[PATCH v3 3/3] PCI/DPC: Enumerate the devices after DPC trigger event

From: Oza Pawandeep
Date: Mon Jan 08 2018 - 02:55:38 EST


Implement error_resume callback in DPC, which, after DPC trigger event
enumerates the devices beneath.

Signed-off-by: Oza Pawandeep <poza@xxxxxxxxxxxxxx>

diff --git a/drivers/pci/pcie/pcie-dpc.c b/drivers/pci/pcie/pcie-dpc.c
index 68296ec..4c6bef3 100644
--- a/drivers/pci/pcie/pcie-dpc.c
+++ b/drivers/pci/pcie/pcie-dpc.c
@@ -161,6 +161,43 @@ static void dpc_wait_link_inactive(struct dpc_dev *dpc)
dev_warn(dev, "Link state not disabled for DPC event\n");
}

+static bool dpc_wait_link_active(struct pci_dev *pdev)
+{
+ unsigned long timeout = jiffies + HZ;
+ u16 lnk_status;
+ bool ret = true;
+
+ pcie_capability_read_word(pdev, PCI_EXP_LNKSTA, &lnk_status);
+
+ while (!(lnk_status & PCI_EXP_LNKSTA_DLLLA) &&
+ !time_after(jiffies, timeout)) {
+ msleep(10);
+ pcie_capability_read_word(pdev, PCI_EXP_LNKSTA, &lnk_status);
+ }
+
+ if (!(lnk_status & PCI_EXP_LNKSTA_DLLLA)) {
+ dev_warn(&pdev->dev, "Link state not enabled after DPC event\n");
+ ret = false;
+ }
+
+ return ret;
+}
+
+/**
+ * dpc_error_resume - enumerate the devices beneath
+ * @dev: pointer to Root Port's pci_dev data structure
+ *
+ * Invoked by Port Bus driver during nonfatal recovery.
+ */
+static void dpc_error_resume(struct pci_dev *pdev)
+{
+ if (dpc_wait_link_active(pdev)) {
+ pci_lock_rescan_remove();
+ pci_rescan_bus(pdev->bus);
+ pci_unlock_rescan_remove();
+ }
+}
+
/**
* dpc_reset_link - reset link DPC routine
* @dev: pointer to Root Port's pci_dev data structure
@@ -419,6 +456,7 @@ static void dpc_remove(struct pcie_device *dev)
.service = PCIE_PORT_SERVICE_DPC,
.probe = dpc_probe,
.remove = dpc_remove,
+ .error_resume = dpc_error_resume,
.reset_link = dpc_reset_link,
};

diff --git a/drivers/pci/pcie/pcie-err.c b/drivers/pci/pcie/pcie-err.c
index 858c94c..7d1d20e 100644
--- a/drivers/pci/pcie/pcie-err.c
+++ b/drivers/pci/pcie/pcie-err.c
@@ -238,7 +238,8 @@ pci_ers_result_t pci_reset_link(struct pci_dev *dev, int severity)
pci_ers_result_t pci_broadcast_error_message(struct pci_dev *dev,
enum pci_channel_state state,
char *error_mesg,
- int (*cb)(struct pci_dev *, void *))
+ int (*cb)(struct pci_dev *, void *),
+ int severity)
{
struct pci_err_broadcast_data result_data;

@@ -250,6 +251,15 @@ pci_ers_result_t pci_broadcast_error_message(struct pci_dev *dev,
result_data.result = PCI_ERS_RESULT_RECOVERED;

if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE) {
+ /* If DPC is triggered, call resume error hanlder
+ * because, at this point we can safely assume that
+ * link recovery has happened.
+ */
+ if ((severity == PCI_ERR_DPC_FATAL) &&
+ (cb == pci_report_resume)) {
+ cb(dev, NULL);
+ return PCI_ERS_RESULT_RECOVERED;
+ }
/*
* If the error is reported by a bridge, we think this error
* is related to the downstream link of the bridge, so we
@@ -299,7 +309,8 @@ void pci_do_recovery(struct pci_dev *dev, int severity)
status = pci_broadcast_error_message(dev,
state,
"error_detected",
- pci_report_error_detected);
+ pci_report_error_detected,
+ severity);

if ((severity == PCI_ERR_AER_FATAL) ||
(severity == PCI_ERR_DPC_FATAL)) {
@@ -308,11 +319,15 @@ void pci_do_recovery(struct pci_dev *dev, int severity)
goto failed;
}

+ if (severity == PCI_ERR_DPC_FATAL)
+ goto resume;
+
if (status == PCI_ERS_RESULT_CAN_RECOVER)
status = pci_broadcast_error_message(dev,
state,
"mmio_enabled",
- pci_report_mmio_enabled);
+ pci_report_mmio_enabled,
+ severity);

if (status == PCI_ERS_RESULT_NEED_RESET) {
/*
@@ -323,16 +338,19 @@ void pci_do_recovery(struct pci_dev *dev, int severity)
status = pci_broadcast_error_message(dev,
state,
"slot_reset",
- pci_report_slot_reset);
+ pci_report_slot_reset,
+ severity);
}

if (status != PCI_ERS_RESULT_RECOVERED)
goto failed;

+resume:
pci_broadcast_error_message(dev,
state,
"resume",
- pci_report_resume);
+ pci_report_resume,
+ severity);

dev_info(&dev->dev, "Device recovery successful\n");
mutex_unlock(&pci_err_recovery_lock);
diff --git a/include/linux/pci.h b/include/linux/pci.h
index 123ee15..46e2526 100644
--- a/include/linux/pci.h
+++ b/include/linux/pci.h
@@ -2010,7 +2010,8 @@ static inline resource_size_t pci_iov_resource_size(struct pci_dev *dev, int res
pci_ers_result_t pci_broadcast_error_message(struct pci_dev *dev,
enum pci_channel_state state,
char *error_mesg,
- int (*cb)(struct pci_dev *, void *));
+ int (*cb)(struct pci_dev *, void *),
+ int severity);
int pci_report_mmio_enabled(struct pci_dev *dev, void *data);
int pci_report_slot_reset(struct pci_dev *dev, void *data);
int pci_report_resume(struct pci_dev *dev, void *data);
--
Qualcomm Datacenter Technologies, Inc. as an affiliate of Qualcomm Technologies, Inc.,
a Qualcomm Technologies, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project.