Commit | Line | Data |
---|---|---|
68648ed1 AD |
1 | /* |
2 | * Generic Bit Block Transfer for frame buffers located in system RAM with | |
3 | * packed pixels of any depth. | |
4 | * | |
5 | * Based almost entirely from cfbcopyarea.c (which is based almost entirely | |
6 | * on Geert Uytterhoeven's copyarea routine) | |
7 | * | |
8 | * Copyright (C) 2007 Antonino Daplas <adaplas@pol.net> | |
9 | * | |
10 | * This file is subject to the terms and conditions of the GNU General Public | |
11 | * License. See the file COPYING in the main directory of this archive for | |
12 | * more details. | |
13 | * | |
14 | */ | |
15 | #include <linux/module.h> | |
16 | #include <linux/kernel.h> | |
17 | #include <linux/string.h> | |
18 | #include <linux/fb.h> | |
68648ed1 AD |
19 | #include <asm/types.h> |
20 | #include <asm/io.h> | |
dc0e6e05 | 21 | #include "fb_draw.h" |
68648ed1 AD |
22 | |
23 | /* | |
24 | * Generic bitwise copy algorithm | |
25 | */ | |
26 | ||
27 | static void | |
e4c690e0 AV |
28 | bitcpy(struct fb_info *p, unsigned long *dst, int dst_idx, |
29 | const unsigned long *src, int src_idx, int bits, unsigned n) | |
68648ed1 AD |
30 | { |
31 | unsigned long first, last; | |
32 | int const shift = dst_idx-src_idx; | |
33 | int left, right; | |
34 | ||
e4c690e0 AV |
35 | first = FB_SHIFT_HIGH(p, ~0UL, dst_idx); |
36 | last = ~(FB_SHIFT_HIGH(p, ~0UL, (dst_idx+n) % bits)); | |
68648ed1 AD |
37 | |
38 | if (!shift) { | |
39 | /* Same alignment for source and dest */ | |
40 | if (dst_idx+n <= bits) { | |
41 | /* Single word */ | |
42 | if (last) | |
43 | first &= last; | |
44 | *dst = comp(*src, *dst, first); | |
45 | } else { | |
46 | /* Multiple destination words */ | |
47 | /* Leading bits */ | |
48 | if (first != ~0UL) { | |
49 | *dst = comp(*src, *dst, first); | |
50 | dst++; | |
51 | src++; | |
52 | n -= bits - dst_idx; | |
53 | } | |
54 | ||
55 | /* Main chunk */ | |
56 | n /= bits; | |
57 | while (n >= 8) { | |
58 | *dst++ = *src++; | |
59 | *dst++ = *src++; | |
60 | *dst++ = *src++; | |
61 | *dst++ = *src++; | |
62 | *dst++ = *src++; | |
63 | *dst++ = *src++; | |
64 | *dst++ = *src++; | |
65 | *dst++ = *src++; | |
66 | n -= 8; | |
67 | } | |
68 | while (n--) | |
69 | *dst++ = *src++; | |
70 | ||
71 | /* Trailing bits */ | |
72 | if (last) | |
73 | *dst = comp(*src, *dst, last); | |
74 | } | |
75 | } else { | |
76 | unsigned long d0, d1; | |
77 | int m; | |
78 | ||
79 | /* Different alignment for source and dest */ | |
80 | right = shift & (bits - 1); | |
81 | left = -shift & (bits - 1); | |
82 | ||
83 | if (dst_idx+n <= bits) { | |
84 | /* Single destination word */ | |
85 | if (last) | |
86 | first &= last; | |
87 | if (shift > 0) { | |
88 | /* Single source word */ | |
89 | *dst = comp(*src >> right, *dst, first); | |
90 | } else if (src_idx+n <= bits) { | |
91 | /* Single source word */ | |
92 | *dst = comp(*src << left, *dst, first); | |
93 | } else { | |
