1 /* SPDX-License-Identifier: GPL-2.0+ */
2 /*
3  * Copyright (C) 2013-2014, 2020 Synopsys, Inc. All rights reserved.
4  */
5 
6 #ifndef __ASM_ARC_IO_H
7 #define __ASM_ARC_IO_H
8 
9 #include <linux/types.h>
10 #include <asm/byteorder.h>
11 
12 /*
13  * Compiler barrier. It prevents compiler from reordering instructions before
14  * and after it. It doesn't prevent HW (CPU) from any reordering though.
15  */
16 #define __comp_b()		asm volatile("" : : : "memory")
17 
18 #ifdef __ARCHS__
19 
20 /*
21  * ARCv2 based HS38 cores are in-order issue, but still weakly ordered
22  * due to micro-arch buffering/queuing of load/store, cache hit vs. miss ...
23  *
24  * Explicit barrier provided by DMB instruction
25  *  - Operand supports fine grained load/store/load+store semantics
26  *  - Ensures that selected memory operation issued before it will complete
27  *    before any subsequent memory operation of same type
28  *  - DMB guarantees SMP as well as local barrier semantics
29  *    (asm-generic/barrier.h ensures sane smp_*mb if not defined here, i.e.
30  *    UP: barrier(), SMP: smp_*mb == *mb)
31  *  - DSYNC provides DMB+completion_of_cache_bpu_maintenance_ops hence not needed
32  *    in the general case. Plus it only provides full barrier.
33  */
34 
35 #define mb()	asm volatile("dmb 3\n" : : : "memory")
36 #define rmb()	asm volatile("dmb 1\n" : : : "memory")
37 #define wmb()	asm volatile("dmb 2\n" : : : "memory")
38 
39 #else
40 
41 /*
42  * ARCompact based cores (ARC700) only have SYNC instruction which is super
43  * heavy weight as it flushes the pipeline as well.
44  * There are no real SMP implementations of such cores.
45  */
46 
47 #define mb()	asm volatile("sync\n" : : : "memory")
48 #endif
49 
50 #ifdef __ARCHS__
51 #define __iormb()		rmb()
52 #define __iowmb()		wmb()
53 #else
54 #define __iormb()		__comp_b()
55 #define __iowmb()		__comp_b()
56 #endif
57 
sync(void)58 static inline void sync(void)
59 {
60 	/* Not yet implemented */
61 }
62 
63 /*
64  * We must use 'volatile' in C-version read/write IO accessors implementation
65  * to avoid merging several reads (writes) into one read (write), or optimizing
66  * them out by compiler.
67  * We must use compiler barriers before and after operation (read or write) so
68  * it won't be reordered by compiler.
69  */
70 #define __arch_getb(a)		({ u8  __v; __comp_b(); __v = *(volatile u8  *)(a); __comp_b(); __v; })
71 #define __arch_getw(a)		({ u16 __v; __comp_b(); __v = *(volatile u16 *)(a); __comp_b(); __v; })
72 #define __arch_getl(a)		({ u32 __v; __comp_b(); __v = *(volatile u32 *)(a); __comp_b(); __v; })
73 #define __arch_getq(a)		({ u64 __v; __comp_b(); __v = *(volatile u64 *)(a); __comp_b(); __v; })
74 
75 #define __arch_putb(v, a)	({ __comp_b(); *(volatile u8  *)(a) = (v); __comp_b(); })
76 #define __arch_putw(v, a)	({ __comp_b(); *(volatile u16 *)(a) = (v); __comp_b(); })
77 #define __arch_putl(v, a)	({ __comp_b(); *(volatile u32 *)(a) = (v); __comp_b(); })
78 #define __arch_putq(v, a)	({ __comp_b(); *(volatile u64 *)(a) = (v); __comp_b(); })
79 
80 /*
81  * We add memory barriers for __raw_readX / __raw_writeX accessors same way as
82  * it is done for readX and writeX accessors as lots of U-Boot driver uses
83  * __raw_readX / __raw_writeX instead of proper accessor with barrier.
