2 * APEI Generic Hardware Error Source support
4 * Generic Hardware Error Source provides a way to report platform
5 * hardware errors (such as that from chipset). It works in so called
6 * "Firmware First" mode, that is, hardware errors are reported to
7 * firmware firstly, then reported to Linux by firmware. This way,
8 * some non-standard hardware error registers or non-standard hardware
9 * link can be checked by firmware to produce more hardware error
10 * information for Linux.
12 * For more information about Generic Hardware Error Source, please
13 * refer to ACPI Specification version 4.0, section 17.3.2.6
15 * Copyright 2010,2011 Intel Corp.
16 * Author: Huang Ying <ying.huang@intel.com>
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License version
20 * 2 as published by the Free Software Foundation;
22 * This program is distributed in the hope that it will be useful,
23 * but WITHOUT ANY WARRANTY; without even the implied warranty of
24 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25 * GNU General Public License for more details.
27 * You should have received a copy of the GNU General Public License
28 * along with this program; if not, write to the Free Software
29 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
32 #include <linux/kernel.h>
33 #include <linux/module.h>
34 #include <linux/init.h>
35 #include <linux/acpi.h>
37 #include <linux/interrupt.h>
38 #include <linux/timer.h>
39 #include <linux/cper.h>
40 #include <linux/kdebug.h>
41 #include <linux/platform_device.h>
42 #include <linux/mutex.h>
43 #include <linux/ratelimit.h>
44 #include <linux/vmalloc.h>
45 #include <linux/irq_work.h>
46 #include <linux/llist.h>
47 #include <linux/genalloc.h>
48 #include <acpi/apei.h>
49 #include <acpi/atomicio.h>
52 #include <asm/tlbflush.h>
54 #include "apei-internal.h"
56 #define GHES_PFX "GHES: "
58 #define GHES_ESTATUS_MAX_SIZE 65536
59 #define GHES_ESOURCE_PREALLOC_MAX_SIZE 65536
61 #define GHES_ESTATUS_POOL_MIN_ALLOC_ORDER 3
63 /* This is just an estimation for memory pool allocation */
64 #define GHES_ESTATUS_CACHE_AVG_SIZE 512
66 #define GHES_ESTATUS_CACHES_SIZE 4
68 #define GHES_ESTATUS_IN_CACHE_MAX_NSEC (10 * NSEC_PER_SEC)
69 /* Prevent too many caches are allocated because of RCU */
70 #define GHES_ESTATUS_CACHE_ALLOCED_MAX (GHES_ESTATUS_CACHES_SIZE * 3 / 2)
72 #define GHES_ESTATUS_CACHE_LEN(estatus_len) \
73 (sizeof(struct ghes_estatus_cache) + (estatus_len))
74 #define GHES_ESTATUS_FROM_CACHE(estatus_cache) \
75 ((struct acpi_hest_generic_status *) \
76 ((struct ghes_estatus_cache *)(estatus_cache) + 1))
78 #define GHES_ESTATUS_NODE_LEN(estatus_len) \
79 (sizeof(struct ghes_estatus_node) + (estatus_len))
80 #define GHES_ESTATUS_FROM_NODE(estatus_node) \
81 ((struct acpi_hest_generic_status *) \
82 ((struct ghes_estatus_node *)(estatus_node) + 1))
85 * One struct ghes is created for each generic hardware error source.
86 * It provides the context for APEI hardware error timer/IRQ/SCI/NMI
89 * estatus: memory buffer for error status block, allocated during
92 #define GHES_TO_CLEAR 0x0001
93 #define GHES_EXITING 0x0002
96 struct acpi_hest_generic *generic;
97 struct acpi_hest_generic_status *estatus;
101 struct list_head list;
102 struct timer_list timer;
107 struct ghes_estatus_node {
108 struct llist_node llnode;
109 struct acpi_hest_generic *generic;
112 struct ghes_estatus_cache {
115 struct acpi_hest_generic *generic;
116 unsigned long long time_in;
121 module_param_named(disable, ghes_disable, bool, 0);
123 static int ghes_panic_timeout __read_mostly = 30;
126 * All error sources notified with SCI shares one notifier function,
127 * so they need to be linked and checked one by one. This is applied
130 * RCU is used for these lists, so ghes_list_mutex is only used for
131 * list changing, not for traversing.
