[Bug 583867] New: Soft lockups on domU with high CPU load om dom0
http://bugzilla.novell.com/show_bug.cgi?id=583867 http://bugzilla.novell.com/show_bug.cgi?id=583867#c0 Summary: Soft lockups on domU with high CPU load om dom0 Classification: openSUSE Product: openSUSE 11.2 Version: Final Platform: x86-64 OS/Version: openSUSE 11.2 Status: NEW Severity: Major Priority: P5 - None Component: Xen AssignedTo: jdouglas@novell.com ReportedBy: erik@brakkee.org QAContact: qa@suse.de Found By: --- Blocker: --- User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.1.8) Gecko/20100206 SUSE/2.0.3-0.1.1 SeaMonkey/2.0.3 If I run an opensuse 11.2 xen kernel as dom0 and create an opensuse 11.2 guest (in my case a 32 bit guest), then putting a high load on dom0 results in messages about 'soft lockups' and 'CPU#0 stuck' in the domU. Reproducible: Always Steps to Reproduce: 1. Install 64 bit opensuse 11.2 2. Create HVM guest with 32 bit opensuse 11.2 3. start the guest using xm start and open a console to it 4. Compile a kernel in dom0 with 'make -j 8' to generate some load. Actual Results: DomU freezes as soon as the kernel compilation is under way. Dom0 itself remains responsive. Messages about soft lockups and CPU#<n> stuck start appearing. Expected Results: DomU should have been responsive, albeit a bit slower, but not completely unresponsive.. Using stock kernel 2.6.31.12-0.1-xen (dom0) and 2.6.31.12-0.1-desktop (domU). Hardware Core i7 720QM, 8gb memory, 500GB hard disk. Sony Vaio F11 laptop. Problem appears with both credit scheduler and sedf scheduler. Xen boot option noapic does not help. Manipulating caps and weights with the credit scheduler also does not have any effect. This feature in effect renders xen useless. I am evaluating Xen on my laptop for more serious uses on a server and the current behavior of xen is blocking for that. Will have to consider other virtualization options if this problem is not solved. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c1
--- Comment #1 from Erik Brakkee
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c
Jan Beulich
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c2
Jan Beulich
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c3
--- Comment #3 from Erik Brakkee
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c4
--- Comment #4 from Erik Brakkee
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c5
--- Comment #5 from Jan Beulich
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c6
--- Comment #6 from Erik Brakkee
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c7
--- Comment #7 from Erik Brakkee
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c8
--- Comment #8 from Erik Brakkee
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c9
--- Comment #9 from Erik Brakkee
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c10
--- Comment #10 from Erik Brakkee
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c11
Erik Brakkee
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c12
--- Comment #12 from Erik Brakkee
Hmm, okay - so far, partly from your post on xen-devel, I was under the impression that Dom0 is also misbehaving for you.
Because of the issues with the stock xen kernel of opensuse 11.2 I was trying the xen/master kernel. However, that one gave a problem booting dom0. I am able to boot that kernel now but when I use the xen 3.4.2 hyper visor that I downloaded and compiled (with kernel 2.6.18) I get another problem in starting xend (xend keeps on recursively starting xend), so I can only use the xen-3.4.1 version from SuSE with the xen/master kernel. The problem is then, however, still there. It would also help if I could compile a new xen version and try that, but I don't know how much the xen hypervisor is tied to the kernel version. I would expect that there would only be a dependency of the kernel running as dom0 on the hypervisor interface but that the hypervisor would be independent of the kernel. In summary, I can bootup with xen-3.4.2 but in that case the /etc/init.d/xend script fails horribly. Would it be safe to use the xend binary from the 3.4.2 distribution together with the /etc/init.d/xend script from SuSE (xen-tools-3.4.1_19718_04-2.1.x86_64)? -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c13
--- Comment #13 from Erik Brakkee
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c14
--- Comment #14 from Erik Brakkee
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c15
--- Comment #15 from Jan Beulich
It would also help if I could compile a new xen version and try that, but I don't know how much the xen hypervisor is tied to the kernel version. I would expect that there would only be a dependency of the kernel running as dom0 on the hypervisor interface but that the hypervisor would be independent of the kernel.
Generally, replacing the hypervisor with a newer one should cause no problems (apart from the system becoming completely unsupportable if you do this for other than problem analysis) with either Dom0 or DomU kernels - anything else would be a bug. Xend, however, has to match the underlying hypervisor, so if you update that one you'll also have to make sure you use the corresponding newer tools. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c
Jan Beulich
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c16
Jan Beulich
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c
Jan Beulich
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c17
--- Comment #17 from Erik Brakkee
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c18
Jan Beulich
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c19
--- Comment #19 from Erik Brakkee
A tentative fix for this was just committed, will be available in a future kernel maintenance update - please re-test against that.
I have tried in the past to build my own xen kernel but as I understand it Xen is very closely integrated with the kernel. If you can provide me with a reference to a kernel version with Xen included then I could try (and some instructions on how to build). However, my best hope is in trying it out in a new opensuse 11.3 version. Can I expect the patch to be in there? -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c20
--- Comment #20 from Erik Brakkee
http://bugzilla.novell.com/show_bug.cgi?id=583867
http://bugzilla.novell.com/show_bug.cgi?id=583867#c21
Jan Beulich
https://bugzilla.novell.com/show_bug.cgi?id=583867
https://bugzilla.novell.com/show_bug.cgi?id=583867#c22
Swamp Workflow Management
participants (1)
-
bugzilla_noreply@novell.com