In RDMADispatcher::handle_async_event, when case
IBV_EVENT_QP_LAST_WQE_REACHED matched, enqueue_dead_qp wants the lock,
result in deadlock here. Change enqueue_dead_qp to lockless.
Fixes: https://tracker.ceph.com/issues/44298
Signed-off-by: Chunsong Feng <fengchunsong@huawei.com>
Signed-off-by: luorixin <luorixin@huawei.com>
Generally, the conflict resolution performed manually is always to do a
union: automate this. This avoids pointless rebases of PRs.
Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
* refs/pull/36641/head:
mds : skip show_cache if level of log is lower than 7
Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
Reviewed-by: Zheng Yan <zyan@redhat.com>
MMonMetadata is sent to pre-luminous monitors and handled only for
backward compatibility with pre-luminous monitors. and luminous is
the last release we were still sending this message. but there is no
need for pacific to be compatible with pre-luminous monitors because of
the +2 policy.
so let's drop this message from the source tree.
Signed-off-by: Kefu Chai <kchai@redhat.com>
test/multi_stress_watch: use condition_variable for level triggering
Reviewed-By: Josh Durgin <jdurgin@redhat.com>
Reviewed-by: Neha Ojha <nojha@redhat.com>
Reviewed-by: Samuel Just <sjust@redhat.com>
* use the latest stable sphinx
* unpin breathe, sphinx-autodoc-typehints, so pip can choose the
compatible versions without worrying about the compile failures.
Signed-off-by: Kefu Chai <kchai@redhat.com>
otherwise we could have following errors:
Invalid C declaration: Expected identifier in nested name, got keyword: int [error at 18]
CEPH_RADOS_API int rados_aio_append (rados_ioctx_t io, const char *oid, rados_completion_t completion, const char *buf, size_t len)
------------------^
Signed-off-by: Kefu Chai <kchai@redhat.com>
Make parsing of keyword arguments a bit safer by
makeing sure the key is actually a valid parameter name
Fixes: https://tracker.ceph.com/issues/47185
Signed-off-by: Sebastian Wagner <sebastian.wagner@suse.com>
Othwerise we end up with being unable to create
new containers, due to the name being already used
Fixes: https://tracker.ceph.com/issues/47170
Signed-off-by: Sebastian Wagner <sebastian.wagner@suse.com>
MMonMetadata is sent to pre-luminous monitors and handled only for
backward compatibility with pre-luminous monitors. and luminous is the
last release we were still sending this message. but there is no need
for pacific to be compatible with pre-luminous monitors because of the
+2 policy.
so let's stop handle this message in pacific and up.
Signed-off-by: Kefu Chai <kchai@redhat.com>
Using Type=Forking and setting the PIDFile=
directive is the supported configuration for
running podman containers with systemd as it
more accuractely models what is actually
happening. Type=Simple assumes podman is the
direct parent of the container processes which
is not true.
Fixes: https://tracker.ceph.com/issues/46654
Signed-off-by: Adam King <adking@redhat.com>
The get_active_daemon functions for monitoring stack daemons
were just returning the first or last daemon in the given list
without checking the list actually contained any daemons
Fixes: https://tracker.ceph.com/issues/47171
Signed-off-by: Adam King <adking@redhat.com>
If we don't emit the select event, the action button can become stale and not
update when the data changes.
This fixes a regression introduced in 31e9bbb9b308277d2f89390d5ff62f13457d69cb
Fixes: https://tracker.ceph.com/issues/47175
Signed-off-by: Tiago Melo <tmelo@suse.com>
This is particularly useful for debugging purposes when clean_temps()
takes abnormally high amount of time due to flaws in crush rules etc.
Fixes: https://tracker.ceph.com/issues/47159
Signed-off-by: Neha Ojha <nojha@redhat.com>
