blob: 663157b3a0d4bdf38bfc7137a8f0e474f1220f9e [file] [log] [blame]
Frank Haverkamp67f4add2013-12-09 13:30:41 +01001/**
2 * IBM Accelerator Family 'GenWQE'
3 *
4 * (C) Copyright IBM Corp. 2013
5 *
6 * Author: Frank Haverkamp <haver@linux.vnet.ibm.com>
7 * Author: Joerg-Stephan Vogt <jsvogt@de.ibm.com>
8 * Author: Michael Jung <mijung@de.ibm.com>
9 * Author: Michael Ruettger <michael@ibmra.de>
10 *
11 * This program is free software; you can redistribute it and/or modify
12 * it under the terms of the GNU General Public License (version 2 only)
13 * as published by the Free Software Foundation.
14 *
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 * GNU General Public License for more details.
19 */
20
21/*
22 * Miscelanous functionality used in the other GenWQE driver parts.
23 */
24
25#include <linux/kernel.h>
26#include <linux/dma-mapping.h>
27#include <linux/sched.h>
28#include <linux/vmalloc.h>
29#include <linux/page-flags.h>
30#include <linux/scatterlist.h>
31#include <linux/hugetlb.h>
32#include <linux/iommu.h>
33#include <linux/delay.h>
34#include <linux/pci.h>
35#include <linux/dma-mapping.h>
36#include <linux/ctype.h>
37#include <linux/module.h>
38#include <linux/platform_device.h>
39#include <linux/delay.h>
40#include <asm/pgtable.h>
41
42#include "genwqe_driver.h"
43#include "card_base.h"
44#include "card_ddcb.h"
45
46/**
47 * __genwqe_writeq() - Write 64-bit register
48 * @cd: genwqe device descriptor
49 * @byte_offs: byte offset within BAR
50 * @val: 64-bit value
51 *
52 * Return: 0 if success; < 0 if error
53 */
54int __genwqe_writeq(struct genwqe_dev *cd, u64 byte_offs, u64 val)
55{
56 if (cd->err_inject & GENWQE_INJECT_HARDWARE_FAILURE)
57 return -EIO;
58
59 if (cd->mmio == NULL)
60 return -EIO;
61
Frank Haverkamp58d66ce2013-12-20 16:26:10 +010062 __raw_writeq((__force u32)cpu_to_be64(val), cd->mmio + byte_offs);
Frank Haverkamp67f4add2013-12-09 13:30:41 +010063 return 0;
64}
65
66/**
67 * __genwqe_readq() - Read 64-bit register
68 * @cd: genwqe device descriptor
69 * @byte_offs: offset within BAR
70 *
71 * Return: value from register
72 */
73u64 __genwqe_readq(struct genwqe_dev *cd, u64 byte_offs)
74{
Frank Haverkamp67f4add2013-12-09 13:30:41 +010075 if (cd->err_inject & GENWQE_INJECT_HARDWARE_FAILURE)
76 return 0xffffffffffffffffull;
77
78 if ((cd->err_inject & GENWQE_INJECT_GFIR_FATAL) &&
79 (byte_offs == IO_SLC_CFGREG_GFIR))
80 return 0x000000000000ffffull;
81
82 if ((cd->err_inject & GENWQE_INJECT_GFIR_INFO) &&
83 (byte_offs == IO_SLC_CFGREG_GFIR))
84 return 0x00000000ffff0000ull;
85
86 if (cd->mmio == NULL)
87 return 0xffffffffffffffffull;
88
Frank Haverkamp58d66ce2013-12-20 16:26:10 +010089 return be64_to_cpu((__force __be64)__raw_readq(cd->mmio + byte_offs));
Frank Haverkamp67f4add2013-12-09 13:30:41 +010090}
91
92/**
93 * __genwqe_writel() - Write 32-bit register
94 * @cd: genwqe device descriptor
95 * @byte_offs: byte offset within BAR
96 * @val: 32-bit value
97 *
98 * Return: 0 if success; < 0 if error
99 */
100int __genwqe_writel(struct genwqe_dev *cd, u64 byte_offs, u32 val)
101{
102 if (cd->err_inject & GENWQE_INJECT_HARDWARE_FAILURE)
103 return -EIO;
104
105 if (cd->mmio == NULL)
106 return -EIO;
107
