3 * Purpose: Generic MCA handling layer
5 * Copyright (C) 2004 FUJITSU LIMITED
6 * Copyright (C) Hidetoshi Seto (seto.hidetoshi@jp.fujitsu.com)
7 * Copyright (C) 2005 Silicon Graphics, Inc
8 * Copyright (C) 2005 Keith Owens <kaos@sgi.com>
10 #include <linux/config.h>
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>
27 #include <asm/delay.h>
28 #include <asm/machvec.h>
30 #include <asm/ptrace.h>
31 #include <asm/system.h>
36 #include <asm/hw_irq.h>
40 /* max size of SAL error record (default) */
41 static int sal_rec_max
= 10000;
43 /* from mca_drv_asm.S */
44 extern void *mca_handler_bhhook(void);
46 static DEFINE_SPINLOCK(mca_bh_lock
);
53 #define MAX_PAGE_ISOLATE 1024
55 static struct page
*page_isolate
[MAX_PAGE_ISOLATE
];
56 static int num_page_isolate
= 0;
64 * This pool keeps pointers to the section part of SAL error record
67 slidx_list_t
*buffer
; /* section pointer list pool */
68 int cur_idx
; /* Current index of section pointer list pool */
69 int max_idx
; /* Maximum index of section pointer list pool */
73 * mca_page_isolate - isolate a poisoned page in order not to use it later
74 * @paddr: poisoned memory location
77 * ISOLATE_OK / ISOLATE_NG
80 static isolate_status_t
81 mca_page_isolate(unsigned long paddr
)
86 /* whether physical address is valid or not */
87 if (!ia64_phys_addr_valid(paddr
))
90 /* convert physical address to physical page number */
91 p
= pfn_to_page(paddr
>>PAGE_SHIFT
);
93 /* check whether a page number have been already registered or not */
94 for (i
= 0; i
< num_page_isolate
; i
++)
95 if (page_isolate
[i
] == p
)
96 return ISOLATE_OK
; /* already listed */
98 /* limitation check */
99 if (num_page_isolate
== MAX_PAGE_ISOLATE
)
102 /* kick pages having attribute 'SLAB' or 'Reserved' */
103 if (PageSlab(p
) || PageReserved(p
))
106 /* add attribute 'Reserved' and register the page */
108 page_isolate
[num_page_isolate
++] = p
;
114 * mca_hanlder_bh - Kill the process which occurred memory read error
115 * @paddr: poisoned address received from MCA Handler
119 mca_handler_bh(unsigned long paddr
)
121 printk(KERN_DEBUG
"OS_MCA: process [pid: %d](%s) encounters MCA.\n",
122 current
->pid
, current
->comm
);
124 spin_lock(&mca_bh_lock
);
125 if (mca_page_isolate(paddr
) == ISOLATE_OK
) {
126 printk(KERN_DEBUG
"Page isolation: ( %lx ) success.\n", paddr
);
128 printk(KERN_DEBUG
"Page isolation: ( %lx ) failure.\n", paddr
);
130 spin_unlock(&mca_bh_lock
);
132 /* This process is about to be killed itself */
137 * mca_make_peidx - Make index of processor error section
138 * @slpi: pointer to record of processor error section
139 * @peidx: pointer to index of processor error section
143 mca_make_peidx(sal_log_processor_info_t
*slpi
, peidx_table_t
*peidx
)
146 * calculate the start address of
147 * "struct cpuid_info" and "sal_processor_static_info_t".
149 u64 total_check_num
= slpi
->valid
.num_cache_check
150 + slpi
->valid
.num_tlb_check
151 + slpi
->valid
.num_bus_check
152 + slpi
->valid
.num_reg_file_check
153 + slpi
->valid
.num_ms_check
;
154 u64 head_size
= sizeof(sal_log_mod_error_info_t
) * total_check_num
155 + sizeof(sal_log_processor_info_t
);
156 u64 mid_size
= slpi
->valid
.cpuid_info
* sizeof(struct sal_cpuid_info
);
158 peidx_head(peidx
) = slpi
;
159 peidx_mid(peidx
) = (struct sal_cpuid_info
*)
160 (slpi
->valid
.cpuid_info
? ((char*)slpi
+ head_size
) : NULL
);
161 peidx_bottom(peidx
) = (sal_processor_static_info_t
*)
162 (slpi
->valid
.psi_static_struct
?
