?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
thinkpad-acpi: log temperatures on termal alarm (v2)
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / drivers / s390 / s390mach.c
blob92b0417f8e12069403f1d50bbfb5653913b5a8f9
1 /*
2 * drivers/s390/s390mach.c
3 * S/390 machine check handler
4 *
5 * Copyright IBM Corp. 2000,2008
6 * Author(s): Ingo Adlung (adlung@de.ibm.com)
7 * Martin Schwidefsky (schwidefsky@de.ibm.com)
8 * Cornelia Huck <cornelia.huck@de.ibm.com>
9 */
10
11 #include <linux/init.h>
12 #include <linux/sched.h>
13 #include <linux/errno.h>
14 #include <linux/workqueue.h>
15 #include <linux/time.h>
16 #include <linux/device.h>
17 #include <linux/kthread.h>
18 #include <asm/etr.h>
19 #include <asm/lowcore.h>
20 #include <asm/cio.h>
21 #include <asm/cpu.h>
22 #include "s390mach.h"
23
24 static struct semaphore m_sem;
25
26 static NORET_TYPE void
27 s390_handle_damage(char *msg)
28 {
29 #ifdef CONFIG_SMP
30 smp_send_stop();
31 #endif
32 disabled_wait((unsigned long) __builtin_return_address(0));
33 for(;;);
34 }
35
36 static crw_handler_t crw_handlers[NR_RSCS];
37
38 /**
39 * s390_register_crw_handler() - register a channel report word handler
40 * @rsc: reporting source code to handle
41 * @handler: handler to be registered
42 *
43 * Returns %0 on success and a negative error value otherwise.
44 */
45 int s390_register_crw_handler(int rsc, crw_handler_t handler)
46 {
47 if ((rsc < 0) || (rsc >= NR_RSCS))
48 return -EINVAL;
49 if (!cmpxchg(&crw_handlers[rsc], NULL, handler))
50 return 0;
51 return -EBUSY;
52 }
53
54 /**
55 * s390_unregister_crw_handler() - unregister a channel report word handler
56 * @rsc: reporting source code to handle
57 */
58 void s390_unregister_crw_handler(int rsc)
59 {
60 if ((rsc < 0) || (rsc >= NR_RSCS))
61 return;
62 xchg(&crw_handlers[rsc], NULL);
63 synchronize_sched();
64 }
65
66 /*
67 * Retrieve CRWs and call function to handle event.
68 */
69 static int s390_collect_crw_info(void *param)
70 {
71 struct crw crw[2];
72 int ccode;
73 struct semaphore *sem;
74 unsigned int chain;
75 int ignore;
76
77 sem = (struct semaphore *)param;
78 repeat:
79 ignore = down_interruptible(sem);
80 chain = 0;
81 while (1) {
82 if (unlikely(chain > 1)) {
83 struct crw tmp_crw;
84
85 printk(KERN_WARNING"%s: Code does not support more "
86 "than two chained crws; please report to "
87 "linux390@de.ibm.com!\n", __func__);
88 ccode = stcrw(&tmp_crw);
89 printk(KERN_WARNING"%s: crw reports slct=%d, oflw=%d, "
90 "chn=%d, rsc=%X, anc=%d, erc=%X, rsid=%X\n",
91 __func__, tmp_crw.slct, tmp_crw.oflw,
92 tmp_crw.chn, tmp_crw.rsc, tmp_crw.anc,
93 tmp_crw.erc, tmp_crw.rsid);
94 printk(KERN_WARNING"%s: This was crw number %x in the "
95 "chain\n", __func__, chain);
96 if (ccode != 0)
97 break;
98 chain = tmp_crw.chn ? chain + 1 : 0;
99 continue;
100 }
101 ccode = stcrw(&crw[chain]);
102 if (ccode != 0)
103 break;
104 printk(KERN_DEBUG "crw_info : CRW reports slct=%d, oflw=%d, "
105 "chn=%d, rsc=%X, anc=%d, erc=%X, rsid=%X\n",
106 crw[chain].slct, crw[chain].oflw, crw[chain].chn,
107 crw[chain].rsc, crw[chain].anc, crw[chain].erc,
108 crw[chain].rsid);
109 /* Check for overflows. */
110 if (crw[chain].oflw) {
111 int i;
112
113 pr_debug("%s: crw overflow detected!\n", __func__);
114 for (i = 0; i < NR_RSCS; i++) {
115 if (crw_handlers[i])
116 crw_handlers[i](NULL, NULL, 1);
117 }
118 chain = 0;
119 continue;
120 }
121 if (crw[0].chn && !chain) {
122 chain++;
123 continue;
124 }
125 if (crw_handlers[crw[chain].rsc])
126 crw_handlers[crw[chain].rsc](&crw[0],
127 chain ? &crw[1] : NULL,
128 0);
129 /* chain is always 0 or 1 here. */
130 chain = crw[chain].chn ? chain + 1 : 0;
131 }
132 goto repeat;
133 return 0;
134 }
135
136 struct mcck_struct {
137 int kill_task;
138 int channel_report;
139 int warning;
140 unsigned long long mcck_code;
141 };
142
143 static DEFINE_PER_CPU(struct mcck_struct, cpu_mcck);
144
145 /*
146 * Main machine check handler function. Will be called with interrupts enabled
147 * or disabled and machine checks enabled or disabled.
