Changcheng Liu [Mon, 1 Jul 2019 02:27:45 +0000 (10:27 +0800)]
msg/async/rdma: use different strategy to reset read/write chunk
When releasing read chunk to pool, the chunk::offset & chunk::bound
should be reset to zero. For write chunk, it's better to reset
chunk::offset to zero and chunk::bound to chunk length which means that
[offset, bound) is writable.
Signed-off-by: Changcheng Liu <changcheng.liu@aliyun.com>
Changcheng Liu [Thu, 27 Jun 2019 05:19:58 +0000 (13:19 +0800)]
msg/async/rdma: cosmetics initialize ibv_send_wr* var
API usage:
int ibv_post_send(struct ibv_qp *qp, struct ibv_send_wr *wr, struct ibv_send_wr **bad_wr)
Input Parameters:
qp struct ibv_qp from ibv_create_qp
wr first work request (WR)
Output Parameters:
bad_wr pointer to first rejected WR
Return Value:
0 on success, -1 on error.
If the call fails, errno will be set to indicate the reason for the failure.
To avoid wrong checking return value, it's better to initialize the
value to be nullptr.
Signed-off-by: Changcheng Liu <changcheng.liu@aliyun.com>
1. There's no need to get stack & dispatcher from RDMAStack again
since RDMAWorker has stored the value.
2. cache the Infiniband object to be used in local scope.
Signed-off-by: Changcheng Liu <changcheng.liu@aliyun.com>
Changcheng Liu [Thu, 27 Jun 2019 03:03:44 +0000 (11:03 +0800)]
msg/async/rdma: refine Chunk construction function
1. all values are initialized in construction function
In this way, it's easy to construct Chunk object in
PoolAllocator::malloc function.
2. For read chunk, member bound is initialized to be 0.
3. For send chunk, member bound is initialzied to be full space size.
Signed-off-by: Changcheng Liu <changcheng.liu@aliyun.com>
After reading one chunk, the chunk could be pushed into buffer list if its
effecitve content size is not zero. In this case, it also means that the
caller has got the required read length. Then all the continuous chunk will
be pushed into buffer list since the effective content size is not zero.
Signed-off-by: Changcheng Liu <changcheng.liu@aliyun.com>
Keep same logic:
1. If parameter block_size is zero, then allocate all the free chunks
to parameter std::vector<Chunk*> &chunks. i.e.
chunk_buffer_number = free_chunks.size()
2. If paramter block_size is not zero, then allocate the requested or
all the free chunks to paramter std::vector<Chunk*> &chunks.
Signed-off-by: Changcheng Liu <changcheng.liu@aliyun.com>
Changcheng Liu [Thu, 13 Jun 2019 11:04:40 +0000 (19:04 +0800)]
msg/async/rdma: use Chunk::get_size to get chunk size
remove Chunk::over interface and add Chunk::get_size interface
1) It's not clear when reading "over" function name.
2) Some places need know the current chunk block effective content size.
3) "Chunk::over()" could be replaced by "Chunk::get_size() == 0"
Signed-off-by: Changcheng Liu <changcheng.liu@aliyun.com>
Changcheng Liu [Thu, 13 Jun 2019 10:34:44 +0000 (18:34 +0800)]
msg/async/rdma: seperate Device construction if rdma_cm is used
If ms_async_rdma_cm is false, there's no need to call the api
rdma_get_device. If rdma_get_device is called, the devices remain
opened while librdmacm is loaded. This is not what we want when
ms_async_rdma_cm is false.
Signed-off-by: Changcheng Liu <changcheng.liu@aliyun.com>
Changcheng Liu [Fri, 28 Jun 2019 06:26:41 +0000 (14:26 +0800)]
msg/async/rdma: fix error argument to get right qp state
1. It's wrong to use "-1" as argument to query queue state.
In rdma library, ibv_query_qp will call ibv_cmd_query_qp to query
queue state. If "-1" is used as attr_mask, ibv_cmd_query_qp will
return error EOPNOTSUPP which means query failed.
2. In class QueuePair, is_error() could use member function get_state()
to get the queue pair state.
3. It's better to use qp_state as queue pair state according to
ibv_query_qp manual guide.
struct ibv_qp_attr {
enum ibv_qp_state qp_state; /* Current QP state */
enum ibv_qp_state cur_qp_state; /* Current QP state - irrelevant for ibv_query_qp */
...
Signed-off-by: Changcheng Liu <changcheng.liu@aliyun.com>
Changcheng Liu [Mon, 3 Jun 2019 05:31:09 +0000 (13:31 +0800)]
msg/async/rdma: export RDMAV_HUGEPAGES_SAFE before ibv_fork_init
In rdma-core library, ibv_fork_init will check environment variable
RDMAV_HUGEPAGES_SAFE to decide whether huge page is usable in system.
