lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <1430154256-26853-7-git-send-email-vkuznets@redhat.com>
Date:	Mon, 27 Apr 2015 19:04:16 +0200
From:	Vitaly Kuznetsov <vkuznets@...hat.com>
To:	"K. Y. Srinivasan" <kys@...rosoft.com>
Cc:	Haiyang Zhang <haiyangz@...rosoft.com>,
	devel@...uxdriverproject.org, linux-kernel@...r.kernel.org,
	Dexuan Cui <decui@...rosoft.com>
Subject: [PATCH v2 6/6] Drivers: hv: vmbus: improve selection of an outgoing channel

vmbus_get_outgoing_channel() implements the following algorithm for selecting
an outgoing channel (despite the comment before the function saying it
distributes the load equally):
1) If we have no subchannels return the primary channel;
2) If primary->next_oc is grater than primary->num_sc reset the primary->next_oc
   to 0 and return the primary channel;
3) Aim for the primary->next_oc subchannel, increment primary->next_oc;
4) Loop through all opened subchannels. If we see a channel which has
 target_cpu == current_cpu return it. If we reached the primary->next_oc'th
open subchannel return it;
5) Return the primary channel.
The implementation also skips the subchannel No. 0 unless it matches the current
cpu as we assign i to 1 in the initialization.

Improve the algorithm to do the full scan searching for a (sub)channel with the
target_cpu == current_cpu, use round-robin as a fallback.

Signed-off-by: Vitaly Kuznetsov <vkuznets@...hat.com>
---
 drivers/hv/channel_mgmt.c | 39 +++++++++++++++++++--------------------
 1 file changed, 19 insertions(+), 20 deletions(-)

diff --git a/drivers/hv/channel_mgmt.c b/drivers/hv/channel_mgmt.c
index 1f1417d..cdc3914 100644
--- a/drivers/hv/channel_mgmt.c
+++ b/drivers/hv/channel_mgmt.c
@@ -823,46 +823,45 @@ cleanup:
 }
 
 /*
- * Retrieve the (sub) channel on which to send an outgoing request.
- * When a primary channel has multiple sub-channels, we try to
- * distribute the load equally amongst all available channels.
+ * Retrieve the (sub) channel on which to send an outgoing request. When a
+ * primary channel has multiple sub-channels, we try to find a (sub)channel
+ * with target_cpu == current CPU and we try to distribute the load equally
+ * amongst all available channels when we fail.
  */
 struct vmbus_channel *vmbus_get_outgoing_channel(struct vmbus_channel *primary)
 {
 	struct list_head *cur, *tmp;
-	int cur_cpu;
 	struct vmbus_channel *cur_channel;
-	struct vmbus_channel *outgoing_channel = primary;
-	int next_channel;
-	int i = 1;
+	struct vmbus_channel *out_channel_rr = NULL;
+	int i = 0;
 
 	if (list_empty(&primary->sc_list))
-		return outgoing_channel;
+		return primary;
 
-	next_channel = primary->next_oc++;
-
-	if (next_channel > (primary->num_sc)) {
+	if (primary->next_oc >= primary->num_sc) {
 		primary->next_oc = 0;
-		return outgoing_channel;
+		return primary;
 	}
 
-	cur_cpu = hv_context.vp_index[get_cpu()];
-	put_cpu();
+	if (primary->target_cpu == smp_processor_id())
+		return primary;
+
 	list_for_each_safe(cur, tmp, &primary->sc_list) {
+		i++;
 		cur_channel = list_entry(cur, struct vmbus_channel, sc_list);
 		if (cur_channel->state != CHANNEL_OPENED_STATE)
 			continue;
 
-		if (cur_channel->target_vp == cur_cpu)
+		if (cur_channel->target_cpu == smp_processor_id())
 			return cur_channel;
 
-		if (i == next_channel)
-			return cur_channel;
-
-		i++;
+		if (!out_channel_rr && i > primary->next_oc) {
+			primary->next_oc = i;
+			out_channel_rr = cur_channel;
+		}
 	}
 
-	return outgoing_channel;
+	return out_channel_rr ? out_channel_rr : primary;
 }
 EXPORT_SYMBOL_GPL(vmbus_get_outgoing_channel);
 
-- 
1.9.3

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