aboutsummaryrefslogtreecommitdiffstats
path: root/arch/mips/lib/csum_partial.S
blob: a6adffbb4e5f0a5ccc5ec268c18c718b69c5a8da (plain) (blame)
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
/*
 * This file is subject to the terms and conditions of the GNU General Public
 * License.  See the file "COPYING" in the main directory of this archive
 * for more details.
 *
 * Quick'n'dirty IP checksum ...
 *
 * Copyright (C) 1998, 1999 Ralf Baechle
 * Copyright (C) 1999 Silicon Graphics, Inc.
 * Copyright (C) 2007  Maciej W. Rozycki
 */
#include <linux/errno.h>
#include <asm/asm.h>
#include <asm/asm-offsets.h>
#include <asm/regdef.h>

#ifdef CONFIG_64BIT
/*
 * As we are sharing code base with the mips32 tree (which use the o32 ABI
 * register definitions). We need to redefine the register definitions from
 * the n64 ABI register naming to the o32 ABI register naming.
 */
#undef t0
#undef t1
#undef t2
#undef t3
#define t0	$8
#define t1	$9
#define t2	$10
#define t3	$11
#define t4	$12
#define t5	$13
#define t6	$14
#define t7	$15

#define USE_DOUBLE
#endif

#ifdef USE_DOUBLE

#define LOAD   ld
#define LOAD32 lwu
#define ADD    daddu
#define NBYTES 8

#else

#define LOAD   lw
#define LOAD32 lw
#define ADD    addu
#define NBYTES 4

#endif /* USE_DOUBLE */

#define UNIT(unit)  ((unit)*NBYTES)

#define ADDC(sum,reg)						\
	ADD	sum, reg;					\
	sltu	v1, sum, reg;					\
	ADD	sum, v1;					\

#define ADDC32(sum,reg)						\
	addu	sum, reg;					\
	sltu	v1, sum, reg;					\
	addu	sum, v1;					\

#define CSUM_BIGCHUNK1(src, offset, sum, _t0, _t1, _t2, _t3)	\
	LOAD	_t0, (offset + UNIT(0))(src);			\
	LOAD	_t1, (offset + UNIT(1))(src);			\
	LOAD	_t2, (offset + UNIT(2))(src);			\
	LOAD	_t3, (offset + UNIT(3))(src);			\
	ADDC(sum, _t0);						\
	ADDC(sum, _t1);						\
	ADDC(sum, _t2);						\
	ADDC(sum, _t3)

#ifdef USE_DOUBLE
#define CSUM_BIGCHUNK(src, offset, sum, _t0, _t1, _t2, _t3)	\
	CSUM_BIGCHUNK1(src, offset, sum, _t0, _t1, _t2, _t3)
#else
#define CSUM_BIGCHUNK(src, offset, sum, _t0, _t1, _t2, _t3)	\
	CSUM_BIGCHUNK1(src, offset, sum, _t0, _t1, _t2, _t3);	\
	CSUM_BIGCHUNK1(src, offset + 0x10, sum, _t0, _t1, _t2, _t3)
#endif

/*
 * a0: source address
 * a1: length of the area to checksum
 * a2: partial checksum
 */

#define src a0
#define sum v0

	.text
	.set	noreorder
	.align	5
LEAF(csum_partial)
	move	sum, zero
	move	t7, zero

	sltiu	t8, a1, 0x8
	bnez	t8, .Lsmall_csumcpy		/* < 8 bytes to copy */
	 move	t2, a1

	andi	t7, src, 0x1			/* odd buffer? */

.Lhword_align:
	beqz	t7, .Lword_align
	 andi	t8, src, 0x2

	lbu	t0, (src)
	LONG_SUBU	a1, a1, 0x1
#ifdef __MIPSEL__
	sll	t0, t0, 8
#endif
	ADDC(sum, t0)
	PTR_ADDU	src, src, 0x1
	andi	t8, src, 0x2

