forked from ~ljy/RK356X_SDK_RELEASE
blame | history | patch | commit | commitdiff
hc
2024-09-20 cf4ce59b3b70238352c7f1729f0f7223214828ad 
 kernel/arch/powerpc/kernel/mce.c
....@@ -1,19 +1,6 @@
1
+// SPDX-License-Identifier: GPL-2.0-or-later
12 /*
23  * Machine check exception handling.
3
- *
4
- * This program is free software; you can redistribute it and/or modify
5
- * it under the terms of the GNU General Public License as published by
6
- * the Free Software Foundation; either version 2 of the License, or
7
- * (at your option) any later version.
8
- *
9
- * This program is distributed in the hope that it will be useful,
10
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
11
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12
- * GNU General Public License for more details.
13
- *
14
- * You should have received a copy of the GNU General Public License
15
- * along with this program; if not, write to the Free Software
16
- * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
174  *
185  * Copyright 2013 IBM Corporation
196  * Author: Mahesh Salgaonkar <mahesh@linux.vnet.ibm.com>
....@@ -28,9 +15,12 @@
2815 #include <linux/percpu.h>
2916 #include <linux/export.h>
3017 #include <linux/irq_work.h>
18
+#include <linux/extable.h>
19
+#include <linux/ftrace.h>
3120 
3221 #include <asm/machdep.h>
3322 #include <asm/mce.h>
23
+#include <asm/nmi.h>
3424 
3525 static DEFINE_PER_CPU(int, mce_nest_count);
3626 static DEFINE_PER_CPU(struct machine_check_event[MAX_MC_EVT], mce_event);
....@@ -46,7 +36,7 @@
4636 
4737 static void machine_check_process_queued_event(struct irq_work *work);
4838 static void machine_check_ue_irq_work(struct irq_work *work);
49
-void machine_check_ue_event(struct machine_check_event *evt);
39
+static void machine_check_ue_event(struct machine_check_event *evt);
5040 static void machine_process_ue_event(struct work_struct *work);
5141 
5242 static struct irq_work mce_event_process_work = {
....@@ -58,6 +48,20 @@
5848 };
5949 
6050 DECLARE_WORK(mce_ue_event_work, machine_process_ue_event);
51
+
52
+static BLOCKING_NOTIFIER_HEAD(mce_notifier_list);
53
+
54
+int mce_register_notifier(struct notifier_block *nb)
55
+{
56
+	return blocking_notifier_chain_register(&mce_notifier_list, nb);
57
+}
58
+EXPORT_SYMBOL_GPL(mce_register_notifier);
59
+
60
+int mce_unregister_notifier(struct notifier_block *nb)
61
+{
62
+	return blocking_notifier_chain_unregister(&mce_notifier_list, nb);
63
+}
64
+EXPORT_SYMBOL_GPL(mce_unregister_notifier);
6165 
6266 static void mce_set_error_info(struct machine_check_event *mce,
6367 			       struct mce_error_info *mce_err)
....@@ -116,6 +120,7 @@
116120 	mce->srr1 = regs->msr;
117121 	mce->gpr3 = regs->gpr[3];
118122 	mce->in_use = 1;
123
+	mce->cpu = get_paca()->paca_index;
119124 
120125 	/* Mark it recovered if we have handled it and MSR(RI=1). */
121126 	if (handled && (regs->msr & MSR_RI))
....@@ -125,6 +130,8 @@
125130 
126131 	mce->initiator = mce_err->initiator;
127132 	mce->severity = mce_err->severity;
133
+	mce->sync_error = mce_err->sync_error;
134
+	mce->error_class = mce_err->error_class;
128135 
129136 	/*
130137 	 * Populate the mce error_type and type-specific error_type.
