aboutsummaryrefslogtreecommitdiffstats
path: root/arch/sh/include
diff options
context:
space:
mode:
authorPaul Mundt <lethal@linux-sh.org>2010-12-24 05:19:23 -0500
committerPaul Mundt <lethal@linux-sh.org>2010-12-24 05:19:23 -0500
commit1dee92bba36b491dbcc15e2cba40501403237f96 (patch)
tree5cf2800692ed722a72a89f965937c9cb368159db /arch/sh/include
parent03aa18f550900855c1d3d17ac83c14a3d668d344 (diff)
sh: Tidy up SH-4A unaligned load support.
The current implementation was rather tied to the packed_struct.h definitions, which immediately began to clash when the packed_struct.h types changed and drivers began to include packed_struct.h directly. In order to support this sort of use it's necessary to get out of the way with regards to namespace collisions, and at the same time we can also kill off some duplicate code now that the unaligned headers are a bit more broken out. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Diffstat (limited to 'arch/sh/include')
-rw-r--r--arch/sh/include/asm/unaligned-sh4a.h164
1 files changed, 52 insertions, 112 deletions
diff --git a/arch/sh/include/asm/unaligned-sh4a.h b/arch/sh/include/asm/unaligned-sh4a.h
index 9f4dd252c981..c48a9c3420da 100644
--- a/arch/sh/include/asm/unaligned-sh4a.h
+++ b/arch/sh/include/asm/unaligned-sh4a.h
@@ -18,10 +18,20 @@
18 * of spill registers and blowing up when building at low optimization 18 * of spill registers and blowing up when building at low optimization
19 * levels. See http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34777. 19 * levels. See http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34777.
20 */ 20 */
21#include <linux/unaligned/packed_struct.h>
21#include <linux/types.h> 22#include <linux/types.h>
22#include <asm/byteorder.h> 23#include <asm/byteorder.h>
23 24
24static __always_inline u32 __get_unaligned_cpu32(const u8 *p) 25static inline u16 sh4a_get_unaligned_cpu16(const u8 *p)
26{
27#ifdef __LITTLE_ENDIAN
28 return p[0] | p[1] << 8;
29#else
30 return p[0] << 8 | p[1];
31#endif
32}
33
34static __always_inline u32 sh4a_get_unaligned_cpu32(const u8 *p)
25{ 35{
26 unsigned long unaligned; 36 unsigned long unaligned;
27 37
@@ -34,218 +44,148 @@ static __always_inline u32 __get_unaligned_cpu32(const u8 *p)
34 return unaligned; 44 return unaligned;
35} 45}
36 46
37struct __una_u16 { u16 x __attribute__((packed)); };
38struct __una_u32 { u32 x __attribute__((packed)); };
39struct __una_u64 { u64 x __attribute__((packed)); };
40
41static inline u16 __get_unaligned_cpu16(const u8 *p)
42{
43#ifdef __LITTLE_ENDIAN
44 return p[0] | p[1] << 8;
45#else
46 return p[0] << 8 | p[1];
47#endif
48}
49
50/* 47/*
51 * Even though movua.l supports auto-increment on the read side, it can 48 * Even though movua.l supports auto-increment on the read side, it can
52 * only store to r0 due to instruction encoding constraints, so just let 49 * only store to r0 due to instruction encoding constraints, so just let
53 * the compiler sort it out on its own. 50 * the compiler sort it out on its own.
