edac: e752x fix too loud on nonmemory errors

This module harvests more than just memory errors, it also harvests various bus and dma errors that the Chipset detects. Previously, it would report all such errors, which would cause output to be TOO loud. This patches therefore adds a parameter which is used to turn off NON-MEMORY error reports by default. Or the reporting can be enabled via the parameter Also did code style cleanup: less than 80 characters per line rule Signed-off-by: Doug Thompson <dougthompson@xmission.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

edac: e752x fix too loud on nonmemory errors
This module harvests more than just memory errors, it also harvests various bus and dma errors that the Chipset detects. Previously, it would report all such errors, which would cause output to be TOO loud. This patches therefore adds a parameter which is used to turn off NON-MEMORY error reports by default. Or the reporting can be enabled via the parameter Also did code style cleanup: less than 80 characters per line rule Signed-off-by: Doug Thompson <dougthompson@xmission.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Doug Thompson · Linus Torvalds
1 parent 124682c785
Showing 1 changed file with 42 additions and 17 deletions Side-by-side Diff
drivers/edac/e752x_edac.c
@@ -28,6 +28,7 @@
 #define E752X_REVISION	" Ver: 2.0.2 " __DATE__
 #define EDAC_MOD_STR	"e752x_edac"
  
+static int report_non_memory_errors;
 static int force_function_unhide;
 static int sysbus_parity = -1;
  
@@ -117,7 +118,7 @@
 #define E752X_BUF_FERR		0x70	/* Memory buffer first error reg (8b) */
 #define E752X_BUF_NERR		0x72	/* Memory buffer next error reg (8b) */
 #define E752X_BUF_ERRMASK	0x74	/* Memory buffer error mask reg (8b) */
-#define E752X_BUF_SMICMD	0x7A	/* Memory buffer SMI command reg (8b) */
+#define E752X_BUF_SMICMD	0x7A	/* Memory buffer SMI cmd reg (8b) */
 #define E752X_DRAM_FERR		0x80	/* DRAM first error register (16b) */
 #define E752X_DRAM_NERR		0x82	/* DRAM next error register (16b) */
 #define E752X_DRAM_ERRMASK	0x84	/* DRAM error mask register (8b) */
@@ -127,7 +128,7 @@
 					/*     error address register (32b) */
 					/*
 					 * 31    Reserved
-					 * 30:2  CE address (64 byte block 34:6)
+					 * 30:2  CE address (64 byte block 34:6
 					 * 1     Reserved
 					 * 0     HiLoCS
 					 */
  
@@ -147,11 +148,11 @@
 					 * 1     Reserved
 					 * 0     HiLoCS
 					 */
-#define E752X_DRAM_SCRB_ADD	0xA8	/* DRAM first uncorrectable scrub memory */
+#define E752X_DRAM_SCRB_ADD	0xA8	/* DRAM 1st uncorrectable scrub mem */
 					/*     error address register (32b) */
 					/*
 					 * 31    Reserved
-					 * 30:2  CE address (64 byte block 34:6)
+					 * 30:2  CE address (64 byte block 34:6
 					 * 1     Reserved
 					 * 0     HiLoCS
 					 */
  
@@ -394,9 +395,12 @@
 	struct e752x_pvt *pvt = (struct e752x_pvt *)mci->pvt_info;
  
 	error_1b = retry_add;
-	page = error_1b >> (PAGE_SHIFT - 4);	/* convert the addr to 4k page */
-	row = pvt->mc_symmetric ? ((page >> 1) & 3) :	/* chip select are bits 14 & 13 */
+	page = error_1b >> (PAGE_SHIFT - 4);  /* convert the addr to 4k page */
+
+	/* chip select are bits 14 & 13 */
+	row = pvt->mc_symmetric ? ((page >> 1) & 3) :
 		edac_mc_find_csrow_by_page(mci, page);
+
 	e752x_mc_printk(mci, KERN_WARNING,
 			"CE page 0x%lx, row %d : Memory read retry\n",
 			(long unsigned int)page, row);
  
@@ -422,12 +426,21 @@
 }
  
 static char *global_message[11] = {
-	"PCI Express C1", "PCI Express C", "PCI Express B1",
-	"PCI Express B", "PCI Express A1", "PCI Express A",
-	"DMA Controler", "HUB or NS Interface", "System Bus",
-	"DRAM Controler", "Internal Buffer"
+	"PCI Express C1",
+	"PCI Express C",
+	"PCI Express B1",
+	"PCI Express B",
+	"PCI Express A1",
+	"PCI Express A",
+	"DMA Controller",
+	"HUB or NS Interface",
+	"System Bus",
+	"DRAM Controller",  /* 9th entry */
+	"Internal Buffer"
 };
  