163 ((char*)slpi
+ head_size
+ mid_size
) : NULL
);
167 * mca_make_slidx - Make index of SAL error record
168 * @buffer: pointer to SAL error record
169 * @slidx: pointer to index of SAL error record
172 * 1 if record has platform error / 0 if not
174 #define LOG_INDEX_ADD_SECT_PTR(sect, ptr) \
175 {slidx_list_t *hl = &slidx_pool.buffer[slidx_pool.cur_idx]; \
177 list_add(&hl->list, &(sect)); \
178 slidx_pool.cur_idx = (slidx_pool.cur_idx + 1)%slidx_pool.max_idx; }
181 mca_make_slidx(void *buffer
, slidx_table_t
*slidx
)
183 int platform_err
= 0;
184 int record_len
= ((sal_log_record_header_t
*)buffer
)->len
;
187 sal_log_section_hdr_t
*sp
;
190 * Initialize index referring current record
192 INIT_LIST_HEAD(&(slidx
->proc_err
));
193 INIT_LIST_HEAD(&(slidx
->mem_dev_err
));
194 INIT_LIST_HEAD(&(slidx
->sel_dev_err
));
195 INIT_LIST_HEAD(&(slidx
->pci_bus_err
));
196 INIT_LIST_HEAD(&(slidx
->smbios_dev_err
));
197 INIT_LIST_HEAD(&(slidx
->pci_comp_err
));
198 INIT_LIST_HEAD(&(slidx
->plat_specific_err
));
199 INIT_LIST_HEAD(&(slidx
->host_ctlr_err
));
200 INIT_LIST_HEAD(&(slidx
->plat_bus_err
));
201 INIT_LIST_HEAD(&(slidx
->unsupported
));
204 * Extract a Record Header
206 slidx
->header
= buffer
;
209 * Extract each section records
210 * (arranged from "int ia64_log_platform_info_print()")
212 for (ercd_pos
= sizeof(sal_log_record_header_t
), sects
= 0;
213 ercd_pos
< record_len
; ercd_pos
+= sp
->len
, sects
++) {
214 sp
= (sal_log_section_hdr_t
*)((char*)buffer
+ ercd_pos
);
215 if (!efi_guidcmp(sp
->guid
, SAL_PROC_DEV_ERR_SECT_GUID
)) {
216 LOG_INDEX_ADD_SECT_PTR(slidx
->proc_err
, sp
);
217 } else if (!efi_guidcmp(sp
->guid
,
218 SAL_PLAT_MEM_DEV_ERR_SECT_GUID
)) {
220 LOG_INDEX_ADD_SECT_PTR(slidx
->mem_dev_err
, sp
);
221 } else if (!efi_guidcmp(sp
->guid
,
222 SAL_PLAT_SEL_DEV_ERR_SECT_GUID
)) {
224 LOG_INDEX_ADD_SECT_PTR(slidx
->sel_dev_err
, sp
);
225 } else if (!efi_guidcmp(sp
->guid
,
226 SAL_PLAT_PCI_BUS_ERR_SECT_GUID
)) {
228 LOG_INDEX_ADD_SECT_PTR(slidx
->pci_bus_err
, sp
);
229 } else if (!efi_guidcmp(sp
->guid
,
230 SAL_PLAT_SMBIOS_DEV_ERR_SECT_GUID
)) {
232 LOG_INDEX_ADD_SECT_PTR(slidx
->smbios_dev_err
, sp
);
233 } else if (!efi_guidcmp(sp
->guid
,
234 SAL_PLAT_PCI_COMP_ERR_SECT_GUID
)) {
236 LOG_INDEX_ADD_SECT_PTR(slidx
->pci_comp_err
, sp
);
237 } else if (!efi_guidcmp(sp
->guid
,
238 SAL_PLAT_SPECIFIC_ERR_SECT_GUID
)) {
240 LOG_INDEX_ADD_SECT_PTR(slidx
->plat_specific_err
, sp
);
241 } else if (!efi_guidcmp(sp
->guid
,
242 SAL_PLAT_HOST_CTLR_ERR_SECT_GUID
)) {
244 LOG_INDEX_ADD_SECT_PTR(slidx
->host_ctlr_err
, sp
);
245 } else if (!efi_guidcmp(sp
->guid
,
246 SAL_PLAT_BUS_ERR_SECT_GUID
)) {
248 LOG_INDEX_ADD_SECT_PTR(slidx
->plat_bus_err
, sp
);
250 LOG_INDEX_ADD_SECT_PTR(slidx
->unsupported
, sp
);
253 slidx
->n_sections
= sects
;
259 * init_record_index_pools - Initialize pool of lists for SAL record index
262 * 0 on Success / -ENOMEM on Failure
265 init_record_index_pools(void)
268 int rec_max_size
; /* Maximum size of SAL error records */
269 int sect_min_size
; /* Minimum size of SAL error sections */
