Add linux-next specific files for 20110831
[linux-2.6/next.git] / drivers / edac / ppc4xx_edac.c
blob38400963e245f93b74beb056f7e5fa10f2209680
1 /*
2 * Copyright (c) 2008 Nuovation System Designs, LLC
3 * Grant Erickson <gerickson@nuovations.com>
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation; version 2 of the
8 * License.
12 #include <linux/edac.h>
13 #include <linux/interrupt.h>
14 #include <linux/irq.h>
15 #include <linux/kernel.h>
16 #include <linux/mm.h>
17 #include <linux/module.h>
18 #include <linux/of_device.h>
19 #include <linux/of_platform.h>
20 #include <linux/types.h>
22 #include <asm/dcr.h>
24 #include "edac_core.h"
25 #include "ppc4xx_edac.h"
28 * This file implements a driver for monitoring and handling events
29 * associated with the IMB DDR2 ECC controller found in the AMCC/IBM
30 * 405EX[r], 440SP, 440SPe, 460EX, 460GT and 460SX.
32 * As realized in the 405EX[r], this controller features:
34 * - Support for registered- and non-registered DDR1 and DDR2 memory.
35 * - 32-bit or 16-bit memory interface with optional ECC.
37 * o ECC support includes:
39 * - 4-bit SEC/DED
40 * - Aligned-nibble error detect
41 * - Bypass mode
43 * - Two (2) memory banks/ranks.
44 * - Up to 1 GiB per bank/rank in 32-bit mode and up to 512 MiB per
45 * bank/rank in 16-bit mode.
47 * As realized in the 440SP and 440SPe, this controller changes/adds:
49 * - 64-bit or 32-bit memory interface with optional ECC.
51 * o ECC support includes:
53 * - 8-bit SEC/DED
54 * - Aligned-nibble error detect
55 * - Bypass mode
57 * - Up to 4 GiB per bank/rank in 64-bit mode and up to 2 GiB
58 * per bank/rank in 32-bit mode.
60 * As realized in the 460EX and 460GT, this controller changes/adds:
62 * - 64-bit or 32-bit memory interface with optional ECC.
64 * o ECC support includes:
66 * - 8-bit SEC/DED
67 * - Aligned-nibble error detect
68 * - Bypass mode
70 * - Four (4) memory banks/ranks.
71 * - Up to 16 GiB per bank/rank in 64-bit mode and up to 8 GiB
72 * per bank/rank in 32-bit mode.
74 * At present, this driver has ONLY been tested against the controller
75 * realization in the 405EX[r] on the AMCC Kilauea and Haleakala
76 * boards (256 MiB w/o ECC memory soldered onto the board) and a
77 * proprietary board based on those designs (128 MiB ECC memory, also
78 * soldered onto the board).
80 * Dynamic feature detection and handling needs to be added for the
81 * other realizations of this controller listed above.
83 * Eventually, this driver will likely be adapted to the above variant
84 * realizations of this controller as well as broken apart to handle
85 * the other known ECC-capable controllers prevalent in other 4xx
86 * processors:
88 * - IBM SDRAM (405GP, 405CR and 405EP) "ibm,sdram-4xx"
89 * - IBM DDR1 (440GP, 440GX, 440EP and 440GR) "ibm,sdram-4xx-ddr"
90 * - Denali DDR1/DDR2 (440EPX and 440GRX) "denali,sdram-4xx-ddr2"
92 * For this controller, unfortunately, correctable errors report
93 * nothing more than the beat/cycle and byte/lane the correction
94 * occurred on and the check bit group that covered the error.
96 * In contrast, uncorrectable errors also report the failing address,
97 * the bus master and the transaction direction (i.e. read or write)
99 * Regardless of whether the error is a CE or a UE, we report the
100 * following pieces of information in the driver-unique message to the
101 * EDAC subsystem:
103 * - Device tree path
104 * - Bank(s)
105 * - Check bit error group
106 * - Beat(s)/lane(s)
109 /* Preprocessor Definitions */
111 #define EDAC_OPSTATE_INT_STR "interrupt"
112 #define EDAC_OPSTATE_POLL_STR "polled"
113 #define EDAC_OPSTATE_UNKNOWN_STR "unknown"
115 #define PPC4XX_EDAC_MODULE_NAME "ppc4xx_edac"
116 #define PPC4XX_EDAC_MODULE_REVISION "v1.0.0"
118 #define PPC4XX_EDAC_MESSAGE_SIZE 256
121 * Kernel logging without an EDAC instance
123 #define ppc4xx_edac_printk(level, fmt, arg...) \
124 edac_printk(level, "PPC4xx MC", fmt, ##arg)
127 * Kernel logging with an EDAC instance
129 #define ppc4xx_edac_mc_printk(level, mci, fmt, arg...) \
130 edac_mc_chipset_printk(mci, level, "PPC4xx", fmt, ##arg)
133 * Macros to convert bank configuration size enumerations into MiB and
134 * page values.
136 #define SDRAM_MBCF_SZ_MiB_MIN 4
137 #define SDRAM_MBCF_SZ_TO_MiB(n) (SDRAM_MBCF_SZ_MiB_MIN \
138 << (SDRAM_MBCF_SZ_DECODE(n)))
139 #define SDRAM_MBCF_SZ_TO_PAGES(n) (SDRAM_MBCF_SZ_MiB_MIN \
140 << (20 - PAGE_SHIFT + \
141 SDRAM_MBCF_SZ_DECODE(n)))
144 * The ibm,sdram-4xx-ddr2 Device Control Registers (DCRs) are
145 * indirectly acccessed and have a base and length defined by the
146 * device tree. The base can be anything; however, we expect the
147 * length to be precisely two registers, the first for the address
148 * window and the second for the data window.
150 #define SDRAM_DCR_RESOURCE_LEN 2
151 #define SDRAM_DCR_ADDR_OFFSET 0
152 #define SDRAM_DCR_DATA_OFFSET 1
155 * Device tree interrupt indices
157 #define INTMAP_ECCDED_INDEX 0 /* Double-bit Error Detect */
158 #define INTMAP_ECCSEC_INDEX 1 /* Single-bit Error Correct */
160 /* Type Definitions */
163 * PPC4xx SDRAM memory controller private instance data
165 struct ppc4xx_edac_pdata {
166 dcr_host_t dcr_host; /* Indirect DCR address/data window mapping */
167 struct {
168 int sec; /* Single-bit correctable error IRQ assigned */
169 int ded; /* Double-bit detectable error IRQ assigned */
170 } irqs;
174 * Various status data gathered and manipulated when checking and
175 * reporting ECC status.