94 | /* 2 source words */ | |
95 | d0 = *src++; | |
96 | d1 = *src; | |
97 | *dst = comp(d0 << left | d1 >> right, *dst, | |
98 | first); | |
99 | } | |
100 | } else { | |
101 | /* Multiple destination words */ | |
102 | /** We must always remember the last value read, | |
103 | because in case SRC and DST overlap bitwise (e.g. | |
104 | when moving just one pixel in 1bpp), we always | |
105 | collect one full long for DST and that might | |
106 | overlap with the current long from SRC. We store | |
107 | this value in 'd0'. */ | |
108 | d0 = *src++; | |
109 | /* Leading bits */ | |
110 | if (shift > 0) { | |
111 | /* Single source word */ | |
112 | *dst = comp(d0 >> right, *dst, first); | |
113 | dst++; | |
114 | n -= bits - dst_idx; | |
115 | } else { | |
116 | /* 2 source words */ | |
117 | d1 = *src++; | |
118 | *dst = comp(d0 << left | *dst >> right, *dst, first); | |
119 | d0 = d1; | |
120 | dst++; | |
121 | n -= bits - dst_idx; | |
122 | } | |
123 | ||
124 | /* Main chunk */ | |
125 | m = n % bits; | |
126 | n /= bits; | |
127 | while (n >= 4) { | |
128 | d1 = *src++; | |
129 | *dst++ = d0 << left | d1 >> right; | |
130 | d0 = d1; | |
131 | d1 = *src++; | |
132 | *dst++ = d0 << left | d1 >> right; | |
133 | d0 = d1; | |
134 | d1 = *src++; | |
135 | *dst++ = d0 << left | d1 >> right; | |
136 | d0 = d1; | |
137 | d1 = *src++; | |
138 | *dst++ = d0 << left | d1 >> right; | |
139 | d0 = d1; | |
140 | n -= 4; | |
141 | } | |
142 | while (n--) { | |
143 | d1 = *src++; | |
144 | *dst++ = d0 << left | d1 >> right; | |
145 | d0 = d1; | |
146 | } | |
147 | ||
148 | /* Trailing bits */ | |
149 | if (last) { | |
150 | if (m <= right) { | |
151 | /* Single source word */ | |
152 | *dst = comp(d0 << left, *dst, last); | |
153 | } else { | |
154 | /* 2 source words */ | |
155 | d1 = *src; | |
156 | *dst = comp(d0 << left | d1 >> right, | |
157 | *dst, last); | |
158 | } | |
159 | } | |
160 | } | |
161 | } | |
162 | } | |
163 | ||
164 | /* | |
165 | * Generic bitwise copy algorithm, operating backward | |
166 | */ | |
167 | ||
168 | static void | |
e4c690e0 AV |
169 | bitcpy_rev(struct fb_info *p, unsigned long *dst, int dst_idx, |
170 | const unsigned long *src, int src_idx, int bits, unsigned n) | |
68648ed1 AD |
171 | { |
172 | unsigned long first, last; | |
173 | int shift; | |
174 | ||
175 | dst += (n-1)/bits; | |
176 | src += (n-1)/bits; | |
177 | if ((n-1) % bits) { | |
178 | dst_idx += (n-1) % bits; | |
179 | dst += dst_idx >> (ffs(bits) - 1); | |
180 | dst_idx &= bits - 1; | |
181 | src_idx += (n-1) % bits; | |
182 | src += src_idx >> (ffs(bits) - 1); | |
183 | src_idx &= bits - 1; | |
184 | } | |
185 | ||
186 | shift = dst_idx-src_idx; | |
187 | ||
e4c690e0 AV |
188 | first = FB_SHIFT_LOW(p, ~0UL, bits - 1 - dst_idx); |
189 | last = ~(FB_SHIFT_LOW(p, ~0UL, bits - 1 - ((dst_idx-n) % bits))); | |
68648ed1 AD |
190 | |
191 | if (!shift) { | |
192 | /* Same alignment for source and dest */ | |
