1 | /**************************************************************************************** |
---|
2 | File : drivers.c |
---|
3 | Written by Alain Greiner & Nicolas Pouillon |
---|
4 | Date : december 2010 |
---|
5 | |
---|
6 | Basic drivers used by the GIET, that is running |
---|
7 | on the MIPS32 processor architecture. |
---|
8 | |
---|
9 | The supported peripherals are: |
---|
10 | - the SoClib pibus_multi_tty |
---|
11 | - the SocLib pibus_timer |
---|
12 | - the SocLib pibus_dma |
---|
13 | - The SoCLib pibus_icu |
---|
14 | - The SoCLib pibus_gcd |
---|
15 | - The SoCLib pibus_frame_buffer |
---|
16 | - The SoCLib pibus_block_device |
---|
17 | |
---|
18 | The following global parameters must be defined in the ldscript. |
---|
19 | - NB_CLUSTERS : number of clusters |
---|
20 | - NB_PROCS : number of processor per cluster |
---|
21 | - NB_NTASKS : max number of tasks per processor |
---|
22 | - NB_LOCKS : max number of supported spin_locks |
---|
23 | - NB_TIMERS : max number of timers per processor |
---|
24 | |
---|
25 | The follobing base addresses must be defined in the ldscript |
---|
26 | - seg_icu_base |
---|
27 | - seg_timer_base |
---|
28 | - seg_tty_base |
---|
29 | - seg_gcd_base |
---|
30 | - seg_dma_base |
---|
31 | - seg_locks_base |
---|
32 | - seg_fb_base |
---|
33 | - seg_ioc_base |
---|
34 | ****************************************************************************************/ |
---|
35 | |
---|
36 | #include "drivers.h" |
---|
37 | #include "icu.h" |
---|
38 | #include "block_device.h" |
---|
39 | #include "dma.h" |
---|
40 | |
---|
41 | struct plouf; |
---|
42 | |
---|
43 | ////////////////////////////////////////////////////////////// |
---|
44 | // various informations that must be defined in ldscript |
---|
45 | ////////////////////////////////////////////////////////////// |
---|
46 | extern struct plouf seg_icu_base; |
---|
47 | extern struct plouf seg_timer_base; |
---|
48 | extern struct plouf seg_tty_base; |
---|
49 | extern struct plouf seg_gcd_base; |
---|
50 | extern struct plouf seg_dma_base; |
---|
51 | extern struct plouf seg_locks_base; |
---|
52 | extern struct plouf seg_fb_base; |
---|
53 | extern struct plouf seg_ioc_base; |
---|
54 | |
---|
55 | extern struct plouf NB_CLUSTERS; |
---|
56 | extern struct plouf NB_PROCS; |
---|
57 | extern struct plouf NB_TASKS; |
---|
58 | extern struct plouf NB_TIMERS; |
---|
59 | extern struct plouf NB_LOCKS; |
---|
60 | |
---|
61 | #define in_drivers __attribute__((section (".drivers"))) |
---|
62 | #define in_unckdata __attribute__((section (".unckdata"))) |
---|
63 | |
---|
64 | //////////////////////////////////////////////////////////////////////////////////////// |
---|
65 | // Global uncachable variables for synchronization between drivers and ISRs |
---|
66 | //////////////////////////////////////////////////////////////////////////////////////// |
---|
67 | |
---|
68 | in_unckdata int volatile _dma_status[256]; |
---|
69 | in_unckdata int volatile _dma_busy[256] = { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
70 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
71 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
72 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
73 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
74 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
75 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
76 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
77 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
78 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
79 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
80 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
81 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
82 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
83 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
84 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0 }; |
---|
85 | |
---|
86 | in_unckdata int volatile _ioc_lock = 0; |
---|
87 | in_unckdata int volatile _ioc_done = 0; |
---|
88 | in_unckdata int volatile _ioc_status; |
---|
89 | |
---|
90 | in_unckdata char volatile _tty_get_buf[256]; |
---|
91 | in_unckdata int volatile _tty_get_full[256] = { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
92 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
93 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
94 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
95 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
96 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
97 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
98 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
99 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
100 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
101 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
102 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
103 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
104 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
105 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
106 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0 }; |
---|
107 | |
---|
108 | in_unckdata char volatile _tty_put_buf[256]; |
---|
109 | in_unckdata int volatile _tty_put_full[256] = { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
110 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
111 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
112 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
113 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
114 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
115 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
116 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
117 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
118 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
