On Wed, 2011-11-09 at 12:40 -0500, Mark Lord wrote:

On 11-11-09 12:31 PM, Ben Hutchings wrote:

quoted

On Wed, 2011-11-09 at 12:20 -0500, Mark Lord wrote:

quoted

On 11-11-09 11:57 AM, Mark Lord wrote:

quoted

On 11-11-09 11:47 AM, Mark Lord wrote:
..

quoted

Note:  I'm looking at smsc95xx.c and smsc75xx.c for examples,
and they both have the same problem I'll have here:

How to update the csum settings atomically.
A spinlock is no good, because config register access is over USB.

Nevermind.. a slight change in the logic and all is well again.

..

Or even simpler (below).  I don't think this method requires any
extra locking, but I'm still open to persuasion.  :)

Looks reasonable, but...

quoted

static int ax88772b_set_features(struct net_device *netdev, u32 features)
{
        struct usbnet *dev = netdev_priv(netdev);
        struct ax88772b_data *priv = (struct ax88772b_data *)dev->driver_priv;
        u16 tx_csum = 0, rx_csum = 0;

        priv->features = features & (NETIF_F_HW_CSUM | NETIF_F_RXCSUM);

...why do you need priv->features at all?

There's code elsewhere that takes action under some conditions
based on the current setting of the NETIF_F_RXCSUM flag.

I don't claim to fully understand what's going on,
but it doesn't care much about races on set/clear of the flag.

And it can use dev->features.

Ben.

-- 
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.