nbd: Don't mishandle unaligned client requests
authorEric Blake <eblake@redhat.com>
Thu, 21 Apr 2016 14:42:30 +0000 (08:42 -0600)
committerPeter Maydell <peter.maydell@linaro.org>
Fri, 22 Apr 2016 10:55:35 +0000 (11:55 +0100)
commitdf7b97ff89319ccf392a16748081482a3d22b35a
tree39fa13b93ee515257d1115af97ab5678ed065fd8
parent8d0d9b9f67d6bdee9eaec1e8c1222ad91dc4ac01
nbd: Don't mishandle unaligned client requests

The NBD protocol does not (yet) force any alignment constraints
on clients.  Even though qemu NBD clients always send requests
that are aligned to 512 bytes, we must be prepared for non-qemu
clients that don't care about alignment (even if it means they
are less efficient).  Our use of blk_read() and blk_write() was
silently operating on the wrong file offsets when the client
made an unaligned request, corrupting the client's data (but
as the client already has control over the file we are serving,
I don't think it is a security hole, per se, just a data
corruption bug).

Note that in the case of NBD_CMD_READ, an unaligned length could
cause us to return up to 511 bytes of uninitialized trailing
garbage from blk_try_blockalign() - hopefully nothing sensitive
from the heap's prior usage is ever leaked in that manner.

Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Tested-by: Kevin Wolf <kwolf@redhat.com>
Message-id: 1461249750-31928-1-git-send-email-eblake@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
nbd/server.c