119 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
120 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
121 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
122 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
123 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
124 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0 }; |
---|
125 | |
---|
126 | //////////////////////////////////////////////////////////////////////////////////////// |
---|
127 | // Global uncachable variables for inter-task barriers |
---|
128 | //////////////////////////////////////////////////////////////////////////////////////// |
---|
129 | |
---|
130 | in_unckdata int volatile _barrier_initial_value[16] = { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0 }; |
---|
131 | in_unckdata int volatile _barrier_count[16] = { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0 }; |
---|
132 | in_unckdata int volatile _barrier_lock[16] = { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0 }; |
---|
133 | |
---|
134 | //////////////////////////////////////////////////////////////////////////////////////// |
---|
135 | // Global uncachable variables for spin_locks using LL/C instructions |
---|
136 | //////////////////////////////////////////////////////////////////////////////////////// |
---|
137 | |
---|
138 | in_unckdata int volatile _spin_lock[256] = { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
139 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
140 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
141 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
142 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
143 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
144 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
145 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
146 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
147 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
148 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
149 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
150 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
151 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
152 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, |
---|
153 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0 }; |
---|
154 | |
---|
155 | //////////////////////////////////////////////////////////////////////////////////////// |
---|
156 | // mempcy() |
---|
157 | // GCC requires this function. Taken from MutekH. |
---|
158 | //////////////////////////////////////////////////////////////////////////////////////// |
---|
159 | __attribute((used)) |
---|
160 | in_drivers static void *memcpy(void *_dst, const void *_src, unsigned int size) |
---|
161 | { |
---|
162 | unsigned int *dst = _dst; |
---|
163 | const unsigned int *src = _src; |
---|
164 | if ( ! ((unsigned int)dst & 3) && ! ((unsigned int)src & 3) ) |
---|
165 | while (size > 3) { |
---|
166 | *dst++ = *src++; |
---|
167 | size -= 4; |
---|
168 | } |
---|
169 | |
---|
170 | unsigned char *cdst = (unsigned char*)dst; |
---|
171 | unsigned char *csrc = (unsigned char*)src; |
---|
172 | |
---|
173 | while (size--) { |
---|
174 | *cdst++ = *csrc++; |
---|
175 | } |
---|
176 | return _dst; |
---|
177 | } |
---|
178 | |
---|
179 | //////////////////////////////////////////////////////////////////////////////////////// |
---|
180 | // _procid() |
---|
181 | // Access CP0 and returns processor ident |
---|
182 | // No more than 1024 processors... |
---|
183 | //////////////////////////////////////////////////////////////////////////////////////// |
---|
184 | in_drivers unsigned int _procid() |
---|
185 | { |
---|
186 | unsigned int ret; |
---|
187 | asm volatile( "mfc0 %0, $15, 1": "=r"(ret) ); |
---|
188 | return (ret & 0x3FF); |
---|
189 | } |
---|
190 | //////////////////////////////////////////////////////////////////////////////////////// |
---|
191 | // _segment_increment() |
---|
192 | // Access CP0 to get the procid, and returns the address increment to access |
---|
193 | // various peripherals (TTY, TIMER, ICU, DMA), in case of multiprocessors architectures. |
---|
194 | // It uses the NB_PROCS and NB_CLUSTERS parameters to compute this increment: |
---|
195 | // - increment = cluster_id*cluster_increment + local_id*local_increment |
---|
196 | // - cluster_id = procid / NB_PROCS |
---|
197 | // - local_id = procid % NB_PROCS |
---|
198 | // - cluster_increment = 4G / NB_CLUSTERS |
---|
199 | //////////////////////////////////////////////////////////////////////////////////////// |
---|
200 | in_drivers unsigned int _segment_increment(unsigned int local_increment) |
---|
201 | { |
---|
202 | unsigned int nprocs = (unsigned int)&NB_PROCS; |
---|
203 | unsigned int nclusters = (unsigned int)&NB_CLUSTERS; |
---|
204 | unsigned int cluster_increment = (0x80000000/nclusters)*2; |
---|
205 | unsigned int pid = _procid(); |
---|
206 | return (pid / nprocs)*cluster_increment + (pid % nprocs)*local_increment; |
---|
207 | } |
---|
208 | //////////////////////////////////////////////////////////////////////////////////////// |
---|
209 | // _proctime() |
---|
210 | // Access CP0 and returns processor time |
---|
211 | //////////////////////////////////////////////////////////////////////////////////////// |
---|
212 | in_drivers unsigned int _proctime() |
---|
213 | { |
---|
214 | unsigned int ret; |
---|
215 | asm volatile( "mfc0 %0, $9": "=r"(ret) ); |
---|
216 | return ret; |
---|
217 | } |
---|
218 | //////////////////////////////////////////////////////////////////////////////////////// |
---|
219 | // _procnumber() |
---|
220 | // Returns the number of processsors controled by the GIET |
---|
221 | //////////////////////////////////////////////////////////////////////////////////////// |
---|
222 | in_drivers unsigned int _procnumber() |
