drivers/edac: updated PCI monitoring

Moving PCI to a per-instance device model

This should include the correct sysfs setup as well. Please review.

Signed-off-by: Dave Jiang <[email protected]>
Signed-off-by: Douglas Thompson <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Author: Dave Jiang

drivers/edac/Makefile

@@ -11,9 +11,12 @@ obj-$(CONFIG_EDAC)			:= edac_stub.o
 obj-$(CONFIG_EDAC_MM_EDAC)		+= edac_core.o
 
 edac_core-objs	:= edac_mc.o edac_device.o edac_mc_sysfs.o edac_pci_sysfs.o
-
 edac_core-objs	+= edac_module.o edac_device_sysfs.o
 
+ifdef CONFIG_PCI
+edac_core-objs	+= edac_pci.o edac_pci_sysfs.o
+endif
+
 obj-$(CONFIG_EDAC_AMD76X)		+= amd76x_edac.o
 obj-$(CONFIG_EDAC_I5000)		+= i5000_edac.o
 obj-$(CONFIG_EDAC_E7XXX)		+= e7xxx_edac.o

drivers/edac/e752x_edac.c

@@ -30,6 +30,8 @@
 
 static int force_function_unhide;
 
+static struct edac_pci_ctl_info *e752x_pci;
+
 #define e752x_printk(level, fmt, arg...) \
 	edac_printk(level, "e752x", fmt, ##arg)
 
@@ -1040,6 +1042,17 @@ static int e752x_probe1(struct pci_dev *pdev, int dev_idx)
 	e752x_init_error_reporting_regs(pvt);
 	e752x_get_error_info(mci, &discard); /* clear other MCH errors */
 
+	/* allocating generic PCI control info */
+	e752x_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
+	if (!e752x_pci) {
+		printk(KERN_WARNING
+			"%s(): Unable to create PCI control\n",
+			__func__);
+		printk(KERN_WARNING
+			"%s(): PCI error report via EDAC not setup\n",
+			__func__);
+	}
+
 	/* get this far and it's successful */
 	debugf3("%s(): success\n", __func__);
 	return 0;
@@ -1073,6 +1086,9 @@ static void __devexit e752x_remove_one(struct pci_dev *pdev)
 
 	debugf0("%s()\n", __func__);
 
+	if (e752x_pci)
+		edac_pci_release_generic_ctl(e752x_pci);
+
 	if ((mci = edac_mc_del_mc(&pdev->dev)) == NULL)
 		return;
 

drivers/edac/edac_core.h

@@ -60,6 +60,10 @@
 #define edac_device_printk(ctl, level, fmt, arg...) \
 	printk(level "EDAC DEVICE%d: " fmt, ctl->dev_idx, ##arg)
 
+/* edac_pci printk */
+#define edac_pci_printk(ctl, level, fmt, arg...) \
+	printk(level "EDAC PCI%d: " fmt, ctl->pci_idx, ##arg)
+
 /* prefixes for edac_printk() and edac_mc_printk() */
 #define EDAC_MC "MC"
 #define EDAC_PCI "PCI"
@@ -200,6 +204,13 @@ enum scrub_type {
 
 /* FIXME - should have notify capabilities: NMI, LOG, PROC, etc */
 
+/* EDAC internal operation states */
+#define	OP_ALLOC		0x100
+#define OP_RUNNING_POLL		0x201
+#define OP_RUNNING_INTERRUPT	0x202
+#define OP_RUNNING_POLL_INTR	0x203
+#define OP_OFFLINE		0x300
+
 extern char * edac_align_ptr(void *ptr, unsigned size);
 
 /*
@@ -520,12 +531,6 @@ struct edac_device_ctl_info {
 
 	/* the internal state of this controller instance */
 	int op_state;
-#define	OP_ALLOC		0x100
-#define OP_RUNNING_POLL		0x201
-#define OP_RUNNING_INTERRUPT	0x202
-#define OP_RUNNING_POLL_INTR	0x203
-#define OP_OFFLINE		0x300
-
 	/* work struct for this instance */
 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,20))
 	struct delayed_work work;
@@ -626,6 +631,84 @@ extern void edac_device_free_ctl_info( struct edac_device_ctl_info *ctl_info);
 