Frank Haverkamp58d66ce2013-12-20 16:26:10 +0100108 __raw_writel((__force u32)cpu_to_be32(val), cd->mmio + byte_offs);
Frank Haverkamp67f4add2013-12-09 13:30:41 +0100109 return 0;
110}
111
112/**
113 * __genwqe_readl() - Read 32-bit register
114 * @cd: genwqe device descriptor
115 * @byte_offs: offset within BAR
116 *
117 * Return: Value from register
118 */
119u32 __genwqe_readl(struct genwqe_dev *cd, u64 byte_offs)
120{
121 if (cd->err_inject & GENWQE_INJECT_HARDWARE_FAILURE)
122 return 0xffffffff;
123
124 if (cd->mmio == NULL)
125 return 0xffffffff;
126
Frank Haverkamp58d66ce2013-12-20 16:26:10 +0100127 return be32_to_cpu((__force __be32)__raw_readl(cd->mmio + byte_offs));
Frank Haverkamp67f4add2013-12-09 13:30:41 +0100128}
129
130/**
131 * genwqe_read_app_id() - Extract app_id
132 *
133 * app_unitcfg need to be filled with valid data first
134 */
135int genwqe_read_app_id(struct genwqe_dev *cd, char *app_name, int len)
136{
137 int i, j;
138 u32 app_id = (u32)cd->app_unitcfg;
139
140 memset(app_name, 0, len);
141 for (i = 0, j = 0; j < min(len, 4); j++) {
142 char ch = (char)((app_id >> (24 - j*8)) & 0xff);
143 if (ch == ' ')
144 continue;
145 app_name[i++] = isprint(ch) ? ch : 'X';
146 }
147 return i;
148}
149
150/**
151 * genwqe_init_crc32() - Prepare a lookup table for fast crc32 calculations
152 *
153 * Existing kernel functions seem to use a different polynom,
154 * therefore we could not use them here.
155 *
156 * Genwqe's Polynomial = 0x20044009
157 */
158#define CRC32_POLYNOMIAL 0x20044009
159static u32 crc32_tab[256]; /* crc32 lookup table */
160
161void genwqe_init_crc32(void)
162{
163 int i, j;
164 u32 crc;
165
166 for (i = 0; i < 256; i++) {
167 crc = i << 24;
168 for (j = 0; j < 8; j++) {
169 if (crc & 0x80000000)
170 crc = (crc << 1) ^ CRC32_POLYNOMIAL;
171 else
172 crc = (crc << 1);
173 }
174 crc32_tab[i] = crc;
175 }
176}
177
178/**
179 * genwqe_crc32() - Generate 32-bit crc as required for DDCBs
180 * @buff: pointer to data buffer
181 * @len: length of data for calculation
182 * @init: initial crc (0xffffffff at start)
183 *
184 * polynomial = x^32 * + x^29 + x^18 + x^14 + x^3 + 1 (0x20044009)
185
186 * Example: 4 bytes 0x01 0x02 0x03 0x04 with init=0xffffffff should
187 * result in a crc32 of 0xf33cb7d3.
188 *
189 * The existing kernel crc functions did not cover this polynom yet.
190 *
191 * Return: crc32 checksum.
192 */
193u32 genwqe_crc32(u8 *buff, size_t len, u32 init)
194{
195 int i;
196 u32 crc;
197
198 crc = init;
199 while (len--) {
200 i = ((crc >> 24) ^ *buff++) & 0xFF;
201 crc = (crc << 8) ^ crc32_tab[i];
202 }
203 return crc;
204}
205
206void *__genwqe_alloc_consistent(struct genwqe_dev *cd, size_t size,
207 dma_addr_t *dma_handle)
208{
209 if (get_order(size) > MAX_ORDER)
210 return NULL;
211
212 return pci_alloc_consistent(cd->pci_dev, size, dma_handle);
213}
214
215void __genwqe_free_consistent(struct genwqe_dev *cd, size_t size,
216 void *vaddr, dma_addr_t dma_handle)
217{
218 if (vaddr == NULL)
219 return;
220
221 pci_free_consistent(cd->pci_dev, size, vaddr, dma_handle);
222}
223
224static void genwqe_unmap_pages(struct genwqe_dev *cd, dma_addr_t *dma_list,
225 int num_pages)
226{
227 int i;
228 struct pci_dev *pci_dev = cd->pci_dev;
229