270 /* minimum size table of each section */
271 static int sal_log_sect_min_sizes
[] = {
272 sizeof(sal_log_processor_info_t
)
273 + sizeof(sal_processor_static_info_t
),
274 sizeof(sal_log_mem_dev_err_info_t
),
275 sizeof(sal_log_sel_dev_err_info_t
),
276 sizeof(sal_log_pci_bus_err_info_t
),
277 sizeof(sal_log_smbios_dev_err_info_t
),
278 sizeof(sal_log_pci_comp_err_info_t
),
279 sizeof(sal_log_plat_specific_err_info_t
),
280 sizeof(sal_log_host_ctlr_err_info_t
),
281 sizeof(sal_log_plat_bus_err_info_t
),
285 * MCA handler cannot allocate new memory on flight,
286 * so we preallocate enough memory to handle a SAL record.
288 * Initialize a handling set of slidx_pool:
289 * 1. Pick up the max size of SAL error records
290 * 2. Pick up the min size of SAL error sections
291 * 3. Allocate the pool as enough to 2 SAL records
292 * (now we can estimate the maxinum of section in a record.)
296 rec_max_size
= sal_rec_max
;
299 sect_min_size
= sal_log_sect_min_sizes
[0];
300 for (i
= 1; i
< sizeof sal_log_sect_min_sizes
/sizeof(size_t); i
++)
301 if (sect_min_size
> sal_log_sect_min_sizes
[i
])
302 sect_min_size
= sal_log_sect_min_sizes
[i
];
305 slidx_pool
.max_idx
= (rec_max_size
/sect_min_size
) * 2 + 1;
306 slidx_pool
.buffer
= (slidx_list_t
*)
307 kmalloc(slidx_pool
.max_idx
* sizeof(slidx_list_t
), GFP_KERNEL
);
309 return slidx_pool
.buffer
? 0 : -ENOMEM
;
313 /*****************************************************************************
314 * Recovery functions *
315 *****************************************************************************/
318 * is_mca_global - Check whether this MCA is global or not
319 * @peidx: pointer of index of processor error section
320 * @pbci: pointer to pal_bus_check_info_t
321 * @sos: pointer to hand off struct between SAL and OS
324 * MCA_IS_LOCAL / MCA_IS_GLOBAL
328 is_mca_global(peidx_table_t
*peidx
, pal_bus_check_info_t
*pbci
,
329 struct ia64_sal_os_state
*sos
)
331 pal_processor_state_info_t
*psp
=
332 (pal_processor_state_info_t
*)peidx_psp(peidx
);
335 * PAL can request a rendezvous, if the MCA has a global scope.
336 * If "rz_always" flag is set, SAL requests MCA rendezvous
337 * in spite of global MCA.
338 * Therefore it is local MCA when rendezvous has not been requested.
339 * Failed to rendezvous, the system must be down.
341 switch (sos
->rv_rc
) {
342 case -1: /* SAL rendezvous unsuccessful */
343 return MCA_IS_GLOBAL
;
344 case 0: /* SAL rendezvous not required */
346 case 1: /* SAL rendezvous successful int */
347 case 2: /* SAL rendezvous successful int with init */
353 * If One or more Cache/TLB/Reg_File/Uarch_Check is here,
354 * it would be a local MCA. (i.e. processor internal error)
356 if (psp
->tc
|| psp
->cc
|| psp
->rc
|| psp
->uc
)
360 * Bus_Check structure with Bus_Check.ib (internal bus error) flag set
361 * would be a global MCA. (e.g. a system bus address parity error)
363 if (!pbci
|| pbci
->ib
)
364 return MCA_IS_GLOBAL
;
367 * Bus_Check structure with Bus_Check.eb (external bus error) flag set
368 * could be either a local MCA or a global MCA.
