2 * Machine check exception handling.
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.
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.
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.
18 * Copyright 2013 IBM Corporation
19 * Author: Mahesh Salgaonkar <mahesh@linux.vnet.ibm.com>
23 #define pr_fmt(fmt) "mce: " fmt
25 #include <linux/types.h>
26 #include <linux/ptrace.h>
27 #include <linux/percpu.h>
28 #include <linux/export.h>
29 #include <linux/irq_work.h>
32 static DEFINE_PER_CPU(int, mce_nest_count
);
33 static DEFINE_PER_CPU(struct machine_check_event
[MAX_MC_EVT
], mce_event
);
35 /* Queue for delayed MCE events. */
36 static DEFINE_PER_CPU(int, mce_queue_count
);
37 static DEFINE_PER_CPU(struct machine_check_event
[MAX_MC_EVT
], mce_event_queue
);
39 static void machine_check_process_queued_event(struct irq_work
*work
);
40 static struct irq_work mce_event_process_work
= {
41 .func
= machine_check_process_queued_event
,
44 static void mce_set_error_info(struct machine_check_event
*mce
,
45 struct mce_error_info
*mce_err
)
47 mce
->error_type
= mce_err
->error_type
;
48 switch (mce_err
->error_type
) {
49 case MCE_ERROR_TYPE_UE
:
50 mce
->u
.ue_error
.ue_error_type
= mce_err
->u
.ue_error_type
;
52 case MCE_ERROR_TYPE_SLB
:
53 mce
->u
.slb_error
.slb_error_type
= mce_err
->u
.slb_error_type
;
55 case MCE_ERROR_TYPE_ERAT
:
56 mce
->u
.erat_error
.erat_error_type
= mce_err
->u
.erat_error_type
;
58 case MCE_ERROR_TYPE_TLB
:
59 mce
->u
.tlb_error
.tlb_error_type
= mce_err
->u
.tlb_error_type
;
61 case MCE_ERROR_TYPE_UNKNOWN
:
68 * Decode and save high level MCE information into per cpu buffer which
69 * is an array of machine_check_event structure.
71 void save_mce_event(struct pt_regs
*regs
, long handled
,
72 struct mce_error_info
*mce_err
,
73 uint64_t nip
, uint64_t addr
)
75 int index
= __this_cpu_inc_return(mce_nest_count
) - 1;
76 struct machine_check_event
*mce
= this_cpu_ptr(&mce_event
[index
]);
79 * Return if we don't have enough space to log mce event.
80 * mce_nest_count may go beyond MAX_MC_EVT but that's ok,
81 * the check below will stop buffer overrun.
83 if (index
>= MAX_MC_EVT
)
86 /* Populate generic machine check info */
87 mce
->version
= MCE_V1
;
89 mce
->srr1
= regs
->msr
;
90 mce
->gpr3
= regs
->gpr
[3];
93 mce
->initiator
= MCE_INITIATOR_CPU
;
94 /* Mark it recovered if we have handled it and MSR(RI=1). */
95 if (handled
&& (regs
->msr
& MSR_RI
))
96 mce
->disposition
= MCE_DISPOSITION_RECOVERED
;
98 mce
->disposition
= MCE_DISPOSITION_NOT_RECOVERED
;
99 mce
->severity
= MCE_SEV_ERROR_SYNC
;
102 * Populate the mce error_type and type-specific error_type.
104 mce_set_error_info(mce
, mce_err
);
109 if (mce
->error_type
== MCE_ERROR_TYPE_TLB
) {
110 mce
->u
.tlb_error
.effective_address_provided
= true;
111 mce
->u
.tlb_error
.effective_address
= addr
;
112 } else if (mce
->error_type
== MCE_ERROR_TYPE_SLB
) {
113 mce
->u
.slb_error
.effective_address_provided
= true;
114 mce
->u
.slb_error
.effective_address
= addr
;
115 } else if (mce
->error_type
== MCE_ERROR_TYPE_ERAT
) {
116 mce
->u
.erat_error
.effective_address_provided
= true;
117 mce
->u
.erat_error
.effective_address
= addr
;
118 } else if (mce
->error_type
== MCE_ERROR_TYPE_UE
) {
119 mce
->u
.ue_error
.effective_address_provided
= true;
120 mce
->u
.ue_error
.effective_address
= addr
;
127 * mce Pointer to machine_check_event structure to be filled.
128 * release Flag to indicate whether to free the event slot or not.
129 * 0 <= do not release the mce event. Caller will invoke
130 * release_mce_event() once event has been consumed.
131 * 1 <= release the slot.
136 * get_mce_event() will be called by platform specific machine check
137 * handle routine and in KVM.
138 * When we call get_mce_event(), we are still in interrupt context and
139 * preemption will not be scheduled until ret_from_expect() routine
142 int get_mce_event(struct machine_check_event
*mce
, bool release
)
144 int index
= __this_cpu_read(mce_nest_count
) - 1;
145 struct machine_check_event
*mc_evt
;
152 /* Check if we have MCE info to process. */
153 if (index
< MAX_MC_EVT
) {
154 mc_evt
= this_cpu_ptr(&mce_event
[index
]);
155 /* Copy the event structure and release the original */
162 /* Decrement the count to free the slot. */
164 __this_cpu_dec(mce_nest_count
);
169 void release_mce_event(void)
171 get_mce_event(NULL
, true);
175 * Queue up the MCE event which then can be handled later.
