The individual drivers'' remove functions could legitimately attempt to
access this information (for logging messages if nothing else). Note
that I did not in fact observe a problem anywhere, but I came across
this while looking into the reasons for what turned out to need the
fix at https://lkml.org/lkml/2012/3/5/336 to vsprintf().
Signed-off-by: Jan Beulich <jbeulich@suse.com>
---
drivers/xen/xenbus/xenbus_probe.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
--- 3.3-rc6/drivers/xen/xenbus/xenbus_probe.c
+++ 3.3-rc6-xenbus-remove-details/drivers/xen/xenbus/xenbus_probe.c
@@ -257,11 +257,12 @@ int xenbus_dev_remove(struct device *_de
DPRINTK("%s", dev->nodename);
free_otherend_watch(dev);
- free_otherend_details(dev);
if (drv->remove)
drv->remove(dev);
+ free_otherend_details(dev);
+
xenbus_switch_state(dev, XenbusStateClosed);
return 0;
}
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
Konrad Rzeszutek Wilk
2012-Mar-14 16:33 UTC
Re: [PATCH] xenbus: don''t free other end details too early
On Mon, Mar 05, 2012 at 05:11:31PM +0000, Jan Beulich wrote:> The individual drivers'' remove functions could legitimately attempt to > access this information (for logging messages if nothing else). Note > that I did not in fact observe a problem anywhere, but I came across > this while looking into the reasons for what turned out to need the > fix at https://lkml.org/lkml/2012/3/5/336 to vsprintf().applied.> > Signed-off-by: Jan Beulich <jbeulich@suse.com> > > --- > drivers/xen/xenbus/xenbus_probe.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > --- 3.3-rc6/drivers/xen/xenbus/xenbus_probe.c > +++ 3.3-rc6-xenbus-remove-details/drivers/xen/xenbus/xenbus_probe.c > @@ -257,11 +257,12 @@ int xenbus_dev_remove(struct device *_de > DPRINTK("%s", dev->nodename); > > free_otherend_watch(dev); > - free_otherend_details(dev); > > if (drv->remove) > drv->remove(dev); > > + free_otherend_details(dev); > + > xenbus_switch_state(dev, XenbusStateClosed); > return 0; > } > > >> The individual drivers'' remove functions could legitimately attempt to > access this information (for logging messages if nothing else). Note > that I did not in fact observe a problem anywhere, but I came across > this while looking into the reasons for what turned out to need the > fix at https://lkml.org/lkml/2012/3/5/336 to vsprintf(). > > Signed-off-by: Jan Beulich <jbeulich@suse.com> > > --- > drivers/xen/xenbus/xenbus_probe.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > --- 3.3-rc6/drivers/xen/xenbus/xenbus_probe.c > +++ 3.3-rc6-xenbus-remove-details/drivers/xen/xenbus/xenbus_probe.c > @@ -257,11 +257,12 @@ int xenbus_dev_remove(struct device *_de > DPRINTK("%s", dev->nodename); > > free_otherend_watch(dev); > - free_otherend_details(dev); > > if (drv->remove) > drv->remove(dev); > > + free_otherend_details(dev); > + > xenbus_switch_state(dev, XenbusStateClosed); > return 0; > }