| Message ID | 1540624524-118822-1-git-send-email-oulijun@huawei.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | hns misc updates for 4.20 | expand |
On Sat, Oct 27, 2018 at 03:15:14PM +0800, Lijun Ou wrote: > This patch series add two new features, include SRQ > and DCQCN. > From first to the fourth patch implement the SRQ(Share > Receive Queue) verbs and update the poll cq verbs to > deal with SRQ complentions as well as add the process > flow of asynchronous event. 4.20 has been done for a few weeks now, this will be considered for 4.21 (or whatever we call it) Jason
在 2018/10/29 4:04, Jason Gunthorpe 写道: > On Sat, Oct 27, 2018 at 03:15:14PM +0800, Lijun Ou wrote: >> This patch series add two new features, include SRQ >> and DCQCN. >> From first to the fourth patch implement the SRQ(Share >> Receive Queue) verbs and update the poll cq verbs to >> deal with SRQ complentions as well as add the process >> flow of asynchronous event. > > 4.20 has been done for a few weeks now, this will be considered for > 4.21 (or whatever we call it) > > Jason > > Hi, Jason Do I need to resend this patchset? thanks Lijun Ou
On Wed, Nov 07, 2018 at 09:44:04AM +0800, oulijun wrote: > 在 2018/10/29 4:04, Jason Gunthorpe 写道: > > On Sat, Oct 27, 2018 at 03:15:14PM +0800, Lijun Ou wrote: > >> This patch series add two new features, include SRQ > >> and DCQCN. > >> From first to the fourth patch implement the SRQ(Share > >> Receive Queue) verbs and update the poll cq verbs to > >> deal with SRQ complentions as well as add the process > >> flow of asynchronous event. > > > > 4.20 has been done for a few weeks now, this will be considered for > > 4.21 (or whatever we call it) > > > > Jason > > > > > Hi, Jason > Do I need to resend this patchset? No, you should watch patchworks to answer questions like that. If the patches are in patchworks still in then they do not need resending https://patchwork.kernel.org/project/linux-rdma/list/ Jason
On Sat, Oct 27, 2018 at 03:15:14PM +0800, Lijun Ou wrote: > This patch series add two new features, include SRQ > and DCQCN. > >From first to the fourth patch implement the SRQ(Share > Receive Queue) verbs and update the poll cq verbs to > deal with SRQ complentions as well as add the process > flow of asynchronous event. > > The next three patches mainly add DCQCN(Data Center QCN) > and timer support for hip08. > > The remaining patches includes two bugfixes and a > Code optimization > > Lijun Ou (7): > RDMA/hns: Eanble SRQ capacity for hip08 > RDMA/hns: Init SRQ table for hip08 > RDMA/hns: Add SRQ support for hip08 kernel mode > RDMA/hns: Add SRQ asynchronous event support > RDMA/hns: Init qp context when modify qp from reset to init > RDMA/hns: Use the macro instead of qp state transition support > RDMA/hns: Fix the bug while use multi-hop of pbl These 3 look unrelated, they should go in a followup series > Yangyang Li (3): > RDMA/hns: Add SCC context allocation support for hip08 > RDMA/hns: Add SCC context clr support for hip08 > RDMA/hns: Add timer allocation support for hip08 And SCC is related to DCQCN? Why is it being sent in the same series? One topic per series please. > drivers/infiniband/hw/hns/Makefile | 2 +- > drivers/infiniband/hw/hns/hns_roce_alloc.c | 2 + > drivers/infiniband/hw/hns/hns_roce_cmd.h | 16 + > drivers/infiniband/hw/hns/hns_roce_device.h | 130 +++++- > drivers/infiniband/hw/hns/hns_roce_hem.c | 109 ++++- > drivers/infiniband/hw/hns/hns_roce_hem.h | 5 + > drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 613 ++++++++++++++++++++++++++-- > drivers/infiniband/hw/hns/hns_roce_hw_v2.h | 178 +++++++- > drivers/infiniband/hw/hns/hns_roce_main.c | 137 +++++++ > drivers/infiniband/hw/hns/hns_roce_mr.c | 139 ++++++- > drivers/infiniband/hw/hns/hns_roce_qp.c | 50 ++- > drivers/infiniband/hw/hns/hns_roce_srq.c | 490 ++++++++++++++++++++++ > include/uapi/rdma/hns-abi.h | 6 + > 13 files changed, 1799 insertions(+), 78 deletions(-) > create mode 100644 drivers/infiniband/hw/hns/hns_roce_srq.c Other than the radix tree thing this code looked OK to me. If it had been split into 3 series some of it could have been applied already. Please try to write better commit messages. Empty phrases like 'this patch' and 'this patch mainly' are discouraged, we already know the commit message is talking about the patch. Explain better what improvement the patch brings to the kernel. Thanks, Jason
在 2018/11/23 2:17, Jason Gunthorpe 写道: > On Sat, Oct 27, 2018 at 03:15:14PM +0800, Lijun Ou wrote: >> This patch series add two new features, include SRQ >> and DCQCN. >> >From first to the fourth patch implement the SRQ(Share >> Receive Queue) verbs and update the poll cq verbs to >> deal with SRQ complentions as well as add the process >> flow of asynchronous event. >> >> The next three patches mainly add DCQCN(Data Center QCN) >> and timer support for hip08. >> >> The remaining patches includes two bugfixes and a >> Code optimization >> >> Lijun Ou (7): >> RDMA/hns: Eanble SRQ capacity for hip08 >> RDMA/hns: Init SRQ table for hip08 >> RDMA/hns: Add SRQ support for hip08 kernel mode >> RDMA/hns: Add SRQ asynchronous event support >> RDMA/hns: Init qp context when modify qp from reset to init >> RDMA/hns: Use the macro instead of qp state transition support >> RDMA/hns: Fix the bug while use multi-hop of pbl > These 3 look unrelated, they should go in a followup series Yes, it is some optization and bug fix. I will split it from the patchset. >> Yangyang Li (3): >> RDMA/hns: Add SCC context allocation support for hip08 >> RDMA/hns: Add SCC context clr support for hip08 >> RDMA/hns: Add timer allocation support for hip08 > And SCC is related to DCQCN? Why is it being sent in the same series? Yes. > One topic per series please. > >> drivers/infiniband/hw/hns/Makefile | 2 +- >> drivers/infiniband/hw/hns/hns_roce_alloc.c | 2 + >> drivers/infiniband/hw/hns/hns_roce_cmd.h | 16 + >> drivers/infiniband/hw/hns/hns_roce_device.h | 130 +++++- >> drivers/infiniband/hw/hns/hns_roce_hem.c | 109 ++++- >> drivers/infiniband/hw/hns/hns_roce_hem.h | 5 + >> drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 613 ++++++++++++++++++++++++++-- >> drivers/infiniband/hw/hns/hns_roce_hw_v2.h | 178 +++++++- >> drivers/infiniband/hw/hns/hns_roce_main.c | 137 +++++++ >> drivers/infiniband/hw/hns/hns_roce_mr.c | 139 ++++++- >> drivers/infiniband/hw/hns/hns_roce_qp.c | 50 ++- >> drivers/infiniband/hw/hns/hns_roce_srq.c | 490 ++++++++++++++++++++++ >> include/uapi/rdma/hns-abi.h | 6 + >> 13 files changed, 1799 insertions(+), 78 deletions(-) >> create mode 100644 drivers/infiniband/hw/hns/hns_roce_srq.c > Other than the radix tree thing this code looked OK to me. If it had > been split into 3 series some of it could have been applied already. > > Please try to write better commit messages. Empty phrases like 'this > patch' and 'this patch mainly' are discouraged, we already know the > commit message is talking about the patch. Explain better what > improvement the patch brings to the kernel. > > Thanks, > Jason Ok, thank your review. I will analysis the reviews for radix tree and send 3 series in next patchset > . >