forked from torvalds/linux
-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathsba_iommu.c
2109 lines (1770 loc) · 58.1 KB
/
sba_iommu.c
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/*
** System Bus Adapter (SBA) I/O MMU manager
**
** (c) Copyright 2000-2004 Grant Grundler <grundler @ parisc-linux x org>
** (c) Copyright 2004 Naresh Kumar Inna <knaresh at india x hp x com>
** (c) Copyright 2000-2004 Hewlett-Packard Company
**
** Portions (c) 1999 Dave S. Miller (from sparc64 I/O MMU code)
**
** This program is free software; you can redistribute it and/or modify
** it under the terms of the GNU General Public License as published by
** the Free Software Foundation; either version 2 of the License, or
** (at your option) any later version.
**
**
** This module initializes the IOC (I/O Controller) found on B1000/C3000/
** J5000/J7000/N-class/L-class machines and their successors.
**
** FIXME: add DMA hint support programming in both sba and lba modules.
*/
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/spinlock.h>
#include <linux/slab.h>
#include <linux/init.h>
#include <linux/mm.h>
#include <linux/string.h>
#include <linux/pci.h>
#include <linux/scatterlist.h>
#include <linux/iommu-helper.h>
#include <asm/byteorder.h>
#include <asm/io.h>
#include <asm/dma.h> /* for DMA_CHUNK_SIZE */
#include <asm/hardware.h> /* for register_parisc_driver() stuff */
#include <linux/proc_fs.h>
#include <linux/seq_file.h>
#include <linux/module.h>
#include <asm/ropes.h>
#include <asm/mckinley.h> /* for proc_mckinley_root */
#include <asm/runway.h> /* for proc_runway_root */
#include <asm/page.h> /* for PAGE0 */
#include <asm/pdc.h> /* for PDC_MODEL_* */
#include <asm/pdcpat.h> /* for is_pdc_pat() */
#include <asm/parisc-device.h>
#define MODULE_NAME "SBA"
/*
** The number of debug flags is a clue - this code is fragile.
** Don't even think about messing with it unless you have
** plenty of 710's to sacrifice to the computer gods. :^)
*/
#undef DEBUG_SBA_INIT
#undef DEBUG_SBA_RUN
#undef DEBUG_SBA_RUN_SG
#undef DEBUG_SBA_RESOURCE
#undef ASSERT_PDIR_SANITY
#undef DEBUG_LARGE_SG_ENTRIES
#undef DEBUG_DMB_TRAP
#ifdef DEBUG_SBA_INIT
#define DBG_INIT(x...) printk(x)
#else
#define DBG_INIT(x...)
#endif
#ifdef DEBUG_SBA_RUN
#define DBG_RUN(x...) printk(x)
#else
#define DBG_RUN(x...)
#endif
#ifdef DEBUG_SBA_RUN_SG
#define DBG_RUN_SG(x...) printk(x)
#else
#define DBG_RUN_SG(x...)
#endif
#ifdef DEBUG_SBA_RESOURCE
#define DBG_RES(x...) printk(x)
#else
#define DBG_RES(x...)
#endif
#define SBA_INLINE __inline__
#define DEFAULT_DMA_HINT_REG 0
struct sba_device *sba_list;
EXPORT_SYMBOL_GPL(sba_list);
static unsigned long ioc_needs_fdc = 0;
/* global count of IOMMUs in the system */
static unsigned int global_ioc_cnt = 0;
/* PA8700 (Piranha 2.2) bug workaround */
static unsigned long piranha_bad_128k = 0;
/* Looks nice and keeps the compiler happy */
#define SBA_DEV(d) ((struct sba_device *) (d))
#ifdef CONFIG_AGP_PARISC
#define SBA_AGP_SUPPORT
#endif /*CONFIG_AGP_PARISC*/
#ifdef SBA_AGP_SUPPORT
static int sba_reserve_agpgart = 1;
module_param(sba_reserve_agpgart, int, 0444);
MODULE_PARM_DESC(sba_reserve_agpgart, "Reserve half of IO pdir as AGPGART");
