Merge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/parisc-2.6
[linux-2.6/cjktty.git] / drivers / edac / ppc4xx_edac.c
blobb9f0c20df1aa587a8e1c066b3f7cc52983427fb1
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 " __DATE__
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 const struct of_device_id *device_id);
189 static int ppc4xx_edac_remove(struct platform_device *device);
191 /* Global Variables */
194 * Device tree node type and compatible tuples this driver can match
195 * on.
197 static struct of_device_id ppc4xx_edac_match[] = {
199 .compatible = "ibm,sdram-4xx-ddr2"
204 static struct of_platform_driver ppc4xx_edac_driver = {
205 .probe = ppc4xx_edac_probe,
206 .remove = ppc4xx_edac_remove,
207 .driver = {
208 .owner = THIS_MODULE,
209 .name = PPC4XX_EDAC_MODULE_NAME
210 .of_match_table = ppc4xx_edac_match,
215 * TODO: The row and channel parameters likely need to be dynamically
216 * set based on the aforementioned variant controller realizations.
218 static const unsigned ppc4xx_edac_nr_csrows = 2;
219 static const unsigned ppc4xx_edac_nr_chans = 1;
222 * Strings associated with PLB master IDs capable of being posted in
223 * SDRAM_BESR or SDRAM_WMIRQ on uncorrectable ECC errors.
225 static const char * const ppc4xx_plb_masters[9] = {
226 [SDRAM_PLB_M0ID_ICU] = "ICU",
227 [SDRAM_PLB_M0ID_PCIE0] = "PCI-E 0",
228 [SDRAM_PLB_M0ID_PCIE1] = "PCI-E 1",
229 [SDRAM_PLB_M0ID_DMA] = "DMA",
230 [SDRAM_PLB_M0ID_DCU] = "DCU",
231 [SDRAM_PLB_M0ID_OPB] = "OPB",
232 [SDRAM_PLB_M0ID_MAL] = "MAL",
233 [SDRAM_PLB_M0ID_SEC] = "SEC",
234 [SDRAM_PLB_M0ID_AHB] = "AHB"
238 * mfsdram - read and return controller register data
239 * @dcr_host: A pointer to the DCR mapping.
240 * @idcr_n: The indirect DCR register to read.
242 * This routine reads and returns the data associated with the
243 * controller's specified indirect DCR register.
245 * Returns the read data.
247 static inline u32
248 mfsdram(const dcr_host_t *dcr_host, unsigned int idcr_n)
250 return __mfdcri(dcr_host->base + SDRAM_DCR_ADDR_OFFSET,
251 dcr_host->base + SDRAM_DCR_DATA_OFFSET,
252 idcr_n);
256 * mtsdram - write controller register data
257 * @dcr_host: A pointer to the DCR mapping.
258 * @idcr_n: The indirect DCR register to write.
259 * @value: The data to write.
261 * This routine writes the provided data to the controller's specified
262 * indirect DCR register.
264 static inline void
265 mtsdram(const dcr_host_t *dcr_host, unsigned int idcr_n, u32 value)
267 return __mtdcri(dcr_host->base + SDRAM_DCR_ADDR_OFFSET,
268 dcr_host->base + SDRAM_DCR_DATA_OFFSET,
269 idcr_n,
270 value);
274 * ppc4xx_edac_check_bank_error - check a bank for an ECC bank error
275 * @status: A pointer to the ECC status structure to check for an
276 * ECC bank error.
277 * @bank: The bank to check for an ECC error.
279 * This routine determines whether the specified bank has an ECC
280 * error.
282 * Returns true if the specified bank has an ECC error; otherwise,
283 * false.
285 static bool
286 ppc4xx_edac_check_bank_error(const struct ppc4xx_ecc_status *status,
287 unsigned int bank)
289 switch (bank) {
290 case 0:
291 return status->ecces & SDRAM_ECCES_BK0ER;
292 case 1:
293 return status->ecces & SDRAM_ECCES_BK1ER;
294 default:
295 return false;
300 * ppc4xx_edac_generate_bank_message - generate interpretted bank status message
301 * @mci: A pointer to the EDAC memory controller instance associated
302 * with the bank message being generated.
303 * @status: A pointer to the ECC status structure to generate the
304 * message from.
305 * @buffer: A pointer to the buffer in which to generate the
306 * message.
307 * @size: The size, in bytes, of space available in buffer.
309 * This routine generates to the provided buffer the portion of the
310 * driver-unique report message associated with the ECCESS[BKNER]
311 * field of the specified ECC status.
313 * Returns the number of characters generated on success; otherwise, <
314 * 0 on error.
316 static int
317 ppc4xx_edac_generate_bank_message(const struct mem_ctl_info *mci,
318 const struct ppc4xx_ecc_status *status,
319 char *buffer,
320 size_t size)
322 int n, total = 0;
323 unsigned int row, rows;
325 n = snprintf(buffer, size, "%s: Banks: ", mci->dev_name);
327 if (n < 0 || n >= size)
328 goto fail;
330 buffer += n;
331 size -= n;
332 total += n;
334 for (rows = 0, row = 0; row < mci->nr_csrows; row++) {
335 if (ppc4xx_edac_check_bank_error(status, row)) {
336 n = snprintf(buffer, size, "%s%u",
337 (rows++ ? ", " : ""), row);
339 if (n < 0 || n >= size)
340 goto fail;
342 buffer += n;
343 size -= n;
344 total += n;
348 n = snprintf(buffer, size, "%s; ", rows ? "" : "None");
350 if (n < 0 || n >= size)
351 goto fail;
353 buffer += n;
354 size -= n;
355 total += n;
357 fail:
358 return total;
362 * ppc4xx_edac_generate_checkbit_message - generate interpretted checkbit message
363 * @mci: A pointer to the EDAC memory controller instance associated
364 * with the checkbit message being generated.
