net: hns3: fix to stop multiple HNS reset due to the AER changes
authorShiju Jose <shiju.jose@huawei.com>
Sun, 10 Mar 2019 06:47:51 +0000 (14:47 +0800)
committerDavid S. Miller <davem@davemloft.net>
Sun, 10 Mar 2019 06:58:47 +0000 (22:58 -0800)
The commit bfcb79fca19d
("PCI/ERR: Run error recovery callbacks for all affected devices")
affected the non-fatal error recovery logic for the HNS and RDMA devices.
This is because each HNS PF under PCIe bus receive callbacks
from the AER driver when an error is reported for one of the PF.
This causes unwanted PF resets because
the HNS decides which PF to reset based on the reset type set.
The HNS error handling code sets the reset type based on the hw error
type detected.

This patch provides fix for the above issue for the recovery of
the hw errors in the HNS and RDMA devices.

This patch needs backporting to the kernel v5.0+

Fixes: 332fbf576579 ("net: hns3: add handling of hw ras errors using new set of commands")
Reported-by: Xiaofei Tan <tanxiaofei@huawei.com>
Signed-off-by: Shiju Jose <shiju.jose@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/hisilicon/hns3/hnae3.h
drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_err.c

index 66d7a8b80e763ff07f1da0bdca5a42bda09aa5ab..38b430f11fc14f6d474faa4df7a28f2ef19fc287 100644 (file)
@@ -194,6 +194,7 @@ struct hnae3_ae_dev {
        const struct hnae3_ae_ops *ops;
        struct list_head node;
        u32 flag;
+       u8 override_pci_need_reset; /* fix to stop multiple reset happening */
        enum hnae3_dev_type dev_type;
        enum hnae3_reset_type reset_type;
        void *priv;
index 0d1ae15a5927d707b6de3938472db180b0ff4ee6..1c1f17ec6be2dfa11cd1208b0b04273f07855c10 100644 (file)
@@ -1850,7 +1850,9 @@ static pci_ers_result_t hns3_slot_reset(struct pci_dev *pdev)
 
        /* request the reset */
        if (ae_dev->ops->reset_event) {
-               ae_dev->ops->reset_event(pdev, NULL);
+               if (!ae_dev->override_pci_need_reset)
+                       ae_dev->ops->reset_event(pdev, NULL);
+
                return PCI_ERS_RESULT_RECOVERED;
        }
 
index 1feceff1477cea9700ede728e473dca6765483d4..1f52d11f77b55c58d837918f384e172158caaca0 100644 (file)
@@ -1317,8 +1317,10 @@ pci_ers_result_t hclge_handle_hw_ras_error(struct hnae3_ae_dev *ae_dev)
                hclge_handle_all_ras_errors(hdev);
        } else {
                if (test_bit(HCLGE_STATE_RST_HANDLING, &hdev->state) ||
-                   hdev->pdev->revision < 0x21)
+                   hdev->pdev->revision < 0x21) {
+                       ae_dev->override_pci_need_reset = 1;
                        return PCI_ERS_RESULT_RECOVERED;
+               }
        }
 
        if (status & HCLGE_RAS_REG_ROCEE_ERR_MASK) {
@@ -1327,8 +1329,11 @@ pci_ers_result_t hclge_handle_hw_ras_error(struct hnae3_ae_dev *ae_dev)
        }
 
        if (status & HCLGE_RAS_REG_NFE_MASK ||
-           status & HCLGE_RAS_REG_ROCEE_ERR_MASK)
+           status & HCLGE_RAS_REG_ROCEE_ERR_MASK) {
+               ae_dev->override_pci_need_reset = 0;
                return PCI_ERS_RESULT_NEED_RESET;
+       }
+       ae_dev->override_pci_need_reset = 1;
 
        return PCI_ERS_RESULT_RECOVERED;
 }