[PATCH 7/9] PCIe, AER, add aer_recover_queue

[Huang Ying]

In addition to native PCIe AER, now APEI (ACPI Platform Error
Interface) GHES (Generic Hardware Error Source) can be used to report
PCIe AER errors too.  To add support to APEI GHES PCIe AER recovery,
aer_recover_queue is added to export the recovery function in native
PCIe AER driver.

Recoverable PCIe AER errors are reported via NMI in APEI GHES.  Then
APEI GHES uses irq_work to delay the error processing into an IRQ
handler.  But PCIe AER recovery can be very time-consuming, so
aer_recover_queue, which can be used in IRQ handler, delays the real
recovery action into the process context, that is, work queue.

Signed-off-by: Huang Ying <ying.huang@xxxxxxxxx>
CC: Jesse Barnes <jbarnes@xxxxxxxxxxxxxxxx>
CC: Zhang Yanmin <yanmin.zhang@xxxxxxxxx>
---
 drivers/pci/pcie/aer/aerdrv_core.c     |   76 +++++++++++++++++++++++++++++----
 drivers/pci/pcie/aer/aerdrv_errprint.c |    3 -
 include/linux/aer.h                    |    3 +
 3 files changed, 74 insertions(+), 8 deletions(-)

--- a/drivers/pci/pcie/aer/aerdrv_core.c
+++ b/drivers/pci/pcie/aer/aerdrv_core.c
@@ -24,6 +24,7 @@
 #include <linux/suspend.h>
 #include <linux/delay.h>
 #include <linux/slab.h>
+#include <linux/kfifo.h>
 #include "aerdrv.h"
 
 static int forceload;
@@ -445,8 +446,7 @@ static struct pcie_port_service_driver *
 	return drv;
 }
 
-static pci_ers_result_t reset_link(struct pcie_device *aerdev,
-		struct pci_dev *dev)
+static pci_ers_result_t reset_link(struct pci_dev *dev)
 {
 	struct pci_dev *udev;
 	pci_ers_result_t status;
@@ -486,7 +486,6 @@ static pci_ers_result_t reset_link(struc
 
 /**
  * do_recovery - handle nonfatal/fatal error recovery process
- * @aerdev: pointer to a pcie_device data structure of root port
  * @dev: pointer to a pci_dev data structure of agent detecting an error
  * @severity: error severity type
  *
@@ -494,8 +493,7 @@ static pci_ers_result_t reset_link(struc
  * error detected message to all downstream drivers within a hierarchy in
  * question and return the returned code.
  */
-static void do_recovery(struct pcie_device *aerdev, struct pci_dev *dev,
-		int severity)
+static void do_recovery(struct pci_dev *dev, int severity)
 {
 	pci_ers_result_t status, result = PCI_ERS_RESULT_RECOVERED;
 	enum pci_channel_state state;
@@ -511,7 +509,7 @@ static void do_recovery(struct pcie_devi
 			report_error_detected);
 
 	if (severity == AER_FATAL) {
-		result = reset_link(aerdev, dev);
+		result = reset_link(dev);
 		if (result != PCI_ERS_RESULT_RECOVERED)
 			goto failed;
 	}
@@ -576,9 +574,73 @@ static void handle_error_source(struct p
 			pci_write_config_dword(dev, pos + PCI_ERR_COR_STATUS,
 					info->status);
 	} else
-		do_recovery(aerdev, dev, info->severity);
+		do_recovery(dev, info->severity);
 }
 
+#ifdef CONFIG_ACPI_APEI_PCIEAER
+static void aer_recover_work_func(struct work_struct *work);
+
+#define AER_RECOVER_RING_ORDER		4
+#define AER_RECOVER_RING_SIZE		(1 << AER_RECOVER_RING_ORDER)
+
+struct aer_recover_entry
+{
+	u8	bus;
+	u8	devfn;
+	u16	domain;
+	int	severity;
+};
+
+static DEFINE_KFIFO(aer_recover_ring, struct aer_recover_entry,
+		    AER_RECOVER_RING_SIZE);
+/*
+ * Mutual exclusion for writers of aer_recover_ring, reader side don't
+ * need lock, because there is only one reader and lock is not needed
+ * between reader and writer.
+ */
+static DEFINE_SPINLOCK(aer_recover_ring_lock);
+static DECLARE_WORK(aer_recover_work, aer_recover_work_func);
+
+void aer_recover_queue(int domain, unsigned int bus, unsigned int devfn,
+		       int severity)
+{
+	unsigned long flags;
+	struct aer_recover_entry entry = {
+		.bus		= bus,
+		.devfn		= devfn,
+		.domain		= domain,
+		.severity	= severity,
+	};
+
+	spin_lock_irqsave(&aer_recover_ring_lock, flags);
+	if (kfifo_put(&aer_recover_ring, &entry))
+		schedule_work(&aer_recover_work);
+	else
+		pr_err("AER recover: Buffer overflow when recovering AER for %04x:%02x:%02x:%x\n",
+		       domain, bus, PCI_SLOT(devfn), PCI_FUNC(devfn));
+	spin_unlock_irqrestore(&aer_recover_ring_lock, flags);
+}
+EXPORT_SYMBOL_GPL(aer_recover_queue);
+
+static void aer_recover_work_func(struct work_struct *work)
+{
+	struct aer_recover_entry entry;
+	struct pci_dev *pdev;
+
+	while (kfifo_get(&aer_recover_ring, &entry)) {
+		pdev = pci_get_domain_bus_and_slot(entry.domain, entry.bus,
+						   entry.devfn);
+		if (!pdev) {
+			pr_err("AER recover: Can not find pci_dev for %04x:%02x:%02x:%x\n",
+			       entry.domain, entry.bus,
+			       PCI_SLOT(entry.devfn), PCI_FUNC(entry.devfn));
+			continue;
+		}
+		do_recovery(pdev, entry.severity);
+	}
+}
+#endif
+
 /**
  * get_device_error_info - read error status from dev and store it to info
  * @dev: pointer to the device expected to have a error record
--- a/drivers/pci/pcie/aer/aerdrv_errprint.c
+++ b/drivers/pci/pcie/aer/aerdrv_errprint.c
@@ -204,7 +204,7 @@ void aer_print_port_info(struct pci_dev
 }
 
 #ifdef CONFIG_ACPI_APEI_PCIEAER
-static int cper_severity_to_aer(int cper_severity)
+int cper_severity_to_aer(int cper_severity)
 {
 	switch (cper_severity) {
 	case CPER_SEV_RECOVERABLE:
@@ -215,6 +215,7 @@ static int cper_severity_to_aer(int cper
 		return AER_CORRECTABLE;
 	}
 }
+EXPORT_SYMBOL_GPL(cper_severity_to_aer);
 
 void cper_print_aer(const char *prefix, int cper_severity,
 		    struct aer_capability_regs *aer)
--- a/include/linux/aer.h
+++ b/include/linux/aer.h
@@ -51,5 +51,8 @@ static inline int pci_cleanup_aer_uncorr
 
 extern void cper_print_aer(const char *prefix, int cper_severity,
 			   struct aer_capability_regs *aer);
+extern int cper_severity_to_aer(int cper_severity);
+extern void aer_recover_queue(int domain, unsigned int bus, unsigned int devfn,
+			      int severity);
 #endif //_AER_H_
 
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/