133 static LIST_HEAD(ghes_sci);
134 static LIST_HEAD(ghes_nmi);
135 static DEFINE_MUTEX(ghes_list_mutex);
138 * NMI may be triggered on any CPU, so ghes_nmi_lock is used for
141 static DEFINE_RAW_SPINLOCK(ghes_nmi_lock);
144 * Because the memory area used to transfer hardware error information
145 * from BIOS to Linux can be determined only in NMI, IRQ or timer
146 * handler, but general ioremap can not be used in atomic context, so
147 * a special version of atomic ioremap is implemented for that.
151 * Two virtual pages are used, one for NMI context, the other for
152 * IRQ/PROCESS context
154 #define GHES_IOREMAP_PAGES 2
155 #define GHES_IOREMAP_NMI_PAGE(base) (base)
156 #define GHES_IOREMAP_IRQ_PAGE(base) ((base) + PAGE_SIZE)
158 /* virtual memory area for atomic ioremap */
159 static struct vm_struct *ghes_ioremap_area;
161 * These 2 spinlock is used to prevent atomic ioremap virtual memory
162 * area from being mapped simultaneously.
164 static DEFINE_RAW_SPINLOCK(ghes_ioremap_lock_nmi);
165 static DEFINE_SPINLOCK(ghes_ioremap_lock_irq);
168 * printk is not safe in NMI context. So in NMI handler, we allocate
169 * required memory from lock-less memory allocator
170 * (ghes_estatus_pool), save estatus into it, put them into lock-less
171 * list (ghes_estatus_llist), then delay printk into IRQ context via
172 * irq_work (ghes_proc_irq_work). ghes_estatus_size_request record
173 * required pool size by all NMI error source.
175 static struct gen_pool *ghes_estatus_pool;
176 static unsigned long ghes_estatus_pool_size_request;
177 static struct llist_head ghes_estatus_llist;
178 static struct irq_work ghes_proc_irq_work;
180 struct ghes_estatus_cache *ghes_estatus_caches[GHES_ESTATUS_CACHES_SIZE];
181 static atomic_t ghes_estatus_cache_alloced;
183 static int ghes_ioremap_init(void)
185 ghes_ioremap_area = __get_vm_area(PAGE_SIZE * GHES_IOREMAP_PAGES,
186 VM_IOREMAP, VMALLOC_START, VMALLOC_END);
187 if (!ghes_ioremap_area) {
188 pr_err(GHES_PFX "Failed to allocate virtual memory area for atomic ioremap.\n");
195 static void ghes_ioremap_exit(void)
197 free_vm_area(ghes_ioremap_area);
200 static void __iomem *ghes_ioremap_pfn_nmi(u64 pfn)
204 vaddr = (unsigned long)GHES_IOREMAP_NMI_PAGE(ghes_ioremap_area->addr);
205 ioremap_page_range(vaddr, vaddr + PAGE_SIZE,
206 pfn << PAGE_SHIFT, PAGE_KERNEL);
208 return (void __iomem *)vaddr;
211 static void __iomem *ghes_ioremap_pfn_irq(u64 pfn)
215 vaddr = (unsigned long)GHES_IOREMAP_IRQ_PAGE(ghes_ioremap_area->addr);
216 ioremap_page_range(vaddr, vaddr + PAGE_SIZE,
217 pfn << PAGE_SHIFT, PAGE_KERNEL);
219 return (void __iomem *)vaddr;
222 static void ghes_iounmap_nmi(void __iomem *vaddr_ptr)
224 unsigned long vaddr = (unsigned long __force)vaddr_ptr;
225 void *base = ghes_ioremap_area->addr;
227 BUG_ON(vaddr != (unsigned long)GHES_IOREMAP_NMI_PAGE(base));
228 unmap_kernel_range_noflush(vaddr, PAGE_SIZE);
229 __flush_tlb_one(vaddr);
232 static void ghes_iounmap_irq(void __iomem *vaddr_ptr)
234 unsigned long vaddr = (unsigned long __force)vaddr_ptr;
