From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752944AbbB1SX4 (ORCPT ); Sat, 28 Feb 2015 13:23:56 -0500 Received: from p3plsmtps2ded01.prod.phx3.secureserver.net ([208.109.80.58]:39262 "EHLO p3plsmtps2ded01.prod.phx3.secureserver.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752552AbbB1SXt (ORCPT ); Sat, 28 Feb 2015 13:23:49 -0500 x-originating-ip: 72.167.245.219 From: "K. Y. Srinivasan" To: gregkh@linuxfoundation.org, linux-kernel@vger.kernel.org, devel@linuxdriverproject.org, olaf@aepfle.de, apw@canonical.com, vkuznets@redhat.com Cc: "K. Y. Srinivasan" Subject: [PATCH 5/7] Drivers: hv: vmbus: Use a round-robin algorithm for picking the outgoing channel Date: Sat, 28 Feb 2015 11:39:02 -0800 Message-Id: <1425152344-4953-5-git-send-email-kys@microsoft.com> X-Mailer: git-send-email 1.7.4.1 In-Reply-To: <1425152344-4953-1-git-send-email-kys@microsoft.com> References: <1425152321-4914-1-git-send-email-kys@microsoft.com> <1425152344-4953-1-git-send-email-kys@microsoft.com> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org The current algorithm for picking an outgoing channel was not distributing the load well. Implement a simple round-robin scheme to ensure good distribution of the outgoing traffic. Signed-off-by: K. Y. Srinivasan Reviewed-by: Long Li --- drivers/hv/channel_mgmt.c | 30 +++++++++++++++--------------- include/linux/hyperv.h | 3 +++ 2 files changed, 18 insertions(+), 15 deletions(-) diff --git a/drivers/hv/channel_mgmt.c b/drivers/hv/channel_mgmt.c index b1e5a5f..6117891 100644 --- a/drivers/hv/channel_mgmt.c +++ b/drivers/hv/channel_mgmt.c @@ -350,6 +350,7 @@ static void vmbus_process_offer(struct vmbus_channel *newchannel) } newchannel->state = CHANNEL_OPEN_STATE; + channel->num_sc++; if (channel->sc_creation_callback != NULL) /* * We need to invoke the sub-channel creation @@ -862,9 +863,8 @@ cleanup: /* * Retrieve the (sub) channel on which to send an outgoing request. - * When a primary channel has multiple sub-channels, we choose a - * channel whose VCPU binding is closest to the VCPU on which - * this call is being made. + * When a primary channel has multiple sub-channels, we try to + * distribute the load equally amongst all available channels. */ struct vmbus_channel *vmbus_get_outgoing_channel(struct vmbus_channel *primary) { @@ -872,11 +872,19 @@ struct vmbus_channel *vmbus_get_outgoing_channel(struct vmbus_channel *primary) int cur_cpu; struct vmbus_channel *cur_channel; struct vmbus_channel *outgoing_channel = primary; - int cpu_distance, new_cpu_distance; + int next_channel; + int i = 1; if (list_empty(&primary->sc_list)) return outgoing_channel; + next_channel = primary->next_oc++; + + if (next_channel > (primary->num_sc)) { + primary->next_oc = 0; + return outgoing_channel; + } + cur_cpu = hv_context.vp_index[get_cpu()]; put_cpu(); list_for_each_safe(cur, tmp, &primary->sc_list) { @@ -887,18 +895,10 @@ struct vmbus_channel *vmbus_get_outgoing_channel(struct vmbus_channel *primary) if (cur_channel->target_vp == cur_cpu) return cur_channel; - cpu_distance = ((outgoing_channel->target_vp > cur_cpu) ? - (outgoing_channel->target_vp - cur_cpu) : - (cur_cpu - outgoing_channel->target_vp)); - - new_cpu_distance = ((cur_channel->target_vp > cur_cpu) ? - (cur_channel->target_vp - cur_cpu) : - (cur_cpu - cur_channel->target_vp)); - - if (cpu_distance < new_cpu_distance) - continue; + if (i == next_channel) + return cur_channel; - outgoing_channel = cur_channel; + i++; } return outgoing_channel; diff --git a/include/linux/hyperv.h b/include/linux/hyperv.h index dd92a85..1ca5824 100644 --- a/include/linux/hyperv.h +++ b/include/linux/hyperv.h @@ -761,6 +761,9 @@ struct vmbus_channel { * link up channels based on their CPU affinity. */ struct list_head percpu_list; + + int num_sc; + int next_oc; }; static inline void set_channel_read_state(struct vmbus_channel *c, bool state) -- 1.7.4.1