[PATCH iwl-net 0/10] ice: bug fixes -- no link, tc-flower, VF rate

[Aleksandr Loktionov <aleksandr.loktionov@intel.com>]

Good day,

Ten stand-alone bug fixes for the ice driver, all targeting functional
regressions or silent failures that affect connectivity.

Service task / unload races:
  - A scheduled service-task work item could fire after ice_remove()
    started, racing against freed per-PF state.  Check ICE_SHUTTING_DOWN
    at the top of the task body to close the window.

  - The DPLL input-pin callback accepted DPLL_PIN_STATE_CONNECTED and
    silently converted it to DISCONNECTED instead of returning an error.

Mirror / switch:
  - Mirror rules whose action is "to VSI list" had VSI count never set
    to 1, breaking subsequent deletions when a VSI list was involved.

DCB / LLDP:
  - Two missing `need_reconfig = true` assignments caused FW updates to
    be silently skipped when a DCB change only touched the TC bandwidth
    or TSA tables.

  - ice_set_dflt_mib() sent the initial default MIB to firmware with
    the ETS willing bit clear, preventing LLDP negotiation with a peer
    when the FW LLDP agent is active.

ethtool / link modes:
  - 10000baseCR_Full was absent from ice_adv_lnk_speed_10000[], causing
    an ethtool autoneg advertisement of this mode to be silently ignored
    (user gets "nothing changed" even on capable hardware).

  - ice_set_link_ksettings() derived the Autoneg-capable flag only from
    ice_phy_type_to_ethtool(); PHYs that report autoneg solely through
    PHY capabilities were missed.  Pull the flag from
    ice_is_phy_caps_an_enabled() to fix this.
    Fixes: 5cd349c349d6 ("ice: report supported and advertised autoneg using PHY capabilities")

  - 50G single-lane modes (50000baseCR/KR/SR/LR_ER_FR_Full) were absent
    from ice_adv_lnk_speed_50000[], so those link modes were silently
    rejected by ethtool.
    Fixes: 982b0192db45 ("ice: Refactor finding advertised link speed")

TC flower / switch:
  - ice_find_dummy_packet() chose the UDP inner template for rules that
    match ICE_IPV4_IL with ip_proto==TCP but no explicit L4 field,
    causing tc-flower rule installs to fail.
    Fixes: e33163a40d1a ("ice: switch: convert packet template match code to rodata")

SR-IOV / VF:
  - ice_set_vf_bw() refused any min_tx_rate value when total guaranteed
    bandwidth was already oversubscribed, making it impossible to clear
    the rate via "ip link set <pf> vf <id> min_tx_rate 0".  Allow a
    zero to bypass the oversubscription check so users can recover.

With the best regards,
Alex

---
Michal Swiatkowski (2):
  ice: fix mirroring to VSI list
  ice: select inner TCP dummy packet when matching on ip_proto TCP without explicit L4 field

Dave Ertman (2):
  ice: update FW on all DCB changes
  ice: disallow service task to run while driver is unloading

Arkadiusz Kubalewski (1):
  ice: error out on CONNECTED state for input pin

Voon Weifeng (1):
  ice: add 10000baseCR_Full to advertised link speed map

Jan Glaza (1):
  ice: check PHY autoneg capability before rejecting ethtool autoneg setting

Grzegorz Nitka (1):
  ice: fix missing 50G single-lane ethtool link speed mappings

Yochai Hagvi (1):
  ice: set ETS TLV willing bit in default MIB sent to firmware

Sudheer Mogilappagari (1):
  ice: allow setting min_tx_rate to 0 to resolve VF bandwidth oversubscription

 drivers/net/ethernet/intel/ice/ice_dcb.h     |  2 +
 drivers/net/ethernet/intel/ice/ice_dcb_lib.c |  6 ++-
 drivers/net/ethernet/intel/ice/ice_dcb_nl.c  |  0
 drivers/net/ethernet/intel/ice/ice_dpll.c    |  2 ++
 drivers/net/ethernet/intel/ice/ice_ethtool.c | 29 +++++++++--
 drivers/net/ethernet/intel/ice/ice_main.c    |  2 +
 drivers/net/ethernet/intel/ice/ice_sriov.c   |  8 +++
 drivers/net/ethernet/intel/ice/ice_switch.c  |  8 ++-
 8 files changed, 52 insertions(+), 5 deletions(-)