On 3/17/20 3:12 AM, Richard W.M. Jones wrote:
On Mon, Mar 16, 2020 at 10:36:15PM -0500, Eric Blake wrote:
Although nbdkit documents that any positive value should be treated as success to the .can_* callbacks, we had a window of releases where anything other than 1 could trigger an assertion failure, fixed in the previous patch. Update what we return to avoid tripping the bug in broken nbdkit.
Our return value has been latent since 0e8e8eb11d (v1.1.17), but the assertion failures did not occur until c306fa93ab and neighboring commits (v1.15.1). As v1.13.6 was the point where we started preferring builds against libnbd, where we always returned 1 on success, the problem was not detected until now; but it is still in the wild for any user mixing old plugins with new libnbd.
Signed-off-by: Eric Blake <eblake@redhat.com> --- plugins/nbd/nbd-standalone.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-)
While I don't like the idea of modifying plugins to fix broken nbdkit behaviour, you make a good case for this being safer, so:
ACK
Think of it as working around broken nbdkit behavior, rather than fixing it (as the fix is in patch 1/4). How many stable branches should we backport this to? Going all the way back to stable-1.2 seems like it might be overkill; maybe back to stable-1.14 (where nbd-standalone was first split off) or stable-1.12 (the last release that did not use libnbd) is sufficient? -- Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3226 Virtualization: qemu.org | libvirt.org