[v4] x86/PoD: tie together P2M update and increment of entry count

Commit Message

Jan Beulich March 13, 2024, 2 p.m. UTC

When not holding the PoD lock across the entire region covering P2M
update and stats update, the entry count - if to be incorrect at all -
should indicate too large a value in preference to a too small one, to
avoid functions bailing early when they find the count is zero. However,
instead of moving the increment ahead (and adjust back upon failure),
extend the PoD-locked region.

Fixes: 99af3cd40b6e ("x86/mm: Rework locking in the PoD layer")
Signed-off-by: Jan Beulich <[e-mail]>
---
v4: Shrink locked region a little again, where possible.
v3: Extend locked region instead. Add Fixes: tag.
v2: Add comments.

Comments

George Dunlap March 13, 2024, 4:31 p.m. UTC | #1

On Wed, Mar 13, 2024 at 2:00 PM Jan Beulich <[e-mail]> wrote:
>
> When not holding the PoD lock across the entire region covering P2M
> update and stats update, the entry count - if to be incorrect at all -
> should indicate too large a value in preference to a too small one, to
> avoid functions bailing early when they find the count is zero. However,
> instead of moving the increment ahead (and adjust back upon failure),
> extend the PoD-locked region.
>
> Fixes: 99af3cd40b6e ("x86/mm: Rework locking in the PoD layer")
> Signed-off-by: Jan Beulich <[e-mail]>

Would you mind commenting on why you went with multiple unlocks,
rather than multiple if statements?

e.g.,

```
rc = p2m_set_entry(...);

/* Do the pod entry adjustment while holding the lock on success */
if ( rc == 0 ) {
 /* adjust pod entries */
}

pod_unlock(p2m);

/* Do the rest of the clean-up and error handling */
if (rc == 0 ) {
```

Just right now the multiple unlocks makes me worry that we may forget
one at some point.

 -George

Jan Beulich March 14, 2024, 6:59 a.m. UTC | #2

On 13.03.2024 17:31, George Dunlap wrote:
> On Wed, Mar 13, 2024 at 2:00 PM Jan Beulich <[e-mail]> wrote:
>>
>> When not holding the PoD lock across the entire region covering P2M
>> update and stats update, the entry count - if to be incorrect at all -
>> should indicate too large a value in preference to a too small one, to
>> avoid functions bailing early when they find the count is zero. However,
>> instead of moving the increment ahead (and adjust back upon failure),
>> extend the PoD-locked region.
>>
>> Fixes: 99af3cd40b6e ("x86/mm: Rework locking in the PoD layer")
>> Signed-off-by: Jan Beulich <[e-mail]>
> 
> Would you mind commenting on why you went with multiple unlocks,
> rather than multiple if statements?

Simply because what I did I view as more logical a code structure
than ...

> e.g.,
> 
> ```
> rc = p2m_set_entry(...);
> 
> /* Do the pod entry adjustment while holding the lock on success */
> if ( rc == 0 ) {
>  /* adjust pod entries */
> }
> 
> pod_unlock(p2m);
> 
> /* Do the rest of the clean-up and error handling */
> if (rc == 0 ) {

... this, ...

> Just right now the multiple unlocks makes me worry that we may forget
> one at some point.

... despite this possible concern. But well, if going the other route
is what it takes to finally get this in, so be it.

Jan

Patch

--- a/xen/arch/x86/mm/p2m-pod.c
+++ b/xen/arch/x86/mm/p2m-pod.c
@@ -1348,12 +1348,19 @@  mark_populate_on_demand(struct domain *d
         }
     }
 
+    /*
+     * P2M update and stats increment need to collectively be under PoD lock,
+     * to prevent code elsewhere observing PoD entry count being zero despite
+     * there actually still being PoD entries (created by the p2m_set_entry()
+     * invocation below).
+     */
+    pod_lock(p2m);
+
     /* Now, actually do the two-way mapping */
     rc = p2m_set_entry(p2m, gfn, INVALID_MFN, order,
                        p2m_populate_on_demand, p2m->default_access);
     if ( rc == 0 )
     {
-        pod_lock(p2m);
         p2m->pod.entry_count += 1UL << order;
         p2m->pod.entry_count -= pod_count;
         BUG_ON(p2m->pod.entry_count < 0);
@@ -1363,6 +1370,8 @@  mark_populate_on_demand(struct domain *d
     }
     else if ( order )
     {
+        pod_unlock(p2m);
+
         /*
          * If this failed, we can't tell how much of the range was changed.
          * Best to crash the domain.
@@ -1372,6 +1381,8 @@  mark_populate_on_demand(struct domain *d
                d, gfn_l, order, rc);
         domain_crash(d);
     }
+    else
+        pod_unlock(p2m);
 
 out:
     gfn_unlock(p2m, gfn, order);