Re: [PATCH 2.6.18-rc4] aoe [04/13]: zero copy write 1 of 2

["Ed L. Cashin" <ecashin@coraid.com>]

On Sat, Aug 19, 2006 at 11:18:12AM +0100, Alan Cox wrote:
> Ar Gwe, 2006-08-18 am 13:39 -0400, ysgrifennodd Ed L. Cashin:
> > Signed-off-by: "Ed L. Cashin" <ecashin@coraid.com>
> 
> > +	skb->len = sizeof *h + sizeof *ah;
> > +	memset(h, 0, skb->len);
> 
> Never play with skb->len directly. Use skb_put/skb_trim

These are skbs pre-allocated by the aoe driver that will always have
enough room to accomodate this much data, and we are really setting
the packet header length.

To use skb_put here seems awkward.  We'd have to do things like shown
below throughout the driver instead of just setting the length.  Is
that what you'd like to see?

diff -upr 2.6.18-rc4-orig/drivers/block/aoe/aoecmd.c 2.6.18-rc4-aoe/drivers/block/aoe/aoecmd.c
--- 2.6.18-rc4-orig/drivers/block/aoe/aoecmd.c	2006-08-22 12:48:18.000000000 -0400
+++ 2.6.18-rc4-aoe/drivers/block/aoe/aoecmd.c	2006-08-22 17:03:23.000000000 -0400
@@ -314,7 +315,9 @@ rexmit(struct aoedev *d, struct frame *f
 		if (ah->aflags & AOEAFL_WRITE) {
 			skb_fill_page_desc(skb, 0, virt_to_page(f->bufaddr),
 				offset_in_page(f->bufaddr), DEFAULTBCNT);
-			skb->len = sizeof *h + sizeof *ah + DEFAULTBCNT;
+			skb->data_len = 0;
+			skb_trim(skb, 0);
+			skb_put(skb, sizeof *h + sizeof *ah + DEFAULTBCNT);
 			skb->data_len = DEFAULTBCNT;
 		}
 		if (++d->lostjumbo > (d->nframes << 1))


-- 
  Ed L Cashin <ecashin@coraid.com>