[Qemu-devel] [PATCH] virtio-net: fix wrong size of vlan filter table

[Qemu-devel] [PATCH] virtio-net: fix wrong size of vlan filter table

[Amos Kong]

The MAX_VLAN is 4096, currently the vlan filter table has 512
(4096 >> 3) entries, it's wrong.

One entry in vlan filter table can indicate 32(1 << 5) vlans,
so the table should have 128 (4096 >> 5) entries.

Signed-off-by: Amos Kong <akong@redhat.com>
---
btw, it would be simple to use an uint32 number to indicate
all vlans (same as igbvf). I found e1000 uses the same table.
Nothing needs to change here, it would cause migration issue.
---
 hw/net/virtio-net.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c
index d4ef21f..6dedb97 100644
--- a/hw/net/virtio-net.c
+++ b/hw/net/virtio-net.c
@@ -293,7 +293,7 @@ static void virtio_net_reset(VirtIODevice *vdev)
     n->mac_table.uni_overflow = 0;
     memset(n->mac_table.macs, 0, MAC_TABLE_ENTRIES * ETH_ALEN);
     memcpy(&n->mac[0], &n->nic->conf->macaddr, sizeof(n->mac));
-    memset(n->vlans, 0, MAX_VLAN >> 3);
+    memset(n->vlans, 0, MAX_VLAN >> 5);
 }
 
 static void peer_test_vnet_hdr(VirtIONet *n)
@@ -1246,7 +1246,7 @@ static void virtio_net_save(QEMUFile *f, void *opaque)
     qemu_put_byte(f, n->allmulti);
     qemu_put_be32(f, n->mac_table.in_use);
     qemu_put_buffer(f, n->mac_table.macs, n->mac_table.in_use * ETH_ALEN);
-    qemu_put_buffer(f, (uint8_t *)n->vlans, MAX_VLAN >> 3);
+    qemu_put_buffer(f, (uint8_t *)n->vlans, MAX_VLAN >> 5);
     qemu_put_be32(f, n->has_vnet_hdr);
     qemu_put_byte(f, n->mac_table.multi_overflow);
     qemu_put_byte(f, n->mac_table.uni_overflow);
@@ -1316,7 +1316,7 @@ static int virtio_net_load(QEMUFile *f, void *opaque, int version_id)
     }
  
     if (version_id >= 6)
-        qemu_get_buffer(f, (uint8_t *)n->vlans, MAX_VLAN >> 3);
+        qemu_get_buffer(f, (uint8_t *)n->vlans, MAX_VLAN >> 5);
 
     if (version_id >= 7) {
         if (qemu_get_be32(f) && !peer_has_vnet_hdr(n)) {
-- 
1.8.1.4

Re: [Qemu-devel] [PATCH] virtio-net: fix wrong size of vlan filter table

[Amos Kong]

On Wed, Jun 05, 2013 at 02:13:48PM +0800, Amos Kong wrote:
> The MAX_VLAN is 4096, currently the vlan filter table has 512
> (4096 >> 3) entries, it's wrong.
> 
> One entry in vlan filter table can indicate 32(1 << 5) vlans,
> so the table should have 128 (4096 >> 5) entries.
> 
> Signed-off-by: Amos Kong <akong@redhat.com>
> ---

This patch breaks the migration.

> btw, it would be simple to use an uint32 number to indicate
> all vlans (same as igbvf). I found e1000 uses the same table.
> Nothing needs to change here, it would cause migration issue.

The comment is wrong, igbvf also uses an array.

Will post a v2 with migration fix.

> ---
>  hw/net/virtio-net.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 

-- 
			Amos.

Re: [Qemu-devel] [PATCH] virtio-net: fix wrong size of vlan filter table

[Amos Kong]

On Wed, Jun 05, 2013 at 03:21:30PM +0800, Amos Kong wrote:
> On Wed, Jun 05, 2013 at 02:13:48PM +0800, Amos Kong wrote:
> > The MAX_VLAN is 4096, currently the vlan filter table has 512
> > (4096 >> 3) entries, it's wrong.
> > 
> > One entry in vlan filter table can indicate 32(1 << 5) vlans,
> > so the table should have 128 (4096 >> 5) entries.

The patch is wrong, one entry takes 4 bytes. Sorry for the noise.

-- 
			Amos.

Re: [Qemu-devel] [PATCH] virtio-net: fix wrong size of vlan filter table

[Stefan Hajnoczi]

On Wed, Jun 05, 2013 at 02:13:48PM +0800, Amos Kong wrote:
> The MAX_VLAN is 4096, currently the vlan filter table has 512
> (4096 >> 3) entries, it's wrong.
> 
> One entry in vlan filter table can indicate 32(1 << 5) vlans,
> so the table should have 128 (4096 >> 5) entries.
> 
> Signed-off-by: Amos Kong <akong@redhat.com>
> ---
> btw, it would be simple to use an uint32 number to indicate
> all vlans (same as igbvf). I found e1000 uses the same table.
> Nothing needs to change here, it would cause migration issue.
> ---
>  hw/net/virtio-net.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)

I don't understand this patch.  memset() and qemu_put_buffer() work in
bytes, therefore MAX_VLAN >> 3.

MAX_VLAN >> 3 == MAX_VLAN / BITS_PER_BYTE
MAX_VLAN >> 5 == MAX_VLAN / (sizeof(uint32_t) * BITS_PER_BYTE)

What bug are you trying to fix?