177 struct ppc4xx_ecc_status {
178 u32 ecces;
179 u32 besr;
180 u32 bearh;
181 u32 bearl;
182 u32 wmirq;
185 /* Function Prototypes */
187 static int ppc4xx_edac_probe(struct platform_device *device)
188 static int ppc4xx_edac_remove(struct platform_device *device);
190 /* Global Variables */
193 * Device tree node type and compatible tuples this driver can match
194 * on.
196 static struct of_device_id ppc4xx_edac_match[] = {
198 .compatible = "ibm,sdram-4xx-ddr2"
203 static struct platform_driver ppc4xx_edac_driver = {
204 .probe = ppc4xx_edac_probe,
205 .remove = ppc4xx_edac_remove,
206 .driver = {
207 .owner = THIS_MODULE,
208 .name = PPC4XX_EDAC_MODULE_NAME,
209 .of_match_table = ppc4xx_edac_match,
214 * TODO: The row and channel parameters likely need to be dynamically
215 * set based on the aforementioned variant controller realizations.
217 static const unsigned ppc4xx_edac_nr_csrows = 2;
218 static const unsigned ppc4xx_edac_nr_chans = 1;
221 * Strings associated with PLB master IDs capable of being posted in
222 * SDRAM_BESR or SDRAM_WMIRQ on uncorrectable ECC errors.
224 static const char * const ppc4xx_plb_masters[9] = {
225 [SDRAM_PLB_M0ID_ICU] = "ICU",
226 [SDRAM_PLB_M0ID_PCIE0] = "PCI-E 0",
227 [SDRAM_PLB_M0ID_PCIE1] = "PCI-E 1",
228 [SDRAM_PLB_M0ID_DMA] = "DMA",
229 [SDRAM_PLB_M0ID_DCU] = "DCU",
230 [SDRAM_PLB_M0ID_OPB] = "OPB",
231 [SDRAM_PLB_M0ID_MAL] = "MAL",
232 [SDRAM_PLB_M0ID_SEC] = "SEC",
233 [SDRAM_PLB_M0ID_AHB] = "AHB"
237 * mfsdram - read and return controller register data
238 * @dcr_host: A pointer to the DCR mapping.
239 * @idcr_n: The indirect DCR register to read.
241 * This routine reads and returns the data associated with the
242 * controller's specified indirect DCR register.
244 * Returns the read data.
246 static inline u32
247 mfsdram(const dcr_host_t *dcr_host, unsigned int idcr_n)
249 return __mfdcri(dcr_host->base + SDRAM_DCR_ADDR_OFFSET,
250 dcr_host->base + SDRAM_DCR_DATA_OFFSET,
251 idcr_n);
255 * mtsdram - write controller register data
256 * @dcr_host: A pointer to the DCR mapping.
257 * @idcr_n: The indirect DCR register to write.
258 * @value: The data to write.
260 * This routine writes the provided data to the controller's specified
261 * indirect DCR register.
263 static inline void
264 mtsdram(const dcr_host_t *dcr_host, unsigned int idcr_n, u32 value)
266 return __mtdcri(dcr_host->base + SDRAM_DCR_ADDR_OFFSET,
267 dcr_host->base + SDRAM_DCR_DATA_OFFSET,
268 idcr_n,
269 value);
273 * ppc4xx_edac_check_bank_error - check a bank for an ECC bank error
274 * @status: A pointer to the ECC status structure to check for an
275 * ECC bank error.
276 * @bank: The bank to check for an ECC error.
278 * This routine determines whether the specified bank has an ECC
279 * error.
281 * Returns true if the specified bank has an ECC error; otherwise,
282 * false.
284 static bool
285 ppc4xx_edac_check_bank_error(const struct ppc4xx_ecc_status *status,
286 unsigned int bank)
288 switch (bank) {
289 case 0:
290 return status->ecces & SDRAM_ECCES_BK0ER;
291 case 1:
292 return status->ecces & SDRAM_ECCES_BK1ER;
293 default:
294 return false;
299 * ppc4xx_edac_generate_bank_message - generate interpretted bank status message
300 * @mci: A pointer to the EDAC memory controller instance associated
301 * with the bank message being generated.
302 * @status: A pointer to the ECC status structure to generate the
303 * message from.
304 * @buffer: A pointer to the buffer in which to generate the
305 * message.
306 * @size: The size, in bytes, of space available in buffer.
308 * This routine generates to the provided buffer the portion of the
309 * driver-unique report message associated with the ECCESS[BKNER]
310 * field of the specified ECC status.
312 * Returns the number of characters generated on success; otherwise, <
313 * 0 on error.
315 static int
316 ppc4xx_edac_generate_bank_message(const struct mem_ctl_info *mci,
317 const struct ppc4xx_ecc_status *status,
318 char *buffer,
319 size_t size)
321 int n, total = 0;
322 unsigned int row, rows;
324 n = snprintf(buffer, size, "%s: Banks: ", mci->dev_name);
326 if (n < 0 || n >= size)
327 goto fail;
329 buffer += n;
330 size -= n;
331 total += n;
333 for (rows = 0, row = 0; row < mci->nr_csrows; row++) {
334 if (ppc4xx_edac_check_bank_error(status, row)) {
335 n = snprintf(buffer, size, "%s%u",
336 (rows++ ? ", " : ""), row);
338 if (n < 0 || n >= size)
339 goto fail;
341 buffer += n;
342 size -= n;
343 total += n;
347 n = snprintf(buffer, size, "%s; ", rows ? "" : "None");
349 if (n < 0 || n >= size)
350 goto fail;
352 buffer += n;
353 size -= n;
354 total += n;
356 fail:
357 return total;
361 * ppc4xx_edac_generate_checkbit_message - generate interpretted checkbit message
362 * @mci: A pointer to the EDAC memory controller instance associated
363 * with the checkbit message being generated.
364 * @status: A pointer to the ECC status structure to generate the
365 * message from.
366 * @buffer: A pointer to the buffer in which to generate the
367 * message.
