igb writes to doorbells to post transmit and receive descriptors;
after writing descriptors to memory but before writing to doorbells,
use dma_wmb() rather than wmb(). wmb() is more heavyweight than
necessary before doorbell writes.

On x86, this avoids SFENCEs before doorbell writes in both the
tx and rx refill paths.
Signed-off-by: NVenkatesh Srinivas <venkateshs@google.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>