#endif
/************************************
** SBA register read and write support
**
** BE WARNED: register writes are posted.
** (ie follow writes which must reach HW with a read)
**
** Superdome (in particular, REO) allows only 64-bit CSR accesses.
*/
#define READ_REG32(addr) readl(addr)
#define READ_REG64(addr) readq(addr)
#define WRITE_REG32(val, addr) writel((val), (addr))
#define WRITE_REG64(val, addr) writeq((val), (addr))
#ifdef CONFIG_64BIT
#define READ_REG(addr) READ_REG64(addr)
#define WRITE_REG(value, addr) WRITE_REG64(value, addr)
#else
#define READ_REG(addr) READ_REG32(addr)
#define WRITE_REG(value, addr) WRITE_REG32(value, addr)
#endif
#ifdef DEBUG_SBA_INIT
/* NOTE: When CONFIG_64BIT isn't defined, READ_REG64() is two 32-bit reads */
/**
* sba_dump_ranges - debugging only - print ranges assigned to this IOA
* @hpa: base address of the sba
*
* Print the MMIO and IO Port address ranges forwarded by an Astro/Ike/RIO
* IO Adapter (aka Bus Converter).
*/
static void
sba_dump_ranges(void __iomem *hpa)
{
DBG_INIT("SBA at 0x%p\n", hpa);
DBG_INIT("IOS_DIST_BASE : %Lx\n", READ_REG64(hpa+IOS_DIST_BASE));
DBG_INIT("IOS_DIST_MASK : %Lx\n", READ_REG64(hpa+IOS_DIST_MASK));
DBG_INIT("IOS_DIST_ROUTE : %Lx\n", READ_REG64(hpa+IOS_DIST_ROUTE));
DBG_INIT("\n");
DBG_INIT("IOS_DIRECT_BASE : %Lx\n", READ_REG64(hpa+IOS_DIRECT_BASE));
DBG_INIT("IOS_DIRECT_MASK : %Lx\n", READ_REG64(hpa+IOS_DIRECT_MASK));
DBG_INIT("IOS_DIRECT_ROUTE: %Lx\n", READ_REG64(hpa+IOS_DIRECT_ROUTE));
}
/**
* sba_dump_tlb - debugging only - print IOMMU operating parameters
* @hpa: base address of the IOMMU
*
* Print the size/location of the IO MMU PDIR.
*/
static void sba_dump_tlb(void __iomem *hpa)
{
DBG_INIT("IO TLB at 0x%p\n", hpa);
DBG_INIT("IOC_IBASE : 0x%Lx\n", READ_REG64(hpa+IOC_IBASE));
DBG_INIT("IOC_IMASK : 0x%Lx\n", READ_REG64(hpa+IOC_IMASK));
DBG_INIT("IOC_TCNFG : 0x%Lx\n", READ_REG64(hpa+IOC_TCNFG));
DBG_INIT("IOC_PDIR_BASE: 0x%Lx\n", READ_REG64(hpa+IOC_PDIR_BASE));
DBG_INIT("\n");
}
#else
#define sba_dump_ranges(x)
#define sba_dump_tlb(x)
#endif /* DEBUG_SBA_INIT */
#ifdef ASSERT_PDIR_SANITY
/**
* sba_dump_pdir_entry - debugging only - print one IOMMU PDIR entry
* @ioc: IO MMU structure which owns the pdir we are interested in.
* @msg: text to print ont the output line.
* @pide: pdir index.
*
* Print one entry of the IO MMU PDIR in human readable form.
*/
static void
sba_dump_pdir_entry(struct ioc *ioc, char *msg, uint pide)
{
/* start printing from lowest pde in rval */
u64 *ptr = &(ioc->pdir_base[pide & (~0U * BITS_PER_LONG)]);
unsigned long *rptr = (unsigned long *) &(ioc->res_map[(pide >>3) & ~(sizeof(unsigned long) - 1)]);
uint rcnt;
printk(KERN_DEBUG "SBA: %s rp %p bit %d rval 0x%lx\n",
msg,
rptr, pide & (BITS_PER_LONG - 1), *rptr);
rcnt = 0;
while (rcnt < BITS_PER_LONG) {
printk(KERN_DEBUG "%s %2d %p %016Lx\n",
(rcnt == (pide & (BITS_PER_LONG - 1)))
? " -->" : " ",
rcnt, ptr, *ptr );
rcnt++;
ptr++;
}
printk(KERN_DEBUG "%s", msg);
}
/**
* sba_check_pdir - debugging only - consistency checker
* @ioc: IO MMU structure which owns the pdir we are interested in.
* @msg: text to print ont the output line.