....@@ -158,6 +165,7 @@
158165 		if (phys_addr != ULONG_MAX) {
159166 			mce->u.ue_error.physical_address_provided = true;
160167 			mce->u.ue_error.physical_address = phys_addr;
168
+			mce->u.ue_error.ignore_event = mce_err->ignore_event;
161169 			machine_check_ue_event(mce);
162170 		}
163171 	}
....@@ -221,7 +229,7 @@
221229 /*
222230  * Queue up the MCE event which then can be handled later.
223231  */
224
-void machine_check_ue_event(struct machine_check_event *evt)
232
+static void machine_check_ue_event(struct machine_check_event *evt)
225233 {
226234 	int index;
227235 
....@@ -259,6 +267,19 @@
259267 	/* Queue irq work to process this event later. */
260268 	irq_work_queue(&mce_event_process_work);
261269 }
270
+
271
+void mce_common_process_ue(struct pt_regs *regs,
272
+			   struct mce_error_info *mce_err)
273
+{
274
+	const struct exception_table_entry *entry;
275
+
276
+	entry = search_kernel_exception_table(regs->nip);
277
+	if (entry) {
278
+		mce_err->ignore_event = true;
279
+		regs->nip = extable_fixup(entry);
280
+	}
281
+}
282
+
262283 /*
263284  * process pending MCE event from the mce event queue. This function will be
264285  * called during syscall exit.
....@@ -271,12 +292,22 @@
271292 	while (__this_cpu_read(mce_ue_count) > 0) {
272293 		index = __this_cpu_read(mce_ue_count) - 1;
273294 		evt = this_cpu_ptr(&mce_ue_event_queue[index]);
295
+		blocking_notifier_call_chain(&mce_notifier_list, 0, evt);
274296 #ifdef CONFIG_MEMORY_FAILURE
275297 		/*
276298 		 * This should probably queued elsewhere, but
277299 		 * oh! well
300
+		 *
301
+		 * Don't report this machine check because the caller has a
302
+		 * asked us to ignore the event, it has a fixup handler which
303
+		 * will do the appropriate error handling and reporting.
278304 		 */
279305 		if (evt->error_type == MCE_ERROR_TYPE_UE) {
306
+			if (evt->u.ue_error.ignore_event) {
307
+				__this_cpu_dec(mce_ue_count);
308
+				continue;
309
+			}
310
+
280311 			if (evt->u.ue_error.physical_address_provided) {
281312 				unsigned long pfn;
282313 
....@@ -310,15 +341,25 @@
310341 	while (__this_cpu_read(mce_queue_count) > 0) {
311342 		index = __this_cpu_read(mce_queue_count) - 1;
312343 		evt = this_cpu_ptr(&mce_event_queue[index]);
313
-		machine_check_print_event_info(evt, false);
344
+
345
+		if (evt->error_type == MCE_ERROR_TYPE_UE &&
346
+		    evt->u.ue_error.ignore_event) {
347
+			__this_cpu_dec(mce_queue_count);
348
+			continue;
349
+		}
350
+		machine_check_print_event_info(evt, false, false);
314351 		__this_cpu_dec(mce_queue_count);
315352 	}
316353 }
317354 
318355 void machine_check_print_event_info(struct machine_check_event *evt,
319
-				    bool user_mode)
356
+				    bool user_mode, bool in_guest)
320357 {
321
-	const char *level, *sevstr, *subtype;
358
+	const char *level, *sevstr, *subtype, *err_type, *initiator;
359
+	uint64_t ea = 0, pa = 0;
360
+	int n = 0;
361
+	char dar_str[50];
362
+	char pa_str[50];
322363 	static const char *mc_ue_types[] = {
323364 		"Indeterminate",
324365 		"Instruction fetch",
....@@ -344,6 +385,7 @@
344385 	static const char *mc_user_types[] = {
345386 		"Indeterminate",
346387 		"tlbie(l) invalid",
388
+		"scv invalid",
347389 	};
348390 	static const char *mc_ra_types[] = {
349391 		"Indeterminate",
....@@ -365,6 +407,13 @@
365407 		"Store (timeout)",
366408 		"Page table walk Load/Store (timeout)",
367409 	};
410
+	static const char *mc_error_class[] = {
411
+		"Unknown",
412
+		"Hardware error",
413
+		"Probable Hardware error (some chance of software cause)",
414
+		"Software error",
415
+		"Probable Software error (some chance of hardware cause)",
416
+	};
368417 
369418 	/* Print things out */
370419 	if (evt->version != MCE_V1) {
....@@ -379,9 +428,9 @@
379428 		break;
380429 	case MCE_SEV_WARNING:
381430 		level = KERN_WARNING;
382
-		sevstr = "";
431
+		sevstr = "Warning";
383432 		break;
384
-	case MCE_SEV_ERROR_SYNC:
433
+	case MCE_SEV_SEVERE:
385434 		level = KERN_ERR;
386435 		sevstr = "Severe";
387436 		break;
....@@ -392,99 +441,145 @@
392441 		break;
393442 	}
394443 
395
-	printk("%s%s Machine check interrupt [%s]\n", level, sevstr,
396
-	       evt->disposition == MCE_DISPOSITION_RECOVERED ?
