linux-kernel - [PATCH] selftests: kvm/set_memory_region

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-Id: <0fdddb94bb0e31b7da129a809a308d91c10c0b5e.1606941224.git.maciej.szmigiero@oracle.com>
Date:   Wed,  2 Dec 2020 21:35:36 +0100
From:   "Maciej S. Szmigiero" <mail@...iej.szmigiero.name>
To:     Paolo Bonzini <pbonzini@...hat.com>, Shuah Khan <shuah@...nel.org>
Cc:     Sean Christopherson <seanjc@...gle.com>, kvm@...r.kernel.org,
        linux-kselftest@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: [PATCH] selftests: kvm/set_memory_region_test: Fix race in move region test

From: "Maciej S. Szmigiero" <maciej.szmigiero@...cle.com>

The current memory region move test correctly handles the situation that
the second (realigning) memslot move operation would temporarily trigger
MMIO until it completes, however it does not handle the case in which the
first (misaligning) move operation does this, too.
This results in false test assertions in case it does so.

Fix this by handling temporary MMIO from the first memslot move operation
in the test guest code, too.

Fixes: 8a0639fe9201 ("KVM: sefltests: Add explicit synchronization to move mem region test")
Signed-off-by: Maciej S. Szmigiero <maciej.szmigiero@...cle.com>
---
    The race is pretty hard to trigger on the current KVM memslot code,
    to trigger it reliably an extra delay in memslot move op is needed:
    --- a/virt/kvm/kvm_main.c
    +++ b/virt/kvm/kvm_main.c
    @@ -1173,7 +1173,7 @@ static struct kvm_memslots *kvm_dup_memslots(struct kvm_memslots *old,
    
            return slots;
     }
    -
    +#include <linux/delay.h>
     static int kvm_set_memslot(struct kvm *kvm,
                               const struct kvm_userspace_memory_region *mem,
                               struct kvm_memory_slot *old,
    @@ -1212,6 +1212,8 @@ static int kvm_set_memslot(struct kvm *kvm,
                     *      - kvm_is_visible_gfn (mmu_check_root)
                     */
                    kvm_arch_flush_shadow_memslot(kvm, slot);
    +
    +               if (change == KVM_MR_MOVE) mdelay(100);
            }
    
            r = kvm_arch_prepare_memory_region(kvm, new, mem, change);

 .../selftests/kvm/set_memory_region_test.c      | 17 +++++++++++++----
 1 file changed, 13 insertions(+), 4 deletions(-)

diff --git a/tools/testing/selftests/kvm/set_memory_region_test.c b/tools/testing/selftests/kvm/set_memory_region_test.c
index b3ece55a2da6..6f441dd9f33c 100644
--- a/tools/testing/selftests/kvm/set_memory_region_test.c
+++ b/tools/testing/selftests/kvm/set_memory_region_test.c
@@ -156,14 +156,23 @@ static void guest_code_move_memory_region(void)
 	GUEST_SYNC(0);
 
 	/*
-	 * Spin until the memory region is moved to a misaligned address.  This
-	 * may or may not trigger MMIO, as the window where the memslot is
-	 * invalid is quite small.
+	 * Spin until the memory region starts getting moved to a
+	 * misaligned address.
+	 * Every region move may or may not trigger MMIO, as the
+	 * window where the memslot is invalid is usually quite small.
 	 */
 	val = guest_spin_on_val(0);
 	GUEST_ASSERT_1(val == 1 || val == MMIO_VAL, val);
 
-	/* Spin until the memory region is realigned. */
+	/* Spin until the misaligning memory region move completes. */
+	val = guest_spin_on_val(MMIO_VAL);
+	GUEST_ASSERT_1(val == 1 || val == 0, val);
+
+	/* Spin until the memory region starts to get re-aligned. */
+	val = guest_spin_on_val(0);
+	GUEST_ASSERT_1(val == 1 || val == MMIO_VAL, val);
+
+	/* Spin until the re-aligning memory region move completes. */
 	val = guest_spin_on_val(MMIO_VAL);
 	GUEST_ASSERT_1(val == 1, val);