On Sat, 2012-08-18 at 23:04 +0300, Michael S. Tsirkin wrote:

Hi Nicholas,
I just noticed this problem in the interface:

+#include <linux/vhost.h>
+
+/*
+ * Used by QEMU userspace to ensure a consistent vhost-scsi ABI.
+ *
+ * ABI Rev 0: July 2012 version starting point for v3.6-rc merge
candidate +
+ *            RFC-v2 vhost-scsi userspace.  Add GET_ABI_VERSION ioctl
usage
+ */
+
+#define VHOST_SCSI_ABI_VERSION 0
+
+struct vhost_scsi_target {
+       int abi_version;
+       unsigned char vhost_wwpn[TRANSPORT_IQN_LEN];
+       unsigned short vhost_tpgt;
+};
+

Here TRANSPORT_IQN_LEN is 224, which is a multiple of 4.
Since vhost_tpgt is 2 bytes and abi_version is 4, the total size would
be 230.  But gcc needs struct size be aligned to first field size, which
is 4 bytes, so it pads the structure by extra 2 bytes to the total of
232.

This padding is very undesirable in an ABI:
- it can not be initialized easily
- it can not be checked easily
- it can leak information between kernel and userspace

Hmmmm, yes.  Very good reasons to avoid ABI ambiguity  ..

Simplest solution is probably just to make the padding
explicit:

+struct vhost_scsi_target {
+       int abi_version;
+       unsigned char vhost_wwpn[TRANSPORT_IQN_LEN];
+       unsigned short vhost_tpgt;
+       unsigned short reserved;
+};
+

I think we should fix this buglet before it goes out to users.

<nod>, fixing this up in target-pending/master now w/ your reported-by
+signoff, and will change vhost-scsi's copy of these defs for next
week's RFC-v3 posting.

Thanks MST!

--nab