230 for (i = 0; (i < num_pages) && (dma_list[i] != 0x0); i++) {
231 pci_unmap_page(pci_dev, dma_list[i],
232 PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
233 dma_list[i] = 0x0;
234 }
235}
236
237static int genwqe_map_pages(struct genwqe_dev *cd,
238 struct page **page_list, int num_pages,
239 dma_addr_t *dma_list)
240{
241 int i;
242 struct pci_dev *pci_dev = cd->pci_dev;
243
244 /* establish DMA mapping for requested pages */
245 for (i = 0; i < num_pages; i++) {
246 dma_addr_t daddr;
247
248 dma_list[i] = 0x0;
249 daddr = pci_map_page(pci_dev, page_list[i],
250 0, /* map_offs */
251 PAGE_SIZE,
252 PCI_DMA_BIDIRECTIONAL); /* FIXME rd/rw */
253
254 if (pci_dma_mapping_error(pci_dev, daddr)) {
255 dev_err(&pci_dev->dev,
256 "[%s] err: no dma addr daddr=%016llx!\n",
257 __func__, (long long)daddr);
258 goto err;
259 }
260
261 dma_list[i] = daddr;
262 }
263 return 0;
264
265 err:
266 genwqe_unmap_pages(cd, dma_list, num_pages);
267 return -EIO;
268}
269
270static int genwqe_sgl_size(int num_pages)
271{
272 int len, num_tlb = num_pages / 7;
273
274 len = sizeof(struct sg_entry) * (num_pages+num_tlb + 1);
275 return roundup(len, PAGE_SIZE);
276}
277
278struct sg_entry *genwqe_alloc_sgl(struct genwqe_dev *cd, int num_pages,
279 dma_addr_t *dma_addr, size_t *sgl_size)
280{
281 struct pci_dev *pci_dev = cd->pci_dev;
282 struct sg_entry *sgl;
283
284 *sgl_size = genwqe_sgl_size(num_pages);
285 if (get_order(*sgl_size) > MAX_ORDER) {
286 dev_err(&pci_dev->dev,
287 "[%s] err: too much memory requested!\n", __func__);
288 return NULL;
289 }
290
291 sgl = __genwqe_alloc_consistent(cd, *sgl_size, dma_addr);
292 if (sgl == NULL) {
293 dev_err(&pci_dev->dev,
294 "[%s] err: no memory available!\n", __func__);
295 return NULL;
296 }
297
298 return sgl;
299}
300
301int genwqe_setup_sgl(struct genwqe_dev *cd,
302 unsigned long offs,
303 unsigned long size,
304 struct sg_entry *sgl,
305 dma_addr_t dma_addr, size_t sgl_size,
306 dma_addr_t *dma_list, int page_offs, int num_pages)
307{
308 int i = 0, j = 0, p;
309 unsigned long dma_offs, map_offs;
310 struct pci_dev *pci_dev = cd->pci_dev;
311 dma_addr_t prev_daddr = 0;
312 struct sg_entry *s, *last_s = NULL;
313
314 /* sanity checks */
315 if (offs > PAGE_SIZE) {
316 dev_err(&pci_dev->dev,
317 "[%s] too large start offs %08lx\n", __func__, offs);
318 return -EFAULT;
319 }
320 if (sgl_size < genwqe_sgl_size(num_pages)) {
321 dev_err(&pci_dev->dev,
322 "[%s] sgl_size too small %08lx for %d pages\n",
323 __func__, sgl_size, num_pages);
324 return -EFAULT;
325 }
326
327 dma_offs = 128; /* next block if needed/dma_offset */
328 map_offs = offs; /* offset in first page */
329
330 s = &sgl[0]; /* first set of 8 entries */
331 p = 0; /* page */
332 while (p < num_pages) {
333 dma_addr_t daddr;
334 unsigned int size_to_map;
335
336 /* always write the chaining entry, cleanup is done later */
337 j = 0;
338 s[j].target_addr = cpu_to_be64(dma_addr + dma_offs);
339 s[j].len = cpu_to_be32(128);
340 s[j].flags = cpu_to_be32(SG_CHAINED);
341 j++;
342
343 while (j < 8) {
344 /* DMA mapping for requested page, offs, size */
345 size_to_map = min(size, PAGE_SIZE - map_offs);
346 daddr = dma_list[page_offs + p] + map_offs;
347 size -= size_to_map;
348 map_offs = 0;
349
