Hi all! We've faced crash bug, which is reproducing on master branch as well. The case is described in 01, where fix is suggested. New iotest in 02 crashes without that fix.
v3: resend for convenience, as all preparatory patches are merged. 01-02: add Eric's r-b and t-b marks ==== This is a crash-fix, so it would be good to fix in 5.1. Still neither Eric nor I are sure in patch 01: is AIO_WAIT_WHILE used correctly? ==== Side note: this AIO_WAIT_WHILE may be long, if nbd reconnect is enabled and connection failed recently. Still it's another story: I think we actually should disable reconnect in bdrv_close, before drain. Vladimir Sementsov-Ogievskiy (2): nbd: make nbd_export_close_all() synchronous iotests: test shutdown when bitmap is exported through NBD nbd/server.c | 8 +++++ tests/qemu-iotests/299 | 65 ++++++++++++++++++++++++++++++++++++++ tests/qemu-iotests/299.out | 10 ++++++ tests/qemu-iotests/group | 1 + 4 files changed, 84 insertions(+) create mode 100644 tests/qemu-iotests/299 create mode 100644 tests/qemu-iotests/299.out -- 2.21.0
