Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * File: mca_drv.c | |
3 | * Purpose: Generic MCA handling layer | |
4 | * | |
5 | * Copyright (C) 2004 FUJITSU LIMITED | |
6 | * Copyright (C) Hidetoshi Seto (seto.hidetoshi@jp.fujitsu.com) | |
7f613c7d KO |
7 | * Copyright (C) 2005 Silicon Graphics, Inc |
8 | * Copyright (C) 2005 Keith Owens <kaos@sgi.com> | |
d2a28ad9 | 9 | * Copyright (C) 2006 Russ Anderson <rja@sgi.com> |
1da177e4 | 10 | */ |
1da177e4 LT |
11 | #include <linux/types.h> |
12 | #include <linux/init.h> | |
13 | #include <linux/sched.h> | |
14 | #include <linux/interrupt.h> | |
15 | #include <linux/irq.h> | |
16 | #include <linux/kallsyms.h> | |
17 | #include <linux/smp_lock.h> | |
18 | #include <linux/bootmem.h> | |
19 | #include <linux/acpi.h> | |
20 | #include <linux/timer.h> | |
21 | #include <linux/module.h> | |
22 | #include <linux/kernel.h> | |
23 | #include <linux/smp.h> | |
24 | #include <linux/workqueue.h> | |
25 | #include <linux/mm.h> | |
26 | ||
27 | #include <asm/delay.h> | |
28 | #include <asm/machvec.h> | |
29 | #include <asm/page.h> | |
30 | #include <asm/ptrace.h> | |
31 | #include <asm/system.h> | |
32 | #include <asm/sal.h> | |
33 | #include <asm/mca.h> | |
34 | ||
35 | #include <asm/irq.h> | |
36 | #include <asm/hw_irq.h> | |
37 | ||
38 | #include "mca_drv.h" | |
39 | ||
40 | /* max size of SAL error record (default) */ | |
41 | static int sal_rec_max = 10000; | |
42 | ||
1da177e4 LT |
43 | /* from mca_drv_asm.S */ |
44 | extern void *mca_handler_bhhook(void); | |
45 | ||
46 | static DEFINE_SPINLOCK(mca_bh_lock); | |
47 | ||
48 | typedef enum { | |
49 | MCA_IS_LOCAL = 0, | |
50 | MCA_IS_GLOBAL = 1 | |
51 | } mca_type_t; | |
52 | ||
53 | #define MAX_PAGE_ISOLATE 1024 | |
54 | ||
55 | static struct page *page_isolate[MAX_PAGE_ISOLATE]; | |
56 | static int num_page_isolate = 0; | |
57 | ||
58 | typedef enum { | |
4881e2cd HS |
59 | ISOLATE_NG, |
60 | ISOLATE_OK, | |
61 | ISOLATE_NONE | |
1da177e4 LT |
62 | } isolate_status_t; |
63 | ||
18997961 RA |
64 | typedef enum { |
65 | MCA_NOT_RECOVERED = 0, | |
66 | MCA_RECOVERED = 1 | |
67 | } recovery_status_t; | |
68 | ||
1da177e4 LT |
69 | /* |
70 | * This pool keeps pointers to the section part of SAL error record | |
71 | */ | |
72 | static struct { | |
73 | slidx_list_t *buffer; /* section pointer list pool */ | |
74 | int cur_idx; /* Current index of section pointer list pool */ | |
75 | int max_idx; /* Maximum index of section pointer list pool */ | |
76 | } slidx_pool; | |
77 | ||
18997961 RA |
78 | static int |
79 | fatal_mca(const char *fmt, ...) | |
80 | { | |
81 | va_list args; | |
82 | ||
83 | va_start(args, fmt); | |
84 | vprintk(fmt, args); | |
85 | va_end(args); | |
86 | ||
87 | return MCA_NOT_RECOVERED; | |
88 | } | |
89 | ||
1da177e4 LT |
90 | /** |
91 | * mca_page_isolate - isolate a poisoned page in order not to use it later | |
92 | * @paddr: poisoned memory location | |
93 | * | |
94 | * Return value: | |
4881e2cd | 95 | * one of isolate_status_t, ISOLATE_OK/NG/NONE. |
1da177e4 LT |
96 | */ |
97 | ||
98 | static isolate_status_t | |
99 | mca_page_isolate(unsigned long paddr) | |
100 | { | |
101 | int i; | |
102 | struct page *p; | |
103 | ||
104 | /* whether physical address is valid or not */ | |
20305e59 | 105 | if (!ia64_phys_addr_valid(paddr)) |
4881e2cd HS |
106 | return ISOLATE_NONE; |
107 | ||
56f87b82 | 108 | if (!pfn_valid(paddr >> PAGE_SHIFT)) |
4881e2cd | 109 | return ISOLATE_NONE; |