148 */
149 void
150 s390_handle_mcck(void)
151 {
152 unsigned long flags;
153 struct mcck_struct mcck;
154
155 /*
156 * Disable machine checks and get the current state of accumulated
157 * machine checks. Afterwards delete the old state and enable machine
158 * checks again.
159 */
160 local_irq_save(flags);
161 local_mcck_disable();
162 mcck = __get_cpu_var(cpu_mcck);
163 memset(&__get_cpu_var(cpu_mcck), 0, sizeof(struct mcck_struct));
164 clear_thread_flag(TIF_MCCK_PENDING);
165 local_mcck_enable();
166 local_irq_restore(flags);
167
168 if (mcck.channel_report)
169 up(&m_sem);
170
171 #ifdef CONFIG_MACHCHK_WARNING
172 /*
173 * The warning may remain for a prolonged period on the bare iron.
174 * (actually till the machine is powered off, or until the problem is gone)
175 * So we just stop listening for the WARNING MCH and prevent continuously
176 * being interrupted. One caveat is however, that we must do this per
177 * processor and cannot use the smp version of ctl_clear_bit().
178 * On VM we only get one interrupt per virtally presented machinecheck.
179 * Though one suffices, we may get one interrupt per (virtual) processor.
180 */
181 if (mcck.warning) { /* WARNING pending ? */
182 static int mchchk_wng_posted = 0;
183 /*
184 * Use single machine clear, as we cannot handle smp right now
185 */
186 __ctl_clear_bit(14, 24); /* Disable WARNING MCH */
187 if (xchg(&mchchk_wng_posted, 1) == 0)
188 kill_cad_pid(SIGPWR, 1);
189 }
190 #endif
191
192 if (mcck.kill_task) {
193 local_irq_enable();
194 printk(KERN_EMERG "mcck: Terminating task because of machine "
195 "malfunction (code 0x%016llx).\n", mcck.mcck_code);
196 printk(KERN_EMERG "mcck: task: %s, pid: %d.\n",
197 current->comm, current->pid);
198 do_exit(SIGSEGV);
199 }
200 }
201 EXPORT_SYMBOL_GPL(s390_handle_mcck);
202
203 /*
204 * returns 0 if all registers could be validated
205 * returns 1 otherwise
206 */
207 static int
208 s390_revalidate_registers(struct mci *mci)
209 {
210 int kill_task;
211 u64 tmpclock;
212 u64 zero;
213 void *fpt_save_area, *fpt_creg_save_area;
214
215 kill_task = 0;
216 zero = 0;
217 /* General purpose registers */
218 if (!mci->gr)
219 /*
220 * General purpose registers couldn't be restored and have
221 * unknown contents. Process needs to be terminated.
222 */
223 kill_task = 1;
224
225 /* Revalidate floating point registers */
226 if (!mci->fp)
227 /*
228 * Floating point registers can't be restored and
229 * therefore the process needs to be terminated.
230 */
231 kill_task = 1;
232
233 #ifndef CONFIG_64BIT
234 asm volatile(
235 " ld 0,0(%0)\n"
236 " ld 2,8(%0)\n"
237 " ld 4,16(%0)\n"
238 " ld 6,24(%0)"
239 : : "a" (&S390_lowcore.floating_pt_save_area));
240 #endif
241
242 if (MACHINE_HAS_IEEE) {
243 #ifdef CONFIG_64BIT
244 fpt_save_area = &S390_lowcore.floating_pt_save_area;
245 fpt_creg_save_area = &S390_lowcore.fpt_creg_save_area;
246 #else
247 fpt_save_area = (void *) S390_lowcore.extended_save_area_addr;
248 fpt_creg_save_area = fpt_save_area+128;
249 #endif
250 /* Floating point control register */
251 if (!mci->fc) {
252 /*
253 * Floating point control register can't be restored.
254 * Task will be terminated.