365 * @status: A pointer to the ECC status structure to generate the
366 * message from.
367 * @buffer: A pointer to the buffer in which to generate the
368 * message.
369 * @size: The size, in bytes, of space available in buffer.
371 * This routine generates to the provided buffer the portion of the
372 * driver-unique report message associated with the ECCESS[CKBER]
373 * field of the specified ECC status.
375 * Returns the number of characters generated on success; otherwise, <
376 * 0 on error.
378 static int
379 ppc4xx_edac_generate_checkbit_message(const struct mem_ctl_info *mci,
380 const struct ppc4xx_ecc_status *status,
381 char *buffer,
382 size_t size)
384 const struct ppc4xx_edac_pdata *pdata = mci->pvt_info;
385 const char *ckber = NULL;
387 switch (status->ecces & SDRAM_ECCES_CKBER_MASK) {
388 case SDRAM_ECCES_CKBER_NONE:
389 ckber = "None";
390 break;
391 case SDRAM_ECCES_CKBER_32_ECC_0_3:
392 ckber = "ECC0:3";
393 break;
394 case SDRAM_ECCES_CKBER_32_ECC_4_8:
395 switch (mfsdram(&pdata->dcr_host, SDRAM_MCOPT1) &
396 SDRAM_MCOPT1_WDTH_MASK) {
397 case SDRAM_MCOPT1_WDTH_16:
398 ckber = "ECC0:3";
399 break;
400 case SDRAM_MCOPT1_WDTH_32:
401 ckber = "ECC4:8";
402 break;
403 default:
404 ckber = "Unknown";
405 break;
407 break;
408 case SDRAM_ECCES_CKBER_32_ECC_0_8:
409 ckber = "ECC0:8";
410 break;
411 default:
412 ckber = "Unknown";
413 break;
416 return snprintf(buffer, size, "Checkbit Error: %s", ckber);
420 * ppc4xx_edac_generate_lane_message - generate interpretted byte lane message
421 * @mci: A pointer to the EDAC memory controller instance associated
422 * with the byte lane message being generated.
423 * @status: A pointer to the ECC status structure to generate the
424 * message from.
425 * @buffer: A pointer to the buffer in which to generate the
426 * message.
427 * @size: The size, in bytes, of space available in buffer.
429 * This routine generates to the provided buffer the portion of the
430 * driver-unique report message associated with the ECCESS[BNCE]
431 * field of the specified ECC status.
433 * Returns the number of characters generated on success; otherwise, <
434 * 0 on error.
436 static int
437 ppc4xx_edac_generate_lane_message(const struct mem_ctl_info *mci,
438 const struct ppc4xx_ecc_status *status,
439 char *buffer,
440 size_t size)
442 int n, total = 0;
443 unsigned int lane, lanes;
444 const unsigned int first_lane = 0;
445 const unsigned int lane_count = 16;
447 n = snprintf(buffer, size, "; Byte Lane Errors: ");
449 if (n < 0 || n >= size)
450 goto fail;
452 buffer += n;
453 size -= n;
454 total += n;
456 for (lanes = 0, lane = first_lane; lane < lane_count; lane++) {
457 if ((status->ecces & SDRAM_ECCES_BNCE_ENCODE(lane)) != 0) {
458 n = snprintf(buffer, size,
459 "%s%u",
460 (lanes++ ? ", " : ""), lane);
462 if (n < 0 || n >= size)
463 goto fail;
465 buffer += n;
466 size -= n;
467 total += n;
471 n = snprintf(buffer, size, "%s; ", lanes ? "" : "None");
473 if (n < 0 || n >= size)
474 goto fail;
476 buffer += n;
477 size -= n;
478 total += n;
480 fail:
481 return total;
485 * ppc4xx_edac_generate_ecc_message - generate interpretted ECC status message
486 * @mci: A pointer to the EDAC memory controller instance associated
487 * with the ECCES message being generated.
488 * @status: A pointer to the ECC status structure to generate the
489 * message from.
490 * @buffer: A pointer to the buffer in which to generate the
491 * message.
492 * @size: The size, in bytes, of space available in buffer.
494 * This routine generates to the provided buffer the portion of the
495 * driver-unique report message associated with the ECCESS register of
496 * the specified ECC status.
498 * Returns the number of characters generated on success; otherwise, <
499 * 0 on error.
501 static int
502 ppc4xx_edac_generate_ecc_message(const struct mem_ctl_info *mci,
503 const struct ppc4xx_ecc_status *status,
504 char *buffer,
505 size_t size)
507 int n, total = 0;
509 n = ppc4xx_edac_generate_bank_message(mci, status, buffer, size);
511 if (n < 0 || n >= size)
512 goto fail;
514 buffer += n;
515 size -= n;
516 total += n;
518 n = ppc4xx_edac_generate_checkbit_message(mci, status, buffer, size);
520 if (n < 0 || n >= size)
521 goto fail;
523 buffer += n;
524 size -= n;
525 total += n;
527 n = ppc4xx_edac_generate_lane_message(mci, status, buffer, size);
529 if (n < 0 || n >= size)
530 goto fail;
532 buffer += n;
533 size -= n;
534 total += n;
536 fail:
537 return total;
541 * ppc4xx_edac_generate_plb_message - generate interpretted PLB status message
542 * @mci: A pointer to the EDAC memory controller instance associated
543 * with the PLB message being generated.