---|
223 | { |
---|
224 | return (unsigned int)&NB_PROCS * (unsigned int)&NB_CLUSTERS; |
---|
225 | } |
---|
226 | //////////////////////////////////////////////////////////////////////////////////////// |
---|
227 | // _it_mask() |
---|
228 | // Access CP0 and mask IRQs |
---|
229 | //////////////////////////////////////////////////////////////////////////////////////// |
---|
230 | in_drivers void _it_mask() |
---|
231 | { |
---|
232 | int tmp; |
---|
233 | asm volatile("mfc0 %0, $12" : "=r" (tmp) ); |
---|
234 | asm volatile("ori %0, %0, 1" : "=r" (tmp) ); |
---|
235 | asm volatile("mtc0 %0, $12" : "=r" (tmp) ); |
---|
236 | } |
---|
237 | //////////////////////////////////////////////////////////////////////////////////////// |
---|
238 | // _it_enable() |
---|
239 | // Access CP0 and enable IRQs |
---|
240 | //////////////////////////////////////////////////////////////////////////////////////// |
---|
241 | in_drivers void _it_enable() |
---|
242 | { |
---|
243 | int tmp; |
---|
244 | asm volatile("mfc0 %0, $12" : "=r" (tmp) ); |
---|
245 | asm volatile("addi %0, %0, -1" : "=r" (tmp) ); |
---|
246 | asm volatile("mtc0 %0, $12" : "=r" (tmp) ); |
---|
247 | } |
---|
248 | ////////////////////////////////////////////////////////////////////// |
---|
249 | // _dcache_buf_invalidate() |
---|
250 | // Invalidate all cache lines corresponding to a memory buffer. |
---|
251 | // This is used by the block_device driver. |
---|
252 | ///////////////////////////////////////////////////////////////////////// |
---|
253 | in_drivers void _dcache_buf_invalidate(const void * buffer, size_t size) |
---|
254 | { |
---|
255 | size_t i; |
---|
256 | size_t dcache_line_size; |
---|
257 | |
---|
258 | // retrieve dcache line size from config register (bits 12:10) |
---|
259 | asm volatile("mfc0 %0, $16, 1" : "=r" (dcache_line_size)); |
---|
260 | |
---|
261 | dcache_line_size = 2 << ((dcache_line_size>>10) & 0x7); |
---|
262 | |
---|
263 | // iterate on lines to invalidate each one of them |
---|
264 | for ( i=0; i<size; i+=dcache_line_size ) |
---|
265 | asm volatile(" cache %0, %1" |
---|
266 | : |
---|
267 | :"i" (0x11), "R" (*((char*)buffer+i))); |
---|
268 | } |
---|
269 | |
---|
270 | ///////////////////////////////////////////////////////////////////////// |
---|
271 | // _itoa_dec() |
---|
272 | // convert a 32 bits unsigned int to a string of 10 decimal characters. |
---|
273 | ///////////////////////////////////////////////////////////////////////// |
---|
274 | in_drivers void _itoa_dec(unsigned val, char* buf) |
---|
275 | { |
---|
276 | const char DecTab[] = "0123456789"; |
---|
277 | unsigned int i; |
---|
278 | for( i=0 ; i<10 ; i++ ) |
---|
279 | { |
---|
280 | if( (val!=0) || (i==0) ) buf[9-i] = DecTab[val % 10]; |
---|
281 | else buf[9-i] = 0x20; |
---|
282 | val /= 10; |
---|
283 | } |
---|
284 | } |
---|
285 | ////////////////////////////////////////////////////////////////////////// |
---|
286 | // _itoa_hex() |
---|
287 | // convert a 32 bits unsigned int to a string of 8 hexadecimal characters. |
---|
288 | /////////////////////////////////////////////////////////////////////////// |
---|
289 | in_drivers void _itoa_hex(int val, char* buf) |
---|
290 | { |
---|
291 | const char HexaTab[] = "0123456789ABCD"; |
---|
292 | unsigned int i; |
---|
293 | for( i=0 ; i<8 ; i++ ) |
---|
294 | { |
---|
295 | buf[7-i] = HexaTab[val % 16]; |
---|
296 | val /= 16; |
---|
297 | } |
---|
298 | } |
---|
299 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
300 | // MULTI_TIMER component |
---|
301 | // Each processor can handle up to NB_TIMERS independant timers. |
---|
302 | // The segment base address is defined as |
---|
303 | // seg_timer_base + segment_increment(NB_TIMERS*16) + index*16 |
---|
304 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
305 | // _timer_write() |
---|
306 | // Write a 32 bits word in a memory mapped register of the MULTI_TIMER |
---|
307 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
308 | in_drivers int _timer_write(size_t timer_index, size_t register_index, int value) |
---|
309 | { |
---|
310 | int* timer_address; |
---|
311 | size_t ntimers = (size_t)&NB_TIMERS; |
---|
312 | unsigned int base = (unsigned int)&seg_timer_base; |
---|
313 | unsigned int increment = _segment_increment(ntimers*TIMER_SPAN*4); |
---|
314 | |
---|
315 | if( timer_index >= ntimers) return -1; |
---|
316 | if( register_index >= TIMER_SPAN ) return -1; |
---|
317 | |
---|
318 | timer_address = (int*)(base + increment + timer_index*TIMER_SPAN*4); |
---|
319 | timer_address[register_index] = value; // write word |
---|
320 | return 0; |
---|
321 | } |
---|
322 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
323 | // _timer_read() |
---|
324 | // Read a 32 bits word in a memory mapped register of the MULTI_TIMER |
---|
325 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
326 | in_drivers int _timer_read(size_t timer_index, size_t register_index, int* buffer) |
---|
327 | { |
---|
328 | int* timer_address; |
---|
329 | size_t ntimers = (size_t)&NB_TIMERS; |
---|
330 | unsigned int base = (unsigned int)&seg_timer_base; |
---|
331 | unsigned int increment = _segment_increment(ntimers*TIMER_SPAN*4); |
---|
332 | |
---|
333 | if( timer_index >= ntimers) return -1; |
---|
334 | if( register_index >= TIMER_SPAN ) return -1; |
---|
335 | |
---|
336 | if( timer_index >= ntimers) return -1; |
---|
337 | if( register_index >= TIMER_SPAN ) return -1; |
---|
338 | |
---|
339 | timer_address = (int*)(base + increment + timer_index*TIMER_SPAN*4); |
---|
340 | *buffer = timer_address[register_index]; // read word |
---|
341 | return 0; |
---|
342 | } |
---|
343 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
344 | // MULTI_TTY COMPONENT |
---|
345 | // The total number of TTYs is equal to NB_CLUSTERS * NB_PROCS * NB_TASKS. |