.Lword_align:
	beqz	t8, .Ldword_align
	 sltiu	t8, a1, 56

	lhu	t0, (src)
	LONG_SUBU	a1, a1, 0x2
	ADDC(sum, t0)
	sltiu	t8, a1, 56
	PTR_ADDU	src, src, 0x2

.Ldword_align:
	bnez	t8, .Ldo_end_words
	 move	t8, a1

	andi	t8, src, 0x4
	beqz	t8, .Lqword_align
	 andi	t8, src, 0x8

	LOAD32	t0, 0x00(src)
	LONG_SUBU	a1, a1, 0x4
	ADDC(sum, t0)
	PTR_ADDU	src, src, 0x4
	andi	t8, src, 0x8

.Lqword_align:
	beqz	t8, .Loword_align
	 andi	t8, src, 0x10

#ifdef USE_DOUBLE
	ld	t0, 0x00(src)
	LONG_SUBU	a1, a1, 0x8
	ADDC(sum, t0)
#else
	lw	t0, 0x00(src)
	lw	t1, 0x04(src)
	LONG_SUBU	a1, a1, 0x8
	ADDC(sum, t0)
	ADDC(sum, t1)
#endif
	PTR_ADDU	src, src, 0x8
	andi	t8, src, 0x10

.Loword_align:
	beqz	t8, .Lbegin_movement
	 LONG_SRL	t8, a1, 0x7

#ifdef USE_DOUBLE
	ld	t0, 0x00(src)
	ld	t1, 0x08(src)
	ADDC(sum, t0)
	ADDC(sum, t1)
#else
	CSUM_BIGCHUNK1(src, 0x00, sum, t0, t1, t3, t4)
#endif
	LONG_SUBU	a1, a1, 0x10
	PTR_ADDU	src, src, 0x10
	LONG_SRL	t8, a1, 0x7

.Lbegin_movement:
	beqz	t8, 1f
	 andi	t2, a1, 0x40

.Lmove_128bytes:
	CSUM_BIGCHUNK(src, 0x00, sum, t0, t1, t3, t4)
	CSUM_BIGCHUNK(src, 0x20, sum, t0, t1, t3, t4)
	CSUM_BIGCHUNK(src, 0x40, sum, t0, t1, t3, t4)
	CSUM_BIGCHUNK(src, 0x60, sum, t0, t1, t3, t4)
	LONG_SUBU	t8, t8, 0x01
	.set	reorder				/* DADDI_WAR */
	PTR_ADDU	src, src, 0x80
	bnez	t8, .Lmove_128bytes
	.set	noreorder

1:
	beqz	t2, 1f
	 andi	t2, a1, 0x20

.Lmove_64bytes:
	CSUM_BIGCHUNK(src, 0x00, sum, t0, t1, t3, t4)
	CSUM_BIGCHUNK(src, 0x20, sum, t0, t1, t3, t4)
	PTR_ADDU	src, src, 0x40

1:
	beqz	t2, .Ldo_end_words
	 andi	t8, a1, 0x1c

.Lmove_32bytes:
	CSUM_BIGCHUNK(src, 0x00, sum, t0, t1, t3, t4)
	andi	t8, a1, 0x1c
	PTR_ADDU	src, src, 0x20

.Ldo_end_words:
	beqz	t8, .Lsmall_csumcpy
	 andi	t2, a1, 0x3
	LONG_SRL	t8, t8, 0x2

.Lend_words:
	LOAD32	t0, (src)
	LONG_SUBU	t8, t8, 0x1
	ADDC(sum, t0)
	.set	reorder				/* DADDI_WAR */
	PTR_ADDU	src, src, 0x4
	bnez	t8, .Lend_words
	.set	noreorder

/* unknown src alignment and < 8 bytes to go  */
.Lsmall_csumcpy:
	move	a1, t2

	andi	t0, a1, 4
	beqz	t0, 1f
	 andi	t0, a1, 2

	/* Still a full word to go  */
	ulw	t1, (src)
	PTR_ADDIU	src, 4
#ifdef USE_DOUBLE
	dsll	t1, t1, 32			/* clear lower 32bit */
#endif
	ADDC(sum, t1)

