Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 13 May 2016 06:47:47 +0000 (UTC)
From:      Andrew Rybchenko <arybchik@FreeBSD.org>
To:        src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org
Subject:   svn commit: r299595 - head/sys/dev/sfxge/common
Message-ID:  <201605130647.u4D6llCU030763@repo.freebsd.org>

next in thread | raw e-mail | index | archive | help
Author: arybchik
Date: Fri May 13 06:47:47 2016
New Revision: 299595
URL: https://svnweb.freebsd.org/changeset/base/299595

Log:
  sfxge(4): comment on when we assume multicast chaining is available
  
  It's the same on Medford as Huntington.
  
  Multicast chaining is not always on, even with Medford, as it's not
  supported by low latency firmware.
  
  Unlike the Linux driver, we don't need to support virtulization with
  firmware released before support for multicast chaining was added.
  
  Submitted by:   Mark Spender <mspender at solarflare.com>
  Sponsored by:   Solarflare Communications, Inc.
  MFC after:      1 week
  Differential Revision:  https://reviews.freebsd.org/D6319

Modified:
  head/sys/dev/sfxge/common/hunt_filter.c

Modified: head/sys/dev/sfxge/common/hunt_filter.c
==============================================================================
--- head/sys/dev/sfxge/common/hunt_filter.c	Fri May 13 06:47:07 2016	(r299594)
+++ head/sys/dev/sfxge/common/hunt_filter.c	Fri May 13 06:47:47 2016	(r299595)
@@ -1191,7 +1191,7 @@ ef10_filter_remove_old(
 
 
 static	__checkReturn	efx_rc_t
-hunt_filter_get_workarounds(
+ef10_filter_get_workarounds(
 	__in				efx_nic_t *enp)
 {
 	efx_nic_cfg_t *encp = &enp->en_nic_cfg;
@@ -1319,18 +1319,28 @@ ef10_filter_reconfigure(
 	 * filters, and can only be enabled or disabled when the hardware filter
 	 * table is empty.
 	 *
+	 * Chained multicast filters require support from the datapath firmware,
+	 * and may not be available (e.g. low-latency variants or old Huntington
+	 * firmware).
+	 *
 	 * Firmware will reset (FLR) functions which have inserted filters in
 	 * the hardware filter table when the workaround is enabled/disabled.
 	 * Functions without any hardware filters are not reset.
 	 *
 	 * Re-check if the workaround is enabled after adding unicast hardware
-	 * filters. This ensures that encp->enc_workaround_bug26807 matches the
+	 * filters. This ensures that encp->enc_bug26807_workaround matches the
 	 * firmware state, and that later changes to enable/disable the
 	 * workaround will result in this function seeing a reset (FLR).
 	 *
-	 * FIXME: On Medford multicast chaining should always be on.
+	 * In common-code drivers, we only support multiple PCI function
+	 * scenarios with firmware that supports multicast chaining, so we can
+	 * assume it is enabled for such cases and hence simplify the filter
+	 * insertion logic. Firmware that does not support multicast chaining
+	 * does not support multiple PCI function configurations either, so
+	 * filter insertion is much simpler and the same strategies can still be
+	 * used.
 	 */
-	if ((rc = hunt_filter_get_workarounds(enp)) != 0)
+	if ((rc = ef10_filter_get_workarounds(enp)) != 0)
 		goto fail2;
 
 	if ((table->eft_using_all_mulcst != all_mulcst) &&



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201605130647.u4D6llCU030763>