---|
346 | // - tty_address = seg_tty_base + _segment_increment(NB_TASKS*16) + task_id*16 |
---|
347 | // - tty_index = proc_id*NB_TASKS + task_id |
---|
348 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
349 | // _tty_write() |
---|
350 | // Write one or several characters directly from a fixed length user buffer |
---|
351 | // to the TTY_WRITE register of the TTY controler. |
---|
352 | // It doesn't use the TTY_PUT_IRQ interrupt and the associated kernel buffer. |
---|
353 | // This is a non blocking call : it test the TTY_STATUS register. |
---|
354 | // If the TTY_STATUS_WRITE bit is set, the transfer stops and the function |
---|
355 | // returns the number of characters that have been actually written. |
---|
356 | // It returns -1 in case of error (proc_id or task index too large) |
---|
357 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
358 | in_drivers int _tty_write(char* buffer, int length) |
---|
359 | { |
---|
360 | char* tty_address; |
---|
361 | size_t ntasks = (size_t)&NB_TASKS; |
---|
362 | size_t nprocs = (size_t)&NB_PROCS; |
---|
363 | size_t nclusters = (size_t)&NB_CLUSTERS; |
---|
364 | unsigned int base = (unsigned int)&seg_tty_base; |
---|
365 | unsigned int increment = _segment_increment(ntasks*TTY_SPAN*4); |
---|
366 | size_t pid = _procid(); |
---|
367 | int nwritten = 0; |
---|
368 | size_t tid; |
---|
369 | int i; |
---|
370 | |
---|
371 | if( ntasks == 0 ) tid = 0; |
---|
372 | else tid = _current_task_array[pid]; |
---|
373 | |
---|
374 | if( tid >= ntasks ) return -1; |
---|
375 | if( pid >= nprocs*nclusters ) return -1; |
---|
376 | |
---|
377 | tty_address = (char*)(base + increment + tid*TTY_SPAN*4); |
---|
378 | //tty_address = (char*)(base + tid*TTY_SPAN*4); |
---|
379 | |
---|
380 | for ( i=0 ; i < length ; i++ ) |
---|
381 | { |
---|
382 | if((tty_address[TTY_STATUS*4] & 0x2) == 0x2) break; |
---|
383 | else |
---|
384 | { |
---|
385 | tty_address[TTY_WRITE*4] = buffer[i]; // write character |
---|
386 | nwritten++; |
---|
387 | } |
---|
388 | } |
---|
389 | return nwritten; |
---|
390 | } |
---|
391 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
392 | // _tty_read() |
---|
393 | // Fetch one character directly from the TTY_READ register of the TTY controler, |
---|
394 | // and writes this character to the user buffer. |
---|
395 | // It doesn't use the TTY_GET_IRQ interrupt and the associated kernel buffer. |
---|
396 | // This is a non blocking call : it returns 0 if the register is empty, |
---|
397 | // and returns 1 if the register is full. |
---|
398 | // It returns -1 in case of error (proc_id or task_id too large or length != 1) |
---|
399 | // The length argument is not used in this implementation, and has been |
---|
400 | // introduced for future implementations. |
---|
401 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
402 | in_drivers int _tty_read(char* buffer, int length) |
---|
403 | { |
---|
404 | char* tty_address; |
---|
405 | size_t ntasks = (size_t)&NB_TASKS; |
---|
406 | size_t nprocs = (size_t)&NB_PROCS; |
---|
407 | size_t nclusters = (size_t)&NB_CLUSTERS; |
---|
408 | unsigned int base = (unsigned int)&seg_tty_base; |
---|
409 | unsigned int increment = _segment_increment(ntasks*TTY_SPAN*4); |
---|
410 | size_t pid = _procid(); |
---|
411 | size_t tid; |
---|
412 | |
---|
413 | if( pid > 7 ) tid = 0; |
---|
414 | else tid = _current_task_array[pid]; |
---|
415 | |
---|
416 | if( length != 1) return -1; |
---|
417 | if( pid >= nprocs*nclusters ) return -1; |
---|
418 | if( tid >= ntasks ) return -1; |
---|
419 | |
---|
420 | tty_address = (char*)(base + increment + tid*TTY_SPAN*4); |
---|
421 | |
---|
422 | if((tty_address[TTY_STATUS*4] & 0x1) == 0x1) |
---|
423 | { |
---|
424 | buffer[0] = tty_address[TTY_READ*4]; |
---|
425 | return 1; |
---|
426 | } |
---|
427 | else |
---|
428 | { |
---|
429 | return 0; |
---|
430 | } |
---|
431 | } |
---|
432 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
433 | // _tty_read_irq() |
---|
434 | // iAS it uses the TTY_GET_IRQ interrupt and the associated kernel buffer, |
---|
435 | // that has been written by the ISR, this function does not access the TTY registers. |
---|
436 | // It fetch one single character from the _tty_get_buf[tty_index] kernel buffer, writes |
---|
437 | // this character to the user buffer, and reset the _tty_get_full[tty_index] buffer. |
---|
438 | // This is a non blocking call : it returns 0 if the kernel buffer is empty, |
---|
439 | // and returns 1 if the buffer is full. |
---|
440 | // It returns -1 in case of error (proc_id or task_id too large, or length != 1) |
---|
441 | // The length argument is not used in this implementation, and has been |
---|
442 | // introduced for future implementations. |
---|
443 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
444 | in_drivers int _tty_read_irq(char* buffer, int length) |
---|
445 | { |
---|
446 | int pid = _procid(); |
---|
447 | int ntasks = (int)&NB_TASKS; |
---|
448 | int nprocs = (int)&NB_PROCS; |
---|
449 | int nclusters = (int)&NB_CLUSTERS; |
---|
450 | int tty_index; |
---|
451 | int tid; |
---|
452 | |
---|
453 | if( pid > 7 ) tid = 0; |
---|
454 | else tid = _current_task_array[pid]; |
---|
455 | |
---|
456 | if( length != 1) return -1; |
---|
457 | if( pid >= nprocs*nclusters ) return -1; |
---|
458 | if( tid >= ntasks ) return -1; |
---|
459 | |
---|
460 | tty_index = pid*ntasks + tid; |
---|
461 | if( _tty_get_full[tty_index] == 0 ) return 0; |
---|
462 | |
---|
463 | *buffer = _tty_get_buf[tty_index]; |
---|
464 | _tty_get_full[tty_index] = 0; |
---|
465 | return 1; |
---|
466 | } |
---|
467 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
468 | // _exit() |
---|
469 | // Exit (suicide) after printing message on a TTY terminal. |
---|
470 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