177 void machine_check_queue_event(void)
180 struct machine_check_event evt
;
182 if (!get_mce_event(&evt
, MCE_EVENT_RELEASE
))
185 index
= __this_cpu_inc_return(mce_queue_count
) - 1;
186 /* If queue is full, just return for now. */
187 if (index
>= MAX_MC_EVT
) {
188 __this_cpu_dec(mce_queue_count
);
191 memcpy(this_cpu_ptr(&mce_event_queue
[index
]), &evt
, sizeof(evt
));
193 /* Queue irq work to process this event later. */
194 irq_work_queue(&mce_event_process_work
);
198 * process pending MCE event from the mce event queue. This function will be
199 * called during syscall exit.
201 static void machine_check_process_queued_event(struct irq_work
*work
)
206 * For now just print it to console.
207 * TODO: log this error event to FSP or nvram.
209 while (__this_cpu_read(mce_queue_count
) > 0) {
210 index
= __this_cpu_read(mce_queue_count
) - 1;
211 machine_check_print_event_info(
212 this_cpu_ptr(&mce_event_queue
[index
]));
213 __this_cpu_dec(mce_queue_count
);
217 void machine_check_print_event_info(struct machine_check_event
*evt
)
219 const char *level
, *sevstr
, *subtype
;
220 static const char *mc_ue_types
[] = {
223 "Page table walk ifetch",
225 "Page table walk Load/Store",
227 static const char *mc_slb_types
[] = {
232 static const char *mc_erat_types
[] = {
237 static const char *mc_tlb_types
[] = {
243 /* Print things out */
244 if (evt
->version
!= MCE_V1
) {
245 pr_err("Machine Check Exception, Unknown event version %d !\n",
249 switch (evt
->severity
) {
250 case MCE_SEV_NO_ERROR
:
254 case MCE_SEV_WARNING
:
255 level
= KERN_WARNING
;
258 case MCE_SEV_ERROR_SYNC
:
269 printk("%s%s Machine check interrupt [%s]\n", level
, sevstr
,
270 evt
->disposition
== MCE_DISPOSITION_RECOVERED
?
271 "Recovered" : "[Not recovered");
272 printk("%s Initiator: %s\n", level
,
273 evt
->initiator
== MCE_INITIATOR_CPU
? "CPU" : "Unknown");
274 switch (evt
->error_type
) {
275 case MCE_ERROR_TYPE_UE
:
276 subtype
= evt
->u
.ue_error
.ue_error_type
<
277 ARRAY_SIZE(mc_ue_types
) ?
278 mc_ue_types
[evt
->u
.ue_error
.ue_error_type
]
280 printk("%s Error type: UE [%s]\n", level
, subtype
);
281 if (evt
->u
.ue_error
.effective_address_provided
)
282 printk("%s Effective address: %016llx\n",
283 level
, evt
->u
.ue_error
.effective_address
);
284 if (evt
->u
.ue_error
.physical_address_provided
)
285 printk("%s Physical address: %016llx\n",
286 level
, evt
->u
.ue_error
.physical_address
);
288 case MCE_ERROR_TYPE_SLB
:
289 subtype
= evt
->u
.slb_error
.slb_error_type
<
290 ARRAY_SIZE(mc_slb_types
) ?
291 mc_slb_types
[evt
->u
.slb_error
.slb_error_type
]
293 printk("%s Error type: SLB [%s]\n", level
, subtype
);
294 if (evt
->u
.slb_error
.effective_address_provided
)
295 printk("%s Effective address: %016llx\n",
296 level
, evt
->u
.slb_error
.effective_address
);
298 case MCE_ERROR_TYPE_ERAT
:
299 subtype
= evt
->u
.erat_error
.erat_error_type
<
300 ARRAY_SIZE(mc_erat_types
) ?
301 mc_erat_types
[evt
->u
.erat_error
.erat_error_type
]
303 printk("%s Error type: ERAT [%s]\n", level
, subtype
);
304 if (evt
->u
.erat_error
.effective_address_provided
)
305 printk("%s Effective address: %016llx\n",
306 level
, evt
->u
.erat_error
.effective_address
);
308 case MCE_ERROR_TYPE_TLB
:
309 subtype
= evt
->u
.tlb_error
.tlb_error_type
<
310 ARRAY_SIZE(mc_tlb_types
) ?
311 mc_tlb_types
[evt
->u
.tlb_error
.tlb_error_type
]
313 printk("%s Error type: TLB [%s]\n", level
, subtype
);
314 if (evt
->u
.tlb_error
.effective_address_provided
)
315 printk("%s Effective address: %016llx\n",
316 level
, evt
->u
.tlb_error
.effective_address
);
319 case MCE_ERROR_TYPE_UNKNOWN
:
320 printk("%s Error type: Unknown\n", level
);
325 uint64_t get_mce_fault_addr(struct machine_check_event
*evt
)
327 switch (evt
->error_type
) {
328 case MCE_ERROR_TYPE_UE
:
329 if (evt
->u
.ue_error
.effective_address_provided
)
330 return evt
->u
.ue_error
.effective_address
;
332 case MCE_ERROR_TYPE_SLB
:
333 if (evt
->u
.slb_error
.effective_address_provided
)
334 return evt
->u
.slb_error
.effective_address
;
336 case MCE_ERROR_TYPE_ERAT
:
337 if (evt
->u
.erat_error
.effective_address_provided
)
338 return evt
->u
.erat_error
.effective_address
;
340 case MCE_ERROR_TYPE_TLB
:
341 if (evt
->u
.tlb_error
.effective_address_provided
)
342 return evt
->u
.tlb_error
.effective_address
;
345 case MCE_ERROR_TYPE_UNKNOWN
:
350 EXPORT_SYMBOL(get_mce_fault_addr
);