This makes it easier for consumers to detect a path, or a device that
might be mounted or might have more than one mount, including tmpfs and
devtmpfs support, which was just not possible before.
Signed-off-by: Alfredo Deza <adeza@redhat.com>
Fixes the coverity issues:
** 1402628 Uninitialized scalar field
CID 1402628 (#1 of 1): Uninitialized scalar field (UNINIT_CTOR)
2. uninit_member: Non-static class member epoch is not initialized
in this constructor nor in any functions that it calls.
** 1409841 Uninitialized scalar field
CID 1409841 (#1 of 1): Uninitialized scalar field (UNINIT_CTOR)
2. uninit_member: Non-static class member len is not initialized
in this constructor nor in any functions that it calls.
** 1416594 Uninitialized scalar field
CID 1416594 (#1 of 1): Uninitialized scalar field (UNINIT_CTOR)
2. uninit_member: Non-static class member field fh.fh_hk is not
initialized in this constructor nor in any functions that it calls.
Signed-off-by: Amit Kumar <amitkuma@redhat.com>
some times, we only want scrub at out permitted time, avoid
scrub mainly deep scrub affecting business io. this patch
make no scrub deadline when osd_scrub_max_interval == 0, then
we can make sure scrub can only do in expect time.
Signed-off-by: kungf <yang.wang@easystack.cn>
os/bluestore: allow reconstruction of osd data dir from bluestore bdev label
Reviewed-By: Alfredo Deza <adeza@redhat.com>
Reviewed-by: Kefu Chai <kchai@redhat.com>
to include the change of https://github.com/ceph/dmclock/pull/41
Merge commit '060a7777cb6fe3f052259e1324490cdbf7b49980'
Signed-off-by: Kefu Chai <kchai@redhat.com>
1bbd42b8fe Merge pull request #41 from tchaikov/wip-client-info-ptr
cb7901e99a ClientRec: instead of a copy of ClientInfo, keep a pointer of it
git-subtree-dir: src/dmclock
git-subtree-split: 1bbd42b8fe212dbff68c2b470813e936859705d6
This guarantees replayed unsafe requests (on other mds) and mksnap/setlayout
get processed in proper order.
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Auth mds may take xlock on the lock and change the object when replaying
unsafe requests. To guarantee new requests and replayed unsafe requests
(on auth mds) get processed in proper order, we shouldn't rdlock locks in
replica object while auth mds of the object is recovering
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
This reverts commit c11485e1b3a58631628644152816d9b22a17d8bd.
The original sort order was compliant with amazon. Failures observed
in s3tests turned out to be a bug in boto2 - boto3 and other s3 clients
produce the correct sort order.
Fixes: http://tracker.ceph.com/issues/21832
Signed-off-by: Casey Bodley <cbodley@redhat.com>
a. upstream motivating behavior no longer reproduces
b. incorrect stat behavior observed, traced to exact match enforcement
Signed-off-by: Matt Benjamin <mbenjamin@redhat.com>
Introduce new rgw_readdir2(...), which in which continues
from an arbitrary dirent name, which presumably has been
seen in a prior partial enumeration.
Add single-file unit test for READDIR cases, librgw_file_marker.cc.
Signed-off-by: Matt Benjamin <mbenjamin@redhat.com>
Require full size notation for LVM (e.g. 50G or 100M) and do not
"translate" tags by pre-fixing them with 'ceph'. That is already how the
rest of the API works, no need to make that translation in one place
Signed-off-by: Alfredo Deza <adeza@redhat.com>
