| Message ID | 1348216993-6948-3-git-send-email-wangyijing@huawei.com |
|---|---|
| State | Superseded |
| Headers | show |
diff --git a/drivers/pci/pcie/aer/aer_inject.c b/drivers/pci/pcie/aer/aer_inject.c index fdab3bb..f499f01 100644 --- a/drivers/pci/pcie/aer/aer_inject.c +++ b/drivers/pci/pcie/aer/aer_inject.c @@ -543,10 +543,11 @@ static void __exit aer_inject_exit(void) misc_deregister(&aer_inject_device); - while ((bus_ops = pci_bus_ops_pop())) { + list_for_each_entry(bus_ops, &pci_bus_ops_list, list) pci_bus_set_ops(bus_ops->bus, bus_ops->ops); + + while ((bus_ops = pci_bus_ops_pop())) kfree(bus_ops); - } spin_lock_irqsave(&inject_lock, flags); list_for_each_entry_safe(err, err_next, &einjected, list) {
When we rmmod aer_inject module, there is a small race condition window between pci_bus_ops_pop() and pci_bus_set_ops() in aer_inject_exit, eg. pci_read_aer/pci_write_aer was called between them. So keep pci_bus_ops_list until pci_bus_set_ops complete, use list_for_each_entry() instead of pci_bus_ops_pop to get bus_ops. Signed-off-by: Yijing Wang <wangyijing@huawei.com> --- drivers/pci/pcie/aer/aer_inject.c | 5 +++-- 1 files changed, 3 insertions(+), 2 deletions(-)