 #ifdef CONFIG_PCI
 
+struct edac_pci_counter {
+	atomic_t	pe_count;
+	atomic_t	npe_count;
+};
+
+/*
+ * Abstract edac_pci control info structure
+ *
+ */
+struct edac_pci_ctl_info {
+	/* for global list of edac_pci_ctl_info structs */
+	struct list_head link;
+
+	int pci_idx;
+
+	/* Per instance controls for this edac_device */
+	int check_parity_error;	/* boolean for checking parity errs */
+	int log_parity_error;	/* boolean for logging parity errs */
+	int panic_on_pe;	/* boolean for panic'ing on a PE */
+	unsigned poll_msec;	/* number of milliseconds to poll interval */
+	unsigned long delay;	/* number of jiffies for poll_msec */
+
+	struct sysdev_class *edac_class;	/* pointer to class */
+
+	/* the internal state of this controller instance */
+	int op_state;
+	/* work struct for this instance */
+#if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,20))
+	struct delayed_work work;
+#else
+	struct work_struct work;
+#endif
+
+	/* pointer to edac polling checking routine:
+	 *	If NOT NULL: points to polling check routine
+	 *	If NULL: Then assumes INTERRUPT operation, where
+	 *		MC driver will receive events
+	 */
+	void (*edac_check) (struct edac_pci_ctl_info * edac_dev);
+
+	struct device *dev;	/* pointer to device structure */
+
+	const char *mod_name;	/* module name */
+	const char *ctl_name;	/* edac controller  name */
+	const char *dev_name;	/* pci/platform/etc... name */
+
+	void *pvt_info;		/* pointer to 'private driver' info */
+
+	unsigned long start_time;/* edac_pci load start time (jiffies)*/
+
+	/* these are for safe removal of devices from global list while
+	 * NMI handlers may be traversing list
+	 */
+	struct rcu_head rcu;
+	struct completion complete;
+
+	/* sysfs top name under 'edac' directory
+	 * and instance name:
+	 *	cpu/cpu0/...
+	 *	cpu/cpu1/...
+	 *	cpu/cpu2/...
+	 *	...
+	 */
+	char name[EDAC_DEVICE_NAME_LEN + 1];
+
+	/* Event counters for the this whole EDAC Device */
+	struct edac_pci_counter counters;
+
+	/* edac sysfs device control for the 'name'
+	 * device this structure controls
+	 */
+	struct kobject kobj;
+	struct completion kobj_complete;
+};
+
+#define to_edac_pci_ctl_work(w) \
+		container_of(w, struct edac_pci_ctl_info,work)
+
 /* write all or some bits in a byte-register*/
 static inline void pci_write_bits8(struct pci_dev *pdev, int offset, u8 value,
 		u8 mask)
@@ -726,5 +809,30 @@ extern void edac_device_handle_ue(struct edac_device_ctl_info *edac_dev,
 extern void edac_device_handle_ce(struct edac_device_ctl_info *edac_dev,
 		int inst_nr, int block_nr, const char *msg);
 
+/*
+ * edac_pci APIs
+ */
+extern struct edac_pci_ctl_info *
+edac_pci_alloc_ctl_info(unsigned int sz_pvt, const char *edac_pci_name);
+
+extern void edac_pci_free_ctl_info(struct edac_pci_ctl_info *pci);
+
+extern void
+edac_pci_reset_delay_period(struct edac_pci_ctl_info *pci, unsigned long value);
+
+extern int edac_pci_add_device(struct edac_pci_ctl_info *pci, int edac_idx);
+extern struct edac_pci_ctl_info * edac_pci_del_device(struct device *dev);
+
+extern struct edac_pci_ctl_info *
+edac_pci_create_generic_ctl(struct device *dev, const char *mod_name);
+
+extern void edac_pci_release_generic_ctl(struct edac_pci_ctl_info *pci);
+extern int edac_pci_create_sysfs(struct edac_pci_ctl_info *pci);
+extern void edac_pci_remove_sysfs(struct edac_pci_ctl_info *pci);
+
+/*
+ * edac misc APIs
+ */
+extern char * edac_op_state_toString(int op_state);
 