1da177e4 LT |
110 | |
111 | /* convert physical address to physical page number */ | |
112 | p = pfn_to_page(paddr>>PAGE_SHIFT); | |
113 | ||
114 | /* check whether a page number have been already registered or not */ | |
20305e59 HS |
115 | for (i = 0; i < num_page_isolate; i++) |
116 | if (page_isolate[i] == p) | |
1da177e4 LT |
117 | return ISOLATE_OK; /* already listed */ |
118 | ||
119 | /* limitation check */ | |
20305e59 | 120 | if (num_page_isolate == MAX_PAGE_ISOLATE) |
1da177e4 LT |
121 | return ISOLATE_NG; |
122 | ||
123 | /* kick pages having attribute 'SLAB' or 'Reserved' */ | |
20305e59 | 124 | if (PageSlab(p) || PageReserved(p)) |
1da177e4 LT |
125 | return ISOLATE_NG; |
126 | ||
127 | /* add attribute 'Reserved' and register the page */ | |
cbb92144 | 128 | get_page(p); |
1da177e4 LT |
129 | SetPageReserved(p); |
130 | page_isolate[num_page_isolate++] = p; | |
131 | ||
132 | return ISOLATE_OK; | |
133 | } | |
134 | ||
135 | /** | |
136 | * mca_hanlder_bh - Kill the process which occurred memory read error | |
137 | * @paddr: poisoned address received from MCA Handler | |
138 | */ | |
139 | ||
140 | void | |
d2a28ad9 | 141 | mca_handler_bh(unsigned long paddr, void *iip, unsigned long ipsr) |
1da177e4 | 142 | { |
d2a28ad9 RA |
143 | printk(KERN_ERR "OS_MCA: process [cpu %d, pid: %d, uid: %d, " |
144 | "iip: %p, psr: 0x%lx,paddr: 0x%lx](%s) encounters MCA.\n", | |
145 | raw_smp_processor_id(), current->pid, current->uid, | |
146 | iip, ipsr, paddr, current->comm); | |
1da177e4 LT |
147 | |
148 | spin_lock(&mca_bh_lock); | |
4881e2cd HS |
149 | switch (mca_page_isolate(paddr)) { |
150 | case ISOLATE_OK: | |
1da177e4 | 151 | printk(KERN_DEBUG "Page isolation: ( %lx ) success.\n", paddr); |
4881e2cd HS |
152 | break; |
153 | case ISOLATE_NG: | |
ea0e92a6 | 154 | printk(KERN_CRIT "Page isolation: ( %lx ) failure.\n", paddr); |
4881e2cd HS |
155 | break; |
156 | default: | |
157 | break; | |
1da177e4 LT |
158 | } |
159 | spin_unlock(&mca_bh_lock); | |
160 | ||
161 | /* This process is about to be killed itself */ | |
b1b901c2 | 162 | do_exit(SIGKILL); |
1da177e4 LT |
163 | } |
164 | ||
165 | /** | |
166 | * mca_make_peidx - Make index of processor error section | |
167 | * @slpi: pointer to record of processor error section | |
168 | * @peidx: pointer to index of processor error section | |
169 | */ | |
170 | ||
20305e59 | 171 | static void |
1da177e4 LT |
172 | mca_make_peidx(sal_log_processor_info_t *slpi, peidx_table_t *peidx) |
173 | { | |
20305e59 | 174 | /* |
1da177e4 LT |
175 | * calculate the start address of |
176 | * "struct cpuid_info" and "sal_processor_static_info_t". | |
177 | */ | |
178 | u64 total_check_num = slpi->valid.num_cache_check | |
179 | + slpi->valid.num_tlb_check | |
180 | + slpi->valid.num_bus_check | |
181 | + slpi->valid.num_reg_file_check | |
182 | + slpi->valid.num_ms_check; | |
183 | u64 head_size = sizeof(sal_log_mod_error_info_t) * total_check_num | |
184 | + sizeof(sal_log_processor_info_t); | |
185 | u64 mid_size = slpi->valid.cpuid_info * sizeof(struct sal_cpuid_info); | |
186 | ||
187 | peidx_head(peidx) = slpi; | |
188 | peidx_mid(peidx) = (struct sal_cpuid_info *) | |
189 | (slpi->valid.cpuid_info ? ((char*)slpi + head_size) : NULL); | |
190 | peidx_bottom(peidx) = (sal_processor_static_info_t *) | |
191 | (slpi->valid.psi_static_struct ? | |
192 | ((char*)slpi + head_size + mid_size) : NULL); | |
193 | } | |
194 | ||
195 | /** | |
