]> www.infradead.org Git - users/willy/pagecache.git/commit
Merge branch 'net-ethtool-fixes-for-hds-threshold'
authorJakub Kicinski <kuba@kernel.org>
Mon, 20 Jan 2025 19:45:01 +0000 (11:45 -0800)
committerJakub Kicinski <kuba@kernel.org>
Mon, 20 Jan 2025 19:45:01 +0000 (11:45 -0800)
commit01f5f35ae4c96f127b997be2af7cf91f21a92d7f
tree49f9dc74373711be6ce2aebabfa60564977b0d68
parente81fdf7078e950d3a5c78cc640a2ad58a523092d
parent99d028c63457773152964ba3abc0d8fb8cf220bf
Merge branch 'net-ethtool-fixes-for-hds-threshold'

Jakub Kicinski says:

====================
net: ethtool: fixes for HDS threshold

Quick follow up on the HDS threshold work, since the merge window
is upon us.

Fix the bnxt implementation to apply the settings right away,
because we update the parameters _after_ configuring HW user
needed to reconfig the device twice to get the settings to stick.

For this I took the liberty of moving the config to a separate
struct. This follows my original thinking for the queue API.
It should also fit more neatly into how many drivers which
support safe config update operate. Drivers can allocate
new objects using the "pending" struct.

netdevsim:

  KTAP version 1
  1..7
  ok 1 hds.get_hds
  ok 2 hds.get_hds_thresh
  ok 3 hds.set_hds_disable
  ok 4 hds.set_hds_enable
  ok 5 hds.set_hds_thresh_zero
  ok 6 hds.set_hds_thresh_max
  ok 7 hds.set_hds_thresh_gt
  # Totals: pass:7 fail:0 xfail:0 xpass:0 skip:0 error:0

bnxt:

  KTAP version 1
  1..7
  ok 1 hds.get_hds
  ok 2 hds.get_hds_thresh
  ok 3 hds.set_hds_disable # SKIP disabling of HDS not supported by the device
  ok 4 hds.set_hds_enable
  ok 5 hds.set_hds_thresh_zero
  ok 6 hds.set_hds_thresh_max
  ok 7 hds.set_hds_thresh_gt
  # Totals: pass:6 fail:0 xfail:0 xpass:0 skip:1 error:0

v1: https://lore.kernel.org/20250117194815.1514410-1-kuba@kernel.org
====================

Link: https://patch.msgid.link/20250119020518.1962249-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>