255 */
256 asm volatile("lfpc 0(%0)" : : "a" (&zero), "m" (zero));
257 kill_task = 1;
258
259 } else
260 asm volatile("lfpc 0(%0)" : : "a" (fpt_creg_save_area));
261
262 asm volatile(
263 " ld 0,0(%0)\n"
264 " ld 1,8(%0)\n"
265 " ld 2,16(%0)\n"
266 " ld 3,24(%0)\n"
267 " ld 4,32(%0)\n"
268 " ld 5,40(%0)\n"
269 " ld 6,48(%0)\n"
270 " ld 7,56(%0)\n"
271 " ld 8,64(%0)\n"
272 " ld 9,72(%0)\n"
273 " ld 10,80(%0)\n"
274 " ld 11,88(%0)\n"
275 " ld 12,96(%0)\n"
276 " ld 13,104(%0)\n"
277 " ld 14,112(%0)\n"
278 " ld 15,120(%0)\n"
279 : : "a" (fpt_save_area));
280 }
281
282 /* Revalidate access registers */
283 asm volatile(
284 " lam 0,15,0(%0)"
285 : : "a" (&S390_lowcore.access_regs_save_area));
286 if (!mci->ar)
287 /*
288 * Access registers have unknown contents.
289 * Terminating task.
290 */
291 kill_task = 1;
292
293 /* Revalidate control registers */
294 if (!mci->cr)
295 /*
296 * Control registers have unknown contents.
297 * Can't recover and therefore stopping machine.
298 */
299 s390_handle_damage("invalid control registers.");
300 else
301 #ifdef CONFIG_64BIT
302 asm volatile(
303 " lctlg 0,15,0(%0)"
304 : : "a" (&S390_lowcore.cregs_save_area));
305 #else
306 asm volatile(
307 " lctl 0,15,0(%0)"
308 : : "a" (&S390_lowcore.cregs_save_area));
309 #endif
310
311 /*
312 * We don't even try to revalidate the TOD register, since we simply
313 * can't write something sensible into that register.
314 */
315
316 #ifdef CONFIG_64BIT
317 /*
318 * See if we can revalidate the TOD programmable register with its
319 * old contents (should be zero) otherwise set it to zero.
320 */
321 if (!mci->pr)
322 asm volatile(
323 " sr 0,0\n"
324 " sckpf"
325 : : : "0", "cc");
326 else
327 asm volatile(
328 " l 0,0(%0)\n"
329 " sckpf"
330 : : "a" (&S390_lowcore.tod_progreg_save_area)
331 : "0", "cc");
332 #endif
333
334 /* Revalidate clock comparator register */
335 asm volatile(
336 " stck 0(%1)\n"
337 " sckc 0(%1)"
338 : "=m" (tmpclock) : "a" (&(tmpclock)) : "cc", "memory");
339
340 /* Check if old PSW is valid */
341 if (!mci->wp)
342 /*
343 * Can't tell if we come from user or kernel mode
344 * -> stopping machine.
345 */
346 s390_handle_damage("old psw invalid.");
347
348 if (!mci->ms || !mci->pm || !mci->ia)
349 kill_task = 1;
350
351 return kill_task;
352 }
353
354 #define MAX_IPD_COUNT 29
355 #define MAX_IPD_TIME (5 * 60 * USEC_PER_SEC) /* 5 minutes */
356
357 /*
358 * machine check handler.
359 */
360 void
361 s390_do_machine_check(struct pt_regs *regs)
362 {
363 static DEFINE_SPINLOCK(ipd_lock);
364 static unsigned long long last_ipd;
365 static int ipd_count;
366 unsigned long long tmp;
367 struct mci *mci;
368 struct mcck_struct *mcck;
369 int umode;
370
371 lockdep_off();
372
373 s390_idle_check();
374
375 mci = (struct mci *) &S390_lowcore.mcck_interruption_code;
376 mcck = &__get_cpu_var(cpu_mcck);
377 umode = user_mode(regs);
378
379 if (mci->sd)
380 /* System damage -> stopping machine */
381 s390_handle_damage("received system damage machine check.");
382
383 if (mci->pd) {
384 if (mci->b) {
385 /* Processing backup -> verify if we can survive this */
386 u64 z_mcic, o_mcic, t_mcic;
387 #ifdef CONFIG_64BIT
388 z_mcic = (1ULL<<63 | 1ULL<<59 | 1ULL<<29);
389 o_mcic = (1ULL<<43 | 1ULL<<42 | 1ULL<<41 | 1ULL<<40 |
390 1ULL<<36 | 1ULL<<35 | 1ULL<<34 | 1ULL<<32 |
391 1ULL<<30 | 1ULL<<21 | 1ULL<<20 | 1ULL<<17 |
392 1ULL<<16);
393 #else
394 z_mcic = (1ULL<<63 | 1ULL<<59 | 1ULL<<57 | 1ULL<<50 |
395 1ULL<<29);
396 o_mcic = (1ULL<<43 | 1ULL<<42 | 1ULL<<41 | 1ULL<<40 |
397 1ULL<<36 | 1ULL<<35 | 1ULL<<34 | 1ULL<<32 |
398 1ULL<<30 | 1ULL<<20 | 1ULL<<17 | 1ULL<<16);
399 #endif
400 t_mcic = *(u64 *)mci;
401
402 if (((t_mcic & z_mcic) != 0) ||
403 ((t_mcic & o_mcic) != o_mcic)) {
404 s390_handle_damage("processing backup machine "
405 "check with damage.");
406 }
407
408 /*
409 * Nullifying exigent condition, therefore we might
410 * retry this instruction.