235 void *base = ghes_ioremap_area->addr;
237 BUG_ON(vaddr != (unsigned long)GHES_IOREMAP_IRQ_PAGE(base));
238 unmap_kernel_range_noflush(vaddr, PAGE_SIZE);
239 __flush_tlb_one(vaddr);
242 static int ghes_estatus_pool_init(void)
244 ghes_estatus_pool = gen_pool_create(GHES_ESTATUS_POOL_MIN_ALLOC_ORDER, -1);
245 if (!ghes_estatus_pool)
250 static void ghes_estatus_pool_free_chunk_page(struct gen_pool *pool,
251 struct gen_pool_chunk *chunk,
254 free_page(chunk->start_addr);
257 static void ghes_estatus_pool_exit(void)
259 gen_pool_for_each_chunk(ghes_estatus_pool,
260 ghes_estatus_pool_free_chunk_page, NULL);
261 gen_pool_destroy(ghes_estatus_pool);
264 static int ghes_estatus_pool_expand(unsigned long len)
266 unsigned long i, pages, size, addr;
269 ghes_estatus_pool_size_request += PAGE_ALIGN(len);
270 size = gen_pool_size(ghes_estatus_pool);
271 if (size >= ghes_estatus_pool_size_request)
273 pages = (ghes_estatus_pool_size_request - size) / PAGE_SIZE;
274 for (i = 0; i < pages; i++) {
275 addr = __get_free_page(GFP_KERNEL);
278 ret = gen_pool_add(ghes_estatus_pool, addr, PAGE_SIZE, -1);
286 static void ghes_estatus_pool_shrink(unsigned long len)
288 ghes_estatus_pool_size_request -= PAGE_ALIGN(len);
291 static struct ghes *ghes_new(struct acpi_hest_generic *generic)
294 unsigned int error_block_length;
297 ghes = kzalloc(sizeof(*ghes), GFP_KERNEL);
299 return ERR_PTR(-ENOMEM);
300 ghes->generic = generic;
301 rc = acpi_pre_map_gar(&generic->error_status_address);
304 error_block_length = generic->error_block_length;
305 if (error_block_length > GHES_ESTATUS_MAX_SIZE) {
306 pr_warning(FW_WARN GHES_PFX
307 "Error status block length is too long: %u for "
308 "generic hardware error source: %d.\n",
309 error_block_length, generic->header.source_id);
310 error_block_length = GHES_ESTATUS_MAX_SIZE;
312 ghes->estatus = kmalloc(error_block_length, GFP_KERNEL);
313 if (!ghes->estatus) {
321 acpi_post_unmap_gar(&generic->error_status_address);
327 static void ghes_fini(struct ghes *ghes)
329 kfree(ghes->estatus);
330 acpi_post_unmap_gar(&ghes->generic->error_status_address);
335 GHES_SEV_CORRECTED = 0x1,
336 GHES_SEV_RECOVERABLE = 0x2,
337 GHES_SEV_PANIC = 0x3,
340 static inline int ghes_severity(int severity)
343 case CPER_SEV_INFORMATIONAL:
345 case CPER_SEV_CORRECTED:
346 return GHES_SEV_CORRECTED;
347 case CPER_SEV_RECOVERABLE:
348 return GHES_SEV_RECOVERABLE;
350 return GHES_SEV_PANIC;
352 /* Unknown, go panic */
353 return GHES_SEV_PANIC;
357 static void ghes_copy_tofrom_phys(void *buffer, u64 paddr, u32 len,
361 unsigned long flags = 0;
362 int in_nmi = in_nmi();
367 offset = paddr - (paddr & PAGE_MASK);
369 raw_spin_lock(&ghes_ioremap_lock_nmi);
370 vaddr = ghes_ioremap_pfn_nmi(paddr >> PAGE_SHIFT);
372 spin_lock_irqsave(&ghes_ioremap_lock_irq, flags);
373 vaddr = ghes_ioremap_pfn_irq(paddr >> PAGE_SHIFT);
375 trunk = PAGE_SIZE - offset;
376 trunk = min(trunk, len);
378 memcpy_fromio(buffer, vaddr + offset, trunk);
380 memcpy_toio(vaddr + offset, buffer, trunk);
385 ghes_iounmap_nmi(vaddr);
