linux-kernel - Re: propagating vmgenid outward and upward

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

Date:   Wed, 2 Mar 2022 16:36:49 +0100
From:   "Jason A. Donenfeld" <Jason@...c4.com>
To:     "Michael S. Tsirkin" <mst@...hat.com>
Cc:     Laszlo Ersek <lersek@...hat.com>,
        LKML <linux-kernel@...r.kernel.org>,
        KVM list <kvm@...r.kernel.org>,
        QEMU Developers <qemu-devel@...gnu.org>,
        linux-hyperv@...r.kernel.org,
        Linux Crypto Mailing List <linux-crypto@...r.kernel.org>,
        Alexander Graf <graf@...zon.com>,
        "Michael Kelley (LINUX)" <mikelley@...rosoft.com>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        adrian@...ity.io,
        Daniel P. Berrangé <berrange@...hat.com>,
        Dominik Brodowski <linux@...inikbrodowski.net>,
        Jann Horn <jannh@...gle.com>,
        "Rafael J. Wysocki" <rafael@...nel.org>,
        "Brown, Len" <len.brown@...el.com>, Pavel Machek <pavel@....cz>,
        Linux PM <linux-pm@...r.kernel.org>,
        Colm MacCarthaigh <colmmacc@...zon.com>,
        Theodore Ts'o <tytso@....edu>, Arnd Bergmann <arnd@...db.de>
Subject: Re: propagating vmgenid outward and upward

Hi Michael,

On Wed, Mar 02, 2022 at 10:20:25AM -0500, Michael S. Tsirkin wrote:
> So writing some code:
> 
> 1:
> 	put plaintext in a buffer
> 	put a key in a buffer
> 	put the nonce for that encryption in a buffer
> 
> 	if vm gen id != stored vm gen id
> 		stored vm gen id = vm gen id
> 		goto 1
> 
> I think this is race free, but I don't see why does it matter whether we
> read gen id atomically or not.

Because that 16 byte read of vmgenid is not atomic. Let's say you read
the first 8 bytes, and then the VM is forked. In the forked VM, the next
8 bytes are the same as last time, but the first 8 bytes, which you
already read, have changed. In that case, your != becomes a ==, and the
test fails.

This is one of those fundamental things of "unique ID" vs "generation
counter word".

Anyway, per your request in your last email, I wrote some code for this,
which may or may not be totally broken, and only works on 64-bit x86,
which is really the best possible case in terms of performance. And even
so, it's not great.

Jason

--------8<------------------------

diff --git a/drivers/net/wireguard/noise.c b/drivers/net/wireguard/noise.c
index 720952b92e78..250b8973007d 100644
--- a/drivers/net/wireguard/noise.c
+++ b/drivers/net/wireguard/noise.c
@@ -106,6 +106,7 @@ static struct noise_keypair *keypair_create(struct wg_peer *peer)
 	keypair->entry.type = INDEX_HASHTABLE_KEYPAIR;
 	keypair->entry.peer = peer;
 	kref_init(&keypair->refcount);
+	keypair->vmgenid = vmgenid_read_atomic();
 	return keypair;
 }

diff --git a/drivers/net/wireguard/noise.h b/drivers/net/wireguard/noise.h
index c527253dba80..0add240a14a0 100644
--- a/drivers/net/wireguard/noise.h
+++ b/drivers/net/wireguard/noise.h
@@ -27,10 +27,13 @@ struct noise_symmetric_key {
 	bool is_valid;
 };

+extern __uint128_t vmgenid_read_atomic(void);
+
 struct noise_keypair {
 	struct index_hashtable_entry entry;
 	struct noise_symmetric_key sending;
 	atomic64_t sending_counter;
+	__uint128_t vmgenid;
 	struct noise_symmetric_key receiving;
 	struct noise_replay_counter receiving_counter;
 	__le32 remote_index;
diff --git a/drivers/net/wireguard/send.c b/drivers/net/wireguard/send.c
index 5368f7c35b4b..40d016be59e3 100644
--- a/drivers/net/wireguard/send.c
+++ b/drivers/net/wireguard/send.c
@@ -381,6 +381,9 @@ void wg_packet_send_staged_packets(struct wg_peer *peer)
 			goto out_invalid;
 	}

+	if (keypair->vmgenid != vmgenid_read_atomic())
+		goto out_invalid;
+
 	packets.prev->next = NULL;
 	wg_peer_get(keypair->entry.peer);
 	PACKET_CB(packets.next)->keypair = keypair;
diff --git a/drivers/virt/vmgenid.c b/drivers/virt/vmgenid.c
index 0ae1a39f2e28..c122fae1d494 100644
--- a/drivers/virt/vmgenid.c
+++ b/drivers/virt/vmgenid.c
@@ -21,6 +21,21 @@ struct vmgenid_state {
 	u8 this_id[VMGENID_SIZE];
 };

+static __uint128_t *val;
+
+__uint128_t vmgenid_read_atomic(void)
+{
+	__uint128_t ret = 0;
+	if (!val)
+		return 0;
+	asm volatile("lock cmpxchg16b %1"
+		     : "+A"(ret)
+		     : "m"(*val), "b"(0), "c"(0)
+		     : "cc");
+	return ret;
+}
+EXPORT_SYMBOL(vmgenid_read_atomic);
+
 static int vmgenid_add(struct acpi_device *device)
 {
 	struct acpi_buffer parsed = { ACPI_ALLOCATE_BUFFER };
@@ -50,6 +65,7 @@ static int vmgenid_add(struct acpi_device *device)
 	phys_addr = (obj->package.elements[0].integer.value << 0) |
 		    (obj->package.elements[1].integer.value << 32);
 	state->next_id = devm_memremap(&device->dev, phys_addr, VMGENID_SIZE, MEMREMAP_WB);
+	val = (__uint128_t *)state->next_id;
 	if (IS_ERR(state->next_id)) {
 		ret = PTR_ERR(state->next_id);
 		goto out;