471 | in_drivers int _exit() |
---|
472 | { |
---|
473 | char buf[] = "\n\n!!! Exit Processor !!!\n"; |
---|
474 | int pid = _procid(); |
---|
475 | |
---|
476 | buf[24] = '0'; |
---|
477 | buf[25] = 'x'; |
---|
478 | buf[26] = (char)((pid>>8) & 0xF) + 0x30; |
---|
479 | buf[27] = (char)((pid>>4) & 0xF) + 0x30; |
---|
480 | buf[28] = (char)(pid & 0xF) + 0x30; |
---|
481 | _tty_write(buf, 36); |
---|
482 | |
---|
483 | while(1) asm volatile("nop"); // infinite loop... |
---|
484 | } |
---|
485 | |
---|
486 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
487 | // _icu_write() |
---|
488 | // Write a 32 bits word in a memory mapped register of the ICU peripheral |
---|
489 | // The base address is defined by the processor ID |
---|
490 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
491 | in_drivers int _icu_write(size_t register_index, int value) |
---|
492 | { |
---|
493 | int* icu_address; |
---|
494 | unsigned int base = (int)&seg_icu_base; |
---|
495 | unsigned int increment = _segment_increment(ICU_SPAN*4); |
---|
496 | |
---|
497 | if( register_index >= ICU_SPAN ) return -1; |
---|
498 | |
---|
499 | icu_address = (int*)(base + increment); |
---|
500 | icu_address[register_index] = value; // write word |
---|
501 | return 0; |
---|
502 | } |
---|
503 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
504 | // _icu_read() |
---|
505 | // Read a 32 bits word in a memory mapped register of the ICU peripheral |
---|
506 | // The ICU base address is defined by the processor ID |
---|
507 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
508 | in_drivers int _icu_read(size_t register_index, int* buffer) |
---|
509 | { |
---|
510 | int* icu_address; |
---|
511 | unsigned int base = (int)&seg_icu_base; |
---|
512 | unsigned int increment = _segment_increment(ICU_SPAN*4); |
---|
513 | |
---|
514 | if( register_index >= ICU_SPAN ) return -1; |
---|
515 | |
---|
516 | icu_address = (int*)(base + increment); |
---|
517 | *buffer = icu_address[register_index]; // read word |
---|
518 | return 0; |
---|
519 | } |
---|
520 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
521 | // _gcd_write() |
---|
522 | // Write a 32 bits word in a memory mapped register of the GCD coprocessor |
---|
523 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
524 | in_drivers int _gcd_write(size_t register_index, int value) |
---|
525 | { |
---|
526 | int* gcd_address; |
---|
527 | if( register_index >= 4 ) return -1; |
---|
528 | |
---|
529 | gcd_address = (int*)&seg_gcd_base; |
---|
530 | gcd_address[register_index] = value; // write word |
---|
531 | return 0; |
---|
532 | } |
---|
533 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
534 | // _gcd_read() |
---|
535 | // Read a 32 bits word in a memory mapped register of the GCD coprocessor |
---|
536 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
537 | in_drivers int _gcd_read(size_t register_index, int* buffer) |
---|
538 | { |
---|
539 | int* gcd_address; |
---|
540 | if( register_index >= 4 ) return -1; |
---|
541 | |
---|
542 | gcd_address = (int*)&seg_gcd_base; |
---|
543 | *buffer = gcd_address[register_index]; // read word |
---|
544 | return 0; |
---|
545 | } |
---|
546 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
547 | // _locks_write() |
---|
548 | // Release a software spin-lock |
---|
549 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
550 | in_drivers int _locks_write(size_t index) |
---|
551 | |
---|
552 | { |
---|
553 | int max = (int)&NB_LOCKS; |
---|
554 | if( index >= max ) return -1; |
---|
555 | |
---|
556 | _spin_lock[index] = 0; |
---|
557 | return 0; |
---|
558 | } |
---|
559 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
560 | // _locks_read() |
---|
561 | // Try to take a software spin-lock. |
---|
562 | // This is a blocking call, as there is a busy-waiting loop, |
---|
563 | // until the lock is granted to the requester. |
---|
564 | // There is an internal delay of about 100 cycles between |
---|
565 | // two successive lock read, to avoid bus saturation. |
---|
566 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
567 | in_drivers int _locks_read(size_t index) |
---|
568 | { |
---|
569 | int max = (int)&NB_LOCKS; |
---|
570 | if( index >= max ) return -1; |
---|
571 | |
---|
572 | register int delay = ((_proctime() +_procid()) & 0xF) << 4; |
---|
573 | register int * plock = (int *) &_spin_lock[index]; |
---|
574 | |
---|
575 | asm volatile ("_locks_llsc: \n" |
---|
576 | "ll $2, 0(%0) \n" // $2 <= _locks_lock |
---|
577 | "bnez $2, _locks_delay \n" // random delay if busy |
---|
578 | "li $3, 1 \n" // prepare argument for sc |
---|
579 | "sc $3, 0(%0) \n" // try to set _locks_busy |
---|
580 | "bnez $3, _locks_ok \n" // exit if atomic |
---|
581 | "_locks_delay: \n" |
---|
582 | "move $4, %1 \n" // $4 <= delay |
---|
583 | "_locks_loop: \n" |
---|
584 | "addi $4, $4, -1 \n" // $4 <= $4 - 1 |
---|
585 | "beqz $4, _locks_loop \n" // test end delay |
---|
586 | "j _locks_llsc \n" // retry |
---|
587 | "_locks_ok: \n" |
---|
588 | ::"r"(plock),"r"(delay):"$2","$3","$4"); |
---|
589 | return 0; |
---|
590 | } |
---|
591 | ////////////////////////////////////////////////////////////////////////////////////////// |
---|
592 | // I/O BLOCK_DEVICE |
---|
593 | // The three functions below use the three variables _ioc_lock _ioc_done, |
---|
594 | // and _ioc_status for synchronisation. |
---|
595 | // - As the IOC component can be used by several programs running in parallel, |
---|
596 | // the _ioc_lock variable guaranties exclusive access to the device. |
---|
597 | // The _ioc_read() and _ioc_write() functions use atomic LL/SC to get the lock. |
---|
598 | // and set _ioc_lock to a non zero value. |
---|
599 | // The _ioc_write() and _ioc_read() functions are blocking, polling the _ioc_lock |
---|
600 | // variable until the device is available. |