386 raw_spin_unlock(&ghes_ioremap_lock_nmi);
388 ghes_iounmap_irq(vaddr);
389 spin_unlock_irqrestore(&ghes_ioremap_lock_irq, flags);
394 static int ghes_read_estatus(struct ghes *ghes, int silent)
396 struct acpi_hest_generic *g = ghes->generic;
401 rc = acpi_atomic_read(&buf_paddr, &g->error_status_address);
403 if (!silent && printk_ratelimit())
404 pr_warning(FW_WARN GHES_PFX
405 "Failed to read error status block address for hardware error source: %d.\n",
406 g->header.source_id);
412 ghes_copy_tofrom_phys(ghes->estatus, buf_paddr,
413 sizeof(*ghes->estatus), 1);
414 if (!ghes->estatus->block_status)
417 ghes->buffer_paddr = buf_paddr;
418 ghes->flags |= GHES_TO_CLEAR;
421 len = apei_estatus_len(ghes->estatus);
422 if (len < sizeof(*ghes->estatus))
424 if (len > ghes->generic->error_block_length)
426 if (apei_estatus_check_header(ghes->estatus))
428 ghes_copy_tofrom_phys(ghes->estatus + 1,
429 buf_paddr + sizeof(*ghes->estatus),
430 len - sizeof(*ghes->estatus), 1);
431 if (apei_estatus_check(ghes->estatus))
436 if (rc && !silent && printk_ratelimit())
437 pr_warning(FW_WARN GHES_PFX
438 "Failed to read error status block!\n");
442 static void ghes_clear_estatus(struct ghes *ghes)
444 ghes->estatus->block_status = 0;
445 if (!(ghes->flags & GHES_TO_CLEAR))
447 ghes_copy_tofrom_phys(ghes->estatus, ghes->buffer_paddr,
448 sizeof(ghes->estatus->block_status), 0);
449 ghes->flags &= ~GHES_TO_CLEAR;
452 static void ghes_do_proc(const struct acpi_hest_generic_status *estatus)
454 int sev, processed = 0;
455 struct acpi_hest_generic_data *gdata;
457 sev = ghes_severity(estatus->error_severity);
458 apei_estatus_for_each_section(estatus, gdata) {
459 #ifdef CONFIG_X86_MCE
460 if (!uuid_le_cmp(*(uuid_le *)gdata->section_type,
461 CPER_SEC_PLATFORM_MEM)) {
462 apei_mce_report_mem_error(
463 sev == GHES_SEV_CORRECTED,
464 (struct cper_sec_mem_err *)(gdata+1));
471 static void __ghes_print_estatus(const char *pfx,
472 const struct acpi_hest_generic *generic,
473 const struct acpi_hest_generic_status *estatus)
476 if (ghes_severity(estatus->error_severity) <=
478 pfx = KERN_WARNING HW_ERR;
480 pfx = KERN_ERR HW_ERR;
482 printk("%s""Hardware error from APEI Generic Hardware Error Source: %d\n",
483 pfx, generic->header.source_id);
484 apei_estatus_print(pfx, estatus);
487 static int ghes_print_estatus(const char *pfx,
488 const struct acpi_hest_generic *generic,
489 const struct acpi_hest_generic_status *estatus)
491 /* Not more than 2 messages every 5 seconds */
492 static DEFINE_RATELIMIT_STATE(ratelimit_corrected, 5*HZ, 2);
493 static DEFINE_RATELIMIT_STATE(ratelimit_uncorrected, 5*HZ, 2);
494 struct ratelimit_state *ratelimit;
496 if (ghes_severity(estatus->error_severity) <= GHES_SEV_CORRECTED)
497 ratelimit = &ratelimit_corrected;
499 ratelimit = &ratelimit_uncorrected;
500 if (__ratelimit(ratelimit)) {
501 __ghes_print_estatus(pfx, generic, estatus);
508 * GHES error status reporting throttle, to report more kinds of
509 * errors, instead of just most frequently occurred errors.
