[Xen-devel] Re: [PATCH RFC] do_settime is backwards?!

[Rik van Riel <riel@xxxxxxxxxx>]

This patch fixes an apparent bug in do_settime() that was found while
auditing the code to fix an unrelated bug.

Signed-off-by: Rik van Riel <riel@xxxxxxxxxx>
--- 
I still hope I am wrong :)

On Wed, 6 Aug 2008 17:37:23 -0400
Rik van Riel <riel@xxxxxxxxxx> wrote:

> I hope I've overlooked some detail, but just in case I am
> right here's a patch to reverse the assignments.
> 
> How did this ever work?

Of course, that is still not enough.  A u32 has some
more bits than a billion, so the wc_nsec variable
could be off by as much as 3 seconds...


diff -up xen/arch/x86/time.c.backwards xen/arch/x86/time.c
--- xen/arch/x86/time.c.backwards       2008-08-06 17:33:26.000000000 -0400
+++ xen/arch/x86/time.c 2008-08-06 17:43:46.000000000 -0400
@@ -819,12 +819,16 @@ void do_settime(unsigned long secs, unsi
     u32 y, _wc_sec, _wc_nsec;
     struct domain *d;
 
+    /* Calculate nanoseconds */
     x = (secs * 1000000000ULL) + (u64)nsecs - system_time_base;
+    /* Calculate seconds. */
     y = do_div(x, 1000000000);
+    /* Leave the remainder for the nanosecond field. */
+    x -= (y * 1000000000);
 
     spin_lock(&wc_lock);
-    wc_sec  = _wc_sec  = (u32)x;
-    wc_nsec = _wc_nsec = (u32)y;
+    wc_sec  = _wc_sec  = (u32)y;
+    wc_nsec = _wc_nsec = (u32)x;
     spin_unlock(&wc_lock);
 
     rcu_read_lock(&domlist_read_lock);

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