---|
601 | // - When the tranfer is completed, the ISR routine activated by the IOC IRQ |
---|
602 | // set the _ioc_done variable to a non-zero value. Possible address errors detected |
---|
603 | // by the IOC peripheral are reported by the ISR in the _ioc_status variable. |
---|
604 | // The _ioc_completed() function is polling the _ioc_done variable, waiting for |
---|
605 | // tranfer conpletion. When the completion is signaled, the _ioc_completed() function |
---|
606 | // reset the _ioc_done variable to zero, and releases the _ioc_lock variable. |
---|
607 | // |
---|
608 | // In a multi-tasks environment, this polling policy must be replaced by a |
---|
609 | // descheduling policy for the requesting process. |
---|
610 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
611 | // _ioc_get_lock() |
---|
612 | // This blocking function is used by the _ioc_read() and _ioc_write() functions |
---|
613 | // to get _ioc_lock using LL/SC. |
---|
614 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
615 | in_drivers void _ioc_get_lock() |
---|
616 | { |
---|
617 | register unsigned int* plock = (unsigned int*)&_ioc_lock; |
---|
618 | |
---|
619 | asm volatile ("_ioc_llsc: \n" |
---|
620 | "ll $2, 0(%0) \n" // $2 <= _ioc_lock |
---|
621 | "bnez $2, _ioc_llsc \n" // retry if busy |
---|
622 | "li $3, 1 \n" // prepare argument for sc |
---|
623 | "sc $3, 0(%0) \n" // try to set _ioc_busy |
---|
624 | "beqz $3, _ioc_llsc \n" // retry if not atomic |
---|
625 | ::"r"(plock):"$2","$3"); |
---|
626 | } |
---|
627 | ////////////////////////////////////////////////////////////////////////////////////// |
---|
628 | // _ioc_write() |
---|
629 | // Transfer data from a memory buffer to a file on the block_device. |
---|
630 | // - lba : first block index on the disk |
---|
631 | // - buffer : base address of the memory buffer |
---|
632 | // - count : number of blocks to be transfered |
---|
633 | // The source buffer must be in user address space. |
---|
634 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
635 | in_drivers int _ioc_write(size_t lba, void* buffer, size_t count) |
---|
636 | { |
---|
637 | volatile unsigned int* ioc_address = (unsigned int*)&seg_ioc_base; |
---|
638 | |
---|
639 | // buffer must be in user space |
---|
640 | // size_t block_size = ioc_address[BLOCK_DEVICE_BLOCK_SIZE]; |
---|
641 | // if( ( (size_t)buffer + block_size*count ) >= 0x80000000 ) return -1; |
---|
642 | // if( ( (size_t)buffer ) >= 0x80000000 ) return -1; |
---|
643 | |
---|
644 | // get the lock |
---|
645 | _ioc_get_lock(); |
---|
646 | |
---|
647 | // block_device configuration |
---|
648 | ioc_address[BLOCK_DEVICE_BUFFER] = (int)buffer; |
---|
649 | ioc_address[BLOCK_DEVICE_COUNT] = count; |
---|
650 | ioc_address[BLOCK_DEVICE_LBA] = lba; |
---|
651 | ioc_address[BLOCK_DEVICE_IRQ_ENABLE] = 1; |
---|
652 | ioc_address[BLOCK_DEVICE_OP] = BLOCK_DEVICE_WRITE; |
---|
653 | return 0; |
---|
654 | } |
---|
655 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
656 | // _ioc_read() |
---|
657 | // Transfer data from a file on the block device to a memory buffer. |
---|
658 | // - lba : first block index on the disk |
---|
659 | // - buffer : base address of the memory buffer |
---|
660 | // - count : number of blocks to be transfered |
---|
661 | // The destination buffer must be in user address space. |
---|
662 | // All cache lines corresponding to the the target buffer must be invalidated |
---|
663 | // for cache coherence. |
---|
664 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
665 | in_drivers int _ioc_read(size_t lba, void* buffer, size_t count) |
---|
666 | { |
---|
667 | volatile unsigned int* ioc_address = (unsigned int*)&seg_ioc_base; |
---|
668 | |
---|
669 | // buffer must be in user space |
---|
670 | // size_t block_size = ioc_address[BLOCK_DEVICE_BLOCK_SIZE]; |
---|
671 | // if( ( (size_t)buffer + block_size*count ) >= 0x80000000 ) return -1; |
---|
672 | // if( ( (size_t)buffer ) >= 0x80000000 ) return -1; |
---|
673 | |
---|
674 | // get the lock |
---|
675 | _ioc_get_lock(); |
---|
676 | |
---|
677 | // block_device configuration |
---|
678 | ioc_address[BLOCK_DEVICE_BUFFER] = (int)buffer; |
---|
679 | ioc_address[BLOCK_DEVICE_COUNT] = count; |
---|
680 | ioc_address[BLOCK_DEVICE_LBA] = lba; |
---|
681 | ioc_address[BLOCK_DEVICE_IRQ_ENABLE] = 1; |
---|
682 | ioc_address[BLOCK_DEVICE_OP] = BLOCK_DEVICE_READ; |
---|
683 | |
---|
684 | return 0; |
---|
685 | } |
---|
686 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
687 | // _ioc_completed() |
---|
688 | // This blocking function cheks completion of an I/O transfer and reports errors. |
---|
689 | // It returns 0 if the transfer is successfully completed. |
---|
690 | // It returns -1 if an error has been reported. |
---|
691 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
692 | in_drivers int _ioc_completed() |
---|
693 | { |
---|
694 | // waiting for completion |
---|
695 | while (_ioc_done == 0) { asm volatile("nop"); } |
---|
696 | |
---|
697 | // reset synchronisation variables |
---|
698 | _ioc_done = 0; |
---|
699 | _ioc_lock = 0; |
---|
700 | |
---|
701 | // return errors |
---|
702 | if((_ioc_status != BLOCK_DEVICE_READ_SUCCESS) && |
---|
703 | (_ioc_status != BLOCK_DEVICE_WRITE_SUCCESS)) return -1; |
---|
704 | else return 0; |
---|
705 | } |
---|
706 | |
---|
707 | ////////////////////////////////////////////////////////////////////////////////////// |
---|
708 | // FRAME_BUFFER |
---|
709 | // The _fb_sync_write & _fb_sync_read functions use a memcpy strategy to implement |
---|
710 | // the transfer between a data buffer (user space) and the frame buffer (kernel space). |
---|
711 | // They are blocking until completion of the transfer. |
---|
712 | ////////////////////////////////////////////////////////////////////////////////////// |
