summaryrefslogtreecommitdiffstats
path: root/results/classifier/zero-shot/105/KVM/1502934
diff options
context:
space:
mode:
Diffstat (limited to 'results/classifier/zero-shot/105/KVM/1502934')
-rw-r--r--results/classifier/zero-shot/105/KVM/150293442
1 files changed, 42 insertions, 0 deletions
diff --git a/results/classifier/zero-shot/105/KVM/1502934 b/results/classifier/zero-shot/105/KVM/1502934
new file mode 100644
index 00000000..9cd7f2ec
--- /dev/null
+++ b/results/classifier/zero-shot/105/KVM/1502934
@@ -0,0 +1,42 @@
+KVM: 0.913
+instruction: 0.822
+mistranslation: 0.817
+device: 0.782
+other: 0.770
+graphic: 0.743
+semantic: 0.716
+boot: 0.529
+network: 0.378
+vnc: 0.371
+socket: 0.307
+assembly: 0.262
+
+QEMU does not start when kvm enabled (SMM issue?)
+
+Hi!
+
+QEMU stopped working after "[355023f2010c4df619d88a0dd7012b4b9c74c12c] pc: add SMM property" on my server. It says "Guest has not initialized the display (yet)." and nothing happens. But only if I use -enable-kvm.
+
+However, the problem gone after I hardcoded pc_machine_is_smm_enabled() to always return false (but I have little to no understanding of what SMM really is).
+
+CMD line that reproduces the issue: qemu-system-x86_64 -enable-kvm -display curses . It doesn't work the server, but works perfectly on my laptop :(.
+
+I'm using Arch Linux with all updates.
+Some info:
+Linux machine 4.2.2-1-ARCH #1 SMP PREEMPT Tue Sep 29 22:21:33 CEST 2015 x86_64 GNU/Linux
+Qemu-2.4.0 (tried HEAD as well)
+CPU: Intel(R) Core(TM) i7 CPU 950 @ 3.07GHz
+Some messages from dmesg, just in case:
+[ 6.996297] kvm: VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL does not work properly. Using workaround
+[ 6381.722990] kvm: zapping shadow pages for mmio generation wraparound
+
+
+I'm more than happy to provide additional information if needed.
+
+Cheers,
+Alex
+
+This is a KVM bug, not a QEMU bug. It only happens with ept=1 and unrestricted_guest=0. A workaround is to disable SMM in QEMU (-machine smm=off) or to set kvm_intel's ept parameter to 0.
+
+I'm closing the bug here because it is the wrong tracker (also I know about it, can reproduce, and hope to provide a fix later this week).
+