From f12ba0c59fb470378fd054bba8ce424156c4f164 Mon Sep 17 00:00:00 2001 From: Anton Lindqvist Date: Sun, 6 Dec 2020 00:19:43 +0100 Subject: pkg/report: make panic oops on openbsd more stringent (#2315) It looks we got a couple of odd crashes where the syz-executor is probably hanging but the VM is still responsive[1]. Does the hypervisor try to diagnose the VM under such circumstances? Since the VM is still responsive, the input written to the console is echoed back which in turns causes a panic to be detected. If this theory is true, we can avoid detecting such false positives by making the oops indicator disjoint from the input written to the console. [1] https://syzkaller.appspot.com/bug?id=af604b59c590384e9faa00dfc958ef87a922ae71 --- pkg/report/testdata/openbsd/report/34 | 24 ++++++++++++++++++++++++ 1 file changed, 24 insertions(+) create mode 100644 pkg/report/testdata/openbsd/report/34 (limited to 'pkg/report/testdata') diff --git a/pkg/report/testdata/openbsd/report/34 b/pkg/report/testdata/openbsd/report/34 new file mode 100644 index 000000000..085a28184 --- /dev/null +++ b/pkg/report/testdata/openbsd/report/34 @@ -0,0 +1,24 @@ + +OpenBSD/amd64 (ci-openbsd-setuid-9.c.syzkaller.internal) (tty00) + +login: set $lines = 0 +Password: +Login incorrect +login: show panic +Password: +Login incorrect +login: show registers +Password: +Login incorrect +login: ps +Password: +Login incorrect +show malloc +login: Password: +Login incorrect +machine ddbcpu 0 +login: Password: +Login incorrect +machine ddbcpu 1 +login: Password: +Login incorrect -- cgit mrf-deployment