+#define DRAM_ENTRY	9
+
 static char *fatal_message[2] = { "Non-Fatal ", "Fatal " };
  
 static void do_global_error(int fatal, u32 errors)
@@ -435,9 +448,16 @@
 	int i;
  
 	for (i = 0; i < 11; i++) {
-		if (errors & (1 << i))
-			e752x_printk(KERN_WARNING, "%sError %s\n",
-				fatal_message[fatal], global_message[i]);
+		if (errors & (1 << i)) {
+			/* If the error is from DRAM Controller OR
+			 * we are to report ALL errors, then
+			 * report the error
+			 */
+			if ((i == DRAM_ENTRY) || report_non_memory_errors)
+				e752x_printk(KERN_WARNING, "%sError %s\n",
+					fatal_message[fatal],
+					global_message[i]);
+		}
 	}
 }
  
@@ -1021,7 +1041,7 @@
 	struct pci_dev *dev;
  
 	pvt->bridge_ck = pci_get_device(PCI_VENDOR_ID_INTEL,
-					pvt->dev_info->err_dev, pvt->bridge_ck);
+				pvt->dev_info->err_dev, pvt->bridge_ck);
  
 	if (pvt->bridge_ck == NULL)
 		pvt->bridge_ck = pci_scan_single_device(pdev->bus,
@@ -1034,8 +1054,9 @@
 		return 1;
 	}
  
-	dev = pci_get_device(PCI_VENDOR_ID_INTEL, e752x_devs[dev_idx].ctl_dev,
-			NULL);
+	dev = pci_get_device(PCI_VENDOR_ID_INTEL,
+				e752x_devs[dev_idx].ctl_dev,
+				NULL);
  
 	if (dev == NULL)
 		goto fail;
@@ -1316,7 +1337,8 @@
  
 module_param(force_function_unhide, int, 0444);
 MODULE_PARM_DESC(force_function_unhide, "if BIOS sets Dev0:Fun1 up as hidden:"
-		 " 1=force unhide and hope BIOS doesn't fight driver for Dev0:Fun1 access");
+		 " 1=force unhide and hope BIOS doesn't fight driver for "
+		"Dev0:Fun1 access");
  