350 if (prev_daddr == daddr) {
351 u32 prev_len = be32_to_cpu(last_s->len);
352
353 /* pr_info("daddr combining: "
354 "%016llx/%08x -> %016llx\n",
355 prev_daddr, prev_len, daddr); */
356
357 last_s->len = cpu_to_be32(prev_len +
358 size_to_map);
359
360 p++; /* process next page */
361 if (p == num_pages)
362 goto fixup; /* nothing to do */
363
364 prev_daddr = daddr + size_to_map;
365 continue;
366 }
367
368 /* start new entry */
369 s[j].target_addr = cpu_to_be64(daddr);
370 s[j].len = cpu_to_be32(size_to_map);
371 s[j].flags = cpu_to_be32(SG_DATA);
372 prev_daddr = daddr + size_to_map;
373 last_s = &s[j];
374 j++;
375
376 p++; /* process next page */
377 if (p == num_pages)
378 goto fixup; /* nothing to do */
379 }
380 dma_offs += 128;
381 s += 8; /* continue 8 elements further */
382 }
383 fixup:
384 if (j == 1) { /* combining happend on last entry! */
385 s -= 8; /* full shift needed on previous sgl block */
386 j = 7; /* shift all elements */
387 }
388
389 for (i = 0; i < j; i++) /* move elements 1 up */
390 s[i] = s[i + 1];
391
392 s[i].target_addr = cpu_to_be64(0);
393 s[i].len = cpu_to_be32(0);
394 s[i].flags = cpu_to_be32(SG_END_LIST);
395 return 0;
396}
397
398void genwqe_free_sgl(struct genwqe_dev *cd, struct sg_entry *sg_list,
399 dma_addr_t dma_addr, size_t size)
400{
401 __genwqe_free_consistent(cd, size, sg_list, dma_addr);
402}
403
404/**
405 * free_user_pages() - Give pinned pages back
406 *
407 * Documentation of get_user_pages is in mm/memory.c:
408 *
409 * If the page is written to, set_page_dirty (or set_page_dirty_lock,
410 * as appropriate) must be called after the page is finished with, and
411 * before put_page is called.
412 *
413 * FIXME Could be of use to others and might belong in the generic
414 * code, if others agree. E.g.
415 * ll_free_user_pages in drivers/staging/lustre/lustre/llite/rw26.c
416 * ceph_put_page_vector in net/ceph/pagevec.c
417 * maybe more?
418 */
419static int free_user_pages(struct page **page_list, unsigned int nr_pages,
420 int dirty)
421{
422 unsigned int i;
423
424 for (i = 0; i < nr_pages; i++) {
425 if (page_list[i] != NULL) {
426 if (dirty)
427 set_page_dirty_lock(page_list[i]);
428 put_page(page_list[i]);
429 }
430 }
431 return 0;
432}
433
434/**
435 * genwqe_user_vmap() - Map user-space memory to virtual kernel memory
436 * @cd: pointer to genwqe device
437 * @m: mapping params
438 * @uaddr: user virtual address
439 * @size: size of memory to be mapped
440 *
441 * We need to think about how we could speed this up. Of course it is
442 * not a good idea to do this over and over again, like we are
443 * currently doing it. Nevertheless, I am curious where on the path
444 * the performance is spend. Most probably within the memory
445 * allocation functions, but maybe also in the DMA mapping code.
446 *
447 * Restrictions: The maximum size of the possible mapping currently depends
448 * on the amount of memory we can get using kzalloc() for the
449 * page_list and pci_alloc_consistent for the sg_list.
450 * The sg_list is currently itself not scattered, which could
451 * be fixed with some effort. The page_list must be split into
452 * PAGE_SIZE chunks too. All that will make the complicated
453 * code more complicated.