It doesn't make sense to export RDMAV_HUGEPAGES_SAFE env after
calling ibv_fork_init.
Signed-off-by: Changcheng Liu <changcheng.liu@aliyun.com>
Changcheng Liu [Mon, 3 Jun 2019 05:00:22 +0000 (13:00 +0800)]
msg/async/rdma: use ibv_port_attr object type in Port class
1. Avoid to do memory management without using pointer to operate
operate the allocated space. Or, it could have memory leak.
2. Since member type has been changed in class Device, it need
to use member domain operator "." to access to the sub-member in
object.
3. There's no need to consider experimental API of ibv_query_port.
So, merge ibv_query_port in the prolog.
Signed-off-by: Changcheng Liu <changcheng.liu@aliyun.com>
Patrick Donnelly [Wed, 21 Aug 2019 17:57:15 +0000 (10:57 -0700)]
Merge PR #28378 into master
* refs/pull/28378/head:
qa/tasks: introduce Thrasher base class
qa/tasks: Fix typo
qa/tasks: manage thrashers
qa/tasks: start DaemonWatchdog when ceph starts
qa/tasks: make watch and bark handle more daemons
qa/tasks: move DaemonWatchdog to new file
Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
Jos Collin [Mon, 5 Aug 2019 10:52:10 +0000 (16:22 +0530)]
qa/tasks: introduce Thrasher base class
* Introduced a Thrasher base class.
* Updated thrashers to inherit from Thrasher.
* Replaced the magic variable e with Thrasher.exception as per the discussion.
Now the exception variable sets by default as the thrashers are inheriting
from the Thrasher class.
Fixes: https://github.com/ceph/ceph/pull/28378#discussion_r309337928 Fixes: https://tracker.ceph.com/issues/41133 Signed-off-by: Jos Collin <jcollin@redhat.com>
Nathan Cutler [Tue, 20 Aug 2019 14:23:06 +0000 (16:23 +0200)]
Merge pull request #29743 from smithfarm/wip-ceph-backport-https
script/ceph-backport.sh: carry https through to logical conclusion
Reviewed-by: Patrick Donnelly <pdonnell@redhat.com> Reviewed-by: Abhishek Lekshmanan <abhishek@suse.com> Reviewed-by: Willem Jan Withagen <wjw@digiware.nl>
Nathan Cutler [Mon, 19 Aug 2019 14:57:07 +0000 (16:57 +0200)]
scripts/ceph-backport.sh: always use https://tracker.ceph.com
Completing the wave of fixes to this script in the wake of
https://tracker.ceph.com/issues/38764, this commit replaces
"http" with "https" in the comments and puts the Redmine endpoint
into a variable, along with some other cleanups.
Kefu Chai [Tue, 20 Aug 2019 08:15:17 +0000 (16:15 +0800)]
osd: always initialize local variable
to silence a GCC warning like:
../src/osd/OSD.cc:4608:24: warning: ‘sender_delta_ub’ may be used
uninitialized in this function [-Wmaybe-uninitialized]
4608 | ceph::signedspan sender_delta_ub;
| ^~~~~~~~~~~~~~~
Kefu Chai [Mon, 19 Aug 2019 07:21:06 +0000 (15:21 +0800)]
cmake,run-make-check.sh,deb,rpm: disable SPDK by default
but we still enable it in `run-make-check.sh`
* cmake: disable SPDK by default
* run-make-check.sh: enable WITH_SPDK so at least we can ensure it
builds
* deb,rpm: add uuid-dev / libuuid-devel as a "make check" dependency
xie xingguo [Mon, 17 Jun 2019 03:05:31 +0000 (11:05 +0800)]
osd: do not invalidate clear_regions of missing item at boot
Seems we'll always mark clear_regions as all dirty
when reading pg logs and missing items off the disk,
which as a result turns incremental recovery off by default.
Also using std::move seems to be a bit more efficient
and robust here.
xie xingguo [Wed, 14 Aug 2019 06:12:43 +0000 (14:12 +0800)]
osd/PrimaryLogPG: fix dirty range of write_full
A write_full operation may implicitly truncate the object down,
hence we need to mark the truncated part as dirty as well since
follow-up randomized writes may still be able to (re)extend the
object size and leave some holes against the truncated part,
which as a result might cause problems during incremental-mode
recovery.
Note that write_update_size_and_usage would reset oi.size
and that's why we move the mark_data_region_dirty call before that.
projects / ceph.git / log