193 | if ((unsigned long)dst_idx+1 >= n) { | |
194 | /* Single word */ | |
195 | if (last) | |
196 | first &= last; | |
197 | *dst = comp(*src, *dst, first); | |
198 | } else { | |
199 | /* Multiple destination words */ | |
200 | ||
201 | /* Leading bits */ | |
202 | if (first != ~0UL) { | |
203 | *dst = comp(*src, *dst, first); | |
204 | dst--; | |
205 | src--; | |
206 | n -= dst_idx+1; | |
207 | } | |
208 | ||
209 | /* Main chunk */ | |
210 | n /= bits; | |
211 | while (n >= 8) { | |
212 | *dst-- = *src--; | |
213 | *dst-- = *src--; | |
214 | *dst-- = *src--; | |
215 | *dst-- = *src--; | |
216 | *dst-- = *src--; | |
217 | *dst-- = *src--; | |
218 | *dst-- = *src--; | |
219 | *dst-- = *src--; | |
220 | n -= 8; | |
221 | } | |
222 | while (n--) | |
223 | *dst-- = *src--; | |
224 | /* Trailing bits */ | |
225 | if (last) | |
226 | *dst = comp(*src, *dst, last); | |
227 | } | |
228 | } else { | |
229 | /* Different alignment for source and dest */ | |
230 | ||
231 | int const left = -shift & (bits-1); | |
232 | int const right = shift & (bits-1); | |
233 | ||
234 | if ((unsigned long)dst_idx+1 >= n) { | |
235 | /* Single destination word */ | |
236 | if (last) | |
237 | first &= last; | |
238 | if (shift < 0) { | |
239 | /* Single source word */ | |
240 | *dst = comp(*src << left, *dst, first); | |
241 | } else if (1+(unsigned long)src_idx >= n) { | |
242 | /* Single source word */ | |
243 | *dst = comp(*src >> right, *dst, first); | |
244 | } else { | |
245 | /* 2 source words */ | |
246 | *dst = comp(*src >> right | *(src-1) << left, | |
247 | *dst, first); | |
248 | } | |
249 | } else { | |
250 | /* Multiple destination words */ | |
251 | /** We must always remember the last value read, | |
252 | because in case SRC and DST overlap bitwise (e.g. | |
253 | when moving just one pixel in 1bpp), we always | |
254 | collect one full long for DST and that might | |
255 | overlap with the current long from SRC. We store | |
256 | this value in 'd0'. */ | |
257 | unsigned long d0, d1; | |
258 | int m; | |
259 | ||
260 | d0 = *src--; | |
261 | /* Leading bits */ | |
262 | if (shift < 0) { | |
263 | /* Single source word */ | |
264 | *dst = comp(d0 << left, *dst, first); | |
265 | } else { | |
266 | /* 2 source words */ | |
267 | d1 = *src--; | |
268 | *dst = comp(d0 >> right | d1 << left, *dst, | |
269 | first); | |
270 | d0 = d1; | |
271 | } | |
272 | dst--; | |
273 | n -= dst_idx+1; | |
274 | ||
275 | /* Main chunk */ | |
276 | m = n % bits; | |
277 | n /= bits; | |
278 | while (n >= 4) { | |
279 | d1 = *src--; | |
280 | *dst-- = d0 >> right | d1 << left; | |
281 | d0 = d1; | |
282 | d1 = *src--; | |
283 | *dst-- = d0 >> right | d1 << left; | |
284 | d0 = d1; | |
285 | d1 = *src--; | |
286 | *dst-- = d0 >> right | d1 << left; | |
287 | d0 = d1; | |
288 | d1 = *src--; | |
289 | *dst-- = d0 >> right | d1 << left; | |
290 | d0 = d1; | |
291 | n -= 4; | |
292 | } | |
293 | while (n--) { | |