370 * Referring Bus_Check.bsi:
371 * 0: Unknown/unclassified
375 * (FIXME: Are these SGI specific or generic bsi values?)
380 /* e.g. a load from poisoned memory */
385 return MCA_IS_GLOBAL
;
388 return MCA_IS_GLOBAL
;
392 * recover_from_read_error - Try to recover the errors which type are "read"s.
393 * @slidx: pointer of index of SAL error record
394 * @peidx: pointer of index of processor error section
395 * @pbci: pointer of pal_bus_check_info
396 * @sos: pointer to hand off struct between SAL and OS
399 * 1 on Success / 0 on Failure
403 recover_from_read_error(slidx_table_t
*slidx
,
404 peidx_table_t
*peidx
, pal_bus_check_info_t
*pbci
,
405 struct ia64_sal_os_state
*sos
)
407 sal_log_mod_error_info_t
*smei
;
408 pal_min_state_area_t
*pmsa
;
409 struct ia64_psr
*psr1
, *psr2
;
410 ia64_fptr_t
*mca_hdlr_bh
= (ia64_fptr_t
*)mca_handler_bhhook
;
412 /* Is target address valid? */
417 * cpu read or memory-mapped io read
419 * offending process affected process OS MCA do
420 * kernel mode kernel mode down system
421 * kernel mode user mode kill the process
422 * user mode kernel mode down system (*)
423 * user mode user mode kill the process
425 * (*) You could terminate offending user-mode process
426 * if (pbci->pv && pbci->pl != 0) *and* if you sure
427 * the process not have any locks of kernel.
430 psr1
=(struct ia64_psr
*)&(peidx_minstate_area(peidx
)->pmsa_ipsr
);
433 * Check the privilege level of interrupted context.
434 * If it is user-mode, then terminate affected process.
436 if (psr1
->cpl
!= 0) {
437 smei
= peidx_bus_check(peidx
, 0);
438 if (smei
->valid
.target_identifier
) {
440 * setup for resume to bottom half of MCA,
441 * "mca_handler_bhhook"
443 pmsa
= sos
->pal_min_state
;
444 /* pass to bhhook as 1st argument (gr8) */
445 pmsa
->pmsa_gr
[8-1] = smei
->target_identifier
;
446 /* set interrupted return address (but no use) */
447 pmsa
->pmsa_br0
= pmsa
->pmsa_iip
;
448 /* change resume address to bottom half */
449 pmsa
->pmsa_iip
= mca_hdlr_bh
->fp
;
450 pmsa
->pmsa_gr
[1-1] = mca_hdlr_bh
->gp
;
451 /* set cpl with kernel mode */
452 psr2
= (struct ia64_psr
*)&pmsa
->pmsa_ipsr
;
466 * recover_from_platform_error - Recover from platform error.
467 * @slidx: pointer of index of SAL error record
468 * @peidx: pointer of index of processor error section
469 * @pbci: pointer of pal_bus_check_info
470 * @sos: pointer to hand off struct between SAL and OS
473 * 1 on Success / 0 on Failure
477 recover_from_platform_error(slidx_table_t
*slidx
, peidx_table_t
*peidx
,
478 pal_bus_check_info_t
*pbci
,
479 struct ia64_sal_os_state
*sos
)
482 pal_processor_state_info_t
*psp
=
483 (pal_processor_state_info_t
*)peidx_psp(peidx
);
485 if (psp
->bc
&& pbci
->eb
&& pbci
->bsi
== 0) {
487 case 1: /* partial read */
488 case 3: /* full line(cpu) read */
489 case 9: /* I/O space read */
490 status
= recover_from_read_error(slidx
, peidx
, pbci
,
493 case 0: /* unknown */
494 case 2: /* partial write */
495 case 4: /* full line write */
496 case 5: /* implicit or explicit write-back operation */
497 case 6: /* snoop probe */
498 case 7: /* incoming or outgoing ptc.g */
499 case 8: /* write coalescing transactions */
500 case 10: /* I/O space write */
501 case 11: /* inter-processor interrupt message(IPI) */
502 case 12: /* interrupt acknowledge or
503 external task priority cycle */
513 * recover_from_processor_error
514 * @platform: whether there are some platform error section or not
515 * @slidx: pointer of index of SAL error record
516 * @peidx: pointer of index of processor error section
517 * @pbci: pointer of pal_bus_check_info
518 * @sos: pointer to hand off struct between SAL and OS
521 * 1 on Success / 0 on Failure
524 * Later we try to recover when below all conditions are satisfied.
