1 files changed, 94 insertions, 0 deletions
diff --git a/results/classifier/118/TCG/1565 b/results/classifier/118/TCG/1565
new file mode 100644
index 00000000..d518c5c2
--- /dev/null
+++ b/results/classifier/118/TCG/1565
@@ -0,0 +1,94 @@
+TCG: 0.930
+KVM: 0.910
+device: 0.793
+graphic: 0.722
+performance: 0.671
+kernel: 0.625
+network: 0.613
+register: 0.613
+PID: 0.600
+vnc: 0.596
+semantic: 0.590
+socket: 0.568
+files: 0.549
+permissions: 0.545
+ppc: 0.545
+risc-v: 0.484
+architecture: 0.473
+hypervisor: 0.464
+boot: 0.423
+debug: 0.413
+VMM: 0.409
+arm: 0.395
+peripherals: 0.346
+i386: 0.248
+x86: 0.235
+virtual: 0.233
+user-level: 0.208
+mistranslation: 0.201
+assembly: 0.174
+--------------------
+TCG: 0.959
+debug: 0.922
+kernel: 0.557
+KVM: 0.385
+files: 0.105
+user-level: 0.083
+hypervisor: 0.039
+performance: 0.022
+register: 0.021
+PID: 0.014
+architecture: 0.012
+assembly: 0.009
+device: 0.007
+semantic: 0.006
+virtual: 0.004
+VMM: 0.004
+network: 0.002
+graphic: 0.002
+boot: 0.002
+peripherals: 0.002
+risc-v: 0.001
+permissions: 0.001
+vnc: 0.001
+socket: 0.001
+mistranslation: 0.001
+arm: 0.000
+ppc: 0.000
+x86: 0.000
+i386: 0.000
+
+s390x TCG migration failure
+Description of problem:
+We're seeing failures running s390x migration kvm-unit-tests tests with TCG.
+
+Some initial findings:
+
+What seems to be happening is that after migration a control block header accessed by the test code is all zeros which causes an unexpected exception.
+
+I did a bisection which points to c8df4a7aef ("migration: Split save_live_pending() into state_pending_*") as the culprit.
+The migration issue persists after applying the fix e264705012 ("migration: I messed state_pending_exact/estimate") on top of c8df4a7aef.
+
+Applying
+
+```
+diff --git a/migration/ram.c b/migration/ram.c
+index 56ff9cd29d..2dc546cf28 100644
+--- a/migration/ram.c
++++ b/migration/ram.c
+@@ -3437,7 +3437,7 @@ static void ram_state_pending_exact(void *opaque, uint64_t max_size,
+ 
+     uint64_t remaining_size = rs->migration_dirty_pages * TARGET_PAGE_SIZE;
+ 
+-    if (!migration_in_postcopy()) {
++    if (!migration_in_postcopy() && remaining_size < max_size) {
+         qemu_mutex_lock_iothread();
+         WITH_RCU_READ_LOCK_GUARD() {
+             migration_bitmap_sync_precopy(rs);
+```
+on top fixes or hides the issue. (The comparison was removed by c8df4a7aef.)
+
+I arrived at this by experimentation, I haven't looked into why this makes a difference.
+Steps to reproduce:
+1. Run ACCEL=tcg ./run_tests.sh migration-skey-sequential with current QEMU master
+2. Repeat until the test fails (doesn't happen every time, but still easy to reproduce)