20305e59 | 196 | * mca_make_slidx - Make index of SAL error record |
1da177e4 LT |
197 | * @buffer: pointer to SAL error record |
198 | * @slidx: pointer to index of SAL error record | |
199 | * | |
200 | * Return value: | |
201 | * 1 if record has platform error / 0 if not | |
202 | */ | |
203 | #define LOG_INDEX_ADD_SECT_PTR(sect, ptr) \ | |
20305e59 HS |
204 | {slidx_list_t *hl = &slidx_pool.buffer[slidx_pool.cur_idx]; \ |
205 | hl->hdr = ptr; \ | |
206 | list_add(&hl->list, &(sect)); \ | |
207 | slidx_pool.cur_idx = (slidx_pool.cur_idx + 1)%slidx_pool.max_idx; } | |
1da177e4 | 208 | |
20305e59 | 209 | static int |
1da177e4 LT |
210 | mca_make_slidx(void *buffer, slidx_table_t *slidx) |
211 | { | |
212 | int platform_err = 0; | |
213 | int record_len = ((sal_log_record_header_t*)buffer)->len; | |
214 | u32 ercd_pos; | |
215 | int sects; | |
216 | sal_log_section_hdr_t *sp; | |
217 | ||
218 | /* | |
219 | * Initialize index referring current record | |
220 | */ | |
221 | INIT_LIST_HEAD(&(slidx->proc_err)); | |
222 | INIT_LIST_HEAD(&(slidx->mem_dev_err)); | |
223 | INIT_LIST_HEAD(&(slidx->sel_dev_err)); | |
224 | INIT_LIST_HEAD(&(slidx->pci_bus_err)); | |
225 | INIT_LIST_HEAD(&(slidx->smbios_dev_err)); | |
226 | INIT_LIST_HEAD(&(slidx->pci_comp_err)); | |
227 | INIT_LIST_HEAD(&(slidx->plat_specific_err)); | |
228 | INIT_LIST_HEAD(&(slidx->host_ctlr_err)); | |
229 | INIT_LIST_HEAD(&(slidx->plat_bus_err)); | |
230 | INIT_LIST_HEAD(&(slidx->unsupported)); | |
231 | ||
232 | /* | |
233 | * Extract a Record Header | |
234 | */ | |
235 | slidx->header = buffer; | |
236 | ||
237 | /* | |
238 | * Extract each section records | |
239 | * (arranged from "int ia64_log_platform_info_print()") | |
240 | */ | |
241 | for (ercd_pos = sizeof(sal_log_record_header_t), sects = 0; | |
242 | ercd_pos < record_len; ercd_pos += sp->len, sects++) { | |
243 | sp = (sal_log_section_hdr_t *)((char*)buffer + ercd_pos); | |
244 | if (!efi_guidcmp(sp->guid, SAL_PROC_DEV_ERR_SECT_GUID)) { | |
245 | LOG_INDEX_ADD_SECT_PTR(slidx->proc_err, sp); | |
20305e59 HS |
246 | } else if (!efi_guidcmp(sp->guid, |
247 | SAL_PLAT_MEM_DEV_ERR_SECT_GUID)) { | |
1da177e4 LT |
248 | platform_err = 1; |
249 | LOG_INDEX_ADD_SECT_PTR(slidx->mem_dev_err, sp); | |
20305e59 HS |
250 | } else if (!efi_guidcmp(sp->guid, |
251 | SAL_PLAT_SEL_DEV_ERR_SECT_GUID)) { | |
1da177e4 LT |
252 | platform_err = 1; |
253 | LOG_INDEX_ADD_SECT_PTR(slidx->sel_dev_err, sp); | |
20305e59 HS |
254 | } else if (!efi_guidcmp(sp->guid, |
255 | SAL_PLAT_PCI_BUS_ERR_SECT_GUID)) { | |
1da177e4 LT |
256 | platform_err = 1; |
257 | LOG_INDEX_ADD_SECT_PTR(slidx->pci_bus_err, sp); | |
20305e59 HS |
258 | } else if (!efi_guidcmp(sp->guid, |
259 | SAL_PLAT_SMBIOS_DEV_ERR_SECT_GUID)) { | |
1da177e4 LT |
260 | platform_err = 1; |
261 | LOG_INDEX_ADD_SECT_PTR(slidx->smbios_dev_err, sp); | |
20305e59 HS |
262 | } else if (!efi_guidcmp(sp->guid, |
263 | SAL_PLAT_PCI_COMP_ERR_SECT_GUID)) { | |
1da177e4 LT |
264 | platform_err = 1; |
265 | LOG_INDEX_ADD_SECT_PTR(slidx->pci_comp_err, sp); | |
20305e59 HS |
266 | } else if (!efi_guidcmp(sp->guid, |
267 | SAL_PLAT_SPECIFIC_ERR_SECT_GUID)) { | |
1da177e4 LT |
268 | platform_err = 1; |
269 | LOG_INDEX_ADD_SECT_PTR(slidx->plat_specific_err, sp); | |
20305e59 HS |
270 | } else if (!efi_guidcmp(sp->guid, |
271 | SAL_PLAT_HOST_CTLR_ERR_SECT_GUID)) { | |
1da177e4 LT |
272 | platform_err = 1; |
273 | LOG_INDEX_ADD_SECT_PTR(slidx->host_ctlr_err, sp); | |
20305e59 HS |