544 * @status: A pointer to the ECC status structure to generate the
545 * message from.
546 * @buffer: A pointer to the buffer in which to generate the
547 * message.
548 * @size: The size, in bytes, of space available in buffer.
550 * This routine generates to the provided buffer the portion of the
551 * driver-unique report message associated with the PLB-related BESR
552 * and/or WMIRQ registers of the specified ECC status.
554 * Returns the number of characters generated on success; otherwise, <
555 * 0 on error.
557 static int
558 ppc4xx_edac_generate_plb_message(const struct mem_ctl_info *mci,
559 const struct ppc4xx_ecc_status *status,
560 char *buffer,
561 size_t size)
563 unsigned int master;
564 bool read;
566 if ((status->besr & SDRAM_BESR_MASK) == 0)
567 return 0;
569 if ((status->besr & SDRAM_BESR_M0ET_MASK) == SDRAM_BESR_M0ET_NONE)
570 return 0;
572 read = ((status->besr & SDRAM_BESR_M0RW_MASK) == SDRAM_BESR_M0RW_READ);
574 master = SDRAM_BESR_M0ID_DECODE(status->besr);
576 return snprintf(buffer, size,
577 "%s error w/ PLB master %u \"%s\"; ",
578 (read ? "Read" : "Write"),
579 master,
580 (((master >= SDRAM_PLB_M0ID_FIRST) &&
581 (master <= SDRAM_PLB_M0ID_LAST)) ?
582 ppc4xx_plb_masters[master] : "UNKNOWN"));
586 * ppc4xx_edac_generate_message - generate interpretted status message
587 * @mci: A pointer to the EDAC memory controller instance associated
588 * with the driver-unique message being generated.
589 * @status: A pointer to the ECC status structure to generate the
590 * message from.
591 * @buffer: A pointer to the buffer in which to generate the
592 * message.
593 * @size: The size, in bytes, of space available in buffer.
595 * This routine generates to the provided buffer the driver-unique
596 * EDAC report message from the specified ECC status.
598 static void
599 ppc4xx_edac_generate_message(const struct mem_ctl_info *mci,
600 const struct ppc4xx_ecc_status *status,
601 char *buffer,
602 size_t size)
604 int n;
606 if (buffer == NULL || size == 0)
607 return;
609 n = ppc4xx_edac_generate_ecc_message(mci, status, buffer, size);
611 if (n < 0 || n >= size)
612 return;
614 buffer += n;
615 size -= n;
617 ppc4xx_edac_generate_plb_message(mci, status, buffer, size);
620 #ifdef DEBUG
622 * ppc4xx_ecc_dump_status - dump controller ECC status registers
623 * @mci: A pointer to the EDAC memory controller instance
624 * associated with the status being dumped.
625 * @status: A pointer to the ECC status structure to generate the
626 * dump from.
628 * This routine dumps to the kernel log buffer the raw and
629 * interpretted specified ECC status.
631 static void
632 ppc4xx_ecc_dump_status(const struct mem_ctl_info *mci,
633 const struct ppc4xx_ecc_status *status)
635 char message[PPC4XX_EDAC_MESSAGE_SIZE];
637 ppc4xx_edac_generate_message(mci, status, message, sizeof(message));
639 ppc4xx_edac_mc_printk(KERN_INFO, mci,
640 "\n"
641 "\tECCES: 0x%08x\n"
642 "\tWMIRQ: 0x%08x\n"
643 "\tBESR: 0x%08x\n"
644 "\tBEAR: 0x%08x%08x\n"
645 "\t%s\n",
646 status->ecces,
647 status->wmirq,
648 status->besr,
649 status->bearh,
650 status->bearl,
651 message);
653 #endif /* DEBUG */
656 * ppc4xx_ecc_get_status - get controller ECC status
657 * @mci: A pointer to the EDAC memory controller instance
658 * associated with the status being retrieved.
659 * @status: A pointer to the ECC status structure to populate the
660 * ECC status with.
662 * This routine reads and masks, as appropriate, all the relevant
663 * status registers that deal with ibm,sdram-4xx-ddr2 ECC errors.
664 * While we read all of them, for correctable errors, we only expect
665 * to deal with ECCES. For uncorrectable errors, we expect to deal
666 * with all of them.
668 static void
669 ppc4xx_ecc_get_status(const struct mem_ctl_info *mci,
670 struct ppc4xx_ecc_status *status)
672 const struct ppc4xx_edac_pdata *pdata = mci->pvt_info;
673 const dcr_host_t *dcr_host = &pdata->dcr_host;
675 status->ecces = mfsdram(dcr_host, SDRAM_ECCES) & SDRAM_ECCES_MASK;
676 status->wmirq = mfsdram(dcr_host, SDRAM_WMIRQ) & SDRAM_WMIRQ_MASK;
677 status->besr = mfsdram(dcr_host, SDRAM_BESR) & SDRAM_BESR_MASK;
678 status->bearl = mfsdram(dcr_host, SDRAM_BEARL);
679 status->bearh = mfsdram(dcr_host, SDRAM_BEARH);
683 * ppc4xx_ecc_clear_status - clear controller ECC status
684 * @mci: A pointer to the EDAC memory controller instance
685 * associated with the status being cleared.
686 * @status: A pointer to the ECC status structure containing the
687 * values to write to clear the ECC status.