511 static int ghes_estatus_cached(struct acpi_hest_generic_status *estatus)
515 unsigned long long now;
516 struct ghes_estatus_cache *cache;
517 struct acpi_hest_generic_status *cache_estatus;
519 len = apei_estatus_len(estatus);
521 for (i = 0; i < GHES_ESTATUS_CACHES_SIZE; i++) {
522 cache = rcu_dereference(ghes_estatus_caches[i]);
525 if (len != cache->estatus_len)
527 cache_estatus = GHES_ESTATUS_FROM_CACHE(cache);
528 if (memcmp(estatus, cache_estatus, len))
530 atomic_inc(&cache->count);
532 if (now - cache->time_in < GHES_ESTATUS_IN_CACHE_MAX_NSEC)
540 static struct ghes_estatus_cache *ghes_estatus_cache_alloc(
541 struct acpi_hest_generic *generic,
542 struct acpi_hest_generic_status *estatus)
546 struct ghes_estatus_cache *cache;
547 struct acpi_hest_generic_status *cache_estatus;
549 alloced = atomic_add_return(1, &ghes_estatus_cache_alloced);
550 if (alloced > GHES_ESTATUS_CACHE_ALLOCED_MAX) {
551 atomic_dec(&ghes_estatus_cache_alloced);
554 len = apei_estatus_len(estatus);
555 cache_len = GHES_ESTATUS_CACHE_LEN(len);
556 cache = (void *)gen_pool_alloc(ghes_estatus_pool, cache_len);
558 atomic_dec(&ghes_estatus_cache_alloced);
561 cache_estatus = GHES_ESTATUS_FROM_CACHE(cache);
562 memcpy(cache_estatus, estatus, len);
563 cache->estatus_len = len;
564 atomic_set(&cache->count, 0);
565 cache->generic = generic;
566 cache->time_in = sched_clock();
570 static void ghes_estatus_cache_free(struct ghes_estatus_cache *cache)
574 len = apei_estatus_len(GHES_ESTATUS_FROM_CACHE(cache));
575 len = GHES_ESTATUS_CACHE_LEN(len);
576 gen_pool_free(ghes_estatus_pool, (unsigned long)cache, len);
577 atomic_dec(&ghes_estatus_cache_alloced);
580 static void ghes_estatus_cache_rcu_free(struct rcu_head *head)
582 struct ghes_estatus_cache *cache;
584 cache = container_of(head, struct ghes_estatus_cache, rcu);
585 ghes_estatus_cache_free(cache);
588 static void ghes_estatus_cache_add(
589 struct acpi_hest_generic *generic,
590 struct acpi_hest_generic_status *estatus)
592 int i, slot = -1, count;
593 unsigned long long now, duration, period, max_period = 0;
594 struct ghes_estatus_cache *cache, *slot_cache = NULL, *new_cache;
596 new_cache = ghes_estatus_cache_alloc(generic, estatus);
597 if (new_cache == NULL)
601 for (i = 0; i < GHES_ESTATUS_CACHES_SIZE; i++) {
602 cache = rcu_dereference(ghes_estatus_caches[i]);
608 duration = now - cache->time_in;
609 if (duration >= GHES_ESTATUS_IN_CACHE_MAX_NSEC) {
614 count = atomic_read(&cache->count);
615 period = duration / (count + 1);
616 if (period > max_period) {
622 /* new_cache must be put into array after its contents are written */
624 if (slot != -1 && cmpxchg(ghes_estatus_caches + slot,
625 slot_cache, new_cache) == slot_cache) {
627 call_rcu(&slot_cache->rcu, ghes_estatus_cache_rcu_free);
629 ghes_estatus_cache_free(new_cache);
633 static int ghes_proc(struct ghes *ghes)
637 rc = ghes_read_estatus(ghes, 0);
640 if (!ghes_estatus_cached(ghes->estatus)) {
641 if (ghes_print_estatus(NULL, ghes->generic, ghes->estatus))
