| Message ID | 1435730034-18284-1-git-send-email-_govind@gmx.com |
|---|---|
| State | Changes Requested, archived |
| Delegated to: | David Miller |
| Headers | show |
From: Govindarajulu Varadarajan <_govind@gmx.com> Date: Wed, 1 Jul 2015 11:23:54 +0530 > Current code checks only rx work done to complete napi. It completely ignores > tx work done. If we have only tx packets to clean and no rq work, we always > napi complete instead of re-poll. Change this behavior to re-poll until > tx work_done + rx work_done is not 0. The existing TX behavior is correct, please do not change it. You should never count TX work against the NAPI poll budget, it is only for RX work. -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/net/ethernet/cisco/enic/enic_main.c b/drivers/net/ethernet/cisco/enic/enic_main.c index da2004e..6d7ce79 100644 --- a/drivers/net/ethernet/cisco/enic/enic_main.c +++ b/drivers/net/ethernet/cisco/enic/enic_main.c @@ -1170,7 +1170,7 @@ static int enic_poll(struct napi_struct *napi, int budget) wq_work_done, 0 /* dont unmask intr */, 0 /* dont reset intr timer */); - return rq_work_done; + return budget; } if (budget > 0) @@ -1191,26 +1191,27 @@ static int enic_poll(struct napi_struct *napi, int budget) 0 /* don't reset intr timer */); err = vnic_rq_fill(&enic->rq[0], enic_rq_alloc_buf); + enic_poll_unlock_napi(&enic->rq[cq_rq], napi); /* Buffer allocation failed. Stay in polling * mode so we can try to fill the ring again. */ if (err) - rq_work_done = rq_work_to_do; + return budget; - if (rq_work_done < rq_work_to_do) { + if (!work_done) { - /* Some work done, but not enough to stay in polling, - * exit polling + /* No work to do, exit polling */ napi_complete(napi); vnic_intr_unmask(&enic->intr[intr]); + + return work_done; } - enic_poll_unlock_napi(&enic->rq[cq_rq], napi); - return rq_work_done; + return budget; } static void enic_set_int_moderation(struct enic *enic, struct vnic_rq *rq)
In enic_poll, we clean tx and rx queues, when low latency busy socket polling is happening, enic_poll will only clean tx queue. After cleaning tx, it should return total budget for re-poll. Current code checks only rx work done to complete napi. It completely ignores tx work done. If we have only tx packets to clean and no rq work, we always napi complete instead of re-poll. Change this behavior to re-poll until tx work_done + rx work_done is not 0. There is a small window between vnic_intr_unmask() and enic_poll_unlock_napi(). In this window if an irq occurs and napi is scheduled on different cpu, it tries to acquire enic_poll_lock_napi() and fails. Unlock napi_poll before unmasking the interrupt. Signed-off-by: Govindarajulu Varadarajan <_govind@gmx.com> --- drivers/net/ethernet/cisco/enic/enic_main.c | 15 ++++++++------- 1 file changed, 8 insertions(+), 7 deletions(-)