Change Details

By default, lazy_tx_credit_flush will process packets by batch of 15 packets or less. With this change, batches or 64 packets are directly processed. Under high load, this prevents to fill up completely the TX ring and then to flush it entirely. Submitted by: Marc De La Gueronniere <mdelagueronniere@verisign.com> MFC after: 1 week Sponsored by: Verisign, Inc.

Previously, when transmitting short runs of packets via cxgbe_nm_tx(), we would wait until a large number of packets were buffered before scheduling a task to clean transmit buffers. Obtained from: np

By defaultPreviously, lazy_tx_credit_flush will processwhen transmitting short runs of packets by batch of 15via cxgbe_nm_tx(), packets or less. With this change, batches or 64we would wait until a large number of packets are directlywere buffered before processed. Under high load, this prevents to fill up completely the TX ring and then to flush it entirelyscheduling a task to clean transmit buffers. Submitted by: Marc De La Gueronniere <mdelagueronniere@verisign.com> MFC after: 1 week Sponsored by: Verisign, Inc.Obtained from: np