642 ghes_estatus_cache_add(ghes->generic, ghes->estatus);
644 ghes_do_proc(ghes->estatus);
646 ghes_clear_estatus(ghes);
650 static void ghes_add_timer(struct ghes *ghes)
652 struct acpi_hest_generic *g = ghes->generic;
653 unsigned long expire;
655 if (!g->notify.poll_interval) {
656 pr_warning(FW_WARN GHES_PFX "Poll interval is 0 for generic hardware error source: %d, disabled.\n",
657 g->header.source_id);
660 expire = jiffies + msecs_to_jiffies(g->notify.poll_interval);
661 ghes->timer.expires = round_jiffies_relative(expire);
662 add_timer(&ghes->timer);
665 static void ghes_poll_func(unsigned long data)
667 struct ghes *ghes = (void *)data;
670 if (!(ghes->flags & GHES_EXITING))
671 ghes_add_timer(ghes);
674 static irqreturn_t ghes_irq_func(int irq, void *data)
676 struct ghes *ghes = data;
679 rc = ghes_proc(ghes);
686 static int ghes_notify_sci(struct notifier_block *this,
687 unsigned long event, void *data)
690 int ret = NOTIFY_DONE;
693 list_for_each_entry_rcu(ghes, &ghes_sci, list) {
694 if (!ghes_proc(ghes))
702 static void ghes_proc_in_irq(struct irq_work *irq_work)
704 struct llist_node *llnode, *next, *tail = NULL;
705 struct ghes_estatus_node *estatus_node;
706 struct acpi_hest_generic *generic;
707 struct acpi_hest_generic_status *estatus;
711 * Because the time order of estatus in list is reversed,
712 * revert it back to proper order.
714 llnode = llist_del_all(&ghes_estatus_llist);
724 estatus_node = llist_entry(llnode, struct ghes_estatus_node,
726 estatus = GHES_ESTATUS_FROM_NODE(estatus_node);
727 len = apei_estatus_len(estatus);
728 node_len = GHES_ESTATUS_NODE_LEN(len);
729 ghes_do_proc(estatus);
730 if (!ghes_estatus_cached(estatus)) {
731 generic = estatus_node->generic;
732 if (ghes_print_estatus(NULL, generic, estatus))
733 ghes_estatus_cache_add(generic, estatus);
735 gen_pool_free(ghes_estatus_pool, (unsigned long)estatus_node,
741 static int ghes_notify_nmi(struct notifier_block *this,
742 unsigned long cmd, void *data)
744 struct ghes *ghes, *ghes_global = NULL;
745 int sev, sev_global = -1;
746 int ret = NOTIFY_DONE;
751 raw_spin_lock(&ghes_nmi_lock);
752 list_for_each_entry_rcu(ghes, &ghes_nmi, list) {
753 if (ghes_read_estatus(ghes, 1)) {
754 ghes_clear_estatus(ghes);
757 sev = ghes_severity(ghes->estatus->error_severity);
758 if (sev > sev_global) {
765 if (ret == NOTIFY_DONE)
768 if (sev_global >= GHES_SEV_PANIC) {
770 __ghes_print_estatus(KERN_EMERG HW_ERR, ghes_global->generic,
771 ghes_global->estatus);
772 /* reboot to log the error! */
773 if (panic_timeout == 0)
774 panic_timeout = ghes_panic_timeout;
775 panic("Fatal hardware error!");
778 list_for_each_entry_rcu(ghes, &ghes_nmi, list) {
779 #ifdef CONFIG_ARCH_HAVE_NMI_SAFE_CMPXCHG
781 struct ghes_estatus_node *estatus_node;
782 struct acpi_hest_generic_status *estatus;
784 if (!(ghes->flags & GHES_TO_CLEAR))
786 #ifdef CONFIG_ARCH_HAVE_NMI_SAFE_CMPXCHG
787 if (ghes_estatus_cached(ghes->estatus))
789 /* Save estatus for further processing in IRQ context */
