > > > Pasi, > > > > Thanks. I should have said something about that: > > > > I am building the hypervisor about 15 times a day on this box, doing development on the hypervisor, and have no problems doing that.Still, neither gcc nor as should segfault, ever, no matter what the input is. The paths which you mentioned (net/ipv4/blah..) aren''t paths which the Xen support in the really kernel touch. The fact that compiling the hypervisor build does not trigger the issue isn''t really indicative of anything in particular either since the inputs to gcc are different. Are you running the host under Xen when this happens? Which kernel/hypervisor? Is there anything in your kernel logs? Anyway, please provide a complete build log. Ian.> > > > Sincerely, > > > > John > > > > On Mar 18, 2011, at 7:52 AM, Pasi Kärkkäinen wrote: > > > >> On Fri, Mar 18, 2011 at 07:44:54AM -0400, John McDermott wrote: > >>> Xen Developers, > >>> > >>> I downloaded snapshot xen-3d14ce7 (it is currently at the top of the list) yesterday and tried to build it overnight on my Fedora 13 box. It reported > >>> > >>> gcc: Internal error: Segmentation fault (program as) > >>> > >> > >> Hello, > >> > >>> Make reports the error as being in [net/ipv4/ipip.mod.o] but it has "CC net/ipv4/ipip.mod.o" further up the list, above a couple of other files, as successful. The last file listed before the segfault is net/ipv4/netfilter/arptable_filter.mod.o. Sorry I have no further information; I ran it as make -j4. Before I run make clean and re-try as a single thread, to get more error information, is there anything of interest I should look at, to pass along, before I clean? I see there was an update to arp_notify a few commits below this version. > >>> > >> > >> So it fails compiling Linux kernel. > >> Are you sure your hardware is OK? Faulty memory? > >> > >> gcc shouldn''t segfault.. > >> > >> -- Pasi > >> > >>> Sincerely, > >>> > >>> John > >>> ---- > >>> What is the formal meaning of the one-line program > >>> #include "/dev/tty" > >>> > >>> J.P. McDermott building 12 > >>> Code 5542 mcdermott@itd.nrl.navy.mil > >>> Naval Research Laboratory voice: +1 202.404.8301 > >>> Washington, DC 20375, US fax: +1 202.404.7942 > >>> > >>> > >>> > >>> > >>> > >>> > >>> > >>> > >>> _______________________________________________ > >>> Xen-devel mailing list > >>> Xen-devel@lists.xensource.com > >>> http://lists.xensource.com/xen-devel > > > > ---- > > What is the formal meaning of the one-line program > > #include "/dev/tty" > > > > J.P. McDermott building 12 > > Code 5542 mcdermott@itd.nrl.navy.mil > > Naval Research Laboratory voice: +1 202.404.8301 > > Washington, DC 20375, US fax: +1 202.404.7942 > > > > > > > > > > > > > > > > ---- > What is the formal meaning of the one-line program > #include "/dev/tty" > > J.P. McDermott building 12 > Code 5542 mcdermott@itd.nrl.navy.mil > Naval Research Laboratory voice: +1 202.404.8301 > Washington, DC 20375, US fax: +1 202.404.7942 > > > > > > > > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xensource.com > http://lists.xensource.com/xen-devel_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
John McDermott (U.S. Navy Employee)
2011-Mar-28 12:41 UTC
Re: [Xen-devel] Segfault When Building Snapshot xen-3d14ce7 kernel From linux/kernel/git/jeremy/xen.git/shortlog
Ian, Sorry for the delay. The development box in question has older hardware, so it takes a long time to compile a kernel to investigate various possibilities. We were running an earlier version of the 2.6.32.25 pvops kernel, on Xen 4.0.1. Apparently the fault is caused by the -j switch in make, but not always. I can compile the newer kernel single-threaded without any problems (other than I have to run it overnight), without ever reproducing the problem, so it cannot be in kernel source or the tools per se. We have been running (and building Xen) on the newer snapshot xen-3d14ce7 kernel with no problems. Debugging this problem uncovered a different issue with the IBM sHype security features (ACM), so we are focused on that for now. I will start logging more, whenever I build a new pvops kernel, to see if I can collect anything useful. We will also have a look at adding a kernel build to our development stress tests. Sincerely, John On Mar 18, 2011, at 9:28 AM, Ian Campbell wrote:> Are you running the host under Xen when this happens? Which > kernel/hypervisor? Is there anything in your kernel logs?---- What is the formal meaning of the one-line program #include "/dev/tty" J.P. McDermott building 12 Code 5542 mcdermott@itd.nrl.navy.mil Naval Research Laboratory voice: +1 202.404.8301 Washington, DC 20375, US fax: +1 202.404.7942 _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Ian Campbell
2011-Mar-29 09:25 UTC
Re: [Xen-devel] Segfault When Building Snapshot xen-3d14ce7 kernel From linux/kernel/git/jeremy/xen.git/shortlog
On Mon, 2011-03-28 at 13:41 +0100, John McDermott (U.S. Navy Employee) wrote:> Ian, > > Sorry for the delay. The development box in question has older > hardware, so it takes a long time to compile a kernel to investigate > various possibilities. > > We were running an earlier version of the 2.6.32.25 pvops kernel, on > Xen 4.0.1. Apparently the fault is caused by the -j switch in make, > but not always.It''s probably an issue exposed only under stress then.> I can compile the newer kernel single-threaded without any problems > (other than I have to run it overnight), without ever reproducing the > problem, so it cannot be in kernel source or the tools per se. We have > been running (and building Xen) on the newer snapshot xen-3d14ce7 > kernel with no problems.The 2.6.32.25 version was 51f9f525d75? In which case there have been a lot of changes between then and 3d14ce7. Fingers crossed that one of them fixes your issue...> Debugging this problem uncovered a different issue with the IBM sHype > security features (ACM), so we are focused on that for now.In case you didn''t know the ACM support in Xen has been effectively unmaintained for quite some time and was removed from the xen-unstable tree last week (and so won''t be in Xen 4.2 as things stand). The flask module for XSM remains. (it''s not clear if this is the same ACM you were referring to since I''m not sure how sHype fits in here...)> I will start logging more, whenever I build a new pvops kernel, to see > if I can collect anything useful. We will also have a look at adding a > kernel build to our development stress tests.Thanks John, hopefully either the bug won''t happen again or something will show up in the logs.> > Sincerely, > > John > > > On Mar 18, 2011, at 9:28 AM, Ian Campbell wrote: > > > Are you running the host under Xen when this happens? Which > > kernel/hypervisor? Is there anything in your kernel logs? > > ---- > What is the formal meaning of the one-line program > #include "/dev/tty" > > J.P. McDermott building 12 > Code 5542 mcdermott@itd.nrl.navy.mil > Naval Research Laboratory voice: +1 202.404.8301 > Washington, DC 20375, US fax: +1 202.404.7942 > > > > > > > > >_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel