virtio-blk: undo destructive iov_discard_*() operations
authorStefan Hajnoczi <stefanha@redhat.com>
Thu, 17 Sep 2020 09:44:54 +0000 (10:44 +0100)
committerStefan Hajnoczi <stefanha@redhat.com>
Wed, 23 Sep 2020 12:41:58 +0000 (13:41 +0100)
commit7bd04a041addcdef6a03e6498aafaea55ca6e88b
treebe7bbcdcab48a8a9b94a1af4be03ca8b33265c9d
parent9dd6f7c28eaa7d001a526d51c74bbf89d5402b8c
virtio-blk: undo destructive iov_discard_*() operations

Fuzzing discovered that virtqueue_unmap_sg() is being called on modified
req->in/out_sg iovecs. This means dma_memory_map() and
dma_memory_unmap() calls do not have matching memory addresses.

Fuzzing discovered that non-RAM addresses trigger a bug:

  void address_space_unmap(AddressSpace *as, void *buffer, hwaddr len,
                           bool is_write, hwaddr access_len)
  {
      if (buffer != bounce.buffer) {
          ^^^^^^^^^^^^^^^^^^^^^^^

A modified iov->iov_base is no longer recognized as a bounce buffer and
the wrong branch is taken.

There are more potential bugs: dirty memory is not tracked correctly and
MemoryRegion refcounts can be leaked.

Use the new iov_discard_undo() API to restore elem->in/out_sg before
virtqueue_push() is called.

Fixes: 827805a2492c1bbf1c0712ed18ee069b4ebf3dd6 ("virtio-blk: Convert VirtIOBlockReq.out to structrue")
Reported-by: Alexander Bulekov <alxndr@bu.edu>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Li Qiang <liq3ea@gmail.com>
Buglink: https://bugs.launchpad.net/qemu/+bug/1890360
Message-Id: <20200917094455.822379-3-stefanha@redhat.com>
hw/block/virtio-blk.c
include/hw/virtio/virtio-blk.h