54 */ 51 */
55static inline u64 __get_unaligned_cpu64(const u8 *p) 52static inline u64 sh4a_get_unaligned_cpu64(const u8 *p)
56{ 53{
57#ifdef __LITTLE_ENDIAN 54#ifdef __LITTLE_ENDIAN
58 return (u64)__get_unaligned_cpu32(p + 4) << 32 | 55 return (u64)sh4a_get_unaligned_cpu32(p + 4) << 32 |
59 __get_unaligned_cpu32(p); 56 sh4a_get_unaligned_cpu32(p);
60#else 57#else
61 return (u64)__get_unaligned_cpu32(p) << 32 | 58 return (u64)sh4a_get_unaligned_cpu32(p) << 32 |
62 __get_unaligned_cpu32(p + 4); 59 sh4a_get_unaligned_cpu32(p + 4);
63#endif 60#endif
64} 61}
65 62
66static inline u16 get_unaligned_le16(const void *p) 63static inline u16 get_unaligned_le16(const void *p)
67{ 64{
68 return le16_to_cpu(__get_unaligned_cpu16(p)); 65 return le16_to_cpu(sh4a_get_unaligned_cpu16(p));
69} 66}
70 67
71static inline u32 get_unaligned_le32(const void *p) 68static inline u32 get_unaligned_le32(const void *p)
72{ 69{
73 return le32_to_cpu(__get_unaligned_cpu32(p)); 70 return le32_to_cpu(sh4a_get_unaligned_cpu32(p));
74} 71}
75 72
76static inline u64 get_unaligned_le64(const void *p) 73static inline u64 get_unaligned_le64(const void *p)
77{ 74{
78 return le64_to_cpu(__get_unaligned_cpu64(p)); 75 return le64_to_cpu(sh4a_get_unaligned_cpu64(p));
79} 76}
80 77
81static inline u16 get_unaligned_be16(const void *p) 78static inline u16 get_unaligned_be16(const void *p)
82{ 79{
83 return be16_to_cpu(__get_unaligned_cpu16(p)); 80 return be16_to_cpu(sh4a_get_unaligned_cpu16(p));
84} 81}
85 82
86static inline u32 get_unaligned_be32(const void *p) 83static inline u32 get_unaligned_be32(const void *p)
87{ 84{
88 return be32_to_cpu(__get_unaligned_cpu32(p)); 85 return be32_to_cpu(sh4a_get_unaligned_cpu32(p));
89} 86}
90 87
91static inline u64 get_unaligned_be64(const void *p) 88static inline u64 get_unaligned_be64(const void *p)
92{ 89{
93 return be64_to_cpu(__get_unaligned_cpu64(p)); 90 return be64_to_cpu(sh4a_get_unaligned_cpu64(p));
94} 91}
95 92
96static inline void __put_le16_noalign(u8 *p, u16 val) 93static inline void nonnative_put_le16(u16 val, u8 *p)
97{ 94{
98 *p++ = val; 95 *p++ = val;
99 *p++ = val >> 8; 96 *p++ = val >> 8;
100} 97}
101 98
102static inline void __put_le32_noalign(u8 *p, u32 val) 99static inline void nonnative_put_le32(u32 val, u8 *p)
103{ 100{
104 __put_le16_noalign(p, val); 101 nonnative_put_le16(val, p);
105 __put_le16_noalign(p + 2, val >> 16); 102 nonnative_put_le16(val >> 16, p + 2);
106} 103}
107 104
108static inline void __put_le64_noalign(u8 *p, u64 val) 105static inline void nonnative_put_le64(u64 val, u8 *p)
109{ 106{
110 __put_le32_noalign(p, val); 107 nonnative_put_le32(val, p);
111 __put_le32_noalign(p + 4, val >> 32); 108 nonnative_put_le32(val >> 32, p + 4);
112} 109}
113 110
114static inline void __put_be16_noalign(u8 *p, u16 val) 111static inline void nonnative_put_be16(u16 val, u8 *p)
115{ 112{
116 *p++ = val >> 8; 113 *p++ = val >> 8;
117 *p++ = val; 114 *p++ = val;
118} 115}
119 116
120static inline void __put_be32_noalign(u8 *p, u32 val) 117static inline void nonnative_put_be32(u32 val, u8 *p)
121{ 118{
122 __put_be16_noalign(p, val >> 16); 119 nonnative_put_be16(val >> 16, p);
123 __put_be16_noalign(p + 2, val); 120 nonnative_put_be16(val, p + 2);
124} 121}
125 122
126static inline void __put_be64_noalign(u8 *p, u64 val) 123static inline void nonnative_put_be64(u64 val, u8 *p)
127{ 124{
128 __put_be32_noalign(p, val >> 32); 125 nonnative_put_be32(val >> 32, p);
129 __put_be32_noalign(p + 4, val); 126 nonnative_put_be32(val, p + 4);
130} 127}
131 128
132static inline void put_unaligned_le16(u16 val, void *p) 129static inline void put_unaligned_le16(u16 val, void *p)
133{ 130{
134#ifdef __LITTLE_ENDIAN 131#ifdef __LITTLE_ENDIAN
135 ((struct __una_u16 *)p)->x = val; 132 __put_unaligned_cpu16(val, p);
136#else 133#else
137 __put_le16_noalign(p, val); 134 nonnative_put_le16(val, p);
138#endif 135#endif
139} 136}
140 137
141static inline void put_unaligned_le32(u32 val, void *p) 138static inline void put_unaligned_le32(u32 val, void *p)
142{ 139{
143#ifdef __LITTLE_ENDIAN 140#ifdef __LITTLE_ENDIAN
