[Intel-wired-lan] [PATCH net v2] ice: Stop processing VF messages during teardown

[Anirudh Venkataramanan]

When VFs are setup and torn down in quick succession, it is possible that
a VF is torn down by the PF while its virtchnl requests are still in the
PF's mailbox ring. When the PF eventually processes a non-existant VF's
virtchnl request, kernel panic is observed. This can be reproduced using
the following script:

for i in $(seq 0 99); do
  echo 128 > /sys/class/net/ens785f0/device/sriov_numvfs
  sleep 1
  echo 0 > /sys/class/net/ens785f0/device/sriov_numvfs
  sleep 1
done

Three different call traces were seen across multiple reproductions:

Call trace 1:

[ 7057.536706] BUG: unable to handle kernel paging request at 00000f02c0000413
...
[ 7057.568476] RIP: 0010:devres_remove+0x47/0xe0
...
[ 7057.670397] Call Trace:
[ 7057.672854]  devres_destroy+0xa/0x20
[ 7057.676431]  devm_kfree+0x2d/0x40
[ 7057.679758]  ice_free_sched_node+0xf2/0x1f0 [ice]
[ 7057.684468]  ice_free_sched_node+0x3c/0x1f0 [ice]
[ 7057.689173]  ice_rm_vsi_lan_cfg+0x168/0x2e0 [ice]
[ 7057.693877]  ice_vsi_release+0x13b/0x280 [ice]
[ 7057.698325]  ice_free_vfs+0x145/0x2b0 [ice]
[ 7057.702510]  ice_sriov_configure+0x84/0x7d0 [ice]
[ 7057.707214]  ? _kstrtoull+0x3b/0x90
[ 7057.710707]  sriov_numvfs_store+0x10a/0x190
[ 7057.714891]  kernfs_fop_write+0x116/0x190
[ 7057.718903]  vfs_write+0xa5/0x1a0
[ 7057.722221]  ksys_write+0x4f/0xb0
[ 7057.725543]  do_syscall_64+0x5b/0x1a0
[ 7057.729208]  entry_SYSCALL_64_after_hwframe+0x65/0xca

Call trace 2:

[ 1387.873125] BUG: unable to handle kernel paging request at ffffd0d945000008
...
[ 1387.904813] RIP: 0010:kfree+0x5c/0x3a0
...
[ 1388.006160] Call Trace:
[ 1388.008619]  ? dev_disable_lro+0xe0/0xe0
[ 1388.012542]  pfifo_fast_destroy+0x22/0x40
[ 1388.016552]  qdisc_destroy+0x3e/0xc0
[ 1388.020129]  dev_shutdown+0x4d/0xa5
[ 1388.023622]  rollback_registered_many+0x2d1/0x630
[ 1388.028328]  rollback_registered+0x56/0x90
[ 1388.032426]  unregister_netdevice_queue+0x73/0xb0
[ 1388.037132]  unregister_netdev+0x18/0x20
[ 1388.041062]  iavf_remove+0x3a0/0x3d0 [iavf]
[ 1388.045247]  pci_device_remove+0x3b/0xc0
[ 1388.049172]  device_release_driver_internal+0x103/0x1f0
[ 1388.054398]  pci_stop_bus_device+0x69/0x90
[ 1388.058494]  pci_stop_and_remove_bus_device+0xe/0x20
[ 1388.063463]  pci_iov_remove_virtfn+0xba/0x120
[ 1388.067819]  sriov_disable+0x2f/0xe0
[ 1388.071407]  ice_free_vfs+0x276/0x2b0 [ice]
[ 1388.075596]  ice_sriov_configure+0x84/0x7d0 [ice]
[ 1388.080301]  ? _kstrtoull+0x3b/0x90
[ 1388.083790]  sriov_numvfs_store+0x10a/0x190
[ 1388.087979]  kernfs_fop_write+0x116/0x190
[ 1388.091991]  vfs_write+0xa5/0x1a0
[ 1388.095309]  ksys_write+0x4f/0xb0
[ 1388.098630]  do_syscall_64+0x5b/0x1a0
[ 1388.102294]  entry_SYSCALL_64_after_hwframe+0x65/0xca

Call trace 3:

[ 2982.326736] BUG: unable to handle kernel NULL pointer dereference at 0000000000000080
...
[ 2982.359718] RIP: 0010:proc_reg_open+0x30/0x130
...
[ 2982.461708] Call Trace:
[ 2982.464156]  ? proc_put_link+0x10/0x10
[ 2982.467910]  do_dentry_open+0x132/0x340
[ 2982.471747]  path_openat+0x53e/0x14f0
[ 2982.475415]  ? modify_irte.isra.10+0xe2/0x100
[ 2982.479770]  do_filp_open+0x93/0x100
[ 2982.483349]  ? getname_flags+0x4a/0x1e0
[ 2982.487190]  ? __check_object_size+0xa8/0x16b
[ 2982.491548]  do_sys_open+0x184/0x220
[ 2982.495129]  do_syscall_64+0x5b/0x1a0
[ 2982.498795]  entry_SYSCALL_64_after_hwframe+0x65/0xca

Fix this by adding a check to stop processing virtchnl requests when VF
teardown is in progress.

Fixes: ddf30f7ff840 ("ice: Add handler to configure SR-IOV")
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan at intel.com>
---
 drivers/net/ethernet/intel/ice/ice.h             | 1 +
 drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c | 7 +++++++
 2 files changed, 8 insertions(+)

diff --git a/drivers/net/ethernet/intel/ice/ice.h b/drivers/net/ethernet/intel/ice/ice.h
index 622b20c..7662dd2 100644
--- a/drivers/net/ethernet/intel/ice/ice.h
+++ b/drivers/net/ethernet/intel/ice/ice.h
@@ -239,6 +239,7 @@ enum ice_pf_state {
 	ICE_VFLR_EVENT_PENDING,
 	ICE_FLTR_OVERFLOW_PROMISC,
 	ICE_VF_DIS,
+	ICE_VF_DEINIT_IN_PROGRESS,
 	ICE_CFG_BUSY,
 	ICE_SERVICE_SCHED,
 	ICE_SERVICE_DIS,
diff --git a/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c b/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c
index 8fde6ea..0ed7f1e 100644
--- a/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c
+++ b/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c
@@ -500,6 +500,8 @@ void ice_free_vfs(struct ice_pf *pf)
 	struct ice_hw *hw = &pf->hw;
 	unsigned int tmp, i;
 
+	set_bit(ICE_VF_DEINIT_IN_PROGRESS, pf->state);
+
 	if (!pf->vf)
 		return;
 
@@ -568,6 +570,7 @@ void ice_free_vfs(struct ice_pf *pf)
 				i);
 
 	clear_bit(ICE_VF_DIS, pf->state);
+	clear_bit(ICE_VF_DEINIT_IN_PROGRESS, pf->state);
 	clear_bit(ICE_FLAG_SRIOV_ENA, pf->flags);
 }
 
@@ -4567,6 +4570,10 @@ void ice_vc_process_vf_msg(struct ice_pf *pf, struct ice_rq_event_info *event)
 	struct device *dev;
 	int err = 0;
 
+	/* if de-init is underway, don't process messages from VF */
+	if (test_bit(ICE_VF_DEINIT_IN_PROGRESS, pf->state))
+		return;
+
 	dev = ice_pf_to_dev(pf);
 	if (ice_validate_vf_id(pf, vf_id)) {
 		err = -EINVAL;
-- 
2.31.1