*
* Verify the resource map and pdir state is consistent
*/
static int
sba_check_pdir(struct ioc *ioc, char *msg)
{
u32 *rptr_end = (u32 *) &(ioc->res_map[ioc->res_size]);
u32 *rptr = (u32 *) ioc->res_map; /* resource map ptr */
u64 *pptr = ioc->pdir_base; /* pdir ptr */
uint pide = 0;
while (rptr < rptr_end) {
u32 rval = *rptr;
int rcnt = 32; /* number of bits we might check */
while (rcnt) {
/* Get last byte and highest bit from that */
u32 pde = ((u32) (((char *)pptr)[7])) << 24;
if ((rval ^ pde) & 0x80000000)
{
/*
** BUMMER! -- res_map != pdir --
** Dump rval and matching pdir entries
*/
sba_dump_pdir_entry(ioc, msg, pide);
return(1);
}
rcnt--;
rval <<= 1; /* try the next bit */
pptr++;
pide++;
}
rptr++; /* look at next word of res_map */
}
/* It'd be nice if we always got here :^) */
return 0;
}
/**
* sba_dump_sg - debugging only - print Scatter-Gather list
* @ioc: IO MMU structure which owns the pdir we are interested in.
* @startsg: head of the SG list
* @nents: number of entries in SG list
*
* print the SG list so we can verify it's correct by hand.
*/
static void
sba_dump_sg( struct ioc *ioc, struct scatterlist *startsg, int nents)
{
while (nents-- > 0) {
printk(KERN_DEBUG " %d : %08lx/%05x %p/%05x\n",
nents,
(unsigned long) sg_dma_address(startsg),
sg_dma_len(startsg),
sg_virt(startsg), startsg->length);
startsg++;
}
}
#endif /* ASSERT_PDIR_SANITY */
/**************************************************************
*
* I/O Pdir Resource Management
*
* Bits set in the resource map are in use.
* Each bit can represent a number of pages.
* LSbs represent lower addresses (IOVA's).
*
***************************************************************/
#define PAGES_PER_RANGE 1 /* could increase this to 4 or 8 if needed */
/* Convert from IOVP to IOVA and vice versa. */
#ifdef ZX1_SUPPORT
/* Pluto (aka ZX1) boxes need to set or clear the ibase bits appropriately */
#define SBA_IOVA(ioc,iovp,offset,hint_reg) ((ioc->ibase) | (iovp) | (offset))
#define SBA_IOVP(ioc,iova) ((iova) & (ioc)->iovp_mask)
#else
/* only support Astro and ancestors. Saves a few cycles in key places */
#define SBA_IOVA(ioc,iovp,offset,hint_reg) ((iovp) | (offset))
#define SBA_IOVP(ioc,iova) (iova)
#endif
#define PDIR_INDEX(iovp) ((iovp)>>IOVP_SHIFT)
#define RESMAP_MASK(n) (~0UL << (BITS_PER_LONG - (n)))
#define RESMAP_IDX_MASK (sizeof(unsigned long) - 1)
static unsigned long ptr_to_pide(struct ioc *ioc, unsigned long *res_ptr,
unsigned int bitshiftcnt)
{
return (((unsigned long)res_ptr - (unsigned long)ioc->res_map) << 3)
+ bitshiftcnt;
}
/**
* sba_search_bitmap - find free space in IO PDIR resource bitmap
* @ioc: IO MMU structure which owns the pdir we are interested in.
* @bits_wanted: number of entries we need.
*
* Find consecutive free bits in resource bitmap.
* Each bit represents one entry in the IO Pdir.
* Cool perf optimization: search for log2(size) bits at a time.
*/
static SBA_INLINE unsigned long
sba_search_bitmap(struct ioc *ioc, struct device *dev,
unsigned long bits_wanted)
{
unsigned long *res_ptr = ioc->res_hint;
unsigned long *res_end = (unsigned long *) &(ioc->res_map[ioc->res_size]);
unsigned long pide = ~0UL, tpide;
unsigned long boundary_size;
unsigned long shift;
int ret;
boundary_size = ALIGN((unsigned long long)dma_get_seg_boundary(dev) + 1,
1ULL << IOVP_SHIFT) >> IOVP_SHIFT;
#if defined(ZX1_SUPPORT)
BUG_ON(ioc->ibase & ~IOVP_MASK);
shift = ioc->ibase >> IOVP_SHIFT;
#else
shift = 0;
#endif
if (bits_wanted > (BITS_PER_LONG/2)) {
/* Search word at a time - no mask needed */
for(; res_ptr < res_end; ++res_ptr) {
tpide = ptr_to_pide(ioc, res_ptr, 0);
ret = iommu_is_span_boundary(tpide, bits_wanted,
shift,
boundary_size);
if ((*res_ptr == 0) && !ret) {
*res_ptr = RESMAP_MASK(bits_wanted);
pide = tpide;
break;
}
}
/* point to the next word on next pass */
res_ptr++;
ioc->res_bitshift = 0;