274 | } else if (!efi_guidcmp(sp->guid, |
275 | SAL_PLAT_BUS_ERR_SECT_GUID)) { | |
1da177e4 LT |
276 | platform_err = 1; |
277 | LOG_INDEX_ADD_SECT_PTR(slidx->plat_bus_err, sp); | |
278 | } else { | |
279 | LOG_INDEX_ADD_SECT_PTR(slidx->unsupported, sp); | |
280 | } | |
281 | } | |
282 | slidx->n_sections = sects; | |
283 | ||
284 | return platform_err; | |
285 | } | |
286 | ||
287 | /** | |
288 | * init_record_index_pools - Initialize pool of lists for SAL record index | |
289 | * | |
290 | * Return value: | |
291 | * 0 on Success / -ENOMEM on Failure | |
292 | */ | |
20305e59 | 293 | static int |
1da177e4 LT |
294 | init_record_index_pools(void) |
295 | { | |
296 | int i; | |
297 | int rec_max_size; /* Maximum size of SAL error records */ | |
298 | int sect_min_size; /* Minimum size of SAL error sections */ | |
299 | /* minimum size table of each section */ | |
20305e59 HS |
300 | static int sal_log_sect_min_sizes[] = { |
301 | sizeof(sal_log_processor_info_t) | |
302 | + sizeof(sal_processor_static_info_t), | |
1da177e4 LT |
303 | sizeof(sal_log_mem_dev_err_info_t), |
304 | sizeof(sal_log_sel_dev_err_info_t), | |
305 | sizeof(sal_log_pci_bus_err_info_t), | |
306 | sizeof(sal_log_smbios_dev_err_info_t), | |
307 | sizeof(sal_log_pci_comp_err_info_t), | |
308 | sizeof(sal_log_plat_specific_err_info_t), | |
309 | sizeof(sal_log_host_ctlr_err_info_t), | |
310 | sizeof(sal_log_plat_bus_err_info_t), | |
311 | }; | |
312 | ||
313 | /* | |
314 | * MCA handler cannot allocate new memory on flight, | |
315 | * so we preallocate enough memory to handle a SAL record. | |
316 | * | |
317 | * Initialize a handling set of slidx_pool: | |
318 | * 1. Pick up the max size of SAL error records | |
319 | * 2. Pick up the min size of SAL error sections | |
320 | * 3. Allocate the pool as enough to 2 SAL records | |
321 | * (now we can estimate the maxinum of section in a record.) | |
322 | */ | |
323 | ||
324 | /* - 1 - */ | |
325 | rec_max_size = sal_rec_max; | |
326 | ||
327 | /* - 2 - */ | |
328 | sect_min_size = sal_log_sect_min_sizes[0]; | |
329 | for (i = 1; i < sizeof sal_log_sect_min_sizes/sizeof(size_t); i++) | |
330 | if (sect_min_size > sal_log_sect_min_sizes[i]) | |
331 | sect_min_size = sal_log_sect_min_sizes[i]; | |
332 | ||
333 | /* - 3 - */ | |
334 | slidx_pool.max_idx = (rec_max_size/sect_min_size) * 2 + 1; | |
20305e59 HS |
335 | slidx_pool.buffer = (slidx_list_t *) |
336 | kmalloc(slidx_pool.max_idx * sizeof(slidx_list_t), GFP_KERNEL); | |
1da177e4 LT |
337 | |
338 | return slidx_pool.buffer ? 0 : -ENOMEM; | |
339 | } | |
340 | ||
341 | ||
342 | /***************************************************************************** | |
343 | * Recovery functions * | |
344 | *****************************************************************************/ | |
345 | ||
346 | /** | |
347 | * is_mca_global - Check whether this MCA is global or not | |
348 | * @peidx: pointer of index of processor error section | |
349 | * @pbci: pointer to pal_bus_check_info_t | |
20305e59 | 350 | * @sos: pointer to hand off struct between SAL and OS |
1da177e4 LT |
351 | * |
352 | * Return value: | |
353 | * MCA_IS_LOCAL / MCA_IS_GLOBAL | |
354 | */ | |
355 | ||
356 | static mca_type_t | |
7f613c7d KO |
357 | is_mca_global(peidx_table_t *peidx, pal_bus_check_info_t *pbci, |
358 | struct ia64_sal_os_state *sos) | |
1da177e4 | 359 | { |
20305e59 HS |
360 | pal_processor_state_info_t *psp = |
361 | (pal_processor_state_info_t*)peidx_psp(peidx); | |