 #endif				/* _EDAC_CORE_H_ */

drivers/edac/edac_device.c

@@ -418,27 +418,6 @@ void edac_device_reset_delay_period(
 	unlock_device_list();
 }
 
-/*
- * edac_op_state_toString(edac_dev)
- */
-static char *edac_op_state_toString(struct edac_device_ctl_info *edac_dev)
-{
-	int opstate = edac_dev->op_state;
-
-	if (opstate == OP_RUNNING_POLL)
-		return "POLLED";
-	else if (opstate == OP_RUNNING_INTERRUPT)
-		return "INTERRUPT";
-	else if (opstate == OP_RUNNING_POLL_INTR)
-		return "POLL-INTR";
-	else if (opstate == OP_ALLOC)
-		return "ALLOC";
-	else if (opstate == OP_OFFLINE)
-		return "OFFLINE";
-
-	return "UNKNOWN";
-}
-
 /**
  * edac_device_add_device: Insert the 'edac_dev' structure into the
  * edac_device global list and create sysfs entries associated with
@@ -496,7 +475,7 @@ int edac_device_add_device(struct edac_device_ctl_info *edac_dev, int edac_idx)
 		edac_dev->mod_name,
 		edac_dev->ctl_name,
 		dev_name(edac_dev),
-		edac_op_state_toString(edac_dev)
+		edac_op_state_toString(edac_dev->op_state)
 		);
 
 	unlock_device_list();

drivers/edac/edac_module.c

@@ -34,6 +34,25 @@ static struct sysdev_class edac_class = {
 };
 static int edac_class_valid = 0;
 
+/*
+ * edac_op_state_toString()
+ */
+char * edac_op_state_toString(int opstate)
+{
+	if (opstate == OP_RUNNING_POLL)
+		return "POLLED";
+	else if (opstate == OP_RUNNING_INTERRUPT)
+		return "INTERRUPT";
+	else if (opstate == OP_RUNNING_POLL_INTR)
+		return "POLL-INTR";
+	else if (opstate == OP_ALLOC)
+		return "ALLOC";
+	else if (opstate == OP_OFFLINE)
+		return "OFFLINE";
+
+	return "UNKNOWN";
+}
+
 /*
  * edac_get_edac_class()
  *
@@ -153,26 +172,16 @@ static int __init edac_init(void)
 		goto error_sysfs;
 	}
 
-	/* Create the PCI parity sysfs entries */
-	if (edac_sysfs_pci_setup()) {
-		edac_printk(KERN_ERR, EDAC_MC,
-			"PCI: Error initializing sysfs code\n");
-		err = -ENODEV;
-		goto error_mem;
-	}
-
 	/* Setup/Initialize the edac_device system */
 	err = edac_workqueue_setup();
 	if (err) {
 		edac_printk(KERN_ERR, EDAC_MC, "init WorkQueue failure\n");
-		goto error_pci;
+		goto error_mem;
 	}
 
 	return 0;
 
 	/* Error teardown stack */
-error_pci:
-	edac_sysfs_pci_teardown();
 error_mem:
 	edac_sysfs_memctrl_teardown();
 error_sysfs:
@@ -192,7 +201,6 @@ static void __exit edac_exit(void)
 	/* tear down the various subsystems*/
 	edac_workqueue_teardown();
 	edac_sysfs_memctrl_teardown();
-	edac_sysfs_pci_teardown();
 	edac_unregister_sysfs_edac_name();
 }