Hello list i am not sure if this is related, but i discovered it after installing xen: i''ve got the following message in dmesg: Badness in local_bh_enable at kernel/softirq.c:140 [<c011f740>] local_bh_enable+0x80/0x90 [<c02572f9>] skb_checksum+0x129/0x2b0 [<c02c7a86>] schedule+0x306/0x530 [<c029036c>] udp_poll+0x9c/0x150 [<c0252769>] sock_poll+0x29/0x40 [<c016a079>] do_pollfd+0x89/0x90 [<c016a0ea>] do_poll+0x6a/0xd0 [<c016a296>] sys_poll+0x146/0x230 [<c011edfb>] sys_gettimeofday+0x3b/0x80 [<c0169670>] __pollwait+0x0/0xd0 [<c01096b8>] syscall_call+0x7/0xb google didnot gives me any useful pointer. i tried different network card i recompiled a fresh kernel. i switch from xen-2.0.4 to xen-2.0.5 i still got this problem. it doesnot seems to be harmful but i just dont like it. any ideas ? regards -- Thomas Constans www.opendoor.fr 06 23 37 87 85 04 78 68 17 34 _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> i am not sure if this is related, but i discovered it after installing > xen: i''ve got the following message in dmesg: > > Badness in local_bh_enable at kernel/softirq.c:140 > [<c011f740>] local_bh_enable+0x80/0x90 [<c02572f9>] > skb_checksum+0x129/0x2b0 [<c02c7a86>] schedule+0x306/0x530 > [<c029036c>] udp_poll+0x9c/0x150 [<c0252769>] > sock_poll+0x29/0x40 [<c016a079>] do_pollfd+0x89/0x90 > [<c016a0ea>] do_poll+0x6a/0xd0 [<c016a296>] > sys_poll+0x146/0x230 [<c011edfb>] sys_gettimeofday+0x3b/0x80 > [<c0169670>] __pollwait+0x0/0xd0 [<c01096b8>] syscall_call+0x7/0xbI''ll wager that you''re trying to use modules compiled for a non-Xen kernel.> it doesnot seems to be harmful but i just dont like it.You''re living very dangerously. When building modules outside of the kernel tree make sure you use ''make ARCH=xen''. Ian _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Le vendredi 01 avril 2005 à 10:08 +0100, Ian Pratt a écrit :> > i am not sure if this is related, but i discovered it after installing > > xen: i''ve got the following message in dmesg: > > > > Badness in local_bh_enable at kernel/softirq.c:140 > > [<c011f740>] local_bh_enable+0x80/0x90 [<c02572f9>] > > skb_checksum+0x129/0x2b0 [<c02c7a86>] schedule+0x306/0x530 > > [<c029036c>] udp_poll+0x9c/0x150 [<c0252769>] > > sock_poll+0x29/0x40 [<c016a079>] do_pollfd+0x89/0x90 > > [<c016a0ea>] do_poll+0x6a/0xd0 [<c016a296>] > > sys_poll+0x146/0x230 [<c011edfb>] sys_gettimeofday+0x3b/0x80 > > [<c0169670>] __pollwait+0x0/0xd0 [<c01096b8>] syscall_call+0x7/0xb > > I''ll wager that you''re trying to use modules compiled for a non-Xen > kernel. > > > it doesnot seems to be harmful but i just dont like it. > > You''re living very dangerously. > > When building modules outside of the kernel tree make sure you use ''make > ARCH=xen''.thanks for the quick answer. so what is the proper procedure for building a xen kernel ? this is what i do, knowing that xen sources are located in /xen/xen-2.0 : # cd /xen/xen-2.0/linux-2.6.10-xen0/ # make menuconfig ARCH=xen ... # cd /xen/xen-2.0/ # make # ./install.sh am i missing something ? regards _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> > I''ll wager that you''re trying to use modules compiled for a non-Xen > > kernel. > so what is the proper procedure for building a xen kernel ?Are you sure you''re not picking up a module from another kernel build?> this is what i do, knowing that xen sources are located in > /xen/xen-2.0 : > > # cd /xen/xen-2.0/linux-2.6.10-xen0/ > # make menuconfig ARCH=xen[It''s normal to use ''make ARCH=xen menuconfig'', but I guess the above works]> # cd /xen/xen-2.0/ > # make > # ./install.sh > > am i missing something ?Are you using any 3rd party modules not in the standard tree? What modifications did you make to the default xen0 config? It''s possible you enabled something that hasn''t been tested. Thanks, Ian _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
I had this problem a while back. Went away when I cleaned out all the modules and compiled them with ARCH=xen. On Apr 1, 2005 9:21 PM, mlist <mlist@opendoor.fr> wrote:> > so what is the proper procedure for building a xen kernel ? > > this is what i do, knowing that xen sources are located > in /xen/xen-2.0 : > > # cd /xen/xen-2.0/linux-2.6.10-xen0/ > # make menuconfig ARCH=xen > ... > # cd /xen/xen-2.0/ > # make > # ./install.sh > > am i missing something ?cp -r /lid/modules/::kernel-version::. For good measure also do a: mv .config _config; make distclean; mv _config .config; make ARCH=xen oldconfig before building your kernel. Nicholas _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Le samedi 02 avril 2005 à 08:54 +1200, Nicholas Lee a écrit :> I had this problem a while back. Went away when I cleaned out all the > modules and compiled them with ARCH=xen.i found out that dom0 kernel starts loggins this error when i start one of my domU i cleaned the whole build tree, erased lib/modules on client, but i still got this error. i recompiled client kernel by doing "make" in xen source tree, after rm -fr linux-2.6.10-xenU so i dont know what i am missing here any idea ? _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> Le samedi 02 avril 2005 à 08:54 +1200, Nicholas Lee a écrit : > > I had this problem a while back. Went away when I cleaned > out all the > > modules and compiled them with ARCH=xen. > > i found out that dom0 