---|
713 | // _fb_sync_write() |
---|
714 | // Transfer data from an user buffer to the frame_buffer device with a memcpy. |
---|
715 | // - offset : offset (in bytes) in the frame buffer |
---|
716 | // - buffer : base address of the memory buffer |
---|
717 | // - length : number of bytes to be transfered |
---|
718 | ////////////////////////////////////////////////////////////////////////////////////// |
---|
719 | in_drivers int _fb_sync_write(size_t offset, void* buffer, size_t length) |
---|
720 | { |
---|
721 | volatile char* fb = (char*)(void*)&seg_fb_base + offset; |
---|
722 | char* ub = buffer; |
---|
723 | size_t i; |
---|
724 | |
---|
725 | // buffer must be in user space |
---|
726 | // if( ( (size_t)buffer + length ) >= 0x80000000 ) return -1; |
---|
727 | // if( ( (size_t)buffer ) >= 0x80000000 ) return -1; |
---|
728 | |
---|
729 | // memory copy |
---|
730 | for(i=0 ; i<length ; i++) fb[i] = ub[i]; |
---|
731 | return 0; |
---|
732 | } |
---|
733 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
734 | // _fb_sync_read() |
---|
735 | // Transfer data from the frame_buffer device to an user buffer with a memcpy. |
---|
736 | // - offset : offset (in bytes) in the frame buffer |
---|
737 | // - buffer : base address of the memory buffer |
---|
738 | // - length : number of bytes to be transfered |
---|
739 | ////////////////////////////////////////////////////////////////////////////////////// |
---|
740 | in_drivers int _fb_sync_read(size_t offset, void* buffer, size_t length) |
---|
741 | { |
---|
742 | volatile char* fb = (char*)(void*)&seg_fb_base + offset; |
---|
743 | char* ub = buffer; |
---|
744 | size_t i; |
---|
745 | |
---|
746 | // buffer must be in user space |
---|
747 | // if( ( (size_t)buffer + length ) >= 0x80000000 ) return -1; |
---|
748 | // if( ( (size_t)buffer ) >= 0x80000000 ) return -1; |
---|
749 | |
---|
750 | // memory copy |
---|
751 | for(i=0 ; i<length ; i++) ub[i] = fb[i]; |
---|
752 | return 0; |
---|
753 | } |
---|
754 | ////////////////////////////////////////////////////////////////////////////////////// |
---|
755 | // The _fb_write() and _fb_read() functions use the MULTI_DMA |
---|
756 | // coprocessor to transfer data between the user buffer and the frame buffer. |
---|
757 | // The _fb_completed() function, use a polling policy to test |
---|
758 | // the global variables _dma_busy[i] and detect the transfer completion. |
---|
759 | // As each processor can have it's private DMA, there is up to 256 _dma_busy[i] |
---|
760 | // set/reset variables that are indexed by the proc_id. |
---|
761 | // The _dma_busy variable is reset by the ISR associated to the DMA IRQ. |
---|
762 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
763 | // _fb_write() |
---|
764 | // Transfer data from an user buffer to the frame_buffer device using DMA. |
---|
765 | // - offset : offset (in bytes) in the frame buffer |
---|
766 | // - buffer : base address of the memory buffer |
---|
767 | // - length : number of bytes to be transfered |
---|
768 | ////////////////////////////////////////////////////////////////////////////////////// |
---|
769 | in_drivers int _fb_write(size_t offset, void* buffer, size_t length) |
---|
770 | { |
---|
771 | int* dma_address; |
---|
772 | unsigned int base = (unsigned int) &seg_dma_base; |
---|
773 | unsigned int increment = _segment_increment(DMA_SPAN*4); |
---|
774 | char * fb = (char *) &seg_fb_base + offset; |
---|
775 | unsigned int delay = (_proctime() & 0xF) << 4; |
---|
776 | unsigned int pid = _procid(); |
---|
777 | unsigned int i; |
---|
778 | |
---|
779 | |
---|
780 | // checking buffer boundaries (bytes) |
---|
781 | // if( ( (size_t)buffer + length ) >= 0x80000000 ) return -1; |
---|
782 | // if( ( (size_t)buffer ) >= 0x80000000 ) return -1; |
---|
783 | |
---|
784 | // waiting until DMA device is available |
---|
785 | while (_dma_busy[pid] != 0) |
---|
786 | { |
---|
787 | for( i=0 ; i<delay ; i++) // busy waiting |
---|
788 | { // with a pseudo random |
---|
789 | asm volatile("nop"); // delay between bus accesses |
---|
790 | } |
---|
791 | } |
---|
792 | _dma_busy[pid] = 1; |
---|
793 | |
---|
794 | dma_address = (int*)(base + increment); |
---|
795 | |
---|
796 | // DMA configuration |
---|
797 | dma_address[DMA_IRQ_DISABLE] = 0; |
---|
798 | dma_address[DMA_SRC] = (int)buffer; |
---|
799 | dma_address[DMA_DST] = (int)fb; |
---|
800 | dma_address[DMA_LEN] = (int)length; |
---|
801 | return 0; |
---|
802 | } |
---|
803 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
804 | // _fb_read() |
---|
805 | // Transfer data from the frame_buffer device to an user buffer using DMA. |
---|
806 | // - offset : offset (in bytes) in the frame buffer |
---|
807 | // - buffer : base address of the memory buffer |
---|
808 | // - length : number of bytes to be transfered |
---|
809 | ////////////////////////////////////////////////////////////////////////////////////// |
---|
810 | in_drivers int _fb_read(size_t offset, void* buffer, size_t length) |
---|
811 | { |
---|
812 | int* dma_address; |
---|
813 | unsigned int base = (unsigned int)&seg_dma_base; |
---|
814 | unsigned int increment = _segment_increment(DMA_SPAN*4); |
---|
815 | char* fb = (char*)&seg_fb_base + offset; |
---|
816 | unsigned int delay = (_proctime() & 0xF) << 4; |
---|
817 | unsigned int pid = _procid(); |
---|
818 | unsigned int i; |
---|
819 | |
---|
820 | // checking buffer boundaries (bytes) |
---|
821 | // if( ( (size_t)buffer + length ) >= 0x80000000 ) return -1; |
---|
822 | // if( ( (size_t)buffer ) >= 0x80000000 ) return -1; |
---|
823 | |
---|
824 | // waiting until DMA device is available |
---|
825 | while (_dma_busy[pid] != 0) |
---|
826 | { |
---|
827 | for( i=0 ; i<delay ; i++) // busy waiting |
---|
828 | { // with a pseudo random |
---|
