This patch series reduces fcoe_pending_queue locking from two to one for
each new skb added to this queue and improves its timer handling to flush
pending skb in from this queue.
I noticed +10% increase in performance from no use of qfull check in
the fcoe-fixes tip, this series not only fixes performance regression
caused by added qfull check but also improves performance by use of quick
2 jiffies timer on demand.
RFC is for:
1. Appreciate your review comments in modified fcoe_pending_queue handling.
2. It would be great, if Mike you could check performance number difference
with these patches for reported regression by you due to added qfull check.
TODO:
I need to test these patches for any out of order tx framing caused in
past in fcoe_pending_queue handling.
Vasu
---
Vasu Dev (2):
fcoe: removes fcoe_watchdog
fcoe: reduces lock cost when adding a new skb to fcoe_pending_queue
drivers/scsi/fcoe/fcoe.c | 87 +++++++++++++++++-----------------------------
drivers/scsi/fcoe/fcoe.h | 1 +
2 files changed, 33 insertions(+), 55 deletions(-)
--
Signature: [email protected]
_______________________________________________
devel mailing list
[email protected]
http://www.open-fcoe.org/mailman/listinfo/devel