} else {
/*
** Search the resource bit map on well-aligned values.
** "o" is the alignment.
** We need the alignment to invalidate I/O TLB using
** SBA HW features in the unmap path.
*/
unsigned long o = 1 << get_order(bits_wanted << PAGE_SHIFT);
uint bitshiftcnt = ALIGN(ioc->res_bitshift, o);
unsigned long mask;
if (bitshiftcnt >= BITS_PER_LONG) {
bitshiftcnt = 0;
res_ptr++;
}
mask = RESMAP_MASK(bits_wanted) >> bitshiftcnt;
DBG_RES("%s() o %ld %p", __func__, o, res_ptr);
while(res_ptr < res_end)
{
DBG_RES(" %p %lx %lx\n", res_ptr, mask, *res_ptr);
WARN_ON(mask == 0);
tpide = ptr_to_pide(ioc, res_ptr, bitshiftcnt);
ret = iommu_is_span_boundary(tpide, bits_wanted,
shift,
boundary_size);
if ((((*res_ptr) & mask) == 0) && !ret) {
*res_ptr |= mask; /* mark resources busy! */
pide = tpide;
break;
}
mask >>= o;
bitshiftcnt += o;
if (mask == 0) {
mask = RESMAP_MASK(bits_wanted);
bitshiftcnt=0;
res_ptr++;
}
}
/* look in the same word on the next pass */
ioc->res_bitshift = bitshiftcnt + bits_wanted;
}
/* wrapped ? */
if (res_end <= res_ptr) {
ioc->res_hint = (unsigned long *) ioc->res_map;
ioc->res_bitshift = 0;
} else {
ioc->res_hint = res_ptr;
}
return (pide);
}
/**
* sba_alloc_range - find free bits and mark them in IO PDIR resource bitmap
* @ioc: IO MMU structure which owns the pdir we are interested in.
* @size: number of bytes to create a mapping for
*
* Given a size, find consecutive unmarked and then mark those bits in the
* resource bit map.
*/
static int
sba_alloc_range(struct ioc *ioc, struct device *dev, size_t size)
{
unsigned int pages_needed = size >> IOVP_SHIFT;
#ifdef SBA_COLLECT_STATS
unsigned long cr_start = mfctl(16);
#endif
unsigned long pide;
pide = sba_search_bitmap(ioc, dev, pages_needed);
if (pide >= (ioc->res_size << 3)) {
pide = sba_search_bitmap(ioc, dev, pages_needed);
if (pide >= (ioc->res_size << 3))
panic("%s: I/O MMU @ %p is out of mapping resources\n",
__FILE__, ioc->ioc_hpa);
}
#ifdef ASSERT_PDIR_SANITY
/* verify the first enable bit is clear */
if(0x00 != ((u8 *) ioc->pdir_base)[pide*sizeof(u64) + 7]) {
sba_dump_pdir_entry(ioc, "sba_search_bitmap() botched it?", pide);
}
#endif
DBG_RES("%s(%x) %d -> %lx hint %x/%x\n",
__func__, size, pages_needed, pide,
(uint) ((unsigned long) ioc->res_hint - (unsigned long) ioc->res_map),
ioc->res_bitshift );
#ifdef SBA_COLLECT_STATS
{
unsigned long cr_end = mfctl(16);
unsigned long tmp = cr_end - cr_start;
/* check for roll over */
cr_start = (cr_end < cr_start) ? -(tmp) : (tmp);
}
ioc->avg_search[ioc->avg_idx++] = cr_start;
ioc->avg_idx &= SBA_SEARCH_SAMPLE - 1;
ioc->used_pages += pages_needed;
#endif
return (pide);
}
/**
* sba_free_range - unmark bits in IO PDIR resource bitmap
* @ioc: IO MMU structure which owns the pdir we are interested in.
* @iova: IO virtual address which was previously allocated.