397
-	       "Recovered" : "Not recovered");
398
-
399
-	if (user_mode) {
400
-		printk("%s  NIP: [%016llx] PID: %d Comm: %s\n", level,
401
-			evt->srr0, current->pid, current->comm);
402
-	} else {
403
-		printk("%s  NIP [%016llx]: %pS\n", level, evt->srr0,
404
-		       (void *)evt->srr0);
444
+	switch(evt->initiator) {
445
+	case MCE_INITIATOR_CPU:
446
+		initiator = "CPU";
447
+		break;
448
+	case MCE_INITIATOR_PCI:
449
+		initiator = "PCI";
450
+		break;
451
+	case MCE_INITIATOR_ISA:
452
+		initiator = "ISA";
453
+		break;
454
+	case MCE_INITIATOR_MEMORY:
455
+		initiator = "Memory";
456
+		break;
457
+	case MCE_INITIATOR_POWERMGM:
458
+		initiator = "Power Management";
459
+		break;
460
+	case MCE_INITIATOR_UNKNOWN:
461
+	default:
462
+		initiator = "Unknown";
463
+		break;
405464 	}
406465 
407
-	printk("%s  Initiator: %s\n", level,
408
-	       evt->initiator == MCE_INITIATOR_CPU ? "CPU" : "Unknown");
409466 	switch (evt->error_type) {
410467 	case MCE_ERROR_TYPE_UE:
468
+		err_type = "UE";
411469 		subtype = evt->u.ue_error.ue_error_type <
412470 			ARRAY_SIZE(mc_ue_types) ?
413471 			mc_ue_types[evt->u.ue_error.ue_error_type]
414472 			: "Unknown";
415
-		printk("%s  Error type: UE [%s]\n", level, subtype);
416473 		if (evt->u.ue_error.effective_address_provided)
417
-			printk("%s    Effective address: %016llx\n",
418
-			       level, evt->u.ue_error.effective_address);
474
+			ea = evt->u.ue_error.effective_address;
419475 		if (evt->u.ue_error.physical_address_provided)
420
-			printk("%s    Physical address:  %016llx\n",
421
-			       level, evt->u.ue_error.physical_address);
476
+			pa = evt->u.ue_error.physical_address;
422477 		break;
423478 	case MCE_ERROR_TYPE_SLB:
479
+		err_type = "SLB";
424480 		subtype = evt->u.slb_error.slb_error_type <
425481 			ARRAY_SIZE(mc_slb_types) ?
426482 			mc_slb_types[evt->u.slb_error.slb_error_type]
427483 			: "Unknown";
428
-		printk("%s  Error type: SLB [%s]\n", level, subtype);
429484 		if (evt->u.slb_error.effective_address_provided)
430
-			printk("%s    Effective address: %016llx\n",
431
-			       level, evt->u.slb_error.effective_address);
485
+			ea = evt->u.slb_error.effective_address;
432486 		break;
433487 	case MCE_ERROR_TYPE_ERAT:
488
+		err_type = "ERAT";