144 ((struct __una_u32 *)p)->x = val; 141 __put_unaligned_cpu32(val, p);
145#else 142#else
146 __put_le32_noalign(p, val); 143 nonnative_put_le32(val, p);
147#endif 144#endif
148} 145}
149 146
150static inline void put_unaligned_le64(u64 val, void *p) 147static inline void put_unaligned_le64(u64 val, void *p)
151{ 148{
152#ifdef __LITTLE_ENDIAN 149#ifdef __LITTLE_ENDIAN
153 ((struct __una_u64 *)p)->x = val; 150 __put_unaligned_cpu64(val, p);
154#else 151#else
155 __put_le64_noalign(p, val); 152 nonnative_put_le64(val, p);
156#endif 153#endif
157} 154}
158 155
159static inline void put_unaligned_be16(u16 val, void *p) 156static inline void put_unaligned_be16(u16 val, void *p)
160{ 157{
161#ifdef __BIG_ENDIAN 158#ifdef __BIG_ENDIAN
162 ((struct __una_u16 *)p)->x = val; 159 __put_unaligned_cpu16(val, p);
163#else 160#else
164 __put_be16_noalign(p, val); 161 nonnative_put_be16(val, p);
165#endif 162#endif
166} 163}
167 164
168static inline void put_unaligned_be32(u32 val, void *p) 165static inline void put_unaligned_be32(u32 val, void *p)
169{ 166{
170#ifdef __BIG_ENDIAN 167#ifdef __BIG_ENDIAN
171 ((struct __una_u32 *)p)->x = val; 168 __put_unaligned_cpu32(val, p);
172#else 169#else
173 __put_be32_noalign(p, val); 170 nonnative_put_be32(val, p);
174#endif 171#endif
175} 172}
176 173
177static inline void put_unaligned_be64(u64 val, void *p) 174static inline void put_unaligned_be64(u64 val, void *p)
178{ 175{
179#ifdef __BIG_ENDIAN 176#ifdef __BIG_ENDIAN
180 ((struct __una_u64 *)p)->x = val; 177 __put_unaligned_cpu64(val, p);
181#else 178#else
182 __put_be64_noalign(p, val); 179 nonnative_put_be64(val, p);
183#endif 180#endif
184} 181}
185 182
186/* 183/*
187 * Cause a link-time error if we try an unaligned access other than 184 * While it's a bit non-obvious, even though the generic le/be wrappers
188 * 1,2,4 or 8 bytes long 185 * use the __get/put_xxx prefixing, they actually wrap in to the
186 * non-prefixed get/put_xxx variants as provided above.
189 */ 187 */
190extern void __bad_unaligned_access_size(void); 188#include <linux/unaligned/generic.h>
191
192#define __get_unaligned_le(ptr) ((__force typeof(*(ptr)))({ \
193 __builtin_choose_expr(sizeof(*(ptr)) == 1, *(ptr), \
194 __builtin_choose_expr(sizeof(*(ptr)) == 2, get_unaligned_le16((ptr)), \
195 __builtin_choose_expr(sizeof(*(ptr)) == 4, get_unaligned_le32((ptr)), \
196 __builtin_choose_expr(sizeof(*(ptr)) == 8, get_unaligned_le64((ptr)), \
197 __bad_unaligned_access_size())))); \
198 }))
199
200#define __get_unaligned_be(ptr) ((__force typeof(*(ptr)))({ \
201 __builtin_choose_expr(sizeof(*(ptr)) == 1, *(ptr), \
202 __builtin_choose_expr(sizeof(*(ptr)) == 2, get_unaligned_be16((ptr)), \
203 __builtin_choose_expr(sizeof(*(ptr)) == 4, get_unaligned_be32((ptr)), \
204 __builtin_choose_expr(sizeof(*(ptr)) == 8, get_unaligned_be64((ptr)), \
205 __bad_unaligned_access_size())))); \
206 }))
207
208#define __put_unaligned_le(val, ptr) ({ \
209 void *__gu_p = (ptr); \
210 switch (sizeof(*(ptr))) { \
211 case 1: \
212 *(u8 *)__gu_p = (__force u8)(val); \
213 break; \
214 case 2: \
215 put_unaligned_le16((__force u16)(val), __gu_p); \
216 break; \
217 case 4: \
218 put_unaligned_le32((__force u32)(val), __gu_p); \
219 break; \
220 case 8: \
221 put_unaligned_le64((__force u64)(val), __gu_p); \
222 break; \
223 default: \
224 __bad_unaligned_access_size(); \
225 break; \
226 } \
227 (void)0; })
228
229#define __put_unaligned_be(val, ptr) ({ \
230 void *__gu_p = (ptr); \
231 switch (sizeof(*(ptr))) { \
232 case 1: \
233 *(u8 *)__gu_p = (__force u8)(val); \
234 break; \
235 case 2: \
236 put_unaligned_be16((__force u16)(val), __gu_p); \
237 break; \
238 case 4: \
239 put_unaligned_be32((__force u32)(val), __gu_p); \
240 break; \
241 case 8: \
242 put_unaligned_be64((__force u64)(val), __gu_p); \
243 break; \
244 default: \
245 __bad_unaligned_access_size(); \
246 break; \
247 } \
248 (void)0; })
249 189
250#ifdef __LITTLE_ENDIAN 190#ifdef __LITTLE_ENDIAN
251# define get_unaligned __get_unaligned_le 191# define get_unaligned __get_unaligned_le