Revert "net/smc: don't req_notify until all CQEs drained"
Subsystems (5)
Dust Li
maintainerPATCH 1/1 Revert "net/smc: don't req_notify until all CQEs drained"
This reverts commit a505cce6f7cfaf2aa2385aab7286063c96444526.
Leon says:
We already discussed that. SMC should be changed to use RDMA CQ pool API
drivers/infiniband/core/cq.c.
ib_poll_handler() has much better implementation (tracing, IRQ rescheduling, proper error handling) than this SMC variant.
Since we will switch to ib_poll_handler() in the future, revert this patch.
--- net/smc/smc_wr.c | 49 +++++++++++++++++++++--------------------------- 1 file changed, 21 insertions(+), 28 deletions(-)
net/smc/smc_wr.c +21 -28
@@ -137,28 +137,25 @@ static void smc_wr_tx_tasklet_fn(struct tasklet_struct *t) {struct smc_ib_device *dev =from_tasklet (dev, t, send_tasklet);struct ib_wc wc[SMC_WR_MAX_POLL_CQE ]; -int i, rc; +int i =0 , rc; +int polled =0 ; again: + polled++;do {memset (&wc,0 ,sizeof (wc)); rc =ib_poll_cq (dev->roce_cq_send,SMC_WR_MAX_POLL_CQE , wc); +if (polled ==1 ) { +ib_req_notify_cq (dev->roce_cq_send, +IB_CQ_NEXT_COMP | +IB_CQ_REPORT_MISSED_EVENTS ); + } +if (!rc) +break ;for (i =0 ; i < rc; i++)smc_wr_tx_process_cqe (&wc[i]); -if (rc <SMC_WR_MAX_POLL_CQE ) -/* If < SMC_WR_MAX_POLL_CQE, the CQ should have been -* drained, no need to poll again. --Guangguan Wang -*/ -break ; }while (rc >0 ); - -/* IB_CQ_REPORT_MISSED_EVENTS make sure if ib_req_notify_cq() returns -* 0, it is safe to wait for the next event. -* Else we must poll the CQ again to make sure we won't miss any event -*/ -if (ib_req_notify_cq (dev->roce_cq_send, -IB_CQ_NEXT_COMP | -IB_CQ_REPORT_MISSED_EVENTS )) +if (polled ==1 ) goto again; }
@@ -481,28 +478,24 @@ static void smc_wr_rx_tasklet_fn(struct tasklet_struct *t) {struct smc_ib_device *dev =from_tasklet (dev, t, recv_tasklet);struct ib_wc wc[SMC_WR_MAX_POLL_CQE ]; +int polled =0 ;int rc; again: + polled++;do {memset (&wc,0 ,sizeof (wc)); rc =ib_poll_cq (dev->roce_cq_recv,SMC_WR_MAX_POLL_CQE , wc); -if (rc >0 ) -smc_wr_rx_process_cqes (&wc[0 ], rc); -if (rc <SMC_WR_MAX_POLL_CQE ) -/* If < SMC_WR_MAX_POLL_CQE, the CQ should have been -* drained, no need to poll again. --Guangguan Wang -*/ +if (polled ==1 ) { +ib_req_notify_cq (dev->roce_cq_recv, +IB_CQ_SOLICITED_MASK + |IB_CQ_REPORT_MISSED_EVENTS ); + } +if (!rc)break ; +smc_wr_rx_process_cqes (&wc[0 ], rc); }while (rc >0 ); - -/* IB_CQ_REPORT_MISSED_EVENTS make sure if ib_req_notify_cq() returns -* 0, it is safe to wait for the next event. -* Else we must poll the CQ again to make sure we won't miss any event -*/ -if (ib_req_notify_cq (dev->roce_cq_recv, -IB_CQ_SOLICITED_MASK | -IB_CQ_REPORT_MISSED_EVENTS )) +if (polled ==1 ) goto again; }
Signature · 1 line
2.19.1.3.ge56e4f7
Quoted 12 lines · Dust Li
This reverts commit
a505cce6f7cfaf2aa2385aab7286063c96444526.Leon says:
We already discussed that. SMC should be changed to use RDMA CQ pool API
drivers/infiniband/core/cq.c.
ib_poll_handler()has much better implementation (tracing, IRQ rescheduling, proper error handling) than this SMC variant.Since we will switch to
ib_poll_handler()in the future, revert this patch.
Quoted 8 lines
Suggested-by Leon Romanovsky <[email protected]>Suggested-by Karsten Graul <[email protected]>Signed-off-by Dust Li <[email protected]>--- net/smc/smc_wr.c | 49 +++++++++++++++++++++--------------------------- 1 file changed, 21 insertions(+), 28 deletions(-)
Thanks,
This patch was applied to netdev/net-next.git (master) by David S. Miller <[email protected]>:
Hello:
This patch was applied to netdev/net-next.git (master) by David S. Miller <[email protected]>:
Quoted 10 lines · On Fri, 4 Mar 2022 17:17:19 +0800 you wrote:
This reverts commit
a505cce6f7cfaf2aa2385aab7286063c96444526.Leon says:
We already discussed that. SMC should be changed to use RDMA CQ pool API
drivers/infiniband/core/cq.c.
ib_poll_handler()has much better implementation (tracing, IRQ rescheduling, proper error handling) than this SMC variant.[...]
Here is the summary with links:
- [net-next] Revert "net/smc: don't req_notify until all CQEs drained"
https://git.kernel.org/netdev/net-next/c/925a24213b5c
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html