 module_param(edac_op_state, int, 0444);
 MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
@@ -1324,4 +1346,7 @@
 module_param(sysbus_parity, int, 0444);
 MODULE_PARM_DESC(sysbus_parity, "0=disable system bus parity checking,"
 		" 1=enable system bus parity checking, default=auto-detect");
+module_param(report_non_memory_errors, int, 0644);
+MODULE_PARM_DESC(report_non_memory_errors, "0=disable non-memory error "
+		"reporting, 1=enable non-memory error reporting");
...	...	@@ -28,6 +28,7 @@
28	28	#define E752X_REVISION " Ver: 2.0.2 " __DATE__
29	29	#define EDAC_MOD_STR "e752x_edac"
30	30
	31	+static int report_non_memory_errors;
31	32	static int force_function_unhide;
32	33	static int sysbus_parity = -1;
33	34
...	...	@@ -117,7 +118,7 @@
117	118	#define E752X_BUF_FERR 0x70 /* Memory buffer first error reg (8b) */
118	119	#define E752X_BUF_NERR 0x72 /* Memory buffer next error reg (8b) */
119	120	#define E752X_BUF_ERRMASK 0x74 /* Memory buffer error mask reg (8b) */
120		-#define E752X_BUF_SMICMD 0x7A /* Memory buffer SMI command reg (8b) */
	121	+#define E752X_BUF_SMICMD 0x7A /* Memory buffer SMI cmd reg (8b) */
121	122	#define E752X_DRAM_FERR 0x80 /* DRAM first error register (16b) */
122	123	#define E752X_DRAM_NERR 0x82 /* DRAM next error register (16b) */
123	124	#define E752X_DRAM_ERRMASK 0x84 /* DRAM error mask register (8b) */
...	...	@@ -127,7 +128,7 @@
127	128	/* error address register (32b) */
128	129	/*
129	130	* 31 Reserved
130		- * 30:2 CE address (64 byte block 34:6)
	131	+ * 30:2 CE address (64 byte block 34:6
131	132	* 1 Reserved
132	133	* 0 HiLoCS
133	134	*/
134	135
...	...	@@ -147,11 +148,11 @@
147	148	* 1 Reserved
148	149	* 0 HiLoCS
149	150	*/
150		-#define E752X_DRAM_SCRB_ADD 0xA8 /* DRAM first uncorrectable scrub memory */
	151	+#define E752X_DRAM_SCRB_ADD 0xA8 /* DRAM 1st uncorrectable scrub mem */
151	152	/* error address register (32b) */
152	153	/*
153	154	* 31 Reserved
154		- * 30:2 CE address (64 byte block 34:6)
	155	+ * 30:2 CE address (64 byte block 34:6
155	156	* 1 Reserved
156	157	* 0 HiLoCS
157	158	*/
158	159
...	...	@@ -394,9 +395,12 @@
394	395	struct e752x_pvt pvt = (struct e752x_pvt )mci->pvt_info;
395	396
396	397	error_1b = retry_add;
397		- page = error_1b >> (PAGE_SHIFT - 4); /* convert the addr to 4k page */
398		- row = pvt->mc_symmetric ? ((page >> 1) & 3) : /* chip select are bits 14 & 13 */
	398	+ page = error_1b >> (PAGE_SHIFT - 4); /* convert the addr to 4k page */
	399	+
	400	+ /* chip select are bits 14 & 13 */
	401	+ row = pvt->mc_symmetric ? ((page >> 1) & 3) :
399	402	edac_mc_find_csrow_by_page(mci, page);
	403	+
400	404	e752x_mc_printk(mci, KERN_WARNING,
401	405	"CE page 0x%lx, row %d : Memory read retry\n",
402	406	(long unsigned int)page, row);
403	407
...	...	@@ -422,12 +426,21 @@
422	426	}
423	427
424	428	static char *global_message[11] = {
425		- "PCI Express C1", "PCI Express C", "PCI Express B1",
426		- "PCI Express B", "PCI Express A1", "PCI Express A",
427		- "DMA Controler", "HUB or NS Interface", "System Bus",
428		- "DRAM Controler", "Internal Buffer"
	429	+ "PCI Express C1",
	430	+ "PCI Express C",
	431	+ "PCI Express B1",
	432	+ "PCI Express B",
	433	+ "PCI Express A1",
	434	+ "PCI Express A",
	435	+ "DMA Controller",
	436	+ "HUB or NS Interface",
	437	+ "System Bus",
	438	+ "DRAM Controller", /* 9th entry */
	439	+ "Internal Buffer"
429	440	};
430	441
	442	+#define DRAM_ENTRY 9
	443	+
431	444	static char *fatal_message[2] = { "Non-Fatal ", "Fatal " };
432	445
433	446	static void do_global_error(int fatal, u32 errors)
...	...	@@ -435,9 +448,16 @@
435	448	int i;
436	449
437	450	for (i = 0; i < 11; i++) {
438		- if (errors & (1 << i))
439		- e752x_printk(KERN_WARNING, "%sError %s\n",
440		- fatal_message[fatal], global_message[i]);
	451	+ if (errors & (1 << i)) {
	452	+ /* If the error is from DRAM Controller OR
	453	+ * we are to report ALL errors, then
	454	+ * report the error
	455	+ */
	456	+ if ((i == DRAM_ENTRY) \|\| report_non_memory_errors)
	457	+ e752x_printk(KERN_WARNING, "%sError %s\n",
	458	+ fatal_message[fatal],
	459	+ global_message[i]);
	460	+ }
441	461	}
442	462	}
443	463
...	...	@@ -1021,7 +1041,7 @@
1021	1041	struct pci_dev *dev;
1022	1042
1023	1043	pvt->bridge_ck = pci_get_device(PCI_VENDOR_ID_INTEL,
1024		- pvt->dev_info->err_dev, pvt->bridge_ck);
	1044	+ pvt->dev_info->err_dev, pvt->bridge_ck);
1025	1045
1026	1046	if (pvt->bridge_ck == NULL)
1027	1047	pvt->bridge_ck = pci_scan_single_device(pdev->bus,
...	...	@@ -1034,8 +1054,9 @@
1034	1054	return 1;
1035	1055	}
1036	1056
1037		- dev = pci_get_device(PCI_VENDOR_ID_INTEL, e752x_devs[dev_idx].ctl_dev,
1038		- NULL);
	1057	+ dev = pci_get_device(PCI_VENDOR_ID_INTEL,
	1058	+ e752x_devs[dev_idx].ctl_dev,
	1059	+ NULL);
1039	1060
1040	1061	if (dev == NULL)
1041	1062	goto fail;
...	...	@@ -1316,7 +1337,8 @@
1316	1337
1317	1338	module_param(force_function_unhide, int, 0444);
1318	1339	MODULE_PARM_DESC(force_function_unhide, "if BIOS sets Dev0:Fun1 up as hidden:"
1319		- " 1=force unhide and hope BIOS doesn't fight driver for Dev0:Fun1 access");
	1340	+ " 1=force unhide and hope BIOS doesn't fight driver for "
	1341	+ "Dev0:Fun1 access");
1320	1342
1321	1343	module_param(edac_op_state, int, 0444);
1322	1344	MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
...	...	@@ -1324,4 +1346,7 @@
1324	1346	module_param(sysbus_parity, int, 0444);
1325	1347	MODULE_PARM_DESC(sysbus_parity, "0=disable system bus parity checking,"
1326	1348	" 1=enable system bus parity checking, default=auto-detect");
	1349	+module_param(report_non_memory_errors, int, 0644);
	1350	+MODULE_PARM_DESC(report_non_memory_errors, "0=disable non-memory error "
	1351	+ "reporting, 1=enable non-memory error reporting");