* @size: number of bytes to create a mapping for
*
* clear bits in the ioc's resource map
*/
static SBA_INLINE void
sba_free_range(struct ioc *ioc, dma_addr_t iova, size_t size)
{
unsigned long iovp = SBA_IOVP(ioc, iova);
unsigned int pide = PDIR_INDEX(iovp);
unsigned int ridx = pide >> 3; /* convert bit to byte address */
unsigned long *res_ptr = (unsigned long *) &((ioc)->res_map[ridx & ~RESMAP_IDX_MASK]);
int bits_not_wanted = size >> IOVP_SHIFT;
/* 3-bits "bit" address plus 2 (or 3) bits for "byte" == bit in word */
unsigned long m = RESMAP_MASK(bits_not_wanted) >> (pide & (BITS_PER_LONG - 1));
DBG_RES("%s( ,%x,%x) %x/%lx %x %p %lx\n",
__func__, (uint) iova, size,
bits_not_wanted, m, pide, res_ptr, *res_ptr);
#ifdef SBA_COLLECT_STATS
ioc->used_pages -= bits_not_wanted;
#endif
*res_ptr &= ~m;
}
/**************************************************************
*
* "Dynamic DMA Mapping" support (aka "Coherent I/O")
*
***************************************************************/
#ifdef SBA_HINT_SUPPORT
#define SBA_DMA_HINT(ioc, val) ((val) << (ioc)->hint_shift_pdir)
#endif
typedef unsigned long space_t;
#define KERNEL_SPACE 0
/**
* sba_io_pdir_entry - fill in one IO PDIR entry
* @pdir_ptr: pointer to IO PDIR entry
* @sid: process Space ID - currently only support KERNEL_SPACE
* @vba: Virtual CPU address of buffer to map
* @hint: DMA hint set to use for this mapping
*
* SBA Mapping Routine
*
* Given a virtual address (vba, arg2) and space id, (sid, arg1)
* sba_io_pdir_entry() loads the I/O PDIR entry pointed to by
* pdir_ptr (arg0).
* Using the bass-ackwards HP bit numbering, Each IO Pdir entry
* for Astro/Ike looks like:
*
*
* 0 19 51 55 63
* +-+---------------------+----------------------------------+----+--------+
* |V| U | PPN[43:12] | U | VI |
* +-+---------------------+----------------------------------+----+--------+
*
* Pluto is basically identical, supports fewer physical address bits:
*
* 0 23 51 55 63
* +-+------------------------+-------------------------------+----+--------+
* |V| U | PPN[39:12] | U | VI |
* +-+------------------------+-------------------------------+----+--------+
*
* V == Valid Bit (Most Significant Bit is bit 0)
* U == Unused
* PPN == Physical Page Number
* VI == Virtual Index (aka Coherent Index)
*
* LPA instruction output is put into PPN field.
* LCI (Load Coherence Index) instruction provides the "VI" bits.
*
* We pre-swap the bytes since PCX-W is Big Endian and the
* IOMMU uses little endian for the pdir.
*/
static void SBA_INLINE
sba_io_pdir_entry(u64 *pdir_ptr, space_t sid, unsigned long vba,
unsigned long hint)
{
u64 pa; /* physical address */
register unsigned ci; /* coherent index */
pa = virt_to_phys(vba);
pa &= IOVP_MASK;
mtsp(sid,1);
asm("lci 0(%%sr1, %1), %0" : "=r" (ci) : "r" (vba));
pa |= (ci >> PAGE_SHIFT) & 0xff; /* move CI (8 bits) into lowest byte */
pa |= SBA_PDIR_VALID_BIT; /* set "valid" bit */
*pdir_ptr = cpu_to_le64(pa); /* swap and store into I/O Pdir */
/*
* If the PDC_MODEL capabilities has Non-coherent IO-PDIR bit set
* (bit #61, big endian), we have to flush and sync every time
* IO-PDIR is changed in Ike/Astro.
*/
if (ioc_needs_fdc)
asm volatile("fdc %%r0(%0)" : : "r" (pdir_ptr));
}
/**
* sba_mark_invalid - invalidate one or more IO PDIR entries
* @ioc: IO MMU structure which owns the pdir we are interested in.
* @iova: IO Virtual Address mapped earlier
* @byte_cnt: number of bytes this mapping covers.
*
* Marking the IO PDIR entry(ies) as Invalid and invalidate
* corresponding IO TLB entry. The Ike PCOM (Purge Command Register)
* is to purge stale entries in the IO TLB when unmapping entries.