525 * 1. Only one processor error section is exist.
526 * 2. BUS_CHECK is exist and the others are not exist.(Except TLB_CHECK)
527 * 3. The entry of BUS_CHECK_INFO is 1.
528 * 4. "External bus error" flag is set and the others are not set.
532 recover_from_processor_error(int platform
, slidx_table_t
*slidx
,
533 peidx_table_t
*peidx
, pal_bus_check_info_t
*pbci
,
534 struct ia64_sal_os_state
*sos
)
536 pal_processor_state_info_t
*psp
=
537 (pal_processor_state_info_t
*)peidx_psp(peidx
);
540 * We cannot recover errors with other than bus_check.
542 if (psp
->cc
|| psp
->rc
|| psp
->uc
)
546 * If there is no bus error, record is weird but we need not to recover.
548 if (psp
->bc
== 0 || pbci
== NULL
)
552 * Sorry, we cannot handle so many.
554 if (peidx_bus_check_num(peidx
) > 1)
557 * Well, here is only one bus error.
559 if (pbci
->ib
|| pbci
->cc
)
561 if (pbci
->eb
&& pbci
->bsi
> 0)
567 * This is a local MCA and estimated as recoverble external bus error.
568 * (e.g. a load from poisoned memory)
569 * This means "there are some platform errors".
572 return recover_from_platform_error(slidx
, peidx
, pbci
, sos
);
574 * On account of strange SAL error record, we cannot recover.
580 * mca_try_to_recover - Try to recover from MCA
581 * @rec: pointer to a SAL error record
582 * @sos: pointer to hand off struct between SAL and OS
585 * 1 on Success / 0 on Failure
589 mca_try_to_recover(void *rec
, struct ia64_sal_os_state
*sos
)
595 pal_bus_check_info_t pbci
;
597 /* Make index of SAL error record */
598 platform_err
= mca_make_slidx(rec
, &slidx
);
600 /* Count processor error sections */
601 n_proc_err
= slidx_count(&slidx
, proc_err
);
603 /* Now, OS can recover when there is one processor error section */
606 else if (n_proc_err
== 0) {
607 /* Weird SAL record ... We need not to recover */
612 /* Make index of processor error section */
613 mca_make_peidx((sal_log_processor_info_t
*)
614 slidx_first_entry(&slidx
.proc_err
)->hdr
, &peidx
);
616 /* Extract Processor BUS_CHECK[0] */
617 *((u64
*)&pbci
) = peidx_check_info(&peidx
, bus_check
, 0);
619 /* Check whether MCA is global or not */
620 if (is_mca_global(&peidx
, &pbci
, sos
))
623 /* Try to recover a processor error */
624 return recover_from_processor_error(platform_err
, &slidx
, &peidx
,
629 * =============================================================================
632 int __init
mca_external_handler_init(void)
634 if (init_record_index_pools())
637 /* register external mca handlers */
638 if (ia64_reg_MCA_extension(mca_try_to_recover
)) {
639 printk(KERN_ERR
"ia64_reg_MCA_extension failed.\n");
640 kfree(slidx_pool
.buffer
);
646 void __exit
mca_external_handler_exit(void)
648 /* unregister external mca handlers */
649 ia64_unreg_MCA_extension();
650 kfree(slidx_pool
.buffer
);
653 module_init(mca_external_handler_init
);
654 module_exit(mca_external_handler_exit
);
656 module_param(sal_rec_max
, int, 0644);
657 MODULE_PARM_DESC(sal_rec_max
, "Max size of SAL error record");
659 MODULE_DESCRIPTION("ia64 platform dependent mca handler driver");
660 MODULE_LICENSE("GPL");