1:	move	t1, zero
	beqz	t0, 1f
	 andi	t0, a1, 1

	/* Still a halfword to go  */
	ulhu	t1, (src)
	PTR_ADDIU	src, 2

1:	beqz	t0, 1f
	 sll	t1, t1, 16

	lbu	t2, (src)
	 nop

#ifdef __MIPSEB__
	sll	t2, t2, 8
#endif
	or	t1, t2

1:	ADDC(sum, t1)

	/* fold checksum */
#ifdef USE_DOUBLE
	dsll32	v1, sum, 0
	daddu	sum, v1
	sltu	v1, sum, v1
	dsra32	sum, sum, 0
	addu	sum, v1
#endif

	/* odd buffer alignment? */
#ifdef CONFIG_CPU_MIPSR2
	wsbh	v1, sum
	movn	sum, v1, t7
#else
	beqz	t7, 1f			/* odd buffer alignment? */
	 lui	v1, 0x00ff
	addu	v1, 0x00ff
	and	t0, sum, v1
	sll	t0, t0, 8
	srl	sum, sum, 8
	and	sum, sum, v1
	or	sum, sum, t0
1:
#endif
	.set	reorder
	/* Add the passed partial csum.	 */
	ADDC32(sum, a2)
	jr	ra
	.set	noreorder
	END(csum_partial)


/*
 * checksum and copy routines based on memcpy.S
 *
 *	csum_partial_copy_nocheck(src, dst, len, sum)
 *	__csum_partial_copy_user(src, dst, len, sum, errp)
 *
 * See "Spec" in memcpy.S for details.	Unlike __copy_user, all
 * function in this file use the standard calling convention.
 */

#define src a0
#define dst a1
#define len a2
#define psum a3
#define sum v0
#define odd t8
#define errptr t9

/*
 * The exception handler for loads requires that:
 *  1- AT contain the address of the byte just past the end of the source
 *     of the copy,
 *  2- src_entry <= src < AT, and
 *  3- (dst - src) == (dst_entry - src_entry),
 * The _entry suffix denotes values when __copy_user was called.
 *
 * (1) is set up up by __csum_partial_copy_from_user and maintained by
 *	not writing AT in __csum_partial_copy
 * (2) is met by incrementing src by the number of bytes copied
 * (3) is met by not doing loads between a pair of increments of dst and src
 *
 * The exception handlers for stores stores -EFAULT to errptr and return.
 * These handlers do not need to overwrite any data.
 */

#define EXC(inst_reg,addr,handler)		\
9:	inst_reg, addr;				\
	.section __ex_table,"a";		\
	PTR	9b, handler;			\
	.previous

#ifdef USE_DOUBLE

#define LOAD   ld
#define LOADL  ldl
#define LOADR  ldr
#define STOREL sdl
#define STORER sdr
#define STORE  sd
#define ADD    daddu
#define SUB    dsubu
#define SRL    dsrl
#define SLL    dsll
#define SLLV   dsllv
#define SRLV   dsrlv
#define NBYTES 8
#define LOG_NBYTES 3

#else

#define LOAD   lw
#define LOADL  lwl
#define LOADR  lwr
#define STOREL swl
#define STORER swr
#define STORE  sw
#define ADD    addu
#define SUB    subu
#define SRL    srl
#define SLL    sll
#define SLLV   sllv
#define SRLV   srlv
#define NBYTES 4
#define LOG_NBYTES 2

#endif /* USE_DOUBLE */

#ifdef CONFIG_CPU_LITTLE_ENDIAN
#define LDFIRST LOADR
#define LDREST	LOADL
#define STFIRST STORER
#define STREST	STOREL
#define SHIFT_DISCARD SLLV
#define SHIFT_DISCARD_REVERT SRLV
#else
#define LDFIRST LOADL
#define LDREST	LOADR
#define STFIRST STOREL
#define STREST	STORER
#define SHIFT_DISCARD SRLV
#define SHIFT_DISCARD_REVERT SLLV
#endif

