The adapter->vfs_lock currently protects critical sections shared between
igb_disable_sriov() and igb_msg_task(). Since igb_msg_task() — which is
invoked solely by the igb_msix_other() ISR—only proceeds when
adapter->vfs_allocated_count > 0, we can reduce the lock scope further.

By moving the assignment adapter->vfs_allocated_count = 0 to the start of the
cleanup code in igb_disable_sriov(), we can restrict the spinlock protection
solely to this assignment. This change removes kfree() calls from within the
locked section, simplifying lock management.

Once kfree() is outside the vfs_lock scope, it becomes possible to safely
convert vfs_lock to a raw_spin_lock.

Signed-off-by: Wander Lairson Costa <[email protected]>
---
 drivers/net/ethernet/intel/igb/igb_main.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/intel/igb/igb_main.c 
b/drivers/net/ethernet/intel/igb/igb_main.c
index 08578980b6518..4ca25660e876e 100644
--- a/drivers/net/ethernet/intel/igb/igb_main.c
+++ b/drivers/net/ethernet/intel/igb/igb_main.c
@@ -3708,12 +3708,12 @@ static int igb_disable_sriov(struct pci_dev *pdev, bool 
reinit)
                        msleep(500);
                }
                spin_lock_irqsave(&adapter->vfs_lock, flags);
+               adapter->vfs_allocated_count = 0;
+               spin_unlock_irqrestore(&adapter->vfs_lock, flags);
                kfree(adapter->vf_mac_list);
                adapter->vf_mac_list = NULL;
                kfree(adapter->vf_data);
                adapter->vf_data = NULL;
-               adapter->vfs_allocated_count = 0;
-               spin_unlock_irqrestore(&adapter->vfs_lock, flags);
                wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
                wrfl();
                msleep(100);
-- 
2.47.0

Reply via email to