689 * This routine clears--by writing the masked (as appropriate) status
690 * values back to--the status registers that deal with
691 * ibm,sdram-4xx-ddr2 ECC errors.
693 static void
694 ppc4xx_ecc_clear_status(const struct mem_ctl_info *mci,
695 const struct ppc4xx_ecc_status *status)
697 const struct ppc4xx_edac_pdata *pdata = mci->pvt_info;
698 const dcr_host_t *dcr_host = &pdata->dcr_host;
700 mtsdram(dcr_host, SDRAM_ECCES, status->ecces & SDRAM_ECCES_MASK);
701 mtsdram(dcr_host, SDRAM_WMIRQ, status->wmirq & SDRAM_WMIRQ_MASK);
702 mtsdram(dcr_host, SDRAM_BESR, status->besr & SDRAM_BESR_MASK);
703 mtsdram(dcr_host, SDRAM_BEARL, 0);
704 mtsdram(dcr_host, SDRAM_BEARH, 0);
708 * ppc4xx_edac_handle_ce - handle controller correctable ECC error (CE)
709 * @mci: A pointer to the EDAC memory controller instance
710 * associated with the correctable error being handled and reported.
711 * @status: A pointer to the ECC status structure associated with
712 * the correctable error being handled and reported.
714 * This routine handles an ibm,sdram-4xx-ddr2 controller ECC
715 * correctable error. Per the aforementioned discussion, there's not
716 * enough status available to use the full EDAC correctable error
717 * interface, so we just pass driver-unique message to the "no info"
718 * interface.
720 static void
721 ppc4xx_edac_handle_ce(struct mem_ctl_info *mci,
722 const struct ppc4xx_ecc_status *status)
724 int row;
725 char message[PPC4XX_EDAC_MESSAGE_SIZE];
727 ppc4xx_edac_generate_message(mci, status, message, sizeof(message));
729 for (row = 0; row < mci->nr_csrows; row++)
730 if (ppc4xx_edac_check_bank_error(status, row))
731 edac_mc_handle_ce_no_info(mci, message);
735 * ppc4xx_edac_handle_ue - handle controller uncorrectable ECC error (UE)
736 * @mci: A pointer to the EDAC memory controller instance
737 * associated with the uncorrectable error being handled and
738 * reported.
739 * @status: A pointer to the ECC status structure associated with
740 * the uncorrectable error being handled and reported.
742 * This routine handles an ibm,sdram-4xx-ddr2 controller ECC
743 * uncorrectable error.
745 static void
746 ppc4xx_edac_handle_ue(struct mem_ctl_info *mci,
747 const struct ppc4xx_ecc_status *status)
749 const u64 bear = ((u64)status->bearh << 32 | status->bearl);
750 const unsigned long page = bear >> PAGE_SHIFT;
751 const unsigned long offset = bear & ~PAGE_MASK;
752 int row;
753 char message[PPC4XX_EDAC_MESSAGE_SIZE];
755 ppc4xx_edac_generate_message(mci, status, message, sizeof(message));
757 for (row = 0; row < mci->nr_csrows; row++)
758 if (ppc4xx_edac_check_bank_error(status, row))
759 edac_mc_handle_ue(mci, page, offset, row, message);
763 * ppc4xx_edac_check - check controller for ECC errors
764 * @mci: A pointer to the EDAC memory controller instance
765 * associated with the ibm,sdram-4xx-ddr2 controller being
766 * checked.
768 * This routine is used to check and post ECC errors and is called by
769 * both the EDAC polling thread and this driver's CE and UE interrupt
770 * handler.
772 static void
773 ppc4xx_edac_check(struct mem_ctl_info *mci)
775 #ifdef DEBUG
776 static unsigned int count;
777 #endif
778 struct ppc4xx_ecc_status status;
780 ppc4xx_ecc_get_status(mci, &status);
782 #ifdef DEBUG
783 if (count++ % 30 == 0)
784 ppc4xx_ecc_dump_status(mci, &status);
785 #endif
787 if (status.ecces & SDRAM_ECCES_UE)
788 ppc4xx_edac_handle_ue(mci, &status);
790 if (status.ecces & SDRAM_ECCES_CE)
791 ppc4xx_edac_handle_ce(mci, &status);
793 ppc4xx_ecc_clear_status(mci, &status);
797 * ppc4xx_edac_isr - SEC (CE) and DED (UE) interrupt service routine
798 * @irq: The virtual interrupt number being serviced.
799 * @dev_id: A pointer to the EDAC memory controller instance
800 * associated with the interrupt being handled.
802 * This routine implements the interrupt handler for both correctable
803 * (CE) and uncorrectable (UE) ECC errors for the ibm,sdram-4xx-ddr2
804 * controller. It simply calls through to the same routine used during
805 * polling to check, report and clear the ECC status.
807 * Unconditionally returns IRQ_HANDLED.
809 static irqreturn_t
810 ppc4xx_edac_isr(int irq, void *dev_id)
812 struct mem_ctl_info *mci = dev_id;
814 ppc4xx_edac_check(mci);
816 return IRQ_HANDLED;
820 * ppc4xx_edac_get_dtype - return the controller memory width
821 * @mcopt1: The 32-bit Memory Controller Option 1 register value
822 * currently set for the controller, from which the width
823 * is derived.
825 * This routine returns the EDAC device type width appropriate for the
826 * current controller configuration.
828 * TODO: This needs to be conditioned dynamically through feature
829 * flags or some such when other controller variants are supported as
830 * the 405EX[r] is 16-/32-bit and the others are 32-/64-bit with the
831 * 16- and 64-bit field definition/value/enumeration (b1) overloaded
832 * among them.