411 */
412
413 spin_lock(&ipd_lock);
414
415 tmp = get_clock();
416
417 if (((tmp - last_ipd) >> 12) < MAX_IPD_TIME)
418 ipd_count++;
419 else
420 ipd_count = 1;
421
422 last_ipd = tmp;
423
424 if (ipd_count == MAX_IPD_COUNT)
425 s390_handle_damage("too many ipd retries.");
426
427 spin_unlock(&ipd_lock);
428 }
429 else {
430 /* Processing damage -> stopping machine */
431 s390_handle_damage("received instruction processing "
432 "damage machine check.");
433 }
434 }
435 if (s390_revalidate_registers(mci)) {
436 if (umode) {
437 /*
438 * Couldn't restore all register contents while in
439 * user mode -> mark task for termination.
440 */
441 mcck->kill_task = 1;
442 mcck->mcck_code = *(unsigned long long *) mci;
443 set_thread_flag(TIF_MCCK_PENDING);
444 }
445 else
446 /*
447 * Couldn't restore all register contents while in
448 * kernel mode -> stopping machine.
449 */
450 s390_handle_damage("unable to revalidate registers.");
451 }
452
453 if (mci->cd) {
454 /* Timing facility damage */
455 s390_handle_damage("TOD clock damaged");
456 }
457
458 if (mci->ed && mci->ec) {
459 /* External damage */
460 if (S390_lowcore.external_damage_code & (1U << ED_ETR_SYNC))
461 etr_sync_check();
462 if (S390_lowcore.external_damage_code & (1U << ED_ETR_SWITCH))
463 etr_switch_to_local();
464 if (S390_lowcore.external_damage_code & (1U << ED_STP_SYNC))
465 stp_sync_check();
466 if (S390_lowcore.external_damage_code & (1U << ED_STP_ISLAND))
467 stp_island_check();
468 }
469
470 if (mci->se)
471 /* Storage error uncorrected */
472 s390_handle_damage("received storage error uncorrected "
473 "machine check.");
474
475 if (mci->ke)
476 /* Storage key-error uncorrected */
477 s390_handle_damage("received storage key-error uncorrected "
478 "machine check.");
479
480 if (mci->ds && mci->fa)
481 /* Storage degradation */
482 s390_handle_damage("received storage degradation machine "
483 "check.");
484
485 if (mci->cp) {
486 /* Channel report word pending */
487 mcck->channel_report = 1;
488 set_thread_flag(TIF_MCCK_PENDING);
489 }
490
491 if (mci->w) {
492 /* Warning pending */
493 mcck->warning = 1;
494 set_thread_flag(TIF_MCCK_PENDING);
495 }
496 lockdep_on();
497 }
498
499 /*
500 * s390_init_machine_check
501 *
502 * initialize machine check handling
503 */
504 static int
505 machine_check_init(void)
506 {
507 init_MUTEX_LOCKED(&m_sem);
508 ctl_set_bit(14, 25); /* enable external damage MCH */
509 ctl_set_bit(14, 27); /* enable system recovery MCH */
510 #ifdef CONFIG_MACHCHK_WARNING
511 ctl_set_bit(14, 24); /* enable warning MCH */
512 #endif
513 return 0;
514 }
515
516 /*
517 * Initialize the machine check handler really early to be able to
518 * catch all machine checks that happen during boot
519 */
520 arch_initcall(machine_check_init);
521
522 /*
523 * Machine checks for the channel subsystem must be enabled
524 * after the channel subsystem is initialized
525 */
526 static int __init
527 machine_check_crw_init (void)
528 {
529 struct task_struct *task;
530
531 task = kthread_run(s390_collect_crw_info, &m_sem, "kmcheck");
532 if (IS_ERR(task))
533 return PTR_ERR(task);
534 ctl_set_bit(14, 28); /* enable channel report MCH */
535 return 0;
536 }
537
538 device_initcall (machine_check_crw_init);
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree