Thanks Vasu, I saw the kfree_skb in fcoe_start_io and had a counter there before. I added the fc_frame_free counter out of curiousity to see the frame count there. It seemed interesting that the numbers were the same.
--Tristan -----Original Message----- From: Vasu Dev [mailto:[email protected]] Sent: Tuesday, July 27, 2010 5:46 PM To: Tristan Cheever (tcheever) Cc: Joe Eykholt (jeykholt); [email protected]; Walter Song (wasong) Subject: Re: [Open-FCoE] Huge memory usage with mkfs.ext3 over FCoE On Tue, 2010-07-27 at 17:28 -0700, Tristan Cheever (tcheever) wrote: > One more note... > > I added printk to fc_frame_alloc and fc_frame_free to show the number > of > frames allocated vs freed. > > Without Direct I/O: Number of allocated frames is much greater than > number freed. > With Direct I/O: They are identical > > When going through the code, it seemed as though not all frames are > freed via fc_frame_free. Yes that is the case, since fc_frame_free is just a wrapper around kfree_skb and mostly data frame would be freed by kfree_skb() called from fcoe_start_io(), so add your frame free counter there as well but then skb passed up by netdev/ixgbe are freed using fc_frame_free. So not sure how can your number matches with direct I/O, I'd expect same mismatch w or w/o direct I/O. Vasu _______________________________________________ devel mailing list [email protected] http://www.open-fcoe.org/mailman/listinfo/devel