368 * @size: The size, in bytes, of space available in buffer.
370 * This routine generates to the provided buffer the portion of the
371 * driver-unique report message associated with the ECCESS[CKBER]
372 * field of the specified ECC status.
374 * Returns the number of characters generated on success; otherwise, <
375 * 0 on error.
377 static int
378 ppc4xx_edac_generate_checkbit_message(const struct mem_ctl_info *mci,
379 const struct ppc4xx_ecc_status *status,
380 char *buffer,
381 size_t size)
383 const struct ppc4xx_edac_pdata *pdata = mci->pvt_info;
384 const char *ckber = NULL;
386 switch (status->ecces & SDRAM_ECCES_CKBER_MASK) {
387 case SDRAM_ECCES_CKBER_NONE:
388 ckber = "None";
389 break;
390 case SDRAM_ECCES_CKBER_32_ECC_0_3:
391 ckber = "ECC0:3";
392 break;
393 case SDRAM_ECCES_CKBER_32_ECC_4_8:
394 switch (mfsdram(&pdata->dcr_host, SDRAM_MCOPT1) &
395 SDRAM_MCOPT1_WDTH_MASK) {
396 case SDRAM_MCOPT1_WDTH_16:
397 ckber = "ECC0:3";
398 break;
399 case SDRAM_MCOPT1_WDTH_32:
400 ckber = "ECC4:8";
401 break;
402 default:
403 ckber = "Unknown";
404 break;
406 break;
407 case SDRAM_ECCES_CKBER_32_ECC_0_8:
408 ckber = "ECC0:8";
409 break;
410 default:
411 ckber = "Unknown";
412 break;
415 return snprintf(buffer, size, "Checkbit Error: %s", ckber);
419 * ppc4xx_edac_generate_lane_message - generate interpretted byte lane message
420 * @mci: A pointer to the EDAC memory controller instance associated
421 * with the byte lane message being generated.
422 * @status: A pointer to the ECC status structure to generate the
423 * message from.
424 * @buffer: A pointer to the buffer in which to generate the
425 * message.
426 * @size: The size, in bytes, of space available in buffer.
428 * This routine generates to the provided buffer the portion of the
429 * driver-unique report message associated with the ECCESS[BNCE]
430 * field of the specified ECC status.
432 * Returns the number of characters generated on success; otherwise, <
433 * 0 on error.
435 static int
436 ppc4xx_edac_generate_lane_message(const struct mem_ctl_info *mci,
437 const struct ppc4xx_ecc_status *status,
438 char *buffer,
439 size_t size)
441 int n, total = 0;
442 unsigned int lane, lanes;
443 const unsigned int first_lane = 0;
444 const unsigned int lane_count = 16;
446 n = snprintf(buffer, size, "; Byte Lane Errors: ");
448 if (n < 0 || n >= size)
449 goto fail;
451 buffer += n;
452 size -= n;
453 total += n;
455 for (lanes = 0, lane = first_lane; lane < lane_count; lane++) {
456 if ((status->ecces & SDRAM_ECCES_BNCE_ENCODE(lane)) != 0) {
457 n = snprintf(buffer, size,
458 "%s%u",
459 (lanes++ ? ", " : ""), lane);
461 if (n < 0 || n >= size)
462 goto fail;
464 buffer += n;
465 size -= n;
466 total += n;
470 n = snprintf(buffer, size, "%s; ", lanes ? "" : "None");
472 if (n < 0 || n >= size)
473 goto fail;
475 buffer += n;
476 size -= n;
477 total += n;
479 fail:
480 return total;
484 * ppc4xx_edac_generate_ecc_message - generate interpretted ECC status message
485 * @mci: A pointer to the EDAC memory controller instance associated
486 * with the ECCES message being generated.
487 * @status: A pointer to the ECC status structure to generate the
488 * message from.
489 * @buffer: A pointer to the buffer in which to generate the
490 * message.
491 * @size: The size, in bytes, of space available in buffer.
493 * This routine generates to the provided buffer the portion of the
494 * driver-unique report message associated with the ECCESS register of
495 * the specified ECC status.
497 * Returns the number of characters generated on success; otherwise, <
498 * 0 on error.
500 static int
501 ppc4xx_edac_generate_ecc_message(const struct mem_ctl_info *mci,
502 const struct ppc4xx_ecc_status *status,
503 char *buffer,
504 size_t size)
506 int n, total = 0;
508 n = ppc4xx_edac_generate_bank_message(mci, status, buffer, size);
510 if (n < 0 || n >= size)
511 goto fail;
513 buffer += n;
514 size -= n;
515 total += n;
517 n = ppc4xx_edac_generate_checkbit_message(mci, status, buffer, size);
519 if (n < 0 || n >= size)
520 goto fail;
522 buffer += n;
523 size -= n;
524 total += n;
526 n = ppc4xx_edac_generate_lane_message(mci, status, buffer, size);
528 if (n < 0 || n >= size)
529 goto fail;
531 buffer += n;
532 size -= n;
533 total += n;
535 fail:
536 return total;
540 * ppc4xx_edac_generate_plb_message - generate interpretted PLB status message
541 * @mci: A pointer to the EDAC memory controller instance associated
542 * with the PLB message being generated.
543 * @status: A pointer to the ECC status structure to generate the
544 * message from.
545 * @buffer: A pointer to the buffer in which to generate the
546 * message.
547 * @size: The size, in bytes, of space available in buffer.
549 * This routine generates to the provided buffer the portion of the
550 * driver-unique report message associated with the PLB-related BESR
551 * and/or WMIRQ registers of the specified ECC status.
553 * Returns the number of characters generated on success; otherwise, <
554 * 0 on error.
556 static int
557 ppc4xx_edac_generate_plb_message(const struct mem_ctl_info *mci,
558 const struct ppc4xx_ecc_status *status,
559 char *buffer,
560 size_t size)
562 unsigned int master;
563 bool read;
565 if ((status->besr & SDRAM_BESR_MASK) == 0)
566 return 0;
568 if ((status->besr & SDRAM_BESR_M0ET_MASK) == SDRAM_BESR_M0ET_NONE)
569 return 0;
571 read = ((status->besr & SDRAM_BESR_M0RW_MASK) == SDRAM_BESR_M0RW_READ);
573 master = SDRAM_BESR_M0ID_DECODE(status->besr);
575 return snprintf(buffer, size,
576 "%s error w/ PLB master %u \"%s\"; ",
577 (read ? "Read" : "Write"),
578 master,
579 (((master >= SDRAM_PLB_M0ID_FIRST) &&
580 (master <= SDRAM_PLB_M0ID_LAST)) ?
