src/sys/dev/cxgbe/t4_netmap.c, branch releng/12.2

MFC r362616:

2020-08-25T02:54:48Z

cxgbe(4): Add a pointer to the adapter softc in vi_info. There were quite a few places where port_info was being accessed only to get to the adapter. Sponsored by: Chelsio Communications

MFC r351444, r357475, r357479, r357481-r357482, r358859, and r364497.

2020-08-25T02:14:36Z

All these are rx improvements in the cxgbe(4) driver. r351444: cxgbe(4): Use the same buffer size for TOE rx queues as the NIC rx queues. This is a minor simplification. r357475: cxgbe(4): Initialize the rx buffer's metadata on first-use and not on allocation. refill_fl doesn't touch any part of a freshly allocated cluster after this change. r357479: cxgbe(4): Avoid ext_arg2 in rxb_free. ext_arg2 is the only item in the third cacheline in an mbuf and could be cold by the time rxb_free runs. Put the information needed by rxb_free in the same line as the refcount, which is very likely to be hot given that rxb_free runs when the refcount is decremented and reaches 0. r357481: cxgbe(4): Retire the allow_mbufs_in_cluster optimization. This simplifies the driver's rx fast path as well as the bookkeeping code that tracks various rx buffer sizes and layouts. r357482: cxgbe(4): Treat NIC rx as special and run its handler directly and not via the t4_cpl_handler dispatch table. r358859: cxgbe(4): Do not try to use 0 as an rx buffer address when the driver is already allocating from the safe zone and the allocation fails. This bug was introduced in r357481. r364497: cxgbe(4): Use large clusters for TOE rx queues when TOE+TLS is enabled. Rx is more efficient within the chip when the receive buffer size matches the TLS PDU size. Sponsored by: Chelsio Communications

MFC r359159:

2020-03-27T03:58:00Z

cxgbe(4): Split sge_nm_rxq into three cachelines. This reduces the lines bouncing around between the driver rx ithread and the netmap rxsync thread. There is no net change in the size of the struct (it continues to waste a lot of space). This kind of split was originally proposed in D17869 by Marc De La Gueronniere @ Verisign, Inc. Sponsored by: Chelsio Communications

MFC r357552:

2020-03-14T02:05:12Z

cxgbe(4): Allow nm_black_hole and nm_cong_drop to be set at any time. The cong_drop setting will apply to queues created after the setting is changed and not to existing queues. Sponsored by: Chelsio Communications

MFC r357551:

2020-03-14T01:59:39Z

cxgbe(4): Report accurate rx_buf_maxsize to netmap. Sponsored by: Chelsio Communications

MFC r355580, r355673, and r357553.

2020-03-14T01:03:48Z

r355580: cxgbe(4): Use TX_PKTS2 work requests in netmap Tx if it's available. TX_PKTS2 is more efficient within the firmware and this improves netmap Tx by a few Mpps in some common scenarios. r355673: cxgbe(4): Never use hardware checksumming in netmap tx. r357553: cxgbe(4): Add a knob to allow netmap tx traffic to be checksummed by the hardware. hw.cxgbe.nm_txcsum=1 Sponsored by: Chelsio Communications

MFC r354106:

2019-11-30T17:56:04Z

cxgbe(4): Use correct FetchBurstMin values for T6. Sponsored by: Chelsio Communications

MFC r340375 (by jch@):

2019-04-30T21:24:44Z

cxgbe/netmap: Fix cxgbe netmap when interface is DOWN A kernel panic can occur if the cxgbe interface is DOWN when activating netmap. This patch prevents the driver from freeing up cxgbe netmap resources when they have not been allocated. Submitted by: Nicolas Witkowski Reviewed by: np Sponsored by: Verisign, Inc. Differential Revision: https://reviews.freebsd.org/D17802

MFC r339717, r339749, and r339809.

2019-03-27T19:37:50Z

r339717: cxgbe(4): Allow "pass" filters to distribute matching traffic using a subset of a VI's RSS indirection table. This makes it possible to make groups out of rx queues and steer different kinds of traffic to different groups. For example, an interface with 8 rx queues could have all non-TCP traffic delivered to queues 0-3 and all TCP traffic to queues 4-7. Note that it is already possible for filters to steer traffic to a particular queue or to distribute it using the full indirection table (much like normal rx does). Sponsored by: Chelsio Communications r339749: cxgbe(4): Add a knob to split the rx queues for a netmap enabled interface into two groups. Filters can be used to match traffic and distribute it across a group. hw.cxgbe.nm_split_rss Sponsored by: Chelsio Communications r339809: cxgbetool(8): Add a subaction (tcbrss ) that can be used with "pass" action to distribute traffic using the half of the VI's RSS indirection table. The value specified should either be the start of the VI's RSS slice (available at dev...rss_base since r339700) or the midpoint (rss_base + rss_size/2). The traffic that hits the filter will use the first or second half of the indirection table respectively. The indirection table can be populated in different ways to achieve different kinds of traffic/load distributions. For example, r339749 allows a netmap interface to have half the rx queues in the first half of the table and the rest in the other. Sponsored by: Chelsio Communications

MFC r339705:

2019-03-27T19:24:37Z

cxgbe(4): Update the VI's default queue when netmap is enabled/disabled. Sponsored by: Chelsio Communications