1da177e4 | 362 | |
20305e59 | 363 | /* |
1da177e4 | 364 | * PAL can request a rendezvous, if the MCA has a global scope. |
20305e59 | 365 | * If "rz_always" flag is set, SAL requests MCA rendezvous |
1da177e4 LT |
366 | * in spite of global MCA. |
367 | * Therefore it is local MCA when rendezvous has not been requested. | |
368 | * Failed to rendezvous, the system must be down. | |
369 | */ | |
7f613c7d | 370 | switch (sos->rv_rc) { |
1da177e4 LT |
371 | case -1: /* SAL rendezvous unsuccessful */ |
372 | return MCA_IS_GLOBAL; | |
373 | case 0: /* SAL rendezvous not required */ | |
374 | return MCA_IS_LOCAL; | |
375 | case 1: /* SAL rendezvous successful int */ | |
376 | case 2: /* SAL rendezvous successful int with init */ | |
377 | default: | |
378 | break; | |
379 | } | |
380 | ||
381 | /* | |
382 | * If One or more Cache/TLB/Reg_File/Uarch_Check is here, | |
383 | * it would be a local MCA. (i.e. processor internal error) | |
384 | */ | |
385 | if (psp->tc || psp->cc || psp->rc || psp->uc) | |
386 | return MCA_IS_LOCAL; | |
387 | ||
388 | /* | |
389 | * Bus_Check structure with Bus_Check.ib (internal bus error) flag set | |
390 | * would be a global MCA. (e.g. a system bus address parity error) | |
391 | */ | |
392 | if (!pbci || pbci->ib) | |
393 | return MCA_IS_GLOBAL; | |
394 | ||
395 | /* | |
396 | * Bus_Check structure with Bus_Check.eb (external bus error) flag set | |
397 | * could be either a local MCA or a global MCA. | |
398 | * | |
399 | * Referring Bus_Check.bsi: | |
400 | * 0: Unknown/unclassified | |
401 | * 1: BERR# | |
402 | * 2: BINIT# | |
403 | * 3: Hard Fail | |
404 | * (FIXME: Are these SGI specific or generic bsi values?) | |
405 | */ | |
406 | if (pbci->eb) | |
407 | switch (pbci->bsi) { | |
408 | case 0: | |
409 | /* e.g. a load from poisoned memory */ | |
410 | return MCA_IS_LOCAL; | |
411 | case 1: | |
412 | case 2: | |
413 | case 3: | |
414 | return MCA_IS_GLOBAL; | |
415 | } | |
416 | ||
417 | return MCA_IS_GLOBAL; | |
418 | } | |
419 | ||
420 | /** | |
421 | * recover_from_read_error - Try to recover the errors which type are "read"s. | |
422 | * @slidx: pointer of index of SAL error record | |
423 | * @peidx: pointer of index of processor error section | |
424 | * @pbci: pointer of pal_bus_check_info | |
20305e59 | 425 | * @sos: pointer to hand off struct between SAL and OS |
1da177e4 LT |
426 | * |
427 | * Return value: | |
428 | * 1 on Success / 0 on Failure | |
429 | */ | |
430 | ||
431 | static int | |
20305e59 HS |
432 | recover_from_read_error(slidx_table_t *slidx, |
433 | peidx_table_t *peidx, pal_bus_check_info_t *pbci, | |
7f613c7d | 434 | struct ia64_sal_os_state *sos) |
1da177e4 LT |
435 | { |
436 | sal_log_mod_error_info_t *smei; | |
437 | pal_min_state_area_t *pmsa; | |
438 | struct ia64_psr *psr1, *psr2; | |
439 | ia64_fptr_t *mca_hdlr_bh = (ia64_fptr_t*)mca_handler_bhhook; | |
440 | ||
441 | /* Is target address valid? */ | |
442 | if (!pbci->tv) | |
18997961 | 443 | return fatal_mca(KERN_ALERT "MCA: target address not valid\n"); |
1da177e4 LT |
444 | |
445 | /* | |
446 | * cpu read or memory-mapped io read | |
447 | * | |
448 | * offending process affected process OS MCA do | |
449 | * kernel mode kernel mode down system | |
450 | * kernel mode user mode kill the process | |
451 | * user mode kernel mode down system (*) | |
452 | * user mode user mode kill the process | |
453 | * | |
454 | * (*) You could terminate offending user-mode process | |