581 ppc4xx_plb_masters[master] : "UNKNOWN"));
585 * ppc4xx_edac_generate_message - generate interpretted status message
586 * @mci: A pointer to the EDAC memory controller instance associated
587 * with the driver-unique message being generated.
588 * @status: A pointer to the ECC status structure to generate the
589 * message from.
590 * @buffer: A pointer to the buffer in which to generate the
591 * message.
592 * @size: The size, in bytes, of space available in buffer.
594 * This routine generates to the provided buffer the driver-unique
595 * EDAC report message from the specified ECC status.
597 static void
598 ppc4xx_edac_generate_message(const struct mem_ctl_info *mci,
599 const struct ppc4xx_ecc_status *status,
600 char *buffer,
601 size_t size)
603 int n;
605 if (buffer == NULL || size == 0)
606 return;
608 n = ppc4xx_edac_generate_ecc_message(mci, status, buffer, size);
610 if (n < 0 || n >= size)
611 return;
613 buffer += n;
614 size -= n;
616 ppc4xx_edac_generate_plb_message(mci, status, buffer, size);
619 #ifdef DEBUG
621 * ppc4xx_ecc_dump_status - dump controller ECC status registers
622 * @mci: A pointer to the EDAC memory controller instance
623 * associated with the status being dumped.
624 * @status: A pointer to the ECC status structure to generate the
625 * dump from.
627 * This routine dumps to the kernel log buffer the raw and
628 * interpretted specified ECC status.
630 static void
631 ppc4xx_ecc_dump_status(const struct mem_ctl_info *mci,
632 const struct ppc4xx_ecc_status *status)
634 char message[PPC4XX_EDAC_MESSAGE_SIZE];
636 ppc4xx_edac_generate_message(mci, status, message, sizeof(message));
638 ppc4xx_edac_mc_printk(KERN_INFO, mci,
639 "\n"
640 "\tECCES: 0x%08x\n"
641 "\tWMIRQ: 0x%08x\n"
642 "\tBESR: 0x%08x\n"
643 "\tBEAR: 0x%08x%08x\n"
644 "\t%s\n",
645 status->ecces,
646 status->wmirq,
647 status->besr,
648 status->bearh,
649 status->bearl,
650 message);
652 #endif /* DEBUG */
655 * ppc4xx_ecc_get_status - get controller ECC status
656 * @mci: A pointer to the EDAC memory controller instance
657 * associated with the status being retrieved.
658 * @status: A pointer to the ECC status structure to populate the
659 * ECC status with.
661 * This routine reads and masks, as appropriate, all the relevant
662 * status registers that deal with ibm,sdram-4xx-ddr2 ECC errors.
663 * While we read all of them, for correctable errors, we only expect
664 * to deal with ECCES. For uncorrectable errors, we expect to deal
665 * with all of them.
667 static void
668 ppc4xx_ecc_get_status(const struct mem_ctl_info *mci,
669 struct ppc4xx_ecc_status *status)
671 const struct ppc4xx_edac_pdata *pdata = mci->pvt_info;
672 const dcr_host_t *dcr_host = &pdata->dcr_host;
674 status->ecces = mfsdram(dcr_host, SDRAM_ECCES) & SDRAM_ECCES_MASK;
675 status->wmirq = mfsdram(dcr_host, SDRAM_WMIRQ) & SDRAM_WMIRQ_MASK;
676 status->besr = mfsdram(dcr_host, SDRAM_BESR) & SDRAM_BESR_MASK;
677 status->bearl = mfsdram(dcr_host, SDRAM_BEARL);
678 status->bearh = mfsdram(dcr_host, SDRAM_BEARH);
682 * ppc4xx_ecc_clear_status - clear controller ECC status
683 * @mci: A pointer to the EDAC memory controller instance
684 * associated with the status being cleared.
685 * @status: A pointer to the ECC status structure containing the
686 * values to write to clear the ECC status.
688 * This routine clears--by writing the masked (as appropriate) status
689 * values back to--the status registers that deal with
690 * ibm,sdram-4xx-ddr2 ECC errors.
692 static void
693 ppc4xx_ecc_clear_status(const struct mem_ctl_info *mci,
694 const struct ppc4xx_ecc_status *status)
696 const struct ppc4xx_edac_pdata *pdata = mci->pvt_info;
697 const dcr_host_t *dcr_host = &pdata->dcr_host;
699 mtsdram(dcr_host, SDRAM_ECCES, status->ecces & SDRAM_ECCES_MASK);
700 mtsdram(dcr_host, SDRAM_WMIRQ, status->wmirq & SDRAM_WMIRQ_MASK);
701 mtsdram(dcr_host, SDRAM_BESR, status->besr & SDRAM_BESR_MASK);
702 mtsdram(dcr_host, SDRAM_BEARL, 0);
703 mtsdram(dcr_host, SDRAM_BEARH, 0);
707 * ppc4xx_edac_handle_ce - handle controller correctable ECC error (CE)
708 * @mci: A pointer to the EDAC memory controller instance
709 * associated with the correctable error being handled and reported.
710 * @status: A pointer to the ECC status structure associated with
711 * the correctable error being handled and reported.
713 * This routine handles an ibm,sdram-4xx-ddr2 controller ECC
714 * correctable error. Per the aforementioned discussion, there's not
715 * enough status available to use the full EDAC correctable error
716 * interface, so we just pass driver-unique message to the "no info"
717 * interface.
719 static void
720 ppc4xx_edac_handle_ce(struct mem_ctl_info *mci,
721 const struct ppc4xx_ecc_status *status)
723 int row;
724 char message[PPC4XX_EDAC_MESSAGE_SIZE];
726 ppc4xx_edac_generate_message(mci, status, message, sizeof(message));
728 for (row = 0; row < mci->nr_csrows; row++)
729 if (ppc4xx_edac_check_bank_error(status, row))
730 edac_mc_handle_ce_no_info(mci, message);
734 * ppc4xx_edac_handle_ue - handle controller uncorrectable ECC error (UE)
735 * @mci: A pointer to the EDAC memory controller instance
736 * associated with the uncorrectable error being handled and
737 * reported.