294 | d1 = *src--; | |
295 | *dst-- = d0 >> right | d1 << left; | |
296 | d0 = d1; | |
297 | } | |
298 | ||
299 | /* Trailing bits */ | |
300 | if (last) { | |
301 | if (m <= left) { | |
302 | /* Single source word */ | |
303 | *dst = comp(d0 >> right, *dst, last); | |
304 | } else { | |
305 | /* 2 source words */ | |
306 | d1 = *src; | |
307 | *dst = comp(d0 >> right | d1 << left, | |
308 | *dst, last); | |
309 | } | |
310 | } | |
311 | } | |
312 | } | |
313 | } | |
314 | ||
315 | void sys_copyarea(struct fb_info *p, const struct fb_copyarea *area) | |
316 | { | |
317 | u32 dx = area->dx, dy = area->dy, sx = area->sx, sy = area->sy; | |
318 | u32 height = area->height, width = area->width; | |
319 | unsigned long const bits_per_line = p->fix.line_length*8u; | |
320 | unsigned long *dst = NULL, *src = NULL; | |
321 | int bits = BITS_PER_LONG, bytes = bits >> 3; | |
322 | int dst_idx = 0, src_idx = 0, rev_copy = 0; | |
323 | ||
324 | if (p->state != FBINFO_STATE_RUNNING) | |
325 | return; | |
326 | ||
327 | /* if the beginning of the target area might overlap with the end of | |
328 | the source area, be have to copy the area reverse. */ | |
329 | if ((dy == sy && dx > sx) || (dy > sy)) { | |
330 | dy += height; | |
331 | sy += height; | |
332 | rev_copy = 1; | |
333 | } | |
334 | ||
335 | /* split the base of the framebuffer into a long-aligned address and | |
336 | the index of the first bit */ | |
337 | dst = src = (unsigned long *)((unsigned long)p->screen_base & | |
338 | ~(bytes-1)); | |
339 | dst_idx = src_idx = 8*((unsigned long)p->screen_base & (bytes-1)); | |
340 | /* add offset of source and target area */ | |
341 | dst_idx += dy*bits_per_line + dx*p->var.bits_per_pixel; | |
342 | src_idx += sy*bits_per_line + sx*p->var.bits_per_pixel; | |
343 | ||
344 | if (p->fbops->fb_sync) | |
345 | p->fbops->fb_sync(p); | |
346 | ||
347 | if (rev_copy) { | |
348 | while (height--) { | |
349 | dst_idx -= bits_per_line; | |
350 | src_idx -= bits_per_line; | |
351 | dst += dst_idx >> (ffs(bits) - 1); | |
352 | dst_idx &= (bytes - 1); | |
353 | src += src_idx >> (ffs(bits) - 1); | |
354 | src_idx &= (bytes - 1); | |
e4c690e0 | 355 | bitcpy_rev(p, dst, dst_idx, src, src_idx, bits, |
68648ed1 AD |
356 | width*p->var.bits_per_pixel); |
357 | } | |
358 | } else { | |
359 | while (height--) { | |
360 | dst += dst_idx >> (ffs(bits) - 1); | |
361 | dst_idx &= (bytes - 1); | |
362 | src += src_idx >> (ffs(bits) - 1); | |
363 | src_idx &= (bytes - 1); | |
e4c690e0 | 364 | bitcpy(p, dst, dst_idx, src, src_idx, bits, |
68648ed1 AD |
365 | width*p->var.bits_per_pixel); |
366 | dst_idx += bits_per_line; | |
367 | src_idx += bits_per_line; | |
368 | } | |
369 | } | |
370 | } | |
371 | ||
372 | EXPORT_SYMBOL(sys_copyarea); | |
373 | ||
374 | MODULE_AUTHOR("Antonino Daplas <adaplas@pol.net>"); | |
375 | MODULE_DESCRIPTION("Generic copyarea (sys-to-sys)"); | |
376 | MODULE_LICENSE("GPL"); | |
377 |