Re: [PATCH v4 06/13] KVM: Move WARN on invalid memslot index to update_memslots()

From: Maciej S. Szmigiero
Date: Wed Aug 18 2021 - 17:43:40 EST


On 18.08.2021 16:35, David Hildenbrand wrote:
On 13.08.21 21:33, Maciej S. Szmigiero wrote:
From: "Maciej S. Szmigiero" <maciej.szmigiero@xxxxxxxxxx>

Since kvm_memslot_move_forward() can theoretically return a negative
memslot index even when kvm_memslot_move_backward() returned a positive one
(and so did not WARN) let's just move the warning to the common code.

Signed-off-by: Maciej S. Szmigiero <maciej.szmigiero@xxxxxxxxxx>
---
  virt/kvm/kvm_main.c | 6 ++++--
  1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index 03ef42d2e421..7000efff1425 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -1293,8 +1293,7 @@ static inline int kvm_memslot_move_backward(struct kvm_memslots *slots,
      struct kvm_memory_slot *mslots = slots->memslots;
      int i;
-    if (WARN_ON_ONCE(slots->id_to_index[memslot->id] == -1) ||
-        WARN_ON_ONCE(!slots->used_slots))
+    if (slots->id_to_index[memslot->id] == -1 || !slots->used_slots)
          return -1;
      /*
@@ -1398,6 +1397,9 @@ static void update_memslots(struct kvm_memslots *slots,
              i = kvm_memslot_move_backward(slots, memslot);
          i = kvm_memslot_move_forward(slots, memslot, i);
+        if (WARN_ON_ONCE(i < 0))
+            return;
+
          /*
           * Copy the memslot to its new position in memslots and update
           * its index accordingly.



Note that WARN_ON_* is frowned upon, because it can result in crashes with panic_on_warn enabled, which is what some distributions do enable.

We tend to work around that by using pr_warn()/pr_warn_once(), avoiding eventually crashing the system when there is a way to continue.


This patch uses WARN_ON_ONCE because:
1) It was used in the old code and the patch merely moves the check
from kvm_memslot_move_backward() to its caller,

2) This chunk of code is wholly replaced by patch 11 from this series
anyway ("Keep memslots in tree-based structures instead of array-based ones").

Thanks,
Maciej