454 *
455 * Return: 0 if success
456 */
457int genwqe_user_vmap(struct genwqe_dev *cd, struct dma_mapping *m, void *uaddr,
458 unsigned long size, struct ddcb_requ *req)
459{
460 int rc = -EINVAL;
461 unsigned long data, offs;
462 struct pci_dev *pci_dev = cd->pci_dev;
463
464 if ((uaddr == NULL) || (size == 0)) {
465 m->size = 0; /* mark unused and not added */
466 return -EINVAL;
467 }
468 m->u_vaddr = uaddr;
469 m->size = size;
470
471 /* determine space needed for page_list. */
472 data = (unsigned long)uaddr;
473 offs = offset_in_page(data);
474 m->nr_pages = DIV_ROUND_UP(offs + size, PAGE_SIZE);
475
476 m->page_list = kcalloc(m->nr_pages,
477 sizeof(struct page *) + sizeof(dma_addr_t),
478 GFP_KERNEL);
479 if (!m->page_list) {
480 dev_err(&pci_dev->dev, "err: alloc page_list failed\n");
481 m->nr_pages = 0;
482 m->u_vaddr = NULL;
483 m->size = 0; /* mark unused and not added */
484 return -ENOMEM;
485 }
486 m->dma_list = (dma_addr_t *)(m->page_list + m->nr_pages);
487
488 /* pin user pages in memory */
489 rc = get_user_pages_fast(data & PAGE_MASK, /* page aligned addr */
490 m->nr_pages,
491 1, /* write by caller */
492 m->page_list); /* ptrs to pages */
493
494 /* assumption: get_user_pages can be killed by signals. */
495 if (rc < m->nr_pages) {
496 free_user_pages(m->page_list, rc, 0);
497 rc = -EFAULT;
498 goto fail_get_user_pages;
499 }
500
501 rc = genwqe_map_pages(cd, m->page_list, m->nr_pages, m->dma_list);
502 if (rc != 0)
503 goto fail_free_user_pages;
504
505 return 0;
506
507 fail_free_user_pages:
508 free_user_pages(m->page_list, m->nr_pages, 0);
509
510 fail_get_user_pages:
511 kfree(m->page_list);
512 m->page_list = NULL;
513 m->dma_list = NULL;
514 m->nr_pages = 0;
515 m->u_vaddr = NULL;
516 m->size = 0; /* mark unused and not added */
517 return rc;
518}
519
520/**
521 * genwqe_user_vunmap() - Undo mapping of user-space mem to virtual kernel
522 * memory
523 * @cd: pointer to genwqe device
524 * @m: mapping params
525 */
526int genwqe_user_vunmap(struct genwqe_dev *cd, struct dma_mapping *m,
527 struct ddcb_requ *req)
528{
529 struct pci_dev *pci_dev = cd->pci_dev;
530
531 if (!dma_mapping_used(m)) {
532 dev_err(&pci_dev->dev, "[%s] err: mapping %p not used!\n",
533 __func__, m);
534 return -EINVAL;
535 }
536
537 if (m->dma_list)
538 genwqe_unmap_pages(cd, m->dma_list, m->nr_pages);
539
540 if (m->page_list) {
541 free_user_pages(m->page_list, m->nr_pages, 1);
542
543 kfree(m->page_list);
544 m->page_list = NULL;
545 m->dma_list = NULL;
546 m->nr_pages = 0;
547 }
548
549 m->u_vaddr = NULL;
550 m->size = 0; /* mark as unused and not added */
551 return 0;
552}
553
554/**
555 * genwqe_card_type() - Get chip type SLU Configuration Register
556 * @cd: pointer to the genwqe device descriptor
557 * Return: 0: Altera Stratix-IV 230
558 * 1: Altera Stratix-IV 530
559 * 2: Altera Stratix-V A4
560 * 3: Altera Stratix-V A7
561 */
562u8 genwqe_card_type(struct genwqe_dev *cd)
563{
564 u64 card_type = cd->slu_unitcfg;
565 return (u8)((card_type & IO_SLU_UNITCFG_TYPE_MASK) >> 20);
566}
567
568/**
569 * genwqe_card_reset() - Reset the card
570 * @cd: pointer to the genwqe device descriptor
571 */
572int genwqe_card_reset(struct genwqe_dev *cd)
573{
574 u64 softrst;
575 struct pci_dev *pci_dev = cd->pci_dev;
576
577 if (!genwqe_is_privileged(cd))
578 return -ENODEV;
579
580 /* new SL */
581 __genwqe_writeq(cd, IO_SLC_CFGREG_SOFTRESET, 0x1ull);
582 msleep(1000);
583 __genwqe_readq(cd, IO_HSU_FIR_CLR);
584 __genwqe_readq(cd, IO_APP_FIR_CLR);
585 __genwqe_readq(cd, IO_SLU_FIR_CLR);
586
587 /*
588 * Read-modify-write to preserve the stealth bits
589 *
590 * For SL >= 039, Stealth WE bit allows removing
591 * the read-modify-wrote.
592 * r-m-w may require a mask 0x3C to avoid hitting hard
593 * reset again for error reset (should be 0, chicken).