kernel starts loggins this error when i > start one of my domU > > i cleaned the whole build tree, erased lib/modules on client, > but i still got this error.It''s possible that the out-of-tree modules you are building are not respecting the ARCH=xen target. I''ll bet if you run them through objdump -d you''ll find ''cli'' instructions. Ian> i recompiled client kernel by doing "make" in xen source > tree, after rm -fr linux-2.6.10-xenU > > so i dont know what i am missing here > > any idea ? > > > _______________________________________________ > Xen-users mailing list > Xen-users@lists.xensource.com > http://lists.xensource.com/xen-users >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Le mardi 05 avril 2005 à 12:42 +0100, Ian Pratt a écrit :> > It''s possible that the out-of-tree modules you are building are not respecting the ARCH=xen target. > > I''ll bet if you run them through objdump -d you''ll find ''cli'' instructions. >i dont have any out-of-tree modules. investigating further, i found ou the following: being a router, the domainU that is the causes for the error logging has got 2 nic declaration in it config file ( vif declaration ). out of curiosity, i removed one of the vif declaration and booted the domain0 that apparently solved the error/log problem. but since i removed one of the nic declaration, how come i still got 2 nic showing in this domainU ???? _______________________________________________ Xen-fr mailing list Xen-fr@lists.xensource.com http://lists.xensource.com/xen-fr
Le mardi 05 avril 2005 à 14:45 +0200, mlist a écrit :> but since i removed one of the nic declaration, how come i still got 2 > nic showing in this domainU ???? >replying to myself: i still got 2 nic because i got "nic=2" in my xen config file and i still got this error. that means if i specify more than one nic for a given domain, i get the following messages in kern.log: Badness in local_bh_enable at kernel/softirq.c:140 [<c011f740>] local_bh_enable+0x80/0x90 [<c02c2fb9>] skb_checksum+0x129/0x2b0 [<c0333746>] schedule+0x306/0x530 [<c02fc02c>] udp_poll+0x9c/0x150 [<c02be429>] sock_poll+0x29/0x40 [<c016a079>] do_pollfd+0x89/0x90 [<c016a0ea>] do_poll+0x6a/0xd0 [<c016a296>] sys_poll+0x146/0x230 [<c011edfb>] sys_gettimeofday+0x3b/0x80 [<c0169670>] __pollwait+0x0/0xd0 [<c01096b8>] syscall_call+0x7/0xb as soon as there is some kind of network activity is this a xen bug ? or am i missing something ?> > _______________________________________________ > Xen-users mailing list > Xen-users@lists.xensource.com > http://lists.xensource.com/xen-users_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Apr 6, 2005 2:00 AM, mlist <mlist@opendoor.fr> wrote:> is this a xen bug ?Probably not. I had a very similar problem to this a while back. Check the mailing list archive. My problem was solved indirectly by upgrading from 2.0.4 to 2.0.5 testing and a clean proper recompile of both dom0 and domU kernels. In the process I removed pretty much all the unneed modules from dom0. Nicholas _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Le jeudi 07 avril 2005 à 09:25 +1200, Nicholas Lee a écrit :> On Apr 6, 2005 2:00 AM, mlist <mlist@opendoor.fr> wrote: > > is this a xen bug ? > > Probably not. I had a very similar problem to this a while back. Check > the mailing list archive. > > My problem was solved indirectly by upgrading from 2.0.4 to 2.0.5 > testing and a clean proper recompile of both dom0 and domU kernels. > In the process I removed pretty much all the unneed modules from dom0.i have already done this; to no avail. i put back all xen hosts in same subnet as dom0, so there is no more multiple virtual nics | routing involved, and the problems disapeared. i soon i as i can get some spare time, i''ll investigate this further cheers _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> i put back all xen hosts in same subnet as dom0, so there is > no more multiple virtual nics | routing involved, and the > problems disapeared. > > i soon i as i can get some spare time, i''ll investigate this furtherI''ll be really interested to hear what''s going on here. I know of a number of people that use multiple NICs and I haven''t heard of such reports before. Thanks, Ian _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Le jeudi 07 avril 2005 à 17:03 +0100, Ian Pratt a écrit :> I''ll be really interested to hear what''s going on here. I know of a > number of people that use multiple NICs and I haven''t heard of such > reports before.i am planning to do some thorough testing next week. cheers _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Le vendredi 08 avril 2005 à 08:24 +0200, mlist a écrit :> Le jeudi 07 avril 2005 à 17:03 +0100, Ian Pratt a écrit : > > I''ll be really interested to hear what''s going on here. I know of a > > number of people that use multiple NICs and I haven''t heard of such > > reports before. > > i am planning to do some thorough testing next week.i have unfortunately not been able to pinpoint the causes of the problem. i either suspect some kind of hardware issue and|or kernel options. i have finally given up and switch all xen domains on an other machines where they are happily serving us dns, groupware, database, web and ldap services keep up the great work _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users