834 * Returns a device type width enumeration.
836 static enum dev_type __devinit
837 ppc4xx_edac_get_dtype(u32 mcopt1)
839 switch (mcopt1 & SDRAM_MCOPT1_WDTH_MASK) {
840 case SDRAM_MCOPT1_WDTH_16:
841 return DEV_X2;
842 case SDRAM_MCOPT1_WDTH_32:
843 return DEV_X4;
844 default:
845 return DEV_UNKNOWN;
850 * ppc4xx_edac_get_mtype - return controller memory type
851 * @mcopt1: The 32-bit Memory Controller Option 1 register value
852 * currently set for the controller, from which the memory type
853 * is derived.
855 * This routine returns the EDAC memory type appropriate for the
856 * current controller configuration.
858 * Returns a memory type enumeration.
860 static enum mem_type __devinit
861 ppc4xx_edac_get_mtype(u32 mcopt1)
863 bool rden = ((mcopt1 & SDRAM_MCOPT1_RDEN_MASK) == SDRAM_MCOPT1_RDEN);
865 switch (mcopt1 & SDRAM_MCOPT1_DDR_TYPE_MASK) {
866 case SDRAM_MCOPT1_DDR2_TYPE:
867 return rden ? MEM_RDDR2 : MEM_DDR2;
868 case SDRAM_MCOPT1_DDR1_TYPE:
869 return rden ? MEM_RDDR : MEM_DDR;
870 default:
871 return MEM_UNKNOWN;
876 * ppc4xx_edac_init_csrows - initialize driver instance rows
877 * @mci: A pointer to the EDAC memory controller instance
878 * associated with the ibm,sdram-4xx-ddr2 controller for which
879 * the csrows (i.e. banks/ranks) are being initialized.
880 * @mcopt1: The 32-bit Memory Controller Option 1 register value
881 * currently set for the controller, from which bank width
882 * and memory typ information is derived.
884 * This routine initializes the virtual "chip select rows" associated
885 * with the EDAC memory controller instance. An ibm,sdram-4xx-ddr2
886 * controller bank/rank is mapped to a row.
888 * Returns 0 if OK; otherwise, -EINVAL if the memory bank size
889 * configuration cannot be determined.
891 static int __devinit
892 ppc4xx_edac_init_csrows(struct mem_ctl_info *mci, u32 mcopt1)
894 const struct ppc4xx_edac_pdata *pdata = mci->pvt_info;
895 int status = 0;
896 enum mem_type mtype;
897 enum dev_type dtype;
898 enum edac_type edac_mode;
899 int row;
900 u32 mbxcf, size;
901 static u32 ppc4xx_last_page;
903 /* Establish the memory type and width */
905 mtype = ppc4xx_edac_get_mtype(mcopt1);
906 dtype = ppc4xx_edac_get_dtype(mcopt1);
908 /* Establish EDAC mode */
910 if (mci->edac_cap & EDAC_FLAG_SECDED)
911 edac_mode = EDAC_SECDED;
912 else if (mci->edac_cap & EDAC_FLAG_EC)
913 edac_mode = EDAC_EC;
914 else
915 edac_mode = EDAC_NONE;
918 * Initialize each chip select row structure which correspond
919 * 1:1 with a controller bank/rank.
922 for (row = 0; row < mci->nr_csrows; row++) {
923 struct csrow_info *csi = &mci->csrows[row];
926 * Get the configuration settings for this
927 * row/bank/rank and skip disabled banks.
930 mbxcf = mfsdram(&pdata->dcr_host, SDRAM_MBXCF(row));
932 if ((mbxcf & SDRAM_MBCF_BE_MASK) != SDRAM_MBCF_BE_ENABLE)
933 continue;
935 /* Map the bank configuration size setting to pages. */
937 size = mbxcf & SDRAM_MBCF_SZ_MASK;
939 switch (size) {
940 case SDRAM_MBCF_SZ_4MB:
941 case SDRAM_MBCF_SZ_8MB:
942 case SDRAM_MBCF_SZ_16MB:
943 case SDRAM_MBCF_SZ_32MB:
944 case SDRAM_MBCF_SZ_64MB:
945 case SDRAM_MBCF_SZ_128MB:
946 case SDRAM_MBCF_SZ_256MB:
947 case SDRAM_MBCF_SZ_512MB:
948 case SDRAM_MBCF_SZ_1GB:
949 case SDRAM_MBCF_SZ_2GB:
950 case SDRAM_MBCF_SZ_4GB:
951 case SDRAM_MBCF_SZ_8GB:
952 csi->nr_pages = SDRAM_MBCF_SZ_TO_PAGES(size);
953 break;
954 default:
955 ppc4xx_edac_mc_printk(KERN_ERR, mci,
956 "Unrecognized memory bank %d "
957 "size 0x%08x\n",
958 row, SDRAM_MBCF_SZ_DECODE(size));
959 status = -EINVAL;
960 goto done;
963 csi->first_page = ppc4xx_last_page;
964 csi->last_page = csi->first_page + csi->nr_pages - 1;
965 csi->page_mask = 0;
968 * It's unclear exactly what grain should be set to
969 * here. The SDRAM_ECCES register allows resolution of
970 * an error down to a nibble which would potentially
971 * argue for a grain of '1' byte, even though we only
972 * know the associated address for uncorrectable
973 * errors. This value is not used at present for
974 * anything other than error reporting so getting it
975 * wrong should be of little consequence. Other
976 * possible values would be the PLB width (16), the
977 * page size (PAGE_SIZE) or the memory width (2 or 4).
