[PATCH] scsi/fcoe: remove fcoe_select_cpu()

[PATCH] scsi/fcoe: remove fcoe_select_cpu()

[Hannes Reinecke]

The function fcoe_select_cpu() is just used to distribute incoming
skbs which start a new FC command sequence. But the network stack
already received (and processed) that skb, and there is a _really_
good chance that all subsequent skbs for this sequence will be
handled with the same CPU. So we should just use the cpu on which
this skb was allocated on and save ourselves some overhead due
to pointless scheduling.

Signed-off-by: Hannes Reinecke <hare@kernel.org>
---
 drivers/scsi/fcoe/fcoe.c | 22 +---------------------
 1 file changed, 1 insertion(+), 21 deletions(-)

diff --git a/drivers/scsi/fcoe/fcoe.c b/drivers/scsi/fcoe/fcoe.c
index b911fdb387f3..4912087de10d 100644
--- a/drivers/scsi/fcoe/fcoe.c
+++ b/drivers/scsi/fcoe/fcoe.c
@@ -1299,26 +1299,6 @@ static void fcoe_thread_cleanup_local(unsigned int cpu)
 	flush_work(&p->work);
 }
 
-/**
- * fcoe_select_cpu() - Selects CPU to handle post-processing of incoming
- *			command.
- *
- * This routine selects next CPU based on cpumask to distribute
- * incoming requests in round robin.
- *
- * Returns: int CPU number
- */
-static inline unsigned int fcoe_select_cpu(void)
-{
-	static unsigned int selected_cpu;
-
-	selected_cpu = cpumask_next(selected_cpu, cpu_online_mask);
-	if (selected_cpu >= nr_cpu_ids)
-		selected_cpu = cpumask_first(cpu_online_mask);
-
-	return selected_cpu;
-}
-
 /**
  * fcoe_rcv() - Receive packets from a net device
  * @skb:    The received packet
@@ -1405,7 +1385,7 @@ static int fcoe_rcv(struct sk_buff *skb, struct net_device *netdev,
 		cpu = ntohs(fh->fh_ox_id) & fc_cpu_mask;
 	else {
 		if (ntohs(fh->fh_rx_id) == FC_XID_UNKNOWN)
-			cpu = fcoe_select_cpu();
+			cpu = skb->alloc_cpu;
 		else
 			cpu = ntohs(fh->fh_rx_id) & fc_cpu_mask;
 	}
-- 
2.35.3

Re: [PATCH] scsi/fcoe: remove fcoe_select_cpu()

[Martin K. Petersen]

Hannes,

> The function fcoe_select_cpu() is just used to distribute incoming
> skbs which start a new FC command sequence. But the network stack
> already received (and processed) that skb, and there is a _really_
> good chance that all subsequent skbs for this sequence will be handled
> with the same CPU. So we should just use the cpu on which this skb was
> allocated on and save ourselves some overhead due to pointless
> scheduling.

Applied to 6.17/scsi-staging, thanks!

-- 
Martin K. Petersen