738 * @status: A pointer to the ECC status structure associated with
739 * the uncorrectable error being handled and reported.
741 * This routine handles an ibm,sdram-4xx-ddr2 controller ECC
742 * uncorrectable error.
744 static void
745 ppc4xx_edac_handle_ue(struct mem_ctl_info *mci,
746 const struct ppc4xx_ecc_status *status)
748 const u64 bear = ((u64)status->bearh << 32 | status->bearl);
749 const unsigned long page = bear >> PAGE_SHIFT;
750 const unsigned long offset = bear & ~PAGE_MASK;
751 int row;
752 char message[PPC4XX_EDAC_MESSAGE_SIZE];
754 ppc4xx_edac_generate_message(mci, status, message, sizeof(message));
756 for (row = 0; row < mci->nr_csrows; row++)
757 if (ppc4xx_edac_check_bank_error(status, row))
758 edac_mc_handle_ue(mci, page, offset, row, message);
762 * ppc4xx_edac_check - check controller for ECC errors
763 * @mci: A pointer to the EDAC memory controller instance
764 * associated with the ibm,sdram-4xx-ddr2 controller being
765 * checked.
767 * This routine is used to check and post ECC errors and is called by
768 * both the EDAC polling thread and this driver's CE and UE interrupt
769 * handler.
771 static void
772 ppc4xx_edac_check(struct mem_ctl_info *mci)
774 #ifdef DEBUG
775 static unsigned int count;
776 #endif
777 struct ppc4xx_ecc_status status;
779 ppc4xx_ecc_get_status(mci, &status);
781 #ifdef DEBUG
782 if (count++ % 30 == 0)
783 ppc4xx_ecc_dump_status(mci, &status);
784 #endif
786 if (status.ecces & SDRAM_ECCES_UE)
787 ppc4xx_edac_handle_ue(mci, &status);
789 if (status.ecces & SDRAM_ECCES_CE)
790 ppc4xx_edac_handle_ce(mci, &status);
792 ppc4xx_ecc_clear_status(mci, &status);
796 * ppc4xx_edac_isr - SEC (CE) and DED (UE) interrupt service routine
797 * @irq: The virtual interrupt number being serviced.
798 * @dev_id: A pointer to the EDAC memory controller instance
799 * associated with the interrupt being handled.
801 * This routine implements the interrupt handler for both correctable
802 * (CE) and uncorrectable (UE) ECC errors for the ibm,sdram-4xx-ddr2
803 * controller. It simply calls through to the same routine used during
804 * polling to check, report and clear the ECC status.
806 * Unconditionally returns IRQ_HANDLED.
808 static irqreturn_t
809 ppc4xx_edac_isr(int irq, void *dev_id)
811 struct mem_ctl_info *mci = dev_id;
813 ppc4xx_edac_check(mci);
815 return IRQ_HANDLED;
819 * ppc4xx_edac_get_dtype - return the controller memory width
820 * @mcopt1: The 32-bit Memory Controller Option 1 register value
821 * currently set for the controller, from which the width
822 * is derived.
824 * This routine returns the EDAC device type width appropriate for the
825 * current controller configuration.
827 * TODO: This needs to be conditioned dynamically through feature
828 * flags or some such when other controller variants are supported as
829 * the 405EX[r] is 16-/32-bit and the others are 32-/64-bit with the
830 * 16- and 64-bit field definition/value/enumeration (b1) overloaded
831 * among them.
833 * Returns a device type width enumeration.
835 static enum dev_type __devinit
836 ppc4xx_edac_get_dtype(u32 mcopt1)
838 switch (mcopt1 & SDRAM_MCOPT1_WDTH_MASK) {
839 case SDRAM_MCOPT1_WDTH_16:
840 return DEV_X2;
841 case SDRAM_MCOPT1_WDTH_32:
842 return DEV_X4;
843 default:
844 return DEV_UNKNOWN;
849 * ppc4xx_edac_get_mtype - return controller memory type
850 * @mcopt1: The 32-bit Memory Controller Option 1 register value
851 * currently set for the controller, from which the memory type
852 * is derived.
854 * This routine returns the EDAC memory type appropriate for the
855 * current controller configuration.
857 * Returns a memory type enumeration.
859 static enum mem_type __devinit
860 ppc4xx_edac_get_mtype(u32 mcopt1)
862 bool rden = ((mcopt1 & SDRAM_MCOPT1_RDEN_MASK) == SDRAM_MCOPT1_RDEN);
864 switch (mcopt1 & SDRAM_MCOPT1_DDR_TYPE_MASK) {
865 case SDRAM_MCOPT1_DDR2_TYPE:
866 return rden ? MEM_RDDR2 : MEM_DDR2;
867 case SDRAM_MCOPT1_DDR1_TYPE:
868 return rden ? MEM_RDDR : MEM_DDR;
869 default:
870 return MEM_UNKNOWN;
875 * ppc4xx_edac_init_csrows - initialize driver instance rows
876 * @mci: A pointer to the EDAC memory controller instance
877 * associated with the ibm,sdram-4xx-ddr2 controller for which
878 * the csrows (i.e. banks/ranks) are being initialized.
879 * @mcopt1: The 32-bit Memory Controller Option 1 register value
880 * currently set for the controller, from which bank width
881 * and memory typ information is derived.
883 * This routine initializes the virtual "chip select rows" associated
884 * with the EDAC memory controller instance. An ibm,sdram-4xx-ddr2
885 * controller bank/rank is mapped to a row.
887 * Returns 0 if OK; otherwise, -EINVAL if the memory bank size
888 * configuration cannot be determined.
