hw/xen: clean up xen_block_find_free_vdev() to avoid Coverity false positive

Coverity couldn't see that nr_existing was always going to be zero when
qemu_xen_xs_directory() returned NULL in the ENOENT case (CID 1523906).

Perhaps more to the point, neither could Peter at first glance. Improve
the code to hopefully make it clearer to Coverity and human reviewers
alike.

Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Paul Durrant <paul@xen.org>
This commit is contained in:
David Woodhouse 2023-11-09 15:21:13 +00:00
parent 87bfffdf75
commit 6f7997e004

View File

@ -91,9 +91,27 @@ static bool xen_block_find_free_vdev(XenBlockDevice *blockdev, Error **errp)
existing_frontends = qemu_xen_xs_directory(xenbus->xsh, XBT_NULL, fe_path,
&nr_existing);
if (!existing_frontends && errno != ENOENT) {
error_setg_errno(errp, errno, "cannot read %s", fe_path);
return false;
if (!existing_frontends) {
if (errno == ENOENT) {
/*
* If the frontend directory doesn't exist because there are
* no existing vbd devices, that's fine. Just ensure that we
* don't dereference the NULL existing_frontends pointer, by
* checking that nr_existing is zero so the loop below is not
* entered.
*
* In fact this is redundant since nr_existing is initialized
* to zero, but setting it again here makes it abundantly clear
* to Coverity, and to the human reader who doesn't know the
* semantics of qemu_xen_xs_directory() off the top of their
* head.
*/
nr_existing = 0;
} else {
/* All other errors accessing the frontend directory are fatal. */
error_setg_errno(errp, errno, "cannot read %s", fe_path);
return false;
}
}
memset(used_devs, 0, sizeof(used_devs));