980 csi->grain = 1;
982 csi->mtype = mtype;
983 csi->dtype = dtype;
985 csi->edac_mode = edac_mode;
987 ppc4xx_last_page += csi->nr_pages;
990 done:
991 return status;
995 * ppc4xx_edac_mc_init - initialize driver instance
996 * @mci: A pointer to the EDAC memory controller instance being
997 * initialized.
998 * @op: A pointer to the OpenFirmware device tree node associated
999 * with the controller this EDAC instance is bound to.
1000 * @match: A pointer to the OpenFirmware device tree match
1001 * information associated with the controller this EDAC instance
1002 * is bound to.
1003 * @dcr_host: A pointer to the DCR data containing the DCR mapping
1004 * for this controller instance.
1005 * @mcopt1: The 32-bit Memory Controller Option 1 register value
1006 * currently set for the controller, from which ECC capabilities
1007 * and scrub mode are derived.
1009 * This routine performs initialization of the EDAC memory controller
1010 * instance and related driver-private data associated with the
1011 * ibm,sdram-4xx-ddr2 memory controller the instance is bound to.
1013 * Returns 0 if OK; otherwise, < 0 on error.
1015 static int __devinit
1016 ppc4xx_edac_mc_init(struct mem_ctl_info *mci,
1017 struct platform_device *op,
1018 const struct of_device_id *match,
1019 const dcr_host_t *dcr_host,
1020 u32 mcopt1)
1022 int status = 0;
1023 const u32 memcheck = (mcopt1 & SDRAM_MCOPT1_MCHK_MASK);
1024 struct ppc4xx_edac_pdata *pdata = NULL;
1025 const struct device_node *np = op->dev.of_node;
1027 if (match == NULL)
1028 return -EINVAL;
1030 /* Initial driver pointers and private data */
1032 mci->dev = &op->dev;
1034 dev_set_drvdata(mci->dev, mci);
1036 pdata = mci->pvt_info;
1038 pdata->dcr_host = *dcr_host;
1039 pdata->irqs.sec = NO_IRQ;
1040 pdata->irqs.ded = NO_IRQ;
1042 /* Initialize controller capabilities and configuration */
1044 mci->mtype_cap = (MEM_FLAG_DDR | MEM_FLAG_RDDR |
1045 MEM_FLAG_DDR2 | MEM_FLAG_RDDR2);
1047 mci->edac_ctl_cap = (EDAC_FLAG_NONE |
1048 EDAC_FLAG_EC |
1049 EDAC_FLAG_SECDED);
1051 mci->scrub_cap = SCRUB_NONE;
1052 mci->scrub_mode = SCRUB_NONE;
1055 * Update the actual capabilites based on the MCOPT1[MCHK]
1056 * settings. Scrubbing is only useful if reporting is enabled.
1059 switch (memcheck) {
1060 case SDRAM_MCOPT1_MCHK_CHK:
1061 mci->edac_cap = EDAC_FLAG_EC;
1062 break;
1063 case SDRAM_MCOPT1_MCHK_CHK_REP:
1064 mci->edac_cap = (EDAC_FLAG_EC | EDAC_FLAG_SECDED);
1065 mci->scrub_mode = SCRUB_SW_SRC;
1066 break;
1067 default:
1068 mci->edac_cap = EDAC_FLAG_NONE;
1069 break;
1072 /* Initialize strings */
1074 mci->mod_name = PPC4XX_EDAC_MODULE_NAME;
1075 mci->mod_ver = PPC4XX_EDAC_MODULE_REVISION;
1076 mci->ctl_name = match->compatible,
1077 mci->dev_name = np->full_name;
1079 /* Initialize callbacks */
1081 mci->edac_check = ppc4xx_edac_check;
1082 mci->ctl_page_to_phys = NULL;
1084 /* Initialize chip select rows */
1086 status = ppc4xx_edac_init_csrows(mci, mcopt1);
1088 if (status)
1089 ppc4xx_edac_mc_printk(KERN_ERR, mci,
1090 "Failed to initialize rows!\n");
1092 return status;
1096 * ppc4xx_edac_register_irq - setup and register controller interrupts
1097 * @op: A pointer to the OpenFirmware device tree node associated
1098 * with the controller this EDAC instance is bound to.
1099 * @mci: A pointer to the EDAC memory controller instance
1100 * associated with the ibm,sdram-4xx-ddr2 controller for which
1101 * interrupts are being registered.
1103 * This routine parses the correctable (CE) and uncorrectable error (UE)
1104 * interrupts from the device tree node and maps and assigns them to
1105 * the associated EDAC memory controller instance.
1107 * Returns 0 if OK; otherwise, -ENODEV if the interrupts could not be
1108 * mapped and assigned.
