[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20220225194614.136571-3-anthony.l.nguyen@intel.com>
Date: Fri, 25 Feb 2022 11:46:08 -0800
From: Tony Nguyen <anthony.l.nguyen@...el.com>
To: davem@...emloft.net, kuba@...nel.org
Cc: Slawomir Laba <slawomirx.laba@...el.com>, netdev@...r.kernel.org,
anthony.l.nguyen@...el.com, sassmann@...hat.com,
Phani Burra <phani.r.burra@...el.com>,
Jacob Keller <jacob.e.keller@...el.com>,
Mateusz Palczewski <mateusz.palczewski@...el.com>,
Konrad Jankowski <konrad0.jankowski@...el.com>
Subject: [PATCH net 2/8] iavf: Add waiting so the port is initialized in remove
From: Slawomir Laba <slawomirx.laba@...el.com>
There exist races when port is being configured and remove is
triggered.
unregister_netdev is not and can't be called under crit_lock
mutex since it is calling ndo_stop -> iavf_close which requires
this lock. Depending on init state the netdev could be still
unregistered so unregister_netdev never cleans up, when shortly
after that the device could become registered.
Make iavf_remove wait until port finishes initialization.
All critical state changes are atomic (under crit_lock).
Crashes that come from iavf_reset_interrupt_capability and
iavf_free_traffic_irqs should now be solved in a graceful
manner.
Fixes: 605ca7c5c6707 ("iavf: Fix kernel BUG in free_msi_irqs")
Signed-off-by: Slawomir Laba <slawomirx.laba@...el.com>
Signed-off-by: Phani Burra <phani.r.burra@...el.com>
Signed-off-by: Jacob Keller <jacob.e.keller@...el.com>
Signed-off-by: Mateusz Palczewski <mateusz.palczewski@...el.com>
Tested-by: Konrad Jankowski <konrad0.jankowski@...el.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@...el.com>
---
drivers/net/ethernet/intel/iavf/iavf_main.c | 27 ++++++++++++---------
1 file changed, 16 insertions(+), 11 deletions(-)
diff --git a/drivers/net/ethernet/intel/iavf/iavf_main.c b/drivers/net/ethernet/intel/iavf/iavf_main.c
index 84ae96e912d7..5e71b38e9154 100644
--- a/drivers/net/ethernet/intel/iavf/iavf_main.c
+++ b/drivers/net/ethernet/intel/iavf/iavf_main.c
@@ -4558,7 +4558,6 @@ static int __maybe_unused iavf_resume(struct device *dev_d)
static void iavf_remove(struct pci_dev *pdev)
{
struct iavf_adapter *adapter = iavf_pdev_to_adapter(pdev);
- enum iavf_state_t prev_state = adapter->last_state;
struct net_device *netdev = adapter->netdev;
struct iavf_fdir_fltr *fdir, *fdirtmp;
struct iavf_vlan_filter *vlf, *vlftmp;
@@ -4568,6 +4567,22 @@ static void iavf_remove(struct pci_dev *pdev)
struct iavf_hw *hw = &adapter->hw;
int err;
+ /* Wait until port initialization is complete.
+ * There are flows where register/unregister netdev may race.
+ */
+ while (1) {
+ mutex_lock(&adapter->crit_lock);
+ if (adapter->state == __IAVF_RUNNING ||
+ adapter->state == __IAVF_DOWN) {
+ mutex_unlock(&adapter->crit_lock);
+ break;
+ }
+
+ mutex_unlock(&adapter->crit_lock);
+ usleep_range(500, 1000);
+ }
+ cancel_delayed_work_sync(&adapter->watchdog_task);
+
if (adapter->netdev_registered) {
unregister_netdev(netdev);
adapter->netdev_registered = false;
@@ -4605,16 +4620,6 @@ static void iavf_remove(struct pci_dev *pdev)
iavf_free_all_rx_resources(adapter);
iavf_free_misc_irq(adapter);
- /* In case we enter iavf_remove from erroneous state, free traffic irqs
- * here, so as to not cause a kernel crash, when calling
- * iavf_reset_interrupt_capability.
- */
- if ((adapter->last_state == __IAVF_RESETTING &&
- prev_state != __IAVF_DOWN) ||
- (adapter->last_state == __IAVF_RUNNING &&
- !(netdev->flags & IFF_UP)))
- iavf_free_traffic_irqs(adapter);
-
iavf_reset_interrupt_capability(adapter);
iavf_free_q_vectors(adapter);
--
2.31.1
Powered by blists - more mailing lists