Until the kernel supports RBD exclusive locking, this test
has been updated to create shared images (exclusive locking
disabled).
Fixes: #10613
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
Reviewed-by: Josh Durgin <jdurgin@redhat.com>
This lets you put a unique port in .ceph_port in your working dir and
vstart.sh instances will avoid each other without having to pass
CEPH_PORT=... to each one each time.
Signed-off-by: Sage Weil <sage@redhat.com>
On FileStore::umount() when stopping and in ENOSPC condition,
don't let write_thread_entry() get stuck in commit_cond.Wait(write_lock).
Fixes: #10474
Signed-off-by: David Zafman <dzafman@redhat.com>
Originally if memstore run out of space, it will report a very
large positive number as free space. For example:
root@bigmem:~# rados df
pool name KB objects clones degraded unfound rd rd KB wr wr KB
rbd 12366704 92240 0 0 0 0 0 92240 12366704
total used 12375877 92240
total avail 36028797009199167
total space 2611076
st->f_bavail = st->f_blocks - used_bytes / st->f_bsize
This is due to used_bytes is an unsigned value, so compiler make the whole statement unsigned.
Fix it by adding explicit type cast,
st->f_bavail = long(st->f_blocks) - long(used_bytes / st->f_bsize)
Signed-off-by: Xiaoxi Chen <xiaoxi.chen@intel.com>
Until the kernel supports RBD exclusive locking, this test
has been updated to disable the new feature.
Fixes: #10614
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
Previously, auth would end up containing every object without a
self-evident defect -- even if they did not match each other. Instead
of filtering out the non-matching items there, be_compare_scrubmaps now
returns one valid object and be_compare_scrubmaps gathers the other
which match it.
We can be smarter by doing this in be_select_auth_object and selecting
the largest matching set, but for now this is simpler.
Fixes: 10524
Signed-off-by: Samuel Just <sjust@redhat.com>
The buffer provided to aio_write shouldn't be invalidated until
after aio_write has indicated that the operation has completed.
Fixes: #10590
Backport: giant
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
Reviewed-by: Josh Durgin <jdurgin@redhat.com>
This is because running fs/misc on a kclient
should not include quota (quota not supported
on kernel client).
Fixes: #10579
Signed-off-by: John Spray <john.spray@redhat.com>
Fix the typo in doc/rbd/rados-rbd-cmds.rst, also add explanation/example
when creating an image without specifying pool.
Signed-off-by: Frank Yu <flyxiaoyu@gmail.com>
Automatic locking hides the ESHUTDOWN from the caller, which is how
this test detects that blacklisting works.
Fixes: #10592
Signed-off-by: Josh Durgin <jdurgin@redhat.com>
Ceph fork when running as daemon. This fix call xio option to enforce
Accelio to explicitly call ibv_fork_init() to avoid cq overrun issue.
Signed-off-by: Vu Pham <vu@mellanox.com>
d427ca3540 reversed the order of localt
and op_t, while some logic depends on the order of localt and op_t.
This commit revert d427ca3540.
Fix http://tracker.ceph.com/issues/10534.
Change-Id: I5adba561d10274e3d12f89418a755c39e5e5f086
Signed-off-by: Dong Yuan <yuandong1222@gmail.com>
the other.op_bl SHOULD NOT be changes during append operation,
we use additional bufferlist to avoid this problem
Change-Id: I5d4a8c9ed3d0b926ca4bbdcdadf72cb4d5fde53c
Signed-off-by: Dong Yuan <yuandong1222@gmail.com>