1110 static int __devinit
1111 ppc4xx_edac_register_irq(struct platform_device *op, struct mem_ctl_info *mci)
1113 int status = 0;
1114 int ded_irq, sec_irq;
1115 struct ppc4xx_edac_pdata *pdata = mci->pvt_info;
1116 struct device_node *np = op->dev.of_node;
1118 ded_irq = irq_of_parse_and_map(np, INTMAP_ECCDED_INDEX);
1119 sec_irq = irq_of_parse_and_map(np, INTMAP_ECCSEC_INDEX);
1121 if (ded_irq == NO_IRQ || sec_irq == NO_IRQ) {
1122 ppc4xx_edac_mc_printk(KERN_ERR, mci,
1123 "Unable to map interrupts.\n");
1124 status = -ENODEV;
1125 goto fail;
1128 status = request_irq(ded_irq,
1129 ppc4xx_edac_isr,
1130 IRQF_DISABLED,
1131 "[EDAC] MC ECCDED",
1132 mci);
1134 if (status < 0) {
1135 ppc4xx_edac_mc_printk(KERN_ERR, mci,
1136 "Unable to request irq %d for ECC DED",
1137 ded_irq);
1138 status = -ENODEV;
1139 goto fail1;
1142 status = request_irq(sec_irq,
1143 ppc4xx_edac_isr,
1144 IRQF_DISABLED,
1145 "[EDAC] MC ECCSEC",
1146 mci);
1148 if (status < 0) {
1149 ppc4xx_edac_mc_printk(KERN_ERR, mci,
1150 "Unable to request irq %d for ECC SEC",
1151 sec_irq);
1152 status = -ENODEV;
1153 goto fail2;
1156 ppc4xx_edac_mc_printk(KERN_INFO, mci, "ECCDED irq is %d\n", ded_irq);
1157 ppc4xx_edac_mc_printk(KERN_INFO, mci, "ECCSEC irq is %d\n", sec_irq);
1159 pdata->irqs.ded = ded_irq;
1160 pdata->irqs.sec = sec_irq;
1162 return 0;
1164 fail2:
1165 free_irq(sec_irq, mci);
1167 fail1:
1168 free_irq(ded_irq, mci);
1170 fail:
1171 return status;
1175 * ppc4xx_edac_map_dcrs - locate and map controller registers
1176 * @np: A pointer to the device tree node containing the DCR
1177 * resources to map.
1178 * @dcr_host: A pointer to the DCR data to populate with the
1179 * DCR mapping.
1181 * This routine attempts to locate in the device tree and map the DCR
1182 * register resources associated with the controller's indirect DCR
1183 * address and data windows.
1185 * Returns 0 if the DCRs were successfully mapped; otherwise, < 0 on
1186 * error.
1188 static int __devinit
1189 ppc4xx_edac_map_dcrs(const struct device_node *np, dcr_host_t *dcr_host)
1191 unsigned int dcr_base, dcr_len;
1193 if (np == NULL || dcr_host == NULL)
1194 return -EINVAL;
1196 /* Get the DCR resource extent and sanity check the values. */
1198 dcr_base = dcr_resource_start(np, 0);
1199 dcr_len = dcr_resource_len(np, 0);
1201 if (dcr_base == 0 || dcr_len == 0) {
1202 ppc4xx_edac_printk(KERN_ERR,
1203 "Failed to obtain DCR property.\n");
1204 return -ENODEV;
1207 if (dcr_len != SDRAM_DCR_RESOURCE_LEN) {
1208 ppc4xx_edac_printk(KERN_ERR,
1209 "Unexpected DCR length %d, expected %d.\n",
1210 dcr_len, SDRAM_DCR_RESOURCE_LEN);
1211 return -ENODEV;
1214 /* Attempt to map the DCR extent. */
1216 *dcr_host = dcr_map(np, dcr_base, dcr_len);
1218 if (!DCR_MAP_OK(*dcr_host)) {
1219 ppc4xx_edac_printk(KERN_INFO, "Failed to map DCRs.\n");
1220 return -ENODEV;
1223 return 0;
1227 * ppc4xx_edac_probe - check controller and bind driver
1228 * @op: A pointer to the OpenFirmware device tree node associated
1229 * with the controller being probed for driver binding.
1230 * @match: A pointer to the OpenFirmware device tree match
1231 * information associated with the controller being probed
1232 * for driver binding.
1234 * This routine probes a specific ibm,sdram-4xx-ddr2 controller
1235 * instance for binding with the driver.
1237 * Returns 0 if the controller instance was successfully bound to the
1238 * driver; otherwise, < 0 on error.
1240 static int __devinit
1241 ppc4xx_edac_probe(struct platform_device *op, const struct of_device_id *match)
1243 int status = 0;
1244 u32 mcopt1, memcheck;
1245 dcr_host_t dcr_host;
1246 const struct device_node *np = op->dev.of_node;
1247 struct mem_ctl_info *mci = NULL;
1248 static int ppc4xx_edac_instance;
1251 * At this point, we only support the controller realized on
1252 * the AMCC PPC 405EX[r]. Reject anything else.
1255 if (!of_device_is_compatible(np, "ibm,sdram-405ex") &&
1256 !of_device_is_compatible(np, "ibm,sdram-405exr")) {
1257 ppc4xx_edac_printk(KERN_NOTICE,
1258 "Only the PPC405EX[r] is supported.\n");
1259 return -ENODEV;
1263 * Next, get the DCR property and attempt to map it so that we
1264 * can probe the controller.
1267 status = ppc4xx_edac_map_dcrs(np, &dcr_host);
1269 if (status)
1270 return status;
1273 * First determine whether ECC is enabled at all. If not,
1274 * there is no useful checking or monitoring that can be done
1275 * for this controller.
1278 mcopt1 = mfsdram(&dcr_host, SDRAM_MCOPT1);
1279 memcheck = (mcopt1 & SDRAM_MCOPT1_MCHK_MASK);
1281 if (memcheck == SDRAM_MCOPT1_MCHK_NON) {
1282 ppc4xx_edac_printk(KERN_INFO, "%s: No ECC memory detected or "
1283 "ECC is disabled.\n", np->full_name);
1284 status = -ENODEV;
1285 goto done;
1289 * At this point, we know ECC is enabled, allocate an EDAC
1290 * controller instance and perform the appropriate
1291 * initialization.