#define FIRST(unit) ((unit)*NBYTES)
#define REST(unit)  (FIRST(unit)+NBYTES-1)

#define ADDRMASK (NBYTES-1)

#ifndef CONFIG_CPU_DADDI_WORKAROUNDS
	.set	noat
#else
	.set	at=v1
#endif

LEAF(__csum_partial_copy_user)
	PTR_ADDU	AT, src, len	/* See (1) above. */
#ifdef CONFIG_64BIT
	move	errptr, a4
#else
	lw	errptr, 16(sp)
#endif
FEXPORT(csum_partial_copy_nocheck)
	move	sum, zero
	move	odd, zero
	/*
	 * Note: dst & src may be unaligned, len may be 0
	 * Temps
	 */
	/*
	 * The "issue break"s below are very approximate.
	 * Issue delays for dcache fills will perturb the schedule, as will
	 * load queue full replay traps, etc.
	 *
	 * If len < NBYTES use byte operations.
	 */
	sltu	t2, len, NBYTES
	and	t1, dst, ADDRMASK
	bnez	t2, .Lcopy_bytes_checklen
	 and	t0, src, ADDRMASK
	andi	odd, dst, 0x1			/* odd buffer? */
	bnez	t1, .Ldst_unaligned
	 nop
	bnez	t0, .Lsrc_unaligned_dst_aligned
	/*
	 * use delay slot for fall-through
	 * src and dst are aligned; need to compute rem
	 */
.Lboth_aligned:
	 SRL	t0, len, LOG_NBYTES+3	 # +3 for 8 units/iter
	beqz	t0, .Lcleanup_both_aligned # len < 8*NBYTES
	 nop
	SUB	len, 8*NBYTES		# subtract here for bgez loop
	.align	4
1:
EXC(	LOAD	t0, UNIT(0)(src),	.Ll_exc)
EXC(	LOAD	t1, UNIT(1)(src),	.Ll_exc_copy)
EXC(	LOAD	t2, UNIT(2)(src),	.Ll_exc_copy)
EXC(	LOAD	t3, UNIT(3)(src),	.Ll_exc_copy)
EXC(	LOAD	t4, UNIT(4)(src),	.Ll_exc_copy)
EXC(	LOAD	t5, UNIT(5)(src),	.Ll_exc_copy)
EXC(	LOAD	t6, UNIT(6)(src),	.Ll_exc_copy)
EXC(	LOAD	t7, UNIT(7)(src),	.Ll_exc_copy)
	SUB	len, len, 8*NBYTES
	ADD	src, src, 8*NBYTES
EXC(	STORE	t0, UNIT(0)(dst),	.Ls_exc)
	ADDC(sum, t0)
EXC(	STORE	t1, UNIT(1)(dst),	.Ls_exc)
	ADDC(sum, t1)
EXC(	STORE	t2, UNIT(2)(dst),	.Ls_exc)
	ADDC(sum, t2)
EXC(	STORE	t3, UNIT(3)(dst),	.Ls_exc)
	ADDC(sum, t3)
EXC(	STORE	t4, UNIT(4)(dst),	.Ls_exc)
	ADDC(sum, t4)
EXC(	STORE	t5, UNIT(5)(dst),	.Ls_exc)
	ADDC(sum, t5)
EXC(	STORE	t6, UNIT(6)(dst),	.Ls_exc)
	ADDC(sum, t6)
EXC(	STORE	t7, UNIT(7)(dst),	.Ls_exc)
	ADDC(sum, t7)
	.set	reorder				/* DADDI_WAR */
	ADD	dst, dst, 8*NBYTES
	bgez	len, 1b
	.set	noreorder
	ADD	len, 8*NBYTES		# revert len (see above)

