patches/kernel/0007-ice-avoid-bonding-causing-auxiliary-plug-unplug-unde.patch

   1 From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
   2 From: Dave Ertman <david.m.ertman@intel.com>
   3 Date: Fri, 10 Mar 2023 11:48:33 -0800
   4 Subject: [PATCH] ice: avoid bonding causing auxiliary plug/unplug under RTNL
   5  lock
   6
   7 RDMA is not supported in ice on a PF that has been added to a bonded
   8 interface. To enforce this, when an interface enters a bond, we unplug
   9 the auxiliary device that supports RDMA functionality.  This unplug
  10 currently happens in the context of handling the netdev bonding event.
  11 This event is sent to the ice driver under RTNL context.  This is causing
  12 a deadlock where the RDMA driver is waiting for the RTNL lock to complete
  13 the removal.
  14
  15 Defer the unplugging/re-plugging of the auxiliary device to the service
  16 task so that it is not performed under the RTNL lock context.
  17
  18 Cc: stable@vger.kernel.org # 6.1.x
  19 Reported-by: Jaroslav Pulchart <jaroslav.pulchart@gooddata.com>
  20 Link: https://lore.kernel.org/netdev/CAK8fFZ6A_Gphw_3-QMGKEFQk=sfCw1Qmq0TVZK3rtAi7vb621A@mail.gmail.com/
  21 Fixes: 5cb1ebdbc434 ("ice: Fix race condition during interface enslave")
  22 Fixes: 4eace75e0853 ("RDMA/irdma: Report the correct link speed")
  23 Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
  24 Tested-by: Arpana Arland <arpanax.arland@intel.com> (A Contingent worker at Intel)
  25 Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
  26 Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
  27 Link: https://lore.kernel.org/r/20230310194833.3074601-1-anthony.l.nguyen@intel.com
  28 Signed-off-by: Jakub Kicinski <kuba@kernel.org>
  29 Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
  30 ---
  31  drivers/net/ethernet/intel/ice/ice.h      | 14 +++++---------
  32  drivers/net/ethernet/intel/ice/ice_main.c | 19 ++++++++-----------
  33  2 files changed, 13 insertions(+), 20 deletions(-)
  34
  35 diff --git a/drivers/net/ethernet/intel/ice/ice.h b/drivers/net/ethernet/intel/ice/ice.h
  36 index 713069f809ec..3cad5e6b2ad1 100644
  37 --- a/drivers/net/ethernet/intel/ice/ice.h
  38 +++ b/drivers/net/ethernet/intel/ice/ice.h
  39 @@ -506,6 +506,7 @@ enum ice_pf_flags {
  40         ICE_FLAG_VF_VLAN_PRUNING,
  41         ICE_FLAG_LINK_LENIENT_MODE_ENA,
  42         ICE_FLAG_PLUG_AUX_DEV,
  43 +       ICE_FLAG_UNPLUG_AUX_DEV,
  44         ICE_FLAG_MTU_CHANGED,
  45         ICE_FLAG_GNSS,                  /* GNSS successfully initialized */
  46         ICE_PF_FLAGS_NBITS              /* must be last */
  47 @@ -950,16 +951,11 @@ static inline void ice_set_rdma_cap(struct ice_pf *pf)
  48   */
  49  static inline void ice_clear_rdma_cap(struct ice_pf *pf)
  50  {
  51 -       /* We can directly unplug aux device here only if the flag bit
  52 -        * ICE_FLAG_PLUG_AUX_DEV is not set because ice_unplug_aux_dev()
  53 -        * could race with ice_plug_aux_dev() called from
  54 -        * ice_service_task(). In this case we only clear that bit now and
  55 -        * aux device will be unplugged later once ice_plug_aux_device()
  56 -        * called from ice_service_task() finishes (see ice_service_task()).
  57 +       /* defer unplug to service task to avoid RTNL lock and
  58 +        * clear PLUG bit so that pending plugs don't interfere
  59          */
  60 -       if (!test_and_clear_bit(ICE_FLAG_PLUG_AUX_DEV, pf->flags))
  61 -               ice_unplug_aux_dev(pf);
  62 -
  63 +       clear_bit(ICE_FLAG_PLUG_AUX_DEV, pf->flags);
  64 +       set_bit(ICE_FLAG_UNPLUG_AUX_DEV, pf->flags);
  65         clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
  66  }
  67  #endif /* _ICE_H_ */
  68 diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c
  69 index 381146282439..56155a04cc0c 100644
  70 --- a/drivers/net/ethernet/intel/ice/ice_main.c
  71 +++ b/drivers/net/ethernet/intel/ice/ice_main.c
  72 @@ -2316,18 +2316,15 @@ static void ice_service_task(struct work_struct *work)
  73                 }
  74         }
  75
  76 -       if (test_bit(ICE_FLAG_PLUG_AUX_DEV, pf->flags)) {
  77 -               /* Plug aux device per request */
  78 -               ice_plug_aux_dev(pf);
  79 +       /* unplug aux dev per request, if an unplug request came in
  80 +        * while processing a plug request, this will handle it
  81 +        */
  82 +       if (test_and_clear_bit(ICE_FLAG_UNPLUG_AUX_DEV, pf->flags))
  83 +               ice_unplug_aux_dev(pf);
  84
  85 -               /* Mark plugging as done but check whether unplug was
  86 -                * requested during ice_plug_aux_dev() call
  87 -                * (e.g. from ice_clear_rdma_cap()) and if so then
  88 -                * plug aux device.
  89 -                */
  90 -               if (!test_and_clear_bit(ICE_FLAG_PLUG_AUX_DEV, pf->flags))
  91 -                       ice_unplug_aux_dev(pf);
  92 -       }
  93 +       /* Plug aux device per request */
  94 +       if (test_and_clear_bit(ICE_FLAG_PLUG_AUX_DEV, pf->flags))
  95 +               ice_plug_aux_dev(pf);
  96
  97         if (test_and_clear_bit(ICE_FLAG_MTU_CHANGED, pf->flags)) {
  98                 struct iidc_event *event;