790 len = apei_estatus_len(ghes->estatus);
791 node_len = GHES_ESTATUS_NODE_LEN(len);
792 estatus_node = (void *)gen_pool_alloc(ghes_estatus_pool,
795 estatus_node->generic = ghes->generic;
796 estatus = GHES_ESTATUS_FROM_NODE(estatus_node);
797 memcpy(estatus, ghes->estatus, len);
798 llist_add(&estatus_node->llnode, &ghes_estatus_llist);
802 ghes_clear_estatus(ghes);
804 #ifdef CONFIG_ARCH_HAVE_NMI_SAFE_CMPXCHG
805 irq_work_queue(&ghes_proc_irq_work);
809 raw_spin_unlock(&ghes_nmi_lock);
813 static struct notifier_block ghes_notifier_sci = {
814 .notifier_call = ghes_notify_sci,
817 static struct notifier_block ghes_notifier_nmi = {
818 .notifier_call = ghes_notify_nmi,
821 static unsigned long ghes_esource_prealloc_size(
822 const struct acpi_hest_generic *generic)
824 unsigned long block_length, prealloc_records, prealloc_size;
826 block_length = min_t(unsigned long, generic->error_block_length,
827 GHES_ESTATUS_MAX_SIZE);
828 prealloc_records = max_t(unsigned long,
829 generic->records_to_preallocate, 1);
830 prealloc_size = min_t(unsigned long, block_length * prealloc_records,
831 GHES_ESOURCE_PREALLOC_MAX_SIZE);
833 return prealloc_size;
836 static int __devinit ghes_probe(struct platform_device *ghes_dev)
838 struct acpi_hest_generic *generic;
839 struct ghes *ghes = NULL;
843 generic = *(struct acpi_hest_generic **)ghes_dev->dev.platform_data;
844 if (!generic->enabled)
847 switch (generic->notify.type) {
848 case ACPI_HEST_NOTIFY_POLLED:
849 case ACPI_HEST_NOTIFY_EXTERNAL:
850 case ACPI_HEST_NOTIFY_SCI:
851 case ACPI_HEST_NOTIFY_NMI:
853 case ACPI_HEST_NOTIFY_LOCAL:
854 pr_warning(GHES_PFX "Generic hardware error source: %d notified via local interrupt is not supported!\n",
855 generic->header.source_id);
858 pr_warning(FW_WARN GHES_PFX "Unknown notification type: %u for generic hardware error source: %d\n",
859 generic->notify.type, generic->header.source_id);
864 if (generic->error_block_length <
865 sizeof(struct acpi_hest_generic_status)) {
866 pr_warning(FW_BUG GHES_PFX "Invalid error block length: %u for generic hardware error source: %d\n",
867 generic->error_block_length,
868 generic->header.source_id);
871 ghes = ghes_new(generic);
877 switch (generic->notify.type) {
878 case ACPI_HEST_NOTIFY_POLLED:
879 ghes->timer.function = ghes_poll_func;
880 ghes->timer.data = (unsigned long)ghes;
881 init_timer_deferrable(&ghes->timer);
882 ghes_add_timer(ghes);
884 case ACPI_HEST_NOTIFY_EXTERNAL:
885 /* External interrupt vector is GSI */
886 if (acpi_gsi_to_irq(generic->notify.vector, &ghes->irq)) {
887 pr_err(GHES_PFX "Failed to map GSI to IRQ for generic hardware error source: %d\n",
888 generic->header.source_id);
891 if (request_irq(ghes->irq, ghes_irq_func,
892 0, "GHES IRQ", ghes)) {
893 pr_err(GHES_PFX "Failed to register IRQ for generic hardware error source: %d\n",
894 generic->header.source_id);
898 case ACPI_HEST_NOTIFY_SCI:
899 mutex_lock(&ghes_list_mutex);
900 if (list_empty(&ghes_sci))