455 | * if (pbci->pv && pbci->pl != 0) *and* if you sure | |
456 | * the process not have any locks of kernel. | |
457 | */ | |
458 | ||
a9474646 HS |
459 | /* Is minstate valid? */ |
460 | if (!peidx_bottom(peidx) || !(peidx_bottom(peidx)->valid.minstate)) | |
18997961 | 461 | return fatal_mca(KERN_ALERT "MCA: minstate not valid\n"); |
1da177e4 | 462 | psr1 =(struct ia64_psr *)&(peidx_minstate_area(peidx)->pmsa_ipsr); |
d2a28ad9 | 463 | psr2 =(struct ia64_psr *)&(peidx_minstate_area(peidx)->pmsa_xpsr); |
1da177e4 LT |
464 | |
465 | /* | |
466 | * Check the privilege level of interrupted context. | |
467 | * If it is user-mode, then terminate affected process. | |
468 | */ | |
d2a28ad9 RA |
469 | |
470 | pmsa = sos->pal_min_state; | |
471 | if (psr1->cpl != 0 || | |
472 | ((psr2->cpl != 0) && mca_recover_range(pmsa->pmsa_iip))) { | |
1da177e4 LT |
473 | smei = peidx_bus_check(peidx, 0); |
474 | if (smei->valid.target_identifier) { | |
475 | /* | |
476 | * setup for resume to bottom half of MCA, | |
477 | * "mca_handler_bhhook" | |
478 | */ | |
d2a28ad9 | 479 | /* pass to bhhook as argument (gr8, ...) */ |
1da177e4 | 480 | pmsa->pmsa_gr[8-1] = smei->target_identifier; |
d2a28ad9 RA |
481 | pmsa->pmsa_gr[9-1] = pmsa->pmsa_iip; |
482 | pmsa->pmsa_gr[10-1] = pmsa->pmsa_ipsr; | |
1da177e4 LT |
483 | /* set interrupted return address (but no use) */ |
484 | pmsa->pmsa_br0 = pmsa->pmsa_iip; | |
485 | /* change resume address to bottom half */ | |
486 | pmsa->pmsa_iip = mca_hdlr_bh->fp; | |
487 | pmsa->pmsa_gr[1-1] = mca_hdlr_bh->gp; | |
488 | /* set cpl with kernel mode */ | |
489 | psr2 = (struct ia64_psr *)&pmsa->pmsa_ipsr; | |
490 | psr2->cpl = 0; | |
491 | psr2->ri = 0; | |
d2a28ad9 | 492 | psr2->bn = 1; |
b1b901c2 | 493 | psr2->i = 0; |
1da177e4 | 494 | |
18997961 | 495 | return MCA_RECOVERED; |
1da177e4 LT |
496 | } |
497 | ||
498 | } | |
499 | ||
18997961 RA |
500 | return fatal_mca(KERN_ALERT "MCA: kernel context not recovered," |
501 | " iip 0x%lx\n", pmsa->pmsa_iip); | |
1da177e4 LT |
502 | } |
503 | ||
504 | /** | |
505 | * recover_from_platform_error - Recover from platform error. | |
506 | * @slidx: pointer of index of SAL error record | |
507 | * @peidx: pointer of index of processor error section | |
508 | * @pbci: pointer of pal_bus_check_info | |
20305e59 | 509 | * @sos: pointer to hand off struct between SAL and OS |
1da177e4 LT |
510 | * |
511 | * Return value: | |
512 | * 1 on Success / 0 on Failure | |
513 | */ | |
514 | ||
515 | static int | |
20305e59 HS |
516 | recover_from_platform_error(slidx_table_t *slidx, peidx_table_t *peidx, |
517 | pal_bus_check_info_t *pbci, | |
7f613c7d | 518 | struct ia64_sal_os_state *sos) |
1da177e4 LT |
519 | { |
520 | int status = 0; | |
20305e59 HS |
521 | pal_processor_state_info_t *psp = |
522 | (pal_processor_state_info_t*)peidx_psp(peidx); | |
1da177e4 LT |
523 | |
524 | if (psp->bc && pbci->eb && pbci->bsi == 0) { | |
525 | switch(pbci->type) { | |
526 | case 1: /* partial read */ | |
527 | case 3: /* full line(cpu) read */ | |
528 | case 9: /* I/O space read */ | |
20305e59 HS |
529 | status = recover_from_read_error(slidx, peidx, pbci, |
530 | sos); | |
1da177e4 LT |
531 | break; |
532 | case 0: /* unknown */ | |
533 | case 2: /* partial write */ | |
534 | case 4: /* full line write */ | |
535 | case 5: /* implicit or explicit write-back operation */ | |
536 | case 6: /* snoop probe */ | |
537 | case 7: /* incoming or outgoing ptc.g */ | |