890 static int __devinit
891 ppc4xx_edac_init_csrows(struct mem_ctl_info *mci, u32 mcopt1)
893 const struct ppc4xx_edac_pdata *pdata = mci->pvt_info;
894 int status = 0;
895 enum mem_type mtype;
896 enum dev_type dtype;
897 enum edac_type edac_mode;
898 int row;
899 u32 mbxcf, size;
900 static u32 ppc4xx_last_page;
902 /* Establish the memory type and width */
904 mtype = ppc4xx_edac_get_mtype(mcopt1);
905 dtype = ppc4xx_edac_get_dtype(mcopt1);
907 /* Establish EDAC mode */
909 if (mci->edac_cap & EDAC_FLAG_SECDED)
910 edac_mode = EDAC_SECDED;
911 else if (mci->edac_cap & EDAC_FLAG_EC)
912 edac_mode = EDAC_EC;
913 else
914 edac_mode = EDAC_NONE;
917 * Initialize each chip select row structure which correspond
918 * 1:1 with a controller bank/rank.
921 for (row = 0; row < mci->nr_csrows; row++) {
922 struct csrow_info *csi = &mci->csrows[row];
925 * Get the configuration settings for this
926 * row/bank/rank and skip disabled banks.
929 mbxcf = mfsdram(&pdata->dcr_host, SDRAM_MBXCF(row));
931 if ((mbxcf & SDRAM_MBCF_BE_MASK) != SDRAM_MBCF_BE_ENABLE)
932 continue;
934 /* Map the bank configuration size setting to pages. */
936 size = mbxcf & SDRAM_MBCF_SZ_MASK;
938 switch (size) {
939 case SDRAM_MBCF_SZ_4MB:
940 case SDRAM_MBCF_SZ_8MB:
941 case SDRAM_MBCF_SZ_16MB:
942 case SDRAM_MBCF_SZ_32MB:
943 case SDRAM_MBCF_SZ_64MB:
944 case SDRAM_MBCF_SZ_128MB:
945 case SDRAM_MBCF_SZ_256MB:
946 case SDRAM_MBCF_SZ_512MB:
947 case SDRAM_MBCF_SZ_1GB:
948 case SDRAM_MBCF_SZ_2GB:
949 case SDRAM_MBCF_SZ_4GB:
950 case SDRAM_MBCF_SZ_8GB:
951 csi->nr_pages = SDRAM_MBCF_SZ_TO_PAGES(size);
952 break;
953 default:
954 ppc4xx_edac_mc_printk(KERN_ERR, mci,
955 "Unrecognized memory bank %d "
956 "size 0x%08x\n",
957 row, SDRAM_MBCF_SZ_DECODE(size));
958 status = -EINVAL;
959 goto done;
962 csi->first_page = ppc4xx_last_page;
963 csi->last_page = csi->first_page + csi->nr_pages - 1;
964 csi->page_mask = 0;
967 * It's unclear exactly what grain should be set to
968 * here. The SDRAM_ECCES register allows resolution of
969 * an error down to a nibble which would potentially
970 * argue for a grain of '1' byte, even though we only
971 * know the associated address for uncorrectable
972 * errors. This value is not used at present for
973 * anything other than error reporting so getting it
974 * wrong should be of little consequence. Other
975 * possible values would be the PLB width (16), the
976 * page size (PAGE_SIZE) or the memory width (2 or 4).
979 csi->grain = 1;
981 csi->mtype = mtype;
982 csi->dtype = dtype;
984 csi->edac_mode = edac_mode;
986 ppc4xx_last_page += csi->nr_pages;
989 done:
990 return status;
994 * ppc4xx_edac_mc_init - initialize driver instance
995 * @mci: A pointer to the EDAC memory controller instance being
996 * initialized.
997 * @op: A pointer to the OpenFirmware device tree node associated
998 * with the controller this EDAC instance is bound to.
999 * @dcr_host: A pointer to the DCR data containing the DCR mapping
1000 * for this controller instance.
1001 * @mcopt1: The 32-bit Memory Controller Option 1 register value
1002 * currently set for the controller, from which ECC capabilities
1003 * and scrub mode are derived.
1005 * This routine performs initialization of the EDAC memory controller
1006 * instance and related driver-private data associated with the
1007 * ibm,sdram-4xx-ddr2 memory controller the instance is bound to.
1009 * Returns 0 if OK; otherwise, < 0 on error.
1011 static int __devinit
1012 ppc4xx_edac_mc_init(struct mem_ctl_info *mci,
1013 struct platform_device *op,
1014 const dcr_host_t *dcr_host,
1015 u32 mcopt1)
1017 int status = 0;
1018 const u32 memcheck = (mcopt1 & SDRAM_MCOPT1_MCHK_MASK);
1019 struct ppc4xx_edac_pdata *pdata = NULL;
1020 const struct device_node *np = op->dev.of_node;
1022 if (of_match_device(ppc4xx_edac_match, &op->dev) == NULL)
1023 return -EINVAL;
1025 /* Initial driver pointers and private data */
1027 mci->dev = &op->dev;
1029 dev_set_drvdata(mci->dev, mci);
1031 pdata = mci->pvt_info;
1033 pdata->dcr_host = *dcr_host;
1034 pdata->irqs.sec = NO_IRQ;
1035 pdata->irqs.ded = NO_IRQ;
1037 /* Initialize controller capabilities and configuration */
1039 mci->mtype_cap = (MEM_FLAG_DDR | MEM_FLAG_RDDR |
1040 MEM_FLAG_DDR2 | MEM_FLAG_RDDR2);
1042 mci->edac_ctl_cap = (EDAC_FLAG_NONE |
1043 EDAC_FLAG_EC |
1044 EDAC_FLAG_SECDED);
1046 mci->scrub_cap = SCRUB_NONE;
1047 mci->scrub_mode = SCRUB_NONE;
1050 * Update the actual capabilites based on the MCOPT1[MCHK]
1051 * settings. Scrubbing is only useful if reporting is enabled.
1054 switch (memcheck) {
1055 case SDRAM_MCOPT1_MCHK_CHK:
1056 mci->edac_cap = EDAC_FLAG_EC;
1057 break;
1058 case SDRAM_MCOPT1_MCHK_CHK_REP:
1059 mci->edac_cap = (EDAC_FLAG_EC | EDAC_FLAG_SECDED);
1060 mci->scrub_mode = SCRUB_SW_SRC;
1061 break;
1062 default:
1063 mci->edac_cap = EDAC_FLAG_NONE;
1064 break;
1067 /* Initialize strings */
1069 mci->mod_name = PPC4XX_EDAC_MODULE_NAME;
1070 mci->mod_ver = PPC4XX_EDAC_MODULE_REVISION;
1071 mci->ctl_name = match->compatible,
1072 mci->dev_name = np->full_name;
1074 /* Initialize callbacks */
1076 mci->edac_check = ppc4xx_edac_check;
1077 mci->ctl_page_to_phys = NULL;
1079 /* Initialize chip select rows */
1081 status = ppc4xx_edac_init_csrows(mci, mcopt1);
1083 if (status)
1084 ppc4xx_edac_mc_printk(KERN_ERR, mci,
1085 "Failed to initialize rows!\n");
1087 return status;
1091 * ppc4xx_edac_register_irq - setup and register controller interrupts
1092 * @op: A pointer to the OpenFirmware device tree node associated
1093 * with the controller this EDAC instance is bound to.