*
* The PCOM register supports purging of multiple pages, with a minium
* of 1 page and a maximum of 2GB. Hardware requires the address be
* aligned to the size of the range being purged. The size of the range
* must be a power of 2. The "Cool perf optimization" in the
* allocation routine helps keep that true.
*/
static SBA_INLINE void
sba_mark_invalid(struct ioc *ioc, dma_addr_t iova, size_t byte_cnt)
{
u32 iovp = (u32) SBA_IOVP(ioc,iova);
u64 *pdir_ptr = &ioc->pdir_base[PDIR_INDEX(iovp)];
#ifdef ASSERT_PDIR_SANITY
/* Assert first pdir entry is set.
**
** Even though this is a big-endian machine, the entries
** in the iopdir are little endian. That's why we look at
** the byte at +7 instead of at +0.
*/
if (0x80 != (((u8 *) pdir_ptr)[7])) {
sba_dump_pdir_entry(ioc,"sba_mark_invalid()", PDIR_INDEX(iovp));
}
#endif
if (byte_cnt > IOVP_SIZE)
{
#if 0
unsigned long entries_per_cacheline = ioc_needs_fdc ?
L1_CACHE_ALIGN(((unsigned long) pdir_ptr))
- (unsigned long) pdir_ptr;
: 262144;
#endif
/* set "size" field for PCOM */
iovp |= get_order(byte_cnt) + PAGE_SHIFT;
do {
/* clear I/O Pdir entry "valid" bit first */
((u8 *) pdir_ptr)[7] = 0;
if (ioc_needs_fdc) {
asm volatile("fdc %%r0(%0)" : : "r" (pdir_ptr));
#if 0
entries_per_cacheline = L1_CACHE_SHIFT - 3;
#endif
}
pdir_ptr++;
byte_cnt -= IOVP_SIZE;
} while (byte_cnt > IOVP_SIZE);
} else
iovp |= IOVP_SHIFT; /* set "size" field for PCOM */
/*
** clear I/O PDIR entry "valid" bit.
** We have to R/M/W the cacheline regardless how much of the
** pdir entry that we clobber.
** The rest of the entry would be useful for debugging if we
** could dump core on HPMC.
*/
((u8 *) pdir_ptr)[7] = 0;
if (ioc_needs_fdc)
asm volatile("fdc %%r0(%0)" : : "r" (pdir_ptr));
WRITE_REG( SBA_IOVA(ioc, iovp, 0, 0), ioc->ioc_hpa+IOC_PCOM);
}
/**
* sba_dma_supported - PCI driver can query DMA support
* @dev: instance of PCI owned by the driver that's asking
* @mask: number of address bits this PCI device can handle
*
* See Documentation/DMA-API-HOWTO.txt
*/
static int sba_dma_supported( struct device *dev, u64 mask)
{
struct ioc *ioc;
if (dev == NULL) {
printk(KERN_ERR MODULE_NAME ": EISA/ISA/et al not supported\n");
BUG();
return(0);
}
/* Documentation/DMA-API-HOWTO.txt tells drivers to try 64-bit
* first, then fall back to 32-bit if that fails.
* We are just "encouraging" 32-bit DMA masks here since we can
* never allow IOMMU bypass unless we add special support for ZX1.
*/
if (mask > ~0U)
return 0;
ioc = GET_IOC(dev);
/*
* check if mask is >= than the current max IO Virt Address
* The max IO Virt address will *always* < 30 bits.
*/
return((int)(mask >= (ioc->ibase - 1 +
(ioc->pdir_size / sizeof(u64) * IOVP_SIZE) )));
}
/**
* sba_map_single - map one buffer and return IOVA for DMA
* @dev: instance of PCI owned by the driver that's asking.
* @addr: driver buffer to map.
* @size: number of bytes to map in driver buffer.
* @direction: R/W or both.