594 */
595 softrst = __genwqe_readq(cd, IO_SLC_CFGREG_SOFTRESET) & 0x3cull;
596 __genwqe_writeq(cd, IO_SLC_CFGREG_SOFTRESET, softrst | 0x2ull);
597
598 /* give ERRORRESET some time to finish */
599 msleep(50);
600
601 if (genwqe_need_err_masking(cd)) {
602 dev_info(&pci_dev->dev,
603 "[%s] masking errors for old bitstreams\n", __func__);
604 __genwqe_writeq(cd, IO_SLC_MISC_DEBUG, 0x0aull);
605 }
606 return 0;
607}
608
609int genwqe_read_softreset(struct genwqe_dev *cd)
610{
611 u64 bitstream;
612
613 if (!genwqe_is_privileged(cd))
614 return -ENODEV;
615
616 bitstream = __genwqe_readq(cd, IO_SLU_BITSTREAM) & 0x1;
617 cd->softreset = (bitstream == 0) ? 0x8ull : 0xcull;
618 return 0;
619}
620
621/**
622 * genwqe_set_interrupt_capability() - Configure MSI capability structure
623 * @cd: pointer to the device
624 * Return: 0 if no error
625 */
626int genwqe_set_interrupt_capability(struct genwqe_dev *cd, int count)
627{
628 int rc;
629 struct pci_dev *pci_dev = cd->pci_dev;
630
631 rc = pci_enable_msi_block(pci_dev, count);
632 if (rc == 0)
633 cd->flags |= GENWQE_FLAG_MSI_ENABLED;
634 return rc;
635}
636
637/**
638 * genwqe_reset_interrupt_capability() - Undo genwqe_set_interrupt_capability()
639 * @cd: pointer to the device
640 */
641void genwqe_reset_interrupt_capability(struct genwqe_dev *cd)
642{
643 struct pci_dev *pci_dev = cd->pci_dev;
644
645 if (cd->flags & GENWQE_FLAG_MSI_ENABLED) {
646 pci_disable_msi(pci_dev);
647 cd->flags &= ~GENWQE_FLAG_MSI_ENABLED;
648 }
649}
650
651/**
652 * set_reg_idx() - Fill array with data. Ignore illegal offsets.
653 * @cd: card device
654 * @r: debug register array
655 * @i: index to desired entry
656 * @m: maximum possible entries
657 * @addr: addr which is read
658 * @index: index in debug array
659 * @val: read value
660 */
661static int set_reg_idx(struct genwqe_dev *cd, struct genwqe_reg *r,
662 unsigned int *i, unsigned int m, u32 addr, u32 idx,
663 u64 val)
664{
665 if (WARN_ON_ONCE(*i >= m))
666 return -EFAULT;
667
668 r[*i].addr = addr;
669 r[*i].idx = idx;
670 r[*i].val = val;
671 ++*i;
672 return 0;
673}
674
675static int set_reg(struct genwqe_dev *cd, struct genwqe_reg *r,
676 unsigned int *i, unsigned int m, u32 addr, u64 val)
677{
678 return set_reg_idx(cd, r, i, m, addr, 0, val);
679}
680
681int genwqe_read_ffdc_regs(struct genwqe_dev *cd, struct genwqe_reg *regs,
682 unsigned int max_regs, int all)
683{
684 unsigned int i, j, idx = 0;
685 u32 ufir_addr, ufec_addr, sfir_addr, sfec_addr;
686 u64 gfir, sluid, appid, ufir, ufec, sfir, sfec;
687
688 /* Global FIR */
689 gfir = __genwqe_readq(cd, IO_SLC_CFGREG_GFIR);
690 set_reg(cd, regs, &idx, max_regs, IO_SLC_CFGREG_GFIR, gfir);
691
692 /* UnitCfg for SLU */
693 sluid = __genwqe_readq(cd, IO_SLU_UNITCFG); /* 0x00000000 */
694 set_reg(cd, regs, &idx, max_regs, IO_SLU_UNITCFG, sluid);
695
696 /* UnitCfg for APP */
697 appid = __genwqe_readq(cd, IO_APP_UNITCFG); /* 0x02000000 */
698 set_reg(cd, regs, &idx, max_regs, IO_APP_UNITCFG, appid);
699
700 /* Check all chip Units */
701 for (i = 0; i < GENWQE_MAX_UNITS; i++) {
702
703 /* Unit FIR */
704 ufir_addr = (i << 24) | 0x008;
705 ufir = __genwqe_readq(cd, ufir_addr);
706 set_reg(cd, regs, &idx, max_regs, ufir_addr, ufir);
707
708 /* Unit FEC */
709 ufec_addr = (i << 24) | 0x018;
710 ufec = __genwqe_readq(cd, ufec_addr);
711 set_reg(cd, regs, &idx, max_regs, ufec_addr, ufec);
712