	/*
	 * len == the number of bytes left to copy < 8*NBYTES
	 */
.Lcleanup_both_aligned:
#define rem t7
	beqz	len, .Ldone
	 sltu	t0, len, 4*NBYTES
	bnez	t0, .Lless_than_4units
	 and	rem, len, (NBYTES-1)	# rem = len % NBYTES
	/*
	 * len >= 4*NBYTES
	 */
EXC(	LOAD	t0, UNIT(0)(src),	.Ll_exc)
EXC(	LOAD	t1, UNIT(1)(src),	.Ll_exc_copy)
EXC(	LOAD	t2, UNIT(2)(src),	.Ll_exc_copy)
EXC(	LOAD	t3, UNIT(3)(src),	.Ll_exc_copy)
	SUB	len, len, 4*NBYTES
	ADD	src, src, 4*NBYTES
EXC(	STORE	t0, UNIT(0)(dst),	.Ls_exc)
	ADDC(sum, t0)
EXC(	STORE	t1, UNIT(1)(dst),	.Ls_exc)
	ADDC(sum, t1)
EXC(	STORE	t2, UNIT(2)(dst),	.Ls_exc)
	ADDC(sum, t2)
EXC(	STORE	t3, UNIT(3)(dst),	.Ls_exc)
	ADDC(sum, t3)
	.set	reorder				/* DADDI_WAR */
	ADD	dst, dst, 4*NBYTES
	beqz	len, .Ldone
	.set	noreorder
.Lless_than_4units:
	/*
	 * rem = len % NBYTES
	 */
	beq	rem, len, .Lcopy_bytes
	 nop
1:
EXC(	LOAD	t0, 0(src),		.Ll_exc)
	ADD	src, src, NBYTES
	SUB	len, len, NBYTES
EXC(	STORE	t0, 0(dst),		.Ls_exc)
	ADDC(sum, t0)
	.set	reorder				/* DADDI_WAR */
	ADD	dst, dst, NBYTES
	bne	rem, len, 1b
	.set	noreorder

	/*
	 * src and dst are aligned, need to copy rem bytes (rem < NBYTES)
	 * A loop would do only a byte at a time with possible branch
	 * mispredicts.	 Can't do an explicit LOAD dst,mask,or,STORE
	 * because can't assume read-access to dst.  Instead, use
	 * STREST dst, which doesn't require read access to dst.
	 *
	 * This code should perform better than a simple loop on modern,
	 * wide-issue mips processors because the code has fewer branches and
	 * more instruction-level parallelism.
	 */
#define bits t2
	beqz	len, .Ldone
	 ADD	t1, dst, len	# t1 is just past last byte of dst
	li	bits, 8*NBYTES
	SLL	rem, len, 3	# rem = number of bits to keep
EXC(	LOAD	t0, 0(src),		.Ll_exc)
	SUB	bits, bits, rem # bits = number of bits to discard
	SHIFT_DISCARD t0, t0, bits
EXC(	STREST	t0, -1(t1),		.Ls_exc)
	SHIFT_DISCARD_REVERT t0, t0, bits
	.set reorder
	ADDC(sum, t0)
	b	.Ldone
	.set noreorder
.Ldst_unaligned:
	/*
	 * dst is unaligned
	 * t0 = src & ADDRMASK
	 * t1 = dst & ADDRMASK; T1 > 0
	 * len >= NBYTES
	 *
	 * Copy enough bytes to align dst
	 * Set match = (src and dst have same alignment)
	 */
#define match rem
EXC(	LDFIRST t3, FIRST(0)(src),	.Ll_exc)
	ADD	t2, zero, NBYTES
EXC(	LDREST	t3, REST(0)(src),	.Ll_exc_copy)
	SUB	t2, t2, t1	# t2 = number of bytes copied
	xor	match, t0, t1
EXC(	STFIRST t3, FIRST(0)(dst),	.Ls_exc)
	SLL	t4, t1, 3		# t4 = number of bits to discard
	SHIFT_DISCARD t3, t3, t4
	/* no SHIFT_DISCARD_REVERT to handle odd buffer properly */
	ADDC(sum, t3)
	beq	len, t2, .Ldone
	 SUB	len, len, t2
	ADD	dst, dst, t2
	beqz	match, .Lboth_aligned
	 ADD	src, src, t2