*
* See Documentation/DMA-API-HOWTO.txt
*/
static dma_addr_t
sba_map_single(struct device *dev, void *addr, size_t size,
enum dma_data_direction direction)
{
struct ioc *ioc;
unsigned long flags;
dma_addr_t iovp;
dma_addr_t offset;
u64 *pdir_start;
int pide;
ioc = GET_IOC(dev);
/* save offset bits */
offset = ((dma_addr_t) (long) addr) & ~IOVP_MASK;
/* round up to nearest IOVP_SIZE */
size = (size + offset + ~IOVP_MASK) & IOVP_MASK;
spin_lock_irqsave(&ioc->res_lock, flags);
#ifdef ASSERT_PDIR_SANITY
sba_check_pdir(ioc,"Check before sba_map_single()");
#endif
#ifdef SBA_COLLECT_STATS
ioc->msingle_calls++;
ioc->msingle_pages += size >> IOVP_SHIFT;
#endif
pide = sba_alloc_range(ioc, dev, size);
iovp = (dma_addr_t) pide << IOVP_SHIFT;
DBG_RUN("%s() 0x%p -> 0x%lx\n",
__func__, addr, (long) iovp | offset);
pdir_start = &(ioc->pdir_base[pide]);
while (size > 0) {
sba_io_pdir_entry(pdir_start, KERNEL_SPACE, (unsigned long) addr, 0);
DBG_RUN(" pdir 0x%p %02x%02x%02x%02x%02x%02x%02x%02x\n",
pdir_start,
(u8) (((u8 *) pdir_start)[7]),
(u8) (((u8 *) pdir_start)[6]),
(u8) (((u8 *) pdir_start)[5]),
(u8) (((u8 *) pdir_start)[4]),
(u8) (((u8 *) pdir_start)[3]),
(u8) (((u8 *) pdir_start)[2]),
(u8) (((u8 *) pdir_start)[1]),
(u8) (((u8 *) pdir_start)[0])
);
addr += IOVP_SIZE;
size -= IOVP_SIZE;
pdir_start++;
}
/* force FDC ops in io_pdir_entry() to be visible to IOMMU */
if (ioc_needs_fdc)
asm volatile("sync" : : );
#ifdef ASSERT_PDIR_SANITY
sba_check_pdir(ioc,"Check after sba_map_single()");
#endif
spin_unlock_irqrestore(&ioc->res_lock, flags);
/* form complete address */
return SBA_IOVA(ioc, iovp, offset, DEFAULT_DMA_HINT_REG);
}
static dma_addr_t
sba_map_page(struct device *dev, struct page *page, unsigned long offset,
size_t size, enum dma_data_direction direction,
struct dma_attrs *attrs)
{
return sba_map_single(dev, page_address(page) + offset, size,
direction);
}
/**
* sba_unmap_page - unmap one IOVA and free resources
* @dev: instance of PCI owned by the driver that's asking.
* @iova: IOVA of driver buffer previously mapped.
* @size: number of bytes mapped in driver buffer.
* @direction: R/W or both.
*
* See Documentation/DMA-API-HOWTO.txt
*/
static void
sba_unmap_page(struct device *dev, dma_addr_t iova, size_t size,
enum dma_data_direction direction, struct dma_attrs *attrs)
{
struct ioc *ioc;
#if DELAYED_RESOURCE_CNT > 0
struct sba_dma_pair *d;
#endif
unsigned long flags;
dma_addr_t offset;
DBG_RUN("%s() iovp 0x%lx/%x\n", __func__, (long) iova, size);
ioc = GET_IOC(dev);
offset = iova & ~IOVP_MASK;
iova ^= offset; /* clear offset bits */
size += offset;
size = ALIGN(size, IOVP_SIZE);
spin_lock_irqsave(&ioc->res_lock, flags);
#ifdef SBA_COLLECT_STATS
ioc->usingle_calls++;
ioc->usingle_pages += size >> IOVP_SHIFT;
#endif
sba_mark_invalid(ioc, iova, size);
#if DELAYED_RESOURCE_CNT > 0
/* Delaying when we re-use a IO Pdir entry reduces the number
* of MMIO reads needed to flush writes to the PCOM register.
*/
d = &(ioc->saved[ioc->saved_cnt]);
d->iova = iova;
d->size = size;
if (++(ioc->saved_cnt) >= DELAYED_RESOURCE_CNT) {
int cnt = ioc->saved_cnt;
while (cnt--) {
sba_free_range(ioc, d->iova, d->size);
d--;
}
ioc->saved_cnt = 0;
READ_REG(ioc->ioc_hpa+IOC_PCOM); /* flush purges */
}
#else /* DELAYED_RESOURCE_CNT == 0 */
sba_free_range(ioc, iova, size);
/* If fdc's were issued, force fdc's to be visible now */
if (ioc_needs_fdc)
asm volatile("sync" : : );
READ_REG(ioc->ioc_hpa+IOC_PCOM); /* flush purges */
#endif /* DELAYED_RESOURCE_CNT == 0 */
spin_unlock_irqrestore(&ioc->res_lock, flags);
/* XXX REVISIT for 2.5 Linux - need syncdma for zero-copy support.
