On Fri, Jul 29, 2011 at 09:16:42AM +0300, Michael S. Tsirkin wrote:

On Tue, Jul 26, 2011 at 12:05:36PM -0400, Neil Horman wrote:

quoted

Ok, after considering all your comments, Dave suggested this as an alternate
approach:

1) We create a new priv_flag, IFF_SKB_TX_SHARED, to identify drivers capable of
handling shared skbs.  Default is to not set this flag

2) Modify ether_setup to enable this flag, under the assumption that any driver
calling this  function is initalizing a real ethernet device and as such can
handle shared skbs since they don't tend to store state in the skb struct.
Pktgen can then query this flag when a user script attempts to issue the
clone_skb command and decide if it is to be alowed or not.

3) Audit the network drivers calling ether_setup to identify any code doing so
that can't actualy handle shared skbs and manually disable the new flag.  There
are about 10 drivers in this category.

Change notes:
v3) Fixed Erics note in which I tested the length of the passed in string rather
than its converted value for beign > 0

Thoughts/reviews aprpeciated.
Neil

It might be a good idea to disable vhost-net zerocopy for
such devices as well: these skbs are shared with userspace.
Shirley, what do you think?

I don't think thats a problem, since (IIUC) only skbs with (tx_flags &
SKBTX_DEV_ZEROCOPY) set can do that.  The pktgen skbs originate in the kernel
and never have that flag set).
Neil

P.S. Don't call me Shirley :).  Sorry, its not every day you get to use that
line.

-- 
MST