434489 		subtype = evt->u.erat_error.erat_error_type <
435490 			ARRAY_SIZE(mc_erat_types) ?
436491 			mc_erat_types[evt->u.erat_error.erat_error_type]
437492 			: "Unknown";
438
-		printk("%s  Error type: ERAT [%s]\n", level, subtype);
439493 		if (evt->u.erat_error.effective_address_provided)
440
-			printk("%s    Effective address: %016llx\n",
441
-			       level, evt->u.erat_error.effective_address);
494
+			ea = evt->u.erat_error.effective_address;
442495 		break;
443496 	case MCE_ERROR_TYPE_TLB:
497
+		err_type = "TLB";
444498 		subtype = evt->u.tlb_error.tlb_error_type <
445499 			ARRAY_SIZE(mc_tlb_types) ?
446500 			mc_tlb_types[evt->u.tlb_error.tlb_error_type]
447501 			: "Unknown";
448
-		printk("%s  Error type: TLB [%s]\n", level, subtype);
449502 		if (evt->u.tlb_error.effective_address_provided)
450
-			printk("%s    Effective address: %016llx\n",
451
-			       level, evt->u.tlb_error.effective_address);
503
+			ea = evt->u.tlb_error.effective_address;
452504 		break;
453505 	case MCE_ERROR_TYPE_USER:
506
+		err_type = "User";
454507 		subtype = evt->u.user_error.user_error_type <
455508 			ARRAY_SIZE(mc_user_types) ?
456509 			mc_user_types[evt->u.user_error.user_error_type]
457510 			: "Unknown";
458
-		printk("%s  Error type: User [%s]\n", level, subtype);
459511 		if (evt->u.user_error.effective_address_provided)
460
-			printk("%s    Effective address: %016llx\n",
461
-			       level, evt->u.user_error.effective_address);
512
+			ea = evt->u.user_error.effective_address;
462513 		break;
463514 	case MCE_ERROR_TYPE_RA:
515
+		err_type = "Real address";
464516 		subtype = evt->u.ra_error.ra_error_type <
465517 			ARRAY_SIZE(mc_ra_types) ?
466518 			mc_ra_types[evt->u.ra_error.ra_error_type]
467519 			: "Unknown";
468
-		printk("%s  Error type: Real address [%s]\n", level, subtype);
469520 		if (evt->u.ra_error.effective_address_provided)
470
-			printk("%s    Effective address: %016llx\n",
471
-			       level, evt->u.ra_error.effective_address);
521
+			ea = evt->u.ra_error.effective_address;
472522 		break;
473523 	case MCE_ERROR_TYPE_LINK:
524
+		err_type = "Link";
474525 		subtype = evt->u.link_error.link_error_type <
475526 			ARRAY_SIZE(mc_link_types) ?
476527 			mc_link_types[evt->u.link_error.link_error_type]
477528 			: "Unknown";
478
-		printk("%s  Error type: Link [%s]\n", level, subtype);
479529 		if (evt->u.link_error.effective_address_provided)
480
-			printk("%s    Effective address: %016llx\n",
481
-			       level, evt->u.link_error.effective_address);
530
+			ea = evt->u.link_error.effective_address;
531
+		break;
532
+	case MCE_ERROR_TYPE_DCACHE:
533
+		err_type = "D-Cache";
534
+		subtype = "Unknown";
535
+		break;
536
+	case MCE_ERROR_TYPE_ICACHE:
537
+		err_type = "I-Cache";
538
+		subtype = "Unknown";
482539 		break;
483540 	default:
484541 	case MCE_ERROR_TYPE_UNKNOWN:
485
-		printk("%s  Error type: Unknown\n", level);
542
+		err_type = "Unknown";
543
+		subtype = "";
486544 		break;
487545 	}
546
+
547
+	dar_str[0] = pa_str[0] = '\0';
548
+	if (ea && evt->srr0 != ea) {
549
+		/* Load/Store address */
550
+		n = sprintf(dar_str, "DAR: %016llx ", ea);
551
+		if (pa)
552
+			sprintf(dar_str + n, "paddr: %016llx ", pa);
553
+	} else if (pa) {
554
+		sprintf(pa_str, " paddr: %016llx", pa);
555
+	}
556
+
557
+	printk("%sMCE: CPU%d: machine check (%s) %s %s %s %s[%s]\n",
558
+		level, evt->cpu, sevstr, in_guest ? "Guest" : "Host",
559
+		err_type, subtype, dar_str,
560
+		evt->disposition == MCE_DISPOSITION_RECOVERED ?
