[PATCH 2/2] Drivers: hv: vmbus: Implement per-CPU mapping of relid to channel

From: K. Y. Srinivasan
Date: Tue Apr 08 2014 - 20:51:48 EST


Currently the mapping of the relID to channel is done under the protection of a
single spin lock. Starting with ws2012, each channel is bound to a specific VCPU
in the guest. Use this binding to eliminate the spin lock by setting up
per-cpu state for mapping relId to the channel.

Signed-off-by: K. Y. Srinivasan <kys@xxxxxxxxxxxxx>
Reviewed-by: Haiyang Zhang <haiyangz@xxxxxxxxxxxxx>
---
drivers/hv/channel_mgmt.c | 41 ++++++++++++++++++++++++++++++++++++++++-
drivers/hv/connection.c | 24 +++++++++++++++++++++++-
drivers/hv/hv.c | 2 ++
drivers/hv/hyperv_vmbus.h | 5 +++++
include/linux/hyperv.h | 5 +++++
5 files changed, 75 insertions(+), 2 deletions(-)

diff --git a/drivers/hv/channel_mgmt.c b/drivers/hv/channel_mgmt.c
index 6f7fdd9..6c8b032c 100644
--- a/drivers/hv/channel_mgmt.c
+++ b/drivers/hv/channel_mgmt.c
@@ -149,6 +149,7 @@ static struct vmbus_channel *alloc_channel(void)
spin_lock_init(&channel->sc_lock);

INIT_LIST_HEAD(&channel->sc_list);
+ INIT_LIST_HEAD(&channel->percpu_list);

channel->controlwq = create_workqueue("hv_vmbus_ctl");
if (!channel->controlwq) {
@@ -188,7 +189,20 @@ static void free_channel(struct vmbus_channel *channel)
queue_work(vmbus_connection.work_queue, &channel->work);
}

+static void percpu_channel_enq(void *arg)
+{
+ struct vmbus_channel *channel = arg;
+ int cpu = smp_processor_id();
+
+ list_add_tail(&channel->percpu_list, &hv_context.percpu_list[cpu]);
+}

+static void percpu_channel_deq(void *arg)
+{
+ struct vmbus_channel *channel = arg;
+
+ list_del(&channel->percpu_list);
+}

/*
* vmbus_process_rescind_offer -
@@ -210,6 +224,12 @@ static void vmbus_process_rescind_offer(struct work_struct *work)
msg.header.msgtype = CHANNELMSG_RELID_RELEASED;
vmbus_post_msg(&msg, sizeof(struct vmbus_channel_relid_released));

+ if (channel->target_cpu != smp_processor_id())
+ smp_call_function_single(channel->target_cpu,
+ percpu_channel_deq, channel, true);
+ else
+ percpu_channel_deq(channel);
+
if (channel->primary_channel == NULL) {
spin_lock_irqsave(&vmbus_connection.channel_lock, flags);
list_del(&channel->listentry);
@@ -245,6 +265,7 @@ static void vmbus_process_offer(struct work_struct *work)
work);
struct vmbus_channel *channel;
bool fnew = true;
+ bool enq = false;
int ret;
unsigned long flags;

@@ -264,12 +285,22 @@ static void vmbus_process_offer(struct work_struct *work)
}
}

- if (fnew)
+ if (fnew) {
list_add_tail(&newchannel->listentry,
&vmbus_connection.chn_list);
+ enq = true;
+ }

spin_unlock_irqrestore(&vmbus_connection.channel_lock, flags);

+ if (enq) {
+ if (newchannel->target_cpu != smp_processor_id())
+ smp_call_function_single(newchannel->target_cpu,
+ percpu_channel_enq,
+ newchannel, true);
+ else
+ percpu_channel_enq(newchannel);
+ }
if (!fnew) {
/*
* Check to see if this is a sub-channel.
@@ -282,6 +313,14 @@ static void vmbus_process_offer(struct work_struct *work)
spin_lock_irqsave(&channel->sc_lock, flags);
list_add_tail(&newchannel->sc_list, &channel->sc_list);
spin_unlock_irqrestore(&channel->sc_lock, flags);
+
+ if (newchannel->target_cpu != smp_processor_id())
+ smp_call_function_single(newchannel->target_cpu,
+ percpu_channel_enq,
+ newchannel, true);
+ else
+ percpu_channel_enq(newchannel);
+
newchannel->state = CHANNEL_OPEN_STATE;
if (channel->sc_creation_callback != NULL)
channel->sc_creation_callback(newchannel);
diff --git a/drivers/hv/connection.c b/drivers/hv/connection.c
index d484bad..3589236 100644
--- a/drivers/hv/connection.c
+++ b/drivers/hv/connection.c
@@ -231,6 +231,28 @@ cleanup:
return ret;
}

+/*
+ * Map the given relid to the corresponding channel based on the
+ * per-cpu list of channels that have been affinitized to this CPU.
+ * This will be used in the channel callback path as we can do this
+ * mapping in a lock-free fashion.
+ */
+static struct vmbus_channel *pcpu_relid2channel(u32 relid)
+{
+ struct vmbus_channel *channel;
+ struct vmbus_channel *found_channel = NULL;
+ int cpu = smp_processor_id();
+ struct list_head *pcpu_head = &hv_context.percpu_list[cpu];
+
+ list_for_each_entry(channel, pcpu_head, percpu_list) {
+ if (channel->offermsg.child_relid == relid) {
+ found_channel = channel;
+ break;
+ }
+ }
+
+ return found_channel;
+}

/*
* relid2channel - Get the channel object given its
@@ -282,7 +304,7 @@ static void process_chn_event(u32 relid)
* Find the channel based on this relid and invokes the
* channel callback to process the event
*/
- channel = relid2channel(relid);
+ channel = pcpu_relid2channel(relid);

if (!channel) {
pr_err("channel not found for relid - %u\n", relid);
diff --git a/drivers/hv/hv.c b/drivers/hv/hv.c
index bcb4950..edfc848 100644
--- a/drivers/hv/hv.c
+++ b/drivers/hv/hv.c
@@ -383,6 +383,8 @@ void hv_synic_init(void *arg)
*/
rdmsrl(HV_X64_MSR_VP_INDEX, vp_index);
hv_context.vp_index[cpu] = (u32)vp_index;
+
+ INIT_LIST_HEAD(&hv_context.percpu_list[cpu]);
return;
}

diff --git a/drivers/hv/hyperv_vmbus.h b/drivers/hv/hyperv_vmbus.h
index 860134d..18d1a84 100644
--- a/drivers/hv/hyperv_vmbus.h
+++ b/drivers/hv/hyperv_vmbus.h
@@ -510,6 +510,11 @@ struct hv_context {
* basis.
*/
struct tasklet_struct *event_dpc[NR_CPUS];
+ /*
+ * To optimize the mapping of relid to channel, maintain
+ * per-cpu list of the channels based on their CPU affinity.
+ */
+ struct list_head percpu_list[NR_CPUS];
};

extern struct hv_context hv_context;
diff --git a/include/linux/hyperv.h b/include/linux/hyperv.h
index 8b41570..ebd0441 100644
--- a/include/linux/hyperv.h
+++ b/include/linux/hyperv.h
@@ -732,6 +732,11 @@ struct vmbus_channel {
* Support per-channel state for use by vmbus drivers.
*/
void *per_channel_state;
+ /*
+ * To support per-cpu lookup mapping of relid to channel,
+ * link up channels based on their CPU affinity.
+ */
+ struct list_head percpu_list;
};

static inline void set_channel_read_state(struct vmbus_channel *c, bool state)
--
1.7.4.1

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