538 | case 8: /* write coalescing transactions */ | |
539 | case 10: /* I/O space write */ | |
540 | case 11: /* inter-processor interrupt message(IPI) */ | |
20305e59 HS |
541 | case 12: /* interrupt acknowledge or |
542 | external task priority cycle */ | |
1da177e4 LT |
543 | default: |
544 | break; | |
545 | } | |
546 | } | |
547 | ||
548 | return status; | |
549 | } | |
550 | ||
551 | /** | |
552 | * recover_from_processor_error | |
553 | * @platform: whether there are some platform error section or not | |
554 | * @slidx: pointer of index of SAL error record | |
555 | * @peidx: pointer of index of processor error section | |
556 | * @pbci: pointer of pal_bus_check_info | |
20305e59 | 557 | * @sos: pointer to hand off struct between SAL and OS |
1da177e4 LT |
558 | * |
559 | * Return value: | |
560 | * 1 on Success / 0 on Failure | |
561 | */ | |
562 | /* | |
563 | * Later we try to recover when below all conditions are satisfied. | |
564 | * 1. Only one processor error section is exist. | |
565 | * 2. BUS_CHECK is exist and the others are not exist.(Except TLB_CHECK) | |
566 | * 3. The entry of BUS_CHECK_INFO is 1. | |
567 | * 4. "External bus error" flag is set and the others are not set. | |
568 | */ | |
569 | ||
570 | static int | |
20305e59 HS |
571 | recover_from_processor_error(int platform, slidx_table_t *slidx, |
572 | peidx_table_t *peidx, pal_bus_check_info_t *pbci, | |
7f613c7d | 573 | struct ia64_sal_os_state *sos) |
1da177e4 | 574 | { |
20305e59 HS |
575 | pal_processor_state_info_t *psp = |
576 | (pal_processor_state_info_t*)peidx_psp(peidx); | |
1da177e4 | 577 | |
20305e59 | 578 | /* |
a14f25a0 RA |
579 | * Processor recovery status must key off of the PAL recovery |
580 | * status in the Processor State Parameter. | |
1da177e4 | 581 | */ |
a14f25a0 RA |
582 | |
583 | /* | |
584 | * The machine check is corrected. | |
585 | */ | |
586 | if (psp->cm == 1) | |
18997961 | 587 | return MCA_RECOVERED; |
a14f25a0 RA |
588 | |
589 | /* | |
590 | * The error was not contained. Software must be reset. | |
591 | */ | |
592 | if (psp->us || psp->ci == 0) | |
18997961 | 593 | return fatal_mca(KERN_ALERT "MCA: error not contained\n"); |
1da177e4 LT |
594 | |
595 | /* | |
e1c48554 RA |
596 | * The cache check and bus check bits have four possible states |
597 | * cc bc | |
598 | * 0 0 Weird record, not recovered | |
599 | * 1 0 Cache error, not recovered | |
600 | * 0 1 I/O error, attempt recovery | |
601 | * 1 1 Memory error, attempt recovery | |
1da177e4 LT |
602 | */ |
603 | if (psp->bc == 0 || pbci == NULL) | |
18997961 | 604 | return fatal_mca(KERN_ALERT "MCA: No bus check\n"); |
1da177e4 LT |
605 | |
606 | /* | |
607 | * Sorry, we cannot handle so many. | |
608 | */ | |
609 | if (peidx_bus_check_num(peidx) > 1) | |
18997961 | 610 | return fatal_mca(KERN_ALERT "MCA: Too many bus checks\n"); |
1da177e4 LT |
611 | /* |
612 | * Well, here is only one bus error. | |
613 | */ | |
18997961 RA |
614 | if (pbci->ib) |
615 | return fatal_mca(KERN_ALERT "MCA: Internal Bus error\n"); | |
616 | if (pbci->cc) | |
617 | return fatal_mca(KERN_ALERT "MCA: Cache-cache error\n"); | |
1da177e4 | 618 | if (pbci->eb && pbci->bsi > 0) |
18997961 | 619 | return fatal_mca(KERN_ALERT "MCA: External bus check fatal status\n"); |
1da177e4 LT |
620 | |
621 | /* | |
622 | * This is a local MCA and estimated as recoverble external bus error. | |
623 | * (e.g. a load from poisoned memory) | |
