Hi all, Paolo found that commit 27ccdd52598290f introduced a regression in the throttling code.
It can be easily reproduced in scenarios where you have a throttling group with several drives but you only write to one of them. In that case the round-robin algorithm can select the wrong drive all the time and the actual requests are never completed. QEMU 2.7 is affected, here's the patch to fix it, plus a test case. Thanks, Berto Alberto Garcia (2): throttle: Correct access to wrong BlockBackendPublic structures qemu-iotests: Test I/O in a single drive from a throttling group block/throttle-groups.c | 27 +++++++++++++++++++++++---- tests/qemu-iotests/093 | 33 ++++++++++++++++++++++++++++----- tests/qemu-iotests/093.out | 4 ++-- 3 files changed, 53 insertions(+), 11 deletions(-) -- 2.9.3
