[PATCH 3/3] buffer audit msgs directly to skb

David Woodhouse dwmw2 at infradead.org
Wed May 18 08:35:53 UTC 2005


On Fri, 2005-05-06 at 00:55 -0700, Chris Wright wrote:
> Drop the use of a tmp buffer in the audit_buffer, and just buffer
> directly to the skb.  All header data that was temporarily stored in
> the audit_buffer can now be stored directly in the netlink header in
> the skb.  Resize skb as needed.  This eliminates the extra copy (and
> the audit_log_move function which was responsible for copying).

Actually it doesn't, does it? We used to pass an optimally-sized skb
into netlink_unicast() but now we pass one that's larger than it needs
to be.... so netlink_trim() is going to trim it.

I'm going to look at cleaning up this code. Probably the answer will be
to revert to using the tmp buffer, and queue optimally-sized skbs for
netlink.

-- 
dwmw2




More information about the Linux-audit mailing list