.Lsrc_unaligned_dst_aligned:
	SRL	t0, len, LOG_NBYTES+2	 # +2 for 4 units/iter
	beqz	t0, .Lcleanup_src_unaligned
	 and	rem, len, (4*NBYTES-1)	 # rem = len % 4*NBYTES
1:
/*
 * Avoid consecutive LD*'s to the same register since some mips
 * implementations can't issue them in the same cycle.
 * It's OK to load FIRST(N+1) before REST(N) because the two addresses
 * are to the same unit (unless src is aligned, but it's not).
 */
EXC(	LDFIRST t0, FIRST(0)(src),	.Ll_exc)
EXC(	LDFIRST t1, FIRST(1)(src),	.Ll_exc_copy)
	SUB	len, len, 4*NBYTES
EXC(	LDREST	t0, REST(0)(src),	.Ll_exc_copy)
EXC(	LDREST	t1, REST(1)(src),	.Ll_exc_copy)
EXC(	LDFIRST t2, FIRST(2)(src),	.Ll_exc_copy)
EXC(	LDFIRST t3, FIRST(3)(src),	.Ll_exc_copy)
EXC(	LDREST	t2, REST(2)(src),	.Ll_exc_copy)
EXC(	LDREST	t3, REST(3)(src),	.Ll_exc_copy)
	ADD	src, src, 4*NBYTES
#ifdef CONFIG_CPU_SB1
	nop				# improves slotting
#endif
EXC(	STORE	t0, UNIT(0)(dst),	.Ls_exc)
	ADDC(sum, t0)
EXC(	STORE	t1, UNIT(1)(dst),	.Ls_exc)
	ADDC(sum, t1)
EXC(	STORE	t2, UNIT(2)(dst),	.Ls_exc)
	ADDC(sum, t2)
EXC(	STORE	t3, UNIT(3)(dst),	.Ls_exc)
	ADDC(sum, t3)
	.set	reorder				/* DADDI_WAR */
	ADD	dst, dst, 4*NBYTES
	bne	len, rem, 1b
	.set	noreorder

.Lcleanup_src_unaligned:
	beqz	len, .Ldone
	 and	rem, len, NBYTES-1  # rem = len % NBYTES
	beq	rem, len, .Lcopy_bytes
	 nop
1:
EXC(	LDFIRST t0, FIRST(0)(src),	.Ll_exc)
EXC(	LDREST	t0, REST(0)(src),	.Ll_exc_copy)
	ADD	src, src, NBYTES
	SUB	len, len, NBYTES
EXC(	STORE	t0, 0(dst),		.Ls_exc)
	ADDC(sum, t0)
	.set	reorder				/* DADDI_WAR */
	ADD	dst, dst, NBYTES
	bne	len, rem, 1b
	.set	noreorder

.Lcopy_bytes_checklen:
	beqz	len, .Ldone
	 nop
.Lcopy_bytes:
	/* 0 < len < NBYTES  */
#ifdef CONFIG_CPU_LITTLE_ENDIAN
#define SHIFT_START 0
#define SHIFT_INC 8
#else
#define SHIFT_START 8*(NBYTES-1)
#define SHIFT_INC -8
#endif
	move	t2, zero	# partial word
	li	t3, SHIFT_START # shift
/* use .Ll_exc_copy here to return correct sum on fault */
#define COPY_BYTE(N)			\
EXC(	lbu	t0, N(src), .Ll_exc_copy);	\
	SUB	len, len, 1;		\
EXC(	sb	t0, N(dst), .Ls_exc);	\
	SLLV	t0, t0, t3;		\
	addu	t3, SHIFT_INC;		\
	beqz	len, .Lcopy_bytes_done; \
	 or	t2, t0

