1/*
2 *  arch/xtensa/lib/strncpy_user.S
3 *
4 *  This file is subject to the terms and conditions of the GNU General
5 *  Public License.  See the file "COPYING" in the main directory of
6 *  this archive for more details.
7 *
8 *  Returns: -EFAULT if exception before terminator, N if the entire
9 *  buffer filled, else strlen.
10 *
11 *  Copyright (C) 2002 Tensilica Inc.
12 */
13
14#include <linux/errno.h>
15#include <linux/linkage.h>
16#include <asm/asmmacro.h>
17#include <asm/core.h>
18
19/*
20 * char *__strncpy_user(char *dst, const char *src, size_t len)
21 */
22
23#ifdef __XTENSA_EB__
24# define MASK0 0xff000000
25# define MASK1 0x00ff0000
26# define MASK2 0x0000ff00
27# define MASK3 0x000000ff
28#else
29# define MASK0 0x000000ff
30# define MASK1 0x0000ff00
31# define MASK2 0x00ff0000
32# define MASK3 0xff000000
33#endif
34
35# Register use
36#   a0/ return address
37#   a1/ stack pointer
38#   a2/ return value
39#   a3/ src
40#   a4/ len
41#   a5/ mask0
42#   a6/ mask1
43#   a7/ mask2
44#   a8/ mask3
45#   a9/ tmp
46#   a10/ tmp
47#   a11/ dst
48
49.text
50ENTRY(__strncpy_user)
51
52	abi_entry_default
53	# a2/ dst, a3/ src, a4/ len
54	mov	a11, a2		# leave dst in return value register
55	beqz	a4, .Lret	# if len is zero
56	movi	a5, MASK0	# mask for byte 0
57	movi	a6, MASK1	# mask for byte 1
58	movi	a7, MASK2	# mask for byte 2
59	movi	a8, MASK3	# mask for byte 3
60	bbsi.l	a3, 0, .Lsrc1mod2 # if only  8-bit aligned
61	bbsi.l	a3, 1, .Lsrc2mod4 # if only 16-bit aligned
62.Lsrcaligned:	# return here when src is word-aligned
63	srli	a10, a4, 2	# number of loop iterations with 4B per loop
64	movi	a9, 3
65	bnone	a11, a9, .Laligned
66	j	.Ldstunaligned
67
68.Lsrc1mod2:	# src address is odd
69EX(11f)	l8ui	a9, a3, 0		# get byte 0
70	addi	a3, a3, 1		# advance src pointer
71EX(10f)	s8i	a9, a11, 0		# store byte 0
72	beqz	a9, .Lret		# if byte 0 is zero
73	addi	a11, a11, 1		# advance dst pointer
74	addi	a4, a4, -1		# decrement len
75	beqz	a4, .Lret		# if len is zero
76	bbci.l	a3, 1, .Lsrcaligned	# if src is now word-aligned
77
78.Lsrc2mod4:	# src address is 2 mod 4
79EX(11f)	l8ui	a9, a3, 0		# get byte 0
80	/* 1-cycle interlock */
81EX(10f)	s8i	a9, a11, 0		# store byte 0
82	beqz	a9, .Lret		# if byte 0 is zero
83	addi	a11, a11, 1		# advance dst pointer
84	addi	a4, a4, -1		# decrement len
85	beqz	a4, .Lret		# if len is zero
86EX(11f)	l8ui	a9, a3, 1		# get byte 0
87	addi	a3, a3, 2		# advance src pointer
88EX(10f)	s8i	a9, a11, 0		# store byte 0
89	beqz	a9, .Lret		# if byte 0 is zero
90	addi	a11, a11, 1		# advance dst pointer
91	addi	a4, a4, -1		# decrement len
92	bnez	a4, .Lsrcaligned	# if len is nonzero
93.Lret:
94	sub	a2, a11, a2		# compute strlen
95	abi_ret_default
96
97/*
98 * dst is word-aligned, src is word-aligned
99 */
100	.align	4		# 1 mod 4 alignment for LOOPNEZ
101	.byte	0		# (0 mod 4 alignment for LBEG)
102.Laligned:
103#if XCHAL_HAVE_LOOPS
104	loopnez	a10, .Loop1done
105#else
106	beqz	a10, .Loop1done
107	slli	a10, a10, 2
108	add	a10, a10, a11	# a10 = end of last 4B chunck
109#endif
110.Loop1:
111EX(11f)	l32i	a9, a3, 0		# get word from src
112	addi	a3, a3, 4		# advance src pointer
113	bnone	a9, a5, .Lz0		# if byte 0 is zero
114	bnone	a9, a6, .Lz1		# if byte 1 is zero
115	bnone	a9, a7, .Lz2		# if byte 2 is zero
116EX(10f)	s32i	a9, a11, 0		# store word to dst
117	bnone	a9, a8, .Lz3		# if byte 3 is zero
118	addi	a11, a11, 4		# advance dst pointer
119#if !XCHAL_HAVE_LOOPS
120	blt	a11, a10, .Loop1
121#endif
122
123.Loop1done:
124	bbci.l	a4, 1, .L100
125	# copy 2 bytes
126EX(11f)	l16ui	a9, a3, 0
127	addi	a3, a3, 2		# advance src pointer
128#ifdef __XTENSA_EB__
129	bnone	a9, a7, .Lz0		# if byte 2 is zero
130	bnone	a9, a8, .Lz1		# if byte 3 is zero
131#else
132	bnone	a9, a5, .Lz0		# if byte 0 is zero
133	bnone	a9, a6, .Lz1		# if byte 1 is zero
134#endif
135EX(10f)	s16i	a9, a11, 0
136	addi	a11, a11, 2		# advance dst pointer
137.L100:
138	bbci.l	a4, 0, .Lret
139EX(11f)	l8ui	a9, a3, 0
140	/* slot */
141EX(10f)	s8i	a9, a11, 0
142	beqz	a9, .Lret		# if byte is zero
143	addi	a11, a11, 1-3		# advance dst ptr 1, but also cancel
144					# the effect of adding 3 in .Lz3 code
145	/* fall thru to .Lz3 and "retw" */
146
147.Lz3:	# byte 3 is zero
148	addi	a11, a11, 3		# advance dst pointer
149	sub	a2, a11, a2		# compute strlen
150	abi_ret_default
151.Lz0:	# byte 0 is zero
152#ifdef __XTENSA_EB__
153	movi	a9, 0
154#endif /* __XTENSA_EB__ */
155EX(10f)	s8i	a9, a11, 0
156	sub	a2, a11, a2		# compute strlen
157	abi_ret_default
158.Lz1:	# byte 1 is zero
159#ifdef __XTENSA_EB__
160	extui   a9, a9, 16, 16
161#endif /* __XTENSA_EB__ */
162EX(10f)	s16i	a9, a11, 0
163	addi	a11, a11, 1		# advance dst pointer
164	sub	a2, a11, a2		# compute strlen
165	abi_ret_default
166.Lz2:	# byte 2 is zero
167#ifdef __XTENSA_EB__
168	extui   a9, a9, 16, 16
169#endif /* __XTENSA_EB__ */
170EX(10f)	s16i	a9, a11, 0
171	movi	a9, 0
172EX(10f)	s8i	a9, a11, 2
173	addi	a11, a11, 2		# advance dst pointer
174	sub	a2, a11, a2		# compute strlen
175	abi_ret_default
176
177	.align	4		# 1 mod 4 alignment for LOOPNEZ
178	.byte	0		# (0 mod 4 alignment for LBEG)
179.Ldstunaligned:
180/*
181 * for now just use byte copy loop
182 */
183#if XCHAL_HAVE_LOOPS
184	loopnez	a4, .Lunalignedend
185#else
186	beqz	a4, .Lunalignedend
187	add	a10, a11, a4		# a10 = ending address
188#endif /* XCHAL_HAVE_LOOPS */
189.Lnextbyte:
190EX(11f)	l8ui	a9, a3, 0
191	addi	a3, a3, 1
192EX(10f)	s8i	a9, a11, 0
193	beqz	a9, .Lunalignedend
194	addi	a11, a11, 1
195#if !XCHAL_HAVE_LOOPS
196	blt	a11, a10, .Lnextbyte
197#endif
198
199.Lunalignedend:
200	sub	a2, a11, a2		# compute strlen
201	abi_ret_default
202
203ENDPROC(__strncpy_user)
204EXPORT_SYMBOL(__strncpy_user)
205
206	.section .fixup, "ax"
207	.align	4
208
209	/* For now, just return -EFAULT.  Future implementations might
210	 * like to clear remaining kernel space, like the fixup
211	 * implementation in memset().  Thus, we differentiate between
212	 * load/store fixups. */
213
21410:
21511:
216	movi	a2, -EFAULT
217	abi_ret_default
218