** For Astro based systems this isn't a big deal WRT performance.
** As long as 2.4 kernels copyin/copyout data from/to userspace,
** we don't need the syncdma. The issue here is I/O MMU cachelines
** are *not* coherent in all cases. May be hwrev dependent.
** Need to investigate more.
asm volatile("syncdma");
*/
}
/**
* sba_alloc - allocate/map shared mem for DMA
* @hwdev: instance of PCI owned by the driver that's asking.
* @size: number of bytes mapped in driver buffer.
* @dma_handle: IOVA of new buffer.
*
* See Documentation/DMA-API-HOWTO.txt
*/
static void *sba_alloc(struct device *hwdev, size_t size, dma_addr_t *dma_handle,
gfp_t gfp, struct dma_attrs *attrs)
{
void *ret;
if (!hwdev) {
/* only support PCI */
*dma_handle = 0;
return NULL;
}
ret = (void *) __get_free_pages(gfp, get_order(size));
if (ret) {
memset(ret, 0, size);
*dma_handle = sba_map_single(hwdev, ret, size, 0);
}
return ret;
}
/**
* sba_free - free/unmap shared mem for DMA
* @hwdev: instance of PCI owned by the driver that's asking.
* @size: number of bytes mapped in driver buffer.
* @vaddr: virtual address IOVA of "consistent" buffer.
* @dma_handler: IO virtual address of "consistent" buffer.
*
* See Documentation/DMA-API-HOWTO.txt
*/
static void
sba_free(struct device *hwdev, size_t size, void *vaddr,
dma_addr_t dma_handle, struct dma_attrs *attrs)
{
sba_unmap_page(hwdev, dma_handle, size, 0, NULL);
free_pages((unsigned long) vaddr, get_order(size));
}
/*
** Since 0 is a valid pdir_base index value, can't use that
** to determine if a value is valid or not. Use a flag to indicate
** the SG list entry contains a valid pdir index.
*/
#define PIDE_FLAG 0x80000000UL
#ifdef SBA_COLLECT_STATS
#define IOMMU_MAP_STATS
#endif
#include "iommu-helpers.h"
#ifdef DEBUG_LARGE_SG_ENTRIES
int dump_run_sg = 0;
#endif
/**
* sba_map_sg - map Scatter/Gather list
* @dev: instance of PCI owned by the driver that's asking.
* @sglist: array of buffer/length pairs
* @nents: number of entries in list
* @direction: R/W or both.
*
* See Documentation/DMA-API-HOWTO.txt
*/
static int
sba_map_sg(struct device *dev, struct scatterlist *sglist, int nents,
enum dma_data_direction direction, struct dma_attrs *attrs)
{
struct ioc *ioc;
int coalesced, filled = 0;
unsigned long flags;
DBG_RUN_SG("%s() START %d entries\n", __func__, nents);
ioc = GET_IOC(dev);
/* Fast path single entry scatterlists. */
if (nents == 1) {
sg_dma_address(sglist) = sba_map_single(dev, sg_virt(sglist),
sglist->length, direction);
sg_dma_len(sglist) = sglist->length;
return 1;
}
spin_lock_irqsave(&ioc->res_lock, flags);
#ifdef ASSERT_PDIR_SANITY
if (sba_check_pdir(ioc,"Check before sba_map_sg()"))
{
sba_dump_sg(ioc, sglist, nents);
panic("Check before sba_map_sg()");
}
#endif
#ifdef SBA_COLLECT_STATS
ioc->msg_calls++;
#endif
/*
** First coalesce the chunks and allocate I/O pdir space
**
** If this is one DMA stream, we can properly map using the
** correct virtual address associated with each DMA page.
** w/o this association, we wouldn't have coherent DMA!
** Access to the virtual address is what forces a two pass algorithm.
*/
coalesced = iommu_coalesce_chunks(ioc, dev, sglist, nents, sba_alloc_range);
/*
** Program the I/O Pdir
**
** map the virtual addresses to the I/O Pdir
** o dma_address will contain the pdir index
** o dma_len will contain the number of bytes to map
** o address contains the virtual address.
*/
filled = iommu_fill_pdir(ioc, sglist, nents, 0, sba_io_pdir_entry);
/* force FDC ops in io_pdir_entry() to be visible to IOMMU */
if (ioc_needs_fdc)
asm volatile("sync" : : );