| Message ID | 20190614162435.44620.72298.stgit@awfm-01.aw.intel.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | Clean up and refactor some CQ code | expand |
On Fri, Jun 14, 2019 at 12:25:28PM -0400, Dennis Dalessandro wrote: > This is really a resubmit of some code clean up we floated a while back. The > main goal here is to clean up some of the stuff which should be in the uapi > directory vs in in the driver directory. Then to break the single lock for > recv wqe processing. > > The accompanying user bits should already be in a PR on GitHub. > > Change log documented below each commit message. The conflicts with other hfi patches are too big for me, please respin this on wip/for-testing and resend. Thanks, Jason
On 6/25/2019 1:31 PM, Jason Gunthorpe wrote: > On Fri, Jun 14, 2019 at 12:25:28PM -0400, Dennis Dalessandro wrote: >> This is really a resubmit of some code clean up we floated a while back. The >> main goal here is to clean up some of the stuff which should be in the uapi >> directory vs in in the driver directory. Then to break the single lock for >> recv wqe processing. >> >> The accompanying user bits should already be in a PR on GitHub. >> >> Change log documented below each commit message. > > The conflicts with other hfi patches are too big for me, please respin > this on wip/for-testing and resend. > > Thanks, > Jason > Jason, can you pull in f56044d ("IB/rdmavt: Add new completion inline") and 4a9ceb7 ("IB/{rdmavt, qib, hfi1}: Convert to new completion API") from Doug's wip/dl-for-next branch? There are still conflicts in the first patch here which I will respin, but then the rest will apply, the other series will then also apply. I'll post those after I can run some smoke tests. -Denny
On Wed, Jun 26, 2019 at 12:00:54PM -0400, Dennis Dalessandro wrote: > On 6/25/2019 1:31 PM, Jason Gunthorpe wrote: > > On Fri, Jun 14, 2019 at 12:25:28PM -0400, Dennis Dalessandro wrote: > > > This is really a resubmit of some code clean up we floated a while back. The > > > main goal here is to clean up some of the stuff which should be in the uapi > > > directory vs in in the driver directory. Then to break the single lock for > > > recv wqe processing. > > > > > > The accompanying user bits should already be in a PR on GitHub. > > > > > > Change log documented below each commit message. > > > > The conflicts with other hfi patches are too big for me, please respin > > this on wip/for-testing and resend. > > > > Thanks, > > Jason > > > > Jason, can you pull in f56044d ("IB/rdmavt: Add new completion inline") and > 4a9ceb7 ("IB/{rdmavt, qib, hfi1}: Convert to new completion API") from > Doug's wip/dl-for-next branch? This was already done Jason
On 6/26/2019 12:08 PM, Jason Gunthorpe wrote: > On Wed, Jun 26, 2019 at 12:00:54PM -0400, Dennis Dalessandro wrote: >> On 6/25/2019 1:31 PM, Jason Gunthorpe wrote: >>> On Fri, Jun 14, 2019 at 12:25:28PM -0400, Dennis Dalessandro wrote: >>>> This is really a resubmit of some code clean up we floated a while back. The >>>> main goal here is to clean up some of the stuff which should be in the uapi >>>> directory vs in in the driver directory. Then to break the single lock for >>>> recv wqe processing. >>>> >>>> The accompanying user bits should already be in a PR on GitHub. >>>> >>>> Change log documented below each commit message. >>> >>> The conflicts with other hfi patches are too big for me, please respin >>> this on wip/for-testing and resend. >>> >>> Thanks, >>> Jason >>> >> >> Jason, can you pull in f56044d ("IB/rdmavt: Add new completion inline") and >> 4a9ceb7 ("IB/{rdmavt, qib, hfi1}: Convert to new completion API") from >> Doug's wip/dl-for-next branch? > > This was already done > > Jason Hmm... I'm not seeing it: Here they are in Doug's branch: $ git log --pretty=oneline -n 100 rdma/wip/dl-for-next include/rdma | grep -e "Add new completion\|Convert to new" 4a9ceb7dbadf9e1435644b1f49720ee87431ce26 IB/{rdmavt, qib, hfi1}: Convert to new completion API f56044d686c82bd31713fc0398d68e322813dc62 IB/rdmavt: Add new completion inline Here is the for-testing: $ git log --pretty=oneline -n 100 rdma/wip/for-testing include/rdma | grep -e "Add new completion\|Convert to new" awfm-01 $ echo $? 1 -Denny
On Wed, Jun 26, 2019 at 12:21:19PM -0400, Dennis Dalessandro wrote: > Hmm... I'm not seeing it: > > Here they are in Doug's branch: > $ git log --pretty=oneline -n 100 rdma/wip/dl-for-next include/rdma | grep > -e "Add new completion\|Convert to new" > 4a9ceb7dbadf9e1435644b1f49720ee87431ce26 IB/{rdmavt, qib, hfi1}: Convert to > new completion API > f56044d686c82bd31713fc0398d68e322813dc62 IB/rdmavt: Add new completion > inline > > Here is the for-testing: > $ git log --pretty=oneline -n 100 rdma/wip/for-testing include/rdma | grep > -e "Add new completion\|Convert to new" > awfm-01 $ echo $? > 1 Ah the push got botched yesterday, fetch it again Thanks, Jason
This is really a resubmit of some code clean up we floated a while back. The main goal here is to clean up some of the stuff which should be in the uapi directory vs in in the driver directory. Then to break the single lock for recv wqe processing. The accompanying user bits should already be in a PR on GitHub. Change log documented below each commit message. --- Kamenee Arumugam (3): IB/hfi1: Move rvt_cq_wc struct into uapi directory IB/hfi1: Move receive work queue struct into uapi directory IB/rdmavt: Fracture single lock used for posting and processing RWQEs drivers/infiniband/hw/hfi1/qp.c | 4 - drivers/infiniband/sw/rdmavt/cq.c | 191 ++++++++++++++++++----------- drivers/infiniband/sw/rdmavt/qp.c | 239 +++++++++++++++++++++++++----------- drivers/infiniband/sw/rdmavt/qp.h | 2 drivers/infiniband/sw/rdmavt/rc.c | 41 ++++-- drivers/infiniband/sw/rdmavt/srq.c | 69 ++++++---- include/rdma/rdmavt_cq.h | 22 ++- include/rdma/rdmavt_qp.h | 91 +++++++++++--- include/uapi/rdma/rvt-abi.h | 61 +++++++++ 9 files changed, 502 insertions(+), 218 deletions(-) create mode 100644 include/uapi/rdma/rvt-abi.h -- -Denny