901 register_acpi_hed_notifier(&ghes_notifier_sci);
902 list_add_rcu(&ghes->list, &ghes_sci);
903 mutex_unlock(&ghes_list_mutex);
905 case ACPI_HEST_NOTIFY_NMI:
906 len = ghes_esource_prealloc_size(generic);
907 ghes_estatus_pool_expand(len);
908 mutex_lock(&ghes_list_mutex);
909 if (list_empty(&ghes_nmi))
910 register_die_notifier(&ghes_notifier_nmi);
911 list_add_rcu(&ghes->list, &ghes_nmi);
912 mutex_unlock(&ghes_list_mutex);
917 platform_set_drvdata(ghes_dev, ghes);
928 static int __devexit ghes_remove(struct platform_device *ghes_dev)
931 struct acpi_hest_generic *generic;
934 ghes = platform_get_drvdata(ghes_dev);
935 generic = ghes->generic;
937 ghes->flags |= GHES_EXITING;
938 switch (generic->notify.type) {
939 case ACPI_HEST_NOTIFY_POLLED:
940 del_timer_sync(&ghes->timer);
942 case ACPI_HEST_NOTIFY_EXTERNAL:
943 free_irq(ghes->irq, ghes);
945 case ACPI_HEST_NOTIFY_SCI:
946 mutex_lock(&ghes_list_mutex);
947 list_del_rcu(&ghes->list);
948 if (list_empty(&ghes_sci))
949 unregister_acpi_hed_notifier(&ghes_notifier_sci);
950 mutex_unlock(&ghes_list_mutex);
952 case ACPI_HEST_NOTIFY_NMI:
953 mutex_lock(&ghes_list_mutex);
954 list_del_rcu(&ghes->list);
955 if (list_empty(&ghes_nmi))
956 unregister_die_notifier(&ghes_notifier_nmi);
957 mutex_unlock(&ghes_list_mutex);
959 * To synchronize with NMI handler, ghes can only be
960 * freed after NMI handler finishes.
963 len = ghes_esource_prealloc_size(generic);
964 ghes_estatus_pool_shrink(len);
974 platform_set_drvdata(ghes_dev, NULL);
979 static struct platform_driver ghes_platform_driver = {
982 .owner = THIS_MODULE,
985 .remove = ghes_remove,
988 static int __init ghes_init(void)
996 pr_info(GHES_PFX "HEST is not enabled!\n");
1001 pr_info(GHES_PFX "GHES is not enabled!\n");
1005 init_irq_work(&ghes_proc_irq_work, ghes_proc_in_irq);
1007 rc = ghes_ioremap_init();
1011 rc = ghes_estatus_pool_init();
1013 goto err_ioremap_exit;
1015 rc = ghes_estatus_pool_expand(GHES_ESTATUS_CACHE_AVG_SIZE *
1016 GHES_ESTATUS_CACHE_ALLOCED_MAX);
1020 rc = platform_driver_register(&ghes_platform_driver);
1024 rc = apei_osc_setup();
1025 if (rc == 0 && osc_sb_apei_support_acked)
1026 pr_info(GHES_PFX "APEI firmware first mode is enabled by APEI bit and WHEA _OSC.\n");
1027 else if (rc == 0 && !osc_sb_apei_support_acked)
1028 pr_info(GHES_PFX "APEI firmware first mode is enabled by WHEA _OSC.\n");
1029 else if (rc && osc_sb_apei_support_acked)
1030 pr_info(GHES_PFX "APEI firmware first mode is enabled by APEI bit.\n");
1032 pr_info(GHES_PFX "Failed to enable APEI firmware first mode.\n");
1036 ghes_estatus_pool_exit();
1038 ghes_ioremap_exit();
1043 static void __exit ghes_exit(void)
1045 platform_driver_unregister(&ghes_platform_driver);
1046 ghes_estatus_pool_exit();
1047 ghes_ioremap_exit();
1050 module_init(ghes_init);
1051 module_exit(ghes_exit);
1053 MODULE_AUTHOR("Huang Ying");
1054 MODULE_DESCRIPTION("APEI Generic Hardware Error Source support");
1055 MODULE_LICENSE("GPL");
1056 MODULE_ALIAS("platform:GHES");