Xen devel - Oct 2010 - [PATCH 0/2] Xen PV on HVM bug fixes for 2.6.36 rc6

Hi all,
this series is a collection of two serious bug fixes for Linux 2.6.36 rc6:


Stefano Stabellini (2):
      xen: do not set xenstored_ready before xenbus_probe on hvm
      xen: do not initialize PV timers on HVM if !xen_have_vector_callback

 arch/x86/xen/time.c               |    5 +++--
 drivers/xen/xenbus/xenbus_probe.c |    9 ++++++---
 2 files changed, 9 insertions(+), 5 deletions(-)


The first patch fixes a bug in the shutdown_event watcher
registration that causes "xm shutdown" not to work properly for PV on
HVM guests: register_xenstore_notifier should be able to guarantee that
the caller gets notified when xenstore is ready no matter if xenstore is
already up. The right way of fixing register_xenstore_notifier''s bad
behavior is to set xenstored_ready at the right time for PV on HVM
guests too, instead of failing to notify callers when xenstore is
already up.
The second patch fixes an hang at boot that happens when Xen doesn''t
support vector callbacks and the guest has multiple vcpus. Considering
that we cannot get PV timers working on cpus > 0 without vector
callbacks, it is just safer to disable completely PV timers in that
case.

I realize it might be too late for rc6, if that is case then just let me
know and I''ll resubmit the series during the next merge window.

A git branch with these two fixes on top of 2.6.36 rc6 is available
here:

git://xenbits.xen.org/people/sstabellini/linux-pvhvm.git
v2.6.36-rc6-urgent-fixes

Cheers,

Stefano

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

From: Stefano Stabellini <stefano.stabellini@eu.citrix.com>

Register_xenstore_notifier should guarantee that the caller gets
notified even if xenstore is already up.
Therefore we revert "do not notify callers from
register_xenstore_notifier" and set xenstored_read at the right time for
PV on HVM guests too.
In fact in case of PV on HVM guests xenstored is ready only after the
platform pci driver has completed the initialization, so do not set
xenstored_ready before the call to xenbus_probe().

This patch fixes a shutdown_event watcher registration bug that causes
"xm shutdown" not to work properly.

Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Acked-by: Jeremy Fitzhardinge <jeremy@goop.org>
---
 drivers/xen/xenbus/xenbus_probe.c |    9 ++++++---
 1 files changed, 6 insertions(+), 3 deletions(-)

diff --git a/drivers/xen/xenbus/xenbus_probe.c
b/drivers/xen/xenbus/xenbus_probe.c
index 29bac51..d409495 100644
--- a/drivers/xen/xenbus/xenbus_probe.c
+++ b/drivers/xen/xenbus/xenbus_probe.c
@@ -755,7 +755,10 @@ int register_xenstore_notifier(struct notifier_block *nb)
 {
 	int ret = 0;
 
-	blocking_notifier_chain_register(&xenstore_chain, nb);
+	if (xenstored_ready > 0)
+		ret = nb->notifier_call(nb, 0, NULL);
+	else
+		blocking_notifier_chain_register(&xenstore_chain, nb);
 
 	return ret;
 }
@@ -769,7 +772,7 @@ EXPORT_SYMBOL_GPL(unregister_xenstore_notifier);
 
 void xenbus_probe(struct work_struct *unused)
 {
-	BUG_ON((xenstored_ready <= 0));
+	xenstored_ready = 1;
 
 	/* Enumerate devices in xenstore and watch for changes. */
 	xenbus_probe_devices(&xenbus_frontend);
@@ -835,8 +838,8 @@ static int __init xenbus_init(void)
 			xen_store_evtchn = xen_start_info->store_evtchn;
 			xen_store_mfn = xen_start_info->store_mfn;
 			xen_store_interface = mfn_to_virt(xen_store_mfn);
+			xenstored_ready = 1;
 		}
-		xenstored_ready = 1;
 	}
 
 	/* Initialize the interface to xenstore. */
-- 
1.5.6.5


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

From: Stefano Stabellini <stefano.stabellini@eu.citrix.com>

if !xen_have_vector_callback do not initialize PV timer unconditionally
because we still don''t know how many cpus are available and if there is
more than one we won''t be able to receive the timer interrupts on
cpu > 0.

This patch fixes an hang at boot when Xen does not support vector
callbacks and the guest has multiple vcpus.

Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Acked-by: Jeremy Fitzhardinge <jeremy@goop.org>
---
 arch/x86/xen/time.c |    5 +++--
 1 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/arch/x86/xen/time.c b/arch/x86/xen/time.c
index 1a5353a..b2bb5aa 100644
--- a/arch/x86/xen/time.c
+++ b/arch/x86/xen/time.c
@@ -489,8 +489,9 @@ static void xen_hvm_setup_cpu_clockevents(void)
 __init void xen_hvm_init_time_ops(void)
 {
 	/* vector callback is needed otherwise we cannot receive interrupts
-	 * on cpu > 0 */
-	if (!xen_have_vector_callback && num_present_cpus() > 1)
+	 * on cpu > 0 and at this point we don''t know how many cpus are
+	 * available */
+	if (!xen_have_vector_callback)
 		return;
 	if (!xen_feature(XENFEAT_hvm_safe_pvclock)) {
 		printk(KERN_INFO "Xen doesn''t support pvclock on HVM,"
-- 
1.5.6.5


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel