xref: /linux/arch/xtensa/lib/strncpy_user.S (revision c537b994505099b7197e7d3125b942ecbcc51eb6)
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 <asm/variant/core.h>
15#include <linux/errno.h>
16
17/* Load or store instructions that may cause exceptions use the EX macro. */
18
19#define EX(insn,reg1,reg2,offset,handler)	\
209:	insn	reg1, reg2, offset;		\
21	.section __ex_table, "a";		\
22	.word	9b, handler;			\
23	.previous
24
25/*
26 * char *__strncpy_user(char *dst, const char *src, size_t len)
27 */
28.text
29.begin literal
30.align	4
31.Lmask0:
32	.byte	0xff, 0x00, 0x00, 0x00
33.Lmask1:
34	.byte	0x00, 0xff, 0x00, 0x00
35.Lmask2:
36	.byte	0x00, 0x00, 0xff, 0x00
37.Lmask3:
38	.byte	0x00, 0x00, 0x00, 0xff
39.end literal
40
41# Register use
42#   a0/ return address
43#   a1/ stack pointer
44#   a2/ return value
45#   a3/ src
46#   a4/ len
47#   a5/ mask0
48#   a6/ mask1
49#   a7/ mask2
50#   a8/ mask3
51#   a9/ tmp
52#   a10/ tmp
53#   a11/ dst
54#   a12/ tmp
55
56.align	4
57.global	__strncpy_user
58.type	__strncpy_user,@function
59__strncpy_user:
60	entry	sp, 16		# minimal stack frame
61	# a2/ dst, a3/ src, a4/ len
62	mov	a11, a2		# leave dst in return value register
63	beqz	a4, .Lret	# if len is zero
64	l32r	a5, .Lmask0	# mask for byte 0
65	l32r	a6, .Lmask1	# mask for byte 1
66	l32r	a7, .Lmask2	# mask for byte 2
67	l32r	a8, .Lmask3	# mask for byte 3
68	bbsi.l	a3, 0, .Lsrc1mod2 # if only  8-bit aligned
69	bbsi.l	a3, 1, .Lsrc2mod4 # if only 16-bit aligned
70.Lsrcaligned:	# return here when src is word-aligned
71	srli	a12, a4, 2	# number of loop iterations with 4B per loop
72	movi	a9, 3
73	bnone	a11, a9, .Laligned
74	j	.Ldstunaligned
75
76.Lsrc1mod2:	# src address is odd
77	EX(l8ui, a9, a3, 0, fixup_l)	# get byte 0
78	addi	a3, a3, 1		# advance src pointer
79	EX(s8i, a9, a11, 0, fixup_s)	# store byte 0
80	beqz	a9, .Lret		# if byte 0 is zero
81	addi	a11, a11, 1		# advance dst pointer
82	addi	a4, a4, -1		# decrement len
83	beqz	a4, .Lret		# if len is zero
84	bbci.l	a3, 1, .Lsrcaligned	# if src is now word-aligned
85
86.Lsrc2mod4:	# src address is 2 mod 4
87	EX(l8ui, a9, a3, 0, fixup_l)	# get byte 0
88	/* 1-cycle interlock */
89	EX(s8i, a9, a11, 0, fixup_s)	# store byte 0
90	beqz	a9, .Lret		# if byte 0 is zero
91	addi	a11, a11, 1		# advance dst pointer
92	addi	a4, a4, -1		# decrement len
93	beqz	a4, .Lret		# if len is zero
94	EX(l8ui, a9, a3, 1, fixup_l)	# get byte 0
95	addi	a3, a3, 2		# advance src pointer
96	EX(s8i, a9, a11, 0, fixup_s)	# store byte 0
97	beqz	a9, .Lret		# if byte 0 is zero
98	addi	a11, a11, 1		# advance dst pointer
99	addi	a4, a4, -1		# decrement len
100	bnez	a4, .Lsrcaligned	# if len is nonzero
101.Lret:
102	sub	a2, a11, a2		# compute strlen
103	retw
104
105/*
106 * dst is word-aligned, src is word-aligned
107 */
108	.align	4		# 1 mod 4 alignment for LOOPNEZ
109	.byte	0		# (0 mod 4 alignment for LBEG)
110.Laligned:
111#if XCHAL_HAVE_LOOPS
112	loopnez	a12, .Loop1done
113#else
114	beqz	a12, .Loop1done
115	slli	a12, a12, 2
116	add	a12, a12, a11	# a12 = end of last 4B chunck
117#endif
118.Loop1:
119	EX(l32i, a9, a3, 0, fixup_l)	# get word from src
120	addi	a3, a3, 4		# advance src pointer
121	bnone	a9, a5, .Lz0		# if byte 0 is zero
122	bnone	a9, a6, .Lz1		# if byte 1 is zero
123	bnone	a9, a7, .Lz2		# if byte 2 is zero
124	EX(s32i, a9, a11, 0, fixup_s)	# store word to dst
125	bnone	a9, a8, .Lz3		# if byte 3 is zero
126	addi	a11, a11, 4		# advance dst pointer
127#if !XCHAL_HAVE_LOOPS
128	blt	a11, a12, .Loop1
129#endif
130
131.Loop1done:
132	bbci.l	a4, 1, .L100
133	# copy 2 bytes
134	EX(l16ui, a9, a3, 0, fixup_l)
135	addi	a3, a3, 2		# advance src pointer
136#ifdef __XTENSA_EB__
137	bnone	a9, a7, .Lz0		# if byte 2 is zero
138	bnone	a9, a8, .Lz1		# if byte 3 is zero
139#else
140	bnone	a9, a5, .Lz0		# if byte 0 is zero
141	bnone	a9, a6, .Lz1		# if byte 1 is zero
142#endif
143	EX(s16i, a9, a11, 0, fixup_s)
144	addi	a11, a11, 2		# advance dst pointer
145.L100:
146	bbci.l	a4, 0, .Lret
147	EX(l8ui, a9, a3, 0, fixup_l)
148	/* slot */
149	EX(s8i, a9, a11, 0, fixup_s)
150	beqz	a9, .Lret		# if byte is zero
151	addi	a11, a11, 1-3		# advance dst ptr 1, but also cancel
152					# the effect of adding 3 in .Lz3 code
153	/* fall thru to .Lz3 and "retw" */
154
155.Lz3:	# byte 3 is zero
156	addi	a11, a11, 3		# advance dst pointer
157	sub	a2, a11, a2		# compute strlen
158	retw
159.Lz0:	# byte 0 is zero
160#ifdef __XTENSA_EB__
161	movi	a9, 0
162#endif /* __XTENSA_EB__ */
163	EX(s8i, a9, a11, 0, fixup_s)
164	sub	a2, a11, a2		# compute strlen
165	retw
166.Lz1:	# byte 1 is zero
167#ifdef __XTENSA_EB__
168        extui   a9, a9, 16, 16
169#endif /* __XTENSA_EB__ */
170	EX(s16i, a9, a11, 0, fixup_s)
171	addi	a11, a11, 1		# advance dst pointer
172	sub	a2, a11, a2		# compute strlen
173	retw
174.Lz2:	# byte 2 is zero
175#ifdef __XTENSA_EB__
176        extui   a9, a9, 16, 16
177#endif /* __XTENSA_EB__ */
178	EX(s16i, a9, a11, 0, fixup_s)
179	movi	a9, 0
180	EX(s8i, a9, a11, 2, fixup_s)
181	addi	a11, a11, 2		# advance dst pointer
182	sub	a2, a11, a2		# compute strlen
183	retw
184
185	.align	4		# 1 mod 4 alignment for LOOPNEZ
186	.byte	0		# (0 mod 4 alignment for LBEG)
187.Ldstunaligned:
188/*
189 * for now just use byte copy loop
190 */
191#if XCHAL_HAVE_LOOPS
192	loopnez	a4, .Lunalignedend
193#else
194	beqz	a4, .Lunalignedend
195	add	a12, a11, a4		# a12 = ending address
196#endif /* XCHAL_HAVE_LOOPS */
197.Lnextbyte:
198	EX(l8ui, a9, a3, 0, fixup_l)
199	addi	a3, a3, 1
200	EX(s8i, a9, a11, 0, fixup_s)
201	beqz	a9, .Lunalignedend
202	addi	a11, a11, 1
203#if !XCHAL_HAVE_LOOPS
204	blt	a11, a12, .Lnextbyte
205#endif
206
207.Lunalignedend:
208	sub	a2, a11, a2		# compute strlen
209	retw
210
211
212	.section .fixup, "ax"
213	.align	4
214
215	/* For now, just return -EFAULT.  Future implementations might
216	 * like to clear remaining kernel space, like the fixup
217	 * implementation in memset().  Thus, we differentiate between
218	 * load/store fixups. */
219
220fixup_s:
221fixup_l:
222	movi	a2, -EFAULT
223	retw
224
225