713 for (j = 0; j < 64; j++) {
714 /* wherever there is a primary 1, read the 2ndary */
715 if (!all && (!(ufir & (1ull << j))))
716 continue;
717
718 sfir_addr = (i << 24) | (0x100 + 8 * j);
719 sfir = __genwqe_readq(cd, sfir_addr);
720 set_reg(cd, regs, &idx, max_regs, sfir_addr, sfir);
721
722 sfec_addr = (i << 24) | (0x300 + 8 * j);
723 sfec = __genwqe_readq(cd, sfec_addr);
724 set_reg(cd, regs, &idx, max_regs, sfec_addr, sfec);
725 }
726 }
727
728 /* fill with invalid data until end */
729 for (i = idx; i < max_regs; i++) {
730 regs[i].addr = 0xffffffff;
731 regs[i].val = 0xffffffffffffffffull;
732 }
733 return idx;
734}
735
736/**
737 * genwqe_ffdc_buff_size() - Calculates the number of dump registers
738 */
739int genwqe_ffdc_buff_size(struct genwqe_dev *cd, int uid)
740{
741 int entries = 0, ring, traps, traces, trace_entries;
742 u32 eevptr_addr, l_addr, d_len, d_type;
743 u64 eevptr, val, addr;
744
745 eevptr_addr = GENWQE_UID_OFFS(uid) | IO_EXTENDED_ERROR_POINTER;
746 eevptr = __genwqe_readq(cd, eevptr_addr);
747
748 if ((eevptr != 0x0) && (eevptr != -1ull)) {
749 l_addr = GENWQE_UID_OFFS(uid) | eevptr;
750
751 while (1) {
752 val = __genwqe_readq(cd, l_addr);
753
754 if ((val == 0x0) || (val == -1ull))
755 break;
756
757 /* 38:24 */
758 d_len = (val & 0x0000007fff000000ull) >> 24;
759
760 /* 39 */
761 d_type = (val & 0x0000008000000000ull) >> 36;
762
763 if (d_type) { /* repeat */
764 entries += d_len;
765 } else { /* size in bytes! */
766 entries += d_len >> 3;
767 }
768
769 l_addr += 8;
770 }
771 }
772
773 for (ring = 0; ring < 8; ring++) {
774 addr = GENWQE_UID_OFFS(uid) | IO_EXTENDED_DIAG_MAP(ring);
775 val = __genwqe_readq(cd, addr);
776
777 if ((val == 0x0ull) || (val == -1ull))
778 continue;
779
780 traps = (val >> 24) & 0xff;
781 traces = (val >> 16) & 0xff;
782 trace_entries = val & 0xffff;
783
784 entries += traps + (traces * trace_entries);
785 }
786 return entries;
787}
788
789/**
790 * genwqe_ffdc_buff_read() - Implements LogoutExtendedErrorRegisters procedure
791 */
792int genwqe_ffdc_buff_read(struct genwqe_dev *cd, int uid,
793 struct genwqe_reg *regs, unsigned int max_regs)
794{
795 int i, traps, traces, trace, trace_entries, trace_entry, ring;
796 unsigned int idx = 0;
797 u32 eevptr_addr, l_addr, d_addr, d_len, d_type;
798 u64 eevptr, e, val, addr;
799
800 eevptr_addr = GENWQE_UID_OFFS(uid) | IO_EXTENDED_ERROR_POINTER;
801 eevptr = __genwqe_readq(cd, eevptr_addr);
802
803 if ((eevptr != 0x0) && (eevptr != 0xffffffffffffffffull)) {
804 l_addr = GENWQE_UID_OFFS(uid) | eevptr;
805 while (1) {
806 e = __genwqe_readq(cd, l_addr);
807 if ((e == 0x0) || (e == 0xffffffffffffffffull))
808 break;
809
810 d_addr = (e & 0x0000000000ffffffull); /* 23:0 */
811 d_len = (e & 0x0000007fff000000ull) >> 24; /* 38:24 */
812 d_type = (e & 0x0000008000000000ull) >> 36; /* 39 */
813 d_addr |= GENWQE_UID_OFFS(uid);
814
815 if (d_type) {
816 for (i = 0; i < (int)d_len; i++) {
817 val = __genwqe_readq(cd, d_addr);
818 set_reg_idx(cd, regs, &idx, max_regs,
819 d_addr, i, val);
820 }
821 } else {
822 d_len >>= 3; /* Size in bytes! */
823 for (i = 0; i < (int)d_len; i++, d_addr += 8) {
824 val = __genwqe_readq(cd, d_addr);
825 set_reg_idx(cd, regs, &idx, max_regs,
826 d_addr, 0, val);
827 }
828 }
829 l_addr += 8;
830 }
831 }
832
833 /*
834 * To save time, there are only 6 traces poplulated on Uid=2,
835 * Ring=1. each with iters=512.