561
+		"Recovered" : "Not recovered");
562
+
563
+	if (in_guest || user_mode) {
564
+		printk("%sMCE: CPU%d: PID: %d Comm: %s %sNIP: [%016llx]%s\n",
565
+			level, evt->cpu, current->pid, current->comm,
566
+			in_guest ? "Guest " : "", evt->srr0, pa_str);
567
+	} else {
568
+		printk("%sMCE: CPU%d: NIP: [%016llx] %pS%s\n",
569
+			level, evt->cpu, evt->srr0, (void *)evt->srr0, pa_str);
570
+	}
571
+
572
+	printk("%sMCE: CPU%d: Initiator %s\n", level, evt->cpu, initiator);
573
+
574
+	subtype = evt->error_class < ARRAY_SIZE(mc_error_class) ?
575
+		mc_error_class[evt->error_class] : "Unknown";
576
+	printk("%sMCE: CPU%d: %s\n", level, evt->cpu, subtype);
577
+
578
+#ifdef CONFIG_PPC_BOOK3S_64
579
+	/* Display faulty slb contents for SLB errors. */
580
+	if (evt->error_type == MCE_ERROR_TYPE_SLB)
581
+		slb_dump_contents(local_paca->mce_faulty_slbs);
582
+#endif
488583 }
489584 EXPORT_SYMBOL_GPL(machine_check_print_event_info);
490585 
....@@ -493,14 +588,29 @@
493588  *
494589  * regs->nip and regs->msr contains srr0 and ssr1.
495590  */
496
-long machine_check_early(struct pt_regs *regs)
591
+long notrace machine_check_early(struct pt_regs *regs)
497592 {
498593 	long handled = 0;
594
+	u8 ftrace_enabled = this_cpu_get_ftrace_enabled();
499595 
500
-	__this_cpu_inc(irq_stat.mce_exceptions);
596
+	this_cpu_set_ftrace_enabled(0);
597
+	/* Do not use nmi_enter/exit for pseries hpte guest */
598
+	if (radix_enabled() || !firmware_has_feature(FW_FEATURE_LPAR))
599
+		nmi_enter();
501600 
502
-	if (cur_cpu_spec && cur_cpu_spec->machine_check_early)
503
-		handled = cur_cpu_spec->machine_check_early(regs);
601
+	hv_nmi_check_nonrecoverable(regs);
602
+
603
+	/*
604
+	 * See if platform is capable of handling machine check.
605
+	 */
606
+	if (ppc_md.machine_check_early)
607
+		handled = ppc_md.machine_check_early(regs);
608
+
609
+	if (radix_enabled() || !firmware_has_feature(FW_FEATURE_LPAR))
610
+		nmi_exit();
611
+
612
+	this_cpu_set_ftrace_enabled(ftrace_enabled);
613
+
504614 	return handled;
505615 }
506616 
....@@ -616,7 +726,7 @@
616726 {	
617727 	int ret;
618728 
619
-	__this_cpu_inc(irq_stat.hmi_exceptions);
729
+	local_paca->hmi_irqs++;
620730 
621731 	ret = hmi_handle_debugtrig(regs);
622732 	if (ret >= 0)