624 | * This means "there are some platform errors". | |
625 | */ | |
20305e59 | 626 | if (platform) |
7f613c7d | 627 | return recover_from_platform_error(slidx, peidx, pbci, sos); |
20305e59 HS |
628 | /* |
629 | * On account of strange SAL error record, we cannot recover. | |
1da177e4 | 630 | */ |
18997961 | 631 | return fatal_mca(KERN_ALERT "MCA: Strange SAL record\n"); |
1da177e4 LT |
632 | } |
633 | ||
634 | /** | |
635 | * mca_try_to_recover - Try to recover from MCA | |
636 | * @rec: pointer to a SAL error record | |
20305e59 | 637 | * @sos: pointer to hand off struct between SAL and OS |
1da177e4 LT |
638 | * |
639 | * Return value: | |
640 | * 1 on Success / 0 on Failure | |
641 | */ | |
642 | ||
643 | static int | |
20305e59 | 644 | mca_try_to_recover(void *rec, struct ia64_sal_os_state *sos) |
1da177e4 LT |
645 | { |
646 | int platform_err; | |
647 | int n_proc_err; | |
648 | slidx_table_t slidx; | |
649 | peidx_table_t peidx; | |
650 | pal_bus_check_info_t pbci; | |
651 | ||
1da177e4 LT |
652 | /* Make index of SAL error record */ |
653 | platform_err = mca_make_slidx(rec, &slidx); | |
654 | ||
655 | /* Count processor error sections */ | |
656 | n_proc_err = slidx_count(&slidx, proc_err); | |
657 | ||
658 | /* Now, OS can recover when there is one processor error section */ | |
659 | if (n_proc_err > 1) | |
18997961 RA |
660 | return fatal_mca(KERN_ALERT "MCA: Too Many Errors\n"); |
661 | else if (n_proc_err == 0) | |
1da177e4 | 662 | /* Weird SAL record ... We need not to recover */ |
18997961 | 663 | return fatal_mca(KERN_ALERT "MCA: Weird SAL record\n"); |
1da177e4 LT |
664 | |
665 | /* Make index of processor error section */ | |
20305e59 HS |
666 | mca_make_peidx((sal_log_processor_info_t*) |
667 | slidx_first_entry(&slidx.proc_err)->hdr, &peidx); | |
1da177e4 LT |
668 | |
669 | /* Extract Processor BUS_CHECK[0] */ | |
670 | *((u64*)&pbci) = peidx_check_info(&peidx, bus_check, 0); | |
671 | ||
672 | /* Check whether MCA is global or not */ | |
7f613c7d | 673 | if (is_mca_global(&peidx, &pbci, sos)) |
18997961 | 674 | return fatal_mca(KERN_ALERT "MCA: global MCA\n"); |
1da177e4 LT |
675 | |
676 | /* Try to recover a processor error */ | |
20305e59 HS |
677 | return recover_from_processor_error(platform_err, &slidx, &peidx, |
678 | &pbci, sos); | |
1da177e4 LT |
679 | } |
680 | ||
681 | /* | |
682 | * ============================================================================= | |
683 | */ | |
684 | ||
685 | int __init mca_external_handler_init(void) | |
686 | { | |
687 | if (init_record_index_pools()) | |
688 | return -ENOMEM; | |
689 | ||
690 | /* register external mca handlers */ | |
20305e59 | 691 | if (ia64_reg_MCA_extension(mca_try_to_recover)) { |
1da177e4 LT |
692 | printk(KERN_ERR "ia64_reg_MCA_extension failed.\n"); |
693 | kfree(slidx_pool.buffer); | |
694 | return -EFAULT; | |
695 | } | |
696 | return 0; | |
697 | } | |
698 | ||
699 | void __exit mca_external_handler_exit(void) | |
700 | { | |
701 | /* unregister external mca handlers */ | |
702 | ia64_unreg_MCA_extension(); | |
703 | kfree(slidx_pool.buffer); | |
704 | } | |
705 | ||
706 | module_init(mca_external_handler_init); | |
707 | module_exit(mca_external_handler_exit); | |
708 | ||
709 | module_param(sal_rec_max, int, 0644); | |
710 | MODULE_PARM_DESC(sal_rec_max, "Max size of SAL error record"); | |
711 | ||
712 | MODULE_DESCRIPTION("ia64 platform dependent mca handler driver"); | |
713 | MODULE_LICENSE("GPL"); |