d05ab380db
Draining recursively traverses the graph, therefore we need to make sure that also such accesses to the graph are protected by the graph rdlock. There are 3 different drain callers to consider: 1. drain in the main loop: no issue at all, rdlock is nop. 2. drain in an iothread: rdlock only works in main loop or coroutines, so disallow it. 3. drain in a coroutine (regardless of AioContext): the drain mechanism takes care of scheduling a BH in the bs->aio_context that will then take care of perform the actual draining. This is wrong, because as pointed in (2) if bs->aio_context is an iothread then rdlock won't work. Therefore change bdrv_co_yield_to_drain to schedule the BH in the main loop. Caller (2) also implies that we need to modify test-bdrv-drain.c to disallow draining in the iothreads. For some places, we know that they will hold the lock, but we don't have the GRAPH_RDLOCK annotations yet. In this case, add assume_graph_lock() with a FIXME comment. These places will be removed once everything is properly annotated. Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> Message-ID: <20230929145157.45443-6-kwolf@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> |
||
|---|---|---|
| .. | ||
| avocado | ||
| bench | ||
| data | ||
| decode | ||
| docker | ||
| fp | ||
| guest-debug | ||
| image-fuzzer | ||
| include | ||
| keys | ||
| lcitool | ||
| migration | ||
| multiboot | ||
| perf/block/qcow2 | ||
| plugin | ||
| qapi-schema | ||
| qemu-iotests | ||
| qtest | ||
| rocker | ||
| tcg | ||
| tsan | ||
| uefi-test-tools | ||
| unit | ||
| vm | ||
| vmstate-static-checker-data | ||
| dbus-vmstate-daemon.sh | ||
| Makefile.include | ||
| meson.build | ||
| test-qht-par.c | ||
| vhost-user-bridge.c | ||