829 | asm volatile("nop"); // delay between bus accesses |
---|
830 | } |
---|
831 | } |
---|
832 | _dma_busy[pid] = 1; |
---|
833 | |
---|
834 | dma_address = (int*)(base + increment); |
---|
835 | |
---|
836 | // DMA configuration |
---|
837 | dma_address[DMA_IRQ_DISABLE] = 0; |
---|
838 | dma_address[DMA_SRC] = (int)fb; |
---|
839 | dma_address[DMA_DST] = (int)buffer; |
---|
840 | dma_address[DMA_LEN] = (int)length; |
---|
841 | return 0; |
---|
842 | } |
---|
843 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
844 | // _fb_completed() |
---|
845 | // This blocking function cheks completion of a DMA transfer to or fom the frame buffer. |
---|
846 | // The MIPS32 wait instruction stall the processor until the next interrupt. |
---|
847 | // It returns 0 if the transfer is successfully completed |
---|
848 | // It returns -1 if an error has been reported. |
---|
849 | /////////////////////////////////////////////////////////////////////////////////////// |
---|
850 | in_drivers int _fb_completed() |
---|
851 | { |
---|
852 | unsigned int pid = _procid(); |
---|
853 | |
---|
854 | while (_dma_busy[pid] != 0) |
---|
855 | { |
---|
856 | asm volatile("nop"); |
---|
857 | } |
---|
858 | if(_dma_status[pid] == DMA_SUCCESS) return 0; |
---|
859 | else return _dma_status[pid]; |
---|
860 | } |
---|
861 | ////////////////////////////////////////////////////////////////////////////////////// |
---|
862 | // _barrier_init() |
---|
863 | // This function makes a cooperative initialisation of the barrier: |
---|
864 | // - barrier_count[index] <= N |
---|
865 | // - barrier_lock[index] <= 0 |
---|
866 | // All tasks try to initialize the barrier, but the initialisation |
---|
867 | // is done by only one task, using LL/SC instructions. |
---|
868 | // This cooperative initialisation is questionnable, |
---|
869 | // bcause the barrier can ony be initialised once... |
---|
870 | ////////////////////////////////////////////////////////////////////////////////////// |
---|
871 | in_drivers int _barrier_init(unsigned int index, unsigned int value) |
---|
872 | { |
---|
873 | |
---|
874 | register int* pinit = (int*)&_barrier_initial_value[index]; |
---|
875 | register int* pcount = (int*)&_barrier_count[index]; |
---|
876 | register int* plock = (int*)&_barrier_lock[index]; |
---|
877 | |
---|
878 | if ( index > 7 ) return 1; |
---|
879 | |
---|
880 | // parallel initialisation using atomic instructions LL/SC |
---|
881 | asm volatile ("_barrier_init_test: \n" |
---|
882 | "ll $2, 0(%0) \n" // read barrier_inital_value |
---|
883 | "bnez $2, _barrier_init_done \n" |
---|
884 | "move $3, %3 \n" |
---|
885 | "sc $3, 0(%0) \n" // try to write barrier_initial_value |
---|
886 | "beqz $3, _barrier_init_test \n" |
---|
887 | "move $3, %3 \n" |
---|
888 | "sw $3, 0(%1) \n" // barrier_count <= barrier_initial_value |
---|
889 | "move $3, $0 \n" // |
---|
890 | "sw $3, 0(%2) \n" // barrier_lock <= 0 |
---|
891 | "_barrier_init_done: \n" |
---|
892 | ::"r"(pinit),"r"(pcount),"r"(plock),"r"(value):"$2","$3"); |
---|
893 | return 0 ; |
---|
894 | } |
---|
895 | ////////////////////////////////////////////////////////////////////////////////////// |
---|
896 | // _barrier_wait() |
---|
897 | // This blocking function uses a busy_wait technics (on the barrier_lock value), |
---|
898 | // because the GIET does not support dynamic scheduling/descheduling of tasks. |
---|
899 | // The barrier state is actually defined by two variables: |
---|
900 | // _barrier_count[index] define the number of particpants that are waiting |
---|
901 | // _barrier_lock[index] define the bool variable whose value is polled |
---|
902 | // The last participant change the value of _barrier_lock[index] to release the barrier... |
---|
903 | // There is at most 16 independant barriers, and an error is returned |
---|
904 | // if the barrier index is larger than 15. |
---|
905 | ////////////////////////////////////////////////////////////////////////////////////// |
---|
906 | in_drivers int _barrier_wait(unsigned int index) |
---|
907 | { |
---|
908 | register int* pcount = (int*)&_barrier_count[index]; |
---|
909 | register int count; |
---|
910 | |
---|
911 | int lock = _barrier_lock[index]; |
---|
912 | |
---|
913 | if ( index > 15 ) return 1; |
---|
914 | |
---|
915 | // parallel decrement _barrier_count[index] using atomic instructions LL/SC |
---|
916 | // input : pointer on _barrier_count[index] |
---|
917 | // output : count = _barrier_count[index] (before decrementation) |
---|
918 | asm volatile ("_barrier_decrement: \n" |
---|
919 | "ll %0, 0(%1) \n" |
---|
920 | "addi $3, %0, -1 \n" |
---|
921 | "sc $3, 0(%1) \n" |
---|
922 | "beqz $3, _barrier_decrement \n" |
---|
923 | :"=&r"(count) |
---|
924 | :"r"(pcount) |
---|
925 | :"$2","$3"); |
---|
926 | |
---|
927 | // the last task re-initializes the barrier_ count variable |
---|
928 | // and the barrier_lock variable, waking up all other waiting tasks |
---|
929 | |
---|
930 | if ( count == 1 ) // last task |
---|
931 | { |
---|
932 | _barrier_count[index] = _barrier_initial_value[index]; |
---|
933 | asm volatile( "sync" ); |
---|
934 | _barrier_lock[index] = (lock == 0) ? 1 : 0; |
---|
935 | return 0 ; |
---|
936 | } |
---|
937 | else // other tasks |
---|
938 | { |
---|
939 | while ( lock == _barrier_lock[index] ) { } // busy waiting |
---|
940 | return 0 ; |
---|
941 | } |
---|
942 | } |
---|
943 | ////////////////////////////////////////////////////////////////////////////////////// |
---|
944 | |
---|
945 | |
---|
946 | // Local Variables: |
---|
947 | // tab-width: 4; |
---|
948 | // c-basic-offset: 4; |
---|
949 | // c-file-offsets:((innamespace . 0)(inline-open . 0)); |
---|
950 | // indent-tabs-mode: nil; |
---|
951 | // End: |
---|
952 | // |
---|
953 | // vim: filetype=cpp:expandtab:shiftwidth=4:tabstop=4:softtabstop=4 |
---|
954 | |
---|