1094 * @mci: A pointer to the EDAC memory controller instance
1095 * associated with the ibm,sdram-4xx-ddr2 controller for which
1096 * interrupts are being registered.
1098 * This routine parses the correctable (CE) and uncorrectable error (UE)
1099 * interrupts from the device tree node and maps and assigns them to
1100 * the associated EDAC memory controller instance.
1102 * Returns 0 if OK; otherwise, -ENODEV if the interrupts could not be
1103 * mapped and assigned.
1105 static int __devinit
1106 ppc4xx_edac_register_irq(struct platform_device *op, struct mem_ctl_info *mci)
1108 int status = 0;
1109 int ded_irq, sec_irq;
1110 struct ppc4xx_edac_pdata *pdata = mci->pvt_info;
1111 struct device_node *np = op->dev.of_node;
1113 ded_irq = irq_of_parse_and_map(np, INTMAP_ECCDED_INDEX);
1114 sec_irq = irq_of_parse_and_map(np, INTMAP_ECCSEC_INDEX);
1116 if (ded_irq == NO_IRQ || sec_irq == NO_IRQ) {
1117 ppc4xx_edac_mc_printk(KERN_ERR, mci,
1118 "Unable to map interrupts.\n");
1119 status = -ENODEV;
1120 goto fail;
1123 status = request_irq(ded_irq,
1124 ppc4xx_edac_isr,
1125 IRQF_DISABLED,
1126 "[EDAC] MC ECCDED",
1127 mci);
1129 if (status < 0) {
1130 ppc4xx_edac_mc_printk(KERN_ERR, mci,
1131 "Unable to request irq %d for ECC DED",
1132 ded_irq);
1133 status = -ENODEV;
1134 goto fail1;
1137 status = request_irq(sec_irq,
1138 ppc4xx_edac_isr,
1139 IRQF_DISABLED,
1140 "[EDAC] MC ECCSEC",
1141 mci);
1143 if (status < 0) {
1144 ppc4xx_edac_mc_printk(KERN_ERR, mci,
1145 "Unable to request irq %d for ECC SEC",
1146 sec_irq);
1147 status = -ENODEV;
1148 goto fail2;
1151 ppc4xx_edac_mc_printk(KERN_INFO, mci, "ECCDED irq is %d\n", ded_irq);
1152 ppc4xx_edac_mc_printk(KERN_INFO, mci, "ECCSEC irq is %d\n", sec_irq);
1154 pdata->irqs.ded = ded_irq;
1155 pdata->irqs.sec = sec_irq;
1157 return 0;
1159 fail2:
1160 free_irq(sec_irq, mci);
1162 fail1:
1163 free_irq(ded_irq, mci);
1165 fail:
1166 return status;
1170 * ppc4xx_edac_map_dcrs - locate and map controller registers
1171 * @np: A pointer to the device tree node containing the DCR
1172 * resources to map.
1173 * @dcr_host: A pointer to the DCR data to populate with the
1174 * DCR mapping.
1176 * This routine attempts to locate in the device tree and map the DCR
1177 * register resources associated with the controller's indirect DCR
1178 * address and data windows.
1180 * Returns 0 if the DCRs were successfully mapped; otherwise, < 0 on
1181 * error.
1183 static int __devinit
1184 ppc4xx_edac_map_dcrs(const struct device_node *np, dcr_host_t *dcr_host)
1186 unsigned int dcr_base, dcr_len;
1188 if (np == NULL || dcr_host == NULL)
1189 return -EINVAL;
1191 /* Get the DCR resource extent and sanity check the values. */
1193 dcr_base = dcr_resource_start(np, 0);
1194 dcr_len = dcr_resource_len(np, 0);
1196 if (dcr_base == 0 || dcr_len == 0) {
1197 ppc4xx_edac_printk(KERN_ERR,
1198 "Failed to obtain DCR property.\n");
1199 return -ENODEV;
1202 if (dcr_len != SDRAM_DCR_RESOURCE_LEN) {
1203 ppc4xx_edac_printk(KERN_ERR,
1204 "Unexpected DCR length %d, expected %d.\n",
1205 dcr_len, SDRAM_DCR_RESOURCE_LEN);
1206 return -ENODEV;
1209 /* Attempt to map the DCR extent. */
1211 *dcr_host = dcr_map(np, dcr_base, dcr_len);
1213 if (!DCR_MAP_OK(*dcr_host)) {
1214 ppc4xx_edac_printk(KERN_INFO, "Failed to map DCRs.\n");
1215 return -ENODEV;
1218 return 0;
1222 * ppc4xx_edac_probe - check controller and bind driver
1223 * @op: A pointer to the OpenFirmware device tree node associated
1224 * with the controller being probed for driver binding.
1226 * This routine probes a specific ibm,sdram-4xx-ddr2 controller
1227 * instance for binding with the driver.
1229 * Returns 0 if the controller instance was successfully bound to the
1230 * driver; otherwise, < 0 on error.
1232 static int __devinit ppc4xx_edac_probe(struct platform_device *op)
1234 int status = 0;
1235 u32 mcopt1, memcheck;
1236 dcr_host_t dcr_host;
1237 const struct device_node *np = op->dev.of_node;
1238 struct mem_ctl_info *mci = NULL;
1239 static int ppc4xx_edac_instance;
1242 * At this point, we only support the controller realized on
1243 * the AMCC PPC 405EX[r]. Reject anything else.
1246 if (!of_device_is_compatible(np, "ibm,sdram-405ex") &&
1247 !of_device_is_compatible(np, "ibm,sdram-405exr")) {
1248 ppc4xx_edac_printk(KERN_NOTICE,
1249 "Only the PPC405EX[r] is supported.\n");
1250 return -ENODEV;
1254 * Next, get the DCR property and attempt to map it so that we
1255 * can probe the controller.