84  */
85 #define __raw_writeb(v, c)	({ __iowmb(); __arch_putb(v, c); })
86 #define __raw_writew(v, c)	({ __iowmb(); __arch_putw(v, c); })
87 #define __raw_writel(v, c)	({ __iowmb(); __arch_putl(v, c); })
88 #define __raw_writeq(v, c)	({ __iowmb(); __arch_putq(v, c); })
89 
90 #define __raw_readb(c)		({ u8  __v = __arch_getb(c); __iormb(); __v; })
91 #define __raw_readw(c)		({ u16 __v = __arch_getw(c); __iormb(); __v; })
92 #define __raw_readl(c)		({ u32 __v = __arch_getl(c); __iormb(); __v; })
93 #define __raw_readq(c)		({ u64 __v = __arch_getq(c); __iormb(); __v; })
94 
__raw_writesb(unsigned long addr,const void * data,int bytelen)95 static inline void __raw_writesb(unsigned long addr, const void *data,
96 				 int bytelen)
97 {
98 	u8 *buf = (uint8_t *)data;
99 
100 	__iowmb();
101 
102 	while (bytelen--)
103 		__arch_putb(*buf++, addr);
104 }
105 
__raw_writesw(unsigned long addr,const void * data,int wordlen)106 static inline void __raw_writesw(unsigned long addr, const void *data,
107 				 int wordlen)
108 {
109 	u16 *buf = (uint16_t *)data;
110 
111 	__iowmb();
112 
113 	while (wordlen--)
114 		__arch_putw(*buf++, addr);
115 }
116 
__raw_writesl(unsigned long addr,const void * data,int longlen)117 static inline void __raw_writesl(unsigned long addr, const void *data,
118 				 int longlen)
119 {
120 	u32 *buf = (uint32_t *)data;
121 
122 	__iowmb();
123 
124 	while (longlen--)
125 		__arch_putl(*buf++, addr);
126 }
127 
__raw_readsb(unsigned long addr,void * data,int bytelen)128 static inline void __raw_readsb(unsigned long addr, void *data, int bytelen)
129 {
130 	u8 *buf = (uint8_t *)data;
131 
132 	while (bytelen--)
133 		*buf++ = __arch_getb(addr);
134 
135 	__iormb();
136 }
137 
__raw_readsw(unsigned long addr,void * data,int wordlen)138 static inline void __raw_readsw(unsigned long addr, void *data, int wordlen)
139 {
140 	u16 *buf = (uint16_t *)data;
141 
142 	while (wordlen--)
143 		*buf++ = __arch_getw(addr);
144 
145 	__iormb();
146 }
147 
__raw_readsl(unsigned long addr,void * data,int longlen)148 static inline void __raw_readsl(unsigned long addr, void *data, int longlen)
149 {
150 	u32 *buf = (uint32_t *)data;
151 
152 	while (longlen--)
153 		*buf++ = __arch_getl(addr);
154 
155 	__iormb();
156 }
157 
158 /*
159  * Relaxed I/O memory access primitives. These follow the Device memory
160  * ordering rules but do not guarantee any ordering relative to Normal memory
161  * accesses.