1294 mci = edac_mc_alloc(sizeof(struct ppc4xx_edac_pdata),
1295 ppc4xx_edac_nr_csrows,
1296 ppc4xx_edac_nr_chans,
1297 ppc4xx_edac_instance);
1299 if (mci == NULL) {
1300 ppc4xx_edac_printk(KERN_ERR, "%s: "
1301 "Failed to allocate EDAC MC instance!\n",
1302 np->full_name);
1303 status = -ENOMEM;
1304 goto done;
1307 status = ppc4xx_edac_mc_init(mci, op, match, &dcr_host, mcopt1);
1309 if (status) {
1310 ppc4xx_edac_mc_printk(KERN_ERR, mci,
1311 "Failed to initialize instance!\n");
1312 goto fail;
1316 * We have a valid, initialized EDAC instance bound to the
1317 * controller. Attempt to register it with the EDAC subsystem
1318 * and, if necessary, register interrupts.
1321 if (edac_mc_add_mc(mci)) {
1322 ppc4xx_edac_mc_printk(KERN_ERR, mci,
1323 "Failed to add instance!\n");
1324 status = -ENODEV;
1325 goto fail;
1328 if (edac_op_state == EDAC_OPSTATE_INT) {
1329 status = ppc4xx_edac_register_irq(op, mci);
1331 if (status)
1332 goto fail1;
1335 ppc4xx_edac_instance++;
1337 return 0;
1339 fail1:
1340 edac_mc_del_mc(mci->dev);
1342 fail:
1343 edac_mc_free(mci);
1345 done:
1346 return status;
1350 * ppc4xx_edac_remove - unbind driver from controller
1351 * @op: A pointer to the OpenFirmware device tree node associated
1352 * with the controller this EDAC instance is to be unbound/removed
1353 * from.
1355 * This routine unbinds the EDAC memory controller instance associated
1356 * with the specified ibm,sdram-4xx-ddr2 controller described by the
1357 * OpenFirmware device tree node passed as a parameter.
1359 * Unconditionally returns 0.
1361 static int
1362 ppc4xx_edac_remove(struct platform_device *op)
1364 struct mem_ctl_info *mci = dev_get_drvdata(&op->dev);
1365 struct ppc4xx_edac_pdata *pdata = mci->pvt_info;
1367 if (edac_op_state == EDAC_OPSTATE_INT) {
1368 free_irq(pdata->irqs.sec, mci);
1369 free_irq(pdata->irqs.ded, mci);
1372 dcr_unmap(pdata->dcr_host, SDRAM_DCR_RESOURCE_LEN);
1374 edac_mc_del_mc(mci->dev);
1375 edac_mc_free(mci);
1377 return 0;
1381 * ppc4xx_edac_opstate_init - initialize EDAC reporting method
1383 * This routine ensures that the EDAC memory controller reporting
1384 * method is mapped to a sane value as the EDAC core defines the value
1385 * to EDAC_OPSTATE_INVAL by default. We don't call the global
1386 * opstate_init as that defaults to polling and we want interrupt as
1387 * the default.
1389 static inline void __init
1390 ppc4xx_edac_opstate_init(void)
1392 switch (edac_op_state) {
1393 case EDAC_OPSTATE_POLL:
1394 case EDAC_OPSTATE_INT:
1395 break;
1396 default:
1397 edac_op_state = EDAC_OPSTATE_INT;
1398 break;
1401 ppc4xx_edac_printk(KERN_INFO, "Reporting type: %s\n",
1402 ((edac_op_state == EDAC_OPSTATE_POLL) ?
1403 EDAC_OPSTATE_POLL_STR :
1404 ((edac_op_state == EDAC_OPSTATE_INT) ?
1405 EDAC_OPSTATE_INT_STR :
1406 EDAC_OPSTATE_UNKNOWN_STR)));
1410 * ppc4xx_edac_init - driver/module insertion entry point
1412 * This routine is the driver/module insertion entry point. It
1413 * initializes the EDAC memory controller reporting state and
1414 * registers the driver as an OpenFirmware device tree platform
1415 * driver.
1417 static int __init
1418 ppc4xx_edac_init(void)
1420 ppc4xx_edac_printk(KERN_INFO, PPC4XX_EDAC_MODULE_REVISION "\n");
1422 ppc4xx_edac_opstate_init();
1424 return of_register_platform_driver(&ppc4xx_edac_driver);
1428 * ppc4xx_edac_exit - driver/module removal entry point
1430 * This routine is the driver/module removal entry point. It
1431 * unregisters the driver as an OpenFirmware device tree platform
1432 * driver.
1434 static void __exit
1435 ppc4xx_edac_exit(void)
1437 of_unregister_platform_driver(&ppc4xx_edac_driver);
1440 module_init(ppc4xx_edac_init);
1441 module_exit(ppc4xx_edac_exit);
1443 MODULE_LICENSE("GPL v2");
1444 MODULE_AUTHOR("Grant Erickson <gerickson@nuovations.com>");
1445 MODULE_DESCRIPTION("EDAC MC Driver for the PPC4xx IBM DDR2 Memory Controller");
1446 module_param(edac_op_state, int, 0444);
1447 MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting State: "
1448 "0=" EDAC_OPSTATE_POLL_STR ", 2=" EDAC_OPSTATE_INT_STR);