836 */
837 for (ring = 0; ring < 8; ring++) { /* 0 is fls, 1 is fds,
838 2...7 are ASI rings */
839 addr = GENWQE_UID_OFFS(uid) | IO_EXTENDED_DIAG_MAP(ring);
840 val = __genwqe_readq(cd, addr);
841
842 if ((val == 0x0ull) || (val == -1ull))
843 continue;
844
845 traps = (val >> 24) & 0xff; /* Number of Traps */
846 traces = (val >> 16) & 0xff; /* Number of Traces */
847 trace_entries = val & 0xffff; /* Entries per trace */
848
849 /* Note: This is a combined loop that dumps both the traps */
850 /* (for the trace == 0 case) as well as the traces 1 to */
851 /* 'traces'. */
852 for (trace = 0; trace <= traces; trace++) {
853 u32 diag_sel =
854 GENWQE_EXTENDED_DIAG_SELECTOR(ring, trace);
855
856 addr = (GENWQE_UID_OFFS(uid) |
857 IO_EXTENDED_DIAG_SELECTOR);
858 __genwqe_writeq(cd, addr, diag_sel);
859
860 for (trace_entry = 0;
861 trace_entry < (trace ? trace_entries : traps);
862 trace_entry++) {
863 addr = (GENWQE_UID_OFFS(uid) |
864 IO_EXTENDED_DIAG_READ_MBX);
865 val = __genwqe_readq(cd, addr);
866 set_reg_idx(cd, regs, &idx, max_regs, addr,
867 (diag_sel<<16) | trace_entry, val);
868 }
869 }
870 }
871 return 0;
872}
873
874/**
875 * genwqe_write_vreg() - Write register in virtual window
876 *
877 * Note, these registers are only accessible to the PF through the
878 * VF-window. It is not intended for the VF to access.
879 */
880int genwqe_write_vreg(struct genwqe_dev *cd, u32 reg, u64 val, int func)
881{
882 __genwqe_writeq(cd, IO_PF_SLC_VIRTUAL_WINDOW, func & 0xf);
883 __genwqe_writeq(cd, reg, val);
884 return 0;
885}
886
887/**
888 * genwqe_read_vreg() - Read register in virtual window
889 *
890 * Note, these registers are only accessible to the PF through the
891 * VF-window. It is not intended for the VF to access.
892 */
893u64 genwqe_read_vreg(struct genwqe_dev *cd, u32 reg, int func)
894{
895 __genwqe_writeq(cd, IO_PF_SLC_VIRTUAL_WINDOW, func & 0xf);
896 return __genwqe_readq(cd, reg);
897}
898
899/**
900 * genwqe_base_clock_frequency() - Deteremine base clock frequency of the card
901 *
902 * Note: From a design perspective it turned out to be a bad idea to
903 * use codes here to specifiy the frequency/speed values. An old
904 * driver cannot understand new codes and is therefore always a
905 * problem. Better is to measure out the value or put the
906 * speed/frequency directly into a register which is always a valid
907 * value for old as well as for new software.
908 *
909 * Return: Card clock in MHz
910 */
911int genwqe_base_clock_frequency(struct genwqe_dev *cd)
912{
913 u16 speed; /* MHz MHz MHz MHz */
914 static const int speed_grade[] = { 250, 200, 166, 175 };
915
916 speed = (u16)((cd->slu_unitcfg >> 28) & 0x0full);
917 if (speed >= ARRAY_SIZE(speed_grade))
918 return 0; /* illegal value */
919
920 return speed_grade[speed];
921}
922
923/**
924 * genwqe_stop_traps() - Stop traps
925 *
926 * Before reading out the analysis data, we need to stop the traps.
927 */
928void genwqe_stop_traps(struct genwqe_dev *cd)
929{
930 __genwqe_writeq(cd, IO_SLC_MISC_DEBUG_SET, 0xcull);
931}
932
933/**
934 * genwqe_start_traps() - Start traps
935 *
936 * After having read the data, we can/must enable the traps again.
937 */
938void genwqe_start_traps(struct genwqe_dev *cd)
939{
940 __genwqe_writeq(cd, IO_SLC_MISC_DEBUG_CLR, 0xcull);
941
942 if (genwqe_need_err_masking(cd))
943 __genwqe_writeq(cd, IO_SLC_MISC_DEBUG, 0x0aull);
944}