a217ee0e |
From: Dave Jones <davej@redhat.com>
Date: Tue, 24 Jun 2014 08:43:34 -0400
Subject: [PATCH] watchdog: Disable watchdog on virtual machines.
For various reasons, VMs seem to trigger the soft lockup detector a lot,
in cases where it's just not possible for a lockup to occur.
(Example: https://bugzilla.redhat.com/show_bug.cgi?id=971139)
In some cases it seems that the host just never scheduled the app running
the VM for a very long time (Could be the host was under heavy load).
Just disable the detector on VMs.
Bugzilla: 971139
Upstream-status: Fedora mustard for now
Signed-off-by: Dave Jones <davej@redhat.com>
---
kernel/watchdog.c | 29 +++++++++++++++++++++++++++++
1 file changed, 29 insertions(+)
diff --git a/kernel/watchdog.c b/kernel/watchdog.c
index a6ffa43f2993..b378b762844a 100644
--- a/kernel/watchdog.c
+++ b/kernel/watchdog.c |
a217ee0e |
__setup("hardlockup_all_cpu_backtrace=", hardlockup_all_cpu_backtrace_setup);
#endif
+static int disable_watchdog(const struct dmi_system_id *d)
+{
+ printk(KERN_INFO "watchdog: disabled (inside virtual machine)\n");
+ watchdog_user_enabled = 0;
+ return 0;
+}
+
+static const struct dmi_system_id watchdog_virt_dmi_table[] = {
+ {
+ .callback = disable_watchdog,
+ .ident = "VMware",
+ .matches = {
+ DMI_MATCH(DMI_SYS_VENDOR, "VMware, Inc."),
+ },
+ },
+ {
+ .callback = disable_watchdog,
+ .ident = "Bochs",
+ .matches = {
+ DMI_MATCH(DMI_SYS_VENDOR, "Bochs"),
+ },
+ },
+ {}
+};
+
+
/*
* Hard-lockup warnings should be triggered after just a few seconds. Soft-
* lockups can have false positives under extreme conditions. So we generally |