162  */
163 #define readb_relaxed(c)	({ u8  __r = __arch_getb(c); __r; })
164 #define readw_relaxed(c)	({ u16 __r = le16_to_cpu((__force __le16)__arch_getw(c)); __r; })
165 #define readl_relaxed(c)	({ u32 __r = le32_to_cpu((__force __le32)__arch_getl(c)); __r; })
166 #define readq_relaxed(c)	({ u64 __r = le64_to_cpu((__force __le64)__arch_getq(c)); __r; })
167 
168 #define writeb_relaxed(v, c)	((void)__arch_putb((v), (c)))
169 #define writew_relaxed(v, c)	((void)__arch_putw((__force u16)cpu_to_le16(v), (c)))
170 #define writel_relaxed(v, c)	((void)__arch_putl((__force u32)cpu_to_le32(v), (c)))
171 #define writeq_relaxed(v, c)	((void)__arch_putq((__force u64)cpu_to_le64(v), (c)))
172 
173 /*
174  * MMIO can also get buffered/optimized in micro-arch, so barriers needed
175  * Based on ARM model for the typical use case
176  *
177  *	<ST [DMA buffer]>
178  *	<writel MMIO "go" reg>
179  *  or:
180  *	<readl MMIO "status" reg>
181  *	<LD [DMA buffer]>
182  *
183  * http://lkml.kernel.org/r/20150622133656.GG1583@arm.com
184  */
185 #define readb(c)	({ u8  __v = readb_relaxed(c); __iormb(); __v; })
186 #define readw(c)	({ u16 __v = readw_relaxed(c); __iormb(); __v; })
187 #define readl(c)	({ u32 __v = readl_relaxed(c); __iormb(); __v; })
188 #define readq(c)	({ u64 __v = readq_relaxed(c); __iormb(); __v; })
189 
190 #define writeb(v, c)	({ __iowmb(); writeb_relaxed(v, c); })
191 #define writew(v, c)	({ __iowmb(); writew_relaxed(v, c); })
192 #define writel(v, c)	({ __iowmb(); writel_relaxed(v, c); })
193 #define writeq(v, c)	({ __iowmb(); writeq_relaxed(v, c); })
194 
195 #define out_arch(type, endian, a, v)	__raw_write##type(cpu_to_##endian(v), a)
196 #define in_arch(type, endian, a)	endian##_to_cpu(__raw_read##type(a))
197 
198 #define out_le32(a, v)	out_arch(l, le32, a, v)
199 #define out_le16(a, v)	out_arch(w, le16, a, v)
200 
201 #define in_le32(a)	in_arch(l, le32, a)
202 #define in_le16(a)	in_arch(w, le16, a)
203 
204 #define out_be32(a, v)	out_arch(l, be32, a, v)
205 #define out_be16(a, v)	out_arch(w, be16, a, v)
206 
207 #define in_be32(a)	in_arch(l, be32, a)
208 #define in_be16(a)	in_arch(w, be16, a)
209 
210 #define out_8(a, v)	__raw_writeb(v, a)
211 #define in_8(a)		__raw_readb(a)
212 
213 /*
214  * Clear and set bits in one shot. These macros can be used to clear and
215  * set multiple bits in a register using a single call. These macros can
216  * also be used to set a multiple-bit bit pattern using a mask, by
217  * specifying the mask in the 'clear' parameter and the new bit pattern
218  * in the 'set' parameter.
219  */
220 
221 #define clrbits(type, addr, clear) \
222 	out_##type((addr), in_##type(addr) & ~(clear))
223 
224 #define setbits(type, addr, set) \
225 	out_##type((addr), in_##type(addr) | (set))
226 
227 #define clrsetbits(type, addr, clear, set) \
228 	out_##type((addr), (in_##type(addr) & ~(clear)) | (set))
229 
230 #define clrbits_be32(addr, clear) clrbits(be32, addr, clear)
231 #define setbits_be32(addr, set) setbits(be32, addr, set)
232 #define clrsetbits_be32(addr, clear, set) clrsetbits(be32, addr, clear, set)
233 
234 #define clrbits_le32(addr, clear) clrbits(le32, addr, clear)
235 #define setbits_le32(addr, set) setbits(le32, addr, set)
236 #define clrsetbits_le32(addr, clear, set) clrsetbits(le32, addr, clear, set)
237 
238 #define clrbits_be16(addr, clear) clrbits(be16, addr, clear)
239 #define setbits_be16(addr, set) setbits(be16, addr, set)
240 #define clrsetbits_be16(addr, clear, set) clrsetbits(be16, addr, clear, set)
241 
242 #define clrbits_le16(addr, clear) clrbits(le16, addr, clear)
243 #define setbits_le16(addr, set) setbits(le16, addr, set)
244 #define clrsetbits_le16(addr, clear, set) clrsetbits(le16, addr, clear, set)
245 
246 #define clrbits_8(addr, clear) clrbits(8, addr, clear)
247 #define setbits_8(addr, set) setbits(8, addr, set)
248 #define clrsetbits_8(addr, clear, set) clrsetbits(8, addr, clear, set)
249 
250 #include <asm-generic/io.h>
251 
252 #endif	/* __ASM_ARC_IO_H */
253