1258 status = ppc4xx_edac_map_dcrs(np, &dcr_host);
1260 if (status)
1261 return status;
1264 * First determine whether ECC is enabled at all. If not,
1265 * there is no useful checking or monitoring that can be done
1266 * for this controller.
1269 mcopt1 = mfsdram(&dcr_host, SDRAM_MCOPT1);
1270 memcheck = (mcopt1 & SDRAM_MCOPT1_MCHK_MASK);
1272 if (memcheck == SDRAM_MCOPT1_MCHK_NON) {
1273 ppc4xx_edac_printk(KERN_INFO, "%s: No ECC memory detected or "
1274 "ECC is disabled.\n", np->full_name);
1275 status = -ENODEV;
1276 goto done;
1280 * At this point, we know ECC is enabled, allocate an EDAC
1281 * controller instance and perform the appropriate
1282 * initialization.
1285 mci = edac_mc_alloc(sizeof(struct ppc4xx_edac_pdata),
1286 ppc4xx_edac_nr_csrows,
1287 ppc4xx_edac_nr_chans,
1288 ppc4xx_edac_instance);
1290 if (mci == NULL) {
1291 ppc4xx_edac_printk(KERN_ERR, "%s: "
1292 "Failed to allocate EDAC MC instance!\n",
1293 np->full_name);
1294 status = -ENOMEM;
1295 goto done;
1298 status = ppc4xx_edac_mc_init(mci, op, &dcr_host, mcopt1);
1300 if (status) {
1301 ppc4xx_edac_mc_printk(KERN_ERR, mci,
1302 "Failed to initialize instance!\n");
1303 goto fail;
1307 * We have a valid, initialized EDAC instance bound to the
1308 * controller. Attempt to register it with the EDAC subsystem
1309 * and, if necessary, register interrupts.
1312 if (edac_mc_add_mc(mci)) {
1313 ppc4xx_edac_mc_printk(KERN_ERR, mci,
1314 "Failed to add instance!\n");
1315 status = -ENODEV;
1316 goto fail;
1319 if (edac_op_state == EDAC_OPSTATE_INT) {
1320 status = ppc4xx_edac_register_irq(op, mci);
1322 if (status)
1323 goto fail1;
1326 ppc4xx_edac_instance++;
1328 return 0;
1330 fail1:
1331 edac_mc_del_mc(mci->dev);
1333 fail:
1334 edac_mc_free(mci);
1336 done:
1337 return status;
1341 * ppc4xx_edac_remove - unbind driver from controller
1342 * @op: A pointer to the OpenFirmware device tree node associated
1343 * with the controller this EDAC instance is to be unbound/removed
1344 * from.
1346 * This routine unbinds the EDAC memory controller instance associated
1347 * with the specified ibm,sdram-4xx-ddr2 controller described by the
1348 * OpenFirmware device tree node passed as a parameter.
1350 * Unconditionally returns 0.
1352 static int
1353 ppc4xx_edac_remove(struct platform_device *op)
1355 struct mem_ctl_info *mci = dev_get_drvdata(&op->dev);
1356 struct ppc4xx_edac_pdata *pdata = mci->pvt_info;
1358 if (edac_op_state == EDAC_OPSTATE_INT) {
1359 free_irq(pdata->irqs.sec, mci);
1360 free_irq(pdata->irqs.ded, mci);
1363 dcr_unmap(pdata->dcr_host, SDRAM_DCR_RESOURCE_LEN);
1365 edac_mc_del_mc(mci->dev);
1366 edac_mc_free(mci);
1368 return 0;
1372 * ppc4xx_edac_opstate_init - initialize EDAC reporting method
1374 * This routine ensures that the EDAC memory controller reporting
1375 * method is mapped to a sane value as the EDAC core defines the value
1376 * to EDAC_OPSTATE_INVAL by default. We don't call the global
1377 * opstate_init as that defaults to polling and we want interrupt as
1378 * the default.
1380 static inline void __init
1381 ppc4xx_edac_opstate_init(void)
1383 switch (edac_op_state) {
1384 case EDAC_OPSTATE_POLL:
1385 case EDAC_OPSTATE_INT:
1386 break;
1387 default:
1388 edac_op_state = EDAC_OPSTATE_INT;
1389 break;
1392 ppc4xx_edac_printk(KERN_INFO, "Reporting type: %s\n",
1393 ((edac_op_state == EDAC_OPSTATE_POLL) ?
1394 EDAC_OPSTATE_POLL_STR :
1395 ((edac_op_state == EDAC_OPSTATE_INT) ?
1396 EDAC_OPSTATE_INT_STR :
1397 EDAC_OPSTATE_UNKNOWN_STR)));
1401 * ppc4xx_edac_init - driver/module insertion entry point
1403 * This routine is the driver/module insertion entry point. It
1404 * initializes the EDAC memory controller reporting state and
1405 * registers the driver as an OpenFirmware device tree platform
1406 * driver.
1408 static int __init
1409 ppc4xx_edac_init(void)
1411 ppc4xx_edac_printk(KERN_INFO, PPC4XX_EDAC_MODULE_REVISION "\n");
1413 ppc4xx_edac_opstate_init();
1415 return platform_driver_register(&ppc4xx_edac_driver);
1419 * ppc4xx_edac_exit - driver/module removal entry point
1421 * This routine is the driver/module removal entry point. It
1422 * unregisters the driver as an OpenFirmware device tree platform
1423 * driver.
1425 static void __exit
1426 ppc4xx_edac_exit(void)
1428 platform_driver_unregister(&ppc4xx_edac_driver);
1431 module_init(ppc4xx_edac_init);
1432 module_exit(ppc4xx_edac_exit);
1434 MODULE_LICENSE("GPL v2");
1435 MODULE_AUTHOR("Grant Erickson <gerickson@nuovations.com>");
1436 MODULE_DESCRIPTION("EDAC MC Driver for the PPC4xx IBM DDR2 Memory Controller");
1437 module_param(edac_op_state, int, 0444);
1438 MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting State: "
1439 "0=" EDAC_OPSTATE_POLL_STR ", 2=" EDAC_OPSTATE_INT_STR);