	COPY_BYTE(0)
	COPY_BYTE(1)
#ifdef USE_DOUBLE
	COPY_BYTE(2)
	COPY_BYTE(3)
	COPY_BYTE(4)
	COPY_BYTE(5)
#endif
EXC(	lbu	t0, NBYTES-2(src), .Ll_exc_copy)
	SUB	len, len, 1
EXC(	sb	t0, NBYTES-2(dst), .Ls_exc)
	SLLV	t0, t0, t3
	or	t2, t0
.Lcopy_bytes_done:
	ADDC(sum, t2)
.Ldone:
	/* fold checksum */
#ifdef USE_DOUBLE
	dsll32	v1, sum, 0
	daddu	sum, v1
	sltu	v1, sum, v1
	dsra32	sum, sum, 0
	addu	sum, v1
#endif

#ifdef CONFIG_CPU_MIPSR2
	wsbh	v1, sum
	movn	sum, v1, odd
#else
	beqz	odd, 1f			/* odd buffer alignment? */
	 lui	v1, 0x00ff
	addu	v1, 0x00ff
	and	t0, sum, v1
	sll	t0, t0, 8
	srl	sum, sum, 8
	and	sum, sum, v1
	or	sum, sum, t0
1:
#endif
	.set reorder
	ADDC32(sum, psum)
	jr	ra
	.set noreorder

.Ll_exc_copy:
	/*
	 * Copy bytes from src until faulting load address (or until a
	 * lb faults)
	 *
	 * When reached by a faulting LDFIRST/LDREST, THREAD_BUADDR($28)
	 * may be more than a byte beyond the last address.
	 * Hence, the lb below may get an exception.
	 *
	 * Assumes src < THREAD_BUADDR($28)
	 */
	LOAD	t0, TI_TASK($28)
	 li	t2, SHIFT_START
	LOAD	t0, THREAD_BUADDR(t0)
1:
EXC(	lbu	t1, 0(src),	.Ll_exc)
	ADD	src, src, 1
	sb	t1, 0(dst)	# can't fault -- we're copy_from_user
	SLLV	t1, t1, t2
	addu	t2, SHIFT_INC
	ADDC(sum, t1)
	.set	reorder				/* DADDI_WAR */
	ADD	dst, dst, 1
	bne	src, t0, 1b
	.set	noreorder
.Ll_exc:
	LOAD	t0, TI_TASK($28)
	 nop
	LOAD	t0, THREAD_BUADDR(t0)	# t0 is just past last good address
	 nop
	SUB	len, AT, t0		# len number of uncopied bytes
	/*
	 * Here's where we rely on src and dst being incremented in tandem,
	 *   See (3) above.
	 * dst += (fault addr - src) to put dst at first byte to clear
	 */
	ADD	dst, t0			# compute start address in a1
	SUB	dst, src
	/*
	 * Clear len bytes starting at dst.  Can't call __bzero because it
	 * might modify len.  An inefficient loop for these rare times...
	 */
	.set	reorder				/* DADDI_WAR */
	SUB	src, len, 1
	beqz	len, .Ldone
	.set	noreorder
1:	sb	zero, 0(dst)
	ADD	dst, dst, 1
	.set	push
	.set	noat
#ifndef CONFIG_CPU_DADDI_WORKAROUNDS
	bnez	src, 1b
	 SUB	src, src, 1
#else
	li	v1, 1
	bnez	src, 1b
	 SUB	src, src, v1
#endif
	li	v1, -EFAULT
	b	.Ldone
	 sw	v1, (errptr)

.Ls_exc:
	li	v0, -1 /* invalid checksum */
	li	v1, -EFAULT
	jr	